Re: [Xen-devel] [PATCH RFC 3/4] Implement save and restore for gic (template impl)

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

> @@ -65,3 +66,53 @@ long do_hvm_op(unsigned long op, 
> XEN_GUEST_HANDLE_PARAM(void) arg)
>  
>      return rc;
>  }
> +
> +static int gic_save(struct domain *d, hvm_domain_context_t *h)
> +{
> +    struct hvm_hw_gic ctxt;
> +    struct vcpu *v;
> +
> +    /* Save the state of GICs */
> +    for_each_vcpu( d, v )
> +    {
> +        ctxt.gic_hcr = v->arch.gic_hcr;
> +        ctxt.gic_vmcr = v->arch.gic_vmcr;
> +        ctxt.gic_apr = v->arch.gic_apr;
> +        memcpy( ctxt.gic_lr, v->arch.gic_lr, sizeof(v->arch.gic_lr) );
> +        ctxt.event_mask = v->arch.event_mask;
> +        ctxt.lr_mask = v->arch.lr_mask;

In general on x86 the policy is that hvm headers should store only
architectural state. I think this is a good policy to carry over to ARM.

With that in mind I don't think either event_mask or lr_mask are
architectural state but are actually internal state of the vgic
"emulation" which can and should be reconstructed when loading the
architectural state.

Ian


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