linux-kernel - Re: [PATCH] mm: fix unsafe page -> lruvec lookups with cgroup charge migration

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 20 Nov 2019 16:39:31 -0500
From:   Johannes Weiner <hannes@...xchg.org>
To:     Shakeel Butt <shakeelb@...gle.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        Hugh Dickins <hughd@...gle.com>,
        Michal Hocko <mhocko@...e.com>,
        Alex Shi <alex.shi@...ux.alibaba.com>,
        Roman Gushchin <guro@...com>, Linux MM <linux-mm@...ck.org>,
        Cgroups <cgroups@...r.kernel.org>,
        LKML <linux-kernel@...r.kernel.org>,
        Kernel Team <kernel-team@...com>
Subject: Re: [PATCH] mm: fix unsafe page -> lruvec lookups with cgroup charge
 migration

On Wed, Nov 20, 2019 at 12:31:06PM -0800, Shakeel Butt wrote:
> On Wed, Nov 20, 2019 at 8:58 AM Johannes Weiner <hannes@...xchg.org> wrote:
> >
> > While reviewing the "per lruvec lru_lock for memcg" series, Hugh and I
> > noticed two places in the existing code where the page -> memcg ->
> > lruvec lookup can result in a use-after-free bug. This affects cgroup1
> > setups that have charge migration enabled.
> >
> > To pin page->mem_cgroup, callers need to either have the page locked,
> > an exclusive refcount (0), or hold the lru_lock and "own" PageLRU
> > (either ensure it's set, or be the one to hold the page in isolation)
> > to make cgroup migration fail the isolation step.
> 
> I think we should add the above para in the comments for better visibility.

Good idea. I'm attaching a delta patch below.

> > Reported-by: Hugh Dickins <hughd@...gle.com>
> > Signed-off-by: Johannes Weiner <hannes@...xchg.org>
> 
> Reviewed-by: Shakeel Butt <shakeelb@...gle.com>

Thanks!

---
>From 73b58ce09009cce668ea97d9e047611c60e95bd6 Mon Sep 17 00:00:00 2001
From: Johannes Weiner <hannes@...xchg.org>
Date: Wed, 20 Nov 2019 16:36:03 -0500
Subject: [PATCH] mm: fix unsafe page -> lruvec lookups with cgroup charge
 migration fix

Better document the mem_cgroup_page_lruvec() caller requirements.

Suggested-by: Shakeel Butt <shakeelb@...gle.com>
Signed-off-by: Johannes Weiner <hannes@...xchg.org>
---
 mm/memcontrol.c | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 50f5bc55fcec..2d700fa0d7f4 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1202,9 +1202,18 @@ int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
  * @page: the page
  * @pgdat: pgdat of the page
  *
- * This function is only safe when following the LRU page isolation
- * and putback protocol: the LRU lock must be held, and the page must
- * either be PageLRU() or the caller must have isolated/allocated it.
+ * NOTE: The returned lruvec is only stable if the calling context has
+ * the page->mem_cgroup pinned! This is accomplished by satisfying one
+ * of the following criteria:
+ *
+ *    a) have the @page locked
+ *    b) have an exclusive reference to @page (e.g. refcount 0)
+ *    c) hold the lru_lock and "own" the PageLRU (meaning either ensure
+ *       it's set, or be the one to hold the page in isolation)
+ *
+ * Otherwise, the page could be freed or moved out of the memcg,
+ * thereby releasing its reference on the memcg and potentially
+ * freeing it and its lruvecs in the process.
  */
 struct lruvec *mem_cgroup_page_lruvec(struct page *page, struct pglist_data *pgdat)
 {
-- 
2.24.0