With classification being pushed down into the DLS layer for
softrings, we're facing a point where we're doing double the
amount of work in sanity checking IP packets.

Whilst the PEF project would solve this for us, PEF has been
put in suspended animation for the time being :-(

Does anyone have any thoughts on how we might architect
DLS/IP to try and eliminate this double-work effort in the
mean time?

This is similar to another idea that Garrett brought up
recently where we make the entry requirements for
ip_input() quite different for those of ip_rput() and use
mac/dls layer to cover the those bases that ip_rput()
would do for legacy device drivers.

Thoughts?

(This was motivated by the extract of DLS from Brendan's
email showing the OK_PTR and msgpullup.)

Darren

_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to