exspes@gdr.bath.ac.uk (P E Smee) (11/08/90)
Not really sure this is the best group for this, but it's the best-looking I can find. Any other groups I should try? We have a question regarding DECNET use of MAC addresses, which has become important as we are rewiring ourself into a bunch of local ethers interconnected by clever managed bridges. We can't determine the answer experimentally, as we are not running DECNET at the moment, but we do anticipate local use of it in the near future. We are told that DEC, rather than assigning a single MAC address to a given machine, assigns a group of MAC address to each box, so that the DECNET protocol can use some of the 'address' bits to send small control messages around. (I find this distasteful, but apparently it's 'legal' as long as they stick to their assigned address range.) Can anyone confirm that this is true? Even more useful, if it is true can anyone tell me how many, and which, bits of the MAC address in an IEEE802.3 (Ethernet) packet are used by DEC in this way? (Hex mask on the 12-hex-digit MAC address format would be best to avoid ambiguity.) Do they do this to both the source and destination addresses? Or only to one (and if so, which)? Vaguely related issue, one of the switches available on our bridges is one to tell it whether to forward or discard Ether packets whose source and destination MAC addresses are the same (excluding broadcast packets). Is anyone aware of any protocol under which it would be necessary or sensible to forward such packets? We aren't, and offhand it feels like if the source and destination are the same, then they are by definition the same box on the same side of the bridge and so shouldn't be passed around. Any ideas? -- Paul Smee, Computing Service, University of Bristol, Bristol BS8 1UD, UK P.Smee@bristol.ac.uk - ..!uunet!ukc!bsmail!p.smee - Tel +44 272 303132
doelz@urz.unibas.ch (11/09/90)
In article <1990Nov8.125106.18326@gdr.bath.ac.uk>, exspes@gdr.bath.ac.uk (P E Smee) writes: > > We have a question regarding DECNET use of MAC addresses, which has > become important as we are rewiring ourself into a bunch of local > ethers interconnected by clever managed bridges. We can't determine > the answer experimentally, as we are not running DECNET at the moment, > but we do anticipate local use of it in the near future. > My mac is running TSSnet Ver 2.0 and it reassigns the node address to the ethernet card - becoming the typical aa:... :node:area type. I am not aware of and I don't think it is permitted in the Ethernet spec to have more than one machine with the same enet address. Its something else with gateways. At the days where it was still kinetics Fast Path 4, we had one on testing which was permitting decnet tunneling on local talk. This box had one address, and the macs on the *local talk* cable could share it. This is something completely else than the ethernet, and local talk protocols take the part of addressing - its encapsulated. The LANworks might work differently. We ordered it long ago and are still waiting for delivery... Regards Reinhard
ted@blia.BLI.COM (Ted Marshall) (11/10/90)
In article <1990Nov8.125106.18326@gdr.bath.ac.uk>, exspes@gdr.bath.ac.uk (P E Smee) writes: > We are told that DEC, rather than assigning a single MAC address to a > given machine, assigns a group of MAC address to each box, so that the > DECNET protocol can use some of the 'address' bits to send small > control messages around. (I find this distasteful, but apparently it's > 'legal' as long as they stick to their assigned address range.) Whomever told you that is dead wrong! A machine running DECnet has its Ethernet address set based on the DECnet address (see below). Each machine has a single address only. Multicast addresses in the range AB-00... are also used but not as station addresses. The station address is set as follows: The first 4 bytes are AA-00-04-00. The last two bytes are computed as (area * 1024 + node), byte swapped. Thus, DECnet address 8.56 would compute as (8 * 1024 + 56) = 8248 = 0x2038. Thus, the corresponding Ethernet address would be AA-00-04-00-38-20. Hope this helps. -- Ted Marshall ...!ucbvax!mtxinu!blia!ted <or> ted@blia.bli.com ShareBase Corp., 14600 Winchester Blvd, Los Gatos, Ca 95030 (408)378-7000 The opinions expressed above are those of the poster and not his employer.
exspes@gdr.bath.ac.uk (P E Smee) (11/12/90)
In article <12925@blia.BLI.COM> ted@blia.BLI.COM (Ted Marshall) writes: >In article <1990Nov8.125106.18326@gdr.bath.ac.uk>, exspes@gdr.bath.ac.uk (P E Smee) writes: >> We are told that DEC, rather than assigning a single MAC address to a >> given machine, assigns a group of MAC address to each box, so that the >> DECNET protocol can use some of the 'address' bits to send small >> control messages around. (I find this distasteful, but apparently it's >> 'legal' as long as they stick to their assigned address range.) > >Whomever told you that is dead wrong! A machine running DECnet has its >Ethernet address set based on the DECnet address (see below). Each machine >has a single address only. Multicast addresses in the range AB-00... are >also used but not as station addresses. > >The station address is set as follows: The first 4 bytes are AA-00-04-00. >The last two bytes are computed as (area * 1024 + node), byte swapped. >Thus, DECnet address 8.56 would compute as (8 * 1024 + 56) = 8248 = 0x2038. >Thus, the corresponding Ethernet address would be AA-00-04-00-38-20. How about the following as a closer guess? We are running a number of DEC machines with hardware MAC addresses of 08-00-2b-... (and a Silicon Graphics box 08-00-69-...). We are told that when they start wanting to use DECnet across our bridges, we will have to tell the bridges that their addresses are AA-00-04-<something else>. It appears that this latter address holds when they are running DECNET protocol; and we believe that they will still continue to claim to be 08-00-2b-whatever when they are running other protocols. Is this any better? (Still means that from the point of view of the bridge, the box has two addresses.) We also hear rumours that AA-00-04-... machines may under some circumstances use AA-00-03-<same thing>. (AA-00-03-... is registered as 'Global physical address for some DEC machines' -- or, in other words, is assigned to DEC.) -- Paul Smee, Computing Service, University of Bristol, Bristol BS8 1UD, UK P.Smee@bristol.ac.uk - ..!uunet!ukc!bsmail!p.smee - Tel +44 272 303132
iglesias@draco.acs.uci.edu (Mike Iglesias) (11/13/90)
In article <1990Nov12.155501.23342@gdr.bath.ac.uk> P.Smee@bristol.ac.uk (Paul Smee) writes: >In article <12925@blia.BLI.COM> ted@blia.BLI.COM (Ted Marshall) writes: >>The station address is set as follows: The first 4 bytes are AA-00-04-00. >>The last two bytes are computed as (area * 1024 + node), byte swapped. >>Thus, DECnet address 8.56 would compute as (8 * 1024 + 56) = 8248 = 0x2038. >>Thus, the corresponding Ethernet address would be AA-00-04-00-38-20. > >How about the following as a closer guess? We are running a number of >DEC machines with hardware MAC addresses of 08-00-2b-... (and a Silicon >Graphics box 08-00-69-...). We are told that when they start wanting >to use DECnet across our bridges, we will have to tell the bridges that >their addresses are AA-00-04-<something else>. It appears that this >latter address holds when they are running DECNET protocol; and we >believe that they will still continue to claim to be 08-00-2b-whatever >when they are running other protocols. Is this any better? (Still >means that from the point of view of the bridge, the box has two >addresses.) No, they will use the AA-00-04 address for TCP/IP, etc. If the bridge learns addresses by looking at the net, you shouldn't have any problems. If you have to tell your bridge what system is on each side of the bridge, they you need to tell it the AA-00-04 address, not the 08-00-2B address (assuming you're running DECNET). You also have to start DECNET before any other protocol so the other protocols get the right ethernet address. This is true for VMS and Ultrix. Otherwise, you may confuse other systems on the network! >We also hear rumours that AA-00-04-... machines may under some >circumstances use AA-00-03-<same thing>. (AA-00-03-... is registered >as 'Global physical address for some DEC machines' -- or, in other >words, is assigned to DEC.) I've never seen this - maybe a DEC expert knows? Mike Iglesias University of California, Irvine Internet: iglesias@draco.acs.uci.edu BITNET: iglesias@uci uucp: ...!ucbvax!ucivax!iglesias
terry@spcvxb.spc.edu (Terry Kennedy, Operations Mgr) (11/13/90)
In article <273F4A7D.27274@orion.oac.uci.edu>, iglesias@draco.acs.uci.edu (Mike Iglesias) writes: > In article <1990Nov12.155501.23342@gdr.bath.ac.uk> P.Smee@bristol.ac.uk (Paul Smee) writes: >>We also hear rumours that AA-00-04-... machines may under some >>circumstances use AA-00-03-<same thing>. > > I've never seen this - maybe a DEC expert knows? Who, me? Well, if you insist... AA-00-03-... is the hardware address on the older DEC interfaces, such as the DEUNA. It is changed to AA-00-04-... when DECnet is started, just as on the other boards. I don't know why DEC changed to 08-00-2B-... when they came out with the newer interfaces. Perhaps people were being confused by the -03- vs. -04-? Terry Kennedy Operations Manager, Academic Computing terry@spcvxa.bitnet St. Peter's College, US terry@spcvxa.spc.edu (201) 915-9381
doelz@urz.unibas.ch (11/13/90)
In article <1990Nov12.155501.23342@gdr.bath.ac.uk>, exspes@gdr.bath.ac.uk (P E Smee) writes: > > How about the following as a closer guess? We are running a number of > DEC machines with hardware MAC addresses of 08-00-2b-... (and a Silicon > Graphics box 08-00-69-...). We are told that when they start wanting > to use DECnet across our bridges, we will have to tell the bridges that > their addresses are AA-00-04-<something else>. It appears that this > latter address holds when they are running DECNET protocol; and we > believe that they will still continue to claim to be 08-00-2b-whatever > when they are running other protocols. Is this any better? (Still > means that from the point of view of the bridge, the box has two > addresses.) Still wrong. As soon as any machine boots decnet, the (hardwired) board address will be reconfigured to read aa:00:... :node:area - and there is no 08:00: .. any longer. If you want to gain the old addresses back, the decnet must be shut down, and the hardware rebooted. However, the appletalk will also run on aa:00... addresses, because its a different layer! The layer we are currently talking about is so low that there is no need to know which protocol is running... its decoded much higher in the hierarchy of OSI layers. There is a protocol code for that in each packet, and the hardware takes care of that... Reinhard (The above saying applies to: VAX, SGI, CONVEX, MAC, and I have reasons to believe that other machines behave similar)
moroney@ramblr.enet.dec.com (11/14/90)
In article <1990Nov12.155501.23342@gdr.bath.ac.uk> P.Smee@bristol.ac.uk (Paul Smee) writes: >We also hear rumours that AA-00-04-... machines may under some >use AA-00-03-<same thing>. (AA-00-03-... is registered >as 'Global physical address for some DEC machines' -- or, in other >words, is assigned to DEC.) Very old DEUNAs and DEQNAs may have address ROMs that have AA-00-03 (and maybe some other AA-00-xx addresses). Since IEEE has redefined the second bit of the first byte as "locally administrated address space" DEC no longer uses these address for new address ROMs, since these addresses have this bit set. -Mike Moroney
exspes@gdr.bath.ac.uk (P E Smee) (11/14/90)
In article <1990Nov13.025835.759@spcvxb.spc.edu> terry@spcvxb.spc.edu (Terry Kennedy, Operations Mgr) writes: > AA-00-03-... is the hardware address on the older DEC interfaces, such as >the DEUNA. It is changed to AA-00-04-... when DECnet is started, just as on >the other boards. I don't know why DEC changed to 08-00-2B-... when they came >out with the newer interfaces. Perhaps people were being confused by the -03- >vs. -04-? Thanks everyone who responded, that's been a great help. Reason for picking on this one to 'F' to is the question of the 'change to 08-00-2B-...' Our copy of the IEEE assignments list mentions that one of the bits in AA- is the 'locally administered' bit, which means that in general any site can nick an AA-... address for locally produced stuff which has not been given a company registration. (Of course, if you are clever and know about existing uses you will not use AA-00-03-... (or -04-...), but if you were working from the protocol description alone, without a list of assigned manufacturers' address ranges, you could accidentally end up assigning a home-brew device to an address which would match one of the addresses being used by DEC. Murphy's law says that you would then order a box from DEC and end up on the same number.) Or, in other words and basically, 08-00-2B-... is ASSIGNED to DEC. The two AA- addresses are listed as DEC because DEC nabbed them before things were properly organised. In theory, 'locally administered' addresses shouldn't be used for any equipment which might be connected to anything which is NOT on your site, which means that according to modern theology, they shouldn't be used by vendors. -- Paul Smee, Computing Service, University of Bristol, Bristol BS8 1UD, UK P.Smee@bristol.ac.uk - ..!uunet!ukc!bsmail!p.smee - Tel +44 272 303132