[PATCH v4] io_uring: don't wait when under-submitting

[Pavel Begunkov]

There is no reliable way to submit and wait in a single syscall, as
io_submit_sqes() may under-consume sqes (in case of an early error).
Then it will wait for not-yet-submitted requests, deadlocking the user
in most cases.

In such cases adjust min_complete, so it won't wait for more than
what have been submitted in the current io_uring_enter() call. It
may be less than total in-flight, but that up to a user to handle.

Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
---

v2: cap min_complete if submitted partially (Jens Axboe)
v3: update commit message (Jens Axboe)
v4: fix behavioural change when submitting more than have (Jens Axboe)

 fs/io_uring.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 5ad652fa24b8..167fbcc8be0b 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -4463,11 +4463,15 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
 		submitted = io_submit_sqes(ctx, to_submit, f.file, fd,
 					   &cur_mm, false);
 		mutex_unlock(&ctx->uring_lock);
+		if (submitted <= 0)
+			goto done;
 	}
 	if (flags & IORING_ENTER_GETEVENTS) {
 		unsigned nr_events = 0;
 
 		min_complete = min(min_complete, ctx->cq_entries);
+		if (submitted != to_submit)
+			min_complete = min(min_complete, (u32)submitted);
 
 		if (ctx->flags & IORING_SETUP_IOPOLL) {
 			ret = io_iopoll_check(ctx, &nr_events, min_complete);
@@ -4475,7 +4479,7 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
 			ret = io_cqring_wait(ctx, min_complete, sig, sigsz);
 		}
 	}
-
+done:
 	percpu_ref_put(&ctx->refs);
 out_fput:
 	fdput(f);
-- 
2.24.0