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>
54 * UNIX password, and DES, encryption.
55 * By Tom Truscott, trt@rti.rti.org,
56 * from algorithms by Robert W. Baldwin and James Gillogly.
59 * "Mathematical Cryptology for Computer Scientists and Mathematicians,"
60 * by Wayne Patterson, 1987, ISBN 0-8476-7438-X.
62 * "Password Security: A Case History," R. Morris and Ken Thompson,
63 * Communications of the ACM, vol. 22, pp. 594-597, Nov. 1979.
65 * "DES will be Totally Insecure within Ten Years," M.E. Hellman,
66 * IEEE Spectrum, vol. 16, pp. 32-39, July 1979.
69 /* ===== Configuration ==================== */
72 * define "MUST_ALIGN" if your compiler cannot load/store
73 * long integers at arbitrary (e.g. odd) memory locations.
74 * (Either that or never pass unaligned addresses to des_cipher!)
82 #error C_block structure assumes 8 bit characters
87 * define "LONG_IS_32_BITS" only if sizeof(long)==4.
88 * This avoids use of bit fields (your compiler may be sloppy with them).
91 /* XXX shouldn't this be !AFS_64BIT_ENV ? */
92 #if !defined(cray) && !defined(AFS_ALPHA_LINUX20_ENV) && !defined(AFS_IA64_LINUX20_ENV) && !defined(AFS_AMD64_LINUX20_ENV) && !defined(AFS_PPC64_LINUX20_ENV) & !defined(AFS_S390X_LINUX20_ENV)
93 #define LONG_IS_32_BITS
97 * define "B64" to be the declaration for a 64 bit integer.
98 * XXX this feature is currently unused, see "endian" comment below.
104 #define B64 long long
108 * define "LARGEDATA" to get faster permutations, by using about 72 kilobytes
109 * of lookup tables. This speeds up des_setkey() and des_cipher(), but has
110 * little effect on crypt().
116 /* compile with "-DSTATIC=int" when profiling */
118 #define STATIC static
124 /* ==================================== */
127 * Cipher-block representation (Bob Baldwin):
129 * DES operates on groups of 64 bits, numbered 1..64 (sigh). One
130 * representation is to store one bit per byte in an array of bytes. Bit N of
131 * the NBS spec is stored as the LSB of the Nth byte (index N-1) in the array.
132 * Another representation stores the 64 bits in 8 bytes, with bits 1..8 in the
133 * first byte, 9..16 in the second, and so on. The DES spec apparently has
134 * bit 1 in the MSB of the first byte, but that is particularly noxious so we
135 * bit-reverse each byte so that bit 1 is the LSB of the first byte, bit 8 is
136 * the MSB of the first byte. Specifically, the 64-bit input data and key are
137 * converted to LSB format, and the output 64-bit block is converted back into
140 * DES operates internally on groups of 32 bits which are expanded to 48 bits
141 * by permutation E and shrunk back to 32 bits by the S boxes. To speed up
142 * the computation, the expansion is applied only once, the expanded
143 * representation is maintained during the encryption, and a compression
144 * permutation is applied only at the end. To speed up the S-box lookups,
145 * the 48 bits are maintained as eight 6 bit groups, one per byte, which
146 * directly feed the eight S-boxes. Within each byte, the 6 bits are the
147 * most significant ones. The low two bits of each byte are zero. (Thus,
148 * bit 1 of the 48 bit E expansion is stored as the "4"-valued bit of the
149 * first byte in the eight byte representation, bit 2 of the 48 bit value is
150 * the "8"-valued bit, and so on.) In fact, a combined "SPE"-box lookup is
151 * used, in which the output is the 64 bit result of an S-box lookup which
152 * has been permuted by P and expanded by E, and is ready for use in the next
153 * iteration. Two 32-bit wide tables, SPE[0] and SPE[1], are used for this
154 * lookup. Since each byte in the 48 bit path is a multiple of four, indexed
155 * lookup of SPE[0] and SPE[1] is simple and fast. The key schedule and
156 * "salt" are also converted to this 8*(6+2) format. The SPE table size is
159 * To speed up bit-parallel operations (such as XOR), the 8 byte
160 * representation is "union"ed with 32 bit values "i0" and "i1", and, on
161 * machines which support it, a 64 bit value "b64". This data structure,
162 * "C_block", has two problems. First, alignment restrictions must be
163 * honored. Second, the byte-order (e.g. little-endian or big-endian) of
164 * the architecture becomes visible.
166 * The byte-order problem is unfortunate, since on the one hand it is good
167 * to have a machine-independent C_block representation (bits 1..8 in the
168 * first byte, etc.), and on the other hand it is good for the LSB of the
169 * first byte to be the LSB of i0. We cannot have both these things, so we
170 * currently use the "little-endian" representation and avoid any multi-byte
171 * operations that depend on byte order. This largely precludes use of the
172 * 64-bit datatype since the relative order of i0 and i1 are unknown. It
173 * also inhibits grouping the SPE table to look up 12 bits at a time. (The
174 * 12 bits can be stored in a 16-bit field with 3 low-order zeroes and 1
175 * high-order zero, providing fast indexing into a 64-bit wide SPE.) On the
176 * other hand, 64-bit datatypes are currently rare, and a 12-bit SPE lookup
177 * requires a 128 kilobyte table, so perhaps this is not a big loss.
179 * Permutation representation (Jim Gillogly):
181 * A transformation is defined by its effect on each of the 8 bytes of the
182 * 64-bit input. For each byte we give a 64-bit output that has the bits in
183 * the input distributed appropriately. The transformation is then the OR
184 * of the 8 sets of 64-bits. This uses 8*256*8 = 16K bytes of storage for
185 * each transformation. Unless LARGEDATA is defined, however, a more compact
186 * table is used which looks up 16 4-bit "chunks" rather than 8 8-bit chunks.
187 * The smaller table uses 16*16*8 = 2K bytes for each transformation. This
188 * is slower but tolerable, particularly for password encryption in which
189 * the SPE transformation is iterated many times. The small tables total 9K
190 * bytes, the large tables total 72K bytes.
192 * The transformations used are:
193 * IE3264: MSB->LSB conversion, initial permutation, and expansion.
194 * This is done by collecting the 32 even-numbered bits and applying
195 * a 32->64 bit transformation, and then collecting the 32 odd-numbered
196 * bits and applying the same transformation. Since there are only
197 * 32 input bits, the IE3264 transformation table is half the size of
199 * CF6464: Compression, final permutation, and LSB->MSB conversion.
200 * This is done by two trivial 48->32 bit compressions to obtain
201 * a 64-bit block (the bit numbering is given in the "CIFP" table)
202 * followed by a 64->64 bit "cleanup" transformation. (It would
203 * be possible to group the bits in the 64-bit block so that 2
204 * identical 32->32 bit transformations could be used instead,
205 * saving a factor of 4 in space and possibly 2 in time, but
206 * byte-ordering and other complications rear their ugly head.
207 * Similar opportunities/problems arise in the key schedule
209 * PC1ROT: MSB->LSB, PC1 permutation, rotate, and PC2 permutation.
210 * This admittedly baroque 64->64 bit transformation is used to
211 * produce the first code (in 8*(6+2) format) of the key schedule.
212 * PC2ROT[0]: Inverse PC2 permutation, rotate, and PC2 permutation.
213 * It would be possible to define 15 more transformations, each
214 * with a different rotation, to generate the entire key schedule.
215 * To save space, however, we instead permute each code into the
216 * next by using a transformation that "undoes" the PC2 permutation,
217 * rotates the code, and then applies PC2. Unfortunately, PC2
218 * transforms 56 bits into 48 bits, dropping 8 bits, so PC2 is not
219 * invertible. We get around that problem by using a modified PC2
220 * which retains the 8 otherwise-lost bits in the unused low-order
221 * bits of each byte. The low-order bits are cleared when the
222 * codes are stored into the key schedule.
223 * PC2ROT[1]: Same as PC2ROT[0], but with two rotations.
224 * This is faster than applying PC2ROT[0] twice,
226 * The Bell Labs "salt" (Bob Baldwin):
228 * The salting is a simple permutation applied to the 48-bit result of E.
229 * Specifically, if bit i (1 <= i <= 24) of the salt is set then bits i and
230 * i+24 of the result are swapped. The salt is thus a 24 bit number, with
231 * 16777216 possible values. (The original salt was 12 bits and could not
232 * swap bits 13..24 with 36..48.)
234 * It is possible, but ugly, to warp the SPE table to account for the salt
235 * permutation. Fortunately, the conditional bit swapping requires only
236 * about four machine instructions and can be done on-the-fly with about an
237 * 8% performance penalty.
243 #if defined(LONG_IS_32_BITS)
244 /* long is often faster than a 32-bit bit field */
258 * Convert twenty-four-bit long in host-order
259 * to six bits (and 2 low-order zeroes) per char little-endian format.
261 #define TO_SIX_BIT(rslt, src) { \
263 cvt.b[0] = (unsigned char) src; src >>= 6; \
264 cvt.b[1] = (unsigned char) src; src >>= 6; \
265 cvt.b[2] = (unsigned char) src; src >>= 6; \
266 cvt.b[3] = (unsigned char) src; \
267 rslt = (cvt.b32.i0 & 0x3f3f3f3fL) << 2; \
271 * These macros may someday permit efficient use of 64-bit integers.
273 #define ZERO(d,d0,d1) d0 = 0, d1 = 0
274 #define LOAD(d,d0,d1,bl) d0 = (bl).b32.i0, d1 = (bl).b32.i1
275 #define LOADREG(d,d0,d1,s,s0,s1) d0 = s0, d1 = s1
276 #define OR(d,d0,d1,bl) d0 |= (bl).b32.i0, d1 |= (bl).b32.i1
277 #define STORE(s,s0,s1,bl) (bl).b32.i0 = (s0), (bl).b32.i1 = (s1)
278 #define DCL_BLOCK(d,d0,d1) long d0, d1
280 #if defined(LARGEDATA)
281 /* Waste memory like crazy. Also, do permutations in line */
282 #define LGCHUNKBITS 3
283 #define CHUNKBITS (1<<LGCHUNKBITS)
284 #define PERM6464(d,d0,d1,cpp,p) \
285 LOAD(d,d0,d1,(p)[(0<<CHUNKBITS)+(cpp)[0]]); \
286 OR (d,d0,d1,(p)[(1<<CHUNKBITS)+(cpp)[1]]); \
287 OR (d,d0,d1,(p)[(2<<CHUNKBITS)+(cpp)[2]]); \
288 OR (d,d0,d1,(p)[(3<<CHUNKBITS)+(cpp)[3]]); \
289 OR (d,d0,d1,(p)[(4<<CHUNKBITS)+(cpp)[4]]); \
290 OR (d,d0,d1,(p)[(5<<CHUNKBITS)+(cpp)[5]]); \
291 OR (d,d0,d1,(p)[(6<<CHUNKBITS)+(cpp)[6]]); \
292 OR (d,d0,d1,(p)[(7<<CHUNKBITS)+(cpp)[7]]);
293 #define PERM3264(d,d0,d1,cpp,p) \
294 LOAD(d,d0,d1,(p)[(0<<CHUNKBITS)+(cpp)[0]]); \
295 OR (d,d0,d1,(p)[(1<<CHUNKBITS)+(cpp)[1]]); \
296 OR (d,d0,d1,(p)[(2<<CHUNKBITS)+(cpp)[2]]); \
297 OR (d,d0,d1,(p)[(3<<CHUNKBITS)+(cpp)[3]]);
300 #define LGCHUNKBITS 2
301 #define CHUNKBITS (1<<LGCHUNKBITS)
302 #define PERM6464(d,d0,d1,cpp,p) \
303 { C_block tblk; permute(cpp,&tblk,p,8); LOAD (d,d0,d1,tblk); }
304 #define PERM3264(d,d0,d1,cpp,p) \
305 { C_block tblk; permute(cpp,&tblk,p,4); LOAD (d,d0,d1,tblk); }
308 permute(unsigned char *cp, C_block *out, register C_block *p, int chars_in)
310 register DCL_BLOCK(D, D0, D1);
311 register C_block *tp;
319 p += (1 << CHUNKBITS);
322 p += (1 << CHUNKBITS);
323 } while (--chars_in > 0);
324 STORE(D, D0, D1, *out);
326 #endif /* LARGEDATA */
328 STATIC void init_des(void);
329 STATIC void init_perm(C_block [64 / CHUNKBITS][1 << CHUNKBITS],
330 unsigned char [64], int, int);
331 STATIC int des_setkey(const char *key);
332 STATIC int des_cipher(const char *in, char *out, long salt, int num_iter);
336 /* ===== (mostly) Standard DES Tables ==================== */
338 static unsigned char IP[] = { /* initial permutation */
339 58, 50, 42, 34, 26, 18, 10, 2,
340 60, 52, 44, 36, 28, 20, 12, 4,
341 62, 54, 46, 38, 30, 22, 14, 6,
342 64, 56, 48, 40, 32, 24, 16, 8,
343 57, 49, 41, 33, 25, 17, 9, 1,
344 59, 51, 43, 35, 27, 19, 11, 3,
345 61, 53, 45, 37, 29, 21, 13, 5,
346 63, 55, 47, 39, 31, 23, 15, 7,
349 /* The final permutation is the inverse of IP - no table is necessary */
351 static unsigned char ExpandTr[] = { /* expansion operation */
354 8, 9, 10, 11, 12, 13,
355 12, 13, 14, 15, 16, 17,
356 16, 17, 18, 19, 20, 21,
357 20, 21, 22, 23, 24, 25,
358 24, 25, 26, 27, 28, 29,
359 28, 29, 30, 31, 32, 1,
362 static unsigned char PC1[] = { /* permuted choice table 1 */
363 57, 49, 41, 33, 25, 17, 9,
364 1, 58, 50, 42, 34, 26, 18,
365 10, 2, 59, 51, 43, 35, 27,
366 19, 11, 3, 60, 52, 44, 36,
368 63, 55, 47, 39, 31, 23, 15,
369 7, 62, 54, 46, 38, 30, 22,
370 14, 6, 61, 53, 45, 37, 29,
371 21, 13, 5, 28, 20, 12, 4,
374 static unsigned char Rotates[] = { /* PC1 rotation schedule */
375 1, 1, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 1,
378 /* note: each "row" of PC2 is left-padded with bits that make it invertible */
379 static unsigned char PC2[] = { /* permuted choice table 2 */
380 9, 18, 14, 17, 11, 24, 1, 5,
381 22, 25, 3, 28, 15, 6, 21, 10,
382 35, 38, 23, 19, 12, 4, 26, 8,
383 43, 54, 16, 7, 27, 20, 13, 2,
385 0, 0, 41, 52, 31, 37, 47, 55,
386 0, 0, 30, 40, 51, 45, 33, 48,
387 0, 0, 44, 49, 39, 56, 34, 53,
388 0, 0, 46, 42, 50, 36, 29, 32,
391 static unsigned char S[8][64] = { /* 48->32 bit substitution tables */
393 {14, 4, 13, 1, 2, 15, 11, 8, 3, 10, 6, 12, 5, 9, 0, 7,
394 0, 15, 7, 4, 14, 2, 13, 1, 10, 6, 12, 11, 9, 5, 3, 8,
395 4, 1, 14, 8, 13, 6, 2, 11, 15, 12, 9, 7, 3, 10, 5, 0,
396 15, 12, 8, 2, 4, 9, 1, 7, 5, 11, 3, 14, 10, 0, 6, 13,},
398 {15, 1, 8, 14, 6, 11, 3, 4, 9, 7, 2, 13, 12, 0, 5, 10,
399 3, 13, 4, 7, 15, 2, 8, 14, 12, 0, 1, 10, 6, 9, 11, 5,
400 0, 14, 7, 11, 10, 4, 13, 1, 5, 8, 12, 6, 9, 3, 2, 15,
401 13, 8, 10, 1, 3, 15, 4, 2, 11, 6, 7, 12, 0, 5, 14, 9,},
403 {10, 0, 9, 14, 6, 3, 15, 5, 1, 13, 12, 7, 11, 4, 2, 8,
404 13, 7, 0, 9, 3, 4, 6, 10, 2, 8, 5, 14, 12, 11, 15, 1,
405 13, 6, 4, 9, 8, 15, 3, 0, 11, 1, 2, 12, 5, 10, 14, 7,
406 1, 10, 13, 0, 6, 9, 8, 7, 4, 15, 14, 3, 11, 5, 2, 12,},
408 {7, 13, 14, 3, 0, 6, 9, 10, 1, 2, 8, 5, 11, 12, 4, 15,
409 13, 8, 11, 5, 6, 15, 0, 3, 4, 7, 2, 12, 1, 10, 14, 9,
410 10, 6, 9, 0, 12, 11, 7, 13, 15, 1, 3, 14, 5, 2, 8, 4,
411 3, 15, 0, 6, 10, 1, 13, 8, 9, 4, 5, 11, 12, 7, 2, 14,},
413 {2, 12, 4, 1, 7, 10, 11, 6, 8, 5, 3, 15, 13, 0, 14, 9,
414 14, 11, 2, 12, 4, 7, 13, 1, 5, 0, 15, 10, 3, 9, 8, 6,
415 4, 2, 1, 11, 10, 13, 7, 8, 15, 9, 12, 5, 6, 3, 0, 14,
416 11, 8, 12, 7, 1, 14, 2, 13, 6, 15, 0, 9, 10, 4, 5, 3,},
418 {12, 1, 10, 15, 9, 2, 6, 8, 0, 13, 3, 4, 14, 7, 5, 11,
419 10, 15, 4, 2, 7, 12, 9, 5, 6, 1, 13, 14, 0, 11, 3, 8,
420 9, 14, 15, 5, 2, 8, 12, 3, 7, 0, 4, 10, 1, 13, 11, 6,
421 4, 3, 2, 12, 9, 5, 15, 10, 11, 14, 1, 7, 6, 0, 8, 13,},
423 {4, 11, 2, 14, 15, 0, 8, 13, 3, 12, 9, 7, 5, 10, 6, 1,
424 13, 0, 11, 7, 4, 9, 1, 10, 14, 3, 5, 12, 2, 15, 8, 6,
425 1, 4, 11, 13, 12, 3, 7, 14, 10, 15, 6, 8, 0, 5, 9, 2,
426 6, 11, 13, 8, 1, 4, 10, 7, 9, 5, 0, 15, 14, 2, 3, 12,},
428 {13, 2, 8, 4, 6, 15, 11, 1, 10, 9, 3, 14, 5, 0, 12, 7,
429 1, 15, 13, 8, 10, 3, 7, 4, 12, 5, 6, 11, 0, 14, 9, 2,
430 7, 11, 4, 1, 9, 12, 14, 2, 0, 6, 10, 13, 15, 3, 5, 8,
431 2, 1, 14, 7, 4, 10, 8, 13, 15, 12, 9, 0, 3, 5, 6, 11,}
434 static unsigned char P32Tr[] = { /* 32-bit permutation function */
445 static unsigned char CIFP[] = { /* compressed/interleaved permutation */
446 1, 2, 3, 4, 17, 18, 19, 20,
447 5, 6, 7, 8, 21, 22, 23, 24,
448 9, 10, 11, 12, 25, 26, 27, 28,
449 13, 14, 15, 16, 29, 30, 31, 32,
451 33, 34, 35, 36, 49, 50, 51, 52,
452 37, 38, 39, 40, 53, 54, 55, 56,
453 41, 42, 43, 44, 57, 58, 59, 60,
454 45, 46, 47, 48, 61, 62, 63, 64,
457 static unsigned char itoa64[] = /* 0..63 => ascii-64 */
458 "./0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
461 /* ===== Tables that are initialized at run time ==================== */
464 static unsigned char a64toi[128]; /* ascii-64 => 0..63 */
466 /* Initial key schedule permutation */
467 static C_block PC1ROT[64 / CHUNKBITS][1 << CHUNKBITS];
469 /* Subsequent key schedule rotation permutations */
470 static C_block PC2ROT[2][64 / CHUNKBITS][1 << CHUNKBITS];
472 /* Initial permutation/expansion table */
473 static C_block IE3264[32 / CHUNKBITS][1 << CHUNKBITS];
475 /* Table that combines the S, P, and E operations. */
476 static long SPE[2][8][64];
478 /* compressed/interleaved => final permutation table */
479 static C_block CF6464[64 / CHUNKBITS][1 << CHUNKBITS];
482 /* ==================================== */
485 static C_block constdatablock; /* encryption constant */
486 static char cryptresult[1 + 4 + 4 + 11 + 1]; /* encrypted result */
489 * Return a pointer to static data consisting of the "setting"
490 * followed by an encryption produced by the "key" and "setting".
493 crypt(register const char *key, register const char *setting)
499 int num_iter, salt_size;
500 C_block keyblock, rsltblock;
503 for (i = 0; i < 8; i++) {
504 if ((t = 2 * (unsigned char)(*key)) != 0)
508 if (des_setkey((char *)keyblock.b)) /* also initializes "a64toi" */
511 encp = &cryptresult[0];
513 case '_': /* was EFMT1 */
515 * Involve the rest of the password 8 characters at a time.
518 if (des_cipher((char *)&keyblock, (char *)&keyblock, 0L, 1))
520 for (i = 0; i < 8; i++) {
521 if ((t = 2 * (unsigned char)(*key)) != 0)
525 if (des_setkey((char *)keyblock.b))
529 *encp++ = *setting++;
531 /* get iteration count */
533 for (i = 4; --i >= 0;) {
534 if ((t = (unsigned char)setting[i]) == '\0')
537 num_iter = (num_iter << 6) | a64toi[t];
549 for (i = salt_size; --i >= 0;) {
550 if ((t = (unsigned char)setting[i]) == '\0')
553 salt = (salt << 6) | a64toi[t];
557 ((char *)&constdatablock, (char *)&rsltblock, salt, num_iter))
561 * Encode the 64 cipher bits as 11 ascii characters.
563 i = ((long)((rsltblock.b[0] << 8) | rsltblock.b[1]) << 8) | rsltblock.
565 encp[3] = itoa64[i & 0x3f];
567 encp[2] = itoa64[i & 0x3f];
569 encp[1] = itoa64[i & 0x3f];
573 i = ((long)((rsltblock.b[3] << 8) | rsltblock.b[4]) << 8) | rsltblock.
575 encp[3] = itoa64[i & 0x3f];
577 encp[2] = itoa64[i & 0x3f];
579 encp[1] = itoa64[i & 0x3f];
583 i = ((long)((rsltblock.b[6]) << 8) | rsltblock.b[7]) << 2;
584 encp[2] = itoa64[i & 0x3f];
586 encp[1] = itoa64[i & 0x3f];
592 return (cryptresult);
597 * The Key Schedule, filled in by des_setkey() or setkey().
600 static C_block KS[KS_SIZE];
603 * Set up the key schedule from the key.
606 des_setkey(register const char *key)
608 register DCL_BLOCK(K, K0, K1);
609 register C_block *ptabp;
611 static int des_ready = 0;
618 PERM6464(K, K0, K1, (unsigned char *)key, (C_block *) PC1ROT);
619 key = (char *)&KS[0];
620 STORE(K & ~0x03030303L, K0 & ~0x03030303L, K1, *(C_block *) key);
621 for (i = 1; i < 16; i++) {
622 key += sizeof(C_block);
623 STORE(K, K0, K1, *(C_block *) key);
624 ptabp = (C_block *) PC2ROT[Rotates[i] - 1];
625 PERM6464(K, K0, K1, (unsigned char *)key, ptabp);
626 STORE(K & ~0x03030303L, K0 & ~0x03030303L, K1, *(C_block *) key);
632 * Encrypt (or decrypt if num_iter < 0) the 8 chars at "in" with abs(num_iter)
633 * iterations of DES, using the the given 24-bit salt and the pre-computed key
634 * schedule, and store the resulting 8 chars at "out" (in == out is permitted).
636 * NOTE: the performance of this routine is critically dependent on your
637 * compiler and machine architecture.
640 des_cipher(const char *in, char *out, long salt, int num_iter)
642 /* variables that we want in registers, most important first */
646 register long L0, L1, R0, R1, k;
647 register C_block *kp;
648 register int ks_inc, loop_count;
652 TO_SIX_BIT(salt, L0); /* convert to 4*(6+2) format */
654 #if defined(vax) || defined(pdp11)
655 salt = ~salt; /* "x &~ y" is faster than "x & y". */
661 #if defined(MUST_ALIGN)
672 LOAD(L, L0, L1, *(C_block *) in);
674 LOADREG(R, R0, R1, L, L0, L1);
677 L0 = (L0 << 1) | L1; /* L0 is the even-numbered input bits */
679 R1 = (R1 >> 1) & 0x55555555L;
680 L1 = R0 | R1; /* L1 is the odd-numbered input bits */
682 PERM3264(L, L0, L1, B.b, (C_block *) IE3264); /* even bits */
683 PERM3264(R, R0, R1, B.b + 4, (C_block *) IE3264); /* odd bits */
685 if (num_iter >= 0) { /* encryption */
687 ks_inc = sizeof(*kp);
688 } else { /* decryption */
689 num_iter = -num_iter;
690 kp = &KS[KS_SIZE - 1];
691 ks_inc = -((long)sizeof(*kp));
694 while (--num_iter >= 0) {
698 #define SPTAB(t, i) (*(long *)((unsigned char *)t + i*(sizeof(long)/4)))
700 /* use this if B.b[i] is evaluated just once ... */
701 #define DOXOR(x,y,i) x^=SPTAB(SPE[0][i],B.b[i]); y^=SPTAB(SPE[1][i],B.b[i]);
704 /* use this if your "long" int indexing is slow */
705 #define DOXOR(x,y,i) j=B.b[i]; x^=SPTAB(SPE[0][i],j); y^=SPTAB(SPE[1][i],j);
707 /* use this if "k" is allocated to a register ... */
708 #define DOXOR(x,y,i) k=B.b[i]; x^=SPTAB(SPE[0][i],k); y^=SPTAB(SPE[1][i],k);
712 #define CRUNCH(p0, p1, q0, q1) \
713 k = (q0 ^ q1) & SALT; \
714 B.b32.i0 = k ^ q0 ^ kp->b32.i0; \
715 B.b32.i1 = k ^ q1 ^ kp->b32.i1; \
716 kp = (C_block *)((char *)kp+ks_inc); \
727 CRUNCH(L0, L1, R0, R1);
728 CRUNCH(R0, R1, L0, L1);
729 } while (--loop_count != 0);
730 kp = (C_block *) ((char *)kp - (ks_inc * KS_SIZE));
742 /* store the encrypted (or decrypted) result */
743 L0 = ((L0 >> 3) & 0x0f0f0f0fL) | ((L1 << 1) & 0xf0f0f0f0L);
744 L1 = ((R0 >> 3) & 0x0f0f0f0fL) | ((R1 << 1) & 0xf0f0f0f0L);
746 PERM6464(L, L0, L1, B.b, (C_block *) CF6464);
747 #if defined(MUST_ALIGN)
758 STORE(L, L0, L1, *(C_block *) out);
765 * Initialize various tables. This need only be done once. It could even be
766 * done at compile time, if the compiler were capable of that sort of thing.
773 register int tableno;
774 static unsigned char perm[64], tmp32[32]; /* "static" for speed */
777 * table that converts chars "./0-9A-Za-z"to integers 0-63.
779 for (i = 0; i < 64; i++)
780 a64toi[itoa64[i]] = i;
783 * PC1ROT - bit reverse, then PC1, then Rotate, then PC2.
785 for (i = 0; i < 64; i++)
787 for (i = 0; i < 64; i++) {
788 if ((k = PC2[i]) == 0)
791 if ((k % 28) < Rotates[0])
796 k = (k | 07) - (k & 07);
799 perm[i] = (unsigned char)k;
802 prtab("pc1tab", perm, 8);
804 init_perm(PC1ROT, perm, 8, 8);
807 * PC2ROT - PC2 inverse, then Rotate (once or twice), then PC2.
809 for (j = 0; j < 2; j++) {
810 unsigned char pc2inv[64];
811 for (i = 0; i < 64; i++)
812 perm[i] = pc2inv[i] = 0;
813 for (i = 0; i < 64; i++) {
814 if ((k = PC2[i]) == 0)
816 pc2inv[k - 1] = i + 1;
818 for (i = 0; i < 64; i++) {
819 if ((k = PC2[i]) == 0)
827 prtab("pc2tab", perm, 8);
829 init_perm(PC2ROT[j], perm, 8, 8);
833 * Bit reverse, then initial permutation, then expansion.
835 for (i = 0; i < 8; i++) {
836 for (j = 0; j < 8; j++) {
837 k = (j < 2) ? 0 : IP[ExpandTr[i * 6 + j - 2] - 1];
844 k = (k | 07) - (k & 07);
847 perm[i * 8 + j] = (unsigned char)k;
851 prtab("ietab", perm, 8);
853 init_perm(IE3264, perm, 4, 8);
856 * Compression, then final permutation, then bit reverse.
858 for (i = 0; i < 64; i++) {
862 k = (k | 07) - (k & 07);
868 prtab("cftab", perm, 8);
870 init_perm(CF6464, perm, 8, 8);
875 for (i = 0; i < 48; i++)
876 perm[i] = P32Tr[ExpandTr[i] - 1];
877 for (tableno = 0; tableno < 8; tableno++) {
878 for (j = 0; j < 64; j++) {
879 k = (((j >> 0) & 01) << 5) | (((j >> 1) & 01) << 3) |
880 (((j >> 2) & 01) << 2) | (((j >> 3) & 01) << 1) |
881 (((j >> 4) & 01) << 0) | (((j >> 5) & 01) << 4);
883 k = (((k >> 3) & 01) << 0) | (((k >> 2) & 01) << 1) |
884 (((k >> 1) & 01) << 2) | (((k >> 0) & 01) << 3);
885 for (i = 0; i < 32; i++)
887 for (i = 0; i < 4; i++)
888 tmp32[4 * tableno + i] = (k >> i) & 01;
890 for (i = 24; --i >= 0;)
891 k = (k << 1) | tmp32[perm[i] - 1];
892 TO_SIX_BIT(SPE[0][tableno][j], k);
894 for (i = 24; --i >= 0;)
895 k = (k << 1) | tmp32[perm[i + 24] - 1];
896 TO_SIX_BIT(SPE[1][tableno][j], k);
902 * Initialize "perm" to represent transformation "p", which rearranges
903 * (perhaps with expansion and/or contraction) one packed array of bits
904 * (of size "chars_in" characters) into another array (of size "chars_out"
907 * "perm" must be all-zeroes on entry to this routine.
910 init_perm(C_block perm[64 / CHUNKBITS][1 << CHUNKBITS],
911 unsigned char p[64], int chars_in, int chars_out)
913 register int i, j, k, l;
915 for (k = 0; k < chars_out * 8; k++) { /* each output bit position */
916 l = p[k] - 1; /* where this bit comes from */
918 continue; /* output bit is always 0 */
919 i = l >> LGCHUNKBITS; /* which chunk this bit comes from */
920 l = 1 << (l & (CHUNKBITS - 1)); /* mask for this bit */
921 for (j = 0; j < (1 << CHUNKBITS); j++) { /* each chunk value */
923 perm[i][j].b[k >> 3] |= 1 << (k & 07);
929 * "setkey" routine (for backwards compatibility)
931 #if 0 /* static and doesn't appear to be referenced */
934 register const char *key;
936 register int i, j, k;
939 for (i = 0; i < 8; i++) {
941 for (j = 0; j < 8; j++) {
943 k |= (unsigned char)*key++;
947 return (des_setkey((char *)keyblock.b));
953 * "encrypt" routine (for backwards compatibility)
957 register char *block;
960 register int i, j, k;
963 for (i = 0; i < 8; i++) {
965 for (j = 0; j < 8; j++) {
967 k |= (unsigned char)*block++;
971 if (des_cipher((char *)&cblock, (char *)&cblock, 0L, (flag ? -1 : 1)))
973 for (i = 7; i >= 0; i--) {
975 for (j = 7; j >= 0; j--) {
986 prtab(char *s, unsigned char *t, int num_rows)
990 (void)printf("%s:\n", s);
991 for (i = 0; i < num_rows; i++) {
992 for (j = 0; j < 8; j++) {
993 (void)printf("%3d", t[i * 8 + j]);