linux-kernel - Re: [PATCH] mm: fix the inaccurate memory statistics issue for users

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: <72f0dc8c-def3-447c-b54e-c390705f8c26@linux.alibaba.com>
Date: Tue, 3 Jun 2025 16:08:21 +0800
From: Baolin Wang <baolin.wang@...ux.alibaba.com>
To: Michal Hocko <mhocko@...e.com>, Andrew Morton <akpm@...ux-foundation.org>
Cc: david@...hat.com, shakeel.butt@...ux.dev, lorenzo.stoakes@...cle.com,
 Liam.Howlett@...cle.com, vbabka@...e.cz, rppt@...nel.org, surenb@...gle.com,
 donettom@...ux.ibm.com, aboorvad@...ux.ibm.com, sj@...nel.org,
 linux-mm@...ck.org, linux-fsdevel@...r.kernel.org,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH] mm: fix the inaccurate memory statistics issue for users



On 2025/5/30 21:39, Michal Hocko wrote:
> On Thu 29-05-25 20:53:13, Andrew Morton wrote:
>> On Sat, 24 May 2025 09:59:53 +0800 Baolin Wang <baolin.wang@...ux.alibaba.com> wrote:
>>
>>> On some large machines with a high number of CPUs running a 64K pagesize
>>> kernel, we found that the 'RES' field is always 0 displayed by the top
>>> command for some processes, which will cause a lot of confusion for users.
>>>
>>>      PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
>>>   875525 root      20   0   12480      0      0 R   0.3   0.0   0:00.08 top
>>>        1 root      20   0  172800      0      0 S   0.0   0.0   0:04.52 systemd
>>>
>>> The main reason is that the batch size of the percpu counter is quite large
>>> on these machines, caching a significant percpu value, since converting mm's
>>> rss stats into percpu_counter by commit f1a7941243c1 ("mm: convert mm's rss
>>> stats into percpu_counter"). Intuitively, the batch number should be optimized,
>>> but on some paths, performance may take precedence over statistical accuracy.
>>> Therefore, introducing a new interface to add the percpu statistical count
>>> and display it to users, which can remove the confusion. In addition, this
>>> change is not expected to be on a performance-critical path, so the modification
>>> should be acceptable.
>>>
>>> Fixes: f1a7941243c1 ("mm: convert mm's rss stats into percpu_counter")
>>
>> Three years ago.
>>
>>> Tested-by Donet Tom <donettom@...ux.ibm.com>
>>> Reviewed-by: Aboorva Devarajan <aboorvad@...ux.ibm.com>
>>> Tested-by: Aboorva Devarajan <aboorvad@...ux.ibm.com>
>>> Acked-by: Shakeel Butt <shakeel.butt@...ux.dev>
>>> Acked-by: SeongJae Park <sj@...nel.org>
>>> Signed-off-by: Baolin Wang <baolin.wang@...ux.alibaba.com>
>>
>> Thanks, I added cc:stable to this.
> 
> I have only noticed this new posting now. I do not think this is a
> stable material. I am also not convinced that the impact of the pcp lock
> exposure to the userspace has been properly analyzed and documented in
> the changelog. I am not nacking the patch (yet) but I would like to see
> a serious analyses that this has been properly thought through.

Good point. I did a quick measurement on my 32 cores Arm machine. I ran 
two workloads, one is the 'top' command: top -d 1 (updating every 
second). Another workload is kernel building (time make -j32).

 From the following data, I did not see any significant impact of the 
patch changes on the execution of the kernel building workload.

w/o patch:
real	4m33.887s
user	118m24.153s
sys	9m51.402s

w/ patch:
real	4m34.495s
user	118m21.739s
sys	9m39.232s