Annotation of /trunk/mkinitrd-magellan/busybox/archival/libunarchive/decompress_bunzip2.c
Parent Directory | Revision Log
Revision 1123 -
(hide annotations)
(download)
Wed Aug 18 21:56:57 2010 UTC (13 years, 8 months ago) by niro
File MIME type: text/plain
File size: 24548 byte(s)
Wed Aug 18 21:56:57 2010 UTC (13 years, 8 months ago) by niro
File MIME type: text/plain
File size: 24548 byte(s)
-updated to busybox-1.17.1
1 | niro | 532 | /* vi: set sw=4 ts=4: */ |
2 | /* Small bzip2 deflate implementation, by Rob Landley (rob@landley.net). | ||
3 | |||
4 | Based on bzip2 decompression code by Julian R Seward (jseward@acm.org), | ||
5 | which also acknowledges contributions by Mike Burrows, David Wheeler, | ||
6 | Peter Fenwick, Alistair Moffat, Radford Neal, Ian H. Witten, | ||
7 | Robert Sedgewick, and Jon L. Bentley. | ||
8 | |||
9 | Licensed under GPLv2 or later, see file LICENSE in this tarball for details. | ||
10 | */ | ||
11 | |||
12 | /* | ||
13 | Size and speed optimizations by Manuel Novoa III (mjn3@codepoet.org). | ||
14 | |||
15 | More efficient reading of Huffman codes, a streamlined read_bunzip() | ||
16 | function, and various other tweaks. In (limited) tests, approximately | ||
17 | 20% faster than bzcat on x86 and about 10% faster on arm. | ||
18 | |||
19 | Note that about 2/3 of the time is spent in read_unzip() reversing | ||
20 | the Burrows-Wheeler transformation. Much of that time is delay | ||
21 | resulting from cache misses. | ||
22 | |||
23 | I would ask that anyone benefiting from this work, especially those | ||
24 | using it in commercial products, consider making a donation to my local | ||
25 | non-profit hospice organization (www.hospiceacadiana.com) in the name of | ||
26 | the woman I loved, Toni W. Hagan, who passed away Feb. 12, 2003. | ||
27 | |||
28 | Manuel | ||
29 | */ | ||
30 | |||
31 | #include "libbb.h" | ||
32 | #include "unarchive.h" | ||
33 | |||
34 | /* Constants for Huffman coding */ | ||
35 | #define MAX_GROUPS 6 | ||
36 | #define GROUP_SIZE 50 /* 64 would have been more efficient */ | ||
37 | #define MAX_HUFCODE_BITS 20 /* Longest Huffman code allowed */ | ||
38 | #define MAX_SYMBOLS 258 /* 256 literals + RUNA + RUNB */ | ||
39 | #define SYMBOL_RUNA 0 | ||
40 | #define SYMBOL_RUNB 1 | ||
41 | |||
42 | /* Status return values */ | ||
43 | #define RETVAL_OK 0 | ||
44 | #define RETVAL_LAST_BLOCK (-1) | ||
45 | #define RETVAL_NOT_BZIP_DATA (-2) | ||
46 | #define RETVAL_UNEXPECTED_INPUT_EOF (-3) | ||
47 | niro | 816 | #define RETVAL_SHORT_WRITE (-4) |
48 | niro | 532 | #define RETVAL_DATA_ERROR (-5) |
49 | #define RETVAL_OUT_OF_MEMORY (-6) | ||
50 | #define RETVAL_OBSOLETE_INPUT (-7) | ||
51 | |||
52 | /* Other housekeeping constants */ | ||
53 | #define IOBUF_SIZE 4096 | ||
54 | |||
55 | /* This is what we know about each Huffman coding group */ | ||
56 | struct group_data { | ||
57 | niro | 816 | /* We have an extra slot at the end of limit[] for a sentinel value. */ |
58 | int limit[MAX_HUFCODE_BITS+1], base[MAX_HUFCODE_BITS], permute[MAX_SYMBOLS]; | ||
59 | niro | 532 | int minLen, maxLen; |
60 | }; | ||
61 | |||
62 | /* Structure holding all the housekeeping data, including IO buffers and | ||
63 | niro | 816 | * memory that persists between calls to bunzip |
64 | * Found the most used member: | ||
65 | * cat this_file.c | sed -e 's/"/ /g' -e "s/'/ /g" | xargs -n1 \ | ||
66 | * | grep 'bd->' | sed 's/^.*bd->/bd->/' | sort | $PAGER | ||
67 | * and moved it (inbufBitCount) to offset 0. | ||
68 | */ | ||
69 | struct bunzip_data { | ||
70 | /* I/O tracking data (file handles, buffers, positions, etc.) */ | ||
71 | unsigned inbufBitCount, inbufBits; | ||
72 | int in_fd, out_fd, inbufCount, inbufPos /*, outbufPos*/; | ||
73 | unsigned char *inbuf /*,*outbuf*/; | ||
74 | niro | 532 | |
75 | /* State for interrupting output loop */ | ||
76 | niro | 816 | int writeCopies, writePos, writeRunCountdown, writeCount, writeCurrent; |
77 | niro | 532 | |
78 | /* The CRC values stored in the block header and calculated from the data */ | ||
79 | niro | 816 | uint32_t headerCRC, totalCRC, writeCRC; |
80 | niro | 532 | |
81 | /* Intermediate buffer and its size (in bytes) */ | ||
82 | niro | 816 | unsigned *dbuf, dbufSize; |
83 | niro | 532 | |
84 | niro | 816 | /* For I/O error handling */ |
85 | jmp_buf jmpbuf; | ||
86 | niro | 532 | |
87 | niro | 816 | /* Big things go last (register-relative addressing can be larger for big offsets) */ |
88 | uint32_t crc32Table[256]; | ||
89 | niro | 532 | unsigned char selectors[32768]; /* nSelectors=15 bits */ |
90 | struct group_data groups[MAX_GROUPS]; /* Huffman coding tables */ | ||
91 | niro | 816 | }; |
92 | /* typedef struct bunzip_data bunzip_data; -- done in .h file */ | ||
93 | niro | 532 | |
94 | |||
95 | /* Return the next nnn bits of input. All reads from the compressed input | ||
96 | are done through this function. All reads are big endian */ | ||
97 | |||
98 | niro | 816 | static unsigned get_bits(bunzip_data *bd, int bits_wanted) |
99 | niro | 532 | { |
100 | niro | 816 | unsigned bits = 0; |
101 | niro | 532 | |
102 | /* If we need to get more data from the byte buffer, do so. (Loop getting | ||
103 | one byte at a time to enforce endianness and avoid unaligned access.) */ | ||
104 | niro | 816 | while ((int)(bd->inbufBitCount) < bits_wanted) { |
105 | niro | 532 | |
106 | /* If we need to read more data from file into byte buffer, do so */ | ||
107 | niro | 816 | if (bd->inbufPos == bd->inbufCount) { |
108 | /* if "no input fd" case: in_fd == -1, read fails, we jump */ | ||
109 | bd->inbufCount = read(bd->in_fd, bd->inbuf, IOBUF_SIZE); | ||
110 | if (bd->inbufCount <= 0) | ||
111 | longjmp(bd->jmpbuf, RETVAL_UNEXPECTED_INPUT_EOF); | ||
112 | bd->inbufPos = 0; | ||
113 | niro | 532 | } |
114 | |||
115 | /* Avoid 32-bit overflow (dump bit buffer to top of output) */ | ||
116 | niro | 816 | if (bd->inbufBitCount >= 24) { |
117 | bits = bd->inbufBits & ((1 << bd->inbufBitCount) - 1); | ||
118 | bits_wanted -= bd->inbufBitCount; | ||
119 | bits <<= bits_wanted; | ||
120 | bd->inbufBitCount = 0; | ||
121 | niro | 532 | } |
122 | |||
123 | /* Grab next 8 bits of input from buffer. */ | ||
124 | niro | 816 | bd->inbufBits = (bd->inbufBits << 8) | bd->inbuf[bd->inbufPos++]; |
125 | bd->inbufBitCount += 8; | ||
126 | niro | 532 | } |
127 | |||
128 | /* Calculate result */ | ||
129 | niro | 816 | bd->inbufBitCount -= bits_wanted; |
130 | bits |= (bd->inbufBits >> bd->inbufBitCount) & ((1 << bits_wanted) - 1); | ||
131 | niro | 532 | |
132 | return bits; | ||
133 | } | ||
134 | |||
135 | /* Unpacks the next block and sets up for the inverse burrows-wheeler step. */ | ||
136 | static int get_next_block(bunzip_data *bd) | ||
137 | { | ||
138 | struct group_data *hufGroup; | ||
139 | niro | 816 | int dbufCount, nextSym, dbufSize, groupCount, *base, *limit, selector, |
140 | i, j, k, t, runPos, symCount, symTotal, nSelectors, byteCount[256]; | ||
141 | niro | 532 | unsigned char uc, symToByte[256], mtfSymbol[256], *selectors; |
142 | niro | 816 | unsigned *dbuf, origPtr; |
143 | niro | 532 | |
144 | niro | 816 | dbuf = bd->dbuf; |
145 | dbufSize = bd->dbufSize; | ||
146 | selectors = bd->selectors; | ||
147 | niro | 532 | |
148 | /* Reset longjmp I/O error handling */ | ||
149 | niro | 816 | i = setjmp(bd->jmpbuf); |
150 | niro | 532 | if (i) return i; |
151 | |||
152 | /* Read in header signature and CRC, then validate signature. | ||
153 | (last block signature means CRC is for whole file, return now) */ | ||
154 | niro | 816 | i = get_bits(bd, 24); |
155 | j = get_bits(bd, 24); | ||
156 | bd->headerCRC = get_bits(bd, 32); | ||
157 | niro | 532 | if ((i == 0x177245) && (j == 0x385090)) return RETVAL_LAST_BLOCK; |
158 | if ((i != 0x314159) || (j != 0x265359)) return RETVAL_NOT_BZIP_DATA; | ||
159 | |||
160 | /* We can add support for blockRandomised if anybody complains. There was | ||
161 | some code for this in busybox 1.0.0-pre3, but nobody ever noticed that | ||
162 | it didn't actually work. */ | ||
163 | niro | 816 | if (get_bits(bd, 1)) return RETVAL_OBSOLETE_INPUT; |
164 | origPtr = get_bits(bd, 24); | ||
165 | if ((int)origPtr > dbufSize) return RETVAL_DATA_ERROR; | ||
166 | niro | 532 | |
167 | /* mapping table: if some byte values are never used (encoding things | ||
168 | like ascii text), the compression code removes the gaps to have fewer | ||
169 | symbols to deal with, and writes a sparse bitfield indicating which | ||
170 | values were present. We make a translation table to convert the symbols | ||
171 | back to the corresponding bytes. */ | ||
172 | niro | 816 | t = get_bits(bd, 16); |
173 | symTotal = 0; | ||
174 | for (i = 0; i < 16; i++) { | ||
175 | if (t & (1 << (15-i))) { | ||
176 | k = get_bits(bd, 16); | ||
177 | for (j = 0; j < 16; j++) | ||
178 | if (k & (1 << (15-j))) | ||
179 | symToByte[symTotal++] = (16*i) + j; | ||
180 | niro | 532 | } |
181 | } | ||
182 | |||
183 | /* How many different Huffman coding groups does this block use? */ | ||
184 | niro | 816 | groupCount = get_bits(bd, 3); |
185 | if (groupCount < 2 || groupCount > MAX_GROUPS) | ||
186 | return RETVAL_DATA_ERROR; | ||
187 | niro | 532 | |
188 | /* nSelectors: Every GROUP_SIZE many symbols we select a new Huffman coding | ||
189 | group. Read in the group selector list, which is stored as MTF encoded | ||
190 | bit runs. (MTF=Move To Front, as each value is used it's moved to the | ||
191 | start of the list.) */ | ||
192 | niro | 816 | nSelectors = get_bits(bd, 15); |
193 | if (!nSelectors) return RETVAL_DATA_ERROR; | ||
194 | for (i = 0; i < groupCount; i++) mtfSymbol[i] = i; | ||
195 | for (i = 0; i < nSelectors; i++) { | ||
196 | niro | 532 | |
197 | /* Get next value */ | ||
198 | niro | 816 | for (j = 0; get_bits(bd, 1); j++) |
199 | if (j >= groupCount) return RETVAL_DATA_ERROR; | ||
200 | niro | 532 | |
201 | /* Decode MTF to get the next selector */ | ||
202 | uc = mtfSymbol[j]; | ||
203 | for (;j;j--) mtfSymbol[j] = mtfSymbol[j-1]; | ||
204 | niro | 816 | mtfSymbol[0] = selectors[i] = uc; |
205 | niro | 532 | } |
206 | |||
207 | /* Read the Huffman coding tables for each group, which code for symTotal | ||
208 | literal symbols, plus two run symbols (RUNA, RUNB) */ | ||
209 | niro | 816 | symCount = symTotal + 2; |
210 | for (j = 0; j < groupCount; j++) { | ||
211 | unsigned char length[MAX_SYMBOLS]; | ||
212 | /* 8 bits is ALMOST enough for temp[], see below */ | ||
213 | unsigned temp[MAX_HUFCODE_BITS+1]; | ||
214 | int minLen, maxLen, pp; | ||
215 | niro | 532 | |
216 | /* Read Huffman code lengths for each symbol. They're stored in | ||
217 | a way similar to mtf; record a starting value for the first symbol, | ||
218 | and an offset from the previous value for everys symbol after that. | ||
219 | (Subtracting 1 before the loop and then adding it back at the end is | ||
220 | an optimization that makes the test inside the loop simpler: symbol | ||
221 | length 0 becomes negative, so an unsigned inequality catches it.) */ | ||
222 | niro | 816 | t = get_bits(bd, 5) - 1; |
223 | niro | 532 | for (i = 0; i < symCount; i++) { |
224 | for (;;) { | ||
225 | niro | 816 | if ((unsigned)t > (MAX_HUFCODE_BITS-1)) |
226 | niro | 532 | return RETVAL_DATA_ERROR; |
227 | |||
228 | /* If first bit is 0, stop. Else second bit indicates whether | ||
229 | to increment or decrement the value. Optimization: grab 2 | ||
230 | bits and unget the second if the first was 0. */ | ||
231 | niro | 816 | k = get_bits(bd, 2); |
232 | niro | 532 | if (k < 2) { |
233 | bd->inbufBitCount++; | ||
234 | break; | ||
235 | } | ||
236 | |||
237 | /* Add one if second bit 1, else subtract 1. Avoids if/else */ | ||
238 | niro | 816 | t += (((k+1) & 2) - 1); |
239 | niro | 532 | } |
240 | |||
241 | /* Correct for the initial -1, to get the final symbol length */ | ||
242 | niro | 816 | length[i] = t + 1; |
243 | niro | 532 | } |
244 | |||
245 | /* Find largest and smallest lengths in this group */ | ||
246 | niro | 816 | minLen = maxLen = length[0]; |
247 | niro | 532 | for (i = 1; i < symCount; i++) { |
248 | niro | 816 | if (length[i] > maxLen) maxLen = length[i]; |
249 | else if (length[i] < minLen) minLen = length[i]; | ||
250 | niro | 532 | } |
251 | |||
252 | /* Calculate permute[], base[], and limit[] tables from length[]. | ||
253 | * | ||
254 | * permute[] is the lookup table for converting Huffman coded symbols | ||
255 | * into decoded symbols. base[] is the amount to subtract from the | ||
256 | * value of a Huffman symbol of a given length when using permute[]. | ||
257 | * | ||
258 | * limit[] indicates the largest numerical value a symbol with a given | ||
259 | * number of bits can have. This is how the Huffman codes can vary in | ||
260 | * length: each code with a value>limit[length] needs another bit. | ||
261 | */ | ||
262 | niro | 816 | hufGroup = bd->groups + j; |
263 | niro | 532 | hufGroup->minLen = minLen; |
264 | hufGroup->maxLen = maxLen; | ||
265 | |||
266 | /* Note that minLen can't be smaller than 1, so we adjust the base | ||
267 | and limit array pointers so we're not always wasting the first | ||
268 | entry. We do this again when using them (during symbol decoding).*/ | ||
269 | niro | 816 | base = hufGroup->base - 1; |
270 | limit = hufGroup->limit - 1; | ||
271 | niro | 532 | |
272 | /* Calculate permute[]. Concurently, initialize temp[] and limit[]. */ | ||
273 | niro | 816 | pp = 0; |
274 | for (i = minLen; i <= maxLen; i++) { | ||
275 | temp[i] = limit[i] = 0; | ||
276 | for (t = 0; t < symCount; t++) | ||
277 | if (length[t] == i) | ||
278 | hufGroup->permute[pp++] = t; | ||
279 | niro | 532 | } |
280 | |||
281 | /* Count symbols coded for at each bit length */ | ||
282 | niro | 816 | /* NB: in pathological cases, temp[8] can end ip being 256. |
283 | * That's why uint8_t is too small for temp[]. */ | ||
284 | for (i = 0; i < symCount; i++) temp[length[i]]++; | ||
285 | niro | 532 | |
286 | /* Calculate limit[] (the largest symbol-coding value at each bit | ||
287 | * length, which is (previous limit<<1)+symbols at this level), and | ||
288 | * base[] (number of symbols to ignore at each bit length, which is | ||
289 | * limit minus the cumulative count of symbols coded for already). */ | ||
290 | niro | 816 | pp = t = 0; |
291 | for (i = minLen; i < maxLen; i++) { | ||
292 | pp += temp[i]; | ||
293 | niro | 532 | |
294 | /* We read the largest possible symbol size and then unget bits | ||
295 | after determining how many we need, and those extra bits could | ||
296 | be set to anything. (They're noise from future symbols.) At | ||
297 | each level we're really only interested in the first few bits, | ||
298 | so here we set all the trailing to-be-ignored bits to 1 so they | ||
299 | don't affect the value>limit[length] comparison. */ | ||
300 | niro | 816 | limit[i] = (pp << (maxLen - i)) - 1; |
301 | pp <<= 1; | ||
302 | t += temp[i]; | ||
303 | base[i+1] = pp - t; | ||
304 | niro | 532 | } |
305 | niro | 816 | limit[maxLen+1] = INT_MAX; /* Sentinel value for reading next sym. */ |
306 | limit[maxLen] = pp + temp[maxLen] - 1; | ||
307 | base[minLen] = 0; | ||
308 | niro | 532 | } |
309 | |||
310 | /* We've finished reading and digesting the block header. Now read this | ||
311 | block's Huffman coded symbols from the file and undo the Huffman coding | ||
312 | niro | 816 | and run length encoding, saving the result into dbuf[dbufCount++] = uc */ |
313 | niro | 532 | |
314 | /* Initialize symbol occurrence counters and symbol Move To Front table */ | ||
315 | niro | 816 | memset(byteCount, 0, sizeof(byteCount)); /* smaller, maybe slower? */ |
316 | for (i = 0; i < 256; i++) { | ||
317 | //byteCount[i] = 0; | ||
318 | mtfSymbol[i] = (unsigned char)i; | ||
319 | niro | 532 | } |
320 | |||
321 | /* Loop through compressed symbols. */ | ||
322 | |||
323 | niro | 816 | runPos = dbufCount = selector = 0; |
324 | niro | 532 | for (;;) { |
325 | |||
326 | niro | 816 | /* Fetch next Huffman coding group from list. */ |
327 | symCount = GROUP_SIZE - 1; | ||
328 | if (selector >= nSelectors) return RETVAL_DATA_ERROR; | ||
329 | hufGroup = bd->groups + selectors[selector++]; | ||
330 | base = hufGroup->base - 1; | ||
331 | limit = hufGroup->limit - 1; | ||
332 | continue_this_group: | ||
333 | niro | 532 | |
334 | /* Read next Huffman-coded symbol. */ | ||
335 | |||
336 | /* Note: It is far cheaper to read maxLen bits and back up than it is | ||
337 | to read minLen bits and then an additional bit at a time, testing | ||
338 | as we go. Because there is a trailing last block (with file CRC), | ||
339 | there is no danger of the overread causing an unexpected EOF for a | ||
340 | valid compressed file. As a further optimization, we do the read | ||
341 | inline (falling back to a call to get_bits if the buffer runs | ||
342 | dry). The following (up to got_huff_bits:) is equivalent to | ||
343 | niro | 816 | j = get_bits(bd, hufGroup->maxLen); |
344 | niro | 532 | */ |
345 | niro | 816 | while ((int)(bd->inbufBitCount) < hufGroup->maxLen) { |
346 | if (bd->inbufPos == bd->inbufCount) { | ||
347 | j = get_bits(bd, hufGroup->maxLen); | ||
348 | niro | 532 | goto got_huff_bits; |
349 | } | ||
350 | niro | 816 | bd->inbufBits = (bd->inbufBits << 8) | bd->inbuf[bd->inbufPos++]; |
351 | bd->inbufBitCount += 8; | ||
352 | niro | 532 | }; |
353 | niro | 816 | bd->inbufBitCount -= hufGroup->maxLen; |
354 | j = (bd->inbufBits >> bd->inbufBitCount) & ((1 << hufGroup->maxLen) - 1); | ||
355 | niro | 532 | |
356 | niro | 816 | got_huff_bits: |
357 | niro | 532 | |
358 | /* Figure how how many bits are in next symbol and unget extras */ | ||
359 | niro | 816 | i = hufGroup->minLen; |
360 | while (j > limit[i]) ++i; | ||
361 | niro | 532 | bd->inbufBitCount += (hufGroup->maxLen - i); |
362 | |||
363 | /* Huffman decode value to get nextSym (with bounds checking) */ | ||
364 | niro | 816 | if (i > hufGroup->maxLen) |
365 | niro | 532 | return RETVAL_DATA_ERROR; |
366 | niro | 816 | j = (j >> (hufGroup->maxLen - i)) - base[i]; |
367 | if ((unsigned)j >= MAX_SYMBOLS) | ||
368 | return RETVAL_DATA_ERROR; | ||
369 | niro | 532 | nextSym = hufGroup->permute[j]; |
370 | |||
371 | /* We have now decoded the symbol, which indicates either a new literal | ||
372 | byte, or a repeated run of the most recent literal byte. First, | ||
373 | check if nextSym indicates a repeated run, and if so loop collecting | ||
374 | how many times to repeat the last literal. */ | ||
375 | niro | 816 | if ((unsigned)nextSym <= SYMBOL_RUNB) { /* RUNA or RUNB */ |
376 | niro | 532 | |
377 | /* If this is the start of a new run, zero out counter */ | ||
378 | niro | 816 | if (!runPos) { |
379 | niro | 532 | runPos = 1; |
380 | t = 0; | ||
381 | } | ||
382 | |||
383 | /* Neat trick that saves 1 symbol: instead of or-ing 0 or 1 at | ||
384 | each bit position, add 1 or 2 instead. For example, | ||
385 | 1011 is 1<<0 + 1<<1 + 2<<2. 1010 is 2<<0 + 2<<1 + 1<<2. | ||
386 | You can make any bit pattern that way using 1 less symbol than | ||
387 | the basic or 0/1 method (except all bits 0, which would use no | ||
388 | symbols, but a run of length 0 doesn't mean anything in this | ||
389 | context). Thus space is saved. */ | ||
390 | t += (runPos << nextSym); /* +runPos if RUNA; +2*runPos if RUNB */ | ||
391 | niro | 816 | if (runPos < dbufSize) runPos <<= 1; |
392 | niro | 532 | goto end_of_huffman_loop; |
393 | } | ||
394 | |||
395 | /* When we hit the first non-run symbol after a run, we now know | ||
396 | how many times to repeat the last literal, so append that many | ||
397 | copies to our buffer of decoded symbols (dbuf) now. (The last | ||
398 | literal used is the one at the head of the mtfSymbol array.) */ | ||
399 | niro | 816 | if (runPos) { |
400 | runPos = 0; | ||
401 | if (dbufCount + t >= dbufSize) return RETVAL_DATA_ERROR; | ||
402 | niro | 532 | |
403 | uc = symToByte[mtfSymbol[0]]; | ||
404 | byteCount[uc] += t; | ||
405 | niro | 816 | while (t--) dbuf[dbufCount++] = uc; |
406 | niro | 532 | } |
407 | |||
408 | /* Is this the terminating symbol? */ | ||
409 | niro | 816 | if (nextSym > symTotal) break; |
410 | niro | 532 | |
411 | /* At this point, nextSym indicates a new literal character. Subtract | ||
412 | one to get the position in the MTF array at which this literal is | ||
413 | currently to be found. (Note that the result can't be -1 or 0, | ||
414 | because 0 and 1 are RUNA and RUNB. But another instance of the | ||
415 | first symbol in the mtf array, position 0, would have been handled | ||
416 | as part of a run above. Therefore 1 unused mtf position minus | ||
417 | 2 non-literal nextSym values equals -1.) */ | ||
418 | niro | 816 | if (dbufCount >= dbufSize) return RETVAL_DATA_ERROR; |
419 | niro | 532 | i = nextSym - 1; |
420 | uc = mtfSymbol[i]; | ||
421 | |||
422 | /* Adjust the MTF array. Since we typically expect to move only a | ||
423 | * small number of symbols, and are bound by 256 in any case, using | ||
424 | * memmove here would typically be bigger and slower due to function | ||
425 | * call overhead and other assorted setup costs. */ | ||
426 | do { | ||
427 | mtfSymbol[i] = mtfSymbol[i-1]; | ||
428 | } while (--i); | ||
429 | mtfSymbol[0] = uc; | ||
430 | niro | 816 | uc = symToByte[uc]; |
431 | niro | 532 | |
432 | /* We have our literal byte. Save it into dbuf. */ | ||
433 | byteCount[uc]++; | ||
434 | niro | 816 | dbuf[dbufCount++] = (unsigned)uc; |
435 | niro | 532 | |
436 | /* Skip group initialization if we're not done with this group. Done | ||
437 | * this way to avoid compiler warning. */ | ||
438 | niro | 816 | end_of_huffman_loop: |
439 | if (symCount--) goto continue_this_group; | ||
440 | niro | 532 | } |
441 | |||
442 | /* At this point, we've read all the Huffman-coded symbols (and repeated | ||
443 | niro | 816 | runs) for this block from the input stream, and decoded them into the |
444 | niro | 532 | intermediate buffer. There are dbufCount many decoded bytes in dbuf[]. |
445 | Now undo the Burrows-Wheeler transform on dbuf. | ||
446 | See http://dogma.net/markn/articles/bwt/bwt.htm | ||
447 | */ | ||
448 | |||
449 | /* Turn byteCount into cumulative occurrence counts of 0 to n-1. */ | ||
450 | niro | 816 | j = 0; |
451 | for (i = 0; i < 256; i++) { | ||
452 | k = j + byteCount[i]; | ||
453 | niro | 532 | byteCount[i] = j; |
454 | niro | 816 | j = k; |
455 | niro | 532 | } |
456 | |||
457 | /* Figure out what order dbuf would be in if we sorted it. */ | ||
458 | niro | 816 | for (i = 0; i < dbufCount; i++) { |
459 | uc = (unsigned char)(dbuf[i] & 0xff); | ||
460 | niro | 532 | dbuf[byteCount[uc]] |= (i << 8); |
461 | byteCount[uc]++; | ||
462 | } | ||
463 | |||
464 | /* Decode first byte by hand to initialize "previous" byte. Note that it | ||
465 | doesn't get output, and if the first three characters are identical | ||
466 | it doesn't qualify as a run (hence writeRunCountdown=5). */ | ||
467 | niro | 816 | if (dbufCount) { |
468 | if ((int)origPtr >= dbufCount) return RETVAL_DATA_ERROR; | ||
469 | bd->writePos = dbuf[origPtr]; | ||
470 | bd->writeCurrent = (unsigned char)(bd->writePos & 0xff); | ||
471 | bd->writePos >>= 8; | ||
472 | bd->writeRunCountdown = 5; | ||
473 | niro | 532 | } |
474 | niro | 816 | bd->writeCount = dbufCount; |
475 | niro | 532 | |
476 | return RETVAL_OK; | ||
477 | } | ||
478 | |||
479 | /* Undo burrows-wheeler transform on intermediate buffer to produce output. | ||
480 | If start_bunzip was initialized with out_fd=-1, then up to len bytes of | ||
481 | data are written to outbuf. Return value is number of bytes written or | ||
482 | error (all errors are negative numbers). If out_fd!=-1, outbuf and len | ||
483 | are ignored, data is written to out_fd and return is RETVAL_OK or error. | ||
484 | */ | ||
485 | niro | 816 | int FAST_FUNC read_bunzip(bunzip_data *bd, char *outbuf, int len) |
486 | niro | 532 | { |
487 | niro | 816 | const unsigned *dbuf; |
488 | int pos, current, previous, gotcount; | ||
489 | niro | 532 | |
490 | /* If last read was short due to end of file, return last block now */ | ||
491 | niro | 816 | if (bd->writeCount < 0) return bd->writeCount; |
492 | niro | 532 | |
493 | gotcount = 0; | ||
494 | niro | 816 | dbuf = bd->dbuf; |
495 | pos = bd->writePos; | ||
496 | current = bd->writeCurrent; | ||
497 | niro | 532 | |
498 | /* We will always have pending decoded data to write into the output | ||
499 | buffer unless this is the very first call (in which case we haven't | ||
500 | Huffman-decoded a block into the intermediate buffer yet). */ | ||
501 | if (bd->writeCopies) { | ||
502 | |||
503 | /* Inside the loop, writeCopies means extra copies (beyond 1) */ | ||
504 | --bd->writeCopies; | ||
505 | |||
506 | /* Loop outputting bytes */ | ||
507 | for (;;) { | ||
508 | |||
509 | /* If the output buffer is full, snapshot state and return */ | ||
510 | niro | 816 | if (gotcount >= len) { |
511 | bd->writePos = pos; | ||
512 | bd->writeCurrent = current; | ||
513 | niro | 532 | bd->writeCopies++; |
514 | return len; | ||
515 | } | ||
516 | |||
517 | /* Write next byte into output buffer, updating CRC */ | ||
518 | outbuf[gotcount++] = current; | ||
519 | niro | 816 | bd->writeCRC = (bd->writeCRC << 8) |
520 | ^ bd->crc32Table[(bd->writeCRC >> 24) ^ current]; | ||
521 | niro | 532 | |
522 | /* Loop now if we're outputting multiple copies of this byte */ | ||
523 | if (bd->writeCopies) { | ||
524 | --bd->writeCopies; | ||
525 | continue; | ||
526 | } | ||
527 | niro | 816 | decode_next_byte: |
528 | niro | 532 | if (!bd->writeCount--) break; |
529 | /* Follow sequence vector to undo Burrows-Wheeler transform */ | ||
530 | niro | 816 | previous = current; |
531 | pos = dbuf[pos]; | ||
532 | current = pos & 0xff; | ||
533 | pos >>= 8; | ||
534 | niro | 532 | |
535 | niro | 816 | /* After 3 consecutive copies of the same byte, the 4th |
536 | * is a repeat count. We count down from 4 instead | ||
537 | niro | 532 | * of counting up because testing for non-zero is faster */ |
538 | niro | 816 | if (--bd->writeRunCountdown) { |
539 | if (current != previous) | ||
540 | bd->writeRunCountdown = 4; | ||
541 | niro | 532 | } else { |
542 | |||
543 | /* We have a repeated run, this byte indicates the count */ | ||
544 | niro | 816 | bd->writeCopies = current; |
545 | current = previous; | ||
546 | bd->writeRunCountdown = 5; | ||
547 | niro | 532 | |
548 | /* Sometimes there are just 3 bytes (run length 0) */ | ||
549 | niro | 816 | if (!bd->writeCopies) goto decode_next_byte; |
550 | niro | 532 | |
551 | /* Subtract the 1 copy we'd output anyway to get extras */ | ||
552 | --bd->writeCopies; | ||
553 | } | ||
554 | } | ||
555 | |||
556 | /* Decompression of this block completed successfully */ | ||
557 | niro | 816 | bd->writeCRC = ~bd->writeCRC; |
558 | bd->totalCRC = ((bd->totalCRC << 1) | (bd->totalCRC >> 31)) ^ bd->writeCRC; | ||
559 | niro | 532 | |
560 | /* If this block had a CRC error, force file level CRC error. */ | ||
561 | niro | 816 | if (bd->writeCRC != bd->headerCRC) { |
562 | bd->totalCRC = bd->headerCRC + 1; | ||
563 | niro | 532 | return RETVAL_LAST_BLOCK; |
564 | } | ||
565 | } | ||
566 | |||
567 | /* Refill the intermediate buffer by Huffman-decoding next block of input */ | ||
568 | /* (previous is just a convenient unused temp variable here) */ | ||
569 | niro | 816 | previous = get_next_block(bd); |
570 | if (previous) { | ||
571 | bd->writeCount = previous; | ||
572 | return (previous != RETVAL_LAST_BLOCK) ? previous : gotcount; | ||
573 | niro | 532 | } |
574 | niro | 816 | bd->writeCRC = ~0; |
575 | pos = bd->writePos; | ||
576 | current = bd->writeCurrent; | ||
577 | niro | 532 | goto decode_next_byte; |
578 | } | ||
579 | |||
580 | /* Allocate the structure, read file header. If in_fd==-1, inbuf must contain | ||
581 | a complete bunzip file (len bytes long). If in_fd!=-1, inbuf and len are | ||
582 | ignored, and data is read from file handle into temporary buffer. */ | ||
583 | |||
584 | niro | 816 | /* Because bunzip2 is used for help text unpacking, and because bb_show_usage() |
585 | should work for NOFORK applets too, we must be extremely careful to not leak | ||
586 | any allocations! */ | ||
587 | int FAST_FUNC start_bunzip(bunzip_data **bdp, int in_fd, const unsigned char *inbuf, | ||
588 | niro | 532 | int len) |
589 | { | ||
590 | bunzip_data *bd; | ||
591 | niro | 816 | unsigned i; |
592 | enum { | ||
593 | BZh0 = ('B' << 24) + ('Z' << 16) + ('h' << 8) + '0', | ||
594 | h0 = ('h' << 8) + '0', | ||
595 | }; | ||
596 | niro | 532 | |
597 | /* Figure out how much data to allocate */ | ||
598 | niro | 816 | i = sizeof(bunzip_data); |
599 | if (in_fd != -1) i += IOBUF_SIZE; | ||
600 | niro | 532 | |
601 | /* Allocate bunzip_data. Most fields initialize to zero. */ | ||
602 | niro | 816 | bd = *bdp = xzalloc(i); |
603 | niro | 532 | |
604 | /* Setup input buffer */ | ||
605 | niro | 816 | bd->in_fd = in_fd; |
606 | if (-1 == in_fd) { | ||
607 | /* in this case, bd->inbuf is read-only */ | ||
608 | bd->inbuf = (void*)inbuf; /* cast away const-ness */ | ||
609 | bd->inbufCount = len; | ||
610 | } else | ||
611 | bd->inbuf = (unsigned char *)(bd + 1); | ||
612 | niro | 532 | |
613 | /* Init the CRC32 table (big endian) */ | ||
614 | niro | 816 | crc32_filltable(bd->crc32Table, 1); |
615 | niro | 532 | |
616 | /* Setup for I/O error handling via longjmp */ | ||
617 | niro | 816 | i = setjmp(bd->jmpbuf); |
618 | if (i) return i; | ||
619 | niro | 532 | |
620 | /* Ensure that file starts with "BZh['1'-'9']." */ | ||
621 | niro | 816 | /* Update: now caller verifies 1st two bytes, makes .gz/.bz2 |
622 | * integration easier */ | ||
623 | /* was: */ | ||
624 | /* i = get_bits(bd, 32); */ | ||
625 | /* if ((unsigned)(i - BZh0 - 1) >= 9) return RETVAL_NOT_BZIP_DATA; */ | ||
626 | i = get_bits(bd, 16); | ||
627 | if ((unsigned)(i - h0 - 1) >= 9) return RETVAL_NOT_BZIP_DATA; | ||
628 | niro | 532 | |
629 | niro | 816 | /* Fourth byte (ascii '1'-'9') indicates block size in units of 100k of |
630 | niro | 532 | uncompressed data. Allocate intermediate buffer for block. */ |
631 | niro | 816 | /* bd->dbufSize = 100000 * (i - BZh0); */ |
632 | bd->dbufSize = 100000 * (i - h0); | ||
633 | niro | 532 | |
634 | niro | 816 | /* Cannot use xmalloc - may leak bd in NOFORK case! */ |
635 | bd->dbuf = malloc_or_warn(bd->dbufSize * sizeof(int)); | ||
636 | if (!bd->dbuf) { | ||
637 | free(bd); | ||
638 | xfunc_die(); | ||
639 | } | ||
640 | niro | 532 | return RETVAL_OK; |
641 | } | ||
642 | |||
643 | niro | 816 | void FAST_FUNC dealloc_bunzip(bunzip_data *bd) |
644 | { | ||
645 | free(bd->dbuf); | ||
646 | free(bd); | ||
647 | } | ||
648 | niro | 532 | |
649 | niro | 816 | |
650 | /* Decompress src_fd to dst_fd. Stops at end of bzip data, not end of file. */ | ||
651 | niro | 984 | IF_DESKTOP(long long) int FAST_FUNC |
652 | niro | 816 | unpack_bz2_stream(int src_fd, int dst_fd) |
653 | niro | 532 | { |
654 | niro | 984 | IF_DESKTOP(long long total_written = 0;) |
655 | niro | 532 | char *outbuf; |
656 | bunzip_data *bd; | ||
657 | int i; | ||
658 | |||
659 | niro | 816 | outbuf = xmalloc(IOBUF_SIZE); |
660 | i = start_bunzip(&bd, src_fd, NULL, 0); | ||
661 | if (!i) { | ||
662 | niro | 532 | for (;;) { |
663 | niro | 816 | i = read_bunzip(bd, outbuf, IOBUF_SIZE); |
664 | if (i <= 0) break; | ||
665 | if (i != full_write(dst_fd, outbuf, i)) { | ||
666 | i = RETVAL_SHORT_WRITE; | ||
667 | niro | 532 | break; |
668 | } | ||
669 | niro | 984 | IF_DESKTOP(total_written += i;) |
670 | niro | 532 | } |
671 | } | ||
672 | |||
673 | /* Check CRC and release memory */ | ||
674 | |||
675 | niro | 816 | if (i == RETVAL_LAST_BLOCK) { |
676 | if (bd->headerCRC != bd->totalCRC) { | ||
677 | bb_error_msg("CRC error"); | ||
678 | niro | 532 | } else { |
679 | niro | 816 | i = RETVAL_OK; |
680 | niro | 532 | } |
681 | niro | 816 | } else if (i == RETVAL_SHORT_WRITE) { |
682 | bb_error_msg("short write"); | ||
683 | niro | 532 | } else { |
684 | niro | 816 | bb_error_msg("bunzip error %d", i); |
685 | niro | 532 | } |
686 | niro | 816 | dealloc_bunzip(bd); |
687 | niro | 532 | free(outbuf); |
688 | |||
689 | niro | 984 | return i ? i : IF_DESKTOP(total_written) + 0; |
690 | niro | 532 | } |
691 | |||
692 | niro | 984 | IF_DESKTOP(long long) int FAST_FUNC |
693 | niro | 816 | unpack_bz2_stream_prime(int src_fd, int dst_fd) |
694 | { | ||
695 | niro | 1123 | uint16_t magic2; |
696 | xread(src_fd, &magic2, 2); | ||
697 | if (magic2 != BZIP2_MAGIC) { | ||
698 | niro | 816 | bb_error_msg_and_die("invalid magic"); |
699 | } | ||
700 | return unpack_bz2_stream(src_fd, dst_fd); | ||
701 | } | ||
702 | |||
703 | niro | 532 | #ifdef TESTING |
704 | |||
705 | niro | 816 | static char *const bunzip_errors[] = { |
706 | NULL, "Bad file checksum", "Not bzip data", | ||
707 | "Unexpected input EOF", "Unexpected output EOF", "Data error", | ||
708 | "Out of memory", "Obsolete (pre 0.9.5) bzip format not supported" | ||
709 | }; | ||
710 | niro | 532 | |
711 | /* Dumb little test thing, decompress stdin to stdout */ | ||
712 | niro | 816 | int main(int argc, char **argv) |
713 | niro | 532 | { |
714 | niro | 816 | int i; |
715 | niro | 532 | char c; |
716 | |||
717 | niro | 816 | int i = unpack_bz2_stream_prime(0, 1); |
718 | if (i < 0) | ||
719 | fprintf(stderr, "%s\n", bunzip_errors[-i]); | ||
720 | else if (read(STDIN_FILENO, &c, 1)) | ||
721 | fprintf(stderr, "Trailing garbage ignored\n"); | ||
722 | niro | 532 | return -i; |
723 | } | ||
724 | #endif |