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 /* Called twice: once at startup and once in rc_normalize() */
86 static void INIT
rc_read(struct rc
*rc
)
88 rc
->buffer_size
= rc
->fill((char *)rc
->buffer
, LZMA_IOBUF_SIZE
);
89 if (rc
->buffer_size
<= 0)
90 error("unexpected EOF");
92 rc
->buffer_end
= rc
->buffer
+ rc
->buffer_size
;
96 static inline void INIT
rc_init(struct rc
*rc
,
97 int (*fill
)(void*, unsigned int),
98 char *buffer
, int buffer_size
)
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
;
107 rc
->range
= 0xFFFFFFFF;
110 static inline void INIT
rc_init_code(struct rc
*rc
)
114 for (i
= 0; i
< 5; i
++) {
115 if (rc
->ptr
>= rc
->buffer_end
)
117 rc
->code
= (rc
->code
<< 8) | *rc
->ptr
++;
122 /* Called once. TODO: bb_maybe_free() */
123 static inline void INIT
rc_free(struct rc
*rc
)
128 /* Called twice, but one callsite is in inline'd rc_is_bit_0_helper() */
129 static void INIT
rc_do_normalize(struct rc
*rc
)
131 if (rc
->ptr
>= rc
->buffer_end
)
134 rc
->code
= (rc
->code
<< 8) | *rc
->ptr
++;
136 static inline void INIT
rc_normalize(struct rc
*rc
)
138 if (rc
->range
< (1 << RC_TOP_BITS
))
143 /* Why rc_is_bit_0_helper exists?
144 *Because we want to always expose (rc->code < rc->bound) to optimizer
146 static inline uint32_t INIT
rc_is_bit_0_helper(struct rc
*rc
, uint16_t *p
)
149 rc
->bound
= *p
* (rc
->range
>> RC_MODEL_TOTAL_BITS
);
152 static inline int INIT
rc_is_bit_0(struct rc
*rc
, uint16_t *p
)
154 uint32_t t
= rc_is_bit_0_helper(rc
, p
);
158 /* Called ~10 times, but very small, thus inlined */
159 static inline void INIT
rc_update_bit_0(struct rc
*rc
, uint16_t *p
)
161 rc
->range
= rc
->bound
;
162 *p
+= ((1 << RC_MODEL_TOTAL_BITS
) - *p
) >> RC_MOVE_BITS
;
164 static inline void rc_update_bit_1(struct rc
*rc
, uint16_t *p
)
166 rc
->range
-= rc
->bound
;
167 rc
->code
-= rc
->bound
;
168 *p
-= *p
>> RC_MOVE_BITS
;
171 /* Called 4 times in unlzma loop */
172 static int INIT
rc_get_bit(struct rc
*rc
, uint16_t *p
, int *symbol
)
174 if (rc_is_bit_0(rc
, p
)) {
175 rc_update_bit_0(rc
, p
);
179 rc_update_bit_1(rc
, p
);
180 *symbol
= *symbol
* 2 + 1;
186 static inline int INIT
rc_direct_bit(struct rc
*rc
)
190 if (rc
->code
>= rc
->range
) {
191 rc
->code
-= rc
->range
;
198 static inline void INIT
199 rc_bit_tree_decode(struct rc
*rc
, uint16_t *p
, int num_levels
, int *symbol
)
205 rc_get_bit(rc
, p
+ *symbol
, symbol
);
206 *symbol
-= 1 << num_levels
;
211 * Small lzma deflate implementation.
212 * Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
214 * Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
215 * Copyright (C) 1999-2005 Igor Pavlov
223 } __attribute__ ((packed
)) ;
226 #define LZMA_BASE_SIZE 1846
227 #define LZMA_LIT_SIZE 768
229 #define LZMA_NUM_POS_BITS_MAX 4
231 #define LZMA_LEN_NUM_LOW_BITS 3
232 #define LZMA_LEN_NUM_MID_BITS 3
233 #define LZMA_LEN_NUM_HIGH_BITS 8
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))
244 #define LZMA_NUM_STATES 12
245 #define LZMA_NUM_LIT_STATES 7
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))
251 #define LZMA_NUM_POS_SLOT_BITS 6
252 #define LZMA_NUM_LEN_TO_POS_STATES 4
254 #define LZMA_NUM_ALIGN_BITS 4
256 #define LZMA_MATCH_MIN_LEN 2
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)
277 uint8_t previous_byte
;
281 int(*flush
)(void*, unsigned int);
282 struct lzma_header
*header
;
287 uint32_t rep0
, rep1
, rep2
, rep3
;
290 static inline size_t INIT
get_pos(struct writer
*wr
)
293 wr
->global_pos
+ wr
->buffer_pos
;
296 static inline uint8_t INIT
peek_old_byte(struct writer
*wr
,
301 while (offs
> wr
->header
->dict_size
)
302 offs
-= wr
->header
->dict_size
;
303 pos
= wr
->buffer_pos
- offs
;
304 return wr
->buffer
[pos
];
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
];
314 static inline void INIT
write_byte(struct writer
*wr
, uint8_t byte
)
316 wr
->buffer
[wr
->buffer_pos
++] = wr
->previous_byte
= byte
;
317 if (wr
->flush
&& wr
->buffer_pos
== wr
->header
->dict_size
) {
319 wr
->global_pos
+= wr
->header
->dict_size
;
320 wr
->flush((char *)wr
->buffer
, wr
->header
->dict_size
);
325 static inline void INIT
copy_byte(struct writer
*wr
, uint32_t offs
)
327 write_byte(wr
, peek_old_byte(wr
, offs
));
330 static inline void INIT
copy_bytes(struct writer
*wr
,
331 uint32_t rep0
, int len
)
336 } while (len
!= 0 && wr
->buffer_pos
< wr
->header
->dst_size
);
339 static 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
) {
344 rc_update_bit_0(rc
, prob
);
345 prob
= (p
+ LZMA_LITERAL
+
347 * (((get_pos(wr
) & literal_pos_mask
) << lc
)
348 + (wr
->previous_byte
>> (8 - lc
))))
351 if (cst
->state
>= LZMA_NUM_LIT_STATES
) {
352 int match_byte
= peek_old_byte(wr
, cst
->rep0
);
358 bit
= match_byte
& 0x100;
359 prob_lit
= prob
+ 0x100 + bit
+ mi
;
360 if (rc_get_bit(rc
, prob_lit
, &mi
)) {
367 } while (mi
< 0x100);
370 uint16_t *prob_lit
= prob
+ mi
;
371 rc_get_bit(rc
, prob_lit
, &mi
);
376 else if (cst
->state
< 10)
382 static 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
) {
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
;
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
406 LZMA_NUM_POS_BITS_MAX
) +
408 if (rc_is_bit_0(rc
, prob
)) {
409 rc_update_bit_0(rc
, prob
);
411 cst
->state
= cst
->state
< LZMA_NUM_LIT_STATES
?
413 copy_byte(wr
, cst
->rep0
);
416 rc_update_bit_1(rc
, prob
);
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
;
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
;
433 rc_update_bit_1(rc
, prob
);
434 distance
= cst
->rep3
;
435 cst
->rep3
= cst
->rep2
;
437 cst
->rep2
= cst
->rep1
;
439 cst
->rep1
= cst
->rep0
;
440 cst
->rep0
= distance
;
442 cst
->state
= cst
->state
< LZMA_NUM_LIT_STATES
? 8 : 11;
443 prob
= p
+ LZMA_REP_LEN_CODER
;
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
451 LZMA_LEN_NUM_LOW_BITS
));
453 num_bits
= LZMA_LEN_NUM_LOW_BITS
;
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
461 LZMA_LEN_NUM_MID_BITS
));
462 offset
= 1 << LZMA_LEN_NUM_LOW_BITS
;
463 num_bits
= LZMA_LEN_NUM_MID_BITS
;
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
;
473 rc_bit_tree_decode(rc
, prob_len
, num_bits
, &len
);
476 if (cst
->state
< 4) {
479 cst
->state
+= LZMA_NUM_LIT_STATES
;
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
,
489 if (pos_slot
>= LZMA_START_POS_MODEL_INDEX
) {
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;
498 num_bits
-= LZMA_NUM_ALIGN_BITS
;
500 cst
->rep0
= (cst
->rep0
<< 1) |
502 prob
= p
+ LZMA_ALIGN
;
503 cst
->rep0
<<= LZMA_NUM_ALIGN_BITS
;
504 num_bits
= LZMA_NUM_ALIGN_BITS
;
509 if (rc_get_bit(rc
, prob
+ mi
, &mi
))
514 cst
->rep0
= pos_slot
;
515 if (++(cst
->rep0
) == 0)
519 len
+= LZMA_MATCH_MIN_LEN
;
521 copy_bytes(wr
, cst
->rep0
, len
);
526 STATIC
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
,
531 void(*error_fn
)(char *x
)
534 struct lzma_header header
;
536 uint32_t pos_state_mask
;
537 uint32_t literal_pos_mask
;
544 unsigned char *inbuf
;
547 set_error_fn(error_fn
);
552 inbuf
= malloc(LZMA_IOBUF_SIZE
);
554 error("Could not allocate input bufer");
559 cst
.rep0
= cst
.rep1
= cst
.rep2
= cst
.rep3
= 1;
564 wr
.previous_byte
= 0;
567 rc_init(&rc
, fill
, inbuf
, in_len
);
569 for (i
= 0; i
< sizeof(header
); i
++) {
570 if (rc
.ptr
>= rc
.buffer_end
)
572 ((unsigned char *)&header
)[i
] = *rc
.ptr
++;
575 if (header
.pos
>= (9 * 5 * 5))
590 pos_state_mask
= (1 << pb
) - 1;
591 literal_pos_mask
= (1 << lp
) - 1;
593 ENDIAN_CONVERT(header
.dict_size
);
594 ENDIAN_CONVERT(header
.dst_size
);
596 if (header
.dict_size
== 0)
597 header
.dict_size
= 1;
602 wr
.bufsize
= MIN(header
.dst_size
, header
.dict_size
);
603 wr
.buffer
= large_malloc(wr
.bufsize
);
605 if (wr
.buffer
== NULL
)
608 num_probs
= LZMA_BASE_SIZE
+ (LZMA_LIT_SIZE
<< (lc
+ lp
));
609 p
= (uint16_t *) large_malloc(num_probs
* sizeof(*p
));
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;
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
);
626 process_bit1(&wr
, &rc
, &cst
, p
, pos_state
, prob
);
633 *posp
= rc
.ptr
-rc
.buffer
;
635 wr
.flush(wr
.buffer
, wr
.buffer_pos
);
640 large_free(wr
.buffer
);
649 STATIC
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
,
654 void(*error_fn
)(char *x
)
657 return unlzma(buf
, in_len
- 4, fill
, flush
, output
, posp
, error_fn
);