| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Thu, 7 May 2009 15:59:50 +0900
From: Daisuke Nishimura <nishimura@....nes.nec.co.jp>
To: balbir@...ux.vnet.ibm.com
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"linux-mm@...ck.org" <linux-mm@...ck.org>,
"akpm@...ux-foundation.org" <akpm@...ux-foundation.org>,
"hugh@...itas.com" <hugh@...itas.com>,
d-nishimura@....biglobe.ne.jp,
Daisuke Nishimura <nishimura@....nes.nec.co.jp>
Subject: Re: [PATCH] memcg: fix stale swap cache leak v5
On Mon, 4 May 2009 22:08:06 +0530, Balbir Singh <balbir@...ux.vnet.ibm.com> wrote:
> * Daisuke Nishimura <d-nishimura@....biglobe.ne.jp> [2009-05-01 13:33:17]:
>
> > processA | processB
> > -------------------------------------+-------------------------------------
> > (page_remove_rmap()) | (shrink_page_list())
> > mem_cgroup_uncharge_page() |
> > ->uncharged because it's not |
> > PageSwapCache yet. |
> > So, both mem/memsw.usage |
> > are decremented. |
> > | add_to_swap() -> added to swap cache.
> >
> > If this page goes thorough without being freed for some reason, this page
> > doesn't goes back to memcg's LRU because of !PageCgroupUsed.
>
> For some reason could use some clarification.
>
If swap_writepage() is called(via pageout()), try_to_free_swap() in swap_writepage()
will free this unused swap cache.
But there are many checks before swap_writepage() is called.
For example, if page_referenced() returned true, "referenced" will be set,
so we hit the check:
If (PageDirty(page)) {
if (sc->order <= PAGE_ALLOC_COSTLY_ORDER && referenced)
goto keep_locked;
I think there are other cases.
> >
> > These swap cache cannot be freed in memcg's LRU scanning, and swp_entry cannot
> > be freed properly as a result.
> > This patch adds a hook after add_to_swap() to check the page is mapped by a
> > process or not, and frees it if it has been unmapped already.
> >
> > If a page has been on swap cache already when the owner process calls
> > page_remove_rmap() -> mem_cgroup_uncharge_page(), the page is not uncharged.
> > It goes back to memcg's LRU even if it goes through shrink_page_list()
> > without being freed, so this patch ignores these case.
> >
> > Signed-off-by: Daisuke Nishimura <nishimura@....nes.nec.co.jp>
> > ---
> > include/linux/swap.h | 12 ++++++++++++
> > mm/memcontrol.c | 14 ++++++++++++++
> > mm/vmscan.c | 8 ++++++++
> > 3 files changed, 34 insertions(+), 0 deletions(-)
> >
> > diff --git a/include/linux/swap.h b/include/linux/swap.h
> > index caf0767..8e75d7a 100644
> > --- a/include/linux/swap.h
> > +++ b/include/linux/swap.h
> > @@ -336,11 +336,17 @@ static inline void disable_swap_token(void)
> >
> > #ifdef CONFIG_CGROUP_MEM_RES_CTLR
> > extern void mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent);
> > +extern int memcg_free_unused_swapcache(struct page *page);
> > #else
> > static inline void
> > mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent)
> > {
> > }
> > +static inline int
> > +memcg_free_unused_swapcache(struct page *page)
> > +{
> > + return 0;
> > +}
> > #endif
> > #ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
> > extern void mem_cgroup_uncharge_swap(swp_entry_t ent);
> > @@ -431,6 +437,12 @@ static inline swp_entry_t get_swap_page(void)
> > #define has_swap_token(x) 0
> > #define disable_swap_token() do { } while(0)
> >
> > +static inline int
> > +memcg_free_unused_swapcache(struct page *page)
> > +{
> > + return 0;
> > +}
> > +
> > #endif /* CONFIG_SWAP */
> > #endif /* __KERNEL__*/
> > #endif /* _LINUX_SWAP_H */
> > diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> > index 01c2d8f..4f7e5b6 100644
> > --- a/mm/memcontrol.c
> > +++ b/mm/memcontrol.c
> > @@ -1488,6 +1488,7 @@ 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.
> > * memcg information is recorded to swap_cgroup of "ent"
> > @@ -1507,6 +1508,19 @@ void mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent)
> > css_put(&memcg->css);
> > }
> >
> > +int memcg_free_unused_swapcache(struct page *page)
> > +{
> > + VM_BUG_ON(!PageLocked(page));
> > + VM_BUG_ON(!PageSwapCache(page));
> > +
> > + if (mem_cgroup_disabled())
> > + return 0;
> > + if (!PageAnon(page) || page_mapped(page))
> > + return 0;
>
> Do we need these checks? Isn't PageSwapCache() check and
> page_swapcount() check enough in try_to_free_swap()?
>
I think this is needed.
page_swapcount() will return true, because this page has just been added
to swap cache(try_to_unmap() has not been called yet).
> > + return try_to_free_swap(page); /* checks page_swapcount */
>
> try_to_free_swap() marks the page as dirty, do you know why?
>
I'm not sure.
Other callers of delete_from_swap_cache() also set the page dirty.
> > +}
> > +#endif /* CONFIG_SWAP */
> > +
> > #ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
> > /*
> > * called from swap_entry_free(). remove record in swap_cgroup and
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index eac9577..c1a7a6f 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -656,6 +656,14 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> > goto keep_locked;
> > if (!add_to_swap(page))
> > goto activate_locked;
> > + /*
> > + * The owner process might have uncharged the page
> > + * (by page_remove_rmap()) before it has been added
> > + * to swap cache.
> > + * Check it here to avoid making it stale.
> > + */
> > + if (memcg_free_unused_swapcache(page))
> > + goto keep_locked;
>
> Seems reasonable, but I think it is better to check for
> scan_global_lru().. no?
>
hmm, we cannot ensure that additional global LRU scanning will happen after this,
so I think it would be better to free this swap cache if possible
instead of making it stale.
Thanks,
Daisuke Nishimura.
> > may_enter_fs = 1;
> > }
> >
> >
>
> --
> Balbir
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists