[chassis]: Found SAI_SWITCH_ATTR_REGISTER_READ and SAI_SWITCH_ATTR_REGISTER_READ Error messages via LogAnalyzer

[selvipal]

As a part of T2 Log Analyzer enablement/running, we found out following error messages occur on LCs, which led to LogAnalyzer failure:

swss0#orchagent: updateNotifications: pointer for SAI_SWITCH_ATTR_REGISTER_READ is not handled, FIXME!
ERR gbsyncd0#GBSAI[13] updateNotifications: pointer for SAI_SWITCH_ATTR_REGISTER_WRITE is not handled, FIXME!

The message SAI_SWITCH_ATTR_REGISTER_WRITE/ SAI_SWITCH_ATTR_REGISTER_READ trigger and occurrence is from PI (SONiC) specific implementation.
The message occurs from both syncd and gbsyncd container.
This code flow involves set of callbacks which were already implemented by PI and the callbacks for the case of SAI_SWITCH_ATTR_REGISTER_READ and SAI_SWITCH_ATTR_REGISTER_WRITE are not yet implemented by PI.

Code Ref point (the above-mentioned WRITE/READ cases not handled):
https://github.com/sonic-net/sonic-sairedis/blob/master/syncd/NotificationHandler.cpp#L90
https://github.com/sonic-net/sonic-sairedis/blob/master/lib/Switch.cpp#L101

[1] Can you have someone from SONiC gbsyncd/syncd look into this?
[2] Please help update as to what’s the use case of these callback in PI layer

[kcudnik]

im not sure if this was planned to be supported, but if yes, then it will be addressed, but i think there is no timeline for this yet, since read/write pointers must be updated locally inside syncd, they cant be passed via orchagent, since that different process

[anamehra]

Hi @abdosi , raised this for "SAI_SWITCH_ATTR_REGISTER_WRITE" error. If this should be moved to sonic-buildimage, please help to move.

[bmridul]

@abdosi, pls comment. Issue has been open for some time