Added rsmi_version_str_get()

[ROCm/rocm_smi_lib commit: bb73c2607f]

projects/rocm-smi-lib/include/rocm_smi/rocm_smi.h

@@ -144,6 +144,18 @@ typedef rsmi_dev_perf_level_t rsmi_dev_perf_level;
 /**
  * @brief Available clock types.
  */
+
+/**
+ * @brief Software components
+ */
+typedef enum {
+  RSMI_SW_COMP_FIRST = 0x0,
+
+  RSMI_SW_COMP_DRIVER = RSMI_SW_COMP_FIRST,    //!< Driver
+
+  RSMI_SW_COMP_LAST = RSMI_SW_COMP_DRIVER
+} rsmi_sw_component_t;
+
 typedef enum {
   RSMI_CLK_TYPE_SYS = 0x0,            //!< System clock
   RSMI_CLK_TYPE_FIRST = RSMI_CLK_TYPE_SYS,
@@ -174,22 +186,29 @@ typedef enum {
   RSMI_TEMP_MAX,             //!< Temperature max value.
   RSMI_TEMP_MIN,             //!< Temperature min value.
   RSMI_TEMP_MAX_HYST,        //!< Temperature hysteresis value for max limit.
+                             //!< (This is an absolute temperature, not a
+                             //!< delta).
   RSMI_TEMP_MIN_HYST,        //!< Temperature hysteresis value for min limit.
+                             //!< (This is an absolute temperature,
+                             //!<  not a delta).
   RSMI_TEMP_CRITICAL,        //!< Temperature critical max value, typically
                              //!<  greater than corresponding temp_max values.
   RSMI_TEMP_CRITICAL_HYST,   //!< Temperature hysteresis value for critical
-                             //!<  limit.
+                             //!<  limit. (This is an absolute temperature,
+                             //!<  not a delta).
   RSMI_TEMP_EMERGENCY,       //!< Temperature emergency max value, for chips
                              //!<  supporting more than two upper temperature
                              //!<  limits. Must be equal or greater than
                              //!<  corresponding temp_crit values.
   RSMI_TEMP_EMERGENCY_HYST,  //!< Temperature hysteresis value for emergency
-                             //!<  limit.
+                             //!<  limit. (This is an absolute temperature,
+                             //!<  not a delta).
   RSMI_TEMP_CRIT_MIN,        //!< Temperature critical min value, typically
                              //!<  lower than corresponding temperature
                              //!<  minimum values.
   RSMI_TEMP_CRIT_MIN_HYST,   //!< Temperature hysteresis value for critical
-                             //!<  minimum limit.
+                             //!< minimum limit. (This is an absolute
+                             //!< temperature, not a delta).
   RSMI_TEMP_OFFSET,          //!< Temperature offset which is added to the
                              //!  temperature reading by the chip.
   RSMI_TEMP_LOWEST,          //!< Historical minimum temperature.
@@ -440,7 +459,7 @@ typedef rsmi_od_volt_curve_t rsmi_od_volt_curve;
 typedef struct {
   rsmi_range_t curr_sclk_range;          //!< The current SCLK frequency range
   rsmi_range_t curr_mclk_range;          //!< The current MCLK frequency range;
-                                       //!< (upper bound only)
+                                         //!< (upper bound only)
   rsmi_range_t sclk_freq_limits;         //!< The range possible of SCLK values
   rsmi_range_t mclk_freq_limits;         //!< The range possible of MCLK values
 
@@ -1387,6 +1406,30 @@ rsmi_status_t rsmi_dev_od_freq_range_set(uint32_t dv_ind, rsmi_clk_type_t clk,
 rsmi_status_t
 rsmi_version_get(rsmi_version_t *version);
 
+/**
+ * @brief Get the driver version string for the current system.
+ *
+ * @details Given a software component @p component, a pointer to a char
+ * buffer, @p ver_str, this function will write the driver version string
+ * (up to @p len characters) for the current system to @p ver_str. The caller
+ * must ensure that it is safe to write at least @p len characters to @p
+ * ver_str.
+ *
+ * @param[in] component The component for which the version string is being
+ * requested
+ *
+ * @param[inout] ver_str A pointer to a buffer of char's to which the VBIOS
+ * name will be written
+ *
+ * @param[in] len The number of char's pointed to by @p ver_str which can
+ * safely be written to by this function.
+ *
+ * @retval ::RSMI_STATUS_SUCCESS is returned upon successful call.
+ */
+rsmi_status_t
+rsmi_version_str_get(rsmi_sw_component_t component, char *ver_str,
+                                                                uint32_t len);
+
 /**
  * @brief Get the VBIOS identifer string
  *

projects/rocm-smi-lib/src/rocm_smi.cc

@@ -43,6 +43,7 @@
 
 #include <assert.h>
 #include <errno.h>
+#include <sys/utsname.h>
 
 #include <sstream>
 #include <algorithm>
@@ -1231,7 +1232,7 @@ static rsmi_status_t get_dev_name_from_id(uint32_t dv_ind, char *name,
 
   name[std::min(len - 1, ct)] = '\0';
 
-  if (len < val_str.size()) {
+  if (len < (val_str.size() + 1)) {
     return RSMI_STATUS_INSUFFICIENT_SIZE;
   }
 
@@ -1872,11 +1873,18 @@ rsmi_dev_vbios_version_get(uint32_t dv_ind, char *vbios, uint32_t len) {
   std::string val_str;
   int ret = dev->readDevInfo(amd::smi::kDevVBiosVer, &val_str);
 
+  if (ret != 0) {
+    return errno_to_rsmi_status(ret);
+  }
+
   uint32_t ln = val_str.copy(vbios, len);
 
   vbios[std::min(len - 1, ln)] = '\0';
 
-  return errno_to_rsmi_status(ret);
+  if (len < (val_str.size() + 1)) {
+    return RSMI_STATUS_INSUFFICIENT_SIZE;
+  }
+  return RSMI_STATUS_SUCCESS;
 
   CATCH
 }
@@ -1897,3 +1905,53 @@ rsmi_version_get(rsmi_version_t *version) {
 
   CATCH
 }
+
+static const char *kROCmDriverVersionPath = "/sys/module/amdgpu/version";
+
+rsmi_status_t
+rsmi_version_str_get(rsmi_sw_component_t component, char *ver_str,
+                                                               uint32_t len) {
+  if (ver_str == nullptr || len == 0) {
+    return RSMI_STATUS_INVALID_ARGS;
+  }
+
+  TRY
+
+  int err;
+  std::string val_str;
+  std::string ver_path;
+
+  switch (component) {
+    case RSMI_SW_COMP_DRIVER:
+      ver_path = kROCmDriverVersionPath;
+      break;
+
+    default:
+      assert(!"Unexpected component type provided");
+      return RSMI_STATUS_INVALID_ARGS;
+  }
+
+  err = amd::smi::ReadSysfsStr(ver_path, &val_str);
+
+  if (err != 0) {
+    struct utsname buf;
+    err = uname(&buf);
+
+    if (err != 0) {
+      return errno_to_rsmi_status(err);
+    }
+
+    val_str = buf.release;
+  }
+
+  uint32_t ln = val_str.copy(ver_str, len);
+
+  ver_str[std::min(len - 1, ln)] = '\0';
+
+  if (len < (val_str.size() + 1)) {
+    return RSMI_STATUS_INSUFFICIENT_SIZE;
+  }
+  return RSMI_STATUS_SUCCESS;
+
+  CATCH
+}

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

@@ -111,7 +111,7 @@ void TestPowerRead::Run(void) {
       std::cout << "\t**Averge Power Usage: ";
       CHK_RSMI_PERM_ERR(err)
       if (err == RSMI_STATUS_SUCCESS) {
-        std::cout << static_cast<float>(val_ui64)/1000 << " W" << std::endl;
+        std::cout << static_cast<float>(val_ui64)/1000 << " mW" << std::endl;
       }
     }
   }

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

@@ -47,6 +47,7 @@
 #include <stddef.h>
 
 #include <iostream>
+#include <map>
 
 #include "gtest/gtest.h"
 #include "rocm_smi/rocm_smi.h"
@@ -83,6 +84,12 @@ void TestVersionRead::Close() {
   TestBase::Close();
 }
 
+static const uint32_t kVerMaxStrLen = 80;
+
+static const std::map<uint32_t, const char *>
+  kComponentNameMap = {
+      {RSMI_SW_COMP_DRIVER, "Driver Version"},
+};
 
 void TestVersionRead::Run(void) {
   rsmi_status_t err;
@@ -99,4 +106,17 @@ void TestVersionRead::Run(void) {
     std::cout << "\t**RocM SMI Library version: " << ver.major << "." <<
        ver.minor << "." << ver.patch << " (" << ver.build << ")" << std::endl;
   }
+
+  char ver_str[kVerMaxStrLen];
+
+  for (uint32_t cmp = RSMI_SW_COMP_FIRST; cmp <= RSMI_SW_COMP_LAST; ++cmp) {
+    err = rsmi_version_str_get(static_cast<rsmi_sw_component_t>(cmp),
+                                                      ver_str, kVerMaxStrLen);
+    CHK_ERR_ASRT(err)
+
+    IF_VERB(STANDARD) {
+      std::cout << "\t**" << kComponentNameMap.at(cmp) << ": " <<
+                                                         ver_str << std::endl;
+    }
+  }
 }