linux-kernel - Re: [PATCH mm-unstable] mm: folio_add_new_anon_rmap() careful __folio_set

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <9e8d71f4-69dc-4bf4-a40d-e1b89586f5c9@redhat.com>
Date: Tue, 25 Jun 2024 21:45:59 +0200
From: David Hildenbrand <david@...hat.com>
To: Hugh Dickins <hughd@...gle.com>
Cc: Andrew Morton <akpm@...ux-foundation.org>, Barry Song
 <21cnbao@...il.com>, baolin.wang@...ux.alibaba.com, chrisl@...nel.org,
 linux-kernel@...r.kernel.org, linux-mm@...ck.org, mhocko@...e.com,
 ryan.roberts@....com, shy828301@...il.com, surenb@...gle.com,
 v-songbaohua@...o.com, willy@...radead.org, ying.huang@...el.com,
 yosryahmed@...gle.com, yuanshuai@...o.com, yuzhao@...gle.com
Subject: Re: [PATCH mm-unstable] mm: folio_add_new_anon_rmap() careful
 __folio_set_swapbacked()

On 25.06.24 21:37, Hugh Dickins wrote:
> On Tue, 25 Jun 2024, David Hildenbrand wrote:
>> On 25.06.24 07:00, Hugh Dickins wrote:
>>> Commit "mm: use folio_add_new_anon_rmap() if folio_test_anon(folio)==
>>> false" has extended folio_add_new_anon_rmap() to use on non-exclusive
>>> folios, already visible to others in swap cache and on LRU.
>>>
>>> That renders its non-atomic __folio_set_swapbacked() unsafe: it risks
>>> overwriting concurrent atomic operations on folio->flags, losing bits
>>> added or restoring bits cleared.  Since it's only used in this risky
>>> way when folio_test_locked and !folio_test_anon, many such races are
>>> excluded; but, for example, isolations by folio_test_clear_lru() are
>>> vulnerable, and setting or clearing active.
>>>
>>> It could just use the atomic folio_set_swapbacked(); but this function
>>> does try to avoid atomics where it can, so use a branch instead: just
>>> avoid setting swapbacked when it is already set, that is good enough.
>>> (Swapbacked is normally stable once set: lazyfree can undo it, but
>>> only later, when found anon in a page table.)
>>>
>>> This fixes a lot of instability under compaction and swapping loads:
>>> assorted "Bad page"s, VM_BUG_ON_FOLIO()s, apparently even page double
>>> frees - though I've not worked out what races could lead to the latter.
>>>
>>> Signed-off-by: Hugh Dickins <hughd@...gle.com>
>>> ---
>>>    mm/rmap.c | 4 +++-
>>>    1 file changed, 3 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/mm/rmap.c b/mm/rmap.c
>>> index df1a43295c85..5394c1178bf1 100644
>>> --- a/mm/rmap.c
>>> +++ b/mm/rmap.c
>>> @@ -1408,7 +1408,9 @@ void folio_add_new_anon_rmap(struct folio *folio,
>>> struct vm_area_struct *vma,
>>>     VM_WARN_ON_FOLIO(folio_test_hugetlb(folio), folio);
>>>     VM_BUG_ON_VMA(address < vma->vm_start ||
>>>    			address + (nr << PAGE_SHIFT) > vma->vm_end, vma);
>>> -	__folio_set_swapbacked(folio);
>>> +
>>> +	if (!folio_test_swapbacked(folio))
>>> +		__folio_set_swapbacked(folio);
>>>     __folio_set_anon(folio, vma, address, exclusive);
>>>    
>>>     if (likely(!folio_test_large(folio))) {
>>
>> LGTM.
>>
>> I'll point out that it's sufficient for a PFN walker to do a tryget + trylock
>> to cause trouble.
> 
> That surprises me.  I thought a racer's tryget was irrelevant (touching
> a different field) and its trylock not a problem, since "we" hold the
> folio lock throughout.  If my mental model is too naive there, please
> explain in more detail: we all need to understand this better.

Sorry, I was imprecise.

tryget+trylock should indeed not be a problem, tryget+lock would be, 
because IIRC folio_wait_bit_common()->folio_set_waiters() would be 
messing with folio flags.

-- 
Cheers,

David / dhildenb