Re: [edk2-devel] [PATCH 6/6] UefiCpuPkg: Avoid assuming only one smmbasehob

["duntan" <dun.tan@intel.com>]

Updated in original message.

Thanks,
Dun

-----Original Message-----
From: Ni, Ray <ray.ni@intel.com> 
Sent: Wednesday, December 6, 2023 6:15 PM
To: Tan, Dun <dun.tan@intel.com>; devel@edk2.groups.io
Cc: Dong, Eric <eric.dong@intel.com>; Kumar, Rahul R <rahul.r.kumar@intel.com>; Gerd Hoffmann <kraxel@redhat.com>
Subject: RE: [PATCH 6/6] UefiCpuPkg: Avoid assuming only one smmbasehob

> +EFI_STATUS
> +GetSmBaseFromSmmBaseHob (
> +  IN  EFI_HOB_GUID_TYPE  *FirstSmmBaseGuidHob,
> +  IN  UINTN              MaxNumberOfCpus,
> +  OUT UINTN              **SmBaseBufferPointer
> +  )

1. It's a bit strange that caller should locate the first GuidHob.
Can you update the existing code as follows:
mCpuHotPlugData.SmBase = GetSmBase(mMaxNumberOfCpus); if (mCpuHotPlugData.SmBase != NULL) {
  mSmmRelocated = TRUE;
}

Dun: Ok, will change code to this.

> +{
> +  UINTN              HobCount;
> +  EFI_HOB_GUID_TYPE  *GuidHob;
> +  SMM_BASE_HOB_DATA  *SmmBaseHobData;
> +  UINTN              NumberOfProcessors;
> +  SMM_BASE_HOB_DATA  **SmBaseHobPointerBuffer;
> +  UINTN              *SmBaseBuffer;
> +  UINTN              Index;
> +  UINTN              SortBuffer;
> +  UINTN              ProcessorIndex;
> +  UINT64             PrevProcessorIndex;
> +
> +  SmmBaseHobData     = NULL;
> +  Index              = 0;
> +  ProcessorIndex     = 0;
> +  PrevProcessorIndex = 0;
> +  HobCount           = 0;
> +  NumberOfProcessors = 0;
> +  GuidHob            = FirstSmmBaseGuidHob;
> +
> +  while (GuidHob != NULL) {
> +    HobCount++;
> +    SmmBaseHobData      = GET_GUID_HOB_DATA (GuidHob);
> +    NumberOfProcessors += SmmBaseHobData->NumberOfProcessors;
> +    GuidHob             = GetNextGuidHob (&gSmmBaseHobGuid,
> GET_NEXT_HOB (GuidHob));

2. We could break the while-loop when NumberOfProcessors equals to the value we retrieved from MpInfo2Hob. Right?
This can speed up the code when there are lots of HOBs after the last SmmBaseHob instance.

Dun: If the code flow break before finding all potential SmmBaseHob instance, there may be more SmmBaseHob instance covering NumberOfProcessors more than the expected value. The code is to catch this case. Do you think we should also catch this?

> +  }
> +
> +  ASSERT (NumberOfProcessors == MaxNumberOfCpus);

3. ASSERT may fail when HotPlug is TRUE?

Dun: If HotPlug, I think the SmBase count should be PcdCpuMaxLogicalProcessorNumber instead of the NumberOfProcessors extracted from MpInfo2Hob?


> +
> +  SmBaseHobPointerBuffer = AllocatePool (sizeof (SMM_BASE_HOB_DATA *)
> * HobCount);

4. SmBaseHobPointerBuffer -> SmBaseHobs

Dun: will change the naming.

> +  for (Index = 0; Index < HobCount; Index++) {
> +    //
> +    // Make sure no overlap and no gap in the CPU range covered by 
> + each
> HOB
> +    //
> +    ASSERT (SmBaseHobPointerBuffer[Index]->ProcessorIndex ==
> PrevProcessorIndex);

5. similarly, can you move "PrevProcessorIndex" assignment to just above "for"?

Dun: Will change the code

> +
> +    //
> +    // Cache each SmBase in order.
> +    //
> +    if (sizeof (UINTN) == sizeof (UINT64)) {
> +      CopyMem (
> +        SmBaseBuffer + PrevProcessorIndex,
> +        &SmBaseHobPointerBuffer[Index]->SmBase,
> +        sizeof (UINT64) *
> SmBaseHobPointerBuffer[Index]->NumberOfProcessors
> +        );
> +    } else {
> +      for (ProcessorIndex = 0; ProcessorIndex <
> SmBaseHobPointerBuffer[Index]->NumberOfProcessors; ProcessorIndex++) {
> +        SmBaseBuffer[PrevProcessorIndex + ProcessorIndex] =
> (UINTN)SmBaseHobPointerBuffer[Index]->SmBase[ProcessorIndex];
> +      }
> +    }


6. I don't like the "if-else" above. Can you just change SmBaseBuffer to UINT64 *?
Or, you always use for-loop to copy SmBase value for each cpu.

Dun: Ok, will always use for-loop to copy SmBase value for each cpu.

> +
> +    PrevProcessorIndex +=
> SmBaseHobPointerBuffer[Index]->NumberOfProcessors;
> +  }
> +
> +  FreePool (SmBaseHobPointerBuffer);
> +
> +  *SmBaseBufferPointer = SmBaseBuffer;

7. Similarly, how about return SmBaseBuffer?

Dun: Ok, will change the code


-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#112148): https://edk2.groups.io/g/devel/message/112148
Mute This Topic: https://groups.io/mt/102987142/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-