On Sat, Apr 26, 2008 at 4:19 PM, Garrett D'Amore <[EMAIL PROTECTED]> wrote: > > Peter Tribble wrote: > > > I logged a bug in b.o.o - 6691647 is the ID - which has been marked > > incomplete. > > > > (Heaven knows how someone external is meant to supply > > additional information though.) > > > > But anyway, the symptom I'm seeing is that on a Sun Blade 1500 > > that gets its adress using dhcp from a home router, the IP address > > often comes up as 0.0.0.0 after a suspend/resume cycle. > > > > Needless to say, this is pretty useless. It's true with both > > network/physical:default and network/physical:nwam. A couple > > of manual workarounds are to pull the network cable, or to manually > > restart nwam. I've got a cron job running every minute to kick nwam > > if it detects the address being four zeros, which at least means the > > machine can recover itself. > > > > (To be clear, this isn't an nwam problem - it looks like the underlying > > dhcp.) > > > > I put the following comments in the bug: > > I wonder, has the lease expired while the host was suspended? Maybe > an attempt to renew a lease is not possible because the address has > been given out elsewhere while the system was suspended?
Well, it might have expired. But in that case it should try to do something, right? As for something else grabbing the address, I can rule that out. > It strikes me that dhcp is probably one of those few bits of > software that needs to "know" about a S/R cycle. In this case, I > think setting up a signal handler for SIGTHAW could enable it to > know when it has been resumed, and logic added to reacquire an > address (perhaps not just asking for a renewal, but for a whole new > address if necessary). > > The other possible thing to look at, I suppose, is the MAC address > -- ensure that the MAC address on resume is the same as it was on > suspend. If that is not correct, it is indicative of a driver bug. I'm fairly sure that the MAC address is constant and is that of the hardware. > Its probably just a matter of teaching dhcp about resume. Some extra notes. This problem did not occur on this system until recently. I can't be precise, but certainly builds in the 70s worked fine. (And I haven't seen this problem on Sun Blade 150s from S10 FCS up to about nv55 either.) Certainly on older releases, it would just continue using the old address regardless. (Leading to the case where we did have more than one machine with the same IP address.) One thing that does occur to me is that it may be a timing issue - dhcp wakes up a bit too early. But what do I know. -- -Peter Tribble http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/ _______________________________________________ networking-discuss mailing list [email protected]
