Re: [Xen-devel] [RFC v3 2/6] xen/arm: Add save/restore support for ARM GIC V2

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Fri, 2014-05-09 at 09:12 -0500, Wei Huang wrote:
> On 05/08/2014 05:47 PM, Andrew Cooper wrote:
> >> +DECLARE_HVM_SAVE_TYPE(GICH_V2, 3, struct hvm_arm_gich_v2);
> >> +
> >>   /*
> >>    * Largest type-code in use
> >>    */
> >> -#define HVM_SAVE_CODE_MAX 1
> >> +#define HVM_SAVE_CODE_MAX 3
> >>
> >>   #endif
> >>
> >
> > On x86, we require that HVM save records only contain architectural
> > state.  Not knowing arm myself, it is not clear from your comments
> > whether this is the case or not.  Can you confirm whether it is or not?
> Most states are guest architecture states which include core registers, 
> arch timer, memory. GIC states are arguable, given that Xen uses data 
> structures (e.g. struct vgic_irq_rank) to represent GIC states internally.

(note: I've not looked at this series for ages, I plan to look at this
new version next week)

The contents of vgic_irq_rank is still a set of architectural register,
I think (the rank thing is just to account for the fact that some
registers use 1 bit to describe 32-registers, some use 2 bits to
describe 16, etc).

If there is something in a vgic rank which needs saving which isn't in
an architectural form then it needs synthesizing to/from the
architectural state on save/restore.

IOW ARM has the same requirements as x86 here.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel