| Message ID | 20230531083828.8009-1-ed.tsai@mediatek.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | blk-mq: check on cpu id when there is only one ctx mapping | expand |
On Wed, May 31, 2023 at 4:49 PM Ed Tsai <ed.tsai@mediatek.com> wrote: > > commit f168420 ("blk-mq: don't redirect completion for hctx withs only > one ctx mapping") When nvme applies a 1:1 mapping of hctx and ctx, there > will be no remote request. > > But for ufs, the submission and completion queues could be asymmetric. > (e.g. Multiple SQs share one CQ) Therefore, 1:1 mapping of hctx and > ctx won't complete request on the submission cpu. In this situation, > this nr_ctx check could violate the QUEUE_FLAG_SAME_FORCE, as a result, > check on cpu id when there is only one ctx mapping. > > Signed-off-by: Ed Tsai <ed.tsai@mediatek.com> > Signed-off-by: Po-Wen Kao <powen.kao@mediatek.com> Reviewed-by: Stanley Chu <stanley.chu@mediatek.com>
On 5/31/23 01:38, Ed Tsai wrote: > commit f168420 ("blk-mq: don't redirect completion for hctx withs only > one ctx mapping") When nvme applies a 1:1 mapping of hctx and ctx, there > will be no remote request. The above sentence is incomprehensible. Please make it comprehensible. > Signed-off-by: Ed Tsai <ed.tsai@mediatek.com> > Signed-off-by: Po-Wen Kao <powen.kao@mediatek.com> This patch comes from Keith Busch. Why has his name not been mentioned? See also https://lore.kernel.org/linux-block/ZHY2TUrKVBj2xGE2@kbusch-mbp.dhcp.thefacebook.com/. Bart.
diff --git a/block/blk-mq.c b/block/blk-mq.c index 41f7e2b500bd..ecc30ebe9483 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -1176,7 +1176,8 @@ bool blk_mq_complete_request_remote(struct request *rq) * or a polled request, always complete locally, * it's pointless to redirect the completion. */ - if (rq->mq_hctx->nr_ctx == 1 || + if ((rq->mq_hctx->nr_ctx == 1 && + rq->mq_ctx->cpu == raw_smp_processor_id()) || rq->cmd_flags & REQ_POLLED) return false;
commit f168420 ("blk-mq: don't redirect completion for hctx withs only one ctx mapping") When nvme applies a 1:1 mapping of hctx and ctx, there will be no remote request. But for ufs, the submission and completion queues could be asymmetric. (e.g. Multiple SQs share one CQ) Therefore, 1:1 mapping of hctx and ctx won't complete request on the submission cpu. In this situation, this nr_ctx check could violate the QUEUE_FLAG_SAME_FORCE, as a result, check on cpu id when there is only one ctx mapping. Signed-off-by: Ed Tsai <ed.tsai@mediatek.com> Signed-off-by: Po-Wen Kao <powen.kao@mediatek.com> --- block/blk-mq.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)