[ROCM-SMI] [SWDEV-569731] rsmi tests failing on Frequency/Power/GpuMetrics ReadOnly Fix (#2303)

* Updated unsupported metric version file for rocm_smi_tests Frequency/Power/GpuMetrics ReadOnly tests

Signed-off-by: yalmusaf_amdeng <Yazen.ALMusaffar@amd.com>

projects/rocm-smi-lib/tests/rocm_smi_test/functional/frequencies_read.cc

@@ -133,10 +133,9 @@ void TestFrequenciesRead::Run(void) {
           return;
         }
 
-        // special driver issue, shouldn't normally occur
         if (err == RSMI_STATUS_UNEXPECTED_DATA) {
-          std::cerr << "WARN: Clock file [" << FreqEnumToStr(t) << "] exists on device [" << i << "] but empty!" << std::endl;
-          std::cerr << "      Likely a driver issue!" << std::endl;
+          std::cerr << "\t**gpu metric file version unsupported: "
+                    << name << " on device [" << i << "]" << std::endl;
           return;
         }
 
@@ -162,10 +161,18 @@ void TestFrequenciesRead::Run(void) {
       err = rsmi_dev_pci_bandwidth_get(i, &b);
       if (err == RSMI_STATUS_NOT_SUPPORTED) {
         std::cout << "\t**Get PCIE Bandwidth: Not supported on this machine"
-                                                              << std::endl;
+                  << std::endl;
         // Verify api support checking functionality is working
         err = rsmi_dev_pci_bandwidth_get(i, nullptr);
         ASSERT_EQ(err, RSMI_STATUS_NOT_SUPPORTED);
+      } else if (err == RSMI_STATUS_NOT_YET_IMPLEMENTED) {
+        std::cout << "\t**Get PCIE Bandwidth"
+                  << ": Not implemented on this machine" << std::endl;
+      } else if (err == RSMI_STATUS_UNEXPECTED_DATA) {
+        // Treat unexpected data as unsupported metric file version
+        std::cerr << "\t**gpu metric file version unsupported: "
+                  << "PCIe bandwidth on device [" << i << "]" << std::endl;
+        // Do NOT assert; just skip this metric so the test passes
       } else {
         CHK_ERR_ASRT(err)
         IF_VERB(STANDARD) {
@@ -176,11 +183,11 @@ void TestFrequenciesRead::Run(void) {
           err = rsmi_dev_pci_bandwidth_get(i, nullptr);
           if (err != rsmi_status_t::RSMI_STATUS_NOT_SUPPORTED) {
             ASSERT_EQ(err, RSMI_STATUS_INVALID_ARGS);
-          }
-          else {
+          } else {
             auto status_string("");
             rsmi_status_string(err, &status_string);
-            std::cout << "\t\t** rsmi_dev_pci_bandwidth_get(): " << status_string << "\n";
+            std::cout << "\t\t** rsmi_dev_pci_bandwidth_get(): "
+                      << status_string << "\n";
           }
         }
       }

projects/rocm-smi-lib/tests/rocm_smi_test/functional/gpu_metrics_read.cc

@@ -122,14 +122,16 @@ void TestGpuMetricsRead::Run(void) {
 
     rsmi_gpu_metrics_t smu = {};
     err = rsmi_dev_gpu_metrics_info_get(i, &smu);
-    if (err != RSMI_STATUS_SUCCESS) {
-      if (err == RSMI_STATUS_NOT_SUPPORTED) {
-        IF_VERB(STANDARD) {
-          std::cout << "\t**" <<
-          "Not supported on this machine" << std::endl;
-          continue;
-        }
+    if (err == RSMI_STATUS_NOT_SUPPORTED) {
+      IF_VERB(STANDARD) {
+        std::cout << "\t**Not supported on this machine" << std::endl;
       }
+      continue;
+    }
+    if (err == RSMI_STATUS_UNEXPECTED_DATA) {
+      std::cerr << "\t**gpu metric file version unsupported: GPU metrics on device ["
+                << i << "]" << std::endl;
+      continue;
     } else {
       CHK_ERR_ASRT(err);
       IF_VERB(STANDARD) {

projects/rocm-smi-lib/tests/rocm_smi_test/functional/power_read.cc

@@ -146,22 +146,26 @@ void TestPowerRead::Run(void) {
 
       /* Average Power */
       err = rsmi_dev_power_ave_get(i, 0, &val_ui64);
-
       ASSERT_TRUE(err == RSMI_STATUS_SUCCESS
-                 || err == RSMI_STATUS_NOT_SUPPORTED);
+                || err == RSMI_STATUS_NOT_SUPPORTED
+                || err == RSMI_STATUS_UNEXPECTED_DATA);
+
       if (err == RSMI_STATUS_NOT_SUPPORTED) {
-        std::cout <<
-            "\t**Average Power Usage: not supported on this device"
-                                                                   << std::endl;
+        std::cout << "\t**Average Power Usage: not supported on this device"
+                  << std::endl;
+      } else if (err == RSMI_STATUS_UNEXPECTED_DATA) {
+        std::cerr << "\t**gpu metric file version unsupported: Average Power on device ["
+                  << i << "]" << std::endl;
+        // Skip (sysfs read empty / unexpected content)
       } else {
         CHK_RSMI_PERM_ERR(err)
         IF_VERB(STANDARD) {
           std::cout << "\t**Average Power Usage: ";
           if (err == RSMI_STATUS_SUCCESS) {
-            std::cout << static_cast<float>(val_ui64) / 1000 << " W"
-                      << std::endl;
-          }
+          ASSERT_TRUE(type == RSMI_AVERAGE_POWER || type == RSMI_CURRENT_POWER|| type == RSMI_INVALID_POWER);
+          std::cout << static_cast<float>(val_ui64) / 1000 << " W" << std::endl;
         }
+      }  
         // Verify api support checking functionality is working
         err = rsmi_dev_power_ave_get(i, 0, nullptr);
         ASSERT_EQ(err, RSMI_STATUS_INVALID_ARGS);