linux-kernel - Re: [PATCH V2] lockdep: fix deadlock issue between lockdep and rcu

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] [day] [month] [year] [list]

Date: Fri, 2 Feb 2024 16:19:34 +0800
From: Zhiguo Niu <niuzhiguo84@...il.com>
To: Bart Van Assche <bvanassche@....org>
Cc: Boqun Feng <boqun.feng@...il.com>, Zhiguo Niu <zhiguo.niu@...soc.com>, peterz@...radead.org, 
	mingo@...hat.com, will@...nel.org, longman@...hat.com, 
	linux-kernel@...r.kernel.org, ke.wang@...soc.com, xuewen.yan@...soc.com, 
	Carlos Llamas <cmllamas@...gle.com>
Subject: Re: [PATCH V2] lockdep: fix deadlock issue between lockdep and rcu

Dear All,
Thanks for your discussion and suggestions , I update and send patch
version3 according to your comments.
thanks!

On Fri, Feb 2, 2024 at 10:13 AM Bart Van Assche <bvanassche@....org> wrote:
>
> On 2/1/24 13:53, Boqun Feng wrote:
> > Have you checked the change in the patch? Now call_rcu_zapped() has been
> > splitted into two parts: preparing the callback and calling call_rcu(),
> > the preparing part checks and sets the delayed_free.scheduled under
> > graph_lock(), so only one CPU/thread will win and do the actual
> > call_rcu(). And the RCU callback free_zapped_rcu() will unset
> > delayed_free.scheduled, again under graph_lock().
> >
> > If you think it's still possible, could you provide a case where the
> > race may happen?
>
> Yes, I noticed that call_rcu_zapped() has been split but the first time
> I took a look at this patch I wasn't sure that the new code is correct.
> After having taken a second look, the new mechanism for deciding whether
> or not to invoke call_rcu() seems fine to me.
>
> Bart.