linux-kernel - Re: [PATCH] x86/intel/quark: Parameterize the kernel's IMR lock logic

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, 19 Feb 2016 01:07:58 +0000
From:	Bryan O'Donoghue <pure.logic@...us-software.ie>
To:	Ingo Molnar <mingo@...nel.org>
Cc:	linux-kernel@...r.kernel.org, corbet@....net, tglx@...utronix.de,
	mingo@...hat.com, hpa@...or.com, x86@...nel.org,
	andriy.shevchenko@...ux.intel.com, boon.leong.ong@...el.com,
	fengguang.wu@...el.com, linux-doc@...r.kernel.org
Subject: Re: [PATCH] x86/intel/quark: Parameterize the kernel's IMR lock
 logic

On Thu, 2016-02-18 at 19:53 +0100, Ingo Molnar wrote:
> * Bryan O'Donoghue <pure.logic@...us-software.ie> wrote:
> 
> > On Thu, 2016-02-18 at 08:58 +0100, Ingo Molnar wrote:
> > > So why not simply do the patch below? Very few people use boot
> > > parameters, and the 
> > > complexity does not seem to be worth it.
> > > 
> > > Furthermore I think an IMR range in itself is safe enough - it's
> > > not
> > > like such 
> > > register state is going to be randomly corrupted, even with the
> > > 'lock' bit unset. 
> > 
> > 
> > Hi Ingo.
> > 
> > I agree - to flip the lock bit you need to be in ring-0 anyway.
> > 
> > > So it's a perfectly fine protective measure against accidental
> > > memory
> > > corruption 
> > > from the DMA space. It should not try to be more than that.
> > > 
> > > And once we do this, I suggest we get rid of the 'lock' parameter
> > > altogether - 
> > > that will further simplify the code.
> > > 
> > > Thanks,
> > > 
> > >         Ingo
> > 
> > That was the V1 of this patch
> > 
> > https://groups.google.com/forum/#!topic/linux.kernel/6ZuVOF3TJow
> 
> heh ;-)

:)

> > Andriy asked for the boot parameter to control the state of the IMR
> > lock bit, I'm just as happy to go back to that version TBH
> 
> I really think it's over-engineered - especially considering that
> with the kernel 
> lock-down removed there's no other IMR area that is really locked
> down - so we 
> could get rid of the whole 'locked' logic that would simplify the
> code throughout.

I'm in favour of that too. Charitably I think locking a register like
this makes sense only when you talk about it in a meeting room
somewhere; as soon as you go to try to use it in a real situation you
find its far more trouble than its really worth.

So, I'm going to trim that out of this code unless I hear some pushback
from elsewhere in the 1/2 day or so.

> 
> Yeah, it's a nice looking hardware feature - but I don't think it's
> particularly 
> useful in terms of extra protection.
> 
> Thanks,
> 
> 	Ingo