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 if ((endOnInput ? length != RUN_MASK : length <= 8)
147 * strictly "less than" on input, to re-enter
148 * the loop with at least one byte
150 && likely((endOnInput ? ip < shortiend : 1) &
151 (op <= shortoend))) {
152 /* Copy the literals */
153 memcpy(op, ip, endOnInput ? 16 : 8);
154 op += length; ip += length;
158 * prepare for match copying, decode full info.
159 * If it doesn't work out, the info won't be wasted.
161 length = token & ML_MASK; /* match length */
162 offset = LZ4_readLE16(ip);
165 assert(match <= op); /* check overflow */
167 /* Do not deal with overlapping matches. */
168 if ((length != ML_MASK) &&
170 (dict == withPrefix64k || match >= lowPrefix)) {
171 /* Copy the match. */
172 memcpy(op + 0, match + 0, 8);
173 memcpy(op + 8, match + 8, 8);
174 memcpy(op + 16, match + 16, 2);
175 op += length + MINMATCH;
176 /* Both stages worked, load the next token. */
181 * The second stage didn't work out, but the info
182 * is ready. Propel it right to the point of match
188 /* decode literal length */
189 if (length == RUN_MASK) {
192 if (unlikely(endOnInput ? ip >= iend - RUN_MASK : 0)) {
193 /* overflow detection */
199 } while (likely(endOnInput
200 ? ip < iend - RUN_MASK
204 && unlikely((uptrval)(op) +
205 length < (uptrval)(op))) {
206 /* overflow detection */
210 && unlikely((uptrval)(ip) +
211 length < (uptrval)(ip))) {
212 /* overflow detection */
219 LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
221 if (((endOnInput) && ((cpy > oend - MFLIMIT)
222 || (ip + length > iend - (2 + 1 + LASTLITERALS))))
223 || ((!endOnInput) && (cpy > oend - WILDCOPYLENGTH))) {
224 if (partialDecoding) {
228 * stop in the middle of literal segment
234 && (ip + length > iend)) {
237 * read attempt beyond
238 * end of input buffer
247 * block decoding must
253 && ((ip + length != iend)
257 * input must be consumed
264 * supports overlapping memory regions; only matters
265 * for in-place decompression scenarios
267 LZ4_memmove(op, ip, length);
271 /* Necessarily EOF when !partialDecoding.
272 * When partialDecoding, it is EOF if we've either
273 * filled the output buffer or
274 * can't proceed with reading an offset for following match.
276 if (!partialDecoding || (cpy == oend) || (ip >= (iend - 2)))
279 /* may overwrite up to WILDCOPYLENGTH beyond cpy */
280 LZ4_wildCopy(op, ip, cpy);
286 offset = LZ4_readLE16(ip);
290 /* get matchlength */
291 length = token & ML_MASK;
294 if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) {
295 /* Error : offset outside buffers */
299 /* costs ~1%; silence an msan warning when offset == 0 */
301 * note : when partialDecoding, there is no guarantee that
302 * at least 4 bytes remain available in output buffer
304 if (!partialDecoding) {
306 assert(oend - op >= 4);
308 LZ4_write32(op, (U32)offset);
311 if (length == ML_MASK) {
317 if ((endOnInput) && (ip > iend - LASTLITERALS))
325 (uptrval)(op) + length < (uptrval)op)) {
326 /* overflow detection */
333 /* match starting within external dictionary */
334 if ((dict == usingExtDict) && (match < lowPrefix)) {
335 if (unlikely(op + length > oend - LASTLITERALS)) {
336 /* doesn't respect parsing restriction */
337 if (!partialDecoding)
339 length = min(length, (size_t)(oend - op));
342 if (length <= (size_t)(lowPrefix - match)) {
344 * match fits entirely within external
345 * dictionary : just copy
347 memmove(op, dictEnd - (lowPrefix - match),
352 * match stretches into both external
353 * dictionary and current block
355 size_t const copySize = (size_t)(lowPrefix - match);
356 size_t const restSize = length - copySize;
358 memcpy(op, dictEnd - copySize, copySize);
360 if (restSize > (size_t)(op - lowPrefix)) {
362 BYTE * const endOfMatch = op + restSize;
363 const BYTE *copyFrom = lowPrefix;
365 while (op < endOfMatch)
368 memcpy(op, lowPrefix, restSize);
375 /* copy match within block */
380 * may not respect endBlock parsing restrictions
383 if (partialDecoding &&
384 (cpy > oend - MATCH_SAFEGUARD_DISTANCE)) {
385 size_t const mlen = min(length, (size_t)(oend - op));
386 const BYTE * const matchEnd = match + mlen;
387 BYTE * const copyEnd = op + mlen;
394 memcpy(op, match, mlen);
402 if (unlikely(offset < 8)) {
407 match += inc32table[offset];
408 memcpy(op + 4, match, 4);
409 match -= dec64table[offset];
411 LZ4_copy8(op, match);
417 if (unlikely(cpy > oend - MATCH_SAFEGUARD_DISTANCE)) {
418 BYTE * const oCopyLimit = oend - (WILDCOPYLENGTH - 1);
420 if (cpy > oend - LASTLITERALS) {
422 * Error : last LASTLITERALS bytes
423 * must be literals (uncompressed)
428 if (op < oCopyLimit) {
429 LZ4_wildCopy(op, match, oCopyLimit);
430 match += oCopyLimit - op;
436 LZ4_copy8(op, match);
438 LZ4_wildCopy(op + 8, match + 8, cpy);
440 op = cpy; /* wildcopy correction */
443 /* end of decoding */
445 /* Nb of output bytes decoded */
446 return (int) (((char *)op) - dst);
448 /* Nb of input bytes read */
449 return (int) (((const char *)ip) - src);
452 /* Overflow error detected */
454 return (int) (-(((const char *)ip) - src)) - 1;
457 int LZ4_decompress_safe(const char *source, char *dest,
458 int compressedSize, int maxDecompressedSize)
460 return LZ4_decompress_generic(source, dest,
461 compressedSize, maxDecompressedSize,
462 endOnInputSize, decode_full_block,
463 noDict, (BYTE *)dest, NULL, 0);
466 int LZ4_decompress_safe_partial(const char *src, char *dst,
467 int compressedSize, int targetOutputSize, int dstCapacity)
469 dstCapacity = min(targetOutputSize, dstCapacity);
470 return LZ4_decompress_generic(src, dst, compressedSize, dstCapacity,
471 endOnInputSize, partial_decode,
472 noDict, (BYTE *)dst, NULL, 0);
475 int LZ4_decompress_fast(const char *source, char *dest, int originalSize)
477 return LZ4_decompress_generic(source, dest, 0, originalSize,
478 endOnOutputSize, decode_full_block,
480 (BYTE *)dest - 64 * KB, NULL, 0);
483 /* ===== Instantiate a few more decoding cases, used more than once. ===== */
485 int LZ4_decompress_safe_withPrefix64k(const char *source, char *dest,
486 int compressedSize, int maxOutputSize)
488 return LZ4_decompress_generic(source, dest,
489 compressedSize, maxOutputSize,
490 endOnInputSize, decode_full_block,
492 (BYTE *)dest - 64 * KB, NULL, 0);
495 static int LZ4_decompress_safe_withSmallPrefix(const char *source, char *dest,
500 return LZ4_decompress_generic(source, dest,
501 compressedSize, maxOutputSize,
502 endOnInputSize, decode_full_block,
504 (BYTE *)dest - prefixSize, NULL, 0);
507 int LZ4_decompress_safe_forceExtDict(const char *source, char *dest,
508 int compressedSize, int maxOutputSize,
509 const void *dictStart, size_t dictSize)
511 return LZ4_decompress_generic(source, dest,
512 compressedSize, maxOutputSize,
513 endOnInputSize, decode_full_block,
514 usingExtDict, (BYTE *)dest,
515 (const BYTE *)dictStart, dictSize);
518 static int LZ4_decompress_fast_extDict(const char *source, char *dest,
520 const void *dictStart, size_t dictSize)
522 return LZ4_decompress_generic(source, dest,
524 endOnOutputSize, decode_full_block,
525 usingExtDict, (BYTE *)dest,
526 (const BYTE *)dictStart, dictSize);
530 * The "double dictionary" mode, for use with e.g. ring buffers: the first part
531 * of the dictionary is passed as prefix, and the second via dictStart + dictSize.
532 * These routines are used only once, in LZ4_decompress_*_continue().
535 int LZ4_decompress_safe_doubleDict(const char *source, char *dest,
536 int compressedSize, int maxOutputSize,
538 const void *dictStart, size_t dictSize)
540 return LZ4_decompress_generic(source, dest,
541 compressedSize, maxOutputSize,
542 endOnInputSize, decode_full_block,
543 usingExtDict, (BYTE *)dest - prefixSize,
544 (const BYTE *)dictStart, dictSize);
548 int LZ4_decompress_fast_doubleDict(const char *source, char *dest,
549 int originalSize, size_t prefixSize,
550 const void *dictStart, size_t dictSize)
552 return LZ4_decompress_generic(source, dest,
554 endOnOutputSize, decode_full_block,
555 usingExtDict, (BYTE *)dest - prefixSize,
556 (const BYTE *)dictStart, dictSize);
559 /* ===== streaming decompression functions ===== */
561 int LZ4_setStreamDecode(LZ4_streamDecode_t *LZ4_streamDecode,
562 const char *dictionary, int dictSize)
564 LZ4_streamDecode_t_internal *lz4sd =
565 &LZ4_streamDecode->internal_donotuse;
567 lz4sd->prefixSize = (size_t) dictSize;
568 lz4sd->prefixEnd = (const BYTE *) dictionary + dictSize;
569 lz4sd->externalDict = NULL;
570 lz4sd->extDictSize = 0;
576 * These decoding functions allow decompression of multiple blocks
577 * in "streaming" mode.
578 * Previously decoded blocks must still be available at the memory
579 * position where they were decoded.
580 * If it's not possible, save the relevant part of
581 * decoded data into a safe buffer,
582 * and indicate where it stands using LZ4_setStreamDecode()
584 int LZ4_decompress_safe_continue(LZ4_streamDecode_t *LZ4_streamDecode,
585 const char *source, char *dest, int compressedSize, int maxOutputSize)
587 LZ4_streamDecode_t_internal *lz4sd =
588 &LZ4_streamDecode->internal_donotuse;
591 if (lz4sd->prefixSize == 0) {
592 /* The first call, no dictionary yet. */
593 assert(lz4sd->extDictSize == 0);
594 result = LZ4_decompress_safe(source, dest,
595 compressedSize, maxOutputSize);
598 lz4sd->prefixSize = result;
599 lz4sd->prefixEnd = (BYTE *)dest + result;
600 } else if (lz4sd->prefixEnd == (BYTE *)dest) {
601 /* They're rolling the current segment. */
602 if (lz4sd->prefixSize >= 64 * KB - 1)
603 result = LZ4_decompress_safe_withPrefix64k(source, dest,
604 compressedSize, maxOutputSize);
605 else if (lz4sd->extDictSize == 0)
606 result = LZ4_decompress_safe_withSmallPrefix(source,
607 dest, compressedSize, maxOutputSize,
610 result = LZ4_decompress_safe_doubleDict(source, dest,
611 compressedSize, maxOutputSize,
613 lz4sd->externalDict, lz4sd->extDictSize);
616 lz4sd->prefixSize += result;
617 lz4sd->prefixEnd += result;
620 * The buffer wraps around, or they're
621 * switching to another buffer.
623 lz4sd->extDictSize = lz4sd->prefixSize;
624 lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
625 result = LZ4_decompress_safe_forceExtDict(source, dest,
626 compressedSize, maxOutputSize,
627 lz4sd->externalDict, lz4sd->extDictSize);
630 lz4sd->prefixSize = result;
631 lz4sd->prefixEnd = (BYTE *)dest + result;
637 int LZ4_decompress_fast_continue(LZ4_streamDecode_t *LZ4_streamDecode,
638 const char *source, char *dest, int originalSize)
640 LZ4_streamDecode_t_internal *lz4sd = &LZ4_streamDecode->internal_donotuse;
643 if (lz4sd->prefixSize == 0) {
644 assert(lz4sd->extDictSize == 0);
645 result = LZ4_decompress_fast(source, dest, originalSize);
648 lz4sd->prefixSize = originalSize;
649 lz4sd->prefixEnd = (BYTE *)dest + originalSize;
650 } else if (lz4sd->prefixEnd == (BYTE *)dest) {
651 if (lz4sd->prefixSize >= 64 * KB - 1 ||
652 lz4sd->extDictSize == 0)
653 result = LZ4_decompress_fast(source, dest,
656 result = LZ4_decompress_fast_doubleDict(source, dest,
657 originalSize, lz4sd->prefixSize,
658 lz4sd->externalDict, lz4sd->extDictSize);
661 lz4sd->prefixSize += originalSize;
662 lz4sd->prefixEnd += originalSize;
664 lz4sd->extDictSize = lz4sd->prefixSize;
665 lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
666 result = LZ4_decompress_fast_extDict(source, dest,
667 originalSize, lz4sd->externalDict, lz4sd->extDictSize);
670 lz4sd->prefixSize = originalSize;
671 lz4sd->prefixEnd = (BYTE *)dest + originalSize;
676 int LZ4_decompress_safe_usingDict(const char *source, char *dest,
677 int compressedSize, int maxOutputSize,
678 const char *dictStart, int dictSize)
681 return LZ4_decompress_safe(source, dest,
682 compressedSize, maxOutputSize);
683 if (dictStart+dictSize == dest) {
684 if (dictSize >= 64 * KB - 1)
685 return LZ4_decompress_safe_withPrefix64k(source, dest,
686 compressedSize, maxOutputSize);
687 return LZ4_decompress_safe_withSmallPrefix(source, dest,
688 compressedSize, maxOutputSize, dictSize);
690 return LZ4_decompress_safe_forceExtDict(source, dest,
691 compressedSize, maxOutputSize, dictStart, dictSize);
694 int LZ4_decompress_fast_usingDict(const char *source, char *dest,
696 const char *dictStart, int dictSize)
698 if (dictSize == 0 || dictStart + dictSize == dest)
699 return LZ4_decompress_fast(source, dest, originalSize);
701 return LZ4_decompress_fast_extDict(source, dest, originalSize,
702 dictStart, dictSize);
706 EXPORT_SYMBOL(LZ4_decompress_safe);
707 EXPORT_SYMBOL(LZ4_decompress_safe_partial);
708 EXPORT_SYMBOL(LZ4_decompress_fast);
709 EXPORT_SYMBOL(LZ4_setStreamDecode);
710 EXPORT_SYMBOL(LZ4_decompress_safe_continue);
711 EXPORT_SYMBOL(LZ4_decompress_fast_continue);
712 EXPORT_SYMBOL(LZ4_decompress_safe_usingDict);
713 EXPORT_SYMBOL(LZ4_decompress_fast_usingDict);
715 MODULE_LICENSE("Dual BSD/GPL");
716 MODULE_DESCRIPTION("LZ4 decompressor");