On 22.04.2018 20:59, Vladimir Davydov wrote:
> On Tue, Apr 17, 2018 at 09:53:31PM +0300, Kirill Tkhai wrote:
>> Imagine a big node with many cpus, memory cgroups and containers.
>> Let we have 200 containers, every container has 10 mounts,
>> and 10 cgroups. All container tasks don't touch foreign
>> containers mounts. If there is intensive pages write,
>> and global reclaim happens, a writing task has to iterate
>> over all memcgs to shrink slab, before it's able to go
>> to shrink_page_list().
>>
>> Iteration over all the memcg slabs is very expensive:
>> the task has to visit 200 * 10 = 2000 shrinkers
>> for every memcg, and since there are 2000 memcgs,
>> the total calls are 2000 * 2000 = 4000000.
>>
>> So, the shrinker makes 4 million do_shrink_slab() calls
>> just to try to isolate SWAP_CLUSTER_MAX pages in one
>> of the actively writing memcg via shrink_page_list().
>> I've observed a node spending almost 100% in kernel,
>> making useless iteration over already shrinked slab.
>>
>> This patch adds bitmap of memcg-aware shrinkers to memcg.
>> The size of the bitmap depends on bitmap_nr_ids, and during
>> memcg life it's maintained to be enough to fit bitmap_nr_ids
>> shrinkers. Every bit in the map is related to corresponding
>> shrinker id.
>>
>> Next patches will maintain set bit only for really charged
>> memcg. This will allow shrink_slab() to increase its
>> performance in significant way. See the last patch for
>> the numbers.
>>
>> Signed-off-by: Kirill Tkhai <ktkhai@xxxxxxxxxxxxx>
>> ---
>> include/linux/memcontrol.h | 15 +++++
>> mm/memcontrol.c | 125 ++++++++++++++++++++++++++++++++++++++++++++
>> mm/vmscan.c | 21 +++++++
>> 3 files changed, 160 insertions(+), 1 deletion(-)
>>
>> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
>> index af9eed2e3e04..2ec96ab46b01 100644
>> --- a/include/linux/memcontrol.h
>> +++ b/include/linux/memcontrol.h
>> @@ -115,6 +115,7 @@ struct mem_cgroup_per_node {
>> unsigned long lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
>>
>> struct mem_cgroup_reclaim_iter iter[DEF_PRIORITY + 1];
>
>> + struct memcg_shrinker_map __rcu *shrinkers_map;
>
> shrinker_map
>
>>
>> struct rb_node tree_node; /* RB tree node */
>> unsigned long usage_in_excess;/* Set to the value by which */
>> @@ -153,6 +154,11 @@ struct mem_cgroup_thresholds {
>> struct mem_cgroup_threshold_ary *spare;
>> };
>>
>> +struct memcg_shrinker_map {
>> + struct rcu_head rcu;
>> + unsigned long map[0];
>> +};
>> +
>
> This struct should be defined before struct mem_cgroup_per_node.
>
> A comment explaining what this map is for and what it maps would be
> really helpful.
>
>> enum memcg_kmem_state {
>> KMEM_NONE,
>> KMEM_ALLOCATED,
>> @@ -1200,6 +1206,8 @@ extern int memcg_nr_cache_ids;
>> void memcg_get_cache_ids(void);
>> void memcg_put_cache_ids(void);
>>
>> +extern int shrinkers_max_nr;
>> +
>
> memcg_shrinker_id_max?
memcg_shrinker_id_max sounds like an includive value, doesn't it?
While shrinker->id < shrinker_max_nr.
Let's better use memcg_shrinker_nr_max.
>> /*
>> * Helper macro to loop through all memcg-specific caches. Callers must still
>> * check if the cache is valid (it is either valid or NULL).
>> @@ -1223,6 +1231,13 @@ static inline int memcg_cache_id(struct mem_cgroup *memcg)
>> return memcg ? memcg->kmemcg_id : -1;
>> }
>>
>> +extern struct memcg_shrinker_map __rcu *root_shrinkers_map[];
>> +#define SHRINKERS_MAP(memcg, nid) \
>> + (memcg == root_mem_cgroup || !memcg ? \
>> + root_shrinkers_map[nid] : memcg->nodeinfo[nid]->shrinkers_map)
>> +
>> +extern int expand_shrinker_maps(int old_id, int id);
>> +
>
> I'm strongly against using a special map for the root cgroup. I'd prefer
> to disable this optimization for the root cgroup altogether and simply
> iterate over all registered shrinkers when it comes to global reclaim.
> It shouldn't degrade performance as the root cgroup is singular.
>
>> #else
>> #define for_each_memcg_cache_index(_idx) \
>> for (; NULL; )
>> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
>> index 2959a454a072..562dfb1be9ef 100644
>> --- a/mm/memcontrol.c
>> +++ b/mm/memcontrol.c
>> @@ -305,6 +305,113 @@ EXPORT_SYMBOL(memcg_kmem_enabled_key);
>>
>> struct workqueue_struct *memcg_kmem_cache_wq;
>>
>> +static DECLARE_RWSEM(shrinkers_max_nr_rwsem);
>
> Why rwsem? AFAIU you want to synchronize between two code paths: when a
> memory cgroup is allocated (or switched online?) to allocate a shrinker
> map for it and in the function growing shrinker maps for all cgroups.
> A mutex should be enough for this.
>
>> +struct memcg_shrinker_map __rcu *root_shrinkers_map[MAX_NUMNODES] = { 0 };
>> +
>> +static void get_shrinkers_max_nr(void)
>> +{
>> + down_read(&shrinkers_max_nr_rwsem);
>> +}
>> +
>> +static void put_shrinkers_max_nr(void)
>> +{
>> + up_read(&shrinkers_max_nr_rwsem);
>> +}
>> +
>> +static void kvfree_map_rcu(struct rcu_head *head)
>
> free_shrinker_map_rcu
>
>> +{
>> + kvfree(container_of(head, struct memcg_shrinker_map, rcu));
>> +}
>> +
>> +static int memcg_expand_maps(struct mem_cgroup *memcg, int nid,
>
> Bad name: the function reallocates just one map, not many maps; the name
> doesn't indicate that it is about shrinkers; the name is inconsistent
> with alloc_shrinker_maps and free_shrinker_maps. Please fix.
>
>> + int size, int old_size)
>> +{
>> + struct memcg_shrinker_map *new, *old;
>> +
>> + lockdep_assert_held(&shrinkers_max_nr_rwsem);
>> +
>> + new = kvmalloc(sizeof(*new) + size, GFP_KERNEL);
>> + if (!new)
>> + return -ENOMEM;
>> +
>> + /* Set all old bits, clear all new bits */
>> + memset(new->map, (int)0xff, old_size);
>> + memset((void *)new->map + old_size, 0, size - old_size);
>> +
>> + old = rcu_dereference_protected(SHRINKERS_MAP(memcg, nid), true);
>> +
>> + if (memcg)
>> + rcu_assign_pointer(memcg->nodeinfo[nid]->shrinkers_map, new);
>> + else
>> + rcu_assign_pointer(root_shrinkers_map[nid], new);
>> +
>> + if (old)
>> + call_rcu(&old->rcu, kvfree_map_rcu);
>> +
>> + return 0;
>> +}
>> +
>> +static int alloc_shrinker_maps(struct mem_cgroup *memcg, int nid)
>> +{
>> + /* Skip allocation, when we're initializing root_mem_cgroup */
>> + if (!root_mem_cgroup)
>> + return 0;
>> +
>> + return memcg_expand_maps(memcg, nid, shrinkers_max_nr/BITS_PER_BYTE, 0);
>> +}
>> +
>> +static void free_shrinker_maps(struct mem_cgroup *memcg,
>> + struct mem_cgroup_per_node *pn)
>> +{
>> + struct memcg_shrinker_map *map;
>> +
>> + if (memcg == root_mem_cgroup)
>> + return;
>> +
>> + /* IDR unhashed long ago, and expand_shrinker_maps can't race with us */
>> + map = rcu_dereference_protected(pn->shrinkers_map, true);
>> + kvfree_map_rcu(&map->rcu);
>> +}
>> +
>> +static struct idr mem_cgroup_idr;
>> +
>> +int expand_shrinker_maps(int old_nr, int nr)
>> +{
>> + int id, size, old_size, node, ret;
>> + struct mem_cgroup *memcg;
>> +
>> + old_size = old_nr / BITS_PER_BYTE;
>> + size = nr / BITS_PER_BYTE;
>> +
>> + down_write(&shrinkers_max_nr_rwsem);
>> + for_each_node(node) {
>
> Iterating over cgroups first, numa nodes second seems like a better idea
> to me. I think you should fold for_each_node in memcg_expand_maps.
>
>> + idr_for_each_entry(&mem_cgroup_idr, memcg, id) {
>
> Iterating over mem_cgroup_idr looks strange. Why don't you use
> for_each_mem_cgroup?
We want to allocate shrinkers maps in mem_cgroup_css_alloc(), since
mem_cgroup_css_online() mustn't fail (it's a requirement of currently
existing design of memcg_cgroup::id).
A new memcg is added to parent's list between two of these calls:
css_create()
ss->css_alloc()
list_add_tail_rcu(&css->sibling, &parent_css->children)
ss->css_online()
for_each_mem_cgroup() does not see allocated, but not linked children.
>> + if (id == 1)
>> + memcg = NULL;
>> + ret = memcg_expand_maps(memcg, node, size, old_size);
>> + if (ret)
>> + goto unlock;
>> + }
>> +
>> + /* root_mem_cgroup is not initialized yet */
>> + if (id == 0)
>> + ret = memcg_expand_maps(NULL, node, size, old_size);
>> + }
>> +unlock:
>> + up_write(&shrinkers_max_nr_rwsem);
>> + return ret;
>> +}
>> +#else /* CONFIG_SLOB */
>> +static void get_shrinkers_max_nr(void) { }
>> +static void put_shrinkers_max_nr(void) { }
>> +
>> +static int alloc_shrinker_maps(struct mem_cgroup *memcg, int nid)
>> +{
>> + return 0;
>> +}
>> +static void free_shrinker_maps(struct mem_cgroup *memcg,
>> + struct mem_cgroup_per_node *pn) { }
>> +
>> #endif /* !CONFIG_SLOB */
>>
>> /**
>> @@ -3002,6 +3109,8 @@ static u64 mem_cgroup_read_u64(struct cgroup_subsys_state *css,
>> }
>>
>> #ifndef CONFIG_SLOB
>> +int shrinkers_max_nr;
>> +
>> static int memcg_online_kmem(struct mem_cgroup *memcg)
>> {
>> int memcg_id;
>> @@ -4266,7 +4375,10 @@ static DEFINE_IDR(mem_cgroup_idr);
>> static void mem_cgroup_id_remove(struct mem_cgroup *memcg)
>> {
>> if (memcg->id.id > 0) {
>> + /* Removing IDR must be visible for expand_shrinker_maps() */
>> + get_shrinkers_max_nr();
>> idr_remove(&mem_cgroup_idr, memcg->id.id);
>> + put_shrinkers_max_nr();
>> memcg->id.id = 0;
>> }
>> }
>> @@ -4333,12 +4445,17 @@ static int alloc_mem_cgroup_per_node_info(struct mem_cgroup *memcg, int node)
>> if (!pn->lruvec_stat_cpu)
>> goto err_pcpu;
>>
>> + if (alloc_shrinker_maps(memcg, node))
>> + goto err_maps;
>> +
>> lruvec_init(&pn->lruvec);
>> pn->usage_in_excess = 0;
>> pn->on_tree = false;
>> pn->memcg = memcg;
>> return 0;
>>
>> +err_maps:
>> + free_percpu(pn->lruvec_stat_cpu);
>> err_pcpu:
>> memcg->nodeinfo[node] = NULL;
>> kfree(pn);
>> @@ -4352,6 +4469,7 @@ static void free_mem_cgroup_per_node_info(struct mem_cgroup *memcg, int node)
>> if (!pn)
>> return;
>>
>> + free_shrinker_maps(memcg, pn);
>> free_percpu(pn->lruvec_stat_cpu);
>> kfree(pn);
>> }
>> @@ -4407,13 +4525,18 @@ static struct mem_cgroup *mem_cgroup_alloc(void)
>> #ifdef CONFIG_CGROUP_WRITEBACK
>> INIT_LIST_HEAD(&memcg->cgwb_list);
>> #endif
>> +
>> + get_shrinkers_max_nr();
>> for_each_node(node)
>> - if (alloc_mem_cgroup_per_node_info(memcg, node))
>> + if (alloc_mem_cgroup_per_node_info(memcg, node)) {
>> + put_shrinkers_max_nr();
>> goto fail;
>> + }
>>
>> memcg->id.id = idr_alloc(&mem_cgroup_idr, memcg,
>> 1, MEM_CGROUP_ID_MAX,
>> GFP_KERNEL);
>> + put_shrinkers_max_nr();
>> if (memcg->id.id < 0)
>> goto fail;
>>
>> diff --git a/mm/vmscan.c b/mm/vmscan.c
>> index 4f02fe83537e..f63eb5596c35 100644
>> --- a/mm/vmscan.c
>> +++ b/mm/vmscan.c
>> @@ -172,6 +172,22 @@ static DECLARE_RWSEM(shrinker_rwsem);
>> #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
>> static DEFINE_IDR(shrinkers_id_idr);
>>
>> +static int expand_shrinker_id(int id)
>> +{
>> + if (likely(id < shrinkers_max_nr))
>> + return 0;
>> +
>> + id = shrinkers_max_nr * 2;
>> + if (id == 0)
>> + id = BITS_PER_BYTE;
>> +
>> + if (expand_shrinker_maps(shrinkers_max_nr, id))
>> + return -ENOMEM;
>> +
>> + shrinkers_max_nr = id;
>> + return 0;
>> +}
>> +
>
> I think this function should live in memcontrol.c and shrinkers_max_nr
> should be private to memcontrol.c.
>
>> static int add_memcg_shrinker(struct shrinker *shrinker)
>> {
>> int id, ret;
>> @@ -180,6 +196,11 @@ static int add_memcg_shrinker(struct shrinker *shrinker)
>> ret = id = idr_alloc(&shrinkers_id_idr, shrinker, 0, 0, GFP_KERNEL);
>> if (ret < 0)
>> goto unlock;
>> + ret = expand_shrinker_id(id);
>> + if (ret < 0) {
>> + idr_remove(&shrinkers_id_idr, id);
>> + goto unlock;
>> + }
>> shrinker->id = id;
>> ret = 0;
>> unlock:
>>
