1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
|
/*
* Copyright 2011-2018 The OpenSSL Project Authors. All Rights Reserved.
*
* Licensed under the OpenSSL license (the "License"). You may not use
* this file except in compliance with the License. You can obtain a copy
* in the file LICENSE in the source distribution or at
* https://www.openssl.org/source/license.html
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <setjmp.h>
#include <signal.h>
#include <openssl/crypto.h>
#include "internal/cryptlib.h"
#include "arm_arch.h"
unsigned int OPENSSL_armcap_P = 0;
#if __ARM_MAX_ARCH__<7
void OPENSSL_cpuid_setup(void)
{
}
uint32_t OPENSSL_rdtsc(void)
{
return 0;
}
#else
static sigset_t all_masked;
static sigjmp_buf ill_jmp;
static void ill_handler(int sig)
{
siglongjmp(ill_jmp, sig);
}
/*
* Following subroutines could have been inlined, but it's not all
* ARM compilers support inline assembler...
*/
void _armv7_neon_probe(void);
void _armv8_aes_probe(void);
void _armv8_sha1_probe(void);
void _armv8_sha256_probe(void);
void _armv8_pmull_probe(void);
# ifdef __aarch64__
void _armv8_sha512_probe(void);
# endif
uint32_t _armv7_tick(void);
uint32_t OPENSSL_rdtsc(void)
{
if (OPENSSL_armcap_P & ARMV7_TICK)
return _armv7_tick();
else
return 0;
}
# if defined(__GNUC__) && __GNUC__>=2
void OPENSSL_cpuid_setup(void) __attribute__ ((constructor));
# endif
/*
* Use a weak reference to getauxval() so we can use it if it is available but
* don't break the build if it is not.
*/
# if defined(__GNUC__) && __GNUC__>=2 && defined(__ELF__)
extern unsigned long getauxval(unsigned long type) __attribute__ ((weak));
# else
static unsigned long (*getauxval) (unsigned long) = NULL;
# endif
/*
* ARM puts the feature bits for Crypto Extensions in AT_HWCAP2, whereas
* AArch64 used AT_HWCAP.
*/
# if defined(__arm__) || defined (__arm)
# define HWCAP 16
/* AT_HWCAP */
# define HWCAP_NEON (1 << 12)
# define HWCAP_CE 26
/* AT_HWCAP2 */
# define HWCAP_CE_AES (1 << 0)
# define HWCAP_CE_PMULL (1 << 1)
# define HWCAP_CE_SHA1 (1 << 2)
# define HWCAP_CE_SHA256 (1 << 3)
# elif defined(__aarch64__)
# define HWCAP 16
/* AT_HWCAP */
# define HWCAP_NEON (1 << 1)
# define HWCAP_CE HWCAP
# define HWCAP_CE_AES (1 << 3)
# define HWCAP_CE_PMULL (1 << 4)
# define HWCAP_CE_SHA1 (1 << 5)
# define HWCAP_CE_SHA256 (1 << 6)
# define HWCAP_CE_SHA512 (1 << 21)
# endif
void OPENSSL_cpuid_setup(void)
{
const char *e;
struct sigaction ill_oact, ill_act;
sigset_t oset;
static int trigger = 0;
if (trigger)
return;
trigger = 1;
if ((e = getenv("OPENSSL_armcap"))) {
OPENSSL_armcap_P = (unsigned int)strtoul(e, NULL, 0);
return;
}
# if defined(__APPLE__) && !defined(__aarch64__)
/*
* Capability probing by catching SIGILL appears to be problematic
* on iOS. But since Apple universe is "monocultural", it's actually
* possible to simply set pre-defined processor capability mask.
*/
if (1) {
OPENSSL_armcap_P = ARMV7_NEON;
return;
}
/*
* One could do same even for __aarch64__ iOS builds. It's not done
* exclusively for reasons of keeping code unified across platforms.
* Unified code works because it never triggers SIGILL on Apple
* devices...
*/
# endif
sigfillset(&all_masked);
sigdelset(&all_masked, SIGILL);
sigdelset(&all_masked, SIGTRAP);
sigdelset(&all_masked, SIGFPE);
sigdelset(&all_masked, SIGBUS);
sigdelset(&all_masked, SIGSEGV);
OPENSSL_armcap_P = 0;
memset(&ill_act, 0, sizeof(ill_act));
ill_act.sa_handler = ill_handler;
ill_act.sa_mask = all_masked;
sigprocmask(SIG_SETMASK, &ill_act.sa_mask, &oset);
sigaction(SIGILL, &ill_act, &ill_oact);
if (getauxval != NULL) {
if (getauxval(HWCAP) & HWCAP_NEON) {
unsigned long hwcap = getauxval(HWCAP_CE);
OPENSSL_armcap_P |= ARMV7_NEON;
if (hwcap & HWCAP_CE_AES)
OPENSSL_armcap_P |= ARMV8_AES;
if (hwcap & HWCAP_CE_PMULL)
OPENSSL_armcap_P |= ARMV8_PMULL;
if (hwcap & HWCAP_CE_SHA1)
OPENSSL_armcap_P |= ARMV8_SHA1;
if (hwcap & HWCAP_CE_SHA256)
OPENSSL_armcap_P |= ARMV8_SHA256;
# ifdef __aarch64__
if (hwcap & HWCAP_CE_SHA512)
OPENSSL_armcap_P |= ARMV8_SHA512;
# endif
}
} else if (sigsetjmp(ill_jmp, 1) == 0) {
_armv7_neon_probe();
OPENSSL_armcap_P |= ARMV7_NEON;
if (sigsetjmp(ill_jmp, 1) == 0) {
_armv8_pmull_probe();
OPENSSL_armcap_P |= ARMV8_PMULL | ARMV8_AES;
} else if (sigsetjmp(ill_jmp, 1) == 0) {
_armv8_aes_probe();
OPENSSL_armcap_P |= ARMV8_AES;
}
if (sigsetjmp(ill_jmp, 1) == 0) {
_armv8_sha1_probe();
OPENSSL_armcap_P |= ARMV8_SHA1;
}
if (sigsetjmp(ill_jmp, 1) == 0) {
_armv8_sha256_probe();
OPENSSL_armcap_P |= ARMV8_SHA256;
}
# if defined(__aarch64__) && !defined(__APPLE__)
if (sigsetjmp(ill_jmp, 1) == 0) {
_armv8_sha512_probe();
OPENSSL_armcap_P |= ARMV8_SHA512;
}
# endif
}
if (sigsetjmp(ill_jmp, 1) == 0) {
_armv7_tick();
OPENSSL_armcap_P |= ARMV7_TICK;
}
sigaction(SIGILL, &ill_oact, NULL);
sigprocmask(SIG_SETMASK, &oset, NULL);
}
#endif
|