xref: /titanic_44/usr/src/common/bignum/bignumimpl.c (revision 8b2cc8ac894f2d58f38cf2fb7c3ac778f4c57c09)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 /*
27  * Configuration guide
28  * -------------------
29  *
30  * There are 4 preprocessor symbols used to configure the bignum
31  * implementation.  This file contains no logic to configure based on
32  * processor; we leave that to the Makefiles to specify.
33  *
34  * USE_FLOATING_POINT
35  *   Meaning: There is support for a fast floating-point implementation of
36  *   Montgomery multiply.
37  *
38  * PSR_MUL
39  *   Meaning: There are processor-specific versions of the low level
40  *   functions to implement big_mul.  Those functions are: big_mul_set_vec,
41  *   big_mul_add_vec, big_mul_vec, and big_sqr_vec.  PSR_MUL implies support
42  *   for all 4 functions.  You cannot pick and choose which subset of these
43  *   functions to support; that would lead to a rat's nest of #ifdefs.
44  *
45  * HWCAP
46  *   Meaning: Call multiply support functions through a function pointer.
47  *   On x86, there are multiple implementations for different hardware
48  *   capabilities, such as MMX, SSE2, etc.  Tests are made at run-time, when
49  *   a function is first used.  So, the support functions are called through
50  *   a function pointer.  There is no need for that on Sparc, because there
51  *   is only one implementation; support functions are called directly.
52  *   Later, if there were some new VIS instruction, or something, and a
53  *   run-time test were needed, rather than variant kernel modules and
54  *   libraries, then HWCAP would be defined for Sparc, as well.
55  *
56  * UMUL64
57  *   Meaning: It is safe to use generic C code that assumes the existence
58  *   of a 32 x 32 --> 64 bit unsigned multiply.  If this is not defined,
59  *   then the generic code for big_mul_add_vec() must necessarily be very slow,
60  *   because it must fall back to using 16 x 16 --> 32 bit multiplication.
61  *
62  */
63 
64 
65 #include <sys/types.h>
66 #include "bignum.h"
67 
68 #ifdef	_KERNEL
69 #include <sys/ddi.h>
70 #include <sys/mdesc.h>
71 #include <sys/crypto/common.h>
72 
73 #include <sys/kmem.h>
74 #include <sys/param.h>
75 #include <sys/sunddi.h>
76 
77 #else
78 #include <stdlib.h>
79 #include <stdio.h>
80 #include <assert.h>
81 #define	ASSERT	assert
82 #endif	/* _KERNEL */
83 
84 #ifdef	_LP64 /* truncate 64-bit size_t to 32-bits */
85 #define	UI32(ui)	((uint32_t)ui)
86 #else /* size_t already 32-bits */
87 #define	UI32(ui)	(ui)
88 #endif
89 
90 
91 #ifdef	_KERNEL
92 #define	big_malloc(size)	kmem_alloc(size, KM_NOSLEEP)
93 #define	big_free(ptr, size)	kmem_free(ptr, size)
94 
95 void *
96 big_realloc(void *from, size_t oldsize, size_t newsize)
97 {
98 	void *rv;
99 
100 	rv = kmem_alloc(newsize, KM_NOSLEEP);
101 	if (rv != NULL)
102 		bcopy(from, rv, oldsize);
103 	kmem_free(from, oldsize);
104 	return (rv);
105 }
106 
107 #else	/* _KERNEL */
108 
109 #ifndef MALLOC_DEBUG
110 
111 #define	big_malloc(size)	malloc(size)
112 #define	big_free(ptr, size)	free(ptr)
113 
114 #else
115 
116 void
117 big_free(void *ptr, size_t size)
118 {
119 	printf("freed %d bytes at %p\n", size, ptr);
120 	free(ptr);
121 }
122 
123 void *
124 big_malloc(size_t size)
125 {
126 	void *rv;
127 	rv = malloc(size);
128 	printf("malloced %d bytes, addr:%p\n", size, rv);
129 	return (rv);
130 }
131 #endif /* MALLOC_DEBUG */
132 
133 #define	big_realloc(x, y, z) realloc((x), (z))
134 
135 
136 /*
137  * printbignum()
138  * Print a BIGNUM type to stdout.
139  */
140 void
141 printbignum(char *aname, BIGNUM *a)
142 {
143 	int i;
144 
145 	(void) printf("\n%s\n%d\n", aname, a->sign*a->len);
146 	for (i = a->len - 1; i >= 0; i--) {
147 #ifdef BIGNUM_CHUNK_32
148 		(void) printf("%08x ", a->value[i]);
149 		if (((i & (BITSINBYTE - 1)) == 0) && (i != 0)) {
150 			(void) printf("\n");
151 		}
152 #else
153 		(void) printf("%08x %08x ", (uint32_t)((a->value[i]) >> 32),
154 		    (uint32_t)((a->value[i]) & 0xffffffff));
155 		if (((i & 3) == 0) && (i != 0)) { /* end of this chunk */
156 			(void) printf("\n");
157 		}
158 #endif
159 	}
160 	(void) printf("\n");
161 }
162 
163 #endif	/* _KERNEL */
164 
165 
166 /*
167  * big_init()
168  * Initialize and allocate memory for a BIGNUM type.
169  *
170  * big_init(number, size) is equivalent to big_init1(number, size, NULL, 0)
171  *
172  * Note: call big_finish() to free memory allocated by big_init().
173  *
174  * Input:
175  * number	Uninitialized memory for BIGNUM
176  * size		Minimum size, in BIG_CHUNK_SIZE-bit words, required for BIGNUM
177  *
178  * Output:
179  * number	Initialized BIGNUM
180  *
181  * Return BIG_OK on success or BIG_NO_MEM for an allocation error.
182  */
183 BIG_ERR_CODE
184 big_init(BIGNUM *number, int size)
185 {
186 	number->value = big_malloc(BIGNUM_WORDSIZE * size);
187 	if (number->value == NULL) {
188 		return (BIG_NO_MEM);
189 	}
190 	number->size = size;
191 	number->len = 0;
192 	number->sign = 1;
193 	number->malloced = 1;
194 	return (BIG_OK);
195 }
196 
197 
198 /*
199  * big_init1()
200  * Initialize and, if needed, allocate memory for a BIGNUM type.
201  * Use the buffer passed, buf, if any, instad of allocating memory
202  * if it's at least "size" bytes.
203  *
204  * Note: call big_finish() to free memory allocated by big_init().
205  *
206  * Input:
207  * number	Uninitialized memory for BIGNUM
208  * size		Minimum size, in BIG_CHUNK_SIZE-bit words, required for BIGNUM
209  * buf		Buffer for storing a BIGNUM.
210  *		If NULL, big_init1() will allocate a buffer
211  * bufsize	Size, in BIG_CHUNK_SIZE_bit words, of buf
212  *
213  * Output:
214  * number	Initialized BIGNUM
215  *
216  * Return BIG_OK on success or BIG_NO_MEM for an allocation error.
217  */
218 BIG_ERR_CODE
219 big_init1(BIGNUM *number, int size, BIG_CHUNK_TYPE *buf, int bufsize)
220 {
221 	if ((buf == NULL) || (size > bufsize)) {
222 		number->value = big_malloc(BIGNUM_WORDSIZE * size);
223 		if (number->value == NULL) {
224 			return (BIG_NO_MEM);
225 		}
226 		number->size = size;
227 		number->malloced = 1;
228 	} else {
229 		number->value = buf;
230 		number->size = bufsize;
231 		number->malloced = 0;
232 	}
233 	number->len = 0;
234 	number->sign = 1;
235 
236 	return (BIG_OK);
237 }
238 
239 
240 /*
241  * big_finish()
242  * Free memory, if any, allocated by big_init() or big_init1().
243  */
244 void
245 big_finish(BIGNUM *number)
246 {
247 	if (number->malloced == 1) {
248 		big_free(number->value, BIGNUM_WORDSIZE * number->size);
249 		number->malloced = 0;
250 	}
251 }
252 
253 
254 /*
255  * bn->size should be at least
256  * (len + BIGNUM_WORDSIZE - 1) / BIGNUM_WORDSIZE bytes
257  * converts from byte-big-endian format to bignum format (words in
258  * little endian order, but bytes within the words big endian)
259  */
260 void
261 bytestring2bignum(BIGNUM *bn, uchar_t *kn, size_t len)
262 {
263 	int		i, j;
264 	uint32_t	offs;
265 	const uint32_t	slen = UI32(len);
266 	BIG_CHUNK_TYPE	word;
267 	uchar_t		*knwordp;
268 
269 	if (slen == 0) {
270 		bn->len = 1;
271 		bn->value[0] = 0;
272 		return;
273 	}
274 
275 	offs = slen % BIGNUM_WORDSIZE;
276 	bn->len = slen / BIGNUM_WORDSIZE;
277 
278 	for (i = 0; i < slen / BIGNUM_WORDSIZE; i++) {
279 		knwordp = &(kn[slen - BIGNUM_WORDSIZE * (i + 1)]);
280 		word = knwordp[0];
281 		for (j = 1; j < BIGNUM_WORDSIZE; j++) {
282 			word = (word << BITSINBYTE) + knwordp[j];
283 		}
284 		bn->value[i] = word;
285 	}
286 	if (offs > 0) {
287 		word = kn[0];
288 		for (i = 1; i < offs; i++) word = (word << BITSINBYTE) + kn[i];
289 		bn->value[bn->len++] = word;
290 	}
291 	while ((bn->len > 1) && (bn->value[bn->len - 1] == 0)) {
292 		bn->len --;
293 	}
294 }
295 
296 
297 /*
298  * copies the least significant len bytes if
299  * len < bn->len * BIGNUM_WORDSIZE
300  * converts from bignum format to byte-big-endian format.
301  * bignum format is words of type  BIG_CHUNK_TYPE in little endian order.
302  */
303 void
304 bignum2bytestring(uchar_t *kn, BIGNUM *bn, size_t len)
305 {
306 	int		i, j;
307 	uint32_t	offs;
308 	const uint32_t	slen = UI32(len);
309 	BIG_CHUNK_TYPE	word;
310 
311 	if (len < BIGNUM_WORDSIZE * bn->len) {
312 		for (i = 0; i < slen / BIGNUM_WORDSIZE; i++) {
313 			word = bn->value[i];
314 			for (j = 0; j < BIGNUM_WORDSIZE; j++) {
315 				kn[slen - BIGNUM_WORDSIZE * i - j - 1] =
316 				    word & 0xff;
317 				word = word >> BITSINBYTE;
318 			}
319 		}
320 		offs = slen % BIGNUM_WORDSIZE;
321 		if (offs > 0) {
322 			word = bn->value[slen / BIGNUM_WORDSIZE];
323 			for (i =  slen % BIGNUM_WORDSIZE; i > 0; i --) {
324 				kn[i - 1] = word & 0xff;
325 				word = word >> BITSINBYTE;
326 			}
327 		}
328 	} else {
329 		for (i = 0; i < bn->len; i++) {
330 			word = bn->value[i];
331 			for (j = 0; j < BIGNUM_WORDSIZE; j++) {
332 				kn[slen - BIGNUM_WORDSIZE * i - j - 1] =
333 				    word & 0xff;
334 				word = word >> BITSINBYTE;
335 			}
336 		}
337 		for (i = 0; i < slen - BIGNUM_WORDSIZE * bn->len; i++) {
338 			kn[i] = 0;
339 		}
340 	}
341 }
342 
343 
344 int
345 big_bitlength(BIGNUM *a)
346 {
347 	int		l = 0, b = 0;
348 	BIG_CHUNK_TYPE	c;
349 
350 	l = a->len - 1;
351 	while ((l > 0) && (a->value[l] == 0)) {
352 		l--;
353 	}
354 	b = BIG_CHUNK_SIZE;
355 	c = a->value[l];
356 	while ((b > 1) && ((c & BIG_CHUNK_HIGHBIT) == 0)) {
357 		c = c << 1;
358 		b--;
359 	}
360 
361 	return (l * BIG_CHUNK_SIZE + b);
362 }
363 
364 
365 BIG_ERR_CODE
366 big_copy(BIGNUM *dest, BIGNUM *src)
367 {
368 	BIG_CHUNK_TYPE	*newptr;
369 	int		i, len;
370 
371 	len = src->len;
372 	while ((len > 1) && (src->value[len - 1] == 0)) {
373 		len--;
374 	}
375 	src->len = len;
376 	if (dest->size < len) {
377 		if (dest->malloced == 1) {
378 			newptr = (BIG_CHUNK_TYPE *)big_realloc(dest->value,
379 			    BIGNUM_WORDSIZE * dest->size,
380 			    BIGNUM_WORDSIZE * len);
381 		} else {
382 			newptr = (BIG_CHUNK_TYPE *)
383 			    big_malloc(BIGNUM_WORDSIZE * len);
384 			if (newptr != NULL) {
385 				dest->malloced = 1;
386 			}
387 		}
388 		if (newptr == NULL) {
389 			return (BIG_NO_MEM);
390 		}
391 		dest->value = newptr;
392 		dest->size = len;
393 	}
394 	dest->len = len;
395 	dest->sign = src->sign;
396 	for (i = 0; i < len; i++) {
397 		dest->value[i] = src->value[i];
398 	}
399 
400 	return (BIG_OK);
401 }
402 
403 
404 BIG_ERR_CODE
405 big_extend(BIGNUM *number, int size)
406 {
407 	BIG_CHUNK_TYPE	*newptr;
408 	int		i;
409 
410 	if (number->size >= size)
411 		return (BIG_OK);
412 	if (number->malloced) {
413 		number->value = big_realloc(number->value,
414 		    BIGNUM_WORDSIZE * number->size,
415 		    BIGNUM_WORDSIZE * size);
416 	} else {
417 		newptr = big_malloc(BIGNUM_WORDSIZE * size);
418 		if (newptr != NULL) {
419 			for (i = 0; i < number->size; i++) {
420 				newptr[i] = number->value[i];
421 			}
422 		}
423 		number->value = newptr;
424 	}
425 
426 	if (number->value == NULL) {
427 		return (BIG_NO_MEM);
428 	}
429 
430 	number->size = size;
431 	number->malloced = 1;
432 	return (BIG_OK);
433 }
434 
435 
436 /* returns 1 if n == 0 */
437 int
438 big_is_zero(BIGNUM *n)
439 {
440 	int	i, result;
441 
442 	result = 1;
443 	for (i = 0; i < n->len; i++) {
444 		if (n->value[i] != 0) {
445 			result = 0;
446 		}
447 	}
448 	return (result);
449 }
450 
451 
452 BIG_ERR_CODE
453 big_add_abs(BIGNUM *result, BIGNUM *aa, BIGNUM *bb)
454 {
455 	int		i, shorter, longer;
456 	BIG_CHUNK_TYPE	cy, ai;
457 	BIG_CHUNK_TYPE	*r, *a, *b, *c;
458 	BIG_ERR_CODE	err;
459 	BIGNUM		*longerarg;
460 
461 	if (aa->len > bb->len) {
462 		shorter = bb->len;
463 		longer = aa->len;
464 		longerarg = aa;
465 	} else {
466 		shorter = aa->len;
467 		longer = bb->len;
468 		longerarg = bb;
469 	}
470 	if (result->size < longer + 1) {
471 		err = big_extend(result, longer + 1);
472 		if (err != BIG_OK) {
473 			return (err);
474 		}
475 	}
476 
477 	r = result->value;
478 	a = aa->value;
479 	b = bb->value;
480 	c = longerarg->value;
481 	cy = 0;
482 	for (i = 0; i < shorter; i++) {
483 		ai = a[i];
484 		r[i] = ai + b[i] + cy;
485 		if (r[i] > ai) {
486 			cy = 0;
487 		} else if (r[i] < ai) {
488 			cy = 1;
489 		}
490 	}
491 	for (; i < longer; i++) {
492 		ai = c[i];
493 		r[i] = ai + cy;
494 		if (r[i] >= ai) {
495 			cy = 0;
496 		}
497 	}
498 	if (cy == 1) {
499 		r[i] = cy;
500 		result->len = longer + 1;
501 	} else {
502 		result->len = longer;
503 	}
504 	result->sign = 1;
505 	return (BIG_OK);
506 }
507 
508 
509 /* caller must make sure that result has at least len words allocated */
510 void
511 big_sub_vec(BIG_CHUNK_TYPE *r, BIG_CHUNK_TYPE *a, BIG_CHUNK_TYPE *b, int len)
512 {
513 	int		i;
514 	BIG_CHUNK_TYPE	cy, ai;
515 
516 	cy = 1;
517 	for (i = 0; i < len; i++) {
518 		ai = a[i];
519 		r[i] = ai + (~b[i]) + cy;
520 		if (r[i] > ai) {
521 			cy = 0;
522 		} else if (r[i] < ai) {
523 			cy = 1;
524 		}
525 	}
526 }
527 
528 
529 /* result=aa-bb  it is assumed that aa>=bb */
530 BIG_ERR_CODE
531 big_sub_pos(BIGNUM *result, BIGNUM *aa, BIGNUM *bb)
532 {
533 	int		i, shorter;
534 	BIG_CHUNK_TYPE	cy = 1, ai;
535 	BIG_CHUNK_TYPE	*r, *a, *b;
536 	BIG_ERR_CODE	err = BIG_OK;
537 
538 	if (aa->len > bb->len) {
539 		shorter = bb->len;
540 	} else {
541 		shorter = aa->len;
542 	}
543 	if (result->size < aa->len) {
544 		err = big_extend(result, aa->len);
545 		if (err != BIG_OK) {
546 			return (err);
547 		}
548 	}
549 
550 	r = result->value;
551 	a = aa->value;
552 	b = bb->value;
553 	result->len = aa->len;
554 	cy = 1;
555 	for (i = 0; i < shorter; i++) {
556 		ai = a[i];
557 		r[i] = ai + (~b[i]) + cy;
558 		if (r[i] > ai) {
559 			cy = 0;
560 		} else if (r[i] < ai) {
561 			cy = 1;
562 		}
563 	}
564 	for (; i < aa->len; i++) {
565 		ai = a[i];
566 		r[i] = ai + (~0) + cy;
567 		if (r[i] < ai) {
568 			cy = 1;
569 		}
570 	}
571 	result->sign = 1;
572 
573 	if (cy == 0) {
574 		return (BIG_INVALID_ARGS);
575 	} else {
576 		return (BIG_OK);
577 	}
578 }
579 
580 
581 /* returns -1 if |aa|<|bb|, 0 if |aa|==|bb| 1 if |aa|>|bb| */
582 int
583 big_cmp_abs(BIGNUM *aa, BIGNUM *bb)
584 {
585 	int	i;
586 
587 	if (aa->len > bb->len) {
588 		for (i = aa->len - 1; i > bb->len - 1; i--) {
589 			if (aa->value[i] > 0) {
590 				return (1);
591 			}
592 		}
593 	} else if (aa->len < bb->len) {
594 		for (i = bb->len - 1; i > aa->len - 1; i--) {
595 			if (bb->value[i] > 0) {
596 				return (-1);
597 			}
598 		}
599 	} else {
600 		i = aa->len - 1;
601 	}
602 	for (; i >= 0; i--) {
603 		if (aa->value[i] > bb->value[i]) {
604 			return (1);
605 		} else if (aa->value[i] < bb->value[i]) {
606 			return (-1);
607 		}
608 	}
609 
610 	return (0);
611 }
612 
613 
614 BIG_ERR_CODE
615 big_sub(BIGNUM *result, BIGNUM *aa, BIGNUM *bb)
616 {
617 	BIG_ERR_CODE	err;
618 
619 	if ((bb->sign == -1) && (aa->sign == 1)) {
620 		if ((err = big_add_abs(result, aa, bb)) != BIG_OK) {
621 			return (err);
622 		}
623 		result->sign = 1;
624 	} else if ((aa->sign == -1) && (bb->sign == 1)) {
625 		if ((err = big_add_abs(result, aa, bb)) != BIG_OK) {
626 			return (err);
627 		}
628 		result->sign = -1;
629 	} else if ((aa->sign == 1) && (bb->sign == 1)) {
630 		if (big_cmp_abs(aa, bb) >= 0) {
631 			if ((err = big_sub_pos(result, aa, bb)) != BIG_OK) {
632 				return (err);
633 			}
634 			result->sign = 1;
635 		} else {
636 			if ((err = big_sub_pos(result, bb, aa)) != BIG_OK) {
637 				return (err);
638 			}
639 			result->sign = -1;
640 		}
641 	} else {
642 		if (big_cmp_abs(aa, bb) >= 0) {
643 			if ((err = big_sub_pos(result, aa, bb)) != BIG_OK) {
644 				return (err);
645 			}
646 			result->sign = -1;
647 		} else {
648 			if ((err = big_sub_pos(result, bb, aa)) != BIG_OK) {
649 				return (err);
650 			}
651 			result->sign = 1;
652 		}
653 	}
654 	return (BIG_OK);
655 }
656 
657 
658 BIG_ERR_CODE
659 big_add(BIGNUM *result, BIGNUM *aa, BIGNUM *bb)
660 {
661 	BIG_ERR_CODE	err;
662 
663 	if ((bb->sign == -1) && (aa->sign == -1)) {
664 		if ((err = big_add_abs(result, aa, bb)) != BIG_OK) {
665 			return (err);
666 		}
667 		result->sign = -1;
668 	} else if ((aa->sign == 1) && (bb->sign == 1)) {
669 		if ((err = big_add_abs(result, aa, bb)) != BIG_OK) {
670 			return (err);
671 		}
672 		result->sign = 1;
673 	} else if ((aa->sign == 1) && (bb->sign == -1)) {
674 		if (big_cmp_abs(aa, bb) >= 0) {
675 			if ((err = big_sub_pos(result, aa, bb)) != BIG_OK) {
676 				return (err);
677 			}
678 			result->sign = 1;
679 		} else {
680 			if ((err = big_sub_pos(result, bb, aa)) != BIG_OK) {
681 				return (err);
682 			}
683 			result->sign = -1;
684 		}
685 	} else {
686 		if (big_cmp_abs(aa, bb) >= 0) {
687 			if ((err = big_sub_pos(result, aa, bb)) != BIG_OK) {
688 				return (err);
689 			}
690 			result->sign = -1;
691 		} else {
692 			if ((err = big_sub_pos(result, bb, aa)) != BIG_OK) {
693 				return (err);
694 			}
695 			result->sign = 1;
696 		}
697 	}
698 	return (BIG_OK);
699 }
700 
701 
702 /* result = aa/2 */
703 BIG_ERR_CODE
704 big_half_pos(BIGNUM *result, BIGNUM *aa)
705 {
706 	BIG_ERR_CODE	err;
707 	int		i;
708 	BIG_CHUNK_TYPE	cy, cy1;
709 	BIG_CHUNK_TYPE	*a, *r;
710 
711 	if (result->size < aa->len) {
712 		err = big_extend(result, aa->len);
713 		if (err != BIG_OK) {
714 			return (err);
715 		}
716 	}
717 
718 	result->len = aa->len;
719 	a = aa->value;
720 	r = result->value;
721 	cy = 0;
722 	for (i = aa->len - 1; i >= 0; i--) {
723 		cy1 = a[i] << (BIG_CHUNK_SIZE - 1);
724 		r[i] = (cy | (a[i] >> 1));
725 		cy = cy1;
726 	}
727 	if (r[result->len - 1] == 0) {
728 		result->len--;
729 	}
730 
731 	return (BIG_OK);
732 }
733 
734 /* result  =  aa*2 */
735 BIG_ERR_CODE
736 big_double(BIGNUM *result, BIGNUM *aa)
737 {
738 	BIG_ERR_CODE	err;
739 	int		i, rsize;
740 	BIG_CHUNK_TYPE	cy, cy1;
741 	BIG_CHUNK_TYPE	*a, *r;
742 
743 	if ((aa->len > 0) &&
744 	    ((aa->value[aa->len - 1] & BIG_CHUNK_HIGHBIT) != 0)) {
745 		rsize = aa->len + 1;
746 	} else {
747 		rsize = aa->len;
748 	}
749 
750 	if (result->size < rsize) {
751 		err = big_extend(result, rsize);
752 		if (err != BIG_OK)
753 			return (err);
754 	}
755 
756 	a = aa->value;
757 	r = result->value;
758 	if (rsize == aa->len + 1) {
759 		r[rsize - 1] = 1;
760 	}
761 	cy = 0;
762 	for (i = 0; i < aa->len; i++) {
763 		cy1 = a[i] >> (BIG_CHUNK_SIZE - 1);
764 		r[i] = (cy | (a[i] << 1));
765 		cy = cy1;
766 	}
767 	result->len = rsize;
768 	return (BIG_OK);
769 }
770 
771 
772 /*
773  * returns aa mod b, aa must be nonneg, b must be a max
774  * (BIG_CHUNK_SIZE / 2)-bit integer
775  */
776 static uint32_t
777 big_modhalf_pos(BIGNUM *aa, uint32_t b)
778 {
779 	int		i;
780 	BIG_CHUNK_TYPE	rem;
781 
782 	if (aa->len == 0) {
783 		return (0);
784 	}
785 	rem = aa->value[aa->len - 1] % b;
786 	for (i = aa->len - 2; i >= 0; i--) {
787 		rem = ((rem << (BIG_CHUNK_SIZE / 2)) |
788 		    (aa->value[i] >> (BIG_CHUNK_SIZE / 2))) % b;
789 		rem = ((rem << (BIG_CHUNK_SIZE / 2)) |
790 		    (aa->value[i] & BIG_CHUNK_LOWHALFBITS)) % b;
791 	}
792 
793 	return ((uint32_t)rem);
794 }
795 
796 
797 /*
798  * result = aa - (2^BIG_CHUNK_SIZE)^lendiff * bb
799  * result->size should be at least aa->len at entry
800  * aa, bb, and result should be positive
801  */
802 void
803 big_sub_pos_high(BIGNUM *result, BIGNUM *aa, BIGNUM *bb)
804 {
805 	int i, lendiff;
806 	BIGNUM res1, aa1;
807 
808 	lendiff = aa->len - bb->len;
809 	res1.size = result->size - lendiff;
810 	res1.malloced = 0;
811 	res1.value = result->value + lendiff;
812 	aa1.size = aa->size - lendiff;
813 	aa1.value = aa->value + lendiff;
814 	aa1.len = bb->len;
815 	aa1.sign = 1;
816 	(void) big_sub_pos(&res1, &aa1, bb);
817 	if (result->value != aa->value) {
818 		for (i = 0; i < lendiff; i++) {
819 			result->value[i] = aa->value[i];
820 		}
821 	}
822 	result->len = aa->len;
823 }
824 
825 
826 /*
827  * returns 1, 0, or -1 depending on whether |aa| > , ==, or <
828  *					(2^BIG_CHUNK_SIZE)^lendiff * |bb|
829  * aa->len should be >= bb->len
830  */
831 int
832 big_cmp_abs_high(BIGNUM *aa, BIGNUM *bb)
833 {
834 	int		lendiff;
835 	BIGNUM		aa1;
836 
837 	lendiff = aa->len - bb->len;
838 	aa1.len = bb->len;
839 	aa1.size = aa->size - lendiff;
840 	aa1.malloced = 0;
841 	aa1.value = aa->value + lendiff;
842 	return (big_cmp_abs(&aa1, bb));
843 }
844 
845 
846 /*
847  * result = aa * b where b is a max. (BIG_CHUNK_SIZE / 2)-bit positive integer.
848  * result should have enough space allocated.
849  */
850 static void
851 big_mulhalf_low(BIGNUM *result, BIGNUM *aa, BIG_CHUNK_TYPE b)
852 {
853 	int		i;
854 	BIG_CHUNK_TYPE	t1, t2, ai, cy;
855 	BIG_CHUNK_TYPE	*a, *r;
856 
857 	a = aa->value;
858 	r = result->value;
859 	cy = 0;
860 	for (i = 0; i < aa->len; i++) {
861 		ai = a[i];
862 		t1 = (ai & BIG_CHUNK_LOWHALFBITS) * b + cy;
863 		t2 = (ai >> (BIG_CHUNK_SIZE / 2)) * b +
864 		    (t1 >> (BIG_CHUNK_SIZE / 2));
865 		r[i] = (t1 & BIG_CHUNK_LOWHALFBITS) |
866 		    (t2 << (BIG_CHUNK_SIZE / 2));
867 		cy = t2 >> (BIG_CHUNK_SIZE / 2);
868 	}
869 	r[i] = cy;
870 	result->len = aa->len + 1;
871 	result->sign = aa->sign;
872 }
873 
874 
875 /*
876  * result = aa * b * 2^(BIG_CHUNK_SIZE / 2) where b is a max.
877  * (BIG_CHUNK_SIZE / 2)-bit positive integer.
878  * result should have enough space allocated.
879  */
880 static void
881 big_mulhalf_high(BIGNUM *result, BIGNUM *aa, BIG_CHUNK_TYPE b)
882 {
883 	int		i;
884 	BIG_CHUNK_TYPE	t1, t2, ai, cy, ri;
885 	BIG_CHUNK_TYPE	*a, *r;
886 
887 	a = aa->value;
888 	r = result->value;
889 	cy = 0;
890 	ri = 0;
891 	for (i = 0; i < aa->len; i++) {
892 		ai = a[i];
893 		t1 = (ai & BIG_CHUNK_LOWHALFBITS) * b + cy;
894 		t2 = (ai >>  (BIG_CHUNK_SIZE / 2)) * b +
895 		    (t1 >>  (BIG_CHUNK_SIZE / 2));
896 		r[i] = (t1 <<  (BIG_CHUNK_SIZE / 2)) + ri;
897 		ri = t2 & BIG_CHUNK_LOWHALFBITS;
898 		cy = t2 >> (BIG_CHUNK_SIZE / 2);
899 	}
900 	r[i] = (cy <<  (BIG_CHUNK_SIZE / 2)) + ri;
901 	result->len = aa->len + 1;
902 	result->sign = aa->sign;
903 }
904 
905 
906 /* it is assumed that result->size is big enough */
907 void
908 big_shiftleft(BIGNUM *result, BIGNUM *aa, int offs)
909 {
910 	int		i;
911 	BIG_CHUNK_TYPE	cy, ai;
912 
913 	if (offs == 0) {
914 		if (result != aa) {
915 			(void) big_copy(result, aa);
916 		}
917 		return;
918 	}
919 	cy = 0;
920 	for (i = 0; i < aa->len; i++) {
921 		ai = aa->value[i];
922 		result->value[i] = (ai << offs) | cy;
923 		cy = ai >> (BIG_CHUNK_SIZE - offs);
924 	}
925 	if (cy != 0) {
926 		result->len = aa->len + 1;
927 		result->value[result->len - 1] = cy;
928 	} else {
929 		result->len = aa->len;
930 	}
931 	result->sign = aa->sign;
932 }
933 
934 
935 /* it is assumed that result->size is big enough */
936 void
937 big_shiftright(BIGNUM *result, BIGNUM *aa, int offs)
938 {
939 	int		 i;
940 	BIG_CHUNK_TYPE	cy, ai;
941 
942 	if (offs == 0) {
943 		if (result != aa) {
944 			(void) big_copy(result, aa);
945 		}
946 		return;
947 	}
948 	cy = aa->value[0] >> offs;
949 	for (i = 1; i < aa->len; i++) {
950 		ai = aa->value[i];
951 		result->value[i - 1] = (ai << (BIG_CHUNK_SIZE - offs)) | cy;
952 		cy = ai >> offs;
953 	}
954 	result->len = aa->len;
955 	result->value[result->len - 1] = cy;
956 	result->sign = aa->sign;
957 }
958 
959 
960 /*
961  * result = aa/bb   remainder = aa mod bb
962  * it is assumed that aa and bb are positive
963  */
964 BIG_ERR_CODE
965 big_div_pos(BIGNUM *result, BIGNUM *remainder, BIGNUM *aa, BIGNUM *bb)
966 {
967 	BIG_ERR_CODE	err = BIG_OK;
968 	int		i, alen, blen, tlen, rlen, offs;
969 	BIG_CHUNK_TYPE	higha, highb, coeff;
970 	BIG_CHUNK_TYPE	*a, *b;
971 	BIGNUM		bbhigh, bblow, tresult, tmp1, tmp2;
972 	BIG_CHUNK_TYPE	tmp1value[BIGTMPSIZE];
973 	BIG_CHUNK_TYPE	tmp2value[BIGTMPSIZE];
974 	BIG_CHUNK_TYPE	tresultvalue[BIGTMPSIZE];
975 	BIG_CHUNK_TYPE	bblowvalue[BIGTMPSIZE];
976 	BIG_CHUNK_TYPE	bbhighvalue[BIGTMPSIZE];
977 
978 	a = aa->value;
979 	b = bb->value;
980 	alen = aa->len;
981 	blen = bb->len;
982 	while ((alen > 1) && (a[alen - 1] == 0)) {
983 		alen = alen - 1;
984 	}
985 	aa->len = alen;
986 	while ((blen > 1) && (b[blen - 1] == 0)) {
987 		blen = blen - 1;
988 	}
989 	bb->len = blen;
990 	if ((blen == 1) && (b[0] == 0)) {
991 		return (BIG_DIV_BY_0);
992 	}
993 
994 	if (big_cmp_abs(aa, bb) < 0) {
995 		if ((remainder != NULL) &&
996 		    ((err = big_copy(remainder, aa)) != BIG_OK)) {
997 			return (err);
998 		}
999 		if (result != NULL) {
1000 			result->len = 1;
1001 			result->sign = 1;
1002 			result->value[0] = 0;
1003 		}
1004 		return (BIG_OK);
1005 	}
1006 
1007 	if ((err = big_init1(&bblow, blen + 1,
1008 	    bblowvalue, arraysize(bblowvalue))) != BIG_OK)
1009 		return (err);
1010 
1011 	if ((err = big_init1(&bbhigh, blen + 1,
1012 	    bbhighvalue, arraysize(bbhighvalue))) != BIG_OK)
1013 		goto ret1;
1014 
1015 	if ((err = big_init1(&tmp1, alen + 2,
1016 	    tmp1value, arraysize(tmp1value))) != BIG_OK)
1017 		goto ret2;
1018 
1019 	if ((err = big_init1(&tmp2, blen + 2,
1020 	    tmp2value, arraysize(tmp2value))) != BIG_OK)
1021 		goto ret3;
1022 
1023 	if ((err = big_init1(&tresult, alen - blen + 2,
1024 	    tresultvalue, arraysize(tresultvalue))) != BIG_OK)
1025 		goto ret4;
1026 
1027 	offs = 0;
1028 	highb = b[blen - 1];
1029 	if (highb >= (BIG_CHUNK_HALF_HIGHBIT << 1)) {
1030 		highb = highb >> (BIG_CHUNK_SIZE / 2);
1031 		offs = (BIG_CHUNK_SIZE / 2);
1032 	}
1033 	while ((highb & BIG_CHUNK_HALF_HIGHBIT) == 0) {
1034 		highb = highb << 1;
1035 		offs++;
1036 	}
1037 
1038 	big_shiftleft(&bblow, bb, offs);
1039 
1040 	if (offs <= (BIG_CHUNK_SIZE / 2 - 1)) {
1041 		big_shiftleft(&bbhigh, &bblow, BIG_CHUNK_SIZE / 2);
1042 	} else {
1043 		big_shiftright(&bbhigh, &bblow, BIG_CHUNK_SIZE / 2);
1044 	}
1045 	if (bbhigh.value[bbhigh.len - 1] == 0) {
1046 		bbhigh.len--;
1047 	} else {
1048 		bbhigh.value[bbhigh.len] = 0;
1049 	}
1050 
1051 	highb = bblow.value[bblow.len - 1];
1052 
1053 	big_shiftleft(&tmp1, aa, offs);
1054 	rlen = tmp1.len - bblow.len + 1;
1055 	tresult.len = rlen;
1056 
1057 	tmp1.len++;
1058 	tlen = tmp1.len;
1059 	tmp1.value[tmp1.len - 1] = 0;
1060 	for (i = 0; i < rlen; i++) {
1061 		higha = (tmp1.value[tlen - 1] << (BIG_CHUNK_SIZE / 2)) +
1062 		    (tmp1.value[tlen - 2] >> (BIG_CHUNK_SIZE / 2));
1063 		coeff = higha / (highb + 1);
1064 		big_mulhalf_high(&tmp2, &bblow, coeff);
1065 		big_sub_pos_high(&tmp1, &tmp1, &tmp2);
1066 		bbhigh.len++;
1067 		while (tmp1.value[tlen - 1] > 0) {
1068 			big_sub_pos_high(&tmp1, &tmp1, &bbhigh);
1069 			coeff++;
1070 		}
1071 		bbhigh.len--;
1072 		tlen--;
1073 		tmp1.len--;
1074 		while (big_cmp_abs_high(&tmp1, &bbhigh) >= 0) {
1075 			big_sub_pos_high(&tmp1, &tmp1, &bbhigh);
1076 			coeff++;
1077 		}
1078 		tresult.value[rlen - i - 1] = coeff << (BIG_CHUNK_SIZE / 2);
1079 		higha = tmp1.value[tlen - 1];
1080 		coeff = higha / (highb + 1);
1081 		big_mulhalf_low(&tmp2, &bblow, coeff);
1082 		tmp2.len--;
1083 		big_sub_pos_high(&tmp1, &tmp1, &tmp2);
1084 		while (big_cmp_abs_high(&tmp1, &bblow) >= 0) {
1085 			big_sub_pos_high(&tmp1, &tmp1, &bblow);
1086 			coeff++;
1087 		}
1088 		tresult.value[rlen - i - 1] =
1089 		    tresult.value[rlen - i - 1] + coeff;
1090 	}
1091 
1092 	big_shiftright(&tmp1, &tmp1, offs);
1093 
1094 	err = BIG_OK;
1095 
1096 	if ((remainder != NULL) &&
1097 	    ((err = big_copy(remainder, &tmp1)) != BIG_OK))
1098 		goto ret;
1099 
1100 	if (result != NULL)
1101 		err = big_copy(result, &tresult);
1102 
1103 ret:
1104 	big_finish(&tresult);
1105 ret4:
1106 	big_finish(&tmp1);
1107 ret3:
1108 	big_finish(&tmp2);
1109 ret2:
1110 	big_finish(&bbhigh);
1111 ret1:
1112 	big_finish(&bblow);
1113 	return (err);
1114 }
1115 
1116 
1117 /*
1118  * If there is no processor-specific integer implementation of
1119  * the lower level multiply functions, then this code is provided
1120  * for big_mul_set_vec(), big_mul_add_vec(), big_mul_vec() and
1121  * big_sqr_vec().
1122  *
1123  * There are two generic implementations.  One that assumes that
1124  * there is hardware and C compiler support for a 32 x 32 --> 64
1125  * bit unsigned multiply, but otherwise is not specific to any
1126  * processor, platform, or ISA.
1127  *
1128  * The other makes very few assumptions about hardware capabilities.
1129  * It does not even assume that there is any implementation of a
1130  * 32 x 32 --> 64 bit multiply that is accessible to C code and
1131  * appropriate to use.  It falls constructs 32 x 32 --> 64 bit
1132  * multiplies from 16 x 16 --> 32 bit multiplies.
1133  *
1134  */
1135 
1136 #if !defined(PSR_MUL)
1137 
1138 #ifdef UMUL64
1139 
1140 #if (BIG_CHUNK_SIZE == 32)
1141 
1142 #define	UNROLL8
1143 
1144 #define	MUL_SET_VEC_ROUND_PREFETCH(R) \
1145 	p = pf * d; \
1146 	pf = (uint64_t)a[R + 1]; \
1147 	t = p + cy; \
1148 	r[R] = (uint32_t)t; \
1149 	cy = t >> 32
1150 
1151 #define	MUL_SET_VEC_ROUND_NOPREFETCH(R) \
1152 	p = pf * d; \
1153 	t = p + cy; \
1154 	r[R] = (uint32_t)t; \
1155 	cy = t >> 32
1156 
1157 #define	MUL_ADD_VEC_ROUND_PREFETCH(R) \
1158 	t = (uint64_t)r[R]; \
1159 	p = pf * d; \
1160 	pf = (uint64_t)a[R + 1]; \
1161 	t = p + t + cy; \
1162 	r[R] = (uint32_t)t; \
1163 	cy = t >> 32
1164 
1165 #define	MUL_ADD_VEC_ROUND_NOPREFETCH(R) \
1166 	t = (uint64_t)r[R]; \
1167 	p = pf * d; \
1168 	t = p + t + cy; \
1169 	r[R] = (uint32_t)t; \
1170 	cy = t >> 32
1171 
1172 #ifdef UNROLL8
1173 
1174 #define	UNROLL 8
1175 
1176 /*
1177  * r = a * b
1178  * where r and a are vectors; b is a single 32-bit digit
1179  */
1180 
1181 uint32_t
1182 big_mul_set_vec(uint32_t *r, uint32_t *a, int len, uint32_t b)
1183 {
1184 	uint64_t d, pf, p, t, cy;
1185 
1186 	if (len == 0)
1187 		return (0);
1188 	cy = 0;
1189 	d = (uint64_t)b;
1190 	pf = (uint64_t)a[0];
1191 	while (len > UNROLL) {
1192 		MUL_SET_VEC_ROUND_PREFETCH(0);
1193 		MUL_SET_VEC_ROUND_PREFETCH(1);
1194 		MUL_SET_VEC_ROUND_PREFETCH(2);
1195 		MUL_SET_VEC_ROUND_PREFETCH(3);
1196 		MUL_SET_VEC_ROUND_PREFETCH(4);
1197 		MUL_SET_VEC_ROUND_PREFETCH(5);
1198 		MUL_SET_VEC_ROUND_PREFETCH(6);
1199 		MUL_SET_VEC_ROUND_PREFETCH(7);
1200 		r += UNROLL;
1201 		a += UNROLL;
1202 		len -= UNROLL;
1203 	}
1204 	if (len == UNROLL) {
1205 		MUL_SET_VEC_ROUND_PREFETCH(0);
1206 		MUL_SET_VEC_ROUND_PREFETCH(1);
1207 		MUL_SET_VEC_ROUND_PREFETCH(2);
1208 		MUL_SET_VEC_ROUND_PREFETCH(3);
1209 		MUL_SET_VEC_ROUND_PREFETCH(4);
1210 		MUL_SET_VEC_ROUND_PREFETCH(5);
1211 		MUL_SET_VEC_ROUND_PREFETCH(6);
1212 		MUL_SET_VEC_ROUND_NOPREFETCH(7);
1213 		return ((uint32_t)cy);
1214 	}
1215 	while (len > 1) {
1216 		MUL_SET_VEC_ROUND_PREFETCH(0);
1217 		++r;
1218 		++a;
1219 		--len;
1220 	}
1221 	if (len > 0) {
1222 		MUL_SET_VEC_ROUND_NOPREFETCH(0);
1223 	}
1224 	return ((uint32_t)cy);
1225 }
1226 
1227 /*
1228  * r += a * b
1229  * where r and a are vectors; b is a single 32-bit digit
1230  */
1231 
1232 uint32_t
1233 big_mul_add_vec(uint32_t *r, uint32_t *a, int len, uint32_t b)
1234 {
1235 	uint64_t d, pf, p, t, cy;
1236 
1237 	if (len == 0)
1238 		return (0);
1239 	cy = 0;
1240 	d = (uint64_t)b;
1241 	pf = (uint64_t)a[0];
1242 	while (len > 8) {
1243 		MUL_ADD_VEC_ROUND_PREFETCH(0);
1244 		MUL_ADD_VEC_ROUND_PREFETCH(1);
1245 		MUL_ADD_VEC_ROUND_PREFETCH(2);
1246 		MUL_ADD_VEC_ROUND_PREFETCH(3);
1247 		MUL_ADD_VEC_ROUND_PREFETCH(4);
1248 		MUL_ADD_VEC_ROUND_PREFETCH(5);
1249 		MUL_ADD_VEC_ROUND_PREFETCH(6);
1250 		MUL_ADD_VEC_ROUND_PREFETCH(7);
1251 		r += 8;
1252 		a += 8;
1253 		len -= 8;
1254 	}
1255 	if (len == 8) {
1256 		MUL_ADD_VEC_ROUND_PREFETCH(0);
1257 		MUL_ADD_VEC_ROUND_PREFETCH(1);
1258 		MUL_ADD_VEC_ROUND_PREFETCH(2);
1259 		MUL_ADD_VEC_ROUND_PREFETCH(3);
1260 		MUL_ADD_VEC_ROUND_PREFETCH(4);
1261 		MUL_ADD_VEC_ROUND_PREFETCH(5);
1262 		MUL_ADD_VEC_ROUND_PREFETCH(6);
1263 		MUL_ADD_VEC_ROUND_NOPREFETCH(7);
1264 		return ((uint32_t)cy);
1265 	}
1266 	while (len > 1) {
1267 		MUL_ADD_VEC_ROUND_PREFETCH(0);
1268 		++r;
1269 		++a;
1270 		--len;
1271 	}
1272 	if (len > 0) {
1273 		MUL_ADD_VEC_ROUND_NOPREFETCH(0);
1274 	}
1275 	return ((uint32_t)cy);
1276 }
1277 #endif /* UNROLL8 */
1278 
1279 void
1280 big_sqr_vec(uint32_t *r, uint32_t *a, int len)
1281 {
1282 	uint32_t	*tr, *ta;
1283 	int		tlen, row, col;
1284 	uint64_t	p, s, t, t2, cy;
1285 	uint32_t	d;
1286 
1287 	tr = r + 1;
1288 	ta = a;
1289 	tlen = len - 1;
1290 	tr[tlen] = big_mul_set_vec(tr, ta + 1, tlen, ta[0]);
1291 	while (--tlen > 0) {
1292 		tr += 2;
1293 		++ta;
1294 		tr[tlen] = big_mul_add_vec(tr, ta + 1, tlen, ta[0]);
1295 	}
1296 	s = (uint64_t)a[0];
1297 	s = s * s;
1298 	r[0] = (uint32_t)s;
1299 	cy = s >> 32;
1300 	p = ((uint64_t)r[1] << 1) + cy;
1301 	r[1] = (uint32_t)p;
1302 	cy = p >> 32;
1303 	row = 1;
1304 	col = 2;
1305 	while (row < len) {
1306 		s = (uint64_t)a[row];
1307 		s = s * s;
1308 		p = (uint64_t)r[col] << 1;
1309 		t = p + s;
1310 		d = (uint32_t)t;
1311 		t2 = (uint64_t)d + cy;
1312 		r[col] = (uint32_t)t2;
1313 		cy = (t >> 32) + (t2 >> 32);
1314 		if (row == len - 1)
1315 			break;
1316 		p = ((uint64_t)r[col + 1] << 1) + cy;
1317 		r[col + 1] = (uint32_t)p;
1318 		cy = p >> 32;
1319 		++row;
1320 		col += 2;
1321 	}
1322 	r[col + 1] = (uint32_t)cy;
1323 }
1324 
1325 #else /* BIG_CHUNK_SIZE == 64 */
1326 
1327 /*
1328  * r = r + a * digit, r and a are vectors of length len
1329  * returns the carry digit
1330  */
1331 BIG_CHUNK_TYPE
1332 big_mul_add_vec(BIG_CHUNK_TYPE *r, BIG_CHUNK_TYPE *a, int len,
1333     BIG_CHUNK_TYPE digit)
1334 {
1335 	BIG_CHUNK_TYPE	cy, cy1, retcy, dlow, dhigh;
1336 	int		i;
1337 
1338 	cy1 = 0;
1339 	dlow = digit & BIG_CHUNK_LOWHALFBITS;
1340 	dhigh = digit >> (BIG_CHUNK_SIZE / 2);
1341 	for (i = 0; i < len; i++) {
1342 		cy = (cy1 >> (BIG_CHUNK_SIZE / 2)) +
1343 		    dlow * (a[i] & BIG_CHUNK_LOWHALFBITS) +
1344 		    (r[i] & BIG_CHUNK_LOWHALFBITS);
1345 		cy1 = (cy >> (BIG_CHUNK_SIZE / 2)) +
1346 		    dlow * (a[i] >> (BIG_CHUNK_SIZE / 2)) +
1347 		    (r[i] >> (BIG_CHUNK_SIZE / 2));
1348 		r[i] = (cy & BIG_CHUNK_LOWHALFBITS) |
1349 		    (cy1 << (BIG_CHUNK_SIZE / 2));
1350 	}
1351 	retcy = cy1 >> (BIG_CHUNK_SIZE / 2);
1352 
1353 	cy1 = r[0] & BIG_CHUNK_LOWHALFBITS;
1354 	for (i = 0; i < len - 1; i++) {
1355 		cy = (cy1 >> (BIG_CHUNK_SIZE / 2)) +
1356 		    dhigh * (a[i] & BIG_CHUNK_LOWHALFBITS) +
1357 		    (r[i] >> (BIG_CHUNK_SIZE / 2));
1358 		r[i] = (cy1 & BIG_CHUNK_LOWHALFBITS) |
1359 		    (cy << (BIG_CHUNK_SIZE / 2));
1360 		cy1 = (cy >> (BIG_CHUNK_SIZE / 2)) +
1361 		    dhigh * (a[i] >> (BIG_CHUNK_SIZE / 2)) +
1362 		    (r[i + 1] & BIG_CHUNK_LOWHALFBITS);
1363 	}
1364 	cy = (cy1 >> (BIG_CHUNK_SIZE / 2)) +
1365 	    dhigh * (a[len - 1] & BIG_CHUNK_LOWHALFBITS) +
1366 	    (r[len - 1] >> (BIG_CHUNK_SIZE / 2));
1367 	r[len - 1] = (cy1 & BIG_CHUNK_LOWHALFBITS) |
1368 	    (cy << (BIG_CHUNK_SIZE / 2));
1369 	retcy = (cy >> (BIG_CHUNK_SIZE / 2)) +
1370 	    dhigh * (a[len - 1] >> (BIG_CHUNK_SIZE / 2)) + retcy;
1371 
1372 	return (retcy);
1373 }
1374 
1375 
1376 /*
1377  * r = a * digit, r and a are vectors of length len
1378  * returns the carry digit
1379  */
1380 BIG_CHUNK_TYPE
1381 big_mul_set_vec(BIG_CHUNK_TYPE *r, BIG_CHUNK_TYPE *a, int len,
1382     BIG_CHUNK_TYPE digit)
1383 {
1384 	int	i;
1385 
1386 	ASSERT(r != a);
1387 	for (i = 0; i < len; i++) {
1388 		r[i] = 0;
1389 	}
1390 	return (big_mul_add_vec(r, a, len, digit));
1391 }
1392 
1393 void
1394 big_sqr_vec(BIG_CHUNK_TYPE *r, BIG_CHUNK_TYPE *a, int len)
1395 {
1396 	int i;
1397 
1398 	ASSERT(r != a);
1399 	r[len] = big_mul_set_vec(r, a, len, a[0]);
1400 	for (i = 1; i < len; ++i)
1401 		r[len + i] = big_mul_add_vec(r + i, a, len, a[i]);
1402 }
1403 
1404 #endif /* BIG_CHUNK_SIZE == 32/64 */
1405 
1406 
1407 #else /* ! UMUL64 */
1408 
1409 #if (BIG_CHUNK_SIZE != 32)
1410 #error Don't use 64-bit chunks without defining UMUL64
1411 #endif
1412 
1413 
1414 /*
1415  * r = r + a * digit, r and a are vectors of length len
1416  * returns the carry digit
1417  */
1418 uint32_t
1419 big_mul_add_vec(uint32_t *r, uint32_t *a, int len, uint32_t digit)
1420 {
1421 	uint32_t cy, cy1, retcy, dlow, dhigh;
1422 	int i;
1423 
1424 	cy1 = 0;
1425 	dlow = digit & 0xffff;
1426 	dhigh = digit >> 16;
1427 	for (i = 0; i < len; i++) {
1428 		cy = (cy1 >> 16) + dlow * (a[i] & 0xffff) + (r[i] & 0xffff);
1429 		cy1 = (cy >> 16) + dlow * (a[i]>>16) + (r[i] >> 16);
1430 		r[i] = (cy & 0xffff) | (cy1 << 16);
1431 	}
1432 	retcy = cy1 >> 16;
1433 
1434 	cy1 = r[0] & 0xffff;
1435 	for (i = 0; i < len - 1; i++) {
1436 		cy = (cy1 >> 16) + dhigh * (a[i] & 0xffff) + (r[i] >> 16);
1437 		r[i] = (cy1 & 0xffff) | (cy << 16);
1438 		cy1 = (cy >> 16) + dhigh * (a[i] >> 16) + (r[i + 1] & 0xffff);
1439 	}
1440 	cy = (cy1 >> 16) + dhigh * (a[len - 1] & 0xffff) + (r[len - 1] >> 16);
1441 	r[len - 1] = (cy1 & 0xffff) | (cy << 16);
1442 	retcy = (cy >> 16) + dhigh * (a[len - 1] >> 16) + retcy;
1443 
1444 	return (retcy);
1445 }
1446 
1447 
1448 /*
1449  * r = a * digit, r and a are vectors of length len
1450  * returns the carry digit
1451  */
1452 uint32_t
1453 big_mul_set_vec(uint32_t *r, uint32_t *a, int len, uint32_t digit)
1454 {
1455 	int	i;
1456 
1457 	ASSERT(r != a);
1458 	for (i = 0; i < len; i++) {
1459 		r[i] = 0;
1460 	}
1461 
1462 	return (big_mul_add_vec(r, a, len, digit));
1463 }
1464 
1465 void
1466 big_sqr_vec(uint32_t *r, uint32_t *a, int len)
1467 {
1468 	int i;
1469 
1470 	ASSERT(r != a);
1471 	r[len] = big_mul_set_vec(r, a, len, a[0]);
1472 	for (i = 1; i < len; ++i)
1473 		r[len + i] = big_mul_add_vec(r + i, a, len, a[i]);
1474 }
1475 
1476 #endif /* UMUL64 */
1477 
1478 void
1479 big_mul_vec(BIG_CHUNK_TYPE *r, BIG_CHUNK_TYPE *a, int alen,
1480     BIG_CHUNK_TYPE *b, int blen)
1481 {
1482 	int i;
1483 
1484 	r[alen] = big_mul_set_vec(r, a, alen, b[0]);
1485 	for (i = 1; i < blen; ++i)
1486 		r[alen + i] = big_mul_add_vec(r + i, a, alen, b[i]);
1487 }
1488 
1489 
1490 #endif /* ! PSR_MUL */
1491 
1492 
1493 /*
1494  * result = aa * bb  result->value should be big enough to hold the result
1495  *
1496  * Implementation: Standard grammar school algorithm
1497  *
1498  */
1499 BIG_ERR_CODE
1500 big_mul(BIGNUM *result, BIGNUM *aa, BIGNUM *bb)
1501 {
1502 	BIGNUM		tmp1;
1503 	BIG_CHUNK_TYPE	tmp1value[BIGTMPSIZE];
1504 	BIG_CHUNK_TYPE	*r, *t, *a, *b;
1505 	BIG_ERR_CODE	err;
1506 	int		i, alen, blen, rsize, sign, diff;
1507 
1508 	if (aa == bb) {
1509 		diff = 0;
1510 	} else {
1511 		diff = big_cmp_abs(aa, bb);
1512 		if (diff < 0) {
1513 			BIGNUM *tt;
1514 			tt = aa;
1515 			aa = bb;
1516 			bb = tt;
1517 		}
1518 	}
1519 	a = aa->value;
1520 	b = bb->value;
1521 	alen = aa->len;
1522 	blen = bb->len;
1523 	while ((alen > 1) && (a[alen - 1] == 0)) {
1524 		alen--;
1525 	}
1526 	aa->len = alen;
1527 	while ((blen > 1) && (b[blen - 1] == 0)) {
1528 		blen--;
1529 	}
1530 	bb->len = blen;
1531 
1532 	rsize = alen + blen;
1533 	ASSERT(rsize > 0);
1534 	if (result->size < rsize) {
1535 		err = big_extend(result, rsize);
1536 		if (err != BIG_OK) {
1537 			return (err);
1538 		}
1539 		/* aa or bb might be an alias to result */
1540 		a = aa->value;
1541 		b = bb->value;
1542 	}
1543 	r = result->value;
1544 
1545 	if (((alen == 1) && (a[0] == 0)) || ((blen == 1) && (b[0] == 0))) {
1546 		result->len = 1;
1547 		result->sign = 1;
1548 		r[0] = 0;
1549 		return (BIG_OK);
1550 	}
1551 	sign = aa->sign * bb->sign;
1552 	if ((alen == 1) && (a[0] == 1)) {
1553 		for (i = 0; i < blen; i++) {
1554 			r[i] = b[i];
1555 		}
1556 		result->len = blen;
1557 		result->sign = sign;
1558 		return (BIG_OK);
1559 	}
1560 	if ((blen == 1) && (b[0] == 1)) {
1561 		for (i = 0; i < alen; i++) {
1562 			r[i] = a[i];
1563 		}
1564 		result->len = alen;
1565 		result->sign = sign;
1566 		return (BIG_OK);
1567 	}
1568 
1569 	if ((err = big_init1(&tmp1, rsize,
1570 	    tmp1value, arraysize(tmp1value))) != BIG_OK) {
1571 		return (err);
1572 	}
1573 	t = tmp1.value;
1574 
1575 	for (i = 0; i < rsize; i++) {
1576 		t[i] = 0;
1577 	}
1578 
1579 	if (diff == 0 && alen > 2) {
1580 		BIG_SQR_VEC(t, a, alen);
1581 	} else if (blen > 0) {
1582 		BIG_MUL_VEC(t, a, alen, b, blen);
1583 	}
1584 
1585 	if (t[rsize - 1] == 0) {
1586 		tmp1.len = rsize - 1;
1587 	} else {
1588 		tmp1.len = rsize;
1589 	}
1590 
1591 	err = big_copy(result, &tmp1);
1592 
1593 	result->sign = sign;
1594 
1595 	big_finish(&tmp1);
1596 
1597 	return (err);
1598 }
1599 
1600 
1601 /*
1602  * caller must ensure that  a < n,  b < n  and  ret->size >=  2 * n->len + 1
1603  * and that ret is not n
1604  */
1605 BIG_ERR_CODE
1606 big_mont_mul(BIGNUM *ret, BIGNUM *a, BIGNUM *b, BIGNUM *n, BIG_CHUNK_TYPE n0)
1607 {
1608 	int	i, j, nlen, needsubtract;
1609 	BIG_CHUNK_TYPE	*nn, *rr;
1610 	BIG_CHUNK_TYPE	digit, c;
1611 	BIG_ERR_CODE	err;
1612 
1613 	nlen = n->len;
1614 	nn = n->value;
1615 
1616 	rr = ret->value;
1617 
1618 	if ((err = big_mul(ret, a, b)) != BIG_OK) {
1619 		return (err);
1620 	}
1621 
1622 	rr = ret->value;
1623 	for (i = ret->len; i < 2 * nlen + 1; i++) {
1624 		rr[i] = 0;
1625 	}
1626 	for (i = 0; i < nlen; i++) {
1627 		digit = rr[i];
1628 		digit = digit * n0;
1629 
1630 		c = BIG_MUL_ADD_VEC(rr + i, nn, nlen, digit);
1631 		j = i + nlen;
1632 		rr[j] += c;
1633 		while (rr[j] < c) {
1634 			rr[j + 1] += 1;
1635 			j++;
1636 			c = 1;
1637 		}
1638 	}
1639 
1640 	needsubtract = 0;
1641 	if ((rr[2 * nlen]  != 0))
1642 		needsubtract = 1;
1643 	else {
1644 		for (i = 2 * nlen - 1; i >= nlen; i--) {
1645 			if (rr[i] > nn[i - nlen]) {
1646 				needsubtract = 1;
1647 				break;
1648 			} else if (rr[i] < nn[i - nlen]) {
1649 				break;
1650 			}
1651 		}
1652 	}
1653 	if (needsubtract)
1654 		big_sub_vec(rr, rr + nlen, nn, nlen);
1655 	else {
1656 		for (i = 0; i < nlen; i++) {
1657 			rr[i] = rr[i + nlen];
1658 		}
1659 	}
1660 
1661 	/* Remove leading zeros, but keep at least 1 digit: */
1662 	for (i = nlen - 1; (i > 0) && (rr[i] == 0); i--)
1663 		;
1664 	ret->len = i + 1;
1665 
1666 	return (BIG_OK);
1667 }
1668 
1669 
1670 BIG_CHUNK_TYPE
1671 big_n0(BIG_CHUNK_TYPE n)
1672 {
1673 	int		i;
1674 	BIG_CHUNK_TYPE	result, tmp;
1675 
1676 	result = 0;
1677 	tmp = BIG_CHUNK_ALLBITS;
1678 	for (i = 0; i < BIG_CHUNK_SIZE; i++) {
1679 		if ((tmp & 1) == 1) {
1680 			result = (result >> 1) | BIG_CHUNK_HIGHBIT;
1681 			tmp = tmp - n;
1682 		} else {
1683 			result = (result >> 1);
1684 		}
1685 		tmp = tmp >> 1;
1686 	}
1687 
1688 	return (result);
1689 }
1690 
1691 
1692 int
1693 big_numbits(BIGNUM *n)
1694 {
1695 	int		i, j;
1696 	BIG_CHUNK_TYPE	t;
1697 
1698 	for (i = n->len - 1; i > 0; i--) {
1699 		if (n->value[i] != 0) {
1700 			break;
1701 		}
1702 	}
1703 	t = n->value[i];
1704 	for (j = BIG_CHUNK_SIZE; j > 0; j--) {
1705 		if ((t & BIG_CHUNK_HIGHBIT) == 0) {
1706 			t = t << 1;
1707 		} else {
1708 			return (BIG_CHUNK_SIZE * i + j);
1709 		}
1710 	}
1711 	return (0);
1712 }
1713 
1714 
1715 /* caller must make sure that a < n */
1716 BIG_ERR_CODE
1717 big_mont_rr(BIGNUM *result, BIGNUM *n)
1718 {
1719 	BIGNUM		rr;
1720 	BIG_CHUNK_TYPE	rrvalue[BIGTMPSIZE];
1721 	int		len, i;
1722 	BIG_ERR_CODE	err;
1723 
1724 	rr.malloced = 0;
1725 	len = n->len;
1726 
1727 	if ((err = big_init1(&rr, 2 * len + 1,
1728 	    rrvalue, arraysize(rrvalue))) != BIG_OK) {
1729 		return (err);
1730 	}
1731 
1732 	for (i = 0; i < 2 * len; i++) {
1733 		rr.value[i] = 0;
1734 	}
1735 	rr.value[2 * len] = 1;
1736 	rr.len = 2 * len + 1;
1737 	if ((err = big_div_pos(NULL, &rr, &rr, n)) != BIG_OK) {
1738 		goto ret;
1739 	}
1740 	err = big_copy(result, &rr);
1741 ret:
1742 	big_finish(&rr);
1743 	return (err);
1744 }
1745 
1746 
1747 /* caller must make sure that a < n */
1748 BIG_ERR_CODE
1749 big_mont_conv(BIGNUM *result, BIGNUM *a, BIGNUM *n, BIG_CHUNK_TYPE n0,
1750     BIGNUM *n_rr)
1751 {
1752 	BIGNUM		rr;
1753 	BIG_CHUNK_TYPE	rrvalue[BIGTMPSIZE];
1754 	int		len, i;
1755 	BIG_ERR_CODE	err;
1756 
1757 	rr.malloced = 0;
1758 	len = n->len;
1759 
1760 	if ((err = big_init1(&rr, 2 * len + 1, rrvalue, arraysize(rrvalue)))
1761 	    != BIG_OK) {
1762 		return (err);
1763 	}
1764 
1765 	if (n_rr == NULL) {
1766 		for (i = 0; i < 2 * len; i++) {
1767 			rr.value[i] = 0;
1768 		}
1769 		rr.value[2 * len] = 1;
1770 		rr.len = 2 * len + 1;
1771 		if ((err = big_div_pos(NULL, &rr, &rr, n)) != BIG_OK) {
1772 			goto ret;
1773 		}
1774 		n_rr = &rr;
1775 	}
1776 
1777 	if ((err = big_mont_mul(&rr, n_rr, a, n, n0)) != BIG_OK) {
1778 		goto ret;
1779 	}
1780 	err = big_copy(result, &rr);
1781 
1782 ret:
1783 	big_finish(&rr);
1784 	return (err);
1785 }
1786 
1787 
1788 #ifdef	USE_FLOATING_POINT
1789 #define	big_modexp_ncp_float	big_modexp_ncp_sw
1790 #else
1791 #define	big_modexp_ncp_int	big_modexp_ncp_sw
1792 #endif
1793 
1794 #define	MAX_EXP_BIT_GROUP_SIZE 6
1795 #define	APOWERS_MAX_SIZE (1 << (MAX_EXP_BIT_GROUP_SIZE - 1))
1796 
1797 /* ARGSUSED */
1798 static BIG_ERR_CODE
1799 big_modexp_ncp_int(BIGNUM *result, BIGNUM *ma, BIGNUM *e, BIGNUM *n,
1800     BIGNUM *tmp, BIG_CHUNK_TYPE n0)
1801 
1802 {
1803 	BIGNUM		apowers[APOWERS_MAX_SIZE];
1804 	BIGNUM		tmp1;
1805 	BIG_CHUNK_TYPE	tmp1value[BIGTMPSIZE];
1806 	int		i, j, k, l, m, p;
1807 	uint32_t	bit, bitind, bitcount, groupbits, apowerssize;
1808 	uint32_t	nbits;
1809 	BIG_ERR_CODE	err;
1810 
1811 	nbits = big_numbits(e);
1812 	if (nbits < 50) {
1813 		groupbits = 1;
1814 		apowerssize = 1;
1815 	} else {
1816 		groupbits = MAX_EXP_BIT_GROUP_SIZE;
1817 		apowerssize = 1 << (groupbits - 1);
1818 	}
1819 
1820 
1821 	if ((err = big_init1(&tmp1, 2 * n->len + 1,
1822 	    tmp1value, arraysize(tmp1value))) != BIG_OK) {
1823 		return (err);
1824 	}
1825 
1826 	/* clear the malloced bit to help cleanup */
1827 	for (i = 0; i < apowerssize; i++) {
1828 		apowers[i].malloced = 0;
1829 	}
1830 
1831 	for (i = 0; i < apowerssize; i++) {
1832 		if ((err = big_init1(&(apowers[i]), n->len, NULL, 0)) !=
1833 		    BIG_OK) {
1834 			goto ret;
1835 		}
1836 	}
1837 
1838 	(void) big_copy(&(apowers[0]), ma);
1839 
1840 	if ((err = big_mont_mul(&tmp1, ma, ma, n, n0)) != BIG_OK) {
1841 		goto ret;
1842 	}
1843 	(void) big_copy(ma, &tmp1);
1844 
1845 	for (i = 1; i < apowerssize; i++) {
1846 		if ((err = big_mont_mul(&tmp1, ma,
1847 		    &(apowers[i - 1]), n, n0)) != BIG_OK) {
1848 			goto ret;
1849 		}
1850 		(void) big_copy(&apowers[i], &tmp1);
1851 	}
1852 
1853 	bitind = nbits % BIG_CHUNK_SIZE;
1854 	k = 0;
1855 	l = 0;
1856 	p = 0;
1857 	bitcount = 0;
1858 	for (i = nbits / BIG_CHUNK_SIZE; i >= 0; i--) {
1859 		for (j = bitind - 1; j >= 0; j--) {
1860 			bit = (e->value[i] >> j) & 1;
1861 			if ((bitcount == 0) && (bit == 0)) {
1862 				if ((err = big_mont_mul(tmp,
1863 				    tmp, tmp, n, n0)) != BIG_OK) {
1864 					goto ret;
1865 				}
1866 			} else {
1867 				bitcount++;
1868 				p = p * 2 + bit;
1869 				if (bit == 1) {
1870 					k = k + l + 1;
1871 					l = 0;
1872 				} else {
1873 					l++;
1874 				}
1875 				if (bitcount == groupbits) {
1876 					for (m = 0; m < k; m++) {
1877 						if ((err = big_mont_mul(tmp,
1878 						    tmp, tmp, n, n0)) !=
1879 						    BIG_OK) {
1880 							goto ret;
1881 						}
1882 					}
1883 					if ((err = big_mont_mul(tmp, tmp,
1884 					    &(apowers[p >> (l + 1)]),
1885 					    n, n0)) != BIG_OK) {
1886 						goto ret;
1887 					}
1888 					for (m = 0; m < l; m++) {
1889 						if ((err = big_mont_mul(tmp,
1890 						    tmp, tmp, n, n0)) !=
1891 						    BIG_OK) {
1892 							goto ret;
1893 						}
1894 					}
1895 					k = 0;
1896 					l = 0;
1897 					p = 0;
1898 					bitcount = 0;
1899 				}
1900 			}
1901 		}
1902 		bitind = BIG_CHUNK_SIZE;
1903 	}
1904 
1905 	for (m = 0; m < k; m++) {
1906 		if ((err = big_mont_mul(tmp, tmp, tmp, n, n0)) != BIG_OK) {
1907 			goto ret;
1908 		}
1909 	}
1910 	if (p != 0) {
1911 		if ((err = big_mont_mul(tmp, tmp,
1912 		    &(apowers[p >> (l + 1)]), n, n0)) != BIG_OK) {
1913 			goto ret;
1914 		}
1915 	}
1916 	for (m = 0; m < l; m++) {
1917 		if ((err = big_mont_mul(result, tmp, tmp, n, n0)) != BIG_OK) {
1918 			goto ret;
1919 		}
1920 	}
1921 
1922 ret:
1923 	for (i = apowerssize - 1; i >= 0; i--) {
1924 		big_finish(&(apowers[i]));
1925 	}
1926 	big_finish(&tmp1);
1927 
1928 	return (err);
1929 }
1930 
1931 
1932 #ifdef USE_FLOATING_POINT
1933 
1934 #ifdef _KERNEL
1935 
1936 #include <sys/sysmacros.h>
1937 #include <sys/regset.h>
1938 #include <sys/fpu/fpusystm.h>
1939 
1940 /* the alignment for block stores to save fp registers */
1941 #define	FPR_ALIGN	(64)
1942 
1943 extern void big_savefp(kfpu_t *);
1944 extern void big_restorefp(kfpu_t *);
1945 
1946 #endif /* _KERNEL */
1947 
1948 /*
1949  * This version makes use of floating point for performance
1950  */
1951 static BIG_ERR_CODE
1952 big_modexp_ncp_float(BIGNUM *result, BIGNUM *ma, BIGNUM *e, BIGNUM *n,
1953     BIGNUM *tmp, BIG_CHUNK_TYPE n0)
1954 {
1955 
1956 	int		i, j, k, l, m, p;
1957 	uint32_t	bit, bitind, bitcount, nlen;
1958 	double		dn0;
1959 	double		*dn, *dt, *d16r, *d32r;
1960 	uint32_t	*nint, *prod;
1961 	double		*apowers[APOWERS_MAX_SIZE];
1962 	uint32_t	nbits, groupbits, apowerssize;
1963 	BIG_ERR_CODE	err = BIG_OK;
1964 
1965 #ifdef _KERNEL
1966 	uint8_t fpua[sizeof (kfpu_t) + FPR_ALIGN];
1967 	kfpu_t *fpu;
1968 
1969 #ifdef DEBUG
1970 	if (!fpu_exists)
1971 		return (BIG_GENERAL_ERR);
1972 #endif
1973 
1974 	fpu =  (kfpu_t *)P2ROUNDUP((uintptr_t)fpua, FPR_ALIGN);
1975 	big_savefp(fpu);
1976 
1977 #endif /* _KERNEL */
1978 
1979 	nbits = big_numbits(e);
1980 	if (nbits < 50) {
1981 		groupbits = 1;
1982 		apowerssize = 1;
1983 	} else {
1984 		groupbits = MAX_EXP_BIT_GROUP_SIZE;
1985 		apowerssize = 1 << (groupbits - 1);
1986 	}
1987 
1988 	nlen = (BIG_CHUNK_SIZE / 32) * n->len;
1989 	dn0 = (double)(n0 & 0xffff);
1990 
1991 	dn = dt = d16r = d32r = NULL;
1992 	nint = prod = NULL;
1993 	for (i = 0; i < apowerssize; i++) {
1994 		apowers[i] = NULL;
1995 	}
1996 
1997 	if ((dn = big_malloc(nlen * sizeof (double))) == NULL) {
1998 		err = BIG_NO_MEM;
1999 		goto ret;
2000 	}
2001 	if ((dt = big_malloc((4 * nlen + 2) * sizeof (double))) == NULL) {
2002 		err = BIG_NO_MEM;
2003 		goto ret;
2004 	}
2005 	if ((nint = big_malloc(nlen * sizeof (uint32_t))) == NULL) {
2006 		err = BIG_NO_MEM;
2007 		goto ret;
2008 	}
2009 	if ((prod = big_malloc((nlen + 1) * sizeof (uint32_t))) == NULL) {
2010 		err = BIG_NO_MEM;
2011 		goto ret;
2012 	}
2013 	if ((d16r = big_malloc((2 * nlen + 1) * sizeof (double))) == NULL) {
2014 		err = BIG_NO_MEM;
2015 		goto ret;
2016 	}
2017 	if ((d32r = big_malloc(nlen * sizeof (double))) == NULL) {
2018 		err = BIG_NO_MEM;
2019 		goto ret;
2020 	}
2021 	for (i = 0; i < apowerssize; i++) {
2022 		if ((apowers[i] = big_malloc((2 * nlen + 1) *
2023 		    sizeof (double))) == NULL) {
2024 			err = BIG_NO_MEM;
2025 			goto ret;
2026 		}
2027 	}
2028 
2029 #if (BIG_CHUNK_SIZE == 32)
2030 	for (i = 0; i < ma->len; i++) {
2031 		nint[i] = ma->value[i];
2032 	}
2033 	for (; i < nlen; i++) {
2034 		nint[i] = 0;
2035 	}
2036 #else
2037 	for (i = 0; i < ma->len; i++) {
2038 		nint[2 * i] = (uint32_t)(ma->value[i] & 0xffffffffULL);
2039 		nint[2 * i + 1] = (uint32_t)(ma->value[i] >> 32);
2040 	}
2041 	for (i = ma->len * 2; i < nlen; i++) {
2042 		nint[i] = 0;
2043 	}
2044 #endif
2045 	conv_i32_to_d32_and_d16(d32r, apowers[0], nint, nlen);
2046 
2047 #if (BIG_CHUNK_SIZE == 32)
2048 	for (i = 0; i < n->len; i++) {
2049 		nint[i] = n->value[i];
2050 	}
2051 	for (; i < nlen; i++) {
2052 		nint[i] = 0;
2053 	}
2054 #else
2055 	for (i = 0; i < n->len; i++) {
2056 		nint[2 * i] = (uint32_t)(n->value[i] & 0xffffffffULL);
2057 		nint[2 * i + 1] = (uint32_t)(n->value[i] >> 32);
2058 	}
2059 	for (i = n->len * 2; i < nlen; i++) {
2060 		nint[i] = 0;
2061 	}
2062 #endif
2063 	conv_i32_to_d32(dn, nint, nlen);
2064 
2065 	mont_mulf_noconv(prod, d32r, apowers[0], dt, dn, nint, nlen, dn0);
2066 	conv_i32_to_d32(d32r, prod, nlen);
2067 	for (i = 1; i < apowerssize; i++) {
2068 		mont_mulf_noconv(prod, d32r, apowers[i - 1],
2069 		    dt, dn, nint, nlen, dn0);
2070 		conv_i32_to_d16(apowers[i], prod, nlen);
2071 	}
2072 
2073 #if (BIG_CHUNK_SIZE == 32)
2074 	for (i = 0; i < tmp->len; i++) {
2075 		prod[i] = tmp->value[i];
2076 	}
2077 	for (; i < nlen + 1; i++) {
2078 		prod[i] = 0;
2079 	}
2080 #else
2081 	for (i = 0; i < tmp->len; i++) {
2082 		prod[2 * i] = (uint32_t)(tmp->value[i] & 0xffffffffULL);
2083 		prod[2 * i + 1] = (uint32_t)(tmp->value[i] >> 32);
2084 	}
2085 	for (i = tmp->len * 2; i < nlen + 1; i++) {
2086 		prod[i] = 0;
2087 	}
2088 #endif
2089 
2090 	bitind = nbits % BIG_CHUNK_SIZE;
2091 	k = 0;
2092 	l = 0;
2093 	p = 0;
2094 	bitcount = 0;
2095 	for (i = nbits / BIG_CHUNK_SIZE; i >= 0; i--) {
2096 		for (j = bitind - 1; j >= 0; j--) {
2097 			bit = (e->value[i] >> j) & 1;
2098 			if ((bitcount == 0) && (bit == 0)) {
2099 				conv_i32_to_d32_and_d16(d32r, d16r,
2100 				    prod, nlen);
2101 				mont_mulf_noconv(prod, d32r, d16r,
2102 				    dt, dn, nint, nlen, dn0);
2103 			} else {
2104 				bitcount++;
2105 				p = p * 2 + bit;
2106 				if (bit == 1) {
2107 					k = k + l + 1;
2108 					l = 0;
2109 				} else {
2110 					l++;
2111 				}
2112 				if (bitcount == groupbits) {
2113 					for (m = 0; m < k; m++) {
2114 						conv_i32_to_d32_and_d16(d32r,
2115 						    d16r, prod, nlen);
2116 						mont_mulf_noconv(prod, d32r,
2117 						    d16r, dt, dn, nint,
2118 						    nlen, dn0);
2119 					}
2120 					conv_i32_to_d32(d32r, prod, nlen);
2121 					mont_mulf_noconv(prod, d32r,
2122 					    apowers[p >> (l + 1)],
2123 					    dt, dn, nint, nlen, dn0);
2124 					for (m = 0; m < l; m++) {
2125 						conv_i32_to_d32_and_d16(d32r,
2126 						    d16r, prod, nlen);
2127 						mont_mulf_noconv(prod, d32r,
2128 						    d16r, dt, dn, nint,
2129 						    nlen, dn0);
2130 					}
2131 					k = 0;
2132 					l = 0;
2133 					p = 0;
2134 					bitcount = 0;
2135 				}
2136 			}
2137 		}
2138 		bitind = BIG_CHUNK_SIZE;
2139 	}
2140 
2141 	for (m = 0; m < k; m++) {
2142 		conv_i32_to_d32_and_d16(d32r, d16r, prod, nlen);
2143 		mont_mulf_noconv(prod, d32r, d16r, dt, dn, nint, nlen, dn0);
2144 	}
2145 	if (p != 0) {
2146 		conv_i32_to_d32(d32r, prod, nlen);
2147 		mont_mulf_noconv(prod, d32r, apowers[p >> (l + 1)],
2148 		    dt, dn, nint, nlen, dn0);
2149 	}
2150 	for (m = 0; m < l; m++) {
2151 		conv_i32_to_d32_and_d16(d32r, d16r, prod, nlen);
2152 		mont_mulf_noconv(prod, d32r, d16r, dt, dn, nint, nlen, dn0);
2153 	}
2154 
2155 #if (BIG_CHUNK_SIZE == 32)
2156 	for (i = 0; i < nlen; i++) {
2157 		result->value[i] = prod[i];
2158 	}
2159 	for (i = nlen - 1; (i > 0) && (prod[i] == 0); i--)
2160 		;
2161 #else
2162 	for (i = 0; i < nlen / 2; i++) {
2163 		result->value[i] = (uint64_t)(prod[2 * i]) +
2164 		    (((uint64_t)(prod[2 * i + 1])) << 32);
2165 	}
2166 	for (i = nlen / 2 - 1; (i > 0) && (result->value[i] == 0); i--)
2167 		;
2168 #endif
2169 	result->len = i + 1;
2170 
2171 ret:
2172 	for (i = apowerssize - 1; i >= 0; i--) {
2173 		if (apowers[i] != NULL)
2174 			big_free(apowers[i], (2 * nlen + 1) * sizeof (double));
2175 	}
2176 	if (d32r != NULL) {
2177 		big_free(d32r, nlen * sizeof (double));
2178 	}
2179 	if (d16r != NULL) {
2180 		big_free(d16r, (2 * nlen + 1) * sizeof (double));
2181 	}
2182 	if (prod != NULL) {
2183 		big_free(prod, (nlen + 1) * sizeof (uint32_t));
2184 	}
2185 	if (nint != NULL) {
2186 		big_free(nint, nlen * sizeof (uint32_t));
2187 	}
2188 	if (dt != NULL) {
2189 		big_free(dt, (4 * nlen + 2) * sizeof (double));
2190 	}
2191 	if (dn != NULL) {
2192 		big_free(dn, nlen * sizeof (double));
2193 	}
2194 
2195 #ifdef _KERNEL
2196 	big_restorefp(fpu);
2197 #endif
2198 
2199 	return (err);
2200 }
2201 
2202 #endif /* USE_FLOATING_POINT */
2203 
2204 
2205 BIG_ERR_CODE
2206 big_modexp_ext(BIGNUM *result, BIGNUM *a, BIGNUM *e, BIGNUM *n, BIGNUM *n_rr,
2207     big_modexp_ncp_info_t *info)
2208 {
2209 	BIGNUM		ma, tmp, rr;
2210 	BIG_CHUNK_TYPE	mavalue[BIGTMPSIZE];
2211 	BIG_CHUNK_TYPE	tmpvalue[BIGTMPSIZE];
2212 	BIG_CHUNK_TYPE	rrvalue[BIGTMPSIZE];
2213 	BIG_ERR_CODE	err;
2214 	BIG_CHUNK_TYPE	n0;
2215 
2216 	if ((err = big_init1(&ma, n->len, mavalue, arraysize(mavalue)))	!=
2217 	    BIG_OK) {
2218 		return (err);
2219 	}
2220 	ma.len = 1;
2221 	ma.value[0] = 0;
2222 
2223 	if ((err = big_init1(&tmp, 2 * n->len + 1,
2224 	    tmpvalue, arraysize(tmpvalue))) != BIG_OK) {
2225 		goto ret1;
2226 	}
2227 
2228 	/* clear the malloced bit to help cleanup */
2229 	rr.malloced = 0;
2230 
2231 	if (n_rr == NULL) {
2232 		if ((err = big_init1(&rr, 2 * n->len + 1,
2233 		    rrvalue, arraysize(rrvalue))) != BIG_OK) {
2234 			goto ret2;
2235 		}
2236 		if (big_mont_rr(&rr, n) != BIG_OK) {
2237 			goto ret;
2238 		}
2239 		n_rr = &rr;
2240 	}
2241 
2242 	n0 = big_n0(n->value[0]);
2243 
2244 	if (big_cmp_abs(a, n) > 0) {
2245 		if ((err = big_div_pos(NULL, &ma, a, n)) != BIG_OK) {
2246 			goto ret;
2247 		}
2248 		err = big_mont_conv(&ma, &ma, n, n0, n_rr);
2249 	} else {
2250 		err = big_mont_conv(&ma, a, n, n0, n_rr);
2251 	}
2252 	if (err != BIG_OK) {
2253 		goto ret;
2254 	}
2255 
2256 	tmp.len = 1;
2257 	tmp.value[0] = 1;
2258 	if ((err = big_mont_conv(&tmp, &tmp, n, n0, n_rr)) != BIG_OK) {
2259 		goto ret;
2260 	}
2261 
2262 	if ((info != NULL) && (info->func != NULL)) {
2263 		err = (*(info->func))(&tmp, &ma, e, n, &tmp, n0,
2264 		    info->ncp, info->reqp);
2265 	} else {
2266 		err = big_modexp_ncp_sw(&tmp, &ma, e, n, &tmp, n0);
2267 	}
2268 	if (err != BIG_OK) {
2269 		goto ret;
2270 	}
2271 
2272 	ma.value[0] = 1;
2273 	ma.len = 1;
2274 	if ((err = big_mont_mul(&tmp, &tmp, &ma, n, n0)) != BIG_OK) {
2275 		goto ret;
2276 	}
2277 	err = big_copy(result, &tmp);
2278 
2279 ret:
2280 	if (rr.malloced) {
2281 		big_finish(&rr);
2282 	}
2283 ret2:
2284 	big_finish(&tmp);
2285 ret1:
2286 	big_finish(&ma);
2287 
2288 	return (err);
2289 }
2290 
2291 BIG_ERR_CODE
2292 big_modexp(BIGNUM *result, BIGNUM *a, BIGNUM *e, BIGNUM *n, BIGNUM *n_rr)
2293 {
2294 	return (big_modexp_ext(result, a, e, n, n_rr, NULL));
2295 }
2296 
2297 
2298 BIG_ERR_CODE
2299 big_modexp_crt_ext(BIGNUM *result, BIGNUM *a, BIGNUM *dmodpminus1,
2300     BIGNUM *dmodqminus1, BIGNUM *p, BIGNUM *q, BIGNUM *pinvmodq,
2301     BIGNUM *p_rr, BIGNUM *q_rr, big_modexp_ncp_info_t *info)
2302 {
2303 	BIGNUM		ap, aq, tmp;
2304 	int		alen, biglen, sign;
2305 	BIG_ERR_CODE	err;
2306 
2307 	if (p->len > q->len) {
2308 		biglen = p->len;
2309 	} else {
2310 		biglen = q->len;
2311 	}
2312 
2313 	if ((err = big_init1(&ap, p->len, NULL, 0)) != BIG_OK) {
2314 		return (err);
2315 	}
2316 	if ((err = big_init1(&aq, q->len, NULL, 0)) != BIG_OK) {
2317 		goto ret1;
2318 	}
2319 	if ((err = big_init1(&tmp, biglen + q->len + 1, NULL, 0)) != BIG_OK) {
2320 		goto ret2;
2321 	}
2322 
2323 	/*
2324 	 * check whether a is too short - to avoid timing attacks
2325 	 */
2326 	alen = a->len;
2327 	while ((alen > p->len) && (a->value[alen - 1] == 0)) {
2328 		alen--;
2329 	}
2330 	if (alen < p->len + q->len) {
2331 		/*
2332 		 * a is too short, add p*q to it before
2333 		 * taking it modulo p and q
2334 		 * this will also affect timing, but this difference
2335 		 * does not depend on p or q, only on a
2336 		 * (in "normal" operation, this path will never be
2337 		 * taken, so it is not a performance penalty
2338 		 */
2339 		if ((err = big_mul(&tmp, p, q)) != BIG_OK) {
2340 			goto ret;
2341 		}
2342 		if ((err = big_add(&tmp, &tmp, a)) != BIG_OK) {
2343 			goto ret;
2344 		}
2345 		if ((err = big_div_pos(NULL, &ap, &tmp, p)) != BIG_OK) {
2346 			goto ret;
2347 		}
2348 		if ((err = big_div_pos(NULL, &aq, &tmp, q)) != BIG_OK) {
2349 			goto ret;
2350 		}
2351 	} else {
2352 		if ((err = big_div_pos(NULL, &ap, a, p)) != BIG_OK) {
2353 			goto ret;
2354 		}
2355 		if ((err = big_div_pos(NULL, &aq, a, q)) != BIG_OK) {
2356 			goto ret;
2357 		}
2358 	}
2359 
2360 	if ((err = big_modexp_ext(&ap, &ap, dmodpminus1, p, p_rr, info)) !=
2361 	    BIG_OK) {
2362 		goto ret;
2363 	}
2364 	if ((err = big_modexp_ext(&aq, &aq, dmodqminus1, q, q_rr, info)) !=
2365 	    BIG_OK) {
2366 		goto ret;
2367 	}
2368 	if ((err = big_sub(&tmp, &aq, &ap)) != BIG_OK) {
2369 		goto ret;
2370 	}
2371 	if ((err = big_mul(&tmp, &tmp, pinvmodq)) != BIG_OK) {
2372 		goto ret;
2373 	}
2374 	sign = tmp.sign;
2375 	tmp.sign = 1;
2376 	if ((err = big_div_pos(NULL, &aq, &tmp, q)) != BIG_OK) {
2377 		goto ret;
2378 	}
2379 	if ((sign == -1) && (!big_is_zero(&aq))) {
2380 		(void) big_sub_pos(&aq, q, &aq);
2381 	}
2382 	if ((err = big_mul(&tmp, &aq, p)) != BIG_OK) {
2383 		goto ret;
2384 	}
2385 	err = big_add_abs(result, &ap, &tmp);
2386 
2387 ret:
2388 	big_finish(&tmp);
2389 ret2:
2390 	big_finish(&aq);
2391 ret1:
2392 	big_finish(&ap);
2393 
2394 	return (err);
2395 }
2396 
2397 
2398 BIG_ERR_CODE
2399 big_modexp_crt(BIGNUM *result, BIGNUM *a, BIGNUM *dmodpminus1,
2400     BIGNUM *dmodqminus1, BIGNUM *p, BIGNUM *q, BIGNUM *pinvmodq,
2401     BIGNUM *p_rr, BIGNUM *q_rr)
2402 {
2403 	return (big_modexp_crt_ext(result, a, dmodpminus1, dmodqminus1,
2404 	    p, q, pinvmodq, p_rr, q_rr, NULL));
2405 }
2406 
2407 
2408 static BIG_CHUNK_TYPE onearr[1] = {(BIG_CHUNK_TYPE)1};
2409 BIGNUM big_One = {1, 1, 1, 0, onearr};
2410 
2411 static BIG_CHUNK_TYPE twoarr[1] = {(BIG_CHUNK_TYPE)2};
2412 BIGNUM big_Two = {1, 1, 1, 0, twoarr};
2413 
2414 static BIG_CHUNK_TYPE fourarr[1] = {(BIG_CHUNK_TYPE)4};
2415 static BIGNUM big_Four = {1, 1, 1, 0, fourarr};
2416 
2417 
2418 BIG_ERR_CODE
2419 big_sqrt_pos(BIGNUM *result, BIGNUM *n)
2420 {
2421 	BIGNUM		*high, *low, *mid, *t;
2422 	BIGNUM		t1, t2, t3, prod;
2423 	BIG_CHUNK_TYPE	t1value[BIGTMPSIZE];
2424 	BIG_CHUNK_TYPE	t2value[BIGTMPSIZE];
2425 	BIG_CHUNK_TYPE	t3value[BIGTMPSIZE];
2426 	BIG_CHUNK_TYPE	prodvalue[BIGTMPSIZE];
2427 	int		i, diff;
2428 	uint32_t	nbits, nrootbits, highbits;
2429 	BIG_ERR_CODE	err;
2430 
2431 	nbits = big_numbits(n);
2432 
2433 	if ((err = big_init1(&t1, n->len + 1,
2434 	    t1value, arraysize(t1value))) != BIG_OK)
2435 		return (err);
2436 	if ((err = big_init1(&t2, n->len + 1,
2437 	    t2value, arraysize(t2value))) != BIG_OK)
2438 		goto ret1;
2439 	if ((err = big_init1(&t3, n->len + 1,
2440 	    t3value, arraysize(t3value))) != BIG_OK)
2441 		goto ret2;
2442 	if ((err = big_init1(&prod, n->len + 1,
2443 	    prodvalue, arraysize(prodvalue))) != BIG_OK)
2444 		goto ret3;
2445 
2446 	nrootbits = (nbits + 1) / 2;
2447 	t1.len = t2.len = t3.len = (nrootbits - 1) / BIG_CHUNK_SIZE + 1;
2448 	for (i = 0; i < t1.len; i++) {
2449 		t1.value[i] = 0;
2450 		t2.value[i] = BIG_CHUNK_ALLBITS;
2451 	}
2452 	highbits = nrootbits - BIG_CHUNK_SIZE * (t1.len - 1);
2453 	if (highbits == BIG_CHUNK_SIZE) {
2454 		t1.value[t1.len - 1] = BIG_CHUNK_HIGHBIT;
2455 		t2.value[t2.len - 1] = BIG_CHUNK_ALLBITS;
2456 	} else {
2457 		t1.value[t1.len - 1] = (BIG_CHUNK_TYPE)1 << (highbits - 1);
2458 		t2.value[t2.len - 1] = 2 * t1.value[t1.len - 1] - 1;
2459 	}
2460 
2461 	high = &t2;
2462 	low = &t1;
2463 	mid = &t3;
2464 
2465 	if ((err = big_mul(&prod, high, high)) != BIG_OK) {
2466 		goto ret;
2467 	}
2468 	diff = big_cmp_abs(&prod, n);
2469 	if (diff <= 0) {
2470 		err = big_copy(result, high);
2471 		goto ret;
2472 	}
2473 
2474 	(void) big_sub_pos(mid, high, low);
2475 	while (big_cmp_abs(&big_One, mid) != 0) {
2476 		(void) big_add_abs(mid, high, low);
2477 		(void) big_half_pos(mid, mid);
2478 		if ((err = big_mul(&prod, mid, mid)) != BIG_OK)
2479 			goto ret;
2480 		diff = big_cmp_abs(&prod, n);
2481 		if (diff > 0) {
2482 			t = high;
2483 			high = mid;
2484 			mid = t;
2485 		} else if (diff < 0) {
2486 			t = low;
2487 			low = mid;
2488 			mid = t;
2489 		} else {
2490 			err = big_copy(result, low);
2491 			goto ret;
2492 		}
2493 		(void) big_sub_pos(mid, high, low);
2494 	}
2495 
2496 	err = big_copy(result, low);
2497 ret:
2498 	if (prod.malloced) big_finish(&prod);
2499 ret3:
2500 	if (t3.malloced) big_finish(&t3);
2501 ret2:
2502 	if (t2.malloced) big_finish(&t2);
2503 ret1:
2504 	if (t1.malloced) big_finish(&t1);
2505 
2506 	return (err);
2507 }
2508 
2509 
2510 BIG_ERR_CODE
2511 big_Jacobi_pos(int *jac, BIGNUM *nn, BIGNUM *mm)
2512 {
2513 	BIGNUM		*t, *tmp2, *m, *n;
2514 	BIGNUM		t1, t2, t3;
2515 	BIG_CHUNK_TYPE	t1value[BIGTMPSIZE];
2516 	BIG_CHUNK_TYPE	t2value[BIGTMPSIZE];
2517 	BIG_CHUNK_TYPE	t3value[BIGTMPSIZE];
2518 	int		len, err;
2519 
2520 	if (big_is_zero(nn) ||
2521 	    (((nn->value[0] & 1) | (mm->value[0] & 1)) == 0)) {
2522 		*jac = 0;
2523 		return (BIG_OK);
2524 	}
2525 
2526 	if (nn->len > mm->len) {
2527 		len = nn->len;
2528 	} else {
2529 		len = mm->len;
2530 	}
2531 
2532 	if ((err = big_init1(&t1, len,
2533 	    t1value, arraysize(t1value))) != BIG_OK) {
2534 		return (err);
2535 	}
2536 	if ((err = big_init1(&t2, len,
2537 	    t2value, arraysize(t2value))) != BIG_OK) {
2538 		goto ret1;
2539 	}
2540 	if ((err = big_init1(&t3, len,
2541 	    t3value, arraysize(t3value))) != BIG_OK) {
2542 		goto ret2;
2543 	}
2544 
2545 	n = &t1;
2546 	m = &t2;
2547 	tmp2 = &t3;
2548 
2549 	(void) big_copy(n, nn);
2550 	(void) big_copy(m, mm);
2551 
2552 	*jac = 1;
2553 	while (big_cmp_abs(&big_One, m) != 0) {
2554 		if (big_is_zero(n)) {
2555 			*jac = 0;
2556 			goto ret;
2557 		}
2558 		if ((m->value[0] & 1) == 0) {
2559 			if (((n->value[0] & 7) == 3) ||
2560 			    ((n->value[0] & 7) == 5))
2561 				*jac = -*jac;
2562 			(void) big_half_pos(m, m);
2563 		} else if ((n->value[0] & 1) == 0) {
2564 			if (((m->value[0] & 7) == 3) ||
2565 			    ((m->value[0] & 7) == 5))
2566 				*jac = -*jac;
2567 			(void) big_half_pos(n, n);
2568 		} else {
2569 			if (((m->value[0] & 3) == 3) &&
2570 			    ((n->value[0] & 3) == 3)) {
2571 				*jac = -*jac;
2572 			}
2573 			if ((err = big_div_pos(NULL, tmp2, m, n)) != BIG_OK) {
2574 				goto ret;
2575 			}
2576 			t = tmp2;
2577 			tmp2 = m;
2578 			m = n;
2579 			n = t;
2580 		}
2581 	}
2582 	err = BIG_OK;
2583 
2584 ret:
2585 	if (t3.malloced) big_finish(&t3);
2586 ret2:
2587 	if (t2.malloced) big_finish(&t2);
2588 ret1:
2589 	if (t1.malloced) big_finish(&t1);
2590 
2591 	return (err);
2592 }
2593 
2594 
2595 BIG_ERR_CODE
2596 big_Lucas(BIGNUM *Lkminus1, BIGNUM *Lk, BIGNUM *p, BIGNUM *k, BIGNUM *n)
2597 {
2598 	int		i;
2599 	uint32_t	m, w;
2600 	BIG_CHUNK_TYPE	bit;
2601 	BIGNUM		ki, tmp, tmp2;
2602 	BIG_CHUNK_TYPE	kivalue[BIGTMPSIZE];
2603 	BIG_CHUNK_TYPE	tmpvalue[BIGTMPSIZE];
2604 	BIG_CHUNK_TYPE	tmp2value[BIGTMPSIZE];
2605 	BIG_ERR_CODE	err;
2606 
2607 	if (big_cmp_abs(k, &big_One) == 0) {
2608 		(void) big_copy(Lk, p);
2609 		(void) big_copy(Lkminus1, &big_Two);
2610 		return (BIG_OK);
2611 	}
2612 
2613 	if ((err = big_init1(&ki, k->len + 1,
2614 	    kivalue, arraysize(kivalue))) != BIG_OK)
2615 		return (err);
2616 
2617 	if ((err = big_init1(&tmp, 2 * n->len + 1,
2618 	    tmpvalue, arraysize(tmpvalue))) != BIG_OK)
2619 		goto ret1;
2620 
2621 	if ((err = big_init1(&tmp2, n->len,
2622 	    tmp2value, arraysize(tmp2value))) != BIG_OK)
2623 		goto ret2;
2624 
2625 	m = big_numbits(k);
2626 	ki.len = (m - 1) / BIG_CHUNK_SIZE + 1;
2627 	w = (m - 1) / BIG_CHUNK_SIZE;
2628 	bit = (BIG_CHUNK_TYPE)1 << ((m - 1) % BIG_CHUNK_SIZE);
2629 	for (i = 0; i < ki.len; i++) {
2630 		ki.value[i] = 0;
2631 	}
2632 	ki.value[ki.len - 1] = bit;
2633 	if (big_cmp_abs(k, &ki) != 0) {
2634 		(void) big_double(&ki, &ki);
2635 	}
2636 	(void) big_sub_pos(&ki, &ki, k);
2637 
2638 	(void) big_copy(Lk, p);
2639 	(void) big_copy(Lkminus1, &big_Two);
2640 
2641 	for (i = 0; i < m; i++) {
2642 		if ((err = big_mul(&tmp, Lk, Lkminus1)) != BIG_OK) {
2643 			goto ret;
2644 		}
2645 		(void) big_add_abs(&tmp, &tmp, n);
2646 		(void) big_sub_pos(&tmp, &tmp, p);
2647 		if ((err = big_div_pos(NULL, &tmp2, &tmp, n)) != BIG_OK) {
2648 			goto ret;
2649 		}
2650 		if ((ki.value[w] & bit) != 0) {
2651 			if ((err = big_mul(&tmp, Lkminus1, Lkminus1)) !=
2652 			    BIG_OK) {
2653 				goto ret;
2654 			}
2655 			(void) big_add_abs(&tmp, &tmp, n);
2656 			(void) big_sub_pos(&tmp, &tmp, &big_Two);
2657 			if ((err = big_div_pos(NULL, Lkminus1, &tmp, n)) !=
2658 			    BIG_OK) {
2659 				goto ret;
2660 			}
2661 			(void) big_copy(Lk, &tmp2);
2662 		} else {
2663 			if ((err = big_mul(&tmp, Lk, Lk)) != BIG_OK) {
2664 				goto ret;
2665 			}
2666 			(void) big_add_abs(&tmp, &tmp, n);
2667 			(void) big_sub_pos(&tmp, &tmp, &big_Two);
2668 			if ((err = big_div_pos(NULL, Lk, &tmp, n)) != BIG_OK) {
2669 				goto ret;
2670 			}
2671 			(void) big_copy(Lkminus1, &tmp2);
2672 		}
2673 		bit = bit >> 1;
2674 		if (bit == 0) {
2675 			bit = BIG_CHUNK_HIGHBIT;
2676 			w--;
2677 		}
2678 	}
2679 
2680 	err = BIG_OK;
2681 
2682 ret:
2683 	if (tmp2.malloced) big_finish(&tmp2);
2684 ret2:
2685 	if (tmp.malloced) big_finish(&tmp);
2686 ret1:
2687 	if (ki.malloced) big_finish(&ki);
2688 
2689 	return (err);
2690 }
2691 
2692 
2693 BIG_ERR_CODE
2694 big_isprime_pos_ext(BIGNUM *n, big_modexp_ncp_info_t *info)
2695 {
2696 	BIGNUM		o, nminus1, tmp, Lkminus1, Lk;
2697 	BIG_CHUNK_TYPE	ovalue[BIGTMPSIZE];
2698 	BIG_CHUNK_TYPE	nminus1value[BIGTMPSIZE];
2699 	BIG_CHUNK_TYPE	tmpvalue[BIGTMPSIZE];
2700 	BIG_CHUNK_TYPE	Lkminus1value[BIGTMPSIZE];
2701 	BIG_CHUNK_TYPE	Lkvalue[BIGTMPSIZE];
2702 	BIG_ERR_CODE	err;
2703 	int		e, i, jac;
2704 
2705 	if (big_cmp_abs(n, &big_One) == 0) {
2706 		return (BIG_FALSE);
2707 	}
2708 	if (big_cmp_abs(n, &big_Two) == 0) {
2709 		return (BIG_TRUE);
2710 	}
2711 	if ((n->value[0] & 1) == 0) {
2712 		return (BIG_FALSE);
2713 	}
2714 
2715 	if ((err = big_init1(&o, n->len, ovalue, arraysize(ovalue))) !=
2716 	    BIG_OK) {
2717 		return (err);
2718 	}
2719 
2720 	if ((err = big_init1(&nminus1, n->len,
2721 	    nminus1value, arraysize(nminus1value))) != BIG_OK) {
2722 		goto ret1;
2723 	}
2724 
2725 	if ((err = big_init1(&tmp, 2 * n->len,
2726 	    tmpvalue, arraysize(tmpvalue))) != BIG_OK) {
2727 		goto ret2;
2728 	}
2729 
2730 	if ((err = big_init1(&Lkminus1, n->len,
2731 	    Lkminus1value, arraysize(Lkminus1value))) != BIG_OK) {
2732 		goto ret3;
2733 	}
2734 
2735 	if ((err = big_init1(&Lk, n->len,
2736 	    Lkvalue, arraysize(Lkvalue))) != BIG_OK) {
2737 		goto ret4;
2738 	}
2739 
2740 	(void) big_sub_pos(&o, n, &big_One);	/* cannot fail */
2741 	(void) big_copy(&nminus1, &o);		/* cannot fail */
2742 	e = 0;
2743 	while ((o.value[0] & 1) == 0) {
2744 		e++;
2745 		(void) big_half_pos(&o, &o);  /* cannot fail */
2746 	}
2747 	if ((err = big_modexp_ext(&tmp, &big_Two, &o, n, NULL, info)) !=
2748 	    BIG_OK) {
2749 		goto ret;
2750 	}
2751 
2752 	i = 0;
2753 	while ((i < e) &&
2754 	    (big_cmp_abs(&tmp, &big_One) != 0) &&
2755 	    (big_cmp_abs(&tmp, &nminus1) != 0)) {
2756 		if ((err =
2757 		    big_modexp_ext(&tmp, &tmp, &big_Two, n, NULL, info)) !=
2758 		    BIG_OK)
2759 			goto ret;
2760 		i++;
2761 	}
2762 
2763 	if (!((big_cmp_abs(&tmp, &nminus1) == 0) ||
2764 	    ((i == 0) && (big_cmp_abs(&tmp, &big_One) == 0)))) {
2765 		err = BIG_FALSE;
2766 		goto ret;
2767 	}
2768 
2769 	if ((err = big_sqrt_pos(&tmp, n)) != BIG_OK) {
2770 		goto ret;
2771 	}
2772 
2773 	if ((err = big_mul(&tmp, &tmp, &tmp)) != BIG_OK) {
2774 		goto ret;
2775 	}
2776 	if (big_cmp_abs(&tmp, n) == 0) {
2777 		err = BIG_FALSE;
2778 		goto ret;
2779 	}
2780 
2781 	(void) big_copy(&o, &big_Two);
2782 	do {
2783 		(void) big_add_abs(&o, &o, &big_One);
2784 		if ((err = big_mul(&tmp, &o, &o)) != BIG_OK) {
2785 			goto ret;
2786 		}
2787 		(void) big_sub_pos(&tmp, &tmp, &big_Four);
2788 		if ((err = big_Jacobi_pos(&jac, &tmp, n)) != BIG_OK) {
2789 			goto ret;
2790 		}
2791 	} while (jac != -1);
2792 
2793 	(void) big_add_abs(&tmp, n, &big_One);
2794 	if ((err = big_Lucas(&Lkminus1, &Lk, &o, &tmp, n)) != BIG_OK) {
2795 		goto ret;
2796 	}
2797 
2798 	if ((big_cmp_abs(&Lkminus1, &o) == 0) &&
2799 	    (big_cmp_abs(&Lk, &big_Two) == 0)) {
2800 		err = BIG_TRUE;
2801 	} else {
2802 		err = BIG_FALSE;
2803 	}
2804 
2805 ret:
2806 	if (Lk.malloced) big_finish(&Lk);
2807 ret4:
2808 	if (Lkminus1.malloced) big_finish(&Lkminus1);
2809 ret3:
2810 	if (tmp.malloced) big_finish(&tmp);
2811 ret2:
2812 	if (nminus1.malloced) big_finish(&nminus1);
2813 ret1:
2814 	if (o.malloced) big_finish(&o);
2815 
2816 	return (err);
2817 }
2818 
2819 
2820 BIG_ERR_CODE
2821 big_isprime_pos(BIGNUM *n)
2822 {
2823 	return (big_isprime_pos_ext(n, NULL));
2824 }
2825 
2826 
2827 #define	SIEVESIZE 1000
2828 
2829 
2830 BIG_ERR_CODE
2831 big_nextprime_pos_ext(BIGNUM *result, BIGNUM *n, big_modexp_ncp_info_t *info)
2832 {
2833 	static const uint32_t smallprimes[] = {
2834 	    3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47,
2835 	    51, 53, 59, 61, 67, 71, 73, 79, 83, 89, 91, 97 };
2836 	BIG_ERR_CODE	err;
2837 	int		sieve[SIEVESIZE];
2838 	int		i;
2839 	uint32_t	off, p;
2840 
2841 	if ((err = big_copy(result, n)) != BIG_OK) {
2842 		return (err);
2843 	}
2844 	result->value[0] |= 1;
2845 	/* CONSTCOND */
2846 	while (1) {
2847 		for (i = 0; i < SIEVESIZE; i++) sieve[i] = 0;
2848 		for (i = 0;
2849 		    i < sizeof (smallprimes) / sizeof (smallprimes[0]); i++) {
2850 			p = smallprimes[i];
2851 			off = big_modhalf_pos(result, p);
2852 			off = p - off;
2853 			if ((off % 2) == 1) {
2854 				off = off + p;
2855 			}
2856 			off = off / 2;
2857 			while (off < SIEVESIZE) {
2858 				sieve[off] = 1;
2859 				off = off + p;
2860 			}
2861 		}
2862 
2863 		for (i = 0; i < SIEVESIZE; i++) {
2864 			if (sieve[i] == 0) {
2865 				err = big_isprime_pos_ext(result, info);
2866 				if (err != BIG_FALSE) {
2867 					if (err != BIG_TRUE) {
2868 						return (err);
2869 					} else {
2870 						goto out;
2871 					}
2872 				}
2873 
2874 			}
2875 			if ((err = big_add_abs(result, result, &big_Two)) !=
2876 			    BIG_OK) {
2877 				return (err);
2878 			}
2879 		}
2880 	}
2881 
2882 out:
2883 	return (BIG_OK);
2884 }
2885 
2886 
2887 BIG_ERR_CODE
2888 big_nextprime_pos(BIGNUM *result, BIGNUM *n)
2889 {
2890 	return (big_nextprime_pos_ext(result, n, NULL));
2891 }
2892 
2893 
2894 BIG_ERR_CODE
2895 big_nextprime_pos_slow(BIGNUM *result, BIGNUM *n)
2896 {
2897 	BIG_ERR_CODE err;
2898 
2899 
2900 	if ((err = big_copy(result, n)) != BIG_OK)
2901 		return (err);
2902 	result->value[0] |= 1;
2903 	while ((err = big_isprime_pos_ext(result, NULL)) != BIG_TRUE) {
2904 		if (err != BIG_FALSE)
2905 			return (err);
2906 		if ((err = big_add_abs(result, result, &big_Two)) != BIG_OK)
2907 			return (err);
2908 	}
2909 	return (BIG_OK);
2910 }
2911 
2912 
2913 /*
2914  * given m and e, computes the rest in the equation
2915  * gcd(m, e) = cm * m + ce * e
2916  */
2917 BIG_ERR_CODE
2918 big_ext_gcd_pos(BIGNUM *gcd, BIGNUM *cm, BIGNUM *ce, BIGNUM *m, BIGNUM *e)
2919 {
2920 	BIGNUM		*xi, *ri, *riminus1, *riminus2, *t;
2921 	BIGNUM		*vmi, *vei, *vmiminus1, *veiminus1;
2922 	BIGNUM		t1, t2, t3, t4, t5, t6, t7, t8, tmp;
2923 	BIG_CHUNK_TYPE	t1value[BIGTMPSIZE];
2924 	BIG_CHUNK_TYPE	t2value[BIGTMPSIZE];
2925 	BIG_CHUNK_TYPE	t3value[BIGTMPSIZE];
2926 	BIG_CHUNK_TYPE	t4value[BIGTMPSIZE];
2927 	BIG_CHUNK_TYPE	t5value[BIGTMPSIZE];
2928 	BIG_CHUNK_TYPE	t6value[BIGTMPSIZE];
2929 	BIG_CHUNK_TYPE	t7value[BIGTMPSIZE];
2930 	BIG_CHUNK_TYPE	t8value[BIGTMPSIZE];
2931 	BIG_CHUNK_TYPE	tmpvalue[BIGTMPSIZE];
2932 	BIG_ERR_CODE	err;
2933 	int		len;
2934 
2935 	if (big_cmp_abs(m, e) >= 0) {
2936 		len = m->len;
2937 	} else {
2938 		len = e->len;
2939 	}
2940 
2941 	if ((err = big_init1(&t1, len,
2942 	    t1value, arraysize(t1value))) != BIG_OK) {
2943 		return (err);
2944 	}
2945 	if ((err = big_init1(&t2, len,
2946 	    t2value, arraysize(t2value))) != BIG_OK) {
2947 		goto ret1;
2948 	}
2949 	if ((err = big_init1(&t3, len,
2950 	    t3value, arraysize(t3value))) != BIG_OK) {
2951 		goto ret2;
2952 	}
2953 	if ((err = big_init1(&t4, len,
2954 	    t4value, arraysize(t3value))) != BIG_OK) {
2955 		goto ret3;
2956 	}
2957 	if ((err = big_init1(&t5, len,
2958 	    t5value, arraysize(t5value))) != BIG_OK) {
2959 		goto ret4;
2960 	}
2961 	if ((err = big_init1(&t6, len,
2962 	    t6value, arraysize(t6value))) != BIG_OK) {
2963 		goto ret5;
2964 	}
2965 	if ((err = big_init1(&t7, len,
2966 	    t7value, arraysize(t7value))) != BIG_OK) {
2967 		goto ret6;
2968 	}
2969 	if ((err = big_init1(&t8, len,
2970 	    t8value, arraysize(t8value))) != BIG_OK) {
2971 		goto ret7;
2972 	}
2973 
2974 	if ((err = big_init1(&tmp, 2 * len,
2975 	    tmpvalue, arraysize(tmpvalue))) != BIG_OK) {
2976 		goto ret8;
2977 	}
2978 
2979 	ri = &t1;
2980 	ri->value[0] = 1;
2981 	ri->len = 1;
2982 	xi = &t2;
2983 	riminus1 = &t3;
2984 	riminus2 = &t4;
2985 	vmi = &t5;
2986 	vei = &t6;
2987 	vmiminus1 = &t7;
2988 	veiminus1 = &t8;
2989 
2990 	(void) big_copy(vmiminus1, &big_One);
2991 	(void) big_copy(vmi, &big_One);
2992 	(void) big_copy(veiminus1, &big_One);
2993 	(void) big_copy(xi, &big_One);
2994 	vei->len = 1;
2995 	vei->value[0] = 0;
2996 
2997 	(void) big_copy(riminus1, m);
2998 	(void) big_copy(ri, e);
2999 
3000 	while (!big_is_zero(ri)) {
3001 		t = riminus2;
3002 		riminus2 = riminus1;
3003 		riminus1 = ri;
3004 		ri = t;
3005 		if ((err = big_mul(&tmp, vmi, xi)) != BIG_OK) {
3006 			goto ret;
3007 		}
3008 		if ((err = big_sub(vmiminus1, vmiminus1, &tmp)) != BIG_OK) {
3009 			goto ret;
3010 		}
3011 		t = vmiminus1;
3012 		vmiminus1 = vmi;
3013 		vmi = t;
3014 		if ((err = big_mul(&tmp, vei, xi)) != BIG_OK) {
3015 			goto ret;
3016 		}
3017 		if ((err = big_sub(veiminus1, veiminus1, &tmp)) != BIG_OK) {
3018 			goto ret;
3019 		}
3020 		t = veiminus1;
3021 		veiminus1 = vei;
3022 		vei = t;
3023 		if ((err = big_div_pos(xi, ri, riminus2, riminus1)) !=
3024 		    BIG_OK) {
3025 			goto ret;
3026 		}
3027 	}
3028 	if ((gcd != NULL) && ((err = big_copy(gcd, riminus1)) != BIG_OK)) {
3029 		goto ret;
3030 	}
3031 	if ((cm != NULL) && ((err = big_copy(cm, vmi)) != BIG_OK)) {
3032 		goto ret;
3033 	}
3034 	if (ce != NULL) {
3035 		err = big_copy(ce, vei);
3036 	}
3037 ret:
3038 	if (tmp.malloced) big_finish(&tmp);
3039 ret8:
3040 	if (t8.malloced) big_finish(&t8);
3041 ret7:
3042 	if (t7.malloced) big_finish(&t7);
3043 ret6:
3044 	if (t6.malloced) big_finish(&t6);
3045 ret5:
3046 	if (t5.malloced) big_finish(&t5);
3047 ret4:
3048 	if (t4.malloced) big_finish(&t4);
3049 ret3:
3050 	if (t3.malloced) big_finish(&t3);
3051 ret2:
3052 	if (t2.malloced) big_finish(&t2);
3053 ret1:
3054 	if (t1.malloced) big_finish(&t1);
3055 
3056 	return (err);
3057 }
3058