Re: [PATCH V2 2/5] memcg: add function should_reclaim_mem_cgroup()

[Ying Han <yinghan@google.com>]

On Wed, Apr 11, 2012 at 5:19 PM, Johannes Weiner <hannes@cmpxchg.org> wrote:
> On Wed, Apr 11, 2012 at 03:00:06PM -0700, Ying Han wrote:
>> Add the filter function should_reclaim_mem_cgroup() under the common function
>> shrink_zone(). The later one is being called both from per-memcg reclaim as
>> well as global reclaim.
>>
>> Today the softlimit takes effect only under global memory pressure. The memcgs
>> get free run above their softlimit until there is a global memory contention.
>> This patch doesn't change the semantics.
>>
>> Under the global reclaim, we skip reclaiming from a memcg under its softlimit.
>> To prevent reclaim from trying too hard on hitting memcgs (above softlimit) w/
>> only hard-to-reclaim pages, the reclaim proirity is used to skip the softlimit
>> check. This is a trade-off of system performance and resource isolation.
>>
>> Signed-off-by: Ying Han <yinghan@google.com>
>> ---
>>  include/linux/memcontrol.h |    7 +++++++
>>  mm/memcontrol.c            |   10 +++++++++-
>>  mm/vmscan.c                |   25 ++++++++++++++++++++++++-
>>  3 files changed, 40 insertions(+), 2 deletions(-)
>>
>> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
>> index db71193..3d14f90 100644
>> --- a/include/linux/memcontrol.h
>> +++ b/include/linux/memcontrol.h
>> @@ -110,6 +110,8 @@ struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
>>                                  struct mem_cgroup_reclaim_cookie *);
>>  void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
>>
>> +bool mem_cgroup_soft_limit_exceeded(struct mem_cgroup *);
>> +
>>  /*
>>   * For memory reclaim.
>>   */
>> @@ -295,6 +297,11 @@ static inline void mem_cgroup_iter_break(struct mem_cgroup *root,
>>  {
>>  }
>>
>> +static inline bool mem_cgroup_soft_limit_exceeded(struct mem_cgroup *mem)
>> +{
>> +     return true;
>> +}
>> +
>>  static inline int mem_cgroup_get_reclaim_priority(struct mem_cgroup *memcg)
>>  {
>>       return 0;
>> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
>> index 9a64093..cffcded 100644
>> --- a/mm/memcontrol.c
>> +++ b/mm/memcontrol.c
>> @@ -358,12 +358,12 @@ enum charge_type {
>>  static void mem_cgroup_get(struct mem_cgroup *memcg);
>>  static void mem_cgroup_put(struct mem_cgroup *memcg);
>>
>> +static bool mem_cgroup_is_root(struct mem_cgroup *memcg);
>>  /* Writing them here to avoid exposing memcg's inner layout */
>>  #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
>>  #include <net/sock.h>
>>  #include <net/ip.h>
>>
>> -static bool mem_cgroup_is_root(struct mem_cgroup *memcg);
>
> The prototype is hardly shorter than the friggin function itself!
>
> I'll send a patch to remove this thing completely, doing memcg ==
> root_mem_cgroup should be pretty obvious without a helper function.

Ok, I will leave as it is now. Before your post, i will just do memcg
== root_mem_cgroup.

>> @@ -2133,6 +2133,27 @@ restart:
>>       throttle_vm_writeout(sc->gfp_mask);
>>  }
>>
>> +static bool should_reclaim_mem_cgroup(struct mem_cgroup *target_mem_cgroup,
>> +                                   struct mem_cgroup *memcg,
>> +                                   int priority)
>> +{
>> +     /* Reclaim from mem_cgroup if any of these conditions are met:
>> +      * - This is a global reclaim
>> +      * - reclaim priority is higher than DEF_PRIORITY - 3
>> +      * - mem_cgroup exceeds its soft limit
>> +      *
>> +      * The priority check is a balance of how hard to preserve the pages
>> +      * under softlimit. If the memcgs of the zone having trouble to reclaim
>> +      * pages above their softlimit, we have to reclaim under softlimit
>> +      * instead of burning more cpu cycles.
>> +      */
>> +     if (target_mem_cgroup || priority <= DEF_PRIORITY - 3 ||
>> +                     mem_cgroup_soft_limit_exceeded(memcg))
>> +             return true;
>
> The comment is contradicting the code: global reclaim does not scan
> unconditionally, hard limit reclaim does.  Global reclaim scans only
> if the memcg is above soft limit or if the priority level dropped
> sufficiently.
>
> I suppose it's the comment that's wrong, not the code.

You are right. I will fix the comment on next post.

--Ying

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>