Re: [PATCH] total_free_shortage() using zone_free_shortage()

[Marcelo Tosatti]

On Wed, 8 Aug 2001, Linus Torvalds wrote:

> 
> On Wed, 8 Aug 2001, Marcelo Tosatti wrote:
> >
> > Having "zone->pages_low" as the low water mark to when start kswapd does
> > _not_ mean we want "zone->pages_low" as the freetarget (or the "free
> > shortage" indicator).
> 
> Right. We want the free target to obviously be larger than the low target,
> to avoid hysteresis around it.  And at the same time the free target
> obviously has to be smaller than the more-than-enough "plenty" case.
> 
> So we actually have _three_ numbers, not two.
> 
> > Could you be more verbose ?
> 
> I already was, our mails crossed. I think the main thing I have decided to
> be a "Good Feature (tm)" is to consider the free target to be more of a
> global thing, and not a per-zone thing. While the low water mark and
> "plenty" mark obviously have to be per-zone decisions.
> 
> And we've actually done exactly this - this is how "inactive_target" works
> already, and how "global_free_shortage()" ends up working too. We just
> didn't write it down explicitly.

Your change in pre7 makes zone_inactive_shortage() not return the shortage
size, which is needed by refill_inactive().

--- linux.orig/mm/vmscan.c	Wed Aug  8 17:55:26 2001
+++ linux/mm/vmscan.c	Wed Aug  8 18:04:18 2001
@@ -54,16 +54,19 @@
 
 static unsigned int zone_inactive_shortage(zone_t *zone) 
 {
-	unsigned int inactive;
+	unsigned int sum;
 
 	if (!zone->size)
 		return 0;
 
-	inactive = zone->inactive_dirty_pages;
-	inactive += zone->inactive_clean_pages;
-	inactive += zone->free_pages;
-
-	return inactive < zone->pages_high;
+	sum = zone->pages_high;
+	sum -= zone->inactive_dirty_pages;
+	sum -= zone->inactive_clean_pages;
+	sum -= zone->free_pages;
+	
+	if (sum > 0)
+		return sum;
+	return 0;
 }
 
 static unsigned int zone_free_plenty(zone_t *zone)