Re: [PATCH 04/29] cifs: implement get acl method

[Christian Brauner <brauner@kernel.org>]

On Sun, Sep 25, 2022 at 05:53:03PM -0500, Steve French wrote:
> On Fri, Sep 23, 2022 at 3:38 AM Christian Brauner <brauner@kernel.org> wrote:
> >
> > On Thu, Sep 22, 2022 at 10:52:43PM -0500, Steve French wrote:
> > > Looks like the SMB1 Protocol operations for get/set posix ACL were
> > > removed in the companion patch (in SMB3, POSIX ACLs have to be handled
> >
> > Sorry, what companion patch? Is a patch in this series or are you
> > referring to something else?
> 
> I found it - the patch order was confusing (I saw patches 4 and 27,
> but patch 5 was
> missed).  The functions I was asking about were deleted in patch 27 in
> your series but readded in patch 5 which I had missed.

Ok, so we should be good.

> 
> On the more general topic of POSIX ACLs:
> - Note that they are supported for SMB1 (to some servers, including Samba)
> - But ... almost all servers (including modern ones, not just ancient
> SMB1 servers) support "RichACLs" (remember that RichACLs  were
> originally based on SMB/NTFS ACLs and include deny ACEs so cover use
> cases that primitive POSIX ACLs can't handle) but for cifs.ko we have
> to map the local UID to a global unique ID for each ACE (ie id to SID
> translation).  I am interested in the topic for how it is recommended
> to map "POSIX ACLs" to "RichACLs."  I am also interested in making

I think this calls for a session during next years LSFMM but it's a bit
out of scope for this refactoring. :) But we should keep this discussion
in mind!