diff options
author | Jussi Kivilinna <jussi.kivilinna@mbnet.fi> | 2012-06-18 07:07:39 -0400 |
---|---|---|
committer | Herbert Xu <herbert@gondor.apana.org.au> | 2012-06-27 02:42:02 -0400 |
commit | a7378d4e552ac139ae1cbbdfebfeaa9b18c948d0 (patch) | |
tree | dc8960d5ce50ac27f05e02151c7744bb313286b4 /arch/x86/crypto/twofish_avx_glue.c | |
parent | 414cb5e7cc6e258fe36e2c3cc3ef1ff2e246c0e3 (diff) |
crypto: twofish-avx - remove duplicated glue code and use shared glue code from glue_helper
Now that shared glue code is available, convert twofish-avx to use it.
Cc: Johannes Goetzfried <Johannes.Goetzfried@informatik.stud.uni-erlangen.de>
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Diffstat (limited to 'arch/x86/crypto/twofish_avx_glue.c')
-rw-r--r-- | arch/x86/crypto/twofish_avx_glue.c | 556 |
1 files changed, 99 insertions, 457 deletions
diff --git a/arch/x86/crypto/twofish_avx_glue.c b/arch/x86/crypto/twofish_avx_glue.c index cabe058eba14..782b67ddaf6a 100644 --- a/arch/x86/crypto/twofish_avx_glue.c +++ b/arch/x86/crypto/twofish_avx_glue.c | |||
@@ -4,9 +4,6 @@ | |||
4 | * Copyright (C) 2012 Johannes Goetzfried | 4 | * Copyright (C) 2012 Johannes Goetzfried |
5 | * <Johannes.Goetzfried@informatik.stud.uni-erlangen.de> | 5 | * <Johannes.Goetzfried@informatik.stud.uni-erlangen.de> |
6 | * | 6 | * |
7 | * Glue code based on serpent_sse2_glue.c by: | ||
8 | * Copyright (C) 2011 Jussi Kivilinna <jussi.kivilinna@mbnet.fi> | ||
9 | * | ||
10 | * This program is free software; you can redistribute it and/or modify | 7 | * This program is free software; you can redistribute it and/or modify |
11 | * it under the terms of the GNU General Public License as published by | 8 | * it under the terms of the GNU General Public License as published by |
12 | * the Free Software Foundation; either version 2 of the License, or | 9 | * the Free Software Foundation; either version 2 of the License, or |
@@ -39,38 +36,21 @@ | |||
39 | #include <asm/i387.h> | 36 | #include <asm/i387.h> |
40 | #include <asm/xcr.h> | 37 | #include <asm/xcr.h> |
41 | #include <asm/xsave.h> | 38 | #include <asm/xsave.h> |
39 | #include <asm/crypto/twofish.h> | ||
42 | #include <asm/crypto/ablk_helper.h> | 40 | #include <asm/crypto/ablk_helper.h> |
41 | #include <asm/crypto/glue_helper.h> | ||
43 | #include <crypto/scatterwalk.h> | 42 | #include <crypto/scatterwalk.h> |
44 | #include <linux/workqueue.h> | 43 | #include <linux/workqueue.h> |
45 | #include <linux/spinlock.h> | 44 | #include <linux/spinlock.h> |
46 | 45 | ||
47 | |||
48 | #define TWOFISH_PARALLEL_BLOCKS 8 | 46 | #define TWOFISH_PARALLEL_BLOCKS 8 |
49 | 47 | ||
50 | /* regular block cipher functions from twofish_x86_64 module */ | ||
51 | asmlinkage void twofish_enc_blk(struct twofish_ctx *ctx, u8 *dst, | ||
52 | const u8 *src); | ||
53 | asmlinkage void twofish_dec_blk(struct twofish_ctx *ctx, u8 *dst, | ||
54 | const u8 *src); | ||
55 | |||
56 | /* 3-way parallel cipher functions from twofish_x86_64-3way module */ | ||
57 | asmlinkage void __twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst, | ||
58 | const u8 *src, bool xor); | ||
59 | asmlinkage void twofish_dec_blk_3way(struct twofish_ctx *ctx, u8 *dst, | ||
60 | const u8 *src); | ||
61 | |||
62 | static inline void twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst, | 48 | static inline void twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst, |
63 | const u8 *src) | 49 | const u8 *src) |
64 | { | 50 | { |
65 | __twofish_enc_blk_3way(ctx, dst, src, false); | 51 | __twofish_enc_blk_3way(ctx, dst, src, false); |
66 | } | 52 | } |
67 | 53 | ||
68 | static inline void twofish_enc_blk_3way_xor(struct twofish_ctx *ctx, u8 *dst, | ||
69 | const u8 *src) | ||
70 | { | ||
71 | __twofish_enc_blk_3way(ctx, dst, src, true); | ||
72 | } | ||
73 | |||
74 | /* 8-way parallel cipher functions */ | 54 | /* 8-way parallel cipher functions */ |
75 | asmlinkage void __twofish_enc_blk_8way(struct twofish_ctx *ctx, u8 *dst, | 55 | asmlinkage void __twofish_enc_blk_8way(struct twofish_ctx *ctx, u8 *dst, |
76 | const u8 *src, bool xor); | 56 | const u8 *src, bool xor); |
@@ -95,423 +75,142 @@ static inline void twofish_dec_blk_xway(struct twofish_ctx *ctx, u8 *dst, | |||
95 | twofish_dec_blk_8way(ctx, dst, src); | 75 | twofish_dec_blk_8way(ctx, dst, src); |
96 | } | 76 | } |
97 | 77 | ||
98 | 78 | static void twofish_dec_blk_cbc_xway(void *ctx, u128 *dst, const u128 *src) | |
99 | static inline bool twofish_fpu_begin(bool fpu_enabled, unsigned int nbytes) | ||
100 | { | 79 | { |
101 | if (fpu_enabled) | 80 | u128 ivs[TWOFISH_PARALLEL_BLOCKS - 1]; |
102 | return true; | 81 | unsigned int j; |
103 | 82 | ||
104 | /* AVX is only used when chunk to be processed is large enough, so | 83 | for (j = 0; j < TWOFISH_PARALLEL_BLOCKS - 1; j++) |
105 | * do not enable FPU until it is necessary. | 84 | ivs[j] = src[j]; |
106 | */ | ||
107 | if (nbytes < TF_BLOCK_SIZE * TWOFISH_PARALLEL_BLOCKS) | ||
108 | return false; | ||
109 | 85 | ||
110 | kernel_fpu_begin(); | 86 | twofish_dec_blk_xway(ctx, (u8 *)dst, (u8 *)src); |
111 | return true; | ||
112 | } | ||
113 | 87 | ||
114 | static inline void twofish_fpu_end(bool fpu_enabled) | 88 | for (j = 0; j < TWOFISH_PARALLEL_BLOCKS - 1; j++) |
115 | { | 89 | u128_xor(dst + (j + 1), dst + (j + 1), ivs + j); |
116 | if (fpu_enabled) | ||
117 | kernel_fpu_end(); | ||
118 | } | 90 | } |
119 | 91 | ||
120 | static int ecb_crypt(struct blkcipher_desc *desc, struct blkcipher_walk *walk, | 92 | static void twofish_enc_blk_ctr_xway(void *ctx, u128 *dst, const u128 *src, |
121 | bool enc) | 93 | u128 *iv) |
122 | { | 94 | { |
123 | bool fpu_enabled = false; | 95 | be128 ctrblks[TWOFISH_PARALLEL_BLOCKS]; |
124 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | 96 | unsigned int i; |
125 | const unsigned int bsize = TF_BLOCK_SIZE; | ||
126 | unsigned int nbytes; | ||
127 | int err; | ||
128 | 97 | ||
129 | err = blkcipher_walk_virt(desc, walk); | 98 | for (i = 0; i < TWOFISH_PARALLEL_BLOCKS; i++) { |
130 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | 99 | if (dst != src) |
100 | dst[i] = src[i]; | ||
131 | 101 | ||
132 | while ((nbytes = walk->nbytes)) { | 102 | u128_to_be128(&ctrblks[i], iv); |
133 | u8 *wsrc = walk->src.virt.addr; | 103 | u128_inc(iv); |
134 | u8 *wdst = walk->dst.virt.addr; | ||
135 | |||
136 | fpu_enabled = twofish_fpu_begin(fpu_enabled, nbytes); | ||
137 | |||
138 | /* Process multi-block batch */ | ||
139 | if (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS) { | ||
140 | do { | ||
141 | if (enc) | ||
142 | twofish_enc_blk_xway(ctx, wdst, wsrc); | ||
143 | else | ||
144 | twofish_dec_blk_xway(ctx, wdst, wsrc); | ||
145 | |||
146 | wsrc += bsize * TWOFISH_PARALLEL_BLOCKS; | ||
147 | wdst += bsize * TWOFISH_PARALLEL_BLOCKS; | ||
148 | nbytes -= bsize * TWOFISH_PARALLEL_BLOCKS; | ||
149 | } while (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS); | ||
150 | |||
151 | if (nbytes < bsize) | ||
152 | goto done; | ||
153 | } | ||
154 | |||
155 | /* Process three block batch */ | ||
156 | if (nbytes >= bsize * 3) { | ||
157 | do { | ||
158 | if (enc) | ||
159 | twofish_enc_blk_3way(ctx, wdst, wsrc); | ||
160 | else | ||
161 | twofish_dec_blk_3way(ctx, wdst, wsrc); | ||
162 | |||
163 | wsrc += bsize * 3; | ||
164 | wdst += bsize * 3; | ||
165 | nbytes -= bsize * 3; | ||
166 | } while (nbytes >= bsize * 3); | ||
167 | |||
168 | if (nbytes < bsize) | ||
169 | goto done; | ||
170 | } | ||
171 | |||
172 | /* Handle leftovers */ | ||
173 | do { | ||
174 | if (enc) | ||
175 | twofish_enc_blk(ctx, wdst, wsrc); | ||
176 | else | ||
177 | twofish_dec_blk(ctx, wdst, wsrc); | ||
178 | |||
179 | wsrc += bsize; | ||
180 | wdst += bsize; | ||
181 | nbytes -= bsize; | ||
182 | } while (nbytes >= bsize); | ||
183 | |||
184 | done: | ||
185 | err = blkcipher_walk_done(desc, walk, nbytes); | ||
186 | } | 104 | } |
187 | 105 | ||
188 | twofish_fpu_end(fpu_enabled); | 106 | twofish_enc_blk_xway_xor(ctx, (u8 *)dst, (u8 *)ctrblks); |
189 | return err; | ||
190 | } | 107 | } |
191 | 108 | ||
109 | static const struct common_glue_ctx twofish_enc = { | ||
110 | .num_funcs = 3, | ||
111 | .fpu_blocks_limit = TWOFISH_PARALLEL_BLOCKS, | ||
112 | |||
113 | .funcs = { { | ||
114 | .num_blocks = TWOFISH_PARALLEL_BLOCKS, | ||
115 | .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_enc_blk_xway) } | ||
116 | }, { | ||
117 | .num_blocks = 3, | ||
118 | .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_enc_blk_3way) } | ||
119 | }, { | ||
120 | .num_blocks = 1, | ||
121 | .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_enc_blk) } | ||
122 | } } | ||
123 | }; | ||
124 | |||
125 | static const struct common_glue_ctx twofish_ctr = { | ||
126 | .num_funcs = 3, | ||
127 | .fpu_blocks_limit = TWOFISH_PARALLEL_BLOCKS, | ||
128 | |||
129 | .funcs = { { | ||
130 | .num_blocks = TWOFISH_PARALLEL_BLOCKS, | ||
131 | .fn_u = { .ctr = GLUE_CTR_FUNC_CAST(twofish_enc_blk_ctr_xway) } | ||
132 | }, { | ||
133 | .num_blocks = 3, | ||
134 | .fn_u = { .ctr = GLUE_CTR_FUNC_CAST(twofish_enc_blk_ctr_3way) } | ||
135 | }, { | ||
136 | .num_blocks = 1, | ||
137 | .fn_u = { .ctr = GLUE_CTR_FUNC_CAST(twofish_enc_blk_ctr) } | ||
138 | } } | ||
139 | }; | ||
140 | |||
141 | static const struct common_glue_ctx twofish_dec = { | ||
142 | .num_funcs = 3, | ||
143 | .fpu_blocks_limit = TWOFISH_PARALLEL_BLOCKS, | ||
144 | |||
145 | .funcs = { { | ||
146 | .num_blocks = TWOFISH_PARALLEL_BLOCKS, | ||
147 | .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_dec_blk_xway) } | ||
148 | }, { | ||
149 | .num_blocks = 3, | ||
150 | .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_dec_blk_3way) } | ||
151 | }, { | ||
152 | .num_blocks = 1, | ||
153 | .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_dec_blk) } | ||
154 | } } | ||
155 | }; | ||
156 | |||
157 | static const struct common_glue_ctx twofish_dec_cbc = { | ||
158 | .num_funcs = 3, | ||
159 | .fpu_blocks_limit = TWOFISH_PARALLEL_BLOCKS, | ||
160 | |||
161 | .funcs = { { | ||
162 | .num_blocks = TWOFISH_PARALLEL_BLOCKS, | ||
163 | .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(twofish_dec_blk_cbc_xway) } | ||
164 | }, { | ||
165 | .num_blocks = 3, | ||
166 | .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(twofish_dec_blk_cbc_3way) } | ||
167 | }, { | ||
168 | .num_blocks = 1, | ||
169 | .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(twofish_dec_blk) } | ||
170 | } } | ||
171 | }; | ||
172 | |||
192 | static int ecb_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | 173 | static int ecb_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, |
193 | struct scatterlist *src, unsigned int nbytes) | 174 | struct scatterlist *src, unsigned int nbytes) |
194 | { | 175 | { |
195 | struct blkcipher_walk walk; | 176 | return glue_ecb_crypt_128bit(&twofish_enc, desc, dst, src, nbytes); |
196 | |||
197 | blkcipher_walk_init(&walk, dst, src, nbytes); | ||
198 | return ecb_crypt(desc, &walk, true); | ||
199 | } | 177 | } |
200 | 178 | ||
201 | static int ecb_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | 179 | static int ecb_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, |
202 | struct scatterlist *src, unsigned int nbytes) | 180 | struct scatterlist *src, unsigned int nbytes) |
203 | { | 181 | { |
204 | struct blkcipher_walk walk; | 182 | return glue_ecb_crypt_128bit(&twofish_dec, desc, dst, src, nbytes); |
205 | |||
206 | blkcipher_walk_init(&walk, dst, src, nbytes); | ||
207 | return ecb_crypt(desc, &walk, false); | ||
208 | } | ||
209 | |||
210 | static unsigned int __cbc_encrypt(struct blkcipher_desc *desc, | ||
211 | struct blkcipher_walk *walk) | ||
212 | { | ||
213 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | ||
214 | const unsigned int bsize = TF_BLOCK_SIZE; | ||
215 | unsigned int nbytes = walk->nbytes; | ||
216 | u128 *src = (u128 *)walk->src.virt.addr; | ||
217 | u128 *dst = (u128 *)walk->dst.virt.addr; | ||
218 | u128 *iv = (u128 *)walk->iv; | ||
219 | |||
220 | do { | ||
221 | u128_xor(dst, src, iv); | ||
222 | twofish_enc_blk(ctx, (u8 *)dst, (u8 *)dst); | ||
223 | iv = dst; | ||
224 | |||
225 | src += 1; | ||
226 | dst += 1; | ||
227 | nbytes -= bsize; | ||
228 | } while (nbytes >= bsize); | ||
229 | |||
230 | u128_xor((u128 *)walk->iv, (u128 *)walk->iv, iv); | ||
231 | return nbytes; | ||
232 | } | 183 | } |
233 | 184 | ||
234 | static int cbc_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | 185 | static int cbc_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, |
235 | struct scatterlist *src, unsigned int nbytes) | 186 | struct scatterlist *src, unsigned int nbytes) |
236 | { | 187 | { |
237 | struct blkcipher_walk walk; | 188 | return glue_cbc_encrypt_128bit(GLUE_FUNC_CAST(twofish_enc_blk), desc, |
238 | int err; | 189 | dst, src, nbytes); |
239 | |||
240 | blkcipher_walk_init(&walk, dst, src, nbytes); | ||
241 | err = blkcipher_walk_virt(desc, &walk); | ||
242 | |||
243 | while ((nbytes = walk.nbytes)) { | ||
244 | nbytes = __cbc_encrypt(desc, &walk); | ||
245 | err = blkcipher_walk_done(desc, &walk, nbytes); | ||
246 | } | ||
247 | |||
248 | return err; | ||
249 | } | ||
250 | |||
251 | static unsigned int __cbc_decrypt(struct blkcipher_desc *desc, | ||
252 | struct blkcipher_walk *walk) | ||
253 | { | ||
254 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | ||
255 | const unsigned int bsize = TF_BLOCK_SIZE; | ||
256 | unsigned int nbytes = walk->nbytes; | ||
257 | u128 *src = (u128 *)walk->src.virt.addr; | ||
258 | u128 *dst = (u128 *)walk->dst.virt.addr; | ||
259 | u128 ivs[TWOFISH_PARALLEL_BLOCKS - 1]; | ||
260 | u128 last_iv; | ||
261 | int i; | ||
262 | |||
263 | /* Start of the last block. */ | ||
264 | src += nbytes / bsize - 1; | ||
265 | dst += nbytes / bsize - 1; | ||
266 | |||
267 | last_iv = *src; | ||
268 | |||
269 | /* Process multi-block batch */ | ||
270 | if (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS) { | ||
271 | do { | ||
272 | nbytes -= bsize * (TWOFISH_PARALLEL_BLOCKS - 1); | ||
273 | src -= TWOFISH_PARALLEL_BLOCKS - 1; | ||
274 | dst -= TWOFISH_PARALLEL_BLOCKS - 1; | ||
275 | |||
276 | for (i = 0; i < TWOFISH_PARALLEL_BLOCKS - 1; i++) | ||
277 | ivs[i] = src[i]; | ||
278 | |||
279 | twofish_dec_blk_xway(ctx, (u8 *)dst, (u8 *)src); | ||
280 | |||
281 | for (i = 0; i < TWOFISH_PARALLEL_BLOCKS - 1; i++) | ||
282 | u128_xor(dst + (i + 1), dst + (i + 1), ivs + i); | ||
283 | |||
284 | nbytes -= bsize; | ||
285 | if (nbytes < bsize) | ||
286 | goto done; | ||
287 | |||
288 | u128_xor(dst, dst, src - 1); | ||
289 | src -= 1; | ||
290 | dst -= 1; | ||
291 | } while (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS); | ||
292 | |||
293 | if (nbytes < bsize) | ||
294 | goto done; | ||
295 | } | ||
296 | |||
297 | /* Process three block batch */ | ||
298 | if (nbytes >= bsize * 3) { | ||
299 | do { | ||
300 | nbytes -= bsize * (3 - 1); | ||
301 | src -= 3 - 1; | ||
302 | dst -= 3 - 1; | ||
303 | |||
304 | ivs[0] = src[0]; | ||
305 | ivs[1] = src[1]; | ||
306 | |||
307 | twofish_dec_blk_3way(ctx, (u8 *)dst, (u8 *)src); | ||
308 | |||
309 | u128_xor(dst + 1, dst + 1, ivs + 0); | ||
310 | u128_xor(dst + 2, dst + 2, ivs + 1); | ||
311 | |||
312 | nbytes -= bsize; | ||
313 | if (nbytes < bsize) | ||
314 | goto done; | ||
315 | |||
316 | u128_xor(dst, dst, src - 1); | ||
317 | src -= 1; | ||
318 | dst -= 1; | ||
319 | } while (nbytes >= bsize * 3); | ||
320 | |||
321 | if (nbytes < bsize) | ||
322 | goto done; | ||
323 | } | ||
324 | |||
325 | /* Handle leftovers */ | ||
326 | for (;;) { | ||
327 | twofish_dec_blk(ctx, (u8 *)dst, (u8 *)src); | ||
328 | |||
329 | nbytes -= bsize; | ||
330 | if (nbytes < bsize) | ||
331 | break; | ||
332 | |||
333 | u128_xor(dst, dst, src - 1); | ||
334 | src -= 1; | ||
335 | dst -= 1; | ||
336 | } | ||
337 | |||
338 | done: | ||
339 | u128_xor(dst, dst, (u128 *)walk->iv); | ||
340 | *(u128 *)walk->iv = last_iv; | ||
341 | |||
342 | return nbytes; | ||
343 | } | 190 | } |
344 | 191 | ||
345 | static int cbc_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | 192 | static int cbc_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, |
346 | struct scatterlist *src, unsigned int nbytes) | 193 | struct scatterlist *src, unsigned int nbytes) |
347 | { | 194 | { |
348 | bool fpu_enabled = false; | 195 | return glue_cbc_decrypt_128bit(&twofish_dec_cbc, desc, dst, src, |
349 | struct blkcipher_walk walk; | 196 | nbytes); |
350 | int err; | ||
351 | |||
352 | blkcipher_walk_init(&walk, dst, src, nbytes); | ||
353 | err = blkcipher_walk_virt(desc, &walk); | ||
354 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | ||
355 | |||
356 | while ((nbytes = walk.nbytes)) { | ||
357 | fpu_enabled = twofish_fpu_begin(fpu_enabled, nbytes); | ||
358 | nbytes = __cbc_decrypt(desc, &walk); | ||
359 | err = blkcipher_walk_done(desc, &walk, nbytes); | ||
360 | } | ||
361 | |||
362 | twofish_fpu_end(fpu_enabled); | ||
363 | return err; | ||
364 | } | ||
365 | |||
366 | static inline void u128_to_be128(be128 *dst, const u128 *src) | ||
367 | { | ||
368 | dst->a = cpu_to_be64(src->a); | ||
369 | dst->b = cpu_to_be64(src->b); | ||
370 | } | 197 | } |
371 | 198 | ||
372 | static inline void be128_to_u128(u128 *dst, const be128 *src) | 199 | static int ctr_crypt(struct blkcipher_desc *desc, struct scatterlist *dst, |
373 | { | 200 | struct scatterlist *src, unsigned int nbytes) |
374 | dst->a = be64_to_cpu(src->a); | ||
375 | dst->b = be64_to_cpu(src->b); | ||
376 | } | ||
377 | |||
378 | static inline void u128_inc(u128 *i) | ||
379 | { | 201 | { |
380 | i->b++; | 202 | return glue_ctr_crypt_128bit(&twofish_ctr, desc, dst, src, nbytes); |
381 | if (!i->b) | ||
382 | i->a++; | ||
383 | } | 203 | } |
384 | 204 | ||
385 | static void ctr_crypt_final(struct blkcipher_desc *desc, | 205 | static inline bool twofish_fpu_begin(bool fpu_enabled, unsigned int nbytes) |
386 | struct blkcipher_walk *walk) | ||
387 | { | 206 | { |
388 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | 207 | return glue_fpu_begin(TF_BLOCK_SIZE, TWOFISH_PARALLEL_BLOCKS, NULL, |
389 | u8 *ctrblk = walk->iv; | 208 | fpu_enabled, nbytes); |
390 | u8 keystream[TF_BLOCK_SIZE]; | ||
391 | u8 *src = walk->src.virt.addr; | ||
392 | u8 *dst = walk->dst.virt.addr; | ||
393 | unsigned int nbytes = walk->nbytes; | ||
394 | |||
395 | twofish_enc_blk(ctx, keystream, ctrblk); | ||
396 | crypto_xor(keystream, src, nbytes); | ||
397 | memcpy(dst, keystream, nbytes); | ||
398 | |||
399 | crypto_inc(ctrblk, TF_BLOCK_SIZE); | ||
400 | } | 209 | } |
401 | 210 | ||
402 | static unsigned int __ctr_crypt(struct blkcipher_desc *desc, | 211 | static inline void twofish_fpu_end(bool fpu_enabled) |
403 | struct blkcipher_walk *walk) | ||
404 | { | ||
405 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | ||
406 | const unsigned int bsize = TF_BLOCK_SIZE; | ||
407 | unsigned int nbytes = walk->nbytes; | ||
408 | u128 *src = (u128 *)walk->src.virt.addr; | ||
409 | u128 *dst = (u128 *)walk->dst.virt.addr; | ||
410 | u128 ctrblk; | ||
411 | be128 ctrblocks[TWOFISH_PARALLEL_BLOCKS]; | ||
412 | int i; | ||
413 | |||
414 | be128_to_u128(&ctrblk, (be128 *)walk->iv); | ||
415 | |||
416 | /* Process multi-block batch */ | ||
417 | if (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS) { | ||
418 | do { | ||
419 | /* create ctrblks for parallel encrypt */ | ||
420 | for (i = 0; i < TWOFISH_PARALLEL_BLOCKS; i++) { | ||
421 | if (dst != src) | ||
422 | dst[i] = src[i]; | ||
423 | |||
424 | u128_to_be128(&ctrblocks[i], &ctrblk); | ||
425 | u128_inc(&ctrblk); | ||
426 | } | ||
427 | |||
428 | twofish_enc_blk_xway_xor(ctx, (u8 *)dst, | ||
429 | (u8 *)ctrblocks); | ||
430 | |||
431 | src += TWOFISH_PARALLEL_BLOCKS; | ||
432 | dst += TWOFISH_PARALLEL_BLOCKS; | ||
433 | nbytes -= bsize * TWOFISH_PARALLEL_BLOCKS; | ||
434 | } while (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS); | ||
435 | |||
436 | if (nbytes < bsize) | ||
437 | goto done; | ||
438 | } | ||
439 | |||
440 | /* Process three block batch */ | ||
441 | if (nbytes >= bsize * 3) { | ||
442 | do { | ||
443 | if (dst != src) { | ||
444 | dst[0] = src[0]; | ||
445 | dst[1] = src[1]; | ||
446 | dst[2] = src[2]; | ||
447 | } | ||
448 | |||
449 | /* create ctrblks for parallel encrypt */ | ||
450 | u128_to_be128(&ctrblocks[0], &ctrblk); | ||
451 | u128_inc(&ctrblk); | ||
452 | u128_to_be128(&ctrblocks[1], &ctrblk); | ||
453 | u128_inc(&ctrblk); | ||
454 | u128_to_be128(&ctrblocks[2], &ctrblk); | ||
455 | u128_inc(&ctrblk); | ||
456 | |||
457 | twofish_enc_blk_3way_xor(ctx, (u8 *)dst, | ||
458 | (u8 *)ctrblocks); | ||
459 | |||
460 | src += 3; | ||
461 | dst += 3; | ||
462 | nbytes -= bsize * 3; | ||
463 | } while (nbytes >= bsize * 3); | ||
464 | |||
465 | if (nbytes < bsize) | ||
466 | goto done; | ||
467 | } | ||
468 | |||
469 | /* Handle leftovers */ | ||
470 | do { | ||
471 | if (dst != src) | ||
472 | *dst = *src; | ||
473 | |||
474 | u128_to_be128(&ctrblocks[0], &ctrblk); | ||
475 | u128_inc(&ctrblk); | ||
476 | |||
477 | twofish_enc_blk(ctx, (u8 *)ctrblocks, (u8 *)ctrblocks); | ||
478 | u128_xor(dst, dst, (u128 *)ctrblocks); | ||
479 | |||
480 | src += 1; | ||
481 | dst += 1; | ||
482 | nbytes -= bsize; | ||
483 | } while (nbytes >= bsize); | ||
484 | |||
485 | done: | ||
486 | u128_to_be128((be128 *)walk->iv, &ctrblk); | ||
487 | return nbytes; | ||
488 | } | ||
489 | |||
490 | static int ctr_crypt(struct blkcipher_desc *desc, struct scatterlist *dst, | ||
491 | struct scatterlist *src, unsigned int nbytes) | ||
492 | { | 212 | { |
493 | bool fpu_enabled = false; | 213 | glue_fpu_end(fpu_enabled); |
494 | struct blkcipher_walk walk; | ||
495 | int err; | ||
496 | |||
497 | blkcipher_walk_init(&walk, dst, src, nbytes); | ||
498 | err = blkcipher_walk_virt_block(desc, &walk, TF_BLOCK_SIZE); | ||
499 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | ||
500 | |||
501 | while ((nbytes = walk.nbytes) >= TF_BLOCK_SIZE) { | ||
502 | fpu_enabled = twofish_fpu_begin(fpu_enabled, nbytes); | ||
503 | nbytes = __ctr_crypt(desc, &walk); | ||
504 | err = blkcipher_walk_done(desc, &walk, nbytes); | ||
505 | } | ||
506 | |||
507 | twofish_fpu_end(fpu_enabled); | ||
508 | |||
509 | if (walk.nbytes) { | ||
510 | ctr_crypt_final(desc, &walk); | ||
511 | err = blkcipher_walk_done(desc, &walk, 0); | ||
512 | } | ||
513 | |||
514 | return err; | ||
515 | } | 214 | } |
516 | 215 | ||
517 | struct crypt_priv { | 216 | struct crypt_priv { |
@@ -563,26 +262,6 @@ static void decrypt_callback(void *priv, u8 *srcdst, unsigned int nbytes) | |||
563 | twofish_dec_blk(ctx->ctx, srcdst, srcdst); | 262 | twofish_dec_blk(ctx->ctx, srcdst, srcdst); |
564 | } | 263 | } |
565 | 264 | ||
566 | struct twofish_lrw_ctx { | ||
567 | struct lrw_table_ctx lrw_table; | ||
568 | struct twofish_ctx twofish_ctx; | ||
569 | }; | ||
570 | |||
571 | static int lrw_twofish_setkey(struct crypto_tfm *tfm, const u8 *key, | ||
572 | unsigned int keylen) | ||
573 | { | ||
574 | struct twofish_lrw_ctx *ctx = crypto_tfm_ctx(tfm); | ||
575 | int err; | ||
576 | |||
577 | err = __twofish_setkey(&ctx->twofish_ctx, key, | ||
578 | keylen - TF_BLOCK_SIZE, &tfm->crt_flags); | ||
579 | if (err) | ||
580 | return err; | ||
581 | |||
582 | return lrw_init_table(&ctx->lrw_table, key + keylen - | ||
583 | TF_BLOCK_SIZE); | ||
584 | } | ||
585 | |||
586 | static int lrw_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | 265 | static int lrw_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, |
587 | struct scatterlist *src, unsigned int nbytes) | 266 | struct scatterlist *src, unsigned int nbytes) |
588 | { | 267 | { |
@@ -635,43 +314,6 @@ static int lrw_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |||
635 | return ret; | 314 | return ret; |
636 | } | 315 | } |
637 | 316 | ||
638 | static void lrw_exit_tfm(struct crypto_tfm *tfm) | ||
639 | { | ||
640 | struct twofish_lrw_ctx *ctx = crypto_tfm_ctx(tfm); | ||
641 | |||
642 | lrw_free_table(&ctx->lrw_table); | ||
643 | } | ||
644 | |||
645 | struct twofish_xts_ctx { | ||
646 | struct twofish_ctx tweak_ctx; | ||
647 | struct twofish_ctx crypt_ctx; | ||
648 | }; | ||
649 | |||
650 | static int xts_twofish_setkey(struct crypto_tfm *tfm, const u8 *key, | ||
651 | unsigned int keylen) | ||
652 | { | ||
653 | struct twofish_xts_ctx *ctx = crypto_tfm_ctx(tfm); | ||
654 | u32 *flags = &tfm->crt_flags; | ||
655 | int err; | ||
656 | |||
657 | /* key consists of keys of equal size concatenated, therefore | ||
658 | * the length must be even | ||
659 | */ | ||
660 | if (keylen % 2) { | ||
661 | *flags |= CRYPTO_TFM_RES_BAD_KEY_LEN; | ||
662 | return -EINVAL; | ||
663 | } | ||
664 | |||
665 | /* first half of xts-key is for crypt */ | ||
666 | err = __twofish_setkey(&ctx->crypt_ctx, key, keylen / 2, flags); | ||
667 | if (err) | ||
668 | return err; | ||
669 | |||
670 | /* second half of xts-key is for tweak */ | ||
671 | return __twofish_setkey(&ctx->tweak_ctx, | ||
672 | key + keylen / 2, keylen / 2, flags); | ||
673 | } | ||
674 | |||
675 | static int xts_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | 317 | static int xts_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, |
676 | struct scatterlist *src, unsigned int nbytes) | 318 | struct scatterlist *src, unsigned int nbytes) |
677 | { | 319 | { |
@@ -798,7 +440,7 @@ static struct crypto_alg twofish_algs[10] = { { | |||
798 | .cra_type = &crypto_blkcipher_type, | 440 | .cra_type = &crypto_blkcipher_type, |
799 | .cra_module = THIS_MODULE, | 441 | .cra_module = THIS_MODULE, |
800 | .cra_list = LIST_HEAD_INIT(twofish_algs[3].cra_list), | 442 | .cra_list = LIST_HEAD_INIT(twofish_algs[3].cra_list), |
801 | .cra_exit = lrw_exit_tfm, | 443 | .cra_exit = lrw_twofish_exit_tfm, |
802 | .cra_u = { | 444 | .cra_u = { |
803 | .blkcipher = { | 445 | .blkcipher = { |
804 | .min_keysize = TF_MIN_KEY_SIZE + | 446 | .min_keysize = TF_MIN_KEY_SIZE + |