Skip reclaim_mapped determination if we do not swap

Skip reclaim_mapped determination if we do not swap

[Christoph Lameter]

This puts the variables and the way to get to reclaim_mapped in one 
block. And allows zone_reclaim or other things to skip the determination 
(maybe this whole block of code does not belong into 
refill_inactive_zone()?)

Signed-off-by: Christoph Lameter <clameter@sgi.com>

Index: linux-2.6.16-rc2/mm/vmscan.c
===================================================================
--- linux-2.6.16-rc2.orig/mm/vmscan.c	2006-02-11 13:29:51.000000000 -0800
+++ linux-2.6.16-rc2/mm/vmscan.c	2006-02-11 13:31:15.000000000 -0800
@@ -1180,9 +1180,43 @@ refill_inactive_zone(struct zone *zone, 
 	struct page *page;
 	struct pagevec pvec;
 	int reclaim_mapped = 0;
-	long mapped_ratio;
-	long distress;
-	long swap_tendency;
+
+	if (sc->may_swap) {
+		long mapped_ratio;
+		long distress;
+		long swap_tendency;
+
+		/*
+		 * `distress' is a measure of how much trouble we're having reclaiming
+		 * pages.  0 -> no problems.  100 -> great trouble.
+		 */
+		distress = 100 >> zone->prev_priority;
+
+		/*
+		 * The point of this algorithm is to decide when to start reclaiming
+		 * mapped memory instead of just pagecache.  Work out how much memory
+		 * is mapped.
+		 */
+		mapped_ratio = (sc->nr_mapped * 100) / total_memory;
+
+		/*
+		 * Now decide how much we really want to unmap some pages.  The mapped
+		 * ratio is downgraded - just because there's a lot of mapped memory
+		 * doesn't necessarily mean that page reclaim isn't succeeding.
+		 *
+		 * The distress ratio is important - we don't want to start going oom.
+		 *
+		 * A 100% value of vm_swappiness overrides this algorithm altogether.
+		 */
+		swap_tendency = mapped_ratio / 2 + distress + vm_swappiness;
+
+		/*
+		 * Now use this metric to decide whether to start moving mapped memory
+		 * onto the inactive list.
+		 */
+		if (swap_tendency >= 100)
+			reclaim_mapped = 1;
+	}
 
 	lru_add_drain();
 	spin_lock_irq(&zone->lru_lock);
@@ -1192,37 +1226,6 @@ refill_inactive_zone(struct zone *zone, 
 	zone->nr_active -= pgmoved;
 	spin_unlock_irq(&zone->lru_lock);
 
-	/*
-	 * `distress' is a measure of how much trouble we're having reclaiming
-	 * pages.  0 -> no problems.  100 -> great trouble.
-	 */
-	distress = 100 >> zone->prev_priority;
-
-	/*
-	 * The point of this algorithm is to decide when to start reclaiming
-	 * mapped memory instead of just pagecache.  Work out how much memory
-	 * is mapped.
-	 */
-	mapped_ratio = (sc->nr_mapped * 100) / total_memory;
-
-	/*
-	 * Now decide how much we really want to unmap some pages.  The mapped
-	 * ratio is downgraded - just because there's a lot of mapped memory
-	 * doesn't necessarily mean that page reclaim isn't succeeding.
-	 *
-	 * The distress ratio is important - we don't want to start going oom.
-	 *
-	 * A 100% value of vm_swappiness overrides this algorithm altogether.
-	 */
-	swap_tendency = mapped_ratio / 2 + distress + vm_swappiness;
-
-	/*
-	 * Now use this metric to decide whether to start moving mapped memory
-	 * onto the inactive list.
-	 */
-	if (swap_tendency >= 100 && sc->may_swap)
-		reclaim_mapped = 1;
-
 	while (!list_empty(&l_hold)) {
 		cond_resched();
 		page = lru_to_page(&l_hold);

--
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>

Re: Skip reclaim_mapped determination if we do not swap

[Andrew Morton]

Christoph Lameter <clameter@engr.sgi.com> wrote:
>
> This puts the variables and the way to get to reclaim_mapped in one 
> block. And allows zone_reclaim or other things to skip the determination 
> (maybe this whole block of code does not belong into 
> refill_inactive_zone()?)

For your enjoyment, here is a picture of what the resulting code looks like
in an 80-col window:

	http://www.zip.com.au/~akpm/linux/patches/stuff/x.jpeg

It would make things somewhat easier if I didn't have to go fixing up after
you all the time.

--
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>

Re: Skip reclaim_mapped determination if we do not swap

[Christoph Lameter]

On Sat, 11 Feb 2006, Andrew Morton wrote:

> For your enjoyment, here is a picture of what the resulting code looks like
> in an 80-col window:
> 
> 	http://www.zip.com.au/~akpm/linux/patches/stuff/x.jpeg

Ahh. Yes looks really ragged...
 
> It would make things somewhat easier if I didn't have to go fixing up after
> you all the time.

I can fix this if this is the final resting place of the code. But should 
this piece of code really be there? Doesnt it belong ito shrink_zone() or 
even in try_to_free_pages() or balance_pgdat(). Shouldn't we pass 
reclaim_mapped as a parameter to refill_inactive_zone?

--
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>

Re: Skip reclaim_mapped determination if we do not swap

[Christoph Lameter]

Here is a new rev of the earlier patch that moves the determination of
reclaim_mapped into shrink_zone(). This means that refill_inactive does 
not depend on scan control anymore. And its properly formatted for 80 
columns

Signed-off-by: Christoph Lameter <clameter@sgi.com>

Index: linux-2.6.16-rc2/mm/vmscan.c
===================================================================
--- linux-2.6.16-rc2.orig/mm/vmscan.c	2006-02-11 13:35:20.000000000 -0800
+++ linux-2.6.16-rc2/mm/vmscan.c	2006-02-11 14:22:07.000000000 -0800
@@ -1168,21 +1168,16 @@ done:
  * But we had to alter page->flags anyway.
  */
 static void
-refill_inactive_zone(struct zone *zone, struct scan_control *sc)
+refill_inactive_zone(struct zone *zone, unsigned long nr_pages, int reclaim_mapped)
 {
 	int pgmoved;
 	int pgdeactivate = 0;
 	int pgscanned;
-	int nr_pages = sc->nr_to_scan;
 	LIST_HEAD(l_hold);	/* The pages which were snipped off */
 	LIST_HEAD(l_inactive);	/* Pages to go onto the inactive_list */
 	LIST_HEAD(l_active);	/* Pages to go onto the active_list */
 	struct page *page;
 	struct pagevec pvec;
-	int reclaim_mapped = 0;
-	long mapped_ratio;
-	long distress;
-	long swap_tendency;
 
 	lru_add_drain();
 	spin_lock_irq(&zone->lru_lock);
@@ -1192,37 +1187,6 @@ refill_inactive_zone(struct zone *zone, 
 	zone->nr_active -= pgmoved;
 	spin_unlock_irq(&zone->lru_lock);
 
-	/*
-	 * `distress' is a measure of how much trouble we're having reclaiming
-	 * pages.  0 -> no problems.  100 -> great trouble.
-	 */
-	distress = 100 >> zone->prev_priority;
-
-	/*
-	 * The point of this algorithm is to decide when to start reclaiming
-	 * mapped memory instead of just pagecache.  Work out how much memory
-	 * is mapped.
-	 */
-	mapped_ratio = (sc->nr_mapped * 100) / total_memory;
-
-	/*
-	 * Now decide how much we really want to unmap some pages.  The mapped
-	 * ratio is downgraded - just because there's a lot of mapped memory
-	 * doesn't necessarily mean that page reclaim isn't succeeding.
-	 *
-	 * The distress ratio is important - we don't want to start going oom.
-	 *
-	 * A 100% value of vm_swappiness overrides this algorithm altogether.
-	 */
-	swap_tendency = mapped_ratio / 2 + distress + vm_swappiness;
-
-	/*
-	 * Now use this metric to decide whether to start moving mapped memory
-	 * onto the inactive list.
-	 */
-	if (swap_tendency >= 100 && sc->may_swap)
-		reclaim_mapped = 1;
-
 	while (!list_empty(&l_hold)) {
 		cond_resched();
 		page = lru_to_page(&l_hold);
@@ -1304,6 +1268,7 @@ shrink_zone(struct zone *zone, struct sc
 {
 	unsigned long nr_active;
 	unsigned long nr_inactive;
+	int reclaim_mapped = 0;
 
 	atomic_inc(&zone->reclaim_in_progress);
 
@@ -1325,12 +1290,52 @@ shrink_zone(struct zone *zone, struct sc
 	else
 		nr_inactive = 0;
 
+	if (sc->may_swap) {
+		long mapped_ratio;
+		long distress;
+		long swap_tendency;
+
+		/*
+		 * `distress' is a measure of how much trouble we're having
+		 * reclaiming  pages.  0 -> no problems.  100 -> great trouble.
+		 */
+		distress = 100 >> zone->prev_priority;
+
+		/*
+		 * The point of this algorithm is to decide when to start
+		 * reclaiming mapped memory instead of just pagecache.
+		 * Work out how much memory is mapped.
+		 */
+		mapped_ratio = (sc->nr_mapped * 100) / total_memory;
+
+		/*
+		 * Now decide how much we really want to unmap some pages.
+		 * The mappe ratio is downgraded - just because there's a lot
+		 * of mapped memory doesn't necessarily mean that page reclaim
+		 * isn't succeeding.
+		 *
+		 * The distress ratio is important - we don't want to start
+		 * going oom.
+		 *
+		 * A 100% value of vm_swappiness overrides this algorithm
+		 * altogether.
+		 */
+		swap_tendency = mapped_ratio / 2 + distress + vm_swappiness;
+
+		/*
+		 * Now use this metric to decide whether to start moving
+		 * mapped memory onto the inactive list.
+		 */
+		if (swap_tendency >= 100)
+			reclaim_mapped = 1;
+	}
+
 	while (nr_active || nr_inactive) {
 		if (nr_active) {
 			sc->nr_to_scan = min(nr_active,
 					(unsigned long)sc->swap_cluster_max);
 			nr_active -= sc->nr_to_scan;
-			refill_inactive_zone(zone, sc);
+			refill_inactive_zone(zone, sc->nr_to_scan, reclaim_mapped);
 		}
 
 		if (nr_inactive) {

--
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>

Re: Skip reclaim_mapped determination if we do not swap

[Andrew Morton]

Christoph Lameter <clameter@engr.sgi.com> wrote:
>
> But should 
>  this piece of code really be there? Doesnt it belong ito shrink_zone() or 
>  even in try_to_free_pages() or balance_pgdat(). Shouldn't we pass 
>  reclaim_mapped as a parameter to refill_inactive_zone?

Well, it's only used in refill_inactive_zone().  Does it matter much where
it is?

--
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>

Re: Skip reclaim_mapped determination if we do not swap

[Christoph Lameter]

On Sat, 11 Feb 2006, Andrew Morton wrote:

> Christoph Lameter <clameter@engr.sgi.com> wrote:
> >
> > But should 
> >  this piece of code really be there? Doesnt it belong ito shrink_zone() or 
> >  even in try_to_free_pages() or balance_pgdat(). Shouldn't we pass 
> >  reclaim_mapped as a parameter to refill_inactive_zone?
> 
> Well, it's only used in refill_inactive_zone().  Does it matter much where
> it is?

If it is repeatedly evaluated when reclaiming then it matters. It also 
makes refill_inactive_zone() simpler if the statistical considerations are 
in one place in shrink_zone().

--
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>

pluggable reclaim infrastructure

[Marcelo Tosatti]

Hi,

On Sat, Feb 11, 2006 at 02:25:57PM -0800, Christoph Lameter wrote:
> Here is a new rev of the earlier patch that moves the determination of
> reclaim_mapped into shrink_zone(). This means that refill_inactive does 
> not depend on scan control anymore. And its properly formatted for 80 
> columns

Can we please hold off this reclaim tweaks for a while? Peter's
pluggable page-replace infrastructure does more fundamental and
important changes to the reclaim code (along with better organization
and separation of the heuristics such as reclaim_mapped, etc.), and
having vmscan.c to change often (as it is now) is just a pain in the
arse to handle.

Its not going to take long for the patchset to be sent for
review/shaping.

--
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>