721 lines · c
1/*2 * LZ4 - Fast LZ compression algorithm3 * 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 without6 * modification, are permitted provided that the following conditions are7 * met:8 * * Redistributions of source code must retain the above copyright9 * notice, this list of conditions and the following disclaimer.10 * * Redistributions in binary form must reproduce the above11 * copyright notice, this list of conditions and the following disclaimer12 * in the documentation and/or other materials provided with the13 * distribution.14 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS15 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT16 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR17 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT18 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,19 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT20 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE24 * 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.org27 * - LZ4 source repository : https://github.com/lz4/lz428 *29 * Changed for kernel usage by:30 * Sven Schmidt <4sschmid@informatik.uni-hamburg.de>31 */32 33/*-************************************34 * Dependencies35 **************************************/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 <linux/unaligned.h>42 43/*-*****************************44 * Decompression functions45 *******************************/46 47#define DEBUGLOG(l, ...) {} /* disabled */48 49#ifndef assert50#define assert(condition) ((void)0)51#endif52 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 dictSize81 )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 enough136 * space, enter the shortcut and copy 16 bytes on behalf137 * of the literals (in the fast mode, only 8 bytes can be138 * safely copied this way).139 * 2) Further if the match length is 4..18, copy 18 bytes140 * in a similar manner; but we ensure that there's enough141 * space in the output for those 18 bytes earlier, upon142 * entering the shortcut (in other words, there is a143 * combined check for both stages).144 *145 * The & in the likely() below is intentionally not && so that146 * some compilers can produce better parallelized runtime code147 */148 if ((endOnInput ? length != RUN_MASK : length <= 8)149 /*150 * strictly "less than" on input, to re-enter151 * the loop with at least one byte152 */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 info185 * is ready. Propel it right to the point of match186 * 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(endOnInput203 ? ip < iend - RUN_MASK204 : 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 segment232 */233 cpy = oend;234 length = oend - op;235 }236 if ((endOnInput)237 && (ip + length > iend)) {238 /*239 * Error :240 * read attempt beyond241 * end of input buffer242 */243 goto _output_error;244 }245 } else {246 if ((!endOnInput)247 && (cpy != oend)) {248 /*249 * Error :250 * block decoding must251 * stop exactly there252 */253 goto _output_error;254 }255 if ((endOnInput)256 && ((ip + length != iend)257 || (cpy > oend))) {258 /*259 * Error :260 * input must be consumed261 */262 goto _output_error;263 }264 }265 266 /*267 * supports overlapping memory regions; only matters268 * for in-place decompression scenarios269 */270 LZ4_memmove(op, ip, length);271 ip += length;272 op += length;273 274 /* Necessarily EOF when !partialDecoding.275 * When partialDecoding, it is EOF if we've either276 * filled the output buffer or277 * can't proceed with reading an offset for following match.278 */279 if (!partialDecoding || (cpy == oend) || (ip >= (iend - 2)))280 break;281 } else {282 /* may overwrite up to WILDCOPYLENGTH beyond cpy */283 LZ4_wildCopy(op, ip, cpy);284 ip += length;285 op = cpy;286 }287 288 /* get offset */289 offset = LZ4_readLE16(ip);290 ip += 2;291 match = op - offset;292 293 /* get matchlength */294 length = token & ML_MASK;295 296_copy_match:297 if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) {298 /* Error : offset outside buffers */299 goto _output_error;300 }301 302 /* costs ~1%; silence an msan warning when offset == 0 */303 /*304 * note : when partialDecoding, there is no guarantee that305 * at least 4 bytes remain available in output buffer306 */307 if (!partialDecoding) {308 assert(oend > op);309 assert(oend - op >= 4);310 311 LZ4_write32(op, (U32)offset);312 }313 314 if (length == ML_MASK) {315 unsigned int s;316 317 do {318 s = *ip++;319 320 if ((endOnInput) && (ip > iend - LASTLITERALS))321 goto _output_error;322 323 length += s;324 } while (s == 255);325 326 if ((safeDecode)327 && unlikely(328 (uptrval)(op) + length < (uptrval)op)) {329 /* overflow detection */330 goto _output_error;331 }332 }333 334 length += MINMATCH;335 336 /* 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)341 goto _output_error;342 length = min(length, (size_t)(oend - op));343 }344 345 if (length <= (size_t)(lowPrefix - match)) {346 /*347 * match fits entirely within external348 * dictionary : just copy349 */350 memmove(op, dictEnd - (lowPrefix - match),351 length);352 op += length;353 } else {354 /*355 * match stretches into both external356 * dictionary and current block357 */358 size_t const copySize = (size_t)(lowPrefix - match);359 size_t const restSize = length - copySize;360 361 LZ4_memcpy(op, dictEnd - copySize, copySize);362 op += copySize;363 if (restSize > (size_t)(op - lowPrefix)) {364 /* overlap copy */365 BYTE * const endOfMatch = op + restSize;366 const BYTE *copyFrom = lowPrefix;367 368 while (op < endOfMatch)369 *op++ = *copyFrom++;370 } else {371 LZ4_memcpy(op, lowPrefix, restSize);372 op += restSize;373 }374 }375 continue;376 }377 378 /* copy match within block */379 cpy = op + length;380 381 /*382 * partialDecoding :383 * may not respect endBlock parsing restrictions384 */385 assert(op <= oend);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;391 392 if (matchEnd > op) {393 /* overlap copy */394 while (op < copyEnd)395 *op++ = *match++;396 } else {397 LZ4_memcpy(op, match, mlen);398 }399 op = copyEnd;400 if (op == oend)401 break;402 continue;403 }404 405 if (unlikely(offset < 8)) {406 op[0] = match[0];407 op[1] = match[1];408 op[2] = match[2];409 op[3] = match[3];410 match += inc32table[offset];411 LZ4_memcpy(op + 4, match, 4);412 match -= dec64table[offset];413 } else {414 LZ4_copy8(op, match);415 match += 8;416 }417 418 op += 8;419 420 if (unlikely(cpy > oend - MATCH_SAFEGUARD_DISTANCE)) {421 BYTE * const oCopyLimit = oend - (WILDCOPYLENGTH - 1);422 423 if (cpy > oend - LASTLITERALS) {424 /*425 * Error : last LASTLITERALS bytes426 * must be literals (uncompressed)427 */428 goto _output_error;429 }430 431 if (op < oCopyLimit) {432 LZ4_wildCopy(op, match, oCopyLimit);433 match += oCopyLimit - op;434 op = oCopyLimit;435 }436 while (op < cpy)437 *op++ = *match++;438 } else {439 LZ4_copy8(op, match);440 if (length > 16)441 LZ4_wildCopy(op + 8, match + 8, cpy);442 }443 op = cpy; /* wildcopy correction */444 }445 446 /* end of decoding */447 if (endOnInput) {448 /* Nb of output bytes decoded */449 return (int) (((char *)op) - dst);450 } else {451 /* Nb of input bytes read */452 return (int) (((const char *)ip) - src);453 }454 455 /* Overflow error detected */456_output_error:457 return (int) (-(((const char *)ip) - src)) - 1;458}459 460int LZ4_decompress_safe(const char *source, char *dest,461 int compressedSize, int maxDecompressedSize)462{463 return LZ4_decompress_generic(source, dest,464 compressedSize, maxDecompressedSize,465 endOnInputSize, decode_full_block,466 noDict, (BYTE *)dest, NULL, 0);467}468 469int LZ4_decompress_safe_partial(const char *src, char *dst,470 int compressedSize, int targetOutputSize, int dstCapacity)471{472 dstCapacity = min(targetOutputSize, dstCapacity);473 return LZ4_decompress_generic(src, dst, compressedSize, dstCapacity,474 endOnInputSize, partial_decode,475 noDict, (BYTE *)dst, NULL, 0);476}477 478int LZ4_decompress_fast(const char *source, char *dest, int originalSize)479{480 return LZ4_decompress_generic(source, dest, 0, originalSize,481 endOnOutputSize, decode_full_block,482 withPrefix64k,483 (BYTE *)dest - 64 * KB, NULL, 0);484}485 486/* ===== Instantiate a few more decoding cases, used more than once. ===== */487 488static int LZ4_decompress_safe_withPrefix64k(const char *source, char *dest,489 int compressedSize, int maxOutputSize)490{491 return LZ4_decompress_generic(source, dest,492 compressedSize, maxOutputSize,493 endOnInputSize, decode_full_block,494 withPrefix64k,495 (BYTE *)dest - 64 * KB, NULL, 0);496}497 498static int LZ4_decompress_safe_withSmallPrefix(const char *source, char *dest,499 int compressedSize,500 int maxOutputSize,501 size_t prefixSize)502{503 return LZ4_decompress_generic(source, dest,504 compressedSize, maxOutputSize,505 endOnInputSize, decode_full_block,506 noDict,507 (BYTE *)dest - prefixSize, NULL, 0);508}509 510static int LZ4_decompress_safe_forceExtDict(const char *source, char *dest,511 int compressedSize, int maxOutputSize,512 const void *dictStart, size_t dictSize)513{514 return LZ4_decompress_generic(source, dest,515 compressedSize, maxOutputSize,516 endOnInputSize, decode_full_block,517 usingExtDict, (BYTE *)dest,518 (const BYTE *)dictStart, dictSize);519}520 521static int LZ4_decompress_fast_extDict(const char *source, char *dest,522 int originalSize,523 const void *dictStart, size_t dictSize)524{525 return LZ4_decompress_generic(source, dest,526 0, originalSize,527 endOnOutputSize, decode_full_block,528 usingExtDict, (BYTE *)dest,529 (const BYTE *)dictStart, dictSize);530}531 532/*533 * The "double dictionary" mode, for use with e.g. ring buffers: the first part534 * of the dictionary is passed as prefix, and the second via dictStart + dictSize.535 * These routines are used only once, in LZ4_decompress_*_continue().536 */537static FORCE_INLINE538int LZ4_decompress_safe_doubleDict(const char *source, char *dest,539 int compressedSize, int maxOutputSize,540 size_t prefixSize,541 const void *dictStart, size_t dictSize)542{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);548}549 550static FORCE_INLINE551int LZ4_decompress_fast_doubleDict(const char *source, char *dest,552 int originalSize, size_t prefixSize,553 const void *dictStart, size_t dictSize)554{555 return LZ4_decompress_generic(source, dest,556 0, originalSize,557 endOnOutputSize, decode_full_block,558 usingExtDict, (BYTE *)dest - prefixSize,559 (const BYTE *)dictStart, dictSize);560}561 562/* ===== streaming decompression functions ===== */563 564int LZ4_setStreamDecode(LZ4_streamDecode_t *LZ4_streamDecode,565 const char *dictionary, int dictSize)566{567 LZ4_streamDecode_t_internal *lz4sd =568 &LZ4_streamDecode->internal_donotuse;569 570 lz4sd->prefixSize = (size_t) dictSize;571 lz4sd->prefixEnd = (const BYTE *) dictionary + dictSize;572 lz4sd->externalDict = NULL;573 lz4sd->extDictSize = 0;574 return 1;575}576 577/*578 * *_continue() :579 * These decoding functions allow decompression of multiple blocks580 * in "streaming" mode.581 * Previously decoded blocks must still be available at the memory582 * position where they were decoded.583 * If it's not possible, save the relevant part of584 * decoded data into a safe buffer,585 * and indicate where it stands using LZ4_setStreamDecode()586 */587int LZ4_decompress_safe_continue(LZ4_streamDecode_t *LZ4_streamDecode,588 const char *source, char *dest, int compressedSize, int maxOutputSize)589{590 LZ4_streamDecode_t_internal *lz4sd =591 &LZ4_streamDecode->internal_donotuse;592 int result;593 594 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);599 if (result <= 0)600 return result;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,611 lz4sd->prefixSize);612 else613 result = LZ4_decompress_safe_doubleDict(source, dest,614 compressedSize, maxOutputSize,615 lz4sd->prefixSize,616 lz4sd->externalDict, lz4sd->extDictSize);617 if (result <= 0)618 return result;619 lz4sd->prefixSize += result;620 lz4sd->prefixEnd += result;621 } else {622 /*623 * The buffer wraps around, or they're624 * switching to another buffer.625 */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);631 if (result <= 0)632 return result;633 lz4sd->prefixSize = result;634 lz4sd->prefixEnd = (BYTE *)dest + result;635 }636 637 return result;638}639 640int LZ4_decompress_fast_continue(LZ4_streamDecode_t *LZ4_streamDecode,641 const char *source, char *dest, int originalSize)642{643 LZ4_streamDecode_t_internal *lz4sd = &LZ4_streamDecode->internal_donotuse;644 int result;645 646 if (lz4sd->prefixSize == 0) {647 assert(lz4sd->extDictSize == 0);648 result = LZ4_decompress_fast(source, dest, originalSize);649 if (result <= 0)650 return result;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,657 originalSize);658 else659 result = LZ4_decompress_fast_doubleDict(source, dest,660 originalSize, lz4sd->prefixSize,661 lz4sd->externalDict, lz4sd->extDictSize);662 if (result <= 0)663 return result;664 lz4sd->prefixSize += originalSize;665 lz4sd->prefixEnd += originalSize;666 } else {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);671 if (result <= 0)672 return result;673 lz4sd->prefixSize = originalSize;674 lz4sd->prefixEnd = (BYTE *)dest + originalSize;675 }676 return result;677}678 679int LZ4_decompress_safe_usingDict(const char *source, char *dest,680 int compressedSize, int maxOutputSize,681 const char *dictStart, int dictSize)682{683 if (dictSize == 0)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);692 }693 return LZ4_decompress_safe_forceExtDict(source, dest,694 compressedSize, maxOutputSize, dictStart, dictSize);695}696 697int LZ4_decompress_fast_usingDict(const char *source, char *dest,698 int originalSize,699 const char *dictStart, int dictSize)700{701 if (dictSize == 0 || dictStart + dictSize == dest)702 return LZ4_decompress_fast(source, dest, originalSize);703 704 return LZ4_decompress_fast_extDict(source, dest, originalSize,705 dictStart, dictSize);706}707 708#ifndef STATIC709EXPORT_SYMBOL(LZ4_decompress_safe);710EXPORT_SYMBOL(LZ4_decompress_safe_partial);711EXPORT_SYMBOL(LZ4_decompress_fast);712EXPORT_SYMBOL(LZ4_setStreamDecode);713EXPORT_SYMBOL(LZ4_decompress_safe_continue);714EXPORT_SYMBOL(LZ4_decompress_fast_continue);715EXPORT_SYMBOL(LZ4_decompress_safe_usingDict);716EXPORT_SYMBOL(LZ4_decompress_fast_usingDict);717 718MODULE_LICENSE("Dual BSD/GPL");719MODULE_DESCRIPTION("LZ4 decompressor");720#endif721