diff options
Diffstat (limited to 'toolkit/crashreporter/google-breakpad/src/common/windows')
23 files changed, 4937 insertions, 0 deletions
diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/common_windows.gyp b/toolkit/crashreporter/google-breakpad/src/common/windows/common_windows.gyp new file mode 100644 index 0000000000..5f7594b161 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/common_windows.gyp @@ -0,0 +1,112 @@ +# Copyright 2013 Google Inc. All rights reserved. +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are +# met: +# +# * Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# * Redistributions in binary form must reproduce the above +# copyright notice, this list of conditions and the following disclaimer +# in the documentation and/or other materials provided with the +# distribution. +# * Neither the name of Google Inc. nor the names of its +# contributors may be used to endorse or promote products derived from +# this software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +{ + 'includes': [ + '../../build/common.gypi', + ], + 'targets': [ + { + 'target_name': 'dia_sdk', + 'type': 'none', + 'all_dependent_settings': { + 'include_dirs': [ + '<(DEPTH)', + '$(VSInstallDir)/DIA SDK/include', + ], + 'msvs_settings': { + 'VCLinkerTool': { + 'AdditionalDependencies': [ + 'diaguids.lib', + 'imagehlp.lib', + ], + }, + }, + 'configurations': { + 'x86_Base': { + 'msvs_settings': { + 'VCLinkerTool': { + 'AdditionalLibraryDirectories': + ['$(VSInstallDir)/DIA SDK/lib'], + }, + }, + }, + 'x64_Base': { + 'msvs_settings': { + 'VCLinkerTool': { + 'AdditionalLibraryDirectories': + ['$(VSInstallDir)/DIA SDK/lib/amd64'], + }, + }, + }, + }, + }, + }, + { + 'target_name': 'common_windows_lib', + 'type': 'static_library', + 'sources': [ + 'dia_util.cc', + 'dia_util.h', + 'guid_string.cc', + 'guid_string.h', + 'http_upload.cc', + 'http_upload.h', + 'module_info.h', + 'omap.cc', + 'omap.h', + 'omap_internal.h', + 'pdb_source_line_writer.cc', + 'pdb_source_line_writer.h', + 'pe_source_line_writer.cc', + 'pe_source_line_writer.h', + 'pe_util.h', + 'pe_util.cc', + 'string_utils.cc', + 'string_utils-inl.h', + 'symbol_collector_client.cc', + 'symbol_collector_client.h', + ], + 'dependencies': [ + 'dia_sdk', + ], + }, + { + 'target_name': 'common_windows_unittests', + 'type': 'executable', + 'sources': [ + 'omap_unittest.cc', + ], + 'dependencies': [ + '<(DEPTH)/client/windows/unittests/testing.gyp:gmock', + '<(DEPTH)/client/windows/unittests/testing.gyp:gtest', + 'common_windows_lib', + ], + }, + ], +} diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/dia_util.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/dia_util.cc new file mode 100644 index 0000000000..ed8cb5b658 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/dia_util.cc @@ -0,0 +1,92 @@ +// Copyright 2013 Google Inc. All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+// * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+// * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+// * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "common/windows/dia_util.h"
+
+#include <atlbase.h>
+
+namespace google_breakpad {
+
+bool FindDebugStream(const wchar_t* name,
+ IDiaSession* session,
+ IDiaEnumDebugStreamData** debug_stream) {
+ CComPtr<IDiaEnumDebugStreams> enum_debug_streams;
+ if (FAILED(session->getEnumDebugStreams(&enum_debug_streams))) {
+ fprintf(stderr, "IDiaSession::getEnumDebugStreams failed\n");
+ return false;
+ }
+
+ CComPtr<IDiaEnumDebugStreamData> temp_debug_stream;
+ ULONG fetched = 0;
+ while (SUCCEEDED(enum_debug_streams->Next(1, &temp_debug_stream, &fetched)) &&
+ fetched == 1) {
+ CComBSTR stream_name;
+ if (FAILED(temp_debug_stream->get_name(&stream_name))) {
+ fprintf(stderr, "IDiaEnumDebugStreamData::get_name failed\n");
+ return false;
+ }
+
+ // Found the stream?
+ if (wcsncmp((LPWSTR)stream_name, name, stream_name.Length()) == 0) {
+ *debug_stream = temp_debug_stream.Detach();
+ return true;
+ }
+
+ temp_debug_stream.Release();
+ }
+
+ // No table was found.
+ return false;
+}
+
+bool FindTable(REFIID iid, IDiaSession* session, void** table) {
+ // Get the table enumerator.
+ CComPtr<IDiaEnumTables> enum_tables;
+ if (FAILED(session->getEnumTables(&enum_tables))) {
+ fprintf(stderr, "IDiaSession::getEnumTables failed\n");
+ return false;
+ }
+
+ // Iterate through the tables.
+ CComPtr<IDiaTable> temp_table;
+ ULONG fetched = 0;
+ while (SUCCEEDED(enum_tables->Next(1, &temp_table, &fetched)) &&
+ fetched == 1) {
+ void* temp = NULL;
+ if (SUCCEEDED(temp_table->QueryInterface(iid, &temp))) {
+ *table = temp;
+ return true;
+ }
+ temp_table.Release();
+ }
+
+ // The table was not found.
+ return false;
+}
+
+} // namespace google_breakpad
\ No newline at end of file diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/dia_util.h b/toolkit/crashreporter/google-breakpad/src/common/windows/dia_util.h new file mode 100644 index 0000000000..b9e0df2d56 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/dia_util.h @@ -0,0 +1,64 @@ +// Copyright 2013 Google Inc. All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +// Utilities for loading debug streams and tables from a PDB file. + +#ifndef COMMON_WINDOWS_DIA_UTIL_H_ +#define COMMON_WINDOWS_DIA_UTIL_H_ + +#include <Windows.h> +#include <dia2.h> + +namespace google_breakpad { + +// Find the debug stream of the given |name| in the given |session|. Returns +// true on success, false on error of if the stream does not exist. On success +// the stream will be returned via |debug_stream|. +bool FindDebugStream(const wchar_t* name, + IDiaSession* session, + IDiaEnumDebugStreamData** debug_stream); + +// Finds the first table implementing the COM interface with ID |iid| in the +// given |session|. Returns true on success, false on error or if no such +// table is found. On success the table will be returned via |table|. +bool FindTable(REFIID iid, IDiaSession* session, void** table); + +// A templated version of FindTable. Finds the first table implementing type +// |InterfaceType| in the given |session|. Returns true on success, false on +// error or if no such table is found. On success the table will be returned via +// |table|. +template<typename InterfaceType> +bool FindTable(IDiaSession* session, InterfaceType** table) { + return FindTable(__uuidof(InterfaceType), + session, + reinterpret_cast<void**>(table)); +} + +} // namespace google_breakpad + +#endif // COMMON_WINDOWS_DIA_UTIL_H_ diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/guid_string.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/guid_string.cc new file mode 100644 index 0000000000..b7f877e66e --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/guid_string.cc @@ -0,0 +1,76 @@ +// Copyright (c) 2006, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +// guid_string.cc: Convert GUIDs to strings. +// +// See guid_string.h for documentation. + +#include <wchar.h> + +#include "common/windows/string_utils-inl.h" + +#include "common/windows/guid_string.h" + +namespace google_breakpad { + +// static +wstring GUIDString::GUIDToWString(GUID *guid) { + wchar_t guid_string[37]; + swprintf( + guid_string, sizeof(guid_string) / sizeof(guid_string[0]), + L"%08x-%04x-%04x-%02x%02x-%02x%02x%02x%02x%02x%02x", + guid->Data1, guid->Data2, guid->Data3, + guid->Data4[0], guid->Data4[1], guid->Data4[2], + guid->Data4[3], guid->Data4[4], guid->Data4[5], + guid->Data4[6], guid->Data4[7]); + + // remove when VC++7.1 is no longer supported + guid_string[sizeof(guid_string) / sizeof(guid_string[0]) - 1] = L'\0'; + + return wstring(guid_string); +} + +// static +wstring GUIDString::GUIDToSymbolServerWString(GUID *guid) { + wchar_t guid_string[33]; + swprintf( + guid_string, sizeof(guid_string) / sizeof(guid_string[0]), + L"%08X%04X%04X%02X%02X%02X%02X%02X%02X%02X%02X", + guid->Data1, guid->Data2, guid->Data3, + guid->Data4[0], guid->Data4[1], guid->Data4[2], + guid->Data4[3], guid->Data4[4], guid->Data4[5], + guid->Data4[6], guid->Data4[7]); + + // remove when VC++7.1 is no longer supported + guid_string[sizeof(guid_string) / sizeof(guid_string[0]) - 1] = L'\0'; + + return wstring(guid_string); +} + +} // namespace google_breakpad diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/guid_string.h b/toolkit/crashreporter/google-breakpad/src/common/windows/guid_string.h new file mode 100644 index 0000000000..48a5c1d370 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/guid_string.h @@ -0,0 +1,58 @@ +// Copyright (c) 2006, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +// guid_string.cc: Convert GUIDs to strings. + +#ifndef COMMON_WINDOWS_GUID_STRING_H_ +#define COMMON_WINDOWS_GUID_STRING_H_ + +#include <guiddef.h> + +#include <string> + +namespace google_breakpad { + +using std::wstring; + +class GUIDString { + public: + // Converts guid to a string in the format recommended by RFC 4122 and + // returns the string. + static wstring GUIDToWString(GUID *guid); + + // Converts guid to a string formatted as uppercase hexadecimal, with + // no separators, and returns the string. This is the format used for + // symbol server identifiers, although identifiers have an age tacked + // on to the string. + static wstring GUIDToSymbolServerWString(GUID *guid); +}; + +} // namespace google_breakpad + +#endif // COMMON_WINDOWS_GUID_STRING_H_ diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/http_upload.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/http_upload.cc new file mode 100644 index 0000000000..5df17e1a88 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/http_upload.cc @@ -0,0 +1,493 @@ +// Copyright (c) 2006, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#include <assert.h> + +// Disable exception handler warnings. +#pragma warning(disable:4530) + +#include <fstream> +#include <vector> + +#include "common/windows/string_utils-inl.h" + +#include "common/windows/http_upload.h" + +namespace { + using std::string; + using std::wstring; + using std::map; + using std::vector; + using std::ifstream; + using std::ios; + + const wchar_t kUserAgent[] = L"Breakpad/1.0 (Windows)"; + + // Helper class which closes an internet handle when it goes away + class AutoInternetHandle { + public: + explicit AutoInternetHandle(HINTERNET handle) : handle_(handle) {} + ~AutoInternetHandle() { + if (handle_) { + InternetCloseHandle(handle_); + } + } + + HINTERNET get() { return handle_; } + + private: + HINTERNET handle_; + }; + + wstring UTF8ToWide(const string &utf8) { + if (utf8.length() == 0) { + return wstring(); + } + + // compute the length of the buffer we'll need + int charcount = MultiByteToWideChar(CP_UTF8, 0, utf8.c_str(), -1, NULL, 0); + + if (charcount == 0) { + return wstring(); + } + + // convert + wchar_t* buf = new wchar_t[charcount]; + MultiByteToWideChar(CP_UTF8, 0, utf8.c_str(), -1, buf, charcount); + wstring result(buf); + delete[] buf; + return result; + } + + string WideToMBCP(const wstring &wide, unsigned int cp) { + if (wide.length() == 0) { + return string(); + } + + // compute the length of the buffer we'll need + int charcount = WideCharToMultiByte(cp, 0, wide.c_str(), -1, + NULL, 0, NULL, NULL); + if (charcount == 0) { + return string(); + } + + // convert + char *buf = new char[charcount]; + WideCharToMultiByte(cp, 0, wide.c_str(), -1, buf, charcount, + NULL, NULL); + + string result(buf); + delete[] buf; + return result; + } + + bool GetFileContents(const wstring &filename, vector<char> *contents) { + bool rv = false; + // The "open" method on pre-MSVC8 ifstream implementations doesn't accept a + // wchar_t* filename, so use _wfopen directly in that case. For VC8 and + // later, _wfopen has been deprecated in favor of _wfopen_s, which does + // not exist in earlier versions, so let the ifstream open the file itself. + // GCC doesn't support wide file name and opening on FILE* requires ugly + // hacks, so fallback to multi byte file. +#ifdef _MSC_VER + ifstream file; + file.open(filename.c_str(), ios::binary); +#else // GCC + ifstream file(WideToMBCP(filename, CP_ACP).c_str(), ios::binary); +#endif // _MSC_VER >= 1400 + if (file.is_open()) { + file.seekg(0, ios::end); + std::streamoff length = file.tellg(); + // Check for loss of data when converting lenght from std::streamoff into + // std::vector<char>::size_type + std::vector<char>::size_type vector_size = + static_cast<std::vector<char>::size_type>(length); + if (static_cast<std::streamoff>(vector_size) == length) { + contents->resize(vector_size); + if (length != 0) { + file.seekg(0, ios::beg); + file.read(&((*contents)[0]), length); + } + rv = true; + } + file.close(); + } + return rv; + } + + // Converts a UTF16 string to UTF8. + string WideToUTF8(const wstring &wide) { + return WideToMBCP(wide, CP_UTF8); + } + + bool ReadResponse(HINTERNET request, wstring *response) { + bool has_content_length_header = false; + wchar_t content_length[32]; + DWORD content_length_size = sizeof(content_length); + DWORD claimed_size = 0; + string response_body; + + if (HttpQueryInfo(request, HTTP_QUERY_CONTENT_LENGTH, + static_cast<LPVOID>(&content_length), + &content_length_size, 0)) { + has_content_length_header = true; + claimed_size = wcstol(content_length, NULL, 10); + response_body.reserve(claimed_size); + } + + DWORD bytes_available; + DWORD total_read = 0; + BOOL return_code; + + while (((return_code = InternetQueryDataAvailable(request, &bytes_available, + 0, 0)) != 0) && bytes_available > 0) { + vector<char> response_buffer(bytes_available); + DWORD size_read; + + return_code = InternetReadFile(request, + &response_buffer[0], + bytes_available, &size_read); + + if (return_code && size_read > 0) { + total_read += size_read; + response_body.append(&response_buffer[0], size_read); + } + else { + break; + } + } + + bool succeeded = return_code && (!has_content_length_header || + (total_read == claimed_size)); + if (succeeded && response) { + *response = UTF8ToWide(response_body); + } + + return succeeded; + } + + bool SendRequestInner( + const wstring& url, + const wstring& http_method, + const wstring& content_type_header, + const string& request_body, + int* timeout_ms, + wstring* response_body, + int* response_code) { + if (response_code) { + *response_code = 0; + } + + // Break up the URL and make sure we can handle it + wchar_t scheme[16], host[256], path[1024]; + URL_COMPONENTS components; + memset(&components, 0, sizeof(components)); + components.dwStructSize = sizeof(components); + components.lpszScheme = scheme; + components.dwSchemeLength = sizeof(scheme) / sizeof(scheme[0]); + components.lpszHostName = host; + components.dwHostNameLength = sizeof(host) / sizeof(host[0]); + components.lpszUrlPath = path; + components.dwUrlPathLength = sizeof(path) / sizeof(path[0]); + if (!InternetCrackUrl(url.c_str(), static_cast<DWORD>(url.size()), + 0, &components)) { + DWORD err = GetLastError(); + wprintf(L"%d\n", err); + return false; + } + bool secure = false; + if (wcscmp(scheme, L"https") == 0) { + secure = true; + } + else if (wcscmp(scheme, L"http") != 0) { + return false; + } + + AutoInternetHandle internet(InternetOpen(kUserAgent, + INTERNET_OPEN_TYPE_PRECONFIG, + NULL, // proxy name + NULL, // proxy bypass + 0)); // flags + if (!internet.get()) { + return false; + } + + AutoInternetHandle connection(InternetConnect(internet.get(), + host, + components.nPort, + NULL, // user name + NULL, // password + INTERNET_SERVICE_HTTP, + 0, // flags + NULL)); // context + if (!connection.get()) { + return false; + } + + DWORD http_open_flags = secure ? INTERNET_FLAG_SECURE : 0; + http_open_flags |= INTERNET_FLAG_NO_COOKIES; + AutoInternetHandle request(HttpOpenRequest(connection.get(), + http_method.c_str(), + path, + NULL, // version + NULL, // referer + NULL, // agent type + http_open_flags, + NULL)); // context + if (!request.get()) { + return false; + } + + if (!content_type_header.empty()) { + HttpAddRequestHeaders(request.get(), + content_type_header.c_str(), + static_cast<DWORD>(-1), + HTTP_ADDREQ_FLAG_ADD); + } + + if (timeout_ms) { + if (!InternetSetOption(request.get(), + INTERNET_OPTION_SEND_TIMEOUT, + timeout_ms, + sizeof(*timeout_ms))) { + fwprintf(stderr, L"Could not unset send timeout, continuing...\n"); + } + + if (!InternetSetOption(request.get(), + INTERNET_OPTION_RECEIVE_TIMEOUT, + timeout_ms, + sizeof(*timeout_ms))) { + fwprintf(stderr, L"Could not unset receive timeout, continuing...\n"); + } + } + + if (!HttpSendRequest(request.get(), NULL, 0, + const_cast<char *>(request_body.data()), + static_cast<DWORD>(request_body.size()))) { + return false; + } + + // The server indicates a successful upload with HTTP status 200. + wchar_t http_status[4]; + DWORD http_status_size = sizeof(http_status); + if (!HttpQueryInfo(request.get(), HTTP_QUERY_STATUS_CODE, + static_cast<LPVOID>(&http_status), &http_status_size, + 0)) { + return false; + } + + int http_response = wcstol(http_status, NULL, 10); + if (response_code) { + *response_code = http_response; + } + + bool result = (http_response == 200); + + if (result) { + result = ReadResponse(request.get(), response_body); + } + + return result; + } + + wstring GenerateMultipartBoundary() { + // The boundary has 27 '-' characters followed by 16 hex digits + static const wchar_t kBoundaryPrefix[] = L"---------------------------"; + static const int kBoundaryLength = 27 + 16 + 1; + + // Generate some random numbers to fill out the boundary + int r0 = rand(); + int r1 = rand(); + + wchar_t temp[kBoundaryLength]; + swprintf(temp, kBoundaryLength, L"%s%08X%08X", kBoundaryPrefix, r0, r1); + + // remove when VC++7.1 is no longer supported + temp[kBoundaryLength - 1] = L'\0'; + + return wstring(temp); + } + + wstring GenerateMultipartPostRequestHeader(const wstring &boundary) { + wstring header = L"Content-Type: multipart/form-data; boundary="; + header += boundary; + return header; + } + + bool AppendFileToRequestBody( + const wstring& file_part_name, + const wstring& filename, + string* request_body) { + string file_part_name_utf8 = WideToUTF8(file_part_name); + if (file_part_name_utf8.empty()) { + return false; + } + + string filename_utf8 = WideToUTF8(filename); + if (filename_utf8.empty()) { + return false; + } + + request_body->append("Content-Disposition: form-data; " + "name=\"" + file_part_name_utf8 + "\"; " + "filename=\"" + filename_utf8 + "\"\r\n"); + request_body->append("Content-Type: application/octet-stream\r\n"); + request_body->append("\r\n"); + + vector<char> contents; + if (!GetFileContents(filename, &contents)) { + return false; + } + + if (!contents.empty()) { + request_body->append(&(contents[0]), contents.size()); + } + request_body->append("\r\n"); + + return true; + } + + bool GenerateRequestBody(const string ¶meters, + const map<wstring, wstring> &files, + const wstring &boundary, + string *request_body) { + string boundary_str = WideToUTF8(boundary); + if (boundary_str.empty()) { + return false; + } + + request_body->clear(); + + // Append the extra data as a single JSON form entry + request_body->append("--" + boundary_str + "\r\n"); + request_body->append( + "Content-Disposition: form-data; " + "name=\"extra\"; " + "filename=\"extra.json\"\r\n"); + request_body->append("Content-Type: application/json\r\n"); + request_body->append("\r\n"); + + if (!parameters.empty()) { + request_body->append(parameters); + } + request_body->append("\r\n"); + + // Now append each upload file as a binary (octet-stream) part + for (map<wstring, wstring>::const_iterator pos = files.begin(); + pos != files.end(); ++pos) { + request_body->append("--" + boundary_str + "\r\n"); + + if (!AppendFileToRequestBody(pos->first, pos->second, request_body)) { + return false; + } + } + request_body->append("--" + boundary_str + "--\r\n"); + return true; + } +} + +namespace google_breakpad { + bool HTTPUpload::SendPutRequest( + const wstring& url, + const wstring& path, + int* timeout_ms, + wstring* response_body, + int* response_code) { + string request_body; + if (!AppendFileToRequestBody(L"symbol_file", path, &request_body)) { + return false; + } + + return SendRequestInner( + url, + L"PUT", + L"", + request_body, + timeout_ms, + response_body, + response_code); + } + + bool HTTPUpload::SendGetRequest( + const wstring& url, + int* timeout_ms, + wstring* response_body, + int* response_code) { + return SendRequestInner( + url, + L"GET", + L"", + "", + timeout_ms, + response_body, + response_code); + } + + bool HTTPUpload::SendMultipartPostRequest( + const wstring& url, + const string& parameters, + const map<wstring, wstring>& files, + int* timeout_ms, + wstring* response_body, + int* response_code) { + wstring boundary = GenerateMultipartBoundary(); + wstring content_type_header = GenerateMultipartPostRequestHeader(boundary); + + string request_body; + if (!GenerateRequestBody(parameters, files, boundary, &request_body)) { + return false; + } + + return SendRequestInner( + url, + L"POST", + content_type_header, + request_body, + timeout_ms, + response_body, + response_code); + } + + bool HTTPUpload::SendSimplePostRequest( + const wstring& url, + const wstring& body, + const wstring& content_type, + int *timeout_ms, + wstring *response_body, + int *response_code) { + return SendRequestInner( + url, + L"POST", + content_type, + WideToUTF8(body), + timeout_ms, + response_body, + response_code); + } +} // namespace google_breakpad diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/http_upload.h b/toolkit/crashreporter/google-breakpad/src/common/windows/http_upload.h new file mode 100644 index 0000000000..1e47f5829f --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/http_upload.h @@ -0,0 +1,125 @@ +// Copyright (c) 2006, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +// HTTPUpload provides a "nice" API to send a multipart HTTP(S) POST +// request using wininet. It currently supports requests that contain +// parameters encoded in a JSON string, and a file to upload. + +#ifndef COMMON_WINDOWS_HTTP_UPLOAD_H_ +#define COMMON_WINDOWS_HTTP_UPLOAD_H_ + +#pragma warning(push) +// Disable exception handler warnings. +#pragma warning(disable : 4530) + +#include <windows.h> +#include <wininet.h> + +#include <map> + +namespace google_breakpad { + +using std::map; +using std::string; +using std::wstring; + +class HTTPUpload { + public: + // Sends a PUT request containing the data in |path| to the given + // URL. + // Only HTTP(S) URLs are currently supported. Returns true on success. + // If the request is successful and response_body is non-NULL, + // the response body will be returned in response_body. + // If response_code is non-NULL, it will be set to the HTTP response code + // received (or 0 if the request failed before getting an HTTP response). + static bool SendPutRequest( + const wstring& url, + const wstring& path, + int* timeout_ms, + wstring* response_body, + int* response_code); + + // Sends a GET request to the given URL. + // Only HTTP(S) URLs are currently supported. Returns true on success. + // If the request is successful and response_body is non-NULL, + // the response body will be returned in response_body. + // If response_code is non-NULL, it will be set to the HTTP response code + // received (or 0 if the request failed before getting an HTTP response). + static bool SendGetRequest( + const wstring& url, + int* timeout_ms, + wstring* response_body, + int* response_code); + + // Sends the given sets of parameters and files as a multipart POST + // request to the given URL. + // Each key in |files| is the name of the file part of the request + // (i.e. it corresponds to the name= attribute on an <input type="file">. + // Parameters are specified as a JSON-encoded string in |parameters|. + // Only HTTP(S) URLs are currently supported. Returns true on success. + // If the request is successful and response_body is non-NULL, + // the response body will be returned in response_body. + // If response_code is non-NULL, it will be set to the HTTP response code + // received (or 0 if the request failed before getting an HTTP response). + static bool SendMultipartPostRequest( + const wstring& url, + const string& parameters, + const map<wstring, wstring>& files, + int *timeout_ms, + wstring *response_body, + int *response_code); + + // Sends a POST request, with the body set to |body|, to the given URL. + // Only HTTP(S) URLs are currently supported. Returns true on success. + // If the request is successful and response_body is non-NULL, + // the response body will be returned in response_body. + // If response_code is non-NULL, it will be set to the HTTP response code + // received (or 0 if the request failed before getting an HTTP response). + static bool SendSimplePostRequest( + const wstring& url, + const wstring& body, + const wstring& content_type, + int *timeout_ms, + wstring *response_body, + int *response_code); + + private: + // No instances of this class should be created. + // Disallow all constructors, destructors, and operator=. + HTTPUpload(); + explicit HTTPUpload(const HTTPUpload &); + void operator=(const HTTPUpload &); + ~HTTPUpload(); +}; + +} // namespace google_breakpad + +#pragma warning(pop) + +#endif // COMMON_WINDOWS_HTTP_UPLOAD_H_ diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/module_info.h b/toolkit/crashreporter/google-breakpad/src/common/windows/module_info.h new file mode 100644 index 0000000000..3dccc80882 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/module_info.h @@ -0,0 +1,75 @@ +// Copyright (c) 2019, Google Inc.
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+// * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+// * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+// * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef COMMON_WINDOWS_MODULE_INFO_H_
+#define COMMON_WINDOWS_MODULE_INFO_H_
+
+#include <string>
+
+namespace google_breakpad {
+
+using std::wstring;
+// A structure that carries information that identifies a module.
+struct PDBModuleInfo {
+public:
+ // The basename of the pe/pdb file from which information was loaded.
+ wstring debug_file;
+
+ // The module's identifier. For recent pe/pdb files, the identifier consists
+ // of the pe/pdb's guid, in uppercase hexadecimal form without any dashes
+ // or separators, followed immediately by the pe/pdb's age, also in
+ // uppercase hexadecimal form. For older pe/pdb files which have no guid,
+ // the identifier is the pe/pdb's 32-bit signature value, in zero-padded
+ // hexadecimal form, followed immediately by the pe/pdb's age, in lowercase
+ // hexadecimal form.
+ wstring debug_identifier;
+
+ // A string identifying the cpu that the pe/pdb is associated with.
+ // Currently, this may be "x86" or "unknown".
+ wstring cpu;
+};
+
+// A structure that carries information that identifies a PE file,
+// either an EXE or a DLL.
+struct PEModuleInfo {
+ // The basename of the PE file.
+ wstring code_file;
+
+ // The PE file's code identifier, which consists of its timestamp
+ // and file size concatenated together into a single hex string.
+ // (The fields IMAGE_OPTIONAL_HEADER::SizeOfImage and
+ // IMAGE_FILE_HEADER::TimeDateStamp, as defined in the ImageHlp
+ // documentation.) This is not well documented, if it's documented
+ // at all, but it's what symstore does and what DbgHelp supports.
+ wstring code_identifier;
+};
+
+} // namespace google_breakpad
+
+#endif // COMMON_WINDOWS_MODULE_INFO_H_
diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/objs.mozbuild b/toolkit/crashreporter/google-breakpad/src/common/windows/objs.mozbuild new file mode 100644 index 0000000000..5fcdae5a35 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/objs.mozbuild @@ -0,0 +1,15 @@ +# -*- Mode: python; indent-tabs-mode: nil; tab-width: 40 -*- +# vim: set filetype=python: +# This Source Code Form is subject to the terms of the Mozilla Public +# License, v. 2.0. If a copy of the MPL was not distributed with this +# file, You can obtain one at http://mozilla.org/MPL/2.0/. + +lobjs_common = [ + 'guid_string.cc', + 'string_utils.cc', +] + +subdir = 'toolkit/crashreporter/google-breakpad/src/common/windows' +objs_common = [ + '/%s/%s' % (subdir, s) for s in lobjs_common +] diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/omap.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/omap.cc new file mode 100644 index 0000000000..ba3ce86b8b --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/omap.cc @@ -0,0 +1,716 @@ +// Copyright 2013 Google Inc. All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +// This contains a suite of tools for transforming symbol information when +// when that information has been extracted from a PDB containing OMAP +// information. + +// OMAP information is a lightweight description of a mapping between two +// address spaces. It consists of two streams, each of them a vector 2-tuples. +// The OMAPTO stream contains tuples of the form +// +// (RVA in transformed image, RVA in original image) +// +// while the OMAPFROM stream contains tuples of the form +// +// (RVA in original image, RVA in transformed image) +// +// The entries in each vector are sorted by the first value of the tuple, and +// the lengths associated with a mapping are implicit as the distance between +// two successive addresses in the vector. + +// Consider a trivial 10-byte function described by the following symbol: +// +// Function: RVA 0x00001000, length 10, "foo" +// +// Now consider the same function, but with 5-bytes of instrumentation injected +// at offset 5. The OMAP streams describing this would look like: +// +// OMAPTO : [ [0x00001000, 0x00001000], +// [0x00001005, 0xFFFFFFFF], +// [0x0000100a, 0x00001005] ] +// OMAPFROM: [ [0x00001000, 0x00001000], +// [0x00001005, 0x0000100a] ] +// +// In this case the injected code has been marked as not originating in the +// source image, and thus it will have no symbol information at all. However, +// the injected code may also be associated with an original address range; +// for example, when prepending instrumentation to a basic block the +// instrumentation can be labelled as originating from the same source BB such +// that symbol resolution will still find the appropriate source code line +// number. In this case the OMAP stream would look like: +// +// OMAPTO : [ [0x00001000, 0x00001000], +// [0x00001005, 0x00001005], +// [0x0000100a, 0x00001005] ] +// OMAPFROM: [ [0x00001000, 0x00001000], +// [0x00001005, 0x0000100a] ] +// +// Suppose we asked DIA to lookup the symbol at location 0x0000100a of the +// instrumented image. It would first run this through the OMAPTO table and +// translate that address to 0x00001005. It would then lookup the symbol +// at that address and return the symbol for the function "foo". This is the +// correct result. +// +// However, if we query DIA for the length and address of the symbol it will +// tell us that it has length 10 and is at RVA 0x00001000. The location is +// correct, but the length doesn't take into account the 5-bytes of injected +// code. Symbol resolution works (starting from an instrumented address, +// mapping to an original address, and looking up a symbol), but the symbol +// metadata is incorrect. +// +// If we dump the symbols using DIA they will have their addresses +// appropriately transformed and reflect positions in the instrumented image. +// However, if we try to do a lookup using those symbols resolution can fail. +// For example, the address 0x0000100a will not map to the symbol for "foo", +// because DIA tells us it is at location 0x00001000 (correct) with length +// 10 (incorrect). The problem is one of order of operations: in this case +// we're attempting symbol resolution by looking up an instrumented address +// in the table of translated symbols. +// +// One way to handle this is to dump the OMAP information as part of the +// breakpad symbols. This requires the rest of the toolchain to be aware of +// OMAP information and to use it when present prior to performing lookup. The +// other option is to properly transform the symbols (updating length as well as +// position) so that resolution will work as expected for translated addresses. +// This is transparent to the rest of the toolchain. + +#include "common/windows/omap.h" + +#include <atlbase.h> + +#include <algorithm> +#include <cassert> +#include <set> + +#include "common/windows/dia_util.h" + +namespace google_breakpad { + +namespace { + +static const wchar_t kOmapToDebugStreamName[] = L"OMAPTO"; +static const wchar_t kOmapFromDebugStreamName[] = L"OMAPFROM"; + +// Dependending on where this is used in breakpad we sometimes get min/max from +// windef, and other times from algorithm. To get around this we simply +// define our own min/max functions. +template<typename T> +const T& Min(const T& t1, const T& t2) { return t1 < t2 ? t1 : t2; } +template<typename T> +const T& Max(const T& t1, const T& t2) { return t1 > t2 ? t1 : t2; } + +// It makes things more readable to have two different OMAP types. We cast +// normal OMAPs into these. They must be the same size as the OMAP structure +// for this to work, hence the static asserts. +struct OmapOrigToTran { + DWORD rva_original; + DWORD rva_transformed; +}; +struct OmapTranToOrig { + DWORD rva_transformed; + DWORD rva_original; +}; +static_assert(sizeof(OmapOrigToTran) == sizeof(OMAP), + "OmapOrigToTran must have same size as OMAP."); +static_assert(sizeof(OmapTranToOrig) == sizeof(OMAP), + "OmapTranToOrig must have same size as OMAP."); +typedef std::vector<OmapOrigToTran> OmapFromTable; +typedef std::vector<OmapTranToOrig> OmapToTable; + +// Used for sorting and searching through a Mapping. +bool MappedRangeOriginalLess(const MappedRange& lhs, const MappedRange& rhs) { + if (lhs.rva_original < rhs.rva_original) + return true; + if (lhs.rva_original > rhs.rva_original) + return false; + return lhs.length < rhs.length; +} +bool MappedRangeMappedLess(const MappedRange& lhs, const MappedRange& rhs) { + if (lhs.rva_transformed < rhs.rva_transformed) + return true; + if (lhs.rva_transformed > rhs.rva_transformed) + return false; + return lhs.length < rhs.length; +} + +// Used for searching through the EndpointIndexMap. +bool EndpointIndexLess(const EndpointIndex& ei1, const EndpointIndex& ei2) { + return ei1.endpoint < ei2.endpoint; +} + +// Finds the debug stream with the given |name| in the given |session|, and +// populates |table| with its contents. Casts the data directly into OMAP +// structs. +bool FindAndLoadOmapTable(const wchar_t* name, + IDiaSession* session, + OmapTable* table) { + assert(name != NULL); + assert(session != NULL); + assert(table != NULL); + + CComPtr<IDiaEnumDebugStreamData> stream; + if (!FindDebugStream(name, session, &stream)) + return false; + assert(stream.p != NULL); + + LONG count = 0; + if (FAILED(stream->get_Count(&count))) { + fprintf(stderr, "IDiaEnumDebugStreamData::get_Count failed for stream " + "\"%ws\"\n", name); + return false; + } + + // Get the length of the stream in bytes. + DWORD bytes_read = 0; + ULONG count_read = 0; + if (FAILED(stream->Next(count, 0, &bytes_read, NULL, &count_read))) { + fprintf(stderr, "IDiaEnumDebugStreamData::Next failed while reading " + "length of stream \"%ws\"\n", name); + return false; + } + + // Ensure it's consistent with the OMAP data type. + DWORD bytes_expected = count * sizeof(OmapTable::value_type); + if (count * sizeof(OmapTable::value_type) != bytes_read) { + fprintf(stderr, "DIA debug stream \"%ws\" has an unexpected length", name); + return false; + } + + // Read the table. + table->resize(count); + bytes_read = 0; + count_read = 0; + if (FAILED(stream->Next(count, bytes_expected, &bytes_read, + reinterpret_cast<BYTE*>(&table->at(0)), + &count_read))) { + fprintf(stderr, "IDiaEnumDebugStreamData::Next failed while reading " + "data from stream \"%ws\"\n", name); + return false; + } + + return true; +} + +// This determines the original image length by looking through the segment +// table. +bool GetOriginalImageLength(IDiaSession* session, DWORD* image_length) { + assert(session != NULL); + assert(image_length != NULL); + + CComPtr<IDiaEnumSegments> enum_segments; + if (!FindTable(session, &enum_segments)) + return false; + assert(enum_segments.p != NULL); + + DWORD temp_image_length = 0; + CComPtr<IDiaSegment> segment; + ULONG fetched = 0; + while (SUCCEEDED(enum_segments->Next(1, &segment, &fetched)) && + fetched == 1) { + assert(segment.p != NULL); + + DWORD rva = 0; + DWORD length = 0; + DWORD frame = 0; + if (FAILED(segment->get_relativeVirtualAddress(&rva)) || + FAILED(segment->get_length(&length)) || + FAILED(segment->get_frame(&frame))) { + fprintf(stderr, "Failed to get basic properties for IDiaSegment\n"); + return false; + } + + if (frame > 0) { + DWORD segment_end = rva + length; + if (segment_end > temp_image_length) + temp_image_length = segment_end; + } + segment.Release(); + } + + *image_length = temp_image_length; + return true; +} + +// Detects regions of the original image that have been removed in the +// transformed image, and sets the 'removed' property on all mapped ranges +// immediately preceding a gap. The mapped ranges must be sorted by +// 'rva_original'. +void FillInRemovedLengths(Mapping* mapping) { + assert(mapping != NULL); + + // Find and fill gaps. We do this with two sweeps. We first sweep forward + // looking for gaps. When we identify a gap we then sweep forward with a + // second scan and set the 'removed' property for any intervals that + // immediately precede the gap. + // + // Gaps are typically between two successive intervals, but not always: + // + // Range 1: --------------- + // Range 2: ------- + // Range 3: ------------- + // Gap : ****** + // + // In the above example the gap is between range 1 and range 3. A forward + // sweep finds the gap, and a second forward sweep identifies that range 1 + // immediately precedes the gap and sets its 'removed' property. + + size_t fill = 0; + DWORD rva_front = 0; + for (size_t find = 0; find < mapping->size(); ++find) { +#ifndef NDEBUG + // We expect the mapped ranges to be sorted by 'rva_original'. + if (find > 0) { + assert(mapping->at(find - 1).rva_original <= + mapping->at(find).rva_original); + } +#endif + + if (rva_front < mapping->at(find).rva_original) { + // We've found a gap. Fill it in by setting the 'removed' property for + // any affected intervals. + DWORD removed = mapping->at(find).rva_original - rva_front; + for (; fill < find; ++fill) { + if (mapping->at(fill).rva_original + mapping->at(fill).length != + rva_front) { + continue; + } + + // This interval ends right where the gap starts. It needs to have its + // 'removed' information filled in. + mapping->at(fill).removed = removed; + } + } + + // Advance the front that indicates the covered portion of the image. + rva_front = mapping->at(find).rva_original + mapping->at(find).length; + } +} + +// Builds a unified view of the mapping between the original and transformed +// image space by merging OMAPTO and OMAPFROM data. +void BuildMapping(const OmapData& omap_data, Mapping* mapping) { + assert(mapping != NULL); + + mapping->clear(); + + if (omap_data.omap_from.empty() || omap_data.omap_to.empty()) + return; + + // The names 'omap_to' and 'omap_from' are awfully confusing, so we make + // ourselves more explicit here. This cast is only safe because the underlying + // types have the exact same size. + const OmapToTable& tran2orig = + reinterpret_cast<const OmapToTable&>(omap_data.omap_to); + const OmapFromTable& orig2tran = reinterpret_cast<const OmapFromTable&>( + omap_data.omap_from); + + // Handle the range of data at the beginning of the image. This is not usually + // specified by the OMAP data. + if (tran2orig[0].rva_transformed > 0 && orig2tran[0].rva_original > 0) { + DWORD header_transformed = tran2orig[0].rva_transformed; + DWORD header_original = orig2tran[0].rva_original; + DWORD header = Min(header_transformed, header_original); + + MappedRange mr = {}; + mr.length = header; + mr.injected = header_transformed - header; + mr.removed = header_original - header; + mapping->push_back(mr); + } + + // Convert the implied lengths to explicit lengths, and infer which content + // has been injected into the transformed image. Injected content is inferred + // as regions of the transformed address space that does not map back to + // known valid content in the original image. + for (size_t i = 0; i < tran2orig.size(); ++i) { + const OmapTranToOrig& o1 = tran2orig[i]; + + // This maps to content that is outside the original image, thus it + // describes injected content. We can skip this entry. + if (o1.rva_original >= omap_data.length_original) + continue; + + // Calculate the length of the current OMAP entry. This is implicit as the + // distance between successive |rva| values, capped at the end of the + // original image. + DWORD length = 0; + if (i + 1 < tran2orig.size()) { + const OmapTranToOrig& o2 = tran2orig[i + 1]; + + // We expect the table to be sorted by rva_transformed. + assert(o1.rva_transformed <= o2.rva_transformed); + + length = o2.rva_transformed - o1.rva_transformed; + if (o1.rva_original + length > omap_data.length_original) { + length = omap_data.length_original - o1.rva_original; + } + } else { + length = omap_data.length_original - o1.rva_original; + } + + // Zero-length entries don't describe anything and can be ignored. + if (length == 0) + continue; + + // Any gaps in the transformed address-space are due to injected content. + if (!mapping->empty()) { + MappedRange& prev_mr = mapping->back(); + prev_mr.injected += o1.rva_transformed - + (prev_mr.rva_transformed + prev_mr.length); + } + + MappedRange mr = {}; + mr.rva_original = o1.rva_original; + mr.rva_transformed = o1.rva_transformed; + mr.length = length; + mapping->push_back(mr); + } + + // Sort based on the original image addresses. + std::sort(mapping->begin(), mapping->end(), MappedRangeOriginalLess); + + // Fill in the 'removed' lengths by looking for gaps in the coverage of the + // original address space. + FillInRemovedLengths(mapping); + + return; +} + +void BuildEndpointIndexMap(ImageMap* image_map) { + assert(image_map != NULL); + + if (image_map->mapping.size() == 0) + return; + + const Mapping& mapping = image_map->mapping; + EndpointIndexMap& eim = image_map->endpoint_index_map; + + // Get the unique set of interval endpoints. + std::set<DWORD> endpoints; + for (size_t i = 0; i < mapping.size(); ++i) { + endpoints.insert(mapping[i].rva_original); + endpoints.insert(mapping[i].rva_original + + mapping[i].length + + mapping[i].removed); + } + + // Use the endpoints to initialize the secondary search structure for the + // mapping. + eim.resize(endpoints.size()); + std::set<DWORD>::const_iterator it = endpoints.begin(); + for (size_t i = 0; it != endpoints.end(); ++it, ++i) { + eim[i].endpoint = *it; + eim[i].index = mapping.size(); + } + + // For each endpoint we want the smallest index of any interval containing + // it. We iterate over the intervals and update the indices associated with + // each interval endpoint contained in the current interval. In the general + // case of an arbitrary set of intervals this is O(n^2), but the structure of + // OMAP data makes this O(n). + for (size_t i = 0; i < mapping.size(); ++i) { + EndpointIndex ei1 = { mapping[i].rva_original, 0 }; + EndpointIndexMap::iterator it1 = std::lower_bound( + eim.begin(), eim.end(), ei1, EndpointIndexLess); + + EndpointIndex ei2 = { mapping[i].rva_original + mapping[i].length + + mapping[i].removed, 0 }; + EndpointIndexMap::iterator it2 = std::lower_bound( + eim.begin(), eim.end(), ei2, EndpointIndexLess); + + for (; it1 != it2; ++it1) + it1->index = Min(i, it1->index); + } +} + +void BuildSubsequentRVAMap(const OmapData &omap_data, + std::map<DWORD, DWORD> *subsequent) { + assert(subsequent->empty()); + const OmapFromTable &orig2tran = + reinterpret_cast<const OmapFromTable &>(omap_data.omap_from); + + if (orig2tran.empty()) + return; + + for (size_t i = 0; i < orig2tran.size() - 1; ++i) { + // Expect that orig2tran is sorted. + if (orig2tran[i].rva_original >= orig2tran[i + 1].rva_original) { + fprintf(stderr, "OMAP 'from' table unexpectedly unsorted\n"); + subsequent->clear(); + return; + } + subsequent->insert(std::make_pair(orig2tran[i].rva_original, + orig2tran[i + 1].rva_original)); + } +} + +// Clips the given mapped range. +void ClipMappedRangeOriginal(const AddressRange& clip_range, + MappedRange* mapped_range) { + assert(mapped_range != NULL); + + // The clipping range is entirely outside of the mapped range. + if (clip_range.end() <= mapped_range->rva_original || + mapped_range->rva_original + mapped_range->length + + mapped_range->removed <= clip_range.rva) { + mapped_range->length = 0; + mapped_range->injected = 0; + mapped_range->removed = 0; + return; + } + + // Clip the left side. + if (mapped_range->rva_original < clip_range.rva) { + DWORD clip_left = clip_range.rva - mapped_range->rva_original; + mapped_range->rva_original += clip_left; + mapped_range->rva_transformed += clip_left; + + if (clip_left > mapped_range->length) { + // The left clipping boundary entirely erases the content section of the + // range. + DWORD trim = clip_left - mapped_range->length; + mapped_range->length = 0; + mapped_range->injected -= Min(trim, mapped_range->injected); + // We know that trim <= mapped_range->remove. + mapped_range->removed -= trim; + } else { + // The left clipping boundary removes some, but not all, of the content. + // As such it leaves the removed/injected component intact. + mapped_range->length -= clip_left; + } + } + + // Clip the right side. + DWORD end_original = mapped_range->rva_original + mapped_range->length; + if (clip_range.end() < end_original) { + // The right clipping boundary lands in the 'content' section of the range, + // entirely clearing the injected/removed portion. + DWORD clip_right = end_original - clip_range.end(); + mapped_range->length -= clip_right; + mapped_range->injected = 0; + mapped_range->removed = 0; + return; + } else { + // The right clipping boundary is outside of the content, but may affect + // the removed/injected portion of the range. + DWORD end_removed = end_original + mapped_range->removed; + if (clip_range.end() < end_removed) + mapped_range->removed = clip_range.end() - end_original; + + DWORD end_injected = end_original + mapped_range->injected; + if (clip_range.end() < end_injected) + mapped_range->injected = clip_range.end() - end_original; + } + + return; +} + +} // namespace + +int AddressRange::Compare(const AddressRange& rhs) const { + if (end() <= rhs.rva) + return -1; + if (rhs.end() <= rva) + return 1; + return 0; +} + +bool GetOmapDataAndDisableTranslation(IDiaSession* session, + OmapData* omap_data) { + assert(session != NULL); + assert(omap_data != NULL); + + CComPtr<IDiaAddressMap> address_map; + if (FAILED(session->QueryInterface(&address_map))) { + fprintf(stderr, "IDiaSession::QueryInterface(IDiaAddressMap) failed\n"); + return false; + } + assert(address_map.p != NULL); + + BOOL omap_enabled = FALSE; + if (FAILED(address_map->get_addressMapEnabled(&omap_enabled))) { + fprintf(stderr, "IDiaAddressMap::get_addressMapEnabled failed\n"); + return false; + } + + if (!omap_enabled) { + // We indicate the non-presence of OMAP data by returning empty tables. + omap_data->omap_from.clear(); + omap_data->omap_to.clear(); + omap_data->length_original = 0; + return true; + } + + // OMAP data is present. Disable translation. + if (FAILED(address_map->put_addressMapEnabled(FALSE))) { + fprintf(stderr, "IDiaAddressMap::put_addressMapEnabled failed\n"); + return false; + } + + // Read the OMAP streams. + if (!FindAndLoadOmapTable(kOmapFromDebugStreamName, + session, + &omap_data->omap_from)) { + return false; + } + if (!FindAndLoadOmapTable(kOmapToDebugStreamName, + session, + &omap_data->omap_to)) { + return false; + } + + // Get the lengths of the address spaces. + if (!GetOriginalImageLength(session, &omap_data->length_original)) + return false; + + return true; +} + +void BuildImageMap(const OmapData& omap_data, ImageMap* image_map) { + assert(image_map != NULL); + + BuildMapping(omap_data, &image_map->mapping); + BuildEndpointIndexMap(image_map); + BuildSubsequentRVAMap(omap_data, &image_map->subsequent_rva_block); +} + +void MapAddressRange(const ImageMap& image_map, + const AddressRange& original_range, + AddressRangeVector* mapped_ranges) { + assert(mapped_ranges != NULL); + + const Mapping& map = image_map.mapping; + + // Handle the trivial case of an empty image_map. This means that there is + // no transformation to be applied, and we can simply return the original + // range. + if (map.empty()) { + mapped_ranges->push_back(original_range); + return; + } + + // If we get a query of length 0 we need to handle it by using a non-zero + // query length. + AddressRange query_range(original_range); + if (query_range.length == 0) + query_range.length = 1; + + // Find the range of intervals that can potentially intersect our query range. + size_t imin = 0; + size_t imax = 0; + { + // The index of the earliest possible range that can affect is us done by + // searching through the secondary indexing structure. + const EndpointIndexMap& eim = image_map.endpoint_index_map; + EndpointIndex q1 = { query_range.rva, 0 }; + EndpointIndexMap::const_iterator it1 = std::lower_bound( + eim.begin(), eim.end(), q1, EndpointIndexLess); + if (it1 == eim.end()) { + imin = map.size(); + } else { + // Backup to find the interval that contains our query point. + if (it1 != eim.begin() && query_range.rva < it1->endpoint) + --it1; + imin = it1->index; + } + + // The first range that can't possibly intersect us is found by searching + // through the image map directly as it is already sorted by interval start + // point. + MappedRange q2 = { query_range.end(), 0 }; + Mapping::const_iterator it2 = std::lower_bound( + map.begin(), map.end(), q2, MappedRangeOriginalLess); + imax = it2 - map.begin(); + } + + // Find all intervals that intersect the query range. + Mapping temp_map; + for (size_t i = imin; i < imax; ++i) { + MappedRange mr = map[i]; + ClipMappedRangeOriginal(query_range, &mr); + if (mr.length + mr.injected > 0) + temp_map.push_back(mr); + } + + // If there are no intersecting ranges then the query range has been removed + // from the image in question. + if (temp_map.empty()) + return; + + // Sort based on transformed addresses. + std::sort(temp_map.begin(), temp_map.end(), MappedRangeMappedLess); + + // Zero-length queries can't actually be merged. We simply output the set of + // unique RVAs that correspond to the query RVA. + if (original_range.length == 0) { + mapped_ranges->push_back(AddressRange(temp_map[0].rva_transformed, 0)); + for (size_t i = 1; i < temp_map.size(); ++i) { + if (temp_map[i].rva_transformed > mapped_ranges->back().rva) + mapped_ranges->push_back(AddressRange(temp_map[i].rva_transformed, 0)); + } + return; + } + + // Merge any ranges that are consecutive in the mapped image. We merge over + // injected content if it makes ranges contiguous, but we ignore any injected + // content at the tail end of a range. This allows us to detect symbols that + // have been lengthened by injecting content in the middle. However, it + // misses the case where content has been injected at the head or the tail. + // The problem is that it doesn't know whether to attribute it to the + // preceding or following symbol. It is up to the author of the transform to + // output explicit OMAP info in these cases to ensure full coverage of the + // transformed address space. + DWORD rva_begin = temp_map[0].rva_transformed; + DWORD rva_cur_content = rva_begin + temp_map[0].length; + DWORD rva_cur_injected = rva_cur_content + temp_map[0].injected; + for (size_t i = 1; i < temp_map.size(); ++i) { + if (rva_cur_injected < temp_map[i].rva_transformed) { + // This marks the end of a continuous range in the image. Output the + // current range and start a new one. + if (rva_begin < rva_cur_content) { + mapped_ranges->push_back( + AddressRange(rva_begin, rva_cur_content - rva_begin)); + } + rva_begin = temp_map[i].rva_transformed; + } + + rva_cur_content = temp_map[i].rva_transformed + temp_map[i].length; + rva_cur_injected = rva_cur_content + temp_map[i].injected; + } + + // Output the range in progress. + if (rva_begin < rva_cur_content) { + mapped_ranges->push_back( + AddressRange(rva_begin, rva_cur_content - rva_begin)); + } + + return; +} + +} // namespace google_breakpad diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/omap.h b/toolkit/crashreporter/google-breakpad/src/common/windows/omap.h new file mode 100644 index 0000000000..bc293afb51 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/omap.h @@ -0,0 +1,72 @@ +// Copyright 2013 Google Inc. All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +// Provides an API for mapping symbols through OMAP information, if a PDB file +// is augmented with it. This allows breakpad to work with addresses in +// transformed images by transforming the symbols themselves, rather than +// transforming addresses prior to querying symbols (the way it is typically +// done by Windows-native tools, including the DIA). + +#ifndef COMMON_WINDOWS_OMAP_H_ +#define COMMON_WINDOWS_OMAP_H_ + +#include "common/windows/omap_internal.h" + +namespace google_breakpad { + +// If the given session contains OMAP data this extracts it, populating +// |omap_data|, and then disabling automatic translation for the session. +// OMAP data is present in the PDB if |omap_data| is not empty. This returns +// true on success, false otherwise. +bool GetOmapDataAndDisableTranslation(IDiaSession* dia_session, + OmapData* omap_data); + +// Given raw OMAP data builds an ImageMap. This can be used to query individual +// image ranges using MapAddressRange. +// |omap_data|| is the OMAP data extracted from the PDB. +// |image_map| will be populated with a description of the image mapping. If +// |omap_data| is empty then this will also be empty. +void BuildImageMap(const OmapData& omap_data, ImageMap* image_map); + +// Given an address range in the original image space determines how exactly it +// has been tranformed. +// |omap_data| is the OMAP data extracted from the PDB, which must not be +// empty. +// |original_range| is the address range in the original image being queried. +// |mapped_ranges| will be populated with a full description of the mapping. +// They may be disjoint in the transformed image so a vector is needed to +// fully represent the mapping. This will be appended to if it is not +// empty. If |omap_data| is empty then |mapped_ranges| will simply be +// populated with a copy of |original_range| (the identity transform). +void MapAddressRange(const ImageMap& image_map, + const AddressRange& original_range, + AddressRangeVector* mapped_ranges); + +} // namespace google_breakpad + +#endif // COMMON_WINDOWS_OMAP_H_ diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/omap_internal.h b/toolkit/crashreporter/google-breakpad/src/common/windows/omap_internal.h new file mode 100644 index 0000000000..2a4713d934 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/omap_internal.h @@ -0,0 +1,140 @@ +// Copyright 2013 Google Inc. All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +// Declares internal implementation details for functionality in omap.h and +// omap.cc. + +#ifndef COMMON_WINDOWS_OMAP_INTERNAL_H_ +#define COMMON_WINDOWS_OMAP_INTERNAL_H_ + +#include <windows.h> +#include <dia2.h> + +#include <map> +#include <vector> + +namespace google_breakpad { + +// The OMAP struct is defined by debughlp.h, which doesn't play nicely with +// imagehlp.h. We simply redefine it. +struct OMAP { + DWORD rva; + DWORD rvaTo; +}; +static_assert(sizeof(OMAP) == 8, "Wrong size for OMAP structure."); +typedef std::vector<OMAP> OmapTable; + +// This contains the OMAP data extracted from an image. +struct OmapData { + // The table of OMAP entries describing the transformation from the + // original image to the transformed image. + OmapTable omap_from; + // The table of OMAP entries describing the transformation from the + // instrumented image to the original image. + OmapTable omap_to; + // The length of the original untransformed image. + DWORD length_original; + + OmapData() : length_original(0) { } +}; + +// This represents a range of addresses in an image. +struct AddressRange { + DWORD rva; + DWORD length; + + AddressRange() : rva(0), length(0) { } + AddressRange(DWORD rva, DWORD length) : rva(rva), length(length) { } + + // Returns the end address of this range. + DWORD end() const { return rva + length; } + + // Addreses only compare as less-than or greater-than if they are not + // overlapping. Otherwise, they compare equal. + int Compare(const AddressRange& rhs) const; + bool operator<(const AddressRange& rhs) const { return Compare(rhs) == -1; } + bool operator>(const AddressRange& rhs) const { return Compare(rhs) == 1; } + + // Equality operators compare exact values. + bool operator==(const AddressRange& rhs) const { + return rva == rhs.rva && length == rhs.length; + } + bool operator!=(const AddressRange& rhs) const { return !((*this) == rhs); } +}; + +typedef std::vector<AddressRange> AddressRangeVector; + +// This represents an address range in an original image, and its corresponding +// range in the transformed image. +struct MappedRange { + // An address in the original image. + DWORD rva_original; + // The corresponding addresses in the transformed image. + DWORD rva_transformed; + // The length of the address range. + DWORD length; + // It is possible for code to be injected into a transformed image, for which + // there is no corresponding code in the original image. If this range of + // transformed image is immediately followed by such injected code we maintain + // a record of its length here. + DWORD injected; + // It is possible for code to be removed from the original image. This happens + // for things like padding between blocks. There is no actual content lost, + // but the spacing between items may be lost. This keeps track of any removed + // content immediately following the |original| range. + DWORD removed; +}; +// A vector of mapped ranges is used as a more useful representation of +// OMAP data. +typedef std::vector<MappedRange> Mapping; + +// Used as a secondary search structure accompanying a Mapping. +struct EndpointIndex { + DWORD endpoint; + size_t index; +}; +typedef std::vector<EndpointIndex> EndpointIndexMap; + +// An ImageMap is vector of mapped ranges, plus a secondary index into it for +// doing interval searches. (An interval tree would also work, but is overkill +// because we don't need insertion and deletion.) +struct ImageMap { + // This is a description of the mapping between original and transformed + // image, sorted by addresses in the original image. + Mapping mapping; + // For all interval endpoints in |mapping| this stores the minimum index of + // an interval in |mapping| that contains the endpoint. Useful for doing + // interval intersection queries. + EndpointIndexMap endpoint_index_map; + + std::map<DWORD, DWORD> subsequent_rva_block; +}; + +} // namespace google_breakpad + +#endif // COMMON_WINDOWS_OMAP_INTERNAL_H_ diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/omap_unittest.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/omap_unittest.cc new file mode 100644 index 0000000000..7fe66bd406 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/omap_unittest.cc @@ -0,0 +1,329 @@ +// Copyright 2013 Google Inc. All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+// * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+// * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+// * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// Unittests for OMAP related functions.
+
+#include "common/windows/omap.h"
+
+#include "breakpad_googletest_includes.h"
+
+namespace google_breakpad {
+
+// Equality operators for ContainerEq. These must be outside of the anonymous
+// namespace in order for them to be found.
+bool operator==(const MappedRange& mr1, const MappedRange& mr2) {
+ return mr1.rva_original == mr2.rva_original &&
+ mr1.rva_transformed == mr2.rva_transformed &&
+ mr1.length == mr2.length &&
+ mr1.injected == mr2.injected &&
+ mr1.removed == mr2.removed;
+}
+bool operator==(const EndpointIndex& ei1, const EndpointIndex& ei2) {
+ return ei1.endpoint == ei2.endpoint && ei1.index == ei2.index;
+}
+
+// Pretty printers for more meaningful error messages. Also need to be outside
+// the anonymous namespace.
+std::ostream& operator<<(std::ostream& os, const MappedRange& mr) {
+ os << "MappedRange(rva_original=" << mr.rva_original
+ << ", rva_transformed=" << mr.rva_transformed
+ << ", length=" << mr.length
+ << ", injected=" << mr.injected
+ << ", removed=" << mr.removed << ")";
+ return os;
+}
+std::ostream& operator<<(std::ostream& os, const EndpointIndex& ei) {
+ os << "EndpointIndex(endpoint=" << ei.endpoint
+ << ", index=" << ei.index << ")";
+ return os;
+}
+std::ostream& operator<<(std::ostream& os, const AddressRange& ar) {
+ os << "AddressRange(rva=" << ar.rva << ", length=" << ar.length << ")";
+ return os;
+}
+
+namespace {
+
+OMAP CreateOmap(DWORD rva, DWORD rvaTo) {
+ OMAP o = { rva, rvaTo };
+ return o;
+}
+
+MappedRange CreateMappedRange(DWORD rva_original,
+ DWORD rva_transformed,
+ DWORD length,
+ DWORD injected,
+ DWORD removed) {
+ MappedRange mr = { rva_original, rva_transformed, length, injected, removed };
+ return mr;
+}
+
+EndpointIndex CreateEndpointIndex(DWORD endpoint, size_t index) {
+ EndpointIndex ei = { endpoint, index };
+ return ei;
+}
+
+// (C is removed)
+// Original : A B C D E F G H
+// Transformed: A B D F E * H1 G1 G2 H2
+// (* is injected, G is copied, H is split)
+// A is implied.
+
+// Layout of the original image.
+const AddressRange B(100, 15);
+const AddressRange C(B.end(), 10);
+const AddressRange D(C.end(), 25);
+const AddressRange E(D.end(), 10);
+const AddressRange F(E.end(), 40);
+const AddressRange G(F.end(), 3);
+const AddressRange H(G.end(), 7);
+
+// Layout of the transformed image.
+const AddressRange Bt(100, 15);
+const AddressRange Dt(Bt.end(), 20); // D is shortened.
+const AddressRange Ft(Dt.end(), F.length);
+const AddressRange Et(Ft.end(), E.length);
+const AddressRange injected(Et.end(), 5);
+const AddressRange H1t(injected.end(), 4); // H is split.
+const AddressRange G1t(H1t.end(), G.length); // G is copied.
+const AddressRange G2t(G1t.end(), G.length); // G is copied.
+const AddressRange H2t(G2t.end(), 3); // H is split.
+
+class BuildImageMapTest : public testing::Test {
+ public:
+ static const DWORD kInvalidAddress = 0xFFFFFFFF;
+
+ void InitOmapData() {
+ omap_data.length_original = H.end();
+
+ // Build the OMAPTO vector (from transformed to original).
+ omap_data.omap_to.push_back(CreateOmap(Bt.rva, B.rva));
+ omap_data.omap_to.push_back(CreateOmap(Dt.rva, D.rva));
+ omap_data.omap_to.push_back(CreateOmap(Ft.rva, F.rva));
+ omap_data.omap_to.push_back(CreateOmap(Et.rva, E.rva));
+ omap_data.omap_to.push_back(CreateOmap(injected.rva, kInvalidAddress));
+ omap_data.omap_to.push_back(CreateOmap(H1t.rva, H.rva));
+ omap_data.omap_to.push_back(CreateOmap(G1t.rva, G.rva));
+ omap_data.omap_to.push_back(CreateOmap(G2t.rva, G.rva));
+ omap_data.omap_to.push_back(CreateOmap(H2t.rva, H.rva + H1t.length));
+ omap_data.omap_to.push_back(CreateOmap(H2t.end(), kInvalidAddress));
+
+ // Build the OMAPFROM vector (from original to transformed).
+ omap_data.omap_from.push_back(CreateOmap(B.rva, Bt.rva));
+ omap_data.omap_from.push_back(CreateOmap(C.rva, kInvalidAddress));
+ omap_data.omap_from.push_back(CreateOmap(D.rva, Dt.rva));
+ omap_data.omap_from.push_back(CreateOmap(E.rva, Et.rva));
+ omap_data.omap_from.push_back(CreateOmap(F.rva, Ft.rva));
+ omap_data.omap_from.push_back(CreateOmap(G.rva, G1t.rva));
+ omap_data.omap_from.push_back(CreateOmap(H.rva, H1t.rva));
+ omap_data.omap_from.push_back(CreateOmap(H.rva + H1t.length, H2t.rva));
+ omap_data.omap_from.push_back(CreateOmap(H.end(), kInvalidAddress));
+ }
+
+ OmapData omap_data;
+};
+
+} // namespace
+
+TEST_F(BuildImageMapTest, EmptyImageMapOnEmptyOmapData) {
+ ASSERT_EQ(0u, omap_data.omap_from.size());
+ ASSERT_EQ(0u, omap_data.omap_to.size());
+ ASSERT_EQ(0u, omap_data.length_original);
+
+ ImageMap image_map;
+ BuildImageMap(omap_data, &image_map);
+ EXPECT_EQ(0u, image_map.mapping.size());
+ EXPECT_EQ(0u, image_map.endpoint_index_map.size());
+}
+
+TEST_F(BuildImageMapTest, ImageMapIsCorrect) {
+ InitOmapData();
+ ASSERT_LE(0u, omap_data.omap_from.size());
+ ASSERT_LE(0u, omap_data.omap_to.size());
+ ASSERT_LE(0u, omap_data.length_original);
+
+ ImageMap image_map;
+ BuildImageMap(omap_data, &image_map);
+ EXPECT_LE(9u, image_map.mapping.size());
+ EXPECT_LE(9u, image_map.endpoint_index_map.size());
+
+ Mapping mapping;
+ mapping.push_back(CreateMappedRange(0, 0, B.rva, 0, 0));
+ // C is removed, and it originally comes immediately after B.
+ mapping.push_back(CreateMappedRange(B.rva, Bt.rva, B.length, 0, C.length));
+ // D is shortened by a length of 5.
+ mapping.push_back(CreateMappedRange(D.rva, Dt.rva, Dt.length, 0, 5));
+ // The injected content comes immediately after E in the transformed image.
+ mapping.push_back(CreateMappedRange(E.rva, Et.rva, E.length, injected.length,
+ 0));
+ mapping.push_back(CreateMappedRange(F.rva, Ft.rva, F.length, 0, 0));
+ // G is copied so creates two entries.
+ mapping.push_back(CreateMappedRange(G.rva, G1t.rva, G.length, 0, 0));
+ mapping.push_back(CreateMappedRange(G.rva, G2t.rva, G.length, 0, 0));
+ // H is split, so create two entries.
+ mapping.push_back(CreateMappedRange(H.rva, H1t.rva, H1t.length, 0, 0));
+ mapping.push_back(CreateMappedRange(H.rva + H1t.length, H2t.rva, H2t.length,
+ 0, 0));
+ EXPECT_THAT(mapping,
+ testing::ContainerEq(image_map.mapping));
+
+ EndpointIndexMap endpoint_index_map;
+ endpoint_index_map.push_back(CreateEndpointIndex(0, 0));
+ endpoint_index_map.push_back(CreateEndpointIndex(B.rva, 1));
+ endpoint_index_map.push_back(CreateEndpointIndex(D.rva, 2));
+ endpoint_index_map.push_back(CreateEndpointIndex(E.rva, 3));
+ endpoint_index_map.push_back(CreateEndpointIndex(F.rva, 4));
+ // G is duplicated so 2 ranges map back to it, hence the skip from 5 to 7.
+ endpoint_index_map.push_back(CreateEndpointIndex(G.rva, 5));
+ // H is split so we expect 2 endpoints to show up attributed to it.
+ endpoint_index_map.push_back(CreateEndpointIndex(H.rva, 7));
+ endpoint_index_map.push_back(CreateEndpointIndex(H.rva + H1t.length, 8));
+ endpoint_index_map.push_back(CreateEndpointIndex(H.end(), 9));
+ EXPECT_THAT(endpoint_index_map,
+ testing::ContainerEq(image_map.endpoint_index_map));
+}
+
+namespace {
+
+class MapAddressRangeTest : public BuildImageMapTest {
+ public:
+ typedef BuildImageMapTest Super;
+ virtual void SetUp() {
+ Super::SetUp();
+ InitOmapData();
+ BuildImageMap(omap_data, &image_map);
+ }
+
+ ImageMap image_map;
+
+ private:
+ using BuildImageMapTest::InitOmapData;
+ using BuildImageMapTest::omap_data;
+};
+
+} // namespace
+
+TEST_F(MapAddressRangeTest, EmptyImageMapReturnsIdentity) {
+ ImageMap im;
+ AddressRangeVector mapped_ranges;
+ AddressRange ar(0, 1024);
+ MapAddressRange(im, ar, &mapped_ranges);
+ EXPECT_EQ(1u, mapped_ranges.size());
+ EXPECT_EQ(ar, mapped_ranges[0]);
+}
+
+TEST_F(MapAddressRangeTest, MapOutOfImage) {
+ AddressRangeVector mapped_ranges;
+ MapAddressRange(image_map, AddressRange(H.end() + 10, 10), &mapped_ranges);
+ EXPECT_EQ(0u, mapped_ranges.size());
+}
+
+TEST_F(MapAddressRangeTest, MapIdentity) {
+ AddressRangeVector mapped_ranges;
+ MapAddressRange(image_map, B, &mapped_ranges);
+ EXPECT_EQ(1u, mapped_ranges.size());
+ EXPECT_THAT(mapped_ranges, testing::ElementsAre(B));
+}
+
+TEST_F(MapAddressRangeTest, MapReorderedContiguous) {
+ AddressRangeVector mapped_ranges;
+
+ AddressRange DEF(D.rva, F.end() - D.rva);
+ MapAddressRange(image_map, DEF, &mapped_ranges);
+ EXPECT_EQ(1u, mapped_ranges.size());
+
+ AddressRange DFEt(Dt.rva, Et.end() - Dt.rva);
+ EXPECT_THAT(mapped_ranges, testing::ElementsAre(DFEt));
+}
+
+TEST_F(MapAddressRangeTest, MapEmptySingle) {
+ AddressRangeVector mapped_ranges;
+ MapAddressRange(image_map, AddressRange(D.rva, 0), &mapped_ranges);
+ EXPECT_EQ(1u, mapped_ranges.size());
+ EXPECT_THAT(mapped_ranges, testing::ElementsAre(AddressRange(Dt.rva, 0)));
+}
+
+TEST_F(MapAddressRangeTest, MapEmptyCopied) {
+ AddressRangeVector mapped_ranges;
+ MapAddressRange(image_map, AddressRange(G.rva, 0), &mapped_ranges);
+ EXPECT_EQ(2u, mapped_ranges.size());
+ EXPECT_THAT(mapped_ranges, testing::ElementsAre(AddressRange(G1t.rva, 0),
+ AddressRange(G2t.rva, 0)));
+}
+
+TEST_F(MapAddressRangeTest, MapCopiedContiguous) {
+ AddressRangeVector mapped_ranges;
+ MapAddressRange(image_map, G, &mapped_ranges);
+ EXPECT_EQ(1u, mapped_ranges.size());
+ EXPECT_THAT(mapped_ranges, testing::ElementsAre(
+ AddressRange(G1t.rva, G2t.end() - G1t.rva)));
+}
+
+TEST_F(MapAddressRangeTest, MapSplitDiscontiguous) {
+ AddressRangeVector mapped_ranges;
+ MapAddressRange(image_map, H, &mapped_ranges);
+ EXPECT_EQ(2u, mapped_ranges.size());
+ EXPECT_THAT(mapped_ranges, testing::ElementsAre(H1t, H2t));
+}
+
+TEST_F(MapAddressRangeTest, MapInjected) {
+ AddressRangeVector mapped_ranges;
+
+ AddressRange EFGH(E.rva, H.end() - E.rva);
+ MapAddressRange(image_map, EFGH, &mapped_ranges);
+ EXPECT_EQ(1u, mapped_ranges.size());
+
+ AddressRange FEHGGHt(Ft.rva, H2t.end() - Ft.rva);
+ EXPECT_THAT(mapped_ranges, testing::ElementsAre(FEHGGHt));
+}
+
+TEST_F(MapAddressRangeTest, MapRemovedEntirely) {
+ AddressRangeVector mapped_ranges;
+ MapAddressRange(image_map, C, &mapped_ranges);
+ EXPECT_EQ(0u, mapped_ranges.size());
+}
+
+TEST_F(MapAddressRangeTest, MapRemovedPartly) {
+ AddressRangeVector mapped_ranges;
+ MapAddressRange(image_map, D, &mapped_ranges);
+ EXPECT_EQ(1u, mapped_ranges.size());
+ EXPECT_THAT(mapped_ranges, testing::ElementsAre(Dt));
+}
+
+TEST_F(MapAddressRangeTest, MapFull) {
+ AddressRangeVector mapped_ranges;
+
+ AddressRange AH(0, H.end());
+ MapAddressRange(image_map, AH, &mapped_ranges);
+ EXPECT_EQ(1u, mapped_ranges.size());
+
+ AddressRange AHt(0, H2t.end());
+ EXPECT_THAT(mapped_ranges, testing::ElementsAre(AHt));
+}
+
+} // namespace google_breakpad
diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/pdb_source_line_writer.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/pdb_source_line_writer.cc new file mode 100644 index 0000000000..4030a2e953 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/pdb_source_line_writer.cc @@ -0,0 +1,1194 @@ +// Copyright (c) 2006, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#include "common/windows/pdb_source_line_writer.h" + +#include <windows.h> +#include <winnt.h> +#include <atlbase.h> +#include <dia2.h> +#include <diacreate.h> +#include <ImageHlp.h> +#include <stdio.h> + +#include <algorithm> +#include <limits> +#include <map> +#include <set> +#include <utility> + +#include "common/windows/dia_util.h" +#include "common/windows/guid_string.h" +#include "common/windows/pe_util.h" +#include "common/windows/string_utils-inl.h" + +// This constant may be missing from DbgHelp.h. See the documentation for +// IDiaSymbol::get_undecoratedNameEx. +#ifndef UNDNAME_NO_ECSU +#define UNDNAME_NO_ECSU 0x8000 // Suppresses enum/class/struct/union. +#endif // UNDNAME_NO_ECSU + +namespace google_breakpad { + +namespace { + +using std::vector; + +// The symbol (among possibly many) selected to represent an rva. +struct SelectedSymbol { + SelectedSymbol(const CComPtr<IDiaSymbol>& symbol, bool is_public) + : symbol(symbol), is_public(is_public), is_multiple(false) {} + + // The symbol to use for an rva. + CComPtr<IDiaSymbol> symbol; + // Whether this is a public or function symbol. + bool is_public; + // Whether the rva has multiple associated symbols. An rva will correspond to + // multiple symbols in the case of linker identical symbol folding. + bool is_multiple; +}; + +// Maps rva to the symbol to use for that address. +typedef std::map<DWORD, SelectedSymbol> SymbolMap; + +// Record this in the map as the selected symbol for the rva if it satisfies the +// necessary conditions. +void MaybeRecordSymbol(DWORD rva, + const CComPtr<IDiaSymbol> symbol, + bool is_public, + SymbolMap* map) { + SymbolMap::iterator loc = map->find(rva); + if (loc == map->end()) { + map->insert(std::make_pair(rva, SelectedSymbol(symbol, is_public))); + return; + } + + // Prefer function symbols to public symbols. + if (is_public && !loc->second.is_public) { + return; + } + + loc->second.is_multiple = true; + + // Take the 'least' symbol by lexicographical order of the decorated name. We + // use the decorated rather than undecorated name because computing the latter + // is expensive. + BSTR current_name, new_name; + loc->second.symbol->get_name(¤t_name); + symbol->get_name(&new_name); + if (wcscmp(new_name, current_name) < 0) { + loc->second.symbol = symbol; + loc->second.is_public = is_public; + } +} + + + +bool SymbolsMatch(IDiaSymbol* a, IDiaSymbol* b) { + DWORD a_section, a_offset, b_section, b_offset; + if (FAILED(a->get_addressSection(&a_section)) || + FAILED(a->get_addressOffset(&a_offset)) || + FAILED(b->get_addressSection(&b_section)) || + FAILED(b->get_addressOffset(&b_offset))) + return false; + return a_section == b_section && a_offset == b_offset; +} + +bool CreateDiaDataSourceInstance(CComPtr<IDiaDataSource> &data_source) { + if (SUCCEEDED(data_source.CoCreateInstance(CLSID_DiaSource))) { + return true; + } + + class DECLSPEC_UUID("B86AE24D-BF2F-4ac9-B5A2-34B14E4CE11D") DiaSource100; + class DECLSPEC_UUID("761D3BCD-1304-41D5-94E8-EAC54E4AC172") DiaSource110; + class DECLSPEC_UUID("3BFCEA48-620F-4B6B-81F7-B9AF75454C7D") DiaSource120; + class DECLSPEC_UUID("E6756135-1E65-4D17-8576-610761398C3C") DiaSource140; + + // If the CoCreateInstance call above failed, msdia*.dll is not registered. + // We can try loading the DLL corresponding to the #included DIA SDK, but + // the DIA headers don't provide a version. Lets try to figure out which DIA + // version we're compiling against by comparing CLSIDs. + const wchar_t *msdia_dll = nullptr; + if (CLSID_DiaSource == _uuidof(DiaSource100)) { + msdia_dll = L"msdia100.dll"; + } else if (CLSID_DiaSource == _uuidof(DiaSource110)) { + msdia_dll = L"msdia110.dll"; + } else if (CLSID_DiaSource == _uuidof(DiaSource120)) { + msdia_dll = L"msdia120.dll"; + } else if (CLSID_DiaSource == _uuidof(DiaSource140)) { + msdia_dll = L"msdia140.dll"; + } + + if (msdia_dll && + SUCCEEDED(NoRegCoCreate(msdia_dll, CLSID_DiaSource, IID_IDiaDataSource, + reinterpret_cast<void **>(&data_source)))) { + return true; + } + + return false; +} + +} // namespace + +PDBSourceLineWriter::PDBSourceLineWriter() : output_(NULL) { +} + +PDBSourceLineWriter::~PDBSourceLineWriter() { + Close(); +} + +bool PDBSourceLineWriter::SetCodeFile(const wstring &exe_file) { + if (code_file_.empty()) { + code_file_ = exe_file; + return true; + } + // Setting a different code file path is an error. It is success only if the + // file paths are the same. + return exe_file == code_file_; +} + +bool PDBSourceLineWriter::Open(const wstring &file, FileFormat format) { + Close(); + code_file_.clear(); + + if (FAILED(CoInitialize(NULL))) { + fprintf(stderr, "CoInitialize failed\n"); + return false; + } + + CComPtr<IDiaDataSource> data_source; + if (!CreateDiaDataSourceInstance(data_source)) { + const int kGuidSize = 64; + wchar_t classid[kGuidSize] = {0}; + StringFromGUID2(CLSID_DiaSource, classid, kGuidSize); + fprintf(stderr, "CoCreateInstance CLSID_DiaSource %S failed " + "(msdia*.dll unregistered?)\n", classid); + return false; + } + + switch (format) { + case PDB_FILE: + if (FAILED(data_source->loadDataFromPdb(file.c_str()))) { + fprintf(stderr, "loadDataFromPdb failed for %ws\n", file.c_str()); + return false; + } + break; + case EXE_FILE: + if (FAILED(data_source->loadDataForExe(file.c_str(), NULL, NULL))) { + fprintf(stderr, "loadDataForExe failed for %ws\n", file.c_str()); + return false; + } + code_file_ = file; + break; + case ANY_FILE: + if (FAILED(data_source->loadDataFromPdb(file.c_str()))) { + if (FAILED(data_source->loadDataForExe(file.c_str(), NULL, NULL))) { + fprintf(stderr, "loadDataForPdb and loadDataFromExe failed for %ws\n", + file.c_str()); + return false; + } + code_file_ = file; + } + break; + default: + fprintf(stderr, "Unknown file format\n"); + return false; + } + + if (FAILED(data_source->openSession(&session_))) { + fprintf(stderr, "openSession failed\n"); + } + + return true; +} + +bool PDBSourceLineWriter::PrintLines(IDiaEnumLineNumbers *lines) { + // The line number format is: + // <rva> <line number> <source file id> + CComPtr<IDiaLineNumber> line; + ULONG count; + + while (SUCCEEDED(lines->Next(1, &line, &count)) && count == 1) { + DWORD rva; + if (FAILED(line->get_relativeVirtualAddress(&rva))) { + fprintf(stderr, "failed to get line rva\n"); + return false; + } + + DWORD length; + if (FAILED(line->get_length(&length))) { + fprintf(stderr, "failed to get line code length\n"); + return false; + } + + DWORD dia_source_id; + if (FAILED(line->get_sourceFileId(&dia_source_id))) { + fprintf(stderr, "failed to get line source file id\n"); + return false; + } + // duplicate file names are coalesced to share one ID + DWORD source_id = GetRealFileID(dia_source_id); + + DWORD line_num; + if (FAILED(line->get_lineNumber(&line_num))) { + fprintf(stderr, "failed to get line number\n"); + return false; + } + + AddressRangeVector ranges; + MapAddressRange(image_map_, AddressRange(rva, length), &ranges); + for (size_t i = 0; i < ranges.size(); ++i) { + fprintf(output_, "%lx %lx %lu %lu\n", ranges[i].rva, ranges[i].length, + line_num, source_id); + } + line.Release(); + } + return true; +} + +bool PDBSourceLineWriter::PrintFunction(IDiaSymbol *function, + IDiaSymbol *block, + bool has_multiple_symbols) { + // The function format is: + // FUNC <address> <length> <param_stack_size> <function> + DWORD rva; + if (FAILED(block->get_relativeVirtualAddress(&rva))) { + fprintf(stderr, "couldn't get rva\n"); + return false; + } + + ULONGLONG length; + if (FAILED(block->get_length(&length))) { + fprintf(stderr, "failed to get function length\n"); + return false; + } + + if (length == 0) { + // Silently ignore zero-length functions, which can infrequently pop up. + return true; + } + + CComBSTR name; + int stack_param_size; + if (!GetSymbolFunctionName(function, &name, &stack_param_size)) { + return false; + } + + // If the decorated name didn't give the parameter size, try to + // calculate it. + if (stack_param_size < 0) { + stack_param_size = GetFunctionStackParamSize(function); + } + + AddressRangeVector ranges; + MapAddressRange(image_map_, AddressRange(rva, static_cast<DWORD>(length)), + &ranges); + for (size_t i = 0; i < ranges.size(); ++i) { + const char* optional_multiple_field = has_multiple_symbols ? "m " : ""; + fprintf(output_, "FUNC %s%lx %lx %x %ws\n", optional_multiple_field, + ranges[i].rva, ranges[i].length, stack_param_size, name.m_str); + } + + CComPtr<IDiaEnumLineNumbers> lines; + if (FAILED(session_->findLinesByRVA(rva, DWORD(length), &lines))) { + return false; + } + + if (!PrintLines(lines)) { + return false; + } + return true; +} + +bool PDBSourceLineWriter::PrintSourceFiles() { + CComPtr<IDiaSymbol> global; + if (FAILED(session_->get_globalScope(&global))) { + fprintf(stderr, "get_globalScope failed\n"); + return false; + } + + CComPtr<IDiaEnumSymbols> compilands; + if (FAILED(global->findChildren(SymTagCompiland, NULL, + nsNone, &compilands))) { + fprintf(stderr, "findChildren failed\n"); + return false; + } + + CComPtr<IDiaSymbol> compiland; + ULONG count; + while (SUCCEEDED(compilands->Next(1, &compiland, &count)) && count == 1) { + CComPtr<IDiaEnumSourceFiles> source_files; + if (FAILED(session_->findFile(compiland, NULL, nsNone, &source_files))) { + return false; + } + CComPtr<IDiaSourceFile> file; + while (SUCCEEDED(source_files->Next(1, &file, &count)) && count == 1) { + DWORD file_id; + if (FAILED(file->get_uniqueId(&file_id))) { + return false; + } + + CComBSTR file_name; + if (FAILED(file->get_fileName(&file_name))) { + return false; + } + + wstring file_name_string(file_name); + if (!FileIDIsCached(file_name_string)) { + // this is a new file name, cache it and output a FILE line. + CacheFileID(file_name_string, file_id); + fwprintf(output_, L"FILE %d %ws\n", file_id, file_name_string.c_str()); + } else { + // this file name has already been seen, just save this + // ID for later lookup. + StoreDuplicateFileID(file_name_string, file_id); + } + file.Release(); + } + compiland.Release(); + } + return true; +} + +bool PDBSourceLineWriter::PrintFunctions() { + ULONG count = 0; + DWORD rva = 0; + CComPtr<IDiaSymbol> global; + HRESULT hr; + + if (FAILED(session_->get_globalScope(&global))) { + fprintf(stderr, "get_globalScope failed\n"); + return false; + } + + CComPtr<IDiaEnumSymbols> symbols = NULL; + + // Find all function symbols first. + SymbolMap rva_symbol; + hr = global->findChildren(SymTagFunction, NULL, nsNone, &symbols); + + if (SUCCEEDED(hr)) { + CComPtr<IDiaSymbol> symbol = NULL; + + while (SUCCEEDED(symbols->Next(1, &symbol, &count)) && count == 1) { + if (SUCCEEDED(symbol->get_relativeVirtualAddress(&rva))) { + // Potentially record this as the canonical symbol for this rva. + MaybeRecordSymbol(rva, symbol, false, &rva_symbol); + } else { + fprintf(stderr, "get_relativeVirtualAddress failed on the symbol\n"); + return false; + } + + symbol.Release(); + } + + symbols.Release(); + } + + // Find all public symbols and record public symbols that are not also private + // symbols. + hr = global->findChildren(SymTagPublicSymbol, NULL, nsNone, &symbols); + + if (SUCCEEDED(hr)) { + CComPtr<IDiaSymbol> symbol = NULL; + + while (SUCCEEDED(symbols->Next(1, &symbol, &count)) && count == 1) { + if (SUCCEEDED(symbol->get_relativeVirtualAddress(&rva))) { + // Potentially record this as the canonical symbol for this rva. + MaybeRecordSymbol(rva, symbol, true, &rva_symbol); + } else { + fprintf(stderr, "get_relativeVirtualAddress failed on the symbol\n"); + return false; + } + + symbol.Release(); + } + + symbols.Release(); + } + + // For each rva, dump the selected symbol at the address. + SymbolMap::iterator it; + for (it = rva_symbol.begin(); it != rva_symbol.end(); ++it) { + CComPtr<IDiaSymbol> symbol = it->second.symbol; + // Only print public symbols if there is no function symbol for the address. + if (!it->second.is_public) { + if (!PrintFunction(symbol, symbol, it->second.is_multiple)) + return false; + } else { + if (!PrintCodePublicSymbol(symbol, it->second.is_multiple)) + return false; + } + } + + // When building with PGO, the compiler can split functions into + // "hot" and "cold" blocks, and move the "cold" blocks out to separate + // pages, so the function can be noncontiguous. To find these blocks, + // we have to iterate over all the compilands, and then find blocks + // that are children of them. We can then find the lexical parents + // of those blocks and print out an extra FUNC line for blocks + // that are not contained in their parent functions. + CComPtr<IDiaEnumSymbols> compilands; + if (FAILED(global->findChildren(SymTagCompiland, NULL, + nsNone, &compilands))) { + fprintf(stderr, "findChildren failed on the global\n"); + return false; + } + + CComPtr<IDiaSymbol> compiland; + while (SUCCEEDED(compilands->Next(1, &compiland, &count)) && count == 1) { + CComPtr<IDiaEnumSymbols> blocks; + if (FAILED(compiland->findChildren(SymTagBlock, NULL, + nsNone, &blocks))) { + fprintf(stderr, "findChildren failed on a compiland\n"); + return false; + } + + CComPtr<IDiaSymbol> block; + while (SUCCEEDED(blocks->Next(1, &block, &count)) && count == 1) { + // find this block's lexical parent function + CComPtr<IDiaSymbol> parent; + DWORD tag; + if (SUCCEEDED(block->get_lexicalParent(&parent)) && + SUCCEEDED(parent->get_symTag(&tag)) && + tag == SymTagFunction) { + // now get the block's offset and the function's offset and size, + // and determine if the block is outside of the function + DWORD func_rva, block_rva; + ULONGLONG func_length; + if (SUCCEEDED(block->get_relativeVirtualAddress(&block_rva)) && + SUCCEEDED(parent->get_relativeVirtualAddress(&func_rva)) && + SUCCEEDED(parent->get_length(&func_length))) { + if (block_rva < func_rva || block_rva > (func_rva + func_length)) { + if (!PrintFunction(parent, block, false)) { + return false; + } + } + } + } + parent.Release(); + block.Release(); + } + blocks.Release(); + compiland.Release(); + } + + global.Release(); + return true; +} + +#undef max + +bool PDBSourceLineWriter::PrintFrameDataUsingPDB() { + // It would be nice if it were possible to output frame data alongside the + // associated function, as is done with line numbers, but the DIA API + // doesn't make it possible to get the frame data in that way. + + CComPtr<IDiaEnumFrameData> frame_data_enum; + if (!FindTable(session_, &frame_data_enum)) + return false; + + DWORD last_type = std::numeric_limits<DWORD>::max(); + DWORD last_rva = std::numeric_limits<DWORD>::max(); + DWORD last_code_size = 0; + DWORD last_prolog_size = std::numeric_limits<DWORD>::max(); + + CComPtr<IDiaFrameData> frame_data; + ULONG count = 0; + while (SUCCEEDED(frame_data_enum->Next(1, &frame_data, &count)) && + count == 1) { + DWORD type; + if (FAILED(frame_data->get_type(&type))) + return false; + + DWORD rva; + if (FAILED(frame_data->get_relativeVirtualAddress(&rva))) + return false; + + DWORD code_size; + if (FAILED(frame_data->get_lengthBlock(&code_size))) + return false; + + DWORD prolog_size; + if (FAILED(frame_data->get_lengthProlog(&prolog_size))) + return false; + + // parameter_size is the size of parameters passed on the stack. If any + // parameters are not passed on the stack (such as in registers), their + // sizes will not be included in parameter_size. + DWORD parameter_size; + if (FAILED(frame_data->get_lengthParams(¶meter_size))) + return false; + + DWORD saved_register_size; + if (FAILED(frame_data->get_lengthSavedRegisters(&saved_register_size))) + return false; + + DWORD local_size; + if (FAILED(frame_data->get_lengthLocals(&local_size))) + return false; + + // get_maxStack can return S_FALSE, just use 0 in that case. + DWORD max_stack_size = 0; + if (FAILED(frame_data->get_maxStack(&max_stack_size))) + return false; + + // get_programString can return S_FALSE, indicating that there is no + // program string. In that case, check whether %ebp is used. + HRESULT program_string_result; + CComBSTR program_string; + if (FAILED(program_string_result = frame_data->get_program( + &program_string))) { + return false; + } + + // get_allocatesBasePointer can return S_FALSE, treat that as though + // %ebp is not used. + BOOL allocates_base_pointer = FALSE; + if (program_string_result != S_OK) { + if (FAILED(frame_data->get_allocatesBasePointer( + &allocates_base_pointer))) { + return false; + } + } + + // Only print out a line if type, rva, code_size, or prolog_size have + // changed from the last line. It is surprisingly common (especially in + // system library PDBs) for DIA to return a series of identical + // IDiaFrameData objects. For kernel32.pdb from Windows XP SP2 on x86, + // this check reduces the size of the dumped symbol file by a third. + if (type != last_type || rva != last_rva || code_size != last_code_size || + prolog_size != last_prolog_size) { + // The prolog and the code portions of the frame have to be treated + // independently as they may have independently changed in size, or may + // even have been split. + // NOTE: If epilog size is ever non-zero, we have to do something + // similar with it. + + // Figure out where the prolog bytes have landed. + AddressRangeVector prolog_ranges; + if (prolog_size > 0) { + MapAddressRange(image_map_, AddressRange(rva, prolog_size), + &prolog_ranges); + } + + // And figure out where the code bytes have landed. + AddressRangeVector code_ranges; + MapAddressRange(image_map_, + AddressRange(rva + prolog_size, + code_size - prolog_size), + &code_ranges); + + struct FrameInfo { + DWORD rva; + DWORD code_size; + DWORD prolog_size; + }; + std::vector<FrameInfo> frame_infos; + + // Special case: The prolog and the code bytes remain contiguous. This is + // only done for compactness of the symbol file, and we could actually + // be outputting independent frame info for the prolog and code portions. + if (prolog_ranges.size() == 1 && code_ranges.size() == 1 && + prolog_ranges[0].end() == code_ranges[0].rva) { + FrameInfo fi = { prolog_ranges[0].rva, + prolog_ranges[0].length + code_ranges[0].length, + prolog_ranges[0].length }; + frame_infos.push_back(fi); + } else { + // Otherwise we output the prolog and code frame info independently. + for (size_t i = 0; i < prolog_ranges.size(); ++i) { + FrameInfo fi = { prolog_ranges[i].rva, + prolog_ranges[i].length, + prolog_ranges[i].length }; + frame_infos.push_back(fi); + } + for (size_t i = 0; i < code_ranges.size(); ++i) { + FrameInfo fi = { code_ranges[i].rva, code_ranges[i].length, 0 }; + frame_infos.push_back(fi); + } + } + + for (size_t i = 0; i < frame_infos.size(); ++i) { + const FrameInfo& fi(frame_infos[i]); + fprintf(output_, "STACK WIN %lx %lx %lx %lx %x %lx %lx %lx %lx %d ", + type, fi.rva, fi.code_size, fi.prolog_size, + 0 /* epilog_size */, parameter_size, saved_register_size, + local_size, max_stack_size, program_string_result == S_OK); + if (program_string_result == S_OK) { + fprintf(output_, "%ws\n", program_string.m_str); + } else { + fprintf(output_, "%d\n", allocates_base_pointer); + } + } + + last_type = type; + last_rva = rva; + last_code_size = code_size; + last_prolog_size = prolog_size; + } + + frame_data.Release(); + } + + return true; +} + +bool PDBSourceLineWriter::PrintFrameDataUsingEXE() { + if (code_file_.empty() && !FindPEFile()) { + fprintf(stderr, "Couldn't locate EXE or DLL file.\n"); + return false; + } + + return PrintPEFrameData(code_file_, output_); +} + +bool PDBSourceLineWriter::PrintFrameData() { + PDBModuleInfo info; + if (GetModuleInfo(&info) && info.cpu == L"x86_64") { + return PrintFrameDataUsingEXE(); + } else { + return PrintFrameDataUsingPDB(); + } + return false; +} + +bool PDBSourceLineWriter::PrintCodePublicSymbol(IDiaSymbol *symbol, + bool has_multiple_symbols) { + BOOL is_code; + if (FAILED(symbol->get_code(&is_code))) { + return false; + } + if (!is_code) { + return true; + } + + DWORD rva; + if (FAILED(symbol->get_relativeVirtualAddress(&rva))) { + return false; + } + + CComBSTR name; + int stack_param_size; + if (!GetSymbolFunctionName(symbol, &name, &stack_param_size)) { + return false; + } + + AddressRangeVector ranges; + MapAddressRange(image_map_, AddressRange(rva, 1), &ranges); + for (size_t i = 0; i < ranges.size(); ++i) { + const char* optional_multiple_field = has_multiple_symbols ? "m " : ""; + fprintf(output_, "PUBLIC %s%lx %x %ws\n", optional_multiple_field, + ranges[i].rva, stack_param_size > 0 ? stack_param_size : 0, + name.m_str); + } + + // Now walk the function in the original untranslated space, asking DIA + // what function is at that location, stepping through OMAP blocks. If + // we're still in the same function, emit another entry, because the + // symbol could have been split into multiple pieces. If we've gotten to + // another symbol in the original address space, then we're done for + // this symbol. See https://crbug.com/678874. + for (;;) { + // This steps to the next block in the original image. Simply doing + // rva++ would also be correct, but would emit tons of unnecessary + // entries. + rva = image_map_.subsequent_rva_block[rva]; + if (rva == 0) + break; + + CComPtr<IDiaSymbol> next_sym = NULL; + LONG displacement; + if (FAILED(session_->findSymbolByRVAEx(rva, SymTagPublicSymbol, &next_sym, + &displacement))) { + break; + } + + if (!SymbolsMatch(symbol, next_sym)) + break; + + AddressRangeVector next_ranges; + MapAddressRange(image_map_, AddressRange(rva, 1), &next_ranges); + for (size_t i = 0; i < next_ranges.size(); ++i) { + fprintf(output_, "PUBLIC %lx %x %ws\n", next_ranges[i].rva, + stack_param_size > 0 ? stack_param_size : 0, name.m_str); + } + } + + return true; +} + +bool PDBSourceLineWriter::PrintPDBInfo() { + PDBModuleInfo info; + if (!GetModuleInfo(&info)) { + return false; + } + + // Hard-code "windows" for the OS because that's the only thing that makes + // sense for PDB files. (This might not be strictly correct for Windows CE + // support, but we don't care about that at the moment.) + fprintf(output_, "MODULE windows %ws %ws %ws\n", + info.cpu.c_str(), info.debug_identifier.c_str(), + info.debug_file.c_str()); + + return true; +} + +bool PDBSourceLineWriter::PrintPEInfo() { + PEModuleInfo info; + if (!GetPEInfo(&info)) { + return false; + } + + fprintf(output_, "INFO CODE_ID %ws %ws\n", + info.code_identifier.c_str(), + info.code_file.c_str()); + return true; +} + +// wcstol_positive_strict is sort of like wcstol, but much stricter. string +// should be a buffer pointing to a null-terminated string containing only +// decimal digits. If the entire string can be converted to an integer +// without overflowing, and there are no non-digit characters before the +// result is set to the value and this function returns true. Otherwise, +// this function returns false. This is an alternative to the strtol, atoi, +// and scanf families, which are not as strict about input and in some cases +// don't provide a good way for the caller to determine if a conversion was +// successful. +static bool wcstol_positive_strict(wchar_t *string, int *result) { + int value = 0; + for (wchar_t *c = string; *c != '\0'; ++c) { + int last_value = value; + value *= 10; + // Detect overflow. + if (value / 10 != last_value || value < 0) { + return false; + } + if (*c < '0' || *c > '9') { + return false; + } + unsigned int c_value = *c - '0'; + last_value = value; + value += c_value; + // Detect overflow. + if (value < last_value) { + return false; + } + // Forbid leading zeroes unless the string is just "0". + if (value == 0 && *(c+1) != '\0') { + return false; + } + } + *result = value; + return true; +} + +bool PDBSourceLineWriter::FindPEFile() { + CComPtr<IDiaSymbol> global; + if (FAILED(session_->get_globalScope(&global))) { + fprintf(stderr, "get_globalScope failed\n"); + return false; + } + + CComBSTR symbols_file; + if (SUCCEEDED(global->get_symbolsFileName(&symbols_file))) { + wstring file(symbols_file); + + // Look for an EXE or DLL file. + const wchar_t *extensions[] = { L"exe", L"dll" }; + for (size_t i = 0; i < sizeof(extensions) / sizeof(extensions[0]); i++) { + size_t dot_pos = file.find_last_of(L"."); + if (dot_pos != wstring::npos) { + file.replace(dot_pos + 1, wstring::npos, extensions[i]); + // Check if this file exists. + if (GetFileAttributesW(file.c_str()) != INVALID_FILE_ATTRIBUTES) { + code_file_ = file; + return true; + } + } + } + } + + return false; +} + +// static +bool PDBSourceLineWriter::GetSymbolFunctionName(IDiaSymbol *function, + BSTR *name, + int *stack_param_size) { + *stack_param_size = -1; + const DWORD undecorate_options = UNDNAME_NO_MS_KEYWORDS | + UNDNAME_NO_FUNCTION_RETURNS | + UNDNAME_NO_ALLOCATION_MODEL | + UNDNAME_NO_ALLOCATION_LANGUAGE | + UNDNAME_NO_THISTYPE | + UNDNAME_NO_ACCESS_SPECIFIERS | + UNDNAME_NO_THROW_SIGNATURES | + UNDNAME_NO_MEMBER_TYPE | + UNDNAME_NO_RETURN_UDT_MODEL | + UNDNAME_NO_ECSU; + + // Use get_undecoratedNameEx to get readable C++ names with arguments. + if (function->get_undecoratedNameEx(undecorate_options, name) != S_OK) { + if (function->get_name(name) != S_OK) { + fprintf(stderr, "failed to get function name\n"); + return false; + } + + // It's possible for get_name to return an empty string, so + // special-case that. + if (wcscmp(*name, L"") == 0) { + SysFreeString(*name); + // dwarf_cu_to_module.cc uses "<name omitted>", so match that. + *name = SysAllocString(L"<name omitted>"); + return true; + } + + // If a name comes from get_name because no undecorated form existed, + // it's already formatted properly to be used as output. Don't do any + // additional processing. + // + // MSVC7's DIA seems to not undecorate names in as many cases as MSVC8's. + // This will result in calling get_name for some C++ symbols, so + // all of the parameter and return type information may not be included in + // the name string. + } else { + // C++ uses a bogus "void" argument for functions and methods that don't + // take any parameters. Take it out of the undecorated name because it's + // ugly and unnecessary. + const wchar_t *replace_string = L"(void)"; + const size_t replace_length = wcslen(replace_string); + const wchar_t *replacement_string = L"()"; + size_t length = wcslen(*name); + if (length >= replace_length) { + wchar_t *name_end = *name + length - replace_length; + if (wcscmp(name_end, replace_string) == 0) { + WindowsStringUtils::safe_wcscpy(name_end, replace_length, + replacement_string); + length = wcslen(*name); + } + } + + // Undecorate names used for stdcall and fastcall. These names prefix + // the identifier with '_' (stdcall) or '@' (fastcall) and suffix it + // with '@' followed by the number of bytes of parameters, in decimal. + // If such a name is found, take note of the size and undecorate it. + // Only do this for names that aren't C++, which is determined based on + // whether the undecorated name contains any ':' or '(' characters. + if (!wcschr(*name, ':') && !wcschr(*name, '(') && + (*name[0] == '_' || *name[0] == '@')) { + wchar_t *last_at = wcsrchr(*name + 1, '@'); + if (last_at && wcstol_positive_strict(last_at + 1, stack_param_size)) { + // If this function adheres to the fastcall convention, it accepts up + // to the first 8 bytes of parameters in registers (%ecx and %edx). + // We're only interested in the stack space used for parameters, so + // so subtract 8 and don't let the size go below 0. + if (*name[0] == '@') { + if (*stack_param_size > 8) { + *stack_param_size -= 8; + } else { + *stack_param_size = 0; + } + } + + // Undecorate the name by moving it one character to the left in its + // buffer, and terminating it where the last '@' had been. + WindowsStringUtils::safe_wcsncpy(*name, length, + *name + 1, last_at - *name - 1); + } else if (*name[0] == '_') { + // This symbol's name is encoded according to the cdecl rules. The + // name doesn't end in a '@' character followed by a decimal positive + // integer, so it's not a stdcall name. Strip off the leading + // underscore. + WindowsStringUtils::safe_wcsncpy(*name, length, *name + 1, length); + } + } + } + + return true; +} + +// static +int PDBSourceLineWriter::GetFunctionStackParamSize(IDiaSymbol *function) { + // This implementation is highly x86-specific. + + // Gather the symbols corresponding to data. + CComPtr<IDiaEnumSymbols> data_children; + if (FAILED(function->findChildren(SymTagData, NULL, nsNone, + &data_children))) { + return 0; + } + + // lowest_base is the lowest %ebp-relative byte offset used for a parameter. + // highest_end is one greater than the highest offset (i.e. base + length). + // Stack parameters are assumed to be contiguous, because in reality, they + // are. + int lowest_base = INT_MAX; + int highest_end = INT_MIN; + + CComPtr<IDiaSymbol> child; + DWORD count; + while (SUCCEEDED(data_children->Next(1, &child, &count)) && count == 1) { + // If any operation fails at this point, just proceed to the next child. + // Use the next_child label instead of continue because child needs to + // be released before it's reused. Declare constructable/destructable + // types early to avoid gotos that cross initializations. + CComPtr<IDiaSymbol> child_type; + + // DataIsObjectPtr is only used for |this|. Because |this| can be passed + // as a stack parameter, look for it in addition to traditional + // parameters. + DWORD child_kind; + if (FAILED(child->get_dataKind(&child_kind)) || + (child_kind != DataIsParam && child_kind != DataIsObjectPtr)) { + goto next_child; + } + + // Only concentrate on register-relative parameters. Parameters may also + // be enregistered (passed directly in a register), but those don't + // consume any stack space, so they're not of interest. + DWORD child_location_type; + if (FAILED(child->get_locationType(&child_location_type)) || + child_location_type != LocIsRegRel) { + goto next_child; + } + + // Of register-relative parameters, the only ones that make any sense are + // %ebp- or %esp-relative. Note that MSVC's debugging information always + // gives parameters as %ebp-relative even when a function doesn't use a + // traditional frame pointer and stack parameters are accessed relative to + // %esp, so just look for %ebp-relative parameters. If you wanted to + // access parameters, you'd probably want to treat these %ebp-relative + // offsets as if they were relative to %esp before a function's prolog + // executed. + DWORD child_register; + if (FAILED(child->get_registerId(&child_register)) || + child_register != CV_REG_EBP) { + goto next_child; + } + + LONG child_register_offset; + if (FAILED(child->get_offset(&child_register_offset))) { + goto next_child; + } + + // IDiaSymbol::get_type can succeed but still pass back a NULL value. + if (FAILED(child->get_type(&child_type)) || !child_type) { + goto next_child; + } + + ULONGLONG child_length; + if (FAILED(child_type->get_length(&child_length))) { + goto next_child; + } + + // Extra scope to avoid goto jumping over variable initialization + { + int child_end = child_register_offset + static_cast<ULONG>(child_length); + if (child_register_offset < lowest_base) { + lowest_base = child_register_offset; + } + if (child_end > highest_end) { + highest_end = child_end; + } + } + +next_child: + child.Release(); + } + + int param_size = 0; + // Make sure lowest_base isn't less than 4, because [%esp+4] is the lowest + // possible address to find a stack parameter before executing a function's + // prolog (see above). Some optimizations cause parameter offsets to be + // lower than 4, but we're not concerned with those because we're only + // looking for parameters contained in addresses higher than where the + // return address is stored. + if (lowest_base < 4) { + lowest_base = 4; + } + if (highest_end > lowest_base) { + // All stack parameters are pushed as at least 4-byte quantities. If the + // last type was narrower than 4 bytes, promote it. This assumes that all + // parameters' offsets are 4-byte-aligned, which is always the case. Only + // worry about the last type, because we're not summing the type sizes, + // just looking at the lowest and highest offsets. + int remainder = highest_end % 4; + if (remainder) { + highest_end += 4 - remainder; + } + + param_size = highest_end - lowest_base; + } + + return param_size; +} + +bool PDBSourceLineWriter::WriteSymbols(FILE *symbol_file) { + output_ = symbol_file; + + // Load the OMAP information, and disable auto-translation of addresses in + // preference of doing it ourselves. + OmapData omap_data; + if (!GetOmapDataAndDisableTranslation(session_, &omap_data)) + return false; + BuildImageMap(omap_data, &image_map_); + + bool ret = PrintPDBInfo(); + // This is not a critical piece of the symbol file. + PrintPEInfo(); + ret = ret && + PrintSourceFiles() && + PrintFunctions() && + PrintFrameData(); + + output_ = NULL; + return ret; +} + +void PDBSourceLineWriter::Close() { + if (session_ != nullptr) { + session_.Release(); + } +} + +bool PDBSourceLineWriter::GetModuleInfo(PDBModuleInfo *info) { + if (!info) { + return false; + } + + info->debug_file.clear(); + info->debug_identifier.clear(); + info->cpu.clear(); + + CComPtr<IDiaSymbol> global; + if (FAILED(session_->get_globalScope(&global))) { + return false; + } + + DWORD machine_type; + // get_machineType can return S_FALSE. + if (global->get_machineType(&machine_type) == S_OK) { + // The documentation claims that get_machineType returns a value from + // the CV_CPU_TYPE_e enumeration, but that's not the case. + // Instead, it returns one of the IMAGE_FILE_MACHINE values as + // defined here: + // http://msdn.microsoft.com/en-us/library/ms680313%28VS.85%29.aspx + info->cpu = FileHeaderMachineToCpuString(static_cast<WORD>(machine_type)); + } else { + // Unexpected, but handle gracefully. + info->cpu = L"unknown"; + } + + // DWORD* and int* are not compatible. This is clean and avoids a cast. + DWORD age; + if (FAILED(global->get_age(&age))) { + return false; + } + + bool uses_guid; + if (!UsesGUID(&uses_guid)) { + return false; + } + + if (uses_guid) { + GUID guid; + if (FAILED(global->get_guid(&guid))) { + return false; + } + + info->debug_identifier = GenerateDebugIdentifier(age, guid); + } else { + DWORD signature; + if (FAILED(global->get_signature(&signature))) { + return false; + } + + info->debug_identifier = GenerateDebugIdentifier(age, signature); + } + + CComBSTR debug_file_string; + if (FAILED(global->get_symbolsFileName(&debug_file_string))) { + return false; + } + info->debug_file = + WindowsStringUtils::GetBaseName(wstring(debug_file_string)); + + return true; +} + +bool PDBSourceLineWriter::GetPEInfo(PEModuleInfo *info) { + if (!info) { + return false; + } + + if (code_file_.empty() && !FindPEFile()) { + fprintf(stderr, "Couldn't locate EXE or DLL file.\n"); + return false; + } + + return ReadPEInfo(code_file_, info); +} + +bool PDBSourceLineWriter::UsesGUID(bool *uses_guid) { + if (!uses_guid) + return false; + + CComPtr<IDiaSymbol> global; + if (FAILED(session_->get_globalScope(&global))) + return false; + + GUID guid; + if (FAILED(global->get_guid(&guid))) + return false; + + DWORD signature; + if (FAILED(global->get_signature(&signature))) + return false; + + // There are two possibilities for guid: either it's a real 128-bit GUID + // as identified in a code module by a new-style CodeView record, or it's + // a 32-bit signature (timestamp) as identified by an old-style record. + // See MDCVInfoPDB70 and MDCVInfoPDB20 in minidump_format.h. + // + // Because DIA doesn't provide a way to directly determine whether a module + // uses a GUID or a 32-bit signature, this code checks whether the first 32 + // bits of guid are the same as the signature, and if the rest of guid is + // zero. If so, then with a pretty high degree of certainty, there's an + // old-style CodeView record in use. This method will only falsely find an + // an old-style CodeView record if a real 128-bit GUID has its first 32 + // bits set the same as the module's signature (timestamp) and the rest of + // the GUID is set to 0. This is highly unlikely. + + GUID signature_guid = {signature}; // 0-initializes other members + *uses_guid = !IsEqualGUID(guid, signature_guid); + return true; +} + +} // namespace google_breakpad diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/pdb_source_line_writer.h b/toolkit/crashreporter/google-breakpad/src/common/windows/pdb_source_line_writer.h new file mode 100644 index 0000000000..c0adf29f24 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/pdb_source_line_writer.h @@ -0,0 +1,226 @@ +// Copyright (c) 2006, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +// PDBSourceLineWriter uses a pdb file produced by Visual C++ to output +// a line/address map for use with BasicSourceLineResolver. + +#ifndef COMMON_WINDOWS_PDB_SOURCE_LINE_WRITER_H_ +#define COMMON_WINDOWS_PDB_SOURCE_LINE_WRITER_H_ + +#include <atlcomcli.h> + +#include <string> +#include <unordered_map> + +#include "common/windows/module_info.h" +#include "common/windows/omap.h" + +struct IDiaEnumLineNumbers; +struct IDiaSession; +struct IDiaSymbol; + +namespace google_breakpad { + +using std::wstring; +using std::unordered_map; + +class PDBSourceLineWriter { + public: + enum FileFormat { + PDB_FILE, // a .pdb file containing debug symbols + EXE_FILE, // a .exe or .dll file + ANY_FILE // try PDB_FILE and then EXE_FILE + }; + + explicit PDBSourceLineWriter(); + ~PDBSourceLineWriter(); + + // Opens the given file. For executable files, the corresponding pdb + // file must be available; Open will be if it is not. + // If there is already a pdb file open, it is automatically closed. + // Returns true on success. + bool Open(const wstring &file, FileFormat format); + + // Closes the current pdb file and its associated resources. + void Close(); + + // Sets the code file full path. This is optional for 32-bit modules. It is + // also optional for 64-bit modules when there is an executable file stored + // in the same directory as the PDB file. It is only required for 64-bit + // modules when the executable file is not in the same location as the PDB + // file and it must be called after Open() and before WriteMap(). + // If Open() was called for an executable file, then it is an error to call + // SetCodeFile() with a different file path and it will return false. + bool SetCodeFile(const wstring &exe_file); + + // Writes a Breakpad symbol file from the current pdb file to |symbol_file|. + // Returns true on success. + bool WriteSymbols(FILE *symbol_file); + + // Retrieves information about the module's debugging file. Returns + // true on success and false on failure. + bool GetModuleInfo(PDBModuleInfo *info); + + // Retrieves information about the module's PE file. Returns + // true on success and false on failure. + bool GetPEInfo(PEModuleInfo *info); + + // Sets uses_guid to true if the opened file uses a new-style CodeView + // record with a 128-bit GUID, or false if the opened file uses an old-style + // CodeView record. When no GUID is available, a 32-bit signature should be + // used to identify the module instead. If the information cannot be + // determined, this method returns false. + bool UsesGUID(bool *uses_guid); + + private: + // Outputs the line/address pairs for each line in the enumerator. + // Returns true on success. + bool PrintLines(IDiaEnumLineNumbers *lines); + + // Outputs a function address and name, followed by its source line list. + // block can be the same object as function, or it can be a reference to a + // code block that is lexically part of this function, but resides at a + // separate address. If has_multiple_symbols is true, this function's + // instructions correspond to multiple symbols. Returns true on success. + bool PrintFunction(IDiaSymbol *function, IDiaSymbol *block, + bool has_multiple_symbols); + + // Outputs all functions as described above. Returns true on success. + bool PrintFunctions(); + + // Outputs all of the source files in the session's pdb file. + // Returns true on success. + bool PrintSourceFiles(); + + // Outputs all of the frame information necessary to construct stack + // backtraces in the absence of frame pointers. For x86 data stored in + // .pdb files. Returns true on success. + bool PrintFrameDataUsingPDB(); + + // Outputs all of the frame information necessary to construct stack + // backtraces in the absence of frame pointers. For x64 data stored in + // .exe, .dll files. Returns true on success. + bool PrintFrameDataUsingEXE(); + + // Outputs all of the frame information necessary to construct stack + // backtraces in the absence of frame pointers. Returns true on success. + bool PrintFrameData(); + + // Outputs a single public symbol address and name, if the symbol corresponds + // to a code address. Returns true on success. If symbol is does not + // correspond to code, returns true without outputting anything. If + // has_multiple_symbols is true, the symbol corresponds to a code address and + // the instructions correspond to multiple symbols. + bool PrintCodePublicSymbol(IDiaSymbol *symbol, bool has_multiple_symbols); + + // Outputs a line identifying the PDB file that is being dumped, along with + // its uuid and age. + bool PrintPDBInfo(); + + // Outputs a line identifying the PE file corresponding to the PDB + // file that is being dumped, along with its code identifier, + // which consists of its timestamp and file size. + bool PrintPEInfo(); + + // Returns true if this filename has already been seen, + // and an ID is stored for it, or false if it has not. + bool FileIDIsCached(const wstring &file) { + return unique_files_.find(file) != unique_files_.end(); + } + + // Cache this filename and ID for later reuse. + void CacheFileID(const wstring &file, DWORD id) { + unique_files_[file] = id; + } + + // Store this ID in the cache as a duplicate for this filename. + void StoreDuplicateFileID(const wstring &file, DWORD id) { + unordered_map<wstring, DWORD>::iterator iter = unique_files_.find(file); + if (iter != unique_files_.end()) { + // map this id to the previously seen one + file_ids_[id] = iter->second; + } + } + + // Given a file's unique ID, return the ID that should be used to + // reference it. There may be multiple files with identical filenames + // but different unique IDs. The cache attempts to coalesce these into + // one ID per unique filename. + DWORD GetRealFileID(DWORD id) { + unordered_map<DWORD, DWORD>::iterator iter = file_ids_.find(id); + if (iter == file_ids_.end()) + return id; + return iter->second; + } + + // Find the PE file corresponding to the loaded PDB file, and + // set the code_file_ member. Returns false on failure. + bool FindPEFile(); + + // Returns the function name for a symbol. If possible, the name is + // undecorated. If the symbol's decorated form indicates the size of + // parameters on the stack, this information is returned in stack_param_size. + // Returns true on success. If the symbol doesn't encode parameter size + // information, stack_param_size is set to -1. + static bool GetSymbolFunctionName(IDiaSymbol *function, BSTR *name, + int *stack_param_size); + + // Returns the number of bytes of stack space used for a function's + // parameters. function must have the tag SymTagFunction. In the event of + // a failure, returns 0, which is also a valid number of bytes. + static int GetFunctionStackParamSize(IDiaSymbol *function); + + // The filename of the PE file corresponding to the currently-open + // pdb file. + wstring code_file_; + + // The session for the currently-open pdb file. + CComPtr<IDiaSession> session_; + + // The current output file for this WriteMap invocation. + FILE *output_; + + // There may be many duplicate filenames with different IDs. + // This maps from the DIA "unique ID" to a single ID per unique + // filename. + unordered_map<DWORD, DWORD> file_ids_; + // This maps unique filenames to file IDs. + unordered_map<wstring, DWORD> unique_files_; + + // This is used for calculating post-transform symbol addresses and lengths. + ImageMap image_map_; + + // Disallow copy ctor and operator= + PDBSourceLineWriter(const PDBSourceLineWriter&); + void operator=(const PDBSourceLineWriter&); +}; + +} // namespace google_breakpad + +#endif // COMMON_WINDOWS_PDB_SOURCE_LINE_WRITER_H_ diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/pe_source_line_writer.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/pe_source_line_writer.cc new file mode 100644 index 0000000000..cb6cc71396 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/pe_source_line_writer.cc @@ -0,0 +1,77 @@ +// Copyright (c) 2019, Google Inc.
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+// * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+// * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+// * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "common/windows/pe_source_line_writer.h"
+
+#include "common/windows/pe_util.h"
+
+namespace google_breakpad {
+PESourceLineWriter::PESourceLineWriter(const wstring& pe_file) :
+ pe_file_(pe_file) {
+}
+
+PESourceLineWriter::~PESourceLineWriter() {
+}
+
+bool PESourceLineWriter::WriteSymbols(FILE* symbol_file) {
+ PDBModuleInfo module_info;
+ if (!GetModuleInfo(&module_info)) {
+ return false;
+ }
+ // Hard-code "windows" for the OS because that's the only thing that makes
+ // sense for PDB files. (This might not be strictly correct for Windows CE
+ // support, but we don't care about that at the moment.)
+ fprintf(symbol_file, "MODULE windows %ws %ws %ws\n",
+ module_info.cpu.c_str(), module_info.debug_identifier.c_str(),
+ module_info.debug_file.c_str());
+
+ PEModuleInfo pe_info;
+ if (!GetPEInfo(&pe_info)) {
+ return false;
+ }
+ fprintf(symbol_file, "INFO CODE_ID %ws %ws\n",
+ pe_info.code_identifier.c_str(),
+ pe_info.code_file.c_str());
+
+ if (!PrintPEFrameData(pe_file_, symbol_file)) {
+ return false;
+ }
+
+ return true;
+}
+
+bool PESourceLineWriter::GetModuleInfo(PDBModuleInfo* info) {
+ return ReadModuleInfo(pe_file_, info);
+}
+
+bool PESourceLineWriter::GetPEInfo(PEModuleInfo* info) {
+ return ReadPEInfo(pe_file_, info);
+}
+
+} // namespace google_breakpad
diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/pe_source_line_writer.h b/toolkit/crashreporter/google-breakpad/src/common/windows/pe_source_line_writer.h new file mode 100644 index 0000000000..2bf1d4fd2b --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/pe_source_line_writer.h @@ -0,0 +1,69 @@ +// Copyright (c) 2019, Google Inc.
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+// * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+// * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+// * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef COMMON_WINDOWS_PE_SOURCE_LINE_WRITER_H_
+#define COMMON_WINDOWS_PE_SOURCE_LINE_WRITER_H_
+
+#include <string>
+
+#include "common/basictypes.h"
+#include "common/windows/module_info.h"
+
+namespace google_breakpad {
+
+using std::wstring;
+
+// PESourceLineWriter uses a pe file produced by Visual C++ to output
+// a line/address map for use with BasicSourceLineResolver.
+// NOTE: Only supports PE32+ format, ie. a 64bit PE file.
+class PESourceLineWriter {
+public:
+ explicit PESourceLineWriter(const wstring& pe_file);
+ ~PESourceLineWriter();
+
+ // Writes Breakpad symbols from the pe file to |symbol_file|.
+ // Returns true on success.
+ bool WriteSymbols(FILE* symbol_file);
+
+ // Retrieves information about the module. Returns true on success.
+ bool GetModuleInfo(PDBModuleInfo* info);
+
+ // Retrieves information about the module's PE file. Returns + // true on success.
+ bool GetPEInfo(PEModuleInfo* info);
+
+private:
+ const wstring pe_file_;
+
+ DISALLOW_COPY_AND_ASSIGN(PESourceLineWriter);
+};
+
+} // namespace google_breakpad
+
+#endif // COMMON_WINDOWS_PE_SOURCE_LINE_WRITER_H_
diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/pe_util.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/pe_util.cc new file mode 100644 index 0000000000..6fa63fa3b5 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/pe_util.cc @@ -0,0 +1,407 @@ +// Copyright (c) 2019, Google Inc.
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+// * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+// * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+// * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#include "pe_util.h"
+
+#include <windows.h>
+#include <winnt.h>
+#include <atlbase.h>
+#include <ImageHlp.h>
+
+#include <functional>
+
+#include "common/windows/string_utils-inl.h"
+#include "common/windows/guid_string.h"
+
+namespace {
+
+/*
+ * Not defined in WinNT.h for some reason. Definitions taken from:
+ * http://uninformed.org/index.cgi?v=4&a=1&p=13
+ *
+ */
+typedef unsigned char UBYTE;
+
+#if !defined(_WIN64)
+#define UNW_FLAG_EHANDLER 0x01
+#define UNW_FLAG_UHANDLER 0x02
+#define UNW_FLAG_CHAININFO 0x04
+#endif
+
+union UnwindCode {
+ struct {
+ UBYTE offset_in_prolog;
+ UBYTE unwind_operation_code : 4;
+ UBYTE operation_info : 4;
+ };
+ USHORT frame_offset;
+};
+
+enum UnwindOperationCodes {
+ UWOP_PUSH_NONVOL = 0, /* info == register number */
+ UWOP_ALLOC_LARGE, /* no info, alloc size in next 2 slots */
+ UWOP_ALLOC_SMALL, /* info == size of allocation / 8 - 1 */
+ UWOP_SET_FPREG, /* no info, FP = RSP + UNWIND_INFO.FPRegOffset*16 */
+ UWOP_SAVE_NONVOL, /* info == register number, offset in next slot */
+ UWOP_SAVE_NONVOL_FAR, /* info == register number, offset in next 2 slots */
+ // XXX: these are missing from MSDN!
+ // See: http://www.osronline.com/ddkx/kmarch/64bitamd_4rs7.htm
+ UWOP_SAVE_XMM,
+ UWOP_SAVE_XMM_FAR,
+ UWOP_SAVE_XMM128, /* info == XMM reg number, offset in next slot */
+ UWOP_SAVE_XMM128_FAR, /* info == XMM reg number, offset in next 2 slots */
+ UWOP_PUSH_MACHFRAME /* info == 0: no error-code, 1: error-code */
+};
+
+// See: http://msdn.microsoft.com/en-us/library/ddssxxy8.aspx
+// Note: some fields removed as we don't use them.
+struct UnwindInfo {
+ UBYTE version : 3;
+ UBYTE flags : 5;
+ UBYTE size_of_prolog;
+ UBYTE count_of_codes;
+ UBYTE frame_register : 4;
+ UBYTE frame_offset : 4;
+ UnwindCode unwind_code[1];
+};
+
+struct CV_INFO_PDB70 {
+ ULONG cv_signature;
+ GUID signature;
+ ULONG age;
+ CHAR pdb_filename[ANYSIZE_ARRAY];
+};
+
+#define CV_SIGNATURE_RSDS 'SDSR'
+
+// A helper class to scope a PLOADED_IMAGE.
+class AutoImage {
+public:
+ explicit AutoImage(PLOADED_IMAGE img) : img_(img) {}
+ ~AutoImage() {
+ if (img_)
+ ImageUnload(img_);
+ }
+
+ operator PLOADED_IMAGE() { return img_; }
+ PLOADED_IMAGE operator->() { return img_; }
+
+private:
+ PLOADED_IMAGE img_;
+};
+} // namespace
+
+namespace google_breakpad {
+
+using std::unique_ptr;
+using google_breakpad::GUIDString;
+
+bool ReadModuleInfo(const wstring & pe_file, PDBModuleInfo * info) {
+ // Convert wchar to native charset because ImageLoad only takes
+ // a PSTR as input.
+ string img_file;
+ if (!WindowsStringUtils::safe_wcstombs(pe_file, &img_file)) {
+ fprintf(stderr, "Image path '%S' contains unrecognized characters.\n",
+ pe_file.c_str());
+ return false;
+ }
+
+ AutoImage img(ImageLoad((PSTR)img_file.c_str(), NULL));
+ if (!img) {
+ fprintf(stderr, "Failed to load %s\n", img_file.c_str());
+ return false;
+ }
+
+ info->cpu = FileHeaderMachineToCpuString(
+ img->FileHeader->FileHeader.Machine);
+
+ PIMAGE_OPTIONAL_HEADER64 optional_header =
+ &(reinterpret_cast<PIMAGE_NT_HEADERS64>(img->FileHeader))->OptionalHeader;
+
+ // Search debug directories for a guid signature & age
+ DWORD debug_rva = optional_header->
+ DataDirectory[IMAGE_DIRECTORY_ENTRY_DEBUG].VirtualAddress;
+ DWORD debug_size = optional_header->
+ DataDirectory[IMAGE_DIRECTORY_ENTRY_DEBUG].Size;
+ PIMAGE_DEBUG_DIRECTORY debug_directories =
+ static_cast<PIMAGE_DEBUG_DIRECTORY>(
+ ImageRvaToVa(img->FileHeader,
+ img->MappedAddress,
+ debug_rva,
+ &img->LastRvaSection));
+
+ for (DWORD i = 0; i < debug_size / sizeof(*debug_directories); i++) {
+ if (debug_directories[i].Type != IMAGE_DEBUG_TYPE_CODEVIEW ||
+ debug_directories[i].SizeOfData < sizeof(CV_INFO_PDB70)) {
+ continue;
+ }
+
+ struct CV_INFO_PDB70* cv_info = static_cast<CV_INFO_PDB70*>(ImageRvaToVa(
+ img->FileHeader,
+ img->MappedAddress,
+ debug_directories[i].AddressOfRawData,
+ &img->LastRvaSection));
+ if (cv_info->cv_signature != CV_SIGNATURE_RSDS) {
+ continue;
+ }
+
+ info->debug_identifier = GenerateDebugIdentifier(cv_info->age,
+ cv_info->signature);
+
+ // This code assumes that the pdb_filename is stored as ASCII without
+ // multibyte characters, but it's not clear if that's true.
+ size_t debug_file_length = strnlen_s(cv_info->pdb_filename, MAX_PATH);
+ if (debug_file_length < 0 || debug_file_length >= MAX_PATH) {
+ fprintf(stderr, "PE debug directory is corrupt.\n");
+ return false;
+ }
+ std::string debug_file(cv_info->pdb_filename, debug_file_length);
+ if (!WindowsStringUtils::safe_mbstowcs(debug_file, &info->debug_file)) {
+ fprintf(stderr, "PDB filename '%s' contains unrecognized characters.\n",
+ debug_file.c_str());
+ return false;
+ }
+ info->debug_file = WindowsStringUtils::GetBaseName(info->debug_file);
+
+ return true;
+ }
+
+ fprintf(stderr, "Image is missing debug information.\n");
+ return false;
+}
+
+bool ReadPEInfo(const wstring & pe_file, PEModuleInfo * info) {
+ // Convert wchar to native charset because ImageLoad only takes
+ // a PSTR as input.
+ string img_file;
+ if (!WindowsStringUtils::safe_wcstombs(pe_file, &img_file)) {
+ fprintf(stderr, "Image path '%S' contains unrecognized characters.\n",
+ pe_file.c_str());
+ return false;
+ }
+
+ AutoImage img(ImageLoad((PSTR)img_file.c_str(), NULL));
+ if (!img) {
+ fprintf(stderr, "Failed to open PE file: %S\n", pe_file.c_str());
+ return false;
+ }
+
+ info->code_file = WindowsStringUtils::GetBaseName(pe_file);
+
+ // The date and time that the file was created by the linker.
+ DWORD TimeDateStamp = img->FileHeader->FileHeader.TimeDateStamp;
+ // The size of the file in bytes, including all headers.
+ DWORD SizeOfImage = 0;
+ PIMAGE_OPTIONAL_HEADER64 opt =
+ &((PIMAGE_NT_HEADERS64)img->FileHeader)->OptionalHeader;
+ if (opt->Magic == IMAGE_NT_OPTIONAL_HDR64_MAGIC) {
+ // 64-bit PE file.
+ SizeOfImage = opt->SizeOfImage;
+ }
+ else {
+ // 32-bit PE file.
+ SizeOfImage = img->FileHeader->OptionalHeader.SizeOfImage;
+ }
+ wchar_t code_identifier[32];
+ swprintf(code_identifier,
+ sizeof(code_identifier) / sizeof(code_identifier[0]),
+ L"%08X%X", TimeDateStamp, SizeOfImage);
+ info->code_identifier = code_identifier;
+
+ return true;
+}
+
+bool PrintPEFrameData(const wstring & pe_file, FILE * out_file)
+{
+ // Convert wchar to native charset because ImageLoad only takes
+ // a PSTR as input.
+ string img_file;
+ if (!WindowsStringUtils::safe_wcstombs(pe_file, &img_file)) {
+ fprintf(stderr, "Image path '%S' contains unrecognized characters.\n",
+ pe_file.c_str());
+ return false;
+ }
+
+ AutoImage img(ImageLoad((PSTR)img_file.c_str(), NULL));
+ if (!img) {
+ fprintf(stderr, "Failed to load %s\n", img_file.c_str());
+ return false;
+ }
+ PIMAGE_OPTIONAL_HEADER64 optional_header =
+ &(reinterpret_cast<PIMAGE_NT_HEADERS64>(img->FileHeader))->OptionalHeader;
+ if (optional_header->Magic != IMAGE_NT_OPTIONAL_HDR64_MAGIC) {
+ fprintf(stderr, "Not a PE32+ image\n");
+ return false;
+ }
+
+ // Read Exception Directory
+ DWORD exception_rva = optional_header->
+ DataDirectory[IMAGE_DIRECTORY_ENTRY_EXCEPTION].VirtualAddress;
+ DWORD exception_size = optional_header->
+ DataDirectory[IMAGE_DIRECTORY_ENTRY_EXCEPTION].Size;
+ PIMAGE_RUNTIME_FUNCTION_ENTRY funcs =
+ static_cast<PIMAGE_RUNTIME_FUNCTION_ENTRY>(
+ ImageRvaToVa(img->FileHeader,
+ img->MappedAddress,
+ exception_rva,
+ &img->LastRvaSection));
+ for (DWORD i = 0; i < exception_size / sizeof(*funcs); i++) {
+ DWORD unwind_rva = funcs[i].UnwindInfoAddress;
+ // handle chaining
+ while (unwind_rva & 0x1) {
+ unwind_rva ^= 0x1;
+ PIMAGE_RUNTIME_FUNCTION_ENTRY chained_func =
+ static_cast<PIMAGE_RUNTIME_FUNCTION_ENTRY>(
+ ImageRvaToVa(img->FileHeader,
+ img->MappedAddress,
+ unwind_rva,
+ &img->LastRvaSection));
+ unwind_rva = chained_func->UnwindInfoAddress;
+ }
+
+ UnwindInfo *unwind_info = static_cast<UnwindInfo *>(
+ ImageRvaToVa(img->FileHeader,
+ img->MappedAddress,
+ unwind_rva,
+ &img->LastRvaSection));
+
+ DWORD stack_size = 8; // minimal stack size is 8 for RIP
+ DWORD rip_offset = 8;
+ do {
+ for (UBYTE c = 0; c < unwind_info->count_of_codes; c++) {
+ UnwindCode *unwind_code = &unwind_info->unwind_code[c];
+ switch (unwind_code->unwind_operation_code) {
+ case UWOP_PUSH_NONVOL: {
+ stack_size += 8;
+ break;
+ }
+ case UWOP_ALLOC_LARGE: {
+ if (unwind_code->operation_info == 0) {
+ c++;
+ if (c < unwind_info->count_of_codes)
+ stack_size += (unwind_code + 1)->frame_offset * 8;
+ }
+ else {
+ c += 2;
+ if (c < unwind_info->count_of_codes)
+ stack_size += (unwind_code + 1)->frame_offset |
+ ((unwind_code + 2)->frame_offset << 16);
+ }
+ break;
+ }
+ case UWOP_ALLOC_SMALL: {
+ stack_size += unwind_code->operation_info * 8 + 8;
+ break;
+ }
+ case UWOP_SET_FPREG:
+ case UWOP_SAVE_XMM:
+ case UWOP_SAVE_XMM_FAR:
+ break;
+ case UWOP_SAVE_NONVOL:
+ case UWOP_SAVE_XMM128: {
+ c++; // skip slot with offset
+ break;
+ }
+ case UWOP_SAVE_NONVOL_FAR:
+ case UWOP_SAVE_XMM128_FAR: {
+ c += 2; // skip 2 slots with offset
+ break;
+ }
+ case UWOP_PUSH_MACHFRAME: {
+ if (unwind_code->operation_info) {
+ stack_size += 88;
+ }
+ else {
+ stack_size += 80;
+ }
+ rip_offset += 80;
+ break;
+ }
+ }
+ }
+ if (unwind_info->flags & UNW_FLAG_CHAININFO) {
+ PIMAGE_RUNTIME_FUNCTION_ENTRY chained_func =
+ reinterpret_cast<PIMAGE_RUNTIME_FUNCTION_ENTRY>(
+ (unwind_info->unwind_code +
+ ((unwind_info->count_of_codes + 1) & ~1)));
+
+ unwind_info = static_cast<UnwindInfo *>(
+ ImageRvaToVa(img->FileHeader,
+ img->MappedAddress,
+ chained_func->UnwindInfoAddress,
+ &img->LastRvaSection));
+ }
+ else {
+ unwind_info = NULL;
+ }
+ } while (unwind_info);
+ fprintf(out_file, "STACK CFI INIT %lx %lx .cfa: $rsp .ra: .cfa %lu - ^\n",
+ funcs[i].BeginAddress,
+ funcs[i].EndAddress - funcs[i].BeginAddress, rip_offset);
+ fprintf(out_file, "STACK CFI %lx .cfa: $rsp %lu +\n",
+ funcs[i].BeginAddress, stack_size);
+ }
+
+ return true;
+}
+
+wstring GenerateDebugIdentifier(DWORD age, GUID signature)
+{
+ // Use the same format that the MS symbol server uses in filesystem
+ // hierarchies.
+ wchar_t age_string[9];
+ swprintf(age_string, sizeof(age_string) / sizeof(age_string[0]),
+ L"%x", age);
+
+ // remove when VC++7.1 is no longer supported
+ age_string[sizeof(age_string) / sizeof(age_string[0]) - 1] = L'\0';
+
+ wstring debug_identifier = GUIDString::GUIDToSymbolServerWString(&signature);
+ debug_identifier.append(age_string);
+
+ return debug_identifier;
+}
+
+wstring GenerateDebugIdentifier(DWORD age, DWORD signature)
+{
+ // Use the same format that the MS symbol server uses in filesystem
+ // hierarchies.
+ wchar_t identifier_string[17];
+ swprintf(identifier_string,
+ sizeof(identifier_string) / sizeof(identifier_string[0]),
+ L"%08X%x", signature, age);
+
+ // remove when VC++7.1 is no longer supported
+ identifier_string[sizeof(identifier_string) /
+ sizeof(identifier_string[0]) - 1] = L'\0';
+
+ return wstring(identifier_string);
+}
+
+} // namespace google_breakpad
diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/pe_util.h b/toolkit/crashreporter/google-breakpad/src/common/windows/pe_util.h new file mode 100644 index 0000000000..634ba29343 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/pe_util.h @@ -0,0 +1,78 @@ +// Copyright (c) 2019, Google Inc.
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+// * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+// * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+// * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+#ifndef COMMON_WINDOWS_PE_UTIL_H_
+#define COMMON_WINDOWS_PE_UTIL_H_
+
+#include <windows.h>
+
+#include "common/windows/module_info.h"
+
+namespace google_breakpad {
+
+using std::wstring;
+
+// Reads |pe_file| and populates |info|. Returns true on success.
+// Only supports PE32+ format, ie. a 64bit PE file.
+// Will fail if |pe_file| does not contain a valid CodeView record.
+bool ReadModuleInfo(const wstring& pe_file, PDBModuleInfo* info);
+
+// Reads |pe_file| and populates |info|. Returns true on success.
+bool ReadPEInfo(const wstring& pe_file, PEModuleInfo* info);
+
+// Reads |pe_file| and prints frame data (aka. unwind info) to |out_file|.
+// Only supports PE32+ format, ie. a 64bit PE file.
+bool PrintPEFrameData(const wstring& pe_file, FILE* out_file);
+
+// Combines a GUID |signature| and DWORD |age| to create a Breakpad debug
+// identifier.
+wstring GenerateDebugIdentifier(DWORD age, GUID signature);
+
+// Combines a DWORD |signature| and DWORD |age| to create a Breakpad debug
+// identifier.
+wstring GenerateDebugIdentifier(DWORD age, DWORD signature);
+
+// Converts |machine| enum value to the corresponding string used by Breakpad.
+// The enum is IMAGE_FILE_MACHINE_*, contained in winnt.h.
+constexpr const wchar_t* FileHeaderMachineToCpuString(WORD machine) {
+ switch (machine) {
+ case IMAGE_FILE_MACHINE_I386: {
+ return L"x86";
+ }
+ case IMAGE_FILE_MACHINE_IA64:
+ case IMAGE_FILE_MACHINE_AMD64: {
+ return L"x86_64";
+ }
+ default: { return L"unknown"; }
+ }
+}
+
+} // namespace google_breakpad
+
+#endif // COMMON_WINDOWS_PE_UTIL_H_
diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/string_utils-inl.h b/toolkit/crashreporter/google-breakpad/src/common/windows/string_utils-inl.h new file mode 100644 index 0000000000..9b63607268 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/string_utils-inl.h @@ -0,0 +1,142 @@ +// Copyright (c) 2006, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +// string_utils-inl.h: Safer string manipulation on Windows, supporting +// pre-MSVC8 environments. + +#ifndef COMMON_WINDOWS_STRING_UTILS_INL_H_ +#define COMMON_WINDOWS_STRING_UTILS_INL_H_ + +#include <stdarg.h> +#include <wchar.h> + +#include <string> + +// The "ll" printf format size specifier corresponding to |long long| was +// intrudced in MSVC8. Earlier versions did not provide this size specifier, +// but "I64" can be used to print 64-bit types. Don't use "I64" where "ll" +// is available, in the event of oddball systems where |long long| is not +// 64 bits wide. +#if _MSC_VER >= 1400 // MSVC 2005/8 +#define WIN_STRING_FORMAT_LL "ll" +#else // MSC_VER >= 1400 +#define WIN_STRING_FORMAT_LL "I64" +#endif // MSC_VER >= 1400 + +// A nonconforming version of swprintf, without the length argument, was +// included with the CRT prior to MSVC8. Although a conforming version was +// also available via an overload, it is not reliably chosen. _snwprintf +// behaves as a standards-confirming swprintf should, so force the use of +// _snwprintf when using older CRTs. +#if _MSC_VER < 1400 // MSVC 2005/8 +#define swprintf _snwprintf +#else +// For MSVC8 and newer, swprintf_s is the recommended method. Conveniently, +// it takes the same argument list as swprintf. +#define swprintf swprintf_s +#endif // MSC_VER < 1400 + +namespace google_breakpad { + +using std::string; +using std::wstring; + +class WindowsStringUtils { + public: + // Roughly equivalent to MSVC8's wcscpy_s, except pre-MSVC8, this does + // not fail if source is longer than destination_size. The destination + // buffer is always 0-terminated. + static void safe_wcscpy(wchar_t *destination, size_t destination_size, + const wchar_t *source); + + // Roughly equivalent to MSVC8's wcsncpy_s, except that _TRUNCATE cannot + // be passed directly, and pre-MSVC8, this will not fail if source or count + // are longer than destination_size. The destination buffer is always + // 0-terminated. + static void safe_wcsncpy(wchar_t *destination, size_t destination_size, + const wchar_t *source, size_t count); + + // Performs multi-byte to wide character conversion on C++ strings, using + // mbstowcs_s (MSVC8) or mbstowcs (pre-MSVC8). Returns false on failure, + // without setting wcs. + static bool safe_mbstowcs(const string &mbs, wstring *wcs); + + // The inverse of safe_mbstowcs. + static bool safe_wcstombs(const wstring &wcs, string *mbs); + + // Returns the base name of a file, e.g. strips off the path. + static wstring GetBaseName(const wstring &filename); + + private: + // Disallow instantiation and other object-based operations. + WindowsStringUtils(); + WindowsStringUtils(const WindowsStringUtils&); + ~WindowsStringUtils(); + void operator=(const WindowsStringUtils&); +}; + +// static +inline void WindowsStringUtils::safe_wcscpy(wchar_t *destination, + size_t destination_size, + const wchar_t *source) { +#if _MSC_VER >= 1400 // MSVC 2005/8 + wcscpy_s(destination, destination_size, source); +#else // _MSC_VER >= 1400 + // Pre-MSVC 2005/8 doesn't have wcscpy_s. Simulate it with wcsncpy. + // wcsncpy doesn't 0-terminate the destination buffer if the source string + // is longer than size. Ensure that the destination is 0-terminated. + wcsncpy(destination, source, destination_size); + if (destination && destination_size) + destination[destination_size - 1] = 0; +#endif // _MSC_VER >= 1400 +} + +// static +inline void WindowsStringUtils::safe_wcsncpy(wchar_t *destination, + size_t destination_size, + const wchar_t *source, + size_t count) { +#if _MSC_VER >= 1400 // MSVC 2005/8 + wcsncpy_s(destination, destination_size, source, count); +#else // _MSC_VER >= 1400 + // Pre-MSVC 2005/8 doesn't have wcsncpy_s. Simulate it with wcsncpy. + // wcsncpy doesn't 0-terminate the destination buffer if the source string + // is longer than size. Ensure that the destination is 0-terminated. + if (destination_size < count) + count = destination_size; + + wcsncpy(destination, source, count); + if (destination && count) + destination[count - 1] = 0; +#endif // _MSC_VER >= 1400 +} + +} // namespace google_breakpad + +#endif // COMMON_WINDOWS_STRING_UTILS_INL_H_ diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/string_utils.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/string_utils.cc new file mode 100644 index 0000000000..272800035c --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/string_utils.cc @@ -0,0 +1,133 @@ +// Copyright (c) 2006, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#include <cassert> +#include <vector> + +#include "common/windows/string_utils-inl.h" + +namespace google_breakpad { + +// static +wstring WindowsStringUtils::GetBaseName(const wstring &filename) { + wstring base_name(filename); + size_t slash_pos = base_name.find_last_of(L"/\\"); + if (slash_pos != wstring::npos) { + base_name.erase(0, slash_pos + 1); + } + return base_name; +} + +// static +bool WindowsStringUtils::safe_mbstowcs(const string &mbs, wstring *wcs) { + assert(wcs); + + // First, determine the length of the destination buffer. + size_t wcs_length; + +#if _MSC_VER >= 1400 // MSVC 2005/8 + errno_t err; + if ((err = mbstowcs_s(&wcs_length, NULL, 0, mbs.c_str(), _TRUNCATE)) != 0) { + return false; + } + assert(wcs_length > 0); +#else // _MSC_VER >= 1400 + if ((wcs_length = mbstowcs(NULL, mbs.c_str(), mbs.length())) == (size_t)-1) { + return false; + } + + // Leave space for the 0-terminator. + ++wcs_length; +#endif // _MSC_VER >= 1400 + + std::vector<wchar_t> wcs_v(wcs_length); + + // Now, convert. +#if _MSC_VER >= 1400 // MSVC 2005/8 + if ((err = mbstowcs_s(NULL, &wcs_v[0], wcs_length, mbs.c_str(), + _TRUNCATE)) != 0) { + return false; + } +#else // _MSC_VER >= 1400 + if (mbstowcs(&wcs_v[0], mbs.c_str(), mbs.length()) == (size_t)-1) { + return false; + } + + // Ensure presence of 0-terminator. + wcs_v[wcs_length - 1] = '\0'; +#endif // _MSC_VER >= 1400 + + *wcs = &wcs_v[0]; + return true; +} + +// static +bool WindowsStringUtils::safe_wcstombs(const wstring &wcs, string *mbs) { + assert(mbs); + + // First, determine the length of the destination buffer. + size_t mbs_length; + +#if _MSC_VER >= 1400 // MSVC 2005/8 + errno_t err; + if ((err = wcstombs_s(&mbs_length, NULL, 0, wcs.c_str(), _TRUNCATE)) != 0) { + return false; + } + assert(mbs_length > 0); +#else // _MSC_VER >= 1400 + if ((mbs_length = wcstombs(NULL, wcs.c_str(), wcs.length())) == (size_t)-1) { + return false; + } + + // Leave space for the 0-terminator. + ++mbs_length; +#endif // _MSC_VER >= 1400 + + std::vector<char> mbs_v(mbs_length); + + // Now, convert. +#if _MSC_VER >= 1400 // MSVC 2005/8 + if ((err = wcstombs_s(NULL, &mbs_v[0], mbs_length, wcs.c_str(), + _TRUNCATE)) != 0) { + return false; + } +#else // _MSC_VER >= 1400 + if (wcstombs(&mbs_v[0], wcs.c_str(), wcs.length()) == (size_t)-1) { + return false; + } + + // Ensure presence of 0-terminator. + mbs_v[mbs_length - 1] = '\0'; +#endif // _MSC_VER >= 1400 + + *mbs = &mbs_v[0]; + return true; +} + +} // namespace google_breakpad diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/symbol_collector_client.cc b/toolkit/crashreporter/google-breakpad/src/common/windows/symbol_collector_client.cc new file mode 100644 index 0000000000..30c663ed37 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/symbol_collector_client.cc @@ -0,0 +1,155 @@ +#include "common/windows/symbol_collector_client.h" + +#include <stdio.h> + +#include <regex> + +#include "common/windows/http_upload.h" + +namespace google_breakpad { + + // static + bool SymbolCollectorClient::CreateUploadUrl( + wstring& api_url, + wstring& api_key, + UploadUrlResponse *uploadUrlResponse) { + wstring url = api_url + + L"/v1/uploads:create" + L"?key=" + api_key; + wstring response; + int response_code; + + if (!HTTPUpload::SendSimplePostRequest( + url, + L"", + L"", + NULL, + &response, + &response_code)) { + wprintf(L"Failed to create upload url.\n"); + wprintf(L"Response code: %ld\n", response_code); + wprintf(L"Response:\n"); + wprintf(L"%s\n", response.c_str()); + return false; + } + + // Note camel-case rather than underscores. + std::wregex upload_url_regex(L"\"uploadUrl\": \"([^\"]+)\""); + std::wregex upload_key_regex(L"\"uploadKey\": \"([^\"]+)\""); + + std::wsmatch upload_url_match; + if (!std::regex_search(response, upload_url_match, upload_url_regex) || + upload_url_match.size() != 2) { + wprintf(L"Failed to parse create url response."); + wprintf(L"Response:\n"); + wprintf(L"%s\n", response.c_str()); + return false; + } + wstring upload_url = upload_url_match[1].str(); + + std::wsmatch upload_key_match; + if (!std::regex_search(response, upload_key_match, upload_key_regex) || + upload_key_match.size() != 2) { + wprintf(L"Failed to parse create url response."); + wprintf(L"Response:\n"); + wprintf(L"%s\n", response.c_str()); + return false; + } + wstring upload_key = upload_key_match[1].str(); + + uploadUrlResponse->upload_url = upload_url; + uploadUrlResponse->upload_key = upload_key; + return true; + } + + // static + CompleteUploadResult SymbolCollectorClient::CompleteUpload( + wstring& api_url, + wstring& api_key, + const wstring& upload_key, + const wstring& debug_file, + const wstring& debug_id) { + wstring url = api_url + + L"/v1/uploads/" + upload_key + L":complete" + L"?key=" + api_key; + wstring body = + L"{ symbol_id: {" + L"debug_file: \"" + debug_file + L"\", " + L"debug_id: \"" + debug_id + L"\" " + L"} }"; + wstring response; + int response_code; + + if (!HTTPUpload::SendSimplePostRequest( + url, + body, + L"application/json", + NULL, + &response, + &response_code)) { + wprintf(L"Failed to complete upload.\n"); + wprintf(L"Response code: %ld\n", response_code); + wprintf(L"Response:\n"); + wprintf(L"%s\n", response.c_str()); + return CompleteUploadResult::Error; + } + + std::wregex result_regex(L"\"result\": \"([^\"]+)\""); + std::wsmatch result_match; + if (!std::regex_search(response, result_match, result_regex) || + result_match.size() != 2) { + wprintf(L"Failed to parse complete upload response."); + wprintf(L"Response:\n"); + wprintf(L"%s\n", response.c_str()); + return CompleteUploadResult::Error; + } + wstring result = result_match[1].str(); + + if (result.compare(L"DUPLICATE_DATA") == 0) { + return CompleteUploadResult::DuplicateData; + } + + return CompleteUploadResult::Ok; + } + + // static + SymbolStatus SymbolCollectorClient::CheckSymbolStatus( + wstring& api_url, + wstring& api_key, + const wstring& debug_file, + const wstring& debug_id) { + wstring response; + int response_code; + wstring url = api_url + + L"/v1/symbols/" + debug_file + L"/" + debug_id + L":checkStatus" + L"?key=" + api_key; + + if (!HTTPUpload::SendGetRequest( + url, + NULL, + &response, + &response_code)) { + wprintf(L"Failed to check symbol status.\n"); + wprintf(L"Response code: %ld\n", response_code); + wprintf(L"Response:\n"); + wprintf(L"%s\n", response.c_str()); + return SymbolStatus::Unknown; + } + + std::wregex status_regex(L"\"status\": \"([^\"]+)\""); + std::wsmatch status_match; + if (!std::regex_search(response, status_match, status_regex) || + status_match.size() != 2) { + wprintf(L"Failed to parse check symbol status response."); + wprintf(L"Response:\n"); + wprintf(L"%s\n", response.c_str()); + return SymbolStatus::Unknown; + } + wstring status = status_match[1].str(); + + return (status.compare(L"FOUND") == 0) ? + SymbolStatus::Found : + SymbolStatus::Missing; + } + +} // namespace google_breakpad
\ No newline at end of file diff --git a/toolkit/crashreporter/google-breakpad/src/common/windows/symbol_collector_client.h b/toolkit/crashreporter/google-breakpad/src/common/windows/symbol_collector_client.h new file mode 100644 index 0000000000..30e0cb3230 --- /dev/null +++ b/toolkit/crashreporter/google-breakpad/src/common/windows/symbol_collector_client.h @@ -0,0 +1,89 @@ +// Copyright (c) 2019, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#ifndef COMMON_WINDOWS_SYMBOL_COLLECTOR_CLIENT_H_ +#define COMMON_WINDOWS_SYMBOL_COLLECTOR_CLIENT_H_ + +#include <string> + +namespace google_breakpad { + + using std::wstring; + + struct UploadUrlResponse { + // URL at which to HTTP PUT symbol file. + wstring upload_url; + // Unique key used to complete upload of symbol file. + wstring upload_key; + }; + + enum SymbolStatus { + Found, + Missing, + Unknown + }; + + enum CompleteUploadResult { + Ok, + DuplicateData, + Error + }; + + // Client to interact with sym-upload-v2 API server via HTTP/REST. + class SymbolCollectorClient { + public: + // Returns a URL at which a symbol file can be HTTP PUT without + // authentication, along with an upload key that can be used to + // complete the upload process with CompleteUpload. + static bool CreateUploadUrl( + wstring& api_url, + wstring& api_key, + UploadUrlResponse *uploadUrlResponse); + + // Notify the API that symbol file upload is finished and its contents + // are ready to be read and/or used for further processing. + static CompleteUploadResult CompleteUpload( + wstring& api_url, + wstring& api_key, + const wstring& upload_key, + const wstring& debug_file, + const wstring& debug_id); + + // Returns whether or not a symbol file corresponding to the debug_file/ + // debug_id pair is already present in symbol storage. + static SymbolStatus CheckSymbolStatus( + wstring& api_url, + wstring& api_key, + const wstring& debug_file, + const wstring& debug_id); + }; + +} // namespace google_breakpad + +#endif // COMMON_WINDOWS_SYMBOL_COLLECTOR_CLIENT_H_ |