[RFC v5 1/6] gpu: rfc: Proposal for a GPU cgroup controller

[T.J. Mercier]

On Thu, Apr 21, 2022 at 1:34 PM Tejun Heo <tj at kernel.org> wrote:
>
> Hello,
>
> On Wed, Apr 20, 2022 at 11:52:19PM +0000, T.J. Mercier wrote:
> > From: Hridya Valsaraju <hridya at google.com>
> >
> > This patch adds a proposal for a new GPU cgroup controller for
> > accounting/limiting GPU and GPU-related memory allocations.
> > The proposed controller is based on the DRM cgroup controller[1] and
> > follows the design of the RDMA cgroup controller.
> >
> > The new cgroup controller would:
> > * Allow setting per-device limits on the total size of buffers
> >   allocated by device within a cgroup.
> > * Expose a per-device/allocator breakdown of the buffers charged to a
> >   cgroup.
> >
> > The prototype in the following patches is only for memory accounting
> > using the GPU cgroup controller and does not implement limit setting.
> >
> > [1]: https://lore.kernel.org/amd-gfx/20210126214626.16260-1-brian.welty@intel.com/
> >
> > Signed-off-by: Hridya Valsaraju <hridya at google.com>
> > Signed-off-by: T.J. Mercier <tjmercier at google.com>
>
> Looks straight-forward enough from cgroup side. Are gpu folks generally
> happy? David, Daniel, Kenny, what are your thoughts?
>
> >  Documentation/gpu/rfc/gpu-cgroup.rst | 190 +++++++++++++++++++++++++++
>
> Can you fold the important part into cgroup-v2.rst and maybe make the rest
> code comments if necessary?
>
Sure thing, thanks.

> Thanks.
>
> --
> tejun