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]
Date:   Sun, 12 Nov 2023 21:28:21 -0500
From:   Gregory Price <>
To:     "Huang, Ying" <>
Cc:     Gregory Price <>,,,,,,,,,,,,,,
Subject: Re: [RFC PATCH v4 0/3] memcg weighted interleave mempolicy control

On Mon, Nov 13, 2023 at 09:31:07AM +0800, Huang, Ying wrote:
> Gregory Price <> writes:
> > On Fri, Nov 10, 2023 at 02:16:05PM +0800, Huang, Ying wrote:
> >> Gregory Price <> writes:
> >> 
> >
> > weights are also inherited from parent cgroups, so if you set them in
> > parent slices you can automatically set update system settings.
> >
> > by default the parent slice weights will always be 1 until set
> > otherwise.  Once they're set, children inherit naturally.
> >
> > Maybe there's an argument here for including interleave_weights in the
> > root cgroup.
> Even if the interleave_weights is introduced in root cgroup, the initial
> default weight need to be 1 to be back-compatible with the original

Sorry, I am maybe not explaining correctly.  Right now, the weights
are not *replicated* when a child cgroup is created.  Instead, when
weights are requested (during allocation) the requestor searches for the
first cgroup in its family that has weights.

while (!memcg->weights && !memcg_is_root(memcg))
    memcg = parent(memcg)

We only create new weights on each child if the child explicitly has
their weights set.  We manage everything else via RCU to keep it all

This would allow a set of weights in the root cgroup to be set and then
immediately inherited by the entire system, though it does add the
overhead of searching the cgroup family lineage on allocations (which
could be non-trivial, we are still testing it).

> If we don't reuse MPOL_INTERLEAVE, but use a new memory policy mode (say
> MPOL_WEIGHTED_INTERLEAVE).  The default values of the interleave weight
> in root cgroup needn't to be 1.

I agree, and I already have patches that do just this.  Though based on
other feedback, it's looking like I'll be reverting back to implementing
all of this in mempolicy, and maybe trying to pull mempolicy forward
into the procfs world.


Powered by blists - more mailing lists