Re: [Xen-devel] [PATCH 6/7] x86/pagewalk: Consistently use guest_walk_*() helpers for translation

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 01/03/17 16:22, Jan Beulich wrote:
>>>> On 27.02.17 at 15:03, <andrew.cooper3@xxxxxxxxxx> wrote:
>> --- a/xen/arch/x86/mm/hap/guest_walk.c
>> +++ b/xen/arch/x86/mm/hap/guest_walk.c
>> @@ -98,7 +98,7 @@ unsigned long hap_p2m_ga_to_gfn(GUEST_PAGING_LEVELS)(
>>      /* Interpret the answer */
>>      if ( missing == 0 )
>>      {
>> -        gfn_t gfn = guest_l1e_get_gfn(gw.l1e);
>> +        gfn_t gfn = guest_walk_to_gfn(&gw);
>>          struct page_info *page;
>>          page = get_page_from_gfn_p2m(p2m->domain, p2m, gfn_x(gfn), &p2mt,
>>                                       NULL, P2M_ALLOC | P2M_UNSHARE);
>> --- a/xen/arch/x86/mm/shadow/multi.c
>> +++ b/xen/arch/x86/mm/shadow/multi.c
>> @@ -3109,7 +3109,7 @@ static int sh_page_fault(struct vcpu *v,
>>      }
>>  
>>      /* What mfn is the guest trying to access? */
>> -    gfn = guest_l1e_get_gfn(gw.l1e);
>> +    gfn = guest_walk_to_gfn(&gw);
>>      gmfn = get_gfn(d, gfn, &p2mt);
>>  
>>      if ( shadow_mode_refcounts(d) &&
> With these adjusted, guest_l1e_get_gfn() uses left are in shadow's
> multi.c (presumably okay to stay)

Yes - they are purposefully looking at the L1 entry.

> and at the end of guest_walk_tables() itself. Is that latter one not also in 
> need of
> conversion, considering the PSE36 reference in the description?

Well - it gets swallowed up in "Re-implement the pagetable walker", but
inside guest_walk_tables() itself, it is reasonable to rely on the
internal details.

~Andrew

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