On 7/10/21 2:40 PM, Jens Axboe wrote: > On Fri, Jul 9, 2021 at 7:46 PM Pavel Begunkov <asml.silence@xxxxxxxxx> wrote: >> When we use delayed_work for fallback execution of requests, current >> will be not of the submitter task, and so checks in io_req_task_submit() >> may not behave as expected. Currently, it leaves inline completions not >> flushed, so making io_ring_exit_work() to hang. Use the submitter task >> for all those checks. >> >> Cc: stable@xxxxxxxxxxxxxxx >> Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx> >> --- >> fs/io_uring.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/fs/io_uring.c b/fs/io_uring.c >> index 7167c61c6d1b..770fdcd7d3e4 100644 >> --- a/fs/io_uring.c >> +++ b/fs/io_uring.c >> @@ -2016,7 +2016,7 @@ static void io_req_task_submit(struct io_kiocb *req) >> >> /* ctx stays valid until unlock, even if we drop all ours ctx->refs */ >> mutex_lock(&ctx->uring_lock); >> - if (!(current->flags & PF_EXITING) && !current->in_execve) >> + if (!(req->task->flags & PF_EXITING) && !req->task->in_execve) >> __io_queue_sqe(req); >> else >> io_req_complete_failed(req, -EFAULT); > > I don't think that ->in_execve check is useful anymore now that we don't > have weak references to the files table, so it should probably just go > away. Had such a thought but from the premise that on exec we wait / cancel all requests. But I'd rather to leave it to a separate commit for-next, don't you think so? -- Pavel Begunkov