Received-SPF: pass (google.com: domain of linux-kernel+bounces-48798-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Feedback-ID: iad51458e:Fastmail
Date: Thu, 1 Feb 2024 11:58:13 -0800
From: Boqun Feng <boqun.feng@gmail.com>
To: Bart Van Assche <bvanassche@acm.org>
Cc: Zhiguo Niu <zhiguo.niu@unisoc.com>, peterz@infradead.org,
	mingo@redhat.com, will@kernel.org, longman@redhat.com,
	linux-kernel@vger.kernel.org, niuzhiguo84@gmail.com,
	ke.wang@unisoc.com, xuewen.yan@unisoc.com,
	Carlos Llamas <cmllamas@google.com>
Subject: Re: [PATCH V2] lockdep: fix deadlock issue between lockdep and rcu
Message-ID: <Zbv31aq6aZuSXpY1@boqun-archlinux>
References: <1705477714-10467-1-git-send-email-zhiguo.niu@unisoc.com>
 <2d1f4f16-6d66-4449-a7b7-5e2b2a4eaeb3@acm.org>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <2d1f4f16-6d66-4449-a7b7-5e2b2a4eaeb3@acm.org>

On Thu, Feb 01, 2024 at 09:22:20AM -0800, Bart Van Assche wrote:
> On 1/16/24 23:48, Zhiguo Niu wrote:
> > There is a deadlock scenario between lockdep and rcu when
> > rcu nocb feature is enabled, just as following call stack:
> 
> Is it necessary to support lockdep for this kernel configuration or should we
> rather forbid this combination by changing lib/Kconfig.debug?
> 

RCU nocb is a quite common configuration for RCU, so I think lockdep
should support this.

> >   /*
> > - * Schedule an RCU callback if no RCU callback is pending. Must be called with
> > - * the graph lock held.
> > - */
> > -static void call_rcu_zapped(struct pending_free *pf)
> > +* See if we need to queue an RCU callback, must called with
> > +* the lockdep lock held, returns false if either we don't have
> > +* any pending free or the callback is already scheduled.
> > +* Otherwise, a call_rcu() must follow this function call.
> > +*/
> 
> Why has the name been changed from "graph lock" into "lockdep lock"? I think
> elsewhere in this source file it is called the "graph lock".
> 
> >   	/*
> > -	 * If there's anything on the open list, close and start a new callback.
> > -	 */
> > -	call_rcu_zapped(delayed_free.pf + delayed_free.index);
> > +	* If there's anything on the open list, close and start a new callback.
> > +	*/
> > +	if (need_callback)
> > +		call_rcu(&delayed_free.rcu_head, free_zapped_rcu);
> 
> The comment above the if-statement refers to the call_rcu_zapped() function
> while call_rcu_zapped() has been changed into call_rcu(). So the comment is
> now incorrect.
> 
> Additionally, what guarantees that the above code won't be triggered
> concurrently from two different threads? As you may know calling call_rcu()
> twice before the callback has been started is not allowed. I think that can
> happen with the above code.
> 

No, it's synchronized by the delayed_free.schedule. Only one thread/CPU
can schedule at a time. Or am I missing something subtle?

Regards,
Boqun

> Bart.