[PATCH stable-6.6 0/3] provided buffer recycling fixes

[PATCH stable-6.6 0/3] provided buffer recycling fixes

[Pavel Begunkov]

Fixes for the provided buffers for not allowing kbufs to cross a single
execution section. Upstream had most of it already fixed by chance,
which is why all 3 patches refer to a single upstream commit.

Pavel Begunkov (3):
  io_uring: fix multishots with selected buffers
  io_uring: fix io_req_prep_async with provided buffers
  io_uring/rw: commit provided buffer state on async

 io_uring/io_uring.c |  5 ++++-
 io_uring/poll.c     |  2 ++
 io_uring/rw.c       | 10 ++++++++++
 3 files changed, 16 insertions(+), 1 deletion(-)

-- 
2.47.1

[PATCH stable-6.6 1/3] io_uring: fix multishots with selected buffers

[Pavel Begunkov]

[ upstream commit d63b0e8a628e62ca85a0f7915230186bb92f8bb4 ]

We do io_kbuf_recycle() when arming a poll but every iteration of a
multishot can grab more buffers, which is why we need to flush the kbuf
ring state before continuing with waiting.

Cc: stable@vger.kernel.org
Fixes: b3fdea6ecb55c ("io_uring: multishot recv")
Reported-by: Muhammad Ramdhan <ramdhan@starlabs.sg>
Reported-by: Bing-Jhong Billy Jheng <billy@starlabs.sg>
Reported-by: Jacob Soo <jacob.soo@starlabs.sg>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/1bfc9990fe435f1fc6152ca9efeba5eb3e68339c.1738025570.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 io_uring/poll.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/io_uring/poll.c b/io_uring/poll.c
index 5cf4fffe8b6c..2824a3560245 100644
--- a/io_uring/poll.c
+++ b/io_uring/poll.c
@@ -350,8 +350,10 @@ void io_poll_task_func(struct io_kiocb *req, struct io_tw_state *ts)
 
 	ret = io_poll_check_events(req, ts);
 	if (ret == IOU_POLL_NO_ACTION) {
+		io_kbuf_recycle(req, 0);
 		return;
 	} else if (ret == IOU_POLL_REQUEUE) {
+		io_kbuf_recycle(req, 0);
 		__io_poll_execute(req, 0);
 		return;
 	}
-- 
2.47.1

Re: [PATCH stable-6.6 1/3] io_uring: fix multishots with selected buffers

[Greg KH]

On Mon, Feb 10, 2025 at 03:21:36PM +0000, Pavel Begunkov wrote:
> [ upstream commit d63b0e8a628e62ca85a0f7915230186bb92f8bb4 ]
> 
> We do io_kbuf_recycle() when arming a poll but every iteration of a
> multishot can grab more buffers, which is why we need to flush the kbuf
> ring state before continuing with waiting.
> 
> Cc: stable@vger.kernel.org
> Fixes: b3fdea6ecb55c ("io_uring: multishot recv")
> Reported-by: Muhammad Ramdhan <ramdhan@starlabs.sg>
> Reported-by: Bing-Jhong Billy Jheng <billy@starlabs.sg>
> Reported-by: Jacob Soo <jacob.soo@starlabs.sg>
> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
> Link: https://lore.kernel.org/r/1bfc9990fe435f1fc6152ca9efeba5eb3e68339c.1738025570.git.asml.silence@gmail.com
> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> ---
>  io_uring/poll.c | 2 ++
>  1 file changed, 2 insertions(+)

This is already in our 6.6.y queue, thanks!

greg k-h

[PATCH stable-6.6 2/3] io_uring: fix io_req_prep_async with provided buffers

[Pavel Begunkov]

[ upstream commit d63b0e8a628e62ca85a0f7915230186bb92f8bb4 ]

io_req_prep_async() can import provided buffers, commit the ring state
by giving up on that before, it'll be reimported later if needed.

Reported-by: Muhammad Ramdhan <ramdhan@starlabs.sg>
Reported-by: Bing-Jhong Billy Jheng <billy@starlabs.sg>
Reported-by: Jacob Soo <jacob.soo@starlabs.sg>
Fixes: c7fb19428d67d ("io_uring: add support for ring mapped supplied buffers")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 io_uring/io_uring.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index c7198fbcf734..b61637dad442 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -1779,6 +1779,7 @@ int io_req_prep_async(struct io_kiocb *req)
 {
 	const struct io_cold_def *cdef = &io_cold_defs[req->opcode];
 	const struct io_issue_def *def = &io_issue_defs[req->opcode];
+	int ret;
 
 	/* assign early for deferred execution for non-fixed file */
 	if (def->needs_file && !(req->flags & REQ_F_FIXED_FILE) && !req->file)
@@ -1791,7 +1792,9 @@ int io_req_prep_async(struct io_kiocb *req)
 		if (io_alloc_async_data(req))
 			return -EAGAIN;
 	}
-	return cdef->prep_async(req);
+	ret = cdef->prep_async(req);
+	io_kbuf_recycle(req, 0);
+	return ret;
 }
 
 static u32 io_get_sequence(struct io_kiocb *req)
-- 
2.47.1

Re: [PATCH stable-6.6 2/3] io_uring: fix io_req_prep_async with provided buffers

[Greg KH]

On Mon, Feb 10, 2025 at 03:21:37PM +0000, Pavel Begunkov wrote:
> [ upstream commit d63b0e8a628e62ca85a0f7915230186bb92f8bb4 ]

Nope, wrong git id :(

Can you redo this series with the right one?

thanks,

greg k-h

[PATCH stable-6.6 3/3] io_uring/rw: commit provided buffer state on async

[Pavel Begunkov]

[ upstream commit d63b0e8a628e62ca85a0f7915230186bb92f8bb4 ]

When we get -EIOCBQUEUED, we need to ensure that the buffer is consumed
from the provided buffer ring, which can be done with io_kbuf_recycle()
+ REQ_F_PARTIAL_IO.

Reported-by: Muhammad Ramdhan <ramdhan@starlabs.sg>
Reported-by: Bing-Jhong Billy Jheng <billy@starlabs.sg>
Reported-by: Jacob Soo <jacob.soo@starlabs.sg>
Fixes: c7fb19428d67d ("io_uring: add support for ring mapped supplied buffers")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 io_uring/rw.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/io_uring/rw.c b/io_uring/rw.c
index a62f84e28bac..75b001febb4d 100644
--- a/io_uring/rw.c
+++ b/io_uring/rw.c
@@ -793,6 +793,8 @@ static int __io_read(struct io_kiocb *req, unsigned int issue_flags)
 			goto done;
 		ret = 0;
 	} else if (ret == -EIOCBQUEUED) {
+		req->flags |= REQ_F_PARTIAL_IO;
+		io_kbuf_recycle(req, issue_flags);
 		if (iovec)
 			kfree(iovec);
 		return IOU_ISSUE_SKIP_COMPLETE;
@@ -816,6 +818,9 @@ static int __io_read(struct io_kiocb *req, unsigned int issue_flags)
 		goto done;
 	}
 
+	req->flags |= REQ_F_PARTIAL_IO;
+	io_kbuf_recycle(req, issue_flags);
+
 	io = req->async_data;
 	s = &io->s;
 	/*
@@ -956,6 +961,11 @@ int io_write(struct io_kiocb *req, unsigned int issue_flags)
 	else
 		ret2 = -EINVAL;
 
+	if (ret2 == -EIOCBQUEUED) {
+		req->flags |= REQ_F_PARTIAL_IO;
+		io_kbuf_recycle(req, issue_flags);
+	}
+
 	if (req->flags & REQ_F_REISSUE) {
 		req->flags &= ~REQ_F_REISSUE;
 		ret2 = -EAGAIN;
-- 
2.47.1

Re: [PATCH stable-6.6 3/3] io_uring/rw: commit provided buffer state on async

[Greg KH]

On Mon, Feb 10, 2025 at 03:21:38PM +0000, Pavel Begunkov wrote:
> [ upstream commit d63b0e8a628e62ca85a0f7915230186bb92f8bb4 ]

Wrong commit id :(

Let me figure out what went wrong here...

Re: [PATCH stable-6.6 0/3] provided buffer recycling fixes

[Greg KH]

On Mon, Feb 10, 2025 at 03:21:35PM +0000, Pavel Begunkov wrote:
> Fixes for the provided buffers for not allowing kbufs to cross a single
> execution section. Upstream had most of it already fixed by chance,
> which is why all 3 patches refer to a single upstream commit.

Ah.  Ok, that makes more sense, nevermind, I should have read patch 0/X
first...

I'll drop the upstream commit reference here as it's just confusing.

thanks,

greg k-h

Re: [PATCH stable-6.6 0/3] provided buffer recycling fixes

[Pavel Begunkov]

On 2/11/25 10:18, Greg KH wrote:
> On Mon, Feb 10, 2025 at 03:21:35PM +0000, Pavel Begunkov wrote:
>> Fixes for the provided buffers for not allowing kbufs to cross a single
>> execution section. Upstream had most of it already fixed by chance,
>> which is why all 3 patches refer to a single upstream commit.
> 
> Ah.  Ok, that makes more sense, nevermind, I should have read patch 0/X
> first...
> 
> I'll drop the upstream commit reference here as it's just confusing.

Got it, thanks

-- 
Pavel Begunkov