Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

From: Gleb Natapov (glebn_at_[hidden])
Date: 2007-05-20 09:02:27


On Fri, May 18, 2007 at 06:04:07PM -0400, Patrick Geoffray wrote:
> Hi Gleb,
>
> Gleb Natapov wrote:
> > new madvice flag was implemented that allows userspace to mark certain
> > memory to not be copied to a child process. This memory is not mapped in
> > a child at all, no even VMA created for it. In the parent this memory is
>
> Ah, that explain your previous mention of segfault. For static
> registrations, the ones that are the real problem with fork because of
> the infinite exposure, it's much simpler to use MAP_SHARED...
>
Yes, for internal allocations MAP_SHARED should solve the problem. But the
real problem is silent data corruption after fork() with user allocations.
Segfault is preferable to data corruption.

--
			Gleb.