ppsspp/Core/HLE/sceKernelMutex.cpp
Unknown W. Brackets 4bfa24eca8 Fix sceKernelCreateMutex() to match tests.
Still have an issue where reusing threads doesn't work.
2012-11-20 02:19:21 -08:00

312 lines
No EOL
7.6 KiB
C++

// Copyright (c) 2012- PPSSPP Project.
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, version 2.0 or later versions.
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License 2.0 for more details.
// A copy of the GPL 2.0 should have been included with the program.
// If not, see http://www.gnu.org/licenses/
// Official git repository and contact information can be found at
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
// UNFINISHED
#include "HLE.h"
#include "../MIPS/MIPS.h"
#include "sceKernel.h"
#include "sceKernelMutex.h"
#include "sceKernelThread.h"
#define PSP_MUTEX_ATTR_FIFO 0
#define PSP_MUTEX_ATTR_PRIORITY 0x100
#define PSP_MUTEX_ATTR_ALLOW_RECURSIVE 0x200
// Not sure about the names of these
#define PSP_MUTEX_ERROR_NOT_LOCKED 0x800201C5
#define PSP_MUTEX_ERROR_NO_SUCH_MUTEX 0x800201C3
#define PSP_MUTEX_ERROR_UNLOCK_UNDERFLOW 0x800201C7
// Guesswork - not exposed anyway
struct NativeMutex
{
SceSize size;
char name[32];
SceUInt attr;
int lockLevel;
int lockThread; // The thread holding the lock
};
struct Mutex : public KernelObject
{
const char *GetName() {return nm.name;}
const char *GetTypeName() {return "Mutex";}
static u32 GetMissingErrorCode() { return PSP_MUTEX_ERROR_NO_SUCH_MUTEX; } // Not sure?
int GetIDType() const { return SCE_KERNEL_TMID_Mutex; }
NativeMutex nm;
std::vector<SceUID> waitingThreads;
};
struct LWMutex : public KernelObject
{
const char *GetName() {return nm.name;}
const char *GetTypeName() {return "LWMutex";}
static u32 GetMissingErrorCode() { return SCE_KERNEL_ERROR_UNKNOWN_SEMID; } // Not sure?
int GetIDType() const { return SCE_KERNEL_TMID_LwMutex; }
NativeMutex nm;
std::vector<SceUID> waitingThreads;
};
void sceKernelCreateMutex(const char *name, u32 attr, int initialCount, u32 optionsPtr)
{
u32 error = 0;
if (!error && !name)
error = SCE_KERNEL_ERROR_ERROR;
if (!error && initialCount < 0)
error = SCE_KERNEL_ERROR_ILLEGAL_COUNT;
if (!error && (attr & PSP_MUTEX_ATTR_ALLOW_RECURSIVE) == 0 && initialCount > 1)
error = SCE_KERNEL_ERROR_ILLEGAL_COUNT;
if (error)
{
RETURN(error);
return;
}
DEBUG_LOG(HLE,"sceKernelCreateMutex(%s, %08x, %d, %08x)", name, attr, initialCount, optionsPtr);
Mutex *mutex = new Mutex();
SceUID id = kernelObjects.Create(mutex);
mutex->nm.size = sizeof(mutex);
strncpy(mutex->nm.name, name, 31);
mutex->nm.name[31] = 0;
mutex->nm.attr = attr;
mutex->nm.lockLevel = initialCount;
mutex->nm.lockThread = __KernelGetCurThread();
RETURN(id);
__KernelReSchedule("mutex created");
}
void sceKernelDeleteMutex(SceUID id)
{
DEBUG_LOG(HLE,"sceKernelDeleteMutex(%i)", id);
u32 error;
Mutex *mutex = kernelObjects.Get<Mutex>(id, error);
if (mutex)
{
std::vector<SceUID>::iterator iter, end;
for (iter = mutex->waitingThreads.begin(), end = mutex->waitingThreads.end(); iter != end; ++iter)
{
SceUID threadID = *iter;
// TODO: Set returnValue?
__KernelResumeThreadFromWait(threadID);
}
mutex->waitingThreads.empty();
RETURN(kernelObjects.Destroy<Mutex>(id));
__KernelReSchedule("mutex deleted");
}
else
RETURN(error);
}
// int sceKernelLockMutex(SceUID id, int count, int *timeout)
// void because it changes threads.
void sceKernelLockMutex(SceUID id, int count, u32 timeoutPtr)
{
DEBUG_LOG(HLE,"sceKernelLockMutex(%i, %i, %08x)", id, count, timeoutPtr);
u32 error;
Mutex *mutex = kernelObjects.Get<Mutex>(id, error);
if (!mutex)
{
RETURN(PSP_MUTEX_ERROR_NO_SUCH_MUTEX);
return;
}
if (count <= 0)
{
RETURN(SCE_KERNEL_ERROR_ILLEGAL_COUNT);
return;
}
if (count > 1 && !(mutex->nm.attr & PSP_MUTEX_ATTR_ALLOW_RECURSIVE))
{
RETURN(SCE_KERNEL_ERROR_ILLEGAL_COUNT);
return;
}
RETURN(0);
if (mutex->nm.lockLevel == 0)
{
mutex->nm.lockLevel += count;
mutex->nm.lockThread = __KernelGetCurThread();
// Needed to get the proper order per real PSP.
__KernelReSchedule("mutex locked");
}
else if ((mutex->nm.attr & PSP_MUTEX_ATTR_ALLOW_RECURSIVE) && mutex->nm.lockThread == __KernelGetCurThread())
{
// Recursive mutex, let's just increase the lock count and keep going
mutex->nm.lockLevel += count;
}
else
{
mutex->waitingThreads.push_back(__KernelGetCurThread());
__KernelWaitCurThread(WAITTYPE_MUTEX, id, count, 0, false);
}
}
// int sceKernelLockMutexCB(SceUID id, int count, int *timeout)
// void because it changes threads.
void sceKernelLockMutexCB(SceUID id, int count, u32 timeoutPtr)
{
DEBUG_LOG(HLE,"sceKernelLockMutexCB(%i, %i, %08x)", id, count, timeoutPtr);
u32 error;
Mutex *mutex = kernelObjects.Get<Mutex>(id, error);
if (!mutex)
{
RETURN(PSP_MUTEX_ERROR_NO_SUCH_MUTEX);
return;
}
if (count <= 0)
{
RETURN(SCE_KERNEL_ERROR_ILLEGAL_COUNT);
return;
}
if (count > 1 && !(mutex->nm.attr & PSP_MUTEX_ATTR_ALLOW_RECURSIVE))
{
RETURN(SCE_KERNEL_ERROR_ILLEGAL_COUNT);
return;
}
RETURN(0);
if (mutex->nm.lockLevel == 0)
{
mutex->nm.lockLevel += count;
mutex->nm.lockThread = __KernelGetCurThread();
// Nobody had it locked - no need to block
}
else if ((mutex->nm.attr & PSP_MUTEX_ATTR_ALLOW_RECURSIVE) && mutex->nm.lockThread == __KernelGetCurThread())
{
// Recursive mutex, let's just increase the lock count and keep going
mutex->nm.lockLevel += count;
}
else
{
mutex->waitingThreads.push_back(__KernelGetCurThread());
__KernelWaitCurThread(WAITTYPE_MUTEX, id, count, 0, true);
__KernelCheckCallbacks();
}
__KernelReSchedule("mutex locked");
}
// int sceKernelUnlockMutex(SceUID id, int count)
// void because it changes threads.
void sceKernelUnlockMutex(SceUID id, int count)
{
DEBUG_LOG(HLE,"sceKernelUnlockMutex(%i, %i)", id, count);
u32 error;
Mutex *mutex = kernelObjects.Get<Mutex>(id, error);
if (!mutex)
{
RETURN(PSP_MUTEX_ERROR_NO_SUCH_MUTEX);
return;
}
if (mutex->nm.lockLevel == 0)
{
RETURN(PSP_MUTEX_ERROR_NOT_LOCKED);
return;
}
if (mutex->nm.lockLevel < count)
{
RETURN(PSP_MUTEX_ERROR_UNLOCK_UNDERFLOW);
return;
}
mutex->nm.lockLevel -= count;
RETURN(0);
if (mutex->nm.lockLevel == 0)
{
mutex->nm.lockThread = -1;
// TODO: PSP_MUTEX_ATTR_PRIORITY
bool wokeThreads = false;
std::vector<SceUID>::iterator iter, end;
for (iter = mutex->waitingThreads.begin(), end = mutex->waitingThreads.end(); iter != end; ++iter)
{
SceUID threadID = *iter;
int wVal = (int)__KernelGetWaitValue(threadID, error);
mutex->nm.lockThread = threadID;
mutex->nm.lockLevel = wVal;
__KernelResumeThreadFromWait(threadID);
wokeThreads = true;
mutex->waitingThreads.erase(iter);
break;
}
// Not sure if this should actually resched, need to test.
if (wokeThreads)
__KernelReSchedule("mutex unlocked");
}
}
struct NativeLwMutex
{
SceSize size;
char name[32];
SceUInt attr;
SceUID mutexUid;
SceUInt opaqueWorkAreaAddr;
int numWaitThreads;
int locked;
int threadid; // thread holding the lock
};
void sceKernelCreateLwMutex()
{
ERROR_LOG(HLE,"UNIMPL sceKernelCreateLwMutex()");
RETURN(0);
}
void sceKernelDeleteLwMutex()
{
ERROR_LOG(HLE,"UNIMPL sceKernelDeleteLwMutex()");
RETURN(0);
}
void sceKernelTryLockLwMutex()
{
ERROR_LOG(HLE,"UNIMPL sceKernelTryLockLwMutex()");
RETURN(0);
}
void sceKernelLockLwMutex()
{
ERROR_LOG(HLE,"UNIMPL sceKernelLockLwMutex()");
RETURN(0);
}
void sceKernelLockLwMutexCB()
{
ERROR_LOG(HLE,"UNIMPL sceKernelLockLwMutexCB()");
RETURN(0);
}
void sceKernelUnlockLwMutex()
{
ERROR_LOG(HLE,"UNIMPL void sceKernelUnlockLwMutex()");
RETURN(0);
}