Messages in this thread | | | Date | Thu, 23 Jun 2011 22:51:28 +0900 | Subject | Re: [PATCH] mm: preallocate page before lock_page at filemap COW. (WasRe: [PATCH V2] mm: Do not keep page locked during page fault while charging it for memcg | From | Hiroyuki Kamezawa <> |
| |
2011/6/23 Michal Hocko <mhocko@suse.cz>: > On Thu 23-06-11 22:01:40, Hiroyuki Kamezawa wrote: >> 2011/6/23 Michal Hocko <mhocko@suse.cz>: >> > On Thu 23-06-11 19:01:57, KAMEZAWA Hiroyuki wrote: >> >> On Thu, 23 Jun 2011 11:02:04 +0200 >> >> Michal Hocko <mhocko@suse.cz> wrote: >> >> >> >> > On Thu 23-06-11 17:08:11, KAMEZAWA Hiroyuki wrote: >> >> > > On Thu, 23 Jun 2011 09:41:33 +0200 >> >> > > Michal Hocko <mhocko@suse.cz> wrote: >> >> > [...] >> >> > > > Other than that: >> >> > > > Reviewed-by: Michal Hocko <mhocko@suse.cz> >> >> > > > >> >> > > >> >> > > I found the page is added to LRU before charging. (In this case, >> >> > > memcg's LRU is ignored.) I'll post a new version with a fix. >> >> > >> >> > Yes, you are right. I have missed that. >> >> > This means that we might race with reclaim which could evict the COWed >> >> > page wich in turn would uncharge that page even though we haven't >> >> > charged it yet. >> >> > >> >> > Can we postpone page_add_new_anon_rmap to the charging path or it would >> >> > just race somewhere else? >> >> > >> >> >> >> I got a different idea. How about this ? >> >> I think this will have benefit for non-memcg users under OOM, too. >> > >> > Could you be more specific? I do not see how preallocation which might >> > turn out to be pointless could help under OOM. >> > >> >> We'll have no page allocation under lock_page() held in this path. >> I think it is good. > > But it can also cause that the page, we are about to fault in, is evicted > due to allocation so we would have to do a major fault... This is > probably not that serious, though.
For other purpose, I have(had) other patch to prevent it (and planned to post it.)
The basic logic is...
1. add a new member variable to vm_area_struct as
vma->vm_faulting_to
2. at __do_fault(), set vm_faulting_to as
vma->vm_faulting_to = pgoff.
3. chec vma->vm_faulting_to at page_referenced_file() as
if (pgoff (Was page->index) == vma->vm_faulting_to) referenced++
Then, the page which someone is waiting for page-fault will be marked as referenced and go KEEP_LOCKED.
(vm_faulting_to can be cleared after we got lock_page()).
In corner case, several threads which shares vma may fault into a vma. But this will help typical case and have no overheads, I think.
Thanks, -Kame
| |