Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Sender: Tejun Heo <htejun@gmail.com>
Date:   Thu, 19 May 2022 22:11:26 -1000
From:   Tejun Heo <tj@kernel.org>
To:     Yosry Ahmed <yosryahmed@google.com>
Cc:     Alexei Starovoitov <ast@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Andrii Nakryiko <andrii@kernel.org>,
        Martin KaFai Lau <kafai@fb.com>,
        Song Liu <songliubraving@fb.com>, Yonghong Song <yhs@fb.com>,
        John Fastabend <john.fastabend@gmail.com>,
        KP Singh <kpsingh@kernel.org>, Hao Luo <haoluo@google.com>,
        Zefan Li <lizefan.x@bytedance.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Shuah Khan <shuah@kernel.org>,
        Roman Gushchin <roman.gushchin@linux.dev>,
        Michal Hocko <mhocko@kernel.org>,
        Stanislav Fomichev <sdf@google.com>,
        David Rientjes <rientjes@google.com>,
        Greg Thelen <gthelen@google.com>,
        Shakeel Butt <shakeelb@google.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Networking <netdev@vger.kernel.org>, bpf <bpf@vger.kernel.org>,
        Cgroups <cgroups@vger.kernel.org>
Subject: Re: [PATCH bpf-next v1 3/5] bpf: Introduce cgroup iter
Message-ID: <YodNLpxut+Zddnre@slm.duckdns.org>
References: <20220520012133.1217211-1-yosryahmed@google.com>
 <20220520012133.1217211-4-yosryahmed@google.com>
 <YodGI73xq8aIBrNM@slm.duckdns.org>
 <CAJD7tkbvMcMWESMcWi6TtdCKLr6keBNGgZTnqcHZvBrPa1qWPw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAJD7tkbvMcMWESMcWi6TtdCKLr6keBNGgZTnqcHZvBrPa1qWPw@mail.gmail.com>
Precedence: bulk

Hello,

On Fri, May 20, 2022 at 12:58:52AM -0700, Yosry Ahmed wrote:
> On Fri, May 20, 2022 at 12:41 AM Tejun Heo <tj@kernel.org> wrote:
> >
> > On Fri, May 20, 2022 at 01:21:31AM +0000, Yosry Ahmed wrote:
> > > From: Hao Luo <haoluo@google.com>
> > >
> > > Introduce a new type of iter prog: cgroup. Unlike other bpf_iter, this
> > > iter doesn't iterate a set of kernel objects. Instead, it is supposed to
> > > be parameterized by a cgroup id and prints only that cgroup. So one
> > > needs to specify a target cgroup id when attaching this iter. The target
> > > cgroup's state can be read out via a link of this iter.
> > >
> > > Signed-off-by: Hao Luo <haoluo@google.com>
> > > Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
> >
> > This could be me not understanding why it's structured this way but it keeps
> > bothering me that this is adding a cgroup iterator which doesn't iterate
> > cgroups. If all that's needed is extracting information from a specific
> > cgroup, why does this need to be an iterator? e.g. why can't I use
> > BPF_PROG_TEST_RUN which looks up the cgroup with the provided ID, flushes
> > rstat, retrieves whatever information necessary and returns that as the
> > result?
> 
> I will let Hao and Yonghong reply here as they have a lot more
> context, and they had previous discussions about cgroup_iter. I just
> want to say that exposing the stats in a file is extremely convenient
> for userspace apps. It becomes very similar to reading stats from
> cgroupfs. It also makes migrating cgroup stats that we have
> implemented in the kernel to BPF a lot easier.

So, if it were upto me, I'd rather direct energy towards making retrieving
information through TEST_RUN_PROG easier rather than clinging to making
kernel output text. I get that text interface is familiar but it kinda
sucks in many ways.

> AFAIK there are also discussions about using overlayfs to have links
> to the bpffs files in cgroupfs, which makes it even better. So I would
> really prefer keeping the approach we have here of reading stats
> through a file from userspace. As for how we go about this (and why a
> cgroup iterator doesn't iterate cgroups) I will leave this for Hao and
> Yonghong to explain the rationale behind it. Ideally we can keep the
> same functionality under a more descriptive name/type.

My answer would be the same here. You guys seem dead set on making the
kernel emulate cgroup1. I'm not gonna explicitly block that but would
strongly suggest having a longer term view.

If you *must* do the iterator, can you at least make it a proper iterator
which supports seeking? AFAICS there's nothing fundamentally preventing bpf
iterators from supporting seeking. Or is it that you need something which is
pinned to a cgroup so that you can emulate the directory structure?

Thanks.

-- 
tejun