about summary refs log tree commit diff
path: root/absl/base/internal/unaligned_access.h
blob: 5c7517abd71a6bf5250719d1d58cc7a05a906e1a (plain) (blame)
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
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
//
// Copyright 2017 The Abseil Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//      http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//

#ifndef ABSL_BASE_INTERNAL_UNALIGNED_ACCESS_H_
#define ABSL_BASE_INTERNAL_UNALIGNED_ACCESS_H_

#include <string.h>
#include <cstdint>

#include "absl/base/attributes.h"

// unaligned APIs

// Portable handling of unaligned loads, stores, and copies.
// On some platforms, like ARM, the copy functions can be more efficient
// then a load and a store.
//
// It is possible to implement all of these these using constant-length memcpy
// calls, which is portable and will usually be inlined into simple loads and
// stores if the architecture supports it. However, such inlining usually
// happens in a pass that's quite late in compilation, which means the resulting
// loads and stores cannot participate in many other optimizations, leading to
// overall worse code.

// The unaligned API is C++ only.  The declarations use C++ features
// (namespaces, inline) which are absent or incompatible in C.
#if defined(__cplusplus)

#if defined(ADDRESS_SANITIZER) || defined(THREAD_SANITIZER) ||\
    defined(MEMORY_SANITIZER)
// Consider we have an unaligned load/store of 4 bytes from address 0x...05.
// AddressSanitizer will treat it as a 3-byte access to the range 05:07 and
// will miss a bug if 08 is the first unaddressable byte.
// ThreadSanitizer will also treat this as a 3-byte access to 05:07 and will
// miss a race between this access and some other accesses to 08.
// MemorySanitizer will correctly propagate the shadow on unaligned stores
// and correctly report bugs on unaligned loads, but it may not properly
// update and report the origin of the uninitialized memory.
// For all three tools, replacing an unaligned access with a tool-specific
// callback solves the problem.

// Make sure uint16_t/uint32_t/uint64_t are defined.
#include <stdint.h>

extern "C" {
uint16_t __sanitizer_unaligned_load16(const void *p);
uint32_t __sanitizer_unaligned_load32(const void *p);
uint64_t __sanitizer_unaligned_load64(const void *p);
void __sanitizer_unaligned_store16(void *p, uint16_t v);
void __sanitizer_unaligned_store32(void *p, uint32_t v);
void __sanitizer_unaligned_store64(void *p, uint64_t v);
}  // extern "C"

namespace absl {

inline uint16_t UnalignedLoad16(const void *p) {
  return __sanitizer_unaligned_load16(p);
}

inline uint32_t UnalignedLoad32(const void *p) {
  return __sanitizer_unaligned_load32(p);
}

inline uint64_t UnalignedLoad64(const void *p) {
  return __sanitizer_unaligned_load64(p);
}

inline void UnalignedStore16(void *p, uint16_t v) {
  __sanitizer_unaligned_store16(p, v);
}

inline void UnalignedStore32(void *p, uint32_t v) {
  __sanitizer_unaligned_store32(p, v);
}

inline void UnalignedStore64(void *p, uint64_t v) {
  __sanitizer_unaligned_store64(p, v);
}

}  // namespace absl

#define ABSL_INTERNAL_UNALIGNED_LOAD16(_p) (absl::UnalignedLoad16(_p))
#define ABSL_INTERNAL_UNALIGNED_LOAD32(_p) (absl::UnalignedLoad32(_p))
#define ABSL_INTERNAL_UNALIGNED_LOAD64(_p) (absl::UnalignedLoad64(_p))

#define ABSL_INTERNAL_UNALIGNED_STORE16(_p, _val) \
  (absl::UnalignedStore16(_p, _val))
#define ABSL_INTERNAL_UNALIGNED_STORE32(_p, _val) \
  (absl::UnalignedStore32(_p, _val))
#define ABSL_INTERNAL_UNALIGNED_STORE64(_p, _val) \
  (absl::UnalignedStore64(_p, _val))

#elif defined(UNDEFINED_BEHAVIOR_SANITIZER)

namespace absl {

inline uint16_t UnalignedLoad16(const void *p) {
  uint16_t t;
  memcpy(&t, p, sizeof t);
  return t;
}

inline uint32_t UnalignedLoad32(const void *p) {
  uint32_t t;
  memcpy(&t, p, sizeof t);
  return t;
}

inline uint64_t UnalignedLoad64(const void *p) {
  uint64_t t;
  memcpy(&t, p, sizeof t);
  return t;
}

inline void UnalignedStore16(void *p, uint16_t v) { memcpy(p, &v, sizeof v); }

inline void UnalignedStore32(void *p, uint32_t v) { memcpy(p, &v, sizeof v); }

inline void UnalignedStore64(void *p, uint64_t v) { memcpy(p, &v, sizeof v); }

}  // namespace absl

#define ABSL_INTERNAL_UNALIGNED_LOAD16(_p) (absl::UnalignedLoad16(_p))
#define ABSL_INTERNAL_UNALIGNED_LOAD32(_p) (absl::UnalignedLoad32(_p))
#define ABSL_INTERNAL_UNALIGNED_LOAD64(_p) (absl::UnalignedLoad64(_p))

#define ABSL_INTERNAL_UNALIGNED_STORE16(_p, _val) \
  (absl::UnalignedStore16(_p, _val))
#define ABSL_INTERNAL_UNALIGNED_STORE32(_p, _val) \
  (absl::UnalignedStore32(_p, _val))
#define ABSL_INTERNAL_UNALIGNED_STORE64(_p, _val) \
  (absl::UnalignedStore64(_p, _val))

#elif defined(__x86_64__) || defined(_M_X64) || defined(__i386) || \
    defined(_M_IX86) || defined(__ppc__) || defined(__PPC__) ||    \
    defined(__ppc64__) || defined(__PPC64__)

// x86 and x86-64 can perform unaligned loads/stores directly;
// modern PowerPC hardware can also do unaligned integer loads and stores;
// but note: the FPU still sends unaligned loads and stores to a trap handler!

#define ABSL_INTERNAL_UNALIGNED_LOAD16(_p) \
  (*reinterpret_cast<const uint16_t *>(_p))
#define ABSL_INTERNAL_UNALIGNED_LOAD32(_p) \
  (*reinterpret_cast<const uint32_t *>(_p))
#define ABSL_INTERNAL_UNALIGNED_LOAD64(_p) \
  (*reinterpret_cast<const uint64_t *>(_p))

#define ABSL_INTERNAL_UNALIGNED_STORE16(_p, _val) \
  (*reinterpret_cast<uint16_t *>(_p) = (_val))
#define ABSL_INTERNAL_UNALIGNED_STORE32(_p, _val) \
  (*reinterpret_cast<uint32_t *>(_p) = (_val))
#define ABSL_INTERNAL_UNALIGNED_STORE64(_p, _val) \
  (*reinterpret_cast<uint64_t *>(_p) = (_val))

#elif defined(__arm__) && \
      !defined(__ARM_ARCH_5__) && \
      !defined(__ARM_ARCH_5T__) && \
      !defined(__ARM_ARCH_5TE__) && \
      !defined(__ARM_ARCH_5TEJ__) && \
      !defined(__ARM_ARCH_6__) && \
      !defined(__ARM_ARCH_6J__) && \
      !defined(__ARM_ARCH_6K__) && \
      !defined(__ARM_ARCH_6Z__) && \
      !defined(__ARM_ARCH_6ZK__) && \
      !defined(__ARM_ARCH_6T2__)


// ARMv7 and newer support native unaligned accesses, but only of 16-bit
// and 32-bit values (not 64-bit); older versions either raise a fatal signal,
// do an unaligned read and rotate the words around a bit, or do the reads very
// slowly (trip through kernel mode). There's no simple #define that says just
// "ARMv7 or higher", so we have to filter away all ARMv5 and ARMv6
// sub-architectures. Newer gcc (>= 4.6) set an __ARM_FEATURE_ALIGNED #define,
// so in time, maybe we can move on to that.
//
// This is a mess, but there's not much we can do about it.
//
// To further complicate matters, only LDR instructions (single reads) are
// allowed to be unaligned, not LDRD (two reads) or LDM (many reads). Unless we
// explicitly tell the compiler that these accesses can be unaligned, it can and
// will combine accesses. On armcc, the way to signal this is done by accessing
// through the type (uint32_t __packed *), but GCC has no such attribute
// (it ignores __attribute__((packed)) on individual variables). However,
// we can tell it that a _struct_ is unaligned, which has the same effect,
// so we do that.

namespace absl {
namespace internal {

struct Unaligned16Struct {
  uint16_t value;
  uint8_t dummy;  // To make the size non-power-of-two.
} ABSL_ATTRIBUTE_PACKED;

struct Unaligned32Struct {
  uint32_t value;
  uint8_t dummy;  // To make the size non-power-of-two.
} ABSL_ATTRIBUTE_PACKED;

}  // namespace internal
}  // namespace absl

#define ABSL_INTERNAL_UNALIGNED_LOAD16(_p) \
  ((reinterpret_cast<const ::absl::internal::Unaligned16Struct *>(_p))->value)
#define ABSL_INTERNAL_UNALIGNED_LOAD32(_p) \
  ((reinterpret_cast<const ::absl::internal::Unaligned32Struct *>(_p))->value)

#define ABSL_INTERNAL_UNALIGNED_STORE16(_p, _val)                          \
  ((reinterpret_cast< ::absl::internal::Unaligned16Struct *>(_p))->value = \
       (_val))
#define ABSL_INTERNAL_UNALIGNED_STORE32(_p, _val)                          \
  ((reinterpret_cast< ::absl::internal::Unaligned32Struct *>(_p))->value = \
       (_val))

namespace absl {

inline uint64_t UnalignedLoad64(const void *p) {
  uint64_t t;
  memcpy(&t, p, sizeof t);
  return t;
}

inline void UnalignedStore64(void *p, uint64_t v) { memcpy(p, &v, sizeof v); }

}  // namespace absl

#define ABSL_INTERNAL_UNALIGNED_LOAD64(_p) (absl::UnalignedLoad64(_p))
#define ABSL_INTERNAL_UNALIGNED_STORE64(_p, _val) \
  (absl::UnalignedStore64(_p, _val))

#else

// ABSL_INTERNAL_NEED_ALIGNED_LOADS is defined when the underlying platform
// doesn't support unaligned access.
#define ABSL_INTERNAL_NEED_ALIGNED_LOADS

// These functions are provided for architectures that don't support
// unaligned loads and stores.

namespace absl {

inline uint16_t UnalignedLoad16(const void *p) {
  uint16_t t;
  memcpy(&t, p, sizeof t);
  return t;
}

inline uint32_t UnalignedLoad32(const void *p) {
  uint32_t t;
  memcpy(&t, p, sizeof t);
  return t;
}

inline uint64_t UnalignedLoad64(const void *p) {
  uint64_t t;
  memcpy(&t, p, sizeof t);
  return t;
}

inline void UnalignedStore16(void *p, uint16_t v) { memcpy(p, &v, sizeof v); }

inline void UnalignedStore32(void *p, uint32_t v) { memcpy(p, &v, sizeof v); }

inline void UnalignedStore64(void *p, uint64_t v) { memcpy(p, &v, sizeof v); }

}  // namespace absl

#define ABSL_INTERNAL_UNALIGNED_LOAD16(_p) (absl::UnalignedLoad16(_p))
#define ABSL_INTERNAL_UNALIGNED_LOAD32(_p) (absl::UnalignedLoad32(_p))
#define ABSL_INTERNAL_UNALIGNED_LOAD64(_p) (absl::UnalignedLoad64(_p))

#define ABSL_INTERNAL_UNALIGNED_STORE16(_p, _val) \
  (absl::UnalignedStore16(_p, _val))
#define ABSL_INTERNAL_UNALIGNED_STORE32(_p, _val) \
  (absl::UnalignedStore32(_p, _val))
#define ABSL_INTERNAL_UNALIGNED_STORE64(_p, _val) \
  (absl::UnalignedStore64(_p, _val))

#endif

#endif  // defined(__cplusplus), end of unaligned API

#endif  // ABSL_BASE_INTERNAL_UNALIGNED_ACCESS_H_