public inbox for [email protected]
 help / color / mirror / Atom feed
* [PATCH for-next 0/3] sendzc dedup / simplification
@ 2024-04-07 23:54 Pavel Begunkov
  2024-04-07 23:54 ` [PATCH for-next 1/3] io_uring/net: merge ubuf sendzc callbacks Pavel Begunkov
                   ` (3 more replies)
  0 siblings, 4 replies; 5+ messages in thread
From: Pavel Begunkov @ 2024-04-07 23:54 UTC (permalink / raw)
  To: io-uring; +Cc: Jens Axboe, asml.silence

Consolidate two pairs of duplicated callbacks for zerocopy send, which
was a premature optimisation, and it's cleaner and simpler to keep
them unified.

Pavel Begunkov (3):
  io_uring/net: merge ubuf sendzc callbacks
  io_uring/net: get rid of io_notif_complete_tw_ext
  io_uring/net: set MSG_ZEROCOPY for sendzc in advance

 io_uring/net.c   | 16 ++++++++--------
 io_uring/notif.c | 32 +++++++-------------------------
 io_uring/notif.h |  6 ++++--
 3 files changed, 19 insertions(+), 35 deletions(-)

-- 
2.44.0


^ permalink raw reply	[flat|nested] 5+ messages in thread

* [PATCH for-next 1/3] io_uring/net: merge ubuf sendzc callbacks
  2024-04-07 23:54 [PATCH for-next 0/3] sendzc dedup / simplification Pavel Begunkov
@ 2024-04-07 23:54 ` Pavel Begunkov
  2024-04-07 23:54 ` [PATCH for-next 2/3] io_uring/net: get rid of io_notif_complete_tw_ext Pavel Begunkov
                   ` (2 subsequent siblings)
  3 siblings, 0 replies; 5+ messages in thread
From: Pavel Begunkov @ 2024-04-07 23:54 UTC (permalink / raw)
  To: io-uring; +Cc: Jens Axboe, asml.silence

Splitting io_tx_ubuf_callback_ext from io_tx_ubuf_callback is a pre
mature optimisation that doesn't give us much. Merge the functions into
one and reclaim some simplicity back.

Signed-off-by: Pavel Begunkov <[email protected]>
---
 io_uring/notif.c | 26 ++++++++------------------
 1 file changed, 8 insertions(+), 18 deletions(-)

diff --git a/io_uring/notif.c b/io_uring/notif.c
index d3e703c37aba..47ff2da8c421 100644
--- a/io_uring/notif.c
+++ b/io_uring/notif.c
@@ -30,36 +30,24 @@ static void io_tx_ubuf_callback(struct sk_buff *skb, struct ubuf_info *uarg,
 	struct io_notif_data *nd = container_of(uarg, struct io_notif_data, uarg);
 	struct io_kiocb *notif = cmd_to_io_kiocb(nd);
 
-	if (refcount_dec_and_test(&uarg->refcnt))
-		__io_req_task_work_add(notif, IOU_F_TWQ_LAZY_WAKE);
-}
-
-static void io_tx_ubuf_callback_ext(struct sk_buff *skb, struct ubuf_info *uarg,
-			     bool success)
-{
-	struct io_notif_data *nd = container_of(uarg, struct io_notif_data, uarg);
-
 	if (nd->zc_report) {
 		if (success && !nd->zc_used && skb)
 			WRITE_ONCE(nd->zc_used, true);
 		else if (!success && !nd->zc_copied)
 			WRITE_ONCE(nd->zc_copied, true);
 	}
-	io_tx_ubuf_callback(skb, uarg, success);
+
+	if (refcount_dec_and_test(&uarg->refcnt))
+		__io_req_task_work_add(notif, IOU_F_TWQ_LAZY_WAKE);
 }
 
 void io_notif_set_extended(struct io_kiocb *notif)
 {
 	struct io_notif_data *nd = io_notif_to_data(notif);
 
-	if (nd->uarg.callback != io_tx_ubuf_callback_ext) {
-		nd->account_pages = 0;
-		nd->zc_report = false;
-		nd->zc_used = false;
-		nd->zc_copied = false;
-		nd->uarg.callback = io_tx_ubuf_callback_ext;
-		notif->io_task_work.func = io_notif_complete_tw_ext;
-	}
+	nd->zc_used = false;
+	nd->zc_copied = false;
+	notif->io_task_work.func = io_notif_complete_tw_ext;
 }
 
 struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx)
@@ -79,6 +67,8 @@ struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx)
 	notif->io_task_work.func = io_req_task_complete;
 
 	nd = io_notif_to_data(notif);
+	nd->zc_report = false;
+	nd->account_pages = 0;
 	nd->uarg.flags = IO_NOTIF_UBUF_FLAGS;
 	nd->uarg.callback = io_tx_ubuf_callback;
 	refcount_set(&nd->uarg.refcnt, 1);
-- 
2.44.0


^ permalink raw reply related	[flat|nested] 5+ messages in thread

* [PATCH for-next 2/3] io_uring/net: get rid of io_notif_complete_tw_ext
  2024-04-07 23:54 [PATCH for-next 0/3] sendzc dedup / simplification Pavel Begunkov
  2024-04-07 23:54 ` [PATCH for-next 1/3] io_uring/net: merge ubuf sendzc callbacks Pavel Begunkov
@ 2024-04-07 23:54 ` Pavel Begunkov
  2024-04-07 23:54 ` [PATCH for-next 3/3] io_uring/net: set MSG_ZEROCOPY for sendzc in advance Pavel Begunkov
  2024-04-09 21:00 ` [PATCH for-next 0/3] sendzc dedup / simplification Jens Axboe
  3 siblings, 0 replies; 5+ messages in thread
From: Pavel Begunkov @ 2024-04-07 23:54 UTC (permalink / raw)
  To: io-uring; +Cc: Jens Axboe, asml.silence

io_notif_complete_tw_ext() can be removed and combined with
io_notif_complete_tw to make it simpler without sacrificing
anything.

Signed-off-by: Pavel Begunkov <[email protected]>
---
 io_uring/net.c   | 10 +++++-----
 io_uring/notif.c | 18 +++++-------------
 io_uring/notif.h |  6 ++++--
 3 files changed, 14 insertions(+), 20 deletions(-)

diff --git a/io_uring/net.c b/io_uring/net.c
index 27b853033d6f..f97014566b52 100644
--- a/io_uring/net.c
+++ b/io_uring/net.c
@@ -1019,8 +1019,11 @@ int io_send_zc_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 		if (zc->flags & ~IO_ZC_FLAGS_VALID)
 			return -EINVAL;
 		if (zc->flags & IORING_SEND_ZC_REPORT_USAGE) {
-			io_notif_set_extended(notif);
-			io_notif_to_data(notif)->zc_report = true;
+			struct io_notif_data *nd = io_notif_to_data(notif);
+
+			nd->zc_report = true;
+			nd->zc_used = false;
+			nd->zc_copied = false;
 		}
 	}
 
@@ -1129,7 +1132,6 @@ static int io_send_zc_import(struct io_kiocb *req, struct io_async_msghdr *kmsg)
 			return ret;
 		kmsg->msg.sg_from_iter = io_sg_from_iter;
 	} else {
-		io_notif_set_extended(sr->notif);
 		ret = import_ubuf(ITER_SOURCE, sr->buf, sr->len, &kmsg->msg.msg_iter);
 		if (unlikely(ret))
 			return ret;
@@ -1218,8 +1220,6 @@ int io_sendmsg_zc(struct io_kiocb *req, unsigned int issue_flags)
 	unsigned flags;
 	int ret, min_ret = 0;
 
-	io_notif_set_extended(sr->notif);
-
 	sock = sock_from_file(req->file);
 	if (unlikely(!sock))
 		return -ENOTSOCK;
diff --git a/io_uring/notif.c b/io_uring/notif.c
index 47ff2da8c421..b561bd763435 100644
--- a/io_uring/notif.c
+++ b/io_uring/notif.c
@@ -9,12 +9,12 @@
 #include "notif.h"
 #include "rsrc.h"
 
-static void io_notif_complete_tw_ext(struct io_kiocb *notif, struct io_tw_state *ts)
+void io_notif_tw_complete(struct io_kiocb *notif, struct io_tw_state *ts)
 {
 	struct io_notif_data *nd = io_notif_to_data(notif);
 	struct io_ring_ctx *ctx = notif->ctx;
 
-	if (nd->zc_report && (nd->zc_copied || !nd->zc_used))
+	if (unlikely(nd->zc_report) && (nd->zc_copied || !nd->zc_used))
 		notif->cqe.res |= IORING_NOTIF_USAGE_ZC_COPIED;
 
 	if (nd->account_pages && ctx->user) {
@@ -37,17 +37,10 @@ static void io_tx_ubuf_callback(struct sk_buff *skb, struct ubuf_info *uarg,
 			WRITE_ONCE(nd->zc_copied, true);
 	}
 
-	if (refcount_dec_and_test(&uarg->refcnt))
+	if (refcount_dec_and_test(&uarg->refcnt)) {
+		notif->io_task_work.func = io_notif_tw_complete;
 		__io_req_task_work_add(notif, IOU_F_TWQ_LAZY_WAKE);
-}
-
-void io_notif_set_extended(struct io_kiocb *notif)
-{
-	struct io_notif_data *nd = io_notif_to_data(notif);
-
-	nd->zc_used = false;
-	nd->zc_copied = false;
-	notif->io_task_work.func = io_notif_complete_tw_ext;
+	}
 }
 
 struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx)
@@ -64,7 +57,6 @@ struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx)
 	notif->task = current;
 	io_get_task_refs(1);
 	notif->rsrc_node = NULL;
-	notif->io_task_work.func = io_req_task_complete;
 
 	nd = io_notif_to_data(notif);
 	nd->zc_report = false;
diff --git a/io_uring/notif.h b/io_uring/notif.h
index 86d32bd9f856..52e124a9957c 100644
--- a/io_uring/notif.h
+++ b/io_uring/notif.h
@@ -20,7 +20,7 @@ struct io_notif_data {
 };
 
 struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx);
-void io_notif_set_extended(struct io_kiocb *notif);
+void io_notif_tw_complete(struct io_kiocb *notif, struct io_tw_state *ts);
 
 static inline struct io_notif_data *io_notif_to_data(struct io_kiocb *notif)
 {
@@ -33,8 +33,10 @@ static inline void io_notif_flush(struct io_kiocb *notif)
 	struct io_notif_data *nd = io_notif_to_data(notif);
 
 	/* drop slot's master ref */
-	if (refcount_dec_and_test(&nd->uarg.refcnt))
+	if (refcount_dec_and_test(&nd->uarg.refcnt)) {
+		notif->io_task_work.func = io_notif_tw_complete;
 		__io_req_task_work_add(notif, IOU_F_TWQ_LAZY_WAKE);
+	}
 }
 
 static inline int io_notif_account_mem(struct io_kiocb *notif, unsigned len)
-- 
2.44.0


^ permalink raw reply related	[flat|nested] 5+ messages in thread

* [PATCH for-next 3/3] io_uring/net: set MSG_ZEROCOPY for sendzc in advance
  2024-04-07 23:54 [PATCH for-next 0/3] sendzc dedup / simplification Pavel Begunkov
  2024-04-07 23:54 ` [PATCH for-next 1/3] io_uring/net: merge ubuf sendzc callbacks Pavel Begunkov
  2024-04-07 23:54 ` [PATCH for-next 2/3] io_uring/net: get rid of io_notif_complete_tw_ext Pavel Begunkov
@ 2024-04-07 23:54 ` Pavel Begunkov
  2024-04-09 21:00 ` [PATCH for-next 0/3] sendzc dedup / simplification Jens Axboe
  3 siblings, 0 replies; 5+ messages in thread
From: Pavel Begunkov @ 2024-04-07 23:54 UTC (permalink / raw)
  To: io-uring; +Cc: Jens Axboe, asml.silence

We can set MSG_ZEROCOPY at the preparation step, do it so we don't have
to care about it later in the issue callback.

Signed-off-by: Pavel Begunkov <[email protected]>
---
 io_uring/net.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/io_uring/net.c b/io_uring/net.c
index f97014566b52..5e153a3f89b1 100644
--- a/io_uring/net.c
+++ b/io_uring/net.c
@@ -1051,7 +1051,7 @@ int io_send_zc_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 
 	zc->buf = u64_to_user_ptr(READ_ONCE(sqe->addr));
 	zc->len = READ_ONCE(sqe->len);
-	zc->msg_flags = READ_ONCE(sqe->msg_flags) | MSG_NOSIGNAL;
+	zc->msg_flags = READ_ONCE(sqe->msg_flags) | MSG_NOSIGNAL | MSG_ZEROCOPY;
 	if (zc->msg_flags & MSG_DONTWAIT)
 		req->flags |= REQ_F_NOWAIT;
 
@@ -1168,7 +1168,7 @@ int io_send_zc(struct io_kiocb *req, unsigned int issue_flags)
 			return ret;
 	}
 
-	msg_flags = zc->msg_flags | MSG_ZEROCOPY;
+	msg_flags = zc->msg_flags;
 	if (issue_flags & IO_URING_F_NONBLOCK)
 		msg_flags |= MSG_DONTWAIT;
 	if (msg_flags & MSG_WAITALL)
@@ -1230,7 +1230,7 @@ int io_sendmsg_zc(struct io_kiocb *req, unsigned int issue_flags)
 	    (sr->flags & IORING_RECVSEND_POLL_FIRST))
 		return -EAGAIN;
 
-	flags = sr->msg_flags | MSG_ZEROCOPY;
+	flags = sr->msg_flags;
 	if (issue_flags & IO_URING_F_NONBLOCK)
 		flags |= MSG_DONTWAIT;
 	if (flags & MSG_WAITALL)
-- 
2.44.0


^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: [PATCH for-next 0/3] sendzc dedup / simplification
  2024-04-07 23:54 [PATCH for-next 0/3] sendzc dedup / simplification Pavel Begunkov
                   ` (2 preceding siblings ...)
  2024-04-07 23:54 ` [PATCH for-next 3/3] io_uring/net: set MSG_ZEROCOPY for sendzc in advance Pavel Begunkov
@ 2024-04-09 21:00 ` Jens Axboe
  3 siblings, 0 replies; 5+ messages in thread
From: Jens Axboe @ 2024-04-09 21:00 UTC (permalink / raw)
  To: io-uring, Pavel Begunkov


On Mon, 08 Apr 2024 00:54:54 +0100, Pavel Begunkov wrote:
> Consolidate two pairs of duplicated callbacks for zerocopy send, which
> was a premature optimisation, and it's cleaner and simpler to keep
> them unified.
> 
> Pavel Begunkov (3):
>   io_uring/net: merge ubuf sendzc callbacks
>   io_uring/net: get rid of io_notif_complete_tw_ext
>   io_uring/net: set MSG_ZEROCOPY for sendzc in advance
> 
> [...]

Applied, thanks!

[1/3] io_uring/net: merge ubuf sendzc callbacks
      (no commit info)
[2/3] io_uring/net: get rid of io_notif_complete_tw_ext
      (no commit info)
[3/3] io_uring/net: set MSG_ZEROCOPY for sendzc in advance
      (no commit info)

Best regards,
-- 
Jens Axboe




^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2024-04-09 21:00 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-04-07 23:54 [PATCH for-next 0/3] sendzc dedup / simplification Pavel Begunkov
2024-04-07 23:54 ` [PATCH for-next 1/3] io_uring/net: merge ubuf sendzc callbacks Pavel Begunkov
2024-04-07 23:54 ` [PATCH for-next 2/3] io_uring/net: get rid of io_notif_complete_tw_ext Pavel Begunkov
2024-04-07 23:54 ` [PATCH for-next 3/3] io_uring/net: set MSG_ZEROCOPY for sendzc in advance Pavel Begunkov
2024-04-09 21:00 ` [PATCH for-next 0/3] sendzc dedup / simplification Jens Axboe

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox