Re: [PATCH v3 2/3] hfsplus: add implementation of the ACLs support

["J. Bruce Fields" <bfields@fieldses.org>]

On Tue, Mar 05, 2013 at 10:52:56AM +0400, Vyacheslav Dubeyko wrote:
> On Mon, 2013-03-04 at 16:26 -0500, J. Bruce Fields wrote:
> > On Sun, Mar 03, 2013 at 07:29:32PM +0400, Vyacheslav Dubeyko wrote:
> > > From: Vyacheslav Dubeyko <slava@dubeyko.com>
> > > Subject: [PATCH v3 2/3] hfsplus: add implementation of the ACLs support
> > > 
> > > This patch adds implementation of the ACLs support for hfsplus driver.
> > 
> > Much of this actually appears to have started as a cut-and-paste of
> > fs/nfsd/nfs4acl.c.
> > 
> > Could you explain what you needed to change from the nfsd code?
> > 
> > It looks like we should be able to share at least some of this code
> > instead of copying it.  The posix<->NFSv4 ACL mapping is a bit tricky
> > and I'd prefer we not have to fix bugs in two different instances of it.
> > 
> 
> Yes, I totally agree with you. I used fs/nfsd/nfs4acl.c as a basis for
> implementation of ACLs support in hfsplus. But it was a first step
> because I can't used implemented in nfsd functionality without changes.
> The ACLs in hfsplus have some peculiarities (special raw ACLs
> representation and slightly special set of flags) and I needed to
> achieve first of all working implementation of ACLs support. Now it is
> done and I think over how to share common code between hfsplus and nfsd.
> Maybe it can be useful and for other drivers. As a preliminary vision, I
> think that we can have something like pair of files:
> include/linux/nfsv4_acl.h (declarations) and fs/nfsv4_acl.c
> (implementation). The implementation should share common code with
> possibility to implement specialized part functionality in the concrete
> driver.

Sure.

> > Also: have you looked at the latest rich ACL patches?  That would
> > provide a native interface to NFSv4-like ACLs, which would really be
> > much more useful.
> > 
> 
> Could you point out about what patches you are talking? Maybe these
> patches correct my vision of sharing code between hfsplus and nfsd.

They implement NFSv4-like ACLs for ext4, and hook them up to nfs and
nfsd:

	http://www.bestbits.at/richacl/

Ideally hfsplus would use the same interface.  The translation wouldn't
be perfect, but it'd be easier than posix<->nfsv4.

The last review I can find is:

	http://marc.info/?l=linux-fsdevel&m=131944979121959&w=2

which comes across as a rant, but I think there are some concrete things
to work on there that just need a volunteer to take on.

--b.