Re: [PATCH 08/10] mm: Set bit in memcg shrinker bitmap on first list_lru item apearance

[Kirill Tkhai <ktkhai@virtuozzo.com>]

On 24.03.2018 22:45, Vladimir Davydov wrote:
> On Wed, Mar 21, 2018 at 04:22:40PM +0300, Kirill Tkhai wrote:
>> Introduce set_shrinker_bit() function to set shrinker-related
>> bit in memcg shrinker bitmap, and set the bit after the first
>> item is added and in case of reparenting destroyed memcg's items.
>>
>> This will allow next patch to make shrinkers be called only,
>> in case of they have charged objects at the moment, and
>> to improve shrink_slab() performance.
>>
>> Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
>> ---
>>  include/linux/shrinker.h |    7 +++++++
>>  mm/list_lru.c            |   22 ++++++++++++++++++++--
>>  mm/vmscan.c              |    7 +++++++
>>  3 files changed, 34 insertions(+), 2 deletions(-)
>>
>> diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h
>> index 738de8ef5246..24aeed1bc332 100644
>> --- a/include/linux/shrinker.h
>> +++ b/include/linux/shrinker.h
>> @@ -78,4 +78,11 @@ struct shrinker {
>>  
>>  extern __must_check int register_shrinker(struct shrinker *);
>>  extern void unregister_shrinker(struct shrinker *);
>> +#if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
>> +extern void set_shrinker_bit(struct mem_cgroup *, int, int);
>> +#else
>> +static inline void set_shrinker_bit(struct mem_cgroup *memcg, int node, int id)
>> +{
>> +}
>> +#endif
> 
> IMO this function, as well as other shrinker bitmap manipulation
> functions, should be defined in memcontrol.[hc] and have mem_cgroup_
> prefix.
> 
>> diff --git a/mm/vmscan.c b/mm/vmscan.c
>> index 9d1df5d90eca..265cf069b470 100644
>> --- a/mm/vmscan.c
>> +++ b/mm/vmscan.c
>> @@ -378,6 +378,13 @@ static void del_shrinker(struct shrinker *shrinker)
>>  	list_del(&shrinker->list);
>>  	up_write(&shrinker_rwsem);
>>  }
>> +
>> +void set_shrinker_bit(struct mem_cgroup *memcg, int nid, int nr)
>> +{
>> +	struct shrinkers_map *map = SHRINKERS_MAP(memcg);
>> +
>> +	set_bit(nr, map->map[nid]);
>> +}
> 
> Shouldn't we use rcu_read_lock here? After all, the map can be
> reallocated right from under our feet.

We do have to do that! Thanks for pointing.

Kirill