[openib-general] Re: Re: madvise MADV_DONTFORK/MADV_DOFORK

[Gleb Natapov]

On Mon, Feb 20, 2006 at 08:50:30AM -0800, Roland Dreier wrote:
> Probably the best thing to do would be to have libibverbs handle the
> madvise() stuff transparently by default, but perhaps add some hooks
> for more intelligent applications or MPI libraries to do their own thing.
> 
I also think that transparent handling is the best. I am not sure about
hook though. I think a call that will enable/disable madvise() handling
from inside reg_mr might be added to libibverbs so the application can 
handle madvice() by itself if it wants to. I can't think why MPI will want 
to do it by itself.

--
			Gleb.