home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!sun-barr!ames!pacbell.com!att!ucbvax!compuserve.com!72510.553
- From: 72510.553@compuserve.com ("Richard V. Ford")
- Newsgroups: comp.protocols.appletalk
- Subject: Re: Evil Packets
- Message-ID: <920722174148_72510.553_DHB66-2@CompuServe.COM>
- Date: 22 Jul 92 17:41:49 GMT
- Sender: daemon@ucbvax.BERKELEY.EDU
- Distribution: world
- Organization: The Internet
- Lines: 164
-
- Sorry for the delay in sending a response...
-
- I posted several months ago about a problem with DDP broadcast storms. The
- problem has been tracked down to a known bug in the Rival anti-virus program.
- There is a fixed version available if you ask the vendor. Only machnes
- running the Rival software that have ethernet cards cause the problem.
-
- You can find out if the problem is happening on your network by using a
- protocol analyzer and looking for packets of DDP type=0 with the sequence
- 09876543210987654321 in the data portion of the packet. A trace packet
- was included in the original message and is appended below for reference.
-
- I've not yet tracked down what network conditions trigger the storm, but the
- problem has been resolved by installing the newer version of the software.
-
- Log this one in your tech support databases.
-
- -Richard
-
- ------------------------------------------------------------------------------
- Richard Ford
- Falcon Microsystems
- Landover MD 20785 USA
- 301-386-6440 (voice) 341-0187 (fax)
-
- Mail: 72510.553@compuserve.com
-
-
-
-
-
- *****************************************************************************
- original message follows
- *****************************************************************************
-
- To: >internet:info-appletalk@andrew.cmu.edu
- Subject: Evil Packets
-
- Hi -
-
- I'm trying to track down the source of a DDP broadcast storm that takes down
- the ethernet on which it occurs. Any insights from those of you who might have
- seen this before would be greatly appreciated.
-
- Background:
-
- The network topology abstracts to two segments separated by a cisco IGS, one
- side with a FastPath5 (KSTAR 9.1) and two VAXes running PacerShare (AppleTalk
- Routers), and a QuickMail 2.5 server running on a IIci. The other side of the
- cisco is an ethernet with a dozen Fastpaths, an AppleShare Server, and a
- Meeting Maker Server. The network is exclusively Phase II, and most of the
- Fastpaths are FP 4's with upgraded proms and KSTAR 9.1. Each segment has a
- network number range of 1, and a single associated zone name. Extensive
- analysis reveals no routing problems.
-
- The problem:
-
- For some not-yet-explained reason, random ethernet-connected Mac's will start
- streaming packets to address 0.255 (dump appended below) repeatedly at a rate
- exceeding 200/sec. This DDP broadcast storm can only be stopped by powering
- off the offending Mac. During the storm, the ethernet side LED on the
- Fastpaths go solid (LED on) and the PBN drops ccounter in RouterCheck increases
- dramatically. All routing grinds to a halt (no RTMP's are exchanged) and no
- ethernet-connected users see zones. When the storm is stopped, the network
- recovers, one sees a series of AARP's, RTMP's and ZIP's as the routers
- converge. Apparently if a storm is long enough, all the learned routes age out
- of the routers.
-
- Raw and decoded versions of the packet are appended to this message. It was
- interesting to ot that DDP type 0 was used (Inside AppleTalk says don't use
- type 0), and also the pattern 0,9,8,7,6,5,4,3,2,1,0,9,8,7,6,5,4,3,2,1,0 in the
- data portion of the packet (thanks to Tom Evans for noticing the pattern). I
- checked the network with Inter*Poll and no applications are registered as using
- socket 67.
-
- The source machine:
-
- The problem has occured with Macs containing ethernet boards from multiple
- manufacturers (both Asante and 3com boards are on the network), and from
- multiple CPU types (IIsi, IIci etc...). Both system 7.0 and 6.0x machines
- have had this happen. The machine from which the packet in the data capture
- started had an Asante board, Asante ethernet drivers, QuickMail 2.5
- workstation, Meeting Maker and Status*Mac on it. The problem pre-dated the
- Status*Mac installation. The problem does not occur at any "fixed" time
- (i.e. bootup, 10 min after bootup etc...) but does happen every day or two,
- sometimes more often.
-
- I have looked at the data set closely and also found at least one other machine
- sending out packets with DDP type zero, and the 987654321 pattern but have not
- yet been able to isolate the software package sending out the packets. It was
- on the LocalTalk side of a Fastpath and sending the packets out every minute or
- two. I have not yet determined whether streaming occurs on LocalTalk Macs, and
- whether there is some type of "trigger" packet that comes in from the ethernet
- that effects ethernet-connected Mac's.
-
- Does there exist a package that can be run on a Mac and list all sockets in
- use, both those with registered NBP entities and those without? This type of
- software would simplify determing what network package is sending out the
- packets. Is there an easy way to do this with MacsBug?
-
- ---
- raw packet
- ---
-
- Total packets = 1
-
-
- Packet #1 Packet size = 64 Timestamp = 16 (ms)
- 09 00 07 ff ff ff 00 00 94 04 bb e1 00 21 aa aa .............!..
- 03 08 00 07 80 9b 00 19 11 34 00 00 00 01 ff 38 .........4.....8
- 43 43 00 00 00 09 87 65 43 21 09 87 65 43 21 00 CC.....eC!..eC!.
- 00 00 00 00 00 00 00 00 00 00 00 00 14 fe 5b b0 ..............[.
-
- ---
- decoded packet/ NetMinder EN
- ---
- Packet 1
- Size: 64
- T (ms)= 16 (4/15/92 4:12:42 PM)
- Errors: None
-
- Ethernet Header
- Destination =ATalk_BCast
- Source Asante04bbe1
- Length $0021
-
- ATalk Phase 2 Header
- DSAP $aa
- SSAP $aa
- Control $03
- Protocol ATalk
-
- Long DDP Header
- Length 25 (Hops=0)
- Checksum $1134
- Dest Net 0
- Src Net 1
- Dest Node 255
- Src Node 56
- Dest Socket 67
- Src Socket 67
- Type $0
-
- Packet Data
- 00 00 09 87 65 43 21 09 ....eC!.
- 87 65 43 21 .eC!
-
- Pad/CRC Bytes
- 00 00 00 00 00 00 00 00 ........
- 00 00 00 00 00 14 fe 5b .......[
- b0 .
-
- -----
-
- Thank's in advance.
-
- -Richard
-
- Richard Ford
- Falcon Microsystems
- Landover MD 20785 USA
- 301-386-6440 (voice) 341-0187 (fax)
-
- Mail: 72510.553@compuserve.com
-