mitch asked>

We just ran across a strange happening on our F760 filer. One of our folks did a "du -s" on an RCS directory to see how much space it was using:

=> du -s RCS du: RCS/ParseNVP.pm,v: Value too large for defined data type

I've seen this error on Solaris boxes running 32-bit code, such as GNU ls, when they are accessing Netapp volumes and directories with very large files (>2GB.)

I've also seen this on Solaris clients when accessing files on a Linux fileserver, when the dates were outrageously wrong on the file. Like a year of 1935!

If you see this again you might want to look at atime, mtime and ctime to see if something is strange. I've appended a Perl script below called `stat' that will show you all sorts of good things (or bad things), and that may help.

But I don't have a silver bullet for you -- your description does not match what I've seen, unless you did not notice wacky file dates. (ls -l, ls -cl, ls -ul) Was the system clock off on the client?

And got a strange error. In fact, pretty much anything that would do a stat(2) call on this file in this directory would generate the same error message. There were no errors logged on the filer and all weekly messages seem to indicate that all is well, but it still puzzles me.

I ran off for the requisite caffine boost and came back and am now unable to reproduce the problem. I asked the user if they perhaps did something such as recreating the RCS directory from another source... they hadn't.

The file which was complained about is a mere 3422 bytes, and all timestamps indicate that the file has not changed recently.

Has anybody else seen this problem or have any ideas on how/why this may have occurred?

For what it's worth, the NFS client accessing the files is running Solaris 5.7. I unfortunately didn't pop over to one of my BSD boxes before the problem worked itself out.

~mitch

#!/tool/pandora/.package/perl-5.6.0/bin/perl # # Last edited: Mon Oct 5 11:48:02 1992 by Tim Wilson # $Id: stat,v 1.1 1992/10/05 10:53:13 tdw Exp $ # # Stat(2) files and print results. # # Usage: stat [-?] [-l] [-n] [-h] [--] files... # # -l: use lstat(2) instead of stat(2) # -n: print output in numeric format (instead of readable format) # -h: if in numeric format, also print header line # -?: print help message # --: remaining arguments are file names # ###################################################################### # # Copyright (C) 1992 T D Wilson. All rights reserved. # # Permission to copy without fee all or part of this material is # granted provided that the copies are not made or distributed for # direct commercial advantage, the copyright notice and the title and # date appear, and notice is given that copying is by permission of # the author. To copy otherwise, or to republish, requires specific # permission. # # T D Wilson makes no representations about the suitability of this # software for any purpose. It is provided "as is" without express or # implied warranty. # ###################################################################### # # Modification summary: # 5 Oct 1992 Tim Wilson Altered comments; posted to # alt.sources. Revision 1.1. # 11 Jul 1991 Tim Wilson Created. # # Tim Wilson, University of Cambridge Computer Laboratory, UK. # tdw@cl.cam.ac.uk #

sub usage { die "Usage: $progname [-l] [-n] [-h] [--] files...\n"; }

sub Lstat { ($st_dev,$st_ino,$st_mode,$st_nlink,$st_uid,$st_gid,$st_rdev,$st_size, $st_atime,$st_mtime,$st_ctime,$st_blksize,$st_blocks) = lstat(shift(@_)); }

require "stat.pl"; require "ctime.pl";

###################################################################### # # p_perm -- Return part permissions symbolically # # perm: Permissions. Only the low three bits are significant # extra: Extra bit to be encoded with execute (suid, sgid, sticky) # c: Character to be used instead of `x' (`s' or `t') # # Returns -- three character string # # Side effects -- Output # ######################################################################

sub p_perm { local ($perm,$extra,$c) = @_; local ($lower,$upper) = ($c, $c); local ($result); $lower =~ tr/ST/st/; # Guaranteed lower case `s' or `t' $upper =~ tr/st/ST/; # Guaranteed upper case `S' or `T'

$result = ($perm & 04) ? "r" : "-"; $result .= ($perm & 02) ? "w" : "-"; # Want 2D array; do index arithmetic ourselves $result .= ("-", "x", $upper, $lower)[!!$extra * 2 + ($perm & 01)]; }

###################################################################### # # p_mode -- Return symbolic $st_mode (and/or memorized stat structure) # # Arguments are global variables. Ahem! # # Returns -- ten character string # # Side effects -- Output # ######################################################################

sub p_mode { local ($result); # First letter according to type of file TYPE: { -p _ && do {$result = "p"; last TYPE;}; -c _ && do {$result = "c"; last TYPE;}; -d _ && do {$result = "d"; last TYPE;}; -b _ && do {$result = "b"; last TYPE;}; -l $file && do {$result = "l"; last TYPE;};# lstat problems -f _ && do {$result = "-"; last TYPE;}; -S _ && do {$result = "s"; last TYPE;}; warn ("unknown type of file, mode %#o\n", $st_mode); $result = "?"; }

$result .= &p_perm ($st_mode >> 6, -u _, "s"); # User $result .= &p_perm ($st_mode >> 3, -g _, "s"); # Group $result .= &p_perm ($st_mode >> 0, -k _, "t"); # Others }

###################################################################### # # uid, gid -- Convert numeric uid, gid to symbolic # # arg: numeric id # # Returns -- Name associated with id if found, else number # # Side effects -- Accesses /etc/passwd or replacement # ######################################################################

sub uid { local ($uid) = @_; $uid{$uid} = getpwuid($uid) unless defined $uid{$uid}; $uid{$uid} = "#$uid" unless defined $uid{$uid}; $uid{$uid}; }

sub gid { local ($gid) = @_; $gid{$gid} = getgrgid($gid) unless defined $gid{$gid}; $gid{$gid} = "#$gid" unless defined $gid{$gid}; $gid{$gid}; }

###################################################################### # # verbstat -- Print $st_* variable (and/or remembered stat structure) # in multiline verbose format # # Arguments are global variables. Ahem! # # Returns -- nothing # # Side effects -- Output # ######################################################################

sub verbstat { local ($atime, $mtime, $ctime) = (&ctime ($st_atime), &ctime ($st_mtime), &ctime ($st_ctime)); # local ($mode) = &p_mode;

printf ("%s\n", $file); printf ("\tdevice\t\t%#x\n", $st_dev); printf ("\tinode\t\t%d\t\t%#x\n", $st_ino, $st_ino); printf ("\tmode\t\t%s\t%#o\n", &p_mode, $st_mode); printf ("\tnlink\t\t%d\n", $st_nlink); printf ("\towner\t\t%s\n", &uid ($st_uid)); printf ("\tgroup\t\t%s\n", &gid ($st_gid)); printf ("\trdev\t\t%d\n", $st_rdev); printf ("\tsize\t\t%d\n", $st_size); chop ($atime); chop ($mtime); chop ($ctime); printf "\tatime\t\t%d %s\n\tmtime\t\t%d %s\n\tctime\t\t%d %s\n", $st_atime, $atime, $st_mtime, $mtime, $st_ctime, $ctime;

printf ("\tblksize\t\t%d\n", $st_blksize); printf ("\tblocks\t\t%d\n", $st_blocks); }

###################################################################### # Main program ######################################################################

$_ = $0; $progname = m|.*/([^/]+)| ? $1 : $_;

$opt_lstat = 0; # If true use lstat not stat $opt_header = 0; # Add header line $opt_numeric = 0; # One-line numeric format

# No switch clustering with this simple parser

SWITCH: while ($_ = shift) { /^-?/ && &usage; # stat -? ... gives usage /^--$/ && last SWITCH; # -- terminates switches /^-l$/ && ($opt_lstat = 1, next SWITCH); /^-h$/ && ($opt_header = 1, next SWITCH); /^-n$/ && ($opt_numeric = 1, next SWITCH);

unshift (@ARGV, $_); # We are looking at the first file last SWITCH; }

($#ARGV < $[) && &usage; # Must have at least one arg

# Print out header if requested and numeric format

$opt_numeric && $opt_header && print "dev ino mode nlink uid gid rdev size atime mtime ctime blksize blocks\n";

# Stat remaining args and print result

foreach $file (@ARGV) { if ($opt_lstat) { &Lstat ($file); } else { &Stat ($file); }

if (defined ($st_dev)) { if ($opt_numeric) { # fn dev ino md nl ud gd rdv siz atm mtm ctm bks nbk printf "%s %#x %#x %#o %d %d %d %#x %ld %ld %ld %ld %d %d\n", $file, $st_dev, $st_ino, $st_mode, $st_nlink, $st_uid, $st_gid, $st_rdev, $st_size, $st_atime, $st_mtime, $st_ctime, $st_blksize, $st_blocks; } else { &verbstat; } } else { warn "$0: can't ", $opt_lstat ? "lstat" : "stat", " $file: $!\n"; } }

# End of stat

-- Quentin Fennessy Quentin.Fennessy@amd.com Voice: 512.602.3873 Pager: 512.622.6316

+-- On Monday, December 04, 2000 14:54:05 -0800 mitch@netline.com wrote: | We just ran across a strange happening on our F760 filer. One of our | folks did a "du -s" on an RCS directory to see how much space it was | using: | | => du -s RCS | du: RCS/ParseNVP.pm,v: Value too large for defined data type | | And got a strange error. In fact, pretty much anything that would | do a stat(2) call on this file in this directory would generate | the same error message. There were no errors logged on the filer | and all weekly messages seem to indicate that all is well, but it | still puzzles me. | | I ran off for the requisite caffine boost and came back and am now | unable to reproduce the problem. I asked the user if they perhaps | did something such as recreating the RCS directory from another | source... they hadn't. | | The file which was complained about is a mere 3422 bytes, and all | timestamps indicate that the file has not changed recently. | | Has anybody else seen this problem or have any ideas on how/why this | may have occurred? | | For what it's worth, the NFS client accessing the files is running | Solaris 5.7. I unfortunately didn't pop over to one of my BSD boxes | before the problem worked itself out. +--

NFS defines all dates to be 32 bit unsigned values but many clients, including older Solaris clients ignore this and treat it as a 32 bit signed value. Starting with Solaris 7 Sun began enforcing the standard and refuse to accept a date after 2038 as the standard requires. Presumably filers use a 32 bit date and hence can accept it and still follow the standard.

/Michael