Re: [Phishing Risk] [External] Re: [PATCH] cgroup/cpuset: Add a new isolated mems.policy type.

[Zhongkun He]

Hi Tejun, thanks for your reply.

We usually use numactl to set the memory policy, but it cannot be 
changed dynamically. In addition, the mempolicy of cpuset can provide a 
more convenient interface for management and control panel.

Sorry,I don't quite understand the meaning of "don't enforce anything 
resource related". Does it mean mempolicy, such as "prefer:2" must 
specify node? Or "cpuset.mems.policy" need to specify a default value?
(cpuset.mems.policy does not require a default value.)

Thanks.


> Hello,
>
> On Fri, Sep 02, 2022 at 02:33:03PM +0800, hezhongkun wrote:
> > From: Zhongkun He <hezhongkun.hzk@xxxxxxxxxxxxx>
> >
> > Mempolicy is difficult to use because it is set in-process
> > via a system call. We want to make it easier to use mempolicy
> > in cpuset, and  we can control low-priority cgroups to
> > allocate memory in specified nodes. So this patch want to
> > adds the mempolicy interface in cpuset.
> >
> > The mempolicy priority of cpuset is lower than the task.
> > The order of getting the policy is:
> > 	1) vma mempolicy
> > 	2) task->mempolicy
> > 	3) cpuset->mempolicy
> > 	4) default policy.
> >
> > cpuset's policy is owned by itself, but descendants will
> > get the default mempolicy from parent.
> >
> > How to use the mempolicy interface:
> > 	echo prefer:2 > /sys/fs/cgroup/zz/cpuset.mems.policy
> > 	echo bind:1-3 > /sys/fs/cgroup/zz/cpuset.mems.policy
> >          echo interleave:0,1,2,3 >/sys/fs/cgroup/zz/cpuset.mems.policy
> > Show the policy:
> > 	cat /sys/fs/cgroup/zz/cpuset.mems.policy
> > 		prefer:2
> > 	cat /sys/fs/cgroup/zz/cpuset.mems.policy
> > 		bind:1-3
> > 	cat /sys/fs/cgroup/zz/cpuset.mems.policy
> > 		interleave:0-3
> > Clear the policy:
> > 	echo default > /sys/fs/cgroup/zz/cpuset.mems.policy
>
> So, I'm a fan of adding cgroup functionalities which don't enforce anything
> resource related. What you're proposing can easily be achieved with userland
> tooling, right?
>
> Thanks.