Re: [PATCH 5/8] x86, microcode, intel: don't check extsig entry checksum

From: Borislav Petkov <bp@alien8.de>
To: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Cc: Fenghua Yu <fenghua.yu@intel.com>,
	linux-kernel@vger.kernel.org, H Peter Anvin <hpa@zytor.com>
Subject: Re: [PATCH 5/8] x86, microcode, intel: don't check extsig entry checksum
Date: Fri, 7 Nov 2014 18:49:17 +0100	[thread overview]
Message-ID: <20141107174917.GB5180@pd.tnic> (raw)
In-Reply-To: <20141031171440.GA27028@khazad-dum.debian.net>

On Fri, Oct 31, 2014 at 03:14:40PM -0200, Henrique de Moraes Holschuh wrote:
> > I don't know where you come up with this but the code you're removing
> > was added in 2013:
> > 
> > 	e666dfa273db ("x86/microcode_intel_lib.c: Early update ucode on Intel's CPU")
> 
> Was it?  As far as I know, it is "heavily based" (i.e. copied and adapted)
> on code from the regular microcode driver, and not rewritten from scrach.
> 
> If I am correct, this really is code from 2003, changed by 11 years worth of
> file renaming, file splits, and refactoring.
> 
> It took me about 30 minutes to track the origin of this code in the regular
> Intel microcode driver.  As far as I can tell, it was added to version 1.12
> of the original driver, somewhere between v2.6.0-test7 and v2.6.0-test8.
> 
> The Intel engineers credited for the original change are: Nitin Kamble
> <nitin.a.kamble@intel.com> and Jun Nakajima <jun.nakajima@intel.com>.
> (information from the changelog at the top of the driver source).
> 
> You can see one of the earliest versions of the code here:
> https://git.kernel.org/cgit/linux/kernel/git/history/history.git/tree/arch/i386/kernel/microcode.c?id=v2.6.0-test8
> 
> The extended table checksum gets calculated around line 323.  The possibly
> problematic logic is in line 341 and 345, and matches what the modern driver
> does.

Ok so e666dfa273db only adds code so I assumed it was new.

> I am not so sure.  Besides, it is entirely possible that the Intel SDM got
> updated in that area after the september 2003 code drop.  This is very hard
> to verify outside of Intel.   Or, maybe, the issue was detected internally
> at Intel, and it was decided that the issue should be shelved until the day
> extended microcode tables are actually required (which didn't happen yet).

Yeah, I wouldn't want to remove code from the loader without Intel
people saying something first.

-- 
Regards/Gruss,
    Boris.

Sent from a fat crate under my desk. Formatting is fine.
--