Re: 2.6.39-rc5+ BUG at scsi_run_queue+0x24/0xe3

["Jim Schutt" <jaschut@sandia.gov>]

James Bottomley wrote:
> On Tue, 2011-05-03 at 11:27 -0600, Jim Schutt wrote:
>> James Bottomley wrote:
>>> On Tue, 2011-05-03 at 10:53 -0600, Jim Schutt wrote:
>>>> Please let me know if what further information you need, or if there is
>>>> anything I can do, to help resolve this.
>>> I think this is the fix (already in rc-fixes):
>>>
>>> James
>>>
>>> ---
>>> From 3e85ea868dbd60a84240be5c1eebc36841b9c568 Mon Sep 17 00:00:00 2001
>>> From: James Bottomley <James.Bottomley@suse.de>
>>> Date: Sun, 1 May 2011 09:42:07 -0500
>>> Subject: [PATCH] [SCSI] fix oops in scsi_run_queue()
>>>
>>> The recent commit closing the race window in device teardown:
>>>
>>> commit 86cbfb5607d4b81b1a993ff689bbd2addd5d3a9b
>>> Author: James Bottomley <James.Bottomley@suse.de>
>>> Date:   Fri Apr 22 10:39:59 2011 -0500
>>>
>>>     [SCSI] put stricter guards on queue dead checks
>>>
>>> is causing a potential NULL deref in scsi_run_queue() because the
>>> q->queuedata may already be NULL by the time this function is called.
>>> Since we shouldn't be running a queue that is being torn down, simply
>>> add a NULL check in scsi_run_queue() to forestall this.
>>>
>>> Signed-off-by: James Bottomley <James.Bottomley@suse.de>
>>>
>>> diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
>>> index e9901b8..03979f4 100644
>>> --- a/drivers/scsi/scsi_lib.c
>>> +++ b/drivers/scsi/scsi_lib.c
>>> @@ -404,6 +404,10 @@ static void scsi_run_queue(struct request_queue *q)
>>>  	LIST_HEAD(starved_list);
>>>  	unsigned long flags;
>>>  
>>> +	/* if the device is dead, sdev will be NULL, so no queue to run */
>>> +	if (!sdev)
>>> +		return;
>>> +
>>>  	if (scsi_target(sdev)->single_lun)
>>>  		scsi_single_lun_run(sdev);
>>>  
>> Hmmm, with the above added, I still get BUGs.  Here's an
>> example:
>>
>> [   17.142931] BUG: unable to handle kernel NULL pointer dereference at           (null)
>> [   17.143002] IP: [<ffffffffa01cf8c5>] scsi_run_queue+0x24/0xec [scsi_mod]
> 
> Ooh, compiler optimisation, I think; try this instead
> 
> James
> 
> ---
> 
> diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
> index e9901b8..0bac91e 100644
> --- a/drivers/scsi/scsi_lib.c
> +++ b/drivers/scsi/scsi_lib.c
> @@ -400,10 +400,15 @@ static inline int scsi_host_is_busy(struct Scsi_Host *shost)
>  static void scsi_run_queue(struct request_queue *q)
>  {
>  	struct scsi_device *sdev = q->queuedata;
> -	struct Scsi_Host *shost = sdev->host;
> +	struct Scsi_Host *shost;
>  	LIST_HEAD(starved_list);
>  	unsigned long flags;
>  
> +	/* if the device is dead, sdev will be NULL, so no queue to run */
> +	if (!sdev)
> +		return;
> +
> +	shost = sdev->host;
>  	if (scsi_target(sdev)->single_lun)
>  		scsi_single_lun_run(sdev);
>  

Yes, that definitely fixes things for me.

Thanks!!

-- Jim