aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/libs/clang14-rt/lib/scudo/standalone/tsd_shared.h
blob: 4b3b1a07c12fb4ab728d693c70d3e310bf36b026 (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
//===-- tsd_shared.h --------------------------------------------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//

#ifndef SCUDO_TSD_SHARED_H_
#define SCUDO_TSD_SHARED_H_

#include "tsd.h"

#if SCUDO_HAS_PLATFORM_TLS_SLOT
// This is a platform-provided header that needs to be on the include path when
// Scudo is compiled. It must declare a function with the prototype:
//   uintptr_t *getPlatformAllocatorTlsSlot()
// that returns the address of a thread-local word of storage reserved for
// Scudo, that must be zero-initialized in newly created threads.
#error #include "scudo_platform_tls_slot.h"
#endif

namespace scudo {

template <class Allocator, u32 TSDsArraySize, u32 DefaultTSDCount>
struct TSDRegistrySharedT {
  void init(Allocator *Instance) {
    DCHECK(!Initialized);
    Instance->init();
    for (u32 I = 0; I < TSDsArraySize; I++)
      TSDs[I].init(Instance);
    const u32 NumberOfCPUs = getNumberOfCPUs();
    setNumberOfTSDs((NumberOfCPUs == 0) ? DefaultTSDCount
                                        : Min(NumberOfCPUs, DefaultTSDCount));
    Initialized = true;
  }

  void initOnceMaybe(Allocator *Instance) {
    ScopedLock L(Mutex);
    if (LIKELY(Initialized))
      return;
    init(Instance); // Sets Initialized.
  }

  void unmapTestOnly(Allocator *Instance) {
    for (u32 I = 0; I < TSDsArraySize; I++) {
      TSDs[I].commitBack(Instance);
      TSDs[I] = {};
    }
    setCurrentTSD(nullptr);
    Initialized = false;
  }

  ALWAYS_INLINE void initThreadMaybe(Allocator *Instance,
                                     UNUSED bool MinimalInit) {
    if (LIKELY(getCurrentTSD()))
      return;
    initThread(Instance);
  }

  ALWAYS_INLINE TSD<Allocator> *getTSDAndLock(bool *UnlockRequired) {
    TSD<Allocator> *TSD = getCurrentTSD();
    DCHECK(TSD);
    *UnlockRequired = true;
    // Try to lock the currently associated context.
    if (TSD->tryLock())
      return TSD;
    // If that fails, go down the slow path.
    if (TSDsArraySize == 1U) {
      // Only 1 TSD, not need to go any further.
      // The compiler will optimize this one way or the other.
      TSD->lock();
      return TSD;
    }
    return getTSDAndLockSlow(TSD);
  }

  void disable() {
    Mutex.lock();
    for (u32 I = 0; I < TSDsArraySize; I++)
      TSDs[I].lock();
  }

  void enable() {
    for (s32 I = static_cast<s32>(TSDsArraySize - 1); I >= 0; I--)
      TSDs[I].unlock();
    Mutex.unlock();
  }

  bool setOption(Option O, sptr Value) {
    if (O == Option::MaxTSDsCount)
      return setNumberOfTSDs(static_cast<u32>(Value));
    if (O == Option::ThreadDisableMemInit)
      setDisableMemInit(Value);
    // Not supported by the TSD Registry, but not an error either.
    return true;
  }

  bool getDisableMemInit() const { return *getTlsPtr() & 1; }

private:
  ALWAYS_INLINE uptr *getTlsPtr() const {
#if SCUDO_HAS_PLATFORM_TLS_SLOT
    return reinterpret_cast<uptr *>(getPlatformAllocatorTlsSlot());
#else
    static thread_local uptr ThreadTSD;
    return &ThreadTSD;
#endif
  }

  static_assert(alignof(TSD<Allocator>) >= 2, "");

  ALWAYS_INLINE void setCurrentTSD(TSD<Allocator> *CurrentTSD) {
    *getTlsPtr() &= 1;
    *getTlsPtr() |= reinterpret_cast<uptr>(CurrentTSD);
  }

  ALWAYS_INLINE TSD<Allocator> *getCurrentTSD() {
    return reinterpret_cast<TSD<Allocator> *>(*getTlsPtr() & ~1ULL);
  }

  bool setNumberOfTSDs(u32 N) {
    ScopedLock L(MutexTSDs);
    if (N < NumberOfTSDs)
      return false;
    if (N > TSDsArraySize)
      N = TSDsArraySize;
    NumberOfTSDs = N;
    NumberOfCoPrimes = 0;
    // Compute all the coprimes of NumberOfTSDs. This will be used to walk the
    // array of TSDs in a random order. For details, see:
    // https://lemire.me/blog/2017/09/18/visiting-all-values-in-an-array-exactly-once-in-random-order/
    for (u32 I = 0; I < N; I++) {
      u32 A = I + 1;
      u32 B = N;
      // Find the GCD between I + 1 and N. If 1, they are coprimes.
      while (B != 0) {
        const u32 T = A;
        A = B;
        B = T % B;
      }
      if (A == 1)
        CoPrimes[NumberOfCoPrimes++] = I + 1;
    }
    return true;
  }

  void setDisableMemInit(bool B) {
    *getTlsPtr() &= ~1ULL;
    *getTlsPtr() |= B;
  }

  NOINLINE void initThread(Allocator *Instance) {
    initOnceMaybe(Instance);
    // Initial context assignment is done in a plain round-robin fashion.
    const u32 Index = atomic_fetch_add(&CurrentIndex, 1U, memory_order_relaxed);
    setCurrentTSD(&TSDs[Index % NumberOfTSDs]);
    Instance->callPostInitCallback();
  }

  NOINLINE TSD<Allocator> *getTSDAndLockSlow(TSD<Allocator> *CurrentTSD) {
    // Use the Precedence of the current TSD as our random seed. Since we are
    // in the slow path, it means that tryLock failed, and as a result it's
    // very likely that said Precedence is non-zero.
    const u32 R = static_cast<u32>(CurrentTSD->getPrecedence());
    u32 N, Inc;
    {
      ScopedLock L(MutexTSDs);
      N = NumberOfTSDs;
      DCHECK_NE(NumberOfCoPrimes, 0U);
      Inc = CoPrimes[R % NumberOfCoPrimes];
    }
    if (N > 1U) {
      u32 Index = R % N;
      uptr LowestPrecedence = UINTPTR_MAX;
      TSD<Allocator> *CandidateTSD = nullptr;
      // Go randomly through at most 4 contexts and find a candidate.
      for (u32 I = 0; I < Min(4U, N); I++) {
        if (TSDs[Index].tryLock()) {
          setCurrentTSD(&TSDs[Index]);
          return &TSDs[Index];
        }
        const uptr Precedence = TSDs[Index].getPrecedence();
        // A 0 precedence here means another thread just locked this TSD.
        if (Precedence && Precedence < LowestPrecedence) {
          CandidateTSD = &TSDs[Index];
          LowestPrecedence = Precedence;
        }
        Index += Inc;
        if (Index >= N)
          Index -= N;
      }
      if (CandidateTSD) {
        CandidateTSD->lock();
        setCurrentTSD(CandidateTSD);
        return CandidateTSD;
      }
    }
    // Last resort, stick with the current one.
    CurrentTSD->lock();
    return CurrentTSD;
  }

  atomic_u32 CurrentIndex = {};
  u32 NumberOfTSDs = 0;
  u32 NumberOfCoPrimes = 0;
  u32 CoPrimes[TSDsArraySize] = {};
  bool Initialized = false;
  HybridMutex Mutex;
  HybridMutex MutexTSDs;
  TSD<Allocator> TSDs[TSDsArraySize];
};

} // namespace scudo

#endif // SCUDO_TSD_SHARED_H_