/src/mbedtls/library/common.h
Line | Count | Source (jump to first uncovered line) |
1 | | /** |
2 | | * \file common.h |
3 | | * |
4 | | * \brief Utility macros for internal use in the library |
5 | | */ |
6 | | /* |
7 | | * Copyright The Mbed TLS Contributors |
8 | | * SPDX-License-Identifier: Apache-2.0 OR GPL-2.0-or-later |
9 | | */ |
10 | | |
11 | | #ifndef MBEDTLS_LIBRARY_COMMON_H |
12 | | #define MBEDTLS_LIBRARY_COMMON_H |
13 | | |
14 | | #include "mbedtls/build_info.h" |
15 | | #include "alignment.h" |
16 | | |
17 | | #include <assert.h> |
18 | | #include <stddef.h> |
19 | | #include <stdint.h> |
20 | | #include <stddef.h> |
21 | | |
22 | | #if defined(__ARM_NEON) |
23 | | #include <arm_neon.h> |
24 | | #define MBEDTLS_HAVE_NEON_INTRINSICS |
25 | | #elif defined(MBEDTLS_PLATFORM_IS_WINDOWS_ON_ARM64) |
26 | | #include <arm64_neon.h> |
27 | | #define MBEDTLS_HAVE_NEON_INTRINSICS |
28 | | #endif |
29 | | |
30 | | /** Helper to define a function as static except when building invasive tests. |
31 | | * |
32 | | * If a function is only used inside its own source file and should be |
33 | | * declared `static` to allow the compiler to optimize for code size, |
34 | | * but that function has unit tests, define it with |
35 | | * ``` |
36 | | * MBEDTLS_STATIC_TESTABLE int mbedtls_foo(...) { ... } |
37 | | * ``` |
38 | | * and declare it in a header in the `library/` directory with |
39 | | * ``` |
40 | | * #if defined(MBEDTLS_TEST_HOOKS) |
41 | | * int mbedtls_foo(...); |
42 | | * #endif |
43 | | * ``` |
44 | | */ |
45 | | #if defined(MBEDTLS_TEST_HOOKS) |
46 | | #define MBEDTLS_STATIC_TESTABLE |
47 | | #else |
48 | | #define MBEDTLS_STATIC_TESTABLE static |
49 | | #endif |
50 | | |
51 | | #if defined(MBEDTLS_TEST_HOOKS) |
52 | | extern void (*mbedtls_test_hook_test_fail)(const char *test, int line, const char *file); |
53 | | #define MBEDTLS_TEST_HOOK_TEST_ASSERT(TEST) \ |
54 | | do { \ |
55 | | if ((!(TEST)) && ((*mbedtls_test_hook_test_fail) != NULL)) \ |
56 | | { \ |
57 | | (*mbedtls_test_hook_test_fail)( #TEST, __LINE__, __FILE__); \ |
58 | | } \ |
59 | | } while (0) |
60 | | #else |
61 | | #define MBEDTLS_TEST_HOOK_TEST_ASSERT(TEST) |
62 | | #endif /* defined(MBEDTLS_TEST_HOOKS) */ |
63 | | |
64 | | /** \def ARRAY_LENGTH |
65 | | * Return the number of elements of a static or stack array. |
66 | | * |
67 | | * \param array A value of array (not pointer) type. |
68 | | * |
69 | | * \return The number of elements of the array. |
70 | | */ |
71 | | /* A correct implementation of ARRAY_LENGTH, but which silently gives |
72 | | * a nonsensical result if called with a pointer rather than an array. */ |
73 | | #define ARRAY_LENGTH_UNSAFE(array) \ |
74 | | (sizeof(array) / sizeof(*(array))) |
75 | | |
76 | | #if defined(__GNUC__) |
77 | | /* Test if arg and &(arg)[0] have the same type. This is true if arg is |
78 | | * an array but not if it's a pointer. */ |
79 | | #define IS_ARRAY_NOT_POINTER(arg) \ |
80 | | (!__builtin_types_compatible_p(__typeof__(arg), \ |
81 | | __typeof__(&(arg)[0]))) |
82 | | /* A compile-time constant with the value 0. If `const_expr` is not a |
83 | | * compile-time constant with a nonzero value, cause a compile-time error. */ |
84 | | #define STATIC_ASSERT_EXPR(const_expr) \ |
85 | | (0 && sizeof(struct { unsigned int STATIC_ASSERT : 1 - 2 * !(const_expr); })) |
86 | | |
87 | | /* Return the scalar value `value` (possibly promoted). This is a compile-time |
88 | | * constant if `value` is. `condition` must be a compile-time constant. |
89 | | * If `condition` is false, arrange to cause a compile-time error. */ |
90 | | #define STATIC_ASSERT_THEN_RETURN(condition, value) \ |
91 | | (STATIC_ASSERT_EXPR(condition) ? 0 : (value)) |
92 | | |
93 | | #define ARRAY_LENGTH(array) \ |
94 | | (STATIC_ASSERT_THEN_RETURN(IS_ARRAY_NOT_POINTER(array), \ |
95 | | ARRAY_LENGTH_UNSAFE(array))) |
96 | | |
97 | | #else |
98 | | /* If we aren't sure the compiler supports our non-standard tricks, |
99 | | * fall back to the unsafe implementation. */ |
100 | | #define ARRAY_LENGTH(array) ARRAY_LENGTH_UNSAFE(array) |
101 | | #endif |
102 | | /** Allow library to access its structs' private members. |
103 | | * |
104 | | * Although structs defined in header files are publicly available, |
105 | | * their members are private and should not be accessed by the user. |
106 | | */ |
107 | | #define MBEDTLS_ALLOW_PRIVATE_ACCESS |
108 | | |
109 | | /** |
110 | | * \brief Securely zeroize a buffer then free it. |
111 | | * |
112 | | * Similar to making consecutive calls to |
113 | | * \c mbedtls_platform_zeroize() and \c mbedtls_free(), but has |
114 | | * code size savings, and potential for optimisation in the future. |
115 | | * |
116 | | * Guaranteed to be a no-op if \p buf is \c NULL and \p len is 0. |
117 | | * |
118 | | * \param buf Buffer to be zeroized then freed. |
119 | | * \param len Length of the buffer in bytes |
120 | | */ |
121 | | void mbedtls_zeroize_and_free(void *buf, size_t len); |
122 | | |
123 | | /** Return an offset into a buffer. |
124 | | * |
125 | | * This is just the addition of an offset to a pointer, except that this |
126 | | * function also accepts an offset of 0 into a buffer whose pointer is null. |
127 | | * (`p + n` has undefined behavior when `p` is null, even when `n == 0`. |
128 | | * A null pointer is a valid buffer pointer when the size is 0, for example |
129 | | * as the result of `malloc(0)` on some platforms.) |
130 | | * |
131 | | * \param p Pointer to a buffer of at least n bytes. |
132 | | * This may be \p NULL if \p n is zero. |
133 | | * \param n An offset in bytes. |
134 | | * \return Pointer to offset \p n in the buffer \p p. |
135 | | * Note that this is only a valid pointer if the size of the |
136 | | * buffer is at least \p n + 1. |
137 | | */ |
138 | | static inline unsigned char *mbedtls_buffer_offset( |
139 | | unsigned char *p, size_t n) |
140 | 8 | { |
141 | 8 | return p == NULL ? NULL : p + n; |
142 | 8 | } Unexecuted instantiation: platform.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_aead.c:mbedtls_buffer_offset psa_crypto_cipher.c:mbedtls_buffer_offset Line | Count | Source | 140 | 8 | { | 141 | 8 | return p == NULL ? NULL : p + n; | 142 | 8 | } |
Unexecuted instantiation: psa_crypto_client.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_driver_wrappers_no_static.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_ecp.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_ffdh.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_hash.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_mac.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_pake.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_rsa.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_slot_management.c:mbedtls_buffer_offset Unexecuted instantiation: psa_crypto_storage.c:mbedtls_buffer_offset Unexecuted instantiation: psa_its_file.c:mbedtls_buffer_offset Unexecuted instantiation: psa_util.c:mbedtls_buffer_offset Unexecuted instantiation: ripemd160.c:mbedtls_buffer_offset Unexecuted instantiation: rsa.c:mbedtls_buffer_offset Unexecuted instantiation: rsa_alt_helpers.c:mbedtls_buffer_offset Unexecuted instantiation: sha1.c:mbedtls_buffer_offset Unexecuted instantiation: sha256.c:mbedtls_buffer_offset Unexecuted instantiation: sha3.c:mbedtls_buffer_offset Unexecuted instantiation: sha512.c:mbedtls_buffer_offset Unexecuted instantiation: asn1parse.c:mbedtls_buffer_offset Unexecuted instantiation: asn1write.c:mbedtls_buffer_offset Unexecuted instantiation: bignum.c:mbedtls_buffer_offset Unexecuted instantiation: bignum_core.c:mbedtls_buffer_offset Unexecuted instantiation: ccm.c:mbedtls_buffer_offset Unexecuted instantiation: chachapoly.c:mbedtls_buffer_offset Unexecuted instantiation: cipher.c:mbedtls_buffer_offset Unexecuted instantiation: cipher_wrap.c:mbedtls_buffer_offset Unexecuted instantiation: cmac.c:mbedtls_buffer_offset Unexecuted instantiation: constant_time.c:mbedtls_buffer_offset Unexecuted instantiation: ctr_drbg.c:mbedtls_buffer_offset Unexecuted instantiation: des.c:mbedtls_buffer_offset Unexecuted instantiation: ecdh.c:mbedtls_buffer_offset Unexecuted instantiation: ecdsa.c:mbedtls_buffer_offset Unexecuted instantiation: ecjpake.c:mbedtls_buffer_offset Unexecuted instantiation: ecp.c:mbedtls_buffer_offset Unexecuted instantiation: ecp_curves.c:mbedtls_buffer_offset Unexecuted instantiation: entropy.c:mbedtls_buffer_offset Unexecuted instantiation: entropy_poll.c:mbedtls_buffer_offset Unexecuted instantiation: gcm.c:mbedtls_buffer_offset Unexecuted instantiation: hmac_drbg.c:mbedtls_buffer_offset Unexecuted instantiation: md.c:mbedtls_buffer_offset Unexecuted instantiation: md5.c:mbedtls_buffer_offset Unexecuted instantiation: nist_kw.c:mbedtls_buffer_offset Unexecuted instantiation: oid.c:mbedtls_buffer_offset Unexecuted instantiation: platform_util.c:mbedtls_buffer_offset Unexecuted instantiation: poly1305.c:mbedtls_buffer_offset Unexecuted instantiation: aes.c:mbedtls_buffer_offset Unexecuted instantiation: aesni.c:mbedtls_buffer_offset Unexecuted instantiation: aria.c:mbedtls_buffer_offset Unexecuted instantiation: camellia.c:mbedtls_buffer_offset Unexecuted instantiation: chacha20.c:mbedtls_buffer_offset Unexecuted instantiation: hkdf.c:mbedtls_buffer_offset Unexecuted instantiation: pkcs12.c:mbedtls_buffer_offset Unexecuted instantiation: pkcs5.c:mbedtls_buffer_offset |
143 | | |
144 | | /** Return an offset into a read-only buffer. |
145 | | * |
146 | | * Similar to mbedtls_buffer_offset(), but for const pointers. |
147 | | * |
148 | | * \param p Pointer to a buffer of at least n bytes. |
149 | | * This may be \p NULL if \p n is zero. |
150 | | * \param n An offset in bytes. |
151 | | * \return Pointer to offset \p n in the buffer \p p. |
152 | | * Note that this is only a valid pointer if the size of the |
153 | | * buffer is at least \p n + 1. |
154 | | */ |
155 | | static inline const unsigned char *mbedtls_buffer_offset_const( |
156 | | const unsigned char *p, size_t n) |
157 | 8 | { |
158 | 8 | return p == NULL ? NULL : p + n; |
159 | 8 | } Unexecuted instantiation: platform.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_aead.c:mbedtls_buffer_offset_const psa_crypto_cipher.c:mbedtls_buffer_offset_const Line | Count | Source | 157 | 8 | { | 158 | 8 | return p == NULL ? NULL : p + n; | 159 | 8 | } |
Unexecuted instantiation: psa_crypto_client.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_driver_wrappers_no_static.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_ecp.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_ffdh.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_hash.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_mac.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_pake.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_rsa.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_slot_management.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_crypto_storage.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_its_file.c:mbedtls_buffer_offset_const Unexecuted instantiation: psa_util.c:mbedtls_buffer_offset_const Unexecuted instantiation: ripemd160.c:mbedtls_buffer_offset_const Unexecuted instantiation: rsa.c:mbedtls_buffer_offset_const Unexecuted instantiation: rsa_alt_helpers.c:mbedtls_buffer_offset_const Unexecuted instantiation: sha1.c:mbedtls_buffer_offset_const Unexecuted instantiation: sha256.c:mbedtls_buffer_offset_const Unexecuted instantiation: sha3.c:mbedtls_buffer_offset_const Unexecuted instantiation: sha512.c:mbedtls_buffer_offset_const Unexecuted instantiation: asn1parse.c:mbedtls_buffer_offset_const Unexecuted instantiation: asn1write.c:mbedtls_buffer_offset_const Unexecuted instantiation: bignum.c:mbedtls_buffer_offset_const Unexecuted instantiation: bignum_core.c:mbedtls_buffer_offset_const Unexecuted instantiation: ccm.c:mbedtls_buffer_offset_const Unexecuted instantiation: chachapoly.c:mbedtls_buffer_offset_const Unexecuted instantiation: cipher.c:mbedtls_buffer_offset_const Unexecuted instantiation: cipher_wrap.c:mbedtls_buffer_offset_const Unexecuted instantiation: cmac.c:mbedtls_buffer_offset_const Unexecuted instantiation: constant_time.c:mbedtls_buffer_offset_const Unexecuted instantiation: ctr_drbg.c:mbedtls_buffer_offset_const Unexecuted instantiation: des.c:mbedtls_buffer_offset_const Unexecuted instantiation: ecdh.c:mbedtls_buffer_offset_const Unexecuted instantiation: ecdsa.c:mbedtls_buffer_offset_const Unexecuted instantiation: ecjpake.c:mbedtls_buffer_offset_const Unexecuted instantiation: ecp.c:mbedtls_buffer_offset_const Unexecuted instantiation: ecp_curves.c:mbedtls_buffer_offset_const Unexecuted instantiation: entropy.c:mbedtls_buffer_offset_const Unexecuted instantiation: entropy_poll.c:mbedtls_buffer_offset_const Unexecuted instantiation: gcm.c:mbedtls_buffer_offset_const Unexecuted instantiation: hmac_drbg.c:mbedtls_buffer_offset_const Unexecuted instantiation: md.c:mbedtls_buffer_offset_const Unexecuted instantiation: md5.c:mbedtls_buffer_offset_const Unexecuted instantiation: nist_kw.c:mbedtls_buffer_offset_const Unexecuted instantiation: oid.c:mbedtls_buffer_offset_const Unexecuted instantiation: platform_util.c:mbedtls_buffer_offset_const Unexecuted instantiation: poly1305.c:mbedtls_buffer_offset_const Unexecuted instantiation: aes.c:mbedtls_buffer_offset_const Unexecuted instantiation: aesni.c:mbedtls_buffer_offset_const Unexecuted instantiation: aria.c:mbedtls_buffer_offset_const Unexecuted instantiation: camellia.c:mbedtls_buffer_offset_const Unexecuted instantiation: chacha20.c:mbedtls_buffer_offset_const Unexecuted instantiation: hkdf.c:mbedtls_buffer_offset_const Unexecuted instantiation: pkcs12.c:mbedtls_buffer_offset_const Unexecuted instantiation: pkcs5.c:mbedtls_buffer_offset_const |
160 | | |
161 | | /* Always inline mbedtls_xor() for similar reasons as mbedtls_xor_no_simd(). */ |
162 | | #if defined(__IAR_SYSTEMS_ICC__) |
163 | | #pragma inline = forced |
164 | | #elif defined(__GNUC__) |
165 | | __attribute__((always_inline)) |
166 | | #endif |
167 | | /** |
168 | | * Perform a fast block XOR operation, such that |
169 | | * r[i] = a[i] ^ b[i] where 0 <= i < n |
170 | | * |
171 | | * \param r Pointer to result (buffer of at least \p n bytes). \p r |
172 | | * may be equal to either \p a or \p b, but behaviour when |
173 | | * it overlaps in other ways is undefined. |
174 | | * \param a Pointer to input (buffer of at least \p n bytes) |
175 | | * \param b Pointer to input (buffer of at least \p n bytes) |
176 | | * \param n Number of bytes to process. |
177 | | * |
178 | | * \note Depending on the situation, it may be faster to use either mbedtls_xor() or |
179 | | * mbedtls_xor_no_simd() (these are functionally equivalent). |
180 | | * If the result is used immediately after the xor operation in non-SIMD code (e.g, in |
181 | | * AES-CBC), there may be additional latency to transfer the data from SIMD to scalar |
182 | | * registers, and in this case, mbedtls_xor_no_simd() may be faster. In other cases where |
183 | | * the result is not used immediately (e.g., in AES-CTR), mbedtls_xor() may be faster. |
184 | | * For targets without SIMD support, they will behave the same. |
185 | | */ |
186 | | static inline void mbedtls_xor(unsigned char *r, |
187 | | const unsigned char *a, |
188 | | const unsigned char *b, |
189 | | size_t n) |
190 | 72.9k | { |
191 | 72.9k | size_t i = 0; |
192 | 72.9k | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) |
193 | | #if defined(MBEDTLS_HAVE_NEON_INTRINSICS) && \ |
194 | | (!(defined(MBEDTLS_COMPILER_IS_GCC) && MBEDTLS_GCC_VERSION < 70300)) |
195 | | /* Old GCC versions generate a warning here, so disable the NEON path for these compilers */ |
196 | | for (; (i + 16) <= n; i += 16) { |
197 | | uint8x16_t v1 = vld1q_u8(a + i); |
198 | | uint8x16_t v2 = vld1q_u8(b + i); |
199 | | uint8x16_t x = veorq_u8(v1, v2); |
200 | | vst1q_u8(r + i, x); |
201 | | } |
202 | | #if defined(__IAR_SYSTEMS_ICC__) |
203 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case |
204 | | * where n is a constant multiple of 16. |
205 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time |
206 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ |
207 | | if (n % 16 == 0) { |
208 | | return; |
209 | | } |
210 | | #endif |
211 | | #elif defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) |
212 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ |
213 | 344k | for (; (i + 8) <= n; i += 8) { |
214 | 271k | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); |
215 | 271k | mbedtls_put_unaligned_uint64(r + i, x); |
216 | 271k | } |
217 | | #if defined(__IAR_SYSTEMS_ICC__) |
218 | | if (n % 8 == 0) { |
219 | | return; |
220 | | } |
221 | | #endif |
222 | | #else |
223 | | for (; (i + 4) <= n; i += 4) { |
224 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); |
225 | | mbedtls_put_unaligned_uint32(r + i, x); |
226 | | } |
227 | | #if defined(__IAR_SYSTEMS_ICC__) |
228 | | if (n % 4 == 0) { |
229 | | return; |
230 | | } |
231 | | #endif |
232 | | #endif |
233 | 72.9k | #endif |
234 | 177k | for (; i < n; i++) { |
235 | 104k | r[i] = a[i] ^ b[i]; |
236 | 104k | } |
237 | 72.9k | } Unexecuted instantiation: platform.c:mbedtls_xor Unexecuted instantiation: psa_crypto.c:mbedtls_xor Unexecuted instantiation: psa_crypto_aead.c:mbedtls_xor Unexecuted instantiation: psa_crypto_cipher.c:mbedtls_xor Unexecuted instantiation: psa_crypto_client.c:mbedtls_xor Unexecuted instantiation: psa_crypto_driver_wrappers_no_static.c:mbedtls_xor Unexecuted instantiation: psa_crypto_ecp.c:mbedtls_xor Unexecuted instantiation: psa_crypto_ffdh.c:mbedtls_xor Unexecuted instantiation: psa_crypto_hash.c:mbedtls_xor Unexecuted instantiation: psa_crypto_mac.c:mbedtls_xor Unexecuted instantiation: psa_crypto_pake.c:mbedtls_xor Unexecuted instantiation: psa_crypto_rsa.c:mbedtls_xor Unexecuted instantiation: psa_crypto_slot_management.c:mbedtls_xor Unexecuted instantiation: psa_crypto_storage.c:mbedtls_xor Unexecuted instantiation: psa_its_file.c:mbedtls_xor Unexecuted instantiation: psa_util.c:mbedtls_xor Unexecuted instantiation: ripemd160.c:mbedtls_xor Unexecuted instantiation: rsa.c:mbedtls_xor Unexecuted instantiation: rsa_alt_helpers.c:mbedtls_xor Unexecuted instantiation: sha1.c:mbedtls_xor Unexecuted instantiation: sha256.c:mbedtls_xor Unexecuted instantiation: sha3.c:mbedtls_xor Unexecuted instantiation: sha512.c:mbedtls_xor Unexecuted instantiation: asn1parse.c:mbedtls_xor Unexecuted instantiation: asn1write.c:mbedtls_xor Unexecuted instantiation: bignum.c:mbedtls_xor Unexecuted instantiation: bignum_core.c:mbedtls_xor Line | Count | Source | 190 | 52 | { | 191 | 52 | size_t i = 0; | 192 | 52 | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 193 | | #if defined(MBEDTLS_HAVE_NEON_INTRINSICS) && \ | 194 | | (!(defined(MBEDTLS_COMPILER_IS_GCC) && MBEDTLS_GCC_VERSION < 70300)) | 195 | | /* Old GCC versions generate a warning here, so disable the NEON path for these compilers */ | 196 | | for (; (i + 16) <= n; i += 16) { | 197 | | uint8x16_t v1 = vld1q_u8(a + i); | 198 | | uint8x16_t v2 = vld1q_u8(b + i); | 199 | | uint8x16_t x = veorq_u8(v1, v2); | 200 | | vst1q_u8(r + i, x); | 201 | | } | 202 | | #if defined(__IAR_SYSTEMS_ICC__) | 203 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 204 | | * where n is a constant multiple of 16. | 205 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 206 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 207 | | if (n % 16 == 0) { | 208 | | return; | 209 | | } | 210 | | #endif | 211 | | #elif defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 212 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 213 | 126 | for (; (i + 8) <= n; i += 8) { | 214 | 74 | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 215 | 74 | mbedtls_put_unaligned_uint64(r + i, x); | 216 | 74 | } | 217 | | #if defined(__IAR_SYSTEMS_ICC__) | 218 | | if (n % 8 == 0) { | 219 | | return; | 220 | | } | 221 | | #endif | 222 | | #else | 223 | | for (; (i + 4) <= n; i += 4) { | 224 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 225 | | mbedtls_put_unaligned_uint32(r + i, x); | 226 | | } | 227 | | #if defined(__IAR_SYSTEMS_ICC__) | 228 | | if (n % 4 == 0) { | 229 | | return; | 230 | | } | 231 | | #endif | 232 | | #endif | 233 | 52 | #endif | 234 | 74 | for (; i < n; i++) { | 235 | 22 | r[i] = a[i] ^ b[i]; | 236 | 22 | } | 237 | 52 | } |
Unexecuted instantiation: chachapoly.c:mbedtls_xor Unexecuted instantiation: cipher.c:mbedtls_xor Unexecuted instantiation: cipher_wrap.c:mbedtls_xor Line | Count | Source | 190 | 4 | { | 191 | 4 | size_t i = 0; | 192 | 4 | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 193 | | #if defined(MBEDTLS_HAVE_NEON_INTRINSICS) && \ | 194 | | (!(defined(MBEDTLS_COMPILER_IS_GCC) && MBEDTLS_GCC_VERSION < 70300)) | 195 | | /* Old GCC versions generate a warning here, so disable the NEON path for these compilers */ | 196 | | for (; (i + 16) <= n; i += 16) { | 197 | | uint8x16_t v1 = vld1q_u8(a + i); | 198 | | uint8x16_t v2 = vld1q_u8(b + i); | 199 | | uint8x16_t x = veorq_u8(v1, v2); | 200 | | vst1q_u8(r + i, x); | 201 | | } | 202 | | #if defined(__IAR_SYSTEMS_ICC__) | 203 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 204 | | * where n is a constant multiple of 16. | 205 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 206 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 207 | | if (n % 16 == 0) { | 208 | | return; | 209 | | } | 210 | | #endif | 211 | | #elif defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 212 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 213 | 12 | for (; (i + 8) <= n; i += 8) { | 214 | 8 | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 215 | 8 | mbedtls_put_unaligned_uint64(r + i, x); | 216 | 8 | } | 217 | | #if defined(__IAR_SYSTEMS_ICC__) | 218 | | if (n % 8 == 0) { | 219 | | return; | 220 | | } | 221 | | #endif | 222 | | #else | 223 | | for (; (i + 4) <= n; i += 4) { | 224 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 225 | | mbedtls_put_unaligned_uint32(r + i, x); | 226 | | } | 227 | | #if defined(__IAR_SYSTEMS_ICC__) | 228 | | if (n % 4 == 0) { | 229 | | return; | 230 | | } | 231 | | #endif | 232 | | #endif | 233 | 4 | #endif | 234 | 4 | for (; i < n; i++) { | 235 | 0 | r[i] = a[i] ^ b[i]; | 236 | 0 | } | 237 | 4 | } |
Unexecuted instantiation: constant_time.c:mbedtls_xor Line | Count | Source | 190 | 160 | { | 191 | 160 | size_t i = 0; | 192 | 160 | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 193 | | #if defined(MBEDTLS_HAVE_NEON_INTRINSICS) && \ | 194 | | (!(defined(MBEDTLS_COMPILER_IS_GCC) && MBEDTLS_GCC_VERSION < 70300)) | 195 | | /* Old GCC versions generate a warning here, so disable the NEON path for these compilers */ | 196 | | for (; (i + 16) <= n; i += 16) { | 197 | | uint8x16_t v1 = vld1q_u8(a + i); | 198 | | uint8x16_t v2 = vld1q_u8(b + i); | 199 | | uint8x16_t x = veorq_u8(v1, v2); | 200 | | vst1q_u8(r + i, x); | 201 | | } | 202 | | #if defined(__IAR_SYSTEMS_ICC__) | 203 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 204 | | * where n is a constant multiple of 16. | 205 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 206 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 207 | | if (n % 16 == 0) { | 208 | | return; | 209 | | } | 210 | | #endif | 211 | | #elif defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 212 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 213 | 520 | for (; (i + 8) <= n; i += 8) { | 214 | 360 | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 215 | 360 | mbedtls_put_unaligned_uint64(r + i, x); | 216 | 360 | } | 217 | | #if defined(__IAR_SYSTEMS_ICC__) | 218 | | if (n % 8 == 0) { | 219 | | return; | 220 | | } | 221 | | #endif | 222 | | #else | 223 | | for (; (i + 4) <= n; i += 4) { | 224 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 225 | | mbedtls_put_unaligned_uint32(r + i, x); | 226 | | } | 227 | | #if defined(__IAR_SYSTEMS_ICC__) | 228 | | if (n % 4 == 0) { | 229 | | return; | 230 | | } | 231 | | #endif | 232 | | #endif | 233 | 160 | #endif | 234 | 160 | for (; i < n; i++) { | 235 | 0 | r[i] = a[i] ^ b[i]; | 236 | 0 | } | 237 | 160 | } |
Line | Count | Source | 190 | 8 | { | 191 | 8 | size_t i = 0; | 192 | 8 | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 193 | | #if defined(MBEDTLS_HAVE_NEON_INTRINSICS) && \ | 194 | | (!(defined(MBEDTLS_COMPILER_IS_GCC) && MBEDTLS_GCC_VERSION < 70300)) | 195 | | /* Old GCC versions generate a warning here, so disable the NEON path for these compilers */ | 196 | | for (; (i + 16) <= n; i += 16) { | 197 | | uint8x16_t v1 = vld1q_u8(a + i); | 198 | | uint8x16_t v2 = vld1q_u8(b + i); | 199 | | uint8x16_t x = veorq_u8(v1, v2); | 200 | | vst1q_u8(r + i, x); | 201 | | } | 202 | | #if defined(__IAR_SYSTEMS_ICC__) | 203 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 204 | | * where n is a constant multiple of 16. | 205 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 206 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 207 | | if (n % 16 == 0) { | 208 | | return; | 209 | | } | 210 | | #endif | 211 | | #elif defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 212 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 213 | 16 | for (; (i + 8) <= n; i += 8) { | 214 | 8 | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 215 | 8 | mbedtls_put_unaligned_uint64(r + i, x); | 216 | 8 | } | 217 | | #if defined(__IAR_SYSTEMS_ICC__) | 218 | | if (n % 8 == 0) { | 219 | | return; | 220 | | } | 221 | | #endif | 222 | | #else | 223 | | for (; (i + 4) <= n; i += 4) { | 224 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 225 | | mbedtls_put_unaligned_uint32(r + i, x); | 226 | | } | 227 | | #if defined(__IAR_SYSTEMS_ICC__) | 228 | | if (n % 4 == 0) { | 229 | | return; | 230 | | } | 231 | | #endif | 232 | | #endif | 233 | 8 | #endif | 234 | 8 | for (; i < n; i++) { | 235 | 0 | r[i] = a[i] ^ b[i]; | 236 | 0 | } | 237 | 8 | } |
Unexecuted instantiation: ecdh.c:mbedtls_xor Unexecuted instantiation: ecdsa.c:mbedtls_xor Unexecuted instantiation: ecjpake.c:mbedtls_xor Unexecuted instantiation: ecp.c:mbedtls_xor Unexecuted instantiation: ecp_curves.c:mbedtls_xor Unexecuted instantiation: entropy.c:mbedtls_xor Unexecuted instantiation: entropy_poll.c:mbedtls_xor Line | Count | Source | 190 | 13.6k | { | 191 | 13.6k | size_t i = 0; | 192 | 13.6k | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 193 | | #if defined(MBEDTLS_HAVE_NEON_INTRINSICS) && \ | 194 | | (!(defined(MBEDTLS_COMPILER_IS_GCC) && MBEDTLS_GCC_VERSION < 70300)) | 195 | | /* Old GCC versions generate a warning here, so disable the NEON path for these compilers */ | 196 | | for (; (i + 16) <= n; i += 16) { | 197 | | uint8x16_t v1 = vld1q_u8(a + i); | 198 | | uint8x16_t v2 = vld1q_u8(b + i); | 199 | | uint8x16_t x = veorq_u8(v1, v2); | 200 | | vst1q_u8(r + i, x); | 201 | | } | 202 | | #if defined(__IAR_SYSTEMS_ICC__) | 203 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 204 | | * where n is a constant multiple of 16. | 205 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 206 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 207 | | if (n % 16 == 0) { | 208 | | return; | 209 | | } | 210 | | #endif | 211 | | #elif defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 212 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 213 | 40.5k | for (; (i + 8) <= n; i += 8) { | 214 | 26.8k | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 215 | 26.8k | mbedtls_put_unaligned_uint64(r + i, x); | 216 | 26.8k | } | 217 | | #if defined(__IAR_SYSTEMS_ICC__) | 218 | | if (n % 8 == 0) { | 219 | | return; | 220 | | } | 221 | | #endif | 222 | | #else | 223 | | for (; (i + 4) <= n; i += 4) { | 224 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 225 | | mbedtls_put_unaligned_uint32(r + i, x); | 226 | | } | 227 | | #if defined(__IAR_SYSTEMS_ICC__) | 228 | | if (n % 4 == 0) { | 229 | | return; | 230 | | } | 231 | | #endif | 232 | | #endif | 233 | 13.6k | #endif | 234 | 15.0k | for (; i < n; i++) { | 235 | 1.38k | r[i] = a[i] ^ b[i]; | 236 | 1.38k | } | 237 | 13.6k | } |
Unexecuted instantiation: hmac_drbg.c:mbedtls_xor Line | Count | Source | 190 | 48.5k | { | 191 | 48.5k | size_t i = 0; | 192 | 48.5k | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 193 | | #if defined(MBEDTLS_HAVE_NEON_INTRINSICS) && \ | 194 | | (!(defined(MBEDTLS_COMPILER_IS_GCC) && MBEDTLS_GCC_VERSION < 70300)) | 195 | | /* Old GCC versions generate a warning here, so disable the NEON path for these compilers */ | 196 | | for (; (i + 16) <= n; i += 16) { | 197 | | uint8x16_t v1 = vld1q_u8(a + i); | 198 | | uint8x16_t v2 = vld1q_u8(b + i); | 199 | | uint8x16_t x = veorq_u8(v1, v2); | 200 | | vst1q_u8(r + i, x); | 201 | | } | 202 | | #if defined(__IAR_SYSTEMS_ICC__) | 203 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 204 | | * where n is a constant multiple of 16. | 205 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 206 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 207 | | if (n % 16 == 0) { | 208 | | return; | 209 | | } | 210 | | #endif | 211 | | #elif defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 212 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 213 | 259k | for (; (i + 8) <= n; i += 8) { | 214 | 211k | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 215 | 211k | mbedtls_put_unaligned_uint64(r + i, x); | 216 | 211k | } | 217 | | #if defined(__IAR_SYSTEMS_ICC__) | 218 | | if (n % 8 == 0) { | 219 | | return; | 220 | | } | 221 | | #endif | 222 | | #else | 223 | | for (; (i + 4) <= n; i += 4) { | 224 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 225 | | mbedtls_put_unaligned_uint32(r + i, x); | 226 | | } | 227 | | #if defined(__IAR_SYSTEMS_ICC__) | 228 | | if (n % 4 == 0) { | 229 | | return; | 230 | | } | 231 | | #endif | 232 | | #endif | 233 | 48.5k | #endif | 234 | 129k | for (; i < n; i++) { | 235 | 80.8k | r[i] = a[i] ^ b[i]; | 236 | 80.8k | } | 237 | 48.5k | } |
Unexecuted instantiation: md5.c:mbedtls_xor Unexecuted instantiation: nist_kw.c:mbedtls_xor Unexecuted instantiation: oid.c:mbedtls_xor Unexecuted instantiation: platform_util.c:mbedtls_xor Unexecuted instantiation: poly1305.c:mbedtls_xor Unexecuted instantiation: aes.c:mbedtls_xor Unexecuted instantiation: aesni.c:mbedtls_xor Unexecuted instantiation: aria.c:mbedtls_xor Unexecuted instantiation: camellia.c:mbedtls_xor Line | Count | Source | 190 | 88 | { | 191 | 88 | size_t i = 0; | 192 | 88 | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 193 | | #if defined(MBEDTLS_HAVE_NEON_INTRINSICS) && \ | 194 | | (!(defined(MBEDTLS_COMPILER_IS_GCC) && MBEDTLS_GCC_VERSION < 70300)) | 195 | | /* Old GCC versions generate a warning here, so disable the NEON path for these compilers */ | 196 | | for (; (i + 16) <= n; i += 16) { | 197 | | uint8x16_t v1 = vld1q_u8(a + i); | 198 | | uint8x16_t v2 = vld1q_u8(b + i); | 199 | | uint8x16_t x = veorq_u8(v1, v2); | 200 | | vst1q_u8(r + i, x); | 201 | | } | 202 | | #if defined(__IAR_SYSTEMS_ICC__) | 203 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 204 | | * where n is a constant multiple of 16. | 205 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 206 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 207 | | if (n % 16 == 0) { | 208 | | return; | 209 | | } | 210 | | #endif | 211 | | #elif defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 212 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 213 | 615 | for (; (i + 8) <= n; i += 8) { | 214 | 527 | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 215 | 527 | mbedtls_put_unaligned_uint64(r + i, x); | 216 | 527 | } | 217 | | #if defined(__IAR_SYSTEMS_ICC__) | 218 | | if (n % 8 == 0) { | 219 | | return; | 220 | | } | 221 | | #endif | 222 | | #else | 223 | | for (; (i + 4) <= n; i += 4) { | 224 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 225 | | mbedtls_put_unaligned_uint32(r + i, x); | 226 | | } | 227 | | #if defined(__IAR_SYSTEMS_ICC__) | 228 | | if (n % 4 == 0) { | 229 | | return; | 230 | | } | 231 | | #endif | 232 | | #endif | 233 | 88 | #endif | 234 | 273 | for (; i < n; i++) { | 235 | 185 | r[i] = a[i] ^ b[i]; | 236 | 185 | } | 237 | 88 | } |
Unexecuted instantiation: hkdf.c:mbedtls_xor Unexecuted instantiation: pkcs12.c:mbedtls_xor Line | Count | Source | 190 | 10.4k | { | 191 | 10.4k | size_t i = 0; | 192 | 10.4k | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 193 | | #if defined(MBEDTLS_HAVE_NEON_INTRINSICS) && \ | 194 | | (!(defined(MBEDTLS_COMPILER_IS_GCC) && MBEDTLS_GCC_VERSION < 70300)) | 195 | | /* Old GCC versions generate a warning here, so disable the NEON path for these compilers */ | 196 | | for (; (i + 16) <= n; i += 16) { | 197 | | uint8x16_t v1 = vld1q_u8(a + i); | 198 | | uint8x16_t v2 = vld1q_u8(b + i); | 199 | | uint8x16_t x = veorq_u8(v1, v2); | 200 | | vst1q_u8(r + i, x); | 201 | | } | 202 | | #if defined(__IAR_SYSTEMS_ICC__) | 203 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 204 | | * where n is a constant multiple of 16. | 205 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 206 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 207 | | if (n % 16 == 0) { | 208 | | return; | 209 | | } | 210 | | #endif | 211 | | #elif defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 212 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 213 | 42.3k | for (; (i + 8) <= n; i += 8) { | 214 | 31.8k | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 215 | 31.8k | mbedtls_put_unaligned_uint64(r + i, x); | 216 | 31.8k | } | 217 | | #if defined(__IAR_SYSTEMS_ICC__) | 218 | | if (n % 8 == 0) { | 219 | | return; | 220 | | } | 221 | | #endif | 222 | | #else | 223 | | for (; (i + 4) <= n; i += 4) { | 224 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 225 | | mbedtls_put_unaligned_uint32(r + i, x); | 226 | | } | 227 | | #if defined(__IAR_SYSTEMS_ICC__) | 228 | | if (n % 4 == 0) { | 229 | | return; | 230 | | } | 231 | | #endif | 232 | | #endif | 233 | 10.4k | #endif | 234 | 32.8k | for (; i < n; i++) { | 235 | 22.3k | r[i] = a[i] ^ b[i]; | 236 | 22.3k | } | 237 | 10.4k | } |
|
238 | | |
239 | | /* Always inline mbedtls_xor_no_simd() as we see significant perf regressions when it does not get |
240 | | * inlined (e.g., observed about 3x perf difference in gcm_mult_largetable with gcc 7 - 12) */ |
241 | | #if defined(__IAR_SYSTEMS_ICC__) |
242 | | #pragma inline = forced |
243 | | #elif defined(__GNUC__) |
244 | | __attribute__((always_inline)) |
245 | | #endif |
246 | | /** |
247 | | * Perform a fast block XOR operation, such that |
248 | | * r[i] = a[i] ^ b[i] where 0 <= i < n |
249 | | * |
250 | | * In some situations, this can perform better than mbedtls_xor() (e.g., it's about 5% |
251 | | * better in AES-CBC). |
252 | | * |
253 | | * \param r Pointer to result (buffer of at least \p n bytes). \p r |
254 | | * may be equal to either \p a or \p b, but behaviour when |
255 | | * it overlaps in other ways is undefined. |
256 | | * \param a Pointer to input (buffer of at least \p n bytes) |
257 | | * \param b Pointer to input (buffer of at least \p n bytes) |
258 | | * \param n Number of bytes to process. |
259 | | * |
260 | | * \note Depending on the situation, it may be faster to use either mbedtls_xor() or |
261 | | * mbedtls_xor_no_simd() (these are functionally equivalent). |
262 | | * If the result is used immediately after the xor operation in non-SIMD code (e.g, in |
263 | | * AES-CBC), there may be additional latency to transfer the data from SIMD to scalar |
264 | | * registers, and in this case, mbedtls_xor_no_simd() may be faster. In other cases where |
265 | | * the result is not used immediately (e.g., in AES-CTR), mbedtls_xor() may be faster. |
266 | | * For targets without SIMD support, they will behave the same. |
267 | | */ |
268 | | static inline void mbedtls_xor_no_simd(unsigned char *r, |
269 | | const unsigned char *a, |
270 | | const unsigned char *b, |
271 | | size_t n) |
272 | 158 | { |
273 | 158 | size_t i = 0; |
274 | 158 | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) |
275 | 158 | #if defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) |
276 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ |
277 | 474 | for (; (i + 8) <= n; i += 8) { |
278 | 316 | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); |
279 | 316 | mbedtls_put_unaligned_uint64(r + i, x); |
280 | 316 | } |
281 | | #if defined(__IAR_SYSTEMS_ICC__) |
282 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case |
283 | | * where n is a constant multiple of 8. |
284 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time |
285 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ |
286 | | if (n % 8 == 0) { |
287 | | return; |
288 | | } |
289 | | #endif |
290 | | #else |
291 | | for (; (i + 4) <= n; i += 4) { |
292 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); |
293 | | mbedtls_put_unaligned_uint32(r + i, x); |
294 | | } |
295 | | #if defined(__IAR_SYSTEMS_ICC__) |
296 | | if (n % 4 == 0) { |
297 | | return; |
298 | | } |
299 | | #endif |
300 | | #endif |
301 | 158 | #endif |
302 | 158 | for (; i < n; i++) { |
303 | 0 | r[i] = a[i] ^ b[i]; |
304 | 0 | } |
305 | 158 | } Unexecuted instantiation: platform.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_aead.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_cipher.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_client.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_driver_wrappers_no_static.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_ecp.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_ffdh.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_hash.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_mac.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_pake.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_rsa.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_slot_management.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_crypto_storage.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_its_file.c:mbedtls_xor_no_simd Unexecuted instantiation: psa_util.c:mbedtls_xor_no_simd Unexecuted instantiation: ripemd160.c:mbedtls_xor_no_simd Unexecuted instantiation: rsa.c:mbedtls_xor_no_simd Unexecuted instantiation: rsa_alt_helpers.c:mbedtls_xor_no_simd Unexecuted instantiation: sha1.c:mbedtls_xor_no_simd Unexecuted instantiation: sha256.c:mbedtls_xor_no_simd Unexecuted instantiation: sha3.c:mbedtls_xor_no_simd Unexecuted instantiation: sha512.c:mbedtls_xor_no_simd Unexecuted instantiation: asn1parse.c:mbedtls_xor_no_simd Unexecuted instantiation: asn1write.c:mbedtls_xor_no_simd Unexecuted instantiation: bignum.c:mbedtls_xor_no_simd Unexecuted instantiation: bignum_core.c:mbedtls_xor_no_simd Unexecuted instantiation: ccm.c:mbedtls_xor_no_simd Unexecuted instantiation: chachapoly.c:mbedtls_xor_no_simd Unexecuted instantiation: cipher.c:mbedtls_xor_no_simd Unexecuted instantiation: cipher_wrap.c:mbedtls_xor_no_simd cmac.c:mbedtls_xor_no_simd Line | Count | Source | 272 | 156 | { | 273 | 156 | size_t i = 0; | 274 | 156 | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 275 | 156 | #if defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 276 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 277 | 468 | for (; (i + 8) <= n; i += 8) { | 278 | 312 | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 279 | 312 | mbedtls_put_unaligned_uint64(r + i, x); | 280 | 312 | } | 281 | | #if defined(__IAR_SYSTEMS_ICC__) | 282 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 283 | | * where n is a constant multiple of 8. | 284 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 285 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 286 | | if (n % 8 == 0) { | 287 | | return; | 288 | | } | 289 | | #endif | 290 | | #else | 291 | | for (; (i + 4) <= n; i += 4) { | 292 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 293 | | mbedtls_put_unaligned_uint32(r + i, x); | 294 | | } | 295 | | #if defined(__IAR_SYSTEMS_ICC__) | 296 | | if (n % 4 == 0) { | 297 | | return; | 298 | | } | 299 | | #endif | 300 | | #endif | 301 | 156 | #endif | 302 | 156 | for (; i < n; i++) { | 303 | 0 | r[i] = a[i] ^ b[i]; | 304 | 0 | } | 305 | 156 | } |
Unexecuted instantiation: constant_time.c:mbedtls_xor_no_simd Unexecuted instantiation: ctr_drbg.c:mbedtls_xor_no_simd Unexecuted instantiation: des.c:mbedtls_xor_no_simd Unexecuted instantiation: ecdh.c:mbedtls_xor_no_simd Unexecuted instantiation: ecdsa.c:mbedtls_xor_no_simd Unexecuted instantiation: ecjpake.c:mbedtls_xor_no_simd Unexecuted instantiation: ecp.c:mbedtls_xor_no_simd Unexecuted instantiation: ecp_curves.c:mbedtls_xor_no_simd Unexecuted instantiation: entropy.c:mbedtls_xor_no_simd Unexecuted instantiation: entropy_poll.c:mbedtls_xor_no_simd Unexecuted instantiation: gcm.c:mbedtls_xor_no_simd Unexecuted instantiation: hmac_drbg.c:mbedtls_xor_no_simd Unexecuted instantiation: md.c:mbedtls_xor_no_simd Unexecuted instantiation: md5.c:mbedtls_xor_no_simd Unexecuted instantiation: nist_kw.c:mbedtls_xor_no_simd Unexecuted instantiation: oid.c:mbedtls_xor_no_simd Unexecuted instantiation: platform_util.c:mbedtls_xor_no_simd Unexecuted instantiation: poly1305.c:mbedtls_xor_no_simd aes.c:mbedtls_xor_no_simd Line | Count | Source | 272 | 2 | { | 273 | 2 | size_t i = 0; | 274 | 2 | #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS) | 275 | 2 | #if defined(MBEDTLS_ARCH_IS_X64) || defined(MBEDTLS_ARCH_IS_ARM64) | 276 | | /* This codepath probably only makes sense on architectures with 64-bit registers */ | 277 | 6 | for (; (i + 8) <= n; i += 8) { | 278 | 4 | uint64_t x = mbedtls_get_unaligned_uint64(a + i) ^ mbedtls_get_unaligned_uint64(b + i); | 279 | 4 | mbedtls_put_unaligned_uint64(r + i, x); | 280 | 4 | } | 281 | | #if defined(__IAR_SYSTEMS_ICC__) | 282 | | /* This if statement helps some compilers (e.g., IAR) optimise out the byte-by-byte tail case | 283 | | * where n is a constant multiple of 8. | 284 | | * For other compilers (e.g. recent gcc and clang) it makes no difference if n is a compile-time | 285 | | * constant, and is a very small perf regression if n is not a compile-time constant. */ | 286 | | if (n % 8 == 0) { | 287 | | return; | 288 | | } | 289 | | #endif | 290 | | #else | 291 | | for (; (i + 4) <= n; i += 4) { | 292 | | uint32_t x = mbedtls_get_unaligned_uint32(a + i) ^ mbedtls_get_unaligned_uint32(b + i); | 293 | | mbedtls_put_unaligned_uint32(r + i, x); | 294 | | } | 295 | | #if defined(__IAR_SYSTEMS_ICC__) | 296 | | if (n % 4 == 0) { | 297 | | return; | 298 | | } | 299 | | #endif | 300 | | #endif | 301 | 2 | #endif | 302 | 2 | for (; i < n; i++) { | 303 | 0 | r[i] = a[i] ^ b[i]; | 304 | 0 | } | 305 | 2 | } |
Unexecuted instantiation: aesni.c:mbedtls_xor_no_simd Unexecuted instantiation: aria.c:mbedtls_xor_no_simd Unexecuted instantiation: camellia.c:mbedtls_xor_no_simd Unexecuted instantiation: chacha20.c:mbedtls_xor_no_simd Unexecuted instantiation: hkdf.c:mbedtls_xor_no_simd Unexecuted instantiation: pkcs12.c:mbedtls_xor_no_simd Unexecuted instantiation: pkcs5.c:mbedtls_xor_no_simd |
306 | | |
307 | | /* Fix MSVC C99 compatible issue |
308 | | * MSVC support __func__ from visual studio 2015( 1900 ) |
309 | | * Use MSVC predefine macro to avoid name check fail. |
310 | | */ |
311 | | #if (defined(_MSC_VER) && (_MSC_VER <= 1900)) |
312 | | #define /*no-check-names*/ __func__ __FUNCTION__ |
313 | | #endif |
314 | | |
315 | | /* Define `asm` for compilers which don't define it. */ |
316 | | /* *INDENT-OFF* */ |
317 | | #ifndef asm |
318 | | #if defined(__IAR_SYSTEMS_ICC__) |
319 | | #define asm __asm |
320 | | #else |
321 | 1.91G | #define asm __asm__ |
322 | | #endif |
323 | | #endif |
324 | | /* *INDENT-ON* */ |
325 | | |
326 | | /* |
327 | | * Define the constraint used for read-only pointer operands to aarch64 asm. |
328 | | * |
329 | | * This is normally the usual "r", but for aarch64_32 (aka ILP32, |
330 | | * as found in watchos), "p" is required to avoid warnings from clang. |
331 | | * |
332 | | * Note that clang does not recognise '+p' or '=p', and armclang |
333 | | * does not recognise 'p' at all. Therefore, to update a pointer from |
334 | | * aarch64 assembly, it is necessary to use something like: |
335 | | * |
336 | | * uintptr_t uptr = (uintptr_t) ptr; |
337 | | * asm( "ldr x4, [%x0], #8" ... : "+r" (uptr) : : ) |
338 | | * ptr = (void*) uptr; |
339 | | * |
340 | | * Note that the "x" in "%x0" is neccessary; writing "%0" will cause warnings. |
341 | | */ |
342 | | #if defined(__aarch64__) && defined(MBEDTLS_HAVE_ASM) |
343 | | #if UINTPTR_MAX == 0xfffffffful |
344 | | /* ILP32: Specify the pointer operand slightly differently, as per #7787. */ |
345 | | #define MBEDTLS_ASM_AARCH64_PTR_CONSTRAINT "p" |
346 | | #elif UINTPTR_MAX == 0xfffffffffffffffful |
347 | | /* Normal case (64-bit pointers): use "r" as the constraint for pointer operands to asm */ |
348 | | #define MBEDTLS_ASM_AARCH64_PTR_CONSTRAINT "r" |
349 | | #else |
350 | | #error "Unrecognised pointer size for aarch64" |
351 | | #endif |
352 | | #endif |
353 | | |
354 | | /* Always provide a static assert macro, so it can be used unconditionally. |
355 | | * It does nothing on systems where we don't know how to define a static assert. |
356 | | */ |
357 | | /* Can't use the C11-style `defined(static_assert)` on FreeBSD, since it |
358 | | * defines static_assert even with -std=c99, but then complains about it. |
359 | | */ |
360 | | #if defined(static_assert) && !defined(__FreeBSD__) |
361 | | #define MBEDTLS_STATIC_ASSERT(expr, msg) static_assert(expr, msg) |
362 | | #else |
363 | | /* Make sure `MBEDTLS_STATIC_ASSERT(expr, msg);` is valid both inside and |
364 | | * outside a function. We choose a struct declaration, which can be repeated |
365 | | * any number of times and does not need a matching definition. */ |
366 | | #define MBEDTLS_STATIC_ASSERT(expr, msg) \ |
367 | | struct ISO_C_does_not_allow_extra_semicolon_outside_of_a_function |
368 | | #endif |
369 | | |
370 | | #if defined(__has_builtin) |
371 | | #define MBEDTLS_HAS_BUILTIN(x) __has_builtin(x) |
372 | | #else |
373 | | #define MBEDTLS_HAS_BUILTIN(x) 0 |
374 | | #endif |
375 | | |
376 | | /* Define compiler branch hints */ |
377 | | #if MBEDTLS_HAS_BUILTIN(__builtin_expect) |
378 | | #define MBEDTLS_LIKELY(x) __builtin_expect(!!(x), 1) |
379 | 0 | #define MBEDTLS_UNLIKELY(x) __builtin_expect(!!(x), 0) |
380 | | #else |
381 | | #define MBEDTLS_LIKELY(x) x |
382 | | #define MBEDTLS_UNLIKELY(x) x |
383 | | #endif |
384 | | |
385 | | /* MBEDTLS_ASSUME may be used to provide additional information to the compiler |
386 | | * which can result in smaller code-size. */ |
387 | | #if MBEDTLS_HAS_BUILTIN(__builtin_assume) |
388 | | /* clang provides __builtin_assume */ |
389 | 41 | #define MBEDTLS_ASSUME(x) __builtin_assume(x) |
390 | | #elif MBEDTLS_HAS_BUILTIN(__builtin_unreachable) |
391 | | /* gcc and IAR can use __builtin_unreachable */ |
392 | | #define MBEDTLS_ASSUME(x) do { if (!(x)) __builtin_unreachable(); } while (0) |
393 | | #elif defined(_MSC_VER) |
394 | | /* Supported by MSVC since VS 2005 */ |
395 | | #define MBEDTLS_ASSUME(x) __assume(x) |
396 | | #else |
397 | | #define MBEDTLS_ASSUME(x) do { } while (0) |
398 | | #endif |
399 | | |
400 | | /* For gcc -Os, override with -O2 for a given function. |
401 | | * |
402 | | * This will not affect behaviour for other optimisation settings, e.g. -O0. |
403 | | */ |
404 | | #if defined(MBEDTLS_COMPILER_IS_GCC) && defined(__OPTIMIZE_SIZE__) |
405 | | #define MBEDTLS_OPTIMIZE_FOR_PERFORMANCE __attribute__((optimize("-O2"))) |
406 | | #else |
407 | | #define MBEDTLS_OPTIMIZE_FOR_PERFORMANCE |
408 | | #endif |
409 | | |
410 | | /* Suppress compiler warnings for unused functions and variables. */ |
411 | | #if !defined(MBEDTLS_MAYBE_UNUSED) && defined(__has_attribute) |
412 | | # if __has_attribute(unused) |
413 | | # define MBEDTLS_MAYBE_UNUSED __attribute__((unused)) |
414 | | # endif |
415 | | #endif |
416 | | #if !defined(MBEDTLS_MAYBE_UNUSED) && defined(__GNUC__) |
417 | | # define MBEDTLS_MAYBE_UNUSED __attribute__((unused)) |
418 | | #endif |
419 | | #if !defined(MBEDTLS_MAYBE_UNUSED) && defined(__IAR_SYSTEMS_ICC__) && defined(__VER__) |
420 | | /* IAR does support __attribute__((unused)), but only if the -e flag (extended language support) |
421 | | * is given; the pragma always works. |
422 | | * Unfortunately the pragma affects the rest of the file where it is used, but this is harmless. |
423 | | * Check for version 5.2 or later - this pragma may be supported by earlier versions, but I wasn't |
424 | | * able to find documentation). |
425 | | */ |
426 | | # if (__VER__ >= 5020000) |
427 | | # define MBEDTLS_MAYBE_UNUSED _Pragma("diag_suppress=Pe177") |
428 | | # endif |
429 | | #endif |
430 | | #if !defined(MBEDTLS_MAYBE_UNUSED) && defined(_MSC_VER) |
431 | | # define MBEDTLS_MAYBE_UNUSED __pragma(warning(suppress:4189)) |
432 | | #endif |
433 | | #if !defined(MBEDTLS_MAYBE_UNUSED) |
434 | | # define MBEDTLS_MAYBE_UNUSED |
435 | | #endif |
436 | | |
437 | | #endif /* MBEDTLS_LIBRARY_COMMON_H */ |