Mailing List Archive

On 1/12/06 7:33 am, "He, Qing" <qing.he@intel.com> wrote:

> This patch changes the structure of shared_info and vcpu_info, so that
> its members (event channel related flags) have the same offset in
> x86_32/x86_64 environment. This modification enables 32bit PV-on-HVM
> driver to run transparently on both 32bit and 64bit hypervisor.

Breaks ABI compatibility, and Jan Beulich's PV-PAE-on-64 patches can be
extended for the HVM case.

-- Keir


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

>Breaks ABI compatibility, and Jan Beulich's PV-PAE-on-64 patches can be
>extended for the HVM case.
>
Keir,
I don't quite understand what "break ABI compatibility" mean? Doesn't it
mean old domX kernel can't run with new HV or vice versa? Since we only
made a small change to evtchn structure inside and exported APIs is
untouched, we think it won't affect others.

Thanks,
Xiaowei

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 1/12/06 08:17, "Yang, Xiaowei" <xiaowei.yang@intel.com> wrote:

> I don't quite understand what "break ABI compatibility" mean? Doesn't it
> mean old domX kernel can't run with new HV or vice versa? Since we only
> made a small change to evtchn structure inside and exported APIs is
> untouched, we think it won't affect others.

Yes, that's precisely what ABI compatibility means. Your patch breaks it and
so is unacceptable.

Luckily Jan Beulich has done the hard work of supporting the current 32-bit
ABI on a 64-bit hypervisor. He did this primarily to support paravirtual PAE
guests on 64-bit hypervisor, but I'm sure the work can be (fairly) easily
extended to support 32-bit PV-on-HVM drivers on 64-bit hypervisor.

-- Keir


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel