Strange file system errors
Hi,
if I have to reboot my system without a clean shutdown the recovered filesystems are corrupt. Files which I edited recently and maybe not yet written to the buffer are recovered with no or corrupt contents.
For example:
$ ls -l config -rw------- 1 bond bond 19622 Jan 14 12:04 .config $ stat config File: `config' Size: 19622 Blocks: 0 IO Block: 4096 regular file Device: fe01h/65025d Inode: 20967432 Links: 1 Access: (0600/-rw-------) Uid: ( 1000/ bond) Gid: ( 1000/ bond) Access: 2004-01-14 12:27:01.000000000 +0100 Modify: 2004-01-14 12:04:16.000000000 +0100 Change: 2004-01-14 12:04:16.000000000 +0100 $ cat config $ less config "config" may be a binary file. See it anyway? ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@ [...] $ strace -e write cat config write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096 write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096 write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096 write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096 write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 3238) = 3238
I can force this error by pressing the reset button with an open file in my text editor. It seems to have nothing to do with the underlying file system. I have XFS and ext3 partitions. I would like to know, if it is a LVM problem.
Kernel 2.6.1
LVM version: 2.00.08 (2003-11-14) Library version: 1.00.07-ioctl (2003-11-21) Driver version: 4.0.0
Signature · 3 lines
Kai Weber » [email protected] http://www.glorybox.de
Quoted 30 lines · Kai Weber
Hi,
if I have to reboot my system without a clean shutdown the recovered filesystems are corrupt. Files which I edited recently and maybe not yet written to the buffer are recovered with no or corrupt contents.
For example:
$ ls -l config -rw------- 1 bond bond 19622 Jan 14 12:04 .config $ stat config File: `config' Size: 19622 Blocks: 0 IO Block: 4096 regular file Device: fe01h/65025d Inode: 20967432 Links: 1 Access: (0600/-rw-------) Uid: ( 1000/ bond) Gid: ( 1000/ bond) Access: 2004-01-14 12:27:01.000000000 +0100 Modify: 2004-01-14 12:04:16.000000000 +0100 Change: 2004-01-14 12:04:16.000000000 +0100 $ cat config $ less config "config" may be a binary file. See it anyway? ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@ [...] $ strace -e write cat config write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096 write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096 write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096 write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096 write(1, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 3238) = 3238
Quoted 4 lines
I can force this error by pressing the reset button with an open file in my text editor. It seems to have nothing to do with the underlying file system. I have XFS and ext3 partitions. I would like to know, if it is a LVM problem.
No.
Quoted 16 lines
Kernel 2.6.1
LVM version: 2.00.08 (2003-11-14) Library version: 1.00.07-ioctl (2003-11-21) Driver version: 4.0.0Signature · 9 lines
Kai Weber » [email protected] http://www.glorybox.de _______________________________________________ linux-lvm mailing list [email protected] http://lists.sistina.com/mailman/listinfo/linux-lvm read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
Signature · 15 lines
Regards,
Heinz -- The LVM Guy --
*** Software bugs are stupid.
Nevertheless it needs not so stupid people to solve them ***
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Heinz Mauelshagen Sistina Software Inc.
Senior Consultant/Developer Am Sonnenhang 11
56242 Marienrachdorf
Germany
[email protected] +49 2626 141200
FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-* Heinz Mauelshagen <[email protected]>:
Quoted 4 lines
Quoted 2 lines
system. I have XFS and ext3 partitions. I would like to know, if it is a LVM problem.
No.
Strange! If I do the test on my /boot partition, which is no LVM partition, the journal is recovered and the files are not corrupt.
Any suggestions (if it is no LVM problem) where I could ask about it?
Regards,
Signature · 3 lines
Kai Weber » [email protected] http://www.glorybox.de
Assuming you have IDE drives, it sounds like a write caching problem. This problem is amplified with the large caches on newer drives. A sudden power down causes the the data is the cache to be lost. The kernel doesn't know this has occured, because the drive lies, and claims that the cache has been flushed, resulting in file system _and_ journal corruption.
Try the tests again with write caching off. You can do this with "hdparm -W0 /dev/hdX"
Jord Tanner
Quoted 13 lines · Kai Weber
* Heinz Mauelshagen <[email protected]>:
Quoted 4 lines
Quoted 2 lines
system. I have XFS and ext3 partitions. I would like to know, if it is a LVM problem.
No.
Strange! If I do the test on my /boot partition, which is no LVM partition, the journal is recovered and the files are not corrupt.
Any suggestions (if it is no LVM problem) where I could ask about it?
Regards,
Signature · 1 line
Jord Tanner <[email protected]>
* Jord Tanner <[email protected]>:
Quoted 2 lines
Try the tests again with write caching off. You can do this with "hdparm -W0
/dev/hdX"
Thanks for your help. Until today I can not force the filesystem corruption with write caching off. Your tip seems to work. The downsize of disabled write caching seems to be a slightly slower system when writing large files to the disks.
OK. End Of Topic. and now to something completely different.
Signature · 3 lines
Kai Weber » [email protected] http://www.glorybox.de