xref: /linux/Documentation/filesystems/squashfs.rst (revision a36e9f5cfe9eb3a1dce8769c7058251c42705357)
1.. SPDX-License-Identifier: GPL-2.0
2
3=======================
4Squashfs 4.0 Filesystem
5=======================
6
7Squashfs is a compressed read-only filesystem for Linux.
8
9It uses zlib, lz4, lzo, or xz compression to compress files, inodes and
10directories.  Inodes in the system are very small and all blocks are packed to
11minimise data overhead. Block sizes greater than 4K are supported up to a
12maximum of 1Mbytes (default block size 128K).
13
14Squashfs is intended for general read-only filesystem use, for archival
15use (i.e. in cases where a .tar.gz file may be used), and in constrained
16block device/memory systems (e.g. embedded systems) where low overhead is
17needed.
18
19Mailing list: squashfs-devel@lists.sourceforge.net
20Web site: www.squashfs.org
21
221. Filesystem Features
23----------------------
24
25Squashfs filesystem features versus Cramfs:
26
27============================== 	=========		==========
28				Squashfs		Cramfs
29============================== 	=========		==========
30Max filesystem size		2^64			256 MiB
31Max file size			~ 2 TiB			16 MiB
32Max files			unlimited		unlimited
33Max directories			unlimited		unlimited
34Max entries per directory	unlimited		unlimited
35Max block size			1 MiB			4 KiB
36Metadata compression		yes			no
37Directory indexes		yes			no
38Sparse file support		yes			no
39Tail-end packing (fragments)	yes			no
40Exportable (NFS etc.)		yes			no
41Hard link support		yes			no
42"." and ".." in readdir		yes			no
43Real inode numbers		yes			no
4432-bit uids/gids		yes			no
45File creation time		yes			no
46Xattr support			yes			no
47ACL support			no			no
48============================== 	=========		==========
49
50Squashfs compresses data, inodes and directories.  In addition, inode and
51directory data are highly compacted, and packed on byte boundaries.  Each
52compressed inode is on average 8 bytes in length (the exact length varies on
53file type, i.e. regular file, directory, symbolic link, and block/char device
54inodes have different sizes).
55
562. Using Squashfs
57-----------------
58
59As squashfs is a read-only filesystem, the mksquashfs program must be used to
60create populated squashfs filesystems.  This and other squashfs utilities
61can be obtained from http://www.squashfs.org.  Usage instructions can be
62obtained from this site also.
63
64The squashfs-tools development tree is now located on kernel.org
65	git://git.kernel.org/pub/scm/fs/squashfs/squashfs-tools.git
66
672.1 Mount options
68-----------------
69===================    =========================================================
70errors=%s              Specify whether squashfs errors trigger a kernel panic
71                       or not
72
73		       ==========  =============================================
74                         continue  errors don't trigger a panic (default)
75                            panic  trigger a panic when errors are encountered,
76                                   similar to several other filesystems (e.g.
77                                   btrfs, ext4, f2fs, GFS2, jfs, ntfs, ubifs)
78
79                                   This allows a kernel dump to be saved,
80                                   useful for analyzing and debugging the
81                                   corruption.
82                       ==========  =============================================
83threads=%s             Select the decompression mode or the number of threads
84
85                       If SQUASHFS_CHOICE_DECOMP_BY_MOUNT is set:
86
87		       ==========  =============================================
88                           single  use single-threaded decompression (default)
89
90                                   Only one block (data or metadata) can be
91                                   decompressed at any one time. This limits
92                                   CPU and memory usage to a minimum, but it
93                                   also gives poor performance on parallel I/O
94                                   workloads when using multiple CPU machines
95                                   due to waiting on decompressor availability.
96                            multi  use up to two parallel decompressors per core
97
98                                   If you have a parallel I/O workload and your
99                                   system has enough memory, using this option
100                                   may improve overall I/O performance. It
101                                   dynamically allocates decompressors on a
102                                   demand basis.
103                           percpu  use a maximum of one decompressor per core
104
105                                   It uses percpu variables to ensure
106                                   decompression is load-balanced across the
107                                   cores.
108                        1|2|3|...  configure the number of threads used for
109                                   decompression
110
111                                   The upper limit is num_online_cpus() * 2.
112                       ==========  =============================================
113
114                       If SQUASHFS_CHOICE_DECOMP_BY_MOUNT is **not** set and
115                       SQUASHFS_DECOMP_MULTI, SQUASHFS_MOUNT_DECOMP_THREADS are
116                       both set:
117
118		       ==========  =============================================
119                          2|3|...  configure the number of threads used for
120                                   decompression
121
122                                   The upper limit is num_online_cpus() * 2.
123                       ==========  =============================================
124
125===================    =========================================================
126
1273. Squashfs Filesystem Design
128-----------------------------
129
130A squashfs filesystem consists of a maximum of nine parts, packed together on a
131byte alignment::
132
133	 ---------------
134	|  superblock 	|
135	|---------------|
136	|  compression  |
137	|    options    |
138	|---------------|
139	|  datablocks   |
140	|  & fragments  |
141	|---------------|
142	|  inode table	|
143	|---------------|
144	|   directory	|
145	|     table     |
146	|---------------|
147	|   fragment	|
148	|    table      |
149	|---------------|
150	|    export     |
151	|    table      |
152	|---------------|
153	|    uid/gid	|
154	|  lookup table	|
155	|---------------|
156	|     xattr     |
157	|     table	|
158	 ---------------
159
160Compressed data blocks are written to the filesystem as files are read from
161the source directory, and checked for duplicates.  Once all file data has been
162written the completed inode, directory, fragment, export, uid/gid lookup and
163xattr tables are written.
164
1653.1 Compression options
166-----------------------
167
168Compressors can optionally support compression specific options (e.g.
169dictionary size).  If non-default compression options have been used, then
170these are stored here.
171
1723.2 Inodes
173----------
174
175Metadata (inodes and directories) are compressed in 8Kbyte blocks.  Each
176compressed block is prefixed by a two byte length, the top bit is set if the
177block is uncompressed.  A block will be uncompressed if the -noI option is set,
178or if the compressed block was larger than the uncompressed block.
179
180Inodes are packed into the metadata blocks, and are not aligned to block
181boundaries, therefore inodes overlap compressed blocks.  Inodes are identified
182by a 48-bit number which encodes the location of the compressed metadata block
183containing the inode, and the byte offset into that block where the inode is
184placed (<block, offset>).
185
186To maximise compression there are different inodes for each file type
187(regular file, directory, device, etc.), the inode contents and length
188varying with the type.
189
190To further maximise compression, two types of regular file inode and
191directory inode are defined: inodes optimised for frequently occurring
192regular files and directories, and extended types where extra
193information has to be stored.
194
1953.3 Directories
196---------------
197
198Like inodes, directories are packed into compressed metadata blocks, stored
199in a directory table.  Directories are accessed using the start address of
200the metablock containing the directory and the offset into the
201decompressed block (<block, offset>).
202
203Directories are organised in a slightly complex way, and are not simply
204a list of file names.  The organisation takes advantage of the
205fact that (in most cases) the inodes of the files will be in the same
206compressed metadata block, and therefore, can share the start block.
207Directories are therefore organised in a two level list, a directory
208header containing the shared start block value, and a sequence of directory
209entries, each of which share the shared start block.  A new directory header
210is written once/if the inode start block changes.  The directory
211header/directory entry list is repeated as many times as necessary.
212
213Directories are sorted, and can contain a directory index to speed up
214file lookup.  Directory indexes store one entry per metablock, each entry
215storing the index/filename mapping to the first directory header
216in each metadata block.  Directories are sorted in alphabetical order,
217and at lookup the index is scanned linearly looking for the first filename
218alphabetically larger than the filename being looked up.  At this point the
219location of the metadata block the filename is in has been found.
220The general idea of the index is to ensure only one metadata block needs to be
221decompressed to do a lookup irrespective of the length of the directory.
222This scheme has the advantage that it doesn't require extra memory overhead
223and doesn't require much extra storage on disk.
224
2253.4 File data
226-------------
227
228Regular files consist of a sequence of contiguous compressed blocks, and/or a
229compressed fragment block (tail-end packed block).   The compressed size
230of each datablock is stored in a block list contained within the
231file inode.
232
233To speed up access to datablocks when reading 'large' files (256 Mbytes or
234larger), the code implements an index cache that caches the mapping from
235block index to datablock location on disk.
236
237The index cache allows Squashfs to handle large files (up to 1.75 TiB) while
238retaining a simple and space-efficient block list on disk.  The cache
239is split into slots, caching up to eight 224 GiB files (128 KiB blocks).
240Larger files use multiple slots, with 1.75 TiB files using all 8 slots.
241The index cache is designed to be memory efficient, and by default uses
24216 KiB.
243
2443.5 Fragment lookup table
245-------------------------
246
247Regular files can contain a fragment index which is mapped to a fragment
248location on disk and compressed size using a fragment lookup table.  This
249fragment lookup table is itself stored compressed into metadata blocks.
250A second index table is used to locate these.  This second index table for
251speed of access (and because it is small) is read at mount time and cached
252in memory.
253
2543.6 Uid/gid lookup table
255------------------------
256
257For space efficiency regular files store uid and gid indexes, which are
258converted to 32-bit uids/gids using an id look up table.  This table is
259stored compressed into metadata blocks.  A second index table is used to
260locate these.  This second index table for speed of access (and because it
261is small) is read at mount time and cached in memory.
262
2633.7 Export table
264----------------
265
266To enable Squashfs filesystems to be exportable (via NFS etc.) filesystems
267can optionally (disabled with the -no-exports Mksquashfs option) contain
268an inode number to inode disk location lookup table.  This is required to
269enable Squashfs to map inode numbers passed in filehandles to the inode
270location on disk, which is necessary when the export code reinstantiates
271expired/flushed inodes.
272
273This table is stored compressed into metadata blocks.  A second index table is
274used to locate these.  This second index table for speed of access (and because
275it is small) is read at mount time and cached in memory.
276
2773.8 Xattr table
278---------------
279
280The xattr table contains extended attributes for each inode.  The xattrs
281for each inode are stored in a list, each list entry containing a type,
282name and value field.  The type field encodes the xattr prefix
283("user.", "trusted." etc) and it also encodes how the name/value fields
284should be interpreted.  Currently the type indicates whether the value
285is stored inline (in which case the value field contains the xattr value),
286or if it is stored out of line (in which case the value field stores a
287reference to where the actual value is stored).  This allows large values
288to be stored out of line improving scanning and lookup performance and it
289also allows values to be de-duplicated, the value being stored once, and
290all other occurrences holding an out of line reference to that value.
291
292The xattr lists are packed into compressed 8K metadata blocks.
293To reduce overhead in inodes, rather than storing the on-disk
294location of the xattr list inside each inode, a 32-bit xattr id
295is stored.  This xattr id is mapped into the location of the xattr
296list using a second xattr id lookup table.
297
2984. TODOs and Outstanding Issues
299-------------------------------
300
3014.1 TODO list
302-------------
303
304Implement ACL support.
305
3064.2 Squashfs Internal Cache
307---------------------------
308
309Blocks in Squashfs are compressed.  To avoid repeatedly decompressing
310recently accessed data Squashfs uses two small metadata and fragment caches.
311
312The cache is not used for file datablocks, these are decompressed and cached in
313the page-cache in the normal way.  The cache is used to temporarily cache
314fragment and metadata blocks which have been read as a result of a metadata
315(i.e. inode or directory) or fragment access.  Because metadata and fragments
316are packed together into blocks (to gain greater compression) the read of a
317particular piece of metadata or fragment will retrieve other metadata/fragments
318which have been packed with it, these because of locality-of-reference may be
319read in the near future. Temporarily caching them ensures they are available
320for near future access without requiring an additional read and decompress.
321
322In the future this internal cache may be replaced with an implementation which
323uses the kernel page cache.  Because the page cache operates on page sized
324units this may introduce additional complexity in terms of locking and
325associated race conditions.
326