| 1 | .. SPDX-License-Identifier: GPL-2.0 |
| 2 | |
| 3 | ====== |
| 4 | NILFS2 |
| 5 | ====== |
| 6 | |
| 7 | NILFS2 is a log-structured file system (LFS) supporting continuous |
| 8 | snapshotting. In addition to versioning capability of the entire file |
| 9 | system, users can even restore files mistakenly overwritten or |
| 10 | destroyed just a few seconds ago. Since NILFS2 can keep consistency |
| 11 | like conventional LFS, it achieves quick recovery after system |
| 12 | crashes. |
| 13 | |
| 14 | NILFS2 creates a number of checkpoints every few seconds or per |
| 15 | synchronous write basis (unless there is no change). Users can select |
| 16 | significant versions among continuously created checkpoints, and can |
| 17 | change them into snapshots which will be preserved until they are |
| 18 | changed back to checkpoints. |
| 19 | |
| 20 | There is no limit on the number of snapshots until the volume gets |
| 21 | full. Each snapshot is mountable as a read-only file system |
| 22 | concurrently with its writable mount, and this feature is convenient |
| 23 | for online backup. |
| 24 | |
| 25 | The userland tools are included in nilfs-utils package, which is |
| 26 | available from the following download page. At least "mkfs.nilfs2", |
| 27 | "mount.nilfs2", "umount.nilfs2", and "nilfs_cleanerd" (so called |
| 28 | cleaner or garbage collector) are required. Details on the tools are |
| 29 | described in the man pages included in the package. |
| 30 | |
| 31 | :Project web page: https://nilfs.sourceforge.io/ |
| 32 | :Download page: https://nilfs.sourceforge.io/en/download.html |
| 33 | :List info: http://vger.kernel.org/vger-lists.html#linux-nilfs |
| 34 | |
| 35 | Caveats |
| 36 | ======= |
| 37 | |
| 38 | Features which NILFS2 does not support yet: |
| 39 | |
| 40 | - atime |
| 41 | - extended attributes |
| 42 | - POSIX ACLs |
| 43 | - quotas |
| 44 | - fsck |
| 45 | - defragmentation |
| 46 | |
| 47 | Mount options |
| 48 | ============= |
| 49 | |
| 50 | NILFS2 supports the following mount options: |
| 51 | (*) == default |
| 52 | |
| 53 | ======================= ======================================================= |
| 54 | barrier(*) This enables/disables the use of write barriers. This |
| 55 | nobarrier requires an IO stack which can support barriers, and |
| 56 | if nilfs gets an error on a barrier write, it will |
| 57 | disable again with a warning. |
| 58 | errors=continue Keep going on a filesystem error. |
| 59 | errors=remount-ro(*) Remount the filesystem read-only on an error. |
| 60 | errors=panic Panic and halt the machine if an error occurs. |
| 61 | cp=n Specify the checkpoint-number of the snapshot to be |
| 62 | mounted. Checkpoints and snapshots are listed by lscp |
| 63 | user command. Only the checkpoints marked as snapshot |
| 64 | are mountable with this option. Snapshot is read-only, |
| 65 | so a read-only mount option must be specified together. |
| 66 | order=relaxed(*) Apply relaxed order semantics that allows modified data |
| 67 | blocks to be written to disk without making a |
| 68 | checkpoint if no metadata update is going. This mode |
| 69 | is equivalent to the ordered data mode of the ext3 |
| 70 | filesystem except for the updates on data blocks still |
| 71 | conserve atomicity. This will improve synchronous |
| 72 | write performance for overwriting. |
| 73 | order=strict Apply strict in-order semantics that preserves sequence |
| 74 | of all file operations including overwriting of data |
| 75 | blocks. That means, it is guaranteed that no |
| 76 | overtaking of events occurs in the recovered file |
| 77 | system after a crash. |
| 78 | norecovery Disable recovery of the filesystem on mount. |
| 79 | This disables every write access on the device for |
| 80 | read-only mounts or snapshots. This option will fail |
| 81 | for r/w mounts on an unclean volume. |
| 82 | discard This enables/disables the use of discard/TRIM commands. |
| 83 | nodiscard(*) The discard/TRIM commands are sent to the underlying |
| 84 | block device when blocks are freed. This is useful |
| 85 | for SSD devices and sparse/thinly-provisioned LUNs. |
| 86 | ======================= ======================================================= |
| 87 | |
| 88 | Ioctls |
| 89 | ====== |
| 90 | |
| 91 | There is some NILFS2 specific functionality which can be accessed by applications |
| 92 | through the system call interfaces. The list of all NILFS2 specific ioctls are |
| 93 | shown in the table below. |
| 94 | |
| 95 | Table of NILFS2 specific ioctls: |
| 96 | |
| 97 | ============================== =============================================== |
| 98 | Ioctl Description |
| 99 | ============================== =============================================== |
| 100 | NILFS_IOCTL_CHANGE_CPMODE Change mode of given checkpoint between |
| 101 | checkpoint and snapshot state. This ioctl is |
| 102 | used in chcp and mkcp utilities. |
| 103 | |
| 104 | NILFS_IOCTL_DELETE_CHECKPOINT Remove checkpoint from NILFS2 file system. |
| 105 | This ioctl is used in rmcp utility. |
| 106 | |
| 107 | NILFS_IOCTL_GET_CPINFO Return info about requested checkpoints. This |
| 108 | ioctl is used in lscp utility and by |
| 109 | nilfs_cleanerd daemon. |
| 110 | |
| 111 | NILFS_IOCTL_GET_CPSTAT Return checkpoints statistics. This ioctl is |
| 112 | used by lscp, rmcp utilities and by |
| 113 | nilfs_cleanerd daemon. |
| 114 | |
| 115 | NILFS_IOCTL_GET_SUINFO Return segment usage info about requested |
| 116 | segments. This ioctl is used in lssu, |
| 117 | nilfs_resize utilities and by nilfs_cleanerd |
| 118 | daemon. |
| 119 | |
| 120 | NILFS_IOCTL_SET_SUINFO Modify segment usage info of requested |
| 121 | segments. This ioctl is used by |
| 122 | nilfs_cleanerd daemon to skip unnecessary |
| 123 | cleaning operation of segments and reduce |
| 124 | performance penalty or wear of flash device |
| 125 | due to redundant move of in-use blocks. |
| 126 | |
| 127 | NILFS_IOCTL_GET_SUSTAT Return segment usage statistics. This ioctl |
| 128 | is used in lssu, nilfs_resize utilities and |
| 129 | by nilfs_cleanerd daemon. |
| 130 | |
| 131 | NILFS_IOCTL_GET_VINFO Return information on virtual block addresses. |
| 132 | This ioctl is used by nilfs_cleanerd daemon. |
| 133 | |
| 134 | NILFS_IOCTL_GET_BDESCS Return information about descriptors of disk |
| 135 | block numbers. This ioctl is used by |
| 136 | nilfs_cleanerd daemon. |
| 137 | |
| 138 | NILFS_IOCTL_CLEAN_SEGMENTS Do garbage collection operation in the |
| 139 | environment of requested parameters from |
| 140 | userspace. This ioctl is used by |
| 141 | nilfs_cleanerd daemon. |
| 142 | |
| 143 | NILFS_IOCTL_SYNC Make a checkpoint. This ioctl is used in |
| 144 | mkcp utility. |
| 145 | |
| 146 | NILFS_IOCTL_RESIZE Resize NILFS2 volume. This ioctl is used |
| 147 | by nilfs_resize utility. |
| 148 | |
| 149 | NILFS_IOCTL_SET_ALLOC_RANGE Define lower limit of segments in bytes and |
| 150 | upper limit of segments in bytes. This ioctl |
| 151 | is used by nilfs_resize utility. |
| 152 | ============================== =============================================== |
| 153 | |
| 154 | NILFS2 usage |
| 155 | ============ |
| 156 | |
| 157 | To use nilfs2 as a local file system, simply:: |
| 158 | |
| 159 | # mkfs -t nilfs2 /dev/block_device |
| 160 | # mount -t nilfs2 /dev/block_device /dir |
| 161 | |
| 162 | This will also invoke the cleaner through the mount helper program |
| 163 | (mount.nilfs2). |
| 164 | |
| 165 | Checkpoints and snapshots are managed by the following commands. |
| 166 | Their manpages are included in the nilfs-utils package above. |
| 167 | |
| 168 | ==== =========================================================== |
| 169 | lscp list checkpoints or snapshots. |
| 170 | mkcp make a checkpoint or a snapshot. |
| 171 | chcp change an existing checkpoint to a snapshot or vice versa. |
| 172 | rmcp invalidate specified checkpoint(s). |
| 173 | ==== =========================================================== |
| 174 | |
| 175 | To mount a snapshot:: |
| 176 | |
| 177 | # mount -t nilfs2 -r -o cp=<cno> /dev/block_device /snap_dir |
| 178 | |
| 179 | where <cno> is the checkpoint number of the snapshot. |
| 180 | |
| 181 | To unmount the NILFS2 mount point or snapshot, simply:: |
| 182 | |
| 183 | # umount /dir |
| 184 | |
| 185 | Then, the cleaner daemon is automatically shut down by the umount |
| 186 | helper program (umount.nilfs2). |
| 187 | |
| 188 | Disk format |
| 189 | =========== |
| 190 | |
| 191 | A nilfs2 volume is equally divided into a number of segments except |
| 192 | for the super block (SB) and segment #0. A segment is the container |
| 193 | of logs. Each log is composed of summary information blocks, payload |
| 194 | blocks, and an optional super root block (SR):: |
| 195 | |
| 196 | ______________________________________________________ |
| 197 | | |SB| | Segment | Segment | Segment | ... | Segment | | |
| 198 | |_|__|_|____0____|____1____|____2____|_____|____N____|_| |
| 199 | 0 +1K +4K +8M +16M +24M +(8MB x N) |
| 200 | . . (Typical offsets for 4KB-block) |
| 201 | . . |
| 202 | .______________________. |
| 203 | | log | log |... | log | |
| 204 | |__1__|__2__|____|__m__| |
| 205 | . . |
| 206 | . . |
| 207 | . . |
| 208 | .______________________________. |
| 209 | | Summary | Payload blocks |SR| |
| 210 | |_blocks__|_________________|__| |
| 211 | |
| 212 | The payload blocks are organized per file, and each file consists of |
| 213 | data blocks and B-tree node blocks:: |
| 214 | |
| 215 | |<--- File-A --->|<--- File-B --->| |
| 216 | _______________________________________________________________ |
| 217 | | Data blocks | B-tree blocks | Data blocks | B-tree blocks | ... |
| 218 | _|_____________|_______________|_____________|_______________|_ |
| 219 | |
| 220 | |
| 221 | Since only the modified blocks are written in the log, it may have |
| 222 | files without data blocks or B-tree node blocks. |
| 223 | |
| 224 | The organization of the blocks is recorded in the summary information |
| 225 | blocks, which contains a header structure (nilfs_segment_summary), per |
| 226 | file structures (nilfs_finfo), and per block structures (nilfs_binfo):: |
| 227 | |
| 228 | _________________________________________________________________________ |
| 229 | | Summary | finfo | binfo | ... | binfo | finfo | binfo | ... | binfo |... |
| 230 | |_blocks__|___A___|_(A,1)_|_____|(A,Na)_|___B___|_(B,1)_|_____|(B,Nb)_|___ |
| 231 | |
| 232 | |
| 233 | The logs include regular files, directory files, symbolic link files |
| 234 | and several meta data files. The mata data files are the files used |
| 235 | to maintain file system meta data. The current version of NILFS2 uses |
| 236 | the following meta data files:: |
| 237 | |
| 238 | 1) Inode file (ifile) -- Stores on-disk inodes |
| 239 | 2) Checkpoint file (cpfile) -- Stores checkpoints |
| 240 | 3) Segment usage file (sufile) -- Stores allocation state of segments |
| 241 | 4) Data address translation file -- Maps virtual block numbers to usual |
| 242 | (DAT) block numbers. This file serves to |
| 243 | make on-disk blocks relocatable. |
| 244 | |
| 245 | The following figure shows a typical organization of the logs:: |
| 246 | |
| 247 | _________________________________________________________________________ |
| 248 | | Summary | regular file | file | ... | ifile | cpfile | sufile | DAT |SR| |
| 249 | |_blocks__|_or_directory_|_______|_____|_______|________|________|_____|__| |
| 250 | |
| 251 | |
| 252 | To stride over segment boundaries, this sequence of files may be split |
| 253 | into multiple logs. The sequence of logs that should be treated as |
| 254 | logically one log, is delimited with flags marked in the segment |
| 255 | summary. The recovery code of nilfs2 looks this boundary information |
| 256 | to ensure atomicity of updates. |
| 257 | |
| 258 | The super root block is inserted for every checkpoints. It includes |
| 259 | three special inodes, inodes for the DAT, cpfile, and sufile. Inodes |
| 260 | of regular files, directories, symlinks and other special files, are |
| 261 | included in the ifile. The inode of ifile itself is included in the |
| 262 | corresponding checkpoint entry in the cpfile. Thus, the hierarchy |
| 263 | among NILFS2 files can be depicted as follows:: |
| 264 | |
| 265 | Super block (SB) |
| 266 | | |
| 267 | v |
| 268 | Super root block (the latest cno=xx) |
| 269 | |-- DAT |
| 270 | |-- sufile |
| 271 | `-- cpfile |
| 272 | |-- ifile (cno=c1) |
| 273 | |-- ifile (cno=c2) ---- file (ino=i1) |
| 274 | : : |-- file (ino=i2) |
| 275 | `-- ifile (cno=xx) |-- file (ino=i3) |
| 276 | : : |
| 277 | `-- file (ino=yy) |
| 278 | ( regular file, directory, or symlink ) |
| 279 | |
| 280 | For detail on the format of each file, please see nilfs2_ondisk.h |
| 281 | located at include/uapi/linux directory. |
| 282 | |
| 283 | There are no patents or other intellectual property that we protect |
| 284 | with regard to the design of NILFS2. It is allowed to replicate the |
| 285 | design in hopes that other operating systems could share (mount, read, |
| 286 | write, etc.) data stored in this format. |