GELI/UFS: strange data corruption

No at SPAM at nospam at
Mon Apr 3 10:06:17 UTC 2006


the environment:
- Intel Server w 1 x XEON 2.8 / 2GB
- HW-RAID1 (Intel SRCS16 w 2xSeagate NL35 400GB)
- FreeBSD 6.1-BETA4
- GELI encryption of whole amrd0 with Sector-Size 4096
- ufs2 filesystems w Softupdates and Sector-Size 4096

the source:
- SuSE 8.2 reiserfs (3.6?) filesystem mounted ro

the target:
- the ufs2 fs mentioned above

content: 300GB

what happened:
created a geli-device with AES256 on amrd0
setup 3 partitions with bsdlabel on it:
  amrd0.elia: 4096MB BSD-rootfs
  amrd0.elib: 4096MB swap
  amrd0.elib: ~350GB data (the rest of the sectors)
installed bsd w boot from usb-drive (because of geli)
  kernel/bootloader on USB
  root-fs get's mounted from HDD
tried copying the data from the reiserfs disk using
  cd /olddata &&  find . | sort -n | cpio -pdvum /newdata

this step took 9hrs to complete.

the problem:
the count of files is exactly the same
cpio/syslog/dmesg logged absolutely no errors (not even warnings)
the files contain almost the same data, but at the beginning and
at the end of the files there is content mixed up from other files
that have been copied!

i compared plain-text files as well as binary files (database, executables)
and they ALL differ. not even 1 file is binary the same!

any ideas!?

br & cu...

More information about the freebsd-questions mailing list