Re: [PATCH v3 2/3] xen/pvh: check permissions when adding MMIO regions

["Roger Pau Monné" <roger.pau@citrix.com>]

El 22/01/15 a les 16.43, Jan Beulich ha escrit:
>>>> On 22.01.15 at 16:19, <roger.pau@citrix.com> wrote:
>> --- a/xen/arch/x86/domain_build.c
>> +++ b/xen/arch/x86/domain_build.c
>> @@ -320,11 +320,24 @@ static __init void pvh_add_mem_mapping(struct domain *d, unsigned long gfn,
>>  {
>>      unsigned long i;
>>      p2m_access_t a;
>> +    mfn_t omfn;
>> +    p2m_type_t t;
>>      int rc;
>>  
>> -    a = p2m_get_hostp2m(d)->default_access;
> 
> Iirc this is rwx.
> 
>>      for ( i = 0; i < nr_mfns; i++ )
>>      {
>> +        if ( !iomem_access_permitted(d, mfn + i, mfn + i) )
>> +        {
>> +            omfn = get_gfn_query_unlocked(d, gfn + i, &t);
>> +            guest_physmap_remove_page(d, gfn + i, mfn_x(omfn), PAGE_ORDER_4K);
>> +            continue;
>> +        }
>> +
>> +        if ( rangeset_contains_singleton(mmio_ro_ranges, mfn + i) )
>> +            a = p2m_access_r;
>> +        else
>> +            a = p2m_access_rw;
> 
> Shouldn't these two therefore be rx and rwx respectively? Or even
> better ->default_access in the else case (albeit that doesn't really
> matter here since nothing can have changed that field from its
> default value)? I'm particularly thinking of ROMs that may be sitting
> in these areas.

Yes, it should be rx and rwx. I would prefer to use the explicit types.
IMHO it's easier to understand to which access type it's set (without
having to dig to what value p2m_get_hostp2m(d)->default_access is set).

Roger.