I am looking at some PV driver issues we are seeing with 32 bit HVM guests hosted on 64 bit hypervisor. It looks like the problem can be traced to the fact that the hypervisor''s view of the shared info structure is different from the PV drivers view - The PV driver is built for a 32 bit environment while the hypervisor is built for a 64 bit environment. Specifically, the offset for the field evtchn_mask is different as seen by the PV drivers compared to what the hypervisor thinks it is. How is this issue supposed to be dealt with. Regards, K. Y _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 23/4/07 17:51, "Ky Srinivasan" <ksrinivasan@novell.com> wrote:> I am looking at some PV driver issues we are seeing with 32 bit HVM guests > hosted on 64 bit hypervisor. It looks like the problem can be traced to the > fact that the hypervisor''s view of the shared info structure is different from > the PV drivers view - The PV driver is built for a 32 bit environment while > the hypervisor is built for a 64 bit environment. Specifically, the offset for > the field evtchn_mask is different as seen by the PV drivers compared to what > the hypervisor thinks it is. How is this issue supposed to be dealt with.We probably need to extend the IS_COMPAT() macro to apply to HVM guests. Set a flag in the domain structure as soon as any VCPU enters long mode. Then use the existing vcpu_info and shared_info access macros that Jan introduced for PV PAE-on-64. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Thanks Kier. I will give this a shot. K. Y>>> On Mon, Apr 23, 2007 at 1:08 PM, in message<C252A68C.625D%Keir.Fraser@cl.cam.ac.uk>, Keir Fraser <Keir.Fraser@cl.cam.ac.uk> wrote:> On 23/4/07 17:51, "Ky Srinivasan" <ksrinivasan@novell.com> wrote: > >> I am looking at some PV driver issues we are seeing with 32 bit HVM guests >> hosted on 64 bit hypervisor. It looks like the problem can be traced to the >> fact that the hypervisor''s view of the shared info structure is different > from >> the PV drivers view - The PV driver is built for a 32 bit environment while >> the hypervisor is built for a 64 bit environment. Specifically, the offset > for >> the field evtchn_mask is different as seen by the PV drivers compared to > what >> the hypervisor thinks it is. How is this issue supposed to be dealt with. > > We probably need to extend the IS_COMPAT() macro to apply to HVM guests. Set > a flag in the domain structure as soon as any VCPU enters long mode. Then > use the existing vcpu_info and shared_info access macros that Jan introduced > for PV PAE- on- 64. > > -- Keir_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel