>>> On 02.10.15 at 20:23, <andrew.coop...@citrix.com> wrote:
> On 02/10/15 16:48, Roger Pau Monne wrote:
>> diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c
>> index bbec0e8..63b7a24 100644
>> --- a/xen/arch/x86/hvm/vmx/vmx.c
>> +++ b/xen/arch/x86/hvm/vmx/vmx.c
>> @@ -2412,7 +2412,8 @@ static void vmx_install_vlapic_mapping(struct vcpu *v)
>> {
>> paddr_t virt_page_ma, apic_page_ma;
>>
>> - if ( !cpu_has_vmx_virtualize_apic_accesses )
>> + if ( !cpu_has_vmx_virtualize_apic_accesses ||
>> + v->domain->arch.hvm_domain.vmx.apic_access_mfn == 0 )
>
> 0 is a valid (albeit very unlikely) mfn. The unused sentinel value
> should be INVALID_MFN. It appears that all the current uses of
> apic_access_mfn are buggy.
>From a purely theoretical pov 0 is indeed a valid MFN. However,
on x86 we don't hand memory below 1M to the allocator, so this
is only theoretically (or one could call it latently) buggy.
Jan
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
