Dear Matt,
Thank you very much. Attached is the re-written patch with comments
included. This has been tested in our platform, where it recovers from
CMD12 cmd_conflict errors.
Regards,
Mahadev
diff -urNd linux-orig/drivers/mmc/card/block.c
linux-current/drivers/mmc/card/block.c
--- linux-orig/drivers/mmc/card/block.c 2009-09-16 15:43:03.667274139 +0530
+++ linux-current/drivers/mmc/card/block.c 2009-09-30 16:41:50.605273759
+0530
@@ -330,6 +330,67 @@
mmc_queue_bounce_post(mq);
+ /* cmd12 error recovery fix */
+ if( (brq.stop.error) && (brq.data.blocks > 1) &&
+ !mmc_host_is_spi(card->host) &&
mmc_card_sd(card) ) {
+
+ /*
+ * if there is an error for CMD12, get the state of the
card by
+ * sending cmd13 and then check the state of the card.
+ * If the card state is "tran", consider the cmd12 is
successfull
+ * If not, then send again CMD12 to stop the data
transfer. This
+ * time, any failure to get response, will result in to
the card
+ * being non-recoverable.
+ */
+ memset(&cmd, 0, sizeof(struct mmc_command));
+ cmd.opcode = MMC_SEND_STATUS;
+ cmd.arg = card->rca << 16;
+ cmd.flags = MMC_RSP_R1 | MMC_CMD_AC;
+ mmc_wait_for_cmd(card->host, &cmd, 0);
+
+ if(!cmd.error && R1_CURRENT_STATE(cmd.resp[0]) == 4) {
+ /* card is in "trans" state, so treat the CMD12
as succesfull */
+ brq.stop.error = 0;
+ } else {
+ /*
+ * card is not still in trans state. Now send
CMD12 again and
+ * repeat the process of checking the card
status using CMD13
+ */
+ memset(&cmd, 0, sizeof(struct mmc_command));
+ cmd.opcode = MMC_STOP_TRANSMISSION;
+ cmd.arg = 0;
+ cmd.flags = MMC_RSP_SPI_R1B | MMC_RSP_R1B |
MMC_CMD_AC;
+ mmc_wait_for_cmd(card->host, &cmd, 0);
+
+ /* get the card status */
+ memset(&cmd, 0, sizeof(struct mmc_command));
+ cmd.opcode = MMC_SEND_STATUS;
+ cmd.arg = card->rca << 16;
+ cmd.flags = MMC_RSP_R1 | MMC_CMD_AC;
+ mmc_wait_for_cmd(card->host, &cmd, 0);
+
+ if(!cmd.error) {
+ if(R1_CURRENT_STATE(cmd.resp[0]) == 4) {
+ /* card is in "trans" state, so
treat the CMD12 as succesfull */
+ brq.stop.error = 0;
+ } else {
+ printk(KERN_ERR "%s: CMD12
error not recoverable\n",
+
req->rq_disk->disk_name );
+ /*
+ * card is not in "trans" state
yet. Just post an error and
+ * error is not recoverable
+ */
+ }
+ } else {
+ printk(KERN_ERR "%s: CMD13 returns an
error!!\n",
+ req->rq_disk->disk_name
);
+ /* error for cmd13!!, then something is
wrong with the card */
+ }
+ }
+ }
+ /* cmd12 error recovery fix */
+
+
/*
* Check for errors here, but don't jump to cmd_err
* until later as we need to wait for the card to leave
2009/9/23 Matt Fleming <[email protected]>:
> On Wed, Sep 16, 2009 at 04:09:11PM +0530, Mahadev Cholachagudda wrote:
>> Dear All,
>>
>> I've a kernel code base of 2.6.28.9, where in I find that CMD12 error
>> recovery is not been implemented. It is a wise idea to implement CMD12
>> error recovery, that involves sending CMD13 to get the status of the
>> card. If the card is in "trans" state, it means CMD12 error shall be
>> treated as successfull. If card state is not in "trans", then we have
>> to send CMD12 again to stop the data transfer and then checking of the
>> card status through CMD13. This sequence is detailed in the
>> "Simplified Host Controller specification" on SDCard website. I
>> believe this is the case for any controller, though I may be wrong.
>>
>> With the modification done to the code which is attached as patch to
>> the latest mmc git, it is working at our site. And we do see some
>> errors with the SDIO controller (proprietary but meets Standard SD
>> host controller specification) that we are using.
>>
>> Request you to comment on the patch.
>>
>
> In principal this patch seems OK. However, I recommend reading
> Documentation/CodingStyle and sending this patch again once you've fixed
> up the coding style and whitespace errors.
>
> Is there a specific reason that you have used mmc_wait_for_req() instead
> of mmc_wait_for_cmd() and reusing the "cmd" local variable? The empty
> "else" clauses can also probably be deleted.
>
--
Cheers,
Mahadev
--
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
