io_uring: correct poll cancel and linked timeout expiration completion

to #26323578

commit b0dd8a412699afe3420a08f841333f3474ad45c5 upstream.

Currently a poll request fills a completion entry of 0, even if it got
cancelled. This is odd, and it makes it harder to support with chains.
Ensure that it returns -ECANCELED in the completions events if it got
cancelled, and furthermore ensure that the linked timeout that triggered
it completes with -ETIME if we did indeed trigger the completions
through a timeout.
Signed-off-by: NJens Axboe <axboe@kernel.dk>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Acked-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>

fs/io_uring.c

@@ -2068,11 +2068,14 @@ static int io_poll_remove(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	return 0;
 }
 
-static void io_poll_complete(struct io_kiocb *req, __poll_t mask)
+static void io_poll_complete(struct io_kiocb *req, __poll_t mask, int error)
 {
 	struct io_ring_ctx *ctx = req->ctx;
 
 	req->poll.done = true;
+	if (error)
+		io_cqring_fill_event(req, error);
+	else
 		io_cqring_fill_event(req, mangle_poll(mask));
 	io_commit_cqring(ctx);
 }
@@ -2086,11 +2089,16 @@ static void io_poll_complete_work(struct io_wq_work **workptr)
 	struct io_ring_ctx *ctx = req->ctx;
 	struct io_kiocb *nxt = NULL;
 	__poll_t mask = 0;
+	int ret = 0;
 
-	if (work->flags & IO_WQ_WORK_CANCEL)
+	if (work->flags & IO_WQ_WORK_CANCEL) {
 		WRITE_ONCE(poll->canceled, true);
+		ret = -ECANCELED;
+	} else if (READ_ONCE(poll->canceled)) {
+		ret = -ECANCELED;
+	}
 
-	if (!READ_ONCE(poll->canceled))
+	if (ret != -ECANCELED)
 		mask = vfs_poll(poll->file, &pt) & poll->events;
 
 	/*
@@ -2101,13 +2109,13 @@ static void io_poll_complete_work(struct io_wq_work **workptr)
 	 * avoid further branches in the fast path.
 	 */
 	spin_lock_irq(&ctx->completion_lock);
-	if (!mask && !READ_ONCE(poll->canceled)) {
+	if (!mask && ret != -ECANCELED) {
 		add_wait_queue(poll->head, &poll->wait);
 		spin_unlock_irq(&ctx->completion_lock);
 		return;
 	}
 	io_poll_remove_req(req);
-	io_poll_complete(req, mask);
+	io_poll_complete(req, mask, ret);
 	spin_unlock_irq(&ctx->completion_lock);
 
 	io_cqring_ev_posted(ctx);
@@ -2141,7 +2149,7 @@ static int io_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
 	 */
 	if (mask && spin_trylock_irqsave(&ctx->completion_lock, flags)) {
 		io_poll_remove_req(req);
-		io_poll_complete(req, mask);
+		io_poll_complete(req, mask, 0);
 		req->flags |= REQ_F_COMP_LOCKED;
 		io_put_req(req);
 		spin_unlock_irqrestore(&ctx->completion_lock, flags);
@@ -2253,7 +2261,7 @@ static int io_poll_add(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 	}
 	if (mask) { /* no async, we'd stolen it */
 		ipt.error = 0;
-		io_poll_complete(req, mask);
+		io_poll_complete(req, mask, 0);
 	}
 	spin_unlock_irq(&ctx->completion_lock);
 
@@ -2505,7 +2513,7 @@ static int io_async_cancel_one(struct io_ring_ctx *ctx, void *sqe_addr)
 
 static void io_async_find_and_cancel(struct io_ring_ctx *ctx,
 				     struct io_kiocb *req, __u64 sqe_addr,
-				     struct io_kiocb **nxt)
+				     struct io_kiocb **nxt, int success_ret)
 {
 	unsigned long flags;
 	int ret;
@@ -2522,6 +2530,8 @@ static void io_async_find_and_cancel(struct io_ring_ctx *ctx,
 		goto done;
 	ret = io_poll_cancel(ctx, sqe_addr);
 done:
+	if (!ret)
+		ret = success_ret;
 	io_cqring_fill_event(req, ret);
 	io_commit_cqring(ctx);
 	spin_unlock_irqrestore(&ctx->completion_lock, flags);
@@ -2543,7 +2553,7 @@ static int io_async_cancel(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 	    sqe->cancel_flags)
 		return -EINVAL;
 
-	io_async_find_and_cancel(ctx, req, READ_ONCE(sqe->addr), nxt);
+	io_async_find_and_cancel(ctx, req, READ_ONCE(sqe->addr), nxt, 0);
 	return 0;
 }
 
@@ -2833,7 +2843,8 @@ static enum hrtimer_restart io_link_timeout_fn(struct hrtimer *timer)
 	spin_unlock_irqrestore(&ctx->completion_lock, flags);
 
 	if (prev) {
-		io_async_find_and_cancel(ctx, req, prev->user_data, NULL);
+		io_async_find_and_cancel(ctx, req, prev->user_data, NULL,
+						-ETIME);
 		io_put_req(prev);
 	} else {
 		io_cqring_add_event(req, -ETIME);