Corruption of in-memory data detected.

[Marc Schmitt <schmitt@xxxxxxxxxxx>]

Hello all,

I`m currently setting up a 1.2TB file server and consider using XFS. The 
configuration is:
- STL2 MB with 512MB RAM, 2 CPUs (866MHz)
- 3 x 3ware RAID Controller
- 20 x 80GB disks attached in blocks of 4 with RAID 5 (5x240GB)
- md0 is a RAID 0 over the 5 RAID 5 devices

I tested a couple of XFS-enabled kernels and got the following results 
when using mongo.pl:
Output of mongo.pl (same for all tests):

reiser_fract_tree is up to date ...
mongo_slinks is up to date ...
mongo_read is up to date ...
map5 is up to date ...
summ is up to date ...
umount: /local: not mounted
meta-data=/dev/md0               isize=512    agcount=280, 
agsize=1048576 blks
data     =                       bsize=4096   blocks=293055600, imaxpct=25
         =                       sunit=8      swidth=40 blks, unwritten=0
         =                       imaxbits=32
naming   =version 2              bsize=4096
log      =internal log           bsize=4096   blocks=32768
realtime =none                   extsz=163840 blocks=0, rtextents=0

mongo_single_process, the_set_of_param.N=0 of 3
Results in file : /root/mongo_pl/results/bla

1.Create files of median size 100 bytes (1 processes)...
Create : open: Invalid argument
real 0.34
user 0.04
sys 0.28

Create : open: No such file or directory
real 0.04
user 0.04
sys 0.00

Create : open: No such file or directory
real 0.04
user 0.04
sys 0.00

total Create time: 0 sec.
Used disk space (df) : 480 KB
Total dirs: 1
Total files: 0
Illegal division by zero at ./mongo.pl line 377.


Console outputs:

Test 1: 2.4.12-xfs #2 SMP (linux-2.4.12-xfs-2001-10-11.patch) (gcc 
version egcs-2.91.66 19990314/Linux (egcs-1.1.2 release))
XFS mounting filesystem md(9,0)
xfs_force_shutdown(md(9,0),0x8) called from line 1020 of file 
xfs_trans.c.  Return address = 0xc01f4409
Corruption of in-memory data detected.  Shutting down filesystem: md(9,0)
Please umount the filesystem, and rectify the problem(s)


Test 2 : 2.4.8-mdk-xfs #10 SMP

xfs_force_shutdown(md(9,0),0x8) called from line 1013 of file 
xfs_trans.c.  Return address = 0xc01f78b9
Corruption of in-memory data detected.  Shutting down filesystem: md(9,0)
Please umount the filesystem, and rectify the problem(s)


Test 3: 2.4.7-10SGI_XFS_PR1smp #1 SMP

XFS mounting filesystem md(9,0)
xfs_force_shutdown(md(9,0),0x8) called from line 1020 of file 
xfs_trans.c.  Return address = 0xc01d53f9
Corruption of in-memory data detected.  Shutting down filesystem: md(9,0)
Please umount the filesystem, and rectify the problem(s)


If I use any of those kernels with only a single partition (240GB) 
instead of running RAID 0 over the 5 partitions, everything works fine.
The "2.4.9 is bad" (by Florin Andrei) thread mentions the same 
"Corruption of in-memory data detected"-type error, but refers to 
xfs_bmap.c, not xfs_trans.c.
Please let me know if you want me to run any other tests.


Regards,
        Marc