On Fri, 2018-03-02 at 10:44 -0800, Stephen Hemminger wrote: > On Fri, 2 Mar 2018 21:16:48 +0300 > > Since this is a generic problem why is not fixed in qdisc_drop instead?
AFAIK only netem and tbf might segment GSO packets so far. I am not sure we want to add code in qdisc_drop() that is used under stress on normal qdiscs and are inline code.