lkml.org 
[lkml]   [2011]   [Jun]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] memcg: do not expose uninitialized mem_cgroup_per_node to world
On 06/01/2011 02:39 PM, Michal Hocko wrote:
> On Wed 01-06-11 12:44:04, Igor Mammedov wrote:
>> Freshly allocated 'mem_cgroup_per_node' list entries must be
>> initialized before the rest of the kernel can see them. Otherwise
>> zero initialized list fields can lead to race condition at
>> mem_cgroup_force_empty_list:
>> pc = list_entry(list->prev, struct page_cgroup, lru);
>> where 'pc' will be something like 0xfffffffc if list->prev is 0
>> and cause page fault later when 'pc' is dereferenced.
> Have you ever seen such a race? I do not see how this could happen.
> mem_cgroup_force_empty_list is called only from
> mem_cgroup_force_empty_write (aka echo whatever> group/force_empty)
> or mem_cgroup_pre_destroy when the group is destroyed.
>
> The initialization code is, however, called before a group is
> given for use AFICS.
>
> I am not saying tha the change is bad, I like it, but I do not think it
> is a fix for potential race condition.
>

Yes I've seen it (RHBZ#700565). It causes random crashes
in virt env ocasionally. It's easier to reproduce if you overcommit
cpu.

>> Signed-off-by: Igor Mammedov<imammedo@redhat.com>
>> ---
>> mm/memcontrol.c | 2 +-
>> 1 files changed, 1 insertions(+), 1 deletions(-)
>>
>> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
>> index bd9052a..ee7cb4c 100644
>> --- a/mm/memcontrol.c
>> +++ b/mm/memcontrol.c
>> @@ -4707,7 +4707,6 @@ static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
>> if (!pn)
>> return 1;
>>
>> - mem->info.nodeinfo[node] = pn;
>> for (zone = 0; zone< MAX_NR_ZONES; zone++) {
>> mz =&pn->zoneinfo[zone];
>> for_each_lru(l)
>> @@ -4716,6 +4715,7 @@ static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
>> mz->on_tree = false;
>> mz->mem = mem;
>> }
>> + mem->info.nodeinfo[node] = pn;
>> return 0;
>> }
>>
>> --
>> 1.7.1
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at http://www.tux.org/lkml/



\
 
 \ /
  Last update: 2011-06-01 15:11    [W:0.164 / U:5.500 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site