If we're exiting the ring, just let the IO fail with -EAGAIN as nobody will care anyway. It's not the right context to reissue from. Cc: stable@xxxxxxxxxxxxxxx Signed-off-by: Jens Axboe <axboe@xxxxxxxxx> --- diff --git a/fs/io_uring.c b/fs/io_uring.c index bf9ad810c621..275ad84e8227 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -2839,6 +2839,13 @@ static bool io_rw_reissue(struct io_kiocb *req) return false; if ((req->flags & REQ_F_NOWAIT) || io_wq_current_is_worker()) return false; + /* + * If ref is dying, we might be running poll reap from the exit work. + * Don't attempt to reissue from that path, just let it fail with + * -EAGAIN. + */ + if (percpu_ref_is_dying(&req->ctx->refs)) + return false; lockdep_assert_held(&req->ctx->uring_lock); -- Jens Axboe
