xfs
[Top] [All Lists]

A possible xfs bug: cp/mv to/from an xfs filesystem

To: <linux-xfs@xxxxxxxxxxx>
Subject: A possible xfs bug: cp/mv to/from an xfs filesystem
From: J Landman <landman@xxxxxxxxxxxx>
Date: Tue, 26 Jun 2001 07:49:44 -0400 (EDT)
In-reply-to: <3B386DF9.AAC07E7@kscanners.com>
Sender: owner-linux-xfs@xxxxxxxxxxx
Folks:

   Let me know if this is not the appropriate forum for this.  I have what
I will call an annoying problem.  It seems that cp/mv have stopped working
between xfs and the other file systems.  Here are the basics:

uname -a : Linux genome.dtw.macsch.com 2.4.5-xfs #2 Thu Jun 21 03:06:30 EDT 
2001 i686
unknown

(plain 2.4.5 using the xfs patches from last week for 2.4.5)

file system /work is a RAID 0 /dev/md0 device, striped across two disks on
one controller.  Other file systems are reiserfs (or some ext2).

Here is the problem

(from a reiserfs filesystem)

[landman@xxxxxxxxxxxxxxxxxxxxx:~]
11 >cp XML-Config-0.2.tar.gz /work/
Segmentation fault (core dumped)

(whoops)

[landman@xxxxxxxxxxxxxxxxxxxxx:~]
12 >cp XML-Config-0.2.tar.gz xml.tgz

(ok)

now cd to /work (the xfs filesystem)
[landman@xxxxxxxxxxxxxxxxxxxxx:/work]
15 >cp installer  i
Segmentation fault

so we see that it may be independent of where it starts from (xfs/reiserfs
to xfs)

so I ran strace on the file, and this is what I got near the end (after
the umask)

umask(0)                                = 02
lstat64("i", {st_mode=S_IFREG|0755, st_size=0, ...}) = 0
stat64("i", {st_mode=S_IFREG|0755, st_size=0, ...}) = 0
stat64("installer", {st_mode=S_IFREG|0755, st_size=4332356, ...}) = 0
stat64("i", {st_mode=S_IFREG|0755, st_size=0, ...}) = 0
open("installer", O_RDONLY|O_LARGEFILE) = 3
open("i", O_WRONLY|O_TRUNC|O_LARGEFILE) = 4
fstat64(4, {st_mode=S_IFREG|0755, st_size=0, ...}) = 0
fstat64(3, {st_mode=S_IFREG|0755, st_size=4332356, ...}) = 0
--- SIGSEGV (Segmentation fault) ---
+++ killed by SIGSEGV +++

looking at /bin/cp, I see

[landman@xxxxxxxxxxxxxxxxxxxxx:/work]
17 >ldd /bin/cp
        libc.so.6 => /lib/libc.so.6 (0x40029000)
        /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)


[landman@xxxxxxxxxxxxxxxxxxxxx:/work]
18 >rpm -qf /lib/libc.so.6
glibc-2.2.2-4mdk

This is a mandrake 8.0 system.  I used the original Russell Mandrake
patched kernel until I started to notice this.  I can move files around
using tar.  Tar uses the same libraries as cp, with the additional librt,
and libpthread.  This is all being done using the tcsh.  This uses the
fileutils-4.0.  I downloaded and handcompiled 4.1 of fileutils, and I get
the same behavior.

What else should I do?  I just want to make this go away.  File system was
built using the 1.0 mkfs.  I have run xfs_check and xfs_repair -n on it.
I can back it up and rebuild the FS if needed.  Let me know what you
think.

Joe

-- 
 Joe Landman,
 landman@xxxxxxxxxxxx
 joe.landman@xxxxxxxxxxxxxxx


<Prev in Thread] Current Thread [Next in Thread>