io_uring: kill io_sq_thread_fork() and return -EOWNERDEAD if the sq_t…

…hread is gone

This brings the behavior back in line with what 5.11 and earlier did,
and this is no longer needed with the improved handling of creds
not needing to do unshare().

Signed-off-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

fs/io_uring.c

@@ -336,7 +336,6 @@ struct io_ring_ctx {
 		unsigned int		drain_next: 1;
 		unsigned int		eventfd_async: 1;
 		unsigned int		restricted: 1;
-		unsigned int		sqo_exec: 1;
 
 		/*
 		 * Ring buffer of indices into array of io_uring_sqe, which is
@@ -6786,7 +6785,6 @@ static int io_sq_thread(void *data)
 
 	sqd->thread = NULL;
 	list_for_each_entry(ctx, &sqd->ctx_list, sqd_list) {
-		ctx->sqo_exec = 1;
 		io_ring_set_wakeup_flag(ctx);
 	}
 
@@ -7846,26 +7844,6 @@ void __io_uring_free(struct task_struct *tsk)
 	tsk->io_uring = NULL;
 }
 
-static int io_sq_thread_fork(struct io_sq_data *sqd, struct io_ring_ctx *ctx)
-{
-	struct task_struct *tsk;
-	int ret;
-
-	clear_bit(IO_SQ_THREAD_SHOULD_STOP, &sqd->state);
-	reinit_completion(&sqd->parked);
-	ctx->sqo_exec = 0;
-	sqd->task_pid = current->pid;
-	tsk = create_io_thread(io_sq_thread, sqd, NUMA_NO_NODE);
-	if (IS_ERR(tsk))
-		return PTR_ERR(tsk);
-	ret = io_uring_alloc_task_context(tsk, ctx);
-	if (ret)
-		set_bit(IO_SQ_THREAD_SHOULD_STOP, &sqd->state);
-	sqd->thread = tsk;
-	wake_up_new_task(tsk);
-	return ret;
-}
-
 static int io_sq_offload_create(struct io_ring_ctx *ctx,
 				struct io_uring_params *p)
 {
@@ -9199,13 +9177,10 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
 	if (ctx->flags & IORING_SETUP_SQPOLL) {
 		io_cqring_overflow_flush(ctx, false, NULL, NULL);
 
-		if (unlikely(ctx->sqo_exec)) {
-			ret = io_sq_thread_fork(ctx->sq_data, ctx);
-			if (ret)
-				goto out;
-			ctx->sqo_exec = 0;
-		}
 		ret = -EOWNERDEAD;
+		if (unlikely(ctx->sq_data->thread == NULL)) {
+			goto out;
+		}
 		if (flags & IORING_ENTER_SQ_WAKEUP)
 			wake_up(&ctx->sq_data->wait);
 		if (flags & IORING_ENTER_SQ_WAIT) {