From: Dave Taht <hidden> Date: 2021-09-30 18:10:49
To me, the symptoms of this problem (a packet storm) smelt like the
lack of bql leading to very bursty tcp behavior
and enormous tcp RTT inflation in this driver which I observed 6+ months back.
now that it seems to be fixed (?), could you re-run the netperf 128
stream test and share a packet capture? (try with pfifo_fast or
fq_codel
at the qdisc), and sch_fq as a control (which was working correctly).
thx.