From: Tomas Winkler <tomas.winkler@intel.com>
To: linville@tuxdriver.com, yi.zhu@intel.com, johannes@sipsolutions.net
Cc: linux-wireless@vger.kernel.org, Ben Cahill <ben.m.cahill@intel.com>
Subject: [PATCH 23/34] iwlwifi: document Rx registers
Date: Mon, 3 Dec 2007 13:43:35 +0200 [thread overview]
Message-ID: <11966822312882-git-send-email-tomas.winkler@intel.com> (raw)
In-Reply-To: <11966822311945-git-send-email-tomas.winkler@intel.com>
From: Ben Cahill <ben.m.cahill@intel.com>
Document Rx registers
Document Tx circular buffer base registers
Signed-off-by: Ben Cahill <ben.m.cahill@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
---
drivers/net/wireless/iwlwifi/iwl-4965-hw.h | 170 +++++++++++++++++++++++++++-
1 files changed, 164 insertions(+), 6 deletions(-)
diff --git a/drivers/net/wireless/iwlwifi/iwl-4965-hw.h b/drivers/net/wireless/iwlwifi/iwl-4965-hw.h
index cc348de..ab88c72 100644
--- a/drivers/net/wireless/iwlwifi/iwl-4965-hw.h
+++ b/drivers/net/wireless/iwlwifi/iwl-4965-hw.h
@@ -1286,7 +1286,7 @@ enum {
/* Flow Handler Definitions */
/****************************/
-/*
+/**
* This I/O area is directly read/writable by driver (e.g. Linux uses writel())
* Addresses are offsets from device's PCI hardware base address.
*/
@@ -1313,35 +1313,193 @@ enum {
#define IWL_FH_KW_MEM_ADDR_REG (FH_MEM_LOWER_BOUND + 0x97C)
-/* CBBC Area - Circular buffers base address cache pointers table */
+/**
+ * TFD Circular Buffers Base (CBBC) addresses
+ *
+ * 4965 has 16 base pointer registers, one for each of 16 host-DRAM-resident
+ * circular buffers (CBs/queues) containing Transmit Frame Descriptors (TFDs)
+ * (see struct iwl_tfd_frame). These 16 pointer registers are offset by 0x04
+ * bytes from one another. Each TFD circular buffer in DRAM must be 256-byte
+ * aligned (address bits 0-7 must be 0).
+ *
+ * Bit fields in each pointer register:
+ * 27-0: TFD CB physical base address [35:8], must be 256-byte aligned
+ */
#define FH_MEM_CBBC_LOWER_BOUND (FH_MEM_LOWER_BOUND + 0x9D0)
#define FH_MEM_CBBC_UPPER_BOUND (FH_MEM_LOWER_BOUND + 0xA10)
-/* queues 0 - 15 */
+
+/* Find TFD CB base pointer for given queue (range 0-15). */
#define FH_MEM_CBBC_QUEUE(x) (FH_MEM_CBBC_LOWER_BOUND + (x) * 0x4)
-/* RSCSR Area */
+
+/**
+ * Rx SRAM Control and Status Registers (RSCSR)
+ *
+ * These registers provide handshake between driver and 4965 for the Rx queue
+ * (this queue handles *all* command responses, notifications, Rx data, etc.
+ * sent from 4965 uCode to host driver). Unlike Tx, there is only one Rx
+ * queue, and only one Rx DMA/FIFO channel. Also unlike Tx, which can
+ * concatenate up to 20 DRAM buffers to form a Tx frame, each Receive Buffer
+ * Descriptor (RBD) points to only one Rx Buffer (RB); there is a 1:1
+ * mapping between RBDs and RBs.
+ *
+ * Driver must allocate host DRAM memory for the following, and set the
+ * physical address of each into 4965 registers:
+ *
+ * 1) Receive Buffer Descriptor (RBD) circular buffer (CB), typically with 256
+ * entries (although any power of 2, up to 4096, is selectable by driver).
+ * Each entry (1 dword) points to a receive buffer (RB) of consistent size
+ * (typically 4K, although 8K or 16K are also selectable by driver).
+ * Driver sets up RB size and number of RBDs in the CB via Rx config
+ * register FH_MEM_RCSR_CHNL0_CONFIG_REG.
+ *
+ * Bit fields within one RBD:
+ * 27-0: Receive Buffer physical address bits [35:8], 256-byte aligned
+ *
+ * Driver sets physical address [35:8] of base of RBD circular buffer
+ * into FH_RSCSR_CHNL0_RBDCB_BASE_REG [27:0].
+ *
+ * 2) Rx status buffer, 8 bytes, in which 4965 indicates which Rx Buffers
+ * (RBs) have been filled, via a "write pointer", actually the index of
+ * the RB's corresponding RBD within the circular buffer. Driver sets
+ * physical address [35:4] into FH_RSCSR_CHNL0_STTS_WPTR_REG [31:0].
+ *
+ * Bit fields in lower dword of Rx status buffer (upper dword not used
+ * by driver; see struct iwl4965_shared, val0):
+ * 31-12: Not used by driver
+ * 11- 0: Index of last filled Rx buffer descriptor
+ * (4965 writes, driver reads this value)
+ *
+ * As the driver prepares Receive Buffers (RBs) for 4965 to fill, driver must
+ * enter pointers to these RBs into contiguous RBD circular buffer entries,
+ * and update the 4965's "write" index register, FH_RSCSR_CHNL0_RBDCB_WPTR_REG.
+ *
+ * This "write" index corresponds to the *next* RBD that the driver will make
+ * available, i.e. one RBD past the tail of the ready-to-fill RBDs within
+ * the circular buffer. This value should initially be 0 (before preparing any
+ * RBs), should be 8 after preparing the first 8 RBs (for example), and must
+ * wrap back to 0 at the end of the circular buffer (but don't wrap before
+ * "read" index has advanced past 1! See below).
+ * NOTE: 4965 EXPECTS THE WRITE INDEX TO BE INCREMENTED IN MULTIPLES OF 8.
+ *
+ * As the 4965 fills RBs (referenced from contiguous RBDs within the circular
+ * buffer), it updates the Rx status buffer in host DRAM, 2) described above,
+ * to tell the driver the index of the latest filled RBD. The driver must
+ * read this "read" index from DRAM after receiving an Rx interrupt from 4965.
+ *
+ * The driver must also internally keep track of a third index, which is the
+ * next RBD to process. When receiving an Rx interrupt, driver should process
+ * all filled but unprocessed RBs up to, but not including, the RB
+ * corresponding to the "read" index. For example, if "read" index becomes "1",
+ * driver may process the RB pointed to by RBD 0. Depending on volume of
+ * traffic, there may be many RBs to process.
+ *
+ * If read index == write index, 4965 thinks there is no room to put new data.
+ * Due to this, the maximum number of filled RBs is 255, instead of 256. To
+ * be safe, make sure that there is a gap of at least 2 RBDs between "write"
+ * and "read" indexes; that is, make sure that there are no more than 254
+ * buffers waiting to be filled.
+ */
#define FH_MEM_RSCSR_LOWER_BOUND (FH_MEM_LOWER_BOUND + 0xBC0)
#define FH_MEM_RSCSR_UPPER_BOUND (FH_MEM_LOWER_BOUND + 0xC00)
#define FH_MEM_RSCSR_CHNL0 (FH_MEM_RSCSR_LOWER_BOUND)
+/**
+ * Physical base address of 8-byte Rx Status buffer.
+ * Bit fields:
+ * 31-0: Rx status buffer physical base address [35:4], must 16-byte aligned.
+ */
#define FH_RSCSR_CHNL0_STTS_WPTR_REG (FH_MEM_RSCSR_CHNL0)
+
+/**
+ * Physical base address of Rx Buffer Descriptor Circular Buffer.
+ * Bit fields:
+ * 27-0: RBD CD physical base address [35:8], must be 256-byte aligned.
+ */
#define FH_RSCSR_CHNL0_RBDCB_BASE_REG (FH_MEM_RSCSR_CHNL0 + 0x004)
+
+/**
+ * Rx write pointer (index, really!).
+ * Bit fields:
+ * 11-0: Index of driver's most recent prepared-to-be-filled RBD, + 1.
+ * NOTE: For 256-entry circular buffer, use only bits [7:0].
+ */
#define FH_RSCSR_CHNL0_RBDCB_WPTR_REG (FH_MEM_RSCSR_CHNL0 + 0x008)
-/* RCSR Area - Registers address map */
+
+/**
+ * Rx Config/Status Registers (RCSR)
+ * Rx Config Reg for channel 0 (only channel used)
+ *
+ * Driver must initialize FH_MEM_RCSR_CHNL0_CONFIG_REG as follows for
+ * normal operation (see bit fields).
+ *
+ * Clearing FH_MEM_RCSR_CHNL0_CONFIG_REG to 0 turns off Rx DMA.
+ * Driver should poll FH_MEM_RSSR_RX_STATUS_REG for
+ * FH_RSSR_CHNL0_RX_STATUS_CHNL_IDLE (bit 24) before continuing.
+ *
+ * Bit fields:
+ * 31-30: Rx DMA channel enable: '00' off/pause, '01' pause at end of frame,
+ * '10' operate normally
+ * 29-24: reserved
+ * 23-20: # RBDs in circular buffer = 2^value; use "8" for 256 RBDs (normal),
+ * min "5" for 32 RBDs, max "12" for 4096 RBDs.
+ * 19-18: reserved
+ * 17-16: size of each receive buffer; '00' 4K (normal), '01' 8K,
+ * '10' 12K, '11' 16K.
+ * 15-14: reserved
+ * 13-12: IRQ destination; '00' none, '01' host driver (normal operation)
+ * 11- 4: timeout for closing Rx buffer and interrupting host (units 32 usec)
+ * typical value 0x10 (about 1/2 msec)
+ * 3- 0: reserved
+ */
#define FH_MEM_RCSR_LOWER_BOUND (FH_MEM_LOWER_BOUND + 0xC00)
#define FH_MEM_RCSR_UPPER_BOUND (FH_MEM_LOWER_BOUND + 0xCC0)
#define FH_MEM_RCSR_CHNL0 (FH_MEM_RCSR_LOWER_BOUND)
#define FH_MEM_RCSR_CHNL0_CONFIG_REG (FH_MEM_RCSR_CHNL0)
-/* RSSR Area - Rx shared ctrl & status registers */
+#define FH_RCSR_CHNL0_RX_CONFIG_RB_TIMEOUT_MASK (0x00000FF0) /* bit 4-11 */
+#define FH_RCSR_CHNL0_RX_CONFIG_IRQ_DEST_MASK (0x00001000) /* bit 12 */
+#define FH_RCSR_CHNL0_RX_CONFIG_SINGLE_FRAME_MASK (0x00008000) /* bit 15 */
+#define FH_RCSR_CHNL0_RX_CONFIG_RB_SIZE_MASK (0x00030000) /* bits 16-17 */
+#define FH_RCSR_CHNL0_RX_CONFIG_RBDBC_SIZE_MASK (0x00F00000) /* bits 20-23 */
+#define FH_RCSR_CHNL0_RX_CONFIG_DMA_CHNL_EN_MASK (0xC0000000) /* bits 30-31 */
+
+#define FH_RCSR_RX_CONFIG_RBDCB_SIZE_BITSHIFT (20)
+#define FH_RCSR_RX_CONFIG_RB_SIZE_BITSHIFT (16)
+
+#define FH_RCSR_RX_CONFIG_CHNL_EN_PAUSE_VAL (0x00000000)
+#define FH_RCSR_RX_CONFIG_CHNL_EN_PAUSE_EOF_VAL (0x40000000)
+#define FH_RCSR_RX_CONFIG_CHNL_EN_ENABLE_VAL (0x80000000)
+
+#define IWL_FH_RCSR_RX_CONFIG_REG_VAL_RB_SIZE_4K (0x00000000)
+
+#define FH_RCSR_CHNL0_RX_CONFIG_IRQ_DEST_NO_INT_VAL (0x00000000)
+#define FH_RCSR_CHNL0_RX_CONFIG_IRQ_DEST_INT_HOST_VAL (0x00001000)
+
+
+/**
+ * Rx Shared Status Registers (RSSR)
+ *
+ * After stopping Rx DMA channel (writing 0 to FH_MEM_RCSR_CHNL0_CONFIG_REG),
+ * driver must poll FH_MEM_RSSR_RX_STATUS_REG until Rx channel is idle.
+ *
+ * Bit fields:
+ * 24: 1 = Channel 0 is idle
+ *
+ * FH_MEM_RSSR_SHARED_CTRL_REG and FH_MEM_RSSR_RX_ENABLE_ERR_IRQ2DRV contain
+ * default values that should not be altered by the driver.
+ */
#define FH_MEM_RSSR_LOWER_BOUND (FH_MEM_LOWER_BOUND + 0xC40)
#define FH_MEM_RSSR_UPPER_BOUND (FH_MEM_LOWER_BOUND + 0xD00)
+
#define FH_MEM_RSSR_SHARED_CTRL_REG (FH_MEM_RSSR_LOWER_BOUND)
#define FH_MEM_RSSR_RX_STATUS_REG (FH_MEM_RSSR_LOWER_BOUND + 0x004)
#define FH_MEM_RSSR_RX_ENABLE_ERR_IRQ2DRV (FH_MEM_RSSR_LOWER_BOUND + 0x008)
+#define FH_RSSR_CHNL0_RX_STATUS_CHNL_IDLE (0x01000000)
+
/* TCSR */
#define IWL_FH_TCSR_LOWER_BOUND (FH_MEM_LOWER_BOUND + 0xD00)
#define IWL_FH_TCSR_UPPER_BOUND (FH_MEM_LOWER_BOUND + 0xE60)
--
1.5.2.2
---------------------------------------------------------------------
Intel Israel (74) Limited
This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.
next prev parent reply other threads:[~2007-12-03 11:48 UTC|newest]
Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-12-03 11:43 [PATCH 0/34 V2] iwlwifi (version 1.1.22) Tomas Winkler
2007-12-03 11:43 ` [PATCH 01/34] iwlwifi: Support for uCode without init and bsm section Tomas Winkler
2007-12-03 11:43 ` [PATCH 02/34] iwlwifi: continue namespace changes - fix CONFIG variables Tomas Winkler
2007-12-03 11:43 ` [PATCH 03/34] iwl3945: re-align 3945 event log data Tomas Winkler
2007-12-03 11:43 ` [PATCH 04/34] iwl4965: add comments to rate scaling code Tomas Winkler
2007-12-03 11:43 ` [PATCH 05/34] iwlwifi: add comments to EEPROM stuff Tomas Winkler
2007-12-03 11:43 ` [PATCH 06/34] iwl-4965-hw.h: clean up unused eeprom structures and definitions Tomas Winkler
2007-12-03 11:43 ` [PATCH 07/34] iwlwifi: clean up and clarify some comments after 3945/4965 split Tomas Winkler
2007-12-03 11:43 ` [PATCH 08/34] iwlwifi: Move is_legacy() macro family from iwl-4965-hw.h to iwl-4965-rs.h Tomas Winkler
2007-12-03 11:43 ` [PATCH 09/34] iwlwifi: Add comments to some driver data structures Tomas Winkler
2007-12-03 11:43 ` [PATCH 10/34] iwlwifi: Document 4965 rate_n_flags bits Tomas Winkler
2007-12-03 11:43 ` [PATCH 11/34] iwlwifi: Document Rx calibration Tomas Winkler
2007-12-03 11:43 ` [PATCH 12/34] iwlwifi: Partially clean-up, add comments to iwl-*-hw.h Tomas Winkler
2007-12-03 11:43 ` [PATCH 13/34] iwlwifi: clean up some unused definitions in iwl-4965.h and iwl-3945.h Tomas Winkler
2007-12-03 11:43 ` [PATCH 14/34] iwlwifi: add comments, mostly on Tx queues Tomas Winkler
2007-12-03 11:43 ` [PATCH 15/34] iwlwifi: add comments to iwl*-base.c Tomas Winkler
2007-12-03 11:43 ` [PATCH 16/34] iwlwifi: Clean up unused definitions in iwl-3945-hw.h Tomas Winkler
2007-12-03 11:43 ` [PATCH 17/34] iwlwifi: clean up unused definitions in iwl-4965-hw.h Tomas Winkler
2007-12-03 11:43 ` [PATCH 18/34] iwlwifi: move uCode API definitions to iwl-4965-commands.h Tomas Winkler
2007-12-03 11:43 ` [PATCH 19/34] iwlwifi: move HT_IE_EXT_CHANNEL_* driver definitions to iwl-4965.h Tomas Winkler
2007-12-03 11:43 ` [PATCH 20/34] iwlwifi: document temperature calculation Tomas Winkler
2007-12-03 11:43 ` [PATCH 21/34] iwlwifi: document txpower calculations Tomas Winkler
2007-12-03 11:43 ` [PATCH 22/34] iwlwifi: document keep-warm buffer Tomas Winkler
2007-12-03 11:43 ` Tomas Winkler [this message]
2007-12-03 11:43 ` [PATCH 24/34] iwlwifi: document shared Tx structures Tomas Winkler
2007-12-03 11:43 ` [PATCH 25/34] iwlwifi: document 4965 Tx scheduler Tomas Winkler
2007-12-03 11:43 ` [PATCH 26/34] iwlwifi: document command header and "alive" responses Tomas Winkler
2007-12-03 11:43 ` [PATCH 27/34] iwlwifi: add comments to RXON command and txpower formats Tomas Winkler
2007-12-03 11:43 ` [PATCH 28/34] iwlwifi: add comments to QOS and ADD_STA commands Tomas Winkler
2007-12-03 11:43 ` [PATCH 29/34] iwlwifi: add comments to Tx commands Tomas Winkler
2007-12-03 11:43 ` [PATCH 30/34] iwlwifi: document 4965 rate scaling Tomas Winkler
2007-12-03 11:43 ` [PATCH 31/34] iwlwifi: enhance WPA authenication stability Tomas Winkler
2007-12-03 11:43 ` [PATCH 32/34] iwlwifi: fix ucode assertion for RX queue overrun Tomas Winkler
2007-12-03 11:43 ` [PATCH 33/34] iwlwifi: avoid firmware command sending if rfkill is enabled Tomas Winkler
2007-12-03 11:43 ` [PATCH 34/34] iwlwifi: update version number to 1.2.22 Tomas Winkler
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=11966822312882-git-send-email-tomas.winkler@intel.com \
--to=tomas.winkler@intel.com \
--cc=ben.m.cahill@intel.com \
--cc=johannes@sipsolutions.net \
--cc=linux-wireless@vger.kernel.org \
--cc=linville@tuxdriver.com \
--cc=yi.zhu@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).