[xcvr api] cmis.get_error_description does not show correct error status when there is split

[Junchao-Mellanox]

Description

cmis.get_error_description always checks all lanes of the module. If any lane returns an error state, the function returns error state. For example, the port are splitted to 8, only the first sub port is admin up.

Ethernet496      496     100G   9100     rs   etp63a   trunk      up       up  OSFP 8X Pluggable Transceiver         off
Ethernet497      497     100G   9100     rs   etp63b  routed    down     down  OSFP 8X Pluggable Transceiver         off
Ethernet498      498     100G   9100     rs   etp63c  routed    down     down  OSFP 8X Pluggable Transceiver         off
Ethernet499      499     100G   9100     rs   etp63d  routed    down     down  OSFP 8X Pluggable Transceiver         off
Ethernet500      500     100G   9100     rs   etp63e  routed    down     down  OSFP 8X Pluggable Transceiver         off
Ethernet501      501     100G   9100     rs   etp63f  routed    down     down  OSFP 8X Pluggable Transceiver         off
Ethernet502      502     100G   9100     rs   etp63g  routed    down     down  OSFP 8X Pluggable Transceiver         off
Ethernet503      503     100G   9100     rs   etp63h  routed    down     down  OSFP 8X Pluggable Transceiver         off

get_error_description would return DataPathInitialized for all sub ports:

sudo sfputil show error-status --fetch-from-hardware
Ethernet496  DataPathInitialized
Ethernet497  DataPathInitialized
Ethernet498  DataPathInitialized
Ethernet499  DataPathInitialized
Ethernet500  DataPathInitialized
Ethernet501  DataPathInitialized
Ethernet502  DataPathInitialized
Ethernet503  DataPathInitialized

However, the actual datapath state of Ethernet496 is DataPathActivated:

>>> import sonic_platform
>>> c=sonic_platform.platform.Platform().get_chassis()
>>> s = c.get_sfp(63)
>>> api = s.get_xcvr_api()
>>> d = api.get_datapath_state()
>>> d
{'DP1State': 'DataPathActivated', 'DP2State': 'DataPathInitialized', 'DP3State': 'DataPathInitialized', 'DP4State': 'DataPathInitialized', 'DP5State': 'DataPathInitialized', 'DP6State': 'DataPathInitialized', 'DP7State': 'DataPathInitialized', 'DP8State': 'DataPathInitialized'}

Reproduce Steps

1. Split port to 2 or 4 or 8
2. Set first sub port admin up, rest of them are admin down
3. Call sudo sfputil show error-status --fetch-from-hardware

[bingwang-ms]

Hi @Junchao-Mellanox, so if any subport is admin down, the CLI will return error state?
I remember we discussed the issue about this CLI earlier, and I was surprised that I can't repro the issue some how. Maybe this is the root cause?

[mihirpat1]

We will need to modify the API definition to fix the current behavior.
The get_error_description currently iterates over all lanes of the physical port and returns if any lane is not in DataPathActivated or ConfigSuccess state.

Hence, we need to decide on if we want the API to care about the lane mask and then return accordingly or do we want further changes to the behavior of this API for breakout ports.

[Junchao-Mellanox]

Hi @bingwang-ms , it could be.

[r12f]

@Pterosaur and @Janetxxx for viz

[dgsudharsan]

@mihirpat1 Do you have an ETA?

[mihirpat1]

@prgeor for viz
@dgsudharsan Can you please work with @Junchao-Mellanox to close on this?
Please refer to the below for further information
#21878 (comment)

[Junchao-Mellanox]

Hi @mihirpat1 , I am not sure I understand your previous comment. Do you have a PR for this fix?

[mihirpat1]

@Junchao-Mellanox I don't have a PR for this.
Would you be able to fix this GI?

[Junchao-Mellanox]

Hi @mihirpat1 , I am fully booked recently, not sure I can handle this issue anytime soon...

[mihirpat1]

@prgeor
@Junchao-Mellanox Can you please discuss internally on this issue since I am not sure on the use case of this CLI/API and this was initially added via sonic-net/SONiC#586