rocr/driver: add GetDeviceHandle to driver interface

This commit introduces a new GetDeviceHandle API to the driver
interface, allowing retrieval of the device handle for a
specific node.

- Implemented GetDeviceHandle in KfdDriver to fetch the AMD GPU
  device handle using hsaKmtGetAMDGPUDeviceHandle.
- Added a stub implementation of GetDeviceHandle in XdnaDriver
  that returns HSA_STATUS_ERROR.
- Modified GpuAgent::InitLibDrm to use driver().GetDeviceHandle
  instead of directly calling hsaKmtGetAMDGPUDeviceHandle.

Signed-off-by: Honglei Huang <Honglei1.Huang@amd.com>


[ROCm/ROCR-Runtime commit: 05b83e72d9]

projects/rocr-runtime/runtime/hsa-runtime/core/driver/kfd/amd_kfd_driver.cpp

@@ -574,6 +574,15 @@ hsa_status_t KfdDriver::SetTrapHandler(uint32_t node_id, const void* base, uint6
   return HSA_STATUS_SUCCESS;
 }
 
+hsa_status_t KfdDriver::GetDeviceHandle(uint32_t node_id, void** device_handle) const {
+  assert(device_handle);
+
+  if (HSAKMT_CALL(hsaKmtGetAMDGPUDeviceHandle(node_id, reinterpret_cast<HsaAMDGPUDeviceHandle*>(device_handle))) != HSAKMT_STATUS_SUCCESS)
+    return HSA_STATUS_ERROR;
+
+  return HSA_STATUS_SUCCESS;
+}
+
 hsa_status_t KfdDriver::IsModelEnabled(bool* enable) const {
   // AIE does not support streaming performance monitor.
   HSAKMT_STATUS status = HSAKMT_STATUS_ERROR;

projects/rocr-runtime/runtime/hsa-runtime/core/driver/xdna/amd_xdna_driver.cpp

@@ -877,5 +877,9 @@ hsa_status_t XdnaDriver::SetTrapHandler(uint32_t node_id, const void* base, uint
   return HSA_STATUS_ERROR;
 }
 
+hsa_status_t XdnaDriver::GetDeviceHandle(uint32_t node_id, void** device_handle) const {
+  return HSA_STATUS_ERROR;
+}
+
 } // namespace AMD
 } // namespace rocr

projects/rocr-runtime/runtime/hsa-runtime/core/inc/amd_kfd_driver.h

@@ -124,6 +124,7 @@ public:
                                 bool* is_spm_data_loss) const override;
   hsa_status_t SetTrapHandler(uint32_t node_id, const void* base, uint64_t base_size,
                               const void* buffer_base, uint64_t buffer_base_size) const override;
+  hsa_status_t GetDeviceHandle(uint32_t node_id, void** device_handle) const override;
 
   hsa_status_t OpenSMI(uint32_t node_id, int* fd) const override;
 

projects/rocr-runtime/runtime/hsa-runtime/core/inc/amd_xdna_driver.h

@@ -238,6 +238,7 @@ public:
                                 bool* is_spm_data_loss) const override;
   hsa_status_t SetTrapHandler(uint32_t node_id, const void* base, uint64_t base_size,
                               const void* buffer_base, uint64_t buffer_base_size) const override;
+  hsa_status_t GetDeviceHandle(uint32_t node_id, void** device_handle) const override;
 
   hsa_status_t IsModelEnabled(bool* enable) const override;
 

projects/rocr-runtime/runtime/hsa-runtime/core/inc/driver.h

@@ -269,6 +269,13 @@ public:
   virtual hsa_status_t SetTrapHandler(uint32_t node_id, const void* base, uint64_t base_size,
                                       const void* buffer_base, uint64_t buffer_base_size) const = 0;
 
+  /// @brief Gets the device handle for a specific node.
+  /// @param node_id Node ID of the agent
+  /// @param device_handle Device handle
+  /// @return HSA_STATUS_SUCCESS if the driver successfully returns the device
+  virtual hsa_status_t GetDeviceHandle(uint32_t node_id, void** device_handle) const = 0;
+
+
   /// @brief Check if the HSA KMT Model is enabled
   /// @param[out] enable True if the model is enabled, false otherwise
   virtual hsa_status_t IsModelEnabled(bool* enable) const = 0;

projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_gpu_agent.cpp

@@ -601,12 +601,12 @@ void GpuAgent::InitCacheList() {
 }
 
 void GpuAgent::InitLibDrm() {
-  HSAKMT_STATUS status;
+  hsa_status_t status;
 
   HsaAMDGPUDeviceHandle device_handle;
-  status = HSAKMT_CALL(hsaKmtGetAMDGPUDeviceHandle(node_id(), &device_handle));
+  status = driver().GetDeviceHandle(node_id(), &device_handle);
-  if (status != HSAKMT_STATUS_SUCCESS)
+  if (status != HSA_STATUS_SUCCESS)
-    throw AMD::hsa_exception(HSA_STATUS_ERROR,
+    throw AMD::hsa_exception(status,
                              "Agent creation failed.\nlibdrm get device handle failed.\n");
 
   ldrm_dev_ = (amdgpu_device_handle)device_handle;