Root/lib/decompress_unlzma.c

1/* 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
32#ifdef STATIC
33#define PREBOOT
34#else
35#include <linux/decompress/unlzma.h>
36#include <linux/slab.h>
37#endif /* STATIC */
38
39#include <linux/decompress/mm.h>
40
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
85static int nofill(void *buffer, unsigned int len)
86{
87    return -1;
88}
89
90/* Called twice: once at startup and once in rc_normalize() */
91static void INIT rc_read(struct rc *rc)
92{
93    rc->buffer_size = rc->fill((char *)rc->buffer, LZMA_IOBUF_SIZE);
94    if (rc->buffer_size <= 0)
95        error("unexpected EOF");
96    rc->ptr = rc->buffer;
97    rc->buffer_end = rc->buffer + rc->buffer_size;
98}
99
100/* Called once */
101static inline void INIT rc_init(struct rc *rc,
102                       int (*fill)(void*, unsigned int),
103                       char *buffer, int buffer_size)
104{
105    if (fill)
106        rc->fill = fill;
107    else
108        rc->fill = nofill;
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;
113
114    rc->code = 0;
115    rc->range = 0xFFFFFFFF;
116}
117
118static inline void INIT rc_init_code(struct rc *rc)
119{
120    int i;
121
122    for (i = 0; i < 5; i++) {
123        if (rc->ptr >= rc->buffer_end)
124            rc_read(rc);
125        rc->code = (rc->code << 8) | *rc->ptr++;
126    }
127}
128
129
130/* Called once. TODO: bb_maybe_free() */
131static inline void INIT rc_free(struct rc *rc)
132{
133    free(rc->buffer);
134}
135
136/* Called twice, but one callsite is in inline'd rc_is_bit_0_helper() */
137static void INIT rc_do_normalize(struct rc *rc)
138{
139    if (rc->ptr >= rc->buffer_end)
140        rc_read(rc);
141    rc->range <<= 8;
142    rc->code = (rc->code << 8) | *rc->ptr++;
143}
144static inline void INIT rc_normalize(struct rc *rc)
145{
146    if (rc->range < (1 << RC_TOP_BITS))
147        rc_do_normalize(rc);
148}
149
150/* Called 9 times */
151/* Why rc_is_bit_0_helper exists?
152 *Because we want to always expose (rc->code < rc->bound) to optimizer
153 */
154static inline uint32_t INIT rc_is_bit_0_helper(struct rc *rc, uint16_t *p)
155{
156    rc_normalize(rc);
157    rc->bound = *p * (rc->range >> RC_MODEL_TOTAL_BITS);
158    return rc->bound;
159}
160static inline int INIT rc_is_bit_0(struct rc *rc, uint16_t *p)
161{
162    uint32_t t = rc_is_bit_0_helper(rc, p);
163    return rc->code < t;
164}
165
166/* Called ~10 times, but very small, thus inlined */
167static inline void INIT rc_update_bit_0(struct rc *rc, uint16_t *p)
168{
169    rc->range = rc->bound;
170    *p += ((1 << RC_MODEL_TOTAL_BITS) - *p) >> RC_MOVE_BITS;
171}
172static inline void rc_update_bit_1(struct rc *rc, uint16_t *p)
173{
174    rc->range -= rc->bound;
175    rc->code -= rc->bound;
176    *p -= *p >> RC_MOVE_BITS;
177}
178
179/* Called 4 times in unlzma loop */
180static int INIT rc_get_bit(struct rc *rc, uint16_t *p, int *symbol)
181{
182    if (rc_is_bit_0(rc, p)) {
183        rc_update_bit_0(rc, p);
184        *symbol *= 2;
185        return 0;
186    } else {
187        rc_update_bit_1(rc, p);
188        *symbol = *symbol * 2 + 1;
189        return 1;
190    }
191}
192
193/* Called once */
194static inline int INIT rc_direct_bit(struct rc *rc)
195{
196    rc_normalize(rc);
197    rc->range >>= 1;
198    if (rc->code >= rc->range) {
199        rc->code -= rc->range;
200        return 1;
201    }
202    return 0;
203}
204
205/* Called twice */
206static inline void INIT
207rc_bit_tree_decode(struct rc *rc, uint16_t *p, int num_levels, int *symbol)
208{
209    int i = num_levels;
210
211    *symbol = 1;
212    while (i--)
213        rc_get_bit(rc, p + *symbol, symbol);
214    *symbol -= 1 << num_levels;
215}
216
217
218/*
219 * Small lzma deflate implementation.
220 * Copyright (C) 2006 Aurelien Jacobs < aurel@gnuage.org >
221 *
222 * Based on LzmaDecode.c from the LZMA SDK 4.22 (http://www.7-zip.org/)
223 * Copyright (C) 1999-2005 Igor Pavlov
224 */
225
226
227struct lzma_header {
228    uint8_t pos;
229    uint32_t dict_size;
230    uint64_t dst_size;
231} __attribute__ ((packed)) ;
232
233
234#define LZMA_BASE_SIZE 1846
235#define LZMA_LIT_SIZE 768
236
237#define LZMA_NUM_POS_BITS_MAX 4
238
239#define LZMA_LEN_NUM_LOW_BITS 3
240#define LZMA_LEN_NUM_MID_BITS 3
241#define LZMA_LEN_NUM_HIGH_BITS 8
242
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))
251
252#define LZMA_NUM_STATES 12
253#define LZMA_NUM_LIT_STATES 7
254
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))
258
259#define LZMA_NUM_POS_SLOT_BITS 6
260#define LZMA_NUM_LEN_TO_POS_STATES 4
261
262#define LZMA_NUM_ALIGN_BITS 4
263
264#define LZMA_MATCH_MIN_LEN 2
265
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)
281
282
283struct writer {
284    uint8_t *buffer;
285    uint8_t previous_byte;
286    size_t buffer_pos;
287    int bufsize;
288    size_t global_pos;
289    int(*flush)(void*, unsigned int);
290    struct lzma_header *header;
291};
292
293struct cstate {
294    int state;
295    uint32_t rep0, rep1, rep2, rep3;
296};
297
298static inline size_t INIT get_pos(struct writer *wr)
299{
300    return
301        wr->global_pos + wr->buffer_pos;
302}
303
304static inline uint8_t INIT peek_old_byte(struct writer *wr,
305                        uint32_t offs)
306{
307    if (!wr->flush) {
308        int32_t pos;
309        while (offs > wr->header->dict_size)
310            offs -= wr->header->dict_size;
311        pos = wr->buffer_pos - offs;
312        return wr->buffer[pos];
313    } else {
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];
318    }
319
320}
321
322static inline void INIT write_byte(struct writer *wr, uint8_t byte)
323{
324    wr->buffer[wr->buffer_pos++] = wr->previous_byte = byte;
325    if (wr->flush && wr->buffer_pos == wr->header->dict_size) {
326        wr->buffer_pos = 0;
327        wr->global_pos += wr->header->dict_size;
328        wr->flush((char *)wr->buffer, wr->header->dict_size);
329    }
330}
331
332
333static inline void INIT copy_byte(struct writer *wr, uint32_t offs)
334{
335    write_byte(wr, peek_old_byte(wr, offs));
336}
337
338static inline void INIT copy_bytes(struct writer *wr,
339                     uint32_t rep0, int len)
340{
341    do {
342        copy_byte(wr, rep0);
343        len--;
344    } while (len != 0 && wr->buffer_pos < wr->header->dst_size);
345}
346
347static 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) {
351    int mi = 1;
352    rc_update_bit_0(rc, prob);
353    prob = (p + LZMA_LITERAL +
354        (LZMA_LIT_SIZE
355         * (((get_pos(wr) & literal_pos_mask) << lc)
356            + (wr->previous_byte >> (8 - lc))))
357        );
358
359    if (cst->state >= LZMA_NUM_LIT_STATES) {
360        int match_byte = peek_old_byte(wr, cst->rep0);
361        do {
362            int bit;
363            uint16_t *prob_lit;
364
365            match_byte <<= 1;
366            bit = match_byte & 0x100;
367            prob_lit = prob + 0x100 + bit + mi;
368            if (rc_get_bit(rc, prob_lit, &mi)) {
369                if (!bit)
370                    break;
371            } else {
372                if (bit)
373                    break;
374            }
375        } while (mi < 0x100);
376    }
377    while (mi < 0x100) {
378        uint16_t *prob_lit = prob + mi;
379        rc_get_bit(rc, prob_lit, &mi);
380    }
381    write_byte(wr, mi);
382    if (cst->state < 4)
383        cst->state = 0;
384    else if (cst->state < 10)
385        cst->state -= 3;
386    else
387        cst->state -= 6;
388}
389
390static 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) {
393  int offset;
394    uint16_t *prob_len;
395    int num_bits;
396    int len;
397
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;
407    } else {
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
413                + (cst->state <<
414                   LZMA_NUM_POS_BITS_MAX) +
415                pos_state);
416            if (rc_is_bit_0(rc, prob)) {
417                rc_update_bit_0(rc, prob);
418
419                cst->state = cst->state < LZMA_NUM_LIT_STATES ?
420                    9 : 11;
421                copy_byte(wr, cst->rep0);
422                return;
423            } else {
424                rc_update_bit_1(rc, prob);
425            }
426        } else {
427            uint32_t distance;
428
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;
434            } else {
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;
440                } else {
441                    rc_update_bit_1(rc, prob);
442                    distance = cst->rep3;
443                    cst->rep3 = cst->rep2;
444                }
445                cst->rep2 = cst->rep1;
446            }
447            cst->rep1 = cst->rep0;
448            cst->rep0 = distance;
449        }
450        cst->state = cst->state < LZMA_NUM_LIT_STATES ? 8 : 11;
451        prob = p + LZMA_REP_LEN_CODER;
452    }
453
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
458                + (pos_state <<
459                   LZMA_LEN_NUM_LOW_BITS));
460        offset = 0;
461        num_bits = LZMA_LEN_NUM_LOW_BITS;
462    } else {
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
468                    + (pos_state <<
469                       LZMA_LEN_NUM_MID_BITS));
470            offset = 1 << LZMA_LEN_NUM_LOW_BITS;
471            num_bits = LZMA_LEN_NUM_MID_BITS;
472        } else {
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;
478        }
479    }
480
481    rc_bit_tree_decode(rc, prob_len, num_bits, &len);
482    len += offset;
483
484    if (cst->state < 4) {
485        int pos_slot;
486
487        cst->state += LZMA_NUM_LIT_STATES;
488        prob =
489            p + LZMA_POS_SLOT +
490            ((len <
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,
496                   &pos_slot);
497        if (pos_slot >= LZMA_START_POS_MODEL_INDEX) {
498            int i, mi;
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;
505            } else {
506                num_bits -= LZMA_NUM_ALIGN_BITS;
507                while (num_bits--)
508                    cst->rep0 = (cst->rep0 << 1) |
509                        rc_direct_bit(rc);
510                prob = p + LZMA_ALIGN;
511                cst->rep0 <<= LZMA_NUM_ALIGN_BITS;
512                num_bits = LZMA_NUM_ALIGN_BITS;
513            }
514            i = 1;
515            mi = 1;
516            while (num_bits--) {
517                if (rc_get_bit(rc, prob + mi, &mi))
518                    cst->rep0 |= i;
519                i <<= 1;
520            }
521        } else
522            cst->rep0 = pos_slot;
523        if (++(cst->rep0) == 0)
524            return;
525    }
526
527    len += LZMA_MATCH_MIN_LEN;
528
529    copy_bytes(wr, cst->rep0, len);
530}
531
532
533
534STATIC 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,
538                  int *posp,
539                  void(*error_fn)(char *x)
540    )
541{
542    struct lzma_header header;
543    int lc, pb, lp;
544    uint32_t pos_state_mask;
545    uint32_t literal_pos_mask;
546    uint16_t *p;
547    int num_probs;
548    struct rc rc;
549    int i, mi;
550    struct writer wr;
551    struct cstate cst;
552    unsigned char *inbuf;
553    int ret = -1;
554
555    set_error_fn(error_fn);
556
557    if (buf)
558        inbuf = buf;
559    else
560        inbuf = malloc(LZMA_IOBUF_SIZE);
561    if (!inbuf) {
562        error("Could not allocate input bufer");
563        goto exit_0;
564    }
565
566    cst.state = 0;
567    cst.rep0 = cst.rep1 = cst.rep2 = cst.rep3 = 1;
568
569    wr.header = &header;
570    wr.flush = flush;
571    wr.global_pos = 0;
572    wr.previous_byte = 0;
573    wr.buffer_pos = 0;
574
575    rc_init(&rc, fill, inbuf, in_len);
576
577    for (i = 0; i < sizeof(header); i++) {
578        if (rc.ptr >= rc.buffer_end)
579            rc_read(&rc);
580        ((unsigned char *)&header)[i] = *rc.ptr++;
581    }
582
583    if (header.pos >= (9 * 5 * 5))
584        error("bad header");
585
586    mi = 0;
587    lc = header.pos;
588    while (lc >= 9) {
589        mi++;
590        lc -= 9;
591    }
592    pb = 0;
593    lp = mi;
594    while (lp >= 5) {
595        pb++;
596        lp -= 5;
597    }
598    pos_state_mask = (1 << pb) - 1;
599    literal_pos_mask = (1 << lp) - 1;
600
601    ENDIAN_CONVERT(header.dict_size);
602    ENDIAN_CONVERT(header.dst_size);
603
604    if (header.dict_size == 0)
605        header.dict_size = 1;
606
607    if (output)
608        wr.buffer = output;
609    else {
610        wr.bufsize = MIN(header.dst_size, header.dict_size);
611        wr.buffer = large_malloc(wr.bufsize);
612    }
613    if (wr.buffer == NULL)
614        goto exit_1;
615
616    num_probs = LZMA_BASE_SIZE + (LZMA_LIT_SIZE << (lc + lp));
617    p = (uint16_t *) large_malloc(num_probs * sizeof(*p));
618    if (p == 0)
619        goto exit_2;
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;
623
624    rc_init_code(&rc);
625
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);
633        else {
634            process_bit1(&wr, &rc, &cst, p, pos_state, prob);
635            if (cst.rep0 == 0)
636                break;
637        }
638    }
639
640    if (posp)
641        *posp = rc.ptr-rc.buffer;
642    if (wr.flush)
643        wr.flush(wr.buffer, wr.buffer_pos);
644    ret = 0;
645    large_free(p);
646exit_2:
647    if (!output)
648        large_free(wr.buffer);
649exit_1:
650    if (!buf)
651        free(inbuf);
652exit_0:
653    return ret;
654}
655
656#ifdef PREBOOT
657STATIC 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,
661                  int *posp,
662                  void(*error_fn)(char *x)
663    )
664{
665    return unlzma(buf, in_len - 4, fill, flush, output, posp, error_fn);
666}
667#endif
668

Archive Download this file



interactive