Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | initramfs buffer format |
| 2 | ----------------------- |
| 3 | |
| 4 | Al Viro, H. Peter Anvin |
| 5 | Last revision: 2002-01-13 |
| 6 | |
| 7 | Starting with kernel 2.5.x, the old "initial ramdisk" protocol is |
| 8 | getting {replaced/complemented} with the new "initial ramfs" |
| 9 | (initramfs) protocol. The initramfs contents is passed using the same |
| 10 | memory buffer protocol used by the initrd protocol, but the contents |
| 11 | is different. The initramfs buffer contains an archive which is |
| 12 | expanded into a ramfs filesystem; this document details the format of |
| 13 | the initramfs buffer format. |
| 14 | |
| 15 | The initramfs buffer format is based around the "newc" or "crc" CPIO |
| 16 | formats, and can be created with the cpio(1) utility. The cpio |
| 17 | archive can be compressed using gzip(1). One valid version of an |
| 18 | initramfs buffer is thus a single .cpio.gz file. |
| 19 | |
| 20 | The full format of the initramfs buffer is defined by the following |
| 21 | grammar, where: |
| 22 | * is used to indicate "0 or more occurrences of" |
| 23 | (|) indicates alternatives |
| 24 | + indicates concatenation |
| 25 | GZIP() indicates the gzip(1) of the operand |
| 26 | ALGN(n) means padding with null bytes to an n-byte boundary |
| 27 | |
| 28 | initramfs := ("\0" | cpio_archive | cpio_gzip_archive)* |
| 29 | |
| 30 | cpio_gzip_archive := GZIP(cpio_archive) |
| 31 | |
| 32 | cpio_archive := cpio_file* + (<nothing> | cpio_trailer) |
| 33 | |
| 34 | cpio_file := ALGN(4) + cpio_header + filename + "\0" + ALGN(4) + data |
| 35 | |
| 36 | cpio_trailer := ALGN(4) + cpio_header + "TRAILER!!!\0" + ALGN(4) |
| 37 | |
| 38 | |
| 39 | In human terms, the initramfs buffer contains a collection of |
| 40 | compressed and/or uncompressed cpio archives (in the "newc" or "crc" |
| 41 | formats); arbitrary amounts zero bytes (for padding) can be added |
| 42 | between members. |
| 43 | |
| 44 | The cpio "TRAILER!!!" entry (cpio end-of-archive) is optional, but is |
| 45 | not ignored; see "handling of hard links" below. |
| 46 | |
| 47 | The structure of the cpio_header is as follows (all fields contain |
| 48 | hexadecimal ASCII numbers fully padded with '0' on the left to the |
| 49 | full width of the field, for example, the integer 4780 is represented |
| 50 | by the ASCII string "000012ac"): |
| 51 | |
| 52 | Field name Field size Meaning |
| 53 | c_magic 6 bytes The string "070701" or "070702" |
| 54 | c_ino 8 bytes File inode number |
| 55 | c_mode 8 bytes File mode and permissions |
| 56 | c_uid 8 bytes File uid |
| 57 | c_gid 8 bytes File gid |
| 58 | c_nlink 8 bytes Number of links |
| 59 | c_mtime 8 bytes Modification time |
| 60 | c_filesize 8 bytes Size of data field |
| 61 | c_maj 8 bytes Major part of file device number |
| 62 | c_min 8 bytes Minor part of file device number |
| 63 | c_rmaj 8 bytes Major part of device node reference |
| 64 | c_rmin 8 bytes Minor part of device node reference |
| 65 | c_namesize 8 bytes Length of filename, including final \0 |
| 66 | c_chksum 8 bytes Checksum of data field if c_magic is 070702; |
| 67 | otherwise zero |
| 68 | |
| 69 | The c_mode field matches the contents of st_mode returned by stat(2) |
| 70 | on Linux, and encodes the file type and file permissions. |
| 71 | |
| 72 | The c_filesize should be zero for any file which is not a regular file |
| 73 | or symlink. |
| 74 | |
| 75 | The c_chksum field contains a simple 32-bit unsigned sum of all the |
| 76 | bytes in the data field. cpio(1) refers to this as "crc", which is |
| 77 | clearly incorrect (a cyclic redundancy check is a different and |
| 78 | significantly stronger integrity check), however, this is the |
| 79 | algorithm used. |
| 80 | |
| 81 | If the filename is "TRAILER!!!" this is actually an end-of-archive |
| 82 | marker; the c_filesize for an end-of-archive marker must be zero. |
| 83 | |
| 84 | |
| 85 | *** Handling of hard links |
| 86 | |
| 87 | When a nondirectory with c_nlink > 1 is seen, the (c_maj,c_min,c_ino) |
| 88 | tuple is looked up in a tuple buffer. If not found, it is entered in |
| 89 | the tuple buffer and the entry is created as usual; if found, a hard |
| 90 | link rather than a second copy of the file is created. It is not |
| 91 | necessary (but permitted) to include a second copy of the file |
| 92 | contents; if the file contents is not included, the c_filesize field |
| 93 | should be set to zero to indicate no data section follows. If data is |
| 94 | present, the previous instance of the file is overwritten; this allows |
| 95 | the data-carrying instance of a file to occur anywhere in the sequence |
| 96 | (GNU cpio is reported to attach the data to the last instance of a |
| 97 | file only.) |
| 98 | |
| 99 | c_filesize must not be zero for a symlink. |
| 100 | |
| 101 | When a "TRAILER!!!" end-of-archive marker is seen, the tuple buffer is |
| 102 | reset. This permits archives which are generated independently to be |
| 103 | concatenated. |
| 104 | |
| 105 | To combine file data from different sources (without having to |
| 106 | regenerate the (c_maj,c_min,c_ino) fields), therefore, either one of |
| 107 | the following techniques can be used: |
| 108 | |
| 109 | a) Separate the different file data sources with a "TRAILER!!!" |
| 110 | end-of-archive marker, or |
| 111 | |
| 112 | b) Make sure c_nlink == 1 for all nondirectory entries. |