Bug #14267
openena panic triggered by static networking
0%
Description
Testing Tribblix m25.2 on an EC2 t3.micro instance. All going well until I try to enable static networking:
/usr/sbin/svcadm disable -s svc:/network/physical:nwam
/usr/sbin/svcadm enable -s svc:/network/physical:default
/usr/sbin/ipadm create-if ${IFACE}
/usr/sbin/ipadm create-addr -T static -a $IP1 ${IFACE}/v4
/usr/sbin/route -p add net default $IPROUTE
At least, that's what the script was supposed to do - I don't know how far it got.
At which point the system hung and rebooted, ending up in a reboot loop. (There might have been a brief "could not initialize" or similar flashed up before it went, I couldn't guarantee what it said.)
At boot I see:
ena: WARNING: ena0: unexpected status code: 6
ena: WARNING: ena0: failed Destroy SQ: 5
ena: WARNING: ena0: failed to destroy Rx SQ : 5
(So the 6 is ENAHW_RESP_UNKNOWN_ERROR which the code converts to EIO which is the 5 in the last 2 messages.)
I think I've copied these correctly, including the lack of a 2nd number before the ":" in the last message. But not 100% sure.
Managed to get to the console and drop into kmdb, image of stack trace attached.
The messages logged appear to align with the ena_cleanup_rxq() and ena_destroy_sq() in the stack trace. It's not immediately obvious to me where the ena_free_host_info() comes from though.
Files
Related issues
Updated by Ryan Zezeski about 2 years ago
Thanks for the report Peter. I'm slammed this week and will probably not have time to dig into this at all this week. However, if you, or anyone else, wants to take a stab at fixing this I'll be happy to CR .
Updated by Andy Fiddaman over 1 year ago
- Related to Feature #13689: Want AWS ENA driver added