On Tue, Jun 29, 2010 at 08:37:22AM -0400, Christoph Hellwig wrote: > I don't see a patch in this set which refuses writeback from the memcg > context, which we identified as having large stack footprint in hte > discussion of the last patch set. > It wasn't clear to me what the right approach was there and should have noted that in the intro. The last note I have on it is this message http://kerneltrap.org/mailarchive/linux-kernel/2010/6/17/4584087 which might avoid the deep stack usage but I wasn't 100% sure. As kswapd doesn't clean pages for memcg, I left memcg being able to direct writeback to see what the memcg people preferred. > Meanwhile I've submitted a patch to xfs to allow reclaim from kswapd, > and just prevent it from direct and memcg reclaim. > Good stuff. > Btw, it might be worth to also allow kswap to all writeout on ext4, > but doing that will be a bit more complicated than the btrfs and xfs > variants as the code is rather convoluted. > Fully agreed. I looked into it and got caught in its twisty web so postponed it until this much can be finalised, agreed upon or rejected - all pre-requisities to making the ext4 work worthwhile. -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>