Re: [Linaro-mm-sig] [PATCHv23 12/16] mm: trigger page reclaim in alloc_contig_range() to stabilise watermarks

[Barry Song <21cnbao@gmail.com>]

2012/2/23 Marek Szyprowski <m.szyprowski@samsung.com>:
> alloc_contig_range() performs memory allocation so it also should keep
> track on keeping the correct level of memory watermarks. This commit adds
> a call to *_slowpath style reclaim to grab enough pages to make sure that
> the final collection of contiguous pages from freelists will not starve
> the system.
>
> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
> CC: Michal Nazarewicz <mina86@mina86.com>
> Tested-by: Rob Clark <rob.clark@linaro.org>
> Tested-by: Ohad Ben-Cohen <ohad@wizery.com>
> Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
> Tested-by: Robert Nelson <robertcnelson@gmail.com>
> ---
>  include/linux/mmzone.h |    9 +++++++
>  mm/page_alloc.c        |   62 ++++++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 71 insertions(+), 0 deletions(-)
>
> diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
> index 4781f30..77db8c0 100644
> --- a/include/linux/mmzone.h
> +++ b/include/linux/mmzone.h
> @@ -63,8 +63,10 @@ enum {
>
>  #ifdef CONFIG_CMA
>  #  define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
> +#  define cma_wmark_pages(zone)        zone->min_cma_pages
>  #else
>  #  define is_migrate_cma(migratetype) false
> +#  define cma_wmark_pages(zone) 0
>  #endif
>
>  #define for_each_migratetype_order(order, type) \
> @@ -371,6 +373,13 @@ struct zone {
>        /* see spanned/present_pages for more description */
>        seqlock_t               span_seqlock;
>  #endif
> +#ifdef CONFIG_CMA
> +       /*
> +        * CMA needs to increase watermark levels during the allocation
> +        * process to make sure that the system is not starved.
> +        */
> +       unsigned long           min_cma_pages;
> +#endif
>        struct free_area        free_area[MAX_ORDER];
>
>  #ifndef CONFIG_SPARSEMEM
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 7a0d286..39cd74f 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -5092,6 +5092,11 @@ static void __setup_per_zone_wmarks(void)
>                                        low + (min >> 2);
>                zone->watermark[WMARK_HIGH] = min_wmark_pages(zone) +
>                                        low + (min >> 1);
> +
> +               zone->watermark[WMARK_MIN] += cma_wmark_pages(zone);
> +               zone->watermark[WMARK_LOW] += cma_wmark_pages(zone);
> +               zone->watermark[WMARK_HIGH] += cma_wmark_pages(zone);
> +
>                setup_zone_migrate_reserve(zone);
>                spin_unlock_irqrestore(&zone->lock, flags);
>        }
> @@ -5695,6 +5700,56 @@ static int __alloc_contig_migrate_range(unsigned long start, unsigned long end)
>        return ret > 0 ? 0 : ret;
>  }
>
> +/*
> + * Update zone's cma pages counter used for watermark level calculation.
> + */
> +static inline void __update_cma_watermarks(struct zone *zone, int count)
> +{
> +       unsigned long flags;
> +       spin_lock_irqsave(&zone->lock, flags);
> +       zone->min_cma_pages += count;
> +       spin_unlock_irqrestore(&zone->lock, flags);
> +       setup_per_zone_wmarks();
> +}
> +
> +/*
> + * Trigger memory pressure bump to reclaim some pages in order to be able to
> + * allocate 'count' pages in single page units. Does similar work as
> + *__alloc_pages_slowpath() function.
> + */
> +static int __reclaim_pages(struct zone *zone, gfp_t gfp_mask, int count)
> +{
> +       enum zone_type high_zoneidx = gfp_zone(gfp_mask);
> +       struct zonelist *zonelist = node_zonelist(0, gfp_mask);
> +       int did_some_progress = 0;
> +       int order = 1;
> +       unsigned long watermark;
> +
> +       /*
> +        * Increase level of watermarks to force kswapd do his job
> +        * to stabilise at new watermark level.
> +        */
> +       __update_cma_watermarks(zone, count);
> +
> +       /* Obey watermarks as if the page was being allocated */
> +       watermark = low_wmark_pages(zone) + count;
> +       while (!zone_watermark_ok(zone, 0, watermark, 0, 0)) {
> +               wake_all_kswapd(order, zonelist, high_zoneidx, zone_idx(zone));
> +
> +               did_some_progress = __perform_reclaim(gfp_mask, order, zonelist,
> +                                                     NULL);
> +               if (!did_some_progress) {
> +                       /* Exhausted what can be done so it's blamo time */
> +                       out_of_memory(zonelist, gfp_mask, order, NULL);

out_of_memory() has got another param in the newest next/master tree,
out_of_memory(zonelist, gfp_mask, order, NULL, false) should be OK.

-barry

--
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/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>