Re: [patch] mm, thp: always direct reclaim for MADV_HUGEPAGE even when deferred

From: Mel Gorman
Date: Fri Dec 30 2016 - 09:09:00 EST

On Fri, Dec 30, 2016 at 01:56:16PM +0100, Michal Hocko wrote:
> On Fri 30-12-16 12:36:20, Mel Gorman wrote:
> [...]
> > I'll neither ack nor nak this patch. However, I would much prefer an
> > additional option be added to sysfs called defer-fault that would avoid
> > all fault-based stalls but still potentially stall for MADV_HUGEPAGE.
> Would you consider changing the semantic of defer=madvise to invoke
> KSWAPD for !madvised vmas as acceptable. It would be a change in
> semantic but I am wondering what would be a risk and potential
> regression space.

I'd worry a little, but not a lot. The concern would be that kswapd waking
up would reclaim pages and cause major faults that would have remained
resident with the current semantics.

Mel Gorman