On 5/18/23 01:30, Bart Van Assche wrote:
> On 5/16/23 18:06, Damien Le Moal wrote:
>> On 5/17/23 07:33, Bart Van Assche wrote:
>>> Make deadline_skip_seq_writes() do what its name suggests, namely to
>>> skip sequential writes.
>>>
>>> Reviewed-by: Christoph Hellwig <hch@xxxxxx>
>>> Cc: Damien Le Moal <damien.lemoal@xxxxxxxxxxxxxxxxxx>
>>> Cc: Ming Lei <ming.lei@xxxxxxxxxx>
>>> Signed-off-by: Bart Van Assche <bvanassche@xxxxxxx>
>>> ---
>>> block/mq-deadline.c | 2 +-
>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/block/mq-deadline.c b/block/mq-deadline.c
>>> index 6276afede9cd..dbc0feca963e 100644
>>> --- a/block/mq-deadline.c
>>> +++ b/block/mq-deadline.c
>>> @@ -308,7 +308,7 @@ static struct request *deadline_skip_seq_writes(struct deadline_data *dd,
>>> do {
>>> pos += blk_rq_sectors(rq);
>>> rq = deadline_latter_request(rq);
>>> - } while (rq && blk_rq_pos(rq) == pos);
>>> + } while (rq && blk_rq_pos(rq) == pos && blk_rq_is_seq_zoned_write(rq));
>>
>> No ! The "seq write" skip here is to skip writes that are contiguous/sequential
>> to ensure that we keep issuing contiguous/sequential writes belonging to
>> different zones, regardless of the target zone type.
>>
>> So drop this change please.
>
> Hi Damien,
>
> I'm fine with dropping this patch. I came up with this patch because it
> surprised me to see that deadline_skip_seq_writes() does not check the
> type of the requests that it is skipping. If e.g. a WRITE is followed by
> two contiguous READs, all three requests are skipped. Is this intentional?
Hmmm... mq-deadline is not supposed to mix up reads and writes in the same
scheduling batch, and there is one sort_list (rbtree) per data direction. So we
should not be seeing different data directions in this loop, no ?
>
> Thanks,
>
> Bart.
>
--
Damien Le Moal
Western Digital Research
