core: Many things (#194)

* video_core: Add a few missed things

* libkernel: More proper memory mapped files

* memory: Fix tessellation buffer mapping

* Cuphead work

* sceKernelPollSema fix

* clang format

* fixed ngs2 lle loading and rtc lib

* draft pthreads keys implementation

* fixed return codes

* return error code if sceKernelLoadStartModule module is invalid

* re-enabled system modules and disable debug in libs.h

* Improve linux support

* fix windows build

* kernel: Rework keys

---------

Co-authored-by: georgemoralis <giorgosmrls@gmail.com>

CMakeLists.txt

@@ -105,8 +105,10 @@ set(KERNEL_LIB
                src/core/libraries/kernel/event_flag/event_flag.h
                src/core/libraries/kernel/event_flag/event_flag_obj.cpp
                src/core/libraries/kernel/event_flag/event_flag_obj.h
-               src/core/libraries/kernel/threads/kernel_threads.h
-               src/core/libraries/kernel/threads/kernel_threads_rwlock.cpp
+               src/core/libraries/kernel/threads/rwlock.cpp
+               src/core/libraries/kernel/threads/semaphore.cpp
+               src/core/libraries/kernel/threads/keys.cpp
+               src/core/libraries/kernel/threads/threads.h
                src/core/libraries/kernel/cpu_management.cpp
                src/core/libraries/kernel/cpu_management.h
                src/core/libraries/kernel/event_queue.cpp
@@ -226,8 +228,6 @@ set(COMMON src/common/logging/backend.cpp
            src/common/debug.h
            src/common/disassembler.cpp
            src/common/disassembler.h
-           src/common/discord.cpp
-           src/common/discord.h
            src/common/endian.h
            src/common/enum.h
            src/common/io_file.cpp
@@ -278,6 +278,8 @@ set(CORE src/core/aerolib/stubs.cpp
          src/core/file_sys/fs.h  
          src/core/loader.cpp
          src/core/loader.h
+         src/core/loader/dwarf.cpp
+         src/core/loader/dwarf.h
          src/core/loader/elf.cpp
          src/core/loader/elf.h
          src/core/loader/symbols_resolver.h
@@ -500,13 +502,17 @@ else()
         ${VIDEO_CORE}
         ${EMULATOR}
         src/main.cpp
+        src/emulator.cpp
+        src/emulator.h
+        src/sdl_window.h
+        src/sdl_window.cpp
     )
 endif()
 
 create_target_directory_groups(shadps4)
 
 target_link_libraries(shadps4 PRIVATE magic_enum::magic_enum fmt::fmt toml11::toml11 tsl::robin_map xbyak Tracy::TracyClient)
-target_link_libraries(shadps4 PRIVATE discord-rpc boost vma sirit vulkan-headers xxhash Zydis SPIRV glslang SDL3-shared)
+target_link_libraries(shadps4 PRIVATE boost vma sirit vulkan-headers xxhash Zydis SPIRV glslang SDL3-shared)
 
 if (NOT ENABLE_QT_GUI)
   target_link_libraries(shadps4 PRIVATE SDL3-shared)

src/common/io_file.cpp

@@ -3,6 +3,9 @@
 
 #include <vector>
 
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/error.h"
 #include "common/io_file.h"
 #include "common/logging/log.h"
 #include "common/path_util.h"
@@ -10,6 +13,7 @@
 #ifdef _WIN32
 #include <io.h>
 #include <share.h>
+#include <windows.h>
 #else
 #include <unistd.h>
 #endif
@@ -204,6 +208,30 @@ void IOFile::Close() {
     }
 
     file = nullptr;
+
+#ifdef _WIN64
+    if (file_mapping) {
+        CloseHandle(std::bit_cast<HANDLE>(file_mapping));
+    }
+#endif
+}
+
+uintptr_t IOFile::GetFileMapping() {
+    if (file_mapping) {
+        return file_mapping;
+    }
+#ifdef _WIN64
+    const int fd = fileno(file);
+    HANDLE hfile = reinterpret_cast<HANDLE>(_get_osfhandle(fd));
+    HANDLE mapping =
+        CreateFileMapping2(hfile, NULL, FILE_MAP_READ, PAGE_READONLY, SEC_COMMIT, 0, NULL, NULL, 0);
+    file_mapping = std::bit_cast<uintptr_t>(mapping);
+    ASSERT_MSG(file_mapping, "{}", Common::GetLastErrorMsg());
+    return file_mapping;
+#else
+    file_mapping = fileno(file);
+    return file_mapping;
+#endif
 }
 
 std::string IOFile::ReadString(size_t length) const {

src/common/io_file.h

@@ -100,6 +100,8 @@ public:
         return file != nullptr;
     }
 
+    uintptr_t GetFileMapping();
+
     void Open(const std::filesystem::path& path, FileAccessMode mode,
               FileType type = FileType::BinaryFile,
               FileShareFlag flag = FileShareFlag::ShareReadOnly);
@@ -212,6 +214,7 @@ private:
     FileType file_type{};
 
     std::FILE* file = nullptr;
+    uintptr_t file_mapping = 0;
 };
 
 } // namespace Common::FS

src/core/address_space.cpp

@@ -85,7 +85,7 @@ struct AddressSpace::Impl {
         }
     }
 
-    void* Map(VAddr virtual_addr, PAddr phys_addr, size_t size, ULONG prot) {
+    void* Map(VAddr virtual_addr, PAddr phys_addr, size_t size, ULONG prot, uintptr_t fd = 0) {
         const auto it = placeholders.find(virtual_addr);
         ASSERT_MSG(it != placeholders.end(), "Cannot map already mapped region");
         ASSERT_MSG(virtual_addr >= it->lower() && virtual_addr + size <= it->upper(),
@@ -117,8 +117,9 @@ struct AddressSpace::Impl {
         // Perform the map.
         void* ptr = nullptr;
         if (phys_addr != -1) {
-            ptr = MapViewOfFile3(backing_handle, process, reinterpret_cast<PVOID>(virtual_addr),
-                                 phys_addr, size, MEM_REPLACE_PLACEHOLDER, prot, nullptr, 0);
+            HANDLE backing = fd ? reinterpret_cast<HANDLE>(fd) : backing_handle;
+            ptr = MapViewOfFile3(backing, process, reinterpret_cast<PVOID>(virtual_addr), phys_addr,
+                                 size, MEM_REPLACE_PLACEHOLDER, prot, nullptr, 0);
         } else {
             ptr =
                 VirtualAlloc2(process, reinterpret_cast<PVOID>(virtual_addr), size,
@@ -128,9 +129,9 @@ struct AddressSpace::Impl {
         return ptr;
     }
 
-    void Unmap(VAddr virtual_addr, PAddr phys_addr, size_t size) {
+    void Unmap(VAddr virtual_addr, size_t size, bool has_backing) {
         bool ret;
-        if (phys_addr != -1) {
+        if (has_backing) {
             ret = UnmapViewOfFile2(process, reinterpret_cast<PVOID>(virtual_addr),
                                    MEM_PRESERVE_PLACEHOLDER);
         } else {
@@ -254,13 +255,14 @@ struct AddressSpace::Impl {
         m_free_regions.insert({start_addr, start_addr + virtual_size});
     }
 
-    void* Map(VAddr virtual_addr, PAddr phys_addr, size_t size, PosixPageProtection prot) {
+    void* Map(VAddr virtual_addr, PAddr phys_addr, size_t size, PosixPageProtection prot,
+              int fd = -1) {
         m_free_regions.subtract({virtual_addr, virtual_addr + size});
-        const int fd = phys_addr != -1 ? backing_fd : -1;
-        const int host_offset = phys_addr != -1 ? phys_addr : 0;
+        const int handle = phys_addr != -1 ? (fd == -1 ? backing_fd : fd) : -1;
+        const off_t host_offset = phys_addr != -1 ? phys_addr : 0;
         const int flag = phys_addr != -1 ? MAP_SHARED : (MAP_ANONYMOUS | MAP_PRIVATE);
-        void* ret = mmap(reinterpret_cast<void*>(virtual_addr), size, prot, MAP_FIXED | flag, fd,
-                         host_offset);
+        void* ret = mmap(reinterpret_cast<void*>(virtual_addr), size, prot, MAP_FIXED | flag,
+                         handle, host_offset);
         ASSERT_MSG(ret != MAP_FAILED, "mmap failed: {}", strerror(errno));
         return ret;
     }
@@ -323,8 +325,12 @@ void* AddressSpace::Map(VAddr virtual_addr, size_t size, u64 alignment, PAddr ph
                      is_exec ? PAGE_EXECUTE_READWRITE : PAGE_READWRITE);
 }
 
-void AddressSpace::Unmap(VAddr virtual_addr, size_t size, PAddr phys_addr) {
-    return impl->Unmap(virtual_addr, phys_addr, size);
+void* AddressSpace::MapFile(VAddr virtual_addr, size_t size, size_t offset, uintptr_t fd) {
+    return impl->Map(virtual_addr, offset, size, fd ? PAGE_READONLY : PAGE_READWRITE, fd);
+}
+
+void AddressSpace::Unmap(VAddr virtual_addr, size_t size, bool has_backing) {
+    return impl->Unmap(virtual_addr, size, has_backing);
 }
 
 void AddressSpace::Protect(VAddr virtual_addr, size_t size, MemoryPermission perms) {

src/core/address_space.h

@@ -20,7 +20,7 @@ DECLARE_ENUM_FLAG_OPERATORS(MemoryPermission)
 
 constexpr VAddr SYSTEM_RESERVED = 0x800000000ULL;
 constexpr VAddr CODE_BASE_OFFSET = 0x100000000ULL;
-constexpr VAddr SYSTEM_MANAGED_MIN = 0x0000040000ULL;
+constexpr VAddr SYSTEM_MANAGED_MIN = 0x00000400000ULL;
 constexpr VAddr SYSTEM_MANAGED_MAX = 0x07FFFFBFFFULL;
 constexpr VAddr USER_MIN = 0x1000000000ULL;
 constexpr VAddr USER_MAX = 0xFBFFFFFFFFULL;
@@ -62,8 +62,11 @@ public:
     void* Map(VAddr virtual_addr, size_t size, u64 alignment = 0, PAddr phys_addr = -1,
               bool exec = false);
 
+    /// Memory maps a specified file descriptor.
+    void* MapFile(VAddr virtual_addr, size_t size, size_t offset, uintptr_t fd);
+
     /// Unmaps specified virtual memory area.
-    void Unmap(VAddr virtual_addr, size_t size, PAddr phys_addr);
+    void Unmap(VAddr virtual_addr, size_t size, bool has_backing);
 
     void Protect(VAddr virtual_addr, size_t size, MemoryPermission perms);
 

src/core/aerolib/stubs.cpp

@@ -19,7 +19,7 @@ namespace Core::AeroLib {
 // and to longer compile / CI times
 //
 // Must match STUBS_LIST define
-constexpr u32 MAX_STUBS = 512;
+constexpr u32 MAX_STUBS = 1024;
 
 u64 UnresolvedStub() {
     LOG_ERROR(Core, "Returning zero to {}", __builtin_return_address(0));
@@ -60,8 +60,9 @@ static u32 UsedStubEntries;
 #define XREP_128(x) XREP_64(x) XREP_64(x + 64)
 #define XREP_256(x) XREP_128(x) XREP_128(x + 128)
 #define XREP_512(x) XREP_256(x) XREP_256(x + 256)
+#define XREP_1024(x) XREP_512(x) XREP_512(x + 512)
 
-#define STUBS_LIST XREP_512(0)
+#define STUBS_LIST XREP_1024(0)
 
 static u64 (*stub_handlers[MAX_STUBS])() = {STUBS_LIST};
 

src/core/file_sys/fs.cpp

@@ -97,4 +97,5 @@ File* HandleTable::getFile(const std::string& host_name) {
     }
     return nullptr;
 }
+
 } // namespace Core::FileSys

src/core/libraries/gnmdriver/gnmdriver.cpp

@@ -66,6 +66,9 @@ static inline u32* WriteTrailingNop(u32* cmdbuf) {
 
 s32 PS4_SYSV_ABI sceGnmAddEqEvent(SceKernelEqueue eq, u64 id, void* udata) {
     LOG_TRACE(Lib_GnmDriver, "called");
+    if (id != SceKernelEvent::Type::GfxEop) {
+        return ORBIS_OK;
+    }
     ASSERT_MSG(id == SceKernelEvent::Type::GfxEop);
 
     if (!eq) {
@@ -986,7 +989,7 @@ s32 PS4_SYSV_ABI sceGnmSetEmbeddedVsShader(u32* cmdbuf, u32 size, u32 shader_id,
     // a check for zero in the upper part of shader address. In our case, the address is a
     // pointer to a stack memory, so the check will likely fail. To workaround it we will
     // repeat set shader functionality here as it is trivial.
-    cmdbuf = PM4CmdSetData::SetShReg(cmdbuf, 0x48u, vs_regs[0], 0u); // SPI_SHADER_PGM_LO_VS
+    cmdbuf = PM4CmdSetData::SetShReg(cmdbuf, 0x48u, vs_regs[0], vs_regs[1]); // SPI_SHADER_PGM_LO_VS
     cmdbuf =
         PM4CmdSetData::SetShReg(cmdbuf, 0x4au, vs_regs[2], vs_regs[3]); // SPI_SHADER_PGM_RSRC1_VS
     cmdbuf = PM4CmdSetData::SetContextReg(cmdbuf, 0x207u, vs_regs[6]);  // PA_CL_VS_OUT_CNTL

src/core/libraries/kernel/event_flag/event_flag.cpp

@@ -73,7 +73,8 @@ int PS4_SYSV_ABI sceKernelCloseEventFlag() {
     return ORBIS_OK;
 }
 int PS4_SYSV_ABI sceKernelClearEventFlag(OrbisKernelEventFlag ef, u64 bitPattern) {
-    LOG_ERROR(Kernel_Event, "(STUBBED) called");
+    LOG_ERROR(Kernel_Event, "called");
+    ef->Clear(bitPattern);
     return ORBIS_OK;
 }
 int PS4_SYSV_ABI sceKernelCancelEventFlag(OrbisKernelEventFlag ef, u64 setPattern,

src/core/libraries/kernel/event_flag/event_flag_obj.cpp

@@ -90,4 +90,15 @@ void EventFlagInternal::Set(u64 bits) {
     m_cond_var.notify_all();
 }
 
+void EventFlagInternal::Clear(u64 bits) {
+    std::unique_lock lock{m_mutex};
+    while (m_status != Status::Set) {
+        m_mutex.unlock();
+        std::this_thread::sleep_for(std::chrono::microseconds(10));
+        m_mutex.lock();
+    }
+
+    m_bits &= bits;
+}
+
 } // namespace Libraries::Kernel

src/core/libraries/kernel/event_flag/event_flag_obj.h

@@ -25,6 +25,7 @@ public:
     int Wait(u64 bits, WaitMode wait_mode, ClearMode clear_mode, u64* result, u32* ptr_micros);
     int Poll(u64 bits, WaitMode wait_mode, ClearMode clear_mode, u64* result);
     void Set(u64 bits);
+    void Clear(u64 bits);
 
 private:
     enum class Status { Set, Canceled, Deleted };

src/core/libraries/kernel/event_queue.h

@@ -34,6 +34,13 @@ constexpr s16 EVFILT_GPU_SYSTEM_EXCEPTION = -21;
 constexpr s16 EVFILT_GPU_DBGGC_EV = -22;
 constexpr s16 EVFILT_SYSCOUNT = 22;
 
+constexpr u16 EV_ONESHOT = 0x10;    // only report one occurrence
+constexpr u16 EV_CLEAR = 0x20;      // clear event state after reporting
+constexpr u16 EV_RECEIPT = 0x40;    // force EV_ERROR on success, data=0
+constexpr u16 EV_DISPATCH = 0x80;   // disable event after reporting
+constexpr u16 EV_SYSFLAGS = 0xF000; // reserved by system
+constexpr u16 EV_FLAG1 = 0x2000;    // filter-specific flag
+
 class EqueueInternal;
 struct EqueueEvent;
 

src/core/libraries/kernel/event_queues.cpp

@@ -88,6 +88,24 @@ int PS4_SYSV_ABI sceKernelAddUserEvent(SceKernelEqueue eq, int id) {
     event.event.ident = id;
     event.event.filter = Kernel::EVFILT_USER;
     event.event.udata = 0;
+    event.event.flags = 1;
+    event.event.fflags = 0;
+    event.event.data = 0;
+
+    return eq->addEvent(event);
+}
+
+int PS4_SYSV_ABI sceKernelAddUserEventEdge(SceKernelEqueue eq, int id) {
+    if (eq == nullptr) {
+        return ORBIS_KERNEL_ERROR_EBADF;
+    }
+
+    Kernel::EqueueEvent event{};
+    event.isTriggered = false;
+    event.event.ident = id;
+    event.event.filter = Kernel::EVFILT_USER;
+    event.event.udata = 0;
+    event.event.flags = 0x21;
     event.event.fflags = 0;
     event.event.data = 0;
 
@@ -111,4 +129,5 @@ int PS4_SYSV_ABI sceKernelDeleteUserEvent(SceKernelEqueue eq, int id) {
     eq->removeEvent(id);
     return ORBIS_OK;
 }
+
 } // namespace Libraries::Kernel

src/core/libraries/kernel/event_queues.h

@@ -18,5 +18,6 @@ void* PS4_SYSV_ABI sceKernelGetEventUserData(const SceKernelEvent* ev);
 int PS4_SYSV_ABI sceKernelTriggerUserEvent(SceKernelEqueue eq, int id, void* udata);
 int PS4_SYSV_ABI sceKernelDeleteUserEvent(SceKernelEqueue eq, int id);
 int PS4_SYSV_ABI sceKernelAddUserEvent(SceKernelEqueue eq, int id);
+int PS4_SYSV_ABI sceKernelAddUserEventEdge(SceKernelEqueue eq, int id);
 
 } // namespace Libraries::Kernel

src/core/libraries/kernel/file_system.cpp

@@ -31,6 +31,10 @@ int PS4_SYSV_ABI sceKernelOpen(const char* path, int flags, u16 mode) {
     bool direct = (flags & ORBIS_KERNEL_O_DIRECT) != 0;
     bool directory = (flags & ORBIS_KERNEL_O_DIRECTORY) != 0;
 
+    if (std::string_view{path} == "/dev/console" || std::string_view{path} == "/dev/deci_tty6") {
+        return ORBIS_OK;
+    }
+
     if (directory) {
         LOG_ERROR(Kernel_Fs, "called on directory");
     } else {
@@ -89,6 +93,11 @@ int PS4_SYSV_ABI sceKernelClose(int d) {
     return SCE_OK;
 }
 
+int PS4_SYSV_ABI posix_close(int d) {
+    ASSERT(sceKernelClose(d) == 0);
+    return ORBIS_OK;
+}
+
 size_t PS4_SYSV_ABI sceKernelWrite(int d, void* buf, size_t nbytes) {
     if (buf == nullptr) {
         return SCE_KERNEL_ERROR_EFAULT;
@@ -282,10 +291,22 @@ int PS4_SYSV_ABI sceKernelFStat(int fd, OrbisKernelStat* sb) {
     return ORBIS_OK;
 }
 
+int PS4_SYSV_ABI posix_fstat(int fd, OrbisKernelStat* sb) {
+    return sceKernelFStat(fd, sb);
+}
+
+s32 PS4_SYSV_ABI sceKernelFsync(int fd) {
+    auto* h = Common::Singleton<Core::FileSys::HandleTable>::Instance();
+    auto* file = h->GetFile(fd);
+    file->f.Flush();
+    return ORBIS_OK;
+}
+
 void fileSystemSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("1G3lF1Gg1k8", "libkernel", 1, "libkernel", 1, 1, sceKernelOpen);
     LIB_FUNCTION("wuCroIGjt2g", "libScePosix", 1, "libkernel", 1, 1, posix_open);
     LIB_FUNCTION("UK2Tl2DWUns", "libkernel", 1, "libkernel", 1, 1, sceKernelClose);
+    LIB_FUNCTION("bY-PO6JhzhQ", "libScePosix", 1, "libkernel", 1, 1, posix_close);
     LIB_FUNCTION("4wSze92BhLI", "libkernel", 1, "libkernel", 1, 1, sceKernelWrite);
 
     LIB_FUNCTION("+WRlkKjZvag", "libkernel", 1, "libkernel", 1, 1, _readv);
@@ -295,10 +316,12 @@ void fileSystemSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("1-LFLmRFxxM", "libkernel", 1, "libkernel", 1, 1, sceKernelMkdir);
     LIB_FUNCTION("eV9wAD2riIA", "libkernel", 1, "libkernel", 1, 1, sceKernelStat);
     LIB_FUNCTION("kBwCPsYX-m4", "libkernel", 1, "libkernel", 1, 1, sceKernelFStat);
+    LIB_FUNCTION("mqQMh1zPPT8", "libScePosix", 1, "libkernel", 1, 1, posix_fstat);
 
     LIB_FUNCTION("E6ao34wPw+U", "libScePosix", 1, "libkernel", 1, 1, posix_stat);
     LIB_FUNCTION("+r3rMFwItV4", "libkernel", 1, "libkernel", 1, 1, sceKernelPread);
     LIB_FUNCTION("uWyW3v98sU4", "libkernel", 1, "libkernel", 1, 1, sceKernelCheckReachability);
+    LIB_FUNCTION("fTx66l5iWIA", "libkernel", 1, "libkernel", 1, 1, sceKernelFsync);
 
     // openOrbis (to check if it is valid out of OpenOrbis
     LIB_FUNCTION("6c3rCVE-fTU", "libkernel", 1, "libkernel", 1, 1,

src/core/libraries/kernel/libkernel.cpp

@@ -62,62 +62,35 @@ size_t PS4_SYSV_ABI _writev(int fd, const struct iovec* iov, int iovcn) {
     return total_written;
 }
 
-static thread_local int libc_error;
+static thread_local int libc_error{};
 int* PS4_SYSV_ABI __Error() {
     return &libc_error;
 }
 
-#define PROT_READ 0x1
-#define PROT_WRITE 0x2
-
-int PS4_SYSV_ABI sceKernelMmap(void* addr, u64 len, int prot, int flags, int fd, off_t offset,
+int PS4_SYSV_ABI sceKernelMmap(void* addr, u64 len, int prot, int flags, int fd, size_t offset,
                                void** res) {
-#ifdef _WIN64
-    LOG_INFO(Kernel_Vmm, "called");
-    if (prot > 3) {
-        LOG_ERROR(Kernel_Vmm, "prot = {} not supported", prot);
+    LOG_INFO(Kernel_Vmm, "called addr = {}, len = {}, prot = {}, flags = {}, fd = {}, offset = {}",
+             fmt::ptr(addr), len, prot, flags, fd, offset);
+    auto* h = Common::Singleton<Core::FileSys::HandleTable>::Instance();
+    auto* memory = Core::Memory::Instance();
+    const auto mem_prot = static_cast<Core::MemoryProt>(prot);
+    const auto mem_flags = static_cast<Core::MemoryMapFlags>(flags);
+    if (fd == -1) {
+        return memory->MapMemory(res, std::bit_cast<VAddr>(addr), len, mem_prot, mem_flags,
+                                 Core::VMAType::Flexible);
+    } else {
+        const uintptr_t handle = h->GetFile(fd)->f.GetFileMapping();
+        return memory->MapFile(res, std::bit_cast<VAddr>(addr), len, mem_prot, mem_flags, handle,
+                               offset);
     }
-    DWORD flProtect;
-    if (prot & PROT_WRITE) {
-        flProtect = PAGE_READWRITE;
-    }
-    off_t end = len + offset;
-    HANDLE mmap_fd, h;
-    if (fd == -1)
-        mmap_fd = INVALID_HANDLE_VALUE;
-    else
-        mmap_fd = (HANDLE)_get_osfhandle(fd);
-    h = CreateFileMapping(mmap_fd, NULL, flProtect, 0, end, NULL);
-    int k = GetLastError();
-    if (NULL == h)
-        return -1;
-    DWORD dwDesiredAccess;
-    if (prot & PROT_WRITE)
-        dwDesiredAccess = FILE_MAP_WRITE;
-    else
-        dwDesiredAccess = FILE_MAP_READ;
-    void* ret = MapViewOfFile(h, dwDesiredAccess, 0, offset, len);
-    if (ret == NULL) {
-        CloseHandle(h);
-        ret = nullptr;
-    }
-    *res = ret;
-    return 0;
-#else
-    void* result = mmap(addr, len, prot, flags, fd, offset);
-    if (result != MAP_FAILED) {
-        *res = result;
-        return 0;
-    }
-    std::abort();
-#endif
 }
 
-PS4_SYSV_ABI void* posix_mmap(void* addr, u64 len, int prot, int flags, int fd, u64 offset) {
+void* PS4_SYSV_ABI posix_mmap(void* addr, u64 len, int prot, int flags, int fd, u64 offset) {
     void* ptr;
     LOG_INFO(Kernel_Vmm, "posix mmap redirect to sceKernelMmap\n");
     // posix call the difference is that there is a different behaviour when it doesn't return 0 or
     // SCE_OK
+    const VAddr ret_addr = (VAddr)__builtin_return_address(0);
     int result = sceKernelMmap(addr, len, prot, flags, fd, offset, &ptr);
     ASSERT(result == 0);
     return ptr;
@@ -201,11 +174,19 @@ s32 PS4_SYSV_ABI sceKernelLoadStartModule(const char* moduleFileName, size_t arg
     auto* mnt = Common::Singleton<Core::FileSys::MntPoints>::Instance();
     const auto path = mnt->GetHostFile(moduleFileName);
 
-    // Load PRX module.
+    // Load PRX module and relocate any modules that import it.
     auto* linker = Common::Singleton<Core::Linker>::Instance();
     u32 handle = linker->LoadModule(path);
+    if (handle == -1) {
+        return ORBIS_KERNEL_ERROR_EINVAL;
+    }
     auto* module = linker->GetModule(handle);
-    linker->Relocate(module);
+    linker->RelocateAnyImports(module);
+
+    // If the new module has a TLS image, trigger its load when TlsGetAddr is called.
+    if (module->tls.image_size != 0) {
+        linker->AdvanceGenerationCounter();
+    }
 
     // Retrieve and verify proc param according to libkernel.
     u64* param = module->GetProcParam<u64*>();
@@ -225,10 +206,84 @@ s32 PS4_SYSV_ABI sceKernelDlsym(s32 handle, const char* symbol, void** addrp) {
     return ORBIS_OK;
 }
 
+static constexpr size_t ORBIS_DBG_MAX_NAME_LENGTH = 256;
+
+struct OrbisModuleInfoForUnwind {
+    u64 st_size;
+    std::array<char, ORBIS_DBG_MAX_NAME_LENGTH> name;
+    VAddr eh_frame_hdr_addr;
+    VAddr eh_frame_addr;
+    u64 eh_frame_size;
+    VAddr seg0_addr;
+    u64 seg0_size;
+};
+
+s32 PS4_SYSV_ABI sceKernelGetModuleInfoForUnwind(VAddr addr, int flags,
+                                                 OrbisModuleInfoForUnwind* info) {
+    if (flags >= 3) {
+        std::memset(info, 0, sizeof(OrbisModuleInfoForUnwind));
+        return SCE_KERNEL_ERROR_EINVAL;
+    }
+    if (!info) {
+        return ORBIS_KERNEL_ERROR_EFAULT;
+    }
+    if (info->st_size <= sizeof(OrbisModuleInfoForUnwind)) {
+        return ORBIS_KERNEL_ERROR_EINVAL;
+    }
+
+    // Find module that contains specified address.
+    LOG_INFO(Lib_Kernel, "called addr = {:#x}, flags = {:#x}", addr, flags);
+    auto* linker = Common::Singleton<Core::Linker>::Instance();
+    auto* module = linker->FindByAddress(addr);
+    const auto mod_info = module->GetModuleInfoEx();
+
+    // Fill in module info.
+    info->name = mod_info.name;
+    info->eh_frame_hdr_addr = mod_info.eh_frame_hdr_addr;
+    info->eh_frame_addr = mod_info.eh_frame_addr;
+    info->eh_frame_size = mod_info.eh_frame_size;
+    info->seg0_addr = mod_info.segments[0].address;
+    info->seg0_size = mod_info.segments[0].size;
+    return ORBIS_OK;
+}
+
+int PS4_SYSV_ABI sceKernelGetModuleInfoFromAddr(VAddr addr, int flags,
+                                                Core::OrbisKernelModuleInfoEx* info) {
+    LOG_INFO(Lib_Kernel, "called addr = {:#x}, flags = {:#x}", addr, flags);
+    auto* linker = Common::Singleton<Core::Linker>::Instance();
+    auto* module = linker->FindByAddress(addr);
+    *info = module->GetModuleInfoEx();
+    return ORBIS_OK;
+}
+
+int PS4_SYSV_ABI sceKernelDebugRaiseException() {
+    UNREACHABLE();
+    return 0;
+}
+
+char PS4_SYSV_ABI _is_signal_return(s64* param_1) {
+    char cVar1;
+
+    if (((*param_1 != 0x48006a40247c8d48ULL) || (param_1[1] != 0x50f000001a1c0c7ULL)) ||
+        (cVar1 = '\x01', (param_1[2] & 0xffffffU) != 0xfdebf4)) {
+        cVar1 = ((*(u64*)((s64)param_1 + -5) & 0xffffffffff) == 0x50fca8949) * '\x02';
+    }
+    return cVar1;
+}
+
+int PS4_SYSV_ABI sceKernelGetCpumode() {
+    return 5;
+}
+
+void PS4_SYSV_ABI sched_yield() {
+    return std::this_thread::yield();
+}
+
 void LibKernel_Register(Core::Loader::SymbolsResolver* sym) {
     // obj
     LIB_OBJ("f7uOxY9mM1U", "libkernel", 1, "libkernel", 1, 1, &g_stack_chk_guard);
     // memory
+    LIB_FUNCTION("OMDRKKAZ8I4", "libkernel", 1, "libkernel", 1, 1, sceKernelDebugRaiseException);
     LIB_FUNCTION("rTXw65xmLIA", "libkernel", 1, "libkernel", 1, 1, sceKernelAllocateDirectMemory);
     LIB_FUNCTION("B+vc2AO2Zrc", "libkernel", 1, "libkernel", 1, 1,
                  sceKernelAllocateMainDirectMemory);
@@ -248,6 +303,9 @@ void LibKernel_Register(Core::Loader::SymbolsResolver* sym) {
                  _sceKernelRtldSetApplicationHeapAPI);
     LIB_FUNCTION("wzvqT4UqKX8", "libkernel", 1, "libkernel", 1, 1, sceKernelLoadStartModule);
     LIB_FUNCTION("LwG8g3niqwA", "libkernel", 1, "libkernel", 1, 1, sceKernelDlsym);
+    LIB_FUNCTION("RpQJJVKTiFM", "libkernel", 1, "libkernel", 1, 1, sceKernelGetModuleInfoForUnwind);
+    LIB_FUNCTION("f7KBOafysXo", "libkernel", 1, "libkernel", 1, 1, sceKernelGetModuleInfoFromAddr);
+    LIB_FUNCTION("VOx8NGmHXTs", "libkernel", 1, "libkernel", 1, 1, sceKernelGetCpumode);
 
     // equeue
     LIB_FUNCTION("D0OdFMjp46I", "libkernel", 1, "libkernel", 1, 1, sceKernelCreateEqueue);
@@ -255,6 +313,7 @@ void LibKernel_Register(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("fzyMKs9kim0", "libkernel", 1, "libkernel", 1, 1, sceKernelWaitEqueue);
     LIB_FUNCTION("vz+pg2zdopI", "libkernel", 1, "libkernel", 1, 1, sceKernelGetEventUserData);
     LIB_FUNCTION("4R6-OvI2cEA", "libkernel", 1, "libkernel", 1, 1, sceKernelAddUserEvent);
+    LIB_FUNCTION("WDszmSbWuDk", "libkernel", 1, "libkernel", 1, 1, sceKernelAddUserEventEdge);
     LIB_FUNCTION("F6e0kwo4cnk", "libkernel", 1, "libkernel", 1, 1, sceKernelTriggerUserEvent);
     LIB_FUNCTION("LJDwdSNTnDg", "libkernel", 1, "libkernel", 1, 1, sceKernelDeleteUserEvent);
 
@@ -263,11 +322,13 @@ void LibKernel_Register(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("Ou3iL1abvng", "libkernel", 1, "libkernel", 1, 1, stack_chk_fail);
     LIB_FUNCTION("9BcDykPmo1I", "libkernel", 1, "libkernel", 1, 1, __Error);
     LIB_FUNCTION("BPE9s9vQQXo", "libkernel", 1, "libkernel", 1, 1, posix_mmap);
+    LIB_FUNCTION("BPE9s9vQQXo", "libScePosix", 1, "libkernel", 1, 1, posix_mmap);
     LIB_FUNCTION("YSHRBRLn2pI", "libkernel", 1, "libkernel", 1, 1, _writev);
     LIB_FUNCTION("959qrazPIrg", "libkernel", 1, "libkernel", 1, 1, sceKernelGetProcParam);
     LIB_FUNCTION("-o5uEDpN+oY", "libkernel", 1, "libkernel", 1, 1, sceKernelConvertUtcToLocaltime);
     LIB_FUNCTION("WB66evu8bsU", "libkernel", 1, "libkernel", 1, 1, sceKernelGetCompiledSdkVersion);
     LIB_FUNCTION("DRuBt2pvICk", "libkernel", 1, "libkernel", 1, 1, ps4__read);
+    LIB_FUNCTION("crb5j7mkk1c", "libkernel", 1, "libkernel", 1, 1, _is_signal_return);
 
     Libraries::Kernel::fileSystemSymbolsRegister(sym);
     Libraries::Kernel::timeSymbolsRegister(sym);
@@ -278,6 +339,7 @@ void LibKernel_Register(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("NWtTN10cJzE", "libSceLibcInternalExt", 1, "libSceLibcInternal", 1, 1,
                  sceLibcHeapGetTraceInfo);
     LIB_FUNCTION("FxVZqBAA7ks", "libkernel", 1, "libkernel", 1, 1, ps4__write);
+    LIB_FUNCTION("6XG4B33N09g", "libScePosix", 1, "libkernel", 1, 1, sched_yield);
 }
 
 } // namespace Libraries::Kernel

src/core/libraries/kernel/memory_management.cpp

@@ -80,6 +80,9 @@ s32 PS4_SYSV_ABI sceKernelAvailableDirectMemorySize(u64 searchStart, u64 searchE
 s32 PS4_SYSV_ABI sceKernelVirtualQuery(const void* addr, int flags, OrbisVirtualQueryInfo* info,
                                        size_t infoSize) {
     LOG_INFO(Kernel_Vmm, "called addr = {}, flags = {:#x}", fmt::ptr(addr), flags);
+    if (!addr) {
+        return SCE_KERNEL_ERROR_EACCES;
+    }
     auto* memory = Core::Memory::Instance();
     return memory->VirtualQuery(std::bit_cast<VAddr>(addr), flags, info);
 }
@@ -87,10 +90,10 @@ s32 PS4_SYSV_ABI sceKernelVirtualQuery(const void* addr, int flags, OrbisVirtual
 int PS4_SYSV_ABI sceKernelMapNamedDirectMemory(void** addr, u64 len, int prot, int flags,
                                                s64 directMemoryStart, u64 alignment,
                                                const char* name) {
-    LOG_INFO(
-        Kernel_Vmm,
-        "len = {:#x}, prot = {:#x}, flags = {:#x}, directMemoryStart = {:#x}, alignment = {:#x}",
-        len, prot, flags, directMemoryStart, alignment);
+    LOG_INFO(Kernel_Vmm,
+             "addr = {}, len = {:#x}, prot = {:#x}, flags = {:#x}, directMemoryStart = {:#x}, "
+             "alignment = {:#x}",
+             fmt::ptr(*addr), len, prot, flags, directMemoryStart, alignment);
 
     if (len == 0 || !Common::Is16KBAligned(len)) {
         LOG_ERROR(Kernel_Vmm, "Map size is either zero or not 16KB aligned!");
@@ -117,11 +120,7 @@ int PS4_SYSV_ABI sceKernelMapNamedDirectMemory(void** addr, u64 len, int prot, i
 
 int PS4_SYSV_ABI sceKernelMapDirectMemory(void** addr, u64 len, int prot, int flags,
                                           s64 directMemoryStart, u64 alignment) {
-    LOG_INFO(Kernel_Vmm,
-             "redirected to sceKernelMapNamedDirectMemory: "
-             "len = {:#x}, prot = {:#x}, flags = {:#x}, directMemoryStart = {:#x}, alignment = "
-             "{:#x}",
-             len, prot, flags, directMemoryStart, alignment);
+    LOG_INFO(Kernel_Vmm, "called, redirected to sceKernelMapNamedDirectMemory");
     return sceKernelMapNamedDirectMemory(addr, len, prot, flags, directMemoryStart, alignment, "");
 }
 
@@ -169,7 +168,7 @@ int PS4_SYSV_ABI sceKernelQueryMemoryProtection(void* addr, void** start, void**
 
 int PS4_SYSV_ABI sceKernelDirectMemoryQuery(u64 offset, int flags, OrbisQueryInfo* query_info,
                                             size_t infoSize) {
-    LOG_WARNING(Kernel_Vmm, "called");
+    LOG_WARNING(Kernel_Vmm, "called offset = {:#x}, flags = {:#x}", offset, flags);
     auto* memory = Core::Memory::Instance();
     return memory->DirectMemoryQuery(offset, flags == 1, query_info);
 }

src/core/libraries/kernel/thread_management.cpp

@@ -3,18 +3,20 @@
 
 #include <mutex>
 #include <thread>
+#include <semaphore.h>
 #include "common/assert.h"
+#include "common/error.h"
 #include "common/logging/log.h"
 #include "common/singleton.h"
 #include "common/thread.h"
 #include "core/libraries/error_codes.h"
 #include "core/libraries/kernel/thread_management.h"
+#include "core/libraries/kernel/threads/threads.h"
 #include "core/libraries/libs.h"
 #include "core/linker.h"
 #ifdef _WIN64
 #include <windows.h>
 #endif
-#include "core/libraries/kernel/threads/kernel_threads.h"
 
 namespace Libraries::Kernel {
 
@@ -40,6 +42,7 @@ void init_pthreads() {
     scePthreadRwlockattrInit(&default_rwattr);
     g_pthread_cxt->setDefaultRwattr(default_rwattr);
 
+    g_pthread_cxt->setPthreadKeys(new PthreadKeys);
     g_pthread_cxt->SetPthreadPool(new PThreadPool);
 }
 
@@ -136,9 +139,8 @@ int PS4_SYSV_ABI scePthreadAttrGetdetachstate(const ScePthreadAttr* attr, int* s
         return SCE_KERNEL_ERROR_EINVAL;
     }
 
-    // int result = pthread_attr_getdetachstate(&(*attr)->p, state);
+    // int result = pthread_attr_getdetachstate(&(*attr)->pth_attr, state);
     int result = 0;
-
     *state = ((*attr)->detached ? PTHREAD_CREATE_DETACHED : PTHREAD_CREATE_JOINABLE);
 
     switch (*state) {
@@ -172,12 +174,9 @@ int PS4_SYSV_ABI scePthreadAttrSetdetachstate(ScePthreadAttr* attr, int detachst
         UNREACHABLE_MSG("Invalid detachstate: {}", detachstate);
     }
 
-    // int result = pthread_attr_setdetachstate(&(*attr)->pth_attr, pstate); doesn't seem to work
-    // correctly
+    // int result = pthread_attr_setdetachstate(&(*attr)->pth_attr, pstate);
     int result = 0;
-
     (*attr)->detached = (pstate == PTHREAD_CREATE_DETACHED);
-
     return result == 0 ? SCE_OK : SCE_KERNEL_ERROR_EINVAL;
 }
 
@@ -244,7 +243,6 @@ int PS4_SYSV_ABI scePthreadAttrSetschedparam(ScePthreadAttr* attr,
 }
 
 int PS4_SYSV_ABI scePthreadAttrGetschedpolicy(const ScePthreadAttr* attr, int* policy) {
-
     if (policy == nullptr || attr == nullptr || *attr == nullptr) {
         return SCE_KERNEL_ERROR_EINVAL;
     }
@@ -273,16 +271,26 @@ int PS4_SYSV_ABI scePthreadAttrSetschedpolicy(ScePthreadAttr* attr, int policy)
         return SCE_KERNEL_ERROR_EINVAL;
     }
 
-    int ppolicy = SCHED_OTHER; // winpthreads only supports SCHED_OTHER
-    if (policy != SCHED_OTHER) {
-        LOG_ERROR(Kernel_Pthread, "policy={} not supported by winpthreads\n", policy);
+    int ppolicy = SCHED_OTHER;
+    switch (policy) {
+    case 0:
+        ppolicy = SCHED_OTHER;
+        break;
+    case 1:
+        ppolicy = SCHED_FIFO;
+        break;
+    case 3:
+        ppolicy = SCHED_OTHER;
+        break;
+    default:
+        UNREACHABLE();
     }
+
     (*attr)->policy = policy;
-
     int result = pthread_attr_setschedpolicy(&(*attr)->pth_attr, ppolicy);
-
     return result == 0 ? SCE_OK : SCE_KERNEL_ERROR_EINVAL;
 }
+
 ScePthread PS4_SYSV_ABI scePthreadSelf() {
     return g_pthread_self;
 }
@@ -296,7 +304,6 @@ int PS4_SYSV_ABI scePthreadAttrSetaffinity(ScePthreadAttr* pattr,
     }
 
     (*pattr)->affinity = mask;
-
     return SCE_OK;
 }
 
@@ -389,16 +396,18 @@ int PS4_SYSV_ABI scePthreadSetaffinity(ScePthread thread, const /*SceKernelCpuma
     return result;
 }
 
-void* createMutex(void* addr) {
-    if (addr == nullptr || *static_cast<ScePthreadMutex*>(addr) != nullptr) {
+ScePthreadMutex* createMutex(ScePthreadMutex* addr) {
+    if (addr == nullptr || *addr != nullptr) {
         return addr;
     }
     static std::mutex mutex;
     std::scoped_lock lk{mutex};
-    auto vaddr = reinterpret_cast<u64>(addr);
-
+    if (*addr != nullptr) {
+        return addr;
+    }
+    const VAddr vaddr = reinterpret_cast<VAddr>(addr);
     std::string name = fmt::format("mutex{:#x}", vaddr);
-    scePthreadMutexInit(static_cast<ScePthreadMutex*>(addr), nullptr, name.c_str());
+    scePthreadMutexInit(addr, nullptr, name.c_str());
     return addr;
 }
 
@@ -466,7 +475,7 @@ int PS4_SYSV_ABI scePthreadMutexattrInit(ScePthreadMutexattr* attr) {
 
     int result = pthread_mutexattr_init(&(*attr)->pth_mutex_attr);
 
-    result = (result == 0 ? scePthreadMutexattrSettype(attr, 2) : result);
+    result = (result == 0 ? scePthreadMutexattrSettype(attr, 1) : result);
     result = (result == 0 ? scePthreadMutexattrSetprotocol(attr, 0) : result);
 
     switch (result) {
@@ -517,22 +526,20 @@ int PS4_SYSV_ABI scePthreadMutexattrSetprotocol(ScePthreadMutexattr* attr, int p
         UNREACHABLE_MSG("Invalid protocol: {}", protocol);
     }
 
-    int result = 0; // pthread_mutexattr_setprotocol(&(*attr)->p, pprotocol); //it appears that
-                    // pprotocol has issues in winpthreads
+    int result = pthread_mutexattr_setprotocol(&(*attr)->pth_mutex_attr, pprotocol);
     (*attr)->pprotocol = pprotocol;
-
     return result == 0 ? SCE_OK : SCE_KERNEL_ERROR_EINVAL;
 }
-int PS4_SYSV_ABI scePthreadMutexLock(ScePthreadMutex* mutex) {
-    mutex = static_cast<ScePthreadMutex*>(createMutex(mutex));
 
+int PS4_SYSV_ABI scePthreadMutexLock(ScePthreadMutex* mutex) {
+    mutex = createMutex(mutex);
     if (mutex == nullptr) {
         return SCE_KERNEL_ERROR_EINVAL;
     }
 
     int result = pthread_mutex_lock(&(*mutex)->pth_mutex);
     if (result != 0) {
-        LOG_TRACE(Kernel_Pthread, "name={}, result={}", (*mutex)->name, result);
+        LOG_TRACE(Kernel_Pthread, "Locked name={}, result={}", (*mutex)->name, result);
     }
     switch (result) {
     case 0:
@@ -547,20 +554,20 @@ int PS4_SYSV_ABI scePthreadMutexLock(ScePthreadMutex* mutex) {
         return SCE_KERNEL_ERROR_EINVAL;
     }
 }
+
 int PS4_SYSV_ABI scePthreadMutexUnlock(ScePthreadMutex* mutex) {
-    mutex = static_cast<ScePthreadMutex*>(createMutex(mutex));
+    mutex = createMutex(mutex);
     if (mutex == nullptr) {
         return SCE_KERNEL_ERROR_EINVAL;
     }
 
     int result = pthread_mutex_unlock(&(*mutex)->pth_mutex);
     if (result != 0) {
-        LOG_TRACE(Kernel_Pthread, "name={}, result={}", (*mutex)->name, result);
+        LOG_TRACE(Kernel_Pthread, "Unlocking name={}, result={}", (*mutex)->name, result);
     }
     switch (result) {
     case 0:
         return SCE_OK;
-
     case EINVAL:
         return SCE_KERNEL_ERROR_EINVAL;
     case EPERM:
@@ -571,7 +578,6 @@ int PS4_SYSV_ABI scePthreadMutexUnlock(ScePthreadMutex* mutex) {
 }
 
 int PS4_SYSV_ABI scePthreadMutexattrDestroy(ScePthreadMutexattr* attr) {
-
     int result = pthread_mutexattr_destroy(&(*attr)->pth_mutex_attr);
 
     delete *attr;
@@ -587,12 +593,16 @@ int PS4_SYSV_ABI scePthreadMutexattrDestroy(ScePthreadMutexattr* attr) {
     }
 }
 
-void* createCond(void* addr) {
-    if (addr == nullptr || *static_cast<ScePthreadCond*>(addr) != nullptr) {
+ScePthreadCond* createCond(ScePthreadCond* addr) {
+    if (addr == nullptr || *addr != nullptr) {
         return addr;
     }
-    auto vaddr = reinterpret_cast<u64>(addr);
-
+    static std::mutex mutex;
+    std::scoped_lock lk{mutex};
+    if (*addr != nullptr) {
+        return addr;
+    }
+    const VAddr vaddr = reinterpret_cast<VAddr>(addr);
     std::string name = fmt::format("cond{:#x}", vaddr);
     scePthreadCondInit(static_cast<ScePthreadCond*>(addr), nullptr, name.c_str());
     return addr;
@@ -652,22 +662,20 @@ int PS4_SYSV_ABI scePthreadCondattrInit(ScePthreadCondattr* attr) {
 }
 
 int PS4_SYSV_ABI scePthreadCondBroadcast(ScePthreadCond* cond) {
-    LOG_INFO(Kernel_Pthread, "called");
-    cond = static_cast<ScePthreadCond*>(createCond(cond));
-
+    cond = createCond(cond);
     if (cond == nullptr) {
         return SCE_KERNEL_ERROR_EINVAL;
     }
 
     int result = pthread_cond_broadcast(&(*cond)->cond);
 
-    LOG_INFO(Kernel_Pthread, "name={}, result={}", (*cond)->name, result);
+    LOG_TRACE(Kernel_Pthread, "called name={}, result={}", (*cond)->name, result);
 
     return (result == 0 ? SCE_OK : SCE_KERNEL_ERROR_EINVAL);
 }
 
 int PS4_SYSV_ABI scePthreadCondTimedwait(ScePthreadCond* cond, ScePthreadMutex* mutex, u64 usec) {
-    cond = static_cast<ScePthreadCond*>(createCond(cond));
+    cond = createCond(cond);
     if (cond == nullptr) {
         return SCE_KERNEL_ERROR_EINVAL;
     }
@@ -873,6 +881,11 @@ int PS4_SYSV_ABI scePthreadAttrGet(ScePthread thread, ScePthreadAttr* attr) {
 
 static void cleanup_thread(void* arg) {
     auto* thread = static_cast<ScePthread>(arg);
+    for (const auto& [key, destructor] : thread->key_destructors) {
+        if (void* value = pthread_getspecific(key); value != nullptr) {
+            destructor(value);
+        }
+    }
     thread->is_almost_done = true;
 }
 
@@ -891,7 +904,7 @@ static void* run_thread(void* arg) {
 }
 
 int PS4_SYSV_ABI scePthreadCreate(ScePthread* thread, const ScePthreadAttr* attr,
-                                  pthreadEntryFunc start_routine, void* arg, const char* name) {
+                                  PthreadEntryFunc start_routine, void* arg, const char* name) {
     if (thread == nullptr) {
         return SCE_KERNEL_ERROR_EINVAL;
     }
@@ -1031,9 +1044,9 @@ int PS4_SYSV_ABI scePthreadCondSignal(ScePthreadCond* cond) {
 }
 
 int PS4_SYSV_ABI scePthreadCondWait(ScePthreadCond* cond, ScePthreadMutex* mutex) {
-    if (cond == nullptr || *cond == nullptr) {
-        // return SCE_KERNEL_ERROR_EINVAL;
-        cond = static_cast<ScePthreadCond*>(createCond(cond)); // check this. Kero Blaster.
+    cond = createCond(cond);
+    if (cond == nullptr) {
+        return SCE_KERNEL_ERROR_EINVAL;
     }
     if (mutex == nullptr || *mutex == nullptr) {
         return SCE_KERNEL_ERROR_EINVAL;
@@ -1073,7 +1086,7 @@ int PS4_SYSV_ABI scePthreadCondattrDestroy(ScePthreadCondattr* attr) {
 }
 
 int PS4_SYSV_ABI scePthreadMutexTrylock(ScePthreadMutex* mutex) {
-
+    mutex = createMutex(mutex);
     if (mutex == nullptr) {
         return ORBIS_KERNEL_ERROR_EINVAL;
     }
@@ -1124,7 +1137,7 @@ int PS4_SYSV_ABI posix_sched_get_priority_min() {
 int PS4_SYSV_ABI posix_pthread_mutex_trylock(ScePthreadMutex* mutex) {
     int result = scePthreadMutexTrylock(mutex);
     if (result < 0) {
-        UNREACHABLE();
+        // UNREACHABLE();
     }
     return result;
 }
@@ -1153,7 +1166,7 @@ int PS4_SYSV_ABI posix_pthread_attr_setdetachstate(ScePthreadAttr* attr, int det
 }
 
 int PS4_SYSV_ABI posix_pthread_create_name_np(ScePthread* thread, const ScePthreadAttr* attr,
-                                              pthreadEntryFunc start_routine, void* arg,
+                                              PthreadEntryFunc start_routine, void* arg,
                                               const char* name) {
     LOG_INFO(Kernel_Pthread, "posix pthread_create redirect to scePthreadCreate: name = {}", name);
 
@@ -1167,15 +1180,30 @@ int PS4_SYSV_ABI posix_pthread_create_name_np(ScePthread* thread, const ScePthre
     return result;
 }
 
-int PS4_SYSV_ABI scePthreadOnce(int* once_control, void (*init_routine)(void)) {
-    return pthread_once(reinterpret_cast<pthread_once_t*>(once_control), init_routine);
-}
 int PS4_SYSV_ABI posix_pthread_create(ScePthread* thread, const ScePthreadAttr* attr,
-                                      pthreadEntryFunc start_routine, void* arg) {
-    LOG_INFO(Kernel_Pthread, "posix pthread_create redirect to scePthreadCreate");
+                                      PthreadEntryFunc start_routine, void* arg) {
+    return posix_pthread_create_name_np(thread, attr, start_routine, arg, "NoName");
+}
 
-    int result = scePthreadCreate(thread, attr, start_routine, arg, "");
-    if (result != 0) {
+using Destructor = void (*)(void*);
+
+int PS4_SYSV_ABI posix_pthread_key_create(u32* key, Destructor func) {
+    return pthread_key_create(key, func);
+}
+
+int PS4_SYSV_ABI posix_pthread_setspecific(int key, const void* value) {
+    return pthread_setspecific(key, value);
+}
+
+void* PS4_SYSV_ABI posix_pthread_getspecific(int key) {
+    return pthread_getspecific(key);
+}
+
+int PS4_SYSV_ABI posix_pthread_cond_init(ScePthreadCond* cond, const ScePthreadCondattr* attr,
+                                         const char* name) {
+    // LOG_INFO(Kernel_Pthread, "posix pthread_mutex_init redirect to scePthreadMutexInit");
+    int result = scePthreadCondInit(cond, attr, name);
+    if (result < 0) {
         int rt = result > SCE_KERNEL_ERROR_UNKNOWN && result <= SCE_KERNEL_ERROR_ESTOP
                      ? result + -SCE_KERNEL_ERROR_UNKNOWN
                      : POSIX_EOTHER;
@@ -1183,7 +1211,47 @@ int PS4_SYSV_ABI posix_pthread_create(ScePthread* thread, const ScePthreadAttr*
     }
     return result;
 }
+
+int PS4_SYSV_ABI posix_pthread_setcancelstate(int state, int* oldstate) {
+    return pthread_setcancelstate(state, oldstate);
+}
+
+int PS4_SYSV_ABI posix_pthread_detach(ScePthread thread) {
+    return pthread_detach(thread->pth);
+}
+
+int PS4_SYSV_ABI posix_sem_init(sem_t* sem, int pshared, unsigned int value) {
+    return sem_init(sem, pshared, value);
+}
+
+int PS4_SYSV_ABI posix_sem_wait(sem_t* sem) {
+    return sem_wait(sem);
+}
+
+int PS4_SYSV_ABI posix_sem_post(sem_t* sem) {
+    return sem_post(sem);
+}
+
+int PS4_SYSV_ABI scePthreadGetschedparam(ScePthread thread, int* policy,
+                                         SceKernelSchedParam* param) {
+    return pthread_getschedparam(thread->pth, policy, param);
+}
+
+int PS4_SYSV_ABI scePthreadSetschedparam(ScePthread thread, int policy,
+                                         const SceKernelSchedParam* param) {
+    return pthread_setschedparam(thread->pth, policy, param);
+}
+
+int PS4_SYSV_ABI scePthreadOnce(int* once_control, void (*init_routine)(void)) {
+    return pthread_once(reinterpret_cast<pthread_once_t*>(once_control), init_routine);
+}
+
 void pthreadSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
+    LIB_FUNCTION("lZzFeSxPl08", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_setcancelstate);
+    LIB_FUNCTION("0TyVk4MSLt0", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_cond_init);
+    LIB_FUNCTION("mqULNdimTn0", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_key_create);
+    LIB_FUNCTION("0-KXaS70xy4", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_getspecific);
+    LIB_FUNCTION("WrOLvHU0yQM", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_setspecific);
     LIB_FUNCTION("4+h9EzwKF4I", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetschedpolicy);
     LIB_FUNCTION("-Wreprtu0Qs", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetdetachstate);
     LIB_FUNCTION("eXbUSpEaTsA", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetinheritsched);
@@ -1194,7 +1262,7 @@ void pthreadSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("4qGrR6eoP9Y", "libkernel", 1, "libkernel", 1, 1, scePthreadDetach);
     LIB_FUNCTION("3PtV6p3QNX4", "libkernel", 1, "libkernel", 1, 1, scePthreadEqual);
     LIB_FUNCTION("7Xl257M4VNI", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_equal);
-    LIB_FUNCTION("7Xl257M4VNI", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_join);
+    LIB_FUNCTION("h9CcP3J0oVM", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_join);
 
     LIB_FUNCTION("aI+OeCz8xrQ", "libkernel", 1, "libkernel", 1, 1, scePthreadSelf);
     LIB_FUNCTION("EotR8a3ASf4", "libkernel", 1, "libkernel", 1, 1, posix_pthread_self);
@@ -1202,6 +1270,9 @@ void pthreadSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("3qxgM4ezETA", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetaffinity);
     LIB_FUNCTION("8+s5BzZjxSg", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGetaffinity);
     LIB_FUNCTION("x1X76arYMxU", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGet);
+    LIB_FUNCTION("FXPWHNk8Of0", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGetschedparam);
+    LIB_FUNCTION("P41kTWUS3EI", "libkernel", 1, "libkernel", 1, 1, scePthreadGetschedparam);
+    LIB_FUNCTION("oIRFTjoILbg", "libkernel", 1, "libkernel", 1, 1, scePthreadSetschedparam);
     LIB_FUNCTION("UTXzJbWhhTE", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetstacksize);
     LIB_FUNCTION("vNe1w4diLCs", "libkernel", 1, "libkernel", 1, 1, __tls_get_addr);
     LIB_FUNCTION("OxhIB8LB-PQ", "libkernel", 1, "libkernel", 1, 1, posix_pthread_create);
@@ -1210,6 +1281,8 @@ void pthreadSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("6UgtwV+0zb4", "libkernel", 1, "libkernel", 1, 1, scePthreadCreate);
     LIB_FUNCTION("T72hz6ffq08", "libkernel", 1, "libkernel", 1, 1, scePthreadYield);
     LIB_FUNCTION("-quPa4SEJUw", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGetstack);
+    LIB_FUNCTION("Ru36fiTtJzA", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGetstackaddr);
+    LIB_FUNCTION("-fA+7ZlGDQs", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGetstacksize);
     LIB_FUNCTION("14bOACANTBo", "libkernel", 1, "libkernel", 1, 1, scePthreadOnce);
 
     // mutex calls
@@ -1259,10 +1332,17 @@ void pthreadSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
                  posix_pthread_attr_setdetachstate);
     LIB_FUNCTION("zHchY8ft5pk", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_attr_destroy);
     LIB_FUNCTION("Jmi+9w9u0E4", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_create_name_np);
+    LIB_FUNCTION("OxhIB8LB-PQ", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_create);
+    LIB_FUNCTION("+U1R4WtXvoc", "libScePosix", 1, "libkernel", 1, 1, posix_pthread_detach);
     LIB_FUNCTION("CBNtXOoef-E", "libScePosix", 1, "libkernel", 1, 1, posix_sched_get_priority_max);
     LIB_FUNCTION("m0iS6jNsXds", "libScePosix", 1, "libkernel", 1, 1, posix_sched_get_priority_min);
+    LIB_FUNCTION("pDuPEf3m4fI", "libScePosix", 1, "libkernel", 1, 1, posix_sem_init);
+    LIB_FUNCTION("YCV5dGGBcCo", "libScePosix", 1, "libkernel", 1, 1, posix_sem_wait);
+    LIB_FUNCTION("IKP8typ0QUk", "libScePosix", 1, "libkernel", 1, 1, posix_sem_post);
     // libs
-    ThreadsRwlockSymbolsRegister(sym);
+    RwlockSymbolsRegister(sym);
+    SemaphoreSymbolsRegister(sym);
+    KeySymbolsRegister(sym);
 }
 
 } // namespace Libraries::Kernel

src/core/libraries/kernel/thread_management.h

@@ -28,6 +28,7 @@ struct PthreadCondInternal;
 struct PthreadCondAttrInternal;
 struct PthreadRwInternal;
 struct PthreadRwLockAttrInternal;
+class PthreadKeys;
 
 using SceKernelSchedParam = ::sched_param;
 using ScePthread = PthreadInternal*;
@@ -38,20 +39,24 @@ using ScePthreadCond = PthreadCondInternal*;
 using ScePthreadCondattr = PthreadCondAttrInternal*;
 using OrbisPthreadRwlock = PthreadRwInternal*;
 using OrbisPthreadRwlockattr = PthreadRwLockAttrInternal*;
+using OrbisPthreadKey = u32;
 
-using pthreadEntryFunc = PS4_SYSV_ABI void* (*)(void*);
+using PthreadKeyDestructor = PS4_SYSV_ABI void (*)(void*);
+using PthreadEntryFunc = PS4_SYSV_ABI void* (*)(void*);
 
 struct PthreadInternal {
     u8 reserved[4096];
     std::string name;
     pthread_t pth;
     ScePthreadAttr attr;
-    pthreadEntryFunc entry;
+    PthreadEntryFunc entry;
     void* arg;
     std::atomic_bool is_started;
     std::atomic_bool is_detached;
     std::atomic_bool is_almost_done;
     std::atomic_bool is_free;
+    using Destructor = std::pair<OrbisPthreadKey, PthreadKeyDestructor>;
+    std::vector<Destructor> key_destructors;
 };
 
 struct PthreadAttrInternal {
@@ -106,6 +111,30 @@ private:
     std::mutex m_mutex;
 };
 
+class PthreadKeys {
+public:
+    PthreadKeys() {}
+    virtual ~PthreadKeys() {}
+
+    bool CreateKey(int* key, PthreadKeyDestructor destructor);
+    bool GetKey(int key, int thread_id, void** data);
+    bool SetKey(int key, int thread_id, void* data);
+
+private:
+    struct Map {
+        int thread_id = -1;
+        void* data = nullptr;
+    };
+
+    struct Key {
+        bool used = false;
+        PthreadKeyDestructor destructor = nullptr;
+        std::vector<Map> specific_values;
+    };
+
+    std::mutex m_mutex;
+    Key m_keys[256];
+};
 class PThreadCxt {
 public:
     ScePthreadMutexattr* getDefaultMutexattr() {
@@ -138,6 +167,12 @@ public:
     void setDefaultRwattr(OrbisPthreadRwlockattr attr) {
         m_default_Rwattr = attr;
     }
+    PthreadKeys* getPthreadKeys() {
+        return m_pthread_keys;
+    }
+    void setPthreadKeys(PthreadKeys* keys) {
+        m_pthread_keys = keys;
+    }
 
 private:
     ScePthreadMutexattr m_default_mutexattr = nullptr;
@@ -145,6 +180,7 @@ private:
     ScePthreadAttr m_default_attr = nullptr;
     PThreadPool* m_pthread_pool = nullptr;
     OrbisPthreadRwlockattr m_default_Rwattr = nullptr;
+    PthreadKeys* m_pthread_keys = nullptr;
 };
 
 void init_pthreads();
@@ -161,7 +197,7 @@ int PS4_SYSV_ABI scePthreadAttrSetaffinity(ScePthreadAttr* pattr,
                                            const /*SceKernelCpumask*/ u64 mask);
 int PS4_SYSV_ABI scePthreadSetaffinity(ScePthread thread, const /*SceKernelCpumask*/ u64 mask);
 int PS4_SYSV_ABI scePthreadCreate(ScePthread* thread, const ScePthreadAttr* attr,
-                                  pthreadEntryFunc start_routine, void* arg, const char* name);
+                                  PthreadEntryFunc start_routine, void* arg, const char* name);
 
 /***
  * Mutex calls

src/core/libraries/kernel/threads/kernel_threads.h

@@ -1,53 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#pragma once
-
-#include "core/libraries/kernel/thread_management.h"
-
-namespace Core::Loader {
-class SymbolsResolver;
-}
-
-namespace Libraries::Kernel {
-/****
- * rwlock calls
- */
-int PS4_SYSV_ABI posix_pthread_rwlock_destroy(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI posix_pthread_rwlock_init(OrbisPthreadRwlock* rwlock,
-                                           const OrbisPthreadRwlockattr* attr, const char* name);
-int PS4_SYSV_ABI posix_pthread_rwlock_rdlock(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI posix_pthread_rwlock_reltimedrdlock_np();
-int PS4_SYSV_ABI posix_pthread_rwlock_reltimedwrlock_np();
-int PS4_SYSV_ABI posix_pthread_rwlock_setname_np();
-int PS4_SYSV_ABI posix_pthread_rwlock_timedrdlock();
-int PS4_SYSV_ABI posix_pthread_rwlock_timedwrlock();
-int PS4_SYSV_ABI posix_pthread_rwlock_tryrdlock(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI posix_pthread_rwlock_trywrlock(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI posix_pthread_rwlock_unlock(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI posix_pthread_rwlock_wrlock(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI posix_pthread_rwlockattr_destroy(OrbisPthreadRwlockattr* attr);
-int PS4_SYSV_ABI posix_pthread_rwlockattr_getpshared();
-int PS4_SYSV_ABI posix_pthread_rwlockattr_gettype_np();
-int PS4_SYSV_ABI posix_pthread_rwlockattr_init(OrbisPthreadRwlockattr* attr);
-int PS4_SYSV_ABI posix_pthread_rwlockattr_setpshared();
-int PS4_SYSV_ABI posix_pthread_rwlockattr_settype_np();
-int PS4_SYSV_ABI scePthreadRwlockattrDestroy(OrbisPthreadRwlockattr* attr);
-int PS4_SYSV_ABI scePthreadRwlockattrGetpshared();
-int PS4_SYSV_ABI scePthreadRwlockattrGettype();
-int PS4_SYSV_ABI scePthreadRwlockattrInit(OrbisPthreadRwlockattr* attr);
-int PS4_SYSV_ABI scePthreadRwlockattrSetpshared();
-int PS4_SYSV_ABI scePthreadRwlockattrSettype();
-int PS4_SYSV_ABI scePthreadRwlockDestroy(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI scePthreadRwlockInit(OrbisPthreadRwlock* rwlock,
-                                      const OrbisPthreadRwlockattr* attr, const char* name);
-int PS4_SYSV_ABI scePthreadRwlockRdlock(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI scePthreadRwlockTimedrdlock();
-int PS4_SYSV_ABI scePthreadRwlockTimedwrlock();
-int PS4_SYSV_ABI scePthreadRwlockTryrdlock(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI scePthreadRwlockTrywrlock(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI scePthreadRwlockUnlock(OrbisPthreadRwlock* rwlock);
-int PS4_SYSV_ABI scePthreadRwlockWrlock(OrbisPthreadRwlock* rwlock);
-
-void ThreadsRwlockSymbolsRegister(Core::Loader::SymbolsResolver* sym);
-} // namespace Libraries::Kernel

src/core/libraries/kernel/threads/keys.cpp

@@ -0,0 +1,47 @@
+// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/libraries/error_codes.h"
+#include "core/libraries/kernel/thread_management.h"
+#include "core/libraries/libs.h"
+
+namespace Libraries::Kernel {
+
+int PS4_SYSV_ABI scePthreadKeyCreate(OrbisPthreadKey* key, PthreadKeyDestructor destructor) {
+    if (key == nullptr) {
+        return ORBIS_KERNEL_ERROR_EINVAL;
+    }
+
+    int result = pthread_key_create(key, nullptr);
+    if (destructor) {
+        auto thread = scePthreadSelf();
+        thread->key_destructors.emplace_back(*key, destructor);
+    }
+
+    if (result != 0) {
+        LOG_ERROR(Kernel_Pthread, "scePthreadKeyCreate: error = {}", result);
+        result += ORBIS_KERNEL_ERROR_UNKNOWN;
+    }
+    return result;
+}
+
+void* PS4_SYSV_ABI scePthreadGetspecific(OrbisPthreadKey key) {
+    return pthread_getspecific(key);
+}
+
+int PS4_SYSV_ABI scePthreadSetspecific(OrbisPthreadKey key, /* const*/ void* value) {
+    int result = pthread_setspecific(key, value);
+    if (result != 0) {
+        LOG_ERROR(Kernel_Pthread, "scePthreadSetspecific: error = {}", result);
+        result += ORBIS_KERNEL_ERROR_UNKNOWN;
+    }
+    return result;
+}
+
+void KeySymbolsRegister(Core::Loader::SymbolsResolver* sym) {
+    LIB_FUNCTION("geDaqgH9lTg", "libkernel", 1, "libkernel", 1, 1, scePthreadKeyCreate);
+    LIB_FUNCTION("eoht7mQOCmo", "libkernel", 1, "libkernel", 1, 1, scePthreadGetspecific);
+    LIB_FUNCTION("+BzXYkqYeLE", "libkernel", 1, "libkernel", 1, 1, scePthreadSetspecific);
+}
+
+} // namespace Libraries::Kernel

src/core/libraries/kernel/threads/rwlock.cpp

@@ -4,7 +4,7 @@
 #include "common/logging/log.h"
 #include "core/libraries/error_codes.h"
 #include "core/libraries/libs.h"
-#include "kernel_threads.h"
+#include "threads.h"
 
 namespace Libraries::Kernel {
 
@@ -34,7 +34,23 @@ int PS4_SYSV_ABI posix_pthread_rwlock_init(OrbisPthreadRwlock* rwlock,
     return ORBIS_OK;
 }
 
+OrbisPthreadRwlock* createRwlock(OrbisPthreadRwlock* rwlock) {
+    if (rwlock == nullptr || *rwlock != nullptr) {
+        return rwlock;
+    }
+    static std::mutex mutex;
+    std::scoped_lock lk{mutex};
+    if (*rwlock != nullptr) {
+        return rwlock;
+    }
+    const VAddr addr = std::bit_cast<VAddr>(rwlock);
+    const auto name = fmt::format("rwlock{:#x}", addr);
+    posix_pthread_rwlock_init(rwlock, nullptr, name.c_str());
+    return rwlock;
+}
+
 int PS4_SYSV_ABI posix_pthread_rwlock_rdlock(OrbisPthreadRwlock* rwlock) {
+    rwlock = createRwlock(rwlock);
     int result = pthread_rwlock_rdlock(&(*rwlock)->pth_rwlock);
     if (result != 0) {
         LOG_ERROR(Kernel_Pthread, "posix_pthread_rwlock_rdlock: error = {}", result);
@@ -69,6 +85,10 @@ int PS4_SYSV_ABI posix_pthread_rwlock_timedwrlock() {
 }
 
 int PS4_SYSV_ABI posix_pthread_rwlock_tryrdlock(OrbisPthreadRwlock* rwlock) {
+    rwlock = createRwlock(rwlock);
+    if (rwlock == nullptr) {
+        return ORBIS_KERNEL_ERROR_EINVAL;
+    }
     int result = pthread_rwlock_tryrdlock(&(*rwlock)->pth_rwlock);
     if (result != 0) {
         LOG_ERROR(Kernel_Pthread, "posix_pthread_rwlock_tryrdlock: error = {}", result);
@@ -77,6 +97,10 @@ int PS4_SYSV_ABI posix_pthread_rwlock_tryrdlock(OrbisPthreadRwlock* rwlock) {
 }
 
 int PS4_SYSV_ABI posix_pthread_rwlock_trywrlock(OrbisPthreadRwlock* rwlock) {
+    rwlock = createRwlock(rwlock);
+    if (rwlock == nullptr) {
+        return ORBIS_KERNEL_ERROR_EINVAL;
+    }
     int result = pthread_rwlock_trywrlock(&(*rwlock)->pth_rwlock);
     if (result != 0) {
         LOG_ERROR(Kernel_Pthread, "posix_pthread_rwlock_trywrlock: error = {}", result);
@@ -85,6 +109,10 @@ int PS4_SYSV_ABI posix_pthread_rwlock_trywrlock(OrbisPthreadRwlock* rwlock) {
 }
 
 int PS4_SYSV_ABI posix_pthread_rwlock_unlock(OrbisPthreadRwlock* rwlock) {
+    rwlock = createRwlock(rwlock);
+    if (rwlock == nullptr) {
+        return ORBIS_KERNEL_ERROR_EINVAL;
+    }
     int result = pthread_rwlock_unlock(&(*rwlock)->pth_rwlock);
     if (result != 0) {
         LOG_ERROR(Kernel_Pthread, "posix_pthread_rwlock_unlock: error = {}", result);
@@ -93,6 +121,10 @@ int PS4_SYSV_ABI posix_pthread_rwlock_unlock(OrbisPthreadRwlock* rwlock) {
 }
 
 int PS4_SYSV_ABI posix_pthread_rwlock_wrlock(OrbisPthreadRwlock* rwlock) {
+    rwlock = createRwlock(rwlock);
+    if (rwlock == nullptr) {
+        return ORBIS_KERNEL_ERROR_EINVAL;
+    }
     int result = pthread_rwlock_wrlock(&(*rwlock)->pth_rwlock);
     if (result != 0) {
         LOG_ERROR(Kernel_Pthread, "posix_pthread_rwlock_wrlock: error = {}", result);
@@ -271,9 +303,7 @@ int PS4_SYSV_ABI scePthreadRwlockUnlock(OrbisPthreadRwlock* rwlock) {
 }
 
 int PS4_SYSV_ABI scePthreadRwlockWrlock(OrbisPthreadRwlock* rwlock) {
-    if (rwlock == nullptr || *rwlock == nullptr) {
-        return ORBIS_KERNEL_ERROR_EINVAL;
-    }
+    rwlock = createRwlock(rwlock);
     int result = pthread_rwlock_wrlock(&(*rwlock)->pth_rwlock);
     if (result != 0) {
         LOG_ERROR(Kernel_Pthread, "scePthreadRwlockWrlock: error = {}", result);
@@ -282,7 +312,7 @@ int PS4_SYSV_ABI scePthreadRwlockWrlock(OrbisPthreadRwlock* rwlock) {
     return result;
 }
 
-void ThreadsRwlockSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
+void RwlockSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("1471ajPzxh0", "libkernel", 1, "libkernel", 1, 1, posix_pthread_rwlock_destroy);
     LIB_FUNCTION("ytQULN-nhL4", "libkernel", 1, "libkernel", 1, 1, posix_pthread_rwlock_init);
     LIB_FUNCTION("iGjsr1WAtI0", "libkernel", 1, "libkernel", 1, 1, posix_pthread_rwlock_rdlock);

src/core/libraries/kernel/threads/semaphore.cpp

@@ -0,0 +1,180 @@
+// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <condition_variable>
+#include <mutex>
+#include <boost/intrusive/list.hpp>
+#include <pthread.h>
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "common/scope_exit.h"
+#include "core/libraries/error_codes.h"
+#include "core/libraries/libs.h"
+
+namespace Libraries::Kernel {
+
+using ListBaseHook =
+    boost::intrusive::list_base_hook<boost::intrusive::link_mode<boost::intrusive::normal_link>>;
+
+class Semaphore {
+public:
+    Semaphore(s32 init_count, s32 max_count, const char* name, bool is_fifo)
+        : name{name}, token_count{init_count}, max_count{max_count}, is_fifo{is_fifo} {}
+
+    bool Wait(bool can_block, s32 need_count, u64* timeout) {
+        if (HasAvailableTokens(need_count)) {
+            return true;
+        }
+        if (!can_block) {
+            return false;
+        }
+
+        // Create waiting thread object and add it into the list of waiters.
+        WaitingThread waiter{need_count, is_fifo};
+        AddWaiter(waiter);
+        SCOPE_EXIT {
+            PopWaiter(waiter);
+        };
+
+        // Perform the wait.
+        return waiter.Wait(timeout);
+    }
+
+    bool Signal(s32 signal_count) {
+        std::scoped_lock lk{mutex};
+        if (token_count + signal_count > max_count) {
+            return false;
+        }
+        token_count += signal_count;
+
+        // Wake up threads in order of priority.
+        for (auto& waiter : wait_list) {
+            if (waiter.need_count > token_count) {
+                continue;
+            }
+            token_count -= waiter.need_count;
+            waiter.cv.notify_one();
+        }
+
+        return true;
+    }
+
+private:
+    struct WaitingThread : public ListBaseHook {
+        std::mutex mutex;
+        std::condition_variable cv;
+        u32 priority;
+        s32 need_count;
+
+        explicit WaitingThread(s32 need_count, bool is_fifo) : need_count{need_count} {
+            if (is_fifo) {
+                return;
+            }
+            // Retrieve calling thread priority for sorting into waiting threads list.
+            s32 policy;
+            sched_param param;
+            pthread_getschedparam(pthread_self(), &policy, &param);
+            priority = param.sched_priority;
+        }
+
+        bool Wait(u64* timeout) {
+            std::unique_lock lk{mutex};
+            if (!timeout) {
+                // Wait indefinitely until we are woken up.
+                cv.wait(lk);
+                return true;
+            }
+            // Wait until timeout runs out, recording how much remaining time there was.
+            const auto start = std::chrono::high_resolution_clock::now();
+            const auto status = cv.wait_for(lk, std::chrono::microseconds(*timeout));
+            const auto end = std::chrono::high_resolution_clock::now();
+            const auto time =
+                std::chrono::duration_cast<std::chrono::microseconds>(end - start).count();
+            *timeout -= time;
+            return status != std::cv_status::timeout;
+        }
+
+        bool operator<(const WaitingThread& other) const {
+            return priority < other.priority;
+        }
+    };
+
+    void AddWaiter(WaitingThread& waiter) {
+        std::scoped_lock lk{mutex};
+        // Insert at the end of the list for FIFO order.
+        if (is_fifo) {
+            wait_list.push_back(waiter);
+            return;
+        }
+        // Find the first with priority less then us and insert right before it.
+        auto it = wait_list.begin();
+        while (it != wait_list.end() && it->priority > waiter.priority) {
+            it++;
+        }
+        wait_list.insert(it, waiter);
+    }
+
+    void PopWaiter(WaitingThread& waiter) {
+        std::scoped_lock lk{mutex};
+        wait_list.erase(WaitingThreads::s_iterator_to(waiter));
+    }
+
+    bool HasAvailableTokens(s32 need_count) {
+        std::scoped_lock lk{mutex};
+        if (token_count >= need_count) {
+            token_count -= need_count;
+            return true;
+        }
+        return false;
+    }
+
+    using WaitingThreads =
+        boost::intrusive::list<WaitingThread, boost::intrusive::base_hook<ListBaseHook>,
+                               boost::intrusive::constant_time_size<false>>;
+    WaitingThreads wait_list;
+    std::string name;
+    std::atomic<s32> token_count;
+    std::mutex mutex;
+    s32 max_count;
+    bool is_fifo;
+};
+
+using OrbisKernelSema = Semaphore*;
+
+s32 PS4_SYSV_ABI sceKernelCreateSema(OrbisKernelSema* sem, const char* pName, u32 attr,
+                                     s32 initCount, s32 maxCount, const void* pOptParam) {
+    if (!pName || attr > 2 || initCount < 0 || maxCount <= 0 || initCount > maxCount) {
+        LOG_ERROR(Lib_Kernel, "Semaphore creation parameters are invalid!");
+        return ORBIS_KERNEL_ERROR_EINVAL;
+    }
+    *sem = new Semaphore(initCount, maxCount, pName, attr == 1);
+    return ORBIS_OK;
+}
+
+s32 PS4_SYSV_ABI sceKernelWaitSema(OrbisKernelSema sem, s32 needCount, u64* pTimeout) {
+    ASSERT(sem->Wait(true, needCount, pTimeout));
+    return ORBIS_OK;
+}
+
+s32 PS4_SYSV_ABI sceKernelSignalSema(OrbisKernelSema sem, s32 signalCount) {
+    if (!sem->Signal(signalCount)) {
+        return ORBIS_KERNEL_ERROR_EINVAL;
+    }
+    return ORBIS_OK;
+}
+
+s32 PS4_SYSV_ABI sceKernelPollSema(OrbisKernelSema sem, s32 needCount) {
+    if (!sem->Wait(false, needCount, nullptr)) {
+        return ORBIS_KERNEL_ERROR_EBUSY;
+    }
+    return ORBIS_OK;
+}
+
+void SemaphoreSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
+    LIB_FUNCTION("188x57JYp0g", "libkernel", 1, "libkernel", 1, 1, sceKernelCreateSema);
+    LIB_FUNCTION("Zxa0VhQVTsk", "libkernel", 1, "libkernel", 1, 1, sceKernelWaitSema);
+    LIB_FUNCTION("4czppHBiriw", "libkernel", 1, "libkernel", 1, 1, sceKernelSignalSema);
+    LIB_FUNCTION("12wOHk8ywb0", "libkernel", 1, "libkernel", 1, 1, sceKernelPollSema);
+}
+
+} // namespace Libraries::Kernel

src/core/libraries/kernel/threads/threads.h

@@ -0,0 +1,20 @@
+// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "core/libraries/kernel/thread_management.h"
+
+namespace Core::Loader {
+class SymbolsResolver;
+}
+
+namespace Libraries::Kernel {
+
+int PS4_SYSV_ABI scePthreadRwlockattrInit(OrbisPthreadRwlockattr* attr);
+
+void SemaphoreSymbolsRegister(Core::Loader::SymbolsResolver* sym);
+void RwlockSymbolsRegister(Core::Loader::SymbolsResolver* sym);
+void KeySymbolsRegister(Core::Loader::SymbolsResolver* sym);
+
+} // namespace Libraries::Kernel

src/core/libraries/kernel/time_management.cpp

@@ -153,6 +153,26 @@ int PS4_SYSV_ABI gettimeofday(OrbisKernelTimeval* tp, OrbisKernelTimezone* tz) {
     return sceKernelGettimeofday(tp);
 }
 
+s32 PS4_SYSV_ABI sceKernelGettimezone(OrbisKernelTimezone* tz) {
+#ifdef _WIN64
+    ASSERT(tz);
+    static int tzflag = 0;
+    if (!tzflag) {
+        _tzset();
+        tzflag++;
+    }
+    tz->tz_minuteswest = _timezone / 60;
+    tz->tz_dsttime = _daylight;
+#else
+    struct timezone tzz;
+    struct timeval tv;
+    gettimeofday(&tv, &tzz);
+    tz->tz_dsttime = tzz.tz_dsttime;
+    tz->tz_minuteswest = tzz.tz_minuteswest;
+#endif
+    return ORBIS_OK;
+}
+
 int PS4_SYSV_ABI posix_clock_getres(u32 clock_id, OrbisKernelTimespec* res) {
     if (res == nullptr) {
         return SCE_KERNEL_ERROR_EFAULT;
@@ -198,12 +218,14 @@ void timeSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("n88vx3C5nW8", "libScePosix", 1, "libkernel", 1, 1, gettimeofday);
     LIB_FUNCTION("QvsZxomvUHs", "libkernel", 1, "libkernel", 1, 1, sceKernelNanosleep);
     LIB_FUNCTION("1jfXLRVzisc", "libkernel", 1, "libkernel", 1, 1, sceKernelUsleep);
+    LIB_FUNCTION("QcteRwbsnV0", "libkernel", 1, "libkernel", 1, 1, posix_usleep);
     LIB_FUNCTION("QcteRwbsnV0", "libScePosix", 1, "libkernel", 1, 1, posix_usleep);
     LIB_FUNCTION("-ZR+hG7aDHw", "libkernel", 1, "libkernel", 1, 1, sceKernelSleep);
     LIB_FUNCTION("0wu33hunNdE", "libScePosix", 1, "libkernel", 1, 1, sceKernelSleep);
     LIB_FUNCTION("yS8U2TGCe1A", "libkernel", 1, "libkernel", 1, 1, posix_nanosleep);
     LIB_FUNCTION("yS8U2TGCe1A", "libScePosix", 1, "libkernel", 1, 1, posix_nanosleep);
     LIB_FUNCTION("QBi7HCK03hw", "libkernel", 1, "libkernel", 1, 1, sceKernelClockGettime);
+    LIB_FUNCTION("kOcnerypnQA", "libkernel", 1, "libkernel", 1, 1, sceKernelGettimezone);
     LIB_FUNCTION("lLMT9vJAck0", "libkernel", 1, "libkernel", 1, 1, posix_clock_gettime);
     LIB_FUNCTION("lLMT9vJAck0", "libScePosix", 1, "libkernel", 1, 1, posix_clock_gettime);
     LIB_FUNCTION("smIj7eqzZE8", "libScePosix", 1, "libkernel", 1, 1, posix_clock_getres);

src/core/libraries/libc/libc.cpp

@@ -463,10 +463,10 @@ void libcSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
 
     // stdio functions
     LIB_FUNCTION("xeYO4u7uyJ0", "libc", 1, "libc", 1, 1, ps4_fopen);
-    LIB_FUNCTION("hcuQgD53UxM", "libc", 1, "libc", 1, 1, ps4_printf);
+    // LIB_FUNCTION("hcuQgD53UxM", "libc", 1, "libc", 1, 1, ps4_printf);
     LIB_FUNCTION("Q2V+iqvjgC0", "libc", 1, "libc", 1, 1, ps4_vsnprintf);
     LIB_FUNCTION("YQ0navp+YIc", "libc", 1, "libc", 1, 1, ps4_puts);
-    LIB_FUNCTION("fffwELXNVFA", "libc", 1, "libc", 1, 1, ps4_fprintf);
+    // LIB_FUNCTION("fffwELXNVFA", "libc", 1, "libc", 1, 1, ps4_fprintf);
     LIB_FUNCTION("QMFyLoqNxIg", "libc", 1, "libc", 1, 1, ps4_setvbuf);
     LIB_FUNCTION("uodLYyUip20", "libc", 1, "libc", 1, 1, ps4_fclose);
     LIB_FUNCTION("rQFVBXp-Cxg", "libc", 1, "libc", 1, 1, ps4_fseek);

src/core/libraries/libc_internal/libc_internal.cpp

@@ -39,10 +39,26 @@ int PS4_SYSV_ABI internal_memcmp(const void* s1, const void* s2, size_t n) {
     return std::memcmp(s1, s2, n);
 }
 
+int PS4_SYSV_ABI internal_strncmp(const char* str1, const char* str2, size_t num) {
+    return std::strncmp(str1, str2, num);
+}
+
+int PS4_SYSV_ABI internal_strlen(const char* str) {
+    return std::strlen(str);
+}
+
 float PS4_SYSV_ABI internal_expf(float x) {
     return expf(x);
 }
 
+void* PS4_SYSV_ABI internal_malloc(size_t size) {
+    return std::malloc(size);
+}
+
+char* PS4_SYSV_ABI internal_strncpy(char* dest, const char* src, std::size_t count) {
+    return std::strncpy(dest, src, count);
+}
+
 void RegisterlibSceLibcInternal(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("NFLs+dRJGNg", "libSceLibcInternal", 1, "libSceLibcInternal", 1, 1,
                  internal_memcpy_s);
@@ -55,6 +71,14 @@ void RegisterlibSceLibcInternal(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("DfivPArhucg", "libSceLibcInternal", 1, "libSceLibcInternal", 1, 1,
                  internal_memcmp);
     LIB_FUNCTION("8zsu04XNsZ4", "libSceLibcInternal", 1, "libSceLibcInternal", 1, 1, internal_expf);
+    LIB_FUNCTION("aesyjrHVWy4", "libSceLibcInternal", 1, "libSceLibcInternal", 1, 1,
+                 internal_strncmp);
+    LIB_FUNCTION("j4ViWNHEgww", "libSceLibcInternal", 1, "libSceLibcInternal", 1, 1,
+                 internal_strlen);
+    LIB_FUNCTION("6sJWiWSRuqk", "libSceLibcInternal", 1, "libSceLibcInternal", 1, 1,
+                 internal_strncpy);
+    LIB_FUNCTION("gQX+4GDQjpM", "libSceLibcInternal", 1, "libSceLibcInternal", 1, 1,
+                 internal_malloc);
 };
 
 } // namespace Libraries::LibcInternal

src/core/libraries/libs.h

@@ -3,9 +3,48 @@
 
 #pragma once
 
+#include <functional>
+#include "common/logging/log.h"
 #include "core/loader/elf.h"
 #include "core/loader/symbols_resolver.h"
 
+template <size_t N>
+struct StringLiteral {
+    constexpr StringLiteral(const char (&str)[N]) {
+        std::copy_n(str, N, value);
+    }
+
+    char value[N];
+};
+
+template <StringLiteral name, class F, F f>
+struct wrapper_impl;
+
+template <StringLiteral name, class R, class... Args, PS4_SYSV_ABI R (*f)(Args...)>
+struct wrapper_impl<name, PS4_SYSV_ABI R (*)(Args...), f> {
+    static R PS4_SYSV_ABI wrap(Args... args) {
+        if (std::string_view(name.value) != "scePthreadEqual" &&
+            std::string_view(name.value) != "sceUserServiceGetEvent") {
+            LOG_WARNING(Core_Linker, "Function {} called", name.value);
+        }
+        if constexpr (std::is_same_v<R, s32> || std::is_same_v<R, u32>) {
+            const u32 ret = f(args...);
+            if (ret != 0 && std::string_view(name.value) != "scePthreadEqual") {
+                LOG_WARNING(Core_Linker, "Function {} returned {:#x}", name.value, ret);
+            }
+            return ret;
+        }
+        // stuff
+        return f(args...);
+    }
+};
+
+template <StringLiteral name, class F, F f>
+constexpr auto wrapper = wrapper_impl<name, F, f>::wrap;
+
+// #define W(foo) wrapper<#foo, decltype(&foo), foo>
+#define W(foo) foo
+
 #define LIB_FUNCTION(nid, lib, libversion, mod, moduleVersionMajor, moduleVersionMinor, function)  \
     {                                                                                              \
         Core::Loader::SymbolResolver sr{};                                                         \
@@ -16,7 +55,7 @@
         sr.module_version_major = moduleVersionMajor;                                              \
         sr.module_version_minor = moduleVersionMinor;                                              \
         sr.type = Core::Loader::SymbolType::Function;                                              \
-        auto func = reinterpret_cast<u64>(function);                                               \
+        auto func = reinterpret_cast<u64>(W(function));                                            \
         sym->AddSymbol(sr, func);                                                                  \
     }
 

src/core/libraries/system/userservice.cpp

@@ -104,7 +104,7 @@ int PS4_SYSV_ABI sceUserServiceGetDiscPlayerFlag() {
 }
 
 s32 PS4_SYSV_ABI sceUserServiceGetEvent(OrbisUserServiceEvent* event) {
-    LOG_INFO(Lib_UserService, "(DUMMY) called");
+    LOG_TRACE(Lib_UserService, "(DUMMY) called");
     // fake a loggin event
     static bool logged_in = false;
 

src/core/libraries/videoout/driver.cpp

@@ -201,18 +201,25 @@ void VideoOutDriver::Flip(std::chrono::microseconds timeout) {
     }
 
     // Reset flip label
+    if (req.index != -1) {
         req.port->buffer_labels[req.index] = 0;
+    }
 }
 
 bool VideoOutDriver::SubmitFlip(VideoOutPort* port, s32 index, s64 flip_arg,
                                 bool is_eop /*= false*/) {
+    Vulkan::Frame* frame;
+    if (index == -1) {
+        frame = renderer->PrepareBlankFrame();
+    } else {
         const auto& buffer = port->buffer_slots[index];
         const auto& group = port->groups[buffer.group_index];
-    auto* frame = renderer->PrepareFrame(group, buffer.address_left);
+        frame = renderer->PrepareFrame(group, buffer.address_left);
+    }
 
     std::scoped_lock lock{mutex};
 
-    if (requests.size() >= port->NumRegisteredBuffers()) {
+    if (index != -1 && requests.size() >= port->NumRegisteredBuffers()) {
         LOG_ERROR(Lib_VideoOut, "Flip queue is full");
         return false;
     }

src/core/libraries/videoout/video_out.cpp

@@ -124,14 +124,12 @@ s32 PS4_SYSV_ABI sceVideoOutSubmitFlip(s32 handle, s32 bufferIndex, s32 flipMode
         LOG_WARNING(Lib_VideoOut, "flipmode = {}", flipMode);
     }
 
-    ASSERT_MSG(bufferIndex != -1, "Blank output not supported");
-
     if (bufferIndex < -1 || bufferIndex > 15) {
         LOG_ERROR(Lib_VideoOut, "Invalid bufferIndex = {}", bufferIndex);
         return ORBIS_VIDEO_OUT_ERROR_INVALID_INDEX;
     }
 
-    if (port->buffer_slots[bufferIndex].group_index < 0) {
+    if (bufferIndex != -1 && port->buffer_slots[bufferIndex].group_index < 0) {
         LOG_ERROR(Lib_VideoOut, "Slot in bufferIndex = {} is not registered", bufferIndex);
         return ORBIS_VIDEO_OUT_ERROR_INVALID_INDEX;
     }
@@ -196,7 +194,6 @@ s32 PS4_SYSV_ABI sceVideoOutOpen(SceUserServiceUserId userId, s32 busType, s32 i
     LOG_INFO(Lib_VideoOut, "called");
     ASSERT(userId == UserService::ORBIS_USER_SERVICE_USER_ID_SYSTEM || userId == 0);
     ASSERT(busType == SCE_VIDEO_OUT_BUS_TYPE_MAIN);
-    ASSERT(param == nullptr);
 
     if (index != 0) {
         LOG_ERROR(Lib_VideoOut, "Index != 0");
@@ -259,6 +256,12 @@ s32 sceVideoOutSubmitEopFlip(s32 handle, u32 buf_id, u32 mode, u32 arg, void** u
     return ORBIS_OK;
 }
 
+s32 PS4_SYSV_ABI sceVideoOutGetDeviceCapabilityInfo(
+    s32 handle, SceVideoOutDeviceCapabilityInfo* pDeviceCapabilityInfo) {
+    pDeviceCapabilityInfo->capability = 0;
+    return ORBIS_OK;
+}
+
 void RegisterLib(Core::Loader::SymbolsResolver* sym) {
     driver = std::make_unique<VideoOutDriver>(Config::getScreenWidth(), Config::getScreenHeight());
 
@@ -299,6 +302,8 @@ void RegisterLib(Core::Loader::SymbolsResolver* sym) {
     LIB_FUNCTION("U46NwOiJpys", "libSceVideoOut", 1, "libSceVideoOut", 1, 1, sceVideoOutSubmitFlip);
     LIB_FUNCTION("SbU3dwp80lQ", "libSceVideoOut", 1, "libSceVideoOut", 1, 1,
                  sceVideoOutGetFlipStatus);
+    LIB_FUNCTION("kGVLc3htQE8", "libSceVideoOut", 1, "libSceVideoOut", 1, 1,
+                 sceVideoOutGetDeviceCapabilityInfo);
 }
 
 } // namespace Libraries::VideoOut

src/core/libraries/videoout/video_out.h

@@ -84,6 +84,10 @@ struct SceVideoOutVblankStatus {
     u8 pad1[7] = {};
 };
 
+struct SceVideoOutDeviceCapabilityInfo {
+    u64 capability;
+};
+
 void PS4_SYSV_ABI sceVideoOutSetBufferAttribute(BufferAttribute* attribute, PixelFormat pixelFormat,
                                                 u32 tilingMode, u32 aspectRatio, u32 width,
                                                 u32 height, u32 pitchInPixel);

src/core/linker.cpp

@@ -57,9 +57,6 @@ void Linker::Execute() {
 
     // Calculate static TLS size.
     for (const auto& module : m_modules) {
-        if (module->tls.image_size != 0) {
-            module->tls.modid = ++max_tls_index;
-        }
         static_tls_size += module->tls.image_size;
         module->tls.offset = static_tls_size;
     }
@@ -101,7 +98,7 @@ s32 Linker::LoadModule(const std::filesystem::path& elf_name) {
         return -1;
     }
 
-    auto module = std::make_unique<Module>(elf_name);
+    auto module = std::make_unique<Module>(elf_name, max_tls_index);
     if (!module->IsValid()) {
         LOG_ERROR(Core_Linker, "Provided file {} is not valid ELF file", elf_name.string());
         return -1;
@@ -111,8 +108,24 @@ s32 Linker::LoadModule(const std::filesystem::path& elf_name) {
     return m_modules.size() - 1;
 }
 
+Module* Linker::FindByAddress(VAddr address) {
+    for (auto& module : m_modules) {
+        const VAddr base = module->GetBaseAddress();
+        if (address >= base && address < base + module->aligned_base_size) {
+            return module.get();
+        }
+    }
+    return nullptr;
+}
+
 void Linker::Relocate(Module* module) {
-    module->ForEachRelocation([&](elf_relocation* rel, bool isJmpRel) {
+    module->ForEachRelocation([&](elf_relocation* rel, u32 i, bool isJmpRel) {
+        const u32 bit_idx =
+            (isJmpRel ? module->dynamic_info.relocation_table_size / sizeof(elf_relocation) : 0) +
+            i;
+        if (module->TestRelaBit(bit_idx)) {
+            return;
+        }
         auto type = rel->GetType();
         auto symbol = rel->GetSymbol();
         auto addend = rel->rel_addend;
@@ -167,11 +180,15 @@ void Linker::Relocate(Module* module) {
             switch (sym_bind) {
             case STB_LOCAL:
                 symbol_vitrual_addr = rel_base_virtual_addr + sym.st_value;
+                module->SetRelaBit(bit_idx);
                 break;
             case STB_GLOBAL:
             case STB_WEAK: {
                 rel_name = namesTlb + sym.st_name;
-                Resolve(rel_name, rel_sym_type, module, &symrec);
+                if (Resolve(rel_name, rel_sym_type, module, &symrec)) {
+                    // Only set the rela bit if the symbol was actually resolved and not stubbed.
+                    module->SetRelaBit(bit_idx);
+                }
                 symbol_vitrual_addr = symrec.virtual_address;
                 break;
             }
@@ -203,14 +220,14 @@ const Module* Linker::FindExportedModule(const ModuleInfo& module, const Library
     return it == m_modules.end() ? nullptr : it->get();
 }
 
-void Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Module* m,
+bool Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Module* m,
                      Loader::SymbolRecord* return_info) {
     const auto ids = Common::SplitString(name, '#');
     if (ids.size() != 3) {
         return_info->virtual_address = 0;
         return_info->name = name;
         LOG_ERROR(Core_Linker, "Not Resolved {}", name);
-        return;
+        return false;
     }
 
     const LibraryInfo* library = m->FindLibrary(ids[1]);
@@ -236,7 +253,7 @@ void Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Modul
     }
     if (record) {
         *return_info = *record;
-        return;
+        return true;
     }
 
     const auto aeronid = AeroLib::FindByNid(sr.name.c_str());
@@ -249,18 +266,42 @@ void Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Modul
     }
     LOG_ERROR(Core_Linker, "Linker: Stub resolved {} as {} (lib: {}, mod: {})", sr.name,
               return_info->name, library->name, module->name);
+    return false;
 }
 
 void* Linker::TlsGetAddr(u64 module_index, u64 offset) {
     std::scoped_lock lk{mutex};
 
     DtvEntry* dtv_table = GetTcbBase()->tcb_dtv;
-    ASSERT_MSG(dtv_table[0].counter == dtv_generation_counter,
-               "Reallocation of DTV table is not supported");
+    if (dtv_table[0].counter != dtv_generation_counter) {
+        // Generation counter changed, a dynamic module was either loaded or unloaded.
+        const u32 old_num_dtvs = dtv_table[1].counter;
+        ASSERT_MSG(max_tls_index > old_num_dtvs, "Module unloading unsupported");
+        // Module was loaded, increase DTV table size.
+        DtvEntry* new_dtv_table = new DtvEntry[max_tls_index + 2];
+        std::memcpy(new_dtv_table + 2, dtv_table + 2, old_num_dtvs * sizeof(DtvEntry));
+        new_dtv_table[0].counter = dtv_generation_counter;
+        new_dtv_table[1].counter = max_tls_index;
+        delete[] dtv_table;
 
-    void* module = (u8*)dtv_table[module_index + 1].pointer + offset;
-    ASSERT_MSG(module, "DTV allocation is not supported");
-    return module;
+        // Update TCB pointer.
+        GetTcbBase()->tcb_dtv = new_dtv_table;
+        dtv_table = new_dtv_table;
+    }
+
+    u8* addr = dtv_table[module_index + 1].pointer;
+    if (!addr) {
+        // Module was just loaded by above code. Allocate TLS block for it.
+        Module* module = m_modules[module_index - 1].get();
+        const u32 init_image_size = module->tls.init_image_size;
+        u8* dest = reinterpret_cast<u8*>(heap_api_func(module->tls.image_size));
+        const u8* src = reinterpret_cast<const u8*>(module->tls.image_virtual_addr);
+        std::memcpy(dest, src, init_image_size);
+        std::memset(dest + init_image_size, 0, module->tls.image_size - init_image_size);
+        dtv_table[module_index + 1].pointer = dest;
+        addr = dest;
+    }
+    return addr + offset;
 }
 
 void Linker::InitTlsForThread(bool is_primary) {

src/core/linker.h

@@ -3,6 +3,7 @@
 
 #pragma once
 
+#include <algorithm>
 #include <mutex>
 #include <vector>
 #include "core/module.h"
@@ -37,17 +38,32 @@ public:
         return m_modules.at(index).get();
     }
 
+    void RelocateAnyImports(Module* m) {
+        Relocate(m);
+        for (auto& module : m_modules) {
+            const auto imports = module->GetImportModules();
+            if (std::ranges::contains(imports, m->name, &ModuleInfo::name)) {
+                Relocate(module.get());
+            }
+        }
+    }
+
     void SetHeapApiFunc(void* func) {
         heap_api_func = *reinterpret_cast<HeapApiFunc*>(func);
     }
 
+    void AdvanceGenerationCounter() noexcept {
+        dtv_generation_counter++;
+    }
+
     void* TlsGetAddr(u64 module_index, u64 offset);
     void InitTlsForThread(bool is_primary = false);
 
     s32 LoadModule(const std::filesystem::path& elf_name);
+    Module* FindByAddress(VAddr address);
 
     void Relocate(Module* module);
-    void Resolve(const std::string& name, Loader::SymbolType type, Module* module,
+    bool Resolve(const std::string& name, Loader::SymbolType type, Module* module,
                  Loader::SymbolRecord* return_info);
     void Execute();
     void DebugDump();
@@ -58,7 +74,7 @@ private:
     std::mutex mutex;
     u32 dtv_generation_counter{1};
     size_t static_tls_size{};
-    size_t max_tls_index{};
+    u32 max_tls_index{};
     HeapApiFunc heap_api_func{};
     std::vector<std::unique_ptr<Module>> m_modules;
     Loader::SymbolsResolver m_hle_symbols{};

src/core/loader/dwarf.cpp

@@ -0,0 +1,137 @@
+// SPDX-FileCopyrightText: Copyright (C) 2001-2024 Free Software Foundation, Inc.
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "core/loader/dwarf.h"
+
+namespace Dwarf {
+
+template <typename T>
+T get(uintptr_t addr) {
+    T val;
+    memcpy(&val, reinterpret_cast<void*>(addr), sizeof(T));
+    return val;
+}
+
+static uintptr_t getEncodedP(uintptr_t& addr, uintptr_t end, u8 encoding, uintptr_t datarelBase) {
+    const uintptr_t startAddr = addr;
+    const u8* p = (u8*)addr;
+    uintptr_t result;
+
+    // First get value
+    switch (encoding & 0x0F) {
+    case DW_EH_PE_ptr:
+        result = get<uintptr_t>(addr);
+        p += sizeof(uintptr_t);
+        addr = (uintptr_t)p;
+        break;
+    case DW_EH_PE_udata2:
+        result = get<u16>(addr);
+        p += sizeof(u16);
+        addr = (uintptr_t)p;
+        break;
+    case DW_EH_PE_udata4:
+        result = get<u32>(addr);
+        p += sizeof(u32);
+        addr = (uintptr_t)p;
+        break;
+    case DW_EH_PE_udata8:
+        result = get<u64>(addr);
+        p += sizeof(u64);
+        addr = (uintptr_t)p;
+        break;
+    case DW_EH_PE_sdata2:
+        // Sign extend from signed 16-bit value.
+        result = get<s16>(addr);
+        p += sizeof(s16);
+        addr = (uintptr_t)p;
+        break;
+    case DW_EH_PE_sdata4:
+        // Sign extend from signed 32-bit value.
+        result = get<s32>(addr);
+        p += sizeof(s32);
+        addr = (uintptr_t)p;
+        break;
+    case DW_EH_PE_sdata8:
+        result = get<s64>(addr);
+        p += sizeof(s64);
+        addr = (uintptr_t)p;
+        break;
+    default:
+        UNREACHABLE_MSG("unknown pointer encoding");
+    }
+
+    // Then add relative offset
+    switch (encoding & 0x70) {
+    case DW_EH_PE_absptr:
+        // do nothing
+        break;
+    case DW_EH_PE_pcrel:
+        result += startAddr;
+        break;
+    case DW_EH_PE_textrel:
+        UNREACHABLE_MSG("DW_EH_PE_textrel pointer encoding not supported");
+        break;
+    case DW_EH_PE_datarel:
+        // DW_EH_PE_datarel is only valid in a few places, so the parameter has a
+        // default value of 0, and we abort in the event that someone calls this
+        // function with a datarelBase of 0 and DW_EH_PE_datarel encoding.
+        if (datarelBase == 0)
+            UNREACHABLE_MSG("DW_EH_PE_datarel is invalid with a datarelBase of 0");
+        result += datarelBase;
+        break;
+    case DW_EH_PE_funcrel:
+        UNREACHABLE_MSG("DW_EH_PE_funcrel pointer encoding not supported");
+        break;
+    case DW_EH_PE_aligned:
+        UNREACHABLE_MSG("DW_EH_PE_aligned pointer encoding not supported");
+        break;
+    default:
+        UNREACHABLE_MSG("unknown pointer encoding");
+        break;
+    }
+
+    if (encoding & DW_EH_PE_indirect) {
+        result = get<uintptr_t>(result);
+    }
+
+    return result;
+}
+
+bool DecodeEHHdr(uintptr_t ehHdrStart, uintptr_t ehHdrEnd, EHHeaderInfo& ehHdrInfo) {
+    auto p = ehHdrStart;
+    // Ensure that we don't read data beyond the end of .eh_frame_hdr
+    if (ehHdrEnd - ehHdrStart < 4) {
+        // Don't print a message for an empty .eh_frame_hdr (this can happen if
+        // the linker script defines symbols for it even in the empty case).
+        if (ehHdrEnd == ehHdrStart) {
+            return false;
+        }
+        LOG_ERROR(Core_Linker,
+                  "Unsupported .eh_frame_hdr at {:#x} "
+                  "need at least 4 bytes of data but only got {:#x}",
+                  ehHdrStart, ehHdrEnd - ehHdrStart);
+        return false;
+    }
+
+    const u8 version = get<u8>(p++);
+    if (version != 1) {
+        LOG_CRITICAL(Core_Linker, "Unsupported .eh_frame_hdr version: {:#x} at {:#x}", version,
+                     ehHdrStart);
+        return false;
+    }
+
+    const u8 eh_frame_ptr_enc = get<u8>(p++);
+    const u8 fde_count_enc = get<u8>(p++);
+    ehHdrInfo.table_enc = get<u8>(p++);
+
+    ehHdrInfo.eh_frame_ptr = getEncodedP(p, ehHdrEnd, eh_frame_ptr_enc, ehHdrStart);
+    ehHdrInfo.fde_count =
+        fde_count_enc == DW_EH_PE_omit ? 0 : getEncodedP(p, ehHdrEnd, fde_count_enc, ehHdrStart);
+    ehHdrInfo.table = p;
+
+    return true;
+}
+
+} // namespace Dwarf

src/core/loader/dwarf.h

@@ -0,0 +1,41 @@
+// SPDX-FileCopyrightText: Copyright (C) 2001-2024 Free Software Foundation, Inc.
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/types.h"
+
+namespace Dwarf {
+
+enum {
+    DW_EH_PE_ptr = 0x00,
+    DW_EH_PE_uleb128 = 0x01,
+    DW_EH_PE_udata2 = 0x02,
+    DW_EH_PE_udata4 = 0x03,
+    DW_EH_PE_udata8 = 0x04,
+    DW_EH_PE_signed = 0x08,
+    DW_EH_PE_sleb128 = 0x09,
+    DW_EH_PE_sdata2 = 0x0A,
+    DW_EH_PE_sdata4 = 0x0B,
+    DW_EH_PE_sdata8 = 0x0C,
+    DW_EH_PE_absptr = 0x00,
+    DW_EH_PE_pcrel = 0x10,
+    DW_EH_PE_textrel = 0x20,
+    DW_EH_PE_datarel = 0x30,
+    DW_EH_PE_funcrel = 0x40,
+    DW_EH_PE_aligned = 0x50,
+    DW_EH_PE_indirect = 0x80,
+    DW_EH_PE_omit = 0xFF
+};
+
+/// Information encoded in the EH frame header.
+struct EHHeaderInfo {
+    uintptr_t eh_frame_ptr;
+    size_t fde_count;
+    uintptr_t table;
+    u8 table_enc;
+};
+
+bool DecodeEHHdr(uintptr_t ehHdrStart, uintptr_t ehHdrEnd, EHHeaderInfo& ehHdrInfo);
+
+} // namespace Dwarf

src/core/loader/elf.h

@@ -449,6 +449,15 @@ constexpr u32 R_X86_64_JUMP_SLOT = 7; // Create PLT entry
 constexpr u32 R_X86_64_RELATIVE = 8;  // Adjust by program base
 constexpr u32 R_X86_64_DTPMOD64 = 16;
 
+struct eh_frame_hdr {
+    uint8_t version;
+    uint8_t eh_frame_ptr_enc;
+    uint8_t fde_count_enc;
+    uint8_t table_enc;
+    uint32_t eh_frame_ptr;
+    uint32_t fde_count;
+};
+
 namespace Core::Loader {
 
 class Elf {

src/core/memory.cpp

@@ -84,10 +84,14 @@ int MemoryManager::MapMemory(void** out_addr, VAddr virtual_addr, size_t size, M
                              MemoryMapFlags flags, VMAType type, std::string_view name,
                              bool is_exec, PAddr phys_addr, u64 alignment) {
     std::scoped_lock lk{mutex};
+    if (type == VMAType::Flexible && total_flexible_usage + size > 448_MB) {
+        return SCE_KERNEL_ERROR_ENOMEM;
+    }
 
     // When virtual addr is zero, force it to virtual_base. The guest cannot pass Fixed
     // flag so we will take the branch that searches for free (or reserved) mappings.
     virtual_addr = (virtual_addr == 0) ? impl.VirtualBase() : virtual_addr;
+    alignment = alignment > 0 ? alignment : 16_KB;
 
     VAddr mapped_addr = alignment > 0 ? Common::AlignUp(virtual_addr, alignment) : virtual_addr;
     SCOPE_EXIT {
@@ -101,6 +105,9 @@ int MemoryManager::MapMemory(void** out_addr, VAddr virtual_addr, size_t size, M
             new_vma.phys_base = phys_addr;
             MapVulkanMemory(mapped_addr, size);
         }
+        if (type == VMAType::Flexible) {
+            total_flexible_usage += size;
+        }
     };
 
     // Fixed mapping means the virtual address must exactly match the provided one.
@@ -114,12 +121,18 @@ int MemoryManager::MapMemory(void** out_addr, VAddr virtual_addr, size_t size, M
     // Find the first free area starting with provided virtual address.
     if (False(flags & MemoryMapFlags::Fixed)) {
         auto it = FindVMA(mapped_addr);
+        // If the VMA is free and contains the requested mapping we are done.
+        if (it->second.type == VMAType::Free && it->second.Contains(virtual_addr, size)) {
+            mapped_addr = virtual_addr;
+        } else {
+            // Search for the first free VMA that fits our mapping.
             while (it->second.type != VMAType::Free || it->second.size < size) {
                 it++;
             }
             ASSERT(it != vma_map.end());
-        const VAddr base = it->second.base;
-        mapped_addr = alignment > 0 ? Common::AlignUp(base, alignment) : base;
+            const auto& vma = it->second;
+            mapped_addr = alignment > 0 ? Common::AlignUp(vma.base, alignment) : vma.base;
+        }
     }
 
     // Perform the mapping.
@@ -128,6 +141,35 @@ int MemoryManager::MapMemory(void** out_addr, VAddr virtual_addr, size_t size, M
     return ORBIS_OK;
 }
 
+int MemoryManager::MapFile(void** out_addr, VAddr virtual_addr, size_t size, MemoryProt prot,
+                           MemoryMapFlags flags, uintptr_t fd, size_t offset) {
+    ASSERT(virtual_addr == 0);
+    virtual_addr = impl.VirtualBase();
+    const size_t size_aligned = Common::AlignUp(size, 16_KB);
+
+    // Find first free area to map the file.
+    auto it = FindVMA(virtual_addr);
+    while (it->second.type != VMAType::Free || it->second.size < size_aligned) {
+        it++;
+    }
+    ASSERT(it != vma_map.end());
+
+    // Map the file.
+    const VAddr mapped_addr = it->second.base;
+    impl.MapFile(mapped_addr, size, offset, fd);
+
+    // Add virtual memory area
+    auto& new_vma = AddMapping(mapped_addr, size_aligned);
+    new_vma.disallow_merge = True(flags & MemoryMapFlags::NoCoalesce);
+    new_vma.prot = prot;
+    new_vma.name = "File";
+    new_vma.fd = fd;
+    new_vma.type = VMAType::File;
+
+    *out_addr = std::bit_cast<void*>(mapped_addr);
+    return ORBIS_OK;
+}
+
 void MemoryManager::UnmapMemory(VAddr virtual_addr, size_t size) {
     std::scoped_lock lk{mutex};
 
@@ -137,10 +179,13 @@ void MemoryManager::UnmapMemory(VAddr virtual_addr, size_t size) {
                "Attempting to unmap partially mapped range");
 
     const auto type = it->second.type;
-    const PAddr phys_addr = type == VMAType::Direct ? it->second.phys_base : -1;
+    const bool has_backing = type == VMAType::Direct || type == VMAType::File;
     if (type == VMAType::Direct) {
         UnmapVulkanMemory(virtual_addr, size);
     }
+    if (type == VMAType::Flexible) {
+        total_flexible_usage -= size;
+    }
 
     // Mark region as free and attempt to coalesce it with neighbours.
     auto& vma = it->second;
@@ -150,7 +195,7 @@ void MemoryManager::UnmapMemory(VAddr virtual_addr, size_t size) {
     MergeAdjacent(vma_map, it);
 
     // Unmap the memory region.
-    impl.Unmap(virtual_addr, size, phys_addr);
+    impl.Unmap(virtual_addr, size, has_backing);
     TRACK_FREE(virtual_addr, "VMEM");
 }
 
@@ -200,7 +245,7 @@ int MemoryManager::DirectMemoryQuery(PAddr addr, bool find_next,
     std::scoped_lock lk{mutex};
 
     auto dmem_area = FindDmemArea(addr);
-    if (dmem_area->second.is_free && find_next) {
+    while (dmem_area != dmem_map.end() && dmem_area->second.is_free && find_next) {
         dmem_area++;
     }
 

src/core/memory.h

@@ -36,8 +36,12 @@ enum class MemoryProt : u32 {
 
 enum class MemoryMapFlags : u32 {
     NoFlags = 0,
+    Shared = 1,
+    Private = 2,
     Fixed = 0x10,
     NoOverwrite = 0x0080,
+    NoSync = 0x800,
+    NoCore = 0x20000,
     NoCoalesce = 0x400000,
 };
 DECLARE_ENUM_FLAG_OPERATORS(MemoryMapFlags)
@@ -50,6 +54,7 @@ enum class VMAType : u32 {
     Pooled = 4,
     Stack = 5,
     Code = 6,
+    File = 7,
 };
 
 struct DirectMemoryArea {
@@ -81,6 +86,11 @@ struct VirtualMemoryArea {
     MemoryProt prot = MemoryProt::NoAccess;
     bool disallow_merge = false;
     std::string name = "";
+    uintptr_t fd = 0;
+
+    bool Contains(VAddr addr, size_t size) const {
+        return addr >= base && (addr + size) < (base + this->size);
+    }
 
     bool CanMergeWith(const VirtualMemoryArea& next) const {
         if (disallow_merge || next.disallow_merge) {
@@ -123,6 +133,9 @@ public:
                   MemoryMapFlags flags, VMAType type, std::string_view name = "",
                   bool is_exec = false, PAddr phys_addr = -1, u64 alignment = 0);
 
+    int MapFile(void** out_addr, VAddr virtual_addr, size_t size, MemoryProt prot,
+                MemoryMapFlags flags, uintptr_t fd, size_t offset);
+
     void UnmapMemory(VAddr virtual_addr, size_t size);
 
     int QueryProtection(VAddr addr, void** start, void** end, u32* prot);
@@ -182,6 +195,7 @@ private:
     DMemMap dmem_map;
     VMAMap vma_map;
     std::recursive_mutex mutex;
+    size_t total_flexible_usage{};
 
     struct MappedMemory {
         vk::UniqueBuffer buffer;

src/core/module.cpp

@@ -7,6 +7,7 @@
 #include "common/logging/log.h"
 #include "common/string_util.h"
 #include "core/aerolib/aerolib.h"
+#include "core/loader/dwarf.h"
 #include "core/memory.h"
 #include "core/module.h"
 #include "core/tls.h"
@@ -54,10 +55,11 @@ static std::string EncodeId(u64 nVal) {
     return enc;
 }
 
-Module::Module(const std::filesystem::path& file_) : file{file_} {
+Module::Module(const std::filesystem::path& file_, u32& max_tls_index)
+    : file{file_}, name{file.stem().string()} {
     elf.Open(file);
     if (elf.IsElfFile()) {
-        LoadModuleToMemory();
+        LoadModuleToMemory(max_tls_index);
         LoadDynamicInfo();
         LoadSymbols();
     }
@@ -65,13 +67,13 @@ Module::Module(const std::filesystem::path& file_) : file{file_} {
 
 Module::~Module() = default;
 
-void Module::Start(size_t args, const void* argp, void* param) {
-    LOG_INFO(Core_Linker, "Module started : {}", file.filename().string());
+s32 Module::Start(size_t args, const void* argp, void* param) {
+    LOG_INFO(Core_Linker, "Module started : {}", name);
     const VAddr addr = dynamic_info.init_virtual_addr + GetBaseAddress();
-    reinterpret_cast<EntryFunc>(addr)(args, argp, param);
+    return reinterpret_cast<EntryFunc>(addr)(args, argp, param);
 }
 
-void Module::LoadModuleToMemory() {
+void Module::LoadModuleToMemory(u32& max_tls_index) {
     static constexpr size_t BlockAlign = 0x1000;
     static constexpr u64 TrampolineSize = 8_MB;
 
@@ -84,7 +86,6 @@ void Module::LoadModuleToMemory() {
     // Map module segments (and possible TLS trampolines)
     auto* memory = Core::Memory::Instance();
     void** out_addr = reinterpret_cast<void**>(&base_virtual_addr);
-    const auto name = file.filename().string();
     memory->MapMemory(out_addr, LoadAddress, aligned_base_size + TrampolineSize,
                       MemoryProt::CpuReadWrite, MemoryMapFlags::Fixed, VMAType::Code, name, true);
     LoadAddress += CODE_BASE_INCR * (1 + aligned_base_size / CODE_BASE_INCR);
@@ -98,6 +99,17 @@ void Module::LoadModuleToMemory() {
     LOG_INFO(Core_Linker, "base_size ..............: {:#018x}", base_size);
     LOG_INFO(Core_Linker, "aligned_base_size ......: {:#018x}", aligned_base_size);
 
+    const auto add_segment = [this](const elf_program_header& phdr, bool do_map = true) {
+        const VAddr segment_addr = base_virtual_addr + phdr.p_vaddr;
+        if (do_map) {
+            elf.LoadSegment(segment_addr, phdr.p_offset, phdr.p_filesz);
+        }
+        auto& segment = info.segments[info.num_segments++];
+        segment.address = segment_addr;
+        segment.prot = phdr.p_flags;
+        segment.size = GetAlignedSize(phdr);
+    };
+
     for (u16 i = 0; i < elf_header.e_phnum; i++) {
         const auto header_type = elf.ElfPheaderTypeStr(elf_pheader[i].p_type);
         switch (elf_pheader[i].p_type) {
@@ -118,13 +130,14 @@ void Module::LoadModuleToMemory() {
             LOG_INFO(Core_Linker, "segment_memory_size ...: {}", segment_memory_size);
             LOG_INFO(Core_Linker, "segment_mode ..........: {}", segment_mode);
 
-            elf.LoadSegment(segment_addr, elf_pheader[i].p_offset, segment_file_size);
+            add_segment(elf_pheader[i]);
             if (elf_pheader[i].p_flags & PF_EXEC) {
-                PatchTLS(segment_addr, segment_memory_size, c);
+                PatchTLS(segment_addr, segment_file_size, c);
             }
             break;
         }
         case PT_DYNAMIC:
+            add_segment(elf_pheader[i], false);
             if (elf_pheader[i].p_filesz != 0) {
                 m_dynamic.resize(elf_pheader[i].p_filesz);
                 const VAddr segment_addr = std::bit_cast<VAddr>(m_dynamic.data());
@@ -147,12 +160,31 @@ void Module::LoadModuleToMemory() {
             tls.align = elf_pheader[i].p_align;
             tls.image_virtual_addr = elf_pheader[i].p_vaddr + base_virtual_addr;
             tls.image_size = GetAlignedSize(elf_pheader[i]);
+            if (tls.image_size != 0) {
+                tls.modid = ++max_tls_index;
+            }
             LOG_INFO(Core_Linker, "TLS virtual address = {:#x}", tls.image_virtual_addr);
             LOG_INFO(Core_Linker, "TLS image size      = {}", tls.image_size);
             break;
         case PT_SCE_PROCPARAM:
             proc_param_virtual_addr = elf_pheader[i].p_vaddr + base_virtual_addr;
             break;
+        case PT_GNU_EH_FRAME: {
+            eh_frame_hdr_addr = elf_pheader[i].p_vaddr;
+            eh_frame_hdr_size = elf_pheader[i].p_memsz;
+            const VAddr eh_hdr_start = base_virtual_addr + eh_frame_hdr_addr;
+            const VAddr eh_hdr_end = eh_hdr_start + eh_frame_hdr_size;
+            Dwarf::EHHeaderInfo hdr_info;
+            if (Dwarf::DecodeEHHdr(eh_hdr_start, eh_hdr_end, hdr_info)) {
+                eh_frame_addr = hdr_info.eh_frame_ptr - base_virtual_addr;
+                if (eh_frame_hdr_addr > eh_frame_addr) {
+                    eh_frame_size = (eh_frame_hdr_addr - eh_frame_addr);
+                } else {
+                    eh_frame_size = (aligned_base_size - eh_frame_hdr_addr);
+                }
+            }
+            break;
+        }
         default:
             LOG_ERROR(Core_Linker, "Unimplemented type {}", header_type);
         }
@@ -287,8 +319,8 @@ void Module::LoadDynamicInfo() {
             // the given app. How exactly this is generated isn't known, however it is not necessary
             // to have a valid fingerprint. While an invalid fingerprint will cause a warning to be
             // printed to the kernel log, the ELF will still load and run.
-            LOG_INFO(Core_Linker, "unsupported DT_SCE_FINGERPRINT value = ..........: {:#018x}",
-                     dyn->d_un.d_val);
+            LOG_INFO(Core_Linker, "DT_SCE_FINGERPRINT value = {:#018x}", dyn->d_un.d_val);
+            std::memcpy(info.fingerprint.data(), &dyn->d_un.d_val, sizeof(SCE_DBG_NUM_FINGERPRINT));
             break;
         case DT_SCE_IMPORT_LIB_ATTR:
             // The upper 32-bits should contain the module index multiplied by 0x10000. The lower
@@ -304,6 +336,8 @@ void Module::LoadDynamicInfo() {
             info.value = dyn->d_un.d_val;
             info.name = dynamic_info.str_table + info.name_offset;
             info.enc_id = EncodeId(info.id);
+            const std::string full_name = info.name + ".sprx";
+            full_name.copy(this->info.name.data(), full_name.size());
             break;
         };
         case DT_SCE_MODULE_ATTR:
@@ -321,6 +355,9 @@ void Module::LoadDynamicInfo() {
             LOG_INFO(Core_Linker, "unsupported dynamic tag ..........: {:#018x}", dyn->d_tag);
         }
     }
+    const u32 relabits_num = dynamic_info.relocation_table_size / sizeof(elf_relocation) +
+                             dynamic_info.jmp_relocation_table_size / sizeof(elf_relocation);
+    rela_bits.resize((relabits_num + 7) / 8);
 }
 
 void Module::LoadSymbols() {
@@ -384,6 +421,26 @@ void Module::LoadSymbols() {
     symbol_database(import_sym, false);
 }
 
+OrbisKernelModuleInfoEx Module::GetModuleInfoEx() const {
+    return OrbisKernelModuleInfoEx{
+        .name = info.name,
+        .tls_index = tls.modid,
+        .tls_init_addr = tls.image_virtual_addr,
+        .tls_init_size = tls.init_image_size,
+        .tls_size = tls.image_size,
+        .tls_offset = tls.offset,
+        .tls_align = tls.align,
+        .init_proc_addr = base_virtual_addr + dynamic_info.init_virtual_addr,
+        .fini_proc_addr = base_virtual_addr + dynamic_info.fini_virtual_addr,
+        .eh_frame_hdr_addr = eh_frame_hdr_addr,
+        .eh_frame_addr = eh_frame_addr,
+        .eh_frame_hdr_size = eh_frame_hdr_size,
+        .eh_frame_size = eh_frame_size,
+        .segments = info.segments,
+        .segment_count = info.num_segments,
+    };
+}
+
 const ModuleInfo* Module::FindModule(std::string_view id) {
     const auto& import_modules = dynamic_info.import_modules;
     for (u32 i = 0; const auto& mod : import_modules) {

src/core/module.h

@@ -11,6 +11,46 @@
 
 namespace Core {
 
+static constexpr size_t SCE_DBG_MAX_NAME_LENGTH = 256;
+static constexpr size_t SCE_DBG_MAX_SEGMENTS = 4;
+static constexpr size_t SCE_DBG_NUM_FINGERPRINT = 20;
+
+struct OrbisKernelModuleSegmentInfo {
+    VAddr address;
+    u32 size;
+    s32 prot;
+};
+
+struct OrbisKernelModuleInfo {
+    u64 st_size = sizeof(OrbisKernelModuleInfo);
+    std::array<char, SCE_DBG_MAX_NAME_LENGTH> name;
+    std::array<OrbisKernelModuleSegmentInfo, SCE_DBG_MAX_SEGMENTS> segments;
+    u32 num_segments;
+    std::array<u8, SCE_DBG_NUM_FINGERPRINT> fingerprint;
+};
+
+struct OrbisKernelModuleInfoEx {
+    u64 st_size = sizeof(OrbisKernelModuleInfoEx);
+    std::array<char, SCE_DBG_MAX_NAME_LENGTH> name;
+    s32 id;
+    u32 tls_index;
+    VAddr tls_init_addr;
+    u32 tls_init_size;
+    u32 tls_size;
+    u32 tls_offset;
+    u32 tls_align;
+    VAddr init_proc_addr;
+    VAddr fini_proc_addr;
+    u64 reserved1;
+    u64 reserved2;
+    VAddr eh_frame_hdr_addr;
+    VAddr eh_frame_addr;
+    u32 eh_frame_hdr_size;
+    u32 eh_frame_size;
+    std::array<OrbisKernelModuleSegmentInfo, SCE_DBG_MAX_SEGMENTS> segments;
+    u32 segment_count;
+};
+
 struct ModuleInfo {
     bool operator==(const ModuleInfo& other) const {
         return version_major == other.version_major && version_minor == other.version_minor &&
@@ -46,12 +86,12 @@ struct LibraryInfo {
 };
 
 struct ThreadLocalImage {
-    u64 align;
-    u64 image_size;
-    u64 offset;
+    u32 align;
+    u32 image_size;
+    u32 offset;
     u32 modid;
     VAddr image_virtual_addr;
-    u64 init_image_size;
+    u32 init_image_size;
 };
 
 struct DynamicModuleInfo {
@@ -100,7 +140,7 @@ using ModuleFunc = int (*)(size_t, const void*);
 
 class Module {
 public:
-    explicit Module(const std::filesystem::path& file);
+    explicit Module(const std::filesystem::path& file, u32& max_tls_index);
     ~Module();
 
     VAddr GetBaseAddress() const noexcept {
@@ -111,6 +151,10 @@ public:
         return base_virtual_addr + elf.GetElfEntry();
     }
 
+    OrbisKernelModuleInfo GetModuleInfo() const noexcept {
+        return info;
+    }
+
     bool IsValid() const noexcept {
         return base_virtual_addr != 0;
     }
@@ -151,33 +195,49 @@ public:
 
     void ForEachRelocation(auto&& func) {
         for (u32 i = 0; i < dynamic_info.relocation_table_size / sizeof(elf_relocation); i++) {
-            func(&dynamic_info.relocation_table[i], false);
+            func(&dynamic_info.relocation_table[i], i, false);
         }
         for (u32 i = 0; i < dynamic_info.jmp_relocation_table_size / sizeof(elf_relocation); i++) {
-            func(&dynamic_info.jmp_relocation_table[i], true);
+            func(&dynamic_info.jmp_relocation_table[i], i, true);
         }
     }
 
-    void Start(size_t args, const void* argp, void* param);
-    void LoadModuleToMemory();
+    void SetRelaBit(u32 index) {
+        rela_bits[index >> 3] |= (1 << (index & 7));
+    }
+
+    bool TestRelaBit(u32 index) const {
+        return (rela_bits[index >> 3] >> (index & 7)) & 1;
+    }
+
+    s32 Start(size_t args, const void* argp, void* param);
+    void LoadModuleToMemory(u32& max_tls_index);
     void LoadDynamicInfo();
     void LoadSymbols();
 
+    OrbisKernelModuleInfoEx GetModuleInfoEx() const;
     const ModuleInfo* FindModule(std::string_view id);
     const LibraryInfo* FindLibrary(std::string_view id);
 
 public:
     std::filesystem::path file;
+    std::string name;
     Loader::Elf elf;
     u64 aligned_base_size{};
     VAddr base_virtual_addr{};
     VAddr proc_param_virtual_addr{};
+    VAddr eh_frame_hdr_addr{};
+    VAddr eh_frame_addr{};
+    u32 eh_frame_hdr_size{};
+    u32 eh_frame_size{};
     DynamicModuleInfo dynamic_info{};
     std::vector<u8> m_dynamic;
     std::vector<u8> m_dynamic_data;
     Loader::SymbolsResolver export_sym;
     Loader::SymbolsResolver import_sym;
     ThreadLocalImage tls{};
+    OrbisKernelModuleInfo info{};
+    std::vector<u8> rela_bits;
 };
 
 } // namespace Core

src/core/tls.cpp

@@ -125,7 +125,8 @@ static void PatchFsAccess(u8* code, const TLSPattern& tls_pattern, Xbyak::CodeGe
     const auto target_reg = Xbyak::Reg64(tls_pattern.target_reg);
     c.putSeg(fs);
     c.mov(target_reg, qword[SelfInTcbheadOffset]); // Load self member pointer of tcbhead_t.
-    c.add(target_reg, SpecificFirstBlockOffset + sizeof(uintptr_t) + slot * PthreadKeyDataSize);
+    c.add(target_reg, SpecificFirstBlockOffset + sizeof(uintptr_t) * 2 + slot * PthreadKeyDataSize);
+    c.mov(target_reg, qword[target_reg]);
     c.jmp(code + total_size); // Return to the instruction right after the mov.
 }
 

src/core/tls.h

@@ -13,7 +13,7 @@ namespace Core {
 
 union DtvEntry {
     size_t counter;
-    void* pointer;
+    u8* pointer;
 };
 
 struct Tcb {

src/emulator.cpp

@@ -94,7 +94,11 @@ void Emulator::Run(const std::filesystem::path& file) {
         if (std::filesystem::is_directory(sce_module_folder)) {
             for (const auto& entry : std::filesystem::directory_iterator(sce_module_folder)) {
                 if (entry.path().filename() == "libc.prx" ||
-                    entry.path().filename() == "libSceFios2.prx") {
+                    entry.path().filename() == "libSceFios2.prx" ||
+                    entry.path().filename() == "libSceAudioLatencyEstimation.prx" ||
+                    entry.path().filename() == "libSceJobManager.prx" ||
+                    entry.path().filename() == "libSceNpToolkit2.prx" ||
+                    entry.path().filename() == "libSceS3DConversion.prx") {
                     found = true;
                     LOG_INFO(Loader, "Loading {}", entry.path().string().c_str());
                     linker->LoadModule(entry.path());
@@ -126,7 +130,8 @@ void Emulator::Run(const std::filesystem::path& file) {
 void Emulator::LoadSystemModules(const std::filesystem::path& file) {
     const auto& sys_module_path = Common::FS::GetUserPath(Common::FS::PathType::SysModuleDir);
     for (const auto& entry : std::filesystem::directory_iterator(sys_module_path)) {
-        if (entry.path().filename() == "libSceNgs2.sprx") {
+        if (entry.path().filename() == "libSceNgs2.sprx" ||
+            entry.path().filename() == "libSceLibcInternal.sprx") {
             LOG_INFO(Loader, "Loading {}", entry.path().string().c_str());
             linker->LoadModule(entry.path());
         }

src/shader_recompiler/frontend/translate/scalar_alu.cpp

@@ -160,7 +160,7 @@ void Translator::S_OR_B64(bool negate, const GcnInst& inst) {
     }
 }
 
-void Translator::S_AND_B64(const GcnInst& inst) {
+void Translator::S_AND_B64(bool negate, const GcnInst& inst) {
     const auto get_src = [&](const InstOperand& operand) {
         switch (operand.field) {
         case OperandField::VccLo:
@@ -175,7 +175,10 @@ void Translator::S_AND_B64(const GcnInst& inst) {
     };
     const IR::U1 src0{get_src(inst.src[0])};
     const IR::U1 src1{get_src(inst.src[1])};
-    const IR::U1 result = ir.LogicalAnd(src0, src1);
+    IR::U1 result = ir.LogicalAnd(src0, src1);
+    if (negate) {
+        result = ir.LogicalNot(result);
+    }
     ir.SetScc(result);
     switch (inst.dst[0].field) {
     case OperandField::VccLo:

src/shader_recompiler/frontend/translate/translate.cpp

@@ -121,6 +121,9 @@ IR::U32F32 Translator::GetSrc(const InstOperand& operand, bool force_flt) {
     case OperandField::ConstFloatNeg_2_0:
         value = ir.Imm32(-2.0f);
         break;
+    case OperandField::ConstFloatNeg_4_0:
+        value = ir.Imm32(-4.0f);
+        break;
     case OperandField::VccLo:
         if (force_flt) {
             value = ir.BitCast<IR::F32>(ir.GetVccLo());
@@ -304,6 +307,7 @@ void Translate(IR::Block* block, std::span<const GcnInst> inst_list, Info& info)
         case Opcode::V_MADAK_F32: // Yes these can share the opcode
             translator.V_FMA_F32(inst);
             break;
+        case Opcode::IMAGE_SAMPLE_LZ_O:
         case Opcode::IMAGE_SAMPLE_C_LZ:
         case Opcode::IMAGE_SAMPLE_LZ:
         case Opcode::IMAGE_SAMPLE:
@@ -372,6 +376,9 @@ void Translate(IR::Block* block, std::span<const GcnInst> inst_list, Info& info)
         case Opcode::V_CMP_NLE_F32:
             translator.V_CMP_F32(ConditionOp::GT, false, inst);
             break;
+        case Opcode::V_CMP_NLT_F32:
+            translator.V_CMP_F32(ConditionOp::GE, false, inst);
+            break;
         case Opcode::S_CMP_LG_U32:
             translator.S_CMP(ConditionOp::LG, false, inst);
             break;
@@ -563,7 +570,10 @@ void Translate(IR::Block* block, std::span<const GcnInst> inst_list, Info& info)
             translator.S_OR_B64(true, inst);
             break;
         case Opcode::S_AND_B64:
-            translator.S_AND_B64(inst);
+            translator.S_AND_B64(false, inst);
+            break;
+        case Opcode::S_NAND_B64:
+            translator.S_AND_B64(true, inst);
             break;
         case Opcode::V_LSHRREV_B32:
             translator.V_LSHRREV_B32(inst);

src/shader_recompiler/frontend/translate/translate.h

@@ -41,7 +41,7 @@ public:
     void S_AND_SAVEEXEC_B64(const GcnInst& inst);
     void S_MOV_B64(const GcnInst& inst);
     void S_OR_B64(bool negate, const GcnInst& inst);
-    void S_AND_B64(const GcnInst& inst);
+    void S_AND_B64(bool negate, const GcnInst& inst);
     void S_ADD_I32(const GcnInst& inst);
     void S_AND_B32(const GcnInst& inst);
     void S_LSHR_B32(const GcnInst& inst);

src/shader_recompiler/frontend/translate/vector_alu.cpp

@@ -15,7 +15,8 @@ void Translator::V_SAD(const GcnInst& inst) {
 }
 
 void Translator::V_MAC_F32(const GcnInst& inst) {
-    SetDst(inst.dst[0], ir.FPFma(GetSrc(inst.src[0]), GetSrc(inst.src[1]), GetSrc(inst.dst[0])));
+    SetDst(inst.dst[0], ir.FPFma(GetSrc(inst.src[0], true), GetSrc(inst.src[1], true),
+                                 GetSrc(inst.dst[0], true)));
 }
 
 void Translator::V_CVT_PKRTZ_F16_F32(const GcnInst& inst) {
@@ -134,13 +135,13 @@ void Translator::V_FLOOR_F32(const GcnInst& inst) {
 }
 
 void Translator::V_SUB_F32(const GcnInst& inst) {
-    const IR::F32 src0{GetSrc(inst.src[0])};
-    const IR::F32 src1{GetSrc(inst.src[1])};
+    const IR::F32 src0{GetSrc(inst.src[0], true)};
+    const IR::F32 src1{GetSrc(inst.src[1], true)};
     SetDst(inst.dst[0], ir.FPSub(src0, src1));
 }
 
 void Translator::V_RCP_F32(const GcnInst& inst) {
-    const IR::F32 src0{GetSrc(inst.src[0])};
+    const IR::F32 src0{GetSrc(inst.src[0], true)};
     SetDst(inst.dst[0], ir.FPRecip(src0));
 }
 

src/video_core/amdgpu/liverpool.cpp

@@ -22,8 +22,6 @@ Liverpool::Liverpool() {
 
 Liverpool::~Liverpool() {
     process_thread.request_stop();
-    num_submits = -1;
-    num_submits.notify_one();
     process_thread.join();
 }
 
@@ -31,8 +29,10 @@ void Liverpool::Process(std::stop_token stoken) {
     Common::SetCurrentThreadName("GPU_CommandProcessor");
 
     while (!stoken.stop_requested()) {
-        num_submits.wait(0);
-
+        {
+            std::unique_lock lk{submit_mutex};
+            submit_cv.wait(lk, stoken, [this] { return num_submits != 0; });
+        }
         if (stoken.stop_requested()) {
             break;
         }
@@ -67,7 +67,8 @@ void Liverpool::Process(std::stop_token stoken) {
         }
 
         if (submit_done) {
-            num_submits.notify_all();
+            std::scoped_lock lk{submit_mutex};
+            submit_cv.notify_all();
             submit_done = false;
         }
     }
@@ -76,9 +77,8 @@ void Liverpool::Process(std::stop_token stoken) {
 void Liverpool::WaitGpuIdle() {
     RENDERER_TRACE;
 
-    while (const auto old = num_submits.load()) {
-        num_submits.wait(old);
-    }
+    std::unique_lock lk{submit_mutex};
+    submit_cv.wait(lk, [this] { return num_submits == 0; });
 }
 
 Liverpool::Task Liverpool::ProcessCeUpdate(std::span<const u32> ccb) {
@@ -369,7 +369,6 @@ Liverpool::Task Liverpool::ProcessGraphics(std::span<const u32> dcb, std::span<c
             UNREACHABLE_MSG("Unknown PM4 type 3 opcode {:#x} with count {}",
                             static_cast<u32>(opcode), count);
         }
-
         dcb = dcb.subspan(header->type3.NumWords() + 1);
     }
 
@@ -415,8 +414,9 @@ void Liverpool::SubmitGfx(std::span<const u32> dcb, std::span<const u32> ccb) {
         queue.submits.emplace(task.handle);
     }
 
+    std::scoped_lock lk{submit_mutex};
     ++num_submits;
-    num_submits.notify_one();
+    submit_cv.notify_one();
 }
 
 void Liverpool::SubmitAsc(u32 vqid, std::span<const u32> acb) {
@@ -429,8 +429,9 @@ void Liverpool::SubmitAsc(u32 vqid, std::span<const u32> acb) {
         queue.submits.emplace(task.handle);
     }
 
+    std::scoped_lock lk{submit_mutex};
     ++num_submits;
-    num_submits.notify_one();
+    submit_cv.notify_one();
 }
 
 } // namespace AmdGpu

src/video_core/amdgpu/liverpool.h

@@ -10,6 +10,7 @@
 #include "video_core/amdgpu/pixel_format.h"
 
 #include <array>
+#include <condition_variable>
 #include <coroutine>
 #include <mutex>
 #include <span>
@@ -479,9 +480,9 @@ struct Liverpool {
     };
 
     struct Scissor {
-        union {
-            BitField<0, 16, s32> top_left_x;
-            BitField<16, 16, s32> top_left_y;
+        struct {
+            s16 top_left_x;
+            s16 top_left_y;
         };
         union {
             BitField<0, 15, u32> bottom_right_x;
@@ -865,13 +866,15 @@ public:
     void SubmitAsc(u32 vqid, std::span<const u32> acb);
 
     void WaitGpuIdle();
+
     bool IsGpuIdle() const {
         return num_submits == 0;
     }
 
     void NotifySubmitDone() {
+        std::scoped_lock lk{submit_mutex};
         submit_done = true;
-        num_submits.notify_all();
+        submit_cv.notify_all();
     }
 
     void BindRasterizer(Vulkan::Rasterizer* rasterizer_) {
@@ -939,7 +942,9 @@ private:
 
     Vulkan::Rasterizer* rasterizer{};
     std::jthread process_thread{};
-    std::atomic<u32> num_submits{};
+    u32 num_submits{};
+    std::mutex submit_mutex;
+    std::condition_variable_any submit_cv;
     std::atomic<bool> submit_done{};
 };
 

src/video_core/amdgpu/resource.h

@@ -281,7 +281,8 @@ struct Sampler {
     };
 
     float LodBias() const noexcept {
-        return static_cast<float>(lod_bias);
+        return static_cast<float>(static_cast<int16_t>((lod_bias.Value() ^ 0x2000u) - 0x2000u)) /
+               256.0f;
     }
 
     float MinLod() const noexcept {

src/video_core/renderer_vulkan/liverpool_to_vk.cpp

@@ -347,6 +347,9 @@ vk::Format SurfaceFormat(AmdGpu::DataFormat data_format, AmdGpu::NumberFormat nu
     if (data_format == AmdGpu::DataFormat::Format8_8 && num_format == AmdGpu::NumberFormat::Unorm) {
         return vk::Format::eR8G8Unorm;
     }
+    if (data_format == AmdGpu::DataFormat::FormatBc2 && num_format == AmdGpu::NumberFormat::Unorm) {
+        return vk::Format::eBc2UnormBlock;
+    }
     UNREACHABLE_MSG("Unknown data_format={} and num_format={}", u32(data_format), u32(num_format));
 }
 
@@ -367,6 +370,10 @@ vk::Format DepthFormat(DepthBuffer::ZFormat z_format, DepthBuffer::StencilFormat
         stencil_format == DepthBuffer::StencilFormat::Stencil8) {
         return vk::Format::eD16UnormS8Uint;
     }
+    if (z_format == DepthBuffer::ZFormat::Invald &&
+        stencil_format == DepthBuffer::StencilFormat::Invalid) {
+        return vk::Format::eUndefined;
+    }
     UNREACHABLE();
 }
 

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -9,6 +9,7 @@
 #include "sdl_window.h"
 #include "video_core/renderer_vulkan/renderer_vulkan.h"
 #include "video_core/renderer_vulkan/vk_rasterizer.h"
+#include "video_core/texture_cache/image.h"
 
 #include <vk_mem_alloc.h>
 
@@ -199,6 +200,11 @@ Frame* RendererVulkan::PrepareFrame(const Libraries::VideoOut::BufferAttributeGr
     return PrepareFrameInternal(image);
 }
 
+Frame* RendererVulkan::PrepareBlankFrame() {
+    auto& image = texture_cache.GetImage(VideoCore::NULL_IMAGE_ID);
+    return PrepareFrameInternal(image);
+}
+
 Frame* RendererVulkan::PrepareFrameInternal(VideoCore::Image& image) {
     // Request a free presentation frame.
     Frame* frame = GetRenderFrame();

src/video_core/renderer_vulkan/renderer_vulkan.h

@@ -39,6 +39,7 @@ public:
 
     Frame* PrepareFrame(const Libraries::VideoOut::BufferAttributeGroup& attribute,
                         VAddr cpu_address);
+    Frame* PrepareBlankFrame();
 
     bool ShowSplash(Frame* frame = nullptr);
     void Present(Frame* frame);

src/video_core/renderer_vulkan/vk_compute_pipeline.cpp

@@ -36,7 +36,8 @@ ComputePipeline::ComputePipeline(const Instance& instance_, Scheduler& scheduler
     for (const auto& image : info.images) {
         bindings.push_back({
             .binding = binding++,
-            .descriptorType = vk::DescriptorType::eSampledImage,
+            .descriptorType = image.is_storage ? vk::DescriptorType::eStorageImage
+                                               : vk::DescriptorType::eSampledImage,
             .descriptorCount = 1,
             .stageFlags = vk::ShaderStageFlagBits::eCompute,
         });

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -78,7 +78,7 @@ GraphicsPipeline::GraphicsPipeline(const Instance& instance_, Scheduler& schedul
         .depthClampEnable = false,
         .rasterizerDiscardEnable = false,
         .polygonMode = LiverpoolToVK::PolygonMode(key.polygon_mode),
-        .cullMode = LiverpoolToVK::CullMode(key.cull_mode),
+        .cullMode = vk::CullModeFlagBits::eNone /*LiverpoolToVK::CullMode(key.cull_mode)*/,
         .frontFace = key.front_face == Liverpool::FrontFace::Clockwise
                          ? vk::FrontFace::eClockwise
                          : vk::FrontFace::eCounterClockwise,
@@ -289,7 +289,8 @@ void GraphicsPipeline::BuildDescSetLayout() {
         for (const auto& image : stage.images) {
             bindings.push_back({
                 .binding = binding++,
-                .descriptorType = vk::DescriptorType::eSampledImage,
+                .descriptorType = image.is_storage ? vk::DescriptorType::eStorageImage
+                                                   : vk::DescriptorType::eSampledImage,
                 .descriptorCount = 1,
                 .stageFlags = vk::ShaderStageFlagBits::eVertex | vk::ShaderStageFlagBits::eFragment,
             });
@@ -316,8 +317,8 @@ void GraphicsPipeline::BindResources(Core::MemoryManager* memory, StreamBuffer&
     BindVertexBuffers(staging);
 
     // Bind resource buffers and textures.
-    boost::container::static_vector<vk::DescriptorBufferInfo, 4> buffer_infos;
-    boost::container::static_vector<vk::DescriptorImageInfo, 8> image_infos;
+    boost::container::static_vector<vk::DescriptorBufferInfo, 16> buffer_infos;
+    boost::container::static_vector<vk::DescriptorImageInfo, 16> image_infos;
     boost::container::small_vector<vk::WriteDescriptorSet, 16> set_writes;
     u32 binding{};
 

src/video_core/renderer_vulkan/vk_graphics_pipeline.h

@@ -79,6 +79,10 @@ public:
         return key.write_masks;
     }
 
+    [[nodiscard]] bool IsDepthEnabled() const {
+        return key.depth.depth_enable.Value();
+    }
+
 private:
     void BuildDescSetLayout();
     void BindVertexBuffers(StreamBuffer& staging) const;

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -6,6 +6,7 @@
 #include "common/io_file.h"
 #include "common/path_util.h"
 #include "shader_recompiler/backend/spirv/emit_spirv.h"
+#include "shader_recompiler/exception.h"
 #include "shader_recompiler/recompiler.h"
 #include "shader_recompiler/runtime_info.h"
 #include "video_core/renderer_vulkan/vk_instance.h"
@@ -88,6 +89,8 @@ void PipelineCache::RefreshGraphicsKey() {
     auto& key = graphics_key;
 
     key.depth = regs.depth_control;
+    key.depth.depth_write_enable.Assign(regs.depth_control.depth_write_enable.Value() &&
+                                        !regs.depth_render_control.depth_clear_enable);
     key.depth_bounds_min = regs.depth_bounds_min;
     key.depth_bounds_max = regs.depth_bounds_max;
     key.depth_bias_enable = regs.polygon_control.enable_polygon_offset_back ||
@@ -111,9 +114,10 @@ void PipelineCache::RefreshGraphicsKey() {
     key.front_face = regs.polygon_control.front_face;
 
     const auto& db = regs.depth_buffer;
-    key.depth_format = key.depth.depth_enable
-                           ? LiverpoolToVK::DepthFormat(db.z_info.format, db.stencil_info.format)
-                           : vk::Format::eUndefined;
+    if (key.depth.depth_enable) {
+        key.depth_format = LiverpoolToVK::DepthFormat(db.z_info.format, db.stencil_info.format);
+        key.depth.depth_enable.Assign(key.depth_format != vk::Format::eUndefined);
+    }
     // `RenderingInfo` is assumed to be initialized with a contiguous array of valid color
     // attachments. This might be not a case as HW color buffers can be bound in an arbitrary order.
     // We need to do some arrays compaction at this stage
@@ -180,6 +184,7 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline() {
         inst_pool.ReleaseContents();
 
         // Recompile shader to IR.
+        LOG_INFO(Render_Vulkan, "Compiling {} shader {:#X}", stage, hash);
         const Shader::Info info = MakeShaderInfo(stage, pgm->user_data, regs);
         programs[i] = Shader::TranslateProgram(inst_pool, block_pool, code, std::move(info));
 

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -62,7 +62,7 @@ void Rasterizer::Draw(bool is_indexed, u32 index_offset) {
             .storeOp = vk::AttachmentStoreOp::eStore,
         });
     }
-    if (regs.depth_control.depth_enable && regs.depth_buffer.Address() != 0) {
+    if (pipeline->IsDepthEnabled() && regs.depth_buffer.Address() != 0) {
         const bool is_clear = regs.depth_render_control.depth_clear_enable;
         const auto& image_view =
             texture_cache.DepthTarget(regs.depth_buffer, liverpool->last_db_extent);

src/video_core/texture_cache/image.h

@@ -88,6 +88,8 @@ private:
     vk::Image image{};
 };
 
+constexpr SlotId NULL_IMAGE_ID{0};
+
 struct Image {
     explicit Image(const Vulkan::Instance& instance, Vulkan::Scheduler& scheduler,
                    const ImageInfo& info, VAddr cpu_addr);

src/video_core/texture_cache/image_view.cpp

@@ -46,17 +46,20 @@ vk::ComponentSwizzle ConvertComponentSwizzle(u32 dst_sel) {
     }
 }
 
-ImageViewInfo::ImageViewInfo(const AmdGpu::Image& image) noexcept {
+ImageViewInfo::ImageViewInfo(const AmdGpu::Image& image, bool is_storage) noexcept
+    : is_storage{is_storage} {
     type = ConvertImageViewType(image.type);
     format = Vulkan::LiverpoolToVK::SurfaceFormat(image.GetDataFmt(), image.GetNumberFmt());
     range.base.level = 0;
     range.base.layer = 0;
     range.extent.levels = image.NumLevels();
     range.extent.layers = image.NumLayers();
+    if (!is_storage) {
         mapping.r = ConvertComponentSwizzle(image.dst_sel_x);
         mapping.g = ConvertComponentSwizzle(image.dst_sel_y);
         mapping.b = ConvertComponentSwizzle(image.dst_sel_z);
         mapping.a = ConvertComponentSwizzle(image.dst_sel_w);
+    }
 }
 
 ImageView::ImageView(const Vulkan::Instance& instance, const ImageViewInfo& info_, Image& image,
@@ -74,7 +77,7 @@ ImageView::ImageView(const Vulkan::Instance& instance, const ImageViewInfo& info
     }
 
     const vk::ImageViewCreateInfo image_view_ci = {
-        .pNext = usage_override.has_value() ? &usage_ci : nullptr,
+        .pNext = nullptr,
         .image = image.image,
         .viewType = info.type,
         .format = format,

src/video_core/texture_cache/image_view.h

@@ -18,12 +18,13 @@ namespace VideoCore {
 
 struct ImageViewInfo {
     explicit ImageViewInfo() = default;
-    explicit ImageViewInfo(const AmdGpu::Image& image) noexcept;
+    explicit ImageViewInfo(const AmdGpu::Image& image, bool is_storage) noexcept;
 
     vk::ImageViewType type = vk::ImageViewType::e2D;
     vk::Format format = vk::Format::eR8G8B8A8Unorm;
     SubresourceRange range;
     vk::ComponentMapping mapping{};
+    bool is_storage;
 
     auto operator<=>(const ImageViewInfo&) const = default;
 };

src/video_core/texture_cache/texture_cache.cpp

@@ -169,14 +169,14 @@ ImageView& TextureCache::FindImageView(const AmdGpu::Image& desc, bool is_storag
         image.Transit(vk::ImageLayout::eShaderReadOnlyOptimal, vk::AccessFlagBits::eShaderRead);
     }
 
-    const ImageViewInfo view_info{desc};
+    const ImageViewInfo view_info{desc, is_storage};
     return RegisterImageView(image, view_info);
 }
 
 ImageView& TextureCache::RenderTarget(const AmdGpu::Liverpool::ColorBuffer& buffer,
                                       const AmdGpu::Liverpool::CbDbExtent& hint) {
     const ImageInfo info{buffer, hint};
-    auto& image = FindImage(info, buffer.Address());
+    auto& image = FindImage(info, buffer.Address(), false);
     image.flags &= ~ImageFlagBits::CpuModified;
 
     image.Transit(vk::ImageLayout::eColorAttachmentOptimal,

src/video_core/texture_cache/texture_cache.h

@@ -55,6 +55,11 @@ public:
     /// Retrieves the sampler that matches the provided S# descriptor.
     [[nodiscard]] vk::Sampler GetSampler(const AmdGpu::Sampler& sampler);
 
+    /// Retrieves the image with the specified id.
+    [[nodiscard]] Image& GetImage(ImageId id) {
+        return slot_images[id];
+    }
+
 private:
     ImageView& RegisterImageView(Image& image, const ImageViewInfo& view_info);