1 Notes on Filesystem Layout
2 --------------------------
4 These notes describe what mkcramfs generates. Kernel requirements are
5 a bit looser, e.g. it doesn't care if the <file_data> items are
6 swapped around (though it does care that directory entries (inodes) in
7 a given directory are contiguous, as this is used by readdir).
9 All data is currently in host-endian format; neither mkcramfs nor the
10 kernel ever do swabbing. (See section `Block Size' below.)
17 <superblock>: struct cramfs_super (see cramfs_fs.h).
19 <directory_structure>:
21 struct cramfs_inode (see cramfs_fs.h).
22 Filename. Not generally null-terminated, but it is
23 null-padded to a multiple of 4 bytes.
25 The order of inode traversal is described as "width-first" (not to be
26 confused with breadth-first); i.e. like depth-first but listing all of
27 a directory's entries before recursing down its subdirectories: the
28 same order as `ls -AUR' (but without the /^\..*:$/ directory header
29 lines); put another way, the same order as `find -type d -exec
32 Beginning in 2.4.7, directory entries are sorted. This optimization
33 allows cramfs_lookup to return more quickly when a filename does not
34 exist, speeds up user-space directory sorts, etc.
37 One <file_data> for each file that's either a symlink or a
38 regular file of non-zero st_size.
41 nblocks * <block_pointer>
42 (where nblocks = (st_size - 1) / blksize + 1)
44 padding to multiple of 4 bytes
46 The i'th <block_pointer> for a file stores the byte offset of the
47 *end* of the i'th <block> (i.e. one past the last byte, which is the
48 same as the start of the (i+1)'th <block> if there is one). The first
49 <block> immediately follows the last <block_pointer> for the file.
50 <block_pointer>s are each 32 bits long.
52 When the CRAMFS_FLAG_EXT_BLOCK_POINTERS capability bit is set, each
53 <block_pointer>'s top bits may contain special flags as follows:
55 CRAMFS_BLK_FLAG_UNCOMPRESSED (bit 31):
56 The block data is not compressed and should be copied verbatim.
58 CRAMFS_BLK_FLAG_DIRECT_PTR (bit 30):
59 The <block_pointer> stores the actual block start offset and not
60 its end, shifted right by 2 bits. The block must therefore be
61 aligned to a 4-byte boundary. The block size is either blksize
62 if CRAMFS_BLK_FLAG_UNCOMPRESSED is also specified, otherwise
63 the compressed data length is included in the first 2 bytes of
64 the block data. This is used to allow discontiguous data layout
65 and specific data block alignments e.g. for XIP applications.
68 The order of <file_data>'s is a depth-first descent of the directory
69 tree, i.e. the same order as `find -size +0 \( -type f -o -type l \)
73 <block>: The i'th <block> is the output of zlib's compress function
74 applied to the i'th blksize-sized chunk of the input data if the
75 corresponding CRAMFS_BLK_FLAG_UNCOMPRESSED <block_ptr> bit is not set,
76 otherwise it is the input data directly.
77 (For the last <block> of the file, the input may of course be smaller.)
78 Each <block> may be a different size. (See <block_pointer> above.)
80 <block>s are merely byte-aligned, not generally u32-aligned.
82 When CRAMFS_BLK_FLAG_DIRECT_PTR is specified then the corresponding
83 <block> may be located anywhere and not necessarily contiguous with
84 the previous/next blocks. In that case it is minimally u32-aligned.
85 If CRAMFS_BLK_FLAG_UNCOMPRESSED is also specified then the size is always
86 blksize except for the last block which is limited by the file length.
87 If CRAMFS_BLK_FLAG_DIRECT_PTR is set and CRAMFS_BLK_FLAG_UNCOMPRESSED
88 is not set then the first 2 bytes of the block contains the size of the
89 remaining block data as this cannot be determined from the placement of
90 logically adjacent blocks.
96 This kernel supports cramfs holes (i.e. [efficient representation of]
97 blocks in uncompressed data consisting entirely of NUL bytes), but by
98 default mkcramfs doesn't test for & create holes, since cramfs in
99 kernels up to at least 2.3.39 didn't support holes. Run mkcramfs
100 with -z if you want it to create files that can have holes in them.
106 The cramfs user-space tools, including mkcramfs and cramfsck, are
107 located at <http://sourceforge.net/projects/cramfs/>.
116 (Block size in cramfs refers to the size of input data that is
117 compressed at a time. It's intended to be somewhere around
118 PAGE_SIZE for cramfs_readpage's convenience.)
120 The superblock ought to indicate the block size that the fs was
121 written for, since comments in <linux/pagemap.h> indicate that
122 PAGE_SIZE may grow in future (if I interpret the comment
125 Currently, mkcramfs #define's PAGE_SIZE as 4096 and uses that
126 for blksize, whereas Linux-2.3.39 uses its PAGE_SIZE, which in
127 turn is defined as PAGE_SIZE (which can be as large as 32KB on arm).
128 This discrepancy is a bug, though it's not clear which should be
131 One option is to change mkcramfs to take its PAGE_SIZE from
132 <asm/page.h>. Personally I don't like this option, but it does
133 require the least amount of change: just change `#define
134 PAGE_SIZE (4096)' to `#include <asm/page.h>'. The disadvantage
135 is that the generated cramfs cannot always be shared between different
136 kernels, not even necessarily kernels of the same architecture if
137 PAGE_SIZE is subject to change between kernel versions
138 (currently possible with arm and ia64).
140 The remaining options try to make cramfs more sharable.
142 One part of that is addressing endianness. The two options here are
143 `always use little-endian' (like ext2fs) or `writer chooses
144 endianness; kernel adapts at runtime'. Little-endian wins because of
145 code simplicity and little CPU overhead even on big-endian machines.
147 The cost of swabbing is changing the code to use the le32_to_cpu
148 etc. macros as used by ext2fs. We don't need to swab the compressed
149 data, only the superblock, inodes and block pointers.
152 The other part of making cramfs more sharable is choosing a block
153 size. The options are:
155 1. Always 4096 bytes.
157 2. Writer chooses blocksize; kernel adapts but rejects blocksize >
160 3. Writer chooses blocksize; kernel adapts even to blocksize >
163 It's easy enough to change the kernel to use a smaller value than
164 PAGE_SIZE: just make cramfs_readpage read multiple blocks.
166 The cost of option 1 is that kernels with a larger PAGE_SIZE
167 value don't get as good compression as they can.
169 The cost of option 2 relative to option 1 is that the code uses
170 variables instead of #define'd constants. The gain is that people
171 with kernels having larger PAGE_SIZE can make use of that if
172 they don't mind their cramfs being inaccessible to kernels with
173 smaller PAGE_SIZE values.
175 Option 3 is easy to implement if we don't mind being CPU-inefficient:
176 e.g. get readpage to decompress to a buffer of size MAX_BLKSIZE (which
177 must be no larger than 32KB) and discard what it doesn't need.
178 Getting readpage to read into all the covered pages is harder.
180 The main advantage of option 3 over 1, 2, is better compression. The
181 cost is greater complexity. Probably not worth it, but I hope someone
182 will disagree. (If it is implemented, then I'll re-use that code in
186 Another cost of 2 and 3 over 1 is making mkcramfs use a different
187 block size, but that just means adding and parsing a -b option.
193 Given that cramfs will probably be used for CDs etc. as well as just
194 silicon ROMs, it might make sense to expand the inode a little from
195 its current 12 bytes. Inodes other than the root inode are followed
196 by filename, so the expansion doesn't even have to be a multiple of 4