anh@mit-amt.MEDIA.MIT.EDU (Viet Anh Ho) (11/23/89)
Two of our HP9000/320s think that they broadcast the addresses 18.85.255.255, when instead, according to the outside world they are actually 18.0.0.0! # /etc/ifconfig lan0 lan0: flags=64b<UP,BROADCAST,ROUTE,RUNNING,IEEE,ETHER> inet 18.85.0.30 netmask ffff0000 broadcast 18.85.255.255 We couldn't find any reference on how to re-set their broadcasting addresses. The following command won't do the job: # /etc/ifconfig lan0 $NODENAME netmask 255.255.0.0 up We appreciate any suggestion and reference to the problem. Thanks. -anh (anh@media-lab.media.mit.edu) -- Viet Anh @ The MIT Media Lab E15-353, 253-0302
dfc@hpindda.HP.COM (Don Coolidge) (11/30/89)
If you're asking if HP-UX provides a way to arbitrarily configure your interface IP broadcast address (a la 4.3BSD), the answer is, "not yet". That capability will be available as a patch to the 7.0 HP-UX release, and will be an integral part of all subsequent releases. The interface broadcast address is currently defaulted to all ones (4.3-ish, and conforming to the Internet Hosts Requirements Document), subject to the interface netmask and IP address. In your case (netmask 255.255.0.0; IP address 18.85.0.30), that would result in 18.85.255.255 (as in your example). If other nodes on the LAN are seeing something else, you have a problem that I don't understand and that should never be able to happen. If that's the case, please contact me via email at : dfc@hpindaw.hp.com There's only one way I can conceive of that other nodes on your LAN could be seeing 18.0.0.0 - if you have an application using sendto() with 18.0.0.0 specified as the target IP address. By the way, even though the interface broadcast address is defaulted to all ones, HP-UX will also recognize inbound and outbound datagrams with all zeroes as valid broadcast packets. - Don Coolidge
dfc@hpindda.HP.COM (Don Coolidge) (12/02/89)
>There's only one way I can conceive of that other nodes on your LAN could >be seeing 18.0.0.0 - if you have an application using sendto() with 18.0.0.0 >specified as the target IP address. >By the way, even though the interface broadcast address is defaulted to all >ones, HP-UX will also recognize inbound and outbound datagrams with all zeroes >as valid broadcast packets. I've received some private email describing the problem in greater detail, and wanted to post the final answer. The problem is rwhod, which is using sendto() as I described above. HP's released networking code is based on 4.2BSD, which used zeroes as a broadcast address. We've made many 4.3-ish upgrades, but rwhod has not yet been one of them. Unforunately, the address used by rwhod's sendto() is hard-coded in 4.2-ish zeroes. So, for the systems mentioned in the original posting (6.0, 6.2), rwhod will continue to pump zero-based packets onto the net unless you invoke it with the -r flag (receive data only; don't send any). Administrators of large LANs might want to keep rwhod generally inactive, anyway (on all machines, not only on HP systems) - large numbers of broadcasts (one/node/3 minutes) unnecessarily eat up a lot of LAN and machine bandwidth on LANs with lots of nodes. - Don Coolidge
burzio@mmlai.UUCP (Tony Burzio) (12/05/89)
In article <4310077@hpindda.HP.COM>, dfc@hpindda.HP.COM (Don Coolidge) writes: > Administrators of large LANs might want to keep rwhod generally inactive, > anyway (on all machines, not only on HP systems) - large numbers of broadcasts > (one/node/3 minutes) unnecessarily eat up a lot of LAN and machine bandwidth > on LANs with lots of nodes. You haven't seen network overload until you coexist with a VMS/VAX cluster! The poor things constantly chatter amongst themselves. For a short time the VAXen herd were down today, and BOY did my HP-UX cluster speed up! :-) ********************************************************************* Tony Burzio * Oh no! Not again! Martin Marietta Labs * - the flower mmlai!burzio@uunet.uu.net * *********************************************************************