Re: [RFC v7 26/41] richacl: Apply the file masks to a richacl

[J. Bruce Fields]

On Thu, Sep 24, 2015 at 12:14:40AM +0200, Andreas Gruenbacher wrote:
> 2015-09-23 23:05 GMT+02:00 J. Bruce Fields <bfields@xxxxxxxxxxxx>:
> > On Wed, Sep 23, 2015 at 10:40:18PM +0200, Andreas Gruenbacher wrote:
> >> 2015-09-23 22:33 GMT+02:00 J. Bruce Fields <bfields@xxxxxxxxxxxx>:
> >> > The same could be said if there's a group-i-belong-to:rwx::allow entry,
> >> > do we make that exception too?
> >>
> >> We cannot because that would be incorrect for all other group members.
> >
> > OK.  So people have to learn how the group mask works anyway, and now
> > they have to learn a special exception to that rule.
> >
> > I don't like having this exception.  Or making the richacl->v4acl
> > translation dependent on the owner.
> >
> > But I admit it's surprising to that an 0700 mask with
> > "bfields:rwx::allow" ACL denies access to a bfields-owned file.
> 
> I fully understand your point. This kind of acl is one of the the
> first things people will try, and nobody is going to accept when
> access is denied in this case though.
> 
> Things are made worse by the fact that Windows has the concept of
> owner@ or group@ entries for inheritable permissions but not for
> effective ones; it will always produce and expect "bfields:rwx::allow"
> type entries instead of "owner@:rwx::allow" type entries. I'm not sure
> if Samba could bridge that gap.

I guess Samba's only choice on reading an ACL will be to split OWNER@
ACEs into inheritable and effective parts and then replace the "who" on
the latter by the current owner.


On writing do you think it should try to translate ACEs for users
matching the current owner to OWNER@ ACEs, or are you assuming it should
leave those untouched?

Sambas needs here seem most likely to be the determining factor, so I
just want to make sure I understand.

--b.

> The fact that we cannot handle entries for groups the owner is in in a
> similar way is not a big deal; it's not surprising that changing the
> group file mode permission bits affects group entries.
> 
> Thanks,
> Andreas
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/