Hi,
I have run a simple test on my JFS partition (in "bigdir" directory):
for i in `seq 1 1000000`; do touch $i; done
There were no problems, but then I decided to delete the bigdir directory
(it was rather useless ;-> ). Simple 'rm -r bigdir' and my disk-led is red.
When the led went out I switched to the proper console and... the rm was
still running. Hm, I will see how many files are still there 'ls bigdir'
and... nothing happend (it worked before 'rm'). ls becoms dead and I was not
able to kill it (even with 'kill -9'). Ok, so 'strace -p `pidof rm`'...
nothing. Reboot.
% mount /dev/hda3 /mnt
mount: wrong fs type, bad option, bad superblock on /dev/hda3,
or too many mounted file systems
% dmesg | tail -4
Jan 27 22:47:47 moon kernel: JFS development version: $Name: v1_0_13 $
Jan 27 22:47:47 moon kernel: jfs_mount: Mount Failure: File System Dirty.
Jan 27 22:47:47 moon kernel: Mount JFS Failure: 22
Jan 27 22:47:47 moon kernel: jfs_mount failed w/return code = 22
% fsck.jfs -v /dev/hda3
fsck.jfs version 1.0.13, 25-Jan-2002
The current device is: /dev/hda3
(chklog) FSCK Open(...READ/WRITE EXCLUSIVE...) returned rc = 0
(chklog) FSCK Primary superblock is valid.
(chklog) FSCK The type of file system for the device is JFS.
Block size in bytes: 4096
File system size in blocks: 1048241
Phase 0 - Replay Journal Log
(chklog) FSCK:LOGREDO: Journal log is now open. (major = 0x00, minor =
0x00)
(chklog) FSCK:LOGREDO: Allocating for ReDoPage: (d) 4096 bytes
(chklog) FSCK:LOGREDO: Allocating for NoDoFile: (d) 4096 bytes
(chklog) FSCK:LOGREDO: Allocating for BMap: (d) 26704 bytes
(chklog) FSCK:LOGREDO: Allocating for IMap: (d) 20336 bytes
(chklog) FSCK:LOGREDO: Allocating for IMap: (d) 2048 bytes
(chklog) FSCK:LOGREDO: Allocating for IMap: (d) 2048 bytes
(chklog) FSCK:LOGREDO: Log record for Sync Point at: 0x02eaab8
(chklog) FSCK:LOGREDO: Beginning to update the Inode Allocation Map.
(chklog) FSCK:LOGREDO: Done updating the Inode Allocation Map.
(chklog) FSCK:LOGREDO: Beginning to update the Block Map.
(chklog) FSCK:LOGREDO: Done updating the Block Map.
(chklog) FSCK:LOGREDO: End of log found at logend = 0x033fcc0
(chklog) FSCK:LOGREDO: Synch point record number: 0x02eaab8
(chklog) FSCK:LOGREDO: Synch point record address: 0x02ea7c8
(chklog) FSCK:LOGREDO: Number of log records: (d) 1571
(chklog) FSCK:LOGREDO: Number of Do blocks: (d) 51
(chklog) FSCK:LOGREDO: Number of NoDo blocks: (d) 36
(chklog) FSCK logredo returned rc = 0
File system is clean.
(chklog) FSCK All observed inconsistencies have been repaired.
(chklog) FSCK Filesystem has been marked clean.
(chklog) FSCK **** Filesystem was modified. ****
(chklog) FSCK processing terminated: 1/27/2002 23.11.30 with return code: 0.
% mount /dev/hda3 /mnt
%
Mounted, but:
% strace rm -r bigdir
--- big cut ----
lstat("1,023e+06", {st_mode=S_IFREG|0644, st_size=0, ...}) = 0
access("1,023e+06", W_OK) = 0
unlink("1,023e+06") = 0
lstat("1,02401e+06", {st_mode=S_IFREG|0644, st_size=0, ...}) = 0
access("1,02401e+06", W_OK) = 0
unlink("1,02401e+06") = 0
lstat("1,02402e+06",
Not good, ^C: Ok. Once again:
% strace rm -r bigdir
--- big cut ----
lstat("bigdir", {st_mode=S_IFDIR|0755, st_size=8040448, ...}) = 0
access("bigdir", W_OK) = 0
chdir("bigdir") = 0
lstat(".", {st_mode=S_IFDIR|0755, st_size=8040448, ...}) = 0
brk(0x8054000) = 0x8054000
open("/dev/null", O_RDONLY|O_NONBLOCK|O_DIRECTORY) = -1 ENOTDIR (Not a
directory)
open(".", O_RDONLY|O_NONBLOCK|O_LARGEFILE|O_DIRECTORY) = 3
fstat(3, {st_mode=S_IFDIR|0755, st_size=8040448, ...}) = 0
fcntl(3, F_SETFD, FD_CLOEXEC) = 0
brk(0x8056000) = 0x8056000
getdents(3,
I'm here. I don't know what to do. It is absolutely non-production
partition, so I can do what you want with it. Just let me know what to
do to find out where is the bug (I suppose bug in JFS, don't you?).
Linux 2.4.17 with (only) jfs-1.0.13 patch
disk: Seagate ST320011A 20GB UDMA100
partition: 4GB
mainboard: Soltek SL-75DRV
Hardware is rather new, but pretty tested, so I don't suppose a hardware
problem. I hope you understand me and can help. Thanks,
--
Michal Kuratczyk <[EMAIL PROTECTED]>
_______________________________________________
Jfs-discussion mailing list
[EMAIL PROTECTED]
http://www-124.ibm.com/developerworks/oss/mailman/listinfo/jfs-discussion
