linux-kernel - Re: [PATCH] intel_rapl: Correct hotplug correction

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]

Date:	Thu, 22 May 2014 12:08:20 +0200
From:	Borislav Petkov <bp@...en8.de>
To:	"Srivatsa S. Bhat" <srivatsa.bhat@...ux.vnet.ibm.com>
Cc:	Srinivas Pandruvada <srinivas.pandruvada@...ux.intel.com>,
	Jacob Pan <jacob.jun.pan@...ux.intel.com>,
	LKML <linux-kernel@...r.kernel.org>,
	Borislav Petkov <bp@...e.de>, Ingo Molnar <mingo@...nel.org>,
	"Rafael J. Wysocki" <rafael.j.wysocki@...el.com>,
	Peter Zijlstra <peterz@...radead.org>,
	Thomas Gleixner <tglx@...utronix.de>
Subject: Re: [PATCH] intel_rapl: Correct hotplug correction

On Thu, May 22, 2014 at 03:13:46PM +0530, Srivatsa S. Bhat wrote:
> On 05/22/2014 02:53 PM, Borislav Petkov wrote:
> > From: Borislav Petkov <bp@...e.de>
> > 
> > So 009f225ef050 ("powercap, intel-rapl: Fix CPU hotplug callback
> > registration") says how get_/put_online_cpus() should be replaced with
> > this cpu_notifier_register_begin/_done().
> > 
> > But they're still there so what's up?
> > 
> 
> Ok, so I retained that because the comments in the code said that
> the caller of rapl_cleanup_data() should hold the hotplug lock.
> 
> Here is the snippet from the patch's changelog:
> 
>     ...
>     Fix the intel-rapl code in the powercap driver by using this latter form
>     of callback registration. But retain the calls to get/put_online_cpus(),
>     since they also protect the function rapl_cleanup_data(). By nesting
>     get/put_online_cpus() *inside* cpu_notifier_register_begin/done(), we avoid
>     the ABBA deadlock possibility mentioned above.

My bad, I missed that part.

> But looking closer at the code, I think the only requirement is that
> rapl_cleanup_data() should be protected against CPU hotplug, and we
> don't actually need to hold the cpu_hotplug.lock per-se.

What is the difference between "CPU hotplug" and cpu_hotplug.lock?
>From looking at the code those are two different mutexes with
cpu_hotplug.lock, i.e. get_online_cpus() having a preemption point.

And yet, you want to replace get_/put_online_cpus() with
cpu_notifier_register_begin/_done() which is kinda the same thing but
not really. The one protects against hotplug operations and the other
protects against cpu hotplug notifier registration.

Oh, and there's a third one, aliased to the notifier one, which
is "attempting to serialize the updates to cpu_online_mask &
cpu_present_mask."

So why, oh why do we need three? This is absolutely insane. Do we have
at least one sensible reason why cpu hotplug users should need to know
all that gunk?

> cpu_notifier_register_begin()/end() also provide equivalent protection
> against CPU hotplug. So we should be able to remove the get/put_online_cpus()
> from intel-rapl driver.

Btw, rapl_exit() has both calls too :-\.

-- 
Regards/Gruss,
    Boris.

Sent from a fat crate under my desk. Formatting is fine.
--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/