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 <linux/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
267 * supports overlapping memory regions; only matters
268 * for in-place decompression scenarios
270 LZ4_memmove(op
, ip
, length
);
274 /* Necessarily EOF when !partialDecoding.
275 * When partialDecoding, it is EOF if we've either
276 * filled the output buffer or
277 * can't proceed with reading an offset for following match.
279 if (!partialDecoding
|| (cpy
== oend
) || (ip
>= (iend
- 2)))
282 /* may overwrite up to WILDCOPYLENGTH beyond cpy */
283 LZ4_wildCopy(op
, ip
, cpy
);
289 offset
= LZ4_readLE16(ip
);
293 /* get matchlength */
294 length
= token
& ML_MASK
;
297 if ((checkOffset
) && (unlikely(match
+ dictSize
< lowPrefix
))) {
298 /* Error : offset outside buffers */
302 /* costs ~1%; silence an msan warning when offset == 0 */
304 * note : when partialDecoding, there is no guarantee that
305 * at least 4 bytes remain available in output buffer
307 if (!partialDecoding
) {
309 assert(oend
- op
>= 4);
311 LZ4_write32(op
, (U32
)offset
);
314 if (length
== ML_MASK
) {
320 if ((endOnInput
) && (ip
> iend
- LASTLITERALS
))
328 (uptrval
)(op
) + length
< (uptrval
)op
)) {
329 /* overflow detection */
336 /* match starting within external dictionary */
337 if ((dict
== usingExtDict
) && (match
< lowPrefix
)) {
338 if (unlikely(op
+ length
> oend
- LASTLITERALS
)) {
339 /* doesn't respect parsing restriction */
340 if (!partialDecoding
)
342 length
= min(length
, (size_t)(oend
- op
));
345 if (length
<= (size_t)(lowPrefix
- match
)) {
347 * match fits entirely within external
348 * dictionary : just copy
350 memmove(op
, dictEnd
- (lowPrefix
- match
),
355 * match stretches into both external
356 * dictionary and current block
358 size_t const copySize
= (size_t)(lowPrefix
- match
);
359 size_t const restSize
= length
- copySize
;
361 LZ4_memcpy(op
, dictEnd
- copySize
, copySize
);
363 if (restSize
> (size_t)(op
- lowPrefix
)) {
365 BYTE
* const endOfMatch
= op
+ restSize
;
366 const BYTE
*copyFrom
= lowPrefix
;
368 while (op
< endOfMatch
)
371 LZ4_memcpy(op
, lowPrefix
, restSize
);
378 /* copy match within block */
383 * may not respect endBlock parsing restrictions
386 if (partialDecoding
&&
387 (cpy
> oend
- MATCH_SAFEGUARD_DISTANCE
)) {
388 size_t const mlen
= min(length
, (size_t)(oend
- op
));
389 const BYTE
* const matchEnd
= match
+ mlen
;
390 BYTE
* const copyEnd
= op
+ mlen
;
397 LZ4_memcpy(op
, match
, mlen
);
405 if (unlikely(offset
< 8)) {
410 match
+= inc32table
[offset
];
411 LZ4_memcpy(op
+ 4, match
, 4);
412 match
-= dec64table
[offset
];
414 LZ4_copy8(op
, match
);
420 if (unlikely(cpy
> oend
- MATCH_SAFEGUARD_DISTANCE
)) {
421 BYTE
* const oCopyLimit
= oend
- (WILDCOPYLENGTH
- 1);
423 if (cpy
> oend
- LASTLITERALS
) {
425 * Error : last LASTLITERALS bytes
426 * must be literals (uncompressed)
431 if (op
< oCopyLimit
) {
432 LZ4_wildCopy(op
, match
, oCopyLimit
);
433 match
+= oCopyLimit
- op
;
439 LZ4_copy8(op
, match
);
441 LZ4_wildCopy(op
+ 8, match
+ 8, cpy
);
443 op
= cpy
; /* wildcopy correction */
446 /* end of decoding */
448 /* Nb of output bytes decoded */
449 return (int) (((char *)op
) - dst
);
451 /* Nb of input bytes read */
452 return (int) (((const char *)ip
) - src
);
455 /* Overflow error detected */
457 return (int) (-(((const char *)ip
) - src
)) - 1;
460 int LZ4_decompress_safe(const char *source
, char *dest
,
461 int compressedSize
, int maxDecompressedSize
)
463 return LZ4_decompress_generic(source
, dest
,
464 compressedSize
, maxDecompressedSize
,
465 endOnInputSize
, decode_full_block
,
466 noDict
, (BYTE
*)dest
, NULL
, 0);
469 int LZ4_decompress_safe_partial(const char *src
, char *dst
,
470 int compressedSize
, int targetOutputSize
, int dstCapacity
)
472 dstCapacity
= min(targetOutputSize
, dstCapacity
);
473 return LZ4_decompress_generic(src
, dst
, compressedSize
, dstCapacity
,
474 endOnInputSize
, partial_decode
,
475 noDict
, (BYTE
*)dst
, NULL
, 0);
478 int LZ4_decompress_fast(const char *source
, char *dest
, int originalSize
)
480 return LZ4_decompress_generic(source
, dest
, 0, originalSize
,
481 endOnOutputSize
, decode_full_block
,
483 (BYTE
*)dest
- 64 * KB
, NULL
, 0);
486 /* ===== Instantiate a few more decoding cases, used more than once. ===== */
488 static int LZ4_decompress_safe_withPrefix64k(const char *source
, char *dest
,
489 int compressedSize
, int maxOutputSize
)
491 return LZ4_decompress_generic(source
, dest
,
492 compressedSize
, maxOutputSize
,
493 endOnInputSize
, decode_full_block
,
495 (BYTE
*)dest
- 64 * KB
, NULL
, 0);
498 static int LZ4_decompress_safe_withSmallPrefix(const char *source
, char *dest
,
503 return LZ4_decompress_generic(source
, dest
,
504 compressedSize
, maxOutputSize
,
505 endOnInputSize
, decode_full_block
,
507 (BYTE
*)dest
- prefixSize
, NULL
, 0);
510 static int LZ4_decompress_safe_forceExtDict(const char *source
, char *dest
,
511 int compressedSize
, int maxOutputSize
,
512 const void *dictStart
, size_t dictSize
)
514 return LZ4_decompress_generic(source
, dest
,
515 compressedSize
, maxOutputSize
,
516 endOnInputSize
, decode_full_block
,
517 usingExtDict
, (BYTE
*)dest
,
518 (const BYTE
*)dictStart
, dictSize
);
521 static int LZ4_decompress_fast_extDict(const char *source
, char *dest
,
523 const void *dictStart
, size_t dictSize
)
525 return LZ4_decompress_generic(source
, dest
,
527 endOnOutputSize
, decode_full_block
,
528 usingExtDict
, (BYTE
*)dest
,
529 (const BYTE
*)dictStart
, dictSize
);
533 * The "double dictionary" mode, for use with e.g. ring buffers: the first part
534 * of the dictionary is passed as prefix, and the second via dictStart + dictSize.
535 * These routines are used only once, in LZ4_decompress_*_continue().
538 int LZ4_decompress_safe_doubleDict(const char *source
, char *dest
,
539 int compressedSize
, int maxOutputSize
,
541 const void *dictStart
, size_t dictSize
)
543 return LZ4_decompress_generic(source
, dest
,
544 compressedSize
, maxOutputSize
,
545 endOnInputSize
, decode_full_block
,
546 usingExtDict
, (BYTE
*)dest
- prefixSize
,
547 (const BYTE
*)dictStart
, dictSize
);
551 int LZ4_decompress_fast_doubleDict(const char *source
, char *dest
,
552 int originalSize
, size_t prefixSize
,
553 const void *dictStart
, size_t dictSize
)
555 return LZ4_decompress_generic(source
, dest
,
557 endOnOutputSize
, decode_full_block
,
558 usingExtDict
, (BYTE
*)dest
- prefixSize
,
559 (const BYTE
*)dictStart
, dictSize
);
562 /* ===== streaming decompression functions ===== */
564 int LZ4_setStreamDecode(LZ4_streamDecode_t
*LZ4_streamDecode
,
565 const char *dictionary
, int dictSize
)
567 LZ4_streamDecode_t_internal
*lz4sd
=
568 &LZ4_streamDecode
->internal_donotuse
;
570 lz4sd
->prefixSize
= (size_t) dictSize
;
571 lz4sd
->prefixEnd
= (const BYTE
*) dictionary
+ dictSize
;
572 lz4sd
->externalDict
= NULL
;
573 lz4sd
->extDictSize
= 0;
579 * These decoding functions allow decompression of multiple blocks
580 * in "streaming" mode.
581 * Previously decoded blocks must still be available at the memory
582 * position where they were decoded.
583 * If it's not possible, save the relevant part of
584 * decoded data into a safe buffer,
585 * and indicate where it stands using LZ4_setStreamDecode()
587 int LZ4_decompress_safe_continue(LZ4_streamDecode_t
*LZ4_streamDecode
,
588 const char *source
, char *dest
, int compressedSize
, int maxOutputSize
)
590 LZ4_streamDecode_t_internal
*lz4sd
=
591 &LZ4_streamDecode
->internal_donotuse
;
594 if (lz4sd
->prefixSize
== 0) {
595 /* The first call, no dictionary yet. */
596 assert(lz4sd
->extDictSize
== 0);
597 result
= LZ4_decompress_safe(source
, dest
,
598 compressedSize
, maxOutputSize
);
601 lz4sd
->prefixSize
= result
;
602 lz4sd
->prefixEnd
= (BYTE
*)dest
+ result
;
603 } else if (lz4sd
->prefixEnd
== (BYTE
*)dest
) {
604 /* They're rolling the current segment. */
605 if (lz4sd
->prefixSize
>= 64 * KB
- 1)
606 result
= LZ4_decompress_safe_withPrefix64k(source
, dest
,
607 compressedSize
, maxOutputSize
);
608 else if (lz4sd
->extDictSize
== 0)
609 result
= LZ4_decompress_safe_withSmallPrefix(source
,
610 dest
, compressedSize
, maxOutputSize
,
613 result
= LZ4_decompress_safe_doubleDict(source
, dest
,
614 compressedSize
, maxOutputSize
,
616 lz4sd
->externalDict
, lz4sd
->extDictSize
);
619 lz4sd
->prefixSize
+= result
;
620 lz4sd
->prefixEnd
+= result
;
623 * The buffer wraps around, or they're
624 * switching to another buffer.
626 lz4sd
->extDictSize
= lz4sd
->prefixSize
;
627 lz4sd
->externalDict
= lz4sd
->prefixEnd
- lz4sd
->extDictSize
;
628 result
= LZ4_decompress_safe_forceExtDict(source
, dest
,
629 compressedSize
, maxOutputSize
,
630 lz4sd
->externalDict
, lz4sd
->extDictSize
);
633 lz4sd
->prefixSize
= result
;
634 lz4sd
->prefixEnd
= (BYTE
*)dest
+ result
;
640 int LZ4_decompress_fast_continue(LZ4_streamDecode_t
*LZ4_streamDecode
,
641 const char *source
, char *dest
, int originalSize
)
643 LZ4_streamDecode_t_internal
*lz4sd
= &LZ4_streamDecode
->internal_donotuse
;
646 if (lz4sd
->prefixSize
== 0) {
647 assert(lz4sd
->extDictSize
== 0);
648 result
= LZ4_decompress_fast(source
, dest
, originalSize
);
651 lz4sd
->prefixSize
= originalSize
;
652 lz4sd
->prefixEnd
= (BYTE
*)dest
+ originalSize
;
653 } else if (lz4sd
->prefixEnd
== (BYTE
*)dest
) {
654 if (lz4sd
->prefixSize
>= 64 * KB
- 1 ||
655 lz4sd
->extDictSize
== 0)
656 result
= LZ4_decompress_fast(source
, dest
,
659 result
= LZ4_decompress_fast_doubleDict(source
, dest
,
660 originalSize
, lz4sd
->prefixSize
,
661 lz4sd
->externalDict
, lz4sd
->extDictSize
);
664 lz4sd
->prefixSize
+= originalSize
;
665 lz4sd
->prefixEnd
+= originalSize
;
667 lz4sd
->extDictSize
= lz4sd
->prefixSize
;
668 lz4sd
->externalDict
= lz4sd
->prefixEnd
- lz4sd
->extDictSize
;
669 result
= LZ4_decompress_fast_extDict(source
, dest
,
670 originalSize
, lz4sd
->externalDict
, lz4sd
->extDictSize
);
673 lz4sd
->prefixSize
= originalSize
;
674 lz4sd
->prefixEnd
= (BYTE
*)dest
+ originalSize
;
679 int LZ4_decompress_safe_usingDict(const char *source
, char *dest
,
680 int compressedSize
, int maxOutputSize
,
681 const char *dictStart
, int dictSize
)
684 return LZ4_decompress_safe(source
, dest
,
685 compressedSize
, maxOutputSize
);
686 if (dictStart
+dictSize
== dest
) {
687 if (dictSize
>= 64 * KB
- 1)
688 return LZ4_decompress_safe_withPrefix64k(source
, dest
,
689 compressedSize
, maxOutputSize
);
690 return LZ4_decompress_safe_withSmallPrefix(source
, dest
,
691 compressedSize
, maxOutputSize
, dictSize
);
693 return LZ4_decompress_safe_forceExtDict(source
, dest
,
694 compressedSize
, maxOutputSize
, dictStart
, dictSize
);
697 int LZ4_decompress_fast_usingDict(const char *source
, char *dest
,
699 const char *dictStart
, int dictSize
)
701 if (dictSize
== 0 || dictStart
+ dictSize
== dest
)
702 return LZ4_decompress_fast(source
, dest
, originalSize
);
704 return LZ4_decompress_fast_extDict(source
, dest
, originalSize
,
705 dictStart
, dictSize
);
709 EXPORT_SYMBOL(LZ4_decompress_safe
);
710 EXPORT_SYMBOL(LZ4_decompress_safe_partial
);
711 EXPORT_SYMBOL(LZ4_decompress_fast
);
712 EXPORT_SYMBOL(LZ4_setStreamDecode
);
713 EXPORT_SYMBOL(LZ4_decompress_safe_continue
);
714 EXPORT_SYMBOL(LZ4_decompress_fast_continue
);
715 EXPORT_SYMBOL(LZ4_decompress_safe_usingDict
);
716 EXPORT_SYMBOL(LZ4_decompress_fast_usingDict
);
718 MODULE_LICENSE("Dual BSD/GPL");
719 MODULE_DESCRIPTION("LZ4 decompressor");