Merge 'master' into 'amd-master'

Change-Id: I4bf7ea780e1c74ece3ead10be1eb941128ce2d9e


[ROCm/amdsmi commit: 8ab76fd0cc]

projects/amdsmi/CMakeLists.txt

@@ -106,13 +106,15 @@ endif ()
 set(SRC_DIR "src")
 set(INC_DIR "include/rocm_smi")
 
-include_directories(${CMAKE_CURRENT_SOURCE_DIR}/include)
+include_directories(${CMAKE_CURRENT_SOURCE_DIR}/include
+                     ${CMAKE_CURRENT_SOURCE_DIR}/src/shared_mutex)
 set(SMI_SRC_LIST "${SRC_DIR}/rocm_smi_device.cc")
 set(SMI_SRC_LIST ${SMI_SRC_LIST} "${SRC_DIR}/rocm_smi_main.cc")
 set(SMI_SRC_LIST ${SMI_SRC_LIST} "${SRC_DIR}/rocm_smi_monitor.cc")
 set(SMI_SRC_LIST ${SMI_SRC_LIST} "${SRC_DIR}/rocm_smi.cc")
 set(SMI_SRC_LIST ${SMI_SRC_LIST} "${SRC_DIR}/rocm_smi_power_mon.cc")
 set(SMI_SRC_LIST ${SMI_SRC_LIST} "${SRC_DIR}/rocm_smi_utils.cc")
+set(SMI_SRC_LIST ${SMI_SRC_LIST} "${SRC_DIR}/shared_mutex/shared_mutex.c")
 
 set(SMI_INC_LIST "${INC_DIR}/rocm_smi_device.h")
 set(SMI_INC_LIST ${SMI_INC_LIST} "${INC_DIR}/rocm_smi_main.h")
@@ -121,12 +123,14 @@ set(SMI_INC_LIST ${SMI_INC_LIST} "${INC_DIR}/rocm_smi_power_mon.h")
 set(SMI_INC_LIST ${SMI_INC_LIST} "${INC_DIR}/rocm_smi_utils.h")
 set(SMI_INC_LIST ${SMI_INC_LIST} "${INC_DIR}/rocm_smi_common.h")
 set(SMI_INC_LIST ${SMI_INC_LIST} "${INC_DIR}/rocm_smi_exception.h")
+set(SMI_INC_LIST ${SMI_INC_LIST} "${SRC_DIR}/shared_mutex/shared_mutex.h")
 
 set(SMI_EXAMPLE_EXE "rocm_smi_ex")
 
 add_executable(${SMI_EXAMPLE_EXE} "example/rocm_smi_example.cc")
 target_link_libraries(${SMI_EXAMPLE_EXE} ${ROCM_SMI_TARGET})
 add_library(${ROCM_SMI_TARGET} SHARED ${SMI_SRC_LIST} ${SMI_INC_LIST})
+target_link_libraries(${ROCM_SMI_TARGET} pthread rt)
 
 ## Set the VERSION and SOVERSION values
 set_property(TARGET ${ROCM_SMI_TARGET}

projects/amdsmi/include/rocm_smi/rocm_smi.h

@@ -744,6 +744,24 @@ rsmi_status_t rsmi_dev_pci_id_get(uint32_t dv_ind, uint64_t *bdfid);
 rsmi_status_t rsmi_dev_pci_throughput_get(uint32_t dv_ind, uint64_t *sent,
                                     uint64_t *received, uint64_t *max_pkt_sz);
 
+/**
+ *  @brief Get PCIe replay counter
+ *
+ *  @details Given a device index @p dv_ind and a pointer to a uint64_t @p
+ *  counter, this function will write the sum of the number of NAK's received
+ *  by the GPU and the NAK's generated by the GPU to memory pointed to by @p
+ *  counter.
+ *
+ *  @param[in] dv_ind a device index
+ *
+ *  @param[inout] counter a pointer to uint64_t to which the sum of the NAK's
+ *  received and generated by the GPU is written
+ *
+ *  @retval ::RSMI_STATUS_SUCCESS is returned upon successful call.
+ */
+rsmi_status_t rsmi_dev_pci_replay_counter_get(uint32_t dv_ind,
+                                                           uint64_t *counter);
+
 /** @} */  // end of PCIeQuer
 /*****************************************************************************/
 /** @defgroup PCIeCont PCIe Control

projects/amdsmi/include/rocm_smi/rocm_smi_device.h

@@ -42,6 +42,9 @@
  */
 #ifndef INCLUDE_ROCM_SMI_ROCM_SMI_DEVICE_H_
 #define INCLUDE_ROCM_SMI_ROCM_SMI_DEVICE_H_
+
+#include <pthread.h>
+
 #include <string>
 #include <memory>
 #include <utility>
@@ -52,6 +55,9 @@
 #include "rocm_smi/rocm_smi_power_mon.h"
 #include "rocm_smi/rocm_smi_common.h"
 #include "rocm_smi/rocm_smi.h"
+extern "C" {
+#include "shared_mutex.h"   // NOLINT
+};
 
 namespace amd {
 namespace smi {
@@ -84,6 +90,7 @@ enum DevInfoTypes {
   kDevMemUsedGTT,
   kDevMemUsedVisVRAM,
   kDevMemUsedVRAM,
+  kDevPCIEReplayCount,
 };
 
 class Device {
@@ -109,11 +116,13 @@ class Device {
     uint64_t bdfid(void) const {return bdfid_;}
     void set_bdfid(uint64_t val) {bdfid_ = val;}
     uint64_t get_bdfid(void) const {return bdfid_;}
+    pthread_mutex_t *mutex(void) {return mutex_.ptr;}
 
  private:
     std::shared_ptr<Monitor> monitor_;
     std::shared_ptr<PowerMon> power_monitor_;
     std::string path_;
+    shared_mutex_t mutex_;
     uint32_t index_;
     const RocmSMI_env_vars *env_;
     template <typename T> int openSysfsFileStream(DevInfoTypes type, T *fs,

projects/amdsmi/include/rocm_smi/rocm_smi_utils.h

@@ -43,6 +43,8 @@
 #ifndef INCLUDE_ROCM_SMI_ROCM_SMI_UTILS_H_
 #define INCLUDE_ROCM_SMI_ROCM_SMI_UTILS_H_
 
+#include <pthread.h>
+
 #include <string>
 #include <cstdint>
 
@@ -63,6 +65,28 @@ namespace smi {
 int ReadSysfsStr(std::string path, std::string *retStr);
 int WriteSysfsStr(std::string path, std::string val);
 
+struct pthread_wrap {
+ public:
+        explicit pthread_wrap(pthread_mutex_t &p_mut) : mutex_(p_mut) {}
+
+        void Acquire() { pthread_mutex_lock(&mutex_);   }
+        void Release() { pthread_mutex_unlock(&mutex_); }
+ private:
+        pthread_mutex_t& mutex_;
+};
+struct ScopedPthread {
+     explicit ScopedPthread(pthread_wrap& mutex) : pthrd_ref_(mutex) {
+       pthrd_ref_.Acquire();
+     }
+
+     ~ScopedPthread() {
+       pthrd_ref_.Release();
+     }
+ private:
+     ScopedPthread(const ScopedPthread&);
+
+     pthread_wrap& pthrd_ref_;
+};
 }  // namespace smi
 }  // namespace amd
 

projects/amdsmi/src/rocm_smi.cc

@@ -44,6 +44,7 @@
 #include <assert.h>
 #include <errno.h>
 #include <sys/utsname.h>
+#include <pthread.h>
 
 #include <sstream>
 #include <algorithm>
@@ -100,12 +101,27 @@ static rsmi_status_t handleException() {
   std::shared_ptr<amd::smi::Device> dev = smi.monitor_devices()[dv_ind]; \
   assert(dev != nullptr);
 
+#define DEVICE_MUTEX \
+    amd::smi::pthread_wrap _pw(*get_mutex(dv_ind)); \
+    amd::smi::ScopedPthread _lock(_pw);
+
+static pthread_mutex_t *get_mutex(uint32_t dv_ind) {
+  amd::smi::RocmSMI smi = amd::smi::RocmSMI::getInstance();
+  if (dv_ind >= smi.monitor_devices().size()) {
+    return nullptr;
+  }
+  std::shared_ptr<amd::smi::Device> dev = smi.monitor_devices()[dv_ind];
+  assert(dev != nullptr);
+
+  return dev->mutex();
+}
+
 static rsmi_status_t errno_to_rsmi_status(uint32_t err) {
   switch (err) {
     case 0:      return RSMI_STATUS_SUCCESS;
     case EACCES: return RSMI_STATUS_PERMISSION;
-    case EPERM:  return RSMI_STATUS_NOT_SUPPORTED;
-    case ENOENT:
+    case EPERM:
+    case ENOENT: return RSMI_STATUS_NOT_SUPPORTED;
     case EISDIR: return RSMI_STATUS_FILE_ERROR;
     default:     return RSMI_STATUS_UNKNOWN_ERROR;
   }
@@ -440,6 +456,9 @@ rsmi_status_t rsmi_dev_ecc_enabled_get(uint32_t dv_ind,
   if (enabled_mask == nullptr) {
     return RSMI_STATUS_INVALID_ARGS;
   }
+
+  DEVICE_MUTEX
+
   std::vector<std::string> val_vec;
 
   ret = get_dev_value_vec(amd::smi::kDevErrCntFeatures, dv_ind, &val_vec);
@@ -508,6 +527,8 @@ rsmi_status_t rsmi_dev_ecc_status_get(uint32_t dv_ind, rsmi_gpu_block_t block,
   rsmi_status_t ret;
   std::vector<std::string> val_vec;
 
+  DEVICE_MUTEX
+
   ret = get_dev_value_vec(amd::smi::kDevErrCntFeatures, dv_ind, &val_vec);
 
   if (ret == RSMI_STATUS_FILE_ERROR) {
@@ -569,6 +590,9 @@ rsmi_dev_ecc_count_get(uint32_t dv_ind, rsmi_gpu_block_t block,
     default:
       return RSMI_STATUS_NOT_SUPPORTED;
   }
+
+  DEVICE_MUTEX
+
   ret = get_dev_value_vec(type, dv_ind, &val_vec);
 
   if (ret == RSMI_STATUS_FILE_ERROR) {
@@ -605,6 +629,8 @@ rsmi_dev_pci_id_get(uint32_t dv_ind, uint64_t *bdfid) {
   }
   GET_DEV_FROM_INDX
 
+  DEVICE_MUTEX
+
   *bdfid = dev->get_bdfid();
   return RSMI_STATUS_SUCCESS;
   CATCH
@@ -614,6 +640,9 @@ static rsmi_status_t
 get_id(uint32_t dv_ind, amd::smi::DevInfoTypes typ, uint16_t *id) {
   TRY
   std::string val_str;
+
+  DEVICE_MUTEX
+
   rsmi_status_t ret = get_dev_value_str(typ, dv_ind, &val_str);
 
   if (ret != RSMI_STATUS_SUCCESS) {
@@ -630,21 +659,25 @@ get_id(uint32_t dv_ind, amd::smi::DevInfoTypes typ, uint16_t *id) {
 
 rsmi_status_t
 rsmi_dev_id_get(uint32_t dv_ind, uint16_t *id) {
+  DEVICE_MUTEX
   return get_id(dv_ind, amd::smi::kDevDevID, id);
 }
 
 rsmi_status_t
 rsmi_dev_subsystem_id_get(uint32_t dv_ind, uint16_t *id) {
+  DEVICE_MUTEX
   return get_id(dv_ind, amd::smi::kDevSubSysDevID, id);
 }
 
 rsmi_status_t
 rsmi_dev_vendor_id_get(uint32_t dv_ind, uint16_t *id) {
+  DEVICE_MUTEX
   return get_id(dv_ind, amd::smi::kDevVendorID, id);
 }
 
 rsmi_status_t
 rsmi_dev_subsystem_vendor_id_get(uint32_t dv_ind, uint16_t *id) {
+  DEVICE_MUTEX
   return get_id(dv_ind, amd::smi::kDevSubSysVendorID, id);
 }
 
@@ -652,6 +685,8 @@ rsmi_status_t
 rsmi_dev_perf_level_get(uint32_t dv_ind, rsmi_dev_perf_level_t *perf) {
   TRY
   std::string val_str;
+  DEVICE_MUTEX
+
   rsmi_status_t ret = get_dev_value_str(amd::smi::kDevPerfLevel, dv_ind,
                                                                     &val_str);
   if (ret != RSMI_STATUS_SUCCESS) {
@@ -668,6 +703,8 @@ rsmi_status_t
 rsmi_dev_overdrive_level_get(uint32_t dv_ind, uint32_t *od) {
   TRY
   std::string val_str;
+  DEVICE_MUTEX
+
   rsmi_status_t ret = get_dev_value_str(amd::smi::kDevOverDriveLevel, dv_ind,
                                                                     &val_str);
   if (ret != RSMI_STATUS_SUCCESS) {
@@ -688,7 +725,7 @@ rsmi_dev_overdrive_level_set(int32_t dv_ind, uint32_t od) {
   if (od > kMaxOverdriveLevel) {
     return RSMI_STATUS_INVALID_ARGS;
   }
-
+  DEVICE_MUTEX
   return set_dev_value(amd::smi::kDevOverDriveLevel, dv_ind, od);
   CATCH
 }
@@ -700,6 +737,7 @@ rsmi_dev_perf_level_set(int32_t dv_ind, rsmi_dev_perf_level_t perf_level) {
     return RSMI_STATUS_INVALID_ARGS;
   }
 
+  DEVICE_MUTEX
   return set_dev_value(amd::smi::kDevPerfLevel, dv_ind, perf_level);
   CATCH
 }
@@ -1009,6 +1047,8 @@ rsmi_dev_gpu_clk_freq_get(uint32_t dv_ind, rsmi_clk_type_t clk_type,
       return RSMI_STATUS_INVALID_ARGS;
   }
 
+  DEVICE_MUTEX
+
   return get_frequencies(dev_type, dv_ind, f);
 
   CATCH
@@ -1035,6 +1075,9 @@ rsmi_dev_gpu_clk_freq_set(uint32_t dv_ind,
   rsmi_frequencies_t freqs;
 
   TRY
+
+  DEVICE_MUTEX
+
   ret = rsmi_dev_gpu_clk_freq_get(dv_ind, clk_type, &freqs);
 
   if (ret != RSMI_STATUS_SUCCESS) {
@@ -1247,6 +1290,8 @@ rsmi_dev_name_get(uint32_t dv_ind, char *name, size_t len) {
     return RSMI_STATUS_INVALID_ARGS;
   }
 
+  DEVICE_MUTEX
+
   ret = get_dev_name_from_id(dv_ind, name, len, NAME_STR_DEVICE);
   if (ret != RSMI_STATUS_SUCCESS) {
     return ret;
@@ -1265,6 +1310,8 @@ rsmi_dev_subsystem_name_get(uint32_t dv_ind, char *name, size_t len) {
     return RSMI_STATUS_INVALID_ARGS;
   }
 
+  DEVICE_MUTEX
+
   ret = get_dev_name_from_id(dv_ind, name, len, NAME_STR_SUBSYS);
   return ret;
   CATCH
@@ -1279,6 +1326,7 @@ rsmi_dev_vendor_name_get(uint32_t dv_ind, char *name, size_t len) {
     return RSMI_STATUS_INVALID_ARGS;
   }
 
+  DEVICE_MUTEX
   ret = get_dev_name_from_id(dv_ind, name, len, NAME_STR_VENDOR);
   return ret;
   CATCH
@@ -1294,6 +1342,8 @@ rsmi_dev_pci_bandwidth_get(uint32_t dv_ind, rsmi_pcie_bandwidth_t *b) {
     return RSMI_STATUS_INVALID_ARGS;
   }
 
+  DEVICE_MUTEX
+
   return get_frequencies(amd::smi::kDevPCIEClk, dv_ind,
                                         &b->transfer_rate, b->lanes);
 
@@ -1306,6 +1356,8 @@ rsmi_dev_pci_bandwidth_set(uint32_t dv_ind, uint64_t bw_bitmask) {
   rsmi_pcie_bandwidth_t bws;
 
   TRY
+
+  DEVICE_MUTEX
   ret = rsmi_dev_pci_bandwidth_get(dv_ind, &bws);
 
   if (ret != RSMI_STATUS_SUCCESS) {
@@ -1346,6 +1398,9 @@ rsmi_dev_pci_throughput_get(uint32_t dv_ind, uint64_t *sent,
   rsmi_status_t ret;
 
   std::string val_str;
+
+  DEVICE_MUTEX
+
   ret = get_dev_value_line(amd::smi::kDevPCIEThruPut, dv_ind, &val_str);
 
   if (ret != RSMI_STATUS_SUCCESS) {
@@ -1435,6 +1490,8 @@ rsmi_dev_temp_metric_get(uint32_t dv_ind, uint32_t sensor_ind,
       mon_type = amd::smi::kMonInvalid;
   }
 
+  DEVICE_MUTEX
+
   ret = get_dev_mon_value(mon_type, dv_ind, sensor_ind, temperature);
 
   return ret;
@@ -1453,6 +1510,8 @@ rsmi_dev_fan_speed_get(uint32_t dv_ind, uint32_t sensor_ind, int64_t *speed) {
 
   ++sensor_ind;  // fan sysfs files have 1-based indices
 
+  DEVICE_MUTEX
+
   ret = get_dev_mon_value(amd::smi::kMonFanSpeed, dv_ind, sensor_ind, speed);
 
   return ret;
@@ -1470,6 +1529,8 @@ rsmi_dev_fan_rpms_get(uint32_t dv_ind, uint32_t sensor_ind, int64_t *speed) {
 
   rsmi_status_t ret;
 
+  DEVICE_MUTEX
+
   ret = get_dev_mon_value(amd::smi::kMonFanRPMs, dv_ind, sensor_ind, speed);
 
   return ret;
@@ -1484,6 +1545,8 @@ rsmi_dev_fan_reset(uint32_t dv_ind, uint32_t sensor_ind) {
 
   ++sensor_ind;  // fan sysfs files have 1-based indices
 
+  DEVICE_MUTEX
+
   ret = set_dev_mon_value<uint64_t>(amd::smi::kMonFanCntrlEnable,
                                                        dv_ind, sensor_ind, 2);
 
@@ -1499,6 +1562,7 @@ rsmi_dev_fan_speed_set(uint32_t dv_ind, uint32_t sensor_ind, uint64_t speed) {
   rsmi_status_t ret;
   uint64_t max_speed;
 
+  DEVICE_MUTEX
 
   ret = rsmi_dev_fan_speed_max_get(dv_ind, sensor_ind, &max_speed);
 
@@ -1541,6 +1605,8 @@ rsmi_dev_fan_speed_max_get(uint32_t dv_ind, uint32_t sensor_ind,
 
   rsmi_status_t ret;
 
+  DEVICE_MUTEX
+
   ret = get_dev_mon_value(amd::smi::kMonMaxFanSpeed, dv_ind, sensor_ind,
                                       reinterpret_cast<int64_t *>(max_speed));
 
@@ -1551,6 +1617,8 @@ rsmi_dev_fan_speed_max_get(uint32_t dv_ind, uint32_t sensor_ind,
 rsmi_status_t
 rsmi_dev_od_volt_info_get(uint32_t dv_ind, rsmi_od_volt_freq_data_t *odv) {
   TRY
+  DEVICE_MUTEX
+
   rsmi_status_t ret = get_od_clk_volt_info(dv_ind, odv);
 
   return ret;
@@ -1564,6 +1632,8 @@ rsmi_status_t rsmi_dev_od_volt_curve_regions_get(uint32_t dv_ind,
   if (buffer == nullptr || num_regions == nullptr || *num_regions == 0) {
     return RSMI_STATUS_INVALID_ARGS;
   }
+
+  DEVICE_MUTEX
   rsmi_status_t ret = get_od_clk_volt_curve_regions(dv_ind, num_regions,
                                                                       buffer);
   return ret;
@@ -1582,6 +1652,8 @@ rsmi_dev_power_max_get(uint32_t dv_ind, uint32_t sensor_ind, uint64_t *power) {
   // ++sensor_ind;  // power sysfs files have 1-based indices
 
   rsmi_status_t ret;
+
+  DEVICE_MUTEX
   ret = get_power_mon_value(amd::smi::kPowerMaxGPUPower, dv_ind, power);
 
   return ret;
@@ -1598,6 +1670,8 @@ rsmi_dev_power_ave_get(uint32_t dv_ind, uint32_t sensor_ind, uint64_t *power) {
   ++sensor_ind;  // power sysfs files have 1-based indices
 
   rsmi_status_t ret;
+
+  DEVICE_MUTEX
   ret = get_dev_mon_value(amd::smi::kMonPowerAve, dv_ind, sensor_ind, power);
 
   return ret;
@@ -1615,6 +1689,8 @@ rsmi_dev_power_cap_get(uint32_t dv_ind, uint32_t sensor_ind, uint64_t *cap) {
   ++sensor_ind;  // power sysfs files have 1-based indices
 
   rsmi_status_t ret;
+
+  DEVICE_MUTEX
   ret = get_dev_mon_value(amd::smi::kMonPowerCap, dv_ind, sensor_ind, cap);
 
   return ret;
@@ -1633,6 +1709,8 @@ rsmi_dev_power_cap_range_get(uint32_t dv_ind, uint32_t sensor_ind,
   ++sensor_ind;  // power sysfs files have 1-based indices
 
   rsmi_status_t ret;
+
+  DEVICE_MUTEX
   ret = get_dev_mon_value(amd::smi::kMonPowerCapMax, dv_ind, sensor_ind, max);
 
   if (ret == RSMI_STATUS_SUCCESS) {
@@ -1650,6 +1728,7 @@ rsmi_dev_power_cap_set(uint32_t dv_ind, uint32_t sensor_ind, uint64_t cap) {
   rsmi_status_t ret;
   uint64_t min, max;
 
+  DEVICE_MUTEX
 
   ret = rsmi_dev_power_cap_range_get(dv_ind, sensor_ind, &max, &min);
   if (ret != RSMI_STATUS_SUCCESS) {
@@ -1678,6 +1757,7 @@ rsmi_dev_power_profile_presets_get(uint32_t dv_ind, uint32_t sensor_ind,
 
   ++sensor_ind;  // power sysfs files have 1-based indices
 
+  DEVICE_MUTEX
   rsmi_status_t ret = get_power_profiles(dv_ind, status, nullptr);
   return ret;
   CATCH
@@ -1689,6 +1769,7 @@ rsmi_dev_power_profile_set(uint32_t dv_ind, uint32_t sensor_ind,
   TRY
   ++sensor_ind;  // power sysfs files have 1-based indices
 
+  DEVICE_MUTEX
   rsmi_status_t ret = set_power_profile(dv_ind, profile);
   return ret;
   CATCH
@@ -1722,6 +1803,8 @@ rsmi_dev_memory_total_get(uint32_t dv_ind, rsmi_memory_type_t mem_type,
       assert(!"Unexpected memory type");
       return RSMI_STATUS_INVALID_ARGS;
   }
+
+  DEVICE_MUTEX
   ret = get_dev_value_int(mem_type_file, dv_ind, total);
 
   return ret;
@@ -1755,6 +1838,8 @@ rsmi_dev_memory_usage_get(uint32_t dv_ind, rsmi_memory_type_t mem_type,
       assert(!"Unexpected memory type");
       return RSMI_STATUS_INVALID_ARGS;
   }
+
+  DEVICE_MUTEX
   ret = get_dev_value_int(mem_type_file, dv_ind, used);
 
   return ret;
@@ -1847,6 +1932,8 @@ rsmi_status_t
 rsmi_dev_busy_percent_get(uint32_t dv_ind, uint32_t *busy_percent) {
   TRY
   std::string val_str;
+
+  DEVICE_MUTEX
   rsmi_status_t ret = get_dev_value_str(amd::smi::kDevUsage, dv_ind,
                                                                     &val_str);
   if (ret != RSMI_STATUS_SUCCESS) {
@@ -1871,6 +1958,8 @@ rsmi_dev_vbios_version_get(uint32_t dv_ind, char *vbios, uint32_t len) {
   TRY
   GET_DEV_FROM_INDX
   std::string val_str;
+
+  DEVICE_MUTEX
   int ret = dev->readDevInfo(amd::smi::kDevVBiosVer, &val_str);
 
   if (ret != 0) {
@@ -1955,3 +2044,16 @@ rsmi_version_str_get(rsmi_sw_component_t component, char *ver_str,
 
   CATCH
 }
+
+rsmi_status_t
+rsmi_dev_pci_replay_counter_get(uint32_t dv_ind, uint64_t *counter) {
+  TRY
+
+  DEVICE_MUTEX
+  rsmi_status_t ret;
+
+  ret = get_dev_value_int(amd::smi::kDevPCIEReplayCount, dv_ind, counter);
+  return ret;
+
+  CATCH
+}

projects/amdsmi/src/rocm_smi_device.cc

@@ -41,6 +41,10 @@
  *
  */
 
+#include <pthread.h>
+#include <unistd.h>
+#include <sys/types.h>
+
 #include <assert.h>
 #include <sys/stat.h>
 #include <stdint.h>
@@ -55,6 +59,11 @@
 #include "rocm_smi/rocm_smi_main.h"
 #include "rocm_smi/rocm_smi_device.h"
 #include "rocm_smi/rocm_smi.h"
+#include "rocm_smi/rocm_smi_exception.h"
+
+extern "C" {
+#include "shared_mutex.h"  // NOLINT
+};
 
 namespace amd {
 namespace smi {
@@ -87,6 +96,7 @@ static const char *kDevMemTotVRAMFName = "mem_info_vram_total";
 static const char *kDevMemUsedGTTFName = "mem_info_gtt_used";
 static const char *kDevMemUsedVisVRAMFName = "mem_info_vis_vram_used";
 static const char *kDevMemUsedVRAMFName = "mem_info_vram_used";
+static const char *kDevPCIEReplayCountFName = "pcie_replay_count";
 
 // Strings that are found within sysfs files
 static const char *kDevPerfLevelAutoStr = "auto";
@@ -127,6 +137,7 @@ static const std::map<DevInfoTypes, const char *> kDevAttribNameMap = {
     {kDevMemUsedGTT, kDevMemUsedGTTFName},
     {kDevMemUsedVisVRAM, kDevMemUsedVisVRAMFName},
     {kDevMemUsedVRAM, kDevMemUsedVRAMFName},
+    {kDevPCIEReplayCount, kDevPCIEReplayCountFName},
 };
 
 static const std::map<rsmi_dev_perf_level, const char *> kDevPerfLvlMap = {
@@ -154,9 +165,26 @@ static bool isRegularFile(std::string fname) {
 
 Device::Device(std::string p, RocmSMI_env_vars const *e) : path_(p), env_(e) {
   monitor_ = nullptr;
+
+  // Get the device name
+  size_t i = path_.rfind('/', path_.length());
+  std::string dev = path_.substr(i + 1, path_.length() - i);
+
+  std::string m_name("/rocm_smi_");
+  m_name += dev;
+  m_name += '_';
+  m_name += std::to_string(geteuid());
+
+  mutex_ = shared_mutex_init(m_name.c_str(), 0777);
+
+  if (mutex_.ptr == nullptr) {
+    throw amd::smi::rsmi_exception(RSMI_INITIALIZATION_ERROR,
+                                       "Failed to create shared mem. mutex.");
+  }
 }
 
 Device:: ~Device() {
+  shared_mutex_close(mutex_);
 }
 
 template <typename T>
@@ -176,7 +204,7 @@ int Device::openSysfsFileStream(DevInfoTypes type, T *fs, const char *str) {
 
   DBG_FILE_ERROR(sysfs_path, str);
   if (!isRegularFile(sysfs_path)) {
-    return EISDIR;
+    return ENOENT;
   }
 
   fs->open(sysfs_path);
@@ -341,6 +369,7 @@ int Device::readDevInfo(DevInfoTypes type, uint64_t *val) {
     case kDevMemUsedGTT:
     case kDevMemUsedVisVRAM:
     case kDevMemUsedVRAM:
+    case kDevPCIEReplayCount:
       ret = readDevInfoStr(type, &tempStr);
       RET_IF_NONZERO(ret);
       *val = std::stoul(tempStr, 0);

projects/amdsmi/src/shared_mutex/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2018 Oleg Yamnikov
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

projects/amdsmi/src/shared_mutex/shared_mutex.c

@@ -0,0 +1,145 @@
+#include "shared_mutex.h"
+#include <errno.h> // errno, ENOENT
+#include <fcntl.h> // O_RDWR, O_CREATE
+#include <linux/limits.h> // NAME_MAX
+#include <sys/mman.h> // shm_open, shm_unlink, mmap, munmap,
+                      // PROT_READ, PROT_WRITE, MAP_SHARED, MAP_FAILED
+#include <unistd.h> // ftruncate, close
+#include <stdio.h> // perror
+#include <stdlib.h> // malloc, free
+#include <string.h> // strcpy
+#include <time.h>  // clock_gettime
+
+shared_mutex_t shared_mutex_init(const char *name, mode_t mode) {
+  shared_mutex_t mutex = {NULL, 0, NULL, 0};
+  errno = 0;
+
+  // Open existing shared memory object, or create one.
+  // Two separate calls are needed here, to mark fact of creation
+  // for later initialization of pthread mutex.
+  mutex.shm_fd = shm_open(name, O_RDWR, mode);
+  if (errno == ENOENT) {
+    mutex.shm_fd = shm_open(name, O_RDWR|O_CREAT, mode);
+    mutex.created = 1;
+    // Change permissions of shared memory, so every body can access it. Avoiding the umask of shm_open
+    if (fchmod(mutex.shm_fd, mode) != 0) {
+      perror("fchmod");
+    }
+  }
+  if (mutex.shm_fd == -1) {
+    perror("shm_open");
+    return mutex;
+  }
+
+  // Truncate shared memory segment so it would contain
+  // pthread_mutex_t AND the ref. count
+  if (ftruncate(mutex.shm_fd, sizeof(pthread_mutex_t)) != 0) {
+    perror("ftruncate");
+    return mutex;
+  }
+
+  // Map pthread mutex into the shared memory.
+  void *addr = mmap(
+    NULL,
+    sizeof(pthread_mutex_t),
+    PROT_READ|PROT_WRITE,
+    MAP_SHARED,
+    mutex.shm_fd,
+    0
+  );
+  if (addr == MAP_FAILED) {
+    perror("mmap");
+    return mutex;
+  }
+
+  pthread_mutex_t *mutex_ptr =  (pthread_mutex_t *)addr;
+
+  // Make sure the mutex wasn't left in a locked state. If we can't
+  // acquire it in 3 sec., re-do everything.
+  struct timespec expireTime;
+  clock_gettime(CLOCK_REALTIME, &expireTime);
+  expireTime.tv_sec += 3;
+
+  int ret = pthread_mutex_timedlock(mutex_ptr, &expireTime);
+  
+  if (ret || (mutex.created == 0 && ((shared_mutex_t *)addr)->ptr == NULL)) {
+    // Something is out of sync. Unlink shm and start over.
+    if (shm_unlink(name)) {
+      mutex.shm_fd = 0;
+      perror("shm_unlink");
+    }
+    free(mutex.name);
+
+    return shared_mutex_init(name, mode);
+  } else {
+    if (pthread_mutex_unlock(mutex_ptr)) {
+      perror("pthread_mutex_unlock");
+    }
+  }
+
+
+  if (mutex.created) {
+    pthread_mutexattr_t attr;
+    if (pthread_mutexattr_init(&attr)) {
+      perror("pthread_mutexattr_init");
+      return mutex;
+    }
+    if (pthread_mutexattr_setpshared(&attr, PTHREAD_PROCESS_SHARED)) {
+      perror("pthread_mutexattr_setpshared");
+      return mutex;
+    }
+    
+    if (pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE)) {
+      perror("pthread_mutexattr_settype");
+      return mutex;
+    }
+    if (pthread_mutex_init(mutex_ptr, &attr)) {
+      perror("pthread_mutex_init");
+      return mutex;
+    }
+  }
+  
+  mutex.ptr = mutex_ptr;
+  mutex.name = (char *)malloc(NAME_MAX+1);
+  strcpy(mutex.name, name);
+  return mutex;
+}
+
+int shared_mutex_close(shared_mutex_t mutex) {  
+  if (munmap((void *)mutex.ptr, sizeof(pthread_mutex_t))) {
+    perror("munmap");
+    return -1;
+  }
+  mutex.ptr = NULL;
+  if (close(mutex.shm_fd)) {
+    perror("close");
+    return -1;
+  }
+  mutex.shm_fd = 0;
+  free(mutex.name);
+
+  return 0;
+}
+
+int shared_mutex_destroy(shared_mutex_t mutex) {
+  if ((errno = pthread_mutex_destroy(mutex.ptr))) {
+    perror("pthread_mutex_destroy");
+    return -1;
+  }
+  if (munmap((void *)mutex.ptr, sizeof(pthread_mutex_t))) {
+    perror("munmap");
+    return -1;
+  }
+  mutex.ptr = NULL;
+  if (close(mutex.shm_fd)) {
+    perror("close");
+    return -1;
+  }
+  mutex.shm_fd = 0;
+  if (shm_unlink(mutex.name)) {
+    perror("shm_unlink");
+    return -1;
+  }
+  free(mutex.name);
+  return 0;
+}

projects/amdsmi/src/shared_mutex/shared_mutex.h

@@ -0,0 +1,70 @@
+// NOLINT(legal/copyright)
+// See LICENSE file
+
+#ifndef SRC_SHARED_MUTEX_SHARED_MUTEX_H_
+#define SRC_SHARED_MUTEX_SHARED_MUTEX_H_
+
+#include <sys/stat.h>
+
+#include <pthread.h>  // pthread_mutex_t, pthread_mutexattr_t,
+                      // pthread_mutexattr_init, pthread_mutexattr_setpshared,
+                      // pthread_mutex_init, pthread_mutex_destroy
+
+// Structure of a shared mutex.
+typedef struct shared_mutex_t {
+  pthread_mutex_t *ptr;  // Pointer to the pthread mutex and
+                         // shared memory segment.
+  int shm_fd;            // Descriptor of shared memory object.
+  char* name;            // Name of the mutex and associated
+                         // shared memory object.
+  int created;           // Equals 1 (true) if initialization
+                         // of this structure caused creation
+                         // of a new shared mutex.
+                         // Equals 0 (false) if this mutex was
+                         // just retrieved from shared memory.
+} shared_mutex_t;
+
+// Initialize a new shared mutex with given `name`. If a mutex
+// with such name exists in the system, it will be loaded.
+// Otherwise a new mutes will by created.
+//
+// In case of any error, it will be printed into the standard output
+// and the returned structure will have `ptr` equal `NULL`.
+// `errno` wil not be reset in such case, so you may used it.
+//
+// **NOTE:** In case when the mutex appears to be uncreated,
+// this function becomes *non-thread-safe*. If multiple threads
+// call it at one moment, there occur several race conditions,
+// in which one call might recreate another's shared memory
+// object or rewrite another's pthread mutex in the shared memory.
+// There is no workaround currently, except to run first
+// initialization only before multi-threaded or multi-process
+// functionality.
+shared_mutex_t shared_mutex_init(const char *name, mode_t mode);
+
+// Close access to the shared mutex and free all the resources,
+// used by the structure.
+//
+// Returns 0 in case of success. If any error occurs, it will be
+// printed into the standard output and the function will return -1.
+// `errno` wil not be reset in such case, so you may used it.
+//
+// **NOTE:** It will not destroy the mutex. The mutex would not
+// only be available to other processes using it right now,
+// but also to any process which might want to use it later on.
+// For complete desctruction use `shared_mutex_destroy` instead.
+//
+// **NOTE:** It will not unlock locked mutex.
+int shared_mutex_close(shared_mutex_t mutex);
+
+// Close and destroy shared mutex.
+// Any open pointers to it will be invalidated.
+//
+// Returns 0 in case of success. If any error occurs, it will be
+// printed into the standard output and the function will return -1.
+// `errno` wil not be reset in such case, so you may used it.
+//
+// **NOTE:** It will not unlock locked mutex.
+int shared_mutex_destroy(shared_mutex_t mutex);
+
+#endif  // SRC_SHARED_MUTEX_SHARED_MUTEX_H_

projects/amdsmi/tests/rocm_smi_test/functional/err_cnt_read.cc

@@ -96,7 +96,7 @@ void TestErrCntRead::Run(void) {
     err = rsmi_dev_ecc_enabled_get(i, &enabled_mask);
     if (err == RSMI_STATUS_NOT_SUPPORTED) {
       std::cout <<
-          "\t**Error Count Enabled Mask for is not supported on this machine"
+          "\t**Error Count Enabled Mask get is not supported on this machine"
                                                                  << std::endl;
     } else {
       CHK_ERR_ASRT(err)

projects/amdsmi/tests/rocm_smi_test/functional/pci_read_write.cc

@@ -89,13 +89,26 @@ void TestPciReadWrite::Run(void) {
   rsmi_status_t ret;
   rsmi_pcie_bandwidth_t bw;
   uint32_t freq_bitmask;
-  uint64_t sent, received, max_pkt_sz;
+  uint64_t sent, received, max_pkt_sz, u64int;
 
   TestBase::Run();
 
   for (uint32_t dv_ind = 0; dv_ind < num_monitor_devs(); ++dv_ind) {
     PrintDeviceHeader(dv_ind);
 
+    ret = rsmi_dev_pci_replay_counter_get(dv_ind, &u64int);
+
+     if (ret == RSMI_STATUS_NOT_SUPPORTED) {
+        std::cout <<
+            "\t**rsmi_dev_pci_replay_counter_get() is not supported"
+            " on this machine" << std::endl;
+      } else {
+        CHK_ERR_ASRT(ret)
+        IF_VERB(STANDARD) {
+          std::cout << "\tPCIe Replay Counter: " << u64int << std::endl;
+        }
+      }
+
     ret = rsmi_dev_pci_throughput_get(dv_ind, &sent, &received, &max_pkt_sz);
     if (ret == RSMI_STATUS_NOT_SUPPORTED) {
       std::cout << "TEST FAILURE: Current PCIe throughput is not detected. "
@@ -106,7 +119,7 @@ void TestPciReadWrite::Run(void) {
     CHK_ERR_ASRT(ret)
 
     IF_VERB(STANDARD) {
-      std::cout << "PCIe Throughput (1 sec.): " << std::endl;
+      std::cout << "\tPCIe Throughput (1 sec.): " << std::endl;
       std::cout << "\t\tSent: " << sent << " bytes" << std::endl;
       std::cout << "\t\tReceived: " << received << " bytes" << std::endl;
       std::cout << "\t\tMax Packet Size: " << max_pkt_sz << " bytes" <<
@@ -125,7 +138,8 @@ void TestPciReadWrite::Run(void) {
     CHK_ERR_ASRT(ret)
 
     IF_VERB(STANDARD) {
-      std::cout << "Initial PCIe is " << bw.transfer_rate.current << std::endl;
+      std::cout << "\tInitial PCIe is " << bw.transfer_rate.current <<
+                                                                    std::endl;
     }
 
     // First set the bitmask to all supported bandwidths
@@ -141,7 +155,7 @@ void TestPciReadWrite::Run(void) {
                                                        freq_bm_str.size()-1));
 
     IF_VERB(STANDARD) {
-    std::cout << "Setting bandwidth mask to " << "0b" << freq_bm_str <<
+    std::cout << "\tSetting bandwidth mask to " << "0b" << freq_bm_str <<
                                                             " ..." << std::endl;
     }
     ret = rsmi_dev_pci_bandwidth_set(dv_ind, freq_bitmask);
@@ -151,9 +165,9 @@ void TestPciReadWrite::Run(void) {
     CHK_ERR_ASRT(ret)
 
     IF_VERB(STANDARD) {
-      std::cout << "Bandwidth is now index " << bw.transfer_rate.current <<
+      std::cout << "\tBandwidth is now index " << bw.transfer_rate.current <<
                                                                       std::endl;
-      std::cout << "Resetting mask to all bandwidths." << std::endl;
+      std::cout << "\tResetting mask to all bandwidths." << std::endl;
     }
     ret = rsmi_dev_pci_bandwidth_set(dv_ind, 0xFFFFFFFF);
     CHK_ERR_ASRT(ret)

projects/amdsmi/tests/rocm_smi_test/functional/temp_read.cc

@@ -99,7 +99,7 @@ void TestTempRead::Run(void) {
       err = rsmi_dev_temp_metric_get(i, 0, met, &val_i64);
 
       if (err != RSMI_STATUS_SUCCESS) {
-        if (err == RSMI_STATUS_FILE_ERROR) {
+        if (err == RSMI_STATUS_NOT_SUPPORTED) {
           IF_VERB(STANDARD) {
             std::cout << "\t**" << label << ": " <<
                                "Not supported on this machine" << std::endl;

projects/amdsmi/tests/rocm_smi_test/functional/volt_freq_curv_read.cc

@@ -155,7 +155,7 @@ void TestVoltCurvRead::Run(void) {
 
     err = rsmi_dev_od_volt_info_get(i, &odv);
     if (err == RSMI_STATUS_FILE_ERROR ||
-                                   err == RSMI_STATUS_NOT_YET_IMPLEMENTED) {
+                                   err == RSMI_STATUS_NOT_SUPPORTED) {
       IF_VERB(STANDARD) {
         std::cout <<
             "\t**rsmi_dev_od_volt_info_get: Not supported on this machine"