On Fri, 12 Sept 2025 at 12:03, Ilya Maximets <[email protected]> wrote:
> On 9/10/25 4:46 PM, David Marchand wrote:
> > +AT_CHECK([ovs-vsctl set Interface p1 options:ol_tso_segsz=460])
> > +
> > +AT_CHECK([ovs-appctl netdev-dummy/receive p1 "${flow_s}"])
> > +AT_CHECK([ovs-pcap p2-tx.pcap | wc -l], [0], [2
> > +])
> > +AT_CHECK([ovs-pcap p3-tx.pcap | wc -l], [0], [0
> > +])
> > +
> > +AT_CHECK([ovs-appctl netdev-dummy/receive p1 "${flow_s}" --len 974])
> > +AT_CHECK([ovs-pcap p2-tx.pcap | wc -l], [0], [4
> > +])
> > +AT_CHECK([ovs-pcap p3-tx.pcap | wc -l], [0], [0
> > +])
> > +
> > +AT_CHECK([ovs-vsctl set Interface p1 options:ol_tso_segsz=500])
>
> I think, 462 is a more appropriate value here to test the boundaries.
> It's just above the threshold and the second segment will still fit,
> but should be forwarded along with the first one.
>
> I made these two changes to save some iterations and applied the fix.
> Backported down to 3.3.  Thanks!

Ok for me.

>
> Note: backporting this kind of changes is a bit of a pain due to each
> branch having different implementation of offloads.

I had prepared the 3.5 backport (for Maxime), but thanks for handling
it, and down to 3.3 no less :-).


-- 
David Marchand

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to