Re: [PATCH 1/5] vmscan: remove all_unreclaimable check from direct reclaim path completely

From: KOSAKI Motohiro
Date: Thu Mar 24 2011 - 01:35:23 EST


> On Thu, 24 Mar 2011 11:48:19 +0900 (JST) KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx> wrote:
>
> > > On Thu, 24 Mar 2011 11:11:46 +0900 (JST) KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx> wrote:
> > >
> > > > Subject: [PATCH] vmscan: remove all_unreclaimable check from direct reclaim path completely
> > >
> > > zone.all_unreclaimable is there to prevent reclaim from wasting CPU
> > > cycles scanning a zone which has no reclaimable pages. When originally
> > > implemented it did this very well.
> > >
> > > That you guys keep breaking it, or don't feel like improving it is not a
> > > reason to remove it!
> > >
> > > If the code is unneeded and the kernel now reliably solves this problem
> > > by other means then this should have been fully explained in the
> > > changelog, but it was not even mentioned.
> >
> > The changelog says, the logic was removed at 2008. three years ago.
> > even though it's unintentionally. and I and minchan tried to resurrect
> > the broken logic and resurrected a bug in the logic too. then, we
> > are discussed it should die or alive.
> >
> > Which part is hard to understand for you?
>
> The part which isn't there: how does the kernel now address the problem
> which that code fixed?

Ah, got it.
The history says the problem haven't occur for three years. thus I
meant

past: code exist, but broken and don't work for three years.
new: code removed.

What's different? But last minchan's mail pointed out recent
drain_all_pages() stuff depend on a return value of try_to_free_pages.

thus, I've made new patch and sent it. please see it?




--
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/