Re: Better communication about spectre/meltdown

[Fabian Grünbichler <fabian.gruenbichler@student.tuwien.ac.at>]

On Sat, Mar 03, 2018 at 11:07:12AM -0500, Roberto C. Sánchez wrote:
> On Sat, Mar 03, 2018 at 03:22:14PM +0000, Ben Hutchings wrote:
> > 
> > I think that backporting gcc-4.9 and building the kernel with it (for
> > x86) is lower risk than backporting the retpoline patches to gcc-4.7
> > and building the kernel with that.  (In fact it's not just the kernel;
> > if you change gcc-4.7 that has the potential to affect most updates to
> > wheezy, even though use of retpoline should be disabled by defaul.)
> > 
> Thanks for clarifying.
> 
> I will work on backporting your patched gcc-4.9 to wheezy. Should I also
> start working on getting the wheezy kernel building with gcc-4.9?  I
> understand that an upload must wait for the kernel microde update you
> mentioned previously.

FWIW / to prevent potential confusion, I am almost certain Ben did not
mean "microcode" (which is not part of the kernel except for the loading
infrastructure for CPU and other/firmware microcode), but actually
assembler, as in those parts of the kernel that are not written in C,
and thus not compiled with GCC, and thus need to be retpolined in some
other way if necessary (the kernel contains a fancy mechanism to select
between different variants of assembler code which has been heavily
utilized for this in current mainline kernels).