On Wed, May 07, 2008 at 06:12:32PM -0700, Christoph Lameter wrote:
quoted text
> Andrea's mm_lock could have wider impact. It is the first effective 
> way that I have seen of temporarily holding off reclaim from an address 
> space. It sure is a brute force approach.

The only improvement I can imagine on mm_lock, is after changing the
name to global_mm_lock() to reestablish the signal_pending check in
the loop that takes the spinlock and to backoff and put the cap to 512
vmas so the ram wasted on anon-vmas wouldn't save more than 10-100usec
at most (plus the vfree that may be a bigger cost but we're ok to pay
it and it surely isn't security related).

Then on the long term we need to talk to Matt on returning a parameter
to the sort function to break the loop. After that we remove the 512
vma cap and mm_lock is free to run as long as it wants like
/dev/urandom, nobody can care less how long it will run before
returning as long as it reacts to signals.

This is the right way if we want to support XPMEM/GRU efficiently and
without introducing unnecessary regressions in the VM fastpaths and VM
footprint.
--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/