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

[Jens Axboe <axboe@kernel.dk>]

[-- 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