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>
56 * UNIX password, and DES, encryption.
57 * By Tom Truscott, trt@rti.rti.org,
58 * from algorithms by Robert W. Baldwin and James Gillogly.
61 * "Mathematical Cryptology for Computer Scientists and Mathematicians,"
62 * by Wayne Patterson, 1987, ISBN 0-8476-7438-X.
64 * "Password Security: A Case History," R. Morris and Ken Thompson,
65 * Communications of the ACM, vol. 22, pp. 594-597, Nov. 1979.
67 * "DES will be Totally Insecure within Ten Years," M.E. Hellman,
68 * IEEE Spectrum, vol. 16, pp. 32-39, July 1979.
71 /* ===== Configuration ==================== */
74 * define "MUST_ALIGN" if your compiler cannot load/store
75 * long integers at arbitrary (e.g. odd) memory locations.
76 * (Either that or never pass unaligned addresses to des_cipher!)
84 #error C_block structure assumes 8 bit characters
89 * define "LONG_IS_32_BITS" only if sizeof(long)==4.
90 * This avoids use of bit fields (your compiler may be sloppy with them).
93 /* XXX shouldn't this be !AFS_64BIT_ENV ? */
94 #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)
95 #define LONG_IS_32_BITS
99 * define "B64" to be the declaration for a 64 bit integer.
100 * XXX this feature is currently unused, see "endian" comment below.
106 #define B64 long long
110 * define "LARGEDATA" to get faster permutations, by using about 72 kilobytes
111 * of lookup tables. This speeds up des_setkey() and des_cipher(), but has
112 * little effect on crypt().
118 /* compile with "-DSTATIC=int" when profiling */
120 #define STATIC static
122 STATIC void init_des();
123 STATIC void permute();
124 STATIC void init_perm();
126 STATIC int des_setkey(const char *key);
127 STATIC int des_cipher(const char *in, char *out, long salt, int num_iter);
133 /* ==================================== */
136 * Cipher-block representation (Bob Baldwin):
138 * DES operates on groups of 64 bits, numbered 1..64 (sigh). One
139 * representation is to store one bit per byte in an array of bytes. Bit N of
140 * the NBS spec is stored as the LSB of the Nth byte (index N-1) in the array.
141 * Another representation stores the 64 bits in 8 bytes, with bits 1..8 in the
142 * first byte, 9..16 in the second, and so on. The DES spec apparently has
143 * bit 1 in the MSB of the first byte, but that is particularly noxious so we
144 * bit-reverse each byte so that bit 1 is the LSB of the first byte, bit 8 is
145 * the MSB of the first byte. Specifically, the 64-bit input data and key are
146 * converted to LSB format, and the output 64-bit block is converted back into
149 * DES operates internally on groups of 32 bits which are expanded to 48 bits
150 * by permutation E and shrunk back to 32 bits by the S boxes. To speed up
151 * the computation, the expansion is applied only once, the expanded
152 * representation is maintained during the encryption, and a compression
153 * permutation is applied only at the end. To speed up the S-box lookups,
154 * the 48 bits are maintained as eight 6 bit groups, one per byte, which
155 * directly feed the eight S-boxes. Within each byte, the 6 bits are the
156 * most significant ones. The low two bits of each byte are zero. (Thus,
157 * bit 1 of the 48 bit E expansion is stored as the "4"-valued bit of the
158 * first byte in the eight byte representation, bit 2 of the 48 bit value is
159 * the "8"-valued bit, and so on.) In fact, a combined "SPE"-box lookup is
160 * used, in which the output is the 64 bit result of an S-box lookup which
161 * has been permuted by P and expanded by E, and is ready for use in the next
162 * iteration. Two 32-bit wide tables, SPE[0] and SPE[1], are used for this
163 * lookup. Since each byte in the 48 bit path is a multiple of four, indexed
164 * lookup of SPE[0] and SPE[1] is simple and fast. The key schedule and
165 * "salt" are also converted to this 8*(6+2) format. The SPE table size is
168 * To speed up bit-parallel operations (such as XOR), the 8 byte
169 * representation is "union"ed with 32 bit values "i0" and "i1", and, on
170 * machines which support it, a 64 bit value "b64". This data structure,
171 * "C_block", has two problems. First, alignment restrictions must be
172 * honored. Second, the byte-order (e.g. little-endian or big-endian) of
173 * the architecture becomes visible.
175 * The byte-order problem is unfortunate, since on the one hand it is good
176 * to have a machine-independent C_block representation (bits 1..8 in the
177 * first byte, etc.), and on the other hand it is good for the LSB of the
178 * first byte to be the LSB of i0. We cannot have both these things, so we
179 * currently use the "little-endian" representation and avoid any multi-byte
180 * operations that depend on byte order. This largely precludes use of the
181 * 64-bit datatype since the relative order of i0 and i1 are unknown. It
182 * also inhibits grouping the SPE table to look up 12 bits at a time. (The
183 * 12 bits can be stored in a 16-bit field with 3 low-order zeroes and 1
184 * high-order zero, providing fast indexing into a 64-bit wide SPE.) On the
185 * other hand, 64-bit datatypes are currently rare, and a 12-bit SPE lookup
186 * requires a 128 kilobyte table, so perhaps this is not a big loss.
188 * Permutation representation (Jim Gillogly):
190 * A transformation is defined by its effect on each of the 8 bytes of the
191 * 64-bit input. For each byte we give a 64-bit output that has the bits in
192 * the input distributed appropriately. The transformation is then the OR
193 * of the 8 sets of 64-bits. This uses 8*256*8 = 16K bytes of storage for
194 * each transformation. Unless LARGEDATA is defined, however, a more compact
195 * table is used which looks up 16 4-bit "chunks" rather than 8 8-bit chunks.
196 * The smaller table uses 16*16*8 = 2K bytes for each transformation. This
197 * is slower but tolerable, particularly for password encryption in which
198 * the SPE transformation is iterated many times. The small tables total 9K
199 * bytes, the large tables total 72K bytes.
201 * The transformations used are:
202 * IE3264: MSB->LSB conversion, initial permutation, and expansion.
203 * This is done by collecting the 32 even-numbered bits and applying
204 * a 32->64 bit transformation, and then collecting the 32 odd-numbered
205 * bits and applying the same transformation. Since there are only
206 * 32 input bits, the IE3264 transformation table is half the size of
208 * CF6464: Compression, final permutation, and LSB->MSB conversion.
209 * This is done by two trivial 48->32 bit compressions to obtain
210 * a 64-bit block (the bit numbering is given in the "CIFP" table)
211 * followed by a 64->64 bit "cleanup" transformation. (It would
212 * be possible to group the bits in the 64-bit block so that 2
213 * identical 32->32 bit transformations could be used instead,
214 * saving a factor of 4 in space and possibly 2 in time, but
215 * byte-ordering and other complications rear their ugly head.
216 * Similar opportunities/problems arise in the key schedule
218 * PC1ROT: MSB->LSB, PC1 permutation, rotate, and PC2 permutation.
219 * This admittedly baroque 64->64 bit transformation is used to
220 * produce the first code (in 8*(6+2) format) of the key schedule.
221 * PC2ROT[0]: Inverse PC2 permutation, rotate, and PC2 permutation.
222 * It would be possible to define 15 more transformations, each
223 * with a different rotation, to generate the entire key schedule.
224 * To save space, however, we instead permute each code into the
225 * next by using a transformation that "undoes" the PC2 permutation,
226 * rotates the code, and then applies PC2. Unfortunately, PC2
227 * transforms 56 bits into 48 bits, dropping 8 bits, so PC2 is not
228 * invertible. We get around that problem by using a modified PC2
229 * which retains the 8 otherwise-lost bits in the unused low-order
230 * bits of each byte. The low-order bits are cleared when the
231 * codes are stored into the key schedule.
232 * PC2ROT[1]: Same as PC2ROT[0], but with two rotations.
233 * This is faster than applying PC2ROT[0] twice,
235 * The Bell Labs "salt" (Bob Baldwin):
237 * The salting is a simple permutation applied to the 48-bit result of E.
238 * Specifically, if bit i (1 <= i <= 24) of the salt is set then bits i and
239 * i+24 of the result are swapped. The salt is thus a 24 bit number, with
240 * 16777216 possible values. (The original salt was 12 bits and could not
241 * swap bits 13..24 with 36..48.)
243 * It is possible, but ugly, to warp the SPE table to account for the salt
244 * permutation. Fortunately, the conditional bit swapping requires only
245 * about four machine instructions and can be done on-the-fly with about an
246 * 8% performance penalty.
252 #if defined(LONG_IS_32_BITS)
253 /* long is often faster than a 32-bit bit field */
267 * Convert twenty-four-bit long in host-order
268 * to six bits (and 2 low-order zeroes) per char little-endian format.
270 #define TO_SIX_BIT(rslt, src) { \
272 cvt.b[0] = (unsigned char) src; src >>= 6; \
273 cvt.b[1] = (unsigned char) src; src >>= 6; \
274 cvt.b[2] = (unsigned char) src; src >>= 6; \
275 cvt.b[3] = (unsigned char) src; \
276 rslt = (cvt.b32.i0 & 0x3f3f3f3fL) << 2; \
280 * These macros may someday permit efficient use of 64-bit integers.
282 #define ZERO(d,d0,d1) d0 = 0, d1 = 0
283 #define LOAD(d,d0,d1,bl) d0 = (bl).b32.i0, d1 = (bl).b32.i1
284 #define LOADREG(d,d0,d1,s,s0,s1) d0 = s0, d1 = s1
285 #define OR(d,d0,d1,bl) d0 |= (bl).b32.i0, d1 |= (bl).b32.i1
286 #define STORE(s,s0,s1,bl) (bl).b32.i0 = (s0), (bl).b32.i1 = (s1)
287 #define DCL_BLOCK(d,d0,d1) long d0, d1
289 #if defined(LARGEDATA)
290 /* Waste memory like crazy. Also, do permutations in line */
291 #define LGCHUNKBITS 3
292 #define CHUNKBITS (1<<LGCHUNKBITS)
293 #define PERM6464(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]]); \
298 OR (d,d0,d1,(p)[(4<<CHUNKBITS)+(cpp)[4]]); \
299 OR (d,d0,d1,(p)[(5<<CHUNKBITS)+(cpp)[5]]); \
300 OR (d,d0,d1,(p)[(6<<CHUNKBITS)+(cpp)[6]]); \
301 OR (d,d0,d1,(p)[(7<<CHUNKBITS)+(cpp)[7]]);
302 #define PERM3264(d,d0,d1,cpp,p) \
303 LOAD(d,d0,d1,(p)[(0<<CHUNKBITS)+(cpp)[0]]); \
304 OR (d,d0,d1,(p)[(1<<CHUNKBITS)+(cpp)[1]]); \
305 OR (d,d0,d1,(p)[(2<<CHUNKBITS)+(cpp)[2]]); \
306 OR (d,d0,d1,(p)[(3<<CHUNKBITS)+(cpp)[3]]);
309 #define LGCHUNKBITS 2
310 #define CHUNKBITS (1<<LGCHUNKBITS)
311 #define PERM6464(d,d0,d1,cpp,p) \
312 { C_block tblk; permute(cpp,&tblk,p,8); LOAD (d,d0,d1,tblk); }
313 #define PERM3264(d,d0,d1,cpp,p) \
314 { C_block tblk; permute(cpp,&tblk,p,4); LOAD (d,d0,d1,tblk); }
317 permute(cp, out, p, chars_in)
323 register DCL_BLOCK(D, D0, D1);
324 register C_block *tp;
332 p += (1 << CHUNKBITS);
335 p += (1 << CHUNKBITS);
336 } while (--chars_in > 0);
337 STORE(D, D0, D1, *out);
339 #endif /* LARGEDATA */
342 /* ===== (mostly) Standard DES Tables ==================== */
344 static unsigned char IP[] = { /* initial permutation */
345 58, 50, 42, 34, 26, 18, 10, 2,
346 60, 52, 44, 36, 28, 20, 12, 4,
347 62, 54, 46, 38, 30, 22, 14, 6,
348 64, 56, 48, 40, 32, 24, 16, 8,
349 57, 49, 41, 33, 25, 17, 9, 1,
350 59, 51, 43, 35, 27, 19, 11, 3,
351 61, 53, 45, 37, 29, 21, 13, 5,
352 63, 55, 47, 39, 31, 23, 15, 7,
355 /* The final permutation is the inverse of IP - no table is necessary */
357 static unsigned char ExpandTr[] = { /* expansion operation */
360 8, 9, 10, 11, 12, 13,
361 12, 13, 14, 15, 16, 17,
362 16, 17, 18, 19, 20, 21,
363 20, 21, 22, 23, 24, 25,
364 24, 25, 26, 27, 28, 29,
365 28, 29, 30, 31, 32, 1,
368 static unsigned char PC1[] = { /* permuted choice table 1 */
369 57, 49, 41, 33, 25, 17, 9,
370 1, 58, 50, 42, 34, 26, 18,
371 10, 2, 59, 51, 43, 35, 27,
372 19, 11, 3, 60, 52, 44, 36,
374 63, 55, 47, 39, 31, 23, 15,
375 7, 62, 54, 46, 38, 30, 22,
376 14, 6, 61, 53, 45, 37, 29,
377 21, 13, 5, 28, 20, 12, 4,
380 static unsigned char Rotates[] = { /* PC1 rotation schedule */
381 1, 1, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 1,
384 /* note: each "row" of PC2 is left-padded with bits that make it invertible */
385 static unsigned char PC2[] = { /* permuted choice table 2 */
386 9, 18, 14, 17, 11, 24, 1, 5,
387 22, 25, 3, 28, 15, 6, 21, 10,
388 35, 38, 23, 19, 12, 4, 26, 8,
389 43, 54, 16, 7, 27, 20, 13, 2,
391 0, 0, 41, 52, 31, 37, 47, 55,
392 0, 0, 30, 40, 51, 45, 33, 48,
393 0, 0, 44, 49, 39, 56, 34, 53,
394 0, 0, 46, 42, 50, 36, 29, 32,
397 static unsigned char S[8][64] = { /* 48->32 bit substitution tables */
399 {14, 4, 13, 1, 2, 15, 11, 8, 3, 10, 6, 12, 5, 9, 0, 7,
400 0, 15, 7, 4, 14, 2, 13, 1, 10, 6, 12, 11, 9, 5, 3, 8,
401 4, 1, 14, 8, 13, 6, 2, 11, 15, 12, 9, 7, 3, 10, 5, 0,
402 15, 12, 8, 2, 4, 9, 1, 7, 5, 11, 3, 14, 10, 0, 6, 13,},
404 {15, 1, 8, 14, 6, 11, 3, 4, 9, 7, 2, 13, 12, 0, 5, 10,
405 3, 13, 4, 7, 15, 2, 8, 14, 12, 0, 1, 10, 6, 9, 11, 5,
406 0, 14, 7, 11, 10, 4, 13, 1, 5, 8, 12, 6, 9, 3, 2, 15,
407 13, 8, 10, 1, 3, 15, 4, 2, 11, 6, 7, 12, 0, 5, 14, 9,},
409 {10, 0, 9, 14, 6, 3, 15, 5, 1, 13, 12, 7, 11, 4, 2, 8,
410 13, 7, 0, 9, 3, 4, 6, 10, 2, 8, 5, 14, 12, 11, 15, 1,
411 13, 6, 4, 9, 8, 15, 3, 0, 11, 1, 2, 12, 5, 10, 14, 7,
412 1, 10, 13, 0, 6, 9, 8, 7, 4, 15, 14, 3, 11, 5, 2, 12,},
414 {7, 13, 14, 3, 0, 6, 9, 10, 1, 2, 8, 5, 11, 12, 4, 15,
415 13, 8, 11, 5, 6, 15, 0, 3, 4, 7, 2, 12, 1, 10, 14, 9,
416 10, 6, 9, 0, 12, 11, 7, 13, 15, 1, 3, 14, 5, 2, 8, 4,
417 3, 15, 0, 6, 10, 1, 13, 8, 9, 4, 5, 11, 12, 7, 2, 14,},
419 {2, 12, 4, 1, 7, 10, 11, 6, 8, 5, 3, 15, 13, 0, 14, 9,
420 14, 11, 2, 12, 4, 7, 13, 1, 5, 0, 15, 10, 3, 9, 8, 6,
421 4, 2, 1, 11, 10, 13, 7, 8, 15, 9, 12, 5, 6, 3, 0, 14,
422 11, 8, 12, 7, 1, 14, 2, 13, 6, 15, 0, 9, 10, 4, 5, 3,},
424 {12, 1, 10, 15, 9, 2, 6, 8, 0, 13, 3, 4, 14, 7, 5, 11,
425 10, 15, 4, 2, 7, 12, 9, 5, 6, 1, 13, 14, 0, 11, 3, 8,
426 9, 14, 15, 5, 2, 8, 12, 3, 7, 0, 4, 10, 1, 13, 11, 6,
427 4, 3, 2, 12, 9, 5, 15, 10, 11, 14, 1, 7, 6, 0, 8, 13,},
429 {4, 11, 2, 14, 15, 0, 8, 13, 3, 12, 9, 7, 5, 10, 6, 1,
430 13, 0, 11, 7, 4, 9, 1, 10, 14, 3, 5, 12, 2, 15, 8, 6,
431 1, 4, 11, 13, 12, 3, 7, 14, 10, 15, 6, 8, 0, 5, 9, 2,
432 6, 11, 13, 8, 1, 4, 10, 7, 9, 5, 0, 15, 14, 2, 3, 12,},
434 {13, 2, 8, 4, 6, 15, 11, 1, 10, 9, 3, 14, 5, 0, 12, 7,
435 1, 15, 13, 8, 10, 3, 7, 4, 12, 5, 6, 11, 0, 14, 9, 2,
436 7, 11, 4, 1, 9, 12, 14, 2, 0, 6, 10, 13, 15, 3, 5, 8,
437 2, 1, 14, 7, 4, 10, 8, 13, 15, 12, 9, 0, 3, 5, 6, 11,}
440 static unsigned char P32Tr[] = { /* 32-bit permutation function */
451 static unsigned char CIFP[] = { /* compressed/interleaved permutation */
452 1, 2, 3, 4, 17, 18, 19, 20,
453 5, 6, 7, 8, 21, 22, 23, 24,
454 9, 10, 11, 12, 25, 26, 27, 28,
455 13, 14, 15, 16, 29, 30, 31, 32,
457 33, 34, 35, 36, 49, 50, 51, 52,
458 37, 38, 39, 40, 53, 54, 55, 56,
459 41, 42, 43, 44, 57, 58, 59, 60,
460 45, 46, 47, 48, 61, 62, 63, 64,
463 static unsigned char itoa64[] = /* 0..63 => ascii-64 */
464 "./0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
467 /* ===== Tables that are initialized at run time ==================== */
470 static unsigned char a64toi[128]; /* ascii-64 => 0..63 */
472 /* Initial key schedule permutation */
473 static C_block PC1ROT[64 / CHUNKBITS][1 << CHUNKBITS];
475 /* Subsequent key schedule rotation permutations */
476 static C_block PC2ROT[2][64 / CHUNKBITS][1 << CHUNKBITS];
478 /* Initial permutation/expansion table */
479 static C_block IE3264[32 / CHUNKBITS][1 << CHUNKBITS];
481 /* Table that combines the S, P, and E operations. */
482 static long SPE[2][8][64];
484 /* compressed/interleaved => final permutation table */
485 static C_block CF6464[64 / CHUNKBITS][1 << CHUNKBITS];
488 /* ==================================== */
491 static C_block constdatablock; /* encryption constant */
492 static char cryptresult[1 + 4 + 4 + 11 + 1]; /* encrypted result */
495 * Return a pointer to static data consisting of the "setting"
496 * followed by an encryption produced by the "key" and "setting".
500 register const char *key;
501 register const char *setting;
507 int num_iter, salt_size;
508 C_block keyblock, rsltblock;
511 for (i = 0; i < 8; i++) {
512 if ((t = 2 * (unsigned char)(*key)) != 0)
516 if (des_setkey((char *)keyblock.b)) /* also initializes "a64toi" */
519 encp = &cryptresult[0];
521 case '_': /* was EFMT1 */
523 * Involve the rest of the password 8 characters at a time.
526 if (des_cipher((char *)&keyblock, (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];
565 ((char *)&constdatablock, (char *)&rsltblock, salt, num_iter))
569 * Encode the 64 cipher bits as 11 ascii characters.
571 i = ((long)((rsltblock.b[0] << 8) | rsltblock.b[1]) << 8) | rsltblock.
573 encp[3] = itoa64[i & 0x3f];
575 encp[2] = itoa64[i & 0x3f];
577 encp[1] = itoa64[i & 0x3f];
581 i = ((long)((rsltblock.b[3] << 8) | rsltblock.b[4]) << 8) | rsltblock.
583 encp[3] = itoa64[i & 0x3f];
585 encp[2] = itoa64[i & 0x3f];
587 encp[1] = itoa64[i & 0x3f];
591 i = ((long)((rsltblock.b[6]) << 8) | rsltblock.b[7]) << 2;
592 encp[2] = itoa64[i & 0x3f];
594 encp[1] = itoa64[i & 0x3f];
600 return (cryptresult);
605 * The Key Schedule, filled in by des_setkey() or setkey().
608 static C_block KS[KS_SIZE];
611 * Set up the key schedule from the key.
615 register const char *key;
617 register DCL_BLOCK(K, K0, K1);
618 register C_block *ptabp;
620 static int des_ready = 0;
627 PERM6464(K, K0, K1, (unsigned char *)key, (C_block *) PC1ROT);
628 key = (char *)&KS[0];
629 STORE(K & ~0x03030303L, K0 & ~0x03030303L, K1, *(C_block *) key);
630 for (i = 1; i < 16; i++) {
631 key += sizeof(C_block);
632 STORE(K, K0, K1, *(C_block *) key);
633 ptabp = (C_block *) PC2ROT[Rotates[i] - 1];
634 PERM6464(K, K0, K1, (unsigned char *)key, ptabp);
635 STORE(K & ~0x03030303L, K0 & ~0x03030303L, K1, *(C_block *) key);
641 * Encrypt (or decrypt if num_iter < 0) the 8 chars at "in" with abs(num_iter)
642 * iterations of DES, using the the given 24-bit salt and the pre-computed key
643 * schedule, and store the resulting 8 chars at "out" (in == out is permitted).
645 * NOTE: the performance of this routine is critically dependent on your
646 * compiler and machine architecture.
649 des_cipher(in, out, salt, num_iter)
655 /* variables that we want in registers, most important first */
659 register long L0, L1, R0, R1, k;
660 register C_block *kp;
661 register int ks_inc, loop_count;
665 TO_SIX_BIT(salt, L0); /* convert to 4*(6+2) format */
667 #if defined(vax) || defined(pdp11)
668 salt = ~salt; /* "x &~ y" is faster than "x & y". */
674 #if defined(MUST_ALIGN)
685 LOAD(L, L0, L1, *(C_block *) in);
687 LOADREG(R, R0, R1, L, L0, L1);
690 L0 = (L0 << 1) | L1; /* L0 is the even-numbered input bits */
692 R1 = (R1 >> 1) & 0x55555555L;
693 L1 = R0 | R1; /* L1 is the odd-numbered input bits */
695 PERM3264(L, L0, L1, B.b, (C_block *) IE3264); /* even bits */
696 PERM3264(R, R0, R1, B.b + 4, (C_block *) IE3264); /* odd bits */
698 if (num_iter >= 0) { /* encryption */
700 ks_inc = sizeof(*kp);
701 } else { /* decryption */
702 num_iter = -num_iter;
703 kp = &KS[KS_SIZE - 1];
704 ks_inc = -((long)sizeof(*kp));
707 while (--num_iter >= 0) {
711 #define SPTAB(t, i) (*(long *)((unsigned char *)t + i*(sizeof(long)/4)))
713 /* use this if B.b[i] is evaluated just once ... */
714 #define DOXOR(x,y,i) x^=SPTAB(SPE[0][i],B.b[i]); y^=SPTAB(SPE[1][i],B.b[i]);
717 /* use this if your "long" int indexing is slow */
718 #define DOXOR(x,y,i) j=B.b[i]; x^=SPTAB(SPE[0][i],j); y^=SPTAB(SPE[1][i],j);
720 /* use this if "k" is allocated to a register ... */
721 #define DOXOR(x,y,i) k=B.b[i]; x^=SPTAB(SPE[0][i],k); y^=SPTAB(SPE[1][i],k);
725 #define CRUNCH(p0, p1, q0, q1) \
726 k = (q0 ^ q1) & SALT; \
727 B.b32.i0 = k ^ q0 ^ kp->b32.i0; \
728 B.b32.i1 = k ^ q1 ^ kp->b32.i1; \
729 kp = (C_block *)((char *)kp+ks_inc); \
740 CRUNCH(L0, L1, R0, R1);
741 CRUNCH(R0, R1, L0, L1);
742 } while (--loop_count != 0);
743 kp = (C_block *) ((char *)kp - (ks_inc * KS_SIZE));
755 /* store the encrypted (or decrypted) result */
756 L0 = ((L0 >> 3) & 0x0f0f0f0fL) | ((L1 << 1) & 0xf0f0f0f0L);
757 L1 = ((R0 >> 3) & 0x0f0f0f0fL) | ((R1 << 1) & 0xf0f0f0f0L);
759 PERM6464(L, L0, L1, B.b, (C_block *) CF6464);
760 #if defined(MUST_ALIGN)
771 STORE(L, L0, L1, *(C_block *) out);
778 * Initialize various tables. This need only be done once. It could even be
779 * done at compile time, if the compiler were capable of that sort of thing.
786 register int tableno;
787 static unsigned char perm[64], tmp32[32]; /* "static" for speed */
790 * table that converts chars "./0-9A-Za-z"to integers 0-63.
792 for (i = 0; i < 64; i++)
793 a64toi[itoa64[i]] = i;
796 * PC1ROT - bit reverse, then PC1, then Rotate, then PC2.
798 for (i = 0; i < 64; i++)
800 for (i = 0; i < 64; i++) {
801 if ((k = PC2[i]) == 0)
804 if ((k % 28) < Rotates[0])
809 k = (k | 07) - (k & 07);
812 perm[i] = (unsigned char)k;
815 prtab("pc1tab", perm, 8);
817 init_perm(PC1ROT, perm, 8, 8);
820 * PC2ROT - PC2 inverse, then Rotate (once or twice), then PC2.
822 for (j = 0; j < 2; j++) {
823 unsigned char pc2inv[64];
824 for (i = 0; i < 64; i++)
825 perm[i] = pc2inv[i] = 0;
826 for (i = 0; i < 64; i++) {
827 if ((k = PC2[i]) == 0)
829 pc2inv[k - 1] = i + 1;
831 for (i = 0; i < 64; i++) {
832 if ((k = PC2[i]) == 0)
840 prtab("pc2tab", perm, 8);
842 init_perm(PC2ROT[j], perm, 8, 8);
846 * Bit reverse, then initial permutation, then expansion.
848 for (i = 0; i < 8; i++) {
849 for (j = 0; j < 8; j++) {
850 k = (j < 2) ? 0 : IP[ExpandTr[i * 6 + j - 2] - 1];
857 k = (k | 07) - (k & 07);
860 perm[i * 8 + j] = (unsigned char)k;
864 prtab("ietab", perm, 8);
866 init_perm(IE3264, perm, 4, 8);
869 * Compression, then final permutation, then bit reverse.
871 for (i = 0; i < 64; i++) {
875 k = (k | 07) - (k & 07);
881 prtab("cftab", perm, 8);
883 init_perm(CF6464, perm, 8, 8);
888 for (i = 0; i < 48; i++)
889 perm[i] = P32Tr[ExpandTr[i] - 1];
890 for (tableno = 0; tableno < 8; tableno++) {
891 for (j = 0; j < 64; j++) {
892 k = (((j >> 0) & 01) << 5) | (((j >> 1) & 01) << 3) |
893 (((j >> 2) & 01) << 2) | (((j >> 3) & 01) << 1) |
894 (((j >> 4) & 01) << 0) | (((j >> 5) & 01) << 4);
896 k = (((k >> 3) & 01) << 0) | (((k >> 2) & 01) << 1) |
897 (((k >> 1) & 01) << 2) | (((k >> 0) & 01) << 3);
898 for (i = 0; i < 32; i++)
900 for (i = 0; i < 4; i++)
901 tmp32[4 * tableno + i] = (k >> i) & 01;
903 for (i = 24; --i >= 0;)
904 k = (k << 1) | tmp32[perm[i] - 1];
905 TO_SIX_BIT(SPE[0][tableno][j], k);
907 for (i = 24; --i >= 0;)
908 k = (k << 1) | tmp32[perm[i + 24] - 1];
909 TO_SIX_BIT(SPE[1][tableno][j], k);
915 * Initialize "perm" to represent transformation "p", which rearranges
916 * (perhaps with expansion and/or contraction) one packed array of bits
917 * (of size "chars_in" characters) into another array (of size "chars_out"
920 * "perm" must be all-zeroes on entry to this routine.
923 init_perm(perm, p, chars_in, chars_out)
924 C_block perm[64 / CHUNKBITS][1 << CHUNKBITS];
926 int chars_in, chars_out;
928 register int i, j, k, l;
930 for (k = 0; k < chars_out * 8; k++) { /* each output bit position */
931 l = p[k] - 1; /* where this bit comes from */
933 continue; /* output bit is always 0 */
934 i = l >> LGCHUNKBITS; /* which chunk this bit comes from */
935 l = 1 << (l & (CHUNKBITS - 1)); /* mask for this bit */
936 for (j = 0; j < (1 << CHUNKBITS); j++) { /* each chunk value */
938 perm[i][j].b[k >> 3] |= 1 << (k & 07);
944 * "setkey" routine (for backwards compatibility)
946 #if 0 /* static and doesn't appear to be referenced */
949 register const char *key;
951 register int i, j, k;
954 for (i = 0; i < 8; i++) {
956 for (j = 0; j < 8; j++) {
958 k |= (unsigned char)*key++;
962 return (des_setkey((char *)keyblock.b));
967 * "encrypt" routine (for backwards compatibility)
971 register char *block;
974 register int i, j, k;
977 for (i = 0; i < 8; i++) {
979 for (j = 0; j < 8; j++) {
981 k |= (unsigned char)*block++;
985 if (des_cipher((char *)&cblock, (char *)&cblock, 0L, (flag ? -1 : 1)))
987 for (i = 7; i >= 0; i--) {
989 for (j = 7; j >= 0; j--) {
999 prtab(s, t, num_rows)
1006 (void)printf("%s:\n", s);
1007 for (i = 0; i < num_rows; i++) {
1008 for (j = 0; j < 8; j++) {
1009 (void)printf("%3d", t[i * 8 + j]);