[PATCH] test5-pre1 vmfix (rev 8)

[PATCH] test5-pre1 vmfix (rev 8)

[Roger Larsson]

[-- Attachment #1: Type: text/plain, Size: 701 bytes --]

Hi,

Since I am responsible for messing up some aspects of vm
(when fixing others)
here is a patch that tries to solve the introduced problems.

* no more periodic wake up of kswapd - not needed anymore
* no more freeing all zones to (free_pages > pages_high)
* always wakes kswapd up after try_to_free_pages
* kswapd starts when all zones gets zone_wake_kswapd
  (runs once for each zone that hits zone_wake_kswapd)
* removed test for more than pages_high in alloc_pages,
  zones will mostly be in the range [pages_high...pages_low]
* I get 10% better throughput than 2.4.0-test4, YMMV

Note: logic of function keep_kswapd_awake has changed.

/RogerL


--
Home page:
  http://www.norran.net/nra02596/

[-- Attachment #2: patch-2.4.0-test5-1-vmfix.8 --]
[-- Type: text/plain, Size: 3628 bytes --]

--- linux/mm/vmscan.c.orig	Sat Jul 15 23:44:34 2000
+++ linux/mm/vmscan.c	Tue Jul 18 02:08:48 2000
@@ -440,7 +440,7 @@ static inline int memory_pressure(void)
 }
 
 /*
- * Check if there recently has been memory pressure (zone_wake_kswapd)
+ * Check if all zones have recently had memory_pressure (zone_wake_kswapd)
  */
 static inline int keep_kswapd_awake(void)
 {
@@ -451,13 +451,13 @@ static inline int keep_kswapd_awake(void
 		for(i = 0; i < MAX_NR_ZONES; i++) {
 			zone_t *zone = pgdat->node_zones+ i;
 			if (zone->size &&
-			    zone->zone_wake_kswapd)
-				return 1;
+			    !zone->zone_wake_kswapd)
+				return 0;
 		}
 		pgdat = pgdat->node_next;
 	} while (pgdat);
 
-	return 0;
+	return 1;
 }
 
 /*
@@ -496,9 +496,7 @@ static int do_try_to_free_pages(unsigned
 				goto done;
 		}
 
-		/* not (been) low on memory - it is
-		 * pointless to try to swap out.
-		 */
+		/* check if mission completed */
 		if (!keep_kswapd_awake())
 			goto done;
 
@@ -596,10 +594,7 @@ int kswapd(void *unused)
 
 	for (;;) {
 		if (!keep_kswapd_awake()) {
-			/* wake up regulary to do an early attempt too free
-			 * pages - pages will not actually be freed.
-			 */
-			interruptible_sleep_on_timeout(&kswapd_wait, HZ);
+			interruptible_sleep_on(&kswapd_wait);
 		}
 
 		do_try_to_free_pages(GFP_KSWAPD);
@@ -631,18 +626,18 @@ int try_to_free_pages(unsigned int gfp_m
 		retval = do_try_to_free_pages(gfp_mask);
 		current->flags &= ~PF_MEMALLOC;
 	}
-	else {
-		/* make sure kswapd runs */
-		if (waitqueue_active(&kswapd_wait))
-			wake_up_interruptible(&kswapd_wait);
-	}
+
+	/* someone needed memory that kswapd had not provided
+	 * make sure kswapd runs, should not happen often */
+	if (waitqueue_active(&kswapd_wait))
+		wake_up_interruptible(&kswapd_wait);
 
 	return retval;
 }
 
 static int __init kswapd_init(void)
 {
-	printk("Starting kswapd v1.6\n");
+	printk("Starting kswapd v1.7\n");
 	swap_setup();
 	kernel_thread(kswapd, NULL, CLONE_FS | CLONE_FILES | CLONE_SIGHAND);
 	return 0;
--- linux/mm/page_alloc.c.orig	Sat Jul 15 23:44:46 2000
+++ linux/mm/page_alloc.c	Tue Jul 18 02:19:30 2000
@@ -217,7 +217,7 @@ static struct page * rmqueue(zone_t *zon
  */
 struct page * __alloc_pages(zonelist_t *zonelist, unsigned long order)
 {
-	zone_t **zone = zonelist->zones;
+	zone_t **zone;
 	extern wait_queue_head_t kswapd_wait;
 
 	/*
@@ -228,21 +228,6 @@ struct page * __alloc_pages(zonelist_t *
 	 * in a higher zone fails.
 	 */
 
-	for (;;) {
-		zone_t *z = *(zone++);
-		if (!z)
-			break;
-		if (!z->size)
-			BUG();
-
-		/* If there are zones with a lot of free memory allocate from them */
-		if (z->free_pages > z->pages_high) {
-			struct page *page = rmqueue(z, order);
-			if (page)
-				return page;
-		}
-	}
-
 	zone = zonelist->zones;
 	for (;;) {
 		zone_t *z = *(zone++);
@@ -263,6 +248,21 @@ struct page * __alloc_pages(zonelist_t *
 				return page;
 		}
 	}
+
+	/* Three possibilities to get here
+	 * - Previous alloc_pages resulted in last zone set to have
+	 *   zone_wake_kswapd and start it. kswapd has not been able
+	 *   to release enough pages so that one zone does not have
+	 *   zone_wake_kswapd set.
+	 * - Different sets of zones (zonelist)
+	 *   previous did not have all zones with zone_wake_kswapd but
+	 *   this one has... should kswapd be woken up? it will run once.
+	 * - SMP race, kswapd went to sleep slightly after it as running
+	 *   in 'if (waitqueue_active(...))' above.
+	 * + anyway the test is very cheap to do...
+	 */
+	if (waitqueue_active(&kswapd_wait))
+		wake_up_interruptible(&kswapd_wait);
 
 	/*
 	 * Ok, we don't have any zones that don't need some

Re: [PATCH] test5-pre1 vmfix (rev 8)

[Mike Galbraith]

On Tue, 18 Jul 2000, Roger Larsson wrote:

> Hi,

Howdy,

(since I tested/commented on last)

> * I get 10% better throughput than 2.4.0-test4, YMMV

Seeing about the same improvement here.. lightly tested.

	Cheers,

	-Mike

--
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.eu.org/Linux-MM/

Re: [PATCH] test5-pre1 vmfix (rev 8)

[Roger Larsson]

Hi again,

Correction:

I had forgot to start a background task during testing:
* Throuput gets up to 2% to 10% better for my tests.

/RogerL

Roger Larsson wrote:
> 
> Hi,
> 
> Since I am responsible for messing up some aspects of vm
> (when fixing others)
> here is a patch that tries to solve the introduced problems.
> 
> * no more periodic wake up of kswapd - not needed anymore
> * no more freeing all zones to (free_pages > pages_high)
> * always wakes kswapd up after try_to_free_pages
> * kswapd starts when all zones gets zone_wake_kswapd
>   (runs once for each zone that hits zone_wake_kswapd)
> * removed test for more than pages_high in alloc_pages,
>   zones will mostly be in the range [pages_high...pages_low]
> * I get 10% better throughput than 2.4.0-test4, YMMV
> 
> Note: logic of function keep_kswapd_awake has changed.
> 
> /RogerL
> 
> --
> Home page:
>   http://www.norran.net/nra02596/
> 
>   ------------------------------------------------------------------------
>                                   Name: patch-2.4.0-test5-1-vmfix.8
>    patch-2.4.0-test5-1-vmfix.8    Type: Plain Text (text/plain)
>                               Encoding: 7bit

--
Home page:
  http://www.norran.net/nra02596/
--
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.eu.org/Linux-MM/

Re: [PATCH] test5-pre1 vmfix (rev 8)

[Rik van Riel]

On Tue, 18 Jul 2000, Mike Galbraith wrote:
> On Tue, 18 Jul 2000, Roger Larsson wrote:
> 
> > * I get 10% better throughput than 2.4.0-test4, YMMV
> 
> Seeing about the same improvement here.. lightly tested.

Part of the code looks bogus though.

Linus, would it be possible to have a rule in place
that nobody submits VM code unless they accompany it
with an explanation of _why_ their code is supposed
to work?

There seem to be too many "random tweaks" in the current
VM code where nobody has an explanation of why the code works and
what exactly it is supposed to do...

regards,

Rik
--
"What you're running that piece of shit Gnome?!?!"
       -- Miguel de Icaza, UKUUG 2000

http://www.conectiva.com/		http://www.surriel.com/

--
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.eu.org/Linux-MM/