Re: [PATCH][BUGFIX] memcg: fix for deadlock betweenlock_page_cgroup and mapping tree_lock

From: Daisuke Nishimura
Date: Tue May 12 2009 - 07:00:26 EST


On Tue, 12 May 2009 17:13:56 +0900
KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:

> On Tue, 12 May 2009 17:00:07 +0900
> Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx> wrote:
> > hmm, I see.
> > cache_charge is outside of tree_lock, so moving uncharge would make sense.
> > IMHO, we should make the period of spinlock as small as possible,
> > and charge/uncharge of pagecache/swapcache is protected by page lock, not tree_lock.
> >
> How about this ?
Looks good conceptually, but it cannot be built :)

It needs a fix like this.
Passed build test with enabling/disabling both CONFIG_MEM_RES_CTLR
and CONFIG_SWAP.

===
include/linux/swap.h | 5 +++++
mm/memcontrol.c | 4 +++-
mm/swap_state.c | 4 +---
mm/vmscan.c | 2 +-
4 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/include/linux/swap.h b/include/linux/swap.h
index caf0767..6ea541d 100644
--- a/include/linux/swap.h
+++ b/include/linux/swap.h
@@ -431,6 +431,11 @@ static inline swp_entry_t get_swap_page(void)
#define has_swap_token(x) 0
#define disable_swap_token() do { } while(0)

+static inline void
+mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent)
+{
+}
+
#endif /* CONFIG_SWAP */
#endif /* __KERNEL__*/
#endif /* _LINUX_SWAP_H */
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 0c9c1ad..89523cf 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1488,8 +1488,9 @@ void mem_cgroup_uncharge_cache_page(struct page *page)
__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
}

+#ifdef CONFIG_SWAP
/*
- * called from __delete_from_swap_cache() and drop "page" account.
+ * called after __delete_from_swap_cache() and drop "page" account.
* memcg information is recorded to swap_cgroup of "ent"
*/
void mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent)
@@ -1506,6 +1507,7 @@ void mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent)
if (memcg)
css_put(&memcg->css);
}
+#endif

#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
/*
diff --git a/mm/swap_state.c b/mm/swap_state.c
index 87f10d4..7624c89 100644
--- a/mm/swap_state.c
+++ b/mm/swap_state.c
@@ -109,8 +109,6 @@ int add_to_swap_cache(struct page *page, swp_entry_t entry, gfp_t gfp_mask)
*/
void __delete_from_swap_cache(struct page *page)
{
- swp_entry_t ent = {.val = page_private(page)};
-
VM_BUG_ON(!PageLocked(page));
VM_BUG_ON(!PageSwapCache(page));
VM_BUG_ON(PageWriteback(page));
@@ -190,7 +188,7 @@ void delete_from_swap_cache(struct page *page)
__delete_from_swap_cache(page);
spin_unlock_irq(&swapper_space.tree_lock);

- mem_cgroup_uncharge_swapcache(page, ent);
+ mem_cgroup_uncharge_swapcache(page, entry);
swap_free(entry);
page_cache_release(page);
}
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 6c5988d..a7d7a06 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -470,7 +470,7 @@ static int __remove_mapping(struct address_space *mapping, struct page *page)
swp_entry_t swap = { .val = page_private(page) };
__delete_from_swap_cache(page);
spin_unlock_irq(&mapping->tree_lock);
- mem_cgroup_uncharge_swapcache(page);
+ mem_cgroup_uncharge_swapcache(page, swap);
swap_free(swap);
} else {
__remove_from_page_cache(page);
===
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/