Joe Thornber
88c63f620f
Lots of work on the btree_damage_visitor
2013-05-13 12:36:57 +01:00
Joe Thornber
0dba0abe88
Bump up the nr of blocks in the cache from 64 to 1024.
2013-05-13 12:34:52 +01:00
Joe Thornber
08c5fde04f
btree_damage_tracker
2013-05-13 11:27:38 +01:00
Joe Thornber
42a4221499
A little optimisation for the core space map.
2013-05-09 14:09:47 +01:00
Joe Thornber
8ecf2f27ed
btree_damage_visitor stuff.
2013-05-09 13:31:04 +01:00
Joe Thornber
ab66e9f8e3
Some more work on btree_damage_visitor
2013-05-08 16:38:38 +01:00
Joe Thornber
a7adefbae8
Move range.h to persistent-data
2013-05-08 16:38:04 +01:00
Joe Thornber
23ef3b387d
Start btree_damage_visitor.
2013-05-08 12:00:24 +01:00
Joe Thornber
f6acd473a7
Move with_directory and with_temporary_directory to test_utils.
2013-05-08 11:47:15 +01:00
Joe Thornber
bac19b3c94
[btree_checker] no longer need to check the checksum, since the block validator has already done this.
2013-05-07 15:47:43 +01:00
Joe Thornber
44d0b1903f
[btree] allow visitors to trap node access failures such as checksum error.
2013-05-07 15:22:13 +01:00
Joe Thornber
8e0271b3bf
[btree] When walking a btree use a validator when taking the locks.
...
We really can't rely on anything in the block if the checksum is
wrong.
2013-05-07 14:54:20 +01:00
Joe Thornber
1e14e1cd10
[metadata_checker_t] fix silly allocation bug in the devices_visitor.
2013-05-07 14:29:54 +01:00
Joe Thornber
d113ff6e5a
thin restorer is now careful to share the mapping tree for empty devices.
...
This saves a lot of metadata space if you're creating lots of empty
devs (4k per dev). Also makes things faster since we don't have to
allocate.
2013-05-07 14:24:59 +01:00
Joe Thornber
edf16f59c7
[restorer] Mappings only get inserted when the device is completed.
...
This saves looking up the device mapping tree for every insert.
2013-05-02 12:59:42 +01:00
Joe Thornber
5a231a7db0
[metadata_checker_t] add damage_should_include() method.
2013-05-02 12:51:55 +01:00
Joe Thornber
c976fc3d6c
[metadata_checker_t] Make device_checker a member of DeviceCheckerTests
2013-05-02 12:43:16 +01:00
Joe Thornber
a828cde96d
[btree] visitor now passed as a reference rather than shared_ptr
2013-05-02 12:31:33 +01:00
Joe Thornber
5913de5e38
[metadata_checker_t] check the range for missing_devices_damage.
2013-05-02 12:22:54 +01:00
Joe Thornber
008a4d495c
[metadata_checker_t] Some more work on the devices_visitor.
2013-05-02 12:04:05 +01:00
Joe Thornber
3832fd5952
More work on metadata_checker_t
2013-05-01 16:33:42 +01:00
Joe Thornber
647cb68b76
Some test utils
2013-05-01 16:33:24 +01:00
Joe Thornber
21603c5a5d
ostream operator << for range template.
2013-05-01 16:30:59 +01:00
Joe Thornber
e1c5d485eb
Tidy up a typedef
2013-05-01 16:29:44 +01:00
Joe Thornber
5d0b23beea
Some btree visitor tidying.
...
Introduce node_location to replace the long list of parameters. Also
add a depth field to keep track of the depth from root.
2013-05-01 16:16:23 +01:00
Joe Thornber
1884233a2b
DeviceCheckerTests.passes_with_valid_metadata_containing_some_devices
2013-04-29 16:35:03 +01:00
Joe Thornber
c782da7a34
Use a mock of the metadata_damage_visitor to confirm the type off damage reported from the device_checker.
2013-04-29 16:13:18 +01:00
Joe Thornber
fa96ec6301
Add operator == for the range template
2013-04-29 16:12:34 +01:00
Joe Thornber
5e26d51372
add equality operators for the metadata_damage classes.
2013-04-29 16:12:05 +01:00
Joe Thornber
969a5f62c6
Working on device_checker
2013-04-29 15:37:53 +01:00
Joe Thornber
524f9474cf
Add a range class
2013-04-29 15:10:01 +01:00
Joe Thornber
6422b22610
Move block_manager member down into checker base class.
2013-04-29 13:31:33 +01:00
Joe Thornber
1a8b1b29e8
Start device_checker
2013-04-29 13:24:19 +01:00
Joe Thornber
c0b14ce280
fluff
2013-04-29 12:49:32 +01:00
Joe Thornber
034abaff1e
Stop metadata:commit printing the superblock out to stderr.
2013-04-29 12:48:58 +01:00
Joe Thornber
2b8b74065b
add fixme
2013-04-29 12:40:56 +01:00
Joe Thornber
925dbdbc53
[block_t] verify any exception thrown by validator->check() gets passed up, and the lock is not consequently taken.
2013-04-29 12:32:04 +01:00
Joe Thornber
a7c0b0c39b
[block, lock_checker] add an is_locked() method for use in unit tests.
2013-04-29 12:31:30 +01:00
Joe Thornber
a0f53f47aa
[block_t] a superblock_zero test was mistakenly calling plain write_lock.
2013-04-29 12:05:55 +01:00
Joe Thornber
1fe0df9f1a
[block_t] make sure validator->check isn't called unexpectedly.
2013-04-29 12:03:47 +01:00
Joe Thornber
441cba7185
Small work on metadata_checker tests.
2013-04-26 15:55:45 +01:00
Joe Thornber
b1e0ca2207
[block] make sure we can change validators
2013-04-26 15:54:15 +01:00
Joe Thornber
5367631c8c
[various data structures] Reuse validators rather than always creating a new one.
2013-04-26 14:15:20 +01:00
Joe Thornber
53ac7b8119
[block_manager] Allow people to change the validator for a block
2013-04-26 14:14:28 +01:00
Joe Thornber
0b4c359a63
[block manager] The dirty flag for blocks was never being cleared, causes redundant writes.
2013-04-26 13:18:24 +01:00
Joe Thornber
a55f6fcf78
A stack of thin_check refactorings
2013-04-25 16:27:07 +01:00
Joe Thornber
20ff78c818
[thin_check] Start refactoring metadata_checker.
2013-04-23 15:21:44 +01:00
Joe Thornber
70fd048426
Add a TODO file
2013-04-23 11:18:21 +01:00
Joe Thornber
9943bd6549
cucumber tests were generating invalid xml metadata.
2013-04-23 11:17:50 +01:00
Joe Thornber
d8a208cbc8
A stack of space map tweaks.
...
new_block() is now a concrete method implemented using the virtual
find_free() and inc() methods.
recursive space map is better at giving correct reference counts.
2013-04-23 10:57:47 +01:00
