Re: [PATCH 4/4] fs: take the ACL checks to common code

[Christoph Hellwig <hch@lst.de>]

On Sat, Jul 23, 2011 at 05:15:16PM +0100, Al Viro wrote:
> > Yes, and if we add a ->set_acl we can take most of the existing boilerplate
> > code completely into posix_acl.c.  I'll see if I can do something like that
> > for v3.2.
>  
> As for ->set_acl(), how are you going to deal with things like ext3_init_acl(),
> where we get transaction handle as argument and pass it down to ext3_set_acl()?
> Or the things like e.g. gfs2_set_mode()...

ext3/4 already store the handle_t (what an awfully misleading name, btw)
in current->journal_info for similar cases where they want to keep to access
it when doing detours through common code.  Similar for gfs2 and gfs2_trans.

I'll have to look into it in more details.  So far my plan is:

 - add a generic xattr method for reading ACLs, we already have can do
   that just based on ->get_acl
 - add a generic xattr method (and thus set of xattr ops) for writing
   ACLs, and add ->set_acl for it.
 - then look into how we can sanely factor the more complicated operations