Re: [Xen-devel] [PATCH v3 1/2] Xen: Fix live migration while enabling APICV

["Li, Jiongxi" <jiongxi.li@xxxxxxxxx>]

> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> Sent: Monday, January 28, 2013 7:22 PM
> To: Li, Jiongxi; 'xen-devel@xxxxxxxxxxxxx'
> Cc: Keir Fraser
> Subject: Re: [Xen-devel] [PATCH v3 1/2] Xen: Fix live migration while enabling
> APICV
> 
> >>> On 28.01.13 at 05:13, "Li, Jiongxi" <jiongxi.li@xxxxxxxxx> wrote:
> > SVI should be restored in case guest is processing virtual interrupt
> > while saving a domain state. Otherwise SVI would be missied when
> > virtual interrupt delivery is enabled.
> 
> Why would RVI not need similar treatment?
Because RVI will be updated at every VMEntry at vmx_intr_assist(). Basically 
SVI update is handled
by APICV hardware, but in live migration case, we have to restore it.


> 
> > --- a/xen/arch/x86/hvm/vmx/vmx.c
> > +++ b/xen/arch/x86/hvm/vmx/vmx.c
> > @@ -1419,6 +1419,28 @@ static int
> vmx_virtual_intr_delivery_enabled(void)
> >      return cpu_has_vmx_virtual_intr_delivery;
> >  }
> >
> > +static void vmx_set_svi(int isr, struct vcpu *v) {
> > +    unsigned long status;
> > +    u8 old;
> > +
> > +    if ( !cpu_has_vmx_virtual_intr_delivery )
> > +        return;
> > +
> > +    if (isr == -1)
> 
> Perhaps better "isr < 0"?
> 
> > +        isr = 0;
> > +
> > +    vmx_vmcs_enter(v);
> > +    status = __vmread(GUEST_INTR_STATUS);
> > +    old = status >> 8;
> > +    if (isr != old) {
> > +        status &= 0x0FF;
> > +        status |= isr << 8;
> 
> Could we get #define-s for all these literal numbers (including using those in
> vmx_intr_assist(), where similar literal numbers exist)?
> 
> > --- a/xen/include/asm-x86/hvm/hvm.h
> > +++ b/xen/include/asm-x86/hvm/hvm.h
> > @@ -183,6 +183,7 @@ struct hvm_function_table {
> >      /* Virtual interrupt delivery */
> >      void (*update_eoi_exit_bitmap)(struct vcpu *v, u8 vector, u8 trig);
> >      int (*virtual_intr_delivery_enabled)(void);
> > +    void (*set_svi)(int isr, struct vcpu *v);
> 
> Please give this a better, vendor neutral name.
> 
> Jan
> 
> >
> >      /*Walk nested p2m  */
> >      int (*nhvm_hap_walk_L1_p2m)(struct vcpu *v, paddr_t L2_gpa,
> 


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