2 * LZ4 - Fast LZ compression algorithm
3 * Copyright (C) 2011 - 2016, Yann Collet.
4 * BSD 2 - Clause License (http://www.opensource.org/licenses/bsd - license.php)
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are
8 * * Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above
11 * copyright notice, this list of conditions and the following disclaimer
12 * in the documentation and/or other materials provided with the
14 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
15 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
16 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
17 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
18 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
19 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
20 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
24 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 * You can contact the author at :
26 * - LZ4 homepage : http://www.lz4.org
27 * - LZ4 source repository : https://github.com/lz4/lz4
29 * Changed for kernel usage by:
30 * Sven Schmidt <4sschmid@informatik.uni-hamburg.de>
33 /*-************************************
35 **************************************/
36 #include <linux/lz4.h>
38 #include <linux/init.h>
39 #include <linux/module.h>
40 #include <linux/kernel.h>
41 #include <asm/unaligned.h>
43 /*-*****************************
44 * Decompression functions
45 *******************************/
47 #define DEBUGLOG(l, ...) {} /* disabled */
50 #define assert(condition) ((void)0)
54 * LZ4_decompress_generic() :
55 * This generic decompression function covers all use cases.
56 * It shall be instantiated several times, using different sets of directives.
57 * Note that it is important for performance that this function really get inlined,
58 * in order to remove useless branches during compilation optimization.
60 static FORCE_INLINE int LZ4_decompress_generic(
61 const char * const src,
65 * If endOnInput == endOnInputSize,
66 * this value is `dstCapacity`
69 /* endOnOutputSize, endOnInputSize */
70 endCondition_directive endOnInput,
72 earlyEnd_directive partialDecoding,
73 /* noDict, withPrefix64k, usingExtDict */
75 /* always <= dst, == dst when no prefix */
76 const BYTE * const lowPrefix,
77 /* only if dict == usingExtDict */
78 const BYTE * const dictStart,
79 /* note : = 0 if noDict */
83 const BYTE *ip = (const BYTE *) src;
84 const BYTE * const iend = ip + srcSize;
86 BYTE *op = (BYTE *) dst;
87 BYTE * const oend = op + outputSize;
90 const BYTE * const dictEnd = (const BYTE *)dictStart + dictSize;
91 static const unsigned int inc32table[8] = {0, 1, 2, 1, 0, 4, 4, 4};
92 static const int dec64table[8] = {0, 0, 0, -1, -4, 1, 2, 3};
94 const int safeDecode = (endOnInput == endOnInputSize);
95 const int checkOffset = ((safeDecode) && (dictSize < (int)(64 * KB)));
97 /* Set up the "end" pointers for the shortcut. */
98 const BYTE *const shortiend = iend -
99 (endOnInput ? 14 : 8) /*maxLL*/ - 2 /*offset*/;
100 const BYTE *const shortoend = oend -
101 (endOnInput ? 14 : 8) /*maxLL*/ - 18 /*maxML*/;
103 DEBUGLOG(5, "%s (srcSize:%i, dstSize:%i)", __func__,
104 srcSize, outputSize);
107 assert(lowPrefix <= op);
110 /* Empty output buffer */
111 if ((endOnInput) && (unlikely(outputSize == 0)))
112 return ((srcSize == 1) && (*ip == 0)) ? 0 : -1;
114 if ((!endOnInput) && (unlikely(outputSize == 0)))
115 return (*ip == 0 ? 1 : -1);
117 if ((endOnInput) && unlikely(srcSize == 0))
120 /* Main Loop : decode sequences */
126 /* get literal length */
127 unsigned int const token = *ip++;
128 length = token>>ML_BITS;
130 /* ip < iend before the increment */
131 assert(!endOnInput || ip <= iend);
134 * A two-stage shortcut for the most common case:
135 * 1) If the literal length is 0..14, and there is enough
136 * space, enter the shortcut and copy 16 bytes on behalf
137 * of the literals (in the fast mode, only 8 bytes can be
138 * safely copied this way).
139 * 2) Further if the match length is 4..18, copy 18 bytes
140 * in a similar manner; but we ensure that there's enough
141 * space in the output for those 18 bytes earlier, upon
142 * entering the shortcut (in other words, there is a
143 * combined check for both stages).
145 * The & in the likely() below is intentionally not && so that
146 * some compilers can produce better parallelized runtime code
148 if ((endOnInput ? length != RUN_MASK : length <= 8)
150 * strictly "less than" on input, to re-enter
151 * the loop with at least one byte
153 && likely((endOnInput ? ip < shortiend : 1) &
154 (op <= shortoend))) {
155 /* Copy the literals */
156 LZ4_memcpy(op, ip, endOnInput ? 16 : 8);
157 op += length; ip += length;
161 * prepare for match copying, decode full info.
162 * If it doesn't work out, the info won't be wasted.
164 length = token & ML_MASK; /* match length */
165 offset = LZ4_readLE16(ip);
168 assert(match <= op); /* check overflow */
170 /* Do not deal with overlapping matches. */
171 if ((length != ML_MASK) &&
173 (dict == withPrefix64k || match >= lowPrefix)) {
174 /* Copy the match. */
175 LZ4_memcpy(op + 0, match + 0, 8);
176 LZ4_memcpy(op + 8, match + 8, 8);
177 LZ4_memcpy(op + 16, match + 16, 2);
178 op += length + MINMATCH;
179 /* Both stages worked, load the next token. */
184 * The second stage didn't work out, but the info
185 * is ready. Propel it right to the point of match
191 /* decode literal length */
192 if (length == RUN_MASK) {
195 if (unlikely(endOnInput ? ip >= iend - RUN_MASK : 0)) {
196 /* overflow detection */
202 } while (likely(endOnInput
203 ? ip < iend - RUN_MASK
207 && unlikely((uptrval)(op) +
208 length < (uptrval)(op))) {
209 /* overflow detection */
213 && unlikely((uptrval)(ip) +
214 length < (uptrval)(ip))) {
215 /* overflow detection */
222 LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
224 if (((endOnInput) && ((cpy > oend - MFLIMIT)
225 || (ip + length > iend - (2 + 1 + LASTLITERALS))))
226 || ((!endOnInput) && (cpy > oend - WILDCOPYLENGTH))) {
227 if (partialDecoding) {
231 * stop in the middle of literal segment
237 && (ip + length > iend)) {
240 * read attempt beyond
241 * end of input buffer
250 * block decoding must
256 && ((ip + length != iend)
260 * input must be consumed
266 LZ4_memcpy(op, ip, length);
270 /* Necessarily EOF, due to parsing restrictions */
271 if (!partialDecoding || (cpy == oend))
274 /* may overwrite up to WILDCOPYLENGTH beyond cpy */
275 LZ4_wildCopy(op, ip, cpy);
281 offset = LZ4_readLE16(ip);
285 /* get matchlength */
286 length = token & ML_MASK;
289 if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) {
290 /* Error : offset outside buffers */
294 /* costs ~1%; silence an msan warning when offset == 0 */
296 * note : when partialDecoding, there is no guarantee that
297 * at least 4 bytes remain available in output buffer
299 if (!partialDecoding) {
301 assert(oend - op >= 4);
303 LZ4_write32(op, (U32)offset);
306 if (length == ML_MASK) {
312 if ((endOnInput) && (ip > iend - LASTLITERALS))
320 (uptrval)(op) + length < (uptrval)op)) {
321 /* overflow detection */
328 /* match starting within external dictionary */
329 if ((dict == usingExtDict) && (match < lowPrefix)) {
330 if (unlikely(op + length > oend - LASTLITERALS)) {
331 /* doesn't respect parsing restriction */
332 if (!partialDecoding)
334 length = min(length, (size_t)(oend - op));
337 if (length <= (size_t)(lowPrefix - match)) {
339 * match fits entirely within external
340 * dictionary : just copy
342 memmove(op, dictEnd - (lowPrefix - match),
347 * match stretches into both external
348 * dictionary and current block
350 size_t const copySize = (size_t)(lowPrefix - match);
351 size_t const restSize = length - copySize;
353 LZ4_memcpy(op, dictEnd - copySize, copySize);
355 if (restSize > (size_t)(op - lowPrefix)) {
357 BYTE * const endOfMatch = op + restSize;
358 const BYTE *copyFrom = lowPrefix;
360 while (op < endOfMatch)
363 LZ4_memcpy(op, lowPrefix, restSize);
370 /* copy match within block */
375 * may not respect endBlock parsing restrictions
378 if (partialDecoding &&
379 (cpy > oend - MATCH_SAFEGUARD_DISTANCE)) {
380 size_t const mlen = min(length, (size_t)(oend - op));
381 const BYTE * const matchEnd = match + mlen;
382 BYTE * const copyEnd = op + mlen;
389 LZ4_memcpy(op, match, mlen);
397 if (unlikely(offset < 8)) {
402 match += inc32table[offset];
403 LZ4_memcpy(op + 4, match, 4);
404 match -= dec64table[offset];
406 LZ4_copy8(op, match);
412 if (unlikely(cpy > oend - MATCH_SAFEGUARD_DISTANCE)) {
413 BYTE * const oCopyLimit = oend - (WILDCOPYLENGTH - 1);
415 if (cpy > oend - LASTLITERALS) {
417 * Error : last LASTLITERALS bytes
418 * must be literals (uncompressed)
423 if (op < oCopyLimit) {
424 LZ4_wildCopy(op, match, oCopyLimit);
425 match += oCopyLimit - op;
431 LZ4_copy8(op, match);
433 LZ4_wildCopy(op + 8, match + 8, cpy);
435 op = cpy; /* wildcopy correction */
438 /* end of decoding */
440 /* Nb of output bytes decoded */
441 return (int) (((char *)op) - dst);
443 /* Nb of input bytes read */
444 return (int) (((const char *)ip) - src);
447 /* Overflow error detected */
449 return (int) (-(((const char *)ip) - src)) - 1;
452 int LZ4_decompress_safe(const char *source, char *dest,
453 int compressedSize, int maxDecompressedSize)
455 return LZ4_decompress_generic(source, dest,
456 compressedSize, maxDecompressedSize,
457 endOnInputSize, decode_full_block,
458 noDict, (BYTE *)dest, NULL, 0);
461 int LZ4_decompress_safe_partial(const char *src, char *dst,
462 int compressedSize, int targetOutputSize, int dstCapacity)
464 dstCapacity = min(targetOutputSize, dstCapacity);
465 return LZ4_decompress_generic(src, dst, compressedSize, dstCapacity,
466 endOnInputSize, partial_decode,
467 noDict, (BYTE *)dst, NULL, 0);
470 int LZ4_decompress_fast(const char *source, char *dest, int originalSize)
472 return LZ4_decompress_generic(source, dest, 0, originalSize,
473 endOnOutputSize, decode_full_block,
475 (BYTE *)dest - 64 * KB, NULL, 0);
478 /* ===== Instantiate a few more decoding cases, used more than once. ===== */
480 int LZ4_decompress_safe_withPrefix64k(const char *source, char *dest,
481 int compressedSize, int maxOutputSize)
483 return LZ4_decompress_generic(source, dest,
484 compressedSize, maxOutputSize,
485 endOnInputSize, decode_full_block,
487 (BYTE *)dest - 64 * KB, NULL, 0);
490 static int LZ4_decompress_safe_withSmallPrefix(const char *source, char *dest,
495 return LZ4_decompress_generic(source, dest,
496 compressedSize, maxOutputSize,
497 endOnInputSize, decode_full_block,
499 (BYTE *)dest - prefixSize, NULL, 0);
502 int LZ4_decompress_safe_forceExtDict(const char *source, char *dest,
503 int compressedSize, int maxOutputSize,
504 const void *dictStart, size_t dictSize)
506 return LZ4_decompress_generic(source, dest,
507 compressedSize, maxOutputSize,
508 endOnInputSize, decode_full_block,
509 usingExtDict, (BYTE *)dest,
510 (const BYTE *)dictStart, dictSize);
513 static int LZ4_decompress_fast_extDict(const char *source, char *dest,
515 const void *dictStart, size_t dictSize)
517 return LZ4_decompress_generic(source, dest,
519 endOnOutputSize, decode_full_block,
520 usingExtDict, (BYTE *)dest,
521 (const BYTE *)dictStart, dictSize);
525 * The "double dictionary" mode, for use with e.g. ring buffers: the first part
526 * of the dictionary is passed as prefix, and the second via dictStart + dictSize.
527 * These routines are used only once, in LZ4_decompress_*_continue().
530 int LZ4_decompress_safe_doubleDict(const char *source, char *dest,
531 int compressedSize, int maxOutputSize,
533 const void *dictStart, size_t dictSize)
535 return LZ4_decompress_generic(source, dest,
536 compressedSize, maxOutputSize,
537 endOnInputSize, decode_full_block,
538 usingExtDict, (BYTE *)dest - prefixSize,
539 (const BYTE *)dictStart, dictSize);
543 int LZ4_decompress_fast_doubleDict(const char *source, char *dest,
544 int originalSize, size_t prefixSize,
545 const void *dictStart, size_t dictSize)
547 return LZ4_decompress_generic(source, dest,
549 endOnOutputSize, decode_full_block,
550 usingExtDict, (BYTE *)dest - prefixSize,
551 (const BYTE *)dictStart, dictSize);
554 /* ===== streaming decompression functions ===== */
556 int LZ4_setStreamDecode(LZ4_streamDecode_t *LZ4_streamDecode,
557 const char *dictionary, int dictSize)
559 LZ4_streamDecode_t_internal *lz4sd =
560 &LZ4_streamDecode->internal_donotuse;
562 lz4sd->prefixSize = (size_t) dictSize;
563 lz4sd->prefixEnd = (const BYTE *) dictionary + dictSize;
564 lz4sd->externalDict = NULL;
565 lz4sd->extDictSize = 0;
571 * These decoding functions allow decompression of multiple blocks
572 * in "streaming" mode.
573 * Previously decoded blocks must still be available at the memory
574 * position where they were decoded.
575 * If it's not possible, save the relevant part of
576 * decoded data into a safe buffer,
577 * and indicate where it stands using LZ4_setStreamDecode()
579 int LZ4_decompress_safe_continue(LZ4_streamDecode_t *LZ4_streamDecode,
580 const char *source, char *dest, int compressedSize, int maxOutputSize)
582 LZ4_streamDecode_t_internal *lz4sd =
583 &LZ4_streamDecode->internal_donotuse;
586 if (lz4sd->prefixSize == 0) {
587 /* The first call, no dictionary yet. */
588 assert(lz4sd->extDictSize == 0);
589 result = LZ4_decompress_safe(source, dest,
590 compressedSize, maxOutputSize);
593 lz4sd->prefixSize = result;
594 lz4sd->prefixEnd = (BYTE *)dest + result;
595 } else if (lz4sd->prefixEnd == (BYTE *)dest) {
596 /* They're rolling the current segment. */
597 if (lz4sd->prefixSize >= 64 * KB - 1)
598 result = LZ4_decompress_safe_withPrefix64k(source, dest,
599 compressedSize, maxOutputSize);
600 else if (lz4sd->extDictSize == 0)
601 result = LZ4_decompress_safe_withSmallPrefix(source,
602 dest, compressedSize, maxOutputSize,
605 result = LZ4_decompress_safe_doubleDict(source, dest,
606 compressedSize, maxOutputSize,
608 lz4sd->externalDict, lz4sd->extDictSize);
611 lz4sd->prefixSize += result;
612 lz4sd->prefixEnd += result;
615 * The buffer wraps around, or they're
616 * switching to another buffer.
618 lz4sd->extDictSize = lz4sd->prefixSize;
619 lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
620 result = LZ4_decompress_safe_forceExtDict(source, dest,
621 compressedSize, maxOutputSize,
622 lz4sd->externalDict, lz4sd->extDictSize);
625 lz4sd->prefixSize = result;
626 lz4sd->prefixEnd = (BYTE *)dest + result;
632 int LZ4_decompress_fast_continue(LZ4_streamDecode_t *LZ4_streamDecode,
633 const char *source, char *dest, int originalSize)
635 LZ4_streamDecode_t_internal *lz4sd = &LZ4_streamDecode->internal_donotuse;
638 if (lz4sd->prefixSize == 0) {
639 assert(lz4sd->extDictSize == 0);
640 result = LZ4_decompress_fast(source, dest, originalSize);
643 lz4sd->prefixSize = originalSize;
644 lz4sd->prefixEnd = (BYTE *)dest + originalSize;
645 } else if (lz4sd->prefixEnd == (BYTE *)dest) {
646 if (lz4sd->prefixSize >= 64 * KB - 1 ||
647 lz4sd->extDictSize == 0)
648 result = LZ4_decompress_fast(source, dest,
651 result = LZ4_decompress_fast_doubleDict(source, dest,
652 originalSize, lz4sd->prefixSize,
653 lz4sd->externalDict, lz4sd->extDictSize);
656 lz4sd->prefixSize += originalSize;
657 lz4sd->prefixEnd += originalSize;
659 lz4sd->extDictSize = lz4sd->prefixSize;
660 lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
661 result = LZ4_decompress_fast_extDict(source, dest,
662 originalSize, lz4sd->externalDict, lz4sd->extDictSize);
665 lz4sd->prefixSize = originalSize;
666 lz4sd->prefixEnd = (BYTE *)dest + originalSize;
671 int LZ4_decompress_safe_usingDict(const char *source, char *dest,
672 int compressedSize, int maxOutputSize,
673 const char *dictStart, int dictSize)
676 return LZ4_decompress_safe(source, dest,
677 compressedSize, maxOutputSize);
678 if (dictStart+dictSize == dest) {
679 if (dictSize >= 64 * KB - 1)
680 return LZ4_decompress_safe_withPrefix64k(source, dest,
681 compressedSize, maxOutputSize);
682 return LZ4_decompress_safe_withSmallPrefix(source, dest,
683 compressedSize, maxOutputSize, dictSize);
685 return LZ4_decompress_safe_forceExtDict(source, dest,
686 compressedSize, maxOutputSize, dictStart, dictSize);
689 int LZ4_decompress_fast_usingDict(const char *source, char *dest,
691 const char *dictStart, int dictSize)
693 if (dictSize == 0 || dictStart + dictSize == dest)
694 return LZ4_decompress_fast(source, dest, originalSize);
696 return LZ4_decompress_fast_extDict(source, dest, originalSize,
697 dictStart, dictSize);
701 EXPORT_SYMBOL(LZ4_decompress_safe);
702 EXPORT_SYMBOL(LZ4_decompress_safe_partial);
703 EXPORT_SYMBOL(LZ4_decompress_fast);
704 EXPORT_SYMBOL(LZ4_setStreamDecode);
705 EXPORT_SYMBOL(LZ4_decompress_safe_continue);
706 EXPORT_SYMBOL(LZ4_decompress_fast_continue);
707 EXPORT_SYMBOL(LZ4_decompress_safe_usingDict);
708 EXPORT_SYMBOL(LZ4_decompress_fast_usingDict);
710 MODULE_LICENSE("Dual BSD/GPL");
711 MODULE_DESCRIPTION("LZ4 decompressor");