On Thu, Mar 11, 2010 at 02:38:54PM +0100, Andre Przywara wrote:
> Hi,
>
> a rebased and refined version of my CPUID cleanup series. This should
> now apply on top of git head. Compared to the last post this leaves
> out the bits already done by John Cooper's patch. Also I added a separate
> patch to deal with an accidently removed line (patch 3/13).
>
> This patchset cleans up the CPUID handling code in QEMU. The biggest change
> is obviously the move of the CPUID function to a separate file (cpuid.c).
> This helps to split up a rather large source file, which's name (helper.c) is
> also a bit misleading.
> Additionally the rest of the patches beautifies or simplifies some code.
> Feature additions are:
> 5/13: add missing CPUID feature bit names (updated)
> 8/13: -cpu host propagates more CPUID leafs, so that the cache topology
> will be visibile in the guest
> 9/13: add CPUID feature bit trimming for TCG: Features not supported by
> the emulator will be masked out.
> 10/13: always show all CPU types: also expose the newer (64bit) CPU types
> for the i386 emulator. 64bit features will be masked out due to 9/13.
> 11/13: add kvm32 CPU model: Per popular request add a counterpart to kvm64
> describing a basic hardware virtualization capable CPU for migration
> purposes.
> 13/13: Set the feature bits of qemu{32,64} to the set supported by TCG.
>
> The last patch is actually debatable, as it changes the default CPUID bits
> seen by guest. On the other hand it is only consequent to do so, as
> qemu{32,64} are meant to be artifical CPUs.
>
> More details in the commit messages.
>
> Please review, comment and apply.
>
Thanks, I have applied all patches except:
- 08: I prefer someone knowing KVM better than me reviews it
- 11: same
- 13: See comment from Jamie
--
Aurelien Jarno GPG: 1024D/F1BCDB73
aurel...@aurel32.net http://www.aurel32.net
