Skip to contentSkip to conversation

MMC:mmc_spi: Recover from CRC error for SD read/write operation over SPI.

PATCH In review sonic zhang 4 participants 7 messages Jul 12, 2010 – Oct 16, 2010

Subsystems (2)
MMC, SD and SDIO Multimedia card (MMC) etc. over SPI
lore
Share & tools
lore

sonic zhang

author
#0 lore raw

PATCH v2 1/1 MMC:mmc_spi: Recover from CRC error for SD read/write operation over SPI.

Quoted 1 line

From 94d1ea57ac2fbe37133e25e99ccb265250f0027d Mon Sep 17 00:00:00 2001

From: Sonic Zhang <[email protected]>
Date: Fri, 9 Jul 2010 15:36:20 +0800
Subject: [PATCH v2] MMC:mmc_spi: Recover from crc error for SD read/write operation over SPI.

SPI bus is not reliable on all platforms when doing large data transfer. Current mmc spi driver fails SD read/write command immediately, if occasional CRC error is reported by SD device. This patch makes the operation recover from the CRC error by doing last SD command again. The retry count is set to 5 to ensure the driver pass any stress test.

Signed-off-by Sonic Zhang <[email protected]>
---
 drivers/mmc/host/mmc_spi.c |   22 ++++++++++++++++++++++
 1 files changed, 22 insertions(+), 0 deletions(-)
drivers/mmc/host/mmc_spi.c +22 -0
index ad847a2..8192434 100644
@@ -1055,6 +1055,8 @@ static void mmc_spi_request(struct mmc_host *mmc, struct mmc_request *mrq)
 {
 	struct mmc_spi_host	*host = mmc_priv(mmc);
 	int			status = -EINVAL;
+	int			crc_retry = 5;
+	struct mmc_command	stop;
 
 #ifdef DEBUG
 	/* MMC core and layered drivers *MUST* issue SPI-aware commands */

@@ -1084,10 +1086,30 @@ static void mmc_spi_request(struct mmc_host *mmc, struct mmc_request *mrq)
 	}
 #endif
 
+crc_recover:
 	/* issue command; then optionally data and stop */
 	status = mmc_spi_command_send(host, mrq, mrq->cmd, mrq->data != NULL);
 	if (status == 0 && mrq->data) {
 		mmc_spi_data_do(host, mrq->cmd, mrq->data, mrq->data->blksz);
+ 
+ 		/*
+ 		 * SPI bus is not reliable when doing large data transfer.
+ 		 * If occasional crc error is reported by SD device when do
+ 		 * data read/write over SPI, it can be recovered by doing
+ 		 * last SD command again. The retry count is set to 5 to
+ 		 * ensure the driver pass any stress test.
+ 		 */
+ 		if (mrq->data->error == -EILSEQ && crc_retry) {
+ 			stop.opcode = MMC_STOP_TRANSMISSION;
+ 			stop.arg = 0;
+ 			stop.flags = MMC_RSP_SPI_R1B | MMC_RSP_R1B | MMC_CMD_AC;
+ 
+ 			status = mmc_spi_command_send(host, mrq, &stop, 0);
+ 			crc_retry--;
+ 			mrq->data->error = 0;
+ 			goto crc_recover;
+ 		}
+ 
 		if (mrq->stop)
 			status = mmc_spi_command_send(host, mrq, mrq->stop, 0);
 		else
Signature · 1 line
1.6.0
author
#1 lore raw

Any comments?

Sonic

Quoted 70 lines · sonic zhang
Quoted 1 line

From 94d1ea57ac2fbe37133e25e99ccb265250f0027d Mon Sep 17 00:00:00 2001

From: Sonic Zhang <[email protected]>
Date: Fri, 9 Jul 2010 15:36:20 +0800
Subject: [PATCH v2] MMC:mmc_spi: Recover from crc error for SD read/write operation over SPI.

SPI bus is not reliable on all platforms when doing large data transfer. Current mmc spi driver fails SD read/write command immediately, if occasional CRC error is reported by SD device. This patch makes the operation recover from the CRC error by doing last SD command again. The retry count is set to 5 to ensure the driver pass any stress test.

Signed-off-by Sonic Zhang <[email protected]>

---
 drivers/mmc/host/mmc_spi.c |   22 ++++++++++++++++++++++

 1 files changed, 22 insertions(+), 0 deletions(-)
drivers/mmc/host/mmc_spi.c +22 -0
index ad847a2..8192434 100644
@@ -1055,6 +1055,8 @@ static void mmc_spi_request(struct mmc_host *mmc, struct mmc_request *mrq)
  {
        struct mmc_spi_host     *host = mmc_priv(mmc);
        int                     status = -EINVAL;
+       int                     crc_retry = 5;
+       struct mmc_command      stop;
 
  #ifdef DEBUG
        /* MMC core and layered drivers *MUST* issue SPI-aware commands */

@@ -1084,10 +1086,30 @@ static void mmc_spi_request(struct mmc_host *mmc, struct mmc_request *mrq)
        }
  #endif
 
+crc_recover:
        /* issue command; then optionally data and stop */
        status = mmc_spi_command_send(host, mrq, mrq->cmd, mrq->data != NULL);
        if (status == 0 && mrq->data) {
                mmc_spi_data_do(host, mrq->cmd, mrq->data, mrq->data->blksz);
+
+               /*
+                * SPI bus is not reliable when doing large data transfer.
+                * If occasional crc error is reported by SD device when do
+                * data read/write over SPI, it can be recovered by doing
+                * last SD command again. The retry count is set to 5 to
+                * ensure the driver pass any stress test.
+                */
+               if (mrq->data->error == -EILSEQ && crc_retry) {
+                       stop.opcode = MMC_STOP_TRANSMISSION;
+                       stop.arg = 0;
+                       stop.flags = MMC_RSP_SPI_R1B | MMC_RSP_R1B | MMC_CMD_AC;
+
+                       status = mmc_spi_command_send(host, mrq, &stop, 0);
+                       crc_retry--;
+                       mrq->data->error = 0;
+                       goto crc_recover;
+               }
+
                if (mrq->stop)
                        status = mmc_spi_command_send(host, mrq, mrq->stop, 0);

               else
--
1.6.0

--
To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

#2 lore raw
Quoted 1 line · Sonic Zhang

Any comments?

might want to CC akpm so he'll pick this up in case the mmc guys miss it -mike

#3 lore raw

PATCH RESEND 1/1 MMC:mmc_spi: Recover from CRC error for SD read/write operation over SPI.

In reply to sonic zhang

Hi,

There were no comments about this patch -- Andrew, please take it if there are no objections.

From: Sonic Zhang <[email protected]>
Date: Mon, 12 Jul 2010 15:50:56 +0800
Subject: [PATCH] mmc_spi: Recover from CRC error for SD read/write operation over SPI.

SPI bus is not reliable on all platforms when doing large data transfer. Current mmc spi driver fails SD read/write command immediately, if occasional CRC error is reported by SD device. This patch makes the operation recover from the CRC error by doing last SD command again. The retry count is set to 5 to ensure the driver pass any stress test.

Signed-off-by Sonic Zhang <[email protected]>
[cjb: fixed merge conflict against current Linus]
Signed-off-by Chris Ball <[email protected]>
---
 drivers/mmc/host/mmc_spi.c |   21 +++++++++++++++++++++
 1 files changed, 21 insertions(+), 0 deletions(-)
drivers/mmc/host/mmc_spi.c +21 -0
index 62a3582..3d48e21 100644
@@ -1055,6 +1055,8 @@ static void mmc_spi_request(struct mmc_host *mmc, struct mmc_request *mrq)
 {
 	struct mmc_spi_host	*host = mmc_priv(mmc);
 	int			status = -EINVAL;
+	int			crc_retry = 5;
+	struct mmc_command	stop;
 
 #ifdef DEBUG
 	/* MMC core and layered drivers *MUST* issue SPI-aware commands */

@@ -1087,10 +1089,29 @@ static void mmc_spi_request(struct mmc_host *mmc, struct mmc_request *mrq)
 	/* request exclusive bus access */
 	spi_bus_lock(host->spi->master);
 
+crc_recover:
 	/* issue command; then optionally data and stop */
 	status = mmc_spi_command_send(host, mrq, mrq->cmd, mrq->data != NULL);
 	if (status == 0 && mrq->data) {
 		mmc_spi_data_do(host, mrq->cmd, mrq->data, mrq->data->blksz);
+
+		/*
+		 * SPI bus is not reliable when doing large data transfer.
+		 * If occasional crc error is reported by SD device when do
+		 * data read/write over SPI, it can be recovered by doing
+		 * last SD command again. The retry count is set to 5 to
+		 * ensure the driver pass any stress test.
+		 */
+		if (mrq->data->error == -EILSEQ && crc_retry) {
+			stop.opcode = MMC_STOP_TRANSMISSION;
+			stop.arg = 0;
+			stop.flags = MMC_RSP_SPI_R1B | MMC_RSP_R1B | MMC_CMD_AC;
+			status = mmc_spi_command_send(host, mrq, &stop, 0);
+			crc_retry--;
+			mrq->data->error = 0;
+			goto crc_recover;
+		}
+
 		if (mrq->stop)
 			status = mmc_spi_command_send(host, mrq, mrq->stop, 0);
 		else
Signature · 5 lines
1.7.0.1

-- 
Chris Ball   <[email protected]>   <http://printf.net/>
One Laptop Per Child
#4 lore raw

PATCH 1/1 MMC:mmc_spi: Recover from CRC error for SD read/write operation over SPI.

From: Sonic Zhang <[email protected]>

SPI bus is not reliable on all platforms when doing large data transfer. Current mmc spi driver fails SD read/write command immediately, if occasional CRC error is reported by SD device. This patch makes the operation recover from the CRC error by doing last SD command again. The retry count is set to 5 to ensure the driver pass any stress test.

Signed-off-by Sonic Zhang <[email protected]>
[cjb: fixed merge conflict against current Linus]
Signed-off-by Chris Ball <[email protected]>
Signed-off-by Mike Frysinger <[email protected]>

---
Andrew: we posted this months ago with no feedback ... could you pick it up ?

 drivers/mmc/host/mmc_spi.c |   22 ++++++++++++++++++++++
 1 files changed, 22 insertions(+), 0 deletions(-)
drivers/mmc/host/mmc_spi.c +22 -0
index 62a3582..fbd673e 100644
@@ -1055,6 +1055,8 @@ static void mmc_spi_request(struct mmc_host *mmc, struct mmc_request *mrq)
 {
 	struct mmc_spi_host	*host = mmc_priv(mmc);
 	int			status = -EINVAL;
+	int			crc_retry = 5;
+	struct mmc_command	stop;
 
 #ifdef DEBUG
 	/* MMC core and layered drivers *MUST* issue SPI-aware commands */

@@ -1087,10 +1089,30 @@ static void mmc_spi_request(struct mmc_host *mmc, struct mmc_request *mrq)
 	/* request exclusive bus access */
 	spi_bus_lock(host->spi->master);
 
+ crc_recover:
 	/* issue command; then optionally data and stop */
 	status = mmc_spi_command_send(host, mrq, mrq->cmd, mrq->data != NULL);
 	if (status == 0 && mrq->data) {
 		mmc_spi_data_do(host, mrq->cmd, mrq->data, mrq->data->blksz);
+
+		/*
+		 * SPI bus is not reliable when doing large data transfer.
+		 * If occasional crc error is reported by SD device when do
+		 * data read/write over SPI, it can be recovered by doing
+		 * last SD command again. The retry count is set to 5 to
+		 * ensure the driver pass any stress test.
+		 */
+		if (mrq->data->error == -EILSEQ && crc_retry) {
+			stop.opcode = MMC_STOP_TRANSMISSION;
+			stop.arg = 0;
+			stop.flags = MMC_RSP_SPI_R1B | MMC_RSP_R1B | MMC_CMD_AC;
+
+			status = mmc_spi_command_send(host, mrq, &stop, 0);
+			crc_retry--;
+			mrq->data->error = 0;
+			goto crc_recover;
+		}
+
 		if (mrq->stop)
 			status = mmc_spi_command_send(host, mrq, mrq->stop, 0);
 		else
Signature · 1 line
1.7.3.1
#5 lore raw
Quoted 1 line · Mike Frysinger

Andrew: we posted this months ago with no feedback ... could you pick it up ?

Sorry, my fault for missing this -- I've pushed it to mmc-next now, with minor language changes.

Thanks,

Signature · 2 lines
Chris Ball   <[email protected]>   <http://printf.net/>
One Laptop Per Child
#6 lore raw
Quoted 5 lines · Chris Ball
Quoted 1 line · Mike Frysinger

Andrew: we posted this months ago with no feedback ... could you pick it up ?

Sorry, my fault for missing this -- I've pushed it to mmc-next now, with minor language changes.

np. thanks !
-mike