Yesterday, July 22, 1992, I wrote:
> Yesterday, I had a panic crash of the type:
>
> panic: error in swapping in u-area
>
> I was told that I should change the added swap area to be rwx for
> owner, group, and world. Since then, I have had SEVEN panic crashes of
> the type:
>
> panic: ufc_putpage hole
>
> I called Sun and was told by the Call Screening Division person that he
> didn't know what it was, but that it should be fixed if I change the
> CPU.
>
> Does anyone know what this is?
>
> System: SS1+ with internal 105 MByte drive (that is real flaky and has
> been replaced 8 times in the past 2 years), 3rd party external box with
> 2 Seagate 655 MByte drives, Exabyte 8200, and 150 MByte tape -- no
> problems ever). I am currently running SunOS 4.1.1B.
1. It was pointed out by almost everyone who responded that the
swap file should have 1600 permission (which is logical and
rational) and not 777 as I was told by Sun. When I called
Sun this morning because I had another 2 "ufc_putpage hole"
crashes last night after they put in a new CPU and a new
disk, they were unable to locate why I was told to make the
swap file on the local disk (sorry I didn't say that
earlier) with 777 permission.
2. The unanimous comment was that I was mistaken to follow
Sun's Systems Administrators Guide and "mkfile -n". The
"mkfile" should be done WITHOUT the "-n" option. I have
done this, and I expect no problems. (When I called Sun
this morning, they had no comment about this and was going
to pass the problem to a software engineer.)
I have attached a particularly interesting reply from Dave Yearke
below.
Thanks for the help and rapid responses from:
jdavis@noao.edu (Jim Davis)
"Casper H.S. Dik" <casper@fwi.uva.nl>
yearke@calspan.com (Dave Yearke)
birger@vest.sdata.no (Birger A. Wathne)
"Susan Thielen" <thielen@irus.rri.uwo.ca>
Dave Mitchell <D.Mitchell@dcs.sheffield.ac.uk>
trinkle@cs.purdue.edu
era@niwot.scd.ucar.EDU (Ed Arnold)
Don Christensen <djc@xanadu.acuson.com>
seidc!gwolsk@mips.com (Guntram Wolski)
Charles <mcgrew@dartagnan.rutgers.edu>
****************************** APPENDED EMAIL ******************************
From: yearke@calspan.com (Dave Yearke)
We were getting these for a while in January.
The "ufs_putpage hole" message is normally generated when a "dirty" page
is flushed to a disk block that hasn't been allocated. The common cause
for this is an improperly created swap file. In my case, I don't have
any swap files on this system, other than the swap partition. A Sun tech
support person was able to examine our crash dump file and determine that
this panic was generated by the kernel "update" process, which is called
by "sync" every 10 seconds or so to flush disk buffers back to disk.
Our problem turned out to be an NFS server on a non-UNIX host that was
creating files with negative file sizes under certain rare conditions.
Apparently, Sun's NFS client doesn't do much error checking and will
happily set the file size to a negative number, which utterly confuses
the kernel's filesystem code.
--
Dave Yearke, yearke@calspan.com
"The First Amendment is a tragic amendment because everyone is
going to have his or her feelings hurt and your government is not here to
protect you from having your feelings hurt." --- Kurt Vonnegut
****************************** APPENDED EMAIL ******************************
--
Terence P. Ma, Ph.D. Department of Anatomy
VOICE: 601-984-1654 University of Mississippi Med. Ctr.
FAX: 601-984-1655 2500 North State Street
INTERNET: tpm@anat.UMSMED.EDU Jackson, MS 39216-4505
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:46 CDT