Re: [PATCH] mm/workingset: do not forget to unlock page

[Andrew Morton]

On Wed, 3 Feb 2016 11:24:00 -0500 Johannes Weiner <hannes@xxxxxxxxxxx> wrote:

> On Wed, Feb 03, 2016 at 07:41:36PM +0900, Sergey Senozhatsky wrote:
> > From 1d6315221f2f81c53c99f9980158f8ae49dbd582 Mon Sep 17 00:00:00 2001
> > From: Sergey Senozhatsky <sergey.senozhatsky@xxxxxxxxx>
> > Date: Wed, 3 Feb 2016 18:49:16 +0900
> > Subject: [PATCH] mm/workingset: do not forget to unlock_page in workingset_activation
> > 
> > Do not return from workingset_activation() with locked rcu and page.
> > 
> > Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@xxxxxxxxx>
> 
> Thanks Sergey. Even though I wrote this function, my brain must have
> gone "it can't be locking anything when it returns NULL, right?" It's
> a dumb interface. Luckily, that's fixed with follow-up patches in -mm.
> 
> As for this one:
> 
> Acked-by: Johannes Weiner <hannes@xxxxxxxxxxx>
> Fixes: mm: workingset: per-cgroup cache thrash detection
> 
> Andrew, can you please fold this?

Yup.  I turned it into a fix against
mm-workingset-per-cgroup-cache-thrash-detection.patch, which is where
the bug was added.  And I did the goto thing instead, so the final
result will be

void workingset_activation(struct page *page)
{
	struct lruvec *lruvec;

	lock_page_memcg(page);
	/*
	 * Filter non-memcg pages here, e.g. unmap can call
	 * mark_page_accessed() on VDSO pages.
	 *
	 * XXX: See workingset_refault() - this should return
	 * root_mem_cgroup even for !CONFIG_MEMCG.
	 */
	if (!mem_cgroup_disabled() && !page_memcg(page))
		goto out;
	lruvec = mem_cgroup_zone_lruvec(page_zone(page), page_memcg(page));
	atomic_long_inc(&lruvec->inactive_age);
out:
	unlock_page_memcg(page);
}