Re: [PATCH] net/smc: Add autocork support

["dust.li" <dust.li@linux.alibaba.com>]

On Fri, Feb 18, 2022 at 05:03:56PM +0100, Karsten Graul wrote:
>On 18/02/2022 08:33, dust.li wrote:
>> On Thu, Feb 17, 2022 at 07:15:54PM +0100, Hendrik Brueckner wrote:
>>> On Thu, Feb 17, 2022 at 09:22:00PM +0800, dust.li wrote:
>>>> On Thu, Feb 17, 2022 at 10:37:28AM +0100, Stefan Raspl wrote:
>>>>> On 2/16/22 16:27, dust.li wrote:
>>>>>> On Wed, Feb 16, 2022 at 02:58:32PM +0100, Stefan Raspl wrote:
>>>>>>> On 2/16/22 04:49, Dust Li wrote:
>>>>>>>
>>>>
>>>>> Now we understand that cloud workloads are a bit different, and the desire to
>>>>> be able to modify the environment of a container while leaving the container
>>>>> image unmodified is understandable. But then again, enabling the base image
>>>>> would be the cloud way to address this. The question to us is: How do other
>>>>> parts of the kernel address this?
>>>>
>>>> I'm not familiar with K8S, but from one of my colleague who has worked
>>>> in that area tells me for resources like CPU/MEM and configurations
>>>> like sysctl, can be set using K8S configuration:
>>>> https://kubernetes.io/docs/tasks/administer-cluster/sysctl-cluster/
>>>
>>> For K8s, this involves container engines like cri-o, containerd, podman,
>>> and others towards the runtimes like runc.  To ensure they operate together,
>>> specifications by the Open Container Initiative (OCI) at
>>> https://opencontainers.org/release-notices/overview/
>>>
>>> For container/pod deployments, there is especially the Container Runtime
>>> Interface (CRI) that defines the interface, e.g., of K8s to cri-o etc.
>>>
>>> CRI includes support for (namespaced) sysctl's:
>>> https://github.com/opencontainers/runtime-spec/releases/tag/v1.0.2
>>>
>>> In essence, the CRI spec would allow users to specify/control a specific
>>> runtime for the container in a declarative way w/o modifying the (base)
>>> container images.
>> 
>> Thanks a lot for your kind explanation !
>> 
>> After a quick look at the OCI spec, I saw the support for file based
>> configuration (Including sysfs/procfs etc.). And unfortunately, no
>> netlink support.
>> 
>> 
>> Hi Karsten & Stefan:
>> Back to the patch itself, do you think I need to add the control switch
>> now ? Or just leave the switch and fix other issues first ?
>
>Hi, looks like we need more time to evaluate possibilities, so if you have 
>additional topics on your desk move on and delay this one.

OK, got it.

>Right now for me it looks like there is no way to use netlink for container runtime
>configuration, which is a pity.
>We continue our discussions about this in the team, and also here on the list.

Many thanks for your time on this topic !