linux-kernel - Re: [PATCH v4 0/3] Fix deadlock caused by cgroup_mutex and cpu_hotplug

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <Zu3LtU9HUY3XH1WV@slm.duckdns.org>
Date: Fri, 20 Sep 2024 09:23:33 -1000
From: Tejun Heo <tj@...nel.org>
To: Chen Ridong <chenridong@...weicloud.com>
Cc: martin.lau@...ux.dev, ast@...nel.org, daniel@...earbox.net,
	andrii@...nel.org, eddyz87@...il.com, song@...nel.org,
	yonghong.song@...ux.dev, john.fastabend@...il.com,
	kpsingh@...nel.org, sdf@...gle.com, haoluo@...gle.com,
	jolsa@...nel.org, lizefan.x@...edance.com, hannes@...xchg.org,
	roman.gushchin@...ux.dev,
	Michal Koutný <mkoutny@...e.com>,
	bpf@...r.kernel.org, cgroups@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH v4 0/3] Fix deadlock caused by cgroup_mutex and
 cpu_hotplug_lock

On Wed, Sep 18, 2024 at 05:44:47PM +0800, Chen Ridong wrote:
...
> >    cgroup: fix deadlock caused by cgroup_mutex and cpu_hotplug_lock
> >    workqueue: doc: Add a note saturating the system_wq is not permitted
> >    workqueue: Adjust WQ_MAX_ACTIVE from 512 to 2048
> > 
> >   Documentation/core-api/workqueue.rst | 8 ++++++--
> >   include/linux/workqueue.h            | 2 +-
> >   kernel/bpf/cgroup.c                  | 2 +-
> >   kernel/cgroup/cgroup-internal.h      | 1 +
> >   kernel/cgroup/cgroup.c               | 2 +-
> >   5 files changed, 10 insertions(+), 5 deletions(-)
> > 
> Friendly ping.

I don't know why but this series isn't in my inbox for some reason. Here are
some feedbacks after looking at the thread from lore:

- Can you create a separate workqueue for cgrp->bpf.release_work instead of
  exporting cgroup_destroy_wq? Workqueues aren't that expensive. No reason
  to share like this.

- The patch title is rather misleading. The deadlock isn't really caused
  between cgroup_mutex and cpu_hotplug_lock. They're just victims of
  system_wq concurrency depletion. Can you please update accordingly?

- Can you add a new line before the note paragraph? Also, I'd say "deadlock"
  rather than "block" to properly convey the imapct of such saturation
  events. I don't think "eg. cgroup release work" is adding much.

Thanks.

-- 
tejun