memcg: remove redundant function calls
KAMEZAWA Hiroyuki [Tue, 29 Apr 2008 08:00:22 +0000 (01:00 -0700)]
remove_list/add_list uses page_cgroup_zoneinfo() in it.

So, it's called twice before and after lock.

mz = page_cgroup_zoneinfo();
lock();
mz = page_cgroup_zoneinfo();
....
unlock();

And address of mz never changes.

This is not good. This patch fixes this behavior.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcontrol.c

index f891876..395fd8e 100644 (file)
@@ -275,10 +275,10 @@ static void unlock_page_cgroup(struct page *page)
        bit_spin_unlock(PAGE_CGROUP_LOCK_BIT, &page->page_cgroup);
 }
 
-static void __mem_cgroup_remove_list(struct page_cgroup *pc)
+static void __mem_cgroup_remove_list(struct mem_cgroup_per_zone *mz,
+                       struct page_cgroup *pc)
 {
        int from = pc->flags & PAGE_CGROUP_FLAG_ACTIVE;
-       struct mem_cgroup_per_zone *mz = page_cgroup_zoneinfo(pc);
 
        if (from)
                MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_ACTIVE) -= 1;
@@ -289,10 +289,10 @@ static void __mem_cgroup_remove_list(struct page_cgroup *pc)
        list_del_init(&pc->lru);
 }
 
-static void __mem_cgroup_add_list(struct page_cgroup *pc)
+static void __mem_cgroup_add_list(struct mem_cgroup_per_zone *mz,
+                               struct page_cgroup *pc)
 {
        int to = pc->flags & PAGE_CGROUP_FLAG_ACTIVE;
-       struct mem_cgroup_per_zone *mz = page_cgroup_zoneinfo(pc);
 
        if (!to) {
                MEM_CGROUP_ZSTAT(mz, MEM_CGROUP_ZSTAT_INACTIVE) += 1;
@@ -618,7 +618,7 @@ retry:
 
        mz = page_cgroup_zoneinfo(pc);
        spin_lock_irqsave(&mz->lru_lock, flags);
-       __mem_cgroup_add_list(pc);
+       __mem_cgroup_add_list(mz, pc);
        spin_unlock_irqrestore(&mz->lru_lock, flags);
 
        unlock_page_cgroup(page);
@@ -674,7 +674,7 @@ void mem_cgroup_uncharge_page(struct page *page)
        if (--(pc->ref_cnt) == 0) {
                mz = page_cgroup_zoneinfo(pc);
                spin_lock_irqsave(&mz->lru_lock, flags);
-               __mem_cgroup_remove_list(pc);
+               __mem_cgroup_remove_list(mz, pc);
                spin_unlock_irqrestore(&mz->lru_lock, flags);
 
                page_assign_page_cgroup(page, NULL);
@@ -736,7 +736,7 @@ void mem_cgroup_page_migration(struct page *page, struct page *newpage)
 
        mz = page_cgroup_zoneinfo(pc);
        spin_lock_irqsave(&mz->lru_lock, flags);
-       __mem_cgroup_remove_list(pc);
+       __mem_cgroup_remove_list(mz, pc);
        spin_unlock_irqrestore(&mz->lru_lock, flags);
 
        page_assign_page_cgroup(page, NULL);
@@ -748,7 +748,7 @@ void mem_cgroup_page_migration(struct page *page, struct page *newpage)
 
        mz = page_cgroup_zoneinfo(pc);
        spin_lock_irqsave(&mz->lru_lock, flags);
-       __mem_cgroup_add_list(pc);
+       __mem_cgroup_add_list(mz, pc);
        spin_unlock_irqrestore(&mz->lru_lock, flags);
 
        unlock_page_cgroup(newpage);