[ros-dev] [ros-diffs] [jgardou] 66334: [NTOSKRNL/MM] - MiIsEntireRangeCommitted: Ensure the PTE we are checking is really faulted in. - Prefer MiPteToPde and MiPdeToPte (which should really be called MiFirstPteInPde) in...

[Jérôme Gardou]

If it takes me 6 months to understand why, then yes.

Hint: you can make this delay shorter.

Also the title is a bit misleading, it's the PDE that is not faulted in
in case we don't terminate the loop iteration early.

Le 19/02/2015 05:46, Alex Ionescu a écrit :
> it might fix an assert but the patch is incorrect. will this also take 6
> months to revert?
> 
> Best regards,
> Alex Ionescu
> 
> On Tue, Feb 17, 2015 at 6:19 AM, <jgardou at svn.reactos.org> wrote:
> 
>> Author: jgardou
>> Date: Tue Feb 17 14:19:05 2015
>> New Revision: 66334
>>
>> URL: http://svn.reactos.org/svn/reactos?rev=66334&view=rev
>> Log:
>> [NTOSKRNL/MM]
>>  - MiIsEntireRangeCommitted: Ensure the PTE we are checking is really
>> faulted in.
>>  - Prefer MiPteToPde and MiPdeToPte (which should really be called
>> MiFirstPteInPde) instead of MiAddressToPte and MiPteToAddress
>> Fixes weird failed ASSERT in page fault handler when using DPH.
>>
>> Modified:
>>     trunk/reactos/ntoskrnl/mm/ARM3/virtual.c
>>
>> Modified: trunk/reactos/ntoskrnl/mm/ARM3/virtual.c
>> URL:
>> http://svn.reactos.org/svn/reactos/trunk/reactos/ntoskrnl/mm/ARM3/virtual.c?rev=66334&r1=66333&r2=66334&view=diff
>>
>> ==============================================================================
>> --- trunk/reactos/ntoskrnl/mm/ARM3/virtual.c    [iso-8859-1] (original)
>> +++ trunk/reactos/ntoskrnl/mm/ARM3/virtual.c    [iso-8859-1] Tue Feb 17
>> 14:19:05 2015
>> @@ -1994,14 +1994,13 @@
>>          if (OnBoundary)
>>          {
>>              /* Is this PDE demand zero? */
>> -            PointerPde = MiAddressToPte(PointerPte);
>> +            PointerPde = MiPteToPde(PointerPte);
>>              if (PointerPde->u.Long != 0)
>>              {
>>                  /* It isn't -- is it valid? */
>>                  if (PointerPde->u.Hard.Valid == 0)
>>                  {
>>                      /* Nope, fault it in */
>> -                    PointerPte = MiPteToAddress(PointerPde);
>>                      MiMakeSystemAddressValid(PointerPte, Process);
>>                  }
>>              }
>> @@ -2009,13 +2008,13 @@
>>              {
>>                  /* The PTE was already valid, so move to the next one */
>>                  PointerPde++;
>> -                PointerPte = MiPteToAddress(PointerPde);
>> +                PointerPte = MiPdeToPte(PointerPde);
>>
>>                  /* Is the entire VAD committed? If not, fail */
>>                  if (!Vad->u.VadFlags.MemCommit) return FALSE;
>>
>> -                /* Everything is committed so far past the range, return
>> true */
>> -                if (PointerPte > LastPte) return TRUE;
>> +                /* New loop iteration with our new, on-boundary PTE. */
>> +                continue;
>>              }
>>          }
>>
>>
>>
>>
> 
> 
> 
> _______________________________________________
> Ros-dev mailing list
> Ros-dev at reactos.org
> http://www.reactos.org/mailman/listinfo/ros-dev
>