Re: [fuse-devel] [RFC v3 0/2] Support for posix acls in fuse

[Michael Theall <mtheall@linux.vnet.ibm.com>]

On Mon, 2016-08-01 at 22:39 -0500, Seth Forshee wrote:
> On Mon, Aug 01, 2016 at 04:03:57PM -0700, Nikolaus Rath wrote:
> > 
> > On Aug 01 2016, Seth Forshee <seth.forshee@canonical.com> wrote:
> > > 
> > > There's also a problem with default acls that I'm not sure
> > > there's
> > > currently a solution for. As far as I can tell FUSE_CREATE
> > > doesn't give
> > > back any indication of whether an existing file was opened or a
> > > new file
> > > was created. Without knowing that I cannot know whether or not
> > > the inode
> > > should inherit default acls from its parent.
> > Would it be possible for the FUSE file system to implement this
> > inheritance by a dumb-copy of the ACL-related xattrs of the parent?
> > 
> > This would solve your problem. But in addition to that, it also
> > seems to
> > me that even if a file system uses default_permissions for ACL
> > handling,
> > it may want implement a different policy for permission
> > inheritance...
> In my opinion it's preferable for the kernel to handle all of this
> and
> for the filesystems to need only xattr support to get support for
> acls.
> 
> The inheritance behavior is standard for default acls. It shouldn't
> be
> left to individual filesystems to decide the inheritance policy.
> 
> Thanks,
> Seth

In case this has any bearing, my filesystem would in fact interpret the
ACLs from the xattrs in order to apply them to the backing filesystem
(which supports ACLs but through a non-xattr interface). In my
particular case, it would be okay for the kernel to assume the
inherited ACLs since it should be the same as if the kernel requested
the ACLs after creation.

Regards,
Michael Theall