Re: 2.6.29 regression: ATA bus errors on resume

[Tejun Heo]

Hello,

Sorry about the delay.

Niel Lambrechts wrote:
> I think I should be able to reproduce when actively using X with 2.6.29,
> and I have an external disk where I could backup to / boot from if the
> corruption became a problem.
> 
> These issues are keeping me from 2.6.29 so I'll gladly help where I can,
> if you can please provide me the patches and the .config settings that
> may be required?

Attached is the debug patch.  Please reproduce the problem with the
patch applied and post full kernel log.

Thanks.

-- 
tejun
diff --git a/drivers/ata/libata-eh.c b/drivers/ata/libata-eh.c
index 0183131..1597874 100644
--- a/drivers/ata/libata-eh.c
+++ b/drivers/ata/libata-eh.c
@@ -1274,7 +1274,7 @@ void ata_eh_about_to_do(struct ata_link *link, struct ata_device *dev,
 {
 	struct ata_port *ap = link->ap;
 	struct ata_eh_info *ehi = &link->eh_info;
-	struct ata_eh_context *ehc = &link->eh_context;
+	//struct ata_eh_context *ehc = &link->eh_context;
 	unsigned long flags;
 
 	spin_lock_irqsave(ap->lock, flags);
@@ -1284,7 +1284,7 @@ void ata_eh_about_to_do(struct ata_link *link, struct ata_device *dev,
 	/* About to take EH action, set RECOVERED.  Ignore actions on
 	 * slave links as master will do them again.
 	 */
-	if (!(ehc->i.flags & ATA_EHI_QUIET) && link != ap->slave_link)
+	if (/*!(ehc->i.flags & ATA_EHI_QUIET) && */link != ap->slave_link)
 		ap->pflags |= ATA_PFLAG_RECOVERED;
 
 	spin_unlock_irqrestore(ap->lock, flags);
@@ -2017,8 +2017,13 @@ static void ata_eh_link_autopsy(struct ata_link *link)
 
 		/* determine whether the command is worth retrying */
 		if (!(qc->err_mask & AC_ERR_INVALID) &&
-		    ((qc->flags & ATA_QCFLAG_IO) || qc->err_mask != AC_ERR_DEV))
+		    ((qc->flags & ATA_QCFLAG_IO) || qc->err_mask != AC_ERR_DEV)) {
+			ata_dev_printk(qc->dev, KERN_INFO,
+				       "XXX setting retry on qc%d\n", tag);
 			qc->flags |= ATA_QCFLAG_RETRY;
+		} else
+			ata_dev_printk(qc->dev, KERN_INFO,
+				       "XXX no retry for qc%d\n", tag);
 
 		/* accumulate error info */
 		ehc->i.dev = qc->dev;
@@ -2126,8 +2131,8 @@ static void ata_eh_link_report(struct ata_link *link)
 	char tries_buf[6];
 	int tag, nr_failed = 0;
 
-	if (ehc->i.flags & ATA_EHI_QUIET)
-		return;
+	/*if (ehc->i.flags & ATA_EHI_QUIET)
+		return;*/
 
 	desc = NULL;
 	if (ehc->i.desc[0] != '\0')
@@ -2147,8 +2152,8 @@ static void ata_eh_link_report(struct ata_link *link)
 		nr_failed++;
 	}
 
-	if (!nr_failed && !ehc->i.err_mask)
-		return;
+	/*if (!nr_failed && !ehc->i.err_mask)
+		return;*/
 
 	frozen = "";
 	if (ap->pflags & ATA_PFLAG_FROZEN)
@@ -3350,16 +3355,23 @@ void ata_eh_finish(struct ata_port *ap)
 			 * generate sense data in this function,
 			 * considering both err_mask and tf.
 			 */
-			if (qc->flags & ATA_QCFLAG_RETRY)
+			if (qc->flags & ATA_QCFLAG_RETRY) {
+				ata_dev_printk(qc->dev, KERN_INFO, "XXX retrying qc%d, retries=%d\n",
+					       tag, qc->scsicmd->retries);
 				ata_eh_qc_retry(qc);
-			else
+			} else {
+				ata_dev_printk(qc->dev, KERN_INFO, "XXX terminating qc%d\n", tag);
 				ata_eh_qc_complete(qc);
+			}
 		} else {
 			if (qc->flags & ATA_QCFLAG_SENSE_VALID) {
+				ata_dev_printk(qc->dev, KERN_INFO, "XXX terminating qc%d (SENSE), retries=%d\n",
+					       tag, qc->scsicmd->retries);
 				ata_eh_qc_complete(qc);
 			} else {
 				/* feed zero TF to sense generation */
 				memset(&qc->result_tf, 0, sizeof(qc->result_tf));
+				ata_dev_printk(qc->dev, KERN_INFO, "XXX retrying qc%d (bogus SENSE)\n", tag);
 				ata_eh_qc_retry(qc);
 			}
 		}