io_uring: fix task_work cap overshooting

A previous commit fixed task_work overrunning by a lot more than what
the user asked for, by adding a retry list. However, it didn't cap the
overall count, hence for multiple task_work runs inside the same wait
loop, it'd still overshoot the target by potentially a large amount.

Cap it generally inside the wait path. Note that this will still
overshoot the default limit of 20, but should overshoot by no more than
limit-1 in addition to the limit. That still provides a ceiling over how
much task_work will be run, rather than still having gaps where it was
uncapped essentially.

Fixes: f46b9cdb22f7 ("io_uring: limit local tw done")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 19 insertions, 15 deletions

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index bfa93888f862..ae199e44da57 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -1277,6 +1277,8 @@ static int __io_run_local_work_loop(struct llist_node **node,
 				    struct io_tw_state *ts,
 				    int events)
 {
+	int ret = 0;
+
 	while (*node) {
 		struct llist_node *next = (*node)->next;
 		struct io_kiocb *req = container_of(*node, struct io_kiocb,
@@ -1285,27 +1287,27 @@ static int __io_run_local_work_loop(struct llist_node **node,
 				io_poll_task_func, io_req_rw_complete,
 				req, ts);
 		*node = next;
-		if (--events <= 0)
+		if (++ret >= events)
 			break;
 	}
 
-	return events;
+	return ret;
 }
 
 static int __io_run_local_work(struct io_ring_ctx *ctx, struct io_tw_state *ts,
-			       int min_events)
+			       int min_events, int max_events)
 {
 	struct llist_node *node;
 	unsigned int loops = 0;
-	int ret, limit;
+	int ret = 0;
 
 	if (WARN_ON_ONCE(ctx->submitter_task != current))
 		return -EEXIST;
 	if (ctx->flags & IORING_SETUP_TASKRUN_FLAG)
 		atomic_andnot(IORING_SQ_TASKRUN, &ctx->rings->sq_flags);
-	limit = max(IO_LOCAL_TW_DEFAULT_MAX, min_events);
 again:
-	ret = __io_run_local_work_loop(&ctx->retry_llist.first, ts, limit);
+	min_events -= ret;
+	ret = __io_run_local_work_loop(&ctx->retry_llist.first, ts, max_events);
 	if (ctx->retry_llist.first)
 		goto retry_done;
 
@@ -1314,11 +1316,10 @@ again:
 	 * running the pending items.
 	 */
 	node = llist_reverse_order(llist_del_all(&ctx->work_llist));
-	ret = __io_run_local_work_loop(&node, ts, ret);
+	ret += __io_run_local_work_loop(&node, ts, max_events - ret);
 	ctx->retry_llist.first = node;
 	loops++;
 
-	ret = limit - ret;
 	if (io_run_local_work_continue(ctx, ret, min_events))
 		goto again;
 retry_done:
@@ -1337,16 +1338,18 @@ static inline int io_run_local_work_locked(struct io_ring_ctx *ctx,
 
 	if (!io_local_work_pending(ctx))
 		return 0;
-	return __io_run_local_work(ctx, &ts, min_events);
+	return __io_run_local_work(ctx, &ts, min_events,
+					max(IO_LOCAL_TW_DEFAULT_MAX, min_events));
 }
 
-static int io_run_local_work(struct io_ring_ctx *ctx, int min_events)
+static int io_run_local_work(struct io_ring_ctx *ctx, int min_events,
+			     int max_events)
 {
 	struct io_tw_state ts = {};
 	int ret;
 
 	mutex_lock(&ctx->uring_lock);
-	ret = __io_run_local_work(ctx, &ts, min_events);
+	ret = __io_run_local_work(ctx, &ts, min_events, max_events);
 	mutex_unlock(&ctx->uring_lock);
 	return ret;
 }
@@ -2352,7 +2355,7 @@ int io_run_task_work_sig(struct io_ring_ctx *ctx)
 {
 	if (io_local_work_pending(ctx)) {
 		__set_current_state(TASK_RUNNING);
-		if (io_run_local_work(ctx, INT_MAX) > 0)
+		if (io_run_local_work(ctx, INT_MAX, IO_LOCAL_TW_DEFAULT_MAX) > 0)
 			return 0;
 	}
 	if (io_run_task_work() > 0)
@@ -2515,7 +2518,8 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, u32 flags,
 	if (!io_allowed_run_tw(ctx))
 		return -EEXIST;
 	if (io_local_work_pending(ctx))
-		io_run_local_work(ctx, min_events);
+		io_run_local_work(ctx, min_events,
+				  max(IO_LOCAL_TW_DEFAULT_MAX, min_events));
 	io_run_task_work();
 
 	if (unlikely(test_bit(IO_CHECK_CQ_OVERFLOW_BIT, &ctx->check_cq)))
@@ -2586,7 +2590,7 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, u32 flags,
 		 * now rather than let the caller do another wait loop.
 		 */
 		if (io_local_work_pending(ctx))
-			io_run_local_work(ctx, nr_wait);
+			io_run_local_work(ctx, nr_wait, nr_wait);
 		io_run_task_work();
 
 		/*
@@ -3098,7 +3102,7 @@ static __cold bool io_uring_try_cancel_requests(struct io_ring_ctx *ctx,
 
 	if ((ctx->flags & IORING_SETUP_DEFER_TASKRUN) &&
 	    io_allowed_defer_tw_run(ctx))
-		ret |= io_run_local_work(ctx, INT_MAX) > 0;
+		ret |= io_run_local_work(ctx, INT_MAX, INT_MAX) > 0;
 	ret |= io_cancel_defer_files(ctx, tctx, cancel_all);
 	mutex_lock(&ctx->uring_lock);
 	ret |= io_poll_remove_all(ctx, tctx, cancel_all);