[PATCH] SCSI: fix scsi_kill_request() busy count handling

[PATCH] SCSI: fix scsi_kill_request() busy count handling

[Tejun Heo]

scsi_kill_request() completes requests via normal SCSI completion path
which decrements busy counts; however, requests which get passed to
scsi_kill_request() aren't holding busy counts and scsi_kill_request()
don't increment them before invoking completion path resulting in
incorrect busy counts.  Bump up busy counts before invoking completion
path.

Signed-off-by: Tejun Heo <htejun@gmail.com>

---

 drivers/scsi/scsi_lib.c |   15 +++++++++++++++
 1 files changed, 15 insertions(+), 0 deletions(-)

a12bf362489393691d147f93fbe1aa5d5a3a513d
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 8f010a3..7b0f9a3 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -1479,6 +1479,8 @@ static inline int scsi_host_queue_ready(
 static void scsi_kill_request(struct request *req, request_queue_t *q)
 {
 	struct scsi_cmnd *cmd = req->special;
+	struct scsi_device *sdev = cmd->device;
+	struct Scsi_Host *shost = sdev->host;
 
 	blkdev_dequeue_request(req);
 
@@ -1491,6 +1493,19 @@ static void scsi_kill_request(struct req
 	scsi_init_cmd_errh(cmd);
 	cmd->result = DID_NO_CONNECT << 16;
 	atomic_inc(&cmd->device->iorequest_cnt);
+
+	/*
+	 * SCSI request completion path will do scsi_device_unbusy(),
+	 * bump busy counts.  To bump the counters, we need to dance
+	 * with the locks as normal issue path does.
+	 */
+	sdev->device_busy++;
+	spin_unlock(sdev->request_queue->queue_lock);
+	spin_lock(shost->host_lock);
+	shost->host_busy++;
+	spin_unlock(shost->host_lock);
+	spin_lock(sdev->request_queue->queue_lock);
+
 	__scsi_done(cmd);
 }
 
-- 
1.2.4

Re: [PATCH] SCSI: fix scsi_kill_request() busy count handling

[Jeff Garzik]

Tejun Heo wrote:
> scsi_kill_request() completes requests via normal SCSI completion path
> which decrements busy counts; however, requests which get passed to
> scsi_kill_request() aren't holding busy counts and scsi_kill_request()
> don't increment them before invoking completion path resulting in
> incorrect busy counts.  Bump up busy counts before invoking completion
> path.
> 
> Signed-off-by: Tejun Heo <htejun@gmail.com>

ACK.

As with the other two SCSI core patches, we're blocked on linux-scsi people.

Same response as the other two:  if nobody complains, and nobody merges 
into scsi-misc-2.6 in a sane amount of time, I'll push it into libata.

	Jeff