[PATCH 0/3] blk-iolatency related ref counting fixes

[PATCH 0/3] blk-iolatency related ref counting fixes

[Josef Bacik]

These are three ref counting issues we found in testing.  They are pretty
straightforward, the first two don't really happen in practice but need to be
fixed, but the last one fixes a real panic we saw pretty regularly in our stress
testing.  They've been running on our test boxes internally under stress tests
and seem to have fixed the remaining issues we were seeing.  Thanks,

Josef

[PATCH 1/3] blk-iolatency: fix blkg leak in timer_fn

[Josef Bacik]

At this point we have a ref on the blkg, we need to drop it if we don't
have a iolat.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
---
 block/blk-iolatency.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/block/blk-iolatency.c b/block/blk-iolatency.c
index bb59b2929e0d..b0dc4fc64b3e 100644
--- a/block/blk-iolatency.c
+++ b/block/blk-iolatency.c
@@ -627,7 +627,7 @@ static void blkiolatency_timer_fn(struct timer_list *t)
 
 		iolat = blkg_to_lat(blkg);
 		if (!iolat)
-			continue;
+			goto next;
 
 		lat_info = &iolat->child_lat;
 		cookie = atomic_read(&lat_info->scale_cookie);
-- 
2.14.3

[PATCH 2/3] blk-cgroup: hold the queue ref during throttling

[Josef Bacik]

The blkg lifetime is protected by the queue lifetime, so we need to put
the queue _after_ we're done using the blkg.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
---
 block/blk-cgroup.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index 1942357d7165..694595b29b8f 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -1759,10 +1759,10 @@ void blkcg_maybe_throttle_current(void)
 	if (!blkg)
 		goto out;
 	rcu_read_unlock();
-	blk_put_queue(q);
 
 	blkcg_maybe_throttle_blkg(blkg, use_memdelay);
 	blkg_put(blkg);
+	blk_put_queue(q);
 	return;
 out:
 	rcu_read_unlock();
-- 
2.14.3

[PATCH 3/3] blk-cgroup: clear the throttle queue on fork

[Josef Bacik]

We were hitting a panic in production where we put too many times on the
request queue.  This is because we'd get the throttle_queue of the
parent if we fork()'ed while we needed to be throttled, but we didn't
have a reference on it.  Instead just clear these flags on fork so the
child doesn't pay for the sins of its father.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
---
 kernel/fork.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/kernel/fork.c b/kernel/fork.c
index 9440d61b925c..694ae0e56866 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -843,6 +843,11 @@ static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
 	tsk->fail_nth = 0;
 #endif
 
+#ifdef CONFIG_BLK_CGROUP
+	tsk->throttle_queue = NULL;
+	tsk->use_memdelay = 0;
+#endif
+
 	return tsk;
 
 free_stack:
-- 
2.14.3

Re: [PATCH 0/3] blk-iolatency related ref counting fixes

[Jens Axboe]

On 7/31/18 10:39 AM, Josef Bacik wrote:
> These are three ref counting issues we found in testing.  They are pretty
> straightforward, the first two don't really happen in practice but need to be
> fixed, but the last one fixes a real panic we saw pretty regularly in our stress
> testing.  They've been running on our test boxes internally under stress tests
> and seem to have fixed the remaining issues we were seeing.  Thanks,

Applied for 4.19, thanks.

-- 
Jens Axboe