aboutsummaryrefslogtreecommitdiffstats
path: root/lib
diff options
context:
space:
mode:
authorAlain Knaff <alain@knaff.lu>2009-01-04 16:46:16 -0500
committerH. Peter Anvin <hpa@zytor.com>2009-01-04 18:53:34 -0500
commitbc22c17e12c130dc929218a95aa347e0f3fd05dc (patch)
treee5dfd433dbf2fec27a033ee729236e63fbe3a1ad /lib
parent7d3b56ba37a95f1f370f50258ed3954c304c524b (diff)
bzip2/lzma: library support for gzip, bzip2 and lzma decompression
Impact: Replaces inflate.c with a wrapper around zlib_inflate; new library code This is the first part of the bzip2/lzma patch The bzip patch is based on an idea by Christian Ludwig, includes support for compressing the kernel with bzip2 or lzma rather than gzip. Both compressors give smaller sizes than gzip. Lzma's decompresses faster than bzip2. It also supports ramdisks and initramfs' compressed using these two compressors. The functionality has been successfully used for a couple of years by the udpcast project This version applies to "tip" kernel 2.6.28 This part contains: - changed inflate.c to accomodate rest of patch - implementation of bzip2 compression (not used at this stage yet) - implementation of lzma compression (not used at this stage yet) - Makefile routines to support bzip2 and lzma kernel compression Signed-off-by: Alain Knaff <alain@knaff.lu> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Diffstat (limited to 'lib')
-rw-r--r--lib/decompress_bunzip2.c735
-rw-r--r--lib/decompress_inflate.c167
-rw-r--r--lib/decompress_unlzma.c647
-rw-r--r--lib/zlib_inflate/inflate.h4
-rw-r--r--lib/zlib_inflate/inftrees.h4
5 files changed, 1557 insertions, 0 deletions
diff --git a/lib/decompress_bunzip2.c b/lib/decompress_bunzip2.c
new file mode 100644
index 000000000000..5d3ddb5fcfd9
--- /dev/null
+++ b/lib/decompress_bunzip2.c
@@ -0,0 +1,735 @@
1/* 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 This code is licensed under the LGPLv2:
10 LGPL (http://www.gnu.org/copyleft/lgpl.html
11*/
12
13/*
14 Size and speed optimizations by Manuel Novoa III (mjn3@codepoet.org).
15
16 More efficient reading of Huffman codes, a streamlined read_bunzip()
17 function, and various other tweaks. In (limited) tests, approximately
18 20% faster than bzcat on x86 and about 10% faster on arm.
19
20 Note that about 2/3 of the time is spent in read_unzip() reversing
21 the Burrows-Wheeler transformation. Much of that time is delay
22 resulting from cache misses.
23
24 I would ask that anyone benefiting from this work, especially those
25 using it in commercial products, consider making a donation to my local
26 non-profit hospice organization in the name of the woman I loved, who
27 passed away Feb. 12, 2003.
28
29 In memory of Toni W. Hagan
30
31 Hospice of Acadiana, Inc.
32 2600 Johnston St., Suite 200
33 Lafayette, LA 70503-3240
34
35 Phone (337) 232-1234 or 1-800-738-2226
36 Fax (337) 232-1297
37
38 http://www.hospiceacadiana.com/
39
40 Manuel
41 */
42
43/*
44 Made it fit for running in Linux Kernel by Alain Knaff (alain@knaff.lu)
45*/
46
47
48#ifndef STATIC
49#include <linux/decompress/bunzip2.h>
50#endif /* !STATIC */
51
52#include <linux/decompress/mm.h>
53
54#ifndef INT_MAX
55#define INT_MAX 0x7fffffff
56#endif
57
58/* Constants for Huffman coding */
59#define MAX_GROUPS 6
60#define GROUP_SIZE 50 /* 64 would have been more efficient */
61#define MAX_HUFCODE_BITS 20 /* Longest Huffman code allowed */
62#define MAX_SYMBOLS 258 /* 256 literals + RUNA + RUNB */
63#define SYMBOL_RUNA 0
64#define SYMBOL_RUNB 1
65
66/* Status return values */
67#define RETVAL_OK 0
68#define RETVAL_LAST_BLOCK (-1)
69#define RETVAL_NOT_BZIP_DATA (-2)
70#define RETVAL_UNEXPECTED_INPUT_EOF (-3)
71#define RETVAL_UNEXPECTED_OUTPUT_EOF (-4)
72#define RETVAL_DATA_ERROR (-5)
73#define RETVAL_OUT_OF_MEMORY (-6)
74#define RETVAL_OBSOLETE_INPUT (-7)
75
76/* Other housekeeping constants */
77#define BZIP2_IOBUF_SIZE 4096
78
79/* This is what we know about each Huffman coding group */
80struct group_data {
81 /* We have an extra slot at the end of limit[] for a sentinal value. */
82 int limit[MAX_HUFCODE_BITS+1];
83 int base[MAX_HUFCODE_BITS];
84 int permute[MAX_SYMBOLS];
85 int minLen, maxLen;
86};
87
88/* Structure holding all the housekeeping data, including IO buffers and
89 memory that persists between calls to bunzip */
90struct bunzip_data {
91 /* State for interrupting output loop */
92 int writeCopies, writePos, writeRunCountdown, writeCount, writeCurrent;
93 /* I/O tracking data (file handles, buffers, positions, etc.) */
94 int (*fill)(void*, unsigned int);
95 int inbufCount, inbufPos /*, outbufPos*/;
96 unsigned char *inbuf /*,*outbuf*/;
97 unsigned int inbufBitCount, inbufBits;
98 /* The CRC values stored in the block header and calculated from the
99 data */
100 unsigned int crc32Table[256], headerCRC, totalCRC, writeCRC;
101 /* Intermediate buffer and its size (in bytes) */
102 unsigned int *dbuf, dbufSize;
103 /* These things are a bit too big to go on the stack */
104 unsigned char selectors[32768]; /* nSelectors = 15 bits */
105 struct group_data groups[MAX_GROUPS]; /* Huffman coding tables */
106 int io_error; /* non-zero if we have IO error */
107};
108
109
110/* Return the next nnn bits of input. All reads from the compressed input
111 are done through this function. All reads are big endian */
112static unsigned int INIT get_bits(struct bunzip_data *bd, char bits_wanted)
113{
114 unsigned int bits = 0;
115
116 /* If we need to get more data from the byte buffer, do so.
117 (Loop getting one byte at a time to enforce endianness and avoid
118 unaligned access.) */
119 while (bd->inbufBitCount < bits_wanted) {
120 /* If we need to read more data from file into byte buffer, do
121 so */
122 if (bd->inbufPos == bd->inbufCount) {
123 if (bd->io_error)
124 return 0;
125 bd->inbufCount = bd->fill(bd->inbuf, BZIP2_IOBUF_SIZE);
126 if (bd->inbufCount <= 0) {
127 bd->io_error = RETVAL_UNEXPECTED_INPUT_EOF;
128 return 0;
129 }
130 bd->inbufPos = 0;
131 }
132 /* Avoid 32-bit overflow (dump bit buffer to top of output) */
133 if (bd->inbufBitCount >= 24) {
134 bits = bd->inbufBits&((1 << bd->inbufBitCount)-1);
135 bits_wanted -= bd->inbufBitCount;
136 bits <<= bits_wanted;
137 bd->inbufBitCount = 0;
138 }
139 /* Grab next 8 bits of input from buffer. */
140 bd->inbufBits = (bd->inbufBits << 8)|bd->inbuf[bd->inbufPos++];
141 bd->inbufBitCount += 8;
142 }
143 /* Calculate result */
144 bd->inbufBitCount -= bits_wanted;
145 bits |= (bd->inbufBits >> bd->inbufBitCount)&((1 << bits_wanted)-1);
146
147 return bits;
148}
149
150/* Unpacks the next block and sets up for the inverse burrows-wheeler step. */
151
152static int INIT get_next_block(struct bunzip_data *bd)
153{
154 struct group_data *hufGroup = NULL;
155 int *base = NULL;
156 int *limit = NULL;
157 int dbufCount, nextSym, dbufSize, groupCount, selector,
158 i, j, k, t, runPos, symCount, symTotal, nSelectors,
159 byteCount[256];
160 unsigned char uc, symToByte[256], mtfSymbol[256], *selectors;
161 unsigned int *dbuf, origPtr;
162
163 dbuf = bd->dbuf;
164 dbufSize = bd->dbufSize;
165 selectors = bd->selectors;
166
167 /* Read in header signature and CRC, then validate signature.
168 (last block signature means CRC is for whole file, return now) */
169 i = get_bits(bd, 24);
170 j = get_bits(bd, 24);
171 bd->headerCRC = get_bits(bd, 32);
172 if ((i == 0x177245) && (j == 0x385090))
173 return RETVAL_LAST_BLOCK;
174 if ((i != 0x314159) || (j != 0x265359))
175 return RETVAL_NOT_BZIP_DATA;
176 /* We can add support for blockRandomised if anybody complains.
177 There was some code for this in busybox 1.0.0-pre3, but nobody ever
178 noticed that it didn't actually work. */
179 if (get_bits(bd, 1))
180 return RETVAL_OBSOLETE_INPUT;
181 origPtr = get_bits(bd, 24);
182 if (origPtr > dbufSize)
183 return RETVAL_DATA_ERROR;
184 /* mapping table: if some byte values are never used (encoding things
185 like ascii text), the compression code removes the gaps to have fewer
186 symbols to deal with, and writes a sparse bitfield indicating which
187 values were present. We make a translation table to convert the
188 symbols back to the corresponding bytes. */
189 t = get_bits(bd, 16);
190 symTotal = 0;
191 for (i = 0; i < 16; i++) {
192 if (t&(1 << (15-i))) {
193 k = get_bits(bd, 16);
194 for (j = 0; j < 16; j++)
195 if (k&(1 << (15-j)))
196 symToByte[symTotal++] = (16*i)+j;
197 }
198 }
199 /* How many different Huffman coding groups does this block use? */
200 groupCount = get_bits(bd, 3);
201 if (groupCount < 2 || groupCount > MAX_GROUPS)
202 return RETVAL_DATA_ERROR;
203 /* nSelectors: Every GROUP_SIZE many symbols we select a new
204 Huffman coding group. Read in the group selector list,
205 which is stored as MTF encoded bit runs. (MTF = Move To
206 Front, as each value is used it's moved to the start of the
207 list.) */
208 nSelectors = get_bits(bd, 15);
209 if (!nSelectors)
210 return RETVAL_DATA_ERROR;
211 for (i = 0; i < groupCount; i++)
212 mtfSymbol[i] = i;
213 for (i = 0; i < nSelectors; i++) {
214 /* Get next value */
215 for (j = 0; get_bits(bd, 1); j++)
216 if (j >= groupCount)
217 return RETVAL_DATA_ERROR;
218 /* Decode MTF to get the next selector */
219 uc = mtfSymbol[j];
220 for (; j; j--)
221 mtfSymbol[j] = mtfSymbol[j-1];
222 mtfSymbol[0] = selectors[i] = uc;
223 }
224 /* Read the Huffman coding tables for each group, which code
225 for symTotal literal symbols, plus two run symbols (RUNA,
226 RUNB) */
227 symCount = symTotal+2;
228 for (j = 0; j < groupCount; j++) {
229 unsigned char length[MAX_SYMBOLS], temp[MAX_HUFCODE_BITS+1];
230 int minLen, maxLen, pp;
231 /* Read Huffman code lengths for each symbol. They're
232 stored in a way similar to mtf; record a starting
233 value for the first symbol, and an offset from the
234 previous value for everys symbol after that.
235 (Subtracting 1 before the loop and then adding it
236 back at the end is an optimization that makes the
237 test inside the loop simpler: symbol length 0
238 becomes negative, so an unsigned inequality catches
239 it.) */
240 t = get_bits(bd, 5)-1;
241 for (i = 0; i < symCount; i++) {
242 for (;;) {
243 if (((unsigned)t) > (MAX_HUFCODE_BITS-1))
244 return RETVAL_DATA_ERROR;
245
246 /* If first bit is 0, stop. Else
247 second bit indicates whether to
248 increment or decrement the value.
249 Optimization: grab 2 bits and unget
250 the second if the first was 0. */
251
252 k = get_bits(bd, 2);
253 if (k < 2) {
254 bd->inbufBitCount++;
255 break;
256 }
257 /* Add one if second bit 1, else
258 * subtract 1. Avoids if/else */
259 t += (((k+1)&2)-1);
260 }
261 /* Correct for the initial -1, to get the
262 * final symbol length */
263 length[i] = t+1;
264 }
265 /* Find largest and smallest lengths in this group */
266 minLen = maxLen = length[0];
267
268 for (i = 1; i < symCount; i++) {
269 if (length[i] > maxLen)
270 maxLen = length[i];
271 else if (length[i] < minLen)
272 minLen = length[i];
273 }
274
275 /* Calculate permute[], base[], and limit[] tables from
276 * length[].
277 *
278 * permute[] is the lookup table for converting
279 * Huffman coded symbols into decoded symbols. base[]
280 * is the amount to subtract from the value of a
281 * Huffman symbol of a given length when using
282 * permute[].
283 *
284 * limit[] indicates the largest numerical value a
285 * symbol with a given number of bits can have. This
286 * is how the Huffman codes can vary in length: each
287 * code with a value > limit[length] needs another
288 * bit.
289 */
290 hufGroup = bd->groups+j;
291 hufGroup->minLen = minLen;
292 hufGroup->maxLen = maxLen;
293 /* Note that minLen can't be smaller than 1, so we
294 adjust the base and limit array pointers so we're
295 not always wasting the first entry. We do this
296 again when using them (during symbol decoding).*/
297 base = hufGroup->base-1;
298 limit = hufGroup->limit-1;
299 /* Calculate permute[]. Concurently, initialize
300 * temp[] and limit[]. */
301 pp = 0;
302 for (i = minLen; i <= maxLen; i++) {
303 temp[i] = limit[i] = 0;
304 for (t = 0; t < symCount; t++)
305 if (length[t] == i)
306 hufGroup->permute[pp++] = t;
307 }
308 /* Count symbols coded for at each bit length */
309 for (i = 0; i < symCount; i++)
310 temp[length[i]]++;
311 /* Calculate limit[] (the largest symbol-coding value
312 *at each bit length, which is (previous limit <<
313 *1)+symbols at this level), and base[] (number of
314 *symbols to ignore at each bit length, which is limit
315 *minus the cumulative count of symbols coded for
316 *already). */
317 pp = t = 0;
318 for (i = minLen; i < maxLen; i++) {
319 pp += temp[i];
320 /* We read the largest possible symbol size
321 and then unget bits after determining how
322 many we need, and those extra bits could be
323 set to anything. (They're noise from
324 future symbols.) At each level we're
325 really only interested in the first few
326 bits, so here we set all the trailing
327 to-be-ignored bits to 1 so they don't
328 affect the value > limit[length]
329 comparison. */
330 limit[i] = (pp << (maxLen - i)) - 1;
331 pp <<= 1;
332 base[i+1] = pp-(t += temp[i]);
333 }
334 limit[maxLen+1] = INT_MAX; /* Sentinal value for
335 * reading next sym. */
336 limit[maxLen] = pp+temp[maxLen]-1;
337 base[minLen] = 0;
338 }
339 /* We've finished reading and digesting the block header. Now
340 read this block's Huffman coded symbols from the file and
341 undo the Huffman coding and run length encoding, saving the
342 result into dbuf[dbufCount++] = uc */
343
344 /* Initialize symbol occurrence counters and symbol Move To
345 * Front table */
346 for (i = 0; i < 256; i++) {
347 byteCount[i] = 0;
348 mtfSymbol[i] = (unsigned char)i;
349 }
350 /* Loop through compressed symbols. */
351 runPos = dbufCount = symCount = selector = 0;
352 for (;;) {
353 /* Determine which Huffman coding group to use. */
354 if (!(symCount--)) {
355 symCount = GROUP_SIZE-1;
356 if (selector >= nSelectors)
357 return RETVAL_DATA_ERROR;
358 hufGroup = bd->groups+selectors[selector++];
359 base = hufGroup->base-1;
360 limit = hufGroup->limit-1;
361 }
362 /* Read next Huffman-coded symbol. */
363 /* Note: It is far cheaper to read maxLen bits and
364 back up than it is to read minLen bits and then an
365 additional bit at a time, testing as we go.
366 Because there is a trailing last block (with file
367 CRC), there is no danger of the overread causing an
368 unexpected EOF for a valid compressed file. As a
369 further optimization, we do the read inline
370 (falling back to a call to get_bits if the buffer
371 runs dry). The following (up to got_huff_bits:) is
372 equivalent to j = get_bits(bd, hufGroup->maxLen);
373 */
374 while (bd->inbufBitCount < hufGroup->maxLen) {
375 if (bd->inbufPos == bd->inbufCount) {
376 j = get_bits(bd, hufGroup->maxLen);
377 goto got_huff_bits;
378 }
379 bd->inbufBits =
380 (bd->inbufBits << 8)|bd->inbuf[bd->inbufPos++];
381 bd->inbufBitCount += 8;
382 };
383 bd->inbufBitCount -= hufGroup->maxLen;
384 j = (bd->inbufBits >> bd->inbufBitCount)&
385 ((1 << hufGroup->maxLen)-1);
386got_huff_bits:
387 /* Figure how how many bits are in next symbol and
388 * unget extras */
389 i = hufGroup->minLen;
390 while (j > limit[i])
391 ++i;
392 bd->inbufBitCount += (hufGroup->maxLen - i);
393 /* Huffman decode value to get nextSym (with bounds checking) */
394 if ((i > hufGroup->maxLen)
395 || (((unsigned)(j = (j>>(hufGroup->maxLen-i))-base[i]))
396 >= MAX_SYMBOLS))
397 return RETVAL_DATA_ERROR;
398 nextSym = hufGroup->permute[j];
399 /* We have now decoded the symbol, which indicates
400 either a new literal byte, or a repeated run of the
401 most recent literal byte. First, check if nextSym
402 indicates a repeated run, and if so loop collecting
403 how many times to repeat the last literal. */
404 if (((unsigned)nextSym) <= SYMBOL_RUNB) { /* RUNA or RUNB */
405 /* If this is the start of a new run, zero out
406 * counter */
407 if (!runPos) {
408 runPos = 1;
409 t = 0;
410 }
411 /* Neat trick that saves 1 symbol: instead of
412 or-ing 0 or 1 at each bit position, add 1
413 or 2 instead. For example, 1011 is 1 << 0
414 + 1 << 1 + 2 << 2. 1010 is 2 << 0 + 2 << 1
415 + 1 << 2. You can make any bit pattern
416 that way using 1 less symbol than the basic
417 or 0/1 method (except all bits 0, which
418 would use no symbols, but a run of length 0
419 doesn't mean anything in this context).
420 Thus space is saved. */
421 t += (runPos << nextSym);
422 /* +runPos if RUNA; +2*runPos if RUNB */
423
424 runPos <<= 1;
425 continue;
426 }
427 /* When we hit the first non-run symbol after a run,
428 we now know how many times to repeat the last
429 literal, so append that many copies to our buffer
430 of decoded symbols (dbuf) now. (The last literal
431 used is the one at the head of the mtfSymbol
432 array.) */
433 if (runPos) {
434 runPos = 0;
435 if (dbufCount+t >= dbufSize)
436 return RETVAL_DATA_ERROR;
437
438 uc = symToByte[mtfSymbol[0]];
439 byteCount[uc] += t;
440 while (t--)
441 dbuf[dbufCount++] = uc;
442 }
443 /* Is this the terminating symbol? */
444 if (nextSym > symTotal)
445 break;
446 /* At this point, nextSym indicates a new literal
447 character. Subtract one to get the position in the
448 MTF array at which this literal is currently to be
449 found. (Note that the result can't be -1 or 0,
450 because 0 and 1 are RUNA and RUNB. But another
451 instance of the first symbol in the mtf array,
452 position 0, would have been handled as part of a
453 run above. Therefore 1 unused mtf position minus 2
454 non-literal nextSym values equals -1.) */
455 if (dbufCount >= dbufSize)
456 return RETVAL_DATA_ERROR;
457 i = nextSym - 1;
458 uc = mtfSymbol[i];
459 /* Adjust the MTF array. Since we typically expect to
460 *move only a small number of symbols, and are bound
461 *by 256 in any case, using memmove here would
462 *typically be bigger and slower due to function call
463 *overhead and other assorted setup costs. */
464 do {
465 mtfSymbol[i] = mtfSymbol[i-1];
466 } while (--i);
467 mtfSymbol[0] = uc;
468 uc = symToByte[uc];
469 /* We have our literal byte. Save it into dbuf. */
470 byteCount[uc]++;
471 dbuf[dbufCount++] = (unsigned int)uc;
472 }
473 /* At this point, we've read all the Huffman-coded symbols
474 (and repeated runs) for this block from the input stream,
475 and decoded them into the intermediate buffer. There are
476 dbufCount many decoded bytes in dbuf[]. Now undo the
477 Burrows-Wheeler transform on dbuf. See
478 http://dogma.net/markn/articles/bwt/bwt.htm
479 */
480 /* Turn byteCount into cumulative occurrence counts of 0 to n-1. */
481 j = 0;
482 for (i = 0; i < 256; i++) {
483 k = j+byteCount[i];
484 byteCount[i] = j;
485 j = k;
486 }
487 /* Figure out what order dbuf would be in if we sorted it. */
488 for (i = 0; i < dbufCount; i++) {
489 uc = (unsigned char)(dbuf[i] & 0xff);
490 dbuf[byteCount[uc]] |= (i << 8);
491 byteCount[uc]++;
492 }
493 /* Decode first byte by hand to initialize "previous" byte.
494 Note that it doesn't get output, and if the first three
495 characters are identical it doesn't qualify as a run (hence
496 writeRunCountdown = 5). */
497 if (dbufCount) {
498 if (origPtr >= dbufCount)
499 return RETVAL_DATA_ERROR;
500 bd->writePos = dbuf[origPtr];
501 bd->writeCurrent = (unsigned char)(bd->writePos&0xff);
502 bd->writePos >>= 8;
503 bd->writeRunCountdown = 5;
504 }
505 bd->writeCount = dbufCount;
506
507 return RETVAL_OK;
508}
509
510/* Undo burrows-wheeler transform on intermediate buffer to produce output.
511 If start_bunzip was initialized with out_fd =-1, then up to len bytes of
512 data are written to outbuf. Return value is number of bytes written or
513 error (all errors are negative numbers). If out_fd!=-1, outbuf and len
514 are ignored, data is written to out_fd and return is RETVAL_OK or error.
515*/
516
517static int INIT read_bunzip(struct bunzip_data *bd, char *outbuf, int len)
518{
519 const unsigned int *dbuf;
520 int pos, xcurrent, previous, gotcount;
521
522 /* If last read was short due to end of file, return last block now */
523 if (bd->writeCount < 0)
524 return bd->writeCount;
525
526 gotcount = 0;
527 dbuf = bd->dbuf;
528 pos = bd->writePos;
529 xcurrent = bd->writeCurrent;
530
531 /* We will always have pending decoded data to write into the output
532 buffer unless this is the very first call (in which case we haven't
533 Huffman-decoded a block into the intermediate buffer yet). */
534
535 if (bd->writeCopies) {
536 /* Inside the loop, writeCopies means extra copies (beyond 1) */
537 --bd->writeCopies;
538 /* Loop outputting bytes */
539 for (;;) {
540 /* If the output buffer is full, snapshot
541 * state and return */
542 if (gotcount >= len) {
543 bd->writePos = pos;
544 bd->writeCurrent = xcurrent;
545 bd->writeCopies++;
546 return len;
547 }
548 /* Write next byte into output buffer, updating CRC */
549 outbuf[gotcount++] = xcurrent;
550 bd->writeCRC = (((bd->writeCRC) << 8)
551 ^bd->crc32Table[((bd->writeCRC) >> 24)
552 ^xcurrent]);
553 /* Loop now if we're outputting multiple
554 * copies of this byte */
555 if (bd->writeCopies) {
556 --bd->writeCopies;
557 continue;
558 }
559decode_next_byte:
560 if (!bd->writeCount--)
561 break;
562 /* Follow sequence vector to undo
563 * Burrows-Wheeler transform */
564 previous = xcurrent;
565 pos = dbuf[pos];
566 xcurrent = pos&0xff;
567 pos >>= 8;
568 /* After 3 consecutive copies of the same
569 byte, the 4th is a repeat count. We count
570 down from 4 instead *of counting up because
571 testing for non-zero is faster */
572 if (--bd->writeRunCountdown) {
573 if (xcurrent != previous)
574 bd->writeRunCountdown = 4;
575 } else {
576 /* We have a repeated run, this byte
577 * indicates the count */
578 bd->writeCopies = xcurrent;
579 xcurrent = previous;
580 bd->writeRunCountdown = 5;
581 /* Sometimes there are just 3 bytes
582 * (run length 0) */
583 if (!bd->writeCopies)
584 goto decode_next_byte;
585 /* Subtract the 1 copy we'd output
586 * anyway to get extras */
587 --bd->writeCopies;
588 }
589 }
590 /* Decompression of this block completed successfully */
591 bd->writeCRC = ~bd->writeCRC;
592 bd->totalCRC = ((bd->totalCRC << 1) |
593 (bd->totalCRC >> 31)) ^ bd->writeCRC;
594 /* If this block had a CRC error, force file level CRC error. */
595 if (bd->writeCRC != bd->headerCRC) {
596 bd->totalCRC = bd->headerCRC+1;
597 return RETVAL_LAST_BLOCK;
598 }
599 }
600
601 /* Refill the intermediate buffer by Huffman-decoding next
602 * block of input */
603 /* (previous is just a convenient unused temp variable here) */
604 previous = get_next_block(bd);
605 if (previous) {
606 bd->writeCount = previous;
607 return (previous != RETVAL_LAST_BLOCK) ? previous : gotcount;
608 }
609 bd->writeCRC = 0xffffffffUL;
610 pos = bd->writePos;
611 xcurrent = bd->writeCurrent;
612 goto decode_next_byte;
613}
614
615static int INIT nofill(void *buf, unsigned int len)
616{
617 return -1;
618}
619
620/* Allocate the structure, read file header. If in_fd ==-1, inbuf must contain
621 a complete bunzip file (len bytes long). If in_fd!=-1, inbuf and len are
622 ignored, and data is read from file handle into temporary buffer. */
623static int INIT start_bunzip(struct bunzip_data **bdp, void *inbuf, int len,
624 int (*fill)(void*, unsigned int))
625{
626 struct bunzip_data *bd;
627 unsigned int i, j, c;
628 const unsigned int BZh0 =
629 (((unsigned int)'B') << 24)+(((unsigned int)'Z') << 16)
630 +(((unsigned int)'h') << 8)+(unsigned int)'0';
631
632 /* Figure out how much data to allocate */
633 i = sizeof(struct bunzip_data);
634
635 /* Allocate bunzip_data. Most fields initialize to zero. */
636 bd = *bdp = malloc(i);
637 memset(bd, 0, sizeof(struct bunzip_data));
638 /* Setup input buffer */
639 bd->inbuf = inbuf;
640 bd->inbufCount = len;
641 if (fill != NULL)
642 bd->fill = fill;
643 else
644 bd->fill = nofill;
645
646 /* Init the CRC32 table (big endian) */
647 for (i = 0; i < 256; i++) {
648 c = i << 24;
649 for (j = 8; j; j--)
650 c = c&0x80000000 ? (c << 1)^0x04c11db7 : (c << 1);
651 bd->crc32Table[i] = c;
652 }
653
654 /* Ensure that file starts with "BZh['1'-'9']." */
655 i = get_bits(bd, 32);
656 if (((unsigned int)(i-BZh0-1)) >= 9)
657 return RETVAL_NOT_BZIP_DATA;
658
659 /* Fourth byte (ascii '1'-'9'), indicates block size in units of 100k of
660 uncompressed data. Allocate intermediate buffer for block. */
661 bd->dbufSize = 100000*(i-BZh0);
662
663 bd->dbuf = large_malloc(bd->dbufSize * sizeof(int));
664 return RETVAL_OK;
665}
666
667/* Example usage: decompress src_fd to dst_fd. (Stops at end of bzip2 data,
668 not end of file.) */
669STATIC int INIT bunzip2(unsigned char *buf, int len,
670 int(*fill)(void*, unsigned int),
671 int(*flush)(void*, unsigned int),
672 unsigned char *outbuf,
673 int *pos,
674 void(*error_fn)(char *x))
675{
676 struct bunzip_data *bd;
677 int i = -1;
678 unsigned char *inbuf;
679
680 set_error_fn(error_fn);
681 if (flush)
682 outbuf = malloc(BZIP2_IOBUF_SIZE);
683 else
684 len -= 4; /* Uncompressed size hack active in pre-boot
685 environment */
686 if (!outbuf) {
687 error("Could not allocate output bufer");
688 return -1;
689 }
690 if (buf)
691 inbuf = buf;
692 else
693 inbuf = malloc(BZIP2_IOBUF_SIZE);
694 if (!inbuf) {
695 error("Could not allocate input bufer");
696 goto exit_0;
697 }
698 i = start_bunzip(&bd, inbuf, len, fill);
699 if (!i) {
700 for (;;) {
701 i = read_bunzip(bd, outbuf, BZIP2_IOBUF_SIZE);
702 if (i <= 0)
703 break;
704 if (!flush)
705 outbuf += i;
706 else
707 if (i != flush(outbuf, i)) {
708 i = RETVAL_UNEXPECTED_OUTPUT_EOF;
709 break;
710 }
711 }
712 }
713 /* Check CRC and release memory */
714 if (i == RETVAL_LAST_BLOCK) {
715 if (bd->headerCRC != bd->totalCRC)
716 error("Data integrity error when decompressing.");
717 else
718 i = RETVAL_OK;
719 } else if (i == RETVAL_UNEXPECTED_OUTPUT_EOF) {
720 error("Compressed file ends unexpectedly");
721 }
722 if (bd->dbuf)
723 large_free(bd->dbuf);
724 if (pos)
725 *pos = bd->inbufPos;
726 free(bd);
727 if (!buf)
728 free(inbuf);
729exit_0:
730 if (flush)
731 free(outbuf);
732 return i;
733}
734
735#define decompress bunzip2
diff --git a/lib/decompress_inflate.c b/lib/decompress_inflate.c
new file mode 100644
index 000000000000..163e66aea5f6
--- /dev/null
+++ b/lib/decompress_inflate.c
@@ -0,0 +1,167 @@
1#ifdef STATIC
2/* Pre-boot environment: included */
3
4/* prevent inclusion of _LINUX_KERNEL_H in pre-boot environment: lots
5 * errors about console_printk etc... on ARM */
6#define _LINUX_KERNEL_H
7
8#include "zlib_inflate/inftrees.c"
9#include "zlib_inflate/inffast.c"
10#include "zlib_inflate/inflate.c"
11
12#else /* STATIC */
13/* initramfs et al: linked */
14
15#include <linux/zutil.h>
16
17#include "zlib_inflate/inftrees.h"
18#include "zlib_inflate/inffast.h"
19#include "zlib_inflate/inflate.h"
20
21#include "zlib_inflate/infutil.h"
22
23#endif /* STATIC */
24
25#include <linux/decompress/mm.h>
26
27#define INBUF_LEN (16*1024)
28
29/* Included from initramfs et al code */
30STATIC int INIT gunzip(unsigned char *buf, int len,
31 int(*fill)(void*, unsigned int),
32 int(*flush)(void*, unsigned int),
33 unsigned char *out_buf,
34 int *pos,
35 void(*error_fn)(char *x)) {
36 u8 *zbuf;
37 struct z_stream_s *strm;
38 int rc;
39 size_t out_len;
40
41 set_error_fn(error_fn);
42 rc = -1;
43 if (flush) {
44 out_len = 0x8100; /* 32 K */
45 out_buf = malloc(out_len);
46 } else {
47 out_len = 0x7fffffff; /* no limit */
48 }
49 if (!out_buf) {
50 error("Out of memory while allocating output buffer");
51 goto gunzip_nomem1;
52 }
53
54 if (buf)
55 zbuf = buf;
56 else {
57 zbuf = malloc(INBUF_LEN);
58 len = 0;
59 }
60 if (!zbuf) {
61 error("Out of memory while allocating input buffer");
62 goto gunzip_nomem2;
63 }
64
65 strm = malloc(sizeof(*strm));
66 if (strm == NULL) {
67 error("Out of memory while allocating z_stream");
68 goto gunzip_nomem3;
69 }
70
71 strm->workspace = malloc(flush ? zlib_inflate_workspacesize() :
72 sizeof(struct inflate_state));
73 if (strm->workspace == NULL) {
74 error("Out of memory while allocating workspace");
75 goto gunzip_nomem4;
76 }
77
78 if (len == 0)
79 len = fill(zbuf, INBUF_LEN);
80
81 /* verify the gzip header */
82 if (len < 10 ||
83 zbuf[0] != 0x1f || zbuf[1] != 0x8b || zbuf[2] != 0x08) {
84 if (pos)
85 *pos = 0;
86 error("Not a gzip file");
87 goto gunzip_5;
88 }
89
90 /* skip over gzip header (1f,8b,08... 10 bytes total +
91 * possible asciz filename)
92 */
93 strm->next_in = zbuf + 10;
94 /* skip over asciz filename */
95 if (zbuf[3] & 0x8) {
96 while (strm->next_in[0])
97 strm->next_in++;
98 strm->next_in++;
99 }
100 strm->avail_in = len - 10;
101
102 strm->next_out = out_buf;
103 strm->avail_out = out_len;
104
105 rc = zlib_inflateInit2(strm, -MAX_WBITS);
106
107 if (!flush) {
108 WS(strm)->inflate_state.wsize = 0;
109 WS(strm)->inflate_state.window = NULL;
110 }
111
112 while (rc == Z_OK) {
113 if (strm->avail_in == 0) {
114 /* TODO: handle case where both pos and fill are set */
115 len = fill(zbuf, INBUF_LEN);
116 if (len < 0) {
117 rc = -1;
118 error("read error");
119 break;
120 }
121 strm->next_in = zbuf;
122 strm->avail_in = len;
123 }
124 rc = zlib_inflate(strm, 0);
125
126 /* Write any data generated */
127 if (flush && strm->next_out > out_buf) {
128 int l = strm->next_out - out_buf;
129 if (l != flush(out_buf, l)) {
130 rc = -1;
131 error("write error");
132 break;
133 }
134 strm->next_out = out_buf;
135 strm->avail_out = out_len;
136 }
137
138 /* after Z_FINISH, only Z_STREAM_END is "we unpacked it all" */
139 if (rc == Z_STREAM_END) {
140 rc = 0;
141 break;
142 } else if (rc != Z_OK) {
143 error("uncompression error");
144 rc = -1;
145 }
146 }
147
148 zlib_inflateEnd(strm);
149 if (pos)
150 /* add + 8 to skip over trailer */
151 *pos = strm->next_in - zbuf+8;
152
153gunzip_5:
154 free(strm->workspace);
155gunzip_nomem4:
156 free(strm);
157gunzip_nomem3:
158 if (!buf)
159 free(zbuf);
160gunzip_nomem2:
161 if (flush)
162 free(out_buf);
163gunzip_nomem1:
164 return rc; /* returns Z_OK (0) if successful */
165}
166
167#define decompress gunzip
diff --git a/lib/decompress_unlzma.c b/lib/decompress_unlzma.c
new file mode 100644
index 000000000000..546f2f4c157e
--- /dev/null
+++ b/lib/decompress_unlzma.c
@@ -0,0 +1,647 @@
1/* Lzma decompressor for Linux kernel. Shamelessly snarfed
2 *from busybox 1.1.1
3 *
4 *Linux kernel adaptation
5 *Copyright (C) 2006 Alain < alain@knaff.lu >
6 *
7 *Based on small lzma deflate implementation/Small range coder
8 *implementation for lzma.
9 *Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
10 *
11 *Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
12 *Copyright (C) 1999-2005 Igor Pavlov
13 *
14 *Copyrights of the parts, see headers below.
15 *
16 *
17 *This program is free software; you can redistribute it and/or
18 *modify it under the terms of the GNU Lesser General Public
19 *License as published by the Free Software Foundation; either
20 *version 2.1 of the License, or (at your option) any later version.
21 *
22 *This program is distributed in the hope that it will be useful,
23 *but WITHOUT ANY WARRANTY; without even the implied warranty of
24 *MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
25 *Lesser General Public License for more details.
26 *
27 *You should have received a copy of the GNU Lesser General Public
28 *License along with this library; if not, write to the Free Software
29 *Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
30 */
31
32#ifndef STATIC
33#include <linux/decompress/unlzma.h>
34#endif /* STATIC */
35
36#include <linux/decompress/mm.h>
37
38#define MIN(a, b) (((a) < (b)) ? (a) : (b))
39
40static long long INIT read_int(unsigned char *ptr, int size)
41{
42 int i;
43 long long ret = 0;
44
45 for (i = 0; i < size; i++)
46 ret = (ret << 8) | ptr[size-i-1];
47 return ret;
48}
49
50#define ENDIAN_CONVERT(x) \
51 x = (typeof(x))read_int((unsigned char *)&x, sizeof(x))
52
53
54/* Small range coder implementation for lzma.
55 *Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
56 *
57 *Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
58 *Copyright (c) 1999-2005 Igor Pavlov
59 */
60
61#include <linux/compiler.h>
62
63#define LZMA_IOBUF_SIZE 0x10000
64
65struct rc {
66 int (*fill)(void*, unsigned int);
67 uint8_t *ptr;
68 uint8_t *buffer;
69 uint8_t *buffer_end;
70 int buffer_size;
71 uint32_t code;
72 uint32_t range;
73 uint32_t bound;
74};
75
76
77#define RC_TOP_BITS 24
78#define RC_MOVE_BITS 5
79#define RC_MODEL_TOTAL_BITS 11
80
81
82/* Called twice: once at startup and once in rc_normalize() */
83static void INIT rc_read(struct rc *rc)
84{
85 rc->buffer_size = rc->fill((char *)rc->buffer, LZMA_IOBUF_SIZE);
86 if (rc->buffer_size <= 0)
87 error("unexpected EOF");
88 rc->ptr = rc->buffer;
89 rc->buffer_end = rc->buffer + rc->buffer_size;
90}
91
92/* Called once */
93static inline void INIT rc_init(struct rc *rc,
94 int (*fill)(void*, unsigned int),
95 char *buffer, int buffer_size)
96{
97 rc->fill = fill;
98 rc->buffer = (uint8_t *)buffer;
99 rc->buffer_size = buffer_size;
100 rc->buffer_end = rc->buffer + rc->buffer_size;
101 rc->ptr = rc->buffer;
102
103 rc->code = 0;
104 rc->range = 0xFFFFFFFF;
105}
106
107static inline void INIT rc_init_code(struct rc *rc)
108{
109 int i;
110
111 for (i = 0; i < 5; i++) {
112 if (rc->ptr >= rc->buffer_end)
113 rc_read(rc);
114 rc->code = (rc->code << 8) | *rc->ptr++;
115 }
116}
117
118
119/* Called once. TODO: bb_maybe_free() */
120static inline void INIT rc_free(struct rc *rc)
121{
122 free(rc->buffer);
123}
124
125/* Called twice, but one callsite is in inline'd rc_is_bit_0_helper() */
126static void INIT rc_do_normalize(struct rc *rc)
127{
128 if (rc->ptr >= rc->buffer_end)
129 rc_read(rc);
130 rc->range <<= 8;
131 rc->code = (rc->code << 8) | *rc->ptr++;
132}
133static inline void INIT rc_normalize(struct rc *rc)
134{
135 if (rc->range < (1 << RC_TOP_BITS))
136 rc_do_normalize(rc);
137}
138
139/* Called 9 times */
140/* Why rc_is_bit_0_helper exists?
141 *Because we want to always expose (rc->code < rc->bound) to optimizer
142 */
143static inline uint32_t INIT rc_is_bit_0_helper(struct rc *rc, uint16_t *p)
144{
145 rc_normalize(rc);
146 rc->bound = *p * (rc->range >> RC_MODEL_TOTAL_BITS);
147 return rc->bound;
148}
149static inline int INIT rc_is_bit_0(struct rc *rc, uint16_t *p)
150{
151 uint32_t t = rc_is_bit_0_helper(rc, p);
152 return rc->code < t;
153}
154
155/* Called ~10 times, but very small, thus inlined */
156static inline void INIT rc_update_bit_0(struct rc *rc, uint16_t *p)
157{
158 rc->range = rc->bound;
159 *p += ((1 << RC_MODEL_TOTAL_BITS) - *p) >> RC_MOVE_BITS;
160}
161static inline void rc_update_bit_1(struct rc *rc, uint16_t *p)
162{
163 rc->range -= rc->bound;
164 rc->code -= rc->bound;
165 *p -= *p >> RC_MOVE_BITS;
166}
167
168/* Called 4 times in unlzma loop */
169static int INIT rc_get_bit(struct rc *rc, uint16_t *p, int *symbol)
170{
171 if (rc_is_bit_0(rc, p)) {
172 rc_update_bit_0(rc, p);
173 *symbol *= 2;
174 return 0;
175 } else {
176 rc_update_bit_1(rc, p);
177 *symbol = *symbol * 2 + 1;
178 return 1;
179 }
180}
181
182/* Called once */
183static inline int INIT rc_direct_bit(struct rc *rc)
184{
185 rc_normalize(rc);
186 rc->range >>= 1;
187 if (rc->code >= rc->range) {
188 rc->code -= rc->range;
189 return 1;
190 }
191 return 0;
192}
193
194/* Called twice */
195static inline void INIT
196rc_bit_tree_decode(struct rc *rc, uint16_t *p, int num_levels, int *symbol)
197{
198 int i = num_levels;
199
200 *symbol = 1;
201 while (i--)
202 rc_get_bit(rc, p + *symbol, symbol);
203 *symbol -= 1 << num_levels;
204}
205
206
207/*
208 * Small lzma deflate implementation.
209 * Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
210 *
211 * Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
212 * Copyright (C) 1999-2005 Igor Pavlov
213 */
214
215
216struct lzma_header {
217 uint8_t pos;
218 uint32_t dict_size;
219 uint64_t dst_size;
220} __attribute__ ((packed)) ;
221
222
223#define LZMA_BASE_SIZE 1846
224#define LZMA_LIT_SIZE 768
225
226#define LZMA_NUM_POS_BITS_MAX 4
227
228#define LZMA_LEN_NUM_LOW_BITS 3
229#define LZMA_LEN_NUM_MID_BITS 3
230#define LZMA_LEN_NUM_HIGH_BITS 8
231
232#define LZMA_LEN_CHOICE 0
233#define LZMA_LEN_CHOICE_2 (LZMA_LEN_CHOICE + 1)
234#define LZMA_LEN_LOW (LZMA_LEN_CHOICE_2 + 1)
235#define LZMA_LEN_MID (LZMA_LEN_LOW \
236 + (1 << (LZMA_NUM_POS_BITS_MAX + LZMA_LEN_NUM_LOW_BITS)))
237#define LZMA_LEN_HIGH (LZMA_LEN_MID \
238 +(1 << (LZMA_NUM_POS_BITS_MAX + LZMA_LEN_NUM_MID_BITS)))
239#define LZMA_NUM_LEN_PROBS (LZMA_LEN_HIGH + (1 << LZMA_LEN_NUM_HIGH_BITS))
240
241#define LZMA_NUM_STATES 12
242#define LZMA_NUM_LIT_STATES 7
243
244#define LZMA_START_POS_MODEL_INDEX 4
245#define LZMA_END_POS_MODEL_INDEX 14
246#define LZMA_NUM_FULL_DISTANCES (1 << (LZMA_END_POS_MODEL_INDEX >> 1))
247
248#define LZMA_NUM_POS_SLOT_BITS 6
249#define LZMA_NUM_LEN_TO_POS_STATES 4
250
251#define LZMA_NUM_ALIGN_BITS 4
252
253#define LZMA_MATCH_MIN_LEN 2
254
255#define LZMA_IS_MATCH 0
256#define LZMA_IS_REP (LZMA_IS_MATCH + (LZMA_NUM_STATES << LZMA_NUM_POS_BITS_MAX))
257#define LZMA_IS_REP_G0 (LZMA_IS_REP + LZMA_NUM_STATES)
258#define LZMA_IS_REP_G1 (LZMA_IS_REP_G0 + LZMA_NUM_STATES)
259#define LZMA_IS_REP_G2 (LZMA_IS_REP_G1 + LZMA_NUM_STATES)
260#define LZMA_IS_REP_0_LONG (LZMA_IS_REP_G2 + LZMA_NUM_STATES)
261#define LZMA_POS_SLOT (LZMA_IS_REP_0_LONG \
262 + (LZMA_NUM_STATES << LZMA_NUM_POS_BITS_MAX))
263#define LZMA_SPEC_POS (LZMA_POS_SLOT \
264 +(LZMA_NUM_LEN_TO_POS_STATES << LZMA_NUM_POS_SLOT_BITS))
265#define LZMA_ALIGN (LZMA_SPEC_POS \
266 + LZMA_NUM_FULL_DISTANCES - LZMA_END_POS_MODEL_INDEX)
267#define LZMA_LEN_CODER (LZMA_ALIGN + (1 << LZMA_NUM_ALIGN_BITS))
268#define LZMA_REP_LEN_CODER (LZMA_LEN_CODER + LZMA_NUM_LEN_PROBS)
269#define LZMA_LITERAL (LZMA_REP_LEN_CODER + LZMA_NUM_LEN_PROBS)
270
271
272struct writer {
273 uint8_t *buffer;
274 uint8_t previous_byte;
275 size_t buffer_pos;
276 int bufsize;
277 size_t global_pos;
278 int(*flush)(void*, unsigned int);
279 struct lzma_header *header;
280};
281
282struct cstate {
283 int state;
284 uint32_t rep0, rep1, rep2, rep3;
285};
286
287static inline size_t INIT get_pos(struct writer *wr)
288{
289 return
290 wr->global_pos + wr->buffer_pos;
291}
292
293static inline uint8_t INIT peek_old_byte(struct writer *wr,
294 uint32_t offs)
295{
296 if (!wr->flush) {
297 int32_t pos;
298 while (offs > wr->header->dict_size)
299 offs -= wr->header->dict_size;
300 pos = wr->buffer_pos - offs;
301 return wr->buffer[pos];
302 } else {
303 uint32_t pos = wr->buffer_pos - offs;
304 while (pos >= wr->header->dict_size)
305 pos += wr->header->dict_size;
306 return wr->buffer[pos];
307 }
308
309}
310
311static inline void INIT write_byte(struct writer *wr, uint8_t byte)
312{
313 wr->buffer[wr->buffer_pos++] = wr->previous_byte = byte;
314 if (wr->flush && wr->buffer_pos == wr->header->dict_size) {
315 wr->buffer_pos = 0;
316 wr->global_pos += wr->header->dict_size;
317 wr->flush((char *)wr->buffer, wr->header->dict_size);
318 }
319}
320
321
322static inline void INIT copy_byte(struct writer *wr, uint32_t offs)
323{
324 write_byte(wr, peek_old_byte(wr, offs));
325}
326
327static inline void INIT copy_bytes(struct writer *wr,
328 uint32_t rep0, int len)
329{
330 do {
331 copy_byte(wr, rep0);
332 len--;
333 } while (len != 0 && wr->buffer_pos < wr->header->dst_size);
334}
335
336static inline void INIT process_bit0(struct writer *wr, struct rc *rc,
337 struct cstate *cst, uint16_t *p,
338 int pos_state, uint16_t *prob,
339 int lc, uint32_t literal_pos_mask) {
340 int mi = 1;
341 rc_update_bit_0(rc, prob);
342 prob = (p + LZMA_LITERAL +
343 (LZMA_LIT_SIZE
344 * (((get_pos(wr) & literal_pos_mask) << lc)
345 + (wr->previous_byte >> (8 - lc))))
346 );
347
348 if (cst->state >= LZMA_NUM_LIT_STATES) {
349 int match_byte = peek_old_byte(wr, cst->rep0);
350 do {
351 int bit;
352 uint16_t *prob_lit;
353
354 match_byte <<= 1;
355 bit = match_byte & 0x100;
356 prob_lit = prob + 0x100 + bit + mi;
357 if (rc_get_bit(rc, prob_lit, &mi)) {
358 if (!bit)
359 break;
360 } else {
361 if (bit)
362 break;
363 }
364 } while (mi < 0x100);
365 }
366 while (mi < 0x100) {
367 uint16_t *prob_lit = prob + mi;
368 rc_get_bit(rc, prob_lit, &mi);
369 }
370 write_byte(wr, mi);
371 if (cst->state < 4)
372 cst->state = 0;
373 else if (cst->state < 10)
374 cst->state -= 3;
375 else
376 cst->state -= 6;
377}
378
379static inline void INIT process_bit1(struct writer *wr, struct rc *rc,
380 struct cstate *cst, uint16_t *p,
381 int pos_state, uint16_t *prob) {
382 int offset;
383 uint16_t *prob_len;
384 int num_bits;
385 int len;
386
387 rc_update_bit_1(rc, prob);
388 prob = p + LZMA_IS_REP + cst->state;
389 if (rc_is_bit_0(rc, prob)) {
390 rc_update_bit_0(rc, prob);
391 cst->rep3 = cst->rep2;
392 cst->rep2 = cst->rep1;
393 cst->rep1 = cst->rep0;
394 cst->state = cst->state < LZMA_NUM_LIT_STATES ? 0 : 3;
395 prob = p + LZMA_LEN_CODER;
396 } else {
397 rc_update_bit_1(rc, prob);
398 prob = p + LZMA_IS_REP_G0 + cst->state;
399 if (rc_is_bit_0(rc, prob)) {
400 rc_update_bit_0(rc, prob);
401 prob = (p + LZMA_IS_REP_0_LONG
402 + (cst->state <<
403 LZMA_NUM_POS_BITS_MAX) +
404 pos_state);
405 if (rc_is_bit_0(rc, prob)) {
406 rc_update_bit_0(rc, prob);
407
408 cst->state = cst->state < LZMA_NUM_LIT_STATES ?
409 9 : 11;
410 copy_byte(wr, cst->rep0);
411 return;
412 } else {
413 rc_update_bit_1(rc, prob);
414 }
415 } else {
416 uint32_t distance;
417
418 rc_update_bit_1(rc, prob);
419 prob = p + LZMA_IS_REP_G1 + cst->state;
420 if (rc_is_bit_0(rc, prob)) {
421 rc_update_bit_0(rc, prob);
422 distance = cst->rep1;
423 } else {
424 rc_update_bit_1(rc, prob);
425 prob = p + LZMA_IS_REP_G2 + cst->state;
426 if (rc_is_bit_0(rc, prob)) {
427 rc_update_bit_0(rc, prob);
428 distance = cst->rep2;
429 } else {
430 rc_update_bit_1(rc, prob);
431 distance = cst->rep3;
432 cst->rep3 = cst->rep2;
433 }
434 cst->rep2 = cst->rep1;
435 }
436 cst->rep1 = cst->rep0;
437 cst->rep0 = distance;
438 }
439 cst->state = cst->state < LZMA_NUM_LIT_STATES ? 8 : 11;
440 prob = p + LZMA_REP_LEN_CODER;
441 }
442
443 prob_len = prob + LZMA_LEN_CHOICE;
444 if (rc_is_bit_0(rc, prob_len)) {
445 rc_update_bit_0(rc, prob_len);
446 prob_len = (prob + LZMA_LEN_LOW
447 + (pos_state <<
448 LZMA_LEN_NUM_LOW_BITS));
449 offset = 0;
450 num_bits = LZMA_LEN_NUM_LOW_BITS;
451 } else {
452 rc_update_bit_1(rc, prob_len);
453 prob_len = prob + LZMA_LEN_CHOICE_2;
454 if (rc_is_bit_0(rc, prob_len)) {
455 rc_update_bit_0(rc, prob_len);
456 prob_len = (prob + LZMA_LEN_MID
457 + (pos_state <<
458 LZMA_LEN_NUM_MID_BITS));
459 offset = 1 << LZMA_LEN_NUM_LOW_BITS;
460 num_bits = LZMA_LEN_NUM_MID_BITS;
461 } else {
462 rc_update_bit_1(rc, prob_len);
463 prob_len = prob + LZMA_LEN_HIGH;
464 offset = ((1 << LZMA_LEN_NUM_LOW_BITS)
465 + (1 << LZMA_LEN_NUM_MID_BITS));
466 num_bits = LZMA_LEN_NUM_HIGH_BITS;
467 }
468 }
469
470 rc_bit_tree_decode(rc, prob_len, num_bits, &len);
471 len += offset;
472
473 if (cst->state < 4) {
474 int pos_slot;
475
476 cst->state += LZMA_NUM_LIT_STATES;
477 prob =
478 p + LZMA_POS_SLOT +
479 ((len <
480 LZMA_NUM_LEN_TO_POS_STATES ? len :
481 LZMA_NUM_LEN_TO_POS_STATES - 1)
482 << LZMA_NUM_POS_SLOT_BITS);
483 rc_bit_tree_decode(rc, prob,
484 LZMA_NUM_POS_SLOT_BITS,
485 &pos_slot);
486 if (pos_slot >= LZMA_START_POS_MODEL_INDEX) {
487 int i, mi;
488 num_bits = (pos_slot >> 1) - 1;
489 cst->rep0 = 2 | (pos_slot & 1);
490 if (pos_slot < LZMA_END_POS_MODEL_INDEX) {
491 cst->rep0 <<= num_bits;
492 prob = p + LZMA_SPEC_POS +
493 cst->rep0 - pos_slot - 1;
494 } else {
495 num_bits -= LZMA_NUM_ALIGN_BITS;
496 while (num_bits--)
497 cst->rep0 = (cst->rep0 << 1) |
498 rc_direct_bit(rc);
499 prob = p + LZMA_ALIGN;
500 cst->rep0 <<= LZMA_NUM_ALIGN_BITS;
501 num_bits = LZMA_NUM_ALIGN_BITS;
502 }
503 i = 1;
504 mi = 1;
505 while (num_bits--) {
506 if (rc_get_bit(rc, prob + mi, &mi))
507 cst->rep0 |= i;
508 i <<= 1;
509 }
510 } else
511 cst->rep0 = pos_slot;
512 if (++(cst->rep0) == 0)
513 return;
514 }
515
516 len += LZMA_MATCH_MIN_LEN;
517
518 copy_bytes(wr, cst->rep0, len);
519}
520
521
522
523STATIC inline int INIT unlzma(unsigned char *buf, int in_len,
524 int(*fill)(void*, unsigned int),
525 int(*flush)(void*, unsigned int),
526 unsigned char *output,
527 int *posp,
528 void(*error_fn)(char *x)
529 )
530{
531 struct lzma_header header;
532 int lc, pb, lp;
533 uint32_t pos_state_mask;
534 uint32_t literal_pos_mask;
535 uint16_t *p;
536 int num_probs;
537 struct rc rc;
538 int i, mi;
539 struct writer wr;
540 struct cstate cst;
541 unsigned char *inbuf;
542 int ret = -1;
543
544 set_error_fn(error_fn);
545 if (!flush)
546 in_len -= 4; /* Uncompressed size hack active in pre-boot
547 environment */
548 if (buf)
549 inbuf = buf;
550 else
551 inbuf = malloc(LZMA_IOBUF_SIZE);
552 if (!inbuf) {
553 error("Could not allocate input bufer");
554 goto exit_0;
555 }
556
557 cst.state = 0;
558 cst.rep0 = cst.rep1 = cst.rep2 = cst.rep3 = 1;
559
560 wr.header = &header;
561 wr.flush = flush;
562 wr.global_pos = 0;
563 wr.previous_byte = 0;
564 wr.buffer_pos = 0;
565
566 rc_init(&rc, fill, inbuf, in_len);
567
568 for (i = 0; i < sizeof(header); i++) {
569 if (rc.ptr >= rc.buffer_end)
570 rc_read(&rc);
571 ((unsigned char *)&header)[i] = *rc.ptr++;
572 }
573
574 if (header.pos >= (9 * 5 * 5))
575 error("bad header");
576
577 mi = 0;
578 lc = header.pos;
579 while (lc >= 9) {
580 mi++;
581 lc -= 9;
582 }
583 pb = 0;
584 lp = mi;
585 while (lp >= 5) {
586 pb++;
587 lp -= 5;
588 }
589 pos_state_mask = (1 << pb) - 1;
590 literal_pos_mask = (1 << lp) - 1;
591
592 ENDIAN_CONVERT(header.dict_size);
593 ENDIAN_CONVERT(header.dst_size);
594
595 if (header.dict_size == 0)
596 header.dict_size = 1;
597
598 if (output)
599 wr.buffer = output;
600 else {
601 wr.bufsize = MIN(header.dst_size, header.dict_size);
602 wr.buffer = large_malloc(wr.bufsize);
603 }
604 if (wr.buffer == NULL)
605 goto exit_1;
606
607 num_probs = LZMA_BASE_SIZE + (LZMA_LIT_SIZE << (lc + lp));
608 p = (uint16_t *) large_malloc(num_probs * sizeof(*p));
609 if (p == 0)
610 goto exit_2;
611 num_probs = LZMA_LITERAL + (LZMA_LIT_SIZE << (lc + lp));
612 for (i = 0; i < num_probs; i++)
613 p[i] = (1 << RC_MODEL_TOTAL_BITS) >> 1;
614
615 rc_init_code(&rc);
616
617 while (get_pos(&wr) < header.dst_size) {
618 int pos_state = get_pos(&wr) & pos_state_mask;
619 uint16_t *prob = p + LZMA_IS_MATCH +
620 (cst.state << LZMA_NUM_POS_BITS_MAX) + pos_state;
621 if (rc_is_bit_0(&rc, prob))
622 process_bit0(&wr, &rc, &cst, p, pos_state, prob,
623 lc, literal_pos_mask);
624 else {
625 process_bit1(&wr, &rc, &cst, p, pos_state, prob);
626 if (cst.rep0 == 0)
627 break;
628 }
629 }
630
631 if (posp)
632 *posp = rc.ptr-rc.buffer;
633 if (wr.flush)
634 wr.flush(wr.buffer, wr.buffer_pos);
635 ret = 0;
636 large_free(p);
637exit_2:
638 if (!output)
639 large_free(wr.buffer);
640exit_1:
641 if (!buf)
642 free(inbuf);
643exit_0:
644 return ret;
645}
646
647#define decompress unlzma
diff --git a/lib/zlib_inflate/inflate.h b/lib/zlib_inflate/inflate.h
index df8a6c92052d..3d17b3d1b21f 100644
--- a/lib/zlib_inflate/inflate.h
+++ b/lib/zlib_inflate/inflate.h
@@ -1,3 +1,6 @@
1#ifndef INFLATE_H
2#define INFLATE_H
3
1/* inflate.h -- internal inflate state definition 4/* inflate.h -- internal inflate state definition
2 * Copyright (C) 1995-2004 Mark Adler 5 * Copyright (C) 1995-2004 Mark Adler
3 * For conditions of distribution and use, see copyright notice in zlib.h 6 * For conditions of distribution and use, see copyright notice in zlib.h
@@ -105,3 +108,4 @@ struct inflate_state {
105 unsigned short work[288]; /* work area for code table building */ 108 unsigned short work[288]; /* work area for code table building */
106 code codes[ENOUGH]; /* space for code tables */ 109 code codes[ENOUGH]; /* space for code tables */
107}; 110};
111#endif
diff --git a/lib/zlib_inflate/inftrees.h b/lib/zlib_inflate/inftrees.h
index 5f5219b1240e..b70b4731ac7a 100644
--- a/lib/zlib_inflate/inftrees.h
+++ b/lib/zlib_inflate/inftrees.h
@@ -1,3 +1,6 @@
1#ifndef INFTREES_H
2#define INFTREES_H
3
1/* inftrees.h -- header to use inftrees.c 4/* inftrees.h -- header to use inftrees.c
2 * Copyright (C) 1995-2005 Mark Adler 5 * Copyright (C) 1995-2005 Mark Adler
3 * For conditions of distribution and use, see copyright notice in zlib.h 6 * For conditions of distribution and use, see copyright notice in zlib.h
@@ -53,3 +56,4 @@ typedef enum {
53extern int zlib_inflate_table (codetype type, unsigned short *lens, 56extern int zlib_inflate_table (codetype type, unsigned short *lens,
54 unsigned codes, code **table, 57 unsigned codes, code **table,
55 unsigned *bits, unsigned short *work); 58 unsigned *bits, unsigned short *work);
59#endif
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-24 15:46:28 -0500