aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat (limited to 'Documentation/device-mapper/persistent-data.txt')
-rw-r--r--Documentation/device-mapper/persistent-data.txt84
1 files changed, 84 insertions, 0 deletions
diff --git a/Documentation/device-mapper/persistent-data.txt b/Documentation/device-mapper/persistent-data.txt
new file mode 100644
index 000000000000..0e5df9b04ad2
--- /dev/null
+++ b/Documentation/device-mapper/persistent-data.txt
@@ -0,0 +1,84 @@
1Introduction
2============
3
4The more-sophisticated device-mapper targets require complex metadata
5that is managed in kernel. In late 2010 we were seeing that various
6different targets were rolling their own data strutures, for example:
7
8- Mikulas Patocka's multisnap implementation
9- Heinz Mauelshagen's thin provisioning target
10- Another btree-based caching target posted to dm-devel
11- Another multi-snapshot target based on a design of Daniel Phillips
12
13Maintaining these data structures takes a lot of work, so if possible
14we'd like to reduce the number.
15
16The persistent-data library is an attempt to provide a re-usable
17framework for people who want to store metadata in device-mapper
18targets. It's currently used by the thin-provisioning target and an
19upcoming hierarchical storage target.
20
21Overview
22========
23
24The main documentation is in the header files which can all be found
25under drivers/md/persistent-data.
26
27The block manager
28-----------------
29
30dm-block-manager.[hc]
31
32This provides access to the data on disk in fixed sized-blocks. There
33is a read/write locking interface to prevent concurrent accesses, and
34keep data that is being used in the cache.
35
36Clients of persistent-data are unlikely to use this directly.
37
38The transaction manager
39-----------------------
40
41dm-transaction-manager.[hc]
42
43This restricts access to blocks and enforces copy-on-write semantics.
44The only way you can get hold of a writable block through the
45transaction manager is by shadowing an existing block (ie. doing
46copy-on-write) or allocating a fresh one. Shadowing is elided within
47the same transaction so performance is reasonable. The commit method
48ensures that all data is flushed before it writes the superblock.
49On power failure your metadata will be as it was when last committed.
50
51The Space Maps
52--------------
53
54dm-space-map.h
55dm-space-map-metadata.[hc]
56dm-space-map-disk.[hc]
57
58On-disk data structures that keep track of reference counts of blocks.
59Also acts as the allocator of new blocks. Currently two
60implementations: a simpler one for managing blocks on a different
61device (eg. thinly-provisioned data blocks); and one for managing
62the metadata space. The latter is complicated by the need to store
63its own data within the space it's managing.
64
65The data structures
66-------------------
67
68dm-btree.[hc]
69dm-btree-remove.c
70dm-btree-spine.c
71dm-btree-internal.h
72
73Currently there is only one data structure, a hierarchical btree.
74There are plans to add more. For example, something with an
75array-like interface would see a lot of use.
76
77The btree is 'hierarchical' in that you can define it to be composed
78of nested btrees, and take multiple keys. For example, the
79thin-provisioning target uses a btree with two levels of nesting.
80The first maps a device id to a mapping tree, and that in turn maps a
81virtual block to a physical block.
82
83Values stored in the btrees can have arbitrary size. Keys are always
8464bits, although nesting allows you to use multiple keys.