xref: /linux/Documentation/filesystems/romfs.rst (revision b77e0ce62d63a761ffb7f7245a215a49f5921c2f)
1.. SPDX-License-Identifier: GPL-2.0
2
3=======================
4ROMFS - ROM File System
5=======================
6
7This is a quite dumb, read only filesystem, mainly for initial RAM
8disks of installation disks.  It has grown up by the need of having
9modules linked at boot time.  Using this filesystem, you get a very
10similar feature, and even the possibility of a small kernel, with a
11file system which doesn't take up useful memory from the router
12functions in the basement of your office.
13
14For comparison, both the older minix and xiafs (the latter is now
15defunct) filesystems, compiled as module need more than 20000 bytes,
16while romfs is less than a page, about 4000 bytes (assuming i586
17code).  Under the same conditions, the msdos filesystem would need
18about 30K (and does not support device nodes or symlinks), while the
19nfs module with nfsroot is about 57K.  Furthermore, as a bit unfair
20comparison, an actual rescue disk used up 3202 blocks with ext2, while
21with romfs, it needed 3079 blocks.
22
23To create such a file system, you'll need a user program named
24genromfs. It is available on http://romfs.sourceforge.net/
25
26As the name suggests, romfs could be also used (space-efficiently) on
27various read-only media, like (E)EPROM disks if someone will have the
28motivation.. :)
29
30However, the main purpose of romfs is to have a very small kernel,
31which has only this filesystem linked in, and then can load any module
32later, with the current module utilities.  It can also be used to run
33some program to decide if you need SCSI devices, and even IDE or
34floppy drives can be loaded later if you use the "initrd"--initial
35RAM disk--feature of the kernel.  This would not be really news
36flash, but with romfs, you can even spare off your ext2 or minix or
37maybe even affs filesystem until you really know that you need it.
38
39For example, a distribution boot disk can contain only the cd disk
40drivers (and possibly the SCSI drivers), and the ISO 9660 filesystem
41module.  The kernel can be small enough, since it doesn't have other
42filesystems, like the quite large ext2fs module, which can then be
43loaded off the CD at a later stage of the installation.  Another use
44would be for a recovery disk, when you are reinstalling a workstation
45from the network, and you will have all the tools/modules available
46from a nearby server, so you don't want to carry two disks for this
47purpose, just because it won't fit into ext2.
48
49romfs operates on block devices as you can expect, and the underlying
50structure is very simple.  Every accessible structure begins on 16
51byte boundaries for fast access.  The minimum space a file will take
52is 32 bytes (this is an empty file, with a less than 16 character
53name).  The maximum overhead for any non-empty file is the header, and
54the 16 byte padding for the name and the contents, also 16+14+15 = 45
55bytes.  This is quite rare however, since most file names are longer
56than 3 bytes, and shorter than 15 bytes.
57
58The layout of the filesystem is the following::
59
60 offset	    content
61
62	+---+---+---+---+
63  0	| - | r | o | m |  \
64	+---+---+---+---+	The ASCII representation of those bytes
65  4	| 1 | f | s | - |  /	(i.e. "-rom1fs-")
66	+---+---+---+---+
67  8	|   full size	|	The number of accessible bytes in this fs.
68	+---+---+---+---+
69 12	|    checksum	|	The checksum of the FIRST 512 BYTES.
70	+---+---+---+---+
71 16	| volume name	|	The zero terminated name of the volume,
72	:               :	padded to 16 byte boundary.
73	+---+---+---+---+
74 xx	|     file	|
75	:    headers	:
76
77Every multi byte value (32 bit words, I'll use the longwords term from
78now on) must be in big endian order.
79
80The first eight bytes identify the filesystem, even for the casual
81inspector.  After that, in the 3rd longword, it contains the number of
82bytes accessible from the start of this filesystem.  The 4th longword
83is the checksum of the first 512 bytes (or the number of bytes
84accessible, whichever is smaller).  The applied algorithm is the same
85as in the AFFS filesystem, namely a simple sum of the longwords
86(assuming bigendian quantities again).  For details, please consult
87the source.  This algorithm was chosen because although it's not quite
88reliable, it does not require any tables, and it is very simple.
89
90The following bytes are now part of the file system; each file header
91must begin on a 16 byte boundary::
92
93 offset	    content
94
95     	+---+---+---+---+
96  0	| next filehdr|X|	The offset of the next file header
97	+---+---+---+---+	  (zero if no more files)
98  4	|   spec.info	|	Info for directories/hard links/devices
99	+---+---+---+---+
100  8	|     size      |	The size of this file in bytes
101	+---+---+---+---+
102 12	|   checksum	|	Covering the meta data, including the file
103	+---+---+---+---+	  name, and padding
104 16	| file name     |	The zero terminated name of the file,
105	:               :	padded to 16 byte boundary
106	+---+---+---+---+
107 xx	| file data	|
108	:		:
109
110Since the file headers begin always at a 16 byte boundary, the lowest
1114 bits would be always zero in the next filehdr pointer.  These four
112bits are used for the mode information.  Bits 0..2 specify the type of
113the file; while bit 4 shows if the file is executable or not.  The
114permissions are assumed to be world readable, if this bit is not set,
115and world executable if it is; except the character and block devices,
116they are never accessible for other than owner.  The owner of every
117file is user and group 0, this should never be a problem for the
118intended use.  The mapping of the 8 possible values to file types is
119the following:
120
121==	=============== ============================================
122	  mapping		spec.info means
123==	=============== ============================================
124 0	hard link	link destination [file header]
125 1	directory	first file's header
126 2	regular file	unused, must be zero [MBZ]
127 3	symbolic link	unused, MBZ (file data is the link content)
128 4	block device	16/16 bits major/minor number
129 5	char device		    - " -
130 6	socket		unused, MBZ
131 7	fifo		unused, MBZ
132==	=============== ============================================
133
134Note that hard links are specifically marked in this filesystem, but
135they will behave as you can expect (i.e. share the inode number).
136Note also that it is your responsibility to not create hard link
137loops, and creating all the . and .. links for directories.  This is
138normally done correctly by the genromfs program.  Please refrain from
139using the executable bits for special purposes on the socket and fifo
140special files, they may have other uses in the future.  Additionally,
141please remember that only regular files, and symlinks are supposed to
142have a nonzero size field; they contain the number of bytes available
143directly after the (padded) file name.
144
145Another thing to note is that romfs works on file headers and data
146aligned to 16 byte boundaries, but most hardware devices and the block
147device drivers are unable to cope with smaller than block-sized data.
148To overcome this limitation, the whole size of the file system must be
149padded to an 1024 byte boundary.
150
151If you have any problems or suggestions concerning this file system,
152please contact me.  However, think twice before wanting me to add
153features and code, because the primary and most important advantage of
154this file system is the small code.  On the other hand, don't be
155alarmed, I'm not getting that much romfs related mail.  Now I can
156understand why Avery wrote poems in the ARCnet docs to get some more
157feedback. :)
158
159romfs has also a mailing list, and to date, it hasn't received any
160traffic, so you are welcome to join it to discuss your ideas. :)
161
162It's run by ezmlm, so you can subscribe to it by sending a message
163to romfs-subscribe@shadow.banki.hu, the content is irrelevant.
164
165Pending issues:
166
167- Permissions and owner information are pretty essential features of a
168  Un*x like system, but romfs does not provide the full possibilities.
169  I have never found this limiting, but others might.
170
171- The file system is read only, so it can be very small, but in case
172  one would want to write _anything_ to a file system, he still needs
173  a writable file system, thus negating the size advantages.  Possible
174  solutions: implement write access as a compile-time option, or a new,
175  similarly small writable filesystem for RAM disks.
176
177- Since the files are only required to have alignment on a 16 byte
178  boundary, it is currently possibly suboptimal to read or execute files
179  from the filesystem.  It might be resolved by reordering file data to
180  have most of it (i.e. except the start and the end) laying at "natural"
181  boundaries, thus it would be possible to directly map a big portion of
182  the file contents to the mm subsystem.
183
184- Compression might be an useful feature, but memory is quite a
185  limiting factor in my eyes.
186
187- Where it is used?
188
189- Does it work on other architectures than intel and motorola?
190
191
192Have fun,
193
194Janos Farkas <chexum@shadow.banki.hu>
195