Re: [git patches] libata updates for 2.6.37

[Tejun Heo <tj@kernel.org>]

Hello,

On 12/06/2010 05:07 PM, Kyle McMartin wrote:
> On Fri, Dec 03, 2010 at 06:59:02AM -0500, Kyle McMartin wrote:
>> On Fri, Dec 03, 2010 at 11:10:12AM +0100, Tejun Heo wrote:
>>>> This gets triggered a bunch on boot-up:
>>>
>>> Yeah, that's expected.  Can you please do the usual testing routine
>>> and trigger the EH giving up message and then attach the full kernel
>>> log?
>>>
>>
> 
> OK, attached is a couple days worth of suspend/resume, hopefully that
> turns up what you're looking for.

Heh, yeah, actually it did.  Can you please apply the following patch
on top and see whether the problem goes away?

diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
index 7f77c67..f23d6d4 100644
--- a/drivers/ata/libata-core.c
+++ b/drivers/ata/libata-core.c
@@ -4807,9 +4807,6 @@ static void ata_verify_xfer(struct ata_queued_cmd *qc)
 {
 	struct ata_device *dev = qc->dev;

-	if (ata_tag_internal(qc->tag))
-		return;
-
 	if (ata_is_nodata(qc->tf.protocol))
 		return;

@@ -4858,14 +4855,23 @@ void ata_qc_complete(struct ata_queued_cmd *qc)
 		if (unlikely(qc->err_mask))
 			qc->flags |= ATA_QCFLAG_FAILED;

-		if (unlikely(qc->flags & ATA_QCFLAG_FAILED)) {
-			/* always fill result TF for failed qc */
+		/*
+		 * Finish internal commands without any further processing
+		 * and always with the result TF filled.
+		 */
+		if (unlikely(ata_tag_internal(qc->tag))) {
 			fill_result_tf(qc);
+			__ata_qc_complete(qc);
+			return;
+		}

-			if (!ata_tag_internal(qc->tag))
-				ata_qc_schedule_eh(qc);
-			else
-				__ata_qc_complete(qc);
+		/*
+		 * Non-internal qc has failed.  Fill the result TF and
+		 * summon EH.
+		 */
+		if (unlikely(qc->flags & ATA_QCFLAG_FAILED)) {
+			fill_result_tf(qc);
+			ata_qc_schedule_eh(qc);
 			return;
 		}