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 (http://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>
36 #include <linux/slab.h>
39 #include <linux/decompress/mm.h>
41 #define MIN(a, b) (((a) < (b)) ? (a) : (b))
43 static long long INIT
read_int(unsigned char *ptr
, int size
)
48 for (i
= 0; i
< size
; i
++)
49 ret
= (ret
<< 8) | ptr
[size
-i
-1];
53 #define ENDIAN_CONVERT(x) \
54 x = (typeof(x))read_int((unsigned char *)&x, sizeof(x))
57 /* Small range coder implementation for lzma.
58 *Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
60 *Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
61 *Copyright (c) 1999-2005 Igor Pavlov
64 #include <linux/compiler.h>
66 #define LZMA_IOBUF_SIZE 0x10000
69 int (*fill
)(void*, unsigned int);
80 #define RC_TOP_BITS 24
81 #define RC_MOVE_BITS 5
82 #define RC_MODEL_TOTAL_BITS 11
85 static int nofill(void *buffer
, unsigned int len
)
90 /* Called twice: once at startup and once in rc_normalize() */
91 static void INIT
rc_read(struct rc
*rc
)
93 rc
->buffer_size
= rc
->fill((char *)rc
->buffer
, LZMA_IOBUF_SIZE
);
94 if (rc
->buffer_size
<= 0)
95 error("unexpected EOF");
97 rc
->buffer_end
= rc
->buffer
+ rc
->buffer_size
;
101 static inline void INIT
rc_init(struct rc
*rc
,
102 int (*fill
)(void*, unsigned int),
103 char *buffer
, int buffer_size
)
109 rc
->buffer
= (uint8_t *)buffer
;
110 rc
->buffer_size
= buffer_size
;
111 rc
->buffer_end
= rc
->buffer
+ rc
->buffer_size
;
112 rc
->ptr
= rc
->buffer
;
115 rc
->range
= 0xFFFFFFFF;
118 static inline void INIT
rc_init_code(struct rc
*rc
)
122 for (i
= 0; i
< 5; i
++) {
123 if (rc
->ptr
>= rc
->buffer_end
)
125 rc
->code
= (rc
->code
<< 8) | *rc
->ptr
++;
130 /* Called once. TODO: bb_maybe_free() */
131 static inline void INIT
rc_free(struct rc
*rc
)
136 /* Called twice, but one callsite is in inline'd rc_is_bit_0_helper() */
137 static void INIT
rc_do_normalize(struct rc
*rc
)
139 if (rc
->ptr
>= rc
->buffer_end
)
142 rc
->code
= (rc
->code
<< 8) | *rc
->ptr
++;
144 static inline void INIT
rc_normalize(struct rc
*rc
)
146 if (rc
->range
< (1 << RC_TOP_BITS
))
151 /* Why rc_is_bit_0_helper exists?
152 *Because we want to always expose (rc->code < rc->bound) to optimizer
154 static inline uint32_t INIT
rc_is_bit_0_helper(struct rc
*rc
, uint16_t *p
)
157 rc
->bound
= *p
* (rc
->range
>> RC_MODEL_TOTAL_BITS
);
160 static inline int INIT
rc_is_bit_0(struct rc
*rc
, uint16_t *p
)
162 uint32_t t
= rc_is_bit_0_helper(rc
, p
);
166 /* Called ~10 times, but very small, thus inlined */
167 static inline void INIT
rc_update_bit_0(struct rc
*rc
, uint16_t *p
)
169 rc
->range
= rc
->bound
;
170 *p
+= ((1 << RC_MODEL_TOTAL_BITS
) - *p
) >> RC_MOVE_BITS
;
172 static inline void rc_update_bit_1(struct rc
*rc
, uint16_t *p
)
174 rc
->range
-= rc
->bound
;
175 rc
->code
-= rc
->bound
;
176 *p
-= *p
>> RC_MOVE_BITS
;
179 /* Called 4 times in unlzma loop */
180 static int INIT
rc_get_bit(struct rc
*rc
, uint16_t *p
, int *symbol
)
182 if (rc_is_bit_0(rc
, p
)) {
183 rc_update_bit_0(rc
, p
);
187 rc_update_bit_1(rc
, p
);
188 *symbol
= *symbol
* 2 + 1;
194 static inline int INIT
rc_direct_bit(struct rc
*rc
)
198 if (rc
->code
>= rc
->range
) {
199 rc
->code
-= rc
->range
;
206 static inline void INIT
207 rc_bit_tree_decode(struct rc
*rc
, uint16_t *p
, int num_levels
, int *symbol
)
213 rc_get_bit(rc
, p
+ *symbol
, symbol
);
214 *symbol
-= 1 << num_levels
;
219 * Small lzma deflate implementation.
220 * Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
222 * Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
223 * Copyright (C) 1999-2005 Igor Pavlov
231 } __attribute__ ((packed
)) ;
234 #define LZMA_BASE_SIZE 1846
235 #define LZMA_LIT_SIZE 768
237 #define LZMA_NUM_POS_BITS_MAX 4
239 #define LZMA_LEN_NUM_LOW_BITS 3
240 #define LZMA_LEN_NUM_MID_BITS 3
241 #define LZMA_LEN_NUM_HIGH_BITS 8
243 #define LZMA_LEN_CHOICE 0
244 #define LZMA_LEN_CHOICE_2 (LZMA_LEN_CHOICE + 1)
245 #define LZMA_LEN_LOW (LZMA_LEN_CHOICE_2 + 1)
246 #define LZMA_LEN_MID (LZMA_LEN_LOW \
247 + (1 << (LZMA_NUM_POS_BITS_MAX + LZMA_LEN_NUM_LOW_BITS)))
248 #define LZMA_LEN_HIGH (LZMA_LEN_MID \
249 +(1 << (LZMA_NUM_POS_BITS_MAX + LZMA_LEN_NUM_MID_BITS)))
250 #define LZMA_NUM_LEN_PROBS (LZMA_LEN_HIGH + (1 << LZMA_LEN_NUM_HIGH_BITS))
252 #define LZMA_NUM_STATES 12
253 #define LZMA_NUM_LIT_STATES 7
255 #define LZMA_START_POS_MODEL_INDEX 4
256 #define LZMA_END_POS_MODEL_INDEX 14
257 #define LZMA_NUM_FULL_DISTANCES (1 << (LZMA_END_POS_MODEL_INDEX >> 1))
259 #define LZMA_NUM_POS_SLOT_BITS 6
260 #define LZMA_NUM_LEN_TO_POS_STATES 4
262 #define LZMA_NUM_ALIGN_BITS 4
264 #define LZMA_MATCH_MIN_LEN 2
266 #define LZMA_IS_MATCH 0
267 #define LZMA_IS_REP (LZMA_IS_MATCH + (LZMA_NUM_STATES << LZMA_NUM_POS_BITS_MAX))
268 #define LZMA_IS_REP_G0 (LZMA_IS_REP + LZMA_NUM_STATES)
269 #define LZMA_IS_REP_G1 (LZMA_IS_REP_G0 + LZMA_NUM_STATES)
270 #define LZMA_IS_REP_G2 (LZMA_IS_REP_G1 + LZMA_NUM_STATES)
271 #define LZMA_IS_REP_0_LONG (LZMA_IS_REP_G2 + LZMA_NUM_STATES)
272 #define LZMA_POS_SLOT (LZMA_IS_REP_0_LONG \
273 + (LZMA_NUM_STATES << LZMA_NUM_POS_BITS_MAX))
274 #define LZMA_SPEC_POS (LZMA_POS_SLOT \
275 +(LZMA_NUM_LEN_TO_POS_STATES << LZMA_NUM_POS_SLOT_BITS))
276 #define LZMA_ALIGN (LZMA_SPEC_POS \
277 + LZMA_NUM_FULL_DISTANCES - LZMA_END_POS_MODEL_INDEX)
278 #define LZMA_LEN_CODER (LZMA_ALIGN + (1 << LZMA_NUM_ALIGN_BITS))
279 #define LZMA_REP_LEN_CODER (LZMA_LEN_CODER + LZMA_NUM_LEN_PROBS)
280 #define LZMA_LITERAL (LZMA_REP_LEN_CODER + LZMA_NUM_LEN_PROBS)
285 uint8_t previous_byte
;
289 int(*flush
)(void*, unsigned int);
290 struct lzma_header
*header
;
295 uint32_t rep0
, rep1
, rep2
, rep3
;
298 static inline size_t INIT
get_pos(struct writer
*wr
)
301 wr
->global_pos
+ wr
->buffer_pos
;
304 static inline uint8_t INIT
peek_old_byte(struct writer
*wr
,
309 while (offs
> wr
->header
->dict_size
)
310 offs
-= wr
->header
->dict_size
;
311 pos
= wr
->buffer_pos
- offs
;
312 return wr
->buffer
[pos
];
314 uint32_t pos
= wr
->buffer_pos
- offs
;
315 while (pos
>= wr
->header
->dict_size
)
316 pos
+= wr
->header
->dict_size
;
317 return wr
->buffer
[pos
];
322 static inline void INIT
write_byte(struct writer
*wr
, uint8_t byte
)
324 wr
->buffer
[wr
->buffer_pos
++] = wr
->previous_byte
= byte
;
325 if (wr
->flush
&& wr
->buffer_pos
== wr
->header
->dict_size
) {
327 wr
->global_pos
+= wr
->header
->dict_size
;
328 wr
->flush((char *)wr
->buffer
, wr
->header
->dict_size
);
333 static inline void INIT
copy_byte(struct writer
*wr
, uint32_t offs
)
335 write_byte(wr
, peek_old_byte(wr
, offs
));
338 static inline void INIT
copy_bytes(struct writer
*wr
,
339 uint32_t rep0
, int len
)
344 } while (len
!= 0 && wr
->buffer_pos
< wr
->header
->dst_size
);
347 static inline void INIT
process_bit0(struct writer
*wr
, struct rc
*rc
,
348 struct cstate
*cst
, uint16_t *p
,
349 int pos_state
, uint16_t *prob
,
350 int lc
, uint32_t literal_pos_mask
) {
352 rc_update_bit_0(rc
, prob
);
353 prob
= (p
+ LZMA_LITERAL
+
355 * (((get_pos(wr
) & literal_pos_mask
) << lc
)
356 + (wr
->previous_byte
>> (8 - lc
))))
359 if (cst
->state
>= LZMA_NUM_LIT_STATES
) {
360 int match_byte
= peek_old_byte(wr
, cst
->rep0
);
366 bit
= match_byte
& 0x100;
367 prob_lit
= prob
+ 0x100 + bit
+ mi
;
368 if (rc_get_bit(rc
, prob_lit
, &mi
)) {
375 } while (mi
< 0x100);
378 uint16_t *prob_lit
= prob
+ mi
;
379 rc_get_bit(rc
, prob_lit
, &mi
);
384 else if (cst
->state
< 10)
390 static inline void INIT
process_bit1(struct writer
*wr
, struct rc
*rc
,
391 struct cstate
*cst
, uint16_t *p
,
392 int pos_state
, uint16_t *prob
) {
398 rc_update_bit_1(rc
, prob
);
399 prob
= p
+ LZMA_IS_REP
+ cst
->state
;
400 if (rc_is_bit_0(rc
, prob
)) {
401 rc_update_bit_0(rc
, prob
);
402 cst
->rep3
= cst
->rep2
;
403 cst
->rep2
= cst
->rep1
;
404 cst
->rep1
= cst
->rep0
;
405 cst
->state
= cst
->state
< LZMA_NUM_LIT_STATES
? 0 : 3;
406 prob
= p
+ LZMA_LEN_CODER
;
408 rc_update_bit_1(rc
, prob
);
409 prob
= p
+ LZMA_IS_REP_G0
+ cst
->state
;
410 if (rc_is_bit_0(rc
, prob
)) {
411 rc_update_bit_0(rc
, prob
);
412 prob
= (p
+ LZMA_IS_REP_0_LONG
414 LZMA_NUM_POS_BITS_MAX
) +
416 if (rc_is_bit_0(rc
, prob
)) {
417 rc_update_bit_0(rc
, prob
);
419 cst
->state
= cst
->state
< LZMA_NUM_LIT_STATES
?
421 copy_byte(wr
, cst
->rep0
);
424 rc_update_bit_1(rc
, prob
);
429 rc_update_bit_1(rc
, prob
);
430 prob
= p
+ LZMA_IS_REP_G1
+ cst
->state
;
431 if (rc_is_bit_0(rc
, prob
)) {
432 rc_update_bit_0(rc
, prob
);
433 distance
= cst
->rep1
;
435 rc_update_bit_1(rc
, prob
);
436 prob
= p
+ LZMA_IS_REP_G2
+ cst
->state
;
437 if (rc_is_bit_0(rc
, prob
)) {
438 rc_update_bit_0(rc
, prob
);
439 distance
= cst
->rep2
;
441 rc_update_bit_1(rc
, prob
);
442 distance
= cst
->rep3
;
443 cst
->rep3
= cst
->rep2
;
445 cst
->rep2
= cst
->rep1
;
447 cst
->rep1
= cst
->rep0
;
448 cst
->rep0
= distance
;
450 cst
->state
= cst
->state
< LZMA_NUM_LIT_STATES
? 8 : 11;
451 prob
= p
+ LZMA_REP_LEN_CODER
;
454 prob_len
= prob
+ LZMA_LEN_CHOICE
;
455 if (rc_is_bit_0(rc
, prob_len
)) {
456 rc_update_bit_0(rc
, prob_len
);
457 prob_len
= (prob
+ LZMA_LEN_LOW
459 LZMA_LEN_NUM_LOW_BITS
));
461 num_bits
= LZMA_LEN_NUM_LOW_BITS
;
463 rc_update_bit_1(rc
, prob_len
);
464 prob_len
= prob
+ LZMA_LEN_CHOICE_2
;
465 if (rc_is_bit_0(rc
, prob_len
)) {
466 rc_update_bit_0(rc
, prob_len
);
467 prob_len
= (prob
+ LZMA_LEN_MID
469 LZMA_LEN_NUM_MID_BITS
));
470 offset
= 1 << LZMA_LEN_NUM_LOW_BITS
;
471 num_bits
= LZMA_LEN_NUM_MID_BITS
;
473 rc_update_bit_1(rc
, prob_len
);
474 prob_len
= prob
+ LZMA_LEN_HIGH
;
475 offset
= ((1 << LZMA_LEN_NUM_LOW_BITS
)
476 + (1 << LZMA_LEN_NUM_MID_BITS
));
477 num_bits
= LZMA_LEN_NUM_HIGH_BITS
;
481 rc_bit_tree_decode(rc
, prob_len
, num_bits
, &len
);
484 if (cst
->state
< 4) {
487 cst
->state
+= LZMA_NUM_LIT_STATES
;
491 LZMA_NUM_LEN_TO_POS_STATES
? len
:
492 LZMA_NUM_LEN_TO_POS_STATES
- 1)
493 << LZMA_NUM_POS_SLOT_BITS
);
494 rc_bit_tree_decode(rc
, prob
,
495 LZMA_NUM_POS_SLOT_BITS
,
497 if (pos_slot
>= LZMA_START_POS_MODEL_INDEX
) {
499 num_bits
= (pos_slot
>> 1) - 1;
500 cst
->rep0
= 2 | (pos_slot
& 1);
501 if (pos_slot
< LZMA_END_POS_MODEL_INDEX
) {
502 cst
->rep0
<<= num_bits
;
503 prob
= p
+ LZMA_SPEC_POS
+
504 cst
->rep0
- pos_slot
- 1;
506 num_bits
-= LZMA_NUM_ALIGN_BITS
;
508 cst
->rep0
= (cst
->rep0
<< 1) |
510 prob
= p
+ LZMA_ALIGN
;
511 cst
->rep0
<<= LZMA_NUM_ALIGN_BITS
;
512 num_bits
= LZMA_NUM_ALIGN_BITS
;
517 if (rc_get_bit(rc
, prob
+ mi
, &mi
))
522 cst
->rep0
= pos_slot
;
523 if (++(cst
->rep0
) == 0)
527 len
+= LZMA_MATCH_MIN_LEN
;
529 copy_bytes(wr
, cst
->rep0
, len
);
534 STATIC
inline int INIT
unlzma(unsigned char *buf
, int in_len
,
535 int(*fill
)(void*, unsigned int),
536 int(*flush
)(void*, unsigned int),
537 unsigned char *output
,
539 void(*error_fn
)(char *x
)
542 struct lzma_header header
;
544 uint32_t pos_state_mask
;
545 uint32_t literal_pos_mask
;
552 unsigned char *inbuf
;
555 set_error_fn(error_fn
);
560 inbuf
= malloc(LZMA_IOBUF_SIZE
);
562 error("Could not allocate input bufer");
567 cst
.rep0
= cst
.rep1
= cst
.rep2
= cst
.rep3
= 1;
572 wr
.previous_byte
= 0;
575 rc_init(&rc
, fill
, inbuf
, in_len
);
577 for (i
= 0; i
< sizeof(header
); i
++) {
578 if (rc
.ptr
>= rc
.buffer_end
)
580 ((unsigned char *)&header
)[i
] = *rc
.ptr
++;
583 if (header
.pos
>= (9 * 5 * 5))
598 pos_state_mask
= (1 << pb
) - 1;
599 literal_pos_mask
= (1 << lp
) - 1;
601 ENDIAN_CONVERT(header
.dict_size
);
602 ENDIAN_CONVERT(header
.dst_size
);
604 if (header
.dict_size
== 0)
605 header
.dict_size
= 1;
610 wr
.bufsize
= MIN(header
.dst_size
, header
.dict_size
);
611 wr
.buffer
= large_malloc(wr
.bufsize
);
613 if (wr
.buffer
== NULL
)
616 num_probs
= LZMA_BASE_SIZE
+ (LZMA_LIT_SIZE
<< (lc
+ lp
));
617 p
= (uint16_t *) large_malloc(num_probs
* sizeof(*p
));
620 num_probs
= LZMA_LITERAL
+ (LZMA_LIT_SIZE
<< (lc
+ lp
));
621 for (i
= 0; i
< num_probs
; i
++)
622 p
[i
] = (1 << RC_MODEL_TOTAL_BITS
) >> 1;
626 while (get_pos(&wr
) < header
.dst_size
) {
627 int pos_state
= get_pos(&wr
) & pos_state_mask
;
628 uint16_t *prob
= p
+ LZMA_IS_MATCH
+
629 (cst
.state
<< LZMA_NUM_POS_BITS_MAX
) + pos_state
;
630 if (rc_is_bit_0(&rc
, prob
))
631 process_bit0(&wr
, &rc
, &cst
, p
, pos_state
, prob
,
632 lc
, literal_pos_mask
);
634 process_bit1(&wr
, &rc
, &cst
, p
, pos_state
, prob
);
641 *posp
= rc
.ptr
-rc
.buffer
;
643 wr
.flush(wr
.buffer
, wr
.buffer_pos
);
648 large_free(wr
.buffer
);
657 STATIC
int INIT
decompress(unsigned char *buf
, int in_len
,
658 int(*fill
)(void*, unsigned int),
659 int(*flush
)(void*, unsigned int),
660 unsigned char *output
,
662 void(*error_fn
)(char *x
)
665 return unlzma(buf
, in_len
- 4, fill
, flush
, output
, posp
, error_fn
);