Mirrors/yuzu
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Kernel: Added real support for thread and event blocking

Browse source 
- SVC: Added ExitThread support
- SVC: Added SignalEvent support
- Thread: Added WAITTYPE_EVENT for waiting threads for event signals
- Thread: Added support for blocking on other threads to finish (e.g. Thread::Join)
- Thread: Added debug function for printing current threads ready for execution
- Thread: Removed hack/broken thread ready state code from Kernel::Reschedule
- Mutex: Moved WaitCurrentThread from SVC to Mutex::WaitSynchronization
- Event: Added support for blocking threads on event signalling

Kernel: Added missing algorithm #include for use of std::find on non-Windows platforms.
This commit is contained in:
bunnei 2014-06-05 22:35:36 -04:00
parent a002abf171
commit f5c7c15434
6 changed files with 196 additions and 76 deletions
Download patch file Download diff file Expand all files Collapse all files

59
src/core/hle/svc.cpp
View file

@ -93,8 +93,8 @@ Result SendSyncRequest(Handle handle) {
bool wait = false;
Kernel::Object* object = Kernel::g_object_pool.GetFast<Kernel::Object>(handle);
DEBUG_LOG(SVC, "called handle=0x%08X", handle);
_assert_msg_(KERNEL, object, "called, but kernel object is NULL!");
DEBUG_LOG(SVC, "called handle=0x%08X(%s)", handle, object->GetTypeName());
Result res = object->SyncRequest(&wait);
if (wait) {
@ -115,29 +115,21 @@ Result CloseHandle(Handle handle) {
Result WaitSynchronization1(Handle handle, s64 nano_seconds) {
// TODO(bunnei): Do something with nano_seconds, currently ignoring this
bool wait = false;
bool wait_infinite = (nano_seconds == -1); // Used to wait until a thread has terminated
Kernel::Object* object = Kernel::g_object_pool.GetFast<Kernel::Object>(handle);
DEBUG_LOG(SVC, "called handle=0x%08X, nanoseconds=%d", handle,
nano_seconds);
DEBUG_LOG(SVC, "called handle=0x%08X(%s:%s), nanoseconds=%d", handle, object->GetTypeName(),
object->GetName(), nano_seconds);
_assert_msg_(KERNEL, object, "called, but kernel object is NULL!");
Result res = object->WaitSynchronization(&wait);
// Check for next thread to schedule
if (wait) {
// Set current thread to wait state if handle was not unlocked
Kernel::WaitCurrentThread(WAITTYPE_SYNCH); // TODO(bunnei): Is this correct?
// Check for next thread to schedule
HLE::Reschedule(__func__);
// Context switch - Function blocked, is not actually returning (will be "called" again)
// TODO(bunnei): This saves handle to R0 so that it's correctly reloaded on context switch
// (otherwise R0 will be set to whatever is returned, and handle will be invalid when this
// thread is resumed). There is probably a better way of keeping track of state so that we
// don't necessarily have to do this.
return (Result)PARAM(0);
return 0;
}
return res;
@ -150,6 +142,7 @@ Result WaitSynchronizationN(void* _out, void* _handles, u32 handle_count, u32 wa
s32* out = (s32*)_out;
Handle* handles = (Handle*)_handles;
bool unlock_all = true;
bool wait_infinite = (nano_seconds == -1); // Used to wait until a thread has terminated
DEBUG_LOG(SVC, "called handle_count=%d, wait_all=%s, nanoseconds=%d",
handle_count, (wait_all ? "true" : "false"), nano_seconds);
@ -162,7 +155,8 @@ Result WaitSynchronizationN(void* _out, void* _handles, u32 handle_count, u32 wa
_assert_msg_(KERNEL, object, "called handle=0x%08X, but kernel object "
"is NULL!", handles[i]);
DEBUG_LOG(SVC, "\thandle[%d] = 0x%08X", i, handles[i]);
DEBUG_LOG(SVC, "\thandle[%d] = 0x%08X(%s:%s)", i, handles[i], object->GetTypeName(),
object->GetName());
Result res = object->WaitSynchronization(&wait);
@ -179,19 +173,10 @@ Result WaitSynchronizationN(void* _out, void* _handles, u32 handle_count, u32 wa
return 0;
}
// Set current thread to wait state if not all handles were unlocked
Kernel::WaitCurrentThread(WAITTYPE_SYNCH); // TODO(bunnei): Is this correct?
// Check for next thread to schedule
HLE::Reschedule(__func__);
// Context switch - Function blocked, is not actually returning (will be "called" again)
// TODO(bunnei): This saves handle to R0 so that it's correctly reloaded on context switch
// (otherwise R0 will be set to whatever is returned, and handle will be invalid when this
// thread is resumed). There is probably a better way of keeping track of state so that we
// don't necessarily have to do this.
return (Result)PARAM(0);
return 0;
}
/// Create an address arbiter (to allocate access to shared resources)
@ -258,6 +243,17 @@ Result CreateThread(u32 priority, u32 entry_point, u32 arg, u32 stack_top, u32 p
return 0;
}
/// Called when a thread exits
u32 ExitThread() {
Handle thread = Kernel::GetCurrentThreadHandle();
DEBUG_LOG(SVC, "called, pc=0x%08X", Core::g_app_core->GetPC()); // PC = 0x0010545C
Kernel::StopThread(thread, __func__);
HLE::Reschedule(__func__);
return 0;
}
/// Gets the priority for the specified thread
Result GetThreadPriority(void* _priority, Handle handle) {
s32* priority = (s32*)_priority;
@ -326,6 +322,13 @@ Result DuplicateHandle(void* _out, Handle handle) {
return 0;
}
/// Signals an event
Result SignalEvent(Handle evt) {
Result res = Kernel::SignalEvent(evt);
DEBUG_LOG(SVC, "called event=0x%08X", evt);
return res;
}
/// Clears an event
Result ClearEvent(Handle evt) {
Result res = Kernel::ClearEvent(evt);
@ -348,7 +351,7 @@ const HLE::FunctionDef SVC_Table[] = {
{0x06, NULL, "GetProcessIdealProcessor"},
{0x07, NULL, "SetProcessIdealProcessor"},
{0x08, WrapI_UUUUU<CreateThread>, "CreateThread"},
{0x09, NULL, "ExitThread"},
{0x09, WrapU_V<ExitThread>, "ExitThread"},
{0x0A, WrapV_S64<SleepThread>, "SleepThread"},
{0x0B, WrapI_VU<GetThreadPriority>, "GetThreadPriority"},
{0x0C, WrapI_UI<SetThreadPriority>, "SetThreadPriority"},
@ -363,7 +366,7 @@ const HLE::FunctionDef SVC_Table[] = {
{0x15, NULL, "CreateSemaphore"},
{0x16, NULL, "ReleaseSemaphore"},
{0x17, WrapI_VU<CreateEvent>, "CreateEvent"},
{0x18, NULL, "SignalEvent"},
{0x18, WrapI_U<SignalEvent>, "SignalEvent"},
{0x19, WrapI_U<ClearEvent>, "ClearEvent"},
{0x1A, NULL, "CreateTimer"},
{0x1B, NULL, "SetTimer"},