]>
Commit | Line | Data |
---|---|---|
7c673cae FG |
1 | /********************************************************************** |
2 | Copyright(c) 2011-2016 Intel Corporation All rights reserved. | |
3 | ||
4 | Redistribution and use in source and binary forms, with or without | |
1e59de90 | 5 | modification, are permitted provided that the following conditions |
7c673cae FG |
6 | are met: |
7 | * Redistributions of source code must retain the above copyright | |
8 | notice, this list of conditions and the following disclaimer. | |
9 | * Redistributions in binary form must reproduce the above copyright | |
10 | notice, this list of conditions and the following disclaimer in | |
11 | the documentation and/or other materials provided with the | |
12 | distribution. | |
13 | * Neither the name of Intel Corporation nor the names of its | |
14 | contributors may be used to endorse or promote products derived | |
15 | from this software without specific prior written permission. | |
16 | ||
17 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | |
18 | "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | |
19 | LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | |
20 | A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | |
21 | OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | |
22 | SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT | |
23 | LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | |
24 | DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | |
25 | THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | |
26 | (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE | |
27 | OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
28 | **********************************************************************/ | |
29 | ||
1e59de90 TL |
30 | #if defined(__clang__) |
31 | # pragma clang attribute push (__attribute__((target("avx"))), apply_to=function) | |
32 | #elif defined(__ICC) | |
33 | # pragma intel optimization_parameter target_arch=AVX | |
34 | #elif defined(__ICL) | |
35 | # pragma [intel] optimization_parameter target_arch=AVX | |
36 | #elif (__GNUC__ >= 5) | |
37 | # pragma GCC target("avx") | |
38 | #endif | |
39 | ||
7c673cae FG |
40 | #include "md5_mb.h" |
41 | #include "memcpy_inline.h" | |
42 | ||
43 | #ifdef _MSC_VER | |
44 | #include <intrin.h> | |
45 | #define inline __inline | |
46 | #endif | |
47 | ||
48 | static inline void hash_init_digest(MD5_WORD_T * digest); | |
1e59de90 | 49 | static inline uint32_t hash_pad(uint8_t padblock[MD5_BLOCK_SIZE * 2], uint64_t total_len); |
7c673cae FG |
50 | static MD5_HASH_CTX *md5_ctx_mgr_resubmit(MD5_HASH_CTX_MGR * mgr, MD5_HASH_CTX * ctx); |
51 | ||
52 | void md5_ctx_mgr_init_avx(MD5_HASH_CTX_MGR * mgr) | |
53 | { | |
54 | md5_mb_mgr_init_avx(&mgr->mgr); | |
55 | } | |
56 | ||
57 | MD5_HASH_CTX *md5_ctx_mgr_submit_avx(MD5_HASH_CTX_MGR * mgr, MD5_HASH_CTX * ctx, | |
58 | const void *buffer, uint32_t len, HASH_CTX_FLAG flags) | |
59 | { | |
60 | if (flags & (~HASH_ENTIRE)) { | |
61 | // User should not pass anything other than FIRST, UPDATE, or LAST | |
62 | ctx->error = HASH_CTX_ERROR_INVALID_FLAGS; | |
63 | return ctx; | |
64 | } | |
65 | ||
66 | if (ctx->status & HASH_CTX_STS_PROCESSING) { | |
67 | // Cannot submit to a currently processing job. | |
68 | ctx->error = HASH_CTX_ERROR_ALREADY_PROCESSING; | |
69 | return ctx; | |
70 | } | |
71 | ||
72 | if ((ctx->status & HASH_CTX_STS_COMPLETE) && !(flags & HASH_FIRST)) { | |
73 | // Cannot update a finished job. | |
74 | ctx->error = HASH_CTX_ERROR_ALREADY_COMPLETED; | |
75 | return ctx; | |
76 | } | |
77 | ||
78 | if (flags & HASH_FIRST) { | |
79 | // Init digest | |
80 | hash_init_digest(ctx->job.result_digest); | |
81 | ||
82 | // Reset byte counter | |
83 | ctx->total_length = 0; | |
84 | ||
85 | // Clear extra blocks | |
86 | ctx->partial_block_buffer_length = 0; | |
87 | } | |
88 | // If we made it here, there were no errors during this call to submit | |
89 | ctx->error = HASH_CTX_ERROR_NONE; | |
90 | ||
91 | // Store buffer ptr info from user | |
92 | ctx->incoming_buffer = buffer; | |
93 | ctx->incoming_buffer_length = len; | |
94 | ||
95 | // Store the user's request flags and mark this ctx as currently being processed. | |
96 | ctx->status = (flags & HASH_LAST) ? | |
97 | (HASH_CTX_STS) (HASH_CTX_STS_PROCESSING | HASH_CTX_STS_LAST) : | |
98 | HASH_CTX_STS_PROCESSING; | |
99 | ||
100 | // Advance byte counter | |
101 | ctx->total_length += len; | |
102 | ||
103 | // If there is anything currently buffered in the extra blocks, append to it until it contains a whole block. | |
104 | // Or if the user's buffer contains less than a whole block, append as much as possible to the extra block. | |
105 | if ((ctx->partial_block_buffer_length) | (len < MD5_BLOCK_SIZE)) { | |
106 | // Compute how many bytes to copy from user buffer into extra block | |
107 | uint32_t copy_len = MD5_BLOCK_SIZE - ctx->partial_block_buffer_length; | |
108 | if (len < copy_len) | |
109 | copy_len = len; | |
110 | ||
111 | if (copy_len) { | |
112 | // Copy and update relevant pointers and counters | |
113 | memcpy_varlen(&ctx->partial_block_buffer | |
114 | [ctx->partial_block_buffer_length], buffer, copy_len); | |
115 | ||
116 | ctx->partial_block_buffer_length += copy_len; | |
117 | ctx->incoming_buffer = (const void *)((const char *)buffer + copy_len); | |
118 | ctx->incoming_buffer_length = len - copy_len; | |
119 | } | |
120 | // The extra block should never contain more than 1 block here | |
121 | assert(ctx->partial_block_buffer_length <= MD5_BLOCK_SIZE); | |
122 | ||
123 | // If the extra block buffer contains exactly 1 block, it can be hashed. | |
124 | if (ctx->partial_block_buffer_length >= MD5_BLOCK_SIZE) { | |
125 | ctx->partial_block_buffer_length = 0; | |
126 | ||
127 | ctx->job.buffer = ctx->partial_block_buffer; | |
128 | ctx->job.len = 1; | |
129 | ctx = (MD5_HASH_CTX *) md5_mb_mgr_submit_avx(&mgr->mgr, &ctx->job); | |
130 | } | |
131 | } | |
132 | ||
133 | return md5_ctx_mgr_resubmit(mgr, ctx); | |
134 | } | |
135 | ||
136 | MD5_HASH_CTX *md5_ctx_mgr_flush_avx(MD5_HASH_CTX_MGR * mgr) | |
137 | { | |
138 | MD5_HASH_CTX *ctx; | |
139 | ||
140 | while (1) { | |
141 | ctx = (MD5_HASH_CTX *) md5_mb_mgr_flush_avx(&mgr->mgr); | |
142 | ||
143 | // If flush returned 0, there are no more jobs in flight. | |
144 | if (!ctx) | |
145 | return NULL; | |
146 | ||
147 | // If flush returned a job, verify that it is safe to return to the user. | |
148 | // If it is not ready, resubmit the job to finish processing. | |
149 | ctx = md5_ctx_mgr_resubmit(mgr, ctx); | |
150 | ||
151 | // If md5_ctx_mgr_resubmit returned a job, it is ready to be returned. | |
152 | if (ctx) | |
153 | return ctx; | |
154 | ||
155 | // Otherwise, all jobs currently being managed by the HASH_CTX_MGR still need processing. Loop. | |
156 | } | |
157 | } | |
158 | ||
159 | static MD5_HASH_CTX *md5_ctx_mgr_resubmit(MD5_HASH_CTX_MGR * mgr, MD5_HASH_CTX * ctx) | |
160 | { | |
161 | while (ctx) { | |
162 | ||
163 | if (ctx->status & HASH_CTX_STS_COMPLETE) { | |
164 | ctx->status = HASH_CTX_STS_COMPLETE; // Clear PROCESSING bit | |
165 | return ctx; | |
166 | } | |
167 | // If the extra blocks are empty, begin hashing what remains in the user's buffer. | |
168 | if (ctx->partial_block_buffer_length == 0 && ctx->incoming_buffer_length) { | |
169 | const void *buffer = ctx->incoming_buffer; | |
170 | uint32_t len = ctx->incoming_buffer_length; | |
171 | ||
172 | // Only entire blocks can be hashed. Copy remainder to extra blocks buffer. | |
173 | uint32_t copy_len = len & (MD5_BLOCK_SIZE - 1); | |
174 | ||
175 | if (copy_len) { | |
176 | len -= copy_len; | |
177 | //memcpy(ctx->partial_block_buffer, ((const char*)buffer + len), copy_len); | |
178 | memcpy_varlen(ctx->partial_block_buffer, | |
179 | ((const char *)buffer + len), copy_len); | |
180 | ctx->partial_block_buffer_length = copy_len; | |
181 | } | |
182 | ||
183 | ctx->incoming_buffer_length = 0; | |
184 | ||
185 | // len should be a multiple of the block size now | |
186 | assert((len % MD5_BLOCK_SIZE) == 0); | |
187 | ||
188 | // Set len to the number of blocks to be hashed in the user's buffer | |
189 | len >>= MD5_LOG2_BLOCK_SIZE; | |
190 | ||
191 | if (len) { | |
192 | ctx->job.buffer = (uint8_t *) buffer; | |
193 | ctx->job.len = len; | |
194 | ctx = (MD5_HASH_CTX *) md5_mb_mgr_submit_avx(&mgr->mgr, | |
195 | &ctx->job); | |
196 | continue; | |
197 | } | |
198 | } | |
199 | // If the extra blocks are not empty, then we are either on the last block(s) | |
200 | // or we need more user input before continuing. | |
201 | if (ctx->status & HASH_CTX_STS_LAST) { | |
202 | ||
203 | uint8_t *buf = ctx->partial_block_buffer; | |
204 | uint32_t n_extra_blocks = hash_pad(buf, ctx->total_length); | |
205 | ||
206 | ctx->status = | |
207 | (HASH_CTX_STS) (HASH_CTX_STS_PROCESSING | HASH_CTX_STS_COMPLETE); | |
208 | ||
209 | ctx->job.buffer = buf; | |
210 | ctx->job.len = (uint32_t) n_extra_blocks; | |
211 | ctx = (MD5_HASH_CTX *) md5_mb_mgr_submit_avx(&mgr->mgr, &ctx->job); | |
212 | continue; | |
213 | } | |
214 | ||
215 | if (ctx) | |
216 | ctx->status = HASH_CTX_STS_IDLE; | |
217 | return ctx; | |
218 | } | |
219 | ||
220 | return NULL; | |
221 | } | |
222 | ||
223 | static inline void hash_init_digest(MD5_WORD_T * digest) | |
224 | { | |
225 | static const MD5_WORD_T hash_initial_digest[MD5_DIGEST_NWORDS] = | |
226 | { MD5_INITIAL_DIGEST }; | |
227 | //memcpy(digest, hash_initial_digest, sizeof(hash_initial_digest)); | |
228 | memcpy_fixedlen(digest, hash_initial_digest, sizeof(hash_initial_digest)); | |
229 | } | |
230 | ||
1e59de90 | 231 | static inline uint32_t hash_pad(uint8_t padblock[MD5_BLOCK_SIZE * 2], uint64_t total_len) |
7c673cae | 232 | { |
1e59de90 | 233 | uint32_t i = (uint32_t) (total_len & (MD5_BLOCK_SIZE - 1)); |
7c673cae FG |
234 | |
235 | // memset(&padblock[i], 0, MD5_BLOCK_SIZE); | |
236 | memclr_fixedlen(&padblock[i], MD5_BLOCK_SIZE); | |
237 | padblock[i] = 0x80; | |
238 | ||
239 | i += ((MD5_BLOCK_SIZE - 1) & (0 - (total_len + MD5_PADLENGTHFIELD_SIZE + 1))) + 1 + | |
240 | MD5_PADLENGTHFIELD_SIZE; | |
241 | ||
242 | *((uint64_t *) & padblock[i - 8]) = ((uint64_t) total_len << 3); | |
243 | ||
244 | return i >> MD5_LOG2_BLOCK_SIZE; // Number of extra blocks to hash | |
245 | } | |
246 | ||
247 | struct slver { | |
248 | uint16_t snum; | |
249 | uint8_t ver; | |
250 | uint8_t core; | |
251 | }; | |
252 | struct slver md5_ctx_mgr_init_avx_slver_02020183; | |
253 | struct slver md5_ctx_mgr_init_avx_slver = { 0x0183, 0x02, 0x02 }; | |
254 | ||
255 | struct slver md5_ctx_mgr_submit_avx_slver_02020184; | |
256 | struct slver md5_ctx_mgr_submit_avx_slver = { 0x0184, 0x02, 0x02 }; | |
257 | ||
258 | struct slver md5_ctx_mgr_flush_avx_slver_02020185; | |
259 | struct slver md5_ctx_mgr_flush_avx_slver = { 0x0185, 0x02, 0x02 }; | |
1e59de90 TL |
260 | |
261 | #if defined(__clang__) | |
262 | # pragma clang attribute pop | |
263 | #endif |