FAILED: patch "[PATCH] io_uring: lock overflowing for IOPOLL" failed to apply to 5.15-stable tree

FAILED: patch "[PATCH] io_uring: lock overflowing for IOPOLL" failed to apply to 5.15-stable tree

[gregkh]

The patch below does not apply to the 5.15-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable@vger.kernel.org>.

Possible dependencies:

544d163d659d ("io_uring: lock overflowing for IOPOLL")
a8cf95f93610 ("io_uring: fix overflow handling regression")
fa18fa2272c7 ("io_uring: inline __io_req_complete_put()")
f9d567c75ec2 ("io_uring: inline __io_req_complete_post()")
52120f0fadcb ("io_uring: add allow_overflow to io_post_aux_cqe")
e6130eba8a84 ("io_uring: add support for passing fixed file descriptors")
253993210bd8 ("io_uring: introduce locking helpers for CQE posting")
305bef988708 ("io_uring: hide eventfd assumptions in eventfd paths")
d9dee4302a7c ("io_uring: remove ->flush_cqes optimisation")
a830ffd28780 ("io_uring: move io_eventfd_signal()")
9046c6415be6 ("io_uring: reshuffle io_uring/io_uring.h")
d142c3ec8d16 ("io_uring: remove extra io_commit_cqring()")
68494a65d0e2 ("io_uring: introduce io_req_cqe_overflow()")
faf88dde060f ("io_uring: don't inline __io_get_cqe()")
d245bca6375b ("io_uring: don't expose io_fill_cqe_aux()")
aa1e90f64ee5 ("io_uring: move small helpers to headers")
aa1e90f64ee5 ("io_uring: move small helpers to headers")
aa1e90f64ee5 ("io_uring: move small helpers to headers")
aa1e90f64ee5 ("io_uring: move small helpers to headers")
aa1e90f64ee5 ("io_uring: move small helpers to headers")

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 544d163d659d45a206d8929370d5a2984e546cb7 Mon Sep 17 00:00:00 2001
From: Pavel Begunkov <asml.silence@gmail.com>
Date: Thu, 12 Jan 2023 13:08:56 +0000
Subject: [PATCH] io_uring: lock overflowing for IOPOLL
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

syzbot reports an issue with overflow filling for IOPOLL:

WARNING: CPU: 0 PID: 28 at io_uring/io_uring.c:734 io_cqring_event_overflow+0x1c0/0x230 io_uring/io_uring.c:734
CPU: 0 PID: 28 Comm: kworker/u4:1 Not tainted 6.2.0-rc3-syzkaller-16369-g358a161a6a9e #0
Workqueue: events_unbound io_ring_exit_work
Call trace:
 io_cqring_event_overflow+0x1c0/0x230 io_uring/io_uring.c:734
 io_req_cqe_overflow+0x5c/0x70 io_uring/io_uring.c:773
 io_fill_cqe_req io_uring/io_uring.h:168 [inline]
 io_do_iopoll+0x474/0x62c io_uring/rw.c:1065
 io_iopoll_try_reap_events+0x6c/0x108 io_uring/io_uring.c:1513
 io_uring_try_cancel_requests+0x13c/0x258 io_uring/io_uring.c:3056
 io_ring_exit_work+0xec/0x390 io_uring/io_uring.c:2869
 process_one_work+0x2d8/0x504 kernel/workqueue.c:2289
 worker_thread+0x340/0x610 kernel/workqueue.c:2436
 kthread+0x12c/0x158 kernel/kthread.c:376
 ret_from_fork+0x10/0x20 arch/arm64/kernel/entry.S:863

There is no real problem for normal IOPOLL as flush is also called with
uring_lock taken, but it's getting more complicated for IOPOLL|SQPOLL,
for which __io_cqring_overflow_flush() happens from the CQ waiting path.

Reported-and-tested-by: syzbot+6805087452d72929404e@syzkaller.appspotmail.com
Cc: stable@vger.kernel.org # 5.10+
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/io_uring/rw.c b/io_uring/rw.c
index 8227af2e1c0f..9c3ddd46a1ad 100644
--- a/io_uring/rw.c
+++ b/io_uring/rw.c
@@ -1062,7 +1062,11 @@ int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin)
 			continue;
 
 		req->cqe.flags = io_put_kbuf(req, 0);
-		io_fill_cqe_req(req->ctx, req);
+		if (unlikely(!__io_fill_cqe_req(ctx, req))) {
+			spin_lock(&ctx->completion_lock);
+			io_req_cqe_overflow(req);
+			spin_unlock(&ctx->completion_lock);
+		}
 	}
 
 	if (unlikely(!nr_events))

Re: FAILED: patch "[PATCH] io_uring: lock overflowing for IOPOLL" failed to apply to 5.15-stable tree

[Jens Axboe]

[-- Attachment #1: Type: text/plain, Size: 479 bytes --]

On 1/14/23 2:51 AM, gregkh@linuxfoundation.org wrote:
> 
> The patch below does not apply to the 5.15-stable tree.
> If someone wants it applied there, or to any other stable or longterm
> tree, then please email the backport, including the original git commit
> id to <stable@vger.kernel.org>.

This has to be done a bit differently, but this one should work. I tested
it on 5.10-stable, but should apply to 5.15-stable as well as they are
the same base now.

-- 
Jens Axboe


[-- Attachment #2: 0001-io_uring-lock-overflowing-for-IOPOLL.patch --]
[-- Type: text/x-patch, Size: 2784 bytes --]

From 313cc3d4c3f6a1c6d0df525f0368a2c5a7ce97c8 Mon Sep 17 00:00:00 2001
From: Pavel Begunkov <asml.silence@gmail.com>
Date: Sat, 14 Jan 2023 09:14:03 -0700
Subject: [PATCH] io_uring: lock overflowing for IOPOLL
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

commit 544d163d659d45a206d8929370d5a2984e546cb7 upstream.

syzbot reports an issue with overflow filling for IOPOLL:

WARNING: CPU: 0 PID: 28 at io_uring/io_uring.c:734 io_cqring_event_overflow+0x1c0/0x230 io_uring/io_uring.c:734
CPU: 0 PID: 28 Comm: kworker/u4:1 Not tainted 6.2.0-rc3-syzkaller-16369-g358a161a6a9e #0
Workqueue: events_unbound io_ring_exit_work
Call trace:
 io_cqring_event_overflow+0x1c0/0x230 io_uring/io_uring.c:734
 io_req_cqe_overflow+0x5c/0x70 io_uring/io_uring.c:773
 io_fill_cqe_req io_uring/io_uring.h:168 [inline]
 io_do_iopoll+0x474/0x62c io_uring/rw.c:1065
 io_iopoll_try_reap_events+0x6c/0x108 io_uring/io_uring.c:1513
 io_uring_try_cancel_requests+0x13c/0x258 io_uring/io_uring.c:3056
 io_ring_exit_work+0xec/0x390 io_uring/io_uring.c:2869
 process_one_work+0x2d8/0x504 kernel/workqueue.c:2289
 worker_thread+0x340/0x610 kernel/workqueue.c:2436
 kthread+0x12c/0x158 kernel/kthread.c:376
 ret_from_fork+0x10/0x20 arch/arm64/kernel/entry.S:863

There is no real problem for normal IOPOLL as flush is also called with
uring_lock taken, but it's getting more complicated for IOPOLL|SQPOLL,
for which __io_cqring_overflow_flush() happens from the CQ waiting path.

Reported-and-tested-by: syzbot+6805087452d72929404e@syzkaller.appspotmail.com
Cc: stable@vger.kernel.org # 5.10+
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 io_uring/io_uring.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 0c4d16afb9ef..1a3f360af373 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -2478,10 +2478,24 @@ static void io_iopoll_complete(struct io_ring_ctx *ctx, unsigned int *nr_events,
 
 	io_init_req_batch(&rb);
 	while (!list_empty(done)) {
+		struct io_uring_cqe *cqe;
+		unsigned cflags;
+
 		req = list_first_entry(done, struct io_kiocb, inflight_entry);
 		list_del(&req->inflight_entry);
+		cflags = io_put_rw_kbuf(req);
 
-		io_fill_cqe_req(req, req->result, io_put_rw_kbuf(req));
+		cqe = io_get_cqe(ctx);
+		if (unlikely(!cqe)) {
+			spin_lock(&ctx->completion_lock);
+			io_cqring_event_overflow(ctx, req->user_data,
+							req->result, cflags);
+			spin_unlock(&ctx->completion_lock);
+			continue;
+		}
+		WRITE_ONCE(cqe->user_data, req->user_data);
+		WRITE_ONCE(cqe->res, req->result);
+		WRITE_ONCE(cqe->flags, cflags);
 		(*nr_events)++;
 
 		if (req_ref_put_and_test(req))
-- 
2.39.0

Re: FAILED: patch "[PATCH] io_uring: lock overflowing for IOPOLL" failed to apply to 5.15-stable tree

[Jens Axboe]

[-- Attachment #1: Type: text/plain, Size: 619 bytes --]

On 1/14/23 9:15 AM, Jens Axboe wrote:
> On 1/14/23 2:51 AM, gregkh@linuxfoundation.org wrote:
>>
>> The patch below does not apply to the 5.15-stable tree.
>> If someone wants it applied there, or to any other stable or longterm
>> tree, then please email the backport, including the original git commit
>> id to <stable@vger.kernel.org>.
> 
> This has to be done a bit differently, but this one should work. I tested
> it on 5.10-stable, but should apply to 5.15-stable as well as they are
> the same base now.

Oops, missed accounting for overflow. Please use this one instead! Sorry
about that.

-- 
Jens Axboe


[-- Attachment #2: 0001-io_uring-lock-overflowing-for-IOPOLL.patch --]
[-- Type: text/x-patch, Size: 2846 bytes --]

From 929e5c54ff2ad116e9dcee967fc5df1dbb6b1d90 Mon Sep 17 00:00:00 2001
From: Pavel Begunkov <asml.silence@gmail.com>
Date: Sat, 14 Jan 2023 09:14:03 -0700
Subject: [PATCH] io_uring: lock overflowing for IOPOLL
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

commit 544d163d659d45a206d8929370d5a2984e546cb7 upstream.

syzbot reports an issue with overflow filling for IOPOLL:

WARNING: CPU: 0 PID: 28 at io_uring/io_uring.c:734 io_cqring_event_overflow+0x1c0/0x230 io_uring/io_uring.c:734
CPU: 0 PID: 28 Comm: kworker/u4:1 Not tainted 6.2.0-rc3-syzkaller-16369-g358a161a6a9e #0
Workqueue: events_unbound io_ring_exit_work
Call trace:
 io_cqring_event_overflow+0x1c0/0x230 io_uring/io_uring.c:734
 io_req_cqe_overflow+0x5c/0x70 io_uring/io_uring.c:773
 io_fill_cqe_req io_uring/io_uring.h:168 [inline]
 io_do_iopoll+0x474/0x62c io_uring/rw.c:1065
 io_iopoll_try_reap_events+0x6c/0x108 io_uring/io_uring.c:1513
 io_uring_try_cancel_requests+0x13c/0x258 io_uring/io_uring.c:3056
 io_ring_exit_work+0xec/0x390 io_uring/io_uring.c:2869
 process_one_work+0x2d8/0x504 kernel/workqueue.c:2289
 worker_thread+0x340/0x610 kernel/workqueue.c:2436
 kthread+0x12c/0x158 kernel/kthread.c:376
 ret_from_fork+0x10/0x20 arch/arm64/kernel/entry.S:863

There is no real problem for normal IOPOLL as flush is also called with
uring_lock taken, but it's getting more complicated for IOPOLL|SQPOLL,
for which __io_cqring_overflow_flush() happens from the CQ waiting path.

Reported-and-tested-by: syzbot+6805087452d72929404e@syzkaller.appspotmail.com
Cc: stable@vger.kernel.org # 5.10+
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 io_uring/io_uring.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 0c4d16afb9ef..c3838a2c5c6f 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -2478,12 +2478,26 @@ static void io_iopoll_complete(struct io_ring_ctx *ctx, unsigned int *nr_events,
 
 	io_init_req_batch(&rb);
 	while (!list_empty(done)) {
+		struct io_uring_cqe *cqe;
+		unsigned cflags;
+
 		req = list_first_entry(done, struct io_kiocb, inflight_entry);
 		list_del(&req->inflight_entry);
-
-		io_fill_cqe_req(req, req->result, io_put_rw_kbuf(req));
+		cflags = io_put_rw_kbuf(req);
 		(*nr_events)++;
 
+		cqe = io_get_cqe(ctx);
+		if (unlikely(!cqe)) {
+			spin_lock(&ctx->completion_lock);
+			io_cqring_event_overflow(ctx, req->user_data,
+							req->result, cflags);
+			spin_unlock(&ctx->completion_lock);
+			continue;
+		}
+		WRITE_ONCE(cqe->user_data, req->user_data);
+		WRITE_ONCE(cqe->res, req->result);
+		WRITE_ONCE(cqe->flags, cflags);
+
 		if (req_ref_put_and_test(req))
 			io_req_free_batch(&rb, req, &ctx->submit_state);
 	}
-- 
2.39.0

Re: FAILED: patch "[PATCH] io_uring: lock overflowing for IOPOLL" failed to apply to 5.15-stable tree

[Jens Axboe]

[-- Attachment #1: Type: text/plain, Size: 795 bytes --]

On 1/14/23 9:21?AM, Jens Axboe wrote:
> On 1/14/23 9:15?AM, Jens Axboe wrote:
>> On 1/14/23 2:51?AM, gregkh@linuxfoundation.org wrote:
>>>
>>> The patch below does not apply to the 5.15-stable tree.
>>> If someone wants it applied there, or to any other stable or longterm
>>> tree, then please email the backport, including the original git commit
>>> id to <stable@vger.kernel.org>.
>>
>> This has to be done a bit differently, but this one should work. I tested
>> it on 5.10-stable, but should apply to 5.15-stable as well as they are
>> the same base now.
> 
> Oops, missed accounting for overflow. Please use this one instead! Sorry
> about that.

Not batting 1000 on this one today. Wrote a test case for this
specifically now and verified it, this one is good. For real.

-- 
Jens Axboe

[-- Attachment #2: 0001-io_uring-lock-overflowing-for-IOPOLL.patch --]
[-- Type: text/x-patch, Size: 2836 bytes --]

From 31f897396bbbd37010260119293e84007736dbc7 Mon Sep 17 00:00:00 2001
From: Pavel Begunkov <asml.silence@gmail.com>
Date: Sat, 14 Jan 2023 09:14:03 -0700
Subject: [PATCH] io_uring: lock overflowing for IOPOLL
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

commit 544d163d659d45a206d8929370d5a2984e546cb7 upstream.

syzbot reports an issue with overflow filling for IOPOLL:

WARNING: CPU: 0 PID: 28 at io_uring/io_uring.c:734 io_cqring_event_overflow+0x1c0/0x230 io_uring/io_uring.c:734
CPU: 0 PID: 28 Comm: kworker/u4:1 Not tainted 6.2.0-rc3-syzkaller-16369-g358a161a6a9e #0
Workqueue: events_unbound io_ring_exit_work
Call trace:
 io_cqring_event_overflow+0x1c0/0x230 io_uring/io_uring.c:734
 io_req_cqe_overflow+0x5c/0x70 io_uring/io_uring.c:773
 io_fill_cqe_req io_uring/io_uring.h:168 [inline]
 io_do_iopoll+0x474/0x62c io_uring/rw.c:1065
 io_iopoll_try_reap_events+0x6c/0x108 io_uring/io_uring.c:1513
 io_uring_try_cancel_requests+0x13c/0x258 io_uring/io_uring.c:3056
 io_ring_exit_work+0xec/0x390 io_uring/io_uring.c:2869
 process_one_work+0x2d8/0x504 kernel/workqueue.c:2289
 worker_thread+0x340/0x610 kernel/workqueue.c:2436
 kthread+0x12c/0x158 kernel/kthread.c:376
 ret_from_fork+0x10/0x20 arch/arm64/kernel/entry.S:863

There is no real problem for normal IOPOLL as flush is also called with
uring_lock taken, but it's getting more complicated for IOPOLL|SQPOLL,
for which __io_cqring_overflow_flush() happens from the CQ waiting path.

Reported-and-tested-by: syzbot+6805087452d72929404e@syzkaller.appspotmail.com
Cc: stable@vger.kernel.org # 5.10+
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 io_uring/io_uring.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 0c4d16afb9ef..cb83ba50b017 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -2478,12 +2478,26 @@ static void io_iopoll_complete(struct io_ring_ctx *ctx, unsigned int *nr_events,
 
 	io_init_req_batch(&rb);
 	while (!list_empty(done)) {
+		struct io_uring_cqe *cqe;
+		unsigned cflags;
+
 		req = list_first_entry(done, struct io_kiocb, inflight_entry);
 		list_del(&req->inflight_entry);
-
-		io_fill_cqe_req(req, req->result, io_put_rw_kbuf(req));
+		cflags = io_put_rw_kbuf(req);
 		(*nr_events)++;
 
+		cqe = io_get_cqe(ctx);
+		if (cqe) {
+			WRITE_ONCE(cqe->user_data, req->user_data);
+			WRITE_ONCE(cqe->res, req->result);
+			WRITE_ONCE(cqe->flags, cflags);
+		} else {
+			spin_lock(&ctx->completion_lock);
+			io_cqring_event_overflow(ctx, req->user_data,
+							req->result, cflags);
+			spin_unlock(&ctx->completion_lock);
+		}
+
 		if (req_ref_put_and_test(req))
 			io_req_free_batch(&rb, req, &ctx->submit_state);
 	}
-- 
2.39.0

Re: FAILED: patch "[PATCH] io_uring: lock overflowing for IOPOLL" failed to apply to 5.15-stable tree

[Greg KH]

On Sat, Jan 14, 2023 at 09:48:34AM -0700, Jens Axboe wrote:
> On 1/14/23 9:21?AM, Jens Axboe wrote:
> > On 1/14/23 9:15?AM, Jens Axboe wrote:
> >> On 1/14/23 2:51?AM, gregkh@linuxfoundation.org wrote:
> >>>
> >>> The patch below does not apply to the 5.15-stable tree.
> >>> If someone wants it applied there, or to any other stable or longterm
> >>> tree, then please email the backport, including the original git commit
> >>> id to <stable@vger.kernel.org>.
> >>
> >> This has to be done a bit differently, but this one should work. I tested
> >> it on 5.10-stable, but should apply to 5.15-stable as well as they are
> >> the same base now.
> > 
> > Oops, missed accounting for overflow. Please use this one instead! Sorry
> > about that.
> 
> Not batting 1000 on this one today. Wrote a test case for this
> specifically now and verified it, this one is good. For real.

Looks like Sasha grabbed all of these already, thanks for the backports!

greg k-h