First of all, 1.0.14 seems to be the most stable release of unionfs that I
worked with. I found two problems and maybe somebody with more experience
in debugging can reproduce those (or tell me what to do to narrow down the
source of the problem)

I am using uniondbg -g $UNION to invalidate directory caches whenever
something changes a layer underneath the union. in 1.0.14 this command does
not work any more (freezes) and the filesystem is not accessible any more,
causing the machine to halt eventually.

Also I experienced filesystem corruption with the following scenario
mkdir layer1
mkdir layer2
mount -t unionfs -o dirs=layer2=rw:layer1=ro union-dir
rpm -i --root union-dir (some large rpm with many files and dirs)
umount union-dir
rm -r layer2
rm -r layer1

try this 5-6 times and eventually, the directory structure in layer2 gets
corrupted and linux is not able to delete it. the only way to fix this is a
reiserfsck --rebuild-tree. There is no OOPS or ASSERT, just the directory
is un-deleteable (rm -rf says "file not found" and "directory not empty")

Kernel is 2.4.31 and both layers are on the same reiserfs filesystem.

if i rmmod unionfs after each umount, it takes many more tries to trigger
this problem.

Mit freundlichem Gruß/ with kind regards

Heiko Helmle


_______________________________________________
unionfs mailing list
[email protected]
http://www.fsl.cs.sunysb.edu/mailman/listinfo/unionfs

Reply via email to