On Thu, 18 Sep 2008 16:56:17 -0700
Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> Thanks, Kame! I'll look at the lazy lru patches and see if I can find anything.
> Do you have a unified patch anywhere, I seem to get confused with the patches, I
> see 10/9, 11/9 and 12/9. I'll do some analysis when I find some free time, I am
> currently at plumbers conference.
> 
I need some more tests and cleaning up, adjustments to new mmotm.
If you want to see temporal one, try attachted tgz. (deleted CC: list)
Enjoy conference :)
I think I should avoid vmalloc(), at first....

Brief description. page_count_fix.patch may confuse you.

special_mapping_fix.patch
	=> don't account special pages which is not on LRU because we make use of
	   usual LRU page's protocol to do our jobs.
swap-cache-under-lock.patch
	=> move swap-cache accounting under lock for avoid unnecessary race.
nolimit_root.patch
	=> make root cgroup to be unlimited. no changes from previous one.
atomic_flags.patch
	=> make page_cgroup->flags to allow atomic ops.
account_move.patch
	=> implements "move account" logic for new force_empty.
new_force_empty.patch
	=> New force_empty, which moves account rather than discards.
make_mapping_null.patch
	=> garauntee page->mapping is NULL before calling uncharge_cache_page.
stat.patch
	=> optimize statitstics and add "unlikely" to charge loop.
page_cgroup.patch
	=> pre-allocate all page_cgroup at boot. access by hash.
pcg_boost.patch
	=> speed up access to page_cgroup vie per-cpu lookaside buffer
page_count_fix.patch
	=> get/put page's count at charge/uncharge.
	   This is for preventing the page from begin reused while it's on memcg.
lazy_lru_free.patch
	=> free page_cgroup from memcg's LRU in lazy way.
lazy_lru_add.patch
	=> add page_cgroup to memcg's LRU in lazy way.
misc.patch
	=> some optimzation now under test.

Thanks,
-Kame