Re: [PATCH bpf-next v1 3/5] bpf: Introduce cgroup iter

[Yonghong Song]

On 5/20/22 2:49 PM, Hao Luo wrote:
> Hi Tejun and Yonghong,
>
> On Fri, May 20, 2022 at 12:42 PM Hao Luo <haoluo@xxxxxxxxxx> wrote:
> > Hi Tejun and Yonghong,
> >
> > On Fri, May 20, 2022 at 9:45 AM Tejun Heo <tj@xxxxxxxxxx> wrote:
> > > On Fri, May 20, 2022 at 09:29:43AM -0700, Yonghong Song wrote:
> > > >     <various stats interested by the user>
> > > >
> > > > This way, user space can easily construct the cgroup hierarchy stat like
> > > >                             cpu   mem   cpu pressure   mem pressure ...
> > > >     cgroup1                 ...
> > > >        child1               ...
> > > >          grandchild1        ...
> > > >        child2               ...
> > > >     cgroup 2                ...
> > > >        child 3              ...
> > > >          ...                ...
> > > >
> > > > the bpf iterator can have additional parameter like
> > > > cgroup_id = ... to only call bpf program once with that
> > > > cgroup_id if specified.
> >
> > Yep, this should work. We just need to make the cgroup_id parameter
> > optional. If it is specified when creating bpf_iter_link, we print for
> > that cgroup only. If it is not specified, we iterate over all cgroups.
> > If I understand correctly, sounds doable.
>
> Yonghong, I realized that seek() which Tejun has been calling out, can
> be used to specify the target cgroup, rather than adding a new
> parameter. Maybe, we can pass cgroup_id to seek() on cgroup bpf_iter,
> which will instruct read() to return the corresponding cgroup's stats.
> On the other hand, reading without calling seek() beforehand will
> return all the cgroups.

Currently, seek is not supported for bpf_iter.

const struct file_operations bpf_iter_fops = {
        .open           = iter_open,
        .llseek         = no_llseek,
        .read           = bpf_seq_read,
        .release        = iter_release,
};

But if seek() works, I don't mind to remove this restriction.
But not sure what to seek. Do you mean to provide a cgroup_fd/cgroup_id
as the seek() syscall parameter? This may work.

But considering we have parameterized example (map_fd) and
in the future, we may have other parameterized bpf_iter
(e.g., for one task). Maybe parameter-based approach is better.

> WDYT?