| 1 | #! /usr/bin/env perl
|
|---|
| 2 | # Copyright 2013-2020 The OpenSSL Project Authors. All Rights Reserved.
|
|---|
| 3 | #
|
|---|
| 4 | # Licensed under the Apache License 2.0 (the "License"). You may not use
|
|---|
| 5 | # this file except in compliance with the License. You can obtain a copy
|
|---|
| 6 | # in the file LICENSE in the source distribution or at
|
|---|
| 7 | # https://www.openssl.org/source/license.html
|
|---|
| 8 |
|
|---|
| 9 |
|
|---|
| 10 | ######################################################################
|
|---|
| 11 | ## Constant-time SSSE3 AES core implementation.
|
|---|
| 12 | ## version 0.1
|
|---|
| 13 | ##
|
|---|
| 14 | ## By Mike Hamburg (Stanford University), 2009
|
|---|
| 15 | ## Public domain.
|
|---|
| 16 | ##
|
|---|
| 17 | ## For details see http://shiftleft.org/papers/vector_aes/ and
|
|---|
| 18 | ## http://crypto.stanford.edu/vpaes/.
|
|---|
| 19 |
|
|---|
| 20 | # CBC encrypt/decrypt performance in cycles per byte processed with
|
|---|
| 21 | # 128-bit key.
|
|---|
| 22 | #
|
|---|
| 23 | # aes-ppc.pl this
|
|---|
| 24 | # PPC74x0/G4e 35.5/52.1/(23.8) 11.9(*)/15.4
|
|---|
| 25 | # PPC970/G5 37.9/55.0/(28.5) 22.2/28.5
|
|---|
| 26 | # POWER6 42.7/54.3/(28.2) 63.0/92.8(**)
|
|---|
| 27 | # POWER7 32.3/42.9/(18.4) 18.5/23.3
|
|---|
| 28 | #
|
|---|
| 29 | # (*) This is ~10% worse than reported in paper. The reason is
|
|---|
| 30 | # twofold. This module doesn't make any assumption about
|
|---|
| 31 | # key schedule (or data for that matter) alignment and handles
|
|---|
| 32 | # it in-line. Secondly it, being transliterated from
|
|---|
| 33 | # vpaes-x86_64.pl, relies on "nested inversion" better suited
|
|---|
| 34 | # for Intel CPUs.
|
|---|
| 35 | # (**) Inadequate POWER6 performance is due to astronomic AltiVec
|
|---|
| 36 | # latency, 9 cycles per simple logical operation.
|
|---|
| 37 |
|
|---|
| 38 | # $output is the last argument if it looks like a file (it has an extension)
|
|---|
| 39 | # $flavour is the first argument if it doesn't look like a file
|
|---|
| 40 | $output = $#ARGV >= 0 && $ARGV[$#ARGV] =~ m|\.\w+$| ? pop : undef;
|
|---|
| 41 | $flavour = $#ARGV >= 0 && $ARGV[0] !~ m|\.| ? shift : undef;
|
|---|
| 42 |
|
|---|
| 43 | if ($flavour =~ /64/) {
|
|---|
| 44 | $SIZE_T =8;
|
|---|
| 45 | $LRSAVE =2*$SIZE_T;
|
|---|
| 46 | $STU ="stdu";
|
|---|
| 47 | $POP ="ld";
|
|---|
| 48 | $PUSH ="std";
|
|---|
| 49 | $UCMP ="cmpld";
|
|---|
| 50 | } elsif ($flavour =~ /32/) {
|
|---|
| 51 | $SIZE_T =4;
|
|---|
| 52 | $LRSAVE =$SIZE_T;
|
|---|
| 53 | $STU ="stwu";
|
|---|
| 54 | $POP ="lwz";
|
|---|
| 55 | $PUSH ="stw";
|
|---|
| 56 | $UCMP ="cmplw";
|
|---|
| 57 | } else { die "nonsense $flavour"; }
|
|---|
| 58 |
|
|---|
| 59 | $sp="r1";
|
|---|
| 60 | $FRAME=6*$SIZE_T+13*16; # 13*16 is for v20-v31 offload
|
|---|
| 61 |
|
|---|
| 62 | $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
|
|---|
| 63 | ( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
|
|---|
| 64 | ( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
|
|---|
| 65 | die "can't locate ppc-xlate.pl";
|
|---|
| 66 |
|
|---|
| 67 | open STDOUT,"| $^X $xlate $flavour \"$output\""
|
|---|
| 68 | || die "can't call $xlate: $!";
|
|---|
| 69 |
|
|---|
| 70 | $code.=<<___;
|
|---|
| 71 | .machine "any"
|
|---|
| 72 |
|
|---|
| 73 | .text
|
|---|
| 74 |
|
|---|
| 75 | .align 7 # totally strategic alignment
|
|---|
| 76 | _vpaes_consts:
|
|---|
| 77 | Lk_mc_forward: # mc_forward
|
|---|
| 78 | .long 0x01020300, 0x05060704, 0x090a0b08, 0x0d0e0f0c ?inv
|
|---|
| 79 | .long 0x05060704, 0x090a0b08, 0x0d0e0f0c, 0x01020300 ?inv
|
|---|
| 80 | .long 0x090a0b08, 0x0d0e0f0c, 0x01020300, 0x05060704 ?inv
|
|---|
| 81 | .long 0x0d0e0f0c, 0x01020300, 0x05060704, 0x090a0b08 ?inv
|
|---|
| 82 | Lk_mc_backward: # mc_backward
|
|---|
| 83 | .long 0x03000102, 0x07040506, 0x0b08090a, 0x0f0c0d0e ?inv
|
|---|
| 84 | .long 0x0f0c0d0e, 0x03000102, 0x07040506, 0x0b08090a ?inv
|
|---|
| 85 | .long 0x0b08090a, 0x0f0c0d0e, 0x03000102, 0x07040506 ?inv
|
|---|
| 86 | .long 0x07040506, 0x0b08090a, 0x0f0c0d0e, 0x03000102 ?inv
|
|---|
| 87 | Lk_sr: # sr
|
|---|
| 88 | .long 0x00010203, 0x04050607, 0x08090a0b, 0x0c0d0e0f ?inv
|
|---|
| 89 | .long 0x00050a0f, 0x04090e03, 0x080d0207, 0x0c01060b ?inv
|
|---|
| 90 | .long 0x0009020b, 0x040d060f, 0x08010a03, 0x0c050e07 ?inv
|
|---|
| 91 | .long 0x000d0a07, 0x04010e0b, 0x0805020f, 0x0c090603 ?inv
|
|---|
| 92 |
|
|---|
| 93 | ##
|
|---|
| 94 | ## "Hot" constants
|
|---|
| 95 | ##
|
|---|
| 96 | Lk_inv: # inv, inva
|
|---|
| 97 | .long 0xf001080d, 0x0f06050e, 0x020c0b0a, 0x09030704 ?rev
|
|---|
| 98 | .long 0xf0070b0f, 0x060a0401, 0x09080502, 0x0c0e0d03 ?rev
|
|---|
| 99 | Lk_ipt: # input transform (lo, hi)
|
|---|
| 100 | .long 0x00702a5a, 0x98e8b2c2, 0x08782252, 0x90e0baca ?rev
|
|---|
| 101 | .long 0x004d7c31, 0x7d30014c, 0x81ccfdb0, 0xfcb180cd ?rev
|
|---|
| 102 | Lk_sbo: # sbou, sbot
|
|---|
| 103 | .long 0x00c7bd6f, 0x176dd2d0, 0x78a802c5, 0x7abfaa15 ?rev
|
|---|
| 104 | .long 0x006abb5f, 0xa574e4cf, 0xfa352b41, 0xd1901e8e ?rev
|
|---|
| 105 | Lk_sb1: # sb1u, sb1t
|
|---|
| 106 | .long 0x0023e2fa, 0x15d41836, 0xefd92e0d, 0xc1ccf73b ?rev
|
|---|
| 107 | .long 0x003e50cb, 0x8fe19bb1, 0x44f52a14, 0x6e7adfa5 ?rev
|
|---|
| 108 | Lk_sb2: # sb2u, sb2t
|
|---|
| 109 | .long 0x0029e10a, 0x4088eb69, 0x4a2382ab, 0xc863a1c2 ?rev
|
|---|
| 110 | .long 0x0024710b, 0xc6937ae2, 0xcd2f98bc, 0x55e9b75e ?rev
|
|---|
| 111 |
|
|---|
| 112 | ##
|
|---|
| 113 | ## Decryption stuff
|
|---|
| 114 | ##
|
|---|
| 115 | Lk_dipt: # decryption input transform
|
|---|
| 116 | .long 0x005f540b, 0x045b500f, 0x1a454e11, 0x1e414a15 ?rev
|
|---|
| 117 | .long 0x00650560, 0xe683e386, 0x94f191f4, 0x72177712 ?rev
|
|---|
| 118 | Lk_dsbo: # decryption sbox final output
|
|---|
| 119 | .long 0x0040f97e, 0x53ea8713, 0x2d3e94d4, 0xb96daac7 ?rev
|
|---|
| 120 | .long 0x001d4493, 0x0f56d712, 0x9c8ec5d8, 0x59814bca ?rev
|
|---|
| 121 | Lk_dsb9: # decryption sbox output *9*u, *9*t
|
|---|
| 122 | .long 0x00d6869a, 0x53031c85, 0xc94c994f, 0x501fd5ca ?rev
|
|---|
| 123 | .long 0x0049d7ec, 0x89173bc0, 0x65a5fbb2, 0x9e2c5e72 ?rev
|
|---|
| 124 | Lk_dsbd: # decryption sbox output *D*u, *D*t
|
|---|
| 125 | .long 0x00a2b1e6, 0xdfcc577d, 0x39442a88, 0x139b6ef5 ?rev
|
|---|
| 126 | .long 0x00cbc624, 0xf7fae23c, 0xd3efde15, 0x0d183129 ?rev
|
|---|
| 127 | Lk_dsbb: # decryption sbox output *B*u, *B*t
|
|---|
| 128 | .long 0x0042b496, 0x926422d0, 0x04d4f2b0, 0xf6462660 ?rev
|
|---|
| 129 | .long 0x006759cd, 0xa69894c1, 0x6baa5532, 0x3e0cfff3 ?rev
|
|---|
| 130 | Lk_dsbe: # decryption sbox output *E*u, *E*t
|
|---|
| 131 | .long 0x00d0d426, 0x9692f246, 0xb0f6b464, 0x04604222 ?rev
|
|---|
| 132 | .long 0x00c1aaff, 0xcda6550c, 0x323e5998, 0x6bf36794 ?rev
|
|---|
| 133 |
|
|---|
| 134 | ##
|
|---|
| 135 | ## Key schedule constants
|
|---|
| 136 | ##
|
|---|
| 137 | Lk_dksd: # decryption key schedule: invskew x*D
|
|---|
| 138 | .long 0x0047e4a3, 0x5d1ab9fe, 0xf9be1d5a, 0xa4e34007 ?rev
|
|---|
| 139 | .long 0x008336b5, 0xf477c241, 0x1e9d28ab, 0xea69dc5f ?rev
|
|---|
| 140 | Lk_dksb: # decryption key schedule: invskew x*B
|
|---|
| 141 | .long 0x00d55085, 0x1fca4f9a, 0x994cc91c, 0x8653d603 ?rev
|
|---|
| 142 | .long 0x004afcb6, 0xa7ed5b11, 0xc882347e, 0x6f2593d9 ?rev
|
|---|
| 143 | Lk_dkse: # decryption key schedule: invskew x*E + 0x63
|
|---|
| 144 | .long 0x00d6c91f, 0xca1c03d5, 0x86504f99, 0x4c9a8553 ?rev
|
|---|
| 145 | .long 0xe87bdc4f, 0x059631a2, 0x8714b320, 0x6af95ecd ?rev
|
|---|
| 146 | Lk_dks9: # decryption key schedule: invskew x*9
|
|---|
| 147 | .long 0x00a7d97e, 0xc86f11b6, 0xfc5b2582, 0x3493ed4a ?rev
|
|---|
| 148 | .long 0x00331427, 0x62517645, 0xcefddae9, 0xac9fb88b ?rev
|
|---|
| 149 |
|
|---|
| 150 | Lk_rcon: # rcon
|
|---|
| 151 | .long 0xb6ee9daf, 0xb991831f, 0x817d7c4d, 0x08982a70 ?asis
|
|---|
| 152 | Lk_s63:
|
|---|
| 153 | .long 0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b ?asis
|
|---|
| 154 |
|
|---|
| 155 | Lk_opt: # output transform
|
|---|
| 156 | .long 0x0060b6d6, 0x29499fff, 0x0868bede, 0x214197f7 ?rev
|
|---|
| 157 | .long 0x00ecbc50, 0x51bded01, 0xe00c5cb0, 0xb15d0de1 ?rev
|
|---|
| 158 | Lk_deskew: # deskew tables: inverts the sbox's "skew"
|
|---|
| 159 | .long 0x00e3a447, 0x40a3e407, 0x1af9be5d, 0x5ab9fe1d ?rev
|
|---|
| 160 | .long 0x0069ea83, 0xdcb5365f, 0x771e9df4, 0xabc24128 ?rev
|
|---|
| 161 | .align 5
|
|---|
| 162 | Lconsts:
|
|---|
| 163 | mflr r0
|
|---|
| 164 | bcl 20,31,\$+4
|
|---|
| 165 | mflr r12 #vvvvv "distance between . and _vpaes_consts
|
|---|
| 166 | addi r12,r12,-0x308
|
|---|
| 167 | mtlr r0
|
|---|
| 168 | blr
|
|---|
| 169 | .long 0
|
|---|
| 170 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 171 | .asciz "Vector Permutation AES for AltiVec, Mike Hamburg (Stanford University)"
|
|---|
| 172 | .align 6
|
|---|
| 173 | ___
|
|---|
| 174 | |
|---|
| 175 |
|
|---|
| 176 | my ($inptail,$inpperm,$outhead,$outperm,$outmask,$keyperm) = map("v$_",(26..31));
|
|---|
| 177 | {
|
|---|
| 178 | my ($inp,$out,$key) = map("r$_",(3..5));
|
|---|
| 179 |
|
|---|
| 180 | my ($invlo,$invhi,$iptlo,$ipthi,$sbou,$sbot) = map("v$_",(10..15));
|
|---|
| 181 | my ($sb1u,$sb1t,$sb2u,$sb2t) = map("v$_",(16..19));
|
|---|
| 182 | my ($sb9u,$sb9t,$sbdu,$sbdt,$sbbu,$sbbt,$sbeu,$sbet)=map("v$_",(16..23));
|
|---|
| 183 |
|
|---|
| 184 | $code.=<<___;
|
|---|
| 185 | ##
|
|---|
| 186 | ## _aes_preheat
|
|---|
| 187 | ##
|
|---|
| 188 | ## Fills register %r10 -> .aes_consts (so you can -fPIC)
|
|---|
| 189 | ## and %xmm9-%xmm15 as specified below.
|
|---|
| 190 | ##
|
|---|
| 191 | .align 4
|
|---|
| 192 | _vpaes_encrypt_preheat:
|
|---|
| 193 | mflr r8
|
|---|
| 194 | bl Lconsts
|
|---|
| 195 | mtlr r8
|
|---|
| 196 | li r11, 0xc0 # Lk_inv
|
|---|
| 197 | li r10, 0xd0
|
|---|
| 198 | li r9, 0xe0 # Lk_ipt
|
|---|
| 199 | li r8, 0xf0
|
|---|
| 200 | vxor v7, v7, v7 # 0x00..00
|
|---|
| 201 | vspltisb v8,4 # 0x04..04
|
|---|
| 202 | vspltisb v9,0x0f # 0x0f..0f
|
|---|
| 203 | lvx $invlo, r12, r11
|
|---|
| 204 | li r11, 0x100
|
|---|
| 205 | lvx $invhi, r12, r10
|
|---|
| 206 | li r10, 0x110
|
|---|
| 207 | lvx $iptlo, r12, r9
|
|---|
| 208 | li r9, 0x120
|
|---|
| 209 | lvx $ipthi, r12, r8
|
|---|
| 210 | li r8, 0x130
|
|---|
| 211 | lvx $sbou, r12, r11
|
|---|
| 212 | li r11, 0x140
|
|---|
| 213 | lvx $sbot, r12, r10
|
|---|
| 214 | li r10, 0x150
|
|---|
| 215 | lvx $sb1u, r12, r9
|
|---|
| 216 | lvx $sb1t, r12, r8
|
|---|
| 217 | lvx $sb2u, r12, r11
|
|---|
| 218 | lvx $sb2t, r12, r10
|
|---|
| 219 | blr
|
|---|
| 220 | .long 0
|
|---|
| 221 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 222 |
|
|---|
| 223 | ##
|
|---|
| 224 | ## _aes_encrypt_core
|
|---|
| 225 | ##
|
|---|
| 226 | ## AES-encrypt %xmm0.
|
|---|
| 227 | ##
|
|---|
| 228 | ## Inputs:
|
|---|
| 229 | ## %xmm0 = input
|
|---|
| 230 | ## %xmm9-%xmm15 as in _vpaes_preheat
|
|---|
| 231 | ## (%rdx) = scheduled keys
|
|---|
| 232 | ##
|
|---|
| 233 | ## Output in %xmm0
|
|---|
| 234 | ## Clobbers %xmm1-%xmm6, %r9, %r10, %r11, %rax
|
|---|
| 235 | ##
|
|---|
| 236 | ##
|
|---|
| 237 | .align 5
|
|---|
| 238 | _vpaes_encrypt_core:
|
|---|
| 239 | lwz r8, 240($key) # pull rounds
|
|---|
| 240 | li r9, 16
|
|---|
| 241 | lvx v5, 0, $key # vmovdqu (%r9), %xmm5 # round0 key
|
|---|
| 242 | li r11, 0x10
|
|---|
| 243 | lvx v6, r9, $key
|
|---|
| 244 | addi r9, r9, 16
|
|---|
| 245 | ?vperm v5, v5, v6, $keyperm # align round key
|
|---|
| 246 | addi r10, r11, 0x40
|
|---|
| 247 | vsrb v1, v0, v8 # vpsrlb \$4, %xmm0, %xmm0
|
|---|
| 248 | vperm v0, $iptlo, $iptlo, v0 # vpshufb %xmm1, %xmm2, %xmm1
|
|---|
| 249 | vperm v1, $ipthi, $ipthi, v1 # vpshufb %xmm0, %xmm3, %xmm2
|
|---|
| 250 | vxor v0, v0, v5 # vpxor %xmm5, %xmm1, %xmm0
|
|---|
| 251 | vxor v0, v0, v1 # vpxor %xmm2, %xmm0, %xmm0
|
|---|
| 252 | mtctr r8
|
|---|
| 253 | b Lenc_entry
|
|---|
| 254 |
|
|---|
| 255 | .align 4
|
|---|
| 256 | Lenc_loop:
|
|---|
| 257 | # middle of middle round
|
|---|
| 258 | vperm v4, $sb1t, v7, v2 # vpshufb %xmm2, %xmm13, %xmm4 # 4 = sb1u
|
|---|
| 259 | lvx v1, r12, r11 # vmovdqa -0x40(%r11,%r10), %xmm1 # .Lk_mc_forward[]
|
|---|
| 260 | addi r11, r11, 16
|
|---|
| 261 | vperm v0, $sb1u, v7, v3 # vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t
|
|---|
| 262 | vxor v4, v4, v5 # vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k
|
|---|
| 263 | andi. r11, r11, 0x30 # and \$0x30, %r11 # ... mod 4
|
|---|
| 264 | vperm v5, $sb2t, v7, v2 # vpshufb %xmm2, %xmm15, %xmm5 # 4 = sb2u
|
|---|
| 265 | vxor v0, v0, v4 # vpxor %xmm4, %xmm0, %xmm0 # 0 = A
|
|---|
| 266 | vperm v2, $sb2u, v7, v3 # vpshufb %xmm3, %xmm14, %xmm2 # 2 = sb2t
|
|---|
| 267 | lvx v4, r12, r10 # vmovdqa (%r11,%r10), %xmm4 # .Lk_mc_backward[]
|
|---|
| 268 | addi r10, r11, 0x40
|
|---|
| 269 | vperm v3, v0, v7, v1 # vpshufb %xmm1, %xmm0, %xmm3 # 0 = B
|
|---|
| 270 | vxor v2, v2, v5 # vpxor %xmm5, %xmm2, %xmm2 # 2 = 2A
|
|---|
| 271 | vperm v0, v0, v7, v4 # vpshufb %xmm4, %xmm0, %xmm0 # 3 = D
|
|---|
| 272 | vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3 # 0 = 2A+B
|
|---|
| 273 | vperm v4, v3, v7, v1 # vpshufb %xmm1, %xmm3, %xmm4 # 0 = 2B+C
|
|---|
| 274 | vxor v0, v0, v3 # vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D
|
|---|
| 275 | vxor v0, v0, v4 # vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D
|
|---|
| 276 |
|
|---|
| 277 | Lenc_entry:
|
|---|
| 278 | # top of round
|
|---|
| 279 | vsrb v1, v0, v8 # vpsrlb \$4, %xmm0, %xmm0 # 1 = i
|
|---|
| 280 | vperm v5, $invhi, $invhi, v0 # vpshufb %xmm1, %xmm11, %xmm5 # 2 = a/k
|
|---|
| 281 | vxor v0, v0, v1 # vpxor %xmm0, %xmm1, %xmm1 # 0 = j
|
|---|
| 282 | vperm v3, $invlo, $invlo, v1 # vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
|
|---|
| 283 | vperm v4, $invlo, $invlo, v0 # vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j
|
|---|
| 284 | vand v0, v0, v9
|
|---|
| 285 | vxor v3, v3, v5 # vpxor %xmm5, %xmm3, %xmm3 # 3 = iak = 1/i + a/k
|
|---|
| 286 | vxor v4, v4, v5 # vpxor %xmm5, %xmm4, %xmm4 # 4 = jak = 1/j + a/k
|
|---|
| 287 | vperm v2, $invlo, v7, v3 # vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak
|
|---|
| 288 | vmr v5, v6
|
|---|
| 289 | lvx v6, r9, $key # vmovdqu (%r9), %xmm5
|
|---|
| 290 | vperm v3, $invlo, v7, v4 # vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak
|
|---|
| 291 | addi r9, r9, 16
|
|---|
| 292 | vxor v2, v2, v0 # vpxor %xmm1, %xmm2, %xmm2 # 2 = io
|
|---|
| 293 | ?vperm v5, v5, v6, $keyperm # align round key
|
|---|
| 294 | vxor v3, v3, v1 # vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
|
|---|
| 295 | bdnz Lenc_loop
|
|---|
| 296 |
|
|---|
| 297 | # middle of last round
|
|---|
| 298 | addi r10, r11, 0x80
|
|---|
| 299 | # vmovdqa -0x60(%r10), %xmm4 # 3 : sbou .Lk_sbo
|
|---|
| 300 | # vmovdqa -0x50(%r10), %xmm0 # 0 : sbot .Lk_sbo+16
|
|---|
| 301 | vperm v4, $sbou, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou
|
|---|
| 302 | lvx v1, r12, r10 # vmovdqa 0x40(%r11,%r10), %xmm1 # .Lk_sr[]
|
|---|
| 303 | vperm v0, $sbot, v7, v3 # vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t
|
|---|
| 304 | vxor v4, v4, v5 # vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k
|
|---|
| 305 | vxor v0, v0, v4 # vpxor %xmm4, %xmm0, %xmm0 # 0 = A
|
|---|
| 306 | vperm v0, v0, v7, v1 # vpshufb %xmm1, %xmm0, %xmm0
|
|---|
| 307 | blr
|
|---|
| 308 | .long 0
|
|---|
| 309 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 310 |
|
|---|
| 311 | .globl .vpaes_encrypt
|
|---|
| 312 | .align 5
|
|---|
| 313 | .vpaes_encrypt:
|
|---|
| 314 | $STU $sp,-$FRAME($sp)
|
|---|
| 315 | li r10,`15+6*$SIZE_T`
|
|---|
| 316 | li r11,`31+6*$SIZE_T`
|
|---|
| 317 | mflr r6
|
|---|
| 318 | mfspr r7, 256 # save vrsave
|
|---|
| 319 | stvx v20,r10,$sp
|
|---|
| 320 | addi r10,r10,32
|
|---|
| 321 | stvx v21,r11,$sp
|
|---|
| 322 | addi r11,r11,32
|
|---|
| 323 | stvx v22,r10,$sp
|
|---|
| 324 | addi r10,r10,32
|
|---|
| 325 | stvx v23,r11,$sp
|
|---|
| 326 | addi r11,r11,32
|
|---|
| 327 | stvx v24,r10,$sp
|
|---|
| 328 | addi r10,r10,32
|
|---|
| 329 | stvx v25,r11,$sp
|
|---|
| 330 | addi r11,r11,32
|
|---|
| 331 | stvx v26,r10,$sp
|
|---|
| 332 | addi r10,r10,32
|
|---|
| 333 | stvx v27,r11,$sp
|
|---|
| 334 | addi r11,r11,32
|
|---|
| 335 | stvx v28,r10,$sp
|
|---|
| 336 | addi r10,r10,32
|
|---|
| 337 | stvx v29,r11,$sp
|
|---|
| 338 | addi r11,r11,32
|
|---|
| 339 | stvx v30,r10,$sp
|
|---|
| 340 | stvx v31,r11,$sp
|
|---|
| 341 | stw r7,`$FRAME-4`($sp) # save vrsave
|
|---|
| 342 | li r0, -1
|
|---|
| 343 | $PUSH r6,`$FRAME+$LRSAVE`($sp)
|
|---|
| 344 | mtspr 256, r0 # preserve all AltiVec registers
|
|---|
| 345 |
|
|---|
| 346 | bl _vpaes_encrypt_preheat
|
|---|
| 347 |
|
|---|
| 348 | ?lvsl $inpperm, 0, $inp # prepare for unaligned access
|
|---|
| 349 | lvx v0, 0, $inp
|
|---|
| 350 | addi $inp, $inp, 15 # 15 is not a typo
|
|---|
| 351 | ?lvsr $outperm, 0, $out
|
|---|
| 352 | ?lvsl $keyperm, 0, $key # prepare for unaligned access
|
|---|
| 353 | lvx $inptail, 0, $inp # redundant in aligned case
|
|---|
| 354 | ?vperm v0, v0, $inptail, $inpperm
|
|---|
| 355 |
|
|---|
| 356 | bl _vpaes_encrypt_core
|
|---|
| 357 |
|
|---|
| 358 | andi. r8, $out, 15
|
|---|
| 359 | li r9, 16
|
|---|
| 360 | beq Lenc_out_aligned
|
|---|
| 361 |
|
|---|
| 362 | vperm v0, v0, v0, $outperm # rotate right/left
|
|---|
| 363 | mtctr r9
|
|---|
| 364 | Lenc_out_unaligned:
|
|---|
| 365 | stvebx v0, 0, $out
|
|---|
| 366 | addi $out, $out, 1
|
|---|
| 367 | bdnz Lenc_out_unaligned
|
|---|
| 368 | b Lenc_done
|
|---|
| 369 |
|
|---|
| 370 | .align 4
|
|---|
| 371 | Lenc_out_aligned:
|
|---|
| 372 | stvx v0, 0, $out
|
|---|
| 373 | Lenc_done:
|
|---|
| 374 |
|
|---|
| 375 | li r10,`15+6*$SIZE_T`
|
|---|
| 376 | li r11,`31+6*$SIZE_T`
|
|---|
| 377 | mtlr r6
|
|---|
| 378 | mtspr 256, r7 # restore vrsave
|
|---|
| 379 | lvx v20,r10,$sp
|
|---|
| 380 | addi r10,r10,32
|
|---|
| 381 | lvx v21,r11,$sp
|
|---|
| 382 | addi r11,r11,32
|
|---|
| 383 | lvx v22,r10,$sp
|
|---|
| 384 | addi r10,r10,32
|
|---|
| 385 | lvx v23,r11,$sp
|
|---|
| 386 | addi r11,r11,32
|
|---|
| 387 | lvx v24,r10,$sp
|
|---|
| 388 | addi r10,r10,32
|
|---|
| 389 | lvx v25,r11,$sp
|
|---|
| 390 | addi r11,r11,32
|
|---|
| 391 | lvx v26,r10,$sp
|
|---|
| 392 | addi r10,r10,32
|
|---|
| 393 | lvx v27,r11,$sp
|
|---|
| 394 | addi r11,r11,32
|
|---|
| 395 | lvx v28,r10,$sp
|
|---|
| 396 | addi r10,r10,32
|
|---|
| 397 | lvx v29,r11,$sp
|
|---|
| 398 | addi r11,r11,32
|
|---|
| 399 | lvx v30,r10,$sp
|
|---|
| 400 | lvx v31,r11,$sp
|
|---|
| 401 | addi $sp,$sp,$FRAME
|
|---|
| 402 | blr
|
|---|
| 403 | .long 0
|
|---|
| 404 | .byte 0,12,0x04,1,0x80,0,3,0
|
|---|
| 405 | .long 0
|
|---|
| 406 | .size .vpaes_encrypt,.-.vpaes_encrypt
|
|---|
| 407 |
|
|---|
| 408 | .align 4
|
|---|
| 409 | _vpaes_decrypt_preheat:
|
|---|
| 410 | mflr r8
|
|---|
| 411 | bl Lconsts
|
|---|
| 412 | mtlr r8
|
|---|
| 413 | li r11, 0xc0 # Lk_inv
|
|---|
| 414 | li r10, 0xd0
|
|---|
| 415 | li r9, 0x160 # Ldipt
|
|---|
| 416 | li r8, 0x170
|
|---|
| 417 | vxor v7, v7, v7 # 0x00..00
|
|---|
| 418 | vspltisb v8,4 # 0x04..04
|
|---|
| 419 | vspltisb v9,0x0f # 0x0f..0f
|
|---|
| 420 | lvx $invlo, r12, r11
|
|---|
| 421 | li r11, 0x180
|
|---|
| 422 | lvx $invhi, r12, r10
|
|---|
| 423 | li r10, 0x190
|
|---|
| 424 | lvx $iptlo, r12, r9
|
|---|
| 425 | li r9, 0x1a0
|
|---|
| 426 | lvx $ipthi, r12, r8
|
|---|
| 427 | li r8, 0x1b0
|
|---|
| 428 | lvx $sbou, r12, r11
|
|---|
| 429 | li r11, 0x1c0
|
|---|
| 430 | lvx $sbot, r12, r10
|
|---|
| 431 | li r10, 0x1d0
|
|---|
| 432 | lvx $sb9u, r12, r9
|
|---|
| 433 | li r9, 0x1e0
|
|---|
| 434 | lvx $sb9t, r12, r8
|
|---|
| 435 | li r8, 0x1f0
|
|---|
| 436 | lvx $sbdu, r12, r11
|
|---|
| 437 | li r11, 0x200
|
|---|
| 438 | lvx $sbdt, r12, r10
|
|---|
| 439 | li r10, 0x210
|
|---|
| 440 | lvx $sbbu, r12, r9
|
|---|
| 441 | lvx $sbbt, r12, r8
|
|---|
| 442 | lvx $sbeu, r12, r11
|
|---|
| 443 | lvx $sbet, r12, r10
|
|---|
| 444 | blr
|
|---|
| 445 | .long 0
|
|---|
| 446 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 447 |
|
|---|
| 448 | ##
|
|---|
| 449 | ## Decryption core
|
|---|
| 450 | ##
|
|---|
| 451 | ## Same API as encryption core.
|
|---|
| 452 | ##
|
|---|
| 453 | .align 4
|
|---|
| 454 | _vpaes_decrypt_core:
|
|---|
| 455 | lwz r8, 240($key) # pull rounds
|
|---|
| 456 | li r9, 16
|
|---|
| 457 | lvx v5, 0, $key # vmovdqu (%r9), %xmm4 # round0 key
|
|---|
| 458 | li r11, 0x30
|
|---|
| 459 | lvx v6, r9, $key
|
|---|
| 460 | addi r9, r9, 16
|
|---|
| 461 | ?vperm v5, v5, v6, $keyperm # align round key
|
|---|
| 462 | vsrb v1, v0, v8 # vpsrlb \$4, %xmm0, %xmm0
|
|---|
| 463 | vperm v0, $iptlo, $iptlo, v0 # vpshufb %xmm1, %xmm2, %xmm2
|
|---|
| 464 | vperm v1, $ipthi, $ipthi, v1 # vpshufb %xmm0, %xmm1, %xmm0
|
|---|
| 465 | vxor v0, v0, v5 # vpxor %xmm4, %xmm2, %xmm2
|
|---|
| 466 | vxor v0, v0, v1 # vpxor %xmm2, %xmm0, %xmm0
|
|---|
| 467 | mtctr r8
|
|---|
| 468 | b Ldec_entry
|
|---|
| 469 |
|
|---|
| 470 | .align 4
|
|---|
| 471 | Ldec_loop:
|
|---|
| 472 | #
|
|---|
| 473 | # Inverse mix columns
|
|---|
| 474 | #
|
|---|
| 475 | lvx v0, r12, r11 # v5 and v0 are flipped
|
|---|
| 476 | # vmovdqa -0x20(%r10),%xmm4 # 4 : sb9u
|
|---|
| 477 | # vmovdqa -0x10(%r10),%xmm1 # 0 : sb9t
|
|---|
| 478 | vperm v4, $sb9u, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sb9u
|
|---|
| 479 | subi r11, r11, 16
|
|---|
| 480 | vperm v1, $sb9t, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb9t
|
|---|
| 481 | andi. r11, r11, 0x30
|
|---|
| 482 | vxor v5, v5, v4 # vpxor %xmm4, %xmm0, %xmm0
|
|---|
| 483 | # vmovdqa 0x00(%r10),%xmm4 # 4 : sbdu
|
|---|
| 484 | vxor v5, v5, v1 # vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
|
|---|
| 485 | # vmovdqa 0x10(%r10),%xmm1 # 0 : sbdt
|
|---|
| 486 |
|
|---|
| 487 | vperm v4, $sbdu, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbdu
|
|---|
| 488 | vperm v5, v5, v7, v0 # vpshufb %xmm5, %xmm0, %xmm0 # MC ch
|
|---|
| 489 | vperm v1, $sbdt, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbdt
|
|---|
| 490 | vxor v5, v5, v4 # vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
|
|---|
| 491 | # vmovdqa 0x20(%r10), %xmm4 # 4 : sbbu
|
|---|
| 492 | vxor v5, v5, v1 # vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
|
|---|
| 493 | # vmovdqa 0x30(%r10), %xmm1 # 0 : sbbt
|
|---|
| 494 |
|
|---|
| 495 | vperm v4, $sbbu, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbbu
|
|---|
| 496 | vperm v5, v5, v7, v0 # vpshufb %xmm5, %xmm0, %xmm0 # MC ch
|
|---|
| 497 | vperm v1, $sbbt, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbbt
|
|---|
| 498 | vxor v5, v5, v4 # vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
|
|---|
| 499 | # vmovdqa 0x40(%r10), %xmm4 # 4 : sbeu
|
|---|
| 500 | vxor v5, v5, v1 # vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
|
|---|
| 501 | # vmovdqa 0x50(%r10), %xmm1 # 0 : sbet
|
|---|
| 502 |
|
|---|
| 503 | vperm v4, $sbeu, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbeu
|
|---|
| 504 | vperm v5, v5, v7, v0 # vpshufb %xmm5, %xmm0, %xmm0 # MC ch
|
|---|
| 505 | vperm v1, $sbet, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbet
|
|---|
| 506 | vxor v0, v5, v4 # vpxor %xmm4, %xmm0, %xmm0 # 4 = ch
|
|---|
| 507 | vxor v0, v0, v1 # vpxor %xmm1, %xmm0, %xmm0 # 0 = ch
|
|---|
| 508 |
|
|---|
| 509 | Ldec_entry:
|
|---|
| 510 | # top of round
|
|---|
| 511 | vsrb v1, v0, v8 # vpsrlb \$4, %xmm0, %xmm0 # 1 = i
|
|---|
| 512 | vperm v2, $invhi, $invhi, v0 # vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k
|
|---|
| 513 | vxor v0, v0, v1 # vpxor %xmm0, %xmm1, %xmm1 # 0 = j
|
|---|
| 514 | vperm v3, $invlo, $invlo, v1 # vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
|
|---|
| 515 | vperm v4, $invlo, $invlo, v0 # vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j
|
|---|
| 516 | vand v0, v0, v9
|
|---|
| 517 | vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k
|
|---|
| 518 | vxor v4, v4, v2 # vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k
|
|---|
| 519 | vperm v2, $invlo, v7, v3 # vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak
|
|---|
| 520 | vmr v5, v6
|
|---|
| 521 | lvx v6, r9, $key # vmovdqu (%r9), %xmm0
|
|---|
| 522 | vperm v3, $invlo, v7, v4 # vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak
|
|---|
| 523 | addi r9, r9, 16
|
|---|
| 524 | vxor v2, v2, v0 # vpxor %xmm1, %xmm2, %xmm2 # 2 = io
|
|---|
| 525 | ?vperm v5, v5, v6, $keyperm # align round key
|
|---|
| 526 | vxor v3, v3, v1 # vpxor %xmm0, %xmm3, %xmm3 # 3 = jo
|
|---|
| 527 | bdnz Ldec_loop
|
|---|
| 528 |
|
|---|
| 529 | # middle of last round
|
|---|
| 530 | addi r10, r11, 0x80
|
|---|
| 531 | # vmovdqa 0x60(%r10), %xmm4 # 3 : sbou
|
|---|
| 532 | vperm v4, $sbou, v7, v2 # vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou
|
|---|
| 533 | # vmovdqa 0x70(%r10), %xmm1 # 0 : sbot
|
|---|
| 534 | lvx v2, r12, r10 # vmovdqa -0x160(%r11), %xmm2 # .Lk_sr-.Lk_dsbd=-0x160
|
|---|
| 535 | vperm v1, $sbot, v7, v3 # vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb1t
|
|---|
| 536 | vxor v4, v4, v5 # vpxor %xmm0, %xmm4, %xmm4 # 4 = sb1u + k
|
|---|
| 537 | vxor v0, v1, v4 # vpxor %xmm4, %xmm1, %xmm0 # 0 = A
|
|---|
| 538 | vperm v0, v0, v7, v2 # vpshufb %xmm2, %xmm0, %xmm0
|
|---|
| 539 | blr
|
|---|
| 540 | .long 0
|
|---|
| 541 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 542 |
|
|---|
| 543 | .globl .vpaes_decrypt
|
|---|
| 544 | .align 5
|
|---|
| 545 | .vpaes_decrypt:
|
|---|
| 546 | $STU $sp,-$FRAME($sp)
|
|---|
| 547 | li r10,`15+6*$SIZE_T`
|
|---|
| 548 | li r11,`31+6*$SIZE_T`
|
|---|
| 549 | mflr r6
|
|---|
| 550 | mfspr r7, 256 # save vrsave
|
|---|
| 551 | stvx v20,r10,$sp
|
|---|
| 552 | addi r10,r10,32
|
|---|
| 553 | stvx v21,r11,$sp
|
|---|
| 554 | addi r11,r11,32
|
|---|
| 555 | stvx v22,r10,$sp
|
|---|
| 556 | addi r10,r10,32
|
|---|
| 557 | stvx v23,r11,$sp
|
|---|
| 558 | addi r11,r11,32
|
|---|
| 559 | stvx v24,r10,$sp
|
|---|
| 560 | addi r10,r10,32
|
|---|
| 561 | stvx v25,r11,$sp
|
|---|
| 562 | addi r11,r11,32
|
|---|
| 563 | stvx v26,r10,$sp
|
|---|
| 564 | addi r10,r10,32
|
|---|
| 565 | stvx v27,r11,$sp
|
|---|
| 566 | addi r11,r11,32
|
|---|
| 567 | stvx v28,r10,$sp
|
|---|
| 568 | addi r10,r10,32
|
|---|
| 569 | stvx v29,r11,$sp
|
|---|
| 570 | addi r11,r11,32
|
|---|
| 571 | stvx v30,r10,$sp
|
|---|
| 572 | stvx v31,r11,$sp
|
|---|
| 573 | stw r7,`$FRAME-4`($sp) # save vrsave
|
|---|
| 574 | li r0, -1
|
|---|
| 575 | $PUSH r6,`$FRAME+$LRSAVE`($sp)
|
|---|
| 576 | mtspr 256, r0 # preserve all AltiVec registers
|
|---|
| 577 |
|
|---|
| 578 | bl _vpaes_decrypt_preheat
|
|---|
| 579 |
|
|---|
| 580 | ?lvsl $inpperm, 0, $inp # prepare for unaligned access
|
|---|
| 581 | lvx v0, 0, $inp
|
|---|
| 582 | addi $inp, $inp, 15 # 15 is not a typo
|
|---|
| 583 | ?lvsr $outperm, 0, $out
|
|---|
| 584 | ?lvsl $keyperm, 0, $key
|
|---|
| 585 | lvx $inptail, 0, $inp # redundant in aligned case
|
|---|
| 586 | ?vperm v0, v0, $inptail, $inpperm
|
|---|
| 587 |
|
|---|
| 588 | bl _vpaes_decrypt_core
|
|---|
| 589 |
|
|---|
| 590 | andi. r8, $out, 15
|
|---|
| 591 | li r9, 16
|
|---|
| 592 | beq Ldec_out_aligned
|
|---|
| 593 |
|
|---|
| 594 | vperm v0, v0, v0, $outperm # rotate right/left
|
|---|
| 595 | mtctr r9
|
|---|
| 596 | Ldec_out_unaligned:
|
|---|
| 597 | stvebx v0, 0, $out
|
|---|
| 598 | addi $out, $out, 1
|
|---|
| 599 | bdnz Ldec_out_unaligned
|
|---|
| 600 | b Ldec_done
|
|---|
| 601 |
|
|---|
| 602 | .align 4
|
|---|
| 603 | Ldec_out_aligned:
|
|---|
| 604 | stvx v0, 0, $out
|
|---|
| 605 | Ldec_done:
|
|---|
| 606 |
|
|---|
| 607 | li r10,`15+6*$SIZE_T`
|
|---|
| 608 | li r11,`31+6*$SIZE_T`
|
|---|
| 609 | mtlr r6
|
|---|
| 610 | mtspr 256, r7 # restore vrsave
|
|---|
| 611 | lvx v20,r10,$sp
|
|---|
| 612 | addi r10,r10,32
|
|---|
| 613 | lvx v21,r11,$sp
|
|---|
| 614 | addi r11,r11,32
|
|---|
| 615 | lvx v22,r10,$sp
|
|---|
| 616 | addi r10,r10,32
|
|---|
| 617 | lvx v23,r11,$sp
|
|---|
| 618 | addi r11,r11,32
|
|---|
| 619 | lvx v24,r10,$sp
|
|---|
| 620 | addi r10,r10,32
|
|---|
| 621 | lvx v25,r11,$sp
|
|---|
| 622 | addi r11,r11,32
|
|---|
| 623 | lvx v26,r10,$sp
|
|---|
| 624 | addi r10,r10,32
|
|---|
| 625 | lvx v27,r11,$sp
|
|---|
| 626 | addi r11,r11,32
|
|---|
| 627 | lvx v28,r10,$sp
|
|---|
| 628 | addi r10,r10,32
|
|---|
| 629 | lvx v29,r11,$sp
|
|---|
| 630 | addi r11,r11,32
|
|---|
| 631 | lvx v30,r10,$sp
|
|---|
| 632 | lvx v31,r11,$sp
|
|---|
| 633 | addi $sp,$sp,$FRAME
|
|---|
| 634 | blr
|
|---|
| 635 | .long 0
|
|---|
| 636 | .byte 0,12,0x04,1,0x80,0,3,0
|
|---|
| 637 | .long 0
|
|---|
| 638 | .size .vpaes_decrypt,.-.vpaes_decrypt
|
|---|
| 639 |
|
|---|
| 640 | .globl .vpaes_cbc_encrypt
|
|---|
| 641 | .align 5
|
|---|
| 642 | .vpaes_cbc_encrypt:
|
|---|
| 643 | ${UCMP}i r5,16
|
|---|
| 644 | bltlr-
|
|---|
| 645 |
|
|---|
| 646 | $STU $sp,-`($FRAME+2*$SIZE_T)`($sp)
|
|---|
| 647 | mflr r0
|
|---|
| 648 | li r10,`15+6*$SIZE_T`
|
|---|
| 649 | li r11,`31+6*$SIZE_T`
|
|---|
| 650 | mfspr r12, 256
|
|---|
| 651 | stvx v20,r10,$sp
|
|---|
| 652 | addi r10,r10,32
|
|---|
| 653 | stvx v21,r11,$sp
|
|---|
| 654 | addi r11,r11,32
|
|---|
| 655 | stvx v22,r10,$sp
|
|---|
| 656 | addi r10,r10,32
|
|---|
| 657 | stvx v23,r11,$sp
|
|---|
| 658 | addi r11,r11,32
|
|---|
| 659 | stvx v24,r10,$sp
|
|---|
| 660 | addi r10,r10,32
|
|---|
| 661 | stvx v25,r11,$sp
|
|---|
| 662 | addi r11,r11,32
|
|---|
| 663 | stvx v26,r10,$sp
|
|---|
| 664 | addi r10,r10,32
|
|---|
| 665 | stvx v27,r11,$sp
|
|---|
| 666 | addi r11,r11,32
|
|---|
| 667 | stvx v28,r10,$sp
|
|---|
| 668 | addi r10,r10,32
|
|---|
| 669 | stvx v29,r11,$sp
|
|---|
| 670 | addi r11,r11,32
|
|---|
| 671 | stvx v30,r10,$sp
|
|---|
| 672 | stvx v31,r11,$sp
|
|---|
| 673 | stw r12,`$FRAME-4`($sp) # save vrsave
|
|---|
| 674 | $PUSH r30,`$FRAME+$SIZE_T*0`($sp)
|
|---|
| 675 | $PUSH r31,`$FRAME+$SIZE_T*1`($sp)
|
|---|
| 676 | li r9, -16
|
|---|
| 677 | $PUSH r0, `$FRAME+$SIZE_T*2+$LRSAVE`($sp)
|
|---|
| 678 |
|
|---|
| 679 | and r30, r5, r9 # copy length&-16
|
|---|
| 680 | andi. r9, $out, 15 # is $out aligned?
|
|---|
| 681 | mr r5, r6 # copy pointer to key
|
|---|
| 682 | mr r31, r7 # copy pointer to iv
|
|---|
| 683 | li r6, -1
|
|---|
| 684 | mcrf cr1, cr0 # put aside $out alignment flag
|
|---|
| 685 | mr r7, r12 # copy vrsave
|
|---|
| 686 | mtspr 256, r6 # preserve all AltiVec registers
|
|---|
| 687 |
|
|---|
| 688 | lvx v24, 0, r31 # load [potentially unaligned] iv
|
|---|
| 689 | li r9, 15
|
|---|
| 690 | ?lvsl $inpperm, 0, r31
|
|---|
| 691 | lvx v25, r9, r31
|
|---|
| 692 | ?vperm v24, v24, v25, $inpperm
|
|---|
| 693 |
|
|---|
| 694 | cmpwi r8, 0 # test direction
|
|---|
| 695 | neg r8, $inp # prepare for unaligned access
|
|---|
| 696 | vxor v7, v7, v7
|
|---|
| 697 | ?lvsl $keyperm, 0, $key
|
|---|
| 698 | ?lvsr $outperm, 0, $out
|
|---|
| 699 | ?lvsr $inpperm, 0, r8 # -$inp
|
|---|
| 700 | vnor $outmask, v7, v7 # 0xff..ff
|
|---|
| 701 | lvx $inptail, 0, $inp
|
|---|
| 702 | ?vperm $outmask, v7, $outmask, $outperm
|
|---|
| 703 | addi $inp, $inp, 15 # 15 is not a typo
|
|---|
| 704 |
|
|---|
| 705 | beq Lcbc_decrypt
|
|---|
| 706 |
|
|---|
| 707 | bl _vpaes_encrypt_preheat
|
|---|
| 708 | li r0, 16
|
|---|
| 709 |
|
|---|
| 710 | beq cr1, Lcbc_enc_loop # $out is aligned
|
|---|
| 711 |
|
|---|
| 712 | vmr v0, $inptail
|
|---|
| 713 | lvx $inptail, 0, $inp
|
|---|
| 714 | addi $inp, $inp, 16
|
|---|
| 715 | ?vperm v0, v0, $inptail, $inpperm
|
|---|
| 716 | vxor v0, v0, v24 # ^= iv
|
|---|
| 717 |
|
|---|
| 718 | bl _vpaes_encrypt_core
|
|---|
| 719 |
|
|---|
| 720 | andi. r8, $out, 15
|
|---|
| 721 | vmr v24, v0 # put aside iv
|
|---|
| 722 | sub r9, $out, r8
|
|---|
| 723 | vperm $outhead, v0, v0, $outperm # rotate right/left
|
|---|
| 724 |
|
|---|
| 725 | Lcbc_enc_head:
|
|---|
| 726 | stvebx $outhead, r8, r9
|
|---|
| 727 | cmpwi r8, 15
|
|---|
| 728 | addi r8, r8, 1
|
|---|
| 729 | bne Lcbc_enc_head
|
|---|
| 730 |
|
|---|
| 731 | sub. r30, r30, r0 # len -= 16
|
|---|
| 732 | addi $out, $out, 16
|
|---|
| 733 | beq Lcbc_unaligned_done
|
|---|
| 734 |
|
|---|
| 735 | Lcbc_enc_loop:
|
|---|
| 736 | vmr v0, $inptail
|
|---|
| 737 | lvx $inptail, 0, $inp
|
|---|
| 738 | addi $inp, $inp, 16
|
|---|
| 739 | ?vperm v0, v0, $inptail, $inpperm
|
|---|
| 740 | vxor v0, v0, v24 # ^= iv
|
|---|
| 741 |
|
|---|
| 742 | bl _vpaes_encrypt_core
|
|---|
| 743 |
|
|---|
| 744 | vmr v24, v0 # put aside iv
|
|---|
| 745 | sub. r30, r30, r0 # len -= 16
|
|---|
| 746 | vperm v0, v0, v0, $outperm # rotate right/left
|
|---|
| 747 | vsel v1, $outhead, v0, $outmask
|
|---|
| 748 | vmr $outhead, v0
|
|---|
| 749 | stvx v1, 0, $out
|
|---|
| 750 | addi $out, $out, 16
|
|---|
| 751 | bne Lcbc_enc_loop
|
|---|
| 752 |
|
|---|
| 753 | b Lcbc_done
|
|---|
| 754 |
|
|---|
| 755 | .align 5
|
|---|
| 756 | Lcbc_decrypt:
|
|---|
| 757 | bl _vpaes_decrypt_preheat
|
|---|
| 758 | li r0, 16
|
|---|
| 759 |
|
|---|
| 760 | beq cr1, Lcbc_dec_loop # $out is aligned
|
|---|
| 761 |
|
|---|
| 762 | vmr v0, $inptail
|
|---|
| 763 | lvx $inptail, 0, $inp
|
|---|
| 764 | addi $inp, $inp, 16
|
|---|
| 765 | ?vperm v0, v0, $inptail, $inpperm
|
|---|
| 766 | vmr v25, v0 # put aside input
|
|---|
| 767 |
|
|---|
| 768 | bl _vpaes_decrypt_core
|
|---|
| 769 |
|
|---|
| 770 | andi. r8, $out, 15
|
|---|
| 771 | vxor v0, v0, v24 # ^= iv
|
|---|
| 772 | vmr v24, v25
|
|---|
| 773 | sub r9, $out, r8
|
|---|
| 774 | vperm $outhead, v0, v0, $outperm # rotate right/left
|
|---|
| 775 |
|
|---|
| 776 | Lcbc_dec_head:
|
|---|
| 777 | stvebx $outhead, r8, r9
|
|---|
| 778 | cmpwi r8, 15
|
|---|
| 779 | addi r8, r8, 1
|
|---|
| 780 | bne Lcbc_dec_head
|
|---|
| 781 |
|
|---|
| 782 | sub. r30, r30, r0 # len -= 16
|
|---|
| 783 | addi $out, $out, 16
|
|---|
| 784 | beq Lcbc_unaligned_done
|
|---|
| 785 |
|
|---|
| 786 | Lcbc_dec_loop:
|
|---|
| 787 | vmr v0, $inptail
|
|---|
| 788 | lvx $inptail, 0, $inp
|
|---|
| 789 | addi $inp, $inp, 16
|
|---|
| 790 | ?vperm v0, v0, $inptail, $inpperm
|
|---|
| 791 | vmr v25, v0 # put aside input
|
|---|
| 792 |
|
|---|
| 793 | bl _vpaes_decrypt_core
|
|---|
| 794 |
|
|---|
| 795 | vxor v0, v0, v24 # ^= iv
|
|---|
| 796 | vmr v24, v25
|
|---|
| 797 | sub. r30, r30, r0 # len -= 16
|
|---|
| 798 | vperm v0, v0, v0, $outperm # rotate right/left
|
|---|
| 799 | vsel v1, $outhead, v0, $outmask
|
|---|
| 800 | vmr $outhead, v0
|
|---|
| 801 | stvx v1, 0, $out
|
|---|
| 802 | addi $out, $out, 16
|
|---|
| 803 | bne Lcbc_dec_loop
|
|---|
| 804 |
|
|---|
| 805 | Lcbc_done:
|
|---|
| 806 | beq cr1, Lcbc_write_iv # $out is aligned
|
|---|
| 807 |
|
|---|
| 808 | Lcbc_unaligned_done:
|
|---|
| 809 | andi. r8, $out, 15
|
|---|
| 810 | sub $out, $out, r8
|
|---|
| 811 | li r9, 0
|
|---|
| 812 | Lcbc_tail:
|
|---|
| 813 | stvebx $outhead, r9, $out
|
|---|
| 814 | addi r9, r9, 1
|
|---|
| 815 | cmpw r9, r8
|
|---|
| 816 | bne Lcbc_tail
|
|---|
| 817 |
|
|---|
| 818 | Lcbc_write_iv:
|
|---|
| 819 | neg r8, r31 # write [potentially unaligned] iv
|
|---|
| 820 | li r10, 4
|
|---|
| 821 | ?lvsl $outperm, 0, r8
|
|---|
| 822 | li r11, 8
|
|---|
| 823 | li r12, 12
|
|---|
| 824 | vperm v24, v24, v24, $outperm # rotate right/left
|
|---|
| 825 | stvewx v24, 0, r31 # ivp is at least 32-bit aligned
|
|---|
| 826 | stvewx v24, r10, r31
|
|---|
| 827 | stvewx v24, r11, r31
|
|---|
| 828 | stvewx v24, r12, r31
|
|---|
| 829 |
|
|---|
| 830 | mtspr 256, r7 # restore vrsave
|
|---|
| 831 | li r10,`15+6*$SIZE_T`
|
|---|
| 832 | li r11,`31+6*$SIZE_T`
|
|---|
| 833 | lvx v20,r10,$sp
|
|---|
| 834 | addi r10,r10,32
|
|---|
| 835 | lvx v21,r11,$sp
|
|---|
| 836 | addi r11,r11,32
|
|---|
| 837 | lvx v22,r10,$sp
|
|---|
| 838 | addi r10,r10,32
|
|---|
| 839 | lvx v23,r11,$sp
|
|---|
| 840 | addi r11,r11,32
|
|---|
| 841 | lvx v24,r10,$sp
|
|---|
| 842 | addi r10,r10,32
|
|---|
| 843 | lvx v25,r11,$sp
|
|---|
| 844 | addi r11,r11,32
|
|---|
| 845 | lvx v26,r10,$sp
|
|---|
| 846 | addi r10,r10,32
|
|---|
| 847 | lvx v27,r11,$sp
|
|---|
| 848 | addi r11,r11,32
|
|---|
| 849 | lvx v28,r10,$sp
|
|---|
| 850 | addi r10,r10,32
|
|---|
| 851 | lvx v29,r11,$sp
|
|---|
| 852 | addi r11,r11,32
|
|---|
| 853 | lvx v30,r10,$sp
|
|---|
| 854 | lvx v31,r11,$sp
|
|---|
| 855 | Lcbc_abort:
|
|---|
| 856 | $POP r0, `$FRAME+$SIZE_T*2+$LRSAVE`($sp)
|
|---|
| 857 | $POP r30,`$FRAME+$SIZE_T*0`($sp)
|
|---|
| 858 | $POP r31,`$FRAME+$SIZE_T*1`($sp)
|
|---|
| 859 | mtlr r0
|
|---|
| 860 | addi $sp,$sp,`$FRAME+$SIZE_T*2`
|
|---|
| 861 | blr
|
|---|
| 862 | .long 0
|
|---|
| 863 | .byte 0,12,0x04,1,0x80,2,6,0
|
|---|
| 864 | .long 0
|
|---|
| 865 | .size .vpaes_cbc_encrypt,.-.vpaes_cbc_encrypt
|
|---|
| 866 | ___
|
|---|
| 867 | } |
|---|
| 868 |
|
|---|
| 869 | {
|
|---|
| 870 | my ($inp,$bits,$out)=map("r$_",(3..5));
|
|---|
| 871 | my $dir="cr1";
|
|---|
| 872 | my ($invlo,$invhi,$iptlo,$ipthi,$rcon) = map("v$_",(10..13,24));
|
|---|
| 873 |
|
|---|
| 874 | $code.=<<___;
|
|---|
| 875 | ########################################################
|
|---|
| 876 | ## ##
|
|---|
| 877 | ## AES key schedule ##
|
|---|
| 878 | ## ##
|
|---|
| 879 | ########################################################
|
|---|
| 880 | .align 4
|
|---|
| 881 | _vpaes_key_preheat:
|
|---|
| 882 | mflr r8
|
|---|
| 883 | bl Lconsts
|
|---|
| 884 | mtlr r8
|
|---|
| 885 | li r11, 0xc0 # Lk_inv
|
|---|
| 886 | li r10, 0xd0
|
|---|
| 887 | li r9, 0xe0 # L_ipt
|
|---|
| 888 | li r8, 0xf0
|
|---|
| 889 |
|
|---|
| 890 | vspltisb v8,4 # 0x04..04
|
|---|
| 891 | vxor v9,v9,v9 # 0x00..00
|
|---|
| 892 | lvx $invlo, r12, r11 # Lk_inv
|
|---|
| 893 | li r11, 0x120
|
|---|
| 894 | lvx $invhi, r12, r10
|
|---|
| 895 | li r10, 0x130
|
|---|
| 896 | lvx $iptlo, r12, r9 # Lk_ipt
|
|---|
| 897 | li r9, 0x220
|
|---|
| 898 | lvx $ipthi, r12, r8
|
|---|
| 899 | li r8, 0x230
|
|---|
| 900 |
|
|---|
| 901 | lvx v14, r12, r11 # Lk_sb1
|
|---|
| 902 | li r11, 0x240
|
|---|
| 903 | lvx v15, r12, r10
|
|---|
| 904 | li r10, 0x250
|
|---|
| 905 |
|
|---|
| 906 | lvx v16, r12, r9 # Lk_dksd
|
|---|
| 907 | li r9, 0x260
|
|---|
| 908 | lvx v17, r12, r8
|
|---|
| 909 | li r8, 0x270
|
|---|
| 910 | lvx v18, r12, r11 # Lk_dksb
|
|---|
| 911 | li r11, 0x280
|
|---|
| 912 | lvx v19, r12, r10
|
|---|
| 913 | li r10, 0x290
|
|---|
| 914 | lvx v20, r12, r9 # Lk_dkse
|
|---|
| 915 | li r9, 0x2a0
|
|---|
| 916 | lvx v21, r12, r8
|
|---|
| 917 | li r8, 0x2b0
|
|---|
| 918 | lvx v22, r12, r11 # Lk_dks9
|
|---|
| 919 | lvx v23, r12, r10
|
|---|
| 920 |
|
|---|
| 921 | lvx v24, r12, r9 # Lk_rcon
|
|---|
| 922 | lvx v25, 0, r12 # Lk_mc_forward[0]
|
|---|
| 923 | lvx v26, r12, r8 # Lks63
|
|---|
| 924 | blr
|
|---|
| 925 | .long 0
|
|---|
| 926 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 927 |
|
|---|
| 928 | .align 4
|
|---|
| 929 | _vpaes_schedule_core:
|
|---|
| 930 | mflr r7
|
|---|
| 931 |
|
|---|
| 932 | bl _vpaes_key_preheat # load the tables
|
|---|
| 933 |
|
|---|
| 934 | #lvx v0, 0, $inp # vmovdqu (%rdi), %xmm0 # load key (unaligned)
|
|---|
| 935 | neg r8, $inp # prepare for unaligned access
|
|---|
| 936 | lvx v0, 0, $inp
|
|---|
| 937 | addi $inp, $inp, 15 # 15 is not typo
|
|---|
| 938 | ?lvsr $inpperm, 0, r8 # -$inp
|
|---|
| 939 | lvx v6, 0, $inp # v6 serves as inptail
|
|---|
| 940 | addi $inp, $inp, 8
|
|---|
| 941 | ?vperm v0, v0, v6, $inpperm
|
|---|
| 942 |
|
|---|
| 943 | # input transform
|
|---|
| 944 | vmr v3, v0 # vmovdqa %xmm0, %xmm3
|
|---|
| 945 | bl _vpaes_schedule_transform
|
|---|
| 946 | vmr v7, v0 # vmovdqa %xmm0, %xmm7
|
|---|
| 947 |
|
|---|
| 948 | bne $dir, Lschedule_am_decrypting
|
|---|
| 949 |
|
|---|
| 950 | # encrypting, output zeroth round key after transform
|
|---|
| 951 | li r8, 0x30 # mov \$0x30,%r8d
|
|---|
| 952 | li r9, 4
|
|---|
| 953 | li r10, 8
|
|---|
| 954 | li r11, 12
|
|---|
| 955 |
|
|---|
| 956 | ?lvsr $outperm, 0, $out # prepare for unaligned access
|
|---|
| 957 | vnor $outmask, v9, v9 # 0xff..ff
|
|---|
| 958 | ?vperm $outmask, v9, $outmask, $outperm
|
|---|
| 959 |
|
|---|
| 960 | #stvx v0, 0, $out # vmovdqu %xmm0, (%rdx)
|
|---|
| 961 | vperm $outhead, v0, v0, $outperm # rotate right/left
|
|---|
| 962 | stvewx $outhead, 0, $out # some are superfluous
|
|---|
| 963 | stvewx $outhead, r9, $out
|
|---|
| 964 | stvewx $outhead, r10, $out
|
|---|
| 965 | addi r10, r12, 0x80 # lea .Lk_sr(%rip),%r10
|
|---|
| 966 | stvewx $outhead, r11, $out
|
|---|
| 967 | b Lschedule_go
|
|---|
| 968 |
|
|---|
| 969 | Lschedule_am_decrypting:
|
|---|
| 970 | srwi r8, $bits, 1 # shr \$1,%r8d
|
|---|
| 971 | andi. r8, r8, 32 # and \$32,%r8d
|
|---|
| 972 | xori r8, r8, 32 # xor \$32,%r8d # nbits==192?0:32
|
|---|
| 973 | addi r10, r12, 0x80 # lea .Lk_sr(%rip),%r10
|
|---|
| 974 | # decrypting, output zeroth round key after shiftrows
|
|---|
| 975 | lvx v1, r8, r10 # vmovdqa (%r8,%r10), %xmm1
|
|---|
| 976 | li r9, 4
|
|---|
| 977 | li r10, 8
|
|---|
| 978 | li r11, 12
|
|---|
| 979 | vperm v4, v3, v3, v1 # vpshufb %xmm1, %xmm3, %xmm3
|
|---|
| 980 |
|
|---|
| 981 | neg r0, $out # prepare for unaligned access
|
|---|
| 982 | ?lvsl $outperm, 0, r0
|
|---|
| 983 | vnor $outmask, v9, v9 # 0xff..ff
|
|---|
| 984 | ?vperm $outmask, $outmask, v9, $outperm
|
|---|
| 985 |
|
|---|
| 986 | #stvx v4, 0, $out # vmovdqu %xmm3, (%rdx)
|
|---|
| 987 | vperm $outhead, v4, v4, $outperm # rotate right/left
|
|---|
| 988 | stvewx $outhead, 0, $out # some are superfluous
|
|---|
| 989 | stvewx $outhead, r9, $out
|
|---|
| 990 | stvewx $outhead, r10, $out
|
|---|
| 991 | addi r10, r12, 0x80 # lea .Lk_sr(%rip),%r10
|
|---|
| 992 | stvewx $outhead, r11, $out
|
|---|
| 993 | addi $out, $out, 15 # 15 is not typo
|
|---|
| 994 | xori r8, r8, 0x30 # xor \$0x30, %r8
|
|---|
| 995 |
|
|---|
| 996 | Lschedule_go:
|
|---|
| 997 | cmplwi $bits, 192 # cmp \$192, %esi
|
|---|
| 998 | bgt Lschedule_256
|
|---|
| 999 | beq Lschedule_192
|
|---|
| 1000 | # 128: fall though
|
|---|
| 1001 |
|
|---|
| 1002 | ##
|
|---|
| 1003 | ## .schedule_128
|
|---|
| 1004 | ##
|
|---|
| 1005 | ## 128-bit specific part of key schedule.
|
|---|
| 1006 | ##
|
|---|
| 1007 | ## This schedule is really simple, because all its parts
|
|---|
| 1008 | ## are accomplished by the subroutines.
|
|---|
| 1009 | ##
|
|---|
| 1010 | Lschedule_128:
|
|---|
| 1011 | li r0, 10 # mov \$10, %esi
|
|---|
| 1012 | mtctr r0
|
|---|
| 1013 |
|
|---|
| 1014 | Loop_schedule_128:
|
|---|
| 1015 | bl _vpaes_schedule_round
|
|---|
| 1016 | bdz Lschedule_mangle_last # dec %esi
|
|---|
| 1017 | bl _vpaes_schedule_mangle # write output
|
|---|
| 1018 | b Loop_schedule_128
|
|---|
| 1019 |
|
|---|
| 1020 | ##
|
|---|
| 1021 | ## .aes_schedule_192
|
|---|
| 1022 | ##
|
|---|
| 1023 | ## 192-bit specific part of key schedule.
|
|---|
| 1024 | ##
|
|---|
| 1025 | ## The main body of this schedule is the same as the 128-bit
|
|---|
| 1026 | ## schedule, but with more smearing. The long, high side is
|
|---|
| 1027 | ## stored in %xmm7 as before, and the short, low side is in
|
|---|
| 1028 | ## the high bits of %xmm6.
|
|---|
| 1029 | ##
|
|---|
| 1030 | ## This schedule is somewhat nastier, however, because each
|
|---|
| 1031 | ## round produces 192 bits of key material, or 1.5 round keys.
|
|---|
| 1032 | ## Therefore, on each cycle we do 2 rounds and produce 3 round
|
|---|
| 1033 | ## keys.
|
|---|
| 1034 | ##
|
|---|
| 1035 | .align 4
|
|---|
| 1036 | Lschedule_192:
|
|---|
| 1037 | li r0, 4 # mov \$4, %esi
|
|---|
| 1038 | lvx v0, 0, $inp
|
|---|
| 1039 | ?vperm v0, v6, v0, $inpperm
|
|---|
| 1040 | ?vsldoi v0, v3, v0, 8 # vmovdqu 8(%rdi),%xmm0 # load key part 2 (very unaligned)
|
|---|
| 1041 | bl _vpaes_schedule_transform # input transform
|
|---|
| 1042 | ?vsldoi v6, v0, v9, 8
|
|---|
| 1043 | ?vsldoi v6, v9, v6, 8 # clobber "low" side with zeros
|
|---|
| 1044 | mtctr r0
|
|---|
| 1045 |
|
|---|
| 1046 | Loop_schedule_192:
|
|---|
| 1047 | bl _vpaes_schedule_round
|
|---|
| 1048 | ?vsldoi v0, v6, v0, 8 # vpalignr \$8,%xmm6,%xmm0,%xmm0
|
|---|
| 1049 | bl _vpaes_schedule_mangle # save key n
|
|---|
| 1050 | bl _vpaes_schedule_192_smear
|
|---|
| 1051 | bl _vpaes_schedule_mangle # save key n+1
|
|---|
| 1052 | bl _vpaes_schedule_round
|
|---|
| 1053 | bdz Lschedule_mangle_last # dec %esi
|
|---|
| 1054 | bl _vpaes_schedule_mangle # save key n+2
|
|---|
| 1055 | bl _vpaes_schedule_192_smear
|
|---|
| 1056 | b Loop_schedule_192
|
|---|
| 1057 |
|
|---|
| 1058 | ##
|
|---|
| 1059 | ## .aes_schedule_256
|
|---|
| 1060 | ##
|
|---|
| 1061 | ## 256-bit specific part of key schedule.
|
|---|
| 1062 | ##
|
|---|
| 1063 | ## The structure here is very similar to the 128-bit
|
|---|
| 1064 | ## schedule, but with an additional "low side" in
|
|---|
| 1065 | ## %xmm6. The low side's rounds are the same as the
|
|---|
| 1066 | ## high side's, except no rcon and no rotation.
|
|---|
| 1067 | ##
|
|---|
| 1068 | .align 4
|
|---|
| 1069 | Lschedule_256:
|
|---|
| 1070 | li r0, 7 # mov \$7, %esi
|
|---|
| 1071 | addi $inp, $inp, 8
|
|---|
| 1072 | lvx v0, 0, $inp # vmovdqu 16(%rdi),%xmm0 # load key part 2 (unaligned)
|
|---|
| 1073 | ?vperm v0, v6, v0, $inpperm
|
|---|
| 1074 | bl _vpaes_schedule_transform # input transform
|
|---|
| 1075 | mtctr r0
|
|---|
| 1076 |
|
|---|
| 1077 | Loop_schedule_256:
|
|---|
| 1078 | bl _vpaes_schedule_mangle # output low result
|
|---|
| 1079 | vmr v6, v0 # vmovdqa %xmm0, %xmm6 # save cur_lo in xmm6
|
|---|
| 1080 |
|
|---|
| 1081 | # high round
|
|---|
| 1082 | bl _vpaes_schedule_round
|
|---|
| 1083 | bdz Lschedule_mangle_last # dec %esi
|
|---|
| 1084 | bl _vpaes_schedule_mangle
|
|---|
| 1085 |
|
|---|
| 1086 | # low round. swap xmm7 and xmm6
|
|---|
| 1087 | ?vspltw v0, v0, 3 # vpshufd \$0xFF, %xmm0, %xmm0
|
|---|
| 1088 | vmr v5, v7 # vmovdqa %xmm7, %xmm5
|
|---|
| 1089 | vmr v7, v6 # vmovdqa %xmm6, %xmm7
|
|---|
| 1090 | bl _vpaes_schedule_low_round
|
|---|
| 1091 | vmr v7, v5 # vmovdqa %xmm5, %xmm7
|
|---|
| 1092 |
|
|---|
| 1093 | b Loop_schedule_256
|
|---|
| 1094 | ##
|
|---|
| 1095 | ## .aes_schedule_mangle_last
|
|---|
| 1096 | ##
|
|---|
| 1097 | ## Mangler for last round of key schedule
|
|---|
| 1098 | ## Mangles %xmm0
|
|---|
| 1099 | ## when encrypting, outputs out(%xmm0) ^ 63
|
|---|
| 1100 | ## when decrypting, outputs unskew(%xmm0)
|
|---|
| 1101 | ##
|
|---|
| 1102 | ## Always called right before return... jumps to cleanup and exits
|
|---|
| 1103 | ##
|
|---|
| 1104 | .align 4
|
|---|
| 1105 | Lschedule_mangle_last:
|
|---|
| 1106 | # schedule last round key from xmm0
|
|---|
| 1107 | li r11, 0x2e0 # lea .Lk_deskew(%rip),%r11
|
|---|
| 1108 | li r9, 0x2f0
|
|---|
| 1109 | bne $dir, Lschedule_mangle_last_dec
|
|---|
| 1110 |
|
|---|
| 1111 | # encrypting
|
|---|
| 1112 | lvx v1, r8, r10 # vmovdqa (%r8,%r10),%xmm1
|
|---|
| 1113 | li r11, 0x2c0 # lea .Lk_opt(%rip), %r11 # prepare to output transform
|
|---|
| 1114 | li r9, 0x2d0 # prepare to output transform
|
|---|
| 1115 | vperm v0, v0, v0, v1 # vpshufb %xmm1, %xmm0, %xmm0 # output permute
|
|---|
| 1116 |
|
|---|
| 1117 | lvx $iptlo, r11, r12 # reload $ipt
|
|---|
| 1118 | lvx $ipthi, r9, r12
|
|---|
| 1119 | addi $out, $out, 16 # add \$16, %rdx
|
|---|
| 1120 | vxor v0, v0, v26 # vpxor .Lk_s63(%rip), %xmm0, %xmm0
|
|---|
| 1121 | bl _vpaes_schedule_transform # output transform
|
|---|
| 1122 |
|
|---|
| 1123 | #stvx v0, r0, $out # vmovdqu %xmm0, (%rdx) # save last key
|
|---|
| 1124 | vperm v0, v0, v0, $outperm # rotate right/left
|
|---|
| 1125 | li r10, 4
|
|---|
| 1126 | vsel v2, $outhead, v0, $outmask
|
|---|
| 1127 | li r11, 8
|
|---|
| 1128 | stvx v2, 0, $out
|
|---|
| 1129 | li r12, 12
|
|---|
| 1130 | stvewx v0, 0, $out # some (or all) are redundant
|
|---|
| 1131 | stvewx v0, r10, $out
|
|---|
| 1132 | stvewx v0, r11, $out
|
|---|
| 1133 | stvewx v0, r12, $out
|
|---|
| 1134 | b Lschedule_mangle_done
|
|---|
| 1135 |
|
|---|
| 1136 | .align 4
|
|---|
| 1137 | Lschedule_mangle_last_dec:
|
|---|
| 1138 | lvx $iptlo, r11, r12 # reload $ipt
|
|---|
| 1139 | lvx $ipthi, r9, r12
|
|---|
| 1140 | addi $out, $out, -16 # add \$-16, %rdx
|
|---|
| 1141 | vxor v0, v0, v26 # vpxor .Lk_s63(%rip), %xmm0, %xmm0
|
|---|
| 1142 | bl _vpaes_schedule_transform # output transform
|
|---|
| 1143 |
|
|---|
| 1144 | #stvx v0, r0, $out # vmovdqu %xmm0, (%rdx) # save last key
|
|---|
| 1145 | addi r9, $out, -15 # -15 is not typo
|
|---|
| 1146 | vperm v0, v0, v0, $outperm # rotate right/left
|
|---|
| 1147 | li r10, 4
|
|---|
| 1148 | vsel v2, $outhead, v0, $outmask
|
|---|
| 1149 | li r11, 8
|
|---|
| 1150 | stvx v2, 0, $out
|
|---|
| 1151 | li r12, 12
|
|---|
| 1152 | stvewx v0, 0, r9 # some (or all) are redundant
|
|---|
| 1153 | stvewx v0, r10, r9
|
|---|
| 1154 | stvewx v0, r11, r9
|
|---|
| 1155 | stvewx v0, r12, r9
|
|---|
| 1156 |
|
|---|
| 1157 |
|
|---|
| 1158 | Lschedule_mangle_done:
|
|---|
| 1159 | mtlr r7
|
|---|
| 1160 | # cleanup
|
|---|
| 1161 | vxor v0, v0, v0 # vpxor %xmm0, %xmm0, %xmm0
|
|---|
| 1162 | vxor v1, v1, v1 # vpxor %xmm1, %xmm1, %xmm1
|
|---|
| 1163 | vxor v2, v2, v2 # vpxor %xmm2, %xmm2, %xmm2
|
|---|
| 1164 | vxor v3, v3, v3 # vpxor %xmm3, %xmm3, %xmm3
|
|---|
| 1165 | vxor v4, v4, v4 # vpxor %xmm4, %xmm4, %xmm4
|
|---|
| 1166 | vxor v5, v5, v5 # vpxor %xmm5, %xmm5, %xmm5
|
|---|
| 1167 | vxor v6, v6, v6 # vpxor %xmm6, %xmm6, %xmm6
|
|---|
| 1168 | vxor v7, v7, v7 # vpxor %xmm7, %xmm7, %xmm7
|
|---|
| 1169 |
|
|---|
| 1170 | blr
|
|---|
| 1171 | .long 0
|
|---|
| 1172 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 1173 |
|
|---|
| 1174 | ##
|
|---|
| 1175 | ## .aes_schedule_192_smear
|
|---|
| 1176 | ##
|
|---|
| 1177 | ## Smear the short, low side in the 192-bit key schedule.
|
|---|
| 1178 | ##
|
|---|
| 1179 | ## Inputs:
|
|---|
| 1180 | ## %xmm7: high side, b a x y
|
|---|
| 1181 | ## %xmm6: low side, d c 0 0
|
|---|
| 1182 | ## %xmm13: 0
|
|---|
| 1183 | ##
|
|---|
| 1184 | ## Outputs:
|
|---|
| 1185 | ## %xmm6: b+c+d b+c 0 0
|
|---|
| 1186 | ## %xmm0: b+c+d b+c b a
|
|---|
| 1187 | ##
|
|---|
| 1188 | .align 4
|
|---|
| 1189 | _vpaes_schedule_192_smear:
|
|---|
| 1190 | ?vspltw v0, v7, 3
|
|---|
| 1191 | ?vsldoi v1, v9, v6, 12 # vpshufd \$0x80, %xmm6, %xmm1 # d c 0 0 -> c 0 0 0
|
|---|
| 1192 | ?vsldoi v0, v7, v0, 8 # vpshufd \$0xFE, %xmm7, %xmm0 # b a _ _ -> b b b a
|
|---|
| 1193 | vxor v6, v6, v1 # vpxor %xmm1, %xmm6, %xmm6 # -> c+d c 0 0
|
|---|
| 1194 | vxor v6, v6, v0 # vpxor %xmm0, %xmm6, %xmm6 # -> b+c+d b+c b a
|
|---|
| 1195 | vmr v0, v6
|
|---|
| 1196 | ?vsldoi v6, v6, v9, 8
|
|---|
| 1197 | ?vsldoi v6, v9, v6, 8 # clobber low side with zeros
|
|---|
| 1198 | blr
|
|---|
| 1199 | .long 0
|
|---|
| 1200 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 1201 |
|
|---|
| 1202 | ##
|
|---|
| 1203 | ## .aes_schedule_round
|
|---|
| 1204 | ##
|
|---|
| 1205 | ## Runs one main round of the key schedule on %xmm0, %xmm7
|
|---|
| 1206 | ##
|
|---|
| 1207 | ## Specifically, runs subbytes on the high dword of %xmm0
|
|---|
| 1208 | ## then rotates it by one byte and xors into the low dword of
|
|---|
| 1209 | ## %xmm7.
|
|---|
| 1210 | ##
|
|---|
| 1211 | ## Adds rcon from low byte of %xmm8, then rotates %xmm8 for
|
|---|
| 1212 | ## next rcon.
|
|---|
| 1213 | ##
|
|---|
| 1214 | ## Smears the dwords of %xmm7 by xoring the low into the
|
|---|
| 1215 | ## second low, result into third, result into highest.
|
|---|
| 1216 | ##
|
|---|
| 1217 | ## Returns results in %xmm7 = %xmm0.
|
|---|
| 1218 | ## Clobbers %xmm1-%xmm4, %r11.
|
|---|
| 1219 | ##
|
|---|
| 1220 | .align 4
|
|---|
| 1221 | _vpaes_schedule_round:
|
|---|
| 1222 | # extract rcon from xmm8
|
|---|
| 1223 | #vxor v4, v4, v4 # vpxor %xmm4, %xmm4, %xmm4
|
|---|
| 1224 | ?vsldoi v1, $rcon, v9, 15 # vpalignr \$15, %xmm8, %xmm4, %xmm1
|
|---|
| 1225 | ?vsldoi $rcon, $rcon, $rcon, 15 # vpalignr \$15, %xmm8, %xmm8, %xmm8
|
|---|
| 1226 | vxor v7, v7, v1 # vpxor %xmm1, %xmm7, %xmm7
|
|---|
| 1227 |
|
|---|
| 1228 | # rotate
|
|---|
| 1229 | ?vspltw v0, v0, 3 # vpshufd \$0xFF, %xmm0, %xmm0
|
|---|
| 1230 | ?vsldoi v0, v0, v0, 1 # vpalignr \$1, %xmm0, %xmm0, %xmm0
|
|---|
| 1231 |
|
|---|
| 1232 | # fall through...
|
|---|
| 1233 |
|
|---|
| 1234 | # low round: same as high round, but no rotation and no rcon.
|
|---|
| 1235 | _vpaes_schedule_low_round:
|
|---|
| 1236 | # smear xmm7
|
|---|
| 1237 | ?vsldoi v1, v9, v7, 12 # vpslldq \$4, %xmm7, %xmm1
|
|---|
| 1238 | vxor v7, v7, v1 # vpxor %xmm1, %xmm7, %xmm7
|
|---|
| 1239 | vspltisb v1, 0x0f # 0x0f..0f
|
|---|
| 1240 | ?vsldoi v4, v9, v7, 8 # vpslldq \$8, %xmm7, %xmm4
|
|---|
| 1241 |
|
|---|
| 1242 | # subbytes
|
|---|
| 1243 | vand v1, v1, v0 # vpand %xmm9, %xmm0, %xmm1 # 0 = k
|
|---|
| 1244 | vsrb v0, v0, v8 # vpsrlb \$4, %xmm0, %xmm0 # 1 = i
|
|---|
| 1245 | vxor v7, v7, v4 # vpxor %xmm4, %xmm7, %xmm7
|
|---|
| 1246 | vperm v2, $invhi, v9, v1 # vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k
|
|---|
| 1247 | vxor v1, v1, v0 # vpxor %xmm0, %xmm1, %xmm1 # 0 = j
|
|---|
| 1248 | vperm v3, $invlo, v9, v0 # vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i
|
|---|
| 1249 | vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k
|
|---|
| 1250 | vperm v4, $invlo, v9, v1 # vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j
|
|---|
| 1251 | vxor v7, v7, v26 # vpxor .Lk_s63(%rip), %xmm7, %xmm7
|
|---|
| 1252 | vperm v3, $invlo, v9, v3 # vpshufb %xmm3, %xmm10, %xmm3 # 2 = 1/iak
|
|---|
| 1253 | vxor v4, v4, v2 # vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k
|
|---|
| 1254 | vperm v2, $invlo, v9, v4 # vpshufb %xmm4, %xmm10, %xmm2 # 3 = 1/jak
|
|---|
| 1255 | vxor v3, v3, v1 # vpxor %xmm1, %xmm3, %xmm3 # 2 = io
|
|---|
| 1256 | vxor v2, v2, v0 # vpxor %xmm0, %xmm2, %xmm2 # 3 = jo
|
|---|
| 1257 | vperm v4, v15, v9, v3 # vpshufb %xmm3, %xmm13, %xmm4 # 4 = sbou
|
|---|
| 1258 | vperm v1, v14, v9, v2 # vpshufb %xmm2, %xmm12, %xmm1 # 0 = sb1t
|
|---|
| 1259 | vxor v1, v1, v4 # vpxor %xmm4, %xmm1, %xmm1 # 0 = sbox output
|
|---|
| 1260 |
|
|---|
| 1261 | # add in smeared stuff
|
|---|
| 1262 | vxor v0, v1, v7 # vpxor %xmm7, %xmm1, %xmm0
|
|---|
| 1263 | vxor v7, v1, v7 # vmovdqa %xmm0, %xmm7
|
|---|
| 1264 | blr
|
|---|
| 1265 | .long 0
|
|---|
| 1266 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 1267 |
|
|---|
| 1268 | ##
|
|---|
| 1269 | ## .aes_schedule_transform
|
|---|
| 1270 | ##
|
|---|
| 1271 | ## Linear-transform %xmm0 according to tables at (%r11)
|
|---|
| 1272 | ##
|
|---|
| 1273 | ## Requires that %xmm9 = 0x0F0F... as in preheat
|
|---|
| 1274 | ## Output in %xmm0
|
|---|
| 1275 | ## Clobbers %xmm2
|
|---|
| 1276 | ##
|
|---|
| 1277 | .align 4
|
|---|
| 1278 | _vpaes_schedule_transform:
|
|---|
| 1279 | #vand v1, v0, v9 # vpand %xmm9, %xmm0, %xmm1
|
|---|
| 1280 | vsrb v2, v0, v8 # vpsrlb \$4, %xmm0, %xmm0
|
|---|
| 1281 | # vmovdqa (%r11), %xmm2 # lo
|
|---|
| 1282 | vperm v0, $iptlo, $iptlo, v0 # vpshufb %xmm1, %xmm2, %xmm2
|
|---|
| 1283 | # vmovdqa 16(%r11), %xmm1 # hi
|
|---|
| 1284 | vperm v2, $ipthi, $ipthi, v2 # vpshufb %xmm0, %xmm1, %xmm0
|
|---|
| 1285 | vxor v0, v0, v2 # vpxor %xmm2, %xmm0, %xmm0
|
|---|
| 1286 | blr
|
|---|
| 1287 | .long 0
|
|---|
| 1288 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 1289 |
|
|---|
| 1290 | ##
|
|---|
| 1291 | ## .aes_schedule_mangle
|
|---|
| 1292 | ##
|
|---|
| 1293 | ## Mangle xmm0 from (basis-transformed) standard version
|
|---|
| 1294 | ## to our version.
|
|---|
| 1295 | ##
|
|---|
| 1296 | ## On encrypt,
|
|---|
| 1297 | ## xor with 0x63
|
|---|
| 1298 | ## multiply by circulant 0,1,1,1
|
|---|
| 1299 | ## apply shiftrows transform
|
|---|
| 1300 | ##
|
|---|
| 1301 | ## On decrypt,
|
|---|
| 1302 | ## xor with 0x63
|
|---|
| 1303 | ## multiply by "inverse mixcolumns" circulant E,B,D,9
|
|---|
| 1304 | ## deskew
|
|---|
| 1305 | ## apply shiftrows transform
|
|---|
| 1306 | ##
|
|---|
| 1307 | ##
|
|---|
| 1308 | ## Writes out to (%rdx), and increments or decrements it
|
|---|
| 1309 | ## Keeps track of round number mod 4 in %r8
|
|---|
| 1310 | ## Preserves xmm0
|
|---|
| 1311 | ## Clobbers xmm1-xmm5
|
|---|
| 1312 | ##
|
|---|
| 1313 | .align 4
|
|---|
| 1314 | _vpaes_schedule_mangle:
|
|---|
| 1315 | #vmr v4, v0 # vmovdqa %xmm0, %xmm4 # save xmm0 for later
|
|---|
| 1316 | # vmovdqa .Lk_mc_forward(%rip),%xmm5
|
|---|
| 1317 | bne $dir, Lschedule_mangle_dec
|
|---|
| 1318 |
|
|---|
| 1319 | # encrypting
|
|---|
| 1320 | vxor v4, v0, v26 # vpxor .Lk_s63(%rip), %xmm0, %xmm4
|
|---|
| 1321 | addi $out, $out, 16 # add \$16, %rdx
|
|---|
| 1322 | vperm v4, v4, v4, v25 # vpshufb %xmm5, %xmm4, %xmm4
|
|---|
| 1323 | vperm v1, v4, v4, v25 # vpshufb %xmm5, %xmm4, %xmm1
|
|---|
| 1324 | vperm v3, v1, v1, v25 # vpshufb %xmm5, %xmm1, %xmm3
|
|---|
| 1325 | vxor v4, v4, v1 # vpxor %xmm1, %xmm4, %xmm4
|
|---|
| 1326 | lvx v1, r8, r10 # vmovdqa (%r8,%r10), %xmm1
|
|---|
| 1327 | vxor v3, v3, v4 # vpxor %xmm4, %xmm3, %xmm3
|
|---|
| 1328 |
|
|---|
| 1329 | vperm v3, v3, v3, v1 # vpshufb %xmm1, %xmm3, %xmm3
|
|---|
| 1330 | addi r8, r8, -16 # add \$-16, %r8
|
|---|
| 1331 | andi. r8, r8, 0x30 # and \$0x30, %r8
|
|---|
| 1332 |
|
|---|
| 1333 | #stvx v3, 0, $out # vmovdqu %xmm3, (%rdx)
|
|---|
| 1334 | vperm v1, v3, v3, $outperm # rotate right/left
|
|---|
| 1335 | vsel v2, $outhead, v1, $outmask
|
|---|
| 1336 | vmr $outhead, v1
|
|---|
| 1337 | stvx v2, 0, $out
|
|---|
| 1338 | blr
|
|---|
| 1339 |
|
|---|
| 1340 | .align 4
|
|---|
| 1341 | Lschedule_mangle_dec:
|
|---|
| 1342 | # inverse mix columns
|
|---|
| 1343 | # lea .Lk_dksd(%rip),%r11
|
|---|
| 1344 | vsrb v1, v0, v8 # vpsrlb \$4, %xmm4, %xmm1 # 1 = hi
|
|---|
| 1345 | #and v4, v0, v9 # vpand %xmm9, %xmm4, %xmm4 # 4 = lo
|
|---|
| 1346 |
|
|---|
| 1347 | # vmovdqa 0x00(%r11), %xmm2
|
|---|
| 1348 | vperm v2, v16, v16, v0 # vpshufb %xmm4, %xmm2, %xmm2
|
|---|
| 1349 | # vmovdqa 0x10(%r11), %xmm3
|
|---|
| 1350 | vperm v3, v17, v17, v1 # vpshufb %xmm1, %xmm3, %xmm3
|
|---|
| 1351 | vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3
|
|---|
| 1352 | vperm v3, v3, v9, v25 # vpshufb %xmm5, %xmm3, %xmm3
|
|---|
| 1353 |
|
|---|
| 1354 | # vmovdqa 0x20(%r11), %xmm2
|
|---|
| 1355 | vperm v2, v18, v18, v0 # vpshufb %xmm4, %xmm2, %xmm2
|
|---|
| 1356 | vxor v2, v2, v3 # vpxor %xmm3, %xmm2, %xmm2
|
|---|
| 1357 | # vmovdqa 0x30(%r11), %xmm3
|
|---|
| 1358 | vperm v3, v19, v19, v1 # vpshufb %xmm1, %xmm3, %xmm3
|
|---|
| 1359 | vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3
|
|---|
| 1360 | vperm v3, v3, v9, v25 # vpshufb %xmm5, %xmm3, %xmm3
|
|---|
| 1361 |
|
|---|
| 1362 | # vmovdqa 0x40(%r11), %xmm2
|
|---|
| 1363 | vperm v2, v20, v20, v0 # vpshufb %xmm4, %xmm2, %xmm2
|
|---|
| 1364 | vxor v2, v2, v3 # vpxor %xmm3, %xmm2, %xmm2
|
|---|
| 1365 | # vmovdqa 0x50(%r11), %xmm3
|
|---|
| 1366 | vperm v3, v21, v21, v1 # vpshufb %xmm1, %xmm3, %xmm3
|
|---|
| 1367 | vxor v3, v3, v2 # vpxor %xmm2, %xmm3, %xmm3
|
|---|
| 1368 |
|
|---|
| 1369 | # vmovdqa 0x60(%r11), %xmm2
|
|---|
| 1370 | vperm v2, v22, v22, v0 # vpshufb %xmm4, %xmm2, %xmm2
|
|---|
| 1371 | vperm v3, v3, v9, v25 # vpshufb %xmm5, %xmm3, %xmm3
|
|---|
| 1372 | # vmovdqa 0x70(%r11), %xmm4
|
|---|
| 1373 | vperm v4, v23, v23, v1 # vpshufb %xmm1, %xmm4, %xmm4
|
|---|
| 1374 | lvx v1, r8, r10 # vmovdqa (%r8,%r10), %xmm1
|
|---|
| 1375 | vxor v2, v2, v3 # vpxor %xmm3, %xmm2, %xmm2
|
|---|
| 1376 | vxor v3, v4, v2 # vpxor %xmm2, %xmm4, %xmm3
|
|---|
| 1377 |
|
|---|
| 1378 | addi $out, $out, -16 # add \$-16, %rdx
|
|---|
| 1379 |
|
|---|
| 1380 | vperm v3, v3, v3, v1 # vpshufb %xmm1, %xmm3, %xmm3
|
|---|
| 1381 | addi r8, r8, -16 # add \$-16, %r8
|
|---|
| 1382 | andi. r8, r8, 0x30 # and \$0x30, %r8
|
|---|
| 1383 |
|
|---|
| 1384 | #stvx v3, 0, $out # vmovdqu %xmm3, (%rdx)
|
|---|
| 1385 | vperm v1, v3, v3, $outperm # rotate right/left
|
|---|
| 1386 | vsel v2, $outhead, v1, $outmask
|
|---|
| 1387 | vmr $outhead, v1
|
|---|
| 1388 | stvx v2, 0, $out
|
|---|
| 1389 | blr
|
|---|
| 1390 | .long 0
|
|---|
| 1391 | .byte 0,12,0x14,0,0,0,0,0
|
|---|
| 1392 |
|
|---|
| 1393 | .globl .vpaes_set_encrypt_key
|
|---|
| 1394 | .align 5
|
|---|
| 1395 | .vpaes_set_encrypt_key:
|
|---|
| 1396 | $STU $sp,-$FRAME($sp)
|
|---|
| 1397 | li r10,`15+6*$SIZE_T`
|
|---|
| 1398 | li r11,`31+6*$SIZE_T`
|
|---|
| 1399 | mflr r0
|
|---|
| 1400 | mfspr r6, 256 # save vrsave
|
|---|
| 1401 | stvx v20,r10,$sp
|
|---|
| 1402 | addi r10,r10,32
|
|---|
| 1403 | stvx v21,r11,$sp
|
|---|
| 1404 | addi r11,r11,32
|
|---|
| 1405 | stvx v22,r10,$sp
|
|---|
| 1406 | addi r10,r10,32
|
|---|
| 1407 | stvx v23,r11,$sp
|
|---|
| 1408 | addi r11,r11,32
|
|---|
| 1409 | stvx v24,r10,$sp
|
|---|
| 1410 | addi r10,r10,32
|
|---|
| 1411 | stvx v25,r11,$sp
|
|---|
| 1412 | addi r11,r11,32
|
|---|
| 1413 | stvx v26,r10,$sp
|
|---|
| 1414 | addi r10,r10,32
|
|---|
| 1415 | stvx v27,r11,$sp
|
|---|
| 1416 | addi r11,r11,32
|
|---|
| 1417 | stvx v28,r10,$sp
|
|---|
| 1418 | addi r10,r10,32
|
|---|
| 1419 | stvx v29,r11,$sp
|
|---|
| 1420 | addi r11,r11,32
|
|---|
| 1421 | stvx v30,r10,$sp
|
|---|
| 1422 | stvx v31,r11,$sp
|
|---|
| 1423 | stw r6,`$FRAME-4`($sp) # save vrsave
|
|---|
| 1424 | li r7, -1
|
|---|
| 1425 | $PUSH r0, `$FRAME+$LRSAVE`($sp)
|
|---|
| 1426 | mtspr 256, r7 # preserve all AltiVec registers
|
|---|
| 1427 |
|
|---|
| 1428 | srwi r9, $bits, 5 # shr \$5,%eax
|
|---|
| 1429 | addi r9, r9, 6 # add \$5,%eax
|
|---|
| 1430 | stw r9, 240($out) # mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5;
|
|---|
| 1431 |
|
|---|
| 1432 | cmplw $dir, $bits, $bits # set encrypt direction
|
|---|
| 1433 | li r8, 0x30 # mov \$0x30,%r8d
|
|---|
| 1434 | bl _vpaes_schedule_core
|
|---|
| 1435 |
|
|---|
| 1436 | $POP r0, `$FRAME+$LRSAVE`($sp)
|
|---|
| 1437 | li r10,`15+6*$SIZE_T`
|
|---|
| 1438 | li r11,`31+6*$SIZE_T`
|
|---|
| 1439 | mtspr 256, r6 # restore vrsave
|
|---|
| 1440 | mtlr r0
|
|---|
| 1441 | xor r3, r3, r3
|
|---|
| 1442 | lvx v20,r10,$sp
|
|---|
| 1443 | addi r10,r10,32
|
|---|
| 1444 | lvx v21,r11,$sp
|
|---|
| 1445 | addi r11,r11,32
|
|---|
| 1446 | lvx v22,r10,$sp
|
|---|
| 1447 | addi r10,r10,32
|
|---|
| 1448 | lvx v23,r11,$sp
|
|---|
| 1449 | addi r11,r11,32
|
|---|
| 1450 | lvx v24,r10,$sp
|
|---|
| 1451 | addi r10,r10,32
|
|---|
| 1452 | lvx v25,r11,$sp
|
|---|
| 1453 | addi r11,r11,32
|
|---|
| 1454 | lvx v26,r10,$sp
|
|---|
| 1455 | addi r10,r10,32
|
|---|
| 1456 | lvx v27,r11,$sp
|
|---|
| 1457 | addi r11,r11,32
|
|---|
| 1458 | lvx v28,r10,$sp
|
|---|
| 1459 | addi r10,r10,32
|
|---|
| 1460 | lvx v29,r11,$sp
|
|---|
| 1461 | addi r11,r11,32
|
|---|
| 1462 | lvx v30,r10,$sp
|
|---|
| 1463 | lvx v31,r11,$sp
|
|---|
| 1464 | addi $sp,$sp,$FRAME
|
|---|
| 1465 | blr
|
|---|
| 1466 | .long 0
|
|---|
| 1467 | .byte 0,12,0x04,1,0x80,0,3,0
|
|---|
| 1468 | .long 0
|
|---|
| 1469 | .size .vpaes_set_encrypt_key,.-.vpaes_set_encrypt_key
|
|---|
| 1470 |
|
|---|
| 1471 | .globl .vpaes_set_decrypt_key
|
|---|
| 1472 | .align 4
|
|---|
| 1473 | .vpaes_set_decrypt_key:
|
|---|
| 1474 | $STU $sp,-$FRAME($sp)
|
|---|
| 1475 | li r10,`15+6*$SIZE_T`
|
|---|
| 1476 | li r11,`31+6*$SIZE_T`
|
|---|
| 1477 | mflr r0
|
|---|
| 1478 | mfspr r6, 256 # save vrsave
|
|---|
| 1479 | stvx v20,r10,$sp
|
|---|
| 1480 | addi r10,r10,32
|
|---|
| 1481 | stvx v21,r11,$sp
|
|---|
| 1482 | addi r11,r11,32
|
|---|
| 1483 | stvx v22,r10,$sp
|
|---|
| 1484 | addi r10,r10,32
|
|---|
| 1485 | stvx v23,r11,$sp
|
|---|
| 1486 | addi r11,r11,32
|
|---|
| 1487 | stvx v24,r10,$sp
|
|---|
| 1488 | addi r10,r10,32
|
|---|
| 1489 | stvx v25,r11,$sp
|
|---|
| 1490 | addi r11,r11,32
|
|---|
| 1491 | stvx v26,r10,$sp
|
|---|
| 1492 | addi r10,r10,32
|
|---|
| 1493 | stvx v27,r11,$sp
|
|---|
| 1494 | addi r11,r11,32
|
|---|
| 1495 | stvx v28,r10,$sp
|
|---|
| 1496 | addi r10,r10,32
|
|---|
| 1497 | stvx v29,r11,$sp
|
|---|
| 1498 | addi r11,r11,32
|
|---|
| 1499 | stvx v30,r10,$sp
|
|---|
| 1500 | stvx v31,r11,$sp
|
|---|
| 1501 | stw r6,`$FRAME-4`($sp) # save vrsave
|
|---|
| 1502 | li r7, -1
|
|---|
| 1503 | $PUSH r0, `$FRAME+$LRSAVE`($sp)
|
|---|
| 1504 | mtspr 256, r7 # preserve all AltiVec registers
|
|---|
| 1505 |
|
|---|
| 1506 | srwi r9, $bits, 5 # shr \$5,%eax
|
|---|
| 1507 | addi r9, r9, 6 # add \$5,%eax
|
|---|
| 1508 | stw r9, 240($out) # mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5;
|
|---|
| 1509 |
|
|---|
| 1510 | slwi r9, r9, 4 # shl \$4,%eax
|
|---|
| 1511 | add $out, $out, r9 # lea (%rdx,%rax),%rdx
|
|---|
| 1512 |
|
|---|
| 1513 | cmplwi $dir, $bits, 0 # set decrypt direction
|
|---|
| 1514 | srwi r8, $bits, 1 # shr \$1,%r8d
|
|---|
| 1515 | andi. r8, r8, 32 # and \$32,%r8d
|
|---|
| 1516 | xori r8, r8, 32 # xor \$32,%r8d # nbits==192?0:32
|
|---|
| 1517 | bl _vpaes_schedule_core
|
|---|
| 1518 |
|
|---|
| 1519 | $POP r0, `$FRAME+$LRSAVE`($sp)
|
|---|
| 1520 | li r10,`15+6*$SIZE_T`
|
|---|
| 1521 | li r11,`31+6*$SIZE_T`
|
|---|
| 1522 | mtspr 256, r6 # restore vrsave
|
|---|
| 1523 | mtlr r0
|
|---|
| 1524 | xor r3, r3, r3
|
|---|
| 1525 | lvx v20,r10,$sp
|
|---|
| 1526 | addi r10,r10,32
|
|---|
| 1527 | lvx v21,r11,$sp
|
|---|
| 1528 | addi r11,r11,32
|
|---|
| 1529 | lvx v22,r10,$sp
|
|---|
| 1530 | addi r10,r10,32
|
|---|
| 1531 | lvx v23,r11,$sp
|
|---|
| 1532 | addi r11,r11,32
|
|---|
| 1533 | lvx v24,r10,$sp
|
|---|
| 1534 | addi r10,r10,32
|
|---|
| 1535 | lvx v25,r11,$sp
|
|---|
| 1536 | addi r11,r11,32
|
|---|
| 1537 | lvx v26,r10,$sp
|
|---|
| 1538 | addi r10,r10,32
|
|---|
| 1539 | lvx v27,r11,$sp
|
|---|
| 1540 | addi r11,r11,32
|
|---|
| 1541 | lvx v28,r10,$sp
|
|---|
| 1542 | addi r10,r10,32
|
|---|
| 1543 | lvx v29,r11,$sp
|
|---|
| 1544 | addi r11,r11,32
|
|---|
| 1545 | lvx v30,r10,$sp
|
|---|
| 1546 | lvx v31,r11,$sp
|
|---|
| 1547 | addi $sp,$sp,$FRAME
|
|---|
| 1548 | blr
|
|---|
| 1549 | .long 0
|
|---|
| 1550 | .byte 0,12,0x04,1,0x80,0,3,0
|
|---|
| 1551 | .long 0
|
|---|
| 1552 | .size .vpaes_set_decrypt_key,.-.vpaes_set_decrypt_key
|
|---|
| 1553 | ___
|
|---|
| 1554 | }
|
|---|
| 1555 |
|
|---|
| 1556 | my $consts=1;
|
|---|
| 1557 | foreach (split("\n",$code)) {
|
|---|
| 1558 | s/\`([^\`]*)\`/eval $1/geo;
|
|---|
| 1559 |
|
|---|
| 1560 | # constants table endian-specific conversion
|
|---|
| 1561 | if ($consts && m/\.long\s+(.+)\s+(\?[a-z]*)$/o) {
|
|---|
| 1562 | my $conv=$2;
|
|---|
| 1563 | my @bytes=();
|
|---|
| 1564 |
|
|---|
| 1565 | # convert to endian-agnostic format
|
|---|
| 1566 | foreach (split(/,\s+/,$1)) {
|
|---|
| 1567 | my $l = /^0/?oct:int;
|
|---|
| 1568 | push @bytes,($l>>24)&0xff,($l>>16)&0xff,($l>>8)&0xff,$l&0xff;
|
|---|
| 1569 | }
|
|---|
| 1570 |
|
|---|
| 1571 | # little-endian conversion
|
|---|
| 1572 | if ($flavour =~ /le$/o) {
|
|---|
| 1573 | SWITCH: for($conv) {
|
|---|
| 1574 | /\?inv/ && do { @bytes=map($_^0xf,@bytes); last; };
|
|---|
| 1575 | /\?rev/ && do { @bytes=reverse(@bytes); last; };
|
|---|
| 1576 | }
|
|---|
| 1577 | }
|
|---|
| 1578 |
|
|---|
| 1579 | #emit
|
|---|
| 1580 | print ".byte\t",join(',',map (sprintf("0x%02x",$_),@bytes)),"\n";
|
|---|
| 1581 | next;
|
|---|
| 1582 | }
|
|---|
| 1583 | $consts=0 if (m/Lconsts:/o); # end of table
|
|---|
| 1584 |
|
|---|
| 1585 | # instructions prefixed with '?' are endian-specific and need
|
|---|
| 1586 | # to be adjusted accordingly...
|
|---|
| 1587 | if ($flavour =~ /le$/o) { # little-endian
|
|---|
| 1588 | s/\?lvsr/lvsl/o or
|
|---|
| 1589 | s/\?lvsl/lvsr/o or
|
|---|
| 1590 | s/\?(vperm\s+v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+)/$1$3$2$4/o or
|
|---|
| 1591 | s/\?(vsldoi\s+v[0-9]+,\s*)(v[0-9]+,)\s*(v[0-9]+,\s*)([0-9]+)/$1$3$2 16-$4/o or
|
|---|
| 1592 | s/\?(vspltw\s+v[0-9]+,\s*)(v[0-9]+,)\s*([0-9])/$1$2 3-$3/o;
|
|---|
| 1593 | } else { # big-endian
|
|---|
| 1594 | s/\?([a-z]+)/$1/o;
|
|---|
| 1595 | }
|
|---|
| 1596 |
|
|---|
| 1597 | print $_,"\n";
|
|---|
| 1598 | }
|
|---|
| 1599 |
|
|---|
| 1600 | close STDOUT or die "error closing STDOUT: $!";
|
|---|
