home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.unix.sys5.r4:1115 comp.dcom.modems:18868
- Newsgroups: comp.unix.sys5.r4,comp.dcom.modems
- Path: sparky!uunet!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!spool.mu.edu!umn.edu!csus.edu!netcom.com!gerg
- From: gerg@netcom.com (Greg Andrews)
- Subject: Re: uucp woes
- Message-ID: <1993Jan4.081829.17615@netcom.com>
- Summary: The packet was good, but the computer rejected it.
- Keywords: uucp esix worldblazer sas
- Organization: Netcom Online Communications Services (408-241-9760 login: guest)
- References: <1993Jan01.191128.4566@wisdom.bubble.org>
- Date: Mon, 4 Jan 1993 08:18:29 GMT
- Lines: 90
-
- jeff@wisdom.bubble.org (Jeff Ross) writes:
- >
- >here is the saga, running (or trying to run) Esix SVR4 (4.0.4) however I am
- >plaiged by problems with uucico and/or SAS-1.25. To make matters worse
- >there is a telebit worldblazer (LA7.01 roms problems where present with
- >5.01 as well)
- >
- >when I call out from the Esix machine and download files from a remote
- >host the transfer will fail, no if ands or buts about it, all I get
- >are alarms, the alarms will come at random times durring the transfer,
- >but always within the first 40k or so.
- >
- >after borrowing an hp datascope from work, I monitored the line between
- >the modem and computer sent it to telebit for evaluation, their
- >assesment was that the worldblazer was functioning properly and that the
- >unix machine was not, more specifically, the unix machine was
- >"supposedly" dropping a packet and requesting a retransmission which the
- >worldblazer "supposedly" retransmitted correctly. according to telebit
- >they even examined the packets that the telebit was sending to the
- >unix machine to verify the checksum was correct. (it was) thus they
- >concluded that the worldblazer is fine, but the unix machine is at
- >fault.
- >
-
- Jeff, what other conclusion can be drawn? Your analyzer trace showed
- the bytes that the WorldBlazer and your ESIX system exchanged. It logged
- everything the WB sent to the computer. That trace showed a perfect
- transfer up to the failure point. Suddenly, the ESIX uucico started
- rejecting a packet from the WB. Examination of that packet shows:
-
- o The packet had the correct sequence number (0) and carried
- the correct ack for the data packet last sent by the ESIX
- uucico (2).
-
- o The packet passed the header XOR check.
-
- o The packet had the correct number of data bytes (64).
- I.e., the packet delivered by the modem didn't have any
- extra characters or missing ones.
-
- o The packet passed the data checksum test.
-
- o There were no extra characters between packets that could
- have confused uucico.
-
-
- The trace shows that the modem placed a correctly formed data packet onto
- the RS232 cable, but uucico rejected it (paused for several seconds, then
- acked the previous packet number). The modem re-sent that same packet
- 10 times until uucico abandoned the transfer.
-
- If the WB sent a good packet but uucico rejected it, how can any other
- conclusion be drawn but the WB worked correctly and uucico didn't?
-
-
- Tech support's statement that your computer was dropping bytes is just
- a theory. (How do I know that? That theory originated with me.) There's
- no way to look at a trace from the RS232 cable and know what happened to
- the data after it entered the computer.
-
- That theory (dropping bytes) isn't airtight. Usually a packet corrupted
- by dropped bytes will be accepted on the first retransmission. This one
- was rejected every time. Bytes would have to be dropped EVERY time the
- packet is re-sent, and that's very rare. Plus, that kind of problem
- usually manifests itself much sooner in the transfer, especially if it's
- bad enough to occur within the first 70 bytes of the data burst (each
- retransmission started with the rejected packet).
-
- Why did I suggest the computer was dropping bytes? That's the most common
- cause of rejected packets on spoofed uucp transfers. The only other
- theories I have are in areas outside my expertise - for example, something
- is munging the variable that holds the packet number your uucico is
- expecting to receive. It gets a correct packet, but thinks the sequence
- number is wrong, so rejects it. Another variable holds the packet number
- of the last correctly received packet, and that variable isn't munged, so
- the NAK looks normal. (actually, it's an ACK for the previous packet)
-
-
- I can't explain the success of the transfers with the T2500 and direct
- connections except to say the pattern of data flow will be different
- from a WB, and that difference might be uncovering a problem in the
- computer. The analyzer trace shows that the WB sent out a good packet
- and the computer rejected it. The problem isn't in the WB.
-
- --
- ::::::::::::::::::: Greg Andrews gerg@netcom.com :::::::::::::::::::
- ObGuindon: Henry Schultz is worried. He's found a slip of paper
- in his pocket which says he has been passed by inspector
- number 7. He figures they did it while he was asleep.
- :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
-