[Bug 202] XFS data corruption

[bugzilla-daemon@oss.sgi.com]

http://oss.sgi.com/bugzilla/show_bug.cgi?id=202





------- Additional Comments From xhejtman@hell.ascs.muni.cz  2002-12-11 10:14 -------
this script creates always correct files:
----
#!/bin/sh

i=0; while [ $i != 100 ]; do 
dd if=/dev/urandom of=/opt/tmp/xfstest.$i bs=1k count=40; sleep 1; echo $i;
i=$[$i+1]; done

dd if=/dev/zero of=/opt/tmp/flush bs=1M count=4

sleep 40

echo "finished"
----

while this does not:
----
#!/bin/sh

i=0; while [ $i != 100 ]; do echo "sjdfhakfj" > /opt/tmp/xfstest.$i; sleep 1;
echo $i; i=$[$i+1]; done

dd if=/dev/zero of=/opt/tmp/flush bs=1M count=4

sleep 40

echo "finished"
----

I've tried the second one to run 2 times and all 200 files were corrupted and
the last one file 'flush' had zero length.

By the way how can on disc cache to cache only data but not metadata?
I've found out that on disc cache cannot be turned off with hdparm -W0 /dev/hdc,
it says turnin off write cache but hdparm -i /dev/hdc says writecache enabled.

Is is possible to do ordered writes as ext3 does? I've reformated /home
partition to ext3 to avoid data looses and so far it looks pretty good however I
would rather prefer xfs.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.