RFC CART Implementation ver 2

Hi All,

Marcelo was ofcourse right, and I needed to do scary stuff to avoid calling page_referenced() while holding zone->lru_lock. Thanks Rik for making my thik head see that :-)

So here is an update of my code.

Changes:
- Changed the hash algo. for the nonresident buckets It is worse now, so it has to change again :-(
- Made the nonresident code blank out a paged in cookie.
- Rewrote the cart code (yet again) to work on temp lists much like refill_inactive and co.

I put all that code in cart.c instead of swap.c to get a clearer distinction on what was general page reclaim logic and what was part of the page cache.

Just before mailing (and ofcourse untested) I changed the calls to cart_rebalance_*() to take nr_dst/2 + 1 as target because I got too much OOMs, no idea if this solved it. When i put nr_scan = sc->swap_cluster_max (in shrink_cache) I go no OOMs, work left there.

Kind regards,

Peter Zijlstra

--
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>