io_uring: fix race between timeout flush and removal

io_flush_timeouts() assumes the timeout isn't in progress of triggering
or being removed/canceled, so it unconditionally removes it from the
timeout list and attempts to cancel it.

Leave it on the list and let the normal timeout cancelation take care
of it.

Cc: stable@vger.kernel.org # 5.5+
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/fs/io_uring.c b/fs/io_uring.c
index fafd1ca..659f8ec 100644 (file)
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -1736,12 +1736,11 @@ static __cold void io_flush_timeouts(struct io_ring_ctx *ctx)
        __must_hold(&ctx->completion_lock)
 {
        u32 seq = ctx->cached_cq_tail - atomic_read(&ctx->cq_timeouts);
+       struct io_kiocb *req, *tmp;
 
        spin_lock_irq(&ctx->timeout_lock);
-       while (!list_empty(&ctx->timeout_list)) {
+       list_for_each_entry_safe(req, tmp, &ctx->timeout_list, timeout.list) {
                u32 events_needed, events_got;
-               struct io_kiocb *req = list_first_entry(&ctx->timeout_list,
-                                               struct io_kiocb, timeout.list);
 
                if (io_is_timeout_noseq(req))
                        break;
@@ -1758,7 +1757,6 @@ static __cold void io_flush_timeouts(struct io_ring_ctx *ctx)
                if (events_got < events_needed)
                        break;
 
-               list_del_init(&req->timeout.list);
                io_kill_timeout(req, 0);
        }
        ctx->cq_last_tm_flush = seq;
@@ -6628,6 +6626,7 @@ static int io_timeout_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe,
        if (data->ts.tv_sec < 0 || data->ts.tv_nsec < 0)
                return -EINVAL;
 
+       INIT_LIST_HEAD(&req->timeout.list);
        data->mode = io_translate_timeout_mode(flags);
        hrtimer_init(&data->timer, io_timeout_get_clock(data), data->mode);