memcg: fix swap accounting leak

Fix swapin charge operation of memcg.

Now, memcg has hooks to swap-out operation and checks SwapCache is really
unused or not.  That check depends on contents of struct page.  I.e.  If
PageAnon(page) && page_mapped(page), the page is recoginized as
still-in-use.

Now, reuse_swap_page() calles delete_from_swap_cache() before establishment
of any rmap. Then, in followinig sequence

	(Page fault with WRITE)
	try_charge() (charge += PAGESIZE)
	commit_charge() (Check page_cgroup is used or not..)
	reuse_swap_page()
		-> delete_from_swapcache()
			-> mem_cgroup_uncharge_swapcache() (charge -= PAGESIZE)
	......
New charge is uncharged soon....
To avoid this,  move commit_charge() after page_mapcount() goes up to 1.
By this,

	try_charge()		(usage += PAGESIZE)
	reuse_swap_page()	(may usage -= PAGESIZE if PCG_USED is set)
	commit_charge()		(If page_cgroup is not marked as PCG_USED,
				 add new charge.)
Accounting will be correct.

Changelog (v2) -> (v3)
  - fixed invalid charge to swp_entry==0.
  - updated documentation.
Changelog (v1) -> (v2)
  - fixed comment.

[nishimura@mxp.nes.nec.co.jp: swap accounting leak doc fix]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Tested-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/Documentation/controllers/memcg_test.txt b/Documentation/controllers/memcg_test.txt
index c91f69b..08d4d3e 100644 (file)
--- a/Documentation/controllers/memcg_test.txt
+++ b/Documentation/controllers/memcg_test.txt
@@ -1,6 +1,6 @@
 Memory Resource Controller(Memcg)  Implementation Memo.
-Last Updated: 2008/12/10
-Base Kernel Version: based on 2.6.28-rc7-mm.
+Last Updated: 2008/12/15
+Base Kernel Version: based on 2.6.28-rc8-mm.
 
 Because VM is getting complex (one of reasons is memcg...), memcg's behavior
 is complex. This is a document for memcg's internal behavior.
@@ -111,9 +111,40 @@ Under below explanation, we assume CONFIG_MEM_RES_CTRL_SWAP=y.
        (b) If the SwapCache has been mapped by processes, it has been
            charged already.
 
-       In case (a), we charge it. In case (b), we don't charge it.
-       (But racy state between (a) and (b) exists. We do check it.)
-       At charging, a charge recorded in swap_cgroup is moved to page_cgroup.
+       This swap-in is one of the most complicated work. In do_swap_page(),
+       following events occur when pte is unchanged.
+
+       (1) the page (SwapCache) is looked up.
+       (2) lock_page()
+       (3) try_charge_swapin()
+       (4) reuse_swap_page() (may call delete_swap_cache())
+       (5) commit_charge_swapin()
+       (6) swap_free().
+
+       Considering following situation for example.
+
+       (A) The page has not been charged before (2) and reuse_swap_page()
+           doesn't call delete_from_swap_cache().
+       (B) The page has not been charged before (2) and reuse_swap_page()
+           calls delete_from_swap_cache().
+       (C) The page has been charged before (2) and reuse_swap_page() doesn't
+           call delete_from_swap_cache().
+       (D) The page has been charged before (2) and reuse_swap_page() calls
+           delete_from_swap_cache().
+
+           memory.usage/memsw.usage changes to this page/swp_entry will be
+        Case          (A)      (B)       (C)     (D)
+         Event
+       Before (2)     0/ 1     0/ 1      1/ 1    1/ 1
+          ===========================================
+          (3)        +1/+1    +1/+1     +1/+1   +1/+1
+          (4)          -       0/ 0       -     -1/ 0
+          (5)         0/-1     0/ 0     -1/-1    0/ 0
+          (6)          -       0/-1       -      0/-1
+          ===========================================
+       Result         1/ 1     1/ 1      1/ 1    1/ 1
+
+       In any cases, charges to this page should be 1/ 1.
 
        4.2 Swap-out.
        At swap-out, typical state transition is below.

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index a7ecf23..0ed61e2 100644 (file)
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1169,10 +1169,11 @@ void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr)
        /*
         * Now swap is on-memory. This means this page may be
         * counted both as mem and swap....double count.
-        * Fix it by uncharging from memsw. This SwapCache is stable
-        * because we're still under lock_page().
+        * Fix it by uncharging from memsw. Basically, this SwapCache is stable
+        * under lock_page(). But in do_swap_page()::memory.c, reuse_swap_page()
+        * may call delete_from_swap_cache() before reach here.
         */
-       if (do_swap_account) {
+       if (do_swap_account && PageSwapCache(page)) {
                swp_entry_t ent = {.val = page_private(page)};
                struct mem_cgroup *memcg;
                memcg = swap_cgroup_record(ent, NULL);

diff --git a/mm/memory.c b/mm/memory.c
index e5bfbe6..e009ce8 100644 (file)
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2457,22 +2457,23 @@ static int do_swap_page(struct mm_struct *mm, struct vm_area_struct *vma,
         * while the page is counted on swap but not yet in mapcount i.e.
         * before page_add_anon_rmap() and swap_free(); try_to_free_swap()
         * must be called after the swap_free(), or it will never succeed.
-        * And mem_cgroup_commit_charge_swapin(), which uses the swp_entry
-        * in page->private, must be called before reuse_swap_page(),
-        * which may delete_from_swap_cache().
+        * Because delete_from_swap_page() may be called by reuse_swap_page(),
+        * mem_cgroup_commit_charge_swapin() may not be able to find swp_entry
+        * in page->private. In this case, a record in swap_cgroup  is silently
+        * discarded at swap_free().
         */
 
-       mem_cgroup_commit_charge_swapin(page, ptr);
        inc_mm_counter(mm, anon_rss);
        pte = mk_pte(page, vma->vm_page_prot);
        if (write_access && reuse_swap_page(page)) {
                pte = maybe_mkwrite(pte_mkdirty(pte), vma);
                write_access = 0;
        }
-
        flush_icache_page(vma, page);
        set_pte_at(mm, address, page_table, pte);
        page_add_anon_rmap(page, vma, address);
+       /* It's better to call commit-charge after rmap is established */
+       mem_cgroup_commit_charge_swapin(page, ptr);
 
        swap_free(entry);
        if (vm_swap_full() || (vma->vm_flags & VM_LOCKED) || PageMlocked(page))