Fix unit issue in policy feature (#78)

1. For temperature the unit in milli Celsius
2. For power the unit in microwatts.
3. Fix second register call to rdcd doesn't functional because start flag

Co-authored-by: Chao Fei <chao.fei@amd.com>

[ROCm/rdc commit: bd7d7c99c1]

projects/rdc/include/rdc/rdc.h

@@ -593,7 +593,7 @@ typedef enum {
   RDC_POLICY_COND_FIRST = 0,
   RDC_POLICY_COND_MAX_PAGE_RETRIED = RDC_POLICY_COND_FIRST,  //!< Max number of page retired
   RDC_POLICY_COND_THERMAL,  //!< Temperature threshold, millidegree Celsius
-  RDC_POLICY_COND_POWER,    //!< Power threshold, unit milliwatt
+  RDC_POLICY_COND_POWER,    //!< Power threshold, unit microwatt
   RDC_POLICY_COND_LAST = RDC_POLICY_COND_POWER,
   RDC_POLICY_COND_MAX
 } rdc_policy_condition_type_t;

projects/rdc/rdci/src/RdciPolicySubSystem.cc

@@ -185,8 +185,18 @@ int rdc_policy_callback(rdc_policy_callback_response_t* userData) {
     return 0;
   }
 
+  int64_t value = userData->value;
+  int64_t threshold = userData->condition.value;
+  if (userData->condition.type == RDC_POLICY_COND_THERMAL) {
+    value /= 1000;
+    threshold /= 1000;
+  } else if (userData->condition.type == RDC_POLICY_COND_POWER) {
+    value /= 1000000;
+    threshold /= 1000000;
+  }
+
   std::cout << "A " << condition_type_to_str(userData->condition.type) << " exceeds the threshold "
-            << userData->condition.value << " with the value " << userData->value << std::endl;
+            << threshold << " with the value " << value << std::endl;
   last_time = now;  // update the last time
   return 0;
 }
@@ -216,10 +226,10 @@ void RdciPolicySubSystem::process() {
             policy.condition = {RDC_POLICY_COND_MAX_PAGE_RETRIED, option.second};
             break;
           case POLICY_OPT_TEMP:
-            policy.condition = {RDC_POLICY_COND_THERMAL, option.second};
+            policy.condition = {RDC_POLICY_COND_THERMAL, option.second * 1000};
             break;
           case POLICY_OPT_POWER:
-            policy.condition = {RDC_POLICY_COND_POWER, option.second};
+            policy.condition = {RDC_POLICY_COND_POWER, option.second * 1000000};
             break;
           case POLICY_OPT_ACTION:
             if (option.second == 0) {
@@ -264,15 +274,21 @@ void RdciPolicySubSystem::process() {
           if (policies[i].condition.type == RDC_POLICY_COND_MAX_PAGE_RETRIED) {
             std::cout << "| Page Retirement\t "
                       << "|   " << policies[i].condition.value;
+            if (policies[i].condition.value < 100) {
+              std::cout << "\t";
+            }
           } else if (policies[i].condition.type == RDC_POLICY_COND_THERMAL) {
             std::cout << "| Temperature Limit\t "
-                      << "|   " << policies[i].condition.value;
+                      << "|   " << policies[i].condition.value / 1000;
+            if (policies[i].condition.value / 1000 < 100) {
+              std::cout << "\t";
+            }
           } else if (policies[i].condition.type == RDC_POLICY_COND_POWER) {
             std::cout << "| Power Limit     \t "
-                      << "|   " << policies[i].condition.value;
-          }
-          if (policies[i].condition.value < 100) {
-            std::cout << "\t";
+                      << "|   " << policies[i].condition.value / 1000000;
+            if (policies[i].condition.value / 1000000 < 100) {
+              std::cout << "\t";
+            }
           }
           if (policies[i].action == 0) {
             std::cout << "\t\t| Notify\t\t|\n";

projects/rdc/server/src/rdc_api_service.cc

@@ -886,11 +886,13 @@ int RdcAPIServiceImpl::PolicyCallback(rdc_policy_callback_response_t* userData)
     return ::grpc::Status(::grpc::StatusCode::INTERNAL, "Empty contents");
   }
 
-  policy_thread_context* data = new policy_thread_context;
-  data->mutex = PTHREAD_MUTEX_INITIALIZER;
-  data->cond = PTHREAD_COND_INITIALIZER;
-  data->start = true;
-  policy_threads_.insert(std::make_pair(request->group_id(), data));
+  if (policy_threads_.size() == 0) {
+    policy_thread_context* data = new policy_thread_context;
+    data->mutex = PTHREAD_MUTEX_INITIALIZER;
+    data->cond = PTHREAD_COND_INITIALIZER;
+    data->start = true;
+    policy_threads_.insert(std::make_pair(request->group_id(), data));
+  }
 
   auto updater = std::async(std::launch::async, [this, request, writer]() {
     rdc_status_t result = rdc_policy_register(rdc_handle_, request->group_id(), PolicyCallback);
@@ -898,6 +900,7 @@ int RdcAPIServiceImpl::PolicyCallback(rdc_policy_callback_response_t* userData)
       auto it = policy_threads_.find(request->group_id());
       if (it != policy_threads_.end()) {
         policy_thread_context* ctx = it->second;
+        ctx->start = true;
         while (ctx->start) {
           struct timespec ts;
           clock_gettime(CLOCK_REALTIME, &ts);