From: Vladimir Davydov <vdavydov.dev@xxxxxxxxx> synchronize_sched() is a heavy operation and calling it per each cache owned by a memory cgroup being destroyed may take quite some time. What is worse, it's currently called under the slab_mutex, stalling all works doing cache creation/destruction. Actually, there isn't much point in calling synchronize_sched() for each cache - it's enough to call it just once - after setting cpu_partial for all caches and before shrinking them. This way, we can also move it out of the slab_mutex, which we have to hold for iterating over the slab cache list. Link: https://bugzilla.kernel.org/show_bug.cgi?id=172991 Link: http://lkml.kernel.org/r/0a10d71ecae3db00fb4421bcd3f82bcc911f4be4.1475329751.git.vdavydov.dev@xxxxxxxxx Signed-off-by: Vladimir Davydov <vdavydov.dev@xxxxxxxxx> Reported-by: Doug Smythies <dsmythies@xxxxxxxxx> Acked-by: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx> Cc: Christoph Lameter <cl@xxxxxxxxx> Cc: David Rientjes <rientjes@xxxxxxxxxx> Cc: Johannes Weiner <hannes@xxxxxxxxxxx> Cc: Michal Hocko <mhocko@xxxxxxxxxx> Cc: Pekka Enberg <penberg@xxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> Signed-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> Signed-off-by: Sasha Levin <alexander.levin@xxxxxxxxxxx> --- mm/slab.c | 4 ++-- mm/slab.h | 2 +- mm/slab_common.c | 27 +++++++++++++++++++++++++-- mm/slob.c | 2 +- mm/slub.c | 19 ++----------------- 5 files changed, 31 insertions(+), 23 deletions(-) diff --git a/mm/slab.c b/mm/slab.c index bd878f0..1f82d16 100644 --- a/mm/slab.c +++ b/mm/slab.c @@ -2332,7 +2332,7 @@ static int drain_freelist(struct kmem_cache *cache, return nr_freed; } -int __kmem_cache_shrink(struct kmem_cache *cachep, bool deactivate) +int __kmem_cache_shrink(struct kmem_cache *cachep) { int ret = 0; int node; @@ -2352,7 +2352,7 @@ int __kmem_cache_shrink(struct kmem_cache *cachep, bool deactivate) int __kmem_cache_shutdown(struct kmem_cache *cachep) { - return __kmem_cache_shrink(cachep, false); + return __kmem_cache_shrink(cachep); } void __kmem_cache_release(struct kmem_cache *cachep) diff --git a/mm/slab.h b/mm/slab.h index bc05fdc..ceb7d70 100644 --- a/mm/slab.h +++ b/mm/slab.h @@ -146,7 +146,7 @@ static inline unsigned long kmem_cache_flags(unsigned long object_size, int __kmem_cache_shutdown(struct kmem_cache *); void __kmem_cache_release(struct kmem_cache *); -int __kmem_cache_shrink(struct kmem_cache *, bool); +int __kmem_cache_shrink(struct kmem_cache *); void slab_kmem_cache_release(struct kmem_cache *); struct seq_file; diff --git a/mm/slab_common.c b/mm/slab_common.c index 329b038..5d2f24f 100644 --- a/mm/slab_common.c +++ b/mm/slab_common.c @@ -573,6 +573,29 @@ void memcg_deactivate_kmem_caches(struct mem_cgroup *memcg) get_online_cpus(); get_online_mems(); +#ifdef CONFIG_SLUB + /* + * In case of SLUB, we need to disable empty slab caching to + * avoid pinning the offline memory cgroup by freeable kmem + * pages charged to it. SLAB doesn't need this, as it + * periodically purges unused slabs. + */ + mutex_lock(&slab_mutex); + list_for_each_entry(s, &slab_caches, list) { + c = is_root_cache(s) ? cache_from_memcg_idx(s, idx) : NULL; + if (c) { + c->cpu_partial = 0; + c->min_partial = 0; + } + } + mutex_unlock(&slab_mutex); + /* + * kmem_cache->cpu_partial is checked locklessly (see + * put_cpu_partial()). Make sure the change is visible. + */ + synchronize_sched(); +#endif + mutex_lock(&slab_mutex); list_for_each_entry(s, &slab_caches, list) { if (!is_root_cache(s)) @@ -584,7 +607,7 @@ void memcg_deactivate_kmem_caches(struct mem_cgroup *memcg) if (!c) continue; - __kmem_cache_shrink(c, true); + __kmem_cache_shrink(c); arr->entries[idx] = NULL; } mutex_unlock(&slab_mutex); @@ -755,7 +778,7 @@ int kmem_cache_shrink(struct kmem_cache *cachep) get_online_cpus(); get_online_mems(); kasan_cache_shrink(cachep); - ret = __kmem_cache_shrink(cachep, false); + ret = __kmem_cache_shrink(cachep); put_online_mems(); put_online_cpus(); return ret; diff --git a/mm/slob.c b/mm/slob.c index 5ec1580..eac04d43 100644 --- a/mm/slob.c +++ b/mm/slob.c @@ -634,7 +634,7 @@ void __kmem_cache_release(struct kmem_cache *c) { } -int __kmem_cache_shrink(struct kmem_cache *d, bool deactivate) +int __kmem_cache_shrink(struct kmem_cache *d) { return 0; } diff --git a/mm/slub.c b/mm/slub.c index 7aa0e97..58c7526 100644 --- a/mm/slub.c +++ b/mm/slub.c @@ -3887,7 +3887,7 @@ EXPORT_SYMBOL(kfree); * being allocated from last increasing the chance that the last objects * are freed in them. */ -int __kmem_cache_shrink(struct kmem_cache *s, bool deactivate) +int __kmem_cache_shrink(struct kmem_cache *s) { int node; int i; @@ -3899,21 +3899,6 @@ int __kmem_cache_shrink(struct kmem_cache *s, bool deactivate) unsigned long flags; int ret = 0; - if (deactivate) { - /* - * Disable empty slabs caching. Used to avoid pinning offline - * memory cgroups by kmem pages that can be freed. - */ - s->cpu_partial = 0; - s->min_partial = 0; - - /* - * s->cpu_partial is checked locklessly (see put_cpu_partial), - * so we have to make sure the change is visible. - */ - synchronize_sched(); - } - flush_all(s); for_each_kmem_cache_node(s, node, n) { INIT_LIST_HEAD(&discard); @@ -3970,7 +3955,7 @@ static int slab_mem_going_offline_callback(void *arg) mutex_lock(&slab_mutex); list_for_each_entry(s, &slab_caches, list) - __kmem_cache_shrink(s, false); + __kmem_cache_shrink(s); mutex_unlock(&slab_mutex); return 0; -- 2.9.3