abee79f vfs: Use posix_sys_acl_blob_get_file in vfs_gpfs for posix ACLs

Jeremy Allison jra at samba.org
Thu Oct 11 10:06:57 MDT 2012

On Thu, Oct 11, 2012 at 01:32:27PM +0200, Volker Lendecke wrote:
> On Thu, Oct 11, 2012 at 10:16:04PM +1100, Andrew Bartlett wrote:
> > On Thu, 2012-10-11 at 13:02 +0200, Christian Ambach wrote:
> > > Andrew,
> > 
> > > I do not really understand what the 
> > > expected behavior of of gpfsacl_sys_acl_blob_get_fd() and 
> > > gpfsacl_sys_acl_blob_get_fd() should be. GPFS supports both the posix 
> > > and NFSv4 ACL model, so I guess it will take special care that both 
> > > cases are treated correctly.
> > 
> > The idea is simply:
> > 
> > If there is an NFSv4 ACL, then we give an error.  The layers above this
> > cope with errors, and when the patch to vfs_acl_common is applied
> > (currently there are no callers) it will simply fall back to the hash of
> > the NT ACL.  (And given the simple mapping, this is less important to
> > implement). 
> > 
> > If it is a posix ACL, then call the posix helper function that
> > linearlises the posix ACL into a blob, along with the owner, group and
> > mode.
> > 
> > What was the error BTW?
> You DID notice that you broke code, right? AGAIN?

Ok. I think it's time to revive Volker's original idea of reviewer-must-push
for *all* changes, even for master.

The fault is not entirely Andrew's, both Simo and I missed this
problem (along with the subsequent crash bug) in the review, so
this means we really didn't do a good enough job on this review.

Enforcing reviewer-must-push will concentrate the mind wonderfully
when doing reviews, as you'll have to add a signed-off-by line
in the push to master :-).

What do Team members think ? Do I have anyone to second this ?


More information about the samba-technical mailing list