[PATCH 5.10 126/199] io_uring: iopoll requests should also wake task ->in_idle state

[Greg Kroah-Hartman]

From: Jens Axboe <ax...@kernel.dk>

commit c93cc9e16d88e0f5ea95d2d65d58a8a4dab258bc upstream.

If we're freeing/finishing iopoll requests, ensure we check if the task
is in idling in terms of cancelation. Otherwise we could end up waiting
forever in __io_uring_task_cancel() if the task has active iopoll
requests that need cancelation.

Cc: sta...@vger.kernel.org # 5.9+
Signed-off-by: Jens Axboe <ax...@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gre...@linuxfoundation.org>

---
 fs/io_uring.c |    4 ++++
 1 file changed, 4 insertions(+)

--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2167,6 +2167,8 @@ static void io_req_free_batch_finish(str
                struct io_uring_task *tctx = rb->task->io_uring;
 
                percpu_counter_sub(&tctx->inflight, rb->task_refs);
+               if (atomic_read(&tctx->in_idle))
+                       wake_up(&tctx->wait);
                put_task_struct_many(rb->task, rb->task_refs);
                rb->task = NULL;
        }
@@ -2186,6 +2188,8 @@ static void io_req_free_batch(struct req
                        struct io_uring_task *tctx = rb->task->io_uring;
 
                        percpu_counter_sub(&tctx->inflight, rb->task_refs);
+                       if (atomic_read(&tctx->in_idle))
+                               wake_up(&tctx->wait);
                        put_task_struct_many(rb->task, rb->task_refs);
                }
                rb->task = req->task;