roode@anansi.acs.uci.edu (Dana Roode) (07/19/90)
We had an IBM RS/6000 on loan for awhile - 9013S1 came while it was here and I installed it. The TCP/IP configuration menus in SMIT worked fine, and I got all the right tcp/ip parameters set for our network. One of our faculty has now taken delivery of the first purchased RS6000 on campus - it occurs to me now I dont know what O/S version he has, but it's whatever IBM is shipping. I went over to get the network setup using SMIT. I used the "minimal (basic?) setup" option under the communications menu (my memory is a little fuzzy) and ran through the menu that sets hostname, netmask, static routes, etc.. Everything was obvious except for the name of the interface. IBM calls "en0" "standard ethernet" and "et0" "IEEE 802.3 ethernet" - Im not sure what I want as I thought 802.3 was "Standard". To add to my confusion, whenever I make a change to either interface it says "inet0" changed, and the /dev/xxx file has still a different name. What is the correct interface to use? I ended up deleting en0 after it would not work (using SMIT), and started over using et0. Still didnt work. I noticed that all my routes (as per netstat -n -r) where going through the loopback interface lo0 and that "ifconfig lo0" reported the loopback address was set not to 127.0.0.1 but to the host ip address. No wonder things were not working, all routing was through the loopback, or so it seems. I manually reset lo0 and got the network working to a single system on the connected subnet, but it refused to work to any other system on that subnet. How can I start over and do the configuration right? Can someone explain how this all works? (I have managed Unix system for years and on most others getting the initial ifconfig correct in rc.whatever as well as a route command or two is all that is ever needed to make the system work). Dana Roode UC Irvine DRoode@UCI.EDU, Bitnet: DRoode@UCI
rbn@umd5.umd.edu (Ron Natalie) (07/20/90)
Standard Ethernet is the DIX Ethernet I/II encapsulation. IEEE is the 802.3 encapsulation. The difference is whether the field following the source address is the protocol type (standard) or the length (ieee). Everything that you are likely to have, unless you are using HP equipment is probably Ethernet I. Actually on receive, the software ought to be able to tell. -Ron
RAH@IBM.COM ("Russell A. Heise") (07/24/90)
roode@anansi.acs.uci.edu (Dana Roode) writes: > ... > > One of our faculty has now taken delivery of the first purchased RS6000 on > campus - it occurs to me now I dont know what O/S version he has, > but it's whatever IBM is shipping. I went over to get the network > setup using SMIT. I used the "minimal (basic?) setup" option under the > communications menu (my memory is a little fuzzy) and ran through > the menu that sets hostname, netmask, static routes, etc.. Everything was > obvious except for the name of the interface. IBM calls "en0" > "standard ethernet" and "et0" "IEEE 802.3 ethernet" - Im not sure > what I want as I thought 802.3 was "Standard". To add to my confusion, > whenever I make a change to either interface it says "inet0" changed, > and the /dev/xxx file has still a different name. > > What is the correct interface to use? To understand the IBM naming scheme, you have to know a little history of the Ethernet (trademark Xerox Corp.) development. When Xerox and its associates originally published the Ethernet standard, it was known simply as Ethernet. Later, Xerox made some modification/improvements in the protocol and published a new standard, called Ethernet II. This protocol has become the de facto industry-standard Ethernet protocol and is supported under AIX 3 as "standard ethernet" or en0. Later still, the IEEE decided to develop some manufacturer-independent standards for networking. One of these standards, IEEE 802.3, defines a CSMA/CD baseband bus network very similar to Ethernet II. AIX 3 supports this standard as "IEEE 802.3 ethernet" or "et0". In fact, machines using IEEE 802.3 can communicate with each other on the same bus as machines using the Ethernet II. However, the two sets of machines cannot directly intercommunicate. To do that, you need one machine that can talk both protocols to act as a gateway. > > I ended up deleting en0 after it would not work (using SMIT), and > started over using et0. Still didnt work. I noticed that all my > routes (as per netstat -n -r) where going through the loopback > interface lo0 and that "ifconfig lo0" reported the loopback address > was set not to 127.0.0.1 but to the host ip address. No wonder things > were not working, all routing was through the loopback, or so it seems. > > I manually reset lo0 and got the network working to a single system > on the connected subnet, but it refused to work to any other system > on that subnet. > > How can I start over and do the configuration right? Can someone > explain how this all works? (I have managed Unix system for years > and on most others getting the initial ifconfig correct in rc.whatever > as well as a route command or two is all that is ever needed to make > the system work). I recommend the following steps: - Reboot the RISC System/6000 - Use SMIT and remove all those unusual routes - Use SMIT to remove any and all traces of both en0 and et0 - Talk to the administrators of the other machines on your network and determine which network protocol those machines support - Reboot the RS/6000 to start anew - Use SMIT to add the interface used by the other machines - Use SMIT to add whetever routes you need Always use SMIT to make these changes. SMIT automatically updates the ODM (Object Database Manager) data base when you make changes. AIX uses the ODM data base to configure the system each time it starts. If you manually make changes without updating the data base, the changes will be lost when you re-boot. Russ Heise, AIX Technical Support, IBM
roode@anansi.acs.uci.edu (Dana Roode) (07/27/90)
Thanks to those that replied to my query about problems getting tcp/ip configured right. It turns out that en0 is the interface of choice, and the O/S version is 9013s6 not 9021. I followed Russell A. Heise's advice and undid everything, rebooted, and redid it all. It worked better - could connect to system on the same subnet, but not off it. My default route was still going through the loopback interface, lo0. "ifconfig lo0" still reported an ip address of the Host ip#, not 127.0.0.1 as it should be. I saw no way in smit to delete or change lo0, so I punted. I commented out the cfgmgr/ODM commands in /etc/rc.net, and used the traditional UNIX commands. They worked great. If anyone has a clue how lo0 got the Host ip number, or how to undo it, drop me a line. Otherwise I'll consider it a fluke and will see to it we upgrade to 9021 soon. Dana Roode UC Irvine