Fix rare timing issue in test_CmisManagerTask_task_worker_decommission

[longhuan-cisco]

Description

If the CPU of build server is very fast, testcase test_CmisManagerTask_task_worker_decommission might have a chance to fail at

sonic-platform-daemons/sonic-xcvrd/tests/test_xcvrd.py

Line 3441 in d555def

assert get_cmis_state_from_state_db('Ethernet0', task.xcvr_table_helper.get_status_sw_tbl(task.get_asic_id('Ethernet0'))) == CMIS_STATE_FAILED

The reason is:

• Test code incorrectly assumes update_cmis_state_expiration_time(lport, duration_seconds=0.0) will set cmis_expired as current_time (i.e. current_time + 0.0, meaning expired immediately), however there's always a CMIS_EXPIRATION_BUFFER_MS (2ms) getting added on top of it.
  

  sonic-platform-daemons/sonic-xcvrd/xcvrd/xcvrd.py

  Lines 1084 to 1094 in d555def

  	def update_cmis_state_expiration_time(self, lport, duration_seconds):
  	"""
  	Set the CMIS expiration time for the given logical port
  	in the port dictionary.
  	Args:
  	lport: Logical port name
  	duration_seconds: Duration in seconds for the expiration
  	"""
  	self.port_dict[lport]['cmis_expired'] = datetime.datetime.now() + \
  	datetime.timedelta(seconds=duration_seconds) + \
  	datetime.timedelta(milliseconds=self.CMIS_EXPIRATION_BUFFER_MS)

e.g. if current_time is 01:36:25.528, calling self.update_cmis_state_expiration_time(lport, 0.0), will set cmis_expired to 01:36:25.530, i.e. 2ms ahead of current time.

When testcase is testing timeout for 'ConfigSuccess' scenario:

sonic-platform-daemons/sonic-xcvrd/xcvrd/xcvrd.py

Lines 1435 to 1441 in d555def

	elif state == CMIS_STATE_DP_INIT:
	if not self.check_config_error(api, host_lanes_mask, ['ConfigSuccess']):
	if self.is_timer_expired(expired):
	self.log_notice("{}: timeout for 'ConfigSuccess', current ConfigStatus: "
	"{}".format(lport, list(api.get_config_datapath_hostlane_status().values())))
	self.force_cmis_reinit(lport, retries + 1)
	continue

• On my build server, this testcase always pass. It constantly takes more than 2ms to run through one while/for loop iteration, so after self.update_cmis_state_expiration_time(lport, 0.0), it costs only two while/loop to become is_timer_expired()==True reaching timeout for 'ConfigSuccess', in total it constantly costs only 17 while/for loop iterations to reach CMIS_STATE_FAILED after retries, in which case task_stopping_event.is_set = MagicMock(side_effect=[False]*50 + [True]*2) is more than enough.
• But if the CPU of build server is very fast, it has a chance to be able to run through multiple while/for loop iterations within 2ms, costing more while/for loop iterations to reach timeout for 'ConfigSuccess', which might end up running out of task_stopping_event.is_set = MagicMock(side_effect=[False]*50 + [True]*2) too soon , not being able to even reach CMIS_STATE_FAILED, causing testcase assertion failure.

In below example, CPU is fast, within 2ms, it's able to run through 5 while/for loop iterations. In other words, it costs 5 while/for loop iterations to reach timeout for 'ConfigSuccess' after self.update_cmis_state_expiration_time(lport, 0.0):

2025-07-26 07:45:08.244422 CMIS: Ethernet0: DpDeinit duration 0.0 secs, modulePwrUp duration 0.0 secs
2025-07-26 07:45:08.244469 CMIS: update_cmis_state_expiration_time(0.0) set expired_time to 2025-07-26 07:45:08.246441, current_time 2025-07-26 07:45:08.244441
2025-07-26 07:45:08.244838 CMIS: Ethernet0: 100G, lanemask=0xff, CMIS state=AP_CONFIGURED(decommission*), ... retries=2
2025-07-26 07:45:08.245373 CMIS: Ethernet0: 100G, lanemask=0xff, CMIS state=DP_INIT(decommission*), ... retries=2
2025-07-26 07:45:08.245446 CMIS: is_timer_expired() = False
2025-07-26 07:45:08.245809 CMIS: Ethernet0: 100G, lanemask=0xff, CMIS state=DP_INIT(decommission*), ... retries=2
2025-07-26 07:45:08.245885 CMIS: is_timer_expired() = False
2025-07-26 07:45:08.246266 CMIS: Ethernet0: 100G, lanemask=0xff, CMIS state=DP_INIT(decommission*), ... retries=2
2025-07-26 07:45:08.246345 CMIS: is_timer_expired() = False
2025-07-26 07:45:08.246715 CMIS: Ethernet0: 100G, lanemask=0xff, CMIS state=DP_INIT(decommission*), ... retries=2
2025-07-26 07:45:08.246795 CMIS: is_timer_expired() = True
2025-07-26 07:45:08.246857 CMIS: Ethernet0: timeout for 'ConfigSuccess', current ConfigStatus: ['ConfigRejected', 'ConfigRejected', 'ConfigRejected', 'ConfigRejected', 'ConfigRejected', 'ConfigRejected', 'ConfigRejected', 'ConfigRejected']
2025-07-26 07:45:08.247248 CMIS: Ethernet0: 100G, lanemask=0xff, CMIS state=INSERTED(decommission*), ... retries=3

Motivation and Context

How Has This Been Tested?

Testcase passed

Additional Information (Optional)

[mssonicbld]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[Copilot]

Pull Request Overview

This PR fixes a rare timing issue in the test_CmisManagerTask_task_worker_decommission test that could cause failures on fast CPUs. The issue occurred because the test incorrectly assumed that setting a duration of 0.0 seconds would immediately expire a timer, but there's always a 2ms buffer added to expiration times.

• Replaces duration-based timer mocking with direct timer expiration mocking
• Removes three mock configurations for duration methods that are no longer needed
• Adds a direct mock for is_timer_expired to return True immediately

Comments suppressed due to low confidence (1)

sonic-xcvrd/tests/test_xcvrd.py:3429

• The mock always returns True, which may not adequately test the timer logic. Consider using a side_effect to simulate timer progression or test both expired and non-expired states.

        task.is_timer_expired = MagicMock(return_value=True)

[longhuan-cisco]

Thanks @mihirpat1 for the quick review/approval.

Hi @prgeor, could you please review
Current test_CmisManagerTask_task_worker_decommission() has a low chance to fail if the CPU of build/CICD server is very fast. And it's a simple fix in test code, nothing wrong with xcvrd.py code.

cc @AnoopKamath