memcg: remove a redundant check
KAMEZAWA Hiroyuki [Fri, 25 Jul 2008 08:47:17 +0000 (01:47 -0700)]
Because of remove refcnt patch, it's very rare case to that
mem_cgroup_charge_common() is called against a page which is accounted.

mem_cgroup_charge_common() is called when.
 1. a page is added into file cache.
 2. an anon page is _newly_ mapped.

A racy case is that a newly-swapped-in anonymous page is referred from
prural threads in do_swap_page() at the same time.
(a page is not Locked when mem_cgroup_charge() is called from do_swap_page.)

Another case is shmem. It charges its page before calling add_to_page_cache().
Then, mem_cgroup_charge_cache() is called twice. This case is handled in
mem_cgroup_cache_charge(). But this check may be too hacky...

Signed-off-by : KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcontrol.c

index 04ded27..5b3759b 100644 (file)
@@ -536,28 +536,6 @@ static int mem_cgroup_charge_common(struct page *page, struct mm_struct *mm,
        if (mem_cgroup_subsys.disabled)
                return 0;
 
-       /*
-        * Should page_cgroup's go to their own slab?
-        * One could optimize the performance of the charging routine
-        * by saving a bit in the page_flags and using it as a lock
-        * to see if the cgroup page already has a page_cgroup associated
-        * with it
-        */
-retry:
-       lock_page_cgroup(page);
-       pc = page_get_page_cgroup(page);
-       /*
-        * The page_cgroup exists and
-        * the page has already been accounted.
-        */
-       if (unlikely(pc)) {
-               VM_BUG_ON(pc->page != page);
-               VM_BUG_ON(!pc->mem_cgroup);
-               unlock_page_cgroup(page);
-               goto done;
-       }
-       unlock_page_cgroup(page);
-
        pc = kmem_cache_alloc(page_cgroup_cache, gfp_mask);
        if (unlikely(pc == NULL))
                goto err;
@@ -618,15 +596,10 @@ retry:
        lock_page_cgroup(page);
        if (unlikely(page_get_page_cgroup(page))) {
                unlock_page_cgroup(page);
-               /*
-                * Another charge has been added to this page already.
-                * We take lock_page_cgroup(page) again and read
-                * page->cgroup, increment refcnt.... just retry is OK.
-                */
                res_counter_uncharge(&mem->res, PAGE_SIZE);
                css_put(&mem->css);
                kmem_cache_free(page_cgroup_cache, pc);
-               goto retry;
+               goto done;
        }
        page_assign_page_cgroup(page, pc);
 
@@ -665,8 +638,32 @@ int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask)
 int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
                                gfp_t gfp_mask)
 {
+       /*
+        * Corner case handling. This is called from add_to_page_cache()
+        * in usual. But some FS (shmem) precharges this page before calling it
+        * and call add_to_page_cache() with GFP_NOWAIT.
+        *
+        * For GFP_NOWAIT case, the page may be pre-charged before calling
+        * add_to_page_cache(). (See shmem.c) check it here and avoid to call
+        * charge twice. (It works but has to pay a bit larger cost.)
+        */
+       if (!(gfp_mask & __GFP_WAIT)) {
+               struct page_cgroup *pc;
+
+               lock_page_cgroup(page);
+               pc = page_get_page_cgroup(page);
+               if (pc) {
+                       VM_BUG_ON(pc->page != page);
+                       VM_BUG_ON(!pc->mem_cgroup);
+                       unlock_page_cgroup(page);
+                       return 0;
+               }
+               unlock_page_cgroup(page);
+       }
+
        if (unlikely(!mm))
                mm = &init_mm;
+
        return mem_cgroup_charge_common(page, mm, gfp_mask,
                                MEM_CGROUP_CHARGE_TYPE_CACHE, NULL);
 }