video_core: Implement maxwell inline_index method

general: compress png images

.ci/scripts/transifex/docker.sh

@@ -3,15 +3,6 @@
 # SPDX-FileCopyrightText: 2021 yuzu Emulator Project
 # SPDX-License-Identifier: GPL-2.0-or-later
 
-# Setup RC file for tx
-cat << EOF > ~/.transifexrc
-[https://www.transifex.com]
-hostname = https://www.transifex.com
-username = api
-password = $TRANSIFEX_API_TOKEN
-EOF
-
-
 set -x
 
 echo -e "\e[1m\e[33mBuild tools information:\e[0m"
@@ -19,9 +10,6 @@ cmake --version
 gcc -v
 tx --version
 
-# vcpkg needs these: curl zip unzip tar, have tar
-apt-get install -y curl zip unzip
-
 mkdir build && cd build
 cmake .. -DENABLE_QT_TRANSLATION=ON -DGENERATE_QT_TRANSLATION=ON -DCMAKE_BUILD_TYPE=Release -DENABLE_SDL2=OFF -DYUZU_TESTS=OFF -DYUZU_USE_BUNDLED_VCPKG=ON
 make translation

.github/workflows/ci.yml

@@ -19,11 +19,11 @@ jobs:
       - uses: actions/checkout@v3
         with:
           submodules: recursive
-          fetch-depth: 0   
+          fetch-depth: 0
       - name: Update Translation
         run: ./.ci/scripts/transifex/docker.sh
         env:
-          TRANSIFEX_API_TOKEN: ${{ secrets.TRANSIFEX_API_TOKEN }}
+          TX_TOKEN: ${{ secrets.TRANSIFEX_API_TOKEN }}
 
   reuse:
     runs-on: ubuntu-latest

dist/languages/.tx/config

@@ -1,7 +1,7 @@
 [main]
 host = https://www.transifex.com
 
-[yuzu.emulator]
+[o:yuzu-emulator:p:yuzu:r:emulator]
 file_filter = <lang>.ts
 source_file = en.ts
 source_lang = en

dist/languages/README.md

@@ -1 +1,3 @@
-This directory stores translation patches (TS files) for yuzu Qt frontend. This directory is linked with [yuzu project on transifex](https://www.transifex.com/yuzu-emulator/yuzu), so you can update the translation by executing `tx pull -a`. If you want to contribute to the translation, please go the transifex link and submit your translation there. This directory on the main repo will be synchronized with transifex periodically. Do not directly open PRs on github to modify the translation.
+This directory stores translation patches (TS files) for yuzu Qt frontend. This directory is linked with [yuzu project on transifex](https://www.transifex.com/yuzu-emulator/yuzu), so you can update the translation by executing `tx pull -t -a`. If you want to contribute to the translation, please go the transifex link and submit your translation there. This directory on the main repo will be synchronized with transifex periodically.
+
+Do not directly open PRs on github to modify the translation.

src/audio_core/renderer/system.cpp

@@ -98,9 +98,8 @@ System::System(Core::System& core_, Kernel::KEvent* adsp_rendered_event_)
     : core{core_}, adsp{core.AudioCore().GetADSP()}, adsp_rendered_event{adsp_rendered_event_} {}
 
 Result System::Initialize(const AudioRendererParameterInternal& params,
-                          Kernel::KTransferMemory* transfer_memory, const u64 transfer_memory_size,
-                          const u32 process_handle_, const u64 applet_resource_user_id_,
-                          const s32 session_id_) {
+                          Kernel::KTransferMemory* transfer_memory, u64 transfer_memory_size,
+                          u32 process_handle_, u64 applet_resource_user_id_, s32 session_id_) {
     if (!CheckValidRevision(params.revision)) {
         return Service::Audio::ERR_INVALID_REVISION;
     }
@@ -354,6 +353,8 @@ Result System::Initialize(const AudioRendererParameterInternal& params,
 
     render_time_limit_percent = 100;
     drop_voice = params.voice_drop_enabled && params.execution_mode == ExecutionMode::Auto;
+    drop_voice_param = 1.0f;
+    num_voices_dropped = 0;
 
     allocator.Align(0x40);
     command_workbuffer_size = allocator.GetRemainingSize();
@@ -547,7 +548,7 @@ u32 System::GetRenderingTimeLimit() const {
     return render_time_limit_percent;
 }
 
-void System::SetRenderingTimeLimit(const u32 limit) {
+void System::SetRenderingTimeLimit(u32 limit) {
     render_time_limit_percent = limit;
 }
 
@@ -635,7 +636,7 @@ void System::SendCommandToDsp() {
 }
 
 u64 System::GenerateCommand(std::span<u8> in_command_buffer,
-                            [[maybe_unused]] const u64 command_buffer_size_) {
+                            [[maybe_unused]] u64 command_buffer_size_) {
     PoolMapper::ClearUseState(memory_pool_workbuffer, memory_pool_count);
     const auto start_time{core.CoreTiming().GetClockTicks()};
 
@@ -693,7 +694,8 @@ u64 System::GenerateCommand(std::span<u8> in_command_buffer,
 
     voice_context.SortInfo();
 
-    const auto start_estimated_time{command_buffer.estimated_process_time};
+    const auto start_estimated_time{drop_voice_param *
+                                    static_cast<f32>(command_buffer.estimated_process_time)};
 
     command_generator.GenerateVoiceCommands();
     command_generator.GenerateSubMixCommands();
@@ -712,11 +714,16 @@ u64 System::GenerateCommand(std::span<u8> in_command_buffer,
             render_context.behavior->IsAudioRendererProcessingTimeLimit70PercentSupported();
             time_limit_percent = 70.0f;
         }
+
+        const auto end_estimated_time{drop_voice_param *
+                                      static_cast<f32>(command_buffer.estimated_process_time)};
+        const auto estimated_time{start_estimated_time - end_estimated_time};
+
         const auto time_limit{static_cast<u32>(
-            static_cast<f32>(start_estimated_time - command_buffer.estimated_process_time) +
-            (((time_limit_percent / 100.0f) * 2'880'000.0) *
-             (static_cast<f32>(render_time_limit_percent) / 100.0f)))};
-        num_voices_dropped = DropVoices(command_buffer, start_estimated_time, time_limit);
+            estimated_time + (((time_limit_percent / 100.0f) * 2'880'000.0) *
+                              (static_cast<f32>(render_time_limit_percent) / 100.0f)))};
+        num_voices_dropped =
+            DropVoices(command_buffer, static_cast<u32>(start_estimated_time), time_limit);
     }
 
     command_list_header->buffer_size = command_buffer.size;
@@ -737,24 +744,33 @@ u64 System::GenerateCommand(std::span<u8> in_command_buffer,
     return command_buffer.size;
 }
 
-u32 System::DropVoices(CommandBuffer& command_buffer, const u32 estimated_process_time,
-                       const u32 time_limit) {
+f32 System::GetVoiceDropParameter() const {
+    return drop_voice_param;
+}
+
+void System::SetVoiceDropParameter(f32 voice_drop_) {
+    drop_voice_param = voice_drop_;
+}
+
+u32 System::DropVoices(CommandBuffer& command_buffer, u32 estimated_process_time, u32 time_limit) {
     u32 i{0};
     auto command_list{command_buffer.command_list.data() + sizeof(CommandListHeader)};
-    ICommand* cmd{};
+    ICommand* cmd{nullptr};
 
-    for (; i < command_buffer.count; i++) {
+    // Find a first valid voice to drop
+    while (i < command_buffer.count) {
         cmd = reinterpret_cast<ICommand*>(command_list);
-        if (cmd->type != CommandId::Performance &&
-            cmd->type != CommandId::DataSourcePcmInt16Version1 &&
-            cmd->type != CommandId::DataSourcePcmInt16Version2 &&
-            cmd->type != CommandId::DataSourcePcmFloatVersion1 &&
-            cmd->type != CommandId::DataSourcePcmFloatVersion2 &&
-            cmd->type != CommandId::DataSourceAdpcmVersion1 &&
-            cmd->type != CommandId::DataSourceAdpcmVersion2) {
+        if (cmd->type == CommandId::Performance ||
+            cmd->type == CommandId::DataSourcePcmInt16Version1 ||
+            cmd->type == CommandId::DataSourcePcmInt16Version2 ||
+            cmd->type == CommandId::DataSourcePcmFloatVersion1 ||
+            cmd->type == CommandId::DataSourcePcmFloatVersion2 ||
+            cmd->type == CommandId::DataSourceAdpcmVersion1 ||
+            cmd->type == CommandId::DataSourceAdpcmVersion2) {
             break;
         }
         command_list += cmd->size;
+        i++;
     }
 
     if (cmd == nullptr || command_buffer.count == 0 || i >= command_buffer.count) {
@@ -767,6 +783,7 @@ u32 System::DropVoices(CommandBuffer& command_buffer, const u32 estimated_proces
         const auto node_id_type{cmd->node_id >> 28};
         const auto node_id_base{cmd->node_id & 0xFFF};
 
+        // If the new estimated process time falls below the limit, we're done dropping.
         if (estimated_process_time <= time_limit) {
             break;
         }
@@ -775,6 +792,7 @@ u32 System::DropVoices(CommandBuffer& command_buffer, const u32 estimated_proces
             break;
         }
 
+        // Don't drop voices marked with the highest priority.
         auto& voice_info{voice_context.GetInfo(node_id_base)};
         if (voice_info.priority == HighestVoicePriority) {
             break;
@@ -783,18 +801,23 @@ u32 System::DropVoices(CommandBuffer& command_buffer, const u32 estimated_proces
         voices_dropped++;
         voice_info.voice_dropped = true;
 
-        if (i < command_buffer.count) {
-            while (cmd->node_id == node_id) {
-                if (cmd->type == CommandId::DepopPrepare) {
-                    cmd->enabled = true;
-                } else if (cmd->type == CommandId::Performance || !cmd->enabled) {
-                    cmd->enabled = false;
-                }
-                i++;
-                command_list += cmd->size;
-                cmd = reinterpret_cast<ICommand*>(command_list);
+        // First iteration should drop the voice, and then iterate through all of the commands tied
+        // to the voice. We don't need reverb on a voice which we've just removed, for example.
+        // Depops can't be removed otherwise we'll introduce audio popping, and we don't
+        // remove perf commands. Lower the estimated time for each command dropped.
+        while (i < command_buffer.count && cmd->node_id == node_id) {
+            if (cmd->type == CommandId::DepopPrepare) {
+                cmd->enabled = true;
+            } else if (cmd->enabled && cmd->type != CommandId::Performance) {
+                cmd->enabled = false;
+                estimated_process_time -= static_cast<u32>(
+                    drop_voice_param * static_cast<f32>(cmd->estimated_process_time));
             }
+            command_list += cmd->size;
+            cmd = reinterpret_cast<ICommand*>(command_list);
+            i++;
         }
+        i++;
     }
     return voices_dropped;
 }

src/audio_core/renderer/system.h

@@ -196,6 +196,20 @@ public:
      */
     u32 DropVoices(CommandBuffer& command_buffer, u32 estimated_process_time, u32 time_limit);
 
+    /**
+     * Get the current voice drop parameter.
+     *
+     * @return The current voice drop.
+     */
+    f32 GetVoiceDropParameter() const;
+
+    /**
+     * Set the voice drop parameter.
+     *
+     * @param The new voice drop.
+     */
+    void SetVoiceDropParameter(f32 voice_drop);
+
 private:
     /// Core system
     Core::System& core;
@@ -301,6 +315,8 @@ private:
     u32 num_voices_dropped{};
     /// Tick that rendering started
     u64 render_start_tick{};
+    /// Parameter to control the threshold for dropping voices if the audio graph gets too large
+    f32 drop_voice_param{1.0f};
 };
 
 } // namespace AudioRenderer

src/audio_core/renderer/voice/voice_context.cpp

@@ -74,8 +74,8 @@ void VoiceContext::SortInfo() {
     }
 
     std::ranges::sort(sorted_voice_info, [](const VoiceInfo* a, const VoiceInfo* b) {
-        return a->priority != b->priority ? a->priority < b->priority
-                                          : a->sort_order < b->sort_order;
+        return a->priority != b->priority ? a->priority > b->priority
+                                          : a->sort_order > b->sort_order;
     });
 }
 

src/audio_core/sink/sdl2_sink.cpp

@@ -230,7 +230,9 @@ std::vector<std::string> ListSDLSinkDevices(bool capture) {
 
     const int device_count = SDL_GetNumAudioDevices(capture);
     for (int i = 0; i < device_count; ++i) {
-        device_list.emplace_back(SDL_GetAudioDeviceName(i, 0));
+        if (const char* name = SDL_GetAudioDeviceName(i, capture)) {
+            device_list.emplace_back(name);
+        }
     }
 
     return device_list;

src/common/CMakeLists.txt

@@ -169,7 +169,11 @@ endif()
 create_target_directory_groups(common)
 
 target_link_libraries(common PUBLIC ${Boost_LIBRARIES} fmt::fmt microprofile Threads::Threads)
-target_link_libraries(common PRIVATE lz4::lz4)
+if (TARGET lz4::lz4)
+  target_link_libraries(common PRIVATE lz4::lz4)
+else()
+  target_link_libraries(common PRIVATE LZ4::lz4_shared)
+endif()
 if (TARGET zstd::zstd)
   target_link_libraries(common PRIVATE zstd::zstd)
 else()

src/common/concepts.h

@@ -34,4 +34,12 @@ concept DerivedFrom = requires {
 template <typename From, typename To>
 concept ConvertibleTo = std::is_convertible_v<From, To>;
 
+// No equivalents in the stdlib
+
+template <typename T>
+concept IsArithmetic = std::is_arithmetic_v<T>;
+
+template <typename T>
+concept IsIntegral = std::is_integral_v<T>;
+
 } // namespace Common

src/common/fixed_point.h

@@ -4,14 +4,7 @@
 // From: https://github.com/eteran/cpp-utilities/blob/master/fixed/include/cpp-utilities/fixed.h
 // See also: http://stackoverflow.com/questions/79677/whats-the-best-way-to-do-fixed-point-math
 
-#ifndef FIXED_H_
-#define FIXED_H_
-
-#if __cplusplus >= 201402L
-#define CONSTEXPR14 constexpr
-#else
-#define CONSTEXPR14
-#endif
+#pragma once
 
 #include <cstddef> // for size_t
 #include <cstdint>
@@ -19,6 +12,8 @@
 #include <ostream>
 #include <type_traits>
 
+#include <common/concepts.h>
+
 namespace Common {
 
 template <size_t I, size_t F>
@@ -57,8 +52,8 @@ struct type_from_size<64> {
     static constexpr size_t size = 64;
 
     using value_type = int64_t;
-    using unsigned_type = std::make_unsigned<value_type>::type;
-    using signed_type = std::make_signed<value_type>::type;
+    using unsigned_type = std::make_unsigned_t<value_type>;
+    using signed_type = std::make_signed_t<value_type>;
     using next_size = type_from_size<128>;
 };
 
@@ -68,8 +63,8 @@ struct type_from_size<32> {
     static constexpr size_t size = 32;
 
     using value_type = int32_t;
-    using unsigned_type = std::make_unsigned<value_type>::type;
-    using signed_type = std::make_signed<value_type>::type;
+    using unsigned_type = std::make_unsigned_t<value_type>;
+    using signed_type = std::make_signed_t<value_type>;
     using next_size = type_from_size<64>;
 };
 
@@ -79,8 +74,8 @@ struct type_from_size<16> {
     static constexpr size_t size = 16;
 
     using value_type = int16_t;
-    using unsigned_type = std::make_unsigned<value_type>::type;
-    using signed_type = std::make_signed<value_type>::type;
+    using unsigned_type = std::make_unsigned_t<value_type>;
+    using signed_type = std::make_signed_t<value_type>;
     using next_size = type_from_size<32>;
 };
 
@@ -90,8 +85,8 @@ struct type_from_size<8> {
     static constexpr size_t size = 8;
 
     using value_type = int8_t;
-    using unsigned_type = std::make_unsigned<value_type>::type;
-    using signed_type = std::make_signed<value_type>::type;
+    using unsigned_type = std::make_unsigned_t<value_type>;
+    using signed_type = std::make_signed_t<value_type>;
     using next_size = type_from_size<16>;
 };
 
@@ -106,9 +101,9 @@ constexpr B next_to_base(N rhs) {
 struct divide_by_zero : std::exception {};
 
 template <size_t I, size_t F>
-CONSTEXPR14 FixedPoint<I, F> divide(
+constexpr FixedPoint<I, F> divide(
     FixedPoint<I, F> numerator, FixedPoint<I, F> denominator, FixedPoint<I, F>& remainder,
-    typename std::enable_if<type_from_size<I + F>::next_size::is_specialized>::type* = nullptr) {
+    std::enable_if_t<type_from_size<I + F>::next_size::is_specialized>* = nullptr) {
 
     using next_type = typename FixedPoint<I, F>::next_type;
     using base_type = typename FixedPoint<I, F>::base_type;
@@ -126,9 +121,9 @@ CONSTEXPR14 FixedPoint<I, F> divide(
 }
 
 template <size_t I, size_t F>
-CONSTEXPR14 FixedPoint<I, F> divide(
+constexpr FixedPoint<I, F> divide(
     FixedPoint<I, F> numerator, FixedPoint<I, F> denominator, FixedPoint<I, F>& remainder,
-    typename std::enable_if<!type_from_size<I + F>::next_size::is_specialized>::type* = nullptr) {
+    std::enable_if_t<!type_from_size<I + F>::next_size::is_specialized>* = nullptr) {
 
     using unsigned_type = typename FixedPoint<I, F>::unsigned_type;
 
@@ -196,9 +191,9 @@ CONSTEXPR14 FixedPoint<I, F> divide(
 
 // this is the usual implementation of multiplication
 template <size_t I, size_t F>
-CONSTEXPR14 FixedPoint<I, F> multiply(
+constexpr FixedPoint<I, F> multiply(
     FixedPoint<I, F> lhs, FixedPoint<I, F> rhs,
-    typename std::enable_if<type_from_size<I + F>::next_size::is_specialized>::type* = nullptr) {
+    std::enable_if_t<type_from_size<I + F>::next_size::is_specialized>* = nullptr) {
 
     using next_type = typename FixedPoint<I, F>::next_type;
     using base_type = typename FixedPoint<I, F>::base_type;
@@ -215,9 +210,9 @@ CONSTEXPR14 FixedPoint<I, F> multiply(
 // it is slightly slower, but is more robust since it doesn't
 // require and upgraded type
 template <size_t I, size_t F>
-CONSTEXPR14 FixedPoint<I, F> multiply(
+constexpr FixedPoint<I, F> multiply(
     FixedPoint<I, F> lhs, FixedPoint<I, F> rhs,
-    typename std::enable_if<!type_from_size<I + F>::next_size::is_specialized>::type* = nullptr) {
+    std::enable_if_t<!type_from_size<I + F>::next_size::is_specialized>* = nullptr) {
 
     using base_type = typename FixedPoint<I, F>::base_type;
 
@@ -272,19 +267,20 @@ public:
     static constexpr base_type one = base_type(1) << fractional_bits;
 
 public: // constructors
-    FixedPoint() = default;
-    FixedPoint(const FixedPoint&) = default;
-    FixedPoint(FixedPoint&&) = default;
-    FixedPoint& operator=(const FixedPoint&) = default;
+    constexpr FixedPoint() = default;
 
-    template <class Number>
-    constexpr FixedPoint(
-        Number n, typename std::enable_if<std::is_arithmetic<Number>::value>::type* = nullptr)
-        : data_(static_cast<base_type>(n * one)) {}
+    constexpr FixedPoint(const FixedPoint&) = default;
+    constexpr FixedPoint& operator=(const FixedPoint&) = default;
+
+    constexpr FixedPoint(FixedPoint&&) noexcept = default;
+    constexpr FixedPoint& operator=(FixedPoint&&) noexcept = default;
+
+    template <IsArithmetic Number>
+    constexpr FixedPoint(Number n) : data_(static_cast<base_type>(n * one)) {}
 
 public: // conversion
     template <size_t I2, size_t F2>
-    CONSTEXPR14 explicit FixedPoint(FixedPoint<I2, F2> other) {
+    constexpr explicit FixedPoint(FixedPoint<I2, F2> other) {
         static_assert(I2 <= I && F2 <= F, "Scaling conversion can only upgrade types");
         using T = FixedPoint<I2, F2>;
 
@@ -308,36 +304,14 @@ public:
     }
 
 public: // comparison operators
-    constexpr bool operator==(FixedPoint rhs) const {
-        return data_ == rhs.data_;
-    }
-
-    constexpr bool operator!=(FixedPoint rhs) const {
-        return data_ != rhs.data_;
-    }
-
-    constexpr bool operator<(FixedPoint rhs) const {
-        return data_ < rhs.data_;
-    }
-
-    constexpr bool operator>(FixedPoint rhs) const {
-        return data_ > rhs.data_;
-    }
-
-    constexpr bool operator<=(FixedPoint rhs) const {
-        return data_ <= rhs.data_;
-    }
-
-    constexpr bool operator>=(FixedPoint rhs) const {
-        return data_ >= rhs.data_;
-    }
+    friend constexpr auto operator<=>(FixedPoint lhs, FixedPoint rhs) = default;
 
 public: // unary operators
-    constexpr bool operator!() const {
+    [[nodiscard]] constexpr bool operator!() const {
         return !data_;
     }
 
-    constexpr FixedPoint operator~() const {
+    [[nodiscard]] constexpr FixedPoint operator~() const {
         // NOTE(eteran): this will often appear to "just negate" the value
         // that is not an error, it is because -x == (~x+1)
         // and that "+1" is adding an infinitesimally small fraction to the
@@ -345,89 +319,87 @@ public: // unary operators
         return FixedPoint::from_base(~data_);
     }
 
-    constexpr FixedPoint operator-() const {
+    [[nodiscard]] constexpr FixedPoint operator-() const {
         return FixedPoint::from_base(-data_);
     }
 
-    constexpr FixedPoint operator+() const {
+    [[nodiscard]] constexpr FixedPoint operator+() const {
         return FixedPoint::from_base(+data_);
     }
 
-    CONSTEXPR14 FixedPoint& operator++() {
+    constexpr FixedPoint& operator++() {
         data_ += one;
         return *this;
     }
 
-    CONSTEXPR14 FixedPoint& operator--() {
+    constexpr FixedPoint& operator--() {
         data_ -= one;
         return *this;
     }
 
-    CONSTEXPR14 FixedPoint operator++(int) {
+    constexpr FixedPoint operator++(int) {
         FixedPoint tmp(*this);
         data_ += one;
         return tmp;
     }
 
-    CONSTEXPR14 FixedPoint operator--(int) {
+    constexpr FixedPoint operator--(int) {
         FixedPoint tmp(*this);
         data_ -= one;
         return tmp;
     }
 
 public: // basic math operators
-    CONSTEXPR14 FixedPoint& operator+=(FixedPoint n) {
+    constexpr FixedPoint& operator+=(FixedPoint n) {
         data_ += n.data_;
         return *this;
     }
 
-    CONSTEXPR14 FixedPoint& operator-=(FixedPoint n) {
+    constexpr FixedPoint& operator-=(FixedPoint n) {
         data_ -= n.data_;
         return *this;
     }
 
-    CONSTEXPR14 FixedPoint& operator*=(FixedPoint n) {
+    constexpr FixedPoint& operator*=(FixedPoint n) {
         return assign(detail::multiply(*this, n));
     }
 
-    CONSTEXPR14 FixedPoint& operator/=(FixedPoint n) {
+    constexpr FixedPoint& operator/=(FixedPoint n) {
         FixedPoint temp;
         return assign(detail::divide(*this, n, temp));
     }
 
 private:
-    CONSTEXPR14 FixedPoint& assign(FixedPoint rhs) {
+    constexpr FixedPoint& assign(FixedPoint rhs) {
         data_ = rhs.data_;
         return *this;
     }
 
 public: // binary math operators, effects underlying bit pattern since these
         // don't really typically make sense for non-integer values
-    CONSTEXPR14 FixedPoint& operator&=(FixedPoint n) {
+    constexpr FixedPoint& operator&=(FixedPoint n) {
         data_ &= n.data_;
         return *this;
     }
 
-    CONSTEXPR14 FixedPoint& operator|=(FixedPoint n) {
+    constexpr FixedPoint& operator|=(FixedPoint n) {
         data_ |= n.data_;
         return *this;
     }
 
-    CONSTEXPR14 FixedPoint& operator^=(FixedPoint n) {
+    constexpr FixedPoint& operator^=(FixedPoint n) {
         data_ ^= n.data_;
         return *this;
     }
 
-    template <class Integer,
-              class = typename std::enable_if<std::is_integral<Integer>::value>::type>
-    CONSTEXPR14 FixedPoint& operator>>=(Integer n) {
+    template <IsIntegral Integer>
+    constexpr FixedPoint& operator>>=(Integer n) {
         data_ >>= n;
         return *this;
     }
 
-    template <class Integer,
-              class = typename std::enable_if<std::is_integral<Integer>::value>::type>
-    CONSTEXPR14 FixedPoint& operator<<=(Integer n) {
+    template <IsIntegral Integer>
+    constexpr FixedPoint& operator<<=(Integer n) {
         data_ <<= n;
         return *this;
     }
@@ -437,42 +409,42 @@ public: // conversion to basic types
         data_ += (data_ & fractional_mask) >> 1;
     }
 
-    constexpr int to_int() {
+    [[nodiscard]] constexpr int to_int() {
         round_up();
         return static_cast<int>((data_ & integer_mask) >> fractional_bits);
     }
 
-    constexpr unsigned int to_uint() const {
+    [[nodiscard]] constexpr unsigned int to_uint() {
         round_up();
         return static_cast<unsigned int>((data_ & integer_mask) >> fractional_bits);
     }
 
-    constexpr int64_t to_long() {
+    [[nodiscard]] constexpr int64_t to_long() {
         round_up();
         return static_cast<int64_t>((data_ & integer_mask) >> fractional_bits);
     }
 
-    constexpr int to_int_floor() const {
+    [[nodiscard]] constexpr int to_int_floor() const {
         return static_cast<int>((data_ & integer_mask) >> fractional_bits);
     }
 
-    constexpr int64_t to_long_floor() {
+    [[nodiscard]] constexpr int64_t to_long_floor() const {
         return static_cast<int64_t>((data_ & integer_mask) >> fractional_bits);
     }
 
-    constexpr unsigned int to_uint_floor() const {
+    [[nodiscard]] constexpr unsigned int to_uint_floor() const {
         return static_cast<unsigned int>((data_ & integer_mask) >> fractional_bits);
     }
 
-    constexpr float to_float() const {
+    [[nodiscard]] constexpr float to_float() const {
         return static_cast<float>(data_) / FixedPoint::one;
     }
 
-    constexpr double to_double() const {
+    [[nodiscard]] constexpr double to_double() const {
         return static_cast<double>(data_) / FixedPoint::one;
     }
 
-    constexpr base_type to_raw() const {
+    [[nodiscard]] constexpr base_type to_raw() const {
         return data_;
     }
 
@@ -480,27 +452,27 @@ public: // conversion to basic types
         data_ &= fractional_mask;
     }
 
-    constexpr base_type get_frac() const {
+    [[nodiscard]] constexpr base_type get_frac() const {
         return data_ & fractional_mask;
     }
 
 public:
-    CONSTEXPR14 void swap(FixedPoint& rhs) {
+    constexpr void swap(FixedPoint& rhs) noexcept {
         using std::swap;
         swap(data_, rhs.data_);
     }
 
 public:
-    base_type data_;
+    base_type data_{};
 };
 
 // if we have the same fractional portion, but differing integer portions, we trivially upgrade the
 // smaller type
 template <size_t I1, size_t I2, size_t F>
-CONSTEXPR14 typename std::conditional<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>::type
-operator+(FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
+constexpr std::conditional_t<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>> operator+(
+    FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
 
-    using T = typename std::conditional<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>::type;
+    using T = std::conditional_t<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>;
 
     const T l = T::from_base(lhs.to_raw());
     const T r = T::from_base(rhs.to_raw());
@@ -508,10 +480,10 @@ operator+(FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
 }
 
 template <size_t I1, size_t I2, size_t F>
-CONSTEXPR14 typename std::conditional<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>::type
-operator-(FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
+constexpr std::conditional_t<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>> operator-(
+    FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
 
-    using T = typename std::conditional<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>::type;
+    using T = std::conditional_t<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>;
 
     const T l = T::from_base(lhs.to_raw());
     const T r = T::from_base(rhs.to_raw());
@@ -519,10 +491,10 @@ operator-(FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
 }
 
 template <size_t I1, size_t I2, size_t F>
-CONSTEXPR14 typename std::conditional<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>::type
-operator*(FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
+constexpr std::conditional_t<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>> operator*(
+    FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
 
-    using T = typename std::conditional<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>::type;
+    using T = std::conditional_t<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>;
 
     const T l = T::from_base(lhs.to_raw());
     const T r = T::from_base(rhs.to_raw());
@@ -530,10 +502,10 @@ operator*(FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
 }
 
 template <size_t I1, size_t I2, size_t F>
-CONSTEXPR14 typename std::conditional<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>::type
-operator/(FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
+constexpr std::conditional_t<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>> operator/(
+    FixedPoint<I1, F> lhs, FixedPoint<I2, F> rhs) {
 
-    using T = typename std::conditional<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>::type;
+    using T = std::conditional_t<I1 >= I2, FixedPoint<I1, F>, FixedPoint<I2, F>>;
 
     const T l = T::from_base(lhs.to_raw());
     const T r = T::from_base(rhs.to_raw());
@@ -548,159 +520,133 @@ std::ostream& operator<<(std::ostream& os, FixedPoint<I, F> f) {
 
 // basic math operators
 template <size_t I, size_t F>
-CONSTEXPR14 FixedPoint<I, F> operator+(FixedPoint<I, F> lhs, FixedPoint<I, F> rhs) {
+constexpr FixedPoint<I, F> operator+(FixedPoint<I, F> lhs, FixedPoint<I, F> rhs) {
     lhs += rhs;
     return lhs;
 }
 template <size_t I, size_t F>
-CONSTEXPR14 FixedPoint<I, F> operator-(FixedPoint<I, F> lhs, FixedPoint<I, F> rhs) {
+constexpr FixedPoint<I, F> operator-(FixedPoint<I, F> lhs, FixedPoint<I, F> rhs) {
     lhs -= rhs;
     return lhs;
 }
 template <size_t I, size_t F>
-CONSTEXPR14 FixedPoint<I, F> operator*(FixedPoint<I, F> lhs, FixedPoint<I, F> rhs) {
+constexpr FixedPoint<I, F> operator*(FixedPoint<I, F> lhs, FixedPoint<I, F> rhs) {
     lhs *= rhs;
     return lhs;
 }
 template <size_t I, size_t F>
-CONSTEXPR14 FixedPoint<I, F> operator/(FixedPoint<I, F> lhs, FixedPoint<I, F> rhs) {
+constexpr FixedPoint<I, F> operator/(FixedPoint<I, F> lhs, FixedPoint<I, F> rhs) {
     lhs /= rhs;
     return lhs;
 }
 
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator+(FixedPoint<I, F> lhs, Number rhs) {
+template <size_t I, size_t F, IsArithmetic Number>
+constexpr FixedPoint<I, F> operator+(FixedPoint<I, F> lhs, Number rhs) {
     lhs += FixedPoint<I, F>(rhs);
     return lhs;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator-(FixedPoint<I, F> lhs, Number rhs) {
+template <size_t I, size_t F, IsArithmetic Number>
+constexpr FixedPoint<I, F> operator-(FixedPoint<I, F> lhs, Number rhs) {
     lhs -= FixedPoint<I, F>(rhs);
     return lhs;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator*(FixedPoint<I, F> lhs, Number rhs) {
+template <size_t I, size_t F, IsArithmetic Number>
+constexpr FixedPoint<I, F> operator*(FixedPoint<I, F> lhs, Number rhs) {
     lhs *= FixedPoint<I, F>(rhs);
     return lhs;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator/(FixedPoint<I, F> lhs, Number rhs) {
+template <size_t I, size_t F, IsArithmetic Number>
+constexpr FixedPoint<I, F> operator/(FixedPoint<I, F> lhs, Number rhs) {
     lhs /= FixedPoint<I, F>(rhs);
     return lhs;
 }
 
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator+(Number lhs, FixedPoint<I, F> rhs) {
+template <size_t I, size_t F, IsArithmetic Number>
+constexpr FixedPoint<I, F> operator+(Number lhs, FixedPoint<I, F> rhs) {
     FixedPoint<I, F> tmp(lhs);
     tmp += rhs;
     return tmp;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator-(Number lhs, FixedPoint<I, F> rhs) {
+template <size_t I, size_t F, IsArithmetic Number>
+constexpr FixedPoint<I, F> operator-(Number lhs, FixedPoint<I, F> rhs) {
     FixedPoint<I, F> tmp(lhs);
     tmp -= rhs;
     return tmp;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator*(Number lhs, FixedPoint<I, F> rhs) {
+template <size_t I, size_t F, IsArithmetic Number>
+constexpr FixedPoint<I, F> operator*(Number lhs, FixedPoint<I, F> rhs) {
     FixedPoint<I, F> tmp(lhs);
     tmp *= rhs;
     return tmp;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator/(Number lhs, FixedPoint<I, F> rhs) {
+template <size_t I, size_t F, IsArithmetic Number>
+constexpr FixedPoint<I, F> operator/(Number lhs, FixedPoint<I, F> rhs) {
     FixedPoint<I, F> tmp(lhs);
     tmp /= rhs;
     return tmp;
 }
 
 // shift operators
-template <size_t I, size_t F, class Integer,
-          class = typename std::enable_if<std::is_integral<Integer>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator<<(FixedPoint<I, F> lhs, Integer rhs) {
+template <size_t I, size_t F, IsIntegral Integer>
+constexpr FixedPoint<I, F> operator<<(FixedPoint<I, F> lhs, Integer rhs) {
     lhs <<= rhs;
     return lhs;
 }
-template <size_t I, size_t F, class Integer,
-          class = typename std::enable_if<std::is_integral<Integer>::value>::type>
-CONSTEXPR14 FixedPoint<I, F> operator>>(FixedPoint<I, F> lhs, Integer rhs) {
+template <size_t I, size_t F, IsIntegral Integer>
+constexpr FixedPoint<I, F> operator>>(FixedPoint<I, F> lhs, Integer rhs) {
     lhs >>= rhs;
     return lhs;
 }
 
 // comparison operators
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator>(FixedPoint<I, F> lhs, Number rhs) {
     return lhs > FixedPoint<I, F>(rhs);
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator<(FixedPoint<I, F> lhs, Number rhs) {
     return lhs < FixedPoint<I, F>(rhs);
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator>=(FixedPoint<I, F> lhs, Number rhs) {
     return lhs >= FixedPoint<I, F>(rhs);
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator<=(FixedPoint<I, F> lhs, Number rhs) {
     return lhs <= FixedPoint<I, F>(rhs);
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator==(FixedPoint<I, F> lhs, Number rhs) {
     return lhs == FixedPoint<I, F>(rhs);
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator!=(FixedPoint<I, F> lhs, Number rhs) {
     return lhs != FixedPoint<I, F>(rhs);
 }
 
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator>(Number lhs, FixedPoint<I, F> rhs) {
     return FixedPoint<I, F>(lhs) > rhs;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator<(Number lhs, FixedPoint<I, F> rhs) {
     return FixedPoint<I, F>(lhs) < rhs;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator>=(Number lhs, FixedPoint<I, F> rhs) {
     return FixedPoint<I, F>(lhs) >= rhs;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator<=(Number lhs, FixedPoint<I, F> rhs) {
     return FixedPoint<I, F>(lhs) <= rhs;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator==(Number lhs, FixedPoint<I, F> rhs) {
     return FixedPoint<I, F>(lhs) == rhs;
 }
-template <size_t I, size_t F, class Number,
-          class = typename std::enable_if<std::is_arithmetic<Number>::value>::type>
+template <size_t I, size_t F, IsArithmetic Number>
 constexpr bool operator!=(Number lhs, FixedPoint<I, F> rhs) {
     return FixedPoint<I, F>(lhs) != rhs;
 }
 
 } // namespace Common
-
-#undef CONSTEXPR14
-
-#endif

src/core/CMakeLists.txt

@@ -190,6 +190,9 @@ add_library(core STATIC
     hle/kernel/k_code_memory.h
     hle/kernel/k_condition_variable.cpp
     hle/kernel/k_condition_variable.h
+    hle/kernel/k_dynamic_page_manager.h
+    hle/kernel/k_dynamic_resource_manager.h
+    hle/kernel/k_dynamic_slab_heap.h
     hle/kernel/k_event.cpp
     hle/kernel/k_event.h
     hle/kernel/k_handle_table.cpp
@@ -240,6 +243,8 @@ add_library(core STATIC
     hle/kernel/k_server_session.h
     hle/kernel/k_session.cpp
     hle/kernel/k_session.h
+    hle/kernel/k_session_request.cpp
+    hle/kernel/k_session_request.h
     hle/kernel/k_shared_memory.cpp
     hle/kernel/k_shared_memory.h
     hle/kernel/k_shared_memory_info.h

src/core/arm/arm_interface.cpp

@@ -134,6 +134,14 @@ void ARM_Interface::Run() {
         }
         system.ExitDynarmicProfile();
 
+        // If the thread is scheduled for termination, exit the thread.
+        if (current_thread->HasDpc()) {
+            if (current_thread->IsTerminationRequested()) {
+                current_thread->Exit();
+                UNREACHABLE();
+            }
+        }
+
         // Notify the debugger and go to sleep if a breakpoint was hit,
         // or if the thread is unable to continue for any reason.
         if (Has(hr, breakpoint) || Has(hr, no_execute)) {

src/core/core.cpp

@@ -133,6 +133,50 @@ struct System::Impl {
         : kernel{system}, fs_controller{system}, memory{system}, hid_core{}, room_network{},
           cpu_manager{system}, reporter{system}, applet_manager{system}, time_manager{system} {}
 
+    void Initialize(System& system) {
+        device_memory = std::make_unique<Core::DeviceMemory>();
+
+        is_multicore = Settings::values.use_multi_core.GetValue();
+
+        core_timing.SetMulticore(is_multicore);
+        core_timing.Initialize([&system]() { system.RegisterHostThread(); });
+
+        const auto posix_time = std::chrono::system_clock::now().time_since_epoch();
+        const auto current_time =
+            std::chrono::duration_cast<std::chrono::seconds>(posix_time).count();
+        Settings::values.custom_rtc_differential =
+            Settings::values.custom_rtc.value_or(current_time) - current_time;
+
+        // Create a default fs if one doesn't already exist.
+        if (virtual_filesystem == nullptr) {
+            virtual_filesystem = std::make_shared<FileSys::RealVfsFilesystem>();
+        }
+        if (content_provider == nullptr) {
+            content_provider = std::make_unique<FileSys::ContentProviderUnion>();
+        }
+
+        // Create default implementations of applets if one is not provided.
+        applet_manager.SetDefaultAppletsIfMissing();
+
+        is_async_gpu = Settings::values.use_asynchronous_gpu_emulation.GetValue();
+
+        kernel.SetMulticore(is_multicore);
+        cpu_manager.SetMulticore(is_multicore);
+        cpu_manager.SetAsyncGpu(is_async_gpu);
+    }
+
+    void ReinitializeIfNecessary(System& system) {
+        if (is_multicore == Settings::values.use_multi_core.GetValue()) {
+            return;
+        }
+
+        LOG_DEBUG(Kernel, "Re-initializing");
+
+        is_multicore = Settings::values.use_multi_core.GetValue();
+
+        Initialize(system);
+    }
+
     SystemResultStatus Run() {
         std::unique_lock<std::mutex> lk(suspend_guard);
         status = SystemResultStatus::Success;
@@ -178,37 +222,14 @@ struct System::Impl {
         debugger = std::make_unique<Debugger>(system, port);
     }
 
-    SystemResultStatus Init(System& system, Frontend::EmuWindow& emu_window) {
+    SystemResultStatus SetupForMainProcess(System& system, Frontend::EmuWindow& emu_window) {
         LOG_DEBUG(Core, "initialized OK");
 
-        device_memory = std::make_unique<Core::DeviceMemory>();
-
-        is_multicore = Settings::values.use_multi_core.GetValue();
-        is_async_gpu = Settings::values.use_asynchronous_gpu_emulation.GetValue();
-
-        kernel.SetMulticore(is_multicore);
-        cpu_manager.SetMulticore(is_multicore);
-        cpu_manager.SetAsyncGpu(is_async_gpu);
-        core_timing.SetMulticore(is_multicore);
+        // Setting changes may require a full system reinitialization (e.g., disabling multicore).
+        ReinitializeIfNecessary(system);
 
         kernel.Initialize();
         cpu_manager.Initialize();
-        core_timing.Initialize([&system]() { system.RegisterHostThread(); });
-
-        const auto posix_time = std::chrono::system_clock::now().time_since_epoch();
-        const auto current_time =
-            std::chrono::duration_cast<std::chrono::seconds>(posix_time).count();
-        Settings::values.custom_rtc_differential =
-            Settings::values.custom_rtc.value_or(current_time) - current_time;
-
-        // Create a default fs if one doesn't already exist.
-        if (virtual_filesystem == nullptr)
-            virtual_filesystem = std::make_shared<FileSys::RealVfsFilesystem>();
-        if (content_provider == nullptr)
-            content_provider = std::make_unique<FileSys::ContentProviderUnion>();
-
-        /// Create default implementations of applets if one is not provided.
-        applet_manager.SetDefaultAppletsIfMissing();
 
         /// Reset all glue registrations
         arp_manager.ResetAll();
@@ -253,11 +274,11 @@ struct System::Impl {
             return SystemResultStatus::ErrorGetLoader;
         }
 
-        SystemResultStatus init_result{Init(system, emu_window)};
+        SystemResultStatus init_result{SetupForMainProcess(system, emu_window)};
         if (init_result != SystemResultStatus::Success) {
             LOG_CRITICAL(Core, "Failed to initialize system (Error {})!",
                          static_cast<int>(init_result));
-            Shutdown();
+            ShutdownMainProcess();
             return init_result;
         }
 
@@ -276,7 +297,7 @@ struct System::Impl {
         const auto [load_result, load_parameters] = app_loader->Load(*main_process, system);
         if (load_result != Loader::ResultStatus::Success) {
             LOG_CRITICAL(Core, "Failed to load ROM (Error {})!", load_result);
-            Shutdown();
+            ShutdownMainProcess();
 
             return static_cast<SystemResultStatus>(
                 static_cast<u32>(SystemResultStatus::ErrorLoader) + static_cast<u32>(load_result));
@@ -335,7 +356,7 @@ struct System::Impl {
         return status;
     }
 
-    void Shutdown() {
+    void ShutdownMainProcess() {
         SetShuttingDown(true);
 
         // Log last frame performance stats if game was loded
@@ -369,7 +390,7 @@ struct System::Impl {
         cheat_engine.reset();
         telemetry_session.reset();
         time_manager.Shutdown();
-        core_timing.Shutdown();
+        core_timing.ClearPendingEvents();
         app_loader.reset();
         audio_core.reset();
         gpu_core.reset();
@@ -377,7 +398,6 @@ struct System::Impl {
         perf_stats.reset();
         kernel.Shutdown();
         memory.Reset();
-        applet_manager.ClearAll();
 
         if (auto room_member = room_network.GetRoomMember().lock()) {
             Network::GameInfo game_info{};
@@ -520,6 +540,10 @@ const CpuManager& System::GetCpuManager() const {
     return impl->cpu_manager;
 }
 
+void System::Initialize() {
+    impl->Initialize(*this);
+}
+
 SystemResultStatus System::Run() {
     return impl->Run();
 }
@@ -540,8 +564,8 @@ void System::InvalidateCpuInstructionCacheRange(VAddr addr, std::size_t size) {
     impl->kernel.InvalidateCpuInstructionCacheRange(addr, size);
 }
 
-void System::Shutdown() {
-    impl->Shutdown();
+void System::ShutdownMainProcess() {
+    impl->ShutdownMainProcess();
 }
 
 bool System::IsShuttingDown() const {

src/core/core.h

@@ -142,6 +142,12 @@ public:
     System(System&&) = delete;
     System& operator=(System&&) = delete;
 
+    /**
+     * Initializes the system
+     * This function will initialize core functionaility used for system emulation
+     */
+    void Initialize();
+
     /**
      * Run the OS and Application
      * This function will start emulation and run the relevant devices
@@ -166,8 +172,8 @@ public:
 
     void InvalidateCpuInstructionCacheRange(VAddr addr, std::size_t size);
 
-    /// Shutdown the emulated system.
-    void Shutdown();
+    /// Shutdown the main emulated process.
+    void ShutdownMainProcess();
 
     /// Check if the core is shutting down.
     [[nodiscard]] bool IsShuttingDown() const;

src/core/core_timing.cpp

@@ -40,7 +40,9 @@ struct CoreTiming::Event {
 CoreTiming::CoreTiming()
     : clock{Common::CreateBestMatchingClock(Hardware::BASE_CLOCK_RATE, Hardware::CNTFREQ)} {}
 
-CoreTiming::~CoreTiming() = default;
+CoreTiming::~CoreTiming() {
+    Reset();
+}
 
 void CoreTiming::ThreadEntry(CoreTiming& instance) {
     constexpr char name[] = "HostTiming";
@@ -53,6 +55,7 @@ void CoreTiming::ThreadEntry(CoreTiming& instance) {
 }
 
 void CoreTiming::Initialize(std::function<void()>&& on_thread_init_) {
+    Reset();
     on_thread_init = std::move(on_thread_init_);
     event_fifo_id = 0;
     shutting_down = false;
@@ -65,17 +68,8 @@ void CoreTiming::Initialize(std::function<void()>&& on_thread_init_) {
     }
 }
 
-void CoreTiming::Shutdown() {
-    paused = true;
-    shutting_down = true;
-    pause_event.Set();
-    event.Set();
-    if (timer_thread) {
-        timer_thread->join();
-    }
-    ClearPendingEvents();
-    timer_thread.reset();
-    has_started = false;
+void CoreTiming::ClearPendingEvents() {
+    event_queue.clear();
 }
 
 void CoreTiming::Pause(bool is_paused) {
@@ -196,10 +190,6 @@ u64 CoreTiming::GetClockTicks() const {
     return CpuCyclesToClockCycles(ticks);
 }
 
-void CoreTiming::ClearPendingEvents() {
-    event_queue.clear();
-}
-
 void CoreTiming::RemoveEvent(const std::shared_ptr<EventType>& event_type) {
     std::scoped_lock lock{basic_lock};
 
@@ -307,6 +297,18 @@ void CoreTiming::ThreadLoop() {
     }
 }
 
+void CoreTiming::Reset() {
+    paused = true;
+    shutting_down = true;
+    pause_event.Set();
+    event.Set();
+    if (timer_thread) {
+        timer_thread->join();
+    }
+    timer_thread.reset();
+    has_started = false;
+}
+
 std::chrono::nanoseconds CoreTiming::GetGlobalTimeNs() const {
     if (is_multicore) {
         return clock->GetTimeNS();

src/core/core_timing.h

@@ -61,19 +61,14 @@ public:
     /// required to end slice - 1 and start slice 0 before the first cycle of code is executed.
     void Initialize(std::function<void()>&& on_thread_init_);
 
-    /// Tears down all timing related functionality.
-    void Shutdown();
+    /// Clear all pending events. This should ONLY be done on exit.
+    void ClearPendingEvents();
 
     /// Sets if emulation is multicore or single core, must be set before Initialize
     void SetMulticore(bool is_multicore_) {
         is_multicore = is_multicore_;
     }
 
-    /// Check if it's using host timing.
-    bool IsHostTiming() const {
-        return is_multicore;
-    }
-
     /// Pauses/Unpauses the execution of the timer thread.
     void Pause(bool is_paused);
 
@@ -136,12 +131,11 @@ public:
 private:
     struct Event;
 
-    /// Clear all pending events. This should ONLY be done on exit.
-    void ClearPendingEvents();
-
     static void ThreadEntry(CoreTiming& instance);
     void ThreadLoop();
 
+    void Reset();
+
     std::unique_ptr<Common::WallClock> clock;
 
     s64 global_timer = 0;

src/core/device_memory.h

@@ -31,12 +31,14 @@ public:
                DramMemoryMap::Base;
     }
 
-    u8* GetPointer(PAddr addr) {
-        return buffer.BackingBasePointer() + (addr - DramMemoryMap::Base);
+    template <typename T>
+    T* GetPointer(PAddr addr) {
+        return reinterpret_cast<T*>(buffer.BackingBasePointer() + (addr - DramMemoryMap::Base));
     }
 
-    const u8* GetPointer(PAddr addr) const {
-        return buffer.BackingBasePointer() + (addr - DramMemoryMap::Base);
+    template <typename T>
+    const T* GetPointer(PAddr addr) const {
+        return reinterpret_cast<T*>(buffer.BackingBasePointer() + (addr - DramMemoryMap::Base));
     }
 
     Common::HostMemory buffer;

src/core/file_sys/savedata_factory.cpp

@@ -5,6 +5,7 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
+#include "common/uuid.h"
 #include "core/core.h"
 #include "core/file_sys/savedata_factory.h"
 #include "core/file_sys/vfs.h"
@@ -59,6 +60,36 @@ bool ShouldSaveDataBeAutomaticallyCreated(SaveDataSpaceId space, const SaveDataA
             attr.title_id == 0 && attr.save_id == 0);
 }
 
+std::string GetFutureSaveDataPath(SaveDataSpaceId space_id, SaveDataType type, u64 title_id,
+                                  u128 user_id) {
+    // Only detect nand user saves.
+    const auto space_id_path = [space_id]() -> std::string_view {
+        switch (space_id) {
+        case SaveDataSpaceId::NandUser:
+            return "/user/save";
+        default:
+            return "";
+        }
+    }();
+
+    if (space_id_path.empty()) {
+        return "";
+    }
+
+    Common::UUID uuid;
+    std::memcpy(uuid.uuid.data(), user_id.data(), sizeof(Common::UUID));
+
+    // Only detect account/device saves from the future location.
+    switch (type) {
+    case SaveDataType::SaveData:
+        return fmt::format("{}/account/{}/{:016X}/1", space_id_path, uuid.RawString(), title_id);
+    case SaveDataType::DeviceSaveData:
+        return fmt::format("{}/device/{:016X}/1", space_id_path, title_id);
+    default:
+        return "";
+    }
+}
+
 } // Anonymous namespace
 
 std::string SaveDataAttribute::DebugInfo() const {
@@ -82,7 +113,7 @@ ResultVal<VirtualDir> SaveDataFactory::Create(SaveDataSpaceId space,
     PrintSaveDataAttributeWarnings(meta);
 
     const auto save_directory =
-        GetFullPath(system, space, meta.type, meta.title_id, meta.user_id, meta.save_id);
+        GetFullPath(system, dir, space, meta.type, meta.title_id, meta.user_id, meta.save_id);
 
     auto out = dir->CreateDirectoryRelative(save_directory);
 
@@ -99,7 +130,7 @@ ResultVal<VirtualDir> SaveDataFactory::Open(SaveDataSpaceId space,
                                             const SaveDataAttribute& meta) const {
 
     const auto save_directory =
-        GetFullPath(system, space, meta.type, meta.title_id, meta.user_id, meta.save_id);
+        GetFullPath(system, dir, space, meta.type, meta.title_id, meta.user_id, meta.save_id);
 
     auto out = dir->GetDirectoryRelative(save_directory);
 
@@ -134,9 +165,9 @@ std::string SaveDataFactory::GetSaveDataSpaceIdPath(SaveDataSpaceId space) {
     }
 }
 
-std::string SaveDataFactory::GetFullPath(Core::System& system, SaveDataSpaceId space,
-                                         SaveDataType type, u64 title_id, u128 user_id,
-                                         u64 save_id) {
+std::string SaveDataFactory::GetFullPath(Core::System& system, VirtualDir dir,
+                                         SaveDataSpaceId space, SaveDataType type, u64 title_id,
+                                         u128 user_id, u64 save_id) {
     // According to switchbrew, if a save is of type SaveData and the title id field is 0, it should
     // be interpreted as the title id of the current process.
     if (type == SaveDataType::SaveData || type == SaveDataType::DeviceSaveData) {
@@ -145,6 +176,17 @@ std::string SaveDataFactory::GetFullPath(Core::System& system, SaveDataSpaceId s
         }
     }
 
+    // For compat with a future impl.
+    if (std::string future_path =
+            GetFutureSaveDataPath(space, type, title_id & ~(0xFFULL), user_id);
+        !future_path.empty()) {
+        // Check if this location exists, and prefer it over the old.
+        if (const auto future_dir = dir->GetDirectoryRelative(future_path); future_dir != nullptr) {
+            LOG_INFO(Service_FS, "Using save at new location: {}", future_path);
+            return future_path;
+        }
+    }
+
     std::string out = GetSaveDataSpaceIdPath(space);
 
     switch (type) {
@@ -167,7 +209,8 @@ std::string SaveDataFactory::GetFullPath(Core::System& system, SaveDataSpaceId s
 
 SaveDataSize SaveDataFactory::ReadSaveDataSize(SaveDataType type, u64 title_id,
                                                u128 user_id) const {
-    const auto path = GetFullPath(system, SaveDataSpaceId::NandUser, type, title_id, user_id, 0);
+    const auto path =
+        GetFullPath(system, dir, SaveDataSpaceId::NandUser, type, title_id, user_id, 0);
     const auto relative_dir = GetOrCreateDirectoryRelative(dir, path);
 
     const auto size_file = relative_dir->GetFile(SAVE_DATA_SIZE_FILENAME);
@@ -185,7 +228,8 @@ SaveDataSize SaveDataFactory::ReadSaveDataSize(SaveDataType type, u64 title_id,
 
 void SaveDataFactory::WriteSaveDataSize(SaveDataType type, u64 title_id, u128 user_id,
                                         SaveDataSize new_value) const {
-    const auto path = GetFullPath(system, SaveDataSpaceId::NandUser, type, title_id, user_id, 0);
+    const auto path =
+        GetFullPath(system, dir, SaveDataSpaceId::NandUser, type, title_id, user_id, 0);
     const auto relative_dir = GetOrCreateDirectoryRelative(dir, path);
 
     const auto size_file = relative_dir->CreateFile(SAVE_DATA_SIZE_FILENAME);

src/core/file_sys/savedata_factory.h

@@ -95,8 +95,8 @@ public:
     VirtualDir GetSaveDataSpaceDirectory(SaveDataSpaceId space) const;
 
     static std::string GetSaveDataSpaceIdPath(SaveDataSpaceId space);
-    static std::string GetFullPath(Core::System& system, SaveDataSpaceId space, SaveDataType type,
-                                   u64 title_id, u128 user_id, u64 save_id);
+    static std::string GetFullPath(Core::System& system, VirtualDir dir, SaveDataSpaceId space,
+                                   SaveDataType type, u64 title_id, u128 user_id, u64 save_id);
 
     SaveDataSize ReadSaveDataSize(SaveDataType type, u64 title_id, u128 user_id) const;
     void WriteSaveDataSize(SaveDataType type, u64 title_id, u128 user_id,

src/core/hid/irs_types.h

@@ -14,7 +14,7 @@ enum class CameraAmbientNoiseLevel : u32 {
     Low,
     Medium,
     High,
-    Unkown3, // This level can't be reached
+    Unknown3, // This level can't be reached
 };
 
 // This is nn::irsensor::CameraLightTarget
@@ -75,9 +75,9 @@ enum class IrCameraStatus : u32 {
 enum class IrCameraInternalStatus : u32 {
     Stopped,
     FirmwareUpdateNeeded,
-    Unkown2,
-    Unkown3,
-    Unkown4,
+    Unknown2,
+    Unknown3,
+    Unknown4,
     FirmwareVersionRequested,
     FirmwareVersionIsInvalid,
     Ready,
@@ -121,20 +121,20 @@ enum class IrSensorFunctionLevel : u8 {
 
 // This is nn::irsensor::MomentProcessorPreprocess
 enum class MomentProcessorPreprocess : u32 {
-    Unkown0,
-    Unkown1,
+    Unknown0,
+    Unknown1,
 };
 
 // This is nn::irsensor::PackedMomentProcessorPreprocess
 enum class PackedMomentProcessorPreprocess : u8 {
-    Unkown0,
-    Unkown1,
+    Unknown0,
+    Unknown1,
 };
 
 // This is nn::irsensor::PointingStatus
 enum class PointingStatus : u32 {
-    Unkown0,
-    Unkown1,
+    Unknown0,
+    Unknown1,
 };
 
 struct IrsRect {

src/core/hle/ipc_helpers.h

@@ -86,13 +86,13 @@ public:
         u32 num_domain_objects{};
         const bool always_move_handles{
             (static_cast<u32>(flags) & static_cast<u32>(Flags::AlwaysMoveHandles)) != 0};
-        if (!ctx.Session()->IsDomain() || always_move_handles) {
+        if (!ctx.Session()->GetSessionRequestManager()->IsDomain() || always_move_handles) {
             num_handles_to_move = num_objects_to_move;
         } else {
             num_domain_objects = num_objects_to_move;
         }
 
-        if (ctx.Session()->IsDomain()) {
+        if (ctx.Session()->GetSessionRequestManager()->IsDomain()) {
             raw_data_size +=
                 static_cast<u32>(sizeof(DomainMessageHeader) / sizeof(u32) + num_domain_objects);
             ctx.write_size += num_domain_objects;
@@ -125,7 +125,8 @@ public:
         if (!ctx.IsTipc()) {
             AlignWithPadding();
 
-            if (ctx.Session()->IsDomain() && ctx.HasDomainMessageHeader()) {
+            if (ctx.Session()->GetSessionRequestManager()->IsDomain() &&
+                ctx.HasDomainMessageHeader()) {
                 IPC::DomainMessageHeader domain_header{};
                 domain_header.num_objects = num_domain_objects;
                 PushRaw(domain_header);
@@ -145,7 +146,7 @@ public:
 
     template <class T>
     void PushIpcInterface(std::shared_ptr<T> iface) {
-        if (context->Session()->IsDomain()) {
+        if (context->Session()->GetSessionRequestManager()->IsDomain()) {
             context->AddDomainObject(std::move(iface));
         } else {
             kernel.CurrentProcess()->GetResourceLimit()->Reserve(
@@ -386,7 +387,7 @@ public:
 
     template <class T>
     std::weak_ptr<T> PopIpcInterface() {
-        ASSERT(context->Session()->IsDomain());
+        ASSERT(context->Session()->GetSessionRequestManager()->IsDomain());
         ASSERT(context->GetDomainMessageHeader().input_object_count > 0);
         return context->GetDomainHandler<T>(Pop<u32>() - 1);
     }

src/core/hle/kernel/hle_ipc.cpp

@@ -19,6 +19,7 @@
 #include "core/hle/kernel/k_server_session.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/service_thread.h"
 #include "core/memory.h"
 
 namespace Kernel {
@@ -56,16 +57,103 @@ bool SessionRequestManager::HasSessionRequestHandler(const HLERequestContext& co
     }
 }
 
+Result SessionRequestManager::CompleteSyncRequest(KServerSession* server_session,
+                                                  HLERequestContext& context) {
+    Result result = ResultSuccess;
+
+    // If the session has been converted to a domain, handle the domain request
+    if (this->HasSessionRequestHandler(context)) {
+        if (IsDomain() && context.HasDomainMessageHeader()) {
+            result = HandleDomainSyncRequest(server_session, context);
+            // If there is no domain header, the regular session handler is used
+        } else if (this->HasSessionHandler()) {
+            // If this manager has an associated HLE handler, forward the request to it.
+            result = this->SessionHandler().HandleSyncRequest(*server_session, context);
+        }
+    } else {
+        ASSERT_MSG(false, "Session handler is invalid, stubbing response!");
+        IPC::ResponseBuilder rb(context, 2);
+        rb.Push(ResultSuccess);
+    }
+
+    if (convert_to_domain) {
+        ASSERT_MSG(!IsDomain(), "ServerSession is already a domain instance.");
+        this->ConvertToDomain();
+        convert_to_domain = false;
+    }
+
+    return result;
+}
+
+Result SessionRequestManager::HandleDomainSyncRequest(KServerSession* server_session,
+                                                      HLERequestContext& context) {
+    if (!context.HasDomainMessageHeader()) {
+        return ResultSuccess;
+    }
+
+    // Set domain handlers in HLE context, used for domain objects (IPC interfaces) as inputs
+    context.SetSessionRequestManager(server_session->GetSessionRequestManager());
+
+    // If there is a DomainMessageHeader, then this is CommandType "Request"
+    const auto& domain_message_header = context.GetDomainMessageHeader();
+    const u32 object_id{domain_message_header.object_id};
+    switch (domain_message_header.command) {
+    case IPC::DomainMessageHeader::CommandType::SendMessage:
+        if (object_id > this->DomainHandlerCount()) {
+            LOG_CRITICAL(IPC,
+                         "object_id {} is too big! This probably means a recent service call "
+                         "needed to return a new interface!",
+                         object_id);
+            ASSERT(false);
+            return ResultSuccess; // Ignore error if asserts are off
+        }
+        if (auto strong_ptr = this->DomainHandler(object_id - 1).lock()) {
+            return strong_ptr->HandleSyncRequest(*server_session, context);
+        } else {
+            ASSERT(false);
+            return ResultSuccess;
+        }
+
+    case IPC::DomainMessageHeader::CommandType::CloseVirtualHandle: {
+        LOG_DEBUG(IPC, "CloseVirtualHandle, object_id=0x{:08X}", object_id);
+
+        this->CloseDomainHandler(object_id - 1);
+
+        IPC::ResponseBuilder rb{context, 2};
+        rb.Push(ResultSuccess);
+        return ResultSuccess;
+    }
+    }
+
+    LOG_CRITICAL(IPC, "Unknown domain command={}", domain_message_header.command.Value());
+    ASSERT(false);
+    return ResultSuccess;
+}
+
+Result SessionRequestManager::QueueSyncRequest(KSession* parent,
+                                               std::shared_ptr<HLERequestContext>&& context) {
+    // Ensure we have a session request handler
+    if (this->HasSessionRequestHandler(*context)) {
+        if (auto strong_ptr = this->GetServiceThread().lock()) {
+            strong_ptr->QueueSyncRequest(*parent, std::move(context));
+        } else {
+            ASSERT_MSG(false, "strong_ptr is nullptr!");
+        }
+    } else {
+        ASSERT_MSG(false, "handler is invalid!");
+    }
+
+    return ResultSuccess;
+}
+
 void SessionRequestHandler::ClientConnected(KServerSession* session) {
-    session->ClientConnected(shared_from_this());
+    session->GetSessionRequestManager()->SetSessionHandler(shared_from_this());
 
     // Ensure our server session is tracked globally.
     kernel.RegisterServerObject(session);
 }
 
-void SessionRequestHandler::ClientDisconnected(KServerSession* session) {
-    session->ClientDisconnected();
-}
+void SessionRequestHandler::ClientDisconnected(KServerSession* session) {}
 
 HLERequestContext::HLERequestContext(KernelCore& kernel_, Core::Memory::Memory& memory_,
                                      KServerSession* server_session_, KThread* thread_)
@@ -126,7 +214,7 @@ void HLERequestContext::ParseCommandBuffer(const KHandleTable& handle_table, u32
         // Padding to align to 16 bytes
         rp.AlignWithPadding();
 
-        if (Session()->IsDomain() &&
+        if (Session()->GetSessionRequestManager()->IsDomain() &&
             ((command_header->type == IPC::CommandType::Request ||
               command_header->type == IPC::CommandType::RequestWithContext) ||
              !incoming)) {
@@ -135,7 +223,7 @@ void HLERequestContext::ParseCommandBuffer(const KHandleTable& handle_table, u32
             if (incoming || domain_message_header) {
                 domain_message_header = rp.PopRaw<IPC::DomainMessageHeader>();
             } else {
-                if (Session()->IsDomain()) {
+                if (Session()->GetSessionRequestManager()->IsDomain()) {
                     LOG_WARNING(IPC, "Domain request has no DomainMessageHeader!");
                 }
             }
@@ -228,12 +316,12 @@ Result HLERequestContext::WriteToOutgoingCommandBuffer(KThread& requesting_threa
     // Write the domain objects to the command buffer, these go after the raw untranslated data.
     // TODO(Subv): This completely ignores C buffers.
 
-    if (Session()->IsDomain()) {
+    if (server_session->GetSessionRequestManager()->IsDomain()) {
         current_offset = domain_offset - static_cast<u32>(outgoing_domain_objects.size());
-        for (const auto& object : outgoing_domain_objects) {
-            server_session->AppendDomainHandler(object);
-            cmd_buf[current_offset++] =
-                static_cast<u32_le>(server_session->NumDomainRequestHandlers());
+        for (auto& object : outgoing_domain_objects) {
+            server_session->GetSessionRequestManager()->AppendDomainHandler(std::move(object));
+            cmd_buf[current_offset++] = static_cast<u32_le>(
+                server_session->GetSessionRequestManager()->DomainHandlerCount());
         }
     }
 

src/core/hle/kernel/hle_ipc.h

@@ -121,6 +121,10 @@ public:
         is_domain = true;
     }
 
+    void ConvertToDomainOnRequestEnd() {
+        convert_to_domain = true;
+    }
+
     std::size_t DomainHandlerCount() const {
         return domain_handlers.size();
     }
@@ -164,7 +168,12 @@ public:
 
     bool HasSessionRequestHandler(const HLERequestContext& context) const;
 
+    Result HandleDomainSyncRequest(KServerSession* server_session, HLERequestContext& context);
+    Result CompleteSyncRequest(KServerSession* server_session, HLERequestContext& context);
+    Result QueueSyncRequest(KSession* parent, std::shared_ptr<HLERequestContext>&& context);
+
 private:
+    bool convert_to_domain{};
     bool is_domain{};
     SessionRequestHandlerPtr session_handler;
     std::vector<SessionRequestHandlerPtr> domain_handlers;

src/core/hle/kernel/init/init_slab_setup.cpp

@@ -18,6 +18,7 @@
 #include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_resource_limit.h"
 #include "core/hle/kernel/k_session.h"
+#include "core/hle/kernel/k_session_request.h"
 #include "core/hle/kernel/k_shared_memory.h"
 #include "core/hle/kernel/k_shared_memory_info.h"
 #include "core/hle/kernel/k_system_control.h"
@@ -34,6 +35,7 @@ namespace Kernel::Init {
     HANDLER(KThread, (SLAB_COUNT(KThread)), ##__VA_ARGS__)                                         \
     HANDLER(KEvent, (SLAB_COUNT(KEvent)), ##__VA_ARGS__)                                           \
     HANDLER(KPort, (SLAB_COUNT(KPort)), ##__VA_ARGS__)                                             \
+    HANDLER(KSessionRequest, (SLAB_COUNT(KSession) * 2), ##__VA_ARGS__)                            \
     HANDLER(KSharedMemory, (SLAB_COUNT(KSharedMemory)), ##__VA_ARGS__)                             \
     HANDLER(KSharedMemoryInfo, (SLAB_COUNT(KSharedMemory) * 8), ##__VA_ARGS__)                     \
     HANDLER(KTransferMemory, (SLAB_COUNT(KTransferMemory)), ##__VA_ARGS__)                         \
@@ -94,8 +96,8 @@ VAddr InitializeSlabHeap(Core::System& system, KMemoryLayout& memory_layout, VAd
     // TODO(bunnei): Fix this once we support the kernel virtual memory layout.
 
     if (size > 0) {
-        void* backing_kernel_memory{
-            system.DeviceMemory().GetPointer(TranslateSlabAddrToPhysical(memory_layout, start))};
+        void* backing_kernel_memory{system.DeviceMemory().GetPointer<void>(
+            TranslateSlabAddrToPhysical(memory_layout, start))};
 
         const KMemoryRegion* region = memory_layout.FindVirtual(start + size - 1);
         ASSERT(region != nullptr);
@@ -181,7 +183,7 @@ void InitializeKPageBufferSlabHeap(Core::System& system) {
     ASSERT(slab_address != 0);
 
     // Initialize the slabheap.
-    KPageBuffer::InitializeSlabHeap(kernel, system.DeviceMemory().GetPointer(slab_address),
+    KPageBuffer::InitializeSlabHeap(kernel, system.DeviceMemory().GetPointer<void>(slab_address),
                                     slab_size);
 }
 

src/core/hle/kernel/k_client_session.cpp

@@ -1,6 +1,7 @@
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include "common/scope_exit.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/kernel/k_client_session.h"
 #include "core/hle/kernel/k_server_session.h"
@@ -10,6 +11,8 @@
 
 namespace Kernel {
 
+static constexpr u32 MessageBufferSize = 0x100;
+
 KClientSession::KClientSession(KernelCore& kernel_)
     : KAutoObjectWithSlabHeapAndContainer{kernel_} {}
 KClientSession::~KClientSession() = default;
@@ -22,8 +25,16 @@ void KClientSession::Destroy() {
 void KClientSession::OnServerClosed() {}
 
 Result KClientSession::SendSyncRequest() {
-    // Signal the server session that new data is available
-    return parent->GetServerSession().OnRequest();
+    // Create a session request.
+    KSessionRequest* request = KSessionRequest::Create(kernel);
+    R_UNLESS(request != nullptr, ResultOutOfResource);
+    SCOPE_EXIT({ request->Close(); });
+
+    // Initialize the request.
+    request->Initialize(nullptr, GetCurrentThread(kernel).GetTLSAddress(), MessageBufferSize);
+
+    // Send the request.
+    return parent->GetServerSession().OnRequest(request);
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_code_memory.cpp

@@ -34,7 +34,7 @@ Result KCodeMemory::Initialize(Core::DeviceMemory& device_memory, VAddr addr, si
 
     // Clear the memory.
     for (const auto& block : m_page_group.Nodes()) {
-        std::memset(device_memory.GetPointer(block.GetAddress()), 0xFF, block.GetSize());
+        std::memset(device_memory.GetPointer<void>(block.GetAddress()), 0xFF, block.GetSize());
     }
 
     // Set remaining tracking members.

src/core/hle/kernel/k_dynamic_page_manager.h

@@ -0,0 +1,136 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/alignment.h"
+#include "common/common_types.h"
+#include "core/hle/kernel/k_page_bitmap.h"
+#include "core/hle/kernel/k_spin_lock.h"
+#include "core/hle/kernel/memory_types.h"
+#include "core/hle/kernel/svc_results.h"
+
+namespace Kernel {
+
+class KDynamicPageManager {
+public:
+    class PageBuffer {
+    private:
+        u8 m_buffer[PageSize];
+    };
+    static_assert(sizeof(PageBuffer) == PageSize);
+
+public:
+    KDynamicPageManager() = default;
+
+    template <typename T>
+    T* GetPointer(VAddr addr) {
+        return reinterpret_cast<T*>(m_backing_memory.data() + (addr - m_address));
+    }
+
+    template <typename T>
+    const T* GetPointer(VAddr addr) const {
+        return reinterpret_cast<T*>(m_backing_memory.data() + (addr - m_address));
+    }
+
+    Result Initialize(VAddr addr, size_t sz) {
+        // We need to have positive size.
+        R_UNLESS(sz > 0, ResultOutOfMemory);
+        m_backing_memory.resize(sz);
+
+        // Calculate management overhead.
+        const size_t management_size =
+            KPageBitmap::CalculateManagementOverheadSize(sz / sizeof(PageBuffer));
+        const size_t allocatable_size = sz - management_size;
+
+        // Set tracking fields.
+        m_address = addr;
+        m_size = Common::AlignDown(allocatable_size, sizeof(PageBuffer));
+        m_count = allocatable_size / sizeof(PageBuffer);
+        R_UNLESS(m_count > 0, ResultOutOfMemory);
+
+        // Clear the management region.
+        u64* management_ptr = GetPointer<u64>(m_address + allocatable_size);
+        std::memset(management_ptr, 0, management_size);
+
+        // Initialize the bitmap.
+        m_page_bitmap.Initialize(management_ptr, m_count);
+
+        // Free the pages to the bitmap.
+        for (size_t i = 0; i < m_count; i++) {
+            // Ensure the freed page is all-zero.
+            std::memset(GetPointer<PageBuffer>(m_address) + i, 0, PageSize);
+
+            // Set the bit for the free page.
+            m_page_bitmap.SetBit(i);
+        }
+
+        R_SUCCEED();
+    }
+
+    VAddr GetAddress() const {
+        return m_address;
+    }
+    size_t GetSize() const {
+        return m_size;
+    }
+    size_t GetUsed() const {
+        return m_used;
+    }
+    size_t GetPeak() const {
+        return m_peak;
+    }
+    size_t GetCount() const {
+        return m_count;
+    }
+
+    PageBuffer* Allocate() {
+        // Take the lock.
+        // TODO(bunnei): We should disable interrupts here via KScopedInterruptDisable.
+        KScopedSpinLock lk(m_lock);
+
+        // Find a random free block.
+        s64 soffset = m_page_bitmap.FindFreeBlock(true);
+        if (soffset < 0) [[unlikely]] {
+            return nullptr;
+        }
+
+        const size_t offset = static_cast<size_t>(soffset);
+
+        // Update our tracking.
+        m_page_bitmap.ClearBit(offset);
+        m_peak = std::max(m_peak, (++m_used));
+
+        return GetPointer<PageBuffer>(m_address) + offset;
+    }
+
+    void Free(PageBuffer* pb) {
+        // Ensure all pages in the heap are zero.
+        std::memset(pb, 0, PageSize);
+
+        // Take the lock.
+        // TODO(bunnei): We should disable interrupts here via KScopedInterruptDisable.
+        KScopedSpinLock lk(m_lock);
+
+        // Set the bit for the free page.
+        size_t offset = (reinterpret_cast<uintptr_t>(pb) - m_address) / sizeof(PageBuffer);
+        m_page_bitmap.SetBit(offset);
+
+        // Decrement our used count.
+        --m_used;
+    }
+
+private:
+    KSpinLock m_lock;
+    KPageBitmap m_page_bitmap;
+    size_t m_used{};
+    size_t m_peak{};
+    size_t m_count{};
+    VAddr m_address{};
+    size_t m_size{};
+
+    // TODO(bunnei): Back by host memory until we emulate kernel virtual address space.
+    std::vector<u8> m_backing_memory;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_dynamic_resource_manager.h

@@ -0,0 +1,58 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/common_funcs.h"
+#include "core/hle/kernel/k_dynamic_slab_heap.h"
+#include "core/hle/kernel/k_memory_block.h"
+
+namespace Kernel {
+
+template <typename T, bool ClearNode = false>
+class KDynamicResourceManager {
+    YUZU_NON_COPYABLE(KDynamicResourceManager);
+    YUZU_NON_MOVEABLE(KDynamicResourceManager);
+
+public:
+    using DynamicSlabType = KDynamicSlabHeap<T, ClearNode>;
+
+public:
+    constexpr KDynamicResourceManager() = default;
+
+    constexpr size_t GetSize() const {
+        return m_slab_heap->GetSize();
+    }
+    constexpr size_t GetUsed() const {
+        return m_slab_heap->GetUsed();
+    }
+    constexpr size_t GetPeak() const {
+        return m_slab_heap->GetPeak();
+    }
+    constexpr size_t GetCount() const {
+        return m_slab_heap->GetCount();
+    }
+
+    void Initialize(KDynamicPageManager* page_allocator, DynamicSlabType* slab_heap) {
+        m_page_allocator = page_allocator;
+        m_slab_heap = slab_heap;
+    }
+
+    T* Allocate() const {
+        return m_slab_heap->Allocate(m_page_allocator);
+    }
+
+    void Free(T* t) const {
+        m_slab_heap->Free(t);
+    }
+
+private:
+    KDynamicPageManager* m_page_allocator{};
+    DynamicSlabType* m_slab_heap{};
+};
+
+class KMemoryBlockSlabManager : public KDynamicResourceManager<KMemoryBlock> {};
+
+using KMemoryBlockSlabHeap = typename KMemoryBlockSlabManager::DynamicSlabType;
+
+} // namespace Kernel

src/core/hle/kernel/k_dynamic_slab_heap.h

@@ -0,0 +1,122 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <atomic>
+
+#include "common/common_funcs.h"
+#include "core/hle/kernel/k_dynamic_page_manager.h"
+#include "core/hle/kernel/k_slab_heap.h"
+
+namespace Kernel {
+
+template <typename T, bool ClearNode = false>
+class KDynamicSlabHeap : protected impl::KSlabHeapImpl {
+    YUZU_NON_COPYABLE(KDynamicSlabHeap);
+    YUZU_NON_MOVEABLE(KDynamicSlabHeap);
+
+public:
+    constexpr KDynamicSlabHeap() = default;
+
+    constexpr VAddr GetAddress() const {
+        return m_address;
+    }
+    constexpr size_t GetSize() const {
+        return m_size;
+    }
+    constexpr size_t GetUsed() const {
+        return m_used.load();
+    }
+    constexpr size_t GetPeak() const {
+        return m_peak.load();
+    }
+    constexpr size_t GetCount() const {
+        return m_count.load();
+    }
+
+    constexpr bool IsInRange(VAddr addr) const {
+        return this->GetAddress() <= addr && addr <= this->GetAddress() + this->GetSize() - 1;
+    }
+
+    void Initialize(KDynamicPageManager* page_allocator, size_t num_objects) {
+        ASSERT(page_allocator != nullptr);
+
+        // Initialize members.
+        m_address = page_allocator->GetAddress();
+        m_size = page_allocator->GetSize();
+
+        // Initialize the base allocator.
+        KSlabHeapImpl::Initialize();
+
+        // Allocate until we have the correct number of objects.
+        while (m_count.load() < num_objects) {
+            auto* allocated = reinterpret_cast<T*>(page_allocator->Allocate());
+            ASSERT(allocated != nullptr);
+
+            for (size_t i = 0; i < sizeof(PageBuffer) / sizeof(T); i++) {
+                KSlabHeapImpl::Free(allocated + i);
+            }
+
+            m_count += sizeof(PageBuffer) / sizeof(T);
+        }
+    }
+
+    T* Allocate(KDynamicPageManager* page_allocator) {
+        T* allocated = static_cast<T*>(KSlabHeapImpl::Allocate());
+
+        // If we successfully allocated and we should clear the node, do so.
+        if constexpr (ClearNode) {
+            if (allocated != nullptr) [[likely]] {
+                reinterpret_cast<KSlabHeapImpl::Node*>(allocated)->next = nullptr;
+            }
+        }
+
+        // If we fail to allocate, try to get a new page from our next allocator.
+        if (allocated == nullptr) [[unlikely]] {
+            if (page_allocator != nullptr) {
+                allocated = reinterpret_cast<T*>(page_allocator->Allocate());
+                if (allocated != nullptr) {
+                    // If we succeeded in getting a page, free the rest to our slab.
+                    for (size_t i = 1; i < sizeof(PageBuffer) / sizeof(T); i++) {
+                        KSlabHeapImpl::Free(allocated + i);
+                    }
+                    m_count += sizeof(PageBuffer) / sizeof(T);
+                }
+            }
+        }
+
+        if (allocated != nullptr) [[likely]] {
+            // Construct the object.
+            std::construct_at(allocated);
+
+            // Update our tracking.
+            const size_t used = ++m_used;
+            size_t peak = m_peak.load();
+            while (peak < used) {
+                if (m_peak.compare_exchange_weak(peak, used, std::memory_order_relaxed)) {
+                    break;
+                }
+            }
+        }
+
+        return allocated;
+    }
+
+    void Free(T* t) {
+        KSlabHeapImpl::Free(t);
+        --m_used;
+    }
+
+private:
+    using PageBuffer = KDynamicPageManager::PageBuffer;
+
+private:
+    std::atomic<size_t> m_used{};
+    std::atomic<size_t> m_peak{};
+    std::atomic<size_t> m_count{};
+    VAddr m_address{};
+    size_t m_size{};
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_interrupt_manager.cpp

@@ -11,29 +11,34 @@
 namespace Kernel::KInterruptManager {
 
 void HandleInterrupt(KernelCore& kernel, s32 core_id) {
-    auto* process = kernel.CurrentProcess();
-    if (!process) {
-        return;
-    }
-
     // Acknowledge the interrupt.
     kernel.PhysicalCore(core_id).ClearInterrupt();
 
     auto& current_thread = GetCurrentThread(kernel);
 
-    // If the user disable count is set, we may need to pin the current thread.
-    if (current_thread.GetUserDisableCount() && !process->GetPinnedThread(core_id)) {
-        KScopedSchedulerLock sl{kernel};
+    if (auto* process = kernel.CurrentProcess(); process) {
+        // If the user disable count is set, we may need to pin the current thread.
+        if (current_thread.GetUserDisableCount() && !process->GetPinnedThread(core_id)) {
+            KScopedSchedulerLock sl{kernel};
 
-        // Pin the current thread.
-        process->PinCurrentThread(core_id);
+            // Pin the current thread.
+            process->PinCurrentThread(core_id);
 
-        // Set the interrupt flag for the thread.
-        GetCurrentThread(kernel).SetInterruptFlag();
+            // Set the interrupt flag for the thread.
+            GetCurrentThread(kernel).SetInterruptFlag();
+        }
     }
 
     // Request interrupt scheduling.
     kernel.CurrentScheduler()->RequestScheduleOnInterrupt();
 }
 
+void SendInterProcessorInterrupt(KernelCore& kernel, u64 core_mask) {
+    for (std::size_t core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; ++core_id) {
+        if (core_mask & (1ULL << core_id)) {
+            kernel.PhysicalCore(core_id).Interrupt();
+        }
+    }
+}
+
 } // namespace Kernel::KInterruptManager

src/core/hle/kernel/k_interrupt_manager.h

@@ -11,6 +11,8 @@ class KernelCore;
 
 namespace KInterruptManager {
 void HandleInterrupt(KernelCore& kernel, s32 core_id);
-}
+void SendInterProcessorInterrupt(KernelCore& kernel, u64 core_mask);
+
+} // namespace KInterruptManager
 
 } // namespace Kernel

src/core/hle/kernel/k_linked_list.h

@@ -16,6 +16,7 @@ class KLinkedListNode : public boost::intrusive::list_base_hook<>,
                         public KSlabAllocated<KLinkedListNode> {
 
 public:
+    explicit KLinkedListNode(KernelCore&) {}
     KLinkedListNode() = default;
 
     void Initialize(void* it) {

src/core/hle/kernel/k_memory_block.h

@@ -6,6 +6,7 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/common_types.h"
+#include "common/intrusive_red_black_tree.h"
 #include "core/hle/kernel/memory_types.h"
 #include "core/hle/kernel/svc_types.h"
 
@@ -168,9 +169,8 @@ constexpr KMemoryPermission ConvertToKMemoryPermission(Svc::MemoryPermission per
 
 enum class KMemoryAttribute : u8 {
     None = 0x00,
-    Mask = 0x7F,
-    All = Mask,
-    DontCareMask = 0x80,
+    All = 0xFF,
+    UserMask = All,
 
     Locked = static_cast<u8>(Svc::MemoryAttribute::Locked),
     IpcLocked = static_cast<u8>(Svc::MemoryAttribute::IpcLocked),
@@ -178,76 +178,112 @@ enum class KMemoryAttribute : u8 {
     Uncached = static_cast<u8>(Svc::MemoryAttribute::Uncached),
 
     SetMask = Uncached,
-
-    IpcAndDeviceMapped = IpcLocked | DeviceShared,
-    LockedAndIpcLocked = Locked | IpcLocked,
-    DeviceSharedAndUncached = DeviceShared | Uncached
 };
 DECLARE_ENUM_FLAG_OPERATORS(KMemoryAttribute);
 
-static_assert((static_cast<u8>(KMemoryAttribute::Mask) &
-               static_cast<u8>(KMemoryAttribute::DontCareMask)) == 0);
+enum class KMemoryBlockDisableMergeAttribute : u8 {
+    None = 0,
+    Normal = (1u << 0),
+    DeviceLeft = (1u << 1),
+    IpcLeft = (1u << 2),
+    Locked = (1u << 3),
+    DeviceRight = (1u << 4),
+
+    AllLeft = Normal | DeviceLeft | IpcLeft | Locked,
+    AllRight = DeviceRight,
+};
+DECLARE_ENUM_FLAG_OPERATORS(KMemoryBlockDisableMergeAttribute);
 
 struct KMemoryInfo {
-    VAddr addr{};
-    std::size_t size{};
-    KMemoryState state{};
-    KMemoryPermission perm{};
-    KMemoryAttribute attribute{};
-    KMemoryPermission original_perm{};
-    u16 ipc_lock_count{};
-    u16 device_use_count{};
+    uintptr_t m_address;
+    size_t m_size;
+    KMemoryState m_state;
+    u16 m_device_disable_merge_left_count;
+    u16 m_device_disable_merge_right_count;
+    u16 m_ipc_lock_count;
+    u16 m_device_use_count;
+    u16 m_ipc_disable_merge_count;
+    KMemoryPermission m_permission;
+    KMemoryAttribute m_attribute;
+    KMemoryPermission m_original_permission;
+    KMemoryBlockDisableMergeAttribute m_disable_merge_attribute;
 
     constexpr Svc::MemoryInfo GetSvcMemoryInfo() const {
         return {
-            addr,
-            size,
-            static_cast<Svc::MemoryState>(state & KMemoryState::Mask),
-            static_cast<Svc::MemoryAttribute>(attribute & KMemoryAttribute::Mask),
-            static_cast<Svc::MemoryPermission>(perm & KMemoryPermission::UserMask),
-            ipc_lock_count,
-            device_use_count,
+            .addr = m_address,
+            .size = m_size,
+            .state = static_cast<Svc::MemoryState>(m_state & KMemoryState::Mask),
+            .attr = static_cast<Svc::MemoryAttribute>(m_attribute & KMemoryAttribute::UserMask),
+            .perm = static_cast<Svc::MemoryPermission>(m_permission & KMemoryPermission::UserMask),
+            .ipc_refcount = m_ipc_lock_count,
+            .device_refcount = m_device_use_count,
+            .padding = {},
         };
     }
 
-    constexpr VAddr GetAddress() const {
-        return addr;
+    constexpr uintptr_t GetAddress() const {
+        return m_address;
     }
-    constexpr std::size_t GetSize() const {
-        return size;
+
+    constexpr size_t GetSize() const {
+        return m_size;
     }
-    constexpr std::size_t GetNumPages() const {
-        return GetSize() / PageSize;
+
+    constexpr size_t GetNumPages() const {
+        return this->GetSize() / PageSize;
     }
-    constexpr VAddr GetEndAddress() const {
-        return GetAddress() + GetSize();
+
+    constexpr uintptr_t GetEndAddress() const {
+        return this->GetAddress() + this->GetSize();
     }
-    constexpr VAddr GetLastAddress() const {
-        return GetEndAddress() - 1;
+
+    constexpr uintptr_t GetLastAddress() const {
+        return this->GetEndAddress() - 1;
     }
+
+    constexpr u16 GetIpcLockCount() const {
+        return m_ipc_lock_count;
+    }
+
+    constexpr u16 GetIpcDisableMergeCount() const {
+        return m_ipc_disable_merge_count;
+    }
+
     constexpr KMemoryState GetState() const {
-        return state;
-    }
-    constexpr KMemoryAttribute GetAttribute() const {
-        return attribute;
+        return m_state;
     }
+
     constexpr KMemoryPermission GetPermission() const {
-        return perm;
+        return m_permission;
+    }
+
+    constexpr KMemoryPermission GetOriginalPermission() const {
+        return m_original_permission;
+    }
+
+    constexpr KMemoryAttribute GetAttribute() const {
+        return m_attribute;
+    }
+
+    constexpr KMemoryBlockDisableMergeAttribute GetDisableMergeAttribute() const {
+        return m_disable_merge_attribute;
     }
 };
 
-class KMemoryBlock final {
-    friend class KMemoryBlockManager;
-
+class KMemoryBlock : public Common::IntrusiveRedBlackTreeBaseNode<KMemoryBlock> {
 private:
-    VAddr addr{};
-    std::size_t num_pages{};
-    KMemoryState state{KMemoryState::None};
-    u16 ipc_lock_count{};
-    u16 device_use_count{};
-    KMemoryPermission perm{KMemoryPermission::None};
-    KMemoryPermission original_perm{KMemoryPermission::None};
-    KMemoryAttribute attribute{KMemoryAttribute::None};
+    u16 m_device_disable_merge_left_count;
+    u16 m_device_disable_merge_right_count;
+    VAddr m_address;
+    size_t m_num_pages;
+    KMemoryState m_memory_state;
+    u16 m_ipc_lock_count;
+    u16 m_device_use_count;
+    u16 m_ipc_disable_merge_count;
+    KMemoryPermission m_permission;
+    KMemoryPermission m_original_permission;
+    KMemoryAttribute m_attribute;
+    KMemoryBlockDisableMergeAttribute m_disable_merge_attribute;
 
 public:
     static constexpr int Compare(const KMemoryBlock& lhs, const KMemoryBlock& rhs) {
@@ -261,113 +297,349 @@ public:
     }
 
 public:
-    constexpr KMemoryBlock() = default;
-    constexpr KMemoryBlock(VAddr addr_, std::size_t num_pages_, KMemoryState state_,
-                           KMemoryPermission perm_, KMemoryAttribute attribute_)
-        : addr{addr_}, num_pages(num_pages_), state{state_}, perm{perm_}, attribute{attribute_} {}
-
     constexpr VAddr GetAddress() const {
-        return addr;
+        return m_address;
     }
 
-    constexpr std::size_t GetNumPages() const {
-        return num_pages;
+    constexpr size_t GetNumPages() const {
+        return m_num_pages;
     }
 
-    constexpr std::size_t GetSize() const {
-        return GetNumPages() * PageSize;
+    constexpr size_t GetSize() const {
+        return this->GetNumPages() * PageSize;
     }
 
     constexpr VAddr GetEndAddress() const {
-        return GetAddress() + GetSize();
+        return this->GetAddress() + this->GetSize();
     }
 
     constexpr VAddr GetLastAddress() const {
-        return GetEndAddress() - 1;
+        return this->GetEndAddress() - 1;
+    }
+
+    constexpr u16 GetIpcLockCount() const {
+        return m_ipc_lock_count;
+    }
+
+    constexpr u16 GetIpcDisableMergeCount() const {
+        return m_ipc_disable_merge_count;
+    }
+
+    constexpr KMemoryPermission GetPermission() const {
+        return m_permission;
+    }
+
+    constexpr KMemoryPermission GetOriginalPermission() const {
+        return m_original_permission;
+    }
+
+    constexpr KMemoryAttribute GetAttribute() const {
+        return m_attribute;
     }
 
     constexpr KMemoryInfo GetMemoryInfo() const {
         return {
-            GetAddress(), GetSize(),     state,          perm,
-            attribute,    original_perm, ipc_lock_count, device_use_count,
+            .m_address = this->GetAddress(),
+            .m_size = this->GetSize(),
+            .m_state = m_memory_state,
+            .m_device_disable_merge_left_count = m_device_disable_merge_left_count,
+            .m_device_disable_merge_right_count = m_device_disable_merge_right_count,
+            .m_ipc_lock_count = m_ipc_lock_count,
+            .m_device_use_count = m_device_use_count,
+            .m_ipc_disable_merge_count = m_ipc_disable_merge_count,
+            .m_permission = m_permission,
+            .m_attribute = m_attribute,
+            .m_original_permission = m_original_permission,
+            .m_disable_merge_attribute = m_disable_merge_attribute,
         };
     }
 
-    void ShareToDevice(KMemoryPermission /*new_perm*/) {
-        ASSERT((attribute & KMemoryAttribute::DeviceShared) == KMemoryAttribute::DeviceShared ||
-               device_use_count == 0);
-        attribute |= KMemoryAttribute::DeviceShared;
-        const u16 new_use_count{++device_use_count};
-        ASSERT(new_use_count > 0);
+public:
+    explicit KMemoryBlock() = default;
+
+    constexpr KMemoryBlock(VAddr addr, size_t np, KMemoryState ms, KMemoryPermission p,
+                           KMemoryAttribute attr)
+        : Common::IntrusiveRedBlackTreeBaseNode<KMemoryBlock>(),
+          m_device_disable_merge_left_count(), m_device_disable_merge_right_count(),
+          m_address(addr), m_num_pages(np), m_memory_state(ms), m_ipc_lock_count(0),
+          m_device_use_count(0), m_ipc_disable_merge_count(), m_permission(p),
+          m_original_permission(KMemoryPermission::None), m_attribute(attr),
+          m_disable_merge_attribute() {}
+
+    constexpr void Initialize(VAddr addr, size_t np, KMemoryState ms, KMemoryPermission p,
+                              KMemoryAttribute attr) {
+        m_device_disable_merge_left_count = 0;
+        m_device_disable_merge_right_count = 0;
+        m_address = addr;
+        m_num_pages = np;
+        m_memory_state = ms;
+        m_ipc_lock_count = 0;
+        m_device_use_count = 0;
+        m_permission = p;
+        m_original_permission = KMemoryPermission::None;
+        m_attribute = attr;
+        m_disable_merge_attribute = KMemoryBlockDisableMergeAttribute::None;
     }
 
-    void UnshareToDevice(KMemoryPermission /*new_perm*/) {
-        ASSERT((attribute & KMemoryAttribute::DeviceShared) == KMemoryAttribute::DeviceShared);
-        const u16 prev_use_count{device_use_count--};
-        ASSERT(prev_use_count > 0);
-        if (prev_use_count == 1) {
-            attribute &= ~KMemoryAttribute::DeviceShared;
-        }
-    }
-
-private:
     constexpr bool HasProperties(KMemoryState s, KMemoryPermission p, KMemoryAttribute a) const {
-        constexpr KMemoryAttribute AttributeIgnoreMask{KMemoryAttribute::DontCareMask |
-                                                       KMemoryAttribute::IpcLocked |
-                                                       KMemoryAttribute::DeviceShared};
-        return state == s && perm == p &&
-               (attribute | AttributeIgnoreMask) == (a | AttributeIgnoreMask);
+        constexpr auto AttributeIgnoreMask =
+            KMemoryAttribute::IpcLocked | KMemoryAttribute::DeviceShared;
+        return m_memory_state == s && m_permission == p &&
+               (m_attribute | AttributeIgnoreMask) == (a | AttributeIgnoreMask);
     }
 
     constexpr bool HasSameProperties(const KMemoryBlock& rhs) const {
-        return state == rhs.state && perm == rhs.perm && original_perm == rhs.original_perm &&
-               attribute == rhs.attribute && ipc_lock_count == rhs.ipc_lock_count &&
-               device_use_count == rhs.device_use_count;
+        return m_memory_state == rhs.m_memory_state && m_permission == rhs.m_permission &&
+               m_original_permission == rhs.m_original_permission &&
+               m_attribute == rhs.m_attribute && m_ipc_lock_count == rhs.m_ipc_lock_count &&
+               m_device_use_count == rhs.m_device_use_count;
     }
 
-    constexpr bool Contains(VAddr start) const {
-        return GetAddress() <= start && start <= GetEndAddress();
+    constexpr bool CanMergeWith(const KMemoryBlock& rhs) const {
+        return this->HasSameProperties(rhs) &&
+               (m_disable_merge_attribute & KMemoryBlockDisableMergeAttribute::AllRight) ==
+                   KMemoryBlockDisableMergeAttribute::None &&
+               (rhs.m_disable_merge_attribute & KMemoryBlockDisableMergeAttribute::AllLeft) ==
+                   KMemoryBlockDisableMergeAttribute::None;
     }
 
-    constexpr void Add(std::size_t count) {
-        ASSERT(count > 0);
-        ASSERT(GetAddress() + count * PageSize - 1 < GetEndAddress() + count * PageSize - 1);
-
-        num_pages += count;
+    constexpr bool Contains(VAddr addr) const {
+        return this->GetAddress() <= addr && addr <= this->GetEndAddress();
     }
 
-    constexpr void Update(KMemoryState new_state, KMemoryPermission new_perm,
-                          KMemoryAttribute new_attribute) {
-        ASSERT(original_perm == KMemoryPermission::None);
-        ASSERT((attribute & KMemoryAttribute::IpcLocked) == KMemoryAttribute::None);
+    constexpr void Add(const KMemoryBlock& added_block) {
+        ASSERT(added_block.GetNumPages() > 0);
+        ASSERT(this->GetAddress() + added_block.GetSize() - 1 <
+               this->GetEndAddress() + added_block.GetSize() - 1);
 
-        state = new_state;
-        perm = new_perm;
-
-        attribute = static_cast<KMemoryAttribute>(
-            new_attribute |
-            (attribute & (KMemoryAttribute::IpcLocked | KMemoryAttribute::DeviceShared)));
+        m_num_pages += added_block.GetNumPages();
+        m_disable_merge_attribute = static_cast<KMemoryBlockDisableMergeAttribute>(
+            m_disable_merge_attribute | added_block.m_disable_merge_attribute);
+        m_device_disable_merge_right_count = added_block.m_device_disable_merge_right_count;
     }
 
-    constexpr KMemoryBlock Split(VAddr split_addr) {
-        ASSERT(GetAddress() < split_addr);
-        ASSERT(Contains(split_addr));
-        ASSERT(Common::IsAligned(split_addr, PageSize));
+    constexpr void Update(KMemoryState s, KMemoryPermission p, KMemoryAttribute a,
+                          bool set_disable_merge_attr, u8 set_mask, u8 clear_mask) {
+        ASSERT(m_original_permission == KMemoryPermission::None);
+        ASSERT((m_attribute & KMemoryAttribute::IpcLocked) == KMemoryAttribute::None);
 
-        KMemoryBlock block;
-        block.addr = addr;
-        block.num_pages = (split_addr - GetAddress()) / PageSize;
-        block.state = state;
-        block.ipc_lock_count = ipc_lock_count;
-        block.device_use_count = device_use_count;
-        block.perm = perm;
-        block.original_perm = original_perm;
-        block.attribute = attribute;
+        m_memory_state = s;
+        m_permission = p;
+        m_attribute = static_cast<KMemoryAttribute>(
+            a | (m_attribute & (KMemoryAttribute::IpcLocked | KMemoryAttribute::DeviceShared)));
 
-        addr = split_addr;
-        num_pages -= block.num_pages;
+        if (set_disable_merge_attr && set_mask != 0) {
+            m_disable_merge_attribute = m_disable_merge_attribute |
+                                        static_cast<KMemoryBlockDisableMergeAttribute>(set_mask);
+        }
+        if (clear_mask != 0) {
+            m_disable_merge_attribute = m_disable_merge_attribute &
+                                        static_cast<KMemoryBlockDisableMergeAttribute>(~clear_mask);
+        }
+    }
 
-        return block;
+    constexpr void Split(KMemoryBlock* block, VAddr addr) {
+        ASSERT(this->GetAddress() < addr);
+        ASSERT(this->Contains(addr));
+        ASSERT(Common::IsAligned(addr, PageSize));
+
+        block->m_address = m_address;
+        block->m_num_pages = (addr - this->GetAddress()) / PageSize;
+        block->m_memory_state = m_memory_state;
+        block->m_ipc_lock_count = m_ipc_lock_count;
+        block->m_device_use_count = m_device_use_count;
+        block->m_permission = m_permission;
+        block->m_original_permission = m_original_permission;
+        block->m_attribute = m_attribute;
+        block->m_disable_merge_attribute = static_cast<KMemoryBlockDisableMergeAttribute>(
+            m_disable_merge_attribute & KMemoryBlockDisableMergeAttribute::AllLeft);
+        block->m_ipc_disable_merge_count = m_ipc_disable_merge_count;
+        block->m_device_disable_merge_left_count = m_device_disable_merge_left_count;
+        block->m_device_disable_merge_right_count = 0;
+
+        m_address = addr;
+        m_num_pages -= block->m_num_pages;
+
+        m_ipc_disable_merge_count = 0;
+        m_device_disable_merge_left_count = 0;
+        m_disable_merge_attribute = static_cast<KMemoryBlockDisableMergeAttribute>(
+            m_disable_merge_attribute & KMemoryBlockDisableMergeAttribute::AllRight);
+    }
+
+    constexpr void UpdateDeviceDisableMergeStateForShareLeft(
+        [[maybe_unused]] KMemoryPermission new_perm, bool left, [[maybe_unused]] bool right) {
+        if (left) {
+            m_disable_merge_attribute = static_cast<KMemoryBlockDisableMergeAttribute>(
+                m_disable_merge_attribute | KMemoryBlockDisableMergeAttribute::DeviceLeft);
+            const u16 new_device_disable_merge_left_count = ++m_device_disable_merge_left_count;
+            ASSERT(new_device_disable_merge_left_count > 0);
+        }
+    }
+
+    constexpr void UpdateDeviceDisableMergeStateForShareRight(
+        [[maybe_unused]] KMemoryPermission new_perm, [[maybe_unused]] bool left, bool right) {
+        if (right) {
+            m_disable_merge_attribute = static_cast<KMemoryBlockDisableMergeAttribute>(
+                m_disable_merge_attribute | KMemoryBlockDisableMergeAttribute::DeviceRight);
+            const u16 new_device_disable_merge_right_count = ++m_device_disable_merge_right_count;
+            ASSERT(new_device_disable_merge_right_count > 0);
+        }
+    }
+
+    constexpr void UpdateDeviceDisableMergeStateForShare(KMemoryPermission new_perm, bool left,
+                                                         bool right) {
+        this->UpdateDeviceDisableMergeStateForShareLeft(new_perm, left, right);
+        this->UpdateDeviceDisableMergeStateForShareRight(new_perm, left, right);
+    }
+
+    constexpr void ShareToDevice([[maybe_unused]] KMemoryPermission new_perm, bool left,
+                                 bool right) {
+        // We must either be shared or have a zero lock count.
+        ASSERT((m_attribute & KMemoryAttribute::DeviceShared) == KMemoryAttribute::DeviceShared ||
+               m_device_use_count == 0);
+
+        // Share.
+        const u16 new_count = ++m_device_use_count;
+        ASSERT(new_count > 0);
+
+        m_attribute = static_cast<KMemoryAttribute>(m_attribute | KMemoryAttribute::DeviceShared);
+
+        this->UpdateDeviceDisableMergeStateForShare(new_perm, left, right);
+    }
+
+    constexpr void UpdateDeviceDisableMergeStateForUnshareLeft(
+        [[maybe_unused]] KMemoryPermission new_perm, bool left, [[maybe_unused]] bool right) {
+
+        if (left) {
+            if (!m_device_disable_merge_left_count) {
+                return;
+            }
+            --m_device_disable_merge_left_count;
+        }
+
+        m_device_disable_merge_left_count =
+            std::min(m_device_disable_merge_left_count, m_device_use_count);
+
+        if (m_device_disable_merge_left_count == 0) {
+            m_disable_merge_attribute = static_cast<KMemoryBlockDisableMergeAttribute>(
+                m_disable_merge_attribute & ~KMemoryBlockDisableMergeAttribute::DeviceLeft);
+        }
+    }
+
+    constexpr void UpdateDeviceDisableMergeStateForUnshareRight(
+        [[maybe_unused]] KMemoryPermission new_perm, [[maybe_unused]] bool left, bool right) {
+        if (right) {
+            const u16 old_device_disable_merge_right_count = m_device_disable_merge_right_count--;
+            ASSERT(old_device_disable_merge_right_count > 0);
+            if (old_device_disable_merge_right_count == 1) {
+                m_disable_merge_attribute = static_cast<KMemoryBlockDisableMergeAttribute>(
+                    m_disable_merge_attribute & ~KMemoryBlockDisableMergeAttribute::DeviceRight);
+            }
+        }
+    }
+
+    constexpr void UpdateDeviceDisableMergeStateForUnshare(KMemoryPermission new_perm, bool left,
+                                                           bool right) {
+        this->UpdateDeviceDisableMergeStateForUnshareLeft(new_perm, left, right);
+        this->UpdateDeviceDisableMergeStateForUnshareRight(new_perm, left, right);
+    }
+
+    constexpr void UnshareToDevice([[maybe_unused]] KMemoryPermission new_perm, bool left,
+                                   bool right) {
+        // We must be shared.
+        ASSERT((m_attribute & KMemoryAttribute::DeviceShared) == KMemoryAttribute::DeviceShared);
+
+        // Unhare.
+        const u16 old_count = m_device_use_count--;
+        ASSERT(old_count > 0);
+
+        if (old_count == 1) {
+            m_attribute =
+                static_cast<KMemoryAttribute>(m_attribute & ~KMemoryAttribute::DeviceShared);
+        }
+
+        this->UpdateDeviceDisableMergeStateForUnshare(new_perm, left, right);
+    }
+
+    constexpr void UnshareToDeviceRight([[maybe_unused]] KMemoryPermission new_perm, bool left,
+                                        bool right) {
+
+        // We must be shared.
+        ASSERT((m_attribute & KMemoryAttribute::DeviceShared) == KMemoryAttribute::DeviceShared);
+
+        // Unhare.
+        const u16 old_count = m_device_use_count--;
+        ASSERT(old_count > 0);
+
+        if (old_count == 1) {
+            m_attribute =
+                static_cast<KMemoryAttribute>(m_attribute & ~KMemoryAttribute::DeviceShared);
+        }
+
+        this->UpdateDeviceDisableMergeStateForUnshareRight(new_perm, left, right);
+    }
+
+    constexpr void LockForIpc(KMemoryPermission new_perm, bool left, [[maybe_unused]] bool right) {
+        // We must either be locked or have a zero lock count.
+        ASSERT((m_attribute & KMemoryAttribute::IpcLocked) == KMemoryAttribute::IpcLocked ||
+               m_ipc_lock_count == 0);
+
+        // Lock.
+        const u16 new_lock_count = ++m_ipc_lock_count;
+        ASSERT(new_lock_count > 0);
+
+        // If this is our first lock, update our permissions.
+        if (new_lock_count == 1) {
+            ASSERT(m_original_permission == KMemoryPermission::None);
+            ASSERT((m_permission | new_perm | KMemoryPermission::NotMapped) ==
+                   (m_permission | KMemoryPermission::NotMapped));
+            ASSERT((m_permission & KMemoryPermission::UserExecute) !=
+                       KMemoryPermission::UserExecute ||
+                   (new_perm == KMemoryPermission::UserRead));
+            m_original_permission = m_permission;
+            m_permission = static_cast<KMemoryPermission>(
+                (new_perm & KMemoryPermission::IpcLockChangeMask) |
+                (m_original_permission & ~KMemoryPermission::IpcLockChangeMask));
+        }
+        m_attribute = static_cast<KMemoryAttribute>(m_attribute | KMemoryAttribute::IpcLocked);
+
+        if (left) {
+            m_disable_merge_attribute = static_cast<KMemoryBlockDisableMergeAttribute>(
+                m_disable_merge_attribute | KMemoryBlockDisableMergeAttribute::IpcLeft);
+            const u16 new_ipc_disable_merge_count = ++m_ipc_disable_merge_count;
+            ASSERT(new_ipc_disable_merge_count > 0);
+        }
+    }
+
+    constexpr void UnlockForIpc([[maybe_unused]] KMemoryPermission new_perm, bool left,
+                                [[maybe_unused]] bool right) {
+        // We must be locked.
+        ASSERT((m_attribute & KMemoryAttribute::IpcLocked) == KMemoryAttribute::IpcLocked);
+
+        // Unlock.
+        const u16 old_lock_count = m_ipc_lock_count--;
+        ASSERT(old_lock_count > 0);
+
+        // If this is our last unlock, update our permissions.
+        if (old_lock_count == 1) {
+            ASSERT(m_original_permission != KMemoryPermission::None);
+            m_permission = m_original_permission;
+            m_original_permission = KMemoryPermission::None;
+            m_attribute = static_cast<KMemoryAttribute>(m_attribute & ~KMemoryAttribute::IpcLocked);
+        }
+
+        if (left) {
+            const u16 old_ipc_disable_merge_count = m_ipc_disable_merge_count--;
+            ASSERT(old_ipc_disable_merge_count > 0);
+            if (old_ipc_disable_merge_count == 1) {
+                m_disable_merge_attribute = static_cast<KMemoryBlockDisableMergeAttribute>(
+                    m_disable_merge_attribute & ~KMemoryBlockDisableMergeAttribute::IpcLeft);
+            }
+        }
+    }
+
+    constexpr KMemoryBlockDisableMergeAttribute GetDisableMergeAttribute() const {
+        return m_disable_merge_attribute;
     }
 };
 static_assert(std::is_trivially_destructible<KMemoryBlock>::value);

src/core/hle/kernel/k_memory_block_manager.cpp

@@ -2,221 +2,336 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include "core/hle/kernel/k_memory_block_manager.h"
-#include "core/hle/kernel/memory_types.h"
 
 namespace Kernel {
 
-KMemoryBlockManager::KMemoryBlockManager(VAddr start_addr_, VAddr end_addr_)
-    : start_addr{start_addr_}, end_addr{end_addr_} {
-    const u64 num_pages{(end_addr - start_addr) / PageSize};
-    memory_block_tree.emplace_back(start_addr, num_pages, KMemoryState::Free,
-                                   KMemoryPermission::None, KMemoryAttribute::None);
+KMemoryBlockManager::KMemoryBlockManager() = default;
+
+Result KMemoryBlockManager::Initialize(VAddr st, VAddr nd, KMemoryBlockSlabManager* slab_manager) {
+    // Allocate a block to encapsulate the address space, insert it into the tree.
+    KMemoryBlock* start_block = slab_manager->Allocate();
+    R_UNLESS(start_block != nullptr, ResultOutOfResource);
+
+    // Set our start and end.
+    m_start_address = st;
+    m_end_address = nd;
+    ASSERT(Common::IsAligned(m_start_address, PageSize));
+    ASSERT(Common::IsAligned(m_end_address, PageSize));
+
+    // Initialize and insert the block.
+    start_block->Initialize(m_start_address, (m_end_address - m_start_address) / PageSize,
+                            KMemoryState::Free, KMemoryPermission::None, KMemoryAttribute::None);
+    m_memory_block_tree.insert(*start_block);
+
+    R_SUCCEED();
 }
 
-KMemoryBlockManager::iterator KMemoryBlockManager::FindIterator(VAddr addr) {
-    auto node{memory_block_tree.begin()};
-    while (node != end()) {
-        const VAddr node_end_addr{node->GetNumPages() * PageSize + node->GetAddress()};
-        if (node->GetAddress() <= addr && node_end_addr - 1 >= addr) {
-            return node;
-        }
-        node = std::next(node);
+void KMemoryBlockManager::Finalize(KMemoryBlockSlabManager* slab_manager,
+                                   HostUnmapCallback&& host_unmap_callback) {
+    // Erase every block until we have none left.
+    auto it = m_memory_block_tree.begin();
+    while (it != m_memory_block_tree.end()) {
+        KMemoryBlock* block = std::addressof(*it);
+        it = m_memory_block_tree.erase(it);
+        slab_manager->Free(block);
+        host_unmap_callback(block->GetAddress(), block->GetSize());
     }
-    return end();
+
+    ASSERT(m_memory_block_tree.empty());
 }
 
-VAddr KMemoryBlockManager::FindFreeArea(VAddr region_start, std::size_t region_num_pages,
-                                        std::size_t num_pages, std::size_t align,
-                                        std::size_t offset, std::size_t guard_pages) {
-    if (num_pages == 0) {
-        return {};
-    }
+VAddr KMemoryBlockManager::FindFreeArea(VAddr region_start, size_t region_num_pages,
+                                        size_t num_pages, size_t alignment, size_t offset,
+                                        size_t guard_pages) const {
+    if (num_pages > 0) {
+        const VAddr region_end = region_start + region_num_pages * PageSize;
+        const VAddr region_last = region_end - 1;
+        for (const_iterator it = this->FindIterator(region_start); it != m_memory_block_tree.cend();
+             it++) {
+            const KMemoryInfo info = it->GetMemoryInfo();
+            if (region_last < info.GetAddress()) {
+                break;
+            }
+            if (info.m_state != KMemoryState::Free) {
+                continue;
+            }
 
-    const VAddr region_end{region_start + region_num_pages * PageSize};
-    const VAddr region_last{region_end - 1};
-    for (auto it{FindIterator(region_start)}; it != memory_block_tree.cend(); it++) {
-        const auto info{it->GetMemoryInfo()};
-        if (region_last < info.GetAddress()) {
-            break;
-        }
+            VAddr area = (info.GetAddress() <= region_start) ? region_start : info.GetAddress();
+            area += guard_pages * PageSize;
 
-        if (info.state != KMemoryState::Free) {
-            continue;
-        }
+            const VAddr offset_area = Common::AlignDown(area, alignment) + offset;
+            area = (area <= offset_area) ? offset_area : offset_area + alignment;
 
-        VAddr area{(info.GetAddress() <= region_start) ? region_start : info.GetAddress()};
-        area += guard_pages * PageSize;
+            const VAddr area_end = area + num_pages * PageSize + guard_pages * PageSize;
+            const VAddr area_last = area_end - 1;
 
-        const VAddr offset_area{Common::AlignDown(area, align) + offset};
-        area = (area <= offset_area) ? offset_area : offset_area + align;
-
-        const VAddr area_end{area + num_pages * PageSize + guard_pages * PageSize};
-        const VAddr area_last{area_end - 1};
-
-        if (info.GetAddress() <= area && area < area_last && area_last <= region_last &&
-            area_last <= info.GetLastAddress()) {
-            return area;
+            if (info.GetAddress() <= area && area < area_last && area_last <= region_last &&
+                area_last <= info.GetLastAddress()) {
+                return area;
+            }
         }
     }
 
     return {};
 }
 
-void KMemoryBlockManager::Update(VAddr addr, std::size_t num_pages, KMemoryState prev_state,
-                                 KMemoryPermission prev_perm, KMemoryAttribute prev_attribute,
-                                 KMemoryState state, KMemoryPermission perm,
-                                 KMemoryAttribute attribute) {
-    const VAddr update_end_addr{addr + num_pages * PageSize};
-    iterator node{memory_block_tree.begin()};
+void KMemoryBlockManager::CoalesceForUpdate(KMemoryBlockManagerUpdateAllocator* allocator,
+                                            VAddr address, size_t num_pages) {
+    // Find the iterator now that we've updated.
+    iterator it = this->FindIterator(address);
+    if (address != m_start_address) {
+        it--;
+    }
 
-    prev_attribute |= KMemoryAttribute::IpcAndDeviceMapped;
-
-    while (node != memory_block_tree.end()) {
-        KMemoryBlock* block{&(*node)};
-        iterator next_node{std::next(node)};
-        const VAddr cur_addr{block->GetAddress()};
-        const VAddr cur_end_addr{block->GetNumPages() * PageSize + cur_addr};
-
-        if (addr < cur_end_addr && cur_addr < update_end_addr) {
-            if (!block->HasProperties(prev_state, prev_perm, prev_attribute)) {
-                node = next_node;
-                continue;
-            }
-
-            iterator new_node{node};
-            if (addr > cur_addr) {
-                memory_block_tree.insert(node, block->Split(addr));
-            }
-
-            if (update_end_addr < cur_end_addr) {
-                new_node = memory_block_tree.insert(node, block->Split(update_end_addr));
-            }
-
-            new_node->Update(state, perm, attribute);
-
-            MergeAdjacent(new_node, next_node);
-        }
-
-        if (cur_end_addr - 1 >= update_end_addr - 1) {
+    // Coalesce blocks that we can.
+    while (true) {
+        iterator prev = it++;
+        if (it == m_memory_block_tree.end()) {
             break;
         }
 
-        node = next_node;
-    }
-}
-
-void KMemoryBlockManager::Update(VAddr addr, std::size_t num_pages, KMemoryState state,
-                                 KMemoryPermission perm, KMemoryAttribute attribute) {
-    const VAddr update_end_addr{addr + num_pages * PageSize};
-    iterator node{memory_block_tree.begin()};
-
-    while (node != memory_block_tree.end()) {
-        KMemoryBlock* block{&(*node)};
-        iterator next_node{std::next(node)};
-        const VAddr cur_addr{block->GetAddress()};
-        const VAddr cur_end_addr{block->GetNumPages() * PageSize + cur_addr};
-
-        if (addr < cur_end_addr && cur_addr < update_end_addr) {
-            iterator new_node{node};
-
-            if (addr > cur_addr) {
-                memory_block_tree.insert(node, block->Split(addr));
-            }
-
-            if (update_end_addr < cur_end_addr) {
-                new_node = memory_block_tree.insert(node, block->Split(update_end_addr));
-            }
-
-            new_node->Update(state, perm, attribute);
-
-            MergeAdjacent(new_node, next_node);
+        if (prev->CanMergeWith(*it)) {
+            KMemoryBlock* block = std::addressof(*it);
+            m_memory_block_tree.erase(it);
+            prev->Add(*block);
+            allocator->Free(block);
+            it = prev;
         }
 
-        if (cur_end_addr - 1 >= update_end_addr - 1) {
+        if (address + num_pages * PageSize < it->GetMemoryInfo().GetEndAddress()) {
             break;
         }
-
-        node = next_node;
     }
 }
 
-void KMemoryBlockManager::UpdateLock(VAddr addr, std::size_t num_pages, LockFunc&& lock_func,
+void KMemoryBlockManager::Update(KMemoryBlockManagerUpdateAllocator* allocator, VAddr address,
+                                 size_t num_pages, KMemoryState state, KMemoryPermission perm,
+                                 KMemoryAttribute attr,
+                                 KMemoryBlockDisableMergeAttribute set_disable_attr,
+                                 KMemoryBlockDisableMergeAttribute clear_disable_attr) {
+    // Ensure for auditing that we never end up with an invalid tree.
+    KScopedMemoryBlockManagerAuditor auditor(this);
+    ASSERT(Common::IsAligned(address, PageSize));
+    ASSERT((attr & (KMemoryAttribute::IpcLocked | KMemoryAttribute::DeviceShared)) ==
+           KMemoryAttribute::None);
+
+    VAddr cur_address = address;
+    size_t remaining_pages = num_pages;
+    iterator it = this->FindIterator(address);
+
+    while (remaining_pages > 0) {
+        const size_t remaining_size = remaining_pages * PageSize;
+        KMemoryInfo cur_info = it->GetMemoryInfo();
+        if (it->HasProperties(state, perm, attr)) {
+            // If we already have the right properties, just advance.
+            if (cur_address + remaining_size < cur_info.GetEndAddress()) {
+                remaining_pages = 0;
+                cur_address += remaining_size;
+            } else {
+                remaining_pages =
+                    (cur_address + remaining_size - cur_info.GetEndAddress()) / PageSize;
+                cur_address = cur_info.GetEndAddress();
+            }
+        } else {
+            // If we need to, create a new block before and insert it.
+            if (cur_info.GetAddress() != cur_address) {
+                KMemoryBlock* new_block = allocator->Allocate();
+
+                it->Split(new_block, cur_address);
+                it = m_memory_block_tree.insert(*new_block);
+                it++;
+
+                cur_info = it->GetMemoryInfo();
+                cur_address = cur_info.GetAddress();
+            }
+
+            // If we need to, create a new block after and insert it.
+            if (cur_info.GetSize() > remaining_size) {
+                KMemoryBlock* new_block = allocator->Allocate();
+
+                it->Split(new_block, cur_address + remaining_size);
+                it = m_memory_block_tree.insert(*new_block);
+
+                cur_info = it->GetMemoryInfo();
+            }
+
+            // Update block state.
+            it->Update(state, perm, attr, cur_address == address, static_cast<u8>(set_disable_attr),
+                       static_cast<u8>(clear_disable_attr));
+            cur_address += cur_info.GetSize();
+            remaining_pages -= cur_info.GetNumPages();
+        }
+        it++;
+    }
+
+    this->CoalesceForUpdate(allocator, address, num_pages);
+}
+
+void KMemoryBlockManager::UpdateIfMatch(KMemoryBlockManagerUpdateAllocator* allocator,
+                                        VAddr address, size_t num_pages, KMemoryState test_state,
+                                        KMemoryPermission test_perm, KMemoryAttribute test_attr,
+                                        KMemoryState state, KMemoryPermission perm,
+                                        KMemoryAttribute attr) {
+    // Ensure for auditing that we never end up with an invalid tree.
+    KScopedMemoryBlockManagerAuditor auditor(this);
+    ASSERT(Common::IsAligned(address, PageSize));
+    ASSERT((attr & (KMemoryAttribute::IpcLocked | KMemoryAttribute::DeviceShared)) ==
+           KMemoryAttribute::None);
+
+    VAddr cur_address = address;
+    size_t remaining_pages = num_pages;
+    iterator it = this->FindIterator(address);
+
+    while (remaining_pages > 0) {
+        const size_t remaining_size = remaining_pages * PageSize;
+        KMemoryInfo cur_info = it->GetMemoryInfo();
+        if (it->HasProperties(test_state, test_perm, test_attr) &&
+            !it->HasProperties(state, perm, attr)) {
+            // If we need to, create a new block before and insert it.
+            if (cur_info.GetAddress() != cur_address) {
+                KMemoryBlock* new_block = allocator->Allocate();
+
+                it->Split(new_block, cur_address);
+                it = m_memory_block_tree.insert(*new_block);
+                it++;
+
+                cur_info = it->GetMemoryInfo();
+                cur_address = cur_info.GetAddress();
+            }
+
+            // If we need to, create a new block after and insert it.
+            if (cur_info.GetSize() > remaining_size) {
+                KMemoryBlock* new_block = allocator->Allocate();
+
+                it->Split(new_block, cur_address + remaining_size);
+                it = m_memory_block_tree.insert(*new_block);
+
+                cur_info = it->GetMemoryInfo();
+            }
+
+            // Update block state.
+            it->Update(state, perm, attr, false, 0, 0);
+            cur_address += cur_info.GetSize();
+            remaining_pages -= cur_info.GetNumPages();
+        } else {
+            // If we already have the right properties, just advance.
+            if (cur_address + remaining_size < cur_info.GetEndAddress()) {
+                remaining_pages = 0;
+                cur_address += remaining_size;
+            } else {
+                remaining_pages =
+                    (cur_address + remaining_size - cur_info.GetEndAddress()) / PageSize;
+                cur_address = cur_info.GetEndAddress();
+            }
+        }
+        it++;
+    }
+
+    this->CoalesceForUpdate(allocator, address, num_pages);
+}
+
+void KMemoryBlockManager::UpdateLock(KMemoryBlockManagerUpdateAllocator* allocator, VAddr address,
+                                     size_t num_pages, MemoryBlockLockFunction lock_func,
                                      KMemoryPermission perm) {
-    const VAddr update_end_addr{addr + num_pages * PageSize};
-    iterator node{memory_block_tree.begin()};
+    // Ensure for auditing that we never end up with an invalid tree.
+    KScopedMemoryBlockManagerAuditor auditor(this);
+    ASSERT(Common::IsAligned(address, PageSize));
 
-    while (node != memory_block_tree.end()) {
-        KMemoryBlock* block{&(*node)};
-        iterator next_node{std::next(node)};
-        const VAddr cur_addr{block->GetAddress()};
-        const VAddr cur_end_addr{block->GetNumPages() * PageSize + cur_addr};
+    VAddr cur_address = address;
+    size_t remaining_pages = num_pages;
+    iterator it = this->FindIterator(address);
 
-        if (addr < cur_end_addr && cur_addr < update_end_addr) {
-            iterator new_node{node};
+    const VAddr end_address = address + (num_pages * PageSize);
 
-            if (addr > cur_addr) {
-                memory_block_tree.insert(node, block->Split(addr));
-            }
+    while (remaining_pages > 0) {
+        const size_t remaining_size = remaining_pages * PageSize;
+        KMemoryInfo cur_info = it->GetMemoryInfo();
 
-            if (update_end_addr < cur_end_addr) {
-                new_node = memory_block_tree.insert(node, block->Split(update_end_addr));
-            }
+        // If we need to, create a new block before and insert it.
+        if (cur_info.m_address != cur_address) {
+            KMemoryBlock* new_block = allocator->Allocate();
 
-            lock_func(new_node, perm);
+            it->Split(new_block, cur_address);
+            it = m_memory_block_tree.insert(*new_block);
+            it++;
 
-            MergeAdjacent(new_node, next_node);
+            cur_info = it->GetMemoryInfo();
+            cur_address = cur_info.GetAddress();
         }
 
-        if (cur_end_addr - 1 >= update_end_addr - 1) {
-            break;
+        if (cur_info.GetSize() > remaining_size) {
+            // If we need to, create a new block after and insert it.
+            KMemoryBlock* new_block = allocator->Allocate();
+
+            it->Split(new_block, cur_address + remaining_size);
+            it = m_memory_block_tree.insert(*new_block);
+
+            cur_info = it->GetMemoryInfo();
         }
 
-        node = next_node;
+        // Call the locked update function.
+        (std::addressof(*it)->*lock_func)(perm, cur_info.GetAddress() == address,
+                                          cur_info.GetEndAddress() == end_address);
+        cur_address += cur_info.GetSize();
+        remaining_pages -= cur_info.GetNumPages();
+        it++;
     }
+
+    this->CoalesceForUpdate(allocator, address, num_pages);
 }
 
-void KMemoryBlockManager::IterateForRange(VAddr start, VAddr end, IterateFunc&& func) {
-    const_iterator it{FindIterator(start)};
-    KMemoryInfo info{};
-    do {
-        info = it->GetMemoryInfo();
-        func(info);
-        it = std::next(it);
-    } while (info.addr + info.size - 1 < end - 1 && it != cend());
-}
+// Debug.
+bool KMemoryBlockManager::CheckState() const {
+    // Loop over every block, ensuring that we are sorted and coalesced.
+    auto it = m_memory_block_tree.cbegin();
+    auto prev = it++;
+    while (it != m_memory_block_tree.cend()) {
+        const KMemoryInfo prev_info = prev->GetMemoryInfo();
+        const KMemoryInfo cur_info = it->GetMemoryInfo();
 
-void KMemoryBlockManager::MergeAdjacent(iterator it, iterator& next_it) {
-    KMemoryBlock* block{&(*it)};
-
-    auto EraseIt = [&](const iterator it_to_erase) {
-        if (next_it == it_to_erase) {
-            next_it = std::next(next_it);
+        // Sequential blocks which can be merged should be merged.
+        if (prev->CanMergeWith(*it)) {
+            return false;
         }
-        memory_block_tree.erase(it_to_erase);
-    };
 
-    if (it != memory_block_tree.begin()) {
-        KMemoryBlock* prev{&(*std::prev(it))};
+        // Sequential blocks should be sequential.
+        if (prev_info.GetEndAddress() != cur_info.GetAddress()) {
+            return false;
+        }
 
-        if (block->HasSameProperties(*prev)) {
-            const iterator prev_it{std::prev(it)};
+        // If the block is ipc locked, it must have a count.
+        if ((cur_info.m_attribute & KMemoryAttribute::IpcLocked) != KMemoryAttribute::None &&
+            cur_info.m_ipc_lock_count == 0) {
+            return false;
+        }
 
-            prev->Add(block->GetNumPages());
-            EraseIt(it);
+        // If the block is device shared, it must have a count.
+        if ((cur_info.m_attribute & KMemoryAttribute::DeviceShared) != KMemoryAttribute::None &&
+            cur_info.m_device_use_count == 0) {
+            return false;
+        }
 
-            it = prev_it;
-            block = prev;
+        // Advance the iterator.
+        prev = it++;
+    }
+
+    // Our loop will miss checking the last block, potentially, so check it.
+    if (prev != m_memory_block_tree.cend()) {
+        const KMemoryInfo prev_info = prev->GetMemoryInfo();
+        // If the block is ipc locked, it must have a count.
+        if ((prev_info.m_attribute & KMemoryAttribute::IpcLocked) != KMemoryAttribute::None &&
+            prev_info.m_ipc_lock_count == 0) {
+            return false;
+        }
+
+        // If the block is device shared, it must have a count.
+        if ((prev_info.m_attribute & KMemoryAttribute::DeviceShared) != KMemoryAttribute::None &&
+            prev_info.m_device_use_count == 0) {
+            return false;
         }
     }
 
-    if (it != cend()) {
-        const KMemoryBlock* const next{&(*std::next(it))};
-
-        if (block->HasSameProperties(*next)) {
-            block->Add(next->GetNumPages());
-            EraseIt(std::next(it));
-        }
-    }
+    return true;
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_memory_block_manager.h

@@ -4,63 +4,154 @@
 #pragma once
 
 #include <functional>
-#include <list>
 
+#include "common/common_funcs.h"
 #include "common/common_types.h"
+#include "core/hle/kernel/k_dynamic_resource_manager.h"
 #include "core/hle/kernel/k_memory_block.h"
 
 namespace Kernel {
 
+class KMemoryBlockManagerUpdateAllocator {
+public:
+    static constexpr size_t MaxBlocks = 2;
+
+private:
+    KMemoryBlock* m_blocks[MaxBlocks];
+    size_t m_index;
+    KMemoryBlockSlabManager* m_slab_manager;
+
+private:
+    Result Initialize(size_t num_blocks) {
+        // Check num blocks.
+        ASSERT(num_blocks <= MaxBlocks);
+
+        // Set index.
+        m_index = MaxBlocks - num_blocks;
+
+        // Allocate the blocks.
+        for (size_t i = 0; i < num_blocks && i < MaxBlocks; ++i) {
+            m_blocks[m_index + i] = m_slab_manager->Allocate();
+            R_UNLESS(m_blocks[m_index + i] != nullptr, ResultOutOfResource);
+        }
+
+        R_SUCCEED();
+    }
+
+public:
+    KMemoryBlockManagerUpdateAllocator(Result* out_result, KMemoryBlockSlabManager* sm,
+                                       size_t num_blocks = MaxBlocks)
+        : m_blocks(), m_index(MaxBlocks), m_slab_manager(sm) {
+        *out_result = this->Initialize(num_blocks);
+    }
+
+    ~KMemoryBlockManagerUpdateAllocator() {
+        for (const auto& block : m_blocks) {
+            if (block != nullptr) {
+                m_slab_manager->Free(block);
+            }
+        }
+    }
+
+    KMemoryBlock* Allocate() {
+        ASSERT(m_index < MaxBlocks);
+        ASSERT(m_blocks[m_index] != nullptr);
+        KMemoryBlock* block = nullptr;
+        std::swap(block, m_blocks[m_index++]);
+        return block;
+    }
+
+    void Free(KMemoryBlock* block) {
+        ASSERT(m_index <= MaxBlocks);
+        ASSERT(block != nullptr);
+        if (m_index == 0) {
+            m_slab_manager->Free(block);
+        } else {
+            m_blocks[--m_index] = block;
+        }
+    }
+};
+
 class KMemoryBlockManager final {
 public:
-    using MemoryBlockTree = std::list<KMemoryBlock>;
+    using MemoryBlockTree =
+        Common::IntrusiveRedBlackTreeBaseTraits<KMemoryBlock>::TreeType<KMemoryBlock>;
+    using MemoryBlockLockFunction = void (KMemoryBlock::*)(KMemoryPermission new_perm, bool left,
+                                                           bool right);
     using iterator = MemoryBlockTree::iterator;
     using const_iterator = MemoryBlockTree::const_iterator;
 
 public:
-    KMemoryBlockManager(VAddr start_addr_, VAddr end_addr_);
+    KMemoryBlockManager();
+
+    using HostUnmapCallback = std::function<void(VAddr, u64)>;
+
+    Result Initialize(VAddr st, VAddr nd, KMemoryBlockSlabManager* slab_manager);
+    void Finalize(KMemoryBlockSlabManager* slab_manager, HostUnmapCallback&& host_unmap_callback);
 
     iterator end() {
-        return memory_block_tree.end();
+        return m_memory_block_tree.end();
     }
     const_iterator end() const {
-        return memory_block_tree.end();
+        return m_memory_block_tree.end();
     }
     const_iterator cend() const {
-        return memory_block_tree.cend();
+        return m_memory_block_tree.cend();
     }
 
-    iterator FindIterator(VAddr addr);
+    VAddr FindFreeArea(VAddr region_start, size_t region_num_pages, size_t num_pages,
+                       size_t alignment, size_t offset, size_t guard_pages) const;
 
-    VAddr FindFreeArea(VAddr region_start, std::size_t region_num_pages, std::size_t num_pages,
-                       std::size_t align, std::size_t offset, std::size_t guard_pages);
+    void Update(KMemoryBlockManagerUpdateAllocator* allocator, VAddr address, size_t num_pages,
+                KMemoryState state, KMemoryPermission perm, KMemoryAttribute attr,
+                KMemoryBlockDisableMergeAttribute set_disable_attr,
+                KMemoryBlockDisableMergeAttribute clear_disable_attr);
+    void UpdateLock(KMemoryBlockManagerUpdateAllocator* allocator, VAddr address, size_t num_pages,
+                    MemoryBlockLockFunction lock_func, KMemoryPermission perm);
 
-    void Update(VAddr addr, std::size_t num_pages, KMemoryState prev_state,
-                KMemoryPermission prev_perm, KMemoryAttribute prev_attribute, KMemoryState state,
-                KMemoryPermission perm, KMemoryAttribute attribute);
+    void UpdateIfMatch(KMemoryBlockManagerUpdateAllocator* allocator, VAddr address,
+                       size_t num_pages, KMemoryState test_state, KMemoryPermission test_perm,
+                       KMemoryAttribute test_attr, KMemoryState state, KMemoryPermission perm,
+                       KMemoryAttribute attr);
 
-    void Update(VAddr addr, std::size_t num_pages, KMemoryState state,
-                KMemoryPermission perm = KMemoryPermission::None,
-                KMemoryAttribute attribute = KMemoryAttribute::None);
+    iterator FindIterator(VAddr address) const {
+        return m_memory_block_tree.find(KMemoryBlock(
+            address, 1, KMemoryState::Free, KMemoryPermission::None, KMemoryAttribute::None));
+    }
 
-    using LockFunc = std::function<void(iterator, KMemoryPermission)>;
-    void UpdateLock(VAddr addr, std::size_t num_pages, LockFunc&& lock_func,
-                    KMemoryPermission perm);
+    const KMemoryBlock* FindBlock(VAddr address) const {
+        if (const_iterator it = this->FindIterator(address); it != m_memory_block_tree.end()) {
+            return std::addressof(*it);
+        }
 
-    using IterateFunc = std::function<void(const KMemoryInfo&)>;
-    void IterateForRange(VAddr start, VAddr end, IterateFunc&& func);
+        return nullptr;
+    }
 
-    KMemoryBlock& FindBlock(VAddr addr) {
-        return *FindIterator(addr);
+    // Debug.
+    bool CheckState() const;
+
+private:
+    void CoalesceForUpdate(KMemoryBlockManagerUpdateAllocator* allocator, VAddr address,
+                           size_t num_pages);
+
+    MemoryBlockTree m_memory_block_tree;
+    VAddr m_start_address{};
+    VAddr m_end_address{};
+};
+
+class KScopedMemoryBlockManagerAuditor {
+public:
+    explicit KScopedMemoryBlockManagerAuditor(KMemoryBlockManager* m) : m_manager(m) {
+        ASSERT(m_manager->CheckState());
+    }
+    explicit KScopedMemoryBlockManagerAuditor(KMemoryBlockManager& m)
+        : KScopedMemoryBlockManagerAuditor(std::addressof(m)) {}
+    ~KScopedMemoryBlockManagerAuditor() {
+        ASSERT(m_manager->CheckState());
     }
 
 private:
-    void MergeAdjacent(iterator it, iterator& next_it);
-
-    [[maybe_unused]] const VAddr start_addr;
-    [[maybe_unused]] const VAddr end_addr;
-
-    MemoryBlockTree memory_block_tree;
+    KMemoryBlockManager* m_manager;
 };
 
 } // namespace Kernel

src/core/hle/kernel/k_memory_manager.cpp

@@ -331,7 +331,7 @@ Result KMemoryManager::AllocateAndOpenForProcess(KPageGroup* out, size_t num_pag
 
     // Set all the allocated memory.
     for (const auto& block : out->Nodes()) {
-        std::memset(system.DeviceMemory().GetPointer(block.GetAddress()), fill_pattern,
+        std::memset(system.DeviceMemory().GetPointer<void>(block.GetAddress()), fill_pattern,
                     block.GetSize());
     }
 

src/core/hle/kernel/k_page_buffer.cpp

@@ -12,7 +12,7 @@ namespace Kernel {
 
 KPageBuffer* KPageBuffer::FromPhysicalAddress(Core::System& system, PAddr phys_addr) {
     ASSERT(Common::IsAligned(phys_addr, PageSize));
-    return reinterpret_cast<KPageBuffer*>(system.DeviceMemory().GetPointer(phys_addr));
+    return system.DeviceMemory().GetPointer<KPageBuffer>(phys_addr);
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_page_buffer.h

@@ -13,6 +13,7 @@ namespace Kernel {
 
 class KPageBuffer final : public KSlabAllocated<KPageBuffer> {
 public:
+    explicit KPageBuffer(KernelCore&) {}
     KPageBuffer() = default;
 
     static KPageBuffer* FromPhysicalAddress(Core::System& system, PAddr phys_addr);

src/core/hle/kernel/k_page_table.h

@@ -9,8 +9,10 @@
 #include "common/common_types.h"
 #include "common/page_table.h"
 #include "core/file_sys/program_metadata.h"
+#include "core/hle/kernel/k_dynamic_resource_manager.h"
 #include "core/hle/kernel/k_light_lock.h"
 #include "core/hle/kernel/k_memory_block.h"
+#include "core/hle/kernel/k_memory_block_manager.h"
 #include "core/hle/kernel/k_memory_layout.h"
 #include "core/hle/kernel/k_memory_manager.h"
 #include "core/hle/result.h"
@@ -34,58 +36,66 @@ public:
     ~KPageTable();
 
     Result InitializeForProcess(FileSys::ProgramAddressSpaceType as_type, bool enable_aslr,
-                                VAddr code_addr, std::size_t code_size, KMemoryManager::Pool pool);
-    Result MapProcessCode(VAddr addr, std::size_t pages_count, KMemoryState state,
+                                VAddr code_addr, size_t code_size,
+                                KMemoryBlockSlabManager* mem_block_slab_manager,
+                                KMemoryManager::Pool pool);
+
+    void Finalize();
+
+    Result MapProcessCode(VAddr addr, size_t pages_count, KMemoryState state,
                           KMemoryPermission perm);
-    Result MapCodeMemory(VAddr dst_address, VAddr src_address, std::size_t size);
-    Result UnmapCodeMemory(VAddr dst_address, VAddr src_address, std::size_t size,
+    Result MapCodeMemory(VAddr dst_address, VAddr src_address, size_t size);
+    Result UnmapCodeMemory(VAddr dst_address, VAddr src_address, size_t size,
                            ICacheInvalidationStrategy icache_invalidation_strategy);
-    Result UnmapProcessMemory(VAddr dst_addr, std::size_t size, KPageTable& src_page_table,
+    Result UnmapProcessMemory(VAddr dst_addr, size_t size, KPageTable& src_page_table,
                               VAddr src_addr);
-    Result MapPhysicalMemory(VAddr addr, std::size_t size);
-    Result UnmapPhysicalMemory(VAddr addr, std::size_t size);
-    Result MapMemory(VAddr dst_addr, VAddr src_addr, std::size_t size);
-    Result UnmapMemory(VAddr dst_addr, VAddr src_addr, std::size_t size);
+    Result MapPhysicalMemory(VAddr addr, size_t size);
+    Result UnmapPhysicalMemory(VAddr addr, size_t size);
+    Result MapMemory(VAddr dst_addr, VAddr src_addr, size_t size);
+    Result UnmapMemory(VAddr dst_addr, VAddr src_addr, size_t size);
     Result MapPages(VAddr addr, KPageGroup& page_linked_list, KMemoryState state,
                     KMemoryPermission perm);
-    Result MapPages(VAddr* out_addr, std::size_t num_pages, std::size_t alignment, PAddr phys_addr,
+    Result MapPages(VAddr* out_addr, size_t num_pages, size_t alignment, PAddr phys_addr,
                     KMemoryState state, KMemoryPermission perm) {
-        return this->MapPages(out_addr, num_pages, alignment, phys_addr, true,
-                              this->GetRegionAddress(state), this->GetRegionSize(state) / PageSize,
-                              state, perm);
+        R_RETURN(this->MapPages(out_addr, num_pages, alignment, phys_addr, true,
+                                this->GetRegionAddress(state),
+                                this->GetRegionSize(state) / PageSize, state, perm));
     }
     Result UnmapPages(VAddr addr, KPageGroup& page_linked_list, KMemoryState state);
-    Result UnmapPages(VAddr address, std::size_t num_pages, KMemoryState state);
-    Result SetProcessMemoryPermission(VAddr addr, std::size_t size, Svc::MemoryPermission svc_perm);
+    Result UnmapPages(VAddr address, size_t num_pages, KMemoryState state);
+    Result SetProcessMemoryPermission(VAddr addr, size_t size, Svc::MemoryPermission svc_perm);
     KMemoryInfo QueryInfo(VAddr addr);
-    Result ReserveTransferMemory(VAddr addr, std::size_t size, KMemoryPermission perm);
-    Result ResetTransferMemory(VAddr addr, std::size_t size);
-    Result SetMemoryPermission(VAddr addr, std::size_t size, Svc::MemoryPermission perm);
-    Result SetMemoryAttribute(VAddr addr, std::size_t size, u32 mask, u32 attr);
-    Result SetMaxHeapSize(std::size_t size);
-    Result SetHeapSize(VAddr* out, std::size_t size);
-    ResultVal<VAddr> AllocateAndMapMemory(std::size_t needed_num_pages, std::size_t align,
-                                          bool is_map_only, VAddr region_start,
-                                          std::size_t region_num_pages, KMemoryState state,
-                                          KMemoryPermission perm, PAddr map_addr = 0);
-    Result LockForDeviceAddressSpace(VAddr addr, std::size_t size);
-    Result UnlockForDeviceAddressSpace(VAddr addr, std::size_t size);
-    Result LockForCodeMemory(KPageGroup* out, VAddr addr, std::size_t size);
-    Result UnlockForCodeMemory(VAddr addr, std::size_t size, const KPageGroup& pg);
+    Result SetMemoryPermission(VAddr addr, size_t size, Svc::MemoryPermission perm);
+    Result SetMemoryAttribute(VAddr addr, size_t size, u32 mask, u32 attr);
+    Result SetMaxHeapSize(size_t size);
+    Result SetHeapSize(VAddr* out, size_t size);
+    ResultVal<VAddr> AllocateAndMapMemory(size_t needed_num_pages, size_t align, bool is_map_only,
+                                          VAddr region_start, size_t region_num_pages,
+                                          KMemoryState state, KMemoryPermission perm,
+                                          PAddr map_addr = 0);
+
+    Result LockForMapDeviceAddressSpace(VAddr address, size_t size, KMemoryPermission perm,
+                                        bool is_aligned);
+    Result LockForUnmapDeviceAddressSpace(VAddr address, size_t size);
+
+    Result UnlockForDeviceAddressSpace(VAddr addr, size_t size);
+
+    Result LockForCodeMemory(KPageGroup* out, VAddr addr, size_t size);
+    Result UnlockForCodeMemory(VAddr addr, size_t size, const KPageGroup& pg);
     Result MakeAndOpenPageGroup(KPageGroup* out, VAddr address, size_t num_pages,
                                 KMemoryState state_mask, KMemoryState state,
                                 KMemoryPermission perm_mask, KMemoryPermission perm,
                                 KMemoryAttribute attr_mask, KMemoryAttribute attr);
 
     Common::PageTable& PageTableImpl() {
-        return page_table_impl;
+        return *m_page_table_impl;
     }
 
     const Common::PageTable& PageTableImpl() const {
-        return page_table_impl;
+        return *m_page_table_impl;
     }
 
-    bool CanContain(VAddr addr, std::size_t size, KMemoryState state) const;
+    bool CanContain(VAddr addr, size_t size, KMemoryState state) const;
 
 private:
     enum class OperationType : u32 {
@@ -96,67 +106,65 @@ private:
         ChangePermissionsAndRefresh,
     };
 
-    static constexpr KMemoryAttribute DefaultMemoryIgnoreAttr = KMemoryAttribute::DontCareMask |
-                                                                KMemoryAttribute::IpcLocked |
-                                                                KMemoryAttribute::DeviceShared;
+    static constexpr KMemoryAttribute DefaultMemoryIgnoreAttr =
+        KMemoryAttribute::IpcLocked | KMemoryAttribute::DeviceShared;
 
-    Result InitializeMemoryLayout(VAddr start, VAddr end);
     Result MapPages(VAddr addr, const KPageGroup& page_linked_list, KMemoryPermission perm);
-    Result MapPages(VAddr* out_addr, std::size_t num_pages, std::size_t alignment, PAddr phys_addr,
-                    bool is_pa_valid, VAddr region_start, std::size_t region_num_pages,
+    Result MapPages(VAddr* out_addr, size_t num_pages, size_t alignment, PAddr phys_addr,
+                    bool is_pa_valid, VAddr region_start, size_t region_num_pages,
                     KMemoryState state, KMemoryPermission perm);
     Result UnmapPages(VAddr addr, const KPageGroup& page_linked_list);
-    bool IsRegionMapped(VAddr address, u64 size);
     bool IsRegionContiguous(VAddr addr, u64 size) const;
-    void AddRegionToPages(VAddr start, std::size_t num_pages, KPageGroup& page_linked_list);
+    void AddRegionToPages(VAddr start, size_t num_pages, KPageGroup& page_linked_list);
     KMemoryInfo QueryInfoImpl(VAddr addr);
-    VAddr AllocateVirtualMemory(VAddr start, std::size_t region_num_pages, u64 needed_num_pages,
-                                std::size_t align);
-    Result Operate(VAddr addr, std::size_t num_pages, const KPageGroup& page_group,
+    VAddr AllocateVirtualMemory(VAddr start, size_t region_num_pages, u64 needed_num_pages,
+                                size_t align);
+    Result Operate(VAddr addr, size_t num_pages, const KPageGroup& page_group,
                    OperationType operation);
-    Result Operate(VAddr addr, std::size_t num_pages, KMemoryPermission perm,
-                   OperationType operation, PAddr map_addr = 0);
+    Result Operate(VAddr addr, size_t num_pages, KMemoryPermission perm, OperationType operation,
+                   PAddr map_addr = 0);
     VAddr GetRegionAddress(KMemoryState state) const;
-    std::size_t GetRegionSize(KMemoryState state) const;
+    size_t GetRegionSize(KMemoryState state) const;
 
-    VAddr FindFreeArea(VAddr region_start, std::size_t region_num_pages, std::size_t num_pages,
-                       std::size_t alignment, std::size_t offset, std::size_t guard_pages);
+    VAddr FindFreeArea(VAddr region_start, size_t region_num_pages, size_t num_pages,
+                       size_t alignment, size_t offset, size_t guard_pages);
 
-    Result CheckMemoryStateContiguous(std::size_t* out_blocks_needed, VAddr addr, std::size_t size,
+    Result CheckMemoryStateContiguous(size_t* out_blocks_needed, VAddr addr, size_t size,
                                       KMemoryState state_mask, KMemoryState state,
                                       KMemoryPermission perm_mask, KMemoryPermission perm,
                                       KMemoryAttribute attr_mask, KMemoryAttribute attr) const;
-    Result CheckMemoryStateContiguous(VAddr addr, std::size_t size, KMemoryState state_mask,
+    Result CheckMemoryStateContiguous(VAddr addr, size_t size, KMemoryState state_mask,
                                       KMemoryState state, KMemoryPermission perm_mask,
                                       KMemoryPermission perm, KMemoryAttribute attr_mask,
                                       KMemoryAttribute attr) const {
-        return this->CheckMemoryStateContiguous(nullptr, addr, size, state_mask, state, perm_mask,
-                                                perm, attr_mask, attr);
+        R_RETURN(this->CheckMemoryStateContiguous(nullptr, addr, size, state_mask, state, perm_mask,
+                                                  perm, attr_mask, attr));
     }
 
     Result CheckMemoryState(const KMemoryInfo& info, KMemoryState state_mask, KMemoryState state,
                             KMemoryPermission perm_mask, KMemoryPermission perm,
                             KMemoryAttribute attr_mask, KMemoryAttribute attr) const;
     Result CheckMemoryState(KMemoryState* out_state, KMemoryPermission* out_perm,
-                            KMemoryAttribute* out_attr, std::size_t* out_blocks_needed, VAddr addr,
-                            std::size_t size, KMemoryState state_mask, KMemoryState state,
+                            KMemoryAttribute* out_attr, size_t* out_blocks_needed, VAddr addr,
+                            size_t size, KMemoryState state_mask, KMemoryState state,
                             KMemoryPermission perm_mask, KMemoryPermission perm,
                             KMemoryAttribute attr_mask, KMemoryAttribute attr,
                             KMemoryAttribute ignore_attr = DefaultMemoryIgnoreAttr) const;
-    Result CheckMemoryState(std::size_t* out_blocks_needed, VAddr addr, std::size_t size,
+    Result CheckMemoryState(size_t* out_blocks_needed, VAddr addr, size_t size,
                             KMemoryState state_mask, KMemoryState state,
                             KMemoryPermission perm_mask, KMemoryPermission perm,
                             KMemoryAttribute attr_mask, KMemoryAttribute attr,
                             KMemoryAttribute ignore_attr = DefaultMemoryIgnoreAttr) const {
-        return CheckMemoryState(nullptr, nullptr, nullptr, out_blocks_needed, addr, size,
-                                state_mask, state, perm_mask, perm, attr_mask, attr, ignore_attr);
+        R_RETURN(CheckMemoryState(nullptr, nullptr, nullptr, out_blocks_needed, addr, size,
+                                  state_mask, state, perm_mask, perm, attr_mask, attr,
+                                  ignore_attr));
     }
-    Result CheckMemoryState(VAddr addr, std::size_t size, KMemoryState state_mask,
-                            KMemoryState state, KMemoryPermission perm_mask, KMemoryPermission perm,
+    Result CheckMemoryState(VAddr addr, size_t size, KMemoryState state_mask, KMemoryState state,
+                            KMemoryPermission perm_mask, KMemoryPermission perm,
                             KMemoryAttribute attr_mask, KMemoryAttribute attr,
                             KMemoryAttribute ignore_attr = DefaultMemoryIgnoreAttr) const {
-        return this->CheckMemoryState(nullptr, addr, size, state_mask, state, perm_mask, perm,
-                                      attr_mask, attr, ignore_attr);
+        R_RETURN(this->CheckMemoryState(nullptr, addr, size, state_mask, state, perm_mask, perm,
+                                        attr_mask, attr, ignore_attr));
     }
 
     Result LockMemoryAndOpen(KPageGroup* out_pg, PAddr* out_paddr, VAddr addr, size_t size,
@@ -174,13 +182,13 @@ private:
     bool IsValidPageGroup(const KPageGroup& pg, VAddr addr, size_t num_pages);
 
     bool IsLockedByCurrentThread() const {
-        return general_lock.IsLockedByCurrentThread();
+        return m_general_lock.IsLockedByCurrentThread();
     }
 
     bool IsHeapPhysicalAddress(const KMemoryLayout& layout, PAddr phys_addr) {
         ASSERT(this->IsLockedByCurrentThread());
 
-        return layout.IsHeapPhysicalAddress(cached_physical_heap_region, phys_addr);
+        return layout.IsHeapPhysicalAddress(m_cached_physical_heap_region, phys_addr);
     }
 
     bool GetPhysicalAddressLocked(PAddr* out, VAddr virt_addr) const {
@@ -191,95 +199,93 @@ private:
         return *out != 0;
     }
 
-    mutable KLightLock general_lock;
-    mutable KLightLock map_physical_memory_lock;
-
-    std::unique_ptr<KMemoryBlockManager> block_manager;
+    mutable KLightLock m_general_lock;
+    mutable KLightLock m_map_physical_memory_lock;
 
 public:
     constexpr VAddr GetAddressSpaceStart() const {
-        return address_space_start;
+        return m_address_space_start;
     }
     constexpr VAddr GetAddressSpaceEnd() const {
-        return address_space_end;
+        return m_address_space_end;
     }
-    constexpr std::size_t GetAddressSpaceSize() const {
-        return address_space_end - address_space_start;
+    constexpr size_t GetAddressSpaceSize() const {
+        return m_address_space_end - m_address_space_start;
     }
     constexpr VAddr GetHeapRegionStart() const {
-        return heap_region_start;
+        return m_heap_region_start;
     }
     constexpr VAddr GetHeapRegionEnd() const {
-        return heap_region_end;
+        return m_heap_region_end;
     }
-    constexpr std::size_t GetHeapRegionSize() const {
-        return heap_region_end - heap_region_start;
+    constexpr size_t GetHeapRegionSize() const {
+        return m_heap_region_end - m_heap_region_start;
     }
     constexpr VAddr GetAliasRegionStart() const {
-        return alias_region_start;
+        return m_alias_region_start;
     }
     constexpr VAddr GetAliasRegionEnd() const {
-        return alias_region_end;
+        return m_alias_region_end;
     }
-    constexpr std::size_t GetAliasRegionSize() const {
-        return alias_region_end - alias_region_start;
+    constexpr size_t GetAliasRegionSize() const {
+        return m_alias_region_end - m_alias_region_start;
     }
     constexpr VAddr GetStackRegionStart() const {
-        return stack_region_start;
+        return m_stack_region_start;
     }
     constexpr VAddr GetStackRegionEnd() const {
-        return stack_region_end;
+        return m_stack_region_end;
     }
-    constexpr std::size_t GetStackRegionSize() const {
-        return stack_region_end - stack_region_start;
+    constexpr size_t GetStackRegionSize() const {
+        return m_stack_region_end - m_stack_region_start;
     }
     constexpr VAddr GetKernelMapRegionStart() const {
-        return kernel_map_region_start;
+        return m_kernel_map_region_start;
     }
     constexpr VAddr GetKernelMapRegionEnd() const {
-        return kernel_map_region_end;
+        return m_kernel_map_region_end;
     }
     constexpr VAddr GetCodeRegionStart() const {
-        return code_region_start;
+        return m_code_region_start;
     }
     constexpr VAddr GetCodeRegionEnd() const {
-        return code_region_end;
+        return m_code_region_end;
     }
     constexpr VAddr GetAliasCodeRegionStart() const {
-        return alias_code_region_start;
+        return m_alias_code_region_start;
     }
     constexpr VAddr GetAliasCodeRegionSize() const {
-        return alias_code_region_end - alias_code_region_start;
+        return m_alias_code_region_end - m_alias_code_region_start;
     }
-    std::size_t GetNormalMemorySize() {
-        KScopedLightLock lk(general_lock);
-        return GetHeapSize() + mapped_physical_memory_size;
+    size_t GetNormalMemorySize() {
+        KScopedLightLock lk(m_general_lock);
+        return GetHeapSize() + m_mapped_physical_memory_size;
     }
-    constexpr std::size_t GetAddressSpaceWidth() const {
-        return address_space_width;
+    constexpr size_t GetAddressSpaceWidth() const {
+        return m_address_space_width;
     }
-    constexpr std::size_t GetHeapSize() const {
-        return current_heap_end - heap_region_start;
+    constexpr size_t GetHeapSize() const {
+        return m_current_heap_end - m_heap_region_start;
     }
-    constexpr bool IsInsideAddressSpace(VAddr address, std::size_t size) const {
-        return address_space_start <= address && address + size - 1 <= address_space_end - 1;
+    constexpr bool IsInsideAddressSpace(VAddr address, size_t size) const {
+        return m_address_space_start <= address && address + size - 1 <= m_address_space_end - 1;
     }
-    constexpr bool IsOutsideAliasRegion(VAddr address, std::size_t size) const {
-        return alias_region_start > address || address + size - 1 > alias_region_end - 1;
+    constexpr bool IsOutsideAliasRegion(VAddr address, size_t size) const {
+        return m_alias_region_start > address || address + size - 1 > m_alias_region_end - 1;
     }
-    constexpr bool IsOutsideStackRegion(VAddr address, std::size_t size) const {
-        return stack_region_start > address || address + size - 1 > stack_region_end - 1;
+    constexpr bool IsOutsideStackRegion(VAddr address, size_t size) const {
+        return m_stack_region_start > address || address + size - 1 > m_stack_region_end - 1;
     }
-    constexpr bool IsInvalidRegion(VAddr address, std::size_t size) const {
+    constexpr bool IsInvalidRegion(VAddr address, size_t size) const {
         return address + size - 1 > GetAliasCodeRegionStart() + GetAliasCodeRegionSize() - 1;
     }
-    constexpr bool IsInsideHeapRegion(VAddr address, std::size_t size) const {
-        return address + size > heap_region_start && heap_region_end > address;
+    constexpr bool IsInsideHeapRegion(VAddr address, size_t size) const {
+        return address + size > m_heap_region_start && m_heap_region_end > address;
     }
-    constexpr bool IsInsideAliasRegion(VAddr address, std::size_t size) const {
-        return address + size > alias_region_start && alias_region_end > address;
+    constexpr bool IsInsideAliasRegion(VAddr address, size_t size) const {
+        return address + size > m_alias_region_start && m_alias_region_end > address;
     }
-    constexpr bool IsOutsideASLRRegion(VAddr address, std::size_t size) const {
+    constexpr bool IsOutsideASLRRegion(VAddr address, size_t size) const {
         if (IsInvalidRegion(address, size)) {
             return true;
         }
@@ -291,73 +297,78 @@ public:
         }
         return {};
     }
-    constexpr bool IsInsideASLRRegion(VAddr address, std::size_t size) const {
+    constexpr bool IsInsideASLRRegion(VAddr address, size_t size) const {
         return !IsOutsideASLRRegion(address, size);
     }
-    constexpr std::size_t GetNumGuardPages() const {
+    constexpr size_t GetNumGuardPages() const {
         return IsKernel() ? 1 : 4;
     }
     PAddr GetPhysicalAddr(VAddr addr) const {
-        const auto backing_addr = page_table_impl.backing_addr[addr >> PageBits];
+        const auto backing_addr = m_page_table_impl->backing_addr[addr >> PageBits];
         ASSERT(backing_addr);
         return backing_addr + addr;
     }
     constexpr bool Contains(VAddr addr) const {
-        return address_space_start <= addr && addr <= address_space_end - 1;
+        return m_address_space_start <= addr && addr <= m_address_space_end - 1;
     }
-    constexpr bool Contains(VAddr addr, std::size_t size) const {
-        return address_space_start <= addr && addr < addr + size &&
-               addr + size - 1 <= address_space_end - 1;
+    constexpr bool Contains(VAddr addr, size_t size) const {
+        return m_address_space_start <= addr && addr < addr + size &&
+               addr + size - 1 <= m_address_space_end - 1;
     }
 
 private:
     constexpr bool IsKernel() const {
-        return is_kernel;
+        return m_is_kernel;
     }
     constexpr bool IsAslrEnabled() const {
-        return is_aslr_enabled;
+        return m_enable_aslr;
     }
 
-    constexpr bool ContainsPages(VAddr addr, std::size_t num_pages) const {
-        return (address_space_start <= addr) &&
-               (num_pages <= (address_space_end - address_space_start) / PageSize) &&
-               (addr + num_pages * PageSize - 1 <= address_space_end - 1);
+    constexpr bool ContainsPages(VAddr addr, size_t num_pages) const {
+        return (m_address_space_start <= addr) &&
+               (num_pages <= (m_address_space_end - m_address_space_start) / PageSize) &&
+               (addr + num_pages * PageSize - 1 <= m_address_space_end - 1);
     }
 
 private:
-    VAddr address_space_start{};
-    VAddr address_space_end{};
-    VAddr heap_region_start{};
-    VAddr heap_region_end{};
-    VAddr current_heap_end{};
-    VAddr alias_region_start{};
-    VAddr alias_region_end{};
-    VAddr stack_region_start{};
-    VAddr stack_region_end{};
-    VAddr kernel_map_region_start{};
-    VAddr kernel_map_region_end{};
-    VAddr code_region_start{};
-    VAddr code_region_end{};
-    VAddr alias_code_region_start{};
-    VAddr alias_code_region_end{};
+    VAddr m_address_space_start{};
+    VAddr m_address_space_end{};
+    VAddr m_heap_region_start{};
+    VAddr m_heap_region_end{};
+    VAddr m_current_heap_end{};
+    VAddr m_alias_region_start{};
+    VAddr m_alias_region_end{};
+    VAddr m_stack_region_start{};
+    VAddr m_stack_region_end{};
+    VAddr m_kernel_map_region_start{};
+    VAddr m_kernel_map_region_end{};
+    VAddr m_code_region_start{};
+    VAddr m_code_region_end{};
+    VAddr m_alias_code_region_start{};
+    VAddr m_alias_code_region_end{};
 
-    std::size_t mapped_physical_memory_size{};
-    std::size_t max_heap_size{};
-    std::size_t max_physical_memory_size{};
-    std::size_t address_space_width{};
+    size_t m_mapped_physical_memory_size{};
+    size_t m_max_heap_size{};
+    size_t m_max_physical_memory_size{};
+    size_t m_address_space_width{};
 
-    bool is_kernel{};
-    bool is_aslr_enabled{};
+    KMemoryBlockManager m_memory_block_manager;
 
-    u32 heap_fill_value{};
-    const KMemoryRegion* cached_physical_heap_region{};
+    bool m_is_kernel{};
+    bool m_enable_aslr{};
+    bool m_enable_device_address_space_merge{};
 
-    KMemoryManager::Pool memory_pool{KMemoryManager::Pool::Application};
-    KMemoryManager::Direction allocation_option{KMemoryManager::Direction::FromFront};
+    KMemoryBlockSlabManager* m_memory_block_slab_manager{};
 
-    Common::PageTable page_table_impl;
+    u32 m_heap_fill_value{};
+    const KMemoryRegion* m_cached_physical_heap_region{};
 
-    Core::System& system;
+    KMemoryManager::Pool m_memory_pool{KMemoryManager::Pool::Application};
+    KMemoryManager::Direction m_allocation_option{KMemoryManager::Direction::FromFront};
+
+    std::unique_ptr<Common::PageTable> m_page_table_impl;
+
+    Core::System& m_system;
 };
 
 } // namespace Kernel

src/core/hle/kernel/k_process.cpp

@@ -72,7 +72,8 @@ Result KProcess::Initialize(KProcess* process, Core::System& system, std::string
 
     process->name = std::move(process_name);
     process->resource_limit = res_limit;
-    process->status = ProcessStatus::Created;
+    process->system_resource_address = 0;
+    process->state = State::Created;
     process->program_id = 0;
     process->process_id = type == ProcessType::KernelInternal ? kernel.CreateNewKernelProcessID()
                                                               : kernel.CreateNewUserProcessID();
@@ -92,11 +93,12 @@ Result KProcess::Initialize(KProcess* process, Core::System& system, std::string
     process->exception_thread = nullptr;
     process->is_suspended = false;
     process->schedule_count = 0;
+    process->is_handle_table_initialized = false;
 
     // Open a reference to the resource limit.
     process->resource_limit->Open();
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 void KProcess::DoWorkerTaskImpl() {
@@ -121,9 +123,9 @@ void KProcess::DecrementRunningThreadCount() {
     }
 }
 
-u64 KProcess::GetTotalPhysicalMemoryAvailable() const {
+u64 KProcess::GetTotalPhysicalMemoryAvailable() {
     const u64 capacity{resource_limit->GetFreeValue(LimitableResource::PhysicalMemory) +
-                       page_table->GetNormalMemorySize() + GetSystemResourceSize() + image_size +
+                       page_table.GetNormalMemorySize() + GetSystemResourceSize() + image_size +
                        main_thread_stack_size};
     if (const auto pool_size = kernel.MemoryManager().GetSize(KMemoryManager::Pool::Application);
         capacity != pool_size) {
@@ -135,16 +137,16 @@ u64 KProcess::GetTotalPhysicalMemoryAvailable() const {
     return memory_usage_capacity;
 }
 
-u64 KProcess::GetTotalPhysicalMemoryAvailableWithoutSystemResource() const {
+u64 KProcess::GetTotalPhysicalMemoryAvailableWithoutSystemResource() {
     return GetTotalPhysicalMemoryAvailable() - GetSystemResourceSize();
 }
 
-u64 KProcess::GetTotalPhysicalMemoryUsed() const {
-    return image_size + main_thread_stack_size + page_table->GetNormalMemorySize() +
+u64 KProcess::GetTotalPhysicalMemoryUsed() {
+    return image_size + main_thread_stack_size + page_table.GetNormalMemorySize() +
            GetSystemResourceSize();
 }
 
-u64 KProcess::GetTotalPhysicalMemoryUsedWithoutSystemResource() const {
+u64 KProcess::GetTotalPhysicalMemoryUsedWithoutSystemResource() {
     return GetTotalPhysicalMemoryUsed() - GetSystemResourceUsage();
 }
 
@@ -244,7 +246,7 @@ Result KProcess::AddSharedMemory(KSharedMemory* shmem, [[maybe_unused]] VAddr ad
     shmem->Open();
     shemen_info->Open();
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 void KProcess::RemoveSharedMemory(KSharedMemory* shmem, [[maybe_unused]] VAddr address,
@@ -289,12 +291,12 @@ Result KProcess::Reset() {
     KScopedSchedulerLock sl{kernel};
 
     // Validate that we're in a state that we can reset.
-    R_UNLESS(status != ProcessStatus::Exited, ResultInvalidState);
+    R_UNLESS(state != State::Terminated, ResultInvalidState);
     R_UNLESS(is_signaled, ResultInvalidState);
 
     // Clear signaled.
     is_signaled = false;
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 Result KProcess::SetActivity(ProcessActivity activity) {
@@ -304,15 +306,13 @@ Result KProcess::SetActivity(ProcessActivity activity) {
     KScopedSchedulerLock sl{kernel};
 
     // Validate our state.
-    R_UNLESS(status != ProcessStatus::Exiting, ResultInvalidState);
-    R_UNLESS(status != ProcessStatus::Exited, ResultInvalidState);
+    R_UNLESS(state != State::Terminating, ResultInvalidState);
+    R_UNLESS(state != State::Terminated, ResultInvalidState);
 
     // Either pause or resume.
     if (activity == ProcessActivity::Paused) {
         // Verify that we're not suspended.
-        if (is_suspended) {
-            return ResultInvalidState;
-        }
+        R_UNLESS(!is_suspended, ResultInvalidState);
 
         // Suspend all threads.
         for (auto* thread : GetThreadList()) {
@@ -325,9 +325,7 @@ Result KProcess::SetActivity(ProcessActivity activity) {
         ASSERT(activity == ProcessActivity::Runnable);
 
         // Verify that we're suspended.
-        if (!is_suspended) {
-            return ResultInvalidState;
-        }
+        R_UNLESS(is_suspended, ResultInvalidState);
 
         // Resume all threads.
         for (auto* thread : GetThreadList()) {
@@ -338,7 +336,7 @@ Result KProcess::SetActivity(ProcessActivity activity) {
         SetSuspended(false);
     }
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size) {
@@ -348,35 +346,38 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     system_resource_size = metadata.GetSystemResourceSize();
     image_size = code_size;
 
+    // We currently do not support process-specific system resource
+    UNIMPLEMENTED_IF(system_resource_size != 0);
+
     KScopedResourceReservation memory_reservation(resource_limit, LimitableResource::PhysicalMemory,
                                                   code_size + system_resource_size);
     if (!memory_reservation.Succeeded()) {
         LOG_ERROR(Kernel, "Could not reserve process memory requirements of size {:X} bytes",
                   code_size + system_resource_size);
-        return ResultLimitReached;
+        R_RETURN(ResultLimitReached);
     }
     // Initialize proces address space
-    if (const Result result{page_table->InitializeForProcess(metadata.GetAddressSpaceType(), false,
-                                                             0x8000000, code_size,
-                                                             KMemoryManager::Pool::Application)};
+    if (const Result result{page_table.InitializeForProcess(
+            metadata.GetAddressSpaceType(), false, 0x8000000, code_size,
+            &kernel.GetApplicationMemoryBlockManager(), KMemoryManager::Pool::Application)};
         result.IsError()) {
-        return result;
+        R_RETURN(result);
     }
 
     // Map process code region
-    if (const Result result{page_table->MapProcessCode(page_table->GetCodeRegionStart(),
-                                                       code_size / PageSize, KMemoryState::Code,
-                                                       KMemoryPermission::None)};
+    if (const Result result{page_table.MapProcessCode(page_table.GetCodeRegionStart(),
+                                                      code_size / PageSize, KMemoryState::Code,
+                                                      KMemoryPermission::None)};
         result.IsError()) {
-        return result;
+        R_RETURN(result);
     }
 
     // Initialize process capabilities
     const auto& caps{metadata.GetKernelCapabilities()};
     if (const Result result{
-            capabilities.InitializeForUserProcess(caps.data(), caps.size(), *page_table)};
+            capabilities.InitializeForUserProcess(caps.data(), caps.size(), page_table)};
         result.IsError()) {
-        return result;
+        R_RETURN(result);
     }
 
     // Set memory usage capacity
@@ -384,12 +385,12 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     case FileSys::ProgramAddressSpaceType::Is32Bit:
     case FileSys::ProgramAddressSpaceType::Is36Bit:
     case FileSys::ProgramAddressSpaceType::Is39Bit:
-        memory_usage_capacity = page_table->GetHeapRegionEnd() - page_table->GetHeapRegionStart();
+        memory_usage_capacity = page_table.GetHeapRegionEnd() - page_table.GetHeapRegionStart();
         break;
 
     case FileSys::ProgramAddressSpaceType::Is32BitNoMap:
-        memory_usage_capacity = page_table->GetHeapRegionEnd() - page_table->GetHeapRegionStart() +
-                                page_table->GetAliasRegionEnd() - page_table->GetAliasRegionStart();
+        memory_usage_capacity = page_table.GetHeapRegionEnd() - page_table.GetHeapRegionStart() +
+                                page_table.GetAliasRegionEnd() - page_table.GetAliasRegionStart();
         break;
 
     default:
@@ -397,10 +398,10 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     }
 
     // Create TLS region
-    R_TRY(this->CreateThreadLocalRegion(std::addressof(tls_region_address)));
+    R_TRY(this->CreateThreadLocalRegion(std::addressof(plr_address)));
     memory_reservation.Commit();
 
-    return handle_table.Initialize(capabilities.GetHandleTableSize());
+    R_RETURN(handle_table.Initialize(capabilities.GetHandleTableSize()));
 }
 
 void KProcess::Run(s32 main_thread_priority, u64 stack_size) {
@@ -409,15 +410,15 @@ void KProcess::Run(s32 main_thread_priority, u64 stack_size) {
     resource_limit->Reserve(LimitableResource::PhysicalMemory, main_thread_stack_size);
 
     const std::size_t heap_capacity{memory_usage_capacity - (main_thread_stack_size + image_size)};
-    ASSERT(!page_table->SetMaxHeapSize(heap_capacity).IsError());
+    ASSERT(!page_table.SetMaxHeapSize(heap_capacity).IsError());
 
-    ChangeStatus(ProcessStatus::Running);
+    ChangeState(State::Running);
 
     SetupMainThread(kernel.System(), *this, main_thread_priority, main_thread_stack_top);
 }
 
 void KProcess::PrepareForTermination() {
-    ChangeStatus(ProcessStatus::Exiting);
+    ChangeState(State::Terminating);
 
     const auto stop_threads = [this](const std::vector<KThread*>& in_thread_list) {
         for (auto* thread : in_thread_list) {
@@ -437,15 +438,15 @@ void KProcess::PrepareForTermination() {
 
     stop_threads(kernel.System().GlobalSchedulerContext().GetThreadList());
 
-    this->DeleteThreadLocalRegion(tls_region_address);
-    tls_region_address = 0;
+    this->DeleteThreadLocalRegion(plr_address);
+    plr_address = 0;
 
     if (resource_limit) {
         resource_limit->Release(LimitableResource::PhysicalMemory,
                                 main_thread_stack_size + image_size);
     }
 
-    ChangeStatus(ProcessStatus::Exited);
+    ChangeState(State::Terminated);
 }
 
 void KProcess::Finalize() {
@@ -474,7 +475,7 @@ void KProcess::Finalize() {
     }
 
     // Finalize the page table.
-    page_table.reset();
+    page_table.Finalize();
 
     // Perform inherited finalization.
     KAutoObjectWithSlabHeapAndContainer<KProcess, KWorkerTask>::Finalize();
@@ -499,7 +500,7 @@ Result KProcess::CreateThreadLocalRegion(VAddr* out) {
             }
 
             *out = tlr;
-            return ResultSuccess;
+            R_SUCCEED();
         }
     }
 
@@ -528,7 +529,7 @@ Result KProcess::CreateThreadLocalRegion(VAddr* out) {
     // We succeeded!
     tlp_guard.Cancel();
     *out = tlr;
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 Result KProcess::DeleteThreadLocalRegion(VAddr addr) {
@@ -576,7 +577,7 @@ Result KProcess::DeleteThreadLocalRegion(VAddr addr) {
         KThreadLocalPage::Free(kernel, page_to_free);
     }
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 bool KProcess::InsertWatchpoint(Core::System& system, VAddr addr, u64 size,
@@ -628,7 +629,7 @@ bool KProcess::RemoveWatchpoint(Core::System& system, VAddr addr, u64 size,
 void KProcess::LoadModule(CodeSet code_set, VAddr base_addr) {
     const auto ReprotectSegment = [&](const CodeSet::Segment& segment,
                                       Svc::MemoryPermission permission) {
-        page_table->SetProcessMemoryPermission(segment.addr + base_addr, segment.size, permission);
+        page_table.SetProcessMemoryPermission(segment.addr + base_addr, segment.size, permission);
     };
 
     kernel.System().Memory().WriteBlock(*this, base_addr, code_set.memory.data(),
@@ -645,19 +646,18 @@ bool KProcess::IsSignaled() const {
 }
 
 KProcess::KProcess(KernelCore& kernel_)
-    : KAutoObjectWithSlabHeapAndContainer{kernel_}, page_table{std::make_unique<KPageTable>(
-                                                        kernel_.System())},
+    : KAutoObjectWithSlabHeapAndContainer{kernel_}, page_table{kernel_.System()},
       handle_table{kernel_}, address_arbiter{kernel_.System()}, condition_var{kernel_.System()},
       state_lock{kernel_}, list_lock{kernel_} {}
 
 KProcess::~KProcess() = default;
 
-void KProcess::ChangeStatus(ProcessStatus new_status) {
-    if (status == new_status) {
+void KProcess::ChangeState(State new_state) {
+    if (state == new_state) {
         return;
     }
 
-    status = new_status;
+    state = new_state;
     is_signaled = true;
     NotifyAvailable();
 }
@@ -668,17 +668,17 @@ Result KProcess::AllocateMainThreadStack(std::size_t stack_size) {
     // The kernel always ensures that the given stack size is page aligned.
     main_thread_stack_size = Common::AlignUp(stack_size, PageSize);
 
-    const VAddr start{page_table->GetStackRegionStart()};
-    const std::size_t size{page_table->GetStackRegionEnd() - start};
+    const VAddr start{page_table.GetStackRegionStart()};
+    const std::size_t size{page_table.GetStackRegionEnd() - start};
 
     CASCADE_RESULT(main_thread_stack_top,
-                   page_table->AllocateAndMapMemory(
+                   page_table.AllocateAndMapMemory(
                        main_thread_stack_size / PageSize, PageSize, false, start, size / PageSize,
                        KMemoryState::Stack, KMemoryPermission::UserReadWrite));
 
     main_thread_stack_top += main_thread_stack_size;
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_process.h

@@ -13,6 +13,7 @@
 #include "core/hle/kernel/k_auto_object.h"
 #include "core/hle/kernel/k_condition_variable.h"
 #include "core/hle/kernel/k_handle_table.h"
+#include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_synchronization_object.h"
 #include "core/hle/kernel/k_thread_local_page.h"
 #include "core/hle/kernel/k_worker_task.h"
@@ -31,7 +32,6 @@ class ProgramMetadata;
 namespace Kernel {
 
 class KernelCore;
-class KPageTable;
 class KResourceLimit;
 class KThread;
 class KSharedMemoryInfo;
@@ -45,24 +45,6 @@ enum class MemoryRegion : u16 {
     BASE = 3,
 };
 
-/**
- * Indicates the status of a Process instance.
- *
- * @note These match the values as used by kernel,
- *       so new entries should only be added if RE
- *       shows that a new value has been introduced.
- */
-enum class ProcessStatus {
-    Created,
-    CreatedWithDebuggerAttached,
-    Running,
-    WaitingForDebuggerToAttach,
-    DebuggerAttached,
-    Exiting,
-    Exited,
-    DebugBreak,
-};
-
 enum class ProcessActivity : u32 {
     Runnable,
     Paused,
@@ -89,6 +71,17 @@ public:
     explicit KProcess(KernelCore& kernel_);
     ~KProcess() override;
 
+    enum class State {
+        Created = static_cast<u32>(Svc::ProcessState::Created),
+        CreatedAttached = static_cast<u32>(Svc::ProcessState::CreatedAttached),
+        Running = static_cast<u32>(Svc::ProcessState::Running),
+        Crashed = static_cast<u32>(Svc::ProcessState::Crashed),
+        RunningAttached = static_cast<u32>(Svc::ProcessState::RunningAttached),
+        Terminating = static_cast<u32>(Svc::ProcessState::Terminating),
+        Terminated = static_cast<u32>(Svc::ProcessState::Terminated),
+        DebugBreak = static_cast<u32>(Svc::ProcessState::DebugBreak),
+    };
+
     enum : u64 {
         /// Lowest allowed process ID for a kernel initial process.
         InitialKIPIDMin = 1,
@@ -114,12 +107,12 @@ public:
 
     /// Gets a reference to the process' page table.
     KPageTable& PageTable() {
-        return *page_table;
+        return page_table;
     }
 
     /// Gets const a reference to the process' page table.
     const KPageTable& PageTable() const {
-        return *page_table;
+        return page_table;
     }
 
     /// Gets a reference to the process' handle table.
@@ -145,26 +138,25 @@ public:
     }
 
     Result WaitConditionVariable(VAddr address, u64 cv_key, u32 tag, s64 ns) {
-        return condition_var.Wait(address, cv_key, tag, ns);
+        R_RETURN(condition_var.Wait(address, cv_key, tag, ns));
     }
 
     Result SignalAddressArbiter(VAddr address, Svc::SignalType signal_type, s32 value, s32 count) {
-        return address_arbiter.SignalToAddress(address, signal_type, value, count);
+        R_RETURN(address_arbiter.SignalToAddress(address, signal_type, value, count));
     }
 
     Result WaitAddressArbiter(VAddr address, Svc::ArbitrationType arb_type, s32 value,
                               s64 timeout) {
-        return address_arbiter.WaitForAddress(address, arb_type, value, timeout);
+        R_RETURN(address_arbiter.WaitForAddress(address, arb_type, value, timeout));
     }
 
-    /// Gets the address to the process' dedicated TLS region.
-    VAddr GetTLSRegionAddress() const {
-        return tls_region_address;
+    VAddr GetProcessLocalRegionAddress() const {
+        return plr_address;
     }
 
     /// Gets the current status of the process
-    ProcessStatus GetStatus() const {
-        return status;
+    State GetState() const {
+        return state;
     }
 
     /// Gets the unique ID that identifies this particular process.
@@ -286,18 +278,18 @@ public:
     }
 
     /// Retrieves the total physical memory available to this process in bytes.
-    u64 GetTotalPhysicalMemoryAvailable() const;
+    u64 GetTotalPhysicalMemoryAvailable();
 
     /// Retrieves the total physical memory available to this process in bytes,
     /// without the size of the personal system resource heap added to it.
-    u64 GetTotalPhysicalMemoryAvailableWithoutSystemResource() const;
+    u64 GetTotalPhysicalMemoryAvailableWithoutSystemResource();
 
     /// Retrieves the total physical memory used by this process in bytes.
-    u64 GetTotalPhysicalMemoryUsed() const;
+    u64 GetTotalPhysicalMemoryUsed();
 
     /// Retrieves the total physical memory used by this process in bytes,
     /// without the size of the personal system resource heap added to it.
-    u64 GetTotalPhysicalMemoryUsedWithoutSystemResource() const;
+    u64 GetTotalPhysicalMemoryUsedWithoutSystemResource();
 
     /// Gets the list of all threads created with this process as their owner.
     std::list<KThread*>& GetThreadList() {
@@ -415,19 +407,24 @@ private:
         pinned_threads[core_id] = nullptr;
     }
 
-    /// Changes the process status. If the status is different
-    /// from the current process status, then this will trigger
-    /// a process signal.
-    void ChangeStatus(ProcessStatus new_status);
+    void FinalizeHandleTable() {
+        // Finalize the table.
+        handle_table.Finalize();
+
+        // Note that the table is finalized.
+        is_handle_table_initialized = false;
+    }
+
+    void ChangeState(State new_state);
 
     /// Allocates the main thread stack for the process, given the stack size in bytes.
     Result AllocateMainThreadStack(std::size_t stack_size);
 
     /// Memory manager for this process
-    std::unique_ptr<KPageTable> page_table;
+    KPageTable page_table;
 
     /// Current status of the process
-    ProcessStatus status{};
+    State state{};
 
     /// The ID of this process
     u64 process_id = 0;
@@ -443,6 +440,8 @@ private:
     /// Resource limit descriptor for this process
     KResourceLimit* resource_limit{};
 
+    VAddr system_resource_address{};
+
     /// The ideal CPU core for this process, threads are scheduled on this core by default.
     u8 ideal_core = 0;
 
@@ -469,7 +468,7 @@ private:
     KConditionVariable condition_var;
 
     /// Address indicating the location of the process' dedicated TLS region.
-    VAddr tls_region_address = 0;
+    VAddr plr_address = 0;
 
     /// Random values for svcGetInfo RandomEntropy
     std::array<u64, RANDOM_ENTROPY_SIZE> random_entropy{};
@@ -495,8 +494,12 @@ private:
     /// Schedule count of this process
     s64 schedule_count{};
 
+    size_t memory_release_hint{};
+
     bool is_signaled{};
     bool is_suspended{};
+    bool is_immortal{};
+    bool is_handle_table_initialized{};
     bool is_initialized{};
 
     std::atomic<u16> num_running_threads{};

src/core/hle/kernel/k_server_session.cpp

@@ -22,15 +22,12 @@
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/k_thread_queue.h"
 #include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/service_thread.h"
 #include "core/memory.h"
 
 namespace Kernel {
 
 using ThreadQueueImplForKServerSessionRequest = KThreadQueue;
 
-static constexpr u32 MessageBufferSize = 0x100;
-
 KServerSession::KServerSession(KernelCore& kernel_)
     : KSynchronizationObject{kernel_}, m_lock{kernel_} {}
 
@@ -73,59 +70,7 @@ bool KServerSession::IsSignaled() const {
     }
 
     // Otherwise, we're signaled if we have a request and aren't handling one.
-    return !m_thread_request_list.empty() && m_current_thread_request == nullptr;
-}
-
-void KServerSession::AppendDomainHandler(SessionRequestHandlerPtr handler) {
-    manager->AppendDomainHandler(std::move(handler));
-}
-
-std::size_t KServerSession::NumDomainRequestHandlers() const {
-    return manager->DomainHandlerCount();
-}
-
-Result KServerSession::HandleDomainSyncRequest(Kernel::HLERequestContext& context) {
-    if (!context.HasDomainMessageHeader()) {
-        return ResultSuccess;
-    }
-
-    // Set domain handlers in HLE context, used for domain objects (IPC interfaces) as inputs
-    context.SetSessionRequestManager(manager);
-
-    // If there is a DomainMessageHeader, then this is CommandType "Request"
-    const auto& domain_message_header = context.GetDomainMessageHeader();
-    const u32 object_id{domain_message_header.object_id};
-    switch (domain_message_header.command) {
-    case IPC::DomainMessageHeader::CommandType::SendMessage:
-        if (object_id > manager->DomainHandlerCount()) {
-            LOG_CRITICAL(IPC,
-                         "object_id {} is too big! This probably means a recent service call "
-                         "to {} needed to return a new interface!",
-                         object_id, name);
-            ASSERT(false);
-            return ResultSuccess; // Ignore error if asserts are off
-        }
-        if (auto strong_ptr = manager->DomainHandler(object_id - 1).lock()) {
-            return strong_ptr->HandleSyncRequest(*this, context);
-        } else {
-            ASSERT(false);
-            return ResultSuccess;
-        }
-
-    case IPC::DomainMessageHeader::CommandType::CloseVirtualHandle: {
-        LOG_DEBUG(IPC, "CloseVirtualHandle, object_id=0x{:08X}", object_id);
-
-        manager->CloseDomainHandler(object_id - 1);
-
-        IPC::ResponseBuilder rb{context, 2};
-        rb.Push(ResultSuccess);
-        return ResultSuccess;
-    }
-    }
-
-    LOG_CRITICAL(IPC, "Unknown domain command={}", domain_message_header.command.Value());
-    ASSERT(false);
-    return ResultSuccess;
+    return !m_request_list.empty() && m_current_request == nullptr;
 }
 
 Result KServerSession::QueueSyncRequest(KThread* thread, Core::Memory::Memory& memory) {
@@ -134,43 +79,11 @@ Result KServerSession::QueueSyncRequest(KThread* thread, Core::Memory::Memory& m
 
     context->PopulateFromIncomingCommandBuffer(kernel.CurrentProcess()->GetHandleTable(), cmd_buf);
 
-    // Ensure we have a session request handler
-    if (manager->HasSessionRequestHandler(*context)) {
-        if (auto strong_ptr = manager->GetServiceThread().lock()) {
-            strong_ptr->QueueSyncRequest(*parent, std::move(context));
-        } else {
-            ASSERT_MSG(false, "strong_ptr is nullptr!");
-        }
-    } else {
-        ASSERT_MSG(false, "handler is invalid!");
-    }
-
-    return ResultSuccess;
+    return manager->QueueSyncRequest(parent, std::move(context));
 }
 
 Result KServerSession::CompleteSyncRequest(HLERequestContext& context) {
-    Result result = ResultSuccess;
-
-    // If the session has been converted to a domain, handle the domain request
-    if (manager->HasSessionRequestHandler(context)) {
-        if (IsDomain() && context.HasDomainMessageHeader()) {
-            result = HandleDomainSyncRequest(context);
-            // If there is no domain header, the regular session handler is used
-        } else if (manager->HasSessionHandler()) {
-            // If this ServerSession has an associated HLE handler, forward the request to it.
-            result = manager->SessionHandler().HandleSyncRequest(*this, context);
-        }
-    } else {
-        ASSERT_MSG(false, "Session handler is invalid, stubbing response!");
-        IPC::ResponseBuilder rb(context, 2);
-        rb.Push(ResultSuccess);
-    }
-
-    if (convert_to_domain) {
-        ASSERT_MSG(!IsDomain(), "ServerSession is already a domain instance.");
-        manager->ConvertToDomain();
-        convert_to_domain = false;
-    }
+    Result result = manager->CompleteSyncRequest(this, context);
 
     // The calling thread is waiting for this request to complete, so wake it up.
     context.GetThread().EndWait(result);
@@ -178,7 +91,7 @@ Result KServerSession::CompleteSyncRequest(HLERequestContext& context) {
     return result;
 }
 
-Result KServerSession::OnRequest() {
+Result KServerSession::OnRequest(KSessionRequest* request) {
     // Create the wait queue.
     ThreadQueueImplForKServerSessionRequest wait_queue{kernel};
 
@@ -198,14 +111,13 @@ Result KServerSession::OnRequest() {
             this->QueueSyncRequest(GetCurrentThreadPointer(kernel), memory);
         } else {
             // Non-HLE request.
-            auto* thread{GetCurrentThreadPointer(kernel)};
 
             // Get whether we're empty.
-            const bool was_empty = m_thread_request_list.empty();
+            const bool was_empty = m_request_list.empty();
 
-            // Add the thread to the list.
-            thread->Open();
-            m_thread_request_list.push_back(thread);
+            // Add the request to the list.
+            request->Open();
+            m_request_list.push_back(*request);
 
             // If we were empty, signal.
             if (was_empty) {
@@ -213,6 +125,9 @@ Result KServerSession::OnRequest() {
             }
         }
 
+        // If we have a request event, this is asynchronous, and we don't need to wait.
+        R_SUCCEED_IF(request->GetEvent() != nullptr);
+
         // This is a synchronous request, so we should wait for our request to complete.
         GetCurrentThread(kernel).SetWaitReasonForDebugging(ThreadWaitReasonForDebugging::IPC);
         GetCurrentThread(kernel).BeginWait(&wait_queue);
@@ -223,32 +138,32 @@ Result KServerSession::OnRequest() {
 
 Result KServerSession::SendReply() {
     // Lock the session.
-    KScopedLightLock lk(m_lock);
+    KScopedLightLock lk{m_lock};
 
     // Get the request.
-    KThread* client_thread;
+    KSessionRequest* request;
     {
         KScopedSchedulerLock sl{kernel};
 
         // Get the current request.
-        client_thread = m_current_thread_request;
-        R_UNLESS(client_thread != nullptr, ResultInvalidState);
+        request = m_current_request;
+        R_UNLESS(request != nullptr, ResultInvalidState);
 
         // Clear the current request, since we're processing it.
-        m_current_thread_request = nullptr;
-        if (!m_thread_request_list.empty()) {
+        m_current_request = nullptr;
+        if (!m_request_list.empty()) {
             this->NotifyAvailable();
         }
     }
 
     // Close reference to the request once we're done processing it.
-    SCOPE_EXIT({ client_thread->Close(); });
+    SCOPE_EXIT({ request->Close(); });
 
     // Extract relevant information from the request.
-    // const uintptr_t client_message  = request->GetAddress();
-    // const size_t client_buffer_size = request->GetSize();
-    // KThread *client_thread          = request->GetThread();
-    // KEvent *event                   = request->GetEvent();
+    const uintptr_t client_message = request->GetAddress();
+    const size_t client_buffer_size = request->GetSize();
+    KThread* client_thread = request->GetThread();
+    KEvent* event = request->GetEvent();
 
     // Check whether we're closed.
     const bool closed = (client_thread == nullptr || parent->IsClientClosed());
@@ -261,8 +176,8 @@ Result KServerSession::SendReply() {
         UNIMPLEMENTED_IF(server_thread->GetOwnerProcess() != client_thread->GetOwnerProcess());
 
         auto* src_msg_buffer = memory.GetPointer(server_thread->GetTLSAddress());
-        auto* dst_msg_buffer = memory.GetPointer(client_thread->GetTLSAddress());
-        std::memcpy(dst_msg_buffer, src_msg_buffer, MessageBufferSize);
+        auto* dst_msg_buffer = memory.GetPointer(client_message);
+        std::memcpy(dst_msg_buffer, src_msg_buffer, client_buffer_size);
     } else {
         result = ResultSessionClosed;
     }
@@ -278,11 +193,30 @@ Result KServerSession::SendReply() {
 
     // If there's a client thread, update it.
     if (client_thread != nullptr) {
-        // End the client thread's wait.
-        KScopedSchedulerLock sl{kernel};
+        if (event != nullptr) {
+            // // Get the client process/page table.
+            // KProcess *client_process             = client_thread->GetOwnerProcess();
+            // KPageTable *client_page_table        = &client_process->PageTable();
 
-        if (!client_thread->IsTerminationRequested()) {
-            client_thread->EndWait(client_result);
+            // // If we need to, reply with an async error.
+            // if (R_FAILED(client_result)) {
+            //     ReplyAsyncError(client_process, client_message, client_buffer_size,
+            //     client_result);
+            // }
+
+            // // Unlock the client buffer.
+            // // NOTE: Nintendo does not check the result of this.
+            // client_page_table->UnlockForIpcUserBuffer(client_message, client_buffer_size);
+
+            // Signal the event.
+            event->Signal();
+        } else {
+            // End the client thread's wait.
+            KScopedSchedulerLock sl{kernel};
+
+            if (!client_thread->IsTerminationRequested()) {
+                client_thread->EndWait(client_result);
+            }
         }
     }
 
@@ -291,10 +225,10 @@ Result KServerSession::SendReply() {
 
 Result KServerSession::ReceiveRequest() {
     // Lock the session.
-    KScopedLightLock lk(m_lock);
+    KScopedLightLock lk{m_lock};
 
     // Get the request and client thread.
-    // KSessionRequest *request;
+    KSessionRequest* request;
     KThread* client_thread;
 
     {
@@ -304,35 +238,41 @@ Result KServerSession::ReceiveRequest() {
         R_UNLESS(!parent->IsClientClosed(), ResultSessionClosed);
 
         // Ensure we aren't already servicing a request.
-        R_UNLESS(m_current_thread_request == nullptr, ResultNotFound);
+        R_UNLESS(m_current_request == nullptr, ResultNotFound);
 
         // Ensure we have a request to service.
-        R_UNLESS(!m_thread_request_list.empty(), ResultNotFound);
+        R_UNLESS(!m_request_list.empty(), ResultNotFound);
 
         // Pop the first request from the list.
-        client_thread = m_thread_request_list.front();
-        m_thread_request_list.pop_front();
+        request = &m_request_list.front();
+        m_request_list.pop_front();
 
         // Get the thread for the request.
+        client_thread = request->GetThread();
         R_UNLESS(client_thread != nullptr, ResultSessionClosed);
 
         // Open the client thread.
         client_thread->Open();
     }
 
-    // SCOPE_EXIT({ client_thread->Close(); });
+    SCOPE_EXIT({ client_thread->Close(); });
 
     // Set the request as our current.
-    m_current_thread_request = client_thread;
+    m_current_request = request;
+
+    // Get the client address.
+    uintptr_t client_message = request->GetAddress();
+    size_t client_buffer_size = request->GetSize();
+    // bool recv_list_broken = false;
 
     // Receive the message.
     Core::Memory::Memory& memory{kernel.System().Memory()};
     KThread* server_thread{GetCurrentThreadPointer(kernel)};
     UNIMPLEMENTED_IF(server_thread->GetOwnerProcess() != client_thread->GetOwnerProcess());
 
-    auto* src_msg_buffer = memory.GetPointer(client_thread->GetTLSAddress());
+    auto* src_msg_buffer = memory.GetPointer(client_message);
     auto* dst_msg_buffer = memory.GetPointer(server_thread->GetTLSAddress());
-    std::memcpy(dst_msg_buffer, src_msg_buffer, MessageBufferSize);
+    std::memcpy(dst_msg_buffer, src_msg_buffer, client_buffer_size);
 
     // We succeeded.
     return ResultSuccess;
@@ -344,35 +284,34 @@ void KServerSession::CleanupRequests() {
     // Clean up any pending requests.
     while (true) {
         // Get the next request.
-        // KSessionRequest *request = nullptr;
-        KThread* client_thread = nullptr;
+        KSessionRequest* request = nullptr;
         {
             KScopedSchedulerLock sl{kernel};
 
-            if (m_current_thread_request) {
+            if (m_current_request) {
                 // Choose the current request if we have one.
-                client_thread = m_current_thread_request;
-                m_current_thread_request = nullptr;
-            } else if (!m_thread_request_list.empty()) {
+                request = m_current_request;
+                m_current_request = nullptr;
+            } else if (!m_request_list.empty()) {
                 // Pop the request from the front of the list.
-                client_thread = m_thread_request_list.front();
-                m_thread_request_list.pop_front();
+                request = &m_request_list.front();
+                m_request_list.pop_front();
             }
         }
 
         // If there's no request, we're done.
-        if (client_thread == nullptr) {
+        if (request == nullptr) {
             break;
         }
 
         // Close a reference to the request once it's cleaned up.
-        SCOPE_EXIT({ client_thread->Close(); });
+        SCOPE_EXIT({ request->Close(); });
 
         // Extract relevant information from the request.
         // const uintptr_t client_message  = request->GetAddress();
         // const size_t client_buffer_size = request->GetSize();
-        // KThread *client_thread          = request->GetThread();
-        // KEvent *event                   = request->GetEvent();
+        KThread* client_thread = request->GetThread();
+        KEvent* event = request->GetEvent();
 
         // KProcess *server_process             = request->GetServerProcess();
         // KProcess *client_process             = (client_thread != nullptr) ?
@@ -385,11 +324,24 @@ void KServerSession::CleanupRequests() {
 
         // If there's a client thread, update it.
         if (client_thread != nullptr) {
-            // End the client thread's wait.
-            KScopedSchedulerLock sl{kernel};
+            if (event != nullptr) {
+                // // We need to reply async.
+                // ReplyAsyncError(client_process, client_message, client_buffer_size,
+                //                 (R_SUCCEEDED(result) ? ResultSessionClosed : result));
 
-            if (!client_thread->IsTerminationRequested()) {
-                client_thread->EndWait(ResultSessionClosed);
+                // // Unlock the client buffer.
+                // NOTE: Nintendo does not check the result of this.
+                // client_page_table->UnlockForIpcUserBuffer(client_message, client_buffer_size);
+
+                // Signal the event.
+                event->Signal();
+            } else {
+                // End the client thread's wait.
+                KScopedSchedulerLock sl{kernel};
+
+                if (!client_thread->IsTerminationRequested()) {
+                    client_thread->EndWait(ResultSessionClosed);
+                }
             }
         }
     }

src/core/hle/kernel/k_server_session.h

@@ -12,6 +12,7 @@
 
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/kernel/k_light_lock.h"
+#include "core/hle/kernel/k_session_request.h"
 #include "core/hle/kernel/k_synchronization_object.h"
 #include "core/hle/result.h"
 
@@ -57,44 +58,15 @@ public:
     }
 
     bool IsSignaled() const override;
-
     void OnClientClosed();
 
-    void ClientConnected(SessionRequestHandlerPtr handler) {
-        if (manager) {
-            manager->SetSessionHandler(std::move(handler));
-        }
-    }
-
-    void ClientDisconnected() {
-        manager = nullptr;
-    }
-
-    /// Adds a new domain request handler to the collection of request handlers within
-    /// this ServerSession instance.
-    void AppendDomainHandler(SessionRequestHandlerPtr handler);
-
-    /// Retrieves the total number of domain request handlers that have been
-    /// appended to this ServerSession instance.
-    std::size_t NumDomainRequestHandlers() const;
-
-    /// Returns true if the session has been converted to a domain, otherwise False
-    bool IsDomain() const {
-        return manager && manager->IsDomain();
-    }
-
-    /// Converts the session to a domain at the end of the current command
-    void ConvertToDomain() {
-        convert_to_domain = true;
-    }
-
     /// Gets the session request manager, which forwards requests to the underlying service
     std::shared_ptr<SessionRequestManager>& GetSessionRequestManager() {
         return manager;
     }
 
     /// TODO: flesh these out to match the real kernel
-    Result OnRequest();
+    Result OnRequest(KSessionRequest* request);
     Result SendReply();
     Result ReceiveRequest();
 
@@ -108,10 +80,6 @@ private:
     /// Completes a sync request from the emulated application.
     Result CompleteSyncRequest(HLERequestContext& context);
 
-    /// Handles a SyncRequest to a domain, forwarding the request to the proper object or closing an
-    /// object handle.
-    Result HandleDomainSyncRequest(Kernel::HLERequestContext& context);
-
     /// This session's HLE request handlers; if nullptr, this is not an HLE server
     std::shared_ptr<SessionRequestManager> manager;
 
@@ -122,9 +90,8 @@ private:
     KSession* parent{};
 
     /// List of threads which are pending a reply.
-    /// FIXME: KSessionRequest
-    std::list<KThread*> m_thread_request_list;
-    KThread* m_current_thread_request{};
+    boost::intrusive::list<KSessionRequest> m_request_list;
+    KSessionRequest* m_current_request;
 
     KLightLock m_lock;
 };

src/core/hle/kernel/k_session_request.cpp

@@ -0,0 +1,61 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/hle/kernel/k_page_buffer.h"
+#include "core/hle/kernel/k_session_request.h"
+
+namespace Kernel {
+
+Result KSessionRequest::SessionMappings::PushMap(VAddr client, VAddr server, size_t size,
+                                                 KMemoryState state, size_t index) {
+    // At most 15 buffers of each type (4-bit descriptor counts).
+    ASSERT(index < ((1ul << 4) - 1) * 3);
+
+    // Get the mapping.
+    Mapping* mapping;
+    if (index < NumStaticMappings) {
+        mapping = &m_static_mappings[index];
+    } else {
+        // Allocate a page for the extra mappings.
+        if (m_mappings == nullptr) {
+            KPageBuffer* page_buffer = KPageBuffer::Allocate(kernel);
+            R_UNLESS(page_buffer != nullptr, ResultOutOfMemory);
+
+            m_mappings = reinterpret_cast<Mapping*>(page_buffer);
+        }
+
+        mapping = &m_mappings[index - NumStaticMappings];
+    }
+
+    // Set the mapping.
+    mapping->Set(client, server, size, state);
+
+    return ResultSuccess;
+}
+
+Result KSessionRequest::SessionMappings::PushSend(VAddr client, VAddr server, size_t size,
+                                                  KMemoryState state) {
+    ASSERT(m_num_recv == 0);
+    ASSERT(m_num_exch == 0);
+    return this->PushMap(client, server, size, state, m_num_send++);
+}
+
+Result KSessionRequest::SessionMappings::PushReceive(VAddr client, VAddr server, size_t size,
+                                                     KMemoryState state) {
+    ASSERT(m_num_exch == 0);
+    return this->PushMap(client, server, size, state, m_num_send + m_num_recv++);
+}
+
+Result KSessionRequest::SessionMappings::PushExchange(VAddr client, VAddr server, size_t size,
+                                                      KMemoryState state) {
+    return this->PushMap(client, server, size, state, m_num_send + m_num_recv + m_num_exch++);
+}
+
+void KSessionRequest::SessionMappings::Finalize() {
+    if (m_mappings) {
+        KPageBuffer::Free(kernel, reinterpret_cast<KPageBuffer*>(m_mappings));
+        m_mappings = nullptr;
+    }
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_session_request.h

@@ -0,0 +1,307 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "core/hle/kernel/k_auto_object.h"
+#include "core/hle/kernel/k_event.h"
+#include "core/hle/kernel/k_memory_block.h"
+#include "core/hle/kernel/k_process.h"
+#include "core/hle/kernel/k_thread.h"
+#include "core/hle/kernel/slab_helpers.h"
+
+namespace Kernel {
+
+class KSessionRequest final : public KSlabAllocated<KSessionRequest>,
+                              public KAutoObject,
+                              public boost::intrusive::list_base_hook<> {
+    KERNEL_AUTOOBJECT_TRAITS(KSessionRequest, KAutoObject);
+
+public:
+    class SessionMappings {
+    private:
+        static constexpr size_t NumStaticMappings = 8;
+
+        class Mapping {
+        public:
+            constexpr void Set(VAddr c, VAddr s, size_t sz, KMemoryState st) {
+                m_client_address = c;
+                m_server_address = s;
+                m_size = sz;
+                m_state = st;
+            }
+
+            constexpr VAddr GetClientAddress() const {
+                return m_client_address;
+            }
+            constexpr VAddr GetServerAddress() const {
+                return m_server_address;
+            }
+            constexpr size_t GetSize() const {
+                return m_size;
+            }
+            constexpr KMemoryState GetMemoryState() const {
+                return m_state;
+            }
+
+        private:
+            VAddr m_client_address;
+            VAddr m_server_address;
+            size_t m_size;
+            KMemoryState m_state;
+        };
+
+    public:
+        explicit SessionMappings(KernelCore& kernel_)
+            : kernel(kernel_), m_mappings(nullptr), m_num_send(), m_num_recv(), m_num_exch() {}
+
+        void Initialize() {}
+        void Finalize();
+
+        size_t GetSendCount() const {
+            return m_num_send;
+        }
+        size_t GetReceiveCount() const {
+            return m_num_recv;
+        }
+        size_t GetExchangeCount() const {
+            return m_num_exch;
+        }
+
+        Result PushSend(VAddr client, VAddr server, size_t size, KMemoryState state);
+        Result PushReceive(VAddr client, VAddr server, size_t size, KMemoryState state);
+        Result PushExchange(VAddr client, VAddr server, size_t size, KMemoryState state);
+
+        VAddr GetSendClientAddress(size_t i) const {
+            return GetSendMapping(i).GetClientAddress();
+        }
+        VAddr GetSendServerAddress(size_t i) const {
+            return GetSendMapping(i).GetServerAddress();
+        }
+        size_t GetSendSize(size_t i) const {
+            return GetSendMapping(i).GetSize();
+        }
+        KMemoryState GetSendMemoryState(size_t i) const {
+            return GetSendMapping(i).GetMemoryState();
+        }
+
+        VAddr GetReceiveClientAddress(size_t i) const {
+            return GetReceiveMapping(i).GetClientAddress();
+        }
+        VAddr GetReceiveServerAddress(size_t i) const {
+            return GetReceiveMapping(i).GetServerAddress();
+        }
+        size_t GetReceiveSize(size_t i) const {
+            return GetReceiveMapping(i).GetSize();
+        }
+        KMemoryState GetReceiveMemoryState(size_t i) const {
+            return GetReceiveMapping(i).GetMemoryState();
+        }
+
+        VAddr GetExchangeClientAddress(size_t i) const {
+            return GetExchangeMapping(i).GetClientAddress();
+        }
+        VAddr GetExchangeServerAddress(size_t i) const {
+            return GetExchangeMapping(i).GetServerAddress();
+        }
+        size_t GetExchangeSize(size_t i) const {
+            return GetExchangeMapping(i).GetSize();
+        }
+        KMemoryState GetExchangeMemoryState(size_t i) const {
+            return GetExchangeMapping(i).GetMemoryState();
+        }
+
+    private:
+        Result PushMap(VAddr client, VAddr server, size_t size, KMemoryState state, size_t index);
+
+        const Mapping& GetSendMapping(size_t i) const {
+            ASSERT(i < m_num_send);
+
+            const size_t index = i;
+            if (index < NumStaticMappings) {
+                return m_static_mappings[index];
+            } else {
+                return m_mappings[index - NumStaticMappings];
+            }
+        }
+
+        const Mapping& GetReceiveMapping(size_t i) const {
+            ASSERT(i < m_num_recv);
+
+            const size_t index = m_num_send + i;
+            if (index < NumStaticMappings) {
+                return m_static_mappings[index];
+            } else {
+                return m_mappings[index - NumStaticMappings];
+            }
+        }
+
+        const Mapping& GetExchangeMapping(size_t i) const {
+            ASSERT(i < m_num_exch);
+
+            const size_t index = m_num_send + m_num_recv + i;
+            if (index < NumStaticMappings) {
+                return m_static_mappings[index];
+            } else {
+                return m_mappings[index - NumStaticMappings];
+            }
+        }
+
+    private:
+        KernelCore& kernel;
+        Mapping m_static_mappings[NumStaticMappings];
+        Mapping* m_mappings;
+        u8 m_num_send;
+        u8 m_num_recv;
+        u8 m_num_exch;
+    };
+
+public:
+    explicit KSessionRequest(KernelCore& kernel_)
+        : KAutoObject(kernel_), m_mappings(kernel_), m_thread(nullptr), m_server(nullptr),
+          m_event(nullptr) {}
+
+    static KSessionRequest* Create(KernelCore& kernel) {
+        KSessionRequest* req = KSessionRequest::Allocate(kernel);
+        if (req != nullptr) [[likely]] {
+            KAutoObject::Create(req);
+        }
+        return req;
+    }
+
+    void Destroy() override {
+        this->Finalize();
+        KSessionRequest::Free(kernel, this);
+    }
+
+    void Initialize(KEvent* event, uintptr_t address, size_t size) {
+        m_mappings.Initialize();
+
+        m_thread = GetCurrentThreadPointer(kernel);
+        m_event = event;
+        m_address = address;
+        m_size = size;
+
+        m_thread->Open();
+        if (m_event != nullptr) {
+            m_event->Open();
+        }
+    }
+
+    static void PostDestroy(uintptr_t arg) {}
+
+    KThread* GetThread() const {
+        return m_thread;
+    }
+    KEvent* GetEvent() const {
+        return m_event;
+    }
+    uintptr_t GetAddress() const {
+        return m_address;
+    }
+    size_t GetSize() const {
+        return m_size;
+    }
+    KProcess* GetServerProcess() const {
+        return m_server;
+    }
+
+    void SetServerProcess(KProcess* process) {
+        m_server = process;
+        m_server->Open();
+    }
+
+    void ClearThread() {
+        m_thread = nullptr;
+    }
+    void ClearEvent() {
+        m_event = nullptr;
+    }
+
+    size_t GetSendCount() const {
+        return m_mappings.GetSendCount();
+    }
+    size_t GetReceiveCount() const {
+        return m_mappings.GetReceiveCount();
+    }
+    size_t GetExchangeCount() const {
+        return m_mappings.GetExchangeCount();
+    }
+
+    Result PushSend(VAddr client, VAddr server, size_t size, KMemoryState state) {
+        return m_mappings.PushSend(client, server, size, state);
+    }
+
+    Result PushReceive(VAddr client, VAddr server, size_t size, KMemoryState state) {
+        return m_mappings.PushReceive(client, server, size, state);
+    }
+
+    Result PushExchange(VAddr client, VAddr server, size_t size, KMemoryState state) {
+        return m_mappings.PushExchange(client, server, size, state);
+    }
+
+    VAddr GetSendClientAddress(size_t i) const {
+        return m_mappings.GetSendClientAddress(i);
+    }
+    VAddr GetSendServerAddress(size_t i) const {
+        return m_mappings.GetSendServerAddress(i);
+    }
+    size_t GetSendSize(size_t i) const {
+        return m_mappings.GetSendSize(i);
+    }
+    KMemoryState GetSendMemoryState(size_t i) const {
+        return m_mappings.GetSendMemoryState(i);
+    }
+
+    VAddr GetReceiveClientAddress(size_t i) const {
+        return m_mappings.GetReceiveClientAddress(i);
+    }
+    VAddr GetReceiveServerAddress(size_t i) const {
+        return m_mappings.GetReceiveServerAddress(i);
+    }
+    size_t GetReceiveSize(size_t i) const {
+        return m_mappings.GetReceiveSize(i);
+    }
+    KMemoryState GetReceiveMemoryState(size_t i) const {
+        return m_mappings.GetReceiveMemoryState(i);
+    }
+
+    VAddr GetExchangeClientAddress(size_t i) const {
+        return m_mappings.GetExchangeClientAddress(i);
+    }
+    VAddr GetExchangeServerAddress(size_t i) const {
+        return m_mappings.GetExchangeServerAddress(i);
+    }
+    size_t GetExchangeSize(size_t i) const {
+        return m_mappings.GetExchangeSize(i);
+    }
+    KMemoryState GetExchangeMemoryState(size_t i) const {
+        return m_mappings.GetExchangeMemoryState(i);
+    }
+
+private:
+    // NOTE: This is public and virtual in Nintendo's kernel.
+    void Finalize() {
+        m_mappings.Finalize();
+
+        if (m_thread) {
+            m_thread->Close();
+        }
+        if (m_event) {
+            m_event->Close();
+        }
+        if (m_server) {
+            m_server->Close();
+        }
+    }
+
+private:
+    SessionMappings m_mappings;
+    KThread* m_thread;
+    KProcess* m_server;
+    KEvent* m_event;
+    uintptr_t m_address;
+    size_t m_size;
+};
+
+} // namespace Kernel

src/core/hle/kernel/k_shared_memory.cpp

@@ -50,7 +50,7 @@ Result KSharedMemory::Initialize(Core::DeviceMemory& device_memory_, KProcess* o
     is_initialized = true;
 
     // Clear all pages in the memory.
-    std::memset(device_memory_.GetPointer(physical_address_), 0, size_);
+    std::memset(device_memory_.GetPointer<void>(physical_address_), 0, size_);
 
     return ResultSuccess;
 }

src/core/hle/kernel/k_shared_memory.h

@@ -54,7 +54,7 @@ public:
      * @return A pointer to the shared memory block from the specified offset
      */
     u8* GetPointer(std::size_t offset = 0) {
-        return device_memory->GetPointer(physical_address + offset);
+        return device_memory->GetPointer<u8>(physical_address + offset);
     }
 
     /**
@@ -63,7 +63,7 @@ public:
      * @return A pointer to the shared memory block from the specified offset
      */
     const u8* GetPointer(std::size_t offset = 0) const {
-        return device_memory->GetPointer(physical_address + offset);
+        return device_memory->GetPointer<u8>(physical_address + offset);
     }
 
     void Finalize() override;

src/core/hle/kernel/k_shared_memory_info.h

@@ -15,7 +15,8 @@ class KSharedMemoryInfo final : public KSlabAllocated<KSharedMemoryInfo>,
                                 public boost::intrusive::list_base_hook<> {
 
 public:
-    explicit KSharedMemoryInfo() = default;
+    explicit KSharedMemoryInfo(KernelCore&) {}
+    KSharedMemoryInfo() = default;
 
     constexpr void Initialize(KSharedMemory* shmem) {
         shared_memory = shmem;

src/core/hle/kernel/k_slab_heap.h

@@ -8,6 +8,7 @@
 #include "common/assert.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
+#include "common/spin_lock.h"
 
 namespace Kernel {
 
@@ -36,28 +37,34 @@ public:
     }
 
     void* Allocate() {
-        Node* ret = m_head.load();
+        // KScopedInterruptDisable di;
 
-        do {
-            if (ret == nullptr) {
-                break;
-            }
-        } while (!m_head.compare_exchange_weak(ret, ret->next));
+        m_lock.lock();
 
+        Node* ret = m_head;
+        if (ret != nullptr) [[likely]] {
+            m_head = ret->next;
+        }
+
+        m_lock.unlock();
         return ret;
     }
 
     void Free(void* obj) {
-        Node* node = static_cast<Node*>(obj);
+        // KScopedInterruptDisable di;
 
-        Node* cur_head = m_head.load();
-        do {
-            node->next = cur_head;
-        } while (!m_head.compare_exchange_weak(cur_head, node));
+        m_lock.lock();
+
+        Node* node = static_cast<Node*>(obj);
+        node->next = m_head;
+        m_head = node;
+
+        m_lock.unlock();
     }
 
 private:
     std::atomic<Node*> m_head{};
+    Common::SpinLock m_lock;
 };
 
 } // namespace impl

src/core/hle/kernel/k_thread.cpp

@@ -30,6 +30,7 @@
 #include "core/hle/kernel/k_worker_task_manager.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/svc_results.h"
+#include "core/hle/kernel/svc_types.h"
 #include "core/hle/result.h"
 #include "core/memory.h"
 
@@ -38,6 +39,9 @@
 #endif
 
 namespace {
+
+constexpr inline s32 TerminatingThreadPriority = Kernel::Svc::SystemThreadPriorityHighest - 1;
+
 static void ResetThreadContext32(Core::ARM_Interface::ThreadContext32& context, u32 stack_top,
                                  u32 entry_point, u32 arg) {
     context = {};
@@ -241,7 +245,7 @@ Result KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_stack
         }
     }
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 Result KThread::InitializeThread(KThread* thread, KThreadFunction func, uintptr_t arg,
@@ -254,7 +258,7 @@ Result KThread::InitializeThread(KThread* thread, KThreadFunction func, uintptr_
     thread->host_context = std::make_shared<Common::Fiber>(std::move(init_func));
     thread->is_single_core = !Settings::values.use_multi_core.GetValue();
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 Result KThread::InitializeDummyThread(KThread* thread) {
@@ -264,31 +268,32 @@ Result KThread::InitializeDummyThread(KThread* thread) {
     // Initialize emulation parameters.
     thread->stack_parameters.disable_count = 0;
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 Result KThread::InitializeMainThread(Core::System& system, KThread* thread, s32 virt_core) {
-    return InitializeThread(thread, {}, {}, {}, IdleThreadPriority, virt_core, {}, ThreadType::Main,
-                            system.GetCpuManager().GetGuestActivateFunc());
+    R_RETURN(InitializeThread(thread, {}, {}, {}, IdleThreadPriority, virt_core, {},
+                              ThreadType::Main, system.GetCpuManager().GetGuestActivateFunc()));
 }
 
 Result KThread::InitializeIdleThread(Core::System& system, KThread* thread, s32 virt_core) {
-    return InitializeThread(thread, {}, {}, {}, IdleThreadPriority, virt_core, {}, ThreadType::Main,
-                            system.GetCpuManager().GetIdleThreadStartFunc());
+    R_RETURN(InitializeThread(thread, {}, {}, {}, IdleThreadPriority, virt_core, {},
+                              ThreadType::Main, system.GetCpuManager().GetIdleThreadStartFunc()));
 }
 
 Result KThread::InitializeHighPriorityThread(Core::System& system, KThread* thread,
                                              KThreadFunction func, uintptr_t arg, s32 virt_core) {
-    return InitializeThread(thread, func, arg, {}, {}, virt_core, nullptr, ThreadType::HighPriority,
-                            system.GetCpuManager().GetShutdownThreadStartFunc());
+    R_RETURN(InitializeThread(thread, func, arg, {}, {}, virt_core, nullptr,
+                              ThreadType::HighPriority,
+                              system.GetCpuManager().GetShutdownThreadStartFunc()));
 }
 
 Result KThread::InitializeUserThread(Core::System& system, KThread* thread, KThreadFunction func,
                                      uintptr_t arg, VAddr user_stack_top, s32 prio, s32 virt_core,
                                      KProcess* owner) {
     system.Kernel().GlobalSchedulerContext().AddThread(thread);
-    return InitializeThread(thread, func, arg, user_stack_top, prio, virt_core, owner,
-                            ThreadType::User, system.GetCpuManager().GetGuestThreadFunc());
+    R_RETURN(InitializeThread(thread, func, arg, user_stack_top, prio, virt_core, owner,
+                              ThreadType::User, system.GetCpuManager().GetGuestThreadFunc()));
 }
 
 void KThread::PostDestroy(uintptr_t arg) {
@@ -538,7 +543,7 @@ Result KThread::GetCoreMask(s32* out_ideal_core, u64* out_affinity_mask) {
     *out_ideal_core = virtual_ideal_core_id;
     *out_affinity_mask = virtual_affinity_mask;
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 Result KThread::GetPhysicalCoreMask(s32* out_ideal_core, u64* out_affinity_mask) {
@@ -554,7 +559,7 @@ Result KThread::GetPhysicalCoreMask(s32* out_ideal_core, u64* out_affinity_mask)
         *out_affinity_mask = original_physical_affinity_mask.GetAffinityMask();
     }
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 Result KThread::SetCoreMask(s32 core_id_, u64 v_affinity_mask) {
@@ -666,7 +671,7 @@ Result KThread::SetCoreMask(s32 core_id_, u64 v_affinity_mask) {
         } while (retry_update);
     }
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 void KThread::SetBasePriority(s32 value) {
@@ -839,7 +844,7 @@ Result KThread::SetActivity(Svc::ThreadActivity activity) {
         } while (thread_is_current);
     }
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 Result KThread::GetThreadContext3(std::vector<u8>& out) {
@@ -874,7 +879,7 @@ Result KThread::GetThreadContext3(std::vector<u8>& out) {
         }
     }
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 void KThread::AddWaiterImpl(KThread* thread) {
@@ -1038,7 +1043,7 @@ Result KThread::Run() {
         // Set our state and finish.
         SetState(ThreadState::Runnable);
 
-        return ResultSuccess;
+        R_SUCCEED();
     }
 }
 
@@ -1073,6 +1078,78 @@ void KThread::Exit() {
     UNREACHABLE_MSG("KThread::Exit() would return");
 }
 
+Result KThread::Terminate() {
+    ASSERT(this != GetCurrentThreadPointer(kernel));
+
+    // Request the thread terminate if it hasn't already.
+    if (const auto new_state = this->RequestTerminate(); new_state != ThreadState::Terminated) {
+        // If the thread isn't terminated, wait for it to terminate.
+        s32 index;
+        KSynchronizationObject* objects[] = {this};
+        R_TRY(KSynchronizationObject::Wait(kernel, std::addressof(index), objects, 1,
+                                           Svc::WaitInfinite));
+    }
+
+    R_SUCCEED();
+}
+
+ThreadState KThread::RequestTerminate() {
+    ASSERT(this != GetCurrentThreadPointer(kernel));
+
+    KScopedSchedulerLock sl{kernel};
+
+    // Determine if this is the first termination request.
+    const bool first_request = [&]() -> bool {
+        // Perform an atomic compare-and-swap from false to true.
+        bool expected = false;
+        return termination_requested.compare_exchange_strong(expected, true);
+    }();
+
+    // If this is the first request, start termination procedure.
+    if (first_request) {
+        // If the thread is in initialized state, just change state to terminated.
+        if (this->GetState() == ThreadState::Initialized) {
+            thread_state = ThreadState::Terminated;
+            return ThreadState::Terminated;
+        }
+
+        // Register the terminating dpc.
+        this->RegisterDpc(DpcFlag::Terminating);
+
+        // If the thread is pinned, unpin it.
+        if (this->GetStackParameters().is_pinned) {
+            this->GetOwnerProcess()->UnpinThread(this);
+        }
+
+        // If the thread is suspended, continue it.
+        if (this->IsSuspended()) {
+            suspend_allowed_flags = 0;
+            this->UpdateState();
+        }
+
+        // Change the thread's priority to be higher than any system thread's.
+        if (this->GetBasePriority() >= Svc::SystemThreadPriorityHighest) {
+            this->SetBasePriority(TerminatingThreadPriority);
+        }
+
+        // If the thread is runnable, send a termination interrupt to other cores.
+        if (this->GetState() == ThreadState::Runnable) {
+            if (const u64 core_mask =
+                    physical_affinity_mask.GetAffinityMask() & ~(1ULL << GetCurrentCoreId(kernel));
+                core_mask != 0) {
+                Kernel::KInterruptManager::SendInterProcessorInterrupt(kernel, core_mask);
+            }
+        }
+
+        // Wake up the thread.
+        if (this->GetState() == ThreadState::Waiting) {
+            wait_queue->CancelWait(this, ResultTerminationRequested, true);
+        }
+    }
+
+    return this->GetState();
+}
+
 Result KThread::Sleep(s64 timeout) {
     ASSERT(!kernel.GlobalSchedulerContext().IsLocked());
     ASSERT(this == GetCurrentThreadPointer(kernel));
@@ -1086,7 +1163,7 @@ Result KThread::Sleep(s64 timeout) {
         // Check if the thread should terminate.
         if (this->IsTerminationRequested()) {
             slp.CancelSleep();
-            return ResultTerminationRequested;
+            R_THROW(ResultTerminationRequested);
         }
 
         // Wait for the sleep to end.
@@ -1094,7 +1171,7 @@ Result KThread::Sleep(s64 timeout) {
         SetWaitReasonForDebugging(ThreadWaitReasonForDebugging::Sleep);
     }
 
-    return ResultSuccess;
+    R_SUCCEED();
 }
 
 void KThread::IfDummyThreadTryWait() {

src/core/hle/kernel/k_thread.h

@@ -180,6 +180,10 @@ public:
 
     void Exit();
 
+    Result Terminate();
+
+    ThreadState RequestTerminate();
+
     [[nodiscard]] u32 GetSuspendFlags() const {
         return suspend_allowed_flags & suspend_request_flags;
     }

src/core/hle/kernel/k_thread_local_page.h

@@ -26,7 +26,7 @@ public:
     static_assert(RegionsPerPage > 0);
 
 public:
-    constexpr explicit KThreadLocalPage(VAddr addr = {}) : m_virt_addr(addr) {
+    constexpr explicit KThreadLocalPage(KernelCore&, VAddr addr = {}) : m_virt_addr(addr) {
         m_is_region_free.fill(true);
     }
 

src/core/hle/kernel/kernel.cpp

@@ -24,6 +24,7 @@
 #include "core/hardware_properties.h"
 #include "core/hle/kernel/init/init_slab_setup.h"
 #include "core/hle/kernel/k_client_port.h"
+#include "core/hle/kernel/k_dynamic_resource_manager.h"
 #include "core/hle/kernel/k_handle_table.h"
 #include "core/hle/kernel/k_memory_layout.h"
 #include "core/hle/kernel/k_memory_manager.h"
@@ -73,8 +74,16 @@ struct KernelCore::Impl {
         InitializeMemoryLayout();
         Init::InitializeKPageBufferSlabHeap(system);
         InitializeShutdownThreads();
-        InitializePreemption(kernel);
         InitializePhysicalCores();
+        InitializePreemption(kernel);
+
+        // Initialize the Dynamic Slab Heaps.
+        {
+            const auto& pt_heap_region = memory_layout->GetPageTableHeapRegion();
+            ASSERT(pt_heap_region.GetEndAddress() != 0);
+
+            InitializeResourceManagers(pt_heap_region.GetAddress(), pt_heap_region.GetSize());
+        }
 
         RegisterHostThread();
     }
@@ -86,6 +95,15 @@ struct KernelCore::Impl {
         }
     }
 
+    void CloseCurrentProcess() {
+        (*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 = nullptr;
+    }
+
     void Shutdown() {
         is_shutting_down.store(true, std::memory_order_relaxed);
         SCOPE_EXIT({ is_shutting_down.store(false, std::memory_order_relaxed); });
@@ -99,10 +117,6 @@ struct KernelCore::Impl {
         next_user_process_id = KProcess::ProcessIDMin;
         next_thread_id = 1;
 
-        for (auto& core : cores) {
-            core = nullptr;
-        }
-
         global_handle_table->Finalize();
         global_handle_table.reset();
 
@@ -152,15 +166,7 @@ struct KernelCore::Impl {
             }
         }
 
-        // Shutdown all processes.
-        if (current_process) {
-            (*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 = nullptr;
-        }
+        CloseCurrentProcess();
 
         // Track kernel objects that were not freed on shutdown
         {
@@ -257,6 +263,18 @@ struct KernelCore::Impl {
         system.CoreTiming().ScheduleLoopingEvent(time_interval, time_interval, preemption_event);
     }
 
+    void InitializeResourceManagers(VAddr address, size_t size) {
+        dynamic_page_manager = std::make_unique<KDynamicPageManager>();
+        memory_block_heap = std::make_unique<KMemoryBlockSlabHeap>();
+        app_memory_block_manager = std::make_unique<KMemoryBlockSlabManager>();
+
+        dynamic_page_manager->Initialize(address, size);
+        static constexpr size_t ApplicationMemoryBlockSlabHeapSize = 20000;
+        memory_block_heap->Initialize(dynamic_page_manager.get(),
+                                      ApplicationMemoryBlockSlabHeapSize);
+        app_memory_block_manager->Initialize(nullptr, memory_block_heap.get());
+    }
+
     void InitializeShutdownThreads() {
         for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
             shutdown_threads[core_id] = KThread::Create(system.Kernel());
@@ -344,11 +362,6 @@ struct KernelCore::Impl {
     static inline thread_local KThread* current_thread{nullptr};
 
     KThread* GetCurrentEmuThread() {
-        // If we are shutting down the kernel, none of this is relevant anymore.
-        if (IsShuttingDown()) {
-            return {};
-        }
-
         const auto thread_id = GetCurrentHostThreadID();
         if (thread_id >= Core::Hardware::NUM_CPU_CORES) {
             return GetHostDummyThread();
@@ -770,6 +783,11 @@ struct KernelCore::Impl {
     // Kernel memory management
     std::unique_ptr<KMemoryManager> memory_manager;
 
+    // Dynamic slab managers
+    std::unique_ptr<KDynamicPageManager> dynamic_page_manager;
+    std::unique_ptr<KMemoryBlockSlabHeap> memory_block_heap;
+    std::unique_ptr<KMemoryBlockSlabManager> app_memory_block_manager;
+
     // Shared memory for services
     Kernel::KSharedMemory* hid_shared_mem{};
     Kernel::KSharedMemory* font_shared_mem{};
@@ -853,6 +871,10 @@ const KProcess* KernelCore::CurrentProcess() const {
     return impl->current_process;
 }
 
+void KernelCore::CloseCurrentProcess() {
+    impl->CloseCurrentProcess();
+}
+
 const std::vector<KProcess*>& KernelCore::GetProcessList() const {
     return impl->process_list;
 }
@@ -1041,6 +1063,14 @@ const KMemoryManager& KernelCore::MemoryManager() const {
     return *impl->memory_manager;
 }
 
+KMemoryBlockSlabManager& KernelCore::GetApplicationMemoryBlockManager() {
+    return *impl->app_memory_block_manager;
+}
+
+const KMemoryBlockSlabManager& KernelCore::GetApplicationMemoryBlockManager() const {
+    return *impl->app_memory_block_manager;
+}
+
 Kernel::KSharedMemory& KernelCore::GetHidSharedMem() {
     return *impl->hid_shared_mem;
 }

src/core/hle/kernel/kernel.h

@@ -37,6 +37,7 @@ class KClientSession;
 class KEvent;
 class KHandleTable;
 class KLinkedListNode;
+class KMemoryBlockSlabManager;
 class KMemoryLayout;
 class KMemoryManager;
 class KPageBuffer;
@@ -46,6 +47,7 @@ class KResourceLimit;
 class KScheduler;
 class KServerSession;
 class KSession;
+class KSessionRequest;
 class KSharedMemory;
 class KSharedMemoryInfo;
 class KThread;
@@ -130,6 +132,9 @@ public:
     /// Retrieves a const pointer to the current process.
     const KProcess* CurrentProcess() const;
 
+    /// Closes the current process.
+    void CloseCurrentProcess();
+
     /// Retrieves the list of processes.
     const std::vector<KProcess*>& GetProcessList() const;
 
@@ -238,6 +243,12 @@ public:
     /// Gets the virtual memory manager for the kernel.
     const KMemoryManager& MemoryManager() const;
 
+    /// Gets the application memory block manager for the kernel.
+    KMemoryBlockSlabManager& GetApplicationMemoryBlockManager();
+
+    /// Gets the application memory block manager for the kernel.
+    const KMemoryBlockSlabManager& GetApplicationMemoryBlockManager() const;
+
     /// Gets the shared memory object for HID services.
     Kernel::KSharedMemory& GetHidSharedMem();
 
@@ -350,6 +361,8 @@ public:
             return slab_heap_container->page_buffer;
         } else if constexpr (std::is_same_v<T, KThreadLocalPage>) {
             return slab_heap_container->thread_local_page;
+        } else if constexpr (std::is_same_v<T, KSessionRequest>) {
+            return slab_heap_container->session_request;
         }
     }
 
@@ -412,6 +425,7 @@ private:
         KSlabHeap<KCodeMemory> code_memory;
         KSlabHeap<KPageBuffer> page_buffer;
         KSlabHeap<KThreadLocalPage> thread_local_page;
+        KSlabHeap<KSessionRequest> session_request;
     };
 
     std::unique_ptr<SlabHeapContainer> slab_heap_container;

src/core/hle/kernel/slab_helpers.h

@@ -24,7 +24,7 @@ public:
     }
 
     static Derived* Allocate(KernelCore& kernel) {
-        return kernel.SlabHeap<Derived>().Allocate();
+        return kernel.SlabHeap<Derived>().Allocate(kernel);
     }
 
     static void Free(KernelCore& kernel, Derived* obj) {

src/core/hle/kernel/svc.cpp

@@ -933,7 +933,7 @@ static Result GetInfo(Core::System& system, u64* result, u64 info_id, Handle han
             return ResultSuccess;
 
         case GetInfoType::UserExceptionContextAddr:
-            *result = process->GetTLSRegionAddress();
+            *result = process->GetProcessLocalRegionAddress();
             return ResultSuccess;
 
         case GetInfoType::TotalPhysicalMemoryAvailableWithoutSystemResource:
@@ -1888,7 +1888,7 @@ static void ExitProcess(Core::System& system) {
     auto* current_process = system.Kernel().CurrentProcess();
 
     LOG_INFO(Kernel_SVC, "Process {} exiting", current_process->GetProcessID());
-    ASSERT_MSG(current_process->GetStatus() == ProcessStatus::Running,
+    ASSERT_MSG(current_process->GetState() == KProcess::State::Running,
                "Process has already exited");
 
     system.Exit();
@@ -2557,7 +2557,7 @@ static Result GetProcessInfo(Core::System& system, u64* out, Handle process_hand
         return ResultInvalidEnumValue;
     }
 
-    *out = static_cast<u64>(process->GetStatus());
+    *out = static_cast<u64>(process->GetState());
     return ResultSuccess;
 }
 

src/core/hle/kernel/svc_common.h

@@ -14,8 +14,11 @@ namespace Kernel::Svc {
 
 using namespace Common::Literals;
 
-constexpr s32 ArgumentHandleCountMax = 0x40;
-constexpr u32 HandleWaitMask{1u << 30};
+constexpr inline s32 ArgumentHandleCountMax = 0x40;
+
+constexpr inline u32 HandleWaitMask = 1u << 30;
+
+constexpr inline s64 WaitInfinite = -1;
 
 constexpr inline std::size_t HeapSizeAlignment = 2_MiB;
 

src/core/hle/kernel/svc_types.h

@@ -95,6 +95,19 @@ constexpr inline s32 IdealCoreNoUpdate = -3;
 constexpr inline s32 LowestThreadPriority = 63;
 constexpr inline s32 HighestThreadPriority = 0;
 
+constexpr inline s32 SystemThreadPriorityHighest = 16;
+
+enum class ProcessState : u32 {
+    Created = 0,
+    CreatedAttached = 1,
+    Running = 2,
+    Crashed = 3,
+    RunningAttached = 4,
+    Terminating = 5,
+    Terminated = 6,
+    DebugBreak = 7,
+};
+
 constexpr inline size_t ThreadLocalRegionSize = 0x200;
 
 } // namespace Kernel::Svc

src/core/hle/result.h

@@ -135,6 +135,14 @@ union Result {
     [[nodiscard]] constexpr bool IsFailure() const {
         return !IsSuccess();
     }
+
+    [[nodiscard]] constexpr u32 GetInnerValue() const {
+        return static_cast<u32>(module.Value()) | (description << module.bits);
+    }
+
+    [[nodiscard]] constexpr bool Includes(Result result) const {
+        return GetInnerValue() == result.GetInnerValue();
+    }
 };
 static_assert(std::is_trivial_v<Result>);
 
@@ -462,9 +470,6 @@ constexpr inline Result __TmpCurrentResultReference = ResultSuccess;
 #define R_UNLESS(expr, res)                                                                        \
     {                                                                                              \
         if (!(expr)) {                                                                             \
-            if (res.IsError()) {                                                                   \
-                LOG_ERROR(Kernel, "Failed with result: {}", res.raw);                              \
-            }                                                                                      \
             R_THROW(res);                                                                          \
         }                                                                                          \
     }

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

@@ -45,9 +45,25 @@ AudCtl::AudCtl(Core::System& system_) : ServiceFramework{system_, "audctl"} {
         {32, nullptr, "GetActiveOutputTarget"},
         {33, nullptr, "GetTargetDeviceInfo"},
         {34, nullptr, "AcquireTargetNotification"},
+        {35, nullptr, "SetHearingProtectionSafeguardTimerRemainingTimeForDebug"},
+        {36, nullptr, "GetHearingProtectionSafeguardTimerRemainingTimeForDebug"},
+        {37, nullptr, "SetHearingProtectionSafeguardEnabled"},
+        {38, nullptr, "IsHearingProtectionSafeguardEnabled"},
+        {39, nullptr, "IsHearingProtectionSafeguardMonitoringOutputForDebug"},
+        {40, nullptr, "GetSystemInformationForDebug"},
+        {41, nullptr, "SetVolumeButtonLongPressTime"},
+        {42, nullptr, "SetNativeVolumeForDebug"},
         {10000, nullptr, "NotifyAudioOutputTargetForPlayReport"},
         {10001, nullptr, "NotifyAudioOutputChannelCountForPlayReport"},
         {10002, nullptr, "NotifyUnsupportedUsbOutputDeviceAttachedForPlayReport"},
+        {10100, nullptr, "GetAudioVolumeDataForPlayReport"},
+        {10101, nullptr, "BindAudioVolumeUpdateEventForPlayReport"},
+        {10102, nullptr, "BindAudioOutputTargetUpdateEventForPlayReport"},
+        {10103, nullptr, "GetAudioOutputTargetForPlayReport"},
+        {10104, nullptr, "GetAudioOutputChannelCountForPlayReport"},
+        {10105, nullptr, "BindAudioOutputChannelCountUpdateEventForPlayReport"},
+        {10106, nullptr, "GetDefaultAudioOutputTargetForPlayReport"},
+        {50000, nullptr, "SetAnalogInputBoostGainForPrototyping"},
     };
     // clang-format on
 

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

@@ -52,6 +52,8 @@ public:
             {9, &IAudioRenderer::GetRenderingTimeLimit, "GetRenderingTimeLimit"},
             {10, &IAudioRenderer::RequestUpdate, "RequestUpdateAuto"},
             {11, nullptr, "ExecuteAudioRendererRendering"},
+            {12, &IAudioRenderer::SetVoiceDropParameter, "SetVoiceDropParameter"},
+            {13, &IAudioRenderer::GetVoiceDropParameter, "GetVoiceDropParameter"},
         };
         // clang-format on
         RegisterHandlers(functions);
@@ -205,6 +207,30 @@ private:
         LOG_DEBUG(Service_Audio, "called");
     }
 
+    void SetVoiceDropParameter(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_Audio, "called");
+
+        IPC::RequestParser rp{ctx};
+        auto voice_drop_param{rp.Pop<f32>()};
+
+        auto& system_ = impl->GetSystem();
+        system_.SetVoiceDropParameter(voice_drop_param);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultSuccess);
+    }
+
+    void GetVoiceDropParameter(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_Audio, "called");
+
+        auto& system_ = impl->GetSystem();
+        auto voice_drop_param{system_.GetVoiceDropParameter()};
+
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(ResultSuccess);
+        rb.Push(voice_drop_param);
+    }
+
     KernelHelpers::ServiceContext service_context;
     Kernel::KEvent* rendered_event;
     Manager& manager;

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

@@ -2118,7 +2118,7 @@ void Hid::WritePalmaWaveEntry(Kernel::HLERequestContext& ctx) {
     ASSERT_MSG(t_mem->GetSize() == 0x3000, "t_mem has incorrect size");
 
     LOG_WARNING(Service_HID,
-                "(STUBBED) called, connection_handle={}, wave_set={}, unkown={}, "
+                "(STUBBED) called, connection_handle={}, wave_set={}, unknown={}, "
                 "t_mem_handle=0x{:08X}, t_mem_size={}, size={}",
                 connection_handle.npad_id, wave_set, unknown, t_mem_handle, t_mem_size, size);
 

src/core/hle/service/hid/irsensor/pointing_processor.h

@@ -37,10 +37,10 @@ private:
         u8 pointing_status;
         INSERT_PADDING_BYTES(3);
         u32 unknown;
-        float unkown_float1;
+        float unknown_float1;
         float position_x;
         float position_y;
-        float unkown_float2;
+        float unknown_float2;
         Core::IrSensor::IrsRect window_of_interest;
     };
     static_assert(sizeof(PointingProcessorMarkerData) == 0x20,

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

@@ -290,7 +290,7 @@ public:
         const std::size_t padding_size{page_table.GetNumGuardPages() * Kernel::PageSize};
         const auto start_info{page_table.QueryInfo(start - 1)};
 
-        if (start_info.state != Kernel::KMemoryState::Free) {
+        if (start_info.GetState() != Kernel::KMemoryState::Free) {
             return {};
         }
 
@@ -300,7 +300,7 @@ public:
 
         const auto end_info{page_table.QueryInfo(start + size)};
 
-        if (end_info.state != Kernel::KMemoryState::Free) {
+        if (end_info.GetState() != Kernel::KMemoryState::Free) {
             return {};
         }
 

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp

@@ -311,7 +311,8 @@ NvResult nvhost_as_gpu::Remap(const std::vector<u8>& input, std::vector<u8>& out
                 handle->address +
                 (static_cast<u64>(entry.handle_offset_big_pages) << vm.big_page_size_bits))};
 
-            gmmu->Map(virtual_address, cpu_address, size, use_big_pages);
+            gmmu->Map(virtual_address, cpu_address, size, static_cast<Tegra::PTEKind>(entry.kind),
+                      use_big_pages);
         }
     }
 
@@ -350,7 +351,8 @@ NvResult nvhost_as_gpu::MapBufferEx(const std::vector<u8>& input, std::vector<u8
             u64 gpu_address{static_cast<u64>(params.offset + params.buffer_offset)};
             VAddr cpu_address{mapping->ptr + params.buffer_offset};
 
-            gmmu->Map(gpu_address, cpu_address, params.mapping_size, mapping->big_page);
+            gmmu->Map(gpu_address, cpu_address, params.mapping_size,
+                      static_cast<Tegra::PTEKind>(params.kind), mapping->big_page);
 
             return NvResult::Success;
         } catch (const std::out_of_range&) {
@@ -389,7 +391,8 @@ NvResult nvhost_as_gpu::MapBufferEx(const std::vector<u8>& input, std::vector<u8
         }
 
         const bool use_big_pages = alloc->second.big_pages && big_page;
-        gmmu->Map(params.offset, cpu_address, size, use_big_pages);
+        gmmu->Map(params.offset, cpu_address, size, static_cast<Tegra::PTEKind>(params.kind),
+                  use_big_pages);
 
         auto mapping{std::make_shared<Mapping>(cpu_address, params.offset, size, true,
                                                use_big_pages, alloc->second.sparse)};
@@ -409,7 +412,8 @@ NvResult nvhost_as_gpu::MapBufferEx(const std::vector<u8>& input, std::vector<u8
             return NvResult::InsufficientMemory;
         }
 
-        gmmu->Map(params.offset, cpu_address, Common::AlignUp(size, page_size), big_page);
+        gmmu->Map(params.offset, cpu_address, Common::AlignUp(size, page_size),
+                  static_cast<Tegra::PTEKind>(params.kind), big_page);
 
         auto mapping{
             std::make_shared<Mapping>(cpu_address, params.offset, size, false, big_page, false)};

src/core/hle/service/nvdrv/devices/nvmap.cpp

@@ -128,7 +128,8 @@ NvResult nvmap::IocAlloc(const std::vector<u8>& input, std::vector<u8>& output)
     }
     ASSERT(system.CurrentProcess()
                ->PageTable()
-               .LockForDeviceAddressSpace(handle_description->address, handle_description->size)
+               .LockForMapDeviceAddressSpace(handle_description->address, handle_description->size,
+                                             Kernel::KMemoryPermission::None, true)
                .IsSuccess());
     std::memcpy(output.data(), &params, sizeof(params));
     return result;

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

@@ -15,9 +15,10 @@
 namespace Service::SM {
 
 void Controller::ConvertCurrentObjectToDomain(Kernel::HLERequestContext& ctx) {
-    ASSERT_MSG(!ctx.Session()->IsDomain(), "Session is already a domain");
+    ASSERT_MSG(!ctx.Session()->GetSessionRequestManager()->IsDomain(),
+               "Session is already a domain");
     LOG_DEBUG(Service, "called, server_session={}", ctx.Session()->GetId());
-    ctx.Session()->ConvertToDomain();
+    ctx.Session()->GetSessionRequestManager()->ConvertToDomainOnRequestEnd();
 
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(ResultSuccess);

src/core/hle/service/vi/vi_results.h

@@ -1,6 +1,8 @@
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#pragma once
+
 #include "core/hle/result.h"
 
 namespace Service::VI {

src/core/memory.cpp

@@ -65,7 +65,7 @@ struct Memory::Impl {
             return {};
         }
 
-        return system.DeviceMemory().GetPointer(paddr) + vaddr;
+        return system.DeviceMemory().GetPointer<u8>(paddr) + vaddr;
     }
 
     [[nodiscard]] u8* GetPointerFromDebugMemory(VAddr vaddr) const {
@@ -75,7 +75,7 @@ struct Memory::Impl {
             return {};
         }
 
-        return system.DeviceMemory().GetPointer(paddr) + vaddr;
+        return system.DeviceMemory().GetPointer<u8>(paddr) + vaddr;
     }
 
     u8 Read8(const VAddr addr) {
@@ -499,7 +499,7 @@ struct Memory::Impl {
         } else {
             while (base != end) {
                 page_table.pointers[base].Store(
-                    system.DeviceMemory().GetPointer(target) - (base << YUZU_PAGEBITS), type);
+                    system.DeviceMemory().GetPointer<u8>(target) - (base << YUZU_PAGEBITS), type);
                 page_table.backing_addr[base] = target - (base << YUZU_PAGEBITS);
 
                 ASSERT_MSG(page_table.pointers[base].Pointer(),

src/tests/core/core_timing.cpp

@@ -40,9 +40,6 @@ struct ScopeInit final {
         core_timing.SetMulticore(true);
         core_timing.Initialize([]() {});
     }
-    ~ScopeInit() {
-        core_timing.Shutdown();
-    }
 
     Core::Timing::CoreTiming core_timing;
 };

src/video_core/CMakeLists.txt

@@ -82,6 +82,7 @@ add_library(video_core STATIC
     gpu_thread.h
     memory_manager.cpp
     memory_manager.h
+    pte_kind.h
     query_cache.h
     rasterizer_accelerated.cpp
     rasterizer_accelerated.h

src/video_core/engines/maxwell_3d.cpp

@@ -117,10 +117,18 @@ void Maxwell3D::InitializeRegisterDefaults() {
 
     shadow_state = regs;
 
-    mme_inline[MAXWELL3D_REG_INDEX(draw.end)] = true;
-    mme_inline[MAXWELL3D_REG_INDEX(draw.begin)] = true;
-    mme_inline[MAXWELL3D_REG_INDEX(vertex_buffer.count)] = true;
-    mme_inline[MAXWELL3D_REG_INDEX(index_buffer.count)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(draw.end)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(draw.begin)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(vertex_buffer.first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(vertex_buffer.count)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer.first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer.count)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer32_first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer16_first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer8_first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(draw_inline_index)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(inline_index_2x16.even)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(inline_index_4x8.index0)] = true;
 }
 
 void Maxwell3D::ProcessMacro(u32 method, const u32* base_start, u32 amount, bool is_last_call) {
@@ -208,25 +216,6 @@ void Maxwell3D::ProcessMethodCall(u32 method, u32 argument, u32 nonshadow_argume
         return ProcessCBBind(3);
     case MAXWELL3D_REG_INDEX(bind_groups[4].raw_config):
         return ProcessCBBind(4);
-    case MAXWELL3D_REG_INDEX(draw.end):
-        return DrawArrays();
-    case MAXWELL3D_REG_INDEX(index_buffer32_first):
-        regs.index_buffer.count = regs.index_buffer32_first.count;
-        regs.index_buffer.first = regs.index_buffer32_first.first;
-        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-        return DrawArrays();
-    case MAXWELL3D_REG_INDEX(index_buffer16_first):
-        regs.index_buffer.count = regs.index_buffer16_first.count;
-        regs.index_buffer.first = regs.index_buffer16_first.first;
-        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-        return DrawArrays();
-    case MAXWELL3D_REG_INDEX(index_buffer8_first):
-        regs.index_buffer.count = regs.index_buffer8_first.count;
-        regs.index_buffer.first = regs.index_buffer8_first.first;
-        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-        // a macro calls this one over and over, should it increase instancing?
-        // Used by Hades and likely other Vulkan games.
-        return DrawArrays();
     case MAXWELL3D_REG_INDEX(topology_override):
         use_topology_override = true;
         return;
@@ -261,14 +250,13 @@ void Maxwell3D::CallMacroMethod(u32 method, const std::vector<u32>& parameters)
 
     // Execute the current macro.
     macro_engine->Execute(macro_positions[entry], parameters);
-    if (mme_draw.current_mode != MMEDrawMode::Undefined) {
-        FlushMMEInlineDraw();
-    }
+
+    ProcessDeferredDraw();
 }
 
 void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
-    // It is an error to write to a register other than the current macro's ARG register before it
-    // has finished execution.
+    // It is an error to write to a register other than the current macro's ARG register before
+    // it has finished execution.
     if (executing_macro != 0) {
         ASSERT(method == executing_macro + 1);
     }
@@ -283,9 +271,33 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
     ASSERT_MSG(method < Regs::NUM_REGS,
                "Invalid Maxwell3D register, increase the size of the Regs structure");
 
-    const u32 argument = ProcessShadowRam(method, method_argument);
-    ProcessDirtyRegisters(method, argument);
-    ProcessMethodCall(method, argument, method_argument, is_last_call);
+    if (draw_command[method]) {
+        regs.reg_array[method] = method_argument;
+        deferred_draw_method.push_back(method);
+        auto u32_to_u8 = [&](const u32 argument) {
+            inline_index_draw_indexes.push_back(static_cast<u8>(argument & 0x000000ff));
+            inline_index_draw_indexes.push_back(static_cast<u8>((argument & 0x0000ff00) >> 8));
+            inline_index_draw_indexes.push_back(static_cast<u8>((argument & 0x00ff0000) >> 16));
+            inline_index_draw_indexes.push_back(static_cast<u8>((argument & 0xff000000) >> 24));
+        };
+        if (MAXWELL3D_REG_INDEX(draw_inline_index) == method) {
+            u32_to_u8(method_argument);
+        } else if (MAXWELL3D_REG_INDEX(inline_index_2x16.even) == method) {
+            u32_to_u8(regs.inline_index_2x16.even);
+            u32_to_u8(regs.inline_index_2x16.odd);
+        } else if (MAXWELL3D_REG_INDEX(inline_index_4x8.index0) == method) {
+            u32_to_u8(regs.inline_index_4x8.index0);
+            u32_to_u8(regs.inline_index_4x8.index1);
+            u32_to_u8(regs.inline_index_4x8.index2);
+            u32_to_u8(regs.inline_index_4x8.index3);
+        }
+    } else {
+        ProcessDeferredDraw();
+
+        const u32 argument = ProcessShadowRam(method, method_argument);
+        ProcessDirtyRegisters(method, argument);
+        ProcessMethodCall(method, argument, method_argument, is_last_call);
+    }
 }
 
 void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
@@ -326,55 +338,6 @@ void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
     }
 }
 
-void Maxwell3D::StepInstance(const MMEDrawMode expected_mode, const u32 count) {
-    if (mme_draw.current_mode == MMEDrawMode::Undefined) {
-        if (mme_draw.gl_begin_consume) {
-            mme_draw.current_mode = expected_mode;
-            mme_draw.current_count = count;
-            mme_draw.instance_count = 1;
-            mme_draw.gl_begin_consume = false;
-            mme_draw.gl_end_count = 0;
-        }
-        return;
-    } else {
-        if (mme_draw.current_mode == expected_mode && count == mme_draw.current_count &&
-            mme_draw.instance_mode && mme_draw.gl_begin_consume) {
-            mme_draw.instance_count++;
-            mme_draw.gl_begin_consume = false;
-            return;
-        } else {
-            FlushMMEInlineDraw();
-        }
-    }
-    // Tail call in case it needs to retry.
-    StepInstance(expected_mode, count);
-}
-
-void Maxwell3D::CallMethodFromMME(u32 method, u32 method_argument) {
-    if (mme_inline[method]) {
-        regs.reg_array[method] = method_argument;
-        if (method == MAXWELL3D_REG_INDEX(vertex_buffer.count) ||
-            method == MAXWELL3D_REG_INDEX(index_buffer.count)) {
-            const MMEDrawMode expected_mode = method == MAXWELL3D_REG_INDEX(vertex_buffer.count)
-                                                  ? MMEDrawMode::Array
-                                                  : MMEDrawMode::Indexed;
-            StepInstance(expected_mode, method_argument);
-        } else if (method == MAXWELL3D_REG_INDEX(draw.begin)) {
-            mme_draw.instance_mode =
-                (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Subsequent) ||
-                (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Unchanged);
-            mme_draw.gl_begin_consume = true;
-        } else {
-            mme_draw.gl_end_count++;
-        }
-    } else {
-        if (mme_draw.current_mode != MMEDrawMode::Undefined) {
-            FlushMMEInlineDraw();
-        }
-        CallMethod(method, method_argument, true);
-    }
-}
-
 void Maxwell3D::ProcessTopologyOverride() {
     using PrimitiveTopology = Maxwell3D::Regs::PrimitiveTopology;
     using PrimitiveTopologyOverride = Maxwell3D::Regs::PrimitiveTopologyOverride;
@@ -404,41 +367,6 @@ void Maxwell3D::ProcessTopologyOverride() {
     }
 }
 
-void Maxwell3D::FlushMMEInlineDraw() {
-    LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
-              regs.vertex_buffer.count);
-    ASSERT_MSG(!(regs.index_buffer.count && regs.vertex_buffer.count), "Both indexed and direct?");
-    ASSERT(mme_draw.instance_count == mme_draw.gl_end_count);
-
-    // Both instance configuration registers can not be set at the same time.
-    ASSERT_MSG(regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::First ||
-                   regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged,
-               "Illegal combination of instancing parameters");
-
-    ProcessTopologyOverride();
-
-    const bool is_indexed = mme_draw.current_mode == MMEDrawMode::Indexed;
-    if (ShouldExecute()) {
-        rasterizer->Draw(is_indexed, true);
-    }
-
-    // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
-    // the game is trying to draw indexed or direct mode. This needs to be verified on HW still -
-    // it's possible that it is incorrect and that there is some other register used to specify the
-    // drawing mode.
-    if (is_indexed) {
-        regs.index_buffer.count = 0;
-    } else {
-        regs.vertex_buffer.count = 0;
-    }
-    mme_draw.current_mode = MMEDrawMode::Undefined;
-    mme_draw.current_count = 0;
-    mme_draw.instance_count = 0;
-    mme_draw.instance_mode = false;
-    mme_draw.gl_begin_consume = false;
-    mme_draw.gl_end_count = 0;
-}
-
 void Maxwell3D::ProcessMacroUpload(u32 data) {
     macro_engine->AddCode(regs.load_mme.instruction_ptr++, data);
 }
@@ -576,42 +504,6 @@ void Maxwell3D::ProcessSyncPoint() {
     }
 }
 
-void Maxwell3D::DrawArrays() {
-    LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
-              regs.vertex_buffer.count);
-    ASSERT_MSG(!(regs.index_buffer.count && regs.vertex_buffer.count), "Both indexed and direct?");
-
-    // Both instance configuration registers can not be set at the same time.
-    ASSERT_MSG(regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::First ||
-                   regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged,
-               "Illegal combination of instancing parameters");
-
-    ProcessTopologyOverride();
-
-    if (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Subsequent) {
-        // Increment the current instance *before* drawing.
-        state.current_instance++;
-    } else if (regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged) {
-        // Reset the current instance to 0.
-        state.current_instance = 0;
-    }
-
-    const bool is_indexed{regs.index_buffer.count && !regs.vertex_buffer.count};
-    if (ShouldExecute()) {
-        rasterizer->Draw(is_indexed, false);
-    }
-
-    // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
-    // the game is trying to draw indexed or direct mode. This needs to be verified on HW still -
-    // it's possible that it is incorrect and that there is some other register used to specify the
-    // drawing mode.
-    if (is_indexed) {
-        regs.index_buffer.count = 0;
-    } else {
-        regs.vertex_buffer.count = 0;
-    }
-}
-
 std::optional<u64> Maxwell3D::GetQueryResult() {
     switch (regs.report_semaphore.query.report) {
     case Regs::ReportSemaphore::Report::Payload:
@@ -694,4 +586,95 @@ void Maxwell3D::ProcessClearBuffers() {
     rasterizer->Clear();
 }
 
+void Maxwell3D::ProcessDeferredDraw() {
+    if (deferred_draw_method.empty()) {
+        return;
+    }
+
+    enum class DrawMode {
+        Undefined,
+        General,
+        Instance,
+    };
+    DrawMode draw_mode{DrawMode::Undefined};
+    u32 instance_count = 1;
+
+    auto first_method = deferred_draw_method[0];
+    if (MAXWELL3D_REG_INDEX(draw.begin) == first_method) {
+        // The minimum number of methods for drawing must be greater than or equal to
+        // 3[draw.begin->vertex(index)count->draw.end] to avoid errors in index mode drawing
+        if (deferred_draw_method.size() < 3) {
+            return;
+        }
+        draw_mode = (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Subsequent) ||
+                            (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Unchanged)
+                        ? DrawMode::Instance
+                        : DrawMode::General;
+    } else if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method ||
+               MAXWELL3D_REG_INDEX(index_buffer16_first) == first_method ||
+               MAXWELL3D_REG_INDEX(index_buffer8_first) == first_method) {
+        draw_mode = DrawMode::General;
+    }
+
+    // Drawing will only begin with draw.begin or index_buffer method, other methods directly
+    // clear
+    if (draw_mode == DrawMode::Undefined) {
+        deferred_draw_method.clear();
+        return;
+    }
+
+    if (draw_mode == DrawMode::Instance) {
+        ASSERT_MSG(deferred_draw_method.size() % 4 == 0, "Instance mode method size error");
+        instance_count = static_cast<u32>(deferred_draw_method.size()) / 4;
+    } else {
+        if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method) {
+            regs.index_buffer.count = regs.index_buffer32_first.count;
+            regs.index_buffer.first = regs.index_buffer32_first.first;
+            dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        } else if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method) {
+            regs.index_buffer.count = regs.index_buffer16_first.count;
+            regs.index_buffer.first = regs.index_buffer16_first.first;
+            dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        } else if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method) {
+            regs.index_buffer.count = regs.index_buffer8_first.count;
+            regs.index_buffer.first = regs.index_buffer8_first.first;
+            dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        } else {
+            auto second_method = deferred_draw_method[1];
+            if (MAXWELL3D_REG_INDEX(draw_inline_index) == second_method ||
+                MAXWELL3D_REG_INDEX(inline_index_2x16.even) == second_method ||
+                MAXWELL3D_REG_INDEX(inline_index_4x8.index0) == second_method) {
+                regs.index_buffer.count = static_cast<u32>(inline_index_draw_indexes.size() / 4);
+                regs.index_buffer.format = Regs::IndexFormat::UnsignedInt;
+            }
+        }
+    }
+
+    LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
+              regs.vertex_buffer.count);
+
+    ASSERT_MSG(!(regs.index_buffer.count && regs.vertex_buffer.count), "Both indexed and direct?");
+
+    // Both instance configuration registers can not be set at the same time.
+    ASSERT_MSG(regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::First ||
+                   regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged,
+               "Illegal combination of instancing parameters");
+
+    ProcessTopologyOverride();
+
+    const bool is_indexed = regs.index_buffer.count && !regs.vertex_buffer.count;
+    if (ShouldExecute()) {
+        rasterizer->Draw(is_indexed, instance_count);
+    }
+
+    if (is_indexed) {
+        regs.index_buffer.count = 0;
+    } else {
+        regs.vertex_buffer.count = 0;
+    }
+
+    deferred_draw_method.clear();
+    inline_index_draw_indexes.clear();
+}
+
 } // namespace Tegra::Engines

src/video_core/engines/maxwell_3d.h

@@ -1739,14 +1739,11 @@ public:
             Footprint_1x1_Virtual = 2,
         };
 
-        struct InlineIndex4x8Align {
+        struct InlineIndex4x8 {
             union {
                 BitField<0, 30, u32> count;
                 BitField<30, 2, u32> start;
             };
-        };
-
-        struct InlineIndex4x8Index {
             union {
                 BitField<0, 8, u32> index0;
                 BitField<8, 8, u32> index1;
@@ -2836,8 +2833,7 @@ public:
                 u32 depth_write_enabled;                                               ///< 0x12E8
                 u32 alpha_test_enabled;                                                ///< 0x12EC
                 INSERT_PADDING_BYTES_NOINIT(0x10);
-                InlineIndex4x8Align inline_index_4x8_align;                            ///< 0x1300
-                InlineIndex4x8Index inline_index_4x8_index;                            ///< 0x1304
+                InlineIndex4x8 inline_index_4x8;                                       ///< 0x1300
                 D3DCullMode d3d_cull_mode;                                             ///< 0x1308
                 ComparisonOp depth_test_func;                                          ///< 0x130C
                 f32 alpha_test_ref;                                                    ///< 0x1310
@@ -3048,8 +3044,6 @@ public:
         };
 
         std::array<ShaderStageInfo, Regs::MaxShaderStage> shader_stages;
-
-        u32 current_instance = 0; ///< Current instance to be used to simulate instanced rendering.
     };
 
     State state{};
@@ -3064,11 +3058,6 @@ public:
     void CallMultiMethod(u32 method, const u32* base_start, u32 amount,
                          u32 methods_pending) override;
 
-    /// Write the value to the register identified by method.
-    void CallMethodFromMME(u32 method, u32 method_argument);
-
-    void FlushMMEInlineDraw();
-
     bool ShouldExecute() const {
         return execute_on;
     }
@@ -3081,21 +3070,6 @@ public:
         return *rasterizer;
     }
 
-    enum class MMEDrawMode : u32 {
-        Undefined,
-        Array,
-        Indexed,
-    };
-
-    struct MMEDrawState {
-        MMEDrawMode current_mode{MMEDrawMode::Undefined};
-        u32 current_count{};
-        u32 instance_count{};
-        bool instance_mode{};
-        bool gl_begin_consume{};
-        u32 gl_end_count{};
-    } mme_draw;
-
     struct DirtyState {
         using Flags = std::bitset<std::numeric_limits<u8>::max()>;
         using Table = std::array<u8, Regs::NUM_REGS>;
@@ -3105,6 +3079,8 @@ public:
         Tables tables{};
     } dirty;
 
+    std::vector<u8> inline_index_draw_indexes;
+
 private:
     void InitializeRegisterDefaults();
 
@@ -3164,14 +3140,10 @@ private:
     /// Handles a write to the CB_BIND register.
     void ProcessCBBind(size_t stage_index);
 
-    /// Handles a write to the VERTEX_END_GL register, triggering a draw.
-    void DrawArrays();
-
     /// Handles use of topology overrides (e.g., to avoid using a topology assigned from a macro)
     void ProcessTopologyOverride();
 
-    // Handles a instance drawcall from MME
-    void StepInstance(MMEDrawMode expected_mode, u32 count);
+    void ProcessDeferredDraw();
 
     /// Returns a query's value or an empty object if the value will be deferred through a cache.
     std::optional<u64> GetQueryResult();
@@ -3184,8 +3156,6 @@ private:
     /// Start offsets of each macro in macro_memory
     std::array<u32, 0x80> macro_positions{};
 
-    std::array<bool, Regs::NUM_REGS> mme_inline{};
-
     /// Macro method that is currently being executed / being fed parameters.
     u32 executing_macro = 0;
     /// Parameters that have been submitted to the macro call so far.
@@ -3198,6 +3168,9 @@ private:
 
     bool execute_on{true};
     bool use_topology_override{false};
+
+    std::array<bool, Regs::NUM_REGS> draw_command{};
+    std::vector<u32> deferred_draw_method;
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
@@ -3402,8 +3375,7 @@ ASSERT_REG_POSITION(alpha_to_coverage_dither, 0x12E0);
 ASSERT_REG_POSITION(blend_per_target_enabled, 0x12E4);
 ASSERT_REG_POSITION(depth_write_enabled, 0x12E8);
 ASSERT_REG_POSITION(alpha_test_enabled, 0x12EC);
-ASSERT_REG_POSITION(inline_index_4x8_align, 0x1300);
-ASSERT_REG_POSITION(inline_index_4x8_index, 0x1304);
+ASSERT_REG_POSITION(inline_index_4x8, 0x1300);
 ASSERT_REG_POSITION(d3d_cull_mode, 0x1308);
 ASSERT_REG_POSITION(depth_test_func, 0x130C);
 ASSERT_REG_POSITION(alpha_test_ref, 0x1310);

src/video_core/engines/maxwell_dma.cpp

@@ -56,68 +56,87 @@ void MaxwellDMA::Launch() {
     ASSERT(launch.interrupt_type == LaunchDMA::InterruptType::NONE);
     ASSERT(launch.data_transfer_type == LaunchDMA::DataTransferType::NON_PIPELINED);
 
-    const bool is_src_pitch = launch.src_memory_layout == LaunchDMA::MemoryLayout::PITCH;
-    const bool is_dst_pitch = launch.dst_memory_layout == LaunchDMA::MemoryLayout::PITCH;
+    if (launch.multi_line_enable) {
+        const bool is_src_pitch = launch.src_memory_layout == LaunchDMA::MemoryLayout::PITCH;
+        const bool is_dst_pitch = launch.dst_memory_layout == LaunchDMA::MemoryLayout::PITCH;
 
-    if (!is_src_pitch && !is_dst_pitch) {
-        // If both the source and the destination are in block layout, assert.
-        UNIMPLEMENTED_MSG("Tiled->Tiled DMA transfers are not yet implemented");
-        return;
-    }
+        if (!is_src_pitch && !is_dst_pitch) {
+            // If both the source and the destination are in block layout, assert.
+            UNIMPLEMENTED_MSG("Tiled->Tiled DMA transfers are not yet implemented");
+            return;
+        }
 
-    if (is_src_pitch && is_dst_pitch) {
-        CopyPitchToPitch();
-    } else {
-        ASSERT(launch.multi_line_enable == 1);
-
-        if (!is_src_pitch && is_dst_pitch) {
-            CopyBlockLinearToPitch();
+        if (is_src_pitch && is_dst_pitch) {
+            for (u32 line = 0; line < regs.line_count; ++line) {
+                const GPUVAddr source_line =
+                    regs.offset_in + static_cast<size_t>(line) * regs.pitch_in;
+                const GPUVAddr dest_line =
+                    regs.offset_out + static_cast<size_t>(line) * regs.pitch_out;
+                memory_manager.CopyBlock(dest_line, source_line, regs.line_length_in);
+            }
         } else {
-            CopyPitchToBlockLinear();
+            if (!is_src_pitch && is_dst_pitch) {
+                CopyBlockLinearToPitch();
+            } else {
+                CopyPitchToBlockLinear();
+            }
+        }
+    } else {
+        // TODO: allow multisized components.
+        auto& accelerate = rasterizer->AccessAccelerateDMA();
+        const bool is_const_a_dst = regs.remap_const.dst_x == RemapConst::Swizzle::CONST_A;
+        if (regs.launch_dma.remap_enable != 0 && is_const_a_dst) {
+            ASSERT(regs.remap_const.component_size_minus_one == 3);
+            accelerate.BufferClear(regs.offset_out, regs.line_length_in, regs.remap_consta_value);
+            std::vector<u32> tmp_buffer(regs.line_length_in, regs.remap_consta_value);
+            memory_manager.WriteBlockUnsafe(regs.offset_out,
+                                            reinterpret_cast<u8*>(tmp_buffer.data()),
+                                            regs.line_length_in * sizeof(u32));
+        } else {
+            auto convert_linear_2_blocklinear_addr = [](u64 address) {
+                return (address & ~0x1f0ULL) | ((address & 0x40) >> 2) | ((address & 0x10) << 1) |
+                       ((address & 0x180) >> 1) | ((address & 0x20) << 3);
+            };
+            auto src_kind = memory_manager.GetPageKind(regs.offset_in);
+            auto dst_kind = memory_manager.GetPageKind(regs.offset_out);
+            const bool is_src_pitch = IsPitchKind(static_cast<PTEKind>(src_kind));
+            const bool is_dst_pitch = IsPitchKind(static_cast<PTEKind>(dst_kind));
+            if (!is_src_pitch && is_dst_pitch) {
+                std::vector<u8> tmp_buffer(regs.line_length_in);
+                std::vector<u8> dst_buffer(regs.line_length_in);
+                memory_manager.ReadBlockUnsafe(regs.offset_in, tmp_buffer.data(),
+                                               regs.line_length_in);
+                for (u32 offset = 0; offset < regs.line_length_in; ++offset) {
+                    dst_buffer[offset] =
+                        tmp_buffer[convert_linear_2_blocklinear_addr(regs.offset_in + offset) -
+                                   regs.offset_in];
+                }
+                memory_manager.WriteBlock(regs.offset_out, dst_buffer.data(), regs.line_length_in);
+            } else if (is_src_pitch && !is_dst_pitch) {
+                std::vector<u8> tmp_buffer(regs.line_length_in);
+                std::vector<u8> dst_buffer(regs.line_length_in);
+                memory_manager.ReadBlockUnsafe(regs.offset_in, tmp_buffer.data(),
+                                               regs.line_length_in);
+                for (u32 offset = 0; offset < regs.line_length_in; ++offset) {
+                    dst_buffer[convert_linear_2_blocklinear_addr(regs.offset_out + offset) -
+                               regs.offset_out] = tmp_buffer[offset];
+                }
+                memory_manager.WriteBlock(regs.offset_out, dst_buffer.data(), regs.line_length_in);
+            } else {
+                if (!accelerate.BufferCopy(regs.offset_in, regs.offset_out, regs.line_length_in)) {
+                    std::vector<u8> tmp_buffer(regs.line_length_in);
+                    memory_manager.ReadBlockUnsafe(regs.offset_in, tmp_buffer.data(),
+                                                   regs.line_length_in);
+                    memory_manager.WriteBlock(regs.offset_out, tmp_buffer.data(),
+                                              regs.line_length_in);
+                }
+            }
         }
     }
+
     ReleaseSemaphore();
 }
 
-void MaxwellDMA::CopyPitchToPitch() {
-    // When `multi_line_enable` bit is enabled we copy a 2D image of dimensions
-    // (line_length_in, line_count).
-    // Otherwise the copy is performed as if we were copying a 1D buffer of length line_length_in.
-    const bool remap_enabled = regs.launch_dma.remap_enable != 0;
-    if (regs.launch_dma.multi_line_enable) {
-        UNIMPLEMENTED_IF(remap_enabled);
-
-        // Perform a line-by-line copy.
-        // We're going to take a subrect of size (line_length_in, line_count) from the source
-        // rectangle. There is no need to manually flush/invalidate the regions because CopyBlock
-        // does that for us.
-        for (u32 line = 0; line < regs.line_count; ++line) {
-            const GPUVAddr source_line = regs.offset_in + static_cast<size_t>(line) * regs.pitch_in;
-            const GPUVAddr dest_line = regs.offset_out + static_cast<size_t>(line) * regs.pitch_out;
-            memory_manager.CopyBlock(dest_line, source_line, regs.line_length_in);
-        }
-        return;
-    }
-    // TODO: allow multisized components.
-    auto& accelerate = rasterizer->AccessAccelerateDMA();
-    const bool is_const_a_dst = regs.remap_const.dst_x == RemapConst::Swizzle::CONST_A;
-    const bool is_buffer_clear = remap_enabled && is_const_a_dst;
-    if (is_buffer_clear) {
-        ASSERT(regs.remap_const.component_size_minus_one == 3);
-        accelerate.BufferClear(regs.offset_out, regs.line_length_in, regs.remap_consta_value);
-        std::vector<u32> tmp_buffer(regs.line_length_in, regs.remap_consta_value);
-        memory_manager.WriteBlockUnsafe(regs.offset_out, reinterpret_cast<u8*>(tmp_buffer.data()),
-                                        regs.line_length_in * sizeof(u32));
-        return;
-    }
-    UNIMPLEMENTED_IF(remap_enabled);
-    if (!accelerate.BufferCopy(regs.offset_in, regs.offset_out, regs.line_length_in)) {
-        std::vector<u8> tmp_buffer(regs.line_length_in);
-        memory_manager.ReadBlockUnsafe(regs.offset_in, tmp_buffer.data(), regs.line_length_in);
-        memory_manager.WriteBlock(regs.offset_out, tmp_buffer.data(), regs.line_length_in);
-    }
-}
-
 void MaxwellDMA::CopyBlockLinearToPitch() {
     UNIMPLEMENTED_IF(regs.src_params.block_size.width != 0);
     UNIMPLEMENTED_IF(regs.src_params.layer != 0);

src/video_core/engines/maxwell_dma.h

@@ -219,8 +219,6 @@ private:
     /// registers.
     void Launch();
 
-    void CopyPitchToPitch();
-
     void CopyBlockLinearToPitch();
 
     void CopyPitchToBlockLinear();

src/video_core/macro/macro_hle.cpp

@@ -22,35 +22,29 @@ void HLE_771BB18C62444DA0(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
     maxwell3d.regs.draw.topology.Assign(
         static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0] & 0x3ffffff));
     maxwell3d.regs.global_base_instance_index = parameters[5];
-    maxwell3d.mme_draw.instance_count = instance_count;
     maxwell3d.regs.global_base_vertex_index = parameters[3];
     maxwell3d.regs.index_buffer.count = parameters[1];
     maxwell3d.regs.index_buffer.first = parameters[4];
 
     if (maxwell3d.ShouldExecute()) {
-        maxwell3d.Rasterizer().Draw(true, true);
+        maxwell3d.Rasterizer().Draw(true, instance_count);
     }
     maxwell3d.regs.index_buffer.count = 0;
-    maxwell3d.mme_draw.instance_count = 0;
-    maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
 }
 
 void HLE_0D61FC9FAAC9FCAD(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& parameters) {
-    const u32 count = (maxwell3d.GetRegisterValue(0xD1B) & parameters[2]);
+    const u32 instance_count = (maxwell3d.GetRegisterValue(0xD1B) & parameters[2]);
 
     maxwell3d.regs.vertex_buffer.first = parameters[3];
     maxwell3d.regs.vertex_buffer.count = parameters[1];
     maxwell3d.regs.global_base_instance_index = parameters[4];
     maxwell3d.regs.draw.topology.Assign(
         static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0]));
-    maxwell3d.mme_draw.instance_count = count;
 
     if (maxwell3d.ShouldExecute()) {
-        maxwell3d.Rasterizer().Draw(false, true);
+        maxwell3d.Rasterizer().Draw(false, instance_count);
     }
     maxwell3d.regs.vertex_buffer.count = 0;
-    maxwell3d.mme_draw.instance_count = 0;
-    maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
 }
 
 void HLE_0217920100488FF7(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& parameters) {
@@ -63,24 +57,21 @@ void HLE_0217920100488FF7(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
     maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
     maxwell3d.regs.global_base_vertex_index = element_base;
     maxwell3d.regs.global_base_instance_index = base_instance;
-    maxwell3d.mme_draw.instance_count = instance_count;
-    maxwell3d.CallMethodFromMME(0x8e3, 0x640);
-    maxwell3d.CallMethodFromMME(0x8e4, element_base);
-    maxwell3d.CallMethodFromMME(0x8e5, base_instance);
+    maxwell3d.CallMethod(0x8e3, 0x640, true);
+    maxwell3d.CallMethod(0x8e4, element_base, true);
+    maxwell3d.CallMethod(0x8e5, base_instance, true);
     maxwell3d.regs.draw.topology.Assign(
         static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0]));
     if (maxwell3d.ShouldExecute()) {
-        maxwell3d.Rasterizer().Draw(true, true);
+        maxwell3d.Rasterizer().Draw(true, instance_count);
     }
     maxwell3d.regs.vertex_id_base = 0x0;
     maxwell3d.regs.index_buffer.count = 0;
     maxwell3d.regs.global_base_vertex_index = 0x0;
     maxwell3d.regs.global_base_instance_index = 0x0;
-    maxwell3d.mme_draw.instance_count = 0;
-    maxwell3d.CallMethodFromMME(0x8e3, 0x640);
-    maxwell3d.CallMethodFromMME(0x8e4, 0x0);
-    maxwell3d.CallMethodFromMME(0x8e5, 0x0);
-    maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
+    maxwell3d.CallMethod(0x8e3, 0x640, true);
+    maxwell3d.CallMethod(0x8e4, 0x0, true);
+    maxwell3d.CallMethod(0x8e5, 0x0, true);
 }
 
 // Multidraw Indirect
@@ -91,11 +82,9 @@ void HLE_3F5E74B9C9A50164(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
         maxwell3d.regs.index_buffer.count = 0;
         maxwell3d.regs.global_base_vertex_index = 0x0;
         maxwell3d.regs.global_base_instance_index = 0x0;
-        maxwell3d.mme_draw.instance_count = 0;
-        maxwell3d.CallMethodFromMME(0x8e3, 0x640);
-        maxwell3d.CallMethodFromMME(0x8e4, 0x0);
-        maxwell3d.CallMethodFromMME(0x8e5, 0x0);
-        maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
+        maxwell3d.CallMethod(0x8e3, 0x640, true);
+        maxwell3d.CallMethod(0x8e4, 0x0, true);
+        maxwell3d.CallMethod(0x8e5, 0x0, true);
         maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
     });
     const u32 start_indirect = parameters[0];
@@ -127,15 +116,13 @@ void HLE_3F5E74B9C9A50164(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
         maxwell3d.regs.index_buffer.count = num_vertices;
         maxwell3d.regs.global_base_vertex_index = base_vertex;
         maxwell3d.regs.global_base_instance_index = base_instance;
-        maxwell3d.mme_draw.instance_count = instance_count;
-        maxwell3d.CallMethodFromMME(0x8e3, 0x640);
-        maxwell3d.CallMethodFromMME(0x8e4, base_vertex);
-        maxwell3d.CallMethodFromMME(0x8e5, base_instance);
+        maxwell3d.CallMethod(0x8e3, 0x640, true);
+        maxwell3d.CallMethod(0x8e4, base_vertex, true);
+        maxwell3d.CallMethod(0x8e5, base_instance, true);
         maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
         if (maxwell3d.ShouldExecute()) {
-            maxwell3d.Rasterizer().Draw(true, true);
+            maxwell3d.Rasterizer().Draw(true, instance_count);
         }
-        maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
     }
 }
 

src/video_core/macro/macro_interpreter.cpp

@@ -335,7 +335,7 @@ void MacroInterpreterImpl::SetMethodAddress(u32 address) {
 }
 
 void MacroInterpreterImpl::Send(u32 value) {
-    maxwell3d.CallMethodFromMME(method_address.address, value);
+    maxwell3d.CallMethod(method_address.address, value, true);
     // Increment the method address by the method increment.
     method_address.address.Assign(method_address.address.Value() +
                                   method_address.increment.Value());

src/video_core/macro/macro_jit_x64.cpp

@@ -346,7 +346,7 @@ void MacroJITx64Impl::Compile_Read(Macro::Opcode opcode) {
 }
 
 void Send(Engines::Maxwell3D* maxwell3d, Macro::MethodAddress method_address, u32 value) {
-    maxwell3d->CallMethodFromMME(method_address.address, value);
+    maxwell3d->CallMethod(method_address.address, value, true);
 }
 
 void MacroJITx64Impl::Compile_Send(Xbyak::Reg32 value) {

src/video_core/memory_manager.cpp

@@ -41,7 +41,11 @@ MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64
     big_entries.resize(big_page_table_size / 32, 0);
     big_page_table_cpu.resize(big_page_table_size);
     big_page_continous.resize(big_page_table_size / continous_bits, 0);
+    std::array<PTEKind, 32> kind_valus;
+    kind_valus.fill(PTEKind::INVALID);
+    big_kinds.resize(big_page_table_size / 32, kind_valus);
     entries.resize(page_table_size / 32, 0);
+    kinds.resize(big_page_table_size / 32, kind_valus);
 }
 
 MemoryManager::~MemoryManager() = default;
@@ -78,6 +82,41 @@ void MemoryManager::SetEntry(size_t position, MemoryManager::EntryType entry) {
     }
 }
 
+PTEKind MemoryManager::GetPageKind(GPUVAddr gpu_addr) const {
+    auto entry = GetEntry<true>(gpu_addr);
+    if (entry == EntryType::Mapped || entry == EntryType::Reserved) [[likely]] {
+        return GetKind<true>(gpu_addr);
+    } else {
+        return GetKind<false>(gpu_addr);
+    }
+}
+
+template <bool is_big_page>
+PTEKind MemoryManager::GetKind(size_t position) const {
+    if constexpr (is_big_page) {
+        position = position >> big_page_bits;
+        const size_t sub_index = position % 32;
+        return big_kinds[position / 32][sub_index];
+    } else {
+        position = position >> page_bits;
+        const size_t sub_index = position % 32;
+        return kinds[position / 32][sub_index];
+    }
+}
+
+template <bool is_big_page>
+void MemoryManager::SetKind(size_t position, PTEKind kind) {
+    if constexpr (is_big_page) {
+        position = position >> big_page_bits;
+        const size_t sub_index = position % 32;
+        big_kinds[position / 32][sub_index] = kind;
+    } else {
+        position = position >> page_bits;
+        const size_t sub_index = position % 32;
+        kinds[position / 32][sub_index] = kind;
+    }
+}
+
 inline bool MemoryManager::IsBigPageContinous(size_t big_page_index) const {
     const u64 entry_mask = big_page_continous[big_page_index / continous_bits];
     const size_t sub_index = big_page_index % continous_bits;
@@ -92,8 +131,8 @@ inline void MemoryManager::SetBigPageContinous(size_t big_page_index, bool value
 }
 
 template <MemoryManager::EntryType entry_type>
-GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr,
-                                    size_t size) {
+GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size,
+                                    PTEKind kind) {
     u64 remaining_size{size};
     if constexpr (entry_type == EntryType::Mapped) {
         page_table.ReserveRange(gpu_addr, size);
@@ -102,6 +141,7 @@ GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cp
         const GPUVAddr current_gpu_addr = gpu_addr + offset;
         [[maybe_unused]] const auto current_entry_type = GetEntry<false>(current_gpu_addr);
         SetEntry<false>(current_gpu_addr, entry_type);
+        SetKind<false>(current_gpu_addr, kind);
         if (current_entry_type != entry_type) {
             rasterizer->ModifyGPUMemory(unique_identifier, gpu_addr, page_size);
         }
@@ -118,12 +158,13 @@ GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cp
 
 template <MemoryManager::EntryType entry_type>
 GPUVAddr MemoryManager::BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr,
-                                       size_t size) {
+                                       size_t size, PTEKind kind) {
     u64 remaining_size{size};
     for (u64 offset{}; offset < size; offset += big_page_size) {
         const GPUVAddr current_gpu_addr = gpu_addr + offset;
         [[maybe_unused]] const auto current_entry_type = GetEntry<true>(current_gpu_addr);
         SetEntry<true>(current_gpu_addr, entry_type);
+        SetKind<true>(current_gpu_addr, kind);
         if (current_entry_type != entry_type) {
             rasterizer->ModifyGPUMemory(unique_identifier, gpu_addr, big_page_size);
         }
@@ -159,19 +200,19 @@ void MemoryManager::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_)
     rasterizer = rasterizer_;
 }
 
-GPUVAddr MemoryManager::Map(GPUVAddr gpu_addr, VAddr cpu_addr, std::size_t size,
+GPUVAddr MemoryManager::Map(GPUVAddr gpu_addr, VAddr cpu_addr, std::size_t size, PTEKind kind,
                             bool is_big_pages) {
     if (is_big_pages) [[likely]] {
-        return BigPageTableOp<EntryType::Mapped>(gpu_addr, cpu_addr, size);
+        return BigPageTableOp<EntryType::Mapped>(gpu_addr, cpu_addr, size, kind);
     }
-    return PageTableOp<EntryType::Mapped>(gpu_addr, cpu_addr, size);
+    return PageTableOp<EntryType::Mapped>(gpu_addr, cpu_addr, size, kind);
 }
 
 GPUVAddr MemoryManager::MapSparse(GPUVAddr gpu_addr, std::size_t size, bool is_big_pages) {
     if (is_big_pages) [[likely]] {
-        return BigPageTableOp<EntryType::Reserved>(gpu_addr, 0, size);
+        return BigPageTableOp<EntryType::Reserved>(gpu_addr, 0, size, PTEKind::INVALID);
     }
-    return PageTableOp<EntryType::Reserved>(gpu_addr, 0, size);
+    return PageTableOp<EntryType::Reserved>(gpu_addr, 0, size, PTEKind::INVALID);
 }
 
 void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
@@ -188,8 +229,8 @@ void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
         rasterizer->UnmapMemory(*cpu_addr, map_size);
     }
 
-    BigPageTableOp<EntryType::Free>(gpu_addr, 0, size);
-    PageTableOp<EntryType::Free>(gpu_addr, 0, size);
+    BigPageTableOp<EntryType::Free>(gpu_addr, 0, size, PTEKind::INVALID);
+    PageTableOp<EntryType::Free>(gpu_addr, 0, size, PTEKind::INVALID);
 }
 
 std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) const {

src/video_core/memory_manager.h

@@ -11,6 +11,7 @@
 #include "common/common_types.h"
 #include "common/multi_level_page_table.h"
 #include "common/virtual_buffer.h"
+#include "video_core/pte_kind.h"
 
 namespace VideoCore {
 class RasterizerInterface;
@@ -98,7 +99,8 @@ public:
     std::vector<std::pair<GPUVAddr, std::size_t>> GetSubmappedRange(GPUVAddr gpu_addr,
                                                                     std::size_t size) const;
 
-    GPUVAddr Map(GPUVAddr gpu_addr, VAddr cpu_addr, std::size_t size, bool is_big_pages = true);
+    GPUVAddr Map(GPUVAddr gpu_addr, VAddr cpu_addr, std::size_t size,
+                 PTEKind kind = PTEKind::INVALID, bool is_big_pages = true);
     GPUVAddr MapSparse(GPUVAddr gpu_addr, std::size_t size, bool is_big_pages = true);
     void Unmap(GPUVAddr gpu_addr, std::size_t size);
 
@@ -114,6 +116,8 @@ public:
         return gpu_addr < address_space_size;
     }
 
+    PTEKind GetPageKind(GPUVAddr gpu_addr) const;
+
 private:
     template <bool is_big_pages, typename FuncMapped, typename FuncReserved, typename FuncUnmapped>
     inline void MemoryOperation(GPUVAddr gpu_src_addr, std::size_t size, FuncMapped&& func_mapped,
@@ -166,10 +170,12 @@ private:
     std::vector<u64> big_entries;
 
     template <EntryType entry_type>
-    GPUVAddr PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size);
+    GPUVAddr PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size,
+                         PTEKind kind);
 
     template <EntryType entry_type>
-    GPUVAddr BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size);
+    GPUVAddr BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size,
+                            PTEKind kind);
 
     template <bool is_big_page>
     inline EntryType GetEntry(size_t position) const;
@@ -177,6 +183,15 @@ private:
     template <bool is_big_page>
     inline void SetEntry(size_t position, EntryType entry);
 
+    std::vector<std::array<PTEKind, 32>> kinds;
+    std::vector<std::array<PTEKind, 32>> big_kinds;
+
+    template <bool is_big_page>
+    inline PTEKind GetKind(size_t position) const;
+
+    template <bool is_big_page>
+    inline void SetKind(size_t position, PTEKind kind);
+
     Common::MultiLevelPageTable<u32> page_table;
     Common::VirtualBuffer<u32> big_page_table_cpu;
 

src/video_core/pte_kind.h

@@ -0,0 +1,264 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Tegra {
+
+// https://github.com/NVIDIA/open-gpu-doc/blob/master/manuals/volta/gv100/dev_mmu.ref.txt
+enum class PTEKind : u8 {
+    INVALID = 0xff,
+    PITCH = 0x00,
+    Z16 = 0x01,
+    Z16_2C = 0x02,
+    Z16_MS2_2C = 0x03,
+    Z16_MS4_2C = 0x04,
+    Z16_MS8_2C = 0x05,
+    Z16_MS16_2C = 0x06,
+    Z16_2Z = 0x07,
+    Z16_MS2_2Z = 0x08,
+    Z16_MS4_2Z = 0x09,
+    Z16_MS8_2Z = 0x0a,
+    Z16_MS16_2Z = 0x0b,
+    Z16_2CZ = 0x36,
+    Z16_MS2_2CZ = 0x37,
+    Z16_MS4_2CZ = 0x38,
+    Z16_MS8_2CZ = 0x39,
+    Z16_MS16_2CZ = 0x5f,
+    Z16_4CZ = 0x0c,
+    Z16_MS2_4CZ = 0x0d,
+    Z16_MS4_4CZ = 0x0e,
+    Z16_MS8_4CZ = 0x0f,
+    Z16_MS16_4CZ = 0x10,
+    S8Z24 = 0x11,
+    S8Z24_1Z = 0x12,
+    S8Z24_MS2_1Z = 0x13,
+    S8Z24_MS4_1Z = 0x14,
+    S8Z24_MS8_1Z = 0x15,
+    S8Z24_MS16_1Z = 0x16,
+    S8Z24_2CZ = 0x17,
+    S8Z24_MS2_2CZ = 0x18,
+    S8Z24_MS4_2CZ = 0x19,
+    S8Z24_MS8_2CZ = 0x1a,
+    S8Z24_MS16_2CZ = 0x1b,
+    S8Z24_2CS = 0x1c,
+    S8Z24_MS2_2CS = 0x1d,
+    S8Z24_MS4_2CS = 0x1e,
+    S8Z24_MS8_2CS = 0x1f,
+    S8Z24_MS16_2CS = 0x20,
+    S8Z24_4CSZV = 0x21,
+    S8Z24_MS2_4CSZV = 0x22,
+    S8Z24_MS4_4CSZV = 0x23,
+    S8Z24_MS8_4CSZV = 0x24,
+    S8Z24_MS16_4CSZV = 0x25,
+    V8Z24_MS4_VC12 = 0x26,
+    V8Z24_MS4_VC4 = 0x27,
+    V8Z24_MS8_VC8 = 0x28,
+    V8Z24_MS8_VC24 = 0x29,
+    V8Z24_MS4_VC12_1ZV = 0x2e,
+    V8Z24_MS4_VC4_1ZV = 0x2f,
+    V8Z24_MS8_VC8_1ZV = 0x30,
+    V8Z24_MS8_VC24_1ZV = 0x31,
+    V8Z24_MS4_VC12_2CS = 0x32,
+    V8Z24_MS4_VC4_2CS = 0x33,
+    V8Z24_MS8_VC8_2CS = 0x34,
+    V8Z24_MS8_VC24_2CS = 0x35,
+    V8Z24_MS4_VC12_2CZV = 0x3a,
+    V8Z24_MS4_VC4_2CZV = 0x3b,
+    V8Z24_MS8_VC8_2CZV = 0x3c,
+    V8Z24_MS8_VC24_2CZV = 0x3d,
+    V8Z24_MS4_VC12_2ZV = 0x3e,
+    V8Z24_MS4_VC4_2ZV = 0x3f,
+    V8Z24_MS8_VC8_2ZV = 0x40,
+    V8Z24_MS8_VC24_2ZV = 0x41,
+    V8Z24_MS4_VC12_4CSZV = 0x42,
+    V8Z24_MS4_VC4_4CSZV = 0x43,
+    V8Z24_MS8_VC8_4CSZV = 0x44,
+    V8Z24_MS8_VC24_4CSZV = 0x45,
+    Z24S8 = 0x46,
+    Z24S8_1Z = 0x47,
+    Z24S8_MS2_1Z = 0x48,
+    Z24S8_MS4_1Z = 0x49,
+    Z24S8_MS8_1Z = 0x4a,
+    Z24S8_MS16_1Z = 0x4b,
+    Z24S8_2CS = 0x4c,
+    Z24S8_MS2_2CS = 0x4d,
+    Z24S8_MS4_2CS = 0x4e,
+    Z24S8_MS8_2CS = 0x4f,
+    Z24S8_MS16_2CS = 0x50,
+    Z24S8_2CZ = 0x51,
+    Z24S8_MS2_2CZ = 0x52,
+    Z24S8_MS4_2CZ = 0x53,
+    Z24S8_MS8_2CZ = 0x54,
+    Z24S8_MS16_2CZ = 0x55,
+    Z24S8_4CSZV = 0x56,
+    Z24S8_MS2_4CSZV = 0x57,
+    Z24S8_MS4_4CSZV = 0x58,
+    Z24S8_MS8_4CSZV = 0x59,
+    Z24S8_MS16_4CSZV = 0x5a,
+    Z24V8_MS4_VC12 = 0x5b,
+    Z24V8_MS4_VC4 = 0x5c,
+    Z24V8_MS8_VC8 = 0x5d,
+    Z24V8_MS8_VC24 = 0x5e,
+    YUV_B8C1_2Y = 0x60,
+    YUV_B8C2_2Y = 0x61,
+    YUV_B10C1_2Y = 0x62,
+    YUV_B10C2_2Y = 0x6b,
+    YUV_B12C1_2Y = 0x6c,
+    YUV_B12C2_2Y = 0x6d,
+    Z24V8_MS4_VC12_1ZV = 0x63,
+    Z24V8_MS4_VC4_1ZV = 0x64,
+    Z24V8_MS8_VC8_1ZV = 0x65,
+    Z24V8_MS8_VC24_1ZV = 0x66,
+    Z24V8_MS4_VC12_2CS = 0x67,
+    Z24V8_MS4_VC4_2CS = 0x68,
+    Z24V8_MS8_VC8_2CS = 0x69,
+    Z24V8_MS8_VC24_2CS = 0x6a,
+    Z24V8_MS4_VC12_2CZV = 0x6f,
+    Z24V8_MS4_VC4_2CZV = 0x70,
+    Z24V8_MS8_VC8_2CZV = 0x71,
+    Z24V8_MS8_VC24_2CZV = 0x72,
+    Z24V8_MS4_VC12_2ZV = 0x73,
+    Z24V8_MS4_VC4_2ZV = 0x74,
+    Z24V8_MS8_VC8_2ZV = 0x75,
+    Z24V8_MS8_VC24_2ZV = 0x76,
+    Z24V8_MS4_VC12_4CSZV = 0x77,
+    Z24V8_MS4_VC4_4CSZV = 0x78,
+    Z24V8_MS8_VC8_4CSZV = 0x79,
+    Z24V8_MS8_VC24_4CSZV = 0x7a,
+    ZF32 = 0x7b,
+    ZF32_1Z = 0x7c,
+    ZF32_MS2_1Z = 0x7d,
+    ZF32_MS4_1Z = 0x7e,
+    ZF32_MS8_1Z = 0x7f,
+    ZF32_MS16_1Z = 0x80,
+    ZF32_2CS = 0x81,
+    ZF32_MS2_2CS = 0x82,
+    ZF32_MS4_2CS = 0x83,
+    ZF32_MS8_2CS = 0x84,
+    ZF32_MS16_2CS = 0x85,
+    ZF32_2CZ = 0x86,
+    ZF32_MS2_2CZ = 0x87,
+    ZF32_MS4_2CZ = 0x88,
+    ZF32_MS8_2CZ = 0x89,
+    ZF32_MS16_2CZ = 0x8a,
+    X8Z24_X16V8S8_MS4_VC12 = 0x8b,
+    X8Z24_X16V8S8_MS4_VC4 = 0x8c,
+    X8Z24_X16V8S8_MS8_VC8 = 0x8d,
+    X8Z24_X16V8S8_MS8_VC24 = 0x8e,
+    X8Z24_X16V8S8_MS4_VC12_1CS = 0x8f,
+    X8Z24_X16V8S8_MS4_VC4_1CS = 0x90,
+    X8Z24_X16V8S8_MS8_VC8_1CS = 0x91,
+    X8Z24_X16V8S8_MS8_VC24_1CS = 0x92,
+    X8Z24_X16V8S8_MS4_VC12_1ZV = 0x97,
+    X8Z24_X16V8S8_MS4_VC4_1ZV = 0x98,
+    X8Z24_X16V8S8_MS8_VC8_1ZV = 0x99,
+    X8Z24_X16V8S8_MS8_VC24_1ZV = 0x9a,
+    X8Z24_X16V8S8_MS4_VC12_1CZV = 0x9b,
+    X8Z24_X16V8S8_MS4_VC4_1CZV = 0x9c,
+    X8Z24_X16V8S8_MS8_VC8_1CZV = 0x9d,
+    X8Z24_X16V8S8_MS8_VC24_1CZV = 0x9e,
+    X8Z24_X16V8S8_MS4_VC12_2CS = 0x9f,
+    X8Z24_X16V8S8_MS4_VC4_2CS = 0xa0,
+    X8Z24_X16V8S8_MS8_VC8_2CS = 0xa1,
+    X8Z24_X16V8S8_MS8_VC24_2CS = 0xa2,
+    X8Z24_X16V8S8_MS4_VC12_2CSZV = 0xa3,
+    X8Z24_X16V8S8_MS4_VC4_2CSZV = 0xa4,
+    X8Z24_X16V8S8_MS8_VC8_2CSZV = 0xa5,
+    X8Z24_X16V8S8_MS8_VC24_2CSZV = 0xa6,
+    ZF32_X16V8S8_MS4_VC12 = 0xa7,
+    ZF32_X16V8S8_MS4_VC4 = 0xa8,
+    ZF32_X16V8S8_MS8_VC8 = 0xa9,
+    ZF32_X16V8S8_MS8_VC24 = 0xaa,
+    ZF32_X16V8S8_MS4_VC12_1CS = 0xab,
+    ZF32_X16V8S8_MS4_VC4_1CS = 0xac,
+    ZF32_X16V8S8_MS8_VC8_1CS = 0xad,
+    ZF32_X16V8S8_MS8_VC24_1CS = 0xae,
+    ZF32_X16V8S8_MS4_VC12_1ZV = 0xb3,
+    ZF32_X16V8S8_MS4_VC4_1ZV = 0xb4,
+    ZF32_X16V8S8_MS8_VC8_1ZV = 0xb5,
+    ZF32_X16V8S8_MS8_VC24_1ZV = 0xb6,
+    ZF32_X16V8S8_MS4_VC12_1CZV = 0xb7,
+    ZF32_X16V8S8_MS4_VC4_1CZV = 0xb8,
+    ZF32_X16V8S8_MS8_VC8_1CZV = 0xb9,
+    ZF32_X16V8S8_MS8_VC24_1CZV = 0xba,
+    ZF32_X16V8S8_MS4_VC12_2CS = 0xbb,
+    ZF32_X16V8S8_MS4_VC4_2CS = 0xbc,
+    ZF32_X16V8S8_MS8_VC8_2CS = 0xbd,
+    ZF32_X16V8S8_MS8_VC24_2CS = 0xbe,
+    ZF32_X16V8S8_MS4_VC12_2CSZV = 0xbf,
+    ZF32_X16V8S8_MS4_VC4_2CSZV = 0xc0,
+    ZF32_X16V8S8_MS8_VC8_2CSZV = 0xc1,
+    ZF32_X16V8S8_MS8_VC24_2CSZV = 0xc2,
+    ZF32_X24S8 = 0xc3,
+    ZF32_X24S8_1CS = 0xc4,
+    ZF32_X24S8_MS2_1CS = 0xc5,
+    ZF32_X24S8_MS4_1CS = 0xc6,
+    ZF32_X24S8_MS8_1CS = 0xc7,
+    ZF32_X24S8_MS16_1CS = 0xc8,
+    ZF32_X24S8_2CSZV = 0xce,
+    ZF32_X24S8_MS2_2CSZV = 0xcf,
+    ZF32_X24S8_MS4_2CSZV = 0xd0,
+    ZF32_X24S8_MS8_2CSZV = 0xd1,
+    ZF32_X24S8_MS16_2CSZV = 0xd2,
+    ZF32_X24S8_2CS = 0xd3,
+    ZF32_X24S8_MS2_2CS = 0xd4,
+    ZF32_X24S8_MS4_2CS = 0xd5,
+    ZF32_X24S8_MS8_2CS = 0xd6,
+    ZF32_X24S8_MS16_2CS = 0xd7,
+    S8 = 0x2a,
+    S8_2S = 0x2b,
+    GENERIC_16BX2 = 0xfe,
+    C32_2C = 0xd8,
+    C32_2CBR = 0xd9,
+    C32_2CBA = 0xda,
+    C32_2CRA = 0xdb,
+    C32_2BRA = 0xdc,
+    C32_MS2_2C = 0xdd,
+    C32_MS2_2CBR = 0xde,
+    C32_MS2_4CBRA = 0xcc,
+    C32_MS4_2C = 0xdf,
+    C32_MS4_2CBR = 0xe0,
+    C32_MS4_2CBA = 0xe1,
+    C32_MS4_2CRA = 0xe2,
+    C32_MS4_2BRA = 0xe3,
+    C32_MS4_4CBRA = 0x2c,
+    C32_MS8_MS16_2C = 0xe4,
+    C32_MS8_MS16_2CRA = 0xe5,
+    C64_2C = 0xe6,
+    C64_2CBR = 0xe7,
+    C64_2CBA = 0xe8,
+    C64_2CRA = 0xe9,
+    C64_2BRA = 0xea,
+    C64_MS2_2C = 0xeb,
+    C64_MS2_2CBR = 0xec,
+    C64_MS2_4CBRA = 0xcd,
+    C64_MS4_2C = 0xed,
+    C64_MS4_2CBR = 0xee,
+    C64_MS4_2CBA = 0xef,
+    C64_MS4_2CRA = 0xf0,
+    C64_MS4_2BRA = 0xf1,
+    C64_MS4_4CBRA = 0x2d,
+    C64_MS8_MS16_2C = 0xf2,
+    C64_MS8_MS16_2CRA = 0xf3,
+    C128_2C = 0xf4,
+    C128_2CR = 0xf5,
+    C128_MS2_2C = 0xf6,
+    C128_MS2_2CR = 0xf7,
+    C128_MS4_2C = 0xf8,
+    C128_MS4_2CR = 0xf9,
+    C128_MS8_MS16_2C = 0xfa,
+    C128_MS8_MS16_2CR = 0xfb,
+    X8C24 = 0xfc,
+    PITCH_NO_SWIZZLE = 0xfd,
+    SMSKED_MESSAGE = 0xca,
+    SMHOST_MESSAGE = 0xcb,
+};
+
+constexpr bool IsPitchKind(PTEKind kind) {
+    return kind == PTEKind::PITCH || kind == PTEKind::PITCH_NO_SWIZZLE;
+}
+
+} // namespace Tegra

src/video_core/rasterizer_interface.h

@@ -40,7 +40,7 @@ public:
     virtual ~RasterizerInterface() = default;
 
     /// Dispatches a draw invocation
-    virtual void Draw(bool is_indexed, bool is_instanced) = 0;
+    virtual void Draw(bool is_indexed, u32 instance_count) = 0;
 
     /// Clear the current framebuffer
     virtual void Clear() = 0;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -205,7 +205,7 @@ void RasterizerOpenGL::Clear() {
     ++num_queued_commands;
 }
 
-void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
+void RasterizerOpenGL::Draw(bool is_indexed, u32 instance_count) {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
@@ -222,14 +222,15 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     pipeline->SetEngine(maxwell3d, gpu_memory);
     pipeline->Configure(is_indexed);
 
+    BindInlineIndexBuffer();
+
     SyncState();
 
     const GLenum primitive_mode = MaxwellToGL::PrimitiveTopology(maxwell3d->regs.draw.topology);
     BeginTransformFeedback(pipeline, primitive_mode);
 
     const GLuint base_instance = static_cast<GLuint>(maxwell3d->regs.global_base_instance_index);
-    const GLsizei num_instances =
-        static_cast<GLsizei>(is_instanced ? maxwell3d->mme_draw.instance_count : 1);
+    const GLsizei num_instances = static_cast<GLsizei>(instance_count);
     if (is_indexed) {
         const GLint base_vertex = static_cast<GLint>(maxwell3d->regs.global_base_vertex_index);
         const GLsizei num_vertices = static_cast<GLsizei>(maxwell3d->regs.index_buffer.count);
@@ -1129,6 +1130,16 @@ void RasterizerOpenGL::ReleaseChannel(s32 channel_id) {
     query_cache.EraseChannel(channel_id);
 }
 
+void RasterizerOpenGL::BindInlineIndexBuffer() {
+    if (maxwell3d->inline_index_draw_indexes.empty()) {
+        return;
+    }
+    const auto data_count = static_cast<u32>(maxwell3d->inline_index_draw_indexes.size());
+    auto buffer = Buffer(buffer_cache_runtime, *this, 0, data_count);
+    buffer.ImmediateUpload(0, maxwell3d->inline_index_draw_indexes);
+    buffer_cache_runtime.BindIndexBuffer(buffer, 0, data_count);
+}
+
 AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_) : buffer_cache{buffer_cache_} {}
 
 bool AccelerateDMA::BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -68,7 +68,7 @@ public:
                               StateTracker& state_tracker_);
     ~RasterizerOpenGL() override;
 
-    void Draw(bool is_indexed, bool is_instanced) override;
+    void Draw(bool is_indexed, u32 instance_count) override;
     void Clear() override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
@@ -199,6 +199,8 @@ private:
     /// End a transform feedback
     void EndTransformFeedback();
 
+    void BindInlineIndexBuffer();
+
     Tegra::GPU& gpu;
 
     const Device& device;

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -59,10 +59,11 @@ void QueryPool::Reserve(std::pair<VkQueryPool, u32> query) {
         std::find_if(pools.begin(), pools.end(), [query_pool = query.first](vk::QueryPool& pool) {
             return query_pool == *pool;
         });
-    ASSERT(it != std::end(pools));
 
-    const std::ptrdiff_t pool_index = std::distance(std::begin(pools), it);
-    usage[pool_index * GROW_STEP + static_cast<std::ptrdiff_t>(query.second)] = false;
+    if (it != std::end(pools)) {
+        const std::ptrdiff_t pool_index = std::distance(std::begin(pools), it);
+        usage[pool_index * GROW_STEP + static_cast<std::ptrdiff_t>(query.second)] = false;
+    }
 }
 
 QueryCache::QueryCache(VideoCore::RasterizerInterface& rasterizer_, const Device& device_,

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -127,11 +127,10 @@ VkRect2D GetScissorState(const Maxwell& regs, size_t index, u32 up_scale = 1, u3
     return scissor;
 }
 
-DrawParams MakeDrawParams(const Maxwell& regs, u32 num_instances, bool is_instanced,
-                          bool is_indexed) {
+DrawParams MakeDrawParams(const Maxwell& regs, u32 num_instances, bool is_indexed) {
     DrawParams params{
         .base_instance = regs.global_base_instance_index,
-        .num_instances = is_instanced ? num_instances : 1,
+        .num_instances = num_instances,
         .base_vertex = is_indexed ? regs.global_base_vertex_index : regs.vertex_buffer.first,
         .num_vertices = is_indexed ? regs.index_buffer.count : regs.vertex_buffer.count,
         .first_index = is_indexed ? regs.index_buffer.first : 0,
@@ -177,7 +176,7 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
 
 RasterizerVulkan::~RasterizerVulkan() = default;
 
-void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
+void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
     MICROPROFILE_SCOPE(Vulkan_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
@@ -194,13 +193,15 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
     pipeline->SetEngine(maxwell3d, gpu_memory);
     pipeline->Configure(is_indexed);
 
+    BindInlineIndexBuffer();
+
     BeginTransformFeedback();
 
     UpdateDynamicStates();
 
     const auto& regs{maxwell3d->regs};
-    const u32 num_instances{maxwell3d->mme_draw.instance_count};
-    const DrawParams draw_params{MakeDrawParams(regs, num_instances, is_instanced, is_indexed)};
+    const u32 num_instances{instance_count};
+    const DrawParams draw_params{MakeDrawParams(regs, num_instances, is_indexed)};
     scheduler.Record([draw_params](vk::CommandBuffer cmdbuf) {
         if (draw_params.is_indexed) {
             cmdbuf.DrawIndexed(draw_params.num_vertices, draw_params.num_instances,
@@ -1009,4 +1010,17 @@ void RasterizerVulkan::ReleaseChannel(s32 channel_id) {
     query_cache.EraseChannel(channel_id);
 }
 
+void RasterizerVulkan::BindInlineIndexBuffer() {
+    if (maxwell3d->inline_index_draw_indexes.empty()) {
+        return;
+    }
+    const auto data_count = static_cast<u32>(maxwell3d->inline_index_draw_indexes.size());
+    auto buffer = buffer_cache_runtime.UploadStagingBuffer(data_count);
+    std::memcpy(buffer.mapped_span.data(), maxwell3d->inline_index_draw_indexes.data(), data_count);
+    buffer_cache_runtime.BindIndexBuffer(
+        maxwell3d->regs.draw.topology, maxwell3d->regs.index_buffer.format,
+        maxwell3d->regs.index_buffer.first, maxwell3d->regs.index_buffer.count, buffer.buffer,
+        static_cast<u32>(buffer.offset), data_count);
+}
+
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -64,7 +64,7 @@ public:
                               StateTracker& state_tracker_, Scheduler& scheduler_);
     ~RasterizerVulkan() override;
 
-    void Draw(bool is_indexed, bool is_instanced) override;
+    void Draw(bool is_indexed, u32 instance_count) override;
     void Clear() override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
@@ -141,6 +141,8 @@ private:
 
     void UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs);
 
+    void BindInlineIndexBuffer();
+
     Tegra::GPU& gpu;
 
     ScreenInfo& screen_info;

src/video_core/texture_cache/descriptor_table.h

@@ -18,7 +18,7 @@ class DescriptorTable {
 public:
     explicit DescriptorTable(Tegra::MemoryManager& gpu_memory_) : gpu_memory{gpu_memory_} {}
 
-    [[nodiscard]] bool Synchornize(GPUVAddr gpu_addr, u32 limit) {
+    [[nodiscard]] bool Synchronize(GPUVAddr gpu_addr, u32 limit) {
         [[likely]] if (current_gpu_addr == gpu_addr && current_limit == limit) {
             return false;
         }

src/video_core/texture_cache/texture_cache.h

@@ -193,11 +193,11 @@ void TextureCache<P>::SynchronizeGraphicsDescriptors() {
     const bool linked_tsc = maxwell3d->regs.sampler_binding == SamplerBinding::ViaHeaderBinding;
     const u32 tic_limit = maxwell3d->regs.tex_header.limit;
     const u32 tsc_limit = linked_tsc ? tic_limit : maxwell3d->regs.tex_sampler.limit;
-    if (channel_state->graphics_sampler_table.Synchornize(maxwell3d->regs.tex_sampler.Address(),
+    if (channel_state->graphics_sampler_table.Synchronize(maxwell3d->regs.tex_sampler.Address(),
                                                           tsc_limit)) {
         channel_state->graphics_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID);
     }
-    if (channel_state->graphics_image_table.Synchornize(maxwell3d->regs.tex_header.Address(),
+    if (channel_state->graphics_image_table.Synchronize(maxwell3d->regs.tex_header.Address(),
                                                         tic_limit)) {
         channel_state->graphics_image_view_ids.resize(tic_limit + 1, CORRUPT_ID);
     }
@@ -209,10 +209,10 @@ void TextureCache<P>::SynchronizeComputeDescriptors() {
     const u32 tic_limit = kepler_compute->regs.tic.limit;
     const u32 tsc_limit = linked_tsc ? tic_limit : kepler_compute->regs.tsc.limit;
     const GPUVAddr tsc_gpu_addr = kepler_compute->regs.tsc.Address();
-    if (channel_state->compute_sampler_table.Synchornize(tsc_gpu_addr, tsc_limit)) {
+    if (channel_state->compute_sampler_table.Synchronize(tsc_gpu_addr, tsc_limit)) {
         channel_state->compute_sampler_ids.resize(tsc_limit + 1, CORRUPT_ID);
     }
-    if (channel_state->compute_image_table.Synchornize(kepler_compute->regs.tic.Address(),
+    if (channel_state->compute_image_table.Synchronize(kepler_compute->regs.tic.Address(),
                                                        tic_limit)) {
         channel_state->compute_image_view_ids.resize(tic_limit + 1, CORRUPT_ID);
     }

src/yuzu/applets/qt_controller.ui

@@ -2300,7 +2300,7 @@
             <item>
              <widget class="QRadioButton" name="radioUndocked">
               <property name="text">
-               <string>Undocked</string>
+               <string>Handheld</string>
               </property>
              </widget>
             </item>

src/yuzu/bootmanager.cpp

@@ -120,8 +120,8 @@ void EmuThread::run() {
         }
     }
 
-    // Shutdown the core emulation
-    system.Shutdown();
+    // Shutdown the main emulated process
+    system.ShutdownMainProcess();
 
 #if MICROPROFILE_ENABLED
     MicroProfileOnThreadExit();

src/yuzu/main.cpp

@@ -294,6 +294,7 @@ GMainWindow::GMainWindow(std::unique_ptr<Config> config_, bool has_broken_vulkan
 #ifdef __linux__
     SetupSigInterrupts();
 #endif
+    system->Initialize();
 
     Common::Log::Initialize();
     LoadTranslation();
@@ -1895,6 +1896,8 @@ void GMainWindow::OnGameListOpenFolder(u64 program_id, GameListOpenTarget target
     case GameListOpenTarget::SaveData: {
         open_target = tr("Save Data");
         const auto nand_dir = Common::FS::GetYuzuPath(Common::FS::YuzuPath::NANDDir);
+        auto vfs_nand_dir =
+            vfs->OpenDirectory(Common::FS::PathToUTF8String(nand_dir), FileSys::Mode::Read);
 
         if (has_user_save) {
             // User save data
@@ -1921,15 +1924,15 @@ void GMainWindow::OnGameListOpenFolder(u64 program_id, GameListOpenTarget target
             ASSERT(user_id);
 
             const auto user_save_data_path = FileSys::SaveDataFactory::GetFullPath(
-                *system, FileSys::SaveDataSpaceId::NandUser, FileSys::SaveDataType::SaveData,
-                program_id, user_id->AsU128(), 0);
+                *system, vfs_nand_dir, FileSys::SaveDataSpaceId::NandUser,
+                FileSys::SaveDataType::SaveData, program_id, user_id->AsU128(), 0);
 
             path = Common::FS::ConcatPathSafe(nand_dir, user_save_data_path);
         } else {
             // Device save data
             const auto device_save_data_path = FileSys::SaveDataFactory::GetFullPath(
-                *system, FileSys::SaveDataSpaceId::NandUser, FileSys::SaveDataType::SaveData,
-                program_id, {}, 0);
+                *system, vfs_nand_dir, FileSys::SaveDataSpaceId::NandUser,
+                FileSys::SaveDataType::SaveData, program_id, {}, 0);
 
             path = Common::FS::ConcatPathSafe(nand_dir, device_save_data_path);
         }
@@ -3280,7 +3283,7 @@ void GMainWindow::LoadAmiibo(const QString& filename) {
         QMessageBox::warning(this, title, tr("The current game is not looking for amiibos"));
         break;
     case InputCommon::VirtualAmiibo::Info::Unknown:
-        QMessageBox::warning(this, title, tr("An unkown error occured"));
+        QMessageBox::warning(this, title, tr("An unknown error occurred"));
         break;
     default:
         break;

src/yuzu_cmd/yuzu.cpp

@@ -302,6 +302,8 @@ int main(int argc, char** argv) {
     }
 
     Core::System system{};
+    system.Initialize();
+
     InputCommon::InputSubsystem input_subsystem{};
 
     // Apply the command line arguments
@@ -392,7 +394,7 @@ int main(int argc, char** argv) {
     }
     system.DetachDebugger();
     void(system.Pause());
-    system.Shutdown();
+    system.ShutdownMainProcess();
 
     detached_tasks.WaitForAllTasks();
     return 0;