Merge amd-staging into amd-master 20241106

Signed-off-by: Zhang Ava <niandong.zhang@amd.com>
Change-Id: Ib125dee62e5a893871f5c6df7715177973361a02

include/rocm_smi/rocm_smi_utils.h

@@ -82,6 +82,7 @@ int SameFile(const std::string fileA, const std::string fileB);
 bool FileExists(char const *filename);
 std::vector<std::string> globFilesExist(const std::string& filePattern);
 int isRegularFile(std::string fname, bool *is_reg);
+int isReadOnlyForAll(const std::string& fname, bool *is_read_only);
 int ReadSysfsStr(std::string path, std::string *retStr);
 int WriteSysfsStr(std::string path, std::string val);
 bool IsInteger(const std::string & n_str);

python_smi_tools/rocm_smi.py

@@ -89,14 +89,17 @@ validClockNames.sort()
 def driverInitialized():
     """ Returns true if amdgpu is found in the list of initialized modules
     """
-    driverInitialized = ''
-    try:
-        driverInitialized = str(subprocess.check_output("cat /sys/module/amdgpu/initstate |grep live", shell=True))
-    except subprocess.CalledProcessError:
-        pass
-    if len(driverInitialized) > 0:
-        return True
-    return False
+    driverInitialized = False
+    if os.path.exists("/sys/module/amdgpu") :
+        if os.path.exists("/sys/module/amdgpu/initstate"):
+            # amdgpu is loadable module
+            with open("/sys/module/amdgpu/initstate") as initstate:
+                if 'live' in initstate.read():
+                    driverInitialized = True
+        else:
+            # amdgpu is built into the kernel
+            driverInitialized = True
+    return driverInitialized
 
 
 def formatJson(device, log):

src/rocm_smi.cc

@@ -1873,7 +1873,7 @@ rsmi_dev_gpu_clk_freq_set(uint32_t dv_ind,
     return ret;
   }
 
-  int ret_i;
+  rsmi_status_t status;
   amd::smi::DevInfoTypes dev_type;
 
   const auto & clk_type_it = kClkTypeMap.find(clk_type);
@@ -1883,8 +1883,19 @@ rsmi_dev_gpu_clk_freq_set(uint32_t dv_ind,
     return RSMI_STATUS_INVALID_ARGS;
   }
 
-  ret_i = dev->writeDevInfo(dev_type, freq_enable_str);
-  return amd::smi::ErrnoToRsmiStatus(ret_i);
+  status =  amd::smi::ErrnoToRsmiStatus(dev->writeDevInfo(dev_type, freq_enable_str));
+
+  // If an operation is not supported, the dev file, ie /sys/class/drm/card1/device/pp_dpm_pcie
+  // will have read-only perms, and the OS will deny access, before the request hits the driver level
+  if (status == RSMI_STATUS_PERMISSION){
+    bool read_only = false;
+    int perms = amd::smi::isReadOnlyForAll(dev->path(), &read_only);
+    if(read_only){
+      return RSMI_STATUS_NOT_SUPPORTED;
+    }
+  }
+
+  return status;
 
   CATCH
 }

src/rocm_smi_utils.cc

@@ -172,6 +172,24 @@ int isRegularFile(std::string fname, bool *is_reg) {
   return 0;
 }
 
+int isReadOnlyForAll(const std::string& fname, bool *is_read_only){
+  struct stat file_stat;
+  int ret;
+
+  ret = stat(fname.c_str(), &file_stat);
+  if (ret) {
+    return errno;
+  }
+
+  if (is_read_only != nullptr) {
+    *is_read_only = (file_stat.st_mode & (S_IRUSR | S_IRGRP | S_IROTH)) && !(file_stat.st_mode & (S_IWUSR | S_IWGRP | S_IWOTH));
+  } else {
+    ret = 1;
+  }
+
+  return ret;
+}
+
 int WriteSysfsStr(std::string path, std::string val) {
   //  On success, zero is returned.  On error, -1 is returned, and
   //  errno is set to indicate the error.