Re: [patch 3/3] cpusets: add memory_spread_user option

[Lee Schermerhorn <Lee.Schermerhorn@hp.com>]

On Fri, 2007-10-26 at 10:54 -0700, Paul Jackson wrote:
> > Will it handle the case of MPOL_INTERLEAVE policy on a shm segment that
> > is mapped by tasks in different, possibly disjoint, cpusets.  Local
> > allocation does, and my patch does.  That was one of the primary
> > goals--to address an issue that Christoph has with shared policies.
> > cpusets really muck these up!
> 
> It probably won't handle that.  I don't get along too well with shmem.

Not surprising :).  shmem doesn't get along too well with cpusets.

> 
> Can you to an anti-shmem bigot how MPOL_INTERLEAVE should work with
         ^ explain ?
> shmem segments mapped in diverse ways by different tasks in different
> cpusets?  What would be the key attribute(s) of a proper solution?
> Maybe if we keep it simple enough, I can avoid mucking it up too much
> this time around.

Personally, I'm of the opinion "if it hurts when you do that, don't do
that".  I have uses for shared memory and mempolicies on the same, but
they don't involve sharing shmem [nor mapped files] between cpusets nor
dynamically changing cpusets.  So, my approach would be to document the
issues clearly [another reason I'd like to see cpuset man pages] and
make sure that folks can't accidentally trip over them.  But, I suppose
all the documentation in the world won't stop some people from hurting
themselves.  As my grandmother used to tell me, "children and fools
shouldn't play with sharp tools."  [Then she'd always ask me, "Which one
are you?"  I guess time has answered that question...]

Lee