Re: [edk2-devel] [Patch] MdeModulePkg DxeCore: Fix for missing MAT update

["Laszlo Ersek" <lersek@redhat.com>]

On 08/16/19 17:24, Gao, Liming wrote:
> Laszlo:
>
>> -----Original Message-----
>> From: devel@edk2.groups.io [mailto:devel@edk2.groups.io] On Behalf Of
>> Laszlo Ersek
>> Sent: Friday, August 16, 2019 11:18 PM
>> To: Gao, Liming <liming.gao@intel.com>; devel@edk2.groups.io; Kinney,
>> Michael D <michael.d.kinney@intel.com>
>> Cc: Mike Turner <miketur@microsoft.com>; Wang, Jian J
>> <jian.j.wang@intel.com>; Wu, Hao A <hao.a.wu@intel.com>; Bi, Dandan
>> <dandan.bi@intel.com>
>> Subject: Re: [edk2-devel] [Patch] MdeModulePkg DxeCore: Fix for
>> missing MAT update
>>
>> On 08/14/19 17:55, Gao, Liming wrote:
>>
>>> If Platform PEIM doesn't build HOB, DxeIpl will not build HOB,
>>
>> My reading of the code is the opposite. If the platform PEIM does not
>> build the HOB, then the DXE IPL PEIM will attempt to build the HOB,
>> from the UEFI variable.
>>
>> At commit caa7d3a896f6, in file
>> "MdeModulePkg/Core/DxeIplPeim/DxeLoad.c", function DxeLoadCore(), we
>> have:
>>
>>    363    if (GetFirstGuidHob ((CONST EFI_GUID *)&gEfiMemoryTypeInformationGuid) == NULL) {
>>    364      //
>>    365      // Don't build GuidHob if GuidHob has been installed.
>>    366      //
>>    367      Status = PeiServicesLocatePpi (
>>    368                 &gEfiPeiReadOnlyVariable2PpiGuid,
>>    369                 0,
>>    370                 NULL,
>>    371                 (VOID **)&Variable
>>    372                 );
>>    373      if (!EFI_ERROR (Status)) {
>>    374        DataSize = sizeof (MemoryData);
>>    375        Status = Variable->GetVariable (
>>    376                             Variable,
>>    377                             EFI_MEMORY_TYPE_INFORMATION_VARIABLE_NAME,
>>    378                             &gEfiMemoryTypeInformationGuid,
>>    379                             NULL,
>>    380                             &DataSize,
>>    381                             &MemoryData
>>    382                             );
>>    383        if (!EFI_ERROR (Status) && ValidateMemoryTypeInfoVariable(MemoryData, DataSize)) {
>
> Only when this variable exists, Hob will be built. But, if no PEIM
> builds Hob, BDS will not set the variable.
> So, there is still no HOB.

So how is a platform supposed to enable this feature?

If PlatformPei never builds the HOB, the variable will never be created,
so the DXE IPL PEIM will also not build the HOB, ever.

If PlatformPei builds the HOB with static data, then BDS will set
(update) the variable, yes, but the DXE IPL PEIM will ignore the
variable, because PlatformPei already built the HOB.

So... Is PlatformPei supposed to use the Variable PPI, check if the
variable exists, and create the static HOB only if the variable is
absent?

... Ugh, wait. I've actually implemented this for OVMF almost 2 years
ago! And the answer to my question is "yes":

  https://bugzilla.tianocore.org/show_bug.cgi?id=386
  https://lists.01.org/pipermail/edk2-devel/2017-November/018312.html

See the OnReadOnlyVariable2Available() function:

+  //
+  // Check if the "MemoryTypeInformation" variable exists, in the
+  // gEfiMemoryTypeInformationGuid namespace.
+  //
+  ReadOnlyVariable2 = Ppi;
+  DataSize = 0;
+  Status = ReadOnlyVariable2->GetVariable (
+                                ReadOnlyVariable2,
+                                EFI_MEMORY_TYPE_INFORMATION_VARIABLE_NAME,
+                                &gEfiMemoryTypeInformationGuid,
+                                NULL,
+                                &DataSize,
+                                NULL
+                                );
+  if (Status == EFI_BUFFER_TOO_SMALL) {
+    //
+    // The variable exists; the DXE IPL PEIM will build the HOB from it.
+    //
+    return EFI_SUCCESS;
+  }
+  //
+  // Install the default memory type information HOB.
+  //
+  BuildMemTypeInfoHob ();

Apologies for forgetting about this; you are right.

Too bad my work for TianoCore#386 was rejected. :(

Thanks
Laszlo