lkml.org 
[lkml]   [2011]   [Mar]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
Subject[PATCH v2] memcg: fix leak on wrong LRU with FUSE

fs/fuse/dev.c::fuse_try_move_page() does

(1) remove a page by ->steal()
(2) re-add the page to page cache
(3) link the page to LRU if it was not on LRU at (1)

This implies the page is _on_ LRU when it's added to radix-tree.
So, the page is added to memory cgroup while it's on LRU.
because LRU is lazy and no one flushs it.

This is the same behavior as SwapCache and needs special care as
- remove page from LRU before overwrite pc->mem_cgroup.
- add page to LRU after overwrite pc->mem_cgroup.

And we need to taking care of pagevec.

If PageLRU(page) is set before we add PCG_USED bit, the page
will not be added to memcg's LRU (in short period).
So, regardlress of PageLRU(page) value before commit_charge(),
we need to check PageLRU(page) after commit_charge().

Changelog:
- clean up.
- cover !PageLRU() by pagevec case.


Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
---
mm/memcontrol.c | 53 ++++++++++++++++++++++++++++++++++-------------------
1 file changed, 34 insertions(+), 19 deletions(-)
Index: mmotm-0303/mm/memcontrol.c
===================================================================
--- mmotm-0303.orig/mm/memcontrol.c
+++ mmotm-0303/mm/memcontrol.c
@@ -926,13 +926,12 @@ void mem_cgroup_add_lru_list(struct page
}

/*
- * At handling SwapCache, pc->mem_cgroup may be changed while it's linked to
- * lru because the page may.be reused after it's fully uncharged (because of
- * SwapCache behavior).To handle that, unlink page_cgroup from LRU when charge
- * it again. This function is only used to charge SwapCache. It's done under
- * lock_page and expected that zone->lru_lock is never held.
+ * At handling SwapCache and other FUSE stuff, pc->mem_cgroup may be changed
+ * while it's linked to lru because the page may be reused after it's fully
+ * uncharged. To handle that, unlink page_cgroup from LRU when charge it again.
+ * It's done under lock_page and expected that zone->lru_lock isnever held.
*/
-static void mem_cgroup_lru_del_before_commit_swapcache(struct page *page)
+static void mem_cgroup_lru_del_before_commit(struct page *page)
{
unsigned long flags;
struct zone *zone = page_zone(page);
@@ -948,7 +947,7 @@ static void mem_cgroup_lru_del_before_co
spin_unlock_irqrestore(&zone->lru_lock, flags);
}

-static void mem_cgroup_lru_add_after_commit_swapcache(struct page *page)
+static void mem_cgroup_lru_add_after_commit(struct page *page)
{
unsigned long flags;
struct zone *zone = page_zone(page);
@@ -2431,9 +2430,28 @@ static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
enum charge_type ctype);

+static void
+__mem_cgroup_commit_charge_lrucare(struct page *page, struct mem_cgroup *mem,
+ enum charge_type ctype)
+{
+ struct page_cgroup *pc = lookup_page_cgroup(page);
+ /*
+ * In some case, SwapCache, FUSE(splice_buf->radixtree), the page
+ * is already on LRU. It means the page may on some other page_cgroup's
+ * LRU. Take care of it.
+ */
+ if (unlikely(PageLRU(page)))
+ mem_cgroup_lru_del_before_commit(page);
+ __mem_cgroup_commit_charge(mem, page, 1, pc, ctype);
+ if (unlikely(PageLRU(page)))
+ mem_cgroup_lru_add_after_commit(page);
+ return;
+}
+
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
gfp_t gfp_mask)
{
+ struct mem_cgroup *mem = NULL;
int ret;

if (mem_cgroup_disabled())
@@ -2468,14 +2486,15 @@ int mem_cgroup_cache_charge(struct page
if (unlikely(!mm))
mm = &init_mm;

- if (page_is_file_cache(page))
- return mem_cgroup_charge_common(page, mm, gfp_mask,
- MEM_CGROUP_CHARGE_TYPE_CACHE);
-
+ if (page_is_file_cache(page)) {
+ ret = __mem_cgroup_try_charge(mm, gfp_mask, 1, &mem, true);
+ if (ret || !mem)
+ return ret;
+ __mem_cgroup_commit_charge_lrucare(page, mem,
+ MEM_CGROUP_CHARGE_TYPE_CACHE);
+ }
/* shmem */
if (PageSwapCache(page)) {
- struct mem_cgroup *mem;
-
ret = mem_cgroup_try_charge_swapin(mm, page, gfp_mask, &mem);
if (!ret)
__mem_cgroup_commit_charge_swapin(page, mem,
@@ -2532,17 +2551,13 @@ static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
enum charge_type ctype)
{
- struct page_cgroup *pc;
-
if (mem_cgroup_disabled())
return;
if (!ptr)
return;
cgroup_exclude_rmdir(&ptr->css);
- pc = lookup_page_cgroup(page);
- mem_cgroup_lru_del_before_commit_swapcache(page);
- __mem_cgroup_commit_charge(ptr, page, 1, pc, ctype);
- mem_cgroup_lru_add_after_commit_swapcache(page);
+
+ __mem_cgroup_commit_charge_lrucare(page, ptr, ctype);
/*
* Now swap is on-memory. This means this page may be
* counted both as mem and swap....double count.


\
 
 \ /
  Last update: 2011-03-08 06:05    [from the cache]
©2003-2011 Jasper Spaans