In list.openbsd.tech, you wrote: > now it turns out it is trivial to have ether_output prepend the > ether_vlan_header instead of the regular ethernet header, which makes > the vlan tagging essentially free in most cases. > > you need a very current src tree to test this, relies on the code > shuffling in if_ethersubr.c I did a few hours ago.
ok (I tested both with and without VLAN_HW_TAGGING.) vlan_start() doesn't touch the ethernet header any longer, so you probably want to modify this comment at the top of the file: * capability is set on the parent. In this case, vlan_start() * will not modify the ethernet header. > + /* should we use the tx tagging hw offload at all? */ Tempting. I'm slightly worried that some chip out there will refuse the frame as oversized or the TX checksum engine will be confused. -- Christian "naddy" Weisgerber na...@mips.inka.de