kernel: fix incorrect locking order in suspension
This commit is contained in:
parent
78df1ddce8
commit
5086380a63
3 changed files with 26 additions and 34 deletions
|
@ -1198,27 +1198,34 @@ void KernelCore::Suspend(bool suspended) {
|
|||
const bool should_suspend{exception_exited || suspended};
|
||||
const auto activity = should_suspend ? ProcessActivity::Paused : ProcessActivity::Runnable;
|
||||
|
||||
std::vector<KScopedAutoObject<KThread>> process_threads;
|
||||
{
|
||||
KScopedSchedulerLock sl{*this};
|
||||
|
||||
if (auto* process = CurrentProcess(); process != nullptr) {
|
||||
process->SetActivity(activity);
|
||||
|
||||
if (!should_suspend) {
|
||||
// Runnable now; no need to wait.
|
||||
return;
|
||||
}
|
||||
|
||||
for (auto* thread : process->GetThreadList()) {
|
||||
process_threads.emplace_back(thread);
|
||||
}
|
||||
}
|
||||
//! This refers to the application process, not the current process.
|
||||
KScopedAutoObject<KProcess> process = CurrentProcess();
|
||||
if (process.IsNull()) {
|
||||
return;
|
||||
}
|
||||
|
||||
// Wait for execution to stop.
|
||||
for (auto& thread : process_threads) {
|
||||
thread->WaitUntilSuspended();
|
||||
// Set the new activity.
|
||||
process->SetActivity(activity);
|
||||
|
||||
// Wait for process execution to stop.
|
||||
bool must_wait{should_suspend};
|
||||
|
||||
// KernelCore::Suspend must be called from locked context, or we
|
||||
// could race another call to SetActivity, interfering with waiting.
|
||||
while (must_wait) {
|
||||
KScopedSchedulerLock sl{*this};
|
||||
|
||||
// Assume that all threads have finished running.
|
||||
must_wait = false;
|
||||
|
||||
for (auto i = 0; i < static_cast<s32>(Core::Hardware::NUM_CPU_CORES); ++i) {
|
||||
if (Scheduler(i).GetSchedulerCurrentThread()->GetOwnerProcess() ==
|
||||
process.GetPointerUnsafe()) {
|
||||
// A thread has not finished running yet.
|
||||
// Continue waiting.
|
||||
must_wait = true;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue