[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Better communication about spectre/meltdown



On Mon, 2018-02-26 at 14:40 -0500, Antoine Beaupré wrote:
> On 2018-02-25 13:57:07, Roberto C. Sánchez wrote:
> > On Sun, Feb 25, 2018 at 07:04:12PM +0100, Moritz Mühlenhoff wrote:
> > > On Sun, Feb 25, 2018 at 08:54:06AM -0500, Roberto C. Sánchez wrote:
> > > > Hi all,
> > > > 
> > > > Please see my rather long-winded summary of the current state of the
> > > > gcc-4.6/gcc-4.7 update. The bottom line is that I am looking for opions
> > > > and/or guidance for how to proceed.
> > > 
> > > Why 4.6 _and_ 4.7? Only the compiler used for building the amd64 3.2 kernel
> > > is relevant here?
> > > 
> > 
> > Both are triaged in dla-needed.txt. In any event, I have done no work at
> > all on 4.7 at this point, other than to observe that my investigation
> > into the differences in the option parsing code (which was the only
> > significant difficulty I encountered in backport the 4.9 patches) made
> > me think that backporting the 4.9 patches to 4.7 would be *easier* than
> > the backport to 4.6.
> > 
> > As far as I know, it has not been decided that 4.7 will be patched.
> 
> jessie also has two gcc compilers from what I can tell (4.8 and 4.9)
> yet, the security team is concentrating only on one (4.9). It seems like
> we should do the same (concentrate on a single compiler).
> 
> is there anything blocking the use of the 4.9 compiler in wheezy, short
> of, of course, the backport itself? It's true it's kind of nuts to
> introduce a *third* toolchain in a LTS update, but I wonder how feasible
> it is to maintain the two that are already there in the long term, if
> we're already having trouble with 4.6...
> 
> Can't the wheezy kernel build with 4.7 or 4.9 correctly? I guess that
> involves the buildds as well...?

It will almost certainly build correctly with 4.9 on x86.  AIUI the
Spectre mitigations in gcc are x86-specific, so there's no value in
changing it for ARM and there would be a risk of exceeding code size
limits on armel.  The kernel package already has provision for using
different compiler versions per-architecture.

Ben.

> Note that only the 4.9.x series has seen upstream releases in the last
> ~3 years. The last 4.7 release is 4.7.4, from june 2014, and for 4.6.x,
> 4.6.4 in April 2013. Have anyone tried to contact upstream to see if
> they are backporting those changes in any official capacity?

-- 
Ben Hutchings
This sentence contradicts itself - no actually it doesn't.

Attachment: signature.asc
Description: This is a digitally signed message part


Reply to: