* nishimura@xxxxxxxxxxxxxxxxx <nishimura@xxxxxxxxxxxxxxxxx> [2010-10-07 12:47:06]: > On Thu, 7 Oct 2010 08:44:59 +0530 > Balbir Singh <balbir@xxxxxxxxxxxxxxxxxx> wrote: > > > * nishimura@xxxxxxxxxxxxxxxxx <nishimura@xxxxxxxxxxxxxxxxx> [2010-10-07 09:54:58]: > > > > > On Wed, 6 Oct 2010 19:53:14 +0530 > > > Balbir Singh <balbir@xxxxxxxxxxxxxxxxxx> wrote: > > > > > > > I propose restricting page_cgroup.flags to 16 bits. The patch for the > > > > same is below. Comments? > > > > > > > > > > > > Restrict the bits usage in page_cgroup.flags > > > > > > > > From: Balbir Singh <balbir@xxxxxxxxxxxxxxxxxx> > > > > > > > > Restricting the flags helps control growth of the flags unbound. > > > > Restriciting it to 16 bits gives us the possibility of merging > > > > cgroup id with flags (atomicity permitting) and saving a whole > > > > long word in page_cgroup > > > > > > > I agree that reducing the size of page_cgroup would be good and important. > > > But, wouldn't it be better to remove ->page, if possible ? > > > > > > > Without the page pointer, how do we go from pc to page for reclaim? > > > We store page_cgroups in arrays now, so I suppose we can implement pc_to_pfn() > using the similar calculation as page_to_pfn() does. > IIRC, KAMEZAWA-san talked about it in another thread. > Yes, correct we do. Your suggestions, IIUC is to reuse part of the flags to store the section number where the pc belongs and then use that to remove ->page pointer. -- Three Cheers, Balbir _______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers
