[PATCH] io_uring: fix skipping disabling sqo on exec

[Pavel Begunkov <asml.silence@gmail.com>]

If there are no requests at the time __io_uring_task_cancel() is called,
tctx_inflight() returns zero and and it terminates not getting a chance
to go through __io_uring_files_cancel() and do
io_disable_sqo_submit(). And we absolutely want them disabled by the
time cancellation ends.

Also a fix potential false positive warning because of ctx->sq_data
check before io_disable_sqo_submit().

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

To not grow diffstat now will be cleaned for-next

 fs/io_uring.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index d494c4269fc5..0d50845f1f3f 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -8937,10 +8937,12 @@ static void io_uring_cancel_task_requests(struct io_ring_ctx *ctx,
 {
 	struct task_struct *task = current;
 
-	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
+	if (ctx->flags & IORING_SETUP_SQPOLL) {
 		/* for SQPOLL only sqo_task has task notes */
 		WARN_ON_ONCE(ctx->sqo_task != current);
 		io_disable_sqo_submit(ctx);
+	}
+	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
 		task = ctx->sq_data->thread;
 		atomic_inc(&task->io_uring->in_idle);
 		io_sq_thread_park(ctx->sq_data);
@@ -9085,6 +9087,10 @@ void __io_uring_task_cancel(void)
 	/* make sure overflow events are dropped */
 	atomic_inc(&tctx->in_idle);
 
+	/* trigger io_disable_sqo_submit() */
+	if (tctx->sqpoll)
+		__io_uring_files_cancel(NULL);
+
 	do {
 		/* read completions before cancelations */
 		inflight = tctx_inflight(tctx);
-- 
2.24.0