Annotation of src/usr.bin/ssh/rijndael.c, Revision 1.2.4.1
1.2.4.1 ! jason 1: /* $OpenBSD: rijndael.c,v 1.2 2000/10/15 14:14:01 markus Exp $ */
1.1 markus 2:
3: /* This is an independent implementation of the encryption algorithm: */
4: /* */
5: /* RIJNDAEL by Joan Daemen and Vincent Rijmen */
6: /* */
7: /* which is a candidate algorithm in the Advanced Encryption Standard */
8: /* programme of the US National Institute of Standards and Technology. */
9: /* */
10: /* Copyright in this implementation is held by Dr B R Gladman but I */
11: /* hereby give permission for its free direct or derivative use subject */
12: /* to acknowledgment of its origin and compliance with any conditions */
13: /* that the originators of the algorithm place on its exploitation. */
14: /* */
15: /* Dr Brian Gladman (gladman@seven77.demon.co.uk) 14th January 1999 */
16:
17: /* Timing data for Rijndael (rijndael.c)
18:
19: Algorithm: rijndael (rijndael.c)
20:
21: 128 bit key:
22: Key Setup: 305/1389 cycles (encrypt/decrypt)
23: Encrypt: 374 cycles = 68.4 mbits/sec
24: Decrypt: 352 cycles = 72.7 mbits/sec
25: Mean: 363 cycles = 70.5 mbits/sec
26:
27: 192 bit key:
28: Key Setup: 277/1595 cycles (encrypt/decrypt)
29: Encrypt: 439 cycles = 58.3 mbits/sec
30: Decrypt: 425 cycles = 60.2 mbits/sec
31: Mean: 432 cycles = 59.3 mbits/sec
32:
33: 256 bit key:
34: Key Setup: 374/1960 cycles (encrypt/decrypt)
35: Encrypt: 502 cycles = 51.0 mbits/sec
36: Decrypt: 498 cycles = 51.4 mbits/sec
37: Mean: 500 cycles = 51.2 mbits/sec
38:
39: */
40:
41: #include <sys/types.h>
42: #include "rijndael.h"
43:
44: void gen_tabs __P((void));
45:
46: /* 3. Basic macros for speeding up generic operations */
47:
48: /* Circular rotate of 32 bit values */
49:
50: #define rotr(x,n) (((x) >> ((int)(n))) | ((x) << (32 - (int)(n))))
51: #define rotl(x,n) (((x) << ((int)(n))) | ((x) >> (32 - (int)(n))))
52:
53: /* Invert byte order in a 32 bit variable */
54:
55: #define bswap(x) (rotl(x, 8) & 0x00ff00ff | rotr(x, 8) & 0xff00ff00)
56:
57: /* Extract byte from a 32 bit quantity (little endian notation) */
58:
59: #define byte(x,n) ((u1byte)((x) >> (8 * n)))
60:
61: #if BYTE_ORDER != LITTLE_ENDIAN
62: #define BLOCK_SWAP
63: #endif
64:
65: /* For inverting byte order in input/output 32 bit words if needed */
66:
67: #ifdef BLOCK_SWAP
68: #define BYTE_SWAP
69: #define WORD_SWAP
70: #endif
71:
72: #ifdef BYTE_SWAP
73: #define io_swap(x) bswap(x)
74: #else
75: #define io_swap(x) (x)
76: #endif
77:
78: /* For inverting the byte order of input/output blocks if needed */
79:
80: #ifdef WORD_SWAP
81:
82: #define get_block(x) \
83: ((u4byte*)(x))[0] = io_swap(in_blk[3]); \
84: ((u4byte*)(x))[1] = io_swap(in_blk[2]); \
85: ((u4byte*)(x))[2] = io_swap(in_blk[1]); \
86: ((u4byte*)(x))[3] = io_swap(in_blk[0])
87:
88: #define put_block(x) \
89: out_blk[3] = io_swap(((u4byte*)(x))[0]); \
90: out_blk[2] = io_swap(((u4byte*)(x))[1]); \
91: out_blk[1] = io_swap(((u4byte*)(x))[2]); \
92: out_blk[0] = io_swap(((u4byte*)(x))[3])
93:
94: #define get_key(x,len) \
95: ((u4byte*)(x))[4] = ((u4byte*)(x))[5] = \
96: ((u4byte*)(x))[6] = ((u4byte*)(x))[7] = 0; \
97: switch((((len) + 63) / 64)) { \
98: case 2: \
99: ((u4byte*)(x))[0] = io_swap(in_key[3]); \
100: ((u4byte*)(x))[1] = io_swap(in_key[2]); \
101: ((u4byte*)(x))[2] = io_swap(in_key[1]); \
102: ((u4byte*)(x))[3] = io_swap(in_key[0]); \
103: break; \
104: case 3: \
105: ((u4byte*)(x))[0] = io_swap(in_key[5]); \
106: ((u4byte*)(x))[1] = io_swap(in_key[4]); \
107: ((u4byte*)(x))[2] = io_swap(in_key[3]); \
108: ((u4byte*)(x))[3] = io_swap(in_key[2]); \
109: ((u4byte*)(x))[4] = io_swap(in_key[1]); \
110: ((u4byte*)(x))[5] = io_swap(in_key[0]); \
111: break; \
112: case 4: \
113: ((u4byte*)(x))[0] = io_swap(in_key[7]); \
114: ((u4byte*)(x))[1] = io_swap(in_key[6]); \
115: ((u4byte*)(x))[2] = io_swap(in_key[5]); \
116: ((u4byte*)(x))[3] = io_swap(in_key[4]); \
117: ((u4byte*)(x))[4] = io_swap(in_key[3]); \
118: ((u4byte*)(x))[5] = io_swap(in_key[2]); \
119: ((u4byte*)(x))[6] = io_swap(in_key[1]); \
120: ((u4byte*)(x))[7] = io_swap(in_key[0]); \
121: }
122:
123: #else
124:
125: #define get_block(x) \
126: ((u4byte*)(x))[0] = io_swap(in_blk[0]); \
127: ((u4byte*)(x))[1] = io_swap(in_blk[1]); \
128: ((u4byte*)(x))[2] = io_swap(in_blk[2]); \
129: ((u4byte*)(x))[3] = io_swap(in_blk[3])
130:
131: #define put_block(x) \
132: out_blk[0] = io_swap(((u4byte*)(x))[0]); \
133: out_blk[1] = io_swap(((u4byte*)(x))[1]); \
134: out_blk[2] = io_swap(((u4byte*)(x))[2]); \
135: out_blk[3] = io_swap(((u4byte*)(x))[3])
136:
137: #define get_key(x,len) \
138: ((u4byte*)(x))[4] = ((u4byte*)(x))[5] = \
139: ((u4byte*)(x))[6] = ((u4byte*)(x))[7] = 0; \
140: switch((((len) + 63) / 64)) { \
141: case 4: \
142: ((u4byte*)(x))[6] = io_swap(in_key[6]); \
143: ((u4byte*)(x))[7] = io_swap(in_key[7]); \
144: case 3: \
145: ((u4byte*)(x))[4] = io_swap(in_key[4]); \
146: ((u4byte*)(x))[5] = io_swap(in_key[5]); \
147: case 2: \
148: ((u4byte*)(x))[0] = io_swap(in_key[0]); \
149: ((u4byte*)(x))[1] = io_swap(in_key[1]); \
150: ((u4byte*)(x))[2] = io_swap(in_key[2]); \
151: ((u4byte*)(x))[3] = io_swap(in_key[3]); \
152: }
153:
154: #endif
155:
156: #define LARGE_TABLES
157:
158: u1byte pow_tab[256];
159: u1byte log_tab[256];
160: u1byte sbx_tab[256];
161: u1byte isb_tab[256];
162: u4byte rco_tab[ 10];
163: u4byte ft_tab[4][256];
164: u4byte it_tab[4][256];
165:
166: #ifdef LARGE_TABLES
167: u4byte fl_tab[4][256];
168: u4byte il_tab[4][256];
169: #endif
170:
171: u4byte tab_gen = 0;
172:
173: #define ff_mult(a,b) (a && b ? pow_tab[(log_tab[a] + log_tab[b]) % 255] : 0)
174:
175: #define f_rn(bo, bi, n, k) \
176: bo[n] = ft_tab[0][byte(bi[n],0)] ^ \
177: ft_tab[1][byte(bi[(n + 1) & 3],1)] ^ \
178: ft_tab[2][byte(bi[(n + 2) & 3],2)] ^ \
179: ft_tab[3][byte(bi[(n + 3) & 3],3)] ^ *(k + n)
180:
181: #define i_rn(bo, bi, n, k) \
182: bo[n] = it_tab[0][byte(bi[n],0)] ^ \
183: it_tab[1][byte(bi[(n + 3) & 3],1)] ^ \
184: it_tab[2][byte(bi[(n + 2) & 3],2)] ^ \
185: it_tab[3][byte(bi[(n + 1) & 3],3)] ^ *(k + n)
186:
187: #ifdef LARGE_TABLES
188:
189: #define ls_box(x) \
190: ( fl_tab[0][byte(x, 0)] ^ \
191: fl_tab[1][byte(x, 1)] ^ \
192: fl_tab[2][byte(x, 2)] ^ \
193: fl_tab[3][byte(x, 3)] )
194:
195: #define f_rl(bo, bi, n, k) \
196: bo[n] = fl_tab[0][byte(bi[n],0)] ^ \
197: fl_tab[1][byte(bi[(n + 1) & 3],1)] ^ \
198: fl_tab[2][byte(bi[(n + 2) & 3],2)] ^ \
199: fl_tab[3][byte(bi[(n + 3) & 3],3)] ^ *(k + n)
200:
201: #define i_rl(bo, bi, n, k) \
202: bo[n] = il_tab[0][byte(bi[n],0)] ^ \
203: il_tab[1][byte(bi[(n + 3) & 3],1)] ^ \
204: il_tab[2][byte(bi[(n + 2) & 3],2)] ^ \
205: il_tab[3][byte(bi[(n + 1) & 3],3)] ^ *(k + n)
206:
207: #else
208:
209: #define ls_box(x) \
210: ((u4byte)sbx_tab[byte(x, 0)] << 0) ^ \
211: ((u4byte)sbx_tab[byte(x, 1)] << 8) ^ \
212: ((u4byte)sbx_tab[byte(x, 2)] << 16) ^ \
213: ((u4byte)sbx_tab[byte(x, 3)] << 24)
214:
215: #define f_rl(bo, bi, n, k) \
216: bo[n] = (u4byte)sbx_tab[byte(bi[n],0)] ^ \
217: rotl(((u4byte)sbx_tab[byte(bi[(n + 1) & 3],1)]), 8) ^ \
218: rotl(((u4byte)sbx_tab[byte(bi[(n + 2) & 3],2)]), 16) ^ \
219: rotl(((u4byte)sbx_tab[byte(bi[(n + 3) & 3],3)]), 24) ^ *(k + n)
220:
221: #define i_rl(bo, bi, n, k) \
222: bo[n] = (u4byte)isb_tab[byte(bi[n],0)] ^ \
223: rotl(((u4byte)isb_tab[byte(bi[(n + 3) & 3],1)]), 8) ^ \
224: rotl(((u4byte)isb_tab[byte(bi[(n + 2) & 3],2)]), 16) ^ \
225: rotl(((u4byte)isb_tab[byte(bi[(n + 1) & 3],3)]), 24) ^ *(k + n)
226:
227: #endif
228:
229: void
230: gen_tabs(void)
231: {
232: u4byte i, t;
233: u1byte p, q;
234:
235: /* log and power tables for GF(2**8) finite field with */
236: /* 0x11b as modular polynomial - the simplest prmitive */
237: /* root is 0x11, used here to generate the tables */
238:
239: for(i = 0,p = 1; i < 256; ++i) {
240: pow_tab[i] = (u1byte)p; log_tab[p] = (u1byte)i;
241:
242: p = p ^ (p << 1) ^ (p & 0x80 ? 0x01b : 0);
243: }
244:
245: log_tab[1] = 0; p = 1;
246:
247: for(i = 0; i < 10; ++i) {
248: rco_tab[i] = p;
249:
250: p = (p << 1) ^ (p & 0x80 ? 0x1b : 0);
251: }
252:
253: /* note that the affine byte transformation matrix in */
254: /* rijndael specification is in big endian format with */
255: /* bit 0 as the most significant bit. In the remainder */
256: /* of the specification the bits are numbered from the */
257: /* least significant end of a byte. */
258:
259: for(i = 0; i < 256; ++i) {
260: p = (i ? pow_tab[255 - log_tab[i]] : 0); q = p;
261: q = (q >> 7) | (q << 1); p ^= q;
262: q = (q >> 7) | (q << 1); p ^= q;
263: q = (q >> 7) | (q << 1); p ^= q;
264: q = (q >> 7) | (q << 1); p ^= q ^ 0x63;
265: sbx_tab[i] = (u1byte)p; isb_tab[p] = (u1byte)i;
266: }
267:
268: for(i = 0; i < 256; ++i) {
269: p = sbx_tab[i];
270:
271: #ifdef LARGE_TABLES
272:
273: t = p; fl_tab[0][i] = t;
274: fl_tab[1][i] = rotl(t, 8);
275: fl_tab[2][i] = rotl(t, 16);
276: fl_tab[3][i] = rotl(t, 24);
277: #endif
278: t = ((u4byte)ff_mult(2, p)) |
279: ((u4byte)p << 8) |
280: ((u4byte)p << 16) |
281: ((u4byte)ff_mult(3, p) << 24);
282:
283: ft_tab[0][i] = t;
284: ft_tab[1][i] = rotl(t, 8);
285: ft_tab[2][i] = rotl(t, 16);
286: ft_tab[3][i] = rotl(t, 24);
287:
288: p = isb_tab[i];
289:
290: #ifdef LARGE_TABLES
291:
292: t = p; il_tab[0][i] = t;
293: il_tab[1][i] = rotl(t, 8);
294: il_tab[2][i] = rotl(t, 16);
295: il_tab[3][i] = rotl(t, 24);
296: #endif
297: t = ((u4byte)ff_mult(14, p)) |
298: ((u4byte)ff_mult( 9, p) << 8) |
299: ((u4byte)ff_mult(13, p) << 16) |
300: ((u4byte)ff_mult(11, p) << 24);
301:
302: it_tab[0][i] = t;
303: it_tab[1][i] = rotl(t, 8);
304: it_tab[2][i] = rotl(t, 16);
305: it_tab[3][i] = rotl(t, 24);
306: }
307:
308: tab_gen = 1;
1.2 markus 309: }
1.1 markus 310:
311: #define star_x(x) (((x) & 0x7f7f7f7f) << 1) ^ ((((x) & 0x80808080) >> 7) * 0x1b)
312:
313: #define imix_col(y,x) \
314: u = star_x(x); \
315: v = star_x(u); \
316: w = star_x(v); \
317: t = w ^ (x); \
318: (y) = u ^ v ^ w; \
319: (y) ^= rotr(u ^ t, 8) ^ \
320: rotr(v ^ t, 16) ^ \
321: rotr(t,24)
322:
323: /* initialise the key schedule from the user supplied key */
324:
325: #define loop4(i) \
326: { t = ls_box(rotr(t, 8)) ^ rco_tab[i]; \
327: t ^= e_key[4 * i]; e_key[4 * i + 4] = t; \
328: t ^= e_key[4 * i + 1]; e_key[4 * i + 5] = t; \
329: t ^= e_key[4 * i + 2]; e_key[4 * i + 6] = t; \
330: t ^= e_key[4 * i + 3]; e_key[4 * i + 7] = t; \
331: }
332:
333: #define loop6(i) \
334: { t = ls_box(rotr(t, 8)) ^ rco_tab[i]; \
335: t ^= e_key[6 * i]; e_key[6 * i + 6] = t; \
336: t ^= e_key[6 * i + 1]; e_key[6 * i + 7] = t; \
337: t ^= e_key[6 * i + 2]; e_key[6 * i + 8] = t; \
338: t ^= e_key[6 * i + 3]; e_key[6 * i + 9] = t; \
339: t ^= e_key[6 * i + 4]; e_key[6 * i + 10] = t; \
340: t ^= e_key[6 * i + 5]; e_key[6 * i + 11] = t; \
341: }
342:
343: #define loop8(i) \
344: { t = ls_box(rotr(t, 8)) ^ rco_tab[i]; \
345: t ^= e_key[8 * i]; e_key[8 * i + 8] = t; \
346: t ^= e_key[8 * i + 1]; e_key[8 * i + 9] = t; \
347: t ^= e_key[8 * i + 2]; e_key[8 * i + 10] = t; \
348: t ^= e_key[8 * i + 3]; e_key[8 * i + 11] = t; \
349: t = e_key[8 * i + 4] ^ ls_box(t); \
350: e_key[8 * i + 12] = t; \
351: t ^= e_key[8 * i + 5]; e_key[8 * i + 13] = t; \
352: t ^= e_key[8 * i + 6]; e_key[8 * i + 14] = t; \
353: t ^= e_key[8 * i + 7]; e_key[8 * i + 15] = t; \
354: }
355:
356: rijndael_ctx *
357: rijndael_set_key(rijndael_ctx *ctx, const u4byte *in_key, const u4byte key_len,
358: int encrypt)
359: {
360: u4byte i, t, u, v, w;
361: u4byte *e_key = ctx->e_key;
362: u4byte *d_key = ctx->d_key;
363:
364: ctx->decrypt = !encrypt;
365:
366: if(!tab_gen)
367: gen_tabs();
368:
369: ctx->k_len = (key_len + 31) / 32;
370:
371: e_key[0] = in_key[0]; e_key[1] = in_key[1];
372: e_key[2] = in_key[2]; e_key[3] = in_key[3];
373:
374: switch(ctx->k_len) {
375: case 4: t = e_key[3];
376: for(i = 0; i < 10; ++i)
377: loop4(i);
378: break;
379:
380: case 6: e_key[4] = in_key[4]; t = e_key[5] = in_key[5];
381: for(i = 0; i < 8; ++i)
382: loop6(i);
383: break;
384:
385: case 8: e_key[4] = in_key[4]; e_key[5] = in_key[5];
386: e_key[6] = in_key[6]; t = e_key[7] = in_key[7];
387: for(i = 0; i < 7; ++i)
388: loop8(i);
389: break;
390: }
391:
392: if (!encrypt) {
393: d_key[0] = e_key[0]; d_key[1] = e_key[1];
394: d_key[2] = e_key[2]; d_key[3] = e_key[3];
395:
396: for(i = 4; i < 4 * ctx->k_len + 24; ++i) {
397: imix_col(d_key[i], e_key[i]);
398: }
399: }
400:
401: return ctx;
1.2 markus 402: }
1.1 markus 403:
404: /* encrypt a block of text */
405:
406: #define f_nround(bo, bi, k) \
407: f_rn(bo, bi, 0, k); \
408: f_rn(bo, bi, 1, k); \
409: f_rn(bo, bi, 2, k); \
410: f_rn(bo, bi, 3, k); \
411: k += 4
412:
413: #define f_lround(bo, bi, k) \
414: f_rl(bo, bi, 0, k); \
415: f_rl(bo, bi, 1, k); \
416: f_rl(bo, bi, 2, k); \
417: f_rl(bo, bi, 3, k)
418:
419: void
420: rijndael_encrypt(rijndael_ctx *ctx, const u4byte *in_blk, u4byte *out_blk)
421: {
422: u4byte k_len = ctx->k_len;
423: u4byte *e_key = ctx->e_key;
424: u4byte b0[4], b1[4], *kp;
425:
426: b0[0] = in_blk[0] ^ e_key[0]; b0[1] = in_blk[1] ^ e_key[1];
427: b0[2] = in_blk[2] ^ e_key[2]; b0[3] = in_blk[3] ^ e_key[3];
428:
429: kp = e_key + 4;
430:
431: if(k_len > 6) {
432: f_nround(b1, b0, kp); f_nround(b0, b1, kp);
433: }
434:
435: if(k_len > 4) {
436: f_nround(b1, b0, kp); f_nround(b0, b1, kp);
437: }
438:
439: f_nround(b1, b0, kp); f_nround(b0, b1, kp);
440: f_nround(b1, b0, kp); f_nround(b0, b1, kp);
441: f_nround(b1, b0, kp); f_nround(b0, b1, kp);
442: f_nround(b1, b0, kp); f_nround(b0, b1, kp);
443: f_nround(b1, b0, kp); f_lround(b0, b1, kp);
444:
445: out_blk[0] = b0[0]; out_blk[1] = b0[1];
446: out_blk[2] = b0[2]; out_blk[3] = b0[3];
1.2 markus 447: }
1.1 markus 448:
449: /* decrypt a block of text */
450:
451: #define i_nround(bo, bi, k) \
452: i_rn(bo, bi, 0, k); \
453: i_rn(bo, bi, 1, k); \
454: i_rn(bo, bi, 2, k); \
455: i_rn(bo, bi, 3, k); \
456: k -= 4
457:
458: #define i_lround(bo, bi, k) \
459: i_rl(bo, bi, 0, k); \
460: i_rl(bo, bi, 1, k); \
461: i_rl(bo, bi, 2, k); \
462: i_rl(bo, bi, 3, k)
463:
464: void
465: rijndael_decrypt(rijndael_ctx *ctx, const u4byte *in_blk, u4byte *out_blk)
466: {
467: u4byte b0[4], b1[4], *kp;
468: u4byte k_len = ctx->k_len;
469: u4byte *e_key = ctx->e_key;
470: u4byte *d_key = ctx->d_key;
471:
472: b0[0] = in_blk[0] ^ e_key[4 * k_len + 24]; b0[1] = in_blk[1] ^ e_key[4 * k_len + 25];
473: b0[2] = in_blk[2] ^ e_key[4 * k_len + 26]; b0[3] = in_blk[3] ^ e_key[4 * k_len + 27];
474:
475: kp = d_key + 4 * (k_len + 5);
476:
477: if(k_len > 6) {
478: i_nround(b1, b0, kp); i_nround(b0, b1, kp);
479: }
480:
481: if(k_len > 4) {
482: i_nround(b1, b0, kp); i_nround(b0, b1, kp);
483: }
484:
485: i_nround(b1, b0, kp); i_nround(b0, b1, kp);
486: i_nround(b1, b0, kp); i_nround(b0, b1, kp);
487: i_nround(b1, b0, kp); i_nround(b0, b1, kp);
488: i_nround(b1, b0, kp); i_nround(b0, b1, kp);
489: i_nround(b1, b0, kp); i_lround(b0, b1, kp);
490:
491: out_blk[0] = b0[0]; out_blk[1] = b0[1];
492: out_blk[2] = b0[2]; out_blk[3] = b0[3];
1.2 markus 493: }