| Message ID | 20201110031015.15715-1-songmuchun@bytedance.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v2] mm: memcg/slab: Fix root memcg vmstats | expand |
On Mon, Nov 9, 2020 at 7:10 PM Muchun Song <songmuchun@bytedance.com> wrote: > > If we reparent the slab objects to the root memcg, when we free > the slab object, we need to update the per-memcg vmstats to keep > it correct for the root memcg. Now this at least affects the vmstat > of NR_KERNEL_STACK_KB for !CONFIG_VMAP_STACK when the thread stack > size is smaller than the PAGE_SIZE. > > Fixes: ec9f02384f60 ("mm: workingset: fix vmstat counters for shadow nodes") > Signed-off-by: Muchun Song <songmuchun@bytedance.com> > Acked-by: Roman Gushchin <guro@fb.com> Reviewed-by: Shakeel Butt <shakeelb@google.com>
On Tue, Nov 10, 2020 at 11:10:15AM +0800, Muchun Song wrote: > If we reparent the slab objects to the root memcg, when we free > the slab object, we need to update the per-memcg vmstats to keep > it correct for the root memcg. Now this at least affects the vmstat > of NR_KERNEL_STACK_KB for !CONFIG_VMAP_STACK when the thread stack > size is smaller than the PAGE_SIZE. > > Fixes: ec9f02384f60 ("mm: workingset: fix vmstat counters for shadow nodes") > Signed-off-by: Muchun Song <songmuchun@bytedance.com> > Acked-by: Roman Gushchin <guro@fb.com> Acked-by: Johannes Weiner <hannes@cmpxchg.org>
On Tue, 10 Nov 2020, Muchun Song wrote: > If we reparent the slab objects to the root memcg, when we free > the slab object, we need to update the per-memcg vmstats to keep > it correct for the root memcg. Now this at least affects the vmstat > of NR_KERNEL_STACK_KB for !CONFIG_VMAP_STACK when the thread stack > size is smaller than the PAGE_SIZE. > > Fixes: ec9f02384f60 ("mm: workingset: fix vmstat counters for shadow nodes") > Signed-off-by: Muchun Song <songmuchun@bytedance.com> > Acked-by: Roman Gushchin <guro@fb.com> Acked-by: David Rientjes <rientjes@google.com> I assume that without this fix that the root memcg's vmstat would always be inflated if we reparented? If that's accurate, perhaps this is deserving of a Cc: stable@vger.kernel.org # 5.3+
diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 2dde734df7d1..bbd40c5af61e 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -875,8 +875,13 @@ void __mod_lruvec_kmem_state(void *p, enum node_stat_item idx, int val) rcu_read_lock(); memcg = mem_cgroup_from_obj(p); - /* Untracked pages have no memcg, no lruvec. Update only the node */ - if (!memcg || memcg == root_mem_cgroup) { + /* + * Untracked pages have no memcg, no lruvec. Update only the + * node. If we reparent the slab objects to the root memcg, + * when we free the slab object, we need to update the per-memcg + * vmstats to keep it correct for the root memcg. + */ + if (!memcg) { __mod_node_page_state(pgdat, idx, val); } else { lruvec = mem_cgroup_lruvec(memcg, pgdat);
If we reparent the slab objects to the root memcg, when we free the slab object, we need to update the per-memcg vmstats to keep it correct for the root memcg. Now this at least affects the vmstat of NR_KERNEL_STACK_KB for !CONFIG_VMAP_STACK when the thread stack size is smaller than the PAGE_SIZE. Fixes: ec9f02384f60 ("mm: workingset: fix vmstat counters for shadow nodes") Signed-off-by: Muchun Song <songmuchun@bytedance.com> Acked-by: Roman Gushchin <guro@fb.com> --- v1->v2: - Just resend the patch. mm/memcontrol.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)