linux-kernel - Re: [PATCH v5 7/8] mm/shmem, swap: rework swap entry and index calculation for large swapin

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: <d12fbf3f-3389-429a-b636-849ce611f5a3@linux.alibaba.com>
Date: Mon, 14 Jul 2025 10:39:14 +0800
From: Baolin Wang <baolin.wang@...ux.alibaba.com>
To: Kairui Song <kasong@...cent.com>, linux-mm@...ck.org
Cc: Andrew Morton <akpm@...ux-foundation.org>, Hugh Dickins
 <hughd@...gle.com>, Matthew Wilcox <willy@...radead.org>,
 Kemeng Shi <shikemeng@...weicloud.com>, Chris Li <chrisl@...nel.org>,
 Nhat Pham <nphamcs@...il.com>, Baoquan He <bhe@...hat.com>,
 Barry Song <baohua@...nel.org>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v5 7/8] mm/shmem, swap: rework swap entry and index
 calculation for large swapin



On 2025/7/11 14:36, Baolin Wang wrote:
> 
> 
> On 2025/7/10 11:37, Kairui Song wrote:
>> From: Kairui Song <kasong@...cent.com>
>>
>> Instead of calculating the swap entry differently in different swapin
>> paths, calculate it early before the swap cache lookup and use that
>> for the lookup and later swapin. And after swapin have brought a folio,
>> simply round it down against the size of the folio.
>>
>> This is simple and effective enough to verify the swap value. A folio's
>> swap entry is always aligned by its size. Any kind of parallel split or
>> race is acceptable because the final shmem_add_to_page_cache ensures
>> that all entries covered by the folio are correct, and thus there
>> will be no data corruption.
>>
>> This also prevents false positive cache lookup. If a shmem read
>> request's index points to the middle of a large swap entry,
>> previously, shmem will try the swap cache lookup using the large swap
>> entry's starting value (which is the first sub swap entry of this
>> large entry). This will lead to false positive lookup results if only
>> the first few swap entries are cached but the actual requested swap
>> entry pointed by the index is uncached. This is not a rare event,
>> as swap readahead always tries to cache order 0 folios when possible.
>>
>> And this shouldn't cause any increased repeated faults. Instead, no
>> matter how the shmem mapping is split in parallel, as long as the
>> mapping still contains the right entries, the swapin will succeed.
>>
>> The final object size and stack usage are also reduced due to
>> simplified code:
>>
>> ./scripts/bloat-o-meter mm/shmem.o.old mm/shmem.o
>> add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-233 (-233)
>> Function                                     old     new   delta
>> shmem_swapin_folio                          4040    3807    -233
>> Total: Before=33152, After=32919, chg -0.70%
>>
>> Stack usage (Before vs After):
>> mm/shmem.c:2277:12:shmem_swapin_folio   264     static
>> mm/shmem.c:2277:12:shmem_swapin_folio   256     static
>>
>> And while at it, round down the index too if swap entry is round down.
>> The index is used either for folio reallocation or confirming the
>> mapping content. In either case, it should be aligned with the swap
>> folio.
>>
>> Signed-off-by: Kairui Song <kasong@...cent.com>
> 
> Overall, I don't see any obvious issues, but please give me some time to 
> run some tests and then get back to you. Thanks.

LGTM.
Reviewed-by: Baolin Wang <baolin.wang@...ux.alibaba.com>
Tested-by: Baolin Wang <baolin.wang@...ux.alibaba.com>