Re: [RFC zsmalloc 2/4] zsmalloc: squeeze inuse into page->mapping

From: Sergey Senozhatsky
Date: Mon Aug 10 2015 - 04:53:29 EST


On (08/10/15 16:12), Minchan Kim wrote:
[..]
> diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
> index 491491a..75fefba 100644
> --- a/mm/zsmalloc.c
> +++ b/mm/zsmalloc.c
> @@ -35,7 +35,7 @@
> * metadata.
> * page->lru: links together first pages of various zspages.
> * Basically forming list of zspages in a fullness group.
> - * page->mapping: class index and fullness group of the zspage
> + * page->mapping: override by struct zs_meta
> *
> * Usage of struct page flags:
> * PG_private: identifies the first component page
> @@ -132,6 +132,14 @@
> /* each chunk includes extra space to keep handle */
> #define ZS_MAX_ALLOC_SIZE PAGE_SIZE
>
> +#define CLASS_IDX_BITS 8
> +#define CLASS_IDX_MASK ((1 << CLASS_IDX_BITS) - 1)
> +#define FULLNESS_BITS 2
> +#define FULLNESS_MASK ((1 << FULLNESS_BITS) - 1)
> +#define INUSE_BITS 11
> +#define INUSE_MASK ((1 << INUSE_BITS) - 1)
> +#define ETC_BITS ((sizeof(unsigned long) * 8) - CLASS_IDX_BITS \
> + - FULLNESS_BITS - INUSE_BITS)
> /*
> * On systems with 4K page size, this gives 255 size classes! There is a
> * trader-off here:
> @@ -145,16 +153,15 @@
> * ZS_MIN_ALLOC_SIZE and ZS_SIZE_CLASS_DELTA must be multiple of ZS_ALIGN
> * (reason above)
> */
> -#define ZS_SIZE_CLASS_DELTA (PAGE_SIZE >> 8)
> +#define ZS_SIZE_CLASS_DELTA (PAGE_SIZE >> CLASS_IDX_BITS)
>
> /*
> * We do not maintain any list for completely empty or full pages
> + * Don't reorder.
> */
> enum fullness_group {
> - ZS_ALMOST_FULL,
> + ZS_ALMOST_FULL = 0,
> ZS_ALMOST_EMPTY,
> - _ZS_NR_FULLNESS_GROUPS,
> -
> ZS_EMPTY,
> ZS_FULL
> };
> @@ -198,7 +205,7 @@ static const int fullness_threshold_frac = 4;
>
> struct size_class {
> spinlock_t lock;
> - struct page *fullness_list[_ZS_NR_FULLNESS_GROUPS];
> + struct page *fullness_list[ZS_EMPTY];
> /*
> * Size of objects stored in this class. Must be multiple
> * of ZS_ALIGN.
> @@ -259,13 +266,16 @@ struct zs_pool {
> };
>
> /*
> - * A zspage's class index and fullness group
> - * are encoded in its (first)page->mapping
> + * In this implementation, a zspage's class index, fullness group,
> + * inuse object count are encoded in its (first)page->mapping
> + * sizeof(struct zs_meta) should be equal to sizeof(unsigned long).
> */
> -#define CLASS_IDX_BITS 28
> -#define FULLNESS_BITS 4
> -#define CLASS_IDX_MASK ((1 << CLASS_IDX_BITS) - 1)
> -#define FULLNESS_MASK ((1 << FULLNESS_BITS) - 1)
> +struct zs_meta {
> + unsigned long class_idx:CLASS_IDX_BITS;
> + unsigned long fullness:FULLNESS_BITS;
> + unsigned long inuse:INUSE_BITS;
> + unsigned long etc:ETC_BITS;
> +};
>
> struct mapping_area {
> #ifdef CONFIG_PGTABLE_MAPPING
> @@ -403,26 +413,51 @@ static int is_last_page(struct page *page)
> return PagePrivate2(page);
> }
>
> +static int get_inuse_obj(struct page *page)
> +{
> + struct zs_meta *m;
> +
> + BUG_ON(!is_first_page(page));
> +
> + m = (struct zs_meta *)&page->mapping;
> +
> + return m->inuse;
> +}

a side note. I think there are too many BUG_ON in zsmalloc. some of
them are clearly unnecessary, just because same checks are performed
several lines above. for example:

get_fullness_group()
BUG_ON(!is_first_page(page))
get_inuse_obj()
BUG_ON(!is_first_page(page))


fix_fullness_group()
BUG_ON(!is_first_page(page))
get_zspage_mapping()
BUG_ON(!is_first_page(page))


and so on.


> +static void set_inuse_obj(struct page *page, int inc)
> +{
> + struct zs_meta *m;
> +
> + BUG_ON(!is_first_page(page));
> +
> + m = (struct zs_meta *)&page->mapping;
> + m->inuse += inc;
> +}
> +

{get,set}_zspage_inuse() ?


[..]

> + BUILD_BUG_ON(sizeof(unsigned long) * 8 < (CLASS_IDX_BITS + \
> + FULLNESS_BITS + INUSE_BITS + ETC_BITS));

this build_bug is overwritten in the next patch?

-ss
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/