| 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, 5 Nov 2020 12:52:05 +0800
From: Alex Shi <alex.shi@...ux.alibaba.com>
To: Johannes Weiner <hannes@...xchg.org>
Cc: Matthew Wilcox <willy@...radead.org>, akpm@...ux-foundation.org,
mgorman@...hsingularity.net, tj@...nel.org, hughd@...gle.com,
khlebnikov@...dex-team.ru, daniel.m.jordan@...cle.com,
lkp@...el.com, linux-mm@...ck.org, linux-kernel@...r.kernel.org,
cgroups@...r.kernel.org, shakeelb@...gle.com,
iamjoonsoo.kim@....com, richard.weiyang@...il.com,
kirill@...temov.name, alexander.duyck@...il.com,
rong.a.chen@...el.com, mhocko@...e.com, vdavydov.dev@...il.com,
shy828301@...il.com, Vlastimil Babka <vbabka@...e.cz>,
Minchan Kim <minchan@...nel.org>
Subject: Re: [PATCH v20 08/20] mm: page_idle_get_page() does not need lru_lock
在 2020/11/5 上午1:46, Johannes Weiner 写道:
> On Wed, Nov 04, 2020 at 07:27:21PM +0800, Alex Shi wrote:
>> 在 2020/11/3 上午4:20, Johannes Weiner 写道:
>>> On Mon, Nov 02, 2020 at 02:49:27PM +0000, Matthew Wilcox wrote:
>>>> On Mon, Nov 02, 2020 at 09:41:10AM -0500, Johannes Weiner wrote:
>>>>> On Thu, Oct 29, 2020 at 06:44:53PM +0800, Alex Shi wrote:
>>>>>> From: Hugh Dickins <hughd@...gle.com>
>>>>>>
>>>>>> It is necessary for page_idle_get_page() to recheck PageLRU() after
>>>>>> get_page_unless_zero(), but holding lru_lock around that serves no
>>>>>> useful purpose, and adds to lru_lock contention: delete it.
>>>>>>
>>>>>> See https://lore.kernel.org/lkml/20150504031722.GA2768@blaptop for the
>>>>>> discussion that led to lru_lock there; but __page_set_anon_rmap() now
>>>>>> uses WRITE_ONCE(),
>>>>>
>>>>> That doesn't seem to be the case in Linus's or Andrew's tree. Am I
>>>>> missing a dependent patch series?
>>>>>
>>>>>> and I see no other risk in page_idle_clear_pte_refs() using
>>>>>> rmap_walk() (beyond the risk of racing PageAnon->PageKsm, mostly but
>>>>>> not entirely prevented by page_count() check in ksm.c's
>>>>>> write_protect_page(): that risk being shared with page_referenced()
>>>>>> and not helped by lru_lock).
>>>>>
>>>>> Isn't it possible, as per Minchan's description, for page->mapping to
>>>>> point to a struct anon_vma without PAGE_MAPPING_ANON set, and rmap
>>>>> thinking it's looking at a struct address_space?
>>>>
>>>> I don't think it can point to an anon_vma without the ANON bit set.
>>>> Minchan's concern in that email was that it might still be NULL.
>>>
>>> Hm, no, the thread is a lengthy discussion about whether the store
>>> could be split such that page->mapping is actually pointing to
>>> something invalid (anon_vma without the PageAnon bit).
>>>
>>> From his email:
>>>
>>> CPU 0 CPU 1
>>>
>>> do_anonymous_page
>>> __page_set_anon_rmap
>>> /* out of order happened so SetPageLRU is done ahead */
>>> SetPageLRU(page)
>>
>> This SetPageLRU done in __pagevec_lru_add_fn() which under the lru_lock
>> protection, so the original memory barrier or lock concern isn't
>> correct. that means, the SetPageLRU isn't possible to be here.
>> And then no warry on right side 'CPU 1' problem.
>
> The SetPageLRU is done under lru_lock, but the store to page->mapping
> is not, so what ensures ordering between them? And what prevents the
> compiler from tearing the store to page->mapping?
>
Right, I misunderstand the spin_lock in memory barrier. Thanks a lot
for point out this.
So, is this patch fine to fix the problem?
>From 8427121da195a6a386d1ce71abcb41b31211e21f Mon Sep 17 00:00:00 2001
From: Alex Shi <alex.shi@...ux.alibaba.com>
Date: Thu, 5 Nov 2020 11:38:24 +0800
Subject: [PATCH] mm/rmap: stop store reordering issue on page->mapping
Hugh Dickins and Minchan Kim observed a long time issue which
discussed here, but actully the mentioned fix missed.
https://lore.kernel.org/lkml/20150504031722.GA2768@blaptop/
The store reordering may cause problem in the scenario:
CPU 0 CPU1
do_anonymous_page
page_add_new_anon_rmap()
page->mapping = anon_vma + PAGE_MAPPING_ANON
lru_cache_add_inactive_or_unevictable()
spin_lock(lruvec->lock)
SetPageLRU()
spin_unlock(lruvec->lock)
/* idletacking judged it as LRU
* page so pass the page in
* page_idle_clear_pte_refs
*/
page_idle_clear_pte_refs
rmap_walk
if PageAnon(page)
Johannes give detailed examples how the store reordering could cause
a trouble:
The concern is the SetPageLRU may get reorder before 'page->mapping'
setting, That would make CPU 1 will observe at page->mapping after
observing PageLRU set on the page.
1. anon_vma + PAGE_MAPPING_ANON
That's the in-order scenario and is fine.
2. NULL
That's possible if the page->mapping store gets reordered to occur
after SetPageLRU. That's fine too because we check for it.
3. anon_vma without the PAGE_MAPPING_ANON bit
That would be a problem and could lead to all kinds of undesirable
behavior including crashes and data corruption.
Is it possible? AFAICT the compiler is allowed to tear the store to
page->mapping and I don't see anything that would prevent it.
That said, I also don't see how the reader testing PageLRU under the
lru_lock would prevent that in the first place. AFAICT we need that
WRITE_ONCE() around the page->mapping assignment.
Signed-off-by: Alex Shi <alex.shi@...ux.alibaba.com>
Cc: Johannes Weiner <hannes@...xchg.org>
Cc: Andrew Morton <akpm@...ux-foundation.org>
Cc: Hugh Dickins <hughd@...gle.com>
Cc: Matthew Wilcox <willy@...radead.org>
Cc: Minchan Kim <minchan@...nel.org>
Cc: Vladimir Davydov <vdavydov.dev@...il.com>
Cc: linux-kernel@...r.kernel.org
Cc: linux-mm@...ck.org
---
mm/rmap.c | 21 ++++++++++++++++++++-
1 file changed, 20 insertions(+), 1 deletion(-)
diff --git a/mm/rmap.c b/mm/rmap.c
index c050dab2ae65..56af18aa57de 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1054,8 +1054,27 @@ static void __page_set_anon_rmap(struct page *page,
if (!exclusive)
anon_vma = anon_vma->root;
+ /*
+ * w/o the WRITE_ONCE here the following scenario may happens due to
+ * store reordering.
+ *
+ * CPU 0 CPU 1
+ *
+ * do_anonymous_page page_idle_clear_pte_refs
+ * __page_set_anon_rmap
+ * page->mapping = anon_vma + PAGE_MAPPING_ANON
+ * lru_cache_add_inactive_or_unevictable()
+ * SetPageLRU(page)
+ * rmap_walk
+ * if PageAnon(page)
+ *
+ * The 'SetPageLRU' may reordered before page->mapping setting, and
+ * page->mapping may set with anon_vma, w/o anon bit, then rmap_walk
+ * may goes to rmap_walk_file() for a anon page.
+ */
+
anon_vma = (void *) anon_vma + PAGE_MAPPING_ANON;
- page->mapping = (struct address_space *) anon_vma;
+ WRITE_ONCE(page->mapping, (struct address_space *) anon_vma);
page->index = linear_page_index(vma, address);
}
--
1.8.3.1
> The writer does this:
>
> CPU 0
> page_add_new_anon_rmap()
> page->mapping = anon_vma + PAGE_MAPPING_ANON
> lru_cache_add_inactive_or_unevictable()
> spin_lock(lruvec->lock)
> SetPageLRU()
> spin_unlock(lruvec->lock)
>
> The concern is what CPU 1 will observe at page->mapping after
> observing PageLRU set on the page.
>
> 1. anon_vma + PAGE_MAPPING_ANON
>
> That's the in-order scenario and is fine.
>
> 2. NULL
>
> That's possible if the page->mapping store gets reordered to occur
> after SetPageLRU. That's fine too because we check for it.
>
> 3. anon_vma without the PAGE_MAPPING_ANON bit
>
> That would be a problem and could lead to all kinds of undesirable
> behavior including crashes and data corruption.
>
> Is it possible? AFAICT the compiler is allowed to tear the store to
> page->mapping and I don't see anything that would prevent it.
>
> That said, I also don't see how the reader testing PageLRU under the
> lru_lock would prevent that in the first place. AFAICT we need that
> WRITE_ONCE() around the page->mapping assignment that's referenced in
> the changelog of this patch but I cannot find in any tree or patch.
>
Powered by blists - more mailing lists