video_core: implement formats for N64 emulation

ui: Set Link Color when setting theme

.gitmodules

@@ -7,9 +7,6 @@
 [submodule "dynarmic"]
     path = externals/dynarmic
     url = https://github.com/MerryMage/dynarmic.git
-[submodule "soundtouch"]
-    path = externals/soundtouch
-    url = https://github.com/citra-emu/ext-soundtouch.git
 [submodule "libressl"]
     path = externals/libressl
     url = https://github.com/citra-emu/ext-libressl-portable.git

externals/CMakeLists.txt

@@ -68,9 +68,6 @@ if (YUZU_USE_EXTERNAL_SDL2)
     add_library(SDL2 ALIAS SDL2-static)
 endif()
 
-# SoundTouch
-add_subdirectory(soundtouch)
-
 # Cubeb
 if(ENABLE_CUBEB)
     set(BUILD_TESTS OFF CACHE BOOL "")

src/CMakeLists.txt

@@ -65,12 +65,14 @@ if (MSVC)
         /we4305 # 'context': truncation from 'type1' to 'type2'
         /we4388 # 'expression': signed/unsigned mismatch
         /we4389 # 'operator': signed/unsigned mismatch
+        /we4505 # 'function': unreferenced local function has been removed
         /we4547 # 'operator': operator before comma has no effect; expected operator with side-effect
         /we4549 # 'operator1': operator before comma has no effect; did you intend 'operator2'?
         /we4555 # Expression has no effect; expected expression with side-effect
         /we4715 # 'function': not all control paths return a value
         /we4834 # Discarding return value of function with 'nodiscard' attribute
         /we5038 # data member 'member1' will be initialized after data member 'member2'
+        /we5245 # 'function': unreferenced function with internal linkage has been removed
     )
 
     if (ARCHITECTURE_x86_64)
@@ -103,12 +105,6 @@ else()
         -Wno-unused-parameter
     )
 
-    # TODO: Remove when we update to a GCC compiler that enables this
-    #       by default (i.e. GCC 10 or newer).
-    if (CMAKE_CXX_COMPILER_ID STREQUAL GNU)
-        add_compile_options(-fconcepts)
-    endif()
-
     if (ARCHITECTURE_x86_64)
         add_compile_options("-mcx16")
     endif()

src/audio_core/CMakeLists.txt

@@ -36,8 +36,6 @@ add_library(audio_core STATIC
     splitter_context.h
     stream.cpp
     stream.h
-    time_stretch.cpp
-    time_stretch.h
     voice_context.cpp
     voice_context.h
 
@@ -63,7 +61,6 @@ if (NOT MSVC)
 endif()
 
 target_link_libraries(audio_core PUBLIC common core)
-target_link_libraries(audio_core PRIVATE SoundTouch)
 
 if(ENABLE_CUBEB)
     target_link_libraries(audio_core PRIVATE cubeb)

src/audio_core/cubeb_sink.cpp

@@ -7,7 +7,6 @@
 #include <cstring>
 #include "audio_core/cubeb_sink.h"
 #include "audio_core/stream.h"
-#include "audio_core/time_stretch.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/ring_buffer.h"
@@ -23,8 +22,7 @@ class CubebSinkStream final : public SinkStream {
 public:
     CubebSinkStream(cubeb* ctx_, u32 sample_rate, u32 num_channels_, cubeb_devid output_device,
                     const std::string& name)
-        : ctx{ctx_}, num_channels{std::min(num_channels_, 6u)}, time_stretch{sample_rate,
-                                                                             num_channels} {
+        : ctx{ctx_}, num_channels{std::min(num_channels_, 6u)} {
 
         cubeb_stream_params params{};
         params.rate = sample_rate;
@@ -131,7 +129,6 @@ private:
     Common::RingBuffer<s16, 0x10000> queue;
     std::array<s16, 2> last_frame{};
     std::atomic<bool> should_flush{};
-    TimeStretcher time_stretch;
 
     static long DataCallback(cubeb_stream* stream, void* user_data, const void* input_buffer,
                              void* output_buffer, long num_frames);
@@ -205,25 +202,7 @@ long CubebSinkStream::DataCallback([[maybe_unused]] cubeb_stream* stream, void*
 
     const std::size_t num_channels = impl->GetNumChannels();
     const std::size_t samples_to_write = num_channels * num_frames;
-    std::size_t samples_written;
-
-    /*
-    if (Settings::values.enable_audio_stretching.GetValue()) {
-        const std::vector<s16> in{impl->queue.Pop()};
-        const std::size_t num_in{in.size() / num_channels};
-        s16* const out{reinterpret_cast<s16*>(buffer)};
-        const std::size_t out_frames =
-            impl->time_stretch.Process(in.data(), num_in, out, num_frames);
-        samples_written = out_frames * num_channels;
-
-        if (impl->should_flush) {
-            impl->time_stretch.Flush();
-            impl->should_flush = false;
-        }
-    } else {
-        samples_written = impl->queue.Pop(buffer, samples_to_write);
-    }*/
-    samples_written = impl->queue.Pop(buffer, samples_to_write);
+    const std::size_t samples_written = impl->queue.Pop(buffer, samples_to_write);
 
     if (samples_written >= num_channels) {
         std::memcpy(&impl->last_frame[0], buffer + (samples_written - num_channels) * sizeof(s16),

src/audio_core/sdl2_sink.cpp

@@ -7,7 +7,6 @@
 #include <cstring>
 #include "audio_core/sdl2_sink.h"
 #include "audio_core/stream.h"
-#include "audio_core/time_stretch.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
 //#include "common/settings.h"
@@ -27,7 +26,7 @@ namespace AudioCore {
 class SDLSinkStream final : public SinkStream {
 public:
     SDLSinkStream(u32 sample_rate, u32 num_channels_, const std::string& output_device)
-        : num_channels{std::min(num_channels_, 6u)}, time_stretch{sample_rate, num_channels} {
+        : num_channels{std::min(num_channels_, 6u)} {
 
         SDL_AudioSpec spec;
         spec.freq = sample_rate;
@@ -116,7 +115,6 @@ private:
     SDL_AudioDeviceID dev = 0;
     u32 num_channels{};
     std::atomic<bool> should_flush{};
-    TimeStretcher time_stretch;
 };
 
 SDLSink::SDLSink(std::string_view target_device_name) {

src/audio_core/time_stretch.cpp

@@ -1,68 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <algorithm>
-#include <cmath>
-#include <cstddef>
-#include "audio_core/time_stretch.h"
-#include "common/logging/log.h"
-
-namespace AudioCore {
-
-TimeStretcher::TimeStretcher(u32 sample_rate, u32 channel_count) : m_sample_rate{sample_rate} {
-    m_sound_touch.setChannels(channel_count);
-    m_sound_touch.setSampleRate(sample_rate);
-    m_sound_touch.setPitch(1.0);
-    m_sound_touch.setTempo(1.0);
-}
-
-void TimeStretcher::Clear() {
-    m_sound_touch.clear();
-}
-
-void TimeStretcher::Flush() {
-    m_sound_touch.flush();
-}
-
-std::size_t TimeStretcher::Process(const s16* in, std::size_t num_in, s16* out,
-                                   std::size_t num_out) {
-    const double time_delta = static_cast<double>(num_out) / m_sample_rate; // seconds
-
-    // We were given actual_samples number of samples, and num_samples were requested from us.
-    double current_ratio = static_cast<double>(num_in) / static_cast<double>(num_out);
-
-    const double max_latency = 0.25; // seconds
-    const double max_backlog = m_sample_rate * max_latency;
-    const double backlog_fullness = m_sound_touch.numSamples() / max_backlog;
-    if (backlog_fullness > 4.0) {
-        // Too many samples in backlog: Don't push anymore on
-        num_in = 0;
-    }
-
-    // We ideally want the backlog to be about 50% full.
-    // This gives some headroom both ways to prevent underflow and overflow.
-    // We tweak current_ratio to encourage this.
-    constexpr double tweak_time_scale = 0.05; // seconds
-    const double tweak_correction = (backlog_fullness - 0.5) * (time_delta / tweak_time_scale);
-    current_ratio *= std::pow(1.0 + 2.0 * tweak_correction, tweak_correction < 0 ? 3.0 : 1.0);
-
-    // This low-pass filter smoothes out variance in the calculated stretch ratio.
-    // The time-scale determines how responsive this filter is.
-    constexpr double lpf_time_scale = 0.712; // seconds
-    const double lpf_gain = 1.0 - std::exp(-time_delta / lpf_time_scale);
-    m_stretch_ratio += lpf_gain * (current_ratio - m_stretch_ratio);
-
-    // Place a lower limit of 5% speed. When a game boots up, there will be
-    // many silence samples. These do not need to be timestretched.
-    m_stretch_ratio = std::max(m_stretch_ratio, 0.05);
-    m_sound_touch.setTempo(m_stretch_ratio);
-
-    LOG_TRACE(Audio, "{:5}/{:5} ratio:{:0.6f} backlog:{:0.6f}", num_in, num_out, m_stretch_ratio,
-              backlog_fullness);
-
-    m_sound_touch.putSamples(in, static_cast<u32>(num_in));
-    return m_sound_touch.receiveSamples(out, static_cast<u32>(num_out));
-}
-
-} // namespace AudioCore

src/audio_core/time_stretch.h

@@ -1,34 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <cstddef>
-#include <SoundTouch.h>
-#include "common/common_types.h"
-
-namespace AudioCore {
-
-class TimeStretcher {
-public:
-    TimeStretcher(u32 sample_rate, u32 channel_count);
-
-    /// @param in       Input sample buffer
-    /// @param num_in   Number of input frames in `in`
-    /// @param out      Output sample buffer
-    /// @param num_out  Desired number of output frames in `out`
-    /// @returns Actual number of frames written to `out`
-    std::size_t Process(const s16* in, std::size_t num_in, s16* out, std::size_t num_out);
-
-    void Clear();
-
-    void Flush();
-
-private:
-    u32 m_sample_rate;
-    soundtouch::SoundTouch m_sound_touch;
-    double m_stretch_ratio = 1.0;
-};
-
-} // namespace AudioCore

src/common/atomic_ops.h

@@ -46,6 +46,50 @@ namespace Common {
                                           reinterpret_cast<__int64*>(expected.data())) != 0;
 }
 
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u8* pointer, u8 value, u8 expected,
+                                               u8& actual) {
+    actual =
+        _InterlockedCompareExchange8(reinterpret_cast<volatile char*>(pointer), value, expected);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u16* pointer, u16 value, u16 expected,
+                                               u16& actual) {
+    actual =
+        _InterlockedCompareExchange16(reinterpret_cast<volatile short*>(pointer), value, expected);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u32* pointer, u32 value, u32 expected,
+                                               u32& actual) {
+    actual =
+        _InterlockedCompareExchange(reinterpret_cast<volatile long*>(pointer), value, expected);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u64* pointer, u64 value, u64 expected,
+                                               u64& actual) {
+    actual = _InterlockedCompareExchange64(reinterpret_cast<volatile __int64*>(pointer), value,
+                                           expected);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u64* pointer, u128 value, u128 expected,
+                                               u128& actual) {
+    const bool result =
+        _InterlockedCompareExchange128(reinterpret_cast<volatile __int64*>(pointer), value[1],
+                                       value[0], reinterpret_cast<__int64*>(expected.data())) != 0;
+    actual = expected;
+    return result;
+}
+
+[[nodiscard]] inline u128 AtomicLoad128(volatile u64* pointer) {
+    u128 result{};
+    _InterlockedCompareExchange128(reinterpret_cast<volatile __int64*>(pointer), result[1],
+                                   result[0], reinterpret_cast<__int64*>(result.data()));
+    return result;
+}
+
 #else
 
 [[nodiscard]] inline bool AtomicCompareAndSwap(volatile u8* pointer, u8 value, u8 expected) {
@@ -72,6 +116,52 @@ namespace Common {
     return __sync_bool_compare_and_swap((unsigned __int128*)pointer, expected_a, value_a);
 }
 
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u8* pointer, u8 value, u8 expected,
+                                               u8& actual) {
+    actual = __sync_val_compare_and_swap(pointer, expected, value);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u16* pointer, u16 value, u16 expected,
+                                               u16& actual) {
+    actual = __sync_val_compare_and_swap(pointer, expected, value);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u32* pointer, u32 value, u32 expected,
+                                               u32& actual) {
+    actual = __sync_val_compare_and_swap(pointer, expected, value);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u64* pointer, u64 value, u64 expected,
+                                               u64& actual) {
+    actual = __sync_val_compare_and_swap(pointer, expected, value);
+    return actual == expected;
+}
+
+[[nodiscard]] inline bool AtomicCompareAndSwap(volatile u64* pointer, u128 value, u128 expected,
+                                               u128& actual) {
+    unsigned __int128 value_a;
+    unsigned __int128 expected_a;
+    unsigned __int128 actual_a;
+    std::memcpy(&value_a, value.data(), sizeof(u128));
+    std::memcpy(&expected_a, expected.data(), sizeof(u128));
+    actual_a = __sync_val_compare_and_swap((unsigned __int128*)pointer, expected_a, value_a);
+    std::memcpy(actual.data(), &actual_a, sizeof(u128));
+    return actual_a == expected_a;
+}
+
+[[nodiscard]] inline u128 AtomicLoad128(volatile u64* pointer) {
+    unsigned __int128 zeros_a = 0;
+    unsigned __int128 result_a =
+        __sync_val_compare_and_swap((unsigned __int128*)pointer, zeros_a, zeros_a);
+
+    u128 result;
+    std::memcpy(result.data(), &result_a, sizeof(u128));
+    return result;
+}
+
 #endif
 
 } // namespace Common

src/common/host_memory.cpp

@@ -149,7 +149,7 @@ public:
     }
 
     void Unmap(size_t virtual_offset, size_t length) {
-        std::lock_guard lock{placeholder_mutex};
+        std::scoped_lock lock{placeholder_mutex};
 
         // Unmap until there are no more placeholders
         while (UnmapOnePlaceholder(virtual_offset, length)) {
@@ -169,7 +169,7 @@ public:
         }
         const size_t virtual_end = virtual_offset + length;
 
-        std::lock_guard lock{placeholder_mutex};
+        std::scoped_lock lock{placeholder_mutex};
         auto [it, end] = placeholders.equal_range({virtual_offset, virtual_end});
         while (it != end) {
             const size_t offset = std::max(it->lower(), virtual_offset);

src/common/logging/filter.cpp

@@ -101,6 +101,7 @@ bool ParseFilterRule(Filter& instance, Iterator begin, Iterator end) {
     SUB(Service, GRC)                                                                              \
     SUB(Service, HID)                                                                              \
     SUB(Service, IRS)                                                                              \
+    SUB(Service, JIT)                                                                              \
     SUB(Service, LBL)                                                                              \
     SUB(Service, LDN)                                                                              \
     SUB(Service, LDR)                                                                              \

src/common/logging/types.h

@@ -69,6 +69,7 @@ enum class Class : u8 {
     Service_GRC,       ///< The game recording service
     Service_HID,       ///< The HID (Human interface device) service
     Service_IRS,       ///< The IRS service
+    Service_JIT,       ///< The JIT service
     Service_LBL,       ///< The LBL (LCD backlight) service
     Service_LDN,       ///< The LDN (Local domain network) service
     Service_LDR,       ///< The loader service

src/common/settings.h

@@ -38,6 +38,7 @@ enum class CPUAccuracy : u32 {
     Auto = 0,
     Accurate = 1,
     Unsafe = 2,
+    Paranoid = 3,
 };
 
 enum class FullscreenMode : u32 {
@@ -470,7 +471,7 @@ struct Values {
 
     // Cpu
     RangedSetting<CPUAccuracy> cpu_accuracy{CPUAccuracy::Auto, CPUAccuracy::Auto,
-                                            CPUAccuracy::Unsafe, "cpu_accuracy"};
+                                            CPUAccuracy::Paranoid, "cpu_accuracy"};
     // TODO: remove cpu_accuracy_first_time, migration setting added 8 July 2021
     BasicSetting<bool> cpu_accuracy_first_time{true, "cpu_accuracy_first_time"};
     BasicSetting<bool> cpu_debug_mode{false, "cpu_debug_mode"};

src/common/thread.h

@@ -17,7 +17,7 @@ namespace Common {
 class Event {
 public:
     void Set() {
-        std::lock_guard lk{mutex};
+        std::scoped_lock lk{mutex};
         if (!is_set) {
             is_set = true;
             condvar.notify_one();

src/common/threadsafe_queue.h

@@ -52,7 +52,7 @@ public:
         // line before cv.wait
         // TODO(bunnei): This can be replaced with C++20 waitable atomics when properly supported.
         // See discussion on https://github.com/yuzu-emu/yuzu/pull/3173 for details.
-        std::lock_guard lock{cv_mutex};
+        std::scoped_lock lock{cv_mutex};
         cv.notify_one();
     }
 
@@ -159,7 +159,7 @@ public:
 
     template <typename Arg>
     void Push(Arg&& t) {
-        std::lock_guard lock{write_lock};
+        std::scoped_lock lock{write_lock};
         spsc_queue.Push(t);
     }
 

src/common/x64/native_clock.cpp

@@ -10,25 +10,49 @@
 #include "common/uint128.h"
 #include "common/x64/native_clock.h"
 
+#ifdef _MSC_VER
+#include <intrin.h>
+#endif
+
 namespace Common {
 
+#ifdef _MSC_VER
+__forceinline static u64 FencedRDTSC() {
+    _mm_lfence();
+    _ReadWriteBarrier();
+    const u64 result = __rdtsc();
+    _mm_lfence();
+    _ReadWriteBarrier();
+    return result;
+}
+#else
+static u64 FencedRDTSC() {
+    u64 result;
+    asm volatile("lfence\n\t"
+                 "rdtsc\n\t"
+                 "shl $32, %%rdx\n\t"
+                 "or %%rdx, %0\n\t"
+                 "lfence"
+                 : "=a"(result)
+                 :
+                 : "rdx", "memory", "cc");
+    return result;
+}
+#endif
+
 u64 EstimateRDTSCFrequency() {
     // Discard the first result measuring the rdtsc.
-    _mm_mfence();
-    __rdtsc();
+    FencedRDTSC();
     std::this_thread::sleep_for(std::chrono::milliseconds{1});
-    _mm_mfence();
-    __rdtsc();
+    FencedRDTSC();
 
     // Get the current time.
     const auto start_time = std::chrono::steady_clock::now();
-    _mm_mfence();
-    const u64 tsc_start = __rdtsc();
+    const u64 tsc_start = FencedRDTSC();
     // Wait for 200 milliseconds.
     std::this_thread::sleep_for(std::chrono::milliseconds{200});
     const auto end_time = std::chrono::steady_clock::now();
-    _mm_mfence();
-    const u64 tsc_end = __rdtsc();
+    const u64 tsc_end = FencedRDTSC();
     // Calculate differences.
     const u64 timer_diff = static_cast<u64>(
         std::chrono::duration_cast<std::chrono::nanoseconds>(end_time - start_time).count());
@@ -42,8 +66,7 @@ NativeClock::NativeClock(u64 emulated_cpu_frequency_, u64 emulated_clock_frequen
                          u64 rtsc_frequency_)
     : WallClock(emulated_cpu_frequency_, emulated_clock_frequency_, true), rtsc_frequency{
                                                                                rtsc_frequency_} {
-    _mm_mfence();
-    time_point.inner.last_measure = __rdtsc();
+    time_point.inner.last_measure = FencedRDTSC();
     time_point.inner.accumulated_ticks = 0U;
     ns_rtsc_factor = GetFixedPoint64Factor(NS_RATIO, rtsc_frequency);
     us_rtsc_factor = GetFixedPoint64Factor(US_RATIO, rtsc_frequency);
@@ -55,10 +78,10 @@ NativeClock::NativeClock(u64 emulated_cpu_frequency_, u64 emulated_clock_frequen
 u64 NativeClock::GetRTSC() {
     TimePoint new_time_point{};
     TimePoint current_time_point{};
+
+    current_time_point.pack = Common::AtomicLoad128(time_point.pack.data());
     do {
-        current_time_point.pack = time_point.pack;
-        _mm_mfence();
-        const u64 current_measure = __rdtsc();
+        const u64 current_measure = FencedRDTSC();
         u64 diff = current_measure - current_time_point.inner.last_measure;
         diff = diff & ~static_cast<u64>(static_cast<s64>(diff) >> 63); // max(diff, 0)
         new_time_point.inner.last_measure = current_measure > current_time_point.inner.last_measure
@@ -66,7 +89,7 @@ u64 NativeClock::GetRTSC() {
                                                 : current_time_point.inner.last_measure;
         new_time_point.inner.accumulated_ticks = current_time_point.inner.accumulated_ticks + diff;
     } while (!Common::AtomicCompareAndSwap(time_point.pack.data(), new_time_point.pack,
-                                           current_time_point.pack));
+                                           current_time_point.pack, current_time_point.pack));
     /// The clock cannot be more precise than the guest timer, remove the lower bits
     return new_time_point.inner.accumulated_ticks & inaccuracy_mask;
 }
@@ -75,13 +98,13 @@ void NativeClock::Pause(bool is_paused) {
     if (!is_paused) {
         TimePoint current_time_point{};
         TimePoint new_time_point{};
+
+        current_time_point.pack = Common::AtomicLoad128(time_point.pack.data());
         do {
-            current_time_point.pack = time_point.pack;
             new_time_point.pack = current_time_point.pack;
-            _mm_mfence();
-            new_time_point.inner.last_measure = __rdtsc();
+            new_time_point.inner.last_measure = FencedRDTSC();
         } while (!Common::AtomicCompareAndSwap(time_point.pack.data(), new_time_point.pack,
-                                               current_time_point.pack));
+                                               current_time_point.pack, current_time_point.pack));
     }
 }
 

src/core/CMakeLists.txt

@@ -13,6 +13,8 @@ add_library(core STATIC
     arm/dynarmic/arm_exclusive_monitor.h
     arm/exclusive_monitor.cpp
     arm/exclusive_monitor.h
+    arm/symbols.cpp
+    arm/symbols.h
     constants.cpp
     constants.h
     core.cpp
@@ -458,6 +460,10 @@ add_library(core STATIC
     hle/service/hid/controllers/touchscreen.h
     hle/service/hid/controllers/xpad.cpp
     hle/service/hid/controllers/xpad.h
+    hle/service/jit/jit_context.cpp
+    hle/service/jit/jit_context.h
+    hle/service/jit/jit.cpp
+    hle/service/jit/jit.h
     hle/service/lbl/lbl.cpp
     hle/service/lbl/lbl.h
     hle/service/ldn/errors.h

src/core/arm/arm_interface.cpp

@@ -8,134 +8,13 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/arm/arm_interface.h"
+#include "core/arm/symbols.h"
 #include "core/core.h"
+#include "core/hle/kernel/k_process.h"
 #include "core/loader/loader.h"
 #include "core/memory.h"
 
 namespace Core {
-namespace {
-
-constexpr u64 ELF_DYNAMIC_TAG_NULL = 0;
-constexpr u64 ELF_DYNAMIC_TAG_STRTAB = 5;
-constexpr u64 ELF_DYNAMIC_TAG_SYMTAB = 6;
-constexpr u64 ELF_DYNAMIC_TAG_SYMENT = 11;
-
-enum class ELFSymbolType : u8 {
-    None = 0,
-    Object = 1,
-    Function = 2,
-    Section = 3,
-    File = 4,
-    Common = 5,
-    TLS = 6,
-};
-
-enum class ELFSymbolBinding : u8 {
-    Local = 0,
-    Global = 1,
-    Weak = 2,
-};
-
-enum class ELFSymbolVisibility : u8 {
-    Default = 0,
-    Internal = 1,
-    Hidden = 2,
-    Protected = 3,
-};
-
-struct ELFSymbol {
-    u32 name_index;
-    union {
-        u8 info;
-
-        BitField<0, 4, ELFSymbolType> type;
-        BitField<4, 4, ELFSymbolBinding> binding;
-    };
-    ELFSymbolVisibility visibility;
-    u16 sh_index;
-    u64 value;
-    u64 size;
-};
-static_assert(sizeof(ELFSymbol) == 0x18, "ELFSymbol has incorrect size.");
-
-using Symbols = std::vector<std::pair<ELFSymbol, std::string>>;
-
-Symbols GetSymbols(VAddr text_offset, Core::Memory::Memory& memory) {
-    const auto mod_offset = text_offset + memory.Read32(text_offset + 4);
-
-    if (mod_offset < text_offset || (mod_offset & 0b11) != 0 ||
-        memory.Read32(mod_offset) != Common::MakeMagic('M', 'O', 'D', '0')) {
-        return {};
-    }
-
-    const auto dynamic_offset = memory.Read32(mod_offset + 0x4) + mod_offset;
-
-    VAddr string_table_offset{};
-    VAddr symbol_table_offset{};
-    u64 symbol_entry_size{};
-
-    VAddr dynamic_index = dynamic_offset;
-    while (true) {
-        const u64 tag = memory.Read64(dynamic_index);
-        const u64 value = memory.Read64(dynamic_index + 0x8);
-        dynamic_index += 0x10;
-
-        if (tag == ELF_DYNAMIC_TAG_NULL) {
-            break;
-        }
-
-        if (tag == ELF_DYNAMIC_TAG_STRTAB) {
-            string_table_offset = value;
-        } else if (tag == ELF_DYNAMIC_TAG_SYMTAB) {
-            symbol_table_offset = value;
-        } else if (tag == ELF_DYNAMIC_TAG_SYMENT) {
-            symbol_entry_size = value;
-        }
-    }
-
-    if (string_table_offset == 0 || symbol_table_offset == 0 || symbol_entry_size == 0) {
-        return {};
-    }
-
-    const auto string_table_address = text_offset + string_table_offset;
-    const auto symbol_table_address = text_offset + symbol_table_offset;
-
-    Symbols out;
-
-    VAddr symbol_index = symbol_table_address;
-    while (symbol_index < string_table_address) {
-        ELFSymbol symbol{};
-        memory.ReadBlock(symbol_index, &symbol, sizeof(ELFSymbol));
-
-        VAddr string_offset = string_table_address + symbol.name_index;
-        std::string name;
-        for (u8 c = memory.Read8(string_offset); c != 0; c = memory.Read8(++string_offset)) {
-            name += static_cast<char>(c);
-        }
-
-        symbol_index += symbol_entry_size;
-        out.push_back({symbol, name});
-    }
-
-    return out;
-}
-
-std::optional<std::string> GetSymbolName(const Symbols& symbols, VAddr func_address) {
-    const auto iter =
-        std::find_if(symbols.begin(), symbols.end(), [func_address](const auto& pair) {
-            const auto& symbol = pair.first;
-            const auto end_address = symbol.value + symbol.size;
-            return func_address >= symbol.value && func_address < end_address;
-        });
-
-    if (iter == symbols.end()) {
-        return std::nullopt;
-    }
-
-    return iter->second;
-}
-
-} // Anonymous namespace
 
 constexpr u64 SEGMENT_BASE = 0x7100000000ull;
 
@@ -169,9 +48,11 @@ std::vector<ARM_Interface::BacktraceEntry> ARM_Interface::GetBacktraceFromContex
         return {};
     }
 
-    std::map<std::string, Symbols> symbols;
+    std::map<std::string, Symbols::Symbols> symbols;
     for (const auto& module : modules) {
-        symbols.insert_or_assign(module.second, GetSymbols(module.first, memory));
+        symbols.insert_or_assign(module.second,
+                                 Symbols::GetSymbols(module.first, system.Memory(),
+                                                     system.CurrentProcess()->Is64BitProcess()));
     }
 
     for (auto& entry : out) {
@@ -193,7 +74,7 @@ std::vector<ARM_Interface::BacktraceEntry> ARM_Interface::GetBacktraceFromContex
 
         const auto symbol_set = symbols.find(entry.module);
         if (symbol_set != symbols.end()) {
-            const auto symbol = GetSymbolName(symbol_set->second, entry.offset);
+            const auto symbol = Symbols::GetSymbolName(symbol_set->second, entry.offset);
             if (symbol.has_value()) {
                 // TODO(DarkLordZach): Add demangling of symbol names.
                 entry.name = *symbol;
@@ -225,9 +106,11 @@ std::vector<ARM_Interface::BacktraceEntry> ARM_Interface::GetBacktrace() const {
         return {};
     }
 
-    std::map<std::string, Symbols> symbols;
+    std::map<std::string, Symbols::Symbols> symbols;
     for (const auto& module : modules) {
-        symbols.insert_or_assign(module.second, GetSymbols(module.first, memory));
+        symbols.insert_or_assign(module.second,
+                                 Symbols::GetSymbols(module.first, system.Memory(),
+                                                     system.CurrentProcess()->Is64BitProcess()));
     }
 
     for (auto& entry : out) {
@@ -249,7 +132,7 @@ std::vector<ARM_Interface::BacktraceEntry> ARM_Interface::GetBacktrace() const {
 
         const auto symbol_set = symbols.find(entry.module);
         if (symbol_set != symbols.end()) {
-            const auto symbol = GetSymbolName(symbol_set->second, entry.offset);
+            const auto symbol = Symbols::GetSymbolName(symbol_set->second, entry.offset);
             if (symbol.has_value()) {
                 // TODO(DarkLordZach): Add demangling of symbol names.
                 entry.name = *symbol;

src/core/arm/arm_interface.h

@@ -171,6 +171,9 @@ public:
     /// Prepare core for thread reschedule (if needed to correctly handle state)
     virtual void PrepareReschedule() = 0;
 
+    /// Signal an interrupt and ask the core to halt as soon as possible.
+    virtual void SignalInterrupt() = 0;
+
     struct BacktraceEntry {
         std::string module;
         u64 address;

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -25,6 +25,9 @@ namespace Core {
 
 using namespace Common::Literals;
 
+constexpr Dynarmic::HaltReason break_loop = Dynarmic::HaltReason::UserDefined2;
+constexpr Dynarmic::HaltReason svc_call = Dynarmic::HaltReason::UserDefined3;
+
 class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
 public:
     explicit DynarmicCallbacks32(ARM_Dynarmic_32& parent_)
@@ -70,11 +73,13 @@ public:
     }
 
     void InterpreterFallback(u32 pc, std::size_t num_instructions) override {
+        parent.LogBacktrace();
         UNIMPLEMENTED_MSG("This should never happen, pc = {:08X}, code = {:08X}", pc,
                           MemoryReadCode(pc));
     }
 
     void ExceptionRaised(u32 pc, Dynarmic::A32::Exception exception) override {
+        parent.LogBacktrace();
         LOG_CRITICAL(Core_ARM,
                      "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X}, thumb = {})",
                      exception, pc, MemoryReadCode(pc), parent.IsInThumbMode());
@@ -82,15 +87,13 @@ public:
     }
 
     void CallSVC(u32 swi) override {
-        parent.svc_called = true;
         parent.svc_swi = swi;
-        parent.jit->HaltExecution();
+        parent.jit->HaltExecution(svc_call);
     }
 
     void AddTicks(u64 ticks) override {
-        if (parent.uses_wall_clock) {
-            return;
-        }
+        ASSERT_MSG(!parent.uses_wall_clock, "This should never happen - dynarmic ticking disabled");
+
         // Divide the number of ticks by the amount of CPU cores. TODO(Subv): This yields only a
         // rough approximation of the amount of executed ticks in the system, it may be thrown off
         // if not all cores are doing a similar amount of work. Instead of doing this, we should
@@ -106,12 +109,8 @@ public:
     }
 
     u64 GetTicksRemaining() override {
-        if (parent.uses_wall_clock) {
-            if (!parent.interrupt_handlers[parent.core_index].IsInterrupted()) {
-                return minimum_run_cycles;
-            }
-            return 0U;
-        }
+        ASSERT_MSG(!parent.uses_wall_clock, "This should never happen - dynarmic ticking disabled");
+
         return std::max<s64>(parent.system.CoreTiming().GetDowncount(), 0);
     }
 
@@ -146,6 +145,7 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
 
     // Timing
     config.wall_clock_cntpct = uses_wall_clock;
+    config.enable_cycle_counting = !uses_wall_clock;
 
     // Code cache size
     config.code_cache_size = 512_MiB;
@@ -186,35 +186,41 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
         if (!Settings::values.cpuopt_recompile_exclusives) {
             config.recompile_on_exclusive_fastmem_failure = false;
         }
-    }
+    } else {
+        // Unsafe optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Unsafe) {
+            config.unsafe_optimizations = true;
+            if (Settings::values.cpuopt_unsafe_unfuse_fma) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
+            }
+            if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
+            }
+            if (Settings::values.cpuopt_unsafe_ignore_standard_fpcr) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreStandardFPCRValue;
+            }
+            if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
+            }
+            if (Settings::values.cpuopt_unsafe_ignore_global_monitor) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
+            }
+        }
 
-    // Unsafe optimizations
-    if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Unsafe) {
-        config.unsafe_optimizations = true;
-        if (Settings::values.cpuopt_unsafe_unfuse_fma) {
+        // Curated optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Auto) {
+            config.unsafe_optimizations = true;
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
-        }
-        if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
-            config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
-        }
-        if (Settings::values.cpuopt_unsafe_ignore_standard_fpcr) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreStandardFPCRValue;
-        }
-        if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
-        }
-        if (Settings::values.cpuopt_unsafe_ignore_global_monitor) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
         }
-    }
 
-    // Curated optimizations
-    if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Auto) {
-        config.unsafe_optimizations = true;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreStandardFPCRValue;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
+        // Paranoia mode for debugging optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Paranoid) {
+            config.unsafe_optimizations = false;
+            config.optimizations = Dynarmic::no_optimizations;
+        }
     }
 
     return std::make_unique<Dynarmic::A32::Jit>(config);
@@ -222,13 +228,11 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
 
 void ARM_Dynarmic_32::Run() {
     while (true) {
-        jit->Run();
-        if (!svc_called) {
-            break;
+        const auto hr = jit->Run();
+        if (Has(hr, svc_call)) {
+            Kernel::Svc::Call(system, svc_swi);
         }
-        svc_called = false;
-        Kernel::Svc::Call(system, svc_swi);
-        if (shutdown) {
+        if (Has(hr, break_loop) || !uses_wall_clock) {
             break;
         }
     }
@@ -314,8 +318,11 @@ void ARM_Dynarmic_32::LoadContext(const ThreadContext32& ctx) {
 }
 
 void ARM_Dynarmic_32::PrepareReschedule() {
-    jit->HaltExecution();
-    shutdown = true;
+    jit->HaltExecution(break_loop);
+}
+
+void ARM_Dynarmic_32::SignalInterrupt() {
+    jit->HaltExecution(break_loop);
 }
 
 void ARM_Dynarmic_32::ClearInstructionCache() {

src/core/arm/dynarmic/arm_dynarmic_32.h

@@ -57,6 +57,7 @@ public:
     void LoadContext(const ThreadContext64& ctx) override {}
 
     void PrepareReschedule() override;
+    void SignalInterrupt() override;
     void ClearExclusiveState() override;
 
     void ClearInstructionCache() override;
@@ -83,9 +84,6 @@ private:
 
     // SVC callback
     u32 svc_swi{};
-    bool svc_called{};
-
-    bool shutdown{};
 };
 
 } // namespace Core

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -26,6 +26,9 @@ namespace Core {
 using Vector = Dynarmic::A64::Vector;
 using namespace Common::Literals;
 
+constexpr Dynarmic::HaltReason break_loop = Dynarmic::HaltReason::UserDefined2;
+constexpr Dynarmic::HaltReason svc_call = Dynarmic::HaltReason::UserDefined3;
+
 class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
     explicit DynarmicCallbacks64(ARM_Dynarmic_64& parent_)
@@ -81,6 +84,7 @@ public:
     }
 
     void InterpreterFallback(u64 pc, std::size_t num_instructions) override {
+        parent.LogBacktrace();
         LOG_ERROR(Core_ARM,
                   "Unimplemented instruction @ 0x{:X} for {} instructions (instr = {:08X})", pc,
                   num_instructions, MemoryReadCode(pc));
@@ -105,7 +109,7 @@ public:
             break;
         }
 
-        parent.jit->HaltExecution();
+        parent.jit->HaltExecution(Dynarmic::HaltReason::CacheInvalidation);
     }
 
     void ExceptionRaised(u64 pc, Dynarmic::A64::Exception exception) override {
@@ -118,21 +122,19 @@ public:
             return;
         case Dynarmic::A64::Exception::Breakpoint:
         default:
+            parent.LogBacktrace();
             ASSERT_MSG(false, "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X})",
                        static_cast<std::size_t>(exception), pc, MemoryReadCode(pc));
         }
     }
 
     void CallSVC(u32 swi) override {
-        parent.svc_called = true;
         parent.svc_swi = swi;
-        parent.jit->HaltExecution();
+        parent.jit->HaltExecution(svc_call);
     }
 
     void AddTicks(u64 ticks) override {
-        if (parent.uses_wall_clock) {
-            return;
-        }
+        ASSERT_MSG(!parent.uses_wall_clock, "This should never happen - dynarmic ticking disabled");
 
         // Divide the number of ticks by the amount of CPU cores. TODO(Subv): This yields only a
         // rough approximation of the amount of executed ticks in the system, it may be thrown off
@@ -147,12 +149,8 @@ public:
     }
 
     u64 GetTicksRemaining() override {
-        if (parent.uses_wall_clock) {
-            if (!parent.interrupt_handlers[parent.core_index].IsInterrupted()) {
-                return minimum_run_cycles;
-            }
-            return 0U;
-        }
+        ASSERT_MSG(!parent.uses_wall_clock, "This should never happen - dynarmic ticking disabled");
+
         return std::max<s64>(parent.system.CoreTiming().GetDowncount(), 0);
     }
 
@@ -208,6 +206,7 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
 
     // Timing
     config.wall_clock_cntpct = uses_wall_clock;
+    config.enable_cycle_counting = !uses_wall_clock;
 
     // Code cache size
     config.code_cache_size = 512_MiB;
@@ -248,35 +247,41 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
         if (!Settings::values.cpuopt_recompile_exclusives) {
             config.recompile_on_exclusive_fastmem_failure = false;
         }
-    }
+    } else {
+        // Unsafe optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Unsafe) {
+            config.unsafe_optimizations = true;
+            if (Settings::values.cpuopt_unsafe_unfuse_fma) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
+            }
+            if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
+            }
+            if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
+            }
+            if (Settings::values.cpuopt_unsafe_fastmem_check) {
+                config.fastmem_address_space_bits = 64;
+            }
+            if (Settings::values.cpuopt_unsafe_ignore_global_monitor) {
+                config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
+            }
+        }
 
-    // Unsafe optimizations
-    if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Unsafe) {
-        config.unsafe_optimizations = true;
-        if (Settings::values.cpuopt_unsafe_unfuse_fma) {
+        // Curated optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Auto) {
+            config.unsafe_optimizations = true;
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
-        }
-        if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
-            config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
-        }
-        if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
-        }
-        if (Settings::values.cpuopt_unsafe_fastmem_check) {
             config.fastmem_address_space_bits = 64;
-        }
-        if (Settings::values.cpuopt_unsafe_ignore_global_monitor) {
             config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
         }
-    }
 
-    // Curated optimizations
-    if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Auto) {
-        config.unsafe_optimizations = true;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
-        config.fastmem_address_space_bits = 64;
-        config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor;
+        // Paranoia mode for debugging optimizations
+        if (Settings::values.cpu_accuracy.GetValue() == Settings::CPUAccuracy::Paranoid) {
+            config.unsafe_optimizations = false;
+            config.optimizations = Dynarmic::no_optimizations;
+        }
     }
 
     return std::make_shared<Dynarmic::A64::Jit>(config);
@@ -284,13 +289,11 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
 
 void ARM_Dynarmic_64::Run() {
     while (true) {
-        jit->Run();
-        if (!svc_called) {
-            break;
+        const auto hr = jit->Run();
+        if (Has(hr, svc_call)) {
+            Kernel::Svc::Call(system, svc_swi);
         }
-        svc_called = false;
-        Kernel::Svc::Call(system, svc_swi);
-        if (shutdown) {
+        if (Has(hr, break_loop) || !uses_wall_clock) {
             break;
         }
     }
@@ -381,8 +384,11 @@ void ARM_Dynarmic_64::LoadContext(const ThreadContext64& ctx) {
 }
 
 void ARM_Dynarmic_64::PrepareReschedule() {
-    jit->HaltExecution();
-    shutdown = true;
+    jit->HaltExecution(break_loop);
+}
+
+void ARM_Dynarmic_64::SignalInterrupt() {
+    jit->HaltExecution(break_loop);
 }
 
 void ARM_Dynarmic_64::ClearInstructionCache() {

src/core/arm/dynarmic/arm_dynarmic_64.h

@@ -51,6 +51,7 @@ public:
     void LoadContext(const ThreadContext64& ctx) override;
 
     void PrepareReschedule() override;
+    void SignalInterrupt() override;
     void ClearExclusiveState() override;
 
     void ClearInstructionCache() override;
@@ -77,9 +78,6 @@ private:
 
     // SVC callback
     u32 svc_swi{};
-    bool svc_called{};
-
-    bool shutdown{};
 };
 
 } // namespace Core

src/core/arm/symbols.cpp

@@ -0,0 +1,190 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/bit_field.h"
+#include "common/common_funcs.h"
+#include "core/arm/symbols.h"
+#include "core/core.h"
+#include "core/memory.h"
+
+namespace Core {
+namespace {
+
+constexpr u64 ELF_DYNAMIC_TAG_NULL = 0;
+constexpr u64 ELF_DYNAMIC_TAG_STRTAB = 5;
+constexpr u64 ELF_DYNAMIC_TAG_SYMTAB = 6;
+constexpr u64 ELF_DYNAMIC_TAG_SYMENT = 11;
+
+enum class ELFSymbolType : u8 {
+    None = 0,
+    Object = 1,
+    Function = 2,
+    Section = 3,
+    File = 4,
+    Common = 5,
+    TLS = 6,
+};
+
+enum class ELFSymbolBinding : u8 {
+    Local = 0,
+    Global = 1,
+    Weak = 2,
+};
+
+enum class ELFSymbolVisibility : u8 {
+    Default = 0,
+    Internal = 1,
+    Hidden = 2,
+    Protected = 3,
+};
+
+struct ELF64Symbol {
+    u32 name_index;
+    union {
+        u8 info;
+
+        BitField<0, 4, ELFSymbolType> type;
+        BitField<4, 4, ELFSymbolBinding> binding;
+    };
+    ELFSymbolVisibility visibility;
+    u16 sh_index;
+    u64 value;
+    u64 size;
+};
+static_assert(sizeof(ELF64Symbol) == 0x18, "ELF64Symbol has incorrect size.");
+
+struct ELF32Symbol {
+    u32 name_index;
+    u32 value;
+    u32 size;
+    union {
+        u8 info;
+
+        BitField<0, 4, ELFSymbolType> type;
+        BitField<4, 4, ELFSymbolBinding> binding;
+    };
+    ELFSymbolVisibility visibility;
+    u16 sh_index;
+};
+static_assert(sizeof(ELF32Symbol) == 0x10, "ELF32Symbol has incorrect size.");
+
+} // Anonymous namespace
+
+namespace Symbols {
+
+template <typename Word, typename ELFSymbol, typename ByteReader>
+static Symbols GetSymbols(ByteReader ReadBytes) {
+    const auto Read8{[&](u64 index) {
+        u8 ret;
+        ReadBytes(&ret, index, sizeof(u8));
+        return ret;
+    }};
+
+    const auto Read32{[&](u64 index) {
+        u32 ret;
+        ReadBytes(&ret, index, sizeof(u32));
+        return ret;
+    }};
+
+    const auto ReadWord{[&](u64 index) {
+        Word ret;
+        ReadBytes(&ret, index, sizeof(Word));
+        return ret;
+    }};
+
+    const u32 mod_offset = Read32(4);
+
+    if (Read32(mod_offset) != Common::MakeMagic('M', 'O', 'D', '0')) {
+        return {};
+    }
+
+    VAddr string_table_offset{};
+    VAddr symbol_table_offset{};
+    u64 symbol_entry_size{};
+
+    const auto dynamic_offset = Read32(mod_offset + 0x4) + mod_offset;
+
+    VAddr dynamic_index = dynamic_offset;
+    while (true) {
+        const Word tag = ReadWord(dynamic_index);
+        const Word value = ReadWord(dynamic_index + sizeof(Word));
+        dynamic_index += 2 * sizeof(Word);
+
+        if (tag == ELF_DYNAMIC_TAG_NULL) {
+            break;
+        }
+
+        if (tag == ELF_DYNAMIC_TAG_STRTAB) {
+            string_table_offset = value;
+        } else if (tag == ELF_DYNAMIC_TAG_SYMTAB) {
+            symbol_table_offset = value;
+        } else if (tag == ELF_DYNAMIC_TAG_SYMENT) {
+            symbol_entry_size = value;
+        }
+    }
+
+    if (string_table_offset == 0 || symbol_table_offset == 0 || symbol_entry_size == 0) {
+        return {};
+    }
+
+    Symbols out;
+
+    VAddr symbol_index = symbol_table_offset;
+    while (symbol_index < string_table_offset) {
+        ELFSymbol symbol{};
+        ReadBytes(&symbol, symbol_index, sizeof(ELFSymbol));
+
+        VAddr string_offset = string_table_offset + symbol.name_index;
+        std::string name;
+        for (u8 c = Read8(string_offset); c != 0; c = Read8(++string_offset)) {
+            name += static_cast<char>(c);
+        }
+
+        symbol_index += symbol_entry_size;
+        out[name] = std::make_pair(symbol.value, symbol.size);
+    }
+
+    return out;
+}
+
+Symbols GetSymbols(VAddr base, Core::Memory::Memory& memory, bool is_64) {
+    const auto ReadBytes{
+        [&](void* ptr, size_t offset, size_t size) { memory.ReadBlock(base + offset, ptr, size); }};
+
+    if (is_64) {
+        return GetSymbols<u64, ELF64Symbol>(ReadBytes);
+    } else {
+        return GetSymbols<u32, ELF32Symbol>(ReadBytes);
+    }
+}
+
+Symbols GetSymbols(std::span<const u8> data, bool is_64) {
+    const auto ReadBytes{[&](void* ptr, size_t offset, size_t size) {
+        std::memcpy(ptr, data.data() + offset, size);
+    }};
+
+    if (is_64) {
+        return GetSymbols<u64, ELF64Symbol>(ReadBytes);
+    } else {
+        return GetSymbols<u32, ELF32Symbol>(ReadBytes);
+    }
+}
+
+std::optional<std::string> GetSymbolName(const Symbols& symbols, VAddr addr) {
+    const auto iter = std::find_if(symbols.cbegin(), symbols.cend(), [addr](const auto& pair) {
+        const auto& [name, sym_info] = pair;
+        const auto& [start_address, size] = sym_info;
+        const auto end_address = start_address + size;
+        return addr >= start_address && addr < end_address;
+    });
+
+    if (iter == symbols.cend()) {
+        return std::nullopt;
+    }
+
+    return iter->first;
+}
+
+} // namespace Symbols
+} // namespace Core

src/core/arm/symbols.h

@@ -0,0 +1,27 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <map>
+#include <optional>
+#include <span>
+#include <string>
+#include <utility>
+
+#include "common/common_types.h"
+
+namespace Core::Memory {
+class Memory;
+} // namespace Core::Memory
+
+namespace Core::Symbols {
+
+using Symbols = std::map<std::string, std::pair<VAddr, std::size_t>, std::less<>>;
+
+Symbols GetSymbols(VAddr base, Core::Memory::Memory& memory, bool is_64 = true);
+Symbols GetSymbols(std::span<const u8> data, bool is_64 = true);
+std::optional<std::string> GetSymbolName(const Symbols& symbols, VAddr addr);
+
+} // namespace Core::Symbols

src/core/file_sys/patch_manager.cpp

@@ -148,29 +148,33 @@ VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
 
     // LayeredExeFS
     const auto load_dir = fs_controller.GetModificationLoadRoot(title_id);
+    const auto sdmc_load_dir = fs_controller.GetSDMCModificationLoadRoot(title_id);
+
+    std::vector<VirtualDir> patch_dirs = {sdmc_load_dir};
     if (load_dir != nullptr && load_dir->GetSize() > 0) {
-        auto patch_dirs = load_dir->GetSubdirectories();
-        std::sort(
-            patch_dirs.begin(), patch_dirs.end(),
-            [](const VirtualDir& l, const VirtualDir& r) { return l->GetName() < r->GetName(); });
+        const auto load_patch_dirs = load_dir->GetSubdirectories();
+        patch_dirs.insert(patch_dirs.end(), load_patch_dirs.begin(), load_patch_dirs.end());
+    }
 
-        std::vector<VirtualDir> layers;
-        layers.reserve(patch_dirs.size() + 1);
-        for (const auto& subdir : patch_dirs) {
-            if (std::find(disabled.begin(), disabled.end(), subdir->GetName()) != disabled.end())
-                continue;
+    std::sort(patch_dirs.begin(), patch_dirs.end(),
+              [](const VirtualDir& l, const VirtualDir& r) { return l->GetName() < r->GetName(); });
 
-            auto exefs_dir = FindSubdirectoryCaseless(subdir, "exefs");
-            if (exefs_dir != nullptr)
-                layers.push_back(std::move(exefs_dir));
-        }
-        layers.push_back(exefs);
+    std::vector<VirtualDir> layers;
+    layers.reserve(patch_dirs.size() + 1);
+    for (const auto& subdir : patch_dirs) {
+        if (std::find(disabled.begin(), disabled.end(), subdir->GetName()) != disabled.end())
+            continue;
 
-        auto layered = LayeredVfsDirectory::MakeLayeredDirectory(std::move(layers));
-        if (layered != nullptr) {
-            LOG_INFO(Loader, "    ExeFS: LayeredExeFS patches applied successfully");
-            exefs = std::move(layered);
-        }
+        auto exefs_dir = FindSubdirectoryCaseless(subdir, "exefs");
+        if (exefs_dir != nullptr)
+            layers.push_back(std::move(exefs_dir));
+    }
+    layers.push_back(exefs);
+
+    auto layered = LayeredVfsDirectory::MakeLayeredDirectory(std::move(layers));
+    if (layered != nullptr) {
+        LOG_INFO(Loader, "    ExeFS: LayeredExeFS patches applied successfully");
+        exefs = std::move(layered);
     }
 
     if (Settings::values.dump_exefs) {
@@ -536,11 +540,20 @@ PatchManager::PatchVersionNames PatchManager::GetPatchVersionNames(VirtualFile u
 
     // SDMC mod directory (RomFS LayeredFS)
     const auto sdmc_mod_dir = fs_controller.GetSDMCModificationLoadRoot(title_id);
-    if (sdmc_mod_dir != nullptr && sdmc_mod_dir->GetSize() > 0 &&
-        IsDirValidAndNonEmpty(FindSubdirectoryCaseless(sdmc_mod_dir, "romfs"))) {
-        const auto mod_disabled =
-            std::find(disabled.begin(), disabled.end(), "SDMC") != disabled.end();
-        out.insert_or_assign(mod_disabled ? "[D] SDMC" : "SDMC", "LayeredFS");
+    if (sdmc_mod_dir != nullptr && sdmc_mod_dir->GetSize() > 0) {
+        std::string types;
+        if (IsDirValidAndNonEmpty(FindSubdirectoryCaseless(sdmc_mod_dir, "exefs"))) {
+            AppendCommaIfNotEmpty(types, "LayeredExeFS");
+        }
+        if (IsDirValidAndNonEmpty(FindSubdirectoryCaseless(sdmc_mod_dir, "romfs"))) {
+            AppendCommaIfNotEmpty(types, "LayeredFS");
+        }
+
+        if (!types.empty()) {
+            const auto mod_disabled =
+                std::find(disabled.begin(), disabled.end(), "SDMC") != disabled.end();
+            out.insert_or_assign(mod_disabled ? "[D] SDMC" : "SDMC", types);
+        }
     }
 
     // DLC

src/core/file_sys/registered_cache.cpp

@@ -387,15 +387,17 @@ std::vector<NcaID> RegisteredCache::AccumulateFiles() const {
             continue;
 
         for (const auto& nca_dir : d2_dir->GetSubdirectories()) {
-            if (!FollowsNcaIdFormat(nca_dir->GetName()))
+            if (nca_dir == nullptr || !FollowsNcaIdFormat(nca_dir->GetName())) {
                 continue;
+            }
 
             ids.push_back(Common::HexStringToArray<0x10, true>(nca_dir->GetName().substr(0, 0x20)));
         }
 
         for (const auto& nca_file : d2_dir->GetFiles()) {
-            if (!FollowsNcaIdFormat(nca_file->GetName()))
+            if (nca_file == nullptr || !FollowsNcaIdFormat(nca_file->GetName())) {
                 continue;
+            }
 
             ids.push_back(
                 Common::HexStringToArray<0x10, true>(nca_file->GetName().substr(0, 0x20)));

src/core/hid/emulated_console.cpp

@@ -132,7 +132,7 @@ void EmulatedConsole::SetMotionParam(Common::ParamPackage param) {
 }
 
 void EmulatedConsole::SetMotion(const Common::Input::CallbackStatus& callback) {
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     auto& raw_status = console.motion_values.raw_status;
     auto& emulated = console.motion_values.emulated;
 
@@ -151,6 +151,7 @@ void EmulatedConsole::SetMotion(const Common::Input::CallbackStatus& callback) {
     emulated.UpdateOrientation(raw_status.delta_timestamp);
 
     if (is_configuring) {
+        lock.unlock();
         TriggerOnChange(ConsoleTriggerType::Motion);
         return;
     }
@@ -166,6 +167,7 @@ void EmulatedConsole::SetMotion(const Common::Input::CallbackStatus& callback) {
     // Find what is this value
     motion.verticalization_error = 0.0f;
 
+    lock.unlock();
     TriggerOnChange(ConsoleTriggerType::Motion);
 }
 
@@ -173,11 +175,12 @@ void EmulatedConsole::SetTouch(const Common::Input::CallbackStatus& callback, st
     if (index >= console.touch_values.size()) {
         return;
     }
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
 
     console.touch_values[index] = TransformToTouch(callback);
 
     if (is_configuring) {
+        lock.unlock();
         TriggerOnChange(ConsoleTriggerType::Touch);
         return;
     }
@@ -189,26 +192,32 @@ void EmulatedConsole::SetTouch(const Common::Input::CallbackStatus& callback, st
         .pressed = console.touch_values[index].pressed.value,
     };
 
+    lock.unlock();
     TriggerOnChange(ConsoleTriggerType::Touch);
 }
 
 ConsoleMotionValues EmulatedConsole::GetMotionValues() const {
+    std::scoped_lock lock{mutex};
     return console.motion_values;
 }
 
 TouchValues EmulatedConsole::GetTouchValues() const {
+    std::scoped_lock lock{mutex};
     return console.touch_values;
 }
 
 ConsoleMotion EmulatedConsole::GetMotion() const {
+    std::scoped_lock lock{mutex};
     return console.motion_state;
 }
 
 TouchFingerState EmulatedConsole::GetTouch() const {
+    std::scoped_lock lock{mutex};
     return console.touch_state;
 }
 
 void EmulatedConsole::TriggerOnChange(ConsoleTriggerType type) {
+    std::scoped_lock lock{callback_mutex};
     for (const auto& poller_pair : callback_list) {
         const ConsoleUpdateCallback& poller = poller_pair.second;
         if (poller.on_change) {
@@ -218,13 +227,13 @@ void EmulatedConsole::TriggerOnChange(ConsoleTriggerType type) {
 }
 
 int EmulatedConsole::SetCallback(ConsoleUpdateCallback update_callback) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{callback_mutex};
     callback_list.insert_or_assign(last_callback_key, update_callback);
     return last_callback_key++;
 }
 
 void EmulatedConsole::DeleteCallback(int key) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{callback_mutex};
     const auto& iterator = callback_list.find(key);
     if (iterator == callback_list.end()) {
         LOG_ERROR(Input, "Tried to delete non-existent callback {}", key);

src/core/hid/emulated_console.h

@@ -183,6 +183,7 @@ private:
     TouchDevices touch_devices;
 
     mutable std::mutex mutex;
+    mutable std::mutex callback_mutex;
     std::unordered_map<int, ConsoleUpdateCallback> callback_list;
     int last_callback_key = 0;
 

src/core/hid/emulated_controller.cpp

@@ -353,14 +353,17 @@ void EmulatedController::DisableConfiguration() {
 }
 
 void EmulatedController::EnableSystemButtons() {
+    std::scoped_lock lock{mutex};
     system_buttons_enabled = true;
 }
 
 void EmulatedController::DisableSystemButtons() {
+    std::scoped_lock lock{mutex};
     system_buttons_enabled = false;
 }
 
 void EmulatedController::ResetSystemButtons() {
+    std::scoped_lock lock{mutex};
     controller.home_button_state.home.Assign(false);
     controller.capture_button_state.capture.Assign(false);
 }
@@ -494,139 +497,141 @@ void EmulatedController::SetButton(const Common::Input::CallbackStatus& callback
     if (index >= controller.button_values.size()) {
         return;
     }
-    {
-        std::lock_guard lock{mutex};
-        bool value_changed = false;
-        const auto new_status = TransformToButton(callback);
-        auto& current_status = controller.button_values[index];
+    std::unique_lock lock{mutex};
+    bool value_changed = false;
+    const auto new_status = TransformToButton(callback);
+    auto& current_status = controller.button_values[index];
 
-        // Only read button values that have the same uuid or are pressed once
-        if (current_status.uuid != uuid) {
-            if (!new_status.value) {
-                return;
-            }
-        }
-
-        current_status.toggle = new_status.toggle;
-        current_status.uuid = uuid;
-
-        // Update button status with current
-        if (!current_status.toggle) {
-            current_status.locked = false;
-            if (current_status.value != new_status.value) {
-                current_status.value = new_status.value;
-                value_changed = true;
-            }
-        } else {
-            // Toggle button and lock status
-            if (new_status.value && !current_status.locked) {
-                current_status.locked = true;
-                current_status.value = !current_status.value;
-                value_changed = true;
-            }
-
-            // Unlock button ready for next press
-            if (!new_status.value && current_status.locked) {
-                current_status.locked = false;
-            }
-        }
-
-        if (!value_changed) {
+    // Only read button values that have the same uuid or are pressed once
+    if (current_status.uuid != uuid) {
+        if (!new_status.value) {
             return;
         }
-
-        if (is_configuring) {
-            controller.npad_button_state.raw = NpadButton::None;
-            controller.debug_pad_button_state.raw = 0;
-            TriggerOnChange(ControllerTriggerType::Button, false);
-            return;
-        }
-
-        switch (index) {
-        case Settings::NativeButton::A:
-            controller.npad_button_state.a.Assign(current_status.value);
-            controller.debug_pad_button_state.a.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::B:
-            controller.npad_button_state.b.Assign(current_status.value);
-            controller.debug_pad_button_state.b.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::X:
-            controller.npad_button_state.x.Assign(current_status.value);
-            controller.debug_pad_button_state.x.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::Y:
-            controller.npad_button_state.y.Assign(current_status.value);
-            controller.debug_pad_button_state.y.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::LStick:
-            controller.npad_button_state.stick_l.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::RStick:
-            controller.npad_button_state.stick_r.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::L:
-            controller.npad_button_state.l.Assign(current_status.value);
-            controller.debug_pad_button_state.l.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::R:
-            controller.npad_button_state.r.Assign(current_status.value);
-            controller.debug_pad_button_state.r.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::ZL:
-            controller.npad_button_state.zl.Assign(current_status.value);
-            controller.debug_pad_button_state.zl.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::ZR:
-            controller.npad_button_state.zr.Assign(current_status.value);
-            controller.debug_pad_button_state.zr.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::Plus:
-            controller.npad_button_state.plus.Assign(current_status.value);
-            controller.debug_pad_button_state.plus.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::Minus:
-            controller.npad_button_state.minus.Assign(current_status.value);
-            controller.debug_pad_button_state.minus.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::DLeft:
-            controller.npad_button_state.left.Assign(current_status.value);
-            controller.debug_pad_button_state.d_left.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::DUp:
-            controller.npad_button_state.up.Assign(current_status.value);
-            controller.debug_pad_button_state.d_up.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::DRight:
-            controller.npad_button_state.right.Assign(current_status.value);
-            controller.debug_pad_button_state.d_right.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::DDown:
-            controller.npad_button_state.down.Assign(current_status.value);
-            controller.debug_pad_button_state.d_down.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::SL:
-            controller.npad_button_state.left_sl.Assign(current_status.value);
-            controller.npad_button_state.right_sl.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::SR:
-            controller.npad_button_state.left_sr.Assign(current_status.value);
-            controller.npad_button_state.right_sr.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::Home:
-            if (!system_buttons_enabled) {
-                break;
-            }
-            controller.home_button_state.home.Assign(current_status.value);
-            break;
-        case Settings::NativeButton::Screenshot:
-            if (!system_buttons_enabled) {
-                break;
-            }
-            controller.capture_button_state.capture.Assign(current_status.value);
-            break;
-        }
     }
+
+    current_status.toggle = new_status.toggle;
+    current_status.uuid = uuid;
+
+    // Update button status with current
+    if (!current_status.toggle) {
+        current_status.locked = false;
+        if (current_status.value != new_status.value) {
+            current_status.value = new_status.value;
+            value_changed = true;
+        }
+    } else {
+        // Toggle button and lock status
+        if (new_status.value && !current_status.locked) {
+            current_status.locked = true;
+            current_status.value = !current_status.value;
+            value_changed = true;
+        }
+
+        // Unlock button ready for next press
+        if (!new_status.value && current_status.locked) {
+            current_status.locked = false;
+        }
+    }
+
+    if (!value_changed) {
+        return;
+    }
+
+    if (is_configuring) {
+        controller.npad_button_state.raw = NpadButton::None;
+        controller.debug_pad_button_state.raw = 0;
+        lock.unlock();
+        TriggerOnChange(ControllerTriggerType::Button, false);
+        return;
+    }
+
+    switch (index) {
+    case Settings::NativeButton::A:
+        controller.npad_button_state.a.Assign(current_status.value);
+        controller.debug_pad_button_state.a.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::B:
+        controller.npad_button_state.b.Assign(current_status.value);
+        controller.debug_pad_button_state.b.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::X:
+        controller.npad_button_state.x.Assign(current_status.value);
+        controller.debug_pad_button_state.x.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::Y:
+        controller.npad_button_state.y.Assign(current_status.value);
+        controller.debug_pad_button_state.y.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::LStick:
+        controller.npad_button_state.stick_l.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::RStick:
+        controller.npad_button_state.stick_r.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::L:
+        controller.npad_button_state.l.Assign(current_status.value);
+        controller.debug_pad_button_state.l.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::R:
+        controller.npad_button_state.r.Assign(current_status.value);
+        controller.debug_pad_button_state.r.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::ZL:
+        controller.npad_button_state.zl.Assign(current_status.value);
+        controller.debug_pad_button_state.zl.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::ZR:
+        controller.npad_button_state.zr.Assign(current_status.value);
+        controller.debug_pad_button_state.zr.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::Plus:
+        controller.npad_button_state.plus.Assign(current_status.value);
+        controller.debug_pad_button_state.plus.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::Minus:
+        controller.npad_button_state.minus.Assign(current_status.value);
+        controller.debug_pad_button_state.minus.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::DLeft:
+        controller.npad_button_state.left.Assign(current_status.value);
+        controller.debug_pad_button_state.d_left.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::DUp:
+        controller.npad_button_state.up.Assign(current_status.value);
+        controller.debug_pad_button_state.d_up.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::DRight:
+        controller.npad_button_state.right.Assign(current_status.value);
+        controller.debug_pad_button_state.d_right.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::DDown:
+        controller.npad_button_state.down.Assign(current_status.value);
+        controller.debug_pad_button_state.d_down.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::SL:
+        controller.npad_button_state.left_sl.Assign(current_status.value);
+        controller.npad_button_state.right_sl.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::SR:
+        controller.npad_button_state.left_sr.Assign(current_status.value);
+        controller.npad_button_state.right_sr.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::Home:
+        if (!system_buttons_enabled) {
+            break;
+        }
+        controller.home_button_state.home.Assign(current_status.value);
+        break;
+    case Settings::NativeButton::Screenshot:
+        if (!system_buttons_enabled) {
+            break;
+        }
+        controller.capture_button_state.capture.Assign(current_status.value);
+        break;
+    }
+
+    lock.unlock();
+
     if (!is_connected) {
         if (npad_id_type == NpadIdType::Player1 && npad_type != NpadStyleIndex::Handheld) {
             Connect();
@@ -643,7 +648,7 @@ void EmulatedController::SetStick(const Common::Input::CallbackStatus& callback,
     if (index >= controller.stick_values.size()) {
         return;
     }
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     const auto stick_value = TransformToStick(callback);
 
     // Only read stick values that have the same uuid or are over the threshold to avoid flapping
@@ -659,6 +664,7 @@ void EmulatedController::SetStick(const Common::Input::CallbackStatus& callback,
     if (is_configuring) {
         controller.analog_stick_state.left = {};
         controller.analog_stick_state.right = {};
+        lock.unlock();
         TriggerOnChange(ControllerTriggerType::Stick, false);
         return;
     }
@@ -685,6 +691,7 @@ void EmulatedController::SetStick(const Common::Input::CallbackStatus& callback,
         break;
     }
 
+    lock.unlock();
     TriggerOnChange(ControllerTriggerType::Stick, true);
 }
 
@@ -693,7 +700,7 @@ void EmulatedController::SetTrigger(const Common::Input::CallbackStatus& callbac
     if (index >= controller.trigger_values.size()) {
         return;
     }
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     const auto trigger_value = TransformToTrigger(callback);
 
     // Only read trigger values that have the same uuid or are pressed once
@@ -709,6 +716,7 @@ void EmulatedController::SetTrigger(const Common::Input::CallbackStatus& callbac
     if (is_configuring) {
         controller.gc_trigger_state.left = 0;
         controller.gc_trigger_state.right = 0;
+        lock.unlock();
         TriggerOnChange(ControllerTriggerType::Trigger, false);
         return;
     }
@@ -727,6 +735,7 @@ void EmulatedController::SetTrigger(const Common::Input::CallbackStatus& callbac
         break;
     }
 
+    lock.unlock();
     TriggerOnChange(ControllerTriggerType::Trigger, true);
 }
 
@@ -735,7 +744,7 @@ void EmulatedController::SetMotion(const Common::Input::CallbackStatus& callback
     if (index >= controller.motion_values.size()) {
         return;
     }
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     auto& raw_status = controller.motion_values[index].raw_status;
     auto& emulated = controller.motion_values[index].emulated;
 
@@ -756,6 +765,7 @@ void EmulatedController::SetMotion(const Common::Input::CallbackStatus& callback
     force_update_motion = raw_status.force_update;
 
     if (is_configuring) {
+        lock.unlock();
         TriggerOnChange(ControllerTriggerType::Motion, false);
         return;
     }
@@ -767,6 +777,7 @@ void EmulatedController::SetMotion(const Common::Input::CallbackStatus& callback
     motion.orientation = emulated.GetOrientation();
     motion.is_at_rest = !emulated.IsMoving(motion_sensitivity);
 
+    lock.unlock();
     TriggerOnChange(ControllerTriggerType::Motion, true);
 }
 
@@ -775,10 +786,11 @@ void EmulatedController::SetBattery(const Common::Input::CallbackStatus& callbac
     if (index >= controller.battery_values.size()) {
         return;
     }
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     controller.battery_values[index] = TransformToBattery(callback);
 
     if (is_configuring) {
+        lock.unlock();
         TriggerOnChange(ControllerTriggerType::Battery, false);
         return;
     }
@@ -835,6 +847,8 @@ void EmulatedController::SetBattery(const Common::Input::CallbackStatus& callbac
         };
         break;
     }
+
+    lock.unlock();
     TriggerOnChange(ControllerTriggerType::Battery, true);
 }
 
@@ -932,6 +946,7 @@ void EmulatedController::SetSupportedNpadStyleTag(NpadStyleTag supported_styles)
 }
 
 bool EmulatedController::IsControllerFullkey(bool use_temporary_value) const {
+    std::scoped_lock lock{mutex};
     const auto type = is_configuring && use_temporary_value ? tmp_npad_type : npad_type;
     switch (type) {
     case NpadStyleIndex::ProController:
@@ -947,6 +962,7 @@ bool EmulatedController::IsControllerFullkey(bool use_temporary_value) const {
 }
 
 bool EmulatedController::IsControllerSupported(bool use_temporary_value) const {
+    std::scoped_lock lock{mutex};
     const auto type = is_configuring && use_temporary_value ? tmp_npad_type : npad_type;
     switch (type) {
     case NpadStyleIndex::ProController:
@@ -982,40 +998,44 @@ void EmulatedController::Connect(bool use_temporary_value) {
         LOG_ERROR(Service_HID, "Controller type {} is not supported", type);
         return;
     }
-    {
-        std::lock_guard lock{mutex};
-        if (is_configuring) {
-            tmp_is_connected = true;
-            TriggerOnChange(ControllerTriggerType::Connected, false);
-            return;
-        }
 
-        if (is_connected) {
-            return;
-        }
-        is_connected = true;
+    std::unique_lock lock{mutex};
+    if (is_configuring) {
+        tmp_is_connected = true;
+        lock.unlock();
+        TriggerOnChange(ControllerTriggerType::Connected, false);
+        return;
     }
+
+    if (is_connected) {
+        return;
+    }
+    is_connected = true;
+
+    lock.unlock();
     TriggerOnChange(ControllerTriggerType::Connected, true);
 }
 
 void EmulatedController::Disconnect() {
-    {
-        std::lock_guard lock{mutex};
-        if (is_configuring) {
-            tmp_is_connected = false;
-            TriggerOnChange(ControllerTriggerType::Disconnected, false);
-            return;
-        }
-
-        if (!is_connected) {
-            return;
-        }
-        is_connected = false;
+    std::unique_lock lock{mutex};
+    if (is_configuring) {
+        tmp_is_connected = false;
+        lock.unlock();
+        TriggerOnChange(ControllerTriggerType::Disconnected, false);
+        return;
     }
+
+    if (!is_connected) {
+        return;
+    }
+    is_connected = false;
+
+    lock.unlock();
     TriggerOnChange(ControllerTriggerType::Disconnected, true);
 }
 
 bool EmulatedController::IsConnected(bool get_temporary_value) const {
+    std::scoped_lock lock{mutex};
     if (get_temporary_value && is_configuring) {
         return tmp_is_connected;
     }
@@ -1029,10 +1049,12 @@ bool EmulatedController::IsVibrationEnabled() const {
 }
 
 NpadIdType EmulatedController::GetNpadIdType() const {
+    std::scoped_lock lock{mutex};
     return npad_id_type;
 }
 
 NpadStyleIndex EmulatedController::GetNpadStyleIndex(bool get_temporary_value) const {
+    std::scoped_lock lock{mutex};
     if (get_temporary_value && is_configuring) {
         return tmp_npad_type;
     }
@@ -1040,27 +1062,28 @@ NpadStyleIndex EmulatedController::GetNpadStyleIndex(bool get_temporary_value) c
 }
 
 void EmulatedController::SetNpadStyleIndex(NpadStyleIndex npad_type_) {
-    {
-        std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
 
-        if (is_configuring) {
-            if (tmp_npad_type == npad_type_) {
-                return;
-            }
-            tmp_npad_type = npad_type_;
-            TriggerOnChange(ControllerTriggerType::Type, false);
+    if (is_configuring) {
+        if (tmp_npad_type == npad_type_) {
             return;
         }
-
-        if (npad_type == npad_type_) {
-            return;
-        }
-        if (is_connected) {
-            LOG_WARNING(Service_HID, "Controller {} type changed while it's connected",
-                        NpadIdTypeToIndex(npad_id_type));
-        }
-        npad_type = npad_type_;
+        tmp_npad_type = npad_type_;
+        lock.unlock();
+        TriggerOnChange(ControllerTriggerType::Type, false);
+        return;
     }
+
+    if (npad_type == npad_type_) {
+        return;
+    }
+    if (is_connected) {
+        LOG_WARNING(Service_HID, "Controller {} type changed while it's connected",
+                    NpadIdTypeToIndex(npad_id_type));
+    }
+    npad_type = npad_type_;
+
+    lock.unlock();
     TriggerOnChange(ControllerTriggerType::Type, true);
 }
 
@@ -1088,30 +1111,37 @@ LedPattern EmulatedController::GetLedPattern() const {
 }
 
 ButtonValues EmulatedController::GetButtonsValues() const {
+    std::scoped_lock lock{mutex};
     return controller.button_values;
 }
 
 SticksValues EmulatedController::GetSticksValues() const {
+    std::scoped_lock lock{mutex};
     return controller.stick_values;
 }
 
 TriggerValues EmulatedController::GetTriggersValues() const {
+    std::scoped_lock lock{mutex};
     return controller.trigger_values;
 }
 
 ControllerMotionValues EmulatedController::GetMotionValues() const {
+    std::scoped_lock lock{mutex};
     return controller.motion_values;
 }
 
 ColorValues EmulatedController::GetColorsValues() const {
+    std::scoped_lock lock{mutex};
     return controller.color_values;
 }
 
 BatteryValues EmulatedController::GetBatteryValues() const {
+    std::scoped_lock lock{mutex};
     return controller.battery_values;
 }
 
 HomeButtonState EmulatedController::GetHomeButtons() const {
+    std::scoped_lock lock{mutex};
     if (is_configuring) {
         return {};
     }
@@ -1119,6 +1149,7 @@ HomeButtonState EmulatedController::GetHomeButtons() const {
 }
 
 CaptureButtonState EmulatedController::GetCaptureButtons() const {
+    std::scoped_lock lock{mutex};
     if (is_configuring) {
         return {};
     }
@@ -1126,6 +1157,7 @@ CaptureButtonState EmulatedController::GetCaptureButtons() const {
 }
 
 NpadButtonState EmulatedController::GetNpadButtons() const {
+    std::scoped_lock lock{mutex};
     if (is_configuring) {
         return {};
     }
@@ -1133,6 +1165,7 @@ NpadButtonState EmulatedController::GetNpadButtons() const {
 }
 
 DebugPadButton EmulatedController::GetDebugPadButtons() const {
+    std::scoped_lock lock{mutex};
     if (is_configuring) {
         return {};
     }
@@ -1140,20 +1173,27 @@ DebugPadButton EmulatedController::GetDebugPadButtons() const {
 }
 
 AnalogSticks EmulatedController::GetSticks() const {
+    std::unique_lock lock{mutex};
+
     if (is_configuring) {
         return {};
     }
+
     // Some drivers like stick from buttons need constant refreshing
     for (auto& device : stick_devices) {
         if (!device) {
             continue;
         }
+        lock.unlock();
         device->SoftUpdate();
+        lock.lock();
     }
+
     return controller.analog_stick_state;
 }
 
 NpadGcTriggerState EmulatedController::GetTriggers() const {
+    std::scoped_lock lock{mutex};
     if (is_configuring) {
         return {};
     }
@@ -1161,26 +1201,35 @@ NpadGcTriggerState EmulatedController::GetTriggers() const {
 }
 
 MotionState EmulatedController::GetMotions() const {
+    std::unique_lock lock{mutex};
+
+    // Some drivers like mouse motion need constant refreshing
     if (force_update_motion) {
         for (auto& device : motion_devices) {
             if (!device) {
                 continue;
             }
+            lock.unlock();
             device->ForceUpdate();
+            lock.lock();
         }
     }
+
     return controller.motion_state;
 }
 
 ControllerColors EmulatedController::GetColors() const {
+    std::scoped_lock lock{mutex};
     return controller.colors_state;
 }
 
 BatteryLevelState EmulatedController::GetBattery() const {
+    std::scoped_lock lock{mutex};
     return controller.battery_state;
 }
 
 void EmulatedController::TriggerOnChange(ControllerTriggerType type, bool is_npad_service_update) {
+    std::scoped_lock lock{callback_mutex};
     for (const auto& poller_pair : callback_list) {
         const ControllerUpdateCallback& poller = poller_pair.second;
         if (!is_npad_service_update && poller.is_npad_service) {
@@ -1193,13 +1242,13 @@ void EmulatedController::TriggerOnChange(ControllerTriggerType type, bool is_npa
 }
 
 int EmulatedController::SetCallback(ControllerUpdateCallback update_callback) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{callback_mutex};
     callback_list.insert_or_assign(last_callback_key, std::move(update_callback));
     return last_callback_key++;
 }
 
 void EmulatedController::DeleteCallback(int key) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{callback_mutex};
     const auto& iterator = callback_list.find(key);
     if (iterator == callback_list.end()) {
         LOG_ERROR(Input, "Tried to delete non-existent callback {}", key);

src/core/hid/emulated_controller.h

@@ -400,7 +400,7 @@ private:
      */
     void TriggerOnChange(ControllerTriggerType type, bool is_service_update);
 
-    NpadIdType npad_id_type;
+    const NpadIdType npad_id_type;
     NpadStyleIndex npad_type{NpadStyleIndex::None};
     NpadStyleTag supported_style_tag{NpadStyleSet::All};
     bool is_connected{false};
@@ -434,6 +434,7 @@ private:
     StickDevices tas_stick_devices;
 
     mutable std::mutex mutex;
+    mutable std::mutex callback_mutex;
     std::unordered_map<int, ControllerUpdateCallback> callback_list;
     int last_callback_key = 0;
 

src/core/hid/emulated_devices.cpp

@@ -169,7 +169,7 @@ void EmulatedDevices::SetKeyboardButton(const Common::Input::CallbackStatus& cal
     if (index >= device_status.keyboard_values.size()) {
         return;
     }
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     bool value_changed = false;
     const auto new_status = TransformToButton(callback);
     auto& current_status = device_status.keyboard_values[index];
@@ -201,6 +201,7 @@ void EmulatedDevices::SetKeyboardButton(const Common::Input::CallbackStatus& cal
     }
 
     if (is_configuring) {
+        lock.unlock();
         TriggerOnChange(DeviceTriggerType::Keyboard);
         return;
     }
@@ -208,6 +209,7 @@ void EmulatedDevices::SetKeyboardButton(const Common::Input::CallbackStatus& cal
     // Index should be converted from NativeKeyboard to KeyboardKeyIndex
     UpdateKey(index, current_status.value);
 
+    lock.unlock();
     TriggerOnChange(DeviceTriggerType::Keyboard);
 }
 
@@ -227,7 +229,7 @@ void EmulatedDevices::SetKeyboardModifier(const Common::Input::CallbackStatus& c
     if (index >= device_status.keyboard_moddifier_values.size()) {
         return;
     }
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     bool value_changed = false;
     const auto new_status = TransformToButton(callback);
     auto& current_status = device_status.keyboard_moddifier_values[index];
@@ -259,6 +261,7 @@ void EmulatedDevices::SetKeyboardModifier(const Common::Input::CallbackStatus& c
     }
 
     if (is_configuring) {
+        lock.unlock();
         TriggerOnChange(DeviceTriggerType::KeyboardModdifier);
         return;
     }
@@ -289,6 +292,7 @@ void EmulatedDevices::SetKeyboardModifier(const Common::Input::CallbackStatus& c
         break;
     }
 
+    lock.unlock();
     TriggerOnChange(DeviceTriggerType::KeyboardModdifier);
 }
 
@@ -297,7 +301,7 @@ void EmulatedDevices::SetMouseButton(const Common::Input::CallbackStatus& callba
     if (index >= device_status.mouse_button_values.size()) {
         return;
     }
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     bool value_changed = false;
     const auto new_status = TransformToButton(callback);
     auto& current_status = device_status.mouse_button_values[index];
@@ -329,6 +333,7 @@ void EmulatedDevices::SetMouseButton(const Common::Input::CallbackStatus& callba
     }
 
     if (is_configuring) {
+        lock.unlock();
         TriggerOnChange(DeviceTriggerType::Mouse);
         return;
     }
@@ -351,6 +356,7 @@ void EmulatedDevices::SetMouseButton(const Common::Input::CallbackStatus& callba
         break;
     }
 
+    lock.unlock();
     TriggerOnChange(DeviceTriggerType::Mouse);
 }
 
@@ -359,13 +365,14 @@ void EmulatedDevices::SetMouseAnalog(const Common::Input::CallbackStatus& callba
     if (index >= device_status.mouse_analog_values.size()) {
         return;
     }
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     const auto analog_value = TransformToAnalog(callback);
 
     device_status.mouse_analog_values[index] = analog_value;
 
     if (is_configuring) {
         device_status.mouse_position_state = {};
+        lock.unlock();
         TriggerOnChange(DeviceTriggerType::Mouse);
         return;
     }
@@ -379,17 +386,19 @@ void EmulatedDevices::SetMouseAnalog(const Common::Input::CallbackStatus& callba
         break;
     }
 
+    lock.unlock();
     TriggerOnChange(DeviceTriggerType::Mouse);
 }
 
 void EmulatedDevices::SetMouseStick(const Common::Input::CallbackStatus& callback) {
-    std::lock_guard lock{mutex};
+    std::unique_lock lock{mutex};
     const auto touch_value = TransformToTouch(callback);
 
     device_status.mouse_stick_value = touch_value;
 
     if (is_configuring) {
         device_status.mouse_position_state = {};
+        lock.unlock();
         TriggerOnChange(DeviceTriggerType::Mouse);
         return;
     }
@@ -397,42 +406,52 @@ void EmulatedDevices::SetMouseStick(const Common::Input::CallbackStatus& callbac
     device_status.mouse_position_state.x = touch_value.x.value;
     device_status.mouse_position_state.y = touch_value.y.value;
 
+    lock.unlock();
     TriggerOnChange(DeviceTriggerType::Mouse);
 }
 
 KeyboardValues EmulatedDevices::GetKeyboardValues() const {
+    std::scoped_lock lock{mutex};
     return device_status.keyboard_values;
 }
 
 KeyboardModifierValues EmulatedDevices::GetKeyboardModdifierValues() const {
+    std::scoped_lock lock{mutex};
     return device_status.keyboard_moddifier_values;
 }
 
 MouseButtonValues EmulatedDevices::GetMouseButtonsValues() const {
+    std::scoped_lock lock{mutex};
     return device_status.mouse_button_values;
 }
 
 KeyboardKey EmulatedDevices::GetKeyboard() const {
+    std::scoped_lock lock{mutex};
     return device_status.keyboard_state;
 }
 
 KeyboardModifier EmulatedDevices::GetKeyboardModifier() const {
+    std::scoped_lock lock{mutex};
     return device_status.keyboard_moddifier_state;
 }
 
 MouseButton EmulatedDevices::GetMouseButtons() const {
+    std::scoped_lock lock{mutex};
     return device_status.mouse_button_state;
 }
 
 MousePosition EmulatedDevices::GetMousePosition() const {
+    std::scoped_lock lock{mutex};
     return device_status.mouse_position_state;
 }
 
 AnalogStickState EmulatedDevices::GetMouseWheel() const {
+    std::scoped_lock lock{mutex};
     return device_status.mouse_wheel_state;
 }
 
 void EmulatedDevices::TriggerOnChange(DeviceTriggerType type) {
+    std::scoped_lock lock{callback_mutex};
     for (const auto& poller_pair : callback_list) {
         const InterfaceUpdateCallback& poller = poller_pair.second;
         if (poller.on_change) {
@@ -442,13 +461,13 @@ void EmulatedDevices::TriggerOnChange(DeviceTriggerType type) {
 }
 
 int EmulatedDevices::SetCallback(InterfaceUpdateCallback update_callback) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{callback_mutex};
     callback_list.insert_or_assign(last_callback_key, std::move(update_callback));
     return last_callback_key++;
 }
 
 void EmulatedDevices::DeleteCallback(int key) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{callback_mutex};
     const auto& iterator = callback_list.find(key);
     if (iterator == callback_list.end()) {
         LOG_ERROR(Input, "Tried to delete non-existent callback {}", key);

src/core/hid/emulated_devices.h

@@ -200,6 +200,7 @@ private:
     MouseStickDevice mouse_stick_device;
 
     mutable std::mutex mutex;
+    mutable std::mutex callback_mutex;
     std::unordered_map<int, InterfaceUpdateCallback> callback_list;
     int last_callback_key = 0;
 

src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp

@@ -148,9 +148,9 @@ u64 GenerateUniformRange(u64 min, u64 max, F f) {
 } // Anonymous namespace
 
 u64 KSystemControl::GenerateRandomU64() {
-    static std::random_device device;
-    static std::mt19937 gen(device());
-    static std::uniform_int_distribution<u64> distribution(1, std::numeric_limits<u64>::max());
+    std::random_device device;
+    std::mt19937 gen(device());
+    std::uniform_int_distribution<u64> distribution(1, std::numeric_limits<u64>::max());
     return distribution(gen);
 }
 

src/core/hle/kernel/hle_ipc.cpp

@@ -24,8 +24,15 @@
 
 namespace Kernel {
 
-SessionRequestHandler::SessionRequestHandler(KernelCore& kernel_, const char* service_name_)
-    : kernel{kernel_}, service_thread{kernel.CreateServiceThread(service_name_)} {}
+SessionRequestHandler::SessionRequestHandler(KernelCore& kernel_, const char* service_name_,
+                                             ServiceThreadType thread_type)
+    : kernel{kernel_} {
+    if (thread_type == ServiceThreadType::CreateNew) {
+        service_thread = kernel.CreateServiceThread(service_name_);
+    } else {
+        service_thread = kernel.GetDefaultServiceThread();
+    }
+}
 
 SessionRequestHandler::~SessionRequestHandler() {
     kernel.ReleaseServiceThread(service_thread);
@@ -44,7 +51,7 @@ bool SessionRequestManager::HasSessionRequestHandler(const HLERequestContext& co
             LOG_CRITICAL(IPC, "object_id {} is too big!", object_id);
             return false;
         }
-        return DomainHandler(object_id - 1).lock() != nullptr;
+        return !DomainHandler(object_id - 1).expired();
     } else {
         return session_handler != nullptr;
     }
@@ -52,6 +59,9 @@ bool SessionRequestManager::HasSessionRequestHandler(const HLERequestContext& co
 
 void SessionRequestHandler::ClientConnected(KServerSession* session) {
     session->ClientConnected(shared_from_this());
+
+    // Ensure our server session is tracked globally.
+    kernel.RegisterServerObject(session);
 }
 
 void SessionRequestHandler::ClientDisconnected(KServerSession* session) {

src/core/hle/kernel/hle_ipc.h

@@ -33,6 +33,11 @@ namespace Service {
 class ServiceFrameworkBase;
 }
 
+enum class ServiceThreadType {
+    Default,
+    CreateNew,
+};
+
 namespace Kernel {
 
 class Domain;
@@ -57,7 +62,8 @@ enum class ThreadWakeupReason;
  */
 class SessionRequestHandler : public std::enable_shared_from_this<SessionRequestHandler> {
 public:
-    SessionRequestHandler(KernelCore& kernel, const char* service_name_);
+    SessionRequestHandler(KernelCore& kernel_, const char* service_name_,
+                          ServiceThreadType thread_type);
     virtual ~SessionRequestHandler();
 
     /**

src/core/hle/kernel/k_auto_object.h

@@ -89,9 +89,7 @@ public:
     explicit KAutoObject(KernelCore& kernel_) : kernel(kernel_) {
         RegisterWithKernel();
     }
-    virtual ~KAutoObject() {
-        UnregisterWithKernel();
-    }
+    virtual ~KAutoObject() = default;
 
     static KAutoObject* Create(KAutoObject* ptr);
 
@@ -163,11 +161,12 @@ public:
         do {
             ASSERT(cur_ref_count > 0);
         } while (!m_ref_count.compare_exchange_weak(cur_ref_count, cur_ref_count - 1,
-                                                    std::memory_order_relaxed));
+                                                    std::memory_order_acq_rel));
 
         // If ref count hits zero, destroy the object.
         if (cur_ref_count - 1 == 0) {
             this->Destroy();
+            this->UnregisterWithKernel();
         }
     }
 

src/core/hle/kernel/k_code_memory.cpp

@@ -35,9 +35,14 @@ ResultCode KCodeMemory::Initialize(Core::DeviceMemory& device_memory, VAddr addr
     R_TRY(page_table.LockForCodeMemory(addr, size))
 
     // Clear the memory.
-    for (const auto& block : m_page_group.Nodes()) {
-        std::memset(device_memory.GetPointer(block.GetAddress()), 0xFF, block.GetSize());
-    }
+    //
+    // FIXME: this ends up clobbering address ranges outside the scope of the mapping within
+    // guest memory, and is not specifically required if the guest program is correctly
+    // written, so disable until this is further investigated.
+    //
+    // for (const auto& block : m_page_group.Nodes()) {
+    //     std::memset(device_memory.GetPointer(block.GetAddress()), 0xFF, block.GetSize());
+    // }
 
     // Set remaining tracking members.
     m_address = addr;

src/core/hle/kernel/k_page_table.cpp

@@ -346,7 +346,8 @@ ResultCode KPageTable::MapCodeMemory(VAddr dst_address, VAddr src_address, std::
     return ResultSuccess;
 }
 
-ResultCode KPageTable::UnmapCodeMemory(VAddr dst_address, VAddr src_address, std::size_t size) {
+ResultCode KPageTable::UnmapCodeMemory(VAddr dst_address, VAddr src_address, std::size_t size,
+                                       ICacheInvalidationStrategy icache_invalidation_strategy) {
     // Validate the mapping request.
     R_UNLESS(this->CanContain(dst_address, size, KMemoryState::AliasCode),
              ResultInvalidMemoryRegion);
@@ -396,7 +397,11 @@ ResultCode KPageTable::UnmapCodeMemory(VAddr dst_address, VAddr src_address, std
     bool reprotected_pages = false;
     SCOPE_EXIT({
         if (reprotected_pages && any_code_pages) {
-            system.InvalidateCpuInstructionCacheRange(dst_address, size);
+            if (icache_invalidation_strategy == ICacheInvalidationStrategy::InvalidateRange) {
+                system.InvalidateCpuInstructionCacheRange(dst_address, size);
+            } else {
+                system.InvalidateCpuInstructionCaches();
+            }
         }
     });
 
@@ -563,6 +568,8 @@ ResultCode KPageTable::UnmapProcessMemory(VAddr dst_addr, std::size_t size,
     block_manager->Update(dst_addr, num_pages, KMemoryState::Free, KMemoryPermission::None,
                           KMemoryAttribute::None);
 
+    system.InvalidateCpuInstructionCaches();
+
     return ResultSuccess;
 }
 

src/core/hle/kernel/k_page_table.h

@@ -26,6 +26,8 @@ class KMemoryBlockManager;
 
 class KPageTable final {
 public:
+    enum class ICacheInvalidationStrategy : u32 { InvalidateRange, InvalidateAll };
+
     YUZU_NON_COPYABLE(KPageTable);
     YUZU_NON_MOVEABLE(KPageTable);
 
@@ -38,7 +40,8 @@ public:
     ResultCode MapProcessCode(VAddr addr, std::size_t pages_count, KMemoryState state,
                               KMemoryPermission perm);
     ResultCode MapCodeMemory(VAddr dst_address, VAddr src_address, std::size_t size);
-    ResultCode UnmapCodeMemory(VAddr dst_address, VAddr src_address, std::size_t size);
+    ResultCode UnmapCodeMemory(VAddr dst_address, VAddr src_address, std::size_t size,
+                               ICacheInvalidationStrategy icache_invalidation_strategy);
     ResultCode UnmapProcessMemory(VAddr dst_addr, std::size_t size, KPageTable& src_page_table,
                                   VAddr src_addr);
     ResultCode MapPhysicalMemory(VAddr addr, std::size_t size);

src/core/hle/kernel/k_process.h

@@ -422,7 +422,7 @@ private:
     bool is_64bit_process = true;
 
     /// Total running time for the process in ticks.
-    u64 total_process_running_time_ticks = 0;
+    std::atomic<u64> total_process_running_time_ticks = 0;
 
     /// Per-process handle table for storing created object handles in.
     KHandleTable handle_table;

src/core/hle/kernel/k_scheduler_lock.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <atomic>
 #include "common/assert.h"
 #include "core/hle/kernel/k_spin_lock.h"
 #include "core/hle/kernel/k_thread.h"
@@ -75,7 +76,7 @@ private:
     KernelCore& kernel;
     KAlignedSpinLock spin_lock{};
     s32 lock_count{};
-    KThread* owner_thread{};
+    std::atomic<KThread*> owner_thread{};
 };
 
 } // namespace Kernel

src/core/hle/kernel/k_server_port.cpp

@@ -62,6 +62,12 @@ void KServerPort::Destroy() {
 
     // Close our reference to our parent.
     parent->Close();
+
+    // Release host emulation members.
+    session_handler.reset();
+
+    // Ensure that the global list tracking server objects does not hold on to a reference.
+    kernel.UnregisterServerObject(this);
 }
 
 bool KServerPort::IsSignaled() const {

src/core/hle/kernel/k_server_session.cpp

@@ -49,6 +49,9 @@ void KServerSession::Destroy() {
 
     // Release host emulation members.
     manager.reset();
+
+    // Ensure that the global list tracking server objects does not hold on to a reference.
+    kernel.UnregisterServerObject(this);
 }
 
 void KServerSession::OnClientClosed() {

src/core/hle/kernel/k_thread.cpp

@@ -723,7 +723,7 @@ void KThread::UpdateState() {
     ASSERT(kernel.GlobalSchedulerContext().IsLocked());
 
     // Set our suspend flags in state.
-    const auto old_state = thread_state;
+    const ThreadState old_state = thread_state;
     const auto new_state =
         static_cast<ThreadState>(this->GetSuspendFlags()) | (old_state & ThreadState::Mask);
     thread_state = new_state;
@@ -738,7 +738,7 @@ void KThread::Continue() {
     ASSERT(kernel.GlobalSchedulerContext().IsLocked());
 
     // Clear our suspend flags in state.
-    const auto old_state = thread_state;
+    const ThreadState old_state = thread_state;
     thread_state = old_state & ThreadState::Mask;
 
     // Note the state change in scheduler.

src/core/hle/kernel/k_thread.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <atomic>
 #include <span>
 #include <string>
 #include <utility>
@@ -751,7 +752,7 @@ private:
     KAffinityMask original_physical_affinity_mask{};
     s32 original_physical_ideal_core_id{};
     s32 num_core_migration_disables{};
-    ThreadState thread_state{};
+    std::atomic<ThreadState> thread_state{};
     std::atomic<bool> termination_requested{};
     bool wait_cancelled{};
     bool cancellable{};

src/core/hle/kernel/kernel.cpp

@@ -61,6 +61,7 @@ struct KernelCore::Impl {
         global_scheduler_context = std::make_unique<Kernel::GlobalSchedulerContext>(kernel);
         global_handle_table = std::make_unique<Kernel::KHandleTable>(kernel);
         global_handle_table->Initialize(KHandleTable::MaxTableSize);
+        default_service_thread = CreateServiceThread(kernel, "DefaultServiceThread");
 
         is_phantom_mode_for_singlecore = false;
 
@@ -84,7 +85,7 @@ struct KernelCore::Impl {
 
     void InitializeCores() {
         for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            cores[core_id].Initialize(current_process->Is64BitProcess());
+            cores[core_id].Initialize((*current_process).Is64BitProcess());
             system.Memory().SetCurrentPageTable(*current_process, core_id);
         }
     }
@@ -95,15 +96,15 @@ struct KernelCore::Impl {
 
         process_list.clear();
 
-        // Close all open server ports.
-        std::unordered_set<KServerPort*> server_ports_;
+        // Close all open server sessions and ports.
+        std::unordered_set<KAutoObject*> server_objects_;
         {
-            std::lock_guard lk(server_ports_lock);
-            server_ports_ = server_ports;
-            server_ports.clear();
+            std::scoped_lock lk(server_objects_lock);
+            server_objects_ = server_objects;
+            server_objects.clear();
         }
-        for (auto* server_port : server_ports_) {
-            server_port->Close();
+        for (auto* server_object : server_objects_) {
+            server_object->Close();
         }
 
         // Ensures all service threads gracefully shutdown.
@@ -156,7 +157,7 @@ struct KernelCore::Impl {
 
         // Close kernel objects that were not freed on shutdown
         {
-            std::lock_guard lk(registered_in_use_objects_lock);
+            std::scoped_lock lk{registered_in_use_objects_lock};
             if (registered_in_use_objects.size()) {
                 for (auto& object : registered_in_use_objects) {
                     object->Close();
@@ -167,17 +168,17 @@ struct KernelCore::Impl {
 
         // Shutdown all processes.
         if (current_process) {
-            current_process->Finalize();
+            (*current_process).Finalize();
             // current_process->Close();
             // TODO: The current process should be destroyed based on accurate ref counting after
             // calling Close(). Adding a manual Destroy() call instead to avoid a memory leak.
-            current_process->Destroy();
+            (*current_process).Destroy();
             current_process = nullptr;
         }
 
         // Track kernel objects that were not freed on shutdown
         {
-            std::lock_guard lk(registered_objects_lock);
+            std::scoped_lock lk{registered_objects_lock};
             if (registered_objects.size()) {
                 LOG_DEBUG(Kernel, "{} kernel objects were dangling on shutdown!",
                           registered_objects.size());
@@ -658,13 +659,20 @@ struct KernelCore::Impl {
         }
 
         KClientPort* port = &search->second(system.ServiceManager(), system);
-        {
-            std::lock_guard lk(server_ports_lock);
-            server_ports.insert(&port->GetParent()->GetServerPort());
-        }
+        RegisterServerObject(&port->GetParent()->GetServerPort());
         return port;
     }
 
+    void RegisterServerObject(KAutoObject* server_object) {
+        std::scoped_lock lk(server_objects_lock);
+        server_objects.insert(server_object);
+    }
+
+    void UnregisterServerObject(KAutoObject* server_object) {
+        std::scoped_lock lk(server_objects_lock);
+        server_objects.erase(server_object);
+    }
+
     std::weak_ptr<Kernel::ServiceThread> CreateServiceThread(KernelCore& kernel,
                                                              const std::string& name) {
         auto service_thread = std::make_shared<Kernel::ServiceThread>(kernel, 1, name);
@@ -677,6 +685,12 @@ struct KernelCore::Impl {
 
     void ReleaseServiceThread(std::weak_ptr<Kernel::ServiceThread> service_thread) {
         if (auto strong_ptr = service_thread.lock()) {
+            if (strong_ptr == default_service_thread.lock()) {
+                // Nothing to do here, the service is using default_service_thread, which will be
+                // released on shutdown.
+                return;
+            }
+
             service_threads_manager.QueueWork(
                 [this, strong_ptr{std::move(strong_ptr)}]() { service_threads.erase(strong_ptr); });
         }
@@ -686,7 +700,7 @@ struct KernelCore::Impl {
         service_threads_manager.QueueWork([this]() { service_threads.clear(); });
     }
 
-    std::mutex server_ports_lock;
+    std::mutex server_objects_lock;
     std::mutex registered_objects_lock;
     std::mutex registered_in_use_objects_lock;
 
@@ -697,7 +711,7 @@ struct KernelCore::Impl {
 
     // Lists all processes that exist in the current session.
     std::vector<KProcess*> process_list;
-    KProcess* current_process{};
+    std::atomic<KProcess*> current_process{};
     std::unique_ptr<Kernel::GlobalSchedulerContext> global_scheduler_context;
     Kernel::TimeManager time_manager;
 
@@ -716,7 +730,7 @@ struct KernelCore::Impl {
     /// the ConnectToPort SVC.
     std::unordered_map<std::string, ServiceInterfaceFactory> service_interface_factory;
     NamedPortTable named_ports;
-    std::unordered_set<KServerPort*> server_ports;
+    std::unordered_set<KAutoObject*> server_objects;
     std::unordered_set<KAutoObject*> registered_objects;
     std::unordered_set<KAutoObject*> registered_in_use_objects;
 
@@ -739,7 +753,8 @@ struct KernelCore::Impl {
     std::unique_ptr<KMemoryLayout> memory_layout;
 
     // Threads used for services
-    std::unordered_set<std::shared_ptr<Kernel::ServiceThread>> service_threads;
+    std::unordered_set<std::shared_ptr<ServiceThread>> service_threads;
+    std::weak_ptr<ServiceThread> default_service_thread;
     Common::ThreadWorker service_threads_manager;
 
     std::array<KThread*, Core::Hardware::NUM_CPU_CORES> suspend_threads;
@@ -920,23 +935,31 @@ KClientPort* KernelCore::CreateNamedServicePort(std::string name) {
     return impl->CreateNamedServicePort(std::move(name));
 }
 
+void KernelCore::RegisterServerObject(KAutoObject* server_object) {
+    impl->RegisterServerObject(server_object);
+}
+
+void KernelCore::UnregisterServerObject(KAutoObject* server_object) {
+    impl->UnregisterServerObject(server_object);
+}
+
 void KernelCore::RegisterKernelObject(KAutoObject* object) {
-    std::lock_guard lk(impl->registered_objects_lock);
+    std::scoped_lock lk{impl->registered_objects_lock};
     impl->registered_objects.insert(object);
 }
 
 void KernelCore::UnregisterKernelObject(KAutoObject* object) {
-    std::lock_guard lk(impl->registered_objects_lock);
+    std::scoped_lock lk{impl->registered_objects_lock};
     impl->registered_objects.erase(object);
 }
 
 void KernelCore::RegisterInUseObject(KAutoObject* object) {
-    std::lock_guard lk(impl->registered_in_use_objects_lock);
+    std::scoped_lock lk{impl->registered_in_use_objects_lock};
     impl->registered_in_use_objects.insert(object);
 }
 
 void KernelCore::UnregisterInUseObject(KAutoObject* object) {
-    std::lock_guard lk(impl->registered_in_use_objects_lock);
+    std::scoped_lock lk{impl->registered_in_use_objects_lock};
     impl->registered_in_use_objects.erase(object);
 }
 
@@ -1065,6 +1088,10 @@ std::weak_ptr<Kernel::ServiceThread> KernelCore::CreateServiceThread(const std::
     return impl->CreateServiceThread(*this, name);
 }
 
+std::weak_ptr<Kernel::ServiceThread> KernelCore::GetDefaultServiceThread() const {
+    return impl->default_service_thread;
+}
+
 void KernelCore::ReleaseServiceThread(std::weak_ptr<Kernel::ServiceThread> service_thread) {
     impl->ReleaseServiceThread(service_thread);
 }

src/core/hle/kernel/kernel.h

@@ -195,6 +195,14 @@ public:
     /// Opens a port to a service previously registered with RegisterNamedService.
     KClientPort* CreateNamedServicePort(std::string name);
 
+    /// Registers a server session or port with the gobal emulation state, to be freed on shutdown.
+    /// This is necessary because we do not emulate processes for HLE sessions and ports.
+    void RegisterServerObject(KAutoObject* server_object);
+
+    /// Unregisters a server session or port previously registered with RegisterServerSession when
+    /// it was destroyed during the current emulation session.
+    void UnregisterServerObject(KAutoObject* server_object);
+
     /// Registers all kernel objects with the global emulation state, this is purely for tracking
     /// leaks after emulation has been shutdown.
     void RegisterKernelObject(KAutoObject* object);
@@ -271,15 +279,25 @@ public:
     void ExitSVCProfile();
 
     /**
-     * Creates an HLE service thread, which are used to execute service routines asynchronously.
-     * While these are allocated per ServerSession, these need to be owned and managed outside
-     * of ServerSession to avoid a circular dependency.
+     * Creates a host thread to execute HLE service requests, which are used to execute service
+     * routines asynchronously. While these are allocated per ServerSession, these need to be owned
+     * and managed outside of ServerSession to avoid a circular dependency. In general, most
+     * services can just use the default service thread, and not need their own host service thread.
+     * See GetDefaultServiceThread.
      * @param name String name for the ServerSession creating this thread, used for debug
      * purposes.
      * @returns The a weak pointer newly created service thread.
      */
     std::weak_ptr<Kernel::ServiceThread> CreateServiceThread(const std::string& name);
 
+    /**
+     * Gets the default host service thread, which executes HLE service requests. Unless service
+     * requests need to block on the host, the default service thread should be used in favor of
+     * creating a new service thread.
+     * @returns The a weak pointer for the default service thread.
+     */
+    std::weak_ptr<Kernel::ServiceThread> GetDefaultServiceThread() const;
+
     /**
      * Releases a HLE service thread, instructing KernelCore to free it. This should be called when
      * the ServerSession associated with the thread is destroyed.

src/core/hle/kernel/physical_core.cpp

@@ -58,6 +58,7 @@ bool PhysicalCore::IsInterrupted() const {
 void PhysicalCore::Interrupt() {
     guard->lock();
     interrupts[core_index].SetInterrupt(true);
+    arm_interface->SignalInterrupt();
     guard->unlock();
 }
 

src/core/hle/kernel/svc.cpp

@@ -1713,7 +1713,8 @@ static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_ha
         return ResultInvalidMemoryRegion;
     }
 
-    return page_table.UnmapCodeMemory(dst_address, src_address, size);
+    return page_table.UnmapCodeMemory(dst_address, src_address, size,
+                                      KPageTable::ICacheInvalidationStrategy::InvalidateAll);
 }
 
 /// Exits the current process

src/core/hle/kernel/time_manager.cpp

@@ -24,7 +24,7 @@ TimeManager::TimeManager(Core::System& system_) : system{system_} {
 }
 
 void TimeManager::ScheduleTimeEvent(KThread* thread, s64 nanoseconds) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     if (nanoseconds > 0) {
         ASSERT(thread);
         ASSERT(thread->GetState() != ThreadState::Runnable);
@@ -35,7 +35,7 @@ void TimeManager::ScheduleTimeEvent(KThread* thread, s64 nanoseconds) {
 }
 
 void TimeManager::UnscheduleTimeEvent(KThread* thread) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     system.CoreTiming().UnscheduleEvent(time_manager_event_type,
                                         reinterpret_cast<uintptr_t>(thread));
 }

src/core/hle/service/am/am.cpp

@@ -1337,7 +1337,7 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
         {200, nullptr, "GetLastApplicationExitReason"},
         {500, nullptr, "StartContinuousRecordingFlushForDebug"},
         {1000, nullptr, "CreateMovieMaker"},
-        {1001, nullptr, "PrepareForJit"},
+        {1001, &IApplicationFunctions::PrepareForJit, "PrepareForJit"},
     };
     // clang-format on
 
@@ -1787,6 +1787,13 @@ void IApplicationFunctions::GetHealthWarningDisappearedSystemEvent(Kernel::HLERe
     rb.PushCopyObjects(health_warning_disappeared_system_event->GetReadableEvent());
 }
 
+void IApplicationFunctions::PrepareForJit(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
 void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger& nvflinger,
                        Core::System& system) {
     auto message_queue = std::make_shared<AppletMessageQueue>(system);

src/core/hle/service/am/am.h

@@ -336,6 +336,7 @@ private:
     void TryPopFromFriendInvitationStorageChannel(Kernel::HLERequestContext& ctx);
     void GetNotificationStorageChannelEvent(Kernel::HLERequestContext& ctx);
     void GetHealthWarningDisappearedSystemEvent(Kernel::HLERequestContext& ctx);
+    void PrepareForJit(Kernel::HLERequestContext& ctx);
 
     KernelHelpers::ServiceContext service_context;
 

src/core/hle/service/am/applets/applet_web_browser.cpp

@@ -446,6 +446,14 @@ void WebBrowser::ExecuteLogin() {
 }
 
 void WebBrowser::ExecuteOffline() {
+    // TODO (Morph): This is a hack for WebSession foreground web applets such as those used by
+    //               Super Mario 3D All-Stars.
+    // TODO (Morph): Implement WebSession.
+    if (applet_mode == LibraryAppletMode::AllForegroundInitiallyHidden) {
+        LOG_WARNING(Service_AM, "WebSession is not implemented");
+        return;
+    }
+
     const auto main_url = GetMainURL(Common::FS::PathToUTF8String(offline_document));
 
     if (!Common::FS::Exists(main_url)) {

src/core/hle/service/audio/audout_u.cpp

@@ -41,9 +41,10 @@ public:
     explicit IAudioOut(Core::System& system_, AudoutParams audio_params_,
                        AudioCore::AudioOut& audio_core_, std::string&& device_name_,
                        std::string&& unique_name)
-        : ServiceFramework{system_, "IAudioOut"}, audio_core{audio_core_},
-          device_name{std::move(device_name_)}, audio_params{audio_params_},
-          main_memory{system.Memory()}, service_context{system_, "IAudioOut"} {
+        : ServiceFramework{system_, "IAudioOut", ServiceThreadType::CreateNew},
+          audio_core{audio_core_}, device_name{std::move(device_name_)},
+          audio_params{audio_params_}, main_memory{system.Memory()}, service_context{system_,
+                                                                                     "IAudioOut"} {
         // clang-format off
         static const FunctionInfo functions[] = {
             {0, &IAudioOut::GetAudioOutState, "GetAudioOutState"},

src/core/hle/service/audio/audren_u.cpp

@@ -24,7 +24,8 @@ public:
     explicit IAudioRenderer(Core::System& system_,
                             const AudioCommon::AudioRendererParameter& audren_params,
                             const std::size_t instance_number)
-        : ServiceFramework{system_, "IAudioRenderer"}, service_context{system_, "IAudioRenderer"} {
+        : ServiceFramework{system_, "IAudioRenderer", ServiceThreadType::CreateNew},
+          service_context{system_, "IAudioRenderer"} {
         // clang-format off
         static const FunctionInfo functions[] = {
             {0, &IAudioRenderer::GetSampleRate, "GetSampleRate"},

src/core/hle/service/filesystem/filesystem.cpp

@@ -174,7 +174,7 @@ ResultCode VfsDirectoryServiceWrapper::RenameFile(const std::string& src_path_,
     ASSERT_MSG(dest != nullptr, "Newly created file with success cannot be found.");
 
     ASSERT_MSG(dest->WriteBytes(src->ReadAllBytes()) == src->GetSize(),
-               "Could not write all of the bytes but everything else has succeded.");
+               "Could not write all of the bytes but everything else has succeeded.");
 
     if (!src->GetContainingDirectory()->DeleteFile(Common::FS::GetFilename(src_path))) {
         // TODO(DarkLordZach): Find a better error code for this

src/core/hle/service/filesystem/fsp_srv.cpp

@@ -58,7 +58,8 @@ enum class FileSystemType : u8 {
 class IStorage final : public ServiceFramework<IStorage> {
 public:
     explicit IStorage(Core::System& system_, FileSys::VirtualFile backend_)
-        : ServiceFramework{system_, "IStorage"}, backend(std::move(backend_)) {
+        : ServiceFramework{system_, "IStorage", ServiceThreadType::CreateNew},
+          backend(std::move(backend_)) {
         static const FunctionInfo functions[] = {
             {0, &IStorage::Read, "Read"},
             {1, nullptr, "Write"},
@@ -116,7 +117,8 @@ private:
 class IFile final : public ServiceFramework<IFile> {
 public:
     explicit IFile(Core::System& system_, FileSys::VirtualFile backend_)
-        : ServiceFramework{system_, "IFile"}, backend(std::move(backend_)) {
+        : ServiceFramework{system_, "IFile", ServiceThreadType::CreateNew},
+          backend(std::move(backend_)) {
         static const FunctionInfo functions[] = {
             {0, &IFile::Read, "Read"},
             {1, &IFile::Write, "Write"},
@@ -252,7 +254,8 @@ static void BuildEntryIndex(std::vector<FileSys::Entry>& entries, const std::vec
 class IDirectory final : public ServiceFramework<IDirectory> {
 public:
     explicit IDirectory(Core::System& system_, FileSys::VirtualDir backend_)
-        : ServiceFramework{system_, "IDirectory"}, backend(std::move(backend_)) {
+        : ServiceFramework{system_, "IDirectory", ServiceThreadType::CreateNew},
+          backend(std::move(backend_)) {
         static const FunctionInfo functions[] = {
             {0, &IDirectory::Read, "Read"},
             {1, &IDirectory::GetEntryCount, "GetEntryCount"},
@@ -308,8 +311,8 @@ private:
 class IFileSystem final : public ServiceFramework<IFileSystem> {
 public:
     explicit IFileSystem(Core::System& system_, FileSys::VirtualDir backend_, SizeGetter size_)
-        : ServiceFramework{system_, "IFileSystem"}, backend{std::move(backend_)}, size{std::move(
-                                                                                      size_)} {
+        : ServiceFramework{system_, "IFileSystem", ServiceThreadType::CreateNew},
+          backend{std::move(backend_)}, size{std::move(size_)} {
         static const FunctionInfo functions[] = {
             {0, &IFileSystem::CreateFile, "CreateFile"},
             {1, &IFileSystem::DeleteFile, "DeleteFile"},

src/core/hle/service/hid/controllers/npad.cpp

@@ -262,11 +262,6 @@ void Controller_NPad::OnInit() {
             service_context.CreateEvent(fmt::format("npad:NpadStyleSetChanged_{}", i));
     }
 
-    if (hid_core.GetSupportedStyleTag().raw == Core::HID::NpadStyleSet::None) {
-        // We want to support all controllers
-        hid_core.SetSupportedStyleTag({Core::HID::NpadStyleSet::All});
-    }
-
     supported_npad_id_types.resize(npad_id_list.size());
     std::memcpy(supported_npad_id_types.data(), npad_id_list.data(),
                 npad_id_list.size() * sizeof(Core::HID::NpadIdType));
@@ -288,14 +283,6 @@ void Controller_NPad::OnInit() {
             WriteEmptyEntry(npad);
         }
     }
-
-    // Connect controllers
-    for (auto& controller : controller_data) {
-        const auto& device = controller.device;
-        if (device->IsConnected()) {
-            AddNewControllerAt(device->GetNpadStyleIndex(), device->GetNpadIdType());
-        }
-    }
 }
 
 void Controller_NPad::WriteEmptyEntry(NpadInternalState& npad) {
@@ -320,6 +307,7 @@ void Controller_NPad::WriteEmptyEntry(NpadInternalState& npad) {
 }
 
 void Controller_NPad::OnRelease() {
+    is_controller_initialized = false;
     for (std::size_t i = 0; i < controller_data.size(); ++i) {
         auto& controller = controller_data[i];
         service_context.CloseEvent(controller.styleset_changed_event);
@@ -330,7 +318,7 @@ void Controller_NPad::OnRelease() {
 }
 
 void Controller_NPad::RequestPadStateUpdate(Core::HID::NpadIdType npad_id) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     auto& controller = GetControllerFromNpadIdType(npad_id);
     const auto controller_type = controller.device->GetNpadStyleIndex();
     if (!controller.device->IsConnected()) {
@@ -651,9 +639,27 @@ void Controller_NPad::OnMotionUpdate(const Core::Timing::CoreTiming& core_timing
 
 void Controller_NPad::SetSupportedStyleSet(Core::HID::NpadStyleTag style_set) {
     hid_core.SetSupportedStyleTag(style_set);
+
+    if (is_controller_initialized) {
+        return;
+    }
+
+    // Once SetSupportedStyleSet is called controllers are fully initialized
+    is_controller_initialized = true;
+
+    // Connect all active controllers
+    for (auto& controller : controller_data) {
+        const auto& device = controller.device;
+        if (device->IsConnected()) {
+            AddNewControllerAt(device->GetNpadStyleIndex(), device->GetNpadIdType());
+        }
+    }
 }
 
 Core::HID::NpadStyleTag Controller_NPad::GetSupportedStyleSet() const {
+    if (!is_controller_initialized) {
+        return {Core::HID::NpadStyleSet::None};
+    }
     return hid_core.GetSupportedStyleTag();
 }
 

src/core/hle/service/hid/controllers/npad.h

@@ -191,16 +191,16 @@ private:
 
     // This is nn::hid::detail::NpadFullKeyColorState
     struct NpadFullKeyColorState {
-        ColorAttribute attribute;
-        Core::HID::NpadControllerColor fullkey;
+        ColorAttribute attribute{ColorAttribute::NoController};
+        Core::HID::NpadControllerColor fullkey{};
     };
     static_assert(sizeof(NpadFullKeyColorState) == 0xC, "NpadFullKeyColorState is an invalid size");
 
     // This is nn::hid::detail::NpadJoyColorState
     struct NpadJoyColorState {
-        ColorAttribute attribute;
-        Core::HID::NpadControllerColor left;
-        Core::HID::NpadControllerColor right;
+        ColorAttribute attribute{ColorAttribute::NoController};
+        Core::HID::NpadControllerColor left{};
+        Core::HID::NpadControllerColor right{};
     };
     static_assert(sizeof(NpadJoyColorState) == 0x14, "NpadJoyColorState is an invalid size");
 
@@ -226,11 +226,11 @@ private:
     // This is nn::hid::NpadPalmaState
     // This is nn::hid::NpadSystemExtState
     struct NPadGenericState {
-        s64_le sampling_number;
-        Core::HID::NpadButtonState npad_buttons;
-        Core::HID::AnalogStickState l_stick;
-        Core::HID::AnalogStickState r_stick;
-        NpadAttribute connection_status;
+        s64_le sampling_number{};
+        Core::HID::NpadButtonState npad_buttons{};
+        Core::HID::AnalogStickState l_stick{};
+        Core::HID::AnalogStickState r_stick{};
+        NpadAttribute connection_status{};
         INSERT_PADDING_BYTES(4); // Reserved
     };
     static_assert(sizeof(NPadGenericState) == 0x28, "NPadGenericState is an invalid size");
@@ -253,7 +253,7 @@ private:
         Common::Vec3f gyro{};
         Common::Vec3f rotation{};
         std::array<Common::Vec3f, 3> orientation{};
-        SixAxisSensorAttribute attribute;
+        SixAxisSensorAttribute attribute{};
         INSERT_PADDING_BYTES(4); // Reserved
     };
     static_assert(sizeof(SixAxisSensorState) == 0x60, "SixAxisSensorState is an invalid size");
@@ -325,11 +325,11 @@ private:
 
     // This is nn::hid::detail::NfcXcdDeviceHandleStateImpl
     struct NfcXcdDeviceHandleStateImpl {
-        u64 handle;
-        bool is_available;
-        bool is_activated;
+        u64 handle{};
+        bool is_available{};
+        bool is_activated{};
         INSERT_PADDING_BYTES(0x6); // Reserved
-        u64 sampling_number;
+        u64 sampling_number{};
     };
     static_assert(sizeof(NfcXcdDeviceHandleStateImpl) == 0x18,
                   "NfcXcdDeviceHandleStateImpl is an invalid size");
@@ -366,8 +366,8 @@ private:
     };
 
     struct AppletFooterUi {
-        AppletFooterUiAttributes attributes;
-        AppletFooterUiType type;
+        AppletFooterUiAttributes attributes{};
+        AppletFooterUiType type{AppletFooterUiType::None};
         INSERT_PADDING_BYTES(0x5B); // Reserved
     };
     static_assert(sizeof(AppletFooterUi) == 0x60, "AppletFooterUi is an invalid size");
@@ -404,41 +404,41 @@ private:
 
     // This is nn::hid::detail::NpadInternalState
     struct NpadInternalState {
-        Core::HID::NpadStyleTag style_tag;
-        NpadJoyAssignmentMode assignment_mode;
-        NpadFullKeyColorState fullkey_color;
-        NpadJoyColorState joycon_color;
-        Lifo<NPadGenericState, hid_entry_count> fullkey_lifo;
-        Lifo<NPadGenericState, hid_entry_count> handheld_lifo;
-        Lifo<NPadGenericState, hid_entry_count> joy_dual_lifo;
-        Lifo<NPadGenericState, hid_entry_count> joy_left_lifo;
-        Lifo<NPadGenericState, hid_entry_count> joy_right_lifo;
-        Lifo<NPadGenericState, hid_entry_count> palma_lifo;
-        Lifo<NPadGenericState, hid_entry_count> system_ext_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_fullkey_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_handheld_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_dual_left_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_dual_right_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_left_lifo;
-        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_right_lifo;
-        DeviceType device_type;
+        Core::HID::NpadStyleTag style_tag{Core::HID::NpadStyleSet::None};
+        NpadJoyAssignmentMode assignment_mode{NpadJoyAssignmentMode::Dual};
+        NpadFullKeyColorState fullkey_color{};
+        NpadJoyColorState joycon_color{};
+        Lifo<NPadGenericState, hid_entry_count> fullkey_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> handheld_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> joy_dual_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> joy_left_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> joy_right_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> palma_lifo{};
+        Lifo<NPadGenericState, hid_entry_count> system_ext_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_fullkey_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_handheld_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_dual_left_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_dual_right_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_left_lifo{};
+        Lifo<SixAxisSensorState, hid_entry_count> sixaxis_right_lifo{};
+        DeviceType device_type{};
         INSERT_PADDING_BYTES(0x4); // Reserved
-        NPadSystemProperties system_properties;
-        NpadSystemButtonProperties button_properties;
-        Core::HID::NpadBatteryLevel battery_level_dual;
-        Core::HID::NpadBatteryLevel battery_level_left;
-        Core::HID::NpadBatteryLevel battery_level_right;
+        NPadSystemProperties system_properties{};
+        NpadSystemButtonProperties button_properties{};
+        Core::HID::NpadBatteryLevel battery_level_dual{};
+        Core::HID::NpadBatteryLevel battery_level_left{};
+        Core::HID::NpadBatteryLevel battery_level_right{};
         union {
-            Lifo<NfcXcdDeviceHandleStateImpl, 0x2> nfc_xcd_device_lifo{};
-            AppletFooterUi applet_footer;
+            AppletFooterUi applet_footer{};
+            Lifo<NfcXcdDeviceHandleStateImpl, 0x2> nfc_xcd_device_lifo;
         };
         INSERT_PADDING_BYTES(0x20); // Unknown
-        Lifo<NpadGcTriggerState, hid_entry_count> gc_trigger_lifo;
-        NpadLarkType lark_type_l_and_main;
-        NpadLarkType lark_type_r;
-        NpadLuciaType lucia_type;
-        NpadLagonType lagon_type;
-        NpadLagerType lager_type;
+        Lifo<NpadGcTriggerState, hid_entry_count> gc_trigger_lifo{};
+        NpadLarkType lark_type_l_and_main{};
+        NpadLarkType lark_type_r{};
+        NpadLuciaType lucia_type{};
+        NpadLagonType lagon_type{};
+        NpadLagerType lager_type{};
         // FW 13.x Investigate there is some sort of bitflag related to joycons
         INSERT_PADDING_BYTES(0x4);
         INSERT_PADDING_BYTES(0xc08); // Unknown
@@ -511,7 +511,8 @@ private:
     NpadHandheldActivationMode handheld_activation_mode{NpadHandheldActivationMode::Dual};
     NpadCommunicationMode communication_mode{NpadCommunicationMode::Default};
     bool permit_vibration_session_enabled{false};
-    bool analog_stick_use_center_clamp{};
+    bool analog_stick_use_center_clamp{false};
     bool is_in_lr_assignment_mode{false};
+    bool is_controller_initialized{false};
 };
 } // namespace Service::HID

src/core/hle/service/jit/jit.cpp

@@ -0,0 +1,332 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/arm/symbols.h"
+#include "core/core.h"
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/k_code_memory.h"
+#include "core/hle/kernel/k_transfer_memory.h"
+#include "core/hle/result.h"
+#include "core/hle/service/jit/jit.h"
+#include "core/hle/service/jit/jit_context.h"
+#include "core/hle/service/service.h"
+#include "core/memory.h"
+
+namespace Service::JIT {
+
+struct CodeRange {
+    u64 offset;
+    u64 size;
+};
+
+class IJitEnvironment final : public ServiceFramework<IJitEnvironment> {
+public:
+    explicit IJitEnvironment(Core::System& system_, CodeRange user_rx, CodeRange user_ro)
+        : ServiceFramework{system_, "IJitEnvironment", ServiceThreadType::CreateNew},
+          context{system_.Memory()} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0, &IJitEnvironment::GenerateCode, "GenerateCode"},
+            {1, &IJitEnvironment::Control, "Control"},
+            {1000, &IJitEnvironment::LoadPlugin, "LoadPlugin"},
+            {1001, &IJitEnvironment::GetCodeAddress, "GetCodeAddress"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+
+        // Identity map user code range into sysmodule context
+        configuration.user_ro_memory = user_ro;
+        configuration.user_rx_memory = user_rx;
+        configuration.sys_ro_memory = user_ro;
+        configuration.sys_rx_memory = user_rx;
+    }
+
+    void GenerateCode(Kernel::HLERequestContext& ctx) {
+        struct Parameters {
+            u32 data_size;
+            u64 command;
+            CodeRange cr1;
+            CodeRange cr2;
+            Struct32 data;
+        };
+
+        IPC::RequestParser rp{ctx};
+        const auto parameters{rp.PopRaw<Parameters>()};
+        std::vector<u8> input_buffer{ctx.CanReadBuffer() ? ctx.ReadBuffer() : std::vector<u8>()};
+        std::vector<u8> output_buffer(ctx.CanWriteBuffer() ? ctx.GetWriteBufferSize() : 0);
+
+        const VAddr return_ptr{context.AddHeap(0u)};
+        const VAddr cr1_in_ptr{context.AddHeap(parameters.cr1)};
+        const VAddr cr2_in_ptr{context.AddHeap(parameters.cr2)};
+        const VAddr cr1_out_ptr{
+            context.AddHeap(CodeRange{.offset = parameters.cr1.offset, .size = 0})};
+        const VAddr cr2_out_ptr{
+            context.AddHeap(CodeRange{.offset = parameters.cr2.offset, .size = 0})};
+        const VAddr input_ptr{context.AddHeap(input_buffer.data(), input_buffer.size())};
+        const VAddr output_ptr{context.AddHeap(output_buffer.data(), output_buffer.size())};
+        const VAddr data_ptr{context.AddHeap(parameters.data)};
+        const VAddr configuration_ptr{context.AddHeap(configuration)};
+
+        context.CallFunction(callbacks.GenerateCode, return_ptr, cr1_out_ptr, cr2_out_ptr,
+                             configuration_ptr, parameters.command, input_ptr, input_buffer.size(),
+                             cr1_in_ptr, cr2_in_ptr, data_ptr, parameters.data_size, output_ptr,
+                             output_buffer.size());
+
+        const s32 return_value{context.GetHeap<s32>(return_ptr)};
+
+        if (return_value == 0) {
+            system.InvalidateCpuInstructionCacheRange(configuration.user_rx_memory.offset,
+                                                      configuration.user_rx_memory.size);
+
+            if (ctx.CanWriteBuffer()) {
+                context.GetHeap(output_ptr, output_buffer.data(), output_buffer.size());
+                ctx.WriteBuffer(output_buffer.data(), output_buffer.size());
+            }
+            const auto cr1_out{context.GetHeap<CodeRange>(cr1_out_ptr)};
+            const auto cr2_out{context.GetHeap<CodeRange>(cr2_out_ptr)};
+
+            IPC::ResponseBuilder rb{ctx, 8};
+            rb.Push(ResultSuccess);
+            rb.Push<u64>(return_value);
+            rb.PushRaw(cr1_out);
+            rb.PushRaw(cr2_out);
+        } else {
+            LOG_WARNING(Service_JIT, "plugin GenerateCode callback failed");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+        }
+    };
+
+    void Control(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto command{rp.PopRaw<u64>()};
+        const auto input_buffer{ctx.ReadBuffer()};
+        std::vector<u8> output_buffer(ctx.CanWriteBuffer() ? ctx.GetWriteBufferSize() : 0);
+
+        const VAddr return_ptr{context.AddHeap(0u)};
+        const VAddr configuration_ptr{context.AddHeap(configuration)};
+        const VAddr input_ptr{context.AddHeap(input_buffer.data(), input_buffer.size())};
+        const VAddr output_ptr{context.AddHeap(output_buffer.data(), output_buffer.size())};
+        const u64 wrapper_value{
+            context.CallFunction(callbacks.Control, return_ptr, configuration_ptr, command,
+                                 input_ptr, input_buffer.size(), output_ptr, output_buffer.size())};
+        const s32 return_value{context.GetHeap<s32>(return_ptr)};
+
+        if (wrapper_value == 0 && return_value == 0) {
+            if (ctx.CanWriteBuffer()) {
+                context.GetHeap(output_ptr, output_buffer.data(), output_buffer.size());
+                ctx.WriteBuffer(output_buffer.data(), output_buffer.size());
+            }
+            IPC::ResponseBuilder rb{ctx, 3};
+            rb.Push(ResultSuccess);
+            rb.Push(return_value);
+        } else {
+            LOG_WARNING(Service_JIT, "plugin Control callback failed");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+        }
+    }
+
+    void LoadPlugin(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto tmem_size{rp.PopRaw<u64>()};
+        if (tmem_size == 0) {
+            LOG_ERROR(Service_JIT, "attempted to load plugin with empty transfer memory");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+            return;
+        }
+
+        const auto tmem_handle{ctx.GetCopyHandle(0)};
+        auto tmem{system.CurrentProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
+            tmem_handle)};
+        if (tmem.IsNull()) {
+            LOG_ERROR(Service_JIT, "attempted to load plugin with invalid transfer memory handle");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+            return;
+        }
+
+        configuration.work_memory.offset = tmem->GetSourceAddress();
+        configuration.work_memory.size = tmem_size;
+
+        const auto nro_plugin{ctx.ReadBuffer(1)};
+        auto symbols{Core::Symbols::GetSymbols(nro_plugin, true)};
+        const auto GetSymbol{[&](std::string name) { return symbols[name].first; }};
+
+        callbacks =
+            GuestCallbacks{.rtld_fini = GetSymbol("_fini"),
+                           .rtld_init = GetSymbol("_init"),
+                           .Control = GetSymbol("nnjitpluginControl"),
+                           .ResolveBasicSymbols = GetSymbol("nnjitpluginResolveBasicSymbols"),
+                           .SetupDiagnostics = GetSymbol("nnjitpluginSetupDiagnostics"),
+                           .Configure = GetSymbol("nnjitpluginConfigure"),
+                           .GenerateCode = GetSymbol("nnjitpluginGenerateCode"),
+                           .GetVersion = GetSymbol("nnjitpluginGetVersion"),
+                           .Keeper = GetSymbol("nnjitpluginKeeper"),
+                           .OnPrepared = GetSymbol("nnjitpluginOnPrepared")};
+
+        if (callbacks.GetVersion == 0 || callbacks.Configure == 0 || callbacks.GenerateCode == 0 ||
+            callbacks.OnPrepared == 0) {
+            LOG_ERROR(Service_JIT, "plugin does not implement all necessary functionality");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+            return;
+        }
+
+        if (!context.LoadNRO(nro_plugin)) {
+            LOG_ERROR(Service_JIT, "failed to load plugin");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+            return;
+        }
+
+        context.MapProcessMemory(configuration.sys_ro_memory.offset,
+                                 configuration.sys_ro_memory.size);
+        context.MapProcessMemory(configuration.sys_rx_memory.offset,
+                                 configuration.sys_rx_memory.size);
+        context.MapProcessMemory(configuration.work_memory.offset, configuration.work_memory.size);
+
+        if (callbacks.rtld_init != 0) {
+            context.CallFunction(callbacks.rtld_init);
+        }
+
+        const auto version{context.CallFunction(callbacks.GetVersion)};
+        if (version != 1) {
+            LOG_ERROR(Service_JIT, "unknown plugin version {}", version);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+            return;
+        }
+
+        const auto resolve{context.GetHelper("_resolve")};
+        if (callbacks.ResolveBasicSymbols != 0) {
+            context.CallFunction(callbacks.ResolveBasicSymbols, resolve);
+        }
+        const auto resolve_ptr{context.AddHeap(resolve)};
+        if (callbacks.SetupDiagnostics != 0) {
+            context.CallFunction(callbacks.SetupDiagnostics, 0u, resolve_ptr);
+        }
+
+        context.CallFunction(callbacks.Configure, 0u);
+        const auto configuration_ptr{context.AddHeap(configuration)};
+        context.CallFunction(callbacks.OnPrepared, configuration_ptr);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
+
+    void GetCodeAddress(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 6};
+        rb.Push(ResultSuccess);
+        rb.Push(configuration.user_rx_memory.offset);
+        rb.Push(configuration.user_ro_memory.offset);
+    }
+
+private:
+    using Struct32 = std::array<u8, 32>;
+
+    struct GuestCallbacks {
+        VAddr rtld_fini;
+        VAddr rtld_init;
+        VAddr Control;
+        VAddr ResolveBasicSymbols;
+        VAddr SetupDiagnostics;
+        VAddr Configure;
+        VAddr GenerateCode;
+        VAddr GetVersion;
+        VAddr Keeper;
+        VAddr OnPrepared;
+    };
+
+    struct JITConfiguration {
+        CodeRange user_rx_memory;
+        CodeRange user_ro_memory;
+        CodeRange work_memory;
+        CodeRange sys_rx_memory;
+        CodeRange sys_ro_memory;
+    };
+
+    GuestCallbacks callbacks;
+    JITConfiguration configuration;
+    JITContext context;
+};
+
+class JITU final : public ServiceFramework<JITU> {
+public:
+    explicit JITU(Core::System& system_) : ServiceFramework{system_, "jit:u"} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0, &JITU::CreateJitEnvironment, "CreateJitEnvironment"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+
+    void CreateJitEnvironment(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_JIT, "called");
+
+        struct Parameters {
+            u64 rx_size;
+            u64 ro_size;
+        };
+
+        IPC::RequestParser rp{ctx};
+        const auto parameters{rp.PopRaw<Parameters>()};
+        const auto executable_mem_handle{ctx.GetCopyHandle(1)};
+        const auto readable_mem_handle{ctx.GetCopyHandle(2)};
+
+        if (parameters.rx_size == 0 || parameters.ro_size == 0) {
+            LOG_ERROR(Service_JIT, "attempted to init with empty code regions");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+            return;
+        }
+
+        // The copy handle at index 0 is the process handle, but handle tables are
+        // per-process, so there is no point reading it here until we are multiprocess
+        const auto& process{*system.CurrentProcess()};
+
+        auto executable_mem{
+            process.GetHandleTable().GetObject<Kernel::KCodeMemory>(executable_mem_handle)};
+        if (executable_mem.IsNull()) {
+            LOG_ERROR(Service_JIT, "executable_mem is null for handle=0x{:08X}",
+                      executable_mem_handle);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+            return;
+        }
+
+        auto readable_mem{
+            process.GetHandleTable().GetObject<Kernel::KCodeMemory>(readable_mem_handle)};
+        if (readable_mem.IsNull()) {
+            LOG_ERROR(Service_JIT, "readable_mem is null for handle=0x{:08X}", readable_mem_handle);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ResultUnknown);
+            return;
+        }
+
+        const CodeRange user_rx{
+            .offset = executable_mem->GetSourceAddress(),
+            .size = parameters.rx_size,
+        };
+
+        const CodeRange user_ro{
+            .offset = readable_mem->GetSourceAddress(),
+            .size = parameters.ro_size,
+        };
+
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+        rb.Push(ResultSuccess);
+        rb.PushIpcInterface<IJitEnvironment>(system, user_rx, user_ro);
+    }
+};
+
+void InstallInterfaces(SM::ServiceManager& sm, Core::System& system) {
+    std::make_shared<JITU>(system)->InstallAsService(sm);
+}
+
+} // namespace Service::JIT

src/core/hle/service/jit/jit.h

@@ -0,0 +1,20 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+namespace Core {
+class System;
+}
+
+namespace Service::SM {
+class ServiceManager;
+}
+
+namespace Service::JIT {
+
+/// Registers all JIT services with the specified service manager.
+void InstallInterfaces(SM::ServiceManager& sm, Core::System& system);
+
+} // namespace Service::JIT

src/core/hle/service/jit/jit_context.cpp

@@ -0,0 +1,424 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <map>
+#include <span>
+#include <boost/icl/interval_set.hpp>
+#include <dynarmic/interface/A64/a64.h>
+#include <dynarmic/interface/A64/config.h>
+
+#include "common/alignment.h"
+#include "common/common_funcs.h"
+#include "common/div_ceil.h"
+#include "common/logging/log.h"
+#include "core/hle/service/jit/jit_context.h"
+#include "core/memory.h"
+
+namespace Service::JIT {
+
+constexpr std::array<u8, 4> STOP_ARM64 = {
+    0x01, 0x00, 0x00, 0xd4, // svc  #0
+};
+
+constexpr std::array<u8, 8> RESOLVE_ARM64 = {
+    0x21, 0x00, 0x00, 0xd4, // svc  #1
+    0xc0, 0x03, 0x5f, 0xd6, // ret
+};
+
+constexpr std::array<u8, 4> PANIC_ARM64 = {
+    0x41, 0x00, 0x00, 0xd4, // svc  #2
+};
+
+constexpr std::array<u8, 60> MEMMOVE_ARM64 = {
+    0x1f, 0x00, 0x01, 0xeb, // cmp  x0, x1
+    0x83, 0x01, 0x00, 0x54, // b.lo #+34
+    0x42, 0x04, 0x00, 0xd1, // sub  x2, x2, 1
+    0x22, 0x01, 0xf8, 0xb7, // tbnz x2, #63, #+36
+    0x23, 0x68, 0x62, 0x38, // ldrb w3, [x1, x2]
+    0x03, 0x68, 0x22, 0x38, // strb w3, [x0, x2]
+    0xfc, 0xff, 0xff, 0x17, // b    #-16
+    0x24, 0x68, 0x63, 0x38, // ldrb w4, [x1, x3]
+    0x04, 0x68, 0x23, 0x38, // strb w4, [x0, x3]
+    0x63, 0x04, 0x00, 0x91, // add  x3, x3, 1
+    0x7f, 0x00, 0x02, 0xeb, // cmp  x3, x2
+    0x8b, 0xff, 0xff, 0x54, // b.lt #-16
+    0xc0, 0x03, 0x5f, 0xd6, // ret
+    0x03, 0x00, 0x80, 0xd2, // mov  x3, 0
+    0xfc, 0xff, 0xff, 0x17, // b    #-16
+};
+
+constexpr std::array<u8, 28> MEMSET_ARM64 = {
+    0x03, 0x00, 0x80, 0xd2, // mov  x3, 0
+    0x7f, 0x00, 0x02, 0xeb, // cmp  x3, x2
+    0x4b, 0x00, 0x00, 0x54, // b.lt #+8
+    0xc0, 0x03, 0x5f, 0xd6, // ret
+    0x01, 0x68, 0x23, 0x38, // strb w1, [x0, x3]
+    0x63, 0x04, 0x00, 0x91, // add  x3, x3, 1
+    0xfb, 0xff, 0xff, 0x17, // b    #-20
+};
+
+struct HelperFunction {
+    const char* name;
+    const std::span<const u8> data;
+};
+
+constexpr std::array<HelperFunction, 6> HELPER_FUNCTIONS{{
+    {"_stop", STOP_ARM64},
+    {"_resolve", RESOLVE_ARM64},
+    {"_panic", PANIC_ARM64},
+    {"memcpy", MEMMOVE_ARM64},
+    {"memmove", MEMMOVE_ARM64},
+    {"memset", MEMSET_ARM64},
+}};
+
+struct Elf64_Dyn {
+    u64 d_tag;
+    u64 d_un;
+};
+
+struct Elf64_Rela {
+    u64 r_offset;
+    u64 r_info;
+    s64 r_addend;
+};
+
+static constexpr u32 Elf64_RelaType(const Elf64_Rela* rela) {
+    return static_cast<u32>(rela->r_info);
+}
+
+constexpr int DT_RELA = 7;               /* Address of Rela relocs */
+constexpr int DT_RELASZ = 8;             /* Total size of Rela relocs */
+constexpr int R_AARCH64_RELATIVE = 1027; /* Adjust by program base.  */
+
+constexpr size_t STACK_ALIGN = 16;
+
+class JITContextImpl;
+
+using IntervalSet = boost::icl::interval_set<VAddr>::type;
+using IntervalType = boost::icl::interval_set<VAddr>::interval_type;
+
+class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
+public:
+    explicit DynarmicCallbacks64(Core::Memory::Memory& memory_, std::vector<u8>& local_memory_,
+                                 IntervalSet& mapped_ranges_, JITContextImpl& parent_)
+        : memory{memory_}, local_memory{local_memory_},
+          mapped_ranges{mapped_ranges_}, parent{parent_} {}
+
+    u8 MemoryRead8(u64 vaddr) override {
+        return ReadMemory<u8>(vaddr);
+    }
+    u16 MemoryRead16(u64 vaddr) override {
+        return ReadMemory<u16>(vaddr);
+    }
+    u32 MemoryRead32(u64 vaddr) override {
+        return ReadMemory<u32>(vaddr);
+    }
+    u64 MemoryRead64(u64 vaddr) override {
+        return ReadMemory<u64>(vaddr);
+    }
+    u128 MemoryRead128(u64 vaddr) override {
+        return ReadMemory<u128>(vaddr);
+    }
+    std::string MemoryReadCString(u64 vaddr) {
+        std::string result;
+        u8 next;
+
+        while ((next = MemoryRead8(vaddr++)) != 0) {
+            result += next;
+        }
+
+        return result;
+    }
+
+    void MemoryWrite8(u64 vaddr, u8 value) override {
+        WriteMemory<u8>(vaddr, value);
+    }
+    void MemoryWrite16(u64 vaddr, u16 value) override {
+        WriteMemory<u16>(vaddr, value);
+    }
+    void MemoryWrite32(u64 vaddr, u32 value) override {
+        WriteMemory<u32>(vaddr, value);
+    }
+    void MemoryWrite64(u64 vaddr, u64 value) override {
+        WriteMemory<u64>(vaddr, value);
+    }
+    void MemoryWrite128(u64 vaddr, u128 value) override {
+        WriteMemory<u128>(vaddr, value);
+    }
+
+    bool MemoryWriteExclusive8(u64 vaddr, u8 value, u8) override {
+        return WriteMemory<u8>(vaddr, value);
+    }
+    bool MemoryWriteExclusive16(u64 vaddr, u16 value, u16) override {
+        return WriteMemory<u16>(vaddr, value);
+    }
+    bool MemoryWriteExclusive32(u64 vaddr, u32 value, u32) override {
+        return WriteMemory<u32>(vaddr, value);
+    }
+    bool MemoryWriteExclusive64(u64 vaddr, u64 value, u64) override {
+        return WriteMemory<u64>(vaddr, value);
+    }
+    bool MemoryWriteExclusive128(u64 vaddr, u128 value, u128) override {
+        return WriteMemory<u128>(vaddr, value);
+    }
+
+    void CallSVC(u32 swi) override;
+    void ExceptionRaised(u64 pc, Dynarmic::A64::Exception exception) override;
+    void InterpreterFallback(u64 pc, size_t num_instructions) override;
+
+    void AddTicks(u64 ticks) override {}
+    u64 GetTicksRemaining() override {
+        return std::numeric_limits<u32>::max();
+    }
+    u64 GetCNTPCT() override {
+        return 0;
+    }
+
+    template <class T>
+    T ReadMemory(u64 vaddr) {
+        T ret{};
+        if (boost::icl::contains(mapped_ranges, vaddr)) {
+            memory.ReadBlock(vaddr, &ret, sizeof(T));
+        } else if (vaddr + sizeof(T) > local_memory.size()) {
+            LOG_CRITICAL(Service_JIT, "plugin: unmapped read @ 0x{:016x}", vaddr);
+        } else {
+            std::memcpy(&ret, local_memory.data() + vaddr, sizeof(T));
+        }
+        return ret;
+    }
+
+    template <class T>
+    bool WriteMemory(u64 vaddr, const T value) {
+        if (boost::icl::contains(mapped_ranges, vaddr)) {
+            memory.WriteBlock(vaddr, &value, sizeof(T));
+        } else if (vaddr + sizeof(T) > local_memory.size()) {
+            LOG_CRITICAL(Service_JIT, "plugin: unmapped write @ 0x{:016x}", vaddr);
+        } else {
+            std::memcpy(local_memory.data() + vaddr, &value, sizeof(T));
+        }
+        return true;
+    }
+
+private:
+    Core::Memory::Memory& memory;
+    std::vector<u8>& local_memory;
+    IntervalSet& mapped_ranges;
+    JITContextImpl& parent;
+};
+
+class JITContextImpl {
+public:
+    explicit JITContextImpl(Core::Memory::Memory& memory_) : memory{memory_} {
+        callbacks =
+            std::make_unique<DynarmicCallbacks64>(memory, local_memory, mapped_ranges, *this);
+        user_config.callbacks = callbacks.get();
+        jit = std::make_unique<Dynarmic::A64::Jit>(user_config);
+    }
+
+    bool LoadNRO(std::span<const u8> data) {
+        local_memory.clear();
+        local_memory.insert(local_memory.end(), data.begin(), data.end());
+
+        if (FixupRelocations()) {
+            InsertHelperFunctions();
+            InsertStack();
+            return true;
+        } else {
+            return false;
+        }
+    }
+
+    bool FixupRelocations() {
+        const VAddr mod_offset{callbacks->MemoryRead32(4)};
+        if (callbacks->MemoryRead32(mod_offset) != Common::MakeMagic('M', 'O', 'D', '0')) {
+            return false;
+        }
+
+        VAddr dynamic_offset{mod_offset + callbacks->MemoryRead32(mod_offset + 4)};
+        VAddr rela_dyn = 0;
+        size_t num_rela = 0;
+        while (true) {
+            const auto dyn{callbacks->ReadMemory<Elf64_Dyn>(dynamic_offset)};
+            dynamic_offset += sizeof(Elf64_Dyn);
+
+            if (!dyn.d_tag) {
+                break;
+            }
+            if (dyn.d_tag == DT_RELA) {
+                rela_dyn = dyn.d_un;
+            }
+            if (dyn.d_tag == DT_RELASZ) {
+                num_rela = dyn.d_un / sizeof(Elf64_Rela);
+            }
+        }
+
+        for (size_t i = 0; i < num_rela; i++) {
+            const auto rela{callbacks->ReadMemory<Elf64_Rela>(rela_dyn + i * sizeof(Elf64_Rela))};
+            if (Elf64_RelaType(&rela) != R_AARCH64_RELATIVE) {
+                continue;
+            }
+            const VAddr contents{callbacks->MemoryRead64(rela.r_offset)};
+            callbacks->MemoryWrite64(rela.r_offset, contents + rela.r_addend);
+        }
+
+        return true;
+    }
+
+    void InsertHelperFunctions() {
+        for (const auto& [name, contents] : HELPER_FUNCTIONS) {
+            helpers[name] = local_memory.size();
+            local_memory.insert(local_memory.end(), contents.begin(), contents.end());
+        }
+    }
+
+    void InsertStack() {
+        const u64 pad_amount{Common::AlignUp(local_memory.size(), STACK_ALIGN) -
+                             local_memory.size()};
+        local_memory.insert(local_memory.end(), 0x10000 + pad_amount, 0);
+        top_of_stack = local_memory.size();
+        heap_pointer = top_of_stack;
+    }
+
+    void MapProcessMemory(VAddr dest_address, std::size_t size) {
+        mapped_ranges.add(IntervalType{dest_address, dest_address + size});
+    }
+
+    void PushArgument(const void* data, size_t size) {
+        const size_t num_words = Common::DivCeil(size, sizeof(u64));
+        const size_t current_pos = argument_stack.size();
+        argument_stack.insert(argument_stack.end(), num_words, 0);
+        std::memcpy(argument_stack.data() + current_pos, data, size);
+    }
+
+    void SetupArguments() {
+        for (size_t i = 0; i < 8 && i < argument_stack.size(); i++) {
+            jit->SetRegister(i, argument_stack[i]);
+        }
+        if (argument_stack.size() > 8) {
+            const VAddr new_sp = Common::AlignDown(
+                top_of_stack - (argument_stack.size() - 8) * sizeof(u64), STACK_ALIGN);
+            for (size_t i = 8; i < argument_stack.size(); i++) {
+                callbacks->MemoryWrite64(new_sp + (i - 8) * sizeof(u64), argument_stack[i]);
+            }
+            jit->SetSP(new_sp);
+        }
+        argument_stack.clear();
+        heap_pointer = top_of_stack;
+    }
+
+    u64 CallFunction(VAddr func) {
+        jit->SetRegister(30, helpers["_stop"]);
+        jit->SetSP(top_of_stack);
+        SetupArguments();
+
+        jit->SetPC(func);
+        jit->Run();
+        return jit->GetRegister(0);
+    }
+
+    VAddr GetHelper(const std::string& name) {
+        return helpers[name];
+    }
+
+    VAddr AddHeap(const void* data, size_t size) {
+        const size_t num_bytes{Common::AlignUp(size, STACK_ALIGN)};
+        if (heap_pointer + num_bytes > local_memory.size()) {
+            local_memory.insert(local_memory.end(),
+                                (heap_pointer + num_bytes) - local_memory.size(), 0);
+        }
+        const VAddr location{heap_pointer};
+        std::memcpy(local_memory.data() + location, data, size);
+        heap_pointer += num_bytes;
+        return location;
+    }
+
+    void GetHeap(VAddr location, void* data, size_t size) {
+        std::memcpy(data, local_memory.data() + location, size);
+    }
+
+    std::unique_ptr<DynarmicCallbacks64> callbacks;
+    std::vector<u8> local_memory;
+    std::vector<u64> argument_stack;
+    IntervalSet mapped_ranges;
+    Dynarmic::A64::UserConfig user_config;
+    std::unique_ptr<Dynarmic::A64::Jit> jit;
+    std::map<std::string, VAddr, std::less<>> helpers;
+    Core::Memory::Memory& memory;
+    VAddr top_of_stack;
+    VAddr heap_pointer;
+};
+
+void DynarmicCallbacks64::CallSVC(u32 swi) {
+    switch (swi) {
+    case 0:
+        parent.jit->HaltExecution();
+        break;
+
+    case 1: {
+        // X0 contains a char* for a symbol to resolve
+        std::string name{MemoryReadCString(parent.jit->GetRegister(0))};
+        const auto helper{parent.helpers[name]};
+
+        if (helper != 0) {
+            parent.jit->SetRegister(0, helper);
+        } else {
+            LOG_WARNING(Service_JIT, "plugin requested unknown function {}", name);
+            parent.jit->SetRegister(0, parent.helpers["_panic"]);
+        }
+        break;
+    }
+
+    case 2:
+    default:
+        LOG_CRITICAL(Service_JIT, "plugin panicked!");
+        parent.jit->HaltExecution();
+        break;
+    }
+}
+
+void DynarmicCallbacks64::ExceptionRaised(u64 pc, Dynarmic::A64::Exception exception) {
+    LOG_CRITICAL(Service_JIT, "Illegal operation PC @ {:08x}", pc);
+    parent.jit->HaltExecution();
+}
+
+void DynarmicCallbacks64::InterpreterFallback(u64 pc, size_t num_instructions) {
+    LOG_CRITICAL(Service_JIT, "Unimplemented instruction PC @ {:08x}", pc);
+    parent.jit->HaltExecution();
+}
+
+JITContext::JITContext(Core::Memory::Memory& memory)
+    : impl{std::make_unique<JITContextImpl>(memory)} {}
+
+JITContext::~JITContext() {}
+
+bool JITContext::LoadNRO(std::span<const u8> data) {
+    return impl->LoadNRO(data);
+}
+
+void JITContext::MapProcessMemory(VAddr dest_address, std::size_t size) {
+    impl->MapProcessMemory(dest_address, size);
+}
+
+u64 JITContext::CallFunction(VAddr func) {
+    return impl->CallFunction(func);
+}
+
+void JITContext::PushArgument(const void* data, size_t size) {
+    impl->PushArgument(data, size);
+}
+
+VAddr JITContext::GetHelper(const std::string& name) {
+    return impl->GetHelper(name);
+}
+
+VAddr JITContext::AddHeap(const void* data, size_t size) {
+    return impl->AddHeap(data, size);
+}
+
+void JITContext::GetHeap(VAddr location, void* data, size_t size) {
+    impl->GetHeap(location, data, size);
+}
+
+} // namespace Service::JIT

src/core/hle/service/jit/jit_context.h

@@ -0,0 +1,65 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <span>
+#include <string>
+
+#include "common/common_types.h"
+
+namespace Core::Memory {
+class Memory;
+}
+
+namespace Service::JIT {
+
+class JITContextImpl;
+
+class JITContext {
+public:
+    explicit JITContext(Core::Memory::Memory& memory);
+    ~JITContext();
+
+    [[nodiscard]] bool LoadNRO(std::span<const u8> data);
+    void MapProcessMemory(VAddr dest_address, std::size_t size);
+
+    template <typename T, typename... Ts>
+    u64 CallFunction(VAddr func, T argument, Ts... rest) {
+        static_assert(std::is_trivially_copyable_v<T>);
+        PushArgument(&argument, sizeof(argument));
+
+        if constexpr (sizeof...(rest) > 0) {
+            return CallFunction(func, rest...);
+        } else {
+            return CallFunction(func);
+        }
+    }
+
+    u64 CallFunction(VAddr func);
+    VAddr GetHelper(const std::string& name);
+
+    template <typename T>
+    VAddr AddHeap(T argument) {
+        return AddHeap(&argument, sizeof(argument));
+    }
+    VAddr AddHeap(const void* data, size_t size);
+
+    template <typename T>
+    T GetHeap(VAddr location) {
+        static_assert(std::is_trivially_copyable_v<T>);
+        T result;
+        GetHeap(location, &result, sizeof(result));
+        return result;
+    }
+    void GetHeap(VAddr location, void* data, size_t size);
+
+private:
+    std::unique_ptr<JITContextImpl> impl;
+
+    void PushArgument(const void* data, size_t size);
+};
+
+} // namespace Service::JIT

src/core/hle/service/ldr/ldr.cpp

@@ -389,8 +389,12 @@ public:
 
             if (bss_size) {
                 auto block_guard = detail::ScopeExit([&] {
-                    page_table.UnmapCodeMemory(addr + nro_size, bss_addr, bss_size);
-                    page_table.UnmapCodeMemory(addr, nro_addr, nro_size);
+                    page_table.UnmapCodeMemory(
+                        addr + nro_size, bss_addr, bss_size,
+                        Kernel::KPageTable::ICacheInvalidationStrategy::InvalidateRange);
+                    page_table.UnmapCodeMemory(
+                        addr, nro_addr, nro_size,
+                        Kernel::KPageTable::ICacheInvalidationStrategy::InvalidateRange);
                 });
 
                 const ResultCode result{
@@ -570,17 +574,21 @@ public:
         auto& page_table{system.CurrentProcess()->PageTable()};
 
         if (info.bss_size != 0) {
-            CASCADE_CODE(page_table.UnmapCodeMemory(info.nro_address + info.text_size +
-                                                        info.ro_size + info.data_size,
-                                                    info.bss_address, info.bss_size));
+            CASCADE_CODE(page_table.UnmapCodeMemory(
+                info.nro_address + info.text_size + info.ro_size + info.data_size, info.bss_address,
+                info.bss_size, Kernel::KPageTable::ICacheInvalidationStrategy::InvalidateRange));
         }
 
-        CASCADE_CODE(page_table.UnmapCodeMemory(info.nro_address + info.text_size + info.ro_size,
-                                                info.src_addr + info.text_size + info.ro_size,
-                                                info.data_size));
-        CASCADE_CODE(page_table.UnmapCodeMemory(info.nro_address + info.text_size,
-                                                info.src_addr + info.text_size, info.ro_size));
-        CASCADE_CODE(page_table.UnmapCodeMemory(info.nro_address, info.src_addr, info.text_size));
+        CASCADE_CODE(page_table.UnmapCodeMemory(
+            info.nro_address + info.text_size + info.ro_size,
+            info.src_addr + info.text_size + info.ro_size, info.data_size,
+            Kernel::KPageTable::ICacheInvalidationStrategy::InvalidateRange));
+        CASCADE_CODE(page_table.UnmapCodeMemory(
+            info.nro_address + info.text_size, info.src_addr + info.text_size, info.ro_size,
+            Kernel::KPageTable::ICacheInvalidationStrategy::InvalidateRange));
+        CASCADE_CODE(page_table.UnmapCodeMemory(
+            info.nro_address, info.src_addr, info.text_size,
+            Kernel::KPageTable::ICacheInvalidationStrategy::InvalidateRange));
         return ResultSuccess;
     }
 

src/core/hle/service/nvdrv/nvdrv_interface.cpp

@@ -26,7 +26,7 @@ void NVDRV::Open(Kernel::HLERequestContext& ctx) {
         rb.Push<DeviceFD>(0);
         rb.PushEnum(NvResult::NotInitialized);
 
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -61,7 +61,7 @@ void NVDRV::Ioctl1(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -87,7 +87,7 @@ void NVDRV::Ioctl2(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -114,7 +114,7 @@ void NVDRV::Ioctl3(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -139,7 +139,7 @@ void NVDRV::Close(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -170,7 +170,7 @@ void NVDRV::QueryEvent(Kernel::HLERequestContext& ctx) {
 
     if (!is_initialized) {
         ServiceError(ctx, NvResult::NotInitialized);
-        LOG_ERROR(Service_NVDRV, "NvServices is not initalized!");
+        LOG_ERROR(Service_NVDRV, "NvServices is not initialized!");
         return;
     }
 
@@ -230,7 +230,7 @@ void NVDRV::DumpGraphicsMemoryInfo(Kernel::HLERequestContext& ctx) {
 }
 
 NVDRV::NVDRV(Core::System& system_, std::shared_ptr<Module> nvdrv_, const char* name)
-    : ServiceFramework{system_, name}, nvdrv{std::move(nvdrv_)} {
+    : ServiceFramework{system_, name, ServiceThreadType::CreateNew}, nvdrv{std::move(nvdrv_)} {
     static const FunctionInfo functions[] = {
         {0, &NVDRV::Open, "Open"},
         {1, &NVDRV::Ioctl1, "Ioctl"},

src/core/hle/service/nvflinger/buffer_item_consumer.cpp

@@ -21,7 +21,7 @@ Status BufferItemConsumer::AcquireBuffer(BufferItem* item, std::chrono::nanoseco
         return Status::BadValue;
     }
 
-    std::scoped_lock lock(mutex);
+    std::scoped_lock lock{mutex};
 
     if (const auto status = AcquireBufferLocked(item, present_when); status != Status::NoError) {
         if (status != Status::NoBufferAvailable) {
@@ -40,7 +40,7 @@ Status BufferItemConsumer::AcquireBuffer(BufferItem* item, std::chrono::nanoseco
 }
 
 Status BufferItemConsumer::ReleaseBuffer(const BufferItem& item, Fence& release_fence) {
-    std::scoped_lock lock(mutex);
+    std::scoped_lock lock{mutex};
 
     if (const auto status = AddReleaseFenceLocked(item.buf, item.graphic_buffer, release_fence);
         status != Status::NoError) {

src/core/hle/service/nvflinger/buffer_queue_consumer.cpp

@@ -18,9 +18,8 @@ BufferQueueConsumer::BufferQueueConsumer(std::shared_ptr<BufferQueueCore> core_)
 BufferQueueConsumer::~BufferQueueConsumer() = default;
 
 Status BufferQueueConsumer::AcquireBuffer(BufferItem* out_buffer,
-                                          std::chrono::nanoseconds expected_present,
-                                          u64 max_frame_number) {
-    std::scoped_lock lock(core->mutex);
+                                          std::chrono::nanoseconds expected_present) {
+    std::scoped_lock lock{core->mutex};
 
     // Check that the consumer doesn't currently have the maximum number of buffers acquired.
     const s32 num_acquired_buffers{
@@ -50,12 +49,6 @@ Status BufferQueueConsumer::AcquireBuffer(BufferItem* out_buffer,
         while (core->queue.size() > 1 && !core->queue[0].is_auto_timestamp) {
             const auto& buffer_item{core->queue[1]};
 
-            // If dropping entry[0] would leave us with a buffer that the consumer is not yet ready
-            // for, don't drop it.
-            if (max_frame_number && buffer_item.frame_number > max_frame_number) {
-                break;
-            }
-
             // If entry[1] is timely, drop entry[0] (and repeat).
             const auto desired_present = buffer_item.timestamp;
             if (desired_present < expected_present.count() - MAX_REASONABLE_NSEC ||
@@ -127,7 +120,7 @@ Status BufferQueueConsumer::ReleaseBuffer(s32 slot, u64 frame_number, const Fenc
 
     std::shared_ptr<IProducerListener> listener;
     {
-        std::scoped_lock lock(core->mutex);
+        std::scoped_lock lock{core->mutex};
 
         // If the frame number has changed because the buffer has been reallocated, we can ignore
         // this ReleaseBuffer for the old buffer.
@@ -187,7 +180,7 @@ Status BufferQueueConsumer::Connect(std::shared_ptr<IConsumerListener> consumer_
 
     LOG_DEBUG(Service_NVFlinger, "controlled_by_app={}", controlled_by_app);
 
-    std::scoped_lock lock(core->mutex);
+    std::scoped_lock lock{core->mutex};
 
     if (core->is_abandoned) {
         LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
@@ -200,4 +193,39 @@ Status BufferQueueConsumer::Connect(std::shared_ptr<IConsumerListener> consumer_
     return Status::NoError;
 }
 
+Status BufferQueueConsumer::GetReleasedBuffers(u64* out_slot_mask) {
+    if (out_slot_mask == nullptr) {
+        LOG_ERROR(Service_NVFlinger, "out_slot_mask may not be nullptr");
+        return Status::BadValue;
+    }
+
+    std::scoped_lock lock{core->mutex};
+
+    if (core->is_abandoned) {
+        LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
+        return Status::NoInit;
+    }
+
+    u64 mask = 0;
+    for (int s = 0; s < BufferQueueDefs::NUM_BUFFER_SLOTS; ++s) {
+        if (!slots[s].acquire_called) {
+            mask |= (1ULL << s);
+        }
+    }
+
+    // Remove from the mask queued buffers for which acquire has been called, since the consumer
+    // will not receive their buffer addresses and so must retain their cached information
+    auto current(core->queue.begin());
+    while (current != core->queue.end()) {
+        if (current->acquire_called) {
+            mask &= ~(1ULL << current->slot);
+        }
+        ++current;
+    }
+
+    LOG_DEBUG(Service_NVFlinger, "returning mask {}", mask);
+    *out_slot_mask = mask;
+    return Status::NoError;
+}
+
 } // namespace Service::android

src/core/hle/service/nvflinger/buffer_queue_consumer.h

@@ -24,10 +24,10 @@ public:
     explicit BufferQueueConsumer(std::shared_ptr<BufferQueueCore> core_);
     ~BufferQueueConsumer();
 
-    Status AcquireBuffer(BufferItem* out_buffer, std::chrono::nanoseconds expected_present,
-                         u64 max_frame_number = 0);
+    Status AcquireBuffer(BufferItem* out_buffer, std::chrono::nanoseconds expected_present);
     Status ReleaseBuffer(s32 slot, u64 frame_number, const Fence& release_fence);
     Status Connect(std::shared_ptr<IConsumerListener> consumer_listener, bool controlled_by_app);
+    Status GetReleasedBuffers(u64* out_slot_mask);
 
 private:
     std::shared_ptr<BufferQueueCore> core;

src/core/hle/service/nvflinger/buffer_queue_core.cpp

@@ -15,7 +15,7 @@ BufferQueueCore::BufferQueueCore() = default;
 BufferQueueCore::~BufferQueueCore() = default;
 
 void BufferQueueCore::NotifyShutdown() {
-    std::scoped_lock lock(mutex);
+    std::scoped_lock lock{mutex};
 
     is_shutting_down = true;
 
@@ -95,7 +95,6 @@ void BufferQueueCore::FreeBufferLocked(s32 slot) {
 }
 
 void BufferQueueCore::FreeAllBuffersLocked() {
-    queue.clear();
     buffer_has_been_queued = false;
 
     for (s32 slot = 0; slot < BufferQueueDefs::NUM_BUFFER_SLOTS; ++slot) {

src/core/hle/service/nvflinger/buffer_queue_core.h

@@ -73,8 +73,6 @@ private:
     u32 transform_hint{};
     bool is_allocating{};
     mutable std::condition_variable_any is_allocating_condition;
-    bool allow_allocation{true};
-    u64 buffer_age{};
     bool is_shutting_down{};
 };
 

src/core/hle/service/nvflinger/buffer_queue_producer.cpp

@@ -38,7 +38,7 @@ BufferQueueProducer::~BufferQueueProducer() {
 Status BufferQueueProducer::RequestBuffer(s32 slot, std::shared_ptr<GraphicBuffer>* buf) {
     LOG_DEBUG(Service_NVFlinger, "slot {}", slot);
 
-    std::scoped_lock lock(core->mutex);
+    std::scoped_lock lock{core->mutex};
 
     if (core->is_abandoned) {
         LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
@@ -62,11 +62,12 @@ Status BufferQueueProducer::RequestBuffer(s32 slot, std::shared_ptr<GraphicBuffe
 
 Status BufferQueueProducer::SetBufferCount(s32 buffer_count) {
     LOG_DEBUG(Service_NVFlinger, "count = {}", buffer_count);
-    std::shared_ptr<IConsumerListener> listener;
 
+    std::shared_ptr<IConsumerListener> listener;
     {
-        std::scoped_lock lock(core->mutex);
+        std::scoped_lock lock{core->mutex};
         core->WaitWhileAllocatingLocked();
+
         if (core->is_abandoned) {
             LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
             return Status::NoInit;
@@ -120,7 +121,7 @@ Status BufferQueueProducer::SetBufferCount(s32 buffer_count) {
 }
 
 Status BufferQueueProducer::WaitForFreeSlotThenRelock(bool async, s32* found,
-                                                      Status* returnFlags) const {
+                                                      Status* return_flags) const {
     bool try_again = true;
 
     while (try_again) {
@@ -142,10 +143,12 @@ Status BufferQueueProducer::WaitForFreeSlotThenRelock(bool async, s32* found,
             ASSERT(slots[s].buffer_state == BufferState::Free);
             if (slots[s].graphic_buffer != nullptr) {
                 core->FreeBufferLocked(s);
-                *returnFlags |= Status::ReleaseAllBuffers;
+                *return_flags |= Status::ReleaseAllBuffers;
             }
         }
 
+        // Look for a free buffer to give to the client
+        *found = BufferQueueCore::INVALID_BUFFER_SLOT;
         s32 dequeued_count{};
         s32 acquired_count{};
         for (s32 s{}; s < max_buffer_count; ++s) {
@@ -233,70 +236,52 @@ Status BufferQueueProducer::DequeueBuffer(s32* out_slot, Fence* out_fence, bool
     Status return_flags = Status::NoError;
     bool attached_by_consumer = false;
     {
-        std::scoped_lock lock(core->mutex);
+        std::scoped_lock lock{core->mutex};
         core->WaitWhileAllocatingLocked();
+
         if (format == PixelFormat::NoFormat) {
             format = core->default_buffer_format;
         }
 
         // Enable the usage bits the consumer requested
         usage |= core->consumer_usage_bit;
+
+        s32 found{};
+        Status status = WaitForFreeSlotThenRelock(async, &found, &return_flags);
+        if (status != Status::NoError) {
+            return status;
+        }
+
+        // This should not happen
+        if (found == BufferQueueCore::INVALID_BUFFER_SLOT) {
+            LOG_ERROR(Service_NVFlinger, "no available buffer slots");
+            return Status::Busy;
+        }
+
+        *out_slot = found;
+
+        attached_by_consumer = slots[found].attached_by_consumer;
+
         const bool use_default_size = !width && !height;
         if (use_default_size) {
             width = core->default_width;
             height = core->default_height;
         }
 
-        s32 found = BufferItem::INVALID_BUFFER_SLOT;
-        while (found == BufferItem::INVALID_BUFFER_SLOT) {
-            Status status = WaitForFreeSlotThenRelock(async, &found, &return_flags);
-            if (status != Status::NoError) {
-                return status;
-            }
-
-            // This should not happen
-            if (found == BufferQueueCore::INVALID_BUFFER_SLOT) {
-                LOG_DEBUG(Service_NVFlinger, "no available buffer slots");
-                return Status::Busy;
-            }
-
-            const std::shared_ptr<GraphicBuffer>& buffer(slots[found].graphic_buffer);
-
-            // If we are not allowed to allocate new buffers, WaitForFreeSlotThenRelock must have
-            // returned a slot containing a buffer. If this buffer would require reallocation to
-            // meet the requested attributes, we free it and attempt to get another one.
-            if (!core->allow_allocation) {
-                if (buffer->NeedsReallocation(width, height, format, usage)) {
-                    core->FreeBufferLocked(found);
-                    found = BufferItem::INVALID_BUFFER_SLOT;
-                    continue;
-                }
-            }
-        }
-
-        *out_slot = found;
-        attached_by_consumer = slots[found].attached_by_consumer;
         slots[found].buffer_state = BufferState::Dequeued;
 
         const std::shared_ptr<GraphicBuffer>& buffer(slots[found].graphic_buffer);
-
-        if ((buffer == nullptr) || buffer->NeedsReallocation(width, height, format, usage)) {
+        if ((buffer == nullptr) || (buffer->Width() != width) || (buffer->Height() != height) ||
+            (buffer->Format() != format) || ((buffer->Usage() & usage) != usage)) {
             slots[found].acquire_called = false;
             slots[found].graphic_buffer = nullptr;
             slots[found].request_buffer_called = false;
             slots[found].fence = Fence::NoFence();
-            core->buffer_age = 0;
+
             return_flags |= Status::BufferNeedsReallocation;
-        } else {
-            // We add 1 because that will be the frame number when this buffer
-            // is queued
-            core->buffer_age = core->frame_counter + 1 - slots[found].frame_number;
         }
 
-        LOG_DEBUG(Service_NVFlinger, "setting buffer age to {}", core->buffer_age);
-
         *out_fence = slots[found].fence;
-
         slots[found].fence = Fence::NoFence();
     }
 
@@ -310,7 +295,8 @@ Status BufferQueueProducer::DequeueBuffer(s32* out_slot, Fence* out_fence, bool
         }
 
         {
-            std::scoped_lock lock(core->mutex);
+            std::scoped_lock lock{core->mutex};
+
             if (core->is_abandoned) {
                 LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
                 return Status::NoInit;
@@ -327,13 +313,15 @@ Status BufferQueueProducer::DequeueBuffer(s32* out_slot, Fence* out_fence, bool
 
     LOG_DEBUG(Service_NVFlinger, "returning slot={} frame={}, flags={}", *out_slot,
               slots[*out_slot].frame_number, return_flags);
+
     return return_flags;
 }
 
 Status BufferQueueProducer::DetachBuffer(s32 slot) {
     LOG_DEBUG(Service_NVFlinger, "slot {}", slot);
 
-    std::scoped_lock lock(core->mutex);
+    std::scoped_lock lock{core->mutex};
+
     if (core->is_abandoned) {
         LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
         return Status::NoInit;
@@ -368,8 +356,7 @@ Status BufferQueueProducer::DetachNextBuffer(std::shared_ptr<GraphicBuffer>* out
         return Status::BadValue;
     }
 
-    std::scoped_lock lock(core->mutex);
-
+    std::scoped_lock lock{core->mutex};
     core->WaitWhileAllocatingLocked();
 
     if (core->is_abandoned) {
@@ -412,7 +399,7 @@ Status BufferQueueProducer::AttachBuffer(s32* out_slot,
         return Status::BadValue;
     }
 
-    std::scoped_lock lock(core->mutex);
+    std::scoped_lock lock{core->mutex};
     core->WaitWhileAllocatingLocked();
 
     Status return_flags = Status::NoError;
@@ -423,6 +410,7 @@ Status BufferQueueProducer::AttachBuffer(s32* out_slot,
         return status;
     }
 
+    // This should not happen
     if (found == BufferQueueCore::INVALID_BUFFER_SLOT) {
         LOG_ERROR(Service_NVFlinger, "No available buffer slots");
         return Status::Busy;
@@ -466,13 +454,13 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
         return Status::BadValue;
     }
 
-    std::shared_ptr<IConsumerListener> frameAvailableListener;
-    std::shared_ptr<IConsumerListener> frameReplacedListener;
+    std::shared_ptr<IConsumerListener> frame_available_listener;
+    std::shared_ptr<IConsumerListener> frame_replaced_listener;
     s32 callback_ticket{};
     BufferItem item;
 
     {
-        std::scoped_lock lock(core->mutex);
+        std::scoped_lock lock{core->mutex};
 
         if (core->is_abandoned) {
             LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
@@ -541,12 +529,13 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
         item.fence = fence;
         item.is_droppable = core->dequeue_buffer_cannot_block || async;
         item.swap_interval = swap_interval;
+
         sticky_transform = sticky_transform_;
 
         if (core->queue.empty()) {
             // When the queue is empty, we can simply queue this buffer
             core->queue.push_back(item);
-            frameAvailableListener = core->consumer_listener;
+            frame_available_listener = core->consumer_listener;
         } else {
             // When the queue is not empty, we need to look at the front buffer
             // state to see if we need to replace it
@@ -563,10 +552,10 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
                 }
                 // Overwrite the droppable buffer with the incoming one
                 *front = item;
-                frameReplacedListener = core->consumer_listener;
+                frame_replaced_listener = core->consumer_listener;
             } else {
                 core->queue.push_back(item);
-                frameAvailableListener = core->consumer_listener;
+                frame_available_listener = core->consumer_listener;
             }
         }
 
@@ -587,15 +576,15 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
     // Call back without the main BufferQueue lock held, but with the callback lock held so we can
     // ensure that callbacks occur in order
     {
-        std::scoped_lock lock(callback_mutex);
+        std::scoped_lock lock{callback_mutex};
         while (callback_ticket != current_callback_ticket) {
             callback_condition.wait(callback_mutex);
         }
 
-        if (frameAvailableListener != nullptr) {
-            frameAvailableListener->OnFrameAvailable(item);
-        } else if (frameReplacedListener != nullptr) {
-            frameReplacedListener->OnFrameReplaced(item);
+        if (frame_available_listener != nullptr) {
+            frame_available_listener->OnFrameAvailable(item);
+        } else if (frame_replaced_listener != nullptr) {
+            frame_replaced_listener->OnFrameReplaced(item);
         }
 
         ++current_callback_ticket;
@@ -608,7 +597,7 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
 void BufferQueueProducer::CancelBuffer(s32 slot, const Fence& fence) {
     LOG_DEBUG(Service_NVFlinger, "slot {}", slot);
 
-    std::scoped_lock lock(core->mutex);
+    std::scoped_lock lock{core->mutex};
 
     if (core->is_abandoned) {
         LOG_ERROR(Service_NVFlinger, "BufferQueue has been abandoned");
@@ -634,7 +623,7 @@ void BufferQueueProducer::CancelBuffer(s32 slot, const Fence& fence) {
 }
 
 Status BufferQueueProducer::Query(NativeWindow what, s32* out_value) {
-    std::scoped_lock lock(core->mutex);
+    std::scoped_lock lock{core->mutex};
 
     if (out_value == nullptr) {
         LOG_ERROR(Service_NVFlinger, "outValue was nullptr");
@@ -669,13 +658,6 @@ Status BufferQueueProducer::Query(NativeWindow what, s32* out_value) {
     case NativeWindow::ConsumerUsageBits:
         value = core->consumer_usage_bit;
         break;
-    case NativeWindow::BufferAge:
-        if (core->buffer_age > INT32_MAX) {
-            value = 0;
-        } else {
-            value = static_cast<u32>(core->buffer_age);
-        }
-        break;
     default:
         UNREACHABLE();
         return Status::BadValue;
@@ -691,7 +673,7 @@ Status BufferQueueProducer::Query(NativeWindow what, s32* out_value) {
 Status BufferQueueProducer::Connect(const std::shared_ptr<IProducerListener>& listener,
                                     NativeWindowApi api, bool producer_controlled_by_app,
                                     QueueBufferOutput* output) {
-    std::scoped_lock lock(core->mutex);
+    std::scoped_lock lock{core->mutex};
 
     LOG_DEBUG(Service_NVFlinger, "api = {} producer_controlled_by_app = {}", api,
               producer_controlled_by_app);
@@ -737,7 +719,6 @@ Status BufferQueueProducer::Connect(const std::shared_ptr<IProducerListener>& li
     core->buffer_has_been_queued = false;
     core->dequeue_buffer_cannot_block =
         core->consumer_controlled_by_app && producer_controlled_by_app;
-    core->allow_allocation = true;
 
     return status;
 }
@@ -749,7 +730,7 @@ Status BufferQueueProducer::Disconnect(NativeWindowApi api) {
     std::shared_ptr<IConsumerListener> listener;
 
     {
-        std::scoped_lock lock(core->mutex);
+        std::scoped_lock lock{core->mutex};
 
         core->WaitWhileAllocatingLocked();
 
@@ -770,7 +751,7 @@ Status BufferQueueProducer::Disconnect(NativeWindowApi api) {
                 core->SignalDequeueCondition();
                 buffer_wait_event->GetWritableEvent().Signal();
                 listener = core->consumer_listener;
-            } else if (core->connected_api != NativeWindowApi::NoConnectedApi) {
+            } else {
                 LOG_ERROR(Service_NVFlinger, "still connected to another api (cur = {} req = {})",
                           core->connected_api, api);
                 status = Status::BadValue;
@@ -799,7 +780,7 @@ Status BufferQueueProducer::SetPreallocatedBuffer(s32 slot,
         return Status::BadValue;
     }
 
-    std::scoped_lock lock(core->mutex);
+    std::scoped_lock lock{core->mutex};
 
     slots[slot] = {};
     slots[slot].graphic_buffer = buffer;

src/core/hle/service/nvflinger/buffer_queue_producer.h

@@ -66,7 +66,7 @@ public:
 private:
     BufferQueueProducer(const BufferQueueProducer&) = delete;
 
-    Status WaitForFreeSlotThenRelock(bool async, s32* found, Status* returnFlags) const;
+    Status WaitForFreeSlotThenRelock(bool async, s32* found, Status* return_flags) const;
 
     Kernel::KEvent* buffer_wait_event{};
     Service::KernelHelpers::ServiceContext& service_context;

src/core/hle/service/nvflinger/consumer_base.cpp

@@ -18,7 +18,7 @@ ConsumerBase::ConsumerBase(std::unique_ptr<BufferQueueConsumer> consumer_)
     : consumer{std::move(consumer_)} {}
 
 ConsumerBase::~ConsumerBase() {
-    std::scoped_lock lock(mutex);
+    std::scoped_lock lock{mutex};
 
     ASSERT_MSG(is_abandoned, "consumer is not abandoned!");
 }
@@ -36,38 +36,41 @@ void ConsumerBase::FreeBufferLocked(s32 slot_index) {
 }
 
 void ConsumerBase::OnFrameAvailable(const BufferItem& item) {
-    std::scoped_lock lock(mutex);
     LOG_DEBUG(Service_NVFlinger, "called");
 }
 
 void ConsumerBase::OnFrameReplaced(const BufferItem& item) {
-    std::scoped_lock lock(mutex);
     LOG_DEBUG(Service_NVFlinger, "called");
 }
 
 void ConsumerBase::OnBuffersReleased() {
-    std::scoped_lock lock(mutex);
+    std::scoped_lock lock{mutex};
+
     LOG_DEBUG(Service_NVFlinger, "called");
+
+    if (is_abandoned) {
+        // Nothing to do if we're already abandoned.
+        return;
+    }
+
+    u64 mask = 0;
+    consumer->GetReleasedBuffers(&mask);
+    for (int i = 0; i < BufferQueueDefs::NUM_BUFFER_SLOTS; i++) {
+        if (mask & (1ULL << i)) {
+            FreeBufferLocked(i);
+        }
+    }
 }
 
 void ConsumerBase::OnSidebandStreamChanged() {}
 
-Status ConsumerBase::AcquireBufferLocked(BufferItem* item, std::chrono::nanoseconds present_when,
-                                         u64 max_frame_number) {
-    if (is_abandoned) {
-        LOG_ERROR(Service_NVFlinger, "consumer is abandoned!");
-        return Status::NoInit;
-    }
-
-    Status err = consumer->AcquireBuffer(item, present_when, max_frame_number);
+Status ConsumerBase::AcquireBufferLocked(BufferItem* item, std::chrono::nanoseconds present_when) {
+    Status err = consumer->AcquireBuffer(item, present_when);
     if (err != Status::NoError) {
         return err;
     }
 
     if (item->graphic_buffer != nullptr) {
-        if (slots[item->slot].graphic_buffer != nullptr) {
-            FreeBufferLocked(item->slot);
-        }
         slots[item->slot].graphic_buffer = item->graphic_buffer;
     }
 

src/core/hle/service/nvflinger/consumer_base.h

@@ -35,8 +35,7 @@ protected:
     virtual void OnSidebandStreamChanged() override;
 
     void FreeBufferLocked(s32 slot_index);
-    Status AcquireBufferLocked(BufferItem* item, std::chrono::nanoseconds present_when,
-                               u64 max_frame_number = 0);
+    Status AcquireBufferLocked(BufferItem* item, std::chrono::nanoseconds present_when);
     Status ReleaseBufferLocked(s32 slot, const std::shared_ptr<GraphicBuffer> graphic_buffer);
     bool StillTracking(s32 slot, const std::shared_ptr<GraphicBuffer> graphic_buffer) const;
     Status AddReleaseFenceLocked(s32 slot, const std::shared_ptr<GraphicBuffer> graphic_buffer,

src/core/hle/service/nvflinger/graphic_buffer_producer.cpp

@@ -4,8 +4,6 @@
 // Parts of this implementation were base on:
 // https://cs.android.com/android/platform/superproject/+/android-5.1.1_r38:frameworks/native/libs/gui/IGraphicBufferProducer.cpp
 
-#pragma once
-
 #include "core/hle/service/nvflinger/graphic_buffer_producer.h"
 #include "core/hle/service/nvflinger/parcel.h"
 

src/core/hle/service/nvflinger/hos_binder_driver_server.cpp

@@ -14,7 +14,7 @@ HosBinderDriverServer::HosBinderDriverServer(Core::System& system_)
 HosBinderDriverServer::~HosBinderDriverServer() {}
 
 u64 HosBinderDriverServer::RegisterProducer(std::unique_ptr<android::IBinder>&& binder) {
-    std::lock_guard lk{lock};
+    std::scoped_lock lk{lock};
 
     last_id++;
 
@@ -24,7 +24,7 @@ u64 HosBinderDriverServer::RegisterProducer(std::unique_ptr<android::IBinder>&&
 }
 
 android::IBinder* HosBinderDriverServer::TryGetProducer(u64 id) {
-    std::lock_guard lk{lock};
+    std::scoped_lock lk{lock};
 
     if (auto search = producers.find(id); search != producers.end()) {
         return search->second.get();

src/core/hle/service/nvflinger/nvflinger.cpp

@@ -104,7 +104,7 @@ void NVFlinger::SetNVDrvInstance(std::shared_ptr<Nvidia::Module> instance) {
 std::optional<u64> NVFlinger::OpenDisplay(std::string_view name) {
     const auto lock_guard = Lock();
 
-    LOG_DEBUG(Service, "Opening \"{}\" display", name);
+    LOG_DEBUG(Service_NVFlinger, "Opening \"{}\" display", name);
 
     const auto itr =
         std::find_if(displays.begin(), displays.end(),
@@ -219,7 +219,7 @@ VI::Layer* NVFlinger::FindOrCreateLayer(u64 display_id, u64 layer_id) {
     auto* layer = display->FindLayer(layer_id);
 
     if (layer == nullptr) {
-        LOG_DEBUG(Service, "Layer at id {} not found. Trying to create it.", layer_id);
+        LOG_DEBUG(Service_NVFlinger, "Layer at id {} not found. Trying to create it.", layer_id);
         CreateLayerAtId(*display, layer_id);
         return display->FindLayer(layer_id);
     }

src/core/hle/service/service.cpp

@@ -32,6 +32,7 @@
 #include "core/hle/service/glue/glue.h"
 #include "core/hle/service/grc/grc.h"
 #include "core/hle/service/hid/hid.h"
+#include "core/hle/service/jit/jit.h"
 #include "core/hle/service/lbl/lbl.h"
 #include "core/hle/service/ldn/ldn.h"
 #include "core/hle/service/ldr/ldr.h"
@@ -91,8 +92,9 @@ namespace Service {
 }
 
 ServiceFrameworkBase::ServiceFrameworkBase(Core::System& system_, const char* service_name_,
-                                           u32 max_sessions_, InvokerFn* handler_invoker_)
-    : SessionRequestHandler(system_.Kernel(), service_name_), system{system_},
+                                           ServiceThreadType thread_type, u32 max_sessions_,
+                                           InvokerFn* handler_invoker_)
+    : SessionRequestHandler(system_.Kernel(), service_name_, thread_type), system{system_},
       service_name{service_name_}, max_sessions{max_sessions_}, handler_invoker{handler_invoker_} {}
 
 ServiceFrameworkBase::~ServiceFrameworkBase() {
@@ -261,6 +263,7 @@ Services::Services(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system
     Glue::InstallInterfaces(system);
     GRC::InstallInterfaces(*sm, system);
     HID::InstallInterfaces(*sm, system);
+    JIT::InstallInterfaces(*sm, system);
     LBL::InstallInterfaces(*sm, system);
     LDN::InstallInterfaces(*sm, system);
     LDR::InstallInterfaces(*sm, system);

src/core/hle/service/service.h

@@ -114,7 +114,8 @@ private:
                            Kernel::HLERequestContext& ctx);
 
     explicit ServiceFrameworkBase(Core::System& system_, const char* service_name_,
-                                  u32 max_sessions_, InvokerFn* handler_invoker_);
+                                  ServiceThreadType thread_type, u32 max_sessions_,
+                                  InvokerFn* handler_invoker_);
     ~ServiceFrameworkBase() override;
 
     void RegisterHandlersBase(const FunctionInfoBase* functions, std::size_t n);
@@ -176,14 +177,17 @@ protected:
     /**
      * Initializes the handler with no functions installed.
      *
-     * @param system_       The system context to construct this service under.
+     * @param system_ The system context to construct this service under.
      * @param service_name_ Name of the service.
-     * @param max_sessions_ Maximum number of sessions that can be
-     *                      connected to this service at the same time.
+     * @param thread_type Specifies the thread type for this service. If this is set to CreateNew,
+     *                    it creates a new thread for it, otherwise this uses the default thread.
+     * @param max_sessions_ Maximum number of sessions that can be connected to this service at the
+     * same time.
      */
     explicit ServiceFramework(Core::System& system_, const char* service_name_,
+                              ServiceThreadType thread_type = ServiceThreadType::Default,
                               u32 max_sessions_ = ServerSessionCountMax)
-        : ServiceFrameworkBase(system_, service_name_, max_sessions_, Invoker) {}
+        : ServiceFrameworkBase(system_, service_name_, thread_type, max_sessions_, Invoker) {}
 
     /// Registers handlers in the service.
     template <std::size_t N>

src/core/hle/service/sm/sm.cpp

@@ -153,7 +153,7 @@ ResultVal<Kernel::KClientSession*> SM::GetServiceImpl(Kernel::HLERequestContext&
     auto& port = port_result.Unwrap();
     SCOPE_EXIT({ port->GetClientPort().Close(); });
 
-    server_ports.emplace_back(&port->GetServerPort());
+    kernel.RegisterServerObject(&port->GetServerPort());
 
     // Create a new session.
     Kernel::KClientSession* session{};
@@ -206,7 +206,7 @@ void SM::UnregisterService(Kernel::HLERequestContext& ctx) {
 }
 
 SM::SM(ServiceManager& service_manager_, Core::System& system_)
-    : ServiceFramework{system_, "sm:", 4},
+    : ServiceFramework{system_, "sm:", ServiceThreadType::Default, 4},
       service_manager{service_manager_}, kernel{system_.Kernel()} {
     RegisterHandlers({
         {0, &SM::Initialize, "Initialize"},
@@ -224,10 +224,6 @@ SM::SM(ServiceManager& service_manager_, Core::System& system_)
     });
 }
 
-SM::~SM() {
-    for (auto& server_port : server_ports) {
-        server_port->Close();
-    }
-}
+SM::~SM() = default;
 
 } // namespace Service::SM

src/core/hle/service/sm/sm.h

@@ -22,7 +22,6 @@ class KClientPort;
 class KClientSession;
 class KernelCore;
 class KPort;
-class KServerPort;
 class SessionRequestHandler;
 } // namespace Kernel
 
@@ -48,7 +47,6 @@ private:
     ServiceManager& service_manager;
     bool is_initialized{};
     Kernel::KernelCore& kernel;
-    std::vector<Kernel::KServerPort*> server_ports;
 };
 
 class ServiceManager {

src/core/hle/service/sockets/bsd.cpp

@@ -689,6 +689,9 @@ Errno BSD::SetSockOptImpl(s32 fd, u32 level, OptName optname, size_t optlen, con
     case OptName::REUSEADDR:
         ASSERT(value == 0 || value == 1);
         return Translate(socket->SetReuseAddr(value != 0));
+    case OptName::KEEPALIVE:
+        ASSERT(value == 0 || value == 1);
+        return Translate(socket->SetKeepAlive(value != 0));
     case OptName::BROADCAST:
         ASSERT(value == 0 || value == 1);
         return Translate(socket->SetBroadcast(value != 0));
@@ -837,7 +840,8 @@ void BSD::BuildErrnoResponse(Kernel::HLERequestContext& ctx, Errno bsd_errno) co
     rb.PushEnum(bsd_errno);
 }
 
-BSD::BSD(Core::System& system_, const char* name) : ServiceFramework{system_, name} {
+BSD::BSD(Core::System& system_, const char* name)
+    : ServiceFramework{system_, name, ServiceThreadType::CreateNew} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {0, &BSD::RegisterClient, "RegisterClient"},

src/core/hle/service/sockets/sfdnsres.cpp

@@ -2,8 +2,28 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <string_view>
+#include <utility>
+#include <vector>
+
+#include "common/string_util.h"
+#include "common/swap.h"
+#include "core/core.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/service/sockets/sfdnsres.h"
+#include "core/memory.h"
+
+#ifdef _WIN32
+#include <ws2tcpip.h>
+#elif YUZU_UNIX
+#include <arpa/inet.h>
+#include <netdb.h>
+#include <netinet/in.h>
+#include <sys/socket.h>
+#ifndef EAI_NODATA
+#define EAI_NODATA EAI_NONAME
+#endif
+#endif
 
 namespace Service::Sockets {
 
@@ -21,7 +41,7 @@ SFDNSRES::SFDNSRES(Core::System& system_) : ServiceFramework{system_, "sfdnsres"
         {9, nullptr, "CancelRequest"},
         {10, nullptr, "GetHostByNameRequestWithOptions"},
         {11, nullptr, "GetHostByAddrRequestWithOptions"},
-        {12, nullptr, "GetAddrInfoRequestWithOptions"},
+        {12, &SFDNSRES::GetAddrInfoRequestWithOptions, "GetAddrInfoRequestWithOptions"},
         {13, nullptr, "GetNameInfoRequestWithOptions"},
         {14, nullptr, "ResolverSetOptionRequest"},
         {15, nullptr, "ResolverGetOptionRequest"},
@@ -31,7 +51,142 @@ SFDNSRES::SFDNSRES(Core::System& system_) : ServiceFramework{system_, "sfdnsres"
 
 SFDNSRES::~SFDNSRES() = default;
 
-void SFDNSRES::GetAddrInfoRequest(Kernel::HLERequestContext& ctx) {
+enum class NetDbError : s32 {
+    Internal = -1,
+    Success = 0,
+    HostNotFound = 1,
+    TryAgain = 2,
+    NoRecovery = 3,
+    NoData = 4,
+};
+
+static NetDbError AddrInfoErrorToNetDbError(s32 result) {
+    // Best effort guess to map errors
+    switch (result) {
+    case 0:
+        return NetDbError::Success;
+    case EAI_AGAIN:
+        return NetDbError::TryAgain;
+    case EAI_NODATA:
+        return NetDbError::NoData;
+    default:
+        return NetDbError::HostNotFound;
+    }
+}
+
+static std::vector<u8> SerializeAddrInfo(const addrinfo* addrinfo, s32 result_code,
+                                         std::string_view host) {
+    // Adapted from
+    // https://github.com/switchbrew/libnx/blob/c5a9a909a91657a9818a3b7e18c9b91ff0cbb6e3/nx/source/runtime/resolver.c#L190
+    std::vector<u8> data;
+
+    auto* current = addrinfo;
+    while (current != nullptr) {
+        struct SerializedResponseHeader {
+            u32 magic;
+            s32 flags;
+            s32 family;
+            s32 socket_type;
+            s32 protocol;
+            u32 address_length;
+        };
+        static_assert(sizeof(SerializedResponseHeader) == 0x18,
+                      "Response header size must be 0x18 bytes");
+
+        constexpr auto header_size = sizeof(SerializedResponseHeader);
+        const auto addr_size =
+            current->ai_addr && current->ai_addrlen > 0 ? current->ai_addrlen : 4;
+        const auto canonname_size = current->ai_canonname ? strlen(current->ai_canonname) + 1 : 1;
+
+        const auto last_size = data.size();
+        data.resize(last_size + header_size + addr_size + canonname_size);
+
+        // Header in network byte order
+        SerializedResponseHeader header{};
+
+        constexpr auto HEADER_MAGIC = 0xBEEFCAFE;
+        header.magic = htonl(HEADER_MAGIC);
+        header.family = htonl(current->ai_family);
+        header.flags = htonl(current->ai_flags);
+        header.socket_type = htonl(current->ai_socktype);
+        header.protocol = htonl(current->ai_protocol);
+        header.address_length = current->ai_addr ? htonl((u32)current->ai_addrlen) : 0;
+
+        auto* header_ptr = data.data() + last_size;
+        std::memcpy(header_ptr, &header, header_size);
+
+        if (header.address_length == 0) {
+            std::memset(header_ptr + header_size, 0, 4);
+        } else {
+            switch (current->ai_family) {
+            case AF_INET: {
+                struct SockAddrIn {
+                    s16 sin_family;
+                    u16 sin_port;
+                    u32 sin_addr;
+                    u8 sin_zero[8];
+                };
+
+                SockAddrIn serialized_addr{};
+                const auto addr = *reinterpret_cast<sockaddr_in*>(current->ai_addr);
+                serialized_addr.sin_port = htons(addr.sin_port);
+                serialized_addr.sin_family = htons(addr.sin_family);
+                serialized_addr.sin_addr = htonl(addr.sin_addr.s_addr);
+                std::memcpy(header_ptr + header_size, &serialized_addr, sizeof(SockAddrIn));
+
+                char addr_string_buf[64]{};
+                inet_ntop(AF_INET, &addr.sin_addr, addr_string_buf, std::size(addr_string_buf));
+                LOG_INFO(Service, "Resolved host '{}' to IPv4 address {}", host, addr_string_buf);
+                break;
+            }
+            case AF_INET6: {
+                struct SockAddrIn6 {
+                    s16 sin6_family;
+                    u16 sin6_port;
+                    u32 sin6_flowinfo;
+                    u8 sin6_addr[16];
+                    u32 sin6_scope_id;
+                };
+
+                SockAddrIn6 serialized_addr{};
+                const auto addr = *reinterpret_cast<sockaddr_in6*>(current->ai_addr);
+                serialized_addr.sin6_family = htons(addr.sin6_family);
+                serialized_addr.sin6_port = htons(addr.sin6_port);
+                serialized_addr.sin6_flowinfo = htonl(addr.sin6_flowinfo);
+                serialized_addr.sin6_scope_id = htonl(addr.sin6_scope_id);
+                std::memcpy(serialized_addr.sin6_addr, &addr.sin6_addr,
+                            sizeof(SockAddrIn6::sin6_addr));
+                std::memcpy(header_ptr + header_size, &serialized_addr, sizeof(SockAddrIn6));
+
+                char addr_string_buf[64]{};
+                inet_ntop(AF_INET6, &addr.sin6_addr, addr_string_buf, std::size(addr_string_buf));
+                LOG_INFO(Service, "Resolved host '{}' to IPv6 address {}", host, addr_string_buf);
+                break;
+            }
+            default:
+                std::memcpy(header_ptr + header_size, current->ai_addr, addr_size);
+                break;
+            }
+        }
+        if (current->ai_canonname) {
+            std::memcpy(header_ptr + addr_size, current->ai_canonname, canonname_size);
+        } else {
+            *(header_ptr + header_size + addr_size) = 0;
+        }
+
+        current = current->ai_next;
+    }
+
+    // 4-byte sentinel value
+    data.push_back(0);
+    data.push_back(0);
+    data.push_back(0);
+    data.push_back(0);
+
+    return data;
+}
+
+static std::pair<u32, s32> GetAddrInfoRequestImpl(Kernel::HLERequestContext& ctx) {
     struct Parameters {
         u8 use_nsd_resolve;
         u32 unknown;
@@ -42,11 +197,51 @@ void SFDNSRES::GetAddrInfoRequest(Kernel::HLERequestContext& ctx) {
     const auto parameters = rp.PopRaw<Parameters>();
 
     LOG_WARNING(Service,
-                "(STUBBED) called. use_nsd_resolve={}, unknown=0x{:08X}, process_id=0x{:016X}",
+                "called with ignored parameters: use_nsd_resolve={}, unknown={}, process_id={}",
                 parameters.use_nsd_resolve, parameters.unknown, parameters.process_id);
 
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(ResultSuccess);
+    const auto host_buffer = ctx.ReadBuffer(0);
+    const std::string host = Common::StringFromBuffer(host_buffer);
+
+    const auto service_buffer = ctx.ReadBuffer(1);
+    const std::string service = Common::StringFromBuffer(service_buffer);
+
+    addrinfo* addrinfo;
+    // Pass null for hints. Serialized hints are also passed in a buffer, but are ignored for now
+    s32 result_code = getaddrinfo(host.c_str(), service.c_str(), nullptr, &addrinfo);
+
+    u32 data_size = 0;
+    if (result_code == 0 && addrinfo != nullptr) {
+        const std::vector<u8>& data = SerializeAddrInfo(addrinfo, result_code, host);
+        data_size = static_cast<u32>(data.size());
+        freeaddrinfo(addrinfo);
+
+        ctx.WriteBuffer(data, 0);
+    }
+
+    return std::make_pair(data_size, result_code);
 }
 
-} // namespace Service::Sockets
+void SFDNSRES::GetAddrInfoRequest(Kernel::HLERequestContext& ctx) {
+    auto [data_size, result_code] = GetAddrInfoRequestImpl(ctx);
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(ResultSuccess);
+    rb.Push(static_cast<s32>(AddrInfoErrorToNetDbError(result_code))); // NetDBErrorCode
+    rb.Push(result_code);                                              // errno
+    rb.Push(data_size);                                                // serialized size
+}
+
+void SFDNSRES::GetAddrInfoRequestWithOptions(Kernel::HLERequestContext& ctx) {
+    // Additional options are ignored
+    auto [data_size, result_code] = GetAddrInfoRequestImpl(ctx);
+
+    IPC::ResponseBuilder rb{ctx, 5};
+    rb.Push(ResultSuccess);
+    rb.Push(data_size);                                                // serialized size
+    rb.Push(result_code);                                              // errno
+    rb.Push(static_cast<s32>(AddrInfoErrorToNetDbError(result_code))); // NetDBErrorCode
+    rb.Push(0);
+}
+
+} // namespace Service::Sockets

src/core/hle/service/sockets/sfdnsres.h

@@ -19,6 +19,7 @@ public:
 
 private:
     void GetAddrInfoRequest(Kernel::HLERequestContext& ctx);
+    void GetAddrInfoRequestWithOptions(Kernel::HLERequestContext& ctx);
 };
 
 } // namespace Service::Sockets

src/core/hle/service/sockets/sockets.h

@@ -46,6 +46,7 @@ enum class Protocol : u32 {
 
 enum class OptName : u32 {
     REUSEADDR = 0x4,
+    KEEPALIVE = 0x8,
     BROADCAST = 0x20,
     LINGER = 0x80,
     SNDBUF = 0x1001,

src/core/hle/service/vi/vi.cpp

@@ -77,7 +77,8 @@ static_assert(sizeof(NativeWindow) == 0x28, "NativeWindow has wrong size");
 class IHOSBinderDriver final : public ServiceFramework<IHOSBinderDriver> {
 public:
     explicit IHOSBinderDriver(Core::System& system_, NVFlinger::HosBinderDriverServer& server_)
-        : ServiceFramework{system_, "IHOSBinderDriver"}, server(server_) {
+        : ServiceFramework{system_, "IHOSBinderDriver", ServiceThreadType::CreateNew},
+          server(server_) {
         static const FunctionInfo functions[] = {
             {0, &IHOSBinderDriver::TransactParcel, "TransactParcel"},
             {1, &IHOSBinderDriver::AdjustRefcount, "AdjustRefcount"},

src/core/network/network.cpp

@@ -600,6 +600,10 @@ Errno Socket::SetReuseAddr(bool enable) {
     return SetSockOpt<u32>(fd, SO_REUSEADDR, enable ? 1 : 0);
 }
 
+Errno Socket::SetKeepAlive(bool enable) {
+    return SetSockOpt<u32>(fd, SO_KEEPALIVE, enable ? 1 : 0);
+}
+
 Errno Socket::SetBroadcast(bool enable) {
     return SetSockOpt<u32>(fd, SO_BROADCAST, enable ? 1 : 0);
 }

src/core/network/sockets.h

@@ -67,6 +67,8 @@ public:
 
     Errno SetReuseAddr(bool enable);
 
+    Errno SetKeepAlive(bool enable);
+
     Errno SetBroadcast(bool enable);
 
     Errno SetSndBuf(u32 value);

src/core/perf_stats.cpp

@@ -53,13 +53,13 @@ PerfStats::~PerfStats() {
 }
 
 void PerfStats::BeginSystemFrame() {
-    std::lock_guard lock{object_mutex};
+    std::scoped_lock lock{object_mutex};
 
     frame_begin = Clock::now();
 }
 
 void PerfStats::EndSystemFrame() {
-    std::lock_guard lock{object_mutex};
+    std::scoped_lock lock{object_mutex};
 
     auto frame_end = Clock::now();
     const auto frame_time = frame_end - frame_begin;
@@ -79,7 +79,7 @@ void PerfStats::EndGameFrame() {
 }
 
 double PerfStats::GetMeanFrametime() const {
-    std::lock_guard lock{object_mutex};
+    std::scoped_lock lock{object_mutex};
 
     if (current_index <= IgnoreFrames) {
         return 0;
@@ -91,7 +91,7 @@ double PerfStats::GetMeanFrametime() const {
 }
 
 PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us) {
-    std::lock_guard lock{object_mutex};
+    std::scoped_lock lock{object_mutex};
 
     const auto now = Clock::now();
     // Walltime elapsed since stats were reset
@@ -120,7 +120,7 @@ PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us
 }
 
 double PerfStats::GetLastFrameTimeScale() const {
-    std::lock_guard lock{object_mutex};
+    std::scoped_lock lock{object_mutex};
 
     constexpr double FRAME_LENGTH = 1.0 / 60;
     return duration_cast<DoubleSecs>(previous_frame_length).count() / FRAME_LENGTH;

src/core/tools/freezer.cpp

@@ -80,7 +80,7 @@ bool Freezer::IsActive() const {
 }
 
 void Freezer::Clear() {
-    std::lock_guard lock{entries_mutex};
+    std::scoped_lock lock{entries_mutex};
 
     LOG_DEBUG(Common_Memory, "Clearing all frozen memory values.");
 
@@ -88,7 +88,7 @@ void Freezer::Clear() {
 }
 
 u64 Freezer::Freeze(VAddr address, u32 width) {
-    std::lock_guard lock{entries_mutex};
+    std::scoped_lock lock{entries_mutex};
 
     const auto current_value = MemoryReadWidth(memory, width, address);
     entries.push_back({address, width, current_value});
@@ -101,7 +101,7 @@ u64 Freezer::Freeze(VAddr address, u32 width) {
 }
 
 void Freezer::Unfreeze(VAddr address) {
-    std::lock_guard lock{entries_mutex};
+    std::scoped_lock lock{entries_mutex};
 
     LOG_DEBUG(Common_Memory, "Unfreezing memory for address={:016X}", address);
 
@@ -109,13 +109,13 @@ void Freezer::Unfreeze(VAddr address) {
 }
 
 bool Freezer::IsFrozen(VAddr address) const {
-    std::lock_guard lock{entries_mutex};
+    std::scoped_lock lock{entries_mutex};
 
     return FindEntry(address) != entries.cend();
 }
 
 void Freezer::SetFrozenValue(VAddr address, u64 value) {
-    std::lock_guard lock{entries_mutex};
+    std::scoped_lock lock{entries_mutex};
 
     const auto iter = FindEntry(address);
 
@@ -132,7 +132,7 @@ void Freezer::SetFrozenValue(VAddr address, u64 value) {
 }
 
 std::optional<Freezer::Entry> Freezer::GetEntry(VAddr address) const {
-    std::lock_guard lock{entries_mutex};
+    std::scoped_lock lock{entries_mutex};
 
     const auto iter = FindEntry(address);
 
@@ -144,7 +144,7 @@ std::optional<Freezer::Entry> Freezer::GetEntry(VAddr address) const {
 }
 
 std::vector<Freezer::Entry> Freezer::GetEntries() const {
-    std::lock_guard lock{entries_mutex};
+    std::scoped_lock lock{entries_mutex};
 
     return entries;
 }
@@ -165,7 +165,7 @@ void Freezer::FrameCallback(std::uintptr_t, std::chrono::nanoseconds ns_late) {
         return;
     }
 
-    std::lock_guard lock{entries_mutex};
+    std::scoped_lock lock{entries_mutex};
 
     for (const auto& entry : entries) {
         LOG_DEBUG(Common_Memory,
@@ -178,7 +178,7 @@ void Freezer::FrameCallback(std::uintptr_t, std::chrono::nanoseconds ns_late) {
 }
 
 void Freezer::FillEntryReads() {
-    std::lock_guard lock{entries_mutex};
+    std::scoped_lock lock{entries_mutex};
 
     LOG_DEBUG(Common_Memory, "Updating memory freeze entries to current values.");
 

src/input_common/drivers/sdl_driver.cpp

@@ -62,7 +62,7 @@ public:
 
     bool UpdateMotion(SDL_ControllerSensorEvent event) {
         constexpr float gravity_constant = 9.80665f;
-        std::lock_guard lock{mutex};
+        std::scoped_lock lock{mutex};
         const u64 time_difference = event.timestamp - last_motion_update;
         last_motion_update = event.timestamp;
         switch (event.sensor) {
@@ -241,7 +241,7 @@ private:
 };
 
 std::shared_ptr<SDLJoystick> SDLDriver::GetSDLJoystickByGUID(const std::string& guid, int port) {
-    std::lock_guard lock{joystick_map_mutex};
+    std::scoped_lock lock{joystick_map_mutex};
     const auto it = joystick_map.find(guid);
 
     if (it != joystick_map.end()) {
@@ -263,7 +263,7 @@ std::shared_ptr<SDLJoystick> SDLDriver::GetSDLJoystickBySDLID(SDL_JoystickID sdl
     auto sdl_joystick = SDL_JoystickFromInstanceID(sdl_id);
     const std::string guid = GetGUID(sdl_joystick);
 
-    std::lock_guard lock{joystick_map_mutex};
+    std::scoped_lock lock{joystick_map_mutex};
     const auto map_it = joystick_map.find(guid);
 
     if (map_it == joystick_map.end()) {
@@ -297,7 +297,7 @@ void SDLDriver::InitJoystick(int joystick_index) {
 
     const std::string guid = GetGUID(sdl_joystick);
 
-    std::lock_guard lock{joystick_map_mutex};
+    std::scoped_lock lock{joystick_map_mutex};
     if (joystick_map.find(guid) == joystick_map.end()) {
         auto joystick = std::make_shared<SDLJoystick>(guid, 0, sdl_joystick, sdl_gamecontroller);
         PreSetController(joystick->GetPadIdentifier());
@@ -326,7 +326,7 @@ void SDLDriver::InitJoystick(int joystick_index) {
 void SDLDriver::CloseJoystick(SDL_Joystick* sdl_joystick) {
     const std::string guid = GetGUID(sdl_joystick);
 
-    std::lock_guard lock{joystick_map_mutex};
+    std::scoped_lock lock{joystick_map_mutex};
     // This call to guid is safe since the joystick is guaranteed to be in the map
     const auto& joystick_guid_list = joystick_map[guid];
     const auto joystick_it = std::find_if(joystick_guid_list.begin(), joystick_guid_list.end(),
@@ -392,7 +392,7 @@ void SDLDriver::HandleGameControllerEvent(const SDL_Event& event) {
 }
 
 void SDLDriver::CloseJoysticks() {
-    std::lock_guard lock{joystick_map_mutex};
+    std::scoped_lock lock{joystick_map_mutex};
     joystick_map.clear();
 }
 

src/input_common/input_engine.cpp

@@ -8,37 +8,37 @@
 namespace InputCommon {
 
 void InputEngine::PreSetController(const PadIdentifier& identifier) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     controller_list.try_emplace(identifier);
 }
 
 void InputEngine::PreSetButton(const PadIdentifier& identifier, int button) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     ControllerData& controller = controller_list.at(identifier);
     controller.buttons.try_emplace(button, false);
 }
 
 void InputEngine::PreSetHatButton(const PadIdentifier& identifier, int button) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     ControllerData& controller = controller_list.at(identifier);
     controller.hat_buttons.try_emplace(button, u8{0});
 }
 
 void InputEngine::PreSetAxis(const PadIdentifier& identifier, int axis) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     ControllerData& controller = controller_list.at(identifier);
     controller.axes.try_emplace(axis, 0.0f);
 }
 
 void InputEngine::PreSetMotion(const PadIdentifier& identifier, int motion) {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     ControllerData& controller = controller_list.at(identifier);
     controller.motions.try_emplace(motion);
 }
 
 void InputEngine::SetButton(const PadIdentifier& identifier, int button, bool value) {
     {
-        std::lock_guard lock{mutex};
+        std::scoped_lock lock{mutex};
         ControllerData& controller = controller_list.at(identifier);
         if (!configuring) {
             controller.buttons.insert_or_assign(button, value);
@@ -49,7 +49,7 @@ void InputEngine::SetButton(const PadIdentifier& identifier, int button, bool va
 
 void InputEngine::SetHatButton(const PadIdentifier& identifier, int button, u8 value) {
     {
-        std::lock_guard lock{mutex};
+        std::scoped_lock lock{mutex};
         ControllerData& controller = controller_list.at(identifier);
         if (!configuring) {
             controller.hat_buttons.insert_or_assign(button, value);
@@ -60,7 +60,7 @@ void InputEngine::SetHatButton(const PadIdentifier& identifier, int button, u8 v
 
 void InputEngine::SetAxis(const PadIdentifier& identifier, int axis, f32 value) {
     {
-        std::lock_guard lock{mutex};
+        std::scoped_lock lock{mutex};
         ControllerData& controller = controller_list.at(identifier);
         if (!configuring) {
             controller.axes.insert_or_assign(axis, value);
@@ -71,7 +71,7 @@ void InputEngine::SetAxis(const PadIdentifier& identifier, int axis, f32 value)
 
 void InputEngine::SetBattery(const PadIdentifier& identifier, Common::Input::BatteryLevel value) {
     {
-        std::lock_guard lock{mutex};
+        std::scoped_lock lock{mutex};
         ControllerData& controller = controller_list.at(identifier);
         if (!configuring) {
             controller.battery = value;
@@ -82,7 +82,7 @@ void InputEngine::SetBattery(const PadIdentifier& identifier, Common::Input::Bat
 
 void InputEngine::SetMotion(const PadIdentifier& identifier, int motion, const BasicMotion& value) {
     {
-        std::lock_guard lock{mutex};
+        std::scoped_lock lock{mutex};
         ControllerData& controller = controller_list.at(identifier);
         if (!configuring) {
             controller.motions.insert_or_assign(motion, value);
@@ -92,7 +92,7 @@ void InputEngine::SetMotion(const PadIdentifier& identifier, int motion, const B
 }
 
 bool InputEngine::GetButton(const PadIdentifier& identifier, int button) const {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     const auto controller_iter = controller_list.find(identifier);
     if (controller_iter == controller_list.cend()) {
         LOG_ERROR(Input, "Invalid identifier guid={}, pad={}, port={}", identifier.guid.RawString(),
@@ -109,7 +109,7 @@ bool InputEngine::GetButton(const PadIdentifier& identifier, int button) const {
 }
 
 bool InputEngine::GetHatButton(const PadIdentifier& identifier, int button, u8 direction) const {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     const auto controller_iter = controller_list.find(identifier);
     if (controller_iter == controller_list.cend()) {
         LOG_ERROR(Input, "Invalid identifier guid={}, pad={}, port={}", identifier.guid.RawString(),
@@ -126,7 +126,7 @@ bool InputEngine::GetHatButton(const PadIdentifier& identifier, int button, u8 d
 }
 
 f32 InputEngine::GetAxis(const PadIdentifier& identifier, int axis) const {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     const auto controller_iter = controller_list.find(identifier);
     if (controller_iter == controller_list.cend()) {
         LOG_ERROR(Input, "Invalid identifier guid={}, pad={}, port={}", identifier.guid.RawString(),
@@ -143,7 +143,7 @@ f32 InputEngine::GetAxis(const PadIdentifier& identifier, int axis) const {
 }
 
 Common::Input::BatteryLevel InputEngine::GetBattery(const PadIdentifier& identifier) const {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     const auto controller_iter = controller_list.find(identifier);
     if (controller_iter == controller_list.cend()) {
         LOG_ERROR(Input, "Invalid identifier guid={}, pad={}, port={}", identifier.guid.RawString(),
@@ -155,7 +155,7 @@ Common::Input::BatteryLevel InputEngine::GetBattery(const PadIdentifier& identif
 }
 
 BasicMotion InputEngine::GetMotion(const PadIdentifier& identifier, int motion) const {
-    std::lock_guard lock{mutex};
+    std::scoped_lock lock{mutex};
     const auto controller_iter = controller_list.find(identifier);
     if (controller_iter == controller_list.cend()) {
         LOG_ERROR(Input, "Invalid identifier guid={}, pad={}, port={}", identifier.guid.RawString(),
@@ -186,7 +186,7 @@ void InputEngine::ResetAnalogState() {
 }
 
 void InputEngine::TriggerOnButtonChange(const PadIdentifier& identifier, int button, bool value) {
-    std::lock_guard lock{mutex_callback};
+    std::scoped_lock lock{mutex_callback};
     for (const auto& poller_pair : callback_list) {
         const InputIdentifier& poller = poller_pair.second;
         if (!IsInputIdentifierEqual(poller, identifier, EngineInputType::Button, button)) {
@@ -214,7 +214,7 @@ void InputEngine::TriggerOnButtonChange(const PadIdentifier& identifier, int but
 }
 
 void InputEngine::TriggerOnHatButtonChange(const PadIdentifier& identifier, int button, u8 value) {
-    std::lock_guard lock{mutex_callback};
+    std::scoped_lock lock{mutex_callback};
     for (const auto& poller_pair : callback_list) {
         const InputIdentifier& poller = poller_pair.second;
         if (!IsInputIdentifierEqual(poller, identifier, EngineInputType::HatButton, button)) {
@@ -243,7 +243,7 @@ void InputEngine::TriggerOnHatButtonChange(const PadIdentifier& identifier, int
 }
 
 void InputEngine::TriggerOnAxisChange(const PadIdentifier& identifier, int axis, f32 value) {
-    std::lock_guard lock{mutex_callback};
+    std::scoped_lock lock{mutex_callback};
     for (const auto& poller_pair : callback_list) {
         const InputIdentifier& poller = poller_pair.second;
         if (!IsInputIdentifierEqual(poller, identifier, EngineInputType::Analog, axis)) {
@@ -270,7 +270,7 @@ void InputEngine::TriggerOnAxisChange(const PadIdentifier& identifier, int axis,
 
 void InputEngine::TriggerOnBatteryChange(const PadIdentifier& identifier,
                                          [[maybe_unused]] Common::Input::BatteryLevel value) {
-    std::lock_guard lock{mutex_callback};
+    std::scoped_lock lock{mutex_callback};
     for (const auto& poller_pair : callback_list) {
         const InputIdentifier& poller = poller_pair.second;
         if (!IsInputIdentifierEqual(poller, identifier, EngineInputType::Battery, 0)) {
@@ -284,7 +284,7 @@ void InputEngine::TriggerOnBatteryChange(const PadIdentifier& identifier,
 
 void InputEngine::TriggerOnMotionChange(const PadIdentifier& identifier, int motion,
                                         const BasicMotion& value) {
-    std::lock_guard lock{mutex_callback};
+    std::scoped_lock lock{mutex_callback};
     for (const auto& poller_pair : callback_list) {
         const InputIdentifier& poller = poller_pair.second;
         if (!IsInputIdentifierEqual(poller, identifier, EngineInputType::Motion, motion)) {
@@ -346,18 +346,18 @@ const std::string& InputEngine::GetEngineName() const {
 }
 
 int InputEngine::SetCallback(InputIdentifier input_identifier) {
-    std::lock_guard lock{mutex_callback};
+    std::scoped_lock lock{mutex_callback};
     callback_list.insert_or_assign(last_callback_key, std::move(input_identifier));
     return last_callback_key++;
 }
 
 void InputEngine::SetMappingCallback(MappingCallback callback) {
-    std::lock_guard lock{mutex_callback};
+    std::scoped_lock lock{mutex_callback};
     mapping_callback = std::move(callback);
 }
 
 void InputEngine::DeleteCallback(int key) {
-    std::lock_guard lock{mutex_callback};
+    std::scoped_lock lock{mutex_callback};
     const auto& iterator = callback_list.find(key);
     if (iterator == callback_list.end()) {
         LOG_ERROR(Input, "Tried to delete non-existent callback {}", key);

src/shader_recompiler/backend/glsl/emit_glsl_context_get_set.cpp

@@ -35,6 +35,15 @@ std::string_view OutputVertexIndex(EmitContext& ctx) {
     return ctx.stage == Stage::TessellationControl ? "[gl_InvocationID]" : "";
 }
 
+std::string ChooseCbuf(EmitContext& ctx, const IR::Value& binding, std::string_view index) {
+    if (binding.IsImmediate()) {
+        return fmt::format("{}_cbuf{}[{}]", ctx.stage_name, binding.U32(), index);
+    } else {
+        const auto binding_var{ctx.var_alloc.Consume(binding)};
+        return fmt::format("GetCbufIndirect({},{})", binding_var, index);
+    }
+}
+
 void GetCbuf(EmitContext& ctx, std::string_view ret, const IR::Value& binding,
              const IR::Value& offset, u32 num_bits, std::string_view cast = {},
              std::string_view bit_offset = {}) {
@@ -55,8 +64,8 @@ void GetCbuf(EmitContext& ctx, std::string_view ret, const IR::Value& binding,
     const auto swizzle{is_immediate ? fmt::format(".{}", OffsetSwizzle(offset.U32()))
                                     : fmt::format("[({}>>2)%4]", offset_var)};
 
-    const auto cbuf{fmt::format("{}_cbuf{}", ctx.stage_name, binding.U32())};
-    const auto cbuf_cast{fmt::format("{}({}[{}]{{}})", cast, cbuf, index)};
+    const auto cbuf{ChooseCbuf(ctx, binding, index)};
+    const auto cbuf_cast{fmt::format("{}({}{{}})", cast, cbuf)};
     const auto extraction{num_bits == 32 ? cbuf_cast
                                          : fmt::format("bitfieldExtract({},int({}),{})", cbuf_cast,
                                                        bit_offset, num_bits)};
@@ -140,9 +149,9 @@ void EmitGetCbufF32(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding,
 
 void EmitGetCbufU32x2(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding,
                       const IR::Value& offset) {
-    const auto cbuf{fmt::format("{}_cbuf{}", ctx.stage_name, binding.U32())};
     const auto cast{ctx.profile.has_gl_cbuf_ftou_bug ? "" : "ftou"};
     if (offset.IsImmediate()) {
+        const auto cbuf{fmt::format("{}_cbuf{}", ctx.stage_name, binding.U32())};
         static constexpr u32 cbuf_size{0x10000};
         const u32 u32_offset{offset.U32()};
         const s32 signed_offset{static_cast<s32>(offset.U32())};
@@ -162,17 +171,17 @@ void EmitGetCbufU32x2(EmitContext& ctx, IR::Inst& inst, const IR::Value& binding
         return;
     }
     const auto offset_var{ctx.var_alloc.Consume(offset)};
+    const auto cbuf{ChooseCbuf(ctx, binding, fmt::format("{}>>4", offset_var))};
     if (!ctx.profile.has_gl_component_indexing_bug) {
-        ctx.AddU32x2("{}=uvec2({}({}[{}>>4][({}>>2)%4]),{}({}[({}+4)>>4][(({}+4)>>2)%4]));", inst,
-                     cast, cbuf, offset_var, offset_var, cast, cbuf, offset_var, offset_var);
+        ctx.AddU32x2("{}=uvec2({}({}[({}>>2)%4]),{}({}[(({}+4)>>2)%4]));", inst, cast, cbuf,
+                     offset_var, cast, cbuf, offset_var);
         return;
     }
     const auto ret{ctx.var_alloc.Define(inst, GlslVarType::U32x2)};
     const auto cbuf_offset{fmt::format("{}>>2", offset_var)};
     for (u32 swizzle = 0; swizzle < 4; ++swizzle) {
-        ctx.Add("if(({}&3)=={}){}=uvec2({}({}[{}>>4].{}),{}({}[({}+4)>>4].{}));", cbuf_offset,
-                swizzle, ret, cast, cbuf, offset_var, "xyzw"[swizzle], cast, cbuf, offset_var,
-                "xyzw"[(swizzle + 1) % 4]);
+        ctx.Add("if(({}&3)=={}){}=uvec2({}({}.{}),{}({}.{}));", cbuf_offset, swizzle, ret, cast,
+                cbuf, "xyzw"[swizzle], cast, cbuf, "xyzw"[(swizzle + 1) % 4]);
     }
 }
 

src/shader_recompiler/backend/glsl/glsl_emit_context.cpp

@@ -359,6 +359,7 @@ EmitContext::EmitContext(IR::Program& program, Bindings& bindings, const Profile
         header += "layout(location=0) uniform vec4 scaling;";
     }
     DefineConstantBuffers(bindings);
+    DefineConstantBufferIndirect();
     DefineStorageBuffers(bindings);
     SetupImages(bindings);
     SetupTextures(bindings);
@@ -436,6 +437,24 @@ void EmitContext::DefineConstantBuffers(Bindings& bindings) {
     }
 }
 
+void EmitContext::DefineConstantBufferIndirect() {
+    if (!info.uses_cbuf_indirect) {
+        return;
+    }
+
+    header += profile.has_gl_cbuf_ftou_bug ? "uvec4 " : "vec4 ";
+    header += "GetCbufIndirect(uint binding, uint offset){"
+              "switch(binding){"
+              "default:";
+
+    for (const auto& desc : info.constant_buffer_descriptors) {
+        header +=
+            fmt::format("case {}:return {}_cbuf{}[offset];", desc.index, stage_name, desc.index);
+    }
+
+    header += "}}";
+}
+
 void EmitContext::DefineStorageBuffers(Bindings& bindings) {
     if (info.storage_buffers_descriptors.empty()) {
         return;

src/shader_recompiler/backend/glsl/glsl_emit_context.h

@@ -162,6 +162,7 @@ public:
 private:
     void SetupExtensions();
     void DefineConstantBuffers(Bindings& bindings);
+    void DefineConstantBufferIndirect();
     void DefineStorageBuffers(Bindings& bindings);
     void DefineGenericOutput(size_t index, u32 invocations);
     void DefineHelperFunctions();

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -1043,15 +1043,15 @@ void EmitContext::DefineConstantBufferIndirectFunctions(const Info& info) {
         const Id merge_label{OpLabel()};
         const Id uniform_type{uniform_types.*member_ptr};
 
-        std::array<Id, Info::MAX_CBUFS> buf_labels;
-        std::array<Sirit::Literal, Info::MAX_CBUFS> buf_literals;
-        for (u32 i = 0; i < Info::MAX_CBUFS; i++) {
+        std::array<Id, Info::MAX_INDIRECT_CBUFS> buf_labels;
+        std::array<Sirit::Literal, Info::MAX_INDIRECT_CBUFS> buf_literals;
+        for (u32 i = 0; i < Info::MAX_INDIRECT_CBUFS; i++) {
             buf_labels[i] = OpLabel();
             buf_literals[i] = Sirit::Literal{i};
         }
         OpSelectionMerge(merge_label, spv::SelectionControlMask::MaskNone);
         OpSwitch(binding, buf_labels[0], buf_literals, buf_labels);
-        for (u32 i = 0; i < Info::MAX_CBUFS; i++) {
+        for (u32 i = 0; i < Info::MAX_INDIRECT_CBUFS; i++) {
             AddLabel(buf_labels[i]);
             const Id cbuf{cbufs[i].*member_ptr};
             const Id access_chain{OpAccessChain(uniform_type, cbuf, u32_zero_value, offset)};
@@ -1064,22 +1064,23 @@ void EmitContext::DefineConstantBufferIndirectFunctions(const Info& info) {
         return func;
     }};
     IR::Type types{info.used_indirect_cbuf_types};
-    if (True(types & IR::Type::U8)) {
+    bool supports_aliasing = profile.support_descriptor_aliasing;
+    if (supports_aliasing && True(types & IR::Type::U8)) {
         load_const_func_u8 = make_accessor(U8, &UniformDefinitions::U8);
     }
-    if (True(types & IR::Type::U16)) {
+    if (supports_aliasing && True(types & IR::Type::U16)) {
         load_const_func_u16 = make_accessor(U16, &UniformDefinitions::U16);
     }
-    if (True(types & IR::Type::F32)) {
+    if (supports_aliasing && True(types & IR::Type::F32)) {
         load_const_func_f32 = make_accessor(F32[1], &UniformDefinitions::F32);
     }
-    if (True(types & IR::Type::U32)) {
+    if (supports_aliasing && True(types & IR::Type::U32)) {
         load_const_func_u32 = make_accessor(U32[1], &UniformDefinitions::U32);
     }
-    if (True(types & IR::Type::U32x2)) {
+    if (supports_aliasing && True(types & IR::Type::U32x2)) {
         load_const_func_u32x2 = make_accessor(U32[2], &UniformDefinitions::U32x2);
     }
-    if (True(types & IR::Type::U32x4)) {
+    if (!supports_aliasing || True(types & IR::Type::U32x4)) {
         load_const_func_u32x4 = make_accessor(U32[4], &UniformDefinitions::U32x4);
     }
 }

src/shader_recompiler/frontend/maxwell/translate/impl/texture_load.cpp

@@ -132,7 +132,7 @@ void Impl(TranslatorVisitor& v, u64 insn, bool is_bindless) {
         multisample = v.X(meta_reg++);
     }
     if (tld.clamp != 0) {
-        throw NotImplementedException("TLD.CL - CLAMP is not implmented");
+        throw NotImplementedException("TLD.CL - CLAMP is not implemented");
     }
     IR::TextureInstInfo info{};
     info.type.Assign(GetType(tld.type));