On 11/8/18 10:35 AM, Bart Van Assche wrote:
> On Thu, 2018-11-08 at 09:31 -0700, Jens Axboe wrote:
>> On 11/8/18 9:28 AM, Bart Van Assche wrote:
>>> On Thu, 2018-11-08 at 09:06 -0700, Jens Axboe wrote:
>>>> --- a/block/blk-mq-debugfs.c
>>>> +++ b/block/blk-mq-debugfs.c
>>>> @@ -424,13 +424,15 @@ struct show_busy_params {
>>>> * Note: the state of a request may change while this function is in progress,
>>>> * e.g. due to a concurrent blk_mq_finish_request() call.
>>>> */
>>>> -static void hctx_show_busy_rq(struct request *rq, void *data, bool reserved)
>>>> +static bool hctx_show_busy_rq(struct request *rq, void *data, bool reserved)
>>>> {
>>>
>>> Please update the kdoc header above hctx_show_busy_rq() such that it reflects
>>> the new behavior. I'm referring to the "will be called for each request" part.
>>> Otherwise this patch looks fine to me.
>>
>> Took a look at the comment, and what do you want changed? There's no change
>> in behavior for hctx_show_busy_rq(), it loops all requests just like before.
>> We just return true to ensure we continue iterating.
>
> Oops, I added my reply below the wrong function. I wanted to refer to the
> following comment above blk_mq_queue_tag_busy_iter():
>
> * @fn: Pointer to the function that will be called for each request
>
> Additionally, how about similar comments above bt_for_each(), bt_tags_for_each()
> blk_mq_all_tag_busy_iter() and blk_mq_tagset_busy_iter()? I think this patch
> affects all these functions.
Fair enough, I'll add a documentation patch. Didn't consider it a big deal
since this is how it's always worked internally, but we haven't exposed
this break/continue functionality outside the sbitmap core until now.
--
Jens Axboe
