* [PATCH 2/3] enic: add main netdev file with module infrastructure
@ 2008-09-09 3:57 Scott Feldman
2008-09-12 19:47 ` Ben Hutchings
0 siblings, 1 reply; 7+ messages in thread
From: Scott Feldman @ 2008-09-09 3:57 UTC (permalink / raw)
To: jgarzik, netdev
enic add main netdev file with module infrastructure
enic_main.c includes the module load and unload, PCI device probe, netdev
registration. Also includes stats and link notification.
Signed-off-by: Scott Feldman <scofeldm@cisco.com>
---
diff -Naurp -X linux-2.6.26.3/Documentation/dontdiff linux-2.6.26.3/drivers/net/enic/enic.h linux-2.6.26.3-enic/drivers/net/enic/enic.h
--- linux-2.6.26.3/drivers/net/enic/enic.h 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.26.3-enic/drivers/net/enic/enic.h 2008-09-08 15:16:22.000000000 -0700
@@ -0,0 +1,115 @@
+/*
+ * Copyright 2008 Cisco Systems, Inc. All rights reserved.
+ * Copyright 2007 Nuova Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * 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 _ENIC_H_
+#define _ENIC_H_
+
+#include <linux/inet_lro.h>
+
+#include "vnic_enet.h"
+#include "vnic_dev.h"
+#include "vnic_wq.h"
+#include "vnic_rq.h"
+#include "vnic_cq.h"
+#include "vnic_intr.h"
+#include "vnic_stats.h"
+#include "vnic_rss.h"
+
+#define DRV_NAME "enic"
+#define DRV_DESCRIPTION "Cisco 10G Ethernet Driver"
+#define DRV_VERSION "0.0.1.18163.472"
+#define DRV_COPYRIGHT "Copyright 2008 Cisco Systems, Inc"
+#define PFX DRV_NAME ": "
+
+#define ENIC_LRO_MAX_DESC 8
+#define ENIC_LRO_MAX_AGGR 64
+
+enum enic_cq_index {
+ ENIC_CQ_RQ,
+ ENIC_CQ_WQ,
+ ENIC_CQ_MAX,
+};
+
+enum enic_intx_intr_index {
+ ENIC_INTX_WQ_RQ,
+ ENIC_INTX_ERR,
+ ENIC_INTX_NOTIFY,
+ ENIC_INTX_MAX,
+};
+
+enum enic_msix_intr_index {
+ ENIC_MSIX_RQ,
+ ENIC_MSIX_WQ,
+ ENIC_MSIX_ERR,
+ ENIC_MSIX_NOTIFY,
+ ENIC_MSIX_MAX,
+};
+
+struct enic_msix_entry {
+ int requested;
+ char devname[IFNAMSIZ];
+ irqreturn_t (*isr)(int, void *);
+ void *devid;
+};
+
+/* Per-instance private data structure */
+struct enic {
+ struct net_device *netdev;
+ struct pci_dev *pdev;
+ struct vnic_enet_config config;
+ struct vnic_dev_bar bar0;
+ struct vnic_dev *vdev;
+ struct net_device_stats net_stats;
+ struct timer_list notify_timer;
+ struct work_struct reset;
+ struct msix_entry msix_entry[ENIC_MSIX_MAX];
+ struct enic_msix_entry msix[ENIC_MSIX_MAX];
+ u32 msg_enable;
+ spinlock_t devcmd_lock;
+ u8 mac_addr[ETH_ALEN];
+ u8 mc_addr[ENIC_MULTICAST_PERFECT_FILTERS][ETH_ALEN];
+ unsigned int mc_count;
+ int csum_rx_enabled;
+ u32 port_mtu;
+
+ /* work queue cache line section */
+ ____cacheline_aligned struct vnic_wq wq[1];
+ spinlock_t wq_lock[1];
+ unsigned int wq_count;
+ struct vlan_group *vlan_group;
+
+ /* receive queue cache line section */
+ ____cacheline_aligned struct vnic_rq rq[1];
+ unsigned int rq_count;
+ int (*rq_alloc_buf)(struct vnic_rq *rq);
+ struct napi_struct napi;
+ struct net_lro_mgr lro_mgr;
+ struct net_lro_desc lro_desc[ENIC_LRO_MAX_DESC];
+
+ /* interrupt resource cache line section */
+ ____cacheline_aligned struct vnic_intr intr[ENIC_MSIX_MAX];
+ unsigned int intr_count;
+ u32 __iomem *legacy_pba; /* memory-mapped */
+
+ /* completion queue cache line section */
+ ____cacheline_aligned struct vnic_cq cq[ENIC_CQ_MAX];
+ unsigned int cq_count;
+};
+
+#endif /* _ENIC_H_ */
diff -Naurp -X linux-2.6.26.3/Documentation/dontdiff linux-2.6.26.3/drivers/net/enic/enic_main.c linux-2.6.26.3-enic/drivers/net/enic/enic_main.c
--- linux-2.6.26.3/drivers/net/enic/enic_main.c 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.26.3-enic/drivers/net/enic/enic_main.c 2008-09-08 15:16:23.000000000 -0700
@@ -0,0 +1,1949 @@
+/*
+ * Copyright 2008 Cisco Systems, Inc. All rights reserved.
+ * Copyright 2007 Nuova Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * 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/module.h>
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <linux/init.h>
+#include <linux/workqueue.h>
+#include <linux/pci.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
+#include <linux/ethtool.h>
+#include <linux/in.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/tcp.h>
+
+#include "cq_enet_desc.h"
+#include "vnic_dev.h"
+#include "vnic_intr.h"
+#include "vnic_stats.h"
+#include "enic_res.h"
+#include "enic.h"
+
+#define ENIC_NOTIFY_TIMER_PERIOD (2 * HZ)
+#define ENIC_JUMBO_FIRST_BUF_SIZE 256
+
+/* Supported devices */
+static struct pci_device_id enic_id_table[] = {
+ { PCI_VDEVICE(CISCO, 0x0043) },
+ { 0, } /* end of table */
+};
+
+MODULE_DESCRIPTION(DRV_DESCRIPTION);
+MODULE_AUTHOR("Scott Feldman <scofeldm@cisco.com>");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRV_VERSION);
+MODULE_DEVICE_TABLE(pci, enic_id_table);
+
+struct enic_stat {
+ char name[ETH_GSTRING_LEN];
+ unsigned int offset;
+};
+
+#define ENIC_TX_STAT(stat) \
+ { .name = #stat, .offset = offsetof(struct vnic_tx_stats, stat) / 8 }
+#define ENIC_RX_STAT(stat) \
+ { .name = #stat, .offset = offsetof(struct vnic_rx_stats, stat) / 8 }
+
+static const struct enic_stat enic_tx_stats[] = {
+ ENIC_TX_STAT(tx_frames_ok),
+ ENIC_TX_STAT(tx_unicast_frames_ok),
+ ENIC_TX_STAT(tx_multicast_frames_ok),
+ ENIC_TX_STAT(tx_broadcast_frames_ok),
+ ENIC_TX_STAT(tx_bytes_ok),
+ ENIC_TX_STAT(tx_unicast_bytes_ok),
+ ENIC_TX_STAT(tx_multicast_bytes_ok),
+ ENIC_TX_STAT(tx_broadcast_bytes_ok),
+ ENIC_TX_STAT(tx_drops),
+ ENIC_TX_STAT(tx_errors),
+ ENIC_TX_STAT(tx_tso),
+};
+
+static const struct enic_stat enic_rx_stats[] = {
+ ENIC_RX_STAT(rx_frames_ok),
+ ENIC_RX_STAT(rx_frames_total),
+ ENIC_RX_STAT(rx_unicast_frames_ok),
+ ENIC_RX_STAT(rx_multicast_frames_ok),
+ ENIC_RX_STAT(rx_broadcast_frames_ok),
+ ENIC_RX_STAT(rx_bytes_ok),
+ ENIC_RX_STAT(rx_unicast_bytes_ok),
+ ENIC_RX_STAT(rx_multicast_bytes_ok),
+ ENIC_RX_STAT(rx_broadcast_bytes_ok),
+ ENIC_RX_STAT(rx_drop),
+ ENIC_RX_STAT(rx_no_bufs),
+ ENIC_RX_STAT(rx_errors),
+ ENIC_RX_STAT(rx_rss),
+ ENIC_RX_STAT(rx_crc_errors),
+ ENIC_RX_STAT(rx_frames_64),
+ ENIC_RX_STAT(rx_frames_127),
+ ENIC_RX_STAT(rx_frames_255),
+ ENIC_RX_STAT(rx_frames_511),
+ ENIC_RX_STAT(rx_frames_1023),
+ ENIC_RX_STAT(rx_frames_1518),
+ ENIC_RX_STAT(rx_frames_to_max),
+};
+
+static const unsigned int enic_n_tx_stats = ARRAY_SIZE(enic_tx_stats);
+static const unsigned int enic_n_rx_stats = ARRAY_SIZE(enic_rx_stats);
+
+static int enic_get_settings(struct net_device *netdev,
+ struct ethtool_cmd *ecmd)
+{
+ struct enic *enic = netdev_priv(netdev);
+
+ ecmd->supported = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
+ ecmd->advertising = (ADVERTISED_10000baseT_Full | ADVERTISED_FIBRE);
+ ecmd->port = PORT_FIBRE;
+ ecmd->transceiver = XCVR_EXTERNAL;
+
+ if (netif_carrier_ok(netdev)) {
+ ecmd->speed = vnic_dev_port_speed(enic->vdev);
+ ecmd->duplex = DUPLEX_FULL;
+ } else {
+ ecmd->speed = -1;
+ ecmd->duplex = -1;
+ }
+
+ ecmd->autoneg = AUTONEG_DISABLE;
+
+ return 0;
+}
+
+static void enic_get_drvinfo(struct net_device *netdev,
+ struct ethtool_drvinfo *drvinfo)
+{
+ struct enic *enic = netdev_priv(netdev);
+ struct vnic_devcmd_fw_info *fw_info;
+
+ spin_lock(&enic->devcmd_lock);
+ vnic_dev_fw_info(enic->vdev, &fw_info);
+ spin_unlock(&enic->devcmd_lock);
+
+ strncpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver));
+ strncpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version));
+ strncpy(drvinfo->fw_version, fw_info->fw_version,
+ sizeof(drvinfo->fw_version));
+ strncpy(drvinfo->bus_info, pci_name(enic->pdev),
+ sizeof(drvinfo->bus_info));
+}
+
+static void enic_get_strings(struct net_device *netdev, u32 stringset, u8 *data)
+{
+ unsigned int i;
+
+ switch (stringset) {
+ case ETH_SS_STATS:
+ for (i = 0; i < enic_n_tx_stats; i++) {
+ memcpy(data, enic_tx_stats[i].name, ETH_GSTRING_LEN);
+ data += ETH_GSTRING_LEN;
+ }
+ for (i = 0; i < enic_n_rx_stats; i++) {
+ memcpy(data, enic_rx_stats[i].name, ETH_GSTRING_LEN);
+ data += ETH_GSTRING_LEN;
+ }
+ break;
+ }
+}
+
+static int enic_get_stats_count(struct net_device *netdev)
+{
+ return enic_n_tx_stats + enic_n_rx_stats;
+}
+
+static void enic_get_ethtool_stats(struct net_device *netdev,
+ struct ethtool_stats *stats, u64 *data)
+{
+ struct enic *enic = netdev_priv(netdev);
+ struct vnic_stats *vstats;
+ unsigned int i;
+
+ spin_lock(&enic->devcmd_lock);
+ vnic_dev_stats_dump(enic->vdev, &vstats);
+ spin_unlock(&enic->devcmd_lock);
+
+ for (i = 0; i < enic_n_tx_stats; i++)
+ *(data++) = ((u64 *)&vstats->tx)[enic_tx_stats[i].offset];
+ for (i = 0; i < enic_n_rx_stats; i++)
+ *(data++) = ((u64 *)&vstats->rx)[enic_rx_stats[i].offset];
+}
+
+static u32 enic_get_rx_csum(struct net_device *netdev)
+{
+ struct enic *enic = netdev_priv(netdev);
+ return enic->csum_rx_enabled;
+}
+
+static int enic_set_rx_csum(struct net_device *netdev, u32 data)
+{
+ struct enic *enic = netdev_priv(netdev);
+
+ enic->csum_rx_enabled =
+ (data && ENIC_SETTING(enic, RXCSUM)) ? 1 : 0;
+
+ return 0;
+}
+
+static int enic_set_tx_csum(struct net_device *netdev, u32 data)
+{
+ struct enic *enic = netdev_priv(netdev);
+
+ if (data && ENIC_SETTING(enic, TXCSUM))
+ netdev->features |= NETIF_F_HW_CSUM;
+ else
+ netdev->features &= ~NETIF_F_HW_CSUM;
+
+ return 0;
+}
+
+static int enic_set_tso(struct net_device *netdev, u32 data)
+{
+ struct enic *enic = netdev_priv(netdev);
+
+ if (data && ENIC_SETTING(enic, TSO))
+ netdev->features |=
+ NETIF_F_TSO | NETIF_F_TSO6 | NETIF_F_TSO_ECN;
+ else
+ netdev->features &=
+ ~(NETIF_F_TSO | NETIF_F_TSO6 | NETIF_F_TSO_ECN);
+
+ return 0;
+}
+
+static u32 enic_get_msglevel(struct net_device *netdev)
+{
+ struct enic *enic = netdev_priv(netdev);
+ return enic->msg_enable;
+}
+
+static void enic_set_msglevel(struct net_device *netdev, u32 value)
+{
+ struct enic *enic = netdev_priv(netdev);
+ enic->msg_enable = value;
+}
+
+static struct ethtool_ops enic_ethtool_ops = {
+ .get_settings = enic_get_settings,
+ .get_drvinfo = enic_get_drvinfo,
+ .get_msglevel = enic_get_msglevel,
+ .set_msglevel = enic_set_msglevel,
+ .get_link = ethtool_op_get_link,
+ .get_strings = enic_get_strings,
+ .get_stats_count = enic_get_stats_count,
+ .get_ethtool_stats = enic_get_ethtool_stats,
+ .get_rx_csum = enic_get_rx_csum,
+ .set_rx_csum = enic_set_rx_csum,
+ .get_tx_csum = ethtool_op_get_tx_csum,
+ .set_tx_csum = enic_set_tx_csum,
+ .get_sg = ethtool_op_get_sg,
+ .set_sg = ethtool_op_set_sg,
+ .get_tso = ethtool_op_get_tso,
+ .set_tso = enic_set_tso,
+};
+
+static void enic_free_wq_buf(struct vnic_wq *wq, struct vnic_wq_buf *buf)
+{
+ struct enic *enic = vnic_dev_priv(wq->vdev);
+
+ if (buf->sop)
+ pci_unmap_single(enic->pdev, buf->dma_addr,
+ buf->len, PCI_DMA_TODEVICE);
+ else
+ pci_unmap_page(enic->pdev, buf->dma_addr,
+ buf->len, PCI_DMA_TODEVICE);
+
+ if (buf->os_buf)
+ dev_kfree_skb_any(buf->os_buf);
+}
+
+static void enic_wq_free_buf(struct vnic_wq *wq,
+ struct cq_desc *cq_desc, struct vnic_wq_buf *buf, void *opaque)
+{
+ enic_free_wq_buf(wq, buf);
+}
+
+static int enic_wq_service(struct vnic_dev *vdev, struct cq_desc *cq_desc,
+ u8 type, u16 q_number, u16 completed_index, void *opaque)
+{
+ struct enic *enic = vnic_dev_priv(vdev);
+
+ spin_lock(&enic->wq_lock[q_number]);
+
+ vnic_wq_service(&enic->wq[q_number], cq_desc,
+ completed_index, enic_wq_free_buf,
+ opaque);
+
+ if (netif_queue_stopped(enic->netdev) &&
+ vnic_wq_desc_avail(&enic->wq[q_number]) >= MAX_SKB_FRAGS + 1)
+ netif_wake_queue(enic->netdev);
+
+ spin_unlock(&enic->wq_lock[q_number]);
+
+ return 0;
+}
+
+static void enic_log_q_error(struct enic *enic)
+{
+ unsigned int i;
+ u32 error_status;
+
+ for (i = 0; i < enic->wq_count; i++) {
+ error_status = vnic_wq_error_status(&enic->wq[i]);
+ if (error_status)
+ printk(KERN_ERR PFX "%s: WQ[%d] error_status %d\n",
+ enic->netdev->name, i, error_status);
+ }
+
+ for (i = 0; i < enic->rq_count; i++) {
+ error_status = vnic_rq_error_status(&enic->rq[i]);
+ if (error_status)
+ printk(KERN_ERR PFX "%s: RQ[%d] error_status %d\n",
+ enic->netdev->name, i, error_status);
+ }
+}
+
+static void enic_link_check(struct enic *enic)
+{
+ int link_status = vnic_dev_link_status(enic->vdev);
+ int carrier_ok = netif_carrier_ok(enic->netdev);
+
+ if (link_status && !carrier_ok) {
+ printk(KERN_INFO PFX "%s: Link UP\n", enic->netdev->name);
+ netif_carrier_on(enic->netdev);
+ } else if (!link_status && carrier_ok) {
+ printk(KERN_INFO PFX "%s: Link DOWN\n", enic->netdev->name);
+ netif_carrier_off(enic->netdev);
+ }
+}
+
+static void enic_mtu_check(struct enic *enic)
+{
+ u32 mtu = vnic_dev_mtu(enic->vdev);
+
+ if (mtu != enic->port_mtu) {
+ if (mtu < enic->netdev->mtu)
+ printk(KERN_WARNING PFX
+ "%s: interface MTU (%d) set higher "
+ "than switch port MTU (%d)\n",
+ enic->netdev->name, enic->netdev->mtu, mtu);
+ enic->port_mtu = mtu;
+ }
+}
+
+static void enic_msglvl_check(struct enic *enic)
+{
+ u32 msg_enable = vnic_dev_msg_lvl(enic->vdev);
+
+ if (msg_enable != enic->msg_enable) {
+ printk(KERN_INFO PFX "%s: msg lvl changed from 0x%x to 0x%x\n",
+ enic->netdev->name, enic->msg_enable, msg_enable);
+ enic->msg_enable = msg_enable;
+ }
+}
+
+static void enic_notify_check(struct enic *enic)
+{
+ enic_msglvl_check(enic);
+ enic_mtu_check(enic);
+ enic_link_check(enic);
+}
+
+#define ENIC_TEST_INTR(pba, i) (pba & (1 << i))
+
+static irqreturn_t enic_isr_legacy(int irq, void *data)
+{
+ struct net_device *netdev = data;
+ struct enic *enic = netdev_priv(netdev);
+ u32 pba;
+
+ vnic_intr_mask(&enic->intr[ENIC_INTX_WQ_RQ]);
+
+ pba = vnic_intr_legacy_pba(enic->legacy_pba);
+ if (!pba) {
+ vnic_intr_unmask(&enic->intr[ENIC_INTX_WQ_RQ]);
+ return IRQ_NONE; /* not our interrupt */
+ }
+
+ if (ENIC_TEST_INTR(pba, ENIC_INTX_NOTIFY))
+ enic_notify_check(enic);
+
+ if (ENIC_TEST_INTR(pba, ENIC_INTX_ERR)) {
+ enic_log_q_error(enic);
+ /* schedule recovery from WQ/RQ error */
+ schedule_work(&enic->reset);
+ return IRQ_HANDLED;
+ }
+
+ if (ENIC_TEST_INTR(pba, ENIC_INTX_WQ_RQ)) {
+ if (netif_rx_schedule_prep(netdev, &enic->napi))
+ __netif_rx_schedule(netdev, &enic->napi);
+ } else {
+ vnic_intr_unmask(&enic->intr[ENIC_INTX_WQ_RQ]);
+ }
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t enic_isr_msi(int irq, void *data)
+{
+ struct enic *enic = data;
+
+ /* With MSI, there is no sharing of interrupts, so this is
+ * our interrupt and there is no need to ack it. The device
+ * is not providing per-vector masking, so the OS will not
+ * write to PCI config space to mask/unmask the interrupt.
+ * We're using mask_on_assertion for MSI, so the device
+ * automatically masks the interrupt when the interrupt is
+ * generated. Later, when exiting polling, the interrupt
+ * will be unmasked (see enic_poll).
+ *
+ * Also, the device uses the same PCIe Traffic Class (TC)
+ * for Memory Write data and MSI, so there are no ordering
+ * issues; the MSI will always arrive at the Root Complex
+ * _after_ corresponding Memory Writes (i.e. descriptor
+ * writes).
+ */
+
+ netif_rx_schedule(enic->netdev, &enic->napi);
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t enic_isr_msix_rq(int irq, void *data)
+{
+ struct enic *enic = data;
+
+ /* schedule NAPI polling for RQ cleanup */
+ netif_rx_schedule(enic->netdev, &enic->napi);
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t enic_isr_msix_wq(int irq, void *data)
+{
+ struct enic *enic = data;
+ unsigned int wq_work_to_do = -1; /* no limit */
+ unsigned int wq_work_done;
+
+ wq_work_done = vnic_cq_service(&enic->cq[ENIC_CQ_WQ],
+ wq_work_to_do, enic_wq_service, NULL);
+
+ vnic_intr_return_credits(&enic->intr[ENIC_MSIX_WQ],
+ wq_work_done,
+ 1 /* unmask intr */,
+ 1 /* reset intr timer */);
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t enic_isr_msix_err(int irq, void *data)
+{
+ struct enic *enic = data;
+
+ enic_log_q_error(enic);
+
+ /* schedule recovery from WQ/RQ error */
+ schedule_work(&enic->reset);
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t enic_isr_msix_notify(int irq, void *data)
+{
+ struct enic *enic = data;
+
+ enic_notify_check(enic);
+ vnic_intr_unmask(&enic->intr[ENIC_MSIX_NOTIFY]);
+
+ return IRQ_HANDLED;
+}
+
+static inline void enic_queue_wq_skb_cont(struct enic *enic,
+ struct vnic_wq *wq, struct sk_buff *skb,
+ unsigned int len_left)
+{
+ skb_frag_t *frag;
+
+ /* Queue additional data fragments */
+ for (frag = skb_shinfo(skb)->frags; len_left; frag++) {
+ len_left -= frag->size;
+ enic_queue_wq_desc_cont(wq, skb,
+ pci_map_page(enic->pdev, frag->page,
+ frag->page_offset, frag->size,
+ PCI_DMA_TODEVICE),
+ frag->size,
+ (len_left == 0)); /* EOP? */
+ }
+}
+
+static inline void enic_queue_wq_skb_vlan(struct enic *enic,
+ struct vnic_wq *wq, struct sk_buff *skb,
+ int vlan_tag_insert, unsigned int vlan_tag)
+{
+ unsigned int head_len = skb_headlen(skb);
+ unsigned int len_left = skb->len - head_len;
+ int eop = (len_left == 0);
+
+ /* Queue the main skb fragment */
+ enic_queue_wq_desc(wq, skb,
+ pci_map_single(enic->pdev, skb->data,
+ head_len, PCI_DMA_TODEVICE),
+ head_len,
+ vlan_tag_insert, vlan_tag,
+ eop);
+
+ if (!eop)
+ enic_queue_wq_skb_cont(enic, wq, skb, len_left);
+}
+
+static inline void enic_queue_wq_skb_csum_l4(struct enic *enic,
+ struct vnic_wq *wq, struct sk_buff *skb,
+ int vlan_tag_insert, unsigned int vlan_tag)
+{
+ unsigned int head_len = skb_headlen(skb);
+ unsigned int len_left = skb->len - head_len;
+ unsigned int hdr_len = skb_transport_offset(skb);
+ unsigned int csum_offset = hdr_len + skb->csum_offset;
+ int eop = (len_left == 0);
+
+ /* Queue the main skb fragment */
+ enic_queue_wq_desc_csum_l4(wq, skb,
+ pci_map_single(enic->pdev, skb->data,
+ head_len, PCI_DMA_TODEVICE),
+ head_len,
+ csum_offset,
+ hdr_len,
+ vlan_tag_insert, vlan_tag,
+ eop);
+
+ if (!eop)
+ enic_queue_wq_skb_cont(enic, wq, skb, len_left);
+}
+
+static inline void enic_queue_wq_skb_tso(struct enic *enic,
+ struct vnic_wq *wq, struct sk_buff *skb, unsigned int mss,
+ int vlan_tag_insert, unsigned int vlan_tag)
+{
+ unsigned int head_len = skb_headlen(skb);
+ unsigned int len_left = skb->len - head_len;
+ unsigned int hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
+ int eop = (len_left == 0);
+
+ /* Preload TCP csum field with IP pseudo hdr calculated
+ * with IP length set to zero. HW will later add in length
+ * to each TCP segment resulting from the TSO.
+ */
+
+ if (skb->protocol == __constant_htons(ETH_P_IP)) {
+ ip_hdr(skb)->check = 0;
+ tcp_hdr(skb)->check = ~csum_tcpudp_magic(ip_hdr(skb)->saddr,
+ ip_hdr(skb)->daddr, 0, IPPROTO_TCP, 0);
+ } else if (skb->protocol == __constant_htons(ETH_P_IPV6)) {
+ tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
+ &ipv6_hdr(skb)->daddr, 0, IPPROTO_TCP, 0);
+ }
+
+ /* Queue the main skb fragment */
+ enic_queue_wq_desc_tso(wq, skb,
+ pci_map_single(enic->pdev, skb->data,
+ head_len, PCI_DMA_TODEVICE),
+ head_len,
+ mss, hdr_len,
+ vlan_tag_insert, vlan_tag,
+ eop);
+
+ if (!eop)
+ enic_queue_wq_skb_cont(enic, wq, skb, len_left);
+}
+
+static inline void enic_queue_wq_skb(struct enic *enic,
+ struct vnic_wq *wq, struct sk_buff *skb)
+{
+ unsigned int mss = skb_shinfo(skb)->gso_size;
+ unsigned int vlan_tag = 0;
+ int vlan_tag_insert = 0;
+
+ if (enic->vlan_group && vlan_tx_tag_present(skb)) {
+ /* VLAN tag from trunking driver */
+ vlan_tag_insert = 1;
+ vlan_tag = vlan_tx_tag_get(skb);
+ }
+
+ if (mss)
+ enic_queue_wq_skb_tso(enic, wq, skb, mss,
+ vlan_tag_insert, vlan_tag);
+ else if (skb->ip_summed == CHECKSUM_PARTIAL)
+ enic_queue_wq_skb_csum_l4(enic, wq, skb,
+ vlan_tag_insert, vlan_tag);
+ else
+ enic_queue_wq_skb_vlan(enic, wq, skb,
+ vlan_tag_insert, vlan_tag);
+}
+
+/* netif_tx_lock held, process context with BHs disabled */
+static int enic_hard_start_xmit(struct sk_buff *skb, struct net_device *netdev)
+{
+ struct enic *enic = netdev_priv(netdev);
+ struct vnic_wq *wq = &enic->wq[0];
+ unsigned long flags;
+
+ if (skb->len <= 0) {
+ dev_kfree_skb(skb);
+ return NETDEV_TX_OK;
+ }
+
+ /* Non-TSO sends must fit within ENIC_NON_TSO_MAX_DESC descs,
+ * which is very likely. In the off chance it's going to take
+ * more than * ENIC_NON_TSO_MAX_DESC, linearize the skb.
+ */
+
+ if (skb_shinfo(skb)->gso_size == 0 &&
+ skb_shinfo(skb)->nr_frags + 1 > ENIC_NON_TSO_MAX_DESC &&
+ skb_linearize(skb)) {
+ dev_kfree_skb(skb);
+ return NETDEV_TX_OK;
+ }
+
+ spin_lock_irqsave(&enic->wq_lock[0], flags);
+
+ if (vnic_wq_desc_avail(wq) < skb_shinfo(skb)->nr_frags + 1) {
+ netif_stop_queue(netdev);
+ /* This is a hard error, log it */
+ printk(KERN_ERR PFX "%s: BUG! Tx ring full when "
+ "queue awake!\n", netdev->name);
+ spin_unlock_irqrestore(&enic->wq_lock[0], flags);
+ return NETDEV_TX_BUSY;
+ }
+
+ enic_queue_wq_skb(enic, wq, skb);
+
+ if (vnic_wq_desc_avail(wq) < MAX_SKB_FRAGS + 1)
+ netif_stop_queue(netdev);
+
+ netdev->trans_start = jiffies;
+
+ spin_unlock_irqrestore(&enic->wq_lock[0], flags);
+
+ return NETDEV_TX_OK;
+}
+
+/* dev_base_lock rwlock held, nominally process context */
+static struct net_device_stats *enic_get_stats(struct net_device *netdev)
+{
+ struct enic *enic = netdev_priv(netdev);
+ struct vnic_stats *stats;
+
+ spin_lock(&enic->devcmd_lock);
+ vnic_dev_stats_dump(enic->vdev, &stats);
+ spin_unlock(&enic->devcmd_lock);
+
+ enic->net_stats.tx_packets = stats->tx.tx_frames_ok;
+ enic->net_stats.tx_bytes = stats->tx.tx_bytes_ok;
+ enic->net_stats.tx_errors = stats->tx.tx_errors;
+ enic->net_stats.tx_dropped = stats->tx.tx_drops;
+
+ enic->net_stats.rx_packets = stats->rx.rx_frames_ok;
+ enic->net_stats.rx_bytes = stats->rx.rx_bytes_ok;
+ enic->net_stats.rx_errors = stats->rx.rx_errors;
+ enic->net_stats.multicast = stats->rx.rx_multicast_frames_ok;
+ enic->net_stats.rx_crc_errors = stats->rx.rx_crc_errors;
+ enic->net_stats.rx_dropped = stats->rx.rx_no_bufs;
+
+ return &enic->net_stats;
+}
+
+static void enic_reset_mcaddrs(struct enic *enic)
+{
+ enic->mc_count = 0;
+}
+
+static int enic_set_mac_addr(struct net_device *netdev, char *addr)
+{
+ if (!is_valid_ether_addr(addr))
+ return -EADDRNOTAVAIL;
+
+ memcpy(netdev->dev_addr, addr, netdev->addr_len);
+
+ return 0;
+}
+
+/* netif_tx_lock held, BHs disabled */
+static void enic_set_multicast_list(struct net_device *netdev)
+{
+ struct enic *enic = netdev_priv(netdev);
+ struct dev_mc_list *list = netdev->mc_list;
+ int directed = 1;
+ int multicast = (netdev->flags & IFF_MULTICAST) ? 1 : 0;
+ int broadcast = (netdev->flags & IFF_BROADCAST) ? 1 : 0;
+ int promisc = (netdev->flags & IFF_PROMISC) ? 1 : 0;
+ int allmulti = (netdev->flags & IFF_ALLMULTI) ||
+ (netdev->mc_count > ENIC_MULTICAST_PERFECT_FILTERS);
+ u8 mc_addr[ENIC_MULTICAST_PERFECT_FILTERS][ETH_ALEN];
+ unsigned int mc_count = netdev->mc_count;
+ unsigned int i, j;
+
+ if (mc_count > ENIC_MULTICAST_PERFECT_FILTERS)
+ mc_count = ENIC_MULTICAST_PERFECT_FILTERS;
+
+ spin_lock(&enic->devcmd_lock);
+
+ vnic_dev_packet_filter(enic->vdev, directed,
+ multicast, broadcast, promisc, allmulti);
+
+ /* Is there an easier way? Trying to minimize to
+ * calls to add/del multicast addrs. We keep the
+ * addrs from the last call in enic->mc_addr and
+ * look for changes to add/del.
+ */
+
+ for (i = 0; list && i < mc_count; i++) {
+ memcpy(mc_addr[i], list->dmi_addr, ETH_ALEN);
+ list = list->next;
+ }
+
+ for (i = 0; i < enic->mc_count; i++) {
+ for (j = 0; j < mc_count; j++)
+ if (compare_ether_addr(enic->mc_addr[i],
+ mc_addr[j]) == 0)
+ break;
+ if (j == mc_count)
+ enic_del_multicast_addr(enic, enic->mc_addr[i]);
+ }
+
+ for (i = 0; i < mc_count; i++) {
+ for (j = 0; j < enic->mc_count; j++)
+ if (compare_ether_addr(mc_addr[i],
+ enic->mc_addr[j]) == 0)
+ break;
+ if (j == enic->mc_count)
+ enic_add_multicast_addr(enic, mc_addr[i]);
+ }
+
+ /* Save the list to compare against next time
+ */
+
+ for (i = 0; i < mc_count; i++)
+ memcpy(enic->mc_addr[i], mc_addr[i], ETH_ALEN);
+
+ enic->mc_count = mc_count;
+
+ spin_unlock(&enic->devcmd_lock);
+}
+
+/* rtnl lock is held */
+static void enic_vlan_rx_register(struct net_device *netdev,
+ struct vlan_group *vlan_group)
+{
+ struct enic *enic = netdev_priv(netdev);
+ enic->vlan_group = vlan_group;
+}
+
+/* rtnl lock is held */
+static void enic_vlan_rx_add_vid(struct net_device *netdev, u16 vid)
+{
+ struct enic *enic = netdev_priv(netdev);
+
+ spin_lock(&enic->devcmd_lock);
+ enic_add_vlan(enic, vid);
+ spin_unlock(&enic->devcmd_lock);
+}
+
+/* rtnl lock is held */
+static void enic_vlan_rx_kill_vid(struct net_device *netdev, u16 vid)
+{
+ struct enic *enic = netdev_priv(netdev);
+
+ spin_lock(&enic->devcmd_lock);
+ enic_del_vlan(enic, vid);
+ spin_unlock(&enic->devcmd_lock);
+}
+
+/* netif_tx_lock held, BHs disabled */
+static void enic_tx_timeout(struct net_device *netdev)
+{
+ struct enic *enic = netdev_priv(netdev);
+ schedule_work(&enic->reset);
+}
+
+static void enic_free_rq_buf(struct vnic_rq *rq, struct vnic_rq_buf *buf)
+{
+ struct enic *enic = vnic_dev_priv(rq->vdev);
+
+ if (!buf->os_buf)
+ return;
+
+ pci_unmap_single(enic->pdev, buf->dma_addr,
+ buf->len, PCI_DMA_FROMDEVICE);
+ dev_kfree_skb_any(buf->os_buf);
+}
+
+static inline struct sk_buff *enic_rq_alloc_skb(unsigned int size)
+{
+ struct sk_buff *skb;
+
+ skb = dev_alloc_skb(size + NET_IP_ALIGN);
+
+ if (skb)
+ skb_reserve(skb, NET_IP_ALIGN);
+
+ return skb;
+}
+
+static int enic_rq_alloc_buf(struct vnic_rq *rq)
+{
+ struct enic *enic = vnic_dev_priv(rq->vdev);
+ struct sk_buff *skb;
+ unsigned int len = enic->netdev->mtu + ETH_HLEN;
+ unsigned int os_buf_index = 0;
+ dma_addr_t dma_addr;
+
+ skb = enic_rq_alloc_skb(len);
+ if (!skb)
+ return -ENOMEM;
+
+ dma_addr = pci_map_single(enic->pdev, skb->data,
+ len, PCI_DMA_FROMDEVICE);
+
+ enic_queue_rq_desc(rq, skb, os_buf_index,
+ dma_addr, len);
+
+ return 0;
+}
+
+static int enic_get_skb_header(struct sk_buff *skb, void **iphdr,
+ void **tcph, u64 *hdr_flags, void *priv)
+{
+ struct cq_enet_rq_desc *cq_desc = priv;
+ unsigned int ip_len;
+ struct iphdr *iph;
+
+ u8 type, color, eop, sop, ingress_port, vlan_stripped;
+ u8 fcoe, fcoe_sof, fcoe_fc_crc_ok, fcoe_enc_error, fcoe_eof;
+ u8 tcp_udp_csum_ok, udp, tcp, ipv4_csum_ok;
+ u8 ipv6, ipv4, ipv4_fragment, fcs_ok, rss_type, csum_not_calc;
+ u8 packet_error;
+ u16 q_number, completed_index, bytes_written, vlan, checksum;
+ u32 rss_hash;
+
+ cq_enet_rq_desc_dec(cq_desc,
+ &type, &color, &q_number, &completed_index,
+ &ingress_port, &fcoe, &eop, &sop, &rss_type,
+ &csum_not_calc, &rss_hash, &bytes_written,
+ &packet_error, &vlan_stripped, &vlan, &checksum,
+ &fcoe_sof, &fcoe_fc_crc_ok, &fcoe_enc_error,
+ &fcoe_eof, &tcp_udp_csum_ok, &udp, &tcp,
+ &ipv4_csum_ok, &ipv6, &ipv4, &ipv4_fragment,
+ &fcs_ok);
+
+ if (!(ipv4 && tcp && !ipv4_fragment))
+ return -1;
+
+ skb_reset_network_header(skb);
+ iph = ip_hdr(skb);
+
+ ip_len = ip_hdrlen(skb);
+ skb_set_transport_header(skb, ip_len);
+
+ /* check if ip header and tcp header are complete */
+ if (ntohs(iph->tot_len) < ip_len + tcp_hdrlen(skb))
+ return -1;
+
+ *hdr_flags = LRO_IPV4 | LRO_TCP;
+ *tcph = tcp_hdr(skb);
+ *iphdr = iph;
+
+ return 0;
+}
+
+static void enic_rq_indicate_buf(struct vnic_rq *rq,
+ struct cq_desc *cq_desc, struct vnic_rq_buf *buf,
+ int skipped, void *opaque)
+{
+ struct enic *enic = vnic_dev_priv(rq->vdev);
+ struct sk_buff *skb;
+
+ u8 type, color, eop, sop, ingress_port, vlan_stripped;
+ u8 fcoe, fcoe_sof, fcoe_fc_crc_ok, fcoe_enc_error, fcoe_eof;
+ u8 tcp_udp_csum_ok, udp, tcp, ipv4_csum_ok;
+ u8 ipv6, ipv4, ipv4_fragment, fcs_ok, rss_type, csum_not_calc;
+ u8 packet_error;
+ u16 q_number, completed_index, bytes_written, vlan, checksum;
+ u32 rss_hash;
+
+ if (skipped)
+ return;
+
+ skb = buf->os_buf;
+ prefetch(skb->data - NET_IP_ALIGN);
+ pci_unmap_single(enic->pdev, buf->dma_addr,
+ buf->len, PCI_DMA_FROMDEVICE);
+
+ cq_enet_rq_desc_dec((struct cq_enet_rq_desc *)cq_desc,
+ &type, &color, &q_number, &completed_index,
+ &ingress_port, &fcoe, &eop, &sop, &rss_type,
+ &csum_not_calc, &rss_hash, &bytes_written,
+ &packet_error, &vlan_stripped, &vlan, &checksum,
+ &fcoe_sof, &fcoe_fc_crc_ok, &fcoe_enc_error,
+ &fcoe_eof, &tcp_udp_csum_ok, &udp, &tcp,
+ &ipv4_csum_ok, &ipv6, &ipv4, &ipv4_fragment,
+ &fcs_ok);
+
+ if (packet_error) {
+
+ if (bytes_written > 0 && !fcs_ok) {
+ if (net_ratelimit())
+ printk(KERN_ERR PFX
+ "%s: packet error: bad FCS\n",
+ enic->netdev->name);
+ }
+
+ dev_kfree_skb_any(skb);
+
+ return;
+ }
+
+ if (eop && bytes_written > 0) {
+
+ /* Good receive
+ */
+
+ skb_put(skb, bytes_written);
+ skb->protocol = eth_type_trans(skb, enic->netdev);
+
+ if (enic->csum_rx_enabled && !csum_not_calc) {
+ skb->csum = htons(checksum);
+ skb->ip_summed = CHECKSUM_COMPLETE;
+ }
+
+ skb->dev = enic->netdev;
+ enic->netdev->last_rx = jiffies;
+
+ if (enic->vlan_group && vlan_stripped) {
+
+ if (ENIC_SETTING(enic, LRO))
+ lro_vlan_hwaccel_receive_skb(&enic->lro_mgr,
+ skb, enic->vlan_group,
+ vlan, cq_desc);
+ else
+ vlan_hwaccel_receive_skb(skb,
+ enic->vlan_group, vlan);
+
+ } else {
+
+ if (ENIC_SETTING(enic, LRO))
+ lro_receive_skb(&enic->lro_mgr, skb, cq_desc);
+ else
+ netif_receive_skb(skb);
+
+ }
+
+ } else {
+
+ /* Buffer overflow
+ */
+
+ dev_kfree_skb_any(skb);
+ }
+}
+
+static int enic_rq_service(struct vnic_dev *vdev, struct cq_desc *cq_desc,
+ u8 type, u16 q_number, u16 completed_index, void *opaque)
+{
+ struct enic *enic = vnic_dev_priv(vdev);
+
+ vnic_rq_service(&enic->rq[q_number], cq_desc,
+ completed_index, VNIC_RQ_RETURN_DESC,
+ enic_rq_indicate_buf, opaque);
+
+ return 0;
+}
+
+static void enic_rq_drop_buf(struct vnic_rq *rq,
+ struct cq_desc *cq_desc, struct vnic_rq_buf *buf,
+ int skipped, void *opaque)
+{
+ struct enic *enic = vnic_dev_priv(rq->vdev);
+ struct sk_buff *skb = buf->os_buf;
+
+ if (skipped)
+ return;
+
+ pci_unmap_single(enic->pdev, buf->dma_addr,
+ buf->len, PCI_DMA_FROMDEVICE);
+
+ dev_kfree_skb_any(skb);
+}
+
+static int enic_rq_service_drop(struct vnic_dev *vdev, struct cq_desc *cq_desc,
+ u8 type, u16 q_number, u16 completed_index, void *opaque)
+{
+ struct enic *enic = vnic_dev_priv(vdev);
+
+ vnic_rq_service(&enic->rq[q_number], cq_desc,
+ completed_index, VNIC_RQ_RETURN_DESC,
+ enic_rq_drop_buf, opaque);
+
+ return 0;
+}
+
+static int enic_poll(struct napi_struct *napi, int budget)
+{
+ struct enic *enic = container_of(napi, struct enic, napi);
+ struct net_device *netdev = enic->netdev;
+ unsigned int rq_work_to_do = budget;
+ unsigned int wq_work_to_do = -1; /* no limit */
+ unsigned int work_done, rq_work_done, wq_work_done;
+
+ /* Service RQ (first) and WQ
+ */
+
+ rq_work_done = vnic_cq_service(&enic->cq[ENIC_CQ_RQ],
+ rq_work_to_do, enic_rq_service, NULL);
+
+ wq_work_done = vnic_cq_service(&enic->cq[ENIC_CQ_WQ],
+ wq_work_to_do, enic_wq_service, NULL);
+
+ /* Accumulate intr event credits for this polling
+ * cycle. An intr event is the completion of a
+ * a WQ or RQ packet.
+ */
+
+ work_done = rq_work_done + wq_work_done;
+
+ if (work_done > 0)
+ vnic_intr_return_credits(&enic->intr[ENIC_INTX_WQ_RQ],
+ work_done,
+ 0 /* don't unmask intr */,
+ 0 /* don't reset intr timer */);
+
+ if (rq_work_done > 0) {
+
+ /* Replenish RQ
+ */
+
+ vnic_rq_fill(&enic->rq[0], enic_rq_alloc_buf);
+
+ } else {
+
+ /* If no work done, flush all LROs and exit polling
+ */
+
+ if (ENIC_SETTING(enic, LRO))
+ lro_flush_all(&enic->lro_mgr);
+
+ netif_rx_complete(netdev, napi);
+ vnic_intr_unmask(&enic->intr[ENIC_MSIX_RQ]);
+ }
+
+ return rq_work_done;
+}
+
+static int enic_poll_msix(struct napi_struct *napi, int budget)
+{
+ struct enic *enic = container_of(napi, struct enic, napi);
+ struct net_device *netdev = enic->netdev;
+ unsigned int work_to_do = budget;
+ unsigned int work_done;
+
+ /* Service RQ
+ */
+
+ work_done = vnic_cq_service(&enic->cq[ENIC_CQ_RQ],
+ work_to_do, enic_rq_service, NULL);
+
+ if (work_done > 0) {
+
+ /* Replenish RQ
+ */
+
+ vnic_rq_fill(&enic->rq[0], enic_rq_alloc_buf);
+
+ /* Accumulate intr event credits for this polling
+ * cycle. An intr event is the completion of a
+ * a WQ or RQ packet.
+ */
+
+ vnic_intr_return_credits(&enic->intr[ENIC_MSIX_RQ],
+ work_done,
+ 0 /* don't unmask intr */,
+ 0 /* don't reset intr timer */);
+ } else {
+
+ /* If no work done, flush all LROs and exit polling
+ */
+
+ if (ENIC_SETTING(enic, LRO))
+ lro_flush_all(&enic->lro_mgr);
+
+ netif_rx_complete(netdev, napi);
+ vnic_intr_unmask(&enic->intr[ENIC_MSIX_RQ]);
+ }
+
+ return work_done;
+}
+
+static void enic_notify_timer(unsigned long data)
+{
+ struct enic *enic = (struct enic *)data;
+
+ enic_notify_check(enic);
+
+ mod_timer(&enic->notify_timer, round_jiffies(ENIC_NOTIFY_TIMER_PERIOD));
+}
+
+static void enic_free_intr(struct enic *enic)
+{
+ struct net_device *netdev = enic->netdev;
+ unsigned int i;
+
+ switch (vnic_dev_get_intr_mode(enic->vdev)) {
+ case VNIC_DEV_INTR_MODE_INTX:
+ case VNIC_DEV_INTR_MODE_MSI:
+ free_irq(enic->pdev->irq, netdev);
+ break;
+ case VNIC_DEV_INTR_MODE_MSIX:
+ for (i = 0; i < ARRAY_SIZE(enic->msix); i++)
+ if (enic->msix[i].requested)
+ free_irq(enic->msix_entry[i].vector,
+ enic->msix[i].devid);
+ break;
+ default:
+ break;
+ }
+}
+
+static int enic_request_intr(struct enic *enic)
+{
+ struct net_device *netdev = enic->netdev;
+ unsigned int i;
+ int err = 0;
+
+ switch (vnic_dev_get_intr_mode(enic->vdev)) {
+
+ case VNIC_DEV_INTR_MODE_INTX:
+
+ err = request_irq(enic->pdev->irq, enic_isr_legacy,
+ IRQF_SHARED, netdev->name, netdev);
+ break;
+
+ case VNIC_DEV_INTR_MODE_MSI:
+
+ err = request_irq(enic->pdev->irq, enic_isr_msi,
+ 0, netdev->name, enic);
+ break;
+
+ case VNIC_DEV_INTR_MODE_MSIX:
+
+ sprintf(enic->msix[ENIC_MSIX_RQ].devname,
+ "%.11s-rx", netdev->name);
+ enic->msix[ENIC_MSIX_RQ].isr = enic_isr_msix_rq;
+ enic->msix[ENIC_MSIX_RQ].devid = enic;
+
+ sprintf(enic->msix[ENIC_MSIX_WQ].devname,
+ "%.11s-tx", netdev->name);
+ enic->msix[ENIC_MSIX_WQ].isr = enic_isr_msix_wq;
+ enic->msix[ENIC_MSIX_WQ].devid = enic;
+
+ sprintf(enic->msix[ENIC_MSIX_ERR].devname,
+ "%.11s-err", netdev->name);
+ enic->msix[ENIC_MSIX_ERR].isr = enic_isr_msix_err;
+ enic->msix[ENIC_MSIX_ERR].devid = enic;
+
+ sprintf(enic->msix[ENIC_MSIX_NOTIFY].devname,
+ "%.11s-notify", netdev->name);
+ enic->msix[ENIC_MSIX_NOTIFY].isr = enic_isr_msix_notify;
+ enic->msix[ENIC_MSIX_NOTIFY].devid = enic;
+
+ for (i = 0; i < ARRAY_SIZE(enic->msix); i++) {
+ err = request_irq(enic->msix_entry[i].vector,
+ enic->msix[i].isr, 0,
+ enic->msix[i].devname,
+ enic->msix[i].devid);
+ if (err) {
+ enic_free_intr(enic);
+ break;
+ }
+ enic->msix[i].requested = 1;
+ }
+
+ break;
+
+ default:
+ break;
+ }
+
+ return err;
+}
+
+static int enic_notify_set(struct enic *enic)
+{
+ int err;
+
+ switch (vnic_dev_get_intr_mode(enic->vdev)) {
+ case VNIC_DEV_INTR_MODE_INTX:
+ err = vnic_dev_notify_set(enic->vdev, ENIC_INTX_NOTIFY);
+ break;
+ case VNIC_DEV_INTR_MODE_MSIX:
+ err = vnic_dev_notify_set(enic->vdev, ENIC_MSIX_NOTIFY);
+ break;
+ default:
+ err = vnic_dev_notify_set(enic->vdev, -1 /* no intr */);
+ break;
+ }
+
+ return err;
+}
+
+static void enic_notify_timer_start(struct enic *enic)
+{
+ switch (vnic_dev_get_intr_mode(enic->vdev)) {
+ case VNIC_DEV_INTR_MODE_MSI:
+ mod_timer(&enic->notify_timer, jiffies);
+ break;
+ default:
+ /* Using intr for notification for INTx/MSI-X */
+ break;
+ };
+}
+
+/* rtnl lock is held, process context */
+static int enic_open(struct net_device *netdev)
+{
+ struct enic *enic = netdev_priv(netdev);
+ unsigned int i;
+ int err;
+
+ for (i = 0; i < enic->rq_count; i++) {
+ err = vnic_rq_fill(&enic->rq[i], enic_rq_alloc_buf);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Unable to alloc receive buffers.\n",
+ netdev->name);
+ return err;
+ }
+ }
+
+ for (i = 0; i < enic->wq_count; i++)
+ vnic_wq_enable(&enic->wq[i]);
+ for (i = 0; i < enic->rq_count; i++)
+ vnic_rq_enable(&enic->rq[i]);
+
+ enic_add_station_addr(enic);
+ enic_set_multicast_list(netdev);
+
+ netif_wake_queue(netdev);
+ napi_enable(&enic->napi);
+ vnic_dev_enable(enic->vdev);
+
+ for (i = 0; i < enic->intr_count; i++)
+ vnic_intr_unmask(&enic->intr[i]);
+
+ enic_notify_timer_start(enic);
+
+ return 0;
+}
+
+/* rtnl lock is held, process context */
+static int enic_stop(struct net_device *netdev)
+{
+ struct enic *enic = netdev_priv(netdev);
+ unsigned int i;
+ int err;
+
+ del_timer_sync(&enic->notify_timer);
+
+ vnic_dev_disable(enic->vdev);
+ napi_disable(&enic->napi);
+ netif_stop_queue(netdev);
+
+ for (i = 0; i < enic->intr_count; i++)
+ vnic_intr_mask(&enic->intr[i]);
+
+ for (i = 0; i < enic->wq_count; i++) {
+ err = vnic_wq_disable(&enic->wq[i]);
+ if (err)
+ return err;
+ }
+ for (i = 0; i < enic->rq_count; i++) {
+ err = vnic_rq_disable(&enic->rq[i]);
+ if (err)
+ return err;
+ }
+
+ (void)vnic_cq_service(&enic->cq[ENIC_CQ_RQ],
+ -1, enic_rq_service_drop, NULL);
+ (void)vnic_cq_service(&enic->cq[ENIC_CQ_WQ],
+ -1, enic_wq_service, NULL);
+
+ for (i = 0; i < enic->wq_count; i++)
+ vnic_wq_clean(&enic->wq[i], enic_free_wq_buf);
+ for (i = 0; i < enic->rq_count; i++)
+ vnic_rq_clean(&enic->rq[i], enic_free_rq_buf);
+ for (i = 0; i < enic->cq_count; i++)
+ vnic_cq_clean(&enic->cq[i]);
+ for (i = 0; i < enic->intr_count; i++)
+ vnic_intr_clean(&enic->intr[i]);
+
+ return 0;
+}
+
+static int enic_change_mtu(struct net_device *netdev, int new_mtu)
+{
+ struct enic *enic = netdev_priv(netdev);
+ int running = netif_running(netdev);
+
+ if (running)
+ enic_stop(netdev);
+
+ if (new_mtu < ENIC_MIN_MTU)
+ new_mtu = ENIC_MIN_MTU;
+ if (new_mtu > ENIC_MAX_MTU)
+ new_mtu = ENIC_MAX_MTU;
+
+ netdev->mtu = new_mtu;
+
+ if (netdev->mtu > enic->port_mtu)
+ printk(KERN_WARNING PFX
+ "%s: interface MTU (%d) set higher "
+ "than port MTU (%d)\n",
+ netdev->name, netdev->mtu, enic->port_mtu);
+
+ if (running)
+ enic_open(netdev);
+
+ return 0;
+}
+
+#ifdef CONFIG_NET_POLL_CONTROLLER
+static void enic_poll_controller(struct net_device *netdev)
+{
+ struct enic *enic = netdev_priv(netdev);
+ struct vnic_dev *vdev = enic->vdev;
+
+ switch (vnic_dev_get_intr_mode(vdev)) {
+ case VNIC_DEV_INTR_MODE_MSIX:
+ enic_isr_msix_rq(enic->pdev->irq, enic);
+ enic_isr_msix_wq(enic->pdev->irq, enic);
+ break;
+ case VNIC_DEV_INTR_MODE_MSI:
+ enic_isr_msi(enic->pdev->irq, enic);
+ break;
+ case VNIC_DEV_INTR_MODE_INTX:
+ enic_isr_legacy(enic->pdev->irq, netdev);
+ break;
+ default:
+ break;
+ }
+}
+#endif
+
+static int enic_dev_wait(struct vnic_dev *vdev,
+ int (*start)(struct vnic_dev *, int),
+ int (*finished)(struct vnic_dev *, int *),
+ int arg)
+{
+ unsigned long time;
+ int done;
+ int err;
+
+ BUG_ON(in_interrupt());
+
+ err = start(vdev, arg);
+ if (err)
+ return err;
+
+ /* Wait for func to complete...2 seconds max
+ */
+
+ time = jiffies + (HZ * 2);
+ do {
+
+ err = finished(vdev, &done);
+ if (err)
+ return err;
+
+ if (done)
+ return 0;
+
+ schedule_timeout_uninterruptible(HZ / 10);
+
+ } while (time_after(time, jiffies));
+
+ return -ETIMEDOUT;
+}
+
+static int enic_dev_open(struct enic *enic)
+{
+ int err;
+
+ err = enic_dev_wait(enic->vdev, vnic_dev_open,
+ vnic_dev_open_done, 0);
+ if (err)
+ printk(KERN_ERR PFX
+ "vNIC device open failed, err %d.\n", err);
+
+ return err;
+}
+
+static int enic_dev_soft_reset(struct enic *enic)
+{
+ int err;
+
+ err = enic_dev_wait(enic->vdev, vnic_dev_soft_reset,
+ vnic_dev_soft_reset_done, 0);
+ if (err)
+ printk(KERN_ERR PFX
+ "vNIC soft reset failed, err %d.\n", err);
+
+ return err;
+}
+
+static void enic_reset(struct work_struct *work)
+{
+ struct enic *enic = container_of(work, struct enic, reset);
+
+ if (!netif_running(enic->netdev))
+ return;
+
+ rtnl_lock();
+
+ spin_lock(&enic->devcmd_lock);
+ vnic_dev_hang_notify(enic->vdev);
+ spin_unlock(&enic->devcmd_lock);
+
+ enic_stop(enic->netdev);
+ enic_dev_soft_reset(enic);
+ enic_reset_mcaddrs(enic);
+ enic_init_vnic_resources(enic);
+ enic_open(enic->netdev);
+
+ rtnl_unlock();
+}
+
+static int enic_set_intr_mode(struct enic *enic)
+{
+ unsigned int n = ARRAY_SIZE(enic->rq);
+ unsigned int m = ARRAY_SIZE(enic->wq);
+ unsigned int i;
+
+ /* Set interrupt mode (INTx, MSI, MSI-X) depending
+ * system capabilities.
+ *
+ * Try MSI-X first
+ *
+ * We need n RQs, m WQs, n+m CQs, and n+m+2 INTRs
+ * (the second to last INTR is used for WQ/RQ errors)
+ * (the last INTR is used for notifications)
+ */
+
+ BUG_ON(ARRAY_SIZE(enic->msix_entry) < n + m + 2);
+ for (i = 0; i < n + m + 2; i++)
+ enic->msix_entry[i].entry = i;
+
+ if (enic->config.intr_mode < 1 &&
+ enic->rq_count >= n &&
+ enic->wq_count >= m &&
+ enic->cq_count >= n + m &&
+ enic->intr_count >= n + m + 2 &&
+ !pci_enable_msix(enic->pdev, enic->msix_entry, n + m + 2)) {
+
+ enic->rq_count = n;
+ enic->wq_count = m;
+ enic->cq_count = n + m;
+ enic->intr_count = n + m + 2;
+
+ vnic_dev_set_intr_mode(enic->vdev, VNIC_DEV_INTR_MODE_MSIX);
+
+ return 0;
+ }
+
+ /* Next try MSI
+ *
+ * We need 1 RQ, 1 WQ, 2 CQs, and 1 INTR
+ */
+
+ if (enic->config.intr_mode < 2 &&
+ enic->rq_count >= 1 &&
+ enic->wq_count >= 1 &&
+ enic->cq_count >= 2 &&
+ enic->intr_count >= 1 &&
+ !pci_enable_msi(enic->pdev)) {
+
+ enic->rq_count = 1;
+ enic->wq_count = 1;
+ enic->cq_count = 2;
+ enic->intr_count = 1;
+
+ vnic_dev_set_intr_mode(enic->vdev, VNIC_DEV_INTR_MODE_MSI);
+
+ return 0;
+ }
+
+ /* Next try INTx
+ *
+ * We need 1 RQ, 1 WQ, 2 CQs, and 3 INTRs
+ * (the first INTR is used for WQ/RQ)
+ * (the second INTR is used for WQ/RQ errors)
+ * (the last INTR is used for notifications)
+ */
+
+ if (enic->config.intr_mode < 3 &&
+ enic->rq_count >= 1 &&
+ enic->wq_count >= 1 &&
+ enic->cq_count >= 2 &&
+ enic->intr_count >= 3) {
+
+ enic->rq_count = 1;
+ enic->wq_count = 1;
+ enic->cq_count = 2;
+ enic->intr_count = 3;
+
+ vnic_dev_set_intr_mode(enic->vdev, VNIC_DEV_INTR_MODE_INTX);
+
+ return 0;
+ }
+
+ vnic_dev_set_intr_mode(enic->vdev, VNIC_DEV_INTR_MODE_UNKNOWN);
+
+ return -EINVAL;
+}
+
+static void enic_clear_intr_mode(struct enic *enic)
+{
+ switch (vnic_dev_get_intr_mode(enic->vdev)) {
+ case VNIC_DEV_INTR_MODE_MSIX:
+ pci_disable_msix(enic->pdev);
+ break;
+ case VNIC_DEV_INTR_MODE_MSI:
+ pci_disable_msi(enic->pdev);
+ break;
+ default:
+ break;
+ }
+
+ vnic_dev_set_intr_mode(enic->vdev, VNIC_DEV_INTR_MODE_UNKNOWN);
+}
+
+static void enic_iounmap(struct enic *enic)
+{
+ if (enic->bar0.vaddr)
+ iounmap(enic->bar0.vaddr);
+}
+
+static int __devinit enic_probe(struct pci_dev *pdev,
+ const struct pci_device_id *ent)
+{
+ struct net_device *netdev;
+ struct enic *enic;
+ int using_dac = 0;
+ unsigned int i;
+ int err;
+
+ const u8 rss_default_cpu = 0;
+ const u8 rss_hash_type = 0;
+ const u8 rss_hash_bits = 0;
+ const u8 rss_base_cpu = 0;
+ const u8 rss_enable = 0;
+ const u8 tso_ipid_split_en = 0;
+ const u8 ig_vlan_strip_en = 1;
+
+ /* Allocate net device structure and initialize. Private
+ * instance data is initialized to zero.
+ */
+
+ netdev = alloc_etherdev(sizeof(struct enic));
+ if (!netdev) {
+ printk(KERN_ERR PFX "Etherdev alloc failed, aborting.\n");
+ return -ENOMEM;
+ }
+
+ /* Set the netdev name early so intr vectors are properly
+ * named and any error msgs can include netdev->name
+ */
+
+ rtnl_lock();
+ err = dev_alloc_name(netdev, netdev->name);
+ rtnl_unlock();
+ if (err < 0) {
+ printk(KERN_ERR PFX "Unable to allocate netdev name.\n");
+ goto err_out_free_netdev;
+ }
+
+ pci_set_drvdata(pdev, netdev);
+
+ SET_NETDEV_DEV(netdev, &pdev->dev);
+
+ enic = netdev_priv(netdev);
+ enic->netdev = netdev;
+ enic->pdev = pdev;
+
+ /* Setup PCI resources
+ */
+
+ err = pci_enable_device(pdev);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Cannot enable PCI device, aborting.\n",
+ netdev->name);
+ goto err_out_free_netdev;
+ }
+
+ err = pci_request_regions(pdev, DRV_NAME);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Cannot request PCI regions, aborting.\n",
+ netdev->name);
+ goto err_out_disable_device;
+ }
+
+ pci_set_master(pdev);
+
+ /* Query PCI controller on system for DMA addressing
+ * limitation for the device. Try 40-bit first, and
+ * fail to 32-bit.
+ */
+
+ err = pci_set_dma_mask(pdev, DMA_40BIT_MASK);
+ if (err) {
+ err = pci_set_dma_mask(pdev, DMA_32BIT_MASK);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: No usable DMA configuration, aborting.\n",
+ netdev->name);
+ goto err_out_release_regions;
+ }
+ err = pci_set_consistent_dma_mask(pdev, DMA_32BIT_MASK);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Unable to obtain 32-bit DMA "
+ "for consistent allocations, aborting.\n",
+ netdev->name);
+ goto err_out_release_regions;
+ }
+ } else {
+ err = pci_set_consistent_dma_mask(pdev, DMA_40BIT_MASK);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Unable to obtain 40-bit DMA "
+ "for consistent allocations, aborting.\n",
+ netdev->name);
+ goto err_out_release_regions;
+ }
+ using_dac = 1;
+ }
+
+ /* Map vNIC resources from BAR0
+ */
+
+ if (!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM)) {
+ printk(KERN_ERR PFX
+ "%s: BAR0 not memory-map'able, aborting.\n",
+ netdev->name);
+ err = -ENODEV;
+ goto err_out_release_regions;
+ }
+
+ enic->bar0.vaddr = pci_iomap(pdev, 0, enic->bar0.len);
+ enic->bar0.bus_addr = pci_resource_start(pdev, 0);
+ enic->bar0.len = pci_resource_len(pdev, 0);
+
+ if (!enic->bar0.vaddr) {
+ printk(KERN_ERR PFX
+ "%s: Cannot memory-map BAR0 res hdr, aborting.\n",
+ netdev->name);
+ err = -ENODEV;
+ goto err_out_release_regions;
+ }
+
+ /* Register vNIC device
+ */
+
+ enic->vdev = vnic_dev_register(NULL, enic, pdev, &enic->bar0);
+ if (!enic->vdev) {
+ printk(KERN_ERR PFX
+ "%s: vNIC registration failed, aborting.\n",
+ netdev->name);
+ err = -ENODEV;
+ goto err_out_iounmap;
+ }
+
+ /* Issue device open to get device in known state
+ */
+
+ err = enic_dev_open(enic);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: vNIC dev open failed, aborting.\n",
+ netdev->name);
+ goto err_out_vnic_unregister;
+ }
+
+ /* Issue device init to initialize the vnic-to-switch link.
+ * We'll start with carrier off and wait for link UP
+ * notification later to turn on carrier. We don't need
+ * to wait here for the vnic-to-switch link initialization
+ * to complete; link UP notification is the indication that
+ * the process is complete.
+ */
+
+ netif_carrier_off(netdev);
+
+ err = vnic_dev_init(enic->vdev, 0);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: vNIC dev init failed, aborting.\n",
+ netdev->name);
+ goto err_out_dev_close;
+ }
+
+ /* Get vNIC configuration
+ */
+
+ err = enic_get_vnic_config(enic);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Get vNIC configuration failed, aborting.\n",
+ netdev->name);
+ goto err_out_dev_close;
+ }
+
+ /* Get available resource counts
+ */
+
+ enic_get_res_counts(enic);
+
+ /* Set interrupt mode based on resource counts and system
+ * capabilities
+ */
+
+ err = enic_set_intr_mode(enic);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Failed to set intr mode, aborting.\n",
+ netdev->name);
+ goto err_out_dev_close;
+ }
+
+ /* Request interrupt vector(s)
+ */
+
+ err = enic_request_intr(enic);
+ if (err) {
+ printk(KERN_ERR PFX "%s: Unable to request irq.\n",
+ netdev->name);
+ goto err_out_dev_close;
+ }
+
+ /* Allocate and configure vNIC resources
+ */
+
+ err = enic_alloc_vnic_resources(enic);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Failed to alloc vNIC resources, aborting.\n",
+ netdev->name);
+ goto err_out_free_vnic_resources;
+ }
+
+ enic_init_vnic_resources(enic);
+
+ /* Enable VLAN tag stripping. RSS not enabled (yet).
+ */
+
+ err = enic_set_nic_cfg(enic,
+ rss_default_cpu, rss_hash_type,
+ rss_hash_bits, rss_base_cpu,
+ rss_enable, tso_ipid_split_en,
+ ig_vlan_strip_en);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Failed to config nic, aborting.\n",
+ netdev->name);
+ goto err_out_free_vnic_resources;
+ }
+
+ /* Setup notification buffer area
+ */
+
+ err = enic_notify_set(enic);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Failed to alloc notify buffer, aborting.\n",
+ netdev->name);
+ goto err_out_free_vnic_resources;
+ }
+
+ /* Setup notification timer, HW reset task, and locks
+ */
+
+ init_timer(&enic->notify_timer);
+ enic->notify_timer.function = enic_notify_timer;
+ enic->notify_timer.data = (unsigned long)enic;
+
+ INIT_WORK(&enic->reset, enic_reset);
+
+ for (i = 0; i < enic->wq_count; i++)
+ spin_lock_init(&enic->wq_lock[i]);
+
+ spin_lock_init(&enic->devcmd_lock);
+
+ /* Register net device
+ */
+
+ enic->port_mtu = enic->config.mtu;
+ (void)enic_change_mtu(netdev, enic->port_mtu);
+
+ err = enic_set_mac_addr(netdev, enic->mac_addr);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Invalid MAC address, aborting.\n",
+ netdev->name);
+ goto err_out_notify_unset;
+ }
+
+ netdev->open = enic_open;
+ netdev->stop = enic_stop;
+ netdev->hard_start_xmit = enic_hard_start_xmit;
+ netdev->get_stats = enic_get_stats;
+ netdev->set_multicast_list = enic_set_multicast_list;
+ netdev->change_mtu = enic_change_mtu;
+ netdev->vlan_rx_register = enic_vlan_rx_register;
+ netdev->vlan_rx_add_vid = enic_vlan_rx_add_vid;
+ netdev->vlan_rx_kill_vid = enic_vlan_rx_kill_vid;
+ netdev->tx_timeout = enic_tx_timeout;
+ netdev->watchdog_timeo = 2 * HZ;
+ netdev->ethtool_ops = &enic_ethtool_ops;
+#ifdef CONFIG_NET_POLL_CONTROLLER
+ netdev->poll_controller = enic_poll_controller;
+#endif
+
+ switch (vnic_dev_get_intr_mode(enic->vdev)) {
+ default:
+ netif_napi_add(netdev, &enic->napi, enic_poll, 64);
+ break;
+ case VNIC_DEV_INTR_MODE_MSIX:
+ netif_napi_add(netdev, &enic->napi, enic_poll_msix, 64);
+ break;
+ }
+
+ netdev->features |= NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX;
+ if (ENIC_SETTING(enic, TXCSUM))
+ netdev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
+ if (ENIC_SETTING(enic, TSO))
+ netdev->features |= NETIF_F_TSO |
+ NETIF_F_TSO6 | NETIF_F_TSO_ECN;
+ if (using_dac)
+ netdev->features |= NETIF_F_HIGHDMA;
+
+
+ enic->csum_rx_enabled = ENIC_SETTING(enic, RXCSUM);
+
+ if (ENIC_SETTING(enic, LRO)) {
+ enic->lro_mgr.max_aggr = ENIC_LRO_MAX_AGGR;
+ enic->lro_mgr.max_desc = ENIC_LRO_MAX_DESC;
+ enic->lro_mgr.lro_arr = enic->lro_desc;
+ enic->lro_mgr.get_skb_header = enic_get_skb_header;
+ enic->lro_mgr.features = LRO_F_NAPI | LRO_F_EXTRACT_VLAN_ID;
+ enic->lro_mgr.dev = netdev;
+ enic->lro_mgr.ip_summed = CHECKSUM_COMPLETE;
+ enic->lro_mgr.ip_summed_aggr = CHECKSUM_UNNECESSARY;
+ }
+
+ err = register_netdev(netdev);
+ if (err) {
+ printk(KERN_ERR PFX
+ "%s: Cannot register net device, aborting.\n",
+ netdev->name);
+ goto err_out_notify_unset;
+ }
+
+ return 0;
+
+err_out_notify_unset:
+ vnic_dev_notify_unset(enic->vdev);
+err_out_free_vnic_resources:
+ enic_free_vnic_resources(enic);
+ enic_free_intr(enic);
+err_out_dev_close:
+ vnic_dev_close(enic->vdev);
+err_out_vnic_unregister:
+ enic_clear_intr_mode(enic);
+ vnic_dev_unregister(enic->vdev);
+err_out_iounmap:
+ enic_iounmap(enic);
+err_out_release_regions:
+ pci_release_regions(pdev);
+err_out_disable_device:
+ pci_disable_device(pdev);
+err_out_free_netdev:
+ pci_set_drvdata(pdev, NULL);
+ free_netdev(netdev);
+
+ return err;
+}
+
+static void __devexit enic_remove(struct pci_dev *pdev)
+{
+ struct net_device *netdev = pci_get_drvdata(pdev);
+
+ if (netdev) {
+ struct enic *enic = netdev_priv(netdev);
+
+ flush_scheduled_work();
+ unregister_netdev(netdev);
+ vnic_dev_notify_unset(enic->vdev);
+ enic_free_vnic_resources(enic);
+ enic_free_intr(enic);
+ vnic_dev_close(enic->vdev);
+ enic_clear_intr_mode(enic);
+ vnic_dev_unregister(enic->vdev);
+ enic_iounmap(enic);
+ pci_release_regions(pdev);
+ pci_disable_device(pdev);
+ pci_set_drvdata(pdev, NULL);
+ free_netdev(netdev);
+ }
+}
+
+static struct pci_driver enic_driver = {
+ .name = DRV_NAME,
+ .id_table = enic_id_table,
+ .probe = enic_probe,
+ .remove = __devexit_p(enic_remove),
+};
+
+static int __init enic_init_module(void)
+{
+ printk(KERN_INFO PFX "%s, ver %s\n", DRV_DESCRIPTION, DRV_VERSION);
+
+ return pci_register_driver(&enic_driver);
+}
+
+static void __exit enic_cleanup_module(void)
+{
+ pci_unregister_driver(&enic_driver);
+}
+
+module_init(enic_init_module);
+module_exit(enic_cleanup_module);
diff -Naurp -X linux-2.6.26.3/Documentation/dontdiff linux-2.6.26.3/drivers/net/enic/enic_res.c linux-2.6.26.3-enic/drivers/net/enic/enic_res.c
--- linux-2.6.26.3/drivers/net/enic/enic_res.c 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.26.3-enic/drivers/net/enic/enic_res.c 2008-09-08 15:16:23.000000000 -0700
@@ -0,0 +1,370 @@
+/*
+ * Copyright 2008 Cisco Systems, Inc. All rights reserved.
+ * Copyright 2007 Nuova Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * 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/errno.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/netdevice.h>
+
+#include "wq_enet_desc.h"
+#include "rq_enet_desc.h"
+#include "cq_enet_desc.h"
+#include "vnic_resource.h"
+#include "vnic_enet.h"
+#include "vnic_dev.h"
+#include "vnic_wq.h"
+#include "vnic_rq.h"
+#include "vnic_cq.h"
+#include "vnic_intr.h"
+#include "vnic_stats.h"
+#include "vnic_nic.h"
+#include "vnic_rss.h"
+#include "enic_res.h"
+#include "enic.h"
+
+int enic_get_vnic_config(struct enic *enic)
+{
+ struct vnic_enet_config *c = &enic->config;
+ int err;
+
+ err = vnic_dev_mac_addr(enic->vdev, enic->mac_addr);
+ if (err) {
+ printk(KERN_ERR PFX "Error getting MAC addr, %d\n", err);
+ return err;
+ }
+
+#define GET_CONFIG(m) \
+ do { \
+ err = vnic_dev_spec(enic->vdev, \
+ offsetof(struct vnic_enet_config, m), \
+ sizeof(c->m), &c->m); \
+ if (err) { \
+ printk(KERN_ERR PFX \
+ "Error getting %s, %d\n", #m, err); \
+ return err; \
+ } \
+ } while (0)
+
+ GET_CONFIG(flags);
+ GET_CONFIG(wq_desc_count);
+ GET_CONFIG(rq_desc_count);
+ GET_CONFIG(mtu);
+ GET_CONFIG(intr_timer);
+ GET_CONFIG(intr_timer_type);
+ GET_CONFIG(intr_mode);
+
+ c->wq_desc_count =
+ min_t(u32, ENIC_MAX_WQ_DESCS,
+ max_t(u32, ENIC_MIN_WQ_DESCS,
+ c->wq_desc_count));
+ c->wq_desc_count &= 0xfffffff0; /* must be aligned to groups of 16 */
+
+ c->rq_desc_count =
+ min_t(u32, ENIC_MAX_RQ_DESCS,
+ max_t(u32, ENIC_MIN_RQ_DESCS,
+ c->rq_desc_count));
+ c->rq_desc_count &= 0xfffffff0; /* must be aligned to groups of 16 */
+
+ if (c->mtu == 0)
+ c->mtu = 1500;
+ c->mtu = min_t(u16, ENIC_MAX_MTU,
+ max_t(u16, ENIC_MIN_MTU,
+ c->mtu));
+
+ c->intr_timer = min_t(u16, VNIC_INTR_TIMER_MAX, c->intr_timer);
+
+ printk(KERN_INFO PFX "vNIC MAC addr %02x:%02x:%02x:%02x:%02x:%02x "
+ "wq/rq %d/%d\n",
+ enic->mac_addr[0], enic->mac_addr[1], enic->mac_addr[2],
+ enic->mac_addr[3], enic->mac_addr[4], enic->mac_addr[5],
+ c->wq_desc_count, c->rq_desc_count);
+ printk(KERN_INFO PFX "vNIC mtu %d csum tx/rx %d/%d tso/lro %d/%d "
+ "intr timer %d\n",
+ c->mtu, ENIC_SETTING(enic, TXCSUM),
+ ENIC_SETTING(enic, RXCSUM), ENIC_SETTING(enic, TSO),
+ ENIC_SETTING(enic, LRO), c->intr_timer);
+
+ return 0;
+}
+
+void enic_add_station_addr(struct enic *enic)
+{
+ vnic_dev_add_addr(enic->vdev, enic->mac_addr);
+}
+
+void enic_add_multicast_addr(struct enic *enic, u8 *addr)
+{
+ vnic_dev_add_addr(enic->vdev, addr);
+}
+
+void enic_del_multicast_addr(struct enic *enic, u8 *addr)
+{
+ vnic_dev_del_addr(enic->vdev, addr);
+}
+
+void enic_add_vlan(struct enic *enic, u16 vlanid)
+{
+ u64 a0 = vlanid, a1 = 0;
+ int wait = 1000;
+ int err;
+
+ err = vnic_dev_cmd(enic->vdev, CMD_VLAN_ADD, &a0, &a1, wait);
+ if (err)
+ printk(KERN_ERR PFX "Can't add vlan id, %d\n", err);
+}
+
+void enic_del_vlan(struct enic *enic, u16 vlanid)
+{
+ u64 a0 = vlanid, a1 = 0;
+ int wait = 1000;
+ int err;
+
+ err = vnic_dev_cmd(enic->vdev, CMD_VLAN_DEL, &a0, &a1, wait);
+ if (err)
+ printk(KERN_ERR PFX "Can't delete vlan id, %d\n", err);
+}
+
+int enic_set_nic_cfg(struct enic *enic, u8 rss_default_cpu, u8 rss_hash_type,
+ u8 rss_hash_bits, u8 rss_base_cpu, u8 rss_enable, u8 tso_ipid_split_en,
+ u8 ig_vlan_strip_en)
+{
+ u64 a0, a1;
+ u32 nic_cfg;
+ int wait = 1000;
+
+ vnic_set_nic_cfg(&nic_cfg, rss_default_cpu,
+ rss_hash_type, rss_hash_bits, rss_base_cpu,
+ rss_enable, tso_ipid_split_en, ig_vlan_strip_en);
+
+ a0 = nic_cfg;
+ a1 = 0;
+
+ return vnic_dev_cmd(enic->vdev, CMD_NIC_CFG, &a0, &a1, wait);
+}
+
+void enic_free_vnic_resources(struct enic *enic)
+{
+ unsigned int i;
+
+ for (i = 0; i < enic->wq_count; i++)
+ vnic_wq_free(&enic->wq[i]);
+ for (i = 0; i < enic->rq_count; i++)
+ vnic_rq_free(&enic->rq[i]);
+ for (i = 0; i < enic->cq_count; i++)
+ vnic_cq_free(&enic->cq[i]);
+ for (i = 0; i < enic->intr_count; i++)
+ vnic_intr_free(&enic->intr[i]);
+}
+
+void enic_get_res_counts(struct enic *enic)
+{
+ enic->wq_count = vnic_dev_get_res_count(enic->vdev, RES_TYPE_WQ);
+ enic->rq_count = vnic_dev_get_res_count(enic->vdev, RES_TYPE_RQ);
+ enic->cq_count = vnic_dev_get_res_count(enic->vdev, RES_TYPE_CQ);
+ enic->intr_count = vnic_dev_get_res_count(enic->vdev,
+ RES_TYPE_INTR_CTRL);
+
+ printk(KERN_INFO PFX "vNIC resources avail: "
+ "wq %d rq %d cq %d intr %d\n",
+ enic->wq_count, enic->rq_count,
+ enic->cq_count, enic->intr_count);
+}
+
+void enic_init_vnic_resources(struct enic *enic)
+{
+ enum vnic_dev_intr_mode intr_mode;
+ unsigned int mask_on_assertion;
+ unsigned int interrupt_offset;
+ unsigned int error_interrupt_enable;
+ unsigned int error_interrupt_offset;
+ unsigned int cq_index;
+ unsigned int i;
+
+ intr_mode = vnic_dev_get_intr_mode(enic->vdev);
+
+ /* Init RQ/WQ resources.
+ *
+ * RQ[0 - n-1] point to CQ[0 - n-1]
+ * WQ[0 - m-1] point to CQ[n - n+m-1]
+ *
+ * Error interrupt is not enabled for MSI.
+ */
+
+ switch (intr_mode) {
+ case VNIC_DEV_INTR_MODE_INTX:
+ case VNIC_DEV_INTR_MODE_MSIX:
+ error_interrupt_enable = 1;
+ error_interrupt_offset = enic->intr_count - 2;
+ break;
+ default:
+ error_interrupt_enable = 0;
+ error_interrupt_offset = 0;
+ break;
+ }
+
+ for (i = 0; i < enic->rq_count; i++) {
+ cq_index = i;
+ vnic_rq_init(&enic->rq[i],
+ cq_index,
+ error_interrupt_enable,
+ error_interrupt_offset);
+ }
+
+ for (i = 0; i < enic->wq_count; i++) {
+ cq_index = enic->rq_count + i;
+ vnic_wq_init(&enic->wq[i],
+ cq_index,
+ error_interrupt_enable,
+ error_interrupt_offset);
+ }
+
+ /* Init CQ resources
+ *
+ * CQ[0 - n+m-1] point to INTR[0] for INTx, MSI
+ * CQ[0 - n+m-1] point to INTR[0 - n+m-1] for MSI-X
+ */
+
+ for (i = 0; i < enic->cq_count; i++) {
+
+ switch (intr_mode) {
+ case VNIC_DEV_INTR_MODE_MSIX:
+ interrupt_offset = i;
+ break;
+ default:
+ interrupt_offset = 0;
+ break;
+ }
+
+ vnic_cq_init(&enic->cq[i],
+ 0 /* flow_control_enable */,
+ 1 /* color_enable */,
+ 0 /* cq_head */,
+ 0 /* cq_tail */,
+ 1 /* cq_tail_color */,
+ 1 /* interrupt_enable */,
+ 1 /* cq_entry_enable */,
+ 0 /* cq_message_enable */,
+ interrupt_offset,
+ 0 /* cq_message_addr */);
+ }
+
+ /* Init INTR resources
+ *
+ * mask_on_assertion is not used for INTx due to the level-
+ * triggered nature of INTx
+ */
+
+ switch (intr_mode) {
+ case VNIC_DEV_INTR_MODE_MSI:
+ case VNIC_DEV_INTR_MODE_MSIX:
+ mask_on_assertion = 1;
+ break;
+ default:
+ mask_on_assertion = 0;
+ break;
+ }
+
+ for (i = 0; i < enic->intr_count; i++) {
+ vnic_intr_init(&enic->intr[i],
+ enic->config.intr_timer,
+ enic->config.intr_timer_type,
+ mask_on_assertion);
+ }
+
+ /* Clear LIF stats
+ */
+
+ vnic_dev_stats_clear(enic->vdev);
+}
+
+int enic_alloc_vnic_resources(struct enic *enic)
+{
+ enum vnic_dev_intr_mode intr_mode;
+ unsigned int i;
+ int err;
+
+ intr_mode = vnic_dev_get_intr_mode(enic->vdev);
+
+ printk(KERN_INFO PFX "vNIC resources used: "
+ "wq %d rq %d cq %d intr %d intr mode %s\n",
+ enic->wq_count, enic->rq_count,
+ enic->cq_count, enic->intr_count,
+ intr_mode == VNIC_DEV_INTR_MODE_INTX ? "legacy PCI INTx" :
+ intr_mode == VNIC_DEV_INTR_MODE_MSI ? "MSI" :
+ intr_mode == VNIC_DEV_INTR_MODE_MSIX ? "MSI-X" :
+ "unknown"
+ );
+
+ /* Allocate queue resources
+ */
+
+ for (i = 0; i < enic->wq_count; i++) {
+ err = vnic_wq_alloc(enic->vdev, &enic->wq[i], i,
+ enic->config.wq_desc_count,
+ sizeof(struct wq_enet_desc));
+ if (err)
+ goto err_out_cleanup;
+ }
+
+ for (i = 0; i < enic->rq_count; i++) {
+ err = vnic_rq_alloc(enic->vdev, &enic->rq[i], i,
+ enic->config.rq_desc_count,
+ sizeof(struct rq_enet_desc));
+ if (err)
+ goto err_out_cleanup;
+ }
+
+ for (i = 0; i < enic->cq_count; i++) {
+ if (i < enic->rq_count)
+ err = vnic_cq_alloc(enic->vdev, &enic->cq[i], i,
+ enic->config.rq_desc_count,
+ sizeof(struct cq_enet_rq_desc));
+ else
+ err = vnic_cq_alloc(enic->vdev, &enic->cq[i], i,
+ enic->config.wq_desc_count,
+ sizeof(struct cq_enet_wq_desc));
+ if (err)
+ goto err_out_cleanup;
+ }
+
+ for (i = 0; i < enic->intr_count; i++) {
+ err = vnic_intr_alloc(enic->vdev, &enic->intr[i], i);
+ if (err)
+ goto err_out_cleanup;
+ }
+
+ /* Hook remaining resource
+ */
+
+ enic->legacy_pba = vnic_dev_get_res(enic->vdev,
+ RES_TYPE_INTR_PBA_LEGACY, 0);
+ if (!enic->legacy_pba && intr_mode == VNIC_DEV_INTR_MODE_INTX) {
+ printk(KERN_ERR PFX "Failed to hook legacy pba resource\n");
+ err = -ENODEV;
+ goto err_out_cleanup;
+ }
+
+ return 0;
+
+err_out_cleanup:
+ enic_free_vnic_resources(enic);
+
+ return err;
+}
diff -Naurp -X linux-2.6.26.3/Documentation/dontdiff linux-2.6.26.3/drivers/net/enic/enic_res.h linux-2.6.26.3-enic/drivers/net/enic/enic_res.h
--- linux-2.6.26.3/drivers/net/enic/enic_res.h 1969-12-31 16:00:00.000000000 -0800
+++ linux-2.6.26.3-enic/drivers/net/enic/enic_res.h 2008-09-08 15:16:24.000000000 -0700
@@ -0,0 +1,151 @@
+/*
+ * Copyright 2008 Cisco Systems, Inc. All rights reserved.
+ * Copyright 2007 Nuova Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * 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 _ENIC_RES_H_
+#define _ENIC_RES_H_
+
+#include "wq_enet_desc.h"
+#include "rq_enet_desc.h"
+#include "vnic_wq.h"
+#include "vnic_rq.h"
+
+#define ENIC_MIN_WQ_DESCS 64
+#define ENIC_MAX_WQ_DESCS 4096
+#define ENIC_MIN_RQ_DESCS 64
+#define ENIC_MAX_RQ_DESCS 4096
+
+#define ENIC_MIN_MTU 576 /* minimum for IPv4 */
+#define ENIC_MAX_MTU 9000
+
+#define ENIC_MULTICAST_PERFECT_FILTERS 32
+
+#define ENIC_NON_TSO_MAX_DESC 16
+
+#define ENIC_SETTING(enic, f) ((enic->config.flags & VENETF_##f) ? 1 : 0)
+
+static inline void enic_queue_wq_desc_ex(struct vnic_wq *wq,
+ void *os_buf, dma_addr_t dma_addr, unsigned int len,
+ unsigned int mss_or_csum_offset, unsigned int hdr_len,
+ int vlan_tag_insert, unsigned int vlan_tag,
+ int offload_mode, int cq_entry, int sop, int eop)
+{
+ struct wq_enet_desc *desc = vnic_wq_next_desc(wq);
+
+ wq_enet_desc_enc(desc,
+ (u64)dma_addr | VNIC_PADDR_TARGET,
+ (u16)len,
+ (u16)mss_or_csum_offset,
+ (u16)hdr_len, (u8)offload_mode,
+ (u8)eop, (u8)cq_entry,
+ 0, /* fcoe_encap */
+ (u8)vlan_tag_insert,
+ (u16)vlan_tag,
+ 0 /* loopback */);
+
+ wmb();
+
+ vnic_wq_post(wq, os_buf, dma_addr, len, sop, eop);
+}
+
+static inline void enic_queue_wq_desc_cont(struct vnic_wq *wq,
+ void *os_buf, dma_addr_t dma_addr, unsigned int len, int eop)
+{
+ enic_queue_wq_desc_ex(wq, os_buf, dma_addr, len,
+ 0, 0, 0, 0, 0,
+ eop, 0 /* !SOP */, eop);
+}
+
+static inline void enic_queue_wq_desc(struct vnic_wq *wq, void *os_buf,
+ dma_addr_t dma_addr, unsigned int len, int vlan_tag_insert,
+ unsigned int vlan_tag, int eop)
+{
+ enic_queue_wq_desc_ex(wq, os_buf, dma_addr, len,
+ 0, 0, vlan_tag_insert, vlan_tag,
+ WQ_ENET_OFFLOAD_MODE_CSUM,
+ eop, 1 /* SOP */, eop);
+}
+
+static inline void enic_queue_wq_desc_csum(struct vnic_wq *wq,
+ void *os_buf, dma_addr_t dma_addr, unsigned int len,
+ int ip_csum, int tcpudp_csum, int vlan_tag_insert,
+ unsigned int vlan_tag, int eop)
+{
+ enic_queue_wq_desc_ex(wq, os_buf, dma_addr, len,
+ (ip_csum ? 1 : 0) + (tcpudp_csum ? 2 : 0),
+ 0, vlan_tag_insert, vlan_tag,
+ WQ_ENET_OFFLOAD_MODE_CSUM,
+ eop, 1 /* SOP */, eop);
+}
+
+static inline void enic_queue_wq_desc_csum_l4(struct vnic_wq *wq,
+ void *os_buf, dma_addr_t dma_addr, unsigned int len,
+ unsigned int csum_offset, unsigned int hdr_len,
+ int vlan_tag_insert, unsigned int vlan_tag, int eop)
+{
+ enic_queue_wq_desc_ex(wq, os_buf, dma_addr, len,
+ csum_offset, hdr_len, vlan_tag_insert, vlan_tag,
+ WQ_ENET_OFFLOAD_MODE_CSUM_L4,
+ eop, 1 /* SOP */, eop);
+}
+
+static inline void enic_queue_wq_desc_tso(struct vnic_wq *wq,
+ void *os_buf, dma_addr_t dma_addr, unsigned int len,
+ unsigned int mss, unsigned int hdr_len, int vlan_tag_insert,
+ unsigned int vlan_tag, int eop)
+{
+ enic_queue_wq_desc_ex(wq, os_buf, dma_addr, len,
+ mss, hdr_len, vlan_tag_insert, vlan_tag,
+ WQ_ENET_OFFLOAD_MODE_TSO,
+ eop, 1 /* SOP */, eop);
+}
+
+static inline void enic_queue_rq_desc(struct vnic_rq *rq,
+ void *os_buf, unsigned int os_buf_index,
+ dma_addr_t dma_addr, unsigned int len)
+{
+ struct rq_enet_desc *desc = vnic_rq_next_desc(rq);
+ u8 type = os_buf_index ?
+ RQ_ENET_TYPE_NOT_SOP : RQ_ENET_TYPE_ONLY_SOP;
+
+ rq_enet_desc_enc(desc,
+ (u64)dma_addr | VNIC_PADDR_TARGET,
+ type, (u16)len);
+
+ wmb();
+
+ vnic_rq_post(rq, os_buf, os_buf_index, dma_addr, len);
+}
+
+struct enic;
+
+int enic_get_vnic_config(struct enic *);
+void enic_add_station_addr(struct enic *enic);
+void enic_add_multicast_addr(struct enic *enic, u8 *addr);
+void enic_del_multicast_addr(struct enic *enic, u8 *addr);
+void enic_add_vlan(struct enic *enic, u16 vlanid);
+void enic_del_vlan(struct enic *enic, u16 vlanid);
+int enic_set_nic_cfg(struct enic *enic, u8 rss_default_cpu, u8 rss_hash_type,
+ u8 rss_hash_bits, u8 rss_base_cpu, u8 rss_enable, u8 tso_ipid_split_en,
+ u8 ig_vlan_strip_en);
+void enic_get_res_counts(struct enic *enic);
+void enic_init_vnic_resources(struct enic *enic);
+int enic_alloc_vnic_resources(struct enic *);
+void enic_free_vnic_resources(struct enic *);
+
+#endif /* _ENIC_RES_H_ */
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH 2/3] enic: add main netdev file with module infrastructure
2008-09-09 3:57 [PATCH 2/3] enic: add main netdev file with module infrastructure Scott Feldman
@ 2008-09-12 19:47 ` Ben Hutchings
2008-09-17 1:49 ` [PATCH 2/3] enic: add main netdev file with moduleinfrastructure Scott Feldman
0 siblings, 1 reply; 7+ messages in thread
From: Ben Hutchings @ 2008-09-12 19:47 UTC (permalink / raw)
To: Scott Feldman; +Cc: jgarzik, netdev
On Mon, 2008-09-08 at 20:57 -0700, Scott Feldman wrote:
[...]
> diff -Naurp -X linux-2.6.26.3/Documentation/dontdiff linux-2.6.26.3/drivers/net/enic/enic_main.c linux-2.6.26.3-enic/drivers/net/enic/enic_main.c
> --- linux-2.6.26.3/drivers/net/enic/enic_main.c 1969-12-31 16:00:00.000000000 -0800
> +++ linux-2.6.26.3-enic/drivers/net/enic/enic_main.c 2008-09-08 15:16:23.000000000 -0700
[...]
> +static int enic_set_rx_csum(struct net_device *netdev, u32 data)
> +{
> + struct enic *enic = netdev_priv(netdev);
> +
> + enic->csum_rx_enabled =
> + (data && ENIC_SETTING(enic, RXCSUM)) ? 1 : 0;
> +
> + return 0;
> +}
This should return an error code if it can't enable RX checksum:
if (data && !ENIC_SETTING(enic, RXCSUM))
return -EOPNOTSUPP;
enic->csum_rx_enabled = !!data;
return 0;
(I'm not sure what the correct error code is in this situation though.)
> +static int enic_set_tx_csum(struct net_device *netdev, u32 data)
> +{
> + struct enic *enic = netdev_priv(netdev);
> +
> + if (data && ENIC_SETTING(enic, TXCSUM))
> + netdev->features |= NETIF_F_HW_CSUM;
> + else
> + netdev->features &= ~NETIF_F_HW_CSUM;
> +
> + return 0;
> +}
Similarly here.
[...]
> +static struct ethtool_ops enic_ethtool_ops = {
> + .get_settings = enic_get_settings,
> + .get_drvinfo = enic_get_drvinfo,
> + .get_msglevel = enic_get_msglevel,
> + .set_msglevel = enic_set_msglevel,
> + .get_link = ethtool_op_get_link,
> + .get_strings = enic_get_strings,
> + .get_stats_count = enic_get_stats_count,
get_stats_count is deprecated; you should implement get_sset_count.
[...]
> +static int enic_wq_service(struct vnic_dev *vdev, struct cq_desc *cq_desc,
> + u8 type, u16 q_number, u16 completed_index, void *opaque)
> +{
> + struct enic *enic = vnic_dev_priv(vdev);
> +
> + spin_lock(&enic->wq_lock[q_number]);
> +
> + vnic_wq_service(&enic->wq[q_number], cq_desc,
> + completed_index, enic_wq_free_buf,
> + opaque);
> +
> + if (netif_queue_stopped(enic->netdev) &&
> + vnic_wq_desc_avail(&enic->wq[q_number]) >= MAX_SKB_FRAGS + 1)
> + netif_wake_queue(enic->netdev);
The call to netif_queue_stopped() seems redundant since
netif_wake_queue() starts by checking the same thing.
> + spin_unlock(&enic->wq_lock[q_number]);
> +
> + return 0;
> +}
[...]
> +/* dev_base_lock rwlock held, nominally process context */
> +static struct net_device_stats *enic_get_stats(struct net_device *netdev)
> +{
> + struct enic *enic = netdev_priv(netdev);
> + struct vnic_stats *stats;
> +
> + spin_lock(&enic->devcmd_lock);
> + vnic_dev_stats_dump(enic->vdev, &stats);
> + spin_unlock(&enic->devcmd_lock);
> +
> + enic->net_stats.tx_packets = stats->tx.tx_frames_ok;
> + enic->net_stats.tx_bytes = stats->tx.tx_bytes_ok;
> + enic->net_stats.tx_errors = stats->tx.tx_errors;
> + enic->net_stats.tx_dropped = stats->tx.tx_drops;
> +
> + enic->net_stats.rx_packets = stats->rx.rx_frames_ok;
> + enic->net_stats.rx_bytes = stats->rx.rx_bytes_ok;
> + enic->net_stats.rx_errors = stats->rx.rx_errors;
> + enic->net_stats.multicast = stats->rx.rx_multicast_frames_ok;
> + enic->net_stats.rx_crc_errors = stats->rx.rx_crc_errors;
> + enic->net_stats.rx_dropped = stats->rx.rx_no_bufs;
> +
> + return &enic->net_stats;
> +}
Why not use netdev->stats?
> +/* netif_tx_lock held, BHs disabled */
> +static void enic_set_multicast_list(struct net_device *netdev)
> +{
> + struct enic *enic = netdev_priv(netdev);
> + struct dev_mc_list *list = netdev->mc_list;
> + int directed = 1;
> + int multicast = (netdev->flags & IFF_MULTICAST) ? 1 : 0;
> + int broadcast = (netdev->flags & IFF_BROADCAST) ? 1 : 0;
> + int promisc = (netdev->flags & IFF_PROMISC) ? 1 : 0;
> + int allmulti = (netdev->flags & IFF_ALLMULTI) ||
> + (netdev->mc_count > ENIC_MULTICAST_PERFECT_FILTERS);
> + u8 mc_addr[ENIC_MULTICAST_PERFECT_FILTERS][ETH_ALEN];
> + unsigned int mc_count = netdev->mc_count;
> + unsigned int i, j;
> +
> + if (mc_count > ENIC_MULTICAST_PERFECT_FILTERS)
> + mc_count = ENIC_MULTICAST_PERFECT_FILTERS;
> +
> + spin_lock(&enic->devcmd_lock);
> +
> + vnic_dev_packet_filter(enic->vdev, directed,
> + multicast, broadcast, promisc, allmulti);
> +
> + /* Is there an easier way? Trying to minimize to
> + * calls to add/del multicast addrs. We keep the
> + * addrs from the last call in enic->mc_addr and
> + * look for changes to add/del.
> + */
If you sort the table then the following comparison can be done with a
single loop. But then you move the complexity to the sort itself. And
it doesn't really seem worth doing that for at most 32 addresses.
> +static int enic_get_skb_header(struct sk_buff *skb, void **iphdr,
> + void **tcph, u64 *hdr_flags, void *priv)
> +{
> + struct cq_enet_rq_desc *cq_desc = priv;
> + unsigned int ip_len;
> + struct iphdr *iph;
> +
> + u8 type, color, eop, sop, ingress_port, vlan_stripped;
> + u8 fcoe, fcoe_sof, fcoe_fc_crc_ok, fcoe_enc_error, fcoe_eof;
> + u8 tcp_udp_csum_ok, udp, tcp, ipv4_csum_ok;
> + u8 ipv6, ipv4, ipv4_fragment, fcs_ok, rss_type, csum_not_calc;
> + u8 packet_error;
> + u16 q_number, completed_index, bytes_written, vlan, checksum;
> + u32 rss_hash;
> +
> + cq_enet_rq_desc_dec(cq_desc,
> + &type, &color, &q_number, &completed_index,
> + &ingress_port, &fcoe, &eop, &sop, &rss_type,
> + &csum_not_calc, &rss_hash, &bytes_written,
> + &packet_error, &vlan_stripped, &vlan, &checksum,
> + &fcoe_sof, &fcoe_fc_crc_ok, &fcoe_enc_error,
> + &fcoe_eof, &tcp_udp_csum_ok, &udp, &tcp,
> + &ipv4_csum_ok, &ipv6, &ipv4, &ipv4_fragment,
> + &fcs_ok);
I don't know where that function is defined, but it's badly designed.
It will be easy to get the order of arguments wrong and the compiler
will probably not help because most of them have the same type. It is
probably better to define a structure to hold all these attributes.
> + if (!(ipv4 && tcp && !ipv4_fragment))
> + return -1;
> +
> + skb_reset_network_header(skb);
> + iph = ip_hdr(skb);
> +
> + ip_len = ip_hdrlen(skb);
> + skb_set_transport_header(skb, ip_len);
> +
> + /* check if ip header and tcp header are complete */
> + if (ntohs(iph->tot_len) < ip_len + tcp_hdrlen(skb))
> + return -1;
> +
> + *hdr_flags = LRO_IPV4 | LRO_TCP;
> + *tcph = tcp_hdr(skb);
> + *iphdr = iph;
> +
> + return 0;
> +}
[...]
> +static void enic_rq_indicate_buf(struct vnic_rq *rq,
> + struct cq_desc *cq_desc, struct vnic_rq_buf *buf,
> + int skipped, void *opaque)
> +{
> + struct enic *enic = vnic_dev_priv(rq->vdev);
> + struct sk_buff *skb;
> +
> + u8 type, color, eop, sop, ingress_port, vlan_stripped;
> + u8 fcoe, fcoe_sof, fcoe_fc_crc_ok, fcoe_enc_error, fcoe_eof;
> + u8 tcp_udp_csum_ok, udp, tcp, ipv4_csum_ok;
> + u8 ipv6, ipv4, ipv4_fragment, fcs_ok, rss_type, csum_not_calc;
> + u8 packet_error;
> + u16 q_number, completed_index, bytes_written, vlan, checksum;
> + u32 rss_hash;
> +
> + if (skipped)
> + return;
> +
> + skb = buf->os_buf;
> + prefetch(skb->data - NET_IP_ALIGN);
> + pci_unmap_single(enic->pdev, buf->dma_addr,
> + buf->len, PCI_DMA_FROMDEVICE);
> +
> + cq_enet_rq_desc_dec((struct cq_enet_rq_desc *)cq_desc,
> + &type, &color, &q_number, &completed_index,
> + &ingress_port, &fcoe, &eop, &sop, &rss_type,
> + &csum_not_calc, &rss_hash, &bytes_written,
> + &packet_error, &vlan_stripped, &vlan, &checksum,
> + &fcoe_sof, &fcoe_fc_crc_ok, &fcoe_enc_error,
> + &fcoe_eof, &tcp_udp_csum_ok, &udp, &tcp,
> + &ipv4_csum_ok, &ipv6, &ipv4, &ipv4_fragment,
> + &fcs_ok);
> +
> + if (packet_error) {
> +
> + if (bytes_written > 0 && !fcs_ok) {
> + if (net_ratelimit())
> + printk(KERN_ERR PFX
> + "%s: packet error: bad FCS\n",
> + enic->netdev->name);
> + }
> +
> + dev_kfree_skb_any(skb);
> +
> + return;
> + }
> +
> + if (eop && bytes_written > 0) {
> +
> + /* Good receive
> + */
> +
> + skb_put(skb, bytes_written);
> + skb->protocol = eth_type_trans(skb, enic->netdev);
> +
> + if (enic->csum_rx_enabled && !csum_not_calc) {
> + skb->csum = htons(checksum);
> + skb->ip_summed = CHECKSUM_COMPLETE;
> + }
> +
> + skb->dev = enic->netdev;
> + enic->netdev->last_rx = jiffies;
> +
> + if (enic->vlan_group && vlan_stripped) {
> +
> + if (ENIC_SETTING(enic, LRO))
> + lro_vlan_hwaccel_receive_skb(&enic->lro_mgr,
> + skb, enic->vlan_group,
> + vlan, cq_desc);
> + else
> + vlan_hwaccel_receive_skb(skb,
> + enic->vlan_group, vlan);
> +
> + } else {
> +
> + if (ENIC_SETTING(enic, LRO))
> + lro_receive_skb(&enic->lro_mgr, skb, cq_desc);
> + else
> + netif_receive_skb(skb);
> +
> + }
> +
> + } else {
> +
> + /* Buffer overflow
> + */
> +
> + dev_kfree_skb_any(skb);
> + }
Why so much vertical whitespace? It doesn't make this easier to read;
on the contrary, it's harder to see the whole function.
[...]
> +static void enic_notify_timer(unsigned long data)
> +{
> + struct enic *enic = (struct enic *)data;
> +
> + enic_notify_check(enic);
> +
> + mod_timer(&enic->notify_timer, round_jiffies(ENIC_NOTIFY_TIMER_PERIOD));
You want round_jiffies_relative() not round_jiffies().
> +}
[...]
> +static void enic_notify_timer_start(struct enic *enic)
> +{
> + switch (vnic_dev_get_intr_mode(enic->vdev)) {
> + case VNIC_DEV_INTR_MODE_MSI:
> + mod_timer(&enic->notify_timer, jiffies);
Do you really want this to run immediately?
> + break;
> + default:
> + /* Using intr for notification for INTx/MSI-X */
> + break;
> + };
> +}
[...]
> +static int enic_change_mtu(struct net_device *netdev, int new_mtu)
> +{
> + struct enic *enic = netdev_priv(netdev);
> + int running = netif_running(netdev);
> +
> + if (running)
> + enic_stop(netdev);
> +
> + if (new_mtu < ENIC_MIN_MTU)
> + new_mtu = ENIC_MIN_MTU;
> + if (new_mtu > ENIC_MAX_MTU)
> + new_mtu = ENIC_MAX_MTU;
Don't silently adjust an MTU that's out of the supported range; return
-EINVAL.
> + netdev->mtu = new_mtu;
> +
> + if (netdev->mtu > enic->port_mtu)
> + printk(KERN_WARNING PFX
> + "%s: interface MTU (%d) set higher "
> + "than port MTU (%d)\n",
> + netdev->name, netdev->mtu, enic->port_mtu);
> +
> + if (running)
> + enic_open(netdev);
> +
> + return 0;
> +}
[...]
> +static int enic_dev_soft_reset(struct enic *enic)
> +{
> + int err;
> +
> + err = enic_dev_wait(enic->vdev, vnic_dev_soft_reset,
> + vnic_dev_soft_reset_done, 0);
> + if (err)
> + printk(KERN_ERR PFX
> + "vNIC soft reset failed, err %d.\n", err);
> +
> + return err;
> +}
> +
> +static void enic_reset(struct work_struct *work)
> +{
> + struct enic *enic = container_of(work, struct enic, reset);
> +
> + if (!netif_running(enic->netdev))
> + return;
> +
> + rtnl_lock();
> +
> + spin_lock(&enic->devcmd_lock);
> + vnic_dev_hang_notify(enic->vdev);
> + spin_unlock(&enic->devcmd_lock);
> +
> + enic_stop(enic->netdev);
> + enic_dev_soft_reset(enic);
What if the soft reset fails?
> + enic_reset_mcaddrs(enic);
> + enic_init_vnic_resources(enic);
> + enic_open(enic->netdev);
> +
> + rtnl_unlock();
> +}
[...]
Ben.
--
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH 2/3] enic: add main netdev file with moduleinfrastructure
2008-09-12 19:47 ` Ben Hutchings
@ 2008-09-17 1:49 ` Scott Feldman
2008-09-17 11:47 ` Ben Hutchings
0 siblings, 1 reply; 7+ messages in thread
From: Scott Feldman @ 2008-09-17 1:49 UTC (permalink / raw)
To: Ben Hutchings; +Cc: netdev
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: TEXT/PLAIN; charset=X-UNKNOWN; format=flowed, Size: 5992 bytes --]
Thanks for the feedback Ben. Responses inline.
On Fri, 12 Sep 2008, Ben Hutchings wrote:
> On Mon, 2008-09-08 at 20:57 -0700, Scott Feldman wrote:
> [...]
>> diff -Naurp -X linux-2.6.26.3/Documentation/dontdiff linux-2.6.26.3/drivers/net/enic/enic_main.c linux-2.6.26.3-enic/drivers/net/enic/enic_main.c
>> --- linux-2.6.26.3/drivers/net/enic/enic_main.c 1969-12-31 16:00:00.000000000 -0800
>> +++ linux-2.6.26.3-enic/drivers/net/enic/enic_main.c 2008-09-08 15:16:23.000000000 -0700
> [...]
>> +static int enic_set_rx_csum(struct net_device *netdev, u32 data)
>> +{
>> + struct enic *enic = netdev_priv(netdev);
>> +
>> + enic->csum_rx_enabled =
>> + (data && ENIC_SETTING(enic, RXCSUM)) ? 1 : 0;
>> +
>> + return 0;
>> +}
>
> This should return an error code if it can't enable RX checksum:
>
> if (data && !ENIC_SETTING(enic, RXCSUM))
> return -EOPNOTSUPP;
> enic->csum_rx_enabled = !!data;
> return 0;
>
> (I'm not sure what the correct error code is in this situation though.)
Fixed
>> +static int enic_set_tx_csum(struct net_device *netdev, u32 data)
>> +{
>> + struct enic *enic = netdev_priv(netdev);
>> +
>> + if (data && ENIC_SETTING(enic, TXCSUM))
>> + netdev->features |= NETIF_F_HW_CSUM;
>> + else
>> + netdev->features &= ~NETIF_F_HW_CSUM;
>> +
>> + return 0;
>> +}
>
> Similarly here.
Fixed.
> [...]
>> +static struct ethtool_ops enic_ethtool_ops = {
>> + .get_settings = enic_get_settings,
>> + .get_drvinfo = enic_get_drvinfo,
>> + .get_msglevel = enic_get_msglevel,
>> + .set_msglevel = enic_set_msglevel,
>> + .get_link = ethtool_op_get_link,
>> + .get_strings = enic_get_strings,
>> + .get_stats_count = enic_get_stats_count,
>
> get_stats_count is deprecated; you should implement get_sset_count.
Not changed yet...need to figure out the backward compatibility story.
>> +/* dev_base_lock rwlock held, nominally process context */
>> +static struct net_device_stats *enic_get_stats(struct net_device *netdev)
>> +{
>> + struct enic *enic = netdev_priv(netdev);
>> + struct vnic_stats *stats;
>> +
>> + spin_lock(&enic->devcmd_lock);
>> + vnic_dev_stats_dump(enic->vdev, &stats);
>> + spin_unlock(&enic->devcmd_lock);
>> +
>> + enic->net_stats.tx_packets = stats->tx.tx_frames_ok;
>> + enic->net_stats.tx_bytes = stats->tx.tx_bytes_ok;
>> + enic->net_stats.tx_errors = stats->tx.tx_errors;
>> + enic->net_stats.tx_dropped = stats->tx.tx_drops;
>> +
>> + enic->net_stats.rx_packets = stats->rx.rx_frames_ok;
>> + enic->net_stats.rx_bytes = stats->rx.rx_bytes_ok;
>> + enic->net_stats.rx_errors = stats->rx.rx_errors;
>> + enic->net_stats.multicast = stats->rx.rx_multicast_frames_ok;
>> + enic->net_stats.rx_crc_errors = stats->rx.rx_crc_errors;
>> + enic->net_stats.rx_dropped = stats->rx.rx_no_bufs;
>> +
>> + return &enic->net_stats;
>> +}
>
> Why not use netdev->stats?
Another backward compatibility issue. I think I'll leave this one as-is.
>> +static int enic_get_skb_header(struct sk_buff *skb, void **iphdr,
>> + void **tcph, u64 *hdr_flags, void *priv)
>> +{
>> + struct cq_enet_rq_desc *cq_desc = priv;
>> + unsigned int ip_len;
>> + struct iphdr *iph;
>> +
>> + u8 type, color, eop, sop, ingress_port, vlan_stripped;
>> + u8 fcoe, fcoe_sof, fcoe_fc_crc_ok, fcoe_enc_error, fcoe_eof;
>> + u8 tcp_udp_csum_ok, udp, tcp, ipv4_csum_ok;
>> + u8 ipv6, ipv4, ipv4_fragment, fcs_ok, rss_type, csum_not_calc;
>> + u8 packet_error;
>> + u16 q_number, completed_index, bytes_written, vlan, checksum;
>> + u32 rss_hash;
>> +
>> + cq_enet_rq_desc_dec(cq_desc,
>> + &type, &color, &q_number, &completed_index,
>> + &ingress_port, &fcoe, &eop, &sop, &rss_type,
>> + &csum_not_calc, &rss_hash, &bytes_written,
>> + &packet_error, &vlan_stripped, &vlan, &checksum,
>> + &fcoe_sof, &fcoe_fc_crc_ok, &fcoe_enc_error,
>> + &fcoe_eof, &tcp_udp_csum_ok, &udp, &tcp,
>> + &ipv4_csum_ok, &ipv6, &ipv4, &ipv4_fragment,
>> + &fcs_ok);
>
> I don't know where that function is defined, but it's badly designed.
> It will be easy to get the order of arguments wrong and the compiler
> will probably not help because most of them have the same type. It is
> probably better to define a structure to hold all these attributes.
I know it's not ideal as-is but a structure overlay on the descriptor
doesn't work either because some of the fields are bit-fields and a
structure with bit-fields isn't that portable.
+ /* Buffer overflow
>> + */
>> +
>> + dev_kfree_skb_any(skb);
>> + }
>
> Why so much vertical whitespace? It doesn't make this easier to read;
> on the contrary, it's harder to see the whole function.
I'm snickering because I've been accused in the past of using too little
whitepace. It is the longest function and I didn't want to pack it to
help with readability. I guess I failed.
> [...]
>> +static void enic_notify_timer(unsigned long data)
>> +{
>> + struct enic *enic = (struct enic *)data;
>> +
>> + enic_notify_check(enic);
>> +
>> + mod_timer(&enic->notify_timer, round_jiffies(ENIC_NOTIFY_TIMER_PERIOD));
>
> You want ÿÿround_jiffies_relative() not round_jiffies().
No, I want round_jiffies().
>> + case VNIC_DEV_INTR_MODE_MSI:
>> + mod_timer(&enic->notify_timer, jiffies);
>
> Do you really want this to run immediately?
Yes.
>> +static int enic_change_mtu(struct net_device *netdev, int new_mtu)
>> +{
>> + struct enic *enic = netdev_priv(netdev);
>> + int running = netif_running(netdev);
>> +
>> + if (running)
>> + enic_stop(netdev);
>> +
>> + if (new_mtu < ENIC_MIN_MTU)
>> + new_mtu = ENIC_MIN_MTU;
>> + if (new_mtu > ENIC_MAX_MTU)
>> + new_mtu = ENIC_MAX_MTU;
>
> Don't silently adjust an MTU that's out of the supported range; return
> -EINVAL.
Fixed.
>> + if (!netif_running(enic->netdev))
>> + return;
>> +
>> + rtnl_lock();
>> +
>> + spin_lock(&enic->devcmd_lock);
>> + vnic_dev_hang_notify(enic->vdev);
>> + spin_unlock(&enic->devcmd_lock);
>> +
>> + enic_stop(enic->netdev);
>> + enic_dev_soft_reset(enic);
>
> What if the soft reset fails?
If soft_reset fails, it's pretty much game over.
-scott
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH 2/3] enic: add main netdev file with moduleinfrastructure
2008-09-17 1:49 ` [PATCH 2/3] enic: add main netdev file with moduleinfrastructure Scott Feldman
@ 2008-09-17 11:47 ` Ben Hutchings
2008-09-17 18:51 ` [PATCH 2/3] enic: add main netdev file withmoduleinfrastructure Scott Feldman
0 siblings, 1 reply; 7+ messages in thread
From: Ben Hutchings @ 2008-09-17 11:47 UTC (permalink / raw)
To: Scott Feldman; +Cc: netdev
On Tue, 2008-09-16 at 18:49 -0700, Scott Feldman wrote:
[...]
> >> +static int enic_get_skb_header(struct sk_buff *skb, void **iphdr,
> >> + void **tcph, u64 *hdr_flags, void *priv)
> >> +{
> >> + struct cq_enet_rq_desc *cq_desc = priv;
> >> + unsigned int ip_len;
> >> + struct iphdr *iph;
> >> +
> >> + u8 type, color, eop, sop, ingress_port, vlan_stripped;
> >> + u8 fcoe, fcoe_sof, fcoe_fc_crc_ok, fcoe_enc_error, fcoe_eof;
> >> + u8 tcp_udp_csum_ok, udp, tcp, ipv4_csum_ok;
> >> + u8 ipv6, ipv4, ipv4_fragment, fcs_ok, rss_type, csum_not_calc;
> >> + u8 packet_error;
> >> + u16 q_number, completed_index, bytes_written, vlan, checksum;
> >> + u32 rss_hash;
> >> +
> >> + cq_enet_rq_desc_dec(cq_desc,
> >> + &type, &color, &q_number, &completed_index,
> >> + &ingress_port, &fcoe, &eop, &sop, &rss_type,
> >> + &csum_not_calc, &rss_hash, &bytes_written,
> >> + &packet_error, &vlan_stripped, &vlan, &checksum,
> >> + &fcoe_sof, &fcoe_fc_crc_ok, &fcoe_enc_error,
> >> + &fcoe_eof, &tcp_udp_csum_ok, &udp, &tcp,
> >> + &ipv4_csum_ok, &ipv6, &ipv4, &ipv4_fragment,
> >> + &fcs_ok);
> >
> > I don't know where that function is defined, but it's badly designed.
> > It will be easy to get the order of arguments wrong and the compiler
> > will probably not help because most of them have the same type. It is
> > probably better to define a structure to hold all these attributes.
>
> I know it's not ideal as-is but a structure overlay on the descriptor
> doesn't work either because some of the fields are bit-fields and a
> structure with bit-fields isn't that portable.
I wasn't suggesting a structure overlay as I'm aware that's a recipe for
portability problems, but something like:
struct cq_enet_rq_desc {
u8 type, color, eop, sop, ...;
u16 q_number, completed_index, ...;
u32 rss_hash;
};
struct cq_enet_rq_desc desc_fields;
cq_enet_rq_desc_dec(cq_desc, &desc_fields);
> > [...]
> >> +static void enic_notify_timer(unsigned long data)
> >> +{
> >> + struct enic *enic = (struct enic *)data;
> >> +
> >> + enic_notify_check(enic);
> >> +
> >> + mod_timer(&enic->notify_timer, round_jiffies(ENIC_NOTIFY_TIMER_PERIOD));
> >
> > You want round_jiffies_relative() not round_jiffies().
>
> No, I want round_jiffies().
[...]
The kernel-doc says:
"round_jiffies() rounds an absolute time in the future (in jiffies)
up or down to (approximately) full seconds. This is useful for timers
for which the exact time they fire does not matter too much, as long as
they fire approximately every X seconds."
You're passing in ENIC_NOTIFY_TIMER_PERIOD which is a relative time.
Ben.
--
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH 2/3] enic: add main netdev file withmoduleinfrastructure
2008-09-17 11:47 ` Ben Hutchings
@ 2008-09-17 18:51 ` Scott Feldman
2008-09-17 19:03 ` Ben Hutchings
0 siblings, 1 reply; 7+ messages in thread
From: Scott Feldman @ 2008-09-17 18:51 UTC (permalink / raw)
To: Ben Hutchings; +Cc: Scott Feldman, netdev
On Wed, 17 Sep 2008, Ben Hutchings wrote:
> On Tue, 2008-09-16 at 18:49 -0700, Scott Feldman wrote:
>>>> +static void enic_notify_timer(unsigned long data)
>>>> +{
>>>> + struct enic *enic = (struct enic *)data;
>>>> +
>>>> + enic_notify_check(enic);
>>>> +
>>>> + mod_timer(&enic->notify_timer, round_jiffies(ENIC_NOTIFY_TIMER_PERIOD));
>>>
>>> You want round_jiffies_relative() not round_jiffies().
>>
>> No, I want round_jiffies().
> [...]
>
> The kernel-doc says:
> "round_jiffies() rounds an absolute time in the future (in jiffies)
> up or down to (approximately) full seconds. This is useful for timers
> for which the exact time they fire does not matter too much, as long as
> they fire approximately every X seconds."
>
> You're passing in ENIC_NOTIFY_TIMER_PERIOD which is a relative time.
I tried both and round_jiffies(2) gave the desired result of calling the timer
every two seconds. Using round_jiffies_relative(2) called the timer
continuously.
-scott
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH 2/3] enic: add main netdev file withmoduleinfrastructure
2008-09-17 18:51 ` [PATCH 2/3] enic: add main netdev file withmoduleinfrastructure Scott Feldman
@ 2008-09-17 19:03 ` Ben Hutchings
2008-09-17 19:17 ` Scott Feldman
0 siblings, 1 reply; 7+ messages in thread
From: Ben Hutchings @ 2008-09-17 19:03 UTC (permalink / raw)
To: Scott Feldman; +Cc: netdev
On Wed, 2008-09-17 at 11:51 -0700, Scott Feldman wrote:
> On Wed, 17 Sep 2008, Ben Hutchings wrote:
>
> > On Tue, 2008-09-16 at 18:49 -0700, Scott Feldman wrote:
> >>>> +static void enic_notify_timer(unsigned long data)
> >>>> +{
> >>>> + struct enic *enic = (struct enic *)data;
> >>>> +
> >>>> + enic_notify_check(enic);
> >>>> +
> >>>> + mod_timer(&enic->notify_timer, round_jiffies(ENIC_NOTIFY_TIMER_PERIOD));
> >>>
> >>> You want round_jiffies_relative() not round_jiffies().
> >>
> >> No, I want round_jiffies().
> > [...]
> >
> > The kernel-doc says:
> > "round_jiffies() rounds an absolute time in the future (in jiffies)
> > up or down to (approximately) full seconds. This is useful for timers
> > for which the exact time they fire does not matter too much, as long as
> > they fire approximately every X seconds."
> >
> > You're passing in ENIC_NOTIFY_TIMER_PERIOD which is a relative time.
>
> I tried both and round_jiffies(2) gave the desired result of calling the timer
> every two seconds. Using round_jiffies_relative(2) called the timer
> continuously.
Sorry, yes, round_jiffies_relative() not only takes but also returns a
relative time - which if interpreted as absolute is always in the past.
So calling round_jiffies() is right, but the argument should be jiffies
+ ENIC_NOTIFY_TIMER_PERIOD.
Ben.
--
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH 2/3] enic: add main netdev file withmoduleinfrastructure
2008-09-17 19:03 ` Ben Hutchings
@ 2008-09-17 19:17 ` Scott Feldman
0 siblings, 0 replies; 7+ messages in thread
From: Scott Feldman @ 2008-09-17 19:17 UTC (permalink / raw)
To: Ben Hutchings; +Cc: Scott Feldman, netdev
On Wed, 17 Sep 2008, Ben Hutchings wrote:
> On Wed, 2008-09-17 at 11:51 -0700, Scott Feldman wrote:
>> On Wed, 17 Sep 2008, Ben Hutchings wrote:
>>
>>> On Tue, 2008-09-16 at 18:49 -0700, Scott Feldman wrote:
>>>>>> +static void enic_notify_timer(unsigned long data)
>>>>>> +{
>>>>>> + struct enic *enic = (struct enic *)data;
>>>>>> +
>>>>>> + enic_notify_check(enic);
>>>>>> +
>>>>>> + mod_timer(&enic->notify_timer, round_jiffies(ENIC_NOTIFY_TIMER_PERIOD));
>>>>>
>>>>> You want round_jiffies_relative() not round_jiffies().
>>>>
>>>> No, I want round_jiffies().
>>> [...]
>>>
>>> The kernel-doc says:
>>> "round_jiffies() rounds an absolute time in the future (in jiffies)
>>> up or down to (approximately) full seconds. This is useful for timers
>>> for which the exact time they fire does not matter too much, as long as
>>> they fire approximately every X seconds."
>>>
>>> You're passing in ENIC_NOTIFY_TIMER_PERIOD which is a relative time.
>>
>> I tried both and round_jiffies(2) gave the desired result of calling the timer
>> every two seconds. Using round_jiffies_relative(2) called the timer
>> continuously.
>
> Sorry, yes, round_jiffies_relative() not only takes but also returns a
> relative time - which if interpreted as absolute is always in the past.
> So calling round_jiffies() is right, but the argument should be jiffies
> + ENIC_NOTIFY_TIMER_PERIOD.
Indeed, I'll make that change to add in jiffies.
-scott
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2008-09-17 19:17 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-09-09 3:57 [PATCH 2/3] enic: add main netdev file with module infrastructure Scott Feldman
2008-09-12 19:47 ` Ben Hutchings
2008-09-17 1:49 ` [PATCH 2/3] enic: add main netdev file with moduleinfrastructure Scott Feldman
2008-09-17 11:47 ` Ben Hutchings
2008-09-17 18:51 ` [PATCH 2/3] enic: add main netdev file withmoduleinfrastructure Scott Feldman
2008-09-17 19:03 ` Ben Hutchings
2008-09-17 19:17 ` Scott Feldman
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).