2006-02-02 17:30:00

by Brian King

Subject: [PATCH 1/1] blk: Fix SG_IO ioctl failure retry looping

When issuing an SG_IO ioctl through sd that resulted
in an unrecoverable error, a nearly infinite retry loop
was discovered. This is due to the fact that the block
layer SG_IO code is not setting up rq->retries. This
patch also fixes up the sg_scsi_ioctl path.


Signed-off-by: Brian King <[email protected]>
---

linux-2.6-bjking1/block/scsi_ioctl.c | 2 ++
1 files changed, 2 insertions(+)

diff -puN block/scsi_ioctl.c~scsi_ioctl_retries block/scsi_ioctl.c
--- linux-2.6/block/scsi_ioctl.c~scsi_ioctl_retries 2006-02-02 11:07:50.000000000 -0600
+++ linux-2.6-bjking1/block/scsi_ioctl.c 2006-02-02 11:09:11.000000000 -0600
@@ -309,6 +309,7 @@ static int sg_io(struct file *file, requ
rq->timeout = q->sg_timeout;
if (!rq->timeout)
rq->timeout = BLK_DEFAULT_TIMEOUT;
+ rq->retries = 0;

start_time = jiffies;

@@ -427,6 +428,7 @@ static int sg_scsi_ioctl(struct file *fi
rq->data = buffer;
rq->data_len = bytes;
rq->flags |= REQ_BLOCK_PC;
+ rq->retries = 0;

blk_execute_rq(q, bd_disk, rq, 0);
err = rq->errors & 0xff; /* only 8 bit SCSI status */
_

2006-02-02 17:38:26

by Jens Axboe

Subject: Re: [PATCH 1/1] blk: Fix SG_IO ioctl failure retry looping

On Thu, Feb 02 2006, Brian King wrote:
>
> When issuing an SG_IO ioctl through sd that resulted
> in an unrecoverable error, a nearly infinite retry loop
> was discovered. This is due to the fact that the block
> layer SG_IO code is not setting up rq->retries. This
> patch also fixes up the sg_scsi_ioctl path.

Looks good, applied. Thanks!

--
Jens Axboe