On Tue, Mar 26, 2002 at 11:18:00AM +0100, Thomas Winkler wrote:
> Am Mon, 2002-03-25 um 06.08 schrieb Timothy Shimmin:
> > Looking at getfacl(1), the "#effective" comment refers
> > to the effect the mask ACE has on all groups and named user
> > ACEs whose permissions are reduced.
> > In your case, your mask ACE is "r-x" so this will
> > potentially reduce permissions for group and named-user ACEs.
> > I guess the comment is there as a reminder of what the mask ACE is doing.
> > So did you have a different command to set the acl for "otheruser"
> > which had a different mask ACE ???
> since we are running many systems on xfs for quite a while now i had
> some more machines i was able to test. it seems as if this problem only
> occurs on 1 out of 3 systems i tried with. all are running the same
Perhaps these 1 out of 3 are incorrectly set up (but maybe not,
just a theory)...
> kernel. first i thought i did forget to update all xfs related progs
> (acl, attr, xfsprogs...) but nothing changed after an update to the
> versions from the 2.4.16 kernel. now i am kind of lost. it seems to work
2.4.16 or 2.4.18? either way, you have version problems below...
> the way i want it to, but unfortunatelay not on the system it has to.
> i also checked kernel settings and umask entries for the users.
> what could cause this behaviour?
> now i am using (all devels installed):
if your kernel version == 2.4.16, your "attr" package is too new.
if your kernel version == 2.4.18, everything except for "attr" is too old.
ie. using kernel < 2.4.18, all tools must be < 2.0.0;
and using kernel >=2.4.18, all tools must be >=2.0.0.
Perhaps 1 in 3 of your machines have mismatched kernel/userspace?
> are there any known issues? could find any on my list archives...
> > BTW, general userland ACL questions are best sent to
> > acl-devel@xxxxxxxxxxx (check out: http://acl.bestbits.at/)
> > now that we're using common userspace code for ext2, ext3 and XFS.
> yes, but i searched lots of places and could only find solaris
> questions, concerning the same behaviour. in this cases it where bugs,
> thats why i asked here. i am sorry...
It's better to ask there because the author of the acl tools hangs
out over there, whereas we will have to go dig through his code
to try to answer your questions, which is inefficient for everyone.