Re: [PATCH v2 0/2] block,scsi: fixup blk_get_request dead queue scenarios

[Jeff Moyer <jmoyer@redhat.com>]

Joe Lawrence <joe.lawrence@stratus.com> writes:

> On Thu, 26 Jun 2014, Jeff Moyer wrote:
>> Thanks, Jens.  Joe, the patches don't apply anymore... would you mind
>> sending an updated set so Jens doesn't have to do the fixups?  I think
>> you also need to convert blk_mq_alloc_request, I didn't see that in your
>> patch.
>
> Hi Jeff,
>
> I have a new set that applies on top of 3.16-rc2... but before posting, I 
> wasn't sure what needs to be converted w/regard to blk_mq_alloc_request.  
> As far as I can tell, the new multi-queue block IO queueing implementation 
> via blk_mq_alloc_request should be safe from this bug as it doesn't 
> include any dead queue checks that introduce early exits.

Hi, Joe,

blk_mq_alloc_request will return NULL if blk_mq_queue_enter() returns
non-zero.

Cheers,
Jeff