Re: [PATCHv2] mm: optimization on page allocation when CMA enabled

[Andrew Morton]

On Thu, 4 May 2023 18:09:54 +0800 "zhaoyang.huang" <zhaoyang.huang@xxxxxxxxxx> wrote:

> From: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
> 
> Let us look at the series of scenarios below with WMARK_LOW=25MB,WMARK_MIN=5MB
> (managed pages 1.9GB). We can know that current 'fixed 1/2 ratio' start to use
> CMA since C which actually has caused U&R lower than WMARK_LOW (this should be
> deemed as against current memory policy, that is, U&R should either stay around
> WATERMARK_LOW when no allocation or do reclaim via enter slowpath)
> 
> free_cma/free_pages(MB)      A(12/30)     B(12/25)     C(12/20)
> fixed 1/2 ratio                 N             N           Y
> this commit                     Y             Y           Y

A few style issues.

> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -3071,6 +3071,34 @@ static bool unreserve_highatomic_pageblock(const struct alloc_context *ac,
>  
>  }
>  
> +#ifdef CONFIG_CMA
> +static bool __if_use_cma_first(struct zone *zone, unsigned int order, unsigned int alloc_flags)

This function would benefit from a nice covering comment.  Explain what
it does and, especially, why it does it.

I'm not sure that "__if_use_cma_first" is a good name - I'd need to see
that explanation to decide.

> +{
> +	unsigned long cma_proportion = 0;
> +	unsigned long cma_free_proportion = 0;
> +	unsigned long watermark = 0;
> +	long count = 0;
> +	bool cma_first = false;

We seems to have some unnecessary initializations here.

> +	watermark = wmark_pages(zone, alloc_flags & ALLOC_WMARK_MASK);
> +	/*check if GFP_MOVABLE pass previous watermark check via the help of CMA*/

Space after /* and before */

	/*
	 * Check if GFP_MOVABLE passed the previous watermark check with the
	 * help of CMA
	 */

> +	if (!zone_watermark_ok(zone, order, watermark, 0, alloc_flags & (~ALLOC_CMA)))
> +		/* WMARK_LOW failed lead to using cma first, this helps U&R stay
> +		 * around low when being drained by GFP_MOVABLE
> +		 */

Unusual layout, text is hard to understand.  Maybe something like

		/*
		 * WMARK_LOW failed, leading to the use cma first.  This helps
		 * U&R stay low when <something> is being drained by
		 * GFP_MOVABLE
		 */

Also, please expand "U&R" into full words.  I don't recognize that
abbreviation.

> +		cma_first = true;
> +	else {
> +		/*check proportion when zone_watermark_ok*/

		/* check ... _ok */

Comments should seek to explain *why* a thing is being done, rather than
*what* is being done.

> +		count = atomic_long_read(&zone->managed_pages);
> +		cma_proportion = zone->cma_pages * 100 / count;
> +		cma_free_proportion = zone_page_state(zone, NR_FREE_CMA_PAGES) * 100
> +			/  zone_page_state(zone, NR_FREE_PAGES);
> +		cma_first = (cma_free_proportion >= cma_proportion * 2
> +				|| cma_free_proportion >= 50);
> +	}
> +	return cma_first;
> +}
> +#endif
>  /*
>   * Do the hard work of removing an element from the buddy allocator.
>   * Call me with the zone->lock already held.
> @@ -3087,10 +3115,10 @@ static bool unreserve_highatomic_pageblock(const struct alloc_context *ac,

I wonder why git decided this hunk is unreserve_highatomic_pageblock().
It's actually in __rmqueue().

>  		 * allocating from CMA when over half of the zone's free memory
>  		 * is in the CMA area.
>  		 */
> -		if (alloc_flags & ALLOC_CMA &&
> -		    zone_page_state(zone, NR_FREE_CMA_PAGES) >
> -		    zone_page_state(zone, NR_FREE_PAGES) / 2) {
> -			page = __rmqueue_cma_fallback(zone, order);
> +		if (migratetype == MIGRATE_MOVABLE) {
> +			bool cma_first = __if_use_cma_first(zone, order, alloc_flags);
> +
> +			page = cma_first ? __rmqueue_cma_fallback(zone, order) : NULL;

Code could be tidier and could avoid needless 80-column wordwrapping
and an unneeded local.

			page = NULL;
			if (__if_use_cma_first(zone, order, alloc_flags))
				page = __rmqueue_cma_fallback(zone, order);

>  			if (page)
>  				return page;
>  		}

Anyway, please take a look, fix the build error and send us a v3.  I
suggest you cc Minchan Kim, who might review it for us.