[PATCH] mm, memcg: fix potential undefined when for page stat accounting

From: Michal Hocko
Date: Thu Oct 30 2014 - 13:35:19 EST

since d7365e783edb (mm: memcontrol: fix missed end-writeback page
accounting) mem_cgroup_end_page_stat consumes locked and flags variables
directly rather than via pointers which might trigger C undefined
behavior as those variables are initialized only in the slow path of
mem_cgroup_begin_page_stat.
Although mem_cgroup_end_page_stat handles parameters correctly and
touches them only when they hold a sensible value it is caller which
loads a potentially uninitialized value which then might allow compiler
to do crazy things.

I haven't seen any warning from gcc and it seems that the current
version (4.9) doesn't exploit this type undefined behavior but Sasha has
reported the following:
[ 26.868116] ================================================================================
[ 26.870376] UBSan: Undefined behaviour in mm/rmap.c:1084:2
[ 26.871792] load of value 255 is not a valid value for type '_Bool'
[ 26.873256] CPU: 4 PID: 8304 Comm: rngd Not tainted 3.18.0-rc2-next-20141029-sasha-00039-g77ed13d-dirty #1427
[ 26.875636] ffff8800cac17ff0 0000000000000000 0000000000000000 ffff880069ffbb28
[ 26.877611] ffffffffaf010c16 0000000000000037 ffffffffb1c0d050 ffff880069ffbb38
[ 26.879140] ffffffffa6e97899 ffff880069ffbbb8 ffffffffa6e97cc7 ffff880069ffbbb8
[ 26.880765] Call Trace:
[ 26.881185] dump_stack (lib/dump_stack.c:52)
[ 26.882755] ubsan_epilogue (lib/ubsan.c:159)
[ 26.883555] __ubsan_handle_load_invalid_value (lib/ubsan.c:482)
[ 26.884492] ? mem_cgroup_begin_page_stat (mm/memcontrol.c:1962)
[ 26.885441] ? unmap_page_range (./arch/x86/include/asm/paravirt.h:694 mm/memory.c:1091 mm/memory.c:1258 mm/memory.c:1279 mm/memory.c:1303)
[ 26.886242] page_remove_rmap (mm/rmap.c:1084 mm/rmap.c:1096)
[ 26.886922] unmap_page_range (./arch/x86/include/asm/atomic.h:27 include/linux/mm.h:463 mm/memory.c:1146 mm/memory.c:1258 mm/memory.c:1279 mm/memory.c:1303)
[ 26.887824] unmap_single_vma (mm/memory.c:1348)
[ 26.888582] unmap_vmas (mm/memory.c:1377 (discriminator 3))
[ 26.889430] exit_mmap (mm/mmap.c:2837)
[ 26.890060] mmput (kernel/fork.c:659)
[ 26.890656] do_exit (./arch/x86/include/asm/thread_info.h:168 kernel/exit.c:462 kernel/exit.c:747)
[ 26.891359] ? __this_cpu_preempt_check (lib/smp_processor_id.c:63)
[ 26.892287] ? trace_hardirqs_on_caller (kernel/locking/lockdep.c:2559 kernel/locking/lockdep.c:2601)
[ 26.893107] ? syscall_trace_enter_phase2 (arch/x86/kernel/ptrace.c:1598 (discriminator 2))
[ 26.893974] do_group_exit (include/linux/sched.h:775 kernel/exit.c:873)
[ 26.894695] SyS_exit_group (kernel/exit.c:901)
[ 26.895433] tracesys_phase2 (arch/x86/kernel/entry_64.S:529)
[ 26.896134] ================================================================================

Fix this by using pointer parameters for both locked and flags and be
more robust for future compiler changes even though the current code is
implemented correctly.

Reported-by: Sasha Levin <sasha.levin@xxxxxxxxxx>
Signed-off-by: Michal Hocko <mhocko@xxxxxxx>
---
include/linux/memcontrol.h | 6 +++---
mm/memcontrol.c | 8 ++++----
mm/page-writeback.c | 4 ++--
mm/rmap.c | 4 ++--
4 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index d4575a1d6e99..de018766be45 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -141,8 +141,8 @@ static inline bool mem_cgroup_disabled(void)

struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page, bool *locked,
unsigned long *flags);
-void mem_cgroup_end_page_stat(struct mem_cgroup *memcg, bool locked,
- unsigned long flags);
+void mem_cgroup_end_page_stat(struct mem_cgroup *memcg, bool *locked,
+ unsigned long *flags);
void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
enum mem_cgroup_stat_index idx, int val);

@@ -297,7 +297,7 @@ static inline struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page,
}

static inline void mem_cgroup_end_page_stat(struct mem_cgroup *memcg,
- bool locked, unsigned long flags)
+ bool *locked, unsigned long *flags)
{
}

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index b841bf430179..031ca345677b 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -2053,11 +2053,11 @@ again:
* @locked: value received from mem_cgroup_begin_page_stat()
* @flags: value received from mem_cgroup_begin_page_stat()
*/
-void mem_cgroup_end_page_stat(struct mem_cgroup *memcg, bool locked,
- unsigned long flags)
+void mem_cgroup_end_page_stat(struct mem_cgroup *memcg, bool *locked,
+ unsigned long *flags)
{
- if (memcg && locked)
- spin_unlock_irqrestore(&memcg->move_lock, flags);
+ if (memcg && *locked)
+ spin_unlock_irqrestore(&memcg->move_lock, *flags);

rcu_read_unlock();
}
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 19ceae87522d..d5d81f5384d1 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -2357,7 +2357,7 @@ int test_clear_page_writeback(struct page *page)
dec_zone_page_state(page, NR_WRITEBACK);
inc_zone_page_state(page, NR_WRITTEN);
}
- mem_cgroup_end_page_stat(memcg, locked, memcg_flags);
+ mem_cgroup_end_page_stat(memcg, &locked, &memcg_flags);
return ret;
}

@@ -2399,7 +2399,7 @@ int __test_set_page_writeback(struct page *page, bool keep_write)
mem_cgroup_inc_page_stat(memcg, MEM_CGROUP_STAT_WRITEBACK);
inc_zone_page_state(page, NR_WRITEBACK);
}
- mem_cgroup_end_page_stat(memcg, locked, memcg_flags);
+ mem_cgroup_end_page_stat(memcg, &locked, &memcg_flags);
return ret;

}
diff --git a/mm/rmap.c b/mm/rmap.c
index 740dd7d15806..f782b6c2ae48 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1051,7 +1051,7 @@ void page_add_file_rmap(struct page *page)
__inc_zone_page_state(page, NR_FILE_MAPPED);
mem_cgroup_inc_page_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
}
- mem_cgroup_end_page_stat(memcg, locked, flags);
+ mem_cgroup_end_page_stat(memcg, &locked, &flags);
}

static void page_remove_file_rmap(struct page *page)
@@ -1081,7 +1081,7 @@ static void page_remove_file_rmap(struct page *page)
if (unlikely(PageMlocked(page)))
clear_page_mlock(page);
out:
- mem_cgroup_end_page_stat(memcg, locked, flags);
+ mem_cgroup_end_page_stat(memcg, &locked, &flags);
}

/**
--
2.1.1

--
Michal Hocko
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/