[SWDEV-560044]: [AMDSMI][CPU] Update AMDSMI as per latest ESMI Driver (#763)

[AMDSMI][CPU] Update AMDSMI as per latest ESMI Driver, 1) hsmp_acpi 2) amd_hsmp 3) hsmp_common Signed-off-by: Muthusamy Ramalingam <muthusamy.ramalingam@amd.com> Signed-off-by: Arif, Maisam <Maisam.Arif@amd.com> Co-authored-by: ssaka_amdeng <SitharamMurthy.Saka@amd.com> [ROCm/amdsmi commit: b4b3539631]
2025-11-18 01:15:43 +05:30
@@ -67,16 +67,29 @@ Full documentation for amd_smi_lib is available at [https://rocm.docs.amd.com/pr

 ### Changed

- **`amd-smi set --power-cap` now requires specification of the power cap type**.  
-  - Command now takes the form: `amd-smi set --power-cap <power-cap-type> <new-cap>`. Acceptable power cap types are "ppt0" and "ppt1".  
-  Ex.
+- **The `amd-smi` command now shows hsmp rather than amd_hsmp**.  
+  - The hsmp driver version can be shown without the amdgpu version using `amd-smi version -c`
+
+  ```console
+   $ amd-smi version
+   AMDSMI Tool: 24.7.1+b446d6c-dirty | AMDSMI Library version: 24.7.2.0 | ROCm version: N/A | amdgpu version: 6.10.10 | hsmp version: 2.2
+
+   $ amd-smi version -c
+   AMDSMI Tool: 24.7.1+b446d6c-dirty | AMDSMI Library version: 24.7.2.0 | ROCm version: N/A | hsmp version: 2.2
+   ...
+  ```
+
+- **`amd-smi set --power-cap` now requires sepcification of the power cap type**.  
+  - Command now takes the form: `amd-smi set --power-cap <power-cap-type> <new-cap>`
+  - Acceptable power cap types are "ppt0" and "ppt1"
+
+  ```console
+  $ sudo amd-smi set --power-cap ppt1 1150
+  GPU: 0
+    POWERCAP: Successfully set ppt1 power cap to 1150W
+    ...
+  ```

-    ```console
-    $ sudo amd-smi set --power-cap ppt1 1150
-    GPU: 0
-      POWERCAP: Successfully set PPT1 power cap to 1150W
-      ...
-    ```  
 - **`amd-smi reset --power-cap` will attempt to reset both power caps**.  
  - When using the reset command, both PPT0 and PPT1 power caps will be reset to their default values. If a device only has PPT0, then only PPT0 will be reset.  
    Ex.
@@ -1402,7 +1415,7 @@ Functions affected by struct change are:
 - **Corrected CLI CPU argument name**.  
  - `--cpu-pwr-svi-telemtry-rails` to `--cpu-pwr-svi-telemetry-rails`

- **Added amdgpu driver version and amd_hsmp driver version to `amd-smi version` command**.  
+- **Added amdgpu driver version and amd_hsmp driver version to `amd-smi version` command**.
  - The `amd-smi version` command can now also display the amdgpu driver version using the `-g` flag.
  - The amd_hsmp driver version can also be displayed using the `-c` flag.
  - The new default for the `version` command is to display all the version information, including both amdgpu and amd_hsmp driver versions.
@@ -96,7 +96,7 @@ class AMDSMICommands():
            except amdsmi_exception.AmdSmiLibraryException as e:
                if e.err_code in (amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_INIT,
                                amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NO_DRV):
-                    logging.info('Unable to detect any CPU devices, check amd_hsmp version and module status (sudo modprobe amd_hsmp)')
+                    logging.info('Unable to detect any CPU devices, check amd_hsmp (or) hsmp_acpi version and module status (sudo modprobe amd_hsmp (or) sudo modprobe hsmp_acpi)')
                else:
                    raise e

@@ -112,7 +112,7 @@ class AMDSMICommands():

            if len(self.cpu_handles) == 0 and len(self.core_handles) == 0:
                # No CPU's found post amd_hsmp driver initialization
-                logging.error('Unable to detect any CPU devices, check amd_hsmp version and module status (sudo modprobe amd_hsmp)')
+                logging.error('Unable to detect any CPU devices, check amd_hsmp (or) hsmp_acpi version and module status (sudo modprobe amd_hsmp (or) sudo modprobe hsmp_acpi)')
                exit_flag = True

        self.convert_clock_type = {
@@ -200,7 +200,7 @@ class AMDSMICommands():
            if args.gpu_version:
                human_readable_output = human_readable_output + f" | amdgpu version: {gpu_version_str}"
            if args.cpu_version:
-                human_readable_output = human_readable_output + f" | amd_hsmp version: {cpu_version_str}"
+                human_readable_output = human_readable_output + f" | hsmp version: {cpu_version_str}"
            # Custom human readable handling for version
            if self.logger.destination == 'stdout':
                print(human_readable_output)
@@ -2988,7 +2988,7 @@ class AMDSMICommands():
            try:
                bandwidth = amdsmi_interface.amdsmi_get_cpu_current_io_bandwidth(args.cpu,
                                                                                    int(args.cpu_io_bandwidth[0][0]),
-                                                                                    args.cpu_io_bandwidth[0][1])
+                                                                                    args.cpu_io_bandwidth[0][1].upper())
                static_dict["io_bandwidth"]["band_width"] = bandwidth
            except amdsmi_exception.AmdSmiLibraryException as e:
                static_dict["io_bandwidth"]["band_width"] = "N/A"
@@ -2998,7 +2998,7 @@ class AMDSMICommands():
            try:
                bandwidth = amdsmi_interface.amdsmi_get_cpu_current_xgmi_bw(args.cpu,
                                                                            int(args.cpu_xgmi_bandwidth[0][0]),
-                                                                            args.cpu_xgmi_bandwidth[0][1])
+                                                                            args.cpu_xgmi_bandwidth[0][1].upper())
                static_dict["xgmi_bandwidth"]["band_width"] = bandwidth
            except amdsmi_exception.AmdSmiLibraryException as e:
                static_dict["xgmi_bandwidth"]["band_width"] = "N/A"
@@ -239,11 +239,11 @@ class AMDSMIHelpers():
        except amdsmi_interface.AmdSmiLibraryException as e:
            if e.err_code in (amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_INIT,
                              amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_DRIVER_NOT_LOADED):
-                logging.info('Unable to get device choices, driver not initialized (amd_hsmp not found in modules)')
+                logging.info('Unable to get device choices, driver not initialized (amd_hsmp  or hsmp_acpi not found in modules)')
            else:
                raise e
        if len(cpu_handles) == 0:
-            logging.info('Unable to find any devices, check if driver is initialized (amd_hsmp not found in modules)')
+            logging.info('Unable to find any devices, check if driver is initialized (amd_hsmp or hsmp_acpi not found in modules)')
        else:
            # Handle spacing for the gpu_choices_str
            max_padding = int(math.log10(len(cpu_handles))) + 1
@@ -285,11 +285,11 @@ class AMDSMIHelpers():
        except amdsmi_interface.AmdSmiLibraryException as e:
            if e.err_code in (amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_INIT,
                              amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_DRIVER_NOT_LOADED):
-                logging.info('Unable to get device choices, driver not initialized (amd_hsmp not found in modules)')
+                logging.info('Unable to get device choices, driver not initialized (amd_hsmp  or hsmp_acpi not found in modules)')
            else:
                raise e
        if len(core_handles) == 0:
-            logging.info('Unable to find any devices, check if driver is initialized (amd_hsmp not found in modules)')
+            logging.info('Unable to find any devices, check if driver is initialized (amd_hsmp or hsmp_acpi  not found in modules)')
        else:
            # Handle spacing for the gpu_choices_str
            max_padding = int(math.log10(len(core_handles))) + 1
@@ -62,10 +62,9 @@ def check_amdgpu_driver():


 def check_amd_hsmp_driver():
-    """ Returns true if amd_hsmp is found in the list of initialized modules """
-    amd_cpu_status_file = Path("/sys/module/amd_hsmp/initstate")
+    """ Returns true if amd_hsmp or hsmp_acpi is found in the list of initialized modules """
+    amd_cpu_status_file = Path("/dev/hsmp")
    if amd_cpu_status_file.exists():
-        if amd_cpu_status_file.read_text(encoding="ascii").strip() == "live":
            return True
    return False

@@ -73,7 +72,7 @@ def check_amd_hsmp_driver():
 def amdsmi_cli_init():
    """ Initializes AMDSMI Library for the CLI

-    Checks for the presence of the amdgpu and amd_hsmp drivers and initializes the
+    Checks for the presence of the amdgpu, amd_hsmp or hsmp_acpi drivers and initializes the
    AMD SMI library based on the live drivers found.

    Return:
@@ -85,13 +84,13 @@ def amdsmi_cli_init():
    init_flag = amdsmi_interface.AmdSmiInitFlags.INIT_ALL_PROCESSORS
    if check_amdgpu_driver() and check_amd_hsmp_driver():
        init_flag = amdsmi_interface.AmdSmiInitFlags.INIT_AMD_APUS
-        logging.debug("Both amdgpu and amd_hsmp driver's initstate is live")
+        logging.debug("Both amdgpu , amd_hsmp or hsmp_acpi  driver's initstate is live")
        try:
            amdsmi_interface.amdsmi_init(init_flag)
        except (amdsmi_interface.AmdSmiLibraryException, amdsmi_interface.AmdSmiParameterException) as e:
            if e.err_code in (amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_INIT,
                              amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_DRIVER_NOT_LOADED):
-                logging.error("Drivers not loaded (amdgpu and amd_hsmp drivers not found in modules)")
+                logging.error("Drivers not loaded (amdgpu, amd_hsmp or hsmp_acpi drivers not found in modules)")
                sys.exit(-1)
            else:
                raise e
@@ -107,20 +106,20 @@ def amdsmi_cli_init():
                sys.exit(-1)
            else:
                raise e
-        logging.debug("amdgpu driver initialized successfully, but amd_hsmp initstate was not live")
+        logging.debug("amdgpu driver initialized successfully, but amd_hsmp or hsmp_acpi initstate was not live")
    elif check_amd_hsmp_driver():
        init_flag = amdsmi_interface.AmdSmiInitFlags.INIT_AMD_CPUS
-        logging.debug("amd_hsmp driver initstate is live")
+        logging.debug("amd_hsmp or hsmp_acpi driver initstate is live")
        try:
            amdsmi_interface.amdsmi_init(init_flag)
        except (amdsmi_interface.AmdSmiLibraryException, amdsmi_interface.AmdSmiParameterException) as e:
            if e.err_code in (amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_NOT_INIT,
                              amdsmi_interface.amdsmi_wrapper.AMDSMI_STATUS_DRIVER_NOT_LOADED):
-                logging.error("Driver not loaded (amd_hsmp not found in modules)")
+                logging.error("Driver not loaded (amd_hsmp or hsmp_acpi not found in modules)")
                sys.exit(-1)
            else:
                raise e
-        logging.debug("amd_hsmp driver initialized successfully, but amdgpu initstate was not live")
+        logging.debug("amd_hsmp or hsmp_acpi driver initialized successfully, but amdgpu initstate was not live")

    logging.debug(f"AMDSMI initialized with atleast one driver successfully | init flag: {init_flag}")

@@ -754,7 +754,7 @@ class AMDSMIParser(argparse.ArgumentParser):

        # help info
        gpu_version_help = "Display the current amdgpu driver version"
-        cpu_version_help = "Display the current amd_hsmp driver version"
+        cpu_version_help = "Display the current amd_hsmp or hsmp_acpi driver version"

        # Add GPU and CPU version Arguments
        version_parser.add_argument('-g', '--gpu_version', action='store_true', required=False, help=gpu_version_help, default=None)
@@ -37,8 +37,7 @@ AMD SMI library can run on AMD ROCm supported platforms. Refer to
 for more information.
 <!--https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/system-requirements.html-->

-To run the AMD SMI library, the `amdgpu` driver and the `amd_hsmp` driver need
-to be installed. Optionally, `libdrm` can be installed to query firmware
+To run the AMD SMI library, the `amdgpu` driver and the `amd_hsmp` or `hsmp_acpi` driver need to be installed. Optionally, `libdrm` can be installed to query firmware
 information and hardware IPs.

 ### Python interface and CLI tool prerequisites
@@ -45,7 +45,7 @@
 #define AMDGPU_INITSTATE_FILE  "/sys/module/amdgpu/initstate"

 #define AMDHSMP_DRIVER_NAME    "AMDHSMPDriver"
-#define AMDHSMP_INITSTATE_FILE "/sys/module/amd_hsmp/initstate"
+#define AMDHSMP_INITSTATE_FILE "/dev/hsmp"

 static uint32_t num_apuSockets              = GOAMDSMI_VALUE_0;
 static uint32_t num_cpuSockets              = GOAMDSMI_VALUE_0;
@@ -98,12 +98,13 @@ void TestBase::SetUp(uint64_t init_flags) {
      }
    }

-    // Returns true if amd_hsmp is found in the list of initialized modules
-    bool found_amd_hsmp = CheckModule("/sys/module/amd_hsmp/initstate", "live");
+    // Returns true if amd_hsmp or hsmp_acpi is found in the list of initialized modules
+    struct stat buffer;
+    bool found_amd_hsmp = (stat("/dev/hsmp", &buffer)==0);
    if (!found_amd_hsmp) {
      IF_VERB(STANDARD) {
        std::cerr << "ERROR: Unable to get devices, driver not initialized (amd_hsmp not found in modules)" << std::endl;
-        std::cerr << "ERROR: Unable to detect any CPU devices, check amd_hsmp version and module status (sudo modprobe amd_hsmp)" << std::endl;
+        std::cerr << "ERROR: Unable to detect any CPU devices, check amd_hsmp (or) hsmp_acpi version and module status (sudo modprobe amd_hsmp (or) sudo modprobe hsmp_acpi)" << std::endl;
      }
    }