From: Christoph Hellwig <[email protected]>
To: Jens Axboe <[email protected]>
Cc: "Michael S. Tsirkin" <[email protected]>,
Jason Wang <[email protected]>, Keith Busch <[email protected]>,
Sagi Grimberg <[email protected]>,
Pavel Begunkov <[email protected]>,
[email protected], [email protected],
[email protected], [email protected]
Subject: [PATCH 1/6] nvme-pci: reverse request order in nvme_queue_rqs
Date: Wed, 13 Nov 2024 16:20:41 +0100 [thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>
blk_mq_flush_plug_list submits requests in the reverse order that they
were submitted, which leads to a rather suboptimal I/O pattern especially
in rotational devices. Fix this by rewriting nvme_queue_rqs so that it
always pops the requests from the passed in request list, and then adds
them to the head of a local submit list. This actually simplifies the
code a bit as it removes the complicated list splicing, at the cost of
extra updates of the rq_next pointer. As that should be cache hot
anyway it should be an easy price to pay.
Fixes: d62cbcf62f2f ("nvme: add support for mq_ops->queue_rqs()")
Signed-off-by: Christoph Hellwig <[email protected]>
---
drivers/nvme/host/pci.c | 39 +++++++++++++++++----------------------
1 file changed, 17 insertions(+), 22 deletions(-)
diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index 4b9fda0b1d9a..c66ea2c23690 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -904,9 +904,10 @@ static blk_status_t nvme_queue_rq(struct blk_mq_hw_ctx *hctx,
static void nvme_submit_cmds(struct nvme_queue *nvmeq, struct request **rqlist)
{
+ struct request *req;
+
spin_lock(&nvmeq->sq_lock);
- while (!rq_list_empty(*rqlist)) {
- struct request *req = rq_list_pop(rqlist);
+ while ((req = rq_list_pop(rqlist))) {
struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
nvme_sq_copy_cmd(nvmeq, &iod->cmd);
@@ -931,31 +932,25 @@ static bool nvme_prep_rq_batch(struct nvme_queue *nvmeq, struct request *req)
static void nvme_queue_rqs(struct request **rqlist)
{
- struct request *req, *next, *prev = NULL;
+ struct request *submit_list = NULL;
struct request *requeue_list = NULL;
+ struct request **requeue_lastp = &requeue_list;
+ struct nvme_queue *nvmeq = NULL;
+ struct request *req;
- rq_list_for_each_safe(rqlist, req, next) {
- struct nvme_queue *nvmeq = req->mq_hctx->driver_data;
-
- if (!nvme_prep_rq_batch(nvmeq, req)) {
- /* detach 'req' and add to remainder list */
- rq_list_move(rqlist, &requeue_list, req, prev);
-
- req = prev;
- if (!req)
- continue;
- }
+ while ((req = rq_list_pop(rqlist))) {
+ if (nvmeq && nvmeq != req->mq_hctx->driver_data)
+ nvme_submit_cmds(nvmeq, &submit_list);
+ nvmeq = req->mq_hctx->driver_data;
- if (!next || req->mq_hctx != next->mq_hctx) {
- /* detach rest of list, and submit */
- req->rq_next = NULL;
- nvme_submit_cmds(nvmeq, rqlist);
- *rqlist = next;
- prev = NULL;
- } else
- prev = req;
+ if (nvme_prep_rq_batch(nvmeq, req))
+ rq_list_add(&submit_list, req); /* reverse order */
+ else
+ rq_list_add_tail(&requeue_lastp, req);
}
+ if (nvmeq)
+ nvme_submit_cmds(nvmeq, &submit_list);
*rqlist = requeue_list;
}
--
2.45.2
next prev parent reply other threads:[~2024-11-13 15:21 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-11-13 15:20 don't reorder requests passed to ->queue_rqs Christoph Hellwig
2024-11-13 15:20 ` Christoph Hellwig [this message]
2024-11-13 19:10 ` [PATCH 1/6] nvme-pci: reverse request order in nvme_queue_rqs Keith Busch
2024-11-13 15:20 ` [PATCH 2/6] virtio_blk: reverse request order in virtio_queue_rqs Christoph Hellwig
2024-11-13 19:03 ` Keith Busch
2024-11-13 19:05 ` Jens Axboe
2024-11-13 23:25 ` Michael S. Tsirkin
2024-11-13 15:20 ` [PATCH 3/6] block: remove rq_list_move Christoph Hellwig
2024-11-13 15:20 ` [PATCH 4/6] block: add a rq_list type Christoph Hellwig
2024-11-14 20:11 ` Nathan Chancellor
2024-11-15 9:10 ` Christoph Hellwig
2024-11-15 12:49 ` Jens Axboe
2024-11-15 19:38 ` Jens Axboe
2024-11-16 0:56 ` Nathan Chancellor
2024-11-13 15:20 ` [PATCH 5/6] block: don't reorder requests in blk_add_rq_to_plug Christoph Hellwig
2024-11-13 15:20 ` [PATCH 6/6] block: don't reorder requests in blk_mq_add_to_batch Christoph Hellwig
2024-11-13 18:33 ` don't reorder requests passed to ->queue_rqs Bart Van Assche
2024-11-13 18:39 ` Jens Axboe
2024-11-13 18:46 ` Jens Axboe
2024-11-13 20:36 ` Chaitanya Kulkarni
2024-11-13 20:51 ` Jens Axboe
2024-11-13 22:23 ` Chaitanya Kulkarni
2024-11-13 22:27 ` Jens Axboe
2024-11-14 4:16 ` Christoph Hellwig
2024-11-14 15:14 ` Jens Axboe
2024-11-18 15:23 ` Christoph Hellwig
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
[email protected] \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox