Last Modified: Jul 12, 2023
Known Affected Versions:
13.1.0, 188.8.131.52, 184.108.40.206, 220.127.116.11, 18.104.22.168, 22.214.171.124, 126.96.36.199, 188.8.131.52, 184.108.40.206, 13.1.1, 220.127.116.11, 18.104.22.168, 22.214.171.124, 126.96.36.199, 13.1.3, 188.8.131.52, 184.108.40.206, 220.127.116.11, 18.104.22.168, 22.214.171.124, 126.96.36.199, 13.1.4, 188.8.131.52, 13.1.5, 184.108.40.206
Opened: Jan 02, 2018 Severity: 3-Major
If using vmxnet3 or virtio tmm drivers (not unic), if TCP segmentation offload is disabled while processing traffic, a tmm assert may be observed and tmm will restart. The tmm assert appears similar to the following: notice tso ioctl for member:0 of /Common/internal failed with value: 32tso ioctl for member:0 of /Common/external failed with value: 32panic: ./local/net/packet.h:328: Assertion "packet_ref: not free" failed.
Sometimes tmm will restart, which will lead to traffic group failovers. This is a fairly rare situation, and disabling TCP segmentation offloading is not something routinely done, but if done while under load, this failure might occur.
-- Deploy Virtual Edition (VE) using vmxnet3 or virtio tmm drivers. -- Continuously pass traffic through the system. -- Disable the DB variable tm.tcpsegmentationoffload.
There is no workaround at this time.
This issue no longer occurs.