rainwatr@ucqais.uc.edu (Donald J. Rainwater) (05/04/89)
I have an IBM PC RT running AIX 2.2.1 that I would like to connect to our Ethernet network. I've run an Ethernet transceiver drop cable (2 10 meter sections) to a port on a DELNI, and the RT TCP/IP configuration files have been properly set up (as far as I can tell). There are VAXen on the Ethernet that run SRI Multinet (TCP/IP), and one of them is set up as my domain name server. When the RT is started, it tries to configure the ethernet device (net0). After a period of time, I get a console message saying that net0 has timed out, and that the device has been removed from the configuration. Has anyone done what I'm attempting? Any information would be greatly appreciated. Thanks in advance. -- Don Rainwater, University of Cincinnati Computer Center rainwatr@ucbeh.san.uc.edu rainwatr@ucbeh.bitnet rainwatr@ucqais.uc.edu
vlruo02@dutrun.UUCP (Ge van Geldorp) (05/04/89)
In article <1667@ucqais.uc.edu> rainwatr@ucqais.uc.edu (Donald J. Rainwater) writes: >When the RT is started, it tries to configure the ethernet device (net0). >After a period of time, I get a console message saying that net0 has timed >out, and that the device has been removed from the configuration. Could it be that the jumper settings on your Baseband adapter don't match the values given to `devices'? For some reason, the memory address and interrupt level jumper settings on the adapter as shipped by IBM don't match the default values used by `devices'. Being a software type, I changed the `devices' values: rsa from 80000 to 98000 rea from 81fff to 99fff brsa from 82000 to 9a000 brea from 87fff to 9ffff il1 from 9 to 3 Hope this solves your problem. Ge van Geldorp. (...!uunet!hp4nl!dutrun!vlruo02 or ...!uunet!hp4nl!dutlru2!ge)
shair@uxe.cso.uiuc.edu (05/05/89)
The usual reason that I have seen why net0 times out on an RT is that the interrupt level set with the jumpers on the "Baseband Adapter for Ethernet" card is not the same as the level defined in the "devices" command. I'm going from memory, but I think that the card is set to level 9 as shipped from our factory (may be marked as level 2 just to confuse things) and the default in "devices" is level 3. Bob Shair Scientific and Technical Computing Support IBM Champaign
francis@hslrswi.UUCP (Francis Demierre) (05/05/89)
In article <1667@ucqais.uc.edu> rainwatr@ucqais.uc.edu (Donald J. Rainwater) writes: >I have an IBM PC RT running AIX 2.2.1 that I would like to connect to our >Ethernet network. I've run an Ethernet transceiver drop cable (2 10 meter >sections) to a port on a DELNI, and the RT TCP/IP configuration files >have been properly set up (as far as I can tell). There are VAXen on the >Ethernet that run SRI Multinet (TCP/IP), and one of them is set up as my >domain name server. > >When the RT is started, it tries to configure the ethernet device (net0). >After a period of time, I get a console message saying that net0 has timed >out, and that the device has been removed from the configuration. > >Has anyone done what I'm attempting? Any information would be greatly >appreciated. Thanks in advance. >-- >Don Rainwater, University of Cincinnati Computer Center >rainwatr@ucbeh.san.uc.edu >rainwatr@ucbeh.bitnet >rainwatr@ucqais.uc.edu - check the switches on you Ethernet card to determine the interrupt level and the memory address to which it is configured. The values shown below in the 'devices' output are working for me and 'il1 9' is a safe interrupt level (it is not to be used exclusively by any devices as far as I know). You could try these values. I cannot remember now where the switch positions are described, it is probably in the 'Hardware reference Manual' or in one of the Red books - Check that the interrupt level selected dos not conflict with another of your adapters that want to uses the same level exclusively. Required interrupt level are described somewhere under heading "Conflicting Adapter interrupt level". I am sorry to be so vague, but I'm writing this from home and do not have any RT books at hand :-). - I got some problems when I tried to set the 'inetlen' and 'r_inetlen' to a small value (timed out). I recommend to used the values shown below in the '/etc/net' file. - In the version 2.2.1 of the 'devices' command, it was not possible to set the 'il1' value to '9'. You have to manually edit the '/etc/ddi/enet' file to set the interrupt value to 9 and to allow interrupt level 9 (see below). Then run the devices to set your 'net0' parameters. ------------------------------------------------------------------------------ This is the description of the adapter 'net0' as shown with the 'devices' command: Current Possible Name Description Choice Choices rsa Bus ROM Start Address 98000 80000 - f8000 rea Bus ROM End Address 99fff 81fff - f9fff brsa Bus RAM Start Address 9a000 82000 - fa000 brea Bus RAM End Address 9ffff 87fff - fffff [ ... some text deleted ... ] noi Num. of Interupt Levels Used 1 1 il1 Intrpt Level # of 1st Intrpt 9 3,4,5,7,9 si1 1st Interupt Level is Shared false true,false [ ... some text deleted ... ] nidl Net ID Length 2 1 - 255 nidd Net ID Displacement 12 0 - 255 srbt SLIH Ring Buffer Threshold 10 0 - 255 [ ... some text deleted ... ] ------------------------------------------------------------------------------ This is the content of the '/etc/ddi/enet' file (used to check arguments given in the 'devices' command as well as to fix the default values) : default: [ ... some text deleted ... ] il1 = 9 <------------------------------------<manually patched to 9 * Intrpt Level # of 1st Intrpt * 3,4,5,7,9 <--------< " " [ ... some text deleted ... ] denet0: rsa = 98000 * Bus ROM Start Address * 80000 - f8000 rea = 99fff * Bus ROM End Address * 81fff - f9fff brsa = 9a000 * Bus RAM Start Address * 82000 - fa000 brea = 9ffff * Bus RAM End Address * 87fff - fffff pn = 0 * I/O Port Number on Adapter * 0,1,2,3 [ ... some text deleted ... ] il1: syschg = none vtype = 1 map = mil1 display = true required = true [ ... some text deleted ... ] mil1: m3 = 0 m4 = 0 m5 = 0 m7 = 0 m9 = 0 <--------------------------<Manually added to allow level 9 [ ... some text deleted ... ] [ EOF ] ------------------------------------------------------------------------------ The '/dev/net0' file: crw-rw-rwT 1 root system 22, 0 05 May 14.21 net0 ------------------------------------------------------------------------------ The '/etc/net' file for 'net0' entry: * /etc/net net0: netaddr = 1.6.0.5 <--- Your address here inetlen = 1518 r_inetlen = 1518 rip = input <--- It is a new keyword in AIX 2.2 ------------------------------------------------------------------------------ I hope that all this can help you, anyway, you can email me if you have any other questions, I will try to help. ------------------------------------------------------------------------------- ASCOM Hasler AG (Tlx: 911392 hsw ch Phn: +41-31 632111 Fax: +41-31 632182) Francis Demierre (UUCP: ...!mcvax!cernvax!hslrswi!francis Dpt 81B2 Phn: +41-31 633662 and all above numbers Belpstrasse 23 Unix mail: francis@hslrswi CH 3000 Bern 14 -------------------------------------------------------------------------------
njs@scifi.UUCP (Nicholas J. Simicich) (05/06/89)
Running diagnostics and looking at your system configuration will tell you what interrupt level and memory addresses your ethernet card is set to. Then you can match that in devices. So you don't have to take the box apart and pull the card. It will also tell you if you have any other adapters which have overlaps or conflicts in terms of interrupt levels, and I/O space or AT bus memory overlaps, at least for your IBM supplied adapters. Diagnostics will also (look under communications utilities) bounce a message off of the ethernet, so that you can verify the connections you have back to the DELNI. Yes, I know that if diagnostics can do it, devices should..... :-) -- Nick Simicich --- uunet!bywater!scifi!njs --- njs@ibm.com (Internet)
griefer@ibmarc.uucp (Allan D. Griefer) (05/07/89)
In a recent append, rainwatr@ucqais.uc.edu (Donald J. Rainwater) writes... >... >When the RT is started, it tries to configure the ethernet device (net0). >After a period of time, I get a console message saying that net0 has timed >out, and that the device has been removed from the configuration. ... Offhand, it sounds as if the plugging on the Ethernet adapter is set differently from the specification in the devices command. Check out the values set by devices by issuing: showdev adapt net0 from within devices and printing the results. Compare the rsa and il1 fields to the interrupt level and starting address on the card. Note that rea, brsa, and brea must be consecutive to rsa. Hope this helps. Opinions are strictly my own, Allan D. Griefer, IBM Almaden Research Center, San Jose, CA BITNET: GRIEFER at ALMADEN Internet: griefer@ibm.com UUCP: ...!uunet!ibmarc!griefer
jim@applix.UUCP (Jim Morton) (05/08/89)
In article <934@hslrswi.UUCP>, francis@hslrswi.UUCP (Francis Demierre) writes: > /etc/net file: > net0: > netaddr = 1.6.0.5 <--- Your address here > inetlen = 1518 > r_inetlen = 1518 > rip = input <--- It is a new keyword in AIX 2.2 Watch the "inetlen" parameters -- I have not been able to use 1518 on my Sun/VAX/RT network with NFS. I have been using 1500 successfully -- Guy Harris talked about this issue about a month ago -- at 1518, the packets going out on the network are too large and the Suns print "ie0: huge packet" -- Jim Morton, APPLiX Inc., Westboro, MA UUCP: ...harvard!m2c!applix!jim jim@applix.m2c.org
guy@auspex.auspex.com (Guy Harris) (05/09/89)
>Watch the "inetlen" parameters -- I have not been able to use 1518 on my >Sun/VAX/RT network with NFS. I have been using 1500 successfully -- Guy Harris >talked about this issue about a month ago -- at 1518, the packets going >out on the network are too large and the Suns print "ie0: huge packet" I was pretty much just guessing that "1518" means they were putting out Ethernet packets with 1518 bytes of data. Is this actually the case? If so, it's a Bad Thing because the Ethernet standard says Ethernet packets should have at most 1500 bytes of data; in that case, IBM should stop doing it - end of problem. (Did the VAXes complain about the packets as well? If not, it could just be because the Intel 82586 that Sun uses, or the driver than Sun provides for it, is more willing to complain about what it thinks of as Bad Things than the VAXes hardware or driver.)