mm: memcontrol: teach uncharge_list to deal with kmem pages
authorVladimir Davydov <vdavydov@virtuozzo.com>
Tue, 26 Jul 2016 22:24:27 +0000 (15:24 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 26 Jul 2016 23:19:19 +0000 (16:19 -0700)
Page table pages are batched-freed in release_pages on most
architectures.  If we want to charge them to kmemcg (this is what is
done later in this series), we need to teach mem_cgroup_uncharge_list to
handle kmem pages.

Link: http://lkml.kernel.org/r/18d5c09e97f80074ed25b97a7d0f32b95d875717.1464079538.git.vdavydov@virtuozzo.com
Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/memcontrol.c

index 089ef36141555c415062b6ebb8787c6d8774ea8d..ff53e348c4bbf7d0b55f9cd8b42c18375688275f 100644 (file)
@@ -5505,15 +5505,18 @@ void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
 
 static void uncharge_batch(struct mem_cgroup *memcg, unsigned long pgpgout,
                           unsigned long nr_anon, unsigned long nr_file,
-                          unsigned long nr_huge, struct page *dummy_page)
+                          unsigned long nr_huge, unsigned long nr_kmem,
+                          struct page *dummy_page)
 {
-       unsigned long nr_pages = nr_anon + nr_file;
+       unsigned long nr_pages = nr_anon + nr_file + nr_kmem;
        unsigned long flags;
 
        if (!mem_cgroup_is_root(memcg)) {
                page_counter_uncharge(&memcg->memory, nr_pages);
                if (do_memsw_account())
                        page_counter_uncharge(&memcg->memsw, nr_pages);
+               if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && nr_kmem)
+                       page_counter_uncharge(&memcg->kmem, nr_kmem);
                memcg_oom_recover(memcg);
        }
 
@@ -5536,6 +5539,7 @@ static void uncharge_list(struct list_head *page_list)
        unsigned long nr_anon = 0;
        unsigned long nr_file = 0;
        unsigned long nr_huge = 0;
+       unsigned long nr_kmem = 0;
        unsigned long pgpgout = 0;
        struct list_head *next;
        struct page *page;
@@ -5546,8 +5550,6 @@ static void uncharge_list(struct list_head *page_list)
         */
        next = page_list->next;
        do {
-               unsigned int nr_pages = 1;
-
                page = list_entry(next, struct page, lru);
                next = page->lru.next;
 
@@ -5566,31 +5568,35 @@ static void uncharge_list(struct list_head *page_list)
                if (memcg != page->mem_cgroup) {
                        if (memcg) {
                                uncharge_batch(memcg, pgpgout, nr_anon, nr_file,
-                                              nr_huge, page);
-                               pgpgout = nr_anon = nr_file = nr_huge = 0;
+                                              nr_huge, nr_kmem, page);
+                               pgpgout = nr_anon = nr_file =
+                                       nr_huge = nr_kmem = 0;
                        }
                        memcg = page->mem_cgroup;
                }
 
-               if (PageTransHuge(page)) {
-                       nr_pages <<= compound_order(page);
-                       VM_BUG_ON_PAGE(!PageTransHuge(page), page);
-                       nr_huge += nr_pages;
-               }
+               if (!PageKmemcg(page)) {
+                       unsigned int nr_pages = 1;
 
-               if (PageAnon(page))
-                       nr_anon += nr_pages;
-               else
-                       nr_file += nr_pages;
+                       if (PageTransHuge(page)) {
+                               nr_pages <<= compound_order(page);
+                               VM_BUG_ON_PAGE(!PageTransHuge(page), page);
+                               nr_huge += nr_pages;
+                       }
+                       if (PageAnon(page))
+                               nr_anon += nr_pages;
+                       else
+                               nr_file += nr_pages;
+                       pgpgout++;
+               } else
+                       nr_kmem += 1 << compound_order(page);
 
                page->mem_cgroup = NULL;
-
-               pgpgout++;
        } while (next != page_list);
 
        if (memcg)
                uncharge_batch(memcg, pgpgout, nr_anon, nr_file,
-                              nr_huge, page);
+                              nr_huge, nr_kmem, page);
 }
 
 /**