On Wed, Aug 13, 2025 at 5:55 PM Martin KaFai Lau <martin.lau@xxxxxxxxx> wrote: > > On 8/13/25 1:53 PM, Shakeel Butt wrote: > > What I think is the right approach is to have BPF struct ops based > > approach with possible callback 'is this socket under pressure' or maybe > > 'is this socket isolated' and then you can do whatever you want in those > > callbacks. In this way your can follow the same approach of caching the > > result in kernel (lower bits of sk->sk_memcg). > > > > I am CCing bpf list to get some suggestions or concerns on this > > approach. > > I have quickly looked at the set. In patch 11, it sets a bit in sk->sk_memcg. > > On the bpf side, there are already cgroup bpf progs that can do bpf_setsockopt > on a sk, so the same can be done here. The bpf_setsockopt does not have to set > option/knob that is only available in the uapi in case we don't want to expose > this to the user space. > > The cgroup bpf prog (BPF_CGROUP_INET_SOCK_CREATE) can already be run when a > "inet" sock is created. This hook (i.e. attach_type) does not have access to > bpf_setsockopt but should be easy to add. Okay, I will try the bpf_setsockopt() approach. Should I post patch 1-10 to net-next separately ? They are pure net material to gather memcg code under CONFIG_MEMCG. > > For more comprehensive mem charge policy that needs new bpf hook, that probably > will need struct_ops instead of another cgroup attach_type but that will be > implementation details.
