summaryrefslogtreecommitdiff
path: root/src/lib/libcrypto/bn
diff options
context:
space:
mode:
authormarkus <>2003-05-11 21:36:58 +0000
committermarkus <>2003-05-11 21:36:58 +0000
commit1c98a87f0daac81245653c227eb2f2508a22a965 (patch)
tree3de6d603296ec563b936da4e6a8a1e33d48f8884 /src/lib/libcrypto/bn
parent31392c89d1135cf2a416f97295f6d21681b3fbc4 (diff)
downloadopenbsd-1c98a87f0daac81245653c227eb2f2508a22a965.tar.gz
openbsd-1c98a87f0daac81245653c227eb2f2508a22a965.tar.bz2
openbsd-1c98a87f0daac81245653c227eb2f2508a22a965.zip
import 0.9.7b (without idea and rc5)
Diffstat (limited to 'src/lib/libcrypto/bn')
-rw-r--r--src/lib/libcrypto/bn/asm/ia64.S235
-rw-r--r--src/lib/libcrypto/bn/asm/pa-risc2.s36
-rw-r--r--src/lib/libcrypto/bn/asm/x86_64-gcc.c575
-rw-r--r--src/lib/libcrypto/bn/bn.h2
-rw-r--r--src/lib/libcrypto/bn/bn_div.c28
-rw-r--r--src/lib/libcrypto/bn/bn_lcl.h26
-rw-r--r--src/lib/libcrypto/bn/bn_lib.c4
-rw-r--r--src/lib/libcrypto/bn/bn_mul.c529
-rw-r--r--src/lib/libcrypto/bn/bn_prime.c2
-rw-r--r--src/lib/libcrypto/bn/bn_rand.c2
-rw-r--r--src/lib/libcrypto/bn/bn_word.c5
11 files changed, 905 insertions, 539 deletions
diff --git a/src/lib/libcrypto/bn/asm/ia64.S b/src/lib/libcrypto/bn/asm/ia64.S
index ae56066310..7dfda85566 100644
--- a/src/lib/libcrypto/bn/asm/ia64.S
+++ b/src/lib/libcrypto/bn/asm/ia64.S
@@ -1,6 +1,6 @@
1.explicit 1.explicit
2.text 2.text
3.ident "ia64.S, Version 1.1" 3.ident "ia64.S, Version 2.0"
4.ident "IA-64 ISA artwork by Andy Polyakov <appro@fy.chalmers.se>" 4.ident "IA-64 ISA artwork by Andy Polyakov <appro@fy.chalmers.se>"
5 5
6// 6//
@@ -13,6 +13,35 @@
13// disclaimed. 13// disclaimed.
14// ==================================================================== 14// ====================================================================
15// 15//
16// Version 2.x is Itanium2 re-tune. Few words about how Itanum2 is
17// different from Itanium to this module viewpoint. Most notably, is it
18// "wider" than Itanium? Can you experience loop scalability as
19// discussed in commentary sections? Not really:-( Itanium2 has 6
20// integer ALU ports, i.e. it's 2 ports wider, but it's not enough to
21// spin twice as fast, as I need 8 IALU ports. Amount of floating point
22// ports is the same, i.e. 2, while I need 4. In other words, to this
23// module Itanium2 remains effectively as "wide" as Itanium. Yet it's
24// essentially different in respect to this module, and a re-tune was
25// required. Well, because some intruction latencies has changed. Most
26// noticeably those intensively used:
27//
28// Itanium Itanium2
29// ldf8 9 6 L2 hit
30// ld8 2 1 L1 hit
31// getf 2 5
32// xma[->getf] 7[+1] 4[+0]
33// add[->st8] 1[+1] 1[+0]
34//
35// What does it mean? You might ratiocinate that the original code
36// should run just faster... Because sum of latencies is smaller...
37// Wrong! Note that getf latency increased. This means that if a loop is
38// scheduled for lower latency (and they are), then it will suffer from
39// stall condition and the code will therefore turn anti-scalable, e.g.
40// original bn_mul_words spun at 5*n or 2.5 times slower than expected
41// on Itanium2! What to do? Reschedule loops for Itanium2? But then
42// Itanium would exhibit anti-scalability. So I've chosen to reschedule
43// for worst latency for every instruction aiming for best *all-round*
44// performance.
16 45
17// Q. How much faster does it get? 46// Q. How much faster does it get?
18// A. Here is the output from 'openssl speed rsa dsa' for vanilla 47// A. Here is the output from 'openssl speed rsa dsa' for vanilla
@@ -149,12 +178,27 @@ bn_add_words:
149 brp.loop.imp .L_bn_add_words_ctop,.L_bn_add_words_cend-16 178 brp.loop.imp .L_bn_add_words_ctop,.L_bn_add_words_cend-16
150 } 179 }
151 .body 180 .body
152{ .mib; mov r14=r32 // rp 181{ .mib;
182#if defined(_HPUX_SOURCE) && defined(_ILP32)
183 addp4 r14=0,r32 // rp
184#else
185 mov r14=r32 // rp
186#endif
153 mov r9=pr };; 187 mov r9=pr };;
154{ .mii; mov r15=r33 // ap 188{ .mii;
189#if defined(_HPUX_SOURCE) && defined(_ILP32)
190 addp4 r15=0,r33 // ap
191#else
192 mov r15=r33 // ap
193#endif
155 mov ar.lc=r10 194 mov ar.lc=r10
156 mov ar.ec=6 } 195 mov ar.ec=6 }
157{ .mib; mov r16=r34 // bp 196{ .mib;
197#if defined(_HPUX_SOURCE) && defined(_ILP32)
198 addp4 r16=0,r34 // bp
199#else
200 mov r16=r34 // bp
201#endif
158 mov pr.rot=1<<16 };; 202 mov pr.rot=1<<16 };;
159 203
160.L_bn_add_words_ctop: 204.L_bn_add_words_ctop:
@@ -174,7 +218,7 @@ bn_add_words:
174 218
175{ .mii; 219{ .mii;
176(p59) add r8=1,r8 // return value 220(p59) add r8=1,r8 // return value
177 mov pr=r9,-1 221 mov pr=r9,0x1ffff
178 mov ar.lc=r3 } 222 mov ar.lc=r3 }
179{ .mbb; nop.b 0x0 223{ .mbb; nop.b 0x0
180 br.ret.sptk.many b0 };; 224 br.ret.sptk.many b0 };;
@@ -202,12 +246,27 @@ bn_sub_words:
202 brp.loop.imp .L_bn_sub_words_ctop,.L_bn_sub_words_cend-16 246 brp.loop.imp .L_bn_sub_words_ctop,.L_bn_sub_words_cend-16
203 } 247 }
204 .body 248 .body
205{ .mib; mov r14=r32 // rp 249{ .mib;
250#if defined(_HPUX_SOURCE) && defined(_ILP32)
251 addp4 r14=0,r32 // rp
252#else
253 mov r14=r32 // rp
254#endif
206 mov r9=pr };; 255 mov r9=pr };;
207{ .mii; mov r15=r33 // ap 256{ .mii;
257#if defined(_HPUX_SOURCE) && defined(_ILP32)
258 addp4 r15=0,r33 // ap
259#else
260 mov r15=r33 // ap
261#endif
208 mov ar.lc=r10 262 mov ar.lc=r10
209 mov ar.ec=6 } 263 mov ar.ec=6 }
210{ .mib; mov r16=r34 // bp 264{ .mib;
265#if defined(_HPUX_SOURCE) && defined(_ILP32)
266 addp4 r16=0,r34 // bp
267#else
268 mov r16=r34 // bp
269#endif
211 mov pr.rot=1<<16 };; 270 mov pr.rot=1<<16 };;
212 271
213.L_bn_sub_words_ctop: 272.L_bn_sub_words_ctop:
@@ -227,7 +286,7 @@ bn_sub_words:
227 286
228{ .mii; 287{ .mii;
229(p59) add r8=1,r8 // return value 288(p59) add r8=1,r8 // return value
230 mov pr=r9,-1 289 mov pr=r9,0x1ffff
231 mov ar.lc=r3 } 290 mov ar.lc=r3 }
232{ .mbb; nop.b 0x0 291{ .mbb; nop.b 0x0
233 br.ret.sptk.many b0 };; 292 br.ret.sptk.many b0 };;
@@ -253,7 +312,7 @@ bn_mul_words:
253#ifdef XMA_TEMPTATION 312#ifdef XMA_TEMPTATION
254{ .mfi; alloc r2=ar.pfs,4,0,0,0 };; 313{ .mfi; alloc r2=ar.pfs,4,0,0,0 };;
255#else 314#else
256{ .mfi; alloc r2=ar.pfs,4,4,0,8 };; 315{ .mfi; alloc r2=ar.pfs,4,12,0,16 };;
257#endif 316#endif
258{ .mib; mov r8=r0 // return value 317{ .mib; mov r8=r0 // return value
259 cmp4.le p6,p0=r34,r0 318 cmp4.le p6,p0=r34,r0
@@ -266,24 +325,30 @@ bn_mul_words:
266 325
267 .body 326 .body
268{ .mib; setf.sig f8=r35 // w 327{ .mib; setf.sig f8=r35 // w
269 mov pr.rot=0x400001<<16 328 mov pr.rot=0x800001<<16
270 // ------^----- serves as (p48) at first (p26) 329 // ------^----- serves as (p50) at first (p27)
271 brp.loop.imp .L_bn_mul_words_ctop,.L_bn_mul_words_cend-16 330 brp.loop.imp .L_bn_mul_words_ctop,.L_bn_mul_words_cend-16
272 } 331 }
273 332
274#ifndef XMA_TEMPTATION 333#ifndef XMA_TEMPTATION
275 334
276{ .mii; mov r14=r32 // rp 335{ .mii;
277 mov r15=r33 // ap 336#if defined(_HPUX_SOURCE) && defined(_ILP32)
337 addp4 r14=0,r32 // rp
338 addp4 r15=0,r33 // ap
339#else
340 mov r14=r32 // rp
341 mov r15=r33 // ap
342#endif
278 mov ar.lc=r10 } 343 mov ar.lc=r10 }
279{ .mii; mov r39=0 // serves as r33 at first (p26) 344{ .mii; mov r40=0 // serves as r35 at first (p27)
280 mov ar.ec=12 };; 345 mov ar.ec=13 };;
281 346
282// This loop spins in 2*(n+11) ticks. It's scheduled for data in L2 347// This loop spins in 2*(n+12) ticks. It's scheduled for data in Itanium
283// cache (i.e. 9 ticks away) as floating point load/store instructions 348// L2 cache (i.e. 9 ticks away) as floating point load/store instructions
284// bypass L1 cache and L2 latency is actually best-case scenario for 349// bypass L1 cache and L2 latency is actually best-case scenario for
285// ldf8. The loop is not scalable and shall run in 2*(n+11) even on 350// ldf8. The loop is not scalable and shall run in 2*(n+12) even on
286// "wider" IA-64 implementations. It's a trade-off here. n+22 loop 351// "wider" IA-64 implementations. It's a trade-off here. n+24 loop
287// would give us ~5% in *overall* performance improvement on "wider" 352// would give us ~5% in *overall* performance improvement on "wider"
288// IA-64, but would hurt Itanium for about same because of longer 353// IA-64, but would hurt Itanium for about same because of longer
289// epilogue. As it's a matter of few percents in either case I've 354// epilogue. As it's a matter of few percents in either case I've
@@ -291,25 +356,25 @@ bn_mul_words:
291// this very instruction sequence in bn_mul_add_words loop which in 356// this very instruction sequence in bn_mul_add_words loop which in
292// turn is scalable). 357// turn is scalable).
293.L_bn_mul_words_ctop: 358.L_bn_mul_words_ctop:
294{ .mfi; (p25) getf.sig r36=f49 // low 359{ .mfi; (p25) getf.sig r36=f52 // low
295 (p21) xmpy.lu f45=f37,f8 360 (p21) xmpy.lu f48=f37,f8
296 (p27) cmp.ltu p52,p48=r39,r38 } 361 (p28) cmp.ltu p54,p50=r41,r39 }
297{ .mfi; (p16) ldf8 f32=[r15],8 362{ .mfi; (p16) ldf8 f32=[r15],8
298 (p21) xmpy.hu f38=f37,f8 363 (p21) xmpy.hu f40=f37,f8
299 (p0) nop.i 0x0 };; 364 (p0) nop.i 0x0 };;
300{ .mii; (p26) getf.sig r32=f43 // high 365{ .mii; (p25) getf.sig r32=f44 // high
301 .pred.rel "mutex",p48,p52 366 .pred.rel "mutex",p50,p54
302 (p48) add r38=r37,r33 // (p26) 367 (p50) add r40=r38,r35 // (p27)
303 (p52) add r38=r37,r33,1 } // (p26) 368 (p54) add r40=r38,r35,1 } // (p27)
304{ .mfb; (p27) st8 [r14]=r39,8 369{ .mfb; (p28) st8 [r14]=r41,8
305 (p0) nop.f 0x0 370 (p0) nop.f 0x0
306 br.ctop.sptk .L_bn_mul_words_ctop };; 371 br.ctop.sptk .L_bn_mul_words_ctop };;
307.L_bn_mul_words_cend: 372.L_bn_mul_words_cend:
308 373
309{ .mii; nop.m 0x0 374{ .mii; nop.m 0x0
310.pred.rel "mutex",p49,p53 375.pred.rel "mutex",p51,p55
311(p49) add r8=r34,r0 376(p51) add r8=r36,r0
312(p53) add r8=r34,r0,1 } 377(p55) add r8=r36,r0,1 }
313{ .mfb; nop.m 0x0 378{ .mfb; nop.m 0x0
314 nop.f 0x0 379 nop.f 0x0
315 nop.b 0x0 } 380 nop.b 0x0 }
@@ -344,7 +409,7 @@ bn_mul_words:
344#endif // XMA_TEMPTATION 409#endif // XMA_TEMPTATION
345 410
346{ .mii; nop.m 0x0 411{ .mii; nop.m 0x0
347 mov pr=r9,-1 412 mov pr=r9,0x1ffff
348 mov ar.lc=r3 } 413 mov ar.lc=r3 }
349{ .mfb; rum 1<<5 // clear um.mfh 414{ .mfb; rum 1<<5 // clear um.mfh
350 nop.f 0x0 415 nop.f 0x0
@@ -376,59 +441,69 @@ bn_mul_add_words:
376 441
377 .body 442 .body
378{ .mib; setf.sig f8=r35 // w 443{ .mib; setf.sig f8=r35 // w
379 mov pr.rot=0x400001<<16 444 mov pr.rot=0x800001<<16
380 // ------^----- serves as (p48) at first (p26) 445 // ------^----- serves as (p50) at first (p27)
381 brp.loop.imp .L_bn_mul_add_words_ctop,.L_bn_mul_add_words_cend-16 446 brp.loop.imp .L_bn_mul_add_words_ctop,.L_bn_mul_add_words_cend-16
382 } 447 }
383{ .mii; mov r14=r32 // rp 448{ .mii;
384 mov r15=r33 // ap 449#if defined(_HPUX_SOURCE) && defined(_ILP32)
450 addp4 r14=0,r32 // rp
451 addp4 r15=0,r33 // ap
452#else
453 mov r14=r32 // rp
454 mov r15=r33 // ap
455#endif
385 mov ar.lc=r10 } 456 mov ar.lc=r10 }
386{ .mii; mov r39=0 // serves as r33 at first (p26) 457{ .mii; mov r40=0 // serves as r35 at first (p27)
387 mov r18=r32 // rp copy 458#if defined(_HPUX_SOURCE) && defined(_ILP32)
388 mov ar.ec=14 };; 459 addp4 r18=0,r32 // rp copy
460#else
461 mov r18=r32 // rp copy
462#endif
463 mov ar.ec=15 };;
389 464
390// This loop spins in 3*(n+13) ticks on Itanium and should spin in 465// This loop spins in 3*(n+14) ticks on Itanium and should spin in
391// 2*(n+13) on "wider" IA-64 implementations (to be verified with new 466// 2*(n+14) on "wider" IA-64 implementations (to be verified with new
392// µ-architecture manuals as they become available). As usual it's 467// µ-architecture manuals as they become available). As usual it's
393// possible to compress the epilogue, down to 10 in this case, at the 468// possible to compress the epilogue, down to 10 in this case, at the
394// cost of scalability. Compressed (and therefore non-scalable) loop 469// cost of scalability. Compressed (and therefore non-scalable) loop
395// running at 3*(n+10) would buy you ~10% on Itanium but take ~35% 470// running at 3*(n+11) would buy you ~10% on Itanium but take ~35%
396// from "wider" IA-64 so let it be scalable! Special attention was 471// from "wider" IA-64 so let it be scalable! Special attention was
397// paid for having the loop body split at 64-byte boundary. ld8 is 472// paid for having the loop body split at 64-byte boundary. ld8 is
398// scheduled for L1 cache as the data is more than likely there. 473// scheduled for L1 cache as the data is more than likely there.
399// Indeed, bn_mul_words has put it there a moment ago:-) 474// Indeed, bn_mul_words has put it there a moment ago:-)
400.L_bn_mul_add_words_ctop: 475.L_bn_mul_add_words_ctop:
401{ .mfi; (p25) getf.sig r36=f49 // low 476{ .mfi; (p25) getf.sig r36=f52 // low
402 (p21) xmpy.lu f45=f37,f8 477 (p21) xmpy.lu f48=f37,f8
403 (p27) cmp.ltu p52,p48=r39,r38 } 478 (p28) cmp.ltu p54,p50=r41,r39 }
404{ .mfi; (p16) ldf8 f32=[r15],8 479{ .mfi; (p16) ldf8 f32=[r15],8
405 (p21) xmpy.hu f38=f37,f8 480 (p21) xmpy.hu f40=f37,f8
406 (p27) add r43=r43,r39 };; 481 (p28) add r45=r45,r41 };;
407{ .mii; (p26) getf.sig r32=f43 // high 482{ .mii; (p25) getf.sig r32=f44 // high
408 .pred.rel "mutex",p48,p52 483 .pred.rel "mutex",p50,p54
409 (p48) add r38=r37,r33 // (p26) 484 (p50) add r40=r38,r35 // (p27)
410 (p52) add r38=r37,r33,1 } // (p26) 485 (p54) add r40=r38,r35,1 } // (p27)
411{ .mfb; (p27) cmp.ltu.unc p56,p0=r43,r39 486{ .mfb; (p28) cmp.ltu.unc p60,p0=r45,r41
412 (p0) nop.f 0x0 487 (p0) nop.f 0x0
413 (p0) nop.b 0x0 } 488 (p0) nop.b 0x0 }
414{ .mii; (p26) ld8 r42=[r18],8 489{ .mii; (p27) ld8 r44=[r18],8
415 (p58) cmp.eq.or p57,p0=-1,r44 490 (p62) cmp.eq.or p61,p0=-1,r46
416 (p58) add r44=1,r44 } 491 (p62) add r46=1,r46 }
417{ .mfb; (p29) st8 [r14]=r45,8 492{ .mfb; (p30) st8 [r14]=r47,8
418 (p0) nop.f 0x0 493 (p0) nop.f 0x0
419 br.ctop.sptk .L_bn_mul_add_words_ctop};; 494 br.ctop.sptk .L_bn_mul_add_words_ctop};;
420.L_bn_mul_add_words_cend: 495.L_bn_mul_add_words_cend:
421 496
422{ .mii; nop.m 0x0 497{ .mii; nop.m 0x0
423.pred.rel "mutex",p51,p55 498.pred.rel "mutex",p53,p57
424(p51) add r8=r36,r0 499(p53) add r8=r38,r0
425(p55) add r8=r36,r0,1 } 500(p57) add r8=r38,r0,1 }
426{ .mfb; nop.m 0x0 501{ .mfb; nop.m 0x0
427 nop.f 0x0 502 nop.f 0x0
428 nop.b 0x0 };; 503 nop.b 0x0 };;
429{ .mii; 504{ .mii;
430(p59) add r8=1,r8 505(p63) add r8=1,r8
431 mov pr=r9,-1 506 mov pr=r9,0x1ffff
432 mov ar.lc=r3 } 507 mov ar.lc=r3 }
433{ .mfb; rum 1<<5 // clear um.mfh 508{ .mfb; rum 1<<5 // clear um.mfh
434 nop.f 0x0 509 nop.f 0x0
@@ -461,6 +536,10 @@ bn_sqr_words:
461 mov r9=pr };; 536 mov r9=pr };;
462 537
463 .body 538 .body
539#if defined(_HPUX_SOURCE) && defined(_ILP32)
540{ .mii; addp4 r32=0,r32
541 addp4 r33=0,r33 };;
542#endif
464{ .mib; 543{ .mib;
465 mov pr.rot=1<<16 544 mov pr.rot=1<<16
466 brp.loop.imp .L_bn_sqr_words_ctop,.L_bn_sqr_words_cend-16 545 brp.loop.imp .L_bn_sqr_words_ctop,.L_bn_sqr_words_cend-16
@@ -492,7 +571,7 @@ bn_sqr_words:
492.L_bn_sqr_words_cend: 571.L_bn_sqr_words_cend:
493 572
494{ .mii; nop.m 0x0 573{ .mii; nop.m 0x0
495 mov pr=r9,-1 574 mov pr=r9,0x1ffff
496 mov ar.lc=r3 } 575 mov ar.lc=r3 }
497{ .mfb; rum 1<<5 // clear um.mfh 576{ .mfb; rum 1<<5 // clear um.mfh
498 nop.f 0x0 577 nop.f 0x0
@@ -526,7 +605,14 @@ bn_sqr_comba8:
526 .prologue 605 .prologue
527 .fframe 0 606 .fframe 0
528 .save ar.pfs,r2 607 .save ar.pfs,r2
608#if defined(_HPUX_SOURCE) && defined(_ILP32)
529{ .mii; alloc r2=ar.pfs,2,1,0,0 609{ .mii; alloc r2=ar.pfs,2,1,0,0
610 addp4 r33=0,r33
611 addp4 r32=0,r32 };;
612{ .mii;
613#else
614{ .mii; alloc r2=ar.pfs,2,1,0,0
615#endif
530 mov r34=r33 616 mov r34=r33
531 add r14=8,r33 };; 617 add r14=8,r33 };;
532 .body 618 .body
@@ -587,7 +673,14 @@ bn_mul_comba8:
587 .prologue 673 .prologue
588 .fframe 0 674 .fframe 0
589 .save ar.pfs,r2 675 .save ar.pfs,r2
676#if defined(_HPUX_SOURCE) && defined(_ILP32)
590{ .mii; alloc r2=ar.pfs,3,0,0,0 677{ .mii; alloc r2=ar.pfs,3,0,0,0
678 addp4 r33=0,r33
679 addp4 r34=0,r34 };;
680{ .mii; addp4 r32=0,r32
681#else
682{ .mii; alloc r2=ar.pfs,3,0,0,0
683#endif
591 add r14=8,r33 684 add r14=8,r33
592 add r17=8,r34 } 685 add r17=8,r34 }
593 .body 686 .body
@@ -1138,7 +1231,14 @@ bn_sqr_comba4:
1138 .prologue 1231 .prologue
1139 .fframe 0 1232 .fframe 0
1140 .save ar.pfs,r2 1233 .save ar.pfs,r2
1234#if defined(_HPUX_SOURCE) && defined(_ILP32)
1235{ .mii; alloc r2=ar.pfs,2,1,0,0
1236 addp4 r32=0,r32
1237 addp4 r33=0,r33 };;
1238{ .mii;
1239#else
1141{ .mii; alloc r2=ar.pfs,2,1,0,0 1240{ .mii; alloc r2=ar.pfs,2,1,0,0
1241#endif
1142 mov r34=r33 1242 mov r34=r33
1143 add r14=8,r33 };; 1243 add r14=8,r33 };;
1144 .body 1244 .body
@@ -1164,7 +1264,14 @@ bn_mul_comba4:
1164 .prologue 1264 .prologue
1165 .fframe 0 1265 .fframe 0
1166 .save ar.pfs,r2 1266 .save ar.pfs,r2
1267#if defined(_HPUX_SOURCE) && defined(_ILP32)
1268{ .mii; alloc r2=ar.pfs,3,0,0,0
1269 addp4 r33=0,r33
1270 addp4 r34=0,r34 };;
1271{ .mii; addp4 r32=0,r32
1272#else
1167{ .mii; alloc r2=ar.pfs,3,0,0,0 1273{ .mii; alloc r2=ar.pfs,3,0,0,0
1274#endif
1168 add r14=8,r33 1275 add r14=8,r33
1169 add r17=8,r34 } 1276 add r17=8,r34 }
1170 .body 1277 .body
@@ -1464,7 +1571,7 @@ bn_div_words:
1464 or r8=r8,r33 1571 or r8=r8,r33
1465 mov ar.pfs=r2 };; 1572 mov ar.pfs=r2 };;
1466{ .mii; shr.u r9=H,I // remainder if anybody wants it 1573{ .mii; shr.u r9=H,I // remainder if anybody wants it
1467 mov pr=r10,-1 } 1574 mov pr=r10,0x1ffff }
1468{ .mfb; br.ret.sptk.many b0 };; 1575{ .mfb; br.ret.sptk.many b0 };;
1469 1576
1470// Unsigned 64 by 32 (well, by 64 for the moment) bit integer division 1577// Unsigned 64 by 32 (well, by 64 for the moment) bit integer division
diff --git a/src/lib/libcrypto/bn/asm/pa-risc2.s b/src/lib/libcrypto/bn/asm/pa-risc2.s
index af9730d062..f3b16290eb 100644
--- a/src/lib/libcrypto/bn/asm/pa-risc2.s
+++ b/src/lib/libcrypto/bn/asm/pa-risc2.s
@@ -747,8 +747,8 @@ bn_div_words
747 .PROC 747 .PROC
748 .EXPORT bn_div_words,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR,RTNVAL=GR,LONG_RETURN 748 .EXPORT bn_div_words,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR,RTNVAL=GR,LONG_RETURN
749 .IMPORT BN_num_bits_word,CODE 749 .IMPORT BN_num_bits_word,CODE
750 .IMPORT __iob,DATA 750 ;--- not PIC .IMPORT __iob,DATA
751 .IMPORT fprintf,CODE 751 ;--- not PIC .IMPORT fprintf,CODE
752 .IMPORT abort,CODE 752 .IMPORT abort,CODE
753 .IMPORT $$div2U,MILLICODE 753 .IMPORT $$div2U,MILLICODE
754 .CALLINFO CALLER,FRAME=144,ENTRY_GR=%r9,SAVE_RP,ARGS_SAVED,ORDERING_AWARE 754 .CALLINFO CALLER,FRAME=144,ENTRY_GR=%r9,SAVE_RP,ARGS_SAVED,ORDERING_AWARE
@@ -844,12 +844,12 @@ $0006001A
844 MOVIB,TR 2,%r8,$0006001C ;offset 0xa18 844 MOVIB,TR 2,%r8,$0006001C ;offset 0xa18
845 EXTRD,U %r3,63,32,%r7 ;offset 0xa1c 845 EXTRD,U %r3,63,32,%r7 ;offset 0xa1c
846$D2 846$D2
847 ADDIL LR'__iob-$global$,%r27,%r1 ;offset 0xa20 847 ;--- not PIC ADDIL LR'__iob-$global$,%r27,%r1 ;offset 0xa20
848 LDIL LR'C$7,%r21 ;offset 0xa24 848 ;--- not PIC LDIL LR'C$7,%r21 ;offset 0xa24
849 LDO RR'__iob-$global$+32(%r1),%r26 ;offset 0xa28 849 ;--- not PIC LDO RR'__iob-$global$+32(%r1),%r26 ;offset 0xa28
850 .CALL ARGW0=GR,ARGW1=GR,ARGW2=GR,RTNVAL=GR ;in=24,25,26;out=28; 850 ;--- not PIC .CALL ARGW0=GR,ARGW1=GR,ARGW2=GR,RTNVAL=GR ;in=24,25,26;out=28;
851 B,L fprintf,%r2 ;offset 0xa2c 851 ;--- not PIC B,L fprintf,%r2 ;offset 0xa2c
852 LDO RR'C$7(%r21),%r25 ;offset 0xa30 852 ;--- not PIC LDO RR'C$7(%r21),%r25 ;offset 0xa30
853 .CALL ; 853 .CALL ;
854 B,L abort,%r2 ;offset 0xa34 854 B,L abort,%r2 ;offset 0xa34
855 NOP ;offset 0xa38 855 NOP ;offset 0xa38
@@ -1605,14 +1605,14 @@ bn_mul_comba4
1605 .PROCEND 1605 .PROCEND
1606 1606
1607 1607
1608 .SPACE $TEXT$ 1608;--- not PIC .SPACE $TEXT$
1609 .SUBSPA $CODE$ 1609;--- not PIC .SUBSPA $CODE$
1610 .SPACE $PRIVATE$,SORT=16 1610;--- not PIC .SPACE $PRIVATE$,SORT=16
1611 .IMPORT $global$,DATA 1611;--- not PIC .IMPORT $global$,DATA
1612 .SPACE $TEXT$ 1612;--- not PIC .SPACE $TEXT$
1613 .SUBSPA $CODE$ 1613;--- not PIC .SUBSPA $CODE$
1614 .SUBSPA $LIT$,ACCESS=0x2c 1614;--- not PIC .SUBSPA $LIT$,ACCESS=0x2c
1615C$7 1615;--- not PIC C$7
1616 .ALIGN 8 1616;--- not PIC .ALIGN 8
1617 .STRINGZ "Division would overflow (%d)\n" 1617;--- not PIC .STRINGZ "Division would overflow (%d)\n"
1618 .END 1618 .END
diff --git a/src/lib/libcrypto/bn/asm/x86_64-gcc.c b/src/lib/libcrypto/bn/asm/x86_64-gcc.c
new file mode 100644
index 0000000000..b97b394661
--- /dev/null
+++ b/src/lib/libcrypto/bn/asm/x86_64-gcc.c
@@ -0,0 +1,575 @@
1/*
2 * x86_64 BIGNUM accelerator version 0.1, December 2002.
3 *
4 * Implemented by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
5 * project.
6 *
7 * Rights for redistribution and usage in source and binary forms are
8 * granted according to the OpenSSL license. Warranty of any kind is
9 * disclaimed.
10 *
11 * Q. Version 0.1? It doesn't sound like Andy, he used to assign real
12 * versions, like 1.0...
13 * A. Well, that's because this code is basically a quick-n-dirty
14 * proof-of-concept hack. As you can see it's implemented with
15 * inline assembler, which means that you're bound to GCC and that
16 * there must be a room for fine-tuning.
17 *
18 * Q. Why inline assembler?
19 * A. x86_64 features own ABI I'm not familiar with. Which is why
20 * I decided to let the compiler take care of subroutine
21 * prologue/epilogue as well as register allocation.
22 *
23 * Q. How much faster does it get?
24 * A. Unfortunately people sitting on x86_64 hardware are prohibited
25 * to disclose the performance numbers, so they (SuSE labs to be
26 * specific) wouldn't tell me. However! Very similar coding technique
27 * (reaching out for 128-bit result from 64x64-bit multiplication)
28 * results in >3 times performance improvement on MIPS and I see no
29 * reason why gain on x86_64 would be so much different:-)
30 */
31
32#define BN_ULONG unsigned long
33
34/*
35 * "m"(a), "+m"(r) is the way to favor DirectPath µ-code;
36 * "g"(0) let the compiler to decide where does it
37 * want to keep the value of zero;
38 */
39#define mul_add(r,a,word,carry) do { \
40 register BN_ULONG high,low; \
41 asm ("mulq %3" \
42 : "=a"(low),"=d"(high) \
43 : "a"(word),"m"(a) \
44 : "cc"); \
45 asm ("addq %2,%0; adcq %3,%1" \
46 : "+r"(carry),"+d"(high)\
47 : "a"(low),"g"(0) \
48 : "cc"); \
49 asm ("addq %2,%0; adcq %3,%1" \
50 : "+m"(r),"+d"(high) \
51 : "r"(carry),"g"(0) \
52 : "cc"); \
53 carry=high; \
54 } while (0)
55
56#define mul(r,a,word,carry) do { \
57 register BN_ULONG high,low; \
58 asm ("mulq %3" \
59 : "=a"(low),"=d"(high) \
60 : "a"(word),"g"(a) \
61 : "cc"); \
62 asm ("addq %2,%0; adcq %3,%1" \
63 : "+r"(carry),"+d"(high)\
64 : "a"(low),"g"(0) \
65 : "cc"); \
66 (r)=carry, carry=high; \
67 } while (0)
68
69#define sqr(r0,r1,a) \
70 asm ("mulq %2" \
71 : "=a"(r0),"=d"(r1) \
72 : "a"(a) \
73 : "cc");
74
75BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
76 {
77 BN_ULONG c1=0;
78
79 if (num <= 0) return(c1);
80
81 while (num&~3)
82 {
83 mul_add(rp[0],ap[0],w,c1);
84 mul_add(rp[1],ap[1],w,c1);
85 mul_add(rp[2],ap[2],w,c1);
86 mul_add(rp[3],ap[3],w,c1);
87 ap+=4; rp+=4; num-=4;
88 }
89 if (num)
90 {
91 mul_add(rp[0],ap[0],w,c1); if (--num==0) return c1;
92 mul_add(rp[1],ap[1],w,c1); if (--num==0) return c1;
93 mul_add(rp[2],ap[2],w,c1); return c1;
94 }
95
96 return(c1);
97 }
98
99BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
100 {
101 BN_ULONG c1=0;
102
103 if (num <= 0) return(c1);
104
105 while (num&~3)
106 {
107 mul(rp[0],ap[0],w,c1);
108 mul(rp[1],ap[1],w,c1);
109 mul(rp[2],ap[2],w,c1);
110 mul(rp[3],ap[3],w,c1);
111 ap+=4; rp+=4; num-=4;
112 }
113 if (num)
114 {
115 mul(rp[0],ap[0],w,c1); if (--num == 0) return c1;
116 mul(rp[1],ap[1],w,c1); if (--num == 0) return c1;
117 mul(rp[2],ap[2],w,c1);
118 }
119 return(c1);
120 }
121
122void bn_sqr_words(BN_ULONG *r, BN_ULONG *a, int n)
123 {
124 if (n <= 0) return;
125
126 while (n&~3)
127 {
128 sqr(r[0],r[1],a[0]);
129 sqr(r[2],r[3],a[1]);
130 sqr(r[4],r[5],a[2]);
131 sqr(r[6],r[7],a[3]);
132 a+=4; r+=8; n-=4;
133 }
134 if (n)
135 {
136 sqr(r[0],r[1],a[0]); if (--n == 0) return;
137 sqr(r[2],r[3],a[1]); if (--n == 0) return;
138 sqr(r[4],r[5],a[2]);
139 }
140 }
141
142BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
143{ BN_ULONG ret,waste;
144
145 asm ("divq %3"
146 : "=a"(ret),"=d"(waste)
147 : "a"(l),"d"(h),"g"(d)
148 : "cc");
149
150 return ret;
151}
152
153BN_ULONG bn_add_words (BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,int n)
154{ BN_ULONG ret,i;
155
156 if (n <= 0) return 0;
157
158 asm (
159 " subq %2,%2 \n"
160 ".align 16 \n"
161 "1: movq (%4,%2,8),%0 \n"
162 " adcq (%5,%2,8),%0 \n"
163 " movq %0,(%3,%2,8) \n"
164 " leaq 1(%2),%2 \n"
165 " loop 1b \n"
166 " sbbq %0,%0 \n"
167 : "+a"(ret),"+c"(n),"+r"(i)
168 : "r"(rp),"r"(ap),"r"(bp)
169 : "cc"
170 );
171
172 return ret&1;
173}
174
175#ifndef SIMICS
176BN_ULONG bn_sub_words (BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,int n)
177{ BN_ULONG ret,i;
178
179 if (n <= 0) return 0;
180
181 asm (
182 " subq %2,%2 \n"
183 ".align 16 \n"
184 "1: movq (%4,%2,8),%0 \n"
185 " sbbq (%5,%2,8),%0 \n"
186 " movq %0,(%3,%2,8) \n"
187 " leaq 1(%2),%2 \n"
188 " loop 1b \n"
189 " sbbq %0,%0 \n"
190 : "+a"(ret),"+c"(n),"+r"(i)
191 : "r"(rp),"r"(ap),"r"(bp)
192 : "cc"
193 );
194
195 return ret&1;
196}
197#else
198/* Simics 1.4<7 has buggy sbbq:-( */
199#define BN_MASK2 0xffffffffffffffffL
200BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
201 {
202 BN_ULONG t1,t2;
203 int c=0;
204
205 if (n <= 0) return((BN_ULONG)0);
206
207 for (;;)
208 {
209 t1=a[0]; t2=b[0];
210 r[0]=(t1-t2-c)&BN_MASK2;
211 if (t1 != t2) c=(t1 < t2);
212 if (--n <= 0) break;
213
214 t1=a[1]; t2=b[1];
215 r[1]=(t1-t2-c)&BN_MASK2;
216 if (t1 != t2) c=(t1 < t2);
217 if (--n <= 0) break;
218
219 t1=a[2]; t2=b[2];
220 r[2]=(t1-t2-c)&BN_MASK2;
221 if (t1 != t2) c=(t1 < t2);
222 if (--n <= 0) break;
223
224 t1=a[3]; t2=b[3];
225 r[3]=(t1-t2-c)&BN_MASK2;
226 if (t1 != t2) c=(t1 < t2);
227 if (--n <= 0) break;
228
229 a+=4;
230 b+=4;
231 r+=4;
232 }
233 return(c);
234 }
235#endif
236
237/* mul_add_c(a,b,c0,c1,c2) -- c+=a*b for three word number c=(c2,c1,c0) */
238/* mul_add_c2(a,b,c0,c1,c2) -- c+=2*a*b for three word number c=(c2,c1,c0) */
239/* sqr_add_c(a,i,c0,c1,c2) -- c+=a[i]^2 for three word number c=(c2,c1,c0) */
240/* sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number c=(c2,c1,c0) */
241
242#if 0
243/* original macros are kept for reference purposes */
244#define mul_add_c(a,b,c0,c1,c2) { \
245 BN_ULONG ta=(a),tb=(b); \
246 t1 = ta * tb; \
247 t2 = BN_UMULT_HIGH(ta,tb); \
248 c0 += t1; t2 += (c0<t1)?1:0; \
249 c1 += t2; c2 += (c1<t2)?1:0; \
250 }
251
252#define mul_add_c2(a,b,c0,c1,c2) { \
253 BN_ULONG ta=(a),tb=(b),t0; \
254 t1 = BN_UMULT_HIGH(ta,tb); \
255 t0 = ta * tb; \
256 t2 = t1+t1; c2 += (t2<t1)?1:0; \
257 t1 = t0+t0; t2 += (t1<t0)?1:0; \
258 c0 += t1; t2 += (c0<t1)?1:0; \
259 c1 += t2; c2 += (c1<t2)?1:0; \
260 }
261#else
262#define mul_add_c(a,b,c0,c1,c2) do { \
263 asm ("mulq %3" \
264 : "=a"(t1),"=d"(t2) \
265 : "a"(a),"m"(b) \
266 : "cc"); \
267 asm ("addq %2,%0; adcq %3,%1" \
268 : "+r"(c0),"+d"(t2) \
269 : "a"(t1),"g"(0) \
270 : "cc"); \
271 asm ("addq %2,%0; adcq %3,%1" \
272 : "+r"(c1),"+r"(c2) \
273 : "d"(t2),"g"(0) \
274 : "cc"); \
275 } while (0)
276
277#define sqr_add_c(a,i,c0,c1,c2) do { \
278 asm ("mulq %2" \
279 : "=a"(t1),"=d"(t2) \
280 : "a"(a[i]) \
281 : "cc"); \
282 asm ("addq %2,%0; adcq %3,%1" \
283 : "+r"(c0),"+d"(t2) \
284 : "a"(t1),"g"(0) \
285 : "cc"); \
286 asm ("addq %2,%0; adcq %3,%1" \
287 : "+r"(c1),"+r"(c2) \
288 : "d"(t2),"g"(0) \
289 : "cc"); \
290 } while (0)
291
292#define mul_add_c2(a,b,c0,c1,c2) do { \
293 asm ("mulq %3" \
294 : "=a"(t1),"=d"(t2) \
295 : "a"(a),"m"(b) \
296 : "cc"); \
297 asm ("addq %0,%0; adcq %2,%1" \
298 : "+d"(t2),"+r"(c2) \
299 : "g"(0) \
300 : "cc"); \
301 asm ("addq %0,%0; adcq %2,%1" \
302 : "+a"(t1),"+d"(t2) \
303 : "g"(0) \
304 : "cc"); \
305 asm ("addq %2,%0; adcq %3,%1" \
306 : "+r"(c0),"+d"(t2) \
307 : "a"(t1),"g"(0) \
308 : "cc"); \
309 asm ("addq %2,%0; adcq %3,%1" \
310 : "+r"(c1),"+r"(c2) \
311 : "d"(t2),"g"(0) \
312 : "cc"); \
313 } while (0)
314#endif
315
316#define sqr_add_c2(a,i,j,c0,c1,c2) \
317 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
318
319void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
320 {
321 BN_ULONG bl,bh;
322 BN_ULONG t1,t2;
323 BN_ULONG c1,c2,c3;
324
325 c1=0;
326 c2=0;
327 c3=0;
328 mul_add_c(a[0],b[0],c1,c2,c3);
329 r[0]=c1;
330 c1=0;
331 mul_add_c(a[0],b[1],c2,c3,c1);
332 mul_add_c(a[1],b[0],c2,c3,c1);
333 r[1]=c2;
334 c2=0;
335 mul_add_c(a[2],b[0],c3,c1,c2);
336 mul_add_c(a[1],b[1],c3,c1,c2);
337 mul_add_c(a[0],b[2],c3,c1,c2);
338 r[2]=c3;
339 c3=0;
340 mul_add_c(a[0],b[3],c1,c2,c3);
341 mul_add_c(a[1],b[2],c1,c2,c3);
342 mul_add_c(a[2],b[1],c1,c2,c3);
343 mul_add_c(a[3],b[0],c1,c2,c3);
344 r[3]=c1;
345 c1=0;
346 mul_add_c(a[4],b[0],c2,c3,c1);
347 mul_add_c(a[3],b[1],c2,c3,c1);
348 mul_add_c(a[2],b[2],c2,c3,c1);
349 mul_add_c(a[1],b[3],c2,c3,c1);
350 mul_add_c(a[0],b[4],c2,c3,c1);
351 r[4]=c2;
352 c2=0;
353 mul_add_c(a[0],b[5],c3,c1,c2);
354 mul_add_c(a[1],b[4],c3,c1,c2);
355 mul_add_c(a[2],b[3],c3,c1,c2);
356 mul_add_c(a[3],b[2],c3,c1,c2);
357 mul_add_c(a[4],b[1],c3,c1,c2);
358 mul_add_c(a[5],b[0],c3,c1,c2);
359 r[5]=c3;
360 c3=0;
361 mul_add_c(a[6],b[0],c1,c2,c3);
362 mul_add_c(a[5],b[1],c1,c2,c3);
363 mul_add_c(a[4],b[2],c1,c2,c3);
364 mul_add_c(a[3],b[3],c1,c2,c3);
365 mul_add_c(a[2],b[4],c1,c2,c3);
366 mul_add_c(a[1],b[5],c1,c2,c3);
367 mul_add_c(a[0],b[6],c1,c2,c3);
368 r[6]=c1;
369 c1=0;
370 mul_add_c(a[0],b[7],c2,c3,c1);
371 mul_add_c(a[1],b[6],c2,c3,c1);
372 mul_add_c(a[2],b[5],c2,c3,c1);
373 mul_add_c(a[3],b[4],c2,c3,c1);
374 mul_add_c(a[4],b[3],c2,c3,c1);
375 mul_add_c(a[5],b[2],c2,c3,c1);
376 mul_add_c(a[6],b[1],c2,c3,c1);
377 mul_add_c(a[7],b[0],c2,c3,c1);
378 r[7]=c2;
379 c2=0;
380 mul_add_c(a[7],b[1],c3,c1,c2);
381 mul_add_c(a[6],b[2],c3,c1,c2);
382 mul_add_c(a[5],b[3],c3,c1,c2);
383 mul_add_c(a[4],b[4],c3,c1,c2);
384 mul_add_c(a[3],b[5],c3,c1,c2);
385 mul_add_c(a[2],b[6],c3,c1,c2);
386 mul_add_c(a[1],b[7],c3,c1,c2);
387 r[8]=c3;
388 c3=0;
389 mul_add_c(a[2],b[7],c1,c2,c3);
390 mul_add_c(a[3],b[6],c1,c2,c3);
391 mul_add_c(a[4],b[5],c1,c2,c3);
392 mul_add_c(a[5],b[4],c1,c2,c3);
393 mul_add_c(a[6],b[3],c1,c2,c3);
394 mul_add_c(a[7],b[2],c1,c2,c3);
395 r[9]=c1;
396 c1=0;
397 mul_add_c(a[7],b[3],c2,c3,c1);
398 mul_add_c(a[6],b[4],c2,c3,c1);
399 mul_add_c(a[5],b[5],c2,c3,c1);
400 mul_add_c(a[4],b[6],c2,c3,c1);
401 mul_add_c(a[3],b[7],c2,c3,c1);
402 r[10]=c2;
403 c2=0;
404 mul_add_c(a[4],b[7],c3,c1,c2);
405 mul_add_c(a[5],b[6],c3,c1,c2);
406 mul_add_c(a[6],b[5],c3,c1,c2);
407 mul_add_c(a[7],b[4],c3,c1,c2);
408 r[11]=c3;
409 c3=0;
410 mul_add_c(a[7],b[5],c1,c2,c3);
411 mul_add_c(a[6],b[6],c1,c2,c3);
412 mul_add_c(a[5],b[7],c1,c2,c3);
413 r[12]=c1;
414 c1=0;
415 mul_add_c(a[6],b[7],c2,c3,c1);
416 mul_add_c(a[7],b[6],c2,c3,c1);
417 r[13]=c2;
418 c2=0;
419 mul_add_c(a[7],b[7],c3,c1,c2);
420 r[14]=c3;
421 r[15]=c1;
422 }
423
424void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
425 {
426 BN_ULONG bl,bh;
427 BN_ULONG t1,t2;
428 BN_ULONG c1,c2,c3;
429
430 c1=0;
431 c2=0;
432 c3=0;
433 mul_add_c(a[0],b[0],c1,c2,c3);
434 r[0]=c1;
435 c1=0;
436 mul_add_c(a[0],b[1],c2,c3,c1);
437 mul_add_c(a[1],b[0],c2,c3,c1);
438 r[1]=c2;
439 c2=0;
440 mul_add_c(a[2],b[0],c3,c1,c2);
441 mul_add_c(a[1],b[1],c3,c1,c2);
442 mul_add_c(a[0],b[2],c3,c1,c2);
443 r[2]=c3;
444 c3=0;
445 mul_add_c(a[0],b[3],c1,c2,c3);
446 mul_add_c(a[1],b[2],c1,c2,c3);
447 mul_add_c(a[2],b[1],c1,c2,c3);
448 mul_add_c(a[3],b[0],c1,c2,c3);
449 r[3]=c1;
450 c1=0;
451 mul_add_c(a[3],b[1],c2,c3,c1);
452 mul_add_c(a[2],b[2],c2,c3,c1);
453 mul_add_c(a[1],b[3],c2,c3,c1);
454 r[4]=c2;
455 c2=0;
456 mul_add_c(a[2],b[3],c3,c1,c2);
457 mul_add_c(a[3],b[2],c3,c1,c2);
458 r[5]=c3;
459 c3=0;
460 mul_add_c(a[3],b[3],c1,c2,c3);
461 r[6]=c1;
462 r[7]=c2;
463 }
464
465void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
466 {
467 BN_ULONG bl,bh;
468 BN_ULONG t1,t2;
469 BN_ULONG c1,c2,c3;
470
471 c1=0;
472 c2=0;
473 c3=0;
474 sqr_add_c(a,0,c1,c2,c3);
475 r[0]=c1;
476 c1=0;
477 sqr_add_c2(a,1,0,c2,c3,c1);
478 r[1]=c2;
479 c2=0;
480 sqr_add_c(a,1,c3,c1,c2);
481 sqr_add_c2(a,2,0,c3,c1,c2);
482 r[2]=c3;
483 c3=0;
484 sqr_add_c2(a,3,0,c1,c2,c3);
485 sqr_add_c2(a,2,1,c1,c2,c3);
486 r[3]=c1;
487 c1=0;
488 sqr_add_c(a,2,c2,c3,c1);
489 sqr_add_c2(a,3,1,c2,c3,c1);
490 sqr_add_c2(a,4,0,c2,c3,c1);
491 r[4]=c2;
492 c2=0;
493 sqr_add_c2(a,5,0,c3,c1,c2);
494 sqr_add_c2(a,4,1,c3,c1,c2);
495 sqr_add_c2(a,3,2,c3,c1,c2);
496 r[5]=c3;
497 c3=0;
498 sqr_add_c(a,3,c1,c2,c3);
499 sqr_add_c2(a,4,2,c1,c2,c3);
500 sqr_add_c2(a,5,1,c1,c2,c3);
501 sqr_add_c2(a,6,0,c1,c2,c3);
502 r[6]=c1;
503 c1=0;
504 sqr_add_c2(a,7,0,c2,c3,c1);
505 sqr_add_c2(a,6,1,c2,c3,c1);
506 sqr_add_c2(a,5,2,c2,c3,c1);
507 sqr_add_c2(a,4,3,c2,c3,c1);
508 r[7]=c2;
509 c2=0;
510 sqr_add_c(a,4,c3,c1,c2);
511 sqr_add_c2(a,5,3,c3,c1,c2);
512 sqr_add_c2(a,6,2,c3,c1,c2);
513 sqr_add_c2(a,7,1,c3,c1,c2);
514 r[8]=c3;
515 c3=0;
516 sqr_add_c2(a,7,2,c1,c2,c3);
517 sqr_add_c2(a,6,3,c1,c2,c3);
518 sqr_add_c2(a,5,4,c1,c2,c3);
519 r[9]=c1;
520 c1=0;
521 sqr_add_c(a,5,c2,c3,c1);
522 sqr_add_c2(a,6,4,c2,c3,c1);
523 sqr_add_c2(a,7,3,c2,c3,c1);
524 r[10]=c2;
525 c2=0;
526 sqr_add_c2(a,7,4,c3,c1,c2);
527 sqr_add_c2(a,6,5,c3,c1,c2);
528 r[11]=c3;
529 c3=0;
530 sqr_add_c(a,6,c1,c2,c3);
531 sqr_add_c2(a,7,5,c1,c2,c3);
532 r[12]=c1;
533 c1=0;
534 sqr_add_c2(a,7,6,c2,c3,c1);
535 r[13]=c2;
536 c2=0;
537 sqr_add_c(a,7,c3,c1,c2);
538 r[14]=c3;
539 r[15]=c1;
540 }
541
542void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
543 {
544 BN_ULONG bl,bh;
545 BN_ULONG t1,t2;
546 BN_ULONG c1,c2,c3;
547
548 c1=0;
549 c2=0;
550 c3=0;
551 sqr_add_c(a,0,c1,c2,c3);
552 r[0]=c1;
553 c1=0;
554 sqr_add_c2(a,1,0,c2,c3,c1);
555 r[1]=c2;
556 c2=0;
557 sqr_add_c(a,1,c3,c1,c2);
558 sqr_add_c2(a,2,0,c3,c1,c2);
559 r[2]=c3;
560 c3=0;
561 sqr_add_c2(a,3,0,c1,c2,c3);
562 sqr_add_c2(a,2,1,c1,c2,c3);
563 r[3]=c1;
564 c1=0;
565 sqr_add_c(a,2,c2,c3,c1);
566 sqr_add_c2(a,3,1,c2,c3,c1);
567 r[4]=c2;
568 c2=0;
569 sqr_add_c2(a,3,2,c3,c1,c2);
570 r[5]=c3;
571 c3=0;
572 sqr_add_c(a,3,c1,c2,c3);
573 r[6]=c1;
574 r[7]=c2;
575 }
diff --git a/src/lib/libcrypto/bn/bn.h b/src/lib/libcrypto/bn/bn.h
index b40682f831..3da6d8ced9 100644
--- a/src/lib/libcrypto/bn/bn.h
+++ b/src/lib/libcrypto/bn/bn.h
@@ -248,6 +248,8 @@ typedef struct bn_blinding_st
248 BIGNUM *A; 248 BIGNUM *A;
249 BIGNUM *Ai; 249 BIGNUM *Ai;
250 BIGNUM *mod; /* just a reference */ 250 BIGNUM *mod; /* just a reference */
251 unsigned long thread_id; /* added in OpenSSL 0.9.6j and 0.9.7b;
252 * used only by crypto/rsa/rsa_eay.c, rsa_lib.c */
251 } BN_BLINDING; 253 } BN_BLINDING;
252 254
253/* Used for montgomery multiplication */ 255/* Used for montgomery multiplication */
diff --git a/src/lib/libcrypto/bn/bn_div.c b/src/lib/libcrypto/bn/bn_div.c
index f9a095e3b3..580d1201bc 100644
--- a/src/lib/libcrypto/bn/bn_div.c
+++ b/src/lib/libcrypto/bn/bn_div.c
@@ -150,6 +150,20 @@ int BN_div(BIGNUM *dv, BIGNUM *rem, const BIGNUM *m, const BIGNUM *d,
150 q; \ 150 q; \
151 }) 151 })
152# define REMAINDER_IS_ALREADY_CALCULATED 152# define REMAINDER_IS_ALREADY_CALCULATED
153# elif defined(__x86_64) && defined(SIXTY_FOUR_BIT_LONG)
154 /*
155 * Same story here, but it's 128-bit by 64-bit division. Wow!
156 * <appro@fy.chalmers.se>
157 */
158# define bn_div_words(n0,n1,d0) \
159 ({ asm volatile ( \
160 "divq %4" \
161 : "=a"(q), "=d"(rem) \
162 : "a"(n1), "d"(n0), "g"(d0) \
163 : "cc"); \
164 q; \
165 })
166# define REMAINDER_IS_ALREADY_CALCULATED
153# endif /* __<cpu> */ 167# endif /* __<cpu> */
154# endif /* __GNUC__ */ 168# endif /* __GNUC__ */
155#endif /* OPENSSL_NO_ASM */ 169#endif /* OPENSSL_NO_ASM */
@@ -268,6 +282,11 @@ int BN_div(BIGNUM *dv, BIGNUM *rm, const BIGNUM *num, const BIGNUM *divisor,
268 q=(BN_ULONG)(((((BN_ULLONG)n0)<<BN_BITS2)|n1)/d0); 282 q=(BN_ULONG)(((((BN_ULLONG)n0)<<BN_BITS2)|n1)/d0);
269#else 283#else
270 q=bn_div_words(n0,n1,d0); 284 q=bn_div_words(n0,n1,d0);
285#ifdef BN_DEBUG_LEVITTE
286 fprintf(stderr,"DEBUG: bn_div_words(0x%08X,0x%08X,0x%08\
287X) -> 0x%08X\n",
288 n0, n1, d0, q);
289#endif
271#endif 290#endif
272 291
273#ifndef REMAINDER_IS_ALREADY_CALCULATED 292#ifndef REMAINDER_IS_ALREADY_CALCULATED
@@ -292,11 +311,18 @@ int BN_div(BIGNUM *dv, BIGNUM *rm, const BIGNUM *num, const BIGNUM *divisor,
292 BN_ULONG t2l,t2h,ql,qh; 311 BN_ULONG t2l,t2h,ql,qh;
293 312
294 q=bn_div_words(n0,n1,d0); 313 q=bn_div_words(n0,n1,d0);
314#ifdef BN_DEBUG_LEVITTE
315 fprintf(stderr,"DEBUG: bn_div_words(0x%08X,0x%08X,0x%08\
316X) -> 0x%08X\n",
317 n0, n1, d0, q);
318#endif
295#ifndef REMAINDER_IS_ALREADY_CALCULATED 319#ifndef REMAINDER_IS_ALREADY_CALCULATED
296 rem=(n1-q*d0)&BN_MASK2; 320 rem=(n1-q*d0)&BN_MASK2;
297#endif 321#endif
298 322
299#ifdef BN_UMULT_HIGH 323#if defined(BN_UMULT_LOHI)
324 BN_UMULT_LOHI(t2l,t2h,d1,q);
325#elif defined(BN_UMULT_HIGH)
300 t2l = d1 * q; 326 t2l = d1 * q;
301 t2h = BN_UMULT_HIGH(d1,q); 327 t2h = BN_UMULT_HIGH(d1,q);
302#else 328#else
diff --git a/src/lib/libcrypto/bn/bn_lcl.h b/src/lib/libcrypto/bn/bn_lcl.h
index 8a4dba375a..5614bc6164 100644
--- a/src/lib/libcrypto/bn/bn_lcl.h
+++ b/src/lib/libcrypto/bn/bn_lcl.h
@@ -230,6 +230,21 @@ struct bignum_ctx
230 : "r"(a), "r"(b)); \ 230 : "r"(a), "r"(b)); \
231 ret; }) 231 ret; })
232# endif /* compiler */ 232# endif /* compiler */
233# elif defined(__x86_64) && defined(SIXTY_FOUR_BIT_LONG)
234# if defined(__GNUC__)
235# define BN_UMULT_HIGH(a,b) ({ \
236 register BN_ULONG ret,discard; \
237 asm ("mulq %3" \
238 : "=a"(discard),"=d"(ret) \
239 : "a"(a), "g"(b) \
240 : "cc"); \
241 ret; })
242# define BN_UMULT_LOHI(low,high,a,b) \
243 asm ("mulq %3" \
244 : "=a"(low),"=d"(high) \
245 : "a"(a),"g"(b) \
246 : "cc");
247# endif
233# endif /* cpu */ 248# endif /* cpu */
234#endif /* OPENSSL_NO_ASM */ 249#endif /* OPENSSL_NO_ASM */
235 250
@@ -337,7 +352,7 @@ struct bignum_ctx
337 352
338#define LBITS(a) ((a)&BN_MASK2l) 353#define LBITS(a) ((a)&BN_MASK2l)
339#define HBITS(a) (((a)>>BN_BITS4)&BN_MASK2l) 354#define HBITS(a) (((a)>>BN_BITS4)&BN_MASK2l)
340#define L2HBITS(a) ((BN_ULONG)((a)&BN_MASK2l)<<BN_BITS4) 355#define L2HBITS(a) (((a)<<BN_BITS4)&BN_MASK2)
341 356
342#define LLBITS(a) ((a)&BN_MASKl) 357#define LLBITS(a) ((a)&BN_MASKl)
343#define LHBITS(a) (((a)>>BN_BITS2)&BN_MASKl) 358#define LHBITS(a) (((a)>>BN_BITS2)&BN_MASKl)
@@ -353,7 +368,7 @@ struct bignum_ctx
353 lt=(bl)*(lt); \ 368 lt=(bl)*(lt); \
354 m1=(bl)*(ht); \ 369 m1=(bl)*(ht); \
355 ht =(bh)*(ht); \ 370 ht =(bh)*(ht); \
356 m=(m+m1)&BN_MASK2; if (m < m1) ht+=L2HBITS(1L); \ 371 m=(m+m1)&BN_MASK2; if (m < m1) ht+=L2HBITS((BN_ULONG)1); \
357 ht+=HBITS(m); \ 372 ht+=HBITS(m); \
358 m1=L2HBITS(m); \ 373 m1=L2HBITS(m); \
359 lt=(lt+m1)&BN_MASK2; if (lt < m1) ht++; \ 374 lt=(lt+m1)&BN_MASK2; if (lt < m1) ht++; \
@@ -418,20 +433,19 @@ void bn_sqr_comba4(BN_ULONG *r,const BN_ULONG *a);
418int bn_cmp_words(const BN_ULONG *a,const BN_ULONG *b,int n); 433int bn_cmp_words(const BN_ULONG *a,const BN_ULONG *b,int n);
419int bn_cmp_part_words(const BN_ULONG *a, const BN_ULONG *b, 434int bn_cmp_part_words(const BN_ULONG *a, const BN_ULONG *b,
420 int cl, int dl); 435 int cl, int dl);
436#if 0
437/* bn_mul.c rollback <appro> */
421void bn_mul_recursive(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,int n2, 438void bn_mul_recursive(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,int n2,
422 int dna,int dnb,BN_ULONG *t); 439 int dna,int dnb,BN_ULONG *t);
423void bn_mul_part_recursive(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b, 440void bn_mul_part_recursive(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,
424 int n,int tna,int tnb,BN_ULONG *t); 441 int n,int tna,int tnb,BN_ULONG *t);
442#endif
425void bn_sqr_recursive(BN_ULONG *r,const BN_ULONG *a, int n2, BN_ULONG *t); 443void bn_sqr_recursive(BN_ULONG *r,const BN_ULONG *a, int n2, BN_ULONG *t);
426void bn_mul_low_normal(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b, int n); 444void bn_mul_low_normal(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b, int n);
427void bn_mul_low_recursive(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,int n2, 445void bn_mul_low_recursive(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,int n2,
428 BN_ULONG *t); 446 BN_ULONG *t);
429void bn_mul_high(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,BN_ULONG *l,int n2, 447void bn_mul_high(BN_ULONG *r,BN_ULONG *a,BN_ULONG *b,BN_ULONG *l,int n2,
430 BN_ULONG *t); 448 BN_ULONG *t);
431BN_ULONG bn_add_part_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
432 int cl, int dl);
433BN_ULONG bn_sub_part_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
434 int cl, int dl);
435 449
436#ifdef __cplusplus 450#ifdef __cplusplus
437} 451}
diff --git a/src/lib/libcrypto/bn/bn_lib.c b/src/lib/libcrypto/bn/bn_lib.c
index 8abe095af2..fa0ff485ad 100644
--- a/src/lib/libcrypto/bn/bn_lib.c
+++ b/src/lib/libcrypto/bn/bn_lib.c
@@ -263,12 +263,12 @@ void BN_clear_free(BIGNUM *a)
263 if (a == NULL) return; 263 if (a == NULL) return;
264 if (a->d != NULL) 264 if (a->d != NULL)
265 { 265 {
266 memset(a->d,0,a->dmax*sizeof(a->d[0])); 266 OPENSSL_cleanse(a->d,a->dmax*sizeof(a->d[0]));
267 if (!(BN_get_flags(a,BN_FLG_STATIC_DATA))) 267 if (!(BN_get_flags(a,BN_FLG_STATIC_DATA)))
268 OPENSSL_free(a->d); 268 OPENSSL_free(a->d);
269 } 269 }
270 i=BN_get_flags(a,BN_FLG_MALLOCED); 270 i=BN_get_flags(a,BN_FLG_MALLOCED);
271 memset(a,0,sizeof(BIGNUM)); 271 OPENSSL_cleanse(a,sizeof(BIGNUM));
272 if (i) 272 if (i)
273 OPENSSL_free(a); 273 OPENSSL_free(a);
274 } 274 }
diff --git a/src/lib/libcrypto/bn/bn_mul.c b/src/lib/libcrypto/bn/bn_mul.c
index b03458d002..cb93ac3356 100644
--- a/src/lib/libcrypto/bn/bn_mul.c
+++ b/src/lib/libcrypto/bn/bn_mul.c
@@ -56,325 +56,10 @@
56 * [including the GNU Public Licence.] 56 * [including the GNU Public Licence.]
57 */ 57 */
58 58
59#ifndef BN_DEBUG
60# undef NDEBUG /* avoid conflicting definitions */
61# define NDEBUG
62#endif
63
64#include <stdio.h> 59#include <stdio.h>
65#include <assert.h>
66#include "cryptlib.h" 60#include "cryptlib.h"
67#include "bn_lcl.h" 61#include "bn_lcl.h"
68 62
69#if defined(OPENSSL_NO_ASM) || !(defined(__i386) || defined(__i386__)) || defined(__DJGPP__) /* Assembler implementation exists only for x86 */
70/* Here follows specialised variants of bn_add_words() and
71 bn_sub_words(). They have the property performing operations on
72 arrays of different sizes. The sizes of those arrays is expressed through
73 cl, which is the common length ( basicall, min(len(a),len(b)) ), and dl,
74 which is the delta between the two lengths, calculated as len(a)-len(b).
75 All lengths are the number of BN_ULONGs... For the operations that require
76 a result array as parameter, it must have the length cl+abs(dl).
77 These functions should probably end up in bn_asm.c as soon as there are
78 assembler counterparts for the systems that use assembler files. */
79
80BN_ULONG bn_sub_part_words(BN_ULONG *r,
81 const BN_ULONG *a, const BN_ULONG *b,
82 int cl, int dl)
83 {
84 BN_ULONG c, t;
85
86 assert(cl >= 0);
87 c = bn_sub_words(r, a, b, cl);
88
89 if (dl == 0)
90 return c;
91
92 r += cl;
93 a += cl;
94 b += cl;
95
96 if (dl < 0)
97 {
98#ifdef BN_COUNT
99 fprintf(stderr, " bn_sub_part_words %d + %d (dl < 0, c = %d)\n", cl, dl, c);
100#endif
101 for (;;)
102 {
103 t = b[0];
104 r[0] = (0-t-c)&BN_MASK2;
105 if (t != 0) c=1;
106 if (++dl >= 0) break;
107
108 t = b[1];
109 r[1] = (0-t-c)&BN_MASK2;
110 if (t != 0) c=1;
111 if (++dl >= 0) break;
112
113 t = b[2];
114 r[2] = (0-t-c)&BN_MASK2;
115 if (t != 0) c=1;
116 if (++dl >= 0) break;
117
118 t = b[3];
119 r[3] = (0-t-c)&BN_MASK2;
120 if (t != 0) c=1;
121 if (++dl >= 0) break;
122
123 b += 4;
124 r += 4;
125 }
126 }
127 else
128 {
129 int save_dl = dl;
130#ifdef BN_COUNT
131 fprintf(stderr, " bn_sub_part_words %d + %d (dl > 0, c = %d)\n", cl, dl, c);
132#endif
133 while(c)
134 {
135 t = a[0];
136 r[0] = (t-c)&BN_MASK2;
137 if (t != 0) c=0;
138 if (--dl <= 0) break;
139
140 t = a[1];
141 r[1] = (t-c)&BN_MASK2;
142 if (t != 0) c=0;
143 if (--dl <= 0) break;
144
145 t = a[2];
146 r[2] = (t-c)&BN_MASK2;
147 if (t != 0) c=0;
148 if (--dl <= 0) break;
149
150 t = a[3];
151 r[3] = (t-c)&BN_MASK2;
152 if (t != 0) c=0;
153 if (--dl <= 0) break;
154
155 save_dl = dl;
156 a += 4;
157 r += 4;
158 }
159 if (dl > 0)
160 {
161#ifdef BN_COUNT
162 fprintf(stderr, " bn_sub_part_words %d + %d (dl > 0, c == 0)\n", cl, dl);
163#endif
164 if (save_dl > dl)
165 {
166 switch (save_dl - dl)
167 {
168 case 1:
169 r[1] = a[1];
170 if (--dl <= 0) break;
171 case 2:
172 r[2] = a[2];
173 if (--dl <= 0) break;
174 case 3:
175 r[3] = a[3];
176 if (--dl <= 0) break;
177 }
178 a += 4;
179 r += 4;
180 }
181 }
182 if (dl > 0)
183 {
184#ifdef BN_COUNT
185 fprintf(stderr, " bn_sub_part_words %d + %d (dl > 0, copy)\n", cl, dl);
186#endif
187 for(;;)
188 {
189 r[0] = a[0];
190 if (--dl <= 0) break;
191 r[1] = a[1];
192 if (--dl <= 0) break;
193 r[2] = a[2];
194 if (--dl <= 0) break;
195 r[3] = a[3];
196 if (--dl <= 0) break;
197
198 a += 4;
199 r += 4;
200 }
201 }
202 }
203 return c;
204 }
205#endif
206
207BN_ULONG bn_add_part_words(BN_ULONG *r,
208 const BN_ULONG *a, const BN_ULONG *b,
209 int cl, int dl)
210 {
211 BN_ULONG c, l, t;
212
213 assert(cl >= 0);
214 c = bn_add_words(r, a, b, cl);
215
216 if (dl == 0)
217 return c;
218
219 r += cl;
220 a += cl;
221 b += cl;
222
223 if (dl < 0)
224 {
225 int save_dl = dl;
226#ifdef BN_COUNT
227 fprintf(stderr, " bn_add_part_words %d + %d (dl < 0, c = %d)\n", cl, dl, c);
228#endif
229 while (c)
230 {
231 l=(c+b[0])&BN_MASK2;
232 c=(l < c);
233 r[0]=l;
234 if (++dl >= 0) break;
235
236 l=(c+b[1])&BN_MASK2;
237 c=(l < c);
238 r[1]=l;
239 if (++dl >= 0) break;
240
241 l=(c+b[2])&BN_MASK2;
242 c=(l < c);
243 r[2]=l;
244 if (++dl >= 0) break;
245
246 l=(c+b[3])&BN_MASK2;
247 c=(l < c);
248 r[3]=l;
249 if (++dl >= 0) break;
250
251 save_dl = dl;
252 b+=4;
253 r+=4;
254 }
255 if (dl < 0)
256 {
257#ifdef BN_COUNT
258 fprintf(stderr, " bn_add_part_words %d + %d (dl < 0, c == 0)\n", cl, dl);
259#endif
260 if (save_dl < dl)
261 {
262 switch (dl - save_dl)
263 {
264 case 1:
265 r[1] = b[1];
266 if (++dl >= 0) break;
267 case 2:
268 r[2] = b[2];
269 if (++dl >= 0) break;
270 case 3:
271 r[3] = b[3];
272 if (++dl >= 0) break;
273 }
274 b += 4;
275 r += 4;
276 }
277 }
278 if (dl < 0)
279 {
280#ifdef BN_COUNT
281 fprintf(stderr, " bn_add_part_words %d + %d (dl < 0, copy)\n", cl, dl);
282#endif
283 for(;;)
284 {
285 r[0] = b[0];
286 if (++dl >= 0) break;
287 r[1] = b[1];
288 if (++dl >= 0) break;
289 r[2] = b[2];
290 if (++dl >= 0) break;
291 r[3] = b[3];
292 if (++dl >= 0) break;
293
294 b += 4;
295 r += 4;
296 }
297 }
298 }
299 else
300 {
301 int save_dl = dl;
302#ifdef BN_COUNT
303 fprintf(stderr, " bn_add_part_words %d + %d (dl > 0)\n", cl, dl);
304#endif
305 while (c)
306 {
307 t=(a[0]+c)&BN_MASK2;
308 c=(t < c);
309 r[0]=t;
310 if (--dl <= 0) break;
311
312 t=(a[1]+c)&BN_MASK2;
313 c=(t < c);
314 r[1]=t;
315 if (--dl <= 0) break;
316
317 t=(a[2]+c)&BN_MASK2;
318 c=(t < c);
319 r[2]=t;
320 if (--dl <= 0) break;
321
322 t=(a[3]+c)&BN_MASK2;
323 c=(t < c);
324 r[3]=t;
325 if (--dl <= 0) break;
326
327 save_dl = dl;
328 a+=4;
329 r+=4;
330 }
331#ifdef BN_COUNT
332 fprintf(stderr, " bn_add_part_words %d + %d (dl > 0, c == 0)\n", cl, dl);
333#endif
334 if (dl > 0)
335 {
336 if (save_dl > dl)
337 {
338 switch (save_dl - dl)
339 {
340 case 1:
341 r[1] = a[1];
342 if (--dl <= 0) break;
343 case 2:
344 r[2] = a[2];
345 if (--dl <= 0) break;
346 case 3:
347 r[3] = a[3];
348 if (--dl <= 0) break;
349 }
350 a += 4;
351 r += 4;
352 }
353 }
354 if (dl > 0)
355 {
356#ifdef BN_COUNT
357 fprintf(stderr, " bn_add_part_words %d + %d (dl > 0, copy)\n", cl, dl);
358#endif
359 for(;;)
360 {
361 r[0] = a[0];
362 if (--dl <= 0) break;
363 r[1] = a[1];
364 if (--dl <= 0) break;
365 r[2] = a[2];
366 if (--dl <= 0) break;
367 r[3] = a[3];
368 if (--dl <= 0) break;
369
370 a += 4;
371 r += 4;
372 }
373 }
374 }
375 return c;
376 }
377
378#ifdef BN_RECURSION 63#ifdef BN_RECURSION
379/* Karatsuba recursive multiplication algorithm 64/* Karatsuba recursive multiplication algorithm
380 * (cf. Knuth, The Art of Computer Programming, Vol. 2) */ 65 * (cf. Knuth, The Art of Computer Programming, Vol. 2) */
@@ -390,15 +75,14 @@ BN_ULONG bn_add_part_words(BN_ULONG *r,
390 * a[1]*b[1] 75 * a[1]*b[1]
391 */ 76 */
392void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, 77void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
393 int dna, int dnb, BN_ULONG *t) 78 BN_ULONG *t)
394 { 79 {
395 int n=n2/2,c1,c2; 80 int n=n2/2,c1,c2;
396 int tna=n+dna, tnb=n+dnb;
397 unsigned int neg,zero; 81 unsigned int neg,zero;
398 BN_ULONG ln,lo,*p; 82 BN_ULONG ln,lo,*p;
399 83
400# ifdef BN_COUNT 84# ifdef BN_COUNT
401 fprintf(stderr," bn_mul_recursive %d * %d\n",n2,n2); 85 printf(" bn_mul_recursive %d * %d\n",n2,n2);
402# endif 86# endif
403# ifdef BN_MUL_COMBA 87# ifdef BN_MUL_COMBA
404# if 0 88# if 0
@@ -408,40 +92,34 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
408 return; 92 return;
409 } 93 }
410# endif 94# endif
411 /* Only call bn_mul_comba 8 if n2 == 8 and the 95 if (n2 == 8)
412 * two arrays are complete [steve]
413 */
414 if (n2 == 8 && dna == 0 && dnb == 0)
415 { 96 {
416 bn_mul_comba8(r,a,b); 97 bn_mul_comba8(r,a,b);
417 return; 98 return;
418 } 99 }
419# endif /* BN_MUL_COMBA */ 100# endif /* BN_MUL_COMBA */
420 /* Else do normal multiply */
421 if (n2 < BN_MUL_RECURSIVE_SIZE_NORMAL) 101 if (n2 < BN_MUL_RECURSIVE_SIZE_NORMAL)
422 { 102 {
423 bn_mul_normal(r,a,n2+dna,b,n2+dnb); 103 /* This should not happen */
424 if ((dna + dnb) < 0) 104 bn_mul_normal(r,a,n2,b,n2);
425 memset(&r[2*n2 + dna + dnb], 0,
426 sizeof(BN_ULONG) * -(dna + dnb));
427 return; 105 return;
428 } 106 }
429 /* r=(a[0]-a[1])*(b[1]-b[0]) */ 107 /* r=(a[0]-a[1])*(b[1]-b[0]) */
430 c1=bn_cmp_part_words(a,&(a[n]),tna,n-tna); 108 c1=bn_cmp_words(a,&(a[n]),n);
431 c2=bn_cmp_part_words(&(b[n]),b,tnb,tnb-n); 109 c2=bn_cmp_words(&(b[n]),b,n);
432 zero=neg=0; 110 zero=neg=0;
433 switch (c1*3+c2) 111 switch (c1*3+c2)
434 { 112 {
435 case -4: 113 case -4:
436 bn_sub_part_words(t, &(a[n]),a, tna,tna-n); /* - */ 114 bn_sub_words(t, &(a[n]),a, n); /* - */
437 bn_sub_part_words(&(t[n]),b, &(b[n]),tnb,n-tnb); /* - */ 115 bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */
438 break; 116 break;
439 case -3: 117 case -3:
440 zero=1; 118 zero=1;
441 break; 119 break;
442 case -2: 120 case -2:
443 bn_sub_part_words(t, &(a[n]),a, tna,tna-n); /* - */ 121 bn_sub_words(t, &(a[n]),a, n); /* - */
444 bn_sub_part_words(&(t[n]),&(b[n]),b, tnb,tnb-n); /* + */ 122 bn_sub_words(&(t[n]),&(b[n]),b, n); /* + */
445 neg=1; 123 neg=1;
446 break; 124 break;
447 case -1: 125 case -1:
@@ -450,22 +128,21 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
450 zero=1; 128 zero=1;
451 break; 129 break;
452 case 2: 130 case 2:
453 bn_sub_part_words(t, a, &(a[n]),tna,n-tna); /* + */ 131 bn_sub_words(t, a, &(a[n]),n); /* + */
454 bn_sub_part_words(&(t[n]),b, &(b[n]),tnb,n-tnb); /* - */ 132 bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */
455 neg=1; 133 neg=1;
456 break; 134 break;
457 case 3: 135 case 3:
458 zero=1; 136 zero=1;
459 break; 137 break;
460 case 4: 138 case 4:
461 bn_sub_part_words(t, a, &(a[n]),tna,n-tna); 139 bn_sub_words(t, a, &(a[n]),n);
462 bn_sub_part_words(&(t[n]),&(b[n]),b, tnb,tnb-n); 140 bn_sub_words(&(t[n]),&(b[n]),b, n);
463 break; 141 break;
464 } 142 }
465 143
466# ifdef BN_MUL_COMBA 144# ifdef BN_MUL_COMBA
467 if (n == 4 && dna == 0 && dnb == 0) /* XXX: bn_mul_comba4 could take 145 if (n == 4)
468 extra args to do this well */
469 { 146 {
470 if (!zero) 147 if (!zero)
471 bn_mul_comba4(&(t[n2]),t,&(t[n])); 148 bn_mul_comba4(&(t[n2]),t,&(t[n]));
@@ -475,9 +152,7 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
475 bn_mul_comba4(r,a,b); 152 bn_mul_comba4(r,a,b);
476 bn_mul_comba4(&(r[n2]),&(a[n]),&(b[n])); 153 bn_mul_comba4(&(r[n2]),&(a[n]),&(b[n]));
477 } 154 }
478 else if (n == 8 && dna == 0 && dnb == 0) /* XXX: bn_mul_comba8 could 155 else if (n == 8)
479 take extra args to do this
480 well */
481 { 156 {
482 if (!zero) 157 if (!zero)
483 bn_mul_comba8(&(t[n2]),t,&(t[n])); 158 bn_mul_comba8(&(t[n2]),t,&(t[n]));
@@ -492,11 +167,11 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
492 { 167 {
493 p= &(t[n2*2]); 168 p= &(t[n2*2]);
494 if (!zero) 169 if (!zero)
495 bn_mul_recursive(&(t[n2]),t,&(t[n]),n,0,0,p); 170 bn_mul_recursive(&(t[n2]),t,&(t[n]),n,p);
496 else 171 else
497 memset(&(t[n2]),0,n2*sizeof(BN_ULONG)); 172 memset(&(t[n2]),0,n2*sizeof(BN_ULONG));
498 bn_mul_recursive(r,a,b,n,0,0,p); 173 bn_mul_recursive(r,a,b,n,p);
499 bn_mul_recursive(&(r[n2]),&(a[n]),&(b[n]),n,dna,dnb,p); 174 bn_mul_recursive(&(r[n2]),&(a[n]),&(b[n]),n,p);
500 } 175 }
501 176
502 /* t[32] holds (a[0]-a[1])*(b[1]-b[0]), c1 is the sign 177 /* t[32] holds (a[0]-a[1])*(b[1]-b[0]), c1 is the sign
@@ -545,39 +220,39 @@ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
545 220
546/* n+tn is the word length 221/* n+tn is the word length
547 * t needs to be n*4 is size, as does r */ 222 * t needs to be n*4 is size, as does r */
548void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n, 223void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int tn,
549 int tna, int tnb, BN_ULONG *t) 224 int n, BN_ULONG *t)
550 { 225 {
551 int i,j,n2=n*2; 226 int i,j,n2=n*2;
552 unsigned int c1,c2,neg,zero; 227 unsigned int c1,c2,neg,zero;
553 BN_ULONG ln,lo,*p; 228 BN_ULONG ln,lo,*p;
554 229
555# ifdef BN_COUNT 230# ifdef BN_COUNT
556 fprintf(stderr," bn_mul_part_recursive (%d+%d) * (%d+%d)\n", 231 printf(" bn_mul_part_recursive %d * %d\n",tn+n,tn+n);
557 tna, n, tnb, n);
558# endif 232# endif
559 if (n < 8) 233 if (n < 8)
560 { 234 {
561 bn_mul_normal(r,a,n+tna,b,n+tnb); 235 i=tn+n;
236 bn_mul_normal(r,a,i,b,i);
562 return; 237 return;
563 } 238 }
564 239
565 /* r=(a[0]-a[1])*(b[1]-b[0]) */ 240 /* r=(a[0]-a[1])*(b[1]-b[0]) */
566 c1=bn_cmp_part_words(a,&(a[n]),tna,n-tna); 241 c1=bn_cmp_words(a,&(a[n]),n);
567 c2=bn_cmp_part_words(&(b[n]),b,tnb,tnb-n); 242 c2=bn_cmp_words(&(b[n]),b,n);
568 zero=neg=0; 243 zero=neg=0;
569 switch (c1*3+c2) 244 switch (c1*3+c2)
570 { 245 {
571 case -4: 246 case -4:
572 bn_sub_part_words(t, &(a[n]),a, tna,tna-n); /* - */ 247 bn_sub_words(t, &(a[n]),a, n); /* - */
573 bn_sub_part_words(&(t[n]),b, &(b[n]),tnb,n-tnb); /* - */ 248 bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */
574 break; 249 break;
575 case -3: 250 case -3:
576 zero=1; 251 zero=1;
577 /* break; */ 252 /* break; */
578 case -2: 253 case -2:
579 bn_sub_part_words(t, &(a[n]),a, tna,tna-n); /* - */ 254 bn_sub_words(t, &(a[n]),a, n); /* - */
580 bn_sub_part_words(&(t[n]),&(b[n]),b, tnb,tnb-n); /* + */ 255 bn_sub_words(&(t[n]),&(b[n]),b, n); /* + */
581 neg=1; 256 neg=1;
582 break; 257 break;
583 case -1: 258 case -1:
@@ -586,16 +261,16 @@ void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n,
586 zero=1; 261 zero=1;
587 /* break; */ 262 /* break; */
588 case 2: 263 case 2:
589 bn_sub_part_words(t, a, &(a[n]),tna,n-tna); /* + */ 264 bn_sub_words(t, a, &(a[n]),n); /* + */
590 bn_sub_part_words(&(t[n]),b, &(b[n]),tnb,n-tnb); /* - */ 265 bn_sub_words(&(t[n]),b, &(b[n]),n); /* - */
591 neg=1; 266 neg=1;
592 break; 267 break;
593 case 3: 268 case 3:
594 zero=1; 269 zero=1;
595 /* break; */ 270 /* break; */
596 case 4: 271 case 4:
597 bn_sub_part_words(t, a, &(a[n]),tna,n-tna); 272 bn_sub_words(t, a, &(a[n]),n);
598 bn_sub_part_words(&(t[n]),&(b[n]),b, tnb,tnb-n); 273 bn_sub_words(&(t[n]),&(b[n]),b, n);
599 break; 274 break;
600 } 275 }
601 /* The zero case isn't yet implemented here. The speedup 276 /* The zero case isn't yet implemented here. The speedup
@@ -614,59 +289,54 @@ void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n,
614 { 289 {
615 bn_mul_comba8(&(t[n2]),t,&(t[n])); 290 bn_mul_comba8(&(t[n2]),t,&(t[n]));
616 bn_mul_comba8(r,a,b); 291 bn_mul_comba8(r,a,b);
617 bn_mul_normal(&(r[n2]),&(a[n]),tna,&(b[n]),tnb); 292 bn_mul_normal(&(r[n2]),&(a[n]),tn,&(b[n]),tn);
618 memset(&(r[n2+tna+tnb]),0,sizeof(BN_ULONG)*(n2-tna-tnb)); 293 memset(&(r[n2+tn*2]),0,sizeof(BN_ULONG)*(n2-tn*2));
619 } 294 }
620 else 295 else
621 { 296 {
622 p= &(t[n2*2]); 297 p= &(t[n2*2]);
623 bn_mul_recursive(&(t[n2]),t,&(t[n]),n,0,0,p); 298 bn_mul_recursive(&(t[n2]),t,&(t[n]),n,p);
624 bn_mul_recursive(r,a,b,n,0,0,p); 299 bn_mul_recursive(r,a,b,n,p);
625 i=n/2; 300 i=n/2;
626 /* If there is only a bottom half to the number, 301 /* If there is only a bottom half to the number,
627 * just do it */ 302 * just do it */
628 if (tna > tnb) 303 j=tn-i;
629 j = tna - i;
630 else
631 j = tnb - i;
632 if (j == 0) 304 if (j == 0)
633 { 305 {
634 bn_mul_recursive(&(r[n2]),&(a[n]),&(b[n]), 306 bn_mul_recursive(&(r[n2]),&(a[n]),&(b[n]),i,p);
635 i,tna-i,tnb-i,p);
636 memset(&(r[n2+i*2]),0,sizeof(BN_ULONG)*(n2-i*2)); 307 memset(&(r[n2+i*2]),0,sizeof(BN_ULONG)*(n2-i*2));
637 } 308 }
638 else if (j > 0) /* eg, n == 16, i == 8 and tn == 11 */ 309 else if (j > 0) /* eg, n == 16, i == 8 and tn == 11 */
639 { 310 {
640 bn_mul_part_recursive(&(r[n2]),&(a[n]),&(b[n]), 311 bn_mul_part_recursive(&(r[n2]),&(a[n]),&(b[n]),
641 i,tna-i,tnb-i,p); 312 j,i,p);
642 memset(&(r[n2+tna+tnb]),0, 313 memset(&(r[n2+tn*2]),0,
643 sizeof(BN_ULONG)*(n2-tna-tnb)); 314 sizeof(BN_ULONG)*(n2-tn*2));
644 } 315 }
645 else /* (j < 0) eg, n == 16, i == 8 and tn == 5 */ 316 else /* (j < 0) eg, n == 16, i == 8 and tn == 5 */
646 { 317 {
647 memset(&(r[n2]),0,sizeof(BN_ULONG)*n2); 318 memset(&(r[n2]),0,sizeof(BN_ULONG)*n2);
648 if (tna < BN_MUL_RECURSIVE_SIZE_NORMAL 319 if (tn < BN_MUL_RECURSIVE_SIZE_NORMAL)
649 && tnb < BN_MUL_RECURSIVE_SIZE_NORMAL)
650 { 320 {
651 bn_mul_normal(&(r[n2]),&(a[n]),tna,&(b[n]),tnb); 321 bn_mul_normal(&(r[n2]),&(a[n]),tn,&(b[n]),tn);
652 } 322 }
653 else 323 else
654 { 324 {
655 for (;;) 325 for (;;)
656 { 326 {
657 i/=2; 327 i/=2;
658 if (i < tna && i < tnb) 328 if (i < tn)
659 { 329 {
660 bn_mul_part_recursive(&(r[n2]), 330 bn_mul_part_recursive(&(r[n2]),
661 &(a[n]),&(b[n]), 331 &(a[n]),&(b[n]),
662 i,tna-i,tnb-i,p); 332 tn-i,i,p);
663 break; 333 break;
664 } 334 }
665 else if (i <= tna && i <= tnb) 335 else if (i == tn)
666 { 336 {
667 bn_mul_recursive(&(r[n2]), 337 bn_mul_recursive(&(r[n2]),
668 &(a[n]),&(b[n]), 338 &(a[n]),&(b[n]),
669 i,tna-i,tnb-i,p); 339 i,p);
670 break; 340 break;
671 } 341 }
672 } 342 }
@@ -727,10 +397,10 @@ void bn_mul_low_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
727 int n=n2/2; 397 int n=n2/2;
728 398
729# ifdef BN_COUNT 399# ifdef BN_COUNT
730 fprintf(stderr," bn_mul_low_recursive %d * %d\n",n2,n2); 400 printf(" bn_mul_low_recursive %d * %d\n",n2,n2);
731# endif 401# endif
732 402
733 bn_mul_recursive(r,a,b,n,0,0,&(t[0])); 403 bn_mul_recursive(r,a,b,n,&(t[0]));
734 if (n >= BN_MUL_LOW_RECURSIVE_SIZE_NORMAL) 404 if (n >= BN_MUL_LOW_RECURSIVE_SIZE_NORMAL)
735 { 405 {
736 bn_mul_low_recursive(&(t[0]),&(a[0]),&(b[n]),n,&(t[n2])); 406 bn_mul_low_recursive(&(t[0]),&(a[0]),&(b[n]),n,&(t[n2]));
@@ -761,7 +431,7 @@ void bn_mul_high(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, BN_ULONG *l, int n2,
761 BN_ULONG ll,lc,*lp,*mp; 431 BN_ULONG ll,lc,*lp,*mp;
762 432
763# ifdef BN_COUNT 433# ifdef BN_COUNT
764 fprintf(stderr," bn_mul_high %d * %d\n",n2,n2); 434 printf(" bn_mul_high %d * %d\n",n2,n2);
765# endif 435# endif
766 n=n2/2; 436 n=n2/2;
767 437
@@ -814,8 +484,8 @@ void bn_mul_high(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, BN_ULONG *l, int n2,
814 else 484 else
815# endif 485# endif
816 { 486 {
817 bn_mul_recursive(&(t[0]),&(r[0]),&(r[n]),n,0,0,&(t[n2])); 487 bn_mul_recursive(&(t[0]),&(r[0]),&(r[n]),n,&(t[n2]));
818 bn_mul_recursive(r,&(a[n]),&(b[n]),n,0,0,&(t[n2])); 488 bn_mul_recursive(r,&(a[n]),&(b[n]),n,&(t[n2]));
819 } 489 }
820 490
821 /* s0 == low(al*bl) 491 /* s0 == low(al*bl)
@@ -940,19 +610,19 @@ void bn_mul_high(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, BN_ULONG *l, int n2,
940 610
941int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx) 611int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx)
942 { 612 {
943 int ret=0;
944 int top,al,bl; 613 int top,al,bl;
945 BIGNUM *rr; 614 BIGNUM *rr;
615 int ret = 0;
946#if defined(BN_MUL_COMBA) || defined(BN_RECURSION) 616#if defined(BN_MUL_COMBA) || defined(BN_RECURSION)
947 int i; 617 int i;
948#endif 618#endif
949#ifdef BN_RECURSION 619#ifdef BN_RECURSION
950 BIGNUM *t=NULL; 620 BIGNUM *t;
951 int j=0,k; 621 int j,k;
952#endif 622#endif
953 623
954#ifdef BN_COUNT 624#ifdef BN_COUNT
955 fprintf(stderr,"BN_mul %d * %d\n",a->top,b->top); 625 printf("BN_mul %d * %d\n",a->top,b->top);
956#endif 626#endif
957 627
958 bn_check_top(a); 628 bn_check_top(a);
@@ -1005,55 +675,21 @@ int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx)
1005#ifdef BN_RECURSION 675#ifdef BN_RECURSION
1006 if ((al >= BN_MULL_SIZE_NORMAL) && (bl >= BN_MULL_SIZE_NORMAL)) 676 if ((al >= BN_MULL_SIZE_NORMAL) && (bl >= BN_MULL_SIZE_NORMAL))
1007 { 677 {
1008 if (i >= -1 && i <= 1) 678 if (i == 1 && !BN_get_flags(b,BN_FLG_STATIC_DATA) && bl<b->dmax)
1009 { 679 {
1010 int sav_j =0; 680#if 0 /* tribute to const-ification, bl<b->dmax above covers for this */
1011 /* Find out the power of two lower or equal 681 if (bn_wexpand(b,al) == NULL) goto err;
1012 to the longest of the two numbers */ 682#endif
1013 if (i >= 0) 683 b->d[bl]=0;
1014 {
1015 j = BN_num_bits_word((BN_ULONG)al);
1016 }
1017 if (i == -1)
1018 {
1019 j = BN_num_bits_word((BN_ULONG)bl);
1020 }
1021 sav_j = j;
1022 j = 1<<(j-1);
1023 assert(j <= al || j <= bl);
1024 k = j+j;
1025 t = BN_CTX_get(ctx);
1026 if (al > j || bl > j)
1027 {
1028 bn_wexpand(t,k*4);
1029 bn_wexpand(rr,k*4);
1030 bn_mul_part_recursive(rr->d,a->d,b->d,
1031 j,al-j,bl-j,t->d);
1032 }
1033 else /* al <= j || bl <= j */
1034 {
1035 bn_wexpand(t,k*2);
1036 bn_wexpand(rr,k*2);
1037 bn_mul_recursive(rr->d,a->d,b->d,
1038 j,al-j,bl-j,t->d);
1039 }
1040 rr->top=top;
1041 goto end;
1042 }
1043#if 0
1044 if (i == 1 && !BN_get_flags(b,BN_FLG_STATIC_DATA))
1045 {
1046 BIGNUM *tmp_bn = (BIGNUM *)b;
1047 if (bn_wexpand(tmp_bn,al) == NULL) goto err;
1048 tmp_bn->d[bl]=0;
1049 bl++; 684 bl++;
1050 i--; 685 i--;
1051 } 686 }
1052 else if (i == -1 && !BN_get_flags(a,BN_FLG_STATIC_DATA)) 687 else if (i == -1 && !BN_get_flags(a,BN_FLG_STATIC_DATA) && al<a->dmax)
1053 { 688 {
1054 BIGNUM *tmp_bn = (BIGNUM *)a; 689#if 0 /* tribute to const-ification, al<a->dmax above covers for this */
1055 if (bn_wexpand(tmp_bn,bl) == NULL) goto err; 690 if (bn_wexpand(a,bl) == NULL) goto err;
1056 tmp_bn->d[al]=0; 691#endif
692 a->d[al]=0;
1057 al++; 693 al++;
1058 i++; 694 i++;
1059 } 695 }
@@ -1070,17 +706,26 @@ int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx)
1070 if (bn_wexpand(t,k*2) == NULL) goto err; 706 if (bn_wexpand(t,k*2) == NULL) goto err;
1071 if (bn_wexpand(rr,k*2) == NULL) goto err; 707 if (bn_wexpand(rr,k*2) == NULL) goto err;
1072 bn_mul_recursive(rr->d,a->d,b->d,al,t->d); 708 bn_mul_recursive(rr->d,a->d,b->d,al,t->d);
709 rr->top=top;
710 goto end;
1073 } 711 }
712#if 0 /* tribute to const-ification, rsa/dsa performance is not affected */
1074 else 713 else
1075 { 714 {
1076 if (bn_wexpand(t,k*4) == NULL) goto err; 715 if (bn_wexpand(a,k) == NULL ) goto err;
1077 if (bn_wexpand(rr,k*4) == NULL) goto err; 716 if (bn_wexpand(b,k) == NULL ) goto err;
717 if (bn_wexpand(t,k*4) == NULL ) goto err;
718 if (bn_wexpand(rr,k*4) == NULL ) goto err;
719 for (i=a->top; i<k; i++)
720 a->d[i]=0;
721 for (i=b->top; i<k; i++)
722 b->d[i]=0;
1078 bn_mul_part_recursive(rr->d,a->d,b->d,al-j,j,t->d); 723 bn_mul_part_recursive(rr->d,a->d,b->d,al-j,j,t->d);
1079 } 724 }
1080 rr->top=top; 725 rr->top=top;
1081 goto end; 726 goto end;
1082 }
1083#endif 727#endif
728 }
1084 } 729 }
1085#endif /* BN_RECURSION */ 730#endif /* BN_RECURSION */
1086 if (bn_wexpand(rr,top) == NULL) goto err; 731 if (bn_wexpand(rr,top) == NULL) goto err;
@@ -1103,7 +748,7 @@ void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb)
1103 BN_ULONG *rr; 748 BN_ULONG *rr;
1104 749
1105#ifdef BN_COUNT 750#ifdef BN_COUNT
1106 fprintf(stderr," bn_mul_normal %d * %d\n",na,nb); 751 printf(" bn_mul_normal %d * %d\n",na,nb);
1107#endif 752#endif
1108 753
1109 if (na < nb) 754 if (na < nb)
@@ -1116,13 +761,7 @@ void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb)
1116 761
1117 } 762 }
1118 rr= &(r[na]); 763 rr= &(r[na]);
1119 if (nb <= 0) 764 rr[0]=bn_mul_words(r,a,na,b[0]);
1120 {
1121 (void)bn_mul_words(r,a,na,0);
1122 return;
1123 }
1124 else
1125 rr[0]=bn_mul_words(r,a,na,b[0]);
1126 765
1127 for (;;) 766 for (;;)
1128 { 767 {
@@ -1143,7 +782,7 @@ void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb)
1143void bn_mul_low_normal(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) 782void bn_mul_low_normal(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
1144 { 783 {
1145#ifdef BN_COUNT 784#ifdef BN_COUNT
1146 fprintf(stderr," bn_mul_low_normal %d * %d\n",n,n); 785 printf(" bn_mul_low_normal %d * %d\n",n,n);
1147#endif 786#endif
1148 bn_mul_words(r,a,n,b[0]); 787 bn_mul_words(r,a,n,b[0]);
1149 788
diff --git a/src/lib/libcrypto/bn/bn_prime.c b/src/lib/libcrypto/bn/bn_prime.c
index 918b9237c6..e072d9255c 100644
--- a/src/lib/libcrypto/bn/bn_prime.c
+++ b/src/lib/libcrypto/bn/bn_prime.c
@@ -140,6 +140,7 @@ BIGNUM *BN_generate_prime(BIGNUM *ret, int bits, int safe,
140 BN_CTX *ctx; 140 BN_CTX *ctx;
141 int checks = BN_prime_checks_for_size(bits); 141 int checks = BN_prime_checks_for_size(bits);
142 142
143 BN_init(&t);
143 ctx=BN_CTX_new(); 144 ctx=BN_CTX_new();
144 if (ctx == NULL) goto err; 145 if (ctx == NULL) goto err;
145 if (ret == NULL) 146 if (ret == NULL)
@@ -148,7 +149,6 @@ BIGNUM *BN_generate_prime(BIGNUM *ret, int bits, int safe,
148 } 149 }
149 else 150 else
150 rnd=ret; 151 rnd=ret;
151 BN_init(&t);
152loop: 152loop:
153 /* make a random number and set the top and bottom bits */ 153 /* make a random number and set the top and bottom bits */
154 if (add == NULL) 154 if (add == NULL)
diff --git a/src/lib/libcrypto/bn/bn_rand.c b/src/lib/libcrypto/bn/bn_rand.c
index 9e08ccd22e..893c9d2af9 100644
--- a/src/lib/libcrypto/bn/bn_rand.c
+++ b/src/lib/libcrypto/bn/bn_rand.c
@@ -201,7 +201,7 @@ static int bnrand(int pseudorand, BIGNUM *rnd, int bits, int top, int bottom)
201err: 201err:
202 if (buf != NULL) 202 if (buf != NULL)
203 { 203 {
204 memset(buf,0,bytes); 204 OPENSSL_cleanse(buf,bytes);
205 OPENSSL_free(buf); 205 OPENSSL_free(buf);
206 } 206 }
207 return(ret); 207 return(ret);
diff --git a/src/lib/libcrypto/bn/bn_word.c b/src/lib/libcrypto/bn/bn_word.c
index cd59baa2c4..988e0ca7b3 100644
--- a/src/lib/libcrypto/bn/bn_word.c
+++ b/src/lib/libcrypto/bn/bn_word.c
@@ -123,7 +123,10 @@ int BN_add_word(BIGNUM *a, BN_ULONG w)
123 i=0; 123 i=0;
124 for (;;) 124 for (;;)
125 { 125 {
126 l=(a->d[i]+(BN_ULONG)w)&BN_MASK2; 126 if (i >= a->top)
127 l=w;
128 else
129 l=(a->d[i]+(BN_ULONG)w)&BN_MASK2;
127 a->d[i]=l; 130 a->d[i]=l;
128 if (w > l) 131 if (w > l)
129 w=1; 132 w=1;