crashpad/util/win/exception_handler_server_test.cc

214 lines
6.3 KiB
C++
Raw Normal View History

win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
// Copyright 2015 The Crashpad Authors. All rights reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "util/win/exception_handler_server.h"
#include <windows.h>
#include <sys/types.h>
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
#include <string>
#include <vector>
#include "base/macros.h"
#include "base/strings/utf_string_conversions.h"
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
#include "client/crashpad_client.h"
#include "gtest/gtest.h"
#include "test/win/win_child_process.h"
#include "util/thread/thread.h"
#include "util/win/address_types.h"
#include "util/win/registration_protocol_win.h"
#include "util/win/scoped_handle.h"
namespace crashpad {
namespace test {
namespace {
// Runs the ExceptionHandlerServer on a background thread.
class RunServerThread : public Thread {
public:
// Instantiates a thread which will invoke server->Run(delegate).
RunServerThread(ExceptionHandlerServer* server,
ExceptionHandlerServer::Delegate* delegate)
: server_(server), delegate_(delegate) {}
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
~RunServerThread() override {}
private:
// Thread:
void ThreadMain() override { server_->Run(delegate_); }
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
ExceptionHandlerServer* server_;
ExceptionHandlerServer::Delegate* delegate_;
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
DISALLOW_COPY_AND_ASSIGN(RunServerThread);
};
class TestDelegate : public ExceptionHandlerServer::Delegate {
public:
explicit TestDelegate(HANDLE server_ready) : server_ready_(server_ready) {}
~TestDelegate() override {}
void ExceptionHandlerServerStarted() override {
SetEvent(server_ready_);
}
unsigned int ExceptionHandlerServerException(
HANDLE process,
WinVMAddress exception_information_address,
WinVMAddress debug_critical_section_address) override {
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
return 0;
}
void WaitForStart() { WaitForSingleObject(server_ready_, INFINITE); }
private:
HANDLE server_ready_; // weak
bool started_;
DISALLOW_COPY_AND_ASSIGN(TestDelegate);
};
class ExceptionHandlerServerTest : public testing::Test {
public:
ExceptionHandlerServerTest()
: server_(true),
pipe_name_(L"\\\\.\\pipe\\test_name"),
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
server_ready_(CreateEvent(nullptr, false, false, nullptr)),
delegate_(server_ready_.get()),
server_thread_(&server_, &delegate_) {
server_.SetPipeName(pipe_name_);
}
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
TestDelegate& delegate() { return delegate_; }
ExceptionHandlerServer& server() { return server_; }
Thread& server_thread() { return server_thread_; }
const std::wstring& pipe_name() const { return pipe_name_; }
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
private:
ExceptionHandlerServer server_;
std::wstring pipe_name_;
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
ScopedKernelHANDLE server_ready_;
TestDelegate delegate_;
RunServerThread server_thread_;
DISALLOW_COPY_AND_ASSIGN(ExceptionHandlerServerTest);
};
// During destruction, ensures that the server is stopped and the background
// thread joined.
class ScopedStopServerAndJoinThread {
public:
ScopedStopServerAndJoinThread(ExceptionHandlerServer* server, Thread* thread)
: server_(server), thread_(thread) {}
~ScopedStopServerAndJoinThread() {
server_->Stop();
thread_->Join();
}
private:
ExceptionHandlerServer* server_;
Thread* thread_;
DISALLOW_COPY_AND_ASSIGN(ScopedStopServerAndJoinThread);
};
TEST_F(ExceptionHandlerServerTest, Instantiate) {
}
TEST_F(ExceptionHandlerServerTest, StartAndStop) {
server_thread().Start();
ScopedStopServerAndJoinThread scoped_stop_server_and_join_thread(
&server(), &server_thread());
ASSERT_NO_FATAL_FAILURE(delegate().WaitForStart());
}
TEST_F(ExceptionHandlerServerTest, StopWhileConnected) {
server_thread().Start();
ScopedStopServerAndJoinThread scoped_stop_server_and_join_thread(
&server(), &server_thread());
ASSERT_NO_FATAL_FAILURE(delegate().WaitForStart());
CrashpadClient client;
client.SetHandlerIPCPipe(pipe_name());
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
// Leaving this scope causes the server to be stopped, while the connection
// is still open.
}
std::wstring ReadWString(FileHandle handle) {
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
size_t length = 0;
EXPECT_TRUE(LoggingReadFile(handle, &length, sizeof(length)));
std::wstring str(length, L'\0');
if (length > 0) {
EXPECT_TRUE(LoggingReadFile(handle, &str[0], length * sizeof(str[0])));
}
return str;
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
}
void WriteWString(FileHandle handle, const std::wstring& str) {
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
size_t length = str.size();
EXPECT_TRUE(LoggingWriteFile(handle, &length, sizeof(length)));
if (length > 0) {
EXPECT_TRUE(LoggingWriteFile(handle, &str[0], length * sizeof(str[0])));
}
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
}
class TestClient final : public WinChildProcess {
public:
TestClient() : WinChildProcess() {}
~TestClient() {}
private:
int Run() override {
std::wstring pipe_name = ReadWString(ReadPipeHandle());
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
CrashpadClient client;
if (!client.SetHandlerIPCPipe(pipe_name)) {
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
ADD_FAILURE();
return EXIT_FAILURE;
}
WriteWString(WritePipeHandle(), L"OK");
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
return EXIT_SUCCESS;
}
DISALLOW_COPY_AND_ASSIGN(TestClient);
};
TEST_F(ExceptionHandlerServerTest, MultipleConnections) {
WinChildProcess::EntryPoint<TestClient>();
std::unique_ptr<WinChildProcess::Handles> handles_1 =
WinChildProcess::Launch();
std::unique_ptr<WinChildProcess::Handles> handles_2 =
WinChildProcess::Launch();
std::unique_ptr<WinChildProcess::Handles> handles_3 =
WinChildProcess::Launch();
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
// Must ensure the delegate outlasts the server.
{
server_thread().Start();
ScopedStopServerAndJoinThread scoped_stop_server_and_join_thread(
&server(), &server_thread());
ASSERT_NO_FATAL_FAILURE(delegate().WaitForStart());
// Tell all the children where to connect.
WriteWString(handles_1->write.get(), pipe_name());
WriteWString(handles_2->write.get(), pipe_name());
WriteWString(handles_3->write.get(), pipe_name());
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
ASSERT_EQ(L"OK", ReadWString(handles_3->read.get()));
ASSERT_EQ(L"OK", ReadWString(handles_2->read.get()));
ASSERT_EQ(L"OK", ReadWString(handles_1->read.get()));
win: Crash handler server This replaces the registration server, and adds dispatch to a delegate on crash requests. (As you are already aware) we went around in circles on trying to come up with a slightly-too-fancy threading design. All of them seemed to have problems when it comes to out of order events, and orderly shutdown, so I've gone back to something not-too-fancy. Two named pipe instances (that clients connect to) are created. These are used only for registration (which should take <1ms), so 2 should be sufficient to avoid any waits. When a client registers, we duplicate an event to it, which is used to signal when it wants a dump taken. The server registers threadpool waits on that event, and also on the process handle (which will be signalled when the client process exits). These requests (in particular the taking of the dump) are serviced on the threadpool, which avoids us needing to manage those threads, but still allows parallelism in taking dumps. On process termination, we use an IO Completion Port to post a message back to the main thread to request cleanup. This complexity is necessary so that we can unregister the threadpool waits without being on the threadpool, which we need to do synchronously so that we can be sure that no further callbacks will execute (and expect to have the client data around still). In a followup, I will readd support for DumpWithoutCrashing -- I don't think it will be too difficult now that we have an orderly way to clean up client records in the server. R=cpu@chromium.org, mark@chromium.org, jschuh@chromium.org BUG=crashpad:1,crashpad:45 Review URL: https://codereview.chromium.org/1301853002 .
2015-09-03 11:06:17 -07:00
}
}
} // namespace
} // namespace test
} // namespace crashpad