[PATCH for-next 0/2] small tw add improvements

[PATCH for-next 0/2] small tw add improvements

[Pavel Begunkov]

Fiddle with inlining of task_work add so the compiler generates more
concise and efficient code.

Pavel Begunkov (2):
  io_uring: inline io_req_task_work_add()
  io_uring: split tw fallback into a function

 io_uring/io_uring.c | 32 ++++++++++++++++----------------
 io_uring/io_uring.h |  7 ++++++-
 2 files changed, 22 insertions(+), 17 deletions(-)

-- 
2.38.1

[PATCH for-next 1/2] io_uring: inline io_req_task_work_add()

[Pavel Begunkov]

__io_req_task_work_add() is huge but marked inline, that makes compilers
to generate lots of garbage. Inline the wrapper caller
io_req_task_work_add() instead.

before and after:
   text    data     bss     dec     hex filename
  47347   16248       8   63603    f873 io_uring/io_uring.o
   text    data     bss     dec     hex filename
  45303   16248       8   61559    f077 io_uring/io_uring.o

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 io_uring/io_uring.c | 7 +------
 io_uring/io_uring.h | 7 ++++++-
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 19a17d319901..f4420de6ee8b 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -1117,7 +1117,7 @@ static void io_req_local_work_add(struct io_kiocb *req)
 	__io_cqring_wake(ctx);
 }
 
-static inline void __io_req_task_work_add(struct io_kiocb *req, bool allow_local)
+void __io_req_task_work_add(struct io_kiocb *req, bool allow_local)
 {
 	struct io_uring_task *tctx = req->task->io_uring;
 	struct io_ring_ctx *ctx = req->ctx;
@@ -1149,11 +1149,6 @@ static inline void __io_req_task_work_add(struct io_kiocb *req, bool allow_local
 	}
 }
 
-void io_req_task_work_add(struct io_kiocb *req)
-{
-	__io_req_task_work_add(req, true);
-}
-
 static void __cold io_move_task_work_from_local(struct io_ring_ctx *ctx)
 {
 	struct llist_node *node;
diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h
index d14534a2f8e7..0b0620e2bf4b 100644
--- a/io_uring/io_uring.h
+++ b/io_uring/io_uring.h
@@ -48,9 +48,9 @@ static inline bool io_req_ffs_set(struct io_kiocb *req)
 	return req->flags & REQ_F_FIXED_FILE;
 }
 
+void __io_req_task_work_add(struct io_kiocb *req, bool allow_local);
 bool io_is_uring_fops(struct file *file);
 bool io_alloc_async_data(struct io_kiocb *req);
-void io_req_task_work_add(struct io_kiocb *req);
 void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags);
 void io_req_task_queue(struct io_kiocb *req);
 void io_queue_iowq(struct io_kiocb *req, bool *dont_use);
@@ -80,6 +80,11 @@ bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
 bool io_match_task_safe(struct io_kiocb *head, struct task_struct *task,
 			bool cancel_all);
 
+static inline void io_req_task_work_add(struct io_kiocb *req)
+{
+	__io_req_task_work_add(req, true);
+}
+
 #define io_for_each_link(pos, head) \
 	for (pos = (head); pos; pos = pos->link)
 
-- 
2.38.1

[PATCH for-next 2/2] io_uring: split tw fallback into a function

[Pavel Begunkov]

When the target process is dying and so task_work_add() is not allowed
we push all task_work item to the fallback workqueue. Move the part
responsible for moving tw items out of __io_req_task_work_add() into
a separate function. Makes it a bit cleaner and gives the compiler a bit
of extra info.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 io_uring/io_uring.c | 25 +++++++++++++++----------
 1 file changed, 15 insertions(+), 10 deletions(-)

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index f4420de6ee8b..d63f15afa5fd 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -1095,6 +1095,20 @@ void tctx_task_work(struct callback_head *cb)
 	trace_io_uring_task_work_run(tctx, count, loops);
 }
 
+static __cold void io_fallback_tw(struct io_uring_task *tctx)
+{
+	struct llist_node *node = llist_del_all(&tctx->task_list);
+	struct io_kiocb *req;
+
+	while (node) {
+		req = container_of(node, struct io_kiocb, io_task_work.node);
+		node = node->next;
+		if (llist_add(&req->io_task_work.node,
+			      &req->ctx->fallback_llist))
+			schedule_delayed_work(&req->ctx->fallback_work, 1);
+	}
+}
+
 static void io_req_local_work_add(struct io_kiocb *req)
 {
 	struct io_ring_ctx *ctx = req->ctx;
@@ -1121,7 +1135,6 @@ void __io_req_task_work_add(struct io_kiocb *req, bool allow_local)
 {
 	struct io_uring_task *tctx = req->task->io_uring;
 	struct io_ring_ctx *ctx = req->ctx;
-	struct llist_node *node;
 
 	if (allow_local && ctx->flags & IORING_SETUP_DEFER_TASKRUN) {
 		io_req_local_work_add(req);
@@ -1138,15 +1151,7 @@ void __io_req_task_work_add(struct io_kiocb *req, bool allow_local)
 	if (likely(!task_work_add(req->task, &tctx->task_work, ctx->notify_method)))
 		return;
 
-	node = llist_del_all(&tctx->task_list);
-
-	while (node) {
-		req = container_of(node, struct io_kiocb, io_task_work.node);
-		node = node->next;
-		if (llist_add(&req->io_task_work.node,
-			      &req->ctx->fallback_llist))
-			schedule_delayed_work(&req->ctx->fallback_work, 1);
-	}
+	io_fallback_tw(tctx);
 }
 
 static void __cold io_move_task_work_from_local(struct io_ring_ctx *ctx)
-- 
2.38.1

Re: [PATCH for-next 0/2] small tw add improvements

[Jens Axboe]

On Fri, 11 Nov 2022 16:54:07 +0000, Pavel Begunkov wrote:
> Fiddle with inlining of task_work add so the compiler generates more
> concise and efficient code.
> 
> Pavel Begunkov (2):
>   io_uring: inline io_req_task_work_add()
>   io_uring: split tw fallback into a function
> 
> [...]

Applied, thanks!

[1/2] io_uring: inline io_req_task_work_add()
      commit: 912f3f541dd8fb9e987e4bef0f2466b5233ec480
[2/2] io_uring: split tw fallback into a function
      commit: 79fde04791f9a6789ff1a9b90b06e754efb55bd7

Best regards,
-- 
Jens Axboe