1bc3d5698SJohn Baldwin/* Do not modify. This file is auto-generated from vpaes-armv8.pl. */ 2bd9588bcSAndrew Turner#include "arm_arch.h" 3bd9588bcSAndrew Turner 4bc3d5698SJohn Baldwin.text 5bc3d5698SJohn Baldwin 6bc3d5698SJohn Baldwin.type _vpaes_consts,%object 7bc3d5698SJohn Baldwin.align 7 // totally strategic alignment 8bc3d5698SJohn Baldwin_vpaes_consts: 9bc3d5698SJohn Baldwin.Lk_mc_forward: // mc_forward 10bc3d5698SJohn Baldwin.quad 0x0407060500030201, 0x0C0F0E0D080B0A09 11bc3d5698SJohn Baldwin.quad 0x080B0A0904070605, 0x000302010C0F0E0D 12bc3d5698SJohn Baldwin.quad 0x0C0F0E0D080B0A09, 0x0407060500030201 13bc3d5698SJohn Baldwin.quad 0x000302010C0F0E0D, 0x080B0A0904070605 14bc3d5698SJohn Baldwin.Lk_mc_backward: // mc_backward 15bc3d5698SJohn Baldwin.quad 0x0605040702010003, 0x0E0D0C0F0A09080B 16bc3d5698SJohn Baldwin.quad 0x020100030E0D0C0F, 0x0A09080B06050407 17bc3d5698SJohn Baldwin.quad 0x0E0D0C0F0A09080B, 0x0605040702010003 18bc3d5698SJohn Baldwin.quad 0x0A09080B06050407, 0x020100030E0D0C0F 19bc3d5698SJohn Baldwin.Lk_sr: // sr 20bc3d5698SJohn Baldwin.quad 0x0706050403020100, 0x0F0E0D0C0B0A0908 21bc3d5698SJohn Baldwin.quad 0x030E09040F0A0500, 0x0B06010C07020D08 22bc3d5698SJohn Baldwin.quad 0x0F060D040B020900, 0x070E050C030A0108 23bc3d5698SJohn Baldwin.quad 0x0B0E0104070A0D00, 0x0306090C0F020508 24bc3d5698SJohn Baldwin 25bc3d5698SJohn Baldwin// 26bc3d5698SJohn Baldwin// "Hot" constants 27bc3d5698SJohn Baldwin// 28bc3d5698SJohn Baldwin.Lk_inv: // inv, inva 29bc3d5698SJohn Baldwin.quad 0x0E05060F0D080180, 0x040703090A0B0C02 30bc3d5698SJohn Baldwin.quad 0x01040A060F0B0780, 0x030D0E0C02050809 31bc3d5698SJohn Baldwin.Lk_ipt: // input transform (lo, hi) 32bc3d5698SJohn Baldwin.quad 0xC2B2E8985A2A7000, 0xCABAE09052227808 33bc3d5698SJohn Baldwin.quad 0x4C01307D317C4D00, 0xCD80B1FCB0FDCC81 34bc3d5698SJohn Baldwin.Lk_sbo: // sbou, sbot 35bc3d5698SJohn Baldwin.quad 0xD0D26D176FBDC700, 0x15AABF7AC502A878 36bc3d5698SJohn Baldwin.quad 0xCFE474A55FBB6A00, 0x8E1E90D1412B35FA 37bc3d5698SJohn Baldwin.Lk_sb1: // sb1u, sb1t 38bc3d5698SJohn Baldwin.quad 0x3618D415FAE22300, 0x3BF7CCC10D2ED9EF 39bc3d5698SJohn Baldwin.quad 0xB19BE18FCB503E00, 0xA5DF7A6E142AF544 40bc3d5698SJohn Baldwin.Lk_sb2: // sb2u, sb2t 41bc3d5698SJohn Baldwin.quad 0x69EB88400AE12900, 0xC2A163C8AB82234A 42bc3d5698SJohn Baldwin.quad 0xE27A93C60B712400, 0x5EB7E955BC982FCD 43bc3d5698SJohn Baldwin 44bc3d5698SJohn Baldwin// 45bc3d5698SJohn Baldwin// Decryption stuff 46bc3d5698SJohn Baldwin// 47bc3d5698SJohn Baldwin.Lk_dipt: // decryption input transform 48bc3d5698SJohn Baldwin.quad 0x0F505B040B545F00, 0x154A411E114E451A 49bc3d5698SJohn Baldwin.quad 0x86E383E660056500, 0x12771772F491F194 50bc3d5698SJohn Baldwin.Lk_dsbo: // decryption sbox final output 51bc3d5698SJohn Baldwin.quad 0x1387EA537EF94000, 0xC7AA6DB9D4943E2D 52bc3d5698SJohn Baldwin.quad 0x12D7560F93441D00, 0xCA4B8159D8C58E9C 53bc3d5698SJohn Baldwin.Lk_dsb9: // decryption sbox output *9*u, *9*t 54bc3d5698SJohn Baldwin.quad 0x851C03539A86D600, 0xCAD51F504F994CC9 55bc3d5698SJohn Baldwin.quad 0xC03B1789ECD74900, 0x725E2C9EB2FBA565 56bc3d5698SJohn Baldwin.Lk_dsbd: // decryption sbox output *D*u, *D*t 57bc3d5698SJohn Baldwin.quad 0x7D57CCDFE6B1A200, 0xF56E9B13882A4439 58bc3d5698SJohn Baldwin.quad 0x3CE2FAF724C6CB00, 0x2931180D15DEEFD3 59bc3d5698SJohn Baldwin.Lk_dsbb: // decryption sbox output *B*u, *B*t 60bc3d5698SJohn Baldwin.quad 0xD022649296B44200, 0x602646F6B0F2D404 61bc3d5698SJohn Baldwin.quad 0xC19498A6CD596700, 0xF3FF0C3E3255AA6B 62bc3d5698SJohn Baldwin.Lk_dsbe: // decryption sbox output *E*u, *E*t 63bc3d5698SJohn Baldwin.quad 0x46F2929626D4D000, 0x2242600464B4F6B0 64bc3d5698SJohn Baldwin.quad 0x0C55A6CDFFAAC100, 0x9467F36B98593E32 65bc3d5698SJohn Baldwin 66bc3d5698SJohn Baldwin// 67bc3d5698SJohn Baldwin// Key schedule constants 68bc3d5698SJohn Baldwin// 69bc3d5698SJohn Baldwin.Lk_dksd: // decryption key schedule: invskew x*D 70bc3d5698SJohn Baldwin.quad 0xFEB91A5DA3E44700, 0x0740E3A45A1DBEF9 71bc3d5698SJohn Baldwin.quad 0x41C277F4B5368300, 0x5FDC69EAAB289D1E 72bc3d5698SJohn Baldwin.Lk_dksb: // decryption key schedule: invskew x*B 73bc3d5698SJohn Baldwin.quad 0x9A4FCA1F8550D500, 0x03D653861CC94C99 74bc3d5698SJohn Baldwin.quad 0x115BEDA7B6FC4A00, 0xD993256F7E3482C8 75bc3d5698SJohn Baldwin.Lk_dkse: // decryption key schedule: invskew x*E + 0x63 76bc3d5698SJohn Baldwin.quad 0xD5031CCA1FC9D600, 0x53859A4C994F5086 77bc3d5698SJohn Baldwin.quad 0xA23196054FDC7BE8, 0xCD5EF96A20B31487 78bc3d5698SJohn Baldwin.Lk_dks9: // decryption key schedule: invskew x*9 79bc3d5698SJohn Baldwin.quad 0xB6116FC87ED9A700, 0x4AED933482255BFC 80bc3d5698SJohn Baldwin.quad 0x4576516227143300, 0x8BB89FACE9DAFDCE 81bc3d5698SJohn Baldwin 82bc3d5698SJohn Baldwin.Lk_rcon: // rcon 83bc3d5698SJohn Baldwin.quad 0x1F8391B9AF9DEEB6, 0x702A98084D7C7D81 84bc3d5698SJohn Baldwin 85bc3d5698SJohn Baldwin.Lk_opt: // output transform 86bc3d5698SJohn Baldwin.quad 0xFF9F4929D6B66000, 0xF7974121DEBE6808 87bc3d5698SJohn Baldwin.quad 0x01EDBD5150BCEC00, 0xE10D5DB1B05C0CE0 88bc3d5698SJohn Baldwin.Lk_deskew: // deskew tables: inverts the sbox's "skew" 89bc3d5698SJohn Baldwin.quad 0x07E4A34047A4E300, 0x1DFEB95A5DBEF91A 90bc3d5698SJohn Baldwin.quad 0x5F36B5DC83EA6900, 0x2841C2ABF49D1E77 91bc3d5698SJohn Baldwin 92bc3d5698SJohn Baldwin.byte 86,101,99,116,111,114,32,80,101,114,109,117,116,97,116,105,111,110,32,65,69,83,32,102,111,114,32,65,82,77,118,56,44,32,77,105,107,101,32,72,97,109,98,117,114,103,32,40,83,116,97,110,102,111,114,100,32,85,110,105,118,101,114,115,105,116,121,41,0 93bc3d5698SJohn Baldwin.align 2 94bc3d5698SJohn Baldwin.size _vpaes_consts,.-_vpaes_consts 95bc3d5698SJohn Baldwin.align 6 96c0855eaaSJohn Baldwin// 97c0855eaaSJohn Baldwin// _aes_preheat 98c0855eaaSJohn Baldwin// 99c0855eaaSJohn Baldwin// Fills register %r10 -> .aes_consts (so you can -fPIC) 100c0855eaaSJohn Baldwin// and %xmm9-%xmm15 as specified below. 101c0855eaaSJohn Baldwin// 102bc3d5698SJohn Baldwin.type _vpaes_encrypt_preheat,%function 103bc3d5698SJohn Baldwin.align 4 104bc3d5698SJohn Baldwin_vpaes_encrypt_preheat: 105bc3d5698SJohn Baldwin adr x10, .Lk_inv 106bc3d5698SJohn Baldwin movi v17.16b, #0x0f 107bc3d5698SJohn Baldwin ld1 {v18.2d,v19.2d}, [x10],#32 // .Lk_inv 108bc3d5698SJohn Baldwin ld1 {v20.2d,v21.2d,v22.2d,v23.2d}, [x10],#64 // .Lk_ipt, .Lk_sbo 109bc3d5698SJohn Baldwin ld1 {v24.2d,v25.2d,v26.2d,v27.2d}, [x10] // .Lk_sb1, .Lk_sb2 110bc3d5698SJohn Baldwin ret 111bc3d5698SJohn Baldwin.size _vpaes_encrypt_preheat,.-_vpaes_encrypt_preheat 112bc3d5698SJohn Baldwin 113c0855eaaSJohn Baldwin// 114c0855eaaSJohn Baldwin// _aes_encrypt_core 115c0855eaaSJohn Baldwin// 116c0855eaaSJohn Baldwin// AES-encrypt %xmm0. 117c0855eaaSJohn Baldwin// 118c0855eaaSJohn Baldwin// Inputs: 119c0855eaaSJohn Baldwin// %xmm0 = input 120c0855eaaSJohn Baldwin// %xmm9-%xmm15 as in _vpaes_preheat 121c0855eaaSJohn Baldwin// (%rdx) = scheduled keys 122c0855eaaSJohn Baldwin// 123c0855eaaSJohn Baldwin// Output in %xmm0 124c0855eaaSJohn Baldwin// Clobbers %xmm1-%xmm5, %r9, %r10, %r11, %rax 125c0855eaaSJohn Baldwin// Preserves %xmm6 - %xmm8 so you get some local vectors 126c0855eaaSJohn Baldwin// 127c0855eaaSJohn Baldwin// 128bc3d5698SJohn Baldwin.type _vpaes_encrypt_core,%function 129bc3d5698SJohn Baldwin.align 4 130bc3d5698SJohn Baldwin_vpaes_encrypt_core: 131bc3d5698SJohn Baldwin mov x9, x2 132bc3d5698SJohn Baldwin ldr w8, [x2,#240] // pull rounds 133bc3d5698SJohn Baldwin adr x11, .Lk_mc_forward+16 134bc3d5698SJohn Baldwin // vmovdqa .Lk_ipt(%rip), %xmm2 # iptlo 135bc3d5698SJohn Baldwin ld1 {v16.2d}, [x9], #16 // vmovdqu (%r9), %xmm5 # round0 key 136bc3d5698SJohn Baldwin and v1.16b, v7.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 137bc3d5698SJohn Baldwin ushr v0.16b, v7.16b, #4 // vpsrlb $4, %xmm0, %xmm0 138bc3d5698SJohn Baldwin tbl v1.16b, {v20.16b}, v1.16b // vpshufb %xmm1, %xmm2, %xmm1 139bc3d5698SJohn Baldwin // vmovdqa .Lk_ipt+16(%rip), %xmm3 # ipthi 140bc3d5698SJohn Baldwin tbl v2.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm3, %xmm2 141bc3d5698SJohn Baldwin eor v0.16b, v1.16b, v16.16b // vpxor %xmm5, %xmm1, %xmm0 142bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 143bc3d5698SJohn Baldwin b .Lenc_entry 144bc3d5698SJohn Baldwin 145bc3d5698SJohn Baldwin.align 4 146bc3d5698SJohn Baldwin.Lenc_loop: 147bc3d5698SJohn Baldwin // middle of middle round 148bc3d5698SJohn Baldwin add x10, x11, #0x40 149bc3d5698SJohn Baldwin tbl v4.16b, {v25.16b}, v2.16b // vpshufb %xmm2, %xmm13, %xmm4 # 4 = sb1u 150bc3d5698SJohn Baldwin ld1 {v1.2d}, [x11], #16 // vmovdqa -0x40(%r11,%r10), %xmm1 # .Lk_mc_forward[] 151bc3d5698SJohn Baldwin tbl v0.16b, {v24.16b}, v3.16b // vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t 152bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v16.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k 153bc3d5698SJohn Baldwin tbl v5.16b, {v27.16b}, v2.16b // vpshufb %xmm2, %xmm15, %xmm5 # 4 = sb2u 154bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A 155bc3d5698SJohn Baldwin tbl v2.16b, {v26.16b}, v3.16b // vpshufb %xmm3, %xmm14, %xmm2 # 2 = sb2t 156bc3d5698SJohn Baldwin ld1 {v4.2d}, [x10] // vmovdqa (%r11,%r10), %xmm4 # .Lk_mc_backward[] 157bc3d5698SJohn Baldwin tbl v3.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm3 # 0 = B 158bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v5.16b // vpxor %xmm5, %xmm2, %xmm2 # 2 = 2A 159bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b}, v4.16b // vpshufb %xmm4, %xmm0, %xmm0 # 3 = D 160bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 0 = 2A+B 161bc3d5698SJohn Baldwin tbl v4.16b, {v3.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm4 # 0 = 2B+C 162bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v3.16b // vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D 163bc3d5698SJohn Baldwin and x11, x11, #~(1<<6) // and $0x30, %r11 # ... mod 4 164bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D 165bc3d5698SJohn Baldwin sub w8, w8, #1 // nr-- 166bc3d5698SJohn Baldwin 167bc3d5698SJohn Baldwin.Lenc_entry: 168bc3d5698SJohn Baldwin // top of round 169bc3d5698SJohn Baldwin and v1.16b, v0.16b, v17.16b // vpand %xmm0, %xmm9, %xmm1 # 0 = k 170bc3d5698SJohn Baldwin ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i 171bc3d5698SJohn Baldwin tbl v5.16b, {v19.16b}, v1.16b // vpshufb %xmm1, %xmm11, %xmm5 # 2 = a/k 172bc3d5698SJohn Baldwin eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j 173bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i 174bc3d5698SJohn Baldwin tbl v4.16b, {v18.16b}, v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j 175bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v5.16b // vpxor %xmm5, %xmm3, %xmm3 # 3 = iak = 1/i + a/k 176bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v5.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = jak = 1/j + a/k 177bc3d5698SJohn Baldwin tbl v2.16b, {v18.16b}, v3.16b // vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak 178bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b}, v4.16b // vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak 179bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v1.16b // vpxor %xmm1, %xmm2, %xmm2 # 2 = io 180bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo 181bc3d5698SJohn Baldwin ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm5 182bc3d5698SJohn Baldwin cbnz w8, .Lenc_loop 183bc3d5698SJohn Baldwin 184bc3d5698SJohn Baldwin // middle of last round 185bc3d5698SJohn Baldwin add x10, x11, #0x80 186bc3d5698SJohn Baldwin // vmovdqa -0x60(%r10), %xmm4 # 3 : sbou .Lk_sbo 187bc3d5698SJohn Baldwin // vmovdqa -0x50(%r10), %xmm0 # 0 : sbot .Lk_sbo+16 188bc3d5698SJohn Baldwin tbl v4.16b, {v22.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou 189bc3d5698SJohn Baldwin ld1 {v1.2d}, [x10] // vmovdqa 0x40(%r11,%r10), %xmm1 # .Lk_sr[] 190bc3d5698SJohn Baldwin tbl v0.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t 191bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v16.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k 192bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A 193bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm0 194bc3d5698SJohn Baldwin ret 195bc3d5698SJohn Baldwin.size _vpaes_encrypt_core,.-_vpaes_encrypt_core 196bc3d5698SJohn Baldwin 197bc3d5698SJohn Baldwin.globl vpaes_encrypt 198bc3d5698SJohn Baldwin.type vpaes_encrypt,%function 199bc3d5698SJohn Baldwin.align 4 200bc3d5698SJohn Baldwinvpaes_encrypt: 201bd9588bcSAndrew Turner AARCH64_SIGN_LINK_REGISTER 202bc3d5698SJohn Baldwin stp x29,x30,[sp,#-16]! 203bc3d5698SJohn Baldwin add x29,sp,#0 204bc3d5698SJohn Baldwin 205bc3d5698SJohn Baldwin ld1 {v7.16b}, [x0] 206bc3d5698SJohn Baldwin bl _vpaes_encrypt_preheat 207bc3d5698SJohn Baldwin bl _vpaes_encrypt_core 208bc3d5698SJohn Baldwin st1 {v0.16b}, [x1] 209bc3d5698SJohn Baldwin 210bc3d5698SJohn Baldwin ldp x29,x30,[sp],#16 211bd9588bcSAndrew Turner AARCH64_VALIDATE_LINK_REGISTER 212bc3d5698SJohn Baldwin ret 213bc3d5698SJohn Baldwin.size vpaes_encrypt,.-vpaes_encrypt 214bc3d5698SJohn Baldwin 215bc3d5698SJohn Baldwin.type _vpaes_encrypt_2x,%function 216bc3d5698SJohn Baldwin.align 4 217bc3d5698SJohn Baldwin_vpaes_encrypt_2x: 218bc3d5698SJohn Baldwin mov x9, x2 219bc3d5698SJohn Baldwin ldr w8, [x2,#240] // pull rounds 220bc3d5698SJohn Baldwin adr x11, .Lk_mc_forward+16 221bc3d5698SJohn Baldwin // vmovdqa .Lk_ipt(%rip), %xmm2 # iptlo 222bc3d5698SJohn Baldwin ld1 {v16.2d}, [x9], #16 // vmovdqu (%r9), %xmm5 # round0 key 223bc3d5698SJohn Baldwin and v1.16b, v14.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 224bc3d5698SJohn Baldwin ushr v0.16b, v14.16b, #4 // vpsrlb $4, %xmm0, %xmm0 225bc3d5698SJohn Baldwin and v9.16b, v15.16b, v17.16b 226bc3d5698SJohn Baldwin ushr v8.16b, v15.16b, #4 227bc3d5698SJohn Baldwin tbl v1.16b, {v20.16b}, v1.16b // vpshufb %xmm1, %xmm2, %xmm1 228bc3d5698SJohn Baldwin tbl v9.16b, {v20.16b}, v9.16b 229bc3d5698SJohn Baldwin // vmovdqa .Lk_ipt+16(%rip), %xmm3 # ipthi 230bc3d5698SJohn Baldwin tbl v2.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm3, %xmm2 231bc3d5698SJohn Baldwin tbl v10.16b, {v21.16b}, v8.16b 232bc3d5698SJohn Baldwin eor v0.16b, v1.16b, v16.16b // vpxor %xmm5, %xmm1, %xmm0 233bc3d5698SJohn Baldwin eor v8.16b, v9.16b, v16.16b 234bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 235bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v10.16b 236bc3d5698SJohn Baldwin b .Lenc_2x_entry 237bc3d5698SJohn Baldwin 238bc3d5698SJohn Baldwin.align 4 239bc3d5698SJohn Baldwin.Lenc_2x_loop: 240bc3d5698SJohn Baldwin // middle of middle round 241bc3d5698SJohn Baldwin add x10, x11, #0x40 242bc3d5698SJohn Baldwin tbl v4.16b, {v25.16b}, v2.16b // vpshufb %xmm2, %xmm13, %xmm4 # 4 = sb1u 243bc3d5698SJohn Baldwin tbl v12.16b, {v25.16b}, v10.16b 244bc3d5698SJohn Baldwin ld1 {v1.2d}, [x11], #16 // vmovdqa -0x40(%r11,%r10), %xmm1 # .Lk_mc_forward[] 245bc3d5698SJohn Baldwin tbl v0.16b, {v24.16b}, v3.16b // vpshufb %xmm3, %xmm12, %xmm0 # 0 = sb1t 246bc3d5698SJohn Baldwin tbl v8.16b, {v24.16b}, v11.16b 247bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v16.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k 248bc3d5698SJohn Baldwin eor v12.16b, v12.16b, v16.16b 249bc3d5698SJohn Baldwin tbl v5.16b, {v27.16b}, v2.16b // vpshufb %xmm2, %xmm15, %xmm5 # 4 = sb2u 250bc3d5698SJohn Baldwin tbl v13.16b, {v27.16b}, v10.16b 251bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A 252bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v12.16b 253bc3d5698SJohn Baldwin tbl v2.16b, {v26.16b}, v3.16b // vpshufb %xmm3, %xmm14, %xmm2 # 2 = sb2t 254bc3d5698SJohn Baldwin tbl v10.16b, {v26.16b}, v11.16b 255bc3d5698SJohn Baldwin ld1 {v4.2d}, [x10] // vmovdqa (%r11,%r10), %xmm4 # .Lk_mc_backward[] 256bc3d5698SJohn Baldwin tbl v3.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm3 # 0 = B 257bc3d5698SJohn Baldwin tbl v11.16b, {v8.16b}, v1.16b 258bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v5.16b // vpxor %xmm5, %xmm2, %xmm2 # 2 = 2A 259bc3d5698SJohn Baldwin eor v10.16b, v10.16b, v13.16b 260bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b}, v4.16b // vpshufb %xmm4, %xmm0, %xmm0 # 3 = D 261bc3d5698SJohn Baldwin tbl v8.16b, {v8.16b}, v4.16b 262bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 0 = 2A+B 263bc3d5698SJohn Baldwin eor v11.16b, v11.16b, v10.16b 264bc3d5698SJohn Baldwin tbl v4.16b, {v3.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm4 # 0 = 2B+C 265bc3d5698SJohn Baldwin tbl v12.16b, {v11.16b},v1.16b 266bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v3.16b // vpxor %xmm3, %xmm0, %xmm0 # 3 = 2A+B+D 267bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v11.16b 268bc3d5698SJohn Baldwin and x11, x11, #~(1<<6) // and $0x30, %r11 # ... mod 4 269bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = 2A+3B+C+D 270bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v12.16b 271bc3d5698SJohn Baldwin sub w8, w8, #1 // nr-- 272bc3d5698SJohn Baldwin 273bc3d5698SJohn Baldwin.Lenc_2x_entry: 274bc3d5698SJohn Baldwin // top of round 275bc3d5698SJohn Baldwin and v1.16b, v0.16b, v17.16b // vpand %xmm0, %xmm9, %xmm1 # 0 = k 276bc3d5698SJohn Baldwin ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i 277bc3d5698SJohn Baldwin and v9.16b, v8.16b, v17.16b 278bc3d5698SJohn Baldwin ushr v8.16b, v8.16b, #4 279bc3d5698SJohn Baldwin tbl v5.16b, {v19.16b},v1.16b // vpshufb %xmm1, %xmm11, %xmm5 # 2 = a/k 280bc3d5698SJohn Baldwin tbl v13.16b, {v19.16b},v9.16b 281bc3d5698SJohn Baldwin eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j 282bc3d5698SJohn Baldwin eor v9.16b, v9.16b, v8.16b 283bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b},v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i 284bc3d5698SJohn Baldwin tbl v11.16b, {v18.16b},v8.16b 285bc3d5698SJohn Baldwin tbl v4.16b, {v18.16b},v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j 286bc3d5698SJohn Baldwin tbl v12.16b, {v18.16b},v9.16b 287bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v5.16b // vpxor %xmm5, %xmm3, %xmm3 # 3 = iak = 1/i + a/k 288bc3d5698SJohn Baldwin eor v11.16b, v11.16b, v13.16b 289bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v5.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = jak = 1/j + a/k 290bc3d5698SJohn Baldwin eor v12.16b, v12.16b, v13.16b 291bc3d5698SJohn Baldwin tbl v2.16b, {v18.16b},v3.16b // vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak 292bc3d5698SJohn Baldwin tbl v10.16b, {v18.16b},v11.16b 293bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b},v4.16b // vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak 294bc3d5698SJohn Baldwin tbl v11.16b, {v18.16b},v12.16b 295bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v1.16b // vpxor %xmm1, %xmm2, %xmm2 # 2 = io 296bc3d5698SJohn Baldwin eor v10.16b, v10.16b, v9.16b 297bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo 298bc3d5698SJohn Baldwin eor v11.16b, v11.16b, v8.16b 299bc3d5698SJohn Baldwin ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm5 300bc3d5698SJohn Baldwin cbnz w8, .Lenc_2x_loop 301bc3d5698SJohn Baldwin 302bc3d5698SJohn Baldwin // middle of last round 303bc3d5698SJohn Baldwin add x10, x11, #0x80 304bc3d5698SJohn Baldwin // vmovdqa -0x60(%r10), %xmm4 # 3 : sbou .Lk_sbo 305bc3d5698SJohn Baldwin // vmovdqa -0x50(%r10), %xmm0 # 0 : sbot .Lk_sbo+16 306bc3d5698SJohn Baldwin tbl v4.16b, {v22.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou 307bc3d5698SJohn Baldwin tbl v12.16b, {v22.16b}, v10.16b 308bc3d5698SJohn Baldwin ld1 {v1.2d}, [x10] // vmovdqa 0x40(%r11,%r10), %xmm1 # .Lk_sr[] 309bc3d5698SJohn Baldwin tbl v0.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm0, %xmm0 # 0 = sb1t 310bc3d5698SJohn Baldwin tbl v8.16b, {v23.16b}, v11.16b 311bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v16.16b // vpxor %xmm5, %xmm4, %xmm4 # 4 = sb1u + k 312bc3d5698SJohn Baldwin eor v12.16b, v12.16b, v16.16b 313bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 0 = A 314bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v12.16b 315bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b},v1.16b // vpshufb %xmm1, %xmm0, %xmm0 316bc3d5698SJohn Baldwin tbl v1.16b, {v8.16b},v1.16b 317bc3d5698SJohn Baldwin ret 318bc3d5698SJohn Baldwin.size _vpaes_encrypt_2x,.-_vpaes_encrypt_2x 319bc3d5698SJohn Baldwin 320bc3d5698SJohn Baldwin.type _vpaes_decrypt_preheat,%function 321bc3d5698SJohn Baldwin.align 4 322bc3d5698SJohn Baldwin_vpaes_decrypt_preheat: 323bc3d5698SJohn Baldwin adr x10, .Lk_inv 324bc3d5698SJohn Baldwin movi v17.16b, #0x0f 325bc3d5698SJohn Baldwin adr x11, .Lk_dipt 326bc3d5698SJohn Baldwin ld1 {v18.2d,v19.2d}, [x10],#32 // .Lk_inv 327bc3d5698SJohn Baldwin ld1 {v20.2d,v21.2d,v22.2d,v23.2d}, [x11],#64 // .Lk_dipt, .Lk_dsbo 328bc3d5698SJohn Baldwin ld1 {v24.2d,v25.2d,v26.2d,v27.2d}, [x11],#64 // .Lk_dsb9, .Lk_dsbd 329bc3d5698SJohn Baldwin ld1 {v28.2d,v29.2d,v30.2d,v31.2d}, [x11] // .Lk_dsbb, .Lk_dsbe 330bc3d5698SJohn Baldwin ret 331bc3d5698SJohn Baldwin.size _vpaes_decrypt_preheat,.-_vpaes_decrypt_preheat 332bc3d5698SJohn Baldwin 333c0855eaaSJohn Baldwin// 334c0855eaaSJohn Baldwin// Decryption core 335c0855eaaSJohn Baldwin// 336c0855eaaSJohn Baldwin// Same API as encryption core. 337c0855eaaSJohn Baldwin// 338bc3d5698SJohn Baldwin.type _vpaes_decrypt_core,%function 339bc3d5698SJohn Baldwin.align 4 340bc3d5698SJohn Baldwin_vpaes_decrypt_core: 341bc3d5698SJohn Baldwin mov x9, x2 342bc3d5698SJohn Baldwin ldr w8, [x2,#240] // pull rounds 343bc3d5698SJohn Baldwin 344bc3d5698SJohn Baldwin // vmovdqa .Lk_dipt(%rip), %xmm2 # iptlo 345bc3d5698SJohn Baldwin lsl x11, x8, #4 // mov %rax, %r11; shl $4, %r11 346bc3d5698SJohn Baldwin eor x11, x11, #0x30 // xor $0x30, %r11 347bc3d5698SJohn Baldwin adr x10, .Lk_sr 348bc3d5698SJohn Baldwin and x11, x11, #0x30 // and $0x30, %r11 349bc3d5698SJohn Baldwin add x11, x11, x10 350bc3d5698SJohn Baldwin adr x10, .Lk_mc_forward+48 351bc3d5698SJohn Baldwin 352bc3d5698SJohn Baldwin ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm4 # round0 key 353bc3d5698SJohn Baldwin and v1.16b, v7.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 354bc3d5698SJohn Baldwin ushr v0.16b, v7.16b, #4 // vpsrlb $4, %xmm0, %xmm0 355bc3d5698SJohn Baldwin tbl v2.16b, {v20.16b}, v1.16b // vpshufb %xmm1, %xmm2, %xmm2 356bc3d5698SJohn Baldwin ld1 {v5.2d}, [x10] // vmovdqa .Lk_mc_forward+48(%rip), %xmm5 357bc3d5698SJohn Baldwin // vmovdqa .Lk_dipt+16(%rip), %xmm1 # ipthi 358bc3d5698SJohn Baldwin tbl v0.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm1, %xmm0 359bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v16.16b // vpxor %xmm4, %xmm2, %xmm2 360bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 361bc3d5698SJohn Baldwin b .Ldec_entry 362bc3d5698SJohn Baldwin 363bc3d5698SJohn Baldwin.align 4 364bc3d5698SJohn Baldwin.Ldec_loop: 365bc3d5698SJohn Baldwin// 366bc3d5698SJohn Baldwin// Inverse mix columns 367bc3d5698SJohn Baldwin// 368bc3d5698SJohn Baldwin // vmovdqa -0x20(%r10),%xmm4 # 4 : sb9u 369bc3d5698SJohn Baldwin // vmovdqa -0x10(%r10),%xmm1 # 0 : sb9t 370bc3d5698SJohn Baldwin tbl v4.16b, {v24.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sb9u 371bc3d5698SJohn Baldwin tbl v1.16b, {v25.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb9t 372bc3d5698SJohn Baldwin eor v0.16b, v4.16b, v16.16b // vpxor %xmm4, %xmm0, %xmm0 373bc3d5698SJohn Baldwin // vmovdqa 0x00(%r10),%xmm4 # 4 : sbdu 374bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch 375bc3d5698SJohn Baldwin // vmovdqa 0x10(%r10),%xmm1 # 0 : sbdt 376bc3d5698SJohn Baldwin 377bc3d5698SJohn Baldwin tbl v4.16b, {v26.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbdu 378bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch 379bc3d5698SJohn Baldwin tbl v1.16b, {v27.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbdt 380bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch 381bc3d5698SJohn Baldwin // vmovdqa 0x20(%r10), %xmm4 # 4 : sbbu 382bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch 383bc3d5698SJohn Baldwin // vmovdqa 0x30(%r10), %xmm1 # 0 : sbbt 384bc3d5698SJohn Baldwin 385bc3d5698SJohn Baldwin tbl v4.16b, {v28.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbbu 386bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch 387bc3d5698SJohn Baldwin tbl v1.16b, {v29.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbbt 388bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch 389bc3d5698SJohn Baldwin // vmovdqa 0x40(%r10), %xmm4 # 4 : sbeu 390bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch 391bc3d5698SJohn Baldwin // vmovdqa 0x50(%r10), %xmm1 # 0 : sbet 392bc3d5698SJohn Baldwin 393bc3d5698SJohn Baldwin tbl v4.16b, {v30.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbeu 394bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b}, v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch 395bc3d5698SJohn Baldwin tbl v1.16b, {v31.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbet 396bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch 397bc3d5698SJohn Baldwin ext v5.16b, v5.16b, v5.16b, #12 // vpalignr $12, %xmm5, %xmm5, %xmm5 398bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch 399bc3d5698SJohn Baldwin sub w8, w8, #1 // sub $1,%rax # nr-- 400bc3d5698SJohn Baldwin 401bc3d5698SJohn Baldwin.Ldec_entry: 402bc3d5698SJohn Baldwin // top of round 403bc3d5698SJohn Baldwin and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k 404bc3d5698SJohn Baldwin ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i 405bc3d5698SJohn Baldwin tbl v2.16b, {v19.16b}, v1.16b // vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k 406bc3d5698SJohn Baldwin eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j 407bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i 408bc3d5698SJohn Baldwin tbl v4.16b, {v18.16b}, v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j 409bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k 410bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v2.16b // vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k 411bc3d5698SJohn Baldwin tbl v2.16b, {v18.16b}, v3.16b // vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak 412bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b}, v4.16b // vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak 413bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v1.16b // vpxor %xmm1, %xmm2, %xmm2 # 2 = io 414bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo 415bc3d5698SJohn Baldwin ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm0 416bc3d5698SJohn Baldwin cbnz w8, .Ldec_loop 417bc3d5698SJohn Baldwin 418bc3d5698SJohn Baldwin // middle of last round 419bc3d5698SJohn Baldwin // vmovdqa 0x60(%r10), %xmm4 # 3 : sbou 420bc3d5698SJohn Baldwin tbl v4.16b, {v22.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou 421bc3d5698SJohn Baldwin // vmovdqa 0x70(%r10), %xmm1 # 0 : sbot 422bc3d5698SJohn Baldwin ld1 {v2.2d}, [x11] // vmovdqa -0x160(%r11), %xmm2 # .Lk_sr-.Lk_dsbd=-0x160 423bc3d5698SJohn Baldwin tbl v1.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb1t 424bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v16.16b // vpxor %xmm0, %xmm4, %xmm4 # 4 = sb1u + k 425bc3d5698SJohn Baldwin eor v0.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm0 # 0 = A 426bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b}, v2.16b // vpshufb %xmm2, %xmm0, %xmm0 427bc3d5698SJohn Baldwin ret 428bc3d5698SJohn Baldwin.size _vpaes_decrypt_core,.-_vpaes_decrypt_core 429bc3d5698SJohn Baldwin 430bc3d5698SJohn Baldwin.globl vpaes_decrypt 431bc3d5698SJohn Baldwin.type vpaes_decrypt,%function 432bc3d5698SJohn Baldwin.align 4 433bc3d5698SJohn Baldwinvpaes_decrypt: 434bd9588bcSAndrew Turner AARCH64_SIGN_LINK_REGISTER 435bc3d5698SJohn Baldwin stp x29,x30,[sp,#-16]! 436bc3d5698SJohn Baldwin add x29,sp,#0 437bc3d5698SJohn Baldwin 438bc3d5698SJohn Baldwin ld1 {v7.16b}, [x0] 439bc3d5698SJohn Baldwin bl _vpaes_decrypt_preheat 440bc3d5698SJohn Baldwin bl _vpaes_decrypt_core 441bc3d5698SJohn Baldwin st1 {v0.16b}, [x1] 442bc3d5698SJohn Baldwin 443bc3d5698SJohn Baldwin ldp x29,x30,[sp],#16 444bd9588bcSAndrew Turner AARCH64_VALIDATE_LINK_REGISTER 445bc3d5698SJohn Baldwin ret 446bc3d5698SJohn Baldwin.size vpaes_decrypt,.-vpaes_decrypt 447bc3d5698SJohn Baldwin 448bc3d5698SJohn Baldwin// v14-v15 input, v0-v1 output 449bc3d5698SJohn Baldwin.type _vpaes_decrypt_2x,%function 450bc3d5698SJohn Baldwin.align 4 451bc3d5698SJohn Baldwin_vpaes_decrypt_2x: 452bc3d5698SJohn Baldwin mov x9, x2 453bc3d5698SJohn Baldwin ldr w8, [x2,#240] // pull rounds 454bc3d5698SJohn Baldwin 455bc3d5698SJohn Baldwin // vmovdqa .Lk_dipt(%rip), %xmm2 # iptlo 456bc3d5698SJohn Baldwin lsl x11, x8, #4 // mov %rax, %r11; shl $4, %r11 457bc3d5698SJohn Baldwin eor x11, x11, #0x30 // xor $0x30, %r11 458bc3d5698SJohn Baldwin adr x10, .Lk_sr 459bc3d5698SJohn Baldwin and x11, x11, #0x30 // and $0x30, %r11 460bc3d5698SJohn Baldwin add x11, x11, x10 461bc3d5698SJohn Baldwin adr x10, .Lk_mc_forward+48 462bc3d5698SJohn Baldwin 463bc3d5698SJohn Baldwin ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm4 # round0 key 464bc3d5698SJohn Baldwin and v1.16b, v14.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 465bc3d5698SJohn Baldwin ushr v0.16b, v14.16b, #4 // vpsrlb $4, %xmm0, %xmm0 466bc3d5698SJohn Baldwin and v9.16b, v15.16b, v17.16b 467bc3d5698SJohn Baldwin ushr v8.16b, v15.16b, #4 468bc3d5698SJohn Baldwin tbl v2.16b, {v20.16b},v1.16b // vpshufb %xmm1, %xmm2, %xmm2 469bc3d5698SJohn Baldwin tbl v10.16b, {v20.16b},v9.16b 470bc3d5698SJohn Baldwin ld1 {v5.2d}, [x10] // vmovdqa .Lk_mc_forward+48(%rip), %xmm5 471bc3d5698SJohn Baldwin // vmovdqa .Lk_dipt+16(%rip), %xmm1 # ipthi 472bc3d5698SJohn Baldwin tbl v0.16b, {v21.16b},v0.16b // vpshufb %xmm0, %xmm1, %xmm0 473bc3d5698SJohn Baldwin tbl v8.16b, {v21.16b},v8.16b 474bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v16.16b // vpxor %xmm4, %xmm2, %xmm2 475bc3d5698SJohn Baldwin eor v10.16b, v10.16b, v16.16b 476bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 477bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v10.16b 478bc3d5698SJohn Baldwin b .Ldec_2x_entry 479bc3d5698SJohn Baldwin 480bc3d5698SJohn Baldwin.align 4 481bc3d5698SJohn Baldwin.Ldec_2x_loop: 482bc3d5698SJohn Baldwin// 483bc3d5698SJohn Baldwin// Inverse mix columns 484bc3d5698SJohn Baldwin// 485bc3d5698SJohn Baldwin // vmovdqa -0x20(%r10),%xmm4 # 4 : sb9u 486bc3d5698SJohn Baldwin // vmovdqa -0x10(%r10),%xmm1 # 0 : sb9t 487bc3d5698SJohn Baldwin tbl v4.16b, {v24.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sb9u 488bc3d5698SJohn Baldwin tbl v12.16b, {v24.16b}, v10.16b 489bc3d5698SJohn Baldwin tbl v1.16b, {v25.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb9t 490bc3d5698SJohn Baldwin tbl v9.16b, {v25.16b}, v11.16b 491bc3d5698SJohn Baldwin eor v0.16b, v4.16b, v16.16b // vpxor %xmm4, %xmm0, %xmm0 492bc3d5698SJohn Baldwin eor v8.16b, v12.16b, v16.16b 493bc3d5698SJohn Baldwin // vmovdqa 0x00(%r10),%xmm4 # 4 : sbdu 494bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch 495bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v9.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch 496bc3d5698SJohn Baldwin // vmovdqa 0x10(%r10),%xmm1 # 0 : sbdt 497bc3d5698SJohn Baldwin 498bc3d5698SJohn Baldwin tbl v4.16b, {v26.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbdu 499bc3d5698SJohn Baldwin tbl v12.16b, {v26.16b}, v10.16b 500bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch 501bc3d5698SJohn Baldwin tbl v8.16b, {v8.16b},v5.16b 502bc3d5698SJohn Baldwin tbl v1.16b, {v27.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbdt 503bc3d5698SJohn Baldwin tbl v9.16b, {v27.16b}, v11.16b 504bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch 505bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v12.16b 506bc3d5698SJohn Baldwin // vmovdqa 0x20(%r10), %xmm4 # 4 : sbbu 507bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch 508bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v9.16b 509bc3d5698SJohn Baldwin // vmovdqa 0x30(%r10), %xmm1 # 0 : sbbt 510bc3d5698SJohn Baldwin 511bc3d5698SJohn Baldwin tbl v4.16b, {v28.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbbu 512bc3d5698SJohn Baldwin tbl v12.16b, {v28.16b}, v10.16b 513bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch 514bc3d5698SJohn Baldwin tbl v8.16b, {v8.16b},v5.16b 515bc3d5698SJohn Baldwin tbl v1.16b, {v29.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbbt 516bc3d5698SJohn Baldwin tbl v9.16b, {v29.16b}, v11.16b 517bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch 518bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v12.16b 519bc3d5698SJohn Baldwin // vmovdqa 0x40(%r10), %xmm4 # 4 : sbeu 520bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch 521bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v9.16b 522bc3d5698SJohn Baldwin // vmovdqa 0x50(%r10), %xmm1 # 0 : sbet 523bc3d5698SJohn Baldwin 524bc3d5698SJohn Baldwin tbl v4.16b, {v30.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbeu 525bc3d5698SJohn Baldwin tbl v12.16b, {v30.16b}, v10.16b 526bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b},v5.16b // vpshufb %xmm5, %xmm0, %xmm0 # MC ch 527bc3d5698SJohn Baldwin tbl v8.16b, {v8.16b},v5.16b 528bc3d5698SJohn Baldwin tbl v1.16b, {v31.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sbet 529bc3d5698SJohn Baldwin tbl v9.16b, {v31.16b}, v11.16b 530bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v4.16b // vpxor %xmm4, %xmm0, %xmm0 # 4 = ch 531bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v12.16b 532bc3d5698SJohn Baldwin ext v5.16b, v5.16b, v5.16b, #12 // vpalignr $12, %xmm5, %xmm5, %xmm5 533bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v1.16b // vpxor %xmm1, %xmm0, %xmm0 # 0 = ch 534bc3d5698SJohn Baldwin eor v8.16b, v8.16b, v9.16b 535bc3d5698SJohn Baldwin sub w8, w8, #1 // sub $1,%rax # nr-- 536bc3d5698SJohn Baldwin 537bc3d5698SJohn Baldwin.Ldec_2x_entry: 538bc3d5698SJohn Baldwin // top of round 539bc3d5698SJohn Baldwin and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k 540bc3d5698SJohn Baldwin ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i 541bc3d5698SJohn Baldwin and v9.16b, v8.16b, v17.16b 542bc3d5698SJohn Baldwin ushr v8.16b, v8.16b, #4 543bc3d5698SJohn Baldwin tbl v2.16b, {v19.16b},v1.16b // vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k 544bc3d5698SJohn Baldwin tbl v10.16b, {v19.16b},v9.16b 545bc3d5698SJohn Baldwin eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j 546bc3d5698SJohn Baldwin eor v9.16b, v9.16b, v8.16b 547bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b},v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i 548bc3d5698SJohn Baldwin tbl v11.16b, {v18.16b},v8.16b 549bc3d5698SJohn Baldwin tbl v4.16b, {v18.16b},v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j 550bc3d5698SJohn Baldwin tbl v12.16b, {v18.16b},v9.16b 551bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k 552bc3d5698SJohn Baldwin eor v11.16b, v11.16b, v10.16b 553bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v2.16b // vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k 554bc3d5698SJohn Baldwin eor v12.16b, v12.16b, v10.16b 555bc3d5698SJohn Baldwin tbl v2.16b, {v18.16b},v3.16b // vpshufb %xmm3, %xmm10, %xmm2 # 2 = 1/iak 556bc3d5698SJohn Baldwin tbl v10.16b, {v18.16b},v11.16b 557bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b},v4.16b // vpshufb %xmm4, %xmm10, %xmm3 # 3 = 1/jak 558bc3d5698SJohn Baldwin tbl v11.16b, {v18.16b},v12.16b 559bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v1.16b // vpxor %xmm1, %xmm2, %xmm2 # 2 = io 560bc3d5698SJohn Baldwin eor v10.16b, v10.16b, v9.16b 561bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v0.16b // vpxor %xmm0, %xmm3, %xmm3 # 3 = jo 562bc3d5698SJohn Baldwin eor v11.16b, v11.16b, v8.16b 563bc3d5698SJohn Baldwin ld1 {v16.2d}, [x9],#16 // vmovdqu (%r9), %xmm0 564bc3d5698SJohn Baldwin cbnz w8, .Ldec_2x_loop 565bc3d5698SJohn Baldwin 566bc3d5698SJohn Baldwin // middle of last round 567bc3d5698SJohn Baldwin // vmovdqa 0x60(%r10), %xmm4 # 3 : sbou 568bc3d5698SJohn Baldwin tbl v4.16b, {v22.16b}, v2.16b // vpshufb %xmm2, %xmm4, %xmm4 # 4 = sbou 569bc3d5698SJohn Baldwin tbl v12.16b, {v22.16b}, v10.16b 570bc3d5698SJohn Baldwin // vmovdqa 0x70(%r10), %xmm1 # 0 : sbot 571bc3d5698SJohn Baldwin tbl v1.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm1, %xmm1 # 0 = sb1t 572bc3d5698SJohn Baldwin tbl v9.16b, {v23.16b}, v11.16b 573bc3d5698SJohn Baldwin ld1 {v2.2d}, [x11] // vmovdqa -0x160(%r11), %xmm2 # .Lk_sr-.Lk_dsbd=-0x160 574bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v16.16b // vpxor %xmm0, %xmm4, %xmm4 # 4 = sb1u + k 575bc3d5698SJohn Baldwin eor v12.16b, v12.16b, v16.16b 576bc3d5698SJohn Baldwin eor v0.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm0 # 0 = A 577bc3d5698SJohn Baldwin eor v8.16b, v9.16b, v12.16b 578bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b},v2.16b // vpshufb %xmm2, %xmm0, %xmm0 579bc3d5698SJohn Baldwin tbl v1.16b, {v8.16b},v2.16b 580bc3d5698SJohn Baldwin ret 581bc3d5698SJohn Baldwin.size _vpaes_decrypt_2x,.-_vpaes_decrypt_2x 582c0855eaaSJohn Baldwin//////////////////////////////////////////////////////// 583c0855eaaSJohn Baldwin// // 584c0855eaaSJohn Baldwin// AES key schedule // 585c0855eaaSJohn Baldwin// // 586c0855eaaSJohn Baldwin//////////////////////////////////////////////////////// 587bc3d5698SJohn Baldwin.type _vpaes_key_preheat,%function 588bc3d5698SJohn Baldwin.align 4 589bc3d5698SJohn Baldwin_vpaes_key_preheat: 590bc3d5698SJohn Baldwin adr x10, .Lk_inv 591bc3d5698SJohn Baldwin movi v16.16b, #0x5b // .Lk_s63 592bc3d5698SJohn Baldwin adr x11, .Lk_sb1 593bc3d5698SJohn Baldwin movi v17.16b, #0x0f // .Lk_s0F 594bc3d5698SJohn Baldwin ld1 {v18.2d,v19.2d,v20.2d,v21.2d}, [x10] // .Lk_inv, .Lk_ipt 595bc3d5698SJohn Baldwin adr x10, .Lk_dksd 596bc3d5698SJohn Baldwin ld1 {v22.2d,v23.2d}, [x11] // .Lk_sb1 597bc3d5698SJohn Baldwin adr x11, .Lk_mc_forward 598bc3d5698SJohn Baldwin ld1 {v24.2d,v25.2d,v26.2d,v27.2d}, [x10],#64 // .Lk_dksd, .Lk_dksb 599bc3d5698SJohn Baldwin ld1 {v28.2d,v29.2d,v30.2d,v31.2d}, [x10],#64 // .Lk_dkse, .Lk_dks9 600bc3d5698SJohn Baldwin ld1 {v8.2d}, [x10] // .Lk_rcon 601bc3d5698SJohn Baldwin ld1 {v9.2d}, [x11] // .Lk_mc_forward[0] 602bc3d5698SJohn Baldwin ret 603bc3d5698SJohn Baldwin.size _vpaes_key_preheat,.-_vpaes_key_preheat 604bc3d5698SJohn Baldwin 605bc3d5698SJohn Baldwin.type _vpaes_schedule_core,%function 606bc3d5698SJohn Baldwin.align 4 607bc3d5698SJohn Baldwin_vpaes_schedule_core: 608bd9588bcSAndrew Turner AARCH64_SIGN_LINK_REGISTER 609bc3d5698SJohn Baldwin stp x29, x30, [sp,#-16]! 610bc3d5698SJohn Baldwin add x29,sp,#0 611bc3d5698SJohn Baldwin 612bc3d5698SJohn Baldwin bl _vpaes_key_preheat // load the tables 613bc3d5698SJohn Baldwin 614bc3d5698SJohn Baldwin ld1 {v0.16b}, [x0],#16 // vmovdqu (%rdi), %xmm0 # load key (unaligned) 615bc3d5698SJohn Baldwin 616bc3d5698SJohn Baldwin // input transform 617bc3d5698SJohn Baldwin mov v3.16b, v0.16b // vmovdqa %xmm0, %xmm3 618bc3d5698SJohn Baldwin bl _vpaes_schedule_transform 619bc3d5698SJohn Baldwin mov v7.16b, v0.16b // vmovdqa %xmm0, %xmm7 620bc3d5698SJohn Baldwin 621bc3d5698SJohn Baldwin adr x10, .Lk_sr // lea .Lk_sr(%rip),%r10 622bc3d5698SJohn Baldwin add x8, x8, x10 623bc3d5698SJohn Baldwin cbnz w3, .Lschedule_am_decrypting 624bc3d5698SJohn Baldwin 625bc3d5698SJohn Baldwin // encrypting, output zeroth round key after transform 626bc3d5698SJohn Baldwin st1 {v0.2d}, [x2] // vmovdqu %xmm0, (%rdx) 627bc3d5698SJohn Baldwin b .Lschedule_go 628bc3d5698SJohn Baldwin 629bc3d5698SJohn Baldwin.Lschedule_am_decrypting: 630bc3d5698SJohn Baldwin // decrypting, output zeroth round key after shiftrows 631bc3d5698SJohn Baldwin ld1 {v1.2d}, [x8] // vmovdqa (%r8,%r10), %xmm1 632bc3d5698SJohn Baldwin tbl v3.16b, {v3.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 633bc3d5698SJohn Baldwin st1 {v3.2d}, [x2] // vmovdqu %xmm3, (%rdx) 634bc3d5698SJohn Baldwin eor x8, x8, #0x30 // xor $0x30, %r8 635bc3d5698SJohn Baldwin 636bc3d5698SJohn Baldwin.Lschedule_go: 637bc3d5698SJohn Baldwin cmp w1, #192 // cmp $192, %esi 638bc3d5698SJohn Baldwin b.hi .Lschedule_256 639bc3d5698SJohn Baldwin b.eq .Lschedule_192 640bc3d5698SJohn Baldwin // 128: fall though 641bc3d5698SJohn Baldwin 642c0855eaaSJohn Baldwin// 643c0855eaaSJohn Baldwin// .schedule_128 644c0855eaaSJohn Baldwin// 645c0855eaaSJohn Baldwin// 128-bit specific part of key schedule. 646c0855eaaSJohn Baldwin// 647c0855eaaSJohn Baldwin// This schedule is really simple, because all its parts 648c0855eaaSJohn Baldwin// are accomplished by the subroutines. 649c0855eaaSJohn Baldwin// 650bc3d5698SJohn Baldwin.Lschedule_128: 651bc3d5698SJohn Baldwin mov x0, #10 // mov $10, %esi 652bc3d5698SJohn Baldwin 653bc3d5698SJohn Baldwin.Loop_schedule_128: 654bc3d5698SJohn Baldwin sub x0, x0, #1 // dec %esi 655bc3d5698SJohn Baldwin bl _vpaes_schedule_round 656bc3d5698SJohn Baldwin cbz x0, .Lschedule_mangle_last 657bc3d5698SJohn Baldwin bl _vpaes_schedule_mangle // write output 658bc3d5698SJohn Baldwin b .Loop_schedule_128 659bc3d5698SJohn Baldwin 660c0855eaaSJohn Baldwin// 661c0855eaaSJohn Baldwin// .aes_schedule_192 662c0855eaaSJohn Baldwin// 663c0855eaaSJohn Baldwin// 192-bit specific part of key schedule. 664c0855eaaSJohn Baldwin// 665c0855eaaSJohn Baldwin// The main body of this schedule is the same as the 128-bit 666c0855eaaSJohn Baldwin// schedule, but with more smearing. The long, high side is 667c0855eaaSJohn Baldwin// stored in %xmm7 as before, and the short, low side is in 668c0855eaaSJohn Baldwin// the high bits of %xmm6. 669c0855eaaSJohn Baldwin// 670c0855eaaSJohn Baldwin// This schedule is somewhat nastier, however, because each 671c0855eaaSJohn Baldwin// round produces 192 bits of key material, or 1.5 round keys. 672c0855eaaSJohn Baldwin// Therefore, on each cycle we do 2 rounds and produce 3 round 673c0855eaaSJohn Baldwin// keys. 674c0855eaaSJohn Baldwin// 675bc3d5698SJohn Baldwin.align 4 676bc3d5698SJohn Baldwin.Lschedule_192: 677bc3d5698SJohn Baldwin sub x0, x0, #8 678bc3d5698SJohn Baldwin ld1 {v0.16b}, [x0] // vmovdqu 8(%rdi),%xmm0 # load key part 2 (very unaligned) 679bc3d5698SJohn Baldwin bl _vpaes_schedule_transform // input transform 680bc3d5698SJohn Baldwin mov v6.16b, v0.16b // vmovdqa %xmm0, %xmm6 # save short part 681bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v4.16b // vpxor %xmm4, %xmm4, %xmm4 # clear 4 682bc3d5698SJohn Baldwin ins v6.d[0], v4.d[0] // vmovhlps %xmm4, %xmm6, %xmm6 # clobber low side with zeros 683bc3d5698SJohn Baldwin mov x0, #4 // mov $4, %esi 684bc3d5698SJohn Baldwin 685bc3d5698SJohn Baldwin.Loop_schedule_192: 686bc3d5698SJohn Baldwin sub x0, x0, #1 // dec %esi 687bc3d5698SJohn Baldwin bl _vpaes_schedule_round 688bc3d5698SJohn Baldwin ext v0.16b, v6.16b, v0.16b, #8 // vpalignr $8,%xmm6,%xmm0,%xmm0 689bc3d5698SJohn Baldwin bl _vpaes_schedule_mangle // save key n 690bc3d5698SJohn Baldwin bl _vpaes_schedule_192_smear 691bc3d5698SJohn Baldwin bl _vpaes_schedule_mangle // save key n+1 692bc3d5698SJohn Baldwin bl _vpaes_schedule_round 693bc3d5698SJohn Baldwin cbz x0, .Lschedule_mangle_last 694bc3d5698SJohn Baldwin bl _vpaes_schedule_mangle // save key n+2 695bc3d5698SJohn Baldwin bl _vpaes_schedule_192_smear 696bc3d5698SJohn Baldwin b .Loop_schedule_192 697bc3d5698SJohn Baldwin 698c0855eaaSJohn Baldwin// 699c0855eaaSJohn Baldwin// .aes_schedule_256 700c0855eaaSJohn Baldwin// 701c0855eaaSJohn Baldwin// 256-bit specific part of key schedule. 702c0855eaaSJohn Baldwin// 703c0855eaaSJohn Baldwin// The structure here is very similar to the 128-bit 704c0855eaaSJohn Baldwin// schedule, but with an additional "low side" in 705c0855eaaSJohn Baldwin// %xmm6. The low side's rounds are the same as the 706c0855eaaSJohn Baldwin// high side's, except no rcon and no rotation. 707c0855eaaSJohn Baldwin// 708bc3d5698SJohn Baldwin.align 4 709bc3d5698SJohn Baldwin.Lschedule_256: 710bc3d5698SJohn Baldwin ld1 {v0.16b}, [x0] // vmovdqu 16(%rdi),%xmm0 # load key part 2 (unaligned) 711bc3d5698SJohn Baldwin bl _vpaes_schedule_transform // input transform 712bc3d5698SJohn Baldwin mov x0, #7 // mov $7, %esi 713bc3d5698SJohn Baldwin 714bc3d5698SJohn Baldwin.Loop_schedule_256: 715bc3d5698SJohn Baldwin sub x0, x0, #1 // dec %esi 716bc3d5698SJohn Baldwin bl _vpaes_schedule_mangle // output low result 717bc3d5698SJohn Baldwin mov v6.16b, v0.16b // vmovdqa %xmm0, %xmm6 # save cur_lo in xmm6 718bc3d5698SJohn Baldwin 719bc3d5698SJohn Baldwin // high round 720bc3d5698SJohn Baldwin bl _vpaes_schedule_round 721bc3d5698SJohn Baldwin cbz x0, .Lschedule_mangle_last 722bc3d5698SJohn Baldwin bl _vpaes_schedule_mangle 723bc3d5698SJohn Baldwin 724bc3d5698SJohn Baldwin // low round. swap xmm7 and xmm6 725bc3d5698SJohn Baldwin dup v0.4s, v0.s[3] // vpshufd $0xFF, %xmm0, %xmm0 726bc3d5698SJohn Baldwin movi v4.16b, #0 727bc3d5698SJohn Baldwin mov v5.16b, v7.16b // vmovdqa %xmm7, %xmm5 728bc3d5698SJohn Baldwin mov v7.16b, v6.16b // vmovdqa %xmm6, %xmm7 729bc3d5698SJohn Baldwin bl _vpaes_schedule_low_round 730bc3d5698SJohn Baldwin mov v7.16b, v5.16b // vmovdqa %xmm5, %xmm7 731bc3d5698SJohn Baldwin 732bc3d5698SJohn Baldwin b .Loop_schedule_256 733bc3d5698SJohn Baldwin 734c0855eaaSJohn Baldwin// 735c0855eaaSJohn Baldwin// .aes_schedule_mangle_last 736c0855eaaSJohn Baldwin// 737c0855eaaSJohn Baldwin// Mangler for last round of key schedule 738c0855eaaSJohn Baldwin// Mangles %xmm0 739c0855eaaSJohn Baldwin// when encrypting, outputs out(%xmm0) ^ 63 740c0855eaaSJohn Baldwin// when decrypting, outputs unskew(%xmm0) 741c0855eaaSJohn Baldwin// 742c0855eaaSJohn Baldwin// Always called right before return... jumps to cleanup and exits 743c0855eaaSJohn Baldwin// 744bc3d5698SJohn Baldwin.align 4 745bc3d5698SJohn Baldwin.Lschedule_mangle_last: 746bc3d5698SJohn Baldwin // schedule last round key from xmm0 747bc3d5698SJohn Baldwin adr x11, .Lk_deskew // lea .Lk_deskew(%rip),%r11 # prepare to deskew 748bc3d5698SJohn Baldwin cbnz w3, .Lschedule_mangle_last_dec 749bc3d5698SJohn Baldwin 750bc3d5698SJohn Baldwin // encrypting 751bc3d5698SJohn Baldwin ld1 {v1.2d}, [x8] // vmovdqa (%r8,%r10),%xmm1 752bc3d5698SJohn Baldwin adr x11, .Lk_opt // lea .Lk_opt(%rip), %r11 # prepare to output transform 753bc3d5698SJohn Baldwin add x2, x2, #32 // add $32, %rdx 754bc3d5698SJohn Baldwin tbl v0.16b, {v0.16b}, v1.16b // vpshufb %xmm1, %xmm0, %xmm0 # output permute 755bc3d5698SJohn Baldwin 756bc3d5698SJohn Baldwin.Lschedule_mangle_last_dec: 757bc3d5698SJohn Baldwin ld1 {v20.2d,v21.2d}, [x11] // reload constants 758bc3d5698SJohn Baldwin sub x2, x2, #16 // add $-16, %rdx 759bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v16.16b // vpxor .Lk_s63(%rip), %xmm0, %xmm0 760bc3d5698SJohn Baldwin bl _vpaes_schedule_transform // output transform 761bc3d5698SJohn Baldwin st1 {v0.2d}, [x2] // vmovdqu %xmm0, (%rdx) # save last key 762bc3d5698SJohn Baldwin 763bc3d5698SJohn Baldwin // cleanup 764bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v0.16b // vpxor %xmm0, %xmm0, %xmm0 765bc3d5698SJohn Baldwin eor v1.16b, v1.16b, v1.16b // vpxor %xmm1, %xmm1, %xmm1 766bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v2.16b // vpxor %xmm2, %xmm2, %xmm2 767bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v3.16b // vpxor %xmm3, %xmm3, %xmm3 768bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v4.16b // vpxor %xmm4, %xmm4, %xmm4 769bc3d5698SJohn Baldwin eor v5.16b, v5.16b, v5.16b // vpxor %xmm5, %xmm5, %xmm5 770bc3d5698SJohn Baldwin eor v6.16b, v6.16b, v6.16b // vpxor %xmm6, %xmm6, %xmm6 771bc3d5698SJohn Baldwin eor v7.16b, v7.16b, v7.16b // vpxor %xmm7, %xmm7, %xmm7 772bc3d5698SJohn Baldwin ldp x29, x30, [sp],#16 773bd9588bcSAndrew Turner AARCH64_VALIDATE_LINK_REGISTER 774bc3d5698SJohn Baldwin ret 775bc3d5698SJohn Baldwin.size _vpaes_schedule_core,.-_vpaes_schedule_core 776bc3d5698SJohn Baldwin 777c0855eaaSJohn Baldwin// 778c0855eaaSJohn Baldwin// .aes_schedule_192_smear 779c0855eaaSJohn Baldwin// 780c0855eaaSJohn Baldwin// Smear the short, low side in the 192-bit key schedule. 781c0855eaaSJohn Baldwin// 782c0855eaaSJohn Baldwin// Inputs: 783c0855eaaSJohn Baldwin// %xmm7: high side, b a x y 784c0855eaaSJohn Baldwin// %xmm6: low side, d c 0 0 785c0855eaaSJohn Baldwin// %xmm13: 0 786c0855eaaSJohn Baldwin// 787c0855eaaSJohn Baldwin// Outputs: 788c0855eaaSJohn Baldwin// %xmm6: b+c+d b+c 0 0 789c0855eaaSJohn Baldwin// %xmm0: b+c+d b+c b a 790c0855eaaSJohn Baldwin// 791bc3d5698SJohn Baldwin.type _vpaes_schedule_192_smear,%function 792bc3d5698SJohn Baldwin.align 4 793bc3d5698SJohn Baldwin_vpaes_schedule_192_smear: 794bc3d5698SJohn Baldwin movi v1.16b, #0 795bc3d5698SJohn Baldwin dup v0.4s, v7.s[3] 796bc3d5698SJohn Baldwin ins v1.s[3], v6.s[2] // vpshufd $0x80, %xmm6, %xmm1 # d c 0 0 -> c 0 0 0 797bc3d5698SJohn Baldwin ins v0.s[0], v7.s[2] // vpshufd $0xFE, %xmm7, %xmm0 # b a _ _ -> b b b a 798bc3d5698SJohn Baldwin eor v6.16b, v6.16b, v1.16b // vpxor %xmm1, %xmm6, %xmm6 # -> c+d c 0 0 799bc3d5698SJohn Baldwin eor v1.16b, v1.16b, v1.16b // vpxor %xmm1, %xmm1, %xmm1 800bc3d5698SJohn Baldwin eor v6.16b, v6.16b, v0.16b // vpxor %xmm0, %xmm6, %xmm6 # -> b+c+d b+c b a 801bc3d5698SJohn Baldwin mov v0.16b, v6.16b // vmovdqa %xmm6, %xmm0 802bc3d5698SJohn Baldwin ins v6.d[0], v1.d[0] // vmovhlps %xmm1, %xmm6, %xmm6 # clobber low side with zeros 803bc3d5698SJohn Baldwin ret 804bc3d5698SJohn Baldwin.size _vpaes_schedule_192_smear,.-_vpaes_schedule_192_smear 805bc3d5698SJohn Baldwin 806c0855eaaSJohn Baldwin// 807c0855eaaSJohn Baldwin// .aes_schedule_round 808c0855eaaSJohn Baldwin// 809c0855eaaSJohn Baldwin// Runs one main round of the key schedule on %xmm0, %xmm7 810c0855eaaSJohn Baldwin// 811c0855eaaSJohn Baldwin// Specifically, runs subbytes on the high dword of %xmm0 812c0855eaaSJohn Baldwin// then rotates it by one byte and xors into the low dword of 813c0855eaaSJohn Baldwin// %xmm7. 814c0855eaaSJohn Baldwin// 815c0855eaaSJohn Baldwin// Adds rcon from low byte of %xmm8, then rotates %xmm8 for 816c0855eaaSJohn Baldwin// next rcon. 817c0855eaaSJohn Baldwin// 818c0855eaaSJohn Baldwin// Smears the dwords of %xmm7 by xoring the low into the 819c0855eaaSJohn Baldwin// second low, result into third, result into highest. 820c0855eaaSJohn Baldwin// 821c0855eaaSJohn Baldwin// Returns results in %xmm7 = %xmm0. 822c0855eaaSJohn Baldwin// Clobbers %xmm1-%xmm4, %r11. 823c0855eaaSJohn Baldwin// 824bc3d5698SJohn Baldwin.type _vpaes_schedule_round,%function 825bc3d5698SJohn Baldwin.align 4 826bc3d5698SJohn Baldwin_vpaes_schedule_round: 827bc3d5698SJohn Baldwin // extract rcon from xmm8 828bc3d5698SJohn Baldwin movi v4.16b, #0 // vpxor %xmm4, %xmm4, %xmm4 829bc3d5698SJohn Baldwin ext v1.16b, v8.16b, v4.16b, #15 // vpalignr $15, %xmm8, %xmm4, %xmm1 830bc3d5698SJohn Baldwin ext v8.16b, v8.16b, v8.16b, #15 // vpalignr $15, %xmm8, %xmm8, %xmm8 831bc3d5698SJohn Baldwin eor v7.16b, v7.16b, v1.16b // vpxor %xmm1, %xmm7, %xmm7 832bc3d5698SJohn Baldwin 833bc3d5698SJohn Baldwin // rotate 834bc3d5698SJohn Baldwin dup v0.4s, v0.s[3] // vpshufd $0xFF, %xmm0, %xmm0 835bc3d5698SJohn Baldwin ext v0.16b, v0.16b, v0.16b, #1 // vpalignr $1, %xmm0, %xmm0, %xmm0 836bc3d5698SJohn Baldwin 837bc3d5698SJohn Baldwin // fall through... 838bc3d5698SJohn Baldwin 839bc3d5698SJohn Baldwin // low round: same as high round, but no rotation and no rcon. 840bc3d5698SJohn Baldwin_vpaes_schedule_low_round: 841bc3d5698SJohn Baldwin // smear xmm7 842bc3d5698SJohn Baldwin ext v1.16b, v4.16b, v7.16b, #12 // vpslldq $4, %xmm7, %xmm1 843bc3d5698SJohn Baldwin eor v7.16b, v7.16b, v1.16b // vpxor %xmm1, %xmm7, %xmm7 844bc3d5698SJohn Baldwin ext v4.16b, v4.16b, v7.16b, #8 // vpslldq $8, %xmm7, %xmm4 845bc3d5698SJohn Baldwin 846bc3d5698SJohn Baldwin // subbytes 847bc3d5698SJohn Baldwin and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 # 0 = k 848bc3d5698SJohn Baldwin ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 # 1 = i 849bc3d5698SJohn Baldwin eor v7.16b, v7.16b, v4.16b // vpxor %xmm4, %xmm7, %xmm7 850bc3d5698SJohn Baldwin tbl v2.16b, {v19.16b}, v1.16b // vpshufb %xmm1, %xmm11, %xmm2 # 2 = a/k 851bc3d5698SJohn Baldwin eor v1.16b, v1.16b, v0.16b // vpxor %xmm0, %xmm1, %xmm1 # 0 = j 852bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b}, v0.16b // vpshufb %xmm0, %xmm10, %xmm3 # 3 = 1/i 853bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 # 3 = iak = 1/i + a/k 854bc3d5698SJohn Baldwin tbl v4.16b, {v18.16b}, v1.16b // vpshufb %xmm1, %xmm10, %xmm4 # 4 = 1/j 855bc3d5698SJohn Baldwin eor v7.16b, v7.16b, v16.16b // vpxor .Lk_s63(%rip), %xmm7, %xmm7 856bc3d5698SJohn Baldwin tbl v3.16b, {v18.16b}, v3.16b // vpshufb %xmm3, %xmm10, %xmm3 # 2 = 1/iak 857bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v2.16b // vpxor %xmm2, %xmm4, %xmm4 # 4 = jak = 1/j + a/k 858bc3d5698SJohn Baldwin tbl v2.16b, {v18.16b}, v4.16b // vpshufb %xmm4, %xmm10, %xmm2 # 3 = 1/jak 859bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v1.16b // vpxor %xmm1, %xmm3, %xmm3 # 2 = io 860bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v0.16b // vpxor %xmm0, %xmm2, %xmm2 # 3 = jo 861bc3d5698SJohn Baldwin tbl v4.16b, {v23.16b}, v3.16b // vpshufb %xmm3, %xmm13, %xmm4 # 4 = sbou 862bc3d5698SJohn Baldwin tbl v1.16b, {v22.16b}, v2.16b // vpshufb %xmm2, %xmm12, %xmm1 # 0 = sb1t 863bc3d5698SJohn Baldwin eor v1.16b, v1.16b, v4.16b // vpxor %xmm4, %xmm1, %xmm1 # 0 = sbox output 864bc3d5698SJohn Baldwin 865bc3d5698SJohn Baldwin // add in smeared stuff 866bc3d5698SJohn Baldwin eor v0.16b, v1.16b, v7.16b // vpxor %xmm7, %xmm1, %xmm0 867bc3d5698SJohn Baldwin eor v7.16b, v1.16b, v7.16b // vmovdqa %xmm0, %xmm7 868bc3d5698SJohn Baldwin ret 869bc3d5698SJohn Baldwin.size _vpaes_schedule_round,.-_vpaes_schedule_round 870bc3d5698SJohn Baldwin 871c0855eaaSJohn Baldwin// 872c0855eaaSJohn Baldwin// .aes_schedule_transform 873c0855eaaSJohn Baldwin// 874c0855eaaSJohn Baldwin// Linear-transform %xmm0 according to tables at (%r11) 875c0855eaaSJohn Baldwin// 876c0855eaaSJohn Baldwin// Requires that %xmm9 = 0x0F0F... as in preheat 877c0855eaaSJohn Baldwin// Output in %xmm0 878c0855eaaSJohn Baldwin// Clobbers %xmm1, %xmm2 879c0855eaaSJohn Baldwin// 880bc3d5698SJohn Baldwin.type _vpaes_schedule_transform,%function 881bc3d5698SJohn Baldwin.align 4 882bc3d5698SJohn Baldwin_vpaes_schedule_transform: 883bc3d5698SJohn Baldwin and v1.16b, v0.16b, v17.16b // vpand %xmm9, %xmm0, %xmm1 884bc3d5698SJohn Baldwin ushr v0.16b, v0.16b, #4 // vpsrlb $4, %xmm0, %xmm0 885bc3d5698SJohn Baldwin // vmovdqa (%r11), %xmm2 # lo 886bc3d5698SJohn Baldwin tbl v2.16b, {v20.16b}, v1.16b // vpshufb %xmm1, %xmm2, %xmm2 887bc3d5698SJohn Baldwin // vmovdqa 16(%r11), %xmm1 # hi 888bc3d5698SJohn Baldwin tbl v0.16b, {v21.16b}, v0.16b // vpshufb %xmm0, %xmm1, %xmm0 889bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v2.16b // vpxor %xmm2, %xmm0, %xmm0 890bc3d5698SJohn Baldwin ret 891bc3d5698SJohn Baldwin.size _vpaes_schedule_transform,.-_vpaes_schedule_transform 892bc3d5698SJohn Baldwin 893c0855eaaSJohn Baldwin// 894c0855eaaSJohn Baldwin// .aes_schedule_mangle 895c0855eaaSJohn Baldwin// 896c0855eaaSJohn Baldwin// Mangle xmm0 from (basis-transformed) standard version 897c0855eaaSJohn Baldwin// to our version. 898c0855eaaSJohn Baldwin// 899c0855eaaSJohn Baldwin// On encrypt, 900c0855eaaSJohn Baldwin// xor with 0x63 901c0855eaaSJohn Baldwin// multiply by circulant 0,1,1,1 902c0855eaaSJohn Baldwin// apply shiftrows transform 903c0855eaaSJohn Baldwin// 904c0855eaaSJohn Baldwin// On decrypt, 905c0855eaaSJohn Baldwin// xor with 0x63 906c0855eaaSJohn Baldwin// multiply by "inverse mixcolumns" circulant E,B,D,9 907c0855eaaSJohn Baldwin// deskew 908c0855eaaSJohn Baldwin// apply shiftrows transform 909c0855eaaSJohn Baldwin// 910c0855eaaSJohn Baldwin// 911c0855eaaSJohn Baldwin// Writes out to (%rdx), and increments or decrements it 912c0855eaaSJohn Baldwin// Keeps track of round number mod 4 in %r8 913c0855eaaSJohn Baldwin// Preserves xmm0 914c0855eaaSJohn Baldwin// Clobbers xmm1-xmm5 915c0855eaaSJohn Baldwin// 916bc3d5698SJohn Baldwin.type _vpaes_schedule_mangle,%function 917bc3d5698SJohn Baldwin.align 4 918bc3d5698SJohn Baldwin_vpaes_schedule_mangle: 919bc3d5698SJohn Baldwin mov v4.16b, v0.16b // vmovdqa %xmm0, %xmm4 # save xmm0 for later 920bc3d5698SJohn Baldwin // vmovdqa .Lk_mc_forward(%rip),%xmm5 921bc3d5698SJohn Baldwin cbnz w3, .Lschedule_mangle_dec 922bc3d5698SJohn Baldwin 923bc3d5698SJohn Baldwin // encrypting 924bc3d5698SJohn Baldwin eor v4.16b, v0.16b, v16.16b // vpxor .Lk_s63(%rip), %xmm0, %xmm4 925bc3d5698SJohn Baldwin add x2, x2, #16 // add $16, %rdx 926bc3d5698SJohn Baldwin tbl v4.16b, {v4.16b}, v9.16b // vpshufb %xmm5, %xmm4, %xmm4 927bc3d5698SJohn Baldwin tbl v1.16b, {v4.16b}, v9.16b // vpshufb %xmm5, %xmm4, %xmm1 928bc3d5698SJohn Baldwin tbl v3.16b, {v1.16b}, v9.16b // vpshufb %xmm5, %xmm1, %xmm3 929bc3d5698SJohn Baldwin eor v4.16b, v4.16b, v1.16b // vpxor %xmm1, %xmm4, %xmm4 930bc3d5698SJohn Baldwin ld1 {v1.2d}, [x8] // vmovdqa (%r8,%r10), %xmm1 931bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v4.16b // vpxor %xmm4, %xmm3, %xmm3 932bc3d5698SJohn Baldwin 933bc3d5698SJohn Baldwin b .Lschedule_mangle_both 934bc3d5698SJohn Baldwin.align 4 935bc3d5698SJohn Baldwin.Lschedule_mangle_dec: 936bc3d5698SJohn Baldwin // inverse mix columns 937bc3d5698SJohn Baldwin // lea .Lk_dksd(%rip),%r11 938bc3d5698SJohn Baldwin ushr v1.16b, v4.16b, #4 // vpsrlb $4, %xmm4, %xmm1 # 1 = hi 939bc3d5698SJohn Baldwin and v4.16b, v4.16b, v17.16b // vpand %xmm9, %xmm4, %xmm4 # 4 = lo 940bc3d5698SJohn Baldwin 941bc3d5698SJohn Baldwin // vmovdqa 0x00(%r11), %xmm2 942bc3d5698SJohn Baldwin tbl v2.16b, {v24.16b}, v4.16b // vpshufb %xmm4, %xmm2, %xmm2 943bc3d5698SJohn Baldwin // vmovdqa 0x10(%r11), %xmm3 944bc3d5698SJohn Baldwin tbl v3.16b, {v25.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 945bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 946bc3d5698SJohn Baldwin tbl v3.16b, {v3.16b}, v9.16b // vpshufb %xmm5, %xmm3, %xmm3 947bc3d5698SJohn Baldwin 948bc3d5698SJohn Baldwin // vmovdqa 0x20(%r11), %xmm2 949bc3d5698SJohn Baldwin tbl v2.16b, {v26.16b}, v4.16b // vpshufb %xmm4, %xmm2, %xmm2 950bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v3.16b // vpxor %xmm3, %xmm2, %xmm2 951bc3d5698SJohn Baldwin // vmovdqa 0x30(%r11), %xmm3 952bc3d5698SJohn Baldwin tbl v3.16b, {v27.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 953bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 954bc3d5698SJohn Baldwin tbl v3.16b, {v3.16b}, v9.16b // vpshufb %xmm5, %xmm3, %xmm3 955bc3d5698SJohn Baldwin 956bc3d5698SJohn Baldwin // vmovdqa 0x40(%r11), %xmm2 957bc3d5698SJohn Baldwin tbl v2.16b, {v28.16b}, v4.16b // vpshufb %xmm4, %xmm2, %xmm2 958bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v3.16b // vpxor %xmm3, %xmm2, %xmm2 959bc3d5698SJohn Baldwin // vmovdqa 0x50(%r11), %xmm3 960bc3d5698SJohn Baldwin tbl v3.16b, {v29.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 961bc3d5698SJohn Baldwin eor v3.16b, v3.16b, v2.16b // vpxor %xmm2, %xmm3, %xmm3 962bc3d5698SJohn Baldwin 963bc3d5698SJohn Baldwin // vmovdqa 0x60(%r11), %xmm2 964bc3d5698SJohn Baldwin tbl v2.16b, {v30.16b}, v4.16b // vpshufb %xmm4, %xmm2, %xmm2 965bc3d5698SJohn Baldwin tbl v3.16b, {v3.16b}, v9.16b // vpshufb %xmm5, %xmm3, %xmm3 966bc3d5698SJohn Baldwin // vmovdqa 0x70(%r11), %xmm4 967bc3d5698SJohn Baldwin tbl v4.16b, {v31.16b}, v1.16b // vpshufb %xmm1, %xmm4, %xmm4 968bc3d5698SJohn Baldwin ld1 {v1.2d}, [x8] // vmovdqa (%r8,%r10), %xmm1 969bc3d5698SJohn Baldwin eor v2.16b, v2.16b, v3.16b // vpxor %xmm3, %xmm2, %xmm2 970bc3d5698SJohn Baldwin eor v3.16b, v4.16b, v2.16b // vpxor %xmm2, %xmm4, %xmm3 971bc3d5698SJohn Baldwin 972bc3d5698SJohn Baldwin sub x2, x2, #16 // add $-16, %rdx 973bc3d5698SJohn Baldwin 974bc3d5698SJohn Baldwin.Lschedule_mangle_both: 975bc3d5698SJohn Baldwin tbl v3.16b, {v3.16b}, v1.16b // vpshufb %xmm1, %xmm3, %xmm3 976bc3d5698SJohn Baldwin add x8, x8, #64-16 // add $-16, %r8 977bc3d5698SJohn Baldwin and x8, x8, #~(1<<6) // and $0x30, %r8 978bc3d5698SJohn Baldwin st1 {v3.2d}, [x2] // vmovdqu %xmm3, (%rdx) 979bc3d5698SJohn Baldwin ret 980bc3d5698SJohn Baldwin.size _vpaes_schedule_mangle,.-_vpaes_schedule_mangle 981bc3d5698SJohn Baldwin 982bc3d5698SJohn Baldwin.globl vpaes_set_encrypt_key 983bc3d5698SJohn Baldwin.type vpaes_set_encrypt_key,%function 984bc3d5698SJohn Baldwin.align 4 985bc3d5698SJohn Baldwinvpaes_set_encrypt_key: 986bd9588bcSAndrew Turner AARCH64_SIGN_LINK_REGISTER 987bc3d5698SJohn Baldwin stp x29,x30,[sp,#-16]! 988bc3d5698SJohn Baldwin add x29,sp,#0 989bc3d5698SJohn Baldwin stp d8,d9,[sp,#-16]! // ABI spec says so 990bc3d5698SJohn Baldwin 991bc3d5698SJohn Baldwin lsr w9, w1, #5 // shr $5,%eax 992bc3d5698SJohn Baldwin add w9, w9, #5 // $5,%eax 993bc3d5698SJohn Baldwin str w9, [x2,#240] // mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5; 994bc3d5698SJohn Baldwin 995bc3d5698SJohn Baldwin mov w3, #0 // mov $0,%ecx 996bc3d5698SJohn Baldwin mov x8, #0x30 // mov $0x30,%r8d 997bc3d5698SJohn Baldwin bl _vpaes_schedule_core 998bc3d5698SJohn Baldwin eor x0, x0, x0 999bc3d5698SJohn Baldwin 1000bc3d5698SJohn Baldwin ldp d8,d9,[sp],#16 1001bc3d5698SJohn Baldwin ldp x29,x30,[sp],#16 1002bd9588bcSAndrew Turner AARCH64_VALIDATE_LINK_REGISTER 1003bc3d5698SJohn Baldwin ret 1004bc3d5698SJohn Baldwin.size vpaes_set_encrypt_key,.-vpaes_set_encrypt_key 1005bc3d5698SJohn Baldwin 1006bc3d5698SJohn Baldwin.globl vpaes_set_decrypt_key 1007bc3d5698SJohn Baldwin.type vpaes_set_decrypt_key,%function 1008bc3d5698SJohn Baldwin.align 4 1009bc3d5698SJohn Baldwinvpaes_set_decrypt_key: 1010bd9588bcSAndrew Turner AARCH64_SIGN_LINK_REGISTER 1011bc3d5698SJohn Baldwin stp x29,x30,[sp,#-16]! 1012bc3d5698SJohn Baldwin add x29,sp,#0 1013bc3d5698SJohn Baldwin stp d8,d9,[sp,#-16]! // ABI spec says so 1014bc3d5698SJohn Baldwin 1015bc3d5698SJohn Baldwin lsr w9, w1, #5 // shr $5,%eax 1016bc3d5698SJohn Baldwin add w9, w9, #5 // $5,%eax 1017bc3d5698SJohn Baldwin str w9, [x2,#240] // mov %eax,240(%rdx) # AES_KEY->rounds = nbits/32+5; 1018bc3d5698SJohn Baldwin lsl w9, w9, #4 // shl $4,%eax 1019bc3d5698SJohn Baldwin add x2, x2, #16 // lea 16(%rdx,%rax),%rdx 1020bc3d5698SJohn Baldwin add x2, x2, x9 1021bc3d5698SJohn Baldwin 1022bc3d5698SJohn Baldwin mov w3, #1 // mov $1,%ecx 1023bc3d5698SJohn Baldwin lsr w8, w1, #1 // shr $1,%r8d 1024bc3d5698SJohn Baldwin and x8, x8, #32 // and $32,%r8d 1025bc3d5698SJohn Baldwin eor x8, x8, #32 // xor $32,%r8d # nbits==192?0:32 1026bc3d5698SJohn Baldwin bl _vpaes_schedule_core 1027bc3d5698SJohn Baldwin 1028bc3d5698SJohn Baldwin ldp d8,d9,[sp],#16 1029bc3d5698SJohn Baldwin ldp x29,x30,[sp],#16 1030bd9588bcSAndrew Turner AARCH64_VALIDATE_LINK_REGISTER 1031bc3d5698SJohn Baldwin ret 1032bc3d5698SJohn Baldwin.size vpaes_set_decrypt_key,.-vpaes_set_decrypt_key 1033bc3d5698SJohn Baldwin.globl vpaes_cbc_encrypt 1034bc3d5698SJohn Baldwin.type vpaes_cbc_encrypt,%function 1035bc3d5698SJohn Baldwin.align 4 1036bc3d5698SJohn Baldwinvpaes_cbc_encrypt: 1037bd9588bcSAndrew Turner AARCH64_SIGN_LINK_REGISTER 1038bc3d5698SJohn Baldwin cbz x2, .Lcbc_abort 1039bc3d5698SJohn Baldwin cmp w5, #0 // check direction 1040bc3d5698SJohn Baldwin b.eq vpaes_cbc_decrypt 1041bc3d5698SJohn Baldwin 1042bc3d5698SJohn Baldwin stp x29,x30,[sp,#-16]! 1043bc3d5698SJohn Baldwin add x29,sp,#0 1044bc3d5698SJohn Baldwin 1045bc3d5698SJohn Baldwin mov x17, x2 // reassign 1046bc3d5698SJohn Baldwin mov x2, x3 // reassign 1047bc3d5698SJohn Baldwin 1048bc3d5698SJohn Baldwin ld1 {v0.16b}, [x4] // load ivec 1049bc3d5698SJohn Baldwin bl _vpaes_encrypt_preheat 1050bc3d5698SJohn Baldwin b .Lcbc_enc_loop 1051bc3d5698SJohn Baldwin 1052bc3d5698SJohn Baldwin.align 4 1053bc3d5698SJohn Baldwin.Lcbc_enc_loop: 1054bc3d5698SJohn Baldwin ld1 {v7.16b}, [x0],#16 // load input 1055bc3d5698SJohn Baldwin eor v7.16b, v7.16b, v0.16b // xor with ivec 1056bc3d5698SJohn Baldwin bl _vpaes_encrypt_core 1057bc3d5698SJohn Baldwin st1 {v0.16b}, [x1],#16 // save output 1058bc3d5698SJohn Baldwin subs x17, x17, #16 1059bc3d5698SJohn Baldwin b.hi .Lcbc_enc_loop 1060bc3d5698SJohn Baldwin 1061bc3d5698SJohn Baldwin st1 {v0.16b}, [x4] // write ivec 1062bc3d5698SJohn Baldwin 1063bc3d5698SJohn Baldwin ldp x29,x30,[sp],#16 1064bc3d5698SJohn Baldwin.Lcbc_abort: 1065bd9588bcSAndrew Turner AARCH64_VALIDATE_LINK_REGISTER 1066bc3d5698SJohn Baldwin ret 1067bc3d5698SJohn Baldwin.size vpaes_cbc_encrypt,.-vpaes_cbc_encrypt 1068bc3d5698SJohn Baldwin 1069bc3d5698SJohn Baldwin.type vpaes_cbc_decrypt,%function 1070bc3d5698SJohn Baldwin.align 4 1071bc3d5698SJohn Baldwinvpaes_cbc_decrypt: 1072bd9588bcSAndrew Turner // Not adding AARCH64_SIGN_LINK_REGISTER here because vpaes_cbc_decrypt is jumped to 1073bd9588bcSAndrew Turner // only from vpaes_cbc_encrypt which has already signed the return address. 1074bc3d5698SJohn Baldwin stp x29,x30,[sp,#-16]! 1075bc3d5698SJohn Baldwin add x29,sp,#0 1076bc3d5698SJohn Baldwin stp d8,d9,[sp,#-16]! // ABI spec says so 1077bc3d5698SJohn Baldwin stp d10,d11,[sp,#-16]! 1078bc3d5698SJohn Baldwin stp d12,d13,[sp,#-16]! 1079bc3d5698SJohn Baldwin stp d14,d15,[sp,#-16]! 1080bc3d5698SJohn Baldwin 1081bc3d5698SJohn Baldwin mov x17, x2 // reassign 1082bc3d5698SJohn Baldwin mov x2, x3 // reassign 1083bc3d5698SJohn Baldwin ld1 {v6.16b}, [x4] // load ivec 1084bc3d5698SJohn Baldwin bl _vpaes_decrypt_preheat 1085bc3d5698SJohn Baldwin tst x17, #16 1086bc3d5698SJohn Baldwin b.eq .Lcbc_dec_loop2x 1087bc3d5698SJohn Baldwin 1088bc3d5698SJohn Baldwin ld1 {v7.16b}, [x0], #16 // load input 1089bc3d5698SJohn Baldwin bl _vpaes_decrypt_core 1090bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v6.16b // xor with ivec 1091bc3d5698SJohn Baldwin orr v6.16b, v7.16b, v7.16b // next ivec value 1092bc3d5698SJohn Baldwin st1 {v0.16b}, [x1], #16 1093bc3d5698SJohn Baldwin subs x17, x17, #16 1094bc3d5698SJohn Baldwin b.ls .Lcbc_dec_done 1095bc3d5698SJohn Baldwin 1096bc3d5698SJohn Baldwin.align 4 1097bc3d5698SJohn Baldwin.Lcbc_dec_loop2x: 1098bc3d5698SJohn Baldwin ld1 {v14.16b,v15.16b}, [x0], #32 1099bc3d5698SJohn Baldwin bl _vpaes_decrypt_2x 1100bc3d5698SJohn Baldwin eor v0.16b, v0.16b, v6.16b // xor with ivec 1101bc3d5698SJohn Baldwin eor v1.16b, v1.16b, v14.16b 1102bc3d5698SJohn Baldwin orr v6.16b, v15.16b, v15.16b 1103bc3d5698SJohn Baldwin st1 {v0.16b,v1.16b}, [x1], #32 1104bc3d5698SJohn Baldwin subs x17, x17, #32 1105bc3d5698SJohn Baldwin b.hi .Lcbc_dec_loop2x 1106bc3d5698SJohn Baldwin 1107bc3d5698SJohn Baldwin.Lcbc_dec_done: 1108bc3d5698SJohn Baldwin st1 {v6.16b}, [x4] 1109bc3d5698SJohn Baldwin 1110bc3d5698SJohn Baldwin ldp d14,d15,[sp],#16 1111bc3d5698SJohn Baldwin ldp d12,d13,[sp],#16 1112bc3d5698SJohn Baldwin ldp d10,d11,[sp],#16 1113bc3d5698SJohn Baldwin ldp d8,d9,[sp],#16 1114bc3d5698SJohn Baldwin ldp x29,x30,[sp],#16 1115bd9588bcSAndrew Turner AARCH64_VALIDATE_LINK_REGISTER 1116bc3d5698SJohn Baldwin ret 1117bc3d5698SJohn Baldwin.size vpaes_cbc_decrypt,.-vpaes_cbc_decrypt 1118bc3d5698SJohn Baldwin.globl vpaes_ecb_encrypt 1119bc3d5698SJohn Baldwin.type vpaes_ecb_encrypt,%function 1120bc3d5698SJohn Baldwin.align 4 1121bc3d5698SJohn Baldwinvpaes_ecb_encrypt: 1122bd9588bcSAndrew Turner AARCH64_SIGN_LINK_REGISTER 1123bc3d5698SJohn Baldwin stp x29,x30,[sp,#-16]! 1124bc3d5698SJohn Baldwin add x29,sp,#0 1125bc3d5698SJohn Baldwin stp d8,d9,[sp,#-16]! // ABI spec says so 1126bc3d5698SJohn Baldwin stp d10,d11,[sp,#-16]! 1127bc3d5698SJohn Baldwin stp d12,d13,[sp,#-16]! 1128bc3d5698SJohn Baldwin stp d14,d15,[sp,#-16]! 1129bc3d5698SJohn Baldwin 1130bc3d5698SJohn Baldwin mov x17, x2 1131bc3d5698SJohn Baldwin mov x2, x3 1132bc3d5698SJohn Baldwin bl _vpaes_encrypt_preheat 1133bc3d5698SJohn Baldwin tst x17, #16 1134bc3d5698SJohn Baldwin b.eq .Lecb_enc_loop 1135bc3d5698SJohn Baldwin 1136bc3d5698SJohn Baldwin ld1 {v7.16b}, [x0],#16 1137bc3d5698SJohn Baldwin bl _vpaes_encrypt_core 1138bc3d5698SJohn Baldwin st1 {v0.16b}, [x1],#16 1139bc3d5698SJohn Baldwin subs x17, x17, #16 1140bc3d5698SJohn Baldwin b.ls .Lecb_enc_done 1141bc3d5698SJohn Baldwin 1142bc3d5698SJohn Baldwin.align 4 1143bc3d5698SJohn Baldwin.Lecb_enc_loop: 1144bc3d5698SJohn Baldwin ld1 {v14.16b,v15.16b}, [x0], #32 1145bc3d5698SJohn Baldwin bl _vpaes_encrypt_2x 1146bc3d5698SJohn Baldwin st1 {v0.16b,v1.16b}, [x1], #32 1147bc3d5698SJohn Baldwin subs x17, x17, #32 1148bc3d5698SJohn Baldwin b.hi .Lecb_enc_loop 1149bc3d5698SJohn Baldwin 1150bc3d5698SJohn Baldwin.Lecb_enc_done: 1151bc3d5698SJohn Baldwin ldp d14,d15,[sp],#16 1152bc3d5698SJohn Baldwin ldp d12,d13,[sp],#16 1153bc3d5698SJohn Baldwin ldp d10,d11,[sp],#16 1154bc3d5698SJohn Baldwin ldp d8,d9,[sp],#16 1155bc3d5698SJohn Baldwin ldp x29,x30,[sp],#16 1156bd9588bcSAndrew Turner AARCH64_VALIDATE_LINK_REGISTER 1157bc3d5698SJohn Baldwin ret 1158bc3d5698SJohn Baldwin.size vpaes_ecb_encrypt,.-vpaes_ecb_encrypt 1159bc3d5698SJohn Baldwin 1160bc3d5698SJohn Baldwin.globl vpaes_ecb_decrypt 1161bc3d5698SJohn Baldwin.type vpaes_ecb_decrypt,%function 1162bc3d5698SJohn Baldwin.align 4 1163bc3d5698SJohn Baldwinvpaes_ecb_decrypt: 1164bd9588bcSAndrew Turner AARCH64_SIGN_LINK_REGISTER 1165bc3d5698SJohn Baldwin stp x29,x30,[sp,#-16]! 1166bc3d5698SJohn Baldwin add x29,sp,#0 1167bc3d5698SJohn Baldwin stp d8,d9,[sp,#-16]! // ABI spec says so 1168bc3d5698SJohn Baldwin stp d10,d11,[sp,#-16]! 1169bc3d5698SJohn Baldwin stp d12,d13,[sp,#-16]! 1170bc3d5698SJohn Baldwin stp d14,d15,[sp,#-16]! 1171bc3d5698SJohn Baldwin 1172bc3d5698SJohn Baldwin mov x17, x2 1173bc3d5698SJohn Baldwin mov x2, x3 1174bc3d5698SJohn Baldwin bl _vpaes_decrypt_preheat 1175bc3d5698SJohn Baldwin tst x17, #16 1176bc3d5698SJohn Baldwin b.eq .Lecb_dec_loop 1177bc3d5698SJohn Baldwin 1178bc3d5698SJohn Baldwin ld1 {v7.16b}, [x0],#16 1179bc3d5698SJohn Baldwin bl _vpaes_encrypt_core 1180bc3d5698SJohn Baldwin st1 {v0.16b}, [x1],#16 1181bc3d5698SJohn Baldwin subs x17, x17, #16 1182bc3d5698SJohn Baldwin b.ls .Lecb_dec_done 1183bc3d5698SJohn Baldwin 1184bc3d5698SJohn Baldwin.align 4 1185bc3d5698SJohn Baldwin.Lecb_dec_loop: 1186bc3d5698SJohn Baldwin ld1 {v14.16b,v15.16b}, [x0], #32 1187bc3d5698SJohn Baldwin bl _vpaes_decrypt_2x 1188bc3d5698SJohn Baldwin st1 {v0.16b,v1.16b}, [x1], #32 1189bc3d5698SJohn Baldwin subs x17, x17, #32 1190bc3d5698SJohn Baldwin b.hi .Lecb_dec_loop 1191bc3d5698SJohn Baldwin 1192bc3d5698SJohn Baldwin.Lecb_dec_done: 1193bc3d5698SJohn Baldwin ldp d14,d15,[sp],#16 1194bc3d5698SJohn Baldwin ldp d12,d13,[sp],#16 1195bc3d5698SJohn Baldwin ldp d10,d11,[sp],#16 1196bc3d5698SJohn Baldwin ldp d8,d9,[sp],#16 1197bc3d5698SJohn Baldwin ldp x29,x30,[sp],#16 1198bd9588bcSAndrew Turner AARCH64_VALIDATE_LINK_REGISTER 1199bc3d5698SJohn Baldwin ret 1200bc3d5698SJohn Baldwin.size vpaes_ecb_decrypt,.-vpaes_ecb_decrypt 1201