1 /* Lzma decompressor for Linux kernel. Shamelessly snarfed
4 *Linux kernel adaptation
5 *Copyright (C) 2006 Alain < alain@knaff.lu >
7 *Based on small lzma deflate implementation/Small range coder
8 *implementation for lzma.
9 *Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
11 *Based on LzmaDecode.c from the LZMA SDK 4.22 (https://www.7-zip.org/)
12 *Copyright (C) 1999-2005 Igor Pavlov
14 *Copyrights of the parts, see headers below.
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.
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.
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
35 #include <linux/decompress/unlzma.h>
38 #include <linux/decompress/mm.h>
41 #define MIN(a, b) (((a) < (b)) ? (a) : (b))
44 static long long INIT read_int(unsigned char *ptr, int size)
49 for (i = 0; i < size; i++)
50 ret = (ret << 8) | ptr[size-i-1];
54 #define ENDIAN_CONVERT(x) \
55 x = (typeof(x))read_int((unsigned char *)&x, sizeof(x))
58 /* Small range coder implementation for lzma.
59 *Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
61 *Based on LzmaDecode.c from the LZMA SDK 4.22 (https://www.7-zip.org/)
62 *Copyright (c) 1999-2005 Igor Pavlov
65 #include <linux/compiler.h>
67 #define LZMA_IOBUF_SIZE 0x10000
70 long (*fill)(void*, unsigned long);
78 void (*error)(char *);
82 #define RC_TOP_BITS 24
83 #define RC_MOVE_BITS 5
84 #define RC_MODEL_TOTAL_BITS 11
87 static long INIT nofill(void *buffer, unsigned long len)
92 /* Called twice: once at startup and once in rc_normalize() */
93 static void INIT rc_read(struct rc *rc)
95 rc->buffer_size = rc->fill((char *)rc->buffer, LZMA_IOBUF_SIZE);
96 if (rc->buffer_size <= 0)
97 rc->error("unexpected EOF");
99 rc->buffer_end = rc->buffer + rc->buffer_size;
103 static inline void INIT rc_init(struct rc *rc,
104 long (*fill)(void*, unsigned long),
105 char *buffer, long buffer_size)
111 rc->buffer = (uint8_t *)buffer;
112 rc->buffer_size = buffer_size;
113 rc->buffer_end = rc->buffer + rc->buffer_size;
114 rc->ptr = rc->buffer;
117 rc->range = 0xFFFFFFFF;
120 static inline void INIT rc_init_code(struct rc *rc)
124 for (i = 0; i < 5; i++) {
125 if (rc->ptr >= rc->buffer_end)
127 rc->code = (rc->code << 8) | *rc->ptr++;
132 /* Called twice, but one callsite is in inline'd rc_is_bit_0_helper() */
133 static void INIT rc_do_normalize(struct rc *rc)
135 if (rc->ptr >= rc->buffer_end)
138 rc->code = (rc->code << 8) | *rc->ptr++;
140 static inline void INIT rc_normalize(struct rc *rc)
142 if (rc->range < (1 << RC_TOP_BITS))
147 /* Why rc_is_bit_0_helper exists?
148 *Because we want to always expose (rc->code < rc->bound) to optimizer
150 static inline uint32_t INIT rc_is_bit_0_helper(struct rc *rc, uint16_t *p)
153 rc->bound = *p * (rc->range >> RC_MODEL_TOTAL_BITS);
156 static inline int INIT rc_is_bit_0(struct rc *rc, uint16_t *p)
158 uint32_t t = rc_is_bit_0_helper(rc, p);
162 /* Called ~10 times, but very small, thus inlined */
163 static inline void INIT rc_update_bit_0(struct rc *rc, uint16_t *p)
165 rc->range = rc->bound;
166 *p += ((1 << RC_MODEL_TOTAL_BITS) - *p) >> RC_MOVE_BITS;
168 static inline void INIT rc_update_bit_1(struct rc *rc, uint16_t *p)
170 rc->range -= rc->bound;
171 rc->code -= rc->bound;
172 *p -= *p >> RC_MOVE_BITS;
175 /* Called 4 times in unlzma loop */
176 static int INIT rc_get_bit(struct rc *rc, uint16_t *p, int *symbol)
178 if (rc_is_bit_0(rc, p)) {
179 rc_update_bit_0(rc, p);
183 rc_update_bit_1(rc, p);
184 *symbol = *symbol * 2 + 1;
190 static inline int INIT rc_direct_bit(struct rc *rc)
194 if (rc->code >= rc->range) {
195 rc->code -= rc->range;
202 static inline void INIT
203 rc_bit_tree_decode(struct rc *rc, uint16_t *p, int num_levels, int *symbol)
209 rc_get_bit(rc, p + *symbol, symbol);
210 *symbol -= 1 << num_levels;
215 * Small lzma deflate implementation.
216 * Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
218 * Based on LzmaDecode.c from the LZMA SDK 4.22 (https://www.7-zip.org/)
219 * Copyright (C) 1999-2005 Igor Pavlov
227 } __attribute__ ((packed)) ;
230 #define LZMA_BASE_SIZE 1846
231 #define LZMA_LIT_SIZE 768
233 #define LZMA_NUM_POS_BITS_MAX 4
235 #define LZMA_LEN_NUM_LOW_BITS 3
236 #define LZMA_LEN_NUM_MID_BITS 3
237 #define LZMA_LEN_NUM_HIGH_BITS 8
239 #define LZMA_LEN_CHOICE 0
240 #define LZMA_LEN_CHOICE_2 (LZMA_LEN_CHOICE + 1)
241 #define LZMA_LEN_LOW (LZMA_LEN_CHOICE_2 + 1)
242 #define LZMA_LEN_MID (LZMA_LEN_LOW \
243 + (1 << (LZMA_NUM_POS_BITS_MAX + LZMA_LEN_NUM_LOW_BITS)))
244 #define LZMA_LEN_HIGH (LZMA_LEN_MID \
245 +(1 << (LZMA_NUM_POS_BITS_MAX + LZMA_LEN_NUM_MID_BITS)))
246 #define LZMA_NUM_LEN_PROBS (LZMA_LEN_HIGH + (1 << LZMA_LEN_NUM_HIGH_BITS))
248 #define LZMA_NUM_STATES 12
249 #define LZMA_NUM_LIT_STATES 7
251 #define LZMA_START_POS_MODEL_INDEX 4
252 #define LZMA_END_POS_MODEL_INDEX 14
253 #define LZMA_NUM_FULL_DISTANCES (1 << (LZMA_END_POS_MODEL_INDEX >> 1))
255 #define LZMA_NUM_POS_SLOT_BITS 6
256 #define LZMA_NUM_LEN_TO_POS_STATES 4
258 #define LZMA_NUM_ALIGN_BITS 4
260 #define LZMA_MATCH_MIN_LEN 2
262 #define LZMA_IS_MATCH 0
263 #define LZMA_IS_REP (LZMA_IS_MATCH + (LZMA_NUM_STATES << LZMA_NUM_POS_BITS_MAX))
264 #define LZMA_IS_REP_G0 (LZMA_IS_REP + LZMA_NUM_STATES)
265 #define LZMA_IS_REP_G1 (LZMA_IS_REP_G0 + LZMA_NUM_STATES)
266 #define LZMA_IS_REP_G2 (LZMA_IS_REP_G1 + LZMA_NUM_STATES)
267 #define LZMA_IS_REP_0_LONG (LZMA_IS_REP_G2 + LZMA_NUM_STATES)
268 #define LZMA_POS_SLOT (LZMA_IS_REP_0_LONG \
269 + (LZMA_NUM_STATES << LZMA_NUM_POS_BITS_MAX))
270 #define LZMA_SPEC_POS (LZMA_POS_SLOT \
271 +(LZMA_NUM_LEN_TO_POS_STATES << LZMA_NUM_POS_SLOT_BITS))
272 #define LZMA_ALIGN (LZMA_SPEC_POS \
273 + LZMA_NUM_FULL_DISTANCES - LZMA_END_POS_MODEL_INDEX)
274 #define LZMA_LEN_CODER (LZMA_ALIGN + (1 << LZMA_NUM_ALIGN_BITS))
275 #define LZMA_REP_LEN_CODER (LZMA_LEN_CODER + LZMA_NUM_LEN_PROBS)
276 #define LZMA_LITERAL (LZMA_REP_LEN_CODER + LZMA_NUM_LEN_PROBS)
281 uint8_t previous_byte;
285 long (*flush)(void*, unsigned long);
286 struct lzma_header *header;
291 uint32_t rep0, rep1, rep2, rep3;
294 static inline size_t INIT get_pos(struct writer *wr)
297 wr->global_pos + wr->buffer_pos;
300 static inline uint8_t INIT peek_old_byte(struct writer *wr,
305 while (offs > wr->header->dict_size)
306 offs -= wr->header->dict_size;
307 pos = wr->buffer_pos - offs;
308 return wr->buffer[pos];
310 uint32_t pos = wr->buffer_pos - offs;
311 while (pos >= wr->header->dict_size)
312 pos += wr->header->dict_size;
313 return wr->buffer[pos];
318 static inline int INIT write_byte(struct writer *wr, uint8_t byte)
320 wr->buffer[wr->buffer_pos++] = wr->previous_byte = byte;
321 if (wr->flush && wr->buffer_pos == wr->header->dict_size) {
323 wr->global_pos += wr->header->dict_size;
324 if (wr->flush((char *)wr->buffer, wr->header->dict_size)
325 != wr->header->dict_size)
332 static inline int INIT copy_byte(struct writer *wr, uint32_t offs)
334 return write_byte(wr, peek_old_byte(wr, offs));
337 static inline int INIT copy_bytes(struct writer *wr,
338 uint32_t rep0, int len)
341 if (copy_byte(wr, rep0))
344 } while (len != 0 && wr->buffer_pos < wr->header->dst_size);
349 static inline int INIT process_bit0(struct writer *wr, struct rc *rc,
350 struct cstate *cst, uint16_t *p,
351 int pos_state, uint16_t *prob,
352 int lc, uint32_t literal_pos_mask) {
354 rc_update_bit_0(rc, prob);
355 prob = (p + LZMA_LITERAL +
357 * (((get_pos(wr) & literal_pos_mask) << lc)
358 + (wr->previous_byte >> (8 - lc))))
361 if (cst->state >= LZMA_NUM_LIT_STATES) {
362 int match_byte = peek_old_byte(wr, cst->rep0);
368 bit = match_byte & 0x100;
369 prob_lit = prob + 0x100 + bit + mi;
370 if (rc_get_bit(rc, prob_lit, &mi)) {
377 } while (mi < 0x100);
380 uint16_t *prob_lit = prob + mi;
381 rc_get_bit(rc, prob_lit, &mi);
385 else if (cst->state < 10)
390 return write_byte(wr, mi);
393 static inline int INIT process_bit1(struct writer *wr, struct rc *rc,
394 struct cstate *cst, uint16_t *p,
395 int pos_state, uint16_t *prob) {
401 rc_update_bit_1(rc, prob);
402 prob = p + LZMA_IS_REP + cst->state;
403 if (rc_is_bit_0(rc, prob)) {
404 rc_update_bit_0(rc, prob);
405 cst->rep3 = cst->rep2;
406 cst->rep2 = cst->rep1;
407 cst->rep1 = cst->rep0;
408 cst->state = cst->state < LZMA_NUM_LIT_STATES ? 0 : 3;
409 prob = p + LZMA_LEN_CODER;
411 rc_update_bit_1(rc, prob);
412 prob = p + LZMA_IS_REP_G0 + cst->state;
413 if (rc_is_bit_0(rc, prob)) {
414 rc_update_bit_0(rc, prob);
415 prob = (p + LZMA_IS_REP_0_LONG
417 LZMA_NUM_POS_BITS_MAX) +
419 if (rc_is_bit_0(rc, prob)) {
420 rc_update_bit_0(rc, prob);
422 cst->state = cst->state < LZMA_NUM_LIT_STATES ?
424 return copy_byte(wr, cst->rep0);
426 rc_update_bit_1(rc, prob);
431 rc_update_bit_1(rc, prob);
432 prob = p + LZMA_IS_REP_G1 + cst->state;
433 if (rc_is_bit_0(rc, prob)) {
434 rc_update_bit_0(rc, prob);
435 distance = cst->rep1;
437 rc_update_bit_1(rc, prob);
438 prob = p + LZMA_IS_REP_G2 + cst->state;
439 if (rc_is_bit_0(rc, prob)) {
440 rc_update_bit_0(rc, prob);
441 distance = cst->rep2;
443 rc_update_bit_1(rc, prob);
444 distance = cst->rep3;
445 cst->rep3 = cst->rep2;
447 cst->rep2 = cst->rep1;
449 cst->rep1 = cst->rep0;
450 cst->rep0 = distance;
452 cst->state = cst->state < LZMA_NUM_LIT_STATES ? 8 : 11;
453 prob = p + LZMA_REP_LEN_CODER;
456 prob_len = prob + LZMA_LEN_CHOICE;
457 if (rc_is_bit_0(rc, prob_len)) {
458 rc_update_bit_0(rc, prob_len);
459 prob_len = (prob + LZMA_LEN_LOW
461 LZMA_LEN_NUM_LOW_BITS));
463 num_bits = LZMA_LEN_NUM_LOW_BITS;
465 rc_update_bit_1(rc, prob_len);
466 prob_len = prob + LZMA_LEN_CHOICE_2;
467 if (rc_is_bit_0(rc, prob_len)) {
468 rc_update_bit_0(rc, prob_len);
469 prob_len = (prob + LZMA_LEN_MID
471 LZMA_LEN_NUM_MID_BITS));
472 offset = 1 << LZMA_LEN_NUM_LOW_BITS;
473 num_bits = LZMA_LEN_NUM_MID_BITS;
475 rc_update_bit_1(rc, prob_len);
476 prob_len = prob + LZMA_LEN_HIGH;
477 offset = ((1 << LZMA_LEN_NUM_LOW_BITS)
478 + (1 << LZMA_LEN_NUM_MID_BITS));
479 num_bits = LZMA_LEN_NUM_HIGH_BITS;
483 rc_bit_tree_decode(rc, prob_len, num_bits, &len);
486 if (cst->state < 4) {
489 cst->state += LZMA_NUM_LIT_STATES;
493 LZMA_NUM_LEN_TO_POS_STATES ? len :
494 LZMA_NUM_LEN_TO_POS_STATES - 1)
495 << LZMA_NUM_POS_SLOT_BITS);
496 rc_bit_tree_decode(rc, prob,
497 LZMA_NUM_POS_SLOT_BITS,
499 if (pos_slot >= LZMA_START_POS_MODEL_INDEX) {
501 num_bits = (pos_slot >> 1) - 1;
502 cst->rep0 = 2 | (pos_slot & 1);
503 if (pos_slot < LZMA_END_POS_MODEL_INDEX) {
504 cst->rep0 <<= num_bits;
505 prob = p + LZMA_SPEC_POS +
506 cst->rep0 - pos_slot - 1;
508 num_bits -= LZMA_NUM_ALIGN_BITS;
510 cst->rep0 = (cst->rep0 << 1) |
512 prob = p + LZMA_ALIGN;
513 cst->rep0 <<= LZMA_NUM_ALIGN_BITS;
514 num_bits = LZMA_NUM_ALIGN_BITS;
519 if (rc_get_bit(rc, prob + mi, &mi))
524 cst->rep0 = pos_slot;
525 if (++(cst->rep0) == 0)
527 if (cst->rep0 > wr->header->dict_size
528 || cst->rep0 > get_pos(wr))
532 len += LZMA_MATCH_MIN_LEN;
534 return copy_bytes(wr, cst->rep0, len);
539 STATIC inline int INIT unlzma(unsigned char *buf, long in_len,
540 long (*fill)(void*, unsigned long),
541 long (*flush)(void*, unsigned long),
542 unsigned char *output,
544 void(*error)(char *x)
547 struct lzma_header header;
549 uint32_t pos_state_mask;
550 uint32_t literal_pos_mask;
557 unsigned char *inbuf;
565 inbuf = malloc(LZMA_IOBUF_SIZE);
567 error("Could not allocate input buffer");
572 cst.rep0 = cst.rep1 = cst.rep2 = cst.rep3 = 1;
577 wr.previous_byte = 0;
580 rc_init(&rc, fill, inbuf, in_len);
582 for (i = 0; i < sizeof(header); i++) {
583 if (rc.ptr >= rc.buffer_end)
585 ((unsigned char *)&header)[i] = *rc.ptr++;
588 if (header.pos >= (9 * 5 * 5)) {
605 pos_state_mask = (1 << pb) - 1;
606 literal_pos_mask = (1 << lp) - 1;
608 ENDIAN_CONVERT(header.dict_size);
609 ENDIAN_CONVERT(header.dst_size);
611 if (header.dict_size == 0)
612 header.dict_size = 1;
617 wr.bufsize = MIN(header.dst_size, header.dict_size);
618 wr.buffer = large_malloc(wr.bufsize);
620 if (wr.buffer == NULL)
623 num_probs = LZMA_BASE_SIZE + (LZMA_LIT_SIZE << (lc + lp));
624 p = (uint16_t *) large_malloc(num_probs * sizeof(*p));
627 num_probs = LZMA_LITERAL + (LZMA_LIT_SIZE << (lc + lp));
628 for (i = 0; i < num_probs; i++)
629 p[i] = (1 << RC_MODEL_TOTAL_BITS) >> 1;
633 while (get_pos(&wr) < header.dst_size) {
634 int pos_state = get_pos(&wr) & pos_state_mask;
635 uint16_t *prob = p + LZMA_IS_MATCH +
636 (cst.state << LZMA_NUM_POS_BITS_MAX) + pos_state;
637 if (rc_is_bit_0(&rc, prob)) {
638 if (process_bit0(&wr, &rc, &cst, p, pos_state, prob,
639 lc, literal_pos_mask)) {
640 error("LZMA data is corrupt");
644 if (process_bit1(&wr, &rc, &cst, p, pos_state, prob)) {
645 error("LZMA data is corrupt");
651 if (rc.buffer_size <= 0)
656 *posp = rc.ptr-rc.buffer;
657 if (!wr.flush || wr.flush(wr.buffer, wr.buffer_pos) == wr.buffer_pos)
663 large_free(wr.buffer);
672 STATIC int INIT __decompress(unsigned char *buf, long in_len,
673 long (*fill)(void*, unsigned long),
674 long (*flush)(void*, unsigned long),
675 unsigned char *output, long out_len,
677 void (*error)(char *x))
679 return unlzma(buf, in_len - 4, fill, flush, output, posp, error);