2XZ data compression in Linux
   7    XZ is a general purpose data compression format with high compression
   8    ratio and relatively fast decompression. The primary compression
   9    algorithm (filter) is LZMA2. Additional filters can be used to improve
  10    compression ratio even further. E.g. Branch/Call/Jump (BCJ) filters
  11    improve compression ratio of executable data.
  13    The XZ decompressor in Linux is called XZ Embedded. It supports
  14    the LZMA2 filter and optionally also BCJ filters. CRC32 is supported
  15    for integrity checking. The home page of XZ Embedded is at
  16    <>, where you can find the
  17    latest version and also information about using the code outside
  18    the Linux kernel.
  20    For userspace, XZ Utils provide a zlib-like compression library
  21    and a gzip-like command line tool. XZ Utils can be downloaded from
  22    <>.
  24XZ related components in the kernel
  26    The xz_dec module provides XZ decompressor with single-call (buffer
  27    to buffer) and multi-call (stateful) APIs. The usage of the xz_dec
  28    module is documented in include/linux/xz.h.
  30    The xz_dec_test module is for testing xz_dec. xz_dec_test is not
  31    useful unless you are hacking the XZ decompressor. xz_dec_test
  32    allocates a char device major dynamically to which one can write
  33    .xz files from userspace. The decompressed output is thrown away.
  34    Keep an eye on dmesg to see diagnostics printed by xz_dec_test.
  35    See the xz_dec_test source code for the details.
  37    For decompressing the kernel image, initramfs, and initrd, there
  38    is a wrapper function in lib/decompress_unxz.c. Its API is the
  39    same as in other decompress_*.c files, which is defined in
  40    include/linux/decompress/generic.h.
  42    scripts/ is a wrapper for the xz command line tool found
  43    from XZ Utils. The wrapper sets compression options to values suitable
  44    for compressing the kernel image.
  46    For kernel makefiles, two commands are provided for use with
  47    $(call if_needed). The kernel image should be compressed with
  48    $(call if_needed,xzkern) which will use a BCJ filter and a big LZMA2
  49    dictionary. It will also append a four-byte trailer containing the
  50    uncompressed size of the file, which is needed by the boot code.
  51    Other things should be compressed with $(call if_needed,xzmisc)
  52    which will use no BCJ filter and 1 MiB LZMA2 dictionary.
  54Notes on compression options
  56    Since the XZ Embedded supports only streams with no integrity check or
  57    CRC32, make sure that you don't use some other integrity check type
  58    when encoding files that are supposed to be decoded by the kernel. With
  59    liblzma, you need to use either LZMA_CHECK_NONE or LZMA_CHECK_CRC32
  60    when encoding. With the xz command line tool, use --check=none or
  61    --check=crc32.
  63    Using CRC32 is strongly recommended unless there is some other layer
  64    which will verify the integrity of the uncompressed data anyway.
  65    Double checking the integrity would probably be waste of CPU cycles.
  66    Note that the headers will always have a CRC32 which will be validated
  67    by the decoder; you can only change the integrity check type (or
  68    disable it) for the actual uncompressed data.
  70    In userspace, LZMA2 is typically used with dictionary sizes of several
  71    megabytes. The decoder needs to have the dictionary in RAM, thus big
  72    dictionaries cannot be used for files that are intended to be decoded
  73    by the kernel. 1 MiB is probably the maximum reasonable dictionary
  74    size for in-kernel use (maybe more is OK for initramfs). The presets
  75    in XZ Utils may not be optimal when creating files for the kernel,
  76    so don't hesitate to use custom settings. Example:
  78        xz --check=crc32 --lzma2=dict=512KiB inputfile
  80    An exception to above dictionary size limitation is when the decoder
  81    is used in single-call mode. Decompressing the kernel itself is an
  82    example of this situation. In single-call mode, the memory usage
  83    doesn't depend on the dictionary size, and it is perfectly fine to
  84    use a big dictionary: for maximum compression, the dictionary should
  85    be at least as big as the uncompressed data itself.
  87Future plans
  89    Creating a limited XZ encoder may be considered if people think it is
  90    useful. LZMA2 is slower to compress than e.g. Deflate or LZO even at
  91    the fastest settings, so it isn't clear if LZMA2 encoder is wanted
  92    into the kernel.
  94    Support for limited random-access reading is planned for the
  95    decompression code. I don't know if it could have any use in the
  96    kernel, but I know that it would be useful in some embedded projects
  97    outside the Linux kernel.
  99Conformance to the .xz file format specification
 101    There are a couple of corner cases where things have been simplified
 102    at expense of detecting errors as early as possible. These should not
 103    matter in practice all, since they don't cause security issues. But
 104    it is good to know this if testing the code e.g. with the test files
 105    from XZ Utils.
 107Reporting bugs
 109    Before reporting a bug, please check that it's not fixed already
 110    at upstream. See <> to get the
 111    latest code.
 113    Report bugs to <> or visit #tukaani on
 114    Freenode and talk to Larhzu. I don't actively read LKML or other
 115    kernel-related mailing lists, so if there's something I should know,
 116    you should email to me personally or use IRC.
 118    Don't bother Igor Pavlov with questions about the XZ implementation
 119    in the kernel or about XZ Utils. While these two implementations
 120    include essential code that is directly based on Igor Pavlov's code,
 121    these implementations aren't maintained nor supported by him.