io_uring: fix overflowed reqs cancellation

Overflowed requests in io_uring_cancel_files() should be shed only of
inflight and overflowed refs. All other left references are owned by
someone else.

If refcount_sub_and_test() fails, it will go further and put put extra
ref, don't do that. Also, don't need to do io_wq_cancel_work()
for overflowed reqs, they will be let go shortly anyway.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 9f11feb8a3ac3bf46d460d7d506a107d21d4dc31..732ec73ec3c0b49ab6aff58fe6c7383304828853 100644 (file)
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -7449,10 +7449,11 @@ static void io_uring_cancel_files(struct io_ring_ctx *ctx,
                                finish_wait(&ctx->inflight_wait, &wait);
                                continue;
                        }
+               } else {
+                       io_wq_cancel_work(ctx->io_wq, &cancel_req->work);
+                       io_put_req(cancel_req);
                }
 
-               io_wq_cancel_work(ctx->io_wq, &cancel_req->work);
-               io_put_req(cancel_req);
                schedule();
                finish_wait(&ctx->inflight_wait, &wait);
        }