On Sun, 2007-08-26 at 05:35 +0800, Cathy Zhou wrote:
> James Carlson wrote:
> > Sebastien Roy writes:
> >
> >> - VLAN_SIZE_CAPABLE
> >>
> >> The device can handle packets of bigger size but do not have VLAN
> >> PPA access support.
> >>
> >
> > In the interest of perhaps extending this support in the future to
> > handle nested VLANs, RBridges, and the like, could you define this to
> > return an integer (giving the true maximum SDU) as well as the flag?
> >
> > An interface that can handle 1518 header+payload but can't handle 1522
> > would fail to handle nested VLANs, and one that can handle 1522 but
> > not 1528 won't be able to do RBridges with inner VLAN headers.
> >
> Sure. I will think about this. We need a way for GLDv2 driver to tell
> its trun maximum SDU too.
This is interesting. A lot of NIC drivers I've worked with could
support "baby jumbo" frames up to ~2K. But they have software checks
for packets larger than some predetermined value... for GLDv3 I just
changed those predetermined values to add 4 bytes for the VLAN tag.
If we're going to start supporting a wide variety of different MTUs, the
landscape is going to get a lot more complex.
-- Garrett
>
> Thanks
> - Cathy
>
