[Mellanox] Provide default implementation for sfp error description when CMIS host management is enabled (#17294)

- Why I did it
Provide a dummy implementation for SFP error description when CMIS host management is enabled. A future feature shall be raised to implement SFP error description for such mode.

- How I did it
if SFP is under software control, provide "Not supported" as error description
if SFP is under initialization, provide "Initializing" as error description

- How to verify it
unit test

platform/mellanox/mlnx-platform-api/sonic_platform/device_data.py

@@ -234,3 +234,12 @@ class DeviceDataManager:
             # Currently, only fetching BIOS version is supported
             return ComponentCPLDSN2201.get_component_list()
         return ComponentCPLD.get_component_list()
+
+    @classmethod
+    @utils.read_only_cache()
+    def is_independent_mode(cls):
+        from sonic_py_common import device_info
+        _, hwsku_dir = device_info.get_paths_to_platform_and_hwsku_dirs()
+        sai_profile_file = os.path.join(hwsku_dir, 'sai.profile')
+        data = utils.read_key_value_file(sai_profile_file, delimeter='=')
+        return data.get('SAI_INDEPENDENT_MODULE_MODE') == '1'

platform/mellanox/mlnx-platform-api/sonic_platform/sfp.py

@@ -139,6 +139,7 @@ SFP_SYSFS_STATUS = 'status'
 SFP_SYSFS_STATUS_ERROR = 'statuserror'
 SFP_SYSFS_PRESENT = 'present'
 SFP_SYSFS_RESET = 'reset'
+SFP_SYSFS_HWRESET = 'hw_reset'
 SFP_SYSFS_POWER_MODE = 'power_mode'
 SFP_SYSFS_POWER_MODE_POLICY = 'power_mode_policy'
 POWER_MODE_POLICY_HIGH = 1
@@ -638,6 +639,12 @@ class SFP(NvidiaSFPCommon):
         Returns:
             The error description
         """
+        try:
+            if self.is_sw_control():
+                return 'Not supported'
+        except:
+            return self.SFP_STATUS_INITIALIZING
+
         oper_status, error_code = self._get_module_info(self.sdk_index)
         if oper_status == SX_PORT_MODULE_STATUS_INITIALIZING:
             error_description = self.SFP_STATUS_INITIALIZING
@@ -792,6 +799,21 @@ class SFP(NvidiaSFPCommon):
                 self._xcvr_api.get_tx_fault = self.get_tx_fault
         return self._xcvr_api
 
+    def is_sw_control(self):
+        if not DeviceDataManager.is_independent_mode():
+            return False
+
+        db = utils.DbUtils.get_db_instance('STATE_DB')
+        control_type = db.get('STATE_DB', f'TRANSCEIVER_MODULES_MGMT|{self.sdk_index}', 'control_type')
+        control_file_value = utils.read_int_from_file(f'/sys/module/sx_core/asic0/module{self.sdk_index}/control')
+
+        if control_type == 'SW_CONTROL' and control_file_value == 1:
+            return True
+        elif control_type == 'FW_CONTROL' and control_file_value == 0:
+            return False
+        else:
+            raise Exception(f'Module {self.sdk_index} is in initialization, please retry later')
+
 
 class RJ45Port(NvidiaSFPCommon):
     """class derived from SFP, representing RJ45 ports"""

platform/mellanox/mlnx-platform-api/sonic_platform/utils.py

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2020-2021 NVIDIA CORPORATION & AFFILIATES.
+# Copyright (c) 2020-2023 NVIDIA CORPORATION & AFFILIATES.
 # Apache-2.0
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -19,6 +19,7 @@ import functools
 import subprocess
 import json
 import sys
+import threading
 import time
 import os
 from sonic_py_common import device_info
@@ -100,15 +101,15 @@ def read_float_from_file(file_path, default=0.0, raise_exception=False, log_func
     return read_from_file(file_path=file_path, target_type=float, default=default, raise_exception=raise_exception, log_func=log_func)
 
 
-def _key_value_converter(content):
+def _key_value_converter(content, delimeter):
     ret = {}
     for line in content.splitlines():
-        k,v = line.split(':')
+        k,v = line.split(delimeter)
         ret[k.strip()] = v.strip()
     return ret
 
 
-def read_key_value_file(file_path, default={}, raise_exception=False, log_func=logger.log_error):
+def read_key_value_file(file_path, default={}, raise_exception=False, log_func=logger.log_error, delimeter=':'):
     """Read file content and parse the content to a dict. The file content should like:
        key1:value1
        key2:value2
@@ -119,7 +120,8 @@ def read_key_value_file(file_path, default={}, raise_exception=False, log_func=l
         raise_exception (bool, optional): If exception should be raised or hiden. Defaults to False.
         log_func (optional): logger function.. Defaults to logger.log_error.
     """
-    return read_from_file(file_path=file_path, target_type=_key_value_converter, default=default, raise_exception=raise_exception, log_func=log_func)
+    converter = lambda content: _key_value_converter(content, delimeter)
+    return read_from_file(file_path=file_path, target_type=converter, default=default, raise_exception=raise_exception, log_func=log_func)
 
 
 def write_file(file_path, content, raise_exception=False, log_func=logger.log_error):
@@ -285,3 +287,26 @@ def wait_until(predict, timeout, interval=1, *args, **kwargs):
         time.sleep(interval)
         timeout -= interval
     return False
+
+
+class DbUtils:
+    lock = threading.Lock()
+    db_instances = threading.local()
+
+    @classmethod
+    def get_db_instance(cls, db_name, **kargs):
+        try:
+            if not hasattr(cls.db_instances, 'data'):
+                with cls.lock:
+                    if not hasattr(cls.db_instances, 'data'):
+                        cls.db_instances.data = {}
+
+            if db_name not in cls.db_instances.data:
+                from swsscommon.swsscommon import SonicV2Connector
+                db = SonicV2Connector(use_unix_socket_path=True)
+                db.connect(db_name)
+                cls.db_instances.data[db_name] = db
+            return cls.db_instances.data[db_name]
+        except Exception as e:
+            logger.log_error(f'Failed to get DB instance for DB {db_name} - {e}')
+            raise e

platform/mellanox/mlnx-platform-api/tests/test_device_data.py

@@ -52,5 +52,14 @@ class TestDeviceData:
     def test_get_bios_component(self):
         assert DeviceDataManager.get_bios_component() is not None
 
+    @mock.patch('sonic_py_common.device_info.get_paths_to_platform_and_hwsku_dirs', mock.MagicMock(return_value=('', '/tmp')))
+    @mock.patch('sonic_platform.device_data.utils.read_key_value_file')
+    def test_is_independent_mode(self, mock_read):
+        mock_read.return_value = {}
+        assert not DeviceDataManager.is_independent_mode()
+        mock_read.return_value = {'SAI_INDEPENDENT_MODULE_MODE': '1'}
+        assert DeviceDataManager.is_independent_mode()
+
+
 
 

platform/mellanox/mlnx-platform-api/tests/test_sfp.py

@@ -54,17 +54,18 @@ class TestSfp:
         assert sfp.sdk_index == 1
         assert sfp.index == 5
 
+    @mock.patch('sonic_platform.sfp.SFP.is_sw_control')
     @mock.patch('sonic_platform.sfp.SFP.read_eeprom', mock.MagicMock(return_value=None))
     @mock.patch('sonic_platform.sfp.SFP.shared_sdk_handle', mock.MagicMock(return_value=2))
     @mock.patch('sonic_platform.sfp.SFP._get_module_info')
     @mock.patch('sonic_platform.chassis.Chassis.get_num_sfps', mock.MagicMock(return_value=2))
     @mock.patch('sonic_platform.chassis.extract_RJ45_ports_index', mock.MagicMock(return_value=[]))
-    def test_sfp_get_error_status(self, mock_get_error_code):
+    def test_sfp_get_error_status(self, mock_get_error_code, mock_control):
         chassis = Chassis()
 
         # Fetch an SFP module to test
         sfp = chassis.get_sfp(1)
-
+        mock_control.return_value = False
         description_dict = sfp._get_error_description_dict()
         for error in description_dict.keys():
             mock_get_error_code.return_value = (SX_PORT_MODULE_STATUS_PLUGGED_WITH_ERROR, error)
@@ -88,6 +89,14 @@ class TestSfp:
 
             assert description == expected_description
 
+        mock_control.return_value = True
+        description = sfp.get_error_description()
+        assert description == 'Not supported'
+
+        mock_control.side_effect = RuntimeError('')
+        description = sfp.get_error_description()
+        assert description == 'Initializing'
+
     @mock.patch('sonic_platform.sfp.SFP._get_page_and_page_offset')
     @mock.patch('sonic_platform.sfp.SFP._is_write_protected')
     def test_sfp_write_eeprom(self, mock_limited_eeprom, mock_get_page):
@@ -278,6 +287,31 @@ class TestSfp:
         assert sfp.get_transceiver_threshold_info()
         sfp.reinit()
 
+    @mock.patch('sonic_platform.utils.read_int_from_file')
+    @mock.patch('sonic_platform.device_data.DeviceDataManager.is_independent_mode')
+    @mock.patch('sonic_platform.utils.DbUtils.get_db_instance')
+    def test_is_sw_control(self, mock_get_db, mock_mode, mock_read):
+        sfp = SFP(0)
+        mock_mode.return_value = False
+        assert not sfp.is_sw_control()
+        mock_mode.return_value = True
+
+        mock_db = mock.MagicMock()
+        mock_get_db.return_value = mock_db
+        mock_db.get = mock.MagicMock(return_value=None)
+        with pytest.raises(Exception):
+            sfp.is_sw_control()
+
+        mock_read.return_value = 0
+        mock_db.get.return_value = 'FW_CONTROL'
+        assert not sfp.is_sw_control()
+        mock_read.return_value = 1
+        mock_db.get.return_value = 'SW_CONTROL'
+        assert sfp.is_sw_control()
+        mock_read.return_value = 0
+        with pytest.raises(Exception):
+            sfp.is_sw_control()
+
     @mock.patch('sonic_platform.utils.is_host', mock.MagicMock(side_effect = [True, True, False, False]))
     @mock.patch('subprocess.check_output', mock.MagicMock(side_effect = ['True', 'False']))
     @mock.patch('sonic_platform.sfp.SFP._get_lpmode', mock.MagicMock(side_effect = [True, False]))

platform/mellanox/mlnx-platform-api/tests/test_utils.py

@@ -191,3 +191,6 @@ class TestUtils:
         mock_os_open = mock.mock_open(read_data='a:b')
         with mock.patch('sonic_platform.utils.open', mock_os_open):
             assert utils.read_key_value_file('some_file') == {'a':'b'}
+        mock_os_open = mock.mock_open(read_data='a=b')
+        with mock.patch('sonic_platform.utils.open', mock_os_open):
+            assert utils.read_key_value_file('some_file', delimeter='=') == {'a':'b'}