Re: [PATCH 4/5] sock, cgroup: add sock->sk_cgroup

[Daniel Borkmann <daniel@iogearbox.net>]

Hi Tejun,

On 11/17/2015 08:40 PM, Tejun Heo wrote:
...
> While it is possible to solve these issues from controller side by
> implementing hierarchical allowable ranges in both controllers, it
> would involve quite a bit of complexity in the controllers and further
> obfuscate network configuration as it becomes even more difficult to
> tell what's actually being configured looking from the network side.
> While not much can be done for v1 at this point, as membership
> handling is sane on cgroup v2, it'd be better to make cgroup matching
> behave like other network matches and classifiers than introducing
> further complications.
>
> In preparation, this patch adds sock->sk_cgroup which points to the
> associated cgroup.  A sock is associated on creation and stays
> associated to the same cgroup until freed; unfortunately, this ends up
> adding another cgroup field to struct sock on top of sk_cgrp_prioidx
> and sk_classid.  I tried to think of a way to somehow overload the
> existing fields but couldn't come up with a reasonable one.  For the
> longer term, the fields can be rearranged so that disabling prio and
> cls controllers reduce the size of the struct.

Do you see a way forward where the new behavior could be enabled f.e.
as an extra mount option (that long-term would be made default, while
deprecating the current behavior) on net_cls et al? There are various
more users at least on the net_cls side (nft and tc as well). Would be
really great, if sk_cgroup could abstract that somehow away for all of
them w/o adding a second version to all users.

Best,
Daniel