blob: c631253cf85c763de6cfd347a15ce063fd71be07 [file] [log] [blame]
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001What: /sys/fs/ext4/<disk>/mb_stats
2Date: March 2008
3Contact: "Theodore Ts'o" <tytso@mit.edu>
4Description:
5 Controls whether the multiblock allocator should
6 collect statistics, which are shown during the unmount.
7 1 means to collect statistics, 0 means not to collect
8 statistics
9
10What: /sys/fs/ext4/<disk>/mb_group_prealloc
11Date: March 2008
12Contact: "Theodore Ts'o" <tytso@mit.edu>
13Description:
14 The multiblock allocator will round up allocation
15 requests to a multiple of this tuning parameter if the
16 stripe size is not set in the ext4 superblock
17
18What: /sys/fs/ext4/<disk>/mb_max_to_scan
19Date: March 2008
20Contact: "Theodore Ts'o" <tytso@mit.edu>
21Description:
22 The maximum number of extents the multiblock allocator
23 will search to find the best extent
24
25What: /sys/fs/ext4/<disk>/mb_min_to_scan
26Date: March 2008
27Contact: "Theodore Ts'o" <tytso@mit.edu>
28Description:
29 The minimum number of extents the multiblock allocator
30 will search to find the best extent
31
32What: /sys/fs/ext4/<disk>/mb_order2_req
33Date: March 2008
34Contact: "Theodore Ts'o" <tytso@mit.edu>
35Description:
36 Tuning parameter which controls the minimum size for
37 requests (as a power of 2) where the buddy cache is
38 used
39
40What: /sys/fs/ext4/<disk>/mb_stream_req
41Date: March 2008
42Contact: "Theodore Ts'o" <tytso@mit.edu>
43Description:
44 Files which have fewer blocks than this tunable
45 parameter will have their blocks allocated out of a
46 block group specific preallocation pool, so that small
47 files are packed closely together. Each large file
48 will have its blocks allocated out of its own unique
49 preallocation pool.
50
Lukas Czerner6f9524e2011-02-21 20:16:21 -050051What: /sys/fs/ext4/<disk>/inode_readahead_blks
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040052Date: March 2008
53Contact: "Theodore Ts'o" <tytso@mit.edu>
54Description:
55 Tuning parameter which controls the maximum number of
56 inode table blocks that ext4's inode table readahead
57 algorithm will pre-read into the buffer cache
58
59What: /sys/fs/ext4/<disk>/delayed_allocation_blocks
60Date: March 2008
61Contact: "Theodore Ts'o" <tytso@mit.edu>
62Description:
63 This file is read-only and shows the number of blocks
64 that are dirty in the page cache, but which do not
65 have their location in the filesystem allocated yet.
66
67What: /sys/fs/ext4/<disk>/lifetime_write_kbytes
68Date: March 2008
69Contact: "Theodore Ts'o" <tytso@mit.edu>
70Description:
71 This file is read-only and shows the number of kilobytes
72 of data that have been written to this filesystem since it was
73 created.
74
75What: /sys/fs/ext4/<disk>/session_write_kbytes
76Date: March 2008
77Contact: "Theodore Ts'o" <tytso@mit.edu>
78Description:
79 This file is read-only and shows the number of
80 kilobytes of data that have been written to this
81 filesystem since it was mounted.
Andreas Dilger11013912009-06-13 11:45:35 -040082
83What: /sys/fs/ext4/<disk>/inode_goal
84Date: June 2008
85Contact: "Theodore Ts'o" <tytso@mit.edu>
86Description:
87 Tuning parameter which (if non-zero) controls the goal
Lukas Czerner6f9524e2011-02-21 20:16:21 -050088 inode used by the inode allocator in preference to
89 all other allocation heuristics. This is intended for
Andreas Dilger11013912009-06-13 11:45:35 -040090 debugging use only, and should be 0 on production
91 systems.
Lukas Czerner6f9524e2011-02-21 20:16:21 -050092
93What: /sys/fs/ext4/<disk>/max_writeback_mb_bump
94Date: September 2009
95Contact: "Theodore Ts'o" <tytso@mit.edu>
96Description:
97 The maximum number of megabytes the writeback code will
98 try to write out before move on to another inode.
Zheng Liu67a5da52012-08-17 09:54:17 -040099
100What: /sys/fs/ext4/<disk>/extent_max_zeroout_kb
101Date: August 2012
102Contact: "Theodore Ts'o" <tytso@mit.edu>
103Description:
104 The maximum number of kilobytes which will be zeroed
105 out in preference to creating a new uninitialized
106 extent when manipulating an inode's extent tree. Note
107 that using a larger value will increase the
108 variability of time necessary to complete a random
109 write operation (since a 4k random write might turn
110 into a much larger write due to the zeroout
111 operation).