qemu-devel
[Top][All Lists]
Advanced

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

Re: [Qemu-devel] [PATCH 0/5] i386: cpu: remove duplicate feature names


From: Igor Mammedov
Subject: Re: [Qemu-devel] [PATCH 0/5] i386: cpu: remove duplicate feature names
Date: Wed, 26 Sep 2012 15:33:32 +0200

On Thu,  6 Sep 2012 17:05:34 -0300
Eduardo Habkost <address@hidden> wrote:

> The problem:
> 
>  - Some features are report at the same time on both CPUID[1].EDX and
>    CPUID[8000_0001].EDX on AMD CPUs (e.g. fpu, tsc, msr, pae, mmx).
>  - "-cpu <model>,+feature" should enable the bit only on CPUID[1] if
>    it's not an AMD CPU, but it should enable the bit on both CPUID[1] and
>    CPUID[8000_0001] if it's an AMD CPU.
>  - The same should happen when implementing CPU properties: setting the
>    property that enables a feature should set the duplicate
> CPUID[8000_0001].EDX bit only if CPU vendor is AMD.
> 
> Reference: http://article.gmane.org/gmane.comp.emulators.qemu/166024
> 
> The solution implemented by this series is:
>  - On the CPU model table and while parsing CPU options/properties, set the
> bit only on CPUID[1] (the x86_def_t.features field).
>  - When finishing initialization of the CPU cpuid fields, duplicate those
>    feature bits on cpuid_ext2_features if and only if the CPU vendor is AMD.
> 
> This series depends on the "x86 CPU patches that didn't get into 1.2"
> series: http://article.gmane.org/gmane.comp.emulators.qemu/168633
>   Message-Id: <address@hidden>
> 
> 
> Eduardo Habkost (5):
>   i386: kvm: bit 10 of CPUID[8000_0001].EDX is reserved
>   i386: kvm: use a #define for the set of alias feature bits
>   i386: cpu: replace EXT2_FEATURE_MASK with CPUID_EXT2_AMD_ALIASES
>   i386: cpu: eliminate duplicate feature names
>   i386: -cpu help: remove reference to specific CPUID leaves/registers
> 
>  target-i386/cpu.c | 59
> +++++++++++++++++++++++++++++++++++-------------------- target-i386/cpu.h |
> 12 +++++++++++ target-i386/kvm.c |  2 +-
>  3 files changed, 51 insertions(+), 22 deletions(-)
> 

Reviewed-by: Igor Mammedov <address@hidden>



reply via email to

[Prev in Thread] Current Thread [Next in Thread]