[Mellanox] fix sfp eeprom unreadable after switching from SW to FW co…

…ntrol mode (sonic-net#19563) - Why I did it The reading of eeprom instantly after switch the module from sw control to fw control might fail, so add a delay Signed-off-by: Yuanzhe, Liu <[email protected]>
anamehra · Jul 29, 2024 · 66e491d · 66e491d
1 parent 3ff9f4f
commit 66e491d
Show file tree

Hide file tree

Showing 2 changed files with 20 additions and 2 deletions.
diff --git a/platform/mellanox/mlnx-platform-api/sonic_platform/chassis.py b/platform/mellanox/mlnx-platform-api/sonic_platform/chassis.py
@@ -499,16 +499,18 @@ def get_change_event_for_module_host_management_mode(self, timeout):
                     s.on_event(event)
 
                 if s.in_stable_state():
+                    self.sfp_module.SFP.wait_sfp_eeprom_ready([s], 2)
                     s.fill_change_event(port_dict)
                     s.refresh_poll_obj(self.poll_obj, self.registered_fds)
                 else:
                     logger.log_debug(f'SFP {sfp_index} does not reach stable state, state={s.state}')
-                                        
+
             ready_sfp_set = wait_ready_task.get_ready_set()
             for sfp_index in ready_sfp_set:
                 s = self._sfp_list[sfp_index]
                 s.on_event(sfp.EVENT_RESET_DONE)
                 if s.in_stable_state():
+                    self.sfp_module.SFP.wait_sfp_eeprom_ready([s], 2)
                     s.fill_change_event(port_dict)
                     s.refresh_poll_obj(self.poll_obj, self.registered_fds)
                 else:

diff --git a/platform/mellanox/mlnx-platform-api/sonic_platform/sfp.py b/platform/mellanox/mlnx-platform-api/sonic_platform/sfp.py
@@ -500,6 +500,21 @@ def get_presence(self):
             return False
         eeprom_raw = self._read_eeprom(0, 1, log_on_error=False)
         return eeprom_raw is not None
+
+    @classmethod
+    def wait_sfp_eeprom_ready(cls, sfp_list, wait_time):
+        not_ready_list = sfp_list
+
+        while wait_time > 0:
+            not_ready_list = [s for s in not_ready_list if s.state == STATE_FW_CONTROL and s._read_eeprom(0, 2,False) is None]
+            if not_ready_list:
+                time.sleep(0.1)
+                wait_time -= 0.1
+            else:
+                return
+
+        for s in not_ready_list:
+            logger.log_error(f'SFP {s.sdk_index} eeprom is not ready')
 
     # read eeprom specfic bytes beginning from offset with size as num_bytes
     def read_eeprom(self, offset, num_bytes):
@@ -1723,7 +1738,8 @@ def initialize_sfp_modules(cls, sfp_list):
                 logger.log_error(f'SFP {index} is not in stable state after initializing, state={s.state}')
             logger.log_notice(f'SFP {index} is in state {s.state} after module initialization')
 
-
+        cls.wait_sfp_eeprom_ready(sfp_list, 2)
+
 class RJ45Port(NvidiaSFPCommon):
     """class derived from SFP, representing RJ45 ports"""