src/ReturnAddressList.cc

/* -*- Mode: C++; tab-width: 8; c-basic-offset: 2; indent-tabs-mode: nil; -*- */

#include "ReturnAddressList.h"

#include "Task.h"

namespace rr {

static bool read_bytes_no_breakpoints(Task* t, remote_ptr<void> p, ssize_t size,
                                      void* out) {
  if (t->read_bytes_fallible(p, size, out) != size) {
    return false;
  }
  t->vm()->replace_breakpoints_with_original_values(static_cast<uint8_t*>(out),
      size, p.cast<uint8_t>());
  return true;
}

template <typename Arch>
static void return_addresses_x86ish(ReturnAddressList* result, Task* t) {
  // Immediately after a function call the return address is on the stack at
  // SP. After BP is pushed, but before it's initialized for the new stack
  // frame, the return address is on the stack at SP+wordsize. Just
  // capture those words now. We could inspect the code for known prologs/
  // epilogs but that misses cases such as calling into optimized code
  // or PLT stubs (which start with 'jmp'). Since it doesn't matter if we
  // capture addresses that aren't real return addresses, just capture those
  // words unconditionally.
  typename Arch::size_t frame[2];
  int next_address = 0;
  // Make sure the data we fetch here does not depend on where breakpoints have
  // been set. We don't want these results to vary because we call this in
  // some contexts with internal breakpoints set and in other contexts without
  // them set.
  if (read_bytes_no_breakpoints(t, t->regs().sp(), sizeof(frame), frame)) {
    result->addresses[0] = frame[0];
    result->addresses[1] = frame[1];
    next_address = 2;
  }

  remote_ptr<void> bp = t->regs().bp();
  for (int i = next_address; i < ReturnAddressList::X86_COUNT; ++i) {
    if (!read_bytes_no_breakpoints(t, bp, sizeof(frame), frame)) {
      break;
    }
    result->addresses[i] = frame[1];
    bp = frame[0];
  }
}

template <typename Arch>
void return_addresses_arch(ReturnAddressList* result, Task* t);

template <> void return_addresses_arch<X86Arch>(ReturnAddressList* result, Task* t) {
  return_addresses_x86ish<X86Arch>(result, t);
}

template <> void return_addresses_arch<X64Arch>(ReturnAddressList* result, Task* t) {
  return_addresses_x86ish<X64Arch>(result, t);
}

template <> void return_addresses_arch<ARM64Arch>(ReturnAddressList* result, Task*) {
  // On aarch64, we track all taken branches, so tracking return addresses should
  // not be necessary.
  DEBUG_ASSERT(ReturnAddressList::AARCH64_COUNT == 0);
  for (size_t i = 0; i < array_length(result->addresses); ++i) {
    result->addresses[i] = nullptr;
  }
}

static void compute_return_addresses(ReturnAddressList* result, Task* t) {
  RR_ARCH_FUNCTION(return_addresses_arch, t->arch(), result, t);
}

ReturnAddressList::ReturnAddressList(Task* t) {
  compute_return_addresses(this, t);
}

} // namespace rr