catto
/
shadPS4
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

linker: Proper TLS implementation

This commit is contained in:
raphaelthegreat 2024-06-03 03:54:12 +03:00
parent 772891bfa7
commit 2bbe1349c2
29 changed files with 905 additions and 747 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
CMakeLists.txt
View File

@ -287,6 +287,8 @@ set(CORE src/core/aerolib/stubs.cpp
src/core/linker.h src/core/linker.h
src/core/memory.cpp src/core/memory.cpp
src/core/memory.h src/core/memory.h
src/core/module.cpp
src/core/module.h
src/core/platform.h src/core/platform.h
src/core/memory.h src/core/memory.h
src/core/tls.cpp src/core/tls.cpp

6
src/common/alignment.h
View File

@ -9,7 +9,7 @@
namespace Common { namespace Common {
template <typename T> template <typename T>
[[nodiscard]] constexpr T alignUp(T value, std::size_t size) { [[nodiscard]] constexpr T AlignUp(T value, std::size_t size) {
static_assert(std::is_unsigned_v<T>, "T must be an unsigned value."); static_assert(std::is_unsigned_v<T>, "T must be an unsigned value.");
auto mod{static_cast<T>(value % size)}; auto mod{static_cast<T>(value % size)};
value -= mod; value -= mod;
@ -17,14 +17,14 @@ template <typename T>
} }
template <typename T> template <typename T>
[[nodiscard]] constexpr T alignDown(T value, std::size_t size) { [[nodiscard]] constexpr T AlignDown(T value, std::size_t size) {
static_assert(std::is_unsigned_v<T>, "T must be an unsigned value."); static_assert(std::is_unsigned_v<T>, "T must be an unsigned value.");
return static_cast<T>(value - value % size); return static_cast<T>(value - value % size);
} }
template <typename T> template <typename T>
requires std::is_integral_v<T> requires std::is_integral_v<T>
[[nodiscard]] constexpr bool is16KBAligned(T value) { [[nodiscard]] constexpr bool Is16KBAligned(T value) {
return (value & 0x3FFF) == 0; return (value & 0x3FFF) == 0;
} }

5
src/core/libraries/kernel/libkernel.cpp
View File

@ -204,7 +204,8 @@ void LibKernel_Register(Core::Loader::SymbolsResolver* sym) {
LIB_OBJ("f7uOxY9mM1U", "libkernel", 1, "libkernel", 1, 1, &g_stack_chk_guard); LIB_OBJ("f7uOxY9mM1U", "libkernel", 1, "libkernel", 1, 1, &g_stack_chk_guard);
// memory // memory
LIB_FUNCTION("rTXw65xmLIA", "libkernel", 1, "libkernel", 1, 1, sceKernelAllocateDirectMemory); LIB_FUNCTION("rTXw65xmLIA", "libkernel", 1, "libkernel", 1, 1, sceKernelAllocateDirectMemory);
LIB_FUNCTION("B+vc2AO2Zrc", "libkernel", 1, "libkernel", 1, 1, sceKernelAllocateMainDirectMemory); LIB_FUNCTION("B+vc2AO2Zrc", "libkernel", 1, "libkernel", 1, 1,
sceKernelAllocateMainDirectMemory);
LIB_FUNCTION("pO96TwzOm5E", "libkernel", 1, "libkernel", 1, 1, sceKernelGetDirectMemorySize); LIB_FUNCTION("pO96TwzOm5E", "libkernel", 1, "libkernel", 1, 1, sceKernelGetDirectMemorySize);
LIB_FUNCTION("L-Q3LEjIbgA", "libkernel", 1, "libkernel", 1, 1, sceKernelMapDirectMemory); LIB_FUNCTION("L-Q3LEjIbgA", "libkernel", 1, "libkernel", 1, 1, sceKernelMapDirectMemory);
LIB_FUNCTION("WFcfL2lzido", "libkernel", 1, "libkernel", 1, 1, sceKernelQueryMemoryProtection); LIB_FUNCTION("WFcfL2lzido", "libkernel", 1, "libkernel", 1, 1, sceKernelQueryMemoryProtection);
@ -213,6 +214,8 @@ void LibKernel_Register(Core::Loader::SymbolsResolver* sym) {
LIB_FUNCTION("cQke9UuBQOk", "libkernel", 1, "libkernel", 1, 1, sceKernelMunmap); LIB_FUNCTION("cQke9UuBQOk", "libkernel", 1, "libkernel", 1, 1, sceKernelMunmap);
LIB_FUNCTION("mL8NDH86iQI", "libkernel", 1, "libkernel", 1, 1, sceKernelMapNamedFlexibleMemory); LIB_FUNCTION("mL8NDH86iQI", "libkernel", 1, "libkernel", 1, 1, sceKernelMapNamedFlexibleMemory);
LIB_FUNCTION("IWIBBdTHit4", "libkernel", 1, "libkernel", 1, 1, sceKernelMapFlexibleMemory); LIB_FUNCTION("IWIBBdTHit4", "libkernel", 1, "libkernel", 1, 1, sceKernelMapFlexibleMemory);
LIB_FUNCTION("p5EcQeEeJAE", "libkernel", 1, "libkernel", 1, 1, _sceKernelRtldSetApplicationHeapAPI);
// equeue // equeue
LIB_FUNCTION("D0OdFMjp46I", "libkernel", 1, "libkernel", 1, 1, sceKernelCreateEqueue); LIB_FUNCTION("D0OdFMjp46I", "libkernel", 1, "libkernel", 1, 1, sceKernelCreateEqueue);
LIB_FUNCTION("jpFjmgAC5AE", "libkernel", 1, "libkernel", 1, 1, sceKernelDeleteEqueue); LIB_FUNCTION("jpFjmgAC5AE", "libkernel", 1, "libkernel", 1, 1, sceKernelDeleteEqueue);

24
src/core/libraries/kernel/memory_management.cpp
View File

@ -8,6 +8,7 @@
#include "core/libraries/error_codes.h" #include "core/libraries/error_codes.h"
#include "core/libraries/kernel/memory_management.h" #include "core/libraries/kernel/memory_management.h"
#include "core/memory.h" #include "core/memory.h"
#include "core/linker.h"
namespace Libraries::Kernel { namespace Libraries::Kernel {
@ -23,11 +24,11 @@ int PS4_SYSV_ABI sceKernelAllocateDirectMemory(s64 searchStart, s64 searchEnd, u
return SCE_KERNEL_ERROR_EINVAL; return SCE_KERNEL_ERROR_EINVAL;
} }
const bool is_in_range = searchEnd - searchStart >= len; const bool is_in_range = searchEnd - searchStart >= len;
if (len <= 0 || !Common::is16KBAligned(len) || !is_in_range) { if (len <= 0 || !Common::Is16KBAligned(len) || !is_in_range) {
LOG_ERROR(Kernel_Vmm, "Provided address range is invalid!"); LOG_ERROR(Kernel_Vmm, "Provided address range is invalid!");
return SCE_KERNEL_ERROR_EINVAL; return SCE_KERNEL_ERROR_EINVAL;
} }
if ((alignment != 0 || Common::is16KBAligned(alignment)) && !std::has_single_bit(alignment)) { if ((alignment != 0 || Common::Is16KBAligned(alignment)) && !std::has_single_bit(alignment)) {
LOG_ERROR(Kernel_Vmm, "Alignment value is invalid!"); LOG_ERROR(Kernel_Vmm, "Alignment value is invalid!");
return SCE_KERNEL_ERROR_EINVAL; return SCE_KERNEL_ERROR_EINVAL;
} }
@ -48,8 +49,10 @@ int PS4_SYSV_ABI sceKernelAllocateDirectMemory(s64 searchStart, s64 searchEnd, u
return SCE_OK; return SCE_OK;
} }
s32 PS4_SYSV_ABI sceKernelAllocateMainDirectMemory(size_t len, size_t alignment, int memoryType, s64 *physAddrOut) { s32 PS4_SYSV_ABI sceKernelAllocateMainDirectMemory(size_t len, size_t alignment, int memoryType,
return sceKernelAllocateDirectMemory(0, SCE_KERNEL_MAIN_DMEM_SIZE, len, alignment, memoryType, physAddrOut); s64* physAddrOut) {
return sceKernelAllocateDirectMemory(0, SCE_KERNEL_MAIN_DMEM_SIZE, len, alignment, memoryType,
physAddrOut);
} }
int PS4_SYSV_ABI sceKernelMapDirectMemory(void** addr, u64 len, int prot, int flags, int PS4_SYSV_ABI sceKernelMapDirectMemory(void** addr, u64 len, int prot, int flags,
@ -59,16 +62,16 @@ int PS4_SYSV_ABI sceKernelMapDirectMemory(void** addr, u64 len, int prot, int fl
"len = {:#x}, prot = {:#x}, flags = {:#x}, directMemoryStart = {:#x}, alignment = {:#x}", "len = {:#x}, prot = {:#x}, flags = {:#x}, directMemoryStart = {:#x}, alignment = {:#x}",
len, prot, flags, directMemoryStart, alignment); len, prot, flags, directMemoryStart, alignment);
if (len == 0 || !Common::is16KBAligned(len)) { if (len == 0 || !Common::Is16KBAligned(len)) {
LOG_ERROR(Kernel_Vmm, "Map size is either zero or not 16KB aligned!"); LOG_ERROR(Kernel_Vmm, "Map size is either zero or not 16KB aligned!");
return SCE_KERNEL_ERROR_EINVAL; return SCE_KERNEL_ERROR_EINVAL;
} }
if (!Common::is16KBAligned(directMemoryStart)) { if (!Common::Is16KBAligned(directMemoryStart)) {
LOG_ERROR(Kernel_Vmm, "Start address is not 16KB aligned!"); LOG_ERROR(Kernel_Vmm, "Start address is not 16KB aligned!");
return SCE_KERNEL_ERROR_EINVAL; return SCE_KERNEL_ERROR_EINVAL;
} }
if (alignment != 0) { if (alignment != 0) {
if ((!std::has_single_bit(alignment) && !Common::is16KBAligned(alignment))) { if ((!std::has_single_bit(alignment) && !Common::Is16KBAligned(alignment))) {
LOG_ERROR(Kernel_Vmm, "Alignment value is invalid!"); LOG_ERROR(Kernel_Vmm, "Alignment value is invalid!");
return SCE_KERNEL_ERROR_EINVAL; return SCE_KERNEL_ERROR_EINVAL;
} }
@ -85,7 +88,7 @@ int PS4_SYSV_ABI sceKernelMapDirectMemory(void** addr, u64 len, int prot, int fl
s32 PS4_SYSV_ABI sceKernelMapNamedFlexibleMemory(void** addr_in_out, std::size_t len, int prot, s32 PS4_SYSV_ABI sceKernelMapNamedFlexibleMemory(void** addr_in_out, std::size_t len, int prot,
int flags, const char* name) { int flags, const char* name) {
if (len == 0 || !Common::is16KBAligned(len)) { if (len == 0 || !Common::Is16KBAligned(len)) {
LOG_ERROR(Kernel_Vmm, "len is 0 or not 16kb multiple"); LOG_ERROR(Kernel_Vmm, "len is 0 or not 16kb multiple");
return ORBIS_KERNEL_ERROR_EINVAL; return ORBIS_KERNEL_ERROR_EINVAL;
} }
@ -131,4 +134,9 @@ int PS4_SYSV_ABI sceKernelDirectMemoryQuery(u64 offset, int flags, OrbisQueryInf
return memory->DirectMemoryQuery(offset, flags == 1, query_info); return memory->DirectMemoryQuery(offset, flags == 1, query_info);
} }
void PS4_SYSV_ABI _sceKernelRtldSetApplicationHeapAPI(void* func) {
auto* linker = Common::Singleton<Core::Linker>::Instance();
linker->SetHeapApiFunc(func);
}
} // namespace Libraries::Kernel } // namespace Libraries::Kernel

5
src/core/libraries/kernel/memory_management.h
View File

@ -39,7 +39,8 @@ struct OrbisQueryInfo {
u64 PS4_SYSV_ABI sceKernelGetDirectMemorySize(); u64 PS4_SYSV_ABI sceKernelGetDirectMemorySize();
int PS4_SYSV_ABI sceKernelAllocateDirectMemory(s64 searchStart, s64 searchEnd, u64 len, int PS4_SYSV_ABI sceKernelAllocateDirectMemory(s64 searchStart, s64 searchEnd, u64 len,
u64 alignment, int memoryType, s64* physAddrOut); u64 alignment, int memoryType, s64* physAddrOut);
s32 PS4_SYSV_ABI sceKernelAllocateMainDirectMemory(size_t len, size_t alignment, int memoryType, s64 *physAddrOut); s32 PS4_SYSV_ABI sceKernelAllocateMainDirectMemory(size_t len, size_t alignment, int memoryType,
s64* physAddrOut);
int PS4_SYSV_ABI sceKernelMapDirectMemory(void** addr, u64 len, int prot, int flags, int PS4_SYSV_ABI sceKernelMapDirectMemory(void** addr, u64 len, int prot, int flags,
s64 directMemoryStart, u64 alignment); s64 directMemoryStart, u64 alignment);
s32 PS4_SYSV_ABI sceKernelMapNamedFlexibleMemory(void** addrInOut, std::size_t len, int prot, s32 PS4_SYSV_ABI sceKernelMapNamedFlexibleMemory(void** addrInOut, std::size_t len, int prot,
@ -51,4 +52,6 @@ int PS4_SYSV_ABI sceKernelQueryMemoryProtection(void* addr, void** start, void**
int PS4_SYSV_ABI sceKernelDirectMemoryQuery(u64 offset, int flags, OrbisQueryInfo* query_info, int PS4_SYSV_ABI sceKernelDirectMemoryQuery(u64 offset, int flags, OrbisQueryInfo* query_info,
size_t infoSize); size_t infoSize);
void PS4_SYSV_ABI _sceKernelRtldSetApplicationHeapAPI(void* func);
} // namespace Libraries::Kernel } // namespace Libraries::Kernel

2
src/core/libraries/kernel/physical_memory.cpp
View File

@ -20,7 +20,7 @@ bool PhysicalMemory::Alloc(u64 searchStart, u64 searchEnd, u64 len, u64 alignmen
} }
// Align free position // Align free position
find_free_pos = Common::alignUp(find_free_pos, alignment); find_free_pos = Common::AlignUp(find_free_pos, alignment);
// If the new position is between searchStart - searchEnd , allocate a new block // If the new position is between searchStart - searchEnd , allocate a new block
if (find_free_pos >= searchStart && find_free_pos + len <= searchEnd) { if (find_free_pos >= searchStart && find_free_pos + len <= searchEnd) {

17
src/core/libraries/kernel/thread_management.cpp
View File

@ -6,10 +6,12 @@
#include "common/assert.h" #include "common/assert.h"
#include "common/logging/log.h" #include "common/logging/log.h"
#include "common/thread.h" #include "common/thread.h"
#include "common/singleton.h"
#include "core/libraries/error_codes.h" #include "core/libraries/error_codes.h"
#include "core/libraries/kernel/thread_management.h" #include "core/libraries/kernel/thread_management.h"
#include "core/libraries/libs.h" #include "core/libraries/libs.h"
#include "core/tls.h" #include "core/tls.h"
#include "core/linker.h"
#ifdef _WIN64 #ifdef _WIN64
#include <windows.h> #include <windows.h>
#endif #endif
@ -902,7 +904,8 @@ static void cleanup_thread(void* arg) {
static void* run_thread(void* arg) { static void* run_thread(void* arg) {
auto* thread = static_cast<ScePthread>(arg); auto* thread = static_cast<ScePthread>(arg);
Common::SetCurrentThreadName(thread->name.c_str()); Common::SetCurrentThreadName(thread->name.c_str());
Core::SetTLSStorage(0, 0); auto* linker = Common::Singleton<Core::Linker>::Instance();
linker->InitTlsForThread();
void* ret = nullptr; void* ret = nullptr;
g_pthread_self = thread; g_pthread_self = thread;
pthread_cleanup_push(cleanup_thread, thread); pthread_cleanup_push(cleanup_thread, thread);
@ -1109,6 +1112,16 @@ int PS4_SYSV_ABI scePthreadEqual(ScePthread thread1, ScePthread thread2) {
return (thread1 == thread2 ? 1 : 0); return (thread1 == thread2 ? 1 : 0);
} }
struct TlsIndex {
u64 ti_module;
u64 ti_offset;
};
void* PS4_SYSV_ABI __tls_get_addr(TlsIndex* index) {
auto* linker = Common::Singleton<Core::Linker>::Instance();
return linker->TlsGetAddr(index->ti_module, index->ti_offset);
}
void pthreadSymbolsRegister(Core::Loader::SymbolsResolver* sym) { void pthreadSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
LIB_FUNCTION("4+h9EzwKF4I", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetschedpolicy); LIB_FUNCTION("4+h9EzwKF4I", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetschedpolicy);
LIB_FUNCTION("-Wreprtu0Qs", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetdetachstate); LIB_FUNCTION("-Wreprtu0Qs", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetdetachstate);
@ -1121,10 +1134,12 @@ void pthreadSymbolsRegister(Core::Loader::SymbolsResolver* sym) {
LIB_FUNCTION("aI+OeCz8xrQ", "libkernel", 1, "libkernel", 1, 1, scePthreadSelf); LIB_FUNCTION("aI+OeCz8xrQ", "libkernel", 1, "libkernel", 1, 1, scePthreadSelf);
LIB_FUNCTION("EotR8a3ASf4", "libkernel", 1, "libkernel", 1, 1, pthread_self); LIB_FUNCTION("EotR8a3ASf4", "libkernel", 1, "libkernel", 1, 1, pthread_self);
LIB_FUNCTION("EotR8a3ASf4", "libScePosix", 1, "libkernel", 1, 1, pthread_self);
LIB_FUNCTION("3qxgM4ezETA", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetaffinity); LIB_FUNCTION("3qxgM4ezETA", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetaffinity);
LIB_FUNCTION("8+s5BzZjxSg", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGetaffinity); LIB_FUNCTION("8+s5BzZjxSg", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGetaffinity);
LIB_FUNCTION("x1X76arYMxU", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGet); LIB_FUNCTION("x1X76arYMxU", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrGet);
LIB_FUNCTION("UTXzJbWhhTE", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetstacksize); LIB_FUNCTION("UTXzJbWhhTE", "libkernel", 1, "libkernel", 1, 1, scePthreadAttrSetstacksize);
LIB_FUNCTION("vNe1w4diLCs", "libkernel", 1, "libkernel", 1, 1, __tls_get_addr);
LIB_FUNCTION("bt3CTBKmGyI", "libkernel", 1, "libkernel", 1, 1, scePthreadSetaffinity); LIB_FUNCTION("bt3CTBKmGyI", "libkernel", 1, "libkernel", 1, 1, scePthreadSetaffinity);
LIB_FUNCTION("6UgtwV+0zb4", "libkernel", 1, "libkernel", 1, 1, scePthreadCreate); LIB_FUNCTION("6UgtwV+0zb4", "libkernel", 1, "libkernel", 1, 1, scePthreadCreate);

15
src/core/libraries/kernel/time_management.cpp
View File

@ -1,9 +1,9 @@
// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project // SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later // SPDX-License-Identifier: GPL-2.0-or-later
#include <time.h>
#include <thread>
#include <chrono> #include <chrono>
#include <thread>
#include <time.h>
#include "common/native_clock.h" #include "common/native_clock.h"
#include "core/libraries/kernel/time_management.h" #include "core/libraries/kernel/time_management.h"
#include "core/libraries/libs.h" #include "core/libraries/libs.h"
@ -37,9 +37,7 @@ u64 PS4_SYSV_ABI sceKernelReadTsc() {
return clock->GetUptime(); return clock->GetUptime();
} }
int PS4_SYSV_ABI sceKernelGettimeofday(SceKernelTimeval *tp) { int PS4_SYSV_ABI sceKernelGettimeofday(SceKernelTimeval* tp) {}
}
#if defined(_MSC_VER) || defined(_MSC_EXTENSIONS) #if defined(_MSC_VER) || defined(_MSC_EXTENSIONS)
#define DELTA_EPOCH_IN_MICROSECS 11644473600000000Ui64 #define DELTA_EPOCH_IN_MICROSECS 11644473600000000Ui64
@ -47,8 +45,7 @@ int PS4_SYSV_ABI sceKernelGettimeofday(SceKernelTimeval *tp) {
#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL #define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
#endif #endif
struct timezone struct timezone {
{
int tz_minuteswest; /* minutes W of Greenwich */ int tz_minuteswest; /* minutes W of Greenwich */
int tz_dsttime; /* type of dst correction */ int tz_dsttime; /* type of dst correction */
}; };
@ -77,8 +74,8 @@ int PS4_SYSV_ABI gettimeofday(struct timeval * tp, struct timezone * tzp) {
ularge.HighPart = file_time.dwHighDateTime; ularge.HighPart = file_time.dwHighDateTime;
tp->tv_sec = (long)((ularge.QuadPart - epoch) / FILETIME_UNITS_PER_SEC); tp->tv_sec = (long)((ularge.QuadPart - epoch) / FILETIME_UNITS_PER_SEC);
tp->tv_usec = (long) (((ularge.QuadPart - epoch) % FILETIME_UNITS_PER_SEC) tp->tv_usec =
/ FILETIME_UNITS_PER_USEC); (long)(((ularge.QuadPart - epoch) % FILETIME_UNITS_PER_SEC) / FILETIME_UNITS_PER_USEC);
return 0; return 0;
} }

704
src/core/linker.cpp
View File

@ -1,9 +1,8 @@
// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project // SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later // SPDX-License-Identifier: GPL-2.0-or-later
#include <Zydis/Zydis.h> #include "common/assert.h"
#include <common/assert.h> #include "common/alignment.h"
#include <xbyak/xbyak.h>
#include "common/config.h" #include "common/config.h"
#include "common/logging/log.h" #include "common/logging/log.h"
#include "common/path_util.h" #include "common/path_util.h"
@ -12,468 +11,138 @@
#include "core/aerolib/aerolib.h" #include "core/aerolib/aerolib.h"
#include "core/aerolib/stubs.h" #include "core/aerolib/stubs.h"
#include "core/libraries/kernel/thread_management.h" #include "core/libraries/kernel/thread_management.h"
#include "core/libraries/kernel/memory_management.h"
#include "core/linker.h" #include "core/linker.h"
#include "core/tls.h" #include "core/tls.h"
#include "core/virtual_memory.h" #include "core/virtual_memory.h"
#include <windows.h>
namespace Core { namespace Core {
static u64 LoadAddress = SYSTEM_RESERVED + CODE_BASE_OFFSET; using ExitFunc = PS4_SYSV_ABI void (*)();
static constexpr u64 CODE_BASE_INCR = 0x010000000u;
static u64 GetAlignedSize(const elf_program_header& phdr) { static PS4_SYSV_ABI void ProgramExitFunc() {
return (phdr.p_align != 0 ? (phdr.p_memsz + (phdr.p_align - 1)) & ~(phdr.p_align - 1) fmt::print("exit function called\n");
: phdr.p_memsz);
} }
static u64 CalculateBaseSize(const elf_header& ehdr, std::span<const elf_program_header> phdr) { static void RunMainEntry(VAddr addr, EntryParams* params, ExitFunc exit_func) {
u64 base_size = 0; // reinterpret_cast<entry_func_t>(addr)(params, exit_func); // can't be used, stack has to have
for (u16 i = 0; i < ehdr.e_phnum; i++) { // a specific layout
if (phdr[i].p_memsz != 0 && (phdr[i].p_type == PT_LOAD || phdr[i].p_type == PT_SCE_RELRO)) { asm volatile("andq $-16, %%rsp\n" // Align to 16 bytes
u64 last_addr = phdr[i].p_vaddr + GetAlignedSize(phdr[i]); "subq $8, %%rsp\n" // videoout_basic expects the stack to be misaligned
if (last_addr > base_size) {
base_size = last_addr;
}
}
}
return base_size;
}
static std::string EncodeId(u64 nVal) { // Kernel also pushes some more things here during process init
std::string enc; // at least: environment, auxv, possibly other things
const char pCodes[] = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+-";
if (nVal < 0x40u) { "pushq 8(%1)\n" // copy EntryParams to top of stack like the kernel does
enc += pCodes[nVal]; "pushq 0(%1)\n" // OpenOrbis expects to find it there
} else {
if (nVal < 0x1000u) { "movq %1, %%rdi\n" // also pass params and exit func
enc += pCodes[static_cast<u16>(nVal >> 6u) & 0x3fu]; "movq %2, %%rsi\n" // as before
enc += pCodes[nVal & 0x3fu];
} else { "jmp *%0\n" // can't use call here, as that would mangle the prepared stack.
enc += pCodes[static_cast<u16>(nVal >> 12u) & 0x3fu]; // there's no coming back
enc += pCodes[static_cast<u16>(nVal >> 6u) & 0x3fu]; :
enc += pCodes[nVal & 0x3fu]; : "r"(addr), "r"(params), "r"(exit_func)
} : "rax", "rsi", "rdi");
}
return enc;
} }
Linker::Linker() = default; Linker::Linker() = default;
Linker::~Linker() = default; Linker::~Linker() = default;
void Linker::Execute() {
if (Config::debugDump()) {
DebugDump();
}
// Calculate static TLS size.
static constexpr size_t StOff = 0x80; // TODO: What is this offset?
static_tls_size = std::ranges::fold_left(m_modules, StOff, [&](u32 size, auto& module) {
const size_t new_size = size + module->tls.image_size;
module->tls.distance_from_fs = new_size;
return new_size;
});
Common::SetCurrentThreadName("GAME_MainThread");
Libraries::Kernel::pthreadInitSelfMainThread();
// Init primary thread TLS.
InitTlsForThread(true);
// Relocate all modules
for (const auto& m : m_modules) {
Relocate(m.get());
}
// Start shared library modules
for (auto& m : m_modules) {
if (m->IsSharedLib()) {
m->Start(0, nullptr, nullptr);
}
}
// Start main module.
EntryParams p{};
p.argc = 1;
p.argv[0] = "eboot.bin";
for (auto& m : m_modules) {
if (!m->IsSharedLib()) {
RunMainEntry(m->GetEntryAddress(), &p, ProgramExitFunc);
}
}
}
Module* Linker::LoadModule(const std::filesystem::path& elf_name) { Module* Linker::LoadModule(const std::filesystem::path& elf_name) {
std::scoped_lock lock{m_mutex};
if (!std::filesystem::exists(elf_name)) { if (!std::filesystem::exists(elf_name)) {
LOG_ERROR(Core_Linker, "Provided module {} does not exist", elf_name.string()); LOG_ERROR(Core_Linker, "Provided file {} does not exist", elf_name.string());
return nullptr; return nullptr;
} }
auto& m = m_modules.emplace_back(); auto module = std::make_unique<Module>(elf_name);
m = std::make_unique<Module>(); if (!module->IsValid()) {
m->elf.Open(elf_name); LOG_ERROR(Core_Linker, "Provided file {} is not valid ELF file", elf_name.string());
m->file_name = std::filesystem::path(elf_name).filename().string();
if (m->elf.IsElfFile()) {
LoadModuleToMemory(m.get());
LoadDynamicInfo(m.get());
LoadSymbols(m.get());
} else {
m_modules.pop_back();
return nullptr; // It is not a valid elf file //TODO check it why!
}
return m.get();
}
void Linker::LoadModuleToMemory(Module* m) {
// Retrieve elf header and program header
const auto elf_header = m->elf.GetElfHeader();
const auto elf_pheader = m->elf.GetProgramHeader();
u64 base_size = CalculateBaseSize(elf_header, elf_pheader);
m->aligned_base_size = (base_size & ~(static_cast<u64>(0x1000) - 1)) +
0x1000; // align base size to 0x1000 block size (TODO is that the default
// block size or it can be changed?
static constexpr u64 TrampolineSize = 8_MB;
m->base_virtual_addr =
VirtualMemory::memory_alloc(LoadAddress, m->aligned_base_size + TrampolineSize,
VirtualMemory::MemoryMode::ExecuteReadWrite);
LoadAddress += CODE_BASE_INCR * (1 + m->aligned_base_size / CODE_BASE_INCR);
void* trampoline_addr = reinterpret_cast<void*>(m->base_virtual_addr + m->aligned_base_size);
Xbyak::CodeGenerator c(TrampolineSize, trampoline_addr);
LOG_INFO(Core_Linker, "======== Load Module to Memory ========");
LOG_INFO(Core_Linker, "base_virtual_addr ......: {:#018x}", m->base_virtual_addr);
LOG_INFO(Core_Linker, "base_size ..............: {:#018x}", base_size);
LOG_INFO(Core_Linker, "aligned_base_size ......: {:#018x}", m->aligned_base_size);
for (u16 i = 0; i < elf_header.e_phnum; i++) {
switch (elf_pheader[i].p_type) {
case PT_LOAD:
case PT_SCE_RELRO:
if (elf_pheader[i].p_memsz != 0) {
u64 segment_addr = elf_pheader[i].p_vaddr + m->base_virtual_addr;
u64 segment_file_size = elf_pheader[i].p_filesz;
u64 segment_memory_size = GetAlignedSize(elf_pheader[i]);
auto segment_mode = m->elf.ElfPheaderFlagsStr(elf_pheader[i].p_flags);
LOG_INFO(Core_Linker, "program header = [{}] type = {}", i,
m->elf.ElfPheaderTypeStr(elf_pheader[i].p_type));
LOG_INFO(Core_Linker, "segment_addr ..........: {:#018x}", segment_addr);
LOG_INFO(Core_Linker, "segment_file_size .....: {}", segment_file_size);
LOG_INFO(Core_Linker, "segment_memory_size ...: {}", segment_memory_size);
LOG_INFO(Core_Linker, "segment_mode ..........: {}", segment_mode);
m->elf.LoadSegment(segment_addr, elf_pheader[i].p_offset, segment_file_size);
if (elf_pheader[i].p_flags & PF_EXEC) {
PatchTLS(segment_addr, segment_file_size, c);
}
} else {
LOG_ERROR(Core_Linker, "p_memsz==0 in type {}",
m->elf.ElfPheaderTypeStr(elf_pheader[i].p_type));
}
break;
case PT_DYNAMIC:
if (elf_pheader[i].p_filesz != 0) {
m->m_dynamic.resize(elf_pheader[i].p_filesz);
m->elf.LoadSegment(reinterpret_cast<u64>(m->m_dynamic.data()),
elf_pheader[i].p_offset, elf_pheader[i].p_filesz);
} else {
LOG_ERROR(Core_Linker, "p_filesz==0 in type {}",
m->elf.ElfPheaderTypeStr(elf_pheader[i].p_type));
}
break;
case PT_SCE_DYNLIBDATA:
if (elf_pheader[i].p_filesz != 0) {
m->m_dynamic_data.resize(elf_pheader[i].p_filesz);
m->elf.LoadSegment(reinterpret_cast<u64>(m->m_dynamic_data.data()),
elf_pheader[i].p_offset, elf_pheader[i].p_filesz);
} else {
LOG_ERROR(Core_Linker, "p_filesz==0 in type {}",
m->elf.ElfPheaderTypeStr(elf_pheader[i].p_type));
}
break;
case PT_TLS:
m->tls.image_virtual_addr = elf_pheader[i].p_vaddr + m->base_virtual_addr;
m->tls.image_size = GetAlignedSize(elf_pheader[i]);
LOG_INFO(Core_Linker, "TLS virtual address = {:#x}", m->tls.image_virtual_addr);
LOG_INFO(Core_Linker, "TLS image size = {}", m->tls.image_size);
break;
case PT_SCE_PROCPARAM:
m->proc_param_virtual_addr = elf_pheader[i].p_vaddr + m->base_virtual_addr;
break;
default:
LOG_ERROR(Core_Linker, "Unimplemented type {}",
m->elf.ElfPheaderTypeStr(elf_pheader[i].p_type));
}
}
LOG_INFO(Core_Linker, "program entry addr ..........: {:#018x}",
m->elf.GetElfEntry() + m->base_virtual_addr);
}
void Linker::LoadDynamicInfo(Module* m) {
for (const auto* dyn = reinterpret_cast<elf_dynamic*>(m->m_dynamic.data());
dyn->d_tag != DT_NULL; dyn++) {
switch (dyn->d_tag) {
case DT_SCE_HASH: // Offset of the hash table.
m->dynamic_info.hash_table =
reinterpret_cast<void*>(m->m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_HASHSZ: // Size of the hash table
m->dynamic_info.hash_table_size = dyn->d_un.d_val;
break;
case DT_SCE_STRTAB: // Offset of the string table.
m->dynamic_info.str_table =
reinterpret_cast<char*>(m->m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_STRSZ: // Size of the string table.
m->dynamic_info.str_table_size = dyn->d_un.d_val;
break;
case DT_SCE_SYMTAB: // Offset of the symbol table.
m->dynamic_info.symbol_table =
reinterpret_cast<elf_symbol*>(m->m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_SYMTABSZ: // Size of the symbol table.
m->dynamic_info.symbol_table_total_size = dyn->d_un.d_val;
break;
case DT_INIT:
m->dynamic_info.init_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_FINI:
m->dynamic_info.fini_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_SCE_PLTGOT: // Offset of the global offset table.
m->dynamic_info.pltgot_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_SCE_JMPREL: // Offset of the table containing jump slots.
m->dynamic_info.jmp_relocation_table =
reinterpret_cast<elf_relocation*>(m->m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_PLTRELSZ: // Size of the global offset table.
m->dynamic_info.jmp_relocation_table_size = dyn->d_un.d_val;
break;
case DT_SCE_PLTREL: // The type of relocations in the relocation table. Should be DT_RELA
m->dynamic_info.jmp_relocation_type = dyn->d_un.d_val;
if (m->dynamic_info.jmp_relocation_type != DT_RELA) {
LOG_WARNING(Core_Linker, "DT_SCE_PLTREL is NOT DT_RELA should check!");
}
break;
case DT_SCE_RELA: // Offset of the relocation table.
m->dynamic_info.relocation_table =
reinterpret_cast<elf_relocation*>(m->m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_RELASZ: // Size of the relocation table.
m->dynamic_info.relocation_table_size = dyn->d_un.d_val;
break;
case DT_SCE_RELAENT: // The size of relocation table entries.
m->dynamic_info.relocation_table_entries_size = dyn->d_un.d_val;
if (m->dynamic_info.relocation_table_entries_size !=
0x18) // this value should always be 0x18
{
LOG_WARNING(Core_Linker, "DT_SCE_RELAENT is NOT 0x18 should check!");
}
break;
case DT_INIT_ARRAY: // Address of the array of pointers to initialization functions
m->dynamic_info.init_array_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_FINI_ARRAY: // Address of the array of pointers to termination functions
m->dynamic_info.fini_array_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_INIT_ARRAYSZ: // Size in bytes of the array of initialization functions
m->dynamic_info.init_array_size = dyn->d_un.d_val;
break;
case DT_FINI_ARRAYSZ: // Size in bytes of the array of terminationfunctions
m->dynamic_info.fini_array_size = dyn->d_un.d_val;
break;
case DT_PREINIT_ARRAY: // Address of the array of pointers to pre - initialization functions
m->dynamic_info.preinit_array_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_PREINIT_ARRAYSZ: // Size in bytes of the array of pre - initialization functions
m->dynamic_info.preinit_array_size = dyn->d_un.d_val;
break;
case DT_SCE_SYMENT: // The size of symbol table entries
m->dynamic_info.symbol_table_entries_size = dyn->d_un.d_val;
if (m->dynamic_info.symbol_table_entries_size !=
0x18) // this value should always be 0x18
{
LOG_WARNING(Core_Linker, "DT_SCE_SYMENT is NOT 0x18 should check!");
}
break;
case DT_DEBUG:
m->dynamic_info.debug = dyn->d_un.d_val;
break;
case DT_TEXTREL:
m->dynamic_info.textrel = dyn->d_un.d_val;
break;
case DT_FLAGS:
m->dynamic_info.flags = dyn->d_un.d_val;
if (m->dynamic_info.flags != 0x04) // this value should always be DF_TEXTREL (0x04)
{
LOG_WARNING(Core_Linker, "DT_FLAGS is NOT 0x04 should check!");
}
break;
case DT_NEEDED: // Offset of the library string in the string table to be linked in.
if (m->dynamic_info.str_table !=
nullptr) // in theory this should already be filled from about just make a test case
{
m->dynamic_info.needed.push_back(m->dynamic_info.str_table + dyn->d_un.d_val);
} else {
LOG_ERROR(Core_Linker, "DT_NEEDED str table is not loaded should check!");
}
break;
case DT_SCE_NEEDED_MODULE: {
ModuleInfo info{};
info.value = dyn->d_un.d_val;
info.name = m->dynamic_info.str_table + info.name_offset;
info.enc_id = EncodeId(info.id);
m->dynamic_info.import_modules.push_back(info);
} break;
case DT_SCE_IMPORT_LIB: {
LibraryInfo info{};
info.value = dyn->d_un.d_val;
info.name = m->dynamic_info.str_table + info.name_offset;
info.enc_id = EncodeId(info.id);
m->dynamic_info.import_libs.push_back(info);
} break;
case DT_SCE_FINGERPRINT:
// The fingerprint is a 24 byte (0x18) size buffer that contains a unique identifier for
// 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);
break;
case DT_SCE_IMPORT_LIB_ATTR:
// The upper 32-bits should contain the module index multiplied by 0x10000. The lower
// 32-bits should be a constant 0x9.
LOG_INFO(Core_Linker, "unsupported DT_SCE_IMPORT_LIB_ATTR value = ......: {:#018x}",
dyn->d_un.d_val);
break;
case DT_SCE_ORIGINAL_FILENAME:
m->dynamic_info.filename = m->dynamic_info.str_table + dyn->d_un.d_val;
break;
case DT_SCE_MODULE_INFO: // probably only useable in shared modules
{
ModuleInfo info{};
info.value = dyn->d_un.d_val;
info.name = m->dynamic_info.str_table + info.name_offset;
info.enc_id = EncodeId(info.id);
m->dynamic_info.export_modules.push_back(info);
} break;
case DT_SCE_MODULE_ATTR:
// TODO?
LOG_INFO(Core_Linker, "unsupported DT_SCE_MODULE_ATTR value = ..........: {:#018x}",
dyn->d_un.d_val);
break;
case DT_SCE_EXPORT_LIB: {
LibraryInfo info{};
info.value = dyn->d_un.d_val;
info.name = m->dynamic_info.str_table + info.name_offset;
info.enc_id = EncodeId(info.id);
m->dynamic_info.export_libs.push_back(info);
} break;
default:
LOG_INFO(Core_Linker, "unsupported dynamic tag ..........: {:#018x}", dyn->d_tag);
}
}
}
const ModuleInfo* Linker::FindModule(const Module& m, const std::string& id) {
const auto& import_modules = m.dynamic_info.import_modules;
int index = 0;
for (const auto& mod : import_modules) {
if (mod.enc_id.compare(id) == 0) {
return &import_modules.at(index);
}
index++;
}
const auto& export_modules = m.dynamic_info.export_modules;
index = 0;
for (const auto& mod : export_modules) {
if (mod.enc_id.compare(id) == 0) {
return &export_modules.at(index);
}
index++;
}
return nullptr; return nullptr;
} }
const LibraryInfo* Linker::FindLibrary(const Module& m, const std::string& id) { if (!module->IsSharedLib()) {
const auto& import_libs = m.dynamic_info.import_libs; main_proc_param_addr = module->GetProcParam();
int index = 0;
for (const auto& lib : import_libs) {
if (lib.enc_id.compare(id) == 0) {
return &import_libs.at(index);
}
index++;
}
const auto& export_libs = m.dynamic_info.export_libs;
index = 0;
for (const auto& lib : export_libs) {
if (lib.enc_id.compare(id) == 0) {
return &export_libs.at(index);
}
index++;
}
return nullptr;
} }
void Linker::LoadSymbols(Module* m) { return m_modules.emplace_back(std::move(module)).get();
const auto symbol_database = [this, m](Loader::SymbolsResolver* symbol, bool export_func) {
if (m->dynamic_info.symbol_table == nullptr || m->dynamic_info.str_table == nullptr ||
m->dynamic_info.symbol_table_total_size == 0) {
LOG_INFO(Core_Linker, "Symbol table not found!");
return;
}
for (auto* sym = m->dynamic_info.symbol_table;
reinterpret_cast<u8*>(sym) < reinterpret_cast<u8*>(m->dynamic_info.symbol_table) +
m->dynamic_info.symbol_table_total_size;
sym++) {
std::string id = std::string(m->dynamic_info.str_table + sym->st_name);
auto bind = sym->GetBind();
auto type = sym->GetType();
auto visibility = sym->GetVisibility();
const auto ids = Common::SplitString(id, '#');
if (ids.size() == 3) {
const auto* library = FindLibrary(*m, ids.at(1));
const auto* module = FindModule(*m, ids.at(2));
ASSERT_MSG(library && module, "Unable to find library and module");
if ((bind == STB_GLOBAL || bind == STB_WEAK) &&
(type == STT_FUN || type == STT_OBJECT) &&
export_func == (sym->st_value != 0)) {
std::string nidName = "";
auto aeronid = AeroLib::FindByNid(ids.at(0).c_str());
if (aeronid != nullptr) {
nidName = aeronid->name;
} else {
nidName = "UNK";
}
Loader::SymbolResolver sym_r{};
sym_r.name = ids.at(0);
sym_r.nidName = nidName;
sym_r.library = library->name;
sym_r.library_version = library->version;
sym_r.module = module->name;
sym_r.module_version_major = module->version_major;
sym_r.module_version_minor = module->version_minor;
switch (type) {
case STT_NOTYPE:
sym_r.type = Loader::SymbolType::NoType;
break;
case STT_FUN:
sym_r.type = Loader::SymbolType::Function;
break;
case STT_OBJECT:
sym_r.type = Loader::SymbolType::Object;
break;
default:
sym_r.type = Loader::SymbolType::Unknown;
break;
}
symbol->AddSymbol(sym_r,
(export_func ? sym->st_value + m->base_virtual_addr : 0));
}
}
}
};
symbol_database(&m->export_sym, true);
symbol_database(&m->import_sym, false);
} }
void Linker::Relocate(Module* m) { void Linker::Relocate(Module* module) {
const auto relocate = [this](u32 idx, elf_relocation* rel, Module* m, bool isJmpRel) { module->ForEachRelocation([&](elf_relocation* rel, bool isJmpRel) {
auto type = rel->GetType(); auto type = rel->GetType();
auto symbol = rel->GetSymbol(); auto symbol = rel->GetSymbol();
auto addend = rel->rel_addend; auto addend = rel->rel_addend;
auto* symbolsTlb = m->dynamic_info.symbol_table; auto* symbol_table = module->dynamic_info.symbol_table;
auto* namesTlb = m->dynamic_info.str_table; auto* namesTlb = module->dynamic_info.str_table;
const VAddr rel_base_virtual_addr = module->GetBaseAddress();
const VAddr rel_virtual_addr = rel_base_virtual_addr + rel->rel_offset;
bool rel_is_resolved = false;
u64 rel_value = 0; u64 rel_value = 0;
u64 rel_base_virtual_addr = m->base_virtual_addr;
u64 rel_virtual_addr = m->base_virtual_addr + rel->rel_offset;
bool rel_isResolved = false;
Loader::SymbolType rel_sym_type = Loader::SymbolType::Unknown; Loader::SymbolType rel_sym_type = Loader::SymbolType::Unknown;
std::string rel_name; std::string rel_name;
switch (type) { switch (type) {
case R_X86_64_RELATIVE: case R_X86_64_RELATIVE:
rel_value = rel_base_virtual_addr + addend; rel_value = rel_base_virtual_addr + addend;
rel_isResolved = true; rel_is_resolved = true;
break; break;
case R_X86_64_DTPMOD64: case R_X86_64_DTPMOD64:
rel_value = reinterpret_cast<uint64_t>(m); rel_value = reinterpret_cast<uint64_t>(module);
rel_isResolved = true; rel_is_resolved = true;
rel_sym_type = Loader::SymbolType::Tls; rel_sym_type = Loader::SymbolType::Tls;
break; break;
case R_X86_64_GLOB_DAT: case R_X86_64_GLOB_DAT:
case R_X86_64_JUMP_SLOT: case R_X86_64_JUMP_SLOT:
addend = 0; addend = 0;
case R_X86_64_64: { case R_X86_64_64: {
auto sym = symbolsTlb[symbol]; auto sym = symbol_table[symbol];
auto sym_bind = sym.GetBind(); auto sym_bind = sym.GetBind();
auto sym_type = sym.GetType(); auto sym_type = sym.GetType();
auto sym_visibility = sym.GetVisibility(); auto sym_visibility = sym.GetVisibility();
@ -492,10 +161,11 @@ void Linker::Relocate(Module* m) {
default: default:
ASSERT_MSG(0, "unknown symbol type {}", sym_type); ASSERT_MSG(0, "unknown symbol type {}", sym_type);
} }
if (sym_visibility != 0) // should be zero log if else
{ if (sym_visibility != 0) {
LOG_INFO(Core_Linker, "symbol visilibity !=0"); LOG_INFO(Core_Linker, "symbol visilibity !=0");
} }
switch (sym_bind) { switch (sym_bind) {
case STB_LOCAL: case STB_LOCAL:
symbol_vitrual_addr = rel_base_virtual_addr + sym.st_value; symbol_vitrual_addr = rel_base_virtual_addr + sym.st_value;
@ -503,60 +173,36 @@ void Linker::Relocate(Module* m) {
case STB_GLOBAL: case STB_GLOBAL:
case STB_WEAK: { case STB_WEAK: {
rel_name = namesTlb + sym.st_name; rel_name = namesTlb + sym.st_name;
Resolve(rel_name, rel_sym_type, m, &symrec); Resolve(rel_name, rel_sym_type, module, &symrec);
symbol_vitrual_addr = symrec.virtual_address; symbol_vitrual_addr = symrec.virtual_address;
} break; break;
}
default: default:
ASSERT_MSG(0, "unknown bind type {}", sym_bind); ASSERT_MSG(0, "unknown bind type {}", sym_bind);
} }
rel_isResolved = (symbol_vitrual_addr != 0); rel_is_resolved = (symbol_vitrual_addr != 0);
rel_value = (rel_isResolved ? symbol_vitrual_addr + addend : 0); rel_value = (rel_is_resolved ? symbol_vitrual_addr + addend : 0);
rel_name = symrec.name; rel_name = symrec.name;
} break; break;
}
default: default:
LOG_INFO(Core_Linker, "UNK type {:#010x} rel symbol : {:#010x}", type, symbol); LOG_INFO(Core_Linker, "UNK type {:#010x} rel symbol : {:#010x}", type, symbol);
} }
if (rel_isResolved) { if (rel_is_resolved) {
VirtualMemory::memory_patch(rel_virtual_addr, rel_value); VirtualMemory::memory_patch(rel_virtual_addr, rel_value);
} else { } else {
LOG_INFO(Core_Linker, "function not patched! {}", rel_name); LOG_INFO(Core_Linker, "function not patched! {}", rel_name);
} }
}; });
u32 idx = 0;
for (auto* rel = m->dynamic_info.relocation_table;
reinterpret_cast<u8*>(rel) < reinterpret_cast<u8*>(m->dynamic_info.relocation_table) +
m->dynamic_info.relocation_table_size;
rel++, idx++) {
relocate(idx, rel, m, false);
}
idx = 0;
for (auto* rel = m->dynamic_info.jmp_relocation_table;
reinterpret_cast<u8*>(rel) < reinterpret_cast<u8*>(m->dynamic_info.jmp_relocation_table) +
m->dynamic_info.jmp_relocation_table_size;
rel++, idx++) {
relocate(idx, rel, m, true);
}
} }
template <typename T> const Module* Linker::FindExportedModule(const ModuleInfo& module, const LibraryInfo& library) {
bool contains(const std::vector<T>& vecObj, const T& element) { const auto it = std::ranges::find_if(m_modules, [&](const auto& m) {
auto it = std::find(vecObj.begin(), vecObj.end(), element); return std::ranges::contains(m->GetExportLibs(), library) &&
return it != vecObj.end(); std::ranges::contains(m->GetExportModules(), module);
} });
return it == m_modules.end() ? nullptr : it->get();
Module* Linker::FindExportedModule(const ModuleInfo& module, const LibraryInfo& library) {
// std::scoped_lock lock{m_mutex};
for (auto& m : m_modules) {
const auto& export_libs = m->dynamic_info.export_libs;
const auto& export_modules = m->dynamic_info.export_modules;
if (contains(export_libs, library) && contains(export_modules, module)) {
return m.get();
}
}
return nullptr;
} }
void Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Module* m, void Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Module* m,
@ -569,8 +215,8 @@ void Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Modul
return; return;
} }
const auto* library = FindLibrary(*m, ids.at(1)); const LibraryInfo* library = m->FindLibrary(ids[1]);
const auto* module = FindModule(*m, ids.at(2)); const ModuleInfo* module = m->FindModule(ids[2]);
ASSERT_MSG(library && module, "Unable to find library and module"); ASSERT_MSG(library && module, "Unable to find library and module");
Loader::SymbolResolver sr{}; Loader::SymbolResolver sr{};
@ -585,8 +231,8 @@ void Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Modul
const auto* record = m_hle_symbols.FindSymbol(sr); const auto* record = m_hle_symbols.FindSymbol(sr);
if (!record) { if (!record) {
// Check if it an export function // Check if it an export function
if (auto* p = FindExportedModule(*module, *library); const auto* p = FindExportedModule(*module, *library);
p && p->export_sym.GetSize() > 0) { if (p && p->export_sym.GetSize() > 0) {
record = p->export_sym.FindSymbol(sr); record = p->export_sym.FindSymbol(sr);
} }
} }
@ -607,102 +253,72 @@ void Linker::Resolve(const std::string& name, Loader::SymbolType sym_type, Modul
return_info->name, library->name, module->name); return_info->name, library->name, module->name);
} }
u64 Linker::GetProcParam() { void* Linker::TlsGetAddr(u64 module_index, u64 offset) {
for (auto& m : m_modules) { ASSERT_MSG(dtv_table[0].counter == dtv_generation_counter,
if (!m->elf.IsSharedLib()) { "Reallocation of DTV table is not supported");
return m->proc_param_virtual_addr;
} void* module = dtv_table[module_index - 1].pointer;
} ASSERT_MSG(module, "DTV allocation is not supported");
return 0; return module;
} }
using exit_func_t = PS4_SYSV_ABI void (*)(); void Linker::InitTlsForThread(bool is_primary) {
using entry_func_t = PS4_SYSV_ABI void (*)(EntryParams* params, exit_func_t atexit_func); static constexpr size_t TcbSize = 0x40;
using module_ini_func_t = PS4_SYSV_ABI int (*)(size_t args, const void* argp, ModuleFunc func); static constexpr size_t TlsAllocAlign = 0x20;
const size_t total_tls_size = Common::AlignUp(static_tls_size, TlsAllocAlign) + TcbSize;
static PS4_SYSV_ABI int RunModule(uint64_t addr, size_t args, const void* argp, // The kernel module has a few different paths for TLS allocation.
ModuleFunc func) { // For SDK < 1.7 it allocates both main and secondary thread blocks using libc mspace/malloc.
return reinterpret_cast<module_ini_func_t>(addr)(args, argp, func); // In games compiled with newer SDK, the main thread gets mapped from flexible memory,
} // with addr = 0, so system managed area. Here we will only implement the latter.
void* addr_out{};
int Linker::StartModule(Module* m, size_t args, const void* argp, ModuleFunc func) { if (is_primary) {
LOG_INFO(Core_Linker, "Module started : {}", m->file_name); const size_t tls_aligned = Common::AlignUp(total_tls_size, 16_KB);
return RunModule(m->dynamic_info.init_virtual_addr + m->base_virtual_addr, args, argp, func); const int ret = Libraries::Kernel::sceKernelMapNamedFlexibleMemory(&addr_out, tls_aligned,
} 3, 0, "SceKernelPrimaryTcbTls");
ASSERT_MSG(ret == 0, "Unable to allocate TLS+TCB for the primary thread");
void Linker::StartAllModules() { } else {
std::scoped_lock lock{m_mutex}; if (heap_api_func) {
addr_out = heap_api_func(total_tls_size);
for (auto& m : m_modules) { } else {
if (m->elf.IsSharedLib()) { addr_out = std::malloc(total_tls_size);
StartModule(m.get(), 0, nullptr, nullptr);
} }
} }
// Initialize allocated memory and allocate DTV table.
const u32 num_dtvs = m_modules.size() - 1;
std::memset(addr_out, 0, total_tls_size);
dtv_table.resize(num_dtvs + 2);
// Initialize thread control block
u8* addr = reinterpret_cast<u8*>(addr_out);
Tcb* tcb = reinterpret_cast<Tcb*>(addr + static_tls_size);
tcb->tcb_self = tcb;
tcb->tcb_dtv = dtv_table.data();
// Dtv[0] is the generation counter. libkernel puts their number into dtv[1] (why?)
dtv_table[0].counter = dtv_generation_counter;
dtv_table[1].counter = num_dtvs;
// Copy init images to TLS thread blocks and map them to DTV slots.
for (u32 i = 2; const auto& module : m_modules) {
u8* dest = reinterpret_cast<u8*>(addr + static_tls_size - module->tls.distance_from_fs);
const u8* src = reinterpret_cast<const u8*>(module->tls.image_virtual_addr);
std::memcpy(dest, src, module->tls.init_image_size);
tcb->tcb_dtv[i++].pointer = dest;
} }
static PS4_SYSV_ABI void ProgramExitFunc() { // Set pointer to FS base
fmt::print("exit function called\n"); SetTcbBase(tcb);
}
static void RunMainEntry(u64 addr, EntryParams* params, exit_func_t exit_func) {
// reinterpret_cast<entry_func_t>(addr)(params, exit_func); // can't be used, stack has to have
// a specific layout
asm volatile("andq $-16, %%rsp\n" // Align to 16 bytes
"subq $8, %%rsp\n" // videoout_basic expects the stack to be misaligned
// Kernel also pushes some more things here during process init
// at least: environment, auxv, possibly other things
"pushq 8(%1)\n" // copy EntryParams to top of stack like the kernel does
"pushq 0(%1)\n" // OpenOrbis expects to find it there
"movq %1, %%rdi\n" // also pass params and exit func
"movq %2, %%rsi\n" // as before
"jmp *%0\n" // can't use call here, as that would mangle the prepared stack.
// there's no coming back
:
: "r"(addr), "r"(params), "r"(exit_func)
: "rax", "rsi", "rdi");
}
void Linker::Execute() {
if (Config::debugDump()) {
DebugDump();
}
Common::SetCurrentThreadName("GAME_MainThread");
Libraries::Kernel::pthreadInitSelfMainThread();
// Relocate all modules
for (const auto& m : m_modules) {
Relocate(m.get());
}
StartAllModules();
EntryParams p{};
p.argc = 1;
p.argv[0] = "eboot.bin"; // hmm should be ok?
for (auto& m : m_modules) {
if (m->elf.IsSharedLib()) {
continue;
}
if (m->tls.image_virtual_addr != 0) {
SetTLSStorage(m->tls.image_virtual_addr, m->tls.image_size);
}
RunMainEntry(m->elf.GetElfEntry() + m->base_virtual_addr, &p, ProgramExitFunc);
}
} }
void Linker::DebugDump() { void Linker::DebugDump() {
std::scoped_lock lock{m_mutex};
const auto& log_dir = Common::FS::GetUserPath(Common::FS::PathType::LogDir); const auto& log_dir = Common::FS::GetUserPath(Common::FS::PathType::LogDir);
const std::filesystem::path debug(log_dir / "debugdump"); const std::filesystem::path debug(log_dir / "debugdump");
std::filesystem::create_directory(debug); std::filesystem::create_directory(debug);
for (const auto& m : m_modules) { for (const auto& m : m_modules) {
// TODO make a folder with game id for being more unique? // TODO make a folder with game id for being more unique?
const std::filesystem::path filepath(debug / m.get()->file_name); const std::filesystem::path filepath(debug / m.get()->file.stem());
std::filesystem::create_directory(filepath); std::filesystem::create_directory(filepath);
m.get()->import_sym.DebugDump(filepath / "imports.txt"); m.get()->import_sym.DebugDump(filepath / "imports.txt");
m.get()->export_sym.DebugDump(filepath / "exports.txt"); m.get()->export_sym.DebugDump(filepath / "exports.txt");

143
src/core/linker.h
View File

@ -3,10 +3,8 @@
#pragma once #pragma once
#include <mutex>
#include <vector> #include <vector>
#include "core/loader/elf.h" #include "core/module.h"
#include "core/loader/symbols_resolver.h"
namespace Core { namespace Core {
@ -19,133 +17,60 @@ struct EntryParams {
const char* argv[3]; const char* argv[3];
}; };
struct ModuleInfo { union DtvEntry {
bool operator==(const ModuleInfo& other) const {
return version_major == other.version_major && version_minor == other.version_minor &&
name == other.name;
}
std::string name;
union {
u64 value;
struct { struct {
u32 name_offset; size_t counter;
u8 version_minor;
u8 version_major;
u16 id;
}; };
}; void* pointer;
std::string enc_id;
}; };
struct LibraryInfo { struct Tcb {
bool operator==(const LibraryInfo& other) const { Tcb* tcb_self;
return version == other.version && name == other.name; DtvEntry* tcb_dtv;
} void* tcb_thread;
std::string name;
union {
u64 value;
struct {
u32 name_offset;
u16 version;
u16 id;
};
};
std::string enc_id;
}; };
struct PS4ThreadLocal { using HeapApiFunc = PS4_SYSV_ABI void*(*)(size_t);
u64 image_virtual_addr = 0;
u64 image_size = 0;
};
struct DynamicModuleInfo {
void* hash_table = nullptr;
u64 hash_table_size = 0;
char* str_table = nullptr;
u64 str_table_size = 0;
elf_symbol* symbol_table = nullptr;
u64 symbol_table_total_size = 0;
u64 symbol_table_entries_size = 0;
u64 init_virtual_addr = 0;
u64 fini_virtual_addr = 0;
u64 pltgot_virtual_addr = 0;
u64 init_array_virtual_addr = 0;
u64 fini_array_virtual_addr = 0;
u64 preinit_array_virtual_addr = 0;
u64 init_array_size = 0;
u64 fini_array_size = 0;
u64 preinit_array_size = 0;
elf_relocation* jmp_relocation_table = nullptr;
u64 jmp_relocation_table_size = 0;
s64 jmp_relocation_type = 0;
elf_relocation* relocation_table = nullptr;
u64 relocation_table_size = 0;
u64 relocation_table_entries_size = 0;
u64 debug = 0;
u64 textrel = 0;
u64 flags = 0;
std::vector<const char*> needed;
std::vector<ModuleInfo> import_modules;
std::vector<ModuleInfo> export_modules;
std::vector<LibraryInfo> import_libs;
std::vector<LibraryInfo> export_libs;
std::string filename; // Filename with absolute path
};
// This struct keeps neccesary info about loaded modules. Main executeable is included too as well
struct Module {
u64 aligned_base_size = 0;
VAddr base_virtual_addr = 0;
VAddr proc_param_virtual_addr = 0;
std::string file_name;
Loader::Elf elf;
std::vector<u8> m_dynamic;
std::vector<u8> m_dynamic_data;
DynamicModuleInfo dynamic_info{};
Loader::SymbolsResolver export_sym;
Loader::SymbolsResolver import_sym;
PS4ThreadLocal tls;
};
using ModuleFunc = int (*)(size_t, const void*);
class Linker { class Linker {
public: public:
explicit Linker(); explicit Linker();
~Linker(); ~Linker();
Module* LoadModule(const std::filesystem::path& elf_name); Loader::SymbolsResolver& GetHLESymbols() {
void LoadModuleToMemory(Module* m);
void LoadDynamicInfo(Module* m);
void LoadSymbols(Module* m);
Loader::SymbolsResolver& getHLESymbols() {
return m_hle_symbols; return m_hle_symbols;
} }
void Relocate(Module* m);
void Resolve(const std::string& name, Loader::SymbolType Symtype, Module* m, VAddr GetProcParam() const {
Loader::SymbolRecord* return_info); return main_proc_param_addr;
}
void SetHeapApiFunc(void* func) {
heap_api_func = *reinterpret_cast<HeapApiFunc*>(func);
}
void* TlsGetAddr(u64 module_index, u64 offset);
void InitTlsForThread(bool is_primary = false);
Module* LoadModule(const std::filesystem::path& elf_name);
void Relocate(Module* module);
void Resolve(const std::string& name, Loader::SymbolType type,
Module* module, Loader::SymbolRecord* return_info);
void Execute(); void Execute();
void DebugDump(); void DebugDump();
u64 GetProcParam();
private: private:
const ModuleInfo* FindModule(const Module& m, const std::string& id); const Module* FindExportedModule(const ModuleInfo& m, const LibraryInfo& l);
const LibraryInfo* FindLibrary(const Module& program, const std::string& id); void InitTls();
Module* FindExportedModule(const ModuleInfo& m, const LibraryInfo& l);
int StartModule(Module* m, size_t args, const void* argp, ModuleFunc func);
void StartAllModules();
std::vector<DtvEntry> dtv_table;
u32 dtv_generation_counter{1};
size_t static_tls_size{};
HeapApiFunc heap_api_func{};
std::vector<std::unique_ptr<Module>> m_modules; std::vector<std::unique_ptr<Module>> m_modules;
Loader::SymbolsResolver m_hle_symbols{}; Loader::SymbolsResolver m_hle_symbols{};
std::mutex m_mutex; VAddr main_proc_param_addr{};
}; };
} // namespace Core } // namespace Core

6
src/core/loader/elf.cpp
View File

@ -445,7 +445,7 @@ std::string Elf::ElfHeaderStr() {
return header; return header;
} }
std::string Elf::ElfPheaderTypeStr(u32 type) { std::string_view Elf::ElfPheaderTypeStr(u32 type) {
switch (type) { switch (type) {
case PT_NULL: case PT_NULL:
return "Null"; return "Null";
@ -538,8 +538,4 @@ void Elf::LoadSegment(u64 virtual_addr, u64 file_offset, u64 size) {
UNREACHABLE(); UNREACHABLE();
} }
bool Elf::IsSharedLib() {
return m_elf_header.e_type == ET_SCE_DYNAMIC;
}
} // namespace Core::Loader } // namespace Core::Loader

8
src/core/loader/elf.h
View File

@ -3,7 +3,6 @@
#pragma once #pragma once
#include <cinttypes>
#include <span> #include <span>
#include <string> #include <string>
#include <vector> #include <vector>
@ -482,15 +481,18 @@ public:
return m_elf_header.e_entry; return m_elf_header.e_entry;
} }
[[nodiscard]] bool IsSharedLib() const {
return m_elf_header.e_type == ET_SCE_DYNAMIC;
}
std::string SElfHeaderStr(); std::string SElfHeaderStr();
std::string SELFSegHeader(u16 no); std::string SELFSegHeader(u16 no);
std::string ElfHeaderStr(); std::string ElfHeaderStr();
std::string ElfPHeaderStr(u16 no); std::string ElfPHeaderStr(u16 no);
std::string ElfPheaderTypeStr(u32 type); std::string_view ElfPheaderTypeStr(u32 type);
std::string ElfPheaderFlagsStr(u32 flags); std::string ElfPheaderFlagsStr(u32 flags);
void LoadSegment(u64 virtual_addr, u64 file_offset, u64 size); void LoadSegment(u64 virtual_addr, u64 file_offset, u64 size);
bool IsSharedLib();
private: private:
Common::FS::IOFile m_f{}; Common::FS::IOFile m_f{};

18
src/core/loader/symbols_resolver.cpp
View File

@ -1,6 +1,7 @@
// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project // SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later // SPDX-License-Identifier: GPL-2.0-or-later
#include <fmt/format.h>
#include "common/io_file.h" #include "common/io_file.h"
#include "common/string_util.h" #include "common/string_util.h"
#include "common/types.h" #include "common/types.h"
@ -10,9 +11,7 @@
namespace Core::Loader { namespace Core::Loader {
void SymbolsResolver::AddSymbol(const SymbolResolver& s, u64 virtual_addr) { void SymbolsResolver::AddSymbol(const SymbolResolver& s, u64 virtual_addr) {
SymbolRecord& r = m_symbols.emplace_back(); m_symbols.emplace_back(GenerateName(s), virtual_addr);
r.name = GenerateName(s);
r.virtual_address = virtual_addr;
} }
std::string SymbolsResolver::GenerateName(const SymbolResolver& s) { std::string SymbolsResolver::GenerateName(const SymbolResolver& s) {
@ -37,17 +36,12 @@ void SymbolsResolver::DebugDump(const std::filesystem::path& file_name) {
Common::FS::FileType::TextFile}; Common::FS::FileType::TextFile};
for (const auto& symbol : m_symbols) { for (const auto& symbol : m_symbols) {
const auto ids = Common::SplitString(symbol.name, '#'); const auto ids = Common::SplitString(symbol.name, '#');
std::string nidName = ""; const auto aeronid = AeroLib::FindByNid(ids.at(0).c_str());
auto aeronid = AeroLib::FindByNid(ids.at(0).c_str()); const auto nid_name = aeronid ? aeronid->name : "UNK";
if (aeronid != nullptr) {
nidName = aeronid->name;
} else {
nidName = "UNK";
}
f.WriteString( f.WriteString(
fmt::format("0x{:<20x} {:<16} {:<60} {:<30} {:<2} {:<30} {:<2} {:<2} {:<10}\n", fmt::format("0x{:<20x} {:<16} {:<60} {:<30} {:<2} {:<30} {:<2} {:<2} {:<10}\n",
symbol.virtual_address, ids.at(0), nidName, ids.at(1), ids.at(2), ids.at(3), symbol.virtual_address, ids.at(0), nid_name, ids.at(1), ids.at(2),
ids.at(4), ids.at(5), ids.at(6))); ids.at(3), ids.at(4), ids.at(5), ids.at(6)));
} }
} }

4
src/core/memory.cpp
View File

@ -35,7 +35,7 @@ PAddr MemoryManager::Allocate(PAddr search_start, PAddr search_end, size_t size,
} }
// Align free position // Align free position
free_addr = Common::alignUp(free_addr, alignment); free_addr = Common::AlignUp(free_addr, alignment);
ASSERT(free_addr >= search_start && free_addr + size <= search_end); ASSERT(free_addr >= search_start && free_addr + size <= search_end);
// Add the allocated region to the list and commit its pages. // Add the allocated region to the list and commit its pages.
@ -56,7 +56,7 @@ void MemoryManager::Free(PAddr phys_addr, size_t size) {
int MemoryManager::MapMemory(void** out_addr, VAddr virtual_addr, size_t size, MemoryProt prot, int MemoryManager::MapMemory(void** out_addr, VAddr virtual_addr, size_t size, MemoryProt prot,
MemoryMapFlags flags, VMAType type, std::string_view name, MemoryMapFlags flags, VMAType type, std::string_view name,
PAddr phys_addr, u64 alignment) { PAddr phys_addr, u64 alignment) {
VAddr mapped_addr = alignment > 0 ? Common::alignUp(virtual_addr, alignment) : virtual_addr; VAddr mapped_addr = alignment > 0 ? Common::AlignUp(virtual_addr, alignment) : virtual_addr;
SCOPE_EXIT { SCOPE_EXIT {
auto& new_vma = AddMapping(mapped_addr, size); auto& new_vma = AddMapping(mapped_addr, size);
new_vma.disallow_merge = True(flags & MemoryMapFlags::NoCoalesce); new_vma.disallow_merge = True(flags & MemoryMapFlags::NoCoalesce);

419
src/core/module.cpp Normal file
View File

@ -0,0 +1,419 @@
// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#include <xbyak/xbyak.h>
#include "common/alignment.h"
#include "common/assert.h"
#include "common/logging/log.h"
#include "common/string_util.h"
#include "core/aerolib/aerolib.h"
#include "core/module.h"
#include "core/tls.h"
#include "core/virtual_memory.h"
namespace Core {
using EntryFunc = PS4_SYSV_ABI int (*)(size_t args, const void* argp, ModuleFunc func);
static u64 LoadAddress = SYSTEM_RESERVED + CODE_BASE_OFFSET;
static constexpr u64 CODE_BASE_INCR = 0x010000000u;
static u64 GetAlignedSize(const elf_program_header& phdr) {
return (phdr.p_align != 0 ? (phdr.p_memsz + (phdr.p_align - 1)) & ~(phdr.p_align - 1)
: phdr.p_memsz);
}
static u64 CalculateBaseSize(const elf_header& ehdr, std::span<const elf_program_header> phdr) {
u64 base_size = 0;
for (u16 i = 0; i < ehdr.e_phnum; i++) {
if (phdr[i].p_memsz != 0 && (phdr[i].p_type == PT_LOAD || phdr[i].p_type == PT_SCE_RELRO)) {
const u64 last_addr = phdr[i].p_vaddr + GetAlignedSize(phdr[i]);
base_size = std::max(last_addr, base_size);
}
}
return base_size;
}
static std::string EncodeId(u64 nVal) {
std::string enc;
static constexpr std::string_view codes =
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+-";
if (nVal < 0x40u) {
enc += codes[nVal];
} else {
if (nVal < 0x1000u) {
enc += codes[static_cast<u16>(nVal >> 6u) & 0x3fu];
enc += codes[nVal & 0x3fu];
} else {
enc += codes[static_cast<u16>(nVal >> 12u) & 0x3fu];
enc += codes[static_cast<u16>(nVal >> 6u) & 0x3fu];
enc += codes[nVal & 0x3fu];
}
}
return enc;
}
Module::Module(const std::filesystem::path& file_) : file{file_} {
elf.Open(file);
if (elf.IsElfFile()) {
LoadModuleToMemory();
LoadDynamicInfo();
LoadSymbols();
}
}
Module::~Module() = default;
void Module::Start(size_t args, const void* argp, ModuleFunc func) {
LOG_INFO(Core_Linker, "Module started : {}", file.filename().string());
const VAddr addr = dynamic_info.init_virtual_addr + GetBaseAddress();
reinterpret_cast<EntryFunc>(addr)(args, argp, func);
}
void Module::LoadModuleToMemory() {
static constexpr size_t BlockAlign = 0x1000;
static constexpr u64 TrampolineSize = 8_MB;
// Retrieve elf header and program header
const auto elf_header = elf.GetElfHeader();
const auto elf_pheader = elf.GetProgramHeader();
const u64 base_size = CalculateBaseSize(elf_header, elf_pheader);
aligned_base_size = Common::AlignUp(base_size, BlockAlign);
// Map module segments (and possible TLS trampolines)
base_virtual_addr = VirtualMemory::memory_alloc(LoadAddress, aligned_base_size + TrampolineSize,
VirtualMemory::MemoryMode::ExecuteReadWrite);
LoadAddress += CODE_BASE_INCR * (1 + aligned_base_size / CODE_BASE_INCR);
// Initialize trampoline generator.
void* trampoline_addr = std::bit_cast<void*>(base_virtual_addr + aligned_base_size);
Xbyak::CodeGenerator c(TrampolineSize, trampoline_addr);
LOG_INFO(Core_Linker, "======== Load Module to Memory ========");
LOG_INFO(Core_Linker, "base_virtual_addr ......: {:#018x}", base_virtual_addr);
LOG_INFO(Core_Linker, "base_size ..............: {:#018x}", base_size);
LOG_INFO(Core_Linker, "aligned_base_size ......: {:#018x}", aligned_base_size);
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) {
case PT_LOAD:
case PT_SCE_RELRO: {
if (elf_pheader[i].p_memsz == 0) {
LOG_ERROR(Core_Linker, "p_memsz==0 in type {}", header_type);
continue;
}
const u64 segment_addr = elf_pheader[i].p_vaddr + base_virtual_addr;
const u64 segment_file_size = elf_pheader[i].p_filesz;
const u64 segment_memory_size = GetAlignedSize(elf_pheader[i]);
const auto segment_mode = elf.ElfPheaderFlagsStr(elf_pheader[i].p_flags);
LOG_INFO(Core_Linker, "program header = [{}] type = {}", i, header_type);
LOG_INFO(Core_Linker, "segment_addr ..........: {:#018x}", segment_addr);
LOG_INFO(Core_Linker, "segment_file_size .....: {}", segment_file_size);
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);
if (elf_pheader[i].p_flags & PF_EXEC) {
PatchTLS(segment_addr, segment_memory_size, c);
}
break;
}
case PT_DYNAMIC:
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());
elf.LoadSegment(segment_addr, elf_pheader[i].p_offset, elf_pheader[i].p_filesz);
} else {
LOG_ERROR(Core_Linker, "p_filesz==0 in type {}", header_type);
}
break;
case PT_SCE_DYNLIBDATA:
if (elf_pheader[i].p_filesz != 0) {
m_dynamic_data.resize(elf_pheader[i].p_filesz);
const VAddr segment_addr = std::bit_cast<VAddr>(m_dynamic_data.data());
elf.LoadSegment(segment_addr, elf_pheader[i].p_offset, elf_pheader[i].p_filesz);
} else {
LOG_ERROR(Core_Linker, "p_filesz==0 in type {}", header_type);
}
break;
case PT_TLS:
tls.init_image_size = elf_pheader[i].p_filesz;
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]);
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;
default:
LOG_ERROR(Core_Linker, "Unimplemented type {}", header_type);
}
}
const VAddr entry_addr = base_virtual_addr + elf.GetElfEntry();
LOG_INFO(Core_Linker, "program entry addr ..........: {:#018x}", entry_addr);
}
void Module::LoadDynamicInfo() {
for (const auto* dyn = reinterpret_cast<elf_dynamic*>(m_dynamic.data()); dyn->d_tag != DT_NULL;
dyn++) {
switch (dyn->d_tag) {
case DT_SCE_HASH: // Offset of the hash table.
dynamic_info.hash_table =
reinterpret_cast<void*>(m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_HASHSZ: // Size of the hash table
dynamic_info.hash_table_size = dyn->d_un.d_val;
break;
case DT_SCE_STRTAB: // Offset of the string table.
dynamic_info.str_table =
reinterpret_cast<char*>(m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_STRSZ: // Size of the string table.
dynamic_info.str_table_size = dyn->d_un.d_val;
break;
case DT_SCE_SYMTAB: // Offset of the symbol table.
dynamic_info.symbol_table =
reinterpret_cast<elf_symbol*>(m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_SYMTABSZ: // Size of the symbol table.
dynamic_info.symbol_table_total_size = dyn->d_un.d_val;
break;
case DT_INIT:
dynamic_info.init_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_FINI:
dynamic_info.fini_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_SCE_PLTGOT: // Offset of the global offset table.
dynamic_info.pltgot_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_SCE_JMPREL: // Offset of the table containing jump slots.
dynamic_info.jmp_relocation_table =
reinterpret_cast<elf_relocation*>(m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_PLTRELSZ: // Size of the global offset table.
dynamic_info.jmp_relocation_table_size = dyn->d_un.d_val;
break;
case DT_SCE_PLTREL: // The type of relocations in the relocation table. Should be DT_RELA
dynamic_info.jmp_relocation_type = dyn->d_un.d_val;
if (dynamic_info.jmp_relocation_type != DT_RELA) {
LOG_WARNING(Core_Linker, "DT_SCE_PLTREL is NOT DT_RELA should check!");
}
break;
case DT_SCE_RELA: // Offset of the relocation table.
dynamic_info.relocation_table =
reinterpret_cast<elf_relocation*>(m_dynamic_data.data() + dyn->d_un.d_ptr);
break;
case DT_SCE_RELASZ: // Size of the relocation table.
dynamic_info.relocation_table_size = dyn->d_un.d_val;
break;
case DT_SCE_RELAENT: // The size of relocation table entries.
dynamic_info.relocation_table_entries_size = dyn->d_un.d_val;
if (dynamic_info.relocation_table_entries_size != 0x18) {
LOG_WARNING(Core_Linker, "DT_SCE_RELAENT is NOT 0x18 should check!");
}
break;
case DT_INIT_ARRAY: // Address of the array of pointers to initialization functions
dynamic_info.init_array_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_FINI_ARRAY: // Address of the array of pointers to termination functions
dynamic_info.fini_array_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_INIT_ARRAYSZ: // Size in bytes of the array of initialization functions
dynamic_info.init_array_size = dyn->d_un.d_val;
break;
case DT_FINI_ARRAYSZ: // Size in bytes of the array of terminationfunctions
dynamic_info.fini_array_size = dyn->d_un.d_val;
break;
case DT_PREINIT_ARRAY: // Address of the array of pointers to pre - initialization functions
dynamic_info.preinit_array_virtual_addr = dyn->d_un.d_ptr;
break;
case DT_PREINIT_ARRAYSZ: // Size in bytes of the array of pre - initialization functions
dynamic_info.preinit_array_size = dyn->d_un.d_val;
break;
case DT_SCE_SYMENT: // The size of symbol table entries
dynamic_info.symbol_table_entries_size = dyn->d_un.d_val;
if (dynamic_info.symbol_table_entries_size != 0x18) {
LOG_WARNING(Core_Linker, "DT_SCE_SYMENT is NOT 0x18 should check!");
}
break;
case DT_DEBUG:
dynamic_info.debug = dyn->d_un.d_val;
break;
case DT_TEXTREL:
dynamic_info.textrel = dyn->d_un.d_val;
break;
case DT_FLAGS:
dynamic_info.flags = dyn->d_un.d_val;
// This value should always be DF_TEXTREL (0x04)
if (dynamic_info.flags != 0x04) {
LOG_WARNING(Core_Linker, "DT_FLAGS is NOT 0x04 should check!");
}
break;
case DT_NEEDED:
// Offset of the library string in the string table to be linked in.
// In theory this should already be filled from about just make a test case
if (dynamic_info.str_table) {
dynamic_info.needed.push_back(dynamic_info.str_table + dyn->d_un.d_val);
} else {
LOG_ERROR(Core_Linker, "DT_NEEDED str table is not loaded should check!");
}
break;
case DT_SCE_NEEDED_MODULE: {
ModuleInfo& info = dynamic_info.import_modules.emplace_back();
info.value = dyn->d_un.d_val;
info.name = dynamic_info.str_table + info.name_offset;
info.enc_id = EncodeId(info.id);
break;
}
case DT_SCE_IMPORT_LIB: {
LibraryInfo& info = dynamic_info.import_libs.emplace_back();
info.value = dyn->d_un.d_val;
info.name = dynamic_info.str_table + info.name_offset;
info.enc_id = EncodeId(info.id);
break;
}
case DT_SCE_FINGERPRINT:
// The fingerprint is a 24 byte (0x18) size buffer that contains a unique identifier for
// 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);
break;
case DT_SCE_IMPORT_LIB_ATTR:
// The upper 32-bits should contain the module index multiplied by 0x10000. The lower
// 32-bits should be a constant 0x9.
LOG_INFO(Core_Linker, "unsupported DT_SCE_IMPORT_LIB_ATTR value = ......: {:#018x}",
dyn->d_un.d_val);
break;
case DT_SCE_ORIGINAL_FILENAME:
dynamic_info.filename = dynamic_info.str_table + dyn->d_un.d_val;
break;
case DT_SCE_MODULE_INFO: {
ModuleInfo& info = dynamic_info.export_modules.emplace_back();
info.value = dyn->d_un.d_val;
info.name = dynamic_info.str_table + info.name_offset;
info.enc_id = EncodeId(info.id);
break;
};
case DT_SCE_MODULE_ATTR:
LOG_INFO(Core_Linker, "unsupported DT_SCE_MODULE_ATTR value = ..........: {:#018x}",
dyn->d_un.d_val);
break;
case DT_SCE_EXPORT_LIB: {
LibraryInfo& info = dynamic_info.export_libs.emplace_back();
info.value = dyn->d_un.d_val;
info.name = dynamic_info.str_table + info.name_offset;
info.enc_id = EncodeId(info.id);
break;
}
default:
LOG_INFO(Core_Linker, "unsupported dynamic tag ..........: {:#018x}", dyn->d_tag);
}
}
}
void Module::LoadSymbols() {
const auto symbol_database = [this](Loader::SymbolsResolver& symbol, bool export_func) {
if (!dynamic_info.symbol_table || !dynamic_info.str_table ||
dynamic_info.symbol_table_total_size == 0) {
LOG_INFO(Core_Linker, "Symbol table not found!");
return;
}
for (auto* sym = dynamic_info.symbol_table;
reinterpret_cast<u8*>(sym) < reinterpret_cast<u8*>(dynamic_info.symbol_table) +
dynamic_info.symbol_table_total_size;
sym++) {
const u8 bind = sym->GetBind();
const u8 type = sym->GetType();
const u8 visibility = sym->GetVisibility();
const auto id = std::string(dynamic_info.str_table + sym->st_name);
const auto ids = Common::SplitString(id, '#');
if (ids.size() != 3) {
continue;
}
const auto* library = FindLibrary(ids[1]);
const auto* module = FindModule(ids[2]);
ASSERT_MSG(library && module, "Unable to find library and module");
if ((bind != STB_GLOBAL && bind != STB_WEAK) ||
(type != STT_FUN && type != STT_OBJECT) || export_func != (sym->st_value != 0)) {
continue;
}
const auto aeronid = AeroLib::FindByNid(ids.at(0).c_str());
const auto nid_name = aeronid ? aeronid->name : "UNK";
Loader::SymbolResolver sym_r{};
sym_r.name = ids.at(0);
sym_r.nidName = nid_name;
sym_r.library = library->name;
sym_r.library_version = library->version;
sym_r.module = module->name;
sym_r.module_version_major = module->version_major;
sym_r.module_version_minor = module->version_minor;
switch (type) {
case STT_NOTYPE:
sym_r.type = Loader::SymbolType::NoType;
break;
case STT_FUN:
sym_r.type = Loader::SymbolType::Function;
break;
case STT_OBJECT:
sym_r.type = Loader::SymbolType::Object;
break;
default:
sym_r.type = Loader::SymbolType::Unknown;
break;
}
const VAddr sym_addr = export_func ? sym->st_value + base_virtual_addr : 0;
symbol.AddSymbol(sym_r, sym_addr);
}
};
symbol_database(export_sym, true);
symbol_database(import_sym, false);
}
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) {
if (mod.enc_id == id) {
return &import_modules[i];
}
i++;
}
const auto& export_modules = dynamic_info.export_modules;
for (u32 i = 0; const auto& mod : export_modules) {
if (mod.enc_id == id) {
return &export_modules[i];
}
i++;
}
return nullptr;
}
const LibraryInfo* Module::FindLibrary(std::string_view id) {
const auto& import_libs = dynamic_info.import_libs;
for (u32 i = 0; const auto& lib : import_libs) {
if (lib.enc_id == id) {
return &import_libs[i];
}
i++;
}
const auto& export_libs = dynamic_info.export_libs;
for (u32 i = 0; const auto& lib : export_libs) {
if (lib.enc_id == id) {
return &export_libs[i];
}
i++;
}
return nullptr;
}
} // namespace Core

172
src/core/module.h Normal file
View File

@ -0,0 +1,172 @@
// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <string>
#include <vector>
#include "common/types.h"
#include "core/loader/elf.h"
#include "core/loader/symbols_resolver.h"
namespace Core {
struct ModuleInfo {
bool operator==(const ModuleInfo& other) const {
return version_major == other.version_major && version_minor == other.version_minor &&
name == other.name;
}
std::string name;
union {
u64 value;
struct {
u32 name_offset;
u8 version_minor;
u8 version_major;
u16 id;
};
};
std::string enc_id;
};
struct LibraryInfo {
bool operator==(const LibraryInfo& other) const {
return version == other.version && name == other.name;
}
std::string name;
union {
u64 value;
struct {
u32 name_offset;
u16 version;
u16 id;
};
};
std::string enc_id;
};
struct ThreadLocalImage {
u64 align;
VAddr image_virtual_addr;
u64 init_image_size;
u64 image_size;
u64 distance_from_fs;
};
struct DynamicModuleInfo {
void* hash_table = nullptr;
u64 hash_table_size = 0;
char* str_table = nullptr;
u64 str_table_size = 0;
elf_symbol* symbol_table = nullptr;
u64 symbol_table_total_size = 0;
u64 symbol_table_entries_size = 0;
u64 init_virtual_addr = 0;
u64 fini_virtual_addr = 0;
u64 pltgot_virtual_addr = 0;
u64 init_array_virtual_addr = 0;
u64 fini_array_virtual_addr = 0;
u64 preinit_array_virtual_addr = 0;
u64 init_array_size = 0;
u64 fini_array_size = 0;
u64 preinit_array_size = 0;
elf_relocation* jmp_relocation_table = nullptr;
u64 jmp_relocation_table_size = 0;
s64 jmp_relocation_type = 0;
elf_relocation* relocation_table = nullptr;
u64 relocation_table_size = 0;
u64 relocation_table_entries_size = 0;
u64 debug = 0;
u64 textrel = 0;
u64 flags = 0;
std::vector<const char*> needed;
std::vector<ModuleInfo> import_modules;
std::vector<ModuleInfo> export_modules;
std::vector<LibraryInfo> import_libs;
std::vector<LibraryInfo> export_libs;
std::string filename;
};
using ModuleFunc = int (*)(size_t, const void*);
class Module {
public:
explicit Module(const std::filesystem::path& file);
~Module();
VAddr GetBaseAddress() const noexcept {
return base_virtual_addr;
}
VAddr GetEntryAddress() const noexcept {
return base_virtual_addr + elf.GetElfEntry();
}
bool IsValid() const noexcept {
return base_virtual_addr != 0;
}
bool IsSharedLib() const noexcept {
return elf.IsSharedLib();
}
VAddr GetProcParam() const noexcept {
return proc_param_virtual_addr;
}
std::span<const ModuleInfo> GetImportModules() const {
return dynamic_info.import_modules;
}
std::span<const ModuleInfo> GetExportModules() const {
return dynamic_info.export_modules;
}
std::span<const LibraryInfo> GetImportLibs() const {
return dynamic_info.import_libs;
}
std::span<const LibraryInfo> GetExportLibs() const {
return dynamic_info.export_libs;
}
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);
}
for (u32 i = 0; i < dynamic_info.jmp_relocation_table_size / sizeof(elf_relocation); i++) {
func(&dynamic_info.jmp_relocation_table[i], true);
}
}
void Start(size_t args, const void* argp, ModuleFunc func);
void LoadModuleToMemory();
void LoadDynamicInfo();
void LoadSymbols();
const ModuleInfo* FindModule(std::string_view id);
const LibraryInfo* FindLibrary(std::string_view id);
public:
std::filesystem::path file;
Loader::Elf elf;
u64 aligned_base_size{};
VAddr base_virtual_addr{};
VAddr proc_param_virtual_addr{};
DynamicModuleInfo dynamic_info{};
std::vector<u8> m_dynamic;
std::vector<u8> m_dynamic_data;
Loader::SymbolsResolver export_sym;
Loader::SymbolsResolver import_sym;
ThreadLocalImage tls;
};
} // namespace Core

15
src/core/tls.cpp
View File

@ -43,20 +43,9 @@ constexpr static TLSPattern TlsPatterns[] = {
#ifdef _WIN32 #ifdef _WIN32
static DWORD slot = 0; static DWORD slot = 0;
u8* tls_memory{};
u64 tls_image{};
u32 tls_image_size{};
void SetTLSStorage(u64 image_address, u32 image_size) { void SetTcbBase(void* image_address) {
// Guest apps will use both positive and negative offsets to the TLS pointer. const BOOL result = TlsSetValue(slot, image_address);
// User data at probably in negative offsets, while pthread data at positive offset.
if (tls_image == 0) {
tls_image = image_address;
tls_image_size = image_size;
}
tls_memory = (u8*)std::malloc(tls_image_size + 8192);
std::memcpy(tls_memory, reinterpret_cast<LPVOID>(tls_image), tls_image_size);
const BOOL result = TlsSetValue(slot, tls_memory + tls_image_size);
ASSERT(result != 0); ASSERT(result != 0);
} }

5
src/core/tls.h
View File

@ -11,15 +11,14 @@ class CodeGenerator;
namespace Core { namespace Core {
/// Sets the data pointer that contains the TLS image. /// Sets the data pointer to the TCB block.
void SetTLSStorage(u64 image_address, u32 image_size); void SetTcbBase(void* image_address);
/// Patches any instructions that access guest TLS to use provided storage. /// Patches any instructions that access guest TLS to use provided storage.
void PatchTLS(u64 segment_addr, u64 segment_size, Xbyak::CodeGenerator& c); void PatchTLS(u64 segment_addr, u64 segment_size, Xbyak::CodeGenerator& c);
class ThreadLocalStorage { class ThreadLocalStorage {
public: public:
}; };
} // namespace Core } // namespace Core

6
src/main.cpp
View File

@ -76,10 +76,10 @@ int main(int argc, char* argv[]) {
} }
auto linker = Common::Singleton<Core::Linker>::Instance(); auto linker = Common::Singleton<Core::Linker>::Instance();
Libraries::InitHLELibs(&linker->getHLESymbols()); Libraries::InitHLELibs(&linker->GetHLESymbols());
linker->LoadModule(path); linker->LoadModule(path);
// check if we have system modules to load // Check if we have system modules to load
const auto& sys_module_path = Common::FS::GetUserPath(Common::FS::PathType::SysModuleDir); const auto& sys_module_path = Common::FS::GetUserPath(Common::FS::PathType::SysModuleDir);
for (const auto& entry : std::filesystem::directory_iterator(sys_module_path)) { for (const auto& entry : std::filesystem::directory_iterator(sys_module_path)) {
if (entry.path().filename() == "libSceNgs2.sprx") { if (entry.path().filename() == "libSceNgs2.sprx") {
@ -103,7 +103,7 @@ int main(int argc, char* argv[]) {
} }
} }
if (!found) { if (!found) {
Libraries::LibC::libcSymbolsRegister(&linker->getHLESymbols()); Libraries::LibC::libcSymbolsRegister(&linker->GetHLESymbols());
} }
std::thread mainthread([linker]() { linker->Execute(); }); std::thread mainthread([linker]() { linker->Execute(); });
Discord::RPC discordRPC; Discord::RPC discordRPC;

16
src/shader_recompiler/frontend/translate/scalar_alu.cpp
View File

@ -63,18 +63,28 @@ void Translator::S_AND_SAVEEXEC_B64(const GcnInst& inst) {
// This instruction normally operates on 64-bit data (EXEC, VCC, SGPRs) // This instruction normally operates on 64-bit data (EXEC, VCC, SGPRs)
// However here we flatten it to 1-bit EXEC and 1-bit VCC. For the destination // However here we flatten it to 1-bit EXEC and 1-bit VCC. For the destination
// SGPR we have a special IR opcode for SPGRs that act as thread masks. // SGPR we have a special IR opcode for SPGRs that act as thread masks.
ASSERT(inst.src[0].field == OperandField::VccLo);
const IR::U1 exec{ir.GetExec()}; const IR::U1 exec{ir.GetExec()};
const IR::U1 vcc{ir.GetVcc()};
// Mark destination SPGR as an EXEC context. This means we will use 1-bit // Mark destination SPGR as an EXEC context. This means we will use 1-bit
// IR instruction whenever it's loaded. // IR instruction whenever it's loaded.
ASSERT(inst.dst[0].field == OperandField::ScalarGPR); switch (inst.dst[0].field) {
case OperandField::ScalarGPR: {
const u32 reg = inst.dst[0].code; const u32 reg = inst.dst[0].code;
exec_contexts[reg] = true; exec_contexts[reg] = true;
ir.SetThreadBitScalarReg(IR::ScalarReg(reg), exec); ir.SetThreadBitScalarReg(IR::ScalarReg(reg), exec);
break;
}
case OperandField::VccLo:
ir.SetVcc(exec);
break;
default:
UNREACHABLE();
}
// Update EXEC. // Update EXEC.
ASSERT(inst.src[0].field == OperandField::VccLo); ir.SetExec(ir.LogicalAnd(exec, vcc));
ir.SetExec(ir.LogicalAnd(exec, ir.GetVcc()));
} }
void Translator::S_MOV_B64(const GcnInst& inst) { void Translator::S_MOV_B64(const GcnInst& inst) {

4
src/shader_recompiler/frontend/translate/translate.cpp
View File

@ -128,7 +128,11 @@ IR::U1U32F32 Translator::GetSrc(const InstOperand& operand, bool force_flt) {
value = ir.GetExec(); value = ir.GetExec();
break; break;
case OperandField::VccLo: case OperandField::VccLo:
if (force_flt) {
value = ir.BitCast<IR::F32>(ir.GetVccLo());
} else {
value = ir.GetVccLo(); value = ir.GetVccLo();
}
break; break;
case OperandField::VccHi: case OperandField::VccHi:
value = ir.GetVccHi(); value = ir.GetVccHi();

2
src/shader_recompiler/frontend/translate/vector_alu.cpp
View File

@ -33,7 +33,7 @@ void Translator::V_CNDMASK_B32(const GcnInst& inst) {
const IR::VectorReg dst_reg{inst.dst[0].code}; const IR::VectorReg dst_reg{inst.dst[0].code};
const IR::ScalarReg flag_reg{inst.src[2].code}; const IR::ScalarReg flag_reg{inst.src[2].code};
const IR::U1 flag = inst.src[2].field == OperandField::ScalarGPR const IR::U1 flag = inst.src[2].field == OperandField::ScalarGPR
? ir.INotEqual(ir.GetScalarReg(flag_reg), ir.Imm32(0U)) ? ir.GetThreadBitScalarReg(flag_reg)
: ir.GetVcc(); : ir.GetVcc();
// We can treat the instruction as integer most of the time, but when a source is // We can treat the instruction as integer most of the time, but when a source is

1
src/shader_recompiler/ir/ir_emitter.cpp
View File

@ -774,6 +774,7 @@ U1 IREmitter::FPLessThanEqual(const F32F64& lhs, const F32F64& rhs, bool ordered
} }
U1 IREmitter::FPGreaterThanEqual(const F32F64& lhs, const F32F64& rhs, bool ordered) { U1 IREmitter::FPGreaterThanEqual(const F32F64& lhs, const F32F64& rhs, bool ordered) {
ASSERT(lhs.Type() == rhs.Type());
if (lhs.Type() != rhs.Type()) { if (lhs.Type() != rhs.Type()) {
throw InvalidArgument("Mismatching types {} and {}", lhs.Type(), rhs.Type()); throw InvalidArgument("Mismatching types {} and {}", lhs.Type(), rhs.Type());
} }

4
src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
View File

@ -63,8 +63,8 @@ GraphicsPipeline::GraphicsPipeline(const Instance& instance_, Scheduler& schedul
.pVertexAttributeDescriptions = attributes.data(), .pVertexAttributeDescriptions = attributes.data(),
}; };
ASSERT_MSG(key.prim_type != Liverpool::PrimitiveType::RectList || IsEmbeddedVs(), // ASSERT_MSG(key.prim_type != Liverpool::PrimitiveType::RectList || IsEmbeddedVs(),
"Rectangle List primitive type is only supported for embedded VS"); // "Rectangle List primitive type is only supported for embedded VS");
const vk::PipelineInputAssemblyStateCreateInfo input_assembly = { const vk::PipelineInputAssemblyStateCreateInfo input_assembly = {
.topology = LiverpoolToVK::PrimitiveType(key.prim_type), .topology = LiverpoolToVK::PrimitiveType(key.prim_type),

5
src/video_core/renderer_vulkan/vk_rasterizer.cpp
View File

@ -41,6 +41,7 @@ void Rasterizer::Draw(bool is_indexed) {
boost::container::static_vector<vk::RenderingAttachmentInfo, Liverpool::NumColorBuffers> boost::container::static_vector<vk::RenderingAttachmentInfo, Liverpool::NumColorBuffers>
color_attachments{}; color_attachments{};
const std::array color = {0.5f, 0.5f, 0.5f, 1.f};
for (auto col_buf_id = 0u; col_buf_id < Liverpool::NumColorBuffers; ++col_buf_id) { for (auto col_buf_id = 0u; col_buf_id < Liverpool::NumColorBuffers; ++col_buf_id) {
const auto& col_buf = regs.color_buffers[col_buf_id]; const auto& col_buf = regs.color_buffers[col_buf_id];
if (!col_buf) { if (!col_buf) {
@ -53,10 +54,12 @@ void Rasterizer::Draw(bool is_indexed) {
color_attachments.push_back({ color_attachments.push_back({
.imageView = *image_view.image_view, .imageView = *image_view.image_view,
.imageLayout = vk::ImageLayout::eGeneral, .imageLayout = vk::ImageLayout::eGeneral,
.loadOp = vk::AttachmentLoadOp::eLoad, .loadOp = compute_done ? vk::AttachmentLoadOp::eClear : vk::AttachmentLoadOp::eLoad,
.storeOp = vk::AttachmentStoreOp::eStore, .storeOp = vk::AttachmentStoreOp::eStore,
.clearValue = vk::ClearValue{color},
}); });
} }
compute_done = false;
// TODO: Don't restart renderpass every draw // TODO: Don't restart renderpass every draw
const auto& scissor = regs.screen_scissor; const auto& scissor = regs.screen_scissor;

6
src/video_core/renderer_vulkan/vk_stream_buffer.cpp
View File

@ -92,14 +92,14 @@ StreamBuffer::~StreamBuffer() {
std::tuple<u8*, u64, bool> StreamBuffer::Map(u64 size, u64 alignment) { std::tuple<u8*, u64, bool> StreamBuffer::Map(u64 size, u64 alignment) {
if (!is_coherent && type == BufferType::Stream) { if (!is_coherent && type == BufferType::Stream) {
size = Common::alignUp(size, instance.NonCoherentAtomSize()); size = Common::AlignUp(size, instance.NonCoherentAtomSize());
} }
ASSERT(size <= stream_buffer_size); ASSERT(size <= stream_buffer_size);
mapped_size = size; mapped_size = size;
if (alignment > 0) { if (alignment > 0) {
offset = Common::alignUp(offset, alignment); offset = Common::AlignUp(offset, alignment);
} }
bool invalidate{false}; bool invalidate{false};
@ -124,7 +124,7 @@ std::tuple<u8*, u64, bool> StreamBuffer::Map(u64 size, u64 alignment) {
void StreamBuffer::Commit(u64 size) { void StreamBuffer::Commit(u64 size) {
if (!is_coherent && type == BufferType::Stream) { if (!is_coherent && type == BufferType::Stream) {
size = Common::alignUp(size, instance.NonCoherentAtomSize()); size = Common::AlignUp(size, instance.NonCoherentAtomSize());
} }
ASSERT_MSG(size <= mapped_size, "Reserved size {} is too small compared to {}", mapped_size, ASSERT_MSG(size <= mapped_size, "Reserved size {} is too small compared to {}", mapped_size,

2
src/video_core/texture_cache/texture_cache.cpp
View File

@ -338,7 +338,7 @@ void TextureCache::UpdatePagesCachedCount(VAddr addr, u64 size, s32 delta) {
const u32 interval_size = interval_end_addr - interval_start_addr; const u32 interval_size = interval_end_addr - interval_start_addr;
void* addr = reinterpret_cast<void*>(interval_start_addr); void* addr = reinterpret_cast<void*>(interval_start_addr);
if (delta > 0 && count == delta) { if (delta > 0 && count == delta) {
mprotect(addr, interval_size, PAGE_READONLY); // mprotect(addr, interval_size, PAGE_READONLY);
} else if (delta < 0 && count == -delta) { } else if (delta < 0 && count == -delta) {
mprotect(addr, interval_size, PAGE_READWRITE); mprotect(addr, interval_size, PAGE_READWRITE);
} else { } else {