blob: d5c0cef38a7122ed378371acba045f928f891c0b [file] [log] [blame]
Ryusuke Konishi962281a2009-04-06 19:01:20 -07001NILFS2
2------
3
4NILFS2 is a log-structured file system (LFS) supporting continuous
5snapshotting. In addition to versioning capability of the entire file
6system, users can even restore files mistakenly overwritten or
7destroyed just a few seconds ago. Since NILFS2 can keep consistency
8like conventional LFS, it achieves quick recovery after system
9crashes.
10
11NILFS2 creates a number of checkpoints every few seconds or per
12synchronous write basis (unless there is no change). Users can select
13significant versions among continuously created checkpoints, and can
14change them into snapshots which will be preserved until they are
15changed back to checkpoints.
16
17There is no limit on the number of snapshots until the volume gets
18full. Each snapshot is mountable as a read-only file system
19concurrently with its writable mount, and this feature is convenient
20for online backup.
21
22The userland tools are included in nilfs-utils package, which is
23available from the following download page. At least "mkfs.nilfs2",
24"mount.nilfs2", "umount.nilfs2", and "nilfs_cleanerd" (so called
25cleaner or garbage collector) are required. Details on the tools are
26described in the man pages included in the package.
27
28Project web page: http://www.nilfs.org/en/
29Download page: http://www.nilfs.org/en/download.html
30Git tree web page: http://www.nilfs.org/git/
Ryusuke Konishi6aff43f2010-01-02 21:41:53 +090031List info: http://vger.kernel.org/vger-lists.html#linux-nilfs
Ryusuke Konishi962281a2009-04-06 19:01:20 -070032
33Caveats
34=======
35
36Features which NILFS2 does not support yet:
37
38 - atime
39 - extended attributes
40 - POSIX ACLs
41 - quotas
Ryusuke Konishifb6e7112009-05-30 11:27:17 +090042 - fsck
43 - resize
Ryusuke Konishi962281a2009-04-06 19:01:20 -070044 - defragmentation
45
46Mount options
47=============
48
49NILFS2 supports the following mount options:
50(*) == default
51
Ryusuke Konishi773bc4f2010-07-05 13:00:08 +090052barrier(*) This enables/disables the use of write barriers. This
53nobarrier requires an IO stack which can support barriers, and
54 if nilfs gets an error on a barrier write, it will
55 disable again with a warning.
Ryusuke Konishi277a6a32010-04-02 18:02:33 +090056errors=continue Keep going on a filesystem error.
57errors=remount-ro(*) Remount the filesystem read-only on an error.
Ryusuke Konishi962281a2009-04-06 19:01:20 -070058errors=panic Panic and halt the machine if an error occurs.
59cp=n Specify the checkpoint-number of the snapshot to be
60 mounted. Checkpoints and snapshots are listed by lscp
61 user command. Only the checkpoints marked as snapshot
62 are mountable with this option. Snapshot is read-only,
63 so a read-only mount option must be specified together.
64order=relaxed(*) Apply relaxed order semantics that allows modified data
65 blocks to be written to disk without making a
66 checkpoint if no metadata update is going. This mode
67 is equivalent to the ordered data mode of the ext3
68 filesystem except for the updates on data blocks still
69 conserve atomicity. This will improve synchronous
70 write performance for overwriting.
71order=strict Apply strict in-order semantics that preserves sequence
72 of all file operations including overwriting of data
73 blocks. That means, it is guaranteed that no
74 overtaking of events occurs in the recovered file
75 system after a crash.
Ryusuke Konishi02345762009-11-20 03:28:01 +090076norecovery Disable recovery of the filesystem on mount.
77 This disables every write access on the device for
78 read-only mounts or snapshots. This option will fail
79 for r/w mounts on an unclean volume.
Ryusuke Konishi802d3172010-07-05 14:27:04 +090080discard This enables/disables the use of discard/TRIM commands.
81nodiscard(*) The discard/TRIM commands are sent to the underlying
82 block device when blocks are freed. This is useful
83 for SSD devices and sparse/thinly-provisioned LUNs.
Ryusuke Konishi962281a2009-04-06 19:01:20 -070084
85NILFS2 usage
86============
87
88To use nilfs2 as a local file system, simply:
89
90 # mkfs -t nilfs2 /dev/block_device
91 # mount -t nilfs2 /dev/block_device /dir
92
93This will also invoke the cleaner through the mount helper program
94(mount.nilfs2).
95
96Checkpoints and snapshots are managed by the following commands.
97Their manpages are included in the nilfs-utils package above.
98
99 lscp list checkpoints or snapshots.
100 mkcp make a checkpoint or a snapshot.
101 chcp change an existing checkpoint to a snapshot or vice versa.
102 rmcp invalidate specified checkpoint(s).
103
104To mount a snapshot,
105
106 # mount -t nilfs2 -r -o cp=<cno> /dev/block_device /snap_dir
107
108where <cno> is the checkpoint number of the snapshot.
109
110To unmount the NILFS2 mount point or snapshot, simply:
111
112 # umount /dir
113
114Then, the cleaner daemon is automatically shut down by the umount
115helper program (umount.nilfs2).
116
117Disk format
118===========
119
120A nilfs2 volume is equally divided into a number of segments except
121for the super block (SB) and segment #0. A segment is the container
122of logs. Each log is composed of summary information blocks, payload
123blocks, and an optional super root block (SR):
124
125 ______________________________________________________
126 | |SB| | Segment | Segment | Segment | ... | Segment | |
127 |_|__|_|____0____|____1____|____2____|_____|____N____|_|
128 0 +1K +4K +8M +16M +24M +(8MB x N)
129 . . (Typical offsets for 4KB-block)
130 . .
131 .______________________.
132 | log | log |... | log |
133 |__1__|__2__|____|__m__|
134 . .
135 . .
136 . .
137 .______________________________.
138 | Summary | Payload blocks |SR|
139 |_blocks__|_________________|__|
140
141The payload blocks are organized per file, and each file consists of
142data blocks and B-tree node blocks:
143
144 |<--- File-A --->|<--- File-B --->|
145 _______________________________________________________________
146 | Data blocks | B-tree blocks | Data blocks | B-tree blocks | ...
147 _|_____________|_______________|_____________|_______________|_
148
149
150Since only the modified blocks are written in the log, it may have
151files without data blocks or B-tree node blocks.
152
153The organization of the blocks is recorded in the summary information
154blocks, which contains a header structure (nilfs_segment_summary), per
155file structures (nilfs_finfo), and per block structures (nilfs_binfo):
156
157 _________________________________________________________________________
158 | Summary | finfo | binfo | ... | binfo | finfo | binfo | ... | binfo |...
159 |_blocks__|___A___|_(A,1)_|_____|(A,Na)_|___B___|_(B,1)_|_____|(B,Nb)_|___
160
161
162The logs include regular files, directory files, symbolic link files
163and several meta data files. The mata data files are the files used
164to maintain file system meta data. The current version of NILFS2 uses
165the following meta data files:
166
167 1) Inode file (ifile) -- Stores on-disk inodes
168 2) Checkpoint file (cpfile) -- Stores checkpoints
169 3) Segment usage file (sufile) -- Stores allocation state of segments
170 4) Data address translation file -- Maps virtual block numbers to usual
171 (DAT) block numbers. This file serves to
172 make on-disk blocks relocatable.
Ryusuke Konishi962281a2009-04-06 19:01:20 -0700173
174The following figure shows a typical organization of the logs:
175
176 _________________________________________________________________________
177 | Summary | regular file | file | ... | ifile | cpfile | sufile | DAT |SR|
178 |_blocks__|_or_directory_|_______|_____|_______|________|________|_____|__|
179
180
181To stride over segment boundaries, this sequence of files may be split
182into multiple logs. The sequence of logs that should be treated as
183logically one log, is delimited with flags marked in the segment
184summary. The recovery code of nilfs2 looks this boundary information
185to ensure atomicity of updates.
186
187The super root block is inserted for every checkpoints. It includes
188three special inodes, inodes for the DAT, cpfile, and sufile. Inodes
189of regular files, directories, symlinks and other special files, are
190included in the ifile. The inode of ifile itself is included in the
191corresponding checkpoint entry in the cpfile. Thus, the hierarchy
192among NILFS2 files can be depicted as follows:
193
194 Super block (SB)
195 |
196 v
197 Super root block (the latest cno=xx)
198 |-- DAT
199 |-- sufile
200 `-- cpfile
201 |-- ifile (cno=c1)
202 |-- ifile (cno=c2) ---- file (ino=i1)
203 : : |-- file (ino=i2)
204 `-- ifile (cno=xx) |-- file (ino=i3)
205 : :
206 `-- file (ino=yy)
207 ( regular file, directory, or symlink )
208
209For detail on the format of each file, please see include/linux/nilfs2_fs.h.