Bug#410819: Please add VIA C7 processor definition
Aurélien GÉRÔME
Version: 1.72
Severity: wishlist
Tags: patch
Hi,
A popular French datacenter called Dedibox [1] is using a specifically
built motherboard with a VIA C7 processor. Via a secure web interface,
you can install Debian Sarge among other distros (with their untrusty
closed source system).
I am currently setting up a documentation to use the official Debian
Installer with the official Debian kernels on that machine. Once done,
I will put a link on the Debian Wiki.
By default, only the -486 subarch is available at most, because
/proc/cpuinfo is different from other x86 machines, i.e. neither
AMD nor Intel... Therefore, I did a patch to get the -686 subarch
available on that distinct hardware with all the flavours, e.g. xen,
vserver, xen+vserver.
Please consider the attached patch against the current SVN.
Cheers,
[1] <http://www.dedibox.fr/>
--
.''`. Aurélien GÉRÔME
: :' :
`. `'` Free Software Developer
`- Unix Sys & Net Admin
Frans Pop
> Please consider the attached patch against the current SVN.
I've made one correction relative to the patch:
- " GenuineIntel"|" GenuineTMx86"|" CentaurHauls"*)
+ " GenuineIntel"|" GenuineTMx86"*|" CentaurHauls")
Keeping the "*" at the end changes the test for GenuineTMx86, which would
be a regression.
Frans Pop
thanks
On Tuesday 13 February 2007 18:10, Aurélien GÉRÔME wrote:
> I am currently setting up a documentation to use the official Debian
> Installer with the official Debian kernels on that machine. Once done,
> I will put a link on the Debian Wiki.
OK, so that is targeted at supporting Etch?
> By default, only the -486 subarch is available at most, because
> /proc/cpuinfo is different from other x86 machines, i.e. neither
> AMD nor Intel... Therefore, I did a patch to get the -686 subarch
> available on that distinct hardware with all the flavours, e.g. xen,
> vserver, xen+vserver.
I have asked myself a little bit if it really makes sense to support such
a very specific use case. As the patch is minimal and maintenance should
be negligible, I guess it does not do any harm.
> Please consider the attached patch against the current SVN.
Thanks for providing the test case too.
Next time, please make sure your name is UTF-8 encoded in the changelog.
Bonus points for marking the new changelog entry as UNRELEASED.
Cheers,
FJP
Aurélien GÉRÔME
> Keeping the "*" at the end changes the test for GenuineTMx86, which would
> be a regression.
Damn me, nice you spotted that one. I should have tested it back into
qemu afterwards...
Cheers,
Aurélien GÉRÔME
> On Tue, Feb 13, 2007 at 07:00:55PM +0100, Frans Pop wrote:
> > Keeping the "*" at the end changes the test for GenuineTMx86, which would
> > be a regression.
>
> Damn me, nice you spotted that one. I should have tested it back into
> qemu afterwards...
Actually, I remember that I tested it back, but qemu emulates a
GenuineIntel, period. :)
Aurélien GÉRÔME
> > Keeping the "*" at the end changes the test for GenuineTMx86, which would
> > be a regression.
>
> Damn me, nice you spotted that one. I should have tested it back into
> qemu afterwards...
... which leads me to think that the testsuite should have barfed.
$ grep GenuineTMx86 kernel/tests/i386/*
kernel/tests/i386/oqo1.cpuinfo:vendor_id : GenuineTMx86
There is no test case for a " GenuineTMx86"* string...
Aurélien GÉRÔME
> On Tuesday 13 February 2007 18:10, Aurélien GÉRÔME wrote:
> > I am currently setting up a documentation to use the official Debian
> > Installer with the official Debian kernels on that machine. Once done,
> > I will put a link on the Debian Wiki.
>
> OK, so that is targeted at supporting Etch?
Sure, it works well. At least, better than in -486 flavour. There is
a kernel optimisation for VIA C3/C7 processors which is disabled by
default in the Debian kernel package. I will have a look whether it
is fine or not with other 686 machines to activate it...
> > By default, only the -486 subarch is available at most, because
> > /proc/cpuinfo is different from other x86 machines, i.e. neither
> > AMD nor Intel... Therefore, I did a patch to get the -686 subarch
> > available on that distinct hardware with all the flavours, e.g. xen,
> > vserver, xen+vserver.
>
> I have asked myself a little bit if it really makes sense to support such
> a very specific use case. As the patch is minimal and maintenance should
> be negligible, I guess it does not do any harm.
Well, it does no harm. The machine is a 686, it is a bit unsatisfying
not to run the kernel as such.
> > Please consider the attached patch against the current SVN.
>
> Thanks for providing the test case too.
> Next time, please make sure your name is UTF-8 encoded in the changelog.
$ wget "http://bugs.debian.org/cgi-bin/bugreport.cgi/via-c7.patch?bug=410819;msg=5;att=1" && file via*
via-c7.patch?bug=410819;msg=5;att=1: UTF-8 Unicode text
I am always unicode-enabled as you can see... Maybe an issue on
your end?
> Bonus points for marking the new changelog entry as UNRELEASED.
Indeed. :)
Cheers,