From: Roman Gushchin <guro@xxxxxx> Subject: mm: memcg: charge memcg percpu memory to the parent cgroup Memory cgroups are using large chunks of percpu memory to store vmstat data. Yet this memory is not accounted at all, so in the case when there are many (dying) cgroups, it's not exactly clear where all the memory is. Because the size of memory cgroup internal structures can dramatically exceed the size of object or page which is pinning it in the memory, it's not a good idea to simply ignore it. It actually breaks the isolation between cgroups. Let's account the consumed percpu memory to the parent cgroup. [guro@xxxxxx: add WARN_ON_ONCE()s, per Johannes] Link: http://lkml.kernel.org/r/20200811170611.GB1507044@xxxxxxxxxxxxxxxxxxxxxxxxxxx Link: http://lkml.kernel.org/r/20200623184515.4132564-5-guro@xxxxxx Signed-off-by: Roman Gushchin <guro@xxxxxx> Acked-by: Dennis Zhou <dennis@xxxxxxxxxx> Reviewed-by: Shakeel Butt <shakeelb@xxxxxxxxxx> Acked-by: Johannes Weiner <hannes@xxxxxxxxxxx> Cc: Christoph Lameter <cl@xxxxxxxxx> Cc: David Rientjes <rientjes@xxxxxxxxxx> Cc: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx> Cc: Mel Gorman <mgorman@xxxxxxxxxxxxxxxxxxx> Cc: Michal Hocko <mhocko@xxxxxxxxxx> Cc: Pekka Enberg <penberg@xxxxxxxxxx> Cc: Tejun Heo <tj@xxxxxxxxxx> Cc: Tobin C. Harding <tobin@xxxxxxxxxx> Cc: Vlastimil Babka <vbabka@xxxxxxx> Cc: Waiman Long <longman@xxxxxxxxxx> Cc: Bixuan Cui <cuibixuan@xxxxxxxxxx> Cc: Michal Koutný <mkoutny@xxxxxxxx> Cc: Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- mm/memcontrol.c | 20 ++++++++++++++++---- 1 file changed, 16 insertions(+), 4 deletions(-) --- a/mm/memcontrol.c~mm-memcg-charge-memcg-percpu-memory-to-the-parent-cgroup +++ a/mm/memcontrol.c @@ -5131,13 +5131,18 @@ static int alloc_mem_cgroup_per_node_inf if (!pn) return 1; - pn->lruvec_stat_local = alloc_percpu(struct lruvec_stat); + /* We charge the parent cgroup, never the current task */ + WARN_ON_ONCE(!current->active_memcg); + + pn->lruvec_stat_local = alloc_percpu_gfp(struct lruvec_stat, + GFP_KERNEL_ACCOUNT); if (!pn->lruvec_stat_local) { kfree(pn); return 1; } - pn->lruvec_stat_cpu = alloc_percpu(struct lruvec_stat); + pn->lruvec_stat_cpu = alloc_percpu_gfp(struct lruvec_stat, + GFP_KERNEL_ACCOUNT); if (!pn->lruvec_stat_cpu) { free_percpu(pn->lruvec_stat_local); kfree(pn); @@ -5211,11 +5216,16 @@ static struct mem_cgroup *mem_cgroup_all goto fail; } - memcg->vmstats_local = alloc_percpu(struct memcg_vmstats_percpu); + /* We charge the parent cgroup, never the current task */ + WARN_ON_ONCE(!current->active_memcg); + + memcg->vmstats_local = alloc_percpu_gfp(struct memcg_vmstats_percpu, + GFP_KERNEL_ACCOUNT); if (!memcg->vmstats_local) goto fail; - memcg->vmstats_percpu = alloc_percpu(struct memcg_vmstats_percpu); + memcg->vmstats_percpu = alloc_percpu_gfp(struct memcg_vmstats_percpu, + GFP_KERNEL_ACCOUNT); if (!memcg->vmstats_percpu) goto fail; @@ -5264,7 +5274,9 @@ mem_cgroup_css_alloc(struct cgroup_subsy struct mem_cgroup *memcg; long error = -ENOMEM; + memalloc_use_memcg(parent); memcg = mem_cgroup_alloc(); + memalloc_unuse_memcg(); if (IS_ERR(memcg)) return ERR_CAST(memcg); _
