Quelle symbolengine.cpp Sprache: C

/*
* Copyright (c) 2017, 2022, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/

#include "precompiled.hpp"
#include "utilities/globalDefinitions.hpp"
#include "symbolengine.hpp"
#include "utilities/debug.hpp"
#include "utilities/ostream.hpp"
#include "windbghelp.hpp"

#include <windows.h>

#include <imagehlp.h>
#include <psapi.h>

// This code may be invoked normally but also as part of error reporting
// In the latter case, we may run under tight memory constraints (native oom)
// or in a stack overflow situation or the C heap may be corrupted. We may
// run very early before VM initialization or very late when C exit handlers
// run. In all these cases, callstacks would still be nice, so lets be robust.
//
// We need a number of buffers - for the pdb search path, module handle
// lists, for demangled symbols, etc.
//
// These buffers, while typically small, may need to be large for corner
// cases (e.g. templatized C++ symbols, or many DLLs loaded). Where do we
// allocate them?
//
// We may be in error handling for a stack overflow, so lets not put them on
// the stack.
//
// Dynamically allocating them may fail if we are handling a native OOM. It
// is also a bit dangerous, as the C heap may be corrupted already.
//
// That leaves pre-allocating them globally, which is safe and should always
// work (if we synchronize access) but incurs an undesirable footprint for
// non-error cases.
//
// We follow a two-way strategy: Allocate the buffers on the C heap in a
// reasonable large size. Failing that, fall back to static preallocated
// buffers. The size of the latter is large enough to handle common scenarios
// but small enough not to drive up the footprint too much (several kb).
//
// We keep these buffers around once allocated, for subsequent requests. This
// means that by running the initialization early at a safe time - before
// any error happens - buffers can be pre-allocated. This increases the chance
// of useful callstacks in error scenarios in exchange for a some cycles spent
// at startup. This behavior can be controlled with -XX:+InitializeDbgHelpEarly
// and is off by default.

///////

// A simple buffer which attempts to allocate an optimal size but will
// fall back to a static minimally sized array on allocation error.
template <class T, int MINIMAL_CAPACITY, int OPTIMAL_CAPACITY>
class SimpleBufferWithFallback {
  T _fallback_buffer[MINIMAL_CAPACITY];
  T* _p;
  int _capacity;

  // A sentinel at the end of the buffer to catch overflows.
  void imprint_sentinel() {
    assert(_p && _capacity > 0, "Buffer must be allocated");
    _p[_capacity - 1] = (T)'X';
    _capacity --;
  }

public:

  SimpleBufferWithFallback<T, MINIMAL_CAPACITY, OPTIMAL_CAPACITY> ()
    : _p(NULL), _capacity(0)
  {}

  // Note: no destructor because these buffers should, once
  // allocated, live until process end.
  // ~SimpleBufferWithFallback()

  // Note: We use raw ::malloc/::free here instead of os::malloc()/os::free
  // to prevent circularities or secondary crashes during error reporting.
  virtual void initialize () {
    assert(_p == NULL && _capacity == 0, "Only call once.");
    const size_t bytes = OPTIMAL_CAPACITY * sizeof(T);
    T* q = (T*) ::malloc(bytes);
    if (q != NULL) {
      _p = q;
      _capacity = OPTIMAL_CAPACITY;
    } else {
      _p = _fallback_buffer;
      _capacity = (int)(sizeof(_fallback_buffer) / sizeof(T));
    }
    _p[0] = '\0';
    imprint_sentinel();
  }

  // We need a way to reset the buffer to fallback size for one special
  // case, where two buffers need to be of identical capacity.
  void reset_to_fallback_capacity() {
    if (_p != _fallback_buffer) {
      ::free(_p);
    }
    _p = _fallback_buffer;
    _capacity = (int)(sizeof(_fallback_buffer) / sizeof(T));
    _p[0] = '\0';
    imprint_sentinel();
  }

  T* ptr()                { return _p; }
  const T* ptr() const    { return _p; }
  int capacity() const    { return _capacity; }

#ifdef ASSERT
  void check() const {
    assert(_p[_capacity] == (T)'X', "sentinel lost");
  }
#else
  void check() const {}
#endif

};

////

// ModuleHandleArray: a list holding module handles. Needs to be large enough
// to hold one handle per loaded DLL.
// Note: a standard OpenJDK loads normally ~30 libraries, including system
// libraries, without third party libraries.

typedef SimpleBufferWithFallback <HMODULE, 48, 512> ModuleHandleArrayBase;

class ModuleHandleArray : public ModuleHandleArrayBase {

  int _num; // Number of handles in this array (may be < capacity).

public:

  void initialize() {
    ModuleHandleArrayBase::initialize();
    _num = 0;
  }

  int num() const { return _num; }
  void set_num(int n) {
    assert(n <= capacity(), "Too large");
    _num = n;
  }

  // Compare with another list; returns true if all handles are equal (incl.
  // sort order)
  bool equals(const ModuleHandleArray& other) const {
    if (_num != other._num) {
      return false;
    }
    if (::memcmp(ptr(), other.ptr(), _num * sizeof(HMODULE)) != 0) {
      return false;
    }
    return true;
  }

  // Copy content from other list.
  void copy_content_from(ModuleHandleArray& other) {
    assert(capacity() == other.capacity(), "Different capacities.");
    memcpy(ptr(), other.ptr(), other._num * sizeof(HMODULE));
    _num = other._num;
  }

};

////

// PathBuffer: a buffer to hold and work with a pdb search PATH - a concatenation
// of multiple directories separated by ';'.
// A single directory name can be (NTFS) as long as 32K, but in reality is
// seldom larger than the (historical) MAX_PATH of 260.

#define MINIMUM_PDB_PATH_LENGTH  MAX_PATH * 4
#define OPTIMAL_PDB_PATH_LENGTH  MAX_PATH * 64

typedef SimpleBufferWithFallback<char, MINIMUM_PDB_PATH_LENGTH, OPTIMAL_PDB_PATH_LENGTH> PathBufferBase;

class PathBuffer: public PathBufferBase {
public:

  // Search PDB path for a directory. Search is case insensitive. Returns
  // true if directory was found in the path, false otherwise.
  bool contains_directory(const char* directory) {
    if (ptr() == NULL) {
      return false;
    }
    const size_t len = strlen(directory);
    if (len == 0) {
      return false;
    }
    char* p = ptr();
    for(;;) {
      char* q = strchr(p, ';');
      if (q != NULL) {
        if (len == (q - p)) {
          if (_strnicmp(p, directory, len) == 0) {
            return true;
          }
        }
        p = q + 1;
      } else {
        // tail
        return _stricmp(p, directory) == 0;
      }
    }
    return false;
  }

  // Appends the given directory to the path. Returns false if internal
  // buffer size was not sufficient.
  bool append_directory(const char* directory) {
    const size_t len = strlen(directory);
    if (len == 0) {
      return false;
    }
    char* p = ptr();
    const size_t len_now = strlen(p);
    const size_t needs_capacity = len_now + 1 + len + 1; // xxx;yy\0
    if (needs_capacity > (size_t)capacity()) {
      return false; // OOM
    }
    if (len_now > 0) { // Not the first path element.
      p += len_now;
      *p = ';';
      p ++;
    }
    strcpy(p, directory);
    return true;
  }

};

// A simple buffer to hold one single file name. A file name can be (NTFS) as
// long as 32K, but in reality is seldom larger than MAX_PATH.
typedef SimpleBufferWithFallback<char, MAX_PATH, 8 * K> FileNameBuffer;

// A buffer to hold a C++ symbol. Usually small, but symbols may be larger for
// templates.
#define MINIMUM_SYMBOL_NAME_LEN 128
#define OPTIMAL_SYMBOL_NAME_LEN 1024

typedef SimpleBufferWithFallback<uint8_t,
        sizeof(IMAGEHLP_SYMBOL64) + MINIMUM_SYMBOL_NAME_LEN,
        sizeof(IMAGEHLP_SYMBOL64) + OPTIMAL_SYMBOL_NAME_LEN> SymbolBuffer;

static struct {

  // Two buffers to hold lists of loaded modules. handles across invocations of
  // SymbolEngine::recalc_search_path().
  ModuleHandleArray loaded_modules;
  ModuleHandleArray last_loaded_modules;
  // Buffer to retrieve and assemble the pdb search path.
  PathBuffer search_path;
  // Buffer to retrieve directory names for loaded modules.
  FileNameBuffer dir_name;
  // Buffer to retrieve decoded symbol information (in SymbolEngine::decode)
  SymbolBuffer decode_buffer;

  void initialize() {
    search_path.initialize();
    dir_name.initialize();
    decode_buffer.initialize();

    loaded_modules.initialize();
    last_loaded_modules.initialize();

    // Note: both module lists must have the same capacity. If one allocation
    // did fail, let them both fall back to the fallback size.
    if (loaded_modules.capacity() != last_loaded_modules.capacity()) {
      loaded_modules.reset_to_fallback_capacity();
      last_loaded_modules.reset_to_fallback_capacity();
    }

    assert(search_path.capacity() > 0 && dir_name.capacity() > 0 &&
            decode_buffer.capacity() > 0 && loaded_modules.capacity() > 0 &&
            last_loaded_modules.capacity() > 0, "Init error.");
  }

} g_buffers;

// Scan the loaded modules.
//
// For each loaded module, add the directory it is located in to the pdb search
// path, but avoid duplicates. Prior search path content is preserved.
//
// If p_search_path_was_updated is not NULL, points to a bool which, upon
// successful return from the function, contains true if the search path
// was updated, false if no update was needed because no new DLLs were
// loaded or unloaded.
//
// Returns true for success, false for error.
static bool recalc_search_path_locked(bool* p_search_path_was_updated) {

  if (p_search_path_was_updated) {
    *p_search_path_was_updated = false;
  }

  HANDLE hProcess = ::GetCurrentProcess();

  BOOL success = false;

  // 1) Retrieve current set search path.
  //    (PDB search path is a global setting and someone might have modified
  //     it, so take care not to remove directories, just to add our own).

  if (!WindowsDbgHelp::symGetSearchPath(hProcess, g_buffers.search_path.ptr(),
                                       (int)g_buffers.search_path.capacity())) {
    return false;
  }
  DEBUG_ONLY(g_buffers.search_path.check();)

  // 2) Retrieve list of modules handles of all currently loaded modules.
  DWORD bytes_needed = 0;
  const DWORD buffer_capacity_bytes = (DWORD)g_buffers.loaded_modules.capacity() * sizeof(HMODULE);
  success = ::EnumProcessModules(hProcess, g_buffers.loaded_modules.ptr(),
                                 buffer_capacity_bytes, &bytes_needed);
  DEBUG_ONLY(g_buffers.loaded_modules.check();)

  // Note: EnumProcessModules is sloppily defined in terms of whether a
  // too-small output buffer counts as error. Will it truncate but still
  // return TRUE? Nobody knows and the manpage is not telling. So we count
  // truncation it as error, disregarding the return value.
  if (!success || bytes_needed > buffer_capacity_bytes) {
    return false;
  } else {
    const int num_modules = bytes_needed / sizeof(HMODULE);
    g_buffers.loaded_modules.set_num(num_modules);
  }

  // Compare the list of module handles with the last list. If the lists are
  // identical, no additional dlls were loaded and we can stop.
  if (g_buffers.loaded_modules.equals(g_buffers.last_loaded_modules)) {
    return true;
  } else {
    // Remember the new set of module handles and continue.
    g_buffers.last_loaded_modules.copy_content_from(g_buffers.loaded_modules);
  }

  // 3) For each loaded module: retrieve directory from which it was loaded.
  //    Add directory to search path (but avoid duplicates).

  bool did_modify_searchpath = false;

  for (int i = 0; i < (int)g_buffers.loaded_modules.num(); i ++) {

    const HMODULE hMod = g_buffers.loaded_modules.ptr()[i];
    char* const filebuffer = g_buffers.dir_name.ptr();
    const int file_buffer_capacity = g_buffers.dir_name.capacity();
    const int len_returned = (int)::GetModuleFileName(hMod, filebuffer, (DWORD)file_buffer_capacity);
    DEBUG_ONLY(g_buffers.dir_name.check();)
    if (len_returned == 0) {
      // This may happen when a module gets unloaded after our call to EnumProcessModules.
      // It should be rare but may sporadically happen. Just ignore and continue with the
      // next module.
      continue;
    } else if (len_returned == file_buffer_capacity) {
      // Truncation. Just skip this module and continue with the next module.
      continue;
    }

    // Cut file name part off.
    char* last_slash = ::strrchr(filebuffer, '\\');
    if (last_slash == NULL) {
      last_slash = ::strrchr(filebuffer, '/');
    }
    if (last_slash) {
      *last_slash = '\0';
    }

    // If this is already part of the search path, ignore it, otherwise
    // append to search path.
    if (!g_buffers.search_path.contains_directory(filebuffer)) {
      if (!g_buffers.search_path.append_directory(filebuffer)) {
        return false; // oom
      }
      DEBUG_ONLY(g_buffers.search_path.check();)
      did_modify_searchpath = true;
    }

  } // for each loaded module.

  // If we did not modify the search path, nothing further needs to be done.
  if (!did_modify_searchpath) {
    return true;
  }

  // Set the search path to its new value.
  if (!WindowsDbgHelp::symSetSearchPath(hProcess, g_buffers.search_path.ptr())) {
    return false;
  }

  if (p_search_path_was_updated) {
    *p_search_path_was_updated = true;
  }

  return true;

}

static bool demangle_locked(const char* symbol, char *buf, int buflen) {

  return WindowsDbgHelp::unDecorateSymbolName(symbol, buf, buflen, UNDNAME_COMPLETE) > 0;

}

static bool decode_locked(const void* addr, char* buf, int buflen, int* offset, bool do_demangle) {

  assert(g_buffers.decode_buffer.capacity() >= (sizeof(IMAGEHLP_SYMBOL64) + MINIMUM_SYMBOL_NAME_LEN),
         "Decode buffer too small.");
  assert(buf != NULL && buflen > 0 && offset != NULL, "invalid output buffer.");

  DWORD64 displacement;
  PIMAGEHLP_SYMBOL64 pSymbol = NULL;
  bool success = false;

  pSymbol = (PIMAGEHLP_SYMBOL64) g_buffers.decode_buffer.ptr();
  pSymbol->SizeOfStruct = sizeof(IMAGEHLP_SYMBOL64);
  pSymbol->MaxNameLength = (DWORD)(g_buffers.decode_buffer.capacity() - sizeof(IMAGEHLP_SYMBOL64) - 1);

  // It is unclear how SymGetSymFromAddr64 handles truncation. Experiments
  // show it will return TRUE but not zero terminate (which is a really bad
  // combination). Lets be super careful.
  ::memset(pSymbol->Name, 0, pSymbol->MaxNameLength); // To catch truncation.

  if (WindowsDbgHelp::symGetSymFromAddr64(::GetCurrentProcess(), (DWORD64)addr, &displacement, pSymbol)) {
    success = true;
    if (pSymbol->Name[pSymbol->MaxNameLength - 1] != '\0') {
      // Symbol was truncated. Do not attempt to demangle. Instead, zero terminate the
      // truncated string. We still return success - the truncated string may still
      // be usable for the caller.
      pSymbol->Name[pSymbol->MaxNameLength - 1] = '\0';
      do_demangle = false;
    }

    // Attempt to demangle.
    if (do_demangle && demangle_locked(pSymbol->Name, buf, buflen)) {
      // ok.
    } else {
      ::strncpy(buf, pSymbol->Name, buflen - 1);
    }
    buf[buflen - 1] = '\0';

    *offset = (int)displacement;
  }

  DEBUG_ONLY(g_buffers.decode_buffer.check();)

  return success;
}

static enum {
  state_uninitialized = 0,
  state_ready = 1,
  state_error = 2
} g_state = state_uninitialized;

static void initialize() {

  assert(g_state == state_uninitialized, "wrong sequence");
  g_state = state_error;

  // 1) Initialize buffers.
  g_buffers.initialize();

  // 1) Call SymInitialize
  HANDLE hProcess = ::GetCurrentProcess();
  WindowsDbgHelp::symSetOptions(SYMOPT_FAIL_CRITICAL_ERRORS | SYMOPT_DEFERRED_LOADS |
                        SYMOPT_EXACT_SYMBOLS | SYMOPT_LOAD_LINES);
  if (!WindowsDbgHelp::symInitialize(hProcess, NULL, TRUE)) {
    return;
  }

  // Note: we ignore any errors from this point on. The symbol engine may be
  // usable enough.
  g_state = state_ready;

  (void)recalc_search_path_locked(NULL);

}

///////////////////// External functions //////////////////////////

// All outside facing functions are synchronized. Also, we run
// initialization on first touch.

static CRITICAL_SECTION g_cs;

namespace { // Do not export.
  class SymbolEngineEntry {
   public:
    SymbolEngineEntry() {
      ::EnterCriticalSection(&g_cs);
      if (g_state == state_uninitialized) {
        initialize();
      }
    }
    ~SymbolEngineEntry() {
      ::LeaveCriticalSection(&g_cs);
    }
  };
}

// Called at DLL_PROCESS_ATTACH.
void SymbolEngine::pre_initialize() {
  ::InitializeCriticalSection(&g_cs);
}

bool SymbolEngine::decode(const void* addr, char* buf, int buflen, int* offset, bool do_demangle) {

  assert(buf != NULL && buflen > 0 && offset != NULL, "Argument error");
  buf[0] = '\0';
  *offset = -1;

  if (addr == NULL) {
    return false;
  }

  SymbolEngineEntry entry_guard;

  // Try decoding the symbol once. If we fail, attempt to rebuild the
  // symbol search path - maybe the pc points to a dll whose pdb file is
  // outside our search path. Then do attempt the decode again.
  bool success = decode_locked(addr, buf, buflen, offset, do_demangle);
  if (!success) {
    bool did_update_search_path = false;
    if (recalc_search_path_locked(&did_update_search_path)) {
      if (did_update_search_path) {
        success = decode_locked(addr, buf, buflen, offset, do_demangle);
      }
    }
  }

  return success;

}

bool SymbolEngine::demangle(const char* symbol, char *buf, int buflen) {

  SymbolEngineEntry entry_guard;

  return demangle_locked(symbol, buf, buflen);

}

bool SymbolEngine::recalc_search_path(bool* p_search_path_was_updated) {

  SymbolEngineEntry entry_guard;

  return recalc_search_path_locked(p_search_path_was_updated);

}

bool SymbolEngine::get_source_info(const void* addr, char* buf, size_t buflen,
                                   int* line_no)
{
  assert(buf != NULL && buflen > 0 && line_no != NULL, "Argument error");
  buf[0] = '\0';
  *line_no = -1;

  if (addr == NULL) {
    return false;
  }

  SymbolEngineEntry entry_guard;

  IMAGEHLP_LINE64 lineinfo;
  memset(&lineinfo, 0, sizeof(lineinfo));
  lineinfo.SizeOfStruct = sizeof(lineinfo);
  DWORD displacement;
  if (WindowsDbgHelp::symGetLineFromAddr64(::GetCurrentProcess(), (DWORD64)addr,
                                           &displacement, &lineinfo)) {
    if (buf != NULL && buflen > 0 && lineinfo.FileName != NULL) {
      // We only return the file name, not the whole path.
      char* p = lineinfo.FileName;
      char* q = strrchr(lineinfo.FileName, '\\');
      if (q) {
        p = q + 1;
      }
      ::strncpy(buf, p, buflen - 1);
      buf[buflen - 1] = '\0';
    }
    if (line_no != 0) {
      *line_no = lineinfo.LineNumber;
    }
    return true;
  }
  return false;
}

// Print one liner describing state (if library loaded, which functions are
// missing - if any, and the dbhelp API version)
void SymbolEngine::print_state_on(outputStream* st) {

  SymbolEngineEntry entry_guard;

  st->print("symbol engine: ");

  if (g_state == state_uninitialized) {
    st->print("uninitialized.");
  } else if (g_state == state_error) {
    st->print("initialization error.");
  } else {
    st->print("initialized successfully");
    st->print(" - sym options: 0x%X", WindowsDbgHelp::symGetOptions());
    st->print(" - pdb path: ");
    if (WindowsDbgHelp::symGetSearchPath(::GetCurrentProcess(),
                                          g_buffers.search_path.ptr(),
                                          (int)g_buffers.search_path.capacity())) {
      st->print_raw(g_buffers.search_path.ptr());
    } else {
      st->print_raw("(cannot be retrieved)");
    }
  }
  st->cr();

}

quality98%

¤ Dauer der Verarbeitung: 0.38 Sekunden (vorverarbeitet) ¤

Wurzel

Suchen

Beweissystem der NASA

Beweissystem Isabelle

NIST Cobol Testsuite

Cephes Mathematical Library

Wiener Entwicklungsmethode

Haftungshinweis

Die Informationen auf dieser Webseite wurden nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit, noch Qualität der bereit gestellten Informationen zugesichert.

Bemerkung:

Die farbliche Syntaxdarstellung ist noch experimentell.