Re: [PATCH v4 08/15] x86/altp2m: add control of suppress_ve.

[George Dunlap <george.dunlap@eu.citrix.com>]

On 07/10/2015 12:49 PM, Jan Beulich wrote:
>>>> On 10.07.15 at 13:11, <george.dunlap@eu.citrix.com> wrote:
>> On 07/10/2015 10:39 AM, Jan Beulich wrote:
>>>>>> On 10.07.15 at 02:52, <edmund.h.white@intel.com> wrote:
>>>> @@ -1528,16 +1528,17 @@ bool_t p2m_mem_access_check(paddr_t gpa, unsigned long gla,
>>>>      vm_event_request_t *req;
>>>>      int rc;
>>>>      unsigned long eip = guest_cpu_user_regs()->eip;
>>>> +    bool_t sve;
>>>>  
>>>>      /* First, handle rx2rw conversion automatically.
>>>>       * These calls to p2m->set_entry() must succeed: we have the gfn
>>>>       * locked and just did a successful get_entry(). */
>>>>      gfn_lock(p2m, gfn, 0);
>>>> -    mfn = p2m->get_entry(p2m, gfn, &p2mt, &p2ma, 0, NULL);
>>>> +    mfn = p2m->get_entry(p2m, gfn, &p2mt, &p2ma, 0, NULL, &sve);
>>>>  
>>>>      if ( npfec.write_access && p2ma == p2m_access_rx2rw ) 
>>>>      {
>>>> -        rc = p2m->set_entry(p2m, gfn, mfn, PAGE_ORDER_4K, p2mt, p2m_access_rw);
>>>> +        rc = p2m->set_entry(p2m, gfn, mfn, PAGE_ORDER_4K, p2mt, p2m_access_rw, sve);
>>>>          ASSERT(rc == 0);
>>>>          gfn_unlock(p2m, gfn, 0);
>>>>          return 1;
>>>> @@ -1546,7 +1547,7 @@ bool_t p2m_mem_access_check(paddr_t gpa, unsigned long gla,
>>>>      {
>>>>          ASSERT(npfec.write_access || npfec.read_access || npfec.insn_fetch);
>>>>          rc = p2m->set_entry(p2m, gfn, mfn, PAGE_ORDER_4K,
>>>> -                            p2mt, p2m_access_rwx);
>>>> +                            p2mt, p2m_access_rwx, -1);
>>>
>>> So why -1 here ...
>>>
>>>> @@ -1566,14 +1567,14 @@ bool_t p2m_mem_access_check(paddr_t gpa, unsigned long gla,
>>>>          else
>>>>          {
>>>>              gfn_lock(p2m, gfn, 0);
>>>> -            mfn = p2m->get_entry(p2m, gfn, &p2mt, &p2ma, 0, NULL);
>>>> +            mfn = p2m->get_entry(p2m, gfn, &p2mt, &p2ma, 0, NULL, &sve);
>>>>              if ( p2ma != p2m_access_n2rwx )
>>>>              {
>>>>                  /* A listener is not required, so clear the access
>>>>                   * restrictions.  This set must succeed: we have the
>>>>                   * gfn locked and just did a successful get_entry(). */
>>>>                  rc = p2m->set_entry(p2m, gfn, mfn, PAGE_ORDER_4K,
>>>> -                                    p2mt, p2m_access_rwx);
>>>> +                                    p2mt, p2m_access_rwx, sve);
>>>
>>> ... but sve here, when -1 means "retain current setting" anyway?
>>> (Same question applies elsewhere.)
>>
>> This is my code. I considered whether to use -1 here, but since we're
>> reading and retaining gfn, mfn, and p2mt, it seemed more consistent
>> stylistically to just read and re-write it along with the others.
>>
>> In any case I don't have strong opinions.
> 
> I'd suggest the other mechanism so one can easily see which places
> actually want to change the flag (or set it to a specific value). But in
> the end it's your call which way to go.

That does make sense.  In fact, if there were "leave the default"
options for the other values (mfn, p2mt, &c) it would be clearer that
only the page order and the access rights were being changed here.

Anyway that's a minor issue at this point.  Ed / Ravi, feel free to
change it according to Jan's suggestion, or leave it as it is for now.

 -George