Re: [PATCHv2 1/2] block: avoid cpu_hotplug_lock depedency on freeze_lock

[Nilay Shroff <nilay@linux.ibm.com>]

On 8/6/25 6:48 AM, Yu Kuai wrote:
> Hi,
> 
> 在 2025/08/06 1:17, Nilay Shroff 写道:
>>     static inline void rq_qos_done_bio(struct bio *bio)
>>   {
>> -    if (static_branch_unlikely(&block_rq_qos) &&
>> -        bio->bi_bdev && (bio_flagged(bio, BIO_QOS_THROTTLED) ||
>> -                 bio_flagged(bio, BIO_QOS_MERGED))) {
>> -        struct request_queue *q = bdev_get_queue(bio->bi_bdev);
>> -        if (q->rq_qos)
>> -            __rq_qos_done_bio(q->rq_qos, bio);
>> -    }
>> +    struct request_queue *q;
>> +
>> +    if (!bio->bi_bdev || (!bio_flagged(bio, BIO_QOS_THROTTLED) &&
>> +                 !bio_flagged(bio, BIO_QOS_MERGED)))
>> +        return;
>> +
>> +    q = bdev_get_queue(bio->bi_bdev);
>> +    if (unlikely(test_bit(QUEUE_FLAG_QOS_ENABLED, &q->queue_flags)) &&
>> +            q->rq_qos)
> This unlinkey doesn't make sense, BIO_QOS_THROTTLED or BIO_QOS_MERGED
> already indicates rq_qos is enabled while issuing IO, and rq_qos should
> still be valid until this IO is done.
> 
Yes good point! Agreed that having BIO_QOS_THROTTLED or BIO_QOS_MERGED set,
implicitly implies that q->rq_qos must not be NULL.

> Perhaps a prep cleanup patch to remove this checking?
Yes I'd do that but lets wait for Jens or Ming if in case they have any
further comment before I spin v3.

Thanks,
--Nilay