Creating a test packet
Bookmark this page:
 Yahoo!
 Google
 Windows Live
 del.icio.us
 digg
 Netscape
|
|
|||||||||||||||||||
|
Posted by on October 13, 2005, 11:21 am
Please log in for more thread options I've got an ethernet driver I'm modifying that sends out a test packet when it starts to get around a hardware bug. There is no need to ever get a reply, it's sent only to kickstart things. I'm a bit worried that the data format used could be interpreted by some equipment as a valid packet, though this would be rare. What I want is some sort of test packet that is universally ignored. The data as it is now would start with an LLC header of "00 00 00". Is this likely to be misinterpreted, or is there something else I can try? -- Darin Johnson | |||||||||||||||||||
|
Posted by on October 13, 2005, 12:31 pm
Please log in for more thread options It's very unlikely to be missinterpreted. http://www.wildpackets.com/support/compendium/ethernet/frame_formats http://www.ethermanage.com/ethernet/enet-numbers/ieee-lsap-list.html http://www.ethermanage.com/ethernet/enet-numbers/cisco-lsap-list.html http://www.ethermanage.com/ethernet/descript-troubleshoot.html Mention a NULL LSAP with the address fields set to zero. Maybe your zero, zero is just the job. You should get the IEEE 802.2 doc and check yourself though. You used to be able to buy them on-line and download a pdf. Here are a few possible workarounds. Avoid broadcast and m'cast dest MAC. Could you send a packet with an invalid CRC? Then everything would ignore it. What happens if you send it to your own MAC address? Everything else would ignore it, but it might possibly come back to you. Buy a NIC, get its address then destroy it. Use that address as the destination address. Send a 'good' frame eg an ICMP echo request or an ARP request. Just pick a destination MAC, the chances of ever coming into contact with it are rather small. Unless:- you sell millions of them in which case send me my share:) Look through the OUI list and pick an unlikely vendor, say one that has not yet been taken over and is not too likely to be. http://www.iana.org/assignments/ethernet-numbers How about: 080056 Stanford University Not likely to be taken over any day soon and not too likely to go into the mass production of printers, games consoles or refrigerators. Of course they might sell it. | |||||||||||||||||||
|
Posted by News Me on October 13, 2005, 6:06 pm
Please log in for more thread options
anybody43@hotmail.com wrote: [snip] > Just pick a destination MAC, the chances of
> ever coming into contact with it are rather small. > Unless:- you sell millions of them in which case send > me my share:) > > Look through the OUI list and pick an unlikely vendor, say > one that has not yet been taken over and is not too likely to be. > > http://www.iana.org/assignments/ethernet-numbers > > How about: > > 080056 Stanford University > > Not likely to be taken over any day soon and not > too likely to go into the mass production of > printers, games consoles or refrigerators. Rather than hijacking a globally administered MAC address, use a locally administered one. The second bit of the most-significant byte of the MAC is the local/global flag (1 = local, 0 = global). The first bit of the most-significant byte is the multicast/unicast bit (1 = multicast, 0 = unicast), so the second digit MUST be even if you want to prevent switches from possibly forwarding your packet. So as long as the second digit of the OUI is one of you should be OK. E.g., 02-00-00-00-00-01. NM -- convert UPPERCASE NUMBER to a numeral to reply | |||||||||||||||||||
|
Posted by stephen on October 13, 2005, 9:18 pm
Please log in for more thread options
> > I'm a bit worried that the data format used could be interpreted by
> > some > > equipment as a valid packet, though this would be rare. What I want is > > some sort of test packet that is universally ignored. The data as it > > is > > now would start with an LLC header of "00 00 00". Is this likely to be > > misinterpreted, or is there something else I can try? >
> It's very unlikely to be missinterpreted. > > http://www.wildpackets.com/support/compendium/ethernet/frame_formats > > http://www.ethermanage.com/ethernet/enet-numbers/ieee-lsap-list.html > http://www.ethermanage.com/ethernet/enet-numbers/cisco-lsap-list.html > http://www.ethermanage.com/ethernet/descript-troubleshoot.html > > Mention a NULL LSAP with the address fields set to > zero. > > Maybe your zero, zero is just the job. You should get the IEEE > 802.2 doc and check yourself though. You used to be > able to buy them on-line and download a pdf. 00 is the "null" LSAP, so 00,00 should be OK - i think this can be used as an 802.2 type "ping". >
> Here are a few possible workarounds. > > Avoid broadcast and m'cast dest MAC. > > Could you send a packet with an invalid CRC? > Then everything would ignore it. > > What happens if you send it to your own MAC address? > Everything else would ignore it, but it might possibly come back to > you. it shouldnt - but this is what cisco and various others such as the Wellfleet / Bay Nortel routers do for a "test" packet. Also - there is a Ethernet type code 9000 which i think is allocated for "test" - i have seen this "in the wild" with a sniffer. >
--
> Buy a NIC, get its address then destroy it. Use that > address as the destination address. > > Send a 'good' frame eg an ICMP echo request or an ARP request. > > Just pick a destination MAC, the chances of > ever coming into contact with it are rather small. > Unless:- you sell millions of them in which case send > me my share:) > > Look through the OUI list and pick an unlikely vendor, say > one that has not yet been taken over and is not too likely to be. > > http://www.iana.org/assignments/ethernet-numbers > > How about: > > 080056 Stanford University > > Not likely to be taken over any day soon and not > too likely to go into the mass production of > printers, games consoles or refrigerators. > > Of course they might sell it. Regards stephen_hope@xyzworld.com - replace xyz with ntl | |||||||||||||||||||
|
Posted by on October 15, 2005, 11:24 am
Please log in for more thread options
<snip>
You should get the IEEE
> 802.2 doc and check yourself though. You used to be
> able to buy them on-line and download a pdf. 802 group stuff specs are currently free (this is mentioned on charles spurgeon's website) http://standards.ieee.org/getieee802/ | |||||||||||||||||||
| Similar Threads | Posted |
| Creating a test packet | October 13, 2005, 11:21 am |
| Multicast test bed | October 18, 2004, 8:51 am |
| Built-In Bit Error Test? | July 5, 2005, 6:53 pm |
| About link test pulse | October 29, 2004, 4:31 pm |
| How to test network in windows? | September 19, 2008, 2:15 am |
| Software to test network speed? | September 9, 2004, 9:05 pm |
| Utilities to Test Gigabit Ethernet | July 25, 2006, 2:39 am |
| How to test spanning tree protocol | July 28, 2006, 1:03 am |
| switches or other devices which respond to Test frames? | May 23, 2008, 2:31 pm |
| XGMII end of packet | July 20, 2004, 4:19 am |
| Generate packet with bad CRC | June 21, 2007, 9:15 am |
| Packet recorder | August 8, 2007, 11:36 am |
| Packet capture | November 28, 2007, 12:07 am |
| I search a test with benchmark of several ethernet gigabit switch | September 1, 2006, 11:18 am |
| ethernet packet preamble | May 4, 2006, 2:01 pm |
> some
> equipment as a valid packet, though this would be rare. What I want is
> some sort of test packet that is universally ignored. The data as it
> is
> now would start with an LLC header of "00 00 00". Is this likely to be
> misinterpreted, or is there something else I can try?