On Mon, Apr 08, 2002 at 07:41:36PM -0000, eeh@netbsd.org wrote:
> 
> So. getting back to the original point, if you really are locking up
> due to lack of pages, there is nothing that can be done because you 
> simply need more memory than you have on that machin.  However, if
> you want to handle that situation gracefully, you need to track and
> limit overcommit.

Would it be sensible (or is it done already?) to let kernel code
(eg filesystems) specify a 'callback' routine which will be
called when the system is out of swap (or kva, or...) so they
can free stuff they are hanging on to because it is likely to
be needed?

Clearly this will slow a system down, but crawling is better than
being stopped :-)

Also you need (somehow) to work out whether the 'out of memory'
is caused by general overloading, or by one process grabbing
all the space it can find.  In the latter case you do need to
kill the correct process (and its children!).

OTOH if you have a kma leak in the kernel it is easiest to
spot in the system dump where ALL the physical memory
contains the data item being discarded.

	David

-- 
David Laight: david@l8s.co.uk