Commit | Line | Data |
---|---|---|
2fcbc413 MR |
1 | ======================= |
2 | Kernel Samepage Merging | |
3 | ======================= | |
7701c9c0 HD |
4 | |
5 | KSM is a memory-saving de-duplication feature, enabled by CONFIG_KSM=y, | |
2fcbc413 | 6 | added to the Linux kernel in 2.6.32. See ``mm/ksm.c`` for its implementation, |
93431e06 | 7 | and http://lwn.net/Articles/306704/ and https://lwn.net/Articles/330589/ |
7701c9c0 | 8 | |
ee865889 | 9 | The userspace interface of KSM is described in Documentation/admin-guide/mm/ksm.rst |
2c653d0e | 10 | |
064fca37 MR |
11 | Design |
12 | ====== | |
13 | ||
14 | Overview | |
15 | -------- | |
16 | ||
17 | .. kernel-doc:: mm/ksm.c | |
18 | :DOC: Overview | |
19 | ||
20 | Reverse mapping | |
21 | --------------- | |
22 | KSM maintains reverse mapping information for KSM pages in the stable | |
23 | tree. | |
24 | ||
25 | If a KSM page is shared between less than ``max_page_sharing`` VMAs, | |
26 | the node of the stable tree that represents such KSM page points to a | |
21fbd591 | 27 | list of struct ksm_rmap_item and the ``page->mapping`` of the |
064fca37 MR |
28 | KSM page points to the stable tree node. |
29 | ||
30 | When the sharing passes this threshold, KSM adds a second dimension to | |
31 | the stable tree. The tree node becomes a "chain" that links one or | |
32 | more "dups". Each "dup" keeps reverse mapping information for a KSM | |
33 | page with ``page->mapping`` pointing to that "dup". | |
34 | ||
35 | Every "chain" and all "dups" linked into a "chain" enforce the | |
36 | invariant that they represent the same write protected memory content, | |
37 | even if each "dup" will be pointed by a different KSM page copy of | |
38 | that content. | |
39 | ||
40 | This way the stable tree lookup computational complexity is unaffected | |
41 | if compared to an unlimited list of reverse mappings. It is still | |
42 | enforced that there cannot be KSM page content duplicates in the | |
43 | stable tree itself. | |
44 | ||
6570c785 MR |
45 | The deduplication limit enforced by ``max_page_sharing`` is required |
46 | to avoid the virtual memory rmap lists to grow too large. The rmap | |
47 | walk has O(N) complexity where N is the number of rmap_items | |
48 | (i.e. virtual mappings) that are sharing the page, which is in turn | |
49 | capped by ``max_page_sharing``. So this effectively spreads the linear | |
50 | O(N) computational complexity from rmap walk context over different | |
51 | KSM pages. The ksmd walk over the stable_node "chains" is also O(N), | |
52 | but N is the number of stable_node "dups", not the number of | |
53 | rmap_items, so it has not a significant impact on ksmd performance. In | |
54 | practice the best stable_node "dup" candidate will be kept and found | |
55 | at the head of the "dups" list. | |
56 | ||
57 | High values of ``max_page_sharing`` result in faster memory merging | |
58 | (because there will be fewer stable_node dups queued into the | |
59 | stable_node chain->hlist to check for pruning) and higher | |
60 | deduplication factor at the expense of slower worst case for rmap | |
61 | walks for any KSM page which can happen during swapping, compaction, | |
62 | NUMA balancing and page migration. | |
63 | ||
8b898fd1 MR |
64 | The ``stable_node_dups/stable_node_chains`` ratio is also affected by the |
65 | ``max_page_sharing`` tunable, and an high ratio may indicate fragmentation | |
66 | in the stable_node dups, which could be solved by introducing | |
67 | fragmentation algorithms in ksmd which would refile rmap_items from | |
68 | one stable_node dup to another stable_node dup, in order to free up | |
69 | stable_node "dups" with few rmap_items in them, but that may increase | |
70 | the ksmd CPU usage and possibly slowdown the readonly computations on | |
71 | the KSM pages of the applications. | |
72 | ||
2a695ca4 MR |
73 | The whole list of stable_node "dups" linked in the stable_node |
74 | "chains" is scanned periodically in order to prune stale stable_nodes. | |
75 | The frequency of such scans is defined by | |
76 | ``stable_node_chains_prune_millisecs`` sysfs tunable. | |
77 | ||
064fca37 MR |
78 | Reference |
79 | --------- | |
80 | .. kernel-doc:: mm/ksm.c | |
81 | :functions: mm_slot ksm_scan stable_node rmap_item | |
82 | ||
db12c00f | 83 | -- |
7701c9c0 | 84 | Izik Eidus, |
d0f209f6 | 85 | Hugh Dickins, 17 Nov 2009 |