Re: block: add blk_run_queue_async

[Jens Axboe <jaxboe@fusionio.com>]

On 2011-04-18 23:48, Mike Snitzer wrote:
> On Mon, Apr 18 2011 at  4:20pm -0400,
> Mike Snitzer <snitzer@redhat.com> wrote:
> 
>> On Mon, Apr 18 2011 at  3:59pm -0400,
>> Christoph Hellwig <hch@infradead.org> wrote:
>>
>>> On Mon, Apr 18, 2011 at 03:55:04PM -0400, Mike Snitzer wrote:
>>>>> +{
>>>>> + ? ? ? if (likely(!blk_queue_stopped(q)))
>>>>> + ? ? ? ? ? ? ? queue_delayed_work(kblockd_workqueue, &q->delay_work, 0);
>>>>
>>>> I know Jens already queued this up 'for-linus' but why not use
>>>> kblockd_schedule_work(q, &q->delay_work)?
>>>
>>> I don't see what that would buy us.  If we'd absolutely want a wrapper
>>> a blk_delay_queue(q, 0) in Jens' current tree would do it now that is
>>> has been fixed up to use the kblockd workqueue.
>>
>> Right, I missed 4521cc4 block: blk_delay_queue() should use kblockd
>> workqueue.  So why not use blk_delay_queue()?
>>
>> I agree with Jens that it doesn't much matter but I also cannot see it
>> being a bad thing.. I'd prefer it ;)
>>
>> *shrug*
> 
> Also, FYI, I'm seeing a leftover '@force_kblockd: ...' comment in the
> __blk_run_queue's comment block.

Thanks Mike, I've killed that now.

-- 
Jens Axboe