3.1.0-rc4: oops at elv_put_request()

3.1.0-rc4: oops at elv_put_request()

[Alexey Dobriyan]

Caught this one, there was no real load at the moment.
FWIW, setup is ext4 over RAID-6 across 4 disks.

	elv_put_request
	__blk_put_request
	blk_put_request
	scsi_execute
	scsi_execute_req
	scsi_test_unit_ready
	sd_check_events
	disk_events_workfn
	process_one_work
	worker_thread
	kthread
	kernel_thread_helper

$ addr2line -e vmlinux 0xffffffff811bcbbb
block/elevator.c:772

   768	void elv_put_request(struct request_queue *q, struct request *rq)
   769	{
   770		struct elevator_queue *e = q->elevator;
   771	
   772	===>	if (e->ops->elevator_put_req_fn)	<===
   773			e->ops->elevator_put_req_fn(rq);
   774	}

Re: 3.1.0-rc4: oops at elv_put_request()

[Linus Torvalds]

Re-sending selected emails that apparently never went out. But others
are likely just lost forever..

On Sun, Sep 11, 2011 at 2:47 PM, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
> On Sun, Sep 11, 2011 at 2:36 PM, Alexey Dobriyan <adobriyan@gmail.com> wrote:
>> Caught this one, there was no real load at the moment.
>> FWIW, setup is ext4 over RAID-6 across 4 disks.
>
> Do you have the register state and oops address? Is it a NULL pointer,
> or a case of PAGE_ALLOC_DEBUG or what?
>
> There's commit 94007751bb02 that fixes one md-related oops ("Avoid
> dereferencing a 'request_queue' after last close") but that doesn't
> seem to be the case for you (it needs a __blkdev_put()).
>
>                     Linus
>