On Fri, 9 Jun 2000, Philipp Rumpf wrote:
> Rather obvious one :)
>
wasn't obvious to me, I was wondering how could 'm' become NULL for "no
apparent reason" :)
> What happens is we execute cpuid as serializing, but don't tell gcc this
> invalidates %eax, %ebx, %ecx, and %edx. This should be corrected by this
> patch:
>
> diff -ur linux/arch/i386/kernel/microcode.c linux-prumpf/arch/i386/kernel/microcode.c
> --- linux/arch/i386/kernel/microcode.c Fri Jun 9 19:05:09 2000
> +++ linux-prumpf/arch/i386/kernel/microcode.c Fri Jun 9 19:06:19 2000
> @@ -227,7 +227,7 @@
> }
>
> wrmsr(0x79, (unsigned int)(m->bits), 0);
> - __asm__ __volatile__ ("cpuid");
> + __asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx", "cc");
> rdmsr(0x8B, val[0], val[1]);
>
> req->err = 0;
>
>
> Tigran, does this look okay to you ?
>
Simon, that is one bug off the list of your stuff to fix :)
Just to clairfy - the oops here is due to m (which is %ebx) becoming 0
so m->date results in 8(%ebx) which means dereferencing kernel virtual
address=8.
Philipp, thank you for the fix!
Regards,
Tigran
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Jun 10 2000 - 05:29:49 EDT