[device/celestica/x86_64-cel_silverstone-r0] Fix QSFP-DD support #92
Conversation
Signed-off-by: Dante Su <[email protected]>
Signed-off-by: Dante Su <[email protected]>
Signed-off-by: Dante Su <[email protected]>
… != ModuleReady for 3+ seconds Signed-off-by: Dante Su <[email protected]>
ds952811
changed the title
|
Just a heads up, the following xcvrd crash observed after multiple attempts of QSFPDD reinsertions, and hence we'll also need PR5 of sonic-platform-common for this. i.e. Nov 18 00:57:55.488171 sonic INFO pmon#supervisord: xcvrd Traceback (most recent call last): |
…te it if necessary Signed-off-by: Dante Su <[email protected]>
In the current implementation, the SFP change events are actually software simulated by polling the presence flags, and the INSERTION events will be be generated upon pmon restart and reboot. And hence, significant packet drops could be observed upon both pmon restart and a warm-reboot, as we'll always fire a software reset to the transceivers. i.e. It's better to drop the QSFPDD software reset from the insertion event Signed-off-by: Dante Su <[email protected]>
Signed-off-by: Dante Su <[email protected]>
|
Can you please share the testing log for Innolight and Eoptolink optics 4x100G mode with Silverstone-128x100 HWSKU? I would like to see the at least link status (up), plug out/in still works, show interface transceiver eeprom -d etc. |
| if name.upper() != 'INNOLIGHT' or part.upper() != 'T-DP4CNT-N00': | ||
| return True | ||
|
|
||
| log_info("PORT {0}: _init_cmis_module".format(port_num)) |
There was a problem hiding this comment.
log port_num and module name for better visibility.
|
Hi Xu,
Attached is the UT for your reference, please note while InnoLight is in
application 2, the Eoptolink is in application 1 (default mode), and the
individual 100G links could work independently
Note: I have no idea how to attached the log in the Github, so I respond to
this email, hope it works.
Best Wishes,
Dante (Kuo-Jung) Su
…On Thu, Nov 19, 2020 at 6:45 AM zhenggen-xu ***@***.***> wrote:
Can you please share the testing log for Innolight and Eoptolink optics
4x100G mode with Silverstone-128x100 HWSKU? I would like to see the at
least link status (up), plug out/in still works, show interface transceiver
eeprom -d etc.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#92 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AMLISKJ5EGMOHIE5NDYZ563SQRFAJANCNFSM4TWTWJFQ>
.
|
1. Add sanity check to application advertisement decoder to prevent infinite loop 2. Create a function wrapper for module-specific CMIS init Signed-off-by: Dante Su <[email protected]>
|
The UT result of 'sfputil.py: address review comments': $ grep pmon /var/log/syslog |
|
You can drag and drop the txt files into comments or better in PR description, I would like to see the output of "show interface status", "show interface transciever eeprom -d" with multiple optical modules in place. |
Signed-off-by: Dante Su <[email protected]>
Note:
|
|
Not sure why the inline response is not displayed on the conversation page, here is my response to the last question:
The get_transceiver_change_event() is invoked by the infinite loop of the xcvrd, and no matter if a insertion/removal event generated or not, it will always poll the module state and trigger a software reset when the failure count > 2. The insertion/removal event is generated at line 503, there is an in-memory presence table for the change events |
Commits included: * src/sonic-py-swsssdk 748c404...1ea30d2 (1): > Fix bug: ConfigDBConnector.get_table does not work in python3 (#92)
- What I did
i.e. This is a replacement of PR91, and you don't need PR91 with this PR merged.
10.sfputil.py: Monitoring the QSFP-DD initialization state, and reinitiate it if necessary
- How I did it
See - What I did
- How to verify it
- Description for the changelog
- A picture of a cute animal (not mandatory but encouraged)