Re: [-mm][PATCH 8/10] fix shmem page migration incorrectness onmemcgroup

[KAMEZAWA Hiroyuki]

On Fri, 27 Jun 2008 16:57:56 +0900
"MinChan Kim" <minchan.kim@xxxxxxxxx> wrote:

> On Fri, Jun 27, 2008 at 2:41 PM, KOSAKI Motohiro
> <kosaki.motohiro@xxxxxxxxxxxxxx> wrote:
> >> > mem_cgroup_uncharge() against old page is done after radix-tree-replacement.
> >> > And there were special handling to ingore swap-cache page. But, shmem can
> >> > be swap-cache and file-cache at the same time. Chekcing PageSwapCache() is
> >> > not correct here. Check PageAnon() instead.
> >>
> >> When/How shmem can be both swap-cache and file-cache ?
> >> I can't understand that situation.
> >
> > Hi
> >
> > see,
> >
> > shmem_writepage()
> >   -> add_to_swap_cache()
> >      -> SetPageSwapCache()
> >
> >
> > BTW: his file-cache mean !Anon, not mean !SwapBacked.
> 
> Hi KOSAKI-san.
> Thanks for explaining.
> 
> In the migrate_page_move_mapping, the page was already locked in unmap_and_move.
> Also, we have a lock for that page for calling shmem_writepage.
> 
> So I think race problem between shmem_writepage and
> migrate_page_move_mapping don't occur.
> But I am not sure I am right.
> 
> If I am wrong, could you tell me when race problem happen ? :)
> 
You are right. I misundestood the swap/shmem code. there is no race.
Hmm...

But situation is a bit complicated.
- shmem's page is charged as file-cache.
- shmem's swap cache is still charged by mem_cgroup_cache_charge() because
  it's implicitly (to memcg) converted to swap cache. 
- anon's swap cache is charged by mem_cgroup_uncharge_cache_page()

So, uncharging swap-cache of shmem by mem_cgroup_uncharge_cache_page() is valid.
Checking PageSwapCache() was bad and Cheking PageAnon() is good.
(From maintainance view)

I think the patch is valid but my patch description contains wrong information.
Andrew, could you drop this ? I'll rewrite the patch description.

Sorry,
-Kame

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/