io_uring: fix retry handling off iowq

io_req_complete_post() doesn't handle reissue and if called with a
REQ_F_REISSUE request it might post extra unexpected completions. Fix it
by pushing into flush_completion via task work.

Fixes: d803d123948fe ("io_uring/rw: handle -EAGAIN retry at IO completion time")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/badb3d7e462881e7edbfcc2be6301090b07dbe53.1742829388.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index e1128b9551aa870fc23be161ee005172850b93e0..e6c46294827355d071253f2941295067ee78c2b6 100644 (file)
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -904,7 +904,7 @@ static void io_req_complete_post(struct io_kiocb *req, unsigned issue_flags)
         * Handle special CQ sync cases via task_work. DEFER_TASKRUN requires
         * the submitter task context, IOPOLL protects with uring_lock.
         */
-       if (ctx->lockless_cq) {
+       if (ctx->lockless_cq || (req->flags & REQ_F_REISSUE)) {
                req->io_task_work.func = io_req_task_complete;
                io_req_task_work_add(req);
                return;