1 .. SPDX-License-Identifier: GPL-2.0
3 ======================================
4 EROFS - Enhanced Read-Only File System
5 ======================================
10 EROFS filesystem stands for Enhanced Read-Only File System. It aims to form a
11 generic read-only filesystem solution for various read-only use cases instead
12 of just focusing on storage space saving without considering any side effects
13 of runtime performance.
15 It is designed to meet the needs of flexibility, feature extendability and user
16 payload friendly, etc. Apart from those, it is still kept as a simple
17 random-access friendly high-performance filesystem to get rid of unneeded I/O
18 amplification and memory-resident overhead compared to similar approaches.
20 It is implemented to be a better choice for the following scenarios:
22 - read-only storage media or
24 - part of a fully trusted read-only solution, which means it needs to be
25 immutable and bit-for-bit identical to the official golden image for
26 their releases due to security or other considerations and
28 - hope to minimize extra storage space with guaranteed end-to-end performance
29 by using compact layout, transparent file compression and direct access,
30 especially for those embedded devices with limited memory and high-density
31 hosts with numerous containers.
33 Here is the main features of EROFS:
35 - Little endian on-disk design;
37 - 4KiB block size and 32-bit block addresses, therefore 16TiB address space
40 - Two inode layouts for different requirements:
42 ===================== ============ ======================================
43 compact (v1) extended (v2)
44 ===================== ============ ======================================
45 Inode metadata size 32 bytes 64 bytes
46 Max file size 4 GiB 16 EiB (also limited by max. vol size)
47 Max uids/gids 65536 4294967296
48 Per-inode timestamp no yes (64 + 32-bit timestamp)
49 Max hardlinks 65536 4294967296
50 Metadata reserved 8 bytes 18 bytes
51 ===================== ============ ======================================
53 - Metadata and data could be mixed as an option;
55 - Support extended attributes (xattrs) as an option;
57 - Support tailpacking data and xattr inline compared to byte-addressed
58 unaligned metadata or smaller block size alternatives;
60 - Support POSIX.1e ACLs by using xattrs;
62 - Support transparent data compression as an option:
63 LZ4 and MicroLZMA algorithms can be used on a per-file basis; In addition,
64 inplace decompression is also supported to avoid bounce compressed buffers
65 and page cache thrashing.
67 - Support direct I/O on uncompressed files to avoid double caching for loop
70 - Support FSDAX on uncompressed images for secure containers and ramdisks in
71 order to get rid of unnecessary page cache.
73 - Support multiple devices for multi blob container images;
75 - Support file-based on-demand loading with the Fscache infrastructure.
77 The following git tree provides the file system user-space tools under
78 development, such as a formatting tool (mkfs.erofs), an on-disk consistency &
79 compatibility checking tool (fsck.erofs), and a debugging tool (dump.erofs):
81 - git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs-utils.git
83 Bugs and patches are welcome, please kindly help us and send to the following
84 linux-erofs mailing list:
86 - linux-erofs mailing list <linux-erofs@lists.ozlabs.org>
91 =================== =========================================================
92 (no)user_xattr Setup Extended User Attributes. Note: xattr is enabled
93 by default if CONFIG_EROFS_FS_XATTR is selected.
94 (no)acl Setup POSIX Access Control List. Note: acl is enabled
95 by default if CONFIG_EROFS_FS_POSIX_ACL is selected.
96 cache_strategy=%s Select a strategy for cached decompression from now on:
98 ========== =============================================
99 disabled In-place I/O decompression only;
100 readahead Cache the last incomplete compressed physical
101 cluster for further reading. It still does
102 in-place I/O decompression for the rest
103 compressed physical clusters;
104 readaround Cache the both ends of incomplete compressed
105 physical clusters for further reading.
106 It still does in-place I/O decompression
107 for the rest compressed physical clusters.
108 ========== =============================================
109 dax={always,never} Use direct access (no page cache). See
110 Documentation/filesystems/dax.rst.
111 dax A legacy option which is an alias for ``dax=always``.
112 device=%s Specify a path to an extra device to be used together.
113 fsid=%s Specify a filesystem image ID for Fscache back-end.
114 =================== =========================================================
119 Information about mounted erofs file systems can be found in /sys/fs/erofs.
120 Each mounted filesystem will have a directory in /sys/fs/erofs based on its
121 device name (i.e., /sys/fs/erofs/sda).
122 (see also Documentation/ABI/testing/sysfs-fs-erofs)
129 Different from other read-only file systems, an EROFS volume is designed
130 to be as simple as possible::
132 |-> aligned with the block size
133 ____________________________________________________________
134 | |SB| | ... | Metadata | ... | Data | Metadata | ... | Data |
135 |_|__|_|_____|__________|_____|______|__________|_____|______|
138 All data areas should be aligned with the block size, but metadata areas
139 may not. All metadatas can be now observed in two different spaces (views):
141 1. Inode metadata space
143 Each valid inode should be aligned with an inode slot, which is a fixed
144 value (32 bytes) and designed to be kept in line with compact inode size.
146 Each inode can be directly found with the following formula:
147 inode offset = meta_blkaddr * block_size + 32 * nid
153 + meta_blkaddr blocks |-> another slot
154 _____________________________________________________________________
155 | ... | inode | xattrs | extents | data inline | ... | inode ...
156 |________|_______|(optional)|(optional)|__(optional)_|_____|__________
157 |-> aligned with the inode slot size
164 .____________________________________________________|-> aligned with 4B
165 | xattr_ibody_header | shared xattrs | inline xattrs |
166 |____________________|_______________|_______________|
167 |-> 12 bytes <-|->x * 4 bytes<-| .
171 ._______________________________.______________________.
172 | id | id | id | id | ... | id | ent | ... | ent| ... |
173 |____|____|____|____|______|____|_____|_____|____|_____|
177 Inode could be 32 or 64 bytes, which can be distinguished from a common
178 field which all inode versions have -- i_format::
180 __________________ __________________
181 | i_format | | i_format |
182 |__________________| |__________________|
185 |__________________| 32 bytes | |
187 |__________________| 64 bytes
189 Xattrs, extents, data inline are followed by the corresponding inode with
190 proper alignment, and they could be optional for different data mappings.
191 _currently_ total 5 data layouts are supported:
193 == ====================================================================
194 0 flat file data without data inline (no extent);
195 1 fixed-sized output data compression (with non-compacted indexes);
196 2 flat file data with tail packing data inline (no extent);
197 3 fixed-sized output data compression (with compacted indexes, v5.3+);
198 4 chunk-based file (v5.15+).
199 == ====================================================================
201 The size of the optional xattrs is indicated by i_xattr_count in inode
202 header. Large xattrs or xattrs shared by many different files can be
203 stored in shared xattrs metadata rather than inlined right after inode.
205 2. Shared xattrs metadata space
207 Shared xattrs space is similar to the above inode space, started with
208 a specific block indicated by xattr_blkaddr, organized one by one with
211 Each share xattr can also be directly found by the following formula:
212 xattr offset = xattr_blkaddr * block_size + 4 * xattr_id
216 |-> aligned by 4 bytes
217 + xattr_blkaddr blocks |-> aligned with 4 bytes
218 _________________________________________________________________________
219 | ... | xattr_entry | xattr data | ... | xattr_entry | xattr data ...
220 |________|_____________|_____________|_____|______________|_______________
224 All directories are now organized in a compact on-disk format. Note that
225 each directory block is divided into index and name areas in order to support
226 random file lookup, and all directory entries are _strictly_ recorded in
227 alphabetical order in order to support improved prefix binary search
228 algorithm (could refer to the related source code).
232 ___________________________
234 / ______________|________________
235 / / | nameoff1 | nameoffN-1
236 ____________.______________._______________v________________v__________
237 | dirent | dirent | ... | dirent | filename | filename | ... | filename |
238 |___.0___|____1___|_____|___N-1__|____0_____|____1_____|_____|___N-1____|
242 \________________________| nameoff0
245 Note that apart from the offset of the first filename, nameoff0 also indicates
246 the total number of directory entries in this block since it is no need to
247 introduce another on-disk field at all.
251 In order to support chunk-based data deduplication, a new inode data layout has
252 been supported since Linux v5.15: Files are split in equal-sized data chunks
253 with ``extents`` area of the inode metadata indicating how to get the chunk
254 data: these can be simply as a 4-byte block address array or in the 8-byte
255 chunk index form (see struct erofs_inode_chunk_index in erofs_fs.h for more
258 By the way, chunk-based files are all uncompressed for now.
262 EROFS implements LZ4 fixed-sized output compression which generates fixed-sized
263 compressed data blocks from variable-sized input in contrast to other existing
264 fixed-sized input solutions. Relatively higher compression ratios can be gotten
265 by using fixed-sized output compression since nowadays popular data compression
266 algorithms are mostly LZ77-based and such fixed-sized output approach can be
267 benefited from the historical dictionary (aka. sliding window).
269 In details, original (uncompressed) data is turned into several variable-sized
270 extents and in the meanwhile, compressed into physical clusters (pclusters).
271 In order to record each variable-sized extent, logical clusters (lclusters) are
272 introduced as the basic unit of compress indexes to indicate whether a new
273 extent is generated within the range (HEAD) or not (NONHEAD). Lclusters are now
274 fixed in block size, as illustrated below::
276 |<- variable-sized extent ->|<- VLE ->|
277 clusterofs clusterofs clusterofs
279 _________v_________________________________v_______________________v________
280 ... | . | | . | | . ...
281 ____|____._________|______________|________.___ _|______________|__.________
282 |-> lcluster <-|-> lcluster <-|-> lcluster <-|-> lcluster <-|
283 (HEAD) (NONHEAD) (HEAD) (NONHEAD) .
287 _______._____________________________.______________._________________
289 _______|______________|______________|______________|_________________
290 |-> big pcluster <-|-> pcluster <-|
292 A physical cluster can be seen as a container of physical compressed blocks
293 which contains compressed data. Previously, only lcluster-sized (4KB) pclusters
294 were supported. After big pcluster feature is introduced (available since
295 Linux v5.13), pcluster can be a multiple of lcluster size.
297 For each HEAD lcluster, clusterofs is recorded to indicate where a new extent
298 starts and blkaddr is used to seek the compressed data. For each NONHEAD
299 lcluster, delta0 and delta1 are available instead of blkaddr to indicate the
300 distance to its HEAD lcluster and the next HEAD lcluster. A PLAIN lcluster is
301 also a HEAD lcluster except that its data is uncompressed. See the comments
302 around "struct z_erofs_vle_decompressed_index" in erofs_fs.h for more details.
304 If big pcluster is enabled, pcluster size in lclusters needs to be recorded as
305 well. Let the delta0 of the first NONHEAD lcluster store the compressed block
306 count with a special flag as a new called CBLKCNT NONHEAD lcluster. It's easy
307 to understand its delta0 is constantly 1, as illustrated below::
309 __________________________________________________________
310 | HEAD | NONHEAD | NONHEAD | ... | NONHEAD | HEAD | HEAD |
311 |__:___|_(CBLKCNT)_|_________|_____|_________|__:___|____:_|
312 |<----- a big pcluster (with CBLKCNT) ------>|<-- -->|
313 a lcluster-sized pcluster (without CBLKCNT) ^
315 If another HEAD follows a HEAD lcluster, there is no room to record CBLKCNT,
316 but it's easy to know the size of such pcluster is 1 lcluster as well.