Martin Mares wrote:
>
> Hi,
>
> > After Rafael posted his patch, I put some finishing touches to it,
> > converted it to kernel 2.0.34, and actually *tested* the thing.
>
> For 2.1 kernels, __initfunc's should be added and it probably would
> be wise to move everything to setup.c instead.
The patch that was uploaded is specific for 2.0.34+ kernels. It's in
three pieces:
1) A two line patch to /init/main.c that moves the check_bugs call. As
far as I can see, this has no negative effects on kernel operation. OTOH
this affects all architectures, so this seemingly small change could
break something. Testing needed
2) A big patch to setup.c. This patch is _not_ compatible with the
present Cyrix detection code in 2.1.10x kernels. You'll have to rewrite
this if you want to get it into 2.1.10x. Rafael's patch didn't have this
part. BTW this patch includes code to correctly detect and report MII
CPUs (tested, functional).
3) A big patch to bugs.h. This is the Cyrix specific code that was
removed from head.S. I believe it's a Bad Idea to try to move it back to
setup.c. The code is run just once during kernel boot time, it's
specific to 6x86 and 6x86L processors, and it doesn't get in the way of
other x86 processors/chipsets. And it's where Linus wanted it to be. So
please leave it in /include/asm-i386/bugs.h, right next to the F00F
Pentium bug and the K6 bug. It's cosy there.
> I'll look at it after I finish
> some exams.
Colin Plumb just sent me an email, warning me of two typos in the patch
I uploaded, that go undetected by gcc 2.7.2, but could cause a
compilation error on later gcc versions or egcs:
The asm statements that have constraints "=al" should be changed to just
"=a". "al" is not a legal constraint for gcc AFAIK (but is not reported
as an error :( ).
THX, Colin. Keen eye, eh? :-)
Cheers,
------------------------
André Balsa
andrebalsa@altern.org
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu