Re: [PATCH] block: fix blk_queue_end_tag()

[Jens Axboe]

On 2011-12-21 09:22, Williams, Dan J wrote:
> On Wed, Dec 21, 2011 at 12:16 AM, Tao Ma <tm@xxxxxx> wrote:
>> On 12/21/2011 03:30 PM, Williams, Dan J wrote:
>>> On Tue, Dec 20, 2011 at 10:48 PM, Tao Ma <tm@xxxxxx> wrote:
>>>> On 12/21/2011 02:36 PM, Meelis Roos wrote:
>>>>>>> -   if (unlikely(tag >= bqt->max_depth)) {
>>>>>>> +   if (WARN_ONCE(tag >= bqt->real_max_depth,
>>>>>>> +                 "%s: tag %d greater than tag map size: %d\n",
>>>>>>> +                 __func__, tag, bqt->real_max_depth)) {
>>>>>>>             /*
>>>>>>>              * This can happen after tag depth has been reduced.
>>>>>> Please also change the comments here since it should never happen in the
>>>>>> right workload.
>>>>>
>>>>> What do you mean by right workload? Normal workload?
>>>> yeah, so real_max_depth is the maximum depth we ever have. So in normal
>>>> case(shrinking queue depth is also a normal user case), we should never
>>>> arrive here. In another word, if tag >= real_max_depth, we should have a
>>>> bug in the kernel.
>>>
>>> So this is what Ed Nadolski suggested, just cut to the chase and do,
>>> the following.  Seems like the comment is what got us into trouble in
>>> the first place.
>>>
>>> diff --git a/block/blk-tag.c b/block/blk-tag.c
>>> index e74d6d1..e297d9d7 100644
>>> --- a/block/blk-tag.c
>>> +++ b/block/blk-tag.c
>>> @@ -284,16 +284,7 @@ void blk_queue_end_tag(struct request_queue *q,
>>> struct request *rq)
>>>         struct blk_queue_tag *bqt = q->queue_tags;
>>>         int tag = rq->tag;
>>>
>>> -       BUG_ON(tag == -1);
>>> -
>>> -       if (unlikely(tag >= bqt->max_depth)) {
>>> -               /*
>>> -                * This can happen after tag depth has been reduced.
>>> -                * But tag shouldn't be larger than real_max_depth.
>>> -                */
>>> -               WARN_ON(tag >= bqt->real_max_depth);
>>> -               return;
>>> -       }
>>> +       BUG_ON(tag == -1 || tag > bqt->real_max_depth);
>> I guess tag = bqt->real_max_depth should also be a problem.
> 
> Yes, sorry, should have been >=

Can you resend a v2 and I'll get it applied, thanks.

-- 
Jens Axboe

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/