On Wed, Jan 25, 2006 at 04:14:52PM -0800, Sam Leffler wrote: > Robert Watson wrote: > > > >On Tue, 24 Jan 2006, John-Mark Gurney wrote: > > > >>sthaug_at_nethelp.no wrote this message on Tue, Jan 24, 2006 at 23:15 +0100: > >> > >>>>We should probably better document the interface "interface". if we are > >>>>going to (as Sam suggests) > >>>>do some cleanups we might as well consider what other changes should be > >>>>put in at the same time. > >>> > >>> > >>>A couple of other suggestions: > >>> > >>>- For software routers (quagga, zebra etc.) it would be very nice to > >>>have "link up" / "link down" notifications to the routing process(es). > >> > >> > >>already supported, either through kqueue, routing socket, or devd (I > >>believe)... > > > > > >However, I believe some interface drivers don't generate these events. > >Drivers not supporting a specific event delivered by an interrupt > >probably need to learn to poll for state changes so that they can > >generate an event. > > I am not aware of any drivers that do not generate the events; only > drivers that do not properly detect link status changes. We do have some pseudo devices that could simulate a link and don't appear to. For instance tap could pretend to have media and produce notifications based on whether the /dev/tap# device is open. It seems like that could be useful in some scenarios where you're hanging a virtual machine off the other end. -- Brooks -- Any statement of the form "X is the one, true Y" is FALSE. PGP fingerprint 655D 519C 26A7 82E7 2529 9BF0 5D8E 8BE9 F238 1AD4
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:51 UTC