1/*
2 * Cryptographic API.
3 *
4 * Glue code for the SHA512 Secure Hash Algorithm assembler
5 * implementation using supplemental SSE3 / AVX / AVX2 instructions.
6 *
7 * This file is based on sha512_generic.c
8 *
9 * Copyright (C) 2013 Intel Corporation
10 * Author: Tim Chen <tim.c.chen@linux.intel.com>
11 *
12 * This program is free software; you can redistribute it and/or modify it
13 * under the terms of the GNU General Public License as published by the Free
14 * Software Foundation; either version 2 of the License, or (at your option)
15 * any later version.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
18 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
19 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
20 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
21 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
22 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
23 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
24 * SOFTWARE.
25 *
26 */
27
28#define pr_fmt(fmt)	KBUILD_MODNAME ": " fmt
29
30#include <crypto/internal/hash.h>
31#include <linux/init.h>
32#include <linux/module.h>
33#include <linux/mm.h>
34#include <linux/cryptohash.h>
35#include <linux/types.h>
36#include <crypto/sha.h>
37#include <crypto/sha512_base.h>
38#include <asm/i387.h>
39#include <asm/xcr.h>
40#include <asm/xsave.h>
41
42#include <linux/string.h>
43
44asmlinkage void sha512_transform_ssse3(u64 *digest, const char *data,
45				       u64 rounds);
46#ifdef CONFIG_AS_AVX
47asmlinkage void sha512_transform_avx(u64 *digest, const char *data,
48				     u64 rounds);
49#endif
50#ifdef CONFIG_AS_AVX2
51asmlinkage void sha512_transform_rorx(u64 *digest, const char *data,
52				      u64 rounds);
53#endif
54
55static void (*sha512_transform_asm)(u64 *, const char *, u64);
56
57static int sha512_ssse3_update(struct shash_desc *desc, const u8 *data,
58			       unsigned int len)
59{
60	struct sha512_state *sctx = shash_desc_ctx(desc);
61
62	if (!irq_fpu_usable() ||
63	    (sctx->count[0] % SHA512_BLOCK_SIZE) + len < SHA512_BLOCK_SIZE)
64		return crypto_sha512_update(desc, data, len);
65
66	/* make sure casting to sha512_block_fn() is safe */
67	BUILD_BUG_ON(offsetof(struct sha512_state, state) != 0);
68
69	kernel_fpu_begin();
70	sha512_base_do_update(desc, data, len,
71			      (sha512_block_fn *)sha512_transform_asm);
72	kernel_fpu_end();
73
74	return 0;
75}
76
77static int sha512_ssse3_finup(struct shash_desc *desc, const u8 *data,
78			      unsigned int len, u8 *out)
79{
80	if (!irq_fpu_usable())
81		return crypto_sha512_finup(desc, data, len, out);
82
83	kernel_fpu_begin();
84	if (len)
85		sha512_base_do_update(desc, data, len,
86				      (sha512_block_fn *)sha512_transform_asm);
87	sha512_base_do_finalize(desc, (sha512_block_fn *)sha512_transform_asm);
88	kernel_fpu_end();
89
90	return sha512_base_finish(desc, out);
91}
92
93/* Add padding and return the message digest. */
94static int sha512_ssse3_final(struct shash_desc *desc, u8 *out)
95{
96	return sha512_ssse3_finup(desc, NULL, 0, out);
97}
98
99static struct shash_alg algs[] = { {
100	.digestsize	=	SHA512_DIGEST_SIZE,
101	.init		=	sha512_base_init,
102	.update		=	sha512_ssse3_update,
103	.final		=	sha512_ssse3_final,
104	.finup		=	sha512_ssse3_finup,
105	.descsize	=	sizeof(struct sha512_state),
106	.base		=	{
107		.cra_name	=	"sha512",
108		.cra_driver_name =	"sha512-ssse3",
109		.cra_priority	=	150,
110		.cra_flags	=	CRYPTO_ALG_TYPE_SHASH,
111		.cra_blocksize	=	SHA512_BLOCK_SIZE,
112		.cra_module	=	THIS_MODULE,
113	}
114},  {
115	.digestsize	=	SHA384_DIGEST_SIZE,
116	.init		=	sha384_base_init,
117	.update		=	sha512_ssse3_update,
118	.final		=	sha512_ssse3_final,
119	.finup		=	sha512_ssse3_finup,
120	.descsize	=	sizeof(struct sha512_state),
121	.base		=	{
122		.cra_name	=	"sha384",
123		.cra_driver_name =	"sha384-ssse3",
124		.cra_priority	=	150,
125		.cra_flags	=	CRYPTO_ALG_TYPE_SHASH,
126		.cra_blocksize	=	SHA384_BLOCK_SIZE,
127		.cra_module	=	THIS_MODULE,
128	}
129} };
130
131#ifdef CONFIG_AS_AVX
132static bool __init avx_usable(void)
133{
134	u64 xcr0;
135
136	if (!cpu_has_avx || !cpu_has_osxsave)
137		return false;
138
139	xcr0 = xgetbv(XCR_XFEATURE_ENABLED_MASK);
140	if ((xcr0 & (XSTATE_SSE | XSTATE_YMM)) != (XSTATE_SSE | XSTATE_YMM)) {
141		pr_info("AVX detected but unusable.\n");
142
143		return false;
144	}
145
146	return true;
147}
148#endif
149
150static int __init sha512_ssse3_mod_init(void)
151{
152	/* test for SSSE3 first */
153	if (cpu_has_ssse3)
154		sha512_transform_asm = sha512_transform_ssse3;
155
156#ifdef CONFIG_AS_AVX
157	/* allow AVX to override SSSE3, it's a little faster */
158	if (avx_usable()) {
159#ifdef CONFIG_AS_AVX2
160		if (boot_cpu_has(X86_FEATURE_AVX2))
161			sha512_transform_asm = sha512_transform_rorx;
162		else
163#endif
164			sha512_transform_asm = sha512_transform_avx;
165	}
166#endif
167
168	if (sha512_transform_asm) {
169#ifdef CONFIG_AS_AVX
170		if (sha512_transform_asm == sha512_transform_avx)
171			pr_info("Using AVX optimized SHA-512 implementation\n");
172#ifdef CONFIG_AS_AVX2
173		else if (sha512_transform_asm == sha512_transform_rorx)
174			pr_info("Using AVX2 optimized SHA-512 implementation\n");
175#endif
176		else
177#endif
178			pr_info("Using SSSE3 optimized SHA-512 implementation\n");
179		return crypto_register_shashes(algs, ARRAY_SIZE(algs));
180	}
181	pr_info("Neither AVX nor SSSE3 is available/usable.\n");
182
183	return -ENODEV;
184}
185
186static void __exit sha512_ssse3_mod_fini(void)
187{
188	crypto_unregister_shashes(algs, ARRAY_SIZE(algs));
189}
190
191module_init(sha512_ssse3_mod_init);
192module_exit(sha512_ssse3_mod_fini);
193
194MODULE_LICENSE("GPL");
195MODULE_DESCRIPTION("SHA512 Secure Hash Algorithm, Supplemental SSE3 accelerated");
196
197MODULE_ALIAS_CRYPTO("sha512");
198MODULE_ALIAS_CRYPTO("sha384");
199