Re: [PATCH v7 2/2] mm: make lru_add_drain_all() selective

[Tejun Heo <tj@kernel.org>]

Hello,

On Tue, Aug 13, 2013 at 11:46:29PM -0700, Andrew Morton wrote:
> What does "nest" mean?  lru_add_drain_all() calls itself recursively,
> presumably via some ghastly alloc_percpu()->alloc_pages(GFP_KERNEL)
> route?  If that ever happens then we'd certainly want to know about it.
> Hopefully PF_MEMALLOC would prevent infinite recursion.
> 
> If "nest" means something else then please enlighten me!
> 
> As for "doing it simultaneously", I assume we're referring to
> concurrent execution from separate threads.  If so, why would that "buy
> us anything"?  Confused.  As long as each thread sees "all pages which
> were in pagevecs at the time I called lru_add_drain_all() get spilled
> onto the LRU" then we're good.  afaict the implementation will do this.

I was wondering whether we can avoid all allocations by just
pre-allocating all resources.  If it can't call itself if we get rid
of all allocations && running multiple instances of them doesn't buy
us anything, the best solution would be allocating work items
statically and synchronize their use using a mutex.  That way the
whole thing wouldn't need any allocation.

Thanks.

-- 
tejun

--
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>