Re: Re: [PATCH 4 of 4] Nested p2m: rework locking around nested-p2m flushes and updates

[Christoph Egger <Christoph.Egger@amd.com>]

On 06/24/11 12:45, Tim Deegan wrote:
> At 17:08 +0200 on 23 Jun (1308848939), Christoph Egger wrote:
>>> +    /* If this p2m table has been flushed or recycled under our feet,
>>> +     * leave it alone.  We'll pick up the right one as we try to
>>> +     * vmenter the guest. */
>>> +    if ( p2m->cr3 == nhvm_vcpu_hostcr3(v) )
>>> +         rv = set_p2m_entry(p2m, L2_gpa>>   PAGE_SHIFT,
>>> +                            page_to_mfn(maddr_to_page(L0_gpa)),
>>> +                            0 /*4K*/, p2mt, p2ma);
>>
>> The introduction of this check leads to this crash when the L2 guest
>> initializes its third vcpu:
>
> D'oh!
>
> diff -r b40d4bcca0d7 xen/arch/x86/mm/hap/nested_hap.c
> --- a/xen/arch/x86/mm/hap/nested_hap.c	Fri Jun 24 11:16:53 2011 +0100
> +++ b/xen/arch/x86/mm/hap/nested_hap.c	Fri Jun 24 11:43:30 2011 +0100
> @@ -100,7 +100,7 @@ nestedhap_fix_p2m(struct vcpu *v, struct
>                     paddr_t L2_gpa, paddr_t L0_gpa,
>                     p2m_type_t p2mt, p2m_access_t p2ma)
>   {
> -    int rv = 0;
> +    int rv = 1;
>       ASSERT(p2m);
>       ASSERT(p2m->set_entry);
>

Yes, that fixes the crash.

> This is what comes of not being able to test things properly.  Can you
> describe your test setup for me?  I'm having no luck getting
> xen-on-debian-on-xen-on-debian to work.

That should do it. If your problem is that the L1 Dom0 hangs at boot
then disable the TSC Deadline Timer in libxc/xc_cpuid_x86.c.
This feature is broken since its introduction.

Christoph


-- 
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Einsteinring 24, 85689 Dornach b. Muenchen
Geschaeftsfuehrer: Alberto Bozzo, Andrew Bowd
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632