Re: [PATCH] x86/msr: do not warn on writes to OC_MAILBOX

From: Andy Lutomirski <luto@kernel.org>
To: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Cc: Borislav Petkov <bp@alien8.de>,
	Sultan Alsawaf <sultan@kerneltoast.com>,
	"Jason A. Donenfeld" <Jason@zx2c4.com>,
	kitsunyan <kitsunyan@airmail.cc>,
	"Brown, Len" <len.brown@intel.com>, X86 ML <x86@kernel.org>,
	LKML <linux-kernel@vger.kernel.org>,
	Linus Torvalds <torvalds@linux-foundation.org>
Subject: Re: [PATCH] x86/msr: do not warn on writes to OC_MAILBOX
Date: Wed, 9 Sep 2020 17:08:06 -0700	[thread overview]
Message-ID: <CALCETrWAa+WogyaZ+XGsVxQ+3-j3C+VoQnRWz8L+FUgRpY7nfw@mail.gmail.com> (raw)
In-Reply-To: <fa447f6b7c7f03cc0c55573d5736889cee81a1e6.camel@linux.intel.com>

On Tue, Sep 8, 2020 at 6:02 PM Srinivas Pandruvada
<srinivas.pandruvada@linux.intel.com> wrote:
>
> On Tue, 2020-09-08 at 21:30 +0200, Borislav Petkov wrote:
> > On Tue, Sep 08, 2020 at 12:18:38PM -0700, Sultan Alsawaf wrote:
> > > I'd like to point out that on Intel's recent 14nm parts,
> > > undervolting
> > > is not so much for squeezing every last drop of performance out of
> > > the
> > > SoC as it is for necessity.
> >
> > <snip interesting examples>
> >
> > Sounds to me that this undervolting functionality should be part of
> > the kernel and happen automatically. I have no clue, though, whether
> > people who do it, just get lucky and undervolting doesn't cause any
> > other hardware issues, or there's a real reason for this power
> > madness
> > and if not done, power-related failures happen only on some boxes so
> > they decided to do them on all.
> >
> > Or maybe BIOS is nuts, which is not a stretch.
> >
> The whole OC is based on experiments to come to correct values. This
> depends on whole system design not just CPUs.
> https://www.intel.com/content/www/us/en/gaming/resources/how-to-overclock.html
> It warns about system stability.
>
> > Srinivas, what's the story here?
> I checked and there is no public spec. There are several mailbox
> commands with version dependent on the processor.
>
> The actual OC mailbox implementation itself is implemented in Linux in
> intel_turbo_max_3 driver. So that is public.
> So someone can develop a driver and provide some sysfs to send mailbox
> commands, but kernel can't validate commands which can cause any
> security or stability issues. Not sure if this is acceptable standard.
> I don't think there is any precedent of creating such blind sysfs
> entries.
>
>

Having once written a rejected driver to poke at the Intel Xeon memory
controller SMBUS, there is at least precedent for writing drivers for
dangerous interfaces, if not merging said drivers :)

But we have drivers for various EC interfaces, for other SMBUS hosts,
etc, and all of these can cause all kinds of mischief if misused.