Implement rdc_device_get_component_version API related code

Implement an API to obtain the version information of the rdc calling component.
See rdc_component_t for details on available components.
It can be expanded later if necessary.

Change-Id: I03b48f774179c52c57b606704283add74ca39a02
Signed-off-by: Chen Gong <curry.gong@amd.com>


[ROCm/rdc commit: 5a3fd9fbc1]

projects/rdc/include/rdc/rdc.h

@@ -141,6 +141,11 @@ typedef enum { INTEGER = 0, DOUBLE, STRING, BLOB } rdc_field_type_t;
  */
 #define RDC_MAX_NUM_FIELD_GROUPS 64
 
+/**
+ * @brief The max string length occupied by version information
+ */
+#define RDC_MAX_VERSION_STR_LENGTH 60
+
 /**
  * These enums are used to specify a particular field to be retrieved.
  */
@@ -343,6 +348,13 @@ typedef struct {
   char device_name[RDC_MAX_STR_LENGTH];  //!< Name of the device.
 } rdc_device_attributes_t;
 
+/**
+ * @brief Store version information for each component
+ */
+typedef struct {
+  char version[RDC_MAX_VERSION_STR_LENGTH];
+} rdc_component_version_t;
+
 /**
  * @brief The structure to store the group info
  */
@@ -476,6 +488,14 @@ typedef enum {
   RDC_DIAG_TEST_LAST = RDC_DIAG_GPU_PARAMETERS
 } rdc_diag_test_cases_t;
 
+/**
+ * @brief Type of Components
+ */
+typedef enum {
+  RDC_AMDMSI_COMPONENT
+  //If needed later, add them one by one
+} rdc_component_t;
+
 /**
  * @brief The maximum test cases to run
  */
@@ -750,6 +770,21 @@ rdc_status_t rdc_device_get_all(rdc_handle_t p_rdc_handle,
 rdc_status_t rdc_device_get_attributes(rdc_handle_t p_rdc_handle, uint32_t gpu_index,
                                        rdc_device_attributes_t* p_rdc_attr);
 
+/**
+ *  @brief Get version information of components used by rdc.
+ *
+ *  @details Given a component type, return its version information.
+ *
+ *  @param[in] p_rdc_handle The RDC handler.
+ *
+ *  @param[in] component Type of Components. See rdc_component_t definition for details.
+ *
+ *  @param[out] p_rdc_compv Version information of the corresponding component.
+ *
+ *  @retval ::RDC_ST_OK is returned upon successful call.
+ */
+rdc_status_t rdc_device_get_component_version(rdc_handle_t p_rdc_handle, rdc_component_t component, rdc_component_version_t* p_rdc_compv);
+
 /**
  *  @brief Create a group contains multiple GPUs
  *

projects/rdc/include/rdc_lib/RdcHandler.h

@@ -44,6 +44,7 @@ class RdcHandler {
                                           uint32_t* count) = 0;
   virtual rdc_status_t rdc_device_get_attributes(uint32_t gpu_index,
                                                  rdc_device_attributes_t* p_rdc_attr) = 0;
+  virtual rdc_status_t rdc_device_get_component_version(rdc_component_t component, rdc_component_version_t* p_rdc_compv) = 0;
 
   // Group API
   virtual rdc_status_t rdc_group_gpu_create(rdc_group_type_t type, const char* group_name,

projects/rdc/include/rdc_lib/impl/RdcEmbeddedHandler.h

@@ -51,6 +51,7 @@ class RdcEmbeddedHandler final : public RdcHandler {
                                   uint32_t* count) override;
   rdc_status_t rdc_device_get_attributes(uint32_t gpu_index,
                                          rdc_device_attributes_t* p_rdc_attr) override;
+  rdc_status_t rdc_device_get_component_version(rdc_component_t component, rdc_component_version_t* p_rdc_compv) override;
 
   // Group API
   rdc_status_t rdc_group_gpu_create(rdc_group_type_t type, const char* group_name,

projects/rdc/include/rdc_lib/impl/RdcStandaloneHandler.h

@@ -46,6 +46,7 @@ class RdcStandaloneHandler : public RdcHandler {
                                   uint32_t* count) override;
   rdc_status_t rdc_device_get_attributes(uint32_t gpu_index,
                                          rdc_device_attributes_t* p_rdc_attr) override;
+  rdc_status_t rdc_device_get_component_version(rdc_component_t component, rdc_component_version_t* p_rdc_compv) override;
 
   // Group RdcAPI
   rdc_status_t rdc_group_gpu_create(rdc_group_type_t type, const char* group_name,

projects/rdc/rdc_libs/bootstrap/src/RdcBootStrap.cc

@@ -182,6 +182,15 @@ rdc_status_t rdc_device_get_attributes(rdc_handle_t p_rdc_handle, uint32_t gpu_i
       ->rdc_device_get_attributes(gpu_index, p_rdc_attr);
 }
 
+rdc_status_t rdc_device_get_component_version(rdc_handle_t p_rdc_handle, rdc_component_t component, rdc_component_version_t* p_rdc_compv) {
+  if (!p_rdc_handle || !p_rdc_compv) {
+    return RDC_ST_INVALID_HANDLER;
+  }
+
+  return static_cast<amd::rdc::RdcHandler*>(p_rdc_handle)
+      ->rdc_device_get_component_version(component, p_rdc_compv);
+}
+
 rdc_status_t rdc_group_field_create(rdc_handle_t p_rdc_handle, uint32_t num_field_ids,
                                     rdc_field_t* field_ids, const char* field_group_name,
                                     rdc_field_grp_t* rdc_field_group_id) {

projects/rdc/rdc_libs/rdc/src/RdcEmbeddedHandler.cc

@@ -194,6 +194,29 @@ rdc_status_t RdcEmbeddedHandler::rdc_device_get_attributes(uint32_t gpu_index,
   return status;
 }
 
+rdc_status_t RdcEmbeddedHandler::rdc_device_get_component_version(rdc_component_t component, rdc_component_version_t* p_rdc_compv) {
+  if (!p_rdc_compv) {
+    return RDC_ST_BAD_PARAMETER;
+  }
+
+  if (component == RDC_AMDMSI_COMPONENT) {
+    amdsmi_status_t ret;
+    amdsmi_version_t ver = {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, nullptr};
+
+    ret = amdsmi_get_lib_version(&ver);
+
+    if (ret != AMDSMI_STATUS_SUCCESS) {
+      RDC_LOG(RDC_ERROR, "Failed to obtain the version of the server's amd-smi library. reason: " << (ret == AMDSMI_STATUS_INVAL ? "Invalid parameters" : "unknown"));
+      return RDC_ST_MSI_ERROR;
+    }
+
+    strncpy_with_null(p_rdc_compv->version, ver.build, RDC_MAX_VERSION_STR_LENGTH);
+    return RDC_ST_OK;
+  } else {
+    return RDC_ST_BAD_PARAMETER;
+  }
+}
+
 // Group API
 rdc_status_t RdcEmbeddedHandler::rdc_group_gpu_create(rdc_group_type_t type, const char* group_name,
                                                       rdc_gpu_group_t* p_rdc_group_id) {

projects/rdc/rdc_libs/rdc_client/src/RdcStandaloneHandler.cc

@@ -243,6 +243,25 @@ rdc_status_t RdcStandaloneHandler::rdc_device_get_attributes(uint32_t gpu_index,
   return RDC_ST_OK;
 }
 
+rdc_status_t RdcStandaloneHandler::rdc_device_get_component_version(rdc_component_t component, rdc_component_version_t* p_rdc_compv) {
+
+  if (!p_rdc_compv) {
+    return RDC_ST_BAD_PARAMETER;
+  }
+
+  ::rdc::GetComponentVersionRequest request;
+  ::rdc::GetComponentVersionResponse reply;
+  ::grpc::ClientContext context;
+
+  request.set_component_index(component);
+  ::grpc::Status status = stub_->GetComponentVersion(&context, request, &reply);
+  rdc_status_t err_status = error_handle(status, reply.status());
+  if (err_status != RDC_ST_OK) return err_status;
+
+  strncpy_with_null(p_rdc_compv->version, reply.version().c_str(), RDC_MAX_VERSION_STR_LENGTH);
+  return RDC_ST_OK;
+}
+
 // Group RdcAPI
 rdc_status_t RdcStandaloneHandler::rdc_group_gpu_create(rdc_group_type_t type,
                                                         const char* group_name,

projects/rdc/server/include/rdc/rdc_api_service.h

@@ -42,6 +42,9 @@ class RdcAPIServiceImpl final : public ::rdc::RdcAPI::Service {
                                      const ::rdc::GetDeviceAttributesRequest* request,
                                      ::rdc::GetDeviceAttributesResponse* reply) override;
 
+  ::grpc::Status GetComponentVersion(::grpc::ServerContext* context,  const ::rdc::GetComponentVersionRequest* request,
+                               ::rdc::GetComponentVersionResponse* reply) override;
+
   ::grpc::Status CreateGpuGroup(::grpc::ServerContext* context,
                                 const ::rdc::CreateGpuGroupRequest* request,
                                 ::rdc::CreateGpuGroupResponse* reply) override;

projects/rdc/server/src/rdc_api_service.cc

@@ -103,6 +103,23 @@ RdcAPIServiceImpl::~RdcAPIServiceImpl() {
   return ::grpc::Status::OK;
 }
 
+::grpc::Status RdcAPIServiceImpl::GetComponentVersion(::grpc::ServerContext* context,
+                                                const ::rdc::GetComponentVersionRequest* request,
+                                                ::rdc::GetComponentVersionResponse* reply) {
+  (void)(context);
+  if (!reply) {
+    return ::grpc::Status(::grpc::StatusCode::INTERNAL, "Empty reply");
+  }
+
+  rdc_component_t component = static_cast<rdc_component_t>(request->component_index());
+  rdc_component_version_t compv;
+  rdc_status_t result = rdc_device_get_component_version(rdc_handle_, component, &compv);
+
+  reply->set_version(compv.version);
+  reply->set_status(result);
+  return ::grpc::Status::OK;
+}
+
 ::grpc::Status RdcAPIServiceImpl::CreateGpuGroup(::grpc::ServerContext* context,
                                                  const ::rdc::CreateGpuGroupRequest* request,
                                                  ::rdc::CreateGpuGroupResponse* reply) {