From 11541cc28343ea7aed2d4ea6a03a7a024d12671a Mon Sep 17 00:00:00 2001
From: David Yat Sin <David.YatSin@amd.com>
Date: Tue, 11 Apr 2023 21:32:08 +0000
Subject: [PATCH] Add env var to override SRAM ECC

Add HSA_ENABLE_SRAMECC environment variable that can be used to
override SRAM ECC mode reported by KFD

Change-Id: I2b95511820a2d3d146a76b03070659c0695b61fd


[ROCm/ROCR-Runtime commit: a180c9ee78e729a2cd5225392504cb2db52d4b9d]
---
 .../hsa-runtime/core/runtime/amd_gpu_agent.cpp     | 14 ++++++++++++--
 .../runtime/hsa-runtime/core/util/flag.h           |  9 +++++++++
 2 files changed, 21 insertions(+), 2 deletions(-)

diff --git a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_gpu_agent.cpp b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_gpu_agent.cpp
index 00547d16d8..bd07fadab4 100644
--- a/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_gpu_agent.cpp
+++ b/projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_gpu_agent.cpp
@@ -124,8 +124,18 @@ GpuAgent::GpuAgent(HSAuint32 node, const HsaNodeProperties& node_props, bool xna
 
   rocr::core::IsaFeature sramecc = rocr::core::IsaFeature::Unsupported;
   if (isa_base->IsSrameccSupported()) {
-    sramecc = node_props.Capability.ui32.SRAM_EDCSupport == 1 ? core::IsaFeature::Enabled
-                                                              : core::IsaFeature::Disabled;
+    switch (core::Runtime::runtime_singleton_->flag().sramecc_enable()) {
+      case Flag::SRAMECC_DISABLED:
+        sramecc = core::IsaFeature::Disabled;
+        break;
+      case Flag::SRAMECC_ENABLED:
+        sramecc = core::IsaFeature::Enabled;
+        break;
+      case Flag::SRAMECC_DEFAULT:
+        sramecc = node_props.Capability.ui32.SRAM_EDCSupport == 1 ? core::IsaFeature::Enabled
+                                                                  : core::IsaFeature::Disabled;
+        break;
+    }
   }
 
   rocr::core::IsaFeature xnack = rocr::core::IsaFeature::Unsupported;
diff --git a/projects/rocr-runtime/runtime/hsa-runtime/core/util/flag.h b/projects/rocr-runtime/runtime/hsa-runtime/core/util/flag.h
index 5c5e8f3101..2f7f726902 100644
--- a/projects/rocr-runtime/runtime/hsa-runtime/core/util/flag.h
+++ b/projects/rocr-runtime/runtime/hsa-runtime/core/util/flag.h
@@ -57,6 +57,7 @@ namespace rocr {
 class Flag {
  public:
   enum SDMA_OVERRIDE { SDMA_DISABLE, SDMA_ENABLE, SDMA_DEFAULT };
+  enum SRAMECC_ENABLE { SRAMECC_DISABLED, SRAMECC_ENABLED, SRAMECC_DEFAULT };
 
   // The values are meaningful and chosen to satisfy the thunk API.
   enum XNACK_REQUEST { XNACK_DISABLE = 0, XNACK_ENABLE = 1, XNACK_UNCHANGED = 2 };
@@ -175,6 +176,10 @@ class Flag {
     var = os::GetEnvVar("HSA_SVM_PROFILE");
     svm_profile_ = var;
 
+    var = os::GetEnvVar("HSA_ENABLE_SRAMECC");
+    sramecc_enable_ =
+        (var == "0") ? SRAMECC_DISABLED : ((var == "1") ? SRAMECC_ENABLED : SRAMECC_DEFAULT);
+
     var = os::GetEnvVar("HSA_IMAGE_PRINT_SRD");
     image_print_srd_ = (var == "1") ? true : false;
 
@@ -269,6 +274,8 @@ class Flag {
 
   const std::string& svm_profile() const { return svm_profile_; }
 
+  SRAMECC_ENABLE sramecc_enable() const { return sramecc_enable_; }
+
  private:
   bool check_flat_scratch_;
   bool enable_vm_fault_message_;
@@ -312,6 +319,8 @@ class Flag {
   // Indicates user preference for Xnack state.
   XNACK_REQUEST xnack_;
 
+  SRAMECC_ENABLE sramecc_enable_;
+
   // Map GPU index post RVD to its default cu mask.
   std::map<uint32_t, std::vector<uint32_t>> cu_mask_;