UefiCpuPkg/CpuSmm: Add perf-logging for time-consuming BSP procedures

The patch adds perf-logging for the following potential
time-consuming BSP procedures:
* PiCpuSmmEntry
  - SmmRelocateBases
  - ExecuteFirstSmiInit

* BSPHandler
  - SmmWaitForApArrival
  - PerformRemainingTasks
    * InitPaging
    * SetMemMapAttributes
    * SetUefiMemMapAttributes
    * SetPageTableAttributes
    * ConfigSmmCodeAccessCheck
    * SmmCpuFeaturesCompleteSmmReadyToLock

Signed-off-by: Ray Ni <ray.ni@intel.com>
Cc: Rahul Kumar <rahul1.kumar@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Jiaxin Wu <jiaxin.wu@intel.com>
Reviewed-by: Eric Dong <eric.dong@intel.com>

UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c

@@ -351,6 +351,8 @@ SmmWaitForApArrival (
   UINT32   DelayedCount;
   UINT32   BlockedCount;
 
+  PERF_FUNCTION_BEGIN ();
+
   DelayedCount = 0;
   BlockedCount = 0;
 
@@ -439,7 +441,7 @@ SmmWaitForApArrival (
     DEBUG ((DEBUG_INFO, "SmmWaitForApArrival: Delayed AP Count = %d, Blocked AP Count = %d\n", DelayedCount, BlockedCount));
   }
 
-  return;
+  PERF_FUNCTION_END ();
 }
 
 /**
@@ -577,6 +579,8 @@ BSPHandler (
   ASSERT (CpuIndex == mSmmMpSyncData->BspIndex);
   ApCount = 0;
 
+  PERF_FUNCTION_BEGIN ();
+
   //
   // Flag BSP's presence
   //
@@ -792,6 +796,8 @@ BSPHandler (
   *mSmmMpSyncData->Counter                  = 0;
   *mSmmMpSyncData->AllCpusInSync            = FALSE;
   mSmmMpSyncData->AllApArrivedWithException = FALSE;
+
+  PERF_FUNCTION_END ();
 }
 
 /**

UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.c

@@ -410,12 +410,15 @@ ExecuteFirstSmiInit (
 {
   UINTN  Index;
 
+  PERF_FUNCTION_BEGIN ();
+
   if (mSmmInitialized == NULL) {
     mSmmInitialized = (BOOLEAN *)AllocatePool (sizeof (BOOLEAN) * mMaxNumberOfCpus);
   }
 
   ASSERT (mSmmInitialized != NULL);
   if (mSmmInitialized == NULL) {
+    PERF_FUNCTION_END ();
     return;
   }
 
@@ -442,6 +445,8 @@ ExecuteFirstSmiInit (
     while (!(BOOLEAN)mSmmInitialized[Index]) {
     }
   }
+
+  PERF_FUNCTION_END ();
 }
 
 /**
@@ -463,6 +468,8 @@ SmmRelocateBases (
   UINTN                 Index;
   UINTN                 BspIndex;
 
+  PERF_FUNCTION_BEGIN ();
+
   //
   // Make sure the reserved size is large enough for procedure SmmInitTemplate.
   //
@@ -540,6 +547,7 @@ SmmRelocateBases (
   //
   CopyMem (CpuStatePtr, &BakBuf2, sizeof (BakBuf2));
   CopyMem (U8Ptr, BakBuf, sizeof (BakBuf));
+  PERF_FUNCTION_END ();
 }
 
 /**
@@ -617,6 +625,8 @@ PiCpuSmmEntry (
   GuidHob        = NULL;
   SmmBaseHobData = NULL;
 
+  PERF_FUNCTION_BEGIN ();
+
   //
   // Initialize address fixup
   //
@@ -1194,6 +1204,7 @@ PiCpuSmmEntry (
 
   DEBUG ((DEBUG_INFO, "SMM CPU Module exit from SMRAM with EFI_SUCCESS\n"));
 
+  PERF_FUNCTION_END ();
   return EFI_SUCCESS;
 }
 
@@ -1348,12 +1359,15 @@ ConfigSmmCodeAccessCheck (
   UINTN       Index;
   EFI_STATUS  Status;
 
+  PERF_FUNCTION_BEGIN ();
+
   //
   // Check to see if the Feature Control MSR is supported on this CPU
   //
   Index = gSmmCpuPrivate->SmmCoreEntryContext.CurrentlyExecutingCpu;
   if (!SmmCpuFeaturesIsSmmRegisterSupported (Index, SmmRegFeatureControl)) {
     mSmmCodeAccessCheckEnable = FALSE;
+    PERF_FUNCTION_END ();
     return;
   }
 
@@ -1363,6 +1377,7 @@ ConfigSmmCodeAccessCheck (
   //
   if ((AsmReadMsr64 (EFI_MSR_SMM_MCA_CAP) & SMM_CODE_ACCESS_CHK_BIT) == 0) {
     mSmmCodeAccessCheckEnable = FALSE;
+    PERF_FUNCTION_END ();
     return;
   }
 
@@ -1419,6 +1434,8 @@ ConfigSmmCodeAccessCheck (
       ReleaseSpinLock (mConfigSmmCodeAccessCheckLock);
     }
   }
+
+  PERF_FUNCTION_END ();
 }
 
 /**
@@ -1540,6 +1557,8 @@ PerformRemainingTasks (
   )
 {
   if (mSmmReadyToLock) {
+    PERF_FUNCTION_BEGIN ();
+
     //
     // Check if all Aps enter SMM. In Relaxed-AP Sync Mode, BSP will not wait for
     // all Aps arrive. However,PerformRemainingTasks() needs to wait all Aps arrive before calling
@@ -1587,12 +1606,20 @@ PerformRemainingTasks (
     //
     ConfigSmmCodeAccessCheck ();
 
+    //
+    // Measure performance of SmmCpuFeaturesCompleteSmmReadyToLock() from caller side
+    // as the implementation is provided by platform.
+    //
+    PERF_START (NULL, "SmmCompleteReadyToLock", NULL, 0);
     SmmCpuFeaturesCompleteSmmReadyToLock ();
+    PERF_END (NULL, "SmmCompleteReadyToLock", NULL, 0);
 
     //
     // Clean SMM ready to lock flag
     //
     mSmmReadyToLock = FALSE;
+
+    PERF_FUNCTION_END ();
   }
 }
 

UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.h

@@ -50,6 +50,7 @@ SPDX-License-Identifier: BSD-2-Clause-Patent
 #include <Library/SmmCpuFeaturesLib.h>
 #include <Library/PeCoffGetEntryPointLib.h>
 #include <Library/RegisterCpuFeaturesLib.h>
+#include <Library/PerformanceLib.h>
 
 #include <AcpiCpuData.h>
 #include <CpuHotPlugData.h>

UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.inf

@@ -97,6 +97,7 @@
   ReportStatusCodeLib
   SmmCpuFeaturesLib
   PeCoffGetEntryPointLib
+  PerformanceLib
 
 [Protocols]
   gEfiSmmAccess2ProtocolGuid               ## CONSUMES

UefiCpuPkg/PiSmmCpuDxeSmm/SmmCpuMemoryManagement.c

@@ -1,6 +1,6 @@
 /** @file
 
-Copyright (c) 2016 - 2019, Intel Corporation. All rights reserved.<BR>
+Copyright (c) 2016 - 2023, Intel Corporation. All rights reserved.<BR>
 SPDX-License-Identifier: BSD-2-Clause-Patent
 
 **/
@@ -1100,6 +1100,8 @@ SetMemMapAttributes (
     return;
   }
 
+  PERF_FUNCTION_BEGIN ();
+
   DEBUG ((DEBUG_INFO, "MemoryAttributesTable:\n"));
   DEBUG ((DEBUG_INFO, "  Version                   - 0x%08x\n", MemoryAttributesTable->Version));
   DEBUG ((DEBUG_INFO, "  NumberOfEntries           - 0x%08x\n", MemoryAttributesTable->NumberOfEntries));
@@ -1152,7 +1154,7 @@ SetMemMapAttributes (
   PatchSmmSaveStateMap ();
   PatchGdtIdtMap ();
 
-  return;
+  PERF_FUNCTION_END ();
 }
 
 /**
@@ -1454,6 +1456,8 @@ SetUefiMemMapAttributes (
   UINTN                  Index;
   EFI_MEMORY_DESCRIPTOR  *Entry;
 
+  PERF_FUNCTION_BEGIN ();
+
   DEBUG ((DEBUG_INFO, "SetUefiMemMapAttributes\n"));
 
   if (mUefiMemoryMap != NULL) {
@@ -1537,6 +1541,8 @@ SetUefiMemMapAttributes (
   //
   // Do not free mUefiMemoryAttributesTable, it will be checked in IsSmmCommBufferForbiddenAddress().
   //
+
+  PERF_FUNCTION_END ();
 }
 
 /**
@@ -1862,6 +1868,7 @@ SetPageTableAttributes (
     return;
   }
 
+  PERF_FUNCTION_BEGIN ();
   DEBUG ((DEBUG_INFO, "SetPageTableAttributes\n"));
 
   //
@@ -1900,5 +1907,5 @@ SetPageTableAttributes (
     EnableCet ();
   }
 
-  return;
+  PERF_FUNCTION_END ();
 }

UefiCpuPkg/PiSmmCpuDxeSmm/SmmProfile.c

@@ -575,6 +575,8 @@ InitPaging (
   IA32_CR4  Cr4;
   BOOLEAN   Enable5LevelPaging;
 
+  PERF_FUNCTION_BEGIN ();
+
   Cr4.UintN          = AsmReadCr4 ();
   Enable5LevelPaging = (BOOLEAN)(Cr4.Bits.LA57 == 1);
 
@@ -810,7 +812,7 @@ InitPaging (
   //
   mXdEnabled = TRUE;
 
-  return;
+  PERF_FUNCTION_END ();
 }
 
 /**