blob: 64ed63c4f69d6a95e22a1ac766d6d4127a8a3788 [file] [log] [blame]
Minchan Kimd02be502015-04-15 16:15:46 -07001zsmalloc
2--------
3
4This allocator is designed for use with zram. Thus, the allocator is
5supposed to work well under low memory conditions. In particular, it
6never attempts higher order page allocation which is very likely to
7fail under memory pressure. On the other hand, if we just use single
8(0-order) pages, it would suffer from very high fragmentation --
9any object of size PAGE_SIZE/2 or larger would occupy an entire page.
10This was one of the major issues with its predecessor (xvmalloc).
11
12To overcome these issues, zsmalloc allocates a bunch of 0-order pages
13and links them together using various 'struct page' fields. These linked
14pages act as a single higher-order page i.e. an object can span 0-order
15page boundaries. The code refers to these linked pages as a single entity
16called zspage.
17
18For simplicity, zsmalloc can only allocate objects of size up to PAGE_SIZE
19since this satisfies the requirements of all its current users (in the
20worst case, page is incompressible and is thus stored "as-is" i.e. in
21uncompressed form). For allocation requests larger than this size, failure
22is returned (see zs_malloc).
23
24Additionally, zs_malloc() does not return a dereferenceable pointer.
25Instead, it returns an opaque handle (unsigned long) which encodes actual
26location of the allocated object. The reason for this indirection is that
27zsmalloc does not keep zspages permanently mapped since that would cause
28issues on 32-bit systems where the VA region for kernel space mappings
29is very small. So, before using the allocating memory, the object has to
30be mapped using zs_map_object() to get a usable pointer and subsequently
31unmapped using zs_unmap_object().
32
33stat
34----
35
36With CONFIG_ZSMALLOC_STAT, we could see zsmalloc internal information via
37/sys/kernel/debug/zsmalloc/<user name>. Here is a sample of stat output:
38
39# cat /sys/kernel/debug/zsmalloc/zram0/classes
40
41 class size almost_full almost_empty obj_allocated obj_used pages_used pages_per_zspage
42 ..
43 ..
44 9 176 0 1 186 129 8 4
45 10 192 1 0 2880 2872 135 3
46 11 208 0 1 819 795 42 2
47 12 224 0 1 219 159 12 4
48 ..
49 ..
50
51
52class: index
53size: object size zspage stores
54almost_empty: the number of ZS_ALMOST_EMPTY zspages(see below)
55almost_full: the number of ZS_ALMOST_FULL zspages(see below)
56obj_allocated: the number of objects allocated
57obj_used: the number of objects allocated to the user
58pages_used: the number of pages allocated for the class
59pages_per_zspage: the number of 0-order pages to make a zspage
60
61We assign a zspage to ZS_ALMOST_EMPTY fullness group when:
62 n <= N / f, where
63n = number of allocated objects
64N = total number of objects zspage can store
65f = fullness_threshold_frac(ie, 4 at the moment)
66
67Similarly, we assign zspage to:
68 ZS_ALMOST_FULL when n > N / f
69 ZS_EMPTY when n == 0
70 ZS_FULL when n == N