From: mchristi@redhat.com <mchristi@redhat.com>
Date: Thu, 2 Oct 2008 13:58:00 -0500
Subject: [scsi] fix QUEUE_FULL retry handling
Message-id: 1222973881-4441-1-git-send-email-mchristi@redhat.com
O-Subject: [PATCH 1/2] fix QUEUE_FULL retry handling
Bugzilla: 463709
RH-Acked-by: Tomas Henzl <thenzl@redhat.com>

From: Mike Christie <mchristi@redhat.com>

This is for BZ 463709.

When I changed the failfast code I added new error codes
like SCSI_MLQUEUE_DEVICE_BUSY2 which were to be used by new code
and internally where there were not LLD entry points that could return
the old values like SCSI_MLQUEUE_DEVICE_BUSY. When I added that compat
patch I goofed in another place (it looks like this is the last place
I could screw up on at least). We were using the older return values
internally in scsi_error.c but the caller only expected the new values,
so errors like QUEUE_FULL were getting finished instead of retried.

Patch was tested here by forcing devices to return QUEUE_FULLs, then
making sure it was retried.

This is not upstream since it was a workaround for our KABI.

diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c
index e6bc2de..918ec68 100644
--- a/drivers/scsi/scsi_error.c
+++ b/drivers/scsi/scsi_error.c
@@ -1236,7 +1236,7 @@ int scsi_decide_disposition(struct scsi_cmnd *scmd)
 	case DID_IMM_RETRY:
 		return SCSI_MLQUEUE_IMM_RETRY;
 	case DID_REQUEUE:
-		return SCSI_MLQUEUE_DEVICE_BUSY;
+		return SCSI_MLQUEUE_DEVICE_BUSY2;
 	case DID_TRANSPORT_DISRUPTED:
 		/*
 		 * LLD/transport was disrupted during processing of the IO.
@@ -1305,7 +1305,7 @@ int scsi_decide_disposition(struct scsi_cmnd *scmd)
 		 * the empty queue handling to trigger a stall in the
 		 * device.
 		 */
-		return SCSI_MLQUEUE_DEVICE_BUSY;
+		return SCSI_MLQUEUE_DEVICE_BUSY2;
 	case GOOD:
 	case COMMAND_TERMINATED:
 	case TASK_ABORTED: