CITI acl patch is dangerous.
Kevin Coffman
kwc at citi.umich.edu
Fri Oct 26 15:12:42 EDT 2007
On 10/26/07, J. Bruce Fields <bfields at fieldses.org> wrote:
> On Fri, Oct 26, 2007 at 09:50:18AM -0400, Kevin Coffman wrote:
> > I just saw something yesterday that indicated there is a requirement
> > that the groups be sorted. That was the first I had heard of that
> > requirement. We are not doing anything to sort them, so there may be
> > a problem there also?
>
> From examining the code--nfsd_setuser calls set_current_groups() with
> the groups that are passed down, and set_current_groups) calls
> groups_sort(), which should sort them for us.
OK, thanks! I was just thinking about the user-land code. I hadn't
even considered the kernel code.
> It's working on a copy of the group_info associated with the context,
> so it has to redo that sort every time. So I suppose it might be night
> be nice for us to pass them down sorted just for efficiency reasons, but
> it doesn't look like it's required for correctness.
Great. A possible future optimization.
More information about the NFSv4
mailing list