2 ---------------------------------------------------------------------------
3 Copyright (c) 2003, Dr Brian Gladman <brg@gladman.me.uk>, Worcester, UK.
8 The free distribution and use of this software in both source and binary
9 form is allowed (with or without changes) provided that:
11 1. distributions of this source code include the above copyright
12 notice, this list of conditions and the following disclaimer;
14 2. distributions in binary form include the above copyright
15 notice, this list of conditions and the following disclaimer
16 in the documentation and/or other associated materials;
18 3. the copyright holder's name is not used to endorse products
19 built using this software without specific written permission.
21 ALTERNATIVELY, provided that this notice is retained in full, this product
22 may be distributed under the terms of the GNU General Public License (GPL),
23 in which case the provisions of the GPL apply INSTEAD OF those given above.
27 This software is provided 'as is' with no explicit or implied warranties
28 in respect of its properties, including, but not limited to, correctness
29 and/or fitness for purpose.
30 ---------------------------------------------------------------------------
31 Issue Date: 26/08/2003
37 * \brief This file contains the code for implementing encryption and decryption
38 * for AES (Rijndael) for block and key sizes of 16, 24 and 32 bytes. It
39 * can optionally be replaced by code written in assembler using NASM. For
40 * further details see the file aesopt.h
42 * \author Dr Brian Gladman <brg@gladman.me.uk>
47 #if defined(__cplusplus)
52 #define si(y,x,k,c) (s(y,c) = word_in(x, c) ^ (k)[c])
53 #define so(y,x,c) word_out(y, c, s(x,c))
56 #define locals(y,x) x[4],y[4]
58 #define locals(y,x) x##0,x##1,x##2,x##3,y##0,y##1,y##2,y##3
61 #define l_copy(y, x) s(y,0) = s(x,0); s(y,1) = s(x,1); \
62 s(y,2) = s(x,2); s(y,3) = s(x,3);
63 #define state_in(y,x,k) si(y,x,k,0); si(y,x,k,1); si(y,x,k,2); si(y,x,k,3)
64 #define state_out(y,x) so(y,x,0); so(y,x,1); so(y,x,2); so(y,x,3)
65 #define round(rm,y,x,k) rm(y,x,k,0); rm(y,x,k,1); rm(y,x,k,2); rm(y,x,k,3)
67 #if defined(ENCRYPTION) && !defined(AES_ASM)
69 /* Visual C++ .Net v7.1 provides the fastest encryption code when using
70 Pentium optimiation with small code but this is poor for decryption
71 so we need to control this with the following VC++ pragmas
75 #pragma optimize( "s", on )
78 /* Given the column (c) of the output state variable, the following
79 macros give the input state variables which are needed in its
80 computation for each row (r) of the state. All the alternative
81 macros give the same end values but expand into different ways
82 of calculating these values. In particular the complex macro
83 used for dynamically variable block sizes is designed to expand
84 to a compile time constant whenever possible but will expand to
85 conditional clauses on some branches (I am grateful to Frank
86 Yellin for this construction)
89 #define fwd_var(x,r,c)\
90 ( r == 0 ? ( c == 0 ? s(x,0) : c == 1 ? s(x,1) : c == 2 ? s(x,2) : s(x,3))\
91 : r == 1 ? ( c == 0 ? s(x,1) : c == 1 ? s(x,2) : c == 2 ? s(x,3) : s(x,0))\
92 : r == 2 ? ( c == 0 ? s(x,2) : c == 1 ? s(x,3) : c == 2 ? s(x,0) : s(x,1))\
93 : ( c == 0 ? s(x,3) : c == 1 ? s(x,0) : c == 2 ? s(x,1) : s(x,2)))
97 #define fwd_rnd(y,x,k,c) (s(y,c) = (k)[c] ^ four_tables(x,t_use(f,n),fwd_var,rf1,c))
98 #elif defined(FT1_SET)
100 #define fwd_rnd(y,x,k,c) (s(y,c) = (k)[c] ^ one_table(x,upr,t_use(f,n),fwd_var,rf1,c))
102 #define fwd_rnd(y,x,k,c) (s(y,c) = (k)[c] ^ fwd_mcol(no_table(x,t_use(s,box),fwd_var,rf1,c)))
106 #define fwd_lrnd(y,x,k,c) (s(y,c) = (k)[c] ^ four_tables(x,t_use(f,l),fwd_var,rf1,c))
107 #elif defined(FL1_SET)
108 #define fwd_lrnd(y,x,k,c) (s(y,c) = (k)[c] ^ one_table(x,ups,t_use(f,l),fwd_var,rf1,c))
110 #define fwd_lrnd(y,x,k,c) (s(y,c) = (k)[c] ^ no_table(x,t_use(s,box),fwd_var,rf1,c))
113 aes_rval
aes_encrypt(const void *in_blk
, void *out_blk
, const aes_encrypt_ctx cx
[1])
114 { aes_32t
locals(b0
, b1
);
115 const aes_32t
*kp
= cx
->ks
;
117 dec_fmvars
; /* declare variables for fwd_mcol() if needed */
120 aes_32t nr
= (kp
[45] ^ kp
[52] ^ kp
[53] ? kp
[52] : 14);
123 if( (nr
!= 10 || !(kp
[0] | kp
[3] | kp
[4]))
124 && (nr
!= 12 || !(kp
[0] | kp
[5] | kp
[6]))
125 && (nr
!= 14 || !(kp
[0] | kp
[7] | kp
[8])) )
129 state_in(b0
, in_blk
, kp
);
131 #if (ENC_UNROLL == FULL)
136 round(fwd_rnd
, b1
, b0
, kp
+ 1 * N_COLS
);
137 round(fwd_rnd
, b0
, b1
, kp
+ 2 * N_COLS
);
140 round(fwd_rnd
, b1
, b0
, kp
+ 1 * N_COLS
);
141 round(fwd_rnd
, b0
, b1
, kp
+ 2 * N_COLS
);
144 round(fwd_rnd
, b1
, b0
, kp
+ 1 * N_COLS
);
145 round(fwd_rnd
, b0
, b1
, kp
+ 2 * N_COLS
);
146 round(fwd_rnd
, b1
, b0
, kp
+ 3 * N_COLS
);
147 round(fwd_rnd
, b0
, b1
, kp
+ 4 * N_COLS
);
148 round(fwd_rnd
, b1
, b0
, kp
+ 5 * N_COLS
);
149 round(fwd_rnd
, b0
, b1
, kp
+ 6 * N_COLS
);
150 round(fwd_rnd
, b1
, b0
, kp
+ 7 * N_COLS
);
151 round(fwd_rnd
, b0
, b1
, kp
+ 8 * N_COLS
);
152 round(fwd_rnd
, b1
, b0
, kp
+ 9 * N_COLS
);
153 round(fwd_lrnd
, b0
, b1
, kp
+10 * N_COLS
);
158 #if (ENC_UNROLL == PARTIAL)
160 for(rnd
= 0; rnd
< (nr
>> 1) - 1; ++rnd
)
163 round(fwd_rnd
, b1
, b0
, kp
);
165 round(fwd_rnd
, b0
, b1
, kp
);
168 round(fwd_rnd
, b1
, b0
, kp
);
171 for(rnd
= 0; rnd
< nr
- 1; ++rnd
)
174 round(fwd_rnd
, b1
, b0
, kp
);
179 round(fwd_lrnd
, b0
, b1
, kp
);
183 state_out(out_blk
, b0
);
191 #if defined(DECRYPTION) && !defined(AES_ASM)
193 /* Visual C++ .Net v7.1 provides the fastest encryption code when using
194 Pentium optimiation with small code but this is poor for decryption
195 so we need to control this with the following VC++ pragmas
198 #if defined(_MSC_VER)
199 #pragma optimize( "t", on )
202 /* Given the column (c) of the output state variable, the following
203 macros give the input state variables which are needed in its
204 computation for each row (r) of the state. All the alternative
205 macros give the same end values but expand into different ways
206 of calculating these values. In particular the complex macro
207 used for dynamically variable block sizes is designed to expand
208 to a compile time constant whenever possible but will expand to
209 conditional clauses on some branches (I am grateful to Frank
210 Yellin for this construction)
213 #define inv_var(x,r,c)\
214 ( r == 0 ? ( c == 0 ? s(x,0) : c == 1 ? s(x,1) : c == 2 ? s(x,2) : s(x,3))\
215 : r == 1 ? ( c == 0 ? s(x,3) : c == 1 ? s(x,0) : c == 2 ? s(x,1) : s(x,2))\
216 : r == 2 ? ( c == 0 ? s(x,2) : c == 1 ? s(x,3) : c == 2 ? s(x,0) : s(x,1))\
217 : ( c == 0 ? s(x,1) : c == 1 ? s(x,2) : c == 2 ? s(x,3) : s(x,0)))
221 #define inv_rnd(y,x,k,c) (s(y,c) = (k)[c] ^ four_tables(x,t_use(i,n),inv_var,rf1,c))
222 #elif defined(IT1_SET)
224 #define inv_rnd(y,x,k,c) (s(y,c) = (k)[c] ^ one_table(x,upr,t_use(i,n),inv_var,rf1,c))
226 #define inv_rnd(y,x,k,c) (s(y,c) = inv_mcol((k)[c] ^ no_table(x,t_use(i,box),inv_var,rf1,c)))
230 #define inv_lrnd(y,x,k,c) (s(y,c) = (k)[c] ^ four_tables(x,t_use(i,l),inv_var,rf1,c))
231 #elif defined(IL1_SET)
232 #define inv_lrnd(y,x,k,c) (s(y,c) = (k)[c] ^ one_table(x,ups,t_use(i,l),inv_var,rf1,c))
234 #define inv_lrnd(y,x,k,c) (s(y,c) = (k)[c] ^ no_table(x,t_use(i,box),inv_var,rf1,c))
237 aes_rval
aes_decrypt(const void *in_blk
, void *out_blk
, const aes_decrypt_ctx cx
[1])
238 { aes_32t
locals(b0
, b1
);
240 dec_imvars
; /* declare variables for inv_mcol() if needed */
243 aes_32t nr
= (cx
->ks
[45] ^ cx
->ks
[52] ^ cx
->ks
[53] ? cx
->ks
[52] : 14);
244 const aes_32t
*kp
= cx
->ks
+ nr
* N_COLS
;
247 if( (nr
!= 10 || !(cx
->ks
[0] | cx
->ks
[3] | cx
->ks
[4]))
248 && (nr
!= 12 || !(cx
->ks
[0] | cx
->ks
[5] | cx
->ks
[6]))
249 && (nr
!= 14 || !(cx
->ks
[0] | cx
->ks
[7] | cx
->ks
[8])) )
253 state_in(b0
, in_blk
, kp
);
255 #if (DEC_UNROLL == FULL)
260 round(inv_rnd
, b1
, b0
, kp
- 1 * N_COLS
);
261 round(inv_rnd
, b0
, b1
, kp
- 2 * N_COLS
);
264 round(inv_rnd
, b1
, b0
, kp
- 1 * N_COLS
);
265 round(inv_rnd
, b0
, b1
, kp
- 2 * N_COLS
);
268 round(inv_rnd
, b1
, b0
, kp
- 1 * N_COLS
);
269 round(inv_rnd
, b0
, b1
, kp
- 2 * N_COLS
);
270 round(inv_rnd
, b1
, b0
, kp
- 3 * N_COLS
);
271 round(inv_rnd
, b0
, b1
, kp
- 4 * N_COLS
);
272 round(inv_rnd
, b1
, b0
, kp
- 5 * N_COLS
);
273 round(inv_rnd
, b0
, b1
, kp
- 6 * N_COLS
);
274 round(inv_rnd
, b1
, b0
, kp
- 7 * N_COLS
);
275 round(inv_rnd
, b0
, b1
, kp
- 8 * N_COLS
);
276 round(inv_rnd
, b1
, b0
, kp
- 9 * N_COLS
);
277 round(inv_lrnd
, b0
, b1
, kp
- 10 * N_COLS
);
282 #if (DEC_UNROLL == PARTIAL)
284 for(rnd
= 0; rnd
< (nr
>> 1) - 1; ++rnd
)
287 round(inv_rnd
, b1
, b0
, kp
);
289 round(inv_rnd
, b0
, b1
, kp
);
292 round(inv_rnd
, b1
, b0
, kp
);
295 for(rnd
= 0; rnd
< nr
- 1; ++rnd
)
298 round(inv_rnd
, b1
, b0
, kp
);
303 round(inv_lrnd
, b0
, b1
, kp
);
307 state_out(out_blk
, b0
);
315 #if defined(__cplusplus)