[RFC] io_uring: fix possible poll event lost in multi shot mode

[RFC] io_uring: fix possible poll event lost in multi shot mode

[Xiaoguang Wang]

IIUC, IORING_POLL_ADD_MULTI is similar to epoll's edge-triggered mode,
that means once one pure poll request returns one event(cqe), we'll
need to read or write continually until EAGAIN is returned, then I think
there is a possible poll event lost race in multi shot mode:

t1  poll request add |                         |
t2                   |                         |
t3  event happens    |                         |
t4  task work add    |                         |
t5                   | task work run           |
t6                   |   commit one cqe        |
t7                   |                         | user app handles cqe
t8                   |   new event happen      |
t9                   |   add back to waitqueue |
t10                  |

After t6 but before t9, if new event happens, there'll be no wakeup
operation, and if user app has picked up this cqe in t7, read or write
until EAGAIN is returned. In t8, new event happens and will be lost,
though this race window maybe small.

To fix this possible race, add poll request back to waitqueue before
committing cqe.

Signed-off-by: Xiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
---
 fs/io_uring.c | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 2bde732a1183..27608bad2276 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -5098,7 +5098,7 @@ static void io_poll_remove_double(struct io_kiocb *req)
 	}
 }
 
-static bool io_poll_complete(struct io_kiocb *req, __poll_t mask)
+static bool __io_poll_complete(struct io_kiocb *req, __poll_t mask)
 	__must_hold(&req->ctx->completion_lock)
 {
 	struct io_ring_ctx *ctx = req->ctx;
@@ -5120,10 +5120,19 @@ static bool io_poll_complete(struct io_kiocb *req, __poll_t mask)
 	if (flags & IORING_CQE_F_MORE)
 		ctx->cq_extra++;
 
-	io_commit_cqring(ctx);
 	return !(flags & IORING_CQE_F_MORE);
 }
 
+static inline bool io_poll_complete(struct io_kiocb *req, __poll_t mask)
+	__must_hold(&req->ctx->completion_lock)
+{
+	bool done;
+
+	done = __io_poll_complete(req, mask);
+	io_commit_cqring(req->ctx);
+	return done;
+}
+
 static void io_poll_task_func(struct io_kiocb *req, bool *locked)
 {
 	struct io_ring_ctx *ctx = req->ctx;
@@ -5134,7 +5143,7 @@ static void io_poll_task_func(struct io_kiocb *req, bool *locked)
 	} else {
 		bool done;
 
-		done = io_poll_complete(req, req->result);
+		done = __io_poll_complete(req, req->result);
 		if (done) {
 			io_poll_remove_double(req);
 			hash_del(&req->hash_node);
@@ -5142,6 +5151,7 @@ static void io_poll_task_func(struct io_kiocb *req, bool *locked)
 			req->result = 0;
 			add_wait_queue(req->poll.head, &req->poll.wait);
 		}
+		io_commit_cqring(ctx);
 		spin_unlock(&ctx->completion_lock);
 		io_cqring_ev_posted(ctx);
 
-- 
2.14.4.44.g2045bb6

Re: [RFC] io_uring: fix possible poll event lost in multi shot mode

[Jens Axboe]

On 9/2/21 8:48 AM, Xiaoguang Wang wrote:
> IIUC, IORING_POLL_ADD_MULTI is similar to epoll's edge-triggered mode,
> that means once one pure poll request returns one event(cqe), we'll
> need to read or write continually until EAGAIN is returned, then I think
> there is a possible poll event lost race in multi shot mode:
> 
> t1  poll request add |                         |
> t2                   |                         |
> t3  event happens    |                         |
> t4  task work add    |                         |
> t5                   | task work run           |
> t6                   |   commit one cqe        |
> t7                   |                         | user app handles cqe
> t8                   |   new event happen      |
> t9                   |   add back to waitqueue |
> t10                  |
> 
> After t6 but before t9, if new event happens, there'll be no wakeup
> operation, and if user app has picked up this cqe in t7, read or write
> until EAGAIN is returned. In t8, new event happens and will be lost,
> though this race window maybe small.
> 
> To fix this possible race, add poll request back to waitqueue before
> committing cqe.

This looks sane to me. Probably just needs a:

Fixes: 88e41cf928a6 ("io_uring: add multishot mode for IORING_OP_POLL_ADD")

tag.

-- 
Jens Axboe