blob: ddfd5a3fcc5fb8ccc781189c9e4387491eb21517 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Twofish for CryptoAPI
3 *
4 * Originally Twofish for GPG
5 * By Matthew Skala <mskala@ansuz.sooke.bc.ca>, July 26, 1998
6 * 256-bit key length added March 20, 1999
7 * Some modifications to reduce the text size by Werner Koch, April, 1998
8 * Ported to the kerneli patch by Marc Mutz <Marc@Mutz.com>
9 * Ported to CryptoAPI by Colin Slater <hoho@tacomeat.net>
10 *
11 * The original author has disclaimed all copyright interest in this
12 * code and thus put it in the public domain. The subsequent authors
13 * have put this under the GNU General Public License.
14 *
15 * This program is free software; you can redistribute it and/or modify
16 * it under the terms of the GNU General Public License as published by
17 * the Free Software Foundation; either version 2 of the License, or
18 * (at your option) any later version.
19 *
20 * This program is distributed in the hope that it will be useful,
21 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23 * GNU General Public License for more details.
24 *
25 * You should have received a copy of the GNU General Public License
26 * along with this program; if not, write to the Free Software
27 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
28 * USA
29 *
30 * This code is a "clean room" implementation, written from the paper
31 * _Twofish: A 128-Bit Block Cipher_ by Bruce Schneier, John Kelsey,
32 * Doug Whiting, David Wagner, Chris Hall, and Niels Ferguson, available
33 * through http://www.counterpane.com/twofish.html
34 *
35 * For background information on multiplication in finite fields, used for
36 * the matrix operations in the key schedule, see the book _Contemporary
37 * Abstract Algebra_ by Joseph A. Gallian, especially chapter 22 in the
38 * Third Edition.
39 */
Herbert Xu06ace7a2005-10-30 21:25:15 +110040
41#include <asm/byteorder.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#include <linux/module.h>
43#include <linux/init.h>
44#include <linux/types.h>
45#include <linux/errno.h>
46#include <linux/crypto.h>
Denis Vlasenkoa5f8c472006-01-16 17:42:28 +110047#include <linux/bitops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
49
50/* The large precomputed tables for the Twofish cipher (twofish.c)
51 * Taken from the same source as twofish.c
52 * Marc Mutz <Marc@Mutz.com>
53 */
54
55/* These two tables are the q0 and q1 permutations, exactly as described in
56 * the Twofish paper. */
57
58static const u8 q0[256] = {
59 0xA9, 0x67, 0xB3, 0xE8, 0x04, 0xFD, 0xA3, 0x76, 0x9A, 0x92, 0x80, 0x78,
60 0xE4, 0xDD, 0xD1, 0x38, 0x0D, 0xC6, 0x35, 0x98, 0x18, 0xF7, 0xEC, 0x6C,
61 0x43, 0x75, 0x37, 0x26, 0xFA, 0x13, 0x94, 0x48, 0xF2, 0xD0, 0x8B, 0x30,
62 0x84, 0x54, 0xDF, 0x23, 0x19, 0x5B, 0x3D, 0x59, 0xF3, 0xAE, 0xA2, 0x82,
63 0x63, 0x01, 0x83, 0x2E, 0xD9, 0x51, 0x9B, 0x7C, 0xA6, 0xEB, 0xA5, 0xBE,
64 0x16, 0x0C, 0xE3, 0x61, 0xC0, 0x8C, 0x3A, 0xF5, 0x73, 0x2C, 0x25, 0x0B,
65 0xBB, 0x4E, 0x89, 0x6B, 0x53, 0x6A, 0xB4, 0xF1, 0xE1, 0xE6, 0xBD, 0x45,
66 0xE2, 0xF4, 0xB6, 0x66, 0xCC, 0x95, 0x03, 0x56, 0xD4, 0x1C, 0x1E, 0xD7,
67 0xFB, 0xC3, 0x8E, 0xB5, 0xE9, 0xCF, 0xBF, 0xBA, 0xEA, 0x77, 0x39, 0xAF,
68 0x33, 0xC9, 0x62, 0x71, 0x81, 0x79, 0x09, 0xAD, 0x24, 0xCD, 0xF9, 0xD8,
69 0xE5, 0xC5, 0xB9, 0x4D, 0x44, 0x08, 0x86, 0xE7, 0xA1, 0x1D, 0xAA, 0xED,
70 0x06, 0x70, 0xB2, 0xD2, 0x41, 0x7B, 0xA0, 0x11, 0x31, 0xC2, 0x27, 0x90,
71 0x20, 0xF6, 0x60, 0xFF, 0x96, 0x5C, 0xB1, 0xAB, 0x9E, 0x9C, 0x52, 0x1B,
72 0x5F, 0x93, 0x0A, 0xEF, 0x91, 0x85, 0x49, 0xEE, 0x2D, 0x4F, 0x8F, 0x3B,
73 0x47, 0x87, 0x6D, 0x46, 0xD6, 0x3E, 0x69, 0x64, 0x2A, 0xCE, 0xCB, 0x2F,
74 0xFC, 0x97, 0x05, 0x7A, 0xAC, 0x7F, 0xD5, 0x1A, 0x4B, 0x0E, 0xA7, 0x5A,
75 0x28, 0x14, 0x3F, 0x29, 0x88, 0x3C, 0x4C, 0x02, 0xB8, 0xDA, 0xB0, 0x17,
76 0x55, 0x1F, 0x8A, 0x7D, 0x57, 0xC7, 0x8D, 0x74, 0xB7, 0xC4, 0x9F, 0x72,
77 0x7E, 0x15, 0x22, 0x12, 0x58, 0x07, 0x99, 0x34, 0x6E, 0x50, 0xDE, 0x68,
78 0x65, 0xBC, 0xDB, 0xF8, 0xC8, 0xA8, 0x2B, 0x40, 0xDC, 0xFE, 0x32, 0xA4,
79 0xCA, 0x10, 0x21, 0xF0, 0xD3, 0x5D, 0x0F, 0x00, 0x6F, 0x9D, 0x36, 0x42,
80 0x4A, 0x5E, 0xC1, 0xE0
81};
82
83static const u8 q1[256] = {
84 0x75, 0xF3, 0xC6, 0xF4, 0xDB, 0x7B, 0xFB, 0xC8, 0x4A, 0xD3, 0xE6, 0x6B,
85 0x45, 0x7D, 0xE8, 0x4B, 0xD6, 0x32, 0xD8, 0xFD, 0x37, 0x71, 0xF1, 0xE1,
86 0x30, 0x0F, 0xF8, 0x1B, 0x87, 0xFA, 0x06, 0x3F, 0x5E, 0xBA, 0xAE, 0x5B,
87 0x8A, 0x00, 0xBC, 0x9D, 0x6D, 0xC1, 0xB1, 0x0E, 0x80, 0x5D, 0xD2, 0xD5,
88 0xA0, 0x84, 0x07, 0x14, 0xB5, 0x90, 0x2C, 0xA3, 0xB2, 0x73, 0x4C, 0x54,
89 0x92, 0x74, 0x36, 0x51, 0x38, 0xB0, 0xBD, 0x5A, 0xFC, 0x60, 0x62, 0x96,
90 0x6C, 0x42, 0xF7, 0x10, 0x7C, 0x28, 0x27, 0x8C, 0x13, 0x95, 0x9C, 0xC7,
91 0x24, 0x46, 0x3B, 0x70, 0xCA, 0xE3, 0x85, 0xCB, 0x11, 0xD0, 0x93, 0xB8,
92 0xA6, 0x83, 0x20, 0xFF, 0x9F, 0x77, 0xC3, 0xCC, 0x03, 0x6F, 0x08, 0xBF,
93 0x40, 0xE7, 0x2B, 0xE2, 0x79, 0x0C, 0xAA, 0x82, 0x41, 0x3A, 0xEA, 0xB9,
94 0xE4, 0x9A, 0xA4, 0x97, 0x7E, 0xDA, 0x7A, 0x17, 0x66, 0x94, 0xA1, 0x1D,
95 0x3D, 0xF0, 0xDE, 0xB3, 0x0B, 0x72, 0xA7, 0x1C, 0xEF, 0xD1, 0x53, 0x3E,
96 0x8F, 0x33, 0x26, 0x5F, 0xEC, 0x76, 0x2A, 0x49, 0x81, 0x88, 0xEE, 0x21,
97 0xC4, 0x1A, 0xEB, 0xD9, 0xC5, 0x39, 0x99, 0xCD, 0xAD, 0x31, 0x8B, 0x01,
98 0x18, 0x23, 0xDD, 0x1F, 0x4E, 0x2D, 0xF9, 0x48, 0x4F, 0xF2, 0x65, 0x8E,
99 0x78, 0x5C, 0x58, 0x19, 0x8D, 0xE5, 0x98, 0x57, 0x67, 0x7F, 0x05, 0x64,
100 0xAF, 0x63, 0xB6, 0xFE, 0xF5, 0xB7, 0x3C, 0xA5, 0xCE, 0xE9, 0x68, 0x44,
101 0xE0, 0x4D, 0x43, 0x69, 0x29, 0x2E, 0xAC, 0x15, 0x59, 0xA8, 0x0A, 0x9E,
102 0x6E, 0x47, 0xDF, 0x34, 0x35, 0x6A, 0xCF, 0xDC, 0x22, 0xC9, 0xC0, 0x9B,
103 0x89, 0xD4, 0xED, 0xAB, 0x12, 0xA2, 0x0D, 0x52, 0xBB, 0x02, 0x2F, 0xA9,
104 0xD7, 0x61, 0x1E, 0xB4, 0x50, 0x04, 0xF6, 0xC2, 0x16, 0x25, 0x86, 0x56,
105 0x55, 0x09, 0xBE, 0x91
106};
107
108/* These MDS tables are actually tables of MDS composed with q0 and q1,
109 * because it is only ever used that way and we can save some time by
110 * precomputing. Of course the main saving comes from precomputing the
111 * GF(2^8) multiplication involved in the MDS matrix multiply; by looking
112 * things up in these tables we reduce the matrix multiply to four lookups
113 * and three XORs. Semi-formally, the definition of these tables is:
114 * mds[0][i] = MDS (q1[i] 0 0 0)^T mds[1][i] = MDS (0 q0[i] 0 0)^T
115 * mds[2][i] = MDS (0 0 q1[i] 0)^T mds[3][i] = MDS (0 0 0 q0[i])^T
116 * where ^T means "transpose", the matrix multiply is performed in GF(2^8)
117 * represented as GF(2)[x]/v(x) where v(x)=x^8+x^6+x^5+x^3+1 as described
118 * by Schneier et al, and I'm casually glossing over the byte/word
119 * conversion issues. */
120
121static const u32 mds[4][256] = {
122 {0xBCBC3275, 0xECEC21F3, 0x202043C6, 0xB3B3C9F4, 0xDADA03DB, 0x02028B7B,
123 0xE2E22BFB, 0x9E9EFAC8, 0xC9C9EC4A, 0xD4D409D3, 0x18186BE6, 0x1E1E9F6B,
124 0x98980E45, 0xB2B2387D, 0xA6A6D2E8, 0x2626B74B, 0x3C3C57D6, 0x93938A32,
125 0x8282EED8, 0x525298FD, 0x7B7BD437, 0xBBBB3771, 0x5B5B97F1, 0x474783E1,
126 0x24243C30, 0x5151E20F, 0xBABAC6F8, 0x4A4AF31B, 0xBFBF4887, 0x0D0D70FA,
127 0xB0B0B306, 0x7575DE3F, 0xD2D2FD5E, 0x7D7D20BA, 0x666631AE, 0x3A3AA35B,
128 0x59591C8A, 0x00000000, 0xCDCD93BC, 0x1A1AE09D, 0xAEAE2C6D, 0x7F7FABC1,
129 0x2B2BC7B1, 0xBEBEB90E, 0xE0E0A080, 0x8A8A105D, 0x3B3B52D2, 0x6464BAD5,
130 0xD8D888A0, 0xE7E7A584, 0x5F5FE807, 0x1B1B1114, 0x2C2CC2B5, 0xFCFCB490,
131 0x3131272C, 0x808065A3, 0x73732AB2, 0x0C0C8173, 0x79795F4C, 0x6B6B4154,
132 0x4B4B0292, 0x53536974, 0x94948F36, 0x83831F51, 0x2A2A3638, 0xC4C49CB0,
133 0x2222C8BD, 0xD5D5F85A, 0xBDBDC3FC, 0x48487860, 0xFFFFCE62, 0x4C4C0796,
134 0x4141776C, 0xC7C7E642, 0xEBEB24F7, 0x1C1C1410, 0x5D5D637C, 0x36362228,
135 0x6767C027, 0xE9E9AF8C, 0x4444F913, 0x1414EA95, 0xF5F5BB9C, 0xCFCF18C7,
136 0x3F3F2D24, 0xC0C0E346, 0x7272DB3B, 0x54546C70, 0x29294CCA, 0xF0F035E3,
137 0x0808FE85, 0xC6C617CB, 0xF3F34F11, 0x8C8CE4D0, 0xA4A45993, 0xCACA96B8,
138 0x68683BA6, 0xB8B84D83, 0x38382820, 0xE5E52EFF, 0xADAD569F, 0x0B0B8477,
139 0xC8C81DC3, 0x9999FFCC, 0x5858ED03, 0x19199A6F, 0x0E0E0A08, 0x95957EBF,
140 0x70705040, 0xF7F730E7, 0x6E6ECF2B, 0x1F1F6EE2, 0xB5B53D79, 0x09090F0C,
141 0x616134AA, 0x57571682, 0x9F9F0B41, 0x9D9D803A, 0x111164EA, 0x2525CDB9,
142 0xAFAFDDE4, 0x4545089A, 0xDFDF8DA4, 0xA3A35C97, 0xEAEAD57E, 0x353558DA,
143 0xEDEDD07A, 0x4343FC17, 0xF8F8CB66, 0xFBFBB194, 0x3737D3A1, 0xFAFA401D,
144 0xC2C2683D, 0xB4B4CCF0, 0x32325DDE, 0x9C9C71B3, 0x5656E70B, 0xE3E3DA72,
145 0x878760A7, 0x15151B1C, 0xF9F93AEF, 0x6363BFD1, 0x3434A953, 0x9A9A853E,
146 0xB1B1428F, 0x7C7CD133, 0x88889B26, 0x3D3DA65F, 0xA1A1D7EC, 0xE4E4DF76,
147 0x8181942A, 0x91910149, 0x0F0FFB81, 0xEEEEAA88, 0x161661EE, 0xD7D77321,
148 0x9797F5C4, 0xA5A5A81A, 0xFEFE3FEB, 0x6D6DB5D9, 0x7878AEC5, 0xC5C56D39,
149 0x1D1DE599, 0x7676A4CD, 0x3E3EDCAD, 0xCBCB6731, 0xB6B6478B, 0xEFEF5B01,
150 0x12121E18, 0x6060C523, 0x6A6AB0DD, 0x4D4DF61F, 0xCECEE94E, 0xDEDE7C2D,
151 0x55559DF9, 0x7E7E5A48, 0x2121B24F, 0x03037AF2, 0xA0A02665, 0x5E5E198E,
152 0x5A5A6678, 0x65654B5C, 0x62624E58, 0xFDFD4519, 0x0606F48D, 0x404086E5,
153 0xF2F2BE98, 0x3333AC57, 0x17179067, 0x05058E7F, 0xE8E85E05, 0x4F4F7D64,
154 0x89896AAF, 0x10109563, 0x74742FB6, 0x0A0A75FE, 0x5C5C92F5, 0x9B9B74B7,
155 0x2D2D333C, 0x3030D6A5, 0x2E2E49CE, 0x494989E9, 0x46467268, 0x77775544,
156 0xA8A8D8E0, 0x9696044D, 0x2828BD43, 0xA9A92969, 0xD9D97929, 0x8686912E,
157 0xD1D187AC, 0xF4F44A15, 0x8D8D1559, 0xD6D682A8, 0xB9B9BC0A, 0x42420D9E,
158 0xF6F6C16E, 0x2F2FB847, 0xDDDD06DF, 0x23233934, 0xCCCC6235, 0xF1F1C46A,
159 0xC1C112CF, 0x8585EBDC, 0x8F8F9E22, 0x7171A1C9, 0x9090F0C0, 0xAAAA539B,
160 0x0101F189, 0x8B8BE1D4, 0x4E4E8CED, 0x8E8E6FAB, 0xABABA212, 0x6F6F3EA2,
161 0xE6E6540D, 0xDBDBF252, 0x92927BBB, 0xB7B7B602, 0x6969CA2F, 0x3939D9A9,
162 0xD3D30CD7, 0xA7A72361, 0xA2A2AD1E, 0xC3C399B4, 0x6C6C4450, 0x07070504,
163 0x04047FF6, 0x272746C2, 0xACACA716, 0xD0D07625, 0x50501386, 0xDCDCF756,
164 0x84841A55, 0xE1E15109, 0x7A7A25BE, 0x1313EF91},
165
166 {0xA9D93939, 0x67901717, 0xB3719C9C, 0xE8D2A6A6, 0x04050707, 0xFD985252,
167 0xA3658080, 0x76DFE4E4, 0x9A084545, 0x92024B4B, 0x80A0E0E0, 0x78665A5A,
168 0xE4DDAFAF, 0xDDB06A6A, 0xD1BF6363, 0x38362A2A, 0x0D54E6E6, 0xC6432020,
169 0x3562CCCC, 0x98BEF2F2, 0x181E1212, 0xF724EBEB, 0xECD7A1A1, 0x6C774141,
170 0x43BD2828, 0x7532BCBC, 0x37D47B7B, 0x269B8888, 0xFA700D0D, 0x13F94444,
171 0x94B1FBFB, 0x485A7E7E, 0xF27A0303, 0xD0E48C8C, 0x8B47B6B6, 0x303C2424,
172 0x84A5E7E7, 0x54416B6B, 0xDF06DDDD, 0x23C56060, 0x1945FDFD, 0x5BA33A3A,
173 0x3D68C2C2, 0x59158D8D, 0xF321ECEC, 0xAE316666, 0xA23E6F6F, 0x82165757,
174 0x63951010, 0x015BEFEF, 0x834DB8B8, 0x2E918686, 0xD9B56D6D, 0x511F8383,
175 0x9B53AAAA, 0x7C635D5D, 0xA63B6868, 0xEB3FFEFE, 0xA5D63030, 0xBE257A7A,
176 0x16A7ACAC, 0x0C0F0909, 0xE335F0F0, 0x6123A7A7, 0xC0F09090, 0x8CAFE9E9,
177 0x3A809D9D, 0xF5925C5C, 0x73810C0C, 0x2C273131, 0x2576D0D0, 0x0BE75656,
178 0xBB7B9292, 0x4EE9CECE, 0x89F10101, 0x6B9F1E1E, 0x53A93434, 0x6AC4F1F1,
179 0xB499C3C3, 0xF1975B5B, 0xE1834747, 0xE66B1818, 0xBDC82222, 0x450E9898,
180 0xE26E1F1F, 0xF4C9B3B3, 0xB62F7474, 0x66CBF8F8, 0xCCFF9999, 0x95EA1414,
181 0x03ED5858, 0x56F7DCDC, 0xD4E18B8B, 0x1C1B1515, 0x1EADA2A2, 0xD70CD3D3,
182 0xFB2BE2E2, 0xC31DC8C8, 0x8E195E5E, 0xB5C22C2C, 0xE9894949, 0xCF12C1C1,
183 0xBF7E9595, 0xBA207D7D, 0xEA641111, 0x77840B0B, 0x396DC5C5, 0xAF6A8989,
184 0x33D17C7C, 0xC9A17171, 0x62CEFFFF, 0x7137BBBB, 0x81FB0F0F, 0x793DB5B5,
185 0x0951E1E1, 0xADDC3E3E, 0x242D3F3F, 0xCDA47676, 0xF99D5555, 0xD8EE8282,
186 0xE5864040, 0xC5AE7878, 0xB9CD2525, 0x4D049696, 0x44557777, 0x080A0E0E,
187 0x86135050, 0xE730F7F7, 0xA1D33737, 0x1D40FAFA, 0xAA346161, 0xED8C4E4E,
188 0x06B3B0B0, 0x706C5454, 0xB22A7373, 0xD2523B3B, 0x410B9F9F, 0x7B8B0202,
189 0xA088D8D8, 0x114FF3F3, 0x3167CBCB, 0xC2462727, 0x27C06767, 0x90B4FCFC,
190 0x20283838, 0xF67F0404, 0x60784848, 0xFF2EE5E5, 0x96074C4C, 0x5C4B6565,
191 0xB1C72B2B, 0xAB6F8E8E, 0x9E0D4242, 0x9CBBF5F5, 0x52F2DBDB, 0x1BF34A4A,
192 0x5FA63D3D, 0x9359A4A4, 0x0ABCB9B9, 0xEF3AF9F9, 0x91EF1313, 0x85FE0808,
193 0x49019191, 0xEE611616, 0x2D7CDEDE, 0x4FB22121, 0x8F42B1B1, 0x3BDB7272,
194 0x47B82F2F, 0x8748BFBF, 0x6D2CAEAE, 0x46E3C0C0, 0xD6573C3C, 0x3E859A9A,
195 0x6929A9A9, 0x647D4F4F, 0x2A948181, 0xCE492E2E, 0xCB17C6C6, 0x2FCA6969,
196 0xFCC3BDBD, 0x975CA3A3, 0x055EE8E8, 0x7AD0EDED, 0xAC87D1D1, 0x7F8E0505,
197 0xD5BA6464, 0x1AA8A5A5, 0x4BB72626, 0x0EB9BEBE, 0xA7608787, 0x5AF8D5D5,
198 0x28223636, 0x14111B1B, 0x3FDE7575, 0x2979D9D9, 0x88AAEEEE, 0x3C332D2D,
199 0x4C5F7979, 0x02B6B7B7, 0xB896CACA, 0xDA583535, 0xB09CC4C4, 0x17FC4343,
200 0x551A8484, 0x1FF64D4D, 0x8A1C5959, 0x7D38B2B2, 0x57AC3333, 0xC718CFCF,
201 0x8DF40606, 0x74695353, 0xB7749B9B, 0xC4F59797, 0x9F56ADAD, 0x72DAE3E3,
202 0x7ED5EAEA, 0x154AF4F4, 0x229E8F8F, 0x12A2ABAB, 0x584E6262, 0x07E85F5F,
203 0x99E51D1D, 0x34392323, 0x6EC1F6F6, 0x50446C6C, 0xDE5D3232, 0x68724646,
204 0x6526A0A0, 0xBC93CDCD, 0xDB03DADA, 0xF8C6BABA, 0xC8FA9E9E, 0xA882D6D6,
205 0x2BCF6E6E, 0x40507070, 0xDCEB8585, 0xFE750A0A, 0x328A9393, 0xA48DDFDF,
206 0xCA4C2929, 0x10141C1C, 0x2173D7D7, 0xF0CCB4B4, 0xD309D4D4, 0x5D108A8A,
207 0x0FE25151, 0x00000000, 0x6F9A1919, 0x9DE01A1A, 0x368F9494, 0x42E6C7C7,
208 0x4AECC9C9, 0x5EFDD2D2, 0xC1AB7F7F, 0xE0D8A8A8},
209
210 {0xBC75BC32, 0xECF3EC21, 0x20C62043, 0xB3F4B3C9, 0xDADBDA03, 0x027B028B,
211 0xE2FBE22B, 0x9EC89EFA, 0xC94AC9EC, 0xD4D3D409, 0x18E6186B, 0x1E6B1E9F,
212 0x9845980E, 0xB27DB238, 0xA6E8A6D2, 0x264B26B7, 0x3CD63C57, 0x9332938A,
213 0x82D882EE, 0x52FD5298, 0x7B377BD4, 0xBB71BB37, 0x5BF15B97, 0x47E14783,
214 0x2430243C, 0x510F51E2, 0xBAF8BAC6, 0x4A1B4AF3, 0xBF87BF48, 0x0DFA0D70,
215 0xB006B0B3, 0x753F75DE, 0xD25ED2FD, 0x7DBA7D20, 0x66AE6631, 0x3A5B3AA3,
216 0x598A591C, 0x00000000, 0xCDBCCD93, 0x1A9D1AE0, 0xAE6DAE2C, 0x7FC17FAB,
217 0x2BB12BC7, 0xBE0EBEB9, 0xE080E0A0, 0x8A5D8A10, 0x3BD23B52, 0x64D564BA,
218 0xD8A0D888, 0xE784E7A5, 0x5F075FE8, 0x1B141B11, 0x2CB52CC2, 0xFC90FCB4,
219 0x312C3127, 0x80A38065, 0x73B2732A, 0x0C730C81, 0x794C795F, 0x6B546B41,
220 0x4B924B02, 0x53745369, 0x9436948F, 0x8351831F, 0x2A382A36, 0xC4B0C49C,
221 0x22BD22C8, 0xD55AD5F8, 0xBDFCBDC3, 0x48604878, 0xFF62FFCE, 0x4C964C07,
222 0x416C4177, 0xC742C7E6, 0xEBF7EB24, 0x1C101C14, 0x5D7C5D63, 0x36283622,
223 0x672767C0, 0xE98CE9AF, 0x441344F9, 0x149514EA, 0xF59CF5BB, 0xCFC7CF18,
224 0x3F243F2D, 0xC046C0E3, 0x723B72DB, 0x5470546C, 0x29CA294C, 0xF0E3F035,
225 0x088508FE, 0xC6CBC617, 0xF311F34F, 0x8CD08CE4, 0xA493A459, 0xCAB8CA96,
226 0x68A6683B, 0xB883B84D, 0x38203828, 0xE5FFE52E, 0xAD9FAD56, 0x0B770B84,
227 0xC8C3C81D, 0x99CC99FF, 0x580358ED, 0x196F199A, 0x0E080E0A, 0x95BF957E,
228 0x70407050, 0xF7E7F730, 0x6E2B6ECF, 0x1FE21F6E, 0xB579B53D, 0x090C090F,
229 0x61AA6134, 0x57825716, 0x9F419F0B, 0x9D3A9D80, 0x11EA1164, 0x25B925CD,
230 0xAFE4AFDD, 0x459A4508, 0xDFA4DF8D, 0xA397A35C, 0xEA7EEAD5, 0x35DA3558,
231 0xED7AEDD0, 0x431743FC, 0xF866F8CB, 0xFB94FBB1, 0x37A137D3, 0xFA1DFA40,
232 0xC23DC268, 0xB4F0B4CC, 0x32DE325D, 0x9CB39C71, 0x560B56E7, 0xE372E3DA,
233 0x87A78760, 0x151C151B, 0xF9EFF93A, 0x63D163BF, 0x345334A9, 0x9A3E9A85,
234 0xB18FB142, 0x7C337CD1, 0x8826889B, 0x3D5F3DA6, 0xA1ECA1D7, 0xE476E4DF,
235 0x812A8194, 0x91499101, 0x0F810FFB, 0xEE88EEAA, 0x16EE1661, 0xD721D773,
236 0x97C497F5, 0xA51AA5A8, 0xFEEBFE3F, 0x6DD96DB5, 0x78C578AE, 0xC539C56D,
237 0x1D991DE5, 0x76CD76A4, 0x3EAD3EDC, 0xCB31CB67, 0xB68BB647, 0xEF01EF5B,
238 0x1218121E, 0x602360C5, 0x6ADD6AB0, 0x4D1F4DF6, 0xCE4ECEE9, 0xDE2DDE7C,
239 0x55F9559D, 0x7E487E5A, 0x214F21B2, 0x03F2037A, 0xA065A026, 0x5E8E5E19,
240 0x5A785A66, 0x655C654B, 0x6258624E, 0xFD19FD45, 0x068D06F4, 0x40E54086,
241 0xF298F2BE, 0x335733AC, 0x17671790, 0x057F058E, 0xE805E85E, 0x4F644F7D,
242 0x89AF896A, 0x10631095, 0x74B6742F, 0x0AFE0A75, 0x5CF55C92, 0x9BB79B74,
243 0x2D3C2D33, 0x30A530D6, 0x2ECE2E49, 0x49E94989, 0x46684672, 0x77447755,
244 0xA8E0A8D8, 0x964D9604, 0x284328BD, 0xA969A929, 0xD929D979, 0x862E8691,
245 0xD1ACD187, 0xF415F44A, 0x8D598D15, 0xD6A8D682, 0xB90AB9BC, 0x429E420D,
246 0xF66EF6C1, 0x2F472FB8, 0xDDDFDD06, 0x23342339, 0xCC35CC62, 0xF16AF1C4,
247 0xC1CFC112, 0x85DC85EB, 0x8F228F9E, 0x71C971A1, 0x90C090F0, 0xAA9BAA53,
248 0x018901F1, 0x8BD48BE1, 0x4EED4E8C, 0x8EAB8E6F, 0xAB12ABA2, 0x6FA26F3E,
249 0xE60DE654, 0xDB52DBF2, 0x92BB927B, 0xB702B7B6, 0x692F69CA, 0x39A939D9,
250 0xD3D7D30C, 0xA761A723, 0xA21EA2AD, 0xC3B4C399, 0x6C506C44, 0x07040705,
251 0x04F6047F, 0x27C22746, 0xAC16ACA7, 0xD025D076, 0x50865013, 0xDC56DCF7,
252 0x8455841A, 0xE109E151, 0x7ABE7A25, 0x139113EF},
253
254 {0xD939A9D9, 0x90176790, 0x719CB371, 0xD2A6E8D2, 0x05070405, 0x9852FD98,
255 0x6580A365, 0xDFE476DF, 0x08459A08, 0x024B9202, 0xA0E080A0, 0x665A7866,
256 0xDDAFE4DD, 0xB06ADDB0, 0xBF63D1BF, 0x362A3836, 0x54E60D54, 0x4320C643,
257 0x62CC3562, 0xBEF298BE, 0x1E12181E, 0x24EBF724, 0xD7A1ECD7, 0x77416C77,
258 0xBD2843BD, 0x32BC7532, 0xD47B37D4, 0x9B88269B, 0x700DFA70, 0xF94413F9,
259 0xB1FB94B1, 0x5A7E485A, 0x7A03F27A, 0xE48CD0E4, 0x47B68B47, 0x3C24303C,
260 0xA5E784A5, 0x416B5441, 0x06DDDF06, 0xC56023C5, 0x45FD1945, 0xA33A5BA3,
261 0x68C23D68, 0x158D5915, 0x21ECF321, 0x3166AE31, 0x3E6FA23E, 0x16578216,
262 0x95106395, 0x5BEF015B, 0x4DB8834D, 0x91862E91, 0xB56DD9B5, 0x1F83511F,
263 0x53AA9B53, 0x635D7C63, 0x3B68A63B, 0x3FFEEB3F, 0xD630A5D6, 0x257ABE25,
264 0xA7AC16A7, 0x0F090C0F, 0x35F0E335, 0x23A76123, 0xF090C0F0, 0xAFE98CAF,
265 0x809D3A80, 0x925CF592, 0x810C7381, 0x27312C27, 0x76D02576, 0xE7560BE7,
266 0x7B92BB7B, 0xE9CE4EE9, 0xF10189F1, 0x9F1E6B9F, 0xA93453A9, 0xC4F16AC4,
267 0x99C3B499, 0x975BF197, 0x8347E183, 0x6B18E66B, 0xC822BDC8, 0x0E98450E,
268 0x6E1FE26E, 0xC9B3F4C9, 0x2F74B62F, 0xCBF866CB, 0xFF99CCFF, 0xEA1495EA,
269 0xED5803ED, 0xF7DC56F7, 0xE18BD4E1, 0x1B151C1B, 0xADA21EAD, 0x0CD3D70C,
270 0x2BE2FB2B, 0x1DC8C31D, 0x195E8E19, 0xC22CB5C2, 0x8949E989, 0x12C1CF12,
271 0x7E95BF7E, 0x207DBA20, 0x6411EA64, 0x840B7784, 0x6DC5396D, 0x6A89AF6A,
272 0xD17C33D1, 0xA171C9A1, 0xCEFF62CE, 0x37BB7137, 0xFB0F81FB, 0x3DB5793D,
273 0x51E10951, 0xDC3EADDC, 0x2D3F242D, 0xA476CDA4, 0x9D55F99D, 0xEE82D8EE,
274 0x8640E586, 0xAE78C5AE, 0xCD25B9CD, 0x04964D04, 0x55774455, 0x0A0E080A,
275 0x13508613, 0x30F7E730, 0xD337A1D3, 0x40FA1D40, 0x3461AA34, 0x8C4EED8C,
276 0xB3B006B3, 0x6C54706C, 0x2A73B22A, 0x523BD252, 0x0B9F410B, 0x8B027B8B,
277 0x88D8A088, 0x4FF3114F, 0x67CB3167, 0x4627C246, 0xC06727C0, 0xB4FC90B4,
278 0x28382028, 0x7F04F67F, 0x78486078, 0x2EE5FF2E, 0x074C9607, 0x4B655C4B,
279 0xC72BB1C7, 0x6F8EAB6F, 0x0D429E0D, 0xBBF59CBB, 0xF2DB52F2, 0xF34A1BF3,
280 0xA63D5FA6, 0x59A49359, 0xBCB90ABC, 0x3AF9EF3A, 0xEF1391EF, 0xFE0885FE,
281 0x01914901, 0x6116EE61, 0x7CDE2D7C, 0xB2214FB2, 0x42B18F42, 0xDB723BDB,
282 0xB82F47B8, 0x48BF8748, 0x2CAE6D2C, 0xE3C046E3, 0x573CD657, 0x859A3E85,
283 0x29A96929, 0x7D4F647D, 0x94812A94, 0x492ECE49, 0x17C6CB17, 0xCA692FCA,
284 0xC3BDFCC3, 0x5CA3975C, 0x5EE8055E, 0xD0ED7AD0, 0x87D1AC87, 0x8E057F8E,
285 0xBA64D5BA, 0xA8A51AA8, 0xB7264BB7, 0xB9BE0EB9, 0x6087A760, 0xF8D55AF8,
286 0x22362822, 0x111B1411, 0xDE753FDE, 0x79D92979, 0xAAEE88AA, 0x332D3C33,
287 0x5F794C5F, 0xB6B702B6, 0x96CAB896, 0x5835DA58, 0x9CC4B09C, 0xFC4317FC,
288 0x1A84551A, 0xF64D1FF6, 0x1C598A1C, 0x38B27D38, 0xAC3357AC, 0x18CFC718,
289 0xF4068DF4, 0x69537469, 0x749BB774, 0xF597C4F5, 0x56AD9F56, 0xDAE372DA,
290 0xD5EA7ED5, 0x4AF4154A, 0x9E8F229E, 0xA2AB12A2, 0x4E62584E, 0xE85F07E8,
291 0xE51D99E5, 0x39233439, 0xC1F66EC1, 0x446C5044, 0x5D32DE5D, 0x72466872,
292 0x26A06526, 0x93CDBC93, 0x03DADB03, 0xC6BAF8C6, 0xFA9EC8FA, 0x82D6A882,
293 0xCF6E2BCF, 0x50704050, 0xEB85DCEB, 0x750AFE75, 0x8A93328A, 0x8DDFA48D,
294 0x4C29CA4C, 0x141C1014, 0x73D72173, 0xCCB4F0CC, 0x09D4D309, 0x108A5D10,
295 0xE2510FE2, 0x00000000, 0x9A196F9A, 0xE01A9DE0, 0x8F94368F, 0xE6C742E6,
296 0xECC94AEC, 0xFDD25EFD, 0xAB7FC1AB, 0xD8A8E0D8}
297};
298
299/* The exp_to_poly and poly_to_exp tables are used to perform efficient
300 * operations in GF(2^8) represented as GF(2)[x]/w(x) where
301 * w(x)=x^8+x^6+x^3+x^2+1. We care about doing that because it's part of the
302 * definition of the RS matrix in the key schedule. Elements of that field
303 * are polynomials of degree not greater than 7 and all coefficients 0 or 1,
304 * which can be represented naturally by bytes (just substitute x=2). In that
305 * form, GF(2^8) addition is the same as bitwise XOR, but GF(2^8)
306 * multiplication is inefficient without hardware support. To multiply
307 * faster, I make use of the fact x is a generator for the nonzero elements,
308 * so that every element p of GF(2)[x]/w(x) is either 0 or equal to (x)^n for
309 * some n in 0..254. Note that that caret is exponentiation in GF(2^8),
310 * *not* polynomial notation. So if I want to compute pq where p and q are
311 * in GF(2^8), I can just say:
312 * 1. if p=0 or q=0 then pq=0
313 * 2. otherwise, find m and n such that p=x^m and q=x^n
314 * 3. pq=(x^m)(x^n)=x^(m+n), so add m and n and find pq
315 * The translations in steps 2 and 3 are looked up in the tables
316 * poly_to_exp (for step 2) and exp_to_poly (for step 3). To see this
317 * in action, look at the CALC_S macro. As additional wrinkles, note that
318 * one of my operands is always a constant, so the poly_to_exp lookup on it
319 * is done in advance; I included the original values in the comments so
320 * readers can have some chance of recognizing that this *is* the RS matrix
321 * from the Twofish paper. I've only included the table entries I actually
322 * need; I never do a lookup on a variable input of zero and the biggest
323 * exponents I'll ever see are 254 (variable) and 237 (constant), so they'll
324 * never sum to more than 491. I'm repeating part of the exp_to_poly table
325 * so that I don't have to do mod-255 reduction in the exponent arithmetic.
326 * Since I know my constant operands are never zero, I only have to worry
327 * about zero values in the variable operand, and I do it with a simple
328 * conditional branch. I know conditionals are expensive, but I couldn't
329 * see a non-horrible way of avoiding them, and I did manage to group the
330 * statements so that each if covers four group multiplications. */
331
332static const u8 poly_to_exp[255] = {
333 0x00, 0x01, 0x17, 0x02, 0x2E, 0x18, 0x53, 0x03, 0x6A, 0x2F, 0x93, 0x19,
334 0x34, 0x54, 0x45, 0x04, 0x5C, 0x6B, 0xB6, 0x30, 0xA6, 0x94, 0x4B, 0x1A,
335 0x8C, 0x35, 0x81, 0x55, 0xAA, 0x46, 0x0D, 0x05, 0x24, 0x5D, 0x87, 0x6C,
336 0x9B, 0xB7, 0xC1, 0x31, 0x2B, 0xA7, 0xA3, 0x95, 0x98, 0x4C, 0xCA, 0x1B,
337 0xE6, 0x8D, 0x73, 0x36, 0xCD, 0x82, 0x12, 0x56, 0x62, 0xAB, 0xF0, 0x47,
338 0x4F, 0x0E, 0xBD, 0x06, 0xD4, 0x25, 0xD2, 0x5E, 0x27, 0x88, 0x66, 0x6D,
339 0xD6, 0x9C, 0x79, 0xB8, 0x08, 0xC2, 0xDF, 0x32, 0x68, 0x2C, 0xFD, 0xA8,
340 0x8A, 0xA4, 0x5A, 0x96, 0x29, 0x99, 0x22, 0x4D, 0x60, 0xCB, 0xE4, 0x1C,
341 0x7B, 0xE7, 0x3B, 0x8E, 0x9E, 0x74, 0xF4, 0x37, 0xD8, 0xCE, 0xF9, 0x83,
342 0x6F, 0x13, 0xB2, 0x57, 0xE1, 0x63, 0xDC, 0xAC, 0xC4, 0xF1, 0xAF, 0x48,
343 0x0A, 0x50, 0x42, 0x0F, 0xBA, 0xBE, 0xC7, 0x07, 0xDE, 0xD5, 0x78, 0x26,
344 0x65, 0xD3, 0xD1, 0x5F, 0xE3, 0x28, 0x21, 0x89, 0x59, 0x67, 0xFC, 0x6E,
345 0xB1, 0xD7, 0xF8, 0x9D, 0xF3, 0x7A, 0x3A, 0xB9, 0xC6, 0x09, 0x41, 0xC3,
346 0xAE, 0xE0, 0xDB, 0x33, 0x44, 0x69, 0x92, 0x2D, 0x52, 0xFE, 0x16, 0xA9,
347 0x0C, 0x8B, 0x80, 0xA5, 0x4A, 0x5B, 0xB5, 0x97, 0xC9, 0x2A, 0xA2, 0x9A,
348 0xC0, 0x23, 0x86, 0x4E, 0xBC, 0x61, 0xEF, 0xCC, 0x11, 0xE5, 0x72, 0x1D,
349 0x3D, 0x7C, 0xEB, 0xE8, 0xE9, 0x3C, 0xEA, 0x8F, 0x7D, 0x9F, 0xEC, 0x75,
350 0x1E, 0xF5, 0x3E, 0x38, 0xF6, 0xD9, 0x3F, 0xCF, 0x76, 0xFA, 0x1F, 0x84,
351 0xA0, 0x70, 0xED, 0x14, 0x90, 0xB3, 0x7E, 0x58, 0xFB, 0xE2, 0x20, 0x64,
352 0xD0, 0xDD, 0x77, 0xAD, 0xDA, 0xC5, 0x40, 0xF2, 0x39, 0xB0, 0xF7, 0x49,
353 0xB4, 0x0B, 0x7F, 0x51, 0x15, 0x43, 0x91, 0x10, 0x71, 0xBB, 0xEE, 0xBF,
354 0x85, 0xC8, 0xA1
355};
356
357static const u8 exp_to_poly[492] = {
358 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 0x4D, 0x9A, 0x79, 0xF2,
359 0xA9, 0x1F, 0x3E, 0x7C, 0xF8, 0xBD, 0x37, 0x6E, 0xDC, 0xF5, 0xA7, 0x03,
360 0x06, 0x0C, 0x18, 0x30, 0x60, 0xC0, 0xCD, 0xD7, 0xE3, 0x8B, 0x5B, 0xB6,
361 0x21, 0x42, 0x84, 0x45, 0x8A, 0x59, 0xB2, 0x29, 0x52, 0xA4, 0x05, 0x0A,
362 0x14, 0x28, 0x50, 0xA0, 0x0D, 0x1A, 0x34, 0x68, 0xD0, 0xED, 0x97, 0x63,
363 0xC6, 0xC1, 0xCF, 0xD3, 0xEB, 0x9B, 0x7B, 0xF6, 0xA1, 0x0F, 0x1E, 0x3C,
364 0x78, 0xF0, 0xAD, 0x17, 0x2E, 0x5C, 0xB8, 0x3D, 0x7A, 0xF4, 0xA5, 0x07,
365 0x0E, 0x1C, 0x38, 0x70, 0xE0, 0x8D, 0x57, 0xAE, 0x11, 0x22, 0x44, 0x88,
366 0x5D, 0xBA, 0x39, 0x72, 0xE4, 0x85, 0x47, 0x8E, 0x51, 0xA2, 0x09, 0x12,
367 0x24, 0x48, 0x90, 0x6D, 0xDA, 0xF9, 0xBF, 0x33, 0x66, 0xCC, 0xD5, 0xE7,
368 0x83, 0x4B, 0x96, 0x61, 0xC2, 0xC9, 0xDF, 0xF3, 0xAB, 0x1B, 0x36, 0x6C,
369 0xD8, 0xFD, 0xB7, 0x23, 0x46, 0x8C, 0x55, 0xAA, 0x19, 0x32, 0x64, 0xC8,
370 0xDD, 0xF7, 0xA3, 0x0B, 0x16, 0x2C, 0x58, 0xB0, 0x2D, 0x5A, 0xB4, 0x25,
371 0x4A, 0x94, 0x65, 0xCA, 0xD9, 0xFF, 0xB3, 0x2B, 0x56, 0xAC, 0x15, 0x2A,
372 0x54, 0xA8, 0x1D, 0x3A, 0x74, 0xE8, 0x9D, 0x77, 0xEE, 0x91, 0x6F, 0xDE,
373 0xF1, 0xAF, 0x13, 0x26, 0x4C, 0x98, 0x7D, 0xFA, 0xB9, 0x3F, 0x7E, 0xFC,
374 0xB5, 0x27, 0x4E, 0x9C, 0x75, 0xEA, 0x99, 0x7F, 0xFE, 0xB1, 0x2F, 0x5E,
375 0xBC, 0x35, 0x6A, 0xD4, 0xE5, 0x87, 0x43, 0x86, 0x41, 0x82, 0x49, 0x92,
376 0x69, 0xD2, 0xE9, 0x9F, 0x73, 0xE6, 0x81, 0x4F, 0x9E, 0x71, 0xE2, 0x89,
377 0x5F, 0xBE, 0x31, 0x62, 0xC4, 0xC5, 0xC7, 0xC3, 0xCB, 0xDB, 0xFB, 0xBB,
378 0x3B, 0x76, 0xEC, 0x95, 0x67, 0xCE, 0xD1, 0xEF, 0x93, 0x6B, 0xD6, 0xE1,
379 0x8F, 0x53, 0xA6, 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 0x4D,
380 0x9A, 0x79, 0xF2, 0xA9, 0x1F, 0x3E, 0x7C, 0xF8, 0xBD, 0x37, 0x6E, 0xDC,
381 0xF5, 0xA7, 0x03, 0x06, 0x0C, 0x18, 0x30, 0x60, 0xC0, 0xCD, 0xD7, 0xE3,
382 0x8B, 0x5B, 0xB6, 0x21, 0x42, 0x84, 0x45, 0x8A, 0x59, 0xB2, 0x29, 0x52,
383 0xA4, 0x05, 0x0A, 0x14, 0x28, 0x50, 0xA0, 0x0D, 0x1A, 0x34, 0x68, 0xD0,
384 0xED, 0x97, 0x63, 0xC6, 0xC1, 0xCF, 0xD3, 0xEB, 0x9B, 0x7B, 0xF6, 0xA1,
385 0x0F, 0x1E, 0x3C, 0x78, 0xF0, 0xAD, 0x17, 0x2E, 0x5C, 0xB8, 0x3D, 0x7A,
386 0xF4, 0xA5, 0x07, 0x0E, 0x1C, 0x38, 0x70, 0xE0, 0x8D, 0x57, 0xAE, 0x11,
387 0x22, 0x44, 0x88, 0x5D, 0xBA, 0x39, 0x72, 0xE4, 0x85, 0x47, 0x8E, 0x51,
388 0xA2, 0x09, 0x12, 0x24, 0x48, 0x90, 0x6D, 0xDA, 0xF9, 0xBF, 0x33, 0x66,
389 0xCC, 0xD5, 0xE7, 0x83, 0x4B, 0x96, 0x61, 0xC2, 0xC9, 0xDF, 0xF3, 0xAB,
390 0x1B, 0x36, 0x6C, 0xD8, 0xFD, 0xB7, 0x23, 0x46, 0x8C, 0x55, 0xAA, 0x19,
391 0x32, 0x64, 0xC8, 0xDD, 0xF7, 0xA3, 0x0B, 0x16, 0x2C, 0x58, 0xB0, 0x2D,
392 0x5A, 0xB4, 0x25, 0x4A, 0x94, 0x65, 0xCA, 0xD9, 0xFF, 0xB3, 0x2B, 0x56,
393 0xAC, 0x15, 0x2A, 0x54, 0xA8, 0x1D, 0x3A, 0x74, 0xE8, 0x9D, 0x77, 0xEE,
394 0x91, 0x6F, 0xDE, 0xF1, 0xAF, 0x13, 0x26, 0x4C, 0x98, 0x7D, 0xFA, 0xB9,
395 0x3F, 0x7E, 0xFC, 0xB5, 0x27, 0x4E, 0x9C, 0x75, 0xEA, 0x99, 0x7F, 0xFE,
396 0xB1, 0x2F, 0x5E, 0xBC, 0x35, 0x6A, 0xD4, 0xE5, 0x87, 0x43, 0x86, 0x41,
397 0x82, 0x49, 0x92, 0x69, 0xD2, 0xE9, 0x9F, 0x73, 0xE6, 0x81, 0x4F, 0x9E,
398 0x71, 0xE2, 0x89, 0x5F, 0xBE, 0x31, 0x62, 0xC4, 0xC5, 0xC7, 0xC3, 0xCB
399};
400
401
402/* The table constants are indices of
403 * S-box entries, preprocessed through q0 and q1. */
404static const u8 calc_sb_tbl[512] = {
405 0xA9, 0x75, 0x67, 0xF3, 0xB3, 0xC6, 0xE8, 0xF4,
406 0x04, 0xDB, 0xFD, 0x7B, 0xA3, 0xFB, 0x76, 0xC8,
407 0x9A, 0x4A, 0x92, 0xD3, 0x80, 0xE6, 0x78, 0x6B,
408 0xE4, 0x45, 0xDD, 0x7D, 0xD1, 0xE8, 0x38, 0x4B,
409 0x0D, 0xD6, 0xC6, 0x32, 0x35, 0xD8, 0x98, 0xFD,
410 0x18, 0x37, 0xF7, 0x71, 0xEC, 0xF1, 0x6C, 0xE1,
411 0x43, 0x30, 0x75, 0x0F, 0x37, 0xF8, 0x26, 0x1B,
412 0xFA, 0x87, 0x13, 0xFA, 0x94, 0x06, 0x48, 0x3F,
413 0xF2, 0x5E, 0xD0, 0xBA, 0x8B, 0xAE, 0x30, 0x5B,
414 0x84, 0x8A, 0x54, 0x00, 0xDF, 0xBC, 0x23, 0x9D,
415 0x19, 0x6D, 0x5B, 0xC1, 0x3D, 0xB1, 0x59, 0x0E,
416 0xF3, 0x80, 0xAE, 0x5D, 0xA2, 0xD2, 0x82, 0xD5,
417 0x63, 0xA0, 0x01, 0x84, 0x83, 0x07, 0x2E, 0x14,
418 0xD9, 0xB5, 0x51, 0x90, 0x9B, 0x2C, 0x7C, 0xA3,
419 0xA6, 0xB2, 0xEB, 0x73, 0xA5, 0x4C, 0xBE, 0x54,
420 0x16, 0x92, 0x0C, 0x74, 0xE3, 0x36, 0x61, 0x51,
421 0xC0, 0x38, 0x8C, 0xB0, 0x3A, 0xBD, 0xF5, 0x5A,
422 0x73, 0xFC, 0x2C, 0x60, 0x25, 0x62, 0x0B, 0x96,
423 0xBB, 0x6C, 0x4E, 0x42, 0x89, 0xF7, 0x6B, 0x10,
424 0x53, 0x7C, 0x6A, 0x28, 0xB4, 0x27, 0xF1, 0x8C,
425 0xE1, 0x13, 0xE6, 0x95, 0xBD, 0x9C, 0x45, 0xC7,
426 0xE2, 0x24, 0xF4, 0x46, 0xB6, 0x3B, 0x66, 0x70,
427 0xCC, 0xCA, 0x95, 0xE3, 0x03, 0x85, 0x56, 0xCB,
428 0xD4, 0x11, 0x1C, 0xD0, 0x1E, 0x93, 0xD7, 0xB8,
429 0xFB, 0xA6, 0xC3, 0x83, 0x8E, 0x20, 0xB5, 0xFF,
430 0xE9, 0x9F, 0xCF, 0x77, 0xBF, 0xC3, 0xBA, 0xCC,
431 0xEA, 0x03, 0x77, 0x6F, 0x39, 0x08, 0xAF, 0xBF,
432 0x33, 0x40, 0xC9, 0xE7, 0x62, 0x2B, 0x71, 0xE2,
433 0x81, 0x79, 0x79, 0x0C, 0x09, 0xAA, 0xAD, 0x82,
434 0x24, 0x41, 0xCD, 0x3A, 0xF9, 0xEA, 0xD8, 0xB9,
435 0xE5, 0xE4, 0xC5, 0x9A, 0xB9, 0xA4, 0x4D, 0x97,
436 0x44, 0x7E, 0x08, 0xDA, 0x86, 0x7A, 0xE7, 0x17,
437 0xA1, 0x66, 0x1D, 0x94, 0xAA, 0xA1, 0xED, 0x1D,
438 0x06, 0x3D, 0x70, 0xF0, 0xB2, 0xDE, 0xD2, 0xB3,
439 0x41, 0x0B, 0x7B, 0x72, 0xA0, 0xA7, 0x11, 0x1C,
440 0x31, 0xEF, 0xC2, 0xD1, 0x27, 0x53, 0x90, 0x3E,
441 0x20, 0x8F, 0xF6, 0x33, 0x60, 0x26, 0xFF, 0x5F,
442 0x96, 0xEC, 0x5C, 0x76, 0xB1, 0x2A, 0xAB, 0x49,
443 0x9E, 0x81, 0x9C, 0x88, 0x52, 0xEE, 0x1B, 0x21,
444 0x5F, 0xC4, 0x93, 0x1A, 0x0A, 0xEB, 0xEF, 0xD9,
445 0x91, 0xC5, 0x85, 0x39, 0x49, 0x99, 0xEE, 0xCD,
446 0x2D, 0xAD, 0x4F, 0x31, 0x8F, 0x8B, 0x3B, 0x01,
447 0x47, 0x18, 0x87, 0x23, 0x6D, 0xDD, 0x46, 0x1F,
448 0xD6, 0x4E, 0x3E, 0x2D, 0x69, 0xF9, 0x64, 0x48,
449 0x2A, 0x4F, 0xCE, 0xF2, 0xCB, 0x65, 0x2F, 0x8E,
450 0xFC, 0x78, 0x97, 0x5C, 0x05, 0x58, 0x7A, 0x19,
451 0xAC, 0x8D, 0x7F, 0xE5, 0xD5, 0x98, 0x1A, 0x57,
452 0x4B, 0x67, 0x0E, 0x7F, 0xA7, 0x05, 0x5A, 0x64,
453 0x28, 0xAF, 0x14, 0x63, 0x3F, 0xB6, 0x29, 0xFE,
454 0x88, 0xF5, 0x3C, 0xB7, 0x4C, 0x3C, 0x02, 0xA5,
455 0xB8, 0xCE, 0xDA, 0xE9, 0xB0, 0x68, 0x17, 0x44,
456 0x55, 0xE0, 0x1F, 0x4D, 0x8A, 0x43, 0x7D, 0x69,
457 0x57, 0x29, 0xC7, 0x2E, 0x8D, 0xAC, 0x74, 0x15,
458 0xB7, 0x59, 0xC4, 0xA8, 0x9F, 0x0A, 0x72, 0x9E,
459 0x7E, 0x6E, 0x15, 0x47, 0x22, 0xDF, 0x12, 0x34,
460 0x58, 0x35, 0x07, 0x6A, 0x99, 0xCF, 0x34, 0xDC,
461 0x6E, 0x22, 0x50, 0xC9, 0xDE, 0xC0, 0x68, 0x9B,
462 0x65, 0x89, 0xBC, 0xD4, 0xDB, 0xED, 0xF8, 0xAB,
463 0xC8, 0x12, 0xA8, 0xA2, 0x2B, 0x0D, 0x40, 0x52,
464 0xDC, 0xBB, 0xFE, 0x02, 0x32, 0x2F, 0xA4, 0xA9,
465 0xCA, 0xD7, 0x10, 0x61, 0x21, 0x1E, 0xF0, 0xB4,
466 0xD3, 0x50, 0x5D, 0x04, 0x0F, 0xF6, 0x00, 0xC2,
467 0x6F, 0x16, 0x9D, 0x25, 0x36, 0x86, 0x42, 0x56,
468 0x4A, 0x55, 0x5E, 0x09, 0xC1, 0xBE, 0xE0, 0x91
469};
470
471/* Macro to perform one column of the RS matrix multiplication. The
472 * parameters a, b, c, and d are the four bytes of output; i is the index
473 * of the key bytes, and w, x, y, and z, are the column of constants from
474 * the RS matrix, preprocessed through the poly_to_exp table. */
475
476#define CALC_S(a, b, c, d, i, w, x, y, z) \
477 if (key[i]) { \
478 tmp = poly_to_exp[key[i] - 1]; \
479 (a) ^= exp_to_poly[tmp + (w)]; \
480 (b) ^= exp_to_poly[tmp + (x)]; \
481 (c) ^= exp_to_poly[tmp + (y)]; \
482 (d) ^= exp_to_poly[tmp + (z)]; \
483 }
484
485/* Macros to calculate the key-dependent S-boxes for a 128-bit key using
486 * the S vector from CALC_S. CALC_SB_2 computes a single entry in all
487 * four S-boxes, where i is the index of the entry to compute, and a and b
488 * are the index numbers preprocessed through the q0 and q1 tables
489 * respectively. */
490
491#define CALC_SB_2(i, a, b) \
492 ctx->s[0][i] = mds[0][q0[(a) ^ sa] ^ se]; \
493 ctx->s[1][i] = mds[1][q0[(b) ^ sb] ^ sf]; \
494 ctx->s[2][i] = mds[2][q1[(a) ^ sc] ^ sg]; \
495 ctx->s[3][i] = mds[3][q1[(b) ^ sd] ^ sh]
496
497/* Macro exactly like CALC_SB_2, but for 192-bit keys. */
498
499#define CALC_SB192_2(i, a, b) \
500 ctx->s[0][i] = mds[0][q0[q0[(b) ^ sa] ^ se] ^ si]; \
501 ctx->s[1][i] = mds[1][q0[q1[(b) ^ sb] ^ sf] ^ sj]; \
502 ctx->s[2][i] = mds[2][q1[q0[(a) ^ sc] ^ sg] ^ sk]; \
503 ctx->s[3][i] = mds[3][q1[q1[(a) ^ sd] ^ sh] ^ sl];
504
505/* Macro exactly like CALC_SB_2, but for 256-bit keys. */
506
507#define CALC_SB256_2(i, a, b) \
508 ctx->s[0][i] = mds[0][q0[q0[q1[(b) ^ sa] ^ se] ^ si] ^ sm]; \
509 ctx->s[1][i] = mds[1][q0[q1[q1[(a) ^ sb] ^ sf] ^ sj] ^ sn]; \
510 ctx->s[2][i] = mds[2][q1[q0[q0[(a) ^ sc] ^ sg] ^ sk] ^ so]; \
511 ctx->s[3][i] = mds[3][q1[q1[q0[(b) ^ sd] ^ sh] ^ sl] ^ sp];
512
513/* Macros to calculate the whitening and round subkeys. CALC_K_2 computes the
514 * last two stages of the h() function for a given index (either 2i or 2i+1).
515 * a, b, c, and d are the four bytes going into the last two stages. For
516 * 128-bit keys, this is the entire h() function and a and c are the index
517 * preprocessed through q0 and q1 respectively; for longer keys they are the
518 * output of previous stages. j is the index of the first key byte to use.
519 * CALC_K computes a pair of subkeys for 128-bit Twofish, by calling CALC_K_2
520 * twice, doing the Pseudo-Hadamard Transform, and doing the necessary
521 * rotations. Its parameters are: a, the array to write the results into,
522 * j, the index of the first output entry, k and l, the preprocessed indices
523 * for index 2i, and m and n, the preprocessed indices for index 2i+1.
524 * CALC_K192_2 expands CALC_K_2 to handle 192-bit keys, by doing an
525 * additional lookup-and-XOR stage. The parameters a, b, c and d are the
526 * four bytes going into the last three stages. For 192-bit keys, c = d
527 * are the index preprocessed through q0, and a = b are the index
528 * preprocessed through q1; j is the index of the first key byte to use.
529 * CALC_K192 is identical to CALC_K but for using the CALC_K192_2 macro
530 * instead of CALC_K_2.
531 * CALC_K256_2 expands CALC_K192_2 to handle 256-bit keys, by doing an
532 * additional lookup-and-XOR stage. The parameters a and b are the index
533 * preprocessed through q0 and q1 respectively; j is the index of the first
534 * key byte to use. CALC_K256 is identical to CALC_K but for using the
535 * CALC_K256_2 macro instead of CALC_K_2. */
536
537#define CALC_K_2(a, b, c, d, j) \
538 mds[0][q0[a ^ key[(j) + 8]] ^ key[j]] \
539 ^ mds[1][q0[b ^ key[(j) + 9]] ^ key[(j) + 1]] \
540 ^ mds[2][q1[c ^ key[(j) + 10]] ^ key[(j) + 2]] \
541 ^ mds[3][q1[d ^ key[(j) + 11]] ^ key[(j) + 3]]
542
543#define CALC_K(a, j, k, l, m, n) \
544 x = CALC_K_2 (k, l, k, l, 0); \
545 y = CALC_K_2 (m, n, m, n, 4); \
Denis Vlasenkoa5f8c472006-01-16 17:42:28 +1100546 y = rol32(y, 8); \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 x += y; y += x; ctx->a[j] = x; \
Denis Vlasenkoa5f8c472006-01-16 17:42:28 +1100548 ctx->a[(j) + 1] = rol32(y, 9)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549
550#define CALC_K192_2(a, b, c, d, j) \
551 CALC_K_2 (q0[a ^ key[(j) + 16]], \
552 q1[b ^ key[(j) + 17]], \
553 q0[c ^ key[(j) + 18]], \
554 q1[d ^ key[(j) + 19]], j)
555
556#define CALC_K192(a, j, k, l, m, n) \
557 x = CALC_K192_2 (l, l, k, k, 0); \
558 y = CALC_K192_2 (n, n, m, m, 4); \
Denis Vlasenkoa5f8c472006-01-16 17:42:28 +1100559 y = rol32(y, 8); \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 x += y; y += x; ctx->a[j] = x; \
Denis Vlasenkoa5f8c472006-01-16 17:42:28 +1100561 ctx->a[(j) + 1] = rol32(y, 9)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562
563#define CALC_K256_2(a, b, j) \
564 CALC_K192_2 (q1[b ^ key[(j) + 24]], \
565 q1[a ^ key[(j) + 25]], \
566 q0[a ^ key[(j) + 26]], \
567 q0[b ^ key[(j) + 27]], j)
568
569#define CALC_K256(a, j, k, l, m, n) \
570 x = CALC_K256_2 (k, l, 0); \
571 y = CALC_K256_2 (m, n, 4); \
Denis Vlasenkoa5f8c472006-01-16 17:42:28 +1100572 y = rol32(y, 8); \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 x += y; y += x; ctx->a[j] = x; \
Denis Vlasenkoa5f8c472006-01-16 17:42:28 +1100574 ctx->a[(j) + 1] = rol32(y, 9)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575
576
577/* Macros to compute the g() function in the encryption and decryption
578 * rounds. G1 is the straight g() function; G2 includes the 8-bit
579 * rotation for the high 32-bit word. */
580
581#define G1(a) \
582 (ctx->s[0][(a) & 0xFF]) ^ (ctx->s[1][((a) >> 8) & 0xFF]) \
583 ^ (ctx->s[2][((a) >> 16) & 0xFF]) ^ (ctx->s[3][(a) >> 24])
584
585#define G2(b) \
586 (ctx->s[1][(b) & 0xFF]) ^ (ctx->s[2][((b) >> 8) & 0xFF]) \
587 ^ (ctx->s[3][((b) >> 16) & 0xFF]) ^ (ctx->s[0][(b) >> 24])
588
589/* Encryption and decryption Feistel rounds. Each one calls the two g()
590 * macros, does the PHT, and performs the XOR and the appropriate bit
591 * rotations. The parameters are the round number (used to select subkeys),
592 * and the four 32-bit chunks of the text. */
593
594#define ENCROUND(n, a, b, c, d) \
595 x = G1 (a); y = G2 (b); \
596 x += y; y += x + ctx->k[2 * (n) + 1]; \
597 (c) ^= x + ctx->k[2 * (n)]; \
Denis Vlasenkoa5f8c472006-01-16 17:42:28 +1100598 (c) = ror32((c), 1); \
599 (d) = rol32((d), 1) ^ y
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601#define DECROUND(n, a, b, c, d) \
602 x = G1 (a); y = G2 (b); \
603 x += y; y += x; \
604 (d) ^= y + ctx->k[2 * (n) + 1]; \
Denis Vlasenkoa5f8c472006-01-16 17:42:28 +1100605 (d) = ror32((d), 1); \
606 (c) = rol32((c), 1); \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 (c) ^= (x + ctx->k[2 * (n)])
608
609/* Encryption and decryption cycles; each one is simply two Feistel rounds
610 * with the 32-bit chunks re-ordered to simulate the "swap" */
611
612#define ENCCYCLE(n) \
613 ENCROUND (2 * (n), a, b, c, d); \
614 ENCROUND (2 * (n) + 1, c, d, a, b)
615
616#define DECCYCLE(n) \
617 DECROUND (2 * (n) + 1, c, d, a, b); \
618 DECROUND (2 * (n), a, b, c, d)
619
620/* Macros to convert the input and output bytes into 32-bit words,
621 * and simultaneously perform the whitening step. INPACK packs word
622 * number n into the variable named by x, using whitening subkey number m.
623 * OUTUNPACK unpacks word number n from the variable named by x, using
624 * whitening subkey number m. */
625
626#define INPACK(n, x, m) \
Herbert Xu06ace7a2005-10-30 21:25:15 +1100627 x = le32_to_cpu(src[n]) ^ ctx->w[m]
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
629#define OUTUNPACK(n, x, m) \
630 x ^= ctx->w[m]; \
Herbert Xu06ace7a2005-10-30 21:25:15 +1100631 dst[n] = cpu_to_le32(x)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632
633#define TF_MIN_KEY_SIZE 16
634#define TF_MAX_KEY_SIZE 32
635#define TF_BLOCK_SIZE 16
636
637/* Structure for an expanded Twofish key. s contains the key-dependent
638 * S-boxes composed with the MDS matrix; w contains the eight "whitening"
639 * subkeys, K[0] through K[7]. k holds the remaining, "round" subkeys. Note
640 * that k[i] corresponds to what the Twofish paper calls K[i+8]. */
641struct twofish_ctx {
642 u32 s[4][256], w[8], k[32];
643};
644
645/* Perform the key setup. */
646static int twofish_setkey(void *cx, const u8 *key,
647 unsigned int key_len, u32 *flags)
648{
649
650 struct twofish_ctx *ctx = cx;
651
652 int i, j, k;
653
654 /* Temporaries for CALC_K. */
655 u32 x, y;
656
657 /* The S vector used to key the S-boxes, split up into individual bytes.
658 * 128-bit keys use only sa through sh; 256-bit use all of them. */
659 u8 sa = 0, sb = 0, sc = 0, sd = 0, se = 0, sf = 0, sg = 0, sh = 0;
660 u8 si = 0, sj = 0, sk = 0, sl = 0, sm = 0, sn = 0, so = 0, sp = 0;
661
662 /* Temporary for CALC_S. */
663 u8 tmp;
664
665 /* Check key length. */
666 if (key_len != 16 && key_len != 24 && key_len != 32)
667 {
668 *flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
669 return -EINVAL; /* unsupported key length */
670 }
671
672 /* Compute the first two words of the S vector. The magic numbers are
673 * the entries of the RS matrix, preprocessed through poly_to_exp. The
674 * numbers in the comments are the original (polynomial form) matrix
675 * entries. */
676 CALC_S (sa, sb, sc, sd, 0, 0x00, 0x2D, 0x01, 0x2D); /* 01 A4 02 A4 */
677 CALC_S (sa, sb, sc, sd, 1, 0x2D, 0xA4, 0x44, 0x8A); /* A4 56 A1 55 */
678 CALC_S (sa, sb, sc, sd, 2, 0x8A, 0xD5, 0xBF, 0xD1); /* 55 82 FC 87 */
679 CALC_S (sa, sb, sc, sd, 3, 0xD1, 0x7F, 0x3D, 0x99); /* 87 F3 C1 5A */
680 CALC_S (sa, sb, sc, sd, 4, 0x99, 0x46, 0x66, 0x96); /* 5A 1E 47 58 */
681 CALC_S (sa, sb, sc, sd, 5, 0x96, 0x3C, 0x5B, 0xED); /* 58 C6 AE DB */
682 CALC_S (sa, sb, sc, sd, 6, 0xED, 0x37, 0x4F, 0xE0); /* DB 68 3D 9E */
683 CALC_S (sa, sb, sc, sd, 7, 0xE0, 0xD0, 0x8C, 0x17); /* 9E E5 19 03 */
684 CALC_S (se, sf, sg, sh, 8, 0x00, 0x2D, 0x01, 0x2D); /* 01 A4 02 A4 */
685 CALC_S (se, sf, sg, sh, 9, 0x2D, 0xA4, 0x44, 0x8A); /* A4 56 A1 55 */
686 CALC_S (se, sf, sg, sh, 10, 0x8A, 0xD5, 0xBF, 0xD1); /* 55 82 FC 87 */
687 CALC_S (se, sf, sg, sh, 11, 0xD1, 0x7F, 0x3D, 0x99); /* 87 F3 C1 5A */
688 CALC_S (se, sf, sg, sh, 12, 0x99, 0x46, 0x66, 0x96); /* 5A 1E 47 58 */
689 CALC_S (se, sf, sg, sh, 13, 0x96, 0x3C, 0x5B, 0xED); /* 58 C6 AE DB */
690 CALC_S (se, sf, sg, sh, 14, 0xED, 0x37, 0x4F, 0xE0); /* DB 68 3D 9E */
691 CALC_S (se, sf, sg, sh, 15, 0xE0, 0xD0, 0x8C, 0x17); /* 9E E5 19 03 */
692
693 if (key_len == 24 || key_len == 32) { /* 192- or 256-bit key */
694 /* Calculate the third word of the S vector */
695 CALC_S (si, sj, sk, sl, 16, 0x00, 0x2D, 0x01, 0x2D); /* 01 A4 02 A4 */
696 CALC_S (si, sj, sk, sl, 17, 0x2D, 0xA4, 0x44, 0x8A); /* A4 56 A1 55 */
697 CALC_S (si, sj, sk, sl, 18, 0x8A, 0xD5, 0xBF, 0xD1); /* 55 82 FC 87 */
698 CALC_S (si, sj, sk, sl, 19, 0xD1, 0x7F, 0x3D, 0x99); /* 87 F3 C1 5A */
699 CALC_S (si, sj, sk, sl, 20, 0x99, 0x46, 0x66, 0x96); /* 5A 1E 47 58 */
700 CALC_S (si, sj, sk, sl, 21, 0x96, 0x3C, 0x5B, 0xED); /* 58 C6 AE DB */
701 CALC_S (si, sj, sk, sl, 22, 0xED, 0x37, 0x4F, 0xE0); /* DB 68 3D 9E */
702 CALC_S (si, sj, sk, sl, 23, 0xE0, 0xD0, 0x8C, 0x17); /* 9E E5 19 03 */
703 }
704
705 if (key_len == 32) { /* 256-bit key */
706 /* Calculate the fourth word of the S vector */
707 CALC_S (sm, sn, so, sp, 24, 0x00, 0x2D, 0x01, 0x2D); /* 01 A4 02 A4 */
708 CALC_S (sm, sn, so, sp, 25, 0x2D, 0xA4, 0x44, 0x8A); /* A4 56 A1 55 */
709 CALC_S (sm, sn, so, sp, 26, 0x8A, 0xD5, 0xBF, 0xD1); /* 55 82 FC 87 */
710 CALC_S (sm, sn, so, sp, 27, 0xD1, 0x7F, 0x3D, 0x99); /* 87 F3 C1 5A */
711 CALC_S (sm, sn, so, sp, 28, 0x99, 0x46, 0x66, 0x96); /* 5A 1E 47 58 */
712 CALC_S (sm, sn, so, sp, 29, 0x96, 0x3C, 0x5B, 0xED); /* 58 C6 AE DB */
713 CALC_S (sm, sn, so, sp, 30, 0xED, 0x37, 0x4F, 0xE0); /* DB 68 3D 9E */
714 CALC_S (sm, sn, so, sp, 31, 0xE0, 0xD0, 0x8C, 0x17); /* 9E E5 19 03 */
715
716 /* Compute the S-boxes. */
717 for ( i = j = 0, k = 1; i < 256; i++, j += 2, k += 2 ) {
718 CALC_SB256_2( i, calc_sb_tbl[j], calc_sb_tbl[k] );
719 }
720
721 /* Calculate whitening and round subkeys. The constants are
722 * indices of subkeys, preprocessed through q0 and q1. */
723 CALC_K256 (w, 0, 0xA9, 0x75, 0x67, 0xF3);
724 CALC_K256 (w, 2, 0xB3, 0xC6, 0xE8, 0xF4);
725 CALC_K256 (w, 4, 0x04, 0xDB, 0xFD, 0x7B);
726 CALC_K256 (w, 6, 0xA3, 0xFB, 0x76, 0xC8);
727 CALC_K256 (k, 0, 0x9A, 0x4A, 0x92, 0xD3);
728 CALC_K256 (k, 2, 0x80, 0xE6, 0x78, 0x6B);
729 CALC_K256 (k, 4, 0xE4, 0x45, 0xDD, 0x7D);
730 CALC_K256 (k, 6, 0xD1, 0xE8, 0x38, 0x4B);
731 CALC_K256 (k, 8, 0x0D, 0xD6, 0xC6, 0x32);
732 CALC_K256 (k, 10, 0x35, 0xD8, 0x98, 0xFD);
733 CALC_K256 (k, 12, 0x18, 0x37, 0xF7, 0x71);
734 CALC_K256 (k, 14, 0xEC, 0xF1, 0x6C, 0xE1);
735 CALC_K256 (k, 16, 0x43, 0x30, 0x75, 0x0F);
736 CALC_K256 (k, 18, 0x37, 0xF8, 0x26, 0x1B);
737 CALC_K256 (k, 20, 0xFA, 0x87, 0x13, 0xFA);
738 CALC_K256 (k, 22, 0x94, 0x06, 0x48, 0x3F);
739 CALC_K256 (k, 24, 0xF2, 0x5E, 0xD0, 0xBA);
740 CALC_K256 (k, 26, 0x8B, 0xAE, 0x30, 0x5B);
741 CALC_K256 (k, 28, 0x84, 0x8A, 0x54, 0x00);
742 CALC_K256 (k, 30, 0xDF, 0xBC, 0x23, 0x9D);
743 } else if (key_len == 24) { /* 192-bit key */
744 /* Compute the S-boxes. */
745 for ( i = j = 0, k = 1; i < 256; i++, j += 2, k += 2 ) {
746 CALC_SB192_2( i, calc_sb_tbl[j], calc_sb_tbl[k] );
747 }
748
749 /* Calculate whitening and round subkeys. The constants are
750 * indices of subkeys, preprocessed through q0 and q1. */
751 CALC_K192 (w, 0, 0xA9, 0x75, 0x67, 0xF3);
752 CALC_K192 (w, 2, 0xB3, 0xC6, 0xE8, 0xF4);
753 CALC_K192 (w, 4, 0x04, 0xDB, 0xFD, 0x7B);
754 CALC_K192 (w, 6, 0xA3, 0xFB, 0x76, 0xC8);
755 CALC_K192 (k, 0, 0x9A, 0x4A, 0x92, 0xD3);
756 CALC_K192 (k, 2, 0x80, 0xE6, 0x78, 0x6B);
757 CALC_K192 (k, 4, 0xE4, 0x45, 0xDD, 0x7D);
758 CALC_K192 (k, 6, 0xD1, 0xE8, 0x38, 0x4B);
759 CALC_K192 (k, 8, 0x0D, 0xD6, 0xC6, 0x32);
760 CALC_K192 (k, 10, 0x35, 0xD8, 0x98, 0xFD);
761 CALC_K192 (k, 12, 0x18, 0x37, 0xF7, 0x71);
762 CALC_K192 (k, 14, 0xEC, 0xF1, 0x6C, 0xE1);
763 CALC_K192 (k, 16, 0x43, 0x30, 0x75, 0x0F);
764 CALC_K192 (k, 18, 0x37, 0xF8, 0x26, 0x1B);
765 CALC_K192 (k, 20, 0xFA, 0x87, 0x13, 0xFA);
766 CALC_K192 (k, 22, 0x94, 0x06, 0x48, 0x3F);
767 CALC_K192 (k, 24, 0xF2, 0x5E, 0xD0, 0xBA);
768 CALC_K192 (k, 26, 0x8B, 0xAE, 0x30, 0x5B);
769 CALC_K192 (k, 28, 0x84, 0x8A, 0x54, 0x00);
770 CALC_K192 (k, 30, 0xDF, 0xBC, 0x23, 0x9D);
771 } else { /* 128-bit key */
772 /* Compute the S-boxes. */
773 for ( i = j = 0, k = 1; i < 256; i++, j += 2, k += 2 ) {
774 CALC_SB_2( i, calc_sb_tbl[j], calc_sb_tbl[k] );
775 }
776
777 /* Calculate whitening and round subkeys. The constants are
778 * indices of subkeys, preprocessed through q0 and q1. */
779 CALC_K (w, 0, 0xA9, 0x75, 0x67, 0xF3);
780 CALC_K (w, 2, 0xB3, 0xC6, 0xE8, 0xF4);
781 CALC_K (w, 4, 0x04, 0xDB, 0xFD, 0x7B);
782 CALC_K (w, 6, 0xA3, 0xFB, 0x76, 0xC8);
783 CALC_K (k, 0, 0x9A, 0x4A, 0x92, 0xD3);
784 CALC_K (k, 2, 0x80, 0xE6, 0x78, 0x6B);
785 CALC_K (k, 4, 0xE4, 0x45, 0xDD, 0x7D);
786 CALC_K (k, 6, 0xD1, 0xE8, 0x38, 0x4B);
787 CALC_K (k, 8, 0x0D, 0xD6, 0xC6, 0x32);
788 CALC_K (k, 10, 0x35, 0xD8, 0x98, 0xFD);
789 CALC_K (k, 12, 0x18, 0x37, 0xF7, 0x71);
790 CALC_K (k, 14, 0xEC, 0xF1, 0x6C, 0xE1);
791 CALC_K (k, 16, 0x43, 0x30, 0x75, 0x0F);
792 CALC_K (k, 18, 0x37, 0xF8, 0x26, 0x1B);
793 CALC_K (k, 20, 0xFA, 0x87, 0x13, 0xFA);
794 CALC_K (k, 22, 0x94, 0x06, 0x48, 0x3F);
795 CALC_K (k, 24, 0xF2, 0x5E, 0xD0, 0xBA);
796 CALC_K (k, 26, 0x8B, 0xAE, 0x30, 0x5B);
797 CALC_K (k, 28, 0x84, 0x8A, 0x54, 0x00);
798 CALC_K (k, 30, 0xDF, 0xBC, 0x23, 0x9D);
799 }
800
801 return 0;
802}
803
804/* Encrypt one block. in and out may be the same. */
805static void twofish_encrypt(void *cx, u8 *out, const u8 *in)
806{
807 struct twofish_ctx *ctx = cx;
Herbert Xu06ace7a2005-10-30 21:25:15 +1100808 const __le32 *src = (const __le32 *)in;
809 __le32 *dst = (__le32 *)out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
811 /* The four 32-bit chunks of the text. */
812 u32 a, b, c, d;
813
814 /* Temporaries used by the round function. */
815 u32 x, y;
816
817 /* Input whitening and packing. */
818 INPACK (0, a, 0);
819 INPACK (1, b, 1);
820 INPACK (2, c, 2);
821 INPACK (3, d, 3);
822
823 /* Encryption Feistel cycles. */
824 ENCCYCLE (0);
825 ENCCYCLE (1);
826 ENCCYCLE (2);
827 ENCCYCLE (3);
828 ENCCYCLE (4);
829 ENCCYCLE (5);
830 ENCCYCLE (6);
831 ENCCYCLE (7);
832
833 /* Output whitening and unpacking. */
834 OUTUNPACK (0, c, 4);
835 OUTUNPACK (1, d, 5);
836 OUTUNPACK (2, a, 6);
837 OUTUNPACK (3, b, 7);
838
839}
840
841/* Decrypt one block. in and out may be the same. */
842static void twofish_decrypt(void *cx, u8 *out, const u8 *in)
843{
844 struct twofish_ctx *ctx = cx;
Herbert Xu06ace7a2005-10-30 21:25:15 +1100845 const __le32 *src = (const __le32 *)in;
846 __le32 *dst = (__le32 *)out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847
848 /* The four 32-bit chunks of the text. */
849 u32 a, b, c, d;
850
851 /* Temporaries used by the round function. */
852 u32 x, y;
853
854 /* Input whitening and packing. */
855 INPACK (0, c, 4);
856 INPACK (1, d, 5);
857 INPACK (2, a, 6);
858 INPACK (3, b, 7);
859
860 /* Encryption Feistel cycles. */
861 DECCYCLE (7);
862 DECCYCLE (6);
863 DECCYCLE (5);
864 DECCYCLE (4);
865 DECCYCLE (3);
866 DECCYCLE (2);
867 DECCYCLE (1);
868 DECCYCLE (0);
869
870 /* Output whitening and unpacking. */
871 OUTUNPACK (0, a, 0);
872 OUTUNPACK (1, b, 1);
873 OUTUNPACK (2, c, 2);
874 OUTUNPACK (3, d, 3);
875
876}
877
878static struct crypto_alg alg = {
879 .cra_name = "twofish",
880 .cra_flags = CRYPTO_ALG_TYPE_CIPHER,
881 .cra_blocksize = TF_BLOCK_SIZE,
882 .cra_ctxsize = sizeof(struct twofish_ctx),
Herbert Xua429d262006-01-07 16:38:15 +1100883 .cra_alignmask = 3,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 .cra_module = THIS_MODULE,
885 .cra_list = LIST_HEAD_INIT(alg.cra_list),
886 .cra_u = { .cipher = {
887 .cia_min_keysize = TF_MIN_KEY_SIZE,
888 .cia_max_keysize = TF_MAX_KEY_SIZE,
889 .cia_setkey = twofish_setkey,
890 .cia_encrypt = twofish_encrypt,
891 .cia_decrypt = twofish_decrypt } }
892};
893
894static int __init init(void)
895{
896 return crypto_register_alg(&alg);
897}
898
899static void __exit fini(void)
900{
901 crypto_unregister_alg(&alg);
902}
903
904module_init(init);
905module_exit(fini);
906
907MODULE_LICENSE("GPL");
908MODULE_DESCRIPTION ("Twofish Cipher Algorithm");