Christopher Fowler writes: > On Mon, 2007-07-23 at 07:59 -0400, James Carlson wrote: > > On the Annex communications server, we wrote special demand-dial > > management code to look up a dormant interface by address and bring it > > to life when the peer called us. That way, the return traffic (reply > > packets) wouldn't trigger an outbound attempt that would fail. > > > > On Linux, you may be able to get away with it if both links have > > different sets of IP addresses. > > You can do it on Linux. I have hundreds doing this. Yes you do have 2 > interfaces. I'm curious as to details on how you solved it on the > Annex. The advantages of having a single address space unobstructed by an MMU -- we simply had a global table of demand links and matched inbound links against that table before creating a new interface. If it matched, then we'd attach to the interface in that table, otherwise create a new one. The internal design of the Annex is really quite different from Linux. The PPP code there wasn't based on pppd. > They way we solved it is via if-down and if-up. That is a C > program on our device. When pppd calls that program it looks for an > interface that is in the UP state but not our own. It places that > interface in the DOWN state. When if-down is called it looks for an > interface that is DOWN but not our own and it places that in UP. The > effect is that you'll end up with 2 ppp interfaces with the same > address. Only one is UP. Interesting. > I've had this working for a couple years now and I've not had any > problems. I think the only problem I could see is if pppd died and > never called if-down. You would end up with a PPP interface in the DOWN > state and the remote could never call home. Calling from the server > would rectify the situation. I think that process may have race conditions, given that if-up is triggered after the interface is marked 'UP', and nothing in the system actually waits for that script to complete. For that reason, I wouldn't suggest that as a design for a new feature in the pppd code base. Otherwise, though, it seems like a somewhat reasonable solution for a pppd user. -- James Carlson 42.703N 71.076W <carlsonj@xxxxxxxxxxxxxxx> - To unsubscribe from this list: send the line "unsubscribe linux-ppp" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
