From Moon at SCRC-TENEX  Sun Jul 31 00:00:00 1983
From: Moon at SCRC-TENEX (David A. Moon)
Date: Sunday, 31 July 1983, 00:00
Subject: MC crashes
In-Reply-To: The message of 30 Jul 83 19:36-EDT from Christopher C. Stacy <CSTACY at MIT-MC>
Message-ID: <mailman.1352.1479821133.4695.its@nocrew.org>

Somewhere on those scribbled-all-over pieces of paper attached
to the machine it says how to get some information when it seems
to be hung.  There's a KLDCP command with some name I don't
remember ("ALL" maybe) that prints a bunch of stuff, and a command
file (J KLHUNG I think) that prints about three pages of stuff.
Most of the three pages is useless, but buried in there are the
micro and macro PCs and various error status bits.  Somewhere unless
some idiot has thrown it away there is a notebook containing the
KLDCP commands, including how to get the current interrupt level
and other things like that that are in the lights on KAs.  I have
no idea any more what the names of the commands were.  Presumably
all the information it is possible to get is in the KLHUNG output
somewhere.

There should be a microcode listing on-line in the UCODE directory
in which you can look up micro PCs.  They appear embedded in some
syntax in the left margin, maybe "U nnnnn," or something like that.

Microcode hung really means that some timeout inside of KLDCP went
off, I believe, and doesn't really directly imply anything about
the microcode.  I think I have seen this caused by main memory
parity errors sometimes.

"ITS IS DOWN" means that a keep-alive counter incremented by the
60-cycle clock interrupt handler is not being incremented, at least as
far as the pdp11 you are typing to (whichever of the two it is) thinks.
ITS could be halted, looping at interrupt level, not taking interrupts
because the interrupt hardware, I/O bus, or clock is broken, not
running because the processor is broken, or working but not talking to
the pdp11 because the interface is broken.

The page-fault-in-system obviously should be investigated to see whether
there was a reference to a wrong address, or a page fault on a good
address, or code was clobbered, or the machine wasn't executing instructions
correctly, or whatever.  This one could easily be a software bug, so
keep that in mind.

Someone should delete the useless crash dump Taft made after a J NTSDDT.

From CSTACY at MIT-MC  Sun Jul 31 01:36:00 1983
From: CSTACY at MIT-MC (Christopher C. Stacy)
Date: July 30 1983 19:36 EDT
Subject: MC crashes
In-Reply-To: Msg of 07/30/83 16:24:06 from TAFT
Message-ID: <mailman.1353.1479821133.4695.its@nocrew.org>


Dave,

I know you are real busy over there, but if you get a second could you
please shed any light you have on this?  It's pretty much beyond me.


    Date: 07/30/83 16:24:06
    From: TAFT
    To:   CSTACY
    cc:   ELLEN
    Re:   MC crash

       MC crashed twice this afternoon.  The first time was with a
    microcode hung, but the second time the lights were just dead and
    there was no message whatsoever on the console. 

    I tried to take a dump of this, but I am not sure that I got it 
    right.  I hit BREAK on the console, did a "J NTSDDT" and then did:
    $Y CRASH;NO MSG
    It seems to have tried writing out a dump, but perhaps this was
    the wrong way to get into DDT ? (It appeared to load a new one ?)
    Anyway I left the dump in CRASH;NO MSG, maybe someone wants to
    look at it. Otherwise it should be deleted.
    			Jon

The command "DDT" will get you into the current DDT.  J NTSDDT runs a
command file which resets some things and loads up a fresh DDT.  I
don't remember if it clobbers the KL10 state, but I think it does.

Something is very wrong with MC.  It has the following three problems,
sometimes several times a day:

  o Microcode hung.
    Maybe this is not new, maybe it is a known expected ITS bug???
    Is that why it has never been looked into?

  o Machine appears halted, typing on hardware ttys gets ITS IS DOWN,
    no messages on console.
    I guess this acts this way because the machine has either gone into a
    super tight loop (on the order of JRST .), or the microcode is off
    in never-never land and some instruction never returns.  Maybe one 
    thing to do is StoP the machine and look at the PC - I didn't yet.
    If the machine were really halted, I think the console-11 would say so.
    The IO-11 times out doing some protocol to ITS instead, and 
    prints ITS IS DOWN.

  o ITS crashes with PAGE FAULT IN SYSTEM at the exact same address.

Chris

From CSTACY at MIT-MC  Thu Jul 28 10:33:00 1983
From: CSTACY at MIT-MC (Christopher C. Stacy)
Date: July 28 1983 04:33 EDT
Subject: CFTPing from ITS
In-Reply-To: Msg of Jul 28 1983 3:19AM-EDT from Martin David Connor <MARTY at MIT-OZ>
Message-ID: <mailman.1354.1479821134.4695.its@nocrew.org>

    Date: Thursday, July 28, 1983 3:19AM-EDT
    From: Martin David Connor <MARTY at MIT-OZ>
    To:   bug-its, bug-file at MIT-OZ
    Re:   CFTPing from ITS
    I now find that supplying a non-existent username to LOGIN
    when CFTPing to MC or ML now hangs indefinitely.
    I discovered this when running a batch job that CFTPs and
    logs in as OZHOST.
    This used to work.  Was there a change to the file server on MC?



In FILE 520, ITS 1438, on MIT-MC:

KMP and I saw this too, but thought we might be imagining something.
Someone had broken FILE after version 518 by forgetting that on ITS,
ASCII byte pointers must be explicit (440700) and cannot be set up
with HRROI.  It was MPVing because of this in COMST0 when called from
the stuff around GETNAM.  Fixed in the source and installed on the ITS 
systems.

From ED at MIT-MC  Tue Jul 12 07:32:00 1983
From: ED at MIT-MC (Ed Schwalenberg)
Date: July 12 1983 01:32 EDT
Subject: No subject
Message-ID: <mailman.1356.1479821134.4695.its@nocrew.org>

BIL's first problem comes from the archive device handler running out
of room to fit the archive (archives can only hold about 170. blocks of stuff.)
The archive device should report DEVICE FULL rather than seeming to win and
producing 0-length files.  Secondly, file creation dates are randomly munged,
off by an hour or so.  Thirdly, as we all know, the archive device needs to be
done right for a change.

From WGD at MIT-MC  Tue Jul 12 06:20:00 1983
From: WGD at MIT-MC (William G. Dubuque)
Date: July 12 1983 00:20 EDT
Subject: No subject
Message-ID: <mailman.1357.1479821134.4695.its@nocrew.org>

:MOVE <file>,AR0:RAT;  produces a file of length 0 on AR0. Unfortunately
i got screwed and moved a few unbackedup files there before noticing this.
If you list the archive you will notice a few 0 length files there where
this happened. All other files were :MOVEd there previously without such
lossage. Is there a new DDT or ARCDEV or somesuch at the heart of this?
Is there any way of getting around this (I want to preserve the creation
date so :COPY wont do)?

From Ian at MIT-OZ  Mon Jul  4 03:32:00 1983
From: Ian at MIT-OZ (Ian Macky)
Date: Jul 3 1983  21:32 EDT (Sun)
Subject: Artificial Intuition?
In-Reply-To: Msg of 3 Jul 1983  15:57-EDT from Alan Bawden <ALAN at MIT-MC>
Message-ID: <mailman.1358.1479821134.4695.its@nocrew.org>


The source for ARGUS I found on MC was very very old, and is
pretty grim stuff (having been written before I had any idea
what I was doing)... I dunno where a more recent one is;  I'll
keep looking.

From ALAN at MIT-MC  Sun Jul  3 21:57:00 1983
From: ALAN at MIT-MC (Alan Bawden)
Date: July 3 1983 15:57 EDT
Subject: Artificial Intuition?
In-Reply-To: Msg of 3 Jul 1983 09:44 EDT from Christopher C. Stacy <CSTACY>
Message-ID: <mailman.1359.1479821134.4695.its@nocrew.org>

    Date: 3 July 1983 09:44 EDT
    From: Christopher C. Stacy <CSTACY>
    This is CSTACY.  I sat down at DEVON's console here just now, where
    DEVON had a disowned ARGUS looking for any CSTACY's.  I started up an
    EMACS to look at something.  I am sure you can imagine my surprise
    when I saw:  [Here is CSTACY].

    Startled, I ^Zd out of the EMACS I took my hands off the keyboard.
    Ever dutiful, ARGUS barked:  [There goes CSTACY].  
    I suppose this bug has something to do with my having typed CSTACY$^S
    before running the EMACS.  If it does not, congratulations!

'Tain't a bug in my opinion.  

Had always presumed that ARGUS did this by looking at all the XUNAME's in
the system, but looking at the source it looks like it just reads TTY.
Wierd!

From CSTACY at MIT-MC  Sun Jul  3 15:44:00 1983
From: CSTACY at MIT-MC (Christopher C. Stacy)
Date: July 3 1983 09:44 EDT
Subject: Artificial Intuition?
Message-ID: <mailman.1360.1479821134.4695.its@nocrew.org>


This is CSTACY.  I sat down at DEVON's console here just now, where
DEVON had a disowned ARGUS looking for any CSTACY's.  I started up an
EMACS to look at something.  I am sure you can imagine my surprise
when I saw:  [Here is CSTACY].

Startled, I ^Zd out of the EMACS I took my hands off the keyboard.
Ever dutiful, ARGUS barked:  [There goes CSTACY].  
I suppose this bug has something to do with my having typed CSTACY$^S
before running the EMACS.  If it does not, congratulations!