[cifs-discuss] creator/owner - cifs/zfs
Afshin Salek
Afshin.Ardakani at Sun.COM
Thu Apr 23 11:10:35 PDT 2009
Creator Owner/Creator Group ACEs in Windows only take part in
inheritance not in access check and that's how CIFS server works.
This means if you have a Creator Owner ACE in ACL of a directory
and you create a file in that directory, the file's ACL will contain
an ACE for whoever created that file.
Creator Owner/Creator Group are not equivalent of ZFS owner@/group@
entries because owner@/group@ entries participate in access check.
Afshin
Espen Martinsen wrote:
> Hi, does anyone knows the solution to this:
>
> I'm setting up a mixed environment, cifs/nfs on a ZFS filesystem, and are experiencing the following:
>
> In windows, there is a special acl-entry called "CREATOR OWNER", (S-1-3-0), which
> can be assigned permissions.
> (there is also a "CREATOR GROUP" S-1-3-1)
>
> The purpose of this is a bit like "chmod u+rwx", ie it sets the permissions for whoever is
> the owner of the file. When the cifs-server (I'll guess) receives this information, it resolves
> this to an acl called 'user:<myusernam>'
>
> Like this:
> $ ls -lV win.txt
> ----------+ 1 myname mygroup 0 april 23 11:26 win.txt
> user:myname:rwxpdDaARWcCos:-------:allow
> group:mygroup:r-x---a-R-c--s:-------:allow
>
> Now, the problem is when the file is "given" to another user, ie with "chown", then
> the acl for the previous user is still there.
> $pfexec chown coop win.txt
> $ ls -lV win.txt
> ----------+ 1 coop mnemonic 0 april 23 12:11 win.txt
> user:myname:rwxpdDaARWcCos:-------:allow
> group:mygroup:r-x---a-R-c--s:-------:allow
>
> I can't understand it anyway else that it could be
> nice to have a way to map:
>
> creator user : S-1-3-0 : to the owner at -acl (chmod A+owner@:.......)
> creator group: S-1-3-1 : to the group at -acl
>
> The same way that the special-ACL "Everyone" in windows map to the
> everyone at -ACL.
>
> On a ZFS filesystem, we then would have a way to set up the "unix-perm-bits rwx" from
> the windows-gui.
>
> Is all this making any sense ???
>
> yours
> espenM
More information about the cifs-discuss
mailing list