On Wed, Aug 6, 2014 at 10:32 PM, Alex Elder <elder@xxxxxxxx> wrote:
> On 08/05/2014 02:38 AM, Ilya Dryomov wrote:
>> While it was never a good idea to sleep in request_fn(), commit
>> 34c6bc2c919a ("locking/mutexes: Add extra reschedule point") made it
>> a *bad* idea. mutex_lock() since 3.15 may reschedule *before* putting
>> task on the mutex wait queue, which for tasks in !TASK_RUNNING state
>> means block forever. request_fn() may be called with !TASK_RUNNING on
>> the way to schedule() in io_schedule().
>>
>> Offload request handling to a workqueue, one per rbd device, to avoid
>> calling blocking primitives from rbd_request_fn().
>
> Off topic... If you supply "--patience" to your git diff command
> you'll get an easier-to-read result in some cases (like this one).
> If you like that you can just do:
> git config --global --add diff.algorithm patience
>
> I have several comments below, none of which are bugs, just
> suggestions.
>
> I was thinking you could use a single work queue for all rbd
> devices, but I'm not sure that's very important. You'd have
> to stash the rbd_dev pointer in every request somewhere.
That's possible, but workqueues are not threads and are supposed to be
relatively cheap. And for the future, it's probably worth switching to
blk-mq infrastructure, as Christoph suggested.
Thanks,
Ilya
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
