Library:LZO
Appearance
| Developer | Markus F.X.J. Oberhumer |
|---|---|
| Platforms | any (portable ANSI C) |
| Versions | 2.10 (1 Mar 2017) is the latest release in the mirror used |
| License | GPL-2.0-or-later (commercial license available) |
| Website | http://www.oberhumer.com/opensource/lzo/ |
| On this wiki | Facts · Code · Notes · Topics |
Markus Oberhumer's LZO real-time compression library (C, GPL-2.0-or-later or commercial) and its miniLZO subset: raw LZO1X blocks have no magic; the lzop file wrapper starts with 89 4C 5A 4F 00 0D 0A 1A 0A; the library stores a version string and date.
Overview
LZO is a family of very fast LZ77 compressors (LZO1, LZO1A, LZO1B, LZO1C, LZO1F, LZO1X, LZO1Y, LZO1Z, LZO2A). LZO1X-1 is the common one: in the Linux kernel (zram, squashfs, UBIFS, btrfs), game save files, UPX-like packers, and many embedded firmwares. miniLZO is a small LZO1X-only subset in the files minilzo.c/minilzo.h that programs copy into their source tree.
Identification
- Header lzoconf.h defines
LZO_VERSION 0x20a0,LZO_VERSION_STRING "2.10"andLZO_VERSION_DATE "Mar 01 2017". lzo_version(), lzo_version_string() and lzo_version_date() return them, and the string is kept in the library binary, so a search for2.10next toMar 01 2017dates a linked copy. miniLZO has its ownMINILZO_VERSION(0x20a0 for 2.10). - Exported/linked names: lzo_init (checked via the macro
__lzo_init_v2with struct sizes), lzo1x_1_compress, lzo1x_decompress, lzo1x_decompress_safe, lzo1x_999_compress, lzo1x_optimize. - A raw LZO1X block has no magic and no length header. Its end is marked by the bytes
11 00 00(a match with distance 16384 and length 0: instruction 17 followed by two zero bytes). The first byte of an LZO1X block is a literal run or a value above 17, which is how a decoder starts. - Decompression code shape (used by the capa rule 'decompress data using LZO'): immediates 0xFF (length run 255), 0x1F (mask), 0x4000 subtraction (m_pos -= 0x4000), shifts by 2, and the 3-byte constant
0xFFFFFFFCmasking. - lzop file wrapper (the lzop tool, .lzo files): magic
89 4C 5A 4F 00 0D 0A 1A 0A(0x89 'LZO' NUL CR LF Ctrl-Z LF), then version fields, method, level, flags, mode, mtime, name, header checksum (Adler-32 or CRC-32), followed by blocks with big-endian uncompressed/compressed sizes and per-block checksums. The lzop magic is confirmed in the lzop program source (lzop-1.04.tar.gz from lzop.org,src/lzop.cline 627):static const unsigned char lzop_magic[9] = { 0x89, 0x4c, 0x5a, 0x4f, 0x00, 0x0d, 0x0a, 0x1a, 0x0a }. The lzo repository itself does not contain lzop.
Pitfalls
- LZO is GPL: closed-source programs that include it have usually bought a commercial license, or use a re-implementation (e.g. QuickLZ, LZ4, or the clean-room decoders in the Linux kernel and in lzo-rs). Re-implementations lack the version string.
- UCL, NRV and aPLib come from the same family of authors/ideas but have different formats; the capa rules separate 'decompress data using UCL' and 'aplib'.
Code & repositories
See the linked repository (a GitHub mirror of the 2.10 release).
- https://github.com/nemequ/lzo @
0083878c235a89ef96a009d1ff0b500f3a364e4b· license: GPL-2.0
GitHub mirror of LZO 2.10 (lzoconf.h, minilzo)
Tested on: source read, not run: source only · checked 2026-10-18; last activity 2022-04-19
Open questions
- What does lzop write in its header for each lzop version (0x1030 etc.)? Not checked.