The patch titled
Subject: mm/zsmalloc.c: count in handle's size when calculating pages_per_zspage
has been added to the -mm tree. Its filename is
mm-zsmallocc-count-in-handles-size-when-calculating-pages_per_zspage.patch
This patch should soon appear at
http://ozlabs.org/~akpm/mmots/broken-out/mm-zsmallocc-count-in-handles-size-when-calculating-pages_per_zspage.patch
and later at
http://ozlabs.org/~akpm/mmotm/broken-out/mm-zsmallocc-count-in-handles-size-when-calculating-pages_per_zspage.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/SubmitChecklist when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Yinghao Xie <yinghao.xie@xxxxxxxxxxx>
Subject: mm/zsmalloc.c: count in handle's size when calculating pages_per_zspage
1. Fix wastage calculation.
2. Indirect handle introduced extra ZS_HANDLE_SIZE size for each
object,it's transparent for upper function, but a size_class's total
objects will change.
take the 43rd class which class_size = 32 + 43 * 16 = 720 as example:
4096 * 1 % 720 = 496
4096 * 2 % 720 = 272
4096 * 3 % 720 = 48
4096 * 4 %720 = 544
after handle introduced,class_size + ZS_HANDLE_SIZE (4 on 32bit) = 724
4096 * 1 % 724 = 476
4096 * 2 % 724 = 228
4096 * 3 % 724 = 704
4096 * 4 % 724 = 456
Clearly, ZS_HANDLE_SIZE should be considered when calculating
pages_per_zspage;
3. In get_size_class_index(), min(zs_size_classes - 1, idx) insures a
huge class's index <= zs_size_classes - 1, so it's no need to check
again.
Signed-off-by: Yinghao Xie <yinghao.xie@xxxxxxxxxxx>
Cc: Minchan Kim <minchan@xxxxxxxxxx>
Cc: Nitin Gupta <ngupta@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/zsmalloc.c | 15 ++++++---------
1 file changed, 6 insertions(+), 9 deletions(-)
diff -puN mm/zsmalloc.c~mm-zsmallocc-count-in-handles-size-when-calculating-pages_per_zspage mm/zsmalloc.c
--- a/mm/zsmalloc.c~mm-zsmallocc-count-in-handles-size-when-calculating-pages_per_zspage
+++ a/mm/zsmalloc.c
@@ -731,7 +731,8 @@ out:
* to form a zspage for each size class. This is important
* to reduce wastage due to unusable space left at end of
* each zspage which is given as:
- * wastage = Zp - Zp % size_class
+ * wastage = Zp % (class_size + ZS_HANDLE_SIZE)
+ * usage = Zp - wastage
* where Zp = zspage size = k * PAGE_SIZE where k = 1, 2, ...
*
* For example, for size class of 3/8 * PAGE_SIZE, we should
@@ -744,6 +745,9 @@ static int get_pages_per_zspage(int clas
/* zspage order which gives maximum used size per KB */
int max_usedpc_order = 1;
+ if (class_size > ZS_MAX_ALLOC_SIZE)
+ class_size = ZS_MAX_ALLOC_SIZE;
+
for (i = 1; i <= ZS_MAX_PAGES_PER_ZSPAGE; i++) {
int zspage_size;
int waste, usedpc;
@@ -1397,11 +1401,6 @@ unsigned long zs_malloc(struct zs_pool *
/* extra space in chunk to keep the handle */
size += ZS_HANDLE_SIZE;
class = pool->size_class[get_size_class_index(size)];
- /* In huge class size, we store the handle into first_page->private */
- if (class->huge) {
- size -= ZS_HANDLE_SIZE;
- class = pool->size_class[get_size_class_index(size)];
- }
spin_lock(&class->lock);
first_page = find_get_zspage(class);
@@ -1827,9 +1826,7 @@ struct zs_pool *zs_create_pool(char *nam
struct size_class *class;
size = ZS_MIN_ALLOC_SIZE + i * ZS_SIZE_CLASS_DELTA;
- if (size > ZS_MAX_ALLOC_SIZE)
- size = ZS_MAX_ALLOC_SIZE;
- pages_per_zspage = get_pages_per_zspage(size);
+ pages_per_zspage = get_pages_per_zspage(size + ZS_HANDLE_SIZE);
/*
* size_class is used for normal zsmalloc operation such
_
Patches currently in -mm which might be from yinghao.xie@xxxxxxxxxxx are
mm-zsmallocc-count-in-handles-size-when-calculating-pages_per_zspage.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
