blob: 0b954e04bd3015bb08332b2a30a13e1ff6414fea [file] [log] [blame]
Alain Knaffbc22c172009-01-04 22:46:16 +01001/* 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
Phillip Lougherb1af4312009-08-06 15:09:31 -070032#ifdef STATIC
33#define PREBOOT
34#else
Alain Knaffbc22c172009-01-04 22:46:16 +010035#include <linux/decompress/unlzma.h>
Albin Tonnerre9e5cf0c2009-08-06 15:09:32 -070036#include <linux/slab.h>
Alain Knaffbc22c172009-01-04 22:46:16 +010037#endif /* STATIC */
38
39#include <linux/decompress/mm.h>
Alain Knaffbc22c172009-01-04 22:46:16 +010040
41#define MIN(a, b) (((a) < (b)) ? (a) : (b))
42
43static long long INIT read_int(unsigned char *ptr, int size)
44{
45 int i;
46 long long ret = 0;
47
48 for (i = 0; i < size; i++)
49 ret = (ret << 8) | ptr[size-i-1];
50 return ret;
51}
52
53#define ENDIAN_CONVERT(x) \
54 x = (typeof(x))read_int((unsigned char *)&x, sizeof(x))
55
56
57/* Small range coder implementation for lzma.
58 *Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
59 *
60 *Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
61 *Copyright (c) 1999-2005 Igor Pavlov
62 */
63
64#include <linux/compiler.h>
65
66#define LZMA_IOBUF_SIZE 0x10000
67
68struct rc {
69 int (*fill)(void*, unsigned int);
70 uint8_t *ptr;
71 uint8_t *buffer;
72 uint8_t *buffer_end;
73 int buffer_size;
74 uint32_t code;
75 uint32_t range;
76 uint32_t bound;
77};
78
79
80#define RC_TOP_BITS 24
81#define RC_MOVE_BITS 5
82#define RC_MODEL_TOTAL_BITS 11
83
84
85/* Called twice: once at startup and once in rc_normalize() */
86static void INIT rc_read(struct rc *rc)
87{
88 rc->buffer_size = rc->fill((char *)rc->buffer, LZMA_IOBUF_SIZE);
89 if (rc->buffer_size <= 0)
90 error("unexpected EOF");
91 rc->ptr = rc->buffer;
92 rc->buffer_end = rc->buffer + rc->buffer_size;
93}
94
95/* Called once */
96static inline void INIT rc_init(struct rc *rc,
97 int (*fill)(void*, unsigned int),
98 char *buffer, int buffer_size)
99{
100 rc->fill = fill;
101 rc->buffer = (uint8_t *)buffer;
102 rc->buffer_size = buffer_size;
103 rc->buffer_end = rc->buffer + rc->buffer_size;
104 rc->ptr = rc->buffer;
105
106 rc->code = 0;
107 rc->range = 0xFFFFFFFF;
108}
109
110static inline void INIT rc_init_code(struct rc *rc)
111{
112 int i;
113
114 for (i = 0; i < 5; i++) {
115 if (rc->ptr >= rc->buffer_end)
116 rc_read(rc);
117 rc->code = (rc->code << 8) | *rc->ptr++;
118 }
119}
120
121
122/* Called once. TODO: bb_maybe_free() */
123static inline void INIT rc_free(struct rc *rc)
124{
125 free(rc->buffer);
126}
127
128/* Called twice, but one callsite is in inline'd rc_is_bit_0_helper() */
129static void INIT rc_do_normalize(struct rc *rc)
130{
131 if (rc->ptr >= rc->buffer_end)
132 rc_read(rc);
133 rc->range <<= 8;
134 rc->code = (rc->code << 8) | *rc->ptr++;
135}
136static inline void INIT rc_normalize(struct rc *rc)
137{
138 if (rc->range < (1 << RC_TOP_BITS))
139 rc_do_normalize(rc);
140}
141
142/* Called 9 times */
143/* Why rc_is_bit_0_helper exists?
144 *Because we want to always expose (rc->code < rc->bound) to optimizer
145 */
146static inline uint32_t INIT rc_is_bit_0_helper(struct rc *rc, uint16_t *p)
147{
148 rc_normalize(rc);
149 rc->bound = *p * (rc->range >> RC_MODEL_TOTAL_BITS);
150 return rc->bound;
151}
152static inline int INIT rc_is_bit_0(struct rc *rc, uint16_t *p)
153{
154 uint32_t t = rc_is_bit_0_helper(rc, p);
155 return rc->code < t;
156}
157
158/* Called ~10 times, but very small, thus inlined */
159static inline void INIT rc_update_bit_0(struct rc *rc, uint16_t *p)
160{
161 rc->range = rc->bound;
162 *p += ((1 << RC_MODEL_TOTAL_BITS) - *p) >> RC_MOVE_BITS;
163}
164static inline void rc_update_bit_1(struct rc *rc, uint16_t *p)
165{
166 rc->range -= rc->bound;
167 rc->code -= rc->bound;
168 *p -= *p >> RC_MOVE_BITS;
169}
170
171/* Called 4 times in unlzma loop */
172static int INIT rc_get_bit(struct rc *rc, uint16_t *p, int *symbol)
173{
174 if (rc_is_bit_0(rc, p)) {
175 rc_update_bit_0(rc, p);
176 *symbol *= 2;
177 return 0;
178 } else {
179 rc_update_bit_1(rc, p);
180 *symbol = *symbol * 2 + 1;
181 return 1;
182 }
183}
184
185/* Called once */
186static inline int INIT rc_direct_bit(struct rc *rc)
187{
188 rc_normalize(rc);
189 rc->range >>= 1;
190 if (rc->code >= rc->range) {
191 rc->code -= rc->range;
192 return 1;
193 }
194 return 0;
195}
196
197/* Called twice */
198static inline void INIT
199rc_bit_tree_decode(struct rc *rc, uint16_t *p, int num_levels, int *symbol)
200{
201 int i = num_levels;
202
203 *symbol = 1;
204 while (i--)
205 rc_get_bit(rc, p + *symbol, symbol);
206 *symbol -= 1 << num_levels;
207}
208
209
210/*
211 * Small lzma deflate implementation.
212 * Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
213 *
214 * Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
215 * Copyright (C) 1999-2005 Igor Pavlov
216 */
217
218
219struct lzma_header {
220 uint8_t pos;
221 uint32_t dict_size;
222 uint64_t dst_size;
223} __attribute__ ((packed)) ;
224
225
226#define LZMA_BASE_SIZE 1846
227#define LZMA_LIT_SIZE 768
228
229#define LZMA_NUM_POS_BITS_MAX 4
230
231#define LZMA_LEN_NUM_LOW_BITS 3
232#define LZMA_LEN_NUM_MID_BITS 3
233#define LZMA_LEN_NUM_HIGH_BITS 8
234
235#define LZMA_LEN_CHOICE 0
236#define LZMA_LEN_CHOICE_2 (LZMA_LEN_CHOICE + 1)
237#define LZMA_LEN_LOW (LZMA_LEN_CHOICE_2 + 1)
238#define LZMA_LEN_MID (LZMA_LEN_LOW \
239 + (1 << (LZMA_NUM_POS_BITS_MAX + LZMA_LEN_NUM_LOW_BITS)))
240#define LZMA_LEN_HIGH (LZMA_LEN_MID \
241 +(1 << (LZMA_NUM_POS_BITS_MAX + LZMA_LEN_NUM_MID_BITS)))
242#define LZMA_NUM_LEN_PROBS (LZMA_LEN_HIGH + (1 << LZMA_LEN_NUM_HIGH_BITS))
243
244#define LZMA_NUM_STATES 12
245#define LZMA_NUM_LIT_STATES 7
246
247#define LZMA_START_POS_MODEL_INDEX 4
248#define LZMA_END_POS_MODEL_INDEX 14
249#define LZMA_NUM_FULL_DISTANCES (1 << (LZMA_END_POS_MODEL_INDEX >> 1))
250
251#define LZMA_NUM_POS_SLOT_BITS 6
252#define LZMA_NUM_LEN_TO_POS_STATES 4
253
254#define LZMA_NUM_ALIGN_BITS 4
255
256#define LZMA_MATCH_MIN_LEN 2
257
258#define LZMA_IS_MATCH 0
259#define LZMA_IS_REP (LZMA_IS_MATCH + (LZMA_NUM_STATES << LZMA_NUM_POS_BITS_MAX))
260#define LZMA_IS_REP_G0 (LZMA_IS_REP + LZMA_NUM_STATES)
261#define LZMA_IS_REP_G1 (LZMA_IS_REP_G0 + LZMA_NUM_STATES)
262#define LZMA_IS_REP_G2 (LZMA_IS_REP_G1 + LZMA_NUM_STATES)
263#define LZMA_IS_REP_0_LONG (LZMA_IS_REP_G2 + LZMA_NUM_STATES)
264#define LZMA_POS_SLOT (LZMA_IS_REP_0_LONG \
265 + (LZMA_NUM_STATES << LZMA_NUM_POS_BITS_MAX))
266#define LZMA_SPEC_POS (LZMA_POS_SLOT \
267 +(LZMA_NUM_LEN_TO_POS_STATES << LZMA_NUM_POS_SLOT_BITS))
268#define LZMA_ALIGN (LZMA_SPEC_POS \
269 + LZMA_NUM_FULL_DISTANCES - LZMA_END_POS_MODEL_INDEX)
270#define LZMA_LEN_CODER (LZMA_ALIGN + (1 << LZMA_NUM_ALIGN_BITS))
271#define LZMA_REP_LEN_CODER (LZMA_LEN_CODER + LZMA_NUM_LEN_PROBS)
272#define LZMA_LITERAL (LZMA_REP_LEN_CODER + LZMA_NUM_LEN_PROBS)
273
274
275struct writer {
276 uint8_t *buffer;
277 uint8_t previous_byte;
278 size_t buffer_pos;
279 int bufsize;
280 size_t global_pos;
281 int(*flush)(void*, unsigned int);
282 struct lzma_header *header;
283};
284
285struct cstate {
286 int state;
287 uint32_t rep0, rep1, rep2, rep3;
288};
289
290static inline size_t INIT get_pos(struct writer *wr)
291{
292 return
293 wr->global_pos + wr->buffer_pos;
294}
295
296static inline uint8_t INIT peek_old_byte(struct writer *wr,
297 uint32_t offs)
298{
299 if (!wr->flush) {
300 int32_t pos;
301 while (offs > wr->header->dict_size)
302 offs -= wr->header->dict_size;
303 pos = wr->buffer_pos - offs;
304 return wr->buffer[pos];
305 } else {
306 uint32_t pos = wr->buffer_pos - offs;
307 while (pos >= wr->header->dict_size)
308 pos += wr->header->dict_size;
309 return wr->buffer[pos];
310 }
311
312}
313
314static inline void INIT write_byte(struct writer *wr, uint8_t byte)
315{
316 wr->buffer[wr->buffer_pos++] = wr->previous_byte = byte;
317 if (wr->flush && wr->buffer_pos == wr->header->dict_size) {
318 wr->buffer_pos = 0;
319 wr->global_pos += wr->header->dict_size;
320 wr->flush((char *)wr->buffer, wr->header->dict_size);
321 }
322}
323
324
325static inline void INIT copy_byte(struct writer *wr, uint32_t offs)
326{
327 write_byte(wr, peek_old_byte(wr, offs));
328}
329
330static inline void INIT copy_bytes(struct writer *wr,
331 uint32_t rep0, int len)
332{
333 do {
334 copy_byte(wr, rep0);
335 len--;
336 } while (len != 0 && wr->buffer_pos < wr->header->dst_size);
337}
338
339static inline void INIT process_bit0(struct writer *wr, struct rc *rc,
340 struct cstate *cst, uint16_t *p,
341 int pos_state, uint16_t *prob,
342 int lc, uint32_t literal_pos_mask) {
343 int mi = 1;
344 rc_update_bit_0(rc, prob);
345 prob = (p + LZMA_LITERAL +
346 (LZMA_LIT_SIZE
347 * (((get_pos(wr) & literal_pos_mask) << lc)
348 + (wr->previous_byte >> (8 - lc))))
349 );
350
351 if (cst->state >= LZMA_NUM_LIT_STATES) {
352 int match_byte = peek_old_byte(wr, cst->rep0);
353 do {
354 int bit;
355 uint16_t *prob_lit;
356
357 match_byte <<= 1;
358 bit = match_byte & 0x100;
359 prob_lit = prob + 0x100 + bit + mi;
360 if (rc_get_bit(rc, prob_lit, &mi)) {
361 if (!bit)
362 break;
363 } else {
364 if (bit)
365 break;
366 }
367 } while (mi < 0x100);
368 }
369 while (mi < 0x100) {
370 uint16_t *prob_lit = prob + mi;
371 rc_get_bit(rc, prob_lit, &mi);
372 }
373 write_byte(wr, mi);
374 if (cst->state < 4)
375 cst->state = 0;
376 else if (cst->state < 10)
377 cst->state -= 3;
378 else
379 cst->state -= 6;
380}
381
382static inline void INIT process_bit1(struct writer *wr, struct rc *rc,
383 struct cstate *cst, uint16_t *p,
384 int pos_state, uint16_t *prob) {
385 int offset;
386 uint16_t *prob_len;
387 int num_bits;
388 int len;
389
390 rc_update_bit_1(rc, prob);
391 prob = p + LZMA_IS_REP + cst->state;
392 if (rc_is_bit_0(rc, prob)) {
393 rc_update_bit_0(rc, prob);
394 cst->rep3 = cst->rep2;
395 cst->rep2 = cst->rep1;
396 cst->rep1 = cst->rep0;
397 cst->state = cst->state < LZMA_NUM_LIT_STATES ? 0 : 3;
398 prob = p + LZMA_LEN_CODER;
399 } else {
400 rc_update_bit_1(rc, prob);
401 prob = p + LZMA_IS_REP_G0 + cst->state;
402 if (rc_is_bit_0(rc, prob)) {
403 rc_update_bit_0(rc, prob);
404 prob = (p + LZMA_IS_REP_0_LONG
405 + (cst->state <<
406 LZMA_NUM_POS_BITS_MAX) +
407 pos_state);
408 if (rc_is_bit_0(rc, prob)) {
409 rc_update_bit_0(rc, prob);
410
411 cst->state = cst->state < LZMA_NUM_LIT_STATES ?
412 9 : 11;
413 copy_byte(wr, cst->rep0);
414 return;
415 } else {
416 rc_update_bit_1(rc, prob);
417 }
418 } else {
419 uint32_t distance;
420
421 rc_update_bit_1(rc, prob);
422 prob = p + LZMA_IS_REP_G1 + cst->state;
423 if (rc_is_bit_0(rc, prob)) {
424 rc_update_bit_0(rc, prob);
425 distance = cst->rep1;
426 } else {
427 rc_update_bit_1(rc, prob);
428 prob = p + LZMA_IS_REP_G2 + cst->state;
429 if (rc_is_bit_0(rc, prob)) {
430 rc_update_bit_0(rc, prob);
431 distance = cst->rep2;
432 } else {
433 rc_update_bit_1(rc, prob);
434 distance = cst->rep3;
435 cst->rep3 = cst->rep2;
436 }
437 cst->rep2 = cst->rep1;
438 }
439 cst->rep1 = cst->rep0;
440 cst->rep0 = distance;
441 }
442 cst->state = cst->state < LZMA_NUM_LIT_STATES ? 8 : 11;
443 prob = p + LZMA_REP_LEN_CODER;
444 }
445
446 prob_len = prob + LZMA_LEN_CHOICE;
447 if (rc_is_bit_0(rc, prob_len)) {
448 rc_update_bit_0(rc, prob_len);
449 prob_len = (prob + LZMA_LEN_LOW
450 + (pos_state <<
451 LZMA_LEN_NUM_LOW_BITS));
452 offset = 0;
453 num_bits = LZMA_LEN_NUM_LOW_BITS;
454 } else {
455 rc_update_bit_1(rc, prob_len);
456 prob_len = prob + LZMA_LEN_CHOICE_2;
457 if (rc_is_bit_0(rc, prob_len)) {
458 rc_update_bit_0(rc, prob_len);
459 prob_len = (prob + LZMA_LEN_MID
460 + (pos_state <<
461 LZMA_LEN_NUM_MID_BITS));
462 offset = 1 << LZMA_LEN_NUM_LOW_BITS;
463 num_bits = LZMA_LEN_NUM_MID_BITS;
464 } else {
465 rc_update_bit_1(rc, prob_len);
466 prob_len = prob + LZMA_LEN_HIGH;
467 offset = ((1 << LZMA_LEN_NUM_LOW_BITS)
468 + (1 << LZMA_LEN_NUM_MID_BITS));
469 num_bits = LZMA_LEN_NUM_HIGH_BITS;
470 }
471 }
472
473 rc_bit_tree_decode(rc, prob_len, num_bits, &len);
474 len += offset;
475
476 if (cst->state < 4) {
477 int pos_slot;
478
479 cst->state += LZMA_NUM_LIT_STATES;
480 prob =
481 p + LZMA_POS_SLOT +
482 ((len <
483 LZMA_NUM_LEN_TO_POS_STATES ? len :
484 LZMA_NUM_LEN_TO_POS_STATES - 1)
485 << LZMA_NUM_POS_SLOT_BITS);
486 rc_bit_tree_decode(rc, prob,
487 LZMA_NUM_POS_SLOT_BITS,
488 &pos_slot);
489 if (pos_slot >= LZMA_START_POS_MODEL_INDEX) {
490 int i, mi;
491 num_bits = (pos_slot >> 1) - 1;
492 cst->rep0 = 2 | (pos_slot & 1);
493 if (pos_slot < LZMA_END_POS_MODEL_INDEX) {
494 cst->rep0 <<= num_bits;
495 prob = p + LZMA_SPEC_POS +
496 cst->rep0 - pos_slot - 1;
497 } else {
498 num_bits -= LZMA_NUM_ALIGN_BITS;
499 while (num_bits--)
500 cst->rep0 = (cst->rep0 << 1) |
501 rc_direct_bit(rc);
502 prob = p + LZMA_ALIGN;
503 cst->rep0 <<= LZMA_NUM_ALIGN_BITS;
504 num_bits = LZMA_NUM_ALIGN_BITS;
505 }
506 i = 1;
507 mi = 1;
508 while (num_bits--) {
509 if (rc_get_bit(rc, prob + mi, &mi))
510 cst->rep0 |= i;
511 i <<= 1;
512 }
513 } else
514 cst->rep0 = pos_slot;
515 if (++(cst->rep0) == 0)
516 return;
517 }
518
519 len += LZMA_MATCH_MIN_LEN;
520
521 copy_bytes(wr, cst->rep0, len);
522}
523
524
525
526STATIC inline int INIT unlzma(unsigned char *buf, int in_len,
527 int(*fill)(void*, unsigned int),
528 int(*flush)(void*, unsigned int),
529 unsigned char *output,
530 int *posp,
531 void(*error_fn)(char *x)
532 )
533{
534 struct lzma_header header;
535 int lc, pb, lp;
536 uint32_t pos_state_mask;
537 uint32_t literal_pos_mask;
538 uint16_t *p;
539 int num_probs;
540 struct rc rc;
541 int i, mi;
542 struct writer wr;
543 struct cstate cst;
544 unsigned char *inbuf;
545 int ret = -1;
546
547 set_error_fn(error_fn);
Phillip Lougherb1af4312009-08-06 15:09:31 -0700548
Alain Knaffbc22c172009-01-04 22:46:16 +0100549 if (buf)
550 inbuf = buf;
551 else
552 inbuf = malloc(LZMA_IOBUF_SIZE);
553 if (!inbuf) {
554 error("Could not allocate input bufer");
555 goto exit_0;
556 }
557
558 cst.state = 0;
559 cst.rep0 = cst.rep1 = cst.rep2 = cst.rep3 = 1;
560
561 wr.header = &header;
562 wr.flush = flush;
563 wr.global_pos = 0;
564 wr.previous_byte = 0;
565 wr.buffer_pos = 0;
566
567 rc_init(&rc, fill, inbuf, in_len);
568
569 for (i = 0; i < sizeof(header); i++) {
570 if (rc.ptr >= rc.buffer_end)
571 rc_read(&rc);
572 ((unsigned char *)&header)[i] = *rc.ptr++;
573 }
574
575 if (header.pos >= (9 * 5 * 5))
576 error("bad header");
577
578 mi = 0;
579 lc = header.pos;
580 while (lc >= 9) {
581 mi++;
582 lc -= 9;
583 }
584 pb = 0;
585 lp = mi;
586 while (lp >= 5) {
587 pb++;
588 lp -= 5;
589 }
590 pos_state_mask = (1 << pb) - 1;
591 literal_pos_mask = (1 << lp) - 1;
592
593 ENDIAN_CONVERT(header.dict_size);
594 ENDIAN_CONVERT(header.dst_size);
595
596 if (header.dict_size == 0)
597 header.dict_size = 1;
598
599 if (output)
600 wr.buffer = output;
601 else {
602 wr.bufsize = MIN(header.dst_size, header.dict_size);
603 wr.buffer = large_malloc(wr.bufsize);
604 }
605 if (wr.buffer == NULL)
606 goto exit_1;
607
608 num_probs = LZMA_BASE_SIZE + (LZMA_LIT_SIZE << (lc + lp));
609 p = (uint16_t *) large_malloc(num_probs * sizeof(*p));
610 if (p == 0)
611 goto exit_2;
612 num_probs = LZMA_LITERAL + (LZMA_LIT_SIZE << (lc + lp));
613 for (i = 0; i < num_probs; i++)
614 p[i] = (1 << RC_MODEL_TOTAL_BITS) >> 1;
615
616 rc_init_code(&rc);
617
618 while (get_pos(&wr) < header.dst_size) {
619 int pos_state = get_pos(&wr) & pos_state_mask;
620 uint16_t *prob = p + LZMA_IS_MATCH +
621 (cst.state << LZMA_NUM_POS_BITS_MAX) + pos_state;
622 if (rc_is_bit_0(&rc, prob))
623 process_bit0(&wr, &rc, &cst, p, pos_state, prob,
624 lc, literal_pos_mask);
625 else {
626 process_bit1(&wr, &rc, &cst, p, pos_state, prob);
627 if (cst.rep0 == 0)
628 break;
629 }
630 }
631
632 if (posp)
633 *posp = rc.ptr-rc.buffer;
634 if (wr.flush)
635 wr.flush(wr.buffer, wr.buffer_pos);
636 ret = 0;
637 large_free(p);
638exit_2:
639 if (!output)
640 large_free(wr.buffer);
641exit_1:
642 if (!buf)
643 free(inbuf);
644exit_0:
645 return ret;
646}
647
Phillip Lougherb1af4312009-08-06 15:09:31 -0700648#ifdef PREBOOT
649STATIC int INIT decompress(unsigned char *buf, int in_len,
650 int(*fill)(void*, unsigned int),
651 int(*flush)(void*, unsigned int),
652 unsigned char *output,
653 int *posp,
654 void(*error_fn)(char *x)
655 )
656{
657 return unlzma(buf, in_len - 4, fill, flush, output, posp, error_fn);
658}
659#endif