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>
45 #ifdef AFS_PTHREAD_ENV
47 #endif /* AFS_PTHREAD_ENV */
49 #include "des_internal.h"
56 #include "des_prototypes.h"
61 #define DBG_PRINT(s) if (des_debug & 2) \
62 des_debug_print(s,i,L1&0xffff,(L1>>16)&0xffff, \
63 R1&0xffff,(R1>>16)&0xffff)
68 #ifdef AFS_PTHREAD_ENV
69 pthread_mutex_t rxkad_stats_mutex;
70 #endif /* AFS_PTHREAD_ENV */
72 /* encrypt == 0 ==> decrypt, else encrypt */
75 des_ecb_encrypt(void * clear, void * cipher,
76 register des_key_schedule schedule, int encrypt)
78 /* better pass 8 bytes, length not checked here */
80 register afs_uint32 R1 = 0;
81 register afs_uint32 L1 = 0; /* R1 = r10, L1 = r9 */
82 register afs_uint32 R2 = 0, L2 = 0; /* R2 = r8, L2 = r7 */
84 /* one more registers left on VAX, see below P_temp_p */
86 sbox_in_16_a S_in_16_a;
87 sbox_in_16_b S_in_16_b;
88 sbox_in_16_c S_in_16_c;
89 unsigned int *S_in_a_16_p = (unsigned int *)&S_in_16_a;
90 unsigned int *S_in_b_16_p = (unsigned int *)&S_in_16_b;
91 unsigned int *S_in_c_16_p = (unsigned int *)&S_in_16_c;
95 #error dunno how to do this machine type, you lose;
99 register unsigned char *P_temp_p = (unsigned char *)&P_temp;
102 afs_uint32 *S_out_p = (afs_uint32 *) & S_out;
104 afs_uint32 R_save, L_save;
106 afs_uint32 dbg_tmp[2];
110 rxkad_stats.des_encrypts[DES_ENCRYPT]++;
112 rxkad_stats.des_encrypts[DES_DECRYPT]++;
115 * Use L1,R1 and L2,R2 as two sets of "64-bit" registers always
116 * work from L1,R1 input to L2,R2 output; initialize the cleartext
122 * If the alignment is wrong, the programmer really screwed up --
123 * we aren't even getting the right data type. His problem. Keep
124 * this code for debugging.
126 /* Make sure schedule is ok */
127 if ((afs_int32) schedule & 3) {
128 fprintf(stderr, "des.c schedule arg pointer not aligned\n");
132 if ((afs_uint32) clear & 3) {
133 memcpy((char *)(&L_save), (char *)clear, sizeof(L_save));
134 clear=((afs_uint32*)clear)+1;
135 memcpy((char *)(&R_save), (char *)clear, sizeof(R_save));
142 L1 = *((afs_uint32 *)clear);
143 clear=((afs_uint32*)clear)+1;
147 R1 = *((afs_uint32 *)clear);
154 printf("All values printed from low byte (bit 0)");
155 printf(" --> high byte (bit 63)\n");
159 printf("iter = %2d before IP\n\t\tL1 R1 = ", i);
160 des_cblock_print_file(dbg_tmp, stdout);
163 DBG_PRINT("before IP");
168 /* all the Initial Permutation code is in the include file */
170 /* reset input to L1,R1 */
174 /* iterate through the inner loop */
175 for (i = 0; i <= (AUTH_DES_ITER - 1); i++) {
181 printf("iter = %2d start loop\n\t\tL1 R1 = ", i);
182 des_cblock_print_file(dbg_tmp, stdout);
183 DBG_PRINT("start loop");
191 /* apply the E permutation from R1 to L2, R2 */
195 #else /* Bill's fast E */
200 L2 |= (R1 << 3) & 07700;
201 L2 |= (R1 << 5) & 0770000;
202 L2 |= (R1 << 7) & 077000000;
203 L2 |= (R1 << 9) & 07700000000;
204 L2 |= (R1 << 11) & 030000000000;
206 /* now from right to right */
208 R2 = ((R1 >> 17) & 0176000);
212 R2 |= ((R1 >> 21) & 017);
213 R2 |= ((R1 >> 19) & 01760);
218 asm(" rotl $1,r10,r7");
220 L2 |= (R1 << 3) & 07700;
221 L2 |= (R1 << 5) & 0770000;
222 L2 |= (R1 << 7) & 077000000;
223 L2 |= (R1 << 9) & 07700000000;
224 L2 |= (R1 << 11) & 030000000000;
226 asm(" rotl $-17,r10,r8");
228 asm(" rotl $-21,r10,r0");
229 asm(" bicl2 $-16,r0");
231 asm(" rotl $-19,r10,r0");
232 asm(" bicl2 $-1009,r0");
237 /* reset input to L1,R1 */
245 DBG_PRINT("after e");
246 printf("iter = %2d after e\n\t\tL1 R1 = ", i);
247 des_cblock_print_file(dbg_tmp, stdout);
253 * XOR with the key schedule, "schedule"
255 * If this is an encryption operation, use schedule[i],
256 * otherwise use schedule [AUTH_DES_ITER-i-1]
258 * First XOR left half.
261 L1 ^= *(((afs_uint32 *) & schedule[i]) + 0);
263 R1 ^= *(((afs_uint32 *) & schedule[i]) + 1);
265 L1 ^= *(((afs_uint32 *) & schedule[AUTH_DES_ITER - i - 1]) + 0);
267 R1 ^= *(((afs_uint32 *) & schedule[AUTH_DES_ITER - i - 1]) + 1);
270 /* dont have to reset input to L1, R1 */
276 DBG_PRINT("after xor");
277 printf("iter = %2d after xor\n\t\tL1 R1 =", i);
278 des_cblock_print_file(dbg_tmp, stdout);
283 /* apply the S selection from L1, R1 to R2 */
289 /* S operations , cant use registers for bit field stuff */
290 /* from S_in to S_out */
293 *S_in_a_16_p = L1 & 0xffff;
294 *S_in_b_16_p = (L1 >> 16) & 0xffff;
295 *S_in_c_16_p = R1 & 0xffff;
296 (*(afs_uint32 *) & S_out) = (unsigned)S_adj[0][S_in_16_a.b0];
297 S_out.b1 = (unsigned)S_adj[1][S_in_16_a.b1];
298 /* b2 spans two words */
299 S_out.b2 = (unsigned)
300 S_adj[2][(unsigned)S_in_16_a.b2 +
301 (((unsigned)S_in_16_b.b2) << 4)];
302 S_out.b3 = (unsigned)S_adj[3][S_in_16_b.b3];
303 S_out.b4 = (unsigned)S_adj[4][S_in_16_b.b4];
304 /* b5 spans both parts */
305 S_out.b5 = (unsigned)
306 S_adj[5][(unsigned)S_in_16_b.b5 +
307 (((unsigned)S_in_16_c.b5) << 2)];
308 S_out.b6 = (unsigned)S_adj[6][S_in_16_c.b6];
309 S_out.b7 = (unsigned)S_adj[7][S_in_16_c.b7];
312 /* is a 32 bit sys */
314 R2 = (unsigned)S_adj[0][L1 & 077];
315 L2 = (unsigned)S_adj[1][(L1 >> 6) & 077];
317 L2 = (unsigned)S_adj[2][(L1 >> 12) & 077];
319 L2 = (unsigned)S_adj[3][(L1 >> 18) & 077];
321 L2 = (unsigned)S_adj[4][(L1 >> 24) & 077];
323 /* b5 spans both parts */
325 S_adj[5][(unsigned)((L1 >> 30) & 03) + ((R1 & 017) << 2)];
327 L2 = (unsigned)S_adj[6][(R1 >> 4) & 077];
329 L2 = (unsigned)S_adj[7][(R1 >> 10) & 077];
330 R1 = R2 | (L2 << 28);
331 /* reset input to L1, R1 */
334 * this is the c code produced above, with
335 * extzv replaced by rotl
337 asm("bicl3 $-64,r9,r0");
338 asm("movzbl _S_adj[r0],r8");
339 asm("rotl $-6,r9,r0");
340 asm("bicl2 $-64,r0");
341 asm("movzbl _S_adj+64[r0],r7");
342 asm("ashl $4,r7,r0");
344 asm("rotl $-12,r9,r0");
345 asm("bicl2 $-64,r0");
346 asm("movzbl _S_adj+128[r0],r7");
347 asm("ashl $8,r7,r0");
349 asm("rotl $-18,r9,r0");
350 asm("bicl2 $-64,r0");
351 asm("movzbl _S_adj+192[r0],r7");
352 asm("ashl $12,r7,r0");
354 asm("rotl $-24,r9,r0");
355 asm("bicl2 $-64,r0");
356 asm("movzbl _S_adj+256[r0],r7");
357 asm("ashl $16,r7,r0");
359 asm("rotl $-30,r9,r0");
361 asm("bicl3 $-16,r10,r1");
362 asm("ashl $2,r1,r1");
364 asm("movzbl _S_adj+320[r0],r7");
365 asm("ashl $20,r7,r0");
367 asm("rotl $-4,r10,r0");
368 asm("bicl2 $-64,r0");
369 asm("movzbl _S_adj+384[r0],r7");
370 asm("ashl $24,r7,r0");
372 asm("rotl $-10,r10,r0");
373 asm("bicl2 $-64,r0");
374 asm("movzbl _S_adj+448[r0],r7");
375 asm("ashl $28,r7,r0");
386 DBG_PRINT("after s");
387 printf("iter = %2d after s\n\t\tL1 R1 = ", i);
388 des_cblock_print_file(dbg_tmp, stdout);
393 /* and then the p permutation from R1 into R2 */
395 /* reset the input to L1, R1 */
402 DBG_PRINT("after p");
403 printf("iter = %2d after p\n\t\tL1 R1 = ", i);
404 des_cblock_print_file(dbg_tmp, stdout);
408 /* R1 is the output value from the f() */
409 /* move R[iter] to L[iter+1] */
414 /* reset the input */
417 /* flip left and right before final permutation */
420 /* reset the input */
428 DBG_PRINT("before FP");
429 printf("iter = %2d before FP\n\t\tL1 R1 = ", i);
430 des_cblock_print_file(dbg_tmp, stdout);
435 /* do the final permutation from L1R1 to L2R2 */
436 /* all the fp code is in the include file */
439 /* copy the output to the ciphertext string;
440 * can be same as cleartext
444 if ((afs_int32) cipher & 3) {
445 L_save = L2; /* cant bcopy a reg */
447 cipher=((afs_uint32*)cipher)+1;
448 memcpy((char *)cipher, (char *)&L_save, sizeof(L_save));
449 memcpy((char *)cipher, (char *)&R_save, sizeof(R_save));
453 *((afs_uint32*)cipher)= L2;
454 cipher = ((afs_int32 *)cipher)+1;
455 *((afs_uint32 *)cipher) = R2;
465 printf("iter = %2d done\n\t\tL1 R1 = ", i);
466 des_cblock_print_file(dbg_tmp, stdout);
470 /* that's it, no errors can be returned */