linux-kernel - Re: [PATCH 02/13] x86/microcode: Use own MSR accessors

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:   Tue, 17 Jan 2017 20:12:24 +0100 (CET)
From:   Thomas Gleixner <tglx@...utronix.de>
To:     Borislav Petkov <bp@...en8.de>
cc:     X86 ML <x86@...nel.org>, LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 02/13] x86/microcode: Use own MSR accessors

On Tue, 17 Jan 2017, Borislav Petkov wrote:

> On Tue, Jan 17, 2017 at 06:51:06PM +0100, Thomas Gleixner wrote:
> > That would get rid of all the extra microcode specific MSR accessors which
> > are just yet another copy of stuff in msr.h.
> 
> Well, I did think about reusing those but last time I did, they received
> those tracepoints (apparently, we're sprinkling dumb tracepoints left
> and right because good ol' staring at the code is just too hard) which
> simply doesn't work on 32-bit before paging is enabled.
>
> Then, __native_write_msr_notrace() has exception handling which doesn't
> work before paging has been enabled on 32-bit - this is when the 32-bit
> microcode update path happens due to paging hardware bugs in CPUs which
> are fixed in microcode. So we must run that early on 32-bit.

Well, the exception handling is irrelevant in that case. If the MSR access
succeeds then the exception handling is not invoked. If it fails then it
does not matter much whether you die from the MSR #GP or from the exception
handling attempt #GP.

> So before someone decides to add more "functionality" to the generic MSR
> accessors and break the microcode loader once more, I'd really really
> prefer to have private accessors. They're small enough so shouldn't be
> that much of a bloat.

I can understand that, but we better have real native functions in msr.h
which do exactly what they are supposed to do: read or write the MSR,
nothing else. Add a comment to those function which should be immutable and
threaten that offenders will be slapped with stinking trouts or frozen
sharks.

We really want to have a single place for all of this MSR stuff, because
following your argumentation will be a perfect precedence for more private
implementations.

Thanks,

	tglx