2 * Copyright (c) 1989 The Regents of the University of California.
5 * This code is derived from software contributed to Berkeley by
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 #include <afsconfig.h>
38 #include <afs/param.h>
55 * UNIX password, and DES, encryption.
56 * By Tom Truscott, trt@rti.rti.org,
57 * from algorithms by Robert W. Baldwin and James Gillogly.
60 * "Mathematical Cryptology for Computer Scientists and Mathematicians,"
61 * by Wayne Patterson, 1987, ISBN 0-8476-7438-X.
63 * "Password Security: A Case History," R. Morris and Ken Thompson,
64 * Communications of the ACM, vol. 22, pp. 594-597, Nov. 1979.
66 * "DES will be Totally Insecure within Ten Years," M.E. Hellman,
67 * IEEE Spectrum, vol. 16, pp. 32-39, July 1979.
70 /* ===== Configuration ==================== */
73 * define "MUST_ALIGN" if your compiler cannot load/store
74 * long integers at arbitrary (e.g. odd) memory locations.
75 * (Either that or never pass unaligned addresses to des_cipher!)
83 #error C_block structure assumes 8 bit characters
88 * define "LONG_IS_32_BITS" only if sizeof(long)==4.
89 * This avoids use of bit fields (your compiler may be sloppy with them).
92 #define LONG_IS_32_BITS
96 * define "B64" to be the declaration for a 64 bit integer.
97 * XXX this feature is currently unused, see "endian" comment below.
103 #define B64 long long
107 * define "LARGEDATA" to get faster permutations, by using about 72 kilobytes
108 * of lookup tables. This speeds up des_setkey() and des_cipher(), but has
109 * little effect on crypt().
115 /* compile with "-DSTATIC=int" when profiling */
117 #define STATIC static
119 STATIC void init_des();
120 STATIC void permute();
121 STATIC void init_perm();
123 /* Hide these functions for Transarc use; only export crypt() */
124 STATIC int des_setkey(const char *key);
125 STATIC int des_cipher(const char *in, char *out, long salt, int num_iter);
131 /* ==================================== */
134 * Cipher-block representation (Bob Baldwin):
136 * DES operates on groups of 64 bits, numbered 1..64 (sigh). One
137 * representation is to store one bit per byte in an array of bytes. Bit N of
138 * the NBS spec is stored as the LSB of the Nth byte (index N-1) in the array.
139 * Another representation stores the 64 bits in 8 bytes, with bits 1..8 in the
140 * first byte, 9..16 in the second, and so on. The DES spec apparently has
141 * bit 1 in the MSB of the first byte, but that is particularly noxious so we
142 * bit-reverse each byte so that bit 1 is the LSB of the first byte, bit 8 is
143 * the MSB of the first byte. Specifically, the 64-bit input data and key are
144 * converted to LSB format, and the output 64-bit block is converted back into
147 * DES operates internally on groups of 32 bits which are expanded to 48 bits
148 * by permutation E and shrunk back to 32 bits by the S boxes. To speed up
149 * the computation, the expansion is applied only once, the expanded
150 * representation is maintained during the encryption, and a compression
151 * permutation is applied only at the end. To speed up the S-box lookups,
152 * the 48 bits are maintained as eight 6 bit groups, one per byte, which
153 * directly feed the eight S-boxes. Within each byte, the 6 bits are the
154 * most significant ones. The low two bits of each byte are zero. (Thus,
155 * bit 1 of the 48 bit E expansion is stored as the "4"-valued bit of the
156 * first byte in the eight byte representation, bit 2 of the 48 bit value is
157 * the "8"-valued bit, and so on.) In fact, a combined "SPE"-box lookup is
158 * used, in which the output is the 64 bit result of an S-box lookup which
159 * has been permuted by P and expanded by E, and is ready for use in the next
160 * iteration. Two 32-bit wide tables, SPE[0] and SPE[1], are used for this
161 * lookup. Since each byte in the 48 bit path is a multiple of four, indexed
162 * lookup of SPE[0] and SPE[1] is simple and fast. The key schedule and
163 * "salt" are also converted to this 8*(6+2) format. The SPE table size is
166 * To speed up bit-parallel operations (such as XOR), the 8 byte
167 * representation is "union"ed with 32 bit values "i0" and "i1", and, on
168 * machines which support it, a 64 bit value "b64". This data structure,
169 * "C_block", has two problems. First, alignment restrictions must be
170 * honored. Second, the byte-order (e.g. little-endian or big-endian) of
171 * the architecture becomes visible.
173 * The byte-order problem is unfortunate, since on the one hand it is good
174 * to have a machine-independent C_block representation (bits 1..8 in the
175 * first byte, etc.), and on the other hand it is good for the LSB of the
176 * first byte to be the LSB of i0. We cannot have both these things, so we
177 * currently use the "little-endian" representation and avoid any multi-byte
178 * operations that depend on byte order. This largely precludes use of the
179 * 64-bit datatype since the relative order of i0 and i1 are unknown. It
180 * also inhibits grouping the SPE table to look up 12 bits at a time. (The
181 * 12 bits can be stored in a 16-bit field with 3 low-order zeroes and 1
182 * high-order zero, providing fast indexing into a 64-bit wide SPE.) On the
183 * other hand, 64-bit datatypes are currently rare, and a 12-bit SPE lookup
184 * requires a 128 kilobyte table, so perhaps this is not a big loss.
186 * Permutation representation (Jim Gillogly):
188 * A transformation is defined by its effect on each of the 8 bytes of the
189 * 64-bit input. For each byte we give a 64-bit output that has the bits in
190 * the input distributed appropriately. The transformation is then the OR
191 * of the 8 sets of 64-bits. This uses 8*256*8 = 16K bytes of storage for
192 * each transformation. Unless LARGEDATA is defined, however, a more compact
193 * table is used which looks up 16 4-bit "chunks" rather than 8 8-bit chunks.
194 * The smaller table uses 16*16*8 = 2K bytes for each transformation. This
195 * is slower but tolerable, particularly for password encryption in which
196 * the SPE transformation is iterated many times. The small tables total 9K
197 * bytes, the large tables total 72K bytes.
199 * The transformations used are:
200 * IE3264: MSB->LSB conversion, initial permutation, and expansion.
201 * This is done by collecting the 32 even-numbered bits and applying
202 * a 32->64 bit transformation, and then collecting the 32 odd-numbered
203 * bits and applying the same transformation. Since there are only
204 * 32 input bits, the IE3264 transformation table is half the size of
206 * CF6464: Compression, final permutation, and LSB->MSB conversion.
207 * This is done by two trivial 48->32 bit compressions to obtain
208 * a 64-bit block (the bit numbering is given in the "CIFP" table)
209 * followed by a 64->64 bit "cleanup" transformation. (It would
210 * be possible to group the bits in the 64-bit block so that 2
211 * identical 32->32 bit transformations could be used instead,
212 * saving a factor of 4 in space and possibly 2 in time, but
213 * byte-ordering and other complications rear their ugly head.
214 * Similar opportunities/problems arise in the key schedule
216 * PC1ROT: MSB->LSB, PC1 permutation, rotate, and PC2 permutation.
217 * This admittedly baroque 64->64 bit transformation is used to
218 * produce the first code (in 8*(6+2) format) of the key schedule.
219 * PC2ROT[0]: Inverse PC2 permutation, rotate, and PC2 permutation.
220 * It would be possible to define 15 more transformations, each
221 * with a different rotation, to generate the entire key schedule.
222 * To save space, however, we instead permute each code into the
223 * next by using a transformation that "undoes" the PC2 permutation,
224 * rotates the code, and then applies PC2. Unfortunately, PC2
225 * transforms 56 bits into 48 bits, dropping 8 bits, so PC2 is not
226 * invertible. We get around that problem by using a modified PC2
227 * which retains the 8 otherwise-lost bits in the unused low-order
228 * bits of each byte. The low-order bits are cleared when the
229 * codes are stored into the key schedule.
230 * PC2ROT[1]: Same as PC2ROT[0], but with two rotations.
231 * This is faster than applying PC2ROT[0] twice,
233 * The Bell Labs "salt" (Bob Baldwin):
235 * The salting is a simple permutation applied to the 48-bit result of E.
236 * Specifically, if bit i (1 <= i <= 24) of the salt is set then bits i and
237 * i+24 of the result are swapped. The salt is thus a 24 bit number, with
238 * 16777216 possible values. (The original salt was 12 bits and could not
239 * swap bits 13..24 with 36..48.)
241 * It is possible, but ugly, to warp the SPE table to account for the salt
242 * permutation. Fortunately, the conditional bit swapping requires only
243 * about four machine instructions and can be done on-the-fly with about an
244 * 8% performance penalty.
250 #if defined(LONG_IS_32_BITS)
251 /* long is often faster than a 32-bit bit field */
252 #if defined(AFS_IA64_LINUX20_ENV)
270 * Convert twenty-four-bit long in host-order
271 * to six bits (and 2 low-order zeroes) per char little-endian format.
273 #define TO_SIX_BIT(rslt, src) { \
275 cvt.b[0] = (unsigned char) src; src >>= 6; \
276 cvt.b[1] = (unsigned char) src; src >>= 6; \
277 cvt.b[2] = (unsigned char) src; src >>= 6; \
278 cvt.b[3] = (unsigned char) src; \
279 rslt = (cvt.b32.i0 & 0x3f3f3f3fL) << 2; \
283 * These macros may someday permit efficient use of 64-bit integers.
285 #define ZERO(d,d0,d1) d0 = 0, d1 = 0
286 #define LOAD(d,d0,d1,bl) d0 = (bl).b32.i0, d1 = (bl).b32.i1
287 #define LOADREG(d,d0,d1,s,s0,s1) d0 = s0, d1 = s1
288 #define OR(d,d0,d1,bl) d0 |= (bl).b32.i0, d1 |= (bl).b32.i1
289 #define STORE(s,s0,s1,bl) (bl).b32.i0 = s0, (bl).b32.i1 = s1
290 #define DCL_BLOCK(d,d0,d1) long d0, d1
292 #if defined(LARGEDATA)
293 /* Waste memory like crazy. Also, do permutations in line */
294 #define LGCHUNKBITS 3
295 #define CHUNKBITS (1<<LGCHUNKBITS)
296 #define PERM6464(d,d0,d1,cpp,p) \
297 LOAD(d,d0,d1,(p)[(0<<CHUNKBITS)+(cpp)[0]]); \
298 OR (d,d0,d1,(p)[(1<<CHUNKBITS)+(cpp)[1]]); \
299 OR (d,d0,d1,(p)[(2<<CHUNKBITS)+(cpp)[2]]); \
300 OR (d,d0,d1,(p)[(3<<CHUNKBITS)+(cpp)[3]]); \
301 OR (d,d0,d1,(p)[(4<<CHUNKBITS)+(cpp)[4]]); \
302 OR (d,d0,d1,(p)[(5<<CHUNKBITS)+(cpp)[5]]); \
303 OR (d,d0,d1,(p)[(6<<CHUNKBITS)+(cpp)[6]]); \
304 OR (d,d0,d1,(p)[(7<<CHUNKBITS)+(cpp)[7]]);
305 #define PERM3264(d,d0,d1,cpp,p) \
306 LOAD(d,d0,d1,(p)[(0<<CHUNKBITS)+(cpp)[0]]); \
307 OR (d,d0,d1,(p)[(1<<CHUNKBITS)+(cpp)[1]]); \
308 OR (d,d0,d1,(p)[(2<<CHUNKBITS)+(cpp)[2]]); \
309 OR (d,d0,d1,(p)[(3<<CHUNKBITS)+(cpp)[3]]);
312 #define LGCHUNKBITS 2
313 #define CHUNKBITS (1<<LGCHUNKBITS)
314 #define PERM6464(d,d0,d1,cpp,p) \
315 { C_block tblk; permute(cpp,&tblk,p,8); LOAD (d,d0,d1,tblk); }
316 #define PERM3264(d,d0,d1,cpp,p) \
317 { C_block tblk; permute(cpp,&tblk,p,4); LOAD (d,d0,d1,tblk); }
320 void permute(cp, out, p, chars_in)
326 register DCL_BLOCK(D,D0,D1);
327 register C_block *tp;
333 tp = &p[t&0xf]; OR(D,D0,D1,*tp); p += (1<<CHUNKBITS);
334 tp = &p[t>>4]; OR(D,D0,D1,*tp); p += (1<<CHUNKBITS);
335 } while (--chars_in > 0);
338 #endif /* LARGEDATA */
341 /* ===== (mostly) Standard DES Tables ==================== */
343 static unsigned char IP[] = { /* initial permutation */
344 58, 50, 42, 34, 26, 18, 10, 2,
345 60, 52, 44, 36, 28, 20, 12, 4,
346 62, 54, 46, 38, 30, 22, 14, 6,
347 64, 56, 48, 40, 32, 24, 16, 8,
348 57, 49, 41, 33, 25, 17, 9, 1,
349 59, 51, 43, 35, 27, 19, 11, 3,
350 61, 53, 45, 37, 29, 21, 13, 5,
351 63, 55, 47, 39, 31, 23, 15, 7,
354 /* The final permutation is the inverse of IP - no table is necessary */
356 static unsigned char ExpandTr[] = { /* expansion operation */
359 8, 9, 10, 11, 12, 13,
360 12, 13, 14, 15, 16, 17,
361 16, 17, 18, 19, 20, 21,
362 20, 21, 22, 23, 24, 25,
363 24, 25, 26, 27, 28, 29,
364 28, 29, 30, 31, 32, 1,
367 static unsigned char PC1[] = { /* permuted choice table 1 */
368 57, 49, 41, 33, 25, 17, 9,
369 1, 58, 50, 42, 34, 26, 18,
370 10, 2, 59, 51, 43, 35, 27,
371 19, 11, 3, 60, 52, 44, 36,
373 63, 55, 47, 39, 31, 23, 15,
374 7, 62, 54, 46, 38, 30, 22,
375 14, 6, 61, 53, 45, 37, 29,
376 21, 13, 5, 28, 20, 12, 4,
379 static unsigned char Rotates[] = { /* PC1 rotation schedule */
380 1, 1, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 1,
383 /* note: each "row" of PC2 is left-padded with bits that make it invertible */
384 static unsigned char PC2[] = { /* permuted choice table 2 */
385 9, 18, 14, 17, 11, 24, 1, 5,
386 22, 25, 3, 28, 15, 6, 21, 10,
387 35, 38, 23, 19, 12, 4, 26, 8,
388 43, 54, 16, 7, 27, 20, 13, 2,
390 0, 0, 41, 52, 31, 37, 47, 55,
391 0, 0, 30, 40, 51, 45, 33, 48,
392 0, 0, 44, 49, 39, 56, 34, 53,
393 0, 0, 46, 42, 50, 36, 29, 32,
396 static unsigned char S[8][64] = { /* 48->32 bit substitution tables */
398 { 14, 4, 13, 1, 2, 15, 11, 8, 3, 10, 6, 12, 5, 9, 0, 7,
399 0, 15, 7, 4, 14, 2, 13, 1, 10, 6, 12, 11, 9, 5, 3, 8,
400 4, 1, 14, 8, 13, 6, 2, 11, 15, 12, 9, 7, 3, 10, 5, 0,
401 15, 12, 8, 2, 4, 9, 1, 7, 5, 11, 3, 14, 10, 0, 6, 13, },
403 { 15, 1, 8, 14, 6, 11, 3, 4, 9, 7, 2, 13, 12, 0, 5, 10,
404 3, 13, 4, 7, 15, 2, 8, 14, 12, 0, 1, 10, 6, 9, 11, 5,
405 0, 14, 7, 11, 10, 4, 13, 1, 5, 8, 12, 6, 9, 3, 2, 15,
406 13, 8, 10, 1, 3, 15, 4, 2, 11, 6, 7, 12, 0, 5, 14, 9, },
408 { 10, 0, 9, 14, 6, 3, 15, 5, 1, 13, 12, 7, 11, 4, 2, 8,
409 13, 7, 0, 9, 3, 4, 6, 10, 2, 8, 5, 14, 12, 11, 15, 1,
410 13, 6, 4, 9, 8, 15, 3, 0, 11, 1, 2, 12, 5, 10, 14, 7,
411 1, 10, 13, 0, 6, 9, 8, 7, 4, 15, 14, 3, 11, 5, 2, 12, },
413 { 7, 13, 14, 3, 0, 6, 9, 10, 1, 2, 8, 5, 11, 12, 4, 15,
414 13, 8, 11, 5, 6, 15, 0, 3, 4, 7, 2, 12, 1, 10, 14, 9,
415 10, 6, 9, 0, 12, 11, 7, 13, 15, 1, 3, 14, 5, 2, 8, 4,
416 3, 15, 0, 6, 10, 1, 13, 8, 9, 4, 5, 11, 12, 7, 2, 14, },
418 { 2, 12, 4, 1, 7, 10, 11, 6, 8, 5, 3, 15, 13, 0, 14, 9,
419 14, 11, 2, 12, 4, 7, 13, 1, 5, 0, 15, 10, 3, 9, 8, 6,
420 4, 2, 1, 11, 10, 13, 7, 8, 15, 9, 12, 5, 6, 3, 0, 14,
421 11, 8, 12, 7, 1, 14, 2, 13, 6, 15, 0, 9, 10, 4, 5, 3, },
423 { 12, 1, 10, 15, 9, 2, 6, 8, 0, 13, 3, 4, 14, 7, 5, 11,
424 10, 15, 4, 2, 7, 12, 9, 5, 6, 1, 13, 14, 0, 11, 3, 8,
425 9, 14, 15, 5, 2, 8, 12, 3, 7, 0, 4, 10, 1, 13, 11, 6,
426 4, 3, 2, 12, 9, 5, 15, 10, 11, 14, 1, 7, 6, 0, 8, 13, },
428 { 4, 11, 2, 14, 15, 0, 8, 13, 3, 12, 9, 7, 5, 10, 6, 1,
429 13, 0, 11, 7, 4, 9, 1, 10, 14, 3, 5, 12, 2, 15, 8, 6,
430 1, 4, 11, 13, 12, 3, 7, 14, 10, 15, 6, 8, 0, 5, 9, 2,
431 6, 11, 13, 8, 1, 4, 10, 7, 9, 5, 0, 15, 14, 2, 3, 12, },
433 { 13, 2, 8, 4, 6, 15, 11, 1, 10, 9, 3, 14, 5, 0, 12, 7,
434 1, 15, 13, 8, 10, 3, 7, 4, 12, 5, 6, 11, 0, 14, 9, 2,
435 7, 11, 4, 1, 9, 12, 14, 2, 0, 6, 10, 13, 15, 3, 5, 8,
436 2, 1, 14, 7, 4, 10, 8, 13, 15, 12, 9, 0, 3, 5, 6, 11, }
439 static unsigned char P32Tr[] = { /* 32-bit permutation function */
450 static unsigned char CIFP[] = { /* compressed/interleaved permutation */
451 1, 2, 3, 4, 17, 18, 19, 20,
452 5, 6, 7, 8, 21, 22, 23, 24,
453 9, 10, 11, 12, 25, 26, 27, 28,
454 13, 14, 15, 16, 29, 30, 31, 32,
456 33, 34, 35, 36, 49, 50, 51, 52,
457 37, 38, 39, 40, 53, 54, 55, 56,
458 41, 42, 43, 44, 57, 58, 59, 60,
459 45, 46, 47, 48, 61, 62, 63, 64,
462 static unsigned char itoa64[] = /* 0..63 => ascii-64 */
463 "./0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
466 /* ===== Tables that are initialized at run time ==================== */
469 static unsigned char a64toi[128]; /* ascii-64 => 0..63 */
471 /* Initial key schedule permutation */
472 static C_block PC1ROT[64/CHUNKBITS][1<<CHUNKBITS];
474 /* Subsequent key schedule rotation permutations */
475 static C_block PC2ROT[2][64/CHUNKBITS][1<<CHUNKBITS];
477 /* Initial permutation/expansion table */
478 static C_block IE3264[32/CHUNKBITS][1<<CHUNKBITS];
480 /* Table that combines the S, P, and E operations. */
481 static long SPE[2][8][64];
483 /* compressed/interleaved => final permutation table */
484 static C_block CF6464[64/CHUNKBITS][1<<CHUNKBITS];
487 /* ==================================== */
490 static C_block constdatablock; /* encryption constant */
491 static char cryptresult[1+4+4+11+1]; /* encrypted result */
494 * Return a pointer to static data consisting of the "setting"
495 * followed by an encryption produced by the "key" and "setting".
499 register const char *key;
500 register const char *setting;
506 int num_iter, salt_size;
507 C_block keyblock, rsltblock;
510 for (i = 0; i < 8; i++) {
511 if ((t = 2*(unsigned char)(*key)) != 0)
515 if (des_setkey((char *)keyblock.b)) /* also initializes "a64toi" */
518 encp = &cryptresult[0];
520 case '_': /* was EFMT1 */
522 * Involve the rest of the password 8 characters at a time.
525 if (des_cipher((char *)&keyblock,
526 (char *)&keyblock, 0L, 1))
528 for (i = 0; i < 8; i++) {
529 if ((t = 2*(unsigned char)(*key)) != 0)
533 if (des_setkey((char *)keyblock.b))
537 *encp++ = *setting++;
539 /* get iteration count */
541 for (i = 4; --i >= 0; ) {
542 if ((t = (unsigned char)setting[i]) == '\0')
545 num_iter = (num_iter<<6) | a64toi[t];
557 for (i = salt_size; --i >= 0; ) {
558 if ((t = (unsigned char)setting[i]) == '\0')
561 salt = (salt<<6) | a64toi[t];
564 if (des_cipher((char *)&constdatablock, (char *)&rsltblock,
569 * Encode the 64 cipher bits as 11 ascii characters.
571 i = ((long)((rsltblock.b[0]<<8) | rsltblock.b[1])<<8) | rsltblock.b[2];
572 encp[3] = itoa64[i&0x3f]; i >>= 6;
573 encp[2] = itoa64[i&0x3f]; i >>= 6;
574 encp[1] = itoa64[i&0x3f]; i >>= 6;
575 encp[0] = itoa64[i]; encp += 4;
576 i = ((long)((rsltblock.b[3]<<8) | rsltblock.b[4])<<8) | rsltblock.b[5];
577 encp[3] = itoa64[i&0x3f]; i >>= 6;
578 encp[2] = itoa64[i&0x3f]; i >>= 6;
579 encp[1] = itoa64[i&0x3f]; i >>= 6;
580 encp[0] = itoa64[i]; encp += 4;
581 i = ((long)((rsltblock.b[6])<<8) | rsltblock.b[7])<<2;
582 encp[2] = itoa64[i&0x3f]; i >>= 6;
583 encp[1] = itoa64[i&0x3f]; i >>= 6;
588 return (cryptresult);
593 * The Key Schedule, filled in by des_setkey() or setkey().
596 static C_block KS[KS_SIZE];
599 * Set up the key schedule from the key.
603 register const char *key;
605 register DCL_BLOCK(K, K0, K1);
606 register C_block *ptabp;
608 static int des_ready = 0;
615 PERM6464(K,K0,K1,(unsigned char *)key,(C_block *)PC1ROT);
616 key = (char *)&KS[0];
617 STORE(K&~0x03030303L, K0&~0x03030303L, K1, *(C_block *)key);
618 for (i = 1; i < 16; i++) {
619 key += sizeof(C_block);
620 STORE(K,K0,K1,*(C_block *)key);
621 ptabp = (C_block *)PC2ROT[Rotates[i]-1];
622 PERM6464(K,K0,K1,(unsigned char *)key,ptabp);
623 STORE(K&~0x03030303L, K0&~0x03030303L, K1, *(C_block *)key);
629 * Encrypt (or decrypt if num_iter < 0) the 8 chars at "in" with abs(num_iter)
630 * iterations of DES, using the the given 24-bit salt and the pre-computed key
631 * schedule, and store the resulting 8 chars at "out" (in == out is permitted).
633 * NOTE: the performance of this routine is critically dependent on your
634 * compiler and machine architecture.
637 int des_cipher(in, out, salt, num_iter)
643 /* variables that we want in registers, most important first */
647 register long L0, L1, R0, R1, k;
648 register C_block *kp;
649 register int ks_inc, loop_count;
653 TO_SIX_BIT(salt, L0); /* convert to 4*(6+2) format */
655 #if defined(vax) || defined(pdp11)
656 salt = ~salt; /* "x &~ y" is faster than "x & y". */
662 #if defined(MUST_ALIGN)
663 B.b[0] = in[0]; B.b[1] = in[1]; B.b[2] = in[2]; B.b[3] = in[3];
664 B.b[4] = in[4]; B.b[5] = in[5]; B.b[6] = in[6]; B.b[7] = in[7];
667 LOAD(L,L0,L1,*(C_block *)in);
669 LOADREG(R,R0,R1,L,L0,L1);
672 L0 = (L0 << 1) | L1; /* L0 is the even-numbered input bits */
674 R1 = (R1 >> 1) & 0x55555555L;
675 L1 = R0 | R1; /* L1 is the odd-numbered input bits */
677 PERM3264(L,L0,L1,B.b, (C_block *)IE3264); /* even bits */
678 PERM3264(R,R0,R1,B.b+4,(C_block *)IE3264); /* odd bits */
683 ks_inc = sizeof(*kp);
687 num_iter = -num_iter;
689 ks_inc = -((long) sizeof(*kp));
692 while (--num_iter >= 0) {
696 #define SPTAB(t, i) (*(long *)((unsigned char *)t + i*(sizeof(long)/4)))
698 /* use this if B.b[i] is evaluated just once ... */
699 #define DOXOR(x,y,i) x^=SPTAB(SPE[0][i],B.b[i]); y^=SPTAB(SPE[1][i],B.b[i]);
702 /* use this if your "long" int indexing is slow */
703 #define DOXOR(x,y,i) j=B.b[i]; x^=SPTAB(SPE[0][i],j); y^=SPTAB(SPE[1][i],j);
705 /* use this if "k" is allocated to a register ... */
706 #define DOXOR(x,y,i) k=B.b[i]; x^=SPTAB(SPE[0][i],k); y^=SPTAB(SPE[1][i],k);
710 #define CRUNCH(p0, p1, q0, q1) \
711 k = (q0 ^ q1) & SALT; \
712 B.b32.i0 = k ^ q0 ^ kp->b32.i0; \
713 B.b32.i1 = k ^ q1 ^ kp->b32.i1; \
714 kp = (C_block *)((char *)kp+ks_inc); \
725 CRUNCH(L0, L1, R0, R1);
726 CRUNCH(R0, R1, L0, L1);
727 } while (--loop_count != 0);
728 kp = (C_block *)((char *)kp-(ks_inc*KS_SIZE));
737 /* store the encrypted (or decrypted) result */
738 L0 = ((L0 >> 3) & 0x0f0f0f0fL) | ((L1 << 1) & 0xf0f0f0f0L);
739 L1 = ((R0 >> 3) & 0x0f0f0f0fL) | ((R1 << 1) & 0xf0f0f0f0L);
741 PERM6464(L,L0,L1,B.b, (C_block *)CF6464);
742 #if defined(MUST_ALIGN)
744 out[0] = B.b[0]; out[1] = B.b[1]; out[2] = B.b[2]; out[3] = B.b[3];
745 out[4] = B.b[4]; out[5] = B.b[5]; out[6] = B.b[6]; out[7] = B.b[7];
747 STORE(L,L0,L1,*(C_block *)out);
754 * Initialize various tables. This need only be done once. It could even be
755 * done at compile time, if the compiler were capable of that sort of thing.
762 register int tableno;
763 static unsigned char perm[64], tmp32[32]; /* "static" for speed */
766 * table that converts chars "./0-9A-Za-z"to integers 0-63.
768 for (i = 0; i < 64; i++)
769 a64toi[itoa64[i]] = i;
772 * PC1ROT - bit reverse, then PC1, then Rotate, then PC2.
774 for (i = 0; i < 64; i++)
776 for (i = 0; i < 64; i++) {
777 if ((k = PC2[i]) == 0)
780 if ((k%28) < Rotates[0]) k -= 28;
787 perm[i] = (unsigned char) k;
790 prtab("pc1tab", perm, 8);
792 init_perm(PC1ROT, perm, 8, 8);
795 * PC2ROT - PC2 inverse, then Rotate (once or twice), then PC2.
797 for (j = 0; j < 2; j++) {
798 unsigned char pc2inv[64];
799 for (i = 0; i < 64; i++)
800 perm[i] = pc2inv[i] = 0;
801 for (i = 0; i < 64; i++) {
802 if ((k = PC2[i]) == 0)
806 for (i = 0; i < 64; i++) {
807 if ((k = PC2[i]) == 0)
810 if ((k%28) <= j) k -= 28;
814 prtab("pc2tab", perm, 8);
816 init_perm(PC2ROT[j], perm, 8, 8);
820 * Bit reverse, then initial permutation, then expansion.
822 for (i = 0; i < 8; i++) {
823 for (j = 0; j < 8; j++) {
824 k = (j < 2)? 0: IP[ExpandTr[i*6+j-2]-1];
834 perm[i*8+j] = (unsigned char) k;
838 prtab("ietab", perm, 8);
840 init_perm(IE3264, perm, 4, 8);
843 * Compression, then final permutation, then bit reverse.
845 for (i = 0; i < 64; i++) {
855 prtab("cftab", perm, 8);
857 init_perm(CF6464, perm, 8, 8);
862 for (i = 0; i < 48; i++)
863 perm[i] = P32Tr[ExpandTr[i]-1];
864 for (tableno = 0; tableno < 8; tableno++) {
865 for (j = 0; j < 64; j++) {
866 k = (((j >> 0) &01) << 5)|
867 (((j >> 1) &01) << 3)|
868 (((j >> 2) &01) << 2)|
869 (((j >> 3) &01) << 1)|
870 (((j >> 4) &01) << 0)|
871 (((j >> 5) &01) << 4);
873 k = (((k >> 3)&01) << 0)|
874 (((k >> 2)&01) << 1)|
875 (((k >> 1)&01) << 2)|
876 (((k >> 0)&01) << 3);
877 for (i = 0; i < 32; i++)
879 for (i = 0; i < 4; i++)
880 tmp32[4 * tableno + i] = (k >> i) & 01;
882 for (i = 24; --i >= 0; )
883 k = (k<<1) | tmp32[perm[i]-1];
884 TO_SIX_BIT(SPE[0][tableno][j], k);
886 for (i = 24; --i >= 0; )
887 k = (k<<1) | tmp32[perm[i+24]-1];
888 TO_SIX_BIT(SPE[1][tableno][j], k);
894 * Initialize "perm" to represent transformation "p", which rearranges
895 * (perhaps with expansion and/or contraction) one packed array of bits
896 * (of size "chars_in" characters) into another array (of size "chars_out"
899 * "perm" must be all-zeroes on entry to this routine.
902 void init_perm(perm, p, chars_in, chars_out)
903 C_block perm[64/CHUNKBITS][1<<CHUNKBITS];
905 int chars_in, chars_out;
907 register int i, j, k, l;
909 for (k = 0; k < chars_out*8; k++) { /* each output bit position */
910 l = p[k] - 1; /* where this bit comes from */
912 continue; /* output bit is always 0 */
913 i = l>>LGCHUNKBITS; /* which chunk this bit comes from */
914 l = 1<<(l&(CHUNKBITS-1)); /* mask for this bit */
915 for (j = 0; j < (1<<CHUNKBITS); j++) { /* each chunk value */
917 perm[i][j].b[k>>3] |= 1<<(k&07);
923 * "setkey" routine (for backwards compatibility)
925 #if 0 /* static and doesn't appear to be referenced */
928 register const char *key;
930 register int i, j, k;
933 for (i = 0; i < 8; i++) {
935 for (j = 0; j < 8; j++) {
937 k |= (unsigned char)*key++;
941 return (des_setkey((char *)keyblock.b));
946 * "encrypt" routine (for backwards compatibility)
948 int encrypt(block, flag)
949 register char *block;
952 register int i, j, k;
955 for (i = 0; i < 8; i++) {
957 for (j = 0; j < 8; j++) {
959 k |= (unsigned char)*block++;
963 if (des_cipher((char *)&cblock, (char *)&cblock, 0L, (flag ? -1: 1)))
965 for (i = 7; i >= 0; i--) {
967 for (j = 7; j >= 0; j--) {
977 prtab(s, t, num_rows)
984 (void)printf("%s:\n", s);
985 for (i = 0; i < num_rows; i++) {
986 for (j = 0; j < 8; j++) {
987 (void)printf("%3d", t[i*8+j]);