]> git.ipfire.org Git - thirdparty/openssl.git/blob - crypto/aes/asm/aes-x86_64.pl
projects / thirdparty / openssl.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
CHANGES entry: for ECC parameters with NULL or zero cofactor, compute it
[thirdparty/openssl.git] / crypto / aes / asm / aes-x86_64.pl
1 #! /usr/bin/env perl
2 # Copyright 2005-2019 The OpenSSL Project Authors. All Rights Reserved.
3 #
4 # Licensed under the OpenSSL license (the "License"). You may not use
5 # this file except in compliance with the License. You can obtain a copy
6 # in the file LICENSE in the source distribution or at
7 # https://www.openssl.org/source/license.html
8
9 #
10 # ====================================================================
11 # Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
12 # project. The module is, however, dual licensed under OpenSSL and
13 # CRYPTOGAMS licenses depending on where you obtain it. For further
14 # details see http://www.openssl.org/~appro/cryptogams/.
15 # ====================================================================
16 #
17 # Version 2.1.
18 #
19 # aes-*-cbc benchmarks are improved by >70% [compared to gcc 3.3.2 on
20 # Opteron 240 CPU] plus all the bells-n-whistles from 32-bit version
21 # [you'll notice a lot of resemblance], such as compressed S-boxes
22 # in little-endian byte order, prefetch of these tables in CBC mode,
23 # as well as avoiding L1 cache aliasing between stack frame and key
24 # schedule and already mentioned tables, compressed Td4...
25 #
26 # Performance in number of cycles per processed byte for 128-bit key:
27 #
28 # ECB encrypt ECB decrypt CBC large chunk
29 # AMD64 33 43 13.0
30 # EM64T 38 56 18.6(*)
31 # Core 2 30 42 14.5(*)
32 # Atom 65 86 32.1(*)
33 #
34 # (*) with hyper-threading off
35
36 $flavour = shift;
37 $output = shift;
38 if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
39
40 $win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
41
42 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
43 ( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
44 ( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
45 die "can't locate x86_64-xlate.pl";
46
47 open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
48 *STDOUT=*OUT;
49
50 $verticalspin=1; # unlike 32-bit version $verticalspin performs
51 # ~15% better on both AMD and Intel cores
52 $speed_limit=512; # see aes-586.pl for details
53
54 $code=".text\n";
55
56 $s0="%eax";
57 $s1="%ebx";
58 $s2="%ecx";
59 $s3="%edx";
60 $acc0="%esi"; $mask80="%rsi";
61 $acc1="%edi"; $maskfe="%rdi";
62 $acc2="%ebp"; $mask1b="%rbp";
63 $inp="%r8";
64 $out="%r9";
65 $t0="%r10d";
66 $t1="%r11d";
67 $t2="%r12d";
68 $rnds="%r13d";
69 $sbox="%r14";
70 $key="%r15";
71
72 sub hi() { my $r=shift; $r =~ s/%[er]([a-d])x/%\1h/; $r; }
73 sub lo() { my $r=shift; $r =~ s/%[er]([a-d])x/%\1l/;
74 $r =~ s/%[er]([sd]i)/%\1l/;
75 $r =~ s/%(r[0-9]+)[d]?/%\1b/; $r; }
76 sub LO() { my $r=shift; $r =~ s/%r([a-z]+)/%e\1/;
77 $r =~ s/%r([0-9]+)/%r\1d/; $r; }
78 sub _data_word()
79 { my $i;
80 while(defined($i=shift)) { $code.=sprintf".long\t0x%08x,0x%08x\n",$i,$i; }
81 }
82 sub data_word()
83 { my $i;
84 my $last=pop(@_);
85 $code.=".long\t";
86 while(defined($i=shift)) { $code.=sprintf"0x%08x,",$i; }
87 $code.=sprintf"0x%08x\n",$last;
88 }
89
90 sub data_byte()
91 { my $i;
92 my $last=pop(@_);
93 $code.=".byte\t";
94 while(defined($i=shift)) { $code.=sprintf"0x%02x,",$i&0xff; }
95 $code.=sprintf"0x%02x\n",$last&0xff;
96 }
97
98 sub encvert()
99 { my $t3="%r8d"; # zaps $inp!
100
101 $code.=<<___;
102 # favor 3-way issue Opteron pipeline...
103 movzb `&lo("$s0")`,$acc0
104 movzb `&lo("$s1")`,$acc1
105 movzb `&lo("$s2")`,$acc2
106 mov 0($sbox,$acc0,8),$t0
107 mov 0($sbox,$acc1,8),$t1
108 mov 0($sbox,$acc2,8),$t2
109
110 movzb `&hi("$s1")`,$acc0
111 movzb `&hi("$s2")`,$acc1
112 movzb `&lo("$s3")`,$acc2
113 xor 3($sbox,$acc0,8),$t0
114 xor 3($sbox,$acc1,8),$t1
115 mov 0($sbox,$acc2,8),$t3
116
117 movzb `&hi("$s3")`,$acc0
118 shr \$16,$s2
119 movzb `&hi("$s0")`,$acc2
120 xor 3($sbox,$acc0,8),$t2
121 shr \$16,$s3
122 xor 3($sbox,$acc2,8),$t3
123
124 shr \$16,$s1
125 lea 16($key),$key
126 shr \$16,$s0
127
128 movzb `&lo("$s2")`,$acc0
129 movzb `&lo("$s3")`,$acc1
130 movzb `&lo("$s0")`,$acc2
131 xor 2($sbox,$acc0,8),$t0
132 xor 2($sbox,$acc1,8),$t1
133 xor 2($sbox,$acc2,8),$t2
134
135 movzb `&hi("$s3")`,$acc0
136 movzb `&hi("$s0")`,$acc1
137 movzb `&lo("$s1")`,$acc2
138 xor 1($sbox,$acc0,8),$t0
139 xor 1($sbox,$acc1,8),$t1
140 xor 2($sbox,$acc2,8),$t3
141
142 mov 12($key),$s3
143 movzb `&hi("$s1")`,$acc1
144 movzb `&hi("$s2")`,$acc2
145 mov 0($key),$s0
146 xor 1($sbox,$acc1,8),$t2
147 xor 1($sbox,$acc2,8),$t3
148
149 mov 4($key),$s1
150 mov 8($key),$s2
151 xor $t0,$s0
152 xor $t1,$s1
153 xor $t2,$s2
154 xor $t3,$s3
155 ___
156 }
157
158 sub enclastvert()
159 { my $t3="%r8d"; # zaps $inp!
160
161 $code.=<<___;
162 movzb `&lo("$s0")`,$acc0
163 movzb `&lo("$s1")`,$acc1
164 movzb `&lo("$s2")`,$acc2
165 movzb 2($sbox,$acc0,8),$t0
166 movzb 2($sbox,$acc1,8),$t1
167 movzb 2($sbox,$acc2,8),$t2
168
169 movzb `&lo("$s3")`,$acc0
170 movzb `&hi("$s1")`,$acc1
171 movzb `&hi("$s2")`,$acc2
172 movzb 2($sbox,$acc0,8),$t3
173 mov 0($sbox,$acc1,8),$acc1 #$t0
174 mov 0($sbox,$acc2,8),$acc2 #$t1
175
176 and \$0x0000ff00,$acc1
177 and \$0x0000ff00,$acc2
178
179 xor $acc1,$t0
180 xor $acc2,$t1
181 shr \$16,$s2
182
183 movzb `&hi("$s3")`,$acc0
184 movzb `&hi("$s0")`,$acc1
185 shr \$16,$s3
186 mov 0($sbox,$acc0,8),$acc0 #$t2
187 mov 0($sbox,$acc1,8),$acc1 #$t3
188
189 and \$0x0000ff00,$acc0
190 and \$0x0000ff00,$acc1
191 shr \$16,$s1
192 xor $acc0,$t2
193 xor $acc1,$t3
194 shr \$16,$s0
195
196 movzb `&lo("$s2")`,$acc0
197 movzb `&lo("$s3")`,$acc1
198 movzb `&lo("$s0")`,$acc2
199 mov 0($sbox,$acc0,8),$acc0 #$t0
200 mov 0($sbox,$acc1,8),$acc1 #$t1
201 mov 0($sbox,$acc2,8),$acc2 #$t2
202
203 and \$0x00ff0000,$acc0
204 and \$0x00ff0000,$acc1
205 and \$0x00ff0000,$acc2
206
207 xor $acc0,$t0
208 xor $acc1,$t1
209 xor $acc2,$t2
210
211 movzb `&lo("$s1")`,$acc0
212 movzb `&hi("$s3")`,$acc1
213 movzb `&hi("$s0")`,$acc2
214 mov 0($sbox,$acc0,8),$acc0 #$t3
215 mov 2($sbox,$acc1,8),$acc1 #$t0
216 mov 2($sbox,$acc2,8),$acc2 #$t1
217
218 and \$0x00ff0000,$acc0
219 and \$0xff000000,$acc1
220 and \$0xff000000,$acc2
221
222 xor $acc0,$t3
223 xor $acc1,$t0
224 xor $acc2,$t1
225
226 movzb `&hi("$s1")`,$acc0
227 movzb `&hi("$s2")`,$acc1
228 mov 16+12($key),$s3
229 mov 2($sbox,$acc0,8),$acc0 #$t2
230 mov 2($sbox,$acc1,8),$acc1 #$t3
231 mov 16+0($key),$s0
232
233 and \$0xff000000,$acc0
234 and \$0xff000000,$acc1
235
236 xor $acc0,$t2
237 xor $acc1,$t3
238
239 mov 16+4($key),$s1
240 mov 16+8($key),$s2
241 xor $t0,$s0
242 xor $t1,$s1
243 xor $t2,$s2
244 xor $t3,$s3
245 ___
246 }
247
248 sub encstep()
249 { my ($i,@s) = @_;
250 my $tmp0=$acc0;
251 my $tmp1=$acc1;
252 my $tmp2=$acc2;
253 my $out=($t0,$t1,$t2,$s[0])[$i];
254
255 if ($i==3) {
256 $tmp0=$s[1];
257 $tmp1=$s[2];
258 $tmp2=$s[3];
259 }
260 $code.=" movzb ".&lo($s[0]).",$out\n";
261 $code.=" mov $s[2],$tmp1\n" if ($i!=3);
262 $code.=" lea 16($key),$key\n" if ($i==0);
263
264 $code.=" movzb ".&hi($s[1]).",$tmp0\n";
265 $code.=" mov 0($sbox,$out,8),$out\n";
266
267 $code.=" shr \$16,$tmp1\n";
268 $code.=" mov $s[3],$tmp2\n" if ($i!=3);
269 $code.=" xor 3($sbox,$tmp0,8),$out\n";
270
271 $code.=" movzb ".&lo($tmp1).",$tmp1\n";
272 $code.=" shr \$24,$tmp2\n";
273 $code.=" xor 4*$i($key),$out\n";
274
275 $code.=" xor 2($sbox,$tmp1,8),$out\n";
276 $code.=" xor 1($sbox,$tmp2,8),$out\n";
277
278 $code.=" mov $t0,$s[1]\n" if ($i==3);
279 $code.=" mov $t1,$s[2]\n" if ($i==3);
280 $code.=" mov $t2,$s[3]\n" if ($i==3);
281 $code.="\n";
282 }
283
284 sub enclast()
285 { my ($i,@s)=@_;
286 my $tmp0=$acc0;
287 my $tmp1=$acc1;
288 my $tmp2=$acc2;
289 my $out=($t0,$t1,$t2,$s[0])[$i];
290
291 if ($i==3) {
292 $tmp0=$s[1];
293 $tmp1=$s[2];
294 $tmp2=$s[3];
295 }
296 $code.=" movzb ".&lo($s[0]).",$out\n";
297 $code.=" mov $s[2],$tmp1\n" if ($i!=3);
298
299 $code.=" mov 2($sbox,$out,8),$out\n";
300 $code.=" shr \$16,$tmp1\n";
301 $code.=" mov $s[3],$tmp2\n" if ($i!=3);
302
303 $code.=" and \$0x000000ff,$out\n";
304 $code.=" movzb ".&hi($s[1]).",$tmp0\n";
305 $code.=" movzb ".&lo($tmp1).",$tmp1\n";
306 $code.=" shr \$24,$tmp2\n";
307
308 $code.=" mov 0($sbox,$tmp0,8),$tmp0\n";
309 $code.=" mov 0($sbox,$tmp1,8),$tmp1\n";
310 $code.=" mov 2($sbox,$tmp2,8),$tmp2\n";
311
312 $code.=" and \$0x0000ff00,$tmp0\n";
313 $code.=" and \$0x00ff0000,$tmp1\n";
314 $code.=" and \$0xff000000,$tmp2\n";
315
316 $code.=" xor $tmp0,$out\n";
317 $code.=" mov $t0,$s[1]\n" if ($i==3);
318 $code.=" xor $tmp1,$out\n";
319 $code.=" mov $t1,$s[2]\n" if ($i==3);
320 $code.=" xor $tmp2,$out\n";
321 $code.=" mov $t2,$s[3]\n" if ($i==3);
322 $code.="\n";
323 }
324
325 $code.=<<___;
326 .type _x86_64_AES_encrypt,\@abi-omnipotent
327 .align 16
328 _x86_64_AES_encrypt:
329 xor 0($key),$s0 # xor with key
330 xor 4($key),$s1
331 xor 8($key),$s2
332 xor 12($key),$s3
333
334 mov 240($key),$rnds # load key->rounds
335 sub \$1,$rnds
336 jmp .Lenc_loop
337 .align 16
338 .Lenc_loop:
339 ___
340 if ($verticalspin) { &encvert(); }
341 else { &encstep(0,$s0,$s1,$s2,$s3);
342 &encstep(1,$s1,$s2,$s3,$s0);
343 &encstep(2,$s2,$s3,$s0,$s1);
344 &encstep(3,$s3,$s0,$s1,$s2);
345 }
346 $code.=<<___;
347 sub \$1,$rnds
348 jnz .Lenc_loop
349 ___
350 if ($verticalspin) { &enclastvert(); }
351 else { &enclast(0,$s0,$s1,$s2,$s3);
352 &enclast(1,$s1,$s2,$s3,$s0);
353 &enclast(2,$s2,$s3,$s0,$s1);
354 &enclast(3,$s3,$s0,$s1,$s2);
355 $code.=<<___;
356 xor 16+0($key),$s0 # xor with key
357 xor 16+4($key),$s1
358 xor 16+8($key),$s2
359 xor 16+12($key),$s3
360 ___
361 }
362 $code.=<<___;
363 .byte 0xf3,0xc3 # rep ret
364 .size _x86_64_AES_encrypt,.-_x86_64_AES_encrypt
365 ___
366
367 # it's possible to implement this by shifting tN by 8, filling least
368 # significant byte with byte load and finally bswap-ing at the end,
369 # but such partial register load kills Core 2...
370 sub enccompactvert()
371 { my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d");
372
373 $code.=<<___;
374 movzb `&lo("$s0")`,$t0
375 movzb `&lo("$s1")`,$t1
376 movzb `&lo("$s2")`,$t2
377 movzb `&lo("$s3")`,$t3
378 movzb `&hi("$s1")`,$acc0
379 movzb `&hi("$s2")`,$acc1
380 shr \$16,$s2
381 movzb `&hi("$s3")`,$acc2
382 movzb ($sbox,$t0,1),$t0
383 movzb ($sbox,$t1,1),$t1
384 movzb ($sbox,$t2,1),$t2
385 movzb ($sbox,$t3,1),$t3
386
387 movzb ($sbox,$acc0,1),$t4 #$t0
388 movzb `&hi("$s0")`,$acc0
389 movzb ($sbox,$acc1,1),$t5 #$t1
390 movzb `&lo("$s2")`,$acc1
391 movzb ($sbox,$acc2,1),$acc2 #$t2
392 movzb ($sbox,$acc0,1),$acc0 #$t3
393
394 shl \$8,$t4
395 shr \$16,$s3
396 shl \$8,$t5
397 xor $t4,$t0
398 shr \$16,$s0
399 movzb `&lo("$s3")`,$t4
400 shr \$16,$s1
401 xor $t5,$t1
402 shl \$8,$acc2
403 movzb `&lo("$s0")`,$t5
404 movzb ($sbox,$acc1,1),$acc1 #$t0
405 xor $acc2,$t2
406
407 shl \$8,$acc0
408 movzb `&lo("$s1")`,$acc2
409 shl \$16,$acc1
410 xor $acc0,$t3
411 movzb ($sbox,$t4,1),$t4 #$t1
412 movzb `&hi("$s3")`,$acc0
413 movzb ($sbox,$t5,1),$t5 #$t2
414 xor $acc1,$t0
415
416 shr \$8,$s2
417 movzb `&hi("$s0")`,$acc1
418 shl \$16,$t4
419 shr \$8,$s1
420 shl \$16,$t5
421 xor $t4,$t1
422 movzb ($sbox,$acc2,1),$acc2 #$t3
423 movzb ($sbox,$acc0,1),$acc0 #$t0
424 movzb ($sbox,$acc1,1),$acc1 #$t1
425 movzb ($sbox,$s2,1),$s3 #$t3
426 movzb ($sbox,$s1,1),$s2 #$t2
427
428 shl \$16,$acc2
429 xor $t5,$t2
430 shl \$24,$acc0
431 xor $acc2,$t3
432 shl \$24,$acc1
433 xor $acc0,$t0
434 shl \$24,$s3
435 xor $acc1,$t1
436 shl \$24,$s2
437 mov $t0,$s0
438 mov $t1,$s1
439 xor $t2,$s2
440 xor $t3,$s3
441 ___
442 }
443
444 sub enctransform_ref()
445 { my $sn = shift;
446 my ($acc,$r2,$tmp)=("%r8d","%r9d","%r13d");
447
448 $code.=<<___;
449 mov $sn,$acc
450 and \$0x80808080,$acc
451 mov $acc,$tmp
452 shr \$7,$tmp
453 lea ($sn,$sn),$r2
454 sub $tmp,$acc
455 and \$0xfefefefe,$r2
456 and \$0x1b1b1b1b,$acc
457 mov $sn,$tmp
458 xor $acc,$r2
459
460 xor $r2,$sn
461 rol \$24,$sn
462 xor $r2,$sn
463 ror \$16,$tmp
464 xor $tmp,$sn
465 ror \$8,$tmp
466 xor $tmp,$sn
467 ___
468 }
469
470 # unlike decrypt case it does not pay off to parallelize enctransform
471 sub enctransform()
472 { my ($t3,$r20,$r21)=($acc2,"%r8d","%r9d");
473
474 $code.=<<___;
475 mov \$0x80808080,$t0
476 mov \$0x80808080,$t1
477 and $s0,$t0
478 and $s1,$t1
479 mov $t0,$acc0
480 mov $t1,$acc1
481 shr \$7,$t0
482 lea ($s0,$s0),$r20
483 shr \$7,$t1
484 lea ($s1,$s1),$r21
485 sub $t0,$acc0
486 sub $t1,$acc1
487 and \$0xfefefefe,$r20
488 and \$0xfefefefe,$r21
489 and \$0x1b1b1b1b,$acc0
490 and \$0x1b1b1b1b,$acc1
491 mov $s0,$t0
492 mov $s1,$t1
493 xor $acc0,$r20
494 xor $acc1,$r21
495
496 xor $r20,$s0
497 xor $r21,$s1
498 mov \$0x80808080,$t2
499 rol \$24,$s0
500 mov \$0x80808080,$t3
501 rol \$24,$s1
502 and $s2,$t2
503 and $s3,$t3
504 xor $r20,$s0
505 xor $r21,$s1
506 mov $t2,$acc0
507 ror \$16,$t0
508 mov $t3,$acc1
509 ror \$16,$t1
510 lea ($s2,$s2),$r20
511 shr \$7,$t2
512 xor $t0,$s0
513 shr \$7,$t3
514 xor $t1,$s1
515 ror \$8,$t0
516 lea ($s3,$s3),$r21
517 ror \$8,$t1
518 sub $t2,$acc0
519 sub $t3,$acc1
520 xor $t0,$s0
521 xor $t1,$s1
522
523 and \$0xfefefefe,$r20
524 and \$0xfefefefe,$r21
525 and \$0x1b1b1b1b,$acc0
526 and \$0x1b1b1b1b,$acc1
527 mov $s2,$t2
528 mov $s3,$t3
529 xor $acc0,$r20
530 xor $acc1,$r21
531
532 ror \$16,$t2
533 xor $r20,$s2
534 ror \$16,$t3
535 xor $r21,$s3
536 rol \$24,$s2
537 mov 0($sbox),$acc0 # prefetch Te4
538 rol \$24,$s3
539 xor $r20,$s2
540 mov 64($sbox),$acc1
541 xor $r21,$s3
542 mov 128($sbox),$r20
543 xor $t2,$s2
544 ror \$8,$t2
545 xor $t3,$s3
546 ror \$8,$t3
547 xor $t2,$s2
548 mov 192($sbox),$r21
549 xor $t3,$s3
550 ___
551 }
552
553 $code.=<<___;
554 .type _x86_64_AES_encrypt_compact,\@abi-omnipotent
555 .align 16
556 _x86_64_AES_encrypt_compact:
557 .cfi_startproc
558 lea 128($sbox),$inp # size optimization
559 mov 0-128($inp),$acc1 # prefetch Te4
560 mov 32-128($inp),$acc2
561 mov 64-128($inp),$t0
562 mov 96-128($inp),$t1
563 mov 128-128($inp),$acc1
564 mov 160-128($inp),$acc2
565 mov 192-128($inp),$t0
566 mov 224-128($inp),$t1
567 jmp .Lenc_loop_compact
568 .align 16
569 .Lenc_loop_compact:
570 xor 0($key),$s0 # xor with key
571 xor 4($key),$s1
572 xor 8($key),$s2
573 xor 12($key),$s3
574 lea 16($key),$key
575 ___
576 &enccompactvert();
577 $code.=<<___;
578 cmp 16(%rsp),$key
579 je .Lenc_compact_done
580 ___
581 &enctransform();
582 $code.=<<___;
583 jmp .Lenc_loop_compact
584 .align 16
585 .Lenc_compact_done:
586 xor 0($key),$s0
587 xor 4($key),$s1
588 xor 8($key),$s2
589 xor 12($key),$s3
590 .byte 0xf3,0xc3 # rep ret
591 .cfi_endproc
592 .size _x86_64_AES_encrypt_compact,.-_x86_64_AES_encrypt_compact
593 ___
594
595 # void AES_encrypt (const void *inp,void *out,const AES_KEY *key);
596 $code.=<<___;
597 .globl AES_encrypt
598 .type AES_encrypt,\@function,3
599 .align 16
600 .globl asm_AES_encrypt
601 .hidden asm_AES_encrypt
602 asm_AES_encrypt:
603 AES_encrypt:
604 .cfi_startproc
605 mov %rsp,%rax
606 .cfi_def_cfa_register %rax
607 push %rbx
608 .cfi_push %rbx
609 push %rbp
610 .cfi_push %rbp
611 push %r12
612 .cfi_push %r12
613 push %r13
614 .cfi_push %r13
615 push %r14
616 .cfi_push %r14
617 push %r15
618 .cfi_push %r15
619
620 # allocate frame "above" key schedule
621 lea -63(%rdx),%rcx # %rdx is key argument
622 and \$-64,%rsp
623 sub %rsp,%rcx
624 neg %rcx
625 and \$0x3c0,%rcx
626 sub %rcx,%rsp
627 sub \$32,%rsp
628
629 mov %rsi,16(%rsp) # save out
630 mov %rax,24(%rsp) # save original stack pointer
631 .cfi_cfa_expression %rsp+24,deref,+8
632 .Lenc_prologue:
633
634 mov %rdx,$key
635 mov 240($key),$rnds # load rounds
636
637 mov 0(%rdi),$s0 # load input vector
638 mov 4(%rdi),$s1
639 mov 8(%rdi),$s2
640 mov 12(%rdi),$s3
641
642 shl \$4,$rnds
643 lea ($key,$rnds),%rbp
644 mov $key,(%rsp) # key schedule
645 mov %rbp,8(%rsp) # end of key schedule
646
647 # pick Te4 copy which can't "overlap" with stack frame or key schedule
648 lea .LAES_Te+2048(%rip),$sbox
649 lea 768(%rsp),%rbp
650 sub $sbox,%rbp
651 and \$0x300,%rbp
652 lea ($sbox,%rbp),$sbox
653
654 call _x86_64_AES_encrypt_compact
655
656 mov 16(%rsp),$out # restore out
657 mov 24(%rsp),%rsi # restore saved stack pointer
658 .cfi_def_cfa %rsi,8
659 mov $s0,0($out) # write output vector
660 mov $s1,4($out)
661 mov $s2,8($out)
662 mov $s3,12($out)
663
664 mov -48(%rsi),%r15
665 .cfi_restore %r15
666 mov -40(%rsi),%r14
667 .cfi_restore %r14
668 mov -32(%rsi),%r13
669 .cfi_restore %r13
670 mov -24(%rsi),%r12
671 .cfi_restore %r12
672 mov -16(%rsi),%rbp
673 .cfi_restore %rbp
674 mov -8(%rsi),%rbx
675 .cfi_restore %rbx
676 lea (%rsi),%rsp
677 .cfi_def_cfa_register %rsp
678 .Lenc_epilogue:
679 ret
680 .cfi_endproc
681 .size AES_encrypt,.-AES_encrypt
682 ___
683
684 #------------------------------------------------------------------#
685
686 sub decvert()
687 { my $t3="%r8d"; # zaps $inp!
688
689 $code.=<<___;
690 # favor 3-way issue Opteron pipeline...
691 movzb `&lo("$s0")`,$acc0
692 movzb `&lo("$s1")`,$acc1
693 movzb `&lo("$s2")`,$acc2
694 mov 0($sbox,$acc0,8),$t0
695 mov 0($sbox,$acc1,8),$t1
696 mov 0($sbox,$acc2,8),$t2
697
698 movzb `&hi("$s3")`,$acc0
699 movzb `&hi("$s0")`,$acc1
700 movzb `&lo("$s3")`,$acc2
701 xor 3($sbox,$acc0,8),$t0
702 xor 3($sbox,$acc1,8),$t1
703 mov 0($sbox,$acc2,8),$t3
704
705 movzb `&hi("$s1")`,$acc0
706 shr \$16,$s0
707 movzb `&hi("$s2")`,$acc2
708 xor 3($sbox,$acc0,8),$t2
709 shr \$16,$s3
710 xor 3($sbox,$acc2,8),$t3
711
712 shr \$16,$s1
713 lea 16($key),$key
714 shr \$16,$s2
715
716 movzb `&lo("$s2")`,$acc0
717 movzb `&lo("$s3")`,$acc1
718 movzb `&lo("$s0")`,$acc2
719 xor 2($sbox,$acc0,8),$t0
720 xor 2($sbox,$acc1,8),$t1
721 xor 2($sbox,$acc2,8),$t2
722
723 movzb `&hi("$s1")`,$acc0
724 movzb `&hi("$s2")`,$acc1
725 movzb `&lo("$s1")`,$acc2
726 xor 1($sbox,$acc0,8),$t0
727 xor 1($sbox,$acc1,8),$t1
728 xor 2($sbox,$acc2,8),$t3
729
730 movzb `&hi("$s3")`,$acc0
731 mov 12($key),$s3
732 movzb `&hi("$s0")`,$acc2
733 xor 1($sbox,$acc0,8),$t2
734 mov 0($key),$s0
735 xor 1($sbox,$acc2,8),$t3
736
737 xor $t0,$s0
738 mov 4($key),$s1
739 mov 8($key),$s2
740 xor $t2,$s2
741 xor $t1,$s1
742 xor $t3,$s3
743 ___
744 }
745
746 sub declastvert()
747 { my $t3="%r8d"; # zaps $inp!
748
749 $code.=<<___;
750 lea 2048($sbox),$sbox # size optimization
751 movzb `&lo("$s0")`,$acc0
752 movzb `&lo("$s1")`,$acc1
753 movzb `&lo("$s2")`,$acc2
754 movzb ($sbox,$acc0,1),$t0
755 movzb ($sbox,$acc1,1),$t1
756 movzb ($sbox,$acc2,1),$t2
757
758 movzb `&lo("$s3")`,$acc0
759 movzb `&hi("$s3")`,$acc1
760 movzb `&hi("$s0")`,$acc2
761 movzb ($sbox,$acc0,1),$t3
762 movzb ($sbox,$acc1,1),$acc1 #$t0
763 movzb ($sbox,$acc2,1),$acc2 #$t1
764
765 shl \$8,$acc1
766 shl \$8,$acc2
767
768 xor $acc1,$t0
769 xor $acc2,$t1
770 shr \$16,$s3
771
772 movzb `&hi("$s1")`,$acc0
773 movzb `&hi("$s2")`,$acc1
774 shr \$16,$s0
775 movzb ($sbox,$acc0,1),$acc0 #$t2
776 movzb ($sbox,$acc1,1),$acc1 #$t3
777
778 shl \$8,$acc0
779 shl \$8,$acc1
780 shr \$16,$s1
781 xor $acc0,$t2
782 xor $acc1,$t3
783 shr \$16,$s2
784
785 movzb `&lo("$s2")`,$acc0
786 movzb `&lo("$s3")`,$acc1
787 movzb `&lo("$s0")`,$acc2
788 movzb ($sbox,$acc0,1),$acc0 #$t0
789 movzb ($sbox,$acc1,1),$acc1 #$t1
790 movzb ($sbox,$acc2,1),$acc2 #$t2
791
792 shl \$16,$acc0
793 shl \$16,$acc1
794 shl \$16,$acc2
795
796 xor $acc0,$t0
797 xor $acc1,$t1
798 xor $acc2,$t2
799
800 movzb `&lo("$s1")`,$acc0
801 movzb `&hi("$s1")`,$acc1
802 movzb `&hi("$s2")`,$acc2
803 movzb ($sbox,$acc0,1),$acc0 #$t3
804 movzb ($sbox,$acc1,1),$acc1 #$t0
805 movzb ($sbox,$acc2,1),$acc2 #$t1
806
807 shl \$16,$acc0
808 shl \$24,$acc1
809 shl \$24,$acc2
810
811 xor $acc0,$t3
812 xor $acc1,$t0
813 xor $acc2,$t1
814
815 movzb `&hi("$s3")`,$acc0
816 movzb `&hi("$s0")`,$acc1
817 mov 16+12($key),$s3
818 movzb ($sbox,$acc0,1),$acc0 #$t2
819 movzb ($sbox,$acc1,1),$acc1 #$t3
820 mov 16+0($key),$s0
821
822 shl \$24,$acc0
823 shl \$24,$acc1
824
825 xor $acc0,$t2
826 xor $acc1,$t3
827
828 mov 16+4($key),$s1
829 mov 16+8($key),$s2
830 lea -2048($sbox),$sbox
831 xor $t0,$s0
832 xor $t1,$s1
833 xor $t2,$s2
834 xor $t3,$s3
835 ___
836 }
837
838 sub decstep()
839 { my ($i,@s) = @_;
840 my $tmp0=$acc0;
841 my $tmp1=$acc1;
842 my $tmp2=$acc2;
843 my $out=($t0,$t1,$t2,$s[0])[$i];
844
845 $code.=" mov $s[0],$out\n" if ($i!=3);
846 $tmp1=$s[2] if ($i==3);
847 $code.=" mov $s[2],$tmp1\n" if ($i!=3);
848 $code.=" and \$0xFF,$out\n";
849
850 $code.=" mov 0($sbox,$out,8),$out\n";
851 $code.=" shr \$16,$tmp1\n";
852 $tmp2=$s[3] if ($i==3);
853 $code.=" mov $s[3],$tmp2\n" if ($i!=3);
854
855 $tmp0=$s[1] if ($i==3);
856 $code.=" movzb ".&hi($s[1]).",$tmp0\n";
857 $code.=" and \$0xFF,$tmp1\n";
858 $code.=" shr \$24,$tmp2\n";
859
860 $code.=" xor 3($sbox,$tmp0,8),$out\n";
861 $code.=" xor 2($sbox,$tmp1,8),$out\n";
862 $code.=" xor 1($sbox,$tmp2,8),$out\n";
863
864 $code.=" mov $t2,$s[1]\n" if ($i==3);
865 $code.=" mov $t1,$s[2]\n" if ($i==3);
866 $code.=" mov $t0,$s[3]\n" if ($i==3);
867 $code.="\n";
868 }
869
870 sub declast()
871 { my ($i,@s)=@_;
872 my $tmp0=$acc0;
873 my $tmp1=$acc1;
874 my $tmp2=$acc2;
875 my $out=($t0,$t1,$t2,$s[0])[$i];
876
877 $code.=" mov $s[0],$out\n" if ($i!=3);
878 $tmp1=$s[2] if ($i==3);
879 $code.=" mov $s[2],$tmp1\n" if ($i!=3);
880 $code.=" and \$0xFF,$out\n";
881
882 $code.=" movzb 2048($sbox,$out,1),$out\n";
883 $code.=" shr \$16,$tmp1\n";
884 $tmp2=$s[3] if ($i==3);
885 $code.=" mov $s[3],$tmp2\n" if ($i!=3);
886
887 $tmp0=$s[1] if ($i==3);
888 $code.=" movzb ".&hi($s[1]).",$tmp0\n";
889 $code.=" and \$0xFF,$tmp1\n";
890 $code.=" shr \$24,$tmp2\n";
891
892 $code.=" movzb 2048($sbox,$tmp0,1),$tmp0\n";
893 $code.=" movzb 2048($sbox,$tmp1,1),$tmp1\n";
894 $code.=" movzb 2048($sbox,$tmp2,1),$tmp2\n";
895
896 $code.=" shl \$8,$tmp0\n";
897 $code.=" shl \$16,$tmp1\n";
898 $code.=" shl \$24,$tmp2\n";
899
900 $code.=" xor $tmp0,$out\n";
901 $code.=" mov $t2,$s[1]\n" if ($i==3);
902 $code.=" xor $tmp1,$out\n";
903 $code.=" mov $t1,$s[2]\n" if ($i==3);
904 $code.=" xor $tmp2,$out\n";
905 $code.=" mov $t0,$s[3]\n" if ($i==3);
906 $code.="\n";
907 }
908
909 $code.=<<___;
910 .type _x86_64_AES_decrypt,\@abi-omnipotent
911 .align 16
912 _x86_64_AES_decrypt:
913 xor 0($key),$s0 # xor with key
914 xor 4($key),$s1
915 xor 8($key),$s2
916 xor 12($key),$s3
917
918 mov 240($key),$rnds # load key->rounds
919 sub \$1,$rnds
920 jmp .Ldec_loop
921 .align 16
922 .Ldec_loop:
923 ___
924 if ($verticalspin) { &decvert(); }
925 else { &decstep(0,$s0,$s3,$s2,$s1);
926 &decstep(1,$s1,$s0,$s3,$s2);
927 &decstep(2,$s2,$s1,$s0,$s3);
928 &decstep(3,$s3,$s2,$s1,$s0);
929 $code.=<<___;
930 lea 16($key),$key
931 xor 0($key),$s0 # xor with key
932 xor 4($key),$s1
933 xor 8($key),$s2
934 xor 12($key),$s3
935 ___
936 }
937 $code.=<<___;
938 sub \$1,$rnds
939 jnz .Ldec_loop
940 ___
941 if ($verticalspin) { &declastvert(); }
942 else { &declast(0,$s0,$s3,$s2,$s1);
943 &declast(1,$s1,$s0,$s3,$s2);
944 &declast(2,$s2,$s1,$s0,$s3);
945 &declast(3,$s3,$s2,$s1,$s0);
946 $code.=<<___;
947 xor 16+0($key),$s0 # xor with key
948 xor 16+4($key),$s1
949 xor 16+8($key),$s2
950 xor 16+12($key),$s3
951 ___
952 }
953 $code.=<<___;
954 .byte 0xf3,0xc3 # rep ret
955 .size _x86_64_AES_decrypt,.-_x86_64_AES_decrypt
956 ___
957
958 sub deccompactvert()
959 { my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d");
960
961 $code.=<<___;
962 movzb `&lo("$s0")`,$t0
963 movzb `&lo("$s1")`,$t1
964 movzb `&lo("$s2")`,$t2
965 movzb `&lo("$s3")`,$t3
966 movzb `&hi("$s3")`,$acc0
967 movzb `&hi("$s0")`,$acc1
968 shr \$16,$s3
969 movzb `&hi("$s1")`,$acc2
970 movzb ($sbox,$t0,1),$t0
971 movzb ($sbox,$t1,1),$t1
972 movzb ($sbox,$t2,1),$t2
973 movzb ($sbox,$t3,1),$t3
974
975 movzb ($sbox,$acc0,1),$t4 #$t0
976 movzb `&hi("$s2")`,$acc0
977 movzb ($sbox,$acc1,1),$t5 #$t1
978 movzb ($sbox,$acc2,1),$acc2 #$t2
979 movzb ($sbox,$acc0,1),$acc0 #$t3
980
981 shr \$16,$s2
982 shl \$8,$t5
983 shl \$8,$t4
984 movzb `&lo("$s2")`,$acc1
985 shr \$16,$s0
986 xor $t4,$t0
987 shr \$16,$s1
988 movzb `&lo("$s3")`,$t4
989
990 shl \$8,$acc2
991 xor $t5,$t1
992 shl \$8,$acc0
993 movzb `&lo("$s0")`,$t5
994 movzb ($sbox,$acc1,1),$acc1 #$t0
995 xor $acc2,$t2
996 movzb `&lo("$s1")`,$acc2
997
998 shl \$16,$acc1
999 xor $acc0,$t3
1000 movzb ($sbox,$t4,1),$t4 #$t1
1001 movzb `&hi("$s1")`,$acc0
1002 movzb ($sbox,$acc2,1),$acc2 #$t3
1003 xor $acc1,$t0
1004 movzb ($sbox,$t5,1),$t5 #$t2
1005 movzb `&hi("$s2")`,$acc1
1006
1007 shl \$16,$acc2
1008 shl \$16,$t4
1009 shl \$16,$t5
1010 xor $acc2,$t3
1011 movzb `&hi("$s3")`,$acc2
1012 xor $t4,$t1
1013 shr \$8,$s0
1014 xor $t5,$t2
1015
1016 movzb ($sbox,$acc0,1),$acc0 #$t0
1017 movzb ($sbox,$acc1,1),$s1 #$t1
1018 movzb ($sbox,$acc2,1),$s2 #$t2
1019 movzb ($sbox,$s0,1),$s3 #$t3
1020
1021 mov $t0,$s0
1022 shl \$24,$acc0
1023 shl \$24,$s1
1024 shl \$24,$s2
1025 xor $acc0,$s0
1026 shl \$24,$s3
1027 xor $t1,$s1
1028 xor $t2,$s2
1029 xor $t3,$s3
1030 ___
1031 }
1032
1033 # parallelized version! input is pair of 64-bit values: %rax=s1.s0
1034 # and %rcx=s3.s2, output is four 32-bit values in %eax=s0, %ebx=s1,
1035 # %ecx=s2 and %edx=s3.
1036 sub dectransform()
1037 { my ($tp10,$tp20,$tp40,$tp80,$acc0)=("%rax","%r8", "%r9", "%r10","%rbx");
1038 my ($tp18,$tp28,$tp48,$tp88,$acc8)=("%rcx","%r11","%r12","%r13","%rdx");
1039 my $prefetch = shift;
1040
1041 $code.=<<___;
1042 mov $mask80,$tp40
1043 mov $mask80,$tp48
1044 and $tp10,$tp40
1045 and $tp18,$tp48
1046 mov $tp40,$acc0
1047 mov $tp48,$acc8
1048 shr \$7,$tp40
1049 lea ($tp10,$tp10),$tp20
1050 shr \$7,$tp48
1051 lea ($tp18,$tp18),$tp28
1052 sub $tp40,$acc0
1053 sub $tp48,$acc8
1054 and $maskfe,$tp20
1055 and $maskfe,$tp28
1056 and $mask1b,$acc0
1057 and $mask1b,$acc8
1058 xor $acc0,$tp20
1059 xor $acc8,$tp28
1060 mov $mask80,$tp80
1061 mov $mask80,$tp88
1062
1063 and $tp20,$tp80
1064 and $tp28,$tp88
1065 mov $tp80,$acc0
1066 mov $tp88,$acc8
1067 shr \$7,$tp80
1068 lea ($tp20,$tp20),$tp40
1069 shr \$7,$tp88
1070 lea ($tp28,$tp28),$tp48
1071 sub $tp80,$acc0
1072 sub $tp88,$acc8
1073 and $maskfe,$tp40
1074 and $maskfe,$tp48
1075 and $mask1b,$acc0
1076 and $mask1b,$acc8
1077 xor $acc0,$tp40
1078 xor $acc8,$tp48
1079 mov $mask80,$tp80
1080 mov $mask80,$tp88
1081
1082 and $tp40,$tp80
1083 and $tp48,$tp88
1084 mov $tp80,$acc0
1085 mov $tp88,$acc8
1086 shr \$7,$tp80
1087 xor $tp10,$tp20 # tp2^=tp1
1088 shr \$7,$tp88
1089 xor $tp18,$tp28 # tp2^=tp1
1090 sub $tp80,$acc0
1091 sub $tp88,$acc8
1092 lea ($tp40,$tp40),$tp80
1093 lea ($tp48,$tp48),$tp88
1094 xor $tp10,$tp40 # tp4^=tp1
1095 xor $tp18,$tp48 # tp4^=tp1
1096 and $maskfe,$tp80
1097 and $maskfe,$tp88
1098 and $mask1b,$acc0
1099 and $mask1b,$acc8
1100 xor $acc0,$tp80
1101 xor $acc8,$tp88
1102
1103 xor $tp80,$tp10 # tp1^=tp8
1104 xor $tp88,$tp18 # tp1^=tp8
1105 xor $tp80,$tp20 # tp2^tp1^=tp8
1106 xor $tp88,$tp28 # tp2^tp1^=tp8
1107 mov $tp10,$acc0
1108 mov $tp18,$acc8
1109 xor $tp80,$tp40 # tp4^tp1^=tp8
1110 shr \$32,$acc0
1111 xor $tp88,$tp48 # tp4^tp1^=tp8
1112 shr \$32,$acc8
1113 xor $tp20,$tp80 # tp8^=tp8^tp2^tp1=tp2^tp1
1114 rol \$8,`&LO("$tp10")` # ROTATE(tp1^tp8,8)
1115 xor $tp28,$tp88 # tp8^=tp8^tp2^tp1=tp2^tp1
1116 rol \$8,`&LO("$tp18")` # ROTATE(tp1^tp8,8)
1117 xor $tp40,$tp80 # tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2
1118 rol \$8,`&LO("$acc0")` # ROTATE(tp1^tp8,8)
1119 xor $tp48,$tp88 # tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2
1120
1121 rol \$8,`&LO("$acc8")` # ROTATE(tp1^tp8,8)
1122 xor `&LO("$tp80")`,`&LO("$tp10")`
1123 shr \$32,$tp80
1124 xor `&LO("$tp88")`,`&LO("$tp18")`
1125 shr \$32,$tp88
1126 xor `&LO("$tp80")`,`&LO("$acc0")`
1127 xor `&LO("$tp88")`,`&LO("$acc8")`
1128
1129 mov $tp20,$tp80
1130 rol \$24,`&LO("$tp20")` # ROTATE(tp2^tp1^tp8,24)
1131 mov $tp28,$tp88
1132 rol \$24,`&LO("$tp28")` # ROTATE(tp2^tp1^tp8,24)
1133 shr \$32,$tp80
1134 xor `&LO("$tp20")`,`&LO("$tp10")`
1135 shr \$32,$tp88
1136 xor `&LO("$tp28")`,`&LO("$tp18")`
1137 rol \$24,`&LO("$tp80")` # ROTATE(tp2^tp1^tp8,24)
1138 mov $tp40,$tp20
1139 rol \$24,`&LO("$tp88")` # ROTATE(tp2^tp1^tp8,24)
1140 mov $tp48,$tp28
1141 shr \$32,$tp20
1142 xor `&LO("$tp80")`,`&LO("$acc0")`
1143 shr \$32,$tp28
1144 xor `&LO("$tp88")`,`&LO("$acc8")`
1145
1146 `"mov 0($sbox),$mask80" if ($prefetch)`
1147 rol \$16,`&LO("$tp40")` # ROTATE(tp4^tp1^tp8,16)
1148 `"mov 64($sbox),$maskfe" if ($prefetch)`
1149 rol \$16,`&LO("$tp48")` # ROTATE(tp4^tp1^tp8,16)
1150 `"mov 128($sbox),$mask1b" if ($prefetch)`
1151 rol \$16,`&LO("$tp20")` # ROTATE(tp4^tp1^tp8,16)
1152 `"mov 192($sbox),$tp80" if ($prefetch)`
1153 xor `&LO("$tp40")`,`&LO("$tp10")`
1154 rol \$16,`&LO("$tp28")` # ROTATE(tp4^tp1^tp8,16)
1155 xor `&LO("$tp48")`,`&LO("$tp18")`
1156 `"mov 256($sbox),$tp88" if ($prefetch)`
1157 xor `&LO("$tp20")`,`&LO("$acc0")`
1158 xor `&LO("$tp28")`,`&LO("$acc8")`
1159 ___
1160 }
1161
1162 $code.=<<___;
1163 .type _x86_64_AES_decrypt_compact,\@abi-omnipotent
1164 .align 16
1165 _x86_64_AES_decrypt_compact:
1166 .cfi_startproc
1167 lea 128($sbox),$inp # size optimization
1168 mov 0-128($inp),$acc1 # prefetch Td4
1169 mov 32-128($inp),$acc2
1170 mov 64-128($inp),$t0
1171 mov 96-128($inp),$t1
1172 mov 128-128($inp),$acc1
1173 mov 160-128($inp),$acc2
1174 mov 192-128($inp),$t0
1175 mov 224-128($inp),$t1
1176 jmp .Ldec_loop_compact
1177
1178 .align 16
1179 .Ldec_loop_compact:
1180 xor 0($key),$s0 # xor with key
1181 xor 4($key),$s1
1182 xor 8($key),$s2
1183 xor 12($key),$s3
1184 lea 16($key),$key
1185 ___
1186 &deccompactvert();
1187 $code.=<<___;
1188 cmp 16(%rsp),$key
1189 je .Ldec_compact_done
1190
1191 mov 256+0($sbox),$mask80
1192 shl \$32,%rbx
1193 shl \$32,%rdx
1194 mov 256+8($sbox),$maskfe
1195 or %rbx,%rax
1196 or %rdx,%rcx
1197 mov 256+16($sbox),$mask1b
1198 ___
1199 &dectransform(1);
1200 $code.=<<___;
1201 jmp .Ldec_loop_compact
1202 .align 16
1203 .Ldec_compact_done:
1204 xor 0($key),$s0
1205 xor 4($key),$s1
1206 xor 8($key),$s2
1207 xor 12($key),$s3
1208 .byte 0xf3,0xc3 # rep ret
1209 .cfi_endproc
1210 .size _x86_64_AES_decrypt_compact,.-_x86_64_AES_decrypt_compact
1211 ___
1212
1213 # void AES_decrypt (const void *inp,void *out,const AES_KEY *key);
1214 $code.=<<___;
1215 .globl AES_decrypt
1216 .type AES_decrypt,\@function,3
1217 .align 16
1218 .globl asm_AES_decrypt
1219 .hidden asm_AES_decrypt
1220 asm_AES_decrypt:
1221 AES_decrypt:
1222 .cfi_startproc
1223 mov %rsp,%rax
1224 .cfi_def_cfa_register %rax
1225 push %rbx
1226 .cfi_push %rbx
1227 push %rbp
1228 .cfi_push %rbp
1229 push %r12
1230 .cfi_push %r12
1231 push %r13
1232 .cfi_push %r13
1233 push %r14
1234 .cfi_push %r14
1235 push %r15
1236 .cfi_push %r15
1237
1238 # allocate frame "above" key schedule
1239 lea -63(%rdx),%rcx # %rdx is key argument
1240 and \$-64,%rsp
1241 sub %rsp,%rcx
1242 neg %rcx
1243 and \$0x3c0,%rcx
1244 sub %rcx,%rsp
1245 sub \$32,%rsp
1246
1247 mov %rsi,16(%rsp) # save out
1248 mov %rax,24(%rsp) # save original stack pointer
1249 .cfi_cfa_expression %rsp+24,deref,+8
1250 .Ldec_prologue:
1251
1252 mov %rdx,$key
1253 mov 240($key),$rnds # load rounds
1254
1255 mov 0(%rdi),$s0 # load input vector
1256 mov 4(%rdi),$s1
1257 mov 8(%rdi),$s2
1258 mov 12(%rdi),$s3
1259
1260 shl \$4,$rnds
1261 lea ($key,$rnds),%rbp
1262 mov $key,(%rsp) # key schedule
1263 mov %rbp,8(%rsp) # end of key schedule
1264
1265 # pick Td4 copy which can't "overlap" with stack frame or key schedule
1266 lea .LAES_Td+2048(%rip),$sbox
1267 lea 768(%rsp),%rbp
1268 sub $sbox,%rbp
1269 and \$0x300,%rbp
1270 lea ($sbox,%rbp),$sbox
1271 shr \$3,%rbp # recall "magic" constants!
1272 add %rbp,$sbox
1273
1274 call _x86_64_AES_decrypt_compact
1275
1276 mov 16(%rsp),$out # restore out
1277 mov 24(%rsp),%rsi # restore saved stack pointer
1278 .cfi_def_cfa %rsi,8
1279 mov $s0,0($out) # write output vector
1280 mov $s1,4($out)
1281 mov $s2,8($out)
1282 mov $s3,12($out)
1283
1284 mov -48(%rsi),%r15
1285 .cfi_restore %r15
1286 mov -40(%rsi),%r14
1287 .cfi_restore %r14
1288 mov -32(%rsi),%r13
1289 .cfi_restore %r13
1290 mov -24(%rsi),%r12
1291 .cfi_restore %r12
1292 mov -16(%rsi),%rbp
1293 .cfi_restore %rbp
1294 mov -8(%rsi),%rbx
1295 .cfi_restore %rbx
1296 lea (%rsi),%rsp
1297 .cfi_def_cfa_register %rsp
1298 .Ldec_epilogue:
1299 ret
1300 .cfi_endproc
1301 .size AES_decrypt,.-AES_decrypt
1302 ___
1303 #------------------------------------------------------------------#
1304
1305 sub enckey()
1306 {
1307 $code.=<<___;
1308 movz %dl,%esi # rk[i]>>0
1309 movzb -128(%rbp,%rsi),%ebx
1310 movz %dh,%esi # rk[i]>>8
1311 shl \$24,%ebx
1312 xor %ebx,%eax
1313
1314 movzb -128(%rbp,%rsi),%ebx
1315 shr \$16,%edx
1316 movz %dl,%esi # rk[i]>>16
1317 xor %ebx,%eax
1318
1319 movzb -128(%rbp,%rsi),%ebx
1320 movz %dh,%esi # rk[i]>>24
1321 shl \$8,%ebx
1322 xor %ebx,%eax
1323
1324 movzb -128(%rbp,%rsi),%ebx
1325 shl \$16,%ebx
1326 xor %ebx,%eax
1327
1328 xor 1024-128(%rbp,%rcx,4),%eax # rcon
1329 ___
1330 }
1331
1332 # int AES_set_encrypt_key(const unsigned char *userKey, const int bits,
1333 # AES_KEY *key)
1334 $code.=<<___;
1335 .globl AES_set_encrypt_key
1336 .type AES_set_encrypt_key,\@function,3
1337 .align 16
1338 AES_set_encrypt_key:
1339 .cfi_startproc
1340 push %rbx
1341 .cfi_push %rbx
1342 push %rbp
1343 .cfi_push %rbp
1344 push %r12 # redundant, but allows to share
1345 .cfi_push %r12
1346 push %r13 # exception handler...
1347 .cfi_push %r13
1348 push %r14
1349 .cfi_push %r14
1350 push %r15
1351 .cfi_push %r15
1352 sub \$8,%rsp
1353 .cfi_adjust_cfa_offset 8
1354 .Lenc_key_prologue:
1355
1356 call _x86_64_AES_set_encrypt_key
1357
1358 mov 40(%rsp),%rbp
1359 .cfi_restore %rbp
1360 mov 48(%rsp),%rbx
1361 .cfi_restore %rbx
1362 add \$56,%rsp
1363 .cfi_adjust_cfa_offset -56
1364 .Lenc_key_epilogue:
1365 ret
1366 .cfi_endproc
1367 .size AES_set_encrypt_key,.-AES_set_encrypt_key
1368
1369 .type _x86_64_AES_set_encrypt_key,\@abi-omnipotent
1370 .align 16
1371 _x86_64_AES_set_encrypt_key:
1372 .cfi_startproc
1373 mov %esi,%ecx # %ecx=bits
1374 mov %rdi,%rsi # %rsi=userKey
1375 mov %rdx,%rdi # %rdi=key
1376
1377 test \$-1,%rsi
1378 jz .Lbadpointer
1379 test \$-1,%rdi
1380 jz .Lbadpointer
1381
1382 lea .LAES_Te(%rip),%rbp
1383 lea 2048+128(%rbp),%rbp
1384
1385 # prefetch Te4
1386 mov 0-128(%rbp),%eax
1387 mov 32-128(%rbp),%ebx
1388 mov 64-128(%rbp),%r8d
1389 mov 96-128(%rbp),%edx
1390 mov 128-128(%rbp),%eax
1391 mov 160-128(%rbp),%ebx
1392 mov 192-128(%rbp),%r8d
1393 mov 224-128(%rbp),%edx
1394
1395 cmp \$128,%ecx
1396 je .L10rounds
1397 cmp \$192,%ecx
1398 je .L12rounds
1399 cmp \$256,%ecx
1400 je .L14rounds
1401 mov \$-2,%rax # invalid number of bits
1402 jmp .Lexit
1403
1404 .L10rounds:
1405 mov 0(%rsi),%rax # copy first 4 dwords
1406 mov 8(%rsi),%rdx
1407 mov %rax,0(%rdi)
1408 mov %rdx,8(%rdi)
1409
1410 shr \$32,%rdx
1411 xor %ecx,%ecx
1412 jmp .L10shortcut
1413 .align 4
1414 .L10loop:
1415 mov 0(%rdi),%eax # rk[0]
1416 mov 12(%rdi),%edx # rk[3]
1417 .L10shortcut:
1418 ___
1419 &enckey ();
1420 $code.=<<___;
1421 mov %eax,16(%rdi) # rk[4]
1422 xor 4(%rdi),%eax
1423 mov %eax,20(%rdi) # rk[5]
1424 xor 8(%rdi),%eax
1425 mov %eax,24(%rdi) # rk[6]
1426 xor 12(%rdi),%eax
1427 mov %eax,28(%rdi) # rk[7]
1428 add \$1,%ecx
1429 lea 16(%rdi),%rdi
1430 cmp \$10,%ecx
1431 jl .L10loop
1432
1433 movl \$10,80(%rdi) # setup number of rounds
1434 xor %rax,%rax
1435 jmp .Lexit
1436
1437 .L12rounds:
1438 mov 0(%rsi),%rax # copy first 6 dwords
1439 mov 8(%rsi),%rbx
1440 mov 16(%rsi),%rdx
1441 mov %rax,0(%rdi)
1442 mov %rbx,8(%rdi)
1443 mov %rdx,16(%rdi)
1444
1445 shr \$32,%rdx
1446 xor %ecx,%ecx
1447 jmp .L12shortcut
1448 .align 4
1449 .L12loop:
1450 mov 0(%rdi),%eax # rk[0]
1451 mov 20(%rdi),%edx # rk[5]
1452 .L12shortcut:
1453 ___
1454 &enckey ();
1455 $code.=<<___;
1456 mov %eax,24(%rdi) # rk[6]
1457 xor 4(%rdi),%eax
1458 mov %eax,28(%rdi) # rk[7]
1459 xor 8(%rdi),%eax
1460 mov %eax,32(%rdi) # rk[8]
1461 xor 12(%rdi),%eax
1462 mov %eax,36(%rdi) # rk[9]
1463
1464 cmp \$7,%ecx
1465 je .L12break
1466 add \$1,%ecx
1467
1468 xor 16(%rdi),%eax
1469 mov %eax,40(%rdi) # rk[10]
1470 xor 20(%rdi),%eax
1471 mov %eax,44(%rdi) # rk[11]
1472
1473 lea 24(%rdi),%rdi
1474 jmp .L12loop
1475 .L12break:
1476 movl \$12,72(%rdi) # setup number of rounds
1477 xor %rax,%rax
1478 jmp .Lexit
1479
1480 .L14rounds:
1481 mov 0(%rsi),%rax # copy first 8 dwords
1482 mov 8(%rsi),%rbx
1483 mov 16(%rsi),%rcx
1484 mov 24(%rsi),%rdx
1485 mov %rax,0(%rdi)
1486 mov %rbx,8(%rdi)
1487 mov %rcx,16(%rdi)
1488 mov %rdx,24(%rdi)
1489
1490 shr \$32,%rdx
1491 xor %ecx,%ecx
1492 jmp .L14shortcut
1493 .align 4
1494 .L14loop:
1495 mov 0(%rdi),%eax # rk[0]
1496 mov 28(%rdi),%edx # rk[4]
1497 .L14shortcut:
1498 ___
1499 &enckey ();
1500 $code.=<<___;
1501 mov %eax,32(%rdi) # rk[8]
1502 xor 4(%rdi),%eax
1503 mov %eax,36(%rdi) # rk[9]
1504 xor 8(%rdi),%eax
1505 mov %eax,40(%rdi) # rk[10]
1506 xor 12(%rdi),%eax
1507 mov %eax,44(%rdi) # rk[11]
1508
1509 cmp \$6,%ecx
1510 je .L14break
1511 add \$1,%ecx
1512
1513 mov %eax,%edx
1514 mov 16(%rdi),%eax # rk[4]
1515 movz %dl,%esi # rk[11]>>0
1516 movzb -128(%rbp,%rsi),%ebx
1517 movz %dh,%esi # rk[11]>>8
1518 xor %ebx,%eax
1519
1520 movzb -128(%rbp,%rsi),%ebx
1521 shr \$16,%edx
1522 shl \$8,%ebx
1523 movz %dl,%esi # rk[11]>>16
1524 xor %ebx,%eax
1525
1526 movzb -128(%rbp,%rsi),%ebx
1527 movz %dh,%esi # rk[11]>>24
1528 shl \$16,%ebx
1529 xor %ebx,%eax
1530
1531 movzb -128(%rbp,%rsi),%ebx
1532 shl \$24,%ebx
1533 xor %ebx,%eax
1534
1535 mov %eax,48(%rdi) # rk[12]
1536 xor 20(%rdi),%eax
1537 mov %eax,52(%rdi) # rk[13]
1538 xor 24(%rdi),%eax
1539 mov %eax,56(%rdi) # rk[14]
1540 xor 28(%rdi),%eax
1541 mov %eax,60(%rdi) # rk[15]
1542
1543 lea 32(%rdi),%rdi
1544 jmp .L14loop
1545 .L14break:
1546 movl \$14,48(%rdi) # setup number of rounds
1547 xor %rax,%rax
1548 jmp .Lexit
1549
1550 .Lbadpointer:
1551 mov \$-1,%rax
1552 .Lexit:
1553 .byte 0xf3,0xc3 # rep ret
1554 .cfi_endproc
1555 .size _x86_64_AES_set_encrypt_key,.-_x86_64_AES_set_encrypt_key
1556 ___
1557
1558 sub deckey_ref()
1559 { my ($i,$ptr,$te,$td) = @_;
1560 my ($tp1,$tp2,$tp4,$tp8,$acc)=("%eax","%ebx","%edi","%edx","%r8d");
1561 $code.=<<___;
1562 mov $i($ptr),$tp1
1563 mov $tp1,$acc
1564 and \$0x80808080,$acc
1565 mov $acc,$tp4
1566 shr \$7,$tp4
1567 lea 0($tp1,$tp1),$tp2
1568 sub $tp4,$acc
1569 and \$0xfefefefe,$tp2
1570 and \$0x1b1b1b1b,$acc
1571 xor $tp2,$acc
1572 mov $acc,$tp2
1573
1574 and \$0x80808080,$acc
1575 mov $acc,$tp8
1576 shr \$7,$tp8
1577 lea 0($tp2,$tp2),$tp4
1578 sub $tp8,$acc
1579 and \$0xfefefefe,$tp4
1580 and \$0x1b1b1b1b,$acc
1581 xor $tp1,$tp2 # tp2^tp1
1582 xor $tp4,$acc
1583 mov $acc,$tp4
1584
1585 and \$0x80808080,$acc
1586 mov $acc,$tp8
1587 shr \$7,$tp8
1588 sub $tp8,$acc
1589 lea 0($tp4,$tp4),$tp8
1590 xor $tp1,$tp4 # tp4^tp1
1591 and \$0xfefefefe,$tp8
1592 and \$0x1b1b1b1b,$acc
1593 xor $acc,$tp8
1594
1595 xor $tp8,$tp1 # tp1^tp8
1596 rol \$8,$tp1 # ROTATE(tp1^tp8,8)
1597 xor $tp8,$tp2 # tp2^tp1^tp8
1598 xor $tp8,$tp4 # tp4^tp1^tp8
1599 xor $tp2,$tp8
1600 xor $tp4,$tp8 # tp8^(tp8^tp4^tp1)^(tp8^tp2^tp1)=tp8^tp4^tp2
1601
1602 xor $tp8,$tp1
1603 rol \$24,$tp2 # ROTATE(tp2^tp1^tp8,24)
1604 xor $tp2,$tp1
1605 rol \$16,$tp4 # ROTATE(tp4^tp1^tp8,16)
1606 xor $tp4,$tp1
1607
1608 mov $tp1,$i($ptr)
1609 ___
1610 }
1611
1612 # int AES_set_decrypt_key(const unsigned char *userKey, const int bits,
1613 # AES_KEY *key)
1614 $code.=<<___;
1615 .globl AES_set_decrypt_key
1616 .type AES_set_decrypt_key,\@function,3
1617 .align 16
1618 AES_set_decrypt_key:
1619 .cfi_startproc
1620 push %rbx
1621 .cfi_push %rbx
1622 push %rbp
1623 .cfi_push %rbp
1624 push %r12
1625 .cfi_push %r12
1626 push %r13
1627 .cfi_push %r13
1628 push %r14
1629 .cfi_push %r14
1630 push %r15
1631 .cfi_push %r15
1632 push %rdx # save key schedule
1633 .cfi_adjust_cfa_offset 8
1634 .Ldec_key_prologue:
1635
1636 call _x86_64_AES_set_encrypt_key
1637 mov (%rsp),%r8 # restore key schedule
1638 cmp \$0,%eax
1639 jne .Labort
1640
1641 mov 240(%r8),%r14d # pull number of rounds
1642 xor %rdi,%rdi
1643 lea (%rdi,%r14d,4),%rcx
1644 mov %r8,%rsi
1645 lea (%r8,%rcx,4),%rdi # pointer to last chunk
1646 .align 4
1647 .Linvert:
1648 mov 0(%rsi),%rax
1649 mov 8(%rsi),%rbx
1650 mov 0(%rdi),%rcx
1651 mov 8(%rdi),%rdx
1652 mov %rax,0(%rdi)
1653 mov %rbx,8(%rdi)
1654 mov %rcx,0(%rsi)
1655 mov %rdx,8(%rsi)
1656 lea 16(%rsi),%rsi
1657 lea -16(%rdi),%rdi
1658 cmp %rsi,%rdi
1659 jne .Linvert
1660
1661 lea .LAES_Te+2048+1024(%rip),%rax # rcon
1662
1663 mov 40(%rax),$mask80
1664 mov 48(%rax),$maskfe
1665 mov 56(%rax),$mask1b
1666
1667 mov %r8,$key
1668 sub \$1,%r14d
1669 .align 4
1670 .Lpermute:
1671 lea 16($key),$key
1672 mov 0($key),%rax
1673 mov 8($key),%rcx
1674 ___
1675 &dectransform ();
1676 $code.=<<___;
1677 mov %eax,0($key)
1678 mov %ebx,4($key)
1679 mov %ecx,8($key)
1680 mov %edx,12($key)
1681 sub \$1,%r14d
1682 jnz .Lpermute
1683
1684 xor %rax,%rax
1685 .Labort:
1686 mov 8(%rsp),%r15
1687 .cfi_restore %r15
1688 mov 16(%rsp),%r14
1689 .cfi_restore %r14
1690 mov 24(%rsp),%r13
1691 .cfi_restore %r13
1692 mov 32(%rsp),%r12
1693 .cfi_restore %r12
1694 mov 40(%rsp),%rbp
1695 .cfi_restore %rbp
1696 mov 48(%rsp),%rbx
1697 .cfi_restore %rbx
1698 add \$56,%rsp
1699 .cfi_adjust_cfa_offset -56
1700 .Ldec_key_epilogue:
1701 ret
1702 .cfi_endproc
1703 .size AES_set_decrypt_key,.-AES_set_decrypt_key
1704 ___
1705
1706 # void AES_cbc_encrypt (const void char *inp, unsigned char *out,
1707 # size_t length, const AES_KEY *key,
1708 # unsigned char *ivp,const int enc);
1709 {
1710 # stack frame layout
1711 # -8(%rsp) return address
1712 my $keyp="0(%rsp)"; # one to pass as $key
1713 my $keyend="8(%rsp)"; # &(keyp->rd_key[4*keyp->rounds])
1714 my $_rsp="16(%rsp)"; # saved %rsp
1715 my $_inp="24(%rsp)"; # copy of 1st parameter, inp
1716 my $_out="32(%rsp)"; # copy of 2nd parameter, out
1717 my $_len="40(%rsp)"; # copy of 3rd parameter, length
1718 my $_key="48(%rsp)"; # copy of 4th parameter, key
1719 my $_ivp="56(%rsp)"; # copy of 5th parameter, ivp
1720 my $ivec="64(%rsp)"; # ivec[16]
1721 my $aes_key="80(%rsp)"; # copy of aes_key
1722 my $mark="80+240(%rsp)"; # copy of aes_key->rounds
1723
1724 $code.=<<___;
1725 .globl AES_cbc_encrypt
1726 .type AES_cbc_encrypt,\@function,6
1727 .align 16
1728 .extern OPENSSL_ia32cap_P
1729 .globl asm_AES_cbc_encrypt
1730 .hidden asm_AES_cbc_encrypt
1731 asm_AES_cbc_encrypt:
1732 AES_cbc_encrypt:
1733 .cfi_startproc
1734 cmp \$0,%rdx # check length
1735 je .Lcbc_epilogue
1736 pushfq
1737 # This could be .cfi_push 49, but libunwind fails on registers it does not
1738 # recognize. See https://bugzilla.redhat.com/show_bug.cgi?id=217087.
1739 .cfi_adjust_cfa_offset 8
1740 push %rbx
1741 .cfi_push %rbx
1742 push %rbp
1743 .cfi_push %rbp
1744 push %r12
1745 .cfi_push %r12
1746 push %r13
1747 .cfi_push %r13
1748 push %r14
1749 .cfi_push %r14
1750 push %r15
1751 .cfi_push %r15
1752 .Lcbc_prologue:
1753
1754 cld
1755 mov %r9d,%r9d # clear upper half of enc
1756
1757 lea .LAES_Te(%rip),$sbox
1758 lea .LAES_Td(%rip),%r10
1759 cmp \$0,%r9
1760 cmoveq %r10,$sbox
1761
1762 .cfi_remember_state
1763 mov OPENSSL_ia32cap_P(%rip),%r10d
1764 cmp \$$speed_limit,%rdx
1765 jb .Lcbc_slow_prologue
1766 test \$15,%rdx
1767 jnz .Lcbc_slow_prologue
1768 bt \$28,%r10d
1769 jc .Lcbc_slow_prologue
1770
1771 # allocate aligned stack frame...
1772 lea -88-248(%rsp),$key
1773 and \$-64,$key
1774
1775 # ... and make sure it doesn't alias with AES_T[ed] modulo 4096
1776 mov $sbox,%r10
1777 lea 2304($sbox),%r11
1778 mov $key,%r12
1779 and \$0xFFF,%r10 # s = $sbox&0xfff
1780 and \$0xFFF,%r11 # e = ($sbox+2048)&0xfff
1781 and \$0xFFF,%r12 # p = %rsp&0xfff
1782
1783 cmp %r11,%r12 # if (p=>e) %rsp =- (p-e);
1784 jb .Lcbc_te_break_out
1785 sub %r11,%r12
1786 sub %r12,$key
1787 jmp .Lcbc_te_ok
1788 .Lcbc_te_break_out: # else %rsp -= (p-s)&0xfff + framesz
1789 sub %r10,%r12
1790 and \$0xFFF,%r12
1791 add \$320,%r12
1792 sub %r12,$key
1793 .align 4
1794 .Lcbc_te_ok:
1795
1796 xchg %rsp,$key
1797 .cfi_def_cfa_register $key
1798 #add \$8,%rsp # reserve for return address!
1799 mov $key,$_rsp # save %rsp
1800 .cfi_cfa_expression $_rsp,deref,+64
1801 .Lcbc_fast_body:
1802 mov %rdi,$_inp # save copy of inp
1803 mov %rsi,$_out # save copy of out
1804 mov %rdx,$_len # save copy of len
1805 mov %rcx,$_key # save copy of key
1806 mov %r8,$_ivp # save copy of ivp
1807 movl \$0,$mark # copy of aes_key->rounds = 0;
1808 mov %r8,%rbp # rearrange input arguments
1809 mov %r9,%rbx
1810 mov %rsi,$out
1811 mov %rdi,$inp
1812 mov %rcx,$key
1813
1814 mov 240($key),%eax # key->rounds
1815 # do we copy key schedule to stack?
1816 mov $key,%r10
1817 sub $sbox,%r10
1818 and \$0xfff,%r10
1819 cmp \$2304,%r10
1820 jb .Lcbc_do_ecopy
1821 cmp \$4096-248,%r10
1822 jb .Lcbc_skip_ecopy
1823 .align 4
1824 .Lcbc_do_ecopy:
1825 mov $key,%rsi
1826 lea $aes_key,%rdi
1827 lea $aes_key,$key
1828 mov \$240/8,%ecx
1829 .long 0x90A548F3 # rep movsq
1830 mov %eax,(%rdi) # copy aes_key->rounds
1831 .Lcbc_skip_ecopy:
1832 mov $key,$keyp # save key pointer
1833
1834 mov \$18,%ecx
1835 .align 4
1836 .Lcbc_prefetch_te:
1837 mov 0($sbox),%r10
1838 mov 32($sbox),%r11
1839 mov 64($sbox),%r12
1840 mov 96($sbox),%r13
1841 lea 128($sbox),$sbox
1842 sub \$1,%ecx
1843 jnz .Lcbc_prefetch_te
1844 lea -2304($sbox),$sbox
1845
1846 cmp \$0,%rbx
1847 je .LFAST_DECRYPT
1848
1849 #----------------------------- ENCRYPT -----------------------------#
1850 mov 0(%rbp),$s0 # load iv
1851 mov 4(%rbp),$s1
1852 mov 8(%rbp),$s2
1853 mov 12(%rbp),$s3
1854
1855 .align 4
1856 .Lcbc_fast_enc_loop:
1857 xor 0($inp),$s0
1858 xor 4($inp),$s1
1859 xor 8($inp),$s2
1860 xor 12($inp),$s3
1861 mov $keyp,$key # restore key
1862 mov $inp,$_inp # if ($verticalspin) save inp
1863
1864 call _x86_64_AES_encrypt
1865
1866 mov $_inp,$inp # if ($verticalspin) restore inp
1867 mov $_len,%r10
1868 mov $s0,0($out)
1869 mov $s1,4($out)
1870 mov $s2,8($out)
1871 mov $s3,12($out)
1872
1873 lea 16($inp),$inp
1874 lea 16($out),$out
1875 sub \$16,%r10
1876 test \$-16,%r10
1877 mov %r10,$_len
1878 jnz .Lcbc_fast_enc_loop
1879 mov $_ivp,%rbp # restore ivp
1880 mov $s0,0(%rbp) # save ivec
1881 mov $s1,4(%rbp)
1882 mov $s2,8(%rbp)
1883 mov $s3,12(%rbp)
1884
1885 jmp .Lcbc_fast_cleanup
1886
1887 #----------------------------- DECRYPT -----------------------------#
1888 .align 16
1889 .LFAST_DECRYPT:
1890 cmp $inp,$out
1891 je .Lcbc_fast_dec_in_place
1892
1893 mov %rbp,$ivec
1894 .align 4
1895 .Lcbc_fast_dec_loop:
1896 mov 0($inp),$s0 # read input
1897 mov 4($inp),$s1
1898 mov 8($inp),$s2
1899 mov 12($inp),$s3
1900 mov $keyp,$key # restore key
1901 mov $inp,$_inp # if ($verticalspin) save inp
1902
1903 call _x86_64_AES_decrypt
1904
1905 mov $ivec,%rbp # load ivp
1906 mov $_inp,$inp # if ($verticalspin) restore inp
1907 mov $_len,%r10 # load len
1908 xor 0(%rbp),$s0 # xor iv
1909 xor 4(%rbp),$s1
1910 xor 8(%rbp),$s2
1911 xor 12(%rbp),$s3
1912 mov $inp,%rbp # current input, next iv
1913
1914 sub \$16,%r10
1915 mov %r10,$_len # update len
1916 mov %rbp,$ivec # update ivp
1917
1918 mov $s0,0($out) # write output
1919 mov $s1,4($out)
1920 mov $s2,8($out)
1921 mov $s3,12($out)
1922
1923 lea 16($inp),$inp
1924 lea 16($out),$out
1925 jnz .Lcbc_fast_dec_loop
1926 mov $_ivp,%r12 # load user ivp
1927 mov 0(%rbp),%r10 # load iv
1928 mov 8(%rbp),%r11
1929 mov %r10,0(%r12) # copy back to user
1930 mov %r11,8(%r12)
1931 jmp .Lcbc_fast_cleanup
1932
1933 .align 16
1934 .Lcbc_fast_dec_in_place:
1935 mov 0(%rbp),%r10 # copy iv to stack
1936 mov 8(%rbp),%r11
1937 mov %r10,0+$ivec
1938 mov %r11,8+$ivec
1939 .align 4
1940 .Lcbc_fast_dec_in_place_loop:
1941 mov 0($inp),$s0 # load input
1942 mov 4($inp),$s1
1943 mov 8($inp),$s2
1944 mov 12($inp),$s3
1945 mov $keyp,$key # restore key
1946 mov $inp,$_inp # if ($verticalspin) save inp
1947
1948 call _x86_64_AES_decrypt
1949
1950 mov $_inp,$inp # if ($verticalspin) restore inp
1951 mov $_len,%r10
1952 xor 0+$ivec,$s0
1953 xor 4+$ivec,$s1
1954 xor 8+$ivec,$s2
1955 xor 12+$ivec,$s3
1956
1957 mov 0($inp),%r11 # load input
1958 mov 8($inp),%r12
1959 sub \$16,%r10
1960 jz .Lcbc_fast_dec_in_place_done
1961
1962 mov %r11,0+$ivec # copy input to iv
1963 mov %r12,8+$ivec
1964
1965 mov $s0,0($out) # save output [zaps input]
1966 mov $s1,4($out)
1967 mov $s2,8($out)
1968 mov $s3,12($out)
1969
1970 lea 16($inp),$inp
1971 lea 16($out),$out
1972 mov %r10,$_len
1973 jmp .Lcbc_fast_dec_in_place_loop
1974 .Lcbc_fast_dec_in_place_done:
1975 mov $_ivp,%rdi
1976 mov %r11,0(%rdi) # copy iv back to user
1977 mov %r12,8(%rdi)
1978
1979 mov $s0,0($out) # save output [zaps input]
1980 mov $s1,4($out)
1981 mov $s2,8($out)
1982 mov $s3,12($out)
1983
1984 .align 4
1985 .Lcbc_fast_cleanup:
1986 cmpl \$0,$mark # was the key schedule copied?
1987 lea $aes_key,%rdi
1988 je .Lcbc_exit
1989 mov \$240/8,%ecx
1990 xor %rax,%rax
1991 .long 0x90AB48F3 # rep stosq
1992
1993 jmp .Lcbc_exit
1994
1995 #--------------------------- SLOW ROUTINE ---------------------------#
1996 .align 16
1997 .Lcbc_slow_prologue:
1998 .cfi_restore_state
1999 # allocate aligned stack frame...
2000 lea -88(%rsp),%rbp
2001 and \$-64,%rbp
2002 # ... just "above" key schedule
2003 lea -88-63(%rcx),%r10
2004 sub %rbp,%r10
2005 neg %r10
2006 and \$0x3c0,%r10
2007 sub %r10,%rbp
2008
2009 xchg %rsp,%rbp
2010 .cfi_def_cfa_register %rbp
2011 #add \$8,%rsp # reserve for return address!
2012 mov %rbp,$_rsp # save %rsp
2013 .cfi_cfa_expression $_rsp,deref,+64
2014 .Lcbc_slow_body:
2015 #mov %rdi,$_inp # save copy of inp
2016 #mov %rsi,$_out # save copy of out
2017 #mov %rdx,$_len # save copy of len
2018 #mov %rcx,$_key # save copy of key
2019 mov %r8,$_ivp # save copy of ivp
2020 mov %r8,%rbp # rearrange input arguments
2021 mov %r9,%rbx
2022 mov %rsi,$out
2023 mov %rdi,$inp
2024 mov %rcx,$key
2025 mov %rdx,%r10
2026
2027 mov 240($key),%eax
2028 mov $key,$keyp # save key pointer
2029 shl \$4,%eax
2030 lea ($key,%rax),%rax
2031 mov %rax,$keyend
2032
2033 # pick Te4 copy which can't "overlap" with stack frame or key schedule
2034 lea 2048($sbox),$sbox
2035 lea 768-8(%rsp),%rax
2036 sub $sbox,%rax
2037 and \$0x300,%rax
2038 lea ($sbox,%rax),$sbox
2039
2040 cmp \$0,%rbx
2041 je .LSLOW_DECRYPT
2042
2043 #--------------------------- SLOW ENCRYPT ---------------------------#
2044 test \$-16,%r10 # check upon length
2045 mov 0(%rbp),$s0 # load iv
2046 mov 4(%rbp),$s1
2047 mov 8(%rbp),$s2
2048 mov 12(%rbp),$s3
2049 jz .Lcbc_slow_enc_tail # short input...
2050
2051 .align 4
2052 .Lcbc_slow_enc_loop:
2053 xor 0($inp),$s0
2054 xor 4($inp),$s1
2055 xor 8($inp),$s2
2056 xor 12($inp),$s3
2057 mov $keyp,$key # restore key
2058 mov $inp,$_inp # save inp
2059 mov $out,$_out # save out
2060 mov %r10,$_len # save len
2061
2062 call _x86_64_AES_encrypt_compact
2063
2064 mov $_inp,$inp # restore inp
2065 mov $_out,$out # restore out
2066 mov $_len,%r10 # restore len
2067 mov $s0,0($out)
2068 mov $s1,4($out)
2069 mov $s2,8($out)
2070 mov $s3,12($out)
2071
2072 lea 16($inp),$inp
2073 lea 16($out),$out
2074 sub \$16,%r10
2075 test \$-16,%r10
2076 jnz .Lcbc_slow_enc_loop
2077 test \$15,%r10
2078 jnz .Lcbc_slow_enc_tail
2079 mov $_ivp,%rbp # restore ivp
2080 mov $s0,0(%rbp) # save ivec
2081 mov $s1,4(%rbp)
2082 mov $s2,8(%rbp)
2083 mov $s3,12(%rbp)
2084
2085 jmp .Lcbc_exit
2086
2087 .align 4
2088 .Lcbc_slow_enc_tail:
2089 mov %rax,%r11
2090 mov %rcx,%r12
2091 mov %r10,%rcx
2092 mov $inp,%rsi
2093 mov $out,%rdi
2094 .long 0x9066A4F3 # rep movsb
2095 mov \$16,%rcx # zero tail
2096 sub %r10,%rcx
2097 xor %rax,%rax
2098 .long 0x9066AAF3 # rep stosb
2099 mov $out,$inp # this is not a mistake!
2100 mov \$16,%r10 # len=16
2101 mov %r11,%rax
2102 mov %r12,%rcx
2103 jmp .Lcbc_slow_enc_loop # one more spin...
2104 #--------------------------- SLOW DECRYPT ---------------------------#
2105 .align 16
2106 .LSLOW_DECRYPT:
2107 shr \$3,%rax
2108 add %rax,$sbox # recall "magic" constants!
2109
2110 mov 0(%rbp),%r11 # copy iv to stack
2111 mov 8(%rbp),%r12
2112 mov %r11,0+$ivec
2113 mov %r12,8+$ivec
2114
2115 .align 4
2116 .Lcbc_slow_dec_loop:
2117 mov 0($inp),$s0 # load input
2118 mov 4($inp),$s1
2119 mov 8($inp),$s2
2120 mov 12($inp),$s3
2121 mov $keyp,$key # restore key
2122 mov $inp,$_inp # save inp
2123 mov $out,$_out # save out
2124 mov %r10,$_len # save len
2125
2126 call _x86_64_AES_decrypt_compact
2127
2128 mov $_inp,$inp # restore inp
2129 mov $_out,$out # restore out
2130 mov $_len,%r10
2131 xor 0+$ivec,$s0
2132 xor 4+$ivec,$s1
2133 xor 8+$ivec,$s2
2134 xor 12+$ivec,$s3
2135
2136 mov 0($inp),%r11 # load input
2137 mov 8($inp),%r12
2138 sub \$16,%r10
2139 jc .Lcbc_slow_dec_partial
2140 jz .Lcbc_slow_dec_done
2141
2142 mov %r11,0+$ivec # copy input to iv
2143 mov %r12,8+$ivec
2144
2145 mov $s0,0($out) # save output [can zap input]
2146 mov $s1,4($out)
2147 mov $s2,8($out)
2148 mov $s3,12($out)
2149
2150 lea 16($inp),$inp
2151 lea 16($out),$out
2152 jmp .Lcbc_slow_dec_loop
2153 .Lcbc_slow_dec_done:
2154 mov $_ivp,%rdi
2155 mov %r11,0(%rdi) # copy iv back to user
2156 mov %r12,8(%rdi)
2157
2158 mov $s0,0($out) # save output [can zap input]
2159 mov $s1,4($out)
2160 mov $s2,8($out)
2161 mov $s3,12($out)
2162
2163 jmp .Lcbc_exit
2164
2165 .align 4
2166 .Lcbc_slow_dec_partial:
2167 mov $_ivp,%rdi
2168 mov %r11,0(%rdi) # copy iv back to user
2169 mov %r12,8(%rdi)
2170
2171 mov $s0,0+$ivec # save output to stack
2172 mov $s1,4+$ivec
2173 mov $s2,8+$ivec
2174 mov $s3,12+$ivec
2175
2176 mov $out,%rdi
2177 lea $ivec,%rsi
2178 lea 16(%r10),%rcx
2179 .long 0x9066A4F3 # rep movsb
2180 jmp .Lcbc_exit
2181
2182 .align 16
2183 .Lcbc_exit:
2184 mov $_rsp,%rsi
2185 .cfi_def_cfa %rsi,64
2186 mov (%rsi),%r15
2187 .cfi_restore %r15
2188 mov 8(%rsi),%r14
2189 .cfi_restore %r14
2190 mov 16(%rsi),%r13
2191 .cfi_restore %r13
2192 mov 24(%rsi),%r12
2193 .cfi_restore %r12
2194 mov 32(%rsi),%rbp
2195 .cfi_restore %rbp
2196 mov 40(%rsi),%rbx
2197 .cfi_restore %rbx
2198 lea 48(%rsi),%rsp
2199 .cfi_def_cfa %rsp,16
2200 .Lcbc_popfq:
2201 popfq
2202 # This could be .cfi_pop 49, but libunwind fails on registers it does not
2203 # recognize. See https://bugzilla.redhat.com/show_bug.cgi?id=217087.
2204 .cfi_adjust_cfa_offset -8
2205 .Lcbc_epilogue:
2206 ret
2207 .cfi_endproc
2208 .size AES_cbc_encrypt,.-AES_cbc_encrypt
2209 ___
2210 }
2211
2212 $code.=<<___;
2213 .align 64
2214 .LAES_Te:
2215 ___
2216 &_data_word(0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6);
2217 &_data_word(0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591);
2218 &_data_word(0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56);
2219 &_data_word(0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec);
2220 &_data_word(0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa);
2221 &_data_word(0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb);
2222 &_data_word(0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45);
2223 &_data_word(0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b);
2224 &_data_word(0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c);
2225 &_data_word(0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83);
2226 &_data_word(0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9);
2227 &_data_word(0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a);
2228 &_data_word(0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d);
2229 &_data_word(0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f);
2230 &_data_word(0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df);
2231 &_data_word(0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea);
2232 &_data_word(0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34);
2233 &_data_word(0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b);
2234 &_data_word(0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d);
2235 &_data_word(0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413);
2236 &_data_word(0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1);
2237 &_data_word(0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6);
2238 &_data_word(0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972);
2239 &_data_word(0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85);
2240 &_data_word(0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed);
2241 &_data_word(0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511);
2242 &_data_word(0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe);
2243 &_data_word(0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b);
2244 &_data_word(0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05);
2245 &_data_word(0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1);
2246 &_data_word(0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142);
2247 &_data_word(0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf);
2248 &_data_word(0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3);
2249 &_data_word(0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e);
2250 &_data_word(0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a);
2251 &_data_word(0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6);
2252 &_data_word(0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3);
2253 &_data_word(0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b);
2254 &_data_word(0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428);
2255 &_data_word(0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad);
2256 &_data_word(0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14);
2257 &_data_word(0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8);
2258 &_data_word(0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4);
2259 &_data_word(0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2);
2260 &_data_word(0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda);
2261 &_data_word(0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949);
2262 &_data_word(0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf);
2263 &_data_word(0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810);
2264 &_data_word(0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c);
2265 &_data_word(0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697);
2266 &_data_word(0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e);
2267 &_data_word(0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f);
2268 &_data_word(0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc);
2269 &_data_word(0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c);
2270 &_data_word(0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969);
2271 &_data_word(0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27);
2272 &_data_word(0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122);
2273 &_data_word(0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433);
2274 &_data_word(0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9);
2275 &_data_word(0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5);
2276 &_data_word(0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a);
2277 &_data_word(0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0);
2278 &_data_word(0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e);
2279 &_data_word(0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c);
2280
2281 #Te4 # four copies of Te4 to choose from to avoid L1 aliasing
2282 &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2283 &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2284 &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2285 &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2286 &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2287 &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2288 &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2289 &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2290 &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2291 &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2292 &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2293 &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2294 &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2295 &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2296 &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2297 &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2298 &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2299 &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2300 &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2301 &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2302 &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2303 &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2304 &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2305 &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2306 &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2307 &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2308 &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2309 &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2310 &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2311 &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2312 &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2313 &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2314
2315 &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2316 &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2317 &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2318 &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2319 &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2320 &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2321 &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2322 &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2323 &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2324 &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2325 &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2326 &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2327 &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2328 &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2329 &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2330 &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2331 &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2332 &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2333 &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2334 &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2335 &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2336 &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2337 &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2338 &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2339 &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2340 &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2341 &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2342 &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2343 &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2344 &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2345 &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2346 &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2347
2348 &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2349 &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2350 &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2351 &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2352 &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2353 &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2354 &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2355 &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2356 &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2357 &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2358 &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2359 &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2360 &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2361 &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2362 &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2363 &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2364 &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2365 &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2366 &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2367 &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2368 &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2369 &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2370 &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2371 &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2372 &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2373 &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2374 &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2375 &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2376 &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2377 &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2378 &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2379 &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2380
2381 &data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
2382 &data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
2383 &data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
2384 &data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
2385 &data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
2386 &data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
2387 &data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
2388 &data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
2389 &data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
2390 &data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
2391 &data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
2392 &data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
2393 &data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
2394 &data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
2395 &data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
2396 &data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
2397 &data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
2398 &data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
2399 &data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
2400 &data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
2401 &data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
2402 &data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
2403 &data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
2404 &data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
2405 &data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
2406 &data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
2407 &data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
2408 &data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
2409 &data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
2410 &data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
2411 &data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
2412 &data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
2413 #rcon:
2414 $code.=<<___;
2415 .long 0x00000001, 0x00000002, 0x00000004, 0x00000008
2416 .long 0x00000010, 0x00000020, 0x00000040, 0x00000080
2417 .long 0x0000001b, 0x00000036, 0x80808080, 0x80808080
2418 .long 0xfefefefe, 0xfefefefe, 0x1b1b1b1b, 0x1b1b1b1b
2419 ___
2420 $code.=<<___;
2421 .align 64
2422 .LAES_Td:
2423 ___
2424 &_data_word(0x50a7f451, 0x5365417e, 0xc3a4171a, 0x965e273a);
2425 &_data_word(0xcb6bab3b, 0xf1459d1f, 0xab58faac, 0x9303e34b);
2426 &_data_word(0x55fa3020, 0xf66d76ad, 0x9176cc88, 0x254c02f5);
2427 &_data_word(0xfcd7e54f, 0xd7cb2ac5, 0x80443526, 0x8fa362b5);
2428 &_data_word(0x495ab1de, 0x671bba25, 0x980eea45, 0xe1c0fe5d);
2429 &_data_word(0x02752fc3, 0x12f04c81, 0xa397468d, 0xc6f9d36b);
2430 &_data_word(0xe75f8f03, 0x959c9215, 0xeb7a6dbf, 0xda595295);
2431 &_data_word(0x2d83bed4, 0xd3217458, 0x2969e049, 0x44c8c98e);
2432 &_data_word(0x6a89c275, 0x78798ef4, 0x6b3e5899, 0xdd71b927);
2433 &_data_word(0xb64fe1be, 0x17ad88f0, 0x66ac20c9, 0xb43ace7d);
2434 &_data_word(0x184adf63, 0x82311ae5, 0x60335197, 0x457f5362);
2435 &_data_word(0xe07764b1, 0x84ae6bbb, 0x1ca081fe, 0x942b08f9);
2436 &_data_word(0x58684870, 0x19fd458f, 0x876cde94, 0xb7f87b52);
2437 &_data_word(0x23d373ab, 0xe2024b72, 0x578f1fe3, 0x2aab5566);
2438 &_data_word(0x0728ebb2, 0x03c2b52f, 0x9a7bc586, 0xa50837d3);
2439 &_data_word(0xf2872830, 0xb2a5bf23, 0xba6a0302, 0x5c8216ed);
2440 &_data_word(0x2b1ccf8a, 0x92b479a7, 0xf0f207f3, 0xa1e2694e);
2441 &_data_word(0xcdf4da65, 0xd5be0506, 0x1f6234d1, 0x8afea6c4);
2442 &_data_word(0x9d532e34, 0xa055f3a2, 0x32e18a05, 0x75ebf6a4);
2443 &_data_word(0x39ec830b, 0xaaef6040, 0x069f715e, 0x51106ebd);
2444 &_data_word(0xf98a213e, 0x3d06dd96, 0xae053edd, 0x46bde64d);
2445 &_data_word(0xb58d5491, 0x055dc471, 0x6fd40604, 0xff155060);
2446 &_data_word(0x24fb9819, 0x97e9bdd6, 0xcc434089, 0x779ed967);
2447 &_data_word(0xbd42e8b0, 0x888b8907, 0x385b19e7, 0xdbeec879);
2448 &_data_word(0x470a7ca1, 0xe90f427c, 0xc91e84f8, 0x00000000);
2449 &_data_word(0x83868009, 0x48ed2b32, 0xac70111e, 0x4e725a6c);
2450 &_data_word(0xfbff0efd, 0x5638850f, 0x1ed5ae3d, 0x27392d36);
2451 &_data_word(0x64d90f0a, 0x21a65c68, 0xd1545b9b, 0x3a2e3624);
2452 &_data_word(0xb1670a0c, 0x0fe75793, 0xd296eeb4, 0x9e919b1b);
2453 &_data_word(0x4fc5c080, 0xa220dc61, 0x694b775a, 0x161a121c);
2454 &_data_word(0x0aba93e2, 0xe52aa0c0, 0x43e0223c, 0x1d171b12);
2455 &_data_word(0x0b0d090e, 0xadc78bf2, 0xb9a8b62d, 0xc8a91e14);
2456 &_data_word(0x8519f157, 0x4c0775af, 0xbbdd99ee, 0xfd607fa3);
2457 &_data_word(0x9f2601f7, 0xbcf5725c, 0xc53b6644, 0x347efb5b);
2458 &_data_word(0x7629438b, 0xdcc623cb, 0x68fcedb6, 0x63f1e4b8);
2459 &_data_word(0xcadc31d7, 0x10856342, 0x40229713, 0x2011c684);
2460 &_data_word(0x7d244a85, 0xf83dbbd2, 0x1132f9ae, 0x6da129c7);
2461 &_data_word(0x4b2f9e1d, 0xf330b2dc, 0xec52860d, 0xd0e3c177);
2462 &_data_word(0x6c16b32b, 0x99b970a9, 0xfa489411, 0x2264e947);
2463 &_data_word(0xc48cfca8, 0x1a3ff0a0, 0xd82c7d56, 0xef903322);
2464 &_data_word(0xc74e4987, 0xc1d138d9, 0xfea2ca8c, 0x360bd498);
2465 &_data_word(0xcf81f5a6, 0x28de7aa5, 0x268eb7da, 0xa4bfad3f);
2466 &_data_word(0xe49d3a2c, 0x0d927850, 0x9bcc5f6a, 0x62467e54);
2467 &_data_word(0xc2138df6, 0xe8b8d890, 0x5ef7392e, 0xf5afc382);
2468 &_data_word(0xbe805d9f, 0x7c93d069, 0xa92dd56f, 0xb31225cf);
2469 &_data_word(0x3b99acc8, 0xa77d1810, 0x6e639ce8, 0x7bbb3bdb);
2470 &_data_word(0x097826cd, 0xf418596e, 0x01b79aec, 0xa89a4f83);
2471 &_data_word(0x656e95e6, 0x7ee6ffaa, 0x08cfbc21, 0xe6e815ef);
2472 &_data_word(0xd99be7ba, 0xce366f4a, 0xd4099fea, 0xd67cb029);
2473 &_data_word(0xafb2a431, 0x31233f2a, 0x3094a5c6, 0xc066a235);
2474 &_data_word(0x37bc4e74, 0xa6ca82fc, 0xb0d090e0, 0x15d8a733);
2475 &_data_word(0x4a9804f1, 0xf7daec41, 0x0e50cd7f, 0x2ff69117);
2476 &_data_word(0x8dd64d76, 0x4db0ef43, 0x544daacc, 0xdf0496e4);
2477 &_data_word(0xe3b5d19e, 0x1b886a4c, 0xb81f2cc1, 0x7f516546);
2478 &_data_word(0x04ea5e9d, 0x5d358c01, 0x737487fa, 0x2e410bfb);
2479 &_data_word(0x5a1d67b3, 0x52d2db92, 0x335610e9, 0x1347d66d);
2480 &_data_word(0x8c61d79a, 0x7a0ca137, 0x8e14f859, 0x893c13eb);
2481 &_data_word(0xee27a9ce, 0x35c961b7, 0xede51ce1, 0x3cb1477a);
2482 &_data_word(0x59dfd29c, 0x3f73f255, 0x79ce1418, 0xbf37c773);
2483 &_data_word(0xeacdf753, 0x5baafd5f, 0x146f3ddf, 0x86db4478);
2484 &_data_word(0x81f3afca, 0x3ec468b9, 0x2c342438, 0x5f40a3c2);
2485 &_data_word(0x72c31d16, 0x0c25e2bc, 0x8b493c28, 0x41950dff);
2486 &_data_word(0x7101a839, 0xdeb30c08, 0x9ce4b4d8, 0x90c15664);
2487 &_data_word(0x6184cb7b, 0x70b632d5, 0x745c6c48, 0x4257b8d0);
2488
2489 #Td4: # four copies of Td4 to choose from to avoid L1 aliasing
2490 &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2491 &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2492 &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2493 &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2494 &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2495 &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2496 &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2497 &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2498 &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2499 &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2500 &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2501 &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2502 &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2503 &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2504 &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2505 &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2506 &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2507 &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2508 &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2509 &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2510 &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2511 &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2512 &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2513 &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2514 &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2515 &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2516 &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2517 &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2518 &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2519 &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2520 &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2521 &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2522 $code.=<<___;
2523 .long 0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2524 .long 0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2525 ___
2526 &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2527 &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2528 &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2529 &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2530 &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2531 &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2532 &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2533 &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2534 &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2535 &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2536 &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2537 &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2538 &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2539 &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2540 &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2541 &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2542 &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2543 &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2544 &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2545 &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2546 &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2547 &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2548 &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2549 &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2550 &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2551 &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2552 &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2553 &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2554 &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2555 &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2556 &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2557 &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2558 $code.=<<___;
2559 .long 0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2560 .long 0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2561 ___
2562 &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2563 &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2564 &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2565 &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2566 &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2567 &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2568 &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2569 &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2570 &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2571 &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2572 &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2573 &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2574 &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2575 &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2576 &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2577 &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2578 &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2579 &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2580 &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2581 &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2582 &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2583 &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2584 &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2585 &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2586 &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2587 &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2588 &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2589 &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2590 &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2591 &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2592 &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2593 &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2594 $code.=<<___;
2595 .long 0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2596 .long 0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2597 ___
2598 &data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
2599 &data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
2600 &data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
2601 &data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
2602 &data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
2603 &data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
2604 &data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
2605 &data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
2606 &data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
2607 &data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
2608 &data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
2609 &data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
2610 &data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
2611 &data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
2612 &data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
2613 &data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
2614 &data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
2615 &data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
2616 &data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
2617 &data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
2618 &data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
2619 &data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
2620 &data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
2621 &data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
2622 &data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
2623 &data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
2624 &data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
2625 &data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
2626 &data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
2627 &data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
2628 &data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
2629 &data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
2630 $code.=<<___;
2631 .long 0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
2632 .long 0x1b1b1b1b, 0x1b1b1b1b, 0, 0
2633 .asciz "AES for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
2634 .align 64
2635 ___
2636
2637 # EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
2638 # CONTEXT *context,DISPATCHER_CONTEXT *disp)
2639 if ($win64) {
2640 $rec="%rcx";
2641 $frame="%rdx";
2642 $context="%r8";
2643 $disp="%r9";
2644
2645 $code.=<<___;
2646 .extern __imp_RtlVirtualUnwind
2647 .type block_se_handler,\@abi-omnipotent
2648 .align 16
2649 block_se_handler:
2650 push %rsi
2651 push %rdi
2652 push %rbx
2653 push %rbp
2654 push %r12
2655 push %r13
2656 push %r14
2657 push %r15
2658 pushfq
2659 sub \$64,%rsp
2660
2661 mov 120($context),%rax # pull context->Rax
2662 mov 248($context),%rbx # pull context->Rip
2663
2664 mov 8($disp),%rsi # disp->ImageBase
2665 mov 56($disp),%r11 # disp->HandlerData
2666
2667 mov 0(%r11),%r10d # HandlerData[0]
2668 lea (%rsi,%r10),%r10 # prologue label
2669 cmp %r10,%rbx # context->Rip<prologue label
2670 jb .Lin_block_prologue
2671
2672 mov 152($context),%rax # pull context->Rsp
2673
2674 mov 4(%r11),%r10d # HandlerData[1]
2675 lea (%rsi,%r10),%r10 # epilogue label
2676 cmp %r10,%rbx # context->Rip>=epilogue label
2677 jae .Lin_block_prologue
2678
2679 mov 24(%rax),%rax # pull saved real stack pointer
2680
2681 mov -8(%rax),%rbx
2682 mov -16(%rax),%rbp
2683 mov -24(%rax),%r12
2684 mov -32(%rax),%r13
2685 mov -40(%rax),%r14
2686 mov -48(%rax),%r15
2687 mov %rbx,144($context) # restore context->Rbx
2688 mov %rbp,160($context) # restore context->Rbp
2689 mov %r12,216($context) # restore context->R12
2690 mov %r13,224($context) # restore context->R13
2691 mov %r14,232($context) # restore context->R14
2692 mov %r15,240($context) # restore context->R15
2693
2694 .Lin_block_prologue:
2695 mov 8(%rax),%rdi
2696 mov 16(%rax),%rsi
2697 mov %rax,152($context) # restore context->Rsp
2698 mov %rsi,168($context) # restore context->Rsi
2699 mov %rdi,176($context) # restore context->Rdi
2700
2701 jmp .Lcommon_seh_exit
2702 .size block_se_handler,.-block_se_handler
2703
2704 .type key_se_handler,\@abi-omnipotent
2705 .align 16
2706 key_se_handler:
2707 push %rsi
2708 push %rdi
2709 push %rbx
2710 push %rbp
2711 push %r12
2712 push %r13
2713 push %r14
2714 push %r15
2715 pushfq
2716 sub \$64,%rsp
2717
2718 mov 120($context),%rax # pull context->Rax
2719 mov 248($context),%rbx # pull context->Rip
2720
2721 mov 8($disp),%rsi # disp->ImageBase
2722 mov 56($disp),%r11 # disp->HandlerData
2723
2724 mov 0(%r11),%r10d # HandlerData[0]
2725 lea (%rsi,%r10),%r10 # prologue label
2726 cmp %r10,%rbx # context->Rip<prologue label
2727 jb .Lin_key_prologue
2728
2729 mov 152($context),%rax # pull context->Rsp
2730
2731 mov 4(%r11),%r10d # HandlerData[1]
2732 lea (%rsi,%r10),%r10 # epilogue label
2733 cmp %r10,%rbx # context->Rip>=epilogue label
2734 jae .Lin_key_prologue
2735
2736 lea 56(%rax),%rax
2737
2738 mov -8(%rax),%rbx
2739 mov -16(%rax),%rbp
2740 mov -24(%rax),%r12
2741 mov -32(%rax),%r13
2742 mov -40(%rax),%r14
2743 mov -48(%rax),%r15
2744 mov %rbx,144($context) # restore context->Rbx
2745 mov %rbp,160($context) # restore context->Rbp
2746 mov %r12,216($context) # restore context->R12
2747 mov %r13,224($context) # restore context->R13
2748 mov %r14,232($context) # restore context->R14
2749 mov %r15,240($context) # restore context->R15
2750
2751 .Lin_key_prologue:
2752 mov 8(%rax),%rdi
2753 mov 16(%rax),%rsi
2754 mov %rax,152($context) # restore context->Rsp
2755 mov %rsi,168($context) # restore context->Rsi
2756 mov %rdi,176($context) # restore context->Rdi
2757
2758 jmp .Lcommon_seh_exit
2759 .size key_se_handler,.-key_se_handler
2760
2761 .type cbc_se_handler,\@abi-omnipotent
2762 .align 16
2763 cbc_se_handler:
2764 push %rsi
2765 push %rdi
2766 push %rbx
2767 push %rbp
2768 push %r12
2769 push %r13
2770 push %r14
2771 push %r15
2772 pushfq
2773 sub \$64,%rsp
2774
2775 mov 120($context),%rax # pull context->Rax
2776 mov 248($context),%rbx # pull context->Rip
2777
2778 lea .Lcbc_prologue(%rip),%r10
2779 cmp %r10,%rbx # context->Rip<.Lcbc_prologue
2780 jb .Lin_cbc_prologue
2781
2782 lea .Lcbc_fast_body(%rip),%r10
2783 cmp %r10,%rbx # context->Rip<.Lcbc_fast_body
2784 jb .Lin_cbc_frame_setup
2785
2786 lea .Lcbc_slow_prologue(%rip),%r10
2787 cmp %r10,%rbx # context->Rip<.Lcbc_slow_prologue
2788 jb .Lin_cbc_body
2789
2790 lea .Lcbc_slow_body(%rip),%r10
2791 cmp %r10,%rbx # context->Rip<.Lcbc_slow_body
2792 jb .Lin_cbc_frame_setup
2793
2794 .Lin_cbc_body:
2795 mov 152($context),%rax # pull context->Rsp
2796
2797 lea .Lcbc_epilogue(%rip),%r10
2798 cmp %r10,%rbx # context->Rip>=.Lcbc_epilogue
2799 jae .Lin_cbc_prologue
2800
2801 lea 8(%rax),%rax
2802
2803 lea .Lcbc_popfq(%rip),%r10
2804 cmp %r10,%rbx # context->Rip>=.Lcbc_popfq
2805 jae .Lin_cbc_prologue
2806
2807 mov `16-8`(%rax),%rax # biased $_rsp
2808 lea 56(%rax),%rax
2809
2810 .Lin_cbc_frame_setup:
2811 mov -16(%rax),%rbx
2812 mov -24(%rax),%rbp
2813 mov -32(%rax),%r12
2814 mov -40(%rax),%r13
2815 mov -48(%rax),%r14
2816 mov -56(%rax),%r15
2817 mov %rbx,144($context) # restore context->Rbx
2818 mov %rbp,160($context) # restore context->Rbp
2819 mov %r12,216($context) # restore context->R12
2820 mov %r13,224($context) # restore context->R13
2821 mov %r14,232($context) # restore context->R14
2822 mov %r15,240($context) # restore context->R15
2823
2824 .Lin_cbc_prologue:
2825 mov 8(%rax),%rdi
2826 mov 16(%rax),%rsi
2827 mov %rax,152($context) # restore context->Rsp
2828 mov %rsi,168($context) # restore context->Rsi
2829 mov %rdi,176($context) # restore context->Rdi
2830
2831 .Lcommon_seh_exit:
2832
2833 mov 40($disp),%rdi # disp->ContextRecord
2834 mov $context,%rsi # context
2835 mov \$`1232/8`,%ecx # sizeof(CONTEXT)
2836 .long 0xa548f3fc # cld; rep movsq
2837
2838 mov $disp,%rsi
2839 xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
2840 mov 8(%rsi),%rdx # arg2, disp->ImageBase
2841 mov 0(%rsi),%r8 # arg3, disp->ControlPc
2842 mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
2843 mov 40(%rsi),%r10 # disp->ContextRecord
2844 lea 56(%rsi),%r11 # &disp->HandlerData
2845 lea 24(%rsi),%r12 # &disp->EstablisherFrame
2846 mov %r10,32(%rsp) # arg5
2847 mov %r11,40(%rsp) # arg6
2848 mov %r12,48(%rsp) # arg7
2849 mov %rcx,56(%rsp) # arg8, (NULL)
2850 call *__imp_RtlVirtualUnwind(%rip)
2851
2852 mov \$1,%eax # ExceptionContinueSearch
2853 add \$64,%rsp
2854 popfq
2855 pop %r15
2856 pop %r14
2857 pop %r13
2858 pop %r12
2859 pop %rbp
2860 pop %rbx
2861 pop %rdi
2862 pop %rsi
2863 ret
2864 .size cbc_se_handler,.-cbc_se_handler
2865
2866 .section .pdata
2867 .align 4
2868 .rva .LSEH_begin_AES_encrypt
2869 .rva .LSEH_end_AES_encrypt
2870 .rva .LSEH_info_AES_encrypt
2871
2872 .rva .LSEH_begin_AES_decrypt
2873 .rva .LSEH_end_AES_decrypt
2874 .rva .LSEH_info_AES_decrypt
2875
2876 .rva .LSEH_begin_AES_set_encrypt_key
2877 .rva .LSEH_end_AES_set_encrypt_key
2878 .rva .LSEH_info_AES_set_encrypt_key
2879
2880 .rva .LSEH_begin_AES_set_decrypt_key
2881 .rva .LSEH_end_AES_set_decrypt_key
2882 .rva .LSEH_info_AES_set_decrypt_key
2883
2884 .rva .LSEH_begin_AES_cbc_encrypt
2885 .rva .LSEH_end_AES_cbc_encrypt
2886 .rva .LSEH_info_AES_cbc_encrypt
2887
2888 .section .xdata
2889 .align 8
2890 .LSEH_info_AES_encrypt:
2891 .byte 9,0,0,0
2892 .rva block_se_handler
2893 .rva .Lenc_prologue,.Lenc_epilogue # HandlerData[]
2894 .LSEH_info_AES_decrypt:
2895 .byte 9,0,0,0
2896 .rva block_se_handler
2897 .rva .Ldec_prologue,.Ldec_epilogue # HandlerData[]
2898 .LSEH_info_AES_set_encrypt_key:
2899 .byte 9,0,0,0
2900 .rva key_se_handler
2901 .rva .Lenc_key_prologue,.Lenc_key_epilogue # HandlerData[]
2902 .LSEH_info_AES_set_decrypt_key:
2903 .byte 9,0,0,0
2904 .rva key_se_handler
2905 .rva .Ldec_key_prologue,.Ldec_key_epilogue # HandlerData[]
2906 .LSEH_info_AES_cbc_encrypt:
2907 .byte 9,0,0,0
2908 .rva cbc_se_handler
2909 ___
2910 }
2911
2912 $code =~ s/\`([^\`]*)\`/eval($1)/gem;
2913
2914 print $code;
2915
2916 close STDOUT;
A mirror of the official openssl repository