Ard Biesheuvel | e4e7f10 | 2013-09-16 18:31:38 +0200 | [diff] [blame] | 1 | #!/usr/bin/env perl |
| 2 | |
| 3 | # ==================================================================== |
| 4 | # Written by Andy Polyakov <appro@openssl.org> 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 | # Specific modes and adaptation for Linux kernel by Ard Biesheuvel |
| 10 | # <ard.biesheuvel@linaro.org>. Permission to use under GPL terms is |
| 11 | # granted. |
| 12 | # ==================================================================== |
| 13 | |
| 14 | # Bit-sliced AES for ARM NEON |
| 15 | # |
| 16 | # February 2012. |
| 17 | # |
| 18 | # This implementation is direct adaptation of bsaes-x86_64 module for |
| 19 | # ARM NEON. Except that this module is endian-neutral [in sense that |
| 20 | # it can be compiled for either endianness] by courtesy of vld1.8's |
| 21 | # neutrality. Initial version doesn't implement interface to OpenSSL, |
| 22 | # only low-level primitives and unsupported entry points, just enough |
| 23 | # to collect performance results, which for Cortex-A8 core are: |
| 24 | # |
| 25 | # encrypt 19.5 cycles per byte processed with 128-bit key |
| 26 | # decrypt 22.1 cycles per byte processed with 128-bit key |
| 27 | # key conv. 440 cycles per 128-bit key/0.18 of 8x block |
| 28 | # |
| 29 | # Snapdragon S4 encrypts byte in 17.6 cycles and decrypts in 19.7, |
| 30 | # which is [much] worse than anticipated (for further details see |
| 31 | # http://www.openssl.org/~appro/Snapdragon-S4.html). |
| 32 | # |
| 33 | # Cortex-A15 manages in 14.2/16.1 cycles [when integer-only code |
| 34 | # manages in 20.0 cycles]. |
| 35 | # |
| 36 | # When comparing to x86_64 results keep in mind that NEON unit is |
| 37 | # [mostly] single-issue and thus can't [fully] benefit from |
| 38 | # instruction-level parallelism. And when comparing to aes-armv4 |
| 39 | # results keep in mind key schedule conversion overhead (see |
| 40 | # bsaes-x86_64.pl for further details)... |
| 41 | # |
| 42 | # <appro@openssl.org> |
| 43 | |
| 44 | # April-August 2013 |
| 45 | # |
| 46 | # Add CBC, CTR and XTS subroutines, adapt for kernel use. |
| 47 | # |
| 48 | # <ard.biesheuvel@linaro.org> |
| 49 | |
| 50 | while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {} |
| 51 | open STDOUT,">$output"; |
| 52 | |
| 53 | my ($inp,$out,$len,$key)=("r0","r1","r2","r3"); |
| 54 | my @XMM=map("q$_",(0..15)); |
| 55 | |
| 56 | { |
| 57 | my ($key,$rounds,$const)=("r4","r5","r6"); |
| 58 | |
| 59 | sub Dlo() { shift=~m|q([1]?[0-9])|?"d".($1*2):""; } |
| 60 | sub Dhi() { shift=~m|q([1]?[0-9])|?"d".($1*2+1):""; } |
| 61 | |
| 62 | sub Sbox { |
| 63 | # input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb |
| 64 | # output in lsb > [b0, b1, b4, b6, b3, b7, b2, b5] < msb |
| 65 | my @b=@_[0..7]; |
| 66 | my @t=@_[8..11]; |
| 67 | my @s=@_[12..15]; |
| 68 | &InBasisChange (@b); |
| 69 | &Inv_GF256 (@b[6,5,0,3,7,1,4,2],@t,@s); |
| 70 | &OutBasisChange (@b[7,1,4,2,6,5,0,3]); |
| 71 | } |
| 72 | |
| 73 | sub InBasisChange { |
| 74 | # input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb |
| 75 | # output in lsb > [b6, b5, b0, b3, b7, b1, b4, b2] < msb |
| 76 | my @b=@_[0..7]; |
| 77 | $code.=<<___; |
| 78 | veor @b[2], @b[2], @b[1] |
| 79 | veor @b[5], @b[5], @b[6] |
| 80 | veor @b[3], @b[3], @b[0] |
| 81 | veor @b[6], @b[6], @b[2] |
| 82 | veor @b[5], @b[5], @b[0] |
| 83 | |
| 84 | veor @b[6], @b[6], @b[3] |
| 85 | veor @b[3], @b[3], @b[7] |
| 86 | veor @b[7], @b[7], @b[5] |
| 87 | veor @b[3], @b[3], @b[4] |
| 88 | veor @b[4], @b[4], @b[5] |
| 89 | |
| 90 | veor @b[2], @b[2], @b[7] |
| 91 | veor @b[3], @b[3], @b[1] |
| 92 | veor @b[1], @b[1], @b[5] |
| 93 | ___ |
| 94 | } |
| 95 | |
| 96 | sub OutBasisChange { |
| 97 | # input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb |
| 98 | # output in lsb > [b6, b1, b2, b4, b7, b0, b3, b5] < msb |
| 99 | my @b=@_[0..7]; |
| 100 | $code.=<<___; |
| 101 | veor @b[0], @b[0], @b[6] |
| 102 | veor @b[1], @b[1], @b[4] |
| 103 | veor @b[4], @b[4], @b[6] |
| 104 | veor @b[2], @b[2], @b[0] |
| 105 | veor @b[6], @b[6], @b[1] |
| 106 | |
| 107 | veor @b[1], @b[1], @b[5] |
| 108 | veor @b[5], @b[5], @b[3] |
| 109 | veor @b[3], @b[3], @b[7] |
| 110 | veor @b[7], @b[7], @b[5] |
| 111 | veor @b[2], @b[2], @b[5] |
| 112 | |
| 113 | veor @b[4], @b[4], @b[7] |
| 114 | ___ |
| 115 | } |
| 116 | |
| 117 | sub InvSbox { |
| 118 | # input in lsb > [b0, b1, b2, b3, b4, b5, b6, b7] < msb |
| 119 | # output in lsb > [b0, b1, b6, b4, b2, b7, b3, b5] < msb |
| 120 | my @b=@_[0..7]; |
| 121 | my @t=@_[8..11]; |
| 122 | my @s=@_[12..15]; |
| 123 | &InvInBasisChange (@b); |
| 124 | &Inv_GF256 (@b[5,1,2,6,3,7,0,4],@t,@s); |
| 125 | &InvOutBasisChange (@b[3,7,0,4,5,1,2,6]); |
| 126 | } |
| 127 | |
| 128 | sub InvInBasisChange { # OutBasisChange in reverse (with twist) |
| 129 | my @b=@_[5,1,2,6,3,7,0,4]; |
| 130 | $code.=<<___ |
| 131 | veor @b[1], @b[1], @b[7] |
| 132 | veor @b[4], @b[4], @b[7] |
| 133 | |
| 134 | veor @b[7], @b[7], @b[5] |
| 135 | veor @b[1], @b[1], @b[3] |
| 136 | veor @b[2], @b[2], @b[5] |
| 137 | veor @b[3], @b[3], @b[7] |
| 138 | |
| 139 | veor @b[6], @b[6], @b[1] |
| 140 | veor @b[2], @b[2], @b[0] |
| 141 | veor @b[5], @b[5], @b[3] |
| 142 | veor @b[4], @b[4], @b[6] |
| 143 | veor @b[0], @b[0], @b[6] |
| 144 | veor @b[1], @b[1], @b[4] |
| 145 | ___ |
| 146 | } |
| 147 | |
| 148 | sub InvOutBasisChange { # InBasisChange in reverse |
| 149 | my @b=@_[2,5,7,3,6,1,0,4]; |
| 150 | $code.=<<___; |
| 151 | veor @b[1], @b[1], @b[5] |
| 152 | veor @b[2], @b[2], @b[7] |
| 153 | |
| 154 | veor @b[3], @b[3], @b[1] |
| 155 | veor @b[4], @b[4], @b[5] |
| 156 | veor @b[7], @b[7], @b[5] |
| 157 | veor @b[3], @b[3], @b[4] |
| 158 | veor @b[5], @b[5], @b[0] |
| 159 | veor @b[3], @b[3], @b[7] |
| 160 | veor @b[6], @b[6], @b[2] |
| 161 | veor @b[2], @b[2], @b[1] |
| 162 | veor @b[6], @b[6], @b[3] |
| 163 | |
| 164 | veor @b[3], @b[3], @b[0] |
| 165 | veor @b[5], @b[5], @b[6] |
| 166 | ___ |
| 167 | } |
| 168 | |
| 169 | sub Mul_GF4 { |
| 170 | #;************************************************************* |
| 171 | #;* Mul_GF4: Input x0-x1,y0-y1 Output x0-x1 Temp t0 (8) * |
| 172 | #;************************************************************* |
| 173 | my ($x0,$x1,$y0,$y1,$t0,$t1)=@_; |
| 174 | $code.=<<___; |
| 175 | veor $t0, $y0, $y1 |
| 176 | vand $t0, $t0, $x0 |
| 177 | veor $x0, $x0, $x1 |
| 178 | vand $t1, $x1, $y0 |
| 179 | vand $x0, $x0, $y1 |
| 180 | veor $x1, $t1, $t0 |
| 181 | veor $x0, $x0, $t1 |
| 182 | ___ |
| 183 | } |
| 184 | |
| 185 | sub Mul_GF4_N { # not used, see next subroutine |
| 186 | # multiply and scale by N |
| 187 | my ($x0,$x1,$y0,$y1,$t0)=@_; |
| 188 | $code.=<<___; |
| 189 | veor $t0, $y0, $y1 |
| 190 | vand $t0, $t0, $x0 |
| 191 | veor $x0, $x0, $x1 |
| 192 | vand $x1, $x1, $y0 |
| 193 | vand $x0, $x0, $y1 |
| 194 | veor $x1, $x1, $x0 |
| 195 | veor $x0, $x0, $t0 |
| 196 | ___ |
| 197 | } |
| 198 | |
| 199 | sub Mul_GF4_N_GF4 { |
| 200 | # interleaved Mul_GF4_N and Mul_GF4 |
| 201 | my ($x0,$x1,$y0,$y1,$t0, |
| 202 | $x2,$x3,$y2,$y3,$t1)=@_; |
| 203 | $code.=<<___; |
| 204 | veor $t0, $y0, $y1 |
| 205 | veor $t1, $y2, $y3 |
| 206 | vand $t0, $t0, $x0 |
| 207 | vand $t1, $t1, $x2 |
| 208 | veor $x0, $x0, $x1 |
| 209 | veor $x2, $x2, $x3 |
| 210 | vand $x1, $x1, $y0 |
| 211 | vand $x3, $x3, $y2 |
| 212 | vand $x0, $x0, $y1 |
| 213 | vand $x2, $x2, $y3 |
| 214 | veor $x1, $x1, $x0 |
| 215 | veor $x2, $x2, $x3 |
| 216 | veor $x0, $x0, $t0 |
| 217 | veor $x3, $x3, $t1 |
| 218 | ___ |
| 219 | } |
| 220 | sub Mul_GF16_2 { |
| 221 | my @x=@_[0..7]; |
| 222 | my @y=@_[8..11]; |
| 223 | my @t=@_[12..15]; |
| 224 | $code.=<<___; |
| 225 | veor @t[0], @x[0], @x[2] |
| 226 | veor @t[1], @x[1], @x[3] |
| 227 | ___ |
| 228 | &Mul_GF4 (@x[0], @x[1], @y[0], @y[1], @t[2..3]); |
| 229 | $code.=<<___; |
| 230 | veor @y[0], @y[0], @y[2] |
| 231 | veor @y[1], @y[1], @y[3] |
| 232 | ___ |
| 233 | Mul_GF4_N_GF4 (@t[0], @t[1], @y[0], @y[1], @t[3], |
| 234 | @x[2], @x[3], @y[2], @y[3], @t[2]); |
| 235 | $code.=<<___; |
| 236 | veor @x[0], @x[0], @t[0] |
| 237 | veor @x[2], @x[2], @t[0] |
| 238 | veor @x[1], @x[1], @t[1] |
| 239 | veor @x[3], @x[3], @t[1] |
| 240 | |
| 241 | veor @t[0], @x[4], @x[6] |
| 242 | veor @t[1], @x[5], @x[7] |
| 243 | ___ |
| 244 | &Mul_GF4_N_GF4 (@t[0], @t[1], @y[0], @y[1], @t[3], |
| 245 | @x[6], @x[7], @y[2], @y[3], @t[2]); |
| 246 | $code.=<<___; |
| 247 | veor @y[0], @y[0], @y[2] |
| 248 | veor @y[1], @y[1], @y[3] |
| 249 | ___ |
| 250 | &Mul_GF4 (@x[4], @x[5], @y[0], @y[1], @t[2..3]); |
| 251 | $code.=<<___; |
| 252 | veor @x[4], @x[4], @t[0] |
| 253 | veor @x[6], @x[6], @t[0] |
| 254 | veor @x[5], @x[5], @t[1] |
| 255 | veor @x[7], @x[7], @t[1] |
| 256 | ___ |
| 257 | } |
| 258 | sub Inv_GF256 { |
| 259 | #;******************************************************************** |
| 260 | #;* Inv_GF256: Input x0-x7 Output x0-x7 Temp t0-t3,s0-s3 (144) * |
| 261 | #;******************************************************************** |
| 262 | my @x=@_[0..7]; |
| 263 | my @t=@_[8..11]; |
| 264 | my @s=@_[12..15]; |
| 265 | # direct optimizations from hardware |
| 266 | $code.=<<___; |
| 267 | veor @t[3], @x[4], @x[6] |
| 268 | veor @t[2], @x[5], @x[7] |
| 269 | veor @t[1], @x[1], @x[3] |
| 270 | veor @s[1], @x[7], @x[6] |
| 271 | vmov @t[0], @t[2] |
| 272 | veor @s[0], @x[0], @x[2] |
| 273 | |
| 274 | vorr @t[2], @t[2], @t[1] |
| 275 | veor @s[3], @t[3], @t[0] |
| 276 | vand @s[2], @t[3], @s[0] |
| 277 | vorr @t[3], @t[3], @s[0] |
| 278 | veor @s[0], @s[0], @t[1] |
| 279 | vand @t[0], @t[0], @t[1] |
| 280 | veor @t[1], @x[3], @x[2] |
| 281 | vand @s[3], @s[3], @s[0] |
| 282 | vand @s[1], @s[1], @t[1] |
| 283 | veor @t[1], @x[4], @x[5] |
| 284 | veor @s[0], @x[1], @x[0] |
| 285 | veor @t[3], @t[3], @s[1] |
| 286 | veor @t[2], @t[2], @s[1] |
| 287 | vand @s[1], @t[1], @s[0] |
| 288 | vorr @t[1], @t[1], @s[0] |
| 289 | veor @t[3], @t[3], @s[3] |
| 290 | veor @t[0], @t[0], @s[1] |
| 291 | veor @t[2], @t[2], @s[2] |
| 292 | veor @t[1], @t[1], @s[3] |
| 293 | veor @t[0], @t[0], @s[2] |
| 294 | vand @s[0], @x[7], @x[3] |
| 295 | veor @t[1], @t[1], @s[2] |
| 296 | vand @s[1], @x[6], @x[2] |
| 297 | vand @s[2], @x[5], @x[1] |
| 298 | vorr @s[3], @x[4], @x[0] |
| 299 | veor @t[3], @t[3], @s[0] |
| 300 | veor @t[1], @t[1], @s[2] |
| 301 | veor @t[0], @t[0], @s[3] |
| 302 | veor @t[2], @t[2], @s[1] |
| 303 | |
| 304 | @ Inv_GF16 \t0, \t1, \t2, \t3, \s0, \s1, \s2, \s3 |
| 305 | |
| 306 | @ new smaller inversion |
| 307 | |
| 308 | vand @s[2], @t[3], @t[1] |
| 309 | vmov @s[0], @t[0] |
| 310 | |
| 311 | veor @s[1], @t[2], @s[2] |
| 312 | veor @s[3], @t[0], @s[2] |
| 313 | veor @s[2], @t[0], @s[2] @ @s[2]=@s[3] |
| 314 | |
| 315 | vbsl @s[1], @t[1], @t[0] |
| 316 | vbsl @s[3], @t[3], @t[2] |
| 317 | veor @t[3], @t[3], @t[2] |
| 318 | |
| 319 | vbsl @s[0], @s[1], @s[2] |
| 320 | vbsl @t[0], @s[2], @s[1] |
| 321 | |
| 322 | vand @s[2], @s[0], @s[3] |
| 323 | veor @t[1], @t[1], @t[0] |
| 324 | |
| 325 | veor @s[2], @s[2], @t[3] |
| 326 | ___ |
| 327 | # output in s3, s2, s1, t1 |
| 328 | |
| 329 | # Mul_GF16_2 \x0, \x1, \x2, \x3, \x4, \x5, \x6, \x7, \t2, \t3, \t0, \t1, \s0, \s1, \s2, \s3 |
| 330 | |
| 331 | # Mul_GF16_2 \x0, \x1, \x2, \x3, \x4, \x5, \x6, \x7, \s3, \s2, \s1, \t1, \s0, \t0, \t2, \t3 |
| 332 | &Mul_GF16_2(@x,@s[3,2,1],@t[1],@s[0],@t[0,2,3]); |
| 333 | |
| 334 | ### output msb > [x3,x2,x1,x0,x7,x6,x5,x4] < lsb |
| 335 | } |
| 336 | |
| 337 | # AES linear components |
| 338 | |
| 339 | sub ShiftRows { |
| 340 | my @x=@_[0..7]; |
| 341 | my @t=@_[8..11]; |
| 342 | my $mask=pop; |
| 343 | $code.=<<___; |
| 344 | vldmia $key!, {@t[0]-@t[3]} |
| 345 | veor @t[0], @t[0], @x[0] |
| 346 | veor @t[1], @t[1], @x[1] |
| 347 | vtbl.8 `&Dlo(@x[0])`, {@t[0]}, `&Dlo($mask)` |
| 348 | vtbl.8 `&Dhi(@x[0])`, {@t[0]}, `&Dhi($mask)` |
| 349 | vldmia $key!, {@t[0]} |
| 350 | veor @t[2], @t[2], @x[2] |
| 351 | vtbl.8 `&Dlo(@x[1])`, {@t[1]}, `&Dlo($mask)` |
| 352 | vtbl.8 `&Dhi(@x[1])`, {@t[1]}, `&Dhi($mask)` |
| 353 | vldmia $key!, {@t[1]} |
| 354 | veor @t[3], @t[3], @x[3] |
| 355 | vtbl.8 `&Dlo(@x[2])`, {@t[2]}, `&Dlo($mask)` |
| 356 | vtbl.8 `&Dhi(@x[2])`, {@t[2]}, `&Dhi($mask)` |
| 357 | vldmia $key!, {@t[2]} |
| 358 | vtbl.8 `&Dlo(@x[3])`, {@t[3]}, `&Dlo($mask)` |
| 359 | vtbl.8 `&Dhi(@x[3])`, {@t[3]}, `&Dhi($mask)` |
| 360 | vldmia $key!, {@t[3]} |
| 361 | veor @t[0], @t[0], @x[4] |
| 362 | veor @t[1], @t[1], @x[5] |
| 363 | vtbl.8 `&Dlo(@x[4])`, {@t[0]}, `&Dlo($mask)` |
| 364 | vtbl.8 `&Dhi(@x[4])`, {@t[0]}, `&Dhi($mask)` |
| 365 | veor @t[2], @t[2], @x[6] |
| 366 | vtbl.8 `&Dlo(@x[5])`, {@t[1]}, `&Dlo($mask)` |
| 367 | vtbl.8 `&Dhi(@x[5])`, {@t[1]}, `&Dhi($mask)` |
| 368 | veor @t[3], @t[3], @x[7] |
| 369 | vtbl.8 `&Dlo(@x[6])`, {@t[2]}, `&Dlo($mask)` |
| 370 | vtbl.8 `&Dhi(@x[6])`, {@t[2]}, `&Dhi($mask)` |
| 371 | vtbl.8 `&Dlo(@x[7])`, {@t[3]}, `&Dlo($mask)` |
| 372 | vtbl.8 `&Dhi(@x[7])`, {@t[3]}, `&Dhi($mask)` |
| 373 | ___ |
| 374 | } |
| 375 | |
| 376 | sub MixColumns { |
| 377 | # modified to emit output in order suitable for feeding back to aesenc[last] |
| 378 | my @x=@_[0..7]; |
| 379 | my @t=@_[8..15]; |
| 380 | my $inv=@_[16]; # optional |
| 381 | $code.=<<___; |
| 382 | vext.8 @t[0], @x[0], @x[0], #12 @ x0 <<< 32 |
| 383 | vext.8 @t[1], @x[1], @x[1], #12 |
| 384 | veor @x[0], @x[0], @t[0] @ x0 ^ (x0 <<< 32) |
| 385 | vext.8 @t[2], @x[2], @x[2], #12 |
| 386 | veor @x[1], @x[1], @t[1] |
| 387 | vext.8 @t[3], @x[3], @x[3], #12 |
| 388 | veor @x[2], @x[2], @t[2] |
| 389 | vext.8 @t[4], @x[4], @x[4], #12 |
| 390 | veor @x[3], @x[3], @t[3] |
| 391 | vext.8 @t[5], @x[5], @x[5], #12 |
| 392 | veor @x[4], @x[4], @t[4] |
| 393 | vext.8 @t[6], @x[6], @x[6], #12 |
| 394 | veor @x[5], @x[5], @t[5] |
| 395 | vext.8 @t[7], @x[7], @x[7], #12 |
| 396 | veor @x[6], @x[6], @t[6] |
| 397 | |
| 398 | veor @t[1], @t[1], @x[0] |
| 399 | veor @x[7], @x[7], @t[7] |
| 400 | vext.8 @x[0], @x[0], @x[0], #8 @ (x0 ^ (x0 <<< 32)) <<< 64) |
| 401 | veor @t[2], @t[2], @x[1] |
| 402 | veor @t[0], @t[0], @x[7] |
| 403 | veor @t[1], @t[1], @x[7] |
| 404 | vext.8 @x[1], @x[1], @x[1], #8 |
| 405 | veor @t[5], @t[5], @x[4] |
| 406 | veor @x[0], @x[0], @t[0] |
| 407 | veor @t[6], @t[6], @x[5] |
| 408 | veor @x[1], @x[1], @t[1] |
| 409 | vext.8 @t[0], @x[4], @x[4], #8 |
| 410 | veor @t[4], @t[4], @x[3] |
| 411 | vext.8 @t[1], @x[5], @x[5], #8 |
| 412 | veor @t[7], @t[7], @x[6] |
| 413 | vext.8 @x[4], @x[3], @x[3], #8 |
| 414 | veor @t[3], @t[3], @x[2] |
| 415 | vext.8 @x[5], @x[7], @x[7], #8 |
| 416 | veor @t[4], @t[4], @x[7] |
| 417 | vext.8 @x[3], @x[6], @x[6], #8 |
| 418 | veor @t[3], @t[3], @x[7] |
| 419 | vext.8 @x[6], @x[2], @x[2], #8 |
| 420 | veor @x[7], @t[1], @t[5] |
| 421 | ___ |
| 422 | $code.=<<___ if (!$inv); |
| 423 | veor @x[2], @t[0], @t[4] |
| 424 | veor @x[4], @x[4], @t[3] |
| 425 | veor @x[5], @x[5], @t[7] |
| 426 | veor @x[3], @x[3], @t[6] |
| 427 | @ vmov @x[2], @t[0] |
| 428 | veor @x[6], @x[6], @t[2] |
| 429 | @ vmov @x[7], @t[1] |
| 430 | ___ |
| 431 | $code.=<<___ if ($inv); |
| 432 | veor @t[3], @t[3], @x[4] |
| 433 | veor @x[5], @x[5], @t[7] |
| 434 | veor @x[2], @x[3], @t[6] |
| 435 | veor @x[3], @t[0], @t[4] |
| 436 | veor @x[4], @x[6], @t[2] |
| 437 | vmov @x[6], @t[3] |
| 438 | @ vmov @x[7], @t[1] |
| 439 | ___ |
| 440 | } |
| 441 | |
| 442 | sub InvMixColumns_orig { |
| 443 | my @x=@_[0..7]; |
| 444 | my @t=@_[8..15]; |
| 445 | |
| 446 | $code.=<<___; |
| 447 | @ multiplication by 0x0e |
| 448 | vext.8 @t[7], @x[7], @x[7], #12 |
| 449 | vmov @t[2], @x[2] |
| 450 | veor @x[2], @x[2], @x[5] @ 2 5 |
| 451 | veor @x[7], @x[7], @x[5] @ 7 5 |
| 452 | vext.8 @t[0], @x[0], @x[0], #12 |
| 453 | vmov @t[5], @x[5] |
| 454 | veor @x[5], @x[5], @x[0] @ 5 0 [1] |
| 455 | veor @x[0], @x[0], @x[1] @ 0 1 |
| 456 | vext.8 @t[1], @x[1], @x[1], #12 |
| 457 | veor @x[1], @x[1], @x[2] @ 1 25 |
| 458 | veor @x[0], @x[0], @x[6] @ 01 6 [2] |
| 459 | vext.8 @t[3], @x[3], @x[3], #12 |
| 460 | veor @x[1], @x[1], @x[3] @ 125 3 [4] |
| 461 | veor @x[2], @x[2], @x[0] @ 25 016 [3] |
| 462 | veor @x[3], @x[3], @x[7] @ 3 75 |
| 463 | veor @x[7], @x[7], @x[6] @ 75 6 [0] |
| 464 | vext.8 @t[6], @x[6], @x[6], #12 |
| 465 | vmov @t[4], @x[4] |
| 466 | veor @x[6], @x[6], @x[4] @ 6 4 |
| 467 | veor @x[4], @x[4], @x[3] @ 4 375 [6] |
| 468 | veor @x[3], @x[3], @x[7] @ 375 756=36 |
| 469 | veor @x[6], @x[6], @t[5] @ 64 5 [7] |
| 470 | veor @x[3], @x[3], @t[2] @ 36 2 |
| 471 | vext.8 @t[5], @t[5], @t[5], #12 |
| 472 | veor @x[3], @x[3], @t[4] @ 362 4 [5] |
| 473 | ___ |
| 474 | my @y = @x[7,5,0,2,1,3,4,6]; |
| 475 | $code.=<<___; |
| 476 | @ multiplication by 0x0b |
| 477 | veor @y[1], @y[1], @y[0] |
| 478 | veor @y[0], @y[0], @t[0] |
| 479 | vext.8 @t[2], @t[2], @t[2], #12 |
| 480 | veor @y[1], @y[1], @t[1] |
| 481 | veor @y[0], @y[0], @t[5] |
| 482 | vext.8 @t[4], @t[4], @t[4], #12 |
| 483 | veor @y[1], @y[1], @t[6] |
| 484 | veor @y[0], @y[0], @t[7] |
| 485 | veor @t[7], @t[7], @t[6] @ clobber t[7] |
| 486 | |
| 487 | veor @y[3], @y[3], @t[0] |
| 488 | veor @y[1], @y[1], @y[0] |
| 489 | vext.8 @t[0], @t[0], @t[0], #12 |
| 490 | veor @y[2], @y[2], @t[1] |
| 491 | veor @y[4], @y[4], @t[1] |
| 492 | vext.8 @t[1], @t[1], @t[1], #12 |
| 493 | veor @y[2], @y[2], @t[2] |
| 494 | veor @y[3], @y[3], @t[2] |
| 495 | veor @y[5], @y[5], @t[2] |
| 496 | veor @y[2], @y[2], @t[7] |
| 497 | vext.8 @t[2], @t[2], @t[2], #12 |
| 498 | veor @y[3], @y[3], @t[3] |
| 499 | veor @y[6], @y[6], @t[3] |
| 500 | veor @y[4], @y[4], @t[3] |
| 501 | veor @y[7], @y[7], @t[4] |
| 502 | vext.8 @t[3], @t[3], @t[3], #12 |
| 503 | veor @y[5], @y[5], @t[4] |
| 504 | veor @y[7], @y[7], @t[7] |
| 505 | veor @t[7], @t[7], @t[5] @ clobber t[7] even more |
| 506 | veor @y[3], @y[3], @t[5] |
| 507 | veor @y[4], @y[4], @t[4] |
| 508 | |
| 509 | veor @y[5], @y[5], @t[7] |
| 510 | vext.8 @t[4], @t[4], @t[4], #12 |
| 511 | veor @y[6], @y[6], @t[7] |
| 512 | veor @y[4], @y[4], @t[7] |
| 513 | |
| 514 | veor @t[7], @t[7], @t[5] |
| 515 | vext.8 @t[5], @t[5], @t[5], #12 |
| 516 | |
| 517 | @ multiplication by 0x0d |
| 518 | veor @y[4], @y[4], @y[7] |
| 519 | veor @t[7], @t[7], @t[6] @ restore t[7] |
| 520 | veor @y[7], @y[7], @t[4] |
| 521 | vext.8 @t[6], @t[6], @t[6], #12 |
| 522 | veor @y[2], @y[2], @t[0] |
| 523 | veor @y[7], @y[7], @t[5] |
| 524 | vext.8 @t[7], @t[7], @t[7], #12 |
| 525 | veor @y[2], @y[2], @t[2] |
| 526 | |
| 527 | veor @y[3], @y[3], @y[1] |
| 528 | veor @y[1], @y[1], @t[1] |
| 529 | veor @y[0], @y[0], @t[0] |
| 530 | veor @y[3], @y[3], @t[0] |
| 531 | veor @y[1], @y[1], @t[5] |
| 532 | veor @y[0], @y[0], @t[5] |
| 533 | vext.8 @t[0], @t[0], @t[0], #12 |
| 534 | veor @y[1], @y[1], @t[7] |
| 535 | veor @y[0], @y[0], @t[6] |
| 536 | veor @y[3], @y[3], @y[1] |
| 537 | veor @y[4], @y[4], @t[1] |
| 538 | vext.8 @t[1], @t[1], @t[1], #12 |
| 539 | |
| 540 | veor @y[7], @y[7], @t[7] |
| 541 | veor @y[4], @y[4], @t[2] |
| 542 | veor @y[5], @y[5], @t[2] |
| 543 | veor @y[2], @y[2], @t[6] |
| 544 | veor @t[6], @t[6], @t[3] @ clobber t[6] |
| 545 | vext.8 @t[2], @t[2], @t[2], #12 |
| 546 | veor @y[4], @y[4], @y[7] |
| 547 | veor @y[3], @y[3], @t[6] |
| 548 | |
| 549 | veor @y[6], @y[6], @t[6] |
| 550 | veor @y[5], @y[5], @t[5] |
| 551 | vext.8 @t[5], @t[5], @t[5], #12 |
| 552 | veor @y[6], @y[6], @t[4] |
| 553 | vext.8 @t[4], @t[4], @t[4], #12 |
| 554 | veor @y[5], @y[5], @t[6] |
| 555 | veor @y[6], @y[6], @t[7] |
| 556 | vext.8 @t[7], @t[7], @t[7], #12 |
| 557 | veor @t[6], @t[6], @t[3] @ restore t[6] |
| 558 | vext.8 @t[3], @t[3], @t[3], #12 |
| 559 | |
| 560 | @ multiplication by 0x09 |
| 561 | veor @y[4], @y[4], @y[1] |
| 562 | veor @t[1], @t[1], @y[1] @ t[1]=y[1] |
| 563 | veor @t[0], @t[0], @t[5] @ clobber t[0] |
| 564 | vext.8 @t[6], @t[6], @t[6], #12 |
| 565 | veor @t[1], @t[1], @t[5] |
| 566 | veor @y[3], @y[3], @t[0] |
| 567 | veor @t[0], @t[0], @y[0] @ t[0]=y[0] |
| 568 | veor @t[1], @t[1], @t[6] |
| 569 | veor @t[6], @t[6], @t[7] @ clobber t[6] |
| 570 | veor @y[4], @y[4], @t[1] |
| 571 | veor @y[7], @y[7], @t[4] |
| 572 | veor @y[6], @y[6], @t[3] |
| 573 | veor @y[5], @y[5], @t[2] |
| 574 | veor @t[4], @t[4], @y[4] @ t[4]=y[4] |
| 575 | veor @t[3], @t[3], @y[3] @ t[3]=y[3] |
| 576 | veor @t[5], @t[5], @y[5] @ t[5]=y[5] |
| 577 | veor @t[2], @t[2], @y[2] @ t[2]=y[2] |
| 578 | veor @t[3], @t[3], @t[7] |
| 579 | veor @XMM[5], @t[5], @t[6] |
| 580 | veor @XMM[6], @t[6], @y[6] @ t[6]=y[6] |
| 581 | veor @XMM[2], @t[2], @t[6] |
| 582 | veor @XMM[7], @t[7], @y[7] @ t[7]=y[7] |
| 583 | |
| 584 | vmov @XMM[0], @t[0] |
| 585 | vmov @XMM[1], @t[1] |
| 586 | @ vmov @XMM[2], @t[2] |
| 587 | vmov @XMM[3], @t[3] |
| 588 | vmov @XMM[4], @t[4] |
| 589 | @ vmov @XMM[5], @t[5] |
| 590 | @ vmov @XMM[6], @t[6] |
| 591 | @ vmov @XMM[7], @t[7] |
| 592 | ___ |
| 593 | } |
| 594 | |
| 595 | sub InvMixColumns { |
| 596 | my @x=@_[0..7]; |
| 597 | my @t=@_[8..15]; |
| 598 | |
| 599 | # Thanks to Jussi Kivilinna for providing pointer to |
| 600 | # |
| 601 | # | 0e 0b 0d 09 | | 02 03 01 01 | | 05 00 04 00 | |
| 602 | # | 09 0e 0b 0d | = | 01 02 03 01 | x | 00 05 00 04 | |
| 603 | # | 0d 09 0e 0b | | 01 01 02 03 | | 04 00 05 00 | |
| 604 | # | 0b 0d 09 0e | | 03 01 01 02 | | 00 04 00 05 | |
| 605 | |
| 606 | $code.=<<___; |
| 607 | @ multiplication by 0x05-0x00-0x04-0x00 |
| 608 | vext.8 @t[0], @x[0], @x[0], #8 |
| 609 | vext.8 @t[6], @x[6], @x[6], #8 |
| 610 | vext.8 @t[7], @x[7], @x[7], #8 |
| 611 | veor @t[0], @t[0], @x[0] |
| 612 | vext.8 @t[1], @x[1], @x[1], #8 |
| 613 | veor @t[6], @t[6], @x[6] |
| 614 | vext.8 @t[2], @x[2], @x[2], #8 |
| 615 | veor @t[7], @t[7], @x[7] |
| 616 | vext.8 @t[3], @x[3], @x[3], #8 |
| 617 | veor @t[1], @t[1], @x[1] |
| 618 | vext.8 @t[4], @x[4], @x[4], #8 |
| 619 | veor @t[2], @t[2], @x[2] |
| 620 | vext.8 @t[5], @x[5], @x[5], #8 |
| 621 | veor @t[3], @t[3], @x[3] |
| 622 | veor @t[4], @t[4], @x[4] |
| 623 | veor @t[5], @t[5], @x[5] |
| 624 | |
| 625 | veor @x[0], @x[0], @t[6] |
| 626 | veor @x[1], @x[1], @t[6] |
| 627 | veor @x[2], @x[2], @t[0] |
| 628 | veor @x[4], @x[4], @t[2] |
| 629 | veor @x[3], @x[3], @t[1] |
| 630 | veor @x[1], @x[1], @t[7] |
| 631 | veor @x[2], @x[2], @t[7] |
| 632 | veor @x[4], @x[4], @t[6] |
| 633 | veor @x[5], @x[5], @t[3] |
| 634 | veor @x[3], @x[3], @t[6] |
| 635 | veor @x[6], @x[6], @t[4] |
| 636 | veor @x[4], @x[4], @t[7] |
| 637 | veor @x[5], @x[5], @t[7] |
| 638 | veor @x[7], @x[7], @t[5] |
| 639 | ___ |
| 640 | &MixColumns (@x,@t,1); # flipped 2<->3 and 4<->6 |
| 641 | } |
| 642 | |
| 643 | sub swapmove { |
| 644 | my ($a,$b,$n,$mask,$t)=@_; |
| 645 | $code.=<<___; |
| 646 | vshr.u64 $t, $b, #$n |
| 647 | veor $t, $t, $a |
| 648 | vand $t, $t, $mask |
| 649 | veor $a, $a, $t |
| 650 | vshl.u64 $t, $t, #$n |
| 651 | veor $b, $b, $t |
| 652 | ___ |
| 653 | } |
| 654 | sub swapmove2x { |
| 655 | my ($a0,$b0,$a1,$b1,$n,$mask,$t0,$t1)=@_; |
| 656 | $code.=<<___; |
| 657 | vshr.u64 $t0, $b0, #$n |
| 658 | vshr.u64 $t1, $b1, #$n |
| 659 | veor $t0, $t0, $a0 |
| 660 | veor $t1, $t1, $a1 |
| 661 | vand $t0, $t0, $mask |
| 662 | vand $t1, $t1, $mask |
| 663 | veor $a0, $a0, $t0 |
| 664 | vshl.u64 $t0, $t0, #$n |
| 665 | veor $a1, $a1, $t1 |
| 666 | vshl.u64 $t1, $t1, #$n |
| 667 | veor $b0, $b0, $t0 |
| 668 | veor $b1, $b1, $t1 |
| 669 | ___ |
| 670 | } |
| 671 | |
| 672 | sub bitslice { |
| 673 | my @x=reverse(@_[0..7]); |
| 674 | my ($t0,$t1,$t2,$t3)=@_[8..11]; |
| 675 | $code.=<<___; |
| 676 | vmov.i8 $t0,#0x55 @ compose .LBS0 |
| 677 | vmov.i8 $t1,#0x33 @ compose .LBS1 |
| 678 | ___ |
| 679 | &swapmove2x(@x[0,1,2,3],1,$t0,$t2,$t3); |
| 680 | &swapmove2x(@x[4,5,6,7],1,$t0,$t2,$t3); |
| 681 | $code.=<<___; |
| 682 | vmov.i8 $t0,#0x0f @ compose .LBS2 |
| 683 | ___ |
| 684 | &swapmove2x(@x[0,2,1,3],2,$t1,$t2,$t3); |
| 685 | &swapmove2x(@x[4,6,5,7],2,$t1,$t2,$t3); |
| 686 | |
| 687 | &swapmove2x(@x[0,4,1,5],4,$t0,$t2,$t3); |
| 688 | &swapmove2x(@x[2,6,3,7],4,$t0,$t2,$t3); |
| 689 | } |
| 690 | |
| 691 | $code.=<<___; |
| 692 | #ifndef __KERNEL__ |
| 693 | # include "arm_arch.h" |
| 694 | |
| 695 | # define VFP_ABI_PUSH vstmdb sp!,{d8-d15} |
| 696 | # define VFP_ABI_POP vldmia sp!,{d8-d15} |
| 697 | # define VFP_ABI_FRAME 0x40 |
| 698 | #else |
| 699 | # define VFP_ABI_PUSH |
| 700 | # define VFP_ABI_POP |
| 701 | # define VFP_ABI_FRAME 0 |
| 702 | # define BSAES_ASM_EXTENDED_KEY |
| 703 | # define XTS_CHAIN_TWEAK |
Ard Biesheuvel | 001eabf | 2015-02-26 07:22:05 +0000 | [diff] [blame] | 704 | # define __ARM_ARCH__ __LINUX_ARM_ARCH__ |
| 705 | # define __ARM_MAX_ARCH__ 7 |
Ard Biesheuvel | e4e7f10 | 2013-09-16 18:31:38 +0200 | [diff] [blame] | 706 | #endif |
| 707 | |
| 708 | #ifdef __thumb__ |
| 709 | # define adrl adr |
| 710 | #endif |
| 711 | |
Ard Biesheuvel | 001eabf | 2015-02-26 07:22:05 +0000 | [diff] [blame] | 712 | #if __ARM_MAX_ARCH__>=7 |
| 713 | .arch armv7-a |
| 714 | .fpu neon |
| 715 | |
Ard Biesheuvel | e4e7f10 | 2013-09-16 18:31:38 +0200 | [diff] [blame] | 716 | .text |
| 717 | .syntax unified @ ARMv7-capable assembler is expected to handle this |
| 718 | #ifdef __thumb2__ |
| 719 | .thumb |
| 720 | #else |
| 721 | .code 32 |
| 722 | #endif |
| 723 | |
Ard Biesheuvel | e4e7f10 | 2013-09-16 18:31:38 +0200 | [diff] [blame] | 724 | .type _bsaes_decrypt8,%function |
| 725 | .align 4 |
| 726 | _bsaes_decrypt8: |
| 727 | adr $const,_bsaes_decrypt8 |
| 728 | vldmia $key!, {@XMM[9]} @ round 0 key |
| 729 | add $const,$const,#.LM0ISR-_bsaes_decrypt8 |
| 730 | |
| 731 | vldmia $const!, {@XMM[8]} @ .LM0ISR |
| 732 | veor @XMM[10], @XMM[0], @XMM[9] @ xor with round0 key |
| 733 | veor @XMM[11], @XMM[1], @XMM[9] |
| 734 | vtbl.8 `&Dlo(@XMM[0])`, {@XMM[10]}, `&Dlo(@XMM[8])` |
| 735 | vtbl.8 `&Dhi(@XMM[0])`, {@XMM[10]}, `&Dhi(@XMM[8])` |
| 736 | veor @XMM[12], @XMM[2], @XMM[9] |
| 737 | vtbl.8 `&Dlo(@XMM[1])`, {@XMM[11]}, `&Dlo(@XMM[8])` |
| 738 | vtbl.8 `&Dhi(@XMM[1])`, {@XMM[11]}, `&Dhi(@XMM[8])` |
| 739 | veor @XMM[13], @XMM[3], @XMM[9] |
| 740 | vtbl.8 `&Dlo(@XMM[2])`, {@XMM[12]}, `&Dlo(@XMM[8])` |
| 741 | vtbl.8 `&Dhi(@XMM[2])`, {@XMM[12]}, `&Dhi(@XMM[8])` |
| 742 | veor @XMM[14], @XMM[4], @XMM[9] |
| 743 | vtbl.8 `&Dlo(@XMM[3])`, {@XMM[13]}, `&Dlo(@XMM[8])` |
| 744 | vtbl.8 `&Dhi(@XMM[3])`, {@XMM[13]}, `&Dhi(@XMM[8])` |
| 745 | veor @XMM[15], @XMM[5], @XMM[9] |
| 746 | vtbl.8 `&Dlo(@XMM[4])`, {@XMM[14]}, `&Dlo(@XMM[8])` |
| 747 | vtbl.8 `&Dhi(@XMM[4])`, {@XMM[14]}, `&Dhi(@XMM[8])` |
| 748 | veor @XMM[10], @XMM[6], @XMM[9] |
| 749 | vtbl.8 `&Dlo(@XMM[5])`, {@XMM[15]}, `&Dlo(@XMM[8])` |
| 750 | vtbl.8 `&Dhi(@XMM[5])`, {@XMM[15]}, `&Dhi(@XMM[8])` |
| 751 | veor @XMM[11], @XMM[7], @XMM[9] |
| 752 | vtbl.8 `&Dlo(@XMM[6])`, {@XMM[10]}, `&Dlo(@XMM[8])` |
| 753 | vtbl.8 `&Dhi(@XMM[6])`, {@XMM[10]}, `&Dhi(@XMM[8])` |
| 754 | vtbl.8 `&Dlo(@XMM[7])`, {@XMM[11]}, `&Dlo(@XMM[8])` |
| 755 | vtbl.8 `&Dhi(@XMM[7])`, {@XMM[11]}, `&Dhi(@XMM[8])` |
| 756 | ___ |
| 757 | &bitslice (@XMM[0..7, 8..11]); |
| 758 | $code.=<<___; |
| 759 | sub $rounds,$rounds,#1 |
| 760 | b .Ldec_sbox |
| 761 | .align 4 |
| 762 | .Ldec_loop: |
| 763 | ___ |
| 764 | &ShiftRows (@XMM[0..7, 8..12]); |
| 765 | $code.=".Ldec_sbox:\n"; |
| 766 | &InvSbox (@XMM[0..7, 8..15]); |
| 767 | $code.=<<___; |
| 768 | subs $rounds,$rounds,#1 |
| 769 | bcc .Ldec_done |
| 770 | ___ |
| 771 | &InvMixColumns (@XMM[0,1,6,4,2,7,3,5, 8..15]); |
| 772 | $code.=<<___; |
| 773 | vldmia $const, {@XMM[12]} @ .LISR |
| 774 | ite eq @ Thumb2 thing, sanity check in ARM |
| 775 | addeq $const,$const,#0x10 |
| 776 | bne .Ldec_loop |
| 777 | vldmia $const, {@XMM[12]} @ .LISRM0 |
| 778 | b .Ldec_loop |
| 779 | .align 4 |
| 780 | .Ldec_done: |
| 781 | ___ |
| 782 | &bitslice (@XMM[0,1,6,4,2,7,3,5, 8..11]); |
| 783 | $code.=<<___; |
| 784 | vldmia $key, {@XMM[8]} @ last round key |
| 785 | veor @XMM[6], @XMM[6], @XMM[8] |
| 786 | veor @XMM[4], @XMM[4], @XMM[8] |
| 787 | veor @XMM[2], @XMM[2], @XMM[8] |
| 788 | veor @XMM[7], @XMM[7], @XMM[8] |
| 789 | veor @XMM[3], @XMM[3], @XMM[8] |
| 790 | veor @XMM[5], @XMM[5], @XMM[8] |
| 791 | veor @XMM[0], @XMM[0], @XMM[8] |
| 792 | veor @XMM[1], @XMM[1], @XMM[8] |
| 793 | bx lr |
| 794 | .size _bsaes_decrypt8,.-_bsaes_decrypt8 |
| 795 | |
| 796 | .type _bsaes_const,%object |
| 797 | .align 6 |
| 798 | _bsaes_const: |
| 799 | .LM0ISR: @ InvShiftRows constants |
| 800 | .quad 0x0a0e0206070b0f03, 0x0004080c0d010509 |
| 801 | .LISR: |
| 802 | .quad 0x0504070602010003, 0x0f0e0d0c080b0a09 |
| 803 | .LISRM0: |
| 804 | .quad 0x01040b0e0205080f, 0x0306090c00070a0d |
| 805 | .LM0SR: @ ShiftRows constants |
| 806 | .quad 0x0a0e02060f03070b, 0x0004080c05090d01 |
| 807 | .LSR: |
| 808 | .quad 0x0504070600030201, 0x0f0e0d0c0a09080b |
| 809 | .LSRM0: |
| 810 | .quad 0x0304090e00050a0f, 0x01060b0c0207080d |
| 811 | .LM0: |
| 812 | .quad 0x02060a0e03070b0f, 0x0004080c0105090d |
| 813 | .LREVM0SR: |
| 814 | .quad 0x090d01050c000408, 0x03070b0f060a0e02 |
| 815 | .asciz "Bit-sliced AES for NEON, CRYPTOGAMS by <appro\@openssl.org>" |
| 816 | .align 6 |
| 817 | .size _bsaes_const,.-_bsaes_const |
| 818 | |
| 819 | .type _bsaes_encrypt8,%function |
| 820 | .align 4 |
| 821 | _bsaes_encrypt8: |
| 822 | adr $const,_bsaes_encrypt8 |
| 823 | vldmia $key!, {@XMM[9]} @ round 0 key |
| 824 | sub $const,$const,#_bsaes_encrypt8-.LM0SR |
| 825 | |
| 826 | vldmia $const!, {@XMM[8]} @ .LM0SR |
| 827 | _bsaes_encrypt8_alt: |
| 828 | veor @XMM[10], @XMM[0], @XMM[9] @ xor with round0 key |
| 829 | veor @XMM[11], @XMM[1], @XMM[9] |
| 830 | vtbl.8 `&Dlo(@XMM[0])`, {@XMM[10]}, `&Dlo(@XMM[8])` |
| 831 | vtbl.8 `&Dhi(@XMM[0])`, {@XMM[10]}, `&Dhi(@XMM[8])` |
| 832 | veor @XMM[12], @XMM[2], @XMM[9] |
| 833 | vtbl.8 `&Dlo(@XMM[1])`, {@XMM[11]}, `&Dlo(@XMM[8])` |
| 834 | vtbl.8 `&Dhi(@XMM[1])`, {@XMM[11]}, `&Dhi(@XMM[8])` |
| 835 | veor @XMM[13], @XMM[3], @XMM[9] |
| 836 | vtbl.8 `&Dlo(@XMM[2])`, {@XMM[12]}, `&Dlo(@XMM[8])` |
| 837 | vtbl.8 `&Dhi(@XMM[2])`, {@XMM[12]}, `&Dhi(@XMM[8])` |
| 838 | veor @XMM[14], @XMM[4], @XMM[9] |
| 839 | vtbl.8 `&Dlo(@XMM[3])`, {@XMM[13]}, `&Dlo(@XMM[8])` |
| 840 | vtbl.8 `&Dhi(@XMM[3])`, {@XMM[13]}, `&Dhi(@XMM[8])` |
| 841 | veor @XMM[15], @XMM[5], @XMM[9] |
| 842 | vtbl.8 `&Dlo(@XMM[4])`, {@XMM[14]}, `&Dlo(@XMM[8])` |
| 843 | vtbl.8 `&Dhi(@XMM[4])`, {@XMM[14]}, `&Dhi(@XMM[8])` |
| 844 | veor @XMM[10], @XMM[6], @XMM[9] |
| 845 | vtbl.8 `&Dlo(@XMM[5])`, {@XMM[15]}, `&Dlo(@XMM[8])` |
| 846 | vtbl.8 `&Dhi(@XMM[5])`, {@XMM[15]}, `&Dhi(@XMM[8])` |
| 847 | veor @XMM[11], @XMM[7], @XMM[9] |
| 848 | vtbl.8 `&Dlo(@XMM[6])`, {@XMM[10]}, `&Dlo(@XMM[8])` |
| 849 | vtbl.8 `&Dhi(@XMM[6])`, {@XMM[10]}, `&Dhi(@XMM[8])` |
| 850 | vtbl.8 `&Dlo(@XMM[7])`, {@XMM[11]}, `&Dlo(@XMM[8])` |
| 851 | vtbl.8 `&Dhi(@XMM[7])`, {@XMM[11]}, `&Dhi(@XMM[8])` |
| 852 | _bsaes_encrypt8_bitslice: |
| 853 | ___ |
| 854 | &bitslice (@XMM[0..7, 8..11]); |
| 855 | $code.=<<___; |
| 856 | sub $rounds,$rounds,#1 |
| 857 | b .Lenc_sbox |
| 858 | .align 4 |
| 859 | .Lenc_loop: |
| 860 | ___ |
| 861 | &ShiftRows (@XMM[0..7, 8..12]); |
| 862 | $code.=".Lenc_sbox:\n"; |
| 863 | &Sbox (@XMM[0..7, 8..15]); |
| 864 | $code.=<<___; |
| 865 | subs $rounds,$rounds,#1 |
| 866 | bcc .Lenc_done |
| 867 | ___ |
| 868 | &MixColumns (@XMM[0,1,4,6,3,7,2,5, 8..15]); |
| 869 | $code.=<<___; |
| 870 | vldmia $const, {@XMM[12]} @ .LSR |
| 871 | ite eq @ Thumb2 thing, samity check in ARM |
| 872 | addeq $const,$const,#0x10 |
| 873 | bne .Lenc_loop |
| 874 | vldmia $const, {@XMM[12]} @ .LSRM0 |
| 875 | b .Lenc_loop |
| 876 | .align 4 |
| 877 | .Lenc_done: |
| 878 | ___ |
| 879 | # output in lsb > [t0, t1, t4, t6, t3, t7, t2, t5] < msb |
| 880 | &bitslice (@XMM[0,1,4,6,3,7,2,5, 8..11]); |
| 881 | $code.=<<___; |
| 882 | vldmia $key, {@XMM[8]} @ last round key |
| 883 | veor @XMM[4], @XMM[4], @XMM[8] |
| 884 | veor @XMM[6], @XMM[6], @XMM[8] |
| 885 | veor @XMM[3], @XMM[3], @XMM[8] |
| 886 | veor @XMM[7], @XMM[7], @XMM[8] |
| 887 | veor @XMM[2], @XMM[2], @XMM[8] |
| 888 | veor @XMM[5], @XMM[5], @XMM[8] |
| 889 | veor @XMM[0], @XMM[0], @XMM[8] |
| 890 | veor @XMM[1], @XMM[1], @XMM[8] |
| 891 | bx lr |
| 892 | .size _bsaes_encrypt8,.-_bsaes_encrypt8 |
| 893 | ___ |
| 894 | } |
| 895 | { |
| 896 | my ($out,$inp,$rounds,$const)=("r12","r4","r5","r6"); |
| 897 | |
| 898 | sub bitslice_key { |
| 899 | my @x=reverse(@_[0..7]); |
| 900 | my ($bs0,$bs1,$bs2,$t2,$t3)=@_[8..12]; |
| 901 | |
| 902 | &swapmove (@x[0,1],1,$bs0,$t2,$t3); |
| 903 | $code.=<<___; |
| 904 | @ &swapmove(@x[2,3],1,$t0,$t2,$t3); |
| 905 | vmov @x[2], @x[0] |
| 906 | vmov @x[3], @x[1] |
| 907 | ___ |
| 908 | #&swapmove2x(@x[4,5,6,7],1,$t0,$t2,$t3); |
| 909 | |
| 910 | &swapmove2x (@x[0,2,1,3],2,$bs1,$t2,$t3); |
| 911 | $code.=<<___; |
| 912 | @ &swapmove2x(@x[4,6,5,7],2,$t1,$t2,$t3); |
| 913 | vmov @x[4], @x[0] |
| 914 | vmov @x[6], @x[2] |
| 915 | vmov @x[5], @x[1] |
| 916 | vmov @x[7], @x[3] |
| 917 | ___ |
| 918 | &swapmove2x (@x[0,4,1,5],4,$bs2,$t2,$t3); |
| 919 | &swapmove2x (@x[2,6,3,7],4,$bs2,$t2,$t3); |
| 920 | } |
| 921 | |
| 922 | $code.=<<___; |
| 923 | .type _bsaes_key_convert,%function |
| 924 | .align 4 |
| 925 | _bsaes_key_convert: |
| 926 | adr $const,_bsaes_key_convert |
| 927 | vld1.8 {@XMM[7]}, [$inp]! @ load round 0 key |
| 928 | sub $const,$const,#_bsaes_key_convert-.LM0 |
| 929 | vld1.8 {@XMM[15]}, [$inp]! @ load round 1 key |
| 930 | |
| 931 | vmov.i8 @XMM[8], #0x01 @ bit masks |
| 932 | vmov.i8 @XMM[9], #0x02 |
| 933 | vmov.i8 @XMM[10], #0x04 |
| 934 | vmov.i8 @XMM[11], #0x08 |
| 935 | vmov.i8 @XMM[12], #0x10 |
| 936 | vmov.i8 @XMM[13], #0x20 |
| 937 | vldmia $const, {@XMM[14]} @ .LM0 |
| 938 | |
| 939 | #ifdef __ARMEL__ |
| 940 | vrev32.8 @XMM[7], @XMM[7] |
| 941 | vrev32.8 @XMM[15], @XMM[15] |
| 942 | #endif |
| 943 | sub $rounds,$rounds,#1 |
| 944 | vstmia $out!, {@XMM[7]} @ save round 0 key |
| 945 | b .Lkey_loop |
| 946 | |
| 947 | .align 4 |
| 948 | .Lkey_loop: |
| 949 | vtbl.8 `&Dlo(@XMM[7])`,{@XMM[15]},`&Dlo(@XMM[14])` |
| 950 | vtbl.8 `&Dhi(@XMM[7])`,{@XMM[15]},`&Dhi(@XMM[14])` |
| 951 | vmov.i8 @XMM[6], #0x40 |
| 952 | vmov.i8 @XMM[15], #0x80 |
| 953 | |
| 954 | vtst.8 @XMM[0], @XMM[7], @XMM[8] |
| 955 | vtst.8 @XMM[1], @XMM[7], @XMM[9] |
| 956 | vtst.8 @XMM[2], @XMM[7], @XMM[10] |
| 957 | vtst.8 @XMM[3], @XMM[7], @XMM[11] |
| 958 | vtst.8 @XMM[4], @XMM[7], @XMM[12] |
| 959 | vtst.8 @XMM[5], @XMM[7], @XMM[13] |
| 960 | vtst.8 @XMM[6], @XMM[7], @XMM[6] |
| 961 | vtst.8 @XMM[7], @XMM[7], @XMM[15] |
| 962 | vld1.8 {@XMM[15]}, [$inp]! @ load next round key |
| 963 | vmvn @XMM[0], @XMM[0] @ "pnot" |
| 964 | vmvn @XMM[1], @XMM[1] |
| 965 | vmvn @XMM[5], @XMM[5] |
| 966 | vmvn @XMM[6], @XMM[6] |
| 967 | #ifdef __ARMEL__ |
| 968 | vrev32.8 @XMM[15], @XMM[15] |
| 969 | #endif |
| 970 | subs $rounds,$rounds,#1 |
| 971 | vstmia $out!,{@XMM[0]-@XMM[7]} @ write bit-sliced round key |
| 972 | bne .Lkey_loop |
| 973 | |
| 974 | vmov.i8 @XMM[7],#0x63 @ compose .L63 |
| 975 | @ don't save last round key |
| 976 | bx lr |
| 977 | .size _bsaes_key_convert,.-_bsaes_key_convert |
| 978 | ___ |
| 979 | } |
| 980 | |
| 981 | if (0) { # following four functions are unsupported interface |
| 982 | # used for benchmarking... |
| 983 | $code.=<<___; |
| 984 | .globl bsaes_enc_key_convert |
| 985 | .type bsaes_enc_key_convert,%function |
| 986 | .align 4 |
| 987 | bsaes_enc_key_convert: |
| 988 | stmdb sp!,{r4-r6,lr} |
| 989 | vstmdb sp!,{d8-d15} @ ABI specification says so |
| 990 | |
| 991 | ldr r5,[$inp,#240] @ pass rounds |
| 992 | mov r4,$inp @ pass key |
| 993 | mov r12,$out @ pass key schedule |
| 994 | bl _bsaes_key_convert |
| 995 | veor @XMM[7],@XMM[7],@XMM[15] @ fix up last round key |
| 996 | vstmia r12, {@XMM[7]} @ save last round key |
| 997 | |
| 998 | vldmia sp!,{d8-d15} |
| 999 | ldmia sp!,{r4-r6,pc} |
| 1000 | .size bsaes_enc_key_convert,.-bsaes_enc_key_convert |
| 1001 | |
| 1002 | .globl bsaes_encrypt_128 |
| 1003 | .type bsaes_encrypt_128,%function |
| 1004 | .align 4 |
| 1005 | bsaes_encrypt_128: |
| 1006 | stmdb sp!,{r4-r6,lr} |
| 1007 | vstmdb sp!,{d8-d15} @ ABI specification says so |
| 1008 | .Lenc128_loop: |
| 1009 | vld1.8 {@XMM[0]-@XMM[1]}, [$inp]! @ load input |
| 1010 | vld1.8 {@XMM[2]-@XMM[3]}, [$inp]! |
| 1011 | mov r4,$key @ pass the key |
| 1012 | vld1.8 {@XMM[4]-@XMM[5]}, [$inp]! |
| 1013 | mov r5,#10 @ pass rounds |
| 1014 | vld1.8 {@XMM[6]-@XMM[7]}, [$inp]! |
| 1015 | |
| 1016 | bl _bsaes_encrypt8 |
| 1017 | |
| 1018 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1019 | vst1.8 {@XMM[4]}, [$out]! |
| 1020 | vst1.8 {@XMM[6]}, [$out]! |
| 1021 | vst1.8 {@XMM[3]}, [$out]! |
| 1022 | vst1.8 {@XMM[7]}, [$out]! |
| 1023 | vst1.8 {@XMM[2]}, [$out]! |
| 1024 | subs $len,$len,#0x80 |
| 1025 | vst1.8 {@XMM[5]}, [$out]! |
| 1026 | bhi .Lenc128_loop |
| 1027 | |
| 1028 | vldmia sp!,{d8-d15} |
| 1029 | ldmia sp!,{r4-r6,pc} |
| 1030 | .size bsaes_encrypt_128,.-bsaes_encrypt_128 |
| 1031 | |
| 1032 | .globl bsaes_dec_key_convert |
| 1033 | .type bsaes_dec_key_convert,%function |
| 1034 | .align 4 |
| 1035 | bsaes_dec_key_convert: |
| 1036 | stmdb sp!,{r4-r6,lr} |
| 1037 | vstmdb sp!,{d8-d15} @ ABI specification says so |
| 1038 | |
| 1039 | ldr r5,[$inp,#240] @ pass rounds |
| 1040 | mov r4,$inp @ pass key |
| 1041 | mov r12,$out @ pass key schedule |
| 1042 | bl _bsaes_key_convert |
| 1043 | vldmia $out, {@XMM[6]} |
| 1044 | vstmia r12, {@XMM[15]} @ save last round key |
| 1045 | veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key |
| 1046 | vstmia $out, {@XMM[7]} |
| 1047 | |
| 1048 | vldmia sp!,{d8-d15} |
| 1049 | ldmia sp!,{r4-r6,pc} |
| 1050 | .size bsaes_dec_key_convert,.-bsaes_dec_key_convert |
| 1051 | |
| 1052 | .globl bsaes_decrypt_128 |
| 1053 | .type bsaes_decrypt_128,%function |
| 1054 | .align 4 |
| 1055 | bsaes_decrypt_128: |
| 1056 | stmdb sp!,{r4-r6,lr} |
| 1057 | vstmdb sp!,{d8-d15} @ ABI specification says so |
| 1058 | .Ldec128_loop: |
| 1059 | vld1.8 {@XMM[0]-@XMM[1]}, [$inp]! @ load input |
| 1060 | vld1.8 {@XMM[2]-@XMM[3]}, [$inp]! |
| 1061 | mov r4,$key @ pass the key |
| 1062 | vld1.8 {@XMM[4]-@XMM[5]}, [$inp]! |
| 1063 | mov r5,#10 @ pass rounds |
| 1064 | vld1.8 {@XMM[6]-@XMM[7]}, [$inp]! |
| 1065 | |
| 1066 | bl _bsaes_decrypt8 |
| 1067 | |
| 1068 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1069 | vst1.8 {@XMM[6]}, [$out]! |
| 1070 | vst1.8 {@XMM[4]}, [$out]! |
| 1071 | vst1.8 {@XMM[2]}, [$out]! |
| 1072 | vst1.8 {@XMM[7]}, [$out]! |
| 1073 | vst1.8 {@XMM[3]}, [$out]! |
| 1074 | subs $len,$len,#0x80 |
| 1075 | vst1.8 {@XMM[5]}, [$out]! |
| 1076 | bhi .Ldec128_loop |
| 1077 | |
| 1078 | vldmia sp!,{d8-d15} |
| 1079 | ldmia sp!,{r4-r6,pc} |
| 1080 | .size bsaes_decrypt_128,.-bsaes_decrypt_128 |
| 1081 | ___ |
| 1082 | } |
| 1083 | { |
| 1084 | my ($inp,$out,$len,$key, $ivp,$fp,$rounds)=map("r$_",(0..3,8..10)); |
| 1085 | my ($keysched)=("sp"); |
| 1086 | |
| 1087 | $code.=<<___; |
| 1088 | .extern AES_cbc_encrypt |
| 1089 | .extern AES_decrypt |
| 1090 | |
| 1091 | .global bsaes_cbc_encrypt |
| 1092 | .type bsaes_cbc_encrypt,%function |
| 1093 | .align 5 |
| 1094 | bsaes_cbc_encrypt: |
| 1095 | #ifndef __KERNEL__ |
| 1096 | cmp $len, #128 |
| 1097 | #ifndef __thumb__ |
| 1098 | blo AES_cbc_encrypt |
| 1099 | #else |
| 1100 | bhs 1f |
| 1101 | b AES_cbc_encrypt |
| 1102 | 1: |
| 1103 | #endif |
| 1104 | #endif |
| 1105 | |
| 1106 | @ it is up to the caller to make sure we are called with enc == 0 |
| 1107 | |
| 1108 | mov ip, sp |
| 1109 | stmdb sp!, {r4-r10, lr} |
| 1110 | VFP_ABI_PUSH |
| 1111 | ldr $ivp, [ip] @ IV is 1st arg on the stack |
| 1112 | mov $len, $len, lsr#4 @ len in 16 byte blocks |
| 1113 | sub sp, #0x10 @ scratch space to carry over the IV |
| 1114 | mov $fp, sp @ save sp |
| 1115 | |
| 1116 | ldr $rounds, [$key, #240] @ get # of rounds |
| 1117 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1118 | @ allocate the key schedule on the stack |
| 1119 | sub r12, sp, $rounds, lsl#7 @ 128 bytes per inner round key |
| 1120 | add r12, #`128-32` @ sifze of bit-slices key schedule |
| 1121 | |
| 1122 | @ populate the key schedule |
| 1123 | mov r4, $key @ pass key |
| 1124 | mov r5, $rounds @ pass # of rounds |
| 1125 | mov sp, r12 @ sp is $keysched |
| 1126 | bl _bsaes_key_convert |
| 1127 | vldmia $keysched, {@XMM[6]} |
| 1128 | vstmia r12, {@XMM[15]} @ save last round key |
| 1129 | veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key |
| 1130 | vstmia $keysched, {@XMM[7]} |
| 1131 | #else |
| 1132 | ldr r12, [$key, #244] |
| 1133 | eors r12, #1 |
| 1134 | beq 0f |
| 1135 | |
| 1136 | @ populate the key schedule |
| 1137 | str r12, [$key, #244] |
| 1138 | mov r4, $key @ pass key |
| 1139 | mov r5, $rounds @ pass # of rounds |
| 1140 | add r12, $key, #248 @ pass key schedule |
| 1141 | bl _bsaes_key_convert |
| 1142 | add r4, $key, #248 |
| 1143 | vldmia r4, {@XMM[6]} |
| 1144 | vstmia r12, {@XMM[15]} @ save last round key |
| 1145 | veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key |
| 1146 | vstmia r4, {@XMM[7]} |
| 1147 | |
| 1148 | .align 2 |
| 1149 | 0: |
| 1150 | #endif |
| 1151 | |
| 1152 | vld1.8 {@XMM[15]}, [$ivp] @ load IV |
| 1153 | b .Lcbc_dec_loop |
| 1154 | |
| 1155 | .align 4 |
| 1156 | .Lcbc_dec_loop: |
| 1157 | subs $len, $len, #0x8 |
| 1158 | bmi .Lcbc_dec_loop_finish |
| 1159 | |
| 1160 | vld1.8 {@XMM[0]-@XMM[1]}, [$inp]! @ load input |
| 1161 | vld1.8 {@XMM[2]-@XMM[3]}, [$inp]! |
| 1162 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1163 | mov r4, $keysched @ pass the key |
| 1164 | #else |
| 1165 | add r4, $key, #248 |
| 1166 | #endif |
| 1167 | vld1.8 {@XMM[4]-@XMM[5]}, [$inp]! |
| 1168 | mov r5, $rounds |
| 1169 | vld1.8 {@XMM[6]-@XMM[7]}, [$inp] |
| 1170 | sub $inp, $inp, #0x60 |
| 1171 | vstmia $fp, {@XMM[15]} @ put aside IV |
| 1172 | |
| 1173 | bl _bsaes_decrypt8 |
| 1174 | |
| 1175 | vldmia $fp, {@XMM[14]} @ reload IV |
| 1176 | vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input |
| 1177 | veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV |
| 1178 | vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! |
| 1179 | veor @XMM[1], @XMM[1], @XMM[8] |
| 1180 | veor @XMM[6], @XMM[6], @XMM[9] |
| 1181 | vld1.8 {@XMM[12]-@XMM[13]}, [$inp]! |
| 1182 | veor @XMM[4], @XMM[4], @XMM[10] |
| 1183 | veor @XMM[2], @XMM[2], @XMM[11] |
| 1184 | vld1.8 {@XMM[14]-@XMM[15]}, [$inp]! |
| 1185 | veor @XMM[7], @XMM[7], @XMM[12] |
| 1186 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1187 | veor @XMM[3], @XMM[3], @XMM[13] |
| 1188 | vst1.8 {@XMM[6]}, [$out]! |
| 1189 | veor @XMM[5], @XMM[5], @XMM[14] |
| 1190 | vst1.8 {@XMM[4]}, [$out]! |
| 1191 | vst1.8 {@XMM[2]}, [$out]! |
| 1192 | vst1.8 {@XMM[7]}, [$out]! |
| 1193 | vst1.8 {@XMM[3]}, [$out]! |
| 1194 | vst1.8 {@XMM[5]}, [$out]! |
| 1195 | |
| 1196 | b .Lcbc_dec_loop |
| 1197 | |
| 1198 | .Lcbc_dec_loop_finish: |
| 1199 | adds $len, $len, #8 |
| 1200 | beq .Lcbc_dec_done |
| 1201 | |
| 1202 | vld1.8 {@XMM[0]}, [$inp]! @ load input |
| 1203 | cmp $len, #2 |
| 1204 | blo .Lcbc_dec_one |
| 1205 | vld1.8 {@XMM[1]}, [$inp]! |
| 1206 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1207 | mov r4, $keysched @ pass the key |
| 1208 | #else |
| 1209 | add r4, $key, #248 |
| 1210 | #endif |
| 1211 | mov r5, $rounds |
| 1212 | vstmia $fp, {@XMM[15]} @ put aside IV |
| 1213 | beq .Lcbc_dec_two |
| 1214 | vld1.8 {@XMM[2]}, [$inp]! |
| 1215 | cmp $len, #4 |
| 1216 | blo .Lcbc_dec_three |
| 1217 | vld1.8 {@XMM[3]}, [$inp]! |
| 1218 | beq .Lcbc_dec_four |
| 1219 | vld1.8 {@XMM[4]}, [$inp]! |
| 1220 | cmp $len, #6 |
| 1221 | blo .Lcbc_dec_five |
| 1222 | vld1.8 {@XMM[5]}, [$inp]! |
| 1223 | beq .Lcbc_dec_six |
| 1224 | vld1.8 {@XMM[6]}, [$inp]! |
| 1225 | sub $inp, $inp, #0x70 |
| 1226 | |
| 1227 | bl _bsaes_decrypt8 |
| 1228 | |
| 1229 | vldmia $fp, {@XMM[14]} @ reload IV |
| 1230 | vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input |
| 1231 | veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV |
| 1232 | vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! |
| 1233 | veor @XMM[1], @XMM[1], @XMM[8] |
| 1234 | veor @XMM[6], @XMM[6], @XMM[9] |
| 1235 | vld1.8 {@XMM[12]-@XMM[13]}, [$inp]! |
| 1236 | veor @XMM[4], @XMM[4], @XMM[10] |
| 1237 | veor @XMM[2], @XMM[2], @XMM[11] |
| 1238 | vld1.8 {@XMM[15]}, [$inp]! |
| 1239 | veor @XMM[7], @XMM[7], @XMM[12] |
| 1240 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1241 | veor @XMM[3], @XMM[3], @XMM[13] |
| 1242 | vst1.8 {@XMM[6]}, [$out]! |
| 1243 | vst1.8 {@XMM[4]}, [$out]! |
| 1244 | vst1.8 {@XMM[2]}, [$out]! |
| 1245 | vst1.8 {@XMM[7]}, [$out]! |
| 1246 | vst1.8 {@XMM[3]}, [$out]! |
| 1247 | b .Lcbc_dec_done |
| 1248 | .align 4 |
| 1249 | .Lcbc_dec_six: |
| 1250 | sub $inp, $inp, #0x60 |
| 1251 | bl _bsaes_decrypt8 |
| 1252 | vldmia $fp,{@XMM[14]} @ reload IV |
| 1253 | vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input |
| 1254 | veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV |
| 1255 | vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! |
| 1256 | veor @XMM[1], @XMM[1], @XMM[8] |
| 1257 | veor @XMM[6], @XMM[6], @XMM[9] |
| 1258 | vld1.8 {@XMM[12]}, [$inp]! |
| 1259 | veor @XMM[4], @XMM[4], @XMM[10] |
| 1260 | veor @XMM[2], @XMM[2], @XMM[11] |
| 1261 | vld1.8 {@XMM[15]}, [$inp]! |
| 1262 | veor @XMM[7], @XMM[7], @XMM[12] |
| 1263 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1264 | vst1.8 {@XMM[6]}, [$out]! |
| 1265 | vst1.8 {@XMM[4]}, [$out]! |
| 1266 | vst1.8 {@XMM[2]}, [$out]! |
| 1267 | vst1.8 {@XMM[7]}, [$out]! |
| 1268 | b .Lcbc_dec_done |
| 1269 | .align 4 |
| 1270 | .Lcbc_dec_five: |
| 1271 | sub $inp, $inp, #0x50 |
| 1272 | bl _bsaes_decrypt8 |
| 1273 | vldmia $fp, {@XMM[14]} @ reload IV |
| 1274 | vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input |
| 1275 | veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV |
| 1276 | vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! |
| 1277 | veor @XMM[1], @XMM[1], @XMM[8] |
| 1278 | veor @XMM[6], @XMM[6], @XMM[9] |
| 1279 | vld1.8 {@XMM[15]}, [$inp]! |
| 1280 | veor @XMM[4], @XMM[4], @XMM[10] |
| 1281 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1282 | veor @XMM[2], @XMM[2], @XMM[11] |
| 1283 | vst1.8 {@XMM[6]}, [$out]! |
| 1284 | vst1.8 {@XMM[4]}, [$out]! |
| 1285 | vst1.8 {@XMM[2]}, [$out]! |
| 1286 | b .Lcbc_dec_done |
| 1287 | .align 4 |
| 1288 | .Lcbc_dec_four: |
| 1289 | sub $inp, $inp, #0x40 |
| 1290 | bl _bsaes_decrypt8 |
| 1291 | vldmia $fp, {@XMM[14]} @ reload IV |
| 1292 | vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input |
| 1293 | veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV |
| 1294 | vld1.8 {@XMM[10]}, [$inp]! |
| 1295 | veor @XMM[1], @XMM[1], @XMM[8] |
| 1296 | veor @XMM[6], @XMM[6], @XMM[9] |
| 1297 | vld1.8 {@XMM[15]}, [$inp]! |
| 1298 | veor @XMM[4], @XMM[4], @XMM[10] |
| 1299 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1300 | vst1.8 {@XMM[6]}, [$out]! |
| 1301 | vst1.8 {@XMM[4]}, [$out]! |
| 1302 | b .Lcbc_dec_done |
| 1303 | .align 4 |
| 1304 | .Lcbc_dec_three: |
| 1305 | sub $inp, $inp, #0x30 |
| 1306 | bl _bsaes_decrypt8 |
| 1307 | vldmia $fp, {@XMM[14]} @ reload IV |
| 1308 | vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ reload input |
| 1309 | veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV |
| 1310 | vld1.8 {@XMM[15]}, [$inp]! |
| 1311 | veor @XMM[1], @XMM[1], @XMM[8] |
| 1312 | veor @XMM[6], @XMM[6], @XMM[9] |
| 1313 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1314 | vst1.8 {@XMM[6]}, [$out]! |
| 1315 | b .Lcbc_dec_done |
| 1316 | .align 4 |
| 1317 | .Lcbc_dec_two: |
| 1318 | sub $inp, $inp, #0x20 |
| 1319 | bl _bsaes_decrypt8 |
| 1320 | vldmia $fp, {@XMM[14]} @ reload IV |
| 1321 | vld1.8 {@XMM[8]}, [$inp]! @ reload input |
| 1322 | veor @XMM[0], @XMM[0], @XMM[14] @ ^= IV |
| 1323 | vld1.8 {@XMM[15]}, [$inp]! @ reload input |
| 1324 | veor @XMM[1], @XMM[1], @XMM[8] |
| 1325 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1326 | b .Lcbc_dec_done |
| 1327 | .align 4 |
| 1328 | .Lcbc_dec_one: |
| 1329 | sub $inp, $inp, #0x10 |
| 1330 | mov $rounds, $out @ save original out pointer |
| 1331 | mov $out, $fp @ use the iv scratch space as out buffer |
| 1332 | mov r2, $key |
| 1333 | vmov @XMM[4],@XMM[15] @ just in case ensure that IV |
| 1334 | vmov @XMM[5],@XMM[0] @ and input are preserved |
| 1335 | bl AES_decrypt |
| 1336 | vld1.8 {@XMM[0]}, [$fp,:64] @ load result |
| 1337 | veor @XMM[0], @XMM[0], @XMM[4] @ ^= IV |
| 1338 | vmov @XMM[15], @XMM[5] @ @XMM[5] holds input |
| 1339 | vst1.8 {@XMM[0]}, [$rounds] @ write output |
| 1340 | |
| 1341 | .Lcbc_dec_done: |
| 1342 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1343 | vmov.i32 q0, #0 |
| 1344 | vmov.i32 q1, #0 |
| 1345 | .Lcbc_dec_bzero: @ wipe key schedule [if any] |
| 1346 | vstmia $keysched!, {q0-q1} |
| 1347 | cmp $keysched, $fp |
| 1348 | bne .Lcbc_dec_bzero |
| 1349 | #endif |
| 1350 | |
| 1351 | mov sp, $fp |
| 1352 | add sp, #0x10 @ add sp,$fp,#0x10 is no good for thumb |
| 1353 | vst1.8 {@XMM[15]}, [$ivp] @ return IV |
| 1354 | VFP_ABI_POP |
| 1355 | ldmia sp!, {r4-r10, pc} |
| 1356 | .size bsaes_cbc_encrypt,.-bsaes_cbc_encrypt |
| 1357 | ___ |
| 1358 | } |
| 1359 | { |
| 1360 | my ($inp,$out,$len,$key, $ctr,$fp,$rounds)=(map("r$_",(0..3,8..10))); |
| 1361 | my $const = "r6"; # shared with _bsaes_encrypt8_alt |
| 1362 | my $keysched = "sp"; |
| 1363 | |
| 1364 | $code.=<<___; |
| 1365 | .extern AES_encrypt |
| 1366 | .global bsaes_ctr32_encrypt_blocks |
| 1367 | .type bsaes_ctr32_encrypt_blocks,%function |
| 1368 | .align 5 |
| 1369 | bsaes_ctr32_encrypt_blocks: |
| 1370 | cmp $len, #8 @ use plain AES for |
| 1371 | blo .Lctr_enc_short @ small sizes |
| 1372 | |
| 1373 | mov ip, sp |
| 1374 | stmdb sp!, {r4-r10, lr} |
| 1375 | VFP_ABI_PUSH |
| 1376 | ldr $ctr, [ip] @ ctr is 1st arg on the stack |
| 1377 | sub sp, sp, #0x10 @ scratch space to carry over the ctr |
| 1378 | mov $fp, sp @ save sp |
| 1379 | |
| 1380 | ldr $rounds, [$key, #240] @ get # of rounds |
| 1381 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1382 | @ allocate the key schedule on the stack |
| 1383 | sub r12, sp, $rounds, lsl#7 @ 128 bytes per inner round key |
| 1384 | add r12, #`128-32` @ size of bit-sliced key schedule |
| 1385 | |
| 1386 | @ populate the key schedule |
| 1387 | mov r4, $key @ pass key |
| 1388 | mov r5, $rounds @ pass # of rounds |
| 1389 | mov sp, r12 @ sp is $keysched |
| 1390 | bl _bsaes_key_convert |
| 1391 | veor @XMM[7],@XMM[7],@XMM[15] @ fix up last round key |
| 1392 | vstmia r12, {@XMM[7]} @ save last round key |
| 1393 | |
| 1394 | vld1.8 {@XMM[0]}, [$ctr] @ load counter |
| 1395 | add $ctr, $const, #.LREVM0SR-.LM0 @ borrow $ctr |
| 1396 | vldmia $keysched, {@XMM[4]} @ load round0 key |
| 1397 | #else |
| 1398 | ldr r12, [$key, #244] |
| 1399 | eors r12, #1 |
| 1400 | beq 0f |
| 1401 | |
| 1402 | @ populate the key schedule |
| 1403 | str r12, [$key, #244] |
| 1404 | mov r4, $key @ pass key |
| 1405 | mov r5, $rounds @ pass # of rounds |
| 1406 | add r12, $key, #248 @ pass key schedule |
| 1407 | bl _bsaes_key_convert |
| 1408 | veor @XMM[7],@XMM[7],@XMM[15] @ fix up last round key |
| 1409 | vstmia r12, {@XMM[7]} @ save last round key |
| 1410 | |
| 1411 | .align 2 |
| 1412 | 0: add r12, $key, #248 |
| 1413 | vld1.8 {@XMM[0]}, [$ctr] @ load counter |
| 1414 | adrl $ctr, .LREVM0SR @ borrow $ctr |
| 1415 | vldmia r12, {@XMM[4]} @ load round0 key |
| 1416 | sub sp, #0x10 @ place for adjusted round0 key |
| 1417 | #endif |
| 1418 | |
| 1419 | vmov.i32 @XMM[8],#1 @ compose 1<<96 |
| 1420 | veor @XMM[9],@XMM[9],@XMM[9] |
| 1421 | vrev32.8 @XMM[0],@XMM[0] |
| 1422 | vext.8 @XMM[8],@XMM[9],@XMM[8],#4 |
| 1423 | vrev32.8 @XMM[4],@XMM[4] |
| 1424 | vadd.u32 @XMM[9],@XMM[8],@XMM[8] @ compose 2<<96 |
| 1425 | vstmia $keysched, {@XMM[4]} @ save adjusted round0 key |
| 1426 | b .Lctr_enc_loop |
| 1427 | |
| 1428 | .align 4 |
| 1429 | .Lctr_enc_loop: |
| 1430 | vadd.u32 @XMM[10], @XMM[8], @XMM[9] @ compose 3<<96 |
| 1431 | vadd.u32 @XMM[1], @XMM[0], @XMM[8] @ +1 |
| 1432 | vadd.u32 @XMM[2], @XMM[0], @XMM[9] @ +2 |
| 1433 | vadd.u32 @XMM[3], @XMM[0], @XMM[10] @ +3 |
| 1434 | vadd.u32 @XMM[4], @XMM[1], @XMM[10] |
| 1435 | vadd.u32 @XMM[5], @XMM[2], @XMM[10] |
| 1436 | vadd.u32 @XMM[6], @XMM[3], @XMM[10] |
| 1437 | vadd.u32 @XMM[7], @XMM[4], @XMM[10] |
| 1438 | vadd.u32 @XMM[10], @XMM[5], @XMM[10] @ next counter |
| 1439 | |
| 1440 | @ Borrow prologue from _bsaes_encrypt8 to use the opportunity |
| 1441 | @ to flip byte order in 32-bit counter |
| 1442 | |
| 1443 | vldmia $keysched, {@XMM[9]} @ load round0 key |
| 1444 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1445 | add r4, $keysched, #0x10 @ pass next round key |
| 1446 | #else |
| 1447 | add r4, $key, #`248+16` |
| 1448 | #endif |
| 1449 | vldmia $ctr, {@XMM[8]} @ .LREVM0SR |
| 1450 | mov r5, $rounds @ pass rounds |
| 1451 | vstmia $fp, {@XMM[10]} @ save next counter |
| 1452 | sub $const, $ctr, #.LREVM0SR-.LSR @ pass constants |
| 1453 | |
| 1454 | bl _bsaes_encrypt8_alt |
| 1455 | |
| 1456 | subs $len, $len, #8 |
| 1457 | blo .Lctr_enc_loop_done |
| 1458 | |
| 1459 | vld1.8 {@XMM[8]-@XMM[9]}, [$inp]! @ load input |
| 1460 | vld1.8 {@XMM[10]-@XMM[11]}, [$inp]! |
| 1461 | veor @XMM[0], @XMM[8] |
| 1462 | veor @XMM[1], @XMM[9] |
| 1463 | vld1.8 {@XMM[12]-@XMM[13]}, [$inp]! |
| 1464 | veor @XMM[4], @XMM[10] |
| 1465 | veor @XMM[6], @XMM[11] |
| 1466 | vld1.8 {@XMM[14]-@XMM[15]}, [$inp]! |
| 1467 | veor @XMM[3], @XMM[12] |
| 1468 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! @ write output |
| 1469 | veor @XMM[7], @XMM[13] |
| 1470 | veor @XMM[2], @XMM[14] |
| 1471 | vst1.8 {@XMM[4]}, [$out]! |
| 1472 | veor @XMM[5], @XMM[15] |
| 1473 | vst1.8 {@XMM[6]}, [$out]! |
| 1474 | vmov.i32 @XMM[8], #1 @ compose 1<<96 |
| 1475 | vst1.8 {@XMM[3]}, [$out]! |
| 1476 | veor @XMM[9], @XMM[9], @XMM[9] |
| 1477 | vst1.8 {@XMM[7]}, [$out]! |
| 1478 | vext.8 @XMM[8], @XMM[9], @XMM[8], #4 |
| 1479 | vst1.8 {@XMM[2]}, [$out]! |
| 1480 | vadd.u32 @XMM[9],@XMM[8],@XMM[8] @ compose 2<<96 |
| 1481 | vst1.8 {@XMM[5]}, [$out]! |
| 1482 | vldmia $fp, {@XMM[0]} @ load counter |
| 1483 | |
| 1484 | bne .Lctr_enc_loop |
| 1485 | b .Lctr_enc_done |
| 1486 | |
| 1487 | .align 4 |
| 1488 | .Lctr_enc_loop_done: |
| 1489 | add $len, $len, #8 |
| 1490 | vld1.8 {@XMM[8]}, [$inp]! @ load input |
| 1491 | veor @XMM[0], @XMM[8] |
| 1492 | vst1.8 {@XMM[0]}, [$out]! @ write output |
| 1493 | cmp $len, #2 |
| 1494 | blo .Lctr_enc_done |
| 1495 | vld1.8 {@XMM[9]}, [$inp]! |
| 1496 | veor @XMM[1], @XMM[9] |
| 1497 | vst1.8 {@XMM[1]}, [$out]! |
| 1498 | beq .Lctr_enc_done |
| 1499 | vld1.8 {@XMM[10]}, [$inp]! |
| 1500 | veor @XMM[4], @XMM[10] |
| 1501 | vst1.8 {@XMM[4]}, [$out]! |
| 1502 | cmp $len, #4 |
| 1503 | blo .Lctr_enc_done |
| 1504 | vld1.8 {@XMM[11]}, [$inp]! |
| 1505 | veor @XMM[6], @XMM[11] |
| 1506 | vst1.8 {@XMM[6]}, [$out]! |
| 1507 | beq .Lctr_enc_done |
| 1508 | vld1.8 {@XMM[12]}, [$inp]! |
| 1509 | veor @XMM[3], @XMM[12] |
| 1510 | vst1.8 {@XMM[3]}, [$out]! |
| 1511 | cmp $len, #6 |
| 1512 | blo .Lctr_enc_done |
| 1513 | vld1.8 {@XMM[13]}, [$inp]! |
| 1514 | veor @XMM[7], @XMM[13] |
| 1515 | vst1.8 {@XMM[7]}, [$out]! |
| 1516 | beq .Lctr_enc_done |
| 1517 | vld1.8 {@XMM[14]}, [$inp] |
| 1518 | veor @XMM[2], @XMM[14] |
| 1519 | vst1.8 {@XMM[2]}, [$out]! |
| 1520 | |
| 1521 | .Lctr_enc_done: |
| 1522 | vmov.i32 q0, #0 |
| 1523 | vmov.i32 q1, #0 |
| 1524 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1525 | .Lctr_enc_bzero: @ wipe key schedule [if any] |
| 1526 | vstmia $keysched!, {q0-q1} |
| 1527 | cmp $keysched, $fp |
| 1528 | bne .Lctr_enc_bzero |
| 1529 | #else |
| 1530 | vstmia $keysched, {q0-q1} |
| 1531 | #endif |
| 1532 | |
| 1533 | mov sp, $fp |
| 1534 | add sp, #0x10 @ add sp,$fp,#0x10 is no good for thumb |
| 1535 | VFP_ABI_POP |
| 1536 | ldmia sp!, {r4-r10, pc} @ return |
| 1537 | |
| 1538 | .align 4 |
| 1539 | .Lctr_enc_short: |
| 1540 | ldr ip, [sp] @ ctr pointer is passed on stack |
| 1541 | stmdb sp!, {r4-r8, lr} |
| 1542 | |
| 1543 | mov r4, $inp @ copy arguments |
| 1544 | mov r5, $out |
| 1545 | mov r6, $len |
| 1546 | mov r7, $key |
| 1547 | ldr r8, [ip, #12] @ load counter LSW |
| 1548 | vld1.8 {@XMM[1]}, [ip] @ load whole counter value |
| 1549 | #ifdef __ARMEL__ |
| 1550 | rev r8, r8 |
| 1551 | #endif |
| 1552 | sub sp, sp, #0x10 |
| 1553 | vst1.8 {@XMM[1]}, [sp,:64] @ copy counter value |
| 1554 | sub sp, sp, #0x10 |
| 1555 | |
| 1556 | .Lctr_enc_short_loop: |
| 1557 | add r0, sp, #0x10 @ input counter value |
| 1558 | mov r1, sp @ output on the stack |
| 1559 | mov r2, r7 @ key |
| 1560 | |
| 1561 | bl AES_encrypt |
| 1562 | |
| 1563 | vld1.8 {@XMM[0]}, [r4]! @ load input |
| 1564 | vld1.8 {@XMM[1]}, [sp,:64] @ load encrypted counter |
| 1565 | add r8, r8, #1 |
| 1566 | #ifdef __ARMEL__ |
| 1567 | rev r0, r8 |
| 1568 | str r0, [sp, #0x1c] @ next counter value |
| 1569 | #else |
| 1570 | str r8, [sp, #0x1c] @ next counter value |
| 1571 | #endif |
| 1572 | veor @XMM[0],@XMM[0],@XMM[1] |
| 1573 | vst1.8 {@XMM[0]}, [r5]! @ store output |
| 1574 | subs r6, r6, #1 |
| 1575 | bne .Lctr_enc_short_loop |
| 1576 | |
| 1577 | vmov.i32 q0, #0 |
| 1578 | vmov.i32 q1, #0 |
| 1579 | vstmia sp!, {q0-q1} |
| 1580 | |
| 1581 | ldmia sp!, {r4-r8, pc} |
| 1582 | .size bsaes_ctr32_encrypt_blocks,.-bsaes_ctr32_encrypt_blocks |
| 1583 | ___ |
| 1584 | } |
| 1585 | { |
| 1586 | ###################################################################### |
| 1587 | # void bsaes_xts_[en|de]crypt(const char *inp,char *out,size_t len, |
| 1588 | # const AES_KEY *key1, const AES_KEY *key2, |
| 1589 | # const unsigned char iv[16]); |
| 1590 | # |
| 1591 | my ($inp,$out,$len,$key,$rounds,$magic,$fp)=(map("r$_",(7..10,1..3))); |
| 1592 | my $const="r6"; # returned by _bsaes_key_convert |
| 1593 | my $twmask=@XMM[5]; |
| 1594 | my @T=@XMM[6..7]; |
| 1595 | |
| 1596 | $code.=<<___; |
| 1597 | .globl bsaes_xts_encrypt |
| 1598 | .type bsaes_xts_encrypt,%function |
| 1599 | .align 4 |
| 1600 | bsaes_xts_encrypt: |
| 1601 | mov ip, sp |
| 1602 | stmdb sp!, {r4-r10, lr} @ 0x20 |
| 1603 | VFP_ABI_PUSH |
| 1604 | mov r6, sp @ future $fp |
| 1605 | |
| 1606 | mov $inp, r0 |
| 1607 | mov $out, r1 |
| 1608 | mov $len, r2 |
| 1609 | mov $key, r3 |
| 1610 | |
| 1611 | sub r0, sp, #0x10 @ 0x10 |
| 1612 | bic r0, #0xf @ align at 16 bytes |
| 1613 | mov sp, r0 |
| 1614 | |
| 1615 | #ifdef XTS_CHAIN_TWEAK |
| 1616 | ldr r0, [ip] @ pointer to input tweak |
| 1617 | #else |
| 1618 | @ generate initial tweak |
| 1619 | ldr r0, [ip, #4] @ iv[] |
| 1620 | mov r1, sp |
| 1621 | ldr r2, [ip, #0] @ key2 |
| 1622 | bl AES_encrypt |
| 1623 | mov r0,sp @ pointer to initial tweak |
| 1624 | #endif |
| 1625 | |
| 1626 | ldr $rounds, [$key, #240] @ get # of rounds |
| 1627 | mov $fp, r6 |
| 1628 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1629 | @ allocate the key schedule on the stack |
| 1630 | sub r12, sp, $rounds, lsl#7 @ 128 bytes per inner round key |
| 1631 | @ add r12, #`128-32` @ size of bit-sliced key schedule |
| 1632 | sub r12, #`32+16` @ place for tweak[9] |
| 1633 | |
| 1634 | @ populate the key schedule |
| 1635 | mov r4, $key @ pass key |
| 1636 | mov r5, $rounds @ pass # of rounds |
| 1637 | mov sp, r12 |
| 1638 | add r12, #0x90 @ pass key schedule |
| 1639 | bl _bsaes_key_convert |
| 1640 | veor @XMM[7], @XMM[7], @XMM[15] @ fix up last round key |
| 1641 | vstmia r12, {@XMM[7]} @ save last round key |
| 1642 | #else |
| 1643 | ldr r12, [$key, #244] |
| 1644 | eors r12, #1 |
| 1645 | beq 0f |
| 1646 | |
| 1647 | str r12, [$key, #244] |
| 1648 | mov r4, $key @ pass key |
| 1649 | mov r5, $rounds @ pass # of rounds |
| 1650 | add r12, $key, #248 @ pass key schedule |
| 1651 | bl _bsaes_key_convert |
| 1652 | veor @XMM[7], @XMM[7], @XMM[15] @ fix up last round key |
| 1653 | vstmia r12, {@XMM[7]} |
| 1654 | |
| 1655 | .align 2 |
| 1656 | 0: sub sp, #0x90 @ place for tweak[9] |
| 1657 | #endif |
| 1658 | |
| 1659 | vld1.8 {@XMM[8]}, [r0] @ initial tweak |
| 1660 | adr $magic, .Lxts_magic |
| 1661 | |
| 1662 | subs $len, #0x80 |
| 1663 | blo .Lxts_enc_short |
| 1664 | b .Lxts_enc_loop |
| 1665 | |
| 1666 | .align 4 |
| 1667 | .Lxts_enc_loop: |
| 1668 | vldmia $magic, {$twmask} @ load XTS magic |
| 1669 | vshr.s64 @T[0], @XMM[8], #63 |
| 1670 | mov r0, sp |
| 1671 | vand @T[0], @T[0], $twmask |
| 1672 | ___ |
| 1673 | for($i=9;$i<16;$i++) { |
| 1674 | $code.=<<___; |
| 1675 | vadd.u64 @XMM[$i], @XMM[$i-1], @XMM[$i-1] |
| 1676 | vst1.64 {@XMM[$i-1]}, [r0,:128]! |
| 1677 | vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` |
| 1678 | vshr.s64 @T[1], @XMM[$i], #63 |
| 1679 | veor @XMM[$i], @XMM[$i], @T[0] |
| 1680 | vand @T[1], @T[1], $twmask |
| 1681 | ___ |
| 1682 | @T=reverse(@T); |
| 1683 | |
| 1684 | $code.=<<___ if ($i>=10); |
| 1685 | vld1.8 {@XMM[$i-10]}, [$inp]! |
| 1686 | ___ |
| 1687 | $code.=<<___ if ($i>=11); |
| 1688 | veor @XMM[$i-11], @XMM[$i-11], @XMM[$i-3] |
| 1689 | ___ |
| 1690 | } |
| 1691 | $code.=<<___; |
| 1692 | vadd.u64 @XMM[8], @XMM[15], @XMM[15] |
| 1693 | vst1.64 {@XMM[15]}, [r0,:128]! |
| 1694 | vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` |
| 1695 | veor @XMM[8], @XMM[8], @T[0] |
| 1696 | vst1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 1697 | |
| 1698 | vld1.8 {@XMM[6]-@XMM[7]}, [$inp]! |
| 1699 | veor @XMM[5], @XMM[5], @XMM[13] |
| 1700 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1701 | add r4, sp, #0x90 @ pass key schedule |
| 1702 | #else |
| 1703 | add r4, $key, #248 @ pass key schedule |
| 1704 | #endif |
| 1705 | veor @XMM[6], @XMM[6], @XMM[14] |
| 1706 | mov r5, $rounds @ pass rounds |
| 1707 | veor @XMM[7], @XMM[7], @XMM[15] |
| 1708 | mov r0, sp |
| 1709 | |
| 1710 | bl _bsaes_encrypt8 |
| 1711 | |
| 1712 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 1713 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 1714 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 1715 | vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! |
| 1716 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 1717 | veor @XMM[8], @XMM[4], @XMM[10] |
| 1718 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 1719 | veor @XMM[9], @XMM[6], @XMM[11] |
| 1720 | vld1.64 {@XMM[14]-@XMM[15]}, [r0,:128]! |
| 1721 | veor @XMM[10], @XMM[3], @XMM[12] |
| 1722 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 1723 | veor @XMM[11], @XMM[7], @XMM[13] |
| 1724 | veor @XMM[12], @XMM[2], @XMM[14] |
| 1725 | vst1.8 {@XMM[10]-@XMM[11]}, [$out]! |
| 1726 | veor @XMM[13], @XMM[5], @XMM[15] |
| 1727 | vst1.8 {@XMM[12]-@XMM[13]}, [$out]! |
| 1728 | |
| 1729 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 1730 | |
| 1731 | subs $len, #0x80 |
| 1732 | bpl .Lxts_enc_loop |
| 1733 | |
| 1734 | .Lxts_enc_short: |
| 1735 | adds $len, #0x70 |
| 1736 | bmi .Lxts_enc_done |
| 1737 | |
| 1738 | vldmia $magic, {$twmask} @ load XTS magic |
| 1739 | vshr.s64 @T[0], @XMM[8], #63 |
| 1740 | mov r0, sp |
| 1741 | vand @T[0], @T[0], $twmask |
| 1742 | ___ |
| 1743 | for($i=9;$i<16;$i++) { |
| 1744 | $code.=<<___; |
| 1745 | vadd.u64 @XMM[$i], @XMM[$i-1], @XMM[$i-1] |
| 1746 | vst1.64 {@XMM[$i-1]}, [r0,:128]! |
| 1747 | vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` |
| 1748 | vshr.s64 @T[1], @XMM[$i], #63 |
| 1749 | veor @XMM[$i], @XMM[$i], @T[0] |
| 1750 | vand @T[1], @T[1], $twmask |
| 1751 | ___ |
| 1752 | @T=reverse(@T); |
| 1753 | |
| 1754 | $code.=<<___ if ($i>=10); |
| 1755 | vld1.8 {@XMM[$i-10]}, [$inp]! |
| 1756 | subs $len, #0x10 |
| 1757 | bmi .Lxts_enc_`$i-9` |
| 1758 | ___ |
| 1759 | $code.=<<___ if ($i>=11); |
| 1760 | veor @XMM[$i-11], @XMM[$i-11], @XMM[$i-3] |
| 1761 | ___ |
| 1762 | } |
| 1763 | $code.=<<___; |
| 1764 | sub $len, #0x10 |
| 1765 | vst1.64 {@XMM[15]}, [r0,:128] @ next round tweak |
| 1766 | |
| 1767 | vld1.8 {@XMM[6]}, [$inp]! |
| 1768 | veor @XMM[5], @XMM[5], @XMM[13] |
| 1769 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1770 | add r4, sp, #0x90 @ pass key schedule |
| 1771 | #else |
| 1772 | add r4, $key, #248 @ pass key schedule |
| 1773 | #endif |
| 1774 | veor @XMM[6], @XMM[6], @XMM[14] |
| 1775 | mov r5, $rounds @ pass rounds |
| 1776 | mov r0, sp |
| 1777 | |
| 1778 | bl _bsaes_encrypt8 |
| 1779 | |
| 1780 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 1781 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 1782 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 1783 | vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! |
| 1784 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 1785 | veor @XMM[8], @XMM[4], @XMM[10] |
| 1786 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 1787 | veor @XMM[9], @XMM[6], @XMM[11] |
| 1788 | vld1.64 {@XMM[14]}, [r0,:128]! |
| 1789 | veor @XMM[10], @XMM[3], @XMM[12] |
| 1790 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 1791 | veor @XMM[11], @XMM[7], @XMM[13] |
| 1792 | veor @XMM[12], @XMM[2], @XMM[14] |
| 1793 | vst1.8 {@XMM[10]-@XMM[11]}, [$out]! |
| 1794 | vst1.8 {@XMM[12]}, [$out]! |
| 1795 | |
| 1796 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 1797 | b .Lxts_enc_done |
| 1798 | .align 4 |
| 1799 | .Lxts_enc_6: |
| 1800 | vst1.64 {@XMM[14]}, [r0,:128] @ next round tweak |
| 1801 | |
| 1802 | veor @XMM[4], @XMM[4], @XMM[12] |
| 1803 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1804 | add r4, sp, #0x90 @ pass key schedule |
| 1805 | #else |
| 1806 | add r4, $key, #248 @ pass key schedule |
| 1807 | #endif |
| 1808 | veor @XMM[5], @XMM[5], @XMM[13] |
| 1809 | mov r5, $rounds @ pass rounds |
| 1810 | mov r0, sp |
| 1811 | |
| 1812 | bl _bsaes_encrypt8 |
| 1813 | |
| 1814 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 1815 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 1816 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 1817 | vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! |
| 1818 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 1819 | veor @XMM[8], @XMM[4], @XMM[10] |
| 1820 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 1821 | veor @XMM[9], @XMM[6], @XMM[11] |
| 1822 | veor @XMM[10], @XMM[3], @XMM[12] |
| 1823 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 1824 | veor @XMM[11], @XMM[7], @XMM[13] |
| 1825 | vst1.8 {@XMM[10]-@XMM[11]}, [$out]! |
| 1826 | |
| 1827 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 1828 | b .Lxts_enc_done |
| 1829 | |
| 1830 | @ put this in range for both ARM and Thumb mode adr instructions |
| 1831 | .align 5 |
| 1832 | .Lxts_magic: |
| 1833 | .quad 1, 0x87 |
| 1834 | |
| 1835 | .align 5 |
| 1836 | .Lxts_enc_5: |
| 1837 | vst1.64 {@XMM[13]}, [r0,:128] @ next round tweak |
| 1838 | |
| 1839 | veor @XMM[3], @XMM[3], @XMM[11] |
| 1840 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1841 | add r4, sp, #0x90 @ pass key schedule |
| 1842 | #else |
| 1843 | add r4, $key, #248 @ pass key schedule |
| 1844 | #endif |
| 1845 | veor @XMM[4], @XMM[4], @XMM[12] |
| 1846 | mov r5, $rounds @ pass rounds |
| 1847 | mov r0, sp |
| 1848 | |
| 1849 | bl _bsaes_encrypt8 |
| 1850 | |
| 1851 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 1852 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 1853 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 1854 | vld1.64 {@XMM[12]}, [r0,:128]! |
| 1855 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 1856 | veor @XMM[8], @XMM[4], @XMM[10] |
| 1857 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 1858 | veor @XMM[9], @XMM[6], @XMM[11] |
| 1859 | veor @XMM[10], @XMM[3], @XMM[12] |
| 1860 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 1861 | vst1.8 {@XMM[10]}, [$out]! |
| 1862 | |
| 1863 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 1864 | b .Lxts_enc_done |
| 1865 | .align 4 |
| 1866 | .Lxts_enc_4: |
| 1867 | vst1.64 {@XMM[12]}, [r0,:128] @ next round tweak |
| 1868 | |
| 1869 | veor @XMM[2], @XMM[2], @XMM[10] |
| 1870 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1871 | add r4, sp, #0x90 @ pass key schedule |
| 1872 | #else |
| 1873 | add r4, $key, #248 @ pass key schedule |
| 1874 | #endif |
| 1875 | veor @XMM[3], @XMM[3], @XMM[11] |
| 1876 | mov r5, $rounds @ pass rounds |
| 1877 | mov r0, sp |
| 1878 | |
| 1879 | bl _bsaes_encrypt8 |
| 1880 | |
| 1881 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 1882 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 1883 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 1884 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 1885 | veor @XMM[8], @XMM[4], @XMM[10] |
| 1886 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 1887 | veor @XMM[9], @XMM[6], @XMM[11] |
| 1888 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 1889 | |
| 1890 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 1891 | b .Lxts_enc_done |
| 1892 | .align 4 |
| 1893 | .Lxts_enc_3: |
| 1894 | vst1.64 {@XMM[11]}, [r0,:128] @ next round tweak |
| 1895 | |
| 1896 | veor @XMM[1], @XMM[1], @XMM[9] |
| 1897 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1898 | add r4, sp, #0x90 @ pass key schedule |
| 1899 | #else |
| 1900 | add r4, $key, #248 @ pass key schedule |
| 1901 | #endif |
| 1902 | veor @XMM[2], @XMM[2], @XMM[10] |
| 1903 | mov r5, $rounds @ pass rounds |
| 1904 | mov r0, sp |
| 1905 | |
| 1906 | bl _bsaes_encrypt8 |
| 1907 | |
| 1908 | vld1.64 {@XMM[8]-@XMM[9]}, [r0,:128]! |
| 1909 | vld1.64 {@XMM[10]}, [r0,:128]! |
| 1910 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 1911 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 1912 | veor @XMM[8], @XMM[4], @XMM[10] |
| 1913 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 1914 | vst1.8 {@XMM[8]}, [$out]! |
| 1915 | |
| 1916 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 1917 | b .Lxts_enc_done |
| 1918 | .align 4 |
| 1919 | .Lxts_enc_2: |
| 1920 | vst1.64 {@XMM[10]}, [r0,:128] @ next round tweak |
| 1921 | |
| 1922 | veor @XMM[0], @XMM[0], @XMM[8] |
| 1923 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 1924 | add r4, sp, #0x90 @ pass key schedule |
| 1925 | #else |
| 1926 | add r4, $key, #248 @ pass key schedule |
| 1927 | #endif |
| 1928 | veor @XMM[1], @XMM[1], @XMM[9] |
| 1929 | mov r5, $rounds @ pass rounds |
| 1930 | mov r0, sp |
| 1931 | |
| 1932 | bl _bsaes_encrypt8 |
| 1933 | |
| 1934 | vld1.64 {@XMM[8]-@XMM[9]}, [r0,:128]! |
| 1935 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 1936 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 1937 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 1938 | |
| 1939 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 1940 | b .Lxts_enc_done |
| 1941 | .align 4 |
| 1942 | .Lxts_enc_1: |
| 1943 | mov r0, sp |
| 1944 | veor @XMM[0], @XMM[8] |
| 1945 | mov r1, sp |
| 1946 | vst1.8 {@XMM[0]}, [sp,:128] |
| 1947 | mov r2, $key |
| 1948 | mov r4, $fp @ preserve fp |
| 1949 | |
| 1950 | bl AES_encrypt |
| 1951 | |
| 1952 | vld1.8 {@XMM[0]}, [sp,:128] |
| 1953 | veor @XMM[0], @XMM[0], @XMM[8] |
| 1954 | vst1.8 {@XMM[0]}, [$out]! |
| 1955 | mov $fp, r4 |
| 1956 | |
| 1957 | vmov @XMM[8], @XMM[9] @ next round tweak |
| 1958 | |
| 1959 | .Lxts_enc_done: |
| 1960 | #ifndef XTS_CHAIN_TWEAK |
| 1961 | adds $len, #0x10 |
| 1962 | beq .Lxts_enc_ret |
| 1963 | sub r6, $out, #0x10 |
| 1964 | |
| 1965 | .Lxts_enc_steal: |
| 1966 | ldrb r0, [$inp], #1 |
| 1967 | ldrb r1, [$out, #-0x10] |
| 1968 | strb r0, [$out, #-0x10] |
| 1969 | strb r1, [$out], #1 |
| 1970 | |
| 1971 | subs $len, #1 |
| 1972 | bhi .Lxts_enc_steal |
| 1973 | |
| 1974 | vld1.8 {@XMM[0]}, [r6] |
| 1975 | mov r0, sp |
| 1976 | veor @XMM[0], @XMM[0], @XMM[8] |
| 1977 | mov r1, sp |
| 1978 | vst1.8 {@XMM[0]}, [sp,:128] |
| 1979 | mov r2, $key |
| 1980 | mov r4, $fp @ preserve fp |
| 1981 | |
| 1982 | bl AES_encrypt |
| 1983 | |
| 1984 | vld1.8 {@XMM[0]}, [sp,:128] |
| 1985 | veor @XMM[0], @XMM[0], @XMM[8] |
| 1986 | vst1.8 {@XMM[0]}, [r6] |
| 1987 | mov $fp, r4 |
| 1988 | #endif |
| 1989 | |
| 1990 | .Lxts_enc_ret: |
| 1991 | bic r0, $fp, #0xf |
| 1992 | vmov.i32 q0, #0 |
| 1993 | vmov.i32 q1, #0 |
| 1994 | #ifdef XTS_CHAIN_TWEAK |
| 1995 | ldr r1, [$fp, #0x20+VFP_ABI_FRAME] @ chain tweak |
| 1996 | #endif |
| 1997 | .Lxts_enc_bzero: @ wipe key schedule [if any] |
| 1998 | vstmia sp!, {q0-q1} |
| 1999 | cmp sp, r0 |
| 2000 | bne .Lxts_enc_bzero |
| 2001 | |
| 2002 | mov sp, $fp |
| 2003 | #ifdef XTS_CHAIN_TWEAK |
| 2004 | vst1.8 {@XMM[8]}, [r1] |
| 2005 | #endif |
| 2006 | VFP_ABI_POP |
| 2007 | ldmia sp!, {r4-r10, pc} @ return |
| 2008 | |
| 2009 | .size bsaes_xts_encrypt,.-bsaes_xts_encrypt |
| 2010 | |
| 2011 | .globl bsaes_xts_decrypt |
| 2012 | .type bsaes_xts_decrypt,%function |
| 2013 | .align 4 |
| 2014 | bsaes_xts_decrypt: |
| 2015 | mov ip, sp |
| 2016 | stmdb sp!, {r4-r10, lr} @ 0x20 |
| 2017 | VFP_ABI_PUSH |
| 2018 | mov r6, sp @ future $fp |
| 2019 | |
| 2020 | mov $inp, r0 |
| 2021 | mov $out, r1 |
| 2022 | mov $len, r2 |
| 2023 | mov $key, r3 |
| 2024 | |
| 2025 | sub r0, sp, #0x10 @ 0x10 |
| 2026 | bic r0, #0xf @ align at 16 bytes |
| 2027 | mov sp, r0 |
| 2028 | |
| 2029 | #ifdef XTS_CHAIN_TWEAK |
| 2030 | ldr r0, [ip] @ pointer to input tweak |
| 2031 | #else |
| 2032 | @ generate initial tweak |
| 2033 | ldr r0, [ip, #4] @ iv[] |
| 2034 | mov r1, sp |
| 2035 | ldr r2, [ip, #0] @ key2 |
| 2036 | bl AES_encrypt |
| 2037 | mov r0, sp @ pointer to initial tweak |
| 2038 | #endif |
| 2039 | |
| 2040 | ldr $rounds, [$key, #240] @ get # of rounds |
| 2041 | mov $fp, r6 |
| 2042 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 2043 | @ allocate the key schedule on the stack |
| 2044 | sub r12, sp, $rounds, lsl#7 @ 128 bytes per inner round key |
| 2045 | @ add r12, #`128-32` @ size of bit-sliced key schedule |
| 2046 | sub r12, #`32+16` @ place for tweak[9] |
| 2047 | |
| 2048 | @ populate the key schedule |
| 2049 | mov r4, $key @ pass key |
| 2050 | mov r5, $rounds @ pass # of rounds |
| 2051 | mov sp, r12 |
| 2052 | add r12, #0x90 @ pass key schedule |
| 2053 | bl _bsaes_key_convert |
| 2054 | add r4, sp, #0x90 |
| 2055 | vldmia r4, {@XMM[6]} |
| 2056 | vstmia r12, {@XMM[15]} @ save last round key |
| 2057 | veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key |
| 2058 | vstmia r4, {@XMM[7]} |
| 2059 | #else |
| 2060 | ldr r12, [$key, #244] |
| 2061 | eors r12, #1 |
| 2062 | beq 0f |
| 2063 | |
| 2064 | str r12, [$key, #244] |
| 2065 | mov r4, $key @ pass key |
| 2066 | mov r5, $rounds @ pass # of rounds |
| 2067 | add r12, $key, #248 @ pass key schedule |
| 2068 | bl _bsaes_key_convert |
| 2069 | add r4, $key, #248 |
| 2070 | vldmia r4, {@XMM[6]} |
| 2071 | vstmia r12, {@XMM[15]} @ save last round key |
| 2072 | veor @XMM[7], @XMM[7], @XMM[6] @ fix up round 0 key |
| 2073 | vstmia r4, {@XMM[7]} |
| 2074 | |
| 2075 | .align 2 |
| 2076 | 0: sub sp, #0x90 @ place for tweak[9] |
| 2077 | #endif |
| 2078 | vld1.8 {@XMM[8]}, [r0] @ initial tweak |
| 2079 | adr $magic, .Lxts_magic |
| 2080 | |
Ard Biesheuvel | 001eabf | 2015-02-26 07:22:05 +0000 | [diff] [blame] | 2081 | #ifndef XTS_CHAIN_TWEAK |
Ard Biesheuvel | e4e7f10 | 2013-09-16 18:31:38 +0200 | [diff] [blame] | 2082 | tst $len, #0xf @ if not multiple of 16 |
| 2083 | it ne @ Thumb2 thing, sanity check in ARM |
| 2084 | subne $len, #0x10 @ subtract another 16 bytes |
Ard Biesheuvel | 001eabf | 2015-02-26 07:22:05 +0000 | [diff] [blame] | 2085 | #endif |
Ard Biesheuvel | e4e7f10 | 2013-09-16 18:31:38 +0200 | [diff] [blame] | 2086 | subs $len, #0x80 |
| 2087 | |
| 2088 | blo .Lxts_dec_short |
| 2089 | b .Lxts_dec_loop |
| 2090 | |
| 2091 | .align 4 |
| 2092 | .Lxts_dec_loop: |
| 2093 | vldmia $magic, {$twmask} @ load XTS magic |
| 2094 | vshr.s64 @T[0], @XMM[8], #63 |
| 2095 | mov r0, sp |
| 2096 | vand @T[0], @T[0], $twmask |
| 2097 | ___ |
| 2098 | for($i=9;$i<16;$i++) { |
| 2099 | $code.=<<___; |
| 2100 | vadd.u64 @XMM[$i], @XMM[$i-1], @XMM[$i-1] |
| 2101 | vst1.64 {@XMM[$i-1]}, [r0,:128]! |
| 2102 | vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` |
| 2103 | vshr.s64 @T[1], @XMM[$i], #63 |
| 2104 | veor @XMM[$i], @XMM[$i], @T[0] |
| 2105 | vand @T[1], @T[1], $twmask |
| 2106 | ___ |
| 2107 | @T=reverse(@T); |
| 2108 | |
| 2109 | $code.=<<___ if ($i>=10); |
| 2110 | vld1.8 {@XMM[$i-10]}, [$inp]! |
| 2111 | ___ |
| 2112 | $code.=<<___ if ($i>=11); |
| 2113 | veor @XMM[$i-11], @XMM[$i-11], @XMM[$i-3] |
| 2114 | ___ |
| 2115 | } |
| 2116 | $code.=<<___; |
| 2117 | vadd.u64 @XMM[8], @XMM[15], @XMM[15] |
| 2118 | vst1.64 {@XMM[15]}, [r0,:128]! |
| 2119 | vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` |
| 2120 | veor @XMM[8], @XMM[8], @T[0] |
| 2121 | vst1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 2122 | |
| 2123 | vld1.8 {@XMM[6]-@XMM[7]}, [$inp]! |
| 2124 | veor @XMM[5], @XMM[5], @XMM[13] |
| 2125 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 2126 | add r4, sp, #0x90 @ pass key schedule |
| 2127 | #else |
| 2128 | add r4, $key, #248 @ pass key schedule |
| 2129 | #endif |
| 2130 | veor @XMM[6], @XMM[6], @XMM[14] |
| 2131 | mov r5, $rounds @ pass rounds |
| 2132 | veor @XMM[7], @XMM[7], @XMM[15] |
| 2133 | mov r0, sp |
| 2134 | |
| 2135 | bl _bsaes_decrypt8 |
| 2136 | |
| 2137 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 2138 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 2139 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 2140 | vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! |
| 2141 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 2142 | veor @XMM[8], @XMM[6], @XMM[10] |
| 2143 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 2144 | veor @XMM[9], @XMM[4], @XMM[11] |
| 2145 | vld1.64 {@XMM[14]-@XMM[15]}, [r0,:128]! |
| 2146 | veor @XMM[10], @XMM[2], @XMM[12] |
| 2147 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 2148 | veor @XMM[11], @XMM[7], @XMM[13] |
| 2149 | veor @XMM[12], @XMM[3], @XMM[14] |
| 2150 | vst1.8 {@XMM[10]-@XMM[11]}, [$out]! |
| 2151 | veor @XMM[13], @XMM[5], @XMM[15] |
| 2152 | vst1.8 {@XMM[12]-@XMM[13]}, [$out]! |
| 2153 | |
| 2154 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 2155 | |
| 2156 | subs $len, #0x80 |
| 2157 | bpl .Lxts_dec_loop |
| 2158 | |
| 2159 | .Lxts_dec_short: |
| 2160 | adds $len, #0x70 |
| 2161 | bmi .Lxts_dec_done |
| 2162 | |
| 2163 | vldmia $magic, {$twmask} @ load XTS magic |
| 2164 | vshr.s64 @T[0], @XMM[8], #63 |
| 2165 | mov r0, sp |
| 2166 | vand @T[0], @T[0], $twmask |
| 2167 | ___ |
| 2168 | for($i=9;$i<16;$i++) { |
| 2169 | $code.=<<___; |
| 2170 | vadd.u64 @XMM[$i], @XMM[$i-1], @XMM[$i-1] |
| 2171 | vst1.64 {@XMM[$i-1]}, [r0,:128]! |
| 2172 | vswp `&Dhi("@T[0]")`,`&Dlo("@T[0]")` |
| 2173 | vshr.s64 @T[1], @XMM[$i], #63 |
| 2174 | veor @XMM[$i], @XMM[$i], @T[0] |
| 2175 | vand @T[1], @T[1], $twmask |
| 2176 | ___ |
| 2177 | @T=reverse(@T); |
| 2178 | |
| 2179 | $code.=<<___ if ($i>=10); |
| 2180 | vld1.8 {@XMM[$i-10]}, [$inp]! |
| 2181 | subs $len, #0x10 |
| 2182 | bmi .Lxts_dec_`$i-9` |
| 2183 | ___ |
| 2184 | $code.=<<___ if ($i>=11); |
| 2185 | veor @XMM[$i-11], @XMM[$i-11], @XMM[$i-3] |
| 2186 | ___ |
| 2187 | } |
| 2188 | $code.=<<___; |
| 2189 | sub $len, #0x10 |
| 2190 | vst1.64 {@XMM[15]}, [r0,:128] @ next round tweak |
| 2191 | |
| 2192 | vld1.8 {@XMM[6]}, [$inp]! |
| 2193 | veor @XMM[5], @XMM[5], @XMM[13] |
| 2194 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 2195 | add r4, sp, #0x90 @ pass key schedule |
| 2196 | #else |
| 2197 | add r4, $key, #248 @ pass key schedule |
| 2198 | #endif |
| 2199 | veor @XMM[6], @XMM[6], @XMM[14] |
| 2200 | mov r5, $rounds @ pass rounds |
| 2201 | mov r0, sp |
| 2202 | |
| 2203 | bl _bsaes_decrypt8 |
| 2204 | |
| 2205 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 2206 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 2207 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 2208 | vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! |
| 2209 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 2210 | veor @XMM[8], @XMM[6], @XMM[10] |
| 2211 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 2212 | veor @XMM[9], @XMM[4], @XMM[11] |
| 2213 | vld1.64 {@XMM[14]}, [r0,:128]! |
| 2214 | veor @XMM[10], @XMM[2], @XMM[12] |
| 2215 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 2216 | veor @XMM[11], @XMM[7], @XMM[13] |
| 2217 | veor @XMM[12], @XMM[3], @XMM[14] |
| 2218 | vst1.8 {@XMM[10]-@XMM[11]}, [$out]! |
| 2219 | vst1.8 {@XMM[12]}, [$out]! |
| 2220 | |
| 2221 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 2222 | b .Lxts_dec_done |
| 2223 | .align 4 |
| 2224 | .Lxts_dec_6: |
| 2225 | vst1.64 {@XMM[14]}, [r0,:128] @ next round tweak |
| 2226 | |
| 2227 | veor @XMM[4], @XMM[4], @XMM[12] |
| 2228 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 2229 | add r4, sp, #0x90 @ pass key schedule |
| 2230 | #else |
| 2231 | add r4, $key, #248 @ pass key schedule |
| 2232 | #endif |
| 2233 | veor @XMM[5], @XMM[5], @XMM[13] |
| 2234 | mov r5, $rounds @ pass rounds |
| 2235 | mov r0, sp |
| 2236 | |
| 2237 | bl _bsaes_decrypt8 |
| 2238 | |
| 2239 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 2240 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 2241 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 2242 | vld1.64 {@XMM[12]-@XMM[13]}, [r0,:128]! |
| 2243 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 2244 | veor @XMM[8], @XMM[6], @XMM[10] |
| 2245 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 2246 | veor @XMM[9], @XMM[4], @XMM[11] |
| 2247 | veor @XMM[10], @XMM[2], @XMM[12] |
| 2248 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 2249 | veor @XMM[11], @XMM[7], @XMM[13] |
| 2250 | vst1.8 {@XMM[10]-@XMM[11]}, [$out]! |
| 2251 | |
| 2252 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 2253 | b .Lxts_dec_done |
| 2254 | .align 4 |
| 2255 | .Lxts_dec_5: |
| 2256 | vst1.64 {@XMM[13]}, [r0,:128] @ next round tweak |
| 2257 | |
| 2258 | veor @XMM[3], @XMM[3], @XMM[11] |
| 2259 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 2260 | add r4, sp, #0x90 @ pass key schedule |
| 2261 | #else |
| 2262 | add r4, $key, #248 @ pass key schedule |
| 2263 | #endif |
| 2264 | veor @XMM[4], @XMM[4], @XMM[12] |
| 2265 | mov r5, $rounds @ pass rounds |
| 2266 | mov r0, sp |
| 2267 | |
| 2268 | bl _bsaes_decrypt8 |
| 2269 | |
| 2270 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 2271 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 2272 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 2273 | vld1.64 {@XMM[12]}, [r0,:128]! |
| 2274 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 2275 | veor @XMM[8], @XMM[6], @XMM[10] |
| 2276 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 2277 | veor @XMM[9], @XMM[4], @XMM[11] |
| 2278 | veor @XMM[10], @XMM[2], @XMM[12] |
| 2279 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 2280 | vst1.8 {@XMM[10]}, [$out]! |
| 2281 | |
| 2282 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 2283 | b .Lxts_dec_done |
| 2284 | .align 4 |
| 2285 | .Lxts_dec_4: |
| 2286 | vst1.64 {@XMM[12]}, [r0,:128] @ next round tweak |
| 2287 | |
| 2288 | veor @XMM[2], @XMM[2], @XMM[10] |
| 2289 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 2290 | add r4, sp, #0x90 @ pass key schedule |
| 2291 | #else |
| 2292 | add r4, $key, #248 @ pass key schedule |
| 2293 | #endif |
| 2294 | veor @XMM[3], @XMM[3], @XMM[11] |
| 2295 | mov r5, $rounds @ pass rounds |
| 2296 | mov r0, sp |
| 2297 | |
| 2298 | bl _bsaes_decrypt8 |
| 2299 | |
| 2300 | vld1.64 {@XMM[ 8]-@XMM[ 9]}, [r0,:128]! |
| 2301 | vld1.64 {@XMM[10]-@XMM[11]}, [r0,:128]! |
| 2302 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 2303 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 2304 | veor @XMM[8], @XMM[6], @XMM[10] |
| 2305 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 2306 | veor @XMM[9], @XMM[4], @XMM[11] |
| 2307 | vst1.8 {@XMM[8]-@XMM[9]}, [$out]! |
| 2308 | |
| 2309 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 2310 | b .Lxts_dec_done |
| 2311 | .align 4 |
| 2312 | .Lxts_dec_3: |
| 2313 | vst1.64 {@XMM[11]}, [r0,:128] @ next round tweak |
| 2314 | |
| 2315 | veor @XMM[1], @XMM[1], @XMM[9] |
| 2316 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 2317 | add r4, sp, #0x90 @ pass key schedule |
| 2318 | #else |
| 2319 | add r4, $key, #248 @ pass key schedule |
| 2320 | #endif |
| 2321 | veor @XMM[2], @XMM[2], @XMM[10] |
| 2322 | mov r5, $rounds @ pass rounds |
| 2323 | mov r0, sp |
| 2324 | |
| 2325 | bl _bsaes_decrypt8 |
| 2326 | |
| 2327 | vld1.64 {@XMM[8]-@XMM[9]}, [r0,:128]! |
| 2328 | vld1.64 {@XMM[10]}, [r0,:128]! |
| 2329 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 2330 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 2331 | veor @XMM[8], @XMM[6], @XMM[10] |
| 2332 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 2333 | vst1.8 {@XMM[8]}, [$out]! |
| 2334 | |
| 2335 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 2336 | b .Lxts_dec_done |
| 2337 | .align 4 |
| 2338 | .Lxts_dec_2: |
| 2339 | vst1.64 {@XMM[10]}, [r0,:128] @ next round tweak |
| 2340 | |
| 2341 | veor @XMM[0], @XMM[0], @XMM[8] |
| 2342 | #ifndef BSAES_ASM_EXTENDED_KEY |
| 2343 | add r4, sp, #0x90 @ pass key schedule |
| 2344 | #else |
| 2345 | add r4, $key, #248 @ pass key schedule |
| 2346 | #endif |
| 2347 | veor @XMM[1], @XMM[1], @XMM[9] |
| 2348 | mov r5, $rounds @ pass rounds |
| 2349 | mov r0, sp |
| 2350 | |
| 2351 | bl _bsaes_decrypt8 |
| 2352 | |
| 2353 | vld1.64 {@XMM[8]-@XMM[9]}, [r0,:128]! |
| 2354 | veor @XMM[0], @XMM[0], @XMM[ 8] |
| 2355 | veor @XMM[1], @XMM[1], @XMM[ 9] |
| 2356 | vst1.8 {@XMM[0]-@XMM[1]}, [$out]! |
| 2357 | |
| 2358 | vld1.64 {@XMM[8]}, [r0,:128] @ next round tweak |
| 2359 | b .Lxts_dec_done |
| 2360 | .align 4 |
| 2361 | .Lxts_dec_1: |
| 2362 | mov r0, sp |
| 2363 | veor @XMM[0], @XMM[8] |
| 2364 | mov r1, sp |
| 2365 | vst1.8 {@XMM[0]}, [sp,:128] |
| 2366 | mov r2, $key |
| 2367 | mov r4, $fp @ preserve fp |
| 2368 | mov r5, $magic @ preserve magic |
| 2369 | |
| 2370 | bl AES_decrypt |
| 2371 | |
| 2372 | vld1.8 {@XMM[0]}, [sp,:128] |
| 2373 | veor @XMM[0], @XMM[0], @XMM[8] |
| 2374 | vst1.8 {@XMM[0]}, [$out]! |
| 2375 | mov $fp, r4 |
| 2376 | mov $magic, r5 |
| 2377 | |
| 2378 | vmov @XMM[8], @XMM[9] @ next round tweak |
| 2379 | |
| 2380 | .Lxts_dec_done: |
| 2381 | #ifndef XTS_CHAIN_TWEAK |
| 2382 | adds $len, #0x10 |
| 2383 | beq .Lxts_dec_ret |
| 2384 | |
| 2385 | @ calculate one round of extra tweak for the stolen ciphertext |
| 2386 | vldmia $magic, {$twmask} |
| 2387 | vshr.s64 @XMM[6], @XMM[8], #63 |
| 2388 | vand @XMM[6], @XMM[6], $twmask |
| 2389 | vadd.u64 @XMM[9], @XMM[8], @XMM[8] |
| 2390 | vswp `&Dhi("@XMM[6]")`,`&Dlo("@XMM[6]")` |
| 2391 | veor @XMM[9], @XMM[9], @XMM[6] |
| 2392 | |
| 2393 | @ perform the final decryption with the last tweak value |
| 2394 | vld1.8 {@XMM[0]}, [$inp]! |
| 2395 | mov r0, sp |
| 2396 | veor @XMM[0], @XMM[0], @XMM[9] |
| 2397 | mov r1, sp |
| 2398 | vst1.8 {@XMM[0]}, [sp,:128] |
| 2399 | mov r2, $key |
| 2400 | mov r4, $fp @ preserve fp |
| 2401 | |
| 2402 | bl AES_decrypt |
| 2403 | |
| 2404 | vld1.8 {@XMM[0]}, [sp,:128] |
| 2405 | veor @XMM[0], @XMM[0], @XMM[9] |
| 2406 | vst1.8 {@XMM[0]}, [$out] |
| 2407 | |
| 2408 | mov r6, $out |
| 2409 | .Lxts_dec_steal: |
| 2410 | ldrb r1, [$out] |
| 2411 | ldrb r0, [$inp], #1 |
| 2412 | strb r1, [$out, #0x10] |
| 2413 | strb r0, [$out], #1 |
| 2414 | |
| 2415 | subs $len, #1 |
| 2416 | bhi .Lxts_dec_steal |
| 2417 | |
| 2418 | vld1.8 {@XMM[0]}, [r6] |
| 2419 | mov r0, sp |
| 2420 | veor @XMM[0], @XMM[8] |
| 2421 | mov r1, sp |
| 2422 | vst1.8 {@XMM[0]}, [sp,:128] |
| 2423 | mov r2, $key |
| 2424 | |
| 2425 | bl AES_decrypt |
| 2426 | |
| 2427 | vld1.8 {@XMM[0]}, [sp,:128] |
| 2428 | veor @XMM[0], @XMM[0], @XMM[8] |
| 2429 | vst1.8 {@XMM[0]}, [r6] |
| 2430 | mov $fp, r4 |
| 2431 | #endif |
| 2432 | |
| 2433 | .Lxts_dec_ret: |
| 2434 | bic r0, $fp, #0xf |
| 2435 | vmov.i32 q0, #0 |
| 2436 | vmov.i32 q1, #0 |
| 2437 | #ifdef XTS_CHAIN_TWEAK |
| 2438 | ldr r1, [$fp, #0x20+VFP_ABI_FRAME] @ chain tweak |
| 2439 | #endif |
| 2440 | .Lxts_dec_bzero: @ wipe key schedule [if any] |
| 2441 | vstmia sp!, {q0-q1} |
| 2442 | cmp sp, r0 |
| 2443 | bne .Lxts_dec_bzero |
| 2444 | |
| 2445 | mov sp, $fp |
| 2446 | #ifdef XTS_CHAIN_TWEAK |
| 2447 | vst1.8 {@XMM[8]}, [r1] |
| 2448 | #endif |
| 2449 | VFP_ABI_POP |
| 2450 | ldmia sp!, {r4-r10, pc} @ return |
| 2451 | |
| 2452 | .size bsaes_xts_decrypt,.-bsaes_xts_decrypt |
| 2453 | ___ |
| 2454 | } |
| 2455 | $code.=<<___; |
| 2456 | #endif |
| 2457 | ___ |
| 2458 | |
| 2459 | $code =~ s/\`([^\`]*)\`/eval($1)/gem; |
| 2460 | |
| 2461 | open SELF,$0; |
| 2462 | while(<SELF>) { |
| 2463 | next if (/^#!/); |
| 2464 | last if (!s/^#/@/ and !/^$/); |
| 2465 | print; |
| 2466 | } |
| 2467 | close SELF; |
| 2468 | |
| 2469 | print $code; |
| 2470 | |
| 2471 | close STDOUT; |