2.4.23 + Preempt, JFS Corruption.

2.4.23 + Preempt, JFS Corruption.

[Joshua Schmidlkofer]

Howdy,

   We are migrating from ext3 to jfs and had a weird error.  We have a
dual pentium III 1.4 ghz system.  We have 4GB of ram, and a Mylex
AcceleRAID 170 [DAC960] controller. Intel pro/100 nics, nothing else too
special.  The motherboard chipset is ServerWorks.  We added a 40GB ide
drive to the onboard controller.  It has jfs on it, and we are using
this for the migration.  We have been backing up to it for a few days,
and it went read-only on us.  

  I have dtree page corrupt errors, but no hardware errors.  jfs fsck
errors out as well.

  We saw a load of errors like this in the dmesg:

ERROR: (device ide0(3,1)): DT_GETPAGE: dtree page corrupt
ERROR: (device ide0(3,1)): DT_GETPAGE: dtree page corrupt
ERROR: (device ide0(3,1)): DT_GETPAGE: dtree page corrupt
ERROR: (device ide0(3,1)): DT_GETPAGE: dtree page corrupt
ERROR: (device ide0(3,1)): DT_GETPAGE: dtree page corrupt
ERROR: (device ide0(3,1)): DT_GETPAGE: dtree page corrupt
ERROR: (device ide0(3,1)): DT_GETPAGE: dtree page corrupt


jfs fsck reports:
jfs_fsck version 1.1.4, 30-Oct-2003
processing started: 12/9/2003 9.3.38
The current device is:  /dev/hda1
Block size in bytes:  4096
Filesystem size in blocks:  9769520
**Phase 1 - Check Blocks, Files/Directories, and  Directory Entries
**Phase 2 - Count links
Incorrect link counts detected in the aggregate.
**Phase 3 - Duplicate Block Rescan and Directory Connectedness
**Phase 4 - Report Problems
File system object DF1323074 is linked as:
/slash/home/virtual/[innocent protected]
cannot repair the data format error(s) in this directory.
cannot repair DF1323074.
**Phase 5 - Check Connectivity
**Phase 6 - Perform Approved Corrections
**Phase 7 - Verify File/Directory Allocation Maps
**Phase 8 - Verify Disk Allocation Maps
Incorrect data detected in disk allocation structures.
Incorrect data detected in disk allocation control structures.
 39078080 kilobytes total disk space.
   292595 kilobytes in 88974 directories.
 15368020 kilobytes in 1074664 user files.
        0 kilobytes in extended attributes
   633991 kilobytes reserved for system use.
 23368664 kilobytes are available for use.
File system checked READ ONLY.
ERRORS HAVE BEEN DETECTED.  Run fsck with the -f parameter to repair.
Filesystem is dirty.



(the [innocent protected] just obfuscates the path)

Re: 2.4.23 + Preempt, JFS Corruption.

[Dave Kleikamp]

On Tue, 2003-12-09 at 15:34, Joshua Schmidlkofer wrote:
> Howdy,
> 
>    We are migrating from ext3 to jfs and had a weird error.  We have a
> dual pentium III 1.4 ghz system.  We have 4GB of ram, and a Mylex
> AcceleRAID 170 [DAC960] controller. Intel pro/100 nics, nothing else too
> special.  The motherboard chipset is ServerWorks.  We added a 40GB ide
> drive to the onboard controller.  It has jfs on it, and we are using
> this for the migration.  We have been backing up to it for a few days,
> and it went read-only on us.  
> 
>   I have dtree page corrupt errors, but no hardware errors.  jfs fsck
> errors out as well.

jfs_fsck reports the errors, but is running read-only.  Did you run it
with "-n"?  In read-write mode, it should "fix" the problem (by removing
the corrupt directory and placing its contents in lost+found).

I've seen a few scattered reports of corrupted directories lately, so
there may be a race in the JFS code that I haven't found yet.  I'd be
interested if anyone has found a reproducible way of causing any JFS
corruption.

Thanks,
Shaggy
-- 
David Kleikamp
IBM Linux Technology Center