From: Theo Buehler <tb@openbsd.org>
Subject: Re: uvm_fault_upper_lookup: delay pmap_extract()
To: tech@openbsd.org
Date: Thu, 28 Nov 2024 10:44:38 +0100

On Thu, Nov 28, 2024 at 10:35:10AM +0100, Theo Buehler wrote:
> On Thu, Nov 28, 2024 at 10:01:15AM +0100, Martin Pieuchot wrote:
> > uvm_fault_upper_lookup() enters existing neighbor pages found in the
> > amap.
> > 
> > Diff below reorder the checks in the loop to avoid calling the expensive
> > pmap_extract(9) when there is no corresponding anon or if the page is
> > PG_RELEASED|PG_BUSY and won't be entered anyway.
> > 
> > While here call pmap_update() only if, at least, a page has been entered.
> 
> As far as I can see pmap_update() is always a noop. Is this expected to
> change?

Never mind. I should have grepped for '^pmap_update\>'

> ok tb