feat: Log C2C init failure to BMC

Log an error POST code to BMC for each C2C link training failure. Also send the socket, controller number, and C2C status code as an OEM string to BMC. Signed-off-by: Dat Mach <dmach@nvidia.com> Tested-by: Jake Garver <jake@nvidia.com> Reviewed-by: Jake Garver <jake@nvidia.com>
NVIDIA · Feb 11, 2025 · b2fe155 · b2fe155
1 parent 61583ad
commit b2fe155
Show file tree

Hide file tree

Showing 4 changed files with 28 additions and 3 deletions.
diff --git a/Silicon/NVIDIA/Drivers/PcieControllerDxe/PcieControllerDxe.c b/Silicon/NVIDIA/Drivers/PcieControllerDxe/PcieControllerDxe.c
@@ -28,7 +28,9 @@
 #include <Library/PcdLib.h>
 #include <Library/PciHostBridgeLib.h>
 #include <Library/PlatformResourceLib.h>
+#include <Library/PrintLib.h>
 #include <Library/TegraPlatformInfoLib.h>
+#include <Library/ReportStatusCodeLib.h>
 #include <Library/SortLib.h>
 #include <Library/TimerLib.h>
 #include <Library/UefiBootServicesTableLib.h>
@@ -45,6 +47,9 @@
 #include <TH500/TH500Definitions.h>
 #include <TH500/TH500MB1Configuration.h>
 
+#include <NVIDIAStatusCodes.h>
+#include <OemStatusCodes.h>
+
 #include "PcieControllerConfigGPU.h"
 #include "PcieControllerPrivate.h"
 #include <Protocol/PciPlatform.h>
@@ -534,6 +539,7 @@ InitializeController (
   PCI_CAPABILITY_PCIEXP  *PciExpCap = NULL;
   UINT8                  C2cStatus;
   VOID                   *Hob;
+  CHAR8                  OemDesc[sizeof (OEM_EC_DESC_C2C_INIT_FAILULE)];
 
   TEGRABL_EARLY_BOOT_VARIABLES  *Mb1Config = NULL;
   UINTN                         ChipId;
@@ -638,6 +644,20 @@ InitializeController (
           Private->C2cInitSuccessful = TRUE;
         } else {
           DEBUG ((DEBUG_ERROR, "%a: C2C link training failed with error code: 0x%x\r\n", __FUNCTION__, C2cStatus));
+          AsciiSPrint (
+            OemDesc,
+            sizeof (OemDesc),
+            OEM_EC_DESC_C2C_INIT_FAILULE,
+            Private->SocketId,
+            Private->CtrlId,
+            C2cStatus
+            );
+          REPORT_STATUS_CODE_WITH_EXTENDED_DATA (
+            EFI_ERROR_CODE | EFI_ERROR_MAJOR,
+            EFI_CLASS_NV_FIRMWARE | EFI_NV_FW_UEFI_EC_C2C_INIT_FAILED,
+            OemDesc,
+            AsciiStrSize (OemDesc)
+            );
         }
       }
 

diff --git a/Silicon/NVIDIA/Drivers/PcieControllerDxe/PcieControllerDxe.inf b/Silicon/NVIDIA/Drivers/PcieControllerDxe/PcieControllerDxe.inf
@@ -2,7 +2,7 @@
 #
 #  PCIe Controller Driver
 #
-#  SPDX-FileCopyrightText: Copyright (c) 2019-2024 NVIDIA CORPORATION. All rights reserved.
+#  SPDX-FileCopyrightText: Copyright (c) 2019-2025 NVIDIA CORPORATION. All rights reserved.
 #
 #  SPDX-License-Identifier: BSD-2-Clause-Patent
 #
@@ -50,6 +50,7 @@
   DxeServicesTableLib
   DevicePathLib
   PlatformResourceLib
+  ReportStatusCodeLib
 
 [Pcd]
   gNVIDIATokenSpaceGuid.PcdBPMPPCIeControllerEnable

diff --git a/Silicon/NVIDIA/Include/NVIDIAStatusCodes.h b/Silicon/NVIDIA/Include/NVIDIAStatusCodes.h
@@ -1,6 +1,6 @@
 /** @file
 *
-*  SPDX-FileCopyrightText: Copyright (c) 2023-2024 NVIDIA CORPORATION. All rights reserved.
+*  SPDX-FileCopyrightText: Copyright (c) 2023-2025 NVIDIA CORPORATION. All rights reserved.
 *
 *  SPDX-License-Identifier: BSD-2-Clause-Patent
 *
@@ -178,6 +178,7 @@
 #define EFI_NV_FW_UEFI_EC_TPM_PPI_EXECUTE                          (EFI_NV_FIRMWARE_UEFI | 0x00000015)
 #define EFI_NV_FW_UEFI_EC_TPM_CLEAR_FAILED                         (EFI_NV_FIRMWARE_UEFI | 0x00000016)
 #define EFI_NV_FW_UEFI_EC_SECURE_BOOT_FAILED                       (EFI_NV_FIRMWARE_UEFI | 0x00000017)
+#define EFI_NV_FW_UEFI_EC_C2C_INIT_FAILED                          (EFI_NV_FIRMWARE_UEFI | 0x00000018)
 ///@}
 
 ///

diff --git a/Silicon/NVIDIA/Include/Server/OemStatusCodes.h b/Silicon/NVIDIA/Include/Server/OemStatusCodes.h
@@ -1,6 +1,6 @@
 /** @file
 *
-*  SPDX-FileCopyrightText: Copyright (c) 2023-2024 NVIDIA CORPORATION. All rights reserved.
+*  SPDX-FileCopyrightText: Copyright (c) 2023-2025 NVIDIA CORPORATION. All rights reserved.
 *
 *  SPDX-License-Identifier: BSD-2-Clause-Patent
 *
@@ -101,6 +101,9 @@
 #define OEM_EC_DESC_SECURE_BOOT_FAILURE \
         "Secure boot failure "
 
+#define OEM_EC_DESC_C2C_INIT_FAILULE \
+        "S%uC%u C2C init failed - 0x%02X"
+
 ///@}
 
 #endif