[PATCH NG] alloc_pages_limit & pages_min

[PATCH NG] alloc_pages_limit & pages_min

[Roger Larsson]

Hi,

[Note: not tested yet... but it might be hard to trigger]

After discussions with Riel and some additions...

* The original code had a little bug in those cases when
- kreclaimd were not allowed to run for a LONG time...
Lots of kernel activity, RT tasks, or code running
around for a long time allocating memory.
- there were lots of inactive clean pages
- pages were allocated without direct_reclaim (higher order)
(networking might be one candidate)
it could result in using up ALL free pages!

This patch tries to prevent this situation in several ways:
1) Do not allow to alloc a free page when they are critically low.
   [last line of patch]

2) If direct reclaims are allowed do some additional work.
 reclaim & free until pages_min,
 alloc one page,
 reclaim and free until pages_low

Q) Nothing is done to force execution of kreclaimd, if no process
    that can direct_reclaim allocs a page - what will happen then?
    [unlikely but...]

/RogerL

-- 
Roger Larsson
Skelleftea
Sweden


*******************************************
Patch prepared by: roger.larsson@norran.net with comments from Riel

--- linux/mm/page_alloc.c.orig	Wed Aug 22 13:36:57 2001
+++ linux/mm/page_alloc.c	Thu Aug 23 01:15:17 2001
@@ -253,11 +253,35 @@
 
 		if (z->free_pages + z->inactive_clean_pages >= water_mark) {
 			struct page *page = NULL;
-			/* If possible, reclaim a page directly. */
-			if (direct_reclaim)
+
+			/*
+			 * Reclaim a page from the inactive_clean list.
+			 * If needed, refill the free list up to the
+			 * low water mark.
+			 */
+			if (direct_reclaim) {
 				page = reclaim_page(z);
-			/* If that fails, fall back to rmqueue. */
-			if (!page)
+
+				while (page && z->free_pages < z->pages_min) {
+					__free_page(page);
+					page = reclaim_page(z);
+				}
+
+				if (page) {
+					while (z->free_pages < z->pages_low) {
+						struct page *extra = reclaim_page(z);
+						if (!extra)
+							break;
+						__free_page(extra);
+					}
+				}
+
+				/* let kreclaimd handle up to pages_high */
+			}
+			/* If that fails, fall back to rmqueue, but never let
+			*  free_pages go below pages_min...
+			*/
+			if (!page && z->free_pages >= z->pages_min)
 				page = rmqueue(z, order);
 			if (page)
 				return page;
--
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/

Re: [PATCH NG] alloc_pages_limit & pages_min

[Rik van Riel]

On Thu, 23 Aug 2001, Roger Larsson wrote:

> +				if (page) {
> +					while (z->free_pages < z->pages_low) {
> +						struct page *extra = reclaim_page(z);
> +						if (!extra)
> +							break;
> +						__free_page(extra);
> +					}
> +				}

This is a surprise ;)

Why did you introduce this piece of code?
What is it supposed to achieve ?

Rik
-- 
IA64: a worthy successor to i860.

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

Send all your spam to aardvark@nl.linux.org (spam digging piggy)

--
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/

Re: [PATCH NG] alloc_pages_limit & pages_min

[Roger Larsson]

On Thursdayen den 23 August 2001 02:40, Rik van Riel wrote:
> On Thu, 23 Aug 2001, Roger Larsson wrote:
> > +				if (page) {
> > +					while (z->free_pages < z->pages_low) {
> > +						struct page *extra = reclaim_page(z);
> > +						if (!extra)
> > +							break;
> > +						__free_page(extra);
> > +					}
> > +				}
>
> This is a surprise ;)
>
> Why did you introduce this piece of code?
> What is it supposed to achieve ?
>
f we did get one page => we are above pages_min
try to reach pages_low too.
But holding on to the page we got.

It is possible that it should only be done if we started under
pages_min. But it will be faster the closer we start to pages_low

/RogerL

-- 
Roger Larsson
Skelleftea
Sweden
--
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/

Re: [PATCH NG] alloc_pages_limit & pages_min

[Rik van Riel]

On Thu, 23 Aug 2001, Roger Larsson wrote:

> f we did get one page => we are above pages_min
> try to reach pages_low too.

Yeah, but WHY ?

Rik
--
IA64: a worthy successor to the i860.

		http://www.surriel.com/
http://www.conectiva.com/	http://distro.conectiva.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-mm.org/

Re: [PATCH NG] alloc_pages_limit & pages_min

[Roger Larsson]

On Thursdayen den 23 August 2001 20:44, Rik van Riel wrote:
> On Thu, 23 Aug 2001, Roger Larsson wrote:
> > f we did get one page => we are above pages_min
> > try to reach pages_low too.
>
> Yeah, but WHY ?
>

* Historic reasons - I feel good at that limit... :-)
 MIN the limit never crossed
 LOW center, our target of free pages - when all zones time to free.
 HIGH limit were to stop the freeing.

-- 
Roger Larsson
Skelleftea
Sweden
--
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/

Re: [PATCH NG] alloc_pages_limit & pages_min

[Rik van Riel]

On Thu, 23 Aug 2001, Roger Larsson wrote:
> On Thursdayen den 23 August 2001 20:44, Rik van Riel wrote:
> > On Thu, 23 Aug 2001, Roger Larsson wrote:
> > > f we did get one page => we are above pages_min
> > > try to reach pages_low too.
> >
> > Yeah, but WHY ?
>
> * Historic reasons - I feel good at that limit... :-)
>  MIN the limit never crossed

Never crossed by (free + clean) pages. I see no reason why
we couldn't leave the free-only target at this limit...

>  LOW center, our target of free pages - when all zones time to free.

Meaning we'll usually not have any clean pages around but
only free pages if your patch gets applied ;)

>  HIGH limit were to stop the freeing.

Rik
--
IA64: a worthy successor to the i860.

		http://www.surriel.com/
http://www.conectiva.com/	http://distro.conectiva.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-mm.org/

Upd: [PATCH NG] alloc_pages_limit & pages_min

[Roger Larsson]

Riel convinced be to back off a part of the patch.
Here comes an updated one.

-- 
Roger Larsson
Skelleftea
Sweden

*******************************************
Patch prepared by: roger.larsson@norran.net
Name of file: linux-2.4.8-pre3-pages_min-R3

--- linux/mm/page_alloc.c.orig  Thu Aug 23 19:58:55 2001
+++ linux/mm/page_alloc.c       Thu Aug 23 21:19:20 2001
@@ -253,11 +253,26 @@

                if (z->free_pages + z->inactive_clean_pages >= water_mark) {
                        struct page *page = NULL;
-                       /* If possible, reclaim a page directly. */
-                       if (direct_reclaim)
+
+                       /*
+                        * Reclaim a page from the inactive_clean list.
+                        * If needed, refill the free list up to the
+                        * low water mark.
+                        */
+                       if (direct_reclaim) {
                                page = reclaim_page(z);
-                       /* If that fails, fall back to rmqueue. */
-                       if (!page)
+
+                               while (page && z->free_pages < z->pages_min) {
+                                       __free_page(page);
+                                       page = reclaim_page(z);
+                               }
+
+                               /* let kreclaimd handle up to pages_high */
+                       }
+                       /* If that fails, fall back to rmqueue, but never let
+                       *  free_pages go below pages_min...
+                       */
+                       if (!page && z->free_pages >= z->pages_min)
                                page = rmqueue(z, order);
                        if (page)
                                return page;
--
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/

Re: [PATCH NG] alloc_pages_limit & pages_min

[Roger Larsson]

On Thursdayen den 23 August 2001 02:40, Rik van Riel wrote:
> On Thu, 23 Aug 2001, Roger Larsson wrote:
> > +				if (page) {
> > +					while (z->free_pages < z->pages_low) {
> > +						struct page *extra = reclaim_page(z);
> > +						if (!extra)
> > +							break;
> > +						__free_page(extra);
> > +					}
> > +				}
>
> This is a surprise ;)
>
> Why did you introduce this piece of code?
> What is it supposed to achieve ?
>

A lighter alternative would be to reclaim just one extra page...
Then it will move in the right direction but not more, quite
nice actually!

/RogerL

-- 
Roger Larsson
Skelleftea
Sweden
--
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/

Re: [PATCH NG] alloc_pages_limit & pages_min

[Rik van Riel]

On Thu, 23 Aug 2001, Roger Larsson wrote:

> > Why did you introduce this piece of code?
> > What is it supposed to achieve ?
>
> A lighter alternative would be to reclaim just one extra page...
> Then it will move in the right direction but not more, quite
> nice actually!

Why ?

Or, to be more specific, why would we want to throw away
data from the cache all the way up to pages_min when we
know we're running a workload with allocations which can
eat directly from the cache ?

Rik
--
IA64: a worthy successor to the i860.

		http://www.surriel.com/
http://www.conectiva.com/	http://distro.conectiva.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-mm.org/