Loading...
1/*
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
7 * met:
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
13 * distribution.
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
28 *
29 * Changed for kernel usage by:
30 * Sven Schmidt <4sschmid@informatik.uni-hamburg.de>
31 */
32
33/*-************************************
34 * Dependencies
35 **************************************/
36#include <linux/lz4.h>
37#include "lz4defs.h"
38#include <linux/init.h>
39#include <linux/module.h>
40#include <linux/kernel.h>
41#include <asm/unaligned.h>
42
43/*-*****************************
44 * Decompression functions
45 *******************************/
46
47#define DEBUGLOG(l, ...) {} /* disabled */
48
49#ifndef assert
50#define assert(condition) ((void)0)
51#endif
52
53/*
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.
59 */
60static FORCE_INLINE int LZ4_decompress_generic(
61 const char * const src,
62 char * const dst,
63 int srcSize,
64 /*
65 * If endOnInput == endOnInputSize,
66 * this value is `dstCapacity`
67 */
68 int outputSize,
69 /* endOnOutputSize, endOnInputSize */
70 endCondition_directive endOnInput,
71 /* full, partial */
72 earlyEnd_directive partialDecoding,
73 /* noDict, withPrefix64k, usingExtDict */
74 dict_directive dict,
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 */
80 const size_t dictSize
81 )
82{
83 const BYTE *ip = (const BYTE *) src;
84 const BYTE * const iend = ip + srcSize;
85
86 BYTE *op = (BYTE *) dst;
87 BYTE * const oend = op + outputSize;
88 BYTE *cpy;
89
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};
93
94 const int safeDecode = (endOnInput == endOnInputSize);
95 const int checkOffset = ((safeDecode) && (dictSize < (int)(64 * KB)));
96
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*/;
102
103 DEBUGLOG(5, "%s (srcSize:%i, dstSize:%i)", __func__,
104 srcSize, outputSize);
105
106 /* Special cases */
107 assert(lowPrefix <= op);
108 assert(src != NULL);
109
110 /* Empty output buffer */
111 if ((endOnInput) && (unlikely(outputSize == 0)))
112 return ((srcSize == 1) && (*ip == 0)) ? 0 : -1;
113
114 if ((!endOnInput) && (unlikely(outputSize == 0)))
115 return (*ip == 0 ? 1 : -1);
116
117 if ((endOnInput) && unlikely(srcSize == 0))
118 return -1;
119
120 /* Main Loop : decode sequences */
121 while (1) {
122 size_t length;
123 const BYTE *match;
124 size_t offset;
125
126 /* get literal length */
127 unsigned int const token = *ip++;
128 length = token>>ML_BITS;
129
130 /* ip < iend before the increment */
131 assert(!endOnInput || ip <= iend);
132
133 /*
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).
144 *
145 * The & in the likely() below is intentionally not && so that
146 * some compilers can produce better parallelized runtime code
147 */
148 if ((endOnInput ? length != RUN_MASK : length <= 8)
149 /*
150 * strictly "less than" on input, to re-enter
151 * the loop with at least one byte
152 */
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;
158
159 /*
160 * The second stage:
161 * prepare for match copying, decode full info.
162 * If it doesn't work out, the info won't be wasted.
163 */
164 length = token & ML_MASK; /* match length */
165 offset = LZ4_readLE16(ip);
166 ip += 2;
167 match = op - offset;
168 assert(match <= op); /* check overflow */
169
170 /* Do not deal with overlapping matches. */
171 if ((length != ML_MASK) &&
172 (offset >= 8) &&
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. */
180 continue;
181 }
182
183 /*
184 * The second stage didn't work out, but the info
185 * is ready. Propel it right to the point of match
186 * copying.
187 */
188 goto _copy_match;
189 }
190
191 /* decode literal length */
192 if (length == RUN_MASK) {
193 unsigned int s;
194
195 if (unlikely(endOnInput ? ip >= iend - RUN_MASK : 0)) {
196 /* overflow detection */
197 goto _output_error;
198 }
199 do {
200 s = *ip++;
201 length += s;
202 } while (likely(endOnInput
203 ? ip < iend - RUN_MASK
204 : 1) & (s == 255));
205
206 if ((safeDecode)
207 && unlikely((uptrval)(op) +
208 length < (uptrval)(op))) {
209 /* overflow detection */
210 goto _output_error;
211 }
212 if ((safeDecode)
213 && unlikely((uptrval)(ip) +
214 length < (uptrval)(ip))) {
215 /* overflow detection */
216 goto _output_error;
217 }
218 }
219
220 /* copy literals */
221 cpy = op + length;
222 LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
223
224 if (((endOnInput) && ((cpy > oend - MFLIMIT)
225 || (ip + length > iend - (2 + 1 + LASTLITERALS))))
226 || ((!endOnInput) && (cpy > oend - WILDCOPYLENGTH))) {
227 if (partialDecoding) {
228 if (cpy > oend) {
229 /*
230 * Partial decoding :
231 * stop in the middle of literal segment
232 */
233 cpy = oend;
234 length = oend - op;
235 }
236 if ((endOnInput)
237 && (ip + length > iend)) {
238 /*
239 * Error :
240 * read attempt beyond
241 * end of input buffer
242 */
243 goto _output_error;
244 }
245 } else {
246 if ((!endOnInput)
247 && (cpy != oend)) {
248 /*
249 * Error :
250 * block decoding must
251 * stop exactly there
252 */
253 goto _output_error;
254 }
255 if ((endOnInput)
256 && ((ip + length != iend)
257 || (cpy > oend))) {
258 /*
259 * Error :
260 * input must be consumed
261 */
262 goto _output_error;
263 }
264 }
265
266 /*
267 * supports overlapping memory regions; only matters
268 * for in-place decompression scenarios
269 */
270 LZ4_memmove(op, ip, length);
271 ip += length;
272 op += length;
273
274 /* Necessarily EOF, due to parsing restrictions */
275 if (!partialDecoding || (cpy == oend))
276 break;
277 } else {
278 /* may overwrite up to WILDCOPYLENGTH beyond cpy */
279 LZ4_wildCopy(op, ip, cpy);
280 ip += length;
281 op = cpy;
282 }
283
284 /* get offset */
285 offset = LZ4_readLE16(ip);
286 ip += 2;
287 match = op - offset;
288
289 /* get matchlength */
290 length = token & ML_MASK;
291
292_copy_match:
293 if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) {
294 /* Error : offset outside buffers */
295 goto _output_error;
296 }
297
298 /* costs ~1%; silence an msan warning when offset == 0 */
299 /*
300 * note : when partialDecoding, there is no guarantee that
301 * at least 4 bytes remain available in output buffer
302 */
303 if (!partialDecoding) {
304 assert(oend > op);
305 assert(oend - op >= 4);
306
307 LZ4_write32(op, (U32)offset);
308 }
309
310 if (length == ML_MASK) {
311 unsigned int s;
312
313 do {
314 s = *ip++;
315
316 if ((endOnInput) && (ip > iend - LASTLITERALS))
317 goto _output_error;
318
319 length += s;
320 } while (s == 255);
321
322 if ((safeDecode)
323 && unlikely(
324 (uptrval)(op) + length < (uptrval)op)) {
325 /* overflow detection */
326 goto _output_error;
327 }
328 }
329
330 length += MINMATCH;
331
332 /* match starting within external dictionary */
333 if ((dict == usingExtDict) && (match < lowPrefix)) {
334 if (unlikely(op + length > oend - LASTLITERALS)) {
335 /* doesn't respect parsing restriction */
336 if (!partialDecoding)
337 goto _output_error;
338 length = min(length, (size_t)(oend - op));
339 }
340
341 if (length <= (size_t)(lowPrefix - match)) {
342 /*
343 * match fits entirely within external
344 * dictionary : just copy
345 */
346 memmove(op, dictEnd - (lowPrefix - match),
347 length);
348 op += length;
349 } else {
350 /*
351 * match stretches into both external
352 * dictionary and current block
353 */
354 size_t const copySize = (size_t)(lowPrefix - match);
355 size_t const restSize = length - copySize;
356
357 LZ4_memcpy(op, dictEnd - copySize, copySize);
358 op += copySize;
359 if (restSize > (size_t)(op - lowPrefix)) {
360 /* overlap copy */
361 BYTE * const endOfMatch = op + restSize;
362 const BYTE *copyFrom = lowPrefix;
363
364 while (op < endOfMatch)
365 *op++ = *copyFrom++;
366 } else {
367 LZ4_memcpy(op, lowPrefix, restSize);
368 op += restSize;
369 }
370 }
371 continue;
372 }
373
374 /* copy match within block */
375 cpy = op + length;
376
377 /*
378 * partialDecoding :
379 * may not respect endBlock parsing restrictions
380 */
381 assert(op <= oend);
382 if (partialDecoding &&
383 (cpy > oend - MATCH_SAFEGUARD_DISTANCE)) {
384 size_t const mlen = min(length, (size_t)(oend - op));
385 const BYTE * const matchEnd = match + mlen;
386 BYTE * const copyEnd = op + mlen;
387
388 if (matchEnd > op) {
389 /* overlap copy */
390 while (op < copyEnd)
391 *op++ = *match++;
392 } else {
393 LZ4_memcpy(op, match, mlen);
394 }
395 op = copyEnd;
396 if (op == oend)
397 break;
398 continue;
399 }
400
401 if (unlikely(offset < 8)) {
402 op[0] = match[0];
403 op[1] = match[1];
404 op[2] = match[2];
405 op[3] = match[3];
406 match += inc32table[offset];
407 LZ4_memcpy(op + 4, match, 4);
408 match -= dec64table[offset];
409 } else {
410 LZ4_copy8(op, match);
411 match += 8;
412 }
413
414 op += 8;
415
416 if (unlikely(cpy > oend - MATCH_SAFEGUARD_DISTANCE)) {
417 BYTE * const oCopyLimit = oend - (WILDCOPYLENGTH - 1);
418
419 if (cpy > oend - LASTLITERALS) {
420 /*
421 * Error : last LASTLITERALS bytes
422 * must be literals (uncompressed)
423 */
424 goto _output_error;
425 }
426
427 if (op < oCopyLimit) {
428 LZ4_wildCopy(op, match, oCopyLimit);
429 match += oCopyLimit - op;
430 op = oCopyLimit;
431 }
432 while (op < cpy)
433 *op++ = *match++;
434 } else {
435 LZ4_copy8(op, match);
436 if (length > 16)
437 LZ4_wildCopy(op + 8, match + 8, cpy);
438 }
439 op = cpy; /* wildcopy correction */
440 }
441
442 /* end of decoding */
443 if (endOnInput) {
444 /* Nb of output bytes decoded */
445 return (int) (((char *)op) - dst);
446 } else {
447 /* Nb of input bytes read */
448 return (int) (((const char *)ip) - src);
449 }
450
451 /* Overflow error detected */
452_output_error:
453 return (int) (-(((const char *)ip) - src)) - 1;
454}
455
456int LZ4_decompress_safe(const char *source, char *dest,
457 int compressedSize, int maxDecompressedSize)
458{
459 return LZ4_decompress_generic(source, dest,
460 compressedSize, maxDecompressedSize,
461 endOnInputSize, decode_full_block,
462 noDict, (BYTE *)dest, NULL, 0);
463}
464
465int LZ4_decompress_safe_partial(const char *src, char *dst,
466 int compressedSize, int targetOutputSize, int dstCapacity)
467{
468 dstCapacity = min(targetOutputSize, dstCapacity);
469 return LZ4_decompress_generic(src, dst, compressedSize, dstCapacity,
470 endOnInputSize, partial_decode,
471 noDict, (BYTE *)dst, NULL, 0);
472}
473
474int LZ4_decompress_fast(const char *source, char *dest, int originalSize)
475{
476 return LZ4_decompress_generic(source, dest, 0, originalSize,
477 endOnOutputSize, decode_full_block,
478 withPrefix64k,
479 (BYTE *)dest - 64 * KB, NULL, 0);
480}
481
482/* ===== Instantiate a few more decoding cases, used more than once. ===== */
483
484static int LZ4_decompress_safe_withPrefix64k(const char *source, char *dest,
485 int compressedSize, int maxOutputSize)
486{
487 return LZ4_decompress_generic(source, dest,
488 compressedSize, maxOutputSize,
489 endOnInputSize, decode_full_block,
490 withPrefix64k,
491 (BYTE *)dest - 64 * KB, NULL, 0);
492}
493
494static int LZ4_decompress_safe_withSmallPrefix(const char *source, char *dest,
495 int compressedSize,
496 int maxOutputSize,
497 size_t prefixSize)
498{
499 return LZ4_decompress_generic(source, dest,
500 compressedSize, maxOutputSize,
501 endOnInputSize, decode_full_block,
502 noDict,
503 (BYTE *)dest - prefixSize, NULL, 0);
504}
505
506int LZ4_decompress_safe_forceExtDict(const char *source, char *dest,
507 int compressedSize, int maxOutputSize,
508 const void *dictStart, size_t dictSize)
509{
510 return LZ4_decompress_generic(source, dest,
511 compressedSize, maxOutputSize,
512 endOnInputSize, decode_full_block,
513 usingExtDict, (BYTE *)dest,
514 (const BYTE *)dictStart, dictSize);
515}
516
517static int LZ4_decompress_fast_extDict(const char *source, char *dest,
518 int originalSize,
519 const void *dictStart, size_t dictSize)
520{
521 return LZ4_decompress_generic(source, dest,
522 0, originalSize,
523 endOnOutputSize, decode_full_block,
524 usingExtDict, (BYTE *)dest,
525 (const BYTE *)dictStart, dictSize);
526}
527
528/*
529 * The "double dictionary" mode, for use with e.g. ring buffers: the first part
530 * of the dictionary is passed as prefix, and the second via dictStart + dictSize.
531 * These routines are used only once, in LZ4_decompress_*_continue().
532 */
533static FORCE_INLINE
534int LZ4_decompress_safe_doubleDict(const char *source, char *dest,
535 int compressedSize, int maxOutputSize,
536 size_t prefixSize,
537 const void *dictStart, size_t dictSize)
538{
539 return LZ4_decompress_generic(source, dest,
540 compressedSize, maxOutputSize,
541 endOnInputSize, decode_full_block,
542 usingExtDict, (BYTE *)dest - prefixSize,
543 (const BYTE *)dictStart, dictSize);
544}
545
546static FORCE_INLINE
547int LZ4_decompress_fast_doubleDict(const char *source, char *dest,
548 int originalSize, size_t prefixSize,
549 const void *dictStart, size_t dictSize)
550{
551 return LZ4_decompress_generic(source, dest,
552 0, originalSize,
553 endOnOutputSize, decode_full_block,
554 usingExtDict, (BYTE *)dest - prefixSize,
555 (const BYTE *)dictStart, dictSize);
556}
557
558/* ===== streaming decompression functions ===== */
559
560int LZ4_setStreamDecode(LZ4_streamDecode_t *LZ4_streamDecode,
561 const char *dictionary, int dictSize)
562{
563 LZ4_streamDecode_t_internal *lz4sd =
564 &LZ4_streamDecode->internal_donotuse;
565
566 lz4sd->prefixSize = (size_t) dictSize;
567 lz4sd->prefixEnd = (const BYTE *) dictionary + dictSize;
568 lz4sd->externalDict = NULL;
569 lz4sd->extDictSize = 0;
570 return 1;
571}
572
573/*
574 * *_continue() :
575 * These decoding functions allow decompression of multiple blocks
576 * in "streaming" mode.
577 * Previously decoded blocks must still be available at the memory
578 * position where they were decoded.
579 * If it's not possible, save the relevant part of
580 * decoded data into a safe buffer,
581 * and indicate where it stands using LZ4_setStreamDecode()
582 */
583int LZ4_decompress_safe_continue(LZ4_streamDecode_t *LZ4_streamDecode,
584 const char *source, char *dest, int compressedSize, int maxOutputSize)
585{
586 LZ4_streamDecode_t_internal *lz4sd =
587 &LZ4_streamDecode->internal_donotuse;
588 int result;
589
590 if (lz4sd->prefixSize == 0) {
591 /* The first call, no dictionary yet. */
592 assert(lz4sd->extDictSize == 0);
593 result = LZ4_decompress_safe(source, dest,
594 compressedSize, maxOutputSize);
595 if (result <= 0)
596 return result;
597 lz4sd->prefixSize = result;
598 lz4sd->prefixEnd = (BYTE *)dest + result;
599 } else if (lz4sd->prefixEnd == (BYTE *)dest) {
600 /* They're rolling the current segment. */
601 if (lz4sd->prefixSize >= 64 * KB - 1)
602 result = LZ4_decompress_safe_withPrefix64k(source, dest,
603 compressedSize, maxOutputSize);
604 else if (lz4sd->extDictSize == 0)
605 result = LZ4_decompress_safe_withSmallPrefix(source,
606 dest, compressedSize, maxOutputSize,
607 lz4sd->prefixSize);
608 else
609 result = LZ4_decompress_safe_doubleDict(source, dest,
610 compressedSize, maxOutputSize,
611 lz4sd->prefixSize,
612 lz4sd->externalDict, lz4sd->extDictSize);
613 if (result <= 0)
614 return result;
615 lz4sd->prefixSize += result;
616 lz4sd->prefixEnd += result;
617 } else {
618 /*
619 * The buffer wraps around, or they're
620 * switching to another buffer.
621 */
622 lz4sd->extDictSize = lz4sd->prefixSize;
623 lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
624 result = LZ4_decompress_safe_forceExtDict(source, dest,
625 compressedSize, maxOutputSize,
626 lz4sd->externalDict, lz4sd->extDictSize);
627 if (result <= 0)
628 return result;
629 lz4sd->prefixSize = result;
630 lz4sd->prefixEnd = (BYTE *)dest + result;
631 }
632
633 return result;
634}
635
636int LZ4_decompress_fast_continue(LZ4_streamDecode_t *LZ4_streamDecode,
637 const char *source, char *dest, int originalSize)
638{
639 LZ4_streamDecode_t_internal *lz4sd = &LZ4_streamDecode->internal_donotuse;
640 int result;
641
642 if (lz4sd->prefixSize == 0) {
643 assert(lz4sd->extDictSize == 0);
644 result = LZ4_decompress_fast(source, dest, originalSize);
645 if (result <= 0)
646 return result;
647 lz4sd->prefixSize = originalSize;
648 lz4sd->prefixEnd = (BYTE *)dest + originalSize;
649 } else if (lz4sd->prefixEnd == (BYTE *)dest) {
650 if (lz4sd->prefixSize >= 64 * KB - 1 ||
651 lz4sd->extDictSize == 0)
652 result = LZ4_decompress_fast(source, dest,
653 originalSize);
654 else
655 result = LZ4_decompress_fast_doubleDict(source, dest,
656 originalSize, lz4sd->prefixSize,
657 lz4sd->externalDict, lz4sd->extDictSize);
658 if (result <= 0)
659 return result;
660 lz4sd->prefixSize += originalSize;
661 lz4sd->prefixEnd += originalSize;
662 } else {
663 lz4sd->extDictSize = lz4sd->prefixSize;
664 lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
665 result = LZ4_decompress_fast_extDict(source, dest,
666 originalSize, lz4sd->externalDict, lz4sd->extDictSize);
667 if (result <= 0)
668 return result;
669 lz4sd->prefixSize = originalSize;
670 lz4sd->prefixEnd = (BYTE *)dest + originalSize;
671 }
672 return result;
673}
674
675int LZ4_decompress_safe_usingDict(const char *source, char *dest,
676 int compressedSize, int maxOutputSize,
677 const char *dictStart, int dictSize)
678{
679 if (dictSize == 0)
680 return LZ4_decompress_safe(source, dest,
681 compressedSize, maxOutputSize);
682 if (dictStart+dictSize == dest) {
683 if (dictSize >= 64 * KB - 1)
684 return LZ4_decompress_safe_withPrefix64k(source, dest,
685 compressedSize, maxOutputSize);
686 return LZ4_decompress_safe_withSmallPrefix(source, dest,
687 compressedSize, maxOutputSize, dictSize);
688 }
689 return LZ4_decompress_safe_forceExtDict(source, dest,
690 compressedSize, maxOutputSize, dictStart, dictSize);
691}
692
693int LZ4_decompress_fast_usingDict(const char *source, char *dest,
694 int originalSize,
695 const char *dictStart, int dictSize)
696{
697 if (dictSize == 0 || dictStart + dictSize == dest)
698 return LZ4_decompress_fast(source, dest, originalSize);
699
700 return LZ4_decompress_fast_extDict(source, dest, originalSize,
701 dictStart, dictSize);
702}
703
704#ifndef STATIC
705EXPORT_SYMBOL(LZ4_decompress_safe);
706EXPORT_SYMBOL(LZ4_decompress_safe_partial);
707EXPORT_SYMBOL(LZ4_decompress_fast);
708EXPORT_SYMBOL(LZ4_setStreamDecode);
709EXPORT_SYMBOL(LZ4_decompress_safe_continue);
710EXPORT_SYMBOL(LZ4_decompress_fast_continue);
711EXPORT_SYMBOL(LZ4_decompress_safe_usingDict);
712EXPORT_SYMBOL(LZ4_decompress_fast_usingDict);
713
714MODULE_LICENSE("Dual BSD/GPL");
715MODULE_DESCRIPTION("LZ4 decompressor");
716#endif
1/*
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
7 * met:
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
13 * distribution.
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
28 *
29 * Changed for kernel usage by:
30 * Sven Schmidt <4sschmid@informatik.uni-hamburg.de>
31 */
32
33/*-************************************
34 * Dependencies
35 **************************************/
36#include <linux/lz4.h>
37#include "lz4defs.h"
38#include <linux/init.h>
39#include <linux/module.h>
40#include <linux/kernel.h>
41#include <asm/unaligned.h>
42
43/*-*****************************
44 * Decompression functions
45 *******************************/
46/* LZ4_decompress_generic() :
47 * This generic decompression function cover all use cases.
48 * It shall be instantiated several times, using different sets of directives
49 * Note that it is important this generic function is really inlined,
50 * in order to remove useless branches during compilation optimization.
51 */
52static FORCE_INLINE int LZ4_decompress_generic(
53 const char * const source,
54 char * const dest,
55 int inputSize,
56 /*
57 * If endOnInput == endOnInputSize,
58 * this value is the max size of Output Buffer.
59 */
60 int outputSize,
61 /* endOnOutputSize, endOnInputSize */
62 int endOnInput,
63 /* full, partial */
64 int partialDecoding,
65 /* only used if partialDecoding == partial */
66 int targetOutputSize,
67 /* noDict, withPrefix64k, usingExtDict */
68 int dict,
69 /* == dest when no prefix */
70 const BYTE * const lowPrefix,
71 /* only if dict == usingExtDict */
72 const BYTE * const dictStart,
73 /* note : = 0 if noDict */
74 const size_t dictSize
75 )
76{
77 /* Local Variables */
78 const BYTE *ip = (const BYTE *) source;
79 const BYTE * const iend = ip + inputSize;
80
81 BYTE *op = (BYTE *) dest;
82 BYTE * const oend = op + outputSize;
83 BYTE *cpy;
84 BYTE *oexit = op + targetOutputSize;
85 const BYTE * const lowLimit = lowPrefix - dictSize;
86
87 const BYTE * const dictEnd = (const BYTE *)dictStart + dictSize;
88 static const unsigned int dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };
89 static const int dec64table[] = { 0, 0, 0, -1, 0, 1, 2, 3 };
90
91 const int safeDecode = (endOnInput == endOnInputSize);
92 const int checkOffset = ((safeDecode) && (dictSize < (int)(64 * KB)));
93
94 /* Special cases */
95 /* targetOutputSize too high => decode everything */
96 if ((partialDecoding) && (oexit > oend - MFLIMIT))
97 oexit = oend - MFLIMIT;
98
99 /* Empty output buffer */
100 if ((endOnInput) && (unlikely(outputSize == 0)))
101 return ((inputSize == 1) && (*ip == 0)) ? 0 : -1;
102
103 if ((!endOnInput) && (unlikely(outputSize == 0)))
104 return (*ip == 0 ? 1 : -1);
105
106 /* Main Loop : decode sequences */
107 while (1) {
108 size_t length;
109 const BYTE *match;
110 size_t offset;
111
112 /* get literal length */
113 unsigned int const token = *ip++;
114
115 length = token>>ML_BITS;
116
117 if (length == RUN_MASK) {
118 unsigned int s;
119
120 do {
121 s = *ip++;
122 length += s;
123 } while (likely(endOnInput
124 ? ip < iend - RUN_MASK
125 : 1) & (s == 255));
126
127 if ((safeDecode)
128 && unlikely(
129 (size_t)(op + length) < (size_t)(op))) {
130 /* overflow detection */
131 goto _output_error;
132 }
133 if ((safeDecode)
134 && unlikely(
135 (size_t)(ip + length) < (size_t)(ip))) {
136 /* overflow detection */
137 goto _output_error;
138 }
139 }
140
141 /* copy literals */
142 cpy = op + length;
143 if (((endOnInput) && ((cpy > (partialDecoding ? oexit : oend - MFLIMIT))
144 || (ip + length > iend - (2 + 1 + LASTLITERALS))))
145 || ((!endOnInput) && (cpy > oend - WILDCOPYLENGTH))) {
146 if (partialDecoding) {
147 if (cpy > oend) {
148 /*
149 * Error :
150 * write attempt beyond end of output buffer
151 */
152 goto _output_error;
153 }
154 if ((endOnInput)
155 && (ip + length > iend)) {
156 /*
157 * Error :
158 * read attempt beyond
159 * end of input buffer
160 */
161 goto _output_error;
162 }
163 } else {
164 if ((!endOnInput)
165 && (cpy != oend)) {
166 /*
167 * Error :
168 * block decoding must
169 * stop exactly there
170 */
171 goto _output_error;
172 }
173 if ((endOnInput)
174 && ((ip + length != iend)
175 || (cpy > oend))) {
176 /*
177 * Error :
178 * input must be consumed
179 */
180 goto _output_error;
181 }
182 }
183
184 memcpy(op, ip, length);
185 ip += length;
186 op += length;
187 /* Necessarily EOF, due to parsing restrictions */
188 break;
189 }
190
191 LZ4_wildCopy(op, ip, cpy);
192 ip += length;
193 op = cpy;
194
195 /* get offset */
196 offset = LZ4_readLE16(ip);
197 ip += 2;
198 match = op - offset;
199
200 if ((checkOffset) && (unlikely(match < lowLimit))) {
201 /* Error : offset outside buffers */
202 goto _output_error;
203 }
204
205 /* costs ~1%; silence an msan warning when offset == 0 */
206 LZ4_write32(op, (U32)offset);
207
208 /* get matchlength */
209 length = token & ML_MASK;
210 if (length == ML_MASK) {
211 unsigned int s;
212
213 do {
214 s = *ip++;
215
216 if ((endOnInput) && (ip > iend - LASTLITERALS))
217 goto _output_error;
218
219 length += s;
220 } while (s == 255);
221
222 if ((safeDecode)
223 && unlikely(
224 (size_t)(op + length) < (size_t)op)) {
225 /* overflow detection */
226 goto _output_error;
227 }
228 }
229
230 length += MINMATCH;
231
232 /* check external dictionary */
233 if ((dict == usingExtDict) && (match < lowPrefix)) {
234 if (unlikely(op + length > oend - LASTLITERALS)) {
235 /* doesn't respect parsing restriction */
236 goto _output_error;
237 }
238
239 if (length <= (size_t)(lowPrefix - match)) {
240 /*
241 * match can be copied as a single segment
242 * from external dictionary
243 */
244 memmove(op, dictEnd - (lowPrefix - match),
245 length);
246 op += length;
247 } else {
248 /*
249 * match encompass external
250 * dictionary and current block
251 */
252 size_t const copySize = (size_t)(lowPrefix - match);
253 size_t const restSize = length - copySize;
254
255 memcpy(op, dictEnd - copySize, copySize);
256 op += copySize;
257
258 if (restSize > (size_t)(op - lowPrefix)) {
259 /* overlap copy */
260 BYTE * const endOfMatch = op + restSize;
261 const BYTE *copyFrom = lowPrefix;
262
263 while (op < endOfMatch)
264 *op++ = *copyFrom++;
265 } else {
266 memcpy(op, lowPrefix, restSize);
267 op += restSize;
268 }
269 }
270
271 continue;
272 }
273
274 /* copy match within block */
275 cpy = op + length;
276
277 if (unlikely(offset < 8)) {
278 const int dec64 = dec64table[offset];
279
280 op[0] = match[0];
281 op[1] = match[1];
282 op[2] = match[2];
283 op[3] = match[3];
284 match += dec32table[offset];
285 memcpy(op + 4, match, 4);
286 match -= dec64;
287 } else {
288 LZ4_copy8(op, match);
289 match += 8;
290 }
291
292 op += 8;
293
294 if (unlikely(cpy > oend - 12)) {
295 BYTE * const oCopyLimit = oend - (WILDCOPYLENGTH - 1);
296
297 if (cpy > oend - LASTLITERALS) {
298 /*
299 * Error : last LASTLITERALS bytes
300 * must be literals (uncompressed)
301 */
302 goto _output_error;
303 }
304
305 if (op < oCopyLimit) {
306 LZ4_wildCopy(op, match, oCopyLimit);
307 match += oCopyLimit - op;
308 op = oCopyLimit;
309 }
310
311 while (op < cpy)
312 *op++ = *match++;
313 } else {
314 LZ4_copy8(op, match);
315
316 if (length > 16)
317 LZ4_wildCopy(op + 8, match + 8, cpy);
318 }
319
320 op = cpy; /* correction */
321 }
322
323 /* end of decoding */
324 if (endOnInput) {
325 /* Nb of output bytes decoded */
326 return (int) (((char *)op) - dest);
327 } else {
328 /* Nb of input bytes read */
329 return (int) (((const char *)ip) - source);
330 }
331
332 /* Overflow error detected */
333_output_error:
334 return -1;
335}
336
337int LZ4_decompress_safe(const char *source, char *dest,
338 int compressedSize, int maxDecompressedSize)
339{
340 return LZ4_decompress_generic(source, dest, compressedSize,
341 maxDecompressedSize, endOnInputSize, full, 0,
342 noDict, (BYTE *)dest, NULL, 0);
343}
344
345int LZ4_decompress_safe_partial(const char *source, char *dest,
346 int compressedSize, int targetOutputSize, int maxDecompressedSize)
347{
348 return LZ4_decompress_generic(source, dest, compressedSize,
349 maxDecompressedSize, endOnInputSize, partial,
350 targetOutputSize, noDict, (BYTE *)dest, NULL, 0);
351}
352
353int LZ4_decompress_fast(const char *source, char *dest, int originalSize)
354{
355 return LZ4_decompress_generic(source, dest, 0, originalSize,
356 endOnOutputSize, full, 0, withPrefix64k,
357 (BYTE *)(dest - 64 * KB), NULL, 64 * KB);
358}
359
360int LZ4_setStreamDecode(LZ4_streamDecode_t *LZ4_streamDecode,
361 const char *dictionary, int dictSize)
362{
363 LZ4_streamDecode_t_internal *lz4sd = (LZ4_streamDecode_t_internal *) LZ4_streamDecode;
364
365 lz4sd->prefixSize = (size_t) dictSize;
366 lz4sd->prefixEnd = (const BYTE *) dictionary + dictSize;
367 lz4sd->externalDict = NULL;
368 lz4sd->extDictSize = 0;
369 return 1;
370}
371
372/*
373 * *_continue() :
374 * These decoding functions allow decompression of multiple blocks
375 * in "streaming" mode.
376 * Previously decoded blocks must still be available at the memory
377 * position where they were decoded.
378 * If it's not possible, save the relevant part of
379 * decoded data into a safe buffer,
380 * and indicate where it stands using LZ4_setStreamDecode()
381 */
382int LZ4_decompress_safe_continue(LZ4_streamDecode_t *LZ4_streamDecode,
383 const char *source, char *dest, int compressedSize, int maxOutputSize)
384{
385 LZ4_streamDecode_t_internal *lz4sd = &LZ4_streamDecode->internal_donotuse;
386 int result;
387
388 if (lz4sd->prefixEnd == (BYTE *)dest) {
389 result = LZ4_decompress_generic(source, dest,
390 compressedSize,
391 maxOutputSize,
392 endOnInputSize, full, 0,
393 usingExtDict, lz4sd->prefixEnd - lz4sd->prefixSize,
394 lz4sd->externalDict,
395 lz4sd->extDictSize);
396
397 if (result <= 0)
398 return result;
399
400 lz4sd->prefixSize += result;
401 lz4sd->prefixEnd += result;
402 } else {
403 lz4sd->extDictSize = lz4sd->prefixSize;
404 lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
405 result = LZ4_decompress_generic(source, dest,
406 compressedSize, maxOutputSize,
407 endOnInputSize, full, 0,
408 usingExtDict, (BYTE *)dest,
409 lz4sd->externalDict, lz4sd->extDictSize);
410 if (result <= 0)
411 return result;
412 lz4sd->prefixSize = result;
413 lz4sd->prefixEnd = (BYTE *)dest + result;
414 }
415
416 return result;
417}
418
419int LZ4_decompress_fast_continue(LZ4_streamDecode_t *LZ4_streamDecode,
420 const char *source, char *dest, int originalSize)
421{
422 LZ4_streamDecode_t_internal *lz4sd = &LZ4_streamDecode->internal_donotuse;
423 int result;
424
425 if (lz4sd->prefixEnd == (BYTE *)dest) {
426 result = LZ4_decompress_generic(source, dest, 0, originalSize,
427 endOnOutputSize, full, 0,
428 usingExtDict,
429 lz4sd->prefixEnd - lz4sd->prefixSize,
430 lz4sd->externalDict, lz4sd->extDictSize);
431
432 if (result <= 0)
433 return result;
434
435 lz4sd->prefixSize += originalSize;
436 lz4sd->prefixEnd += originalSize;
437 } else {
438 lz4sd->extDictSize = lz4sd->prefixSize;
439 lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
440 result = LZ4_decompress_generic(source, dest, 0, originalSize,
441 endOnOutputSize, full, 0,
442 usingExtDict, (BYTE *)dest,
443 lz4sd->externalDict, lz4sd->extDictSize);
444 if (result <= 0)
445 return result;
446 lz4sd->prefixSize = originalSize;
447 lz4sd->prefixEnd = (BYTE *)dest + originalSize;
448 }
449
450 return result;
451}
452
453/*
454 * Advanced decoding functions :
455 * *_usingDict() :
456 * These decoding functions work the same as "_continue" ones,
457 * the dictionary must be explicitly provided within parameters
458 */
459static FORCE_INLINE int LZ4_decompress_usingDict_generic(const char *source,
460 char *dest, int compressedSize, int maxOutputSize, int safe,
461 const char *dictStart, int dictSize)
462{
463 if (dictSize == 0)
464 return LZ4_decompress_generic(source, dest,
465 compressedSize, maxOutputSize, safe, full, 0,
466 noDict, (BYTE *)dest, NULL, 0);
467 if (dictStart + dictSize == dest) {
468 if (dictSize >= (int)(64 * KB - 1))
469 return LZ4_decompress_generic(source, dest,
470 compressedSize, maxOutputSize, safe, full, 0,
471 withPrefix64k, (BYTE *)dest - 64 * KB, NULL, 0);
472 return LZ4_decompress_generic(source, dest, compressedSize,
473 maxOutputSize, safe, full, 0, noDict,
474 (BYTE *)dest - dictSize, NULL, 0);
475 }
476 return LZ4_decompress_generic(source, dest, compressedSize,
477 maxOutputSize, safe, full, 0, usingExtDict,
478 (BYTE *)dest, (const BYTE *)dictStart, dictSize);
479}
480
481int LZ4_decompress_safe_usingDict(const char *source, char *dest,
482 int compressedSize, int maxOutputSize,
483 const char *dictStart, int dictSize)
484{
485 return LZ4_decompress_usingDict_generic(source, dest,
486 compressedSize, maxOutputSize, 1, dictStart, dictSize);
487}
488
489int LZ4_decompress_fast_usingDict(const char *source, char *dest,
490 int originalSize, const char *dictStart, int dictSize)
491{
492 return LZ4_decompress_usingDict_generic(source, dest, 0,
493 originalSize, 0, dictStart, dictSize);
494}
495
496#ifndef STATIC
497EXPORT_SYMBOL(LZ4_decompress_safe);
498EXPORT_SYMBOL(LZ4_decompress_safe_partial);
499EXPORT_SYMBOL(LZ4_decompress_fast);
500EXPORT_SYMBOL(LZ4_setStreamDecode);
501EXPORT_SYMBOL(LZ4_decompress_safe_continue);
502EXPORT_SYMBOL(LZ4_decompress_fast_continue);
503EXPORT_SYMBOL(LZ4_decompress_safe_usingDict);
504EXPORT_SYMBOL(LZ4_decompress_fast_usingDict);
505
506MODULE_LICENSE("Dual BSD/GPL");
507MODULE_DESCRIPTION("LZ4 decompressor");
508#endif