* [PATCH 1/8] csiostor: Chelsio FCoE offload driver submission (sources part 1).
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
@ 2012-08-23 22:27 ` Naresh Kumar Inna
2012-08-23 22:27 ` [PATCH 2/8] csiostor: Chelsio FCoE offload driver submission (sources part 2) Naresh Kumar Inna
` (8 subsequent siblings)
9 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
This patch contains the hardware interfacing functionality (chip/firmware
initialization/setup) and error handling. It also has slow path event
handling functionality, the Makefile and Kconfig changes.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/Kconfig | 1 +
drivers/scsi/Makefile | 1 +
drivers/scsi/csiostor/Kconfig | 20 +
drivers/scsi/csiostor/Makefile | 11 +
drivers/scsi/csiostor/csio_hw.c | 4385 +++++++++++++++++++++++++++++++++++++++
5 files changed, 4418 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/Kconfig
create mode 100644 drivers/scsi/csiostor/Makefile
create mode 100644 drivers/scsi/csiostor/csio_hw.c
diff --git a/drivers/scsi/Kconfig b/drivers/scsi/Kconfig
index 74bf1aa..af7a3e7 100644
--- a/drivers/scsi/Kconfig
+++ b/drivers/scsi/Kconfig
@@ -1812,6 +1812,7 @@ config SCSI_VIRTIO
This is the virtual HBA driver for virtio. If the kernel will
be used in a virtual machine, say Y or M.
+source "drivers/scsi/csiostor/Kconfig"
endif # SCSI_LOWLEVEL
diff --git a/drivers/scsi/Makefile b/drivers/scsi/Makefile
index 888f73a..8739aa7 100644
--- a/drivers/scsi/Makefile
+++ b/drivers/scsi/Makefile
@@ -90,6 +90,7 @@ obj-$(CONFIG_SCSI_QLA_FC) += qla2xxx/
obj-$(CONFIG_SCSI_QLA_ISCSI) += libiscsi.o qla4xxx/
obj-$(CONFIG_SCSI_LPFC) += lpfc/
obj-$(CONFIG_SCSI_BFA_FC) += bfa/
+obj-$(CONFIG_SCSI_CHELSIO_FCOE) += csiostor/
obj-$(CONFIG_SCSI_PAS16) += pas16.o
obj-$(CONFIG_SCSI_T128) += t128.o
obj-$(CONFIG_SCSI_DMX3191D) += dmx3191d.o
diff --git a/drivers/scsi/csiostor/Kconfig b/drivers/scsi/csiostor/Kconfig
new file mode 100644
index 0000000..c2acf02
--- /dev/null
+++ b/drivers/scsi/csiostor/Kconfig
@@ -0,0 +1,20 @@
+config SCSI_CHELSIO_FCOE
+ tristate "Chelsio Communications FCoE support"
+ depends on PCI && SCSI
+ select SCSI_FC_ATTRS
+ select FW_LOADER
+ help
+ This driver supports FCoE Offload functionality over
+ Chelsio T4-based 10Gb Converged Network Adapters.
+
+ For general information about Chelsio and our products, visit
+ our website at <http://www.chelsio.com>.
+
+ For customer support, please visit our customer support page at
+ <http://www.chelsio.com/support.html>.
+
+ Please send feedback to <linux-bugs@chelsio.com>.
+
+ To compile this driver as a module choose M here; the module
+ will be called csiostor.
+
diff --git a/drivers/scsi/csiostor/Makefile b/drivers/scsi/csiostor/Makefile
new file mode 100644
index 0000000..b581966
--- /dev/null
+++ b/drivers/scsi/csiostor/Makefile
@@ -0,0 +1,11 @@
+#
+## Chelsio FCoE driver
+#
+##
+
+ccflags-y += -I$(srctree)/drivers/net/ethernet/chelsio/cxgb4
+
+obj-$(CONFIG_SCSI_CHELSIO_FCOE) += csiostor.o
+
+csiostor-objs := csio_attr.o csio_init.o csio_lnode.o csio_scsi.o \
+ csio_hw.o csio_isr.o csio_mb.o csio_rnode.o csio_wr.o
diff --git a/drivers/scsi/csiostor/csio_hw.c b/drivers/scsi/csiostor/csio_hw.c
new file mode 100644
index 0000000..a220611
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_hw.c
@@ -0,0 +1,4385 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/pci.h>
+#include <linux/pci_regs.h>
+#include <linux/firmware.h>
+#include <linux/stddef.h>
+#include <linux/delay.h>
+#include <linux/string.h>
+#include <linux/compiler.h>
+#include <linux/jiffies.h>
+#include <linux/log2.h>
+
+#include "csio_hw.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+
+int csio_force_master;
+int csio_dbg_level = 0xFEFF;
+unsigned int csio_port_mask = 0xf;
+
+/* Default FW event queue entries. */
+static uint32_t csio_evtq_sz = CSIO_EVTQ_SIZE;
+
+/* Default MSI param level */
+int csio_msi = 2;
+
+/* FCoE function instances */
+static int dev_num;
+
+/* FCoE Adapter types & its description */
+static struct csio_adap_desc csio_fcoe_adapters[] = {
+ {"T440-Dbg 10G", "Chelsio T440-Dbg 10G [FCoE]"},
+ {"T420-CR 10G", "Chelsio T420-CR 10G [FCoE]"},
+ {"T422-CR 10G/1G", "Chelsio T422-CR 10G/1G [FCoE]"},
+ {"T440-CR 10G", "Chelsio T440-CR 10G [FCoE]"},
+ {"T420-BCH 10G", "Chelsio T420-BCH 10G [FCoE]"},
+ {"T440-BCH 10G", "Chelsio T440-BCH 10G [FCoE]"},
+ {"T440-CH 10G", "Chelsio T440-CH 10G [FCoE]"},
+ {"T420-SO 10G", "Chelsio T420-SO 10G [FCoE]"},
+ {"T420-CX4 10G", "Chelsio T420-CX4 10G [FCoE]"},
+ {"T420-BT 10G", "Chelsio T420-BT 10G [FCoE]"},
+ {"T404-BT 1G", "Chelsio T404-BT 1G [FCoE]"},
+ {"B420-SR 10G", "Chelsio B420-SR 10G [FCoE]"},
+ {"B404-BT 1G", "Chelsio B404-BT 1G [FCoE]"},
+ {"T480-CR 10G", "Chelsio T480-CR 10G [FCoE]"},
+ {"T440-LP-CR 10G", "Chelsio T440-LP-CR 10G [FCoE]"},
+ {"T4 FPGA", "Chelsio T4 FPGA [FCoE]"}
+};
+
+static void csio_mgmtm_cleanup(struct csio_mgmtm *);
+static void csio_hw_mbm_cleanup(struct csio_hw *);
+
+/* State machine forward declarations */
+static void csio_hws_uninit(struct csio_hw *, enum csio_hw_ev);
+static void csio_hws_configuring(struct csio_hw *, enum csio_hw_ev);
+static void csio_hws_initializing(struct csio_hw *, enum csio_hw_ev);
+static void csio_hws_ready(struct csio_hw *, enum csio_hw_ev);
+static void csio_hws_quiescing(struct csio_hw *, enum csio_hw_ev);
+static void csio_hws_quiesced(struct csio_hw *, enum csio_hw_ev);
+static void csio_hws_resetting(struct csio_hw *, enum csio_hw_ev);
+static void csio_hws_removing(struct csio_hw *, enum csio_hw_ev);
+static void csio_hws_pcierr(struct csio_hw *, enum csio_hw_ev);
+
+static void csio_hw_initialize(struct csio_hw *hw);
+static void csio_evtq_stop(struct csio_hw *hw);
+static void csio_evtq_start(struct csio_hw *hw);
+
+int csio_is_hw_ready(struct csio_hw *hw)
+{
+ return csio_match_state(hw, csio_hws_ready);
+}
+
+int csio_is_hw_removing(struct csio_hw *hw)
+{
+ return csio_match_state(hw, csio_hws_removing);
+}
+
+
+/*
+ * csio_hw_wait_op_done_val - wait until an operation is completed
+ * @hw: the HW module
+ * @reg: the register to check for completion
+ * @mask: a single-bit field within @reg that indicates completion
+ * @polarity: the value of the field when the operation is completed
+ * @attempts: number of check iterations
+ * @delay: delay in usecs between iterations
+ * @valp: where to store the value of the register at completion time
+ *
+ * Wait until an operation is completed by checking a bit in a register
+ * up to @attempts times. If @valp is not NULL the value of the register
+ * at the time it indicated completion is stored there. Returns 0 if the
+ * operation completes and -EAGAIN otherwise.
+ */
+static csio_retval_t
+csio_hw_wait_op_done_val(struct csio_hw *hw, int reg, uint32_t mask,
+ int polarity, int attempts, int delay, uint32_t *valp)
+{
+ uint32_t val;
+ while (1) {
+ val = csio_rd_reg32(hw, reg);
+
+ if (!!(val & mask) == polarity) {
+ if (valp)
+ *valp = val;
+ return CSIO_SUCCESS;
+ }
+
+ if (--attempts == 0)
+ return CSIO_RETRY;
+ if (delay)
+ udelay(delay);
+ }
+}
+
+void
+csio_set_reg_field(struct csio_hw *hw, uint32_t reg, uint32_t mask,
+ uint32_t value)
+{
+ uint32_t val = csio_rd_reg32(hw, reg) & ~mask;
+
+ csio_wr_reg32(hw, val | value, reg);
+ /* Flush */
+ csio_rd_reg32(hw, reg);
+
+}
+
+/*
+ * csio_hw_mc_read - read from MC through backdoor accesses
+ * @hw: the hw module
+ * @addr: address of first byte requested
+ * @data: 64 bytes of data containing the requested address
+ * @ecc: where to store the corresponding 64-bit ECC word
+ *
+ * Read 64 bytes of data from MC starting at a 64-byte-aligned address
+ * that covers the requested address @addr. If @parity is not %NULL it
+ * is assigned the 64-bit ECC word for the read data.
+ */
+csio_retval_t
+csio_hw_mc_read(struct csio_hw *hw, uint32_t addr, uint32_t *data,
+ uint64_t *ecc)
+{
+ int i;
+
+ if (csio_rd_reg32(hw, MC_BIST_CMD) & START_BIST)
+ return CSIO_BUSY;
+ csio_wr_reg32(hw, addr & ~0x3fU, MC_BIST_CMD_ADDR);
+ csio_wr_reg32(hw, 64, MC_BIST_CMD_LEN);
+ csio_wr_reg32(hw, 0xc, MC_BIST_DATA_PATTERN);
+ csio_wr_reg32(hw, BIST_OPCODE(1) | START_BIST | BIST_CMD_GAP(1),
+ MC_BIST_CMD);
+ i = csio_hw_wait_op_done_val(hw, MC_BIST_CMD, START_BIST,
+ 0, 10, 1, NULL);
+ if (i)
+ return i;
+
+#define MC_DATA(i) MC_BIST_STATUS_REG(MC_BIST_STATUS_RDATA, i)
+
+ for (i = 15; i >= 0; i--)
+ *data++ = htonl(csio_rd_reg32(hw, MC_DATA(i)));
+ if (ecc)
+ *ecc = csio_rd_reg64(hw, MC_DATA(16));
+#undef MC_DATA
+ return 0;
+}
+
+/*
+ * csio_hw_edc_read - read from EDC through backdoor accesses
+ * @hw: the hw module
+ * @idx: which EDC to access
+ * @addr: address of first byte requested
+ * @data: 64 bytes of data containing the requested address
+ * @ecc: where to store the corresponding 64-bit ECC word
+ *
+ * Read 64 bytes of data from EDC starting at a 64-byte-aligned address
+ * that covers the requested address @addr. If @parity is not %NULL it
+ * is assigned the 64-bit ECC word for the read data.
+ */
+csio_retval_t
+csio_hw_edc_read(struct csio_hw *hw, int idx, uint32_t addr, uint32_t *data,
+ uint64_t *ecc)
+{
+ int i;
+
+ idx *= EDC_STRIDE;
+ if (csio_rd_reg32(hw, EDC_BIST_CMD + idx) & START_BIST)
+ return CSIO_BUSY;
+ csio_wr_reg32(hw, addr & ~0x3fU, EDC_BIST_CMD_ADDR + idx);
+ csio_wr_reg32(hw, 64, EDC_BIST_CMD_LEN + idx);
+ csio_wr_reg32(hw, 0xc, EDC_BIST_DATA_PATTERN + idx);
+ csio_wr_reg32(hw, BIST_OPCODE(1) | BIST_CMD_GAP(1) | START_BIST,
+ EDC_BIST_CMD + idx);
+ i = csio_hw_wait_op_done_val(hw, EDC_BIST_CMD + idx, START_BIST,
+ 0, 10, 1, NULL);
+ if (i)
+ return i;
+
+#define EDC_DATA(i) (EDC_BIST_STATUS_REG(EDC_BIST_STATUS_RDATA, i) + idx)
+
+ for (i = 15; i >= 0; i--)
+ *data++ = htonl(csio_rd_reg32(hw, EDC_DATA(i)));
+ if (ecc)
+ *ecc = csio_rd_reg64(hw, EDC_DATA(16));
+#undef EDC_DATA
+ return 0;
+}
+
+/*
+ * csio_mem_win_rw - read/write memory through PCIE memory window
+ * @hw: the adapter
+ * @addr: address of first byte requested
+ * @data: MEMWIN0_APERTURE bytes of data containing the requested address
+ * @dir: direction of transfer 1 => read, 0 => write
+ *
+ * Read/write MEMWIN0_APERTURE bytes of data from MC starting at a
+ * MEMWIN0_APERTURE-byte-aligned address that covers the requested
+ * address @addr.
+ */
+static int
+csio_mem_win_rw(struct csio_hw *hw, u32 addr, __be32 *data, int dir)
+{
+ int i;
+
+ /*
+ * Setup offset into PCIE memory window. Address must be a
+ * MEMWIN0_APERTURE-byte-aligned address. (Read back MA register to
+ * ensure that changes propagate before we attempt to use the new
+ * values.)
+ */
+ csio_wr_reg32(hw, addr & ~(MEMWIN0_APERTURE - 1),
+ PCIE_MEM_ACCESS_OFFSET);
+ csio_rd_reg32(hw, PCIE_MEM_ACCESS_OFFSET);
+
+ /* Collecting data 4 bytes at a time upto MEMWIN0_APERTURE */
+ for (i = 0; i < MEMWIN0_APERTURE; i = i + sizeof(__be32)) {
+ if (dir)
+ *data++ = csio_rd_reg32(hw, (MEMWIN0_BASE + i));
+ else
+ csio_wr_reg32(hw, *data++, (MEMWIN0_BASE + i));
+ }
+
+ return 0;
+}
+
+/*
+ * csio_memory_rw - read/write EDC 0, EDC 1 or MC via PCIE memory window
+ * @hw: the csio_hw
+ * @mtype: memory type: MEM_EDC0, MEM_EDC1 or MEM_MC
+ * @addr: address within indicated memory type
+ * @len: amount of memory to transfer
+ * @buf: host memory buffer
+ * @dir: direction of transfer 1 => read, 0 => write
+ *
+ * Reads/writes an [almost] arbitrary memory region in the firmware: the
+ * firmware memory address, length and host buffer must be aligned on
+ * 32-bit boudaries. The memory is transferred as a raw byte sequence
+ * from/to the firmware's memory. If this memory contains data
+ * structures which contain multi-byte integers, it's the callers
+ * responsibility to perform appropriate byte order conversions.
+ */
+static csio_retval_t
+csio_memory_rw(struct csio_hw *hw, int mtype, u32 addr, u32 len,
+ uint32_t *buf, int dir)
+{
+ uint32_t pos, start, end, offset, memoffset;
+ int ret;
+ __be32 *data;
+
+ /*
+ * Argument sanity checks ...
+ */
+ if ((addr & 0x3) || (len & 0x3))
+ return CSIO_INVAL;
+
+ data = kmalloc(MEMWIN0_APERTURE, GFP_KERNEL);
+ if (!data)
+ return CSIO_NOMEM;
+
+ /* Offset into the region of memory which is being accessed
+ * MEM_EDC0 = 0
+ * MEM_EDC1 = 1
+ * MEM_MC = 2
+ */
+ memoffset = (mtype * (5 * 1024 * 1024));
+
+ /* Determine the PCIE_MEM_ACCESS_OFFSET */
+ addr = addr + memoffset;
+
+ /*
+ * The underlaying EDC/MC read routines read MEMWIN0_APERTURE bytes
+ * at a time so we need to round down the start and round up the end.
+ * We'll start copying out of the first line at (addr - start) a word
+ * at a time.
+ */
+ start = addr & ~(MEMWIN0_APERTURE-1);
+ end = (addr + len + MEMWIN0_APERTURE-1) & ~(MEMWIN0_APERTURE-1);
+ offset = (addr - start)/sizeof(__be32);
+
+ for (pos = start; pos < end; pos += MEMWIN0_APERTURE, offset = 0) {
+ /*
+ * If we're writing, copy the data from the caller's memory
+ * buffer
+ */
+ if (!dir) {
+ /*
+ * If we're doing a partial write, then we need to do
+ * a read-modify-write ...
+ */
+ if (offset || len < MEMWIN0_APERTURE) {
+ ret = csio_mem_win_rw(hw, pos, data, 1);
+ if (ret) {
+ kfree(data);
+ return ret;
+ }
+ }
+ while (offset < (MEMWIN0_APERTURE/sizeof(__be32)) &&
+ len > 0) {
+ data[offset++] = *buf++;
+ len -= sizeof(__be32);
+ }
+ }
+
+ /*
+ * Transfer a block of memory and bail if there's an error.
+ */
+ ret = csio_mem_win_rw(hw, pos, data, dir);
+ if (ret) {
+ kfree(data);
+ return ret;
+ }
+
+ /*
+ * If we're reading, copy the data into the caller's memory
+ * buffer.
+ */
+ if (dir)
+ while (offset < (MEMWIN0_APERTURE/sizeof(__be32)) &&
+ len > 0) {
+ *buf++ = data[offset++];
+ len -= sizeof(__be32);
+ }
+ }
+
+ kfree(data);
+
+ return 0;
+}
+
+static csio_retval_t
+csio_memory_write(struct csio_hw *hw, int mtype, u32 addr, u32 len, __be32 *buf)
+{
+ return csio_memory_rw(hw, mtype, addr, len, buf, 0);
+}
+
+/*
+ * EEPROM reads take a few tens of us while writes can take a bit over 5 ms.
+ */
+#define EEPROM_MAX_RD_POLL 40
+#define EEPROM_MAX_WR_POLL 6
+#define EEPROM_STAT_ADDR 0x7bfc
+#define VPD_BASE 0x400
+#define VPD_BASE_OLD 0
+#define VPD_LEN 512
+#define VPD_INFO_FLD_HDR_SIZE 3
+
+/*
+ * csio_hw_seeprom_read - read a serial EEPROM location
+ * @hw: hw to read
+ * @addr: EEPROM virtual address
+ * @data: where to store the read data
+ *
+ * Read a 32-bit word from a location in serial EEPROM using the card's PCI
+ * VPD capability. Note that this function must be called with a virtual
+ * address.
+ */
+static int
+csio_hw_seeprom_read(struct csio_hw *hw, uint32_t addr, uint32_t *data)
+{
+ uint16_t val = 0;
+ int attempts = EEPROM_MAX_RD_POLL;
+ uint32_t base = hw->params.pci.vpd_cap_addr;
+
+ if (addr >= EEPROMVSIZE || (addr & 3))
+ return CSIO_INVAL;
+
+ pci_write_config_word(hw->pdev, base + PCI_VPD_ADDR, (uint16_t)addr);
+
+ do {
+ udelay(10);
+ pci_read_config_word(hw->pdev, base + PCI_VPD_ADDR, &val);
+ } while (!(val & PCI_VPD_ADDR_F) && --attempts);
+
+ if (!(val & PCI_VPD_ADDR_F)) {
+ csio_err(hw, "reading EEPROM address 0x%x failed\n", addr);
+ return CSIO_INVAL;
+ }
+
+ pci_read_config_dword(hw->pdev, base + PCI_VPD_DATA, data);
+ *data = le32_to_cpu(*data);
+ return 0;
+}
+
+/*
+ * Partial EEPROM Vital Product Data structure. Includes only the ID and
+ * VPD-R sections.
+ */
+struct t4_vpd_hdr {
+ u8 id_tag;
+ u8 id_len[2];
+ u8 id_data[ID_LEN];
+ u8 vpdr_tag;
+ u8 vpdr_len[2];
+};
+
+/*
+ * csio_hw_get_vpd_keyword_val - Locates an information field keyword in
+ * the VPD
+ * @v: Pointer to buffered vpd data structure
+ * @kw: The keyword to search for
+ *
+ * Returns the value of the information field keyword or
+ * CSIO_INVAL otherwise.
+ */
+static int
+csio_hw_get_vpd_keyword_val(const struct t4_vpd_hdr *v, const char *kw)
+{
+ int32_t i;
+ int32_t offset , len;
+ const uint8_t *buf = &v->id_tag;
+ const uint8_t *vpdr_len = &v->vpdr_tag;
+ offset = sizeof(struct t4_vpd_hdr);
+ len = (uint16_t)vpdr_len[1] + ((uint16_t)vpdr_len[2] << 8);
+
+ if (len + sizeof(struct t4_vpd_hdr) > VPD_LEN)
+ return CSIO_INVAL;
+
+ for (i = offset; (i + VPD_INFO_FLD_HDR_SIZE) <= (offset + len);) {
+ if (memcmp(buf + i , kw, 2) == 0) {
+ i += VPD_INFO_FLD_HDR_SIZE;
+ return i;
+ }
+
+ i += VPD_INFO_FLD_HDR_SIZE + buf[i+2];
+ }
+
+ return CSIO_INVAL;
+}
+
+static int
+csio_pci_capability(struct pci_dev *pdev, int cap, int *pos)
+{
+ *pos = pci_find_capability(pdev, cap);
+ if (*pos)
+ return 0;
+
+ return -1;
+}
+
+/*
+ * csio_hw_get_vpd_params - read VPD parameters from VPD EEPROM
+ * @hw: HW module
+ * @p: where to store the parameters
+ *
+ * Reads card parameters stored in VPD EEPROM.
+ */
+static csio_retval_t
+csio_hw_get_vpd_params(struct csio_hw *hw, struct csio_vpd *p)
+{
+ int i, ret, ec, sn, addr;
+ uint8_t vpd[VPD_LEN], csum;
+ const struct t4_vpd_hdr *v;
+ /* To get around compilation warning from strstrip */
+ char *s;
+
+ if (csio_is_valid_vpd(hw))
+ return CSIO_SUCCESS;
+
+ ret = csio_pci_capability(hw->pdev, PCI_CAP_ID_VPD,
+ &hw->params.pci.vpd_cap_addr);
+ if (ret)
+ return ret;
+
+ /*
+ * Card information normally starts at VPD_BASE but early cards had
+ * it at 0.
+ */
+ ret = csio_hw_seeprom_read(hw, VPD_BASE, (uint32_t *)(vpd));
+ addr = *vpd == 0x82 ? VPD_BASE : VPD_BASE_OLD;
+
+ for (i = 0; i < sizeof(vpd); i += 4) {
+ ret = csio_hw_seeprom_read(hw, addr + i, (uint32_t *)(vpd + i));
+ if (ret)
+ return ret;
+ }
+
+ /* Reset the VPD flag! */
+ hw->flags &= (~CSIO_HWF_VPD_VALID);
+
+ v = (const struct t4_vpd_hdr *)vpd;
+
+#define FIND_VPD_KW(var, name) do { \
+ var = csio_hw_get_vpd_keyword_val(v, name); \
+ if (var < 0) { \
+ csio_err(hw, "missing VPD keyword " name "\n"); \
+ return CSIO_INVAL; \
+ } \
+} while (0)
+
+ FIND_VPD_KW(i, "RV");
+ for (csum = 0; i >= 0; i--)
+ csum += vpd[i];
+
+ if (csum) {
+ csio_err(hw, "corrupted VPD EEPROM, actual csum %u\n", csum);
+ return CSIO_INVAL;
+ }
+ FIND_VPD_KW(ec, "EC");
+ FIND_VPD_KW(sn, "SN");
+#undef FIND_VPD_KW
+
+ memcpy(p->id, v->id_data, ID_LEN);
+ s = strstrip(p->id);
+ memcpy(p->ec, vpd + ec, EC_LEN);
+ s = strstrip(p->ec);
+ i = vpd[sn - VPD_INFO_FLD_HDR_SIZE + 2];
+ memcpy(p->sn, vpd + sn, min(i, SERNUM_LEN));
+ s = strstrip(p->sn);
+
+ csio_valid_vpd_copied(hw);
+ return 0;
+}
+
+/*
+ * csio_hw_sf1_read - read data from the serial flash
+ * @hw: the HW module
+ * @byte_cnt: number of bytes to read
+ * @cont: whether another operation will be chained
+ * @lock: whether to lock SF for PL access only
+ * @valp: where to store the read data
+ *
+ * Reads up to 4 bytes of data from the serial flash. The location of
+ * the read needs to be specified prior to calling this by issuing the
+ * appropriate commands to the serial flash.
+ */
+static csio_retval_t
+csio_hw_sf1_read(struct csio_hw *hw, uint32_t byte_cnt, int32_t cont,
+ int32_t lock, uint32_t *valp)
+{
+ csio_retval_t ret;
+
+ if (!byte_cnt || byte_cnt > 4)
+ return CSIO_INVAL;
+ if (csio_rd_reg32(hw, SF_OP) & SF_BUSY)
+ return CSIO_BUSY;
+
+ cont = cont ? SF_CONT : 0;
+ lock = lock ? SF_LOCK : 0;
+
+ csio_wr_reg32(hw, lock | cont | BYTECNT(byte_cnt - 1), SF_OP);
+ ret = csio_hw_wait_op_done_val(hw, SF_OP, SF_BUSY, 0, SF_ATTEMPTS,
+ 10, NULL);
+ if (!ret)
+ *valp = csio_rd_reg32(hw, SF_DATA);
+ return ret;
+}
+
+/*
+ * csio_hw_sf1_write - write data to the serial flash
+ * @hw: the HW module
+ * @byte_cnt: number of bytes to write
+ * @cont: whether another operation will be chained
+ * @lock: whether to lock SF for PL access only
+ * @val: value to write
+ *
+ * Writes up to 4 bytes of data to the serial flash. The location of
+ * the write needs to be specified prior to calling this by issuing the
+ * appropriate commands to the serial flash.
+ */
+static csio_retval_t
+csio_hw_sf1_write(struct csio_hw *hw, uint32_t byte_cnt, uint32_t cont,
+ int32_t lock, uint32_t val)
+{
+ if (!byte_cnt || byte_cnt > 4)
+ return CSIO_INVAL;
+ if (csio_rd_reg32(hw, SF_OP) & SF_BUSY)
+ return CSIO_BUSY;
+
+ cont = cont ? SF_CONT : 0;
+ lock = lock ? SF_LOCK : 0;
+
+ csio_wr_reg32(hw, val, SF_DATA);
+ csio_wr_reg32(hw, cont | BYTECNT(byte_cnt - 1) | OP_WR | lock, SF_OP);
+
+ return csio_hw_wait_op_done_val(hw, SF_OP, SF_BUSY, 0, SF_ATTEMPTS,
+ 10, NULL);
+}
+
+/*
+ * csio_hw_flash_wait_op - wait for a flash operation to complete
+ * @hw: the HW module
+ * @attempts: max number of polls of the status register
+ * @delay: delay between polls in ms
+ *
+ * Wait for a flash operation to complete by polling the status register.
+ */
+static csio_retval_t
+csio_hw_flash_wait_op(struct csio_hw *hw, int32_t attempts, int32_t delay)
+{
+ csio_retval_t ret;
+ uint32_t status;
+
+ while (1) {
+ ret = csio_hw_sf1_write(hw, 1, 1, 1, SF_RD_STATUS);
+ if (ret != CSIO_SUCCESS)
+ return ret;
+
+ ret = csio_hw_sf1_read(hw, 1, 0, 1, &status);
+ if (ret != CSIO_SUCCESS)
+ return ret;
+
+ if (!(status & 1))
+ return CSIO_SUCCESS;
+ if (--attempts == 0)
+ return CSIO_RETRY;
+ if (delay)
+ msleep(delay);
+ }
+}
+
+/*
+ * csio_hw_read_flash - read words from serial flash
+ * @hw: the HW module
+ * @addr: the start address for the read
+ * @nwords: how many 32-bit words to read
+ * @data: where to store the read data
+ * @byte_oriented: whether to store data as bytes or as words
+ *
+ * Read the specified number of 32-bit words from the serial flash.
+ * If @byte_oriented is set the read data is stored as a byte array
+ * (i.e., big-endian), otherwise as 32-bit words in the platform's
+ * natural endianess.
+ */
+static csio_retval_t
+csio_hw_read_flash(struct csio_hw *hw, uint32_t addr, uint32_t nwords,
+ uint32_t *data, int32_t byte_oriented)
+{
+ csio_retval_t ret;
+
+ if (addr + nwords * sizeof(uint32_t) > hw->params.sf_size || (addr & 3))
+ return CSIO_INVAL;
+
+ addr = swab32(addr) | SF_RD_DATA_FAST;
+
+ ret = csio_hw_sf1_write(hw, 4, 1, 0, addr);
+ if (ret != CSIO_SUCCESS)
+ return ret;
+
+ ret = csio_hw_sf1_read(hw, 1, 1, 0, data);
+ if (ret != CSIO_SUCCESS)
+ return ret;
+
+ for ( ; nwords; nwords--, data++) {
+ ret = csio_hw_sf1_read(hw, 4, nwords > 1, nwords == 1, data);
+ if (nwords == 1)
+ csio_wr_reg32(hw, 0, SF_OP); /* unlock SF */
+ if (ret)
+ return ret;
+ if (byte_oriented)
+ *data = htonl(*data);
+ }
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_hw_write_flash - write up to a page of data to the serial flash
+ * @hw: the hw
+ * @addr: the start address to write
+ * @n: length of data to write in bytes
+ * @data: the data to write
+ *
+ * Writes up to a page of data (256 bytes) to the serial flash starting
+ * at the given address. All the data must be written to the same page.
+ */
+static csio_retval_t
+csio_hw_write_flash(struct csio_hw *hw, uint32_t addr,
+ uint32_t n, const uint8_t *data)
+{
+ csio_retval_t ret = CSIO_INVAL;
+ uint32_t buf[64];
+ uint32_t i, c, left, val, offset = addr & 0xff;
+
+ if (addr >= hw->params.sf_size || offset + n > SF_PAGE_SIZE)
+ return CSIO_INVAL;
+
+ val = swab32(addr) | SF_PROG_PAGE;
+
+ ret = csio_hw_sf1_write(hw, 1, 0, 1, SF_WR_ENABLE);
+ if (ret != CSIO_SUCCESS)
+ goto unlock;
+
+ ret = csio_hw_sf1_write(hw, 4, 1, 1, val);
+ if (ret != CSIO_SUCCESS)
+ goto unlock;
+
+ for (left = n; left; left -= c) {
+ c = min(left, 4U);
+ for (val = 0, i = 0; i < c; ++i)
+ val = (val << 8) + *data++;
+
+ ret = csio_hw_sf1_write(hw, c, c != left, 1, val);
+ if (ret)
+ goto unlock;
+ }
+ ret = csio_hw_flash_wait_op(hw, 8, 1);
+ if (ret)
+ goto unlock;
+
+ csio_wr_reg32(hw, 0, SF_OP); /* unlock SF */
+
+ /* Read the page to verify the write succeeded */
+ ret = csio_hw_read_flash(hw, addr & ~0xff, ARRAY_SIZE(buf), buf, 1);
+ if (ret)
+ return ret;
+
+ if (memcmp(data - n, (uint8_t *)buf + offset, n)) {
+ csio_err(hw,
+ "failed to correctly write the flash page at %#x\n",
+ addr);
+ return CSIO_INVAL;
+ }
+
+ return CSIO_SUCCESS;
+
+unlock:
+ csio_wr_reg32(hw, 0, SF_OP); /* unlock SF */
+ return ret;
+}
+
+/*
+ * csio_hw_flash_erase_sectors - erase a range of flash sectors
+ * @hw: the HW module
+ * @start: the first sector to erase
+ * @end: the last sector to erase
+ *
+ * Erases the sectors in the given inclusive range.
+ */
+static csio_retval_t
+csio_hw_flash_erase_sectors(struct csio_hw *hw, int32_t start, int32_t end)
+{
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ while (start <= end) {
+
+ ret = csio_hw_sf1_write(hw, 1, 0, 1, SF_WR_ENABLE);
+ if (ret != CSIO_SUCCESS)
+ goto out;
+
+ ret = csio_hw_sf1_write(hw, 4, 0, 1,
+ SF_ERASE_SECTOR | (start << 8));
+ if (ret != CSIO_SUCCESS)
+ goto out;
+
+ ret = csio_hw_flash_wait_op(hw, 14, 500);
+ if (ret != CSIO_SUCCESS)
+ goto out;
+
+ start++;
+ }
+out:
+ if (ret)
+ csio_err(hw, "erase of flash sector %d failed, error %d\n",
+ start, ret);
+ csio_wr_reg32(hw, 0, SF_OP); /* unlock SF */
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_hw_flash_cfg_addr - return the address of the flash
+ * configuration file
+ * @hw: the HW module
+ *
+ * Return the address within the flash where the Firmware Configuration
+ * File is stored.
+ */
+static unsigned int
+csio_hw_flash_cfg_addr(struct csio_hw *hw)
+{
+ if (hw->params.sf_size == 0x100000)
+ return FPGA_FLASH_CFG_OFFSET;
+ else
+ return FLASH_CFG_OFFSET;
+}
+
+static void
+csio_hw_print_fw_version(struct csio_hw *hw, char *str)
+{
+ csio_info(hw, "%s: %u.%u.%u.%u\n", str,
+ FW_HDR_FW_VER_MAJOR_GET(hw->fwrev),
+ FW_HDR_FW_VER_MINOR_GET(hw->fwrev),
+ FW_HDR_FW_VER_MICRO_GET(hw->fwrev),
+ FW_HDR_FW_VER_BUILD_GET(hw->fwrev));
+}
+
+/*
+ * csio_hw_get_fw_version - read the firmware version
+ * @hw: HW module
+ * @vers: where to place the version
+ *
+ * Reads the FW version from flash.
+ */
+static csio_retval_t
+csio_hw_get_fw_version(struct csio_hw *hw, uint32_t *vers)
+{
+ return csio_hw_read_flash(hw, FW_IMG_START +
+ offsetof(struct fw_hdr, fw_ver), 1,
+ vers, 0);
+}
+
+/*
+ * csio_hw_get_tp_version - read the TP microcode version
+ * @hw: HW module
+ * @vers: where to place the version
+ *
+ * Reads the TP microcode version from flash.
+ */
+static int
+csio_hw_get_tp_version(struct csio_hw *hw, u32 *vers)
+{
+ return csio_hw_read_flash(hw, FLASH_FW_START +
+ offsetof(struct fw_hdr, tp_microcode_ver), 1,
+ vers, 0);
+}
+
+/*
+ * csio_hw_check_fw_version - check if the FW is compatible with
+ * this driver
+ * @hw: HW module
+ *
+ * Checks if an adapter's FW is compatible with the driver. Returns 0
+ * if there's exact match, a negative error if the version could not be
+ * read or there's a major/minor version mismatch/minor.
+ */
+static int
+csio_hw_check_fw_version(struct csio_hw *hw)
+{
+ int ret, major, minor, micro;
+
+ ret = csio_hw_get_fw_version(hw, &hw->fwrev);
+ if (!ret)
+ ret = csio_hw_get_tp_version(hw, &hw->tp_vers);
+ if (ret)
+ return ret;
+
+ major = FW_HDR_FW_VER_MAJOR_GET(hw->fwrev);
+ minor = FW_HDR_FW_VER_MINOR_GET(hw->fwrev);
+ micro = FW_HDR_FW_VER_MICRO_GET(hw->fwrev);
+
+ if (major != FW_VERSION_MAJOR) { /* major mismatch - fail */
+ csio_err(hw, "card FW has major version %u, driver wants %u\n",
+ major, FW_VERSION_MAJOR);
+ return CSIO_INVAL;
+ }
+
+ if (minor == FW_VERSION_MINOR && micro == FW_VERSION_MICRO)
+ return CSIO_SUCCESS; /* perfect match */
+
+ /* Minor/micro version mismatch */
+ return CSIO_INVAL;
+}
+
+/*
+ * csio_hw_fw_dload - download firmware.
+ * @hw: HW module
+ * @fw_data: firmware image to write.
+ * @size: image size
+ *
+ * Write the supplied firmware image to the card's serial flash.
+ */
+static csio_retval_t
+csio_hw_fw_dload(struct csio_hw *hw, uint8_t *fw_data, uint32_t size)
+{
+ uint32_t csum;
+ int32_t addr;
+ csio_retval_t ret;
+ uint32_t i;
+ uint8_t first_page[SF_PAGE_SIZE];
+ const uint32_t *p = (const uint32_t *)fw_data;
+ struct fw_hdr *hdr = (struct fw_hdr *)fw_data;
+ uint32_t sf_sec_size;
+
+ if ((!hw->params.sf_size) || (!hw->params.sf_nsec)) {
+ csio_err(hw, "Serial Flash data invalid\n");
+ return CSIO_INVAL;
+ }
+
+ if (!size) {
+ csio_err(hw, "FW image has no data\n");
+ return CSIO_INVAL;
+ }
+
+ if (size & 511) {
+ csio_err(hw, "FW image size not multiple of 512 bytes\n");
+ return CSIO_INVAL;
+ }
+
+ if (ntohs(hdr->len512) * 512 != size) {
+ csio_err(hw, "FW image size differs from size in FW header\n");
+ return CSIO_INVAL;
+ }
+
+ if (size > FW_MAX_SIZE) {
+ csio_err(hw, "FW image too large, max is %u bytes\n",
+ FW_MAX_SIZE);
+ return CSIO_INVAL;
+ }
+
+ for (csum = 0, i = 0; i < size / sizeof(csum); i++)
+ csum += ntohl(p[i]);
+
+ if (csum != 0xffffffff) {
+ csio_err(hw, "corrupted firmware image, checksum %#x\n", csum);
+ return CSIO_INVAL;
+ }
+
+ sf_sec_size = hw->params.sf_size / hw->params.sf_nsec;
+ i = CSIO_ROUNDUP(size, sf_sec_size); /* # of sectors spanned */
+
+ csio_dbg(hw, "Erasing sectors... start:%d end:%d\n",
+ FW_START_SEC, FW_START_SEC + i - 1);
+
+ ret = csio_hw_flash_erase_sectors(hw, FW_START_SEC,
+ FW_START_SEC + i - 1);
+ if (ret) {
+ csio_err(hw, "Flash Erase failed\n");
+ goto out;
+ }
+
+ /*
+ * We write the correct version at the end so the driver can see a bad
+ * version if the FW write fails. Start by writing a copy of the
+ * first page with a bad version.
+ */
+ memcpy(first_page, fw_data, SF_PAGE_SIZE);
+ ((struct fw_hdr *)first_page)->fw_ver = htonl(0xffffffff);
+ ret = csio_hw_write_flash(hw, FW_IMG_START, SF_PAGE_SIZE, first_page);
+ if (ret)
+ goto out;
+
+ csio_dbg(hw, "Writing Flash .. start:%d end:%d\n",
+ FW_IMG_START, FW_IMG_START + size);
+
+ addr = FW_IMG_START;
+ for (size -= SF_PAGE_SIZE; size; size -= SF_PAGE_SIZE) {
+ addr += SF_PAGE_SIZE;
+ fw_data += SF_PAGE_SIZE;
+ ret = csio_hw_write_flash(hw, addr, SF_PAGE_SIZE, fw_data);
+ if (ret)
+ goto out;
+ }
+
+ ret = csio_hw_write_flash(hw,
+ FW_IMG_START +
+ offsetof(struct fw_hdr, fw_ver),
+ sizeof(hdr->fw_ver),
+ (const uint8_t *)&hdr->fw_ver);
+
+out:
+ if (ret)
+ csio_err(hw, "firmware download failed, error %d\n", ret);
+ return ret;
+}
+
+static csio_retval_t
+csio_hw_get_flash_params(struct csio_hw *hw)
+{
+ csio_retval_t ret;
+ uint32_t info = 0;
+
+ ret = csio_hw_sf1_write(hw, 1, 1, 0, SF_RD_ID);
+ if (!ret)
+ ret = csio_hw_sf1_read(hw, 3, 0, 1, &info);
+ csio_wr_reg32(hw, 0, SF_OP); /* unlock SF */
+ if (ret != CSIO_SUCCESS)
+ return ret;
+
+ if ((info & 0xff) != 0x20) /* not a Numonix flash */
+ return CSIO_INVAL;
+ info >>= 16; /* log2 of size */
+ if (info >= 0x14 && info < 0x18)
+ hw->params.sf_nsec = 1 << (info - 16);
+ else if (info == 0x18)
+ hw->params.sf_nsec = 64;
+ else
+ return CSIO_INVAL;
+ hw->params.sf_size = 1 << info;
+
+ return CSIO_SUCCESS;
+}
+
+static void
+csio_set_pcie_completion_timeout(struct csio_hw *hw, u8 range)
+{
+ uint16_t val;
+ uint32_t pcie_cap;
+
+ if (!csio_pci_capability(hw->pdev, PCI_CAP_ID_EXP, &pcie_cap)) {
+ pci_read_config_word(hw->pdev,
+ pcie_cap + PCI_EXP_DEVCTL2, &val);
+ val &= 0xfff0;
+ val |= range ;
+ pci_write_config_word(hw->pdev,
+ pcie_cap + PCI_EXP_DEVCTL2, val);
+ }
+}
+
+
+/*
+ * Return the specified PCI-E Configuration Space register from our Physical
+ * Function. We try first via a Firmware LDST Command since we prefer to let
+ * the firmware own all of these registers, but if that fails we go for it
+ * directly ourselves.
+ */
+static uint32_t
+csio_read_pcie_cfg4(struct csio_hw *hw, int reg)
+{
+ u32 val = 0;
+ struct csio_mb *mbp;
+ int rv;
+ struct fw_ldst_cmd *ldst_cmd;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ pci_read_config_dword(hw->pdev, reg, &val);
+ return val;
+ }
+
+ csio_mb_ldst(hw, mbp, CSIO_MB_DEFAULT_TMO, reg);
+
+ rv = csio_mb_issue(hw, mbp);
+
+ /*
+ * If the LDST Command suucceeded, exctract the returned register
+ * value. Otherwise read it directly ourself.
+ */
+ if (rv == 0) {
+ ldst_cmd = (struct fw_ldst_cmd *)(mbp->mb);
+ val = ntohl(ldst_cmd->u.pcie.data[0]);
+ } else
+ pci_read_config_dword(hw->pdev, reg, &val);
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return val;
+} /* csio_read_pcie_cfg4 */
+
+static int
+csio_hw_set_mem_win(struct csio_hw *hw)
+{
+ u32 bar0;
+
+ /*
+ * Truncation intentional: we only read the bottom 32-bits of the
+ * 64-bit BAR0/BAR1 ... We use the hardware backdoor mechanism to
+ * read BAR0 instead of using pci_resource_start() because we could be
+ * operating from within a Virtual Machine which is trapping our
+ * accesses to our Configuration Space and we need to set up the PCI-E
+ * Memory Window decoders with the actual addresses which will be
+ * coming across the PCI-E link.
+ */
+ bar0 = csio_read_pcie_cfg4(hw, PCI_BASE_ADDRESS_0);
+ bar0 &= PCI_BASE_ADDRESS_MEM_MASK;
+
+ /*
+ * Set up memory window for accessing adapter memory ranges. (Read
+ * back MA register to ensure that changes propagate before we attempt
+ * to use the new values.)
+ */
+ csio_wr_reg32(hw, (bar0 + MEMWIN0_BASE) | BIR(0) |
+ WINDOW(ilog2(MEMWIN0_APERTURE) - 10),
+ PCIE_MEM_ACCESS_REG(PCIE_MEM_ACCESS_BASE_WIN, 0));
+ csio_wr_reg32(hw, (bar0 + MEMWIN1_BASE) | BIR(0) |
+ WINDOW(ilog2(MEMWIN1_APERTURE) - 10),
+ PCIE_MEM_ACCESS_REG(PCIE_MEM_ACCESS_BASE_WIN, 1));
+ csio_wr_reg32(hw, (bar0 + MEMWIN2_BASE) | BIR(0) |
+ WINDOW(ilog2(MEMWIN2_APERTURE) - 10),
+ PCIE_MEM_ACCESS_REG(PCIE_MEM_ACCESS_BASE_WIN, 2));
+ csio_rd_reg32(hw, PCIE_MEM_ACCESS_REG(PCIE_MEM_ACCESS_BASE_WIN, 2));
+ return 0;
+} /* csio_hw_set_mem_win */
+
+
+
+/*****************************************************************************/
+/* HW State machine assists */
+/*****************************************************************************/
+
+static csio_retval_t
+csio_hw_dev_ready(struct csio_hw *hw)
+{
+ uint32_t reg;
+ int cnt = 6;
+
+ while (((reg = csio_rd_reg32(hw, PL_WHOAMI)) == 0xFFFFFFFF) &&
+ (--cnt != 0))
+ mdelay(100);
+
+ if ((cnt == 0) && (((int32_t)(SOURCEPF_GET(reg)) < 0) ||
+ (SOURCEPF_GET(reg) >= CSIO_MAX_PFN))) {
+ csio_err(hw, "PL_WHOAMI returned 0x%x, cnt:%d\n", reg, cnt);
+ return CSIO_EIO;
+ }
+
+ hw->pfn = SOURCEPF_GET(reg);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_do_hello - Perform the HELLO FW Mailbox command and process response.
+ * @hw: HW module
+ * @state: Device state
+ *
+ * FW_HELLO_CMD has to be polled for completion.
+ */
+static csio_retval_t
+csio_do_hello(struct csio_hw *hw, enum csio_dev_state *state)
+{
+ struct csio_mb *mbp;
+ csio_retval_t rv = CSIO_SUCCESS;
+ enum csio_dev_master master;
+ enum fw_retval retval;
+ uint8_t mpfn;
+ char state_str[16];
+ int retries = FW_CMD_HELLO_RETRIES;
+
+ memset(state_str, 0, sizeof(state_str));
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ rv = CSIO_NOMEM;
+ csio_inc_stats(hw, n_err_nomem);
+ goto out;
+ }
+
+ master = csio_force_master ? CSIO_MASTER_MUST : CSIO_MASTER_MAY;
+
+retry:
+ csio_mb_hello(hw, mbp, CSIO_MB_DEFAULT_TMO, hw->pfn,
+ hw->pfn, master, NULL);
+
+ rv = csio_mb_issue(hw, mbp);
+ if (rv) {
+ csio_err(hw, "failed to issue HELLO cmd. ret:%d.\n", rv);
+ goto out_free_mb;
+ }
+
+ csio_mb_process_hello_rsp(hw, mbp, &retval, state, &mpfn);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "HELLO cmd failed with ret: %d\n", retval);
+ rv = CSIO_INVAL;
+ goto out_free_mb;
+ }
+
+ /* Firmware has designated us to be master */
+ if (hw->pfn == mpfn) {
+ hw->flags |= CSIO_HWF_MASTER;
+ } else if (*state == CSIO_DEV_STATE_UNINIT) {
+ /*
+ * If we're not the Master PF then we need to wait around for
+ * the Master PF Driver to finish setting up the adapter.
+ *
+ * Note that we also do this wait if we're a non-Master-capable
+ * PF and there is no current Master PF; a Master PF may show up
+ * momentarily and we wouldn't want to fail pointlessly. (This
+ * can happen when an OS loads lots of different drivers rapidly
+ * at the same time). In this case, the Master PF returned by
+ * the firmware will be PCIE_FW_MASTER_MASK so the test below
+ * will work ...
+ */
+
+ int waiting = FW_CMD_HELLO_TIMEOUT;
+
+ /*
+ * Wait for the firmware to either indicate an error or
+ * initialized state. If we see either of these we bail out
+ * and report the issue to the caller. If we exhaust the
+ * "hello timeout" and we haven't exhausted our retries, try
+ * again. Otherwise bail with a timeout error.
+ */
+ for (;;) {
+ uint32_t pcie_fw;
+
+ msleep(50);
+ waiting -= 50;
+
+ /*
+ * If neither Error nor Initialialized are indicated
+ * by the firmware keep waiting till we exaust our
+ * timeout ... and then retry if we haven't exhausted
+ * our retries ...
+ */
+ pcie_fw = csio_rd_reg32(hw, PCIE_FW);
+ if (!(pcie_fw & (PCIE_FW_ERR|PCIE_FW_INIT))) {
+ if (waiting <= 0) {
+ if (retries-- > 0)
+ goto retry;
+
+ rv = CSIO_TIMEOUT;
+ break;
+ }
+ continue;
+ }
+
+ /*
+ * We either have an Error or Initialized condition
+ * report errors preferentially.
+ */
+ if (state) {
+ if (pcie_fw & PCIE_FW_ERR) {
+ *state = CSIO_DEV_STATE_ERR;
+ rv = CSIO_TIMEOUT;
+ } else if (pcie_fw & PCIE_FW_INIT)
+ *state = CSIO_DEV_STATE_INIT;
+ }
+
+ /*
+ * If we arrived before a Master PF was selected and
+ * there's not a valid Master PF, grab its identity
+ * for our caller.
+ */
+ if (mpfn == PCIE_FW_MASTER_MASK &&
+ (pcie_fw & PCIE_FW_MASTER_VLD))
+ mpfn = PCIE_FW_MASTER_GET(pcie_fw);
+ break;
+ }
+ hw->flags &= ~CSIO_HWF_MASTER;
+ }
+
+ switch (*state) {
+ case CSIO_DEV_STATE_UNINIT:
+ strcpy(state_str, "Initializing");
+ break;
+ case CSIO_DEV_STATE_INIT:
+ strcpy(state_str, "Initialized");
+ break;
+ case CSIO_DEV_STATE_ERR:
+ strcpy(state_str, "Error");
+ break;
+ default:
+ strcpy(state_str, "Unknown");
+ break;
+ }
+
+ if (hw->pfn == mpfn)
+ csio_info(hw, "PF: %d, Coming up as MASTER, HW state: %s\n",
+ hw->pfn, state_str);
+ else
+ csio_info(hw,
+ "PF: %d, Coming up as SLAVE, Master PF: %d, HW state: %s\n",
+ hw->pfn, mpfn, state_str);
+
+out_free_mb:
+ mempool_free(mbp, hw->mb_mempool);
+out:
+ return rv;
+}
+
+/*
+ * csio_do_bye - Perform the BYE FW Mailbox command and process response.
+ * @hw: HW module
+ *
+ */
+static csio_retval_t
+csio_do_bye(struct csio_hw *hw)
+{
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ csio_mb_bye(hw, mbp, CSIO_MB_DEFAULT_TMO, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of BYE command failed\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_do_reset- Perform the device reset.
+ * @hw: HW module
+ * @fw_rst: FW reset
+ *
+ * If fw_rst is set, issues FW reset mbox cmd otherwise
+ * does PIO reset.
+ * Performs reset of the function.
+ */
+static csio_retval_t
+csio_do_reset(struct csio_hw *hw, bool fw_rst)
+{
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+
+ if (!fw_rst) {
+ /* PIO reset */
+ csio_wr_reg32(hw, PIORSTMODE | PIORST, PL_RST);
+ mdelay(2000);
+ return CSIO_SUCCESS;
+ }
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ csio_mb_reset(hw, mbp, CSIO_MB_DEFAULT_TMO,
+ PIORSTMODE | PIORST, 0, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of RESET command failed.n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "RESET cmd failed with ret:0x%x.\n", retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+static csio_retval_t
+csio_hw_validate_caps(struct csio_hw *hw, struct csio_mb *mbp)
+{
+ struct fw_caps_config_cmd *rsp = (struct fw_caps_config_cmd *)mbp->mb;
+ uint16_t caps;
+
+ caps = ntohs(rsp->fcoecaps);
+
+ if (!(caps & FW_CAPS_CONFIG_FCOE_INITIATOR)) {
+ csio_err(hw, "No FCoE Initiator capability in the firmware.\n");
+ return CSIO_INVAL;
+ }
+
+ if (!(caps & FW_CAPS_CONFIG_FCOE_CTRL_OFLD)) {
+ csio_err(hw, "No FCoE Control Offload capability\n");
+ return CSIO_INVAL;
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_hw_fw_halt - issue a reset/halt to FW and put uP into RESET
+ * @hw: the HW module
+ * @mbox: mailbox to use for the FW RESET command (if desired)
+ * @force: force uP into RESET even if FW RESET command fails
+ *
+ * Issues a RESET command to firmware (if desired) with a HALT indication
+ * and then puts the microprocessor into RESET state. The RESET command
+ * will only be issued if a legitimate mailbox is provided (mbox <=
+ * PCIE_FW_MASTER_MASK).
+ *
+ * This is generally used in order for the host to safely manipulate the
+ * adapter without fear of conflicting with whatever the firmware might
+ * be doing. The only way out of this state is to RESTART the firmware
+ * ...
+ */
+static csio_retval_t
+csio_hw_fw_halt(struct csio_hw *hw, uint32_t mbox, int32_t force)
+{
+ enum fw_retval retval = 0;
+
+ /*
+ * If a legitimate mailbox is provided, issue a RESET command
+ * with a HALT indication.
+ */
+ if (mbox <= PCIE_FW_MASTER_MASK) {
+ struct csio_mb *mbp;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ csio_mb_reset(hw, mbp, CSIO_MB_DEFAULT_TMO,
+ PIORSTMODE | PIORST, FW_RESET_CMD_HALT,
+ NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of RESET command failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ mempool_free(mbp, hw->mb_mempool);
+ }
+
+ /*
+ * Normally we won't complete the operation if the firmware RESET
+ * command fails but if our caller insists we'll go ahead and put the
+ * uP into RESET. This can be useful if the firmware is hung or even
+ * missing ... We'll have to take the risk of putting the uP into
+ * RESET without the cooperation of firmware in that case.
+ *
+ * We also force the firmware's HALT flag to be on in case we bypassed
+ * the firmware RESET command above or we're dealing with old firmware
+ * which doesn't have the HALT capability. This will serve as a flag
+ * for the incoming firmware to know that it's coming out of a HALT
+ * rather than a RESET ... if it's new enough to understand that ...
+ */
+ if (retval == 0 || force) {
+ csio_set_reg_field(hw, CIM_BOOT_CFG, UPCRST, UPCRST);
+ csio_set_reg_field(hw, PCIE_FW, PCIE_FW_HALT, PCIE_FW_HALT);
+ }
+
+ /*
+ * And we always return the result of the firmware RESET command
+ * even when we force the uP into RESET ...
+ */
+ return retval ? CSIO_INVAL : CSIO_SUCCESS;
+}
+
+/*
+ * csio_hw_fw_restart - restart the firmware by taking the uP out of RESET
+ * @hw: the HW module
+ * @reset: if we want to do a RESET to restart things
+ *
+ * Restart firmware previously halted by csio_hw_fw_halt(). On successful
+ * return the previous PF Master remains as the new PF Master and there
+ * is no need to issue a new HELLO command, etc.
+ *
+ * We do this in two ways:
+ *
+ * 1. If we're dealing with newer firmware we'll simply want to take
+ * the chip's microprocessor out of RESET. This will cause the
+ * firmware to start up from its start vector. And then we'll loop
+ * until the firmware indicates it's started again (PCIE_FW.HALT
+ * reset to 0) or we timeout.
+ *
+ * 2. If we're dealing with older firmware then we'll need to RESET
+ * the chip since older firmware won't recognize the PCIE_FW.HALT
+ * flag and automatically RESET itself on startup.
+ */
+static csio_retval_t
+csio_hw_fw_restart(struct csio_hw *hw, uint32_t mbox, int32_t reset)
+{
+ if (reset) {
+ /*
+ * Since we're directing the RESET instead of the firmware
+ * doing it automatically, we need to clear the PCIE_FW.HALT
+ * bit.
+ */
+ csio_set_reg_field(hw, PCIE_FW, PCIE_FW_HALT, 0);
+
+ /*
+ * If we've been given a valid mailbox, first try to get the
+ * firmware to do the RESET. If that works, great and we can
+ * return success. Otherwise, if we haven't been given a
+ * valid mailbox or the RESET command failed, fall back to
+ * hitting the chip with a hammer.
+ */
+ if (mbox <= PCIE_FW_MASTER_MASK) {
+ csio_set_reg_field(hw, CIM_BOOT_CFG, UPCRST, 0);
+ msleep(100);
+ if (csio_do_reset(hw, CSIO_TRUE) == 0)
+ return CSIO_SUCCESS;
+ }
+
+ csio_wr_reg32(hw, PIORSTMODE | PIORST, PL_RST);
+ msleep(2000);
+ } else {
+ int ms;
+
+ csio_set_reg_field(hw, CIM_BOOT_CFG, UPCRST, 0);
+ for (ms = 0; ms < FW_CMD_MAX_TIMEOUT; ) {
+ if (!(csio_rd_reg32(hw, PCIE_FW) & PCIE_FW_HALT))
+ return CSIO_SUCCESS;
+ msleep(100);
+ ms += 100;
+ }
+ return CSIO_TIMEOUT;
+ }
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_hw_fw_upgrade - perform all of the steps necessary to upgrade FW
+ * @hw: the HW module
+ * @mbox: mailbox to use for the FW RESET command (if desired)
+ * @fw_data: the firmware image to write
+ * @size: image size
+ * @force: force upgrade even if firmware doesn't cooperate
+ *
+ * Perform all of the steps necessary for upgrading an adapter's
+ * firmware image. Normally this requires the cooperation of the
+ * existing firmware in order to halt all existing activities
+ * but if an invalid mailbox token is passed in we skip that step
+ * (though we'll still put the adapter microprocessor into RESET in
+ * that case).
+ *
+ * On successful return the new firmware will have been loaded and
+ * the adapter will have been fully RESET losing all previous setup
+ * state. On unsuccessful return the adapter may be completely hosed ...
+ * positive errno indicates that the adapter is ~probably~ intact, a
+ * negative errno indicates that things are looking bad ...
+ */
+static csio_retval_t
+csio_hw_fw_upgrade(struct csio_hw *hw, uint32_t mbox,
+ const u8 *fw_data, uint32_t size, int32_t force)
+{
+ const struct fw_hdr *fw_hdr = (const struct fw_hdr *)fw_data;
+ csio_retval_t reset, ret;
+
+ ret = csio_hw_fw_halt(hw, mbox, force);
+ if (ret != CSIO_SUCCESS && !force)
+ return ret;
+
+ ret = csio_hw_fw_dload(hw, (uint8_t *) fw_data, size);
+ if (ret != CSIO_SUCCESS)
+ return ret;
+
+ /*
+ * Older versions of the firmware don't understand the new
+ * PCIE_FW.HALT flag and so won't know to perform a RESET when they
+ * restart. So for newly loaded older firmware we'll have to do the
+ * RESET for it so it starts up on a clean slate. We can tell if
+ * the newly loaded firmware will handle this right by checking
+ * its header flags to see if it advertises the capability.
+ */
+ reset = ((ntohl(fw_hdr->flags) & FW_HDR_FLAGS_RESET_HALT) == 0);
+ return csio_hw_fw_restart(hw, mbox, reset);
+}
+
+
+/*
+ * csio_hw_fw_config_file - setup an adapter via a Configuration File
+ * @hw: the HW module
+ * @mbox: mailbox to use for the FW command
+ * @mtype: the memory type where the Configuration File is located
+ * @maddr: the memory address where the Configuration File is located
+ * @finiver: return value for CF [fini] version
+ * @finicsum: return value for CF [fini] checksum
+ * @cfcsum: return value for CF computed checksum
+ *
+ * Issue a command to get the firmware to process the Configuration
+ * File located at the specified mtype/maddress. If the Configuration
+ * File is processed successfully and return value pointers are
+ * provided, the Configuration File "[fini] section version and
+ * checksum values will be returned along with the computed checksum.
+ * It's up to the caller to decide how it wants to respond to the
+ * checksums not matching but it recommended that a prominant warning
+ * be emitted in order to help people rapidly identify changed or
+ * corrupted Configuration Files.
+ *
+ * Also note that it's possible to modify things like "niccaps",
+ * "toecaps",etc. between processing the Configuration File and telling
+ * the firmware to use the new configuration. Callers which want to
+ * do this will need to "hand-roll" their own CAPS_CONFIGS commands for
+ * Configuration Files if they want to do this.
+ */
+static csio_retval_t
+csio_hw_fw_config_file(struct csio_hw *hw,
+ unsigned int mtype, unsigned int maddr,
+ uint32_t *finiver, uint32_t *finicsum, uint32_t *cfcsum)
+{
+ struct csio_mb *mbp;
+ struct fw_caps_config_cmd *caps_cmd;
+ csio_retval_t rv = CSIO_INVAL;
+ enum fw_retval ret;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+ /*
+ * Tell the firmware to process the indicated Configuration File.
+ * If there are no errors and the caller has provided return value
+ * pointers for the [fini] section version, checksum and computed
+ * checksum, pass those back to the caller.
+ */
+ caps_cmd = (struct fw_caps_config_cmd *)(mbp->mb);
+ CSIO_INIT_MBP(mbp, caps_cmd, CSIO_MB_DEFAULT_TMO, hw, NULL, 1);
+ caps_cmd->op_to_write =
+ htonl(FW_CMD_OP(FW_CAPS_CONFIG_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_READ);
+ caps_cmd->cfvalid_to_len16 =
+ htonl(FW_CAPS_CONFIG_CMD_CFVALID |
+ FW_CAPS_CONFIG_CMD_MEMTYPE_CF(mtype) |
+ FW_CAPS_CONFIG_CMD_MEMADDR64K_CF(maddr >> 16) |
+ FW_LEN16(*caps_cmd));
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FW_CAPS_CONFIG_CMD failed!\n");
+ goto out;
+ }
+
+ ret = csio_mb_fw_retval(mbp);
+ if (ret != FW_SUCCESS) {
+ csio_dbg(hw, "FW_CAPS_CONFIG_CMD returned %d!\n", rv);
+ goto out;
+ }
+
+ if (finiver)
+ *finiver = ntohl(caps_cmd->finiver);
+ if (finicsum)
+ *finicsum = ntohl(caps_cmd->finicsum);
+ if (cfcsum)
+ *cfcsum = ntohl(caps_cmd->cfcsum);
+
+ /* Validate device capabilities */
+ if (csio_hw_validate_caps(hw, mbp)) {
+ rv = CSIO_NOSUPP;
+ goto out;
+ }
+
+ /*
+ * And now tell the firmware to use the configuration we just loaded.
+ */
+ caps_cmd->op_to_write =
+ htonl(FW_CMD_OP(FW_CAPS_CONFIG_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_WRITE);
+ caps_cmd->cfvalid_to_len16 = htonl(FW_LEN16(*caps_cmd));
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FW_CAPS_CONFIG_CMD failed!\n");
+ goto out;
+ }
+
+ ret = csio_mb_fw_retval(mbp);
+ if (ret != FW_SUCCESS) {
+ csio_dbg(hw, "FW_CAPS_CONFIG_CMD returned %d!\n", rv);
+ goto out;
+ }
+
+ rv = CSIO_SUCCESS;
+out:
+ mempool_free(mbp, hw->mb_mempool);
+ return rv;
+}
+
+/*
+ * csio_get_device_params - Get device parameters.
+ * @hw: HW module
+ *
+ */
+static csio_retval_t
+csio_get_device_params(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+ u32 param[6];
+ int i, j = 0;
+
+ /* Initialize portids to -1 */
+ for (i = 0; i < CSIO_MAX_PPORTS; i++)
+ hw->pport[i].portid = -1;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /* Get port vec information. */
+ param[0] = FW_PARAM_DEV(PORTVEC);
+
+ /* Get Core clock. */
+ param[1] = FW_PARAM_DEV(CCLK);
+
+ /* Get EQ id start and end. */
+ param[2] = FW_PARAM_PFVF(EQ_START);
+ param[3] = FW_PARAM_PFVF(EQ_END);
+
+ /* Get IQ id start and end. */
+ param[4] = FW_PARAM_PFVF(IQFLINT_START);
+ param[5] = FW_PARAM_PFVF(IQFLINT_END);
+
+ csio_mb_params(hw, mbp, CSIO_MB_DEFAULT_TMO, hw->pfn, 0,
+ ARRAY_SIZE(param), param, NULL, CSIO_FALSE, NULL);
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FW_PARAMS_CMD(read) failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ csio_mb_process_read_params_rsp(hw, mbp, &retval,
+ ARRAY_SIZE(param), param);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_PARAMS_CMD(read) failed with ret:0x%x!\n",
+ retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ /* cache the information. */
+ hw->port_vec = param[0];
+ hw->vpd.cclk = param[1];
+ wrm->fw_eq_start = param[2];
+ wrm->fw_iq_start = param[4];
+
+ /* Using FW configured max iqs & eqs */
+ if ((hw->flags & CSIO_HWF_USING_SOFT_PARAMS) ||
+ !csio_is_hw_master(hw)) {
+ hw->cfg_niq = param[5] - param[4] + 1;
+ hw->cfg_neq = param[3] - param[2] + 1;
+ csio_dbg(hw, "Using fwconfig max niqs %d neqs %d\n",
+ hw->cfg_niq, hw->cfg_neq);
+ }
+
+ hw->port_vec &= csio_port_mask;
+
+ hw->num_pports = hweight32(hw->port_vec);
+
+ csio_dbg(hw, "Port vector: 0x%x, #ports: %d\n",
+ hw->port_vec, hw->num_pports);
+
+ for (i = 0; i < hw->num_pports; i++) {
+ while ((hw->port_vec & (1 << j)) == 0)
+ j++;
+ hw->pport[i].portid = j++;
+ csio_dbg(hw, "Found Port:%d\n", hw->pport[i].portid);
+ }
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+
+/*
+ * csio_config_device_caps - Get and set device capabilities.
+ * @hw: HW module
+ *
+ */
+static csio_retval_t
+csio_config_device_caps(struct csio_hw *hw)
+{
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+ csio_retval_t rv = CSIO_INVAL;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /* Get device capabilities */
+ csio_mb_caps_config(hw, mbp, CSIO_MB_DEFAULT_TMO, 0, 0, 0, 0, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FW_CAPS_CONFIG_CMD(r) failed!\n");
+ goto out;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_CAPS_CONFIG_CMD(r) returned %d!\n", retval);
+ goto out;
+ }
+
+ /* Validate device capabilities */
+ if (csio_hw_validate_caps(hw, mbp))
+ goto out;
+
+ /* Don't config device capabilities if already configured */
+ if (hw->fw_state == CSIO_DEV_STATE_INIT) {
+ rv = CSIO_SUCCESS;
+ goto out;
+ }
+
+ /* Write back desired device capabilities */
+ csio_mb_caps_config(hw, mbp, CSIO_MB_DEFAULT_TMO, CSIO_TRUE, CSIO_TRUE,
+ CSIO_FALSE, CSIO_TRUE, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FW_CAPS_CONFIG_CMD(w) failed!\n");
+ goto out;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_CAPS_CONFIG_CMD(w) returned %d!\n", retval);
+ goto out;
+ }
+
+ rv = CSIO_SUCCESS;
+out:
+ mempool_free(mbp, hw->mb_mempool);
+ return rv;
+}
+
+static csio_retval_t
+csio_config_global_rss(struct csio_hw *hw)
+{
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ csio_rss_glb_config(hw, mbp, CSIO_MB_DEFAULT_TMO,
+ FW_RSS_GLB_CONFIG_CMD_MODE_BASICVIRTUAL,
+ FW_RSS_GLB_CONFIG_CMD_TNLMAPEN |
+ FW_RSS_GLB_CONFIG_CMD_HASHTOEPLITZ |
+ FW_RSS_GLB_CONFIG_CMD_TNLALLLKP,
+ NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FW_RSS_GLB_CONFIG_CMD failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_RSS_GLB_CONFIG_CMD returned 0x%x!\n", retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_config_pfvf - Configure Physical/Virtual functions settings.
+ * @hw: HW module
+ *
+ */
+static csio_retval_t
+csio_config_pfvf(struct csio_hw *hw)
+{
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /*
+ * For now, allow all PFs to access to all ports using a pmask
+ * value of 0xF (M_FW_PFVF_CMD_PMASK). Once we have VFs, we will
+ * need to provide access based on some rule.
+ */
+ csio_mb_pfvf(hw, mbp, CSIO_MB_DEFAULT_TMO, hw->pfn, 0, CSIO_NEQ,
+ CSIO_NETH_CTRL, CSIO_NIQ_FLINT, 0, 0, CSIO_NVI, CSIO_CMASK,
+ CSIO_PMASK, CSIO_NEXACTF, CSIO_R_CAPS, CSIO_WX_CAPS, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FW_PFVF_CMD failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_PFVF_CMD returned 0x%x!\n", retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_enable_ports - Bring up all available ports.
+ * @hw: HW module.
+ *
+ */
+static csio_retval_t
+csio_enable_ports(struct csio_hw *hw)
+{
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+ uint8_t portid;
+ int i;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ for (i = 0; i < hw->num_pports; i++) {
+ portid = hw->pport[i].portid;
+
+ /* Read PORT information */
+ csio_mb_port(hw, mbp, CSIO_MB_DEFAULT_TMO, portid,
+ CSIO_FALSE, 0, 0, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "failed to issue FW_PORT_CMD(r) port:%d\n",
+ portid);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ csio_mb_process_read_port_rsp(hw, mbp, &retval,
+ &hw->pport[i].pcap);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_PORT_CMD(r) port:%d failed: 0x%x\n",
+ portid, retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ /* Write back PORT information */
+ csio_mb_port(hw, mbp, CSIO_MB_DEFAULT_TMO, portid,
+ CSIO_TRUE, (PAUSE_RX | PAUSE_TX),
+ hw->pport[i].pcap, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "failed to issue FW_PORT_CMD(w) port:%d\n",
+ portid);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_PORT_CMD(w) port:%d failed :0x%x\n",
+ portid, retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ } /* For all ports */
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_get_fcoe_resinfo - Read fcoe fw resource info.
+ * @hw: HW module
+ * Issued with lock held.
+ */
+static csio_retval_t
+csio_get_fcoe_resinfo(struct csio_hw *hw)
+{
+ struct csio_fcoe_res_info *res_info = &hw->fres_info;
+ struct fw_fcoe_res_info_cmd *rsp;
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /* Get FCoE FW resource information */
+ csio_fcoe_read_res_info_init_mb(hw, mbp, CSIO_MB_DEFAULT_TMO, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "failed to issue FW_FCOE_RES_INFO_CMD\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ rsp = (struct fw_fcoe_res_info_cmd *)(mbp->mb);
+ retval = FW_CMD_RETVAL_GET(ntohl(rsp->retval_len16));
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_FCOE_RES_INFO_CMD failed with ret x%x\n",
+ retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ res_info->e_d_tov = ntohs(rsp->e_d_tov);
+ res_info->r_a_tov_seq = ntohs(rsp->r_a_tov_seq);
+ res_info->r_a_tov_els = ntohs(rsp->r_a_tov_els);
+ res_info->r_r_tov = ntohs(rsp->r_r_tov);
+ res_info->max_xchgs = ntohl(rsp->max_xchgs);
+ res_info->max_ssns = ntohl(rsp->max_ssns);
+ res_info->used_xchgs = ntohl(rsp->used_xchgs);
+ res_info->used_ssns = ntohl(rsp->used_ssns);
+ res_info->max_fcfs = ntohl(rsp->max_fcfs);
+ res_info->max_vnps = ntohl(rsp->max_vnps);
+ res_info->used_fcfs = ntohl(rsp->used_fcfs);
+ res_info->used_vnps = ntohl(rsp->used_vnps);
+
+ csio_dbg(hw, "max ssns:%d max xchgs:%d\n", res_info->max_ssns,
+ res_info->max_xchgs);
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+static csio_retval_t
+csio_hw_check_fwconfig(struct csio_hw *hw, u32 *param)
+{
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+ u32 _param[1];
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /*
+ * Find out whether we're dealing with a version of
+ * the firmware which has configuration file support.
+ */
+ _param[0] = (FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) |
+ FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_CF));
+
+ csio_mb_params(hw, mbp, CSIO_MB_DEFAULT_TMO, hw->pfn, 0,
+ ARRAY_SIZE(_param), _param, NULL, CSIO_FALSE, NULL);
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FW_PARAMS_CMD(read) failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ csio_mb_process_read_params_rsp(hw, mbp, &retval,
+ ARRAY_SIZE(_param), _param);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_PARAMS_CMD(read) failed with ret:0x%x!\n",
+ retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+ *param = _param[0];
+
+ return CSIO_SUCCESS;
+}
+
+static csio_retval_t
+csio_hw_flash_config(struct csio_hw *hw, u32 *fw_cfg_param, char *path)
+{
+ csio_retval_t ret = CSIO_SUCCESS;
+ const struct firmware *cf;
+ struct pci_dev *pci_dev = hw->pdev;
+ struct device *dev = &pci_dev->dev;
+
+ unsigned int mtype = 0, maddr = 0;
+ uint32_t *cfg_data;
+ int value_to_add = 0;
+
+ if (request_firmware(&cf, CSIO_CF_FNAME, dev) < 0) {
+ csio_err(hw, "could not find config file " CSIO_CF_FNAME
+ ",err: %d\n", ret);
+ return CSIO_NOSUPP;
+ }
+
+ if (cf->size%4 != 0)
+ value_to_add = 4 - (cf->size % 4);
+
+ cfg_data = kzalloc(cf->size+value_to_add, GFP_KERNEL);
+ if (cfg_data == NULL)
+ return CSIO_NOMEM;
+
+ memcpy((void *)cfg_data, (const void *)cf->data, cf->size);
+
+ if (csio_hw_check_fwconfig(hw, fw_cfg_param) != CSIO_SUCCESS)
+ return CSIO_INVAL;
+
+ mtype = FW_PARAMS_PARAM_Y_GET(*fw_cfg_param);
+ maddr = FW_PARAMS_PARAM_Z_GET(*fw_cfg_param) << 16;
+
+ ret = csio_memory_write(hw, mtype, maddr,
+ cf->size + value_to_add, cfg_data);
+ if (ret == CSIO_SUCCESS) {
+ csio_info(hw, "config file upgraded to " CSIO_CF_FNAME "\n");
+ strncpy(path, "/lib/firmware/" CSIO_CF_FNAME, 64);
+ }
+
+ kfree(cfg_data);
+ release_firmware(cf);
+
+ return ret;
+}
+
+/*
+ * HW initialization: contact FW, obtain config, perform basic init.
+ *
+ * If the firmware we're dealing with has Configuration File support, then
+ * we use that to perform all configuration -- either using the configuration
+ * file stored in flash on the adapter or using a filesystem-local file
+ * if available.
+ *
+ * If we don't have configuration file support in the firmware, then we'll
+ * have to set things up the old fashioned way with hard-coded register
+ * writes and firmware commands ...
+ */
+
+/*
+ * Attempt to initialize the HW via a Firmware Configuration File.
+ */
+static csio_retval_t
+csio_hw_use_fwconfig(struct csio_hw *hw, int reset, u32 *fw_cfg_param)
+{
+ unsigned int mtype, maddr;
+ csio_retval_t rv;
+ uint32_t finiver, finicsum, cfcsum;
+ int using_flash;
+ char path[64];
+
+ /*
+ * Reset device if necessary
+ */
+ if (reset) {
+ rv = csio_do_reset(hw, CSIO_TRUE);
+ if (rv != CSIO_SUCCESS)
+ goto bye;
+ }
+
+ /*
+ * If we have a configuration file in host ,
+ * then use that. Otherwise, use the configuration file stored
+ * in the HW flash ...
+ */
+ spin_unlock_irq(&hw->lock);
+ rv = csio_hw_flash_config(hw, fw_cfg_param, path);
+ spin_lock_irq(&hw->lock);
+ if (rv != CSIO_SUCCESS) {
+ if (rv == CSIO_NOSUPP) {
+ /*
+ * config file was not found. Use default
+ * config file from flash.
+ */
+ mtype = FW_MEMTYPE_CF_FLASH;
+ maddr = csio_hw_flash_cfg_addr(hw);
+ using_flash = 1;
+ } else {
+ /*
+ * we revert back to the hardwired config if
+ * flashing failed.
+ */
+ goto bye;
+ }
+ } else {
+ mtype = FW_PARAMS_PARAM_Y_GET(*fw_cfg_param);
+ maddr = FW_PARAMS_PARAM_Z_GET(*fw_cfg_param) << 16;
+ using_flash = 0;
+ }
+
+ hw->cfg_store = (uint8_t)mtype;
+
+ /*
+ * Issue a Capability Configuration command to the firmware to get it
+ * to parse the Configuration File.
+ */
+ rv = csio_hw_fw_config_file(hw, mtype, maddr, &finiver,
+ &finicsum, &cfcsum);
+ if (rv != CSIO_SUCCESS)
+ goto bye;
+
+ hw->cfg_finiver = finiver;
+ hw->cfg_finicsum = finicsum;
+ hw->cfg_cfcsum = cfcsum;
+ hw->cfg_csum_status = CSIO_TRUE;
+
+ if (finicsum != cfcsum) {
+ csio_warn(hw,
+ "Config File checksum mismatch: csum=%#x, computed=%#x\n",
+ finicsum, cfcsum);
+
+ hw->cfg_csum_status = CSIO_FALSE;
+ }
+
+ /*
+ * Note that we're operating with parameters
+ * not supplied by the driver, rather than from hard-wired
+ * initialization constants buried in the driver.
+ */
+ hw->flags |= CSIO_HWF_USING_SOFT_PARAMS;
+
+ /* device parameters */
+ rv = csio_get_device_params(hw);
+ if (rv != CSIO_SUCCESS)
+ goto bye;
+
+ /* Configure SGE */
+ csio_wr_sge_init(hw);
+
+ /*
+ * And finally tell the firmware to initialize itself using the
+ * parameters from the Configuration File.
+ */
+ /* Post event to notify completion of configuration */
+ csio_post_event(&hw->sm, CSIO_HWE_INIT);
+
+ csio_info(hw,
+ "Firmware Configuration File %s, version %#x, computed checksum %#x\n",
+ (using_flash ? "in device FLASH" : path), finiver, cfcsum);
+
+ return 0;
+
+ /*
+ * Something bad happened. Return the error ...
+ */
+bye:
+ hw->flags &= ~CSIO_HWF_USING_SOFT_PARAMS;
+ csio_dbg(hw, "Configuration file error %d\n", rv);
+ return rv;
+}
+
+/*
+ * Attempt to initialize the adapter via hard-coded, driver supplied
+ * parameters ...
+ */
+static int
+csio_hw_no_fwconfig(struct csio_hw *hw, int reset)
+{
+ csio_retval_t rv;
+ /*
+ * Reset device if necessary
+ */
+ if (reset) {
+ rv = csio_do_reset(hw, CSIO_TRUE);
+ if (rv != CSIO_SUCCESS)
+ goto out;
+ }
+
+ /* Get and set device capabilities */
+ rv = csio_config_device_caps(hw);
+ if (rv != CSIO_SUCCESS)
+ goto out;
+
+ /* Config Global RSS command */
+ rv = csio_config_global_rss(hw);
+ if (rv != CSIO_SUCCESS)
+ goto out;
+
+ /* Configure PF/VF capabilities of device */
+ rv = csio_config_pfvf(hw);
+ if (rv != CSIO_SUCCESS)
+ goto out;
+
+ /* device parameters */
+ rv = csio_get_device_params(hw);
+ if (rv != CSIO_SUCCESS)
+ goto out;
+
+ /* Configure SGE */
+ csio_wr_sge_init(hw);
+
+ /* Post event to notify completion of configuration */
+ csio_post_event(&hw->sm, CSIO_HWE_INIT);
+
+out:
+ return rv;
+}
+
+/*
+ * Returns CSIO_INVAL if attempts to flash the firmware failed
+ * else returns CSIO_SUCCESS,
+ * if flashing was not attempted because the card had the
+ * latest firmware CSIO_CANCELLED is returned
+ */
+static int
+csio_hw_flash_fw(struct csio_hw *hw)
+{
+ int ret = CSIO_CANCELLED;
+ const struct firmware *fw;
+ const struct fw_hdr *hdr;
+ u32 fw_ver;
+ struct pci_dev *pci_dev = hw->pdev;
+ struct device *dev = &pci_dev->dev ;
+
+ if (request_firmware(&fw, CSIO_FW_FNAME, dev) < 0) {
+ csio_err(hw, "could not find firmware image " CSIO_FW_FNAME
+ ",err: %d\n", ret);
+ return CSIO_INVAL;
+ }
+
+ hdr = (const struct fw_hdr *)fw->data;
+ fw_ver = ntohl(hdr->fw_ver);
+ if (FW_HDR_FW_VER_MAJOR_GET(fw_ver) != FW_VERSION_MAJOR)
+ return CSIO_INVAL; /* wrong major version, won't do */
+
+ /*
+ * If the flash FW is unusable or we found something newer, load it.
+ */
+ if (FW_HDR_FW_VER_MAJOR_GET(hw->fwrev) != FW_VERSION_MAJOR ||
+ fw_ver > hw->fwrev) {
+ ret = csio_hw_fw_upgrade(hw, hw->pfn, fw->data, fw->size,
+ /*force=*/false);
+ if (!ret)
+ csio_info(hw, "firmware upgraded to version %pI4 from "
+ CSIO_FW_FNAME "\n", &hdr->fw_ver);
+ else
+ csio_err(hw, "firmware upgrade failed! err=%d\n", -ret);
+ }
+
+ release_firmware(fw);
+ return ret;
+}
+
+
+/*
+ * csio_hw_configure - Configure HW
+ * @hw - HW module
+ *
+ */
+static void
+csio_hw_configure(struct csio_hw *hw)
+{
+ int reset = 1;
+ csio_retval_t rv;
+ u32 param[1];
+
+ rv = csio_hw_dev_ready(hw);
+ if (rv != CSIO_SUCCESS) {
+ csio_inc_stats(hw, n_err_fatal);
+ csio_post_event(&hw->sm, CSIO_HWE_FATAL);
+ goto out;
+ }
+
+ /* HW version */
+ hw->chip_ver = (char)csio_rd_reg32(hw, PL_REV);
+
+ /* Needed for FW download */
+ rv = csio_hw_get_flash_params(hw);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, "Failed to get serial flash params rv:%d\n", rv);
+ csio_post_event(&hw->sm, CSIO_HWE_FATAL);
+ goto out;
+ }
+
+ /* Set pci completion timeout value to 4 seconds. */
+ csio_set_pcie_completion_timeout(hw, 0xd);
+
+ csio_hw_set_mem_win(hw);
+
+ rv = csio_hw_get_fw_version(hw, &hw->fwrev);
+ if (rv != CSIO_SUCCESS)
+ goto out;
+
+ csio_hw_print_fw_version(hw, "Firmware revision");
+
+ rv = csio_do_hello(hw, &hw->fw_state);
+ if (rv != CSIO_SUCCESS) {
+ csio_inc_stats(hw, n_err_fatal);
+ csio_post_event(&hw->sm, CSIO_HWE_FATAL);
+ goto out;
+ }
+
+ /* Read vpd */
+ rv = csio_hw_get_vpd_params(hw, &hw->vpd);
+ if (rv != CSIO_SUCCESS)
+ goto out;
+
+ if (csio_is_hw_master(hw) && hw->fw_state != CSIO_DEV_STATE_INIT) {
+ rv = csio_hw_check_fw_version(hw);
+ if (rv == CSIO_INVAL) {
+
+ /* Do firmware update */
+ spin_unlock_irq(&hw->lock);
+ rv = csio_hw_flash_fw(hw);
+ spin_lock_irq(&hw->lock);
+
+ if (rv == CSIO_SUCCESS) {
+ reset = 0;
+ /*
+ * Note that the chip was reset as part of the
+ * firmware upgrade so we don't reset it again
+ * below and grab the new firmware version.
+ */
+ rv = csio_hw_check_fw_version(hw);
+ }
+ }
+ /*
+ * If the firmware doesn't support Configuration
+ * Files, use the old Driver-based, hard-wired
+ * initialization. Otherwise, try using the
+ * Configuration File support and fall back to the
+ * Driver-based initialization if there's no
+ * Configuration File found.
+ */
+ if (csio_hw_check_fwconfig(hw, param) == CSIO_SUCCESS) {
+ rv = csio_hw_use_fwconfig(hw, reset, param);
+ if (rv == CSIO_NOSUPP)
+ goto out;
+ if (rv != CSIO_SUCCESS) {
+ csio_info(hw,
+ "No Configuration File present "
+ "on adapter. Using hard-wired "
+ "configuration parameters.\n");
+ rv = csio_hw_no_fwconfig(hw, reset);
+ }
+ } else {
+ rv = csio_hw_no_fwconfig(hw, reset);
+ }
+
+ if (rv != CSIO_SUCCESS)
+ goto out;
+
+ } else {
+ if (hw->fw_state == CSIO_DEV_STATE_INIT) {
+
+ /* device parameters */
+ rv = csio_get_device_params(hw);
+ if (rv != CSIO_SUCCESS)
+ goto out;
+
+ /* Get device capabilities */
+ rv = csio_config_device_caps(hw);
+ if (rv != CSIO_SUCCESS)
+ goto out;
+
+ /* Configure SGE */
+ csio_wr_sge_init(hw);
+
+ /* Post event to notify completion of configuration */
+ csio_post_event(&hw->sm, CSIO_HWE_INIT);
+ goto out;
+ }
+ } /* if not master */
+
+out:
+ return;
+}
+
+/*
+ * csio_hw_initialize - Initialize HW
+ * @hw - HW module
+ *
+ */
+static void
+csio_hw_initialize(struct csio_hw *hw)
+{
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+ csio_retval_t rv;
+ int i;
+
+ if (csio_is_hw_master(hw) && hw->fw_state != CSIO_DEV_STATE_INIT) {
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp)
+ goto out;
+
+ csio_mb_initialize(hw, mbp, CSIO_MB_DEFAULT_TMO, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FW_INITIALIZE_CMD failed!\n");
+ goto free_and_out;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FW_INITIALIZE_CMD returned 0x%x!\n",
+ retval);
+ goto free_and_out;
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+ }
+
+ rv = csio_get_fcoe_resinfo(hw);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, "Failed to read fcoe resource info: %d\n", rv);
+ goto out;
+ }
+
+ spin_unlock_irq(&hw->lock);
+ rv = csio_config_queues(hw);
+ spin_lock_irq(&hw->lock);
+
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, "Config of queues failed!: %d\n", rv);
+ goto out;
+ }
+
+ for (i = 0; i < hw->num_pports; i++)
+ hw->pport[i].mod_type = FW_PORT_MOD_TYPE_NA;
+
+ if (csio_is_hw_master(hw) && hw->fw_state != CSIO_DEV_STATE_INIT) {
+ rv = csio_enable_ports(hw);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, "Failed to enable ports: %d\n", rv);
+ goto out;
+ }
+ }
+
+ csio_post_event(&hw->sm, CSIO_HWE_INIT_DONE);
+ return;
+
+free_and_out:
+ mempool_free(mbp, hw->mb_mempool);
+out:
+ return;
+}
+
+#define PF_INTR_MASK (PFSW | PFCIM)
+
+/*
+ * csio_hw_intr_enable - Enable HW interrupts
+ * @hw: Pointer to HW module.
+ *
+ * Enable interrupts in HW registers.
+ */
+static void
+csio_hw_intr_enable(struct csio_hw *hw)
+{
+ uint16_t vec = (uint16_t)csio_get_mb_intr_idx(csio_hw_to_mbm(hw));
+ uint32_t pf = SOURCEPF_GET(csio_rd_reg32(hw, PL_WHOAMI));
+ uint32_t pl = csio_rd_reg32(hw, PL_INT_ENABLE);
+
+ /*
+ * Set aivec for MSI/MSIX. PCIE_PF_CFG.INTXType is set up
+ * by FW, so do nothing for INTX.
+ */
+ if (hw->intr_mode == CSIO_IM_MSIX)
+ csio_set_reg_field(hw, MYPF_REG(PCIE_PF_CFG),
+ AIVEC(AIVEC_MASK), vec);
+ else if (hw->intr_mode == CSIO_IM_MSI)
+ csio_set_reg_field(hw, MYPF_REG(PCIE_PF_CFG),
+ AIVEC(AIVEC_MASK), 0);
+
+ csio_wr_reg32(hw, PF_INTR_MASK, MYPF_REG(PL_PF_INT_ENABLE));
+
+ /* Turn on MB interrupts - this will internally flush PIO as well */
+ csio_mb_intr_enable(hw);
+
+ /* These are common registers - only a master can modify them */
+ if (csio_is_hw_master(hw)) {
+ /*
+ * Disable the Serial FLASH interrupt, if enabled!
+ */
+ pl &= (~SF);
+ csio_wr_reg32(hw, pl, PL_INT_ENABLE);
+
+ csio_wr_reg32(hw, ERR_CPL_EXCEED_IQE_SIZE |
+ EGRESS_SIZE_ERR | ERR_INVALID_CIDX_INC |
+ ERR_CPL_OPCODE_0 | ERR_DROPPED_DB |
+ ERR_DATA_CPL_ON_HIGH_QID1 |
+ ERR_DATA_CPL_ON_HIGH_QID0 | ERR_BAD_DB_PIDX3 |
+ ERR_BAD_DB_PIDX2 | ERR_BAD_DB_PIDX1 |
+ ERR_BAD_DB_PIDX0 | ERR_ING_CTXT_PRIO |
+ ERR_EGR_CTXT_PRIO | INGRESS_SIZE_ERR,
+ SGE_INT_ENABLE3);
+ csio_set_reg_field(hw, PL_INT_MAP0, 0, 1 << pf);
+ }
+
+ hw->flags |= CSIO_HWF_HW_INTR_ENABLED;
+
+}
+
+/*
+ * csio_hw_intr_disable - Disable HW interrupts
+ * @hw: Pointer to HW module.
+ *
+ * Turn off Mailbox and PCI_PF_CFG interrupts.
+ */
+void
+csio_hw_intr_disable(struct csio_hw *hw)
+{
+ uint32_t pf = SOURCEPF_GET(csio_rd_reg32(hw, PL_WHOAMI));
+
+ if (!(hw->flags & CSIO_HWF_HW_INTR_ENABLED))
+ return;
+
+ hw->flags &= ~CSIO_HWF_HW_INTR_ENABLED;
+
+ csio_wr_reg32(hw, 0, MYPF_REG(PL_PF_INT_ENABLE));
+ if (csio_is_hw_master(hw))
+ csio_set_reg_field(hw, PL_INT_MAP0, 1 << pf, 0);
+
+ /* Turn off MB interrupts */
+ csio_mb_intr_disable(hw);
+
+}
+
+static void
+csio_hw_fatal_err(struct csio_hw *hw)
+{
+ csio_set_reg_field(hw, SGE_CONTROL, GLOBALENABLE, 0);
+ csio_hw_intr_disable(hw);
+
+ /* Do not reset HW, we may need FW state for debugging */
+ csio_fatal(hw, "HW Fatal error encountered!\n");
+}
+
+/*****************************************************************************/
+/* START: HW SM */
+/*****************************************************************************/
+/*
+ * csio_hws_uninit - Uninit state
+ * @hw - HW module
+ * @evt - Event
+ *
+ */
+static void
+csio_hws_uninit(struct csio_hw *hw, enum csio_hw_ev evt)
+{
+ hw->prev_evt = hw->cur_evt;
+ hw->cur_evt = evt;
+ csio_inc_stats(hw, n_evt_sm[evt]);
+
+ switch (evt) {
+
+ case CSIO_HWE_CFG:
+ csio_set_state(&hw->sm, csio_hws_configuring);
+ csio_hw_configure(hw);
+ break;
+
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_hws_configuring - Configuring state
+ * @hw - HW module
+ * @evt - Event
+ *
+ */
+static void
+csio_hws_configuring(struct csio_hw *hw, enum csio_hw_ev evt)
+{
+ hw->prev_evt = hw->cur_evt;
+ hw->cur_evt = evt;
+ csio_inc_stats(hw, n_evt_sm[evt]);
+
+ switch (evt) {
+
+ case CSIO_HWE_INIT:
+ csio_set_state(&hw->sm, csio_hws_initializing);
+ csio_hw_initialize(hw);
+ break;
+
+ case CSIO_HWE_INIT_DONE:
+ csio_set_state(&hw->sm, csio_hws_ready);
+ /* Fan out event to all lnode SMs */
+ csio_notify_lnodes(hw, CSIO_LN_NOTIFY_HWREADY);
+ break;
+
+ case CSIO_HWE_FATAL:
+ csio_set_state(&hw->sm, csio_hws_uninit);
+ break;
+
+ case CSIO_HWE_PCI_REMOVE:
+ csio_do_bye(hw);
+ break;
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_hws_initializing - Initialiazing state
+ * @hw - HW module
+ * @evt - Event
+ *
+ */
+static void
+csio_hws_initializing(struct csio_hw *hw, enum csio_hw_ev evt)
+{
+ hw->prev_evt = hw->cur_evt;
+ hw->cur_evt = evt;
+ csio_inc_stats(hw, n_evt_sm[evt]);
+
+ switch (evt) {
+
+ case CSIO_HWE_INIT_DONE:
+ csio_set_state(&hw->sm, csio_hws_ready);
+
+ /* Fan out event to all lnode SMs */
+ csio_notify_lnodes(hw, CSIO_LN_NOTIFY_HWREADY);
+
+ /* Enable interrupts */
+ csio_hw_intr_enable(hw);
+ break;
+
+ case CSIO_HWE_FATAL:
+ csio_set_state(&hw->sm, csio_hws_uninit);
+ break;
+
+ case CSIO_HWE_PCI_REMOVE:
+ csio_do_bye(hw);
+ break;
+
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_hws_ready - Ready state
+ * @hw - HW module
+ * @evt - Event
+ *
+ */
+static void
+csio_hws_ready(struct csio_hw *hw, enum csio_hw_ev evt)
+{
+ /* Remember the event */
+ hw->evtflag = evt;
+
+ hw->prev_evt = hw->cur_evt;
+ hw->cur_evt = evt;
+ csio_inc_stats(hw, n_evt_sm[evt]);
+
+ switch (evt) {
+
+ case CSIO_HWE_HBA_RESET:
+ case CSIO_HWE_FW_DLOAD:
+ case CSIO_HWE_SUSPEND:
+ case CSIO_HWE_PCI_REMOVE:
+ case CSIO_HWE_PCIERR_DETECTED:
+ csio_set_state(&hw->sm, csio_hws_quiescing);
+ /* cleanup all outstanding cmds */
+ if (evt == CSIO_HWE_HBA_RESET ||
+ evt == CSIO_HWE_PCIERR_DETECTED)
+ csio_scsim_cleanup_io(csio_hw_to_scsim(hw), CSIO_FALSE);
+ else
+ csio_scsim_cleanup_io(csio_hw_to_scsim(hw), CSIO_TRUE);
+
+ csio_hw_intr_disable(hw);
+ csio_hw_mbm_cleanup(hw);
+ csio_evtq_stop(hw);
+ csio_notify_lnodes(hw, CSIO_LN_NOTIFY_HWSTOP);
+ csio_evtq_flush(hw);
+ csio_mgmtm_cleanup(csio_hw_to_mgmtm(hw));
+ csio_post_event(&hw->sm, CSIO_HWE_QUIESCED);
+ break;
+
+ case CSIO_HWE_FATAL:
+ csio_set_state(&hw->sm, csio_hws_uninit);
+ break;
+
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_hws_quiescing - Quiescing state
+ * @hw - HW module
+ * @evt - Event
+ *
+ */
+static void
+csio_hws_quiescing(struct csio_hw *hw, enum csio_hw_ev evt)
+{
+ hw->prev_evt = hw->cur_evt;
+ hw->cur_evt = evt;
+ csio_inc_stats(hw, n_evt_sm[evt]);
+
+ switch (evt) {
+
+ case CSIO_HWE_QUIESCED:
+ switch (hw->evtflag) {
+
+ case CSIO_HWE_FW_DLOAD:
+ csio_set_state(&hw->sm, csio_hws_resetting);
+ /* Download firmware */
+ /* Fall through */
+
+ case CSIO_HWE_HBA_RESET:
+ csio_set_state(&hw->sm, csio_hws_resetting);
+ /* Start reset of the HBA */
+ csio_notify_lnodes(hw, CSIO_LN_NOTIFY_HWRESET);
+ csio_wr_destroy_queues(hw, CSIO_FALSE);
+ csio_do_reset(hw, CSIO_FALSE);
+ csio_post_event(&hw->sm, CSIO_HWE_HBA_RESET_DONE);
+ break;
+
+ case CSIO_HWE_PCI_REMOVE:
+ csio_set_state(&hw->sm, csio_hws_removing);
+ csio_notify_lnodes(hw, CSIO_LN_NOTIFY_HWREMOVE);
+ csio_wr_destroy_queues(hw, CSIO_TRUE);
+ /* Now send the bye command */
+ csio_do_bye(hw);
+ break;
+
+ case CSIO_HWE_SUSPEND:
+ csio_set_state(&hw->sm, csio_hws_quiesced);
+ break;
+
+ case CSIO_HWE_PCIERR_DETECTED:
+ csio_set_state(&hw->sm, csio_hws_pcierr);
+ csio_wr_destroy_queues(hw, CSIO_FALSE);
+ break;
+
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+
+ }
+ break;
+
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_hws_quiesced - Quiesced state
+ * @hw - HW module
+ * @evt - Event
+ *
+ */
+static void
+csio_hws_quiesced(struct csio_hw *hw, enum csio_hw_ev evt)
+{
+ hw->prev_evt = hw->cur_evt;
+ hw->cur_evt = evt;
+ csio_inc_stats(hw, n_evt_sm[evt]);
+
+ switch (evt) {
+
+ case CSIO_HWE_RESUME:
+ csio_set_state(&hw->sm, csio_hws_configuring);
+ csio_hw_configure(hw);
+ break;
+
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_hws_resetting - HW Resetting state
+ * @hw - HW module
+ * @evt - Event
+ *
+ */
+static void
+csio_hws_resetting(struct csio_hw *hw, enum csio_hw_ev evt)
+{
+ hw->prev_evt = hw->cur_evt;
+ hw->cur_evt = evt;
+ csio_inc_stats(hw, n_evt_sm[evt]);
+
+ switch (evt) {
+
+ case CSIO_HWE_HBA_RESET_DONE:
+ csio_evtq_start(hw);
+ csio_set_state(&hw->sm, csio_hws_configuring);
+ csio_hw_configure(hw);
+ break;
+
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_hws_removing - PCI Hotplug removing state
+ * @hw - HW module
+ * @evt - Event
+ *
+ */
+static void
+csio_hws_removing(struct csio_hw *hw, enum csio_hw_ev evt)
+{
+ hw->prev_evt = hw->cur_evt;
+ hw->cur_evt = evt;
+ csio_inc_stats(hw, n_evt_sm[evt]);
+
+ switch (evt) {
+ case CSIO_HWE_HBA_RESET:
+ if (!csio_is_hw_master(hw))
+ break;
+ /*
+ * The BYE should have alerady been issued, so we cant
+ * use the mailbox interface. Hence we use the PL_RST
+ * register directly.
+ */
+ csio_err(hw, "Resetting HW and waiting 2 seconds...\n");
+ csio_wr_reg32(hw, PIORSTMODE | PIORST, PL_RST);
+ mdelay(2000);
+ break;
+
+ /* Should never receive any new events */
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+
+ }
+}
+
+/*
+ * csio_hws_pcierr - PCI Error state
+ * @hw - HW module
+ * @evt - Event
+ *
+ */
+static void
+csio_hws_pcierr(struct csio_hw *hw, enum csio_hw_ev evt)
+{
+ hw->prev_evt = hw->cur_evt;
+ hw->cur_evt = evt;
+ csio_inc_stats(hw, n_evt_sm[evt]);
+
+ switch (evt) {
+
+ case CSIO_HWE_PCIERR_SLOT_RESET:
+ csio_evtq_start(hw);
+ csio_set_state(&hw->sm, csio_hws_configuring);
+ csio_hw_configure(hw);
+ break;
+
+ default:
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+ }
+}
+
+/*****************************************************************************/
+/* END: HW SM */
+/*****************************************************************************/
+
+/* Slow path handlers */
+struct intr_info {
+ unsigned int mask; /* bits to check in interrupt status */
+ const char *msg; /* message to print or NULL */
+ short stat_idx; /* stat counter to increment or -1 */
+ unsigned short fatal; /* whether the condition reported is fatal */
+};
+
+/*
+ * csio_handle_intr_status - table driven interrupt handler
+ * @hw: HW instance
+ * @reg: the interrupt status register to process
+ * @acts: table of interrupt actions
+ *
+ * A table driven interrupt handler that applies a set of masks to an
+ * interrupt status word and performs the corresponding actions if the
+ * interrupts described by the mask have occured. The actions include
+ * optionally emitting a warning or alert message. The table is terminated
+ * by an entry specifying mask 0. Returns the number of fatal interrupt
+ * conditions.
+ */
+static int
+csio_handle_intr_status(struct csio_hw *hw, unsigned int reg,
+ const struct intr_info *acts)
+{
+ int fatal = 0;
+ unsigned int mask = 0;
+ unsigned int status = csio_rd_reg32(hw, reg);
+
+ for ( ; acts->mask; ++acts) {
+ if (!(status & acts->mask))
+ continue;
+ if (acts->fatal) {
+ fatal++;
+ csio_fatal(hw, "Fatal %s (0x%x)\n",
+ acts->msg, status & acts->mask);
+ } else if (acts->msg)
+ csio_info(hw, "%s (0x%x)\n",
+ acts->msg, status & acts->mask);
+ mask |= acts->mask;
+ }
+ status &= mask;
+ if (status) /* clear processed interrupts */
+ csio_wr_reg32(hw, status, reg);
+ return fatal;
+}
+
+/*
+ * Interrupt handler for the PCIE module.
+ */
+static void
+csio_pcie_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info sysbus_intr_info[] = {
+ { RNPP, "RXNP array parity error", -1, 1 },
+ { RPCP, "RXPC array parity error", -1, 1 },
+ { RCIP, "RXCIF array parity error", -1, 1 },
+ { RCCP, "Rx completions control array parity error", -1, 1 },
+ { RFTP, "RXFT array parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+ static struct intr_info pcie_port_intr_info[] = {
+ { TPCP, "TXPC array parity error", -1, 1 },
+ { TNPP, "TXNP array parity error", -1, 1 },
+ { TFTP, "TXFT array parity error", -1, 1 },
+ { TCAP, "TXCA array parity error", -1, 1 },
+ { TCIP, "TXCIF array parity error", -1, 1 },
+ { RCAP, "RXCA array parity error", -1, 1 },
+ { OTDD, "outbound request TLP discarded", -1, 1 },
+ { RDPE, "Rx data parity error", -1, 1 },
+ { TDUE, "Tx uncorrectable data error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+ static struct intr_info pcie_intr_info[] = {
+ { MSIADDRLPERR, "MSI AddrL parity error", -1, 1 },
+ { MSIADDRHPERR, "MSI AddrH parity error", -1, 1 },
+ { MSIDATAPERR, "MSI data parity error", -1, 1 },
+ { MSIXADDRLPERR, "MSI-X AddrL parity error", -1, 1 },
+ { MSIXADDRHPERR, "MSI-X AddrH parity error", -1, 1 },
+ { MSIXDATAPERR, "MSI-X data parity error", -1, 1 },
+ { MSIXDIPERR, "MSI-X DI parity error", -1, 1 },
+ { PIOCPLPERR, "PCI PIO completion FIFO parity error", -1, 1 },
+ { PIOREQPERR, "PCI PIO request FIFO parity error", -1, 1 },
+ { TARTAGPERR, "PCI PCI target tag FIFO parity error", -1, 1 },
+ { CCNTPERR, "PCI CMD channel count parity error", -1, 1 },
+ { CREQPERR, "PCI CMD channel request parity error", -1, 1 },
+ { CRSPPERR, "PCI CMD channel response parity error", -1, 1 },
+ { DCNTPERR, "PCI DMA channel count parity error", -1, 1 },
+ { DREQPERR, "PCI DMA channel request parity error", -1, 1 },
+ { DRSPPERR, "PCI DMA channel response parity error", -1, 1 },
+ { HCNTPERR, "PCI HMA channel count parity error", -1, 1 },
+ { HREQPERR, "PCI HMA channel request parity error", -1, 1 },
+ { HRSPPERR, "PCI HMA channel response parity error", -1, 1 },
+ { CFGSNPPERR, "PCI config snoop FIFO parity error", -1, 1 },
+ { FIDPERR, "PCI FID parity error", -1, 1 },
+ { INTXCLRPERR, "PCI INTx clear parity error", -1, 1 },
+ { MATAGPERR, "PCI MA tag parity error", -1, 1 },
+ { PIOTAGPERR, "PCI PIO tag parity error", -1, 1 },
+ { RXCPLPERR, "PCI Rx completion parity error", -1, 1 },
+ { RXWRPERR, "PCI Rx write parity error", -1, 1 },
+ { RPLPERR, "PCI replay buffer parity error", -1, 1 },
+ { PCIESINT, "PCI core secondary fault", -1, 1 },
+ { PCIEPINT, "PCI core primary fault", -1, 1 },
+ { UNXSPLCPLERR, "PCI unexpected split completion error", -1,
+ 0 },
+ { 0, NULL, 0, 0 }
+ };
+
+ int fat;
+
+ fat = csio_handle_intr_status(hw,
+ PCIE_CORE_UTL_SYSTEM_BUS_AGENT_STATUS,
+ sysbus_intr_info) +
+ csio_handle_intr_status(hw,
+ PCIE_CORE_UTL_PCI_EXPRESS_PORT_STATUS,
+ pcie_port_intr_info) +
+ csio_handle_intr_status(hw, PCIE_INT_CAUSE, pcie_intr_info);
+ if (fat)
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * TP interrupt handler.
+ */
+static void csio_tp_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info tp_intr_info[] = {
+ { 0x3fffffff, "TP parity error", -1, 1 },
+ { FLMTXFLSTEMPTY, "TP out of Tx pages", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, TP_INT_CAUSE, tp_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * SGE interrupt handler.
+ */
+static void csio_sge_intr_handler(struct csio_hw *hw)
+{
+ uint64_t v;
+
+ static struct intr_info sge_intr_info[] = {
+ { ERR_CPL_EXCEED_IQE_SIZE,
+ "SGE received CPL exceeding IQE size", -1, 1 },
+ { ERR_INVALID_CIDX_INC,
+ "SGE GTS CIDX increment too large", -1, 0 },
+ { ERR_CPL_OPCODE_0, "SGE received 0-length CPL", -1, 0 },
+ { ERR_DROPPED_DB, "SGE doorbell dropped", -1, 0 },
+ { ERR_DATA_CPL_ON_HIGH_QID1 | ERR_DATA_CPL_ON_HIGH_QID0,
+ "SGE IQID > 1023 received CPL for FL", -1, 0 },
+ { ERR_BAD_DB_PIDX3, "SGE DBP 3 pidx increment too large", -1,
+ 0 },
+ { ERR_BAD_DB_PIDX2, "SGE DBP 2 pidx increment too large", -1,
+ 0 },
+ { ERR_BAD_DB_PIDX1, "SGE DBP 1 pidx increment too large", -1,
+ 0 },
+ { ERR_BAD_DB_PIDX0, "SGE DBP 0 pidx increment too large", -1,
+ 0 },
+ { ERR_ING_CTXT_PRIO,
+ "SGE too many priority ingress contexts", -1, 0 },
+ { ERR_EGR_CTXT_PRIO,
+ "SGE too many priority egress contexts", -1, 0 },
+ { INGRESS_SIZE_ERR, "SGE illegal ingress QID", -1, 0 },
+ { EGRESS_SIZE_ERR, "SGE illegal egress QID", -1, 0 },
+ { 0, NULL, 0, 0 }
+ };
+
+ v = (uint64_t)csio_rd_reg32(hw, SGE_INT_CAUSE1) |
+ ((uint64_t)csio_rd_reg32(hw, SGE_INT_CAUSE2) << 32);
+ if (v) {
+ csio_fatal(hw, "SGE parity error (%#llx)\n",
+ (unsigned long long)v);
+ csio_wr_reg32(hw, (uint32_t)(v & 0xFFFFFFFF),
+ SGE_INT_CAUSE1);
+ csio_wr_reg32(hw, (uint32_t)(v >> 32), SGE_INT_CAUSE2);
+ }
+
+ v |= csio_handle_intr_status(hw, SGE_INT_CAUSE3, sge_intr_info);
+
+ if (csio_handle_intr_status(hw, SGE_INT_CAUSE3, sge_intr_info) ||
+ v != 0)
+ csio_hw_fatal_err(hw);
+}
+
+#define CIM_OBQ_INTR (OBQULP0PARERR | OBQULP1PARERR | OBQULP2PARERR |\
+ OBQULP3PARERR | OBQSGEPARERR | OBQNCSIPARERR)
+#define CIM_IBQ_INTR (IBQTP0PARERR | IBQTP1PARERR | IBQULPPARERR |\
+ IBQSGEHIPARERR | IBQSGELOPARERR | IBQNCSIPARERR)
+
+/*
+ * CIM interrupt handler.
+ */
+static void csio_cim_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info cim_intr_info[] = {
+ { PREFDROPINT, "CIM control register prefetch drop", -1, 1 },
+ { CIM_OBQ_INTR, "CIM OBQ parity error", -1, 1 },
+ { CIM_IBQ_INTR, "CIM IBQ parity error", -1, 1 },
+ { MBUPPARERR, "CIM mailbox uP parity error", -1, 1 },
+ { MBHOSTPARERR, "CIM mailbox host parity error", -1, 1 },
+ { TIEQINPARERRINT, "CIM TIEQ outgoing parity error", -1, 1 },
+ { TIEQOUTPARERRINT, "CIM TIEQ incoming parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+ static struct intr_info cim_upintr_info[] = {
+ { RSVDSPACEINT, "CIM reserved space access", -1, 1 },
+ { ILLTRANSINT, "CIM illegal transaction", -1, 1 },
+ { ILLWRINT, "CIM illegal write", -1, 1 },
+ { ILLRDINT, "CIM illegal read", -1, 1 },
+ { ILLRDBEINT, "CIM illegal read BE", -1, 1 },
+ { ILLWRBEINT, "CIM illegal write BE", -1, 1 },
+ { SGLRDBOOTINT, "CIM single read from boot space", -1, 1 },
+ { SGLWRBOOTINT, "CIM single write to boot space", -1, 1 },
+ { BLKWRBOOTINT, "CIM block write to boot space", -1, 1 },
+ { SGLRDFLASHINT, "CIM single read from flash space", -1, 1 },
+ { SGLWRFLASHINT, "CIM single write to flash space", -1, 1 },
+ { BLKWRFLASHINT, "CIM block write to flash space", -1, 1 },
+ { SGLRDEEPROMINT, "CIM single EEPROM read", -1, 1 },
+ { SGLWREEPROMINT, "CIM single EEPROM write", -1, 1 },
+ { BLKRDEEPROMINT, "CIM block EEPROM read", -1, 1 },
+ { BLKWREEPROMINT, "CIM block EEPROM write", -1, 1 },
+ { SGLRDCTLINT , "CIM single read from CTL space", -1, 1 },
+ { SGLWRCTLINT , "CIM single write to CTL space", -1, 1 },
+ { BLKRDCTLINT , "CIM block read from CTL space", -1, 1 },
+ { BLKWRCTLINT , "CIM block write to CTL space", -1, 1 },
+ { SGLRDPLINT , "CIM single read from PL space", -1, 1 },
+ { SGLWRPLINT , "CIM single write to PL space", -1, 1 },
+ { BLKRDPLINT , "CIM block read from PL space", -1, 1 },
+ { BLKWRPLINT , "CIM block write to PL space", -1, 1 },
+ { REQOVRLOOKUPINT , "CIM request FIFO overwrite", -1, 1 },
+ { RSPOVRLOOKUPINT , "CIM response FIFO overwrite", -1, 1 },
+ { TIMEOUTINT , "CIM PIF timeout", -1, 1 },
+ { TIMEOUTMAINT , "CIM PIF MA timeout", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ int fat;
+
+ fat = csio_handle_intr_status(hw, CIM_HOST_INT_CAUSE,
+ cim_intr_info) +
+ csio_handle_intr_status(hw, CIM_HOST_UPACC_INT_CAUSE,
+ cim_upintr_info);
+ if (fat)
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * ULP RX interrupt handler.
+ */
+static void csio_ulprx_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info ulprx_intr_info[] = {
+ { 0x1800000, "ULPRX context error", -1, 1 },
+ { 0x7fffff, "ULPRX parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, ULP_RX_INT_CAUSE, ulprx_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * ULP TX interrupt handler.
+ */
+static void csio_ulptx_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info ulptx_intr_info[] = {
+ { PBL_BOUND_ERR_CH3, "ULPTX channel 3 PBL out of bounds", -1,
+ 0 },
+ { PBL_BOUND_ERR_CH2, "ULPTX channel 2 PBL out of bounds", -1,
+ 0 },
+ { PBL_BOUND_ERR_CH1, "ULPTX channel 1 PBL out of bounds", -1,
+ 0 },
+ { PBL_BOUND_ERR_CH0, "ULPTX channel 0 PBL out of bounds", -1,
+ 0 },
+ { 0xfffffff, "ULPTX parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, ULP_TX_INT_CAUSE, ulptx_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * PM TX interrupt handler.
+ */
+static void csio_pmtx_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info pmtx_intr_info[] = {
+ { PCMD_LEN_OVFL0, "PMTX channel 0 pcmd too large", -1, 1 },
+ { PCMD_LEN_OVFL1, "PMTX channel 1 pcmd too large", -1, 1 },
+ { PCMD_LEN_OVFL2, "PMTX channel 2 pcmd too large", -1, 1 },
+ { ZERO_C_CMD_ERROR, "PMTX 0-length pcmd", -1, 1 },
+ { 0xffffff0, "PMTX framing error", -1, 1 },
+ { OESPI_PAR_ERROR, "PMTX oespi parity error", -1, 1 },
+ { DB_OPTIONS_PAR_ERROR, "PMTX db_options parity error", -1,
+ 1 },
+ { ICSPI_PAR_ERROR, "PMTX icspi parity error", -1, 1 },
+ { C_PCMD_PAR_ERROR, "PMTX c_pcmd parity error", -1, 1},
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, PM_TX_INT_CAUSE, pmtx_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * PM RX interrupt handler.
+ */
+static void csio_pmrx_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info pmrx_intr_info[] = {
+ { ZERO_E_CMD_ERROR, "PMRX 0-length pcmd", -1, 1 },
+ { 0x3ffff0, "PMRX framing error", -1, 1 },
+ { OCSPI_PAR_ERROR, "PMRX ocspi parity error", -1, 1 },
+ { DB_OPTIONS_PAR_ERROR, "PMRX db_options parity error", -1,
+ 1 },
+ { IESPI_PAR_ERROR, "PMRX iespi parity error", -1, 1 },
+ { E_PCMD_PAR_ERROR, "PMRX e_pcmd parity error", -1, 1},
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, PM_RX_INT_CAUSE, pmrx_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * CPL switch interrupt handler.
+ */
+static void csio_cplsw_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info cplsw_intr_info[] = {
+ { CIM_OP_MAP_PERR, "CPLSW CIM op_map parity error", -1, 1 },
+ { CIM_OVFL_ERROR, "CPLSW CIM overflow", -1, 1 },
+ { TP_FRAMING_ERROR, "CPLSW TP framing error", -1, 1 },
+ { SGE_FRAMING_ERROR, "CPLSW SGE framing error", -1, 1 },
+ { CIM_FRAMING_ERROR, "CPLSW CIM framing error", -1, 1 },
+ { ZERO_SWITCH_ERROR, "CPLSW no-switch error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, CPL_INTR_CAUSE, cplsw_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * LE interrupt handler.
+ */
+static void csio_le_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info le_intr_info[] = {
+ { LIPMISS, "LE LIP miss", -1, 0 },
+ { LIP0, "LE 0 LIP error", -1, 0 },
+ { PARITYERR, "LE parity error", -1, 1 },
+ { UNKNOWNCMD, "LE unknown command", -1, 1 },
+ { REQQPARERR, "LE request queue parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, LE_DB_INT_CAUSE, le_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * MPS interrupt handler.
+ */
+static void csio_mps_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info mps_rx_intr_info[] = {
+ { 0xffffff, "MPS Rx parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+ static struct intr_info mps_tx_intr_info[] = {
+ { TPFIFO, "MPS Tx TP FIFO parity error", -1, 1 },
+ { NCSIFIFO, "MPS Tx NC-SI FIFO parity error", -1, 1 },
+ { TXDATAFIFO, "MPS Tx data FIFO parity error", -1, 1 },
+ { TXDESCFIFO, "MPS Tx desc FIFO parity error", -1, 1 },
+ { BUBBLE, "MPS Tx underflow", -1, 1 },
+ { SECNTERR, "MPS Tx SOP/EOP error", -1, 1 },
+ { FRMERR, "MPS Tx framing error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+ static struct intr_info mps_trc_intr_info[] = {
+ { FILTMEM, "MPS TRC filter parity error", -1, 1 },
+ { PKTFIFO, "MPS TRC packet FIFO parity error", -1, 1 },
+ { MISCPERR, "MPS TRC misc parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+ static struct intr_info mps_stat_sram_intr_info[] = {
+ { 0x1fffff, "MPS statistics SRAM parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+ static struct intr_info mps_stat_tx_intr_info[] = {
+ { 0xfffff, "MPS statistics Tx FIFO parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+ static struct intr_info mps_stat_rx_intr_info[] = {
+ { 0xffffff, "MPS statistics Rx FIFO parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+ static struct intr_info mps_cls_intr_info[] = {
+ { MATCHSRAM, "MPS match SRAM parity error", -1, 1 },
+ { MATCHTCAM, "MPS match TCAM parity error", -1, 1 },
+ { HASHSRAM, "MPS hash SRAM parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ int fat;
+
+ fat = csio_handle_intr_status(hw, MPS_RX_PERR_INT_CAUSE,
+ mps_rx_intr_info) +
+ csio_handle_intr_status(hw, MPS_TX_INT_CAUSE,
+ mps_tx_intr_info) +
+ csio_handle_intr_status(hw, MPS_TRC_INT_CAUSE,
+ mps_trc_intr_info) +
+ csio_handle_intr_status(hw, MPS_STAT_PERR_INT_CAUSE_SRAM,
+ mps_stat_sram_intr_info) +
+ csio_handle_intr_status(hw, MPS_STAT_PERR_INT_CAUSE_TX_FIFO,
+ mps_stat_tx_intr_info) +
+ csio_handle_intr_status(hw, MPS_STAT_PERR_INT_CAUSE_RX_FIFO,
+ mps_stat_rx_intr_info) +
+ csio_handle_intr_status(hw, MPS_CLS_INT_CAUSE,
+ mps_cls_intr_info);
+
+ csio_wr_reg32(hw, 0, MPS_INT_CAUSE);
+ csio_rd_reg32(hw, MPS_INT_CAUSE); /* flush */
+ if (fat)
+ csio_hw_fatal_err(hw);
+}
+
+#define MEM_INT_MASK (PERR_INT_CAUSE | ECC_CE_INT_CAUSE | ECC_UE_INT_CAUSE)
+
+/*
+ * EDC/MC interrupt handler.
+ */
+static void csio_mem_intr_handler(struct csio_hw *hw, int idx)
+{
+ static const char name[3][5] = { "EDC0", "EDC1", "MC" };
+
+ unsigned int addr, cnt_addr, v;
+
+ if (idx <= MEM_EDC1) {
+ addr = EDC_REG(EDC_INT_CAUSE, idx);
+ cnt_addr = EDC_REG(EDC_ECC_STATUS, idx);
+ } else {
+ addr = MC_INT_CAUSE;
+ cnt_addr = MC_ECC_STATUS;
+ }
+
+ v = csio_rd_reg32(hw, addr) & MEM_INT_MASK;
+ if (v & PERR_INT_CAUSE)
+ csio_fatal(hw, "%s FIFO parity error\n", name[idx]);
+ if (v & ECC_CE_INT_CAUSE) {
+ uint32_t cnt = ECC_CECNT_GET(csio_rd_reg32(hw, cnt_addr));
+
+ csio_wr_reg32(hw, ECC_CECNT_MASK, cnt_addr);
+ csio_warn(hw, "%u %s correctable ECC data error%s\n",
+ cnt, name[idx], cnt > 1 ? "s" : "");
+ }
+ if (v & ECC_UE_INT_CAUSE)
+ csio_fatal(hw, "%s uncorrectable ECC data error\n", name[idx]);
+
+ csio_wr_reg32(hw, v, addr);
+ if (v & (PERR_INT_CAUSE | ECC_UE_INT_CAUSE))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * MA interrupt handler.
+ */
+static void csio_ma_intr_handler(struct csio_hw *hw)
+{
+ uint32_t v, status = csio_rd_reg32(hw, MA_INT_CAUSE);
+
+ if (status & MEM_PERR_INT_CAUSE)
+ csio_fatal(hw, "MA parity error, parity status %#x\n",
+ csio_rd_reg32(hw, MA_PARITY_ERROR_STATUS));
+ if (status & MEM_WRAP_INT_CAUSE) {
+ v = csio_rd_reg32(hw, MA_INT_WRAP_STATUS);
+ csio_fatal(hw,
+ "MA address wrap-around error by client %u to address %#x\n",
+ MEM_WRAP_CLIENT_NUM_GET(v), MEM_WRAP_ADDRESS_GET(v) << 4);
+ }
+ csio_wr_reg32(hw, status, MA_INT_CAUSE);
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * SMB interrupt handler.
+ */
+static void csio_smb_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info smb_intr_info[] = {
+ { MSTTXFIFOPARINT, "SMB master Tx FIFO parity error", -1, 1 },
+ { MSTRXFIFOPARINT, "SMB master Rx FIFO parity error", -1, 1 },
+ { SLVFIFOPARINT, "SMB slave FIFO parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, SMB_INT_CAUSE, smb_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * NC-SI interrupt handler.
+ */
+static void csio_ncsi_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info ncsi_intr_info[] = {
+ { CIM_DM_PRTY_ERR, "NC-SI CIM parity error", -1, 1 },
+ { MPS_DM_PRTY_ERR, "NC-SI MPS parity error", -1, 1 },
+ { TXFIFO_PRTY_ERR, "NC-SI Tx FIFO parity error", -1, 1 },
+ { RXFIFO_PRTY_ERR, "NC-SI Rx FIFO parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, NCSI_INT_CAUSE, ncsi_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * XGMAC interrupt handler.
+ */
+static void csio_xgmac_intr_handler(struct csio_hw *hw, int port)
+{
+ uint32_t v = csio_rd_reg32(hw, PORT_REG(port, XGMAC_PORT_INT_CAUSE));
+
+ v &= TXFIFO_PRTY_ERR | RXFIFO_PRTY_ERR;
+ if (!v)
+ return;
+
+ if (v & TXFIFO_PRTY_ERR)
+ csio_fatal(hw, "XGMAC %d Tx FIFO parity error\n", port);
+ if (v & RXFIFO_PRTY_ERR)
+ csio_fatal(hw, "XGMAC %d Rx FIFO parity error\n", port);
+ csio_wr_reg32(hw, v, PORT_REG(port, XGMAC_PORT_INT_CAUSE));
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * PL interrupt handler.
+ */
+static void csio_pl_intr_handler(struct csio_hw *hw)
+{
+ static struct intr_info pl_intr_info[] = {
+ { FATALPERR, "T4 fatal parity error", -1, 1 },
+ { PERRVFID, "PL VFID_MAP parity error", -1, 1 },
+ { 0, NULL, 0, 0 }
+ };
+
+ if (csio_handle_intr_status(hw, PL_PL_INT_CAUSE, pl_intr_info))
+ csio_hw_fatal_err(hw);
+}
+
+/*
+ * csio_hw_slow_intr_handler - control path interrupt handler
+ * @hw: HW module
+ *
+ * Interrupt handler for non-data global interrupt events, e.g., errors.
+ * The designation 'slow' is because it involves register reads, while
+ * data interrupts typically don't involve any MMIOs.
+ */
+int
+csio_hw_slow_intr_handler(struct csio_hw *hw)
+{
+ uint32_t cause = csio_rd_reg32(hw, PL_INT_CAUSE);
+
+ if (!(cause & CSIO_GLBL_INTR_MASK)) {
+ csio_inc_stats(hw, n_plint_unexp);
+ return 0;
+ }
+
+ csio_dbg(hw, "Slow interrupt! cause: 0x%x\n", cause);
+
+ csio_inc_stats(hw, n_plint_cnt);
+
+ if (cause & CIM)
+ csio_cim_intr_handler(hw);
+
+ if (cause & MPS)
+ csio_mps_intr_handler(hw);
+
+ if (cause & NCSI)
+ csio_ncsi_intr_handler(hw);
+
+ if (cause & PL)
+ csio_pl_intr_handler(hw);
+
+ if (cause & SMB)
+ csio_smb_intr_handler(hw);
+
+ if (cause & XGMAC0)
+ csio_xgmac_intr_handler(hw, 0);
+
+ if (cause & XGMAC1)
+ csio_xgmac_intr_handler(hw, 1);
+
+ if (cause & XGMAC_KR0)
+ csio_xgmac_intr_handler(hw, 2);
+
+ if (cause & XGMAC_KR1)
+ csio_xgmac_intr_handler(hw, 3);
+
+ if (cause & PCIE)
+ csio_pcie_intr_handler(hw);
+
+ if (cause & MC)
+ csio_mem_intr_handler(hw, MEM_MC);
+
+ if (cause & EDC0)
+ csio_mem_intr_handler(hw, MEM_EDC0);
+
+ if (cause & EDC1)
+ csio_mem_intr_handler(hw, MEM_EDC1);
+
+ if (cause & LE)
+ csio_le_intr_handler(hw);
+
+ if (cause & TP)
+ csio_tp_intr_handler(hw);
+
+ if (cause & MA)
+ csio_ma_intr_handler(hw);
+
+ if (cause & PM_TX)
+ csio_pmtx_intr_handler(hw);
+
+ if (cause & PM_RX)
+ csio_pmrx_intr_handler(hw);
+
+ if (cause & ULP_RX)
+ csio_ulprx_intr_handler(hw);
+
+ if (cause & CPL_SWITCH)
+ csio_cplsw_intr_handler(hw);
+
+ if (cause & SGE)
+ csio_sge_intr_handler(hw);
+
+ if (cause & ULP_TX)
+ csio_ulptx_intr_handler(hw);
+
+ /* Clear the interrupts just processed for which we are the master. */
+ csio_wr_reg32(hw, cause & CSIO_GLBL_INTR_MASK, PL_INT_CAUSE);
+ csio_rd_reg32(hw, PL_INT_CAUSE); /* flush */
+
+ return 1;
+}
+
+/*****************************************************************************
+ * HW <--> mailbox interfacing routines.
+ ****************************************************************************/
+/*
+ * csio_mberr_worker - Worker thread (dpc) for mailbox/error completions
+ *
+ * @data: Private data pointer.
+ *
+ * Called from worker thread context.
+ */
+static void
+csio_mberr_worker(void *data)
+{
+ struct csio_hw *hw = (struct csio_hw *)data;
+ struct csio_mbm *mbm = &hw->mbm;
+ LIST_HEAD(cbfn_q);
+ struct csio_mb *mbp_next;
+ csio_retval_t rv;
+
+ del_timer_sync(&mbm->timer);
+
+ spin_lock_irq(&hw->lock);
+ if (list_empty(&mbm->cbfn_q)) {
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ list_splice_tail_init(&mbm->cbfn_q, &cbfn_q);
+ mbm->stats.n_cbfnq = 0;
+
+ /* Try to start waiting mailboxes */
+ csio_deq_from_head(&mbm->req_q, &mbp_next);
+ if (mbp_next) {
+ rv = csio_mb_issue(hw, mbp_next);
+ if (rv != CSIO_SUCCESS)
+ list_add_tail(&mbp_next->list, &mbm->req_q);
+ else
+ csio_dec_stats(mbm, n_activeq);
+ }
+ spin_unlock_irq(&hw->lock);
+
+ /* Now callback completions */
+ csio_mb_completions(hw, &cbfn_q);
+
+}
+
+/*
+ * csio_hw_mb_timer - Top-level Mailbox timeout handler.
+ *
+ * @data: private data pointer
+ *
+ **/
+static void
+csio_hw_mb_timer(uintptr_t data)
+{
+ struct csio_hw *hw = (struct csio_hw *)data;
+ struct csio_mb *mbp = NULL;
+
+ spin_lock_irq(&hw->lock);
+ mbp = csio_mb_tmo_handler(hw);
+ spin_unlock_irq(&hw->lock);
+
+ /* Call back the function for the timed-out Mailbox */
+ if (mbp)
+ mbp->mb_cbfn(hw, mbp);
+
+}
+
+/*
+ * csio_hw_mbm_cleanup - Cleanup Mailbox module.
+ * @hw: HW module
+ *
+ * Called with lock held, should exit with lock held.
+ * Cancels outstanding mailboxes (waiting, in-flight) and gathers them
+ * into a local queue. Drops lock and calls the completions. Holds
+ * lock and returns.
+ */
+static void
+csio_hw_mbm_cleanup(struct csio_hw *hw)
+{
+ LIST_HEAD(cbfn_q);
+
+ csio_mb_cancel_all(hw, &cbfn_q);
+
+ spin_unlock_irq(&hw->lock);
+ csio_mb_completions(hw, &cbfn_q);
+ spin_lock_irq(&hw->lock);
+}
+
+/*****************************************************************************
+ * Event handling
+ ****************************************************************************/
+csio_retval_t
+csio_enqueue_evt(struct csio_hw *hw, enum csio_evt type, void *evt_msg,
+ uint16_t len)
+{
+ struct csio_evt_msg *evt_entry = NULL;
+
+ if (type >= CSIO_EVT_MAX)
+ return CSIO_INVAL;
+
+ if (len > CSIO_EVT_MSG_SIZE)
+ return CSIO_INVAL;
+
+ if (hw->flags & CSIO_HWF_FWEVT_STOP)
+ return CSIO_INVAL;
+
+ csio_deq_from_head(&hw->evt_free_q, &evt_entry);
+ if (!evt_entry) {
+ csio_err(hw, "Failed to alloc evt entry, msg type %d len %d\n",
+ type, len);
+ return CSIO_NOMEM;
+ }
+
+ /* copy event msg and queue the event */
+ evt_entry->type = type;
+ memcpy((void *)evt_entry->data, evt_msg, len);
+ list_add_tail(&evt_entry->list, &hw->evt_active_q);
+
+ csio_dec_stats(hw, n_evt_freeq);
+ csio_inc_stats(hw, n_evt_activeq);
+
+ return CSIO_SUCCESS;
+}
+
+static csio_retval_t
+csio_enqueue_evt_lock(struct csio_hw *hw, enum csio_evt type, void *evt_msg,
+ uint16_t len, bool msg_sg)
+{
+ struct csio_evt_msg *evt_entry = NULL;
+ struct csio_fl_dma_buf *fl_sg;
+ uint32_t off = 0;
+ unsigned long flags;
+ int n;
+
+ if (type >= CSIO_EVT_MAX)
+ return CSIO_INVAL;
+
+ if (len > CSIO_EVT_MSG_SIZE)
+ return CSIO_INVAL;
+
+ spin_lock_irqsave(&hw->lock, flags);
+ if (hw->flags & CSIO_HWF_FWEVT_STOP) {
+ spin_unlock_irqrestore(&hw->lock, flags);
+ return CSIO_INVAL;
+ }
+ csio_deq_from_head(&hw->evt_free_q, &evt_entry);
+ if (!evt_entry) {
+ csio_err(hw, "Failed to alloc evt entry, msg type %d len %d\n",
+ type, len);
+ spin_unlock_irqrestore(&hw->lock, flags);
+ return CSIO_NOMEM;
+ }
+
+ /* copy event msg and queue the event */
+ evt_entry->type = type;
+
+ /* If Payload in SG list*/
+ if (msg_sg) {
+ fl_sg = (struct csio_fl_dma_buf *) evt_msg;
+ for (n = 0; (n < CSIO_MAX_FLBUF_PER_IQWR && off < len); n++) {
+ memcpy((void *)((uintptr_t)evt_entry->data + off),
+ fl_sg->flbufs[n].vaddr,
+ fl_sg->flbufs[n].len);
+ off += fl_sg->flbufs[n].len;
+ }
+ } else
+ memcpy((void *)evt_entry->data, evt_msg, len);
+
+ list_add_tail(&evt_entry->list, &hw->evt_active_q);
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+ csio_dec_stats(hw, n_evt_freeq);
+ csio_inc_stats(hw, n_evt_activeq);
+
+ return CSIO_SUCCESS;
+}
+
+static void
+csio_free_evt(struct csio_hw *hw, struct csio_evt_msg *evt_entry)
+{
+ if (evt_entry) {
+ spin_lock_irq(&hw->lock);
+ list_del_init(&evt_entry->list);
+ list_add_tail(&evt_entry->list, &hw->evt_free_q);
+ csio_dec_stats(hw, n_evt_activeq);
+ csio_inc_stats(hw, n_evt_freeq);
+ spin_unlock_irq(&hw->lock);
+ }
+}
+
+void
+csio_evtq_flush(struct csio_hw *hw)
+{
+ uint32_t count;
+ count = 30;
+ while (hw->flags & CSIO_HWF_FWEVT_PENDING && count--) {
+ spin_unlock_irq(&hw->lock);
+ msleep(2000);
+ spin_lock_irq(&hw->lock);
+ }
+
+ CSIO_DB_ASSERT(!(hw->flags & CSIO_HWF_FWEVT_PENDING));
+}
+
+static void
+csio_evtq_stop(struct csio_hw *hw)
+{
+ hw->flags |= CSIO_HWF_FWEVT_STOP;
+}
+
+static void
+csio_evtq_start(struct csio_hw *hw)
+{
+ hw->flags &= ~CSIO_HWF_FWEVT_STOP;
+}
+
+static void
+csio_evtq_cleanup(struct csio_hw *hw)
+{
+ struct list_head *evt_entry, *next_entry;
+
+ /* Release outstanding events from activeq to freeq*/
+ if (!list_empty(&hw->evt_active_q))
+ list_splice_tail_init(&hw->evt_active_q, &hw->evt_free_q);
+
+ hw->stats.n_evt_activeq = 0;
+ hw->flags &= ~CSIO_HWF_FWEVT_PENDING;
+
+ /* Freeup event entry */
+ list_for_each_safe(evt_entry, next_entry, &hw->evt_free_q) {
+ kfree(evt_entry);
+ csio_dec_stats(hw, n_evt_freeq);
+ }
+
+ hw->stats.n_evt_freeq = 0;
+}
+
+
+static void
+csio_process_fwevtq_entry(struct csio_hw *hw, void *wr, uint32_t len,
+ struct csio_fl_dma_buf *flb, void *priv)
+{
+ __u8 op;
+ __be64 *data;
+ void *msg = NULL;
+ uint32_t msg_len = 0;
+ bool msg_sg = 0;
+
+ op = ((struct rss_header *) wr)->opcode;
+ if (op == CPL_FW6_PLD) {
+ csio_inc_stats(hw, n_cpl_fw6_pld);
+ if (!flb || !flb->totlen) {
+ csio_inc_stats(hw, n_cpl_unexp);
+ return;
+ }
+
+ msg = (void *) flb;
+ msg_len = flb->totlen;
+ msg_sg = 1;
+
+ data = (__be64 *) msg;
+ } else if (op == CPL_FW6_MSG || op == CPL_FW4_MSG) {
+
+ csio_inc_stats(hw, n_cpl_fw6_msg);
+ /* skip RSS header */
+ msg = (void *)((uintptr_t)wr + sizeof(__be64));
+ msg_len = (op == CPL_FW6_MSG) ? sizeof(struct cpl_fw6_msg) :
+ sizeof(struct cpl_fw4_msg);
+
+ data = (__be64 *) msg;
+ } else {
+ csio_warn(hw, "unexpected CPL %#x on FW event queue\n", op);
+ csio_inc_stats(hw, n_cpl_unexp);
+ return;
+ }
+
+ /*
+ * Enqueue event to EventQ. Events processing happens
+ * in Event worker thread context
+ */
+ if (csio_enqueue_evt_lock(hw, CSIO_EVT_FW, msg,
+ (uint16_t)msg_len, msg_sg))
+ csio_inc_stats(hw, n_evt_drop);
+}
+
+void
+csio_evtq_worker(struct work_struct *work)
+{
+ struct csio_hw *hw = container_of(work, struct csio_hw, evtq_work);
+ struct list_head *evt_entry, *next_entry;
+ LIST_HEAD(evt_q);
+ struct csio_evt_msg *evt_msg;
+ struct cpl_fw6_msg *msg;
+ struct csio_rnode *rn;
+ csio_retval_t rv = 0;
+ uint8_t evtq_stop = 0;
+
+ csio_dbg(hw, "event worker thread active evts#%d\n",
+ hw->stats.n_evt_activeq);
+
+ spin_lock_irq(&hw->lock);
+ while (!list_empty(&hw->evt_active_q)) {
+ list_splice_tail_init(&hw->evt_active_q, &evt_q);
+ spin_unlock_irq(&hw->lock);
+
+ list_for_each_safe(evt_entry, next_entry, &evt_q) {
+ evt_msg = (struct csio_evt_msg *) evt_entry;
+
+ /* Drop events if queue is STOPPED */
+ spin_lock_irq(&hw->lock);
+ if (hw->flags & CSIO_HWF_FWEVT_STOP)
+ evtq_stop = 1;
+ spin_unlock_irq(&hw->lock);
+ if (evtq_stop) {
+ csio_inc_stats(hw, n_evt_drop);
+ goto free_evt;
+ }
+
+ switch (evt_msg->type) {
+
+ case CSIO_EVT_FW:
+ msg = (struct cpl_fw6_msg *)(evt_msg->data);
+
+ if ((msg->opcode == CPL_FW6_MSG ||
+ msg->opcode == CPL_FW4_MSG) &&
+ !msg->type) {
+ rv = csio_mb_fwevt_handler(hw,
+ msg->data);
+ if (!rv)
+ break;
+ /* Handle any remaining fw events */
+ csio_fcoe_fwevt_handler(hw,
+ msg->opcode, msg->data);
+ } else if (msg->opcode == CPL_FW6_PLD) {
+
+ csio_fcoe_fwevt_handler(hw,
+ msg->opcode, msg->data);
+ } else {
+ csio_warn(hw,
+ "Unhandled FW msg op %x type %x\n",
+ msg->opcode, msg->type);
+ csio_inc_stats(hw, n_evt_drop);
+ }
+ break;
+
+ case CSIO_EVT_MBX:
+ csio_mberr_worker(hw);
+ break;
+
+ case CSIO_EVT_DEV_LOSS:
+ memcpy(&rn, evt_msg->data, sizeof(rn));
+ csio_rnode_devloss_handler(rn);
+ break;
+
+ default:
+ csio_warn(hw, "Unhandled event %x on evtq\n",
+ evt_msg->type);
+ csio_inc_stats(hw, n_evt_unexp);
+ break;
+ }
+free_evt:
+ csio_free_evt(hw, evt_msg);
+ }
+
+ spin_lock_irq(&hw->lock);
+ }
+ hw->flags &= ~CSIO_HWF_FWEVT_PENDING;
+ spin_unlock_irq(&hw->lock);
+}
+
+csio_retval_t
+csio_fwevtq_handler(struct csio_hw *hw)
+{
+ csio_retval_t rv;
+
+ if (csio_q_iqid(hw, hw->fwevt_iq_idx) == CSIO_MAX_QID) {
+ csio_inc_stats(hw, n_int_stray);
+ return CSIO_INVAL;
+ }
+
+ rv = csio_wr_process_iq_idx(hw, hw->fwevt_iq_idx,
+ csio_process_fwevtq_entry, NULL);
+ return rv;
+}
+
+/****************************************************************************
+ * Entry points
+ ****************************************************************************/
+
+/* Management module */
+/*
+ * csio_mgmt_req_lookup - Lookup the given IO req exist in Active Q.
+ * mgmt - mgmt module
+ * @io_req - io request
+ *
+ * Return - CSIO_SUCCESS:if given IO Req exists in active Q.
+ * CSIO_INVAL :if lookup fails.
+ */
+csio_retval_t
+csio_mgmt_req_lookup(struct csio_mgmtm *mgmtm, struct csio_ioreq *io_req)
+{
+ struct list_head *tmp;
+
+ /* Lookup ioreq in the ACTIVEQ */
+ list_for_each(tmp, &mgmtm->active_q) {
+ if (io_req == (struct csio_ioreq *)tmp)
+ return CSIO_SUCCESS;
+ }
+ return CSIO_INVAL;
+}
+
+/*
+ * csio_mgmts_tmo_handler - MGMT IO Timeout handler.
+ * @data - Event data.
+ *
+ * Return - none.
+ */
+static void
+csio_mgmt_tmo_handler(uintptr_t data)
+{
+ struct csio_mgmtm *mgmtm = (struct csio_mgmtm *) data;
+ struct list_head *tmp;
+ struct csio_ioreq *io_req;
+
+ csio_dbg(mgmtm->hw, "Mgmt timer invoked!\n");
+
+ spin_lock_irq(&mgmtm->hw->lock);
+
+ list_for_each(tmp, &mgmtm->active_q) {
+ io_req = (struct csio_ioreq *) tmp;
+ io_req->tmo -= min_t(uint32_t, io_req->tmo, ECM_MIN_TMO);
+
+ if (!io_req->tmo) {
+ /* Dequeue the request from retry Q. */
+ tmp = csio_list_prev(tmp);
+ list_del_init(&io_req->sm.sm_list);
+ if (io_req->io_cbfn) {
+ /* io_req will be freed by completion handler */
+ io_req->wr_status = CSIO_TIMEOUT;
+ io_req->io_cbfn(mgmtm->hw, io_req);
+ } else {
+ CSIO_DB_ASSERT(0);
+ }
+ }
+ }
+
+ /* If retry queue is not empty, re-arm timer */
+ if (!list_empty(&mgmtm->active_q))
+ mod_timer(&mgmtm->mgmt_timer,
+ jiffies + msecs_to_jiffies(ECM_MIN_TMO));
+ spin_unlock_irq(&mgmtm->hw->lock);
+}
+
+static void
+csio_mgmtm_cleanup(struct csio_mgmtm *mgmtm)
+{
+ struct csio_hw *hw = mgmtm->hw;
+ struct csio_ioreq *io_req;
+ struct list_head *tmp;
+ uint32_t count;
+
+ count = 30;
+ /* Wait for all outstanding req to complete gracefully */
+ while ((!list_empty(&mgmtm->active_q)) && count--) {
+ spin_unlock_irq(&hw->lock);
+ msleep(2000);
+ spin_lock_irq(&hw->lock);
+ }
+
+ /* release outstanding req from ACTIVEQ */
+ list_for_each(tmp, &mgmtm->active_q) {
+ io_req = (struct csio_ioreq *) tmp;
+ tmp = csio_list_prev(tmp);
+ list_del_init(&io_req->sm.sm_list);
+ mgmtm->stats.n_active--;
+ if (io_req->io_cbfn) {
+ /* io_req will be freed by completion handler */
+ io_req->wr_status = CSIO_TIMEOUT;
+ io_req->io_cbfn(mgmtm->hw, io_req);
+ }
+ }
+}
+
+/*
+ * csio_mgmt_init - Mgmt module init entry point
+ * @mgmtsm - mgmt module
+ * @hw - HW module
+ *
+ * Initialize mgmt timer, resource wait queue, active queue,
+ * completion q. Allocate Egress and Ingress
+ * WR queues and save off the queue index returned by the WR
+ * module for future use. Allocate and save off mgmt reqs in the
+ * mgmt_req_freelist for future use. Make sure their SM is initialized
+ * to uninit state.
+ * Returns: CSIO_SUCCESS - on success
+ * CSIO_NOMEM - on error.
+ */
+static csio_retval_t
+csio_mgmtm_init(struct csio_mgmtm *mgmtm, struct csio_hw *hw)
+{
+ struct timer_list *timer = &mgmtm->mgmt_timer;
+
+ init_timer(timer);
+ timer->function = csio_mgmt_tmo_handler;
+ timer->data = (unsigned long)mgmtm;
+
+ INIT_LIST_HEAD(&mgmtm->active_q);
+ INIT_LIST_HEAD(&mgmtm->cbfn_q);
+
+ mgmtm->hw = hw;
+ /*mgmtm->iq_idx = hw->fwevt_iq_idx;*/
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_mgmtm_exit - MGMT module exit entry point
+ * @mgmtsm - mgmt module
+ *
+ * This function called during MGMT module uninit.
+ * Stop timers, free ioreqs allocated.
+ * Returns: None
+ *
+ */
+static void
+csio_mgmtm_exit(struct csio_mgmtm *mgmtm)
+{
+ del_timer_sync(&mgmtm->mgmt_timer);
+}
+
+
+/**
+ * csio_hw_start - Kicks off the HW State machine
+ * @hw: Pointer to HW module.
+ *
+ * It is assumed that the initialization is a synchronous operation.
+ * So when we return afer posting the event, the HW SM should be in
+ * the ready state, if there were no errors during init.
+ */
+csio_retval_t
+csio_hw_start(struct csio_hw *hw)
+{
+ spin_lock_irq(&hw->lock);
+ csio_post_event(&hw->sm, CSIO_HWE_CFG);
+ spin_unlock_irq(&hw->lock);
+
+ if (csio_is_hw_ready(hw))
+ return CSIO_SUCCESS;
+ else
+ return CSIO_INVAL;
+}
+
+csio_retval_t
+csio_hw_stop(struct csio_hw *hw)
+{
+ csio_post_event(&hw->sm, CSIO_HWE_PCI_REMOVE);
+
+ if (csio_is_hw_removing(hw))
+ return CSIO_SUCCESS;
+ else
+ return CSIO_INVAL;
+}
+
+/* Max reset retries */
+#define CSIO_MAX_RESET_RETRIES 3
+
+/**
+ * csio_hw_reset - Reset the hardware
+ * @hw: HW module.
+ *
+ * Caller should hold lock across this function.
+ */
+csio_retval_t
+csio_hw_reset(struct csio_hw *hw)
+{
+ if (!csio_is_hw_master(hw))
+ return CSIO_NOPERM;
+
+ if (hw->rst_retries >= CSIO_MAX_RESET_RETRIES) {
+ csio_dbg(hw, "Max hw reset attempts reached..");
+ return CSIO_INVAL;
+ }
+
+ hw->rst_retries++;
+ csio_post_event(&hw->sm, CSIO_HWE_HBA_RESET);
+
+ if (csio_is_hw_ready(hw)) {
+ hw->rst_retries = 0;
+ hw->stats.n_reset_start = jiffies_to_msecs(jiffies);
+ return CSIO_SUCCESS;
+ } else
+ return CSIO_INVAL;
+}
+
+/*
+ * csio_hw_get_device_id - Caches the Adapter's vendor & device id.
+ * @hw: HW module.
+ */
+static void
+csio_hw_get_device_id(struct csio_hw *hw)
+{
+ /* Is the adapter device id cached already ?*/
+ if (csio_is_dev_id_cached(hw))
+ return;
+
+ /* Get the PCI vendor & device id */
+ pci_read_config_word(hw->pdev, PCI_VENDOR_ID,
+ &hw->params.pci.vendor_id);
+ pci_read_config_word(hw->pdev, PCI_DEVICE_ID,
+ &hw->params.pci.device_id);
+
+ csio_dev_id_cached(hw);
+
+} /* csio_hw_get_device_id */
+
+/*
+ * csio_hw_set_description - Set the model, description of the hw.
+ * @hw: HW module.
+ * @ven_id: PCI Vendor ID
+ * @dev_id: PCI Device ID
+ */
+static void
+csio_hw_set_description(struct csio_hw *hw, uint16_t ven_id, uint16_t dev_id)
+{
+ uint32_t adap_type, prot_type;
+
+ if (ven_id == CSIO_VENDOR_ID) {
+ prot_type = (dev_id & CSIO_ASIC_DEVID_PROTO_MASK);
+ adap_type = (dev_id & CSIO_ASIC_DEVID_TYPE_MASK);
+
+ if (prot_type == CSIO_FPGA) {
+ memcpy(hw->model_desc,
+ csio_fcoe_adapters[13].description, 32);
+ } else if (prot_type == CSIO_T4_FCOE_ASIC) {
+ memcpy(hw->hw_ver,
+ csio_fcoe_adapters[adap_type].model_no, 16);
+ memcpy(hw->model_desc,
+ csio_fcoe_adapters[adap_type].description, 32);
+ } else {
+ char tempName[32] = "Chelsio FCoE Controller";
+ memcpy(hw->model_desc, tempName, 32);
+
+ CSIO_DB_ASSERT(0);
+ }
+ }
+} /* csio_hw_set_description */
+
+/**
+ * csio_hw_init - Initialize HW module.
+ * @hw: Pointer to HW module.
+ *
+ * Initialize the members of the HW module.
+ */
+csio_retval_t
+csio_hw_init(struct csio_hw *hw)
+{
+ csio_retval_t rv = CSIO_INVAL;
+ uint32_t i;
+ uint16_t ven_id, dev_id;
+ struct csio_evt_msg *evt_entry;
+
+ INIT_LIST_HEAD(&hw->sm.sm_list);
+ csio_init_state(&hw->sm, csio_hws_uninit);
+ spin_lock_init(&hw->lock);
+ INIT_LIST_HEAD(&hw->sln_head);
+
+ /* Get the PCI vendor & device id */
+ csio_hw_get_device_id(hw);
+
+ strcpy(hw->name, CSIO_HW_NAME);
+
+ /* Set the model & its description */
+
+ ven_id = hw->params.pci.vendor_id;
+ dev_id = hw->params.pci.device_id;
+
+ csio_hw_set_description(hw, ven_id, dev_id);
+
+ /* Initialize default log level */
+ hw->params.log_level = (uint32_t) csio_dbg_level;
+
+ csio_set_fwevt_intr_idx(hw, -1);
+ csio_set_nondata_intr_idx(hw, -1);
+
+ /* Init all the modules: Mailbox, WorkRequest and Transport */
+ if (csio_mbm_init(csio_hw_to_mbm(hw), hw, csio_hw_mb_timer))
+ goto err;
+
+ rv = csio_wrm_init(csio_hw_to_wrm(hw), hw);
+ if (rv)
+ goto err_mbm_exit;
+
+ rv = csio_scsim_init(csio_hw_to_scsim(hw), hw);
+ if (rv)
+ goto err_wrm_exit;
+
+ rv = csio_mgmtm_init(csio_hw_to_mgmtm(hw), hw);
+ if (rv)
+ goto err_scsim_exit;
+ /* Pre-allocate evtq and initialize them */
+ INIT_LIST_HEAD(&hw->evt_active_q);
+ INIT_LIST_HEAD(&hw->evt_free_q);
+ for (i = 0; i < csio_evtq_sz; i++) {
+
+ evt_entry = kzalloc(sizeof(struct csio_evt_msg), GFP_KERNEL);
+ if (!evt_entry) {
+ csio_err(hw, "Failed to initialize eventq");
+ goto err_evtq_cleanup;
+ }
+
+ list_add_tail(&evt_entry->list, &hw->evt_free_q);
+ csio_inc_stats(hw, n_evt_freeq);
+ }
+
+ hw->dev_num = dev_num;
+ dev_num++;
+
+ return CSIO_SUCCESS;
+
+err_evtq_cleanup:
+ csio_evtq_cleanup(hw);
+ csio_mgmtm_exit(csio_hw_to_mgmtm(hw));
+err_scsim_exit:
+ csio_scsim_exit(csio_hw_to_scsim(hw));
+err_wrm_exit:
+ csio_wrm_exit(csio_hw_to_wrm(hw), hw);
+err_mbm_exit:
+ csio_mbm_exit(csio_hw_to_mbm(hw));
+err:
+ return rv;
+}
+
+/**
+ * csio_hw_exit - Un-initialize HW module.
+ * @hw: Pointer to HW module.
+ *
+ */
+void
+csio_hw_exit(struct csio_hw *hw)
+{
+ csio_evtq_cleanup(hw);
+ csio_mgmtm_exit(csio_hw_to_mgmtm(hw));
+ csio_scsim_exit(csio_hw_to_scsim(hw));
+ csio_wrm_exit(csio_hw_to_wrm(hw), hw);
+ csio_mbm_exit(csio_hw_to_mbm(hw));
+}
--
1.7.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH 2/8] csiostor: Chelsio FCoE offload driver submission (sources part 2).
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
2012-08-23 22:27 ` [PATCH 1/8] csiostor: Chelsio FCoE offload driver submission (sources part 1) Naresh Kumar Inna
@ 2012-08-23 22:27 ` Naresh Kumar Inna
2012-08-23 22:27 ` [PATCH 3/8] csiostor: Chelsio FCoE offload driver submission (sources part 3) Naresh Kumar Inna
` (7 subsequent siblings)
9 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
This patch contains code for driver initialization, driver resource
allocation, module parameters and the Work Request module functionality.
Driver initialization includes module entry/exit points, registration with
PCI, FC transport and SCSI mid layer subsystems. The Work Request module
provides services for allocation of DMA queues, posting Work Requests on
them and processing completions.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/csiostor/csio_init.c | 1392 +++++++++++++++++++++++++++++++
drivers/scsi/csiostor/csio_wr.c | 1634 +++++++++++++++++++++++++++++++++++++
2 files changed, 3026 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_init.c
create mode 100644 drivers/scsi/csiostor/csio_wr.c
diff --git a/drivers/scsi/csiostor/csio_init.c b/drivers/scsi/csiostor/csio_init.c
new file mode 100644
index 0000000..9be85fe
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_init.c
@@ -0,0 +1,1392 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/aer.h>
+#include <linux/mm.h>
+#include <linux/notifier.h>
+#include <linux/kdebug.h>
+#include <linux/version.h>
+#include <linux/seq_file.h>
+#include <linux/debugfs.h>
+#include <linux/string.h>
+#include <linux/export.h>
+
+#include "csio_init.h"
+#include "csio_defs.h"
+
+#define CSIO_MIN_MEMPOOL_SZ 64
+
+static int pf_counter;
+
+#define csio_param_check(name, def, min, max, typ) \
+static int \
+csio_##name##_check(typ val) \
+{ \
+ if (val >= min && val <= max) { \
+ csio_##name = val; \
+ return 0; \
+ } \
+ \
+ pr_warn("Setting csio_"#name" to default of "#def"\n"); \
+ csio_##name = def; \
+ return -EINVAL; \
+}
+
+#define CSIO_MODULE_PARAM(name, def, min, max, desc, typ) \
+module_param(csio_##name, typ, S_IRUGO|S_IWUSR); \
+MODULE_PARM_DESC(csio_##name, desc); \
+csio_param_check(name, def, min, max, typ)
+
+/*
+ * Normally we're willing to become the firmware's Master PF but will be happy
+ * if another PF has already become the Master. Setting "force_master" will
+ * cause this driver to forcibly establish itself as the Master PF.
+ */
+CSIO_MODULE_PARAM(force_master, 0, 0, 1,
+ " Forcibly become Master PF. 0 - No force, 1 - Force", int);
+
+/* Module parameter - Port vector mask for linkup */
+CSIO_MODULE_PARAM(port_mask, 0xf, 0x1, 0xf,
+ " Port vector mask. (4bits) Default - 0xf (all ports)", uint);
+
+/* Module parameter - Default Log level */
+CSIO_MODULE_PARAM(dbg_level, 0xFEFF, 0, 0xFFFF,
+ " Debug level. Default - 0xFEFF. Module[B15:B8]:Level[B7:B0]",
+ int);
+
+/* Module parameter - lun qdepth */
+CSIO_MODULE_PARAM(lun_qdepth, CSIO_MAX_CMD_PER_LUN, 1, 32,
+ " Lun queue depth (1-32), Default - 32", int);
+
+/* Module parameter - Total number of FCoE rnodes */
+CSIO_MODULE_PARAM(fcoe_rnodes, 512, 8, CSIO_FCOE_MAX_RNODES,
+ " Supported number of FCoE rnodes."
+ " Default - 512, Max - 2048.", int);
+
+/*
+ * Module parameter - Interrupt mode
+ * The driver uses the best interrupt scheme available on a platform in the
+ * order MSI-X, MSI, legacy INTx interrupts. This parameter determines
+ * which
+ * of these schemes the driver may consider as follows:
+ *
+ * csio_msi = 2: choose from among all three options
+ * csio_msi = 1: only consider MSI and INTx interrupts
+ * csio_msi = 0: force INTx interrupts
+ */
+CSIO_MODULE_PARAM(msi, 2, 0, 2, " whether to use MSI-X, MSI or INTx", int);
+
+/* Module parameter - Interrupt coalesce count */
+CSIO_MODULE_PARAM(intr_coalesce_cnt, 0, 0, 16,
+ " Interrupt coalesce count, Disable:0, Range:1-16, Default:0",
+ int);
+
+/* Module parameter - Interrupt coalesce time */
+CSIO_MODULE_PARAM(intr_coalesce_time, 10, 5, 200,
+ " Interrupt coalesce time (us), Min:5, Max:200,"
+ " Default: 10", int);
+
+/* Module parameter - Max scan timeout*/
+CSIO_MODULE_PARAM(max_scan_tmo, 0, 0, 30,
+ " Maximum time (seconds) to wait before declaring scan done.",
+ uint);
+
+/* Module parameter - Delta scan timeout*/
+CSIO_MODULE_PARAM(delta_scan_tmo, 5, 2, 5,
+ " Time (seconds) to wait between two FW scan events."
+ " Default - 5 seconds .", uint);
+
+/* Module parameter - Max SCSI IO request */
+CSIO_MODULE_PARAM(scsi_ioreqs, 2048, 64, 2048,
+ " Max IO req elements per function. Default - 2048.", int);
+
+/* Module parameter - FDMI support */
+CSIO_MODULE_PARAM(fdmi_enable, 1, 0, 1,
+ " fdmi support 1:enable 0:disable Default - 1", int);
+
+static struct dentry *csio_debugfs_root;
+
+static struct scsi_transport_template *csio_fcoe_transport;
+static struct scsi_transport_template *csio_fcoe_transport_vport;
+
+static void
+csio_module_params_check(void)
+{
+ csio_dbg_level_check(csio_dbg_level);
+ csio_force_master_check(csio_force_master);
+ csio_port_mask_check(csio_port_mask);
+ csio_lun_qdepth_check(csio_lun_qdepth);
+ csio_fcoe_rnodes_check(csio_fcoe_rnodes);
+ csio_msi_check(csio_msi);
+ csio_intr_coalesce_cnt_check(csio_intr_coalesce_cnt);
+ csio_intr_coalesce_time_check(csio_intr_coalesce_time);
+ csio_max_scan_tmo_check(csio_max_scan_tmo);
+ csio_delta_scan_tmo_check(csio_delta_scan_tmo);
+ csio_scsi_ioreqs_check(csio_scsi_ioreqs);
+ csio_fdmi_enable_check(csio_fdmi_enable);
+}
+
+/*
+ * debugfs support
+ */
+static int
+csio_mem_open(struct inode *inode, struct file *file)
+{
+ file->private_data = inode->i_private;
+ return 0;
+}
+
+static ssize_t
+csio_mem_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
+{
+ loff_t pos = *ppos;
+ loff_t avail = file->f_path.dentry->d_inode->i_size;
+ unsigned int mem = (uintptr_t)file->private_data & 3;
+ struct csio_hw *hw = file->private_data - mem;
+
+ if (pos < 0)
+ return -EINVAL;
+ if (pos >= avail)
+ return 0;
+ if (count > avail - pos)
+ count = avail - pos;
+
+ while (count) {
+ size_t len;
+ int ret, ofst;
+ __be32 data[16];
+
+ if (mem == MEM_MC)
+ ret = csio_hw_mc_read(hw, pos, data, NULL);
+ else
+ ret = csio_hw_edc_read(hw, mem, pos, data, NULL);
+ if (ret)
+ return ret;
+
+ ofst = pos % sizeof(data);
+ len = min(count, sizeof(data) - ofst);
+ if (copy_to_user(buf, (u8 *)data + ofst, len))
+ return -EFAULT;
+
+ buf += len;
+ pos += len;
+ count -= len;
+ }
+ count = pos - *ppos;
+ *ppos = pos;
+ return count;
+}
+
+static const struct file_operations csio_mem_debugfs_fops = {
+ .owner = THIS_MODULE,
+ .open = csio_mem_open,
+ .read = csio_mem_read,
+ .llseek = default_llseek,
+};
+
+static void __devinit
+csio_add_debugfs_mem(struct csio_hw *hw, const char *name,
+ unsigned int idx, unsigned int size_mb)
+{
+ struct dentry *de;
+
+ de = debugfs_create_file(name, S_IRUSR, hw->debugfs_root,
+ (void *)hw + idx, &csio_mem_debugfs_fops);
+ if (de && de->d_inode)
+ de->d_inode->i_size = size_mb << 20;
+}
+
+static int __devinit
+csio_setup_debugfs(struct csio_hw *hw)
+{
+ int i;
+
+ if (IS_ERR_OR_NULL(hw->debugfs_root))
+ return -1;
+
+ i = csio_rd_reg32(hw, MA_TARGET_MEM_ENABLE);
+ if (i & EDRAM0_ENABLE)
+ csio_add_debugfs_mem(hw, "edc0", MEM_EDC0, 5);
+ if (i & EDRAM1_ENABLE)
+ csio_add_debugfs_mem(hw, "edc1", MEM_EDC1, 5);
+ if (i & EXT_MEM_ENABLE)
+ csio_add_debugfs_mem(hw, "mc", MEM_MC,
+ EXT_MEM_SIZE_GET(csio_rd_reg32(hw, MA_EXT_MEMORY_BAR)));
+ return 0;
+}
+
+/*
+ * csio_dfs_create - Creates debug filesystem and proc fs for the given HW.
+ *
+ */
+static int
+csio_dfs_create(struct csio_hw *hw)
+{
+ if (csio_debugfs_root) {
+ hw->debugfs_root = debugfs_create_dir(pci_name(hw->pdev),
+ csio_debugfs_root);
+ csio_setup_debugfs(hw);
+ }
+
+ return 0;
+}
+
+/*
+ * csio_dfs_destroy - Deletes debugfs and procfs entries for the given HW.
+ *
+ */
+static int
+csio_dfs_destroy(struct csio_hw *hw)
+{
+ if (hw->debugfs_root)
+ debugfs_remove_recursive(hw->debugfs_root);
+
+ return 0;
+}
+
+/*
+ * csio_dfs_init - Debug filesystem initialization.
+ *
+ * This is function is called during driver load to initialize debugfs, procfs
+ * used for debugging.
+ */
+static int
+csio_dfs_init(void)
+{
+ csio_debugfs_root = debugfs_create_dir(KBUILD_MODNAME, NULL);
+ if (!csio_debugfs_root)
+ pr_warn("Could not create debugfs entry, continuing\n");
+
+ return 0;
+}
+
+/*
+ * csio_dfs_exit - Cleans up debugfs and procfs created during driver load.
+ * Function that gets called in the unload path.
+ */
+static void
+csio_dfs_exit(void)
+{
+ debugfs_remove(csio_debugfs_root);
+}
+
+/*
+ * csio_pci_init - PCI initialization.
+ * @pdev: PCI device.
+ * @bars: Bitmask of bars to be requested.
+ *
+ * Initializes the PCI function by enabling MMIO, setting bus
+ * mastership and setting DMA mask.
+ */
+static int
+csio_pci_init(struct pci_dev *pdev, int *bars)
+{
+ int rv = -ENODEV;
+
+ *bars = pci_select_bars(pdev, IORESOURCE_MEM);
+
+ if (pci_enable_device_mem(pdev))
+ goto err;
+
+ if (pci_request_selected_regions(pdev, *bars, KBUILD_MODNAME))
+ goto err_disable_device;
+
+ pci_set_master(pdev);
+ pci_try_set_mwi(pdev);
+
+ if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
+ pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));
+ } else if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(32))) {
+ pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
+ } else {
+ dev_err(&pdev->dev, "No suitable DMA available.\n");
+ goto err_release_regions;
+ }
+
+ return 0;
+
+err_release_regions:
+ pci_release_selected_regions(pdev, *bars);
+err_disable_device:
+ pci_disable_device(pdev);
+err:
+ return rv;
+
+}
+
+/*
+ * csio_pci_exit - PCI unitialization.
+ * @pdev: PCI device.
+ * @bars: Bars to be released.
+ *
+ */
+static void
+csio_pci_exit(struct pci_dev *pdev, int *bars)
+{
+ pci_release_selected_regions(pdev, *bars);
+ pci_disable_device(pdev);
+}
+
+/*
+ * csio_hw_init_workers - Initialize the HW module's worker threads.
+ * @hw: HW module.
+ *
+ */
+static void
+csio_hw_init_workers(struct csio_hw *hw)
+{
+ INIT_WORK(&hw->evtq_work, csio_evtq_worker);
+}
+
+static void
+csio_hw_exit_workers(struct csio_hw *hw)
+{
+ cancel_work_sync(&hw->evtq_work);
+ flush_scheduled_work();
+}
+
+static csio_retval_t
+csio_create_queues(struct csio_hw *hw)
+{
+ int i, j;
+ struct csio_mgmtm *mgmtm = csio_hw_to_mgmtm(hw);
+ csio_retval_t rv;
+ struct csio_scsi_cpu_info *info;
+
+ if (hw->flags & CSIO_HWF_Q_FW_ALLOCED)
+ return CSIO_SUCCESS;
+
+ if (hw->intr_mode != CSIO_IM_MSIX) {
+ rv = csio_wr_iq_create(hw, NULL, hw->intr_iq_idx,
+ 0, hw->pport[0].portid,
+ CSIO_FALSE, NULL);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, " Forward Interrupt IQ failed!: %d\n", rv);
+ return rv;
+ }
+ }
+
+ /* FW event queue */
+ rv = csio_wr_iq_create(hw, NULL, hw->fwevt_iq_idx,
+ csio_get_fwevt_intr_idx(hw),
+ hw->pport[0].portid, CSIO_TRUE, NULL);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, "FW event IQ config failed!: %d\n", rv);
+ return rv;
+ }
+
+ /* Create mgmt queue */
+ rv = csio_wr_eq_create(hw, NULL, mgmtm->eq_idx,
+ mgmtm->iq_idx, hw->pport[0].portid, NULL);
+
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw, "Mgmt EQ create failed!: %d\n", rv);
+ goto err;
+ }
+
+ /* Create SCSI queues */
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+
+ for (j = 0; j < info->max_cpus; j++) {
+ struct csio_scsi_qset *sqset = &hw->sqset[i][j];
+
+ rv = csio_wr_iq_create(hw, NULL, sqset->iq_idx,
+ sqset->intr_idx, i,
+ CSIO_FALSE, NULL);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw,
+ "SCSI module IQ config failed [%d][%d]:%d\n",
+ i, j, rv);
+ goto err;
+ }
+ rv = csio_wr_eq_create(hw, NULL, sqset->eq_idx,
+ sqset->iq_idx, i, NULL);
+ if (rv != CSIO_SUCCESS) {
+ csio_err(hw,
+ "SCSI module EQ config failed [%d][%d]:%d\n",
+ i, j, rv);
+ goto err;
+ }
+ } /* for all CPUs */
+ } /* For all ports */
+
+ hw->flags |= CSIO_HWF_Q_FW_ALLOCED;
+ return CSIO_SUCCESS;
+err:
+ csio_wr_destroy_queues(hw, CSIO_TRUE);
+ return CSIO_INVAL;
+}
+
+/*
+ * csio_config_queues - Configure the DMA queues.
+ * @hw: HW module.
+ *
+ * Allocates memory for queues are registers them with FW.
+ */
+csio_retval_t
+csio_config_queues(struct csio_hw *hw)
+{
+ int i, j, idx, k = 0;
+ csio_retval_t rv;
+ struct csio_scsi_qset *sqset;
+ struct csio_mgmtm *mgmtm = csio_hw_to_mgmtm(hw);
+ struct csio_scsi_qset *orig;
+ struct csio_scsi_cpu_info *info;
+
+ if (hw->flags & CSIO_HWF_Q_MEM_ALLOCED)
+ return csio_create_queues(hw);
+
+ /* Calculate number of SCSI queues for MSIX we would like */
+ hw->num_scsi_msix_cpus = num_online_cpus();
+ hw->num_sqsets = num_online_cpus() * hw->num_pports;
+
+ if (hw->num_sqsets > CSIO_MAX_SCSI_QSETS) {
+ hw->num_sqsets = CSIO_MAX_SCSI_QSETS;
+ hw->num_scsi_msix_cpus = CSIO_MAX_SCSI_CPU;
+ }
+
+ /* Initialize max_cpus, may get reduced during msix allocations */
+ for (i = 0; i < hw->num_pports; i++)
+ hw->scsi_cpu_info[i].max_cpus = hw->num_scsi_msix_cpus;
+
+ csio_dbg(hw, "nsqsets:%d scpus:%d\n",
+ hw->num_sqsets, hw->num_scsi_msix_cpus);
+
+ csio_intr_enable(hw);
+
+ if (hw->intr_mode != CSIO_IM_MSIX) {
+
+ /* Allocate Forward interrupt iq. */
+ hw->intr_iq_idx = csio_wr_alloc_q(hw, CSIO_INTR_IQSIZE,
+ CSIO_INTR_WRSIZE, CSIO_INGRESS,
+ (void *)hw, 0, 0, NULL);
+ if (hw->intr_iq_idx == -1) {
+ csio_err(hw,
+ "Forward interrupt queue creation failed\n");
+ goto intr_disable;
+ }
+ }
+
+ /* Allocate the FW evt queue */
+ hw->fwevt_iq_idx = csio_wr_alloc_q(hw, CSIO_FWEVT_IQSIZE,
+ CSIO_FWEVT_WRSIZE,
+ CSIO_INGRESS, (void *)hw,
+ CSIO_FWEVT_FLBUFS, 0,
+ csio_fwevt_intx_handler);
+ if (hw->fwevt_iq_idx == -1) {
+ csio_err(hw, "FW evt queue creation failed\n");
+ goto intr_disable;
+ }
+
+ /* Allocate the mgmt queue */
+ mgmtm->eq_idx = csio_wr_alloc_q(hw, CSIO_MGMT_EQSIZE,
+ CSIO_MGMT_EQ_WRSIZE,
+ CSIO_EGRESS, (void *)hw, 0, 0, NULL);
+ if (mgmtm->eq_idx == -1) {
+ csio_err(hw, "Failed to alloc egress queue for mgmt module\n");
+ goto intr_disable;
+ }
+
+ /* Use FW IQ for MGMT req completion */
+ mgmtm->iq_idx = hw->fwevt_iq_idx;
+
+ /* Allocate SCSI queues */
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+
+ for (j = 0; j < hw->num_scsi_msix_cpus; j++) {
+ sqset = &hw->sqset[i][j];
+
+ if (j >= info->max_cpus) {
+ k = j % info->max_cpus;
+ orig = &hw->sqset[i][k];
+ sqset->eq_idx = orig->eq_idx;
+ sqset->iq_idx = orig->iq_idx;
+ continue;
+ }
+
+ idx = csio_wr_alloc_q(hw, csio_scsi_eqsize, 0,
+ CSIO_EGRESS, (void *)hw, 0, 0,
+ NULL);
+ if (idx == -1) {
+ csio_err(hw, "EQ creation failed for idx:%d\n",
+ idx);
+ goto intr_disable;
+ }
+
+ sqset->eq_idx = idx;
+
+ idx = csio_wr_alloc_q(hw, CSIO_SCSI_IQSIZE,
+ CSIO_SCSI_IQ_WRSZ, CSIO_INGRESS,
+ (void *)hw, 0, 0,
+ csio_scsi_intx_handler);
+ if (idx == -1) {
+ csio_err(hw, "IQ creation failed for idx:%d\n",
+ idx);
+ goto intr_disable;
+ }
+ sqset->iq_idx = idx;
+ } /* for all CPUs */
+ } /* For all ports */
+
+ hw->flags |= CSIO_HWF_Q_MEM_ALLOCED;
+
+ rv = csio_create_queues(hw);
+ if (rv != CSIO_SUCCESS)
+ goto intr_disable;
+
+ /*
+ * Now request IRQs for the vectors. In the event of a failure,
+ * cleanup is handled internally by this function.
+ */
+ rv = csio_request_irqs(hw);
+ if (rv != CSIO_SUCCESS)
+ return CSIO_INVAL;
+
+ return CSIO_SUCCESS;
+
+intr_disable:
+ csio_intr_disable(hw, CSIO_FALSE);
+
+ return CSIO_INVAL;
+}
+
+static int
+csio_resource_alloc(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ int rv = -ENOMEM;
+
+ wrm->num_q = ((CSIO_MAX_SCSI_QSETS * 2) + CSIO_HW_NIQ +
+ CSIO_HW_NEQ + CSIO_HW_NFLQ + CSIO_HW_NINTXQ);
+
+ hw->mb_mempool = mempool_create_kmalloc_pool(CSIO_MIN_MEMPOOL_SZ,
+ sizeof(struct csio_mb));
+ if (!hw->mb_mempool)
+ goto err;
+
+ hw->rnode_mempool = mempool_create_kmalloc_pool(CSIO_MIN_MEMPOOL_SZ,
+ sizeof(struct csio_rnode));
+ if (!hw->rnode_mempool)
+ goto err_free_mb_mempool;
+
+ hw->scsi_pci_pool = pci_pool_create("csio_scsi_pci_pool", hw->pdev,
+ sizeof(struct csio_fcp_resp), 8, 0);
+ if (!hw->scsi_pci_pool)
+ goto err_free_rn_pool;
+
+ return 0;
+
+err_free_rn_pool:
+ mempool_destroy(hw->rnode_mempool);
+ hw->rnode_mempool = NULL;
+err_free_mb_mempool:
+ mempool_destroy(hw->mb_mempool);
+ hw->mb_mempool = NULL;
+err:
+ return rv;
+}
+
+static void
+csio_resource_free(struct csio_hw *hw)
+{
+ pci_pool_destroy(hw->scsi_pci_pool);
+ hw->scsi_pci_pool = NULL;
+ mempool_destroy(hw->rnode_mempool);
+ hw->rnode_mempool = NULL;
+ mempool_destroy(hw->mb_mempool);
+ hw->mb_mempool = NULL;
+}
+
+/*
+ * csio_hw_alloc - Allocate and initialize the HW module.
+ * @pdev: PCI device.
+ *
+ * Allocates HW structure, DMA, memory resources, maps BARS to
+ * host memory and initializes HW module.
+ */
+static struct csio_hw * __devinit
+csio_hw_alloc(struct pci_dev *pdev)
+{
+ struct csio_hw *hw;
+
+ hw = kzalloc(sizeof(struct csio_hw), GFP_KERNEL);
+ if (!hw)
+ goto err;
+
+ hw->pdev = pdev;
+ strncpy(hw->drv_version, CSIO_DRV_VERSION, 32);
+
+ /* memory pool/DMA pool allocation */
+ if (csio_resource_alloc(hw))
+ goto err_free_hw;
+
+ /* Get the start address of registers from BAR 0 */
+ hw->regstart = ioremap_nocache(pci_resource_start(pdev, 0),
+ pci_resource_len(pdev, 0));
+ if (!hw->regstart) {
+ csio_err(hw, "Could not map BAR 0, regstart = %p\n",
+ hw->regstart);
+ goto err_resource_free;
+ }
+
+ csio_hw_init_workers(hw);
+
+ if (csio_hw_init(hw))
+ goto err_unmap_bar;
+
+ csio_dfs_create(hw);
+
+ csio_dbg(hw, "hw:%p\n", hw);
+
+ return hw;
+
+err_unmap_bar:
+ csio_hw_exit_workers(hw);
+ iounmap(hw->regstart);
+err_resource_free:
+ csio_resource_free(hw);
+err_free_hw:
+ kfree(hw);
+err:
+ return NULL;
+}
+
+/*
+ * csio_hw_free - Uninitialize and free the HW module.
+ * @hw: The HW module
+ *
+ * Disable interrupts, uninit the HW module, free resources, free hw.
+ */
+static void
+csio_hw_free(struct csio_hw *hw)
+{
+ csio_intr_disable(hw, CSIO_TRUE);
+ csio_hw_exit_workers(hw);
+ csio_hw_exit(hw);
+ iounmap(hw->regstart);
+ csio_dfs_destroy(hw);
+ csio_resource_free(hw);
+ kfree(hw);
+}
+
+/**
+ * csio_shost_init - Create and initialize the lnode module.
+ * @hw: The HW module.
+ * @dev: The device associated with this invocation.
+ * @probe: Called from probe context or not?
+ * @os_pln: Parent lnode if any.
+ *
+ * Allocates lnode structure via scsi_host_alloc, initializes
+ * shost, initializes lnode module and registers with SCSI ML
+ * via scsi_host_add. This function is shared between physical and
+ * virtual node ports.
+ */
+struct csio_lnode *
+csio_shost_init(struct csio_hw *hw, struct device *dev,
+ bool probe, struct csio_lnode *pln)
+{
+ struct Scsi_Host *shost = NULL;
+ struct csio_lnode *ln;
+
+ csio_fcoe_shost_template.cmd_per_lun = csio_lun_qdepth;
+ csio_fcoe_shost_vport_template.cmd_per_lun = csio_lun_qdepth;
+
+ /*
+ * hw->pdev is the physical port's PCI dev structure,
+ * which will be different from the NPIV dev structure.
+ */
+ if (dev == &hw->pdev->dev)
+ shost = scsi_host_alloc(
+ &csio_fcoe_shost_template,
+ sizeof(struct csio_lnode));
+ else
+ shost = scsi_host_alloc(
+ &csio_fcoe_shost_vport_template,
+ sizeof(struct csio_lnode));
+
+ if (!shost)
+ goto err;
+
+ ln = shost_priv(shost);
+ memset(ln, 0, sizeof(struct csio_lnode));
+
+ /* Link common lnode to this lnode */
+ ln->dev_num = (shost->host_no << 16);
+
+ shost->can_queue = CSIO_MAX_QUEUE;
+ shost->this_id = -1;
+ shost->unique_id = shost->host_no;
+ shost->max_cmd_len = 16; /* Max CDB length supported */
+ shost->max_id = min_t(uint32_t, csio_fcoe_rnodes,
+ hw->fres_info.max_ssns);
+ shost->max_lun = CSIO_MAX_LUN;
+ if (dev == &hw->pdev->dev)
+ shost->transportt = csio_fcoe_transport;
+ else
+ shost->transportt = csio_fcoe_transport_vport;
+
+ /* root lnode */
+ if (!hw->rln)
+ hw->rln = ln;
+
+ /* Other initialization here: Common, Transport specific */
+ if (csio_lnode_init(ln, hw, pln))
+ goto err_shost_put;
+
+ if (scsi_add_host(shost, dev))
+ goto err_lnode_exit;
+
+ return ln;
+
+err_lnode_exit:
+ csio_lnode_exit(ln);
+err_shost_put:
+ scsi_host_put(shost);
+err:
+ return NULL;
+}
+
+/**
+ * csio_shost_exit - De-instantiate the shost.
+ * @ln: The lnode module corresponding to the shost.
+ *
+ */
+void
+csio_shost_exit(struct csio_lnode *ln)
+{
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ /* Inform transport */
+ fc_remove_host(shost);
+
+ /* Inform SCSI ML */
+ scsi_remove_host(shost);
+
+ /* Flush all the events, so that any rnode removal events
+ * already queued are all handled, before we remove the lnode.
+ */
+ spin_lock_irq(&hw->lock);
+ csio_evtq_flush(hw);
+ spin_unlock_irq(&hw->lock);
+
+ csio_lnode_exit(ln);
+ scsi_host_put(shost);
+}
+
+struct csio_lnode *
+csio_lnode_alloc(struct csio_hw *hw)
+{
+ return csio_shost_init(hw, &hw->pdev->dev, CSIO_FALSE, NULL);
+}
+
+void
+csio_lnodes_block_request(struct csio_hw *hw)
+{
+ struct Scsi_Host *shost;
+ struct csio_lnode *sln;
+ struct csio_lnode *ln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "Failed to allocate lnodes_list");
+ return;
+ }
+
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ lnode_list[cur_cnt++] = sln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "Blocking IOs on lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ shost = csio_ln_to_shost(ln);
+ scsi_block_requests(shost);
+
+ }
+ kfree(lnode_list);
+}
+
+void
+csio_lnodes_unblock_request(struct csio_hw *hw)
+{
+ struct csio_lnode *ln;
+ struct Scsi_Host *shost;
+ struct csio_lnode *sln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "Failed to allocate lnodes_list");
+ return;
+ }
+
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ lnode_list[cur_cnt++] = sln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "unblocking IOs on lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ shost = csio_ln_to_shost(ln);
+ scsi_unblock_requests(shost);
+ }
+ kfree(lnode_list);
+}
+
+void
+csio_lnodes_block_by_port(struct csio_hw *hw, uint8_t portid)
+{
+ struct csio_lnode *ln;
+ struct Scsi_Host *shost;
+ struct csio_lnode *sln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "Failed to allocate lnodes_list");
+ return;
+ }
+
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ if (sln->portid != portid)
+ continue;
+
+ lnode_list[cur_cnt++] = sln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "Blocking IOs on lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ shost = csio_ln_to_shost(ln);
+ scsi_block_requests(shost);
+ }
+ kfree(lnode_list);
+}
+
+void
+csio_lnodes_unblock_by_port(struct csio_hw *hw, uint8_t portid)
+{
+ struct csio_lnode *ln;
+ struct Scsi_Host *shost;
+ struct csio_lnode *sln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "Failed to allocate lnodes_list");
+ return;
+ }
+
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ if (sln->portid != portid)
+ continue;
+ lnode_list[cur_cnt++] = sln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "unblocking IOs on lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ shost = csio_ln_to_shost(ln);
+ scsi_unblock_requests(shost);
+ }
+ kfree(lnode_list);
+}
+
+void
+csio_lnodes_exit(struct csio_hw *hw, bool npiv)
+{
+ struct csio_lnode *sln;
+ struct csio_lnode *ln;
+ struct list_head *cur_ln, *cur_cln;
+ struct csio_lnode **lnode_list;
+ int cur_cnt = 0, ii;
+
+ lnode_list = kzalloc((sizeof(struct csio_lnode *) * hw->num_lns),
+ GFP_KERNEL);
+ if (!lnode_list) {
+ csio_err(hw, "lnodes_exit: Failed to allocate lnodes_list.\n");
+ return;
+ }
+
+ /* Get all child lnodes(NPIV ports) */
+ spin_lock_irq(&hw->lock);
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+
+ /* Traverse children lnodes */
+ list_for_each(cur_cln, &sln->cln_head)
+ lnode_list[cur_cnt++] = (struct csio_lnode *) cur_cln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ /* Delete NPIV lnodes */
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "Deleting child lnode: %p\n", lnode_list[ii]);
+ ln = lnode_list[ii];
+ fc_vport_terminate(ln->fc_vport);
+ }
+
+ /* Delete only npiv lnodes */
+ if (npiv)
+ goto free_lnodes;
+
+ cur_cnt = 0;
+ /* Get all physical lnodes */
+ spin_lock_irq(&hw->lock);
+ /* Traverse sibling lnodes */
+ list_for_each(cur_ln, &hw->sln_head) {
+ sln = (struct csio_lnode *) cur_ln;
+ lnode_list[cur_cnt++] = sln;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ /* Delete physical lnodes */
+ for (ii = 0; ii < cur_cnt; ii++) {
+ csio_dbg(hw, "Deleting parent lnode: %p\n", lnode_list[ii]);
+ csio_shost_exit(lnode_list[ii]);
+ }
+
+free_lnodes:
+ kfree(lnode_list);
+}
+
+/*
+ * csio_lnode_init_post: Set lnode attributes after starting HW.
+ * @ln: lnode.
+ *
+ */
+static void
+csio_lnode_init_post(struct csio_lnode *ln)
+{
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+
+ csio_fchost_attr_init(ln);
+
+ scsi_scan_host(shost);
+}
+
+/*
+ * csio_probe_one - Instantiate this function.
+ * @pdev: PCI device
+ * @id: Device ID
+ *
+ * This is the .probe() callback of the driver. This function:
+ * - Initializes the PCI function by enabling MMIO, setting bus
+ * mastership and setting DMA mask.
+ * - Allocates HW structure, DMA, memory resources, maps BARS to
+ * host memory and initializes HW module.
+ * - Allocates lnode structure via scsi_host_alloc, initializes
+ * shost, initialized lnode module and registers with SCSI ML
+ * via scsi_host_add.
+ * - Enables interrupts, and starts the chip by kicking off the
+ * HW state machine.
+ * - Once hardware is ready, initiated scan of the host via
+ * scsi_scan_host.
+ */
+static int __devinit
+csio_probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
+{
+ int rv;
+ int bars;
+ int i;
+ struct csio_hw *hw;
+ struct csio_lnode *ln;
+
+ if (pf_counter > 0 && (CSIO_IS_T4_FPGA(pdev->device))) {
+ dev_err(&pdev->dev, "Driver claims only one FPGA function\n");
+ return -ENODEV;
+ }
+
+ rv = csio_pci_init(pdev, &bars);
+ if (rv)
+ goto err;
+
+ hw = csio_hw_alloc(pdev);
+ if (!hw) {
+ rv = -ENODEV;
+ goto err_pci_exit;
+ }
+
+ pci_set_drvdata(pdev, hw);
+
+ if (csio_hw_start(hw) != CSIO_SUCCESS) {
+ dev_err(&pdev->dev,
+ "Failed to start FW, continuing in debug mode.\n");
+ return 0;
+ }
+
+ sprintf(hw->fwrev_str, "%u.%u.%u.%u\n",
+ FW_HDR_FW_VER_MAJOR_GET(hw->fwrev),
+ FW_HDR_FW_VER_MINOR_GET(hw->fwrev),
+ FW_HDR_FW_VER_MICRO_GET(hw->fwrev),
+ FW_HDR_FW_VER_BUILD_GET(hw->fwrev));
+
+ for (i = 0; i < hw->num_pports; i++) {
+ ln = csio_shost_init(hw, &pdev->dev, CSIO_TRUE, NULL);
+ if (!ln) {
+ rv = -ENODEV;
+ break;
+ }
+ /* Initialize portid */
+ ln->portid = hw->pport[i].portid;
+
+ spin_lock_irq(&hw->lock);
+ if (csio_lnode_start(ln) != CSIO_SUCCESS)
+ rv = -ENODEV;
+ spin_unlock_irq(&hw->lock);
+
+ if (rv)
+ break;
+
+ csio_lnode_init_post(ln);
+ }
+
+ if (rv)
+ goto err_lnode_exit;
+
+ pf_counter++;
+
+ return 0;
+
+err_lnode_exit:
+ csio_lnodes_block_request(hw);
+ spin_lock_irq(&hw->lock);
+ csio_hw_stop(hw);
+ spin_unlock_irq(&hw->lock);
+ csio_lnodes_unblock_request(hw);
+ pci_set_drvdata(hw->pdev, NULL);
+ csio_lnodes_exit(hw, 0);
+ csio_hw_free(hw);
+err_pci_exit:
+ csio_pci_exit(pdev, &bars);
+err:
+ dev_err(&pdev->dev, "probe of device failed: %d\n", rv);
+ return rv;
+}
+
+/*
+ * csio_remove_one - Remove one instance of the driver at this PCI function.
+ * @pdev: PCI device
+ *
+ * Used during hotplug operation.
+ */
+static void __devexit
+csio_remove_one(struct pci_dev *pdev)
+{
+ struct csio_hw *hw = pci_get_drvdata(pdev);
+ int bars = pci_select_bars(pdev, IORESOURCE_MEM);
+
+ csio_lnodes_block_request(hw);
+ spin_lock_irq(&hw->lock);
+
+ /* Stops lnode, Rnode s/m
+ * Quiesce IOs.
+ * All sessions with remote ports are unregistered.
+ */
+ csio_hw_stop(hw);
+ spin_unlock_irq(&hw->lock);
+ csio_lnodes_unblock_request(hw);
+
+ csio_lnodes_exit(hw, 0);
+ csio_hw_free(hw);
+ pci_set_drvdata(pdev, NULL);
+ csio_pci_exit(pdev, &bars);
+}
+
+/*
+ * csio_pci_error_detected - PCI error was detected
+ * @pdev: PCI device
+ *
+ */
+static pci_ers_result_t
+csio_pci_error_detected(struct pci_dev *pdev, pci_channel_state_t state)
+{
+ struct csio_hw *hw = pci_get_drvdata(pdev);
+
+ csio_lnodes_block_request(hw);
+ spin_lock_irq(&hw->lock);
+
+ /* Post PCI error detected evt to HW s/m
+ * HW s/m handles this evt by quiescing IOs, unregisters rports
+ * and finally takes the device to offline.
+ */
+ csio_post_event(&hw->sm, CSIO_HWE_PCIERR_DETECTED);
+ spin_unlock_irq(&hw->lock);
+ csio_lnodes_unblock_request(hw);
+ csio_lnodes_exit(hw, 0);
+ csio_intr_disable(hw, CSIO_TRUE);
+ pci_disable_device(pdev);
+ return state == pci_channel_io_perm_failure ?
+ PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_NEED_RESET;
+}
+
+/*
+ * csio_pci_slot_reset - PCI slot has been reset.
+ * @pdev: PCI device
+ *
+ */
+static pci_ers_result_t
+csio_pci_slot_reset(struct pci_dev *pdev)
+{
+ struct csio_hw *hw = pci_get_drvdata(pdev);
+
+ if (pci_enable_device(pdev)) {
+ dev_err(&pdev->dev, "cannot re-enable device in slot reset\n");
+ return PCI_ERS_RESULT_DISCONNECT;
+ }
+
+ pci_set_master(pdev);
+ pci_restore_state(pdev);
+ pci_save_state(pdev);
+ pci_cleanup_aer_uncorrect_error_status(pdev);
+
+ /* Bring HW s/m to ready state.
+ * but don't resume IOs.
+ */
+ spin_lock_irq(&hw->lock);
+ csio_post_event(&hw->sm, CSIO_HWE_PCIERR_SLOT_RESET);
+ if (!csio_is_hw_ready(hw)) {
+ spin_unlock_irq(&hw->lock);
+ dev_err(&pdev->dev, "Can't initialize HW when in slot reset\n");
+ return PCI_ERS_RESULT_DISCONNECT;
+ }
+ spin_unlock_irq(&hw->lock);
+ return PCI_ERS_RESULT_RECOVERED;
+}
+
+/*
+ * csio_pci_resume - Resume normal operations
+ * @pdev: PCI device
+ *
+ */
+static void
+csio_pci_resume(struct pci_dev *pdev)
+{
+ struct csio_hw *hw = pci_get_drvdata(pdev);
+ struct csio_lnode *ln;
+ int rv = 0;
+ int i;
+
+ /* Bring the LINK UP and Resume IO */
+
+ for (i = 0; i < hw->num_pports; i++) {
+ ln = csio_shost_init(hw, &pdev->dev, CSIO_TRUE, NULL);
+ if (!ln) {
+ rv = -ENODEV;
+ break;
+ }
+ /* Initialize portid */
+ ln->portid = hw->pport[i].portid;
+
+ spin_lock_irq(&hw->lock);
+ if (csio_lnode_start(ln) != CSIO_SUCCESS)
+ rv = -ENODEV;
+ spin_unlock_irq(&hw->lock);
+
+ if (rv)
+ break;
+
+ csio_lnode_init_post(ln);
+ }
+
+ if (rv)
+ goto err_resume_exit;
+
+ return;
+
+err_resume_exit:
+ csio_lnodes_block_request(hw);
+ spin_lock_irq(&hw->lock);
+ csio_hw_stop(hw);
+ spin_unlock_irq(&hw->lock);
+ csio_lnodes_unblock_request(hw);
+ csio_lnodes_exit(hw, 0);
+ csio_hw_free(hw);
+ dev_err(&pdev->dev, "resume of device failed: %d\n", rv);
+}
+
+static struct pci_error_handlers csio_err_handler = {
+ .error_detected = csio_pci_error_detected,
+ .slot_reset = csio_pci_slot_reset,
+ .resume = csio_pci_resume,
+};
+
+static DEFINE_PCI_DEVICE_TABLE(csio_pci_tbl) = {
+ CSIO_DEVICE(CSIO_DEVID_T440DBG_FCOE, 0), /* T440DBG FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420CR_FCOE, 0), /* T420CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T422CR_FCOE, 0), /* T422CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T440CR_FCOE, 0), /* T440CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420BCH_FCOE, 0), /* T420BCH FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T440BCH_FCOE, 0), /* T440BCH FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T440CH_FCOE, 0), /* T440CH FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420SO_FCOE, 0), /* T420SO FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420CX_FCOE, 0), /* T420CX FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T420BT_FCOE, 0), /* T420BT FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T404BT_FCOE, 0), /* T404BT FCOE */
+ CSIO_DEVICE(CSIO_DEVID_B420_FCOE, 0), /* B420 FCOE */
+ CSIO_DEVICE(CSIO_DEVID_B404_FCOE, 0), /* B404 FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T480CR_FCOE, 0), /* T480 CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_T440LPCR_FCOE, 0), /* T440 LP-CR FCOE */
+ CSIO_DEVICE(CSIO_DEVID_PE10K, 0), /* PE10K FCOE */
+ CSIO_DEVICE(CSIO_DEVID_PE10K_PF1, 0), /* PE10K FCOE on PF1 */
+ { 0, 0, 0, 0, 0, 0, 0 }
+};
+
+
+static struct pci_driver csio_pci_driver = {
+ .name = KBUILD_MODNAME,
+ .driver = {
+ .owner = THIS_MODULE,
+ },
+ .id_table = csio_pci_tbl,
+ .probe = csio_probe_one,
+ .remove = csio_remove_one,
+ .err_handler = &csio_err_handler,
+};
+
+/*
+ * csio_init - Chelsio storage driver initialization function.
+ *
+ */
+static int __init
+csio_init(void)
+{
+ int rv = -ENOMEM;
+
+ pr_info("%s %s\n", CSIO_DRV_DESC, CSIO_DRV_VERSION);
+
+ csio_module_params_check();
+
+ csio_dfs_init();
+
+ csio_fcoe_transport = fc_attach_transport(&csio_fc_transport_funcs);
+ if (!csio_fcoe_transport)
+ goto err;
+
+ csio_fcoe_transport_vport =
+ fc_attach_transport(&csio_fc_transport_vport_funcs);
+ if (!csio_fcoe_transport_vport)
+ goto err_vport;
+
+ rv = pci_register_driver(&csio_pci_driver);
+ if (rv)
+ goto err_pci;
+
+ return 0;
+
+err_pci:
+ fc_release_transport(csio_fcoe_transport_vport);
+err_vport:
+ fc_release_transport(csio_fcoe_transport);
+err:
+ csio_dfs_exit();
+ return rv;
+}
+
+/*
+ * csio_exit - Chelsio storage driver uninitialization .
+ *
+ * Function that gets called in the unload path.
+ */
+static void __exit
+csio_exit(void)
+{
+ pci_unregister_driver(&csio_pci_driver);
+ csio_dfs_exit();
+ fc_release_transport(csio_fcoe_transport_vport);
+ fc_release_transport(csio_fcoe_transport);
+}
+
+module_init(csio_init);
+module_exit(csio_exit);
+MODULE_AUTHOR(CSIO_DRV_AUTHOR);
+MODULE_DESCRIPTION(CSIO_DRV_DESC);
+MODULE_LICENSE(CSIO_DRV_LICENSE);
+MODULE_DEVICE_TABLE(pci, csio_pci_tbl);
+MODULE_VERSION(CSIO_DRV_VERSION);
+MODULE_FIRMWARE(CSIO_FW_FNAME);
diff --git a/drivers/scsi/csiostor/csio_wr.c b/drivers/scsi/csiostor/csio_wr.c
new file mode 100644
index 0000000..dadd486
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_wr.c
@@ -0,0 +1,1634 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/compiler.h>
+#include <linux/slab.h>
+#include <asm/page.h>
+#include <linux/cache.h>
+
+#include "csio_hw.h"
+#include "csio_wr.h"
+#include "csio_mb.h"
+#include "csio_defs.h"
+
+int csio_intr_coalesce_cnt; /* value:SGE_INGRESS_RX_THRESHOLD[0] */
+static int csio_sge_thresh_reg; /* SGE_INGRESS_RX_THRESHOLD[0] */
+
+int csio_intr_coalesce_time = 10; /* value:SGE_TIMER_VALUE_1 */
+static int csio_sge_timer_reg = 1;
+
+#define csio_set_flbuf_size(_hw, _reg, _val) \
+ csio_wr_reg32((_hw), (_val), SGE_FL_BUFFER_SIZE##_reg)
+
+static void
+csio_get_flbuf_size(struct csio_hw *hw, struct csio_sge *sge, uint32_t reg)
+{
+ sge->sge_fl_buf_size[reg] = csio_rd_reg32(hw, SGE_FL_BUFFER_SIZE0 +
+ reg * sizeof(uint32_t));
+}
+
+/* Free list buffer size */
+static inline uint32_t
+csio_wr_fl_bufsz(struct csio_sge *sge, struct csio_dma_buf *buf)
+{
+ return sge->sge_fl_buf_size[buf->paddr & 0xF];
+}
+
+/* Size of the egress queue status page */
+static inline uint32_t
+csio_wr_qstat_pgsz(struct csio_hw *hw)
+{
+ return (hw->wrm.sge.sge_control & EGRSTATUSPAGESIZE(1)) ? 128 : 64;
+}
+
+/* Ring freelist doorbell */
+static inline void
+csio_wr_ring_fldb(struct csio_hw *hw, struct csio_q *flq)
+{
+ /*
+ * Ring the doorbell only when we have atleast CSIO_QCREDIT_SZ
+ * number of bytes in the freelist queue. This translates to atleast
+ * 8 freelist buffer pointers (since each pointer is 8 bytes).
+ */
+ if (flq->inc_idx >= 8) {
+ csio_wr_reg32(hw, DBPRIO(1) | QID(flq->un.fl.flid) |
+ PIDX(flq->inc_idx / 8),
+ MYPF_REG(SGE_PF_KDOORBELL));
+ flq->inc_idx &= 7;
+ }
+}
+
+/* Write a 0 cidx increment value to enable SGE interrupts for this queue */
+static void
+csio_wr_sge_intr_enable(struct csio_hw *hw, uint16_t iqid)
+{
+ csio_wr_reg32(hw, CIDXINC(0) |
+ INGRESSQID(iqid) |
+ TIMERREG(X_TIMERREG_RESTART_COUNTER),
+ MYPF_REG(SGE_PF_GTS));
+}
+
+/*
+ * csio_wr_fill_fl - Populate the FL buffers of a FL queue.
+ * @hw: HW module.
+ * @flq: Freelist queue.
+ *
+ * Fill up freelist buffer entries with buffers of size specified
+ * in the size register.
+ *
+ */
+static csio_retval_t
+csio_wr_fill_fl(struct csio_hw *hw, struct csio_q *flq)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ __be64 *d = (__be64 *)(flq->vstart);
+ struct csio_dma_buf *buf = &flq->un.fl.bufs[0];
+ uint64_t paddr;
+ int sreg = flq->un.fl.sreg;
+ int n = flq->credits;
+
+ while (n--) {
+ buf->len = sge->sge_fl_buf_size[sreg];
+ buf->vaddr = pci_alloc_consistent(hw->pdev, buf->len,
+ &buf->paddr);
+ if (!buf->vaddr) {
+ csio_err(hw, "Could only fill %d buffers!\n", n + 1);
+ return CSIO_NOMEM;
+ }
+
+ paddr = buf->paddr | (sreg & 0xF);
+
+ *d++ = cpu_to_be64(paddr);
+ buf++;
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_update_fl -
+ * @hw: HW module.
+ * @flq: Freelist queue.
+ *
+ *
+ */
+static inline void
+csio_wr_update_fl(struct csio_hw *hw, struct csio_q *flq, uint16_t n)
+{
+
+ flq->inc_idx += n;
+ flq->pidx += n;
+ if (unlikely(flq->pidx >= flq->credits))
+ flq->pidx -= (uint16_t)flq->credits;
+
+ csio_inc_stats(flq, n_flq_refill);
+}
+
+/*
+ * csio_wr_alloc_q - Allocate a WR queue and initialize it.
+ * @hw: HW module
+ * @qsize: Size of the queue in bytes
+ * @wrsize: Since of WR in this queue, if fixed.
+ * @type: Type of queue (Ingress/Egress/Freelist)
+ * @owner: Module that owns this queue.
+ * @nflb: Number of freelist buffers for FL.
+ * @sreg: What is the FL buffer size register?
+ * @iq_int_handler: Ingress queue handler in INTx mode.
+ *
+ * This function allocates and sets up a queue for the caller
+ * of size qsize, aligned at the required boundary. This is subject to
+ * be free entries being available in the queue array. If one is found,
+ * it is initialized with the allocated queue, marked as being used (owner),
+ * and a handle returned to the caller in form of the queue's index
+ * into the q_arr array.
+ * If user has indicated a freelist (by specifying nflb > 0), create
+ * another queue (with its own index into q_arr) for the freelist. Allocate
+ * memory for DMA buffer metadata (vaddr, len etc). Save off the freelist
+ * idx in the ingress queue's flq.idx. This is how a Freelist is associated
+ * with its owning ingress queue.
+ */
+int
+csio_wr_alloc_q(struct csio_hw *hw, uint32_t qsize, uint32_t wrsize,
+ uint16_t type, void *owner, uint32_t nflb, int sreg,
+ iq_handler_t iq_intx_handler)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *q, *flq;
+ int free_idx = wrm->free_qidx;
+ int ret_idx = free_idx;
+ uint32_t qsz;
+ int flq_idx;
+
+ if (free_idx >= wrm->num_q) {
+ csio_err(hw, "No more free queues.\n");
+ return -1;
+ }
+
+ switch (type) {
+ case CSIO_EGRESS:
+ qsz = ALIGN(qsize, CSIO_QCREDIT_SZ) +
+ csio_wr_qstat_pgsz(hw);
+ break;
+ case CSIO_INGRESS:
+ switch (wrsize) {
+ case 16:
+ case 32:
+ case 64:
+ case 128:
+ break;
+ default:
+ csio_err(hw, "Invalid Ingress queue WR size:%d\n",
+ wrsize);
+ return -1;
+ }
+
+ /*
+ * Number of elements must be a multiple of 16
+ * So this includes status page size
+ */
+ qsz = ALIGN(qsize/wrsize, 16) * wrsize;
+
+ break;
+ case CSIO_FREELIST:
+ qsz = ALIGN(qsize/wrsize, 8) * wrsize +
+ csio_wr_qstat_pgsz(hw);
+ break;
+ default:
+ csio_err(hw, "Invalid queue type: 0x%x\n", type);
+ return -1;
+ }
+
+ q = wrm->q_arr[free_idx];
+
+ q->vstart = pci_alloc_consistent(hw->pdev, qsz, &q->pstart);
+ if (!q->vstart) {
+ csio_err(hw,
+ "Failed to allocate DMA memory for "
+ "queue at id: %d size: %d\n", free_idx, qsize);
+ return -1;
+ }
+
+ /*
+ * We need to zero out the contents, importantly for ingress,
+ * since we start with a generatiom bit of 1 for ingress.
+ */
+ memset(q->vstart, 0, qsz);
+
+ q->type = type;
+ q->owner = owner;
+ q->pidx = q->cidx = q->inc_idx = 0;
+ q->size = qsz;
+ q->wr_sz = wrsize; /* If using fixed size WRs */
+
+ wrm->free_qidx++;
+
+ if (type == CSIO_INGRESS) {
+ /* Since queue area is set to zero */
+ q->un.iq.genbit = 1;
+
+ /*
+ * Ingress queue status page size is always the size of
+ * the ingress queue entry.
+ */
+ q->credits = (qsz - q->wr_sz) / q->wr_sz;
+ q->vwrap = (void *)((uintptr_t)(q->vstart) + qsz
+ - q->wr_sz);
+
+ /* Allocate memory for FL if requested */
+ if (nflb > 0) {
+ flq_idx = csio_wr_alloc_q(hw, nflb * sizeof(__be64),
+ sizeof(__be64), CSIO_FREELIST,
+ owner, 0, sreg, NULL);
+ if (flq_idx == -1) {
+ csio_err(hw,
+ "Failed to allocate FL queue"
+ " for IQ idx:%d\n", free_idx);
+ return -1;
+ }
+
+ /* Associate the new FL with the Ingress quue */
+ q->un.iq.flq_idx = flq_idx;
+
+ flq = wrm->q_arr[q->un.iq.flq_idx];
+ flq->un.fl.bufs = kzalloc(flq->credits *
+ sizeof(struct csio_dma_buf),
+ GFP_KERNEL);
+ if (!flq->un.fl.bufs) {
+ csio_err(hw,
+ "Failed to allocate FL queue bufs"
+ " for IQ idx:%d\n", free_idx);
+ return -1;
+ }
+
+ flq->un.fl.packen = 0;
+ flq->un.fl.offset = 0;
+ flq->un.fl.sreg = sreg;
+
+ /* Fill up the free list buffers */
+ if (csio_wr_fill_fl(hw, flq))
+ return -1;
+
+ /*
+ * Make sure in a FLQ, atleast 1 credit (8 FL buffers)
+ * remains unpopulated,otherwise HW thinks
+ * FLQ is empty.
+ */
+ flq->pidx = flq->inc_idx = flq->credits - 8;
+ } else {
+ q->un.iq.flq_idx = -1;
+ }
+
+ /* Associate the IQ INTx handler. */
+ q->un.iq.iq_intx_handler = iq_intx_handler;
+
+ csio_q_iqid(hw, ret_idx) = CSIO_MAX_QID;
+
+ } else if (type == CSIO_EGRESS) {
+ q->credits = (qsz - csio_wr_qstat_pgsz(hw)) / CSIO_QCREDIT_SZ;
+ q->vwrap = (void *)((uintptr_t)(q->vstart) + qsz
+ - csio_wr_qstat_pgsz(hw));
+ csio_q_eqid(hw, ret_idx) = CSIO_MAX_QID;
+ } else { /* Freelist */
+ q->credits = (qsz - csio_wr_qstat_pgsz(hw)) / sizeof(__be64);
+ q->vwrap = (void *)((uintptr_t)(q->vstart) + qsz
+ - csio_wr_qstat_pgsz(hw));
+ csio_q_flid(hw, ret_idx) = CSIO_MAX_QID;
+ }
+
+ return ret_idx;
+}
+
+/*
+ * csio_wr_iq_create_rsp - Response handler for IQ creation.
+ * @hw: The HW module.
+ * @mbp: Mailbox.
+ * @iq_idx: Ingress queue that got created.
+ *
+ * Handle FW_IQ_CMD mailbox completion. Save off the assigned IQ/FL ids.
+ */
+static csio_retval_t
+csio_wr_iq_create_rsp(struct csio_hw *hw, struct csio_mb *mbp, int iq_idx)
+{
+ struct csio_iq_params iqp;
+ enum fw_retval retval;
+ uint32_t iq_id;
+ int flq_idx;
+
+ memset(&iqp, 0, sizeof(struct csio_iq_params));
+
+ csio_mb_iq_alloc_write_rsp(hw, mbp, &retval, &iqp);
+
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "IQ cmd returned 0x%x!\n", retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ csio_q_iqid(hw, iq_idx) = iqp.iqid;
+ csio_q_physiqid(hw, iq_idx) = iqp.physiqid;
+ csio_q_pidx(hw, iq_idx) = csio_q_cidx(hw, iq_idx) = 0;
+ csio_q_inc_idx(hw, iq_idx) = 0;
+
+ /* Actual iq-id. */
+ iq_id = iqp.iqid - hw->wrm.fw_iq_start;
+
+ /* Set the iq-id to iq map table. */
+ if (iq_id >= CSIO_MAX_IQ) {
+ csio_err(hw,
+ "Exceeding MAX_IQ(%d) supported!"
+ " iqid:%d rel_iqid:%d FW iq_start:%d\n",
+ CSIO_MAX_IQ, iq_id, iqp.iqid, hw->wrm.fw_iq_start);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+ csio_q_set_intr_map(hw, iq_idx, iq_id);
+
+ /*
+ * During FW_IQ_CMD, FW sets interrupt_sent bit to 1 in the SGE
+ * ingress context of this queue. This will block interrupts to
+ * this queue until the next GTS write. Therefore, we do a
+ * 0-cidx increment GTS write for this queue just to clear the
+ * interrupt_sent bit. This will re-enable interrupts to this
+ * queue.
+ */
+ csio_wr_sge_intr_enable(hw, iqp.physiqid);
+
+ flq_idx = csio_q_iq_flq_idx(hw, iq_idx);
+ if (flq_idx != -1) {
+ struct csio_q *flq = hw->wrm.q_arr[flq_idx];
+
+ csio_q_flid(hw, flq_idx) = iqp.fl0id;
+ csio_q_cidx(hw, flq_idx) = 0;
+ csio_q_pidx(hw, flq_idx) = csio_q_credits(hw, flq_idx) - 8;
+ csio_q_inc_idx(hw, flq_idx) = csio_q_credits(hw, flq_idx) - 8;
+
+ /* Now update SGE about the buffers allocated during init */
+ csio_wr_ring_fldb(hw, flq);
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_iq_create - Configure an Ingress queue with FW.
+ * @hw: The HW module.
+ * @priv: Private data object.
+ * @iq_idx: Ingress queue index in the WR module.
+ * @vec: MSIX vector.
+ * @portid: PCIE Channel to be associated with this queue.
+ * @async: Is this a FW asynchronous message handling queue?
+ * @cbfn: Completion callback.
+ *
+ * This API configures an ingress queue with FW by issuing a FW_IQ_CMD mailbox
+ * with alloc/write bits set.
+ */
+csio_retval_t
+csio_wr_iq_create(struct csio_hw *hw, void *priv, int iq_idx,
+ uint32_t vec, uint8_t portid, bool async,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct csio_mb *mbp;
+ struct csio_iq_params iqp;
+ int flq_idx;
+
+ memset(&iqp, 0, sizeof(struct csio_iq_params));
+ csio_q_portid(hw, iq_idx) = portid;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_err(hw, "IQ command out of memory!\n");
+ return CSIO_NOMEM;
+ }
+
+ switch (hw->intr_mode) {
+ case CSIO_IM_INTX:
+ case CSIO_IM_MSI:
+ /* For interrupt forwarding queue only */
+ if (hw->intr_iq_idx == iq_idx)
+ iqp.iqandst = X_INTERRUPTDESTINATION_PCIE;
+ else
+ iqp.iqandst = X_INTERRUPTDESTINATION_IQ;
+ iqp.iqandstindex =
+ csio_q_physiqid(hw, hw->intr_iq_idx);
+ break;
+ case CSIO_IM_MSIX:
+ iqp.iqandst = X_INTERRUPTDESTINATION_PCIE;
+ iqp.iqandstindex = (uint16_t)vec;
+ break;
+ case CSIO_IM_NONE:
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ /* Pass in the ingress queue cmd parameters */
+ iqp.pfn = hw->pfn;
+ iqp.vfn = 0;
+ iqp.iq_start = 1;
+ iqp.viid = 0;
+ iqp.type = FW_IQ_TYPE_FL_INT_CAP;
+ iqp.iqasynch = async;
+ if (csio_intr_coalesce_cnt)
+ iqp.iqanus = X_UPDATESCHEDULING_COUNTER_OPTTIMER;
+ else
+ iqp.iqanus = X_UPDATESCHEDULING_TIMER;
+ iqp.iqanud = X_UPDATEDELIVERY_INTERRUPT;
+ iqp.iqpciech = portid;
+ iqp.iqintcntthresh = (uint8_t)csio_sge_thresh_reg;
+
+ switch (csio_q_wr_sz(hw, iq_idx)) {
+ case 16:
+ iqp.iqesize = 0; break;
+ case 32:
+ iqp.iqesize = 1; break;
+ case 64:
+ iqp.iqesize = 2; break;
+ case 128:
+ iqp.iqesize = 3; break;
+ }
+
+ iqp.iqsize = csio_q_size(hw, iq_idx) /
+ csio_q_wr_sz(hw, iq_idx);
+ iqp.iqaddr = csio_q_pstart(hw, iq_idx);
+
+ flq_idx = csio_q_iq_flq_idx(hw, iq_idx);
+ if (flq_idx != -1) {
+ struct csio_q *flq = hw->wrm.q_arr[flq_idx];
+
+ iqp.fl0paden = 1;
+ iqp.fl0packen = flq->un.fl.packen ? 1 : 0;
+ iqp.fl0fbmin = X_FETCHBURSTMIN_64B;
+ iqp.fl0fbmax = X_FETCHBURSTMAX_512B;
+ iqp.fl0size = csio_q_size(hw, flq_idx) / CSIO_QCREDIT_SZ;
+ iqp.fl0addr = csio_q_pstart(hw, flq_idx);
+ }
+
+ csio_mb_iq_alloc_write(hw, mbp, priv, CSIO_MB_DEFAULT_TMO, &iqp, cbfn);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of IQ cmd failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ if (cbfn != NULL)
+ return CSIO_SUCCESS;
+
+ return csio_wr_iq_create_rsp(hw, mbp, iq_idx);
+}
+
+/*
+ * csio_wr_eq_create_rsp - Response handler for EQ creation.
+ * @hw: The HW module.
+ * @mbp: Mailbox.
+ * @eq_idx: Egress queue that got created.
+ *
+ * Handle FW_EQ_OFLD_CMD mailbox completion. Save off the assigned EQ ids.
+ */
+static csio_retval_t
+csio_wr_eq_cfg_rsp(struct csio_hw *hw, struct csio_mb *mbp, int eq_idx)
+{
+ struct csio_eq_params eqp;
+ enum fw_retval retval;
+
+ memset(&eqp, 0, sizeof(struct csio_eq_params));
+
+ csio_mb_eq_ofld_alloc_write_rsp(hw, mbp, &retval, &eqp);
+
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "EQ OFLD cmd returned 0x%x!\n", retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ csio_q_eqid(hw, eq_idx) = (uint16_t)eqp.eqid;
+ csio_q_physeqid(hw, eq_idx) = (uint16_t)eqp.physeqid;
+ csio_q_pidx(hw, eq_idx) = csio_q_cidx(hw, eq_idx) = 0;
+ csio_q_inc_idx(hw, eq_idx) = 0;
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_eq_create - Configure an Egress queue with FW.
+ * @hw: HW module.
+ * @priv: Private data.
+ * @eq_idx: Egress queue index in the WR module.
+ * @iq_idx: Associated ingress queue index.
+ * @cbfn: Completion callback.
+ *
+ * This API configures a offload egress queue with FW by issuing a
+ * FW_EQ_OFLD_CMD (with alloc + write ) mailbox.
+ */
+csio_retval_t
+csio_wr_eq_create(struct csio_hw *hw, void *priv, int eq_idx,
+ int iq_idx, uint8_t portid,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct csio_mb *mbp;
+ struct csio_eq_params eqp;
+
+ memset(&eqp, 0, sizeof(struct csio_eq_params));
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_err(hw, "EQ command out of memory!\n");
+ return CSIO_NOMEM;
+ }
+
+ eqp.pfn = hw->pfn;
+ eqp.vfn = 0;
+ eqp.eqstart = 1;
+ eqp.hostfcmode = X_HOSTFCMODE_STATUS_PAGE;
+ eqp.iqid = csio_q_iqid(hw, iq_idx);
+ eqp.fbmin = X_FETCHBURSTMIN_64B;
+ eqp.fbmax = X_FETCHBURSTMAX_512B;
+ eqp.cidxfthresh = 0;
+ eqp.pciechn = portid;
+ eqp.eqsize = csio_q_size(hw, eq_idx) / CSIO_QCREDIT_SZ;
+ eqp.eqaddr = csio_q_pstart(hw, eq_idx);
+
+ csio_mb_eq_ofld_alloc_write(hw, mbp, priv, CSIO_MB_DEFAULT_TMO,
+ &eqp, cbfn);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of EQ OFLD cmd failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ if (cbfn != NULL)
+ return CSIO_SUCCESS;
+
+ return csio_wr_eq_cfg_rsp(hw, mbp, eq_idx);
+}
+
+/*
+ * csio_wr_iq_destroy_rsp - Response handler for IQ removal.
+ * @hw: The HW module.
+ * @mbp: Mailbox.
+ * @iq_idx: Ingress queue that was freed.
+ *
+ * Handle FW_IQ_CMD (free) mailbox completion.
+ */
+static csio_retval_t
+csio_wr_iq_destroy_rsp(struct csio_hw *hw, struct csio_mb *mbp, int iq_idx)
+{
+ enum fw_retval retval = csio_mb_fw_retval(mbp);
+ csio_retval_t rv = CSIO_SUCCESS;
+
+ if (retval != FW_SUCCESS)
+ rv = CSIO_INVAL;
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return rv;
+}
+
+/*
+ * csio_wr_iq_destroy - Free an ingress queue.
+ * @hw: The HW module.
+ * @priv: Private data object.
+ * @iq_idx: Ingress queue index to destroy
+ * @cbfn: Completion callback.
+ *
+ * This API frees an ingress queue by issuing the FW_IQ_CMD
+ * with the free bit set.
+ */
+static csio_retval_t
+csio_wr_iq_destroy(struct csio_hw *hw, void *priv, int iq_idx,
+ void (*cbfn)(struct csio_hw *, struct csio_mb *))
+{
+ csio_retval_t rv = CSIO_SUCCESS;
+ struct csio_mb *mbp;
+ struct csio_iq_params iqp;
+ int flq_idx;
+
+ memset(&iqp, 0, sizeof(struct csio_iq_params));
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp)
+ return CSIO_NOMEM;
+
+ iqp.pfn = hw->pfn;
+ iqp.vfn = 0;
+ iqp.iqid = csio_q_iqid(hw, iq_idx);
+ iqp.type = FW_IQ_TYPE_FL_INT_CAP;
+
+ flq_idx = csio_q_iq_flq_idx(hw, iq_idx);
+ if (flq_idx != -1)
+ iqp.fl0id = csio_q_flid(hw, flq_idx);
+ else
+ iqp.fl0id = 0xFFFF;
+
+ iqp.fl1id = 0xFFFF;
+
+ csio_mb_iq_free(hw, mbp, priv, CSIO_MB_DEFAULT_TMO, &iqp, cbfn);
+
+ rv = csio_mb_issue(hw, mbp);
+ if (rv != CSIO_SUCCESS) {
+ mempool_free(mbp, hw->mb_mempool);
+ return rv;
+ }
+
+ if (cbfn != NULL)
+ return CSIO_SUCCESS;
+
+ return csio_wr_iq_destroy_rsp(hw, mbp, iq_idx);
+}
+
+/*
+ * csio_wr_eq_destroy_rsp - Response handler for OFLD EQ creation.
+ * @hw: The HW module.
+ * @mbp: Mailbox.
+ * @eq_idx: Egress queue that was freed.
+ *
+ * Handle FW_OFLD_EQ_CMD (free) mailbox completion.
+ */
+static csio_retval_t
+csio_wr_eq_destroy_rsp(struct csio_hw *hw, struct csio_mb *mbp, int eq_idx)
+{
+ enum fw_retval retval = csio_mb_fw_retval(mbp);
+ csio_retval_t rv = CSIO_SUCCESS;
+
+ if (retval != FW_SUCCESS)
+ rv = CSIO_INVAL;
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ return rv;
+}
+
+/*
+ * csio_wr_eq_destroy - Free an Egress queue.
+ * @hw: The HW module.
+ * @priv: Private data object.
+ * @eq_idx: Egress queue index to destroy
+ * @cbfn: Completion callback.
+ *
+ * This API frees an Egress queue by issuing the FW_EQ_OFLD_CMD
+ * with the free bit set.
+ */
+static csio_retval_t
+csio_wr_eq_destroy(struct csio_hw *hw, void *priv, int eq_idx,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ csio_retval_t rv = CSIO_SUCCESS;
+ struct csio_mb *mbp;
+ struct csio_eq_params eqp;
+
+ memset(&eqp, 0, sizeof(struct csio_eq_params));
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp)
+ return CSIO_NOMEM;
+
+ eqp.pfn = hw->pfn;
+ eqp.vfn = 0;
+ eqp.eqid = csio_q_eqid(hw, eq_idx);
+
+ csio_mb_eq_ofld_free(hw, mbp, priv, CSIO_MB_DEFAULT_TMO, &eqp, cbfn);
+
+ rv = csio_mb_issue(hw, mbp);
+ if (rv != CSIO_SUCCESS) {
+ mempool_free(mbp, hw->mb_mempool);
+ return rv;
+ }
+
+ if (cbfn != NULL)
+ return CSIO_SUCCESS;
+
+ return csio_wr_eq_destroy_rsp(hw, mbp, eq_idx);
+}
+
+/*
+ * csio_wr_cleanup_eq_stpg - Cleanup Egress queue status page
+ * @hw: HW module
+ * @qidx: Egress queue index
+ *
+ * Cleanup the Egress queue status page.
+ */
+static void
+csio_wr_cleanup_eq_stpg(struct csio_hw *hw, int qidx)
+{
+ struct csio_q *q = csio_hw_to_wrm(hw)->q_arr[qidx];
+ struct csio_qstatus_page *stp = (struct csio_qstatus_page *)q->vwrap;
+
+ memset(stp, 0, sizeof(*stp));
+}
+
+/*
+ * csio_wr_cleanup_iq_ftr - Cleanup Footer entries in IQ
+ * @hw: HW module
+ * @qidx: Ingress queue index
+ *
+ * Cleanup the footer entries in the given ingress queue,
+ * set to 1 the internal copy of genbit.
+ */
+static void
+csio_wr_cleanup_iq_ftr(struct csio_hw *hw, int qidx)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *q = wrm->q_arr[qidx];
+ void *wr;
+ struct csio_iqwr_footer *ftr;
+ uint32_t i = 0;
+
+ /* set to 1 since we are just about zero out genbit */
+ q->un.iq.genbit = 1;
+
+ for (i = 0; i < q->credits; i++) {
+ /* Get the WR */
+ wr = (void *)((uintptr_t)q->vstart +
+ (i * q->wr_sz));
+ /* Get the footer */
+ ftr = (struct csio_iqwr_footer *)((uintptr_t)wr +
+ (q->wr_sz - sizeof(*ftr)));
+ /* Zero out footer */
+ memset(ftr, 0, sizeof(*ftr));
+ }
+}
+
+csio_retval_t
+csio_wr_destroy_queues(struct csio_hw *hw, bool cmd)
+{
+ int i, flq_idx;
+ struct csio_q *q;
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ csio_retval_t rv;
+
+ for (i = 0; i < wrm->free_qidx; i++) {
+ q = wrm->q_arr[i];
+
+ switch (q->type) {
+ case CSIO_EGRESS:
+ if (csio_q_eqid(hw, i) != CSIO_MAX_QID) {
+ csio_wr_cleanup_eq_stpg(hw, i);
+ if (!cmd) {
+ csio_q_eqid(hw, i) = CSIO_MAX_QID;
+ continue;
+ }
+
+ rv = csio_wr_eq_destroy(hw, NULL, i, NULL);
+ if ((rv == CSIO_BUSY) || (rv == CSIO_TIMEOUT))
+ cmd = CSIO_FALSE;
+
+ csio_q_eqid(hw, i) = CSIO_MAX_QID;
+ }
+ case CSIO_INGRESS:
+ if (csio_q_iqid(hw, i) != CSIO_MAX_QID) {
+ csio_wr_cleanup_iq_ftr(hw, i);
+ if (!cmd) {
+ csio_q_iqid(hw, i) = CSIO_MAX_QID;
+ flq_idx = csio_q_iq_flq_idx(hw, i);
+ if (flq_idx != -1)
+ csio_q_flid(hw, flq_idx) =
+ CSIO_MAX_QID;
+ continue;
+ }
+
+ rv = csio_wr_iq_destroy(hw, NULL, i, NULL);
+ if ((rv == CSIO_BUSY) || (rv == CSIO_TIMEOUT))
+ cmd = CSIO_FALSE;
+
+ csio_q_iqid(hw, i) = CSIO_MAX_QID;
+ flq_idx = csio_q_iq_flq_idx(hw, i);
+ if (flq_idx != -1)
+ csio_q_flid(hw, flq_idx) = CSIO_MAX_QID;
+ }
+ default:
+ break;
+ }
+ }
+
+ hw->flags &= ~CSIO_HWF_Q_FW_ALLOCED;
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_get - Get requested size of WR entry/entries from queue.
+ * @hw: HW module.
+ * @qidx: Index of queue.
+ * @size: Cumulative size of Work request(s).
+ * @wrp: Work request pair.
+ *
+ * If requested credits are available, return the start address of the
+ * work request in the work request pair. Set pidx accordingly and
+ * return.
+ *
+ * NOTE about WR pair:
+ * ==================
+ * A WR can start towards the end of a queue, and then continue at the
+ * beginning, since the queue is considered to be circular. This will
+ * require a pair of address/size to be passed back to the caller -
+ * hence Work request pair format.
+ */
+csio_retval_t
+csio_wr_get(struct csio_hw *hw, int qidx, uint32_t size,
+ struct csio_wr_pair *wrp)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *q = wrm->q_arr[qidx];
+ void *cwr = (void *)((uintptr_t)(q->vstart) +
+ (q->pidx * CSIO_QCREDIT_SZ));
+ struct csio_qstatus_page *stp = (struct csio_qstatus_page *)q->vwrap;
+ uint16_t cidx = q->cidx = ntohs(stp->cidx);
+ uint16_t pidx = q->pidx;
+ uint32_t req_sz = ALIGN(size, CSIO_QCREDIT_SZ);
+ int req_credits = req_sz / CSIO_QCREDIT_SZ;
+ int credits;
+
+ CSIO_DB_ASSERT(q->owner != NULL);
+ CSIO_DB_ASSERT((qidx >= 0) && (qidx < wrm->free_qidx));
+ CSIO_DB_ASSERT(cidx <= q->credits);
+
+ /* Calculate credits */
+ if (pidx > cidx) {
+ credits = q->credits - (pidx - cidx) - 1;
+ } else if (cidx > pidx) {
+ credits = cidx - pidx - 1;
+ } else {
+ /* cidx == pidx, empty queue */
+ credits = q->credits;
+ csio_inc_stats(q, n_qempty);
+ }
+
+ /*
+ * Check if we have enough credits.
+ * credits = 1 implies queue is full.
+ */
+ if (!credits || (req_credits > credits)) {
+ csio_inc_stats(q, n_qfull);
+ return CSIO_BUSY;
+ }
+
+ /*
+ * If we are here, we have enough credits to satisfy the
+ * request. Check if we are near the end of q, and if WR spills over.
+ * If it does, use the first addr/size to cover the queue until
+ * the end. Fit the remainder portion of the request at the top
+ * of queue and return it in the second addr/len. Set pidx
+ * accordingly.
+ */
+ if (unlikely(((uintptr_t)cwr + req_sz) > (uintptr_t)(q->vwrap))) {
+ wrp->addr1 = cwr;
+ wrp->size1 = (uint32_t)((uintptr_t)q->vwrap - (uintptr_t)cwr);
+ wrp->addr2 = q->vstart;
+ wrp->size2 = req_sz - wrp->size1;
+ q->pidx = (uint16_t)(ALIGN(wrp->size2, CSIO_QCREDIT_SZ) /
+ CSIO_QCREDIT_SZ);
+ csio_inc_stats(q, n_qwrap);
+ csio_inc_stats(q, n_eq_wr_split);
+ } else {
+ wrp->addr1 = cwr;
+ wrp->size1 = req_sz;
+ wrp->addr2 = NULL;
+ wrp->size2 = 0;
+ q->pidx += (uint16_t)req_credits;
+
+ /* We are the end of queue, roll back pidx to top of queue */
+ if (unlikely(q->pidx == q->credits)) {
+ q->pidx = 0;
+ csio_inc_stats(q, n_qwrap);
+ }
+ }
+
+ q->inc_idx = (uint16_t)req_credits;
+
+ csio_inc_stats(q, n_tot_reqs);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_wr_copy_to_wrp - Copies given data into WR.
+ * @data_buf - Data buffer
+ * @wrp - Work request pair.
+ * @wr_off - Work request offset.
+ * @data_len - Data length.
+ *
+ * Copies the given data in Work Request. Work request pair(wrp) specifies
+ * address information of Work request.
+ * Returns: none
+ */
+void
+csio_wr_copy_to_wrp(void *data_buf, struct csio_wr_pair *wrp,
+ uint32_t wr_off, uint32_t data_len)
+{
+ uint32_t nbytes;
+
+ /* Number of space available in buffer addr1 of WRP */
+ nbytes = ((wrp->size1 - wr_off) >= data_len) ?
+ data_len : (wrp->size1 - wr_off);
+
+ memcpy((uint8_t *) wrp->addr1 + wr_off, data_buf, nbytes);
+ data_len -= nbytes;
+
+ /* Write the remaining data from the begining of circular buffer */
+ if (data_len) {
+ CSIO_DB_ASSERT(data_len <= wrp->size2);
+ CSIO_DB_ASSERT(wrp->addr2 != NULL);
+ memcpy(wrp->addr2, (uint8_t *) data_buf + nbytes, data_len);
+ }
+}
+
+/*
+ * csio_wr_issue - Notify chip of Work request.
+ * @hw: HW module.
+ * @qidx: Index of queue.
+ * @prio: 0: Low priority, 1: High priority
+ *
+ * Rings the SGE Doorbell by writing the current producer index of the passed
+ * in queue into the register.
+ *
+ */
+csio_retval_t
+csio_wr_issue(struct csio_hw *hw, int qidx, bool prio)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *q = wrm->q_arr[qidx];
+
+ CSIO_DB_ASSERT((qidx >= 0) && (qidx < wrm->free_qidx));
+
+ wmb();
+ /* Ring SGE Doorbell writing q->pidx into it */
+ csio_wr_reg32(hw, DBPRIO(prio) | QID(q->un.eq.physeqid) |
+ PIDX(q->inc_idx), MYPF_REG(SGE_PF_KDOORBELL));
+ q->inc_idx = 0;
+
+ return CSIO_SUCCESS;
+}
+
+static inline uint32_t
+csio_wr_avail_qcredits(struct csio_q *q)
+{
+ if (q->pidx > q->cidx)
+ return q->pidx - q->cidx;
+ else if (q->cidx > q->pidx)
+ return q->credits - (q->cidx - q->pidx);
+ else
+ return 0; /* cidx == pidx, empty queue */
+}
+
+/*
+ * csio_wr_inval_flq_buf - Invalidate a free list buffer entry.
+ * @hw: HW module.
+ * @flq: The freelist queue.
+ *
+ * Invalidate the driver's version of a freelist buffer entry,
+ * without freeing the associated the DMA memory. The entry
+ * to be invalidated is picked up from the current Free list
+ * queue cidx.
+ *
+ */
+static inline void
+csio_wr_inval_flq_buf(struct csio_hw *hw, struct csio_q *flq)
+{
+ flq->cidx++;
+ if (flq->cidx == flq->credits) {
+ flq->cidx = 0;
+ csio_inc_stats(flq, n_qwrap);
+ }
+}
+
+/*
+ * csio_wr_process_fl - Process a freelist completion.
+ * @hw: HW module.
+ * @q: The ingress queue attached to the Freelist.
+ * @wr: The freelist completion WR in the ingress queue.
+ * @len_to_qid: The lower 32-bits of the first flit of the RSP footer
+ * @iq_handler: Caller's handler for this completion.
+ * @priv: Private pointer of caller
+ *
+ */
+static inline void
+csio_wr_process_fl(struct csio_hw *hw, struct csio_q *q,
+ void *wr, uint32_t len_to_qid,
+ void (*iq_handler)(struct csio_hw *, void *,
+ uint32_t, struct csio_fl_dma_buf *,
+ void *),
+ void *priv)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ struct csio_fl_dma_buf flb;
+ struct csio_dma_buf *buf, *fbuf;
+ uint32_t bufsz, len, lastlen = 0;
+ int n;
+ struct csio_q *flq = hw->wrm.q_arr[q->un.iq.flq_idx];
+
+ CSIO_DB_ASSERT(flq != NULL);
+
+ len = len_to_qid;
+
+ if (len & IQWRF_NEWBUF) {
+ if (flq->un.fl.offset > 0) {
+ csio_wr_inval_flq_buf(hw, flq);
+ flq->un.fl.offset = 0;
+ }
+ len = IQWRF_LEN_GET(len);
+ }
+
+ CSIO_DB_ASSERT(len != 0);
+
+ flb.totlen = len;
+
+ /* Consume all freelist buffers used for len bytes */
+ for (n = 0, fbuf = flb.flbufs; ; n++, fbuf++) {
+ buf = &flq->un.fl.bufs[flq->cidx];
+ bufsz = csio_wr_fl_bufsz(sge, buf);
+
+ fbuf->paddr = buf->paddr;
+ fbuf->vaddr = buf->vaddr;
+
+ flb.offset = flq->un.fl.offset;
+ lastlen = min(bufsz, len);
+ fbuf->len = lastlen;
+
+ len -= lastlen;
+ if (!len)
+ break;
+ csio_wr_inval_flq_buf(hw, flq);
+ }
+
+ flb.defer_free = flq->un.fl.packen ? 0 : 1;
+
+ iq_handler(hw, wr, q->wr_sz - sizeof(struct csio_iqwr_footer),
+ &flb, priv);
+
+ if (flq->un.fl.packen)
+ flq->un.fl.offset += ALIGN(lastlen, sge->csio_fl_align);
+ else
+ csio_wr_inval_flq_buf(hw, flq);
+
+}
+
+/*
+ * csio_is_new_iqwr - Is this a new Ingress queue entry ?
+ * @q: Ingress quueue.
+ * @ftr: Ingress queue WR SGE footer.
+ *
+ * The entry is new if our generation bit matches the corresponding
+ * bit in the footer of the current WR.
+ */
+static inline bool
+csio_is_new_iqwr(struct csio_q *q, struct csio_iqwr_footer *ftr)
+{
+ return (q->un.iq.genbit == (ftr->u.type_gen >> IQWRF_GEN_SHIFT));
+}
+
+/*
+ * csio_wr_process_iq - Process elements in Ingress queue.
+ * @hw: HW pointer
+ * @qidx: Index of queue
+ * @iq_handler: Handler for this queue
+ * @priv: Caller's private pointer
+ *
+ * This routine walks through every entry of the ingress queue, calling
+ * the provided iq_handler with the entry, until the generation bit
+ * flips.
+ */
+csio_retval_t
+csio_wr_process_iq(struct csio_hw *hw, struct csio_q *q,
+ void (*iq_handler)(struct csio_hw *, void *,
+ uint32_t, struct csio_fl_dma_buf *,
+ void *),
+ void *priv)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ void *wr = (void *)((uintptr_t)q->vstart + (q->cidx * q->wr_sz));
+ struct csio_iqwr_footer *ftr;
+ uint32_t wr_type, fw_qid, qid;
+ struct csio_q *q_completed;
+ struct csio_q *flq = csio_iq_has_fl(q) ?
+ wrm->q_arr[q->un.iq.flq_idx] : NULL;
+ csio_retval_t rv = CSIO_SUCCESS;
+
+ /* Get the footer */
+ ftr = (struct csio_iqwr_footer *)((uintptr_t)wr +
+ (q->wr_sz - sizeof(*ftr)));
+
+ /*
+ * When q wrapped around last time, driver should have inverted
+ * ic.genbit as well.
+ */
+ while (csio_is_new_iqwr(q, ftr)) {
+
+ CSIO_DB_ASSERT(((uintptr_t)wr + q->wr_sz) <=
+ (uintptr_t)q->vwrap);
+ rmb();
+ wr_type = IQWRF_TYPE_GET(ftr->u.type_gen);
+
+ switch (wr_type) {
+ case X_RSPD_TYPE_CPL:
+ /* Subtract footer from WR len */
+ iq_handler(hw, wr, q->wr_sz - sizeof(*ftr), NULL, priv);
+ break;
+ case X_RSPD_TYPE_FLBUF:
+ csio_wr_process_fl(hw, q, wr,
+ ntohl(ftr->pldbuflen_qid),
+ iq_handler, priv);
+ break;
+ case X_RSPD_TYPE_INTR:
+ fw_qid = ntohl(ftr->pldbuflen_qid);
+ qid = fw_qid - wrm->fw_iq_start;
+ q_completed = hw->wrm.intr_map[qid];
+
+ if (unlikely(qid ==
+ csio_q_physiqid(hw, hw->intr_iq_idx))) {
+ /*
+ * We are already in the Forward Interrupt
+ * Interrupt Queue Service! Do-not service
+ * again!
+ *
+ */
+ } else {
+ CSIO_DB_ASSERT(q_completed);
+ CSIO_DB_ASSERT(
+ q_completed->un.iq.iq_intx_handler);
+
+ /* Call the queue handler. */
+ q_completed->un.iq.iq_intx_handler(hw, NULL,
+ 0, NULL, (void *)q_completed);
+ }
+ break;
+ default:
+ csio_warn(hw, "Unknown resp type 0x%x received\n",
+ wr_type);
+ csio_inc_stats(q, n_rsp_unknown);
+ break;
+ }
+
+ /*
+ * Ingress *always* has fixed size WR entries. Therefore,
+ * there should always be complete WRs towards the end of
+ * queue.
+ */
+ if (((uintptr_t)wr + q->wr_sz) == (uintptr_t)q->vwrap) {
+
+ /* Roll over to start of queue */
+ q->cidx = 0;
+ wr = q->vstart;
+
+ /* Toggle genbit */
+ q->un.iq.genbit ^= 0x1;
+
+ csio_inc_stats(q, n_qwrap);
+ } else {
+ q->cidx++;
+ wr = (void *)((uintptr_t)(q->vstart) +
+ (q->cidx * q->wr_sz));
+ }
+
+ ftr = (struct csio_iqwr_footer *)((uintptr_t)wr +
+ (q->wr_sz - sizeof(*ftr)));
+ q->inc_idx++;
+
+ } /* while (q->un.iq.genbit == hdr->genbit) */
+
+ /*
+ * We need to re-arm SGE interrupts in case we got a stray interrupt,
+ * especially in msix mode. With INTx, this may be a common occurence.
+ */
+ if (unlikely(!q->inc_idx)) {
+ csio_inc_stats(q, n_stray_comp);
+ rv = CSIO_INVAL;
+ goto restart;
+ }
+
+ /* Replenish free list buffers if pending falls below low water mark */
+ if (flq) {
+ uint32_t avail = csio_wr_avail_qcredits(flq);
+ if (avail <= 16) {
+ /* Make sure in FLQ, atleast 1 credit (8 FL buffers)
+ * remains unpopulated otherwise HW thinks
+ * FLQ is empty.
+ */
+ csio_wr_update_fl(hw, flq, (flq->credits - 8) - avail);
+ csio_wr_ring_fldb(hw, flq);
+ }
+ }
+
+restart:
+ /* Now inform SGE about our incremental index value */
+ csio_wr_reg32(hw, CIDXINC(q->inc_idx) |
+ INGRESSQID(q->un.iq.physiqid) |
+ TIMERREG(csio_sge_timer_reg),
+ MYPF_REG(SGE_PF_GTS));
+ q->stats.n_tot_rsps += q->inc_idx;
+
+ q->inc_idx = 0;
+
+ return rv;
+}
+
+csio_retval_t
+csio_wr_process_iq_idx(struct csio_hw *hw, int qidx,
+ void (*iq_handler)(struct csio_hw *, void *,
+ uint32_t, struct csio_fl_dma_buf *,
+ void *),
+ void *priv)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_q *iq = wrm->q_arr[qidx];
+
+ return csio_wr_process_iq(hw, iq, iq_handler, priv);
+}
+
+static int
+csio_closest_timer(struct csio_sge *s, int time)
+{
+ int i, delta, match = 0, min_delta = INT_MAX;
+
+ for (i = 0; i < ARRAY_SIZE(s->timer_val); i++) {
+ delta = time - s->timer_val[i];
+ if (delta < 0)
+ delta = -delta;
+ if (delta < min_delta) {
+ min_delta = delta;
+ match = i;
+ }
+ }
+ return match;
+}
+
+static int
+csio_closest_thresh(struct csio_sge *s, int cnt)
+{
+ int i, delta, match = 0, min_delta = INT_MAX;
+
+ for (i = 0; i < ARRAY_SIZE(s->counter_val); i++) {
+ delta = cnt - s->counter_val[i];
+ if (delta < 0)
+ delta = -delta;
+ if (delta < min_delta) {
+ min_delta = delta;
+ match = i;
+ }
+ }
+ return match;
+}
+
+static void
+csio_wr_fixup_host_params(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ uint32_t clsz = L1_CACHE_BYTES;
+ uint32_t s_hps = PAGE_SHIFT - 10;
+ uint32_t ingpad = 0;
+ uint32_t stat_len = clsz > 64 ? 128 : 64;
+
+ csio_wr_reg32(hw, HOSTPAGESIZEPF0(s_hps) | HOSTPAGESIZEPF1(s_hps) |
+ HOSTPAGESIZEPF2(s_hps) | HOSTPAGESIZEPF3(s_hps) |
+ HOSTPAGESIZEPF4(s_hps) | HOSTPAGESIZEPF5(s_hps) |
+ HOSTPAGESIZEPF6(s_hps) | HOSTPAGESIZEPF7(s_hps),
+ SGE_HOST_PAGE_SIZE);
+
+ sge->csio_fl_align = clsz < 32 ? 32 : clsz;
+ ingpad = ilog2(sge->csio_fl_align) - 5;
+
+ csio_set_reg_field(hw, SGE_CONTROL, INGPADBOUNDARY_MASK |
+ EGRSTATUSPAGESIZE(1),
+ INGPADBOUNDARY(ingpad) |
+ EGRSTATUSPAGESIZE(stat_len != 64));
+
+ /* FL BUFFER SIZE#0 is Page size i,e already aligned to cache line */
+ csio_wr_reg32(hw, PAGE_SIZE, SGE_FL_BUFFER_SIZE0);
+ csio_wr_reg32(hw,
+ (csio_rd_reg32(hw, SGE_FL_BUFFER_SIZE2) +
+ sge->csio_fl_align - 1) & ~(sge->csio_fl_align - 1),
+ SGE_FL_BUFFER_SIZE2);
+ csio_wr_reg32(hw,
+ (csio_rd_reg32(hw, SGE_FL_BUFFER_SIZE3) +
+ sge->csio_fl_align - 1) & ~(sge->csio_fl_align - 1),
+ SGE_FL_BUFFER_SIZE3);
+
+ csio_wr_reg32(hw, HPZ0(PAGE_SHIFT - 12), ULP_RX_TDDP_PSZ);
+
+ /* default value of rx_dma_offset of the NIC driver */
+ csio_set_reg_field(hw, SGE_CONTROL, PKTSHIFT_MASK,
+ PKTSHIFT(CSIO_SGE_RX_DMA_OFFSET));
+}
+
+static void
+csio_init_intr_coalesce_parms(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+
+ csio_sge_thresh_reg = csio_closest_thresh(sge, csio_intr_coalesce_cnt);
+ if (csio_intr_coalesce_cnt) {
+ csio_sge_thresh_reg = 0;
+ csio_sge_timer_reg = X_TIMERREG_RESTART_COUNTER;
+ return;
+ }
+
+ csio_sge_timer_reg = csio_closest_timer(sge, csio_intr_coalesce_time);
+}
+
+/*
+ * csio_wr_get_sge - Get SGE register values.
+ * @hw: HW module.
+ *
+ * Used by non-master functions and by master-functions relying on config file.
+ */
+static void
+csio_wr_get_sge(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ uint32_t ingpad;
+ int i;
+ u32 timer_value_0_and_1, timer_value_2_and_3, timer_value_4_and_5;
+ u32 ingress_rx_threshold;
+
+ sge->sge_control = csio_rd_reg32(hw, SGE_CONTROL);
+
+ ingpad = INGPADBOUNDARY_GET(sge->sge_control);
+
+ switch (ingpad) {
+ case X_INGPCIEBOUNDARY_32B:
+ sge->csio_fl_align = 32; break;
+ case X_INGPCIEBOUNDARY_64B:
+ sge->csio_fl_align = 64; break;
+ case X_INGPCIEBOUNDARY_128B:
+ sge->csio_fl_align = 128; break;
+ case X_INGPCIEBOUNDARY_256B:
+ sge->csio_fl_align = 256; break;
+ case X_INGPCIEBOUNDARY_512B:
+ sge->csio_fl_align = 512; break;
+ case X_INGPCIEBOUNDARY_1024B:
+ sge->csio_fl_align = 1024; break;
+ case X_INGPCIEBOUNDARY_2048B:
+ sge->csio_fl_align = 2048; break;
+ case X_INGPCIEBOUNDARY_4096B:
+ sge->csio_fl_align = 4096; break;
+ }
+
+ for (i = 0; i < CSIO_SGE_FL_SIZE_REGS; i++)
+ csio_get_flbuf_size(hw, sge, i);
+
+ timer_value_0_and_1 = csio_rd_reg32(hw, SGE_TIMER_VALUE_0_AND_1);
+ timer_value_2_and_3 = csio_rd_reg32(hw, SGE_TIMER_VALUE_2_AND_3);
+ timer_value_4_and_5 = csio_rd_reg32(hw, SGE_TIMER_VALUE_4_AND_5);
+
+ sge->timer_val[0] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE0_GET(timer_value_0_and_1));
+ sge->timer_val[1] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE1_GET(timer_value_0_and_1));
+ sge->timer_val[2] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE2_GET(timer_value_2_and_3));
+ sge->timer_val[3] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE3_GET(timer_value_2_and_3));
+ sge->timer_val[4] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE4_GET(timer_value_4_and_5));
+ sge->timer_val[5] = (uint16_t)csio_core_ticks_to_us(hw,
+ TIMERVALUE5_GET(timer_value_4_and_5));
+
+ ingress_rx_threshold = csio_rd_reg32(hw, SGE_INGRESS_RX_THRESHOLD);
+ sge->counter_val[0] = THRESHOLD_0_GET(ingress_rx_threshold);
+ sge->counter_val[1] = THRESHOLD_1_GET(ingress_rx_threshold);
+ sge->counter_val[2] = THRESHOLD_2_GET(ingress_rx_threshold);
+ sge->counter_val[3] = THRESHOLD_3_GET(ingress_rx_threshold);
+
+ csio_init_intr_coalesce_parms(hw);
+}
+
+/*
+ * csio_wr_set_sge - Initialize SGE registers
+ * @hw: HW module.
+ *
+ * Used by Master function to initialize SGE registers in the absence
+ * of a config file.
+ */
+static void
+csio_wr_set_sge(struct csio_hw *hw)
+{
+ struct csio_wrm *wrm = csio_hw_to_wrm(hw);
+ struct csio_sge *sge = &wrm->sge;
+ int i;
+
+ /*
+ * Set up our basic SGE mode to deliver CPL messages to our Ingress
+ * Queue and Packet Date to the Free List.
+ */
+ csio_set_reg_field(hw, SGE_CONTROL, RXPKTCPLMODE, RXPKTCPLMODE);
+
+ sge->sge_control = csio_rd_reg32(hw, SGE_CONTROL);
+
+ /* sge->csio_fl_align is set up by csio_wr_fixup_host_params(). */
+
+ /*
+ * Set up to drop DOORBELL writes when the DOORBELL FIFO overflows
+ * and generate an interrupt when this occurs so we can recover.
+ */
+ csio_set_reg_field(hw, SGE_DBFIFO_STATUS,
+ HP_INT_THRESH(HP_INT_THRESH_MASK) |
+ LP_INT_THRESH(LP_INT_THRESH_MASK),
+ HP_INT_THRESH(CSIO_SGE_DBFIFO_INT_THRESH) |
+ LP_INT_THRESH(CSIO_SGE_DBFIFO_INT_THRESH));
+ csio_set_reg_field(hw, SGE_DOORBELL_CONTROL, ENABLE_DROP,
+ ENABLE_DROP);
+
+ /* SGE_FL_BUFFER_SIZE0 is set up by csio_wr_fixup_host_params(). */
+
+ csio_set_flbuf_size(hw, 1, CSIO_SGE_FLBUF_SIZE1);
+ csio_set_flbuf_size(hw, 2, CSIO_SGE_FLBUF_SIZE2);
+ csio_set_flbuf_size(hw, 3, CSIO_SGE_FLBUF_SIZE3);
+ csio_set_flbuf_size(hw, 4, CSIO_SGE_FLBUF_SIZE4);
+ csio_set_flbuf_size(hw, 5, CSIO_SGE_FLBUF_SIZE5);
+ csio_set_flbuf_size(hw, 6, CSIO_SGE_FLBUF_SIZE6);
+ csio_set_flbuf_size(hw, 7, CSIO_SGE_FLBUF_SIZE7);
+ csio_set_flbuf_size(hw, 8, CSIO_SGE_FLBUF_SIZE8);
+
+ for (i = 0; i < CSIO_SGE_FL_SIZE_REGS; i++)
+ csio_get_flbuf_size(hw, sge, i);
+
+ /* Initialize interrupt coalescing attributes */
+ sge->timer_val[0] = CSIO_SGE_TIMER_VAL_0;
+ sge->timer_val[1] = CSIO_SGE_TIMER_VAL_1;
+ sge->timer_val[2] = CSIO_SGE_TIMER_VAL_2;
+ sge->timer_val[3] = CSIO_SGE_TIMER_VAL_3;
+ sge->timer_val[4] = CSIO_SGE_TIMER_VAL_4;
+ sge->timer_val[5] = CSIO_SGE_TIMER_VAL_5;
+
+ sge->counter_val[0] = CSIO_SGE_INT_CNT_VAL_0;
+ sge->counter_val[1] = CSIO_SGE_INT_CNT_VAL_1;
+ sge->counter_val[2] = CSIO_SGE_INT_CNT_VAL_2;
+ sge->counter_val[3] = CSIO_SGE_INT_CNT_VAL_3;
+
+ csio_wr_reg32(hw, THRESHOLD_0(sge->counter_val[0]) |
+ THRESHOLD_1(sge->counter_val[1]) |
+ THRESHOLD_2(sge->counter_val[2]) |
+ THRESHOLD_3(sge->counter_val[3]),
+ SGE_INGRESS_RX_THRESHOLD);
+
+ csio_wr_reg32(hw,
+ TIMERVALUE0(csio_us_to_core_ticks(hw, sge->timer_val[0])) |
+ TIMERVALUE1(csio_us_to_core_ticks(hw, sge->timer_val[1])),
+ SGE_TIMER_VALUE_0_AND_1);
+
+ csio_wr_reg32(hw,
+ TIMERVALUE2(csio_us_to_core_ticks(hw, sge->timer_val[2])) |
+ TIMERVALUE3(csio_us_to_core_ticks(hw, sge->timer_val[3])),
+ SGE_TIMER_VALUE_2_AND_3);
+
+ csio_wr_reg32(hw,
+ TIMERVALUE4(csio_us_to_core_ticks(hw, sge->timer_val[4])) |
+ TIMERVALUE5(csio_us_to_core_ticks(hw, sge->timer_val[5])),
+ SGE_TIMER_VALUE_4_AND_5);
+
+ csio_init_intr_coalesce_parms(hw);
+}
+
+void
+csio_wr_sge_init(struct csio_hw *hw)
+{
+ /*
+ * If we are master:
+ * - If we plan to use the config file, we need to fixup some
+ * host specific registers, and read the rest of the SGE
+ * configuration.
+ * - If we dont plan to use the config file, we need to initialize
+ * SGE entirely, including fixing the host specific registers.
+ * If we arent the master, we are only allowed to read and work off of
+ * the already initialized SGE values.
+ *
+ * Therefore, before calling this function, we assume that the master-
+ * ship of the card, and whether to use config file or not, have
+ * already been decided. In other words, CSIO_HWF_USING_SOFT_PARAMS and
+ * CSIO_HWF_MASTER should be set/unset.
+ */
+ if (csio_is_hw_master(hw)) {
+ csio_wr_fixup_host_params(hw);
+
+ if (hw->flags & CSIO_HWF_USING_SOFT_PARAMS)
+ csio_wr_get_sge(hw);
+ else
+ csio_wr_set_sge(hw);
+ } else
+ csio_wr_get_sge(hw);
+}
+
+/*
+ * csio_wrm_init - Initialize Work request module.
+ * @wrm: WR module
+ * @hw: HW pointer
+ *
+ * Allocates memory for an array of queue pointers starting at q_arr.
+ */
+csio_retval_t
+csio_wrm_init(struct csio_wrm *wrm, struct csio_hw *hw)
+{
+ int i;
+
+ if (!wrm->num_q) {
+ csio_err(hw, "Num queues is not set\n");
+ return CSIO_INVAL;
+ }
+
+ wrm->q_arr = kzalloc(sizeof(struct csio_q *) * wrm->num_q, GFP_KERNEL);
+ if (!wrm->q_arr)
+ goto err;
+
+ for (i = 0; i < wrm->num_q; i++) {
+ wrm->q_arr[i] = kzalloc(sizeof(struct csio_q), GFP_KERNEL);
+ if (!wrm->q_arr[i]) {
+ while (--i >= 0)
+ kfree(wrm->q_arr[i]);
+ goto err_free_arr;
+ }
+ }
+ wrm->free_qidx = 0;
+
+ return CSIO_SUCCESS;
+
+err_free_arr:
+ kfree(wrm->q_arr);
+err:
+ return CSIO_NOMEM;
+}
+
+/*
+ * csio_wrm_exit - Initialize Work request module.
+ * @wrm: WR module
+ * @hw: HW module
+ *
+ * Uninitialize WR module. Free q_arr and pointers in it.
+ * We have the additional job of freeing the DMA memory associated
+ * with the queues.
+ */
+void
+csio_wrm_exit(struct csio_wrm *wrm, struct csio_hw *hw)
+{
+ int i;
+ uint32_t j;
+ struct csio_q *q;
+ struct csio_dma_buf *buf;
+
+ for (i = 0; i < wrm->num_q; i++) {
+ q = wrm->q_arr[i];
+
+ if (wrm->free_qidx && (i < wrm->free_qidx)) {
+ if (q->type == CSIO_FREELIST) {
+ if (!q->un.fl.bufs)
+ continue;
+ for (j = 0; j < q->credits; j++) {
+ buf = &q->un.fl.bufs[j];
+ if (!buf->vaddr)
+ continue;
+ pci_free_consistent(hw->pdev, buf->len,
+ buf->vaddr,
+ buf->paddr);
+ }
+ kfree(q->un.fl.bufs);
+ }
+ pci_free_consistent(hw->pdev, q->size,
+ q->vstart, q->pstart);
+ }
+ kfree(q);
+ }
+
+ hw->flags &= ~CSIO_HWF_Q_MEM_ALLOCED;
+
+ kfree(wrm->q_arr);
+}
--
1.7.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH 3/8] csiostor: Chelsio FCoE offload driver submission (sources part 3).
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
2012-08-23 22:27 ` [PATCH 1/8] csiostor: Chelsio FCoE offload driver submission (sources part 1) Naresh Kumar Inna
2012-08-23 22:27 ` [PATCH 2/8] csiostor: Chelsio FCoE offload driver submission (sources part 2) Naresh Kumar Inna
@ 2012-08-23 22:27 ` Naresh Kumar Inna
2012-08-23 22:27 ` [PATCH 4/8] csiostor: Chelsio FCoE offload driver submission (sources part 4) Naresh Kumar Inna
` (6 subsequent siblings)
9 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
This patch contains code for the FC transport template callbacks and the
Mailbox module functionality. The FC transport callbacks include Virtual
Node ports creation and deletion, FC session registration, unregistration and
teardown. The Mailbox module provides services to issue/track/cancel
mailbox commands and wrappers for them.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/csiostor/csio_attr.c | 808 +++++++++++++++++
drivers/scsi/csiostor/csio_mb.c | 1768 +++++++++++++++++++++++++++++++++++++
2 files changed, 2576 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_attr.c
create mode 100644 drivers/scsi/csiostor/csio_mb.c
diff --git a/drivers/scsi/csiostor/csio_attr.c b/drivers/scsi/csiostor/csio_attr.c
new file mode 100644
index 0000000..fd30798
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_attr.c
@@ -0,0 +1,808 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/delay.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/mm.h>
+#include <linux/jiffies.h>
+
+#include "csio_init.h"
+
+static void
+csio_vport_set_state(struct csio_lnode *ln);
+
+/*
+ * csio_reg_rnode - Register a remote port with FC transport.
+ * @rn: Rnode representing remote port.
+ *
+ * Call fc_remote_port_add() to register this remote port with FC transport.
+ * If remote port is Initiator OR Target OR both, change the role appropriately.
+ *
+ */
+void
+csio_reg_rnode(struct csio_rnode *rn)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+ struct fc_rport_identifiers ids;
+ struct fc_rport *rport;
+ struct csio_service_parms *sp;
+
+ ids.node_name = wwn_to_u64(csio_rn_wwnn(rn));
+ ids.port_name = wwn_to_u64(csio_rn_wwpn(rn));
+ ids.port_id = rn->nport_id;
+ ids.roles = FC_RPORT_ROLE_UNKNOWN;
+
+ if (rn->role & CSIO_RNFR_INITIATOR || rn->role & CSIO_RNFR_TARGET) {
+ rport = rn->rport;
+ CSIO_ASSERT(rport != NULL);
+ goto update_role;
+ }
+
+ rn->rport = fc_remote_port_add(shost, 0, &ids);
+ if (!rn->rport) {
+ csio_ln_err(ln, "Failed to register rport = 0x%x.\n",
+ rn->nport_id);
+ return;
+ }
+
+ ln->num_reg_rnodes++;
+ rport = rn->rport;
+ spin_lock_irq(shost->host_lock);
+ *((struct csio_rnode **)rport->dd_data) = rn;
+ spin_unlock_irq(shost->host_lock);
+
+ sp = &rn->rn_sparm;
+ rport->maxframe_size = sp->csp.rcv_sz;
+ if (G_SP_CLASS_SUPPORT(sp->clsp[2].serv_option))
+ rport->supported_classes = FC_COS_CLASS3;
+ else
+ rport->supported_classes = FC_COS_UNSPECIFIED;
+update_role:
+ if (rn->role & CSIO_RNFR_INITIATOR)
+ ids.roles |= FC_RPORT_ROLE_FCP_INITIATOR;
+ if (rn->role & CSIO_RNFR_TARGET) {
+ ids.roles |= FC_RPORT_ROLE_FCP_TARGET;
+ ln->n_scsi_tgts++;
+ }
+
+ if (ids.roles != FC_RPORT_ROLE_UNKNOWN)
+ fc_remote_port_rolechg(rport, ids.roles);
+
+ rn->scsi_id = rport->scsi_target_id;
+
+ csio_ln_dbg(ln, "Remote port x%x role 0x%x registered\n",
+ rn->nport_id, ids.roles);
+}
+
+/*
+ * csio_unreg_rnode - Unregister a remote port with FC transport.
+ * @rn: Rnode representing remote port.
+ *
+ * Call fc_remote_port_delete() to unregister this remote port with FC
+ * transport.
+ *
+ */
+void
+csio_unreg_rnode(struct csio_rnode *rn)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ struct fc_rport *rport = rn->rport;
+
+ rn->role &= ~(CSIO_RNFR_INITIATOR | CSIO_RNFR_TARGET);
+ fc_remote_port_delete(rport);
+ ln->num_reg_rnodes--;
+
+ if (ln->n_scsi_tgts)
+ ln->n_scsi_tgts--;
+
+ if (ln->last_scan_ntgts)
+ ln->last_scan_ntgts--;
+
+ csio_ln_dbg(ln, "Remote port x%x un-registered\n", rn->nport_id);
+}
+
+/*
+ * csio_lnode_async_event - Async events from local port.
+ * @ln: lnode representing local port.
+ *
+ * Async events from local node that FC transport/SCSI ML
+ * should be made aware of (Eg: RSCN).
+ */
+void
+csio_lnode_async_event(struct csio_lnode *ln, enum csio_ln_fc_evt fc_evt)
+{
+ switch (fc_evt) {
+ case CSIO_LN_FC_RSCN:
+ /* Get payload of rscn from ln */
+ /* For each RSCN entry */
+ /*
+ * fc_host_post_event(shost,
+ * fc_get_event_number(),
+ * FCH_EVT_RSCN,
+ * rscn_entry);
+ */
+ break;
+ case CSIO_LN_FC_LINKUP:
+ /* send fc_host_post_event */
+ /* set vport state */
+ if (csio_is_npiv_ln(ln))
+ csio_vport_set_state(ln);
+
+ break;
+ case CSIO_LN_FC_LINKDOWN:
+ /* send fc_host_post_event */
+ /* set vport state */
+ if (csio_is_npiv_ln(ln))
+ csio_vport_set_state(ln);
+
+ break;
+ case CSIO_LN_FC_ATTRIB_UPDATE:
+ csio_fchost_attr_init(ln);
+ break;
+ default:
+ break;
+ }
+}
+
+/*
+ * csio_fchost_attr_init - Initialize FC transport attributes
+ * @ln: Lnode.
+ *
+ */
+void
+csio_fchost_attr_init(struct csio_lnode *ln)
+{
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+
+ fc_host_node_name(shost) = wwn_to_u64(csio_ln_wwnn(ln));
+ fc_host_port_name(shost) = wwn_to_u64(csio_ln_wwpn(ln));
+
+ fc_host_supported_classes(shost) = FC_COS_CLASS3;
+ fc_host_max_npiv_vports(shost) =
+ (csio_lnode_to_hw(ln))->fres_info.max_vnps;
+ fc_host_supported_speeds(shost) = FC_PORTSPEED_10GBIT |
+ FC_PORTSPEED_1GBIT;
+
+ fc_host_maxframe_size(shost) = ln->ln_sparm.csp.rcv_sz;
+ memset(fc_host_supported_fc4s(shost), 0,
+ sizeof(fc_host_supported_fc4s(shost)));
+ fc_host_supported_fc4s(shost)[7] = 1;
+
+ memset(fc_host_active_fc4s(shost), 0,
+ sizeof(fc_host_active_fc4s(shost)));
+ fc_host_active_fc4s(shost)[7] = 1;
+}
+
+/*
+ * csio_get_host_port_id - sysfs entries for nport_id is
+ * populated/cached from this function
+ */
+static void
+csio_get_host_port_id(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ fc_host_port_id(shost) = ln->nport_id;
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_port_type - Return FC local port type.
+ * @shost: scsi host.
+ *
+ */
+static void
+csio_get_host_port_type(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ if (csio_is_npiv_ln(ln))
+ fc_host_port_type(shost) = FC_PORTTYPE_NPIV;
+ else
+ fc_host_port_type(shost) = FC_PORTTYPE_NPORT;
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_port_state - Return FC local port state.
+ * @shost: scsi host.
+ *
+ */
+static void
+csio_get_host_port_state(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ char state[16];
+
+ spin_lock_irq(&hw->lock);
+
+ csio_lnode_state_to_str(ln, state);
+ if (!strcmp(state, "READY"))
+ fc_host_port_state(shost) = FC_PORTSTATE_ONLINE;
+ else if (!strcmp(state, "OFFLINE"))
+ fc_host_port_state(shost) = FC_PORTSTATE_LINKDOWN;
+ else
+ fc_host_port_state(shost) = FC_PORTSTATE_UNKNOWN;
+
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_host_speed - Return link speed to FC transport.
+ * @shost: scsi host.
+ *
+ */
+static void
+csio_get_host_speed(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ switch (hw->pport[ln->portid].link_speed) {
+ case FW_PORT_CAP_SPEED_1G:
+ fc_host_speed(shost) = FC_PORTSPEED_1GBIT;
+ break;
+ case FW_PORT_CAP_SPEED_10G:
+ fc_host_speed(shost) = FC_PORTSPEED_10GBIT;
+ break;
+ default:
+ fc_host_speed(shost) = FC_PORTSPEED_UNKNOWN;
+ break;
+ }
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_host_fabric_name - Return fabric name
+ * @shost: scsi host.
+ *
+ */
+static void
+csio_get_host_fabric_name(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_rnode *rn = NULL;
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ rn = csio_rnode_lookup_portid(ln, FABRIC_DID);
+ if (rn)
+ fc_host_fabric_name(shost) = wwn_to_u64(csio_rn_wwnn(rn));
+ else
+ fc_host_fabric_name(shost) = 0;
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_get_host_speed - Return FC transport statistics.
+ * @ln: Lnode.
+ *
+ */
+static struct fc_host_statistics *
+csio_get_stats(struct Scsi_Host *shost)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct fc_host_statistics *fhs = &ln->fch_stats;
+ struct fw_fcoe_port_stats fcoe_port_stats;
+ uint64_t seconds;
+
+ memset(&fcoe_port_stats, 0, sizeof(struct fw_fcoe_port_stats));
+ csio_get_phy_port_stats(hw, ln->portid, &fcoe_port_stats);
+
+ fhs->tx_frames += (fcoe_port_stats.tx_bcast_frames +
+ fcoe_port_stats.tx_mcast_frames +
+ fcoe_port_stats.tx_ucast_frames +
+ fcoe_port_stats.tx_offload_frames);
+ fhs->tx_words += (fcoe_port_stats.tx_bcast_bytes +
+ fcoe_port_stats.tx_mcast_bytes +
+ fcoe_port_stats.tx_ucast_bytes +
+ fcoe_port_stats.tx_offload_bytes) /
+ CSIO_WORD_TO_BYTE;
+ fhs->rx_frames += (fcoe_port_stats.rx_bcast_frames +
+ fcoe_port_stats.rx_mcast_frames +
+ fcoe_port_stats.rx_ucast_frames);
+ fhs->rx_words += (fcoe_port_stats.rx_bcast_bytes +
+ fcoe_port_stats.rx_mcast_bytes +
+ fcoe_port_stats.rx_ucast_bytes) /
+ CSIO_WORD_TO_BYTE;
+ fhs->error_frames += fcoe_port_stats.rx_err_frames;
+ fhs->fcp_input_requests += ln->stats.n_input_requests;
+ fhs->fcp_output_requests += ln->stats.n_output_requests;
+ fhs->fcp_control_requests += ln->stats.n_control_requests;
+ fhs->fcp_input_megabytes += ln->stats.n_input_bytes >> 20;
+ fhs->fcp_output_megabytes += ln->stats.n_output_bytes >> 20;
+ fhs->link_failure_count = ln->stats.n_link_down;
+ /* Reset stats for the device */
+ seconds = jiffies_to_msecs(jiffies);
+ fhs->seconds_since_last_reset =
+ (seconds - hw->stats.n_reset_start) / 1000;
+ return fhs;
+}
+
+/*
+ * csio_set_rport_loss_tmo - Set the rport dev loss timeout
+ * @rport: fc rport.
+ * @timeout: new value for dev loss tmo.
+ *
+ * If timeout is non zero set the dev_loss_tmo to timeout, else set
+ * dev_loss_tmo to one.
+ */
+static void
+csio_set_rport_loss_tmo(struct fc_rport *rport, uint32_t timeout)
+{
+ if (timeout)
+ rport->dev_loss_tmo = timeout;
+ else
+ rport->dev_loss_tmo = 1;
+}
+
+static void
+csio_vport_set_state(struct csio_lnode *ln)
+{
+ struct fc_vport *fc_vport = ln->fc_vport;
+ struct csio_lnode *pln = ln->pln;
+ char state[16];
+
+ /* Set fc vport state based on phyiscal lnode */
+ csio_lnode_state_to_str(pln, state);
+ if (strcmp(state, "READY")) {
+ fc_vport_set_state(fc_vport, FC_VPORT_LINKDOWN);
+ return;
+ }
+
+ if (!(pln->flags & CSIO_LNF_NPIVSUPP)) {
+ fc_vport_set_state(fc_vport, FC_VPORT_NO_FABRIC_SUPP);
+ return;
+ }
+
+ /* Set fc vport state based on virtual lnode */
+ csio_lnode_state_to_str(ln, state);
+ if (strcmp(state, "READY")) {
+ fc_vport_set_state(fc_vport, FC_VPORT_LINKDOWN);
+ return;
+ }
+ fc_vport_set_state(fc_vport, FC_VPORT_ACTIVE);
+}
+
+static csio_retval_t
+csio_fcoe_alloc_vnp(struct csio_hw *hw, struct csio_lnode *ln)
+{
+ struct csio_lnode *pln;
+ struct csio_mb *mbp;
+ struct fw_fcoe_vnp_cmd *rsp;
+ csio_retval_t ret;
+ int retry = 0;
+
+ /* Issue VNP cmd to alloc vport */
+ /* Allocate Mbox request */
+ spin_lock_irq(&hw->lock);
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_NOMEM;
+ }
+
+ pln = ln->pln;
+ ln->fcf_flowid = pln->fcf_flowid;
+ ln->portid = pln->portid;
+
+ csio_fcoe_vnp_alloc_init_mb(ln, mbp, CSIO_MB_DEFAULT_TMO,
+ pln->fcf_flowid, pln->vnp_flowid, 0,
+ csio_ln_wwnn(ln), csio_ln_wwpn(ln),
+ NULL);
+
+ for (retry = 0; retry < 3; retry++) {
+ /* FW is expected to complete vnp cmd in immediate mode
+ * without much delay.
+ * Otherwise, there will be increase in IO latency since HW
+ * lock is held till completion of vnp mbox cmd.
+ */
+ ret = csio_mb_issue(hw, mbp);
+ if (ret != CSIO_BUSY)
+ break;
+
+ /* Retry if mbox returns busy */
+ spin_unlock_irq(&hw->lock);
+ msleep(2000);
+ spin_lock_irq(&hw->lock);
+ }
+
+ if (ret != CSIO_SUCCESS) {
+ csio_ln_err(ln, "Failed to issue mbox FCoE VNP command\n");
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return ret;
+ }
+
+ /* Process Mbox response of VNP command */
+ rsp = (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+ if (FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16)) != FW_SUCCESS) {
+ csio_ln_err(ln, "FCOE VNP ALLOC cmd returned 0x%x!\n",
+ FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16)));
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_NOMEM;
+ }
+
+ ln->vnp_flowid = FW_FCOE_VNP_CMD_VNPI_GET(
+ ntohl(rsp->gen_wwn_to_vnpi));
+ memcpy(csio_ln_wwnn(ln), rsp->vnport_wwnn, 8);
+ memcpy(csio_ln_wwpn(ln), rsp->vnport_wwpn, 8);
+
+ csio_ln_dbg(ln, "FCOE VNPI: 0x%x\n", ln->vnp_flowid);
+ csio_ln_dbg(ln, "\tWWNN: %x%x%x%x%x%x%x%x\n",
+ ln->ln_sparm.wwnn[0], ln->ln_sparm.wwnn[1],
+ ln->ln_sparm.wwnn[2], ln->ln_sparm.wwnn[3],
+ ln->ln_sparm.wwnn[4], ln->ln_sparm.wwnn[5],
+ ln->ln_sparm.wwnn[6], ln->ln_sparm.wwnn[7]);
+ csio_ln_dbg(ln, "\tWWPN: %x%x%x%x%x%x%x%x\n",
+ ln->ln_sparm.wwpn[0], ln->ln_sparm.wwpn[1],
+ ln->ln_sparm.wwpn[2], ln->ln_sparm.wwpn[3],
+ ln->ln_sparm.wwpn[4], ln->ln_sparm.wwpn[5],
+ ln->ln_sparm.wwpn[6], ln->ln_sparm.wwpn[7]);
+
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+
+ return CSIO_SUCCESS;
+}
+
+static csio_retval_t
+csio_fcoe_free_vnp(struct csio_hw *hw, struct csio_lnode *ln)
+{
+ struct csio_lnode *pln;
+ struct csio_mb *mbp;
+ struct fw_fcoe_vnp_cmd *rsp;
+ csio_retval_t ret;
+ int retry = 0;
+
+ /* Issue VNP cmd to free vport */
+ /* Allocate Mbox request */
+
+ spin_lock_irq(&hw->lock);
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_NOMEM;
+ }
+
+ pln = ln->pln;
+
+ csio_fcoe_vnp_free_init_mb(ln, mbp, CSIO_MB_DEFAULT_TMO,
+ ln->fcf_flowid, ln->vnp_flowid,
+ NULL);
+
+ for (retry = 0; retry < 3; retry++) {
+ ret = csio_mb_issue(hw, mbp);
+ if (ret != CSIO_BUSY)
+ break;
+
+ /* Retry if mbox returns busy */
+ spin_unlock_irq(&hw->lock);
+ msleep(2000);
+ spin_lock_irq(&hw->lock);
+ }
+
+ if (ret) {
+ csio_ln_err(ln, "Failed to issue mbox FCoE VNP command\n");
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_INVAL;
+ }
+
+ /* Process Mbox response of VNP command */
+ rsp = (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+ if (FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16)) != FW_SUCCESS) {
+ csio_ln_err(ln, "FCOE VNP FREE cmd returned 0x%x!\n",
+ FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16)));
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return CSIO_NOMEM;
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+
+ return CSIO_SUCCESS;
+}
+
+static int
+csio_vport_create(struct fc_vport *fc_vport, bool disable)
+{
+ struct Scsi_Host *shost = fc_vport->shost;
+ struct csio_lnode *pln = shost_priv(shost);
+ struct csio_lnode *ln = NULL;
+ struct csio_hw *hw = csio_lnode_to_hw(pln);
+ uint8_t wwn[8];
+ int ret = -1;
+
+ ln = csio_shost_init(hw, &fc_vport->dev, CSIO_FALSE, pln);
+ if (!ln)
+ goto error;
+
+ if (fc_vport->node_name != 0) {
+ u64_to_wwn(fc_vport->node_name, wwn);
+
+ if (!csio_valid_wwn(wwn)) {
+ csio_ln_err(ln,
+ "vport create failed. Invalid wwnn\n");
+ goto error;
+ }
+ memcpy(csio_ln_wwnn(ln), wwn, 8);
+ }
+
+ if (fc_vport->port_name != 0) {
+ u64_to_wwn(fc_vport->port_name, wwn);
+
+ if (!csio_valid_wwn(wwn)) {
+ csio_ln_err(ln,
+ "vport create failed. Invalid wwpn\n");
+ goto error;
+ }
+
+ if (csio_lnode_lookup_by_wwpn(hw, wwn)) {
+ csio_ln_err(ln,
+ "vport create failed. wwpn already exists\n");
+ goto error;
+ }
+ memcpy(csio_ln_wwpn(ln), wwn, 8);
+ }
+
+ fc_vport_set_state(fc_vport, FC_VPORT_INITIALIZING);
+
+ if (csio_fcoe_alloc_vnp(hw, ln))
+ goto error;
+
+ *(struct csio_lnode **)fc_vport->dd_data = ln;
+ ln->fc_vport = fc_vport;
+ if (!fc_vport->node_name)
+ fc_vport->node_name = wwn_to_u64(csio_ln_wwnn(ln));
+ if (!fc_vport->port_name)
+ fc_vport->port_name = wwn_to_u64(csio_ln_wwpn(ln));
+ csio_fchost_attr_init(ln);
+ return 0;
+error:
+ if (ln)
+ csio_shost_exit(ln);
+
+ return ret;
+}
+
+static int
+csio_vport_delete(struct fc_vport *fc_vport)
+{
+ struct csio_lnode *ln = *(struct csio_lnode **)fc_vport->dd_data;
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+ if (csio_is_hw_removing(hw)) {
+ spin_unlock_irq(&hw->lock);
+ csio_shost_exit(ln);
+ return 0;
+ }
+ spin_unlock_irq(&hw->lock);
+
+ /* Quiesce ios and send remove event to lnode */
+ scsi_block_requests(shost);
+ spin_lock_irq(&hw->lock);
+ csio_scsim_cleanup_io_lnode(csio_hw_to_scsim(hw), ln);
+ csio_lnode_close(ln);
+ spin_unlock_irq(&hw->lock);
+ scsi_unblock_requests(shost);
+
+ /* Free vnp */
+ if (fc_vport->vport_state != FC_VPORT_DISABLED)
+ csio_fcoe_free_vnp(hw, ln);
+ csio_ln_err(ln, "vport deleted\n");
+ csio_shost_exit(ln);
+ return 0;
+}
+
+static int
+csio_vport_disable(struct fc_vport *fc_vport, bool disable)
+{
+ struct csio_lnode *ln = *(struct csio_lnode **)fc_vport->dd_data;
+ struct Scsi_Host *shost = csio_ln_to_shost(ln);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ /* disable vport */
+ if (disable) {
+ /* Quiesce ios and send stop event to lnode */
+ scsi_block_requests(shost);
+ spin_lock_irq(&hw->lock);
+ csio_scsim_cleanup_io_lnode(csio_hw_to_scsim(hw), ln);
+ csio_lnode_stop(ln);
+ spin_unlock_irq(&hw->lock);
+ scsi_unblock_requests(shost);
+
+ /* Free vnp */
+ csio_fcoe_free_vnp(hw, ln);
+ fc_vport_set_state(fc_vport, FC_VPORT_DISABLED);
+ csio_ln_err(ln, "vport disabled\n");
+ return 0;
+ } else {
+ /* enable vport */
+ fc_vport_set_state(fc_vport, FC_VPORT_INITIALIZING);
+ if (csio_fcoe_alloc_vnp(hw, ln)) {
+ csio_ln_err(ln, "vport enabled failed.\n");
+ return -1;
+ }
+ csio_ln_err(ln, "vport enabled\n");
+ return 0;
+ }
+}
+
+static void
+csio_dev_loss_tmo_callbk(struct fc_rport *rport)
+{
+ struct csio_rnode *rn;
+ struct csio_hw *hw;
+ struct csio_lnode *ln;
+
+ rn = *((struct csio_rnode **)rport->dd_data);
+ ln = csio_rnode_to_lnode(rn);
+ hw = csio_lnode_to_hw(ln);
+
+ spin_lock_irq(&hw->lock);
+
+ /* return if driver is being removed or same rnode comes back online */
+ if (csio_is_hw_removing(hw) || csio_is_rnode_ready(rn)) {
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ csio_ln_dbg(ln, "devloss timeout on rnode:%p portid:x%x flowid:x%x\n",
+ rn, rn->nport_id, csio_rn_flowid(rn));
+
+ csio_inc_stats(ln, n_dev_loss_tmo);
+
+ /*
+ * enqueue devloss event to event worker thread to serialize all
+ * rnode events.
+ */
+ if (csio_enqueue_evt(hw, CSIO_EVT_DEV_LOSS, &rn, sizeof(rn))) {
+ csio_inc_stats(hw, n_evt_drop);
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ if (!(hw->flags & CSIO_HWF_FWEVT_PENDING)) {
+ hw->flags |= CSIO_HWF_FWEVT_PENDING;
+ spin_unlock_irq(&hw->lock);
+ schedule_work(&hw->evtq_work);
+ return;
+ }
+
+ spin_unlock_irq(&hw->lock);
+}
+
+/* FC transport functions template - Physical port */
+struct fc_function_template csio_fc_transport_funcs = {
+ .show_host_node_name = 1,
+ .show_host_port_name = 1,
+ .show_host_supported_classes = 1,
+ .show_host_supported_fc4s = 1,
+ .show_host_maxframe_size = 1,
+
+ .get_host_port_id = csio_get_host_port_id,
+ .show_host_port_id = 1,
+
+ .get_host_port_type = csio_get_host_port_type,
+ .show_host_port_type = 1,
+
+ .get_host_port_state = csio_get_host_port_state,
+ .show_host_port_state = 1,
+
+ .show_host_active_fc4s = 1,
+ .get_host_speed = csio_get_host_speed,
+ .show_host_speed = 1,
+ .get_host_fabric_name = csio_get_host_fabric_name,
+ .show_host_fabric_name = 1,
+
+ .get_fc_host_stats = csio_get_stats,
+
+ .dd_fcrport_size = sizeof(struct csio_rnode *),
+ .show_rport_maxframe_size = 1,
+ .show_rport_supported_classes = 1,
+
+ .set_rport_dev_loss_tmo = csio_set_rport_loss_tmo,
+ .show_rport_dev_loss_tmo = 1,
+
+ .show_starget_port_id = 1,
+ .show_starget_node_name = 1,
+ .show_starget_port_name = 1,
+
+ .dev_loss_tmo_callbk = csio_dev_loss_tmo_callbk,
+ .dd_fcvport_size = sizeof(struct csio_lnode *),
+
+ .vport_create = csio_vport_create,
+ .vport_disable = csio_vport_disable,
+ .vport_delete = csio_vport_delete,
+};
+
+/* FC transport functions template - Virtual port */
+struct fc_function_template csio_fc_transport_vport_funcs = {
+ .show_host_node_name = 1,
+ .show_host_port_name = 1,
+ .show_host_supported_classes = 1,
+ .show_host_supported_fc4s = 1,
+ .show_host_maxframe_size = 1,
+
+ .get_host_port_id = csio_get_host_port_id,
+ .show_host_port_id = 1,
+
+ .get_host_port_type = csio_get_host_port_type,
+ .show_host_port_type = 1,
+
+ .get_host_port_state = csio_get_host_port_state,
+ .show_host_port_state = 1,
+ .show_host_active_fc4s = 1,
+
+ .get_host_speed = csio_get_host_speed,
+ .show_host_speed = 1,
+
+ .get_host_fabric_name = csio_get_host_fabric_name,
+ .show_host_fabric_name = 1,
+
+ .get_fc_host_stats = csio_get_stats,
+
+ .dd_fcrport_size = sizeof(struct csio_rnode *),
+ .show_rport_maxframe_size = 1,
+ .show_rport_supported_classes = 1,
+
+ .set_rport_dev_loss_tmo = csio_set_rport_loss_tmo,
+ .show_rport_dev_loss_tmo = 1,
+
+ .show_starget_port_id = 1,
+ .show_starget_node_name = 1,
+ .show_starget_port_name = 1,
+
+ .dev_loss_tmo_callbk = csio_dev_loss_tmo_callbk,
+
+};
diff --git a/drivers/scsi/csiostor/csio_mb.c b/drivers/scsi/csiostor/csio_mb.c
new file mode 100644
index 0000000..128fa1b
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_mb.c
@@ -0,0 +1,1768 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/delay.h>
+#include <linux/jiffies.h>
+#include <linux/string.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_hw.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+#include "csio_mb.h"
+#include "csio_wr.h"
+
+#define csio_mb_is_host_owner(__owner) ((__owner) == CSIO_MBOWNER_PL)
+
+/* MB Command/Response Helpers */
+/*
+ * csio_mb_fw_retval - FW return value from a mailbox response.
+ * @mbp: Mailbox structure
+ *
+ */
+enum fw_retval
+csio_mb_fw_retval(struct csio_mb *mbp)
+{
+ struct fw_cmd_hdr *hdr;
+
+ hdr = (struct fw_cmd_hdr *)(mbp->mb);
+
+ return FW_CMD_RETVAL_GET(ntohl(hdr->lo));
+}
+
+/*
+ * csio_mb_hello - FW HELLO command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @m_mbox: Master mailbox number, if any.
+ * @a_mbox: Mailbox number for asycn notifications.
+ * @master: Device mastership.
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_hello(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ uint32_t m_mbox, uint32_t a_mbox, enum csio_dev_master master,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_hello_cmd *cmdp = (struct fw_hello_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_HELLO_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->err_to_clearinit = htonl(
+ FW_HELLO_CMD_MASTERDIS(master == CSIO_MASTER_CANT) |
+ FW_HELLO_CMD_MASTERFORCE(master == CSIO_MASTER_MUST) |
+ FW_HELLO_CMD_MBMASTER(master == CSIO_MASTER_MUST ?
+ m_mbox : FW_HELLO_CMD_MBMASTER_MASK) |
+ FW_HELLO_CMD_MBASYNCNOT(a_mbox) |
+ FW_HELLO_CMD_STAGE(FW_HELLO_CMD_STAGE_OS) |
+ FW_HELLO_CMD_CLEARINIT);
+
+}
+
+/*
+ * csio_mb_process_hello_rsp - FW HELLO response processing helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @retval: Mailbox return value from Firmware
+ * @state: State that the function is in.
+ * @mpfn: Master pfn
+ *
+ */
+void
+csio_mb_process_hello_rsp(struct csio_hw *hw, struct csio_mb *mbp,
+ enum fw_retval *retval, enum csio_dev_state *state,
+ uint8_t *mpfn)
+{
+ struct fw_hello_cmd *rsp = (struct fw_hello_cmd *)(mbp->mb);
+ uint32_t value;
+
+ *retval = FW_CMD_RETVAL_GET(ntohl(rsp->retval_len16));
+
+ if (*retval == FW_SUCCESS) {
+ hw->fwrev = ntohl(rsp->fwrev);
+
+ value = ntohl(rsp->err_to_clearinit);
+ *mpfn = FW_HELLO_CMD_MBMASTER_GET(value);
+
+ if (value & FW_HELLO_CMD_INIT)
+ *state = CSIO_DEV_STATE_INIT;
+ else if (value & FW_HELLO_CMD_ERR)
+ *state = CSIO_DEV_STATE_ERR;
+ else
+ *state = CSIO_DEV_STATE_UNINIT;
+ }
+}
+
+/*
+ * csio_mb_bye - FW BYE command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_bye(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_bye_cmd *cmdp = (struct fw_bye_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_BYE_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+}
+
+/*
+ * csio_mb_reset - FW RESET command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @reset: Type of reset.
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_reset(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ int reset, int halt,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_reset_cmd *cmdp = (struct fw_reset_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_RESET_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->val = htonl(reset);
+ cmdp->halt_pkd = htonl(halt);
+
+}
+
+/*
+ * csio_mb_params - FW PARAMS command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @tmo: Command timeout.
+ * @pf: PF number.
+ * @vf: VF number.
+ * @nparams: Number of paramters
+ * @params: Parameter mnemonic array.
+ * @val: Parameter value array.
+ * @wr: Write/Read PARAMS.
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_params(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ unsigned int pf, unsigned int vf, unsigned int nparams,
+ const u32 *params, u32 *val, bool wr,
+ void (*cbfn)(struct csio_hw *, struct csio_mb *))
+{
+ uint32_t i;
+ uint32_t temp_params = 0, temp_val = 0;
+ struct fw_params_cmd *cmdp = (struct fw_params_cmd *)(mbp->mb);
+ __be32 *p = &cmdp->param[0].mnem;
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_PARAMS_CMD) |
+ FW_CMD_REQUEST |
+ (wr ? FW_CMD_WRITE : FW_CMD_READ) |
+ FW_PARAMS_CMD_PFN(pf) |
+ FW_PARAMS_CMD_VFN(vf));
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ /* Write Params */
+ if (wr) {
+ while (nparams--) {
+ temp_params = *params++;
+ temp_val = *val++;
+
+ *p++ = htonl(temp_params);
+ *p++ = htonl(temp_val);
+ }
+ } else {
+ for (i = 0; i < nparams; i++, p += 2) {
+ temp_params = *params++;
+ *p = htonl(temp_params);
+ }
+ }
+
+}
+
+/*
+ * csio_mb_process_read_params_rsp - FW PARAMS response processing helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @retval: Mailbox return value from Firmware
+ * @nparams: Number of parameters
+ * @val: Parameter value array.
+ *
+ */
+void
+csio_mb_process_read_params_rsp(struct csio_hw *hw, struct csio_mb *mbp,
+ enum fw_retval *retval, unsigned int nparams,
+ u32 *val)
+{
+ struct fw_params_cmd *rsp = (struct fw_params_cmd *)(mbp->mb);
+ uint32_t i;
+ __be32 *p = &rsp->param[0].val;
+
+ *retval = FW_CMD_RETVAL_GET(ntohl(rsp->retval_len16));
+
+ if (*retval == FW_SUCCESS)
+ for (i = 0; i < nparams; i++, p += 2)
+ *val++ = ntohl(*p);
+}
+
+/*
+ * csio_mb_ldst - FW LDST command
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @tmo: timeout
+ * @reg: register
+ *
+ */
+void
+csio_mb_ldst(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo, int reg)
+{
+ struct fw_ldst_cmd *ldst_cmd = (struct fw_ldst_cmd *)(mbp->mb);
+ CSIO_INIT_MBP(mbp, ldst_cmd, tmo, hw, NULL, 1);
+
+ /*
+ * Construct and send the Firmware LDST Command to retrieve the
+ * specified PCI-E Configuration Space register.
+ */
+ ldst_cmd->op_to_addrspace =
+ htonl(FW_CMD_OP(FW_LDST_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_READ |
+ FW_LDST_CMD_ADDRSPACE(FW_LDST_ADDRSPC_FUNC_PCIE));
+ ldst_cmd->cycles_to_len16 = htonl(FW_LEN16(struct fw_ldst_cmd));
+ ldst_cmd->u.pcie.select_naccess = FW_LDST_CMD_NACCESS(1);
+ ldst_cmd->u.pcie.ctrl_to_fn =
+ (FW_LDST_CMD_LC | FW_LDST_CMD_FN(hw->pfn));
+ ldst_cmd->u.pcie.r = (uint8_t)reg;
+}
+
+/*
+ *
+ * csio_mb_caps_config - FW Read/Write Capabilities command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @wr: Write if 1, Read if 0
+ * @init: Turn on initiator mode.
+ * @tgt: Turn on target mode.
+ * @cofld: If 1, Control Offload for FCoE
+ * @cbfn: Callback, if any.
+ *
+ * This helper assumes that cmdp has MB payload from a previous CAPS
+ * read command.
+ */
+void
+csio_mb_caps_config(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ bool wr, bool init, bool tgt, bool cofld,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_caps_config_cmd *cmdp =
+ (struct fw_caps_config_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, wr ? 0 : 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_CAPS_CONFIG_CMD) |
+ FW_CMD_REQUEST |
+ (wr ? FW_CMD_WRITE : FW_CMD_READ));
+ cmdp->cfvalid_to_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ /* Read config */
+ if (!wr)
+ return;
+
+ /* Write config */
+ cmdp->fcoecaps = 0;
+
+ if (cofld)
+ cmdp->fcoecaps |= htons(FW_CAPS_CONFIG_FCOE_CTRL_OFLD);
+ if (init)
+ cmdp->fcoecaps |= htons(FW_CAPS_CONFIG_FCOE_INITIATOR);
+ if (tgt)
+ cmdp->fcoecaps |= htons(FW_CAPS_CONFIG_FCOE_TARGET);
+}
+
+void
+csio_rss_glb_config(struct csio_hw *hw, struct csio_mb *mbp,
+ uint32_t tmo, uint8_t mode, unsigned int flags,
+ void (*cbfn)(struct csio_hw *, struct csio_mb *))
+{
+ struct fw_rss_glb_config_cmd *cmdp =
+ (struct fw_rss_glb_config_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_RSS_GLB_CONFIG_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ if (mode == FW_RSS_GLB_CONFIG_CMD_MODE_MANUAL) {
+ cmdp->u.manual.mode_pkd =
+ htonl(FW_RSS_GLB_CONFIG_CMD_MODE(mode));
+ } else if (mode == FW_RSS_GLB_CONFIG_CMD_MODE_BASICVIRTUAL) {
+ cmdp->u.basicvirtual.mode_pkd =
+ htonl(FW_RSS_GLB_CONFIG_CMD_MODE(mode));
+ cmdp->u.basicvirtual.synmapen_to_hashtoeplitz = htonl(flags);
+ }
+}
+
+
+/*
+ * csio_mb_pfvf - FW Write PF/VF capabilities command helper.
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @pf:
+ * @vf:
+ * @txq:
+ * @txq_eht_ctrl:
+ * @rxqi:
+ * @rxq:
+ * @tc:
+ * @vi:
+ * @pmask:
+ * @rcaps:
+ * @wxcaps:
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_pfvf(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ unsigned int pf, unsigned int vf, unsigned int txq,
+ unsigned int txq_eth_ctrl, unsigned int rxqi,
+ unsigned int rxq, unsigned int tc, unsigned int vi,
+ unsigned int cmask, unsigned int pmask, unsigned int nexactf,
+ unsigned int rcaps, unsigned int wxcaps,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_pfvf_cmd *cmdp = (struct fw_pfvf_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_PFVF_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_WRITE |
+ FW_PFVF_CMD_PFN(pf) |
+ FW_PFVF_CMD_VFN(vf));
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->niqflint_niq = htonl(FW_PFVF_CMD_NIQFLINT(rxqi) |
+ FW_PFVF_CMD_NIQ(rxq));
+
+ cmdp->type_to_neq = htonl(FW_PFVF_CMD_TYPE |
+ FW_PFVF_CMD_CMASK(cmask) |
+ FW_PFVF_CMD_PMASK(pmask) |
+ FW_PFVF_CMD_NEQ(txq));
+ cmdp->tc_to_nexactf = htonl(FW_PFVF_CMD_TC(tc) |
+ FW_PFVF_CMD_NVI(vi) |
+ FW_PFVF_CMD_NEXACTF(nexactf));
+ cmdp->r_caps_to_nethctrl = htonl(FW_PFVF_CMD_R_CAPS(rcaps) |
+ FW_PFVF_CMD_WX_CAPS(wxcaps) |
+ FW_PFVF_CMD_NETHCTRL(txq_eth_ctrl));
+}
+
+#define CSIO_ADVERT_MASK (FW_PORT_CAP_SPEED_100M | FW_PORT_CAP_SPEED_1G |\
+ FW_PORT_CAP_SPEED_10G | FW_PORT_CAP_ANEG)
+
+/*
+ * csio_mb_port- FW PORT command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @tmo: COmmand timeout
+ * @portid: Port ID to get/set info
+ * @wr: Write/Read PORT information.
+ * @fc: Flow control
+ * @caps: Port capabilites to set.
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_port(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ uint8_t portid, bool wr, uint32_t fc, uint16_t caps,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_port_cmd *cmdp = (struct fw_port_cmd *)(mbp->mb);
+ unsigned int lfc = 0, mdi = FW_PORT_MDI(FW_PORT_MDI_AUTO);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_portid = htonl(FW_CMD_OP(FW_PORT_CMD) |
+ FW_CMD_REQUEST |
+ (wr ? FW_CMD_EXEC : FW_CMD_READ) |
+ FW_PORT_CMD_PORTID(portid));
+ if (!wr) {
+ cmdp->action_to_len16 = htonl(
+ FW_PORT_CMD_ACTION(FW_PORT_ACTION_GET_PORT_INFO) |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ return;
+ }
+
+ /* Set port */
+ cmdp->action_to_len16 = htonl(
+ FW_PORT_CMD_ACTION(FW_PORT_ACTION_L1_CFG) |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ if (fc & PAUSE_RX)
+ lfc |= FW_PORT_CAP_FC_RX;
+ if (fc & PAUSE_TX)
+ lfc |= FW_PORT_CAP_FC_TX;
+
+ if (!(caps & FW_PORT_CAP_ANEG))
+ cmdp->u.l1cfg.rcap = htonl((caps & CSIO_ADVERT_MASK) | lfc);
+ else
+ cmdp->u.l1cfg.rcap = htonl((caps & CSIO_ADVERT_MASK) |
+ lfc | mdi);
+}
+
+/*
+ * csio_mb_process_read_port_rsp - FW PORT command response processing helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @retval: Mailbox return value from Firmware
+ * @caps: port capabilities
+ *
+ */
+void
+csio_mb_process_read_port_rsp(struct csio_hw *hw, struct csio_mb *mbp,
+ enum fw_retval *retval, uint16_t *caps)
+{
+ struct fw_port_cmd *rsp = (struct fw_port_cmd *)(mbp->mb);
+
+ *retval = FW_CMD_RETVAL_GET(ntohl(rsp->action_to_len16));
+
+ if (*retval == FW_SUCCESS)
+ *caps = ntohs(rsp->u.info.pcap);
+}
+
+/*
+ * csio_mb_initialize - FW INITIALIZE command helper
+ * @hw: The HW structure
+ * @mbp: Mailbox structure
+ * @tmo: COmmand timeout
+ * @cbfn: Callback, if any.
+ *
+ */
+void
+csio_mb_initialize(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_initialize_cmd *cmdp = (struct fw_initialize_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, tmo, hw, cbfn, 1);
+
+ cmdp->op_to_write = htonl(FW_CMD_OP(FW_INITIALIZE_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE);
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+}
+
+/*
+ * csio_mb_iq_alloc - Initializes the mailbox to allocate an
+ * Ingress DMA queue in the firmware.
+ *
+ * @hw: The hw structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private object
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @iq_params: Ingress queue params needed for allocation.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+static void
+csio_mb_iq_alloc(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_iq_params *iq_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_iq_cmd *cmdp = (struct fw_iq_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_IQ_CMD) |
+ FW_CMD_REQUEST | FW_CMD_EXEC |
+ FW_IQ_CMD_PFN(iq_params->pfn) |
+ FW_IQ_CMD_VFN(iq_params->vfn));
+
+ cmdp->alloc_to_len16 = htonl(FW_IQ_CMD_ALLOC |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ cmdp->type_to_iqandstindex = htonl(
+ FW_IQ_CMD_VIID(iq_params->viid) |
+ FW_IQ_CMD_TYPE(iq_params->type) |
+ FW_IQ_CMD_IQASYNCH(iq_params->iqasynch));
+
+ cmdp->fl0size = htons(iq_params->fl0size);
+ cmdp->fl0size = htons(iq_params->fl1size);
+
+} /* csio_mb_iq_alloc */
+
+/*
+ * csio_mb_iq_write - Initializes the mailbox for writing into an
+ * Ingress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private object
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @cascaded_req: TRUE - if this request is cascased with iq-alloc request.
+ * @iq_params: Ingress queue params needed for writing.
+ * @cbfn: The call-back function
+ *
+ * NOTE: We OR relevant bits with cmdp->XXX, instead of just equating,
+ * because this IQ write request can be cascaded with a previous
+ * IQ alloc request, and we dont want to over-write the bits set by
+ * that request. This logic will work even in a non-cascaded case, since the
+ * cmdp structure is zeroed out by CSIO_INIT_MBP.
+ */
+static void
+csio_mb_iq_write(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, bool cascaded_req,
+ struct csio_iq_params *iq_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_iq_cmd *cmdp = (struct fw_iq_cmd *)(mbp->mb);
+
+ uint32_t iq_start_stop = (iq_params->iq_start) ?
+ FW_IQ_CMD_IQSTART(1) :
+ FW_IQ_CMD_IQSTOP(1);
+
+ /*
+ * If this IQ write is cascaded with IQ alloc request, do not
+ * re-initialize with 0's.
+ *
+ */
+ if (!cascaded_req)
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn |= htonl(FW_CMD_OP(FW_IQ_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE |
+ FW_IQ_CMD_PFN(iq_params->pfn) |
+ FW_IQ_CMD_VFN(iq_params->vfn));
+ cmdp->alloc_to_len16 |= htonl(iq_start_stop |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->iqid |= htons(iq_params->iqid);
+ cmdp->fl0id |= htons(iq_params->fl0id);
+ cmdp->fl1id |= htons(iq_params->fl1id);
+ cmdp->type_to_iqandstindex |= htonl(
+ FW_IQ_CMD_IQANDST(iq_params->iqandst) |
+ FW_IQ_CMD_IQANUS(iq_params->iqanus) |
+ FW_IQ_CMD_IQANUD(iq_params->iqanud) |
+ FW_IQ_CMD_IQANDSTINDEX(iq_params->iqandstindex));
+ cmdp->iqdroprss_to_iqesize |= htons(
+ FW_IQ_CMD_IQPCIECH(iq_params->iqpciech) |
+ FW_IQ_CMD_IQDCAEN(iq_params->iqdcaen) |
+ FW_IQ_CMD_IQDCACPU(iq_params->iqdcacpu) |
+ FW_IQ_CMD_IQINTCNTTHRESH(iq_params->iqintcntthresh) |
+ FW_IQ_CMD_IQCPRIO(iq_params->iqcprio) |
+ FW_IQ_CMD_IQESIZE(iq_params->iqesize));
+
+ cmdp->iqsize |= htons(iq_params->iqsize);
+ cmdp->iqaddr |= cpu_to_be64(iq_params->iqaddr);
+
+ if (iq_params->type == 0) {
+ cmdp->iqns_to_fl0congen |= htonl(
+ FW_IQ_CMD_IQFLINTIQHSEN(iq_params->iqflintiqhsen)|
+ FW_IQ_CMD_IQFLINTCONGEN(iq_params->iqflintcongen));
+ }
+
+ if (iq_params->fl0size && iq_params->fl0addr &&
+ (iq_params->fl0id != 0xFFFF)) {
+
+ cmdp->iqns_to_fl0congen |= htonl(
+ FW_IQ_CMD_FL0HOSTFCMODE(iq_params->fl0hostfcmode)|
+ FW_IQ_CMD_FL0CPRIO(iq_params->fl0cprio) |
+ FW_IQ_CMD_FL0PADEN(iq_params->fl0paden) |
+ FW_IQ_CMD_FL0PACKEN(iq_params->fl0packen));
+ cmdp->fl0dcaen_to_fl0cidxfthresh |= htons(
+ FW_IQ_CMD_FL0DCAEN(iq_params->fl0dcaen) |
+ FW_IQ_CMD_FL0DCACPU(iq_params->fl0dcacpu) |
+ FW_IQ_CMD_FL0FBMIN(iq_params->fl0fbmin) |
+ FW_IQ_CMD_FL0FBMAX(iq_params->fl0fbmax) |
+ FW_IQ_CMD_FL0CIDXFTHRESH(iq_params->fl0cidxfthresh));
+ cmdp->fl0size |= htons(iq_params->fl0size);
+ cmdp->fl0addr |= cpu_to_be64(iq_params->fl0addr);
+ }
+} /* csio_mb_iq_write */
+
+/*
+ * csio_mb_iq_alloc_write - Initializes the mailbox for allocating an
+ * Ingress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @iq_params: Ingress queue params needed for allocation & writing.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_mb_iq_alloc_write(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_iq_params *iq_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ csio_mb_iq_alloc(hw, mbp, priv, mb_tmo, iq_params, cbfn);
+ csio_mb_iq_write(hw, mbp, priv, mb_tmo, CSIO_TRUE, iq_params, cbfn);
+} /* csio_mb_iq_alloc_write */
+
+/*
+ * csio_mb_iq_alloc_write_rsp - Process the allocation & writing
+ * of ingress DMA queue mailbox's response.
+ *
+ * @hw: The HW structure.
+ * @mbp: Mailbox structure to initialize.
+ * @retval: Firmware return value.
+ * @iq_params: Ingress queue parameters, after allocation and write.
+ *
+ */
+void
+csio_mb_iq_alloc_write_rsp(struct csio_hw *hw, struct csio_mb *mbp,
+ enum fw_retval *ret_val,
+ struct csio_iq_params *iq_params)
+{
+ struct fw_iq_cmd *rsp = (struct fw_iq_cmd *)(mbp->mb);
+
+ *ret_val = FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16));
+ if (*ret_val == FW_SUCCESS) {
+ iq_params->physiqid = ntohs(rsp->physiqid);
+ iq_params->iqid = ntohs(rsp->iqid);
+ iq_params->fl0id = ntohs(rsp->fl0id);
+ iq_params->fl1id = ntohs(rsp->fl1id);
+ } else {
+ iq_params->physiqid = iq_params->iqid =
+ iq_params->fl0id = iq_params->fl1id = 0;
+ }
+} /* csio_mb_iq_alloc_write_rsp */
+
+/*
+ * csio_mb_iq_free - Initializes the mailbox for freeing a
+ * specified Ingress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @iq_params: Parameters of ingress queue, that is to be freed.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_mb_iq_free(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_iq_params *iq_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_iq_cmd *cmdp = (struct fw_iq_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_IQ_CMD) |
+ FW_CMD_REQUEST | FW_CMD_EXEC |
+ FW_IQ_CMD_PFN(iq_params->pfn) |
+ FW_IQ_CMD_VFN(iq_params->vfn));
+ cmdp->alloc_to_len16 = htonl(FW_IQ_CMD_FREE |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->type_to_iqandstindex = htonl(FW_IQ_CMD_TYPE(iq_params->type));
+
+ cmdp->iqid = htons(iq_params->iqid);
+ cmdp->fl0id = htons(iq_params->fl0id);
+ cmdp->fl1id = htons(iq_params->fl1id);
+
+} /* csio_mb_iq_free */
+
+/*
+ * csio_mb_eq_ofld_alloc - Initializes the mailbox for allocating
+ * an offload-egress queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @eq_ofld_params: (Offload) Egress queue paramters.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+static void
+csio_mb_eq_ofld_alloc(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_eq_params *eq_ofld_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_eq_ofld_cmd *cmdp = (struct fw_eq_ofld_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) |
+ FW_CMD_REQUEST | FW_CMD_EXEC |
+ FW_EQ_OFLD_CMD_PFN(eq_ofld_params->pfn) |
+ FW_EQ_OFLD_CMD_VFN(eq_ofld_params->vfn));
+ cmdp->alloc_to_len16 = htonl(FW_EQ_OFLD_CMD_ALLOC |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+} /* csio_mb_eq_ofld_alloc */
+
+/*
+ * csio_mb_eq_ofld_write - Initializes the mailbox for writing
+ * an alloacted offload-egress queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @cascaded_req: TRUE - if this request is cascased with Eq-alloc request.
+ * @eq_ofld_params: (Offload) Egress queue paramters.
+ * @cbfn: The call-back function
+ *
+ *
+ * NOTE: We OR relevant bits with cmdp->XXX, instead of just equating,
+ * because this EQ write request can be cascaded with a previous
+ * EQ alloc request, and we dont want to over-write the bits set by
+ * that request. This logic will work even in a non-cascaded case, since the
+ * cmdp structure is zeroed out by CSIO_INIT_MBP.
+ */
+static void
+csio_mb_eq_ofld_write(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, bool cascaded_req,
+ struct csio_eq_params *eq_ofld_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_eq_ofld_cmd *cmdp = (struct fw_eq_ofld_cmd *)(mbp->mb);
+
+ uint32_t eq_start_stop = (eq_ofld_params->eqstart) ?
+ FW_EQ_OFLD_CMD_EQSTART : FW_EQ_OFLD_CMD_EQSTOP;
+
+ /*
+ * If this EQ write is cascaded with EQ alloc request, do not
+ * re-initialize with 0's.
+ *
+ */
+ if (!cascaded_req)
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn |= htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) |
+ FW_CMD_REQUEST | FW_CMD_WRITE |
+ FW_EQ_OFLD_CMD_PFN(eq_ofld_params->pfn) |
+ FW_EQ_OFLD_CMD_VFN(eq_ofld_params->vfn));
+ cmdp->alloc_to_len16 |= htonl(eq_start_stop |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ cmdp->eqid_pkd |= htonl(FW_EQ_OFLD_CMD_EQID(eq_ofld_params->eqid));
+
+ cmdp->fetchszm_to_iqid |= htonl(
+ FW_EQ_OFLD_CMD_HOSTFCMODE(eq_ofld_params->hostfcmode) |
+ FW_EQ_OFLD_CMD_CPRIO(eq_ofld_params->cprio) |
+ FW_EQ_OFLD_CMD_PCIECHN(eq_ofld_params->pciechn) |
+ FW_EQ_OFLD_CMD_IQID(eq_ofld_params->iqid));
+
+ cmdp->dcaen_to_eqsize |= htonl(
+ FW_EQ_OFLD_CMD_DCAEN(eq_ofld_params->dcaen) |
+ FW_EQ_OFLD_CMD_DCACPU(eq_ofld_params->dcacpu) |
+ FW_EQ_OFLD_CMD_FBMIN(eq_ofld_params->fbmin) |
+ FW_EQ_OFLD_CMD_FBMAX(eq_ofld_params->fbmax) |
+ FW_EQ_OFLD_CMD_CIDXFTHRESHO(eq_ofld_params->cidxfthresho) |
+ FW_EQ_OFLD_CMD_CIDXFTHRESH(eq_ofld_params->cidxfthresh) |
+ FW_EQ_OFLD_CMD_EQSIZE(eq_ofld_params->eqsize));
+
+ cmdp->eqaddr |= cpu_to_be64(eq_ofld_params->eqaddr);
+
+} /* csio_mb_eq_ofld_write */
+
+/*
+ * csio_mb_eq_ofld_alloc_write - Initializes the mailbox for allocation
+ * writing into an Engress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @eq_ofld_params: (Offload) Egress queue paramters.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_mb_eq_ofld_alloc_write(struct csio_hw *hw, struct csio_mb *mbp,
+ void *priv, uint32_t mb_tmo,
+ struct csio_eq_params *eq_ofld_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ csio_mb_eq_ofld_alloc(hw, mbp, priv, mb_tmo, eq_ofld_params, cbfn);
+ csio_mb_eq_ofld_write(hw, mbp, priv, mb_tmo, CSIO_TRUE,
+ eq_ofld_params, cbfn);
+} /* csio_mb_eq_ofld_alloc_write */
+
+/*
+ * csio_mb_eq_ofld_alloc_write_rsp - Process the allocation
+ * & write egress DMA queue mailbox's response.
+ *
+ * @hw: The HW structure.
+ * @mbp: Mailbox structure to initialize.
+ * @retval: Firmware return value.
+ * @eq_ofld_params: (Offload) Egress queue paramters.
+ *
+ */
+void
+csio_mb_eq_ofld_alloc_write_rsp(struct csio_hw *hw,
+ struct csio_mb *mbp, enum fw_retval *ret_val,
+ struct csio_eq_params *eq_ofld_params)
+{
+ struct fw_eq_ofld_cmd *rsp = (struct fw_eq_ofld_cmd *)(mbp->mb);
+
+ *ret_val = FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16));
+
+ if (*ret_val == FW_SUCCESS) {
+ eq_ofld_params->eqid = FW_EQ_OFLD_CMD_EQID_GET(
+ ntohl(rsp->eqid_pkd));
+ eq_ofld_params->physeqid = FW_EQ_OFLD_CMD_PHYSEQID_GET(
+ ntohl(rsp->physeqid_pkd));
+ } else
+ eq_ofld_params->eqid = 0;
+
+} /* csio_mb_eq_ofld_alloc_write_rsp */
+
+/*
+ * csio_mb_eq_ofld_free - Initializes the mailbox for freeing a
+ * specified Engress DMA Queue.
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @priv: Private data area.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @eq_ofld_params: (Offload) Egress queue paramters, that is to be freed.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_mb_eq_ofld_free(struct csio_hw *hw, struct csio_mb *mbp, void *priv,
+ uint32_t mb_tmo, struct csio_eq_params *eq_ofld_params,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_eq_ofld_cmd *cmdp = (struct fw_eq_ofld_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, priv, cbfn, 1);
+
+ cmdp->op_to_vfn = htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) |
+ FW_CMD_REQUEST | FW_CMD_EXEC |
+ FW_EQ_OFLD_CMD_PFN(eq_ofld_params->pfn) |
+ FW_EQ_OFLD_CMD_VFN(eq_ofld_params->vfn));
+ cmdp->alloc_to_len16 = htonl(FW_EQ_OFLD_CMD_FREE |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->eqid_pkd = htonl(FW_EQ_OFLD_CMD_EQID(eq_ofld_params->eqid));
+
+} /* csio_mb_eq_ofld_free */
+
+/*
+ * csio_write_fcoe_link_cond_init_mb - Initialize Mailbox to write FCoE link
+ * condition.
+ *
+ * @ln: The Lnode structure
+ * @mbp: Mailbox structure to initialize
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @cbfn: The call back function.
+ *
+ *
+ */
+void
+csio_write_fcoe_link_cond_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint8_t port_id, uint32_t sub_opcode,
+ uint8_t cos, bool link_status, uint32_t fcfi,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_link_cmd *cmdp =
+ (struct fw_fcoe_link_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+
+ cmdp->op_to_portid = htonl((
+ FW_CMD_OP(FW_FCOE_LINK_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_WRITE |
+ FW_FCOE_LINK_CMD_PORTID(port_id)));
+ cmdp->sub_opcode_fcfi = htonl(
+ FW_FCOE_LINK_CMD_SUB_OPCODE(sub_opcode) |
+ FW_FCOE_LINK_CMD_FCFI(fcfi));
+ cmdp->lstatus = link_status;
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+} /* csio_write_fcoe_link_cond_init_mb */
+
+/*
+ * csio_fcoe_read_res_info_init_mb - Initializes the mailbox for reading FCoE
+ * resource information(FW_GET_RES_INFO_CMD).
+ *
+ * @hw: The HW structure
+ * @mbp: Mailbox structure to initialize
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_fcoe_read_res_info_init_mb(struct csio_hw *hw, struct csio_mb *mbp,
+ uint32_t mb_tmo,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_res_info_cmd *cmdp =
+ (struct fw_fcoe_res_info_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, hw, cbfn, 1);
+
+ cmdp->op_to_read = htonl((FW_CMD_OP(FW_FCOE_RES_INFO_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_READ));
+
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+} /* csio_fcoe_read_res_info_init_mb */
+
+/*
+ * csio_fcoe_vnp_alloc_init_mb - Initializes the mailbox for allocating VNP
+ * in the firmware (FW_FCOE_VNP_CMD).
+ *
+ * @ln: The Lnode structure.
+ * @mbp: Mailbox structure to initialize.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @fcfi: FCF Index.
+ * @vnpi: vnpi
+ * @iqid: iqid
+ * @vnport_wwnn: vnport WWNN
+ * @vnport_wwpn: vnport WWPN
+ * @cbfn: The call-back function.
+ *
+ *
+ */
+void
+csio_fcoe_vnp_alloc_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint32_t fcfi, uint32_t vnpi, uint16_t iqid,
+ uint8_t vnport_wwnn[8], uint8_t vnport_wwpn[8],
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_vnp_cmd *cmdp =
+ (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+
+ cmdp->op_to_fcfi = htonl((FW_CMD_OP(FW_FCOE_VNP_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_EXEC |
+ FW_FCOE_VNP_CMD_FCFI(fcfi)));
+
+ cmdp->alloc_to_len16 = htonl(FW_FCOE_VNP_CMD_ALLOC |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+ cmdp->gen_wwn_to_vnpi = htonl(FW_FCOE_VNP_CMD_VNPI(vnpi));
+
+ cmdp->iqid = htons(iqid);
+
+ if (!wwn_to_u64(vnport_wwnn) && !wwn_to_u64(vnport_wwpn))
+ cmdp->gen_wwn_to_vnpi |= htonl(FW_FCOE_VNP_CMD_GEN_WWN);
+
+ if (vnport_wwnn)
+ memcpy(cmdp->vnport_wwnn, vnport_wwnn, 8);
+ if (vnport_wwpn)
+ memcpy(cmdp->vnport_wwpn, vnport_wwpn, 8);
+
+} /* csio_fcoe_vnp_alloc_init_mb */
+
+/*
+ * csio_fcoe_vnp_read_init_mb - Prepares VNP read cmd.
+ * @ln: The Lnode structure.
+ * @mbp: Mailbox structure to initialize.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @fcfi: FCF Index.
+ * @vnpi: vnpi
+ * @cbfn: The call-back handler.
+ */
+void
+csio_fcoe_vnp_read_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint32_t fcfi, uint32_t vnpi,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_vnp_cmd *cmdp =
+ (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+ cmdp->op_to_fcfi = htonl(FW_CMD_OP(FW_FCOE_VNP_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_READ |
+ FW_FCOE_VNP_CMD_FCFI(fcfi));
+ cmdp->alloc_to_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->gen_wwn_to_vnpi = htonl(FW_FCOE_VNP_CMD_VNPI(vnpi));
+}
+
+/*
+ * csio_fcoe_vnp_free_init_mb - Initializes the mailbox for freeing an
+ * alloacted VNP in the firmware (FW_FCOE_VNP_CMD).
+ *
+ * @ln: The Lnode structure.
+ * @mbp: Mailbox structure to initialize.
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @fcfi: FCF flow id
+ * @vnpi: VNP flow id
+ * @cbfn: The call-back function.
+ * Return: None
+ */
+void
+csio_fcoe_vnp_free_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint32_t fcfi, uint32_t vnpi,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_vnp_cmd *cmdp =
+ (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+
+ cmdp->op_to_fcfi = htonl(FW_CMD_OP(FW_FCOE_VNP_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_EXEC |
+ FW_FCOE_VNP_CMD_FCFI(fcfi));
+ cmdp->alloc_to_len16 = htonl(FW_FCOE_VNP_CMD_FREE |
+ FW_CMD_LEN16(sizeof(*cmdp) / 16));
+ cmdp->gen_wwn_to_vnpi = htonl(FW_FCOE_VNP_CMD_VNPI(vnpi));
+}
+
+/*
+ * csio_fcoe_read_fcf_init_mb - Initializes the mailbox to read the
+ * FCF records.
+ *
+ * @ln: The Lnode structure
+ * @mbp: Mailbox structure to initialize
+ * @mb_tmo: Mailbox time-out period (in ms).
+ * @fcf_params: FC-Forwarder parameters.
+ * @cbfn: The call-back function
+ *
+ *
+ */
+void
+csio_fcoe_read_fcf_init_mb(struct csio_lnode *ln, struct csio_mb *mbp,
+ uint32_t mb_tmo, uint32_t portid, uint32_t fcfi,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct fw_fcoe_fcf_cmd *cmdp =
+ (struct fw_fcoe_fcf_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, ln, cbfn, 1);
+
+ cmdp->op_to_fcfi = htonl(FW_CMD_OP(FW_FCOE_FCF_CMD) |
+ FW_CMD_REQUEST |
+ FW_CMD_READ |
+ FW_FCOE_FCF_CMD_FCFI(fcfi));
+ cmdp->retval_len16 = htonl(FW_CMD_LEN16(sizeof(*cmdp) / 16));
+
+} /* csio_fcoe_read_fcf_init_mb */
+
+void
+csio_fcoe_read_portparams_init_mb(struct csio_hw *hw, struct csio_mb *mbp,
+ uint32_t mb_tmo,
+ struct fw_fcoe_port_cmd_params *portparams,
+ void (*cbfn)(struct csio_hw *,
+ struct csio_mb *))
+{
+ struct fw_fcoe_stats_cmd *cmdp = (struct fw_fcoe_stats_cmd *)(mbp->mb);
+
+ CSIO_INIT_MBP(mbp, cmdp, mb_tmo, hw, cbfn, 1);
+ mbp->mb_size = 64;
+
+ cmdp->op_to_flowid = htonl(FW_CMD_OP(FW_FCOE_STATS_CMD) |
+ FW_CMD_REQUEST | FW_CMD_READ);
+ cmdp->free_to_len16 = htonl(FW_CMD_LEN16(CSIO_MAX_MB_SIZE/16));
+
+ cmdp->u.ctl.nstats_port = FW_FCOE_STATS_CMD_NSTATS(portparams->nstats) |
+ FW_FCOE_STATS_CMD_PORT(portparams->portid);
+
+ cmdp->u.ctl.port_valid_ix = FW_FCOE_STATS_CMD_IX(portparams->idx) |
+ FW_FCOE_STATS_CMD_PORT_VALID;
+
+} /* csio_fcoe_read_portparams_init_mb */
+
+void
+csio_mb_process_portparams_rsp(
+ struct csio_hw *hw,
+ struct csio_mb *mbp,
+ enum fw_retval *retval,
+ struct fw_fcoe_port_cmd_params *portparams,
+ struct fw_fcoe_port_stats *portstats
+ )
+{
+ struct fw_fcoe_stats_cmd *rsp = (struct fw_fcoe_stats_cmd *)(mbp->mb);
+ struct fw_fcoe_port_stats stats;
+ uint8_t *src;
+ uint8_t *dst;
+
+ *retval = FW_CMD_RETVAL_GET(ntohl(rsp->free_to_len16));
+
+ memset(&stats, 0, sizeof(struct fw_fcoe_port_stats));
+
+ if (*retval == FW_SUCCESS) {
+ dst = (uint8_t *)(&stats) + ((portparams->idx - 1) * 8);
+ src = (uint8_t *)rsp + (CSIO_STATS_OFFSET * 8);
+ memcpy(dst, src, (portparams->nstats * 8));
+ if (portparams->idx == 1) {
+ /* Get the first 6 flits from the Mailbox */
+ portstats->tx_bcast_bytes =
+ be64_to_cpu(stats.tx_bcast_bytes);
+ portstats->tx_bcast_frames =
+ be64_to_cpu(stats.tx_bcast_frames);
+ portstats->tx_mcast_bytes =
+ be64_to_cpu(stats.tx_mcast_bytes);
+ portstats->tx_mcast_frames =
+ be64_to_cpu(stats.tx_mcast_frames);
+ portstats->tx_ucast_bytes =
+ be64_to_cpu(stats.tx_ucast_bytes);
+ portstats->tx_ucast_frames =
+ be64_to_cpu(stats.tx_ucast_frames);
+ }
+ if (portparams->idx == 7) {
+ /* Get the second 6 flits from the Mailbox */
+ portstats->tx_drop_frames =
+ be64_to_cpu(stats.tx_drop_frames);
+ portstats->tx_offload_bytes =
+ be64_to_cpu(stats.tx_offload_bytes);
+ portstats->tx_offload_frames =
+ be64_to_cpu(stats.tx_offload_frames);
+#if 0
+ portstats->rx_pf_bytes =
+ be64_to_cpu(stats.rx_pf_bytes);
+ portstats->rx_pf_frames =
+ be64_to_cpu(stats.rx_pf_frames);
+#endif
+ portstats->rx_bcast_bytes =
+ be64_to_cpu(stats.rx_bcast_bytes);
+ portstats->rx_bcast_frames =
+ be64_to_cpu(stats.rx_bcast_frames);
+ portstats->rx_mcast_bytes =
+ be64_to_cpu(stats.rx_mcast_bytes);
+ }
+ if (portparams->idx == 13) {
+ /* Get the last 4 flits from the Mailbox */
+ portstats->rx_mcast_frames =
+ be64_to_cpu(stats.rx_mcast_frames);
+ portstats->rx_ucast_bytes =
+ be64_to_cpu(stats.rx_ucast_bytes);
+ portstats->rx_ucast_frames =
+ be64_to_cpu(stats.rx_ucast_frames);
+ portstats->rx_err_frames =
+ be64_to_cpu(stats.rx_err_frames);
+ }
+ }
+}
+
+/* Entry points/APIs for MB module */
+/*
+ * csio_mb_intr_enable - Enable Interrupts from mailboxes.
+ * @hw: The HW structure
+ *
+ * Enables CIM interrupt bit in appropriate INT_ENABLE registers.
+ */
+void
+csio_mb_intr_enable(struct csio_hw *hw)
+{
+ csio_wr_reg32(hw, MBMSGRDYINTEN(1), MYPF_REG(CIM_PF_HOST_INT_ENABLE));
+ csio_rd_reg32(hw, MYPF_REG(CIM_PF_HOST_INT_ENABLE));
+}
+
+/*
+ * csio_mb_intr_disable - Disable Interrupts from mailboxes.
+ * @hw: The HW structure
+ *
+ * Disable bit in HostInterruptEnable CIM register.
+ */
+void
+csio_mb_intr_disable(struct csio_hw *hw)
+{
+ csio_wr_reg32(hw, MBMSGRDYINTEN(0), MYPF_REG(CIM_PF_HOST_INT_ENABLE));
+ csio_rd_reg32(hw, MYPF_REG(CIM_PF_HOST_INT_ENABLE));
+}
+
+static void
+csio_mb_dump_fw_dbg(struct csio_hw *hw, __be64 *cmd)
+{
+ struct fw_debug_cmd *dbg = (struct fw_debug_cmd *)cmd;
+
+ if ((FW_DEBUG_CMD_TYPE_GET(ntohl(dbg->op_type))) == 1) {
+ csio_info(hw, "FW print message:\n");
+ csio_info(hw, "\tdebug->dprtstridx = %d\n",
+ ntohs(dbg->u.prt.dprtstridx));
+ csio_info(hw, "\tdebug->dprtstrparam0 = 0x%x\n",
+ ntohl(dbg->u.prt.dprtstrparam0));
+ csio_info(hw, "\tdebug->dprtstrparam1 = 0x%x\n",
+ ntohl(dbg->u.prt.dprtstrparam1));
+ csio_info(hw, "\tdebug->dprtstrparam2 = 0x%x\n",
+ ntohl(dbg->u.prt.dprtstrparam2));
+ csio_info(hw, "\tdebug->dprtstrparam3 = 0x%x\n",
+ ntohl(dbg->u.prt.dprtstrparam3));
+ } else {
+ /* This is a FW assertion */
+ csio_fatal(hw, "FW assertion at %.16s:%u, val0 %#x, val1 %#x\n",
+ dbg->u.assert.filename_0_7,
+ ntohl(dbg->u.assert.line),
+ ntohl(dbg->u.assert.x),
+ ntohl(dbg->u.assert.y));
+ }
+}
+
+static void
+csio_mb_debug_cmd_handler(struct csio_hw *hw)
+{
+ int i;
+ __be64 cmd[CSIO_MB_MAX_REGS];
+ uint32_t ctl_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_CTRL);
+ uint32_t data_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_DATA);
+ int size = sizeof(struct fw_debug_cmd);
+
+ /* Copy mailbox data */
+ for (i = 0; i < size; i += 8)
+ cmd[i / 8] = cpu_to_be64(csio_rd_reg64(hw, data_reg + i));
+
+ csio_mb_dump_fw_dbg(hw, cmd);
+
+ /* Notify FW of mailbox by setting owner as UP */
+ csio_wr_reg32(hw, MBMSGVALID | MBINTREQ | MBOWNER(CSIO_MBOWNER_FW),
+ ctl_reg);
+
+ csio_rd_reg32(hw, ctl_reg);
+ wmb();
+}
+
+/*
+ * csio_mb_issue - generic routine for issuing Mailbox commands.
+ * @hw: The HW structure
+ * @mbp: Mailbox command to issue
+ *
+ * Caller should hold hw lock across this call.
+ */
+csio_retval_t
+csio_mb_issue(struct csio_hw *hw, struct csio_mb *mbp)
+{
+ uint32_t owner, ctl;
+ int i;
+ uint32_t ii;
+ __be64 *cmd = mbp->mb;
+ __be64 hdr;
+ struct csio_mbm *mbm = &hw->mbm;
+ uint32_t ctl_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_CTRL);
+ uint32_t data_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_DATA);
+ int size = mbp->mb_size;
+ csio_retval_t rv = CSIO_INVAL;
+ struct fw_cmd_hdr *fw_hdr;
+
+ /* Determine mode */
+ if (mbp->mb_cbfn == NULL) {
+ /* Need to issue/get results in the same context */
+ if (mbp->tmo < CSIO_MB_POLL_FREQ) {
+ csio_err(hw, "Invalid tmo: 0x%x\n", mbp->tmo);
+ goto error_out;
+ }
+ } else if (!csio_is_host_intr_enabled(hw) ||
+ !csio_is_hw_intr_enabled(hw)) {
+ csio_err(hw, "Cannot issue mailbox in interrupt mode 0x%x\n",
+ *((uint8_t *)mbp->mb));
+ goto error_out;
+ }
+
+ if (mbm->mcurrent != NULL) {
+ /* Queue mbox cmd, if another mbox cmd is active */
+ if (mbp->mb_cbfn == NULL) {
+ rv = CSIO_BUSY;
+ csio_dbg(hw, "Couldnt own Mailbox %x op:0x%x\n",
+ hw->pfn, *((uint8_t *)mbp->mb));
+
+ goto error_out;
+ } else {
+ list_add_tail(&mbp->list, &mbm->req_q);
+ csio_inc_stats(mbm, n_activeq);
+
+ return CSIO_SUCCESS;
+ }
+ }
+
+ /* Now get ownership of mailbox */
+ owner = MBOWNER_GET(csio_rd_reg32(hw, ctl_reg));
+
+ if (!csio_mb_is_host_owner(owner)) {
+
+ for (i = 0; (owner == CSIO_MBOWNER_NONE) && (i < 3); i++)
+ owner = MBOWNER_GET(csio_rd_reg32(hw, ctl_reg));
+ /*
+ * Mailbox unavailable. In immediate mode, fail the command.
+ * In other modes, enqueue the request.
+ */
+ if (!csio_mb_is_host_owner(owner)) {
+ if (mbp->mb_cbfn == NULL) {
+ rv = owner ? CSIO_BUSY : CSIO_TIMEOUT;
+
+ csio_dbg(hw,
+ "Couldnt own Mailbox %x op:0x%x "
+ "owner:%x\n",
+ hw->pfn, *((uint8_t *)mbp->mb), owner);
+ goto error_out;
+ } else {
+ if (mbm->mcurrent == NULL) {
+ csio_err(hw,
+ "Couldnt own Mailbox %x "
+ "op:0x%x owner:%x\n",
+ hw->pfn, *((uint8_t *)mbp->mb),
+ owner);
+ csio_err(hw,
+ "No outstanding driver"
+ " mailbox as well\n");
+ goto error_out;
+ }
+ }
+ }
+ }
+
+ /* Mailbox is available, copy mailbox data into it */
+ for (i = 0; i < size; i += 8) {
+ csio_wr_reg64(hw, be64_to_cpu(*cmd), data_reg + i);
+ cmd++;
+ }
+
+ CSIO_DUMP_MB(hw, hw->pfn, data_reg);
+
+ /* Start completion timers in non-immediate modes and notify FW */
+ if (mbp->mb_cbfn != NULL) {
+ mbm->mcurrent = mbp;
+ mod_timer(&mbm->timer, jiffies + msecs_to_jiffies(mbp->tmo));
+ csio_wr_reg32(hw, MBMSGVALID | MBINTREQ |
+ MBOWNER(CSIO_MBOWNER_FW), ctl_reg);
+ } else
+ csio_wr_reg32(hw, MBMSGVALID | MBOWNER(CSIO_MBOWNER_FW),
+ ctl_reg);
+
+ /* Flush posted writes */
+ csio_rd_reg32(hw, ctl_reg);
+ wmb();
+
+ csio_inc_stats(mbm, n_req);
+
+ if (mbp->mb_cbfn)
+ return CSIO_SUCCESS;
+
+ /* Poll for completion in immediate mode */
+ cmd = mbp->mb;
+
+ for (ii = 0; ii < mbp->tmo; ii += CSIO_MB_POLL_FREQ) {
+ mdelay(CSIO_MB_POLL_FREQ);
+
+ /* Check for response */
+ ctl = csio_rd_reg32(hw, ctl_reg);
+ if (csio_mb_is_host_owner(MBOWNER_GET(ctl))) {
+
+ if (!(ctl & MBMSGVALID)) {
+ csio_wr_reg32(hw, 0, ctl_reg);
+ continue;
+ }
+
+ CSIO_DUMP_MB(hw, hw->pfn, data_reg);
+
+ hdr = cpu_to_be64(csio_rd_reg64(hw, data_reg));
+ fw_hdr = (struct fw_cmd_hdr *)&hdr;
+
+ switch (FW_CMD_OP_GET(ntohl(fw_hdr->hi))) {
+ case FW_DEBUG_CMD:
+ csio_mb_debug_cmd_handler(hw);
+ continue;
+ }
+
+ /* Copy response */
+ for (i = 0; i < size; i += 8)
+ *cmd++ = cpu_to_be64(csio_rd_reg64
+ (hw, data_reg + i));
+ csio_wr_reg32(hw, 0, ctl_reg);
+
+ if (FW_CMD_RETVAL_GET(*(mbp->mb)))
+ csio_inc_stats(mbm, n_err);
+
+ csio_inc_stats(mbm, n_rsp);
+ return CSIO_SUCCESS;
+ }
+ }
+
+ csio_inc_stats(mbm, n_tmo);
+
+ csio_err(hw, "Mailbox %x op:0x%x timed out!\n",
+ hw->pfn, *((uint8_t *)cmd));
+
+ return CSIO_TIMEOUT;
+
+error_out:
+ csio_inc_stats(mbm, n_err);
+ return rv;
+}
+
+/*
+ * csio_mb_completions - Completion handler for Mailbox commands
+ * @hw: The HW structure
+ * @cbfn_q: Completion queue.
+ *
+ */
+void
+csio_mb_completions(struct csio_hw *hw, struct list_head *cbfn_q)
+{
+ struct csio_mb *mbp;
+ struct csio_mbm *mbm = &hw->mbm;
+ enum fw_retval rv;
+
+ while (!list_empty(cbfn_q)) {
+ csio_deq_from_head(cbfn_q, &mbp);
+
+ rv = csio_mb_fw_retval(mbp);
+ if ((rv != FW_SUCCESS) && (rv != FW_HOSTERROR))
+ csio_inc_stats(mbm, n_err);
+ else if (rv != FW_HOSTERROR)
+ csio_inc_stats(mbm, n_rsp);
+
+ if (mbp->mb_cbfn)
+ mbp->mb_cbfn(hw, mbp);
+ }
+}
+
+static void
+csio_mb_portmod_changed(struct csio_hw *hw, uint8_t port_id)
+{
+ static char *mod_str[] = {
+ NULL, "LR", "SR", "ER", "TWINAX", "active TWINAX", "LRM"
+ };
+
+ struct csio_pport *port = &hw->pport[port_id];
+
+ if (port->mod_type == FW_PORT_MOD_TYPE_NONE)
+ csio_info(hw, "Port:%d - port module unplugged\n", port_id);
+ else if (port->mod_type < ARRAY_SIZE(mod_str))
+ csio_info(hw, "Port:%d - %s port module inserted\n", port_id,
+ mod_str[port->mod_type]);
+ else if (port->mod_type == FW_PORT_MOD_TYPE_NOTSUPPORTED)
+ csio_info(hw,
+ "Port:%d - unsupported optical port module "
+ "inserted\n", port_id);
+ else if (port->mod_type == FW_PORT_MOD_TYPE_UNKNOWN)
+ csio_info(hw,
+ "Port:%d - unknown port module inserted, forcing "
+ "TWINAX\n", port_id);
+ else if (port->mod_type == FW_PORT_MOD_TYPE_ERROR)
+ csio_info(hw, "Port:%d - transceiver module error\n", port_id);
+ else
+ csio_info(hw, "Port:%d - unknown module type %d inserted\n",
+ port_id, port->mod_type);
+}
+
+csio_retval_t
+csio_mb_fwevt_handler(struct csio_hw *hw, __be64 *cmd)
+{
+ uint8_t opcode = *(uint8_t *)cmd;
+ struct fw_port_cmd *pcmd;
+ uint8_t port_id;
+ uint32_t link_status;
+ uint16_t action;
+ uint8_t mod_type;
+
+ if (opcode == FW_PORT_CMD) {
+ pcmd = (struct fw_port_cmd *)cmd;
+ port_id = FW_PORT_CMD_PORTID_GET(
+ ntohl(pcmd->op_to_portid));
+ action = FW_PORT_CMD_ACTION_GET(
+ ntohl(pcmd->action_to_len16));
+ if (action != FW_PORT_ACTION_GET_PORT_INFO) {
+ csio_err(hw, "Unhandled FW_PORT_CMD action: %u\n",
+ action);
+ return CSIO_INVAL;
+ }
+
+ link_status = ntohl(pcmd->u.info.lstatus_to_modtype);
+ mod_type = FW_PORT_CMD_MODTYPE_GET(link_status);
+
+ hw->pport[port_id].link_status =
+ FW_PORT_CMD_LSTATUS_GET(link_status);
+ hw->pport[port_id].link_speed =
+ FW_PORT_CMD_LSPEED_GET(link_status);
+
+ csio_info(hw, "Port:%x - LINK %s\n", port_id,
+ FW_PORT_CMD_LSTATUS_GET(link_status) ? "UP" : "DOWN");
+
+ if (mod_type != hw->pport[port_id].mod_type) {
+ hw->pport[port_id].mod_type = mod_type;
+ csio_mb_portmod_changed(hw, port_id);
+ }
+ } else if (opcode == FW_DEBUG_CMD) {
+ csio_mb_dump_fw_dbg(hw, cmd);
+ } else {
+ csio_dbg(hw, "Gen MB can't handle op:0x%x on evtq.\n", opcode);
+ return CSIO_INVAL;
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_mb_isr_handler - Handle mailboxes related interrupts.
+ * @hw: The HW structure
+ *
+ * Called from the ISR to handle Mailbox related interrupts.
+ * HW Lock should be held across this call.
+ */
+csio_retval_t
+csio_mb_isr_handler(struct csio_hw *hw)
+{
+ struct csio_mbm *mbm = &hw->mbm;
+ struct csio_mb *mbp = mbm->mcurrent;
+ __be64 *cmd;
+ uint32_t ctl, cim_cause, pl_cause;
+ int i;
+ uint32_t ctl_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_CTRL);
+ uint32_t data_reg = PF_REG(hw->pfn, CIM_PF_MAILBOX_DATA);
+ int size;
+ __be64 hdr;
+ struct fw_cmd_hdr *fw_hdr;
+
+ pl_cause = csio_rd_reg32(hw, MYPF_REG(PL_PF_INT_CAUSE));
+ cim_cause = csio_rd_reg32(hw, MYPF_REG(CIM_PF_HOST_INT_CAUSE));
+
+ if (!(pl_cause & PFCIM) || !(cim_cause & MBMSGRDYINT)) {
+ csio_inc_stats(hw, n_mbint_unexp);
+ return CSIO_INVAL;
+ }
+
+ /*
+ * The cause registers below HAVE to be cleared in the SAME
+ * order as below: The low level cause register followed by
+ * the upper level cause register. In other words, CIM-cause
+ * first followed by PL-Cause next.
+ */
+ csio_wr_reg32(hw, MBMSGRDYINT, MYPF_REG(CIM_PF_HOST_INT_CAUSE));
+ csio_wr_reg32(hw, PFCIM, MYPF_REG(PL_PF_INT_CAUSE));
+
+ ctl = csio_rd_reg32(hw, ctl_reg);
+
+ if (csio_mb_is_host_owner(MBOWNER_GET(ctl))) {
+
+ CSIO_DUMP_MB(hw, hw->pfn, data_reg);
+
+ if (!(ctl & MBMSGVALID)) {
+ csio_warn(hw,
+ "Stray mailbox interrupt recvd,"
+ " mailbox data not valid\n");
+ csio_wr_reg32(hw, 0, ctl_reg);
+ /* Flush */
+ csio_rd_reg32(hw, ctl_reg);
+ return CSIO_INVAL;
+ }
+
+ hdr = cpu_to_be64(csio_rd_reg64(hw, data_reg));
+ fw_hdr = (struct fw_cmd_hdr *)&hdr;
+
+ switch (FW_CMD_OP_GET(ntohl(fw_hdr->hi))) {
+ case FW_DEBUG_CMD:
+ csio_mb_debug_cmd_handler(hw);
+ return CSIO_INVAL;
+#if 0
+ case FW_ERROR_CMD:
+ case FW_INITIALIZE_CMD: /* When we are not master */
+#endif
+ }
+
+ CSIO_ASSERT(mbp != NULL);
+
+ cmd = mbp->mb;
+ size = mbp->mb_size;
+ /* Get response */
+ for (i = 0; i < size; i += 8)
+ *cmd++ = cpu_to_be64(csio_rd_reg64
+ (hw, data_reg + i));
+
+ csio_wr_reg32(hw, 0, ctl_reg);
+ /* Flush */
+ csio_rd_reg32(hw, ctl_reg);
+
+ mbm->mcurrent = NULL;
+
+ /* Add completion to tail of cbfn queue */
+ list_add_tail(&mbp->list, &mbm->cbfn_q);
+ csio_inc_stats(mbm, n_cbfnq);
+
+ /*
+ * Enqueue event to EventQ. Events processing happens
+ * in Event worker thread context
+ */
+ if (csio_enqueue_evt(hw, CSIO_EVT_MBX, mbp, sizeof(mbp)))
+ csio_inc_stats(hw, n_evt_drop);
+
+ return CSIO_SUCCESS;
+
+ } else {
+ /*
+ * We can get here if mailbox MSIX vector is shared,
+ * or in INTx case. Or a stray interrupt.
+ */
+ csio_dbg(hw, "Host not owner, no mailbox interrupt\n");
+ csio_inc_stats(hw, n_int_stray);
+ return CSIO_INVAL;
+ }
+}
+
+/*
+ * csio_mb_tmo_handler - Timeout handler
+ * @hw: The HW structure
+ *
+ */
+struct csio_mb *
+csio_mb_tmo_handler(struct csio_hw *hw)
+{
+ struct csio_mbm *mbm = &hw->mbm;
+ struct csio_mb *mbp = mbm->mcurrent;
+ struct fw_cmd_hdr *fw_hdr;
+
+ /*
+ * Could be a race b/w the completion handler and the timer
+ * and the completion handler won that race.
+ */
+ if (mbp == NULL) {
+ CSIO_DB_ASSERT(0);
+ return NULL;
+ }
+
+ fw_hdr = (struct fw_cmd_hdr *)(mbp->mb);
+
+ csio_dbg(hw, "Mailbox num:%x op:0x%x timed out\n", hw->pfn,
+ FW_CMD_OP_GET(ntohl(fw_hdr->hi)));
+
+ mbm->mcurrent = NULL;
+ csio_inc_stats(mbm, n_tmo);
+ fw_hdr->lo = htonl(FW_CMD_RETVAL(FW_ETIMEDOUT));
+
+ return mbp;
+}
+
+/*
+ * csio_mb_cancel_all - Cancel all waiting commands.
+ * @hw: The HW structure
+ * @cbfn_q: The callback queue.
+ *
+ * Caller should hold hw lock across this call.
+ */
+void
+csio_mb_cancel_all(struct csio_hw *hw, struct list_head *cbfn_q)
+{
+ struct csio_mb *mbp;
+ struct csio_mbm *mbm = &hw->mbm;
+ struct fw_cmd_hdr *hdr;
+ struct list_head *tmp;
+
+ if (mbm->mcurrent) {
+ mbp = mbm->mcurrent;
+
+ /* Stop mailbox completion timer */
+ del_timer_sync(&mbm->timer);
+
+ /* Add completion to tail of cbfn queue */
+ list_add_tail(&mbp->list, cbfn_q);
+ mbm->mcurrent = NULL;
+ }
+
+ if (!list_empty(&mbm->req_q)) {
+ list_splice_tail_init(&mbm->req_q, cbfn_q);
+ mbm->stats.n_activeq = 0;
+ }
+
+ if (!list_empty(&mbm->cbfn_q)) {
+ list_splice_tail_init(&mbm->cbfn_q, cbfn_q);
+ mbm->stats.n_cbfnq = 0;
+ }
+
+ if (list_empty(cbfn_q))
+ return;
+
+ list_for_each(tmp, cbfn_q) {
+ mbp = (struct csio_mb *)tmp;
+ hdr = (struct fw_cmd_hdr *)(mbp->mb);
+
+ csio_dbg(hw, "Cancelling pending mailbox num %x op:%x\n",
+ hw->pfn, FW_CMD_OP_GET(ntohl(hdr->hi)));
+
+ csio_inc_stats(mbm, n_cancel);
+ hdr->lo = htonl(FW_CMD_RETVAL(FW_HOSTERROR));
+ }
+}
+
+/*
+ * csio_mbm_init - Initialize Mailbox module
+ * @mbm: Mailbox module
+ * @hw: The HW structure
+ * @timer: Timing function for interrupting mailboxes
+ *
+ * Initialize timer and the request/response queues.
+ */
+csio_retval_t
+csio_mbm_init(struct csio_mbm *mbm, struct csio_hw *hw,
+ void (*timer_fn)(uintptr_t))
+{
+ struct timer_list *timer = &mbm->timer;
+
+ init_timer(timer);
+ timer->function = timer_fn;
+ timer->data = (unsigned long)hw;
+
+ INIT_LIST_HEAD(&mbm->req_q);
+ INIT_LIST_HEAD(&mbm->cbfn_q);
+ csio_set_mb_intr_idx(mbm, -1);
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_mbm_exit - Uninitialize mailbox module
+ * @mbm: Mailbox module
+ *
+ * Stop timer.
+ */
+void
+csio_mbm_exit(struct csio_mbm *mbm)
+{
+ del_timer_sync(&mbm->timer);
+
+ CSIO_DB_ASSERT(mbm->mcurrent == NULL);
+ CSIO_DB_ASSERT(list_empty(&mbm->req_q));
+ CSIO_DB_ASSERT(list_empty(&mbm->cbfn_q));
+}
--
1.7.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH 4/8] csiostor: Chelsio FCoE offload driver submission (sources part 4).
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
` (2 preceding siblings ...)
2012-08-23 22:27 ` [PATCH 3/8] csiostor: Chelsio FCoE offload driver submission (sources part 3) Naresh Kumar Inna
@ 2012-08-23 22:27 ` Naresh Kumar Inna
2012-08-23 22:27 ` [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5) Naresh Kumar Inna
` (5 subsequent siblings)
9 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
This patch contains code to implement the local and remote node port
functionality. It includes tracking the firmware events for changes to
the states of these ports.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/csiostor/csio_lnode.c | 2151 ++++++++++++++++++++++++++++++++++++
drivers/scsi/csiostor/csio_rnode.c | 889 +++++++++++++++
2 files changed, 3040 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_lnode.c
create mode 100644 drivers/scsi/csiostor/csio_rnode.c
diff --git a/drivers/scsi/csiostor/csio_lnode.c b/drivers/scsi/csiostor/csio_lnode.c
new file mode 100644
index 0000000..5d5a1a1
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_lnode.c
@@ -0,0 +1,2151 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/delay.h>
+#include <linux/slab.h>
+#include <linux/utsname.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_hw.h"
+#include "csio_mb.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+#include "csio_fcoe_proto.h"
+
+int csio_fcoe_rnodes = 512;
+int csio_fdmi_enable = 1;
+
+#define PORT_ID_PTR(_x) ((uint8_t *)(&_x) + 1)
+
+/* Lnode SM declarations */
+static void csio_lns_uninit(struct csio_lnode *, enum csio_ln_ev);
+static void csio_lns_online(struct csio_lnode *, enum csio_ln_ev);
+static void csio_lns_ready(struct csio_lnode *, enum csio_ln_ev);
+static void csio_lns_offline(struct csio_lnode *, enum csio_ln_ev);
+
+static csio_retval_t csio_ln_mgmt_submit_req(struct csio_ioreq *,
+ void (*io_cbfn) (struct csio_hw *, struct csio_ioreq *),
+ enum fcoe_cmn_type, struct csio_dma_buf *, uint32_t);
+
+/* LN event mapping */
+static enum csio_ln_ev fwevt_to_lnevt[] = {
+ CSIO_LNE_NONE, /* None */
+ CSIO_LNE_NONE, /* PLOGI_ACC_RCVD */
+ CSIO_LNE_NONE, /* PLOGI_RJT_RCVD */
+ CSIO_LNE_NONE, /* PLOGI_RCVD */
+ CSIO_LNE_NONE, /* PLOGO_RCVD */
+ CSIO_LNE_NONE, /* PRLI_ACC_RCVD */
+ CSIO_LNE_NONE, /* PRLI_RJT_RCVD */
+ CSIO_LNE_NONE, /* PRLI_RCVD */
+ CSIO_LNE_NONE, /* PRLO_RCVD */
+ CSIO_LNE_NONE, /* NPORT_ID_CHGD */
+ CSIO_LNE_LOGO, /* FLOGO_RCVD */
+ CSIO_LNE_LOGO, /* CLR_VIRT_LNK_RCVD */
+ CSIO_LNE_FAB_INIT_DONE,/* FLOGI_ACC_RCVD */
+ CSIO_LNE_NONE, /* FLOGI_RJT_RCVD */
+ CSIO_LNE_FAB_INIT_DONE,/* FDISC_ACC_RCVD */
+ CSIO_LNE_NONE, /* FDISC_RJT_RCVD */
+ CSIO_LNE_NONE, /* FLOGI_TMO_MAX_RETRY */
+ CSIO_LNE_NONE, /* IMPL_LOGO_ADISC_ACC */
+ CSIO_LNE_NONE, /* IMPL_LOGO_ADISC_RJT */
+ CSIO_LNE_NONE, /* IMPL_LOGO_ADISC_CNFLT */
+ CSIO_LNE_NONE, /* PRLI_TMO */
+ CSIO_LNE_NONE, /* ADISC_TMO */
+ CSIO_LNE_NONE, /* RSCN_DEV_LOST */
+ CSIO_LNE_NONE, /* SCR_ACC_RCVD */
+ CSIO_LNE_NONE, /* ADISC_RJT_RCVD */
+ CSIO_LNE_NONE, /* LOGO_SNT */
+ CSIO_LNE_NONE, /* PROTO_ERR_IMPL_LOGO */
+};
+
+#define CSIO_FWE_TO_LNE(_evt) ((_evt > PROTO_ERR_IMPL_LOGO) ? \
+ CSIO_LNE_NONE : \
+ fwevt_to_lnevt[_evt])
+/*
+ * csio_ln_match_by_portid - lookup lnode using given portid.
+ * @hw: HW module
+ * @portid: port-id.
+ *
+ * If found, returns lnode matching given portid otherwise returns NULL.
+ */
+static struct csio_lnode *
+csio_ln_lookup_by_portid(struct csio_hw *hw, uint8_t portid)
+{
+ struct csio_lnode *ln = hw->rln;
+ struct list_head *tmp;
+
+ /* Match siblings lnode with portid */
+ list_for_each(tmp, &hw->sln_head) {
+ ln = (struct csio_lnode *) tmp;
+ if (ln->portid == portid)
+ return ln;
+ }
+
+ return NULL;
+}
+
+/*
+ * csio_ln_lookup_by_vnpi - Lookup lnode using given vnp id.
+ * @hw - HW module
+ * @vnpi - vnp index.
+ * Returns - If found, returns lnode matching given vnp id
+ * otherwise returns NULL.
+ */
+static struct csio_lnode *
+csio_ln_lookup_by_vnpi(struct csio_hw *hw, uint32_t vnp_id)
+{
+ struct list_head *tmp1, *tmp2;
+ struct csio_lnode *sln = NULL, *cln = NULL;
+
+ if (list_empty(&hw->sln_head)) {
+ csio_inc_stats(hw, n_lnlkup_miss);
+ return NULL;
+ }
+ /* Traverse sibling lnodes */
+ list_for_each(tmp1, &hw->sln_head) {
+ sln = (struct csio_lnode *) tmp1;
+
+ /* Match sibling lnode */
+ if (sln->vnp_flowid == vnp_id)
+ return sln;
+
+ if (list_empty(&sln->cln_head))
+ continue;
+
+ /* Traverse children lnodes */
+ list_for_each(tmp2, &sln->cln_head) {
+ cln = (struct csio_lnode *) tmp2;
+
+ if (cln->vnp_flowid == vnp_id)
+ return cln;
+ }
+ }
+ csio_inc_stats(hw, n_lnlkup_miss);
+ return NULL;
+}
+
+/**
+ * csio_lnode_lookup_by_wwpn - Lookup lnode using given wwpn.
+ * @hw: HW module.
+ * @wwpn: WWPN.
+ *
+ * If found, returns lnode matching given wwpn, returns NULL otherwise.
+ */
+struct csio_lnode *
+csio_lnode_lookup_by_wwpn(struct csio_hw *hw, uint8_t *wwpn)
+{
+ struct list_head *tmp1, *tmp2;
+ struct csio_lnode *sln = NULL, *cln = NULL;
+
+ if (list_empty(&hw->sln_head)) {
+ csio_inc_stats(hw, n_lnlkup_miss);
+ return NULL;
+ }
+ /* Traverse sibling lnodes */
+ list_for_each(tmp1, &hw->sln_head) {
+ sln = (struct csio_lnode *) tmp1;
+
+ /* Match sibling lnode */
+ if (!memcmp(csio_ln_wwpn(sln), wwpn, 8))
+ return sln;
+
+ if (list_empty(&sln->cln_head))
+ continue;
+
+ /* Traverse children lnodes */
+ list_for_each(tmp2, &sln->cln_head) {
+ cln = (struct csio_lnode *) tmp2;
+
+ if (!memcmp(csio_ln_wwpn(cln), wwpn, 8))
+ return cln;
+ }
+ }
+ return NULL;
+}
+
+static int
+csio_hostname(uint8_t *buf, size_t buf_len)
+{
+ if (sprintf(buf, "%s", init_utsname()->nodename))
+ return 0;
+ return -1;
+}
+
+static int
+csio_osname(uint8_t *buf, size_t buf_len)
+{
+ uint8_t *ptr = buf;
+
+ strcpy(ptr, init_utsname()->sysname);
+ ptr += strlen(init_utsname()->sysname);
+ *ptr = ' '; /* SPACE */
+ strcpy(ptr, init_utsname()->release);
+ ptr += strlen(init_utsname()->release);
+ *ptr = ' '; /* SPACE */
+ strcpy(ptr, init_utsname()->version);
+ ptr += strlen(init_utsname()->version);
+ *ptr = '\0';
+ return 0;
+}
+
+static inline void
+csio_append_string_attrib(uint8_t **ptr, uint16_t type,
+ uint8_t *val, uint16_t len)
+{
+ struct csio_attrib_entry *ae = (struct csio_attrib_entry *) *ptr;
+ ae->type = htons(type);
+ len += 4; /* includes attribute type and length */
+ len = (len + 3) & ~3; /* should be multiple of 4 bytes */
+ ae->len = htons(len);
+ memset(ae->val.string, 0, len - 4);
+ memcpy(ae->val.string, val, len);
+ *ptr += len;
+}
+
+static inline void
+csio_append_int_attrib(uint8_t **ptr, uint16_t type,
+ uint32_t val, uint16_t len)
+{
+ struct csio_attrib_entry *ae = (struct csio_attrib_entry *) *ptr;
+ ae->type = htons(type);
+ len += 4; /* includes attribute type and length */
+ len = (len + 3) & ~3; /* should be multiple of 4 bytes */
+ ae->len = htons(len);
+ ae->val.integer = htonl(val);
+ *ptr += len;
+}
+
+/*
+ * csio_ln_fdmi_done - FDMI registeration completion
+ * @hw: HW context
+ * @fdmi_req: fdmi request
+ */
+static void
+csio_ln_fdmi_done(struct csio_hw *hw, struct csio_ioreq *fdmi_req)
+{
+ void *cmd;
+ struct csio_lnode *ln = fdmi_req->lnode;
+
+ if (fdmi_req->wr_status != FW_SUCCESS) {
+ csio_ln_err(ln, "WR error:%x in processing fdmi rpa cmd\n",
+ fdmi_req->wr_status);
+ csio_inc_stats(ln, n_fdmi_err);
+ }
+
+ cmd = fdmi_req->dma_buf.vaddr;
+ if (ntohs(csio_ct_rsp(cmd)) != CT_RESPONSE_FS_ACC) {
+ csio_ln_dbg(ln, "fdmi rpa cmd rejected reason %x expl %x\n",
+ csio_ct_reason(cmd), csio_ct_expl(cmd));
+ }
+}
+
+/*
+ * csio_ln_fdmi_rhba_cbfn - RHBA completion
+ * @hw: HW context
+ * @fdmi_req: fdmi request
+ */
+static void
+csio_ln_fdmi_rhba_cbfn(struct csio_hw *hw, struct csio_ioreq *fdmi_req)
+{
+ void *cmd;
+ uint8_t *pld;
+ uint32_t len = 0;
+ struct csio_lnode *ln = fdmi_req->lnode;
+ struct csio_attrib_block *attrib_blk;
+ uint8_t buf[64];
+ uint32_t val;
+ uint8_t *fc4_type;
+
+ if (fdmi_req->wr_status != FW_SUCCESS) {
+ csio_ln_err(ln, "WR error:%x in processing fdmi rhba cmd\n",
+ fdmi_req->wr_status);
+ csio_inc_stats(ln, n_fdmi_err);
+ }
+
+ cmd = fdmi_req->dma_buf.vaddr;
+ if (ntohs(csio_ct_rsp(cmd)) != CT_RESPONSE_FS_ACC) {
+ csio_ln_dbg(ln, "fdmi rhba cmd rejected reason %x expl %x\n",
+ csio_ct_reason(cmd), csio_ct_expl(cmd));
+ }
+
+ if (!csio_is_rnode_ready(fdmi_req->rnode)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ return;
+ }
+
+ /* Prepare CT hdr for RPA cmd */
+ memset(cmd, 0, CT_BASIC_IU_LEN);
+ csio_fill_ct_iu(cmd,
+ CT_GS_MGMT_SERVICE,
+ CT_FDMI_HBA_MGMT_SERVER,
+ htons(CT_FDMI_HBA_RPA));
+
+ /* Prepare RPA payload */
+ pld = (uint8_t *) csio_ct_get_pld(cmd);
+ memcpy(pld, csio_ln_wwpn(ln), 8); /* Port name */
+ pld += 8;
+
+ /* Start appending Port attributes */
+ attrib_blk = (struct csio_attrib_block *) pld;
+ attrib_blk->entry_count = 0;
+ len += sizeof(attrib_blk->entry_count);
+ pld += sizeof(attrib_blk->entry_count);
+
+ fc4_type = &buf[0];
+ memset(fc4_type, 0, 32);
+ fc4_type[2] = 1;
+ fc4_type[7] = 1;
+ csio_append_string_attrib(&pld, SUPPORTED_FC4_TYPES, fc4_type , 32);
+ attrib_blk->entry_count++;
+ val = CSIO_HBA_PORTSPEED_1GBIT | CSIO_HBA_PORTSPEED_10GBIT;
+ csio_append_int_attrib(&pld, SUPPORTED_SPEED, val , 4);
+ attrib_blk->entry_count++;
+
+ if (hw->pport[ln->portid].link_speed == FW_PORT_CAP_SPEED_1G)
+ val = CSIO_HBA_PORTSPEED_1GBIT;
+ else if (hw->pport[ln->portid].link_speed ==
+ FW_PORT_CAP_SPEED_10G)
+ val = CSIO_HBA_PORTSPEED_10GBIT;
+ else
+ val = CSIO_HBA_PORTSPEED_UNKNOWN;
+ csio_append_int_attrib(&pld, PORT_SPEED, val , 4);
+ attrib_blk->entry_count++;
+ csio_append_int_attrib(&pld, MAX_FRAME_LEN, ln->ln_sparm.csp.rcv_sz,
+ 4);
+ attrib_blk->entry_count++;
+
+ strcpy(buf, "csiostor");
+ csio_append_string_attrib(&pld, OS_DEVICE_NAME, buf,
+ (uint16_t)strlen(buf));
+ attrib_blk->entry_count++;
+
+ if (!csio_hostname(buf, sizeof(buf))) {
+ csio_append_string_attrib(&pld, HOST_NAME , buf,
+ (uint16_t)strlen(buf));
+ attrib_blk->entry_count++;
+ }
+ attrib_blk->entry_count = ntohl(attrib_blk->entry_count);
+ len = (uint32_t)(pld - (uint8_t *)cmd);
+
+ /* Submit FDMI RPA request */
+ spin_lock_irq(&hw->lock);
+ if (csio_ln_mgmt_submit_req(fdmi_req, csio_ln_fdmi_done,
+ FCOE_CT, &fdmi_req->dma_buf,
+ len)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ csio_ln_err(ln, "Failed to issue fdmi rpa req\n");
+ }
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_ln_fdmi_dprt_cbfn - DPRT completion
+ * @hw: HW context
+ * @fdmi_req: fdmi request
+ */
+static void
+csio_ln_fdmi_dprt_cbfn(struct csio_hw *hw, struct csio_ioreq *fdmi_req)
+{
+ void *cmd;
+ uint8_t *pld;
+ uint32_t len = 0;
+ struct csio_hba_identifier *id;
+ struct csio_lnode *ln = fdmi_req->lnode;
+ struct csio_reg_port_list *reg_pl;
+ struct csio_attrib_block *attrib_blk;
+ uint8_t buf[64];
+
+ if (fdmi_req->wr_status != FW_SUCCESS) {
+ csio_ln_err(ln, "WR error:%x in processing fdmi dprt cmd\n",
+ fdmi_req->wr_status);
+ csio_inc_stats(ln, n_fdmi_err);
+ }
+
+ if (!csio_is_rnode_ready(fdmi_req->rnode)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ return;
+ }
+ cmd = fdmi_req->dma_buf.vaddr;
+ if (ntohs(csio_ct_rsp(cmd)) != CT_RESPONSE_FS_ACC) {
+ csio_ln_dbg(ln, "fdmi dprt cmd rejected reason %x expl %x\n",
+ csio_ct_reason(cmd), csio_ct_expl(cmd));
+ }
+
+ /* Prepare CT hdr for RHBA cmd */
+ memset(cmd, 0, CT_BASIC_IU_LEN);
+ csio_fill_ct_iu(cmd,
+ CT_GS_MGMT_SERVICE,
+ CT_FDMI_HBA_MGMT_SERVER,
+ htons(CT_FDMI_HBA_RHBA));
+ len = CT_BASIC_IU_LEN;
+
+ /* Prepare RHBA payload */
+ pld = (uint8_t *) csio_ct_get_pld(cmd);
+ id = (struct csio_hba_identifier *) pld;
+ memcpy(id->wwpn, csio_ln_wwpn(ln), 8); /* HBA identifer */
+ pld += 8;
+
+ /* Register one port per hba */
+ reg_pl = (struct csio_reg_port_list *) pld;
+ reg_pl->entry_count = ntohl(1);
+ memcpy(reg_pl->entry.wwpn, csio_ln_wwpn(ln), 8);
+ pld += sizeof(*reg_pl);
+
+ /* Start appending HBA attributes hba */
+ attrib_blk = (struct csio_attrib_block *) pld;
+ attrib_blk->entry_count = 0;
+ len += sizeof(attrib_blk->entry_count);
+ pld += sizeof(attrib_blk->entry_count);
+
+ csio_append_string_attrib(&pld, NODE_NAME, csio_ln_wwnn(ln), 8);
+ attrib_blk->entry_count++;
+
+ memset(buf, 0, sizeof(buf));
+
+ strcpy(buf, "Chelsio Communications");
+ csio_append_string_attrib(&pld, MANUFACTURER, buf,
+ (uint16_t)strlen(buf));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, SERIAL_NUMBER, hw->vpd.sn,
+ (uint16_t)sizeof(hw->vpd.sn));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, MODEL, hw->vpd.id,
+ (uint16_t)sizeof(hw->vpd.id));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, MODEL_DESCRIPTION, hw->model_desc,
+ (uint16_t)strlen(hw->model_desc));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, HARDWARE_VERSION, hw->hw_ver,
+ (uint16_t)sizeof(hw->hw_ver));
+ attrib_blk->entry_count++;
+ csio_append_string_attrib(&pld, FIRMWARE_VERSION, hw->fwrev_str,
+ (uint16_t)strlen(hw->fwrev_str));
+ attrib_blk->entry_count++;
+
+ if (!csio_osname(buf, sizeof(buf))) {
+ csio_append_string_attrib(&pld, OS_NAME_VERSION , buf,
+ (uint16_t)strlen(buf));
+ attrib_blk->entry_count++;
+ }
+
+ csio_append_int_attrib(&pld, MAX_CT_PAYLOAD_LEN, 65536, 4);
+ len = (uint32_t)(pld - (uint8_t *)cmd);
+ attrib_blk->entry_count++;
+ attrib_blk->entry_count = ntohl(attrib_blk->entry_count);
+
+ /* Submit FDMI RHBA request */
+ spin_lock_irq(&hw->lock);
+ if (csio_ln_mgmt_submit_req(fdmi_req, csio_ln_fdmi_rhba_cbfn,
+ FCOE_CT, &fdmi_req->dma_buf,
+ len)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ csio_ln_err(ln, "Failed to issue fdmi rhba req\n");
+ }
+ spin_unlock_irq(&hw->lock);
+}
+
+/*
+ * csio_ln_fdmi_dhba_cbfn - DHBA completion
+ * @hw: HW context
+ * @fdmi_req: fdmi request
+ */
+static void
+csio_ln_fdmi_dhba_cbfn(struct csio_hw *hw, struct csio_ioreq *fdmi_req)
+{
+ struct csio_lnode *ln = fdmi_req->lnode;
+ void *cmd;
+ uint8_t *port_name;
+ uint32_t len;
+
+ if (fdmi_req->wr_status != FW_SUCCESS) {
+ csio_ln_err(ln, "WR error:%x in processing fdmi dhba cmd\n",
+ fdmi_req->wr_status);
+ csio_inc_stats(ln, n_fdmi_err);
+ }
+
+ if (!csio_is_rnode_ready(fdmi_req->rnode)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ return;
+ }
+ cmd = fdmi_req->dma_buf.vaddr;
+ if (ntohs(csio_ct_rsp(cmd)) != CT_RESPONSE_FS_ACC) {
+ csio_ln_dbg(ln, "fdmi dhba cmd rejected reason %x expl %x\n",
+ csio_ct_reason(cmd), csio_ct_expl(cmd));
+ }
+
+ /* Send FDMI cmd to de-register any Port attributes if registered
+ * before
+ */
+
+ /* Prepare FDMI DPRT cmd */
+ memset(cmd, 0, CT_BASIC_IU_LEN);
+ csio_fill_ct_iu(cmd,
+ CT_GS_MGMT_SERVICE,
+ CT_FDMI_HBA_MGMT_SERVER,
+ htons(CT_FDMI_HBA_DPRT));
+ len = CT_BASIC_IU_LEN;
+ port_name = (uint8_t *) csio_ct_get_pld(cmd);
+ memcpy(port_name, csio_ln_wwpn(ln), 8);
+ len += 8;
+
+ /* Submit FDMI request */
+ spin_lock_irq(&hw->lock);
+ if (csio_ln_mgmt_submit_req(fdmi_req, csio_ln_fdmi_dprt_cbfn,
+ FCOE_CT, &fdmi_req->dma_buf,
+ len)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ csio_ln_err(ln, "Failed to issue fdmi dprt req\n");
+ }
+ spin_unlock_irq(&hw->lock);
+}
+
+/**
+ * csio_ln_fdmi_start - Start an FDMI request.
+ * @ln: lnode
+ * @context: session context
+ *
+ * Issued with lock held.
+ */
+csio_retval_t
+csio_ln_fdmi_start(struct csio_lnode *ln, void *context)
+{
+ struct csio_ioreq *fdmi_req;
+ struct csio_rnode *fdmi_rn = (struct csio_rnode *) context;
+ void *cmd;
+ struct csio_hba_identifier *id;
+ uint32_t len;
+
+ if (!(ln->flags & CSIO_LNF_FDMI_ENABLE))
+ return CSIO_NOSUPP;
+
+ if (!csio_is_rnode_ready(fdmi_rn))
+ csio_inc_stats(ln, n_fdmi_err);
+
+ /* Send FDMI cmd to de-register any HBA attributes if registered
+ * before
+ */
+
+ fdmi_req = ln->mgmt_req;
+ fdmi_req->lnode = ln;
+ fdmi_req->rnode = fdmi_rn;
+
+ /* Prepare FDMI DHBA cmd */
+ cmd = fdmi_req->dma_buf.vaddr;
+ memset(cmd, 0, CT_BASIC_IU_LEN);
+ csio_fill_ct_iu(cmd,
+ CT_GS_MGMT_SERVICE,
+ CT_FDMI_HBA_MGMT_SERVER,
+ htons(CT_FDMI_HBA_DHBA));
+ len = CT_BASIC_IU_LEN;
+
+ id = (struct csio_hba_identifier *) csio_ct_get_pld(cmd);
+ memcpy(id->wwpn, csio_ln_wwpn(ln), 8);
+ len += sizeof(*id);
+
+ /* Submit FDMI request */
+ if (csio_ln_mgmt_submit_req(fdmi_req, csio_ln_fdmi_dhba_cbfn,
+ FCOE_CT, &fdmi_req->dma_buf,
+ len)) {
+ csio_inc_stats(ln, n_fdmi_err);
+ csio_ln_err(ln, "Failed to issue fdmi dhba req\n");
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_vnp_read_cbfn - vnp read completion handler.
+ * @hw: HW lnode
+ * @cbfn: Completion handler.
+ *
+ * Reads vnp response and updates ln parameters.
+ */
+static void
+csio_ln_vnp_read_cbfn(struct csio_hw *hw, struct csio_mb *mbp)
+{
+ struct csio_lnode *ln = ((struct csio_lnode *)mbp->priv);
+ struct fw_fcoe_vnp_cmd *rsp = (struct fw_fcoe_vnp_cmd *)(mbp->mb);
+ struct csio_cmn_sp *csp;
+ struct csio_class_sp *clsp;
+ enum fw_retval retval;
+
+ spin_lock_irq(&hw->lock);
+
+ retval = FW_CMD_RETVAL_GET(ntohl(rsp->alloc_to_len16));
+ if (retval != FW_SUCCESS) {
+ csio_err(hw, "FCOE VNP read cmd returned error:0x%x\n", retval);
+ spin_unlock_irq(&hw->lock);
+ mempool_free(mbp, hw->mb_mempool);
+ return;
+ }
+
+ memcpy(ln->mac, rsp->vnport_mac, sizeof(ln->mac));
+ memcpy(&ln->nport_id, &rsp->vnport_mac[3],
+ sizeof(uint8_t)*3);
+ ln->nport_id = ntohl(ln->nport_id);
+ ln->nport_id = ln->nport_id>>8;
+
+ /* Update WWNs */
+ /*
+ * This may look like a duplication of what csio_fcoe_enable_link()
+ * does, but is absolutely necessary if the vnpi changes between
+ * a FCOE LINK UP and FCOE LINK DOWN.
+ */
+ memcpy(csio_ln_wwnn(ln), rsp->vnport_wwnn, 8);
+ memcpy(csio_ln_wwpn(ln), rsp->vnport_wwpn, 8);
+
+ /* Copy common sparam */
+ csp = (struct csio_cmn_sp *) rsp->cmn_srv_parms;
+ ln->ln_sparm.csp.hi_ver = csp->hi_ver;
+ ln->ln_sparm.csp.lo_ver = csp->lo_ver;
+ ln->ln_sparm.csp.bb_credit = ntohs(csp->bb_credit);
+ ln->ln_sparm.csp.word1_flags = ntohs(csp->word1_flags);
+ ln->ln_sparm.csp.rcv_sz = ntohs(csp->rcv_sz);
+ ln->ln_sparm.csp.un1.r_a_tov = ntohl(csp->un1.r_a_tov);
+ ln->ln_sparm.csp.e_d_tov = ntohl(csp->e_d_tov);
+
+ /* Copy word 0 & word 1 of class sparam */
+ clsp = (struct csio_class_sp *) rsp->clsp_word_0_1;
+ ln->ln_sparm.clsp[2].serv_option = clsp->serv_option;
+ ln->ln_sparm.clsp[2].init_ctl_option = clsp->init_ctl_option;
+ ln->ln_sparm.clsp[2].rcv_ctl_option = clsp->rcv_ctl_option;
+ ln->ln_sparm.clsp[2].rcv_data_sz = ntohs(clsp->rcv_data_sz);
+
+ spin_unlock_irq(&hw->lock);
+
+ mempool_free(mbp, hw->mb_mempool);
+
+ /* Send an event to update local attribs */
+ csio_lnode_async_event(ln, CSIO_LN_FC_ATTRIB_UPDATE);
+}
+
+/*
+ * csio_ln_vnp_read - Read vnp params.
+ * @ln: lnode
+ * @cbfn: Completion handler.
+ *
+ * Issued with lock held.
+ */
+static csio_retval_t
+csio_ln_vnp_read(struct csio_lnode *ln,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct csio_hw *hw = ln->hwp;
+ struct csio_mb *mbp;
+
+ /* Allocate Mbox request */
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /* Prepare VNP Command */
+ csio_fcoe_vnp_read_init_mb(ln, mbp,
+ CSIO_MB_DEFAULT_TMO,
+ ln->fcf_flowid,
+ ln->vnp_flowid,
+ cbfn);
+
+ /* Issue MBOX cmd */
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Failed to issue mbox FCoE VNP command\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_fcoe_enable_link - Enable fcoe link.
+ * @ln: lnode
+ * @enable: enable/disable
+ * Issued with lock held.
+ * Issues mbox cmd to bring up FCOE link on port associated with given ln.
+ */
+static csio_retval_t
+csio_fcoe_enable_link(struct csio_lnode *ln, bool enable)
+{
+ struct csio_hw *hw = ln->hwp;
+ struct csio_mb *mbp;
+ enum fw_retval retval;
+ uint8_t portid;
+ uint8_t sub_op;
+ struct fw_fcoe_link_cmd *lcmd;
+ int i;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ portid = ln->portid;
+ sub_op = enable ? FCOE_LINK_UP : FCOE_LINK_DOWN;
+
+ csio_dbg(hw, "bringing FCOE LINK %s on Port:%d\n",
+ sub_op ? "UP" : "DOWN", portid);
+
+ csio_write_fcoe_link_cond_init_mb(ln, mbp, CSIO_MB_DEFAULT_TMO,
+ portid, sub_op, 0, 0, 0, NULL);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "failed to issue FCOE LINK cmd on port[%d]\n",
+ portid);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ retval = csio_mb_fw_retval(mbp);
+ if (retval != FW_SUCCESS) {
+ csio_err(hw,
+ "FCOE LINK %s cmd on port[%d] failed with "
+ "ret:x%x\n", sub_op ? "UP" : "DOWN", portid, retval);
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ if (!enable)
+ goto out;
+
+ lcmd = (struct fw_fcoe_link_cmd *)mbp->mb;
+
+ memcpy(csio_ln_wwnn(ln), lcmd->vnport_wwnn, 8);
+ memcpy(csio_ln_wwpn(ln), lcmd->vnport_wwpn, 8);
+
+ for (i = 0; i < CSIO_MAX_PPORTS; i++)
+ if (hw->pport[i].portid == portid)
+ memcpy(hw->pport[i].mac, lcmd->phy_mac, 6);
+
+out:
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_read_fcf_cbfn - Read fcf parameters
+ * @ln: lnode
+ *
+ * read fcf response and Update ln fcf information.
+ */
+static void
+csio_ln_read_fcf_cbfn(struct csio_hw *hw, struct csio_mb *mbp)
+{
+ struct csio_lnode *ln = (struct csio_lnode *)mbp->priv;
+ struct csio_fcf_info *fcf_info;
+ struct fw_fcoe_fcf_cmd *rsp =
+ (struct fw_fcoe_fcf_cmd *)(mbp->mb);
+ enum fw_retval retval;
+
+ spin_lock_irq(&hw->lock);
+
+ retval = FW_CMD_RETVAL_GET(ntohl(rsp->retval_len16));
+ if (retval != FW_SUCCESS) {
+ csio_ln_err(ln, "FCOE FCF cmd failed with ret x%x\n",
+ retval);
+ mempool_free(mbp, hw->mb_mempool);
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ fcf_info = ln->fcfinfo;
+ fcf_info->priority = FW_FCOE_FCF_CMD_PRIORITY_GET(
+ ntohs(rsp->priority_pkd));
+ fcf_info->vf_id = ntohs(rsp->vf_id);
+ fcf_info->vlan_id = rsp->vlan_id;
+ fcf_info->max_fcoe_size = ntohs(rsp->max_fcoe_size);
+ fcf_info->fka_adv = be32_to_cpu(rsp->fka_adv);
+ fcf_info->fcfi = FW_FCOE_FCF_CMD_FCFI_GET(ntohl(rsp->op_to_fcfi));
+ fcf_info->fpma = FW_FCOE_FCF_CMD_FPMA_GET(rsp->fpma_to_portid);
+ fcf_info->spma = FW_FCOE_FCF_CMD_SPMA_GET(rsp->fpma_to_portid);
+ fcf_info->login = FW_FCOE_FCF_CMD_LOGIN_GET(rsp->fpma_to_portid);
+ fcf_info->portid = FW_FCOE_FCF_CMD_PORTID_GET(rsp->fpma_to_portid);
+ memcpy(fcf_info->fc_map, rsp->fc_map, sizeof(fcf_info->fc_map));
+ memcpy(fcf_info->mac, rsp->mac, sizeof(fcf_info->mac));
+ memcpy(fcf_info->name_id, rsp->name_id, sizeof(fcf_info->name_id));
+ memcpy(fcf_info->fabric, rsp->fabric, sizeof(fcf_info->fabric));
+ memcpy(fcf_info->spma_mac, rsp->spma_mac, sizeof(fcf_info->spma_mac));
+
+ spin_unlock_irq(&hw->lock);
+ mempool_free(mbp, hw->mb_mempool);
+
+}
+
+/*
+ * csio_ln_read_fcf_entry - Read fcf entry.
+ * @ln: lnode
+ * @cbfn: Completion handler.
+ *
+ * Issued with lock held.
+ */
+static csio_retval_t
+csio_ln_read_fcf_entry(struct csio_lnode *ln,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *))
+{
+ struct csio_hw *hw = ln->hwp;
+ struct csio_mb *mbp;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /* Get FCoE FCF information */
+ csio_fcoe_read_fcf_init_mb(ln, mbp, CSIO_MB_DEFAULT_TMO,
+ ln->portid, ln->fcf_flowid, cbfn);
+
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "failed to issue FCOE FCF cmd\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+
+ if (cbfn == NULL) {
+ spin_unlock_irq(&hw->lock);
+ csio_ln_read_fcf_cbfn(hw, mbp);
+ spin_lock_irq(&hw->lock);
+ }
+
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_handle_link_up - Logical Linkup event.
+ * @hw - HW module.
+ * @portid - Physical port number
+ * @fcfi - FCF index.
+ * @vnpi - VNP index.
+ * Returns - none.
+ *
+ * This event is received from FW, when virtual link is established between
+ * Physical port[ENode] and FCF. If its new vnpi, then local node object is
+ * created on this FCF and set to [ONLINE] state.
+ * Lnode waits for FW_RDEV_CMD event to be received indicating that
+ * Fabric login is completed and lnode moves to [READY] state.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_handle_link_up(struct csio_hw *hw, uint8_t portid, uint32_t fcfi,
+ uint32_t vnpi)
+{
+ struct csio_lnode *ln = NULL;
+
+ /* Lookup lnode based on vnpi */
+ ln = csio_ln_lookup_by_vnpi(hw, vnpi);
+ if (!ln) {
+ /* Pick lnode based on portid */
+ ln = csio_ln_lookup_by_portid(hw, portid);
+ if (!ln) {
+ csio_err(hw, "failed to lookup fcoe lnode on port:%d\n",
+ portid);
+ CSIO_DB_ASSERT(0);
+ return;
+ }
+
+ /* Check if lnode has valid vnp flowid */
+ if (ln->vnp_flowid != CSIO_INVALID_IDX) {
+ /* New VN-Port */
+ spin_unlock_irq(&hw->lock);
+ csio_lnode_alloc(hw);
+ spin_lock_irq(&hw->lock);
+ if (!ln) {
+ csio_err(hw,
+ "failed to allocate fcoe lnode"
+ "for port:%d vnpi:x%x\n",
+ portid, vnpi);
+ CSIO_DB_ASSERT(0);
+ return;
+ }
+ ln->portid = portid;
+ }
+ ln->vnp_flowid = vnpi;
+ ln->dev_num &= ~0xFFFF;
+ ln->dev_num |= vnpi;
+ }
+
+ /*Initialize fcfi */
+ ln->fcf_flowid = fcfi;
+
+ csio_info(hw, "Port:%d - FCOE LINK UP\n", portid);
+
+ csio_inc_stats(ln, n_link_up);
+
+ /* Send LINKUP event to SM */
+ csio_post_event(&ln->sm, CSIO_LNE_LINKUP);
+}
+
+/*
+ * csio_post_event_rns
+ * @ln - FCOE lnode
+ * @evt - Given rnode event
+ * Returns - none
+ *
+ * Posts given rnode event to all FCOE rnodes connected with given Lnode.
+ * This routine is invoked when lnode receives LINK_DOWN/DOWN_LINK/CLOSE
+ * event.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_post_event_rns(struct csio_lnode *ln, enum csio_rn_ev evt)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp, *next;
+ struct csio_rnode *rn;
+
+ list_for_each_safe(tmp, next, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ csio_post_event(&rn->sm, evt);
+ }
+}
+
+/*
+ * csio_cleanup_rns
+ * @ln - FCOE lnode
+ * Returns - none
+ *
+ * Frees all FCOE rnodes connected with given Lnode.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_cleanup_rns(struct csio_lnode *ln)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp, *next_rn;
+ struct csio_rnode *rn;
+
+ list_for_each_safe(tmp, next_rn, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ csio_put_rnode(ln, rn);
+ }
+
+}
+
+/*
+ * csio_post_event_lns
+ * @ln - FCOE lnode
+ * @evt - Given lnode event
+ * Returns - none
+ *
+ * Posts given lnode event to all FCOE lnodes connected with given Lnode.
+ * This routine is invoked when lnode receives LINK_DOWN/DOWN_LINK/CLOSE
+ * event.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_post_event_lns(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct list_head *tmp;
+ struct csio_lnode *cln, *sln;
+
+ /* If NPIV lnode, send evt only to that and return */
+ if (csio_is_npiv_ln(ln)) {
+ csio_post_event(&ln->sm, evt);
+ return;
+ }
+
+ sln = ln;
+ /* Traverse children lnodes list and send evt */
+ list_for_each(tmp, &sln->cln_head) {
+ cln = (struct csio_lnode *) tmp;
+ csio_post_event(&cln->sm, evt);
+ }
+
+ /* Send evt to parent lnode */
+ csio_post_event(&ln->sm, evt);
+}
+
+/*
+ * csio_ln_down - Lcoal nport is down
+ * @ln - FCOE Lnode
+ * Returns - none
+ *
+ * Sends LINK_DOWN events to Lnode and its associated NPIVs lnodes.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_ln_down(struct csio_lnode *ln)
+{
+ csio_post_event_lns(ln, CSIO_LNE_LINK_DOWN);
+}
+
+/*
+ * csio_handle_link_down - Logical Linkdown event.
+ * @hw - HW module.
+ * @portid - Physical port number
+ * @fcfi - FCF index.
+ * @vnpi - VNP index.
+ * Returns - none
+ *
+ * This event is received from FW, when virtual link goes down between
+ * Physical port[ENode] and FCF. Lnode and its associated NPIVs lnode hosted on
+ * this vnpi[VN-Port] will be de-instantiated.
+ *
+ * This called with hw lock held
+ */
+static void
+csio_handle_link_down(struct csio_hw *hw, uint8_t portid, uint32_t fcfi,
+ uint32_t vnpi)
+{
+ struct csio_fcf_info *fp;
+ struct csio_lnode *ln;
+
+ /* Lookup lnode based on vnpi */
+ ln = csio_ln_lookup_by_vnpi(hw, vnpi);
+ if (ln) {
+ fp = ln->fcfinfo;
+ csio_inc_stats(ln, n_link_down);
+
+ /*Warn if linkdown received if lnode is not in ready state */
+ if (!csio_is_lnode_ready(ln)) {
+ csio_ln_warn(ln,
+ "warn: FCOE link is already in offline "
+ "Ignoring Fcoe linkdown event on portid %d\n",
+ portid);
+ csio_inc_stats(ln, n_evt_drop);
+ return;
+ }
+
+ /* Verify portid */
+ if (fp->portid != portid) {
+ csio_ln_warn(ln,
+ "warn: FCOE linkdown recv with "
+ "invalid port %d\n", portid);
+ csio_inc_stats(ln, n_evt_drop);
+ return;
+ }
+
+ /* verify fcfi */
+ if (ln->fcf_flowid != fcfi) {
+ csio_ln_warn(ln,
+ "warn: FCOE linkdown recv with "
+ "invalid fcfi x%x\n", fcfi);
+ csio_inc_stats(ln, n_evt_drop);
+ return;
+ }
+
+ csio_info(hw, "Port:%d - FCOE LINK DOWN\n", portid);
+
+ /* Send LINK_DOWN event to lnode s/m */
+ csio_ln_down(ln);
+
+ return;
+ } else {
+ csio_warn(hw,
+ "warn: FCOE linkdown recv with invalid vnpi x%x\n",
+ vnpi);
+ csio_inc_stats(hw, n_evt_drop);
+ }
+}
+
+/*
+ * csio_is_lnode_ready - Checks FCOE lnode is in ready state.
+ * @ln: Lnode module
+ *
+ * Returns True if FCOE lnode is in ready state.
+ */
+int
+csio_is_lnode_ready(struct csio_lnode *ln)
+{
+ return (csio_get_state(ln) == ((csio_sm_state_t)csio_lns_ready));
+}
+
+/*****************************************************************************/
+/* START: Lnode SM */
+/*****************************************************************************/
+/*
+ * csio_lns_uninit - The request in uninit state.
+ * @ln - FCOE lnode.
+ * @evt - Event to be processed.
+ *
+ * Process the given lnode event which is currently in "uninit" state.
+ * Invoked with HW lock held.
+ * Return - none.
+ */
+static void
+csio_lns_uninit(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_lnode *rln = hw->rln;
+ csio_retval_t rv;
+
+ csio_inc_stats(ln, n_evt_sm[evt]);
+ switch (evt) {
+
+ case CSIO_LNE_LINKUP:
+ csio_set_state(&ln->sm, csio_lns_online);
+ /* Read FCF only for physical lnode */
+ if (csio_is_phys_ln(ln)) {
+ rv = csio_ln_read_fcf_entry(ln,
+ csio_ln_read_fcf_cbfn);
+ if (rv != CSIO_SUCCESS) {
+ /* TODO: Send HW RESET event */
+ csio_inc_stats(ln, n_err);
+ break;
+ }
+
+ /* Add FCF record */
+ list_add_tail(&ln->fcfinfo->list, &rln->fcf_lsthead);
+ }
+
+ rv = csio_ln_vnp_read(ln, csio_ln_vnp_read_cbfn);
+ if (rv != CSIO_SUCCESS) {
+ /* TODO: Send HW RESET event */
+ csio_inc_stats(ln, n_err);
+ }
+ break;
+
+ case CSIO_LNE_DOWN_LINK:
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "unexp ln event %d recv from did:x%x in "
+ "ln state[uninit].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_unexp);
+ break;
+ } /* switch event */
+}
+
+/*
+ * csio_lns_online - The request in online state.
+ * @ln - FCOE lnode.
+ * @evt - Event to be processed.
+ *
+ * Process the given lnode event which is currently in "online" state.
+ * Invoked with HW lock held.
+ * Return - none.
+ */
+static void
+csio_lns_online(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ csio_inc_stats(ln, n_evt_sm[evt]);
+ switch (evt) {
+
+ case CSIO_LNE_LINKUP:
+ csio_ln_warn(ln,
+ "warn: FCOE link is up already "
+ "Ignoring linkup on port:%d\n", ln->portid);
+ csio_inc_stats(ln, n_evt_drop);
+ break;
+
+ case CSIO_LNE_FAB_INIT_DONE:
+ csio_set_state(&ln->sm, csio_lns_ready);
+
+ spin_unlock_irq(&hw->lock);
+ csio_lnode_async_event(ln, CSIO_LN_FC_LINKUP);
+ spin_lock_irq(&hw->lock);
+
+ break;
+
+ case CSIO_LNE_LINK_DOWN:
+ /* Fall through */
+ case CSIO_LNE_DOWN_LINK:
+ csio_set_state(&ln->sm, csio_lns_uninit);
+ if (csio_is_phys_ln(ln)) {
+ /* Remove FCF entry */
+ list_del_init(&ln->fcfinfo->list);
+ }
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "unexp ln event %d recv from did:x%x in "
+ "ln state[uninit].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_unexp);
+
+ break;
+ } /* switch event */
+}
+
+/*
+ * csio_lns_ready - The request in ready state.
+ * @ln - FCOE lnode.
+ * @evt - Event to be processed.
+ *
+ * Process the given lnode event which is currently in "ready" state.
+ * Invoked with HW lock held.
+ * Return - none.
+ */
+static void
+csio_lns_ready(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ csio_inc_stats(ln, n_evt_sm[evt]);
+ switch (evt) {
+
+ case CSIO_LNE_FAB_INIT_DONE:
+ csio_ln_err(ln,
+ "ignoring event %d recv from did x%x"
+ "in ln state[ready].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_drop);
+ break;
+
+ case CSIO_LNE_LINK_DOWN:
+ csio_set_state(&ln->sm, csio_lns_offline);
+ csio_post_event_rns(ln, CSIO_RNFE_DOWN);
+
+ spin_unlock_irq(&hw->lock);
+ csio_lnode_async_event(ln, CSIO_LN_FC_LINKDOWN);
+ spin_lock_irq(&hw->lock);
+
+ if (csio_is_phys_ln(ln)) {
+ /* Remove FCF entry */
+ list_del_init(&ln->fcfinfo->list);
+ }
+ break;
+
+ case CSIO_LNE_DOWN_LINK:
+ csio_set_state(&ln->sm, csio_lns_offline);
+ csio_post_event_rns(ln, CSIO_RNFE_DOWN);
+
+ /* Host need to issue aborts in case if FW has not returned
+ * WRs with status "ABORTED"
+ */
+ spin_unlock_irq(&hw->lock);
+ csio_lnode_async_event(ln, CSIO_LN_FC_LINKDOWN);
+ spin_lock_irq(&hw->lock);
+
+ if (csio_is_phys_ln(ln)) {
+ /* Remove FCF entry */
+ list_del_init(&ln->fcfinfo->list);
+ }
+ break;
+
+ case CSIO_LNE_CLOSE:
+ csio_set_state(&ln->sm, csio_lns_uninit);
+ csio_post_event_rns(ln, CSIO_RNFE_CLOSE);
+ break;
+
+ case CSIO_LNE_LOGO:
+ csio_set_state(&ln->sm, csio_lns_offline);
+ csio_post_event_rns(ln, CSIO_RNFE_DOWN);
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "unexp ln event %d recv from did:x%x in "
+ "ln state[uninit].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_unexp);
+ CSIO_DB_ASSERT(0);
+ break;
+ } /* switch event */
+}
+
+/*
+ * csio_lns_offline - The request in offline state.
+ * @ln - FCOE lnode.
+ * @evt - Event to be processed.
+ *
+ * Process the given lnode event which is currently in "offline" state.
+ * Invoked with HW lock held.
+ * Return - none.
+ */
+static void
+csio_lns_offline(struct csio_lnode *ln, enum csio_ln_ev evt)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_lnode *rln = hw->rln;
+ csio_retval_t rv;
+
+ csio_inc_stats(ln, n_evt_sm[evt]);
+ switch (evt) {
+
+ case CSIO_LNE_LINKUP:
+ csio_set_state(&ln->sm, csio_lns_online);
+ /* Read FCF only for physical lnode */
+ if (csio_is_phys_ln(ln)) {
+ rv = csio_ln_read_fcf_entry(ln,
+ csio_ln_read_fcf_cbfn);
+ if (rv != CSIO_SUCCESS) {
+ /* TODO: Send HW RESET event */
+ csio_inc_stats(ln, n_err);
+ break;
+ }
+
+ /* Add FCF record */
+ list_add_tail(&ln->fcfinfo->list, &rln->fcf_lsthead);
+ }
+
+ rv = csio_ln_vnp_read(ln, csio_ln_vnp_read_cbfn);
+ if (rv != CSIO_SUCCESS) {
+ /* TODO: Send HW RESET event */
+ csio_inc_stats(ln, n_err);
+ }
+ break;
+
+ case CSIO_LNE_LINK_DOWN:
+ case CSIO_LNE_DOWN_LINK:
+ case CSIO_LNE_LOGO:
+ csio_ln_err(ln,
+ "ignoring event %d recv from did x%x"
+ "in ln state[offline].\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_drop);
+ break;
+
+ case CSIO_LNE_CLOSE:
+ csio_set_state(&ln->sm, csio_lns_uninit);
+ csio_post_event_rns(ln, CSIO_RNFE_CLOSE);
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "unexp ln event %d recv from did:x%x in "
+ "ln state[offline]\n", evt, ln->nport_id);
+ csio_inc_stats(ln, n_evt_unexp);
+ CSIO_DB_ASSERT(0);
+ break;
+ } /* switch event */
+}
+
+/*****************************************************************************/
+/* END: Lnode SM */
+/*****************************************************************************/
+
+static void
+csio_free_fcfinfo(struct kref *kref)
+{
+ struct csio_fcf_info *fcfinfo = container_of(kref,
+ struct csio_fcf_info, kref);
+ kfree(fcfinfo);
+}
+
+/* Helper routines for attributes */
+/*
+ * csio_lnode_state_to_str - Get current state of FCOE lnode.
+ * @ln - lnode
+ * @str - state of lnode.
+ *
+ */
+void
+csio_lnode_state_to_str(struct csio_lnode *ln, int8_t *str)
+{
+ if (csio_get_state(ln) == ((csio_sm_state_t)csio_lns_uninit)) {
+ strcpy(str, "UNINIT");
+ return;
+ }
+ if (csio_get_state(ln) == ((csio_sm_state_t)csio_lns_ready)) {
+ strcpy(str, "READY");
+ return;
+ }
+ if (csio_get_state(ln) == ((csio_sm_state_t)csio_lns_offline)) {
+ strcpy(str, "OFFLINE");
+ return;
+ }
+ strcpy(str, "UNKNOWN");
+} /* csio_lnode_state_to_str */
+
+
+csio_retval_t
+csio_get_phy_port_stats(struct csio_hw *hw, uint8_t portid,
+ struct fw_fcoe_port_stats *port_stats)
+{
+ struct csio_mb *mbp;
+ struct fw_fcoe_port_cmd_params portparams;
+ enum fw_retval retval;
+ int idx;
+
+ mbp = mempool_alloc(hw->mb_mempool, GFP_ATOMIC);
+ if (!mbp) {
+ csio_err(hw, "FCoE FCF PARAMS command out of memory!\n");
+ return CSIO_INVAL;
+ }
+ portparams.portid = portid;
+
+ for (idx = 1; idx <= 3; idx++) {
+ portparams.idx = (idx-1)*6 + 1;
+ portparams.nstats = 6;
+ if (idx == 3)
+ portparams.nstats = 4;
+ csio_fcoe_read_portparams_init_mb(hw, mbp, CSIO_MB_DEFAULT_TMO,
+ &portparams, NULL);
+ if (csio_mb_issue(hw, mbp)) {
+ csio_err(hw, "Issue of FCoE port params failed!\n");
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_INVAL;
+ }
+ csio_mb_process_portparams_rsp(hw, mbp, &retval,
+ &portparams, port_stats);
+ }
+
+ mempool_free(mbp, hw->mb_mempool);
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_mgmt_wr_handler -Mgmt Work Request handler.
+ * @wr - WR.
+ * @len - WR len.
+ * This handler is invoked when an outstanding mgmt WR is completed.
+ * Its invoked in the context of FW event worker thread for every
+ * mgmt event received.
+ * Return - none.
+ */
+
+static void
+csio_ln_mgmt_wr_handler(struct csio_hw *hw, void *wr, uint32_t len)
+{
+ struct csio_mgmtm *mgmtm = csio_hw_to_mgmtm(hw);
+ struct csio_ioreq *io_req = NULL;
+ struct fw_fcoe_els_ct_wr *wr_cmd;
+
+
+ wr_cmd = (struct fw_fcoe_els_ct_wr *) wr;
+
+ if (len < sizeof(struct fw_fcoe_els_ct_wr)) {
+ csio_err(mgmtm->hw,
+ "Invalid ELS CT WR length recvd, len:%x\n", len);
+ mgmtm->stats.n_err++;
+ return;
+ }
+
+ io_req = (struct csio_ioreq *) ((uintptr_t) wr_cmd->cookie);
+ io_req->wr_status = csio_wr_status(wr_cmd);
+
+ /* lookup ioreq exists in our active Q */
+ spin_lock_irq(&hw->lock);
+ if (csio_mgmt_req_lookup(mgmtm, io_req) != CSIO_SUCCESS) {
+ csio_err(mgmtm->hw,
+ "Error- Invalid IO handle recv in WR. handle: %p\n",
+ io_req);
+ mgmtm->stats.n_err++;
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ mgmtm = csio_hw_to_mgmtm(hw);
+
+ /* Dequeue from active queue */
+ list_del_init(&io_req->sm.sm_list);
+ mgmtm->stats.n_active--;
+ spin_unlock_irq(&hw->lock);
+
+ if (io_req->io_cbfn) {
+ /* io_req will be freed by completion handler */
+ io_req->io_cbfn(hw, io_req);
+ }
+}
+
+/**
+ * csio_fcoe_fwevt_handler - Event handler for Firmware FCoE events.
+ * @hw: HW module
+ * @cpl_op: CPL opcode
+ * @cmd: FW cmd/WR.
+ *
+ * Process received FCoE cmd/WR event from FW.
+ */
+void
+csio_fcoe_fwevt_handler(struct csio_hw *hw, __u8 cpl_op, __be64 *cmd)
+{
+ struct csio_lnode *ln;
+ struct csio_rnode *rn;
+ uint8_t portid, opcode = *(uint8_t *)cmd;
+ struct fw_fcoe_link_cmd *lcmd;
+ struct fw_wr_hdr *wr;
+ struct fw_rdev_wr *rdev_wr;
+ enum fw_fcoe_link_status lstatus;
+ uint32_t fcfi, rdev_flowid, vnpi;
+ enum csio_ln_ev evt;
+
+ if (cpl_op == CPL_FW6_MSG && opcode == FW_FCOE_LINK_CMD) {
+
+ lcmd = (struct fw_fcoe_link_cmd *)cmd;
+ lstatus = lcmd->lstatus;
+ portid = FW_FCOE_LINK_CMD_PORTID_GET(
+ ntohl(lcmd->op_to_portid));
+ fcfi = FW_FCOE_LINK_CMD_FCFI_GET(ntohl(lcmd->sub_opcode_fcfi));
+ vnpi = FW_FCOE_LINK_CMD_VNPI_GET(ntohl(lcmd->vnpi_pkd));
+
+ if (lstatus == FCOE_LINKUP) {
+
+ /* HW lock here */
+ spin_lock_irq(&hw->lock);
+ csio_handle_link_up(hw, portid, fcfi, vnpi);
+ spin_unlock_irq(&hw->lock);
+ /* HW un lock here */
+
+ } else if (lstatus == FCOE_LINKDOWN) {
+
+ /* HW lock here */
+ spin_lock_irq(&hw->lock);
+ csio_handle_link_down(hw, portid, fcfi, vnpi);
+ spin_unlock_irq(&hw->lock);
+ /* HW un lock here */
+ } else {
+ csio_warn(hw, "Unexpected FCOE LINK status:0x%x\n",
+ ntohl(lcmd->lstatus));
+ csio_inc_stats(hw, n_cpl_unexp);
+ }
+ } else if (cpl_op == CPL_FW6_PLD) {
+ wr = (struct fw_wr_hdr *) (cmd + 4);
+ if (FW_WR_OP_GET(be32_to_cpu(wr->hi))
+ == FW_RDEV_WR) {
+
+ rdev_wr = (struct fw_rdev_wr *) (cmd + 4);
+
+ rdev_flowid = FW_RDEV_WR_FLOWID_GET(
+ ntohl(rdev_wr->alloc_to_len16));
+ vnpi = FW_RDEV_WR_ASSOC_FLOWID_GET(
+ ntohl(rdev_wr->flags_to_assoc_flowid));
+
+ csio_dbg(hw,
+ "FW_RDEV_WR: flowid:x%x ev_cause:x%x "
+ "vnpi:0x%x\n", rdev_flowid,
+ rdev_wr->event_cause, vnpi);
+
+ if (rdev_wr->protocol != PROT_FCOE) {
+ csio_err(hw,
+ "FW_RDEV_WR: invalid proto:x%x "
+ "received with flowid:x%x\n",
+ rdev_wr->protocol,
+ rdev_flowid);
+ csio_inc_stats(hw, n_evt_drop);
+ return;
+ }
+
+ /* HW lock here */
+ spin_lock_irq(&hw->lock);
+ ln = csio_ln_lookup_by_vnpi(hw, vnpi);
+ if (!ln) {
+ csio_err(hw,
+ "FW_DEV_WR: invalid vnpi:x%x received "
+ "with flowid:x%x\n", vnpi, rdev_flowid);
+ csio_inc_stats(hw, n_evt_drop);
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ rn = csio_confirm_rnode(ln, rdev_flowid,
+ &rdev_wr->u.fcoe_rdev);
+ if (!rn) {
+ csio_ln_dbg(ln,
+ "Failed to confirm rnode "
+ "for flowid:x%x\n", rdev_flowid);
+ csio_inc_stats(hw, n_evt_drop);
+ spin_unlock_irq(&hw->lock);
+ return;
+ }
+
+ /* save previous event for debugging */
+ ln->prev_evt = ln->cur_evt;
+ ln->cur_evt = rdev_wr->event_cause;
+ csio_inc_stats(ln, n_evt_fw[rdev_wr->event_cause]);
+
+ /* Translate all the fabric events to lnode SM events */
+ evt = CSIO_FWE_TO_LNE(rdev_wr->event_cause);
+ if (evt) {
+ csio_ln_dbg(ln,
+ "Posting event to lnode event:%d "
+ "cause:%d flowid:x%x\n", evt,
+ rdev_wr->event_cause, rdev_flowid);
+ csio_post_event(&ln->sm, evt);
+ }
+
+ /* Handover event to rn SM here. */
+ csio_rnode_fwevt_handler(rn, rdev_wr->event_cause);
+
+ spin_unlock_irq(&hw->lock);
+ } else {
+ csio_warn(hw, "unexpected WR op(0x%x) recv\n",
+ FW_WR_OP_GET(be32_to_cpu((wr->hi))));
+ csio_inc_stats(hw, n_cpl_unexp);
+ }
+ } else if (cpl_op == CPL_FW6_MSG) {
+ wr = (struct fw_wr_hdr *) (cmd);
+ if (FW_WR_OP_GET(be32_to_cpu(wr->hi))
+ == FW_FCOE_ELS_CT_WR) {
+ csio_ln_mgmt_wr_handler(hw, wr,
+ sizeof(struct fw_fcoe_els_ct_wr));
+ } else {
+ csio_warn(hw, "unexpected WR op(0x%x) recv\n",
+ FW_WR_OP_GET(be32_to_cpu((wr->hi))));
+ csio_inc_stats(hw, n_cpl_unexp);
+ }
+ } else {
+ csio_warn(hw, "unexpected CPL op(0x%x) recv\n", opcode);
+ csio_inc_stats(hw, n_cpl_unexp);
+ }
+}
+
+/**
+ * csio_lnode_start - Kickstart lnode discovery.
+ * @ln: lnode
+ *
+ * This routine kickstarts the discovery by issuing an FCOE_LINK (up) command.
+ */
+csio_retval_t
+csio_lnode_start(struct csio_lnode *ln)
+{
+ csio_retval_t rv = CSIO_SUCCESS;
+ if (csio_is_phys_ln(ln) && !(ln->flags & CSIO_LNF_LINK_ENABLE)) {
+ rv = csio_fcoe_enable_link(ln, 1);
+ ln->flags |= CSIO_LNF_LINK_ENABLE;
+ }
+
+ return rv;
+}
+
+/**
+ * csio_lnode_stop - Stop the lnode.
+ * @ln: lnode
+ *
+ * This routine is invoked by HW module to stop lnode and its associated NPIV
+ * lnodes.
+ */
+void
+csio_lnode_stop(struct csio_lnode *ln)
+{
+ csio_post_event_lns(ln, CSIO_LNE_DOWN_LINK);
+ if (csio_is_phys_ln(ln) && (ln->flags & CSIO_LNF_LINK_ENABLE)) {
+ csio_fcoe_enable_link(ln, 0);
+ ln->flags &= ~CSIO_LNF_LINK_ENABLE;
+ }
+ csio_ln_dbg(ln, "stopping ln :%p\n", ln);
+}
+
+/**
+ * csio_lnode_close - Close an lnode.
+ * @ln: lnode
+ *
+ * This routine is invoked by HW module to close an lnode and its
+ * associated NPIV lnodes. Lnode and its associated NPIV lnodes are
+ * set to uninitialized state.
+ */
+void
+csio_lnode_close(struct csio_lnode *ln)
+{
+ csio_post_event_lns(ln, CSIO_LNE_CLOSE);
+ if (csio_is_phys_ln(ln))
+ ln->vnp_flowid = CSIO_INVALID_IDX;
+
+ csio_ln_dbg(ln, "closed ln :%p\n", ln);
+}
+
+/*
+ * csio_ln_prep_ecwr - Prepare ELS/CT WR.
+ * @io_req - IO request.
+ * @wr_len - WR len
+ * @immd_len - WR immediate data
+ * @sub_op - Sub opcode
+ * @sid - source portid.
+ * @did - destination portid
+ * @flow_id - flowid
+ * @fw_wr - ELS/CT WR to be prepared.
+ * Returns: CSIO_SUCCESS - on success
+ */
+static csio_retval_t
+csio_ln_prep_ecwr(struct csio_ioreq *io_req, uint32_t wr_len,
+ uint32_t immd_len, uint8_t sub_op, uint32_t sid,
+ uint32_t did, uint32_t flow_id, uint8_t *fw_wr)
+{
+ struct fw_fcoe_els_ct_wr *wr;
+ uint32_t port_id;
+
+ wr = (struct fw_fcoe_els_ct_wr *)fw_wr;
+ wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_FCOE_ELS_CT_WR) |
+ FW_FCOE_ELS_CT_WR_IMMDLEN(immd_len));
+
+ wr_len = CSIO_ROUNDUP(wr_len, 16);
+ wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(flow_id) |
+ FW_WR_LEN16(wr_len));
+ wr->els_ct_type = sub_op;
+ wr->ctl_pri = 0;
+ wr->cp_en_class = 0;
+ wr->cookie = io_req->fw_handle;
+ wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(
+ io_req->lnode->hwp, io_req->iq_idx));
+ wr->fl_to_sp = FW_FCOE_ELS_CT_WR_SP(1);
+ wr->tmo_val = (uint8_t) io_req->tmo;
+ port_id = htonl(sid);
+ memcpy(wr->l_id, PORT_ID_PTR(port_id), 3);
+ port_id = htonl(did);
+ memcpy(wr->r_id, PORT_ID_PTR(port_id), 3);
+
+ /* Prepare RSP SGL */
+ wr->rsp_dmalen = cpu_to_be32(io_req->dma_buf.len);
+ wr->rsp_dmaaddr = cpu_to_be64(io_req->dma_buf.paddr);
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_mgmt_submit_wr - Post elsct work request.
+ * @mgmtm - mgmtm
+ * @io_req - io request.
+ * @sub_op - ELS or CT request type
+ * @pld - Dma Payload buffer
+ * @pld_len - Payload len
+ * Prepares ELSCT Work request and sents it to FW.
+ * Returns: CSIO_SUCCESS - on success
+ */
+static csio_retval_t
+csio_ln_mgmt_submit_wr(struct csio_mgmtm *mgmtm, struct csio_ioreq *io_req,
+ uint8_t sub_op, struct csio_dma_buf *pld,
+ uint32_t pld_len)
+{
+ struct csio_wr_pair wrp;
+ struct csio_lnode *ln = io_req->lnode;
+ struct csio_rnode *rn = io_req->rnode;
+ struct csio_hw *hw = mgmtm->hw;
+ uint8_t fw_wr[64];
+ struct ulptx_sgl dsgl;
+ uint32_t wr_size = 0;
+ uint8_t im_len = 0;
+ uint32_t wr_off = 0;
+
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ /* Calculate WR Size for this ELS REQ */
+ wr_size = sizeof(struct fw_fcoe_els_ct_wr);
+
+ /* Send as immediate data if pld < 256 */
+ if (pld_len < 256) {
+ wr_size += ALIGN(pld_len, 8);
+ im_len = (uint8_t)pld_len;
+ } else
+ wr_size += sizeof(struct ulptx_sgl);
+
+ /* Roundup WR size in units of 16 bytes */
+ wr_size = ALIGN(wr_size, 16);
+
+ /* Get WR to send ELS REQ */
+ ret = csio_wr_get(hw, mgmtm->eq_idx, wr_size, &wrp);
+ if (ret != CSIO_SUCCESS) {
+ csio_err(hw, "Failed to get WR for ec_req %p ret:%d\n",
+ io_req, ret);
+ return ret;
+ }
+
+ /* Prepare Generic WR used by all ELS/CT cmd */
+ csio_ln_prep_ecwr(io_req, wr_size, im_len, sub_op,
+ ln->nport_id, rn->nport_id,
+ csio_rn_flowid(rn),
+ &fw_wr[0]);
+
+ /* Copy ELS/CT WR CMD */
+ csio_wr_copy_to_wrp(&fw_wr[0], &wrp, wr_off,
+ sizeof(struct fw_fcoe_els_ct_wr));
+ wr_off += sizeof(struct fw_fcoe_els_ct_wr);
+
+ /* Copy payload to Immediate section of WR */
+ if (im_len)
+ csio_wr_copy_to_wrp(pld->vaddr, &wrp, wr_off, im_len);
+ else {
+ /* Program DSGL to dma payload */
+ dsgl.cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) |
+ ULPTX_MORE | ULPTX_NSGE(1));
+ dsgl.len0 = cpu_to_be32(pld_len);
+ dsgl.addr0 = cpu_to_be64(pld->paddr);
+ csio_wr_copy_to_wrp(&dsgl, &wrp, ALIGN(wr_off, 8),
+ sizeof(struct ulptx_sgl));
+ }
+
+ /* Issue work request to xmit ELS/CT req to FW */
+ csio_wr_issue(mgmtm->hw, mgmtm->eq_idx, CSIO_FALSE);
+ return ret;
+}
+
+/*
+ * csio_ln_mgmt_submit_req - Submit FCOE Mgmt request.
+ * @io_req - IO Request
+ * @io_cbfn - Completion handler.
+ * @req_type - ELS or CT request type
+ * @pld - Dma Payload buffer
+ * @pld_len - Payload len
+ *
+ *
+ * This API used submit managment ELS/CT request.
+ * This called with hw lock held
+ * Returns: CSIO_SUCCESS - on success
+ * CSIO_NOMEM - on error.
+ */
+static csio_retval_t
+csio_ln_mgmt_submit_req(struct csio_ioreq *io_req,
+ void (*io_cbfn) (struct csio_hw *, struct csio_ioreq *),
+ enum fcoe_cmn_type req_type, struct csio_dma_buf *pld,
+ uint32_t pld_len)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(io_req->lnode);
+ struct csio_mgmtm *mgmtm = csio_hw_to_mgmtm(hw);
+ csio_retval_t rv;
+
+ io_req->io_cbfn = io_cbfn; /* Upper layer callback handler */
+ io_req->fw_handle = (uintptr_t) (io_req);
+ io_req->eq_idx = mgmtm->eq_idx;
+ io_req->iq_idx = mgmtm->iq_idx;
+
+ rv = csio_ln_mgmt_submit_wr(mgmtm, io_req, req_type, pld, pld_len);
+ if (rv == CSIO_SUCCESS) {
+ list_add_tail(&io_req->sm.sm_list, &mgmtm->active_q);
+ mgmtm->stats.n_active++;
+ }
+ return rv;
+}
+
+/*
+ * csio_ln_fdmi_init - FDMI Init entry point.
+ * @ln: lnode
+ */
+static csio_retval_t
+csio_ln_fdmi_init(struct csio_lnode *ln)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_dma_buf *dma_buf;
+
+ /* Allocate MGMT request required for FDMI */
+ ln->mgmt_req = kzalloc(sizeof(struct csio_ioreq), GFP_KERNEL);
+ if (!ln->mgmt_req) {
+ csio_ln_err(ln, "Failed to alloc ioreq for FDMI\n");
+ csio_inc_stats(hw, n_err_nomem);
+ return CSIO_NOMEM;
+ }
+
+ /* Allocate Dma buffers for FDMI response Payload */
+ dma_buf = &ln->mgmt_req->dma_buf;
+ dma_buf->len = 2048;
+ dma_buf->vaddr = pci_alloc_consistent(hw->pdev, dma_buf->len,
+ &dma_buf->paddr);
+ if (!dma_buf->vaddr) {
+ csio_err(hw, "Failed to alloc DMA buffer for FDMI!\n");
+ kfree(ln->mgmt_req);
+ ln->mgmt_req = NULL;
+ return CSIO_NOMEM;
+ }
+
+ ln->flags |= CSIO_LNF_FDMI_ENABLE;
+ return CSIO_SUCCESS;
+}
+
+/*
+ * csio_ln_fdmi_exit - FDMI exit entry point.
+ * @ln: lnode
+ */
+static csio_retval_t
+csio_ln_fdmi_exit(struct csio_lnode *ln)
+{
+ struct csio_dma_buf *dma_buf;
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ if (!ln->mgmt_req)
+ return CSIO_SUCCESS;
+
+ dma_buf = &ln->mgmt_req->dma_buf;
+ if (dma_buf->vaddr)
+ pci_free_consistent(hw->pdev, dma_buf->len, dma_buf->vaddr,
+ dma_buf->paddr);
+
+ kfree(ln->mgmt_req);
+ return CSIO_SUCCESS;
+}
+
+int
+csio_scan_done(struct csio_lnode *ln, unsigned long ticks,
+ unsigned long time, unsigned long max_scan_ticks,
+ unsigned long delta_scan_ticks)
+{
+ int rv = 0;
+
+ if (time >= max_scan_ticks)
+ return 1;
+
+ if (!ln->tgt_scan_tick)
+ ln->tgt_scan_tick = ticks;
+
+ if (((ticks - ln->tgt_scan_tick) >= delta_scan_ticks)) {
+ if (!ln->last_scan_ntgts)
+ ln->last_scan_ntgts = ln->n_scsi_tgts;
+ else {
+ if (ln->last_scan_ntgts == ln->n_scsi_tgts)
+ return 1;
+
+ ln->last_scan_ntgts = ln->n_scsi_tgts;
+ }
+ ln->tgt_scan_tick = ticks;
+ }
+ return rv;
+}
+
+/*
+ * csio_notify_lnodes:
+ * @hw: HW module
+ * @note: Notification
+ *
+ * Called from the HW SM to fan out notifications to the
+ * Lnode SM. Since the HW SM is entered with lock held,
+ * there is no need to hold locks here.
+ *
+ */
+void
+csio_notify_lnodes(struct csio_hw *hw, enum csio_ln_notify note)
+{
+ struct list_head *tmp;
+ struct csio_lnode *ln;
+
+ csio_dbg(hw, "Notifying all nodes of event %d\n", note);
+
+ /* Traverse children lnodes list and send evt */
+ list_for_each(tmp, &hw->sln_head) {
+ ln = (struct csio_lnode *) tmp;
+
+ switch (note) {
+ case CSIO_LN_NOTIFY_HWREADY:
+ csio_lnode_start(ln);
+ break;
+
+ case CSIO_LN_NOTIFY_HWRESET:
+ case CSIO_LN_NOTIFY_HWREMOVE:
+ csio_lnode_close(ln);
+ break;
+
+ case CSIO_LN_NOTIFY_HWSTOP:
+ csio_lnode_stop(ln);
+ break;
+
+ default:
+ break;
+
+ }
+ }
+}
+
+/*
+ * csio_disable_lnodes:
+ * @hw: HW module
+ * @portid:port id
+ * @disable: disable/enable flag.
+ * If disable=1, disables all lnode hosted on given physical port.
+ * otherwise enables all the lnodes on given phsysical port.
+ * This routine need to called with hw lock held.
+ */
+void
+csio_disable_lnodes(struct csio_hw *hw, uint8_t portid, bool disable)
+{
+ struct list_head *tmp;
+ struct csio_lnode *ln;
+
+ csio_dbg(hw, "Notifying event to all nodes of port:%d\n", portid);
+
+ /* Traverse sibling lnodes list and send evt */
+ list_for_each(tmp, &hw->sln_head) {
+ ln = (struct csio_lnode *) tmp;
+ if (ln->portid != portid)
+ continue;
+
+ if (disable)
+ csio_lnode_stop(ln);
+ else
+ csio_lnode_start(ln);
+ }
+}
+
+/*
+ * csio_ln_init - Initialize an lnode.
+ * @ln: lnode
+ *
+ */
+static csio_retval_t
+csio_ln_init(struct csio_lnode *ln)
+{
+ int rv = CSIO_INVAL;
+ struct csio_lnode *rln, *pln;
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ csio_init_state(&ln->sm, csio_lns_uninit);
+ ln->vnp_flowid = CSIO_INVALID_IDX;
+ ln->fcf_flowid = CSIO_INVALID_IDX;
+
+ if (csio_is_root_ln(ln)) {
+
+ /* This is the lnode used during initialization */
+
+ ln->fcfinfo = kzalloc(sizeof(struct csio_fcf_info), GFP_KERNEL);
+ if (!ln->fcfinfo) {
+ csio_ln_err(ln, "Failed to alloc FCF record\n");
+ csio_inc_stats(hw, n_err_nomem);
+ goto err;
+ }
+
+ INIT_LIST_HEAD(&ln->fcf_lsthead);
+ kref_init(&ln->fcfinfo->kref);
+
+ if (csio_fdmi_enable && csio_ln_fdmi_init(ln))
+ goto err;
+
+ } else { /* Either a non-root physical or a virtual lnode */
+
+ /*
+ * THe rest is common for non-root physical and NPIV lnodes.
+ * Just get references to all other modules
+ */
+ rln = csio_root_lnode(ln);
+
+ if (csio_is_npiv_ln(ln)) {
+ /* NPIV */
+ pln = csio_parent_lnode(ln);
+ kref_get(&pln->fcfinfo->kref);
+ ln->fcfinfo = pln->fcfinfo;
+ } else {
+ /* Another non-root physical lnode (FCF) */
+ ln->fcfinfo = kzalloc(sizeof(struct csio_fcf_info),
+ GFP_KERNEL);
+ if (!ln->fcfinfo) {
+ csio_ln_err(ln,
+ "Failed to alloc FCF info\n");
+ csio_inc_stats(hw, n_err_nomem);
+ goto err;
+ }
+
+ kref_init(&ln->fcfinfo->kref);
+
+ if (csio_fdmi_enable && csio_ln_fdmi_init(ln))
+ goto err;
+ }
+
+ } /* if (!csio_is_root_ln(ln)) */
+
+ return CSIO_SUCCESS;
+err:
+ return rv;
+}
+
+static void
+csio_ln_exit(struct csio_lnode *ln)
+{
+ struct csio_lnode *pln;
+
+ csio_cleanup_rns(ln);
+ if (csio_is_npiv_ln(ln)) {
+ pln = csio_parent_lnode(ln);
+ kref_put(&pln->fcfinfo->kref, csio_free_fcfinfo);
+ } else {
+ kref_put(&ln->fcfinfo->kref, csio_free_fcfinfo);
+ if (csio_fdmi_enable)
+ csio_ln_fdmi_exit(ln);
+ }
+ ln->fcfinfo = NULL;
+}
+
+/**
+ * csio_lnode_init - Initialize the members of an lnode.
+ * @ln: lnode
+ *
+ */
+csio_retval_t
+csio_lnode_init(struct csio_lnode *ln, struct csio_hw *hw,
+ struct csio_lnode *pln)
+{
+ int rv = -CSIO_INVAL;
+
+ /* Link this lnode to hw */
+ csio_lnode_to_hw(ln) = hw;
+
+ /* Link child to parent if child lnode */
+ if (pln)
+ ln->pln = pln;
+ else
+ ln->pln = NULL;
+
+ /* Initialize scsi_tgt and timers to zero */
+ ln->n_scsi_tgts = 0;
+ ln->last_scan_ntgts = 0;
+ ln->tgt_scan_tick = 0;
+
+ /* Initialize rnode list */
+ INIT_LIST_HEAD(&ln->rnhead);
+ INIT_LIST_HEAD(&ln->cln_head);
+
+ /* Initialize log level for debug */
+ ln->params.log_level = hw->params.log_level;
+
+ if (csio_ln_init(ln))
+ goto err;
+
+ /* Add lnode to list of sibling or children lnodes */
+ spin_lock_irq(&hw->lock);
+ list_add_tail(&ln->sm.sm_list, pln ? &pln->cln_head : &hw->sln_head);
+ if (pln)
+ pln->num_vports++;
+ spin_unlock_irq(&hw->lock);
+
+ hw->num_lns++;
+
+ return CSIO_SUCCESS;
+err:
+ csio_lnode_to_hw(ln) = NULL;
+ return rv;
+}
+
+/**
+ * csio_lnode_exit - De-instantiate an lnode.
+ * @ln: lnode
+ *
+ */
+void
+csio_lnode_exit(struct csio_lnode *ln)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ csio_ln_exit(ln);
+
+ /* Remove this lnode from hw->sln_head */
+ spin_lock_irq(&hw->lock);
+
+ list_del_init(&ln->sm.sm_list);
+
+ /* If it is children lnode, decrement the
+ * counter in its parent lnode
+ */
+ if (ln->pln)
+ ln->pln->num_vports--;
+
+ /* Update root lnode pointer */
+ if (list_empty(&hw->sln_head))
+ hw->rln = NULL;
+ else
+ hw->rln = (struct csio_lnode *)csio_list_next(&hw->sln_head);
+
+ spin_unlock_irq(&hw->lock);
+
+ csio_lnode_to_hw(ln) = NULL;
+ hw->num_lns--;
+}
diff --git a/drivers/scsi/csiostor/csio_rnode.c b/drivers/scsi/csiostor/csio_rnode.c
new file mode 100644
index 0000000..29e47ac
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_rnode.c
@@ -0,0 +1,889 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/string.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_hw.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+#include "csio_fcoe_proto.h"
+
+static csio_retval_t csio_rnode_init(struct csio_rnode *, struct csio_lnode *);
+static void csio_rnode_exit(struct csio_rnode *);
+
+/* Static machine forward declarations */
+static void csio_rns_uninit(struct csio_rnode *, enum csio_rn_ev);
+static void csio_rns_ready(struct csio_rnode *, enum csio_rn_ev);
+static void csio_rns_offline(struct csio_rnode *, enum csio_rn_ev);
+static void csio_rns_disappeared(struct csio_rnode *, enum csio_rn_ev);
+
+/* RNF event mapping */
+static enum csio_rn_ev fwevt_to_rnevt[] = {
+ CSIO_RNFE_NONE, /* None */
+ CSIO_RNFE_LOGGED_IN, /* PLOGI_ACC_RCVD */
+ CSIO_RNFE_NONE, /* PLOGI_RJT_RCVD */
+ CSIO_RNFE_PLOGI_RECV, /* PLOGI_RCVD */
+ CSIO_RNFE_LOGO_RECV, /* PLOGO_RCVD */
+ CSIO_RNFE_PRLI_DONE, /* PRLI_ACC_RCVD */
+ CSIO_RNFE_NONE, /* PRLI_RJT_RCVD */
+ CSIO_RNFE_PRLI_RECV, /* PRLI_RCVD */
+ CSIO_RNFE_PRLO_RECV, /* PRLO_RCVD */
+ CSIO_RNFE_NONE, /* NPORT_ID_CHGD */
+ CSIO_RNFE_LOGO_RECV, /* FLOGO_RCVD */
+ CSIO_RNFE_NONE, /* CLR_VIRT_LNK_RCVD */
+ CSIO_RNFE_LOGGED_IN, /* FLOGI_ACC_RCVD */
+ CSIO_RNFE_NONE, /* FLOGI_RJT_RCVD */
+ CSIO_RNFE_LOGGED_IN, /* FDISC_ACC_RCVD */
+ CSIO_RNFE_NONE, /* FDISC_RJT_RCVD */
+ CSIO_RNFE_NONE, /* FLOGI_TMO_MAX_RETRY */
+ CSIO_RNFE_NONE, /* IMPL_LOGO_ADISC_ACC */
+ CSIO_RNFE_NONE, /* IMPL_LOGO_ADISC_RJT */
+ CSIO_RNFE_NONE, /* IMPL_LOGO_ADISC_CNFLT */
+ CSIO_RNFE_NONE, /* PRLI_TMO */
+ CSIO_RNFE_NONE, /* ADISC_TMO */
+ CSIO_RNFE_NAME_MISSING, /* RSCN_DEV_LOST */
+ CSIO_RNFE_NONE, /* SCR_ACC_RCVD */
+ CSIO_RNFE_NONE, /* ADISC_RJT_RCVD */
+ CSIO_RNFE_NONE, /* LOGO_SNT */
+ CSIO_RNFE_LOGO_RECV, /* PROTO_ERR_IMPL_LOGO */
+};
+
+#define CSIO_FWE_TO_RNFE(_evt) ((_evt > PROTO_ERR_IMPL_LOGO) ? \
+ CSIO_RNFE_NONE : \
+ fwevt_to_rnevt[_evt])
+int
+csio_is_rnode_ready(struct csio_rnode *rn)
+{
+ return csio_match_state(rn, csio_rns_ready);
+}
+
+static int
+csio_is_rnode_uninit(struct csio_rnode *rn)
+{
+ return csio_match_state(rn, csio_rns_uninit);
+}
+
+/*
+ * csio_rn_lookup - Finds the rnode with the given flowid
+ * @ln - lnode
+ * @flowid - flowid.
+ *
+ * Does the rnode lookup on the given lnode and flowid.If no matching entry
+ * found, NULL is returned.
+ */
+static struct csio_rnode *
+csio_rn_lookup(struct csio_lnode *ln, uint32_t flowid)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp;
+ struct csio_rnode *rn;
+
+ list_for_each(tmp, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ if (rn->flowid == flowid)
+ return rn;
+ }
+
+ return NULL;
+}
+
+/*
+ * csio_rn_lookup_wwpn - Finds the rnode with the given wwpn
+ * @ln: lnode
+ * @wwpn: wwpn
+ *
+ * Does the rnode lookup on the given lnode and wwpn. If no matching entry
+ * found, NULL is returned.
+ */
+static struct csio_rnode *
+csio_rn_lookup_wwpn(struct csio_lnode *ln, uint8_t *wwpn)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp;
+ struct csio_rnode *rn;
+
+ list_for_each(tmp, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ if (!memcmp(csio_rn_wwpn(rn), wwpn, 8))
+ return rn;
+ }
+
+ return NULL;
+}
+
+/**
+ * csio_rnode_lookup_portid - Finds the rnode with the given portid
+ * @ln: lnode
+ * @portid: port id
+ *
+ * Lookup the rnode list for a given portid. If no matching entry
+ * found, NULL is returned.
+ */
+struct csio_rnode *
+csio_rnode_lookup_portid(struct csio_lnode *ln, uint32_t portid)
+{
+ struct csio_rnode *rnhead = (struct csio_rnode *) &ln->rnhead;
+ struct list_head *tmp;
+ struct csio_rnode *rn;
+
+ list_for_each(tmp, &rnhead->sm.sm_list) {
+ rn = (struct csio_rnode *) tmp;
+ if (rn->nport_id == portid)
+ return rn;
+ }
+
+ return NULL;
+}
+
+static int
+csio_rn_dup_flowid(struct csio_lnode *ln, uint32_t rdev_flowid,
+ uint32_t *vnp_flowid)
+{
+ struct csio_rnode *rnhead;
+ struct list_head *tmp, *tmp1;
+ struct csio_rnode *rn;
+ struct csio_lnode *ln_tmp;
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ list_for_each(tmp1, &hw->sln_head) {
+ ln_tmp = (struct csio_lnode *) tmp1;
+ if (ln_tmp == ln)
+ continue;
+
+ rnhead = (struct csio_rnode *)&ln_tmp->rnhead;
+ list_for_each(tmp, &rnhead->sm.sm_list) {
+
+ rn = (struct csio_rnode *) tmp;
+ if (csio_is_rnode_ready(rn)) {
+ if (rn->flowid == rdev_flowid) {
+ *vnp_flowid = csio_ln_flowid(ln_tmp);
+ return 1;
+ }
+ }
+ }
+ }
+
+ return 0;
+}
+
+static struct csio_rnode *
+csio_alloc_rnode(struct csio_lnode *ln)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ struct csio_rnode *rn = mempool_alloc(hw->rnode_mempool, GFP_ATOMIC);
+ if (!rn)
+ goto err;
+
+ memset(rn, 0, sizeof(struct csio_rnode));
+ if (csio_rnode_init(rn, ln))
+ goto err_free;
+
+ csio_inc_stats(ln, n_rnode_alloc);
+
+ return rn;
+
+err_free:
+ mempool_free(rn, hw->rnode_mempool);
+err:
+ csio_inc_stats(ln, n_rnode_nomem);
+ return NULL;
+}
+
+static void
+csio_free_rnode(struct csio_rnode *rn)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(csio_rnode_to_lnode(rn));
+
+ csio_rnode_exit(rn);
+ csio_inc_stats(rn->lnp, n_rnode_free);
+ mempool_free(rn, hw->rnode_mempool);
+}
+
+/*
+ * csio_get_rnode - Gets rnode with the given flowid
+ * @ln - lnode
+ * @flowid - flow id.
+ *
+ * Does the rnode lookup on the given lnode and flowid. If no matching
+ * rnode found, then new rnode with given npid is allocated and returned.
+ */
+static struct csio_rnode *
+csio_get_rnode(struct csio_lnode *ln, uint32_t flowid)
+{
+ struct csio_rnode *rn;
+
+ rn = csio_rn_lookup(ln, flowid);
+ if (!rn) {
+ rn = csio_alloc_rnode(ln);
+ if (!rn)
+ return NULL;
+
+ rn->flowid = flowid;
+ }
+
+ return rn;
+}
+
+/*
+ * csio_put_rnode - Frees the given rnode
+ * @ln - lnode
+ * @flowid - flow id.
+ *
+ * Does the rnode lookup on the given lnode and flowid. If no matching
+ * rnode found, then new rnode with given npid is allocated and returned.
+ */
+void
+csio_put_rnode(struct csio_lnode *ln, struct csio_rnode *rn)
+{
+ CSIO_DB_ASSERT(csio_is_rnode_uninit(rn) != 0);
+ csio_free_rnode(rn);
+}
+
+/*
+ * csio_confirm_rnode - confirms rnode based on wwpn.
+ * @ln: lnode
+ * @rdev_flowid: remote device flowid
+ * @rdevp: remote device params
+ * This routines searches other rnode in list having same wwpn of new rnode.
+ * If there is a match, then matched rnode is returned and otherwise new rnode
+ * is returned.
+ * returns rnode.
+ */
+struct csio_rnode *
+csio_confirm_rnode(struct csio_lnode *ln, uint32_t rdev_flowid,
+ struct fcoe_rdev_entry *rdevp)
+{
+ uint8_t rport_type;
+ struct csio_rnode *rn, *match_rn;
+ uint32_t vnp_flowid;
+
+ rport_type =
+ FW_RDEV_WR_RPORT_TYPE_GET(rdevp->rd_xfer_rdy_to_rport_type);
+ /* Drop rdev event for cntrl port */
+ if (rport_type == FAB_CTLR_VNPORT) {
+ csio_ln_dbg(ln,
+ "Unhandled rport_type:%d recv in rdev evt "
+ "ssni:x%x\n", rport_type, rdev_flowid);
+ return NULL;
+ }
+
+ /* Lookup on flowid */
+ rn = csio_rn_lookup(ln, rdev_flowid);
+ if (!rn) {
+
+ /* Drop events with duplicate flowid */
+ if (csio_rn_dup_flowid(ln, rdev_flowid, &vnp_flowid)) {
+ csio_ln_warn(ln,
+ "ssni:%x already active on vnpi:%x",
+ rdev_flowid, vnp_flowid);
+ return NULL;
+ }
+
+ /* skip wwpn lookup for fabric ports, cntrl port */
+ if (rport_type == FLOGI_VFPORT || rport_type == FDISC_VFPORT
+ || rport_type == FAB_CTLR_VNPORT) {
+ goto alloc_rnode;
+ }
+
+ /* Lookup on wwpn for NPORTs */
+ rn = csio_rn_lookup_wwpn(ln, rdevp->wwpn);
+ if (!rn)
+ goto alloc_rnode;
+
+ /* found rn */
+ goto found_rnode;
+ } else {
+ /* verify rnode found for fabric ports, cntrl port */
+ if (rport_type == FLOGI_VFPORT || rport_type == FDISC_VFPORT
+ || rport_type == FAB_CTLR_VNPORT) {
+
+ /* Rnode role mismatch. Allocate new rnode */
+ if (rn->role == CSIO_RNFR_NS ||
+ rn->role == CSIO_RNFR_NPORT) {
+ csio_ln_dbg(ln,
+ "rnode role mismatch found ssni:x%x "
+ "role:%d new_type:%d\n",
+ rdev_flowid, rn->role, rport_type);
+ if (csio_is_rnode_ready(rn)) {
+ csio_ln_warn(ln,
+ "rnode is already"
+ "active ssni:x%x\n",
+ rdev_flowid);
+ CSIO_DB_ASSERT(0);
+ }
+ csio_rn_flowid(rn) = CSIO_INVALID_IDX;
+ goto alloc_rnode;
+ } else
+ goto found_rnode;
+ }
+
+ /* wwpn match */
+ if (!memcmp(csio_rn_wwpn(rn), rdevp->wwpn, 8)) {
+ /* Update rn */
+ goto found_rnode;
+ }
+
+ /* Search for rnode that have same wwpn */
+ match_rn = csio_rn_lookup_wwpn(ln, rdevp->wwpn);
+ if (match_rn != NULL) {
+ csio_ln_dbg(ln,
+ "ssni:x%x changed for rport name(wwpn):%llx "
+ "did:x%x\n", rdev_flowid,
+ wwn_to_u64(rdevp->wwpn),
+ match_rn->nport_id);
+ csio_rn_flowid(rn) = CSIO_INVALID_IDX;
+ rn = match_rn;
+ csio_inc_stats(ln, n_rnode_match);
+ } else {
+ csio_ln_dbg(ln,
+ "rnode wwpn mismatch found ssni:x%x "
+ "name(wwpn):%llx\n",
+ rdev_flowid,
+ wwn_to_u64(csio_rn_wwpn(rn)));
+ if (csio_is_rnode_ready(rn)) {
+ csio_ln_warn(ln,
+ "rnode is already active "
+ "wwpn:%llx ssni:x%x\n",
+ wwn_to_u64(csio_rn_wwpn(rn)),
+ rdev_flowid);
+ CSIO_DB_ASSERT(0);
+ }
+ csio_rn_flowid(rn) = CSIO_INVALID_IDX;
+ goto alloc_rnode;
+ }
+ }
+
+found_rnode:
+ csio_ln_dbg(ln, "found rnode:%p ssni:x%x name(wwpn):%llx\n",
+ rn, rdev_flowid, wwn_to_u64(rdevp->wwpn));
+
+ /* Update flowid */
+ csio_rn_flowid(rn) = rdev_flowid;
+
+ /* update rdev entry */
+ rn->rdev_entry = rdevp;
+ return rn;
+
+alloc_rnode:
+ rn = csio_get_rnode(ln, rdev_flowid);
+ if (!rn)
+ return NULL;
+
+ csio_ln_dbg(ln, "alloc rnode:%p ssni:x%x name(wwpn):%llx\n",
+ rn, rdev_flowid, wwn_to_u64(rdevp->wwpn));
+
+ /* update rdev entry */
+ rn->rdev_entry = rdevp;
+ return rn;
+}
+
+/*
+ * csio_rn_verify_rparams - verify rparams.
+ * @ln: lnode
+ * @rn: rnode
+ * @rdevp: remote device params
+ * returns success if rparams are verified.
+ */
+static csio_retval_t
+csio_rn_verify_rparams(struct csio_lnode *ln, struct csio_rnode *rn,
+ struct fcoe_rdev_entry *rdevp)
+{
+ uint8_t null[8];
+ uint8_t rport_type;
+ uint8_t fc_class;
+ uint32_t *did;
+
+ did = (uint32_t *) &rdevp->r_id[0];
+ rport_type =
+ FW_RDEV_WR_RPORT_TYPE_GET(rdevp->rd_xfer_rdy_to_rport_type);
+ switch (rport_type) {
+ case FLOGI_VFPORT:
+ rn->role = CSIO_RNFR_FABRIC;
+ if (((ntohl(*did) >> 8) & DID_MASK) != FABRIC_DID) {
+ csio_ln_err(ln, "ssni:x%x invalid fabric portid\n",
+ csio_rn_flowid(rn));
+ return CSIO_INVAL;
+ }
+ /* NPIV support */
+ if (FW_RDEV_WR_NPIV_GET(rdevp->vft_to_qos))
+ ln->flags |= CSIO_LNF_NPIVSUPP;
+
+ break;
+
+ case NS_VNPORT:
+ rn->role = CSIO_RNFR_NS;
+ if (((ntohl(*did) >> 8) & DID_MASK) != NS_DID) {
+ csio_ln_err(ln, "ssni:x%x invalid fabric portid\n",
+ csio_rn_flowid(rn));
+ return CSIO_INVAL;
+ }
+ break;
+
+ case REG_FC4_VNPORT:
+ case REG_VNPORT:
+ rn->role = CSIO_RNFR_NPORT;
+ if (rdevp->event_cause == PRLI_ACC_RCVD ||
+ rdevp->event_cause == PRLI_RCVD) {
+ if (FW_RDEV_WR_TASK_RETRY_ID_GET(
+ rdevp->enh_disc_to_tgt))
+ rn->fcp_flags |= V_PRLI_FCP_DATA_OVERLAY(1);
+
+ if (FW_RDEV_WR_RETRY_GET(rdevp->enh_disc_to_tgt))
+ rn->fcp_flags |= V_PRLI_FCP_RETRY(1);
+
+ if (FW_RDEV_WR_CONF_CMPL_GET(rdevp->enh_disc_to_tgt))
+ rn->fcp_flags |=
+ V_PRLI_FCP_CONF_COMPL_ALLOWED(1);
+
+ if (FW_RDEV_WR_TGT_GET(rdevp->enh_disc_to_tgt))
+ rn->role |= CSIO_RNFR_TARGET;
+
+ if (FW_RDEV_WR_INI_GET(rdevp->enh_disc_to_tgt))
+ rn->role |= CSIO_RNFR_INITIATOR;
+ }
+
+ break;
+
+ case FDMI_VNPORT:
+ case FAB_CTLR_VNPORT:
+ rn->role = 0;
+ break;
+
+ default:
+ csio_ln_err(ln, "ssni:x%x invalid rport type recv x%x\n",
+ csio_rn_flowid(rn), rport_type);
+ return CSIO_INVAL;
+ }
+
+ /* validate wwpn/wwnn for Name server/remote port */
+ if (rport_type == REG_VNPORT || rport_type == NS_VNPORT) {
+ memset(null, 0, 8);
+ if (!memcmp(rdevp->wwnn, null, 8)) {
+ csio_ln_err(ln,
+ "ssni:x%x invalid wwnn received from"
+ " rport did:x%x\n",
+ csio_rn_flowid(rn),
+ (ntohl(*did) & DID_MASK));
+ return CSIO_INVAL;
+ }
+
+ if (!memcmp(rdevp->wwpn, null, 8)) {
+ csio_ln_err(ln,
+ "ssni:x%x invalid wwpn received from"
+ " rport did:x%x\n",
+ csio_rn_flowid(rn),
+ (ntohl(*did) & DID_MASK));
+ return CSIO_INVAL;
+ }
+
+ }
+
+ /* Copy wwnn, wwpn and nport id */
+ rn->nport_id = (ntohl(*did) >> 8) & DID_MASK;
+ memcpy(csio_rn_wwnn(rn), rdevp->wwnn, 8);
+ memcpy(csio_rn_wwpn(rn), rdevp->wwpn, 8);
+ rn->rn_sparm.csp.rcv_sz = ntohs(rdevp->rcv_fr_sz);
+ fc_class = FW_RDEV_WR_CLASS_GET(rdevp->vft_to_qos);
+ rn->rn_sparm.clsp[fc_class - 1].serv_option = V_SP_CLASS_SUPPORT(1);
+ return CSIO_SUCCESS;
+}
+
+static void
+__csio_reg_rnode(struct csio_rnode *rn)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(csio_rnode_to_lnode(rn));
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+
+ spin_unlock_irq(&hw->lock);
+ csio_reg_rnode(rn);
+ spin_lock_irq(&hw->lock);
+
+ if (rn->nport_id == FDMI_DID)
+ csio_ln_fdmi_start(ln, (void *) rn);
+}
+
+static void
+__csio_unreg_rnode(struct csio_rnode *rn)
+{
+ struct csio_hw *hw = csio_lnode_to_hw(csio_rnode_to_lnode(rn));
+ LIST_HEAD(tmp_q);
+ int cmpl = 0;
+
+ if (!list_empty(&rn->host_cmpl_q)) {
+ csio_dbg(hw, "Returning completion queue I/Os\n");
+ list_splice_tail_init(&rn->host_cmpl_q, &tmp_q);
+ cmpl = 1;
+ }
+
+ spin_unlock_irq(&hw->lock);
+ csio_unreg_rnode(rn);
+ spin_lock_irq(&hw->lock);
+
+ /* Cleanup I/Os that were waiting for rnode to unregister */
+ if (cmpl)
+ csio_scsi_cleanup_io_q(csio_hw_to_scsim(hw), &tmp_q);
+
+}
+
+/*****************************************************************************/
+/* START: Rnode SM */
+/*****************************************************************************/
+
+/*
+ * csio_rns_uninit -
+ * @rn - rnode
+ * @evt - SM event.
+ *
+ */
+static void
+csio_rns_uninit(struct csio_rnode *rn, enum csio_rn_ev evt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ csio_inc_stats(rn, n_evt_sm[evt]);
+
+ switch (evt) {
+ case CSIO_RNFE_LOGGED_IN:
+ case CSIO_RNFE_PLOGI_RECV:
+ ret = csio_rn_verify_rparams(ln, rn, rn->rdev_entry);
+ if (!ret) {
+ csio_set_state(&rn->sm, csio_rns_ready);
+ __csio_reg_rnode(rn);
+ } else {
+ csio_inc_stats(rn, n_err_inval);
+ }
+ break;
+ case CSIO_RNFE_LOGO_RECV:
+ csio_ln_dbg(ln,
+ "ssni:x%x Ignoring event %d recv "
+ "in rn state[uninit]\n", csio_rn_flowid(rn), evt);
+ csio_inc_stats(rn, n_evt_drop);
+ break;
+ default:
+ csio_ln_dbg(ln,
+ "ssni:x%x unexp event %d recv "
+ "in rn state[uninit]\n", csio_rn_flowid(rn), evt);
+ csio_inc_stats(rn, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_rns_ready -
+ * @rn - rnode
+ * @evt - SM event.
+ *
+ */
+static void
+csio_rns_ready(struct csio_rnode *rn, enum csio_rn_ev evt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ csio_inc_stats(rn, n_evt_sm[evt]);
+
+ switch (evt) {
+ case CSIO_RNFE_LOGGED_IN:
+ case CSIO_RNFE_PLOGI_RECV:
+ csio_ln_dbg(ln,
+ "ssni:x%x Ignoring event %d recv from did:x%x "
+ "in rn state[ready]\n", csio_rn_flowid(rn), evt,
+ rn->nport_id);
+ csio_inc_stats(rn, n_evt_drop);
+ break;
+
+ case CSIO_RNFE_PRLI_DONE:
+ case CSIO_RNFE_PRLI_RECV:
+ ret = csio_rn_verify_rparams(ln, rn, rn->rdev_entry);
+ if (!ret)
+ __csio_reg_rnode(rn);
+ else
+ csio_inc_stats(rn, n_err_inval);
+
+ break;
+ case CSIO_RNFE_DOWN:
+ csio_set_state(&rn->sm, csio_rns_offline);
+ __csio_unreg_rnode(rn);
+
+ /* FW expected to internally aborted outstanding SCSI WRs
+ * and return all SCSI WRs to host with status "ABORTED".
+ */
+ break;
+
+ case CSIO_RNFE_LOGO_RECV:
+ csio_set_state(&rn->sm, csio_rns_offline);
+
+ __csio_unreg_rnode(rn);
+
+ /* FW expected to internally aborted outstanding SCSI WRs
+ * and return all SCSI WRs to host with status "ABORTED".
+ */
+ break;
+
+ case CSIO_RNFE_CLOSE:
+ /*
+ * Each rnode receives CLOSE event when driver is removed or
+ * device is reset
+ * Note: All outstanding IOs on remote port need to returned
+ * to uppper layer with appropriate error before sending
+ * CLOSE event
+ */
+ csio_set_state(&rn->sm, csio_rns_uninit);
+ __csio_unreg_rnode(rn);
+ break;
+
+ case CSIO_RNFE_NAME_MISSING:
+ csio_set_state(&rn->sm, csio_rns_disappeared);
+ __csio_unreg_rnode(rn);
+
+ /*
+ * FW expected to internally aborted outstanding SCSI WRs
+ * and return all SCSI WRs to host with status "ABORTED".
+ */
+
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "ssni:x%x unexp event %d recv from did:x%x "
+ "in rn state[uninit]\n", csio_rn_flowid(rn), evt,
+ rn->nport_id);
+ csio_inc_stats(rn, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_rns_offline -
+ * @rn - rnode
+ * @evt - SM event.
+ *
+ */
+static void
+csio_rns_offline(struct csio_rnode *rn, enum csio_rn_ev evt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ csio_inc_stats(rn, n_evt_sm[evt]);
+
+ switch (evt) {
+ case CSIO_RNFE_LOGGED_IN:
+ case CSIO_RNFE_PLOGI_RECV:
+ ret = csio_rn_verify_rparams(ln, rn, rn->rdev_entry);
+ if (!ret) {
+ csio_set_state(&rn->sm, csio_rns_ready);
+ __csio_reg_rnode(rn);
+ } else {
+ csio_inc_stats(rn, n_err_inval);
+ csio_post_event(&rn->sm, CSIO_RNFE_CLOSE);
+ }
+ break;
+
+ case CSIO_RNFE_DOWN:
+ csio_ln_dbg(ln,
+ "ssni:x%x Ignoring event %d recv from did:x%x "
+ "in rn state[offline]\n", csio_rn_flowid(rn), evt,
+ rn->nport_id);
+ csio_inc_stats(rn, n_evt_drop);
+ break;
+
+ case CSIO_RNFE_CLOSE:
+ /* Each rnode receives CLOSE event when driver is removed or
+ * device is reset
+ * Note: All outstanding IOs on remote port need to returned
+ * to uppper layer with appropriate error before sending
+ * CLOSE event
+ */
+ csio_set_state(&rn->sm, csio_rns_uninit);
+ break;
+
+ case CSIO_RNFE_NAME_MISSING:
+ csio_set_state(&rn->sm, csio_rns_disappeared);
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "ssni:x%x unexp event %d recv from did:x%x "
+ "in rn state[offline]\n", csio_rn_flowid(rn), evt,
+ rn->nport_id);
+ csio_inc_stats(rn, n_evt_unexp);
+ break;
+ }
+}
+
+/*
+ * csio_rns_disappeared -
+ * @rn - rnode
+ * @evt - SM event.
+ *
+ */
+static void
+csio_rns_disappeared(struct csio_rnode *rn, enum csio_rn_ev evt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ csio_retval_t ret = CSIO_SUCCESS;
+
+ csio_inc_stats(rn, n_evt_sm[evt]);
+
+ switch (evt) {
+ case CSIO_RNFE_LOGGED_IN:
+ case CSIO_RNFE_PLOGI_RECV:
+ ret = csio_rn_verify_rparams(ln, rn, rn->rdev_entry);
+ if (!ret) {
+ csio_set_state(&rn->sm, csio_rns_ready);
+ __csio_reg_rnode(rn);
+ } else {
+ csio_inc_stats(rn, n_err_inval);
+ csio_post_event(&rn->sm, CSIO_RNFE_CLOSE);
+ }
+ break;
+
+ case CSIO_RNFE_CLOSE:
+ /* Each rnode receives CLOSE event when driver is removed or
+ * device is reset.
+ * Note: All outstanding IOs on remote port need to returned
+ * to uppper layer with appropriate error before sending
+ * CLOSE event
+ */
+ csio_set_state(&rn->sm, csio_rns_uninit);
+ break;
+
+ case CSIO_RNFE_DOWN:
+ case CSIO_RNFE_NAME_MISSING:
+ csio_ln_dbg(ln,
+ "ssni:x%x Ignoring event %d recv from did x%x"
+ "in rn state[disappeared]\n", csio_rn_flowid(rn),
+ evt, rn->nport_id);
+ break;
+
+ default:
+ csio_ln_dbg(ln,
+ "ssni:x%x unexp event %d recv from did x%x"
+ "in rn state[disappeared]\n", csio_rn_flowid(rn),
+ evt, rn->nport_id);
+ csio_inc_stats(rn, n_evt_unexp);
+ break;
+ }
+}
+
+/*****************************************************************************/
+/* END: Rnode SM */
+/*****************************************************************************/
+
+/*
+ * csio_rnode_devloss_handler - Device loss event handler
+ * @rn: rnode
+ *
+ * Post event to close rnode SM and free rnode.
+ */
+void
+csio_rnode_devloss_handler(struct csio_rnode *rn)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+
+ /* ignore if same rnode came back as online */
+ if (csio_is_rnode_ready(rn))
+ return;
+
+ csio_post_event(&rn->sm, CSIO_RNFE_CLOSE);
+
+ /* Free rn if in uninit state */
+ if (csio_is_rnode_uninit(rn))
+ csio_put_rnode(ln, rn);
+}
+
+/**
+ * csio_rnode_fwevt_handler - Event handler for firmware rnode events.
+ * @rn: rnode
+ *
+ */
+void
+csio_rnode_fwevt_handler(struct csio_rnode *rn, uint8_t fwevt)
+{
+ struct csio_lnode *ln = csio_rnode_to_lnode(rn);
+ enum csio_rn_ev evt;
+
+ evt = CSIO_FWE_TO_RNFE(fwevt);
+ if (!evt) {
+ csio_ln_err(ln, "ssni:x%x Unhandled FW Rdev event: %d\n",
+ csio_rn_flowid(rn), fwevt);
+ csio_inc_stats(rn, n_evt_unexp);
+ return;
+ }
+ csio_inc_stats(rn, n_evt_fw[fwevt]);
+
+ /* Track previous & current events for debugging */
+ rn->prev_evt = rn->cur_evt;
+ rn->cur_evt = fwevt;
+
+ /* Post event to rnode SM */
+ csio_post_event(&rn->sm, evt);
+
+ /* Free rn if in uninit state */
+ if (csio_is_rnode_uninit(rn))
+ csio_put_rnode(ln, rn);
+}
+
+/*
+ * csio_rnode_init - Initialize rnode.
+ * @rn: RNode
+ * @ln: Associated lnode
+ *
+ * Caller is responsible for holding the lock. The lock is required
+ * to be held for inserting the rnode in ln->rnhead list.
+ */
+static csio_retval_t
+csio_rnode_init(struct csio_rnode *rn, struct csio_lnode *ln)
+{
+ csio_rnode_to_lnode(rn) = ln;
+ csio_init_state(&rn->sm, csio_rns_uninit);
+ INIT_LIST_HEAD(&rn->host_cmpl_q);
+ csio_rn_flowid(rn) = CSIO_INVALID_IDX;
+
+ /* Add rnode to list of lnodes->rnhead */
+ list_add_tail(&rn->sm.sm_list, &ln->rnhead);
+
+ return CSIO_SUCCESS;
+}
+
+static void
+csio_rnode_exit(struct csio_rnode *rn)
+{
+ list_del_init(&rn->sm.sm_list);
+ CSIO_DB_ASSERT(list_empty(&rn->host_cmpl_q));
+}
--
1.7.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5).
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
` (3 preceding siblings ...)
2012-08-23 22:27 ` [PATCH 4/8] csiostor: Chelsio FCoE offload driver submission (sources part 4) Naresh Kumar Inna
@ 2012-08-23 22:27 ` Naresh Kumar Inna
2012-08-23 19:48 ` Nicholas A. Bellinger
2012-08-23 22:27 ` [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1) Naresh Kumar Inna
` (4 subsequent siblings)
9 siblings, 1 reply; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
This patch contains code to implement the interrupt handling and the fast
path I/O functionality. The interrupt handling includes allocation of
MSIX vectors, registering and implemeting the interrupt service routines.
The fast path I/O functionality includes posting the I/O request to firmware
via Work Requests, tracking/completing them, and handling task management
requests. SCSI midlayer host template implementation is also covered by
this patch.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/csiostor/csio_isr.c | 631 ++++++++++
drivers/scsi/csiostor/csio_scsi.c | 2498 +++++++++++++++++++++++++++++++++++++
2 files changed, 3129 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_isr.c
create mode 100644 drivers/scsi/csiostor/csio_scsi.c
diff --git a/drivers/scsi/csiostor/csio_isr.c b/drivers/scsi/csiostor/csio_isr.c
new file mode 100644
index 0000000..96633e9
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_isr.c
@@ -0,0 +1,631 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/pci.h>
+#include <linux/interrupt.h>
+#include <linux/cpumask.h>
+#include <linux/string.h>
+
+#include "csio_init.h"
+#include "csio_hw.h"
+
+static irqreturn_t
+csio_nondata_isr(int irq, void *dev_id)
+{
+ struct csio_hw *hw = (struct csio_hw *) dev_id;
+ csio_retval_t rv;
+ unsigned long flags;
+
+ if (unlikely(!hw))
+ return IRQ_NONE;
+
+ if (unlikely(pci_channel_offline(hw->pdev))) {
+ csio_inc_stats(hw, n_pcich_offline);
+ return IRQ_NONE;
+ }
+
+ spin_lock_irqsave(&hw->lock, flags);
+ csio_hw_slow_intr_handler(hw);
+ rv = csio_mb_isr_handler(hw);
+
+ if (rv == CSIO_SUCCESS && !(hw->flags & CSIO_HWF_FWEVT_PENDING)) {
+ hw->flags |= CSIO_HWF_FWEVT_PENDING;
+ spin_unlock_irqrestore(&hw->lock, flags);
+ schedule_work(&hw->evtq_work);
+ return IRQ_HANDLED;
+ }
+ spin_unlock_irqrestore(&hw->lock, flags);
+ return IRQ_HANDLED;
+}
+
+/*
+ * csio_fwevt_handler - Common FW event handler routine.
+ * @hw: HW module.
+ *
+ * This is the ISR for FW events. It is shared b/w MSIX
+ * and INTx handlers.
+ */
+static void
+csio_fwevt_handler(struct csio_hw *hw)
+{
+ csio_retval_t rv;
+ unsigned long flags;
+
+ rv = csio_fwevtq_handler(hw);
+
+ spin_lock_irqsave(&hw->lock, flags);
+ if (rv == CSIO_SUCCESS && !(hw->flags & CSIO_HWF_FWEVT_PENDING)) {
+ hw->flags |= CSIO_HWF_FWEVT_PENDING;
+ spin_unlock_irqrestore(&hw->lock, flags);
+ schedule_work(&hw->evtq_work);
+ return;
+ }
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+} /* csio_fwevt_handler */
+
+/*
+ * csio_fwevt_isr() - FW events MSIX ISR
+ * @irq:
+ * @dev_id:
+ *
+ * Process WRs on the FW event queue.
+ *
+ */
+static irqreturn_t
+csio_fwevt_isr(int irq, void *dev_id)
+{
+ struct csio_hw *hw = (struct csio_hw *) dev_id;
+
+ if (unlikely(!hw))
+ return IRQ_NONE;
+
+ if (unlikely(pci_channel_offline(hw->pdev))) {
+ csio_inc_stats(hw, n_pcich_offline);
+ return IRQ_NONE;
+ }
+
+ csio_fwevt_handler(hw);
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * csio_fwevt_isr() - INTx wrapper for handling FW events.
+ * @irq:
+ * @dev_id:
+ */
+void
+csio_fwevt_intx_handler(struct csio_hw *hw, void *wr, uint32_t len,
+ struct csio_fl_dma_buf *flb, void *priv)
+{
+ csio_fwevt_handler(hw);
+} /* csio_fwevt_intx_handler */
+
+/*
+ * csio_process_scsi_cmpl - Process a SCSI WR completion.
+ * @hw: HW module.
+ * @wr: The completed WR from the ingress queue.
+ * @len: Length of the WR.
+ * @flb: Freelist buffer array.
+ *
+ */
+static void
+csio_process_scsi_cmpl(struct csio_hw *hw, void *wr, uint32_t len,
+ struct csio_fl_dma_buf *flb, void *cbfn_q)
+{
+ struct csio_ioreq *ioreq;
+ uint8_t *scsiwr;
+ uint8_t subop;
+ void *cmnd;
+ unsigned long flags;
+
+ ioreq = csio_scsi_cmpl_handler(hw, wr, len, flb, NULL, &scsiwr);
+ if (likely(ioreq)) {
+ if (unlikely(*scsiwr == FW_SCSI_ABRT_CLS_WR)) {
+ subop = FW_SCSI_ABRT_CLS_WR_SUB_OPCODE_GET(
+ ((struct fw_scsi_abrt_cls_wr *)
+ scsiwr)->sub_opcode_to_chk_all_io);
+
+ csio_dbg(hw, "%s cmpl recvd ioreq:%p status:%d\n",
+ subop ? "Close" : "Abort",
+ ioreq, ioreq->wr_status);
+
+ spin_lock_irqsave(&hw->lock, flags);
+ if (subop)
+ csio_scsi_closed(ioreq,
+ (struct list_head *)cbfn_q);
+ else
+ csio_scsi_aborted(ioreq,
+ (struct list_head *)cbfn_q);
+ /*
+ * We call scsi_done for I/Os that driver thinks aborts
+ * have timed out. If there is a race caused by FW
+ * completing abort at the exact same time that the
+ * driver has deteced the abort timeout, the following
+ * check prevents calling of scsi_done twice for the
+ * same command: once from the eh_abort_handler, another
+ * from csio_scsi_isr_handler(). This also avoids the
+ * need to check if csio_scsi_cmnd(req) is NULL in the
+ * fast path.
+ */
+ cmnd = csio_scsi_cmnd(ioreq);
+ if (unlikely(cmnd == NULL))
+ list_del_init(&ioreq->sm.sm_list);
+
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+ if (unlikely(cmnd == NULL))
+ csio_put_scsi_ioreq_lock(hw,
+ csio_hw_to_scsim(hw), ioreq);
+ } else {
+ spin_lock_irqsave(&hw->lock, flags);
+ csio_scsi_completed(ioreq, (struct list_head *)cbfn_q);
+ spin_unlock_irqrestore(&hw->lock, flags);
+ }
+ }
+}
+
+/*
+ * csio_scsi_isr_handler() - Common SCSI ISR handler.
+ * @iq: Ingress queue pointer.
+ *
+ * Processes SCSI completions on the SCSI IQ indicated by scm->iq_idx
+ * by calling csio_wr_process_iq_idx. If there are completions on the
+ * isr_cbfn_q, yank them out into a local queue and call their io_cbfns.
+ * Once done, add these completions onto the freelist.
+ * This routine is shared b/w MSIX and INTx.
+ */
+static inline irqreturn_t
+csio_scsi_isr_handler(struct csio_q *iq)
+{
+ struct csio_hw *hw = (struct csio_hw *)iq->owner;
+ LIST_HEAD(cbfn_q);
+ struct list_head *tmp;
+ struct csio_scsim *scm;
+ struct csio_ioreq *ioreq;
+ int isr_completions = 0;
+
+ scm = csio_hw_to_scsim(hw);
+
+ if (unlikely(csio_wr_process_iq(hw, iq, csio_process_scsi_cmpl,
+ &cbfn_q) != CSIO_SUCCESS))
+ return IRQ_NONE;
+
+ /* Call back the completion routines */
+ list_for_each(tmp, &cbfn_q) {
+ ioreq = (struct csio_ioreq *)tmp;
+ isr_completions++;
+ ioreq->io_cbfn(hw, ioreq);
+#ifdef __CSIO_DDP_SUPPORT__
+ /* Release ddp buffer if used for this req */
+ if (unlikely(ioreq->dcopy))
+ csio_put_scsi_ddp_list_lock(hw, scm, &ioreq->gen_list,
+ ioreq->nsge);
+#endif
+ }
+
+ if (isr_completions) {
+ /* Return the ioreqs back to ioreq->freelist */
+ csio_put_scsi_ioreq_list_lock(hw, scm, &cbfn_q,
+ isr_completions);
+ }
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * csio_scsi_isr() - SCSI MSIX handler
+ * @irq:
+ * @dev_id:
+ *
+ * This is the top level SCSI MSIX handler. Calls csio_scsi_isr_handler()
+ * for handling SCSI completions.
+ */
+static irqreturn_t
+csio_scsi_isr(int irq, void *dev_id)
+{
+ struct csio_q *iq = (struct csio_q *) dev_id;
+ struct csio_hw *hw;
+
+ if (unlikely(!iq))
+ return IRQ_NONE;
+
+ hw = (struct csio_hw *)iq->owner;
+
+ if (unlikely(pci_channel_offline(hw->pdev))) {
+ csio_inc_stats(hw, n_pcich_offline);
+ return IRQ_NONE;
+ }
+
+ csio_scsi_isr_handler(iq);
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * csio_scsi_intx_handler() - SCSI INTx handler
+ * @irq:
+ * @dev_id:
+ *
+ * This is the top level SCSI INTx handler. Calls csio_scsi_isr_handler()
+ * for handling SCSI completions.
+ */
+void
+csio_scsi_intx_handler(struct csio_hw *hw, void *wr, uint32_t len,
+ struct csio_fl_dma_buf *flb, void *priv)
+{
+ struct csio_q *iq = priv;
+
+ csio_scsi_isr_handler(iq);
+
+} /* csio_scsi_intx_handler */
+
+/*
+ * csio_fcoe_isr() - INTx/MSI interrupt service routine for FCoE.
+ * @irq:
+ * @dev_id:
+ *
+ *
+ */
+static irqreturn_t
+csio_fcoe_isr(int irq, void *dev_id)
+{
+ struct csio_hw *hw = (struct csio_hw *) dev_id;
+ struct csio_q *intx_q = NULL;
+ csio_retval_t rv;
+ irqreturn_t ret = IRQ_NONE;
+ unsigned long flags;
+
+ if (unlikely(!hw))
+ return IRQ_NONE;
+
+ if (unlikely(pci_channel_offline(hw->pdev))) {
+ csio_inc_stats(hw, n_pcich_offline);
+ return IRQ_NONE;
+ }
+
+ /* Disable the interrupt for this PCI function. */
+ if (hw->intr_mode == CSIO_IM_INTX)
+ csio_wr_reg32(hw, 0, MYPF_REG(PCIE_PF_CLI));
+
+ /*
+ * The read in the following function will flush the
+ * above write.
+ */
+ if (csio_hw_slow_intr_handler(hw))
+ ret = IRQ_HANDLED;
+
+ /* Get the INTx Forward interrupt IQ. */
+ intx_q = csio_get_q(hw, hw->intr_iq_idx);
+
+ CSIO_DB_ASSERT(intx_q);
+
+ /* IQ handler is not possible for intx_q, hence pass in NULL */
+ if (likely(csio_wr_process_iq(hw, intx_q, NULL, NULL) == CSIO_SUCCESS))
+ ret = IRQ_HANDLED;
+
+ spin_lock_irqsave(&hw->lock, flags);
+ rv = csio_mb_isr_handler(hw);
+ if (rv == CSIO_SUCCESS && !(hw->flags & CSIO_HWF_FWEVT_PENDING)) {
+ hw->flags |= CSIO_HWF_FWEVT_PENDING;
+ spin_unlock_irqrestore(&hw->lock, flags);
+ schedule_work(&hw->evtq_work);
+ return IRQ_HANDLED;
+ }
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+ return ret;
+}
+
+#define csio_extra_msix_desc(_desc, _len, _str, _arg1, _arg2, _arg3) \
+do { \
+ memset((_desc), 0, (_len) + 1); \
+ snprintf((_desc), (_len), (_str), (_arg1), (_arg2), (_arg3)); \
+} while (0)
+
+static void
+csio_add_msix_desc(struct csio_hw *hw)
+{
+ int i;
+ struct csio_msix_entries *entryp = &hw->msix_entries[0];
+ int k = CSIO_EXTRA_VECS;
+ int len = sizeof(entryp->desc) - 1;
+ int cnt = hw->num_sqsets + k;
+
+ /* Non-data vector */
+ csio_extra_msix_desc(entryp->desc, len, "csio-%02x:%02x:%x-nondata",
+ CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
+ CSIO_PCI_FUNC(hw));
+ entryp++;
+ csio_extra_msix_desc(entryp->desc, len, "csio-%02x:%02x:%x-fwevt",
+ CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
+ CSIO_PCI_FUNC(hw));
+ entryp++;
+
+ /* Name SCSI vecs */
+ for (i = k; i < cnt; i++, entryp++) {
+ memset(entryp->desc, 0, len + 1);
+ snprintf(entryp->desc, len, "csio-%02x:%02x:%x-scsi%d",
+ CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
+ CSIO_PCI_FUNC(hw), i - CSIO_EXTRA_VECS);
+ }
+}
+
+csio_retval_t
+csio_request_irqs(struct csio_hw *hw)
+{
+ int rv, i, j, k = 0;
+ struct csio_msix_entries *entryp = &hw->msix_entries[0];
+ struct csio_scsi_cpu_info *info;
+
+ if (hw->intr_mode != CSIO_IM_MSIX) {
+ rv = request_irq(hw->pdev->irq, csio_fcoe_isr,
+ (hw->intr_mode == CSIO_IM_MSI) ?
+ 0 : IRQF_SHARED,
+ KBUILD_MODNAME, hw);
+ if (rv) {
+ if (hw->intr_mode == CSIO_IM_MSI)
+ pci_disable_msi(hw->pdev);
+ csio_err(hw, "Failed to allocate interrupt line.\n");
+ return CSIO_INVAL;
+ }
+
+ goto out;
+ }
+
+ /* Add the MSIX vector descriptions */
+ csio_add_msix_desc(hw);
+
+ rv = request_irq(entryp[k].vector, csio_nondata_isr, 0,
+ entryp[k].desc, hw);
+ if (rv) {
+ csio_err(hw, "IRQ request failed for vec %d err:%d\n",
+ entryp[k].vector, rv);
+ goto err;
+ }
+
+ entryp[k++].dev_id = (void *)hw;
+
+ rv = request_irq(entryp[k].vector, csio_fwevt_isr, 0,
+ entryp[k].desc, hw);
+ if (rv) {
+ csio_err(hw, "IRQ request failed for vec %d err:%d\n",
+ entryp[k].vector, rv);
+ goto err;
+ }
+
+ entryp[k++].dev_id = (void *)hw;
+
+ /* Allocate IRQs for SCSI */
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+ for (j = 0; j < info->max_cpus; j++, k++) {
+ struct csio_scsi_qset *sqset = &hw->sqset[i][j];
+ struct csio_q *q = hw->wrm.q_arr[sqset->iq_idx];
+
+ rv = request_irq(entryp[k].vector, csio_scsi_isr, 0,
+ entryp[k].desc, q);
+ if (rv) {
+ csio_err(hw,
+ "IRQ request failed for vec %d err:%d\n",
+ entryp[k].vector, rv);
+ goto err;
+ }
+
+ entryp[k].dev_id = (void *)q;
+
+ } /* for all scsi cpus */
+ } /* for all ports */
+
+out:
+ hw->flags |= CSIO_HWF_HOST_INTR_ENABLED;
+
+ return CSIO_SUCCESS;
+
+err:
+ for (i = 0; i < k; i++) {
+ entryp = &hw->msix_entries[i];
+ free_irq(entryp->vector, entryp->dev_id);
+ }
+ pci_disable_msix(hw->pdev);
+
+ return CSIO_INVAL;
+}
+
+static void
+csio_disable_msix(struct csio_hw *hw, bool free)
+{
+ int i;
+ struct csio_msix_entries *entryp;
+ int cnt = hw->num_sqsets + CSIO_EXTRA_VECS;
+
+ if (free) {
+ for (i = 0; i < cnt; i++) {
+ entryp = &hw->msix_entries[i];
+ free_irq(entryp->vector, entryp->dev_id);
+ }
+ }
+ pci_disable_msix(hw->pdev);
+}
+
+/* Reduce per-port max possible CPUs */
+static void
+csio_reduce_sqsets(struct csio_hw *hw, int cnt)
+{
+ int i;
+ struct csio_scsi_cpu_info *info;
+
+ while (cnt < hw->num_sqsets) {
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+ if (info->max_cpus > 1) {
+ info->max_cpus--;
+ hw->num_sqsets--;
+ if (hw->num_sqsets <= cnt)
+ break;
+ }
+ }
+ }
+
+ csio_dbg(hw, "Reduced sqsets to %d\n", hw->num_sqsets);
+}
+
+static csio_retval_t
+csio_enable_msix(struct csio_hw *hw)
+{
+ int rv, i, j, k, n, min, cnt;
+ struct csio_msix_entries *entryp;
+ struct msix_entry *entries;
+ int extra = CSIO_EXTRA_VECS;
+ struct csio_scsi_cpu_info *info;
+
+ min = hw->num_pports + extra;
+ cnt = hw->num_sqsets + extra;
+
+ /* Max vectors required based on #niqs configured in fw */
+ if (hw->flags & CSIO_HWF_USING_SOFT_PARAMS || !csio_is_hw_master(hw))
+ cnt = min_t(uint8_t, hw->cfg_niq, cnt);
+
+ entries = kzalloc(sizeof(struct msix_entry) * cnt, GFP_KERNEL);
+ if (!entries)
+ return CSIO_NOMEM;
+
+ for (i = 0; i < cnt; i++)
+ entries[i].entry = (uint16_t)i;
+
+ csio_dbg(hw, "FW supp #niq:%d, trying %d msix's\n", hw->cfg_niq, cnt);
+
+ while ((rv = pci_enable_msix(hw->pdev, entries, cnt)) >= min)
+ cnt = rv;
+ if (!rv) {
+ if (cnt < (hw->num_sqsets + extra)) {
+ csio_dbg(hw, "Reducing sqsets to %d\n", cnt - extra);
+ csio_reduce_sqsets(hw, cnt - extra);
+ }
+ } else {
+ if (rv > 0) {
+ pci_disable_msix(hw->pdev);
+ csio_info(hw, "Not using MSI-X, remainder:%d\n", rv);
+ }
+
+ kfree(entries);
+ return CSIO_NOMEM;
+ }
+
+ /* Save off vectors */
+ for (i = 0; i < cnt; i++) {
+ entryp = &hw->msix_entries[i];
+ entryp->vector = entries[i].vector;
+ }
+
+ /* Distribute vectors */
+ k = 0;
+ csio_set_nondata_intr_idx(hw, entries[k].entry);
+ csio_set_mb_intr_idx(csio_hw_to_mbm(hw), entries[k++].entry);
+ csio_set_fwevt_intr_idx(hw, entries[k++].entry);
+
+ for (i = 0; i < hw->num_pports; i++) {
+ info = &hw->scsi_cpu_info[i];
+
+ for (j = 0; j < hw->num_scsi_msix_cpus; j++) {
+ n = (j % info->max_cpus) + k;
+ hw->sqset[i][j].intr_idx = entries[n].entry;
+ }
+
+ k += info->max_cpus;
+ }
+
+ kfree(entries);
+ return CSIO_SUCCESS;
+}
+
+void
+csio_intr_enable(struct csio_hw *hw)
+{
+ hw->intr_mode = CSIO_IM_NONE;
+ hw->flags &= ~CSIO_HWF_HOST_INTR_ENABLED;
+
+ /* Try MSIX, then MSI or fall back to INTx */
+ if ((csio_msi == 2) && !csio_enable_msix(hw))
+ hw->intr_mode = CSIO_IM_MSIX;
+ else {
+ /* Max iqs required based on #niqs configured in fw */
+ if (hw->flags & CSIO_HWF_USING_SOFT_PARAMS ||
+ !csio_is_hw_master(hw)) {
+ int extra = CSIO_EXTRA_MSI_IQS;
+
+ if (hw->cfg_niq < (hw->num_sqsets + extra)) {
+ csio_dbg(hw, "Reducing sqsets to %d\n",
+ hw->cfg_niq - extra);
+ csio_reduce_sqsets(hw, hw->cfg_niq - extra);
+ }
+ }
+
+ if ((csio_msi == 1) && !pci_enable_msi(hw->pdev))
+ hw->intr_mode = CSIO_IM_MSI;
+ else
+ hw->intr_mode = CSIO_IM_INTX;
+ }
+
+ csio_dbg(hw, "Using %s interrupt mode.\n",
+ (hw->intr_mode == CSIO_IM_MSIX) ? "MSIX" :
+ ((hw->intr_mode == CSIO_IM_MSI) ? "MSI" : "INTx"));
+}
+
+void
+csio_intr_disable(struct csio_hw *hw, bool free)
+{
+ csio_hw_intr_disable(hw);
+
+ switch (hw->intr_mode) {
+ case CSIO_IM_MSIX:
+ csio_disable_msix(hw, free);
+ break;
+ case CSIO_IM_MSI:
+ if (free)
+ free_irq(hw->pdev->irq, hw);
+ pci_disable_msi(hw->pdev);
+ break;
+ case CSIO_IM_INTX:
+ if (free)
+ free_irq(hw->pdev->irq, hw);
+ break;
+ default:
+ break;
+ }
+ hw->intr_mode = CSIO_IM_NONE;
+ hw->flags &= ~CSIO_HWF_HOST_INTR_ENABLED;
+}
diff --git a/drivers/scsi/csiostor/csio_scsi.c b/drivers/scsi/csiostor/csio_scsi.c
new file mode 100644
index 0000000..0f87b00
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_scsi.c
@@ -0,0 +1,2498 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/moduleparam.h>
+#include <linux/device.h>
+#include <linux/delay.h>
+#include <linux/ctype.h>
+#include <linux/version.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/compiler.h>
+#include <linux/export.h>
+#include <linux/module.h>
+#include <asm/unaligned.h>
+#include <asm/page.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_hw.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+#include "csio_scsi.h"
+#include "csio_init.h"
+
+int csio_scsi_eqsize = 65536;
+int csio_scsi_iqlen = 128;
+int csio_scsi_ioreqs = 2048;
+uint32_t csio_max_scan_tmo;
+uint32_t csio_delta_scan_tmo = 5;
+int csio_lun_qdepth = 32;
+
+static int csio_ddp_descs = 128;
+
+static csio_retval_t csio_do_abrt_cls(struct csio_hw *,
+ struct csio_ioreq *, bool);
+
+static void csio_scsis_uninit(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_io_active(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_tm_active(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_aborting(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_closing(struct csio_ioreq *, enum csio_scsi_ev);
+static void csio_scsis_shost_cmpl_await(struct csio_ioreq *, enum csio_scsi_ev);
+
+/*
+ * csio_scsi_match_io - Match an ioreq with the given SCSI level data.
+ * @ioreq: The I/O request
+ * @sld: Level information
+ *
+ * Should be called with lock held.
+ *
+ */
+static bool
+csio_scsi_match_io(struct csio_ioreq *ioreq, struct csio_scsi_level_data *sld)
+{
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd(ioreq);
+
+ switch (sld->level) {
+ case CSIO_LEV_LUN:
+ if (scmnd == NULL)
+ return CSIO_FALSE;
+
+ return ((ioreq->lnode == sld->lnode) &&
+ (ioreq->rnode == sld->rnode) &&
+ ((uint64_t)scmnd->device->lun == sld->oslun));
+
+ case CSIO_LEV_RNODE:
+ return ((ioreq->lnode == sld->lnode) &&
+ (ioreq->rnode == sld->rnode));
+ case CSIO_LEV_LNODE:
+ return (ioreq->lnode == sld->lnode);
+ case CSIO_LEV_ALL:
+ return CSIO_TRUE;
+ default:
+ return CSIO_FALSE;
+ }
+}
+
+/*
+ * csio_scsi_gather_active_ios - Gather active I/Os based on level
+ * @scm: SCSI module
+ * @sld: Level information
+ * @dest: The queue where these I/Os have to be gathered.
+ *
+ * Should be called with lock held.
+ */
+static void
+csio_scsi_gather_active_ios(struct csio_scsim *scm,
+ struct csio_scsi_level_data *sld,
+ struct list_head *dest)
+{
+ struct list_head *tmp, *next;
+
+ if (list_empty(&scm->active_q))
+ return;
+
+ /* Just splice the entire active_q into dest */
+ if (sld->level == CSIO_LEV_ALL) {
+ list_splice_tail_init(&scm->active_q, dest);
+ return;
+ }
+
+ list_for_each_safe(tmp, next, &scm->active_q) {
+ if (csio_scsi_match_io((struct csio_ioreq *)tmp, sld)) {
+ list_del_init(tmp);
+ list_add_tail(tmp, dest);
+ }
+ }
+}
+
+static inline bool
+csio_scsi_itnexus_loss_error(uint16_t error)
+{
+ switch (error) {
+ case FW_ERR_LINK_DOWN:
+ case FW_RDEV_NOT_READY:
+ case FW_ERR_RDEV_LOST:
+ case FW_ERR_RDEV_LOGO:
+ case FW_ERR_RDEV_IMPL_LOGO:
+ return 1;
+ }
+ return 0;
+}
+
+static inline void
+csio_scsi_tag(struct scsi_cmnd *scmnd, uint8_t *tag, uint8_t hq,
+ uint8_t oq, uint8_t sq)
+{
+ char stag[2];
+
+ if (scsi_populate_tag_msg(scmnd, stag)) {
+ switch (stag[0]) {
+ case HEAD_OF_QUEUE_TAG:
+ *tag = hq;
+ break;
+ case ORDERED_QUEUE_TAG:
+ *tag = oq;
+ break;
+ default:
+ *tag = sq;
+ break;
+ }
+ } else
+ *tag = 0;
+}
+
+/*
+ * csio_scsi_fcp_cmnd - Frame the SCSI FCP command paylod.
+ * @req: IO req structure.
+ * @addr: DMA location to place the payload.
+ *
+ * This routine is shared between FCP_WRITE, FCP_READ and FCP_CMD requests.
+ */
+static inline void
+csio_scsi_fcp_cmnd(struct csio_ioreq *req, void *addr)
+{
+ struct csio_fcp_cmnd *fcp_cmnd = (struct csio_fcp_cmnd *)addr;
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd(req);
+
+ /* Check for Task Management */
+ if (likely(scmnd->SCp.Message == 0)) {
+ int_to_scsilun(scmnd->device->lun,
+ (struct scsi_lun *)fcp_cmnd->lun);
+ fcp_cmnd->tm_flags = 0;
+ fcp_cmnd->cmdref = 0;
+ fcp_cmnd->pri_ta = 0;
+
+ memcpy(fcp_cmnd->cdb, scmnd->cmnd, 16);
+ csio_scsi_tag(scmnd, &fcp_cmnd->pri_ta,
+ FCP_PTA_HEADQ, FCP_PTA_ORDERED, FCP_PTA_SIMPLE);
+ fcp_cmnd->dl = cpu_to_be32(scsi_bufflen(scmnd));
+
+ if (req->nsge)
+ if (req->datadir == CSIO_IOREQF_DMA_WRITE)
+ fcp_cmnd->flags = FCP_CFL_WRDATA;
+ else
+ fcp_cmnd->flags = FCP_CFL_RDDATA;
+ else
+ fcp_cmnd->flags = 0;
+ } else {
+ memset(fcp_cmnd, 0, sizeof(*fcp_cmnd));
+ int_to_scsilun(scmnd->device->lun,
+ (struct scsi_lun *)fcp_cmnd->lun);
+ fcp_cmnd->tm_flags = (uint8_t)scmnd->SCp.Message;
+ }
+}
+
+/*
+ * csio_scsi_init_cmd_wr - Initialize the SCSI CMD WR.
+ * @req: IO req structure.
+ * @addr: DMA location to place the payload.
+ * @size: Size of WR (including FW WR + immed data + rsp SG entry
+ *
+ * Wrapper for populating fw_scsi_cmd_wr.
+ */
+static inline void
+csio_scsi_init_cmd_wr(struct csio_ioreq *req, void *addr, uint32_t size)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_rnode *rn = req->rnode;
+ struct fw_scsi_cmd_wr *wr = (struct fw_scsi_cmd_wr *)addr;
+ struct csio_dma_buf *dma_buf;
+ uint8_t imm = csio_hw_to_scsim(hw)->proto_cmd_len;
+
+ wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI_CMD_WR) |
+ FW_SCSI_CMD_WR_IMMDLEN(imm));
+ wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(rn->flowid) |
+ FW_WR_LEN16(
+ CSIO_ROUNDUP(size, 16)));
+
+ wr->cookie = (uintptr_t) req;
+ wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(hw, req->iq_idx));
+ wr->tmo_val = (uint8_t) req->tmo;
+ wr->r3 = 0;
+ memset(&wr->r5, 0, 8);
+
+ /* Get RSP DMA buffer */
+ dma_buf = &req->dma_buf;
+
+ /* Prepare RSP SGL */
+ wr->rsp_dmalen = cpu_to_be32(dma_buf->len);
+ wr->rsp_dmaaddr = cpu_to_be64(dma_buf->paddr);
+
+ wr->r6 = 0;
+
+ wr->u.fcoe.ctl_pri = 0;
+ wr->u.fcoe.cp_en_class = 0;
+ wr->u.fcoe.r4_lo[0] = 0;
+ wr->u.fcoe.r4_lo[1] = 0;
+
+ /* Frame a FCP command */
+ csio_scsi_fcp_cmnd(req, (void *)((uintptr_t)addr +
+ sizeof(struct fw_scsi_cmd_wr)));
+}
+
+#define CSIO_SCSI_CMD_WR_SZ(_imm) \
+ (sizeof(struct fw_scsi_cmd_wr) + /* WR size */ \
+ ALIGN((_imm), 16)) /* Immed data */
+
+#define CSIO_SCSI_CMD_WR_SZ_16(_imm) \
+ (ALIGN(CSIO_SCSI_CMD_WR_SZ((_imm)), 16))
+
+/*
+ * csio_scsi_cmd - Create a SCSI CMD WR.
+ * @req: IO req structure.
+ *
+ * Gets a WR slot in the ingress queue and initializes it with SCSI CMD WR.
+ *
+ */
+static inline void
+csio_scsi_cmd(struct csio_ioreq *req)
+{
+ struct csio_wr_pair wrp;
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ uint32_t size = CSIO_SCSI_CMD_WR_SZ_16(scsim->proto_cmd_len);
+
+ req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
+ if (unlikely(req->drv_status != CSIO_SUCCESS))
+ return;
+
+ if (wrp.size1 >= size) {
+ /* Initialize WR in one shot */
+ csio_scsi_init_cmd_wr(req, wrp.addr1, size);
+ } else {
+ uint8_t tmpwr[512];
+ /*
+ * Make a temporary copy of the WR and write back
+ * the copy into the WR pair.
+ */
+ csio_scsi_init_cmd_wr(req, (void *)tmpwr, size);
+ memcpy(wrp.addr1, tmpwr, wrp.size1);
+ memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
+ }
+}
+
+/*
+ * The following is fast path code. Therefore it is inlined with multi-line
+ * macros using name substitution, thus avoiding if-else switches for
+ * operation (read/write), as well as serving the purpose of code re-use.
+ */
+/*
+ * csio_scsi_init_ulptx_dsgl - Fill in a ULP_TX_SC_DSGL
+ * @hw: HW module
+ * @req: IO request
+ * @sgl: ULP TX SGL pointer.
+ *
+ */
+#define csio_scsi_init_ultptx_dsgl(hw, req, sgl) \
+do { \
+ struct ulptx_sge_pair *_sge_pair = NULL; \
+ struct scatterlist *_sgel; \
+ uint32_t _i = 0; \
+ uint32_t _xfer_len; \
+ struct list_head *_tmp; \
+ struct csio_dma_buf *_dma_buf; \
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
+ \
+ (sgl)->cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) | ULPTX_MORE | \
+ ULPTX_NSGE((req)->nsge)); \
+ /* Now add the data SGLs */ \
+ if (likely(!(req)->dcopy)) { \
+ scsi_for_each_sg(scmnd, _sgel, (req)->nsge, _i) { \
+ if (_i == 0) { \
+ (sgl)->addr0 = cpu_to_be64( \
+ sg_dma_address(_sgel)); \
+ (sgl)->len0 = cpu_to_be32( \
+ sg_dma_len(_sgel)); \
+ _sge_pair = \
+ (struct ulptx_sge_pair *)((sgl) + 1); \
+ continue; \
+ } \
+ if ((_i - 1) & 0x1) { \
+ _sge_pair->addr[1] = cpu_to_be64( \
+ sg_dma_address(_sgel)); \
+ _sge_pair->len[1] = cpu_to_be32( \
+ sg_dma_len(_sgel)); \
+ _sge_pair++; \
+ } else { \
+ _sge_pair->addr[0] = cpu_to_be64( \
+ sg_dma_address(_sgel)); \
+ _sge_pair->len[0] = cpu_to_be32( \
+ sg_dma_len(_sgel)); \
+ } \
+ } \
+ } else { \
+ /* Program sg elements with driver's DDP buffer */ \
+ _xfer_len = scsi_bufflen(scmnd); \
+ list_for_each(_tmp, &(req)->gen_list) { \
+ _dma_buf = (struct csio_dma_buf *)_tmp; \
+ if (_i == 0) { \
+ (sgl)->addr0 = cpu_to_be64(_dma_buf->paddr); \
+ (sgl)->len0 = cpu_to_be32( \
+ min(_xfer_len, _dma_buf->len)); \
+ _sge_pair = \
+ (struct ulptx_sge_pair *)((sgl) + 1); \
+ } \
+ else if ((_i - 1) & 0x1) { \
+ _sge_pair->addr[1] = cpu_to_be64( \
+ _dma_buf->paddr); \
+ _sge_pair->len[1] = cpu_to_be32( \
+ min(_xfer_len, _dma_buf->len)); \
+ _sge_pair++; \
+ } else { \
+ _sge_pair->addr[0] = cpu_to_be64( \
+ _dma_buf->paddr); \
+ _sge_pair->len[0] = cpu_to_be32( \
+ min(_xfer_len, _dma_buf->len)); \
+ } \
+ _xfer_len -= min(_xfer_len, _dma_buf->len); \
+ _i++; \
+ } \
+ } \
+} while (0)
+
+/*
+ * csio_scsi_init_data_wr - Initialize the READ/WRITE SCSI WR.
+ * @req: IO req structure.
+ * @oper: read/write
+ * @wrp: DMA location to place the payload.
+ * @size: Size of WR (including FW WR + immed data + rsp SG entry + data SGL
+ * @wrop: _READ_/_WRITE_
+ *
+ * Wrapper for populating fw_scsi_read_wr/fw_scsi_write_wr.
+ */
+#define csio_scsi_init_data_wr(req, oper, wrp, size, wrop) \
+do { \
+ struct csio_hw *_hw = (req)->lnode->hwp; \
+ struct csio_rnode *_rn = (req)->rnode; \
+ struct fw_scsi_##oper##_wr *__wr = (struct fw_scsi_##oper##_wr *)(wrp);\
+ struct ulptx_sgl *_sgl; \
+ struct csio_dma_buf *_dma_buf; \
+ uint8_t _imm = csio_hw_to_scsim(_hw)->proto_cmd_len; \
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
+ \
+ __wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI##wrop##WR) | \
+ FW_SCSI##wrop##WR_IMMDLEN(_imm)); \
+ __wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(_rn->flowid) | \
+ FW_WR_LEN16( \
+ CSIO_ROUNDUP((size), 16))); \
+ __wr->cookie = (uintptr_t) (req); \
+ __wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(_hw, \
+ (req)->iq_idx));\
+ __wr->tmo_val = (uint8_t)((req)->tmo); \
+ __wr->use_xfer_cnt = 1; \
+ __wr->xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
+ __wr->ini_xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
+ /* Get RSP DMA buffer */ \
+ _dma_buf = &(req)->dma_buf; \
+ \
+ /* Prepare RSP SGL */ \
+ __wr->rsp_dmalen = cpu_to_be32(_dma_buf->len); \
+ __wr->rsp_dmaaddr = cpu_to_be64(_dma_buf->paddr); \
+ \
+ __wr->r4 = 0; \
+ \
+ __wr->u.fcoe.ctl_pri = 0; \
+ __wr->u.fcoe.cp_en_class = 0; \
+ __wr->u.fcoe.r3_lo[0] = 0; \
+ __wr->u.fcoe.r3_lo[1] = 0; \
+ csio_scsi_fcp_cmnd((req), (void *)((uintptr_t)(wrp) + \
+ sizeof(struct fw_scsi_##oper##_wr))); \
+ \
+ /* Move WR pointer past command and immediate data */ \
+ _sgl = (struct ulptx_sgl *) ((uintptr_t)(wrp) + \
+ sizeof(struct fw_scsi_##oper##_wr) + \
+ ALIGN(_imm, 16)); \
+ \
+ /* Fill in the DSGL */ \
+ csio_scsi_init_ultptx_dsgl(_hw, (req), _sgl); \
+ \
+} while (0)
+
+/* Calculate WR size needed for fw_scsi_read_wr/fw_scsi_write_wr */
+#define csio_scsi_data_wrsz(req, oper, sz, imm) \
+do { \
+ (sz) = sizeof(struct fw_scsi_##oper##_wr) + /* WR size */ \
+ ALIGN((imm), 16) + /* Immed data */ \
+ sizeof(struct ulptx_sgl); /* ulptx_sgl */ \
+ \
+ if (unlikely((req)->nsge > 1)) \
+ (sz) += (sizeof(struct ulptx_sge_pair) * \
+ (ALIGN(((req)->nsge - 1), 2) / 2)); \
+ /* Data SGE */ \
+} while (0)
+
+/*
+ * csio_scsi_data - Create a SCSI WRITE/READ WR.
+ * @req: IO req structure.
+ * @oper: read/write
+ * @wrop: _READ_/_WRITE_ (string subsitutions to use with the FW bit field
+ * macros).
+ *
+ * Gets a WR slot in the ingress queue and initializes it with
+ * SCSI CMD READ/WRITE WR.
+ *
+ */
+#define csio_scsi_data(req, oper, wrop) \
+do { \
+ struct csio_wr_pair _wrp; \
+ uint32_t _size; \
+ struct csio_hw *_hw = (req)->lnode->hwp; \
+ struct csio_scsim *_scsim = csio_hw_to_scsim(_hw); \
+ \
+ csio_scsi_data_wrsz((req), oper, _size, _scsim->proto_cmd_len); \
+ _size = ALIGN(_size, 16); \
+ \
+ (req)->drv_status = csio_wr_get(_hw, (req)->eq_idx, _size, &_wrp); \
+ if (likely((req)->drv_status == CSIO_SUCCESS)) { \
+ if (likely(_wrp.size1 >= _size)) { \
+ /* Initialize WR in one shot */ \
+ csio_scsi_init_data_wr((req), oper, _wrp.addr1, \
+ _size, wrop); \
+ } else { \
+ uint8_t tmpwr[512]; \
+ /* \
+ * Make a temporary copy of the WR and write back \
+ * the copy into the WR pair. \
+ */ \
+ csio_scsi_init_data_wr((req), oper, (void *)tmpwr, \
+ _size, wrop); \
+ memcpy(_wrp.addr1, tmpwr, _wrp.size1); \
+ memcpy(_wrp.addr2, tmpwr + _wrp.size1, \
+ _size - _wrp.size1); \
+ } \
+ } \
+} while (0)
+
+/*
+ * csio_setup_ddp - Setup DDP buffers for Read request.
+ * @req: IO req structure.
+ *
+ * Checks SGLs/Data buffers are virtually contiguous required for DDP.
+ * If contiguous,driver posts SGLs in the WR otherwise post internal
+ * buffers for such request for DDP.
+ */
+static inline void
+csio_setup_ddp(struct csio_scsim *scsim, struct csio_ioreq *req)
+{
+#ifdef __CSIO_DEBUG__
+ struct csio_hw *hw = req->lnode->hwp;
+#endif
+ struct scatterlist *sgel = NULL;
+ struct scsi_cmnd *scmnd = csio_scsi_cmnd(req);
+ uint64_t sg_addr = 0;
+ uint32_t ddp_pagesz = 4096;
+ uint32_t buf_off;
+ struct csio_dma_buf *dma_buf = NULL;
+ uint32_t alloc_len = 0;
+ uint32_t xfer_len = 0;
+ uint32_t sg_len = 0;
+ uint32_t i;
+
+ scsi_for_each_sg(scmnd, sgel, req->nsge, i) {
+ sg_addr = sg_dma_address(sgel);
+ sg_len = sg_dma_len(sgel);
+
+ buf_off = sg_addr & (ddp_pagesz - 1);
+
+ /* Except 1st buffer,all buffer addr have to be Page aligned */
+ if (i != 0 && buf_off) {
+ csio_dbg(hw, "SGL addr not DDP aligned (%llx:%d)\n",
+ sg_addr, sg_len);
+ goto unaligned;
+ }
+
+ /* Except last buffer,all buffer must end on page boundary */
+ if ((i != (req->nsge - 1)) &&
+ ((buf_off + sg_len) & (ddp_pagesz - 1))) {
+ csio_dbg(hw,
+ "SGL addr not ending on page boundary"
+ "(%llx:%d)\n", sg_addr, sg_len);
+ goto unaligned;
+ }
+ }
+
+ /* SGL's are virtually contiguous. HW will DDP to SGLs */
+ req->dcopy = 0;
+ csio_scsi_data(req, read, _READ_);
+
+ return;
+
+unaligned:
+ csio_inc_stats(scsim, n_unaligned);
+ /*
+ * For unaligned SGLs, driver will allocate internal DDP buffer.
+ * Once command is completed data from DDP buffer copied to SGLs
+ */
+ req->dcopy = 1;
+
+ /* Use gen_list to store the DDP buffers */
+ INIT_LIST_HEAD(&req->gen_list);
+ xfer_len = scsi_bufflen(scmnd);
+
+ i = 0;
+ /* Allocate ddp buffers for this request */
+ while (alloc_len < xfer_len) {
+ dma_buf = csio_get_scsi_ddp(scsim);
+ if (dma_buf == NULL || i > scsim->max_sge) {
+ req->drv_status = CSIO_BUSY;
+ break;
+ }
+ alloc_len += dma_buf->len;
+ /* Added to IO req */
+ list_add_tail(&dma_buf->list, &req->gen_list);
+ i++;
+ }
+
+ if (!req->drv_status) {
+ /* set number of ddp bufs used */
+ req->nsge = i;
+ csio_scsi_data(req, read, _READ_);
+ return;
+ }
+
+ /* release dma descs */
+ if (i > 0)
+ csio_put_scsi_ddp_list(scsim, &req->gen_list, i);
+}
+
+/*
+ * csio_scsi_init_abrt_cls_wr - Initialize an ABORT/CLOSE WR.
+ * @req: IO req structure.
+ * @addr: DMA location to place the payload.
+ * @size: Size of WR
+ * @abort: abort OR close
+ *
+ * Wrapper for populating fw_scsi_cmd_wr.
+ */
+static inline void
+csio_scsi_init_abrt_cls_wr(struct csio_ioreq *req, void *addr, uint32_t size,
+ bool abort)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_rnode *rn = req->rnode;
+ struct fw_scsi_abrt_cls_wr *wr = (struct fw_scsi_abrt_cls_wr *)addr;
+
+ wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI_ABRT_CLS_WR));
+ wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(rn->flowid) |
+ FW_WR_LEN16(
+ CSIO_ROUNDUP(size, 16)));
+
+ wr->cookie = (uintptr_t) req;
+ wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(hw, req->iq_idx));
+ wr->tmo_val = (uint8_t) req->tmo;
+ /* 0 for CHK_ALL_IO tells FW to look up t_cookie */
+ wr->sub_opcode_to_chk_all_io =
+ (FW_SCSI_ABRT_CLS_WR_SUB_OPCODE(abort) |
+ FW_SCSI_ABRT_CLS_WR_CHK_ALL_IO(0));
+ wr->r3[0] = 0;
+ wr->r3[1] = 0;
+ wr->r3[2] = 0;
+ wr->r3[3] = 0;
+ /* Since we re-use the same ioreq for abort as well */
+ wr->t_cookie = (uintptr_t) req;
+}
+
+static inline void
+csio_scsi_abrt_cls(struct csio_ioreq *req, bool abort)
+{
+ struct csio_wr_pair wrp;
+ struct csio_hw *hw = req->lnode->hwp;
+ uint32_t size = ALIGN(sizeof(struct fw_scsi_abrt_cls_wr), 16);
+
+ req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
+ if (req->drv_status != CSIO_SUCCESS)
+ return;
+
+ if (wrp.size1 >= size) {
+ /* Initialize WR in one shot */
+ csio_scsi_init_abrt_cls_wr(req, wrp.addr1, size, abort);
+ } else {
+ uint8_t tmpwr[512];
+ /*
+ * Make a temporary copy of the WR and write back
+ * the copy into the WR pair.
+ */
+ csio_scsi_init_abrt_cls_wr(req, (void *)tmpwr, size, abort);
+ memcpy(wrp.addr1, tmpwr, wrp.size1);
+ memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
+ }
+}
+
+/*****************************************************************************/
+/* START: SCSI SM */
+/*****************************************************************************/
+static void
+csio_scsis_uninit(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+
+ switch (evt) {
+
+ case CSIO_SCSIE_START_IO:
+
+ /* There is data */
+ if (req->nsge) {
+ if (req->datadir == CSIO_IOREQF_DMA_WRITE) {
+ req->dcopy = 0;
+ csio_scsi_data(req, write, _WRITE_);
+ } else
+ csio_setup_ddp(scsim, req);
+ } else {
+ csio_scsi_cmd(req);
+ }
+
+ if (likely(req->drv_status == CSIO_SUCCESS)) {
+ /* change state and enqueue on active_q */
+ csio_set_state(&req->sm, csio_scsis_io_active);
+ list_add_tail(&req->sm.sm_list, &scsim->active_q);
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_inc_stats(scsim, n_active);
+
+ return;
+ }
+ break;
+
+ case CSIO_SCSIE_START_TM:
+ csio_scsi_cmd(req);
+ if (req->drv_status == CSIO_SUCCESS) {
+ /*
+ * NOTE: We collect the affected I/Os prior to issuing
+ * LUN reset, and not after it. This is to prevent
+ * aborting I/Os that get issued after the LUN reset,
+ * but prior to LUN reset completion (in the event that
+ * the host stack has not blocked I/Os to a LUN that is
+ * being reset.
+ */
+ csio_set_state(&req->sm, csio_scsis_tm_active);
+ list_add_tail(&req->sm.sm_list, &scsim->active_q);
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_inc_stats(scsim, n_tm_active);
+ }
+ return;
+
+ case CSIO_SCSIE_ABORT:
+ case CSIO_SCSIE_CLOSE:
+ /*
+ * NOTE:
+ * We could get here due to :
+ * - a window in the cleanup path of the SCSI module
+ * (csio_scsi_abort_io()). Please see NOTE in this function.
+ * - a window in the time we tried to issue an abort/close
+ * of a request to FW, and the FW completed the request
+ * itself.
+ * Print a message for now, and return INVAL either way.
+ */
+ req->drv_status = CSIO_INVAL;
+ csio_warn(hw, "Trying to abort/close completed IO:%p!\n", req);
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_io_active(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+ struct csio_rnode *rn;
+
+ switch (evt) {
+
+ case CSIO_SCSIE_COMPLETED:
+ csio_dec_stats(scm, n_active);
+ list_del_init(&req->sm.sm_list);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ /*
+ * In MSIX mode, with multiple queues, the SCSI compeltions
+ * could reach us sooner than the FW events sent to indicate
+ * I-T nexus loss (link down, remote device logo etc). We
+ * dont want to be returning such I/Os to the upper layer
+ * immediately, since we wouldnt have reported the I-T nexus
+ * loss itself. This forces us to serialize such completions
+ * with the reporting of the I-T nexus loss. Therefore, we
+ * internally queue up such up such completions in the rnode.
+ * The reporting of I-T nexus loss to the upper layer is then
+ * followed by the returning of I/Os in this internal queue.
+ * Having another state alongwith another queue helps us take
+ * actions for events such as ABORT received while we are
+ * in this rnode queue.
+ */
+ if (unlikely(req->wr_status != FW_SUCCESS)) {
+ rn = req->rnode;
+ /*
+ * FW says remote device is lost, but rnode
+ * doesnt reflect it.
+ */
+ if (csio_scsi_itnexus_loss_error(req->wr_status) &&
+ csio_is_rnode_ready(rn)) {
+ csio_set_state(&req->sm,
+ csio_scsis_shost_cmpl_await);
+ list_add_tail(&req->sm.sm_list,
+ &rn->host_cmpl_q);
+ }
+ }
+
+ break;
+
+ case CSIO_SCSIE_ABORT:
+ csio_scsi_abrt_cls(req, SCSI_ABORT);
+ if (req->drv_status == CSIO_SUCCESS) {
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_set_state(&req->sm, csio_scsis_aborting);
+ }
+ break;
+
+ case CSIO_SCSIE_CLOSE:
+ csio_scsi_abrt_cls(req, SCSI_CLOSE);
+ if (req->drv_status == CSIO_SUCCESS) {
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_set_state(&req->sm, csio_scsis_closing);
+ }
+ break;
+
+ case CSIO_SCSIE_DRVCLEANUP:
+ req->wr_status = FW_HOSTERROR;
+ csio_dec_stats(scm, n_active);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_tm_active(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+
+ switch (evt) {
+
+ case CSIO_SCSIE_COMPLETED:
+ csio_dec_stats(scm, n_tm_active);
+ list_del_init(&req->sm.sm_list);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+
+ break;
+
+ case CSIO_SCSIE_ABORT:
+ csio_scsi_abrt_cls(req, SCSI_ABORT);
+ if (req->drv_status == CSIO_SUCCESS) {
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_set_state(&req->sm, csio_scsis_aborting);
+ }
+ break;
+
+
+ case CSIO_SCSIE_CLOSE:
+ csio_scsi_abrt_cls(req, SCSI_CLOSE);
+ if (req->drv_status == CSIO_SUCCESS) {
+ csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
+ csio_set_state(&req->sm, csio_scsis_closing);
+ }
+ break;
+
+ case CSIO_SCSIE_DRVCLEANUP:
+ req->wr_status = FW_HOSTERROR;
+ csio_dec_stats(scm, n_tm_active);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_aborting(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+
+ switch (evt) {
+
+ case CSIO_SCSIE_COMPLETED:
+ csio_dbg(hw,
+ "ioreq %p recvd cmpltd (wr_status:%d) "
+ "in aborting st\n", req, req->wr_status);
+ /*
+ * Use CSIO_CANCELLED to explicitly tell the ABORTED event that
+ * the original I/O was returned to driver by FW.
+ * We dont really care if the I/O was returned with success by
+ * FW (because the ABORT and completion of the I/O crossed each
+ * other), or any other return value. Once we are in aborting
+ * state, the success or failure of the I/O is unimportant to
+ * us.
+ */
+ req->drv_status = CSIO_CANCELLED;
+ break;
+
+ case CSIO_SCSIE_ABORT:
+ csio_inc_stats(scm, n_abrt_dups);
+ break;
+
+ case CSIO_SCSIE_ABORTED:
+
+ csio_dbg(hw, "abort of %p return status:0x%x drv_status:%x\n",
+ req, req->wr_status, req->drv_status);
+ /*
+ * Check if original I/O WR completed before the Abort
+ * completion.
+ */
+ if (req->drv_status != CSIO_CANCELLED) {
+ csio_fatal(hw,
+ "Abort completed before original I/O,"
+ " req:%p\n", req);
+ CSIO_DB_ASSERT(0);
+ }
+
+ /*
+ * There are the following possible scenarios:
+ * 1. The abort completed successfully, FW returned FW_SUCCESS.
+ * 2. The completion of an I/O and the receipt of
+ * abort for that I/O by the FW crossed each other.
+ * The FW returned FW_EINVAL. The original I/O would have
+ * returned with FW_SUCCESS or any other SCSI error.
+ * 3. The FW couldnt sent the abort out on the wire, as there
+ * was an I-T nexus loss (link down, remote device logged
+ * out etc). FW sent back an appropriate IT nexus loss status
+ * for the abort.
+ * 4. FW sent an abort, but abort timed out (remote device
+ * didnt respond). FW replied back with
+ * FW_SCSI_ABORT_TIMEDOUT.
+ * 5. FW couldnt genuinely abort the request for some reason,
+ * and sent us an error.
+ *
+ * The first 3 scenarios are treated as succesful abort
+ * operations by the host, while the last 2 are failed attempts
+ * to abort. Manipulate the return value of the request
+ * appropriately, so that host can convey these results
+ * back to the upper layer.
+ */
+ if ((req->wr_status == FW_SUCCESS) ||
+ (req->wr_status == FW_EINVAL) ||
+ csio_scsi_itnexus_loss_error(req->wr_status))
+ req->wr_status = FW_SCSI_ABORT_REQUESTED;
+
+ csio_dec_stats(scm, n_active);
+ list_del_init(&req->sm.sm_list);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ case CSIO_SCSIE_DRVCLEANUP:
+ req->wr_status = FW_HOSTERROR;
+ csio_dec_stats(scm, n_active);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ case CSIO_SCSIE_CLOSE:
+ /*
+ * We can receive this event from the module
+ * cleanup paths, if the FW forgot to reply to the ABORT WR
+ * and left this ioreq in this state. For now, just ignore
+ * the event. The CLOSE event is sent to this state, as
+ * the LINK may have already gone down.
+ */
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_closing(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ struct csio_hw *hw = req->lnode->hwp;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+
+ switch (evt) {
+
+ case CSIO_SCSIE_COMPLETED:
+ csio_dbg(hw,
+ "ioreq %p recvd cmpltd (wr_status:%d) "
+ "in closing st\n", req, req->wr_status);
+ /*
+ * Use CSIO_CANCELLED to explicitly tell the CLOSED event that
+ * the original I/O was returned to driver by FW.
+ * We dont really care if the I/O was returned with success by
+ * FW (because the CLOSE and completion of the I/O crossed each
+ * other), or any other return value. Once we are in aborting
+ * state, the success or failure of the I/O is unimportant to
+ * us.
+ */
+ req->drv_status = CSIO_CANCELLED;
+ break;
+
+ case CSIO_SCSIE_CLOSED:
+ /*
+ * Check if original I/O WR completed before the Close
+ * completion.
+ */
+ if (req->drv_status != CSIO_CANCELLED) {
+ csio_fatal(hw,
+ "Close completed before original I/O,"
+ " req:%p\n", req);
+ CSIO_DB_ASSERT(0);
+ }
+
+ /*
+ * Either close succeeded, or we issued close to FW at the
+ * same time FW compelted it to us. Either way, the I/O
+ * is closed.
+ */
+ CSIO_DB_ASSERT((req->wr_status == FW_SUCCESS) ||
+ (req->wr_status == FW_EINVAL));
+ req->wr_status = FW_SCSI_CLOSE_REQUESTED;
+
+ csio_dec_stats(scm, n_active);
+ list_del_init(&req->sm.sm_list);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ case CSIO_SCSIE_CLOSE:
+ break;
+
+ case CSIO_SCSIE_DRVCLEANUP:
+ req->wr_status = FW_HOSTERROR;
+ csio_dec_stats(scm, n_active);
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+
+ default:
+ csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+static void
+csio_scsis_shost_cmpl_await(struct csio_ioreq *req, enum csio_scsi_ev evt)
+{
+ switch (evt) {
+ case CSIO_SCSIE_ABORT:
+ case CSIO_SCSIE_CLOSE:
+ /*
+ * Just succeed the abort request, and hope that
+ * the remote device unregister path will cleanup
+ * this I/O to the upper layer within a sane
+ * amount of time.
+ */
+ /*
+ * A close can come in during a LINK DOWN. The FW would have
+ * returned us the I/O back, but not the remote device lost
+ * FW event. In this interval, if the I/O times out at the upper
+ * layer, a close can come in. Take the same action as abort:
+ * return success, and hope that the remote device unregister
+ * path will cleanup this I/O. If the FW still doesnt send
+ * the msg, the close times out, and the upper layer resorts
+ * to the next level of error recovery.
+ */
+ req->drv_status = CSIO_SUCCESS;
+ break;
+ case CSIO_SCSIE_DRVCLEANUP:
+ csio_set_state(&req->sm, csio_scsis_uninit);
+ break;
+ default:
+ csio_dbg(req->lnode->hwp, "Unhandled event:%d sent to req:%p\n",
+ evt, req);
+ CSIO_DB_ASSERT(0);
+ }
+}
+
+/*
+ * csio_scsi_cmpl_handler - WR completion handler for SCSI.
+ * @hw: HW module.
+ * @wr: The completed WR from the ingress queue.
+ * @len: Length of the WR.
+ * @flb: Freelist buffer array.
+ * @priv: Private object
+ * @scsiwr: Pointer to SCSI WR.
+ *
+ * This is the WR completion handler called per completion from the
+ * ISR. It is called with lock held. It walks past the RSS and CPL message
+ * header where the actual WR is present.
+ * It then gets the status, WR handle (ioreq pointer) and the len of
+ * the WR, based on WR opcode. Only on a non-good status is the entire
+ * WR copied into the WR cache (ioreq->fw_wr).
+ * The ioreq corresponding to the WR is returned to the caller.
+ * NOTE: The SCSI queue doesnt allocate a freelist today, hence
+ * no freelist buffer is expected.
+ */
+struct csio_ioreq *
+csio_scsi_cmpl_handler(struct csio_hw *hw, void *wr, uint32_t len,
+ struct csio_fl_dma_buf *flb, void *priv, uint8_t **scsiwr)
+{
+ struct csio_ioreq *ioreq = NULL;
+ struct cpl_fw6_msg *cpl;
+ uint8_t *tempwr;
+ uint8_t status;
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+
+ /* skip RSS header */
+ cpl = (struct cpl_fw6_msg *)((uintptr_t)wr + sizeof(__be64));
+
+ if (unlikely(cpl->opcode != CPL_FW6_MSG)) {
+ csio_warn(hw, "Error: Invalid CPL msg %x recvd on SCSI q\n",
+ cpl->opcode);
+ csio_inc_stats(scm, n_inval_cplop);
+ return NULL;
+ }
+
+ tempwr = (uint8_t *)(cpl->data);
+ status = csio_wr_status(tempwr);
+ *scsiwr = tempwr;
+
+ if (likely((*tempwr == FW_SCSI_READ_WR) ||
+ (*tempwr == FW_SCSI_WRITE_WR) ||
+ (*tempwr == FW_SCSI_CMD_WR))) {
+ ioreq = (struct csio_ioreq *)((uintptr_t)
+ (((struct fw_scsi_read_wr *)tempwr)->cookie));
+ CSIO_DB_ASSERT(virt_addr_valid(ioreq));
+
+ ioreq->wr_status = status;
+
+ return ioreq;
+ }
+
+ if (*tempwr == FW_SCSI_ABRT_CLS_WR) {
+ ioreq = (struct csio_ioreq *)((uintptr_t)
+ (((struct fw_scsi_abrt_cls_wr *)tempwr)->cookie));
+ CSIO_DB_ASSERT(virt_addr_valid(ioreq));
+
+ ioreq->wr_status = status;
+ return ioreq;
+ }
+
+ csio_warn(hw, "WR with invalid opcode in SCSI IQ: %x\n", *tempwr);
+ csio_inc_stats(scm, n_inval_scsiop);
+ return NULL;
+}
+
+/*
+ * csio_scsi_cleanup_io_q - Cleanup the given queue.
+ * @scm: SCSI module.
+ * @q: Queue to be cleaned up.
+ *
+ * Called with lock held. Has to exit with lock held.
+ */
+void
+csio_scsi_cleanup_io_q(struct csio_scsim *scm, struct list_head *q)
+{
+ struct csio_hw *hw = scm->hw;
+ struct csio_ioreq *ioreq;
+ struct list_head *tmp, *next;
+ struct scsi_cmnd *scmnd;
+
+ /* Call back the completion routines of the active_q */
+ list_for_each_safe(tmp, next, q) {
+ ioreq = (struct csio_ioreq *)tmp;
+ csio_scsi_drvcleanup(ioreq);
+ list_del_init(&ioreq->sm.sm_list);
+ scmnd = csio_scsi_cmnd(ioreq);
+ spin_unlock_irq(&hw->lock);
+
+ /*
+ * Upper layers may have cleared this command, hence this
+ * check to avoid accessing stale references.
+ */
+ if (scmnd != NULL)
+ ioreq->io_cbfn(hw, ioreq);
+
+ spin_lock_irq(&scm->freelist_lock);
+ csio_put_scsi_ioreq(scm, ioreq);
+ spin_unlock_irq(&scm->freelist_lock);
+
+ spin_lock_irq(&hw->lock);
+ }
+}
+
+#define CSIO_SCSI_ABORT_Q_POLL_MS 2000
+
+static void
+csio_abrt_cls(struct csio_ioreq *ioreq, struct scsi_cmnd *scmnd)
+{
+ struct csio_lnode *ln = ioreq->lnode;
+ struct csio_hw *hw = ln->hwp;
+ int ready = 0;
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ csio_retval_t rv;
+
+ if (csio_scsi_cmnd(ioreq) != scmnd) {
+ csio_inc_stats(scsim, n_abrt_race_comp);
+ return;
+ }
+
+ ready = csio_is_lnode_ready(ln);
+
+ rv = csio_do_abrt_cls(hw, ioreq, (ready ? SCSI_ABORT : SCSI_CLOSE));
+ if (rv != CSIO_SUCCESS) {
+ if (ready)
+ csio_inc_stats(scsim, n_abrt_busy_error);
+ else
+ csio_inc_stats(scsim, n_cls_busy_error);
+ }
+}
+
+/*
+ * csio_scsi_abort_io_q - Abort all I/Os on given queue
+ * @scm: SCSI module.
+ * @q: Queue to abort.
+ * @tmo: Timeout in ms
+ *
+ * Attempt to abort all I/Os on given queue, and wait for a max
+ * of tmo milliseconds for them to complete. Returns success
+ * if all I/Os are aborted. Else returns CSIO_TIMEOUT.
+ * Should be entered with lock held. Exits with lock held.
+ * NOTE:
+ * Lock has to be held across the loop that aborts I/Os, since dropping the lock
+ * in between can cause the list to be corrupted. As a result, the caller
+ * of this function has to ensure that the number of I/os to be aborted
+ * is finite enough to not cause lock-held-for-too-long issues.
+ */
+static csio_retval_t
+csio_scsi_abort_io_q(struct csio_scsim *scm, struct list_head *q, uint32_t tmo)
+{
+ struct csio_hw *hw = scm->hw;
+ struct list_head *tmp, *next;
+ int count = CSIO_ROUNDUP(tmo, CSIO_SCSI_ABORT_Q_POLL_MS);
+ struct scsi_cmnd *scmnd;
+
+ if (list_empty(q))
+ return CSIO_SUCCESS;
+
+ csio_dbg(hw, "Aborting SCSI I/Os\n");
+
+ /* Now abort/close I/Os in the queue passed */
+ list_for_each_safe(tmp, next, q) {
+ scmnd = csio_scsi_cmnd((struct csio_ioreq *)tmp);
+ csio_abrt_cls((struct csio_ioreq *)tmp, scmnd);
+ }
+
+ /* Wait till all active I/Os are completed/aborted/closed */
+ while (!list_empty(q) && count--) {
+ spin_unlock_irq(&hw->lock);
+ msleep(CSIO_SCSI_ABORT_Q_POLL_MS);
+ spin_lock_irq(&hw->lock);
+ }
+
+ /* all aborts completed */
+ if (list_empty(q))
+ return CSIO_SUCCESS;
+
+ return CSIO_TIMEOUT;
+}
+
+/*
+ * csio_scsim_cleanup_io - Cleanup all I/Os in SCSI module.
+ * @scm: SCSI module.
+ * @abort: abort required.
+ * Called with lock held, should exit with lock held.
+ * Can sleep when waiting for I/Os to complete.
+ */
+csio_retval_t
+csio_scsim_cleanup_io(struct csio_scsim *scm, bool abort)
+{
+ struct csio_hw *hw = scm->hw;
+ csio_retval_t rv = CSIO_SUCCESS;
+ int count = CSIO_ROUNDUP(60 * 1000, CSIO_SCSI_ABORT_Q_POLL_MS);
+
+ /* No I/Os pending */
+ if (list_empty(&scm->active_q))
+ return CSIO_SUCCESS;
+
+ /* Wait until all active I/Os are completed */
+ while (!list_empty(&scm->active_q) && count--) {
+ spin_unlock_irq(&hw->lock);
+ msleep(CSIO_SCSI_ABORT_Q_POLL_MS);
+ spin_lock_irq(&hw->lock);
+ }
+
+ /* all I/Os completed */
+ if (list_empty(&scm->active_q))
+ return CSIO_SUCCESS;
+
+ /* Else abort */
+ if (abort) {
+ rv = csio_scsi_abort_io_q(scm, &scm->active_q, 30000);
+ if (rv == CSIO_SUCCESS)
+ return rv;
+ csio_dbg(hw, "Some I/O aborts timed out, cleaning up..\n");
+ }
+
+ csio_scsi_cleanup_io_q(scm, &scm->active_q);
+
+ CSIO_DB_ASSERT(list_empty(&scm->active_q));
+
+ return rv;
+}
+
+/*
+ * csio_scsim_cleanup_io_lnode - Cleanup all I/Os of given lnode.
+ * @scm: SCSI module.
+ * @lnode: lnode
+ *
+ * Called with lock held, should exit with lock held.
+ * Can sleep (with dropped lock) when waiting for I/Os to complete.
+ */
+csio_retval_t
+csio_scsim_cleanup_io_lnode(struct csio_scsim *scm, struct csio_lnode *ln)
+{
+ struct csio_hw *hw = scm->hw;
+ struct csio_scsi_level_data sld;
+ csio_retval_t rv;
+ int count = CSIO_ROUNDUP(60 * 1000, CSIO_SCSI_ABORT_Q_POLL_MS);
+
+ csio_dbg(hw, "Gathering all SCSI I/Os on lnode %p\n", ln);
+
+ sld.level = CSIO_LEV_LNODE;
+ sld.lnode = ln;
+ INIT_LIST_HEAD(&ln->cmpl_q);
+ csio_scsi_gather_active_ios(scm, &sld, &ln->cmpl_q);
+
+ /* No I/Os pending on this lnode */
+ if (list_empty(&ln->cmpl_q))
+ return CSIO_SUCCESS;
+
+ /* Wait until all active I/Os on this lnode are completed */
+ while (!list_empty(&ln->cmpl_q) && count--) {
+ spin_unlock_irq(&hw->lock);
+ msleep(CSIO_SCSI_ABORT_Q_POLL_MS);
+ spin_lock_irq(&hw->lock);
+ }
+
+ /* all I/Os completed */
+ if (list_empty(&ln->cmpl_q))
+ return CSIO_SUCCESS;
+
+ csio_dbg(hw, "Some I/Os pending on ln:%p, aborting them..\n", ln);
+
+ /* I/Os are pending, abort them */
+ rv = csio_scsi_abort_io_q(scm, &ln->cmpl_q, 30000);
+ if (rv != CSIO_SUCCESS) {
+ csio_dbg(hw, "Some I/O aborts timed out, cleaning up..\n");
+ csio_scsi_cleanup_io_q(scm, &ln->cmpl_q);
+ }
+
+ CSIO_DB_ASSERT(list_empty(&ln->cmpl_q));
+
+ return rv;
+}
+
+static ssize_t
+csio_show_hw_state(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ if (csio_is_hw_ready(hw))
+ return snprintf(buf, PAGE_SIZE, "ready\n");
+ else
+ return snprintf(buf, PAGE_SIZE, "not ready\n");
+}
+
+/* Device reset */
+static ssize_t
+csio_device_reset(struct device *dev,
+ struct device_attribute *attr, const char *buf, size_t count)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+
+ if (*buf != '1')
+ return -EINVAL;
+
+ /* Delete NPIV lnodes */
+ csio_lnodes_exit(hw, 1);
+
+ /* Block upper IOs */
+ csio_lnodes_block_request(hw);
+
+ spin_lock_irq(&hw->lock);
+ csio_hw_reset(hw);
+ spin_unlock_irq(&hw->lock);
+
+ /* Unblock upper IOs */
+ csio_lnodes_unblock_request(hw);
+ return count;
+}
+
+/* disable port */
+static ssize_t
+csio_disable_port(struct device *dev,
+ struct device_attribute *attr, const char *buf, size_t count)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ bool disable;
+
+ if (*buf == '1' || *buf == '0')
+ disable = (*buf == '1') ? CSIO_TRUE : CSIO_FALSE;
+ else
+ return -EINVAL;
+
+ /* Block upper IOs */
+ csio_lnodes_block_by_port(hw, ln->portid);
+
+ spin_lock_irq(&hw->lock);
+ csio_disable_lnodes(hw, ln->portid, disable);
+ spin_unlock_irq(&hw->lock);
+
+ /* Unblock upper IOs */
+ csio_lnodes_unblock_by_port(hw, ln->portid);
+ return count;
+}
+
+/* Show debug level */
+static ssize_t
+csio_show_dbg_level(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+
+ return snprintf(buf, PAGE_SIZE, "%x\n", ln->params.log_level);
+}
+
+/* Store debug level */
+static ssize_t
+csio_store_dbg_level(struct device *dev,
+ struct device_attribute *attr, const char *buf, size_t count)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ uint32_t dbg_level = 0;
+
+ if (!isdigit(buf[0]))
+ return -EINVAL;
+
+ if (sscanf(buf, "%i", &dbg_level))
+ return -EINVAL;
+
+ ln->params.log_level = dbg_level;
+ hw->params.log_level = dbg_level;
+
+ return 0;
+}
+
+static DEVICE_ATTR(hw_state, S_IRUGO, csio_show_hw_state, NULL);
+static DEVICE_ATTR(device_reset, S_IRUGO | S_IWUSR, NULL, csio_device_reset);
+static DEVICE_ATTR(disable_port, S_IRUGO | S_IWUSR, NULL, csio_disable_port);
+static DEVICE_ATTR(dbg_level, S_IRUGO | S_IWUSR, csio_show_dbg_level,
+ csio_store_dbg_level);
+
+static struct device_attribute *csio_fcoe_lport_attrs[] = {
+ &dev_attr_hw_state,
+ &dev_attr_device_reset,
+ &dev_attr_disable_port,
+ &dev_attr_dbg_level,
+ NULL,
+};
+
+static ssize_t
+csio_show_num_reg_rnodes(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct csio_lnode *ln = shost_priv(class_to_shost(dev));
+
+ return snprintf(buf, PAGE_SIZE, "%d\n", ln->num_reg_rnodes);
+}
+
+static DEVICE_ATTR(num_reg_rnodes, S_IRUGO, csio_show_num_reg_rnodes, NULL);
+
+static struct device_attribute *csio_fcoe_vport_attrs[] = {
+ &dev_attr_num_reg_rnodes,
+ &dev_attr_dbg_level,
+ NULL,
+};
+
+static inline uint32_t
+csio_scsi_copy_to_sgl(struct csio_hw *hw, struct csio_ioreq *req)
+{
+ struct scsi_cmnd *scmnd = (struct scsi_cmnd *)csio_scsi_cmnd(req);
+ struct scatterlist *sg;
+ uint32_t bytes_left;
+ uint32_t bytes_copy;
+ uint32_t buf_off = 0;
+ uint32_t start_off = 0;
+ uint32_t sg_off = 0;
+ void *sg_addr;
+ void *buf_addr;
+ struct csio_dma_buf *dma_buf;
+
+ bytes_left = scsi_bufflen(scmnd);
+ sg = scsi_sglist(scmnd);
+ dma_buf = (struct csio_dma_buf *)csio_list_next(&req->gen_list);
+
+ /* Copy data from driver buffer to SGs of SCSI CMD */
+ while (bytes_left > 0 && sg && dma_buf) {
+ if (buf_off >= dma_buf->len) {
+ buf_off = 0;
+ dma_buf = (struct csio_dma_buf *)
+ csio_list_next(dma_buf);
+ continue;
+ }
+
+ if (start_off >= sg->length) {
+ start_off -= sg->length;
+ sg = sg_next(sg);
+ continue;
+ }
+
+ buf_addr = dma_buf->vaddr + buf_off;
+ sg_off = sg->offset + start_off;
+ bytes_copy = min((dma_buf->len - buf_off),
+ sg->length - start_off);
+ bytes_copy = min((uint32_t)(PAGE_SIZE - (sg_off & ~PAGE_MASK)),
+ bytes_copy);
+
+ sg_addr = kmap_atomic(sg_page(sg) + (sg_off >> PAGE_SHIFT));
+ if (!sg_addr) {
+ csio_err(hw, "failed to kmap sg:%p of ioreq:%p\n",
+ sg, req);
+ break;
+ }
+
+ csio_dbg(hw, "copy_to_sgl:sg_addr %p sg_off %d buf %p len %d\n",
+ sg_addr, sg_off, buf_addr, bytes_copy);
+ memcpy(sg_addr + (sg_off & ~PAGE_MASK), buf_addr, bytes_copy);
+ kunmap_atomic(sg_addr);
+
+ start_off += bytes_copy;
+ buf_off += bytes_copy;
+ bytes_left -= bytes_copy;
+ }
+
+ if (bytes_left > 0)
+ return DID_ERROR;
+ else
+ return DID_OK;
+}
+
+/*
+ * csio_scsi_err_handler - SCSI error handler.
+ * @hw: HW module.
+ * @req: IO request.
+ *
+ */
+static inline void
+csio_scsi_err_handler(struct csio_hw *hw, struct csio_ioreq *req)
+{
+ struct scsi_cmnd *cmnd = (struct scsi_cmnd *)csio_scsi_cmnd(req);
+ struct csio_scsim *scm = csio_hw_to_scsim(hw);
+ struct csio_fcp_resp *fcp_resp;
+ struct csio_dma_buf *dma_buf;
+ uint8_t flags, scsi_status = 0;
+ uint32_t host_status = DID_OK;
+ uint32_t rsp_len = 0, sns_len = 0;
+ struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
+
+
+ switch (req->wr_status) {
+ case FW_HOSTERROR:
+ if (unlikely(!csio_is_hw_ready(hw)))
+ return;
+
+ host_status = DID_ERROR;
+ csio_inc_stats(scm, n_hosterror);
+
+ break;
+ case FW_SCSI_RSP_ERR:
+ dma_buf = &req->dma_buf;
+ fcp_resp = (struct csio_fcp_resp *)dma_buf->vaddr;
+ flags = fcp_resp->flags;
+ scsi_status = fcp_resp->scsi_status;
+
+ if (flags & FCP_RSP_LEN_VAL) {
+ rsp_len = be32_to_cpu(fcp_resp->rsp_len);
+ if ((rsp_len != 0 && rsp_len != 4 && rsp_len != 8) ||
+ (fcp_resp->rsp_code != FCP_TMF_CMPL)) {
+ host_status = DID_ERROR;
+ goto out;
+ }
+ }
+
+ if ((flags & FCP_SNS_LEN_VAL) && fcp_resp->sns_len) {
+ sns_len = be32_to_cpu(fcp_resp->sns_len);
+ if (sns_len > SCSI_SENSE_BUFFERSIZE)
+ sns_len = SCSI_SENSE_BUFFERSIZE;
+
+ memcpy(cmnd->sense_buffer, &fcp_resp->rsvd1 + rsp_len,
+ sns_len);
+ csio_inc_stats(scm, n_autosense);
+ }
+
+ scsi_set_resid(cmnd, 0);
+
+ /* Under run */
+ if (flags & FCP_RESID_UNDER) {
+ scsi_set_resid(cmnd, be32_to_cpu(fcp_resp->resid));
+
+ if (!(flags & FCP_SNS_LEN_VAL) &&
+ (scsi_status == SAM_STAT_GOOD) &&
+ ((scsi_bufflen(cmnd) - scsi_get_resid(cmnd))
+ < cmnd->underflow))
+ host_status = DID_ERROR;
+ } else if (flags & FCP_RESID_OVER)
+ host_status = DID_ERROR;
+
+ csio_inc_stats(scm, n_rsperror);
+ break;
+
+ case FW_SCSI_OVER_FLOW_ERR:
+ csio_warn(hw,
+ "Over-flow error,cmnd:0x%x expected len:0x%x"
+ " resid:0x%x\n", cmnd->cmnd[0],
+ scsi_bufflen(cmnd), scsi_get_resid(cmnd));
+ host_status = DID_ERROR;
+ csio_inc_stats(scm, n_ovflerror);
+ break;
+
+ case FW_SCSI_UNDER_FLOW_ERR:
+ csio_warn(hw,
+ "Under-flow error,cmnd:0x%x expected"
+ " len:0x%x resid:0x%x lun:0x%x ssn:0x%x\n",
+ cmnd->cmnd[0], scsi_bufflen(cmnd),
+ scsi_get_resid(cmnd), cmnd->device->lun,
+ rn->flowid);
+ host_status = DID_ERROR;
+ csio_inc_stats(scm, n_unflerror);
+ break;
+
+ case FW_SCSI_ABORT_REQUESTED:
+ case FW_SCSI_ABORTED:
+ case FW_SCSI_CLOSE_REQUESTED:
+ csio_dbg(hw, "Req %p cmd:%p op:%x %s\n", req, cmnd,
+ cmnd->cmnd[0],
+ (req->wr_status == FW_SCSI_CLOSE_REQUESTED) ?
+ "closed" : "aborted");
+ /*
+ * csio_eh_abort_handler checks this value to
+ * succeed or fail the abort request.
+ */
+ host_status = DID_REQUEUE;
+ if (req->wr_status == FW_SCSI_CLOSE_REQUESTED)
+ csio_inc_stats(scm, n_closed);
+ else
+ csio_inc_stats(scm, n_aborted);
+ break;
+
+ case FW_SCSI_ABORT_TIMEDOUT:
+ /* FW timed out the abort itself */
+ csio_dbg(hw, "FW timed out abort req:%p cmnd:%p status:%x\n",
+ req, cmnd, req->wr_status);
+ host_status = DID_ERROR;
+ csio_inc_stats(scm, n_abrt_timedout);
+ break;
+
+ case FW_RDEV_NOT_READY:
+ /*
+ * In firmware, a RDEV can get into this state
+ * temporarily, before moving into dissapeared/lost
+ * state. So, the driver should complete the request equivalent
+ * to device-disappeared!
+ */
+ csio_inc_stats(scm, n_rdev_nr_error);
+ host_status = DID_ERROR;
+ break;
+
+ case FW_ERR_RDEV_LOST:
+ csio_inc_stats(scm, n_rdev_lost_error);
+ host_status = DID_ERROR;
+ break;
+
+ case FW_ERR_RDEV_LOGO:
+ csio_inc_stats(scm, n_rdev_logo_error);
+ host_status = DID_ERROR;
+ break;
+
+ case FW_ERR_RDEV_IMPL_LOGO:
+ host_status = DID_ERROR;
+ break;
+
+ case FW_ERR_LINK_DOWN:
+ csio_inc_stats(scm, n_link_down_error);
+ host_status = DID_ERROR;
+ break;
+
+ case FW_FCOE_NO_XCHG:
+ csio_inc_stats(scm, n_no_xchg_error);
+ host_status = DID_ERROR;
+ break;
+
+ default:
+ csio_err(hw, "Unknown SCSI FW WR status:%d req:%p cmnd:%p\n",
+ req->wr_status, req, cmnd);
+ CSIO_DB_ASSERT(0);
+
+ csio_inc_stats(scm, n_unknown_error);
+ host_status = DID_ERROR;
+ break;
+ }
+
+out:
+ if (req->nsge > 0)
+ scsi_dma_unmap(cmnd);
+
+ cmnd->result = (((host_status) << 16) | scsi_status);
+ cmnd->scsi_done(cmnd);
+
+ /* Wake up waiting threads */
+ csio_scsi_cmnd(req) = NULL;
+ complete_all(&req->cmplobj);
+}
+
+/*
+ * csio_scsi_cbfn - SCSI callback function.
+ * @hw: HW module.
+ * @req: IO request.
+ *
+ */
+static void
+csio_scsi_cbfn(struct csio_hw *hw, struct csio_ioreq *req)
+{
+ struct scsi_cmnd *cmnd = (struct scsi_cmnd *)csio_scsi_cmnd(req);
+ uint8_t scsi_status = SAM_STAT_GOOD;
+ uint32_t host_status = DID_OK;
+
+ if (likely(req->wr_status == FW_SUCCESS)) {
+ if (req->nsge > 0) {
+ scsi_dma_unmap(cmnd);
+ if (req->dcopy)
+ host_status = csio_scsi_copy_to_sgl(hw, req);
+ }
+
+ cmnd->result = (((host_status) << 16) | scsi_status);
+ cmnd->scsi_done(cmnd);
+ csio_scsi_cmnd(req) = NULL;
+ csio_inc_stats(csio_hw_to_scsim(hw), n_tot_success);
+ } else {
+ /* Error handling */
+ csio_scsi_err_handler(hw, req);
+ }
+}
+
+/**
+ * csio_queuecommand_lck - Entry point to kickstart an I/O request.
+ * @cmnd: The I/O request from ML.
+ * @done: The ML callback routine.
+ *
+ * This routine does the following:
+ * - Checks for HW and Rnode module readiness.
+ * - Gets a free ioreq structure (which is already initialized
+ * to uninit during its allocation).
+ * - Maps SG elements.
+ * - Initializes ioreq members.
+ * - Kicks off the SCSI state machine for this IO.
+ * - Returns busy status on error.
+ */
+static int
+csio_queuecommand_lck(struct scsi_cmnd *cmnd, void (*done)(struct scsi_cmnd *))
+{
+ struct csio_lnode *ln = shost_priv(cmnd->device->host);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
+ struct csio_ioreq *ioreq = NULL;
+ unsigned long flags;
+ int nsge = 0;
+ int rv = SCSI_MLQUEUE_HOST_BUSY, nr;
+ csio_retval_t retval;
+ int cpu;
+ struct csio_scsi_qset *sqset;
+ struct fc_rport *rport = starget_to_rport(scsi_target(cmnd->device));
+
+ if (!blk_rq_cpu_valid(cmnd->request))
+ cpu = smp_processor_id();
+ else
+ cpu = cmnd->request->cpu;
+
+ sqset = &hw->sqset[ln->portid][cpu];
+
+ nr = fc_remote_port_chkready(rport);
+ if (nr) {
+ cmnd->result = nr;
+ csio_inc_stats(scsim, n_rn_nr_error);
+ goto err_done;
+ }
+
+ if (unlikely(!csio_is_hw_ready(hw))) {
+ cmnd->result = (DID_REQUEUE << 16);
+ csio_inc_stats(scsim, n_hw_nr_error);
+ goto err_done;
+ }
+
+ /* Get req->nsge, if there are SG elements to be mapped */
+ nsge = scsi_dma_map(cmnd);
+ if (unlikely(nsge < 0)) {
+ csio_inc_stats(scsim, n_dmamap_error);
+ goto err;
+ }
+
+ /* Do we support so many mappings? */
+ if (unlikely(nsge > scsim->max_sge)) {
+ csio_warn(hw,
+ "More SGEs than can be supported."
+ " SGEs: %d, Max SGEs: %d\n", nsge, scsim->max_sge);
+ csio_inc_stats(scsim, n_unsupp_sge_error);
+ goto err_dma_unmap;
+ }
+
+ /* Get a free ioreq structure - SM is already set to uninit */
+ ioreq = csio_get_scsi_ioreq_lock(hw, scsim);
+ if (!ioreq) {
+ csio_err(hw, "Out of I/O request elements. Active #:%d\n",
+ scsim->stats.n_active);
+ csio_inc_stats(scsim, n_no_req_error);
+ goto err_dma_unmap;
+ }
+
+ ioreq->nsge = nsge;
+ ioreq->lnode = ln;
+ ioreq->rnode = rn;
+ ioreq->iq_idx = sqset->iq_idx;
+ ioreq->eq_idx = sqset->eq_idx;
+ ioreq->wr_status = 0;
+ ioreq->drv_status = CSIO_SUCCESS;
+ csio_scsi_cmnd(ioreq) = (void *)cmnd;
+ ioreq->tmo = 0;
+
+ switch (cmnd->sc_data_direction) {
+ case DMA_BIDIRECTIONAL:
+ ioreq->datadir = CSIO_IOREQF_DMA_BIDI;
+ csio_inc_stats(ln, n_control_requests);
+ break;
+ case DMA_TO_DEVICE:
+ ioreq->datadir = CSIO_IOREQF_DMA_WRITE;
+ csio_inc_stats(ln, n_output_requests);
+ ln->stats.n_output_bytes += scsi_bufflen(cmnd);
+ break;
+ case DMA_FROM_DEVICE:
+ ioreq->datadir = CSIO_IOREQF_DMA_READ;
+ csio_inc_stats(ln, n_input_requests);
+ ln->stats.n_input_bytes += scsi_bufflen(cmnd);
+ break;
+ case DMA_NONE:
+ ioreq->datadir = CSIO_IOREQF_DMA_NONE;
+ csio_inc_stats(ln, n_control_requests);
+ break;
+ default:
+ CSIO_DB_ASSERT(0);
+ break;
+ }
+
+ /* Set cbfn */
+ ioreq->io_cbfn = csio_scsi_cbfn;
+
+ /* Needed during abort */
+ cmnd->host_scribble = (unsigned char *)ioreq;
+ cmnd->scsi_done = done;
+ cmnd->SCp.Message = 0;
+
+ /* Kick off SCSI IO SM on the ioreq */
+ spin_lock_irqsave(&hw->lock, flags);
+ retval = csio_scsi_start_io(ioreq);
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+ if (retval != CSIO_SUCCESS) {
+ csio_err(hw, "ioreq: %p couldnt be started, status:%d\n",
+ ioreq, retval);
+ csio_inc_stats(scsim, n_busy_error);
+ goto err_put_req;
+ }
+
+ return 0;
+
+err_put_req:
+ csio_put_scsi_ioreq_lock(hw, scsim, ioreq);
+err_dma_unmap:
+ if (nsge > 0)
+ scsi_dma_unmap(cmnd);
+err:
+ return rv;
+
+err_done:
+ done(cmnd);
+ return 0;
+}
+
+static DEF_SCSI_QCMD(csio_queuecommand);
+
+static csio_retval_t
+csio_do_abrt_cls(struct csio_hw *hw, struct csio_ioreq *ioreq, bool abort)
+{
+ csio_retval_t rv;
+ int cpu = smp_processor_id();
+ struct csio_lnode *ln = ioreq->lnode;
+ struct csio_scsi_qset *sqset = &hw->sqset[ln->portid][cpu];
+
+ ioreq->tmo = CSIO_SCSI_ABRT_TMO_MS;
+ /*
+ * Use current processor queue for posting the abort/close, but retain
+ * the ingress queue ID of the original I/O being aborted/closed - we
+ * need the abort/close completion to be received on the same queue
+ * as the original I/O.
+ */
+ ioreq->eq_idx = sqset->eq_idx;
+
+ if (abort == SCSI_ABORT)
+ rv = csio_scsi_abort(ioreq);
+ else /* close */
+ rv = csio_scsi_close(ioreq);
+
+ return rv;
+}
+
+static int
+csio_eh_abort_handler(struct scsi_cmnd *cmnd)
+{
+ struct csio_ioreq *ioreq;
+ struct csio_lnode *ln = shost_priv(cmnd->device->host);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ int ready = 0, ret;
+ unsigned long tmo = 0;
+ csio_retval_t rv;
+ struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
+
+ ret = fc_block_scsi_eh(cmnd);
+ if (ret)
+ return ret;
+
+ ioreq = (struct csio_ioreq *)cmnd->host_scribble;
+ if (!ioreq)
+ return SUCCESS;
+
+ if (!rn)
+ return FAILED;
+
+ csio_dbg(hw,
+ "Request to abort ioreq:%p cmd:%p cdb:%08llx"
+ " ssni:0x%x lun:%d iq:0x%x\n",
+ ioreq, cmnd, *((uint64_t *)cmnd->cmnd), rn->flowid,
+ cmnd->device->lun, csio_q_physiqid(hw, ioreq->iq_idx));
+
+ if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) != cmnd) {
+ csio_inc_stats(scsim, n_abrt_race_comp);
+ return SUCCESS;
+ }
+
+ ready = csio_is_lnode_ready(ln);
+ tmo = CSIO_SCSI_ABRT_TMO_MS;
+
+ spin_lock_irq(&hw->lock);
+ rv = csio_do_abrt_cls(hw, ioreq, (ready ? SCSI_ABORT : SCSI_CLOSE));
+ spin_unlock_irq(&hw->lock);
+
+ if (rv != CSIO_SUCCESS) {
+ if (rv == CSIO_INVAL) {
+ /* Return success, if abort/close request issued on
+ * already completed IO
+ */
+ return SUCCESS;
+ }
+ if (ready)
+ csio_inc_stats(scsim, n_abrt_busy_error);
+ else
+ csio_inc_stats(scsim, n_cls_busy_error);
+
+ goto inval_scmnd;
+ }
+
+ /* Wait for completion */
+ init_completion(&ioreq->cmplobj);
+ wait_for_completion_timeout(&ioreq->cmplobj, msecs_to_jiffies(tmo));
+
+ /* FW didnt respond to abort within our timeout */
+ if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) == cmnd) {
+
+ csio_err(hw, "Abort timed out -- req: %p\n", ioreq);
+ csio_inc_stats(scsim, n_abrt_timedout);
+
+inval_scmnd:
+ if (ioreq->nsge > 0)
+ scsi_dma_unmap(cmnd);
+
+ spin_lock_irq(&hw->lock);
+ csio_scsi_cmnd(ioreq) = NULL;
+ spin_unlock_irq(&hw->lock);
+
+ cmnd->result = (DID_ERROR << 16);
+ cmnd->scsi_done(cmnd);
+
+ return FAILED;
+ }
+
+ /* FW successfully aborted the request */
+ if (host_byte(cmnd->result) == DID_REQUEUE) {
+ csio_info(hw,
+ "Aborted SCSI command to (%d:%d) serial#:0x%lx\n",
+ cmnd->device->id, cmnd->device->lun,
+ cmnd->serial_number);
+ return SUCCESS;
+ } else {
+ csio_info(hw,
+ "Failed to abort SCSI command, (%d:%d) serial#:0x%lx\n",
+ cmnd->device->id, cmnd->device->lun,
+ cmnd->serial_number);
+ return FAILED;
+ }
+}
+
+/*
+ * csio_tm_cbfn - TM callback function.
+ * @hw: HW module.
+ * @req: IO request.
+ *
+ * Cache the result in 'cmnd', since ioreq will be freed soon
+ * after we return from here, and the waiting thread shouldnt trust
+ * the ioreq contents.
+ */
+static void
+csio_tm_cbfn(struct csio_hw *hw, struct csio_ioreq *req)
+{
+ struct scsi_cmnd *cmnd = (struct scsi_cmnd *)csio_scsi_cmnd(req);
+ struct csio_dma_buf *dma_buf;
+ uint8_t flags = 0;
+ struct csio_fcp_resp *fcp_resp;
+
+ csio_dbg(hw, "req: %p in csio_tm_cbfn status: %d\n",
+ req, req->wr_status);
+
+ /* Cache FW return status */
+ cmnd->SCp.Status = req->wr_status;
+
+ /* Special handling based on FCP response */
+
+ /*
+ * FW returns us this error, if flags were set. FCP4 says
+ * FCP_RSP_LEN_VAL in flags shall be set for TM completions.
+ * So if a target were to set this bit, we expect that the
+ * rsp_code is set to FCP_TMF_CMPL for a successful TM
+ * completion. Any other rsp_code means TM operation failed.
+ * If a target were to just ignore setting flags, we treat
+ * the TM operation as success, and FW returns FW_SUCCESS.
+ */
+ if (req->wr_status == FW_SCSI_RSP_ERR) {
+ dma_buf = &req->dma_buf;
+ fcp_resp = (struct csio_fcp_resp *)dma_buf->vaddr;
+ flags = fcp_resp->flags;
+
+ /* Modify return status if flags indicate success */
+ if (flags & FCP_RSP_LEN_VAL)
+ if (fcp_resp->rsp_code == FCP_TMF_CMPL)
+ cmnd->SCp.Status = FW_SUCCESS;
+
+ csio_dbg(hw, "TM FCP rsp code: %d\n", fcp_resp->rsp_code);
+ }
+
+ /* Wake up the TM handler thread */
+ csio_scsi_cmnd(req) = NULL;
+}
+
+static int
+csio_eh_lun_reset_handler(struct scsi_cmnd *cmnd)
+{
+ struct csio_lnode *ln = shost_priv(cmnd->device->host);
+ struct csio_hw *hw = csio_lnode_to_hw(ln);
+ struct csio_scsim *scsim = csio_hw_to_scsim(hw);
+ struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
+ struct csio_ioreq *ioreq = NULL;
+ struct csio_scsi_qset *sqset;
+ unsigned long flags;
+ csio_retval_t retval;
+ int count, ret;
+ LIST_HEAD(local_q);
+ struct csio_scsi_level_data sld;
+
+ if (!rn)
+ goto fail;
+
+ csio_dbg(hw, "Request to reset LUN:%d (ssni:0x%x tgtid:%d)\n",
+ cmnd->device->lun, rn->flowid, rn->scsi_id);
+
+ if (!csio_is_lnode_ready(ln)) {
+ csio_err(hw,
+ "LUN reset cannot be issued on non-ready"
+ " local node vnpi:0x%x (LUN:%d)\n",
+ ln->vnp_flowid, cmnd->device->lun);
+ goto fail;
+ }
+
+ /* Lnode is ready, now wait on rport node readiness */
+ ret = fc_block_scsi_eh(cmnd);
+ if (ret)
+ return ret;
+
+ /*
+ * If we have blocked in the previous call, at this point, either the
+ * remote node has come back online, or device loss timer has fired
+ * and the remote node is destroyed. Allow the LUN reset only for
+ * the former case, since LUN reset is a TMF I/O on the wire, and we
+ * need a valid session to issue it.
+ */
+ if (fc_remote_port_chkready(rn->rport)) {
+ csio_err(hw,
+ "LUN reset cannot be issued on non-ready"
+ " remote node ssni:0x%x (LUN:%d)\n",
+ rn->flowid, cmnd->device->lun);
+ goto fail;
+ }
+
+ /* Get a free ioreq structure - SM is already set to uninit */
+ ioreq = csio_get_scsi_ioreq_lock(hw, scsim);
+
+ if (!ioreq) {
+ csio_err(hw, "Out of IO request elements. Active # :%d\n",
+ scsim->stats.n_active);
+ goto fail;
+ }
+
+ sqset = &hw->sqset[ln->portid][smp_processor_id()];
+ ioreq->nsge = 0;
+ ioreq->lnode = ln;
+ ioreq->rnode = rn;
+ ioreq->iq_idx = sqset->iq_idx;
+ ioreq->eq_idx = sqset->eq_idx;
+
+ csio_scsi_cmnd(ioreq) = cmnd;
+ cmnd->host_scribble = (unsigned char *)ioreq;
+ cmnd->SCp.Status = 0;
+
+ cmnd->SCp.Message = FCP_TMF_LUN_RESET;
+ ioreq->tmo = CSIO_SCSI_LUNRST_TMO_MS / 1000;
+
+ /*
+ * FW times the LUN reset for ioreq->tmo, so we got to wait a little
+ * longer (10s for now) than that to allow FW to return the timed
+ * out command.
+ */
+ count = CSIO_ROUNDUP((ioreq->tmo + 10) * 1000, CSIO_SCSI_TM_POLL_MS);
+
+ /* Set cbfn */
+ ioreq->io_cbfn = csio_tm_cbfn;
+
+ /* Save of the ioreq info for later use */
+ sld.level = CSIO_LEV_LUN;
+ sld.lnode = ioreq->lnode;
+ sld.rnode = ioreq->rnode;
+ sld.oslun = (uint64_t)cmnd->device->lun;
+
+ spin_lock_irqsave(&hw->lock, flags);
+ /* Kick off TM SM on the ioreq */
+ retval = csio_scsi_start_tm(ioreq);
+ spin_unlock_irqrestore(&hw->lock, flags);
+
+ if (retval != CSIO_SUCCESS) {
+ csio_err(hw, "Failed to issue LUN reset, req:%p, status:%d\n",
+ ioreq, retval);
+ goto fail_ret_ioreq;
+ }
+
+ csio_dbg(hw, "Waiting max %d secs for LUN reset completion\n",
+ count * (CSIO_SCSI_TM_POLL_MS / 1000));
+ /* Wait for completion */
+ while ((((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) == cmnd)
+ && count--)
+ msleep(CSIO_SCSI_TM_POLL_MS);
+
+ /* LUN reset timed-out */
+ if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) == cmnd) {
+ csio_err(hw, "LUN reset (%d:%d) timed out\n",
+ cmnd->device->id, cmnd->device->lun);
+
+ spin_lock_irq(&hw->lock);
+ csio_scsi_drvcleanup(ioreq);
+ list_del_init(&ioreq->sm.sm_list);
+ spin_unlock_irq(&hw->lock);
+
+ goto fail_ret_ioreq;
+ }
+
+ /* LUN reset returned, check cached status */
+ if (cmnd->SCp.Status != FW_SUCCESS) {
+ csio_err(hw, "LUN reset failed (%d:%d), status: %d\n",
+ cmnd->device->id, cmnd->device->lun, cmnd->SCp.Status);
+ goto fail;
+ }
+
+ /* LUN reset succeeded, Start aborting affected I/Os */
+ /*
+ * Since the host guarantees during LUN reset that there
+ * will not be any more I/Os to that LUN, until the LUN reset
+ * completes, we gather pending I/Os after the LUN reset.
+ */
+ spin_lock_irq(&hw->lock);
+ csio_scsi_gather_active_ios(scsim, &sld, &local_q);
+
+ retval = csio_scsi_abort_io_q(scsim, &local_q, 30000);
+ spin_unlock_irq(&hw->lock);
+
+ /* Aborts may have timed out */
+ if (retval != CSIO_SUCCESS) {
+ csio_err(hw,
+ "Attempt to abort I/Os during LUN reset of %d"
+ " returned %d\n", cmnd->device->lun, retval);
+ /* Return I/Os back to active_q */
+ spin_lock_irq(&hw->lock);
+ list_splice_tail_init(&local_q, &scsim->active_q);
+ spin_unlock_irq(&hw->lock);
+ goto fail;
+ }
+
+ csio_inc_stats(rn, n_lun_rst);
+
+ csio_info(hw, "LUN reset occurred (%d:%d)\n",
+ cmnd->device->id, cmnd->device->lun);
+
+ return SUCCESS;
+
+fail_ret_ioreq:
+ csio_put_scsi_ioreq_lock(hw, scsim, ioreq);
+fail:
+ csio_inc_stats(rn, n_lun_rst_fail);
+ return FAILED;
+}
+
+static int
+csio_slave_alloc(struct scsi_device *sdev)
+{
+ struct fc_rport *rport = starget_to_rport(scsi_target(sdev));
+
+ if (!rport || fc_remote_port_chkready(rport))
+ return -ENXIO;
+
+ sdev->hostdata = *((struct csio_lnode **)(rport->dd_data));
+
+ return 0;
+}
+
+static int
+csio_slave_configure(struct scsi_device *sdev)
+{
+ if (sdev->tagged_supported)
+ scsi_activate_tcq(sdev, csio_lun_qdepth);
+ else
+ scsi_deactivate_tcq(sdev, csio_lun_qdepth);
+
+ return 0;
+}
+
+static void
+csio_slave_destroy(struct scsi_device *sdev)
+{
+ sdev->hostdata = NULL;
+}
+
+static int
+csio_scan_finished(struct Scsi_Host *shost, unsigned long time)
+{
+ struct csio_lnode *ln = shost_priv(shost);
+ int rv = 0;
+
+ spin_lock_irq(shost->host_lock);
+ if (!ln->hwp || csio_list_deleted(&ln->sm.sm_list)) {
+ spin_unlock_irq(shost->host_lock);
+ return 1;
+ }
+
+ rv = csio_scan_done(ln, jiffies, time, csio_max_scan_tmo * HZ,
+ csio_delta_scan_tmo * HZ);
+
+ spin_unlock_irq(shost->host_lock);
+
+ return rv;
+}
+
+struct scsi_host_template csio_fcoe_shost_template = {
+ .module = THIS_MODULE,
+ .name = CSIO_DRV_DESC,
+ .proc_name = KBUILD_MODNAME,
+ .queuecommand = csio_queuecommand,
+ .eh_abort_handler = csio_eh_abort_handler,
+ .eh_device_reset_handler = csio_eh_lun_reset_handler,
+ .slave_alloc = csio_slave_alloc,
+ .slave_configure = csio_slave_configure,
+ .slave_destroy = csio_slave_destroy,
+ .scan_finished = csio_scan_finished,
+ .this_id = -1,
+ .sg_tablesize = CSIO_SCSI_MAX_SGE,
+ .cmd_per_lun = CSIO_MAX_CMD_PER_LUN,
+ .use_clustering = ENABLE_CLUSTERING,
+ .shost_attrs = csio_fcoe_lport_attrs,
+ .max_sectors = CSIO_MAX_SECTOR_SIZE,
+};
+
+struct scsi_host_template csio_fcoe_shost_vport_template = {
+ .module = THIS_MODULE,
+ .name = CSIO_DRV_DESC,
+ .proc_name = KBUILD_MODNAME,
+ .queuecommand = csio_queuecommand,
+ .eh_abort_handler = csio_eh_abort_handler,
+ .eh_device_reset_handler = csio_eh_lun_reset_handler,
+ .slave_alloc = csio_slave_alloc,
+ .slave_configure = csio_slave_configure,
+ .slave_destroy = csio_slave_destroy,
+ .scan_finished = csio_scan_finished,
+ .this_id = -1,
+ .sg_tablesize = CSIO_SCSI_MAX_SGE,
+ .cmd_per_lun = CSIO_MAX_CMD_PER_LUN,
+ .use_clustering = ENABLE_CLUSTERING,
+ .shost_attrs = csio_fcoe_vport_attrs,
+ .max_sectors = CSIO_MAX_SECTOR_SIZE,
+};
+
+/*
+ * csio_scsi_alloc_ddp_bufs - Allocate buffers for DDP of unaligned SGLs.
+ * @scm: SCSI Module
+ * @hw: HW device.
+ * @buf_size: buffer size
+ * @num_buf : Number of buffers.
+ *
+ * This routine allocates DMA buffers required for SCSI Data xfer, if
+ * each SGL buffer for a SCSI Read request posted by SCSI midlayer are
+ * not virtually contiguous.
+ */
+static csio_retval_t
+csio_scsi_alloc_ddp_bufs(struct csio_scsim *scm, struct csio_hw *hw,
+ int buf_size, int num_buf)
+{
+ int n = 0;
+ struct list_head *tmp;
+ struct csio_dma_buf *ddp_desc = NULL;
+ uint32_t unit_size = 0;
+
+ if (!num_buf)
+ return CSIO_SUCCESS;
+
+ if (!buf_size)
+ return CSIO_INVAL;
+
+ INIT_LIST_HEAD(&scm->ddp_freelist);
+
+ /* Align buf size to page size */
+ buf_size = (buf_size + PAGE_SIZE - 1) & PAGE_MASK;
+ /* Initialize dma descriptors */
+ for (n = 0; n < num_buf; n++) {
+ /* Set unit size to request size */
+ unit_size = buf_size;
+ ddp_desc = kzalloc(sizeof(struct csio_dma_buf), GFP_KERNEL);
+ if (!ddp_desc) {
+ csio_err(hw,
+ "Failed to allocate ddp descriptors,"
+ " Num allocated = %d.\n",
+ scm->stats.n_free_ddp);
+ goto no_mem;
+ }
+
+ /* Allocate Dma buffers for DDP */
+ ddp_desc->vaddr = pci_alloc_consistent(hw->pdev, unit_size,
+ &ddp_desc->paddr);
+ if (!ddp_desc->vaddr) {
+ csio_err(hw,
+ "SCSI response DMA buffer (ddp) allocation"
+ " failed!\n");
+ kfree(ddp_desc);
+ goto no_mem;
+ }
+
+ ddp_desc->len = unit_size;
+
+ /* Added it to scsi ddp freelist */
+ list_add_tail(&ddp_desc->list, &scm->ddp_freelist);
+ csio_inc_stats(scm, n_free_ddp);
+ }
+
+ return CSIO_SUCCESS;
+no_mem:
+ /* release dma descs back to freelist and free dma memory */
+ list_for_each(tmp, &scm->ddp_freelist) {
+ ddp_desc = (struct csio_dma_buf *) tmp;
+ tmp = csio_list_prev(tmp);
+ pci_free_consistent(hw->pdev, ddp_desc->len, ddp_desc->vaddr,
+ ddp_desc->paddr);
+ list_del_init(&ddp_desc->list);
+ kfree(ddp_desc);
+ }
+ scm->stats.n_free_ddp = 0;
+
+ return CSIO_NOMEM;
+}
+
+/*
+ * csio_scsi_free_ddp_bufs - free DDP buffers of unaligned SGLs.
+ * @scm: SCSI Module
+ * @hw: HW device.
+ *
+ * This routine frees ddp buffers.
+ */
+static csio_retval_t
+csio_scsi_free_ddp_bufs(struct csio_scsim *scm, struct csio_hw *hw)
+{
+ struct list_head *tmp;
+ struct csio_dma_buf *ddp_desc;
+
+ /* release dma descs back to freelist and free dma memory */
+ list_for_each(tmp, &scm->ddp_freelist) {
+ ddp_desc = (struct csio_dma_buf *) tmp;
+ tmp = csio_list_prev(tmp);
+ pci_free_consistent(hw->pdev, ddp_desc->len, ddp_desc->vaddr,
+ ddp_desc->paddr);
+ list_del_init(&ddp_desc->list);
+ kfree(ddp_desc);
+ }
+ scm->stats.n_free_ddp = 0;
+
+ return CSIO_NOMEM;
+}
+
+/**
+ * csio_scsim_init - Initialize SCSI Module
+ * @scm: SCSI Module
+ * @hw: HW module
+ *
+ */
+csio_retval_t
+csio_scsim_init(struct csio_scsim *scm, struct csio_hw *hw)
+{
+ int i;
+ struct csio_ioreq *ioreq;
+ struct csio_dma_buf *dma_buf;
+
+ INIT_LIST_HEAD(&scm->active_q);
+ scm->hw = hw;
+
+ scm->proto_cmd_len = sizeof(struct csio_fcp_cmnd);
+ scm->proto_rsp_len = sizeof(struct csio_fcp_resp);
+ scm->max_sge = CSIO_SCSI_MAX_SGE;
+
+ spin_lock_init(&scm->freelist_lock);
+
+ /* Pre-allocate ioreqs and initialize them */
+ INIT_LIST_HEAD(&scm->ioreq_freelist);
+ for (i = 0; i < csio_scsi_ioreqs; i++) {
+
+ ioreq = kzalloc(sizeof(struct csio_ioreq), GFP_KERNEL);
+ if (!ioreq) {
+ csio_err(hw,
+ "I/O request element allocation failed, "
+ " Num allocated = %d.\n",
+ scm->stats.n_free_ioreq);
+
+ goto free_ioreq;
+ }
+
+ /* Allocate Dma buffers for Response Payload */
+ dma_buf = &ioreq->dma_buf;
+ dma_buf->vaddr = pci_pool_alloc(hw->scsi_pci_pool, GFP_KERNEL,
+ &dma_buf->paddr);
+ if (!dma_buf->vaddr) {
+ csio_err(hw,
+ "SCSI response DMA buffer allocation"
+ " failed!\n");
+ kfree(ioreq);
+ goto free_ioreq;
+ }
+
+ dma_buf->len = scm->proto_rsp_len;
+
+ /* Set state to uninit */
+ csio_init_state(&ioreq->sm, csio_scsis_uninit);
+ INIT_LIST_HEAD(&ioreq->gen_list);
+ init_completion(&ioreq->cmplobj);
+
+ list_add_tail(&ioreq->sm.sm_list, &scm->ioreq_freelist);
+ csio_inc_stats(scm, n_free_ioreq);
+ }
+
+ if (csio_scsi_alloc_ddp_bufs(scm, hw, PAGE_SIZE, csio_ddp_descs))
+ goto free_ioreq;
+
+ return CSIO_SUCCESS;
+
+free_ioreq:
+ /*
+ * Free up existing allocations, since an error
+ * from here means we are returning for good
+ */
+ while (!list_empty(&scm->ioreq_freelist)) {
+ csio_deq_from_head(&scm->ioreq_freelist, &ioreq);
+
+ dma_buf = &ioreq->dma_buf;
+ pci_pool_free(hw->scsi_pci_pool, dma_buf->vaddr,
+ dma_buf->paddr);
+
+ kfree(ioreq);
+ }
+
+ scm->stats.n_free_ioreq = 0;
+
+ return CSIO_NOMEM;
+}
+
+/**
+ * csio_scsim_exit: Uninitialize SCSI Module
+ * @scm: SCSI Module
+ *
+ */
+void
+csio_scsim_exit(struct csio_scsim *scm)
+{
+ struct csio_ioreq *ioreq;
+ struct csio_dma_buf *dma_buf;
+
+ while (!list_empty(&scm->ioreq_freelist)) {
+ csio_deq_from_head(&scm->ioreq_freelist, &ioreq);
+
+ dma_buf = &ioreq->dma_buf;
+ pci_pool_free(scm->hw->scsi_pci_pool, dma_buf->vaddr,
+ dma_buf->paddr);
+
+ kfree(ioreq);
+ }
+
+ scm->stats.n_free_ioreq = 0;
+
+ csio_scsi_free_ddp_bufs(scm, scm->hw);
+}
--
1.7.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* Re: [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5).
2012-08-23 22:27 ` [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5) Naresh Kumar Inna
@ 2012-08-23 19:48 ` Nicholas A. Bellinger
2012-08-24 17:40 ` Naresh Kumar Inna
0 siblings, 1 reply; 33+ messages in thread
From: Nicholas A. Bellinger @ 2012-08-23 19:48 UTC (permalink / raw)
To: Naresh Kumar Inna; +Cc: JBottomley, linux-scsi, dm, netdev, chethan
On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
> This patch contains code to implement the interrupt handling and the fast
> path I/O functionality. The interrupt handling includes allocation of
> MSIX vectors, registering and implemeting the interrupt service routines.
> The fast path I/O functionality includes posting the I/O request to firmware
> via Work Requests, tracking/completing them, and handling task management
> requests. SCSI midlayer host template implementation is also covered by
> this patch.
>
> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
> ---
Hi Naresh,
My review comments are inline below..
> drivers/scsi/csiostor/csio_isr.c | 631 ++++++++++
> drivers/scsi/csiostor/csio_scsi.c | 2498 +++++++++++++++++++++++++++++++++++++
> 2 files changed, 3129 insertions(+), 0 deletions(-)
> create mode 100644 drivers/scsi/csiostor/csio_isr.c
> create mode 100644 drivers/scsi/csiostor/csio_scsi.c
>
> diff --git a/drivers/scsi/csiostor/csio_isr.c b/drivers/scsi/csiostor/csio_isr.c
> new file mode 100644
> index 0000000..96633e9
> --- /dev/null
> +++ b/drivers/scsi/csiostor/csio_isr.c
<SNIP>
> +#define csio_extra_msix_desc(_desc, _len, _str, _arg1, _arg2, _arg3) \
> +do { \
> + memset((_desc), 0, (_len) + 1); \
> + snprintf((_desc), (_len), (_str), (_arg1), (_arg2), (_arg3)); \
> +} while (0)
> +
This type of macro usage is not necessary for just two users below.
Please inline code like this.
> +static void
> +csio_add_msix_desc(struct csio_hw *hw)
> +{
> + int i;
> + struct csio_msix_entries *entryp = &hw->msix_entries[0];
> + int k = CSIO_EXTRA_VECS;
> + int len = sizeof(entryp->desc) - 1;
> + int cnt = hw->num_sqsets + k;
> +
> + /* Non-data vector */
> + csio_extra_msix_desc(entryp->desc, len, "csio-%02x:%02x:%x-nondata",
> + CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
> + CSIO_PCI_FUNC(hw));
> + entryp++;
> + csio_extra_msix_desc(entryp->desc, len, "csio-%02x:%02x:%x-fwevt",
> + CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
> + CSIO_PCI_FUNC(hw));
> + entryp++;
> +
> + /* Name SCSI vecs */
> + for (i = k; i < cnt; i++, entryp++) {
> + memset(entryp->desc, 0, len + 1);
> + snprintf(entryp->desc, len, "csio-%02x:%02x:%x-scsi%d",
> + CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
> + CSIO_PCI_FUNC(hw), i - CSIO_EXTRA_VECS);
> + }
> +}
> +
> diff --git a/drivers/scsi/csiostor/csio_scsi.c b/drivers/scsi/csiostor/csio_scsi.c
> new file mode 100644
> index 0000000..0f87b00
> --- /dev/null
> +++ b/drivers/scsi/csiostor/csio_scsi.c
> +
> +/*
> + * csio_scsi_match_io - Match an ioreq with the given SCSI level data.
> + * @ioreq: The I/O request
> + * @sld: Level information
> + *
> + * Should be called with lock held.
> + *
> + */
> +static bool
> +csio_scsi_match_io(struct csio_ioreq *ioreq, struct csio_scsi_level_data *sld)
> +{
> + struct scsi_cmnd *scmnd = csio_scsi_cmnd(ioreq);
> +
> + switch (sld->level) {
> + case CSIO_LEV_LUN:
> + if (scmnd == NULL)
> + return CSIO_FALSE;
> +
> + return ((ioreq->lnode == sld->lnode) &&
> + (ioreq->rnode == sld->rnode) &&
> + ((uint64_t)scmnd->device->lun == sld->oslun));
> +
> + case CSIO_LEV_RNODE:
> + return ((ioreq->lnode == sld->lnode) &&
> + (ioreq->rnode == sld->rnode));
> + case CSIO_LEV_LNODE:
> + return (ioreq->lnode == sld->lnode);
> + case CSIO_LEV_ALL:
> + return CSIO_TRUE;
> + default:
> + return CSIO_FALSE;
> + }
> +}
> +
Why can't CSIO_[TRUE,FALSE] just use normal Boolean defines..?
> +/*
> + * csio_scsi_fcp_cmnd - Frame the SCSI FCP command paylod.
> + * @req: IO req structure.
> + * @addr: DMA location to place the payload.
> + *
> + * This routine is shared between FCP_WRITE, FCP_READ and FCP_CMD requests.
> + */
> +static inline void
> +csio_scsi_fcp_cmnd(struct csio_ioreq *req, void *addr)
> +{
> + struct csio_fcp_cmnd *fcp_cmnd = (struct csio_fcp_cmnd *)addr;
> + struct scsi_cmnd *scmnd = csio_scsi_cmnd(req);
> +
> + /* Check for Task Management */
> + if (likely(scmnd->SCp.Message == 0)) {
> + int_to_scsilun(scmnd->device->lun,
> + (struct scsi_lun *)fcp_cmnd->lun);
> + fcp_cmnd->tm_flags = 0;
> + fcp_cmnd->cmdref = 0;
> + fcp_cmnd->pri_ta = 0;
> +
> + memcpy(fcp_cmnd->cdb, scmnd->cmnd, 16);
> + csio_scsi_tag(scmnd, &fcp_cmnd->pri_ta,
> + FCP_PTA_HEADQ, FCP_PTA_ORDERED, FCP_PTA_SIMPLE);
> + fcp_cmnd->dl = cpu_to_be32(scsi_bufflen(scmnd));
> +
> + if (req->nsge)
> + if (req->datadir == CSIO_IOREQF_DMA_WRITE)
The same goes for CSIO_IOREQF_DMA_*...
Why can't this just be DMA_* defs from include/linux/dma-direction.h..?
> + fcp_cmnd->flags = FCP_CFL_WRDATA;
> + else
> + fcp_cmnd->flags = FCP_CFL_RDDATA;
> + else
> + fcp_cmnd->flags = 0;
> + } else {
> + memset(fcp_cmnd, 0, sizeof(*fcp_cmnd));
> + int_to_scsilun(scmnd->device->lun,
> + (struct scsi_lun *)fcp_cmnd->lun);
> + fcp_cmnd->tm_flags = (uint8_t)scmnd->SCp.Message;
> + }
> +}
> +
> +
> +#define CSIO_SCSI_CMD_WR_SZ(_imm) \
> + (sizeof(struct fw_scsi_cmd_wr) + /* WR size */ \
> + ALIGN((_imm), 16)) /* Immed data */
> +
> +#define CSIO_SCSI_CMD_WR_SZ_16(_imm) \
> + (ALIGN(CSIO_SCSI_CMD_WR_SZ((_imm)), 16))
> +
> +/*
> + * csio_scsi_cmd - Create a SCSI CMD WR.
> + * @req: IO req structure.
> + *
> + * Gets a WR slot in the ingress queue and initializes it with SCSI CMD WR.
> + *
> + */
> +static inline void
> +csio_scsi_cmd(struct csio_ioreq *req)
> +{
> + struct csio_wr_pair wrp;
> + struct csio_hw *hw = req->lnode->hwp;
> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
> + uint32_t size = CSIO_SCSI_CMD_WR_SZ_16(scsim->proto_cmd_len);
> +
> + req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
> + if (unlikely(req->drv_status != CSIO_SUCCESS))
> + return;
> +
> + if (wrp.size1 >= size) {
> + /* Initialize WR in one shot */
> + csio_scsi_init_cmd_wr(req, wrp.addr1, size);
> + } else {
> + uint8_t tmpwr[512];
Mmmm, putting this large of a buffer on the local stack is probably not
a good idea.
This should become an allocation.. If it's a hot path then you'll
probably want to set this up before-hand.
> + /*
> + * Make a temporary copy of the WR and write back
> + * the copy into the WR pair.
> + */
> + csio_scsi_init_cmd_wr(req, (void *)tmpwr, size);
> + memcpy(wrp.addr1, tmpwr, wrp.size1);
> + memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
> + }
> +}
> +
> +/*
> + * The following is fast path code. Therefore it is inlined with multi-line
> + * macros using name substitution, thus avoiding if-else switches for
> + * operation (read/write), as well as serving the purpose of code re-use.
> + */
> +/*
> + * csio_scsi_init_ulptx_dsgl - Fill in a ULP_TX_SC_DSGL
> + * @hw: HW module
> + * @req: IO request
> + * @sgl: ULP TX SGL pointer.
> + *
> + */
> +#define csio_scsi_init_ultptx_dsgl(hw, req, sgl) \
> +do { \
> + struct ulptx_sge_pair *_sge_pair = NULL; \
> + struct scatterlist *_sgel; \
> + uint32_t _i = 0; \
> + uint32_t _xfer_len; \
> + struct list_head *_tmp; \
> + struct csio_dma_buf *_dma_buf; \
> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
> + \
> + (sgl)->cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) | ULPTX_MORE | \
> + ULPTX_NSGE((req)->nsge)); \
> + /* Now add the data SGLs */ \
> + if (likely(!(req)->dcopy)) { \
> + scsi_for_each_sg(scmnd, _sgel, (req)->nsge, _i) { \
> + if (_i == 0) { \
> + (sgl)->addr0 = cpu_to_be64( \
> + sg_dma_address(_sgel)); \
> + (sgl)->len0 = cpu_to_be32( \
> + sg_dma_len(_sgel)); \
> + _sge_pair = \
> + (struct ulptx_sge_pair *)((sgl) + 1); \
> + continue; \
> + } \
> + if ((_i - 1) & 0x1) { \
> + _sge_pair->addr[1] = cpu_to_be64( \
> + sg_dma_address(_sgel)); \
> + _sge_pair->len[1] = cpu_to_be32( \
> + sg_dma_len(_sgel)); \
> + _sge_pair++; \
> + } else { \
> + _sge_pair->addr[0] = cpu_to_be64( \
> + sg_dma_address(_sgel)); \
> + _sge_pair->len[0] = cpu_to_be32( \
> + sg_dma_len(_sgel)); \
> + } \
> + } \
> + } else { \
> + /* Program sg elements with driver's DDP buffer */ \
> + _xfer_len = scsi_bufflen(scmnd); \
> + list_for_each(_tmp, &(req)->gen_list) { \
> + _dma_buf = (struct csio_dma_buf *)_tmp; \
> + if (_i == 0) { \
> + (sgl)->addr0 = cpu_to_be64(_dma_buf->paddr); \
> + (sgl)->len0 = cpu_to_be32( \
> + min(_xfer_len, _dma_buf->len)); \
> + _sge_pair = \
> + (struct ulptx_sge_pair *)((sgl) + 1); \
> + } \
> + else if ((_i - 1) & 0x1) { \
> + _sge_pair->addr[1] = cpu_to_be64( \
> + _dma_buf->paddr); \
> + _sge_pair->len[1] = cpu_to_be32( \
> + min(_xfer_len, _dma_buf->len)); \
> + _sge_pair++; \
> + } else { \
> + _sge_pair->addr[0] = cpu_to_be64( \
> + _dma_buf->paddr); \
> + _sge_pair->len[0] = cpu_to_be32( \
> + min(_xfer_len, _dma_buf->len)); \
> + } \
> + _xfer_len -= min(_xfer_len, _dma_buf->len); \
> + _i++; \
> + } \
> + } \
> +} while (0)
> +
I don't see any reason why this can't just be a static function..? Why
is the macro usage necessary here..?
> +/*
> + * csio_scsi_init_data_wr - Initialize the READ/WRITE SCSI WR.
> + * @req: IO req structure.
> + * @oper: read/write
> + * @wrp: DMA location to place the payload.
> + * @size: Size of WR (including FW WR + immed data + rsp SG entry + data SGL
> + * @wrop: _READ_/_WRITE_
> + *
> + * Wrapper for populating fw_scsi_read_wr/fw_scsi_write_wr.
> + */
> +#define csio_scsi_init_data_wr(req, oper, wrp, size, wrop) \
> +do { \
> + struct csio_hw *_hw = (req)->lnode->hwp; \
> + struct csio_rnode *_rn = (req)->rnode; \
> + struct fw_scsi_##oper##_wr *__wr = (struct fw_scsi_##oper##_wr *)(wrp);\
> + struct ulptx_sgl *_sgl; \
> + struct csio_dma_buf *_dma_buf; \
> + uint8_t _imm = csio_hw_to_scsim(_hw)->proto_cmd_len; \
> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
> + \
> + __wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI##wrop##WR) | \
> + FW_SCSI##wrop##WR_IMMDLEN(_imm)); \
> + __wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(_rn->flowid) | \
> + FW_WR_LEN16( \
> + CSIO_ROUNDUP((size), 16))); \
> + __wr->cookie = (uintptr_t) (req); \
> + __wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(_hw, \
> + (req)->iq_idx));\
> + __wr->tmo_val = (uint8_t)((req)->tmo); \
> + __wr->use_xfer_cnt = 1; \
> + __wr->xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
> + __wr->ini_xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
> + /* Get RSP DMA buffer */ \
> + _dma_buf = &(req)->dma_buf; \
> + \
> + /* Prepare RSP SGL */ \
> + __wr->rsp_dmalen = cpu_to_be32(_dma_buf->len); \
> + __wr->rsp_dmaaddr = cpu_to_be64(_dma_buf->paddr); \
> + \
> + __wr->r4 = 0; \
> + \
> + __wr->u.fcoe.ctl_pri = 0; \
> + __wr->u.fcoe.cp_en_class = 0; \
> + __wr->u.fcoe.r3_lo[0] = 0; \
> + __wr->u.fcoe.r3_lo[1] = 0; \
> + csio_scsi_fcp_cmnd((req), (void *)((uintptr_t)(wrp) + \
> + sizeof(struct fw_scsi_##oper##_wr))); \
> + \
> + /* Move WR pointer past command and immediate data */ \
> + _sgl = (struct ulptx_sgl *) ((uintptr_t)(wrp) + \
> + sizeof(struct fw_scsi_##oper##_wr) + \
> + ALIGN(_imm, 16)); \
> + \
> + /* Fill in the DSGL */ \
> + csio_scsi_init_ultptx_dsgl(_hw, (req), _sgl); \
> + \
> +} while (0)
> +
This one has four uses of CPP keys. Just turn those into macros, and
leave the rest of the code in a static function.
> +/* Calculate WR size needed for fw_scsi_read_wr/fw_scsi_write_wr */
> +#define csio_scsi_data_wrsz(req, oper, sz, imm) \
> +do { \
> + (sz) = sizeof(struct fw_scsi_##oper##_wr) + /* WR size */ \
> + ALIGN((imm), 16) + /* Immed data */ \
> + sizeof(struct ulptx_sgl); /* ulptx_sgl */ \
> + \
> + if (unlikely((req)->nsge > 1)) \
> + (sz) += (sizeof(struct ulptx_sge_pair) * \
> + (ALIGN(((req)->nsge - 1), 2) / 2)); \
> + /* Data SGE */ \
> +} while (0)
> +
> +/*
> + * csio_scsi_data - Create a SCSI WRITE/READ WR.
> + * @req: IO req structure.
> + * @oper: read/write
> + * @wrop: _READ_/_WRITE_ (string subsitutions to use with the FW bit field
> + * macros).
> + *
> + * Gets a WR slot in the ingress queue and initializes it with
> + * SCSI CMD READ/WRITE WR.
> + *
> + */
> +#define csio_scsi_data(req, oper, wrop) \
> +do { \
> + struct csio_wr_pair _wrp; \
> + uint32_t _size; \
> + struct csio_hw *_hw = (req)->lnode->hwp; \
> + struct csio_scsim *_scsim = csio_hw_to_scsim(_hw); \
> + \
> + csio_scsi_data_wrsz((req), oper, _size, _scsim->proto_cmd_len); \
> + _size = ALIGN(_size, 16); \
> + \
> + (req)->drv_status = csio_wr_get(_hw, (req)->eq_idx, _size, &_wrp); \
> + if (likely((req)->drv_status == CSIO_SUCCESS)) { \
> + if (likely(_wrp.size1 >= _size)) { \
> + /* Initialize WR in one shot */ \
> + csio_scsi_init_data_wr((req), oper, _wrp.addr1, \
> + _size, wrop); \
> + } else { \
> + uint8_t tmpwr[512]; \
> + /* \
> + * Make a temporary copy of the WR and write back \
> + * the copy into the WR pair. \
> + */ \
> + csio_scsi_init_data_wr((req), oper, (void *)tmpwr, \
> + _size, wrop); \
> + memcpy(_wrp.addr1, tmpwr, _wrp.size1); \
> + memcpy(_wrp.addr2, tmpwr + _wrp.size1, \
> + _size - _wrp.size1); \
> + } \
> + } \
> +} while (0)
> +
Ditto on this one, along with the tmpwr[512] stack usage..
> +static inline void
> +csio_scsi_abrt_cls(struct csio_ioreq *req, bool abort)
> +{
> + struct csio_wr_pair wrp;
> + struct csio_hw *hw = req->lnode->hwp;
> + uint32_t size = ALIGN(sizeof(struct fw_scsi_abrt_cls_wr), 16);
> +
> + req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
> + if (req->drv_status != CSIO_SUCCESS)
> + return;
> +
> + if (wrp.size1 >= size) {
> + /* Initialize WR in one shot */
> + csio_scsi_init_abrt_cls_wr(req, wrp.addr1, size, abort);
> + } else {
> + uint8_t tmpwr[512];
Ditto here on local scope stack usage..
> + /*
> + * Make a temporary copy of the WR and write back
> + * the copy into the WR pair.
> + */
> + csio_scsi_init_abrt_cls_wr(req, (void *)tmpwr, size, abort);
> + memcpy(wrp.addr1, tmpwr, wrp.size1);
> + memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
> + }
> +}
> +
> +/*****************************************************************************/
> +/* START: SCSI SM */
> +/*****************************************************************************/
> +static void
> +csio_scsis_uninit(struct csio_ioreq *req, enum csio_scsi_ev evt)
> +{
> + struct csio_hw *hw = req->lnode->hwp;
> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
> +
> + switch (evt) {
> +
> + case CSIO_SCSIE_START_IO:
Extra space between start of first switch case
> +
> + /* There is data */
Point-less comment
> + if (req->nsge) {
> + if (req->datadir == CSIO_IOREQF_DMA_WRITE) {
> + req->dcopy = 0;
> + csio_scsi_data(req, write, _WRITE_);
> + } else
> + csio_setup_ddp(scsim, req);
> + } else {
> + csio_scsi_cmd(req);
> + }
> +
> + if (likely(req->drv_status == CSIO_SUCCESS)) {
> + /* change state and enqueue on active_q */
> + csio_set_state(&req->sm, csio_scsis_io_active);
> + list_add_tail(&req->sm.sm_list, &scsim->active_q);
> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
> + csio_inc_stats(scsim, n_active);
> +
> + return;
> + }
> + break;
> +
> + case CSIO_SCSIE_START_TM:
> + csio_scsi_cmd(req);
> + if (req->drv_status == CSIO_SUCCESS) {
> + /*
> + * NOTE: We collect the affected I/Os prior to issuing
> + * LUN reset, and not after it. This is to prevent
> + * aborting I/Os that get issued after the LUN reset,
> + * but prior to LUN reset completion (in the event that
> + * the host stack has not blocked I/Os to a LUN that is
> + * being reset.
> + */
> + csio_set_state(&req->sm, csio_scsis_tm_active);
> + list_add_tail(&req->sm.sm_list, &scsim->active_q);
> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
> + csio_inc_stats(scsim, n_tm_active);
> + }
> + return;
> +
> + case CSIO_SCSIE_ABORT:
> + case CSIO_SCSIE_CLOSE:
> + /*
> + * NOTE:
> + * We could get here due to :
> + * - a window in the cleanup path of the SCSI module
> + * (csio_scsi_abort_io()). Please see NOTE in this function.
> + * - a window in the time we tried to issue an abort/close
> + * of a request to FW, and the FW completed the request
> + * itself.
> + * Print a message for now, and return INVAL either way.
> + */
> + req->drv_status = CSIO_INVAL;
> + csio_warn(hw, "Trying to abort/close completed IO:%p!\n", req);
> + break;
> +
> + default:
> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
> + CSIO_DB_ASSERT(0);
> + }
> +}
> +
> +static void
> +csio_scsis_io_active(struct csio_ioreq *req, enum csio_scsi_ev evt)
> +{
> + struct csio_hw *hw = req->lnode->hwp;
> + struct csio_scsim *scm = csio_hw_to_scsim(hw);
> + struct csio_rnode *rn;
> +
> + switch (evt) {
> +
> + case CSIO_SCSIE_COMPLETED:
Ditto
> + csio_dec_stats(scm, n_active);
> + list_del_init(&req->sm.sm_list);
> + csio_set_state(&req->sm, csio_scsis_uninit);
> + /*
> + * In MSIX mode, with multiple queues, the SCSI compeltions
> + * could reach us sooner than the FW events sent to indicate
> + * I-T nexus loss (link down, remote device logo etc). We
> + * dont want to be returning such I/Os to the upper layer
> + * immediately, since we wouldnt have reported the I-T nexus
> + * loss itself. This forces us to serialize such completions
> + * with the reporting of the I-T nexus loss. Therefore, we
> + * internally queue up such up such completions in the rnode.
> + * The reporting of I-T nexus loss to the upper layer is then
> + * followed by the returning of I/Os in this internal queue.
> + * Having another state alongwith another queue helps us take
> + * actions for events such as ABORT received while we are
> + * in this rnode queue.
> + */
> + if (unlikely(req->wr_status != FW_SUCCESS)) {
> + rn = req->rnode;
> + /*
> + * FW says remote device is lost, but rnode
> + * doesnt reflect it.
> + */
> + if (csio_scsi_itnexus_loss_error(req->wr_status) &&
> + csio_is_rnode_ready(rn)) {
> + csio_set_state(&req->sm,
> + csio_scsis_shost_cmpl_await);
> + list_add_tail(&req->sm.sm_list,
> + &rn->host_cmpl_q);
> + }
> + }
> +
> + break;
> +
> + case CSIO_SCSIE_ABORT:
> + csio_scsi_abrt_cls(req, SCSI_ABORT);
> + if (req->drv_status == CSIO_SUCCESS) {
> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
> + csio_set_state(&req->sm, csio_scsis_aborting);
> + }
> + break;
> +
> + case CSIO_SCSIE_CLOSE:
> + csio_scsi_abrt_cls(req, SCSI_CLOSE);
> + if (req->drv_status == CSIO_SUCCESS) {
> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
> + csio_set_state(&req->sm, csio_scsis_closing);
> + }
> + break;
> +
> + case CSIO_SCSIE_DRVCLEANUP:
> + req->wr_status = FW_HOSTERROR;
> + csio_dec_stats(scm, n_active);
> + csio_set_state(&req->sm, csio_scsis_uninit);
> + break;
> +
> + default:
> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
> + CSIO_DB_ASSERT(0);
> + }
> +}
> +
> +static void
> +csio_scsis_tm_active(struct csio_ioreq *req, enum csio_scsi_ev evt)
> +{
> + struct csio_hw *hw = req->lnode->hwp;
> + struct csio_scsim *scm = csio_hw_to_scsim(hw);
> +
> + switch (evt) {
> +
> + case CSIO_SCSIE_COMPLETED:
Ditto
> + csio_dec_stats(scm, n_tm_active);
> + list_del_init(&req->sm.sm_list);
> + csio_set_state(&req->sm, csio_scsis_uninit);
> +
> + break;
> +
> + case CSIO_SCSIE_ABORT:
> + csio_scsi_abrt_cls(req, SCSI_ABORT);
> + if (req->drv_status == CSIO_SUCCESS) {
> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
> + csio_set_state(&req->sm, csio_scsis_aborting);
> + }
> + break;
> +
> +
> + case CSIO_SCSIE_CLOSE:
> + csio_scsi_abrt_cls(req, SCSI_CLOSE);
> + if (req->drv_status == CSIO_SUCCESS) {
> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
> + csio_set_state(&req->sm, csio_scsis_closing);
> + }
> + break;
> +
> + case CSIO_SCSIE_DRVCLEANUP:
> + req->wr_status = FW_HOSTERROR;
> + csio_dec_stats(scm, n_tm_active);
> + csio_set_state(&req->sm, csio_scsis_uninit);
> + break;
> +
> + default:
> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
> + CSIO_DB_ASSERT(0);
> + }
> +}
> +
> +static void
> +csio_scsis_aborting(struct csio_ioreq *req, enum csio_scsi_ev evt)
> +{
> + struct csio_hw *hw = req->lnode->hwp;
> + struct csio_scsim *scm = csio_hw_to_scsim(hw);
> +
> + switch (evt) {
> +
> + case CSIO_SCSIE_COMPLETED:
Ditto
> + csio_dbg(hw,
> + "ioreq %p recvd cmpltd (wr_status:%d) "
> + "in aborting st\n", req, req->wr_status);
> + /*
> + * Use CSIO_CANCELLED to explicitly tell the ABORTED event that
> + * the original I/O was returned to driver by FW.
> + * We dont really care if the I/O was returned with success by
> + * FW (because the ABORT and completion of the I/O crossed each
> + * other), or any other return value. Once we are in aborting
> + * state, the success or failure of the I/O is unimportant to
> + * us.
> + */
> + req->drv_status = CSIO_CANCELLED;
> + break;
> +
> + case CSIO_SCSIE_ABORT:
> + csio_inc_stats(scm, n_abrt_dups);
> + break;
> +
> + case CSIO_SCSIE_ABORTED:
> +
> + csio_dbg(hw, "abort of %p return status:0x%x drv_status:%x\n",
> + req, req->wr_status, req->drv_status);
> + /*
> + * Check if original I/O WR completed before the Abort
> + * completion.
> + */
> + if (req->drv_status != CSIO_CANCELLED) {
> + csio_fatal(hw,
> + "Abort completed before original I/O,"
> + " req:%p\n", req);
> + CSIO_DB_ASSERT(0);
> + }
> +
> + /*
> + * There are the following possible scenarios:
> + * 1. The abort completed successfully, FW returned FW_SUCCESS.
> + * 2. The completion of an I/O and the receipt of
> + * abort for that I/O by the FW crossed each other.
> + * The FW returned FW_EINVAL. The original I/O would have
> + * returned with FW_SUCCESS or any other SCSI error.
> + * 3. The FW couldnt sent the abort out on the wire, as there
> + * was an I-T nexus loss (link down, remote device logged
> + * out etc). FW sent back an appropriate IT nexus loss status
> + * for the abort.
> + * 4. FW sent an abort, but abort timed out (remote device
> + * didnt respond). FW replied back with
> + * FW_SCSI_ABORT_TIMEDOUT.
> + * 5. FW couldnt genuinely abort the request for some reason,
> + * and sent us an error.
> + *
> + * The first 3 scenarios are treated as succesful abort
> + * operations by the host, while the last 2 are failed attempts
> + * to abort. Manipulate the return value of the request
> + * appropriately, so that host can convey these results
> + * back to the upper layer.
> + */
> + if ((req->wr_status == FW_SUCCESS) ||
> + (req->wr_status == FW_EINVAL) ||
> + csio_scsi_itnexus_loss_error(req->wr_status))
> + req->wr_status = FW_SCSI_ABORT_REQUESTED;
> +
> + csio_dec_stats(scm, n_active);
> + list_del_init(&req->sm.sm_list);
> + csio_set_state(&req->sm, csio_scsis_uninit);
> + break;
> +
> + case CSIO_SCSIE_DRVCLEANUP:
> + req->wr_status = FW_HOSTERROR;
> + csio_dec_stats(scm, n_active);
> + csio_set_state(&req->sm, csio_scsis_uninit);
> + break;
> +
> + case CSIO_SCSIE_CLOSE:
> + /*
> + * We can receive this event from the module
> + * cleanup paths, if the FW forgot to reply to the ABORT WR
> + * and left this ioreq in this state. For now, just ignore
> + * the event. The CLOSE event is sent to this state, as
> + * the LINK may have already gone down.
> + */
> + break;
> +
> + default:
> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
> + CSIO_DB_ASSERT(0);
> + }
> +}
> +
> +static void
> +csio_scsis_closing(struct csio_ioreq *req, enum csio_scsi_ev evt)
> +{
> + struct csio_hw *hw = req->lnode->hwp;
> + struct csio_scsim *scm = csio_hw_to_scsim(hw);
> +
> + switch (evt) {
> +
> + case CSIO_SCSIE_COMPLETED:
Ditto
> + csio_dbg(hw,
> + "ioreq %p recvd cmpltd (wr_status:%d) "
> + "in closing st\n", req, req->wr_status);
> + /*
> + * Use CSIO_CANCELLED to explicitly tell the CLOSED event that
> + * the original I/O was returned to driver by FW.
> + * We dont really care if the I/O was returned with success by
> + * FW (because the CLOSE and completion of the I/O crossed each
> + * other), or any other return value. Once we are in aborting
> + * state, the success or failure of the I/O is unimportant to
> + * us.
> + */
> + req->drv_status = CSIO_CANCELLED;
> + break;
> +
> + case CSIO_SCSIE_CLOSED:
> + /*
> + * Check if original I/O WR completed before the Close
> + * completion.
> + */
> + if (req->drv_status != CSIO_CANCELLED) {
> + csio_fatal(hw,
> + "Close completed before original I/O,"
> + " req:%p\n", req);
> + CSIO_DB_ASSERT(0);
> + }
> +
> + /*
> + * Either close succeeded, or we issued close to FW at the
> + * same time FW compelted it to us. Either way, the I/O
> + * is closed.
> + */
> + CSIO_DB_ASSERT((req->wr_status == FW_SUCCESS) ||
> + (req->wr_status == FW_EINVAL));
> + req->wr_status = FW_SCSI_CLOSE_REQUESTED;
> +
> + csio_dec_stats(scm, n_active);
> + list_del_init(&req->sm.sm_list);
> + csio_set_state(&req->sm, csio_scsis_uninit);
> + break;
> +
> + case CSIO_SCSIE_CLOSE:
> + break;
> +
> + case CSIO_SCSIE_DRVCLEANUP:
> + req->wr_status = FW_HOSTERROR;
> + csio_dec_stats(scm, n_active);
> + csio_set_state(&req->sm, csio_scsis_uninit);
> + break;
> +
> + default:
> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
> + CSIO_DB_ASSERT(0);
> + }
> +}
> +
> +static void
> +csio_scsis_shost_cmpl_await(struct csio_ioreq *req, enum csio_scsi_ev evt)
> +{
> + switch (evt) {
> + case CSIO_SCSIE_ABORT:
> + case CSIO_SCSIE_CLOSE:
> + /*
> + * Just succeed the abort request, and hope that
> + * the remote device unregister path will cleanup
> + * this I/O to the upper layer within a sane
> + * amount of time.
> + */
> + /*
> + * A close can come in during a LINK DOWN. The FW would have
> + * returned us the I/O back, but not the remote device lost
> + * FW event. In this interval, if the I/O times out at the upper
> + * layer, a close can come in. Take the same action as abort:
> + * return success, and hope that the remote device unregister
> + * path will cleanup this I/O. If the FW still doesnt send
> + * the msg, the close times out, and the upper layer resorts
> + * to the next level of error recovery.
> + */
> + req->drv_status = CSIO_SUCCESS;
> + break;
> + case CSIO_SCSIE_DRVCLEANUP:
> + csio_set_state(&req->sm, csio_scsis_uninit);
> + break;
> + default:
> + csio_dbg(req->lnode->hwp, "Unhandled event:%d sent to req:%p\n",
> + evt, req);
> + CSIO_DB_ASSERT(0);
> + }
> +}
> +
> +
> +/**
> + * csio_queuecommand_lck - Entry point to kickstart an I/O request.
> + * @cmnd: The I/O request from ML.
> + * @done: The ML callback routine.
> + *
> + * This routine does the following:
> + * - Checks for HW and Rnode module readiness.
> + * - Gets a free ioreq structure (which is already initialized
> + * to uninit during its allocation).
> + * - Maps SG elements.
> + * - Initializes ioreq members.
> + * - Kicks off the SCSI state machine for this IO.
> + * - Returns busy status on error.
> + */
> +static int
> +csio_queuecommand_lck(struct scsi_cmnd *cmnd, void (*done)(struct scsi_cmnd *))
> +{
> + struct csio_lnode *ln = shost_priv(cmnd->device->host);
> + struct csio_hw *hw = csio_lnode_to_hw(ln);
> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
> + struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
> + struct csio_ioreq *ioreq = NULL;
> + unsigned long flags;
> + int nsge = 0;
> + int rv = SCSI_MLQUEUE_HOST_BUSY, nr;
> + csio_retval_t retval;
> + int cpu;
> + struct csio_scsi_qset *sqset;
> + struct fc_rport *rport = starget_to_rport(scsi_target(cmnd->device));
> +
> + if (!blk_rq_cpu_valid(cmnd->request))
> + cpu = smp_processor_id();
> + else
> + cpu = cmnd->request->cpu;
> +
> + sqset = &hw->sqset[ln->portid][cpu];
> +
> + nr = fc_remote_port_chkready(rport);
> + if (nr) {
> + cmnd->result = nr;
> + csio_inc_stats(scsim, n_rn_nr_error);
> + goto err_done;
> + }
> +
> + if (unlikely(!csio_is_hw_ready(hw))) {
> + cmnd->result = (DID_REQUEUE << 16);
> + csio_inc_stats(scsim, n_hw_nr_error);
> + goto err_done;
> + }
> +
> + /* Get req->nsge, if there are SG elements to be mapped */
> + nsge = scsi_dma_map(cmnd);
> + if (unlikely(nsge < 0)) {
> + csio_inc_stats(scsim, n_dmamap_error);
> + goto err;
> + }
> +
> + /* Do we support so many mappings? */
> + if (unlikely(nsge > scsim->max_sge)) {
> + csio_warn(hw,
> + "More SGEs than can be supported."
> + " SGEs: %d, Max SGEs: %d\n", nsge, scsim->max_sge);
> + csio_inc_stats(scsim, n_unsupp_sge_error);
> + goto err_dma_unmap;
> + }
> +
> + /* Get a free ioreq structure - SM is already set to uninit */
> + ioreq = csio_get_scsi_ioreq_lock(hw, scsim);
> + if (!ioreq) {
> + csio_err(hw, "Out of I/O request elements. Active #:%d\n",
> + scsim->stats.n_active);
> + csio_inc_stats(scsim, n_no_req_error);
> + goto err_dma_unmap;
> + }
> +
> + ioreq->nsge = nsge;
> + ioreq->lnode = ln;
> + ioreq->rnode = rn;
> + ioreq->iq_idx = sqset->iq_idx;
> + ioreq->eq_idx = sqset->eq_idx;
> + ioreq->wr_status = 0;
> + ioreq->drv_status = CSIO_SUCCESS;
> + csio_scsi_cmnd(ioreq) = (void *)cmnd;
> + ioreq->tmo = 0;
> +
> + switch (cmnd->sc_data_direction) {
> + case DMA_BIDIRECTIONAL:
> + ioreq->datadir = CSIO_IOREQF_DMA_BIDI;
> + csio_inc_stats(ln, n_control_requests);
> + break;
> + case DMA_TO_DEVICE:
> + ioreq->datadir = CSIO_IOREQF_DMA_WRITE;
> + csio_inc_stats(ln, n_output_requests);
> + ln->stats.n_output_bytes += scsi_bufflen(cmnd);
> + break;
> + case DMA_FROM_DEVICE:
> + ioreq->datadir = CSIO_IOREQF_DMA_READ;
> + csio_inc_stats(ln, n_input_requests);
> + ln->stats.n_input_bytes += scsi_bufflen(cmnd);
> + break;
> + case DMA_NONE:
> + ioreq->datadir = CSIO_IOREQF_DMA_NONE;
> + csio_inc_stats(ln, n_control_requests);
> + break;
> + default:
> + CSIO_DB_ASSERT(0);
> + break;
> + }
> +
> + /* Set cbfn */
> + ioreq->io_cbfn = csio_scsi_cbfn;
> +
> + /* Needed during abort */
> + cmnd->host_scribble = (unsigned char *)ioreq;
> + cmnd->scsi_done = done;
> + cmnd->SCp.Message = 0;
> +
> + /* Kick off SCSI IO SM on the ioreq */
> + spin_lock_irqsave(&hw->lock, flags);
> + retval = csio_scsi_start_io(ioreq);
> + spin_unlock_irqrestore(&hw->lock, flags);
> +
> + if (retval != CSIO_SUCCESS) {
> + csio_err(hw, "ioreq: %p couldnt be started, status:%d\n",
> + ioreq, retval);
> + csio_inc_stats(scsim, n_busy_error);
> + goto err_put_req;
> + }
> +
> + return 0;
> +
> +err_put_req:
> + csio_put_scsi_ioreq_lock(hw, scsim, ioreq);
> +err_dma_unmap:
> + if (nsge > 0)
> + scsi_dma_unmap(cmnd);
> +err:
> + return rv;
> +
> +err_done:
> + done(cmnd);
> + return 0;
> +}
> +
> +static DEF_SCSI_QCMD(csio_queuecommand);
> +
This means that your running with the host_lock held.. I'm not sure if
that is really what you want to do as it really end's up killing
multi-lun small packet performance..
How about dropping DEF_SCSI_QCMD usage here, and figure out what
actually needs to be protected by the SCSI host_lock within
csio_queuecommand_lck()..?
> +static csio_retval_t
> +csio_do_abrt_cls(struct csio_hw *hw, struct csio_ioreq *ioreq, bool abort)
> +{
> + csio_retval_t rv;
> + int cpu = smp_processor_id();
> + struct csio_lnode *ln = ioreq->lnode;
> + struct csio_scsi_qset *sqset = &hw->sqset[ln->portid][cpu];
> +
> + ioreq->tmo = CSIO_SCSI_ABRT_TMO_MS;
> + /*
> + * Use current processor queue for posting the abort/close, but retain
> + * the ingress queue ID of the original I/O being aborted/closed - we
> + * need the abort/close completion to be received on the same queue
> + * as the original I/O.
> + */
> + ioreq->eq_idx = sqset->eq_idx;
> +
> + if (abort == SCSI_ABORT)
> + rv = csio_scsi_abort(ioreq);
> + else /* close */
Point-less comment.
> + rv = csio_scsi_close(ioreq);
> +
> + return rv;
> +}
> +
> +static int
> +csio_eh_abort_handler(struct scsi_cmnd *cmnd)
> +{
> + struct csio_ioreq *ioreq;
> + struct csio_lnode *ln = shost_priv(cmnd->device->host);
> + struct csio_hw *hw = csio_lnode_to_hw(ln);
> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
> + int ready = 0, ret;
> + unsigned long tmo = 0;
> + csio_retval_t rv;
> + struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
> +
> + ret = fc_block_scsi_eh(cmnd);
> + if (ret)
> + return ret;
> +
> + ioreq = (struct csio_ioreq *)cmnd->host_scribble;
> + if (!ioreq)
> + return SUCCESS;
> +
> + if (!rn)
> + return FAILED;
> +
> + csio_dbg(hw,
> + "Request to abort ioreq:%p cmd:%p cdb:%08llx"
> + " ssni:0x%x lun:%d iq:0x%x\n",
> + ioreq, cmnd, *((uint64_t *)cmnd->cmnd), rn->flowid,
> + cmnd->device->lun, csio_q_physiqid(hw, ioreq->iq_idx));
> +
> + if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) != cmnd) {
> + csio_inc_stats(scsim, n_abrt_race_comp);
> + return SUCCESS;
> + }
> +
> + ready = csio_is_lnode_ready(ln);
> + tmo = CSIO_SCSI_ABRT_TMO_MS;
> +
> + spin_lock_irq(&hw->lock);
> + rv = csio_do_abrt_cls(hw, ioreq, (ready ? SCSI_ABORT : SCSI_CLOSE));
> + spin_unlock_irq(&hw->lock);
> +
> + if (rv != CSIO_SUCCESS) {
> + if (rv == CSIO_INVAL) {
> + /* Return success, if abort/close request issued on
> + * already completed IO
> + */
> + return SUCCESS;
> + }
> + if (ready)
> + csio_inc_stats(scsim, n_abrt_busy_error);
> + else
> + csio_inc_stats(scsim, n_cls_busy_error);
> +
> + goto inval_scmnd;
> + }
> +
> + /* Wait for completion */
> + init_completion(&ioreq->cmplobj);
> + wait_for_completion_timeout(&ioreq->cmplobj, msecs_to_jiffies(tmo));
> +
> + /* FW didnt respond to abort within our timeout */
> + if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) == cmnd) {
> +
> + csio_err(hw, "Abort timed out -- req: %p\n", ioreq);
> + csio_inc_stats(scsim, n_abrt_timedout);
> +
> +inval_scmnd:
> + if (ioreq->nsge > 0)
> + scsi_dma_unmap(cmnd);
> +
> + spin_lock_irq(&hw->lock);
> + csio_scsi_cmnd(ioreq) = NULL;
> + spin_unlock_irq(&hw->lock);
> +
> + cmnd->result = (DID_ERROR << 16);
> + cmnd->scsi_done(cmnd);
> +
> + return FAILED;
> + }
> +
> + /* FW successfully aborted the request */
> + if (host_byte(cmnd->result) == DID_REQUEUE) {
> + csio_info(hw,
> + "Aborted SCSI command to (%d:%d) serial#:0x%lx\n",
> + cmnd->device->id, cmnd->device->lun,
> + cmnd->serial_number);
> + return SUCCESS;
> + } else {
> + csio_info(hw,
> + "Failed to abort SCSI command, (%d:%d) serial#:0x%lx\n",
> + cmnd->device->id, cmnd->device->lun,
> + cmnd->serial_number);
> + return FAILED;
> + }
> +}
> +
> +struct scsi_host_template csio_fcoe_shost_template = {
> + .module = THIS_MODULE,
> + .name = CSIO_DRV_DESC,
> + .proc_name = KBUILD_MODNAME,
> + .queuecommand = csio_queuecommand,
> + .eh_abort_handler = csio_eh_abort_handler,
> + .eh_device_reset_handler = csio_eh_lun_reset_handler,
> + .slave_alloc = csio_slave_alloc,
> + .slave_configure = csio_slave_configure,
> + .slave_destroy = csio_slave_destroy,
> + .scan_finished = csio_scan_finished,
> + .this_id = -1,
> + .sg_tablesize = CSIO_SCSI_MAX_SGE,
> + .cmd_per_lun = CSIO_MAX_CMD_PER_LUN,
> + .use_clustering = ENABLE_CLUSTERING,
> + .shost_attrs = csio_fcoe_lport_attrs,
> + .max_sectors = CSIO_MAX_SECTOR_SIZE,
> +};
> +
> +struct scsi_host_template csio_fcoe_shost_vport_template = {
> + .module = THIS_MODULE,
> + .name = CSIO_DRV_DESC,
> + .proc_name = KBUILD_MODNAME,
> + .queuecommand = csio_queuecommand,
> + .eh_abort_handler = csio_eh_abort_handler,
> + .eh_device_reset_handler = csio_eh_lun_reset_handler,
> + .slave_alloc = csio_slave_alloc,
> + .slave_configure = csio_slave_configure,
> + .slave_destroy = csio_slave_destroy,
> + .scan_finished = csio_scan_finished,
> + .this_id = -1,
> + .sg_tablesize = CSIO_SCSI_MAX_SGE,
> + .cmd_per_lun = CSIO_MAX_CMD_PER_LUN,
> + .use_clustering = ENABLE_CLUSTERING,
> + .shost_attrs = csio_fcoe_vport_attrs,
> + .max_sectors = CSIO_MAX_SECTOR_SIZE,
> +};
> +
> +/*
> + * csio_scsi_alloc_ddp_bufs - Allocate buffers for DDP of unaligned SGLs.
> + * @scm: SCSI Module
> + * @hw: HW device.
> + * @buf_size: buffer size
> + * @num_buf : Number of buffers.
> + *
> + * This routine allocates DMA buffers required for SCSI Data xfer, if
> + * each SGL buffer for a SCSI Read request posted by SCSI midlayer are
> + * not virtually contiguous.
> + */
> +static csio_retval_t
> +csio_scsi_alloc_ddp_bufs(struct csio_scsim *scm, struct csio_hw *hw,
> + int buf_size, int num_buf)
> +{
> + int n = 0;
> + struct list_head *tmp;
> + struct csio_dma_buf *ddp_desc = NULL;
> + uint32_t unit_size = 0;
> +
> + if (!num_buf)
> + return CSIO_SUCCESS;
Just return 0..?
> +
> + if (!buf_size)
> + return CSIO_INVAL;
Just return -EINVAL..?
> +
> + INIT_LIST_HEAD(&scm->ddp_freelist);
> +
> + /* Align buf size to page size */
> + buf_size = (buf_size + PAGE_SIZE - 1) & PAGE_MASK;
> + /* Initialize dma descriptors */
> + for (n = 0; n < num_buf; n++) {
> + /* Set unit size to request size */
> + unit_size = buf_size;
> + ddp_desc = kzalloc(sizeof(struct csio_dma_buf), GFP_KERNEL);
> + if (!ddp_desc) {
> + csio_err(hw,
> + "Failed to allocate ddp descriptors,"
> + " Num allocated = %d.\n",
> + scm->stats.n_free_ddp);
> + goto no_mem;
> + }
> +
> + /* Allocate Dma buffers for DDP */
> + ddp_desc->vaddr = pci_alloc_consistent(hw->pdev, unit_size,
> + &ddp_desc->paddr);
> + if (!ddp_desc->vaddr) {
> + csio_err(hw,
> + "SCSI response DMA buffer (ddp) allocation"
> + " failed!\n");
> + kfree(ddp_desc);
> + goto no_mem;
> + }
> +
> + ddp_desc->len = unit_size;
> +
> + /* Added it to scsi ddp freelist */
> + list_add_tail(&ddp_desc->list, &scm->ddp_freelist);
> + csio_inc_stats(scm, n_free_ddp);
> + }
> +
> + return CSIO_SUCCESS;
Just return 0..?
> +no_mem:
> + /* release dma descs back to freelist and free dma memory */
> + list_for_each(tmp, &scm->ddp_freelist) {
> + ddp_desc = (struct csio_dma_buf *) tmp;
> + tmp = csio_list_prev(tmp);
> + pci_free_consistent(hw->pdev, ddp_desc->len, ddp_desc->vaddr,
> + ddp_desc->paddr);
> + list_del_init(&ddp_desc->list);
> + kfree(ddp_desc);
> + }
> + scm->stats.n_free_ddp = 0;
> +
> + return CSIO_NOMEM;
This should be just -ENOMEM..?
> +}
> +
> +/*
> + * csio_scsi_free_ddp_bufs - free DDP buffers of unaligned SGLs.
> + * @scm: SCSI Module
> + * @hw: HW device.
> + *
> + * This routine frees ddp buffers.
> + */
> +static csio_retval_t
> +csio_scsi_free_ddp_bufs(struct csio_scsim *scm, struct csio_hw *hw)
> +{
> + struct list_head *tmp;
> + struct csio_dma_buf *ddp_desc;
> +
> + /* release dma descs back to freelist and free dma memory */
> + list_for_each(tmp, &scm->ddp_freelist) {
> + ddp_desc = (struct csio_dma_buf *) tmp;
> + tmp = csio_list_prev(tmp);
> + pci_free_consistent(hw->pdev, ddp_desc->len, ddp_desc->vaddr,
> + ddp_desc->paddr);
> + list_del_init(&ddp_desc->list);
> + kfree(ddp_desc);
> + }
> + scm->stats.n_free_ddp = 0;
> +
> + return CSIO_NOMEM;
> +}
Ditto.. Just -ENOMEM..?
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5).
2012-08-23 19:48 ` Nicholas A. Bellinger
@ 2012-08-24 17:40 ` Naresh Kumar Inna
2012-08-24 18:27 ` Joe Perches
2012-08-24 20:56 ` Nicholas A. Bellinger
0 siblings, 2 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-24 17:40 UTC (permalink / raw)
To: Nicholas A. Bellinger
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/24/2012 1:18 AM, Nicholas A. Bellinger wrote:
> On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
>> This patch contains code to implement the interrupt handling and the fast
>> path I/O functionality. The interrupt handling includes allocation of
>> MSIX vectors, registering and implemeting the interrupt service routines.
>> The fast path I/O functionality includes posting the I/O request to firmware
>> via Work Requests, tracking/completing them, and handling task management
>> requests. SCSI midlayer host template implementation is also covered by
>> this patch.
>>
>> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
>> ---
>
> Hi Naresh,
>
> My review comments are inline below..
Hi Nicholas,
Thanks for taking the time to review the driver. Please find my replies
inline.
Regards,
Naresh.
>
>> drivers/scsi/csiostor/csio_isr.c | 631 ++++++++++
>> drivers/scsi/csiostor/csio_scsi.c | 2498 +++++++++++++++++++++++++++++++++++++
>> 2 files changed, 3129 insertions(+), 0 deletions(-)
>> create mode 100644 drivers/scsi/csiostor/csio_isr.c
>> create mode 100644 drivers/scsi/csiostor/csio_scsi.c
>>
>> diff --git a/drivers/scsi/csiostor/csio_isr.c b/drivers/scsi/csiostor/csio_isr.c
>> new file mode 100644
>> index 0000000..96633e9
>> --- /dev/null
>> +++ b/drivers/scsi/csiostor/csio_isr.c
>
> <SNIP>
>
>> +#define csio_extra_msix_desc(_desc, _len, _str, _arg1, _arg2, _arg3) \
>> +do { \
>> + memset((_desc), 0, (_len) + 1); \
>> + snprintf((_desc), (_len), (_str), (_arg1), (_arg2), (_arg3)); \
>> +} while (0)
>> +
>
> This type of macro usage is not necessary for just two users below.
> Please inline code like this.
>
OK, I will get rid of this macro.
>> +static void
>> +csio_add_msix_desc(struct csio_hw *hw)
>> +{
>> + int i;
>> + struct csio_msix_entries *entryp = &hw->msix_entries[0];
>> + int k = CSIO_EXTRA_VECS;
>> + int len = sizeof(entryp->desc) - 1;
>> + int cnt = hw->num_sqsets + k;
>> +
>> + /* Non-data vector */
>> + csio_extra_msix_desc(entryp->desc, len, "csio-%02x:%02x:%x-nondata",
>> + CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
>> + CSIO_PCI_FUNC(hw));
>> + entryp++;
>> + csio_extra_msix_desc(entryp->desc, len, "csio-%02x:%02x:%x-fwevt",
>> + CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
>> + CSIO_PCI_FUNC(hw));
>> + entryp++;
>> +
>> + /* Name SCSI vecs */
>> + for (i = k; i < cnt; i++, entryp++) {
>> + memset(entryp->desc, 0, len + 1);
>> + snprintf(entryp->desc, len, "csio-%02x:%02x:%x-scsi%d",
>> + CSIO_PCI_BUS(hw), CSIO_PCI_DEV(hw),
>> + CSIO_PCI_FUNC(hw), i - CSIO_EXTRA_VECS);
>> + }
>> +}
>> +
>
>
>> diff --git a/drivers/scsi/csiostor/csio_scsi.c b/drivers/scsi/csiostor/csio_scsi.c
>> new file mode 100644
>> index 0000000..0f87b00
>> --- /dev/null
>> +++ b/drivers/scsi/csiostor/csio_scsi.c
>
>> +
>> +/*
>> + * csio_scsi_match_io - Match an ioreq with the given SCSI level data.
>> + * @ioreq: The I/O request
>> + * @sld: Level information
>> + *
>> + * Should be called with lock held.
>> + *
>> + */
>> +static bool
>> +csio_scsi_match_io(struct csio_ioreq *ioreq, struct csio_scsi_level_data *sld)
>> +{
>> + struct scsi_cmnd *scmnd = csio_scsi_cmnd(ioreq);
>> +
>> + switch (sld->level) {
>> + case CSIO_LEV_LUN:
>> + if (scmnd == NULL)
>> + return CSIO_FALSE;
>> +
>> + return ((ioreq->lnode == sld->lnode) &&
>> + (ioreq->rnode == sld->rnode) &&
>> + ((uint64_t)scmnd->device->lun == sld->oslun));
>> +
>> + case CSIO_LEV_RNODE:
>> + return ((ioreq->lnode == sld->lnode) &&
>> + (ioreq->rnode == sld->rnode));
>> + case CSIO_LEV_LNODE:
>> + return (ioreq->lnode == sld->lnode);
>> + case CSIO_LEV_ALL:
>> + return CSIO_TRUE;
>> + default:
>> + return CSIO_FALSE;
>> + }
>> +}
>> +
>
> Why can't CSIO_[TRUE,FALSE] just use normal Boolean defines..?
>
Is there a TRUE/FALSE define in a standard header file? I see a lot of
kernel code defining their own TRUE/FALSE.
>> +/*
>> + * csio_scsi_fcp_cmnd - Frame the SCSI FCP command paylod.
>> + * @req: IO req structure.
>> + * @addr: DMA location to place the payload.
>> + *
>> + * This routine is shared between FCP_WRITE, FCP_READ and FCP_CMD requests.
>> + */
>> +static inline void
>> +csio_scsi_fcp_cmnd(struct csio_ioreq *req, void *addr)
>> +{
>> + struct csio_fcp_cmnd *fcp_cmnd = (struct csio_fcp_cmnd *)addr;
>> + struct scsi_cmnd *scmnd = csio_scsi_cmnd(req);
>> +
>> + /* Check for Task Management */
>> + if (likely(scmnd->SCp.Message == 0)) {
>> + int_to_scsilun(scmnd->device->lun,
>> + (struct scsi_lun *)fcp_cmnd->lun);
>> + fcp_cmnd->tm_flags = 0;
>> + fcp_cmnd->cmdref = 0;
>> + fcp_cmnd->pri_ta = 0;
>> +
>> + memcpy(fcp_cmnd->cdb, scmnd->cmnd, 16);
>> + csio_scsi_tag(scmnd, &fcp_cmnd->pri_ta,
>> + FCP_PTA_HEADQ, FCP_PTA_ORDERED, FCP_PTA_SIMPLE);
>> + fcp_cmnd->dl = cpu_to_be32(scsi_bufflen(scmnd));
>> +
>> + if (req->nsge)
>> + if (req->datadir == CSIO_IOREQF_DMA_WRITE)
>
> The same goes for CSIO_IOREQF_DMA_*...
>
> Why can't this just be DMA_* defs from include/linux/dma-direction.h..?
OK, I will get rid of them and use the standard defines.
>
>> + fcp_cmnd->flags = FCP_CFL_WRDATA;
>> + else
>> + fcp_cmnd->flags = FCP_CFL_RDDATA;
>> + else
>> + fcp_cmnd->flags = 0;
>> + } else {
>> + memset(fcp_cmnd, 0, sizeof(*fcp_cmnd));
>> + int_to_scsilun(scmnd->device->lun,
>> + (struct scsi_lun *)fcp_cmnd->lun);
>> + fcp_cmnd->tm_flags = (uint8_t)scmnd->SCp.Message;
>> + }
>> +}
>> +
>
>
>
>> +
>> +#define CSIO_SCSI_CMD_WR_SZ(_imm) \
>> + (sizeof(struct fw_scsi_cmd_wr) + /* WR size */ \
>> + ALIGN((_imm), 16)) /* Immed data */
>> +
>> +#define CSIO_SCSI_CMD_WR_SZ_16(_imm) \
>> + (ALIGN(CSIO_SCSI_CMD_WR_SZ((_imm)), 16))
>> +
>> +/*
>> + * csio_scsi_cmd - Create a SCSI CMD WR.
>> + * @req: IO req structure.
>> + *
>> + * Gets a WR slot in the ingress queue and initializes it with SCSI CMD WR.
>> + *
>> + */
>> +static inline void
>> +csio_scsi_cmd(struct csio_ioreq *req)
>> +{
>> + struct csio_wr_pair wrp;
>> + struct csio_hw *hw = req->lnode->hwp;
>> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
>> + uint32_t size = CSIO_SCSI_CMD_WR_SZ_16(scsim->proto_cmd_len);
>> +
>> + req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
>> + if (unlikely(req->drv_status != CSIO_SUCCESS))
>> + return;
>> +
>> + if (wrp.size1 >= size) {
>> + /* Initialize WR in one shot */
>> + csio_scsi_init_cmd_wr(req, wrp.addr1, size);
>> + } else {
>> + uint8_t tmpwr[512];
>
> Mmmm, putting this large of a buffer on the local stack is probably not
> a good idea.
>
> This should become an allocation.. If it's a hot path then you'll
> probably want to set this up before-hand.
>
The else switch above is entered only when we near the end of the DMA
ring. This is not a frequent occurrence. If it is a strict no-no to have
so many on-stack bytes, I have to think of a way re-work it to use
pre-allocated memory. Please let me know.
>> + /*
>> + * Make a temporary copy of the WR and write back
>> + * the copy into the WR pair.
>> + */
>> + csio_scsi_init_cmd_wr(req, (void *)tmpwr, size);
>> + memcpy(wrp.addr1, tmpwr, wrp.size1);
>> + memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
>> + }
>> +}
>> +
>> +/*
>> + * The following is fast path code. Therefore it is inlined with multi-line
>> + * macros using name substitution, thus avoiding if-else switches for
>> + * operation (read/write), as well as serving the purpose of code re-use.
>> + */
>> +/*
>> + * csio_scsi_init_ulptx_dsgl - Fill in a ULP_TX_SC_DSGL
>> + * @hw: HW module
>> + * @req: IO request
>> + * @sgl: ULP TX SGL pointer.
>> + *
>> + */
>> +#define csio_scsi_init_ultptx_dsgl(hw, req, sgl) \
>> +do { \
>> + struct ulptx_sge_pair *_sge_pair = NULL; \
>> + struct scatterlist *_sgel; \
>> + uint32_t _i = 0; \
>> + uint32_t _xfer_len; \
>> + struct list_head *_tmp; \
>> + struct csio_dma_buf *_dma_buf; \
>> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
>> + \
>> + (sgl)->cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) | ULPTX_MORE | \
>> + ULPTX_NSGE((req)->nsge)); \
>> + /* Now add the data SGLs */ \
>> + if (likely(!(req)->dcopy)) { \
>> + scsi_for_each_sg(scmnd, _sgel, (req)->nsge, _i) { \
>> + if (_i == 0) { \
>> + (sgl)->addr0 = cpu_to_be64( \
>> + sg_dma_address(_sgel)); \
>> + (sgl)->len0 = cpu_to_be32( \
>> + sg_dma_len(_sgel)); \
>> + _sge_pair = \
>> + (struct ulptx_sge_pair *)((sgl) + 1); \
>> + continue; \
>> + } \
>> + if ((_i - 1) & 0x1) { \
>> + _sge_pair->addr[1] = cpu_to_be64( \
>> + sg_dma_address(_sgel)); \
>> + _sge_pair->len[1] = cpu_to_be32( \
>> + sg_dma_len(_sgel)); \
>> + _sge_pair++; \
>> + } else { \
>> + _sge_pair->addr[0] = cpu_to_be64( \
>> + sg_dma_address(_sgel)); \
>> + _sge_pair->len[0] = cpu_to_be32( \
>> + sg_dma_len(_sgel)); \
>> + } \
>> + } \
>> + } else { \
>> + /* Program sg elements with driver's DDP buffer */ \
>> + _xfer_len = scsi_bufflen(scmnd); \
>> + list_for_each(_tmp, &(req)->gen_list) { \
>> + _dma_buf = (struct csio_dma_buf *)_tmp; \
>> + if (_i == 0) { \
>> + (sgl)->addr0 = cpu_to_be64(_dma_buf->paddr); \
>> + (sgl)->len0 = cpu_to_be32( \
>> + min(_xfer_len, _dma_buf->len)); \
>> + _sge_pair = \
>> + (struct ulptx_sge_pair *)((sgl) + 1); \
>> + } \
>> + else if ((_i - 1) & 0x1) { \
>> + _sge_pair->addr[1] = cpu_to_be64( \
>> + _dma_buf->paddr); \
>> + _sge_pair->len[1] = cpu_to_be32( \
>> + min(_xfer_len, _dma_buf->len)); \
>> + _sge_pair++; \
>> + } else { \
>> + _sge_pair->addr[0] = cpu_to_be64( \
>> + _dma_buf->paddr); \
>> + _sge_pair->len[0] = cpu_to_be32( \
>> + min(_xfer_len, _dma_buf->len)); \
>> + } \
>> + _xfer_len -= min(_xfer_len, _dma_buf->len); \
>> + _i++; \
>> + } \
>> + } \
>> +} while (0)
>> +
>
> I don't see any reason why this can't just be a static function..? Why
> is the macro usage necessary here..?
OK, I will make this static-inline.
>
>> +/*
>> + * csio_scsi_init_data_wr - Initialize the READ/WRITE SCSI WR.
>> + * @req: IO req structure.
>> + * @oper: read/write
>> + * @wrp: DMA location to place the payload.
>> + * @size: Size of WR (including FW WR + immed data + rsp SG entry + data SGL
>> + * @wrop: _READ_/_WRITE_
>> + *
>> + * Wrapper for populating fw_scsi_read_wr/fw_scsi_write_wr.
>> + */
>> +#define csio_scsi_init_data_wr(req, oper, wrp, size, wrop) \
>> +do { \
>> + struct csio_hw *_hw = (req)->lnode->hwp; \
>> + struct csio_rnode *_rn = (req)->rnode; \
>> + struct fw_scsi_##oper##_wr *__wr = (struct fw_scsi_##oper##_wr *)(wrp);\
>> + struct ulptx_sgl *_sgl; \
>> + struct csio_dma_buf *_dma_buf; \
>> + uint8_t _imm = csio_hw_to_scsim(_hw)->proto_cmd_len; \
>> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
>> + \
>> + __wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI##wrop##WR) | \
>> + FW_SCSI##wrop##WR_IMMDLEN(_imm)); \
>> + __wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(_rn->flowid) | \
>> + FW_WR_LEN16( \
>> + CSIO_ROUNDUP((size), 16))); \
>> + __wr->cookie = (uintptr_t) (req); \
>> + __wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(_hw, \
>> + (req)->iq_idx));\
>> + __wr->tmo_val = (uint8_t)((req)->tmo); \
>> + __wr->use_xfer_cnt = 1; \
>> + __wr->xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
>> + __wr->ini_xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
>> + /* Get RSP DMA buffer */ \
>> + _dma_buf = &(req)->dma_buf; \
>> + \
>> + /* Prepare RSP SGL */ \
>> + __wr->rsp_dmalen = cpu_to_be32(_dma_buf->len); \
>> + __wr->rsp_dmaaddr = cpu_to_be64(_dma_buf->paddr); \
>> + \
>> + __wr->r4 = 0; \
>> + \
>> + __wr->u.fcoe.ctl_pri = 0; \
>> + __wr->u.fcoe.cp_en_class = 0; \
>> + __wr->u.fcoe.r3_lo[0] = 0; \
>> + __wr->u.fcoe.r3_lo[1] = 0; \
>> + csio_scsi_fcp_cmnd((req), (void *)((uintptr_t)(wrp) + \
>> + sizeof(struct fw_scsi_##oper##_wr))); \
>> + \
>> + /* Move WR pointer past command and immediate data */ \
>> + _sgl = (struct ulptx_sgl *) ((uintptr_t)(wrp) + \
>> + sizeof(struct fw_scsi_##oper##_wr) + \
>> + ALIGN(_imm, 16)); \
>> + \
>> + /* Fill in the DSGL */ \
>> + csio_scsi_init_ultptx_dsgl(_hw, (req), _sgl); \
>> + \
>> +} while (0)
>> +
>
> This one has four uses of CPP keys. Just turn those into macros, and
> leave the rest of the code in a static function.
>
So what you are suggesting is to have all the lines of the macro
csio_scsi_init_data_wr() added into a static function, but for the ones
with the 4 keys. csio_scsi_init_data_wr() will then invoke this new
function. Is that correct?
>> +/* Calculate WR size needed for fw_scsi_read_wr/fw_scsi_write_wr */
>> +#define csio_scsi_data_wrsz(req, oper, sz, imm) \
>> +do { \
>> + (sz) = sizeof(struct fw_scsi_##oper##_wr) + /* WR size */ \
>> + ALIGN((imm), 16) + /* Immed data */ \
>> + sizeof(struct ulptx_sgl); /* ulptx_sgl */ \
>> + \
>> + if (unlikely((req)->nsge > 1)) \
>> + (sz) += (sizeof(struct ulptx_sge_pair) * \
>> + (ALIGN(((req)->nsge - 1), 2) / 2)); \
>> + /* Data SGE */ \
>> +} while (0)
>> +
>> +/*
>> + * csio_scsi_data - Create a SCSI WRITE/READ WR.
>> + * @req: IO req structure.
>> + * @oper: read/write
>> + * @wrop: _READ_/_WRITE_ (string subsitutions to use with the FW bit field
>> + * macros).
>> + *
>> + * Gets a WR slot in the ingress queue and initializes it with
>> + * SCSI CMD READ/WRITE WR.
>> + *
>> + */
>> +#define csio_scsi_data(req, oper, wrop) \
>> +do { \
>> + struct csio_wr_pair _wrp; \
>> + uint32_t _size; \
>> + struct csio_hw *_hw = (req)->lnode->hwp; \
>> + struct csio_scsim *_scsim = csio_hw_to_scsim(_hw); \
>> + \
>> + csio_scsi_data_wrsz((req), oper, _size, _scsim->proto_cmd_len); \
>> + _size = ALIGN(_size, 16); \
>> + \
>> + (req)->drv_status = csio_wr_get(_hw, (req)->eq_idx, _size, &_wrp); \
>> + if (likely((req)->drv_status == CSIO_SUCCESS)) { \
>> + if (likely(_wrp.size1 >= _size)) { \
>> + /* Initialize WR in one shot */ \
>> + csio_scsi_init_data_wr((req), oper, _wrp.addr1, \
>> + _size, wrop); \
>> + } else { \
>> + uint8_t tmpwr[512]; \
>> + /* \
>> + * Make a temporary copy of the WR and write back \
>> + * the copy into the WR pair. \
>> + */ \
>> + csio_scsi_init_data_wr((req), oper, (void *)tmpwr, \
>> + _size, wrop); \
>> + memcpy(_wrp.addr1, tmpwr, _wrp.size1); \
>> + memcpy(_wrp.addr2, tmpwr + _wrp.size1, \
>> + _size - _wrp.size1); \
>> + } \
>> + } \
>> +} while (0)
>> +
>
> Ditto on this one, along with the tmpwr[512] stack usage..
>
>> +static inline void
>> +csio_scsi_abrt_cls(struct csio_ioreq *req, bool abort)
>> +{
>> + struct csio_wr_pair wrp;
>> + struct csio_hw *hw = req->lnode->hwp;
>> + uint32_t size = ALIGN(sizeof(struct fw_scsi_abrt_cls_wr), 16);
>> +
>> + req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
>> + if (req->drv_status != CSIO_SUCCESS)
>> + return;
>> +
>> + if (wrp.size1 >= size) {
>> + /* Initialize WR in one shot */
>> + csio_scsi_init_abrt_cls_wr(req, wrp.addr1, size, abort);
>> + } else {
>> + uint8_t tmpwr[512];
>
> Ditto here on local scope stack usage..
>
>> + /*
>> + * Make a temporary copy of the WR and write back
>> + * the copy into the WR pair.
>> + */
>> + csio_scsi_init_abrt_cls_wr(req, (void *)tmpwr, size, abort);
>> + memcpy(wrp.addr1, tmpwr, wrp.size1);
>> + memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
>> + }
>> +}
>> +
>> +/*****************************************************************************/
>> +/* START: SCSI SM */
>> +/*****************************************************************************/
>> +static void
>> +csio_scsis_uninit(struct csio_ioreq *req, enum csio_scsi_ev evt)
>> +{
>> + struct csio_hw *hw = req->lnode->hwp;
>> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
>> +
>> + switch (evt) {
>> +
>> + case CSIO_SCSIE_START_IO:
>
> Extra space between start of first switch case
OK, I will remove it. Is there any tool that catches such deviations?
checkpath.pl didnt report it.
>
>> +
>> + /* There is data */
>
> Point-less comment
I will remove it.
>
>> + if (req->nsge) {
>> + if (req->datadir == CSIO_IOREQF_DMA_WRITE) {
>> + req->dcopy = 0;
>> + csio_scsi_data(req, write, _WRITE_);
>> + } else
>> + csio_setup_ddp(scsim, req);
>> + } else {
>> + csio_scsi_cmd(req);
>> + }
>> +
>> + if (likely(req->drv_status == CSIO_SUCCESS)) {
>> + /* change state and enqueue on active_q */
>> + csio_set_state(&req->sm, csio_scsis_io_active);
>> + list_add_tail(&req->sm.sm_list, &scsim->active_q);
>> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
>> + csio_inc_stats(scsim, n_active);
>> +
>> + return;
>> + }
>> + break;
>> +
>> + case CSIO_SCSIE_START_TM:
>> + csio_scsi_cmd(req);
>> + if (req->drv_status == CSIO_SUCCESS) {
>> + /*
>> + * NOTE: We collect the affected I/Os prior to issuing
>> + * LUN reset, and not after it. This is to prevent
>> + * aborting I/Os that get issued after the LUN reset,
>> + * but prior to LUN reset completion (in the event that
>> + * the host stack has not blocked I/Os to a LUN that is
>> + * being reset.
>> + */
>> + csio_set_state(&req->sm, csio_scsis_tm_active);
>> + list_add_tail(&req->sm.sm_list, &scsim->active_q);
>> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
>> + csio_inc_stats(scsim, n_tm_active);
>> + }
>> + return;
>> +
>> + case CSIO_SCSIE_ABORT:
>> + case CSIO_SCSIE_CLOSE:
>> + /*
>> + * NOTE:
>> + * We could get here due to :
>> + * - a window in the cleanup path of the SCSI module
>> + * (csio_scsi_abort_io()). Please see NOTE in this function.
>> + * - a window in the time we tried to issue an abort/close
>> + * of a request to FW, and the FW completed the request
>> + * itself.
>> + * Print a message for now, and return INVAL either way.
>> + */
>> + req->drv_status = CSIO_INVAL;
>> + csio_warn(hw, "Trying to abort/close completed IO:%p!\n", req);
>> + break;
>> +
>> + default:
>> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
>> + CSIO_DB_ASSERT(0);
>> + }
>> +}
>> +
>> +static void
>> +csio_scsis_io_active(struct csio_ioreq *req, enum csio_scsi_ev evt)
>> +{
>> + struct csio_hw *hw = req->lnode->hwp;
>> + struct csio_scsim *scm = csio_hw_to_scsim(hw);
>> + struct csio_rnode *rn;
>> +
>> + switch (evt) {
>> +
>> + case CSIO_SCSIE_COMPLETED:
>
> Ditto
>
>> + csio_dec_stats(scm, n_active);
>> + list_del_init(&req->sm.sm_list);
>> + csio_set_state(&req->sm, csio_scsis_uninit);
>> + /*
>> + * In MSIX mode, with multiple queues, the SCSI compeltions
>> + * could reach us sooner than the FW events sent to indicate
>> + * I-T nexus loss (link down, remote device logo etc). We
>> + * dont want to be returning such I/Os to the upper layer
>> + * immediately, since we wouldnt have reported the I-T nexus
>> + * loss itself. This forces us to serialize such completions
>> + * with the reporting of the I-T nexus loss. Therefore, we
>> + * internally queue up such up such completions in the rnode.
>> + * The reporting of I-T nexus loss to the upper layer is then
>> + * followed by the returning of I/Os in this internal queue.
>> + * Having another state alongwith another queue helps us take
>> + * actions for events such as ABORT received while we are
>> + * in this rnode queue.
>> + */
>> + if (unlikely(req->wr_status != FW_SUCCESS)) {
>> + rn = req->rnode;
>> + /*
>> + * FW says remote device is lost, but rnode
>> + * doesnt reflect it.
>> + */
>> + if (csio_scsi_itnexus_loss_error(req->wr_status) &&
>> + csio_is_rnode_ready(rn)) {
>> + csio_set_state(&req->sm,
>> + csio_scsis_shost_cmpl_await);
>> + list_add_tail(&req->sm.sm_list,
>> + &rn->host_cmpl_q);
>> + }
>> + }
>> +
>> + break;
>> +
>> + case CSIO_SCSIE_ABORT:
>> + csio_scsi_abrt_cls(req, SCSI_ABORT);
>> + if (req->drv_status == CSIO_SUCCESS) {
>> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
>> + csio_set_state(&req->sm, csio_scsis_aborting);
>> + }
>> + break;
>> +
>> + case CSIO_SCSIE_CLOSE:
>> + csio_scsi_abrt_cls(req, SCSI_CLOSE);
>> + if (req->drv_status == CSIO_SUCCESS) {
>> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
>> + csio_set_state(&req->sm, csio_scsis_closing);
>> + }
>> + break;
>> +
>> + case CSIO_SCSIE_DRVCLEANUP:
>> + req->wr_status = FW_HOSTERROR;
>> + csio_dec_stats(scm, n_active);
>> + csio_set_state(&req->sm, csio_scsis_uninit);
>> + break;
>> +
>> + default:
>> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
>> + CSIO_DB_ASSERT(0);
>> + }
>> +}
>> +
>> +static void
>> +csio_scsis_tm_active(struct csio_ioreq *req, enum csio_scsi_ev evt)
>> +{
>> + struct csio_hw *hw = req->lnode->hwp;
>> + struct csio_scsim *scm = csio_hw_to_scsim(hw);
>> +
>> + switch (evt) {
>> +
>> + case CSIO_SCSIE_COMPLETED:
>
> Ditto
>
>> + csio_dec_stats(scm, n_tm_active);
>> + list_del_init(&req->sm.sm_list);
>> + csio_set_state(&req->sm, csio_scsis_uninit);
>> +
>> + break;
>> +
>> + case CSIO_SCSIE_ABORT:
>> + csio_scsi_abrt_cls(req, SCSI_ABORT);
>> + if (req->drv_status == CSIO_SUCCESS) {
>> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
>> + csio_set_state(&req->sm, csio_scsis_aborting);
>> + }
>> + break;
>> +
>> +
>> + case CSIO_SCSIE_CLOSE:
>> + csio_scsi_abrt_cls(req, SCSI_CLOSE);
>> + if (req->drv_status == CSIO_SUCCESS) {
>> + csio_wr_issue(hw, req->eq_idx, CSIO_FALSE);
>> + csio_set_state(&req->sm, csio_scsis_closing);
>> + }
>> + break;
>> +
>> + case CSIO_SCSIE_DRVCLEANUP:
>> + req->wr_status = FW_HOSTERROR;
>> + csio_dec_stats(scm, n_tm_active);
>> + csio_set_state(&req->sm, csio_scsis_uninit);
>> + break;
>> +
>> + default:
>> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
>> + CSIO_DB_ASSERT(0);
>> + }
>> +}
>> +
>> +static void
>> +csio_scsis_aborting(struct csio_ioreq *req, enum csio_scsi_ev evt)
>> +{
>> + struct csio_hw *hw = req->lnode->hwp;
>> + struct csio_scsim *scm = csio_hw_to_scsim(hw);
>> +
>> + switch (evt) {
>> +
>> + case CSIO_SCSIE_COMPLETED:
>
> Ditto
>
>> + csio_dbg(hw,
>> + "ioreq %p recvd cmpltd (wr_status:%d) "
>> + "in aborting st\n", req, req->wr_status);
>> + /*
>> + * Use CSIO_CANCELLED to explicitly tell the ABORTED event that
>> + * the original I/O was returned to driver by FW.
>> + * We dont really care if the I/O was returned with success by
>> + * FW (because the ABORT and completion of the I/O crossed each
>> + * other), or any other return value. Once we are in aborting
>> + * state, the success or failure of the I/O is unimportant to
>> + * us.
>> + */
>> + req->drv_status = CSIO_CANCELLED;
>> + break;
>> +
>> + case CSIO_SCSIE_ABORT:
>> + csio_inc_stats(scm, n_abrt_dups);
>> + break;
>> +
>> + case CSIO_SCSIE_ABORTED:
>> +
>> + csio_dbg(hw, "abort of %p return status:0x%x drv_status:%x\n",
>> + req, req->wr_status, req->drv_status);
>> + /*
>> + * Check if original I/O WR completed before the Abort
>> + * completion.
>> + */
>> + if (req->drv_status != CSIO_CANCELLED) {
>> + csio_fatal(hw,
>> + "Abort completed before original I/O,"
>> + " req:%p\n", req);
>> + CSIO_DB_ASSERT(0);
>> + }
>> +
>> + /*
>> + * There are the following possible scenarios:
>> + * 1. The abort completed successfully, FW returned FW_SUCCESS.
>> + * 2. The completion of an I/O and the receipt of
>> + * abort for that I/O by the FW crossed each other.
>> + * The FW returned FW_EINVAL. The original I/O would have
>> + * returned with FW_SUCCESS or any other SCSI error.
>> + * 3. The FW couldnt sent the abort out on the wire, as there
>> + * was an I-T nexus loss (link down, remote device logged
>> + * out etc). FW sent back an appropriate IT nexus loss status
>> + * for the abort.
>> + * 4. FW sent an abort, but abort timed out (remote device
>> + * didnt respond). FW replied back with
>> + * FW_SCSI_ABORT_TIMEDOUT.
>> + * 5. FW couldnt genuinely abort the request for some reason,
>> + * and sent us an error.
>> + *
>> + * The first 3 scenarios are treated as succesful abort
>> + * operations by the host, while the last 2 are failed attempts
>> + * to abort. Manipulate the return value of the request
>> + * appropriately, so that host can convey these results
>> + * back to the upper layer.
>> + */
>> + if ((req->wr_status == FW_SUCCESS) ||
>> + (req->wr_status == FW_EINVAL) ||
>> + csio_scsi_itnexus_loss_error(req->wr_status))
>> + req->wr_status = FW_SCSI_ABORT_REQUESTED;
>> +
>> + csio_dec_stats(scm, n_active);
>> + list_del_init(&req->sm.sm_list);
>> + csio_set_state(&req->sm, csio_scsis_uninit);
>> + break;
>> +
>> + case CSIO_SCSIE_DRVCLEANUP:
>> + req->wr_status = FW_HOSTERROR;
>> + csio_dec_stats(scm, n_active);
>> + csio_set_state(&req->sm, csio_scsis_uninit);
>> + break;
>> +
>> + case CSIO_SCSIE_CLOSE:
>> + /*
>> + * We can receive this event from the module
>> + * cleanup paths, if the FW forgot to reply to the ABORT WR
>> + * and left this ioreq in this state. For now, just ignore
>> + * the event. The CLOSE event is sent to this state, as
>> + * the LINK may have already gone down.
>> + */
>> + break;
>> +
>> + default:
>> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
>> + CSIO_DB_ASSERT(0);
>> + }
>> +}
>> +
>> +static void
>> +csio_scsis_closing(struct csio_ioreq *req, enum csio_scsi_ev evt)
>> +{
>> + struct csio_hw *hw = req->lnode->hwp;
>> + struct csio_scsim *scm = csio_hw_to_scsim(hw);
>> +
>> + switch (evt) {
>> +
>> + case CSIO_SCSIE_COMPLETED:
>
> Ditto
>
>> + csio_dbg(hw,
>> + "ioreq %p recvd cmpltd (wr_status:%d) "
>> + "in closing st\n", req, req->wr_status);
>> + /*
>> + * Use CSIO_CANCELLED to explicitly tell the CLOSED event that
>> + * the original I/O was returned to driver by FW.
>> + * We dont really care if the I/O was returned with success by
>> + * FW (because the CLOSE and completion of the I/O crossed each
>> + * other), or any other return value. Once we are in aborting
>> + * state, the success or failure of the I/O is unimportant to
>> + * us.
>> + */
>> + req->drv_status = CSIO_CANCELLED;
>> + break;
>> +
>> + case CSIO_SCSIE_CLOSED:
>> + /*
>> + * Check if original I/O WR completed before the Close
>> + * completion.
>> + */
>> + if (req->drv_status != CSIO_CANCELLED) {
>> + csio_fatal(hw,
>> + "Close completed before original I/O,"
>> + " req:%p\n", req);
>> + CSIO_DB_ASSERT(0);
>> + }
>> +
>> + /*
>> + * Either close succeeded, or we issued close to FW at the
>> + * same time FW compelted it to us. Either way, the I/O
>> + * is closed.
>> + */
>> + CSIO_DB_ASSERT((req->wr_status == FW_SUCCESS) ||
>> + (req->wr_status == FW_EINVAL));
>> + req->wr_status = FW_SCSI_CLOSE_REQUESTED;
>> +
>> + csio_dec_stats(scm, n_active);
>> + list_del_init(&req->sm.sm_list);
>> + csio_set_state(&req->sm, csio_scsis_uninit);
>> + break;
>> +
>> + case CSIO_SCSIE_CLOSE:
>> + break;
>> +
>> + case CSIO_SCSIE_DRVCLEANUP:
>> + req->wr_status = FW_HOSTERROR;
>> + csio_dec_stats(scm, n_active);
>> + csio_set_state(&req->sm, csio_scsis_uninit);
>> + break;
>> +
>> + default:
>> + csio_dbg(hw, "Unhandled event:%d sent to req:%p\n", evt, req);
>> + CSIO_DB_ASSERT(0);
>> + }
>> +}
>> +
>> +static void
>> +csio_scsis_shost_cmpl_await(struct csio_ioreq *req, enum csio_scsi_ev evt)
>> +{
>> + switch (evt) {
>> + case CSIO_SCSIE_ABORT:
>> + case CSIO_SCSIE_CLOSE:
>> + /*
>> + * Just succeed the abort request, and hope that
>> + * the remote device unregister path will cleanup
>> + * this I/O to the upper layer within a sane
>> + * amount of time.
>> + */
>> + /*
>> + * A close can come in during a LINK DOWN. The FW would have
>> + * returned us the I/O back, but not the remote device lost
>> + * FW event. In this interval, if the I/O times out at the upper
>> + * layer, a close can come in. Take the same action as abort:
>> + * return success, and hope that the remote device unregister
>> + * path will cleanup this I/O. If the FW still doesnt send
>> + * the msg, the close times out, and the upper layer resorts
>> + * to the next level of error recovery.
>> + */
>> + req->drv_status = CSIO_SUCCESS;
>> + break;
>> + case CSIO_SCSIE_DRVCLEANUP:
>> + csio_set_state(&req->sm, csio_scsis_uninit);
>> + break;
>> + default:
>> + csio_dbg(req->lnode->hwp, "Unhandled event:%d sent to req:%p\n",
>> + evt, req);
>> + CSIO_DB_ASSERT(0);
>> + }
>> +}
>> +
>
>
>> +
>> +/**
>> + * csio_queuecommand_lck - Entry point to kickstart an I/O request.
>> + * @cmnd: The I/O request from ML.
>> + * @done: The ML callback routine.
>> + *
>> + * This routine does the following:
>> + * - Checks for HW and Rnode module readiness.
>> + * - Gets a free ioreq structure (which is already initialized
>> + * to uninit during its allocation).
>> + * - Maps SG elements.
>> + * - Initializes ioreq members.
>> + * - Kicks off the SCSI state machine for this IO.
>> + * - Returns busy status on error.
>> + */
>> +static int
>> +csio_queuecommand_lck(struct scsi_cmnd *cmnd, void (*done)(struct scsi_cmnd *))
>> +{
>> + struct csio_lnode *ln = shost_priv(cmnd->device->host);
>> + struct csio_hw *hw = csio_lnode_to_hw(ln);
>> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
>> + struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
>> + struct csio_ioreq *ioreq = NULL;
>> + unsigned long flags;
>> + int nsge = 0;
>> + int rv = SCSI_MLQUEUE_HOST_BUSY, nr;
>> + csio_retval_t retval;
>> + int cpu;
>> + struct csio_scsi_qset *sqset;
>> + struct fc_rport *rport = starget_to_rport(scsi_target(cmnd->device));
>> +
>> + if (!blk_rq_cpu_valid(cmnd->request))
>> + cpu = smp_processor_id();
>> + else
>> + cpu = cmnd->request->cpu;
>> +
>> + sqset = &hw->sqset[ln->portid][cpu];
>> +
>> + nr = fc_remote_port_chkready(rport);
>> + if (nr) {
>> + cmnd->result = nr;
>> + csio_inc_stats(scsim, n_rn_nr_error);
>> + goto err_done;
>> + }
>> +
>> + if (unlikely(!csio_is_hw_ready(hw))) {
>> + cmnd->result = (DID_REQUEUE << 16);
>> + csio_inc_stats(scsim, n_hw_nr_error);
>> + goto err_done;
>> + }
>> +
>> + /* Get req->nsge, if there are SG elements to be mapped */
>> + nsge = scsi_dma_map(cmnd);
>> + if (unlikely(nsge < 0)) {
>> + csio_inc_stats(scsim, n_dmamap_error);
>> + goto err;
>> + }
>> +
>> + /* Do we support so many mappings? */
>> + if (unlikely(nsge > scsim->max_sge)) {
>> + csio_warn(hw,
>> + "More SGEs than can be supported."
>> + " SGEs: %d, Max SGEs: %d\n", nsge, scsim->max_sge);
>> + csio_inc_stats(scsim, n_unsupp_sge_error);
>> + goto err_dma_unmap;
>> + }
>> +
>> + /* Get a free ioreq structure - SM is already set to uninit */
>> + ioreq = csio_get_scsi_ioreq_lock(hw, scsim);
>> + if (!ioreq) {
>> + csio_err(hw, "Out of I/O request elements. Active #:%d\n",
>> + scsim->stats.n_active);
>> + csio_inc_stats(scsim, n_no_req_error);
>> + goto err_dma_unmap;
>> + }
>> +
>> + ioreq->nsge = nsge;
>> + ioreq->lnode = ln;
>> + ioreq->rnode = rn;
>> + ioreq->iq_idx = sqset->iq_idx;
>> + ioreq->eq_idx = sqset->eq_idx;
>> + ioreq->wr_status = 0;
>> + ioreq->drv_status = CSIO_SUCCESS;
>> + csio_scsi_cmnd(ioreq) = (void *)cmnd;
>> + ioreq->tmo = 0;
>> +
>> + switch (cmnd->sc_data_direction) {
>> + case DMA_BIDIRECTIONAL:
>> + ioreq->datadir = CSIO_IOREQF_DMA_BIDI;
>> + csio_inc_stats(ln, n_control_requests);
>> + break;
>> + case DMA_TO_DEVICE:
>> + ioreq->datadir = CSIO_IOREQF_DMA_WRITE;
>> + csio_inc_stats(ln, n_output_requests);
>> + ln->stats.n_output_bytes += scsi_bufflen(cmnd);
>> + break;
>> + case DMA_FROM_DEVICE:
>> + ioreq->datadir = CSIO_IOREQF_DMA_READ;
>> + csio_inc_stats(ln, n_input_requests);
>> + ln->stats.n_input_bytes += scsi_bufflen(cmnd);
>> + break;
>> + case DMA_NONE:
>> + ioreq->datadir = CSIO_IOREQF_DMA_NONE;
>> + csio_inc_stats(ln, n_control_requests);
>> + break;
>> + default:
>> + CSIO_DB_ASSERT(0);
>> + break;
>> + }
>> +
>> + /* Set cbfn */
>> + ioreq->io_cbfn = csio_scsi_cbfn;
>> +
>> + /* Needed during abort */
>> + cmnd->host_scribble = (unsigned char *)ioreq;
>> + cmnd->scsi_done = done;
>> + cmnd->SCp.Message = 0;
>> +
>> + /* Kick off SCSI IO SM on the ioreq */
>> + spin_lock_irqsave(&hw->lock, flags);
>> + retval = csio_scsi_start_io(ioreq);
>> + spin_unlock_irqrestore(&hw->lock, flags);
>> +
>> + if (retval != CSIO_SUCCESS) {
>> + csio_err(hw, "ioreq: %p couldnt be started, status:%d\n",
>> + ioreq, retval);
>> + csio_inc_stats(scsim, n_busy_error);
>> + goto err_put_req;
>> + }
>> +
>> + return 0;
>> +
>> +err_put_req:
>> + csio_put_scsi_ioreq_lock(hw, scsim, ioreq);
>> +err_dma_unmap:
>> + if (nsge > 0)
>> + scsi_dma_unmap(cmnd);
>> +err:
>> + return rv;
>> +
>> +err_done:
>> + done(cmnd);
>> + return 0;
>> +}
>> +
>> +static DEF_SCSI_QCMD(csio_queuecommand);
>> +
>
> This means that your running with the host_lock held.. I'm not sure if
> that is really what you want to do as it really end's up killing
> multi-lun small packet performance..
>
> How about dropping DEF_SCSI_QCMD usage here, and figure out what
> actually needs to be protected by the SCSI host_lock within
> csio_queuecommand_lck()..?
It is on my TODO list for the next version of the driver, after the
initial submission. Per the current design, we shouldnt need the
host_lock to be held, but I would like to test this change thoroughly
before I submit it.
>> +static csio_retval_t
>> +csio_do_abrt_cls(struct csio_hw *hw, struct csio_ioreq *ioreq, bool abort)
>> +{
>> + csio_retval_t rv;
>> + int cpu = smp_processor_id();
>> + struct csio_lnode *ln = ioreq->lnode;
>> + struct csio_scsi_qset *sqset = &hw->sqset[ln->portid][cpu];
>> +
>> + ioreq->tmo = CSIO_SCSI_ABRT_TMO_MS;
>> + /*
>> + * Use current processor queue for posting the abort/close, but retain
>> + * the ingress queue ID of the original I/O being aborted/closed - we
>> + * need the abort/close completion to be received on the same queue
>> + * as the original I/O.
>> + */
>> + ioreq->eq_idx = sqset->eq_idx;
>> +
>> + if (abort == SCSI_ABORT)
>> + rv = csio_scsi_abort(ioreq);
>> + else /* close */
>
> Point-less comment.
I will remove it.
>
>> + rv = csio_scsi_close(ioreq);
>> +
>> + return rv;
>> +}
>> +
>> +static int
>> +csio_eh_abort_handler(struct scsi_cmnd *cmnd)
>> +{
>> + struct csio_ioreq *ioreq;
>> + struct csio_lnode *ln = shost_priv(cmnd->device->host);
>> + struct csio_hw *hw = csio_lnode_to_hw(ln);
>> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
>> + int ready = 0, ret;
>> + unsigned long tmo = 0;
>> + csio_retval_t rv;
>> + struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
>> +
>> + ret = fc_block_scsi_eh(cmnd);
>> + if (ret)
>> + return ret;
>> +
>> + ioreq = (struct csio_ioreq *)cmnd->host_scribble;
>> + if (!ioreq)
>> + return SUCCESS;
>> +
>> + if (!rn)
>> + return FAILED;
>> +
>> + csio_dbg(hw,
>> + "Request to abort ioreq:%p cmd:%p cdb:%08llx"
>> + " ssni:0x%x lun:%d iq:0x%x\n",
>> + ioreq, cmnd, *((uint64_t *)cmnd->cmnd), rn->flowid,
>> + cmnd->device->lun, csio_q_physiqid(hw, ioreq->iq_idx));
>> +
>> + if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) != cmnd) {
>> + csio_inc_stats(scsim, n_abrt_race_comp);
>> + return SUCCESS;
>> + }
>> +
>> + ready = csio_is_lnode_ready(ln);
>> + tmo = CSIO_SCSI_ABRT_TMO_MS;
>> +
>> + spin_lock_irq(&hw->lock);
>> + rv = csio_do_abrt_cls(hw, ioreq, (ready ? SCSI_ABORT : SCSI_CLOSE));
>> + spin_unlock_irq(&hw->lock);
>> +
>> + if (rv != CSIO_SUCCESS) {
>> + if (rv == CSIO_INVAL) {
>> + /* Return success, if abort/close request issued on
>> + * already completed IO
>> + */
>> + return SUCCESS;
>> + }
>> + if (ready)
>> + csio_inc_stats(scsim, n_abrt_busy_error);
>> + else
>> + csio_inc_stats(scsim, n_cls_busy_error);
>> +
>> + goto inval_scmnd;
>> + }
>> +
>> + /* Wait for completion */
>> + init_completion(&ioreq->cmplobj);
>> + wait_for_completion_timeout(&ioreq->cmplobj, msecs_to_jiffies(tmo));
>> +
>> + /* FW didnt respond to abort within our timeout */
>> + if (((struct scsi_cmnd *)csio_scsi_cmnd(ioreq)) == cmnd) {
>> +
>> + csio_err(hw, "Abort timed out -- req: %p\n", ioreq);
>> + csio_inc_stats(scsim, n_abrt_timedout);
>> +
>> +inval_scmnd:
>> + if (ioreq->nsge > 0)
>> + scsi_dma_unmap(cmnd);
>> +
>> + spin_lock_irq(&hw->lock);
>> + csio_scsi_cmnd(ioreq) = NULL;
>> + spin_unlock_irq(&hw->lock);
>> +
>> + cmnd->result = (DID_ERROR << 16);
>> + cmnd->scsi_done(cmnd);
>> +
>> + return FAILED;
>> + }
>> +
>> + /* FW successfully aborted the request */
>> + if (host_byte(cmnd->result) == DID_REQUEUE) {
>> + csio_info(hw,
>> + "Aborted SCSI command to (%d:%d) serial#:0x%lx\n",
>> + cmnd->device->id, cmnd->device->lun,
>> + cmnd->serial_number);
>> + return SUCCESS;
>> + } else {
>> + csio_info(hw,
>> + "Failed to abort SCSI command, (%d:%d) serial#:0x%lx\n",
>> + cmnd->device->id, cmnd->device->lun,
>> + cmnd->serial_number);
>> + return FAILED;
>> + }
>> +}
>> +
>
>> +struct scsi_host_template csio_fcoe_shost_template = {
>> + .module = THIS_MODULE,
>> + .name = CSIO_DRV_DESC,
>> + .proc_name = KBUILD_MODNAME,
>> + .queuecommand = csio_queuecommand,
>> + .eh_abort_handler = csio_eh_abort_handler,
>> + .eh_device_reset_handler = csio_eh_lun_reset_handler,
>> + .slave_alloc = csio_slave_alloc,
>> + .slave_configure = csio_slave_configure,
>> + .slave_destroy = csio_slave_destroy,
>> + .scan_finished = csio_scan_finished,
>> + .this_id = -1,
>> + .sg_tablesize = CSIO_SCSI_MAX_SGE,
>> + .cmd_per_lun = CSIO_MAX_CMD_PER_LUN,
>> + .use_clustering = ENABLE_CLUSTERING,
>> + .shost_attrs = csio_fcoe_lport_attrs,
>> + .max_sectors = CSIO_MAX_SECTOR_SIZE,
>> +};
>> +
>> +struct scsi_host_template csio_fcoe_shost_vport_template = {
>> + .module = THIS_MODULE,
>> + .name = CSIO_DRV_DESC,
>> + .proc_name = KBUILD_MODNAME,
>> + .queuecommand = csio_queuecommand,
>> + .eh_abort_handler = csio_eh_abort_handler,
>> + .eh_device_reset_handler = csio_eh_lun_reset_handler,
>> + .slave_alloc = csio_slave_alloc,
>> + .slave_configure = csio_slave_configure,
>> + .slave_destroy = csio_slave_destroy,
>> + .scan_finished = csio_scan_finished,
>> + .this_id = -1,
>> + .sg_tablesize = CSIO_SCSI_MAX_SGE,
>> + .cmd_per_lun = CSIO_MAX_CMD_PER_LUN,
>> + .use_clustering = ENABLE_CLUSTERING,
>> + .shost_attrs = csio_fcoe_vport_attrs,
>> + .max_sectors = CSIO_MAX_SECTOR_SIZE,
>> +};
>> +
>> +/*
>> + * csio_scsi_alloc_ddp_bufs - Allocate buffers for DDP of unaligned SGLs.
>> + * @scm: SCSI Module
>> + * @hw: HW device.
>> + * @buf_size: buffer size
>> + * @num_buf : Number of buffers.
>> + *
>> + * This routine allocates DMA buffers required for SCSI Data xfer, if
>> + * each SGL buffer for a SCSI Read request posted by SCSI midlayer are
>> + * not virtually contiguous.
>> + */
>> +static csio_retval_t
>> +csio_scsi_alloc_ddp_bufs(struct csio_scsim *scm, struct csio_hw *hw,
>> + int buf_size, int num_buf)
>> +{
>> + int n = 0;
>> + struct list_head *tmp;
>> + struct csio_dma_buf *ddp_desc = NULL;
>> + uint32_t unit_size = 0;
>> +
>> + if (!num_buf)
>> + return CSIO_SUCCESS;
>
> Just return 0..?
Since there is a comment about driver-internal return values in your
review of patch 6, I would like to take up this discussion in that thread.
>
>> +
>> + if (!buf_size)
>> + return CSIO_INVAL;
>
> Just return -EINVAL..?
>
>> +
>> + INIT_LIST_HEAD(&scm->ddp_freelist);
>> +
>> + /* Align buf size to page size */
>> + buf_size = (buf_size + PAGE_SIZE - 1) & PAGE_MASK;
>> + /* Initialize dma descriptors */
>> + for (n = 0; n < num_buf; n++) {
>> + /* Set unit size to request size */
>> + unit_size = buf_size;
>> + ddp_desc = kzalloc(sizeof(struct csio_dma_buf), GFP_KERNEL);
>> + if (!ddp_desc) {
>> + csio_err(hw,
>> + "Failed to allocate ddp descriptors,"
>> + " Num allocated = %d.\n",
>> + scm->stats.n_free_ddp);
>> + goto no_mem;
>> + }
>> +
>> + /* Allocate Dma buffers for DDP */
>> + ddp_desc->vaddr = pci_alloc_consistent(hw->pdev, unit_size,
>> + &ddp_desc->paddr);
>> + if (!ddp_desc->vaddr) {
>> + csio_err(hw,
>> + "SCSI response DMA buffer (ddp) allocation"
>> + " failed!\n");
>> + kfree(ddp_desc);
>> + goto no_mem;
>> + }
>> +
>> + ddp_desc->len = unit_size;
>> +
>> + /* Added it to scsi ddp freelist */
>> + list_add_tail(&ddp_desc->list, &scm->ddp_freelist);
>> + csio_inc_stats(scm, n_free_ddp);
>> + }
>> +
>> + return CSIO_SUCCESS;
>
> Just return 0..?
>
>> +no_mem:
>> + /* release dma descs back to freelist and free dma memory */
>> + list_for_each(tmp, &scm->ddp_freelist) {
>> + ddp_desc = (struct csio_dma_buf *) tmp;
>> + tmp = csio_list_prev(tmp);
>> + pci_free_consistent(hw->pdev, ddp_desc->len, ddp_desc->vaddr,
>> + ddp_desc->paddr);
>> + list_del_init(&ddp_desc->list);
>> + kfree(ddp_desc);
>> + }
>> + scm->stats.n_free_ddp = 0;
>> +
>> + return CSIO_NOMEM;
>
> This should be just -ENOMEM..?
>
>> +}
>> +
>> +/*
>> + * csio_scsi_free_ddp_bufs - free DDP buffers of unaligned SGLs.
>> + * @scm: SCSI Module
>> + * @hw: HW device.
>> + *
>> + * This routine frees ddp buffers.
>> + */
>> +static csio_retval_t
>> +csio_scsi_free_ddp_bufs(struct csio_scsim *scm, struct csio_hw *hw)
>> +{
>> + struct list_head *tmp;
>> + struct csio_dma_buf *ddp_desc;
>> +
>> + /* release dma descs back to freelist and free dma memory */
>> + list_for_each(tmp, &scm->ddp_freelist) {
>> + ddp_desc = (struct csio_dma_buf *) tmp;
>> + tmp = csio_list_prev(tmp);
>> + pci_free_consistent(hw->pdev, ddp_desc->len, ddp_desc->vaddr,
>> + ddp_desc->paddr);
>> + list_del_init(&ddp_desc->list);
>> + kfree(ddp_desc);
>> + }
>> + scm->stats.n_free_ddp = 0;
>> +
>> + return CSIO_NOMEM;
>> +}
>
> Ditto.. Just -ENOMEM..?
>
>
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5).
2012-08-24 17:40 ` Naresh Kumar Inna
@ 2012-08-24 18:27 ` Joe Perches
2012-08-24 19:07 ` Naresh Kumar Inna
2012-08-24 20:56 ` Nicholas A. Bellinger
1 sibling, 1 reply; 33+ messages in thread
From: Joe Perches @ 2012-08-24 18:27 UTC (permalink / raw)
To: Naresh Kumar Inna
Cc: Nicholas A. Bellinger, JBottomley@parallels.com,
linux-scsi@vger.kernel.org, Dimitrios Michailidis,
netdev@vger.kernel.org, Chethan Seshadri
On Fri, 2012-08-24 at 23:10 +0530, Naresh Kumar Inna wrote:
> Is there a TRUE/FALSE define in a standard header file?
include/linux/stddef.h:
enum {
false = 0,
true = 1
};
> I see a lot of
> kernel code defining their own TRUE/FALSE.
Those should be changed eventually.
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5).
2012-08-24 18:27 ` Joe Perches
@ 2012-08-24 19:07 ` Naresh Kumar Inna
0 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-24 19:07 UTC (permalink / raw)
To: Joe Perches
Cc: Nicholas A. Bellinger, JBottomley@parallels.com,
linux-scsi@vger.kernel.org, Dimitrios Michailidis,
netdev@vger.kernel.org, Chethan Seshadri
On 8/24/2012 11:57 PM, Joe Perches wrote:
> On Fri, 2012-08-24 at 23:10 +0530, Naresh Kumar Inna wrote:
>> Is there a TRUE/FALSE define in a standard header file?
>
> include/linux/stddef.h:
> enum {
> false = 0,
> true = 1
> };
>
>> I see a lot of
>> kernel code defining their own TRUE/FALSE.
>
> Those should be changed eventually.
>
Thanks! I will switch over to these macros.
^ permalink raw reply [flat|nested] 33+ messages in thread
* Re: [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5).
2012-08-24 17:40 ` Naresh Kumar Inna
2012-08-24 18:27 ` Joe Perches
@ 2012-08-24 20:56 ` Nicholas A. Bellinger
2012-08-25 18:36 ` Naresh Kumar Inna
1 sibling, 1 reply; 33+ messages in thread
From: Nicholas A. Bellinger @ 2012-08-24 20:56 UTC (permalink / raw)
To: Naresh Kumar Inna
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On Fri, 2012-08-24 at 23:10 +0530, Naresh Kumar Inna wrote:
> On 8/24/2012 1:18 AM, Nicholas A. Bellinger wrote:
> > On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
> >> This patch contains code to implement the interrupt handling and the fast
> >> path I/O functionality. The interrupt handling includes allocation of
> >> MSIX vectors, registering and implemeting the interrupt service routines.
> >> The fast path I/O functionality includes posting the I/O request to firmware
> >> via Work Requests, tracking/completing them, and handling task management
> >> requests. SCSI midlayer host template implementation is also covered by
> >> this patch.
> >>
> >> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
> >> ---
> >
> > Hi Naresh,
> >
> > My review comments are inline below..
>
> Hi Nicholas,
>
> Thanks for taking the time to review the driver. Please find my replies
> inline.
>
> Regards,
> Naresh.
>
> >
> >> drivers/scsi/csiostor/csio_isr.c | 631 ++++++++++
> >> drivers/scsi/csiostor/csio_scsi.c | 2498 +++++++++++++++++++++++++++++++++++++
> >> 2 files changed, 3129 insertions(+), 0 deletions(-)
> >> create mode 100644 drivers/scsi/csiostor/csio_isr.c
> >> create mode 100644 drivers/scsi/csiostor/csio_scsi.c
> >>
<SNIP>
> >
> >> diff --git a/drivers/scsi/csiostor/csio_scsi.c b/drivers/scsi/csiostor/csio_scsi.c
> >> new file mode 100644
> >> index 0000000..0f87b00
> >> --- /dev/null
> >> +++ b/drivers/scsi/csiostor/csio_scsi.c
> >> +
> >> +#define CSIO_SCSI_CMD_WR_SZ(_imm) \
> >> + (sizeof(struct fw_scsi_cmd_wr) + /* WR size */ \
> >> + ALIGN((_imm), 16)) /* Immed data */
> >> +
> >> +#define CSIO_SCSI_CMD_WR_SZ_16(_imm) \
> >> + (ALIGN(CSIO_SCSI_CMD_WR_SZ((_imm)), 16))
> >> +
> >> +/*
> >> + * csio_scsi_cmd - Create a SCSI CMD WR.
> >> + * @req: IO req structure.
> >> + *
> >> + * Gets a WR slot in the ingress queue and initializes it with SCSI CMD WR.
> >> + *
> >> + */
> >> +static inline void
> >> +csio_scsi_cmd(struct csio_ioreq *req)
> >> +{
> >> + struct csio_wr_pair wrp;
> >> + struct csio_hw *hw = req->lnode->hwp;
> >> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
> >> + uint32_t size = CSIO_SCSI_CMD_WR_SZ_16(scsim->proto_cmd_len);
> >> +
> >> + req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
> >> + if (unlikely(req->drv_status != CSIO_SUCCESS))
> >> + return;
> >> +
> >> + if (wrp.size1 >= size) {
> >> + /* Initialize WR in one shot */
> >> + csio_scsi_init_cmd_wr(req, wrp.addr1, size);
> >> + } else {
> >> + uint8_t tmpwr[512];
> >
> > Mmmm, putting this large of a buffer on the local stack is probably not
> > a good idea.
> >
> > This should become an allocation.. If it's a hot path then you'll
> > probably want to set this up before-hand.
> >
>
> The else switch above is entered only when we near the end of the DMA
> ring. This is not a frequent occurrence. If it is a strict no-no to have
> so many on-stack bytes, I have to think of a way re-work it to use
> pre-allocated memory. Please let me know.
>
Considering it's used a number of times, it would be better to just
figure out a sensible manner to pre-allocate this, especially if it's
only a single occurrence to an individual ring held under a lock.
> >> + /*
> >> + * Make a temporary copy of the WR and write back
> >> + * the copy into the WR pair.
> >> + */
> >> + csio_scsi_init_cmd_wr(req, (void *)tmpwr, size);
> >> + memcpy(wrp.addr1, tmpwr, wrp.size1);
> >> + memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
> >> + }
> >> +}
> >> +
> >> +/*
> >> + * The following is fast path code. Therefore it is inlined with multi-line
> >> + * macros using name substitution, thus avoiding if-else switches for
> >> + * operation (read/write), as well as serving the purpose of code re-use.
> >> + */
> >> +/*
> >> + * csio_scsi_init_ulptx_dsgl - Fill in a ULP_TX_SC_DSGL
> >> + * @hw: HW module
> >> + * @req: IO request
> >> + * @sgl: ULP TX SGL pointer.
> >> + *
> >> + */
> >> +#define csio_scsi_init_ultptx_dsgl(hw, req, sgl) \
> >> +do { \
> >> + struct ulptx_sge_pair *_sge_pair = NULL; \
> >> + struct scatterlist *_sgel; \
> >> + uint32_t _i = 0; \
> >> + uint32_t _xfer_len; \
> >> + struct list_head *_tmp; \
> >> + struct csio_dma_buf *_dma_buf; \
> >> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
> >> + \
> >> + (sgl)->cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) | ULPTX_MORE | \
> >> + ULPTX_NSGE((req)->nsge)); \
> >> + /* Now add the data SGLs */ \
> >> + if (likely(!(req)->dcopy)) { \
> >> + scsi_for_each_sg(scmnd, _sgel, (req)->nsge, _i) { \
> >> + if (_i == 0) { \
> >> + (sgl)->addr0 = cpu_to_be64( \
> >> + sg_dma_address(_sgel)); \
> >> + (sgl)->len0 = cpu_to_be32( \
> >> + sg_dma_len(_sgel)); \
> >> + _sge_pair = \
> >> + (struct ulptx_sge_pair *)((sgl) + 1); \
> >> + continue; \
> >> + } \
> >> + if ((_i - 1) & 0x1) { \
> >> + _sge_pair->addr[1] = cpu_to_be64( \
> >> + sg_dma_address(_sgel)); \
> >> + _sge_pair->len[1] = cpu_to_be32( \
> >> + sg_dma_len(_sgel)); \
> >> + _sge_pair++; \
> >> + } else { \
> >> + _sge_pair->addr[0] = cpu_to_be64( \
> >> + sg_dma_address(_sgel)); \
> >> + _sge_pair->len[0] = cpu_to_be32( \
> >> + sg_dma_len(_sgel)); \
> >> + } \
> >> + } \
> >> + } else { \
> >> + /* Program sg elements with driver's DDP buffer */ \
> >> + _xfer_len = scsi_bufflen(scmnd); \
> >> + list_for_each(_tmp, &(req)->gen_list) { \
> >> + _dma_buf = (struct csio_dma_buf *)_tmp; \
> >> + if (_i == 0) { \
> >> + (sgl)->addr0 = cpu_to_be64(_dma_buf->paddr); \
> >> + (sgl)->len0 = cpu_to_be32( \
> >> + min(_xfer_len, _dma_buf->len)); \
> >> + _sge_pair = \
> >> + (struct ulptx_sge_pair *)((sgl) + 1); \
> >> + } \
> >> + else if ((_i - 1) & 0x1) { \
> >> + _sge_pair->addr[1] = cpu_to_be64( \
> >> + _dma_buf->paddr); \
> >> + _sge_pair->len[1] = cpu_to_be32( \
> >> + min(_xfer_len, _dma_buf->len)); \
> >> + _sge_pair++; \
> >> + } else { \
> >> + _sge_pair->addr[0] = cpu_to_be64( \
> >> + _dma_buf->paddr); \
> >> + _sge_pair->len[0] = cpu_to_be32( \
> >> + min(_xfer_len, _dma_buf->len)); \
> >> + } \
> >> + _xfer_len -= min(_xfer_len, _dma_buf->len); \
> >> + _i++; \
> >> + } \
> >> + } \
> >> +} while (0)
> >> +
> >
> > I don't see any reason why this can't just be a static function..? Why
> > is the macro usage necessary here..?
>
> OK, I will make this static-inline.
>
> >
> >> +/*
> >> + * csio_scsi_init_data_wr - Initialize the READ/WRITE SCSI WR.
> >> + * @req: IO req structure.
> >> + * @oper: read/write
> >> + * @wrp: DMA location to place the payload.
> >> + * @size: Size of WR (including FW WR + immed data + rsp SG entry + data SGL
> >> + * @wrop: _READ_/_WRITE_
> >> + *
> >> + * Wrapper for populating fw_scsi_read_wr/fw_scsi_write_wr.
> >> + */
> >> +#define csio_scsi_init_data_wr(req, oper, wrp, size, wrop) \
> >> +do { \
> >> + struct csio_hw *_hw = (req)->lnode->hwp; \
> >> + struct csio_rnode *_rn = (req)->rnode; \
> >> + struct fw_scsi_##oper##_wr *__wr = (struct fw_scsi_##oper##_wr *)(wrp);\
> >> + struct ulptx_sgl *_sgl; \
> >> + struct csio_dma_buf *_dma_buf; \
> >> + uint8_t _imm = csio_hw_to_scsim(_hw)->proto_cmd_len; \
> >> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
> >> + \
> >> + __wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI##wrop##WR) | \
> >> + FW_SCSI##wrop##WR_IMMDLEN(_imm)); \
> >> + __wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(_rn->flowid) | \
> >> + FW_WR_LEN16( \
> >> + CSIO_ROUNDUP((size), 16))); \
> >> + __wr->cookie = (uintptr_t) (req); \
> >> + __wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(_hw, \
> >> + (req)->iq_idx));\
> >> + __wr->tmo_val = (uint8_t)((req)->tmo); \
> >> + __wr->use_xfer_cnt = 1; \
> >> + __wr->xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
> >> + __wr->ini_xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
> >> + /* Get RSP DMA buffer */ \
> >> + _dma_buf = &(req)->dma_buf; \
> >> + \
> >> + /* Prepare RSP SGL */ \
> >> + __wr->rsp_dmalen = cpu_to_be32(_dma_buf->len); \
> >> + __wr->rsp_dmaaddr = cpu_to_be64(_dma_buf->paddr); \
> >> + \
> >> + __wr->r4 = 0; \
> >> + \
> >> + __wr->u.fcoe.ctl_pri = 0; \
> >> + __wr->u.fcoe.cp_en_class = 0; \
> >> + __wr->u.fcoe.r3_lo[0] = 0; \
> >> + __wr->u.fcoe.r3_lo[1] = 0; \
> >> + csio_scsi_fcp_cmnd((req), (void *)((uintptr_t)(wrp) + \
> >> + sizeof(struct fw_scsi_##oper##_wr))); \
> >> + \
> >> + /* Move WR pointer past command and immediate data */ \
> >> + _sgl = (struct ulptx_sgl *) ((uintptr_t)(wrp) + \
> >> + sizeof(struct fw_scsi_##oper##_wr) + \
> >> + ALIGN(_imm, 16)); \
> >> + \
> >> + /* Fill in the DSGL */ \
> >> + csio_scsi_init_ultptx_dsgl(_hw, (req), _sgl); \
> >> + \
> >> +} while (0)
> >> +
> >
> > This one has four uses of CPP keys. Just turn those into macros, and
> > leave the rest of the code in a static function.
> >
>
> So what you are suggesting is to have all the lines of the macro
> csio_scsi_init_data_wr() added into a static function, but for the ones
> with the 4 keys. csio_scsi_init_data_wr() will then invoke this new
> function. Is that correct?
>
Not sure how the above should actually look without actually doing it,
but IMHO the usage of macro just obfuscates what is going on..
If it's only used a few times, just inline the code into seperate static
functions. If it's used more than a few times, then use a single static
funciton with macro accessors for the assignment of the various '__wr'
structure members.
The larger problem with all of these macros is that you can't tell what
is a macro and what is a function.
If you need to use a CPP macro, please make sure to capitalize the name
of the macro in order to tell the difference between the two.
> >> +/* Calculate WR size needed for fw_scsi_read_wr/fw_scsi_write_wr */
> >> +#define csio_scsi_data_wrsz(req, oper, sz, imm) \
> >> +do { \
> >> + (sz) = sizeof(struct fw_scsi_##oper##_wr) + /* WR size */ \
> >> + ALIGN((imm), 16) + /* Immed data */ \
> >> + sizeof(struct ulptx_sgl); /* ulptx_sgl */ \
> >> + \
> >> + if (unlikely((req)->nsge > 1)) \
> >> + (sz) += (sizeof(struct ulptx_sge_pair) * \
> >> + (ALIGN(((req)->nsge - 1), 2) / 2)); \
> >> + /* Data SGE */ \
> >> +} while (0)
> >> +
> >> +/*
> >> + * csio_scsi_data - Create a SCSI WRITE/READ WR.
> >> + * @req: IO req structure.
> >> + * @oper: read/write
> >> + * @wrop: _READ_/_WRITE_ (string subsitutions to use with the FW bit field
> >> + * macros).
> >> + *
> >> + * Gets a WR slot in the ingress queue and initializes it with
> >> + * SCSI CMD READ/WRITE WR.
> >> + *
> >> + */
> >> +#define csio_scsi_data(req, oper, wrop) \
> >> +do { \
> >> + struct csio_wr_pair _wrp; \
> >> + uint32_t _size; \
> >> + struct csio_hw *_hw = (req)->lnode->hwp; \
> >> + struct csio_scsim *_scsim = csio_hw_to_scsim(_hw); \
> >> + \
> >> + csio_scsi_data_wrsz((req), oper, _size, _scsim->proto_cmd_len); \
> >> + _size = ALIGN(_size, 16); \
> >> + \
> >> + (req)->drv_status = csio_wr_get(_hw, (req)->eq_idx, _size, &_wrp); \
> >> + if (likely((req)->drv_status == CSIO_SUCCESS)) { \
> >> + if (likely(_wrp.size1 >= _size)) { \
> >> + /* Initialize WR in one shot */ \
> >> + csio_scsi_init_data_wr((req), oper, _wrp.addr1, \
> >> + _size, wrop); \
> >> + } else { \
> >> + uint8_t tmpwr[512]; \
> >> + /* \
> >> + * Make a temporary copy of the WR and write back \
> >> + * the copy into the WR pair. \
> >> + */ \
> >> + csio_scsi_init_data_wr((req), oper, (void *)tmpwr, \
> >> + _size, wrop); \
> >> + memcpy(_wrp.addr1, tmpwr, _wrp.size1); \
> >> + memcpy(_wrp.addr2, tmpwr + _wrp.size1, \
> >> + _size - _wrp.size1); \
> >> + } \
> >> + } \
> >> +} while (0)
> >> +
> >
> > Ditto on this one, along with the tmpwr[512] stack usage..
> >
> >> +static inline void
> >> +csio_scsi_abrt_cls(struct csio_ioreq *req, bool abort)
> >> +{
> >> + struct csio_wr_pair wrp;
> >> + struct csio_hw *hw = req->lnode->hwp;
> >> + uint32_t size = ALIGN(sizeof(struct fw_scsi_abrt_cls_wr), 16);
> >> +
> >> + req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
> >> + if (req->drv_status != CSIO_SUCCESS)
> >> + return;
> >> +
> >> + if (wrp.size1 >= size) {
> >> + /* Initialize WR in one shot */
> >> + csio_scsi_init_abrt_cls_wr(req, wrp.addr1, size, abort);
> >> + } else {
> >> + uint8_t tmpwr[512];
> >
> > Ditto here on local scope stack usage..
> >
> >> + /*
> >> + * Make a temporary copy of the WR and write back
> >> + * the copy into the WR pair.
> >> + */
> >> + csio_scsi_init_abrt_cls_wr(req, (void *)tmpwr, size, abort);
> >> + memcpy(wrp.addr1, tmpwr, wrp.size1);
> >> + memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
> >> + }
> >> +}
> >> +
> >> +/*****************************************************************************/
> >> +/* START: SCSI SM */
> >> +/*****************************************************************************/
> >> +static void
> >> +csio_scsis_uninit(struct csio_ioreq *req, enum csio_scsi_ev evt)
> >> +{
> >> + struct csio_hw *hw = req->lnode->hwp;
> >> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
> >> +
> >> + switch (evt) {
> >> +
> >> + case CSIO_SCSIE_START_IO:
> >
> > Extra space between start of first switch case
>
> OK, I will remove it. Is there any tool that catches such deviations?
> checkpath.pl didnt report it.
>
Sorry, can't help you there.. ;)
> >
> >> +
> >> +/**
> >> + * csio_queuecommand_lck - Entry point to kickstart an I/O request.
> >> + * @cmnd: The I/O request from ML.
> >> + * @done: The ML callback routine.
> >> + *
> >> + * This routine does the following:
> >> + * - Checks for HW and Rnode module readiness.
> >> + * - Gets a free ioreq structure (which is already initialized
> >> + * to uninit during its allocation).
> >> + * - Maps SG elements.
> >> + * - Initializes ioreq members.
> >> + * - Kicks off the SCSI state machine for this IO.
> >> + * - Returns busy status on error.
> >> + */
> >> +static int
> >> +csio_queuecommand_lck(struct scsi_cmnd *cmnd, void (*done)(struct scsi_cmnd *))
> >> +{
> >> + struct csio_lnode *ln = shost_priv(cmnd->device->host);
> >> + struct csio_hw *hw = csio_lnode_to_hw(ln);
> >> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
> >> + struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
> >> + struct csio_ioreq *ioreq = NULL;
> >> + unsigned long flags;
> >> + int nsge = 0;
> >> + int rv = SCSI_MLQUEUE_HOST_BUSY, nr;
> >> + csio_retval_t retval;
> >> + int cpu;
> >> + struct csio_scsi_qset *sqset;
> >> + struct fc_rport *rport = starget_to_rport(scsi_target(cmnd->device));
> >> +
> >> + if (!blk_rq_cpu_valid(cmnd->request))
> >> + cpu = smp_processor_id();
> >> + else
> >> + cpu = cmnd->request->cpu;
> >> +
> >> + sqset = &hw->sqset[ln->portid][cpu];
> >> +
> >> + nr = fc_remote_port_chkready(rport);
> >> + if (nr) {
> >> + cmnd->result = nr;
> >> + csio_inc_stats(scsim, n_rn_nr_error);
> >> + goto err_done;
> >> + }
> >> +
> >> + if (unlikely(!csio_is_hw_ready(hw))) {
> >> + cmnd->result = (DID_REQUEUE << 16);
> >> + csio_inc_stats(scsim, n_hw_nr_error);
> >> + goto err_done;
> >> + }
> >> +
> >> + /* Get req->nsge, if there are SG elements to be mapped */
> >> + nsge = scsi_dma_map(cmnd);
> >> + if (unlikely(nsge < 0)) {
> >> + csio_inc_stats(scsim, n_dmamap_error);
> >> + goto err;
> >> + }
> >> +
> >> + /* Do we support so many mappings? */
> >> + if (unlikely(nsge > scsim->max_sge)) {
> >> + csio_warn(hw,
> >> + "More SGEs than can be supported."
> >> + " SGEs: %d, Max SGEs: %d\n", nsge, scsim->max_sge);
> >> + csio_inc_stats(scsim, n_unsupp_sge_error);
> >> + goto err_dma_unmap;
> >> + }
> >> +
> >> + /* Get a free ioreq structure - SM is already set to uninit */
> >> + ioreq = csio_get_scsi_ioreq_lock(hw, scsim);
> >> + if (!ioreq) {
> >> + csio_err(hw, "Out of I/O request elements. Active #:%d\n",
> >> + scsim->stats.n_active);
> >> + csio_inc_stats(scsim, n_no_req_error);
> >> + goto err_dma_unmap;
> >> + }
> >> +
> >> + ioreq->nsge = nsge;
> >> + ioreq->lnode = ln;
> >> + ioreq->rnode = rn;
> >> + ioreq->iq_idx = sqset->iq_idx;
> >> + ioreq->eq_idx = sqset->eq_idx;
> >> + ioreq->wr_status = 0;
> >> + ioreq->drv_status = CSIO_SUCCESS;
> >> + csio_scsi_cmnd(ioreq) = (void *)cmnd;
> >> + ioreq->tmo = 0;
> >> +
> >> + switch (cmnd->sc_data_direction) {
> >> + case DMA_BIDIRECTIONAL:
> >> + ioreq->datadir = CSIO_IOREQF_DMA_BIDI;
> >> + csio_inc_stats(ln, n_control_requests);
> >> + break;
> >> + case DMA_TO_DEVICE:
> >> + ioreq->datadir = CSIO_IOREQF_DMA_WRITE;
> >> + csio_inc_stats(ln, n_output_requests);
> >> + ln->stats.n_output_bytes += scsi_bufflen(cmnd);
> >> + break;
> >> + case DMA_FROM_DEVICE:
> >> + ioreq->datadir = CSIO_IOREQF_DMA_READ;
> >> + csio_inc_stats(ln, n_input_requests);
> >> + ln->stats.n_input_bytes += scsi_bufflen(cmnd);
> >> + break;
> >> + case DMA_NONE:
> >> + ioreq->datadir = CSIO_IOREQF_DMA_NONE;
> >> + csio_inc_stats(ln, n_control_requests);
> >> + break;
> >> + default:
> >> + CSIO_DB_ASSERT(0);
> >> + break;
> >> + }
> >> +
> >> + /* Set cbfn */
> >> + ioreq->io_cbfn = csio_scsi_cbfn;
> >> +
> >> + /* Needed during abort */
> >> + cmnd->host_scribble = (unsigned char *)ioreq;
> >> + cmnd->scsi_done = done;
> >> + cmnd->SCp.Message = 0;
> >> +
> >> + /* Kick off SCSI IO SM on the ioreq */
> >> + spin_lock_irqsave(&hw->lock, flags);
> >> + retval = csio_scsi_start_io(ioreq);
> >> + spin_unlock_irqrestore(&hw->lock, flags);
> >> +
> >> + if (retval != CSIO_SUCCESS) {
> >> + csio_err(hw, "ioreq: %p couldnt be started, status:%d\n",
> >> + ioreq, retval);
> >> + csio_inc_stats(scsim, n_busy_error);
> >> + goto err_put_req;
> >> + }
> >> +
> >> + return 0;
> >> +
> >> +err_put_req:
> >> + csio_put_scsi_ioreq_lock(hw, scsim, ioreq);
> >> +err_dma_unmap:
> >> + if (nsge > 0)
> >> + scsi_dma_unmap(cmnd);
> >> +err:
> >> + return rv;
> >> +
> >> +err_done:
> >> + done(cmnd);
> >> + return 0;
> >> +}
> >> +
> >> +static DEF_SCSI_QCMD(csio_queuecommand);
> >> +
> >
> > This means that your running with the host_lock held.. I'm not sure if
> > that is really what you want to do as it really end's up killing
> > multi-lun small packet performance..
> >
> > How about dropping DEF_SCSI_QCMD usage here, and figure out what
> > actually needs to be protected by the SCSI host_lock within
> > csio_queuecommand_lck()..?
>
> It is on my TODO list for the next version of the driver, after the
> initial submission. Per the current design, we shouldnt need the
> host_lock to be held, but I would like to test this change thoroughly
> before I submit it.
>
Sure, it's really quite easy to convert and these days the majority of
high performance LLDs do run in host-lock-less mode.
IIRC the libfc based FCoE initiator driver is doing this too..
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5).
2012-08-24 20:56 ` Nicholas A. Bellinger
@ 2012-08-25 18:36 ` Naresh Kumar Inna
2012-08-25 18:43 ` Nicholas A. Bellinger
0 siblings, 1 reply; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-25 18:36 UTC (permalink / raw)
To: Nicholas A. Bellinger
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/25/2012 2:26 AM, Nicholas A. Bellinger wrote:
> On Fri, 2012-08-24 at 23:10 +0530, Naresh Kumar Inna wrote:
>> On 8/24/2012 1:18 AM, Nicholas A. Bellinger wrote:
>>> On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
>>>> This patch contains code to implement the interrupt handling and the fast
>>>> path I/O functionality. The interrupt handling includes allocation of
>>>> MSIX vectors, registering and implemeting the interrupt service routines.
>>>> The fast path I/O functionality includes posting the I/O request to firmware
>>>> via Work Requests, tracking/completing them, and handling task management
>>>> requests. SCSI midlayer host template implementation is also covered by
>>>> this patch.
>>>>
>>>> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
>>>> ---
>>>
>>> Hi Naresh,
>>>
>>> My review comments are inline below..
>>
>> Hi Nicholas,
>>
>> Thanks for taking the time to review the driver. Please find my replies
>> inline.
>>
>> Regards,
>> Naresh.
>>
>>>
>>>> drivers/scsi/csiostor/csio_isr.c | 631 ++++++++++
>>>> drivers/scsi/csiostor/csio_scsi.c | 2498 +++++++++++++++++++++++++++++++++++++
>>>> 2 files changed, 3129 insertions(+), 0 deletions(-)
>>>> create mode 100644 drivers/scsi/csiostor/csio_isr.c
>>>> create mode 100644 drivers/scsi/csiostor/csio_scsi.c
>>>>
>
> <SNIP>
>
>>>
>>>> diff --git a/drivers/scsi/csiostor/csio_scsi.c b/drivers/scsi/csiostor/csio_scsi.c
>>>> new file mode 100644
>>>> index 0000000..0f87b00
>>>> --- /dev/null
>>>> +++ b/drivers/scsi/csiostor/csio_scsi.c
>>>> +
>>>> +#define CSIO_SCSI_CMD_WR_SZ(_imm) \
>>>> + (sizeof(struct fw_scsi_cmd_wr) + /* WR size */ \
>>>> + ALIGN((_imm), 16)) /* Immed data */
>>>> +
>>>> +#define CSIO_SCSI_CMD_WR_SZ_16(_imm) \
>>>> + (ALIGN(CSIO_SCSI_CMD_WR_SZ((_imm)), 16))
>>>> +
>>>> +/*
>>>> + * csio_scsi_cmd - Create a SCSI CMD WR.
>>>> + * @req: IO req structure.
>>>> + *
>>>> + * Gets a WR slot in the ingress queue and initializes it with SCSI CMD WR.
>>>> + *
>>>> + */
>>>> +static inline void
>>>> +csio_scsi_cmd(struct csio_ioreq *req)
>>>> +{
>>>> + struct csio_wr_pair wrp;
>>>> + struct csio_hw *hw = req->lnode->hwp;
>>>> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
>>>> + uint32_t size = CSIO_SCSI_CMD_WR_SZ_16(scsim->proto_cmd_len);
>>>> +
>>>> + req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
>>>> + if (unlikely(req->drv_status != CSIO_SUCCESS))
>>>> + return;
>>>> +
>>>> + if (wrp.size1 >= size) {
>>>> + /* Initialize WR in one shot */
>>>> + csio_scsi_init_cmd_wr(req, wrp.addr1, size);
>>>> + } else {
>>>> + uint8_t tmpwr[512];
>>>
>>> Mmmm, putting this large of a buffer on the local stack is probably not
>>> a good idea.
>>>
>>> This should become an allocation.. If it's a hot path then you'll
>>> probably want to set this up before-hand.
>>>
>>
>> The else switch above is entered only when we near the end of the DMA
>> ring. This is not a frequent occurrence. If it is a strict no-no to have
>> so many on-stack bytes, I have to think of a way re-work it to use
>> pre-allocated memory. Please let me know.
>>
>
> Considering it's used a number of times, it would be better to just
> figure out a sensible manner to pre-allocate this, especially if it's
> only a single occurrence to an individual ring held under a lock.
>
I will see what I can do - I can probably add these bytes into every
queues metadata structure.
>>>> + /*
>>>> + * Make a temporary copy of the WR and write back
>>>> + * the copy into the WR pair.
>>>> + */
>>>> + csio_scsi_init_cmd_wr(req, (void *)tmpwr, size);
>>>> + memcpy(wrp.addr1, tmpwr, wrp.size1);
>>>> + memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
>>>> + }
>>>> +}
>>>> +
>>>> +/*
>>>> + * The following is fast path code. Therefore it is inlined with multi-line
>>>> + * macros using name substitution, thus avoiding if-else switches for
>>>> + * operation (read/write), as well as serving the purpose of code re-use.
>>>> + */
>>>> +/*
>>>> + * csio_scsi_init_ulptx_dsgl - Fill in a ULP_TX_SC_DSGL
>>>> + * @hw: HW module
>>>> + * @req: IO request
>>>> + * @sgl: ULP TX SGL pointer.
>>>> + *
>>>> + */
>>>> +#define csio_scsi_init_ultptx_dsgl(hw, req, sgl) \
>>>> +do { \
>>>> + struct ulptx_sge_pair *_sge_pair = NULL; \
>>>> + struct scatterlist *_sgel; \
>>>> + uint32_t _i = 0; \
>>>> + uint32_t _xfer_len; \
>>>> + struct list_head *_tmp; \
>>>> + struct csio_dma_buf *_dma_buf; \
>>>> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
>>>> + \
>>>> + (sgl)->cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) | ULPTX_MORE | \
>>>> + ULPTX_NSGE((req)->nsge)); \
>>>> + /* Now add the data SGLs */ \
>>>> + if (likely(!(req)->dcopy)) { \
>>>> + scsi_for_each_sg(scmnd, _sgel, (req)->nsge, _i) { \
>>>> + if (_i == 0) { \
>>>> + (sgl)->addr0 = cpu_to_be64( \
>>>> + sg_dma_address(_sgel)); \
>>>> + (sgl)->len0 = cpu_to_be32( \
>>>> + sg_dma_len(_sgel)); \
>>>> + _sge_pair = \
>>>> + (struct ulptx_sge_pair *)((sgl) + 1); \
>>>> + continue; \
>>>> + } \
>>>> + if ((_i - 1) & 0x1) { \
>>>> + _sge_pair->addr[1] = cpu_to_be64( \
>>>> + sg_dma_address(_sgel)); \
>>>> + _sge_pair->len[1] = cpu_to_be32( \
>>>> + sg_dma_len(_sgel)); \
>>>> + _sge_pair++; \
>>>> + } else { \
>>>> + _sge_pair->addr[0] = cpu_to_be64( \
>>>> + sg_dma_address(_sgel)); \
>>>> + _sge_pair->len[0] = cpu_to_be32( \
>>>> + sg_dma_len(_sgel)); \
>>>> + } \
>>>> + } \
>>>> + } else { \
>>>> + /* Program sg elements with driver's DDP buffer */ \
>>>> + _xfer_len = scsi_bufflen(scmnd); \
>>>> + list_for_each(_tmp, &(req)->gen_list) { \
>>>> + _dma_buf = (struct csio_dma_buf *)_tmp; \
>>>> + if (_i == 0) { \
>>>> + (sgl)->addr0 = cpu_to_be64(_dma_buf->paddr); \
>>>> + (sgl)->len0 = cpu_to_be32( \
>>>> + min(_xfer_len, _dma_buf->len)); \
>>>> + _sge_pair = \
>>>> + (struct ulptx_sge_pair *)((sgl) + 1); \
>>>> + } \
>>>> + else if ((_i - 1) & 0x1) { \
>>>> + _sge_pair->addr[1] = cpu_to_be64( \
>>>> + _dma_buf->paddr); \
>>>> + _sge_pair->len[1] = cpu_to_be32( \
>>>> + min(_xfer_len, _dma_buf->len)); \
>>>> + _sge_pair++; \
>>>> + } else { \
>>>> + _sge_pair->addr[0] = cpu_to_be64( \
>>>> + _dma_buf->paddr); \
>>>> + _sge_pair->len[0] = cpu_to_be32( \
>>>> + min(_xfer_len, _dma_buf->len)); \
>>>> + } \
>>>> + _xfer_len -= min(_xfer_len, _dma_buf->len); \
>>>> + _i++; \
>>>> + } \
>>>> + } \
>>>> +} while (0)
>>>> +
>>>
>>> I don't see any reason why this can't just be a static function..? Why
>>> is the macro usage necessary here..?
>>
>> OK, I will make this static-inline.
>>
>>>
>>>> +/*
>>>> + * csio_scsi_init_data_wr - Initialize the READ/WRITE SCSI WR.
>>>> + * @req: IO req structure.
>>>> + * @oper: read/write
>>>> + * @wrp: DMA location to place the payload.
>>>> + * @size: Size of WR (including FW WR + immed data + rsp SG entry + data SGL
>>>> + * @wrop: _READ_/_WRITE_
>>>> + *
>>>> + * Wrapper for populating fw_scsi_read_wr/fw_scsi_write_wr.
>>>> + */
>>>> +#define csio_scsi_init_data_wr(req, oper, wrp, size, wrop) \
>>>> +do { \
>>>> + struct csio_hw *_hw = (req)->lnode->hwp; \
>>>> + struct csio_rnode *_rn = (req)->rnode; \
>>>> + struct fw_scsi_##oper##_wr *__wr = (struct fw_scsi_##oper##_wr *)(wrp);\
>>>> + struct ulptx_sgl *_sgl; \
>>>> + struct csio_dma_buf *_dma_buf; \
>>>> + uint8_t _imm = csio_hw_to_scsim(_hw)->proto_cmd_len; \
>>>> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
>>>> + \
>>>> + __wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI##wrop##WR) | \
>>>> + FW_SCSI##wrop##WR_IMMDLEN(_imm)); \
>>>> + __wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(_rn->flowid) | \
>>>> + FW_WR_LEN16( \
>>>> + CSIO_ROUNDUP((size), 16))); \
>>>> + __wr->cookie = (uintptr_t) (req); \
>>>> + __wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(_hw, \
>>>> + (req)->iq_idx));\
>>>> + __wr->tmo_val = (uint8_t)((req)->tmo); \
>>>> + __wr->use_xfer_cnt = 1; \
>>>> + __wr->xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
>>>> + __wr->ini_xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
>>>> + /* Get RSP DMA buffer */ \
>>>> + _dma_buf = &(req)->dma_buf; \
>>>> + \
>>>> + /* Prepare RSP SGL */ \
>>>> + __wr->rsp_dmalen = cpu_to_be32(_dma_buf->len); \
>>>> + __wr->rsp_dmaaddr = cpu_to_be64(_dma_buf->paddr); \
>>>> + \
>>>> + __wr->r4 = 0; \
>>>> + \
>>>> + __wr->u.fcoe.ctl_pri = 0; \
>>>> + __wr->u.fcoe.cp_en_class = 0; \
>>>> + __wr->u.fcoe.r3_lo[0] = 0; \
>>>> + __wr->u.fcoe.r3_lo[1] = 0; \
>>>> + csio_scsi_fcp_cmnd((req), (void *)((uintptr_t)(wrp) + \
>>>> + sizeof(struct fw_scsi_##oper##_wr))); \
>>>> + \
>>>> + /* Move WR pointer past command and immediate data */ \
>>>> + _sgl = (struct ulptx_sgl *) ((uintptr_t)(wrp) + \
>>>> + sizeof(struct fw_scsi_##oper##_wr) + \
>>>> + ALIGN(_imm, 16)); \
>>>> + \
>>>> + /* Fill in the DSGL */ \
>>>> + csio_scsi_init_ultptx_dsgl(_hw, (req), _sgl); \
>>>> + \
>>>> +} while (0)
>>>> +
>>>
>>> This one has four uses of CPP keys. Just turn those into macros, and
>>> leave the rest of the code in a static function.
>>>
>>
>> So what you are suggesting is to have all the lines of the macro
>> csio_scsi_init_data_wr() added into a static function, but for the ones
>> with the 4 keys. csio_scsi_init_data_wr() will then invoke this new
>> function. Is that correct?
>>
>
> Not sure how the above should actually look without actually doing it,
> but IMHO the usage of macro just obfuscates what is going on..
>
> If it's only used a few times, just inline the code into seperate static
> functions. If it's used more than a few times, then use a single static
> funciton with macro accessors for the assignment of the various '__wr'
> structure members.
>
> The larger problem with all of these macros is that you can't tell what
> is a macro and what is a function.
>
> If you need to use a CPP macro, please make sure to capitalize the name
> of the macro in order to tell the difference between the two.
>
OK, I will see what I can do to convert this macro into a function.
>>>> +/* Calculate WR size needed for fw_scsi_read_wr/fw_scsi_write_wr */
>>>> +#define csio_scsi_data_wrsz(req, oper, sz, imm) \
>>>> +do { \
>>>> + (sz) = sizeof(struct fw_scsi_##oper##_wr) + /* WR size */ \
>>>> + ALIGN((imm), 16) + /* Immed data */ \
>>>> + sizeof(struct ulptx_sgl); /* ulptx_sgl */ \
>>>> + \
>>>> + if (unlikely((req)->nsge > 1)) \
>>>> + (sz) += (sizeof(struct ulptx_sge_pair) * \
>>>> + (ALIGN(((req)->nsge - 1), 2) / 2)); \
>>>> + /* Data SGE */ \
>>>> +} while (0)
>>>> +
>>>> +/*
>>>> + * csio_scsi_data - Create a SCSI WRITE/READ WR.
>>>> + * @req: IO req structure.
>>>> + * @oper: read/write
>>>> + * @wrop: _READ_/_WRITE_ (string subsitutions to use with the FW bit field
>>>> + * macros).
>>>> + *
>>>> + * Gets a WR slot in the ingress queue and initializes it with
>>>> + * SCSI CMD READ/WRITE WR.
>>>> + *
>>>> + */
>>>> +#define csio_scsi_data(req, oper, wrop) \
>>>> +do { \
>>>> + struct csio_wr_pair _wrp; \
>>>> + uint32_t _size; \
>>>> + struct csio_hw *_hw = (req)->lnode->hwp; \
>>>> + struct csio_scsim *_scsim = csio_hw_to_scsim(_hw); \
>>>> + \
>>>> + csio_scsi_data_wrsz((req), oper, _size, _scsim->proto_cmd_len); \
>>>> + _size = ALIGN(_size, 16); \
>>>> + \
>>>> + (req)->drv_status = csio_wr_get(_hw, (req)->eq_idx, _size, &_wrp); \
>>>> + if (likely((req)->drv_status == CSIO_SUCCESS)) { \
>>>> + if (likely(_wrp.size1 >= _size)) { \
>>>> + /* Initialize WR in one shot */ \
>>>> + csio_scsi_init_data_wr((req), oper, _wrp.addr1, \
>>>> + _size, wrop); \
>>>> + } else { \
>>>> + uint8_t tmpwr[512]; \
>>>> + /* \
>>>> + * Make a temporary copy of the WR and write back \
>>>> + * the copy into the WR pair. \
>>>> + */ \
>>>> + csio_scsi_init_data_wr((req), oper, (void *)tmpwr, \
>>>> + _size, wrop); \
>>>> + memcpy(_wrp.addr1, tmpwr, _wrp.size1); \
>>>> + memcpy(_wrp.addr2, tmpwr + _wrp.size1, \
>>>> + _size - _wrp.size1); \
>>>> + } \
>>>> + } \
>>>> +} while (0)
>>>> +
>>>
>>> Ditto on this one, along with the tmpwr[512] stack usage..
>>>
>>>> +static inline void
>>>> +csio_scsi_abrt_cls(struct csio_ioreq *req, bool abort)
>>>> +{
>>>> + struct csio_wr_pair wrp;
>>>> + struct csio_hw *hw = req->lnode->hwp;
>>>> + uint32_t size = ALIGN(sizeof(struct fw_scsi_abrt_cls_wr), 16);
>>>> +
>>>> + req->drv_status = csio_wr_get(hw, req->eq_idx, size, &wrp);
>>>> + if (req->drv_status != CSIO_SUCCESS)
>>>> + return;
>>>> +
>>>> + if (wrp.size1 >= size) {
>>>> + /* Initialize WR in one shot */
>>>> + csio_scsi_init_abrt_cls_wr(req, wrp.addr1, size, abort);
>>>> + } else {
>>>> + uint8_t tmpwr[512];
>>>
>>> Ditto here on local scope stack usage..
>>>
>>>> + /*
>>>> + * Make a temporary copy of the WR and write back
>>>> + * the copy into the WR pair.
>>>> + */
>>>> + csio_scsi_init_abrt_cls_wr(req, (void *)tmpwr, size, abort);
>>>> + memcpy(wrp.addr1, tmpwr, wrp.size1);
>>>> + memcpy(wrp.addr2, tmpwr + wrp.size1, size - wrp.size1);
>>>> + }
>>>> +}
>>>> +
>>>> +/*****************************************************************************/
>>>> +/* START: SCSI SM */
>>>> +/*****************************************************************************/
>>>> +static void
>>>> +csio_scsis_uninit(struct csio_ioreq *req, enum csio_scsi_ev evt)
>>>> +{
>>>> + struct csio_hw *hw = req->lnode->hwp;
>>>> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
>>>> +
>>>> + switch (evt) {
>>>> +
>>>> + case CSIO_SCSIE_START_IO:
>>>
>>> Extra space between start of first switch case
>>
>> OK, I will remove it. Is there any tool that catches such deviations?
>> checkpath.pl didnt report it.
>>
>
> Sorry, can't help you there.. ;)
>
>>>
>>>> +
>>>> +/**
>>>> + * csio_queuecommand_lck - Entry point to kickstart an I/O request.
>>>> + * @cmnd: The I/O request from ML.
>>>> + * @done: The ML callback routine.
>>>> + *
>>>> + * This routine does the following:
>>>> + * - Checks for HW and Rnode module readiness.
>>>> + * - Gets a free ioreq structure (which is already initialized
>>>> + * to uninit during its allocation).
>>>> + * - Maps SG elements.
>>>> + * - Initializes ioreq members.
>>>> + * - Kicks off the SCSI state machine for this IO.
>>>> + * - Returns busy status on error.
>>>> + */
>>>> +static int
>>>> +csio_queuecommand_lck(struct scsi_cmnd *cmnd, void (*done)(struct scsi_cmnd *))
>>>> +{
>>>> + struct csio_lnode *ln = shost_priv(cmnd->device->host);
>>>> + struct csio_hw *hw = csio_lnode_to_hw(ln);
>>>> + struct csio_scsim *scsim = csio_hw_to_scsim(hw);
>>>> + struct csio_rnode *rn = (struct csio_rnode *)(cmnd->device->hostdata);
>>>> + struct csio_ioreq *ioreq = NULL;
>>>> + unsigned long flags;
>>>> + int nsge = 0;
>>>> + int rv = SCSI_MLQUEUE_HOST_BUSY, nr;
>>>> + csio_retval_t retval;
>>>> + int cpu;
>>>> + struct csio_scsi_qset *sqset;
>>>> + struct fc_rport *rport = starget_to_rport(scsi_target(cmnd->device));
>>>> +
>>>> + if (!blk_rq_cpu_valid(cmnd->request))
>>>> + cpu = smp_processor_id();
>>>> + else
>>>> + cpu = cmnd->request->cpu;
>>>> +
>>>> + sqset = &hw->sqset[ln->portid][cpu];
>>>> +
>>>> + nr = fc_remote_port_chkready(rport);
>>>> + if (nr) {
>>>> + cmnd->result = nr;
>>>> + csio_inc_stats(scsim, n_rn_nr_error);
>>>> + goto err_done;
>>>> + }
>>>> +
>>>> + if (unlikely(!csio_is_hw_ready(hw))) {
>>>> + cmnd->result = (DID_REQUEUE << 16);
>>>> + csio_inc_stats(scsim, n_hw_nr_error);
>>>> + goto err_done;
>>>> + }
>>>> +
>>>> + /* Get req->nsge, if there are SG elements to be mapped */
>>>> + nsge = scsi_dma_map(cmnd);
>>>> + if (unlikely(nsge < 0)) {
>>>> + csio_inc_stats(scsim, n_dmamap_error);
>>>> + goto err;
>>>> + }
>>>> +
>>>> + /* Do we support so many mappings? */
>>>> + if (unlikely(nsge > scsim->max_sge)) {
>>>> + csio_warn(hw,
>>>> + "More SGEs than can be supported."
>>>> + " SGEs: %d, Max SGEs: %d\n", nsge, scsim->max_sge);
>>>> + csio_inc_stats(scsim, n_unsupp_sge_error);
>>>> + goto err_dma_unmap;
>>>> + }
>>>> +
>>>> + /* Get a free ioreq structure - SM is already set to uninit */
>>>> + ioreq = csio_get_scsi_ioreq_lock(hw, scsim);
>>>> + if (!ioreq) {
>>>> + csio_err(hw, "Out of I/O request elements. Active #:%d\n",
>>>> + scsim->stats.n_active);
>>>> + csio_inc_stats(scsim, n_no_req_error);
>>>> + goto err_dma_unmap;
>>>> + }
>>>> +
>>>> + ioreq->nsge = nsge;
>>>> + ioreq->lnode = ln;
>>>> + ioreq->rnode = rn;
>>>> + ioreq->iq_idx = sqset->iq_idx;
>>>> + ioreq->eq_idx = sqset->eq_idx;
>>>> + ioreq->wr_status = 0;
>>>> + ioreq->drv_status = CSIO_SUCCESS;
>>>> + csio_scsi_cmnd(ioreq) = (void *)cmnd;
>>>> + ioreq->tmo = 0;
>>>> +
>>>> + switch (cmnd->sc_data_direction) {
>>>> + case DMA_BIDIRECTIONAL:
>>>> + ioreq->datadir = CSIO_IOREQF_DMA_BIDI;
>>>> + csio_inc_stats(ln, n_control_requests);
>>>> + break;
>>>> + case DMA_TO_DEVICE:
>>>> + ioreq->datadir = CSIO_IOREQF_DMA_WRITE;
>>>> + csio_inc_stats(ln, n_output_requests);
>>>> + ln->stats.n_output_bytes += scsi_bufflen(cmnd);
>>>> + break;
>>>> + case DMA_FROM_DEVICE:
>>>> + ioreq->datadir = CSIO_IOREQF_DMA_READ;
>>>> + csio_inc_stats(ln, n_input_requests);
>>>> + ln->stats.n_input_bytes += scsi_bufflen(cmnd);
>>>> + break;
>>>> + case DMA_NONE:
>>>> + ioreq->datadir = CSIO_IOREQF_DMA_NONE;
>>>> + csio_inc_stats(ln, n_control_requests);
>>>> + break;
>>>> + default:
>>>> + CSIO_DB_ASSERT(0);
>>>> + break;
>>>> + }
>>>> +
>>>> + /* Set cbfn */
>>>> + ioreq->io_cbfn = csio_scsi_cbfn;
>>>> +
>>>> + /* Needed during abort */
>>>> + cmnd->host_scribble = (unsigned char *)ioreq;
>>>> + cmnd->scsi_done = done;
>>>> + cmnd->SCp.Message = 0;
>>>> +
>>>> + /* Kick off SCSI IO SM on the ioreq */
>>>> + spin_lock_irqsave(&hw->lock, flags);
>>>> + retval = csio_scsi_start_io(ioreq);
>>>> + spin_unlock_irqrestore(&hw->lock, flags);
>>>> +
>>>> + if (retval != CSIO_SUCCESS) {
>>>> + csio_err(hw, "ioreq: %p couldnt be started, status:%d\n",
>>>> + ioreq, retval);
>>>> + csio_inc_stats(scsim, n_busy_error);
>>>> + goto err_put_req;
>>>> + }
>>>> +
>>>> + return 0;
>>>> +
>>>> +err_put_req:
>>>> + csio_put_scsi_ioreq_lock(hw, scsim, ioreq);
>>>> +err_dma_unmap:
>>>> + if (nsge > 0)
>>>> + scsi_dma_unmap(cmnd);
>>>> +err:
>>>> + return rv;
>>>> +
>>>> +err_done:
>>>> + done(cmnd);
>>>> + return 0;
>>>> +}
>>>> +
>>>> +static DEF_SCSI_QCMD(csio_queuecommand);
>>>> +
>>>
>>> This means that your running with the host_lock held.. I'm not sure if
>>> that is really what you want to do as it really end's up killing
>>> multi-lun small packet performance..
>>>
>>> How about dropping DEF_SCSI_QCMD usage here, and figure out what
>>> actually needs to be protected by the SCSI host_lock within
>>> csio_queuecommand_lck()..?
>>
>> It is on my TODO list for the next version of the driver, after the
>> initial submission. Per the current design, we shouldnt need the
>> host_lock to be held, but I would like to test this change thoroughly
>> before I submit it.
>>
>
> Sure, it's really quite easy to convert and these days the majority of
> high performance LLDs do run in host-lock-less mode.
>
> IIRC the libfc based FCoE initiator driver is doing this too..
>
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5).
2012-08-25 18:36 ` Naresh Kumar Inna
@ 2012-08-25 18:43 ` Nicholas A. Bellinger
2012-08-29 7:47 ` Naresh Kumar Inna
0 siblings, 1 reply; 33+ messages in thread
From: Nicholas A. Bellinger @ 2012-08-25 18:43 UTC (permalink / raw)
To: Naresh Kumar Inna
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On Sun, 2012-08-26 at 00:06 +0530, Naresh Kumar Inna wrote:
> On 8/25/2012 2:26 AM, Nicholas A. Bellinger wrote:
> > On Fri, 2012-08-24 at 23:10 +0530, Naresh Kumar Inna wrote:
> >> On 8/24/2012 1:18 AM, Nicholas A. Bellinger wrote:
> >>> On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
> >>>> This patch contains code to implement the interrupt handling and the fast
> >>>> path I/O functionality. The interrupt handling includes allocation of
> >>>> MSIX vectors, registering and implemeting the interrupt service routines.
> >>>> The fast path I/O functionality includes posting the I/O request to firmware
> >>>> via Work Requests, tracking/completing them, and handling task management
> >>>> requests. SCSI midlayer host template implementation is also covered by
> >>>> this patch.
> >>>>
> >>>> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
> >>>> ---
> >>>
> >>> Hi Naresh,
> >>>
> >>> My review comments are inline below..
> >>
> >> Hi Nicholas,
> >>
> >> Thanks for taking the time to review the driver. Please find my replies
> >> inline.
> >>
> >> Regards,
> >> Naresh.
> >>
> >>>
> >>>> drivers/scsi/csiostor/csio_isr.c | 631 ++++++++++
> >>>> drivers/scsi/csiostor/csio_scsi.c | 2498 +++++++++++++++++++++++++++++++++++++
> >>>> 2 files changed, 3129 insertions(+), 0 deletions(-)
> >>>> create mode 100644 drivers/scsi/csiostor/csio_isr.c
> >>>> create mode 100644 drivers/scsi/csiostor/csio_scsi.c
> >>>>
<SNIP>
> >>>
> >>>> +/*
> >>>> + * csio_scsi_init_data_wr - Initialize the READ/WRITE SCSI WR.
> >>>> + * @req: IO req structure.
> >>>> + * @oper: read/write
> >>>> + * @wrp: DMA location to place the payload.
> >>>> + * @size: Size of WR (including FW WR + immed data + rsp SG entry + data SGL
> >>>> + * @wrop: _READ_/_WRITE_
> >>>> + *
> >>>> + * Wrapper for populating fw_scsi_read_wr/fw_scsi_write_wr.
> >>>> + */
> >>>> +#define csio_scsi_init_data_wr(req, oper, wrp, size, wrop) \
> >>>> +do { \
> >>>> + struct csio_hw *_hw = (req)->lnode->hwp; \
> >>>> + struct csio_rnode *_rn = (req)->rnode; \
> >>>> + struct fw_scsi_##oper##_wr *__wr = (struct fw_scsi_##oper##_wr *)(wrp);\
> >>>> + struct ulptx_sgl *_sgl; \
> >>>> + struct csio_dma_buf *_dma_buf; \
> >>>> + uint8_t _imm = csio_hw_to_scsim(_hw)->proto_cmd_len; \
> >>>> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
> >>>> + \
> >>>> + __wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI##wrop##WR) | \
> >>>> + FW_SCSI##wrop##WR_IMMDLEN(_imm)); \
> >>>> + __wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(_rn->flowid) | \
> >>>> + FW_WR_LEN16( \
> >>>> + CSIO_ROUNDUP((size), 16))); \
> >>>> + __wr->cookie = (uintptr_t) (req); \
> >>>> + __wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(_hw, \
> >>>> + (req)->iq_idx));\
> >>>> + __wr->tmo_val = (uint8_t)((req)->tmo); \
> >>>> + __wr->use_xfer_cnt = 1; \
> >>>> + __wr->xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
> >>>> + __wr->ini_xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
> >>>> + /* Get RSP DMA buffer */ \
> >>>> + _dma_buf = &(req)->dma_buf; \
> >>>> + \
> >>>> + /* Prepare RSP SGL */ \
> >>>> + __wr->rsp_dmalen = cpu_to_be32(_dma_buf->len); \
> >>>> + __wr->rsp_dmaaddr = cpu_to_be64(_dma_buf->paddr); \
> >>>> + \
> >>>> + __wr->r4 = 0; \
> >>>> + \
> >>>> + __wr->u.fcoe.ctl_pri = 0; \
> >>>> + __wr->u.fcoe.cp_en_class = 0; \
> >>>> + __wr->u.fcoe.r3_lo[0] = 0; \
> >>>> + __wr->u.fcoe.r3_lo[1] = 0; \
> >>>> + csio_scsi_fcp_cmnd((req), (void *)((uintptr_t)(wrp) + \
> >>>> + sizeof(struct fw_scsi_##oper##_wr))); \
> >>>> + \
> >>>> + /* Move WR pointer past command and immediate data */ \
> >>>> + _sgl = (struct ulptx_sgl *) ((uintptr_t)(wrp) + \
> >>>> + sizeof(struct fw_scsi_##oper##_wr) + \
> >>>> + ALIGN(_imm, 16)); \
> >>>> + \
> >>>> + /* Fill in the DSGL */ \
> >>>> + csio_scsi_init_ultptx_dsgl(_hw, (req), _sgl); \
> >>>> + \
> >>>> +} while (0)
> >>>> +
> >>>
> >>> This one has four uses of CPP keys. Just turn those into macros, and
> >>> leave the rest of the code in a static function.
> >>>
> >>
> >> So what you are suggesting is to have all the lines of the macro
> >> csio_scsi_init_data_wr() added into a static function, but for the ones
> >> with the 4 keys. csio_scsi_init_data_wr() will then invoke this new
> >> function. Is that correct?
> >>
> >
> > Not sure how the above should actually look without actually doing it,
> > but IMHO the usage of macro just obfuscates what is going on..
> >
> > If it's only used a few times, just inline the code into seperate static
> > functions. If it's used more than a few times, then use a single static
> > funciton with macro accessors for the assignment of the various '__wr'
> > structure members.
> >
> > The larger problem with all of these macros is that you can't tell what
> > is a macro and what is a function.
> >
> > If you need to use a CPP macro, please make sure to capitalize the name
> > of the macro in order to tell the difference between the two.
> >
>
> OK, I will see what I can do to convert this macro into a function.
>
Also please change all of the remaining macro names to be capitalized so
someone reading the code knows the difference between function or macro.
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5).
2012-08-25 18:43 ` Nicholas A. Bellinger
@ 2012-08-29 7:47 ` Naresh Kumar Inna
0 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-29 7:47 UTC (permalink / raw)
To: Nicholas A. Bellinger
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/26/2012 12:13 AM, Nicholas A. Bellinger wrote:
> On Sun, 2012-08-26 at 00:06 +0530, Naresh Kumar Inna wrote:
>> On 8/25/2012 2:26 AM, Nicholas A. Bellinger wrote:
>>> On Fri, 2012-08-24 at 23:10 +0530, Naresh Kumar Inna wrote:
>>>> On 8/24/2012 1:18 AM, Nicholas A. Bellinger wrote:
>>>>> On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
>>>>>> This patch contains code to implement the interrupt handling and the fast
>>>>>> path I/O functionality. The interrupt handling includes allocation of
>>>>>> MSIX vectors, registering and implemeting the interrupt service routines.
>>>>>> The fast path I/O functionality includes posting the I/O request to firmware
>>>>>> via Work Requests, tracking/completing them, and handling task management
>>>>>> requests. SCSI midlayer host template implementation is also covered by
>>>>>> this patch.
>>>>>>
>>>>>> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
>>>>>> ---
>>>>>
>>>>> Hi Naresh,
>>>>>
>>>>> My review comments are inline below..
>>>>
>>>> Hi Nicholas,
>>>>
>>>> Thanks for taking the time to review the driver. Please find my replies
>>>> inline.
>>>>
>>>> Regards,
>>>> Naresh.
>>>>
>>>>>
>>>>>> drivers/scsi/csiostor/csio_isr.c | 631 ++++++++++
>>>>>> drivers/scsi/csiostor/csio_scsi.c | 2498 +++++++++++++++++++++++++++++++++++++
>>>>>> 2 files changed, 3129 insertions(+), 0 deletions(-)
>>>>>> create mode 100644 drivers/scsi/csiostor/csio_isr.c
>>>>>> create mode 100644 drivers/scsi/csiostor/csio_scsi.c
>>>>>>
>
> <SNIP>
>
>>>>>
>>>>>> +/*
>>>>>> + * csio_scsi_init_data_wr - Initialize the READ/WRITE SCSI WR.
>>>>>> + * @req: IO req structure.
>>>>>> + * @oper: read/write
>>>>>> + * @wrp: DMA location to place the payload.
>>>>>> + * @size: Size of WR (including FW WR + immed data + rsp SG entry + data SGL
>>>>>> + * @wrop: _READ_/_WRITE_
>>>>>> + *
>>>>>> + * Wrapper for populating fw_scsi_read_wr/fw_scsi_write_wr.
>>>>>> + */
>>>>>> +#define csio_scsi_init_data_wr(req, oper, wrp, size, wrop) \
>>>>>> +do { \
>>>>>> + struct csio_hw *_hw = (req)->lnode->hwp; \
>>>>>> + struct csio_rnode *_rn = (req)->rnode; \
>>>>>> + struct fw_scsi_##oper##_wr *__wr = (struct fw_scsi_##oper##_wr *)(wrp);\
>>>>>> + struct ulptx_sgl *_sgl; \
>>>>>> + struct csio_dma_buf *_dma_buf; \
>>>>>> + uint8_t _imm = csio_hw_to_scsim(_hw)->proto_cmd_len; \
>>>>>> + struct scsi_cmnd *scmnd = csio_scsi_cmnd((req)); \
>>>>>> + \
>>>>>> + __wr->op_immdlen = cpu_to_be32(FW_WR_OP(FW_SCSI##wrop##WR) | \
>>>>>> + FW_SCSI##wrop##WR_IMMDLEN(_imm)); \
>>>>>> + __wr->flowid_len16 = cpu_to_be32(FW_WR_FLOWID(_rn->flowid) | \
>>>>>> + FW_WR_LEN16( \
>>>>>> + CSIO_ROUNDUP((size), 16))); \
>>>>>> + __wr->cookie = (uintptr_t) (req); \
>>>>>> + __wr->iqid = (uint16_t)cpu_to_be16(csio_q_physiqid(_hw, \
>>>>>> + (req)->iq_idx));\
>>>>>> + __wr->tmo_val = (uint8_t)((req)->tmo); \
>>>>>> + __wr->use_xfer_cnt = 1; \
>>>>>> + __wr->xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
>>>>>> + __wr->ini_xfer_cnt = cpu_to_be32(scsi_bufflen(scmnd)); \
>>>>>> + /* Get RSP DMA buffer */ \
>>>>>> + _dma_buf = &(req)->dma_buf; \
>>>>>> + \
>>>>>> + /* Prepare RSP SGL */ \
>>>>>> + __wr->rsp_dmalen = cpu_to_be32(_dma_buf->len); \
>>>>>> + __wr->rsp_dmaaddr = cpu_to_be64(_dma_buf->paddr); \
>>>>>> + \
>>>>>> + __wr->r4 = 0; \
>>>>>> + \
>>>>>> + __wr->u.fcoe.ctl_pri = 0; \
>>>>>> + __wr->u.fcoe.cp_en_class = 0; \
>>>>>> + __wr->u.fcoe.r3_lo[0] = 0; \
>>>>>> + __wr->u.fcoe.r3_lo[1] = 0; \
>>>>>> + csio_scsi_fcp_cmnd((req), (void *)((uintptr_t)(wrp) + \
>>>>>> + sizeof(struct fw_scsi_##oper##_wr))); \
>>>>>> + \
>>>>>> + /* Move WR pointer past command and immediate data */ \
>>>>>> + _sgl = (struct ulptx_sgl *) ((uintptr_t)(wrp) + \
>>>>>> + sizeof(struct fw_scsi_##oper##_wr) + \
>>>>>> + ALIGN(_imm, 16)); \
>>>>>> + \
>>>>>> + /* Fill in the DSGL */ \
>>>>>> + csio_scsi_init_ultptx_dsgl(_hw, (req), _sgl); \
>>>>>> + \
>>>>>> +} while (0)
>>>>>> +
>>>>>
>>>>> This one has four uses of CPP keys. Just turn those into macros, and
>>>>> leave the rest of the code in a static function.
>>>>>
>>>>
>>>> So what you are suggesting is to have all the lines of the macro
>>>> csio_scsi_init_data_wr() added into a static function, but for the ones
>>>> with the 4 keys. csio_scsi_init_data_wr() will then invoke this new
>>>> function. Is that correct?
>>>>
>>>
>>> Not sure how the above should actually look without actually doing it,
>>> but IMHO the usage of macro just obfuscates what is going on..
>>>
>>> If it's only used a few times, just inline the code into seperate static
>>> functions. If it's used more than a few times, then use a single static
>>> funciton with macro accessors for the assignment of the various '__wr'
>>> structure members.
>>>
>>> The larger problem with all of these macros is that you can't tell what
>>> is a macro and what is a function.
>>>
>>> If you need to use a CPP macro, please make sure to capitalize the name
>>> of the macro in order to tell the difference between the two.
>>>
>>
>> OK, I will see what I can do to convert this macro into a function.
>>
>
> Also please change all of the remaining macro names to be capitalized so
> someone reading the code knows the difference between function or macro.
>
Hi Nicholas,
Did you get a chance to review the rest of the driver patches?
Thanks,
Naresh.
^ permalink raw reply [flat|nested] 33+ messages in thread
* [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
` (4 preceding siblings ...)
2012-08-23 22:27 ` [PATCH 5/8] csiostor: Chelsio FCoE offload driver submission (sources part 5) Naresh Kumar Inna
@ 2012-08-23 22:27 ` Naresh Kumar Inna
2012-08-23 18:15 ` Love, Robert W
2012-08-23 19:58 ` Nicholas A. Bellinger
2012-08-23 22:27 ` [PATCH 7/8] csiostor: Chelsio FCoE offload driver submission (headers part 2) Naresh Kumar Inna
` (3 subsequent siblings)
9 siblings, 2 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
This patch contains the first set of the header files for csiostor driver.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/csiostor/csio_defs.h | 143 ++++++
drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
drivers/scsi/csiostor/csio_hw.h | 668 ++++++++++++++++++++++++
drivers/scsi/csiostor/csio_init.h | 158 ++++++
4 files changed, 1812 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_defs.h
create mode 100644 drivers/scsi/csiostor/csio_fcoe_proto.h
create mode 100644 drivers/scsi/csiostor/csio_hw.h
create mode 100644 drivers/scsi/csiostor/csio_init.h
diff --git a/drivers/scsi/csiostor/csio_defs.h b/drivers/scsi/csiostor/csio_defs.h
new file mode 100644
index 0000000..4f1c713
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_defs.h
@@ -0,0 +1,143 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __CSIO_DEFS_H__
+#define __CSIO_DEFS_H__
+
+#include <linux/kernel.h>
+#include <linux/timer.h>
+#include <linux/list.h>
+#include <linux/bug.h>
+#include <linux/pci.h>
+#include <linux/jiffies.h>
+
+/* Function returns */
+enum csio_retval {
+ CSIO_SUCCESS = 0,
+ CSIO_INVAL = 1,
+ CSIO_BUSY = 2,
+ CSIO_NOSUPP = 3,
+ CSIO_TIMEOUT = 4,
+ CSIO_NOMEM = 5,
+ CSIO_NOPERM = 6,
+ CSIO_RETRY = 7,
+ CSIO_EPROTO = 8,
+ CSIO_EIO = 9,
+ CSIO_CANCELLED = 10,
+};
+
+#define csio_retval_t enum csio_retval
+
+enum {
+ CSIO_FALSE = 0,
+ CSIO_TRUE = 1,
+};
+
+#define CSIO_ROUNDUP(__v, __r) (((__v) + (__r) - 1) / (__r))
+#define CSIO_INVALID_IDX 0xFFFFFFFF
+#define csio_inc_stats(elem, val) ((elem)->stats.val++)
+#define csio_dec_stats(elem, val) ((elem)->stats.val--)
+#define csio_valid_wwn(__n) ((*__n >> 4) == 0x5 ? CSIO_TRUE : \
+ CSIO_FALSE)
+#define CSIO_WORD_TO_BYTE 4
+
+static inline int
+csio_list_deleted(struct list_head *list)
+{
+ return ((list->next == list) && (list->prev == list));
+}
+
+#define csio_list_next(elem) (((struct list_head *)(elem))->next)
+#define csio_list_prev(elem) (((struct list_head *)(elem))->prev)
+
+#define csio_deq_from_head(head, elem) \
+do { \
+ if (!list_empty(head)) { \
+ *((struct list_head **)(elem)) = csio_list_next((head)); \
+ csio_list_next((head)) = \
+ csio_list_next(csio_list_next((head))); \
+ csio_list_prev(csio_list_next((head))) = (head); \
+ INIT_LIST_HEAD(*((struct list_head **)(elem))); \
+ } else \
+ *((struct list_head **)(elem)) = (struct list_head *)NULL;\
+} while (0)
+
+#define csio_deq_from_tail(head, elem) \
+do { \
+ if (!list_empty(head)) { \
+ *((struct list_head **)(elem)) = csio_list_prev((head)); \
+ csio_list_prev((head)) = \
+ csio_list_prev(csio_list_prev((head))); \
+ csio_list_next(csio_list_prev((head))) = (head); \
+ INIT_LIST_HEAD(*((struct list_head **)(elem))); \
+ } else \
+ *((struct list_head **)(elem)) = (struct list_head *)NULL;\
+} while (0)
+
+/* State machine */
+typedef void (*csio_sm_state_t)(void *, uint32_t);
+
+struct csio_sm {
+ struct list_head sm_list;
+ csio_sm_state_t sm_state;
+};
+
+#define csio_init_state(__smp, __state) \
+ (((struct csio_sm *)(__smp))->sm_state = (csio_sm_state_t)(__state))
+
+#define csio_set_state(__smp, __state) \
+ (((struct csio_sm *)(__smp))->sm_state = (csio_sm_state_t)(__state))
+
+
+#define csio_post_event(__smp, __evt) \
+ (((struct csio_sm *)(__smp))->sm_state((__smp), (uint32_t)(__evt)))
+
+#define csio_get_state(__smp) (((struct csio_sm *)(__smp))->sm_state)
+
+#define csio_match_state(__smp, __state) \
+ (csio_get_state((__smp)) == (csio_sm_state_t)(__state))
+
+#define CSIO_ASSERT(cond) \
+do { \
+ if (unlikely(!((cond)))) \
+ BUG(); \
+} while (0)
+
+#ifdef __CSIO_DEBUG__
+#define CSIO_DB_ASSERT(__c) CSIO_ASSERT((__c))
+#else
+#define CSIO_DB_ASSERT(__c)
+#endif
+
+#endif /* ifndef __CSIO_DEFS_H__ */
diff --git a/drivers/scsi/csiostor/csio_fcoe_proto.h b/drivers/scsi/csiostor/csio_fcoe_proto.h
new file mode 100644
index 0000000..32e6f43
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_fcoe_proto.h
@@ -0,0 +1,843 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __CSIO_FCOE_PROTO_H__
+#define __CSIO_FCOE_PROTO_H__
+
+/* FC header type */
+#define FC_TYPE_ELS_DATA 0x1
+#define FC_TYPE_CT_DATA 0x20
+#define FC_TYPE_FCP_DATA 0x8
+
+/* FC header rctl */
+#define FC_RCTL_ELS_REQ 0x22
+#define FC_RCTL_ELS_RSP 0x23
+#define FC_RCTL_FCP_CMND 0x6
+
+/* Well known Fibre channel Address */
+#define FDMI_DID 0xFFFFFA /* Management server */
+#define NS_DID 0xFFFFFC /* Name server */
+#define FABCTL_DID 0xFFFFFD /* Fabric Controller */
+#define FABRIC_DID 0xFFFFFE /* Fabric Login */
+#define BCAST_DID 0xFFFFFF /* Broadcast */
+#define UNKNOWN_DID 0x000000 /* Unknown DID */
+#define DID_MASK 0xFFFFFF /* DID Mask */
+#define WK_DID_MASK 0xFFFFF0 /* Well known did mask */
+
+/* FC4 Device Data Frame - TYPE */
+#define FC4_FCP_TYPE 0x8 /* FCP */
+
+/* MAX FC Payload */
+#define MAX_FC_PAYLOAD 2112
+
+/*Service option: Shift & Mask bits defines */
+#define SP_CLASS_SUPPORT_EN 1 /* Class support enable */
+#define S_SP_CLASS_SUPPORT 7
+#define M_SP_CLASS_SUPPORT 1
+#define V_SP_CLASS_SUPPORT(x) ((x) << S_SP_CLASS_SUPPORT)
+#define G_SP_CLASS_SUPPORT(x) \
+ (((x) >> S_SP_CLASS_SUPPORT) & M_SP_CLASS_SUPPORT)
+
+/* Class service parameters */
+struct csio_class_sp {
+ uint8_t serv_option; /* Service option */
+ uint8_t rsvd1;
+ uint8_t init_ctl_option; /* initiator cntl option */
+ uint8_t rsvd2;
+ uint8_t rcv_ctl_option; /* receiver cntl option */
+ uint8_t rsvd3;
+ uint16_t rcv_data_sz; /* receive data size */
+ uint16_t concurrent_seq; /* Total concurent sequence */
+ uint16_t ee_credit; /* EE credit */
+ uint16_t openseq_per_xchg; /* Open sequence per exch */
+ uint16_t rsvd4;
+};
+
+/* Common service parameters defines */
+
+/* FC Phy version */
+#define FC_PH_VER3 0x20
+
+/* WORD1 (31:16) flags: shift & mask bit defines */
+/* NPIV support */
+#define MULTIPLE_NPORT_ID_SUPPORT_EN 1
+#define S_MULTIPLE_NPORT_ID_SUPPORT 15
+#define M_MULTIPLE_NPORT_ID_SUPPORT 1
+#define V_MULTIPLE_NPORT_ID_SUPPORT(x) ((x) << S_MULTIPLE_NPORT_ID_SUPPORT)
+#define G_MULTIPLE_NPORT_ID_SUPPORT(x) (((x) >> S_MULTIPLE_NPORT_ID_SUPPORT) \
+ & M_MULTIPLE_NPORT_ID_SUPPORT)
+
+/* Continuously increasing relative offset */
+#define CONTI_INCR_OFFSET_SUPPORT_EN 1
+#define S_CONTI_INCR_OFFSET_SUPPORT 15
+#define M_CONTI_INCR_OFFSET_SUPPORT 1
+#define V_CONTI_INCR_OFFSET_SUPPORT(x) ((x) << S_CONTI_INCR_OFFSET_SUPPORT)
+#define G_CONTI_INCR_OFFSET_SUPPORT(x) (((x) >> S_CONTI_INCR_OFFSET_SUPPORT) \
+ & M_CONTI_INCR_OFFSET_SUPPORT)
+/* Continuously increasing relative offset */
+#define CLEAN_ADDR_EN 1
+#define S_CLEAN_ADDR 15
+#define M_CLEAN_ADDR 1
+#define V_CLEAN_ADDR(x) ((x) << S_CLEAN_ADDR)
+#define G_CLEAN_ADDR(x) (((x) >> S_CLEAN_ADDR) & M_CLEAN_ADDR)
+
+/* NPIV supported by Fabric */
+#define NPIV_SUPPORTED_EN 1
+#define S_NPIV_SUPPORTED 13
+#define M_NPIV_SUPPORTED 1
+#define V_NPIV_SUPPORTED(x) ((x) << S_NPIV_SUPPORTED)
+#define G_NPIV_SUPPORTED(x) (((x) >> S_NPIV_SUPPORTED) & M_NPIV_SUPPORTED)
+
+/* N_Port or F_Port */
+#define FABRIC_PORT 1
+#define S_FABRIC_PORT 12
+#define M_FABRIC_PORT 1
+#define V_FABRIC_PORT(x) ((x) << S_FABRIC_PORT)
+#define G_FABRIC_PORT(x) (((x) >> S_FABRIC_PORT) & M_FABRIC_PORT)
+
+/* Alternate B2B credit management support */
+#define ALT_B2B_CREDIT_MGMT_SUPPORT_EN 1
+#define S_ALT_B2B_CREDIT_MGMT_SUPPORT 11
+#define M_ALT_B2B_CREDIT_MGMT_SUPPORT 1
+#define V_ALT_B2B_CREDIT_MGMT_SUPPORT(x) \
+ ((x) << S_ALT_B2B_CREDIT_MGMT_SUPPORT)
+#define G_ALT_B2B_CREDIT_MGMT_SUPPORT(x) \
+ (((x) >> S_ALT_B2B_CREDIT_MGMT_SUPPORT) & M_ALT_B2B_CREDIT_MGMT_SUPPORT)
+
+
+/* WORD2 (31: 0) : shift and mask bit defines */
+#define S_MAX_SEQ_CNT 16
+#define M_MAX_SEQ_CNT 0xFFFF
+#define V_MAX_SEQ_CNT(x) ((x) << S_MAX_SEQ_CNT)
+#define G_MAX_SEQ_CNT(x) (((x) >> S_MAX_SEQ_CNT) & M_MAX_SEQ_CNT)
+
+#define S_REL_OFFSET_BY_CATEGORY 0
+#define M_REL_OFFSET_BY_CATEGORY 0xFFFF
+#define V_REL_OFFSET_BY_CATEGORY(x) ((x) << S_REL_OFFSET_BY_CATEGORY)
+#define G_REL_OFFSET_BY_CATEGORY(x) \
+ (((x) >> S_REL_OFFSET_BY_CATEGORY) & M_REL_OFFSET_BY_CATEGORY)
+
+/* Common service parameters */
+struct csio_cmn_sp {
+ uint8_t hi_ver; /* High PH version */
+ uint8_t lo_ver; /* low PH version */
+ uint16_t bb_credit; /* B2B credit */
+ uint16_t word1_flags; /* Word1 Flags (31:16)*/
+ uint16_t rcv_sz; /* Receive data size */
+ union {
+ uint32_t maxsq_reloff; /* Max seq / Relative offset */
+ uint32_t r_a_tov; /* R_A_TOV */
+ } un1;
+ uint32_t e_d_tov; /*E_D_TOV */
+};
+
+struct csio_service_parms {
+ struct csio_cmn_sp csp; /* Common service parms */
+ uint8_t wwpn[8]; /* WWPN */
+ uint8_t wwnn[8]; /* WWNN */
+ struct csio_class_sp clsp[4]; /* Class service params */
+ uint8_t vvl[16]; /* Vendor version level */
+};
+
+/* Common Transport (CT) defines */
+#define CT_BASIC_IU_LEN 0x10
+#define CT_REVISION 0x1
+
+/* GS Types */
+#define CT_GS_MGMT_SERVICE 0xFA
+#define CT_GS_TIME_SERVICE 0xFB
+#define CT_GS_DIR_SERVICE 0xFC
+#define CT_GS_FABRIC_CNTL_SERVICE 0xFD
+
+/* Directory service Subtypes */
+#define CT_DIR_SERVICE_NAME_SERVER 0x02
+
+/* FDMI MGMT service Subtypes */
+#define CT_FDMI_HBA_MGMT_SERVER 0x10
+
+/* CT Response code */
+#define CT_RESPONSE_FS_RJT 0x8001
+#define CT_RESPONSE_FS_ACC 0x8002
+
+/* CT Reason code */
+#define CT_NO_ADDITIONAL_EXPLANATION 0x00
+#define CT_INVALID_COMMAND 0x01
+#define CT_INVALID_VERSION_LEVEL 0x02
+#define CT_LOGICAL_ERROR 0x03
+#define CT_INVALID_IU_SIZE 0x04
+#define CT_LOGICAL_BUSY 0x05
+#define CT_PROTOCOL_ERROR 0x07
+#define CT_UNABLE_TO_PERFORM_CMD_REQ 0x09
+#define CT_CMD_NOT_SUPPORTED 0x0B
+#define CT_VENDOR_UNIQUE 0xff
+
+/* Name Server explanation for Reason code CT_UNABLE_TO_PERFORM_CMD_REQ */
+#define CT_NS_PORT_ID_NOT_REG 0x01
+#define CT_NS_PORT_NAME_NOT_REG 0x02
+#define CT_NS_NODE_NAME_NOT_REG 0x03
+#define CT_NS_CLASS_OF_SERVICE_NOT_REG 0x04
+#define CT_NS_IP_ADDRESS_NOT_REG 0x05
+#define CT_NS_IPA_NOT_REG 0x06
+#define CT_NS_FC4_TYPES_NOT_REG 0x07
+#define CT_NS_SYMBOLIC_PORT_NAME_NOT_REG 0x08
+#define CT_NS_SYMBOLIC_NODE_NAME_NOT_REG 0x09
+#define CT_NS_PORT_TYPE_NOT_REG 0x0A
+#define CT_NS_ACCESS_DENIED 0x10
+#define CT_NS_INVALID_PORT_ID 0x11
+#define CT_NS_DATABASE_EMPTY 0x12
+
+/* Name Server Command Codes */
+#define CT_NS_GA_NXT 0x0100
+#define CT_NS_GPN_ID 0x0112
+#define CT_NS_GNN_ID 0x0113
+#define CT_NS_GCS_ID 0x0114
+#define CT_NS_GFT_ID 0x0117
+#define CT_NS_GSPN_ID 0x0118
+#define CT_NS_GPT_ID 0x011A
+#define CT_NS_GFF_ID 0x011F
+#define CT_NS_GID_PN 0x0121
+#define CT_NS_GID_NN 0x0131
+#define CT_NS_GIP_NN 0x0135
+#define CT_NS_GIPA_NN 0x0136
+#define CT_NS_GSNN_NN 0x0139
+#define CT_NS_GNN_IP 0x0153
+#define CT_NS_GIPA_IP 0x0156
+#define CT_NS_GID_FT 0x0171
+#define CT_NS_GPN_FT 0x0172
+#define CT_NS_GID_PT 0x01A1
+#define CT_NS_RPN_ID 0x0212
+#define CT_NS_RNN_ID 0x0213
+#define CT_NS_RCS_ID 0x0214
+#define CT_NS_RFT_ID 0x0217
+#define CT_NS_RSPN_ID 0x0218
+#define CT_NS_RPT_ID 0x021A
+#define CT_NS_RFF_ID 0x021F
+#define CT_NS_RIP_NN 0x0235
+#define CT_NS_RIPA_NN 0x0236
+#define CT_NS_RSNN_NN 0x0239
+#define CT_NS_DA_ID 0x0300
+
+/* FDMI HBA management Server Command Codes */
+#define CT_FDMI_HBA_GRHL 0x100 /* Get registered HBA list */
+#define CT_FDMI_HBA_GHAT 0x101 /* Get HBA attributes */
+#define CT_FDMI_HBA_GRPL 0x102 /* Get registered Port list */
+#define CT_FDMI_HBA_GPAT 0x110 /* Get Port attributes */
+#define CT_FDMI_HBA_RHBA 0x200 /* Register HBA */
+#define CT_FDMI_HBA_RHAT 0x201 /* Register HBA atttributes */
+#define CT_FDMI_HBA_RPRT 0x210 /* Register Port */
+#define CT_FDMI_HBA_RPA 0x211 /* Register Port attributes */
+#define CT_FDMI_HBA_DHBA 0x300 /* De-register HBA */
+#define CT_FDMI_HBA_DPRT 0x310 /* De-register Port */
+
+/* HBA Attribute Types */
+#define NODE_NAME 0x1
+#define MANUFACTURER 0x2
+#define SERIAL_NUMBER 0x3
+#define MODEL 0x4
+#define MODEL_DESCRIPTION 0x5
+#define HARDWARE_VERSION 0x6
+#define DRIVER_VERSION 0x7
+#define OPTION_ROM_VERSION 0x8
+#define FIRMWARE_VERSION 0x9
+#define OS_NAME_VERSION 0xa
+#define MAX_CT_PAYLOAD_LEN 0xb
+
+/* Port Attrubute Types */
+#define SUPPORTED_FC4_TYPES 0x1
+#define SUPPORTED_SPEED 0x2
+#define PORT_SPEED 0x3
+#define MAX_FRAME_LEN 0x4
+#define OS_DEVICE_NAME 0x5
+#define HOST_NAME 0x6
+
+#define CSIO_HBA_PORTSPEED_1GBIT 0x0001 /* 1 GBit/sec */
+#define CSIO_HBA_PORTSPEED_2GBIT 0x0002 /* 2 GBit/sec */
+#define CSIO_HBA_PORTSPEED_4GBIT 0x0008 /* 4 GBit/sec */
+#define CSIO_HBA_PORTSPEED_10GBIT 0x0004 /* 10 GBit/sec */
+#define CSIO_HBA_PORTSPEED_8GBIT 0x0010 /* 8 GBit/sec */
+#define CSIO_HBA_PORTSPEED_16GBIT 0x0020 /* 16 GBit/sec */
+#define CSIO_HBA_PORTSPEED_UNKNOWN 0x0800 /* Unknown */
+
+/* Port Types */
+#define CT_PORT_TYPE_N_PORT 0x01
+#define CT_PORT_TYPE_NL_PORT 0x02
+#define CT_PORT_TYPE_FNL_PORT 0x03
+#define CT_PORT_TYPE_IP 0x04
+#define CT_PORT_TYPE_FCP 0x08
+#define CT_PORT_TYPE_NX_PORT 0x7F
+#define CT_PORT_TYPE_F_PORT 0x81
+#define CT_PORT_TYPE_FL_PORT 0x82
+#define CT_PORT_TYPE_E_PORT 0x84
+
+/* FC4 Feature bit defination */
+#define FC4_FEATURE_TARGET_EN 1
+#define S_FC4_FEATURE_TARGET 0
+#define M_FC4_FEATURE_TARGET 1
+#define V_FC4_FEATURE_TARGET(x) ((x) << S_FC4_FEATURE_TARGET)
+#define G_FC4_FEATURE_TARGET(x) \
+ (((x) >> S_FC4_FEATURE_TARGET) & M_FC4_FEATURE_TARGET)
+
+#define FC4_FEATURE_INITIATOR_EN 1
+#define S_FC4_FEATURE_INITIATOR 1
+#define M_FC4_FEATURE_INITIATOR 1
+#define V_FC4_FEATURE_INITIATOR(x) ((x) << S_FC4_FEATURE_INITIATOR)
+#define G_FC4_FEATURE_INITIATOR(x) \
+ (((x) >> S_FC4_FEATURE_INITIATOR) & M_FC4_FEATURE_INITIATOR)
+
+/* GPN_FT ACC Control bit defination */
+#define GPN_FT_ACC_CONTROL_EN 1
+#define S_GPN_FT_ACC_CONTROL 31
+#define M_GPN_FT_ACC_CONTROL 1
+#define V_GPN_FT_ACC_CONTROL(x) ((x) << S_GPN_FT_ACC_CONTROL)
+#define G_GPN_FT_ACC_CONTROL(x) \
+ (((x) >> S_GPN_FT_ACC_CONTROL) & M_GPN_FT_ACC_CONTROL)
+
+/* CT command */
+struct csio_ct_cmd {
+ uint8_t rev; /* Revision */
+ uint8_t in_id[3]; /* Unused */
+ uint8_t gs_type; /* Type of service */
+ uint8_t gs_subtype; /* Sub type */
+ uint8_t opt; /* Options */
+ uint8_t rsvd1;
+ uint16_t op; /* Command or response code */
+ uint16_t size; /* Maximum or Residual size */
+ uint8_t rsvd2;
+ uint8_t reason_code; /* Reason code */
+ uint8_t explanation; /* Explanation code */
+ uint8_t vendor_unique; /* Vendor specific reason code */
+
+ union {
+ uint32_t port_id; /* Port_id list for GID_FT ACC */
+
+ struct gid_ft {
+ uint8_t port_type; /* Port Type */
+ uint8_t domain_scope; /* Domain scope */
+ uint8_t area_scope; /* Area scope */
+ uint8_t fc4_type; /* FC4 Type = FCP(0x8) */
+ } gid_ft;
+
+ struct gpn_ft {
+ uint8_t rsvd;
+ uint8_t domain_scope; /* Domain scope */
+ uint8_t area_scope; /* Area scope */
+ uint8_t fc4_type; /* FC4 Type = FCP(0x8) */
+ } gpn_ft;
+
+ /* Port_id & Port name list for GPN_FT ACC */
+ struct gpn_ft_acc {
+ uint32_t port_id; /* port id */
+ uint32_t rsvd;
+ uint8_t wwpn[8]; /* Port name */
+ } gpn_ft_acc;
+
+ struct rft_id {
+ uint32_t port_id; /* port id */
+ uint16_t rsvd1;
+ uint8_t fcp; /* FCP Type */
+ uint8_t rsvd2;
+ uint8_t rsvd3[28];
+ } rft_id;
+
+ struct rnn_id {
+ uint32_t port_id; /* Port id */
+ uint8_t wwnn[8]; /* Node name */
+ } rnn_id;
+
+ struct da_id {
+ uint32_t port_id; /* Port id */
+ } da_id;
+
+ struct rff_id {
+ uint32_t port_id; /* Port id */
+ uint8_t rsvd1[2];
+ uint8_t fc4_fbits; /* FC4 feature bits */
+ uint8_t fc4_type; /* FC4 Type = FCP(0x8) */
+ } rff_id;
+ } un;
+};
+
+#define csio_ct_rsp(cp) (((struct csio_ct_cmd *) cp)->op)
+#define csio_ct_reason(cp) (((struct csio_ct_cmd *) cp)->reason_code)
+#define csio_ct_expl(cp) (((struct csio_ct_cmd *) cp)->explanation)
+#define csio_ct_get_pld(cp) ((void *)(((uint8_t *)cp) + CT_BASIC_IU_LEN))
+
+static inline void
+csio_fill_ct_iu(void *buf, uint8_t type, uint8_t sub_type,
+ uint16_t op)
+{
+ struct csio_ct_cmd *cmd = (struct csio_ct_cmd *) buf;
+ cmd->rev = CT_REVISION;
+ cmd->gs_type = type;
+ cmd->gs_subtype = sub_type;
+ cmd->op = op;
+}
+
+/* FDMI HBA cmd */
+
+/* Attribute entry */
+struct csio_attrib_entry {
+ uint16_t type; /* Entry type */
+ uint16_t len; /* Entry length */
+ union {
+ uint8_t string[256]; /* Attribute value in string */
+ uint32_t integer; /* Attribute value in integer */
+ } val;
+};
+
+
+/* attribute block */
+struct csio_attrib_block {
+ uint32_t entry_count; /* Entry count */
+ struct csio_attrib_entry entry; /* list of attributes */
+};
+
+/* HBA identifier */
+struct csio_hba_identifier {
+ uint8_t wwpn[8]; /* Port name */
+};
+
+/* Port entry */
+struct csio_port_entry {
+ uint8_t wwpn[8]; /* Port name */
+};
+
+/* register port list */
+struct csio_reg_port_list {
+ uint32_t entry_count; /* Entry count */
+ struct csio_port_entry entry; /* list of port entry */
+};
+
+/* register HBA */
+struct csio_reg_hba {
+ struct csio_hba_identifier id; /* HBA identifier */
+ struct csio_reg_port_list port_list; /* port entry list */
+};
+
+/* register HBA attributes */
+struct csio_reg_hba_attrib {
+ struct csio_hba_identifier id; /* HBA identifier */
+ struct csio_attrib_block attrib_list; /* Attribute list */
+};
+
+/* register Port attributes */
+struct csio_reg_port_attrib {
+ uint8_t wwpn[8]; /* Port name */
+ struct csio_attrib_block attrib_list; /* Attribute list */
+};
+
+/* Get register hba list(GRHL) accept payload */
+struct csio_grhl_acc_pld {
+ uint32_t entry_count; /* Entry count */
+ struct csio_hba_identifier id; /* HBA identifier list */
+};
+
+/* Get register port list(GRPL) accept payload */
+struct csio_grpl_acc_pld {
+ uint32_t entry_count; /* Entry count */
+ struct csio_port_entry entry; /* port entry list */
+};
+
+/* Get port attributes (GPAT) accept payload */
+struct csio_gpat_acc_pld {
+ struct csio_attrib_block attrib_list; /* Attribute list */
+};
+
+/* ELS CMD HDR length */
+#define ELS_CMD_HDR_LEN 0x4
+
+/* ELS COMMAND CODES */
+#define ELS_CMD_CODE_MASK 0xff
+#define ELS_CMD_CODE_LS_RJT 0x01
+#define ELS_CMD_CODE_ACC 0x02
+#define ELS_CMD_CODE_PLOGI 0x03
+#define ELS_CMD_CODE_FLOGI 0x04
+#define ELS_CMD_CODE_LOGO 0x05
+#define ELS_CMD_CODE_RES 0x08
+#define ELS_CMD_CODE_RSS 0x09
+#define ELS_CMD_CODE_RSI 0x0A
+#define ELS_CMD_CODE_ESTS 0x0B
+#define ELS_CMD_CODE_ESTC 0x0C
+#define ELS_CMD_CODE_ADVC 0x0D
+#define ELS_CMD_CODE_RTV 0x0E
+#define ELS_CMD_CODE_RLS 0x0F
+#define ELS_CMD_CODE_ECHO 0x10
+#define ELS_CMD_CODE_TEST 0x11
+#define ELS_CMD_CODE_PRLI 0x20
+#define ELS_CMD_CODE_PRLO 0x21
+#define ELS_CMD_CODE_PDISC 0x50
+#define ELS_CMD_CODE_FDISC 0x51
+#define ELS_CMD_CODE_ADISC 0x52
+#define ELS_CMD_CODE_RPS 0x56
+#define ELS_CMD_CODE_RPL 0x57
+#define ELS_CMD_CODE_RSCN 0x61
+#define ELS_CMD_CODE_SCR 0x62
+#define ELS_CMD_CODE_RNID 0x78
+#define ELS_CMD_CODE_LIRR 0x7A
+
+/* LS_RJT reason codes */
+#define LS_RJT_INVALID_CMD 0x01
+#define LS_RJT_LOGICAL_ERR 0x03
+#define LS_RJT_LOGICAL_BSY 0x05
+#define LS_RJT_PROTOCOL_ERR 0x07
+#define LS_RJT_UNABLE_TPC 0x09
+#define LS_RJT_CMD_UNSUPPORTED 0x0B
+#define LS_RJT_VENDOR_UNIQUE 0xFF
+
+/* LS_RJT reason explanation */
+#define LS_RJT_EXPL_NONE 0x00
+#define LS_RJT_EXPL_SPARM_OPTIONS 0x01
+#define LS_RJT_EXPL_SPARM_ICTL 0x03
+#define LS_RJT_EXPL_SPARM_RCTL 0x05
+#define LS_RJT_EXPL_SPARM_RCV_SIZE 0x07
+#define LS_RJT_EXPL_SPARM_CONCUR_SEQ 0x09
+#define LS_RJT_EXPL_SPARM_CREDIT 0x0B
+#define LS_RJT_EXPL_INVALID_PNAME 0x0D
+#define LS_RJT_EXPL_INVALID_NNAME 0x0E
+#define LS_RJT_EXPL_INVALID_CSP 0x0F
+#define LS_RJT_EXPL_INVALID_ASSOC_HDR 0x11
+#define LS_RJT_EXPL_ASSOC_HDR_REQ 0x13
+#define LS_RJT_EXPL_INVALID_O_SID 0x15
+#define LS_RJT_EXPL_INVALID_OX_RX 0x17
+#define LS_RJT_EXPL_CMD_IN_PROGRESS 0x19
+#define LS_RJT_EXPL_PORT_LOGIN_REQ 0x1E
+#define LS_RJT_EXPL_INVALID_NPORT_ID 0x1F
+#define LS_RJT_EXPL_INVALID_SEQ_ID 0x21
+#define LS_RJT_EXPL_INVALID_XCHG 0x23
+#define LS_RJT_EXPL_INACTIVE_XCHG 0x25
+#define LS_RJT_EXPL_RQ_REQUIRED 0x27
+#define LS_RJT_EXPL_OUT_OF_RESOURCE 0x29
+#define LS_RJT_EXPL_CANT_GIVE_DATA 0x2A
+#define LS_RJT_EXPL_REQ_UNSUPPORTED 0x2C
+
+/* PRLI Page and Payload length */
+#define PRLI_PAGE_LEN 0x10
+#define PRLI_PAYLOAD_LEN 0x14
+
+/* PRLI/PRLO PROCESS FLAGS */
+/* Originator Proc Associator valid */
+#define PRLILO_ORG_PA_VALID 1
+#define S_PRLILO_ORG_PA_VALID 7
+#define M_PRLILO_ORG_PA_VALID 1
+#define V_PRLILO_ORG_PA_VALID(x) ((x) << S_PRLILO_ORG_PA_VALID)
+#define G_PRLILO_ORG_PA_VALID(x) \
+ (((x) >> S_PRLILO_ORG_PA_VALID) & M_PRLILO_ORG_PA_VALID)
+
+/* Responder Proc Associator valid */
+#define PRLILO_RSP_PA_VALID 1
+#define S_PRLILO_RSP_PA_VALID 6
+#define M_PRLILO_RSP_PA_VALID 1
+#define V_PRLILO_RSP_PA_VALID(x) ((x) << S_PRLILO_RSP_PA_VALID)
+#define G_PRLILO_RSP_PA_VALID(x) \
+ (((x) >> S_PRLILO_RSP_PA_VALID) & M_PRLILO_RSP_PA_VALID)
+
+/* Image pair established */
+#define PRLILO_IMG_PAIR_ESTB 1
+#define S_PRLILO_IMG_PAIR_ESTB 5
+#define M_PRLILO_IMG_PAIR_ESTB 1
+#define V_PRLILO_IMG_PAIR_ESTB(x) ((x) << S_PRLILO_IMG_PAIR_ESTB)
+#define G_PRLILO_IMG_PAIR_ESTB(x) \
+ (((x) >> S_PRLILO_IMG_PAIR_ESTB) & M_PRLILO_IMG_PAIR_ESTB)
+
+/* PRLI Response code */
+#define PRLI_RSP_CODE
+#define S_PRLI_RSP 0
+#define M_PRLI_RSP 0xf
+#define V_PRLI_RSP(x) ((x) << S_PRLI_RSP)
+#define G_PRLI_RSP(x) \
+ (((x) >> S_PRLI_RSP) & M_PRLI_RSP)
+
+/* PRLI Service Parameter flags */
+/* FCP Write xfer ready disabled */
+#define PRLI_FCP_WRITE_XFER_RD_DIS 1
+#define S_PRLI_FCP_WRITE_XFER_RD_DIS 0
+#define M_PRLI_FCP_WRITE_XFER_RD_DIS 1
+#define V_PRLI_FCP_WRITE_XFER_RD_DIS(x) ((x) << S_PRLI_FCP_WRITE_XFER_RD_DIS)
+#define G_PRLI_FCP_WRITE_XFER_RD_DIS(x) \
+ (((x) >> S_PRLI_FCP_WRITE_XFER_RD_DIS) & M_PRLI_FCP_WRITE_XFER_RD_DIS)
+
+/* FCP read xfer ready disabled */
+#define PRLI_FCP_READ_XFER_RD_DIS 1
+#define S_PRLI_FCP_READ_XFER_RD_DIS 1
+#define M_PRLI_FCP_READ_XFER_RD_DIS 1
+#define V_PRLI_FCP_READ_XFER_RD_DIS(x) ((x) << S_PRLI_FCP_READ_XFER_RD_DIS)
+#define G_PRLI_FCP_READ_XFER_RD_DIS(x) \
+ (((x) >> S_PRLI_FCP_READ_XFER_RD_DIS) & M_PRLI_FCP_READ_XFER_RD_DIS)
+
+/* FCP data response mix disabled */
+#define PRLI_FCP_DATA_RSP_MIX_DIS 1
+#define S_PRLI_FCP_DATA_RSP_MIX_DIS 2
+#define M_PRLI_FCP_DATA_RSP_MIX_DIS 1
+#define V_PRLI_FCP_DATA_RSP_MIX_DIS(x) ((x) << S_PRLI_FCP_DATA_RSP_MIX_DIS)
+#define G_PRLI_FCP_DATA_RSP_MIX_DIS(x) \
+ (((x) >> S_PRLI_FCP_DATA_RSP_MIX_DIS) & M_PRLI_FCP_DATA_RSP_MIX_DIS)
+
+/* FCP cmd data mix disabled */
+#define PRLI_FCP_CMD_DATA_MIX_DIS 1
+#define S_PRLI_FCP_CMD_DATA_MIX_DIS 3
+#define M_PRLI_FCP_CMD_DATA_MIX_DIS 1
+#define V_PRLI_FCP_CMD_DATA_MIX_DIS(x) ((x) << S_PRLI_FCP_CMD_DATA_MIX_DIS)
+#define G_PRLI_FCP_CMD_DATA_MIX_DIS(x) \
+ (((x) >> S_PRLI_FCP_CMD_DATA_MIX_DIS) & M_PRLI_FCP_CMD_DATA_MIX_DIS)
+
+/* FCP Target function */
+#define PRLI_FCP_TARGET_FUNC 1
+#define S_PRLI_FCP_TARGET_FUNC 4
+#define M_PRLI_FCP_TARGET_FUNC 1
+#define V_PRLI_FCP_TARGET_FUNC(x) ((x) << S_PRLI_FCP_TARGET_FUNC)
+#define G_PRLI_FCP_TARGET_FUNC(x) \
+ (((x) >> S_PRLI_FCP_TARGET_FUNC) & M_PRLI_FCP_TARGET_FUNC)
+
+/* FCP Initiator function */
+#define PRLI_FCP_INITIATOR_FUNC 1
+#define S_PRLI_FCP_INITIATOR_FUNC 5
+#define M_PRLI_FCP_INITIATOR_FUNC 1
+#define V_PRLI_FCP_INITIATOR_FUNC(x) ((x) << S_PRLI_FCP_INITIATOR_FUNC)
+#define G_PRLI_FCP_INITIATOR_FUNC(x) \
+ (((x) >> S_PRLI_FCP_INITIATOR_FUNC) & M_PRLI_FCP_INITIATOR_FUNC)
+
+/* FCP Data overlay */
+#define PRLI_FCP_DATA_OVERLAY 1
+#define S_PRLI_FCP_DATA_OVERLAY 6
+#define M_PRLI_FCP_DATA_OVERLAY 1
+#define V_PRLI_FCP_DATA_OVERLAY(x) ((x) << S_PRLI_FCP_DATA_OVERLAY)
+#define G_PRLI_FCP_DATA_OVERLAY(x) \
+ (((x) >> S_PRLI_FCP_DATA_OVERLAY) & M_PRLI_FCP_DATA_OVERLAY)
+
+/* FCP confirmed completion */
+#define PRLI_FCP_CONF_COMPL_ALLOWED 1
+#define S_PRLI_FCP_CONF_COMPL_ALLOWED 7
+#define M_PRLI_FCP_CONF_COMPL_ALLOWED 1
+#define V_PRLI_FCP_CONF_COMPL_ALLOWED(x) ((x) << S_PRLI_FCP_CONF_COMPL_ALLOWED)
+#define G_PRLI_FCP_CONF_COMPL_ALLOWED(x) \
+ (((x) >> S_PRLI_FCP_CONF_COMPL_ALLOWED) & M_PRLI_FCP_CONF_COMPL_ALLOWED)
+
+/* FCP retry */
+#define PRLI_FCP_RETRY 1
+#define S_PRLI_FCP_RETRY 8
+#define M_PRLI_FCP_RETRY 1
+#define V_PRLI_FCP_RETRY(x) ((x) << S_PRLI_FCP_RETRY)
+#define G_PRLI_FCP_RETRY(x) (((x) >> S_PRLI_FCP_RETRY) & M_PRLI_FCP_RETRY)
+
+/* FCP Task retry id request */
+#define PRLI_FCP_TASK_ID_REQ 1
+#define S_PRLI_FCP_TASK_ID_REQ 9
+#define M_PRLI_FCP_TASK_ID_REQ 1
+#define V_PRLI_FCP_TASK_ID_REQ(x) ((x) << S_PRLI_FCP_TASK_ID_REQ)
+#define G_PRLI_FCP_TASK_ID_REQ(x) \
+ (((x) >> S_PRLI_FCP_TASK_ID_REQ) & M_PRLI_FCP_TASK_ID_REQ)
+
+/* SCR Function */
+#define SCR_FUNCTION_FABRIC 0x01
+#define SCR_FUNCTION_NPORT 0x02
+#define SCR_FUNCTION_FULL 0x03
+#define SCR_FUNCTION_CLEAR 0xFF
+
+/* PRLI accept response code */
+#define PRLI_REQ_COMPLETED 0x1
+#define PRLI_RES_UNAVAIL 0x2
+#define PRLI_INIT_NOT_COMPLETE 0x3
+#define PRLI_RESP_PA_NO_FOUND 0x4
+#define PRLI_REQ_CONDITIONAL 0x5
+#define PRLI_RECP_PRECONFIG 0x6
+#define PRLI_MULTIPAGE_REQ_FAILED 0x7
+#define PRLI_INVALID_SP 0x8
+
+
+/* MAX RETIRES */
+#define MAX_ELS_RETRY 3
+#define ECM_MIN_TMO 1000 /* Minimum timeout value for req */
+
+/* ELS request */
+struct csio_els_cmd {
+ uint8_t op; /* ELS command code*/
+ uint8_t byte1;
+ uint8_t byte2;
+ uint8_t byte3;
+
+ union {
+ struct ls_rjt {
+ uint8_t rsvd1;
+ uint8_t reason_code; /* Reason code */
+ uint8_t reason_exp; /* Explanation */
+ uint8_t vendor_unique; /* Vendor unique code */
+ } ls_rjt;
+
+ struct ls_logi {
+ /* Service Parameters */
+ struct csio_service_parms sp;
+ } ls_logi;
+
+ struct logo {
+ uint32_t nport_id; /* NPort Id */
+ uint8_t wwpn[8]; /* Port name */
+ } logo;
+
+ struct prli {
+ uint8_t type; /* Type code */
+ uint8_t rsvd1;
+ uint8_t proc_flags; /* Process Flags */
+ uint8_t rsvd2;
+
+ /* Originator Process Associator */
+ uint32_t ori_proc_assoc;
+
+ /* Responder Process Associator */
+ uint32_t rsp_proc_assoc;
+
+ /* Service parameter flags */
+ uint32_t serv_parms_flags;
+ } prli;
+
+ struct prlo {
+ uint8_t type; /* Type code */
+ uint8_t rsvd1;
+ uint8_t proc_flags; /* Process flags */
+ uint8_t rsvd2;
+
+ /* Originator Process Associator */
+ uint32_t ori_proc_assoc;
+
+ /* Responder Process Associator */
+ uint32_t rsp_proc_assoc;
+ uint32_t rsvd3;
+ } prlo;
+
+ struct adisc {
+ uint32_t hard_addr; /* Hard address of originator */
+ uint8_t wwpn[8]; /* Port name */
+ uint8_t wwnn[8]; /* Node name */
+ uint32_t nport_id; /* Nport id */
+ } adisc;
+
+ struct scr {
+ uint8_t rsvd[3];
+ uint8_t func; /* SCR Function */
+ } scr;
+
+ struct rscn {
+ uint32_t nport_id; /* Nport id list */
+ } rscn;
+ } un;
+};
+
+/* FCP defines */
+/*
+ * pri_ta.
+ */
+#define FCP_PTA_SIMPLE 0x0 /* simple queue tag */
+#define FCP_PTA_HEADQ 0x1 /* head of queue tag */
+#define FCP_PTA_ORDERED 0x2 /* ordered task attribute */
+#define FCP_PTA_ACA 0x4 /* auto. contigent allegiance */
+#define FCP_PTA_UNTAGGED 0x5
+
+#define FCP_PRI_SHIFT 3 /* priority field starts
+ * in bit 3
+ */
+#define FCP_PRI_RESVD_MASK 0x80 /* reserved bits in priority
+ * field
+ */
+
+/*
+ * tm_flags - task management flags field.
+ */
+#define FCP_TMF_ABT_TASK_SET 0x02 /* abort task set */
+#define FCP_TMF_CLR_TASK_SET 0x04 /* clear task set */
+#define FCP_TMF_BUS_RESET 0x08 /* bus reset */
+#define FCP_TMF_LUN_RESET 0x10 /* LUN reset */
+#define FCP_TMF_TGT_RESET 0x20 /* Target reset */
+#define FCP_TMF_CLR_ACA 0x40 /* clear ACA condition */
+#define FCP_TMF_TERM_TASK 0x80 /* Terminate task */
+
+/*
+ * flags.
+ * Bits 7:2 are the additional FCP_CDB length / 4.
+ */
+#define FCP_CFL_LEN_MASK 0xfc /* mask for additional length */
+#define FCP_CFL_LEN_SHIFT 2 /* shift bits for additional length */
+#define FCP_CFL_RDDATA 0x02 /* read data */
+#define FCP_CFL_WRDATA 0x01 /* write data */
+
+struct csio_fcp_cmnd {
+ uint8_t lun[8]; /* logical unit number */
+ uint8_t cmdref; /* commmand reference number */
+ uint8_t pri_ta; /* priority and task
+ * attribute
+ */
+ uint8_t tm_flags; /* task management flags */
+ uint8_t flags; /* additional len & flags */
+ uint8_t cdb[16]; /* CDB */
+ uint32_t dl; /* data length */
+};
+
+/* Response Flags */
+#define FCP_BIDI_RSP 0x80 /* bidirectional read rsp */
+#define FCP_BIDI_READ_UNDER 0x40 /* bidi read underrun */
+#define FCP_BIDI_READ_OVER 0x20 /* bidi read overrun */
+#define FCP_CONF_REQ 0x10 /* confirmation requested */
+#define FCP_RESID_UNDER 0x08 /* transfer shorter than
+ * expected
+ */
+#define FCP_RESID_OVER 0x04 /* DL insufficient for
+ * full transfer
+ */
+#define FCP_SNS_LEN_VAL 0x02 /* SNS_LEN field is valid */
+#define FCP_RSP_LEN_VAL 0x01 /* RSP_LEN field is valid */
+
+/* Response codes */
+#define FCP_TMF_CMPL 0x00
+#define FCP_DATA_LEN_INVALID 0x01
+#define FCP_CMND_FIELDS_INVALID 0x02
+#define FCP_DATA_PARAM_MISMATCH 0x03
+#define FCP_TMF_REJECTED 0x04
+#define FCP_TMF_FAILED 0x05
+#define FCP_TMF_SUCCEEDED 0x05
+#define FCP_TMF_INVALID_LUN 0x09
+
+struct csio_fcp_resp {
+ uint8_t rsvd0[8];
+ uint16_t retry_delay; /* retry delay timer */
+ uint8_t flags; /* flags */
+ uint8_t scsi_status; /* SCSI status code */
+ uint32_t resid; /* Residual bytes */
+ uint32_t sns_len; /* Length of sense data */
+ uint32_t rsp_len; /* Length of response */
+ uint8_t rsvd1;
+ uint8_t rsvd2;
+ uint8_t rsvd3;
+ uint8_t rsp_code; /* Response code */
+ uint8_t sns_data[128];
+};
+
+#endif /* __CSIO_FCOE_PROTO_H__ */
diff --git a/drivers/scsi/csiostor/csio_hw.h b/drivers/scsi/csiostor/csio_hw.h
new file mode 100644
index 0000000..00b78db
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_hw.h
@@ -0,0 +1,668 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __CSIO_HW_H__
+#define __CSIO_HW_H__
+
+#include <linux/kernel.h>
+#include <linux/pci.h>
+#include <linux/device.h>
+#include <linux/workqueue.h>
+#include <linux/compiler.h>
+#include <linux/cdev.h>
+#include <linux/list.h>
+#include <linux/mempool.h>
+#include <linux/io.h>
+#include <linux/spinlock_types.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_wr.h"
+#include "csio_mb.h"
+#include "csio_scsi.h"
+#include "csio_defs.h"
+#include "t4_regs.h"
+#include "t4_msg.h"
+
+/*
+ * An error value used by host. Should not clash with FW defined return values.
+ */
+#define FW_HOSTERROR 255
+
+#define CSIO_FW_FNAME "cxgb4/t4fw.bin"
+#define CSIO_CF_FNAME "cxgb4/t4-config.txt"
+
+#define FW_VERSION_MAJOR 1
+#define FW_VERSION_MINOR 2
+#define FW_VERSION_MICRO 8
+
+#define CSIO_HW_NAME "Chelsio FCoE Adapter"
+#define CSIO_MAX_PFN 8
+#define CSIO_MAX_PPORTS 4
+
+#define CSIO_MAX_LUN 0xFFFF
+#define CSIO_MAX_QUEUE 2048
+#define CSIO_MAX_CMD_PER_LUN 32
+#define CSIO_MAX_DDP_BUF_SIZE (1024 * 1024)
+#define CSIO_MAX_SECTOR_SIZE 128
+
+/* Interrupts */
+#define CSIO_EXTRA_MSI_IQS 2 /* Extra iqs for INTX/MSI mode
+ * (Forward intr iq + fw iq) */
+#define CSIO_EXTRA_VECS 2 /* non-data + FW evt */
+#define CSIO_MAX_SCSI_CPU 128
+#define CSIO_MAX_SCSI_QSETS (CSIO_MAX_SCSI_CPU * CSIO_MAX_PPORTS)
+#define CSIO_MAX_MSIX_VECS (CSIO_MAX_SCSI_QSETS + CSIO_EXTRA_VECS)
+
+/* Queues */
+enum {
+ CSIO_INTR_WRSIZE = 128,
+ CSIO_INTR_IQSIZE = ((CSIO_MAX_MSIX_VECS + 1) * CSIO_INTR_WRSIZE),
+ CSIO_FWEVT_WRSIZE = 128,
+ CSIO_FWEVT_IQLEN = 128,
+ CSIO_FWEVT_FLBUFS = 64,
+ CSIO_FWEVT_IQSIZE = (CSIO_FWEVT_WRSIZE * CSIO_FWEVT_IQLEN),
+ CSIO_HW_NIQ = 1,
+ CSIO_HW_NFLQ = 1,
+ CSIO_HW_NEQ = 1,
+ CSIO_HW_NINTXQ = 1,
+};
+
+struct csio_msix_entries {
+ unsigned short vector; /* Vector assigned by pci_enable_msix */
+ void *dev_id; /* Priv object associated w/ this msix*/
+ char desc[24]; /* Description of this vector */
+};
+
+struct csio_scsi_qset {
+ int iq_idx; /* Ingress index */
+ int eq_idx; /* Egress index */
+ uint32_t intr_idx; /* MSIX Vector index */
+};
+
+struct csio_scsi_cpu_info {
+ int16_t max_cpus;
+};
+
+extern int csio_dbg_level;
+extern int csio_force_master;
+extern unsigned int csio_port_mask;
+extern int csio_msi;
+
+#define CSIO_VENDOR_ID 0x1425
+#define CSIO_ASIC_DEVID_PROTO_MASK 0xFF00
+#define CSIO_ASIC_DEVID_TYPE_MASK 0x00FF
+#define CSIO_FPGA 0xA000
+#define CSIO_T4_FCOE_ASIC 0x4600
+
+#define CSIO_GLBL_INTR_MASK (CIM | MPS | PL | PCIE | MC | EDC0 | \
+ EDC1 | LE | TP | MA | PM_TX | PM_RX | \
+ ULP_RX | CPL_SWITCH | SGE | \
+ ULP_TX | SF)
+
+/*
+ * Hard parameters used to initialize the card in the absence of a
+ * configuration file.
+ */
+enum {
+ /* General */
+ CSIO_SGE_DBFIFO_INT_THRESH = 10,
+
+ CSIO_SGE_RX_DMA_OFFSET = 2,
+
+ CSIO_SGE_FLBUF_SIZE1 = 65536,
+ CSIO_SGE_FLBUF_SIZE2 = 1536,
+ CSIO_SGE_FLBUF_SIZE3 = 9024,
+ CSIO_SGE_FLBUF_SIZE4 = 9216,
+ CSIO_SGE_FLBUF_SIZE5 = 2048,
+ CSIO_SGE_FLBUF_SIZE6 = 128,
+ CSIO_SGE_FLBUF_SIZE7 = 8192,
+ CSIO_SGE_FLBUF_SIZE8 = 16384,
+
+ CSIO_SGE_TIMER_VAL_0 = 5,
+ CSIO_SGE_TIMER_VAL_1 = 10,
+ CSIO_SGE_TIMER_VAL_2 = 20,
+ CSIO_SGE_TIMER_VAL_3 = 50,
+ CSIO_SGE_TIMER_VAL_4 = 100,
+ CSIO_SGE_TIMER_VAL_5 = 200,
+
+ CSIO_SGE_INT_CNT_VAL_0 = 1,
+ CSIO_SGE_INT_CNT_VAL_1 = 4,
+ CSIO_SGE_INT_CNT_VAL_2 = 8,
+ CSIO_SGE_INT_CNT_VAL_3 = 16,
+
+ /* Storage specific - used by FW_PFVF_CMD */
+ CSIO_WX_CAPS = FW_CMD_CAP_PF, /* w/x all */
+ CSIO_R_CAPS = FW_CMD_CAP_PF, /* r all */
+ CSIO_NVI = 4,
+ CSIO_NIQ_FLINT = 34,
+ CSIO_NETH_CTRL = 32,
+ CSIO_NEQ = 66,
+ CSIO_NEXACTF = 32,
+ CSIO_CMASK = FW_PFVF_CMD_CMASK_MASK,
+ CSIO_PMASK = FW_PFVF_CMD_PMASK_MASK,
+};
+
+/* Slowpath events */
+enum csio_evt {
+ CSIO_EVT_FW = 0, /* FW event */
+ CSIO_EVT_MBX, /* MBX event */
+ CSIO_EVT_SCN, /* State change notification */
+ CSIO_EVT_DEV_LOSS, /* Device loss event */
+ CSIO_EVT_MAX, /* Max supported event */
+};
+
+#define CSIO_EVT_MSG_SIZE 512
+#define CSIO_EVTQ_SIZE 512
+
+/* Event msg */
+struct csio_evt_msg {
+ struct list_head list; /* evt queue*/
+ enum csio_evt type;
+ uint8_t data[CSIO_EVT_MSG_SIZE];
+};
+
+enum {
+ EEPROMVSIZE = 32768, /* Serial EEPROM virtual address space size */
+ SERNUM_LEN = 16, /* Serial # length */
+ EC_LEN = 16, /* E/C length */
+ ID_LEN = 16, /* ID length */
+ TRACE_LEN = 112, /* length of trace data and mask */
+};
+
+enum {
+ SF_PAGE_SIZE = 256, /* serial flash page size */
+ SF_SEC_SIZE = 64 * 1024, /* serial flash sector size */
+ SF_SIZE = SF_SEC_SIZE * 16, /* serial flash size */
+};
+
+enum { MEM_EDC0, MEM_EDC1, MEM_MC };
+
+enum {
+ MEMWIN0_APERTURE = 2048,
+ MEMWIN0_BASE = 0x1b800,
+ MEMWIN1_APERTURE = 32768,
+ MEMWIN1_BASE = 0x28000,
+ MEMWIN2_APERTURE = 65536,
+ MEMWIN2_BASE = 0x30000,
+};
+
+/* serial flash and firmware constants */
+enum {
+ SF_ATTEMPTS = 10, /* max retries for SF operations */
+
+ /* flash command opcodes */
+ SF_PROG_PAGE = 2, /* program page */
+ SF_WR_DISABLE = 4, /* disable writes */
+ SF_RD_STATUS = 5, /* read status register */
+ SF_WR_ENABLE = 6, /* enable writes */
+ SF_RD_DATA_FAST = 0xb, /* read flash */
+ SF_RD_ID = 0x9f, /* read ID */
+ SF_ERASE_SECTOR = 0xd8, /* erase sector */
+
+ FW_START_SEC = 8, /* first flash sector for FW */
+ FW_END_SEC = 15, /* last flash sector for FW */
+ FW_IMG_START = FW_START_SEC * SF_SEC_SIZE,
+ FW_MAX_SIZE = (FW_END_SEC - FW_START_SEC + 1) * SF_SEC_SIZE,
+
+ FLASH_CFG_MAX_SIZE = 0x10000 , /* max size of the flash config file*/
+ FLASH_CFG_OFFSET = 0x1f0000,
+ FLASH_CFG_START_SEC = FLASH_CFG_OFFSET / SF_SEC_SIZE,
+ FPGA_FLASH_CFG_OFFSET = 0xf0000 , /* if FPGA mode, then cfg file is
+ * at 1MB - 64KB */
+ FPGA_FLASH_CFG_START_SEC = FPGA_FLASH_CFG_OFFSET / SF_SEC_SIZE,
+};
+
+/*
+ * Flash layout.
+ */
+#define FLASH_START(start) ((start) * SF_SEC_SIZE)
+#define FLASH_MAX_SIZE(nsecs) ((nsecs) * SF_SEC_SIZE)
+
+enum {
+ /*
+ * Location of firmware image in FLASH.
+ */
+ FLASH_FW_START_SEC = 8,
+ FLASH_FW_NSECS = 8,
+ FLASH_FW_START = FLASH_START(FLASH_FW_START_SEC),
+ FLASH_FW_MAX_SIZE = FLASH_MAX_SIZE(FLASH_FW_NSECS),
+
+};
+
+#undef FLASH_START
+#undef FLASH_MAX_SIZE
+
+/* Management module */
+enum {
+ CSIO_MGMT_EQ_WRSIZE = 512,
+ CSIO_MGMT_IQ_WRSIZE = 128,
+ CSIO_MGMT_EQLEN = 64,
+ CSIO_MGMT_IQLEN = 64,
+};
+
+#define CSIO_MGMT_EQSIZE (CSIO_MGMT_EQLEN * CSIO_MGMT_EQ_WRSIZE)
+#define CSIO_MGMT_IQSIZE (CSIO_MGMT_IQLEN * CSIO_MGMT_IQ_WRSIZE)
+
+/* mgmt module stats */
+struct csio_mgmtm_stats {
+ uint32_t n_abort_req; /* Total abort request */
+ uint32_t n_abort_rsp; /* Total abort response */
+ uint32_t n_close_req; /* Total close request */
+ uint32_t n_close_rsp; /* Total close response */
+ uint32_t n_err; /* Total Errors */
+ uint32_t n_drop; /* Total request dropped */
+ uint32_t n_active; /* Count of active_q */
+ uint32_t n_cbfn; /* Count of cbfn_q */
+};
+
+/* MGMT module */
+struct csio_mgmtm {
+ struct csio_hw *hw; /* Pointer to HW moduel */
+ int eq_idx; /* Egress queue index */
+ int iq_idx; /* Ingress queue index */
+ int msi_vec; /* MSI vector */
+ struct list_head active_q; /* Outstanding ELS/CT */
+ struct list_head abort_q; /* Outstanding abort req */
+ struct list_head cbfn_q; /* Completion queue */
+ struct list_head mgmt_req_freelist; /* Free poll of reqs */
+ /* ELSCT request freelist*/
+ struct timer_list mgmt_timer; /* MGMT timer */
+ struct csio_mgmtm_stats stats; /* ELS/CT stats */
+};
+
+struct csio_adap_desc {
+ char model_no[16];
+ char description[32];
+};
+
+struct pci_params {
+ uint16_t vendor_id;
+ uint16_t device_id;
+ uint32_t vpd_cap_addr;
+ uint16_t speed;
+ uint8_t width;
+};
+
+/* User configurable hw parameters */
+struct csio_hw_params {
+ uint32_t sf_size; /* serial flash
+ * size in bytes
+ */
+ uint32_t sf_nsec; /* # of flash sectors */
+ struct pci_params pci;
+ uint32_t log_level; /* Module-level for
+ * debug log.
+ */
+};
+
+struct csio_vpd {
+ uint32_t cclk;
+ uint8_t ec[EC_LEN + 1];
+ uint8_t sn[SERNUM_LEN + 1];
+ uint8_t id[ID_LEN + 1];
+};
+
+struct csio_pport {
+ uint16_t pcap;
+ uint8_t portid;
+ uint8_t link_status;
+ uint16_t link_speed;
+ uint8_t mac[6];
+ uint8_t mod_type;
+ uint8_t rsvd1;
+ uint8_t rsvd2;
+ uint8_t rsvd3;
+};
+
+/* fcoe resource information */
+struct csio_fcoe_res_info {
+ uint16_t e_d_tov;
+ uint16_t r_a_tov_seq;
+ uint16_t r_a_tov_els;
+ uint16_t r_r_tov;
+ uint32_t max_xchgs;
+ uint32_t max_ssns;
+ uint32_t used_xchgs;
+ uint32_t used_ssns;
+ uint32_t max_fcfs;
+ uint32_t max_vnps;
+ uint32_t used_fcfs;
+ uint32_t used_vnps;
+};
+
+/* HW State machine Events */
+enum csio_hw_ev {
+ CSIO_HWE_CFG = (uint32_t)1, /* Starts off the State machine */
+ CSIO_HWE_INIT, /* Config done, start Init */
+ CSIO_HWE_INIT_DONE, /* Init Mailboxes sent, HW ready */
+ CSIO_HWE_FATAL, /* Fatal error during initialization */
+ CSIO_HWE_PCIERR_DETECTED,/* PCI error recovery detetced */
+ CSIO_HWE_PCIERR_SLOT_RESET, /* Slot reset after PCI recoviery */
+ CSIO_HWE_PCIERR_RESUME, /* Resume after PCI error recovery */
+ CSIO_HWE_QUIESCED, /* HBA quiesced */
+ CSIO_HWE_HBA_RESET, /* HBA reset requested */
+ CSIO_HWE_HBA_RESET_DONE, /* HBA reset completed */
+ CSIO_HWE_FW_DLOAD, /* FW download requested */
+ CSIO_HWE_PCI_REMOVE, /* PCI de-instantiation */
+ CSIO_HWE_SUSPEND, /* HW suspend for Online(hot) replacement */
+ CSIO_HWE_RESUME, /* HW resume for Online(hot) replacement */
+ CSIO_HWE_MAX, /* Max HW event */
+};
+
+/* hw stats */
+struct csio_hw_stats {
+ uint32_t n_evt_activeq; /* Number of event in active Q */
+ uint32_t n_evt_freeq; /* Number of event in free Q */
+ uint32_t n_evt_drop; /* Number of event droped */
+ uint32_t n_evt_unexp; /* Number of unexpected events */
+ uint32_t n_pcich_offline;/* Number of pci channel offline */
+ uint32_t n_lnlkup_miss; /* Number of lnode lookup miss */
+ uint32_t n_cpl_fw6_msg; /* Number of cpl fw6 message*/
+ uint32_t n_cpl_fw6_pld; /* Number of cpl fw6 payload*/
+ uint32_t n_cpl_unexp; /* Number of unexpected cpl */
+ uint32_t n_mbint_unexp; /* Number of unexpected mbox */
+ /* interrupt */
+ uint32_t n_plint_unexp; /* Number of unexpected PL */
+ /* interrupt */
+ uint32_t n_plint_cnt; /* Number of PL interrupt */
+ uint32_t n_int_stray; /* Number of stray interrupt */
+ uint32_t n_err; /* Number of hw errors */
+ uint32_t n_err_fatal; /* Number of fatal errors */
+ uint32_t n_err_nomem; /* Number of memory alloc failure */
+ uint32_t n_err_io; /* Number of IO failure */
+ enum csio_hw_ev n_evt_sm[CSIO_HWE_MAX]; /* Number of sm events */
+ uint64_t n_reset_start; /* Start time after the reset */
+ uint32_t rsvd1;
+};
+
+/* Defines for hw->flags */
+#define CSIO_HWF_MASTER 0x00000001 /* This is the Master
+ * function for the
+ * card.
+ */
+#define CSIO_HWF_HW_INTR_ENABLED 0x00000002 /* Are HW Interrupt
+ * enable bit set?
+ */
+#define CSIO_HWF_FWEVT_PENDING 0x00000004 /* FW events pending */
+#define CSIO_HWF_Q_MEM_ALLOCED 0x00000008 /* Queues have been
+ * allocated memory.
+ */
+#define CSIO_HWF_Q_FW_ALLOCED 0x00000010 /* Queues have been
+ * allocated in FW.
+ */
+#define CSIO_HWF_VPD_VALID 0x00000020 /* Valid VPD copied */
+#define CSIO_HWF_DEVID_CACHED 0X00000040 /* PCI vendor & device
+ * id cached */
+#define CSIO_HWF_FWEVT_STOP 0x00000080 /* Stop processing
+ * FW events
+ */
+#define CSIO_HWF_USING_SOFT_PARAMS 0x00000100 /* Using FW config
+ * params
+ */
+#define CSIO_HWF_HOST_INTR_ENABLED 0x00000200 /* Are host interrupts
+ * enabled?
+ */
+
+#define csio_is_hw_intr_enabled(__hw) \
+ ((__hw)->flags & CSIO_HWF_HW_INTR_ENABLED)
+#define csio_is_host_intr_enabled(__hw) \
+ ((__hw)->flags & CSIO_HWF_HOST_INTR_ENABLED)
+#define csio_is_hw_master(__hw) ((__hw)->flags & CSIO_HWF_MASTER)
+#define csio_is_valid_vpd(__hw) ((__hw)->flags & CSIO_HWF_VPD_VALID)
+#define csio_is_dev_id_cached(__hw) ((__hw)->flags & CSIO_HWF_DEVID_CACHED)
+#define csio_valid_vpd_copied(__hw) ((__hw)->flags |= CSIO_HWF_VPD_VALID)
+#define csio_dev_id_cached(__hw) ((__hw)->flags |= CSIO_HWF_DEVID_CACHED)
+
+/* Defines for intr_mode */
+enum csio_intr_mode {
+ CSIO_IM_NONE = 0,
+ CSIO_IM_INTX = 1,
+ CSIO_IM_MSI = 2,
+ CSIO_IM_MSIX = 3,
+};
+
+/* Master HW structure: One per function */
+struct csio_hw {
+ struct csio_sm sm; /* State machine: should
+ * be the 1st member.
+ */
+ spinlock_t lock; /* Lock for hw */
+
+ struct csio_scsim scsim; /* SCSI module*/
+ struct csio_wrm wrm; /* Work request module*/
+ struct pci_dev *pdev; /* PCI device */
+
+ void __iomem *regstart; /* Virtual address of
+ * register map
+ */
+ /* SCSI queue sets */
+ uint32_t num_sqsets; /* Number of SCSI
+ * queue sets */
+ uint32_t num_scsi_msix_cpus; /* Number of CPUs that
+ * will be used
+ * for ingress
+ * processing.
+ */
+
+ struct csio_scsi_qset sqset[CSIO_MAX_PPORTS][CSIO_MAX_SCSI_CPU];
+ struct csio_scsi_cpu_info scsi_cpu_info[CSIO_MAX_PPORTS];
+
+ uint32_t evtflag; /* Event flag */
+ uint32_t flags; /* HW flags */
+
+ struct csio_mgmtm mgmtm; /* management module */
+ struct csio_mbm mbm; /* Mailbox module */
+
+ /* Lnodes */
+ uint32_t num_lns; /* Number of lnodes */
+ struct csio_lnode *rln; /* Root lnode */
+ struct list_head sln_head; /* Sibling node list
+ * list
+ */
+ int intr_iq_idx; /* Forward interrupt
+ * queue.
+ */
+ int fwevt_iq_idx; /* FW evt queue */
+ struct work_struct evtq_work; /* Worker thread for
+ * HW events.
+ */
+ struct list_head evt_free_q; /* freelist of evt
+ * elements
+ */
+ struct list_head evt_active_q; /* active evt queue*/
+
+ /* board related info */
+ char name[32];
+ char hw_ver[16];
+ char model_desc[32];
+ char drv_version[32];
+ char fwrev_str[32];
+ uint32_t optrom_ver;
+ uint32_t fwrev;
+ uint32_t tp_vers;
+ char chip_ver;
+ uint32_t cfg_finiver;
+ uint32_t cfg_finicsum;
+ uint32_t cfg_cfcsum;
+ uint8_t cfg_csum_status;
+ uint8_t cfg_store;
+ enum csio_dev_state fw_state;
+ struct csio_vpd vpd;
+
+ uint8_t pfn; /* Physical Function
+ * number
+ */
+ uint32_t port_vec; /* Port vector */
+ uint8_t num_pports; /* Number of physical
+ * ports.
+ */
+ uint8_t rst_retries; /* Reset retries */
+ uint8_t cur_evt; /* current s/m evt */
+ uint8_t prev_evt; /* Previous s/m evt */
+ uint32_t dev_num; /* device number */
+ struct csio_pport pport[CSIO_MAX_PPORTS]; /* Ports (XGMACs) */
+ struct csio_hw_params params; /* Hw parameters */
+
+ struct pci_pool *scsi_pci_pool; /* PCI pool for SCSI */
+ mempool_t *mb_mempool; /* Mailbox memory pool*/
+ mempool_t *rnode_mempool; /* rnode memory pool */
+
+ /* Interrupt */
+ enum csio_intr_mode intr_mode; /* INTx, MSI, MSIX */
+ uint32_t fwevt_intr_idx; /* FW evt MSIX/interrupt
+ * index
+ */
+ uint32_t nondata_intr_idx; /* nondata MSIX/intr
+ * idx
+ */
+
+ uint8_t cfg_neq; /* FW configured no of
+ * egress queues
+ */
+ uint8_t cfg_niq; /* FW configured no of
+ * iq queues.
+ */
+
+ struct csio_fcoe_res_info fres_info; /* Fcoe resource info */
+
+ /* MSIX vectors */
+ struct csio_msix_entries msix_entries[CSIO_MAX_MSIX_VECS];
+
+ struct dentry *debugfs_root; /* Debug FS */
+ struct csio_hw_stats stats; /* Hw statistics */
+};
+
+/* Register access macros */
+#define csio_reg(_b, _r) ((_b) + (_r))
+
+#define csio_rd_reg8(_h, _r) readb(csio_reg((_h)->regstart, (_r)))
+#define csio_rd_reg16(_h, _r) readw(csio_reg((_h)->regstart, (_r)))
+#define csio_rd_reg32(_h, _r) readl(csio_reg((_h)->regstart, (_r)))
+#define csio_rd_reg64(_h, _r) readq(csio_reg((_h)->regstart, (_r)))
+
+#define csio_wr_reg8(_h, _v, _r) writeb((_v), \
+ csio_reg((_h)->regstart, (_r)))
+#define csio_wr_reg16(_h, _v, _r) writew((_v), \
+ csio_reg((_h)->regstart, (_r)))
+#define csio_wr_reg32(_h, _v, _r) writel((_v), \
+ csio_reg((_h)->regstart, (_r)))
+#define csio_wr_reg64(_h, _v, _r) writeq((_v), \
+ csio_reg((_h)->regstart, (_r)))
+
+void csio_set_reg_field(struct csio_hw *, uint32_t, uint32_t, uint32_t);
+
+/* Core clocks <==> uSecs */
+static inline uint32_t
+csio_core_ticks_to_us(struct csio_hw *hw, uint32_t ticks)
+{
+ /* add Core Clock / 2 to round ticks to nearest uS */
+ return (ticks * 1000 + hw->vpd.cclk/2) / hw->vpd.cclk;
+}
+
+static inline uint32_t
+csio_us_to_core_ticks(struct csio_hw *hw, uint32_t us)
+{
+ return (us * hw->vpd.cclk) / 1000;
+}
+
+/* Easy access macros */
+#define csio_hw_to_wrm(hw) ((struct csio_wrm *)(&(hw)->wrm))
+#define csio_hw_to_mbm(hw) ((struct csio_mbm *)(&(hw)->mbm))
+#define csio_hw_to_scsim(hw) ((struct csio_scsim *)(&(hw)->scsim))
+#define csio_hw_to_mgmtm(hw) ((struct csio_mgmtm *)(&(hw)->mgmtm))
+
+#define csio_md(hw, idx) (&(hw)->mem_descs[(idx)])
+
+#define CSIO_PCI_BUS(hw) ((hw)->pdev->bus->number)
+#define CSIO_PCI_DEV(hw) (PCI_SLOT((hw)->pdev->devfn))
+#define CSIO_PCI_FUNC(hw) (PCI_FUNC((hw)->pdev->devfn))
+
+#define csio_set_fwevt_intr_idx(_h, _i) ((_h)->fwevt_intr_idx = (_i))
+#define csio_get_fwevt_intr_idx(_h) ((_h)->fwevt_intr_idx)
+#define csio_set_nondata_intr_idx(_h, _i) ((_h)->nondata_intr_idx = (_i))
+#define csio_get_nondata_intr_idx(_h) ((_h)->nondata_intr_idx)
+
+/* Printing/logging */
+#define CSIO_DEVID(__dev) ((__dev)->dev_num)
+#define CSIO_DEVID_LO(__dev) (CSIO_DEVID((__dev)) & 0xFFFF)
+#define CSIO_DEVID_HI(__dev) ((CSIO_DEVID((__dev)) >> 16) & 0xFFFF)
+
+#define csio_info(__hw, __fmt, ...) \
+ dev_info(&(__hw)->pdev->dev, __fmt, ##__VA_ARGS__)
+
+#define csio_fatal(__hw, __fmt, ...) \
+ dev_crit(&(__hw)->pdev->dev, __fmt, ##__VA_ARGS__)
+
+#define csio_err(__hw, __fmt, ...) \
+ dev_err(&(__hw)->pdev->dev, __fmt, ##__VA_ARGS__)
+
+#define csio_warn(__hw, __fmt, ...) \
+ dev_warn(&(__hw)->pdev->dev, __fmt, ##__VA_ARGS__)
+
+#ifdef __CSIO_DEBUG__
+#define csio_dbg(__hw, __fmt, ...) \
+ csio_info((__hw), __fmt, ##__VA_ARGS__);
+#else
+#define csio_dbg(__hw, __fmt, ...)
+#endif
+
+csio_retval_t csio_mgmt_req_lookup(struct csio_mgmtm *, struct csio_ioreq *);
+void csio_hw_intr_disable(struct csio_hw *);
+int csio_hw_slow_intr_handler(struct csio_hw *hw);
+csio_retval_t csio_hw_start(struct csio_hw *);
+csio_retval_t csio_hw_stop(struct csio_hw *);
+csio_retval_t csio_hw_reset(struct csio_hw *);
+int csio_is_hw_ready(struct csio_hw *);
+int csio_is_hw_removing(struct csio_hw *);
+
+csio_retval_t csio_fwevtq_handler(struct csio_hw *);
+void csio_evtq_worker(struct work_struct *);
+csio_retval_t csio_enqueue_evt(struct csio_hw *hw, enum csio_evt type,
+ void *evt_msg, uint16_t len);
+void csio_evtq_flush(struct csio_hw *hw);
+
+csio_retval_t csio_request_irqs(struct csio_hw *);
+void csio_intr_enable(struct csio_hw *);
+void csio_intr_disable(struct csio_hw *, bool);
+
+struct csio_lnode *csio_lnode_alloc(struct csio_hw *);
+csio_retval_t csio_config_queues(struct csio_hw *);
+
+csio_retval_t csio_hw_mc_read(struct csio_hw *, uint32_t,
+ uint32_t *, uint64_t *);
+csio_retval_t csio_hw_edc_read(struct csio_hw *, int, uint32_t, uint32_t *,
+ uint64_t *);
+csio_retval_t csio_hw_init(struct csio_hw *);
+void csio_hw_exit(struct csio_hw *);
+#endif /* ifndef __CSIO_HW_H__ */
diff --git a/drivers/scsi/csiostor/csio_init.h b/drivers/scsi/csiostor/csio_init.h
new file mode 100644
index 0000000..0838fd7
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_init.h
@@ -0,0 +1,158 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __CSIO_INIT_H__
+#define __CSIO_INIT_H__
+
+#include <linux/pci.h>
+#include <linux/if_ether.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi_transport_fc.h>
+
+#include "csio_scsi.h"
+#include "csio_lnode.h"
+#include "csio_rnode.h"
+#include "csio_hw.h"
+
+#define CSIO_DRV_AUTHOR "Chelsio Communications"
+#define CSIO_DRV_LICENSE "Dual BSD/GPL"
+#define CSIO_DRV_DESC "Chelsio FCoE driver"
+#define CSIO_DRV_VERSION "1.0.0"
+
+#define CSIO_DEVICE(devid, idx) \
+{ PCI_VENDOR_ID_CHELSIO, (devid), PCI_ANY_ID, PCI_ANY_ID, 0, 0, (idx) }
+
+#define CSIO_IS_T4_FPGA(_dev) (((_dev) == CSIO_DEVID_PE10K) ||\
+ ((_dev) == CSIO_DEVID_PE10K_PF1))
+
+/* FCoE device IDs */
+#define CSIO_DEVID_PE10K 0xA000
+#define CSIO_DEVID_PE10K_PF1 0xA001
+#define CSIO_DEVID_T440DBG_FCOE 0x4600
+#define CSIO_DEVID_T420CR_FCOE 0x4601
+#define CSIO_DEVID_T422CR_FCOE 0x4602
+#define CSIO_DEVID_T440CR_FCOE 0x4603
+#define CSIO_DEVID_T420BCH_FCOE 0x4604
+#define CSIO_DEVID_T440BCH_FCOE 0x4605
+#define CSIO_DEVID_T440CH_FCOE 0x4606
+#define CSIO_DEVID_T420SO_FCOE 0x4607
+#define CSIO_DEVID_T420CX_FCOE 0x4608
+#define CSIO_DEVID_T420BT_FCOE 0x4609
+#define CSIO_DEVID_T404BT_FCOE 0x460A
+#define CSIO_DEVID_B420_FCOE 0x460B
+#define CSIO_DEVID_B404_FCOE 0x460C
+#define CSIO_DEVID_T480CR_FCOE 0x460D
+#define CSIO_DEVID_T440LPCR_FCOE 0x460E
+
+extern struct fc_function_template csio_fc_transport_funcs;
+extern struct fc_function_template csio_fc_transport_vport_funcs;
+
+void csio_fchost_attr_init(struct csio_lnode *);
+
+/* INTx handlers */
+void csio_scsi_intx_handler(struct csio_hw *, void *, uint32_t,
+ struct csio_fl_dma_buf *, void *);
+
+void csio_fwevt_intx_handler(struct csio_hw *, void *, uint32_t,
+ struct csio_fl_dma_buf *, void *);
+
+/* Common os lnode APIs */
+void csio_lnodes_block_request(struct csio_hw *);
+void csio_lnodes_unblock_request(struct csio_hw *);
+void csio_lnodes_block_by_port(struct csio_hw *, uint8_t);
+void csio_lnodes_unblock_by_port(struct csio_hw *, uint8_t);
+
+struct csio_lnode *csio_shost_init(struct csio_hw *, struct device *, bool,
+ struct csio_lnode *);
+void csio_shost_exit(struct csio_lnode *);
+void csio_lnodes_exit(struct csio_hw *, bool);
+
+static inline struct Scsi_Host *
+csio_ln_to_shost(struct csio_lnode *ln)
+{
+ return container_of((void *)ln, struct Scsi_Host, hostdata[0]);
+}
+
+/* SCSI -- locking version of get/put ioreqs */
+static inline struct csio_ioreq *
+csio_get_scsi_ioreq_lock(struct csio_hw *hw, struct csio_scsim *scsim)
+{
+ struct csio_ioreq *ioreq;
+ unsigned long flags;
+
+ spin_lock_irqsave(&scsim->freelist_lock, flags);
+ ioreq = csio_get_scsi_ioreq(scsim);
+ spin_unlock_irqrestore(&scsim->freelist_lock, flags);
+
+ return ioreq;
+}
+
+static inline void
+csio_put_scsi_ioreq_lock(struct csio_hw *hw, struct csio_scsim *scsim,
+ struct csio_ioreq *ioreq)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&scsim->freelist_lock, flags);
+ csio_put_scsi_ioreq(scsim, ioreq);
+ spin_unlock_irqrestore(&scsim->freelist_lock, flags);
+}
+
+/* Called in interrupt context */
+static inline void
+csio_put_scsi_ioreq_list_lock(struct csio_hw *hw, struct csio_scsim *scsim,
+ struct list_head *reqlist, int n)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&scsim->freelist_lock, flags);
+ csio_put_scsi_ioreq_list(scsim, reqlist, n);
+ spin_unlock_irqrestore(&scsim->freelist_lock, flags);
+}
+
+/* Called in interrupt context */
+static inline void
+csio_put_scsi_ddp_list_lock(struct csio_hw *hw, struct csio_scsim *scsim,
+ struct list_head *reqlist, int n)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&hw->lock, flags);
+ csio_put_scsi_ddp_list(scsim, reqlist, n);
+ spin_unlock_irqrestore(&hw->lock, flags);
+}
+
+#endif /* ifndef __CSIO_INIT_H__ */
--
1.7.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* Re: [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
2012-08-23 22:27 ` [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1) Naresh Kumar Inna
@ 2012-08-23 18:15 ` Love, Robert W
2012-08-24 18:45 ` Naresh Kumar Inna
2012-08-23 19:58 ` Nicholas A. Bellinger
1 sibling, 1 reply; 33+ messages in thread
From: Love, Robert W @ 2012-08-23 18:15 UTC (permalink / raw)
To: Naresh Kumar Inna
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
dm@chelsio.com, netdev@vger.kernel.org, chethan@chelsio.com
On 8/23/2012 3:27 PM, Naresh Kumar Inna wrote:
> This patch contains the first set of the header files for csiostor driver.
>
> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
> ---
> drivers/scsi/csiostor/csio_defs.h | 143 ++++++
> drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
I think most of these protocol definitions can be found in
include/scsi/fc/ and possibly incluse/scsi/scsi_transport_fc.h. I think
any protocol stuff missing should be added to the common header files.
Thanks, //Rob
^ permalink raw reply [flat|nested] 33+ messages in thread
* Re: [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
2012-08-23 18:15 ` Love, Robert W
@ 2012-08-24 18:45 ` Naresh Kumar Inna
0 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-24 18:45 UTC (permalink / raw)
To: Love, Robert W
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/23/2012 11:45 PM, Love, Robert W wrote:
> On 8/23/2012 3:27 PM, Naresh Kumar Inna wrote:
>> This patch contains the first set of the header files for csiostor driver.
>>
>> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
>> ---
>> drivers/scsi/csiostor/csio_defs.h | 143 ++++++
>> drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
>
> I think most of these protocol definitions can be found in
> include/scsi/fc/ and possibly incluse/scsi/scsi_transport_fc.h. I think
> any protocol stuff missing should be added to the common header files.
>
> Thanks, //Rob--
Hi Rob,
Thanks for reviewing this patch. I will look into how best I can re-use
the headers in include/scsi/fc and get back if I have questions.
Regards,
Naresh.
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
^ permalink raw reply [flat|nested] 33+ messages in thread
* Re: [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
2012-08-23 22:27 ` [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1) Naresh Kumar Inna
2012-08-23 18:15 ` Love, Robert W
@ 2012-08-23 19:58 ` Nicholas A. Bellinger
2012-08-24 18:36 ` Naresh Kumar Inna
1 sibling, 1 reply; 33+ messages in thread
From: Nicholas A. Bellinger @ 2012-08-23 19:58 UTC (permalink / raw)
To: Naresh Kumar Inna; +Cc: JBottomley, linux-scsi, dm, netdev, chethan
On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
> This patch contains the first set of the header files for csiostor driver.
>
> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
> ---
> drivers/scsi/csiostor/csio_defs.h | 143 ++++++
> drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
> drivers/scsi/csiostor/csio_hw.h | 668 ++++++++++++++++++++++++
> drivers/scsi/csiostor/csio_init.h | 158 ++++++
> 4 files changed, 1812 insertions(+), 0 deletions(-)
> create mode 100644 drivers/scsi/csiostor/csio_defs.h
> create mode 100644 drivers/scsi/csiostor/csio_fcoe_proto.h
> create mode 100644 drivers/scsi/csiostor/csio_hw.h
> create mode 100644 drivers/scsi/csiostor/csio_init.h
>
Hi Naresh,
Just commenting on csio_defs.h bits here... As Robert mentioned, you'll
need to convert the driver to use (or add to) upstream protocol
definitions and drop the csio_fcoe_proto.h bits..
> diff --git a/drivers/scsi/csiostor/csio_defs.h b/drivers/scsi/csiostor/csio_defs.h
> new file mode 100644
> index 0000000..4f1c713
> --- /dev/null
> +++ b/drivers/scsi/csiostor/csio_defs.h
> @@ -0,0 +1,143 @@
> +/*
> + * This file is part of the Chelsio FCoE driver for Linux.
> + *
> + * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
> + *
> + * This software is available to you under a choice of one of two
> + * licenses. You may choose to be licensed under the terms of the GNU
> + * General Public License (GPL) Version 2, available from the file
> + * COPYING in the main directory of this source tree, or the
> + * OpenIB.org BSD license below:
> + *
> + * Redistribution and use in source and binary forms, with or
> + * without modification, are permitted provided that the following
> + * conditions are met:
> + *
> + * - Redistributions of source code must retain the above
> + * copyright notice, this list of conditions and the following
> + * disclaimer.
> + *
> + * - Redistributions in binary form must reproduce the above
> + * copyright notice, this list of conditions and the following
> + * disclaimer in the documentation and/or other materials
> + * provided with the distribution.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#ifndef __CSIO_DEFS_H__
> +#define __CSIO_DEFS_H__
> +
> +#include <linux/kernel.h>
> +#include <linux/timer.h>
> +#include <linux/list.h>
> +#include <linux/bug.h>
> +#include <linux/pci.h>
> +#include <linux/jiffies.h>
> +
> +/* Function returns */
> +enum csio_retval {
> + CSIO_SUCCESS = 0,
> + CSIO_INVAL = 1,
> + CSIO_BUSY = 2,
> + CSIO_NOSUPP = 3,
> + CSIO_TIMEOUT = 4,
> + CSIO_NOMEM = 5,
> + CSIO_NOPERM = 6,
> + CSIO_RETRY = 7,
> + CSIO_EPROTO = 8,
> + CSIO_EIO = 9,
> + CSIO_CANCELLED = 10,
> +};
> +
Please don't assign macros for errno's, and give them positive values.
> +#define csio_retval_t enum csio_retval
Please get rid of this csio_retval_t nonsense.
> +
> +enum {
> + CSIO_FALSE = 0,
> + CSIO_TRUE = 1,
> +};
> +
Same here, please use normal Boolean macros
> +#define CSIO_ROUNDUP(__v, __r) (((__v) + (__r) - 1) / (__r))
> +#define CSIO_INVALID_IDX 0xFFFFFFFF
> +#define csio_inc_stats(elem, val) ((elem)->stats.val++)
> +#define csio_dec_stats(elem, val) ((elem)->stats.val--)
No reason for either of this stats inc+dec macros. Please drop them.
> +#define csio_valid_wwn(__n) ((*__n >> 4) == 0x5 ? CSIO_TRUE : \
> + CSIO_FALSE)
> +#define CSIO_WORD_TO_BYTE 4
> +
> +static inline int
> +csio_list_deleted(struct list_head *list)
> +{
> + return ((list->next == list) && (list->prev == list));
> +}
> +
> +#define csio_list_next(elem) (((struct list_head *)(elem))->next)
> +#define csio_list_prev(elem) (((struct list_head *)(elem))->prev)
> +
> +#define csio_deq_from_head(head, elem) \
> +do { \
> + if (!list_empty(head)) { \
> + *((struct list_head **)(elem)) = csio_list_next((head)); \
> + csio_list_next((head)) = \
> + csio_list_next(csio_list_next((head))); \
> + csio_list_prev(csio_list_next((head))) = (head); \
> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
> + } else \
> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
> +} while (0)
> +
This code is confusing as hell.. Why can't you just use normal list.h
macros for this..?
> +#define csio_deq_from_tail(head, elem) \
> +do { \
> + if (!list_empty(head)) { \
> + *((struct list_head **)(elem)) = csio_list_prev((head)); \
> + csio_list_prev((head)) = \
> + csio_list_prev(csio_list_prev((head))); \
> + csio_list_next(csio_list_prev((head))) = (head); \
> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
> + } else \
> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
> +} while (0)
> +
Same here.. Please don't use macros like this.
> +/* State machine */
> +typedef void (*csio_sm_state_t)(void *, uint32_t);
> +
> +struct csio_sm {
> + struct list_head sm_list;
> + csio_sm_state_t sm_state;
> +};
> +
> +#define csio_init_state(__smp, __state) \
> + (((struct csio_sm *)(__smp))->sm_state = (csio_sm_state_t)(__state))
> +
> +#define csio_set_state(__smp, __state) \
> + (((struct csio_sm *)(__smp))->sm_state = (csio_sm_state_t)(__state))
> +
> +
> +#define csio_post_event(__smp, __evt) \
> + (((struct csio_sm *)(__smp))->sm_state((__smp), (uint32_t)(__evt)))
> +
> +#define csio_get_state(__smp) (((struct csio_sm *)(__smp))->sm_state)
> +
> +#define csio_match_state(__smp, __state) \
> + (csio_get_state((__smp)) == (csio_sm_state_t)(__state))
> +
Why does any of the sm_state stuff need to be in macros..? Please
inline all of this code.
> +#define CSIO_ASSERT(cond) \
> +do { \
> + if (unlikely(!((cond)))) \
> + BUG(); \
> +} while (0)
> +
> +#ifdef __CSIO_DEBUG__
> +#define CSIO_DB_ASSERT(__c) CSIO_ASSERT((__c))
> +#else
> +#define CSIO_DB_ASSERT(__c)
> +#endif
> +
> +#endif /* ifndef __CSIO_DEFS_H__ */
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
2012-08-23 19:58 ` Nicholas A. Bellinger
@ 2012-08-24 18:36 ` Naresh Kumar Inna
2012-08-24 21:17 ` Nicholas A. Bellinger
0 siblings, 1 reply; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-24 18:36 UTC (permalink / raw)
To: Nicholas A. Bellinger
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/24/2012 1:28 AM, Nicholas A. Bellinger wrote:
> On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
>> This patch contains the first set of the header files for csiostor driver.
>>
>> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
>> ---
>> drivers/scsi/csiostor/csio_defs.h | 143 ++++++
>> drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
>> drivers/scsi/csiostor/csio_hw.h | 668 ++++++++++++++++++++++++
>> drivers/scsi/csiostor/csio_init.h | 158 ++++++
>> 4 files changed, 1812 insertions(+), 0 deletions(-)
>> create mode 100644 drivers/scsi/csiostor/csio_defs.h
>> create mode 100644 drivers/scsi/csiostor/csio_fcoe_proto.h
>> create mode 100644 drivers/scsi/csiostor/csio_hw.h
>> create mode 100644 drivers/scsi/csiostor/csio_init.h
>>
>
> Hi Naresh,
>
> Just commenting on csio_defs.h bits here... As Robert mentioned, you'll
> need to convert the driver to use (or add to) upstream protocol
> definitions and drop the csio_fcoe_proto.h bits..
>
Hi Nicholas,
I would like take up the discussion of the protocol header file in that
email thread. Please find the rest of my replies inline.
Thanks for reviewing,
Naresh.
>> diff --git a/drivers/scsi/csiostor/csio_defs.h b/drivers/scsi/csiostor/csio_defs.h
>> new file mode 100644
>> index 0000000..4f1c713
>> --- /dev/null
>> +++ b/drivers/scsi/csiostor/csio_defs.h
>> @@ -0,0 +1,143 @@
>> +/*
>> + * This file is part of the Chelsio FCoE driver for Linux.
>> + *
>> + * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
>> + *
>> + * This software is available to you under a choice of one of two
>> + * licenses. You may choose to be licensed under the terms of the GNU
>> + * General Public License (GPL) Version 2, available from the file
>> + * COPYING in the main directory of this source tree, or the
>> + * OpenIB.org BSD license below:
>> + *
>> + * Redistribution and use in source and binary forms, with or
>> + * without modification, are permitted provided that the following
>> + * conditions are met:
>> + *
>> + * - Redistributions of source code must retain the above
>> + * copyright notice, this list of conditions and the following
>> + * disclaimer.
>> + *
>> + * - Redistributions in binary form must reproduce the above
>> + * copyright notice, this list of conditions and the following
>> + * disclaimer in the documentation and/or other materials
>> + * provided with the distribution.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#ifndef __CSIO_DEFS_H__
>> +#define __CSIO_DEFS_H__
>> +
>> +#include <linux/kernel.h>
>> +#include <linux/timer.h>
>> +#include <linux/list.h>
>> +#include <linux/bug.h>
>> +#include <linux/pci.h>
>> +#include <linux/jiffies.h>
>> +
>> +/* Function returns */
>> +enum csio_retval {
>> + CSIO_SUCCESS = 0,
>> + CSIO_INVAL = 1,
>> + CSIO_BUSY = 2,
>> + CSIO_NOSUPP = 3,
>> + CSIO_TIMEOUT = 4,
>> + CSIO_NOMEM = 5,
>> + CSIO_NOPERM = 6,
>> + CSIO_RETRY = 7,
>> + CSIO_EPROTO = 8,
>> + CSIO_EIO = 9,
>> + CSIO_CANCELLED = 10,
>> +};
>> +
>
> Please don't assign macros for errno's, and give them positive values.
>
Although some of these return values appear to be mapped to errno
values, there are others that do not have an errno equivalent (example
CSIO_CANCELLED). We may have future needs to have more driver/protocol
specific return values as well. What do you suggest?
>> +#define csio_retval_t enum csio_retval
>
> Please get rid of this csio_retval_t nonsense.
I can get rid of the typedef and use enum csio_retval instead.
>
>> +
>> +enum {
>> + CSIO_FALSE = 0,
>> + CSIO_TRUE = 1,
>> +};
>> +
>
> Same here, please use normal Boolean macros
>
>> +#define CSIO_ROUNDUP(__v, __r) (((__v) + (__r) - 1) / (__r))
>> +#define CSIO_INVALID_IDX 0xFFFFFFFF
>> +#define csio_inc_stats(elem, val) ((elem)->stats.val++)
>> +#define csio_dec_stats(elem, val) ((elem)->stats.val--)
>
> No reason for either of this stats inc+dec macros. Please drop them.
I will get rid of them.
>
>> +#define csio_valid_wwn(__n) ((*__n >> 4) == 0x5 ? CSIO_TRUE : \
>> + CSIO_FALSE)
>> +#define CSIO_WORD_TO_BYTE 4
>> +
>> +static inline int
>> +csio_list_deleted(struct list_head *list)
>> +{
>> + return ((list->next == list) && (list->prev == list));
>> +}
>> +
>> +#define csio_list_next(elem) (((struct list_head *)(elem))->next)
>> +#define csio_list_prev(elem) (((struct list_head *)(elem))->prev)
>> +
>> +#define csio_deq_from_head(head, elem) \
>> +do { \
>> + if (!list_empty(head)) { \
>> + *((struct list_head **)(elem)) = csio_list_next((head)); \
>> + csio_list_next((head)) = \
>> + csio_list_next(csio_list_next((head))); \
>> + csio_list_prev(csio_list_next((head))) = (head); \
>> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
>> + } else \
>> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
>> +} while (0)
>> +
>
> This code is confusing as hell.. Why can't you just use normal list.h
> macros for this..?
I have not found an equivalent function in list.h that does the above
and the following macro. Could you please point me to it? I have seen a
couple of other drivers define their own macros to achieve what this
macro does, hence I assumed there isnt a list.h macro that does this.
>> +#define csio_deq_from_tail(head, elem) \
>> +do { \
>> + if (!list_empty(head)) { \
>> + *((struct list_head **)(elem)) = csio_list_prev((head)); \
>> + csio_list_prev((head)) = \
>> + csio_list_prev(csio_list_prev((head))); \
>> + csio_list_next(csio_list_prev((head))) = (head); \
>> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
>> + } else \
>> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
>> +} while (0)
>> +
>
> Same here.. Please don't use macros like this.
>
>> +/* State machine */
>> +typedef void (*csio_sm_state_t)(void *, uint32_t);
>> +
>> +struct csio_sm {
>> + struct list_head sm_list;
>> + csio_sm_state_t sm_state;
>> +};
>> +
>> +#define csio_init_state(__smp, __state) \
>> + (((struct csio_sm *)(__smp))->sm_state = (csio_sm_state_t)(__state))
>> +
>> +#define csio_set_state(__smp, __state) \
>> + (((struct csio_sm *)(__smp))->sm_state = (csio_sm_state_t)(__state))
>> +
>> +
>> +#define csio_post_event(__smp, __evt) \
>> + (((struct csio_sm *)(__smp))->sm_state((__smp), (uint32_t)(__evt)))
>> +
>> +#define csio_get_state(__smp) (((struct csio_sm *)(__smp))->sm_state)
>> +
>> +#define csio_match_state(__smp, __state) \
>> + (csio_get_state((__smp)) == (csio_sm_state_t)(__state))
>> +
>
> Why does any of the sm_state stuff need to be in macros..? Please
> inline all of this code.
I will change them to static inline.
>
>> +#define CSIO_ASSERT(cond) \
>> +do { \
>> + if (unlikely(!((cond)))) \
>> + BUG(); \
>> +} while (0)
>> +
>> +#ifdef __CSIO_DEBUG__
>> +#define CSIO_DB_ASSERT(__c) CSIO_ASSERT((__c))
>> +#else
>> +#define CSIO_DB_ASSERT(__c)
>> +#endif
>> +
>> +#endif /* ifndef __CSIO_DEFS_H__ */
>
>
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
2012-08-24 18:36 ` Naresh Kumar Inna
@ 2012-08-24 21:17 ` Nicholas A. Bellinger
2012-08-25 18:09 ` Naresh Kumar Inna
0 siblings, 1 reply; 33+ messages in thread
From: Nicholas A. Bellinger @ 2012-08-24 21:17 UTC (permalink / raw)
To: Naresh Kumar Inna
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On Sat, 2012-08-25 at 00:06 +0530, Naresh Kumar Inna wrote:
> On 8/24/2012 1:28 AM, Nicholas A. Bellinger wrote:
> > On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
> >> This patch contains the first set of the header files for csiostor driver.
> >>
> >> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
> >> ---
> >> drivers/scsi/csiostor/csio_defs.h | 143 ++++++
> >> drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
> >> drivers/scsi/csiostor/csio_hw.h | 668 ++++++++++++++++++++++++
> >> drivers/scsi/csiostor/csio_init.h | 158 ++++++
> >> 4 files changed, 1812 insertions(+), 0 deletions(-)
> >> create mode 100644 drivers/scsi/csiostor/csio_defs.h
> >> create mode 100644 drivers/scsi/csiostor/csio_fcoe_proto.h
> >> create mode 100644 drivers/scsi/csiostor/csio_hw.h
> >> create mode 100644 drivers/scsi/csiostor/csio_init.h
> >>
> >
> > Hi Naresh,
> >
> > Just commenting on csio_defs.h bits here... As Robert mentioned, you'll
> > need to convert the driver to use (or add to) upstream protocol
> > definitions and drop the csio_fcoe_proto.h bits..
> >
>
> Hi Nicholas,
>
> I would like take up the discussion of the protocol header file in that
> email thread. Please find the rest of my replies inline.
>
> Thanks for reviewing,
> Naresh.
>
> >> diff --git a/drivers/scsi/csiostor/csio_defs.h b/drivers/scsi/csiostor/csio_defs.h
> >> new file mode 100644
> >> index 0000000..4f1c713
> >> --- /dev/null
> >> +++ b/drivers/scsi/csiostor/csio_defs.h
<SNIP>
> >> +#ifndef __CSIO_DEFS_H__
> >> +#define __CSIO_DEFS_H__
> >> +
> >> +#include <linux/kernel.h>
> >> +#include <linux/timer.h>
> >> +#include <linux/list.h>
> >> +#include <linux/bug.h>
> >> +#include <linux/pci.h>
> >> +#include <linux/jiffies.h>
> >> +
> >> +/* Function returns */
> >> +enum csio_retval {
> >> + CSIO_SUCCESS = 0,
> >> + CSIO_INVAL = 1,
> >> + CSIO_BUSY = 2,
> >> + CSIO_NOSUPP = 3,
> >> + CSIO_TIMEOUT = 4,
> >> + CSIO_NOMEM = 5,
> >> + CSIO_NOPERM = 6,
> >> + CSIO_RETRY = 7,
> >> + CSIO_EPROTO = 8,
> >> + CSIO_EIO = 9,
> >> + CSIO_CANCELLED = 10,
> >> +};
> >> +
> >
> > Please don't assign macros for errno's, and give them positive values.
> >
>
> Although some of these return values appear to be mapped to errno
> values, there are others that do not have an errno equivalent (example
> CSIO_CANCELLED). We may have future needs to have more driver/protocol
> specific return values as well. What do you suggest?
>
Convert all functions aside from CSIO_CANCELLED to use normal negative
return values from include/asm-generic/error[-base].h
For the CSIO_CANCELLED case, propagate this status up to the specific
caller using another method..
> >> +#define csio_retval_t enum csio_retval
> >
> > Please get rid of this csio_retval_t nonsense.
>
> I can get rid of the typedef and use enum csio_retval instead.
>
Using a LLD defined retval where %90 of the items are from errno.h is
code duplication. Please get rid of this.
> >
> >> +
> >> +enum {
> >> + CSIO_FALSE = 0,
> >> + CSIO_TRUE = 1,
> >> +};
> >> +
> >
> > Same here, please use normal Boolean macros
> >
> >> +#define CSIO_ROUNDUP(__v, __r) (((__v) + (__r) - 1) / (__r))
> >> +#define CSIO_INVALID_IDX 0xFFFFFFFF
> >> +#define csio_inc_stats(elem, val) ((elem)->stats.val++)
> >> +#define csio_dec_stats(elem, val) ((elem)->stats.val--)
> >
> > No reason for either of this stats inc+dec macros. Please drop them.
>
> I will get rid of them.
>
> >
> >> +#define csio_valid_wwn(__n) ((*__n >> 4) == 0x5 ? CSIO_TRUE : \
> >> + CSIO_FALSE)
> >> +#define CSIO_WORD_TO_BYTE 4
> >> +
> >> +static inline int
> >> +csio_list_deleted(struct list_head *list)
> >> +{
> >> + return ((list->next == list) && (list->prev == list));
> >> +}
> >> +
> >> +#define csio_list_next(elem) (((struct list_head *)(elem))->next)
> >> +#define csio_list_prev(elem) (((struct list_head *)(elem))->prev)
> >> +
> >> +#define csio_deq_from_head(head, elem) \
> >> +do { \
> >> + if (!list_empty(head)) { \
> >> + *((struct list_head **)(elem)) = csio_list_next((head)); \
> >> + csio_list_next((head)) = \
> >> + csio_list_next(csio_list_next((head))); \
> >> + csio_list_prev(csio_list_next((head))) = (head); \
> >> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
> >> + } else \
> >> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
> >> +} while (0)
> >> +
> >
> > This code is confusing as hell.. Why can't you just use normal list.h
> > macros for this..?
>
> I have not found an equivalent function in list.h that does the above
> and the following macro. Could you please point me to it? I have seen a
> couple of other drivers define their own macros to achieve what this
> macro does, hence I assumed there isnt a list.h macro that does this.
>
AFAICT all that csio_deq_from_head code is supposed to do is pull an
item off a list, right..? Why not just:
while (!list_empty(list)) {
elem = list_first_entry(list, struct elem_type,
elm_list);
list_del_init(&elem->elm_list);
<do work>
<free *elem memory>
}
> >> +#define csio_deq_from_tail(head, elem) \
> >> +do { \
> >> + if (!list_empty(head)) { \
> >> + *((struct list_head **)(elem)) = csio_list_prev((head)); \
> >> + csio_list_prev((head)) = \
> >> + csio_list_prev(csio_list_prev((head))); \
> >> + csio_list_next(csio_list_prev((head))) = (head); \
> >> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
> >> + } else \
> >> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
> >> +} while (0)
> >> +
> >
> > Same here.. Please don't use macros like this.
> >
AFIACT csio_deq_from_tail is unused..?
Please remove it..
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
2012-08-24 21:17 ` Nicholas A. Bellinger
@ 2012-08-25 18:09 ` Naresh Kumar Inna
2012-08-25 18:40 ` Nicholas A. Bellinger
0 siblings, 1 reply; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-25 18:09 UTC (permalink / raw)
To: Nicholas A. Bellinger
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/25/2012 2:47 AM, Nicholas A. Bellinger wrote:
> On Sat, 2012-08-25 at 00:06 +0530, Naresh Kumar Inna wrote:
>> On 8/24/2012 1:28 AM, Nicholas A. Bellinger wrote:
>>> On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
>>>> This patch contains the first set of the header files for csiostor driver.
>>>>
>>>> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
>>>> ---
>>>> drivers/scsi/csiostor/csio_defs.h | 143 ++++++
>>>> drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
>>>> drivers/scsi/csiostor/csio_hw.h | 668 ++++++++++++++++++++++++
>>>> drivers/scsi/csiostor/csio_init.h | 158 ++++++
>>>> 4 files changed, 1812 insertions(+), 0 deletions(-)
>>>> create mode 100644 drivers/scsi/csiostor/csio_defs.h
>>>> create mode 100644 drivers/scsi/csiostor/csio_fcoe_proto.h
>>>> create mode 100644 drivers/scsi/csiostor/csio_hw.h
>>>> create mode 100644 drivers/scsi/csiostor/csio_init.h
>>>>
>>>
>>> Hi Naresh,
>>>
>>> Just commenting on csio_defs.h bits here... As Robert mentioned, you'll
>>> need to convert the driver to use (or add to) upstream protocol
>>> definitions and drop the csio_fcoe_proto.h bits..
>>>
>>
>> Hi Nicholas,
>>
>> I would like take up the discussion of the protocol header file in that
>> email thread. Please find the rest of my replies inline.
>>
>> Thanks for reviewing,
>> Naresh.
>>
>>>> diff --git a/drivers/scsi/csiostor/csio_defs.h b/drivers/scsi/csiostor/csio_defs.h
>>>> new file mode 100644
>>>> index 0000000..4f1c713
>>>> --- /dev/null
>>>> +++ b/drivers/scsi/csiostor/csio_defs.h
>
> <SNIP>
>
>>>> +#ifndef __CSIO_DEFS_H__
>>>> +#define __CSIO_DEFS_H__
>>>> +
>>>> +#include <linux/kernel.h>
>>>> +#include <linux/timer.h>
>>>> +#include <linux/list.h>
>>>> +#include <linux/bug.h>
>>>> +#include <linux/pci.h>
>>>> +#include <linux/jiffies.h>
>>>> +
>>>> +/* Function returns */
>>>> +enum csio_retval {
>>>> + CSIO_SUCCESS = 0,
>>>> + CSIO_INVAL = 1,
>>>> + CSIO_BUSY = 2,
>>>> + CSIO_NOSUPP = 3,
>>>> + CSIO_TIMEOUT = 4,
>>>> + CSIO_NOMEM = 5,
>>>> + CSIO_NOPERM = 6,
>>>> + CSIO_RETRY = 7,
>>>> + CSIO_EPROTO = 8,
>>>> + CSIO_EIO = 9,
>>>> + CSIO_CANCELLED = 10,
>>>> +};
>>>> +
>>>
>>> Please don't assign macros for errno's, and give them positive values.
>>>
>>
>> Although some of these return values appear to be mapped to errno
>> values, there are others that do not have an errno equivalent (example
>> CSIO_CANCELLED). We may have future needs to have more driver/protocol
>> specific return values as well. What do you suggest?
>>
>
> Convert all functions aside from CSIO_CANCELLED to use normal negative
> return values from include/asm-generic/error[-base].h
>
> For the CSIO_CANCELLED case, propagate this status up to the specific
> caller using another method..
>
>>>> +#define csio_retval_t enum csio_retval
>>>
>>> Please get rid of this csio_retval_t nonsense.
>>
>> I can get rid of the typedef and use enum csio_retval instead.
>>
>
> Using a LLD defined retval where %90 of the items are from errno.h is
> code duplication. Please get rid of this.
>
OK I will switch over to the errno values.
>>>
>>>> +
>>>> +enum {
>>>> + CSIO_FALSE = 0,
>>>> + CSIO_TRUE = 1,
>>>> +};
>>>> +
>>>
>>> Same here, please use normal Boolean macros
>>>
>>>> +#define CSIO_ROUNDUP(__v, __r) (((__v) + (__r) - 1) / (__r))
>>>> +#define CSIO_INVALID_IDX 0xFFFFFFFF
>>>> +#define csio_inc_stats(elem, val) ((elem)->stats.val++)
>>>> +#define csio_dec_stats(elem, val) ((elem)->stats.val--)
>>>
>>> No reason for either of this stats inc+dec macros. Please drop them.
>>
>> I will get rid of them.
>>
>>>
>>>> +#define csio_valid_wwn(__n) ((*__n >> 4) == 0x5 ? CSIO_TRUE : \
>>>> + CSIO_FALSE)
>>>> +#define CSIO_WORD_TO_BYTE 4
>>>> +
>>>> +static inline int
>>>> +csio_list_deleted(struct list_head *list)
>>>> +{
>>>> + return ((list->next == list) && (list->prev == list));
>>>> +}
>>>> +
>>>> +#define csio_list_next(elem) (((struct list_head *)(elem))->next)
>>>> +#define csio_list_prev(elem) (((struct list_head *)(elem))->prev)
>>>> +
>>>> +#define csio_deq_from_head(head, elem) \
>>>> +do { \
>>>> + if (!list_empty(head)) { \
>>>> + *((struct list_head **)(elem)) = csio_list_next((head)); \
>>>> + csio_list_next((head)) = \
>>>> + csio_list_next(csio_list_next((head))); \
>>>> + csio_list_prev(csio_list_next((head))) = (head); \
>>>> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
>>>> + } else \
>>>> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
>>>> +} while (0)
>>>> +
>>>
>>> This code is confusing as hell.. Why can't you just use normal list.h
>>> macros for this..?
>>
>> I have not found an equivalent function in list.h that does the above
>> and the following macro. Could you please point me to it? I have seen a
>> couple of other drivers define their own macros to achieve what this
>> macro does, hence I assumed there isnt a list.h macro that does this.
>>
>
> AFAICT all that csio_deq_from_head code is supposed to do is pull an
> item off a list, right..? Why not just:
>
> while (!list_empty(list)) {
> elem = list_first_entry(list, struct elem_type,
> elm_list);
> list_del_init(&elem->elm_list);
>
> <do work>
> <free *elem memory>
> }
>
I will try to come up with a simpler static inline version of the macro.
Would that work?
>>>> +#define csio_deq_from_tail(head, elem) \
>>>> +do { \
>>>> + if (!list_empty(head)) { \
>>>> + *((struct list_head **)(elem)) = csio_list_prev((head)); \
>>>> + csio_list_prev((head)) = \
>>>> + csio_list_prev(csio_list_prev((head))); \
>>>> + csio_list_next(csio_list_prev((head))) = (head); \
>>>> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
>>>> + } else \
>>>> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
>>>> +} while (0)
>>>> +
>>>
>>> Same here.. Please don't use macros like this.
>>>
>
> AFIACT csio_deq_from_tail is unused..?
>
> Please remove it..
>
I will remove it.
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
2012-08-25 18:09 ` Naresh Kumar Inna
@ 2012-08-25 18:40 ` Nicholas A. Bellinger
2012-08-25 19:01 ` Naresh Kumar Inna
0 siblings, 1 reply; 33+ messages in thread
From: Nicholas A. Bellinger @ 2012-08-25 18:40 UTC (permalink / raw)
To: Naresh Kumar Inna
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On Sat, 2012-08-25 at 23:39 +0530, Naresh Kumar Inna wrote:
> On 8/25/2012 2:47 AM, Nicholas A. Bellinger wrote:
> > On Sat, 2012-08-25 at 00:06 +0530, Naresh Kumar Inna wrote:
> >> On 8/24/2012 1:28 AM, Nicholas A. Bellinger wrote:
> >>> On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
> >>>> This patch contains the first set of the header files for csiostor driver.
> >>>>
> >>>> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
> >>>> ---
> >>>> drivers/scsi/csiostor/csio_defs.h | 143 ++++++
> >>>> drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
> >>>> drivers/scsi/csiostor/csio_hw.h | 668 ++++++++++++++++++++++++
> >>>> drivers/scsi/csiostor/csio_init.h | 158 ++++++
> >>>> 4 files changed, 1812 insertions(+), 0 deletions(-)
> >>>> create mode 100644 drivers/scsi/csiostor/csio_defs.h
> >>>> create mode 100644 drivers/scsi/csiostor/csio_fcoe_proto.h
> >>>> create mode 100644 drivers/scsi/csiostor/csio_hw.h
> >>>> create mode 100644 drivers/scsi/csiostor/csio_init.h
> >>>>
> >>>
> >>> Hi Naresh,
> >>>
> >>> Just commenting on csio_defs.h bits here... As Robert mentioned, you'll
> >>> need to convert the driver to use (or add to) upstream protocol
> >>> definitions and drop the csio_fcoe_proto.h bits..
> >>>
> >>
> >> Hi Nicholas,
> >>
> >> I would like take up the discussion of the protocol header file in that
> >> email thread. Please find the rest of my replies inline.
> >>
> >> Thanks for reviewing,
> >> Naresh.
> >>
> >>>> diff --git a/drivers/scsi/csiostor/csio_defs.h b/drivers/scsi/csiostor/csio_defs.h
> >>>> new file mode 100644
> >>>> index 0000000..4f1c713
> >>>> --- /dev/null
> >>>> +++ b/drivers/scsi/csiostor/csio_defs.h
<SNIP>
> >>>> +static inline int
> >>>> +csio_list_deleted(struct list_head *list)
> >>>> +{
> >>>> + return ((list->next == list) && (list->prev == list));
> >>>> +}
> >>>> +
> >>>> +#define csio_list_next(elem) (((struct list_head *)(elem))->next)
> >>>> +#define csio_list_prev(elem) (((struct list_head *)(elem))->prev)
> >>>> +
> >>>> +#define csio_deq_from_head(head, elem) \
> >>>> +do { \
> >>>> + if (!list_empty(head)) { \
> >>>> + *((struct list_head **)(elem)) = csio_list_next((head)); \
> >>>> + csio_list_next((head)) = \
> >>>> + csio_list_next(csio_list_next((head))); \
> >>>> + csio_list_prev(csio_list_next((head))) = (head); \
> >>>> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
> >>>> + } else \
> >>>> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
> >>>> +} while (0)
> >>>> +
> >>>
> >>> This code is confusing as hell.. Why can't you just use normal list.h
> >>> macros for this..?
> >>
> >> I have not found an equivalent function in list.h that does the above
> >> and the following macro. Could you please point me to it? I have seen a
> >> couple of other drivers define their own macros to achieve what this
> >> macro does, hence I assumed there isnt a list.h macro that does this.
> >>
> >
> > AFAICT all that csio_deq_from_head code is supposed to do is pull an
> > item off a list, right..? Why not just:
> >
> > while (!list_empty(list)) {
> > elem = list_first_entry(list, struct elem_type,
> > elm_list);
> > list_del_init(&elem->elm_list);
> >
> > <do work>
> > <free *elem memory>
> > }
> >
>
> I will try to come up with a simpler static inline version of the macro.
> Would that work?
No. The point is that the above code is a disaster, and AFAICT there is
no reason why any of it is necessary to begin with at all.
Why can't csio_deq_from_head() just become list_first_entry() +
list_del_init() to do the exact same thing without all of the extra
overhead of list_head pointer de-reference + assignments..?
--nab
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1).
2012-08-25 18:40 ` Nicholas A. Bellinger
@ 2012-08-25 19:01 ` Naresh Kumar Inna
0 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-25 19:01 UTC (permalink / raw)
To: Nicholas A. Bellinger
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/26/2012 12:10 AM, Nicholas A. Bellinger wrote:
> On Sat, 2012-08-25 at 23:39 +0530, Naresh Kumar Inna wrote:
>> On 8/25/2012 2:47 AM, Nicholas A. Bellinger wrote:
>>> On Sat, 2012-08-25 at 00:06 +0530, Naresh Kumar Inna wrote:
>>>> On 8/24/2012 1:28 AM, Nicholas A. Bellinger wrote:
>>>>> On Fri, 2012-08-24 at 03:57 +0530, Naresh Kumar Inna wrote:
>>>>>> This patch contains the first set of the header files for csiostor driver.
>>>>>>
>>>>>> Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
>>>>>> ---
>>>>>> drivers/scsi/csiostor/csio_defs.h | 143 ++++++
>>>>>> drivers/scsi/csiostor/csio_fcoe_proto.h | 843 +++++++++++++++++++++++++++++++
>>>>>> drivers/scsi/csiostor/csio_hw.h | 668 ++++++++++++++++++++++++
>>>>>> drivers/scsi/csiostor/csio_init.h | 158 ++++++
>>>>>> 4 files changed, 1812 insertions(+), 0 deletions(-)
>>>>>> create mode 100644 drivers/scsi/csiostor/csio_defs.h
>>>>>> create mode 100644 drivers/scsi/csiostor/csio_fcoe_proto.h
>>>>>> create mode 100644 drivers/scsi/csiostor/csio_hw.h
>>>>>> create mode 100644 drivers/scsi/csiostor/csio_init.h
>>>>>>
>>>>>
>>>>> Hi Naresh,
>>>>>
>>>>> Just commenting on csio_defs.h bits here... As Robert mentioned, you'll
>>>>> need to convert the driver to use (or add to) upstream protocol
>>>>> definitions and drop the csio_fcoe_proto.h bits..
>>>>>
>>>>
>>>> Hi Nicholas,
>>>>
>>>> I would like take up the discussion of the protocol header file in that
>>>> email thread. Please find the rest of my replies inline.
>>>>
>>>> Thanks for reviewing,
>>>> Naresh.
>>>>
>>>>>> diff --git a/drivers/scsi/csiostor/csio_defs.h b/drivers/scsi/csiostor/csio_defs.h
>>>>>> new file mode 100644
>>>>>> index 0000000..4f1c713
>>>>>> --- /dev/null
>>>>>> +++ b/drivers/scsi/csiostor/csio_defs.h
>
> <SNIP>
>
>>>>>> +static inline int
>>>>>> +csio_list_deleted(struct list_head *list)
>>>>>> +{
>>>>>> + return ((list->next == list) && (list->prev == list));
>>>>>> +}
>>>>>> +
>>>>>> +#define csio_list_next(elem) (((struct list_head *)(elem))->next)
>>>>>> +#define csio_list_prev(elem) (((struct list_head *)(elem))->prev)
>>>>>> +
>>>>>> +#define csio_deq_from_head(head, elem) \
>>>>>> +do { \
>>>>>> + if (!list_empty(head)) { \
>>>>>> + *((struct list_head **)(elem)) = csio_list_next((head)); \
>>>>>> + csio_list_next((head)) = \
>>>>>> + csio_list_next(csio_list_next((head))); \
>>>>>> + csio_list_prev(csio_list_next((head))) = (head); \
>>>>>> + INIT_LIST_HEAD(*((struct list_head **)(elem))); \
>>>>>> + } else \
>>>>>> + *((struct list_head **)(elem)) = (struct list_head *)NULL;\
>>>>>> +} while (0)
>>>>>> +
>>>>>
>>>>> This code is confusing as hell.. Why can't you just use normal list.h
>>>>> macros for this..?
>>>>
>>>> I have not found an equivalent function in list.h that does the above
>>>> and the following macro. Could you please point me to it? I have seen a
>>>> couple of other drivers define their own macros to achieve what this
>>>> macro does, hence I assumed there isnt a list.h macro that does this.
>>>>
>>>
>>> AFAICT all that csio_deq_from_head code is supposed to do is pull an
>>> item off a list, right..? Why not just:
>>>
>>> while (!list_empty(list)) {
>>> elem = list_first_entry(list, struct elem_type,
>>> elm_list);
>>> list_del_init(&elem->elm_list);
>>>
>>> <do work>
>>> <free *elem memory>
>>> }
>>>
>>
>> I will try to come up with a simpler static inline version of the macro.
>> Would that work?
>
> No. The point is that the above code is a disaster, and AFAICT there is
> no reason why any of it is necessary to begin with at all.
>
> Why can't csio_deq_from_head() just become list_first_entry() +
> list_del_init() to do the exact same thing without all of the extra
> overhead of list_head pointer de-reference + assignments..?
>
> --nab
>
Yes, that's what I was trying to say. csio_deq_from_head() will become
a static function comprising list_first_entry + list_del_init(), with
some checks perhaps.
Thanks,
Naresh.
^ permalink raw reply [flat|nested] 33+ messages in thread
* [PATCH 7/8] csiostor: Chelsio FCoE offload driver submission (headers part 2).
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
` (5 preceding siblings ...)
2012-08-23 22:27 ` [PATCH 6/8] csiostor: Chelsio FCoE offload driver submission (headers part 1) Naresh Kumar Inna
@ 2012-08-23 22:27 ` Naresh Kumar Inna
2012-08-23 22:27 ` [PATCH 8/8] cxgb4: Chelsio FCoE offload driver submission (cxgb4 common header updates) Naresh Kumar Inna
` (2 subsequent siblings)
9 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
This patch contains the second set of the header files for csiostor driver.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/scsi/csiostor/csio_lnode.h | 244 ++++++++++++++
drivers/scsi/csiostor/csio_mb.h | 278 ++++++++++++++++
drivers/scsi/csiostor/csio_rnode.h | 142 ++++++++
drivers/scsi/csiostor/csio_scsi.h | 332 +++++++++++++++++++
drivers/scsi/csiostor/csio_wr.h | 519 +++++++++++++++++++++++++++++
drivers/scsi/csiostor/t4fw_api_stor.h | 578 +++++++++++++++++++++++++++++++++
6 files changed, 2093 insertions(+), 0 deletions(-)
create mode 100644 drivers/scsi/csiostor/csio_lnode.h
create mode 100644 drivers/scsi/csiostor/csio_mb.h
create mode 100644 drivers/scsi/csiostor/csio_rnode.h
create mode 100644 drivers/scsi/csiostor/csio_scsi.h
create mode 100644 drivers/scsi/csiostor/csio_wr.h
create mode 100644 drivers/scsi/csiostor/t4fw_api_stor.h
diff --git a/drivers/scsi/csiostor/csio_lnode.h b/drivers/scsi/csiostor/csio_lnode.h
new file mode 100644
index 0000000..5845856
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_lnode.h
@@ -0,0 +1,244 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __CSIO_LNODE_H__
+#define __CSIO_LNODE_H__
+
+#include <linux/kref.h>
+#include <linux/timer.h>
+#include <linux/workqueue.h>
+
+
+#include "csio_defs.h"
+#include "csio_hw.h"
+#include "csio_fcoe_proto.h"
+
+#define CSIO_FCOE_MAX_NPIV 128
+#define CSIO_FCOE_MAX_RNODES 2048
+
+extern int csio_fcoe_rnodes;
+extern int csio_fdmi_enable;
+
+/* State machine evets */
+enum csio_ln_ev {
+ CSIO_LNE_NONE = (uint32_t)0,
+ CSIO_LNE_LINKUP,
+ CSIO_LNE_FAB_INIT_DONE,
+ CSIO_LNE_LINK_DOWN,
+ CSIO_LNE_DOWN_LINK,
+ CSIO_LNE_LOGO,
+ CSIO_LNE_CLOSE,
+ CSIO_LNE_MAX_EVENT,
+};
+
+
+struct csio_fcf_info {
+ struct list_head list;
+ uint8_t priority;
+ uint8_t mac[6];
+ uint8_t name_id[8];
+ uint8_t fabric[8];
+ uint16_t vf_id;
+ uint8_t vlan_id;
+ uint16_t max_fcoe_size;
+ uint8_t fc_map[3];
+ uint32_t fka_adv;
+ uint32_t fcfi;
+ uint8_t get_next:1;
+ uint8_t link_aff:1;
+ uint8_t fpma:1;
+ uint8_t spma:1;
+ uint8_t login:1;
+ uint8_t portid;
+ uint8_t spma_mac[6];
+ struct kref kref;
+};
+
+/* Defines for flags */
+#define CSIO_LNF_FIPSUPP 0x00000001 /* Fip Supported */
+#define CSIO_LNF_NPIVSUPP 0x00000002 /* NPIV supported */
+#define CSIO_LNF_LINK_ENABLE 0x00000004 /* Link enabled */
+#define CSIO_LNF_FDMI_ENABLE 0x00000008 /* FDMI support */
+
+/* Transport events */
+enum csio_ln_fc_evt {
+ CSIO_LN_FC_LINKUP = 1,
+ CSIO_LN_FC_LINKDOWN,
+ CSIO_LN_FC_RSCN,
+ CSIO_LN_FC_ATTRIB_UPDATE,
+};
+
+/* Lnode stats */
+struct csio_lnode_stats {
+ uint32_t n_link_up; /* Link down */
+ uint32_t n_link_down; /* Link up */
+ uint32_t n_err; /* error */
+ uint32_t n_err_nomem; /* memory not available */
+ uint32_t n_inval_parm; /* Invalid parameters */
+ uint32_t n_evt_unexp; /* unexpected event */
+ uint32_t n_evt_drop; /* dropped event */
+ uint32_t n_rnode_match; /* matched rnode */
+ uint32_t n_dev_loss_tmo; /* Device loss timeout */
+ uint32_t n_fdmi_err; /* fdmi err */
+ uint32_t n_evt_fw[RSCN_DEV_LOST]; /* fw events */
+ enum csio_ln_ev n_evt_sm[CSIO_LNE_MAX_EVENT]; /* State m/c events */
+ uint32_t n_rnode_alloc; /* rnode allocated */
+ uint32_t n_rnode_free; /* rnode freed */
+ uint32_t n_rnode_nomem; /* rnode alloc failure */
+ uint32_t n_input_requests; /* Input Requests */
+ uint32_t n_output_requests; /* Output Requests */
+ uint32_t n_control_requests; /* Control Requests */
+ uint32_t n_input_bytes; /* Input Bytes */
+ uint32_t n_output_bytes; /* Output Bytes */
+ uint32_t rsvd1;
+};
+
+/* Common Lnode params */
+struct csio_lnode_params {
+ uint32_t ra_tov;
+ uint32_t fcfi;
+ uint32_t log_level; /* Module level for debugging */
+};
+
+/* Lnode */
+struct csio_lnode {
+ struct csio_sm sm; /* State machine + sibling
+ * lnode list.
+ */
+ struct csio_hw *hwp; /* Pointer to the HW module */
+ uint8_t portid; /* Port ID */
+ uint8_t rsvd1;
+ uint16_t rsvd2;
+ uint32_t dev_num; /* Device number */
+ uint32_t flags; /* Flags */
+ struct list_head fcf_lsthead; /* FCF entries */
+ struct csio_fcf_info *fcfinfo; /* FCF in use */
+ struct csio_ioreq *mgmt_req; /* MGMT request */
+
+ /* FCoE identifiers */
+ uint8_t mac[6];
+ uint32_t nport_id;
+ struct csio_service_parms ln_sparm; /* Service parms */
+
+ /* Firmware identifiers */
+ uint32_t fcf_flowid; /*fcf flowid */
+ uint32_t vnp_flowid;
+ uint16_t ssn_cnt; /* Registered Session */
+ uint8_t cur_evt; /* Current event */
+ uint8_t prev_evt; /* Previous event */
+
+ /* Children */
+ struct list_head cln_head; /* Head of the children lnode
+ * list.
+ */
+ uint32_t num_vports; /* Total NPIV/children LNodes*/
+ struct csio_lnode *pln; /* Parent lnode of child
+ * lnodes.
+ */
+ struct list_head cmpl_q; /* Pending I/Os on this lnode */
+
+ /* Remote node information */
+ struct list_head rnhead; /* Head of rnode list */
+ uint32_t num_reg_rnodes; /* Number of rnodes registered
+ * with the host.
+ */
+ uint32_t n_scsi_tgts; /* Number of scsi targets
+ * found
+ */
+ uint32_t last_scan_ntgts;/* Number of scsi targets
+ * found per last scan.
+ */
+ uint32_t tgt_scan_tick; /* timer started after
+ * new tgt found
+ */
+ /* FC transport data */
+ struct fc_vport *fc_vport;
+ struct fc_host_statistics fch_stats;
+
+ struct csio_lnode_stats stats; /* Common lnode stats */
+ struct csio_lnode_params params; /* Common lnode params */
+};
+
+#define csio_lnode_to_hw(ln) ((ln)->hwp)
+#define csio_root_lnode(ln) (csio_lnode_to_hw((ln))->rln)
+#define csio_parent_lnode(ln) ((ln)->pln)
+#define csio_ln_flowid(ln) ((ln)->vnp_flowid)
+#define csio_ln_wwpn(ln) ((ln)->ln_sparm.wwpn)
+#define csio_ln_wwnn(ln) ((ln)->ln_sparm.wwnn)
+
+#define csio_is_root_ln(ln) (((ln) == csio_root_lnode((ln))) ? 1 : 0)
+#define csio_is_phys_ln(ln) (((ln)->pln == NULL) ? 1 : 0)
+#define csio_is_npiv_ln(ln) (((ln)->pln != NULL) ? 1 : 0)
+
+
+#define csio_ln_dbg(_ln, _fmt, ...) \
+ csio_dbg(_ln->hwp, "%x:%x "_fmt, CSIO_DEVID_HI(_ln), \
+ CSIO_DEVID_LO(_ln), ##__VA_ARGS__);
+
+#define csio_ln_err(_ln, _fmt, ...) \
+ csio_err(_ln->hwp, "%x:%x "_fmt, CSIO_DEVID_HI(_ln), \
+ CSIO_DEVID_LO(_ln), ##__VA_ARGS__);
+
+#define csio_ln_warn(_ln, _fmt, ...) \
+ csio_warn(_ln->hwp, "%x:%x "_fmt, CSIO_DEVID_HI(_ln), \
+ CSIO_DEVID_LO(_ln), ##__VA_ARGS__);
+
+/* HW->Lnode notifications */
+enum csio_ln_notify {
+ CSIO_LN_NOTIFY_HWREADY = 1,
+ CSIO_LN_NOTIFY_HWSTOP,
+ CSIO_LN_NOTIFY_HWREMOVE,
+ CSIO_LN_NOTIFY_HWRESET,
+};
+
+void csio_fcoe_fwevt_handler(struct csio_hw *, __u8 cpl_op, __be64 *);
+int csio_is_lnode_ready(struct csio_lnode *);
+void csio_lnode_state_to_str(struct csio_lnode *ln, int8_t *str);
+struct csio_lnode *csio_lnode_lookup_by_wwpn(struct csio_hw *, uint8_t *);
+csio_retval_t csio_get_phy_port_stats(struct csio_hw *, uint8_t ,
+ struct fw_fcoe_port_stats *);
+int csio_scan_done(struct csio_lnode *, unsigned long, unsigned long,
+ unsigned long, unsigned long);
+void csio_notify_lnodes(struct csio_hw *, enum csio_ln_notify);
+void csio_disable_lnodes(struct csio_hw *, uint8_t, bool);
+void csio_lnode_async_event(struct csio_lnode *, enum csio_ln_fc_evt);
+csio_retval_t csio_ln_fdmi_start(struct csio_lnode *, void *);
+csio_retval_t csio_lnode_start(struct csio_lnode *);
+void csio_lnode_stop(struct csio_lnode *);
+void csio_lnode_close(struct csio_lnode *);
+csio_retval_t csio_lnode_init(struct csio_lnode *, struct csio_hw *,
+ struct csio_lnode *);
+void csio_lnode_exit(struct csio_lnode *);
+
+#endif /* ifndef __CSIO_LNODE_H__ */
diff --git a/drivers/scsi/csiostor/csio_mb.h b/drivers/scsi/csiostor/csio_mb.h
new file mode 100644
index 0000000..f3e578b
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_mb.h
@@ -0,0 +1,278 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __CSIO_MB_H__
+#define __CSIO_MB_H__
+
+#include <linux/timer.h>
+#include <linux/completion.h>
+
+#include "t4fw_api.h"
+#include "t4fw_api_stor.h"
+#include "csio_defs.h"
+
+#define CSIO_STATS_OFFSET (2)
+#define CSIO_NUM_STATS_PER_MB (6)
+
+struct fw_fcoe_port_cmd_params {
+ uint8_t portid;
+ uint8_t idx;
+ uint8_t nstats;
+};
+
+#define CSIO_DUMP_MB(__hw, __num, __mb) \
+ csio_dbg(__hw, "\t%llx %llx %llx %llx %llx %llx %llx %llx\n", \
+ (unsigned long long)csio_rd_reg64(__hw, __mb), \
+ (unsigned long long)csio_rd_reg64(__hw, __mb + 8), \
+ (unsigned long long)csio_rd_reg64(__hw, __mb + 16), \
+ (unsigned long long)csio_rd_reg64(__hw, __mb + 24), \
+ (unsigned long long)csio_rd_reg64(__hw, __mb + 32), \
+ (unsigned long long)csio_rd_reg64(__hw, __mb + 40), \
+ (unsigned long long)csio_rd_reg64(__hw, __mb + 48), \
+ (unsigned long long)csio_rd_reg64(__hw, __mb + 56))
+
+#define CSIO_MB_MAX_REGS 8
+#define CSIO_MAX_MB_SIZE 64
+#define CSIO_MB_POLL_FREQ 5 /* 5 ms */
+#define CSIO_MB_DEFAULT_TMO FW_CMD_MAX_TIMEOUT
+
+/* Device master in HELLO command */
+enum csio_dev_master { CSIO_MASTER_CANT, CSIO_MASTER_MAY, CSIO_MASTER_MUST };
+
+enum csio_mb_owner { CSIO_MBOWNER_NONE, CSIO_MBOWNER_FW, CSIO_MBOWNER_PL };
+
+enum csio_dev_state {
+ CSIO_DEV_STATE_UNINIT,
+ CSIO_DEV_STATE_INIT,
+ CSIO_DEV_STATE_ERR
+};
+
+#define FW_PARAM_DEV(param) \
+ (FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) | \
+ FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_##param))
+
+#define FW_PARAM_PFVF(param) \
+ (FW_PARAMS_MNEM(FW_PARAMS_MNEM_PFVF) | \
+ FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_PFVF_##param)| \
+ FW_PARAMS_PARAM_Y(0) | \
+ FW_PARAMS_PARAM_Z(0))
+
+enum {
+ PAUSE_RX = 1 << 0,
+ PAUSE_TX = 1 << 1,
+ PAUSE_AUTONEG = 1 << 2
+};
+
+#define CSIO_INIT_MBP(__mbp, __cp, __tmo, __priv, __fn, __clear) \
+do { \
+ if (__clear) \
+ memset((__cp), 0, \
+ CSIO_MB_MAX_REGS * sizeof(__be64)); \
+ INIT_LIST_HEAD(&(__mbp)->list); \
+ (__mbp)->tmo = (__tmo); \
+ (__mbp)->priv = (void *)(__priv); \
+ (__mbp)->mb_cbfn = (__fn); \
+ (__mbp)->mb_size = sizeof(*(__cp)); \
+} while (0)
+
+struct csio_mbm_stats {
+ uint32_t n_req; /* number of mbox req */
+ uint32_t n_rsp; /* number of mbox rsp */
+ uint32_t n_activeq; /* number of mbox req active Q */
+ uint32_t n_cbfnq; /* number of mbox req cbfn Q */
+ uint32_t n_tmo; /* number of mbox timeout */
+ uint32_t n_cancel; /* number of mbox cancel */
+ uint32_t n_err; /* number of mbox error */
+};
+
+/* Driver version of Mailbox */
+struct csio_mb {
+ struct list_head list; /* for req/resp */
+ /* queue in driver */
+ __be64 mb[CSIO_MB_MAX_REGS]; /* MB in HW format */
+ int mb_size; /* Size of this
+ * mailbox.
+ */
+ uint32_t tmo; /* Timeout */
+ struct completion cmplobj; /* MB Completion
+ * object
+ */
+ void (*mb_cbfn) (struct csio_hw *, struct csio_mb *);
+ /* Callback fn */
+ void *priv; /* Owner private ptr */
+};
+
+struct csio_mbm {
+ uint32_t a_mbox; /* Async mbox num */
+ uint32_t intr_idx; /* Interrupt index */
+ struct timer_list timer; /* Mbox timer */
+ struct list_head req_q; /* Mbox request queue */
+ struct list_head cbfn_q; /* Mbox completion q */
+ struct csio_mb *mcurrent; /* Current mailbox */
+ uint32_t req_q_cnt; /* Outstanding mbox
+ * cmds
+ */
+ struct csio_mbm_stats stats; /* Statistics */
+};
+
+#define csio_set_mb_intr_idx(_m, _i) ((_m)->intr_idx = (_i))
+#define csio_get_mb_intr_idx(_m) ((_m)->intr_idx)
+
+struct csio_iq_params;
+struct csio_eq_params;
+
+enum fw_retval csio_mb_fw_retval(struct csio_mb *);
+
+/* MB helpers */
+void csio_mb_hello(struct csio_hw *, struct csio_mb *, uint32_t,
+ uint32_t, uint32_t, enum csio_dev_master,
+ void (*)(struct csio_hw *, struct csio_mb *));
+
+void csio_mb_process_hello_rsp(struct csio_hw *, struct csio_mb *,
+ enum fw_retval *, enum csio_dev_state *,
+ uint8_t *);
+
+void csio_mb_bye(struct csio_hw *, struct csio_mb *, uint32_t,
+ void (*)(struct csio_hw *, struct csio_mb *));
+
+void csio_mb_reset(struct csio_hw *, struct csio_mb *, uint32_t, int, int,
+ void (*)(struct csio_hw *, struct csio_mb *));
+
+void csio_mb_params(struct csio_hw *, struct csio_mb *, uint32_t, unsigned int,
+ unsigned int, unsigned int, const u32 *, u32 *, bool,
+ void (*)(struct csio_hw *, struct csio_mb *));
+
+void csio_mb_process_read_params_rsp(struct csio_hw *, struct csio_mb *,
+ enum fw_retval *, unsigned int , u32 *);
+
+void csio_mb_ldst(struct csio_hw *hw, struct csio_mb *mbp, uint32_t tmo,
+ int reg);
+
+void csio_mb_caps_config(struct csio_hw *, struct csio_mb *, uint32_t,
+ bool, bool, bool, bool,
+ void (*)(struct csio_hw *, struct csio_mb *));
+
+void csio_rss_glb_config(struct csio_hw *, struct csio_mb *,
+ uint32_t, uint8_t, unsigned int,
+ void (*)(struct csio_hw *, struct csio_mb *));
+
+void csio_mb_pfvf(struct csio_hw *, struct csio_mb *, uint32_t,
+ unsigned int, unsigned int, unsigned int,
+ unsigned int, unsigned int, unsigned int,
+ unsigned int, unsigned int, unsigned int,
+ unsigned int, unsigned int, unsigned int,
+ unsigned int, void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_mb_port(struct csio_hw *, struct csio_mb *, uint32_t,
+ uint8_t, bool, uint32_t, uint16_t,
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_mb_process_read_port_rsp(struct csio_hw *, struct csio_mb *,
+ enum fw_retval *, uint16_t *);
+
+void csio_mb_initialize(struct csio_hw *, struct csio_mb *, uint32_t,
+ void (*)(struct csio_hw *, struct csio_mb *));
+
+void csio_mb_iq_alloc_write(struct csio_hw *, struct csio_mb *, void *,
+ uint32_t, struct csio_iq_params *,
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_mb_iq_alloc_write_rsp(struct csio_hw *, struct csio_mb *,
+ enum fw_retval *, struct csio_iq_params *);
+
+void csio_mb_iq_free(struct csio_hw *, struct csio_mb *, void *,
+ uint32_t, struct csio_iq_params *,
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_mb_eq_ofld_alloc_write(struct csio_hw *, struct csio_mb *, void *,
+ uint32_t, struct csio_eq_params *,
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_mb_eq_ofld_alloc_write_rsp(struct csio_hw *, struct csio_mb *,
+ enum fw_retval *, struct csio_eq_params *);
+
+void csio_mb_eq_ofld_free(struct csio_hw *, struct csio_mb *, void *,
+ uint32_t , struct csio_eq_params *,
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_fcoe_read_res_info_init_mb(struct csio_hw *, struct csio_mb *,
+ uint32_t,
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_write_fcoe_link_cond_init_mb(struct csio_lnode *, struct csio_mb *,
+ uint32_t, uint8_t, uint32_t, uint8_t, bool, uint32_t,
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_fcoe_vnp_alloc_init_mb(struct csio_lnode *, struct csio_mb *,
+ uint32_t, uint32_t , uint32_t , uint16_t,
+ uint8_t [8], uint8_t [8],
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_fcoe_vnp_read_init_mb(struct csio_lnode *, struct csio_mb *,
+ uint32_t, uint32_t , uint32_t ,
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_fcoe_vnp_free_init_mb(struct csio_lnode *, struct csio_mb *,
+ uint32_t , uint32_t, uint32_t ,
+ void (*) (struct csio_hw *, struct csio_mb *));
+
+void csio_fcoe_read_fcf_init_mb(struct csio_lnode *, struct csio_mb *,
+ uint32_t, uint32_t, uint32_t,
+ void (*cbfn) (struct csio_hw *, struct csio_mb *));
+
+void csio_fcoe_read_portparams_init_mb(struct csio_hw *hw,
+ struct csio_mb *mbp, uint32_t mb_tmo,
+ struct fw_fcoe_port_cmd_params *portparams,
+ void (*cbfn)(struct csio_hw *, struct csio_mb *));
+
+void csio_mb_process_portparams_rsp(struct csio_hw *hw, struct csio_mb *mbp,
+ enum fw_retval *retval,
+ struct fw_fcoe_port_cmd_params *portparams,
+ struct fw_fcoe_port_stats *portstats);
+
+/* MB module functions */
+csio_retval_t csio_mbm_init(struct csio_mbm *, struct csio_hw *,
+ void (*)(uintptr_t));
+void csio_mbm_exit(struct csio_mbm *);
+void csio_mb_intr_enable(struct csio_hw *);
+void csio_mb_intr_disable(struct csio_hw *);
+
+csio_retval_t csio_mb_issue(struct csio_hw *, struct csio_mb *);
+void csio_mb_completions(struct csio_hw *, struct list_head *);
+csio_retval_t csio_mb_fwevt_handler(struct csio_hw *, __be64 *);
+csio_retval_t csio_mb_isr_handler(struct csio_hw *);
+struct csio_mb *csio_mb_tmo_handler(struct csio_hw *);
+void csio_mb_cancel_all(struct csio_hw *, struct list_head *);
+
+#endif /* ifndef __CSIO_MB_H__ */
diff --git a/drivers/scsi/csiostor/csio_rnode.h b/drivers/scsi/csiostor/csio_rnode.h
new file mode 100644
index 0000000..99f0cd1
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_rnode.h
@@ -0,0 +1,142 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __CSIO_RNODE_H__
+#define __CSIO_RNODE_H__
+
+#include "csio_defs.h"
+#include "csio_fcoe_proto.h"
+
+/* State machine evets */
+enum csio_rn_ev {
+ CSIO_RNFE_NONE = (uint32_t)0, /* None */
+ CSIO_RNFE_LOGGED_IN, /* [N/F]Port login
+ * complete.
+ */
+ CSIO_RNFE_PRLI_DONE, /* PRLI completed */
+ CSIO_RNFE_PLOGI_RECV, /* Received PLOGI */
+ CSIO_RNFE_PRLI_RECV, /* Received PLOGI */
+ CSIO_RNFE_LOGO_RECV, /* Received LOGO */
+ CSIO_RNFE_PRLO_RECV, /* Received PRLO */
+ CSIO_RNFE_DOWN, /* Rnode is down */
+ CSIO_RNFE_CLOSE, /* Close rnode */
+ CSIO_RNFE_NAME_MISSING, /* Rnode name missing
+ * in name server.
+ */
+ CSIO_RNFE_MAX_EVENT,
+};
+
+/* rnode stats */
+struct csio_rnode_stats {
+ uint32_t n_err; /* error */
+ uint32_t n_err_inval; /* invalid parameter */
+ uint32_t n_err_nomem; /* error nomem */
+ uint32_t n_evt_unexp; /* unexpected event */
+ uint32_t n_evt_drop; /* unexpected event */
+ uint32_t n_evt_fw[RSCN_DEV_LOST]; /* fw events */
+ enum csio_rn_ev n_evt_sm[CSIO_RNFE_MAX_EVENT]; /* State m/c events */
+ uint32_t n_lun_rst; /* Number of resets of
+ * of LUNs under this
+ * target
+ */
+ uint32_t n_lun_rst_fail; /* Number of LUN reset
+ * failures.
+ */
+ uint32_t n_tgt_rst; /* Number of target resets */
+ uint32_t n_tgt_rst_fail; /* Number of target reset
+ * failures.
+ */
+};
+
+/* Defines for rnode role */
+#define CSIO_RNFR_INITIATOR 0x1
+#define CSIO_RNFR_TARGET 0x2
+#define CSIO_RNFR_FABRIC 0x4
+#define CSIO_RNFR_NS 0x8
+#define CSIO_RNFR_NPORT 0x10
+
+struct csio_rnode {
+ struct csio_sm sm; /* State machine -
+ * should be the
+ * 1st member
+ */
+ struct csio_lnode *lnp; /* Pointer to owning
+ * Lnode */
+ uint32_t flowid; /* Firmware ID */
+ struct list_head host_cmpl_q; /* SCSI IOs
+ * pending to completed
+ * to Mid-layer.
+ */
+ /* FC identifiers for remote node */
+ uint32_t nport_id;
+ uint16_t fcp_flags; /* FCP Flags */
+ uint8_t cur_evt; /* Current event */
+ uint8_t prev_evt; /* Previous event */
+ uint32_t role; /* Fabric/Target/
+ * Initiator/NS
+ */
+ struct fcoe_rdev_entry *rdev_entry; /* Rdev entry */
+ struct csio_service_parms rn_sparm;
+
+ /* FC transport attributes */
+ struct fc_rport *rport; /* FC transport rport */
+ uint32_t supp_classes; /* Supported FC classes */
+ uint32_t maxframe_size; /* Max Frame size */
+ uint32_t scsi_id; /* Transport given SCSI id */
+
+ struct csio_rnode_stats stats; /* Common rnode stats */
+};
+
+#define csio_rn_flowid(rn) ((rn)->flowid)
+#define csio_rn_wwpn(rn) ((rn)->rn_sparm.wwpn)
+#define csio_rn_wwnn(rn) ((rn)->rn_sparm.wwnn)
+#define csio_rnode_to_lnode(rn) ((rn)->lnp)
+
+int csio_is_rnode_ready(struct csio_rnode *rn);
+void csio_rnode_state_to_str(struct csio_rnode *rn, int8_t *str);
+
+struct csio_rnode *csio_rnode_lookup_portid(struct csio_lnode *, uint32_t);
+struct csio_rnode *csio_confirm_rnode(struct csio_lnode *,
+ uint32_t, struct fcoe_rdev_entry *);
+
+void csio_rnode_fwevt_handler(struct csio_rnode *rn, uint8_t fwevt);
+
+void csio_put_rnode(struct csio_lnode *ln, struct csio_rnode *rn);
+
+void csio_reg_rnode(struct csio_rnode *);
+void csio_unreg_rnode(struct csio_rnode *);
+
+void csio_rnode_devloss_handler(struct csio_rnode *);
+
+#endif /* ifndef __CSIO_RNODE_H__ */
diff --git a/drivers/scsi/csiostor/csio_scsi.h b/drivers/scsi/csiostor/csio_scsi.h
new file mode 100644
index 0000000..c0aa2fb
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_scsi.h
@@ -0,0 +1,332 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __CSIO_SCSI_H__
+#define __CSIO_SCSI_H__
+
+#include <linux/spinlock_types.h>
+#include <linux/completion.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi_eh.h>
+#include <scsi/scsi_tcq.h>
+
+#include "csio_defs.h"
+#include "csio_wr.h"
+
+extern struct scsi_host_template csio_fcoe_shost_template;
+extern struct scsi_host_template csio_fcoe_shost_vport_template;
+
+extern int csio_scsi_eqsize;
+extern int csio_scsi_iqlen;
+extern int csio_scsi_ioreqs;
+extern uint32_t csio_max_scan_tmo;
+extern uint32_t csio_delta_scan_tmo;
+extern int csio_lun_qdepth;
+
+/*
+ **************************** NOTE *******************************
+ * How do we calculate MAX FCoE SCSI SGEs? Here is the math:
+ * Max Egress WR size = 512 bytes
+ * One SCSI egress WR has the following fixed no of bytes:
+ * 48 (sizeof(struct fw_scsi_write[read]_wr)) - FW WR
+ * + 32 (sizeof(struct csio_fcp_cmnd)) - Immediate FCP_CMD
+ * ------
+ * 80
+ * ------
+ * That leaves us with 512 - 96 = 432 bytes for data SGE. Using
+ * struct ulptx_sgl header for the SGE consumes:
+ * - 4 bytes for cmnd_sge.
+ * - 12 bytes for the first SGL.
+ * That leaves us with 416 bytes for the remaining SGE pairs. Which is
+ * is 416 / 24 (size(struct ulptx_sge_pair)) = 17 SGE pairs,
+ * or 34 SGEs. Adding the first SGE fetches us 35 SGEs.
+ */
+#define CSIO_SCSI_MAX_SGE 35
+#define CSIO_SCSI_ABRT_TMO_MS 60000
+#define CSIO_SCSI_LUNRST_TMO_MS 60000
+#define CSIO_SCSI_TM_POLL_MS 2000 /* should be less than
+ * all TM timeouts.
+ */
+#define CSIO_SCSI_IQ_WRSZ 128
+#define CSIO_SCSI_IQSIZE (csio_scsi_iqlen * CSIO_SCSI_IQ_WRSZ)
+
+/* Reference to scsi_cmnd */
+#define csio_scsi_cmnd(req) ((req)->scratch1)
+
+struct csio_scsi_stats {
+ uint64_t n_tot_success; /* Total number of good I/Os */
+ uint32_t n_rn_nr_error; /* No. of remote-node-not-
+ * ready errors
+ */
+ uint32_t n_hw_nr_error; /* No. of hw-module-not-
+ * ready errors
+ */
+ uint32_t n_dmamap_error; /* No. of DMA map erros */
+ uint32_t n_unsupp_sge_error; /* No. of too-many-SGes
+ * errors.
+ */
+ uint32_t n_no_req_error; /* No. of Out-of-ioreqs error */
+ uint32_t n_busy_error; /* No. of CSIO_BUSY errors */
+ uint32_t n_hosterror; /* No. of FW_HOSTERROR I/O */
+ uint32_t n_rsperror; /* No. of response errors */
+ uint32_t n_autosense; /* No. of auto sense replies */
+ uint32_t n_ovflerror; /* No. of overflow errors */
+ uint32_t n_unflerror; /* No. of underflow errors */
+ uint32_t n_rdev_nr_error;/* No. of rdev not
+ * ready errors
+ */
+ uint32_t n_rdev_lost_error;/* No. of rdev lost errors */
+ uint32_t n_rdev_logo_error;/* No. of rdev logo errors */
+ uint32_t n_link_down_error;/* No. of link down errors */
+ uint32_t n_no_xchg_error; /* No. no exchange error */
+ uint32_t n_unknown_error;/* No. of unhandled errors */
+ uint32_t n_aborted; /* No. of aborted I/Os */
+ uint32_t n_abrt_timedout; /* No. of abort timedouts */
+ uint32_t n_abrt_fail; /* No. of abort failures */
+ uint32_t n_abrt_dups; /* No. of duplicate aborts */
+ uint32_t n_abrt_race_comp; /* No. of aborts that raced
+ * with completions.
+ */
+ uint32_t n_abrt_busy_error;/* No. of abort failures
+ * due to CSIO_BUSY.
+ */
+ uint32_t n_closed; /* No. of closed I/Os */
+ uint32_t n_cls_busy_error; /* No. of close failures
+ * due to CSIO_BUSY.
+ */
+ uint32_t n_active; /* No. of IOs in active_q */
+ uint32_t n_tm_active; /* No. of TMs in active_q */
+ uint32_t n_wcbfn; /* No. of I/Os in worker
+ * cbfn q
+ */
+ uint32_t n_free_ioreq; /* No. of freelist entries */
+ uint32_t n_free_ddp; /* No. of DDP freelist */
+ uint32_t n_unaligned; /* No. of Unaligned SGls */
+ uint32_t n_inval_cplop; /* No. invalid CPL op's in IQ */
+ uint32_t n_inval_scsiop; /* No. invalid scsi op's in IQ*/
+};
+
+struct csio_scsim {
+ struct csio_hw *hw; /* Pointer to HW moduel */
+ uint8_t max_sge; /* Max SGE */
+ uint8_t proto_cmd_len; /* Proto specific SCSI
+ * cmd length
+ */
+ uint16_t proto_rsp_len; /* Proto specific SCSI
+ * response length
+ */
+ spinlock_t freelist_lock; /* Lock for ioreq freelist */
+ struct list_head active_q; /* Outstanding SCSI I/Os */
+ struct list_head ioreq_freelist; /* Free list of ioreq's */
+ struct list_head ddp_freelist; /* DDP descriptor freelist */
+ struct csio_scsi_stats stats; /* This module's statistics */
+};
+
+/* State machine defines */
+enum csio_scsi_ev {
+ CSIO_SCSIE_START_IO = 1, /* Start a regular SCSI IO */
+ CSIO_SCSIE_START_TM, /* Start a TM IO */
+ CSIO_SCSIE_COMPLETED, /* IO Completed */
+ CSIO_SCSIE_ABORT, /* Abort IO */
+ CSIO_SCSIE_ABORTED, /* IO Aborted */
+ CSIO_SCSIE_CLOSE, /* Close exchange */
+ CSIO_SCSIE_CLOSED, /* Exchange closed */
+ CSIO_SCSIE_DRVCLEANUP, /* Driver wants to manually
+ * cleanup this I/O.
+ */
+};
+
+enum csio_scsi_lev {
+ CSIO_LEV_ALL = 1,
+ CSIO_LEV_LNODE,
+ CSIO_LEV_RNODE,
+ CSIO_LEV_LUN,
+};
+
+struct csio_scsi_level_data {
+ enum csio_scsi_lev level;
+ struct csio_rnode *rnode;
+ struct csio_lnode *lnode;
+ uint64_t oslun;
+};
+
+static inline struct csio_ioreq *
+csio_get_scsi_ioreq(struct csio_scsim *scm)
+{
+ struct csio_ioreq *ioreq = NULL;
+
+ csio_deq_from_head(&scm->ioreq_freelist, &ioreq);
+ if (ioreq)
+ csio_dec_stats(scm, n_free_ioreq);
+
+ return ioreq;
+}
+
+static inline void
+csio_put_scsi_ioreq(struct csio_scsim *scm, struct csio_ioreq *ioreq)
+{
+ list_add_tail(&ioreq->sm.sm_list, &scm->ioreq_freelist);
+ csio_inc_stats(scm, n_free_ioreq);
+}
+
+static inline void
+csio_put_scsi_ioreq_list(struct csio_scsim *scm, struct list_head *reqlist,
+ int n)
+{
+ list_splice_init(reqlist, &scm->ioreq_freelist);
+ scm->stats.n_free_ioreq += n;
+}
+
+static inline struct csio_dma_buf *
+csio_get_scsi_ddp(struct csio_scsim *scm)
+{
+ struct csio_dma_buf *ddp = NULL;
+
+ csio_deq_from_head(&scm->ddp_freelist, &ddp);
+ if (ddp)
+ csio_dec_stats(scm, n_free_ddp);
+
+ return ddp;
+}
+
+static inline void
+csio_put_scsi_ddp(struct csio_scsim *scm, struct csio_dma_buf *ddp)
+{
+ list_add_tail(&ddp->list, &scm->ddp_freelist);
+ csio_inc_stats(scm, n_free_ddp);
+}
+
+static inline void
+csio_put_scsi_ddp_list(struct csio_scsim *scm, struct list_head *reqlist,
+ int n)
+{
+ list_splice_tail_init(reqlist, &scm->ddp_freelist);
+ scm->stats.n_free_ddp += n;
+}
+
+static inline void
+csio_scsi_completed(struct csio_ioreq *ioreq, struct list_head *cbfn_q)
+{
+ csio_post_event(&ioreq->sm, CSIO_SCSIE_COMPLETED);
+ if (csio_list_deleted(&ioreq->sm.sm_list))
+ list_add_tail(&ioreq->sm.sm_list, cbfn_q);
+}
+
+static inline void
+csio_scsi_aborted(struct csio_ioreq *ioreq, struct list_head *cbfn_q)
+{
+ csio_post_event(&ioreq->sm, CSIO_SCSIE_ABORTED);
+ list_add_tail(&ioreq->sm.sm_list, cbfn_q);
+}
+
+static inline void
+csio_scsi_closed(struct csio_ioreq *ioreq, struct list_head *cbfn_q)
+{
+ csio_post_event(&ioreq->sm, CSIO_SCSIE_CLOSED);
+ list_add_tail(&ioreq->sm.sm_list, cbfn_q);
+}
+
+static inline void
+csio_scsi_drvcleanup(struct csio_ioreq *ioreq)
+{
+ csio_post_event(&ioreq->sm, CSIO_SCSIE_DRVCLEANUP);
+}
+
+/*
+ * csio_scsi_start_io - Kick starts the IO SM.
+ * @req: io request SM.
+ *
+ * needs to be called with lock held.
+ */
+static inline csio_retval_t
+csio_scsi_start_io(struct csio_ioreq *ioreq)
+{
+ csio_post_event(&ioreq->sm, CSIO_SCSIE_START_IO);
+ return ioreq->drv_status;
+}
+
+/*
+ * csio_scsi_start_tm - Kicks off the Task management IO SM.
+ * @req: io request SM.
+ *
+ * needs to be called with lock held.
+ */
+static inline csio_retval_t
+csio_scsi_start_tm(struct csio_ioreq *ioreq)
+{
+ csio_post_event(&ioreq->sm, CSIO_SCSIE_START_TM);
+ return ioreq->drv_status;
+}
+
+/*
+ * csio_scsi_abort - Abort an IO request
+ * @req: io request SM.
+ *
+ * needs to be called with lock held.
+ */
+static inline csio_retval_t
+csio_scsi_abort(struct csio_ioreq *ioreq)
+{
+ csio_post_event(&ioreq->sm, CSIO_SCSIE_ABORT);
+ return ioreq->drv_status;
+}
+
+/*
+ * csio_scsi_close - Close an IO request
+ * @req: io request SM.
+ *
+ * needs to be called with lock held.
+ */
+static inline csio_retval_t
+csio_scsi_close(struct csio_ioreq *ioreq)
+{
+ csio_post_event(&ioreq->sm, CSIO_SCSIE_CLOSE);
+ return ioreq->drv_status;
+}
+
+void csio_scsi_cleanup_io_q(struct csio_scsim *, struct list_head *);
+csio_retval_t csio_scsim_cleanup_io(struct csio_scsim *, bool abort);
+csio_retval_t csio_scsim_cleanup_io_lnode(struct csio_scsim *,
+ struct csio_lnode *);
+struct csio_ioreq *csio_scsi_cmpl_handler(struct csio_hw *, void *, uint32_t,
+ struct csio_fl_dma_buf *,
+ void *, uint8_t **);
+csio_retval_t csio_scsi_qconfig(struct csio_hw *);
+csio_retval_t csio_scsim_init(struct csio_scsim *, struct csio_hw *);
+void csio_scsim_exit(struct csio_scsim *);
+
+#endif /* __CSIO_SCSI_H__ */
diff --git a/drivers/scsi/csiostor/csio_wr.h b/drivers/scsi/csiostor/csio_wr.h
new file mode 100644
index 0000000..004a914
--- /dev/null
+++ b/drivers/scsi/csiostor/csio_wr.h
@@ -0,0 +1,519 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2008-2012 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __CSIO_WR_H__
+#define __CSIO_WR_H__
+
+#include <linux/cache.h>
+
+#include "csio_defs.h"
+#include "t4fw_api.h"
+#include "t4fw_api_stor.h"
+
+/*
+ * SGE register field values.
+ */
+#define X_INGPCIEBOUNDARY_32B 0
+#define X_INGPCIEBOUNDARY_64B 1
+#define X_INGPCIEBOUNDARY_128B 2
+#define X_INGPCIEBOUNDARY_256B 3
+#define X_INGPCIEBOUNDARY_512B 4
+#define X_INGPCIEBOUNDARY_1024B 5
+#define X_INGPCIEBOUNDARY_2048B 6
+#define X_INGPCIEBOUNDARY_4096B 7
+
+/* GTS register */
+#define X_TIMERREG_COUNTER0 0
+#define X_TIMERREG_COUNTER1 1
+#define X_TIMERREG_COUNTER2 2
+#define X_TIMERREG_COUNTER3 3
+#define X_TIMERREG_COUNTER4 4
+#define X_TIMERREG_COUNTER5 5
+#define X_TIMERREG_RESTART_COUNTER 6
+#define X_TIMERREG_UPDATE_CIDX 7
+
+/*
+ * Egress Context field values
+ */
+#define X_FETCHBURSTMIN_16B 0
+#define X_FETCHBURSTMIN_32B 1
+#define X_FETCHBURSTMIN_64B 2
+#define X_FETCHBURSTMIN_128B 3
+
+#define X_FETCHBURSTMAX_64B 0
+#define X_FETCHBURSTMAX_128B 1
+#define X_FETCHBURSTMAX_256B 2
+#define X_FETCHBURSTMAX_512B 3
+
+#define X_HOSTFCMODE_NONE 0
+#define X_HOSTFCMODE_INGRESS_QUEUE 1
+#define X_HOSTFCMODE_STATUS_PAGE 2
+#define X_HOSTFCMODE_BOTH 3
+
+/*
+ * Ingress Context field values
+ */
+#define X_UPDATESCHEDULING_TIMER 0
+#define X_UPDATESCHEDULING_COUNTER_OPTTIMER 1
+
+#define X_UPDATEDELIVERY_NONE 0
+#define X_UPDATEDELIVERY_INTERRUPT 1
+#define X_UPDATEDELIVERY_STATUS_PAGE 2
+#define X_UPDATEDELIVERY_BOTH 3
+
+#define X_INTERRUPTDESTINATION_PCIE 0
+#define X_INTERRUPTDESTINATION_IQ 1
+
+#define X_RSPD_TYPE_FLBUF 0
+#define X_RSPD_TYPE_CPL 1
+#define X_RSPD_TYPE_INTR 2
+
+/* WR status is at the same position as retval in a CMD header */
+#define csio_wr_status(_wr) \
+ (FW_CMD_RETVAL_GET(ntohl(((struct fw_cmd_hdr *)(_wr))->lo)))
+
+struct csio_hw;
+
+extern int csio_intr_coalesce_cnt;
+extern int csio_intr_coalesce_time;
+
+/* Ingress queue params */
+struct csio_iq_params {
+
+ uint8_t iq_start:1;
+ uint8_t iq_stop:1;
+ uint8_t pfn:3;
+
+ uint8_t vfn;
+
+ uint16_t physiqid;
+ uint16_t iqid;
+
+ uint16_t fl0id;
+ uint16_t fl1id;
+
+ uint8_t viid;
+
+ uint8_t type;
+ uint8_t iqasynch;
+ uint8_t reserved4;
+
+ uint8_t iqandst;
+ uint8_t iqanus;
+ uint8_t iqanud;
+
+ uint16_t iqandstindex;
+
+ uint8_t iqdroprss;
+ uint8_t iqpciech;
+ uint8_t iqdcaen;
+
+ uint8_t iqdcacpu;
+ uint8_t iqintcntthresh;
+ uint8_t iqo;
+
+ uint8_t iqcprio;
+ uint8_t iqesize;
+
+ uint16_t iqsize;
+
+ uint64_t iqaddr;
+
+ uint8_t iqflintiqhsen;
+ uint8_t reserved5;
+ uint8_t iqflintcongen;
+ uint8_t iqflintcngchmap;
+
+ uint32_t reserved6;
+
+ uint8_t fl0hostfcmode;
+ uint8_t fl0cprio;
+ uint8_t fl0paden;
+ uint8_t fl0packen;
+ uint8_t fl0congen;
+ uint8_t fl0dcaen;
+
+ uint8_t fl0dcacpu;
+ uint8_t fl0fbmin;
+
+ uint8_t fl0fbmax;
+ uint8_t fl0cidxfthresho;
+ uint8_t fl0cidxfthresh;
+
+ uint16_t fl0size;
+
+ uint64_t fl0addr;
+
+ uint64_t reserved7;
+
+ uint8_t fl1hostfcmode;
+ uint8_t fl1cprio;
+ uint8_t fl1paden;
+ uint8_t fl1packen;
+ uint8_t fl1congen;
+ uint8_t fl1dcaen;
+
+ uint8_t fl1dcacpu;
+ uint8_t fl1fbmin;
+
+ uint8_t fl1fbmax;
+ uint8_t fl1cidxfthresho;
+ uint8_t fl1cidxfthresh;
+
+ uint16_t fl1size;
+
+ uint64_t fl1addr;
+};
+
+/* Egress queue params */
+struct csio_eq_params {
+
+ uint8_t pfn;
+ uint8_t vfn;
+
+ uint8_t eqstart:1;
+ uint8_t eqstop:1;
+
+ uint16_t physeqid;
+ uint32_t eqid;
+
+ uint8_t hostfcmode:2;
+ uint8_t cprio:1;
+ uint8_t pciechn:3;
+
+ uint16_t iqid;
+
+ uint8_t dcaen:1;
+ uint8_t dcacpu:5;
+
+ uint8_t fbmin:3;
+ uint8_t fbmax:3;
+
+ uint8_t cidxfthresho:1;
+ uint8_t cidxfthresh:3;
+
+ uint16_t eqsize;
+
+ uint64_t eqaddr;
+};
+
+struct csio_dma_buf {
+ struct list_head list;
+ void *vaddr; /* Virtual address */
+ dma_addr_t paddr; /* Physical address */
+ uint32_t len; /* Buffer size */
+};
+
+/* Defines for csio_ioreq->datadir */
+enum {
+ CSIO_IOREQF_DMA_BIDI = 0x0,
+ CSIO_IOREQF_DMA_WRITE = 0x1,
+ CSIO_IOREQF_DMA_READ = 0x2,
+ CSIO_IOREQF_DMA_NONE = 0x4,
+};
+
+/* Generic I/O request structure */
+struct csio_ioreq {
+ struct csio_sm sm; /* SM, List
+ * should be the first member
+ */
+ int iq_idx; /* Ingress queue index */
+ int eq_idx; /* Egress queue index */
+ uint32_t nsge; /* Number of SG elements */
+ uint32_t tmo; /* Driver timeout */
+ uint32_t datadir; /* Data direction */
+ struct csio_dma_buf dma_buf; /* Req/resp DMA buffers */
+ uint16_t wr_status; /* WR completion status */
+ uint16_t drv_status; /* Driver internal status */
+ struct csio_lnode *lnode; /* Owner lnode */
+ struct csio_rnode *rnode; /* Src/destination rnode */
+ void (*io_cbfn) (struct csio_hw *, struct csio_ioreq *);
+ /* completion callback */
+ void *scratch1; /* Scratch area 1.
+ */
+ void *scratch2; /* Scratch area 2. */
+ struct list_head gen_list; /* Any list associated with
+ * this ioreq.
+ */
+ uint64_t fw_handle; /* Unique handle passed
+ * to FW
+ */
+ uint8_t dcopy; /* Data copy required */
+ uint8_t reserved1;
+ uint16_t reserved2;
+ struct completion cmplobj; /* ioreq completion object */
+} ____cacheline_aligned_in_smp;
+
+/*
+ * Egress status page for egress cidx updates
+ */
+struct csio_qstatus_page {
+ __be32 qid;
+ __be16 cidx;
+ __be16 pidx;
+};
+
+
+enum {
+ CSIO_MAX_FLBUF_PER_IQWR = 4,
+ CSIO_QCREDIT_SZ = 64, /* pidx/cidx increments
+ * in bytes
+ */
+ CSIO_MAX_QID = 0xFFFF,
+ CSIO_MAX_IQ = 128,
+
+ CSIO_SGE_NTIMERS = 6,
+ CSIO_SGE_NCOUNTERS = 4,
+ CSIO_SGE_FL_SIZE_REGS = 16,
+};
+
+/* Defines for type */
+enum {
+ CSIO_EGRESS = 1,
+ CSIO_INGRESS = 2,
+ CSIO_FREELIST = 3,
+};
+
+/*
+ * Structure for footer (last 2 flits) of Ingress Queue Entry.
+ */
+struct csio_iqwr_footer {
+ __be32 hdrbuflen_pidx;
+ __be32 pldbuflen_qid;
+ union {
+ u8 type_gen;
+ __be64 last_flit;
+ } u;
+};
+
+#define IQWRF_NEWBUF (1 << 31)
+#define IQWRF_LEN_GET(x) (((x) >> 0) & 0x7fffffffU)
+#define IQWRF_GEN_SHIFT 7
+#define IQWRF_TYPE_GET(x) (((x) >> 4) & 0x3U)
+
+
+/*
+ * WR pair:
+ * ========
+ * A WR can start towards the end of a queue, and then continue at the
+ * beginning, since the queue is considered to be circular. This will
+ * require a pair of address/len to be passed back to the caller -
+ * hence the Work request pair structure.
+ */
+struct csio_wr_pair {
+ void *addr1;
+ uint32_t size1;
+ void *addr2;
+ uint32_t size2;
+};
+
+/*
+ * The following structure is used by ingress processing to return the
+ * free list buffers to consumers.
+ */
+struct csio_fl_dma_buf {
+ struct csio_dma_buf flbufs[CSIO_MAX_FLBUF_PER_IQWR];
+ /* Freelist DMA buffers */
+ int offset; /* Offset within the
+ * first FL buf.
+ */
+ uint32_t totlen; /* Total length */
+ uint8_t defer_free; /* Free of buffer can
+ * deferred
+ */
+};
+
+/* Data-types */
+typedef void (*iq_handler_t)(struct csio_hw *, void *, uint32_t,
+ struct csio_fl_dma_buf *, void *);
+
+struct csio_iq {
+ uint16_t iqid; /* Queue ID */
+ uint16_t physiqid; /* Physical Queue ID */
+ uint16_t genbit; /* Generation bit,
+ * initially set to 1
+ */
+ int flq_idx; /* Freelist queue index */
+ iq_handler_t iq_intx_handler; /* IQ INTx handler routine */
+};
+
+struct csio_eq {
+ uint16_t eqid; /* Qid */
+ uint16_t physeqid; /* Physical Queue ID */
+};
+
+struct csio_fl {
+ uint16_t flid; /* Qid */
+ uint16_t packen; /* Packing enabled? */
+ int offset; /* Offset within FL buf */
+ int sreg; /* Size register */
+ struct csio_dma_buf *bufs; /* Free list buffer ptr array
+ * indexed using flq->cidx/pidx
+ */
+};
+
+struct csio_qstats {
+ uint32_t n_tot_reqs; /* Total no. of Requests */
+ uint32_t n_tot_rsps; /* Total no. of responses */
+ uint32_t n_qwrap; /* Queue wraps */
+ uint32_t n_eq_wr_split; /* Number of split EQ WRs */
+ uint32_t n_qentry; /* Queue entry */
+ uint32_t n_qempty; /* Queue empty */
+ uint32_t n_qfull; /* Queue fulls */
+ uint32_t n_rsp_unknown; /* Unknown response type */
+ uint32_t n_stray_comp; /* Stray completion intr */
+ uint32_t n_flq_refill; /* Number of FL refills */
+};
+
+/* Queue metadata */
+struct csio_q {
+ uint16_t type; /* Type: Ingress/Egress/FL */
+ uint16_t pidx; /* producer index */
+ uint16_t cidx; /* consumer index */
+ uint16_t inc_idx; /* Incremental index */
+ uint32_t wr_sz; /* Size of all WRs in this q
+ * if fixed
+ */
+ void *vstart; /* Base virtual address
+ * of queue
+ */
+ void *vwrap; /* Virtual end address to
+ * wrap around at
+ */
+ uint32_t credits; /* Size of queue in credits */
+ void *owner; /* Owner */
+ union { /* Queue contexts */
+ struct csio_iq iq;
+ struct csio_eq eq;
+ struct csio_fl fl;
+ } un;
+
+ dma_addr_t pstart; /* Base physical address of
+ * queue
+ */
+ uint32_t portid; /* PCIE Channel */
+ uint32_t size; /* Size of queue in bytes */
+ struct csio_qstats stats; /* Statistics */
+} ____cacheline_aligned_in_smp;
+
+struct csio_sge {
+ uint32_t csio_fl_align; /* Calculated and cached
+ * for fast path
+ */
+ uint32_t sge_control; /* padding, boundaries,
+ * lengths, etc.
+ */
+ uint32_t sge_host_page_size; /* Host page size */
+ uint32_t sge_fl_buf_size[CSIO_SGE_FL_SIZE_REGS];
+ /* free list buffer sizes */
+ uint16_t timer_val[CSIO_SGE_NTIMERS];
+ uint8_t counter_val[CSIO_SGE_NCOUNTERS];
+};
+
+/* Work request module */
+struct csio_wrm {
+ int num_q; /* Number of queues */
+ struct csio_q **q_arr; /* Array of queue pointers
+ * allocated dynamically
+ * based on configured values
+ */
+ uint32_t fw_iq_start; /* Start ID of IQ for this fn*/
+ uint32_t fw_eq_start; /* Start ID of EQ for this fn*/
+ struct csio_q *intr_map[CSIO_MAX_IQ];
+ /* IQ-id to IQ map table. */
+ int free_qidx; /* queue idx of free queue */
+ struct csio_sge sge; /* SGE params */
+};
+
+#define csio_get_q(__hw, __idx) ((__hw)->wrm.q_arr[__idx])
+#define csio_q_type(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->type)
+#define csio_q_pidx(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->pidx)
+#define csio_q_cidx(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->cidx)
+#define csio_q_inc_idx(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->inc_idx)
+#define csio_q_vstart(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->vstart)
+#define csio_q_pstart(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->pstart)
+#define csio_q_size(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->size)
+#define csio_q_credits(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->credits)
+#define csio_q_portid(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->portid)
+#define csio_q_wr_sz(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->wr_sz)
+#define csio_q_iqid(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->un.iq.iqid)
+#define csio_q_physiqid(__hw, __idx) \
+ ((__hw)->wrm.q_arr[(__idx)]->un.iq.physiqid)
+#define csio_q_iq_flq_idx(__hw, __idx) \
+ ((__hw)->wrm.q_arr[(__idx)]->un.iq.flq_idx)
+#define csio_q_eqid(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->un.eq.eqid)
+#define csio_q_flid(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->un.fl.flid)
+
+#define csio_q_physeqid(__hw, __idx) \
+ ((__hw)->wrm.q_arr[(__idx)]->un.eq.physeqid)
+#define csio_iq_has_fl(__iq) ((__iq)->un.iq.flq_idx != -1)
+#define csio_q_owner(__hw, __idx) ((__hw)->wrm.q_arr[(__idx)]->owner)
+
+#define csio_q_iq_to_flid(__hw, __iq_idx) \
+ csio_q_flid((__hw), (__hw)->wrm.q_arr[(__iq_qidx)]->un.iq.flq_idx)
+#define csio_q_set_intr_map(__hw, __iq_idx, __rel_iq_id) \
+ (__hw)->wrm.intr_map[__rel_iq_id] = csio_get_q(__hw, __iq_idx)
+
+struct csio_mb;
+
+int csio_wr_alloc_q(struct csio_hw *, uint32_t, uint32_t,
+ uint16_t, void *, uint32_t, int, iq_handler_t);
+csio_retval_t csio_wr_iq_create(struct csio_hw *, void *, int,
+ uint32_t, uint8_t, bool,
+ void (*)(struct csio_hw *, struct csio_mb *));
+csio_retval_t csio_wr_eq_create(struct csio_hw *, void *, int, int, uint8_t,
+ void (*)(struct csio_hw *, struct csio_mb *));
+csio_retval_t csio_wr_destroy_queues(struct csio_hw *, bool cmd);
+
+
+csio_retval_t csio_wr_get(struct csio_hw *, int, uint32_t,
+ struct csio_wr_pair *);
+void csio_wr_copy_to_wrp(void *, struct csio_wr_pair *, uint32_t, uint32_t);
+csio_retval_t csio_wr_issue(struct csio_hw *, int, bool);
+csio_retval_t csio_wr_process_iq(struct csio_hw *, struct csio_q *,
+ void (*)(struct csio_hw *, void *,
+ uint32_t, struct csio_fl_dma_buf *,
+ void *),
+ void *);
+csio_retval_t csio_wr_process_iq_idx(struct csio_hw *, int,
+ void (*)(struct csio_hw *, void *,
+ uint32_t, struct csio_fl_dma_buf *,
+ void *),
+ void *);
+
+void csio_wr_sge_init(struct csio_hw *);
+csio_retval_t csio_wrm_init(struct csio_wrm *, struct csio_hw *);
+void csio_wrm_exit(struct csio_wrm *, struct csio_hw *);
+
+#endif /* ifndef __CSIO_WR_H__ */
diff --git a/drivers/scsi/csiostor/t4fw_api_stor.h b/drivers/scsi/csiostor/t4fw_api_stor.h
new file mode 100644
index 0000000..b96903a
--- /dev/null
+++ b/drivers/scsi/csiostor/t4fw_api_stor.h
@@ -0,0 +1,578 @@
+/*
+ * This file is part of the Chelsio FCoE driver for Linux.
+ *
+ * Copyright (c) 2009-2010 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _T4FW_API_STOR_H_
+#define _T4FW_API_STOR_H_
+
+
+/******************************************************************************
+ * R E T U R N V A L U E S
+ ********************************/
+
+enum fw_retval {
+ FW_SUCCESS = 0, /* completed sucessfully */
+ FW_EPERM = 1, /* operation not permitted */
+ FW_ENOENT = 2, /* no such file or directory */
+ FW_EIO = 5, /* input/output error; hw bad */
+ FW_ENOEXEC = 8, /* exec format error; inv microcode */
+ FW_EAGAIN = 11, /* try again */
+ FW_ENOMEM = 12, /* out of memory */
+ FW_EFAULT = 14, /* bad address; fw bad */
+ FW_EBUSY = 16, /* resource busy */
+ FW_EEXIST = 17, /* file exists */
+ FW_EINVAL = 22, /* invalid argument */
+ FW_ENOSPC = 28, /* no space left on device */
+ FW_ENOSYS = 38, /* functionality not implemented */
+ FW_EPROTO = 71, /* protocol error */
+ FW_EADDRINUSE = 98, /* address already in use */
+ FW_EADDRNOTAVAIL = 99, /* cannot assigned requested address */
+ FW_ENETDOWN = 100, /* network is down */
+ FW_ENETUNREACH = 101, /* network is unreachable */
+ FW_ENOBUFS = 105, /* no buffer space available */
+ FW_ETIMEDOUT = 110, /* timeout */
+ FW_EINPROGRESS = 115, /* fw internal */
+ FW_SCSI_ABORT_REQUESTED = 128, /* */
+ FW_SCSI_ABORT_TIMEDOUT = 129, /* */
+ FW_SCSI_ABORTED = 130, /* */
+ FW_SCSI_CLOSE_REQUESTED = 131, /* */
+ FW_ERR_LINK_DOWN = 132, /* */
+ FW_RDEV_NOT_READY = 133, /* */
+ FW_ERR_RDEV_LOST = 134, /* */
+ FW_ERR_RDEV_LOGO = 135, /* */
+ FW_FCOE_NO_XCHG = 136, /* */
+ FW_SCSI_RSP_ERR = 137, /* */
+ FW_ERR_RDEV_IMPL_LOGO = 138, /* */
+ FW_SCSI_UNDER_FLOW_ERR = 139, /* */
+ FW_SCSI_OVER_FLOW_ERR = 140, /* */
+ FW_SCSI_DDP_ERR = 141, /* DDP error*/
+ FW_SCSI_TASK_ERR = 142, /* No SCSI tasks available */
+};
+
+enum fw_fcoe_link_sub_op {
+ FCOE_LINK_DOWN = 0x0,
+ FCOE_LINK_UP = 0x1,
+ FCOE_LINK_COND = 0x2,
+};
+
+enum fw_fcoe_link_status {
+ FCOE_LINKDOWN = 0x0,
+ FCOE_LINKUP = 0x1,
+};
+
+enum fw_ofld_prot {
+ PROT_FCOE = 0x1,
+ PROT_ISCSI = 0x2,
+};
+
+enum rport_type_fcoe {
+ FLOGI_VFPORT = 0x1, /* 0xfffffe */
+ FDISC_VFPORT = 0x2, /* 0xfffffe */
+ NS_VNPORT = 0x3, /* 0xfffffc */
+ REG_FC4_VNPORT = 0x4, /* any FC4 type VN_PORT */
+ REG_VNPORT = 0x5, /* 0xfffxxx - non FC4 port in switch */
+ FDMI_VNPORT = 0x6, /* 0xfffffa */
+ FAB_CTLR_VNPORT = 0x7, /* 0xfffffd */
+};
+
+enum event_cause_fcoe {
+ PLOGI_ACC_RCVD = 0x01,
+ PLOGI_RJT_RCVD = 0x02,
+ PLOGI_RCVD = 0x03,
+ PLOGO_RCVD = 0x04,
+ PRLI_ACC_RCVD = 0x05,
+ PRLI_RJT_RCVD = 0x06,
+ PRLI_RCVD = 0x07,
+ PRLO_RCVD = 0x08,
+ NPORT_ID_CHGD = 0x09,
+ FLOGO_RCVD = 0x0a,
+ CLR_VIRT_LNK_RCVD = 0x0b,
+ FLOGI_ACC_RCVD = 0x0c,
+ FLOGI_RJT_RCVD = 0x0d,
+ FDISC_ACC_RCVD = 0x0e,
+ FDISC_RJT_RCVD = 0x0f,
+ FLOGI_TMO_MAX_RETRY = 0x10,
+ IMPL_LOGO_ADISC_ACC = 0x11,
+ IMPL_LOGO_ADISC_RJT = 0x12,
+ IMPL_LOGO_ADISC_CNFLT = 0x13,
+ PRLI_TMO = 0x14,
+ ADISC_TMO = 0x15,
+ RSCN_DEV_LOST = 0x16,
+ SCR_ACC_RCVD = 0x17,
+ ADISC_RJT_RCVD = 0x18,
+ LOGO_SNT = 0x19,
+ PROTO_ERR_IMPL_LOGO = 0x1a,
+};
+
+enum fcoe_cmn_type {
+ FCOE_ELS,
+ FCOE_CT,
+ FCOE_SCSI_CMD,
+ FCOE_UNSOL_ELS,
+};
+
+enum fw_wr_stor_opcodes {
+ FW_RDEV_WR = 0x38,
+ FW_FCOE_ELS_CT_WR = 0x30,
+ FW_SCSI_WRITE_WR = 0x31,
+ FW_SCSI_READ_WR = 0x32,
+ FW_SCSI_CMD_WR = 0x33,
+ FW_SCSI_ABRT_CLS_WR = 0x34,
+};
+
+struct fw_rdev_wr {
+ __be32 op_to_immdlen;
+ __be32 alloc_to_len16;
+ __be64 cookie;
+ u8 protocol;
+ u8 event_cause;
+ u8 cur_state;
+ u8 prev_state;
+ __be32 flags_to_assoc_flowid;
+ union rdev_entry {
+ struct fcoe_rdev_entry {
+ __be32 flowid;
+ u8 protocol;
+ u8 event_cause;
+ u8 flags;
+ u8 rjt_reason;
+ u8 cur_login_st;
+ u8 prev_login_st;
+ __be16 rcv_fr_sz;
+ u8 rd_xfer_rdy_to_rport_type;
+ u8 vft_to_qos;
+ u8 org_proc_assoc_to_acc_rsp_code;
+ u8 enh_disc_to_tgt;
+ u8 wwnn[8];
+ u8 wwpn[8];
+ __be16 iqid;
+ u8 fc_oui[3];
+ u8 r_id[3];
+ } fcoe_rdev;
+ struct iscsi_rdev_entry {
+ __be32 flowid;
+ u8 protocol;
+ u8 event_cause;
+ u8 flags;
+ u8 r3;
+ __be16 iscsi_opts;
+ __be16 tcp_opts;
+ __be16 ip_opts;
+ __be16 max_rcv_len;
+ __be16 max_snd_len;
+ __be16 first_brst_len;
+ __be16 max_brst_len;
+ __be16 r4;
+ __be16 def_time2wait;
+ __be16 def_time2ret;
+ __be16 nop_out_intrvl;
+ __be16 non_scsi_to;
+ __be16 isid;
+ __be16 tsid;
+ __be16 port;
+ __be16 tpgt;
+ u8 r5[6];
+ __be16 iqid;
+ } iscsi_rdev;
+ } u;
+};
+
+#define FW_RDEV_WR_FLOWID_GET(x) (((x) >> 8) & 0xfffff)
+#define FW_RDEV_WR_ASSOC_FLOWID_GET(x) (((x) >> 0) & 0xfffff)
+#define FW_RDEV_WR_RPORT_TYPE_GET(x) (((x) >> 0) & 0x1f)
+#define FW_RDEV_WR_NPIV_GET(x) (((x) >> 6) & 0x1)
+#define FW_RDEV_WR_CLASS_GET(x) (((x) >> 4) & 0x3)
+#define FW_RDEV_WR_TASK_RETRY_ID_GET(x) (((x) >> 5) & 0x1)
+#define FW_RDEV_WR_RETRY_GET(x) (((x) >> 4) & 0x1)
+#define FW_RDEV_WR_CONF_CMPL_GET(x) (((x) >> 3) & 0x1)
+#define FW_RDEV_WR_INI_GET(x) (((x) >> 1) & 0x1)
+#define FW_RDEV_WR_TGT_GET(x) (((x) >> 0) & 0x1)
+
+struct fw_fcoe_els_ct_wr {
+ __be32 op_immdlen;
+ __be32 flowid_len16;
+ __be64 cookie;
+ __be16 iqid;
+ u8 tmo_val;
+ u8 els_ct_type;
+ u8 ctl_pri;
+ u8 cp_en_class;
+ __be16 xfer_cnt;
+ u8 fl_to_sp;
+ u8 l_id[3];
+ u8 r5;
+ u8 r_id[3];
+ __be64 rsp_dmaaddr;
+ __be32 rsp_dmalen;
+ __be32 r6;
+};
+
+#define FW_FCOE_ELS_CT_WR_OPCODE(x) ((x) << 24)
+#define FW_FCOE_ELS_CT_WR_OPCODE_GET(x) (((x) >> 24) & 0xff)
+#define FW_FCOE_ELS_CT_WR_IMMDLEN(x) ((x) << 0)
+#define FW_FCOE_ELS_CT_WR_IMMDLEN_GET(x) (((x) >> 0) & 0xff)
+#define FW_FCOE_ELS_CT_WR_SP(x) ((x) << 0)
+
+struct fw_scsi_write_wr {
+ __be32 op_immdlen;
+ __be32 flowid_len16;
+ __be64 cookie;
+ __be16 iqid;
+ u8 tmo_val;
+ u8 use_xfer_cnt;
+ union fw_scsi_write_priv {
+ struct fcoe_write_priv {
+ u8 ctl_pri;
+ u8 cp_en_class;
+ u8 r3_lo[2];
+ } fcoe;
+ struct iscsi_write_priv {
+ u8 r3[4];
+ } iscsi;
+ } u;
+ __be32 xfer_cnt;
+ __be32 ini_xfer_cnt;
+ __be64 rsp_dmaaddr;
+ __be32 rsp_dmalen;
+ __be32 r4;
+};
+
+#define FW_SCSI_WRITE_WR_IMMDLEN(x) ((x) << 0)
+
+struct fw_scsi_read_wr {
+ __be32 op_immdlen;
+ __be32 flowid_len16;
+ __be64 cookie;
+ __be16 iqid;
+ u8 tmo_val;
+ u8 use_xfer_cnt;
+ union fw_scsi_read_priv {
+ struct fcoe_read_priv {
+ u8 ctl_pri;
+ u8 cp_en_class;
+ u8 r3_lo[2];
+ } fcoe;
+ struct iscsi_read_priv {
+ u8 r3[4];
+ } iscsi;
+ } u;
+ __be32 xfer_cnt;
+ __be32 ini_xfer_cnt;
+ __be64 rsp_dmaaddr;
+ __be32 rsp_dmalen;
+ __be32 r4;
+};
+
+#define FW_SCSI_READ_WR_IMMDLEN(x) ((x) << 0)
+
+struct fw_scsi_cmd_wr {
+ __be32 op_immdlen;
+ __be32 flowid_len16;
+ __be64 cookie;
+ __be16 iqid;
+ u8 tmo_val;
+ u8 r3;
+ union fw_scsi_cmd_priv {
+ struct fcoe_cmd_priv {
+ u8 ctl_pri;
+ u8 cp_en_class;
+ u8 r4_lo[2];
+ } fcoe;
+ struct iscsi_cmd_priv {
+ u8 r4[4];
+ } iscsi;
+ } u;
+ u8 r5[8];
+ __be64 rsp_dmaaddr;
+ __be32 rsp_dmalen;
+ __be32 r6;
+};
+
+#define FW_SCSI_CMD_WR_IMMDLEN(x) ((x) << 0)
+
+#define SCSI_ABORT 0
+#define SCSI_CLOSE 1
+
+struct fw_scsi_abrt_cls_wr {
+ __be32 op_immdlen;
+ __be32 flowid_len16;
+ __be64 cookie;
+ __be16 iqid;
+ u8 tmo_val;
+ u8 sub_opcode_to_chk_all_io;
+ u8 r3[4];
+ __be64 t_cookie;
+};
+
+#define FW_SCSI_ABRT_CLS_WR_SUB_OPCODE(x) ((x) << 2)
+#define FW_SCSI_ABRT_CLS_WR_SUB_OPCODE_GET(x) (((x) >> 2) & 0x3f)
+#define FW_SCSI_ABRT_CLS_WR_CHK_ALL_IO(x) ((x) << 0)
+
+enum fw_cmd_stor_opcodes {
+ FW_FCOE_RES_INFO_CMD = 0x31,
+ FW_FCOE_LINK_CMD = 0x32,
+ FW_FCOE_VNP_CMD = 0x33,
+ FW_FCOE_SPARAMS_CMD = 0x35,
+ FW_FCOE_STATS_CMD = 0x37,
+ FW_FCOE_FCF_CMD = 0x38,
+};
+
+struct fw_fcoe_res_info_cmd {
+ __be32 op_to_read;
+ __be32 retval_len16;
+ __be16 e_d_tov;
+ __be16 r_a_tov_seq;
+ __be16 r_a_tov_els;
+ __be16 r_r_tov;
+ __be32 max_xchgs;
+ __be32 max_ssns;
+ __be32 used_xchgs;
+ __be32 used_ssns;
+ __be32 max_fcfs;
+ __be32 max_vnps;
+ __be32 used_fcfs;
+ __be32 used_vnps;
+};
+
+struct fw_fcoe_link_cmd {
+ __be32 op_to_portid;
+ __be32 retval_len16;
+ __be32 sub_opcode_fcfi;
+ u8 r3;
+ u8 lstatus;
+ __be16 flags;
+ u8 r4;
+ u8 set_vlan;
+ __be16 vlan_id;
+ __be32 vnpi_pkd;
+ __be16 r6;
+ u8 phy_mac[6];
+ u8 vnport_wwnn[8];
+ u8 vnport_wwpn[8];
+};
+
+#define FW_FCOE_LINK_CMD_PORTID(x) ((x) << 0)
+#define FW_FCOE_LINK_CMD_PORTID_GET(x) (((x) >> 0) & 0xf)
+#define FW_FCOE_LINK_CMD_SUB_OPCODE(x) ((x) << 24U)
+#define FW_FCOE_LINK_CMD_FCFI(x) ((x) << 0)
+#define FW_FCOE_LINK_CMD_FCFI_GET(x) (((x) >> 0) & 0xffffff)
+#define FW_FCOE_LINK_CMD_VNPI_GET(x) (((x) >> 0) & 0xfffff)
+
+struct fw_fcoe_vnp_cmd {
+ __be32 op_to_fcfi;
+ __be32 alloc_to_len16;
+ __be32 gen_wwn_to_vnpi;
+ __be32 vf_id;
+ __be16 iqid;
+ u8 vnport_mac[6];
+ u8 vnport_wwnn[8];
+ u8 vnport_wwpn[8];
+ u8 cmn_srv_parms[16];
+ u8 clsp_word_0_1[8];
+};
+
+#define FW_FCOE_VNP_CMD_FCFI(x) ((x) << 0)
+#define FW_FCOE_VNP_CMD_ALLOC (1U << 31)
+#define FW_FCOE_VNP_CMD_FREE (1U << 30)
+#define FW_FCOE_VNP_CMD_MODIFY (1U << 29)
+#define FW_FCOE_VNP_CMD_GEN_WWN (1U << 22)
+#define FW_FCOE_VNP_CMD_VFID_EN (1U << 20)
+#define FW_FCOE_VNP_CMD_VNPI(x) ((x) << 0)
+#define FW_FCOE_VNP_CMD_VNPI_GET(x) (((x) >> 0) & 0xfffff)
+
+struct fw_fcoe_sparams_cmd {
+ __be32 op_to_portid;
+ __be32 retval_len16;
+ u8 r3[7];
+ u8 cos;
+ u8 lport_wwnn[8];
+ u8 lport_wwpn[8];
+ u8 cmn_srv_parms[16];
+ u8 cls_srv_parms[16];
+};
+
+#define FW_FCOE_SPARAMS_CMD_PORTID(x) ((x) << 0)
+
+struct fw_fcoe_stats_cmd {
+ __be32 op_to_flowid;
+ __be32 free_to_len16;
+ union fw_fcoe_stats {
+ struct fw_fcoe_stats_ctl {
+ u8 nstats_port;
+ u8 port_valid_ix;
+ __be16 r6;
+ __be32 r7;
+ __be64 stat0;
+ __be64 stat1;
+ __be64 stat2;
+ __be64 stat3;
+ __be64 stat4;
+ __be64 stat5;
+ } ctl;
+ struct fw_fcoe_port_stats {
+ __be64 tx_bcast_bytes;
+ __be64 tx_bcast_frames;
+ __be64 tx_mcast_bytes;
+ __be64 tx_mcast_frames;
+ __be64 tx_ucast_bytes;
+ __be64 tx_ucast_frames;
+ __be64 tx_drop_frames;
+ __be64 tx_offload_bytes;
+ __be64 tx_offload_frames;
+ __be64 rx_bcast_bytes;
+ __be64 rx_bcast_frames;
+ __be64 rx_mcast_bytes;
+ __be64 rx_mcast_frames;
+ __be64 rx_ucast_bytes;
+ __be64 rx_ucast_frames;
+ __be64 rx_err_frames;
+ } port_stats;
+ struct fw_fcoe_fcf_stats {
+ __be32 fip_tx_bytes;
+ __be32 fip_tx_fr;
+ __be64 fcf_ka;
+ __be64 mcast_adv_rcvd;
+ __be16 ucast_adv_rcvd;
+ __be16 sol_sent;
+ __be16 vlan_req;
+ __be16 vlan_rpl;
+ __be16 clr_vlink;
+ __be16 link_down;
+ __be16 link_up;
+ __be16 logo;
+ __be16 flogi_req;
+ __be16 flogi_rpl;
+ __be16 fdisc_req;
+ __be16 fdisc_rpl;
+ __be16 fka_prd_chg;
+ __be16 fc_map_chg;
+ __be16 vfid_chg;
+ u8 no_fka_req;
+ u8 no_vnp;
+ } fcf_stats;
+ struct fw_fcoe_pcb_stats {
+ __be64 tx_bytes;
+ __be64 tx_frames;
+ __be64 rx_bytes;
+ __be64 rx_frames;
+ __be32 vnp_ka;
+ __be32 unsol_els_rcvd;
+ __be64 unsol_cmd_rcvd;
+ __be16 implicit_logo;
+ __be16 flogi_inv_sparm;
+ __be16 fdisc_inv_sparm;
+ __be16 flogi_rjt;
+ __be16 fdisc_rjt;
+ __be16 no_ssn;
+ __be16 mac_flt_fail;
+ __be16 inv_fr_rcvd;
+ } pcb_stats;
+ struct fw_fcoe_scb_stats {
+ __be64 tx_bytes;
+ __be64 tx_frames;
+ __be64 rx_bytes;
+ __be64 rx_frames;
+ __be32 host_abrt_req;
+ __be32 adap_auto_abrt;
+ __be32 adap_abrt_rsp;
+ __be32 host_ios_req;
+ __be16 ssn_offl_ios;
+ __be16 ssn_not_rdy_ios;
+ u8 rx_data_ddp_err;
+ u8 ddp_flt_set_err;
+ __be16 rx_data_fr_err;
+ u8 bad_st_abrt_req;
+ u8 no_io_abrt_req;
+ u8 abort_tmo;
+ u8 abort_tmo_2;
+ __be32 abort_req;
+ u8 no_ppod_res_tmo;
+ u8 bp_tmo;
+ u8 adap_auto_cls;
+ u8 no_io_cls_req;
+ __be32 host_cls_req;
+ __be64 unsol_cmd_rcvd;
+ __be32 plogi_req_rcvd;
+ __be32 prli_req_rcvd;
+ __be16 logo_req_rcvd;
+ __be16 prlo_req_rcvd;
+ __be16 plogi_rjt_rcvd;
+ __be16 prli_rjt_rcvd;
+ __be32 adisc_req_rcvd;
+ __be32 rscn_rcvd;
+ __be32 rrq_req_rcvd;
+ __be32 unsol_els_rcvd;
+ u8 adisc_rjt_rcvd;
+ u8 scr_rjt;
+ u8 ct_rjt;
+ u8 inval_bls_rcvd;
+ __be32 ba_rjt_rcvd;
+ } scb_stats;
+ } u;
+};
+
+#define FW_FCOE_STATS_CMD_FLOWID(x) ((x) << 0)
+#define FW_FCOE_STATS_CMD_FREE (1U << 30)
+#define FW_FCOE_STATS_CMD_NSTATS(x) ((x) << 4)
+#define FW_FCOE_STATS_CMD_PORT(x) ((x) << 0)
+#define FW_FCOE_STATS_CMD_PORT_VALID (1U << 7)
+#define FW_FCOE_STATS_CMD_IX(x) ((x) << 0)
+
+struct fw_fcoe_fcf_cmd {
+ __be32 op_to_fcfi;
+ __be32 retval_len16;
+ __be16 priority_pkd;
+ u8 mac[6];
+ u8 name_id[8];
+ u8 fabric[8];
+ __be16 vf_id;
+ __be16 max_fcoe_size;
+ u8 vlan_id;
+ u8 fc_map[3];
+ __be32 fka_adv;
+ __be32 r6;
+ u8 r7_hi;
+ u8 fpma_to_portid;
+ u8 spma_mac[6];
+ __be64 r8;
+};
+
+#define FW_FCOE_FCF_CMD_FCFI(x) ((x) << 0)
+#define FW_FCOE_FCF_CMD_FCFI_GET(x) (((x) >> 0) & 0xfffff)
+#define FW_FCOE_FCF_CMD_PRIORITY_GET(x) (((x) >> 0) & 0xff)
+#define FW_FCOE_FCF_CMD_FPMA_GET(x) (((x) >> 6) & 0x1)
+#define FW_FCOE_FCF_CMD_SPMA_GET(x) (((x) >> 5) & 0x1)
+#define FW_FCOE_FCF_CMD_LOGIN_GET(x) (((x) >> 4) & 0x1)
+#define FW_FCOE_FCF_CMD_PORTID_GET(x) (((x) >> 0) & 0xf)
+
+#endif /* _T4FW_API_STOR_H_ */
--
1.7.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* [PATCH 8/8] cxgb4: Chelsio FCoE offload driver submission (cxgb4 common header updates).
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
` (6 preceding siblings ...)
2012-08-23 22:27 ` [PATCH 7/8] csiostor: Chelsio FCoE offload driver submission (headers part 2) Naresh Kumar Inna
@ 2012-08-23 22:27 ` Naresh Kumar Inna
2012-08-24 17:04 ` [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission David Miller
2012-08-24 21:45 ` Paul Gortmaker
9 siblings, 0 replies; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-23 22:27 UTC (permalink / raw)
To: JBottomley, linux-scsi, dm; +Cc: netdev, naresh, chethan
This patch contains updates to firmware/hardware header files shared
between csiostor and cxgb4, and the resulting changes to the cxgb4 files.
Signed-off-by: Naresh Kumar Inna <naresh@chelsio.com>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 2 +-
drivers/net/ethernet/chelsio/cxgb4/sge.c | 10 +-
drivers/net/ethernet/chelsio/cxgb4/t4_hw.c | 16 ++--
drivers/net/ethernet/chelsio/cxgb4/t4_msg.h | 1 +
drivers/net/ethernet/chelsio/cxgb4/t4_regs.h | 69 ++++++++++++++-
drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h | 104 +++++++++++++++++++---
6 files changed, 170 insertions(+), 32 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 5ed49af..b56d96c 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -3102,7 +3102,7 @@ static int adap_init1(struct adapter *adap, struct fw_caps_config_cmd *c)
memset(c, 0, sizeof(*c));
c->op_to_write = htonl(FW_CMD_OP(FW_CAPS_CONFIG_CMD) |
FW_CMD_REQUEST | FW_CMD_READ);
- c->retval_len16 = htonl(FW_LEN16(*c));
+ c->cfvalid_to_len16 = htonl(FW_LEN16(*c));
ret = t4_wr_mbox(adap, adap->fn, c, sizeof(*c), c);
if (ret < 0)
return ret;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/sge.c b/drivers/net/ethernet/chelsio/cxgb4/sge.c
index 8596aca..6109982 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/sge.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/sge.c
@@ -455,7 +455,7 @@ static inline void ring_fl_db(struct adapter *adap, struct sge_fl *q)
{
if (q->pend_cred >= 8) {
wmb();
- t4_write_reg(adap, MYPF_REG(SGE_PF_KDOORBELL), DBPRIO |
+ t4_write_reg(adap, MYPF_REG(SGE_PF_KDOORBELL), DBPRIO(1) |
QID(q->cntxt_id) | PIDX(q->pend_cred / 8));
q->pend_cred &= 7;
}
@@ -2020,10 +2020,10 @@ int t4_sge_alloc_rxq(struct adapter *adap, struct sge_rspq *iq, bool fwevtq,
goto fl_nomem;
flsz = fl->size / 8 + STAT_LEN / sizeof(struct tx_desc);
- c.iqns_to_fl0congen = htonl(FW_IQ_CMD_FL0PACKEN |
+ c.iqns_to_fl0congen = htonl(FW_IQ_CMD_FL0PACKEN(1) |
FW_IQ_CMD_FL0FETCHRO(1) |
FW_IQ_CMD_FL0DATARO(1) |
- FW_IQ_CMD_FL0PADEN);
+ FW_IQ_CMD_FL0PADEN(1));
c.fl0dcaen_to_fl0cidxfthresh = htons(FW_IQ_CMD_FL0FBMIN(2) |
FW_IQ_CMD_FL0FBMAX(3));
c.fl0size = htons(flsz);
@@ -2416,10 +2416,10 @@ void t4_sge_init(struct adapter *adap)
unsigned int fl_align_log = ilog2(FL_ALIGN);
t4_set_reg_field(adap, SGE_CONTROL, PKTSHIFT_MASK |
- INGPADBOUNDARY_MASK | EGRSTATUSPAGESIZE,
+ INGPADBOUNDARY_MASK | EGRSTATUSPAGESIZE(1),
INGPADBOUNDARY(fl_align_log - 5) | PKTSHIFT(2) |
RXPKTCPLMODE |
- (STAT_LEN == 128 ? EGRSTATUSPAGESIZE : 0));
+ (STAT_LEN == 128 ? EGRSTATUSPAGESIZE(1) : 0));
/*
* Set up to drop DOORBELL writes when the DOORBELL FIFO overflows
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
index fa947df..a943faa 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
@@ -456,12 +456,12 @@ static int sf1_read(struct adapter *adapter, unsigned int byte_cnt, int cont,
if (!byte_cnt || byte_cnt > 4)
return -EINVAL;
- if (t4_read_reg(adapter, SF_OP) & BUSY)
+ if (t4_read_reg(adapter, SF_OP) & SF_BUSY)
return -EBUSY;
cont = cont ? SF_CONT : 0;
lock = lock ? SF_LOCK : 0;
t4_write_reg(adapter, SF_OP, lock | cont | BYTECNT(byte_cnt - 1));
- ret = t4_wait_op_done(adapter, SF_OP, BUSY, 0, SF_ATTEMPTS, 5);
+ ret = t4_wait_op_done(adapter, SF_OP, SF_BUSY, 0, SF_ATTEMPTS, 5);
if (!ret)
*valp = t4_read_reg(adapter, SF_DATA);
return ret;
@@ -484,14 +484,14 @@ static int sf1_write(struct adapter *adapter, unsigned int byte_cnt, int cont,
{
if (!byte_cnt || byte_cnt > 4)
return -EINVAL;
- if (t4_read_reg(adapter, SF_OP) & BUSY)
+ if (t4_read_reg(adapter, SF_OP) & SF_BUSY)
return -EBUSY;
cont = cont ? SF_CONT : 0;
lock = lock ? SF_LOCK : 0;
t4_write_reg(adapter, SF_DATA, val);
t4_write_reg(adapter, SF_OP, lock |
cont | BYTECNT(byte_cnt - 1) | OP_WR);
- return t4_wait_op_done(adapter, SF_OP, BUSY, 0, SF_ATTEMPTS, 5);
+ return t4_wait_op_done(adapter, SF_OP, SF_BUSY, 0, SF_ATTEMPTS, 5);
}
/**
@@ -1972,14 +1972,14 @@ int t4_wol_pat_enable(struct adapter *adap, unsigned int port, unsigned int map,
t4_write_reg(adap, EPIO_REG(DATA0), mask0);
t4_write_reg(adap, EPIO_REG(OP), ADDRESS(i) | EPIOWR);
t4_read_reg(adap, EPIO_REG(OP)); /* flush */
- if (t4_read_reg(adap, EPIO_REG(OP)) & BUSY)
+ if (t4_read_reg(adap, EPIO_REG(OP)) & SF_BUSY)
return -ETIMEDOUT;
/* write CRC */
t4_write_reg(adap, EPIO_REG(DATA0), crc);
t4_write_reg(adap, EPIO_REG(OP), ADDRESS(i + 32) | EPIOWR);
t4_read_reg(adap, EPIO_REG(OP)); /* flush */
- if (t4_read_reg(adap, EPIO_REG(OP)) & BUSY)
+ if (t4_read_reg(adap, EPIO_REG(OP)) & SF_BUSY)
return -ETIMEDOUT;
}
#undef EPIO_REG
@@ -2118,7 +2118,7 @@ int t4_fw_hello(struct adapter *adap, unsigned int mbox, unsigned int evt_mbox,
struct fw_hello_cmd c;
INIT_CMD(c, HELLO, WRITE);
- c.err_to_mbasyncnot = htonl(
+ c.err_to_clearinit = htonl(
FW_HELLO_CMD_MASTERDIS(master == MASTER_CANT) |
FW_HELLO_CMD_MASTERFORCE(master == MASTER_MUST) |
FW_HELLO_CMD_MBMASTER(master == MASTER_MUST ? mbox : 0xff) |
@@ -2126,7 +2126,7 @@ int t4_fw_hello(struct adapter *adap, unsigned int mbox, unsigned int evt_mbox,
ret = t4_wr_mbox(adap, mbox, &c, sizeof(c), &c);
if (ret == 0 && state) {
- u32 v = ntohl(c.err_to_mbasyncnot);
+ u32 v = ntohl(c.err_to_clearinit);
if (v & FW_HELLO_CMD_INIT)
*state = DEV_STATE_INIT;
else if (v & FW_HELLO_CMD_ERR)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h b/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
index eb71b82..b760808 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
@@ -658,6 +658,7 @@ struct ulptx_sgl {
__be32 cmd_nsge;
#define ULPTX_CMD(x) ((x) << 24)
#define ULPTX_NSGE(x) ((x) << 0)
+#define ULPTX_MORE (1U << 23)
__be32 len0;
__be64 addr0;
struct ulptx_sge_pair sge[0];
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
index 111fc32..6bace75 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
@@ -67,7 +67,7 @@
#define QID_MASK 0xffff8000U
#define QID_SHIFT 15
#define QID(x) ((x) << QID_SHIFT)
-#define DBPRIO 0x00004000U
+#define DBPRIO(x) ((x) << 14)
#define PIDX_MASK 0x00003fffU
#define PIDX_SHIFT 0
#define PIDX(x) ((x) << PIDX_SHIFT)
@@ -89,7 +89,7 @@
#define SGE_CONTROL 0x1008
#define DCASYSTYPE 0x00080000U
#define RXPKTCPLMODE 0x00040000U
-#define EGRSTATUSPAGESIZE 0x00020000U
+#define EGRSTATUSPAGESIZE(x) ((x) << 17)
#define PKTSHIFT_MASK 0x00001c00U
#define PKTSHIFT_SHIFT 10
#define PKTSHIFT(x) ((x) << PKTSHIFT_SHIFT)
@@ -111,6 +111,13 @@
#define HOSTPAGESIZEPF0_MASK 0x0000000fU
#define HOSTPAGESIZEPF0_SHIFT 0
#define HOSTPAGESIZEPF0(x) ((x) << HOSTPAGESIZEPF0_SHIFT)
+#define HOSTPAGESIZEPF1(x) ((x) << 4)
+#define HOSTPAGESIZEPF2(x) ((x) << 8)
+#define HOSTPAGESIZEPF3(x) ((x) << 12)
+#define HOSTPAGESIZEPF4(x) ((x) << 16)
+#define HOSTPAGESIZEPF5(x) ((x) << 20)
+#define HOSTPAGESIZEPF6(x) ((x) << 24)
+#define HOSTPAGESIZEPF7(x) ((x) << 28)
#define SGE_EGRESS_QUEUES_PER_PAGE_PF 0x1010
#define QUEUESPERPAGEPF0_MASK 0x0000000fU
@@ -155,6 +162,14 @@
#define SGE_INT_ENABLE3 0x1040
#define SGE_FL_BUFFER_SIZE0 0x1044
#define SGE_FL_BUFFER_SIZE1 0x1048
+#define SGE_FL_BUFFER_SIZE2 0x104c
+#define SGE_FL_BUFFER_SIZE3 0x1050
+#define SGE_FL_BUFFER_SIZE4 0x1054
+#define SGE_FL_BUFFER_SIZE5 0x1058
+#define SGE_FL_BUFFER_SIZE6 0x105c
+#define SGE_FL_BUFFER_SIZE7 0x1060
+#define SGE_FL_BUFFER_SIZE8 0x1064
+
#define SGE_INGRESS_RX_THRESHOLD 0x10a0
#define THRESHOLD_0_MASK 0x3f000000U
#define THRESHOLD_0_SHIFT 24
@@ -173,6 +188,17 @@
#define THRESHOLD_3(x) ((x) << THRESHOLD_3_SHIFT)
#define THRESHOLD_3_GET(x) (((x) & THRESHOLD_3_MASK) >> THRESHOLD_3_SHIFT)
+#define SGE_DBFIFO_STATUS 0x10a4
+#define HP_INT_THRESH_SHIFT 28
+#define HP_INT_THRESH_MASK 0xfU
+#define HP_INT_THRESH(x) ((x) << HP_INT_THRESH_SHIFT)
+#define LP_INT_THRESH_SHIFT 12
+#define LP_INT_THRESH_MASK 0xfU
+#define LP_INT_THRESH(x) ((x) << LP_INT_THRESH_SHIFT)
+
+#define SGE_DOORBELL_CONTROL 0x10a8
+#define ENABLE_DROP (1 << 13)
+
#define SGE_TIMER_VALUE_0_AND_1 0x10b8
#define TIMERVALUE0_MASK 0xffff0000U
#define TIMERVALUE0_SHIFT 16
@@ -184,7 +210,25 @@
#define TIMERVALUE1_GET(x) (((x) & TIMERVALUE1_MASK) >> TIMERVALUE1_SHIFT)
#define SGE_TIMER_VALUE_2_AND_3 0x10bc
+#define TIMERVALUE2_MASK 0xffff0000U
+#define TIMERVALUE2_SHIFT 16
+#define TIMERVALUE2(x) ((x) << TIMERVALUE0_SHIFT)
+#define TIMERVALUE2_GET(x) (((x) & TIMERVALUE0_MASK) >> TIMERVALUE0_SHIFT)
+#define TIMERVALUE3_MASK 0x0000ffffU
+#define TIMERVALUE3_SHIFT 0
+#define TIMERVALUE3(x) ((x) << TIMERVALUE1_SHIFT)
+#define TIMERVALUE3_GET(x) (((x) & TIMERVALUE1_MASK) >> TIMERVALUE1_SHIFT)
+
#define SGE_TIMER_VALUE_4_AND_5 0x10c0
+#define TIMERVALUE4_MASK 0xffff0000U
+#define TIMERVALUE4_SHIFT 16
+#define TIMERVALUE4(x) ((x) << TIMERVALUE0_SHIFT)
+#define TIMERVALUE4_GET(x) (((x) & TIMERVALUE0_MASK) >> TIMERVALUE0_SHIFT)
+#define TIMERVALUE5_MASK 0x0000ffffU
+#define TIMERVALUE5_SHIFT 0
+#define TIMERVALUE5(x) ((x) << TIMERVALUE1_SHIFT)
+#define TIMERVALUE5_GET(x) (((x) & TIMERVALUE1_MASK) >> TIMERVALUE1_SHIFT)
+
#define SGE_DEBUG_INDEX 0x10cc
#define SGE_DEBUG_DATA_HIGH 0x10d0
#define SGE_DEBUG_DATA_LOW 0x10d4
@@ -243,6 +287,10 @@
#define M_HP_INT_THRESH 0xfU
#define M_LP_INT_THRESH 0xfU
+#define PCIE_PF_CFG 0x40
+#define AIVEC(x) ((x) << 4)
+#define AIVEC_MASK 0x3ffU
+
#define PCIE_PF_CLI 0x44
#define PCIE_INT_CAUSE 0x3004
#define UNXSPLCPLERR 0x20000000U
@@ -287,6 +335,15 @@
#define WINDOW(x) ((x) << WINDOW_SHIFT)
#define PCIE_MEM_ACCESS_OFFSET 0x306c
+#define PCIE_FW 0x30b8
+#define PCIE_FW_ERR 0x80000000U
+#define PCIE_FW_INIT 0x40000000U
+#define PCIE_FW_HALT 0x20000000U
+#define PCIE_FW_MASTER_VLD 0x00008000U
+#define PCIE_FW_MASTER(x) ((x) << 12)
+#define PCIE_FW_MASTER_MASK 0x7
+#define PCIE_FW_MASTER_GET(x) (((x) >> 12) & PCIE_FW_MASTER_MASK)
+
#define PCIE_CORE_UTL_SYSTEM_BUS_AGENT_STATUS 0x5908
#define RNPP 0x80000000U
#define RPCP 0x20000000U
@@ -384,6 +441,8 @@
#define EDC_1_BASE_ADDR 0x7980
#define CIM_BOOT_CFG 0x7b00
+#define UPCRST 0x00000001U
+
#define BOOTADDR_MASK 0xffffff00U
#define CIM_PF_MAILBOX_DATA 0x240
@@ -395,6 +454,9 @@
#define MBOWNER(x) ((x) << MBOWNER_SHIFT)
#define MBOWNER_GET(x) (((x) & MBOWNER_MASK) >> MBOWNER_SHIFT)
+#define CIM_PF_HOST_INT_ENABLE 0x288
+#define MBMSGRDYINTEN(x) ((x) << 19)
+
#define CIM_PF_HOST_INT_CAUSE 0x28c
#define MBMSGRDYINT 0x00080000U
@@ -825,7 +887,7 @@
#define SF_DATA 0x193f8
#define SF_OP 0x193fc
-#define BUSY 0x80000000U
+#define SF_BUSY 0x80000000U
#define SF_LOCK 0x00000010U
#define SF_CONT 0x00000008U
#define BYTECNT_MASK 0x00000006U
@@ -884,6 +946,7 @@
#define I2CM 0x00000002U
#define CIM 0x00000001U
+#define PL_INT_ENABLE 0x19410
#define PL_INT_MAP0 0x19414
#define PL_RST 0x19428
#define PIORST 0x00000002U
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h b/drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h
index ad53f79..034ca39 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4fw_api.h
@@ -68,6 +68,7 @@ struct fw_wr_hdr {
};
#define FW_WR_OP(x) ((x) << 24)
+#define FW_WR_OP_GET(x) (((x) >> 24) & 0xff)
#define FW_WR_ATOMIC(x) ((x) << 23)
#define FW_WR_FLUSH(x) ((x) << 22)
#define FW_WR_COMPL(x) ((x) << 21)
@@ -155,6 +156,9 @@ struct fw_eth_tx_pkt_vm_wr {
#define FW_CMD_MAX_TIMEOUT 3000
+#define FW_CMD_HELLO_TIMEOUT (3 * FW_CMD_MAX_TIMEOUT)
+#define FW_CMD_HELLO_RETRIES 3
+
enum fw_cmd_opcodes {
FW_LDST_CMD = 0x01,
FW_RESET_CMD = 0x03,
@@ -209,6 +213,7 @@ struct fw_cmd_hdr {
#define FW_CMD_OP(x) ((x) << 24)
#define FW_CMD_OP_GET(x) (((x) >> 24) & 0xff)
#define FW_CMD_REQUEST (1U << 23)
+#define FW_CMD_REQUEST_GET(x) (((x) >> 23) & 0x1)
#define FW_CMD_READ (1U << 22)
#define FW_CMD_WRITE (1U << 21)
#define FW_CMD_EXEC (1U << 20)
@@ -216,6 +221,7 @@ struct fw_cmd_hdr {
#define FW_CMD_RETVAL(x) ((x) << 8)
#define FW_CMD_RETVAL_GET(x) (((x) >> 8) & 0xff)
#define FW_CMD_LEN16(x) ((x) << 0)
+#define FW_LEN16(fw_struct) FW_CMD_LEN16(sizeof(fw_struct) / 16)
enum fw_ldst_addrspc {
FW_LDST_ADDRSPC_FIRMWARE = 0x0001,
@@ -228,7 +234,8 @@ enum fw_ldst_addrspc {
FW_LDST_ADDRSPC_TP_MIB = 0x0012,
FW_LDST_ADDRSPC_MDIO = 0x0018,
FW_LDST_ADDRSPC_MPS = 0x0020,
- FW_LDST_ADDRSPC_FUNC = 0x0028
+ FW_LDST_ADDRSPC_FUNC = 0x0028,
+ FW_LDST_ADDRSPC_FUNC_PCIE = 0x0029,
};
enum fw_ldst_mps_fid {
@@ -290,6 +297,16 @@ struct fw_ldst_cmd {
__be64 data0;
__be64 data1;
} func;
+ struct fw_ldst_pcie {
+ u8 ctrl_to_fn;
+ u8 bnum;
+ u8 r;
+ u8 ext_r;
+ u8 select_naccess;
+ u8 pcie_fn;
+ __be16 nset_pkd;
+ __be32 data[12];
+ } pcie;
} u;
};
@@ -299,24 +316,42 @@ struct fw_ldst_cmd {
#define FW_LDST_CMD_FID(x) ((x) << 15)
#define FW_LDST_CMD_CTL(x) ((x) << 0)
#define FW_LDST_CMD_RPLCPF(x) ((x) << 0)
+#define FW_LDST_CMD_LC (1U << 4)
+#define FW_LDST_CMD_NACCESS(x) ((x) << 0)
+#define FW_LDST_CMD_FN(x) ((x) << 0)
struct fw_reset_cmd {
__be32 op_to_write;
__be32 retval_len16;
__be32 val;
- __be32 r3;
+ __be32 halt_pkd;
+};
+
+#define FW_RESET_CMD_HALT (1U << 31)
+
+enum {
+ FW_HELLO_CMD_STAGE_OS = 0,
+ FW_HELLO_CMD_STAGE_PREOS0 = 1,
+ FW_HELLO_CMD_STAGE_PREOS1 = 2,
+ FW_HELLO_CMD_STAGE_POSTOS = 3,
};
struct fw_hello_cmd {
__be32 op_to_write;
__be32 retval_len16;
- __be32 err_to_mbasyncnot;
-#define FW_HELLO_CMD_ERR (1U << 31)
-#define FW_HELLO_CMD_INIT (1U << 30)
-#define FW_HELLO_CMD_MASTERDIS(x) ((x) << 29)
-#define FW_HELLO_CMD_MASTERFORCE(x) ((x) << 28)
-#define FW_HELLO_CMD_MBMASTER(x) ((x) << 24)
-#define FW_HELLO_CMD_MBASYNCNOT(x) ((x) << 20)
+ __be32 err_to_clearinit;
+#define FW_HELLO_CMD_ERR (1U << 31)
+#define FW_HELLO_CMD_INIT (1U << 30)
+#define FW_HELLO_CMD_MASTERDIS(x) ((x) << 29)
+#define FW_HELLO_CMD_MASTERFORCE(x) ((x) << 28)
+#define FW_HELLO_CMD_MBMASTER_MASK 0xf
+#define FW_HELLO_CMD_MBMASTER(x) ((x) << 24)
+#define FW_HELLO_CMD_MBMASTER_GET(x) \
+ (((x) >> 24) & FW_HELLO_CMD_MBMASTER_MASK)
+#define FW_HELLO_CMD_MBASYNCNOTINT(x) ((x) << 23)
+#define FW_HELLO_CMD_MBASYNCNOT(x) ((x) << 20)
+#define FW_HELLO_CMD_STAGE(x) ((x) << 17)
+#define FW_HELLO_CMD_CLEARINIT (1U << 16)
__be32 fwrev;
};
@@ -399,11 +434,20 @@ enum fw_caps_config_iscsi {
enum fw_caps_config_fcoe {
FW_CAPS_CONFIG_FCOE_INITIATOR = 0x00000001,
FW_CAPS_CONFIG_FCOE_TARGET = 0x00000002,
+ FW_CAPS_CONFIG_FCOE_CTRL_OFLD = 0x00000004,
+};
+
+enum fw_memtype_cf {
+ FW_MEMTYPE_CF_EDC0 = 0x0,
+ FW_MEMTYPE_CF_EDC1 = 0x1,
+ FW_MEMTYPE_CF_EXTMEM = 0x2,
+ FW_MEMTYPE_CF_FLASH = 0x4,
+ FW_MEMTYPE_CF_INTERNAL = 0x5,
};
struct fw_caps_config_cmd {
__be32 op_to_write;
- __be32 retval_len16;
+ __be32 cfvalid_to_len16;
__be32 r2;
__be32 hwmbitmap;
__be16 nbmcaps;
@@ -416,10 +460,15 @@ struct fw_caps_config_cmd {
__be16 r4;
__be16 iscsicaps;
__be16 fcoecaps;
- __be32 r5;
- __be64 r6;
+ __be32 cfcsum;
+ __be32 finiver;
+ __be32 finicsum;
};
+#define FW_CAPS_CONFIG_CMD_CFVALID (1U << 27)
+#define FW_CAPS_CONFIG_CMD_MEMTYPE_CF(x) ((x) << 24)
+#define FW_CAPS_CONFIG_CMD_MEMADDR64K_CF(x) ((x) << 16)
+
/*
* params command mnemonics
*/
@@ -451,6 +500,7 @@ enum fw_params_param_dev {
FW_PARAMS_PARAM_DEV_INTVER_FCOE = 0x0A,
FW_PARAMS_PARAM_DEV_FWREV = 0x0B,
FW_PARAMS_PARAM_DEV_TPREV = 0x0C,
+ FW_PARAMS_PARAM_DEV_CF = 0x0D,
};
/*
@@ -511,6 +561,9 @@ enum fw_params_param_dmaq {
#define FW_PARAMS_PARAM_Z(x) ((x) << 0)
#define FW_PARAMS_PARAM_XYZ(x) ((x) << 0)
#define FW_PARAMS_PARAM_YZ(x) ((x) << 0)
+#define FW_PARAMS_PARAM_X_GET(x) (((x) >> 16) & 0xff)
+#define FW_PARAMS_PARAM_Y_GET(x) (((x) >> 8) & 0xff)
+#define FW_PARAMS_PARAM_Z_GET(x) (((x) >> 0) & 0xff)
struct fw_params_cmd {
__be32 op_to_vfn;
@@ -648,8 +701,8 @@ struct fw_iq_cmd {
#define FW_IQ_CMD_FL0FETCHRO(x) ((x) << 6)
#define FW_IQ_CMD_FL0HOSTFCMODE(x) ((x) << 4)
#define FW_IQ_CMD_FL0CPRIO(x) ((x) << 3)
-#define FW_IQ_CMD_FL0PADEN (1U << 2)
-#define FW_IQ_CMD_FL0PACKEN (1U << 1)
+#define FW_IQ_CMD_FL0PADEN(x) ((x) << 2)
+#define FW_IQ_CMD_FL0PACKEN(x) ((x) << 1)
#define FW_IQ_CMD_FL0CONGEN (1U << 0)
#define FW_IQ_CMD_FL0DCAEN(x) ((x) << 15)
@@ -1137,6 +1190,14 @@ enum fw_port_dcb_cfg_rc {
FW_PORT_DCB_CFG_ERROR = 0x1
};
+enum fw_port_dcb_type {
+ FW_PORT_DCB_TYPE_PGID = 0x00,
+ FW_PORT_DCB_TYPE_PGRATE = 0x01,
+ FW_PORT_DCB_TYPE_PRIORATE = 0x02,
+ FW_PORT_DCB_TYPE_PFC = 0x03,
+ FW_PORT_DCB_TYPE_APP_ID = 0x04,
+};
+
struct fw_port_cmd {
__be32 op_to_portid;
__be32 action_to_len16;
@@ -1204,6 +1265,7 @@ struct fw_port_cmd {
#define FW_PORT_CMD_TXIPG(x) ((x) << 19)
#define FW_PORT_CMD_LSTATUS (1U << 31)
+#define FW_PORT_CMD_LSTATUS_GET(x) (((x) >> 31) & 0x1)
#define FW_PORT_CMD_LSPEED(x) ((x) << 24)
#define FW_PORT_CMD_LSPEED_GET(x) (((x) >> 24) & 0x3f)
#define FW_PORT_CMD_TXPAUSE (1U << 23)
@@ -1252,6 +1314,9 @@ enum fw_port_module_type {
FW_PORT_MOD_TYPE_TWINAX_PASSIVE,
FW_PORT_MOD_TYPE_TWINAX_ACTIVE,
FW_PORT_MOD_TYPE_LRM,
+ FW_PORT_MOD_TYPE_ERROR = FW_PORT_CMD_MODTYPE_MASK - 3,
+ FW_PORT_MOD_TYPE_UNKNOWN = FW_PORT_CMD_MODTYPE_MASK - 2,
+ FW_PORT_MOD_TYPE_NOTSUPPORTED = FW_PORT_CMD_MODTYPE_MASK - 1,
FW_PORT_MOD_TYPE_NONE = FW_PORT_CMD_MODTYPE_MASK
};
@@ -1613,7 +1678,11 @@ struct fw_hdr {
u8 intfver_iscsi;
u8 intfver_fcoe;
u8 reserved2;
- __be32 reserved3[27];
+ __u32 reserved3;
+ __u32 reserved4;
+ __u32 reserved5;
+ __be32 flags;
+ __be32 reserved6[23];
};
#define FW_HDR_FW_VER_MAJOR_GET(x) (((x) >> 24) & 0xff)
@@ -1621,6 +1690,11 @@ struct fw_hdr {
#define FW_HDR_FW_VER_MICRO_GET(x) (((x) >> 8) & 0xff)
#define FW_HDR_FW_VER_BUILD_GET(x) (((x) >> 0) & 0xff)
+enum fw_hdr_flags {
+ FW_HDR_FLAGS_RESET_HALT = 0x00000001,
+};
+
+
#define S_FW_CMD_OP 24
#define V_FW_CMD_OP(x) ((x) << S_FW_CMD_OP)
--
1.7.1
^ permalink raw reply related [flat|nested] 33+ messages in thread* Re: [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
` (7 preceding siblings ...)
2012-08-23 22:27 ` [PATCH 8/8] cxgb4: Chelsio FCoE offload driver submission (cxgb4 common header updates) Naresh Kumar Inna
@ 2012-08-24 17:04 ` David Miller
2012-08-24 19:04 ` Naresh Kumar Inna
2012-08-24 21:45 ` Paul Gortmaker
9 siblings, 1 reply; 33+ messages in thread
From: David Miller @ 2012-08-24 17:04 UTC (permalink / raw)
To: naresh; +Cc: JBottomley, linux-scsi, dm, netdev, chethan
From: Naresh Kumar Inna <naresh@chelsio.com>
Date: Fri, 24 Aug 2012 03:57:45 +0530
> This is the initial submission of the Chelsio FCoE offload driver (csiostor)
> to the upstream kernel. This driver currently supports FCoE offload
> functionality over Chelsio T4-based 10Gb Converged Network Adapters.
>
> The following patches contain the driver sources for csiostor driver and
> updates to firmware/hardware header files shared between csiostor and
> cxgb4 (Chelsio T4-based NIC driver). The csiostor driver is dependent on these
> header updates. These patches have been generated against scsi 'misc' branch.
>
> csiostor is a low level SCSI driver that interfaces with PCI, SCSI midlayer and
> FC transport subsystems. This driver claims the FCoE PCIe function on the
> Chelsio Converged Network Adapter. It relies on firmware events for slow path
> operations like discovery, thereby offloading session management. The driver
> programs firmware via Work Request interfaces for fast path I/O offload
> features.
You are going to have to get rid of these module parameters.
That have to do with things that are in no way specific to your device,
and therefore should be configured using generic kernel facilities.
Using driver specific module parameters results in a poor user
experience, because in order to make a configuration change the user
has to know exactly what kind of device and driver is underneath,
and then learn what the unique method is to make that configuration
change.
If you use a generic facility, the user only needs to learn one way to
make a configuration change, regardless of device type and driver.
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission
2012-08-24 17:04 ` [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission David Miller
@ 2012-08-24 19:04 ` Naresh Kumar Inna
2012-08-24 19:10 ` David Miller
0 siblings, 1 reply; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-24 19:04 UTC (permalink / raw)
To: David Miller
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/24/2012 10:34 PM, David Miller wrote:
> From: Naresh Kumar Inna <naresh@chelsio.com>
> Date: Fri, 24 Aug 2012 03:57:45 +0530
>
>> This is the initial submission of the Chelsio FCoE offload driver (csiostor)
>> to the upstream kernel. This driver currently supports FCoE offload
>> functionality over Chelsio T4-based 10Gb Converged Network Adapters.
>>
>> The following patches contain the driver sources for csiostor driver and
>> updates to firmware/hardware header files shared between csiostor and
>> cxgb4 (Chelsio T4-based NIC driver). The csiostor driver is dependent on these
>> header updates. These patches have been generated against scsi 'misc' branch.
>>
>> csiostor is a low level SCSI driver that interfaces with PCI, SCSI midlayer and
>> FC transport subsystems. This driver claims the FCoE PCIe function on the
>> Chelsio Converged Network Adapter. It relies on firmware events for slow path
>> operations like discovery, thereby offloading session management. The driver
>> programs firmware via Work Request interfaces for fast path I/O offload
>> features.
>
> You are going to have to get rid of these module parameters.
>
> That have to do with things that are in no way specific to your device,
> and therefore should be configured using generic kernel facilities.
>
> Using driver specific module parameters results in a poor user
> experience, because in order to make a configuration change the user
> has to know exactly what kind of device and driver is underneath,
> and then learn what the unique method is to make that configuration
> change.
>
> If you use a generic facility, the user only needs to learn one way to
> make a configuration change, regardless of device type and driver.
>
Hi Dave,
Thanks for reviewing. Is your comment with respect to any particular
module parameter[s] in this driver or all of them?
Regards,
Naresh.
^ permalink raw reply [flat|nested] 33+ messages in thread
* Re: [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission
2012-08-24 19:04 ` Naresh Kumar Inna
@ 2012-08-24 19:10 ` David Miller
2012-08-25 19:08 ` Naresh Kumar Inna
0 siblings, 1 reply; 33+ messages in thread
From: David Miller @ 2012-08-24 19:10 UTC (permalink / raw)
To: naresh; +Cc: JBottomley, linux-scsi, dm, netdev, chethan
From: Naresh Kumar Inna <naresh@chelsio.com>
Date: Sat, 25 Aug 2012 00:34:35 +0530
> Thanks for reviewing. Is your comment with respect to any particular
> module parameter[s] in this driver or all of them?
All of them.
^ permalink raw reply [flat|nested] 33+ messages in thread
* Re: [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission
2012-08-24 19:10 ` David Miller
@ 2012-08-25 19:08 ` Naresh Kumar Inna
2012-09-04 5:13 ` Naresh Kumar Inna
0 siblings, 1 reply; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-08-25 19:08 UTC (permalink / raw)
To: David Miller
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/25/2012 12:40 AM, David Miller wrote:
> From: Naresh Kumar Inna <naresh@chelsio.com>
> Date: Sat, 25 Aug 2012 00:34:35 +0530
>
>> Thanks for reviewing. Is your comment with respect to any particular
>> module parameter[s] in this driver or all of them?
>
> All of them.
>
So are module parameters being obsoleted? You mentioned about using
generic kernel facilities for tuning driver knobs. Could you please
elaborate a bit more?
Thanks,
Naresh.
^ permalink raw reply [flat|nested] 33+ messages in thread
* Re: [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission
2012-08-25 19:08 ` Naresh Kumar Inna
@ 2012-09-04 5:13 ` Naresh Kumar Inna
2012-09-04 5:34 ` David Miller
0 siblings, 1 reply; 33+ messages in thread
From: Naresh Kumar Inna @ 2012-09-04 5:13 UTC (permalink / raw)
To: David Miller
Cc: JBottomley@parallels.com, linux-scsi@vger.kernel.org,
Dimitrios Michailidis, netdev@vger.kernel.org, Chethan Seshadri
On 8/26/2012 12:38 AM, Naresh Kumar Inna wrote:
> On 8/25/2012 12:40 AM, David Miller wrote:
>> From: Naresh Kumar Inna <naresh@chelsio.com>
>> Date: Sat, 25 Aug 2012 00:34:35 +0530
>>
>>> Thanks for reviewing. Is your comment with respect to any particular
>>> module parameter[s] in this driver or all of them?
>>
>> All of them.
>>
>
> So are module parameters being obsoleted? You mentioned about using
> generic kernel facilities for tuning driver knobs. Could you please
> elaborate a bit more?
>
> Thanks,
> Naresh.
> --
Hi Dave,
Removing all the module parameters from the driver would put us at a
disadvantage with respect to tuning the driver and hardware. I have not
not been able to find alternatives, considering every other driver uses
module parameters. Could you provide pointers to the generic kernel
facilities you mentioned in your last post?
Thanks,
Naresh.
^ permalink raw reply [flat|nested] 33+ messages in thread
* Re: [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission
2012-09-04 5:13 ` Naresh Kumar Inna
@ 2012-09-04 5:34 ` David Miller
0 siblings, 0 replies; 33+ messages in thread
From: David Miller @ 2012-09-04 5:34 UTC (permalink / raw)
To: naresh; +Cc: JBottomley, linux-scsi, dm, netdev, chethan
From: Naresh Kumar Inna <naresh@chelsio.com>
Date: Tue, 4 Sep 2012 10:43:24 +0530
> Removing all the module parameters from the driver would put us at a
> disadvantage with respect to tuning the driver and hardware. I have not
> not been able to find alternatives, considering every other driver uses
> module parameters. Could you provide pointers to the generic kernel
> facilities you mentioned in your last post?
Adding module parameters is a cop-out. It's what you do if you want
to put zero effort into providing good interfaces for your driver.
If an appropriate generic mechanism doesn't exist yet, you simply add
one.
When people add crazy module parameter knobs to networking drivers, we
ask them to extend either ethtool or netlink, as appropriate.
^ permalink raw reply [flat|nested] 33+ messages in thread
* Re: [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission
2012-08-23 22:27 [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission Naresh Kumar Inna
` (8 preceding siblings ...)
2012-08-24 17:04 ` [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission David Miller
@ 2012-08-24 21:45 ` Paul Gortmaker
2012-08-24 21:58 ` James Bottomley
9 siblings, 1 reply; 33+ messages in thread
From: Paul Gortmaker @ 2012-08-24 21:45 UTC (permalink / raw)
To: Naresh Kumar Inna; +Cc: JBottomley, linux-scsi, dm, netdev, chethan
On Thu, Aug 23, 2012 at 6:27 PM, Naresh Kumar Inna <naresh@chelsio.com> wrote:
> This is the initial submission of the Chelsio FCoE offload driver (csiostor)
> to the upstream kernel. This driver currently supports FCoE offload
> functionality over Chelsio T4-based 10Gb Converged Network Adapters.
>
> The following patches contain the driver sources for csiostor driver and
> updates to firmware/hardware header files shared between csiostor and
> cxgb4 (Chelsio T4-based NIC driver). The csiostor driver is dependent on these
> header updates. These patches have been generated against scsi 'misc' branch.
>
> csiostor is a low level SCSI driver that interfaces with PCI, SCSI midlayer and
> FC transport subsystems. This driver claims the FCoE PCIe function on the
> Chelsio Converged Network Adapter. It relies on firmware events for slow path
> operations like discovery, thereby offloading session management. The driver
> programs firmware via Work Request interfaces for fast path I/O offload
> features.
>
> Here is the brief description of patches:
> [PATCH 1/8]: Hardware interface, Makefile and Kconfig changes.
> [PATCH 2/8]: Driver initialization and Work Request services.
> [PATCH 3/8]: FC transport interfaces and mailbox services.
> [PATCH 4/8]: Local and remote port state tracking functionality.
> [PATCH 5/8]: Interrupt handling and fast path I/O functionality.
> [PATCH 6/8]: Header files part 1.
> [PATCH 7/8]: Header files part 2.
Based on the above two, I'm guessing nothing will build and work
on any of steps one through six? Yet you expose the Kconfig
and Makefile linkage into the tree in patch #1? So your patches
as presented are non bisectable.
You need to rethink your breakup in the presentation. Factoring
things just by files alone is not the right approach. You need to
ask yourself whether each commit is a stand-alone entity that does
something independently on its own -- since they generally should.
> [PATCH 8/8]: Updates to header files shared between cxgb4 and csiostor.
>
> Naresh Kumar Inna (8):
> csiostor: Chelsio FCoE offload driver submission (sources part 1).
> csiostor: Chelsio FCoE offload driver submission (sources part 2).
> csiostor: Chelsio FCoE offload driver submission (sources part 3).
> csiostor: Chelsio FCoE offload driver submission (sources part 4).
> csiostor: Chelsio FCoE offload driver submission (sources part 5).
> csiostor: Chelsio FCoE offload driver submission (headers part 1).
> csiostor: Chelsio FCoE offload driver submission (headers part 2).
> cxgb4: Chelsio FCoE offload driver submission (cxgb4 common header
> updates).
Something is wrong here. You offer up semi-informational shortlogs
in your 1st list, but this shows the shortlogs you've used in the actual
commits are largely content-free.
The above are just what I'd happened to notice in reading the 0/8
since I was curious what it was for. I've not looked at the individual
driver code itself. It might be worthwhile to go over some of the
suggestions in Documentation/Submit* files though.
Paul.
^ permalink raw reply [flat|nested] 33+ messages in thread* Re: [PATCH 0/8] csiostor: Chelsio FCoE offload driver submission
2012-08-24 21:45 ` Paul Gortmaker
@ 2012-08-24 21:58 ` James Bottomley
0 siblings, 0 replies; 33+ messages in thread
From: James Bottomley @ 2012-08-24 21:58 UTC (permalink / raw)
To: Paul Gortmaker; +Cc: Naresh Kumar Inna, linux-scsi, dm, netdev, chethan
On Fri, 2012-08-24 at 17:45 -0400, Paul Gortmaker wrote:
> On Thu, Aug 23, 2012 at 6:27 PM, Naresh Kumar Inna <naresh@chelsio.com> wrote:
> > This is the initial submission of the Chelsio FCoE offload driver (csiostor)
> > to the upstream kernel. This driver currently supports FCoE offload
> > functionality over Chelsio T4-based 10Gb Converged Network Adapters.
> >
> > The following patches contain the driver sources for csiostor driver and
> > updates to firmware/hardware header files shared between csiostor and
> > cxgb4 (Chelsio T4-based NIC driver). The csiostor driver is dependent on these
> > header updates. These patches have been generated against scsi 'misc' branch.
> >
> > csiostor is a low level SCSI driver that interfaces with PCI, SCSI midlayer and
> > FC transport subsystems. This driver claims the FCoE PCIe function on the
> > Chelsio Converged Network Adapter. It relies on firmware events for slow path
> > operations like discovery, thereby offloading session management. The driver
> > programs firmware via Work Request interfaces for fast path I/O offload
> > features.
> >
> > Here is the brief description of patches:
> > [PATCH 1/8]: Hardware interface, Makefile and Kconfig changes.
> > [PATCH 2/8]: Driver initialization and Work Request services.
> > [PATCH 3/8]: FC transport interfaces and mailbox services.
> > [PATCH 4/8]: Local and remote port state tracking functionality.
> > [PATCH 5/8]: Interrupt handling and fast path I/O functionality.
> > [PATCH 6/8]: Header files part 1.
> > [PATCH 7/8]: Header files part 2.
>
> Based on the above two, I'm guessing nothing will build and work
> on any of steps one through six? Yet you expose the Kconfig
> and Makefile linkage into the tree in patch #1? So your patches
> as presented are non bisectable.
>
> You need to rethink your breakup in the presentation. Factoring
> things just by files alone is not the right approach. You need to
> ask yourself whether each commit is a stand-alone entity that does
> something independently on its own -- since they generally should.
Actually, this is fine for a new driver ... everything just gets put in
as a single commit. There's really no point adding a driver in pieces
if it's not functional until the last patch, so in this case only, the
breakup is to help the review.
James
^ permalink raw reply [flat|nested] 33+ messages in thread