Re: [Xen-devel] [PATCH v5 03/12] x86/smp: don't online cpu if hypervisor_ap_setup fails

[Wei Liu <wl@xxxxxxx>]

On Fri, Jan 31, 2020 at 02:53:45PM +0100, Jan Beulich wrote:
> On 29.01.2020 21:20, Wei Liu wrote:
> > Push hypervisor_ap_setup down to smp_callin.
> > 
> > Take the chance to replace xen_guest with cpu_has_hypervisor.
> > 
> > Signed-off-by: Wei Liu <liuwe@xxxxxxxxxxxxx>
> > ---
> >  xen/arch/x86/smpboot.c | 10 +++++++---
> >  1 file changed, 7 insertions(+), 3 deletions(-)
> > 
> > diff --git a/xen/arch/x86/smpboot.c b/xen/arch/x86/smpboot.c
> > index c9d1ab4423..93b86a09e9 100644
> > --- a/xen/arch/x86/smpboot.c
> > +++ b/xen/arch/x86/smpboot.c
> > @@ -199,6 +199,13 @@ static void smp_callin(void)
> >          goto halt;
> >      }
> >  
> > +    if ( cpu_has_hypervisor && (rc = hypervisor_ap_setup()) != 0 )
> > +    {
> > +        printk("CPU%d: Failed to initialise hypervisor functions. Not 
> > coming online.\n", cpu);
> > +        cpu_error = rc;
> > +        goto halt;
> > +    }
> 
> There are a few things done up from here which may possibly
> better come after hypervisor interface setup (the two APIC
> related calls in particular). Are you sure you can safely
> move it this far down in the function?

Xen guest's usage of APIC is no different than, say, hvm's usage. If hvm
can be this far down, Xen / Hyper-V can, too.

Furthermore, APIC code has no dependency on guest code.

Wei.

> 
> Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel