linux-kernel - Re: [PATCH v2 3/6] rcu: limit PREEMPT

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: <20241114082534.8Go_kdjY@linutronix.de>
Date: Thu, 14 Nov 2024 09:25:34 +0100
From: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
To: Ankur Arora <ankur.a.arora@...cle.com>
Cc: Frederic Weisbecker <frederic@...nel.org>, linux-kernel@...r.kernel.org,
	peterz@...radead.org, tglx@...utronix.de, paulmck@...nel.org,
	mingo@...nel.org, juri.lelli@...hat.com, vincent.guittot@...aro.org,
	dietmar.eggemann@....com, rostedt@...dmis.org, bsegall@...gle.com,
	mgorman@...e.de, vschneid@...hat.com, efault@....de,
	sshegde@...ux.ibm.com, boris.ostrovsky@...cle.com
Subject: Re: [PATCH v2 3/6] rcu: limit PREEMPT_RCU configurations

On 2024-11-13 16:23:03 [-0800], Ankur Arora wrote:
> > But looking at !CONFIG_PREEMPT_RCU code on tree_plugin.h, I see
> > some issues now that the code can be preemptible. Well I think
> > it has always been preemptible but PREEMPTION && !PREEMPT_RCU
> > has seldom been exerciced (or was it even possible?).
> >
> > For example rcu_read_unlock_strict() can be called with preemption
> > enabled so we need the following otherwise the rdp is unstable, the
> > norm value becomes racy
> 
> I think I see your point about rdp being racy, but given that
> rcu_read_unlock_strict() would always be called with a non-zero
> preemption count (with CONFIG_PREEMPTION), wouldn't the preempt_count()
> check defeat any calls to rcu_read_unlock_strict()?
> 
>     void rcu_read_unlock_strict(void)
>     {
>             struct rcu_data *rdp;
> 
>             if (irqs_disabled() || preempt_count() || !rcu_state.gp_kthread)
>                     return;
> 
> Or am I missing something?

This is indeed broken. By moving preempt_disable() as Frederic suggested
then rcu_read_unlock_strict() becomes a NOP. Keeping this as-is results
in spats due to this_cpu_ptr() in preemptible regions. Looking further
we have "rdp->cpu != smp_processor_id()" as the next candidate.

That preempt_disable() should go to rcu_read_unlock_strict() after the
check.

> Ankur

Sebastian