1/* crc32.c -- compute the CRC-32 of a data stream
2 * Copyright (C) 1995-2006, 2010 Mark Adler
3 * For conditions of distribution and use, see copyright notice in zlib.h
4 *
5 * Thanks to Rodney Brown <rbrown64@csc.com.au> for his contribution of faster
6 * CRC methods: exclusive-oring 32 bits of data at a time, and pre-computing
7 * tables for updating the shift register in one step with three exclusive-ors
8 * instead of four steps with four exclusive-ors.  This results in about a
9 * factor of two increase in speed on a Power PC G4 (PPC7455) using gcc -O3.
10 */
11
12/* @(#) $Id$ */
13
14/*
15  Note on the use of DYNAMIC_CRC_TABLE: there is no mutex or semaphore
16  protection on the static variables used to control the first-use generation
17  of the crc tables.  Therefore, if you #define DYNAMIC_CRC_TABLE, you should
18  first call get_crc_table() to initialize the tables before allowing more than
19  one thread to use crc32().
20 */
21
22
23#ifdef MAKECRCH
24#  include <stdio.h>
25#  ifndef DYNAMIC_CRC_TABLE
26#    define DYNAMIC_CRC_TABLE
27#  endif /* !DYNAMIC_CRC_TABLE */
28#endif /* MAKECRCH */
29
30#include "zutil.h"      /* for STDC and FAR definitions */
31
32#define local static
33
34/* Find a four-byte integer type for crc32_little() and crc32_big(). */
35#ifndef NOBYFOUR
36#  ifdef STDC           /* need ANSI C limits.h to determine sizes */
37#    include <limits.h>
38#    define BYFOUR
39#    if (UINT_MAX == 0xffffffffUL)
40       typedef unsigned int u4;
41#    else
42#      if (ULONG_MAX == 0xffffffffUL)
43         typedef unsigned long u4;
44#      else
45#        if (USHRT_MAX == 0xffffffffUL)
46           typedef unsigned short u4;
47#        else
48#          undef BYFOUR     /* can't find a four-byte integer type! */
49#        endif
50#      endif
51#    endif
52#  endif /* STDC */
53#endif /* !NOBYFOUR */
54
55/* Definitions for doing the crc four data bytes at a time. */
56#ifdef BYFOUR
57#  define REV(w) ((((w)>>24)&0xff)+(((w)>>8)&0xff00)+ \
58                (((w)&0xff00)<<8)+(((w)&0xff)<<24))
59   local unsigned long crc32_little OF((unsigned long,
60                        const unsigned char FAR *, unsigned));
61   local unsigned long crc32_big OF((unsigned long,
62                        const unsigned char FAR *, unsigned));
63#  define TBLS 8
64#else
65#  define TBLS 1
66#endif /* BYFOUR */
67
68/* Local functions for crc concatenation */
69local unsigned long gf2_matrix_times OF((unsigned long *mat,
70                                         unsigned long vec));
71local void gf2_matrix_square OF((unsigned long *square, unsigned long *mat));
72local uLong crc32_combine_(uLong crc1, uLong crc2, z_off64_t len2);
73
74
75#ifdef DYNAMIC_CRC_TABLE
76
77local volatile int crc_table_empty = 1;
78local unsigned long FAR crc_table[TBLS][256];
79local void make_crc_table OF((void));
80#ifdef MAKECRCH
81   local void write_table OF((FILE *, const unsigned long FAR *));
82#endif /* MAKECRCH */
83/*
84  Generate tables for a byte-wise 32-bit CRC calculation on the polynomial:
85  x^32+x^26+x^23+x^22+x^16+x^12+x^11+x^10+x^8+x^7+x^5+x^4+x^2+x+1.
86
87  Polynomials over GF(2) are represented in binary, one bit per coefficient,
88  with the lowest powers in the most significant bit.  Then adding polynomials
89  is just exclusive-or, and multiplying a polynomial by x is a right shift by
90  one.  If we call the above polynomial p, and represent a byte as the
91  polynomial q, also with the lowest power in the most significant bit (so the
92  byte 0xb1 is the polynomial x^7+x^3+x+1), then the CRC is (q*x^32) mod p,
93  where a mod b means the remainder after dividing a by b.
94
95  This calculation is done using the shift-register method of multiplying and
96  taking the remainder.  The register is initialized to zero, and for each
97  incoming bit, x^32 is added mod p to the register if the bit is a one (where
98  x^32 mod p is p+x^32 = x^26+...+1), and the register is multiplied mod p by
99  x (which is shifting right by one and adding x^32 mod p if the bit shifted
100  out is a one).  We start with the highest power (least significant bit) of
101  q and repeat for all eight bits of q.
102
103  The first table is simply the CRC of all possible eight bit values.  This is
104  all the information needed to generate CRCs on data a byte at a time for all
105  combinations of CRC register values and incoming bytes.  The remaining tables
106  allow for word-at-a-time CRC calculation for both big-endian and little-
107  endian machines, where a word is four bytes.
108*/
109local void make_crc_table()
110{
111    unsigned long c;
112    int n, k;
113    unsigned long poly;                 /* polynomial exclusive-or pattern */
114    /* terms of polynomial defining this crc (except x^32): */
115    static volatile int first = 1;      /* flag to limit concurrent making */
116    static const unsigned char p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
117
118    /* See if another task is already doing this (not thread-safe, but better
119       than nothing -- significantly reduces duration of vulnerability in
120       case the advice about DYNAMIC_CRC_TABLE is ignored) */
121    if (first) {
122        first = 0;
123
124        /* make exclusive-or pattern from polynomial (0xedb88320UL) */
125        poly = 0UL;
126        for (n = 0; n < sizeof(p)/sizeof(unsigned char); n++)
127            poly |= 1UL << (31 - p[n]);
128
129        /* generate a crc for every 8-bit value */
130        for (n = 0; n < 256; n++) {
131            c = (unsigned long)n;
132            for (k = 0; k < 8; k++)
133                c = c & 1 ? poly ^ (c >> 1) : c >> 1;
134            crc_table[0][n] = c;
135        }
136
137#ifdef BYFOUR
138        /* generate crc for each value followed by one, two, and three zeros,
139           and then the byte reversal of those as well as the first table */
140        for (n = 0; n < 256; n++) {
141            c = crc_table[0][n];
142            crc_table[4][n] = REV(c);
143            for (k = 1; k < 4; k++) {
144                c = crc_table[0][c & 0xff] ^ (c >> 8);
145                crc_table[k][n] = c;
146                crc_table[k + 4][n] = REV(c);
147            }
148        }
149#endif /* BYFOUR */
150
151        crc_table_empty = 0;
152    }
153    else {      /* not first */
154        /* wait for the other guy to finish (not efficient, but rare) */
155        while (crc_table_empty)
156            ;
157    }
158
159#ifdef MAKECRCH
160    /* write out CRC tables to crc32.h */
161    {
162        FILE *out;
163
164        out = fopen("crc32.h", "w");
165        if (out == NULL) return;
166        fprintf(out, "/* crc32.h -- tables for rapid CRC calculation\n");
167        fprintf(out, " * Generated automatically by crc32.c\n */\n\n");
168        fprintf(out, "local const unsigned long FAR ");
169        fprintf(out, "crc_table[TBLS][256] =\n{\n  {\n");
170        write_table(out, crc_table[0]);
171#  ifdef BYFOUR
172        fprintf(out, "#ifdef BYFOUR\n");
173        for (k = 1; k < 8; k++) {
174            fprintf(out, "  },\n  {\n");
175            write_table(out, crc_table[k]);
176        }
177        fprintf(out, "#endif\n");
178#  endif /* BYFOUR */
179        fprintf(out, "  }\n};\n");
180        fclose(out);
181    }
182#endif /* MAKECRCH */
183}
184
185#ifdef MAKECRCH
186local void write_table(out, table)
187    FILE *out;
188    const unsigned long FAR *table;
189{
190    int n;
191
192    for (n = 0; n < 256; n++)
193        fprintf(out, "%s0x%08lxUL%s", n % 5 ? "" : "    ", table[n],
194                n == 255 ? "\n" : (n % 5 == 4 ? ",\n" : ", "));
195}
196#endif /* MAKECRCH */
197
198#else /* !DYNAMIC_CRC_TABLE */
199/* ========================================================================
200 * Tables of CRC-32s of all single-byte values, made by make_crc_table().
201 */
202#include "crc32.h"
203#endif /* DYNAMIC_CRC_TABLE */
204
205/* =========================================================================
206 * This function can be used by asm versions of crc32()
207 */
208const unsigned long FAR * ZEXPORT get_crc_table()
209{
210#ifdef DYNAMIC_CRC_TABLE
211    if (crc_table_empty)
212        make_crc_table();
213#endif /* DYNAMIC_CRC_TABLE */
214    return (const unsigned long FAR *)crc_table;
215}
216
217/* ========================================================================= */
218#define DO1 crc = crc_table[0][((int)crc ^ (*buf++)) & 0xff] ^ (crc >> 8)
219#define DO8 DO1; DO1; DO1; DO1; DO1; DO1; DO1; DO1
220
221/* ========================================================================= */
222unsigned long ZEXPORT crc32(crc, buf, len)
223    unsigned long crc;
224    const unsigned char FAR *buf;
225    uInt len;
226{
227    if (buf == Z_NULL) return 0UL;
228
229#ifdef DYNAMIC_CRC_TABLE
230    if (crc_table_empty)
231        make_crc_table();
232#endif /* DYNAMIC_CRC_TABLE */
233
234#ifdef BYFOUR
235    if (sizeof(void *) == sizeof(ptrdiff_t)) {
236        u4 endian;
237
238        endian = 1;
239        if (*((unsigned char *)(&endian)))
240            return crc32_little(crc, buf, len);
241        else
242            return crc32_big(crc, buf, len);
243    }
244#endif /* BYFOUR */
245    crc = crc ^ 0xffffffffUL;
246    while (len >= 8) {
247        DO8;
248        len -= 8;
249    }
250    if (len) do {
251        DO1;
252    } while (--len);
253    return crc ^ 0xffffffffUL;
254}
255
256
257#ifdef BYFOUR
258
259/* ========================================================================= */
260#define DOLIT4 c ^= *buf4++; \
261        c = crc_table[3][c & 0xff] ^ crc_table[2][(c >> 8) & 0xff] ^ \
262            crc_table[1][(c >> 16) & 0xff] ^ crc_table[0][c >> 24]
263#define DOLIT32 DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4
264
265/* ========================================================================= */
266local unsigned long crc32_little(crc, buf, len)
267    unsigned long crc;
268    const unsigned char FAR *buf;
269    unsigned len;
270{
271    register u4 c;
272    register const u4 FAR *buf4;
273
274    c = (u4)crc;
275    c = ~c;
276    while (len && ((ptrdiff_t)buf & 3)) {
277        c = crc_table[0][(c ^ *buf++) & 0xff] ^ (c >> 8);
278        len--;
279    }
280
281    buf4 = (const u4 FAR *)(const void FAR *)buf;
282    while (len >= 32) {
283        DOLIT32;
284        len -= 32;
285    }
286    while (len >= 4) {
287        DOLIT4;
288        len -= 4;
289    }
290    buf = (const unsigned char FAR *)buf4;
291
292    if (len) do {
293        c = crc_table[0][(c ^ *buf++) & 0xff] ^ (c >> 8);
294    } while (--len);
295    c = ~c;
296    return (unsigned long)c;
297}
298
299/* ========================================================================= */
300#define DOBIG4 c ^= *++buf4; \
301        c = crc_table[4][c & 0xff] ^ crc_table[5][(c >> 8) & 0xff] ^ \
302            crc_table[6][(c >> 16) & 0xff] ^ crc_table[7][c >> 24]
303#define DOBIG32 DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4
304
305/* ========================================================================= */
306local unsigned long crc32_big(crc, buf, len)
307    unsigned long crc;
308    const unsigned char FAR *buf;
309    unsigned len;
310{
311    register u4 c;
312    register const u4 FAR *buf4;
313
314    c = REV((u4)crc);
315    c = ~c;
316    while (len && ((ptrdiff_t)buf & 3)) {
317        c = crc_table[4][(c >> 24) ^ *buf++] ^ (c << 8);
318        len--;
319    }
320
321    buf4 = (const u4 FAR *)(const void FAR *)buf;
322    buf4--;
323    while (len >= 32) {
324        DOBIG32;
325        len -= 32;
326    }
327    while (len >= 4) {
328        DOBIG4;
329        len -= 4;
330    }
331    buf4++;
332    buf = (const unsigned char FAR *)buf4;
333
334    if (len) do {
335        c = crc_table[4][(c >> 24) ^ *buf++] ^ (c << 8);
336    } while (--len);
337    c = ~c;
338    return (unsigned long)(REV(c));
339}
340
341#endif /* BYFOUR */
342
343#define GF2_DIM 32      /* dimension of GF(2) vectors (length of CRC) */
344
345/* ========================================================================= */
346local unsigned long gf2_matrix_times(mat, vec)
347    unsigned long *mat;
348    unsigned long vec;
349{
350    unsigned long sum;
351
352    sum = 0;
353    while (vec) {
354        if (vec & 1)
355            sum ^= *mat;
356        vec >>= 1;
357        mat++;
358    }
359    return sum;
360}
361
362/* ========================================================================= */
363local void gf2_matrix_square(square, mat)
364    unsigned long *square;
365    unsigned long *mat;
366{
367    int n;
368
369    for (n = 0; n < GF2_DIM; n++)
370        square[n] = gf2_matrix_times(mat, mat[n]);
371}
372
373/* ========================================================================= */
374local uLong crc32_combine_(crc1, crc2, len2)
375    uLong crc1;
376    uLong crc2;
377    z_off64_t len2;
378{
379    int n;
380    unsigned long row;
381    unsigned long even[GF2_DIM];    /* even-power-of-two zeros operator */
382    unsigned long odd[GF2_DIM];     /* odd-power-of-two zeros operator */
383
384    /* degenerate case (also disallow negative lengths) */
385    if (len2 <= 0)
386        return crc1;
387
388    /* put operator for one zero bit in odd */
389    odd[0] = 0xedb88320UL;          /* CRC-32 polynomial */
390    row = 1;
391    for (n = 1; n < GF2_DIM; n++) {
392        odd[n] = row;
393        row <<= 1;
394    }
395
396    /* put operator for two zero bits in even */
397    gf2_matrix_square(even, odd);
398
399    /* put operator for four zero bits in odd */
400    gf2_matrix_square(odd, even);
401
402    /* apply len2 zeros to crc1 (first square will put the operator for one
403       zero byte, eight zero bits, in even) */
404    do {
405        /* apply zeros operator for this bit of len2 */
406        gf2_matrix_square(even, odd);
407        if (len2 & 1)
408            crc1 = gf2_matrix_times(even, crc1);
409        len2 >>= 1;
410
411        /* if no more bits set, then done */
412        if (len2 == 0)
413            break;
414
415        /* another iteration of the loop with odd and even swapped */
416        gf2_matrix_square(odd, even);
417        if (len2 & 1)
418            crc1 = gf2_matrix_times(odd, crc1);
419        len2 >>= 1;
420
421        /* if no more bits set, then done */
422    } while (len2 != 0);
423
424    /* return combined crc */
425    crc1 ^= crc2;
426    return crc1;
427}
428
429/* ========================================================================= */
430uLong ZEXPORT crc32_combine(crc1, crc2, len2)
431    uLong crc1;
432    uLong crc2;
433    z_off_t len2;
434{
435    return crc32_combine_(crc1, crc2, len2);
436}
437
438uLong ZEXPORT crc32_combine64(crc1, crc2, len2)
439    uLong crc1;
440    uLong crc2;
441    z_off64_t len2;
442{
443    return crc32_combine_(crc1, crc2, len2);
444}
445