[mirror_ubuntu-jammy-kernel.git] / Documentation / filesystems / erofs.rst

.. SPDX-License-Identifier: GPL-2.0

======================================
Enhanced Read-Only File System - EROFS
======================================

Overview
========

EROFS file-system stands for Enhanced Read-Only File System. Different
from other read-only file systems, it aims to be designed for flexibility,
scalability, but be kept simple and high performance.

It is designed as a better filesystem solution for the following scenarios:

 - read-only storage media or

 - part of a fully trusted read-only solution, which means it needs to be
   immutable and bit-for-bit identical to the official golden image for
   their releases due to security and other considerations and

 - hope to save some extra storage space with guaranteed end-to-end performance
   by using reduced metadata and transparent file compression, especially
   for those embedded devices with limited memory (ex, smartphone);

Here is the main features of EROFS:

 - Little endian on-disk design;

 - Currently 4KB block size (nobh) and therefore maximum 16TB address space;

 - Metadata & data could be mixed by design;

 - 2 inode versions for different requirements:

   =====================  ============  =====================================
                          compact (v1)  extended (v2)
   =====================  ============  =====================================
   Inode metadata size    32 bytes      64 bytes
   Max file size          4 GB          16 EB (also limited by max. vol size)
   Max uids/gids          65536         4294967296
   File change time       no            yes (64 + 32-bit timestamp)
   Max hardlinks          65536         4294967296
   Metadata reserved      4 bytes       14 bytes
   =====================  ============  =====================================

 - Support extended attributes (xattrs) as an option;

 - Support xattr inline and tail-end data inline for all files;

 - Support POSIX.1e ACLs by using xattrs;

 - Support transparent data compression as an option:
   LZ4 algorithm with the fixed-sized output compression for high performance.

The following git tree provides the file system user-space tools under
development (ex, formatting tool mkfs.erofs):

- git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs-utils.git

Bugs and patches are welcome, please kindly help us and send to the following
linux-erofs mailing list:

- linux-erofs mailing list   <linux-erofs@lists.ozlabs.org>

Mount options
=============

===================    =========================================================
(no)user_xattr         Setup Extended User Attributes. Note: xattr is enabled
                       by default if CONFIG_EROFS_FS_XATTR is selected.
(no)acl                Setup POSIX Access Control List. Note: acl is enabled
                       by default if CONFIG_EROFS_FS_POSIX_ACL is selected.
cache_strategy=%s      Select a strategy for cached decompression from now on:

		       ==========  =============================================
                         disabled  In-place I/O decompression only;
                        readahead  Cache the last incomplete compressed physical
                                   cluster for further reading. It still does
                                   in-place I/O decompression for the rest
                                   compressed physical clusters;
                       readaround  Cache the both ends of incomplete compressed
                                   physical clusters for further reading.
                                   It still does in-place I/O decompression
                                   for the rest compressed physical clusters.
		       ==========  =============================================
dax={always,never}     Use direct access (no page cache).  See
                       Documentation/filesystems/dax.rst.
dax                    A legacy option which is an alias for ``dax=always``.
===================    =========================================================

On-disk details
===============

Summary
-------
Different from other read-only file systems, an EROFS volume is designed
to be as simple as possible::

                                |-> aligned with the block size
   ____________________________________________________________
  | |SB| | ... | Metadata | ... | Data | Metadata | ... | Data |
  |_|__|_|_____|__________|_____|______|__________|_____|______|
  0 +1K

All data areas should be aligned with the block size, but metadata areas
may not. All metadatas can be now observed in two different spaces (views):

 1. Inode metadata space

    Each valid inode should be aligned with an inode slot, which is a fixed
    value (32 bytes) and designed to be kept in line with compact inode size.

    Each inode can be directly found with the following formula:
         inode offset = meta_blkaddr * block_size + 32 * nid

    ::

                                 |-> aligned with 8B
                                            |-> followed closely
     + meta_blkaddr blocks                                      |-> another slot
       _____________________________________________________________________
     |  ...   | inode |  xattrs  | extents  | data inline | ... | inode ...
     |________|_______|(optional)|(optional)|__(optional)_|_____|__________
              |-> aligned with the inode slot size
                   .                   .
                 .                         .
               .                              .
             .                                    .
           .                                         .
         .                                              .
       .____________________________________________________|-> aligned with 4B
       | xattr_ibody_header | shared xattrs | inline xattrs |
       |____________________|_______________|_______________|
       |->    12 bytes    <-|->x * 4 bytes<-|               .
                           .                .                 .
                     .                      .                   .
                .                           .                     .
            ._______________________________.______________________.
            | id | id | id | id |  ... | id | ent | ... | ent| ... |
            |____|____|____|____|______|____|_____|_____|____|_____|
                                            |-> aligned with 4B
                                                        |-> aligned with 4B

    Inode could be 32 or 64 bytes, which can be distinguished from a common
    field which all inode versions have -- i_format::

        __________________               __________________
       |     i_format     |             |     i_format     |
       |__________________|             |__________________|
       |        ...       |             |        ...       |
       |                  |             |                  |
       |__________________| 32 bytes    |                  |
                                        |                  |
                                        |__________________| 64 bytes

    Xattrs, extents, data inline are followed by the corresponding inode with
    proper alignment, and they could be optional for different data mappings.
    _currently_ total 5 data layouts are supported:

    ==  ====================================================================
     0  flat file data without data inline (no extent);
     1  fixed-sized output data compression (with non-compacted indexes);
     2  flat file data with tail packing data inline (no extent);
     3  fixed-sized output data compression (with compacted indexes, v5.3+);
     4  chunk-based file (v5.15+).
    ==  ====================================================================

    The size of the optional xattrs is indicated by i_xattr_count in inode
    header. Large xattrs or xattrs shared by many different files can be
    stored in shared xattrs metadata rather than inlined right after inode.

 2. Shared xattrs metadata space

    Shared xattrs space is similar to the above inode space, started with
    a specific block indicated by xattr_blkaddr, organized one by one with
    proper align.

    Each share xattr can also be directly found by the following formula:
         xattr offset = xattr_blkaddr * block_size + 4 * xattr_id

::

                           |-> aligned by  4 bytes
    + xattr_blkaddr blocks                     |-> aligned with 4 bytes
     _________________________________________________________________________
    |  ...   | xattr_entry |  xattr data | ... |  xattr_entry | xattr data  ...
    |________|_____________|_____________|_____|______________|_______________

Directories
-----------
All directories are now organized in a compact on-disk format. Note that
each directory block is divided into index and name areas in order to support
random file lookup, and all directory entries are _strictly_ recorded in
alphabetical order in order to support improved prefix binary search
algorithm (could refer to the related source code).

::

                  ___________________________
                 /                           |
                /              ______________|________________
               /              /              | nameoff1       | nameoffN-1
  ____________.______________._______________v________________v__________
 | dirent | dirent | ... | dirent | filename | filename | ... | filename |
 |___.0___|____1___|_____|___N-1__|____0_____|____1_____|_____|___N-1____|
      \                           ^
       \                          |                           * could have
        \                         |                             trailing '\0'
         \________________________| nameoff0
                             Directory block

Note that apart from the offset of the first filename, nameoff0 also indicates
the total number of directory entries in this block since it is no need to
introduce another on-disk field at all.

Chunk-based file
----------------
In order to support chunk-based data deduplication, a new inode data layout has
been supported since Linux v5.15: Files are split in equal-sized data chunks
with ``extents`` area of the inode metadata indicating how to get the chunk
data: these can be simply as a 4-byte block address array or in the 8-byte
chunk index form (see struct erofs_inode_chunk_index in erofs_fs.h for more
details.)

By the way, chunk-based files are all uncompressed for now.

Data compression
----------------
EROFS implements LZ4 fixed-sized output compression which generates fixed-sized
compressed data blocks from variable-sized input in contrast to other existing
fixed-sized input solutions. Relatively higher compression ratios can be gotten
by using fixed-sized output compression since nowadays popular data compression
algorithms are mostly LZ77-based and such fixed-sized output approach can be
benefited from the historical dictionary (aka. sliding window).

In details, original (uncompressed) data is turned into several variable-sized
extents and in the meanwhile, compressed into physical clusters (pclusters).
In order to record each variable-sized extent, logical clusters (lclusters) are
introduced as the basic unit of compress indexes to indicate whether a new
extent is generated within the range (HEAD) or not (NONHEAD). Lclusters are now
fixed in block size, as illustrated below::

          |<-    variable-sized extent    ->|<-       VLE         ->|
        clusterofs                        clusterofs              clusterofs
          |                                 |                       |
 _________v_________________________________v_______________________v________
 ... |    .         |              |        .     |              |  .   ...
 ____|____._________|______________|________.___ _|______________|__.________
     |-> lcluster <-|-> lcluster <-|-> lcluster <-|-> lcluster <-|
          (HEAD)        (NONHEAD)       (HEAD)        (NONHEAD)    .
           .             CBLKCNT            .                    .
            .                               .                  .
             .                              .                .
       _______._____________________________.______________._________________
          ... |              |              |              | ...
       _______|______________|______________|______________|_________________
              |->      big pcluster       <-|-> pcluster <-|

A physical cluster can be seen as a container of physical compressed blocks
which contains compressed data. Previously, only lcluster-sized (4KB) pclusters
were supported. After big pcluster feature is introduced (available since
Linux v5.13), pcluster can be a multiple of lcluster size.

For each HEAD lcluster, clusterofs is recorded to indicate where a new extent
starts and blkaddr is used to seek the compressed data. For each NONHEAD
lcluster, delta0 and delta1 are available instead of blkaddr to indicate the
distance to its HEAD lcluster and the next HEAD lcluster. A PLAIN lcluster is
also a HEAD lcluster except that its data is uncompressed. See the comments
around "struct z_erofs_vle_decompressed_index" in erofs_fs.h for more details.

If big pcluster is enabled, pcluster size in lclusters needs to be recorded as
well. Let the delta0 of the first NONHEAD lcluster store the compressed block
count with a special flag as a new called CBLKCNT NONHEAD lcluster. It's easy
to understand its delta0 is constantly 1, as illustrated below::

   __________________________________________________________
  | HEAD |  NONHEAD  | NONHEAD | ... | NONHEAD | HEAD | HEAD |
  |__:___|_(CBLKCNT)_|_________|_____|_________|__:___|____:_|
     |<----- a big pcluster (with CBLKCNT) ------>|<--  -->|
           a lcluster-sized pcluster (without CBLKCNT) ^

If another HEAD follows a HEAD lcluster, there is no room to record CBLKCNT,
but it's easy to know the size of such pcluster is 1 lcluster as well.
Commit	Line	Data
e66d8631 MCC	1	.. SPDX-License-Identifier: GPL-2.0
	2
	3	======================================
	4	Enhanced Read-Only File System - EROFS
	5	======================================
	6
fdb05364 GX	7	Overview
	8	========
	9
	10	EROFS file-system stands for Enhanced Read-Only File System. Different
	11	from other read-only file systems, it aims to be designed for flexibility,
	12	scalability, but be kept simple and high performance.
	13
	14	It is designed as a better filesystem solution for the following scenarios:
e66d8631	15
fdb05364 GX	16	- read-only storage media or
	17
	18	- part of a fully trusted read-only solution, which means it needs to be
	19	immutable and bit-for-bit identical to the official golden image for
	20	their releases due to security and other considerations and
	21
	22	- hope to save some extra storage space with guaranteed end-to-end performance
	23	by using reduced metadata and transparent file compression, especially
	24	for those embedded devices with limited memory (ex, smartphone);
	25
	26	Here is the main features of EROFS:
e66d8631	27
fdb05364 GX	28	- Little endian on-disk design;
	29
	30	- Currently 4KB block size (nobh) and therefore maximum 16TB address space;
	31
	32	- Metadata & data could be mixed by design;
	33
	34	- 2 inode versions for different requirements:
e66d8631 MCC	35
e66d8631 MCC	36	===================== ============ =====================================
ffafde47	37	compact (v1) extended (v2)
e66d8631 MCC	38	===================== ============ =====================================
	39	Inode metadata size 32 bytes 64 bytes
	40	Max file size 4 GB 16 EB (also limited by max. vol size)
	41	Max uids/gids 65536 4294967296
	42	File change time no yes (64 + 32-bit timestamp)
	43	Max hardlinks 65536 4294967296
	44	Metadata reserved 4 bytes 14 bytes
	45	===================== ============ =====================================
fdb05364 GX	46
	47	- Support extended attributes (xattrs) as an option;
	48
	49	- Support xattr inline and tail-end data inline for all files;
	50
516c115c GX	51	- Support POSIX.1e ACLs by using xattrs;
516c115c GX	52
46f2e044 GX	53	- Support transparent data compression as an option:
46f2e044 GX	54	LZ4 algorithm with the fixed-sized output compression for high performance.
fdb05364 GX	55
	56	The following git tree provides the file system user-space tools under
	57	development (ex, formatting tool mkfs.erofs):
e66d8631 MCC	58
e66d8631 MCC	59	- git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs-utils.git
fdb05364 GX	60
	61	Bugs and patches are welcome, please kindly help us and send to the following
	62	linux-erofs mailing list:
e66d8631 MCC	63
e66d8631 MCC	64	- linux-erofs mailing list <linux-erofs@lists.ozlabs.org>
fdb05364	65
fdb05364 GX	66	Mount options
	67	=============
	68
e66d8631	69	=================== =========================================================
fdb05364 GX	70	(no)user_xattr Setup Extended User Attributes. Note: xattr is enabled
	71	by default if CONFIG_EROFS_FS_XATTR is selected.
	72	(no)acl Setup POSIX Access Control List. Note: acl is enabled
	73	by default if CONFIG_EROFS_FS_POSIX_ACL is selected.
4279f3f9	74	cache_strategy=%s Select a strategy for cached decompression from now on:
e66d8631 MCC	75
	76	========== =============================================
	77	disabled In-place I/O decompression only;
	78	readahead Cache the last incomplete compressed physical
4279f3f9 GX	79	cluster for further reading. It still does
	80	in-place I/O decompression for the rest
	81	compressed physical clusters;
e66d8631	82	readaround Cache the both ends of incomplete compressed
4279f3f9 GX	83	physical clusters for further reading.
	84	It still does in-place I/O decompression
	85	for the rest compressed physical clusters.
e66d8631	86	========== =============================================
06252e9c GX	87	dax={always,never} Use direct access (no page cache). See
	88	Documentation/filesystems/dax.rst.
	89	dax A legacy option which is an alias for ``dax=always``.
e66d8631	90	=================== =========================================================
fdb05364 GX	91
	92	On-disk details
	93	===============
	94
	95	Summary
	96	-------
	97	Different from other read-only file systems, an EROFS volume is designed
e66d8631	98	to be as simple as possible::
fdb05364 GX	99
	100	\|-> aligned with the block size
	101	____________________________________________________________
	102	\| \|SB\| \| ... \| Metadata \| ... \| Data \| Metadata \| ... \| Data \|
	103	\|_\|__\|_\|_____\|__________\|_____\|______\|__________\|_____\|______\|
	104	0 +1K
	105
	106	All data areas should be aligned with the block size, but metadata areas
	107	may not. All metadatas can be now observed in two different spaces (views):
e66d8631	108
fdb05364	109	1. Inode metadata space
e66d8631	110
fdb05364	111	Each valid inode should be aligned with an inode slot, which is a fixed
ffafde47	112	value (32 bytes) and designed to be kept in line with compact inode size.
fdb05364 GX	113
	114	Each inode can be directly found with the following formula:
	115	inode offset = meta_blkaddr * block_size + 32 * nid
	116
e66d8631 MCC	117	::
e66d8631 MCC	118
1b55767d GX	119	\|-> aligned with 8B
	120	\|-> followed closely
	121	+ meta_blkaddr blocks \|-> another slot
	122	_____________________________________________________________________
	123	\| ... \| inode \| xattrs \| extents \| data inline \| ... \| inode ...
	124	\|________\|_______\|(optional)\|(optional)\|__(optional)_\|_____\|__________
	125	\|-> aligned with the inode slot size
	126	. .
	127	. .
	128	. .
	129	. .
	130	. .
	131	. .
	132	.____________________________________________________\|-> aligned with 4B
	133	\| xattr_ibody_header \| shared xattrs \| inline xattrs \|
	134	\|____________________\|_______________\|_______________\|
	135	\|-> 12 bytes <-\|->x * 4 bytes<-\| .
	136	. . .
	137	. . .
	138	. . .
	139	._______________________________.______________________.
	140	\| id \| id \| id \| id \| ... \| id \| ent \| ... \| ent\| ... \|
	141	\|____\|____\|____\|____\|______\|____\|_____\|_____\|____\|_____\|
	142	\|-> aligned with 4B
	143	\|-> aligned with 4B
fdb05364 GX	144
fdb05364 GX	145	Inode could be 32 or 64 bytes, which can be distinguished from a common
e66d8631	146	field which all inode versions have -- i_format::
fdb05364 GX	147
fdb05364 GX	148	__________________ __________________
ffafde47	149	\| i_format \| \| i_format \|
fdb05364 GX	150	\|__________________\| \|__________________\|
	151	\| ... \| \| ... \|
	152	\| \| \| \|
	153	\|__________________\| 32 bytes \| \|
	154	\| \|
	155	\|__________________\| 64 bytes
	156
	157	Xattrs, extents, data inline are followed by the corresponding inode with
ffafde47	158	proper alignment, and they could be optional for different data mappings.
2a9dc7a8	159	_currently_ total 5 data layouts are supported:
fdb05364	160
e66d8631	161	== ====================================================================
ffafde47 GX	162	0 flat file data without data inline (no extent);
	163	1 fixed-sized output data compression (with non-compacted indexes);
	164	2 flat file data with tail packing data inline (no extent);
2a9dc7a8 GX	165	3 fixed-sized output data compression (with compacted indexes, v5.3+);
2a9dc7a8 GX	166	4 chunk-based file (v5.15+).
e66d8631	167	== ====================================================================
fdb05364 GX	168
	169	The size of the optional xattrs is indicated by i_xattr_count in inode
	170	header. Large xattrs or xattrs shared by many different files can be
	171	stored in shared xattrs metadata rather than inlined right after inode.
	172
	173	2. Shared xattrs metadata space
e66d8631	174
fdb05364 GX	175	Shared xattrs space is similar to the above inode space, started with
	176	a specific block indicated by xattr_blkaddr, organized one by one with
	177	proper align.
	178
	179	Each share xattr can also be directly found by the following formula:
	180	xattr offset = xattr_blkaddr * block_size + 4 * xattr_id
	181
1b55767d	182	::
e66d8631	183
1b55767d GX	184	\|-> aligned by 4 bytes
	185	+ xattr_blkaddr blocks \|-> aligned with 4 bytes
	186	_________________________________________________________________________
	187	\| ... \| xattr_entry \| xattr data \| ... \| xattr_entry \| xattr data ...
	188	\|________\|_____________\|_____________\|_____\|______________\|_______________
fdb05364 GX	189
	190	Directories
	191	-----------
	192	All directories are now organized in a compact on-disk format. Note that
	193	each directory block is divided into index and name areas in order to support
	194	random file lookup, and all directory entries are _strictly_ recorded in
	195	alphabetical order in order to support improved prefix binary search
	196	algorithm (could refer to the related source code).
	197
e66d8631 MCC	198	::
e66d8631 MCC	199
1b55767d GX	200	___________________________
	201	/ \|
	202	/ ______________\|________________
	203	/ / \| nameoff1 \| nameoffN-1
	204	____________.______________._______________v________________v__________
	205	\| dirent \| dirent \| ... \| dirent \| filename \| filename \| ... \| filename \|
	206	\|___.0___\|____1___\|_____\|___N-1__\|____0_____\|____1_____\|_____\|___N-1____\|
	207	\ ^
	208	\ \| * could have
	209	\ \| trailing '\0'
	210	\________________________\| nameoff0
	211	Directory block
fdb05364 GX	212
	213	Note that apart from the offset of the first filename, nameoff0 also indicates
	214	the total number of directory entries in this block since it is no need to
	215	introduce another on-disk field at all.
	216
2a9dc7a8 GX	217	Chunk-based file
	218	----------------
	219	In order to support chunk-based data deduplication, a new inode data layout has
	220	been supported since Linux v5.15: Files are split in equal-sized data chunks
	221	with ``extents`` area of the inode metadata indicating how to get the chunk
	222	data: these can be simply as a 4-byte block address array or in the 8-byte
	223	chunk index form (see struct erofs_inode_chunk_index in erofs_fs.h for more
	224	details.)
	225
	226	By the way, chunk-based files are all uncompressed for now.
	227
46f2e044 GX	228	Data compression
	229	----------------
	230	EROFS implements LZ4 fixed-sized output compression which generates fixed-sized
	231	compressed data blocks from variable-sized input in contrast to other existing
	232	fixed-sized input solutions. Relatively higher compression ratios can be gotten
	233	by using fixed-sized output compression since nowadays popular data compression
	234	algorithms are mostly LZ77-based and such fixed-sized output approach can be
	235	benefited from the historical dictionary (aka. sliding window).
	236
	237	In details, original (uncompressed) data is turned into several variable-sized
	238	extents and in the meanwhile, compressed into physical clusters (pclusters).
	239	In order to record each variable-sized extent, logical clusters (lclusters) are
	240	introduced as the basic unit of compress indexes to indicate whether a new
	241	extent is generated within the range (HEAD) or not (NONHEAD). Lclusters are now
	242	fixed in block size, as illustrated below::
e66d8631	243
1b55767d GX	244	\|<- variable-sized extent ->\|<- VLE ->\|
	245	clusterofs clusterofs clusterofs
	246	\| \| \|
	247	_________v_________________________________v_______________________v________
	248	... \| . \| \| . \| \| . ...
	249	____\|____._________\|______________\|________.___ _\|______________\|__.________
	250	\|-> lcluster <-\|-> lcluster <-\|-> lcluster <-\|-> lcluster <-\|
46f2e044 GX	251	(HEAD) (NONHEAD) (HEAD) (NONHEAD) .
	252	. CBLKCNT . .
	253	. . .
	254	. . .
	255	_______._____________________________.______________._________________
1b55767d GX	256	... \| \| \| \| ...
1b55767d GX	257	_______\|______________\|______________\|______________\|_________________
46f2e044 GX	258	\|-> big pcluster <-\|-> pcluster <-\|
	259
	260	A physical cluster can be seen as a container of physical compressed blocks
	261	which contains compressed data. Previously, only lcluster-sized (4KB) pclusters
	262	were supported. After big pcluster feature is introduced (available since
	263	Linux v5.13), pcluster can be a multiple of lcluster size.
	264
	265	For each HEAD lcluster, clusterofs is recorded to indicate where a new extent
	266	starts and blkaddr is used to seek the compressed data. For each NONHEAD
	267	lcluster, delta0 and delta1 are available instead of blkaddr to indicate the
	268	distance to its HEAD lcluster and the next HEAD lcluster. A PLAIN lcluster is
	269	also a HEAD lcluster except that its data is uncompressed. See the comments
	270	around "struct z_erofs_vle_decompressed_index" in erofs_fs.h for more details.
	271
	272	If big pcluster is enabled, pcluster size in lclusters needs to be recorded as
	273	well. Let the delta0 of the first NONHEAD lcluster store the compressed block
	274	count with a special flag as a new called CBLKCNT NONHEAD lcluster. It's easy
	275	to understand its delta0 is constantly 1, as illustrated below::
	276
	277	__________________________________________________________
	278	\| HEAD \| NONHEAD \| NONHEAD \| ... \| NONHEAD \| HEAD \| HEAD \|
	279	\|__:___\|_(CBLKCNT)_\|_________\|_____\|_________\|__:___\|____:_\|
	280	\|<----- a big pcluster (with CBLKCNT) ------>\|<-- -->\|
	281	a lcluster-sized pcluster (without CBLKCNT) ^
	282
	283	If another HEAD follows a HEAD lcluster, there is no room to record CBLKCNT,
	284	but it's easy to know the size of such pcluster is 1 lcluster as well.