On 10/6/21 11:55 AM, Bart Van Assche wrote:
> On 10/6/21 9:35 AM, Jens Axboe wrote:
>> Particularly for NVMe with efficient deferred submission for many
>> requests, there are nice benefits to be seen by bumping the default max
>> plug count from 16 to 32. This is especially true for virtualized setups,
>> where the submit part is more expensive. But can be noticed even on
>> native hardware.
>>
>> Reduce the multiple queue factor from 4 to 2, since we're changing the
>> default size.
>>
>> Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
>> ---
>> block/blk-mq.c | 4 ++--
>> include/linux/blkdev.h | 2 +-
>> 2 files changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/block/blk-mq.c b/block/blk-mq.c
>> index a40c94505680..5327abbefbab 100644
>> --- a/block/blk-mq.c
>> +++ b/block/blk-mq.c
>> @@ -2145,14 +2145,14 @@ static void blk_add_rq_to_plug(struct blk_plug *plug, struct request *rq)
>> }
>>
>> /*
>> - * Allow 4x BLK_MAX_REQUEST_COUNT requests on plug queue for multiple
>> + * Allow 2x BLK_MAX_REQUEST_COUNT requests on plug queue for multiple
>> * queues. This is important for md arrays to benefit from merging
>> * requests.
>> */
>> static inline unsigned short blk_plug_max_rq_count(struct blk_plug *plug)
>> {
>> if (plug->multiple_queues)
>> - return BLK_MAX_REQUEST_COUNT * 4;
>> + return BLK_MAX_REQUEST_COUNT * 2;
>> return BLK_MAX_REQUEST_COUNT;
>> }
>>
>> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
>> index b19172db7eef..534298ac73cc 100644
>> --- a/include/linux/blkdev.h
>> +++ b/include/linux/blkdev.h
>> @@ -727,7 +727,7 @@ struct blk_plug {
>> bool multiple_queues;
>> bool nowait;
>> };
>> -#define BLK_MAX_REQUEST_COUNT 16
>> +#define BLK_MAX_REQUEST_COUNT 32
>> #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
>>
>> struct blk_plug_cb;
>
> Since BLK_MAX_REQUEST_COUNT is only used inside the block layer core but
> not by any block driver, can it be moved from include/linux/blkdev.h
> into block/blk-mq.h?
Good point, I'll move it in there as part of the patch.
--
Jens Axboe
