//===- FuzzerTracePC.cpp - PC tracing--------------------------------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
// Trace PCs.
// This module implements __sanitizer_cov_trace_pc_guard[_init],
// the callback required for -fsanitize-coverage=trace-pc-guard instrumentation.
//
//===----------------------------------------------------------------------===//

#include "FuzzerCorpus.h"
#include "FuzzerDefs.h"
#include "FuzzerDictionary.h"
#include "FuzzerTracePC.h"
#include "FuzzerValueBitMap.h"

namespace fuzzer {

TracePC TPC;

void TracePC::HandleTrace(uint32_t *Guard, uintptr_t PC) {
  uint32_t Idx = *Guard;
  if (!Idx) return;
  uint8_t *CounterPtr = &Counters[Idx % kNumCounters];
  uint8_t Counter = *CounterPtr;
  if (Counter == 0) {
    if (!PCs[Idx % kNumPCs]) {
      AddNewPCID(Idx);
      TotalPCCoverage++;
      PCs[Idx % kNumPCs] = PC;
    }
  }
  if (UseCounters) {
    if (Counter < 128)
      *CounterPtr = Counter + 1;
    else
      *Guard = 0;
  } else {
    *CounterPtr = 1;
    *Guard = 0;
  }
}

void TracePC::HandleInit(uint32_t *Start, uint32_t *Stop) {
  if (Start == Stop || *Start) return;
  assert(NumModules < sizeof(Modules) / sizeof(Modules[0]));
  for (uint32_t *P = Start; P < Stop; P++)
    *P = ++NumGuards;
  Modules[NumModules].Start = Start;
  Modules[NumModules].Stop = Stop;
  NumModules++;
}

void TracePC::PrintModuleInfo() {
  Printf("INFO: Loaded %zd modules (%zd guards): ", NumModules, NumGuards);
  for (size_t i = 0; i < NumModules; i++)
    Printf("[%p, %p), ", Modules[i].Start, Modules[i].Stop);
  Printf("\n");
}

void TracePC::ResetGuards() {
  uint32_t N = 0;
  for (size_t M = 0; M < NumModules; M++)
    for (uint32_t *X = Modules[M].Start, *End = Modules[M].Stop; X < End; X++)
      *X = ++N;
  assert(N == NumGuards);
}

size_t TracePC::FinalizeTrace(InputCorpus *C, size_t InputSize, bool Shrink) {
  if (!UsingTracePcGuard()) return 0;
  size_t Res = 0;
  const size_t Step = 8;
  assert(reinterpret_cast<uintptr_t>(Counters) % Step == 0);
  size_t N = Min(kNumCounters, NumGuards + 1);
  N = (N + Step - 1) & ~(Step - 1);  // Round up.
  for (size_t Idx = 0; Idx < N; Idx += Step) {
    uint64_t Bundle = *reinterpret_cast<uint64_t*>(&Counters[Idx]);
    if (!Bundle) continue;
    for (size_t i = Idx; i < Idx + Step; i++) {
      uint8_t Counter = (Bundle >> (i * 8)) & 0xff;
      if (!Counter) continue;
      Counters[i] = 0;
      unsigned Bit = 0;
      /**/ if (Counter >= 128) Bit = 7;
      else if (Counter >= 32) Bit = 6;
      else if (Counter >= 16) Bit = 5;
      else if (Counter >= 8) Bit = 4;
      else if (Counter >= 4) Bit = 3;
      else if (Counter >= 3) Bit = 2;
      else if (Counter >= 2) Bit = 1;
      size_t Feature = (i * 8 + Bit);
      if (C->AddFeature(Feature, InputSize, Shrink))
        Res++;
    }
  }
  if (UseValueProfile)
    ValueProfileMap.ForEach([&](size_t Idx) {
      if (C->AddFeature(NumGuards + Idx, InputSize, Shrink))
        Res++;
    });
  return Res;
}

void TracePC::HandleCallerCallee(uintptr_t Caller, uintptr_t Callee) {
  const uintptr_t kBits = 12;
  const uintptr_t kMask = (1 << kBits) - 1;
  uintptr_t Idx = (Caller & kMask) | ((Callee & kMask) << kBits);
  HandleValueProfile(Idx);
}

void TracePC::PrintCoverage() {
  Printf("COVERAGE:\n");
  for (size_t i = 0; i < Min(NumGuards + 1, kNumPCs); i++) {
    if (PCs[i])
      PrintPC("COVERED: %p %F %L\n", "COVERED: %p\n", PCs[i]);
  }
}

// Value profile.
// We keep track of various values that affect control flow.
// These values are inserted into a bit-set-based hash map.
// Every new bit in the map is treated as a new coverage.
//
// For memcmp/strcmp/etc the interesting value is the length of the common
// prefix of the parameters.
// For cmp instructions the interesting value is a XOR of the parameters.
// The interesting value is mixed up with the PC and is then added to the map.

void TracePC::AddValueForMemcmp(void *caller_pc, const void *s1, const void *s2,
                              size_t n) {
  if (!n) return;
  size_t Len = std::min(n, (size_t)32);
  const uint8_t *A1 = reinterpret_cast<const uint8_t *>(s1);
  const uint8_t *A2 = reinterpret_cast<const uint8_t *>(s2);
  size_t I = 0;
  for (; I < Len; I++)
    if (A1[I] != A2[I])
      break;
  size_t PC = reinterpret_cast<size_t>(caller_pc);
  size_t Idx = I;
  // if (I < Len)
  //  Idx += __builtin_popcountl((A1[I] ^ A2[I])) - 1;
  TPC.HandleValueProfile((PC & 4095) | (Idx << 12));
}

void TracePC::AddValueForStrcmp(void *caller_pc, const char *s1, const char *s2,
                              size_t n) {
  if (!n) return;
  size_t Len = std::min(n, (size_t)32);
  const uint8_t *A1 = reinterpret_cast<const uint8_t *>(s1);
  const uint8_t *A2 = reinterpret_cast<const uint8_t *>(s2);
  size_t I = 0;
  for (; I < Len; I++)
    if (A1[I] != A2[I] || A1[I] == 0)
      break;
  size_t PC = reinterpret_cast<size_t>(caller_pc);
  size_t Idx = I;
  // if (I < Len && A1[I])
  //  Idx += __builtin_popcountl((A1[I] ^ A2[I])) - 1;
  TPC.HandleValueProfile((PC & 4095) | (Idx << 12));
}

template <class T>
ATTRIBUTE_TARGET_POPCNT
#ifdef __clang__  // g++ can't handle this __attribute__ here :(
__attribute__((always_inline))
#endif  // __clang__
void TracePC::HandleCmp(void *PC, T Arg1, T Arg2) {
  uintptr_t PCuint = reinterpret_cast<uintptr_t>(PC);
  uint64_t ArgXor = Arg1 ^ Arg2;
  uint64_t ArgDistance = __builtin_popcountl(ArgXor) + 1; // [1,65]
  uintptr_t Idx = ((PCuint & 4095) + 1) * ArgDistance;
  TORCInsert(ArgXor, Arg1, Arg2);
  HandleValueProfile(Idx);
}

void TracePC::ProcessTORC(Dictionary *Dict, const uint8_t *Data, size_t Size) {
  TORCToDict(TORC8, Dict, Data, Size);
  TORCToDict(TORC4, Dict, Data, Size);
}

template <class T>
void TracePC::TORCToDict(const TableOfRecentCompares<T, kTORCSize> &TORC,
                         Dictionary *Dict, const uint8_t *Data, size_t Size) {
  ScopedDoingMyOwnMemmem scoped_doing_my_own_memmem;
  for (size_t i = 0; i < TORC.kSize; i++) {
    T A[2] = {TORC.Table[i][0], TORC.Table[i][1]};
    if (!A[0] && !A[1]) continue;
    for (int j = 0; j < 2; j++)
      TORCToDict(Dict, A[j], A[!j], Data, Size);
  }
}

template <class T>
void TracePC::TORCToDict(Dictionary *Dict, T FindInData, T Substitute,
                         const uint8_t *Data, size_t Size) {
  if (FindInData == Substitute) return;
  if (sizeof(T) == 4) {
    uint16_t HigherBytes = Substitute >> sizeof(T) * 4;
    if (HigherBytes == 0 || HigherBytes == 0xffff)
      TORCToDict(Dict, static_cast<uint16_t>(FindInData),
                 static_cast<uint16_t>(Substitute), Data, Size);
  }
  const size_t DataSize = sizeof(T);
  const uint8_t *End = Data + Size;
  int Attempts = 3;
  for (int DoSwap = 0; DoSwap <= 1; DoSwap++) {
    for (const uint8_t *Cur = Data; Cur < End && Attempts--; Cur++) {
      Cur = (uint8_t *)memmem(Cur, End - Cur, &FindInData, DataSize);
      if (!Cur)
        break;
      size_t Pos = Cur - Data;
      Word W(reinterpret_cast<uint8_t *>(&Substitute), sizeof(Substitute));
      DictionaryEntry DE(W, Pos);
      // TODO: evict all entries from Dic if it's full.
      Dict->push_back(DE);
      // Printf("Dict[%zd] TORC%zd %llx => %llx pos %zd\n", Dict->size(),
      // sizeof(T),
      //       (uint64_t)FindInData, (uint64_t)Substitute, Pos);
    }
    FindInData = Bswap(FindInData);
    Substitute = Bswap(Substitute);
  }
}

} // namespace fuzzer

extern "C" {
__attribute__((visibility("default")))
void __sanitizer_cov_trace_pc_guard(uint32_t *Guard) {
  uintptr_t PC = (uintptr_t)__builtin_return_address(0);
  fuzzer::TPC.HandleTrace(Guard, PC);
}

__attribute__((visibility("default")))
void __sanitizer_cov_trace_pc_guard_init(uint32_t *Start, uint32_t *Stop) {
  fuzzer::TPC.HandleInit(Start, Stop);
}

__attribute__((visibility("default")))
void __sanitizer_cov_trace_pc_indir(uintptr_t Callee) {
  uintptr_t PC = (uintptr_t)__builtin_return_address(0);
  fuzzer::TPC.HandleCallerCallee(PC, Callee);
}

__attribute__((visibility("default")))
void __sanitizer_cov_trace_cmp8(uint64_t Arg1, uint64_t Arg2) {
  fuzzer::TPC.HandleCmp(__builtin_return_address(0), Arg1, Arg2);
}
__attribute__((visibility("default")))
void __sanitizer_cov_trace_cmp4(uint32_t Arg1, uint32_t Arg2) {
  fuzzer::TPC.HandleCmp(__builtin_return_address(0), Arg1, Arg2);
}
__attribute__((visibility("default")))
void __sanitizer_cov_trace_cmp2(uint16_t Arg1, uint16_t Arg2) {
  fuzzer::TPC.HandleCmp(__builtin_return_address(0), Arg1, Arg2);
}
__attribute__((visibility("default")))
void __sanitizer_cov_trace_cmp1(uint8_t Arg1, uint8_t Arg2) {
  fuzzer::TPC.HandleCmp(__builtin_return_address(0), Arg1, Arg2);
}

__attribute__((visibility("default")))
void __sanitizer_cov_trace_switch(uint64_t Val, uint64_t *Cases) {
  uint64_t N = Cases[0];
  uint64_t *Vals = Cases + 2;
  char *PC = (char*)__builtin_return_address(0);
  for (size_t i = 0; i < N; i++)
    if (Val != Vals[i])
      fuzzer::TPC.HandleCmp(PC + i, Val, Vals[i]);
}

__attribute__((visibility("default")))
void __sanitizer_cov_trace_div4(uint32_t Val) {
  fuzzer::TPC.HandleCmp(__builtin_return_address(0), Val, (uint32_t)0);
}
__attribute__((visibility("default")))
void __sanitizer_cov_trace_div8(uint64_t Val) {
  fuzzer::TPC.HandleCmp(__builtin_return_address(0), Val, (uint64_t)0);
}
__attribute__((visibility("default")))
void __sanitizer_cov_trace_gep(uintptr_t Idx) {
  fuzzer::TPC.HandleCmp(__builtin_return_address(0), Idx, (uintptr_t)0);
}

}  // extern "C"