Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

fix: adapt blocking of handlers for Linux and Windows #109

Merged
merged 1 commit into from
Jul 23, 2024
Merged

fix: adapt blocking of handlers for Linux and Windows #109

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
48 changes: 38 additions & 10 deletions client/crashpad_client_linux.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -207,16 +207,30 @@ class SignalHandler {
static void HandleOrReraiseSignal(int signo,
siginfo_t* siginfo,
void* context) {
if (handler_->first_chance_handler_ &&
handler_->first_chance_handler_(
signo, siginfo, static_cast<ucontext_t*>(context))) {
return;
}

// Only handle the first fatal signal observed. If another thread receives a
// crash signal, it waits for the first dump to complete instead of
// requesting another.
if (!handler_->disabled_.test_and_set()) {
// but we also need to ensure that a handling thread that gets preempted
// with a non-masked signal, can proceed to handle that signal.
SignalHandler::handling_signal_on_thread_ = true;
}
if (SignalHandler::handling_signal_on_thread_) {
// TODO(supervacuus):
// On Linux the first-chance handler is executed in the context of a
// signal-handler, which can be asynchronous (crashpad handles `SIGQUIT`)
// and even interrupt itself if the first-chance handler crashes with a
// non-masked synchronous signal. This means we not only need to be safe
// with multiple threads raising signals with thread-specific signal-
// masks, but also with reentrancy from the same thread. Adapting our
// first-chance handler to this would mean to adapt the pipeline, because
// even if the handler was safe wrt the above scenarios, there is
// currently no way in the event model to correlate crashes this way.
if (handler_->first_chance_handler_ &&
handler_->first_chance_handler_(
signo, siginfo, static_cast<ucontext_t*>(context))) {
return;
}
handler_->HandleCrash(signo, siginfo, context);
handler_->WakeThreads();
if (handler_->last_chance_handler_ &&
Expand Down Expand Up @@ -274,10 +288,12 @@ class SignalHandler {
#else
std::atomic_flag disabled_;
#endif
thread_local static bool handling_signal_on_thread_;

static SignalHandler* handler_;
};
SignalHandler* SignalHandler::handler_ = nullptr;
thread_local bool SignalHandler::handling_signal_on_thread_ = false;

// Launches a single use handler to snapshot this process.
class LaunchAtCrashHandler : public SignalHandler {
Expand Down Expand Up @@ -470,8 +486,14 @@ bool CrashpadClient::StartHandler(
return false;
}

std::vector<std::string> argv = BuildHandlerArgvStrings(
handler, database, metrics_dir, url, http_proxy, annotations, arguments, attachments);
std::vector<std::string> argv = BuildHandlerArgvStrings(handler,
database,
metrics_dir,
url,
http_proxy,
annotations,
arguments,
attachments);

argv.push_back(FormatArgumentInt("initial-client-fd", handler_sock.get()));
argv.push_back("--shared-client-connection");
Expand Down Expand Up @@ -694,8 +716,14 @@ bool CrashpadClient::StartHandlerAtCrash(
const std::map<std::string, std::string>& annotations,
const std::vector<std::string>& arguments,
const std::vector<base::FilePath>& attachments) {
std::vector<std::string> argv = BuildHandlerArgvStrings(
handler, database, metrics_dir, url, http_proxy, annotations, arguments, attachments);
std::vector<std::string> argv = BuildHandlerArgvStrings(handler,
database,
metrics_dir,
url,
http_proxy,
annotations,
arguments,
attachments);

auto signal_handler = LaunchAtCrashHandler::Get();
return signal_handler->Initialize(&argv, nullptr, &unhandled_signals_);
Expand Down
15 changes: 11 additions & 4 deletions client/crashpad_client_win.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -155,10 +155,6 @@ LONG WINAPI UnhandledExceptionHandler(EXCEPTION_POINTERS* exception_pointers) {
return EXCEPTION_CONTINUE_SEARCH;
}

if (first_chance_handler_ && first_chance_handler_(exception_pointers)) {
return EXCEPTION_CONTINUE_SEARCH;
}

// Otherwise, we know the handler startup has succeeded, and we can continue.

// Tracks whether a thread has already entered UnhandledExceptionHandler.
Expand All @@ -180,6 +176,17 @@ LONG WINAPI UnhandledExceptionHandler(EXCEPTION_POINTERS* exception_pointers) {
SleepEx(INFINITE, false);
}

// TODO(supervacuus):
// On Windows the first-chance handler is executed inside the UEF which is
// synchronous with respect to thread it is executing on. However, any other
// running thread can also raise an exception and therefore will also execute
// a UEF, at which point they would run concurrently. Adapting to this would
// mean to adapt the pipeline, because even if the handler was UEF-safe,
// there is currently no way in which to correlate non-nested crashes.
if (first_chance_handler_ && first_chance_handler_(exception_pointers)) {
return EXCEPTION_CONTINUE_SEARCH;
}

// Otherwise, we're the first thread, so record the exception pointer and
// signal the crash handler.
g_crash_exception_information.thread_id = GetCurrentThreadId();
Expand Down