> 2021年06月22日 14:55，Christoph Hellwig <hch@xxxxxx> 写道：
>
> On Wed, Jun 09, 2021 at 08:45:56AM +0800, Ming Lei wrote:
>> diff --git a/block/blk-merge.c b/block/blk-merge.c
>> index bcdff1879c34..65210e9a8efa 100644
>> --- a/block/blk-merge.c
>> +++ b/block/blk-merge.c
>> @@ -724,10 +724,10 @@ static inline bool blk_discard_mergable(struct request *req)
>> static enum elv_merge blk_try_req_merge(struct request *req,
>> struct request *next)
>> {
>> - if (blk_discard_mergable(req))
>> - return ELEVATOR_DISCARD_MERGE;
>> - else if (blk_rq_pos(req) + blk_rq_sectors(req) == blk_rq_pos(next))
>> + if (blk_rq_pos(req) + blk_rq_sectors(req) == blk_rq_pos(next))
>> return ELEVATOR_BACK_MERGE;
>
> Oh well, this breaks my previous suggestions.
>
>> + struct req_discard_range r;
>> +
>> + rq_for_each_discard_range(r, req) {
>
> ... and I can't say I like this rather complex iterator.
>
> What is the problem of just fixing the raid code to not submit stupidly
> small discard requests instead?
Yes, raid code should indeed be fixed. However, the way discard
requests are getting merged is not correct.
I understand your concern about complex iterator. I believe it is because
the difference between discard requests and normal read/write requests.
Discard requests are without data and allowed to be non-contiguous, while
normal requests, which blk-merge was originally dealing with, are with
data buffers and should be contiguous. Actually, correct me if wrong, the
current request struct lacks the ability to express non-contiguous bio's.
Ming's patch, by introducing the rq_for_each_discard_range iterator,
partially enables the request struct to express non-contiguous bio's.
Cheers,
Miao Wang
