Merge pull request #3743 from YossiYad/carry/ampr-pad-launch

feat(libSceAmpr): implement the std::_Pad thread launch helpers

core/libs/prx/libSceAmpr/Export.cpp

@@ -12,6 +12,7 @@
 #include <cstring>
 #include <limits>
 #include <stdexcept>
+#include <string>
 #include <thread>
 
 static constexpr int SCE_AMPR_ERROR_BUFFER_FULL = 0x8002001C;
@@ -224,8 +225,17 @@ static int RecordWriteAddressFromCounterPair(Apr::CommandBufferObject* buffer, v
 }
 
 static constexpr int ThrdSuccess = 0;
+static constexpr int ThrdNomem = 1;
+static constexpr int ThrdTimedout = 2;
+static constexpr int ThrdBusy = 3;
 static constexpr int ThrdError = 4;
 
+static constexpr int CppDeviceOrResourceBusy = 0;
+static constexpr int CppInvalidArgument = 1;
+static constexpr int CppResourceUnavailableTryAgain = 6;
+
+static constexpr int MtxPlain = 0x1;
+
 static std::atomic_flag sharedPtrSpinLock;
 
 struct AmmUsageStatsData {
@@ -240,14 +250,53 @@ static_assert(sizeof(AmmUsageStatsData) == 0x18);
 static constexpr std::uint16_t AmmPageTablePoolEntries = 512;
 static constexpr std::uint32_t AmmRingIdleFlags = 0x7;
 
+struct GuestPad {
+    unsigned (APS5_VABI* const* vtable)(GuestPad*);
+    void* condition;
+    void* mutex;
+    bool started;
+};
+
+static_assert(offsetof(GuestPad, condition) == 0x8 && offsetof(GuestPad, mutex) == 0x10 && offsetof(GuestPad, started) == 0x18 && sizeof(GuestPad) == 0x20);
+
 extern "C" {
+int APS5_VABI pthread_create_nid_postfix(Pthread* thread, const PthreadAttr* attr, PthreadEntry entry, void* arg);
 int APS5_VABI pthread_join_nid_postfix(Pthread thread, void** value);
+int APS5_VABI _Mtx_init_nid_postfix(void** handle, int type);
+void APS5_VABI _Mtx_destroy_nid_postfix(void** handle);
+int APS5_VABI _Mtx_lock_nid_postfix(void** handle);
+int APS5_VABI _Mtx_unlock_nid_postfix(void** handle);
+int APS5_VABI _Cnd_init_nid_postfix(void** handle);
+void APS5_VABI _Cnd_destroy_nid_postfix(void** handle);
+int APS5_VABI _Cnd_wait_nid_postfix(void** condition, void** mutex);
+int APS5_VABI _Cnd_signal_nid_postfix(void** handle);
 int APS5_VABI wcsrtombs_s_nid_postfix(std::size_t* result, char* destination, std::size_t capacity, const std::uint16_t** source, std::size_t limit, void* state);
 int APS5_VABI _Cnd_init_nid_postfix(void** handle);
 int SignalCondition_nid_no_patch(void** handle);
 [[noreturn]] void ThrowCppError_nid_no_patch(int code);
 }
 
+static void RequireThrd(int result, const char* what) {
+    switch (result) {
+    case ThrdSuccess:
+        return;
+    case ThrdNomem:
+    case ThrdTimedout:
+        ThrowCppError_nid_no_patch(CppResourceUnavailableTryAgain);
+    case ThrdBusy:
+        ThrowCppError_nid_no_patch(CppDeviceOrResourceBusy);
+    case ThrdError:
+        ThrowCppError_nid_no_patch(CppInvalidArgument);
+    }
+    throw std::runtime_error(std::string(what) + " failed with " + std::to_string(result));
+}
+
+static void* APS5_VABI PadEntry_nid_no_patch(void* argument) {
+    auto* pad = static_cast<GuestPad*>(argument);
+    const auto go = pad->vtable[0];
+    return reinterpret_cast<void*>(static_cast<std::uintptr_t>(go(pad)));
+}
+
 extern "C" {
 
 int APS5_VABI sceAmprCommandBufferWriteAddressOnCompletion(Apr::CommandBufferObject* buffer, volatile std::uint64_t* address, std::uint64_t value) {
@@ -865,19 +914,36 @@ int APS5_VABI _Thrd_join_nid_postfix(Pthread thread, int* code) {
     return ThrdSuccess;
 }
 
-int APS5_VABI _ZNSt4_Pad8_ReleaseEv_nid_postfix(void) {
-    NotImplemented_nid_no_patch(__func__);
-    return 0;
+void APS5_VABI _ZNSt4_Pad8_ReleaseEv_nid_postfix(GuestPad* pad) {
+    if (!pad) APS5_INVALID_ARG_EX;
+    RequireThrd(_Mtx_lock_nid_postfix(&pad->mutex), "_Pad::_Release: _Mtx_lock");
+    pad->started = true;
+    RequireThrd(_Cnd_signal_nid_postfix(&pad->condition), "_Pad::_Release: _Cnd_signal");
+    RequireThrd(_Mtx_unlock_nid_postfix(&pad->mutex), "_Pad::_Release: _Mtx_unlock");
 }
 
-int APS5_VABI _ZNSt4_PadC2Ev_nid_postfix(void) {
-    NotImplemented_nid_no_patch(__func__);
-    return 0;
+void APS5_VABI _ZNSt4_PadC2Ev_nid_postfix(GuestPad* pad) {
+    if (!pad) APS5_INVALID_ARG_EX;
+    RequireThrd(_Cnd_init_nid_postfix(&pad->condition), "_Pad::_Pad: _Cnd_init");
+    const int created = _Mtx_init_nid_postfix(&pad->mutex, MtxPlain);
+    if (created != ThrdSuccess) {
+        _Cnd_destroy_nid_postfix(&pad->condition);
+        RequireThrd(created, "_Pad::_Pad: _Mtx_init");
+    }
+    pad->started = false;
+    const int locked = _Mtx_lock_nid_postfix(&pad->mutex);
+    if (locked != ThrdSuccess) {
+        _Mtx_destroy_nid_postfix(&pad->mutex);
+        _Cnd_destroy_nid_postfix(&pad->condition);
+        RequireThrd(locked, "_Pad::_Pad: _Mtx_lock");
+    }
 }
 
-int APS5_VABI _ZNSt4_PadD2Ev_nid_postfix(void) {
-    NotImplemented_nid_no_patch(__func__);
-    return 0;
+void APS5_VABI _ZNSt4_PadD2Ev_nid_postfix(GuestPad* pad) {
+    if (!pad) APS5_INVALID_ARG_EX;
+    RequireThrd(_Mtx_unlock_nid_postfix(&pad->mutex), "_Pad::~_Pad: _Mtx_unlock");
+    _Mtx_destroy_nid_postfix(&pad->mutex);
+    _Cnd_destroy_nid_postfix(&pad->condition);
 }
 
 int APS5_VABI _Thrd_id_nid_postfix(void) {
@@ -885,9 +951,11 @@ int APS5_VABI _Thrd_id_nid_postfix(void) {
     return 0;
 }
 
-int APS5_VABI _ZNSt4_Pad7_LaunchEPP7pthread_nid_postfix(void) {
-    NotImplemented_nid_no_patch(__func__);
-    return 0;
+void APS5_VABI _ZNSt4_Pad7_LaunchEPP7pthread_nid_postfix(GuestPad* pad, Pthread* thread) {
+    if (!pad || !thread) APS5_INVALID_ARG_EX;
+    const int created = pthread_create_nid_postfix(thread, nullptr, PadEntry_nid_no_patch, pad);
+    if (created != 0) ThrowCppError_nid_no_patch(CppResourceUnavailableTryAgain);
+    while (!pad->started) RequireThrd(_Cnd_wait_nid_postfix(&pad->condition, &pad->mutex), "_Pad::_Launch: _Cnd_wait");
 }
 
 int APS5_VABI _ZNSt7num_getIcSt19istreambuf_iteratorIcSt11char_traitsIcEEE2idE_nid_postfix(void) {

core/libs/tests/GuestAmprThreads.cpp

@@ -10,13 +10,51 @@
 #include <cstring>
 #include <stdexcept>
 #include <thread>
+#ifndef _WIN32
+#include <sys/wait.h>
+#include <unistd.h>
+#endif
 
 struct GuestXtime {
     std::int64_t sec;
     std::int64_t nsec;
 };
 
+struct PadTarget {
+    std::atomic<int>* order;
+    std::atomic<bool>* destroyed;
+    int beforeRelease;
+    int afterRelease;
+    unsigned exitCode;
+};
+
+struct GuestTypeInfo {
+    const void* vtable;
+    const char* name;
+};
+
+struct GuestSystemError {
+    const void* vtable;
+    const char* message;
+    int code;
+    const void* category;
+};
+
+using GuestTerminate = void (APS5_VABI*)();
+
+struct TestPad {
+    unsigned (APS5_VABI* const* vtable)(TestPad*);
+    void* condition;
+    void* mutex;
+    bool started;
+    PadTarget target;
+};
+
 extern "C" {
+void APS5_VABI _ZNSt4_PadC2Ev_nid_postfix(TestPad* pad);
+void APS5_VABI _ZNSt4_PadD2Ev_nid_postfix(TestPad* pad);
+void APS5_VABI _ZNSt4_Pad7_LaunchEPP7pthread_nid_postfix(TestPad* pad, Pthread* thread);
+void APS5_VABI _ZNSt4_Pad8_ReleaseEv_nid_postfix(TestPad* pad);
 int APS5_VABI _Thrd_join_nid_postfix(Pthread thread, int* code);
 int APS5_VABI _Thrd_sleep_nid_postfix(const GuestXtime* target, GuestXtime* remaining);
 void APS5_VABI _Lock_shared_ptr_spin_lock_nid_postfix(void);
@@ -35,10 +73,14 @@ int APS5_VABI _Mtx_init_nid_postfix(void** handle, int type);
 void APS5_VABI _Mtx_destroy_nid_postfix(void** handle);
 int APS5_VABI _Mtx_lock_nid_postfix(void** handle);
 int APS5_VABI _Mtx_unlock_nid_postfix(void** handle);
+GuestTerminate APS5_VABI _ZSt13set_terminatePFvvE_nid_postfix(GuestTerminate handler);
+const GuestTypeInfo* APS5_VABI __cxa_current_exception_type_nid_postfix();
+void* APS5_VABI __cxa_current_primary_exception_nid_postfix();
 }
 
 static constexpr int ThrdSuccess = 0;
 static constexpr int ThrdError = 4;
+static constexpr int GuestEbusy = 16;
 static constexpr int GuestEinval = 22;
 static constexpr int GuestErange = 34;
 static constexpr int GuestEilseq = 86;
@@ -316,7 +358,181 @@ static void TestConditionSignal() {
     _Mtx_destroy_nid_postfix(&context.mutex);
 }
 
+static unsigned APS5_VABI PadGo(TestPad* pad) {
+    const auto local = pad->target;
+    std::this_thread::sleep_for(std::chrono::milliseconds(20));
+    local.order->store(local.beforeRelease);
+    _ZNSt4_Pad8_ReleaseEv_nid_postfix(pad);
+    while (!local.destroyed->load()) std::this_thread::yield();
+    local.order->store(local.afterRelease);
+    return local.exitCode;
+}
+
+static constexpr std::array<unsigned (APS5_VABI*)(TestPad*), 1> PadVtable{PadGo};
+
+template <typename TException, typename TCall>
+static bool Throws(TCall call) {
+    try {
+        call();
+    } catch (const TException&) {
+        return true;
+    }
+    return false;
+}
+
+static void ConstructPad(TestPad& pad, std::atomic<int>& order, std::atomic<bool>& destroyed, int index, unsigned exitCode) {
+    std::memset(&pad, 0xCD, sizeof(pad));
+    pad.vtable = PadVtable.data();
+    _ZNSt4_PadC2Ev_nid_postfix(&pad);
+    Require(pad.vtable == PadVtable.data());
+    Require(pad.condition && pad.mutex && !pad.started);
+    pad.target = {&order, &destroyed, 2 * index + 1, 2 * index + 2, exitCode};
+}
+
+static void TestPadLaunch() {
+    std::atomic<int> order{0};
+    std::atomic<bool> destroyed{false};
+    TestPad pad;
+    ConstructPad(pad, order, destroyed, 0, 0);
+    Pthread thread = nullptr;
+    _ZNSt4_Pad7_LaunchEPP7pthread_nid_postfix(&pad, &thread);
+    Require(thread != nullptr);
+    Require(pad.started);
+    Require(order.load() == 1);
+    _ZNSt4_PadD2Ev_nid_postfix(&pad);
+    Require(!pad.condition && !pad.mutex);
+    Require(order.load() == 1);
+    destroyed = true;
+    int code = -1;
+    Require(_Thrd_join_nid_postfix(thread, &code) == ThrdSuccess);
+    Require(code == 0);
+    Require(order.load() == 2);
+}
+
+static void TestPadLaunchMany() {
+    constexpr int Count = 4;
+    std::array<std::atomic<int>, Count> orders{};
+    std::atomic<bool> destroyed{false};
+    std::array<TestPad, Count> pads{};
+    std::array<Pthread, Count> threads{};
+    for (int index = 0; index < Count; ++index) {
+        ConstructPad(pads[index], orders[index], destroyed, index, 100u + index);
+        _ZNSt4_Pad7_LaunchEPP7pthread_nid_postfix(&pads[index], &threads[index]);
+        Require(orders[index].load() == 2 * index + 1);
+        _ZNSt4_PadD2Ev_nid_postfix(&pads[index]);
+    }
+    destroyed = true;
+    for (int index = 0; index < Count; ++index) {
+        int code = 0;
+        Require(_Thrd_join_nid_postfix(threads[index], &code) == ThrdSuccess);
+        Require(code == 100 + index);
+        Require(orders[index].load() == 2 * index + 2);
+    }
+}
+
+static unsigned APS5_VABI PadGoImmediate(TestPad* pad) {
+    const auto local = pad->target;
+    local.order->store(local.beforeRelease);
+    _ZNSt4_Pad8_ReleaseEv_nid_postfix(pad);
+    while (!local.destroyed->load()) std::this_thread::yield();
+    local.order->store(local.afterRelease);
+    return local.exitCode;
+}
+
+static void TestPadConcurrentLaunch() {
+    constexpr int Count = 4;
+    constexpr std::array<unsigned (APS5_VABI*)(TestPad*), 1> immediateVtable{PadGoImmediate};
+    std::atomic<int> ready{0};
+    std::array<std::thread, Count> parents;
+    for (int index = 0; index < Count; ++index) {
+        parents[index] = std::thread([&, index] {
+            std::atomic<int> order{0};
+            std::atomic<bool> destroyed{false};
+            TestPad pad;
+            ConstructPad(pad, order, destroyed, index, 0x80000000u + index);
+            pad.vtable = immediateVtable.data();
+            ready.fetch_add(1);
+            while (ready.load() != Count) std::this_thread::yield();
+            Pthread thread = nullptr;
+            _ZNSt4_Pad7_LaunchEPP7pthread_nid_postfix(&pad, &thread);
+            Require(thread != nullptr && pad.started && order.load() == 2 * index + 1);
+            _ZNSt4_PadD2Ev_nid_postfix(&pad);
+            std::memset(&pad, 0xcd, sizeof(pad));
+            destroyed = true;
+            int code = 0;
+            Require(_Thrd_join_nid_postfix(thread, &code) == ThrdSuccess);
+            Require(static_cast<unsigned>(code) == 0x80000000u + index);
+            Require(order.load() == 2 * index + 2);
+        });
+    }
+    for (auto& parent : parents) parent.join();
+}
+
+struct ForeignDestroy {
+    TestPad* pad;
+    bool rejected;
+};
+
+static void* APS5_VABI DestroyPad(void* arg) {
+    auto& context = *static_cast<ForeignDestroy*>(arg);
+    context.rejected = Throws<std::logic_error>([&] { _ZNSt4_PadD2Ev_nid_postfix(context.pad); });
+    return nullptr;
+}
+
+#ifndef _WIN32
+static void APS5_VABI ExitWithSystemErrorCode() {
+    const auto* type = __cxa_current_exception_type_nid_postfix();
+    const auto* error = static_cast<const GuestSystemError*>(__cxa_current_primary_exception_nid_postfix());
+    _exit(type && error && std::strcmp(type->name, "St12system_error") == 0 ? error->code : 1);
+}
+
+template <typename TCall>
+static int GuestSystemErrorCode(TCall call) {
+    const pid_t child = fork();
+    Require(child >= 0);
+    if (child == 0) {
+        _ZSt13set_terminatePFvvE_nid_postfix(ExitWithSystemErrorCode);
+        call();
+        _exit(2);
+    }
+    int status = 0;
+    Require(waitpid(child, &status, 0) == child);
+    return WIFEXITED(status) ? WEXITSTATUS(status) : -1;
+}
+#endif
+
+static void TestPadErrors() {
+    Require(Throws<std::invalid_argument>([] { _ZNSt4_PadC2Ev_nid_postfix(nullptr); }));
+    Require(Throws<std::invalid_argument>([] { _ZNSt4_PadD2Ev_nid_postfix(nullptr); }));
+    Require(Throws<std::invalid_argument>([] { _ZNSt4_Pad8_ReleaseEv_nid_postfix(nullptr); }));
+    Pthread thread = nullptr;
+    Require(Throws<std::invalid_argument>([&] { _ZNSt4_Pad7_LaunchEPP7pthread_nid_postfix(nullptr, &thread); }));
+
+    std::atomic<int> order{0};
+    std::atomic<bool> destroyed{true};
+    TestPad pad;
+    ConstructPad(pad, order, destroyed, 0, 0);
+    Require(Throws<std::invalid_argument>([&] { _ZNSt4_Pad7_LaunchEPP7pthread_nid_postfix(&pad, nullptr); }));
+    Require(!pad.started && order.load() == 0);
+#ifndef _WIN32
+    Require(GuestSystemErrorCode([&] { _ZNSt4_Pad8_ReleaseEv_nid_postfix(&pad); }) == GuestEbusy);
+#endif
+    Require(!pad.started);
+
+    ForeignDestroy context{&pad, false};
+    Require(_Thrd_join_nid_postfix(Start(DestroyPad, &context), nullptr) == ThrdSuccess);
+    Require(context.rejected);
+    Require(pad.condition && pad.mutex);
+    _ZNSt4_PadD2Ev_nid_postfix(&pad);
+    Require(!pad.condition && !pad.mutex);
+    Require(Throws<std::invalid_argument>([&] { _ZNSt4_PadD2Ev_nid_postfix(&pad); }));
+}
+
 int main() {
+    TestPadLaunch();
+    TestPadLaunchMany();
+    TestPadConcurrentLaunch();
+    TestPadErrors();
     TestThreadJoin();
     TestThreadSleep();
     TestSpinLock();

docs/dev/debt/libSceAmpr.md

@@ -13,3 +13,4 @@
 - [sceAmprAmmGetUsageStatsData](../../../core/libs/prx/libSceAmpr/Export.cpp) (libSceAmpr) fills the caller-sized 24-byte statistics structure with 512 page-table pool entries, no allocated entries, a zero high watermark and ring idle flags 0x7, the constants of [KytyPS5](https://github.com/KytyPS5/KytyPS5/blob/main/src/libs/libAmpr.cpp); they are not measured on a console, and the real call is a pass-through to `sceKernelMapperGetUsageStatsData` per [drakmor/ampr_emu](https://github.com/drakmor/ampr_emu/blob/main/src/ampr_emu_amm.cpp), which this port has no kernel-side counterpart for.
 - [Dinkumware thread exports](../../../core/libs/prx/libSceAmpr/Export.cpp) (libSceAmpr) follow the Dinkumware threads library shipped with [Visual C++ 2013](https://github.com/adityaramesh/concurrency_runtime/tree/master/VS2013_SP2/crt/src/thr) (`cthread.c`) and its `atomic.c`, [microsoft/STL `cthread.cpp`](https://github.com/microsoft/STL/blob/main/stl/src/cthread.cpp) and C11 Annex K, without console verification. `_Thrd_t` is taken to be the guest `pthread_t`, as the `std::_Pad::_Launch(pthread**)` export suggests: `_Thrd_join` returns `_Thrd_error` (4) when `pthread_join` fails and stores the thread's return value truncated to `int`. `_Thrd_join` (32-bit store of the result, 0 or 4) and the spin lock (byte exchange loop, no yield) match that disassembly. The shared_ptr spin lock is one process-wide flag that yields while it spins. `wcstombs_s` is `wcsrtombs_s` with a fresh conversion state, so `*retval` excludes the terminating null as in Annex K, not as in the Microsoft CRT, and `_TRUNCATE` has no meaning.
 - [_Cnd_signal, _Cnd_init_with_name](../../../core/libs/prx/libSceAmpr/Export.cpp) (libSceAmpr) - take a pointer to the handle slot like the libc `_Cnd_*` exports from [#3079](https://github.com/boykopovar/AnyPS5/pull/3079) and act on the same conditions. `_Cnd_signal` wakes one waiter and returns `_Thrd_success` (0), as `_Cnd_broadcast` does; whether the console maps a failing internal signal to `_Thrd_error` is not known. The signature `_Cnd_init_with_name(_Cnd_t*, const char* name)` is inferred from [fpPS4 `ps4_mtx_internal.pas`](https://github.com/red-prig/fpPS4/blob/trunk/src/libcinternal/ps4_mtx_internal.pas), where the PS4 `_Mtx_init` is `_Mtx_init_with_name` with the name `SceLibcI_Mtx` and the name only reaches the pthread object; here it is `_Cnd_init` and the name, which may be null, is ignored. Neither is confirmed on a PS5 `libc.prx`. `_Thrd_id` still throws: microsoft/STL returns an `unsigned int` thread id, but whether the PS5 returns the `pthread_t` or a kernel thread id, and at what width, is not known
+- [std::_Pad](../../../core/libs/prx/libSceAmpr/Export.cpp) (libSceAmpr) follows Dinkumware V6.00 `thr/xthread` and `thr/threadcall.cpp` of Visual C++ 2013 ([crt.vc12 `f1d47126`](https://github.com/ganboing/crt.vc12/tree/f1d47126/src/thr)), without console verification. The pad is taken to be that header's class with the console's pointer handles: a vtable pointer whose first slot is the pure virtual `_Go()`, then `_Cnd_t` at 0x8, `_Mtx_t` at 0x10 and `bool _Started` at 0x18, 32 bytes, the derived `_LaunchPad` target following it; the console's `_Pad(const char*)` and `_Launch(const char*, pthread**)`/`_Launch(pthread_attr**, pthread**)` overloads suggest named primitives and could mean another layout or vtable. `_Launch` starts the thread with `pthread_create` and default attributes (the console's `_Thrd_start` name, stack size and priority are unknown), its entry calls `_Go()` and returns the result as the exit code, and it does not run `_Cnd_do_broadcast_at_thread_exit`, which is absent here. A failing primitive throws the guest `std::system_error` that `_Throw_C_error` picks for its code, and a failing thread start the `resource_unavailable_try_again` one of `_Thrd_startX`; any `pthread_create` error counts, since the console's `_Thrd_start` result is unknown. A null pad or thread slot throws