linux-kernel - Re: [PATCH -next] cpuset: add cpuset1_online

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] [day] [month] [year] [list]

Message-ID: <503ec3db-b205-444a-9c92-08eaac14f015@huaweicloud.com>
Date: Wed, 17 Dec 2025 10:18:48 +0800
From: Chen Ridong <chenridong@...weicloud.com>
To: Waiman Long <llong@...hat.com>, Michal Koutný
 <mkoutny@...e.com>
Cc: tj@...nel.org, hannes@...xchg.org, cgroups@...r.kernel.org,
 linux-kernel@...r.kernel.org, lujialin4@...wei.com
Subject: Re: [PATCH -next] cpuset: add cpuset1_online_css helper for
 v1-specific operations



On 2025/12/17 10:03, Waiman Long wrote:
> 
> On 12/16/25 7:53 PM, Chen Ridong wrote:
>>
>> On 2025/12/16 22:58, Waiman Long wrote:
>>> On 12/16/25 9:03 AM, Michal Koutný wrote:
>>>> On Tue, Dec 16, 2025 at 08:13:53PM +0800, Chen Ridong <chenridong@...weicloud.com> wrote:
>>>>> Regarding the lock assertions: cpuset_mutex is defined in cpuset.c and is not visible in
>>>>> cpuset-v1.c. Given that cpuset v1 is deprecated, would you prefer that we add a helper to assert
>>>>> cpuset_mutex is locked? Is that worth?
>>>> It could be un-static'd and defined in cpuset-internal.h. (Hopefully, we
>>>> should not end up with random callers of the helper but it's IMO worth
>>>> it for docs and greater safety.)
>>> I would suggest defining a "assert_cpuset_lock_held(void)" helper function and put the declaration
>>> in include/linux/cpuset.h together with cpuset_lock/unlock() to complete the full set. This will
>>> allow other kernel subsystems to acquire the cpuset_mutex and assert that the mutex was held.
>> Considering potential use by other subsystems, this is worthwhile. I will add the helper.
>>
>>>>> Should we guard with !cpuset_v2() or !is_in_v2mode()?
>>>>>
>>>>> In cgroup v1, if the cpuset is operating in v2 mode, are these flags still valid?
>>>> I have no experience with this transitional option so that made me look
>>>> at the docs and there we specify it only affects behaviors of CPU masks,
>>>> not the extra flags. So I wanted to suggest !cpuset_v2(), correct?
>>> The "cpuset_v2_mode" mount flag is used for making the behavior of cpuset.{cpus,mems}.effective in
>>> v1 behave like in v2. It has no effect on other v1 specific control files. So cpuset1_online_css()
>>> should only be called if "!cpuset_v2()".
>>>
>> If cpuset1_online_css() is only called under the condition !cpuset_v2(), a compile-time option might
>> suffice? When CONFIG_CPUSETS_V1=n, cpuset1_online_css could be defined as an empty inline function.
> 
> cpuset_v2() includes "!IS_ENABLED(CONFIG_CPUSETS_V1)", so the compiler should compile out the call
> to cpuset1_online_css() if CONFIG_CPUSETS_V1 isn't defined. If you want to make cpuset1_online_css()
> conditional on CONFIG_CPUSETS_V1, I am fine with that too.
> 

The cpuset1_online_css() resides in cpuset-v1.c and is only compiled when CONFIG_CPUSETS_V1=y. For
cases where CONFIG_CPUSETS_V1=n, I have defined it as an empty inline function.

-- 
Best regards,
Ridong