[v2] mm/min_free_kbytes: modify min_free_kbytes calculation rules

[PATCH v2] mm/min_free_kbytes: modify min_free_kbytes calculation rules

Posted by liuq 2 years, 7 months ago

The current calculation of min_free_kbytes only uses ZONE_DMA and
ZONE_NORMAL pages,but the ZONE_MOVABLE zone->_watermark[WMARK_MIN]
will also divide part of min_free_kbytes.This will cause the min
watermark of ZONE_NORMAL to be too small in the presence of ZONE_MOVEABLE.

__GFP_HIGH and PF_MEMALLOC allocations usually don't need moveable
zone pages, so just like ZONE_HIGHMEM, cap pages_min to a small
value in __setup_per_zone_wmarks.

Signed-off-by: liuq <liuq131@chinatelecom.cn>
---
 mm/page_alloc.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 47421bedc12b..608384712a89 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -6364,7 +6364,7 @@ static void __setup_per_zone_wmarks(void)
 
 	/* Calculate total number of !ZONE_HIGHMEM pages */
 	for_each_zone(zone) {
-		if (!is_highmem(zone))
+		if (!is_highmem(zone) || zone_idx(zone) != ZONE_MOVABLE)
 			lowmem_pages += zone_managed_pages(zone);
 	}
 
@@ -6374,7 +6374,7 @@ static void __setup_per_zone_wmarks(void)
 		spin_lock_irqsave(&zone->lock, flags);
 		tmp = (u64)pages_min * zone_managed_pages(zone);
 		do_div(tmp, lowmem_pages);
-		if (is_highmem(zone)) {
+		if (is_highmem(zone) || zone_idx(zone) == ZONE_MOVABLE) {
 			/*
 			 * __GFP_HIGH and PF_MEMALLOC allocations usually don't
 			 * need highmem pages, so cap pages_min to a small
-- 
2.27.0

Re: [PATCH v2] mm/min_free_kbytes: modify min_free_kbytes calculation rules

Posted by David Hildenbrand 2 years, 7 months ago

On 21.06.23 09:39, liuq wrote:
> The current calculation of min_free_kbytes only uses ZONE_DMA and
> ZONE_NORMAL pages,but the ZONE_MOVABLE zone->_watermark[WMARK_MIN]
> will also divide part of min_free_kbytes.This will cause the min
> watermark of ZONE_NORMAL to be too small in the presence of ZONE_MOVEABLE.
> 
> __GFP_HIGH and PF_MEMALLOC allocations usually don't need moveable
> zone pages, so just like ZONE_HIGHMEM, cap pages_min to a small
> value in __setup_per_zone_wmarks.
> 
> Signed-off-by: liuq <liuq131@chinatelecom.cn>
> ---
>   mm/page_alloc.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 47421bedc12b..608384712a89 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -6364,7 +6364,7 @@ static void __setup_per_zone_wmarks(void)
>   
>   	/* Calculate total number of !ZONE_HIGHMEM pages */
>   	for_each_zone(zone) {
> -		if (!is_highmem(zone))
> +		if (!is_highmem(zone) || zone_idx(zone) != ZONE_MOVABLE)
>   			lowmem_pages += zone_managed_pages(zone);
>   	}
>   
> @@ -6374,7 +6374,7 @@ static void __setup_per_zone_wmarks(void)
>   		spin_lock_irqsave(&zone->lock, flags);
>   		tmp = (u64)pages_min * zone_managed_pages(zone);
>   		do_div(tmp, lowmem_pages);
> -		if (is_highmem(zone)) {
> +		if (is_highmem(zone) || zone_idx(zone) == ZONE_MOVABLE) {
>   			/*
>   			 * __GFP_HIGH and PF_MEMALLOC allocations usually don't
>   			 * need highmem pages, so cap pages_min to a small

Patch subject should probably be something like

"mm/page_alloc: fix min_free_kbytes calculation regarding ZONE_MOVABLE"

?

-- 
Cheers,

David / dhildenb

Re: [PATCH v2] mm/min_free_kbytes: modify min_free_kbytes calculation rules

Posted by Huang, Ying 2 years, 7 months ago

Hi, Liuq,

Thanks for updated patch.

liuq <liuq131@chinatelecom.cn> writes:

> The current calculation of min_free_kbytes only uses ZONE_DMA and
> ZONE_NORMAL pages,but the ZONE_MOVABLE zone->_watermark[WMARK_MIN]
> will also divide part of min_free_kbytes.This will cause the min
> watermark of ZONE_NORMAL to be too small in the presence of ZONE_MOVEABLE.
>
> __GFP_HIGH and PF_MEMALLOC allocations usually don't need moveable
> zone pages, so just like ZONE_HIGHMEM, cap pages_min to a small
> value in __setup_per_zone_wmarks.

Please list the test result of your patch.  That is, WMARK_MIN before
and after your patch in a system with large ZONE_MOVABLE.

> Signed-off-by: liuq <liuq131@chinatelecom.cn>
> ---
>  mm/page_alloc.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 47421bedc12b..608384712a89 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -6364,7 +6364,7 @@ static void __setup_per_zone_wmarks(void)
>  
>  	/* Calculate total number of !ZONE_HIGHMEM pages */
>  	for_each_zone(zone) {
> -		if (!is_highmem(zone))
> +		if (!is_highmem(zone) || zone_idx(zone) != ZONE_MOVABLE)
>  			lowmem_pages += zone_managed_pages(zone);
>  	}
>  
> @@ -6374,7 +6374,7 @@ static void __setup_per_zone_wmarks(void)
>  		spin_lock_irqsave(&zone->lock, flags);
>  		tmp = (u64)pages_min * zone_managed_pages(zone);
>  		do_div(tmp, lowmem_pages);
> -		if (is_highmem(zone)) {
> +		if (is_highmem(zone) || zone_idx(zone) == ZONE_MOVABLE) {
>  			/*
>  			 * __GFP_HIGH and PF_MEMALLOC allocations usually don't
>  			 * need highmem pages, so cap pages_min to a small
                                ~~~~~~~

Change the comments above too?

Best Regards,
Huang, Ying