Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site brl-tgr.ARPA
Path: utzoo!watmath!clyde!burl!ulysses!allegra!bellcore!decvax!genrad!panda!talcott!harvard!seismo!brl-tgr!tgr!speck@cit-vax
From: speck@cit-vax (Don Speck)
Newsgroups: net.unix-wizards
Subject: cp tbuf par fault -> cache data error?
Message-ID: <8457@brl-tgr.ARPA>
Date: Fri, 22-Feb-85 06:07:30 EST
Article-I.D.: brl-tgr.8457
Posted: Fri Feb 22 06:07:30 1985
Date-Received: Tue, 26-Feb-85 20:41:21 EST
Sender: news@brl-tgr.ARPA
Lines: 16


    One of the VAX/750's that we bought recently, panics with the
message "cp tbuf par fault" every 10-50 minutes of non-idle time.
The error bits seem to indicate a cache data error; the Machine
Check Error Summary Register says "bus error", the MCSR says
"uncorrectable data, error lost", and the Cache Error Register
says "data error".  Our other 750 gets these about once a week.

    The machine check handler in vax/machdep.c has a recovery
action for translation buffer errors, but none for cache errors.
Our 4.2 tape is very old (it says #10); did a patch for this come
out later?  It seems that at the very least it could try disabling
the cache, which would be slow but better than crashing outright.
Is there a better way?	Can the cache be cleared by disabling it
and turning it back on?
						Don Speck