Return to BSD News archive
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msuinfo!uwm.edu!lll-winken.llnl.gov!ames!newsfeed.gsfc.nasa.gov!news!kstailey
From: kstailey@leidecker.gsfc.nasa.gov (Kenneth Stailey)
Newsgroups: comp.os.386bsd.bugs
Subject: NetBSD 1.0 fix 4 small primary swap
Date: 29 Nov 1994 03:03:34 GMT
Organization: NASA Goddard Space Flight Center -- Greenbelt, Maryland USA
Lines: 29
Distribution: world
Message-ID: <KSTAILEY.94Nov28220334@leidecker.gsfc.nasa.gov>
NNTP-Posting-Host: leidecker.gsfc.nasa.gov
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
I have an 8MB primary swap & 16MB core, so if my system ever panics
and dumps into it it will be truncated.
There is a bug in the code that determines if the dump needs to be
truncated. It bubbles up to the surface with this message at boot time:
savecore: read: invalid argument
It's a signed vs. unsigned comparison bug (very popular type, esp w/upgrades)
in /usr/src/sys/arch/i386/i386
--- machdep.c.DIST Fri Oct 14 10:11:44 1994
+++ machdep.c Fri Nov 25 00:31:14 1994
@@ -783,7 +783,7 @@
dumplo &= ~(ctod(1)-1);
/* If it does not fit, truncate it by moving dumplo. */
- if (dumplo < ctod(1)) {
+ if (dumplo < (long)(ctod(1))) {
dumplo = ctod(1);
dumpsize = dtoc(nblks - dumplo);
}
Since ctod() is typecast to unsigned, dumplo will be too! No warnings
even! Since dumplo will start out negative, it will be huge when it
becomes unsigned and the comparison will fail.
Ken