From 38cfb5a45ee013bfab5d1ae4c4738815e744b440 Mon Sep 17 00:00:00 2001
From: Jens Axboe <axboe@kernel.dk>
Date: Sun, 26 Aug 2018 10:10:05 -0600
Subject: [PATCH] blk-wbt: improve waking of tasks
Git-commit: 38cfb5a45ee013bfab5d1ae4c4738815e744b440
Patch-mainline: v4.19-rc2
References: bsc#1135873

We have two potential issues:

1) After commit 2887e41b910b, we only wake one process at the time when
   we finish an IO. We really want to wake up as many tasks as can
   queue IO. Before this commit, we woke up everyone, which could cause
   a thundering herd issue.

2) A task can potentially consume two wakeups, causing us to (in
   practice) miss a wakeup.

Fix both by providing our own wakeup function, which stops
__wake_up_common() from waking up more tasks if we fail to get a
queueing token. With the strict ordering we have on the wait list, this
wakes the right tasks and the right amount of tasks.

Based on a patch from Jianchao Wang <jianchao.w.wang@oracle.com>.

Tested-by: Agarwal, Anchal <anchalag@amazon.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Acked-by: Jan Kara <jack@suse.cz>

---
 block/blk-wbt.c |   62 ++++++++++++++++++++++++++++++++++++++++++++++++++------
 1 file changed, 56 insertions(+), 6 deletions(-)

--- a/block/blk-wbt.c
+++ b/block/blk-wbt.c
@@ -153,7 +152,7 @@ static void wbt_rqw_done(struct rq_wb *r
 		int diff = limit - inflight;
 
 		if (!inflight || diff >= rwb->wb_background / 2)
-			wake_up(&rqw->wait);
+			wake_up_all(&rqw->wait);
 	}
 }
 
@@ -516,6 +515,35 @@ static inline unsigned int get_limit(str
 	return limit;
 }
 
+struct wbt_wait_data {
+	struct wait_queue_entry wq;
+	struct task_struct *task;
+	struct rq_wb *rwb;
+	struct rq_wait *rqw;
+	unsigned long rw;
+	bool got_token;
+};
+
+static int wbt_wake_function(struct wait_queue_entry *curr, unsigned int mode,
+			     int wake_flags, void *key)
+{
+	struct wbt_wait_data *data = container_of(curr, struct wbt_wait_data,
+							wq);
+
+	/*
+	 * If we fail to get a budget, return -1 to interrupt the wake up
+	 * loop in __wake_up_common.
+	 */
+	if (!atomic_inc_below(&data->rqw->inflight,
+			      get_limit(data->rwb, data->rw)))
+		return -1;
+
+	data->got_token = true;
+	list_del_init(&curr->entry);
+	wake_up_process(data->task);
+	return 1;
+}
+
 /*
  * Block if we will exceed our limit, or if we are currently waiting for
  * the timer to kick off queuing again.
@@ -525,7 +553,16 @@ static void __wbt_wait(struct rq_wb *rwb
 	__acquires(lock)
 {
 	struct rq_wait *rqw = get_rq_wait(rwb, wb_acct);
-	DECLARE_WAITQUEUE(wait, current);
+	struct wbt_wait_data data = {
+		.wq = {
+			.func	= wbt_wake_function,
+			.entry	= LIST_HEAD_INIT(data.wq.entry),
+		},
+		.task = current,
+		.rwb = rwb,
+		.rqw = rqw,
+		.rw = rw,
+	};
 	bool has_sleeper;
 
 	has_sleeper = wq_has_sleeper(&rqw->wait);
@@ -565,13 +565,25 @@ static void __wbt_wait(struct rq_wb *rwb
 			atomic_inc_below(&rqw->inflight, get_limit(rwb, rw)))
 		return;
 
-	add_wait_queue_exclusive(&rqw->wait, &wait);
+	prepare_to_wait_exclusive(&rqw->wait, &data.wq, TASK_UNINTERRUPTIBLE);
 	do {
 		set_current_state(TASK_UNINTERRUPTIBLE);
+		if (data.got_token)
+			break;
 
 		if (!has_sleeper &&
-		    atomic_inc_below(&rqw->inflight, get_limit(rwb, rw)))
+		    atomic_inc_below(&rqw->inflight, get_limit(rwb, rw))) {
+			finish_wait(&rqw->wait, &data.wq);
+
+			/*
+			 * We raced with wbt_wake_function() getting a token,
+			 * which means we now have two. Put our local token
+			 * and wake anyone else potentially waiting for one.
+			 */
+			if (data.got_token)
+				wbt_rqw_done(rwb, rqw, wb_acct);
 			break;
+		}
 
 		if (lock) {
 			spin_unlock_irq(lock);
@@ -547,11 +596,11 @@ static void __wbt_wait(struct rq_wb *rwb
 			spin_lock_irq(lock);
 		} else
 			io_schedule();
+
 		has_sleeper = false;
 	} while (1);
 
-	__set_current_state(TASK_RUNNING);
-	remove_wait_queue(&rqw->wait, &wait);
+	finish_wait(&rqw->wait, &data.wq);
 }
 
 static inline bool wbt_should_throttle(struct rq_wb *rwb, struct bio *bio)