From: Yafang Shao <laoar.shao@gmail.com>
To: "Michal Koutný" <mkoutny@suse.com>
Cc: ast@kernel.org, daniel@iogearbox.net, john.fastabend@gmail.com,
andrii@kernel.org, martin.lau@linux.dev, song@kernel.org,
yonghong.song@linux.dev, kpsingh@kernel.org, sdf@google.com,
haoluo@google.com, jolsa@kernel.org, tj@kernel.org,
lizefan.x@bytedance.com, hannes@cmpxchg.org,
yosryahmed@google.com, sinquersw@gmail.com,
cgroups@vger.kernel.org, bpf@vger.kernel.org
Subject: Re: [RFC PATCH bpf-next 1/8] cgroup: Don't have to hold cgroup_mutex in task_cgroup_from_root()
Date: Tue, 10 Oct 2023 11:58:14 +0800 [thread overview]
Message-ID: <CALOAHbC4_0990_HD4=mg8gfU51juk8fs07zYrr6VL9fPOuLOng@mail.gmail.com> (raw)
In-Reply-To: <sdw6rnzbvmktajcxb4svj2kzvttftae2i5nd2lnlxnm3llub37@2q2rlubjzb5a>
On Mon, Oct 9, 2023 at 10:45 PM Michal Koutný <mkoutny@suse.com> wrote:
>
> On Sat, Oct 07, 2023 at 02:02:57PM +0000, Yafang Shao <laoar.shao@gmail.com> wrote:
> > The task cannot modify cgroups if we have already acquired the
> > css_set_lock, thus eliminating the need to hold the cgroup_mutex. Following
> > this change, task_cgroup_from_root() can be employed in non-sleepable contexts.
>
> IIRC, cset_cgroup_from_root() needs cgroup_mutex to make sure the `root`
> doesn't disappear under hands (synchronization with
> cgroup_destroy_root().
current_cgns_cgroup_from_root() doesn't hold the cgroup_mutext as
well. Could this potentially lead to issues, such as triggering the
BUG_ON() in __cset_cgroup_from_root(), if the root has already been
destroyed?
> However, as I look at it now, cgroup_mutex won't synchronize against
> cgroup_kill_sb(), it may worked by accident(?) nowadays (i.e. callers
> pinned the root implicitly in another way).
>
> Still, it'd be good to have an annotation that ensures, the root is around
> when using it. (RCU read lock might be fine but you'd need
> cgroup_get_live() if passing it out of the RCU read section.)
>
> Basically, the code must be made safe against cgroup v1 unmounts.
What we aim to protect against is changes to the `root_list`, which
occur exclusively during cgroup setup and destroy paths. Would it be
beneficial to introduce a dedicated root_list_lock specifically for
this purpose? This approach could potentially reduce the need for the
broader cgroup_mutex in other scenarios.
--
Regards
Yafang
next prev parent reply other threads:[~2023-10-10 3:58 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-10-07 14:02 [RFC PATCH bpf-next 0/8] bpf, cgroup: Add BPF support for cgroup1 hierarchy Yafang Shao
2023-10-07 14:02 ` [RFC PATCH bpf-next 1/8] cgroup: Don't have to hold cgroup_mutex in task_cgroup_from_root() Yafang Shao
2023-10-07 15:50 ` Tejun Heo
2023-10-08 2:32 ` Yafang Shao
2023-10-09 14:45 ` Michal Koutný
2023-10-10 3:58 ` Yafang Shao [this message]
2023-10-10 8:29 ` Michal Koutný
2023-10-10 11:58 ` Yafang Shao
2023-10-07 14:02 ` [RFC PATCH bpf-next 2/8] cgroup: Add new helpers for cgroup1 hierarchy Yafang Shao
2023-10-07 15:55 ` Tejun Heo
2023-10-08 2:36 ` Yafang Shao
2023-10-09 11:32 ` Michal Koutný
2023-10-09 13:10 ` Yafang Shao
2023-10-09 14:48 ` Michal Koutný
2023-10-10 3:59 ` Yafang Shao
2023-10-07 14:02 ` [RFC PATCH bpf-next 3/8] bpf: Add kfuncs " Yafang Shao
2023-10-07 15:57 ` Tejun Heo
2023-10-08 2:37 ` Yafang Shao
2023-10-07 14:03 ` [RFC PATCH bpf-next 4/8] selftests/bpf: Fix issues in setup_classid_environment() Yafang Shao
2023-10-07 14:03 ` [RFC PATCH bpf-next 5/8] selftests/bpf: Add parallel support for classid Yafang Shao
2023-10-07 14:03 ` [RFC PATCH bpf-next 6/8] selftests/bpf: Add a new cgroup helper get_classid_cgroup_id() Yafang Shao
2023-10-07 14:03 ` [RFC PATCH bpf-next 7/8] selftests/bpf: Add a new cgroup helper get_cgroup_hierarchy_id() Yafang Shao
2023-10-07 14:03 ` [RFC PATCH bpf-next 8/8] selftests/bpf: Add selftests for cgroup1 hierarchy Yafang Shao
2023-10-09 11:46 ` [RFC PATCH bpf-next 0/8] bpf, cgroup: Add BPF support " Michal Koutný
2023-10-09 13:11 ` Yafang Shao
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CALOAHbC4_0990_HD4=mg8gfU51juk8fs07zYrr6VL9fPOuLOng@mail.gmail.com' \
--to=laoar.shao@gmail.com \
--cc=andrii@kernel.org \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=cgroups@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=hannes@cmpxchg.org \
--cc=haoluo@google.com \
--cc=john.fastabend@gmail.com \
--cc=jolsa@kernel.org \
--cc=kpsingh@kernel.org \
--cc=lizefan.x@bytedance.com \
--cc=martin.lau@linux.dev \
--cc=mkoutny@suse.com \
--cc=sdf@google.com \
--cc=sinquersw@gmail.com \
--cc=song@kernel.org \
--cc=tj@kernel.org \
--cc=yonghong.song@linux.dev \
--cc=yosryahmed@google.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).