Re: [PATCH bpf-next v9 1/5] bpf: Introduce cgroup iter

From: "Michal Koutný" <mkoutny@suse.com>
To: Hao Luo <haoluo@google.com>
Cc: linux-kernel@vger.kernel.org, bpf@vger.kernel.org,
	cgroups@vger.kernel.org, netdev@vger.kernel.org,
	Alexei Starovoitov <ast@kernel.org>,
	Andrii Nakryiko <andrii@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Martin KaFai Lau <martin.lau@linux.dev>,
	Song Liu <song@kernel.org>, Yonghong Song <yhs@fb.com>,
	Tejun Heo <tj@kernel.org>, Zefan Li <lizefan.x@bytedance.com>,
	KP Singh <kpsingh@kernel.org>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Michal Hocko <mhocko@kernel.org>,
	John Fastabend <john.fastabend@gmail.com>,
	Jiri Olsa <jolsa@kernel.org>,
	Roman Gushchin <roman.gushchin@linux.dev>,
	David Rientjes <rientjes@google.com>,
	Stanislav Fomichev <sdf@google.com>,
	Shakeel Butt <shakeelb@google.com>,
	Yosry Ahmed <yosryahmed@google.com>
Subject: Re: [PATCH bpf-next v9 1/5] bpf: Introduce cgroup iter
Date: Fri, 26 Aug 2022 19:15:29 +0200	[thread overview]
Message-ID: <20220826171529.GB30475@blackbody.suse.cz> (raw)
In-Reply-To: <CA+khW7hKk8yMvsQCQjnEoR3=G9=77F2TgAEDa+uSVedoOE=NsA@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 1066 bytes --]

On Thu, Aug 25, 2022 at 10:58:26AM -0700, Hao Luo <haoluo@google.com> wrote:
> Permission is a valid point about FD. There was discussion in an
> earlier version of this patch series [0].

(I'm sorry, I didn't follow all the version discussions closely.)

I think the permissions are a non-issue when unprivileged BPF is
disabled. If it's allowed, I think it'd be better solved generally
within the BPF iterator framework. (Maybe it's already present, I didn't
check.)

(OT:
> The good thing about ID is that it can be passed across processes 

FDs can be passed too (parent-child trivially, others via SCM_RIGHTS
message).

> and it's meaningful to appear in logs. It's more user-friendly.

I'd say cgroup path wins both in meaning and user friendliness.
(Or maybe you meant different class of users.)
)

> So we decided to support both.

I accept cgroup ids are an establish{ing,ed} way to refer to cgroups
from userspace. Hence my fixups for the BPF cgroup iter (another thread)
for better namespacing consisntency.

Thanks,
Michal

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 228 bytes --]