file corruption with arla

[Magnus Ahltorp]

> >  This is due to the caching done in the Linux 2.0 xfs module. Killing
> >  arlad should empty the cache, and thus release all refcounts on the
> >  inodes.
> 
> It doesn't always; I started tracking it about a month ago and then got
> sidetracked. 

I know it doesn't; the keyword here is "should". Since run very few
machines with Linux 2.0, I don't do the day-to-day testing with the
Linux 2.0 xfs. Quite a few of you people out there seem to run it
though, and your input it very valuable.

/Magnus
map at stacken.kth.se