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
33 #include <linux/decompress/unlzma.h>
36 #include <linux/decompress/mm.h>
38 #define MIN(a, b) (((a) < (b)) ? (a) : (b))
40 static long long INIT
read_int(unsigned char *ptr
, int size
)
45 for (i
= 0; i
< size
; i
++)
46 ret
= (ret
<< 8) | ptr
[size
-i
-1];
50 #define ENDIAN_CONVERT(x) \
51 x = (typeof(x))read_int((unsigned char *)&x, sizeof(x))
54 /* Small range coder implementation for lzma.
55 *Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
57 *Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
58 *Copyright (c) 1999-2005 Igor Pavlov
61 #include <linux/compiler.h>
63 #define LZMA_IOBUF_SIZE 0x10000
66 int (*fill
)(void*, unsigned int);
77 #define RC_TOP_BITS 24
78 #define RC_MOVE_BITS 5
79 #define RC_MODEL_TOTAL_BITS 11
82 /* Called twice: once at startup and once in rc_normalize() */
83 static void INIT
rc_read(struct rc
*rc
)
85 rc
->buffer_size
= rc
->fill((char *)rc
->buffer
, LZMA_IOBUF_SIZE
);
86 if (rc
->buffer_size
<= 0)
87 error("unexpected EOF");
89 rc
->buffer_end
= rc
->buffer
+ rc
->buffer_size
;
93 static inline void INIT
rc_init(struct rc
*rc
,
94 int (*fill
)(void*, unsigned int),
95 char *buffer
, int buffer_size
)
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
;
104 rc
->range
= 0xFFFFFFFF;
107 static inline void INIT
rc_init_code(struct rc
*rc
)
111 for (i
= 0; i
< 5; i
++) {
112 if (rc
->ptr
>= rc
->buffer_end
)
114 rc
->code
= (rc
->code
<< 8) | *rc
->ptr
++;
119 /* Called once. TODO: bb_maybe_free() */
120 static inline void INIT
rc_free(struct rc
*rc
)
125 /* Called twice, but one callsite is in inline'd rc_is_bit_0_helper() */
126 static void INIT
rc_do_normalize(struct rc
*rc
)
128 if (rc
->ptr
>= rc
->buffer_end
)
131 rc
->code
= (rc
->code
<< 8) | *rc
->ptr
++;
133 static inline void INIT
rc_normalize(struct rc
*rc
)
135 if (rc
->range
< (1 << RC_TOP_BITS
))
140 /* Why rc_is_bit_0_helper exists?
141 *Because we want to always expose (rc->code < rc->bound) to optimizer
143 static inline uint32_t INIT
rc_is_bit_0_helper(struct rc
*rc
, uint16_t *p
)
146 rc
->bound
= *p
* (rc
->range
>> RC_MODEL_TOTAL_BITS
);
149 static inline int INIT
rc_is_bit_0(struct rc
*rc
, uint16_t *p
)
151 uint32_t t
= rc_is_bit_0_helper(rc
, p
);
155 /* Called ~10 times, but very small, thus inlined */
156 static inline void INIT
rc_update_bit_0(struct rc
*rc
, uint16_t *p
)
158 rc
->range
= rc
->bound
;
159 *p
+= ((1 << RC_MODEL_TOTAL_BITS
) - *p
) >> RC_MOVE_BITS
;
161 static inline void rc_update_bit_1(struct rc
*rc
, uint16_t *p
)
163 rc
->range
-= rc
->bound
;
164 rc
->code
-= rc
->bound
;
165 *p
-= *p
>> RC_MOVE_BITS
;
168 /* Called 4 times in unlzma loop */
169 static int INIT
rc_get_bit(struct rc
*rc
, uint16_t *p
, int *symbol
)
171 if (rc_is_bit_0(rc
, p
)) {
172 rc_update_bit_0(rc
, p
);
176 rc_update_bit_1(rc
, p
);
177 *symbol
= *symbol
* 2 + 1;
183 static inline int INIT
rc_direct_bit(struct rc
*rc
)
187 if (rc
->code
>= rc
->range
) {
188 rc
->code
-= rc
->range
;
195 static inline void INIT
196 rc_bit_tree_decode(struct rc
*rc
, uint16_t *p
, int num_levels
, int *symbol
)
202 rc_get_bit(rc
, p
+ *symbol
, symbol
);
203 *symbol
-= 1 << num_levels
;
208 * Small lzma deflate implementation.
209 * Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
211 * Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
212 * Copyright (C) 1999-2005 Igor Pavlov
220 } __attribute__ ((packed
)) ;
223 #define LZMA_BASE_SIZE 1846
224 #define LZMA_LIT_SIZE 768
226 #define LZMA_NUM_POS_BITS_MAX 4
228 #define LZMA_LEN_NUM_LOW_BITS 3
229 #define LZMA_LEN_NUM_MID_BITS 3
230 #define LZMA_LEN_NUM_HIGH_BITS 8
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))
241 #define LZMA_NUM_STATES 12
242 #define LZMA_NUM_LIT_STATES 7
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))
248 #define LZMA_NUM_POS_SLOT_BITS 6
249 #define LZMA_NUM_LEN_TO_POS_STATES 4
251 #define LZMA_NUM_ALIGN_BITS 4
253 #define LZMA_MATCH_MIN_LEN 2
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)
274 uint8_t previous_byte
;
278 int(*flush
)(void*, unsigned int);
279 struct lzma_header
*header
;
284 uint32_t rep0
, rep1
, rep2
, rep3
;
287 static inline size_t INIT
get_pos(struct writer
*wr
)
290 wr
->global_pos
+ wr
->buffer_pos
;
293 static inline uint8_t INIT
peek_old_byte(struct writer
*wr
,
298 while (offs
> wr
->header
->dict_size
)
299 offs
-= wr
->header
->dict_size
;
300 pos
= wr
->buffer_pos
- offs
;
301 return wr
->buffer
[pos
];
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
];
311 static inline void INIT
write_byte(struct writer
*wr
, uint8_t byte
)
313 wr
->buffer
[wr
->buffer_pos
++] = wr
->previous_byte
= byte
;
314 if (wr
->flush
&& wr
->buffer_pos
== wr
->header
->dict_size
) {
316 wr
->global_pos
+= wr
->header
->dict_size
;
317 wr
->flush((char *)wr
->buffer
, wr
->header
->dict_size
);
322 static inline void INIT
copy_byte(struct writer
*wr
, uint32_t offs
)
324 write_byte(wr
, peek_old_byte(wr
, offs
));
327 static inline void INIT
copy_bytes(struct writer
*wr
,
328 uint32_t rep0
, int len
)
333 } while (len
!= 0 && wr
->buffer_pos
< wr
->header
->dst_size
);
336 static 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
) {
341 rc_update_bit_0(rc
, prob
);
342 prob
= (p
+ LZMA_LITERAL
+
344 * (((get_pos(wr
) & literal_pos_mask
) << lc
)
345 + (wr
->previous_byte
>> (8 - lc
))))
348 if (cst
->state
>= LZMA_NUM_LIT_STATES
) {
349 int match_byte
= peek_old_byte(wr
, cst
->rep0
);
355 bit
= match_byte
& 0x100;
356 prob_lit
= prob
+ 0x100 + bit
+ mi
;
357 if (rc_get_bit(rc
, prob_lit
, &mi
)) {
364 } while (mi
< 0x100);
367 uint16_t *prob_lit
= prob
+ mi
;
368 rc_get_bit(rc
, prob_lit
, &mi
);
373 else if (cst
->state
< 10)
379 static 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
) {
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
;
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
403 LZMA_NUM_POS_BITS_MAX
) +
405 if (rc_is_bit_0(rc
, prob
)) {
406 rc_update_bit_0(rc
, prob
);
408 cst
->state
= cst
->state
< LZMA_NUM_LIT_STATES
?
410 copy_byte(wr
, cst
->rep0
);
413 rc_update_bit_1(rc
, prob
);
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
;
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
;
430 rc_update_bit_1(rc
, prob
);
431 distance
= cst
->rep3
;
432 cst
->rep3
= cst
->rep2
;
434 cst
->rep2
= cst
->rep1
;
436 cst
->rep1
= cst
->rep0
;
437 cst
->rep0
= distance
;
439 cst
->state
= cst
->state
< LZMA_NUM_LIT_STATES
? 8 : 11;
440 prob
= p
+ LZMA_REP_LEN_CODER
;
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
448 LZMA_LEN_NUM_LOW_BITS
));
450 num_bits
= LZMA_LEN_NUM_LOW_BITS
;
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
458 LZMA_LEN_NUM_MID_BITS
));
459 offset
= 1 << LZMA_LEN_NUM_LOW_BITS
;
460 num_bits
= LZMA_LEN_NUM_MID_BITS
;
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
;
470 rc_bit_tree_decode(rc
, prob_len
, num_bits
, &len
);
473 if (cst
->state
< 4) {
476 cst
->state
+= LZMA_NUM_LIT_STATES
;
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
,
486 if (pos_slot
>= LZMA_START_POS_MODEL_INDEX
) {
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;
495 num_bits
-= LZMA_NUM_ALIGN_BITS
;
497 cst
->rep0
= (cst
->rep0
<< 1) |
499 prob
= p
+ LZMA_ALIGN
;
500 cst
->rep0
<<= LZMA_NUM_ALIGN_BITS
;
501 num_bits
= LZMA_NUM_ALIGN_BITS
;
506 if (rc_get_bit(rc
, prob
+ mi
, &mi
))
511 cst
->rep0
= pos_slot
;
512 if (++(cst
->rep0
) == 0)
516 len
+= LZMA_MATCH_MIN_LEN
;
518 copy_bytes(wr
, cst
->rep0
, len
);
523 STATIC
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
,
528 void(*error_fn
)(char *x
)
531 struct lzma_header header
;
533 uint32_t pos_state_mask
;
534 uint32_t literal_pos_mask
;
541 unsigned char *inbuf
;
544 set_error_fn(error_fn
);
546 in_len
-= 4; /* Uncompressed size hack active in pre-boot
551 inbuf
= malloc(LZMA_IOBUF_SIZE
);
553 error("Could not allocate input bufer");
558 cst
.rep0
= cst
.rep1
= cst
.rep2
= cst
.rep3
= 1;
563 wr
.previous_byte
= 0;
566 rc_init(&rc
, fill
, inbuf
, in_len
);
568 for (i
= 0; i
< sizeof(header
); i
++) {
569 if (rc
.ptr
>= rc
.buffer_end
)
571 ((unsigned char *)&header
)[i
] = *rc
.ptr
++;
574 if (header
.pos
>= (9 * 5 * 5))
589 pos_state_mask
= (1 << pb
) - 1;
590 literal_pos_mask
= (1 << lp
) - 1;
592 ENDIAN_CONVERT(header
.dict_size
);
593 ENDIAN_CONVERT(header
.dst_size
);
595 if (header
.dict_size
== 0)
596 header
.dict_size
= 1;
601 wr
.bufsize
= MIN(header
.dst_size
, header
.dict_size
);
602 wr
.buffer
= large_malloc(wr
.bufsize
);
604 if (wr
.buffer
== NULL
)
607 num_probs
= LZMA_BASE_SIZE
+ (LZMA_LIT_SIZE
<< (lc
+ lp
));
608 p
= (uint16_t *) large_malloc(num_probs
* sizeof(*p
));
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;
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
);
625 process_bit1(&wr
, &rc
, &cst
, p
, pos_state
, prob
);
632 *posp
= rc
.ptr
-rc
.buffer
;
634 wr
.flush(wr
.buffer
, wr
.buffer_pos
);
639 large_free(wr
.buffer
);
647 #define decompress unlzma