"Force Modifier Set", disk noise, and other problems with uVax II

Alan's Home for Wayward Notes File. alan at shodha.enet.dec.com
Sun Jan 6 18:43:19 AEST 1991


In article <71681 at bu.edu.bu.edu>, jdubb at bucsf.bu.edu (jay dubb) writes:
> 
>   I am posting this for a friend of mine, who doesn't have access to
> USENET, so please reply directly to mlevin at jade.tufts.edu.
> 
>     I have a VaxStation II/RC, running Ultrix. It is exhibiting some
> very weird behavior. If any of the following phenomena sound familiar
> to anyone, I'd appreciate hearing ideas on cause and remedy.
> 1) The TK50 tape drive used to work, but now when I use Tar, it begins
>    writing, and then says "tar: I/O write error 0", and hangs
>    (seemingly forever). Nothing (^C, ^Z, etc.) can break it of tar,
>    and if I exit the window, and start a new one, tar refuses to work  
>    saying "/dev/rmt16: no such device or address". This used to only
>    happen on tapes that had been used before, but now happens even on
>    new tapes.

	How long is forever?  I think the TK50 has a fairly long
	reset time.  More than a couple of minutes is to long
	though.  This sounds like hardware in any event.  With
	all the other problems you have a very sick system.

	The ^C part of the problem is simply the driver waiting
	at a higher priority than you can interrupt with the ^C.
	The "no such device or address" is the driver preventing
	two processes from opening the device at the same time.

> 2) The hard-disk makes weird noises, and sometimes makes whistling
>    sounds like a short-wave radio. This is usually followed by a
>    core-dump of DecWrite. DecWrite itself coredumps often, and when it
>    exits (whether it core-dumped or not) leaves "Xlib error 16" messages
>    all over the screen. Sometimes the core-dump is so severe that it
>    kills everything - the windows and all, and you have to start from
>    the beginning "login:" and DIGITAL logo screen.

	Do you know what kind of disk?  Even capacity will tell enough
	if it's a DEC drive.  The whistling (or perhaps chirping) is
	the drive recal noise.  Most likely you're getting some serious
	disk errors.  The DECwrite and Server problems are mostly likely
	errors trying to page.

> 3) Once in a while, the DecWindows screens are invaded by low-level
>    messages saying "re-synching controller", and other cryptic stuff.

	The disk controller expects an answer from the command it
	sends within some amount of time.  The host also expects
	to hear from the controller every now and then.  I think
	the "re-synching controller" message is what the host says
	when it kicks the controller trying to get it's attention.

	This usually indicates a broken controller.

> 4) When accessing a particular file, I get a message saying "Force
>    Modifier Set: LBN = xxx"; the file finally disappeared when I tried
>    to uncompress it. What does this mean???

	When a bad sector is replaced the controller (or host) tries
	very hard to get a good copy of the data.  If it fails it
	sets a field in the sector header saying that the data was
	corrupted.  Once the block gets overwritten it field gets
	cleared.  I would have expected uncompress to get I/O errors
	everything time it touch the block, but it may ignore I/O
	errors.
>  
>     Does anyone have any ideas on what causes this weird stuff to
> happen? 

	You have very sick hardware.  If the system has field
	service support (Customer Services these days) then you 
	should get them to fix it.  Actually you probably need
	them to fix it anyway.  That system sounds real sick.

> Its making my life hell - I've lost several files already, and
> DECWrite's coredumping makes me retype stuff routinely (since for some
> unexplainable reason, the journaling stops as soon as you manually
> save your file!). HELP!!!!

	I think this "feature" of DECwrite is actually documented
	somewhere.  Don't ask my why they do it this way.  I don't
	understand it either.
> 
> 					Mike Levin (mlevin at jade.tufts.edu)

-- 
Alan Rollow				alan at nabeth.enet.dec.com



More information about the Comp.unix.ultrix mailing list