powell@wraith.netops.contel.com (Mike Powell "CFS Net Ops") (03/31/91)
we have and AGS+ with 24 ethernet interfaces being used as a "backbone in a box". I tried to use more than one ethernet interface on one physical ethernet and it failed miserably. I knew before hand this would not work, I've read the arp table more times than I cared to, but at the time I was damned and determined to get something working. My question is: Other than simplicity why was this done? were you going to run out of your assigned block to fast by shipping fully loaded AGS+s? I can understand why this is done if you are routing Decnet. Is there any way around this? Thanks, Mike -- Mike Powell PPASEL "It's hard to find router bits for your cisco." MRP 03/29/91 Disclaimer: I speak for myself. No direct relation to the DUAT folks. internet: powell@wraith.netops.contel.com Usenet: uunet!contel-fss!powell
gah@hood.hood.caltech.edu (Glen Herrmannsfeldt) (04/03/91)
In all TCP/IP implementations I know of, all ports on the same machine use the same ethernet (hardware) address. On initialization, it reads the ROM on the first port it finds, and stores this in all the ports. I believe this is TCP/IP standard, though I can't find the reference. This is interesting, though, because I once put two ports of one machine, with different IP addresses on the same net. It seemed to work fine. I think cisco routers know how to put more than one IP address per port, anyway. At least that is what I was told.
William "Chops" Westfield <BILLW@mathom.cisco.com> (04/03/91)
In all TCP/IP implementations I know of, all ports on the
same machine use the same ethernet (hardware) address. On
initialization, it reads the ROM on the first port it finds,
and stores this in all the ports.
No, this is incorrect. In fact, I don't know of ANY TCP implementations
that do this, although it is common in come other protocols (eg XNS).
In particular, unless you are running some other protocols, cisco routers
will leave the ethernet address of each ethernet interface at its original
(and unique) value.
I think cisco routers know how to put more than one IP
address per port, anyway. At least that is what I
was told.
This is true. I can't think of any reason to connect more than one
ethernet interface in a router to the same cable segment.
Bill Westfield
cisco Systems.
-------wrl@wdl1.wdl.loral.com (Bill Lewandowski) (04/03/91)
In article <33874@boulder.Colorado.EDU> BILLW@mathom.cisco.com (WilliamChops Westfield) writes: > > In all TCP/IP implementations I know of, all ports on the > same machine use the same ethernet (hardware) address. On > initialization, it reads the ROM on the first port it finds, > and stores this in all the ports. > >No, this is incorrect. In fact, I don't know of ANY TCP implementations >that do this, although it is common in come other protocols (eg XNS). >In particular, unless you are running some other protocols, cisco routers >will leave the ethernet address of each ethernet interface at its original >(and unique) value. > Bill, sorry to say but suns have done this to us. Last year we were changing from network 128.5.0.0 to 137.249.0.0. We wanted to have a machine with two ethernet interfaces on the same ethernet cable segment so we could have continous Name Server service (one on each subnet) when the NIC flipped us and until all the hosts.txt got caught up. We had problems because the sun on bootup made both ethernet cards the same hardware address. We had to put a separate hardware address in the "ifconfig" statement for the second ethernet card in order for us not to get duplicate packets. (I got a lot of help on this from the "sun-nets" list) Bill Lewandowski ========================================================================= -- Bill Lewandowski LORAL Western Development Labs (408) 473-4362 Internet: wrl@wdl1.wdl.loral.com FAX: (408) 473-7926 UUCP: wdl1!wrl
tonyf@bwdls49.bnr.ca (Tony Farrow) (04/04/91)
In article <1991Apr3.155233.17045@wdl1.wdl.loral.com>, wrl@wdl1.wdl.loral.com (Bill Lewandowski) writes: |> In article <33874@boulder.Colorado.EDU> BILLW@mathom.cisco.com (WilliamChops Westfield) writes: |> > |> > In all TCP/IP implementations I know of, all ports on the |> > same machine use the same ethernet (hardware) address. On |> > initialization, it reads the ROM on the first port it finds, |> > and stores this in all the ports. |> > |> Bill, sorry to say but suns have done this to us. Last year we were changing |> from network 128.5.0.0 to 137.249.0.0. We wanted to have a machine with |> two ethernet interfaces on the same ethernet cable segment so we could have |> continous Name Server service (one on each subnet) when the NIC flipped us |> and until all the hosts.txt got caught up. |> |> We had problems because the sun on bootup made both ethernet cards the same hardware |> address. We had to put a separate hardware address in the "ifconfig" statement for |> the second ethernet card in order for us not to get duplicate packets. |> (I got a lot of help on this from the "sun-nets" list) |> We had the same problem here with our multi-homed Sun File Servers. We got around the problem by using unique ethernet addresses given to us by Sun for our dual interface servers via ifconfig options. Our problem did not cause us much grief as our unique ethernet addresses were only needed per IP subnet and each of the interfaces were on different IP subnets. regards, tony -- *--------------------------+----------------------------------------* | Tony Farrow | NetNorth/BITNET: TONYF@BNR.CA | | Internet Systems | | | Bell-Northern Research | UUCP: ...uunet!bnrgate!bwdla54!tonyf | | P.O. Box 3511, Station C | Phone: (613)763-4903 ESN: 393-4903 | | Ottawa, Canada K1Y 4H7 | FAX: (613)763-3283 ESN: 393-3283 | *--------------------------+----------------------------------------*
shj@ultra.com (Steve Jay) (04/04/91)
In <33874@boulder.Colorado.EDU> BILLW@mathom.cisco.com (WilliamChops Westfield) writes: > In all TCP/IP implementations I know of, all ports on the > same machine use the same ethernet (hardware) address. On > initialization, it reads the ROM on the first port it finds, > and stores this in all the ports. >No, this is incorrect. In fact, I don't know of ANY TCP implementations >that do this From my neighborhood Sun server: # ifconfig -a le0: flags=63<UP,BROADCAST,NOTRAILERS,RUNNING> inet 139.93.30.1 netmask ffffff80 broadcast 139.93.30.0 ---> ether 8:0:20:7:aa:77 ie1: flags=63<UP,BROADCAST,NOTRAILERS,RUNNING> inet 139.93.1.8 netmask ffffff80 broadcast 139.93.1.0 ---> ether 8:0:20:7:aa:77 Steve Jay shj@ultra.com ...ames!ultra!shj Ultra Network Technologies / 101 Dagget Drive / San Jose, CA 95134 / USA (408) 922-0100 x130 "Home of the 1 Gigabit/Second network"
powell@wraith.netops.contel.com (Mike Powell "CFS Net Ops") (04/06/91)
In article <33874@boulder.Colorado.EDU> BILLW@mathom.cisco.com (WilliamChops Westfield) writes: > >This is true. I can't think of any reason to connect more than one >ethernet interface in a router to the same cable segment. > >Bill Westfield >cisco Systems. >------- To do appletalk phase I <-> II translation. Mike -- Mike Powell PPASEL "It's hard to find router bits for your cisco." MRP 03/29/91 Disclaimer: I speak for myself. No direct relation to the DUAT folks. internet: powell@wraith.netops.contel.com Usenet: uunet!contel-fss!powell
martillo@crackers.clearpoint.com (Martillo) (04/06/91)
In article <33874@boulder.Colorado.EDU> BILLW@mathom.cisco.com (WilliamChops Westfield) writes: > I think cisco routers know how to put more than one IP > address per port, anyway. At least that is what I > was told. >This is true. I can't think of any reason to connect more than one >ethernet interface in a router to the same cable segment. Hardware fault tolerance might be a legitimate reason. >Bill Westfield Joachim Carlo Santos Martillo Ajami
kdenning@pcserver2.naitc.com (Karl Denninger) (04/22/91)
In article <1991Apr3.222527.3422@ultra.com> shj@ultra.com (Steve Jay) writes: >In <33874@boulder.Colorado.EDU> BILLW@mathom.cisco.com (WilliamChops Westfield) writes: > >> In all TCP/IP implementations I know of, all ports on the >> same machine use the same ethernet (hardware) address. On >> initialization, it reads the ROM on the first port it finds, >> and stores this in all the ports. > >>No, this is incorrect. In fact, I don't know of ANY TCP implementations >>that do this It is true on SOME machines -- primarially those which get their ethernet address from the system PROM rather than from the board's PROM. >From my neighborhood Sun server: > > # ifconfig -a > le0: flags=63<UP,BROADCAST,NOTRAILERS,RUNNING> > inet 139.93.30.1 netmask ffffff80 broadcast 139.93.30.0 > ---> ether 8:0:20:7:aa:77 > ie1: flags=63<UP,BROADCAST,NOTRAILERS,RUNNING> > inet 139.93.1.8 netmask ffffff80 broadcast 139.93.1.0 > ---> ether 8:0:20:7:aa:77 Note that you can change that with "ifconfig ie1 ether xx:xx:xx:xx:xx:xx" Not all values are valid for the hardware address, but enough are to make it useful. (I usually change the "20" to "21"). -- Karl Denninger - AC Nielsen, Bannockburn IL (708) 317-3285 kdenning@nis.naitc.com "The most dangerous command on any computer is the carriage return." Disclaimer: The opinions here are solely mine and may or may not reflect those of the company.
ercm20@castle.ed.ac.uk (Sam Wilson) (04/23/91)
In article <1991Apr21.170045.21393@pcserver2.naitc.com> kdenning@pcserver2.naitc.com (Karl Denninger) writes: >In article <1991Apr3.222527.3422@ultra.com> shj@ultra.com (Steve Jay) writes: >>In <33874@boulder.Colorado.EDU> BILLW@mathom.cisco.com (WilliamChops Westfield) writes: >> >>> [ controversy over whether it is ever done to set the MAC address of >>> all ethernet ports on a machine to be the same ] > >It is true on SOME machines -- primarially those which get their ethernet >address from the system PROM rather than from the board's PROM. > >>From my neighborhood Sun server: >> [ example of Sun ifconfig showing the same MAC address for 2 i/fs ] > >Note that you can change that with "ifconfig ie1 ether xx:xx:xx:xx:xx:xx" > >Not all values are valid for the hardware address, but enough are to make it >useful. (I usually change the "20" to "21"). The only systems that I'm aware of that regularly do this are DEC machines running DECnet (which have to because of the way DECnet phase IV works) and Suns. It seems that Sun have an ideosyncratic reading of the IEEE standard which says that each 802.x station must have a unique MAC address - they take 'station' to mean each computer rather than each interface and set all MAC addresses the same. This has interesting implications for mutihomed stations in bridged networks... :-). If you're planning to change the MAC address I believe you ought to set the 'locally administered' bit - bit 1 (value 2) in the first octet, thus a modified Sun address might start 0a-00-20 rather than 08-00-20, but you can set it to anything you like so long as you don't set the low order bit of the first octet, the multicast bit. DEC do it that way: DECnet MAC addresses start aa-. Sam Wilson Network Services, Edinburgh University Computing Service, Scotland, UK