> Somehow, while setting up rsync and doing test transfers (probably while
> testing as user "named"), we managed to change the ownership of a bunch
> of files and directories, to user "named".
> ...
> Could this occur without a bug in the kernel?
As which user is rsyncd running?
(I'm not sure whether you mean it's running as named, vs running as root
and log in as named.)
What you describe does smell like a bug to me,
but with the info so far i'm not entirely certain whether to point finger
first at kernel or at rsyncd.
We have found that rsync creates files with o+r permissions despite setting
an o-r umask on the relevant user ("backup"). Maybe we could resolve that
by the --chmod argument to rsync [aside - Chris: i recently found that one],
but i mention it in case it is related to the chown issue, who knows whether
it is or isnt related (short of debugging).
/Randall
