Messages in this thread | | | Date | Mon, 30 Nov 2009 12:38:33 +0000 (GMT) | From | Hugh Dickins <> | Subject | Re: [PATCH 2/9] ksm: let shared pages be swappable |
| |
On Mon, 30 Nov 2009, KOSAKI Motohiro wrote: > > After this patch, the number of shared swappable page will be unlimited. > > Probably, it doesn't matter. I mean > > - KSM sharing and Shmem sharing are almost same performance characteristics. > - if memroy pressure is low, SplitLRU VM doesn't scan anon list so much. > > if ksm swap is too costly, we need to improve anon list scanning generically.
Yes, we're in agreement that this issue is not new with KSM swapping.
> btw, I'm not sure why bellow kmem_cache_zalloc() is necessary. Why can't we > use stack?
Well, I didn't use stack: partly because I'm so ashamed of the pseudo-vmas on the stack in mm/shmem.c, which have put shmem_getpage() into reports of high stack users (I've unfinished patches to deal with that); and partly because page_referenced_ksm() and try_to_unmap_ksm() are on the page reclaim path, maybe way down deep on a very deep stack.
But it's not something you or I should be worrying about: as the comment says, this is just a temporary hack, to present a patch which gets KSM swapping working in an understandable way, while leaving some corrections and refinements to subsequent patches. This pseudo-vma is removed in the very next patch.
Hugh
> > ---------------------------- > + /* > + * Temporary hack: really we need anon_vma in rmap_item, to > + * provide the correct vma, and to find recently forked instances. > + * Use zalloc to avoid weirdness if any other fields are involved. > + */ > + vma = kmem_cache_zalloc(vm_area_cachep, GFP_ATOMIC); > + if (!vma) { > + spin_lock(&ksm_fallback_vma_lock); > + vma = &ksm_fallback_vma; > + }
| |