Mike Fedyk wrote:
Nick Piggin wrote:
Mike Fedyk wrote:
OK, I'll give that a try.
Is the attached patch the latest version of your alternative patch instead of shrink_slab-for-all-zones.patch?
Yes that looks like it. I am actually starting to like this patch
again
Didn't you like what you wrote in the first place ;)
Hmm well no, it still reclaims slab too hard. I just forgot
why I didn't like it :P
now that lowmem is being properly scanned as a result of
highmem scanning.
That's what zone-balancing-fix.patch does, right?
Yes... but actually the kswapd path needs something similar.
That would be the batching patch, but with a scanning value
derived from the current size of the zone's lrus.
I'd just stick with 2.6.3 for now :)