1 /*
2 * Using hardware provided CRC32 instruction to accelerate the CRC32 disposal.
3 * CRC32C polynomial:0x1EDC6F41(BE)/0x82F63B78(LE)
4 * CRC32 is a new instruction in Intel SSE4.2, the reference can be found at:
5 * http://www.intel.com/products/processor/manuals/
6 * Intel(R) 64 and IA-32 Architectures Software Developer's Manual
7 * Volume 2A: Instruction Set Reference, A-M
8 *
9 * Copyright (C) 2008 Intel Corporation
10 * Authors: Austin Zhang <austin_zhang@linux.intel.com>
11 * Kent Liu <kent.liu@intel.com>
12 *
13 * This program is free software; you can redistribute it and/or modify it
14 * under the terms and conditions of the GNU General Public License,
15 * version 2, as published by the Free Software Foundation.
16 *
17 * This program is distributed in the hope it will be useful, but WITHOUT
18 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
19 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
20 * more details.
21 *
22 * You should have received a copy of the GNU General Public License along with
23 * this program; if not, write to the Free Software Foundation, Inc.,
24 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
25 *
26 */
27 #include <linux/init.h>
28 #include <linux/module.h>
29 #include <linux/string.h>
30 #include <linux/kernel.h>
31 #include <crypto/internal/hash.h>
32
33 #include <asm/cpufeature.h>
34
35 #define CHKSUM_BLOCK_SIZE 1
36 #define CHKSUM_DIGEST_SIZE 4
37
38 #define SCALE_F sizeof(unsigned long)
39
40 #ifdef CONFIG_X86_64
41 #define REX_PRE "0x48, "
42 #else
43 #define REX_PRE
44 #endif
45
crc32c_intel_le_hw_byte(u32 crc,unsigned char const * data,size_t length)46 static u32 crc32c_intel_le_hw_byte(u32 crc, unsigned char const *data, size_t length)
47 {
48 while (length--) {
49 __asm__ __volatile__(
50 ".byte 0xf2, 0xf, 0x38, 0xf0, 0xf1"
51 :"=S"(crc)
52 :"0"(crc), "c"(*data)
53 );
54 data++;
55 }
56
57 return crc;
58 }
59
crc32c_intel_le_hw(u32 crc,unsigned char const * p,size_t len)60 static u32 __pure crc32c_intel_le_hw(u32 crc, unsigned char const *p, size_t len)
61 {
62 unsigned int iquotient = len / SCALE_F;
63 unsigned int iremainder = len % SCALE_F;
64 unsigned long *ptmp = (unsigned long *)p;
65
66 while (iquotient--) {
67 __asm__ __volatile__(
68 ".byte 0xf2, " REX_PRE "0xf, 0x38, 0xf1, 0xf1;"
69 :"=S"(crc)
70 :"0"(crc), "c"(*ptmp)
71 );
72 ptmp++;
73 }
74
75 if (iremainder)
76 crc = crc32c_intel_le_hw_byte(crc, (unsigned char *)ptmp,
77 iremainder);
78
79 return crc;
80 }
81
82 /*
83 * Setting the seed allows arbitrary accumulators and flexible XOR policy
84 * If your algorithm starts with ~0, then XOR with ~0 before you set
85 * the seed.
86 */
crc32c_intel_setkey(struct crypto_shash * hash,const u8 * key,unsigned int keylen)87 static int crc32c_intel_setkey(struct crypto_shash *hash, const u8 *key,
88 unsigned int keylen)
89 {
90 u32 *mctx = crypto_shash_ctx(hash);
91
92 if (keylen != sizeof(u32)) {
93 crypto_shash_set_flags(hash, CRYPTO_TFM_RES_BAD_KEY_LEN);
94 return -EINVAL;
95 }
96 *mctx = le32_to_cpup((__le32 *)key);
97 return 0;
98 }
99
crc32c_intel_init(struct shash_desc * desc)100 static int crc32c_intel_init(struct shash_desc *desc)
101 {
102 u32 *mctx = crypto_shash_ctx(desc->tfm);
103 u32 *crcp = shash_desc_ctx(desc);
104
105 *crcp = *mctx;
106
107 return 0;
108 }
109
crc32c_intel_update(struct shash_desc * desc,const u8 * data,unsigned int len)110 static int crc32c_intel_update(struct shash_desc *desc, const u8 *data,
111 unsigned int len)
112 {
113 u32 *crcp = shash_desc_ctx(desc);
114
115 *crcp = crc32c_intel_le_hw(*crcp, data, len);
116 return 0;
117 }
118
__crc32c_intel_finup(u32 * crcp,const u8 * data,unsigned int len,u8 * out)119 static int __crc32c_intel_finup(u32 *crcp, const u8 *data, unsigned int len,
120 u8 *out)
121 {
122 *(__le32 *)out = ~cpu_to_le32(crc32c_intel_le_hw(*crcp, data, len));
123 return 0;
124 }
125
crc32c_intel_finup(struct shash_desc * desc,const u8 * data,unsigned int len,u8 * out)126 static int crc32c_intel_finup(struct shash_desc *desc, const u8 *data,
127 unsigned int len, u8 *out)
128 {
129 return __crc32c_intel_finup(shash_desc_ctx(desc), data, len, out);
130 }
131
crc32c_intel_final(struct shash_desc * desc,u8 * out)132 static int crc32c_intel_final(struct shash_desc *desc, u8 *out)
133 {
134 u32 *crcp = shash_desc_ctx(desc);
135
136 *(__le32 *)out = ~cpu_to_le32p(crcp);
137 return 0;
138 }
139
crc32c_intel_digest(struct shash_desc * desc,const u8 * data,unsigned int len,u8 * out)140 static int crc32c_intel_digest(struct shash_desc *desc, const u8 *data,
141 unsigned int len, u8 *out)
142 {
143 return __crc32c_intel_finup(crypto_shash_ctx(desc->tfm), data, len,
144 out);
145 }
146
crc32c_intel_cra_init(struct crypto_tfm * tfm)147 static int crc32c_intel_cra_init(struct crypto_tfm *tfm)
148 {
149 u32 *key = crypto_tfm_ctx(tfm);
150
151 *key = ~0;
152
153 return 0;
154 }
155
156 static struct shash_alg alg = {
157 .setkey = crc32c_intel_setkey,
158 .init = crc32c_intel_init,
159 .update = crc32c_intel_update,
160 .final = crc32c_intel_final,
161 .finup = crc32c_intel_finup,
162 .digest = crc32c_intel_digest,
163 .descsize = sizeof(u32),
164 .digestsize = CHKSUM_DIGEST_SIZE,
165 .base = {
166 .cra_name = "crc32c",
167 .cra_driver_name = "crc32c-intel",
168 .cra_priority = 200,
169 .cra_blocksize = CHKSUM_BLOCK_SIZE,
170 .cra_ctxsize = sizeof(u32),
171 .cra_module = THIS_MODULE,
172 .cra_init = crc32c_intel_cra_init,
173 }
174 };
175
176
crc32c_intel_mod_init(void)177 static int __init crc32c_intel_mod_init(void)
178 {
179 if (cpu_has_xmm4_2)
180 return crypto_register_shash(&alg);
181 else
182 return -ENODEV;
183 }
184
crc32c_intel_mod_fini(void)185 static void __exit crc32c_intel_mod_fini(void)
186 {
187 crypto_unregister_shash(&alg);
188 }
189
190 module_init(crc32c_intel_mod_init);
191 module_exit(crc32c_intel_mod_fini);
192
193 MODULE_AUTHOR("Austin Zhang <austin.zhang@intel.com>, Kent Liu <kent.liu@intel.com>");
194 MODULE_DESCRIPTION("CRC32c (Castagnoli) optimization using Intel Hardware.");
195 MODULE_LICENSE("GPL");
196
197 MODULE_ALIAS("crc32c");
198 MODULE_ALIAS("crc32c-intel");
199