1#!/usr/bin/env perl 2 3# ==================================================================== 4# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL 5# project. The module is, however, dual licensed under OpenSSL and 6# CRYPTOGAMS licenses depending on where you obtain it. For further 7# details see http://www.openssl.org/~appro/cryptogams/. 8# ==================================================================== 9 10# I let hardware handle unaligned input, except on page boundaries 11# (see below for details). Otherwise straightforward implementation 12# with X vector in register bank. The module is big-endian [which is 13# not big deal as there're no little-endian targets left around]. 14 15# sha256 | sha512 16# -m64 -m32 | -m64 -m32 17# --------------------------------------+----------------------- 18# PPC970,gcc-4.0.0 +50% +38% | +40% +410%(*) 19# Power6,xlc-7 +150% +90% | +100% +430%(*) 20# 21# (*) 64-bit code in 32-bit application context, which actually is 22# on TODO list. It should be noted that for safe deployment in 23# 32-bit *mutli-threaded* context asyncronous signals should be 24# blocked upon entry to SHA512 block routine. This is because 25# 32-bit signaling procedure invalidates upper halves of GPRs. 26# Context switch procedure preserves them, but not signaling:-( 27 28# Second version is true multi-thread safe. Trouble with the original 29# version was that it was using thread local storage pointer register. 30# Well, it scrupulously preserved it, but the problem would arise the 31# moment asynchronous signal was delivered and signal handler would 32# dereference the TLS pointer. While it's never the case in openssl 33# application or test suite, we have to respect this scenario and not 34# use TLS pointer register. Alternative would be to require caller to 35# block signals prior calling this routine. For the record, in 32-bit 36# context R2 serves as TLS pointer, while in 64-bit context - R13. 37 38$flavour=shift; 39$output =shift; 40 41if ($flavour =~ /64/) { 42 $SIZE_T=8; 43 $LRSAVE=2*$SIZE_T; 44 $STU="stdu"; 45 $UCMP="cmpld"; 46 $SHL="sldi"; 47 $POP="ld"; 48 $PUSH="std"; 49} elsif ($flavour =~ /32/) { 50 $SIZE_T=4; 51 $LRSAVE=$SIZE_T; 52 $STU="stwu"; 53 $UCMP="cmplw"; 54 $SHL="slwi"; 55 $POP="lwz"; 56 $PUSH="stw"; 57} else { die "nonsense $flavour"; } 58 59$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; 60( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or 61( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or 62die "can't locate ppc-xlate.pl"; 63 64open STDOUT,"| $^X $xlate $flavour $output" || die "can't call $xlate: $!"; 65 66if ($output =~ /512/) { 67 $func="sha512_block_data_order"; 68 $SZ=8; 69 @Sigma0=(28,34,39); 70 @Sigma1=(14,18,41); 71 @sigma0=(1, 8, 7); 72 @sigma1=(19,61, 6); 73 $rounds=80; 74 $LD="ld"; 75 $ST="std"; 76 $ROR="rotrdi"; 77 $SHR="srdi"; 78} else { 79 $func="sha256_block_data_order"; 80 $SZ=4; 81 @Sigma0=( 2,13,22); 82 @Sigma1=( 6,11,25); 83 @sigma0=( 7,18, 3); 84 @sigma1=(17,19,10); 85 $rounds=64; 86 $LD="lwz"; 87 $ST="stw"; 88 $ROR="rotrwi"; 89 $SHR="srwi"; 90} 91 92$FRAME=32*$SIZE_T+16*$SZ; 93$LOCALS=6*$SIZE_T; 94 95$sp ="r1"; 96$toc="r2"; 97$ctx="r3"; # zapped by $a0 98$inp="r4"; # zapped by $a1 99$num="r5"; # zapped by $t0 100 101$T ="r0"; 102$a0 ="r3"; 103$a1 ="r4"; 104$t0 ="r5"; 105$t1 ="r6"; 106$Tbl="r7"; 107 108$A ="r8"; 109$B ="r9"; 110$C ="r10"; 111$D ="r11"; 112$E ="r12"; 113$F ="r13"; $F="r2" if ($SIZE_T==8);# reassigned to exempt TLS pointer 114$G ="r14"; 115$H ="r15"; 116 117@V=($A,$B,$C,$D,$E,$F,$G,$H); 118@X=("r16","r17","r18","r19","r20","r21","r22","r23", 119 "r24","r25","r26","r27","r28","r29","r30","r31"); 120 121$inp="r31"; # reassigned $inp! aliases with @X[15] 122 123sub ROUND_00_15 { 124my ($i,$a,$b,$c,$d,$e,$f,$g,$h)=@_; 125$code.=<<___; 126 $LD $T,`$i*$SZ`($Tbl) 127 $ROR $a0,$e,$Sigma1[0] 128 $ROR $a1,$e,$Sigma1[1] 129 and $t0,$f,$e 130 andc $t1,$g,$e 131 add $T,$T,$h 132 xor $a0,$a0,$a1 133 $ROR $a1,$a1,`$Sigma1[2]-$Sigma1[1]` 134 or $t0,$t0,$t1 ; Ch(e,f,g) 135 add $T,$T,@X[$i] 136 xor $a0,$a0,$a1 ; Sigma1(e) 137 add $T,$T,$t0 138 add $T,$T,$a0 139 140 $ROR $a0,$a,$Sigma0[0] 141 $ROR $a1,$a,$Sigma0[1] 142 and $t0,$a,$b 143 and $t1,$a,$c 144 xor $a0,$a0,$a1 145 $ROR $a1,$a1,`$Sigma0[2]-$Sigma0[1]` 146 xor $t0,$t0,$t1 147 and $t1,$b,$c 148 xor $a0,$a0,$a1 ; Sigma0(a) 149 add $d,$d,$T 150 xor $t0,$t0,$t1 ; Maj(a,b,c) 151 add $h,$T,$a0 152 add $h,$h,$t0 153 154___ 155} 156 157sub ROUND_16_xx { 158my ($i,$a,$b,$c,$d,$e,$f,$g,$h)=@_; 159$i-=16; 160$code.=<<___; 161 $ROR $a0,@X[($i+1)%16],$sigma0[0] 162 $ROR $a1,@X[($i+1)%16],$sigma0[1] 163 $ROR $t0,@X[($i+14)%16],$sigma1[0] 164 $ROR $t1,@X[($i+14)%16],$sigma1[1] 165 xor $a0,$a0,$a1 166 $SHR $a1,@X[($i+1)%16],$sigma0[2] 167 xor $t0,$t0,$t1 168 $SHR $t1,@X[($i+14)%16],$sigma1[2] 169 add @X[$i],@X[$i],@X[($i+9)%16] 170 xor $a0,$a0,$a1 ; sigma0(X[(i+1)&0x0f]) 171 xor $t0,$t0,$t1 ; sigma1(X[(i+14)&0x0f]) 172 add @X[$i],@X[$i],$a0 173 add @X[$i],@X[$i],$t0 174___ 175&ROUND_00_15($i,$a,$b,$c,$d,$e,$f,$g,$h); 176} 177 178$code=<<___; 179.machine "any" 180.text 181 182.globl $func 183.align 6 184$func: 185 $STU $sp,-$FRAME($sp) 186 mflr r0 187 $SHL $num,$num,`log(16*$SZ)/log(2)` 188 189 $PUSH $ctx,`$FRAME-$SIZE_T*22`($sp) 190 191 $PUSH $toc,`$FRAME-$SIZE_T*20`($sp) 192 $PUSH r13,`$FRAME-$SIZE_T*19`($sp) 193 $PUSH r14,`$FRAME-$SIZE_T*18`($sp) 194 $PUSH r15,`$FRAME-$SIZE_T*17`($sp) 195 $PUSH r16,`$FRAME-$SIZE_T*16`($sp) 196 $PUSH r17,`$FRAME-$SIZE_T*15`($sp) 197 $PUSH r18,`$FRAME-$SIZE_T*14`($sp) 198 $PUSH r19,`$FRAME-$SIZE_T*13`($sp) 199 $PUSH r20,`$FRAME-$SIZE_T*12`($sp) 200 $PUSH r21,`$FRAME-$SIZE_T*11`($sp) 201 $PUSH r22,`$FRAME-$SIZE_T*10`($sp) 202 $PUSH r23,`$FRAME-$SIZE_T*9`($sp) 203 $PUSH r24,`$FRAME-$SIZE_T*8`($sp) 204 $PUSH r25,`$FRAME-$SIZE_T*7`($sp) 205 $PUSH r26,`$FRAME-$SIZE_T*6`($sp) 206 $PUSH r27,`$FRAME-$SIZE_T*5`($sp) 207 $PUSH r28,`$FRAME-$SIZE_T*4`($sp) 208 $PUSH r29,`$FRAME-$SIZE_T*3`($sp) 209 $PUSH r30,`$FRAME-$SIZE_T*2`($sp) 210 $PUSH r31,`$FRAME-$SIZE_T*1`($sp) 211 $PUSH r0,`$FRAME+$LRSAVE`($sp) 212 213 $LD $A,`0*$SZ`($ctx) 214 mr $inp,r4 ; incarnate $inp 215 $LD $B,`1*$SZ`($ctx) 216 $LD $C,`2*$SZ`($ctx) 217 $LD $D,`3*$SZ`($ctx) 218 $LD $E,`4*$SZ`($ctx) 219 $LD $F,`5*$SZ`($ctx) 220 $LD $G,`6*$SZ`($ctx) 221 $LD $H,`7*$SZ`($ctx) 222 223 bl LPICmeup 224LPICedup: 225 andi. r0,$inp,3 226 bne Lunaligned 227Laligned: 228 add $num,$inp,$num 229 $PUSH $num,`$FRAME-$SIZE_T*24`($sp) ; end pointer 230 $PUSH $inp,`$FRAME-$SIZE_T*23`($sp) ; inp pointer 231 bl Lsha2_block_private 232 b Ldone 233 234; PowerPC specification allows an implementation to be ill-behaved 235; upon unaligned access which crosses page boundary. "Better safe 236; than sorry" principle makes me treat it specially. But I don't 237; look for particular offending word, but rather for the input 238; block which crosses the boundary. Once found that block is aligned 239; and hashed separately... 240.align 4 241Lunaligned: 242 subfic $t1,$inp,4096 243 andi. $t1,$t1,`4096-16*$SZ` ; distance to closest page boundary 244 beq Lcross_page 245 $UCMP $num,$t1 246 ble- Laligned ; didn't cross the page boundary 247 subfc $num,$t1,$num 248 add $t1,$inp,$t1 249 $PUSH $num,`$FRAME-$SIZE_T*25`($sp) ; save real remaining num 250 $PUSH $t1,`$FRAME-$SIZE_T*24`($sp) ; intermediate end pointer 251 $PUSH $inp,`$FRAME-$SIZE_T*23`($sp) ; inp pointer 252 bl Lsha2_block_private 253 ; $inp equals to the intermediate end pointer here 254 $POP $num,`$FRAME-$SIZE_T*25`($sp) ; restore real remaining num 255Lcross_page: 256 li $t1,`16*$SZ/4` 257 mtctr $t1 258 addi r20,$sp,$LOCALS ; aligned spot below the frame 259Lmemcpy: 260 lbz r16,0($inp) 261 lbz r17,1($inp) 262 lbz r18,2($inp) 263 lbz r19,3($inp) 264 addi $inp,$inp,4 265 stb r16,0(r20) 266 stb r17,1(r20) 267 stb r18,2(r20) 268 stb r19,3(r20) 269 addi r20,r20,4 270 bdnz Lmemcpy 271 272 $PUSH $inp,`$FRAME-$SIZE_T*26`($sp) ; save real inp 273 addi $t1,$sp,`$LOCALS+16*$SZ` ; fictitious end pointer 274 addi $inp,$sp,$LOCALS ; fictitious inp pointer 275 $PUSH $num,`$FRAME-$SIZE_T*25`($sp) ; save real num 276 $PUSH $t1,`$FRAME-$SIZE_T*24`($sp) ; end pointer 277 $PUSH $inp,`$FRAME-$SIZE_T*23`($sp) ; inp pointer 278 bl Lsha2_block_private 279 $POP $inp,`$FRAME-$SIZE_T*26`($sp) ; restore real inp 280 $POP $num,`$FRAME-$SIZE_T*25`($sp) ; restore real num 281 addic. $num,$num,`-16*$SZ` ; num-- 282 bne- Lunaligned 283 284Ldone: 285 $POP r0,`$FRAME+$LRSAVE`($sp) 286 $POP $toc,`$FRAME-$SIZE_T*20`($sp) 287 $POP r13,`$FRAME-$SIZE_T*19`($sp) 288 $POP r14,`$FRAME-$SIZE_T*18`($sp) 289 $POP r15,`$FRAME-$SIZE_T*17`($sp) 290 $POP r16,`$FRAME-$SIZE_T*16`($sp) 291 $POP r17,`$FRAME-$SIZE_T*15`($sp) 292 $POP r18,`$FRAME-$SIZE_T*14`($sp) 293 $POP r19,`$FRAME-$SIZE_T*13`($sp) 294 $POP r20,`$FRAME-$SIZE_T*12`($sp) 295 $POP r21,`$FRAME-$SIZE_T*11`($sp) 296 $POP r22,`$FRAME-$SIZE_T*10`($sp) 297 $POP r23,`$FRAME-$SIZE_T*9`($sp) 298 $POP r24,`$FRAME-$SIZE_T*8`($sp) 299 $POP r25,`$FRAME-$SIZE_T*7`($sp) 300 $POP r26,`$FRAME-$SIZE_T*6`($sp) 301 $POP r27,`$FRAME-$SIZE_T*5`($sp) 302 $POP r28,`$FRAME-$SIZE_T*4`($sp) 303 $POP r29,`$FRAME-$SIZE_T*3`($sp) 304 $POP r30,`$FRAME-$SIZE_T*2`($sp) 305 $POP r31,`$FRAME-$SIZE_T*1`($sp) 306 mtlr r0 307 addi $sp,$sp,$FRAME 308 blr 309 .long 0 310 .byte 0,12,4,1,0x80,18,3,0 311 .long 0 312 313.align 4 314Lsha2_block_private: 315___ 316for($i=0;$i<16;$i++) { 317$code.=<<___ if ($SZ==4); 318 lwz @X[$i],`$i*$SZ`($inp) 319___ 320# 64-bit loads are split to 2x32-bit ones, as CPU can't handle 321# unaligned 64-bit loads, only 32-bit ones... 322$code.=<<___ if ($SZ==8); 323 lwz $t0,`$i*$SZ`($inp) 324 lwz @X[$i],`$i*$SZ+4`($inp) 325 insrdi @X[$i],$t0,32,0 326___ 327 &ROUND_00_15($i,@V); 328 unshift(@V,pop(@V)); 329} 330$code.=<<___; 331 li $T,`$rounds/16-1` 332 mtctr $T 333.align 4 334Lrounds: 335 addi $Tbl,$Tbl,`16*$SZ` 336___ 337for(;$i<32;$i++) { 338 &ROUND_16_xx($i,@V); 339 unshift(@V,pop(@V)); 340} 341$code.=<<___; 342 bdnz- Lrounds 343 344 $POP $ctx,`$FRAME-$SIZE_T*22`($sp) 345 $POP $inp,`$FRAME-$SIZE_T*23`($sp) ; inp pointer 346 $POP $num,`$FRAME-$SIZE_T*24`($sp) ; end pointer 347 subi $Tbl,$Tbl,`($rounds-16)*$SZ` ; rewind Tbl 348 349 $LD r16,`0*$SZ`($ctx) 350 $LD r17,`1*$SZ`($ctx) 351 $LD r18,`2*$SZ`($ctx) 352 $LD r19,`3*$SZ`($ctx) 353 $LD r20,`4*$SZ`($ctx) 354 $LD r21,`5*$SZ`($ctx) 355 $LD r22,`6*$SZ`($ctx) 356 addi $inp,$inp,`16*$SZ` ; advance inp 357 $LD r23,`7*$SZ`($ctx) 358 add $A,$A,r16 359 add $B,$B,r17 360 $PUSH $inp,`$FRAME-$SIZE_T*23`($sp) 361 add $C,$C,r18 362 $ST $A,`0*$SZ`($ctx) 363 add $D,$D,r19 364 $ST $B,`1*$SZ`($ctx) 365 add $E,$E,r20 366 $ST $C,`2*$SZ`($ctx) 367 add $F,$F,r21 368 $ST $D,`3*$SZ`($ctx) 369 add $G,$G,r22 370 $ST $E,`4*$SZ`($ctx) 371 add $H,$H,r23 372 $ST $F,`5*$SZ`($ctx) 373 $ST $G,`6*$SZ`($ctx) 374 $UCMP $inp,$num 375 $ST $H,`7*$SZ`($ctx) 376 bne Lsha2_block_private 377 blr 378 .long 0 379 .byte 0,12,0x14,0,0,0,0,0 380___ 381 382# Ugly hack here, because PPC assembler syntax seem to vary too 383# much from platforms to platform... 384$code.=<<___; 385.align 6 386LPICmeup: 387 mflr r0 388 bcl 20,31,\$+4 389 mflr $Tbl ; vvvvvv "distance" between . and 1st data entry 390 addi $Tbl,$Tbl,`64-8` 391 mtlr r0 392 blr 393 .long 0 394 .byte 0,12,0x14,0,0,0,0,0 395 .space `64-9*4` 396___ 397$code.=<<___ if ($SZ==8); 398 .long 0x428a2f98,0xd728ae22,0x71374491,0x23ef65cd 399 .long 0xb5c0fbcf,0xec4d3b2f,0xe9b5dba5,0x8189dbbc 400 .long 0x3956c25b,0xf348b538,0x59f111f1,0xb605d019 401 .long 0x923f82a4,0xaf194f9b,0xab1c5ed5,0xda6d8118 402 .long 0xd807aa98,0xa3030242,0x12835b01,0x45706fbe 403 .long 0x243185be,0x4ee4b28c,0x550c7dc3,0xd5ffb4e2 404 .long 0x72be5d74,0xf27b896f,0x80deb1fe,0x3b1696b1 405 .long 0x9bdc06a7,0x25c71235,0xc19bf174,0xcf692694 406 .long 0xe49b69c1,0x9ef14ad2,0xefbe4786,0x384f25e3 407 .long 0x0fc19dc6,0x8b8cd5b5,0x240ca1cc,0x77ac9c65 408 .long 0x2de92c6f,0x592b0275,0x4a7484aa,0x6ea6e483 409 .long 0x5cb0a9dc,0xbd41fbd4,0x76f988da,0x831153b5 410 .long 0x983e5152,0xee66dfab,0xa831c66d,0x2db43210 411 .long 0xb00327c8,0x98fb213f,0xbf597fc7,0xbeef0ee4 412 .long 0xc6e00bf3,0x3da88fc2,0xd5a79147,0x930aa725 413 .long 0x06ca6351,0xe003826f,0x14292967,0x0a0e6e70 414 .long 0x27b70a85,0x46d22ffc,0x2e1b2138,0x5c26c926 415 .long 0x4d2c6dfc,0x5ac42aed,0x53380d13,0x9d95b3df 416 .long 0x650a7354,0x8baf63de,0x766a0abb,0x3c77b2a8 417 .long 0x81c2c92e,0x47edaee6,0x92722c85,0x1482353b 418 .long 0xa2bfe8a1,0x4cf10364,0xa81a664b,0xbc423001 419 .long 0xc24b8b70,0xd0f89791,0xc76c51a3,0x0654be30 420 .long 0xd192e819,0xd6ef5218,0xd6990624,0x5565a910 421 .long 0xf40e3585,0x5771202a,0x106aa070,0x32bbd1b8 422 .long 0x19a4c116,0xb8d2d0c8,0x1e376c08,0x5141ab53 423 .long 0x2748774c,0xdf8eeb99,0x34b0bcb5,0xe19b48a8 424 .long 0x391c0cb3,0xc5c95a63,0x4ed8aa4a,0xe3418acb 425 .long 0x5b9cca4f,0x7763e373,0x682e6ff3,0xd6b2b8a3 426 .long 0x748f82ee,0x5defb2fc,0x78a5636f,0x43172f60 427 .long 0x84c87814,0xa1f0ab72,0x8cc70208,0x1a6439ec 428 .long 0x90befffa,0x23631e28,0xa4506ceb,0xde82bde9 429 .long 0xbef9a3f7,0xb2c67915,0xc67178f2,0xe372532b 430 .long 0xca273ece,0xea26619c,0xd186b8c7,0x21c0c207 431 .long 0xeada7dd6,0xcde0eb1e,0xf57d4f7f,0xee6ed178 432 .long 0x06f067aa,0x72176fba,0x0a637dc5,0xa2c898a6 433 .long 0x113f9804,0xbef90dae,0x1b710b35,0x131c471b 434 .long 0x28db77f5,0x23047d84,0x32caab7b,0x40c72493 435 .long 0x3c9ebe0a,0x15c9bebc,0x431d67c4,0x9c100d4c 436 .long 0x4cc5d4be,0xcb3e42b6,0x597f299c,0xfc657e2a 437 .long 0x5fcb6fab,0x3ad6faec,0x6c44198c,0x4a475817 438___ 439$code.=<<___ if ($SZ==4); 440 .long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5 441 .long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5 442 .long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3 443 .long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174 444 .long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc 445 .long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da 446 .long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7 447 .long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967 448 .long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13 449 .long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85 450 .long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3 451 .long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070 452 .long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5 453 .long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3 454 .long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208 455 .long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2 456___ 457 458$code =~ s/\`([^\`]*)\`/eval $1/gem; 459print $code; 460close STDOUT; 461