Re: Regression in 2.6.23-rc2-mm2, mounting cpusets causes a hang

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 15 Aug 2007 14:00:29 -0400
Lee Schermerhorn <Lee.Schermerhorn@hp.com> wrote:

> On Wed, 2007-08-15 at 12:12 -0500, Serge E. Hallyn wrote:
> > Quoting Paul Jackson (pj@sgi.com):
> > > Lee wrote:
> > > > [altho' methinks CPUSET should select CONTAINERS rather than
> > > > depend on it...]
> > > 
> > > Good point -- what do you think, Paul Menage?
> > 
> > Paul mentioned (http://www.spinics.net/lists/linux-containers/msg03775.html)
> > that he was asked not to use select in Kconfigs any more.
> 
> That is unfortunate.  I view CPUSET is a "first class" option that
> happens to be built on CONTAINERS.  IMO, I shouldn't have to know this
> implementation detail in order to select CPUSET--indeed, even to be
> offered the optio--in 'make *config'.  However, it will be hidden if I
> haven't already selected containers.  Or do I misunderstand the
> implications of "depends on"?
> 
> Andrew:  is the problem with select something fatal or just too
> difficult to fix?
> 

I don't recall what specifically went wrong when we were using select
in there, but it was pretty ugly.

We're hitting more and more problems with this "select is
broken"-versus-"depends-on user inferface sucks" problem.  I do think that
a suitable fix could be found at the user interface level: show the
"cpusets" option to the user, but don't let them select it.  Instead, tell
the user what needs to be enabled so they can access this option.

But I haven't seen the slightest sign of anyone thinking about doing
anything along those lines.