io_uring: pass in EPOLL_URING_WAKE for eventfd signaling and wakeups

Pass in EPOLL_URING_WAKE when signaling eventfd or doing poll related
wakups, so that we can check for a circular event dependency between
eventfd and epoll. If this flag is set when our wakeup handlers are
called, then we know we have a dependency that needs to terminate
multishot requests.

eventfd and epoll are the only such possible dependencies.

Cc: stable@vger.kernel.org # 6.0
Signed-off-by: Jens Axboe <axboe@kernel.dk>

3 files changed, 21 insertions, 6 deletions

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 1299f9c8567a..762ecab801f2 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -495,7 +495,7 @@ static void io_eventfd_ops(struct rcu_head *rcu)
 	int ops = atomic_xchg(&ev_fd->ops, 0);
 
 	if (ops & BIT(IO_EVENTFD_OP_SIGNAL_BIT))
-		eventfd_signal(ev_fd->cq_ev_fd, 1);
+		eventfd_signal_mask(ev_fd->cq_ev_fd, 1, EPOLL_URING_WAKE);
 
 	/* IO_EVENTFD_OP_FREE_BIT may not be set here depending on callback
 	 * ordering in a race but if references are 0 we know we have to free
@@ -531,7 +531,7 @@ static void io_eventfd_signal(struct io_ring_ctx *ctx)
 		goto out;
 
 	if (likely(eventfd_signal_allowed())) {
-		eventfd_signal(ev_fd->cq_ev_fd, 1);
+		eventfd_signal_mask(ev_fd->cq_ev_fd, 1, EPOLL_URING_WAKE);
 	} else {
 		atomic_inc(&ev_fd->refs);
 		if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_SIGNAL_BIT), &ev_fd->ops))
diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h
index 69fbd27c7577..83013ee584d6 100644
--- a/io_uring/io_uring.h
+++ b/io_uring/io_uring.h
@@ -4,6 +4,7 @@
 #include <linux/errno.h>
 #include <linux/lockdep.h>
 #include <linux/io_uring_types.h>
+#include <uapi/linux/eventpoll.h>
 #include "io-wq.h"
 #include "slist.h"
 #include "filetable.h"
@@ -211,12 +212,18 @@ static inline void io_commit_cqring(struct io_ring_ctx *ctx)
 static inline void __io_cqring_wake(struct io_ring_ctx *ctx)
 {
 	/*
-	 * wake_up_all() may seem excessive, but io_wake_function() and
-	 * io_should_wake() handle the termination of the loop and only
-	 * wake as many waiters as we need to.
+	 * Trigger waitqueue handler on all waiters on our waitqueue. This
+	 * won't necessarily wake up all the tasks, io_should_wake() will make
+	 * that decision.
+	 *
+	 * Pass in EPOLLIN|EPOLL_URING_WAKE as the poll wakeup key. The latter
+	 * set in the mask so that if we recurse back into our own poll
+	 * waitqueue handlers, we know we have a dependency between eventfd or
+	 * epoll and should terminate multishot poll at that point.
 	 */
 	if (waitqueue_active(&ctx->cq_wait))
-		wake_up_all(&ctx->cq_wait);
+		__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
+				poll_to_key(EPOLL_URING_WAKE | EPOLLIN));
 }
 
 static inline void io_cqring_wake(struct io_ring_ctx *ctx)
diff --git a/io_uring/poll.c b/io_uring/poll.c
index 8fb8e781c02d..22c9b2e0944a 100644
--- a/io_uring/poll.c
+++ b/io_uring/poll.c
@@ -389,6 +389,14 @@ static int io_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
 		return 0;
 
 	if (io_poll_get_ownership(req)) {
+		/*
+		 * If we trigger a multishot poll off our own wakeup path,
+		 * disable multishot as there is a circular dependency between
+		 * CQ posting and triggering the event.
+		 */
+		if (mask & EPOLL_URING_WAKE)
+			poll->events |= EPOLLONESHOT;
+
 		/* optional, saves extra locking for removal in tw handler */
 		if (mask && poll->events & EPOLLONESHOT) {
 			list_del_init(&poll->wait.entry);