2 * Copyright 1985, 1986, 1987, 1988 by the Massachusetts Institute
5 * For copying and distribution information, please see the file
8 * These routines perform encryption and decryption using the DES
9 * private key algorithm, or else a subset of it-- fewer inner loops.
10 * (AUTH_DES_ITER defaults to 16, may be less.)
12 * Under U.S. law, this software may not be exported outside the US
13 * without license from the U.S. Commerce department.
15 * The key schedule is passed as an arg, as well as the cleartext or
18 * All registers labeled imply Vax using the Ultrix or 4.2bsd
22 * NOTE: bit and byte numbering:
23 * DES algorithm is defined in terms of bits of L
24 * followed by bits of R.
29 * Always work in register pairs, FROM L1,R1 TO L2,R2 to make
32 * originally written by Steve Miller, MIT Project Athena
34 #include <mit-cpyright.h>
36 #include <afsconfig.h>
37 #include <afs/param.h>
44 #ifdef AFS_PTHREAD_ENV
46 #endif /* AFS_PTHREAD_ENV */
48 #include "des_internal.h"
55 #include "des_prototypes.h"
60 #define DBG_PRINT(s) if (des_debug & 2) \
61 des_debug_print(s,i,L1&0xffff,(L1>>16)&0xffff, \
62 R1&0xffff,(R1>>16)&0xffff)
67 #ifdef AFS_PTHREAD_ENV
68 pthread_mutex_t rxkad_stats_mutex;
69 #endif /* AFS_PTHREAD_ENV */
71 /* encrypt == 0 ==> decrypt, else encrypt */
73 afs_int32 des_ecb_encrypt(afs_uint32 *clear, afs_uint32 *cipher,
74 register des_key_schedule schedule, int encrypt)
76 /* better pass 8 bytes, length not checked here */
78 register afs_uint32 R1, L1; /* R1 = r10, L1 = r9 */
79 register afs_uint32 R2, L2; /* R2 = r8, L2 = r7 */
81 /* one more registers left on VAX, see below P_temp_p */
83 sbox_in_16_a S_in_16_a;
84 sbox_in_16_b S_in_16_b;
85 sbox_in_16_c S_in_16_c;
86 unsigned int *S_in_a_16_p = (unsigned int *) &S_in_16_a;
87 unsigned int *S_in_b_16_p = (unsigned int *) &S_in_16_b;
88 unsigned int *S_in_c_16_p = (unsigned int *) &S_in_16_c;
92 dunno how to do this machine type, you lose;
96 register unsigned char *P_temp_p = (unsigned char *) & P_temp;
99 afs_uint32 *S_out_p = (afs_uint32 *) &S_out;
101 afs_uint32 R_save, L_save;
103 afs_uint32 dbg_tmp[2];
106 if (encrypt) rxkad_stats.des_encrypts[DES_ENCRYPT]++;
107 else rxkad_stats.des_encrypts[DES_DECRYPT]++;
111 * Use L1,R1 and L2,R2 as two sets of "64-bit" registers always
112 * work from L1,R1 input to L2,R2 output; initialize the cleartext
118 * If the alignment is wrong, the programmer really screwed up --
119 * we aren't even getting the right data type. His problem. Keep
120 * this code for debugging.
122 /* Make sure schedule is ok */
123 if ((afs_int32) schedule & 3) {
124 fprintf(stderr,"des.c schedule arg pointer not aligned\n");
128 if ((afs_int32) clear & 3) {
129 memcpy((char *)&L_save, (char *)clear++, sizeof(L_save));
130 memcpy((char *)&R_save, (char *)clear, sizeof(R_save));
137 if (clear) L1 = *clear++;
139 if (clear) R1 = *clear;
145 printf("All values printed from low byte (bit 0)");
146 printf(" --> high byte (bit 63)\n");
150 printf("iter = %2d before IP\n\t\tL1 R1 = ",i);
151 des_cblock_print_file (dbg_tmp, stdout);
154 DBG_PRINT("before IP");
159 /* all the Initial Permutation code is in the include file */
161 /* reset input to L1,R1 */
165 /* iterate through the inner loop */
166 for (i = 0; i <= (AUTH_DES_ITER-1); i++) {
172 printf("iter = %2d start loop\n\t\tL1 R1 = ",i);
173 des_cblock_print_file (dbg_tmp, stdout);
174 DBG_PRINT("start loop");
183 /* apply the E permutation from R1 to L2, R2 */
187 #else /* Bill's fast E */
192 L2 |= (R1 <<3) & 07700;
193 L2 |= (R1 <<5) & 0770000;
194 L2 |= (R1 <<7) & 077000000;
195 L2 |= (R1 <<9) & 07700000000;
196 L2 |= (R1 <<11) & 030000000000;
198 /* now from right to right */
200 R2 = ((R1 >> 17) & 0176000);
201 if (R1 & (1<<0)) R2 |= 1<<15;
203 R2 |= ((R1 >> 21) & 017);
204 R2 |= ((R1 >> 19) & 01760);
209 asm(" rotl $1,r10,r7");
211 L2 |= (R1 <<3) & 07700;
212 L2 |= (R1 <<5) & 0770000;
213 L2 |= (R1 <<7) & 077000000;
214 L2 |= (R1 <<9) & 07700000000;
215 L2 |= (R1 <<11) & 030000000000;
217 asm(" rotl $-17,r10,r8");
219 asm(" rotl $-21,r10,r0");
220 asm(" bicl2 $-16,r0");
222 asm(" rotl $-19,r10,r0");
223 asm(" bicl2 $-1009,r0");
228 /* reset input to L1,R1 */
236 DBG_PRINT("after e");
237 printf("iter = %2d after e\n\t\tL1 R1 = ",i);
238 des_cblock_print_file (dbg_tmp, stdout);
244 * XOR with the key schedule, "schedule"
246 * If this is an encryption operation, use schedule[i],
247 * otherwise use schedule [AUTH_DES_ITER-i-1]
249 * First XOR left half.
252 L1 ^= *(((afs_uint32 *) &schedule[i] )+0);
254 R1 ^= *(((afs_uint32 *) &schedule[i] )+1);
257 L1 ^= *(((afs_uint32 *) &schedule[AUTH_DES_ITER-i-1] )+0);
259 R1 ^= *(((afs_uint32 *) &schedule[AUTH_DES_ITER-i-1] )+1);
262 /* dont have to reset input to L1, R1 */
268 DBG_PRINT("after xor");
269 printf("iter = %2d after xor\n\t\tL1 R1 =",i);
270 des_cblock_print_file (dbg_tmp, stdout);
275 /* apply the S selection from L1, R1 to R2 */
281 /* S operations , cant use registers for bit field stuff */
282 /* from S_in to S_out */
285 *S_in_a_16_p = L1&0xffff;
286 *S_in_b_16_p = (L1>>16)&0xffff;
287 *S_in_c_16_p = R1&0xffff;
288 (*(afs_uint32 *) &S_out) =
289 (unsigned) S_adj[0][S_in_16_a.b0];
290 S_out.b1 = (unsigned) S_adj[1][S_in_16_a.b1];
291 /* b2 spans two words */
292 S_out.b2 = (unsigned)
293 S_adj[2][(unsigned) S_in_16_a.b2
294 + (((unsigned) S_in_16_b.b2) << 4)];
295 S_out.b3 = (unsigned) S_adj[3][S_in_16_b.b3];
296 S_out.b4 = (unsigned) S_adj[4][S_in_16_b.b4];
297 /* b5 spans both parts */
298 S_out.b5 = (unsigned)
299 S_adj[5][(unsigned) S_in_16_b.b5
300 + (((unsigned) S_in_16_c.b5) << 2)];
301 S_out.b6 = (unsigned) S_adj[6][S_in_16_c.b6];
302 S_out.b7 = (unsigned) S_adj[7][S_in_16_c.b7];
305 /* is a 32 bit sys */
307 R2 = (unsigned) S_adj[0][L1 & 077];
308 L2 = (unsigned) S_adj[1][(L1 >> 6) & 077];
310 L2 = (unsigned) S_adj[2][(L1 >> 12) & 077];
312 L2 = (unsigned) S_adj[3][(L1 >> 18) & 077];
314 L2 = (unsigned) S_adj[4][(L1 >> 24) & 077];
316 /* b5 spans both parts */
318 S_adj[5][(unsigned) ((L1 >>30) & 03) + ((R1 & 017) << 2)];
320 L2 = (unsigned) S_adj[6][(R1 >> 4) & 077];
322 L2 = (unsigned) S_adj[7][(R1 >> 10) & 077];
324 /* reset input to L1, R1 */
327 * this is the c code produced above, with
328 * extzv replaced by rotl
330 asm("bicl3 $-64,r9,r0");
331 asm("movzbl _S_adj[r0],r8");
332 asm("rotl $-6,r9,r0");
333 asm("bicl2 $-64,r0");
334 asm("movzbl _S_adj+64[r0],r7");
335 asm("ashl $4,r7,r0");
337 asm("rotl $-12,r9,r0");
338 asm("bicl2 $-64,r0");
339 asm("movzbl _S_adj+128[r0],r7");
340 asm("ashl $8,r7,r0");
342 asm("rotl $-18,r9,r0");
343 asm("bicl2 $-64,r0");
344 asm("movzbl _S_adj+192[r0],r7");
345 asm("ashl $12,r7,r0");
347 asm("rotl $-24,r9,r0");
348 asm("bicl2 $-64,r0");
349 asm("movzbl _S_adj+256[r0],r7");
350 asm("ashl $16,r7,r0");
352 asm("rotl $-30,r9,r0");
354 asm("bicl3 $-16,r10,r1");
355 asm("ashl $2,r1,r1");
357 asm("movzbl _S_adj+320[r0],r7");
358 asm("ashl $20,r7,r0");
360 asm("rotl $-4,r10,r0");
361 asm("bicl2 $-64,r0");
362 asm("movzbl _S_adj+384[r0],r7");
363 asm("ashl $24,r7,r0");
365 asm("rotl $-10,r10,r0");
366 asm("bicl2 $-64,r0");
367 asm("movzbl _S_adj+448[r0],r7");
368 asm("ashl $28,r7,r0");
379 DBG_PRINT("after s");
380 printf("iter = %2d after s\n\t\tL1 R1 = ",i);
381 des_cblock_print_file (dbg_tmp, stdout);
386 /* and then the p permutation from R1 into R2 */
388 /* reset the input to L1, R1 */
395 DBG_PRINT("after p");
396 printf("iter = %2d after p\n\t\tL1 R1 = ",i);
397 des_cblock_print_file (dbg_tmp, stdout);
401 /* R1 is the output value from the f() */
402 /* move R[iter] to L[iter+1] */
407 /* reset the input */
410 /* flip left and right before final permutation */
413 /* reset the input */
421 DBG_PRINT("before FP");
422 printf("iter = %2d before FP\n\t\tL1 R1 = ",i);
423 des_cblock_print_file (dbg_tmp, stdout);
429 /* do the final permutation from L1R1 to L2R2 */
430 /* all the fp code is in the include file */
433 /* copy the output to the ciphertext string;
434 * can be same as cleartext
438 if ((afs_int32) cipher & 3) {
439 L_save = L2; /* cant bcopy a reg */
441 memcpy((char *)cipher++, (char *)&L_save, sizeof(L_save));
442 memcpy((char *)cipher, (char *)&R_save, sizeof(R_save));
458 printf("iter = %2d done\n\t\tL1 R1 = ",i);
459 des_cblock_print_file (dbg_tmp, stdout);
463 /* that's it, no errors can be returned */