Re: Race condition between putback_lru_page and mem_cgroup_move_list

[Lee Schermerhorn <Lee.Schermerhorn@hp.com>]

On Thu, 2008-08-07 at 20:00 +0900, KOSAKI Motohiro wrote:
> Hi
> 
> > If you mean the "active/inactive list transition" in
> > shrink_[in]active_list(), these are already batched under zone lru_lock
> > with batch size determined by the 'release pages' pvec.  So, I think
> > we're OK here.
> 
> No.
> 
> AFAIK shrink_inactive_list batched zone->lru_lock, 
> but it doesn't batched mz->lru_lock.
> 
> then, spin_lock_irqsave is freqently called.

Ah, I see what you mean.  Yes, the mem cgroup zone lru_lock will be
cycled frequently as each back of pages is put back during reclaim.  So,
you'd like to eliminate the mz lru_lock, move the mem cgroup zone info
under the corresponding zone lru_lock and move the page between memcg
lists atomically with adding to global lru lists?  

Lee

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>