On Mon, Aug 04, 2003 at 11:53:11AM +0200, Quentin Garnier wrote:
> [...]
> 
> >From a quick glance at the source for the use of ETHERCAP_VLAN_MTU:
> 
> ic/elinkxl.c:	interface reports an error to be ignored
> ic/gem.c:	interface must be told max frame size it should accept
> ic/hme.c:	interface must be told max frame size it should accept
> ic/i82557.c:	interface must be told to receive bad frames, and some checks
> 		are not skipped when VLAN is enabled
> ic/lance.c:	software sanity check of frame length
> ic/tulip.c:	interface reports an error to be ignored
> ic/atw.c:	interface reports an error to be ignored
> 
> pci/if_ti.c:	interface must be told max frame size it should accept
> pci/if_ste.c:	interface must be told to receive large frames
> pci/if_stge.c:	interface must be told max frame size it should accept
> 
> In ic/lance.c I really think the check should be removed: it's a duplicate of
> what is done after in ether_input, without using the macro.

Yes, probably.

> 
> Good old cheap network cards that let the software do all the sanity checking
> work :) (Well, when they last long enough to qualify as old)
> 
> > [...]
> > It's not different. For now you are missing bits to have the PPPoE frames
> > handled properly by some drivers.
> 
> I wonder if it is worth it, since it is very hard to test. I can add an
> ETHERCAP_PPPOE_MTU for sip(4), but not for any other. Well, we still have a
> lot of bits left for Ethernet capabilities, but it still feels like a
> waste.

Like for vlans, it will be added to ethernet drivers once someone can do the
test. It may be slower than for vlans, though.

> 
> [...]
> > > But anyway, we don't have to check if the NIC can receive the frame at
> > > that time, because obviously it could.
> > 
> > ec_capenable ETHERCAP_VLAN_MTU is there to have the driver allow the nic to
> > receive such frames. 
> 
> I'll do that then.

Thanks.

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
     NetBSD: 24 ans d'experience feront toujours la difference
--