Merge 51d4c772a2 into c0445006af

kernel/scheduler: Take ARM_Interface instances by reference

.travis.yml

@@ -24,7 +24,7 @@ matrix:
     - os: osx
       env: NAME="macos build"
       sudo: false
-      osx_image: xcode9.3
+      osx_image: xcode10
       install: "./.travis/macos/deps.sh"
       script: "./.travis/macos/build.sh"
       after_success: "./.travis/macos/upload.sh"

CMakeLists.txt

@@ -440,8 +440,12 @@ enable_testing()
 add_subdirectory(externals)
 add_subdirectory(src)
 
-# Set yuzu project as default StartUp Project in Visual Studio
-set_property(DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR} PROPERTY VS_STARTUP_PROJECT yuzu)
+# Set yuzu project or yuzu-cmd project as default StartUp Project in Visual Studio depending on whether QT is enabled or not
+if(ENABLE_QT)
+    set_property(DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR} PROPERTY VS_STARTUP_PROJECT yuzu)
+else()
+    set_property(DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR} PROPERTY VS_STARTUP_PROJECT yuzu-cmd)
+endif()
 
 
 # Installation instructions

src/audio_core/audio_renderer.cpp

@@ -79,6 +79,10 @@ u32 AudioRenderer::GetMixBufferCount() const {
     return worker_params.mix_buffer_count;
 }
 
+u32 AudioRenderer::GetState() const {
+    return stream->GetState();
+}
+
 std::vector<u8> AudioRenderer::UpdateAudioRenderer(const std::vector<u8>& input_params) {
     // Copy UpdateDataHeader struct
     UpdateDataHeader config{};

src/audio_core/audio_renderer.h

@@ -170,6 +170,7 @@ public:
     u32 GetSampleRate() const;
     u32 GetSampleCount() const;
     u32 GetMixBufferCount() const;
+    u32 GetState() const;
 
 private:
     class VoiceState;

src/audio_core/stream.cpp

@@ -49,9 +49,14 @@ void Stream::Play() {
 }
 
 void Stream::Stop() {
+    state = State::Stopped;
     ASSERT_MSG(false, "Unimplemented");
 }
 
+u32 Stream::GetState() const {
+    return static_cast<u32>(state);
+}
+
 s64 Stream::GetBufferReleaseCycles(const Buffer& buffer) const {
     const std::size_t num_samples{buffer.GetSamples().size() / GetNumChannels()};
     return CoreTiming::usToCycles((static_cast<u64>(num_samples) * 1000000) / sample_rate);

src/audio_core/stream.h

@@ -72,6 +72,9 @@ public:
     /// Gets the number of channels
     u32 GetNumChannels() const;
 
+    /// Get the state
+    u32 GetState() const;
+
 private:
     /// Current state of the stream
     enum class State {

src/common/common_paths.h

@@ -33,6 +33,8 @@
 #define NAND_DIR "nand"
 #define SYSDATA_DIR "sysdata"
 #define KEYS_DIR "keys"
+#define LOAD_DIR "load"
+#define DUMP_DIR "dump"
 #define LOG_DIR "log"
 
 // Filenames

src/common/file_util.cpp

@@ -705,6 +705,8 @@ const std::string& GetUserPath(UserPath path, const std::string& new_path) {
 #endif
         paths.emplace(UserPath::SDMCDir, user_path + SDMC_DIR DIR_SEP);
         paths.emplace(UserPath::NANDDir, user_path + NAND_DIR DIR_SEP);
+        paths.emplace(UserPath::LoadDir, user_path + LOAD_DIR DIR_SEP);
+        paths.emplace(UserPath::DumpDir, user_path + DUMP_DIR DIR_SEP);
         paths.emplace(UserPath::SysDataDir, user_path + SYSDATA_DIR DIR_SEP);
         paths.emplace(UserPath::KeysDir, user_path + KEYS_DIR DIR_SEP);
         // TODO: Put the logs in a better location for each OS

src/common/file_util.h

@@ -29,6 +29,8 @@ enum class UserPath {
     NANDDir,
     RootDir,
     SDMCDir,
+    LoadDir,
+    DumpDir,
     SysDataDir,
     UserDir,
 };

src/common/logging/backend.cpp

@@ -183,6 +183,7 @@ void FileBackend::Write(const Entry& entry) {
     SUB(Service, FS)                                                                               \
     SUB(Service, GRC)                                                                              \
     SUB(Service, HID)                                                                              \
+    SUB(Service, IRS)                                                                              \
     SUB(Service, LBL)                                                                              \
     SUB(Service, LDN)                                                                              \
     SUB(Service, LDR)                                                                              \

src/common/logging/log.h

@@ -70,6 +70,7 @@ enum class Class : ClassType {
     Service_FS,        ///< The FS (Filesystem) service
     Service_GRC,       ///< The game recording service
     Service_HID,       ///< The HID (Human interface device) service
+    Service_IRS,       ///< The IRS service
     Service_LBL,       ///< The LBL (LCD backlight) service
     Service_LDN,       ///< The LDN (Local domain network) service
     Service_LDR,       ///< The loader service

src/core/CMakeLists.txt

@@ -32,6 +32,8 @@ add_library(core STATIC
     file_sys/control_metadata.h
     file_sys/directory.h
     file_sys/errors.h
+    file_sys/fsmitm_romfsbuild.cpp
+    file_sys/fsmitm_romfsbuild.h
     file_sys/mode.h
     file_sys/nca_metadata.cpp
     file_sys/nca_metadata.h
@@ -59,10 +61,13 @@ add_library(core STATIC
     file_sys/vfs.h
     file_sys/vfs_concat.cpp
     file_sys/vfs_concat.h
+    file_sys/vfs_layered.cpp
+    file_sys/vfs_layered.h
     file_sys/vfs_offset.cpp
     file_sys/vfs_offset.h
     file_sys/vfs_real.cpp
     file_sys/vfs_real.h
+    file_sys/vfs_static.h
     file_sys/vfs_vector.cpp
     file_sys/vfs_vector.h
     file_sys/xts_archive.cpp

src/core/core.cpp

@@ -64,7 +64,7 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
         if (concat.empty())
             return nullptr;
 
-        return FileSys::ConcatenateFiles(concat, dir->GetName());
+        return FileSys::ConcatenatedVfsFile::MakeConcatenatedFile(concat, dir->GetName());
     }
 
     return vfs->OpenFile(path, FileSys::Mode::Read);

src/core/core_cpu.cpp

@@ -55,16 +55,16 @@ Cpu::Cpu(std::shared_ptr<ExclusiveMonitor> exclusive_monitor,
 
     if (Settings::values.use_cpu_jit) {
 #ifdef ARCHITECTURE_x86_64
-        arm_interface = std::make_shared<ARM_Dynarmic>(exclusive_monitor, core_index);
+        arm_interface = std::make_unique<ARM_Dynarmic>(exclusive_monitor, core_index);
 #else
-        arm_interface = std::make_shared<ARM_Unicorn>();
+        arm_interface = std::make_unique<ARM_Unicorn>();
         LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
     } else {
-        arm_interface = std::make_shared<ARM_Unicorn>();
+        arm_interface = std::make_unique<ARM_Unicorn>();
     }
 
-    scheduler = std::make_shared<Kernel::Scheduler>(arm_interface.get());
+    scheduler = std::make_shared<Kernel::Scheduler>(*arm_interface);
 }
 
 Cpu::~Cpu() = default;

src/core/core_cpu.h

@@ -76,7 +76,7 @@ public:
 private:
     void Reschedule();
 
-    std::shared_ptr<ARM_Interface> arm_interface;
+    std::unique_ptr<ARM_Interface> arm_interface;
     std::shared_ptr<CpuBarrier> cpu_barrier;
     std::shared_ptr<Kernel::Scheduler> scheduler;
 

src/core/file_sys/bis_factory.cpp

@@ -2,13 +2,14 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <fmt/format.h>
 #include "core/file_sys/bis_factory.h"
 #include "core/file_sys/registered_cache.h"
 
 namespace FileSys {
 
-BISFactory::BISFactory(VirtualDir nand_root_)
-    : nand_root(std::move(nand_root_)),
+BISFactory::BISFactory(VirtualDir nand_root_, VirtualDir load_root_)
+    : nand_root(std::move(nand_root_)), load_root(std::move(load_root_)),
       sysnand_cache(std::make_shared<RegisteredCache>(
           GetOrCreateDirectoryRelative(nand_root, "/system/Contents/registered"))),
       usrnand_cache(std::make_shared<RegisteredCache>(
@@ -24,4 +25,11 @@ std::shared_ptr<RegisteredCache> BISFactory::GetUserNANDContents() const {
     return usrnand_cache;
 }
 
+VirtualDir BISFactory::GetModificationLoadRoot(u64 title_id) const {
+    // LayeredFS doesn't work on updates and title id-less homebrew
+    if (title_id == 0 || (title_id & 0x800) > 0)
+        return nullptr;
+    return GetOrCreateDirectoryRelative(load_root, fmt::format("/{:016X}", title_id));
+}
+
 } // namespace FileSys

src/core/file_sys/bis_factory.h

@@ -17,14 +17,17 @@ class RegisteredCache;
 /// registered caches.
 class BISFactory {
 public:
-    explicit BISFactory(VirtualDir nand_root);
+    explicit BISFactory(VirtualDir nand_root, VirtualDir load_root);
     ~BISFactory();
 
     std::shared_ptr<RegisteredCache> GetSystemNANDContents() const;
     std::shared_ptr<RegisteredCache> GetUserNANDContents() const;
 
+    VirtualDir GetModificationLoadRoot(u64 title_id) const;
+
 private:
     VirtualDir nand_root;
+    VirtualDir load_root;
 
     std::shared_ptr<RegisteredCache> sysnand_cache;
     std::shared_ptr<RegisteredCache> usrnand_cache;

src/core/file_sys/fsmitm_romfsbuild.cpp

@@ -0,0 +1,367 @@
+/*
+ * Copyright (c) 2018 Atmosphère-NX
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+/*
+ * Adapted by DarkLordZach for use/interaction with yuzu
+ *
+ * Modifications Copyright 2018 yuzu emulator team
+ * Licensed under GPLv2 or any later version
+ * Refer to the license.txt file included.
+ */
+
+#include <cstring>
+#include "common/assert.h"
+#include "core/file_sys/fsmitm_romfsbuild.h"
+#include "core/file_sys/vfs.h"
+#include "core/file_sys/vfs_vector.h"
+
+namespace FileSys {
+
+constexpr u64 FS_MAX_PATH = 0x301;
+
+constexpr u32 ROMFS_ENTRY_EMPTY = 0xFFFFFFFF;
+constexpr u32 ROMFS_FILEPARTITION_OFS = 0x200;
+
+// Types for building a RomFS.
+struct RomFSHeader {
+    u64 header_size;
+    u64 dir_hash_table_ofs;
+    u64 dir_hash_table_size;
+    u64 dir_table_ofs;
+    u64 dir_table_size;
+    u64 file_hash_table_ofs;
+    u64 file_hash_table_size;
+    u64 file_table_ofs;
+    u64 file_table_size;
+    u64 file_partition_ofs;
+};
+static_assert(sizeof(RomFSHeader) == 0x50, "RomFSHeader has incorrect size.");
+
+struct RomFSDirectoryEntry {
+    u32 parent;
+    u32 sibling;
+    u32 child;
+    u32 file;
+    u32 hash;
+    u32 name_size;
+};
+static_assert(sizeof(RomFSDirectoryEntry) == 0x18, "RomFSDirectoryEntry has incorrect size.");
+
+struct RomFSFileEntry {
+    u32 parent;
+    u32 sibling;
+    u64 offset;
+    u64 size;
+    u32 hash;
+    u32 name_size;
+};
+static_assert(sizeof(RomFSFileEntry) == 0x20, "RomFSFileEntry has incorrect size.");
+
+struct RomFSBuildFileContext;
+
+struct RomFSBuildDirectoryContext {
+    std::string path;
+    u32 cur_path_ofs = 0;
+    u32 path_len = 0;
+    u32 entry_offset = 0;
+    std::shared_ptr<RomFSBuildDirectoryContext> parent;
+    std::shared_ptr<RomFSBuildDirectoryContext> child;
+    std::shared_ptr<RomFSBuildDirectoryContext> sibling;
+    std::shared_ptr<RomFSBuildFileContext> file;
+};
+
+struct RomFSBuildFileContext {
+    std::string path;
+    u32 cur_path_ofs = 0;
+    u32 path_len = 0;
+    u32 entry_offset = 0;
+    u64 offset = 0;
+    u64 size = 0;
+    std::shared_ptr<RomFSBuildDirectoryContext> parent;
+    std::shared_ptr<RomFSBuildFileContext> sibling;
+    VirtualFile source;
+};
+
+static u32 romfs_calc_path_hash(u32 parent, std::string path, u32 start, std::size_t path_len) {
+    u32 hash = parent ^ 123456789;
+    for (u32 i = 0; i < path_len; i++) {
+        hash = (hash >> 5) | (hash << 27);
+        hash ^= path[start + i];
+    }
+
+    return hash;
+}
+
+static u64 romfs_get_hash_table_count(u64 num_entries) {
+    if (num_entries < 3) {
+        return 3;
+    }
+
+    if (num_entries < 19) {
+        return num_entries | 1;
+    }
+
+    u64 count = num_entries;
+    while (count % 2 == 0 || count % 3 == 0 || count % 5 == 0 || count % 7 == 0 ||
+           count % 11 == 0 || count % 13 == 0 || count % 17 == 0) {
+        count++;
+    }
+    return count;
+}
+
+void RomFSBuildContext::VisitDirectory(VirtualDir root_romfs,
+                                       std::shared_ptr<RomFSBuildDirectoryContext> parent) {
+    std::vector<std::shared_ptr<RomFSBuildDirectoryContext>> child_dirs;
+
+    VirtualDir dir;
+
+    if (parent->path_len == 0)
+        dir = root_romfs;
+    else
+        dir = root_romfs->GetDirectoryRelative(parent->path);
+
+    const auto entries = dir->GetEntries();
+
+    for (const auto& kv : entries) {
+        if (kv.second == VfsEntryType::Directory) {
+            const auto child = std::make_shared<RomFSBuildDirectoryContext>();
+            // Set child's path.
+            child->cur_path_ofs = parent->path_len + 1;
+            child->path_len = child->cur_path_ofs + static_cast<u32>(kv.first.size());
+            child->path = parent->path + "/" + kv.first;
+
+            // Sanity check on path_len
+            ASSERT(child->path_len < FS_MAX_PATH);
+
+            if (AddDirectory(parent, child)) {
+                child_dirs.push_back(child);
+            }
+        } else {
+            const auto child = std::make_shared<RomFSBuildFileContext>();
+            // Set child's path.
+            child->cur_path_ofs = parent->path_len + 1;
+            child->path_len = child->cur_path_ofs + static_cast<u32>(kv.first.size());
+            child->path = parent->path + "/" + kv.first;
+
+            // Sanity check on path_len
+            ASSERT(child->path_len < FS_MAX_PATH);
+
+            child->source = root_romfs->GetFileRelative(child->path);
+
+            child->size = child->source->GetSize();
+
+            AddFile(parent, child);
+        }
+    }
+
+    for (auto& child : child_dirs) {
+        this->VisitDirectory(root_romfs, child);
+    }
+}
+
+bool RomFSBuildContext::AddDirectory(std::shared_ptr<RomFSBuildDirectoryContext> parent_dir_ctx,
+                                     std::shared_ptr<RomFSBuildDirectoryContext> dir_ctx) {
+    // Check whether it's already in the known directories.
+    const auto existing = directories.find(dir_ctx->path);
+    if (existing != directories.end())
+        return false;
+
+    // Add a new directory.
+    num_dirs++;
+    dir_table_size +=
+        sizeof(RomFSDirectoryEntry) + ((dir_ctx->path_len - dir_ctx->cur_path_ofs + 3) & ~3);
+    dir_ctx->parent = parent_dir_ctx;
+    directories.emplace(dir_ctx->path, dir_ctx);
+
+    return true;
+}
+
+bool RomFSBuildContext::AddFile(std::shared_ptr<RomFSBuildDirectoryContext> parent_dir_ctx,
+                                std::shared_ptr<RomFSBuildFileContext> file_ctx) {
+    // Check whether it's already in the known files.
+    const auto existing = files.find(file_ctx->path);
+    if (existing != files.end()) {
+        return false;
+    }
+
+    // Add a new file.
+    num_files++;
+    file_table_size +=
+        sizeof(RomFSFileEntry) + ((file_ctx->path_len - file_ctx->cur_path_ofs + 3) & ~3);
+    file_ctx->parent = parent_dir_ctx;
+    files.emplace(file_ctx->path, file_ctx);
+
+    return true;
+}
+
+RomFSBuildContext::RomFSBuildContext(VirtualDir base_) : base(std::move(base_)) {
+    root = std::make_shared<RomFSBuildDirectoryContext>();
+    root->path = "\0";
+    directories.emplace(root->path, root);
+    num_dirs = 1;
+    dir_table_size = 0x18;
+
+    VisitDirectory(base, root);
+}
+
+RomFSBuildContext::~RomFSBuildContext() = default;
+
+std::map<u64, VirtualFile> RomFSBuildContext::Build() {
+    const u64 dir_hash_table_entry_count = romfs_get_hash_table_count(num_dirs);
+    const u64 file_hash_table_entry_count = romfs_get_hash_table_count(num_files);
+    dir_hash_table_size = 4 * dir_hash_table_entry_count;
+    file_hash_table_size = 4 * file_hash_table_entry_count;
+
+    // Assign metadata pointers
+    RomFSHeader header{};
+
+    std::vector<u32> dir_hash_table(dir_hash_table_entry_count, ROMFS_ENTRY_EMPTY);
+    std::vector<u32> file_hash_table(file_hash_table_entry_count, ROMFS_ENTRY_EMPTY);
+
+    std::vector<u8> dir_table(dir_table_size);
+    std::vector<u8> file_table(file_table_size);
+
+    std::shared_ptr<RomFSBuildFileContext> cur_file;
+
+    // Determine file offsets.
+    u32 entry_offset = 0;
+    std::shared_ptr<RomFSBuildFileContext> prev_file = nullptr;
+    for (const auto& it : files) {
+        cur_file = it.second;
+        file_partition_size = (file_partition_size + 0xFULL) & ~0xFULL;
+        cur_file->offset = file_partition_size;
+        file_partition_size += cur_file->size;
+        cur_file->entry_offset = entry_offset;
+        entry_offset +=
+            sizeof(RomFSFileEntry) + ((cur_file->path_len - cur_file->cur_path_ofs + 3) & ~3);
+        prev_file = cur_file;
+    }
+    // Assign deferred parent/sibling ownership.
+    for (auto it = files.rbegin(); it != files.rend(); ++it) {
+        cur_file = it->second;
+        cur_file->sibling = cur_file->parent->file;
+        cur_file->parent->file = cur_file;
+    }
+
+    std::shared_ptr<RomFSBuildDirectoryContext> cur_dir;
+
+    // Determine directory offsets.
+    entry_offset = 0;
+    for (const auto& it : directories) {
+        cur_dir = it.second;
+        cur_dir->entry_offset = entry_offset;
+        entry_offset +=
+            sizeof(RomFSDirectoryEntry) + ((cur_dir->path_len - cur_dir->cur_path_ofs + 3) & ~3);
+    }
+    // Assign deferred parent/sibling ownership.
+    for (auto it = directories.rbegin(); it->second != root; ++it) {
+        cur_dir = it->second;
+        cur_dir->sibling = cur_dir->parent->child;
+        cur_dir->parent->child = cur_dir;
+    }
+
+    std::map<u64, VirtualFile> out;
+
+    // Populate file tables.
+    for (const auto& it : files) {
+        cur_file = it.second;
+        RomFSFileEntry cur_entry{};
+
+        cur_entry.parent = cur_file->parent->entry_offset;
+        cur_entry.sibling =
+            cur_file->sibling == nullptr ? ROMFS_ENTRY_EMPTY : cur_file->sibling->entry_offset;
+        cur_entry.offset = cur_file->offset;
+        cur_entry.size = cur_file->size;
+
+        const auto name_size = cur_file->path_len - cur_file->cur_path_ofs;
+        const auto hash = romfs_calc_path_hash(cur_file->parent->entry_offset, cur_file->path,
+                                               cur_file->cur_path_ofs, name_size);
+        cur_entry.hash = file_hash_table[hash % file_hash_table_entry_count];
+        file_hash_table[hash % file_hash_table_entry_count] = cur_file->entry_offset;
+
+        cur_entry.name_size = name_size;
+
+        out.emplace(cur_file->offset + ROMFS_FILEPARTITION_OFS, cur_file->source);
+        std::memcpy(file_table.data() + cur_file->entry_offset, &cur_entry, sizeof(RomFSFileEntry));
+        std::memset(file_table.data() + cur_file->entry_offset + sizeof(RomFSFileEntry), 0,
+                    (cur_entry.name_size + 3) & ~3);
+        std::memcpy(file_table.data() + cur_file->entry_offset + sizeof(RomFSFileEntry),
+                    cur_file->path.data() + cur_file->cur_path_ofs, name_size);
+    }
+
+    // Populate dir tables.
+    for (const auto& it : directories) {
+        cur_dir = it.second;
+        RomFSDirectoryEntry cur_entry{};
+
+        cur_entry.parent = cur_dir == root ? 0 : cur_dir->parent->entry_offset;
+        cur_entry.sibling =
+            cur_dir->sibling == nullptr ? ROMFS_ENTRY_EMPTY : cur_dir->sibling->entry_offset;
+        cur_entry.child =
+            cur_dir->child == nullptr ? ROMFS_ENTRY_EMPTY : cur_dir->child->entry_offset;
+        cur_entry.file = cur_dir->file == nullptr ? ROMFS_ENTRY_EMPTY : cur_dir->file->entry_offset;
+
+        const auto name_size = cur_dir->path_len - cur_dir->cur_path_ofs;
+        const auto hash = romfs_calc_path_hash(cur_dir == root ? 0 : cur_dir->parent->entry_offset,
+                                               cur_dir->path, cur_dir->cur_path_ofs, name_size);
+        cur_entry.hash = dir_hash_table[hash % dir_hash_table_entry_count];
+        dir_hash_table[hash % dir_hash_table_entry_count] = cur_dir->entry_offset;
+
+        cur_entry.name_size = name_size;
+
+        std::memcpy(dir_table.data() + cur_dir->entry_offset, &cur_entry,
+                    sizeof(RomFSDirectoryEntry));
+        std::memcpy(dir_table.data() + cur_dir->entry_offset, &cur_entry,
+                    sizeof(RomFSDirectoryEntry));
+        std::memset(dir_table.data() + cur_dir->entry_offset + sizeof(RomFSDirectoryEntry), 0,
+                    (cur_entry.name_size + 3) & ~3);
+        std::memcpy(dir_table.data() + cur_dir->entry_offset + sizeof(RomFSDirectoryEntry),
+                    cur_dir->path.data() + cur_dir->cur_path_ofs, name_size);
+    }
+
+    // Set header fields.
+    header.header_size = sizeof(RomFSHeader);
+    header.file_hash_table_size = file_hash_table_size;
+    header.file_table_size = file_table_size;
+    header.dir_hash_table_size = dir_hash_table_size;
+    header.dir_table_size = dir_table_size;
+    header.file_partition_ofs = ROMFS_FILEPARTITION_OFS;
+    header.dir_hash_table_ofs = (header.file_partition_ofs + file_partition_size + 3ULL) & ~3ULL;
+    header.dir_table_ofs = header.dir_hash_table_ofs + header.dir_hash_table_size;
+    header.file_hash_table_ofs = header.dir_table_ofs + header.dir_table_size;
+    header.file_table_ofs = header.file_hash_table_ofs + header.file_hash_table_size;
+
+    std::vector<u8> header_data(sizeof(RomFSHeader));
+    std::memcpy(header_data.data(), &header, header_data.size());
+    out.emplace(0, std::make_shared<VectorVfsFile>(header_data));
+
+    std::vector<u8> metadata(file_hash_table_size + file_table_size + dir_hash_table_size +
+                             dir_table_size);
+    std::size_t index = 0;
+    std::memcpy(metadata.data(), dir_hash_table.data(), dir_hash_table.size() * sizeof(u32));
+    index += dir_hash_table.size() * sizeof(u32);
+    std::memcpy(metadata.data() + index, dir_table.data(), dir_table.size());
+    index += dir_table.size();
+    std::memcpy(metadata.data() + index, file_hash_table.data(),
+                file_hash_table.size() * sizeof(u32));
+    index += file_hash_table.size() * sizeof(u32);
+    std::memcpy(metadata.data() + index, file_table.data(), file_table.size());
+    out.emplace(header.dir_hash_table_ofs, std::make_shared<VectorVfsFile>(metadata));
+
+    return out;
+}
+
+} // namespace FileSys

src/core/file_sys/fsmitm_romfsbuild.h

@@ -0,0 +1,70 @@
+/*
+ * Copyright (c) 2018 Atmosphère-NX
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+/*
+ * Adapted by DarkLordZach for use/interaction with yuzu
+ *
+ * Modifications Copyright 2018 yuzu emulator team
+ * Licensed under GPLv2 or any later version
+ * Refer to the license.txt file included.
+ */
+
+#pragma once
+
+#include <map>
+#include <memory>
+#include <string>
+#include <boost/detail/container_fwd.hpp>
+#include "common/common_types.h"
+#include "core/file_sys/vfs.h"
+
+namespace FileSys {
+
+struct RomFSBuildDirectoryContext;
+struct RomFSBuildFileContext;
+struct RomFSDirectoryEntry;
+struct RomFSFileEntry;
+
+class RomFSBuildContext {
+public:
+    explicit RomFSBuildContext(VirtualDir base);
+    ~RomFSBuildContext();
+
+    // This finalizes the context.
+    std::map<u64, VirtualFile> Build();
+
+private:
+    VirtualDir base;
+    std::shared_ptr<RomFSBuildDirectoryContext> root;
+    std::map<std::string, std::shared_ptr<RomFSBuildDirectoryContext>, std::less<>> directories;
+    std::map<std::string, std::shared_ptr<RomFSBuildFileContext>, std::less<>> files;
+    u64 num_dirs = 0;
+    u64 num_files = 0;
+    u64 dir_table_size = 0;
+    u64 file_table_size = 0;
+    u64 dir_hash_table_size = 0;
+    u64 file_hash_table_size = 0;
+    u64 file_partition_size = 0;
+
+    void VisitDirectory(VirtualDir filesys, std::shared_ptr<RomFSBuildDirectoryContext> parent);
+
+    bool AddDirectory(std::shared_ptr<RomFSBuildDirectoryContext> parent_dir_ctx,
+                      std::shared_ptr<RomFSBuildDirectoryContext> dir_ctx);
+    bool AddFile(std::shared_ptr<RomFSBuildDirectoryContext> parent_dir_ctx,
+                 std::shared_ptr<RomFSBuildFileContext> file_ctx);
+};
+
+} // namespace FileSys

src/core/file_sys/patch_manager.cpp

@@ -11,6 +11,7 @@
 #include "core/file_sys/patch_manager.h"
 #include "core/file_sys/registered_cache.h"
 #include "core/file_sys/romfs.h"
+#include "core/file_sys/vfs_layered.h"
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/loader/loader.h"
 
@@ -31,8 +32,9 @@ std::string FormatTitleVersion(u32 version, TitleVersionFormat format) {
     return fmt::format("v{}.{}.{}", bytes[3], bytes[2], bytes[1]);
 }
 
-constexpr std::array<const char*, 1> PATCH_TYPE_NAMES{
+constexpr std::array<const char*, 2> PATCH_TYPE_NAMES{
     "Update",
+    "LayeredFS",
 };
 
 std::string FormatPatchTypeName(PatchType type) {
@@ -66,6 +68,44 @@ VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
     return exefs;
 }
 
+static void ApplyLayeredFS(VirtualFile& romfs, u64 title_id, ContentRecordType type) {
+    const auto load_dir = Service::FileSystem::GetModificationLoadRoot(title_id);
+    if (type != ContentRecordType::Program || load_dir == nullptr || load_dir->GetSize() <= 0) {
+        return;
+    }
+
+    auto extracted = ExtractRomFS(romfs);
+    if (extracted == nullptr) {
+        return;
+    }
+
+    auto patch_dirs = load_dir->GetSubdirectories();
+    std::sort(patch_dirs.begin(), patch_dirs.end(),
+              [](const VirtualDir& l, const VirtualDir& r) { return l->GetName() < r->GetName(); });
+
+    std::vector<VirtualDir> layers;
+    layers.reserve(patch_dirs.size() + 1);
+    for (const auto& subdir : patch_dirs) {
+        auto romfs_dir = subdir->GetSubdirectory("romfs");
+        if (romfs_dir != nullptr)
+            layers.push_back(std::move(romfs_dir));
+    }
+    layers.push_back(std::move(extracted));
+
+    auto layered = LayeredVfsDirectory::MakeLayeredDirectory(std::move(layers));
+    if (layered == nullptr) {
+        return;
+    }
+
+    auto packed = CreateRomFS(std::move(layered));
+    if (packed == nullptr) {
+        return;
+    }
+
+    LOG_INFO(Loader, "    RomFS: LayeredFS patches applied successfully");
+    romfs = std::move(packed);
+}
+
 VirtualFile PatchManager::PatchRomFS(VirtualFile romfs, u64 ivfc_offset,
                                      ContentRecordType type) const {
     LOG_INFO(Loader, "Patching RomFS for title_id={:016X}, type={:02X}", title_id,
@@ -89,6 +129,9 @@ VirtualFile PatchManager::PatchRomFS(VirtualFile romfs, u64 ivfc_offset,
         }
     }
 
+    // LayeredFS
+    ApplyLayeredFS(romfs, title_id, type);
+
     return romfs;
 }
 
@@ -114,6 +157,10 @@ std::map<PatchType, std::string> PatchManager::GetPatchVersionNames() const {
         }
     }
 
+    const auto lfs_dir = Service::FileSystem::GetModificationLoadRoot(title_id);
+    if (lfs_dir != nullptr && lfs_dir->GetSize() > 0)
+        out.insert_or_assign(PatchType::LayeredFS, "");
+
     return out;
 }
 

src/core/file_sys/patch_manager.h

@@ -26,6 +26,7 @@ std::string FormatTitleVersion(u32 version,
 
 enum class PatchType {
     Update,
+    LayeredFS,
 };
 
 std::string FormatPatchTypeName(PatchType type);
@@ -42,6 +43,7 @@ public:
 
     // Currently tracked RomFS patches:
     // - Game Updates
+    // - LayeredFS
     VirtualFile PatchRomFS(VirtualFile base, u64 ivfc_offset,
                            ContentRecordType type = ContentRecordType::Program) const;
 

src/core/file_sys/registered_cache.cpp

@@ -18,6 +18,10 @@
 #include "core/loader/loader.h"
 
 namespace FileSys {
+
+// The size of blocks to use when vfs raw copying into nand.
+constexpr size_t VFS_RC_LARGE_COPY_BLOCK = 0x400000;
+
 std::string RegisteredCacheEntry::DebugInfo() const {
     return fmt::format("title_id={:016X}, content_type={:02X}", title_id, static_cast<u8>(type));
 }
@@ -121,7 +125,7 @@ VirtualFile RegisteredCache::OpenFileOrDirectoryConcat(const VirtualDir& dir,
             if (concat.empty())
                 return nullptr;
 
-            file = FileSys::ConcatenateFiles(concat);
+            file = ConcatenatedVfsFile::MakeConcatenatedFile(concat, concat.front()->GetName());
         }
 
         return file;
@@ -480,7 +484,8 @@ InstallResult RegisteredCache::RawInstallNCA(std::shared_ptr<NCA> nca, const Vfs
     auto out = dir->CreateFileRelative(path);
     if (out == nullptr)
         return InstallResult::ErrorCopyFailed;
-    return copy(in, out) ? InstallResult::Success : InstallResult::ErrorCopyFailed;
+    return copy(in, out, VFS_RC_LARGE_COPY_BLOCK) ? InstallResult::Success
+                                                  : InstallResult::ErrorCopyFailed;
 }
 
 bool RegisteredCache::RawInstallYuzuMeta(const CNMT& cnmt) {

src/core/file_sys/registered_cache.h

@@ -27,7 +27,7 @@ struct ContentRecord;
 
 using NcaID = std::array<u8, 0x10>;
 using RegisteredCacheParsingFunction = std::function<VirtualFile(const VirtualFile&, const NcaID&)>;
-using VfsCopyFunction = std::function<bool(VirtualFile, VirtualFile)>;
+using VfsCopyFunction = std::function<bool(const VirtualFile&, const VirtualFile&, size_t)>;
 
 enum class InstallResult {
     Success,

src/core/file_sys/romfs.cpp

@@ -4,8 +4,10 @@
 
 #include "common/common_types.h"
 #include "common/swap.h"
+#include "core/file_sys/fsmitm_romfsbuild.h"
 #include "core/file_sys/romfs.h"
 #include "core/file_sys/vfs.h"
+#include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_offset.h"
 #include "core/file_sys/vfs_vector.h"
 
@@ -98,7 +100,7 @@ void ProcessDirectory(VirtualFile file, std::size_t dir_offset, std::size_t file
     }
 }
 
-VirtualDir ExtractRomFS(VirtualFile file) {
+VirtualDir ExtractRomFS(VirtualFile file, RomFSExtractionType type) {
     RomFSHeader header{};
     if (file->ReadObject(&header) != sizeof(RomFSHeader))
         return nullptr;
@@ -117,9 +119,22 @@ VirtualDir ExtractRomFS(VirtualFile file) {
 
     VirtualDir out = std::move(root);
 
-    while (out->GetSubdirectory("") != nullptr)
-        out = out->GetSubdirectory("");
+    while (out->GetSubdirectories().size() == 1 && out->GetFiles().empty()) {
+        if (out->GetSubdirectories().front()->GetName() == "data" &&
+            type == RomFSExtractionType::Truncated)
+            break;
+        out = out->GetSubdirectories().front();
+    }
 
     return out;
 }
+
+VirtualFile CreateRomFS(VirtualDir dir) {
+    if (dir == nullptr)
+        return nullptr;
+
+    RomFSBuildContext ctx{dir};
+    return ConcatenatedVfsFile::MakeConcatenatedFile(0, ctx.Build(), dir->GetName());
+}
+
 } // namespace FileSys

src/core/file_sys/romfs.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <map>
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/swap.h"
@@ -12,6 +13,8 @@
 
 namespace FileSys {
 
+struct RomFSHeader;
+
 struct IVFCLevel {
     u64_le offset;
     u64_le size;
@@ -29,8 +32,18 @@ struct IVFCHeader {
 };
 static_assert(sizeof(IVFCHeader) == 0xE0, "IVFCHeader has incorrect size.");
 
+enum class RomFSExtractionType {
+    Full,      // Includes data directory
+    Truncated, // Traverses into data directory
+};
+
 // Converts a RomFS binary blob to VFS Filesystem
 // Returns nullptr on failure
-VirtualDir ExtractRomFS(VirtualFile file);
+VirtualDir ExtractRomFS(VirtualFile file,
+                        RomFSExtractionType type = RomFSExtractionType::Truncated);
+
+// Converts a VFS filesystem into a RomFS binary
+// Returns nullptr on failure
+VirtualFile CreateRomFS(VirtualDir dir);
 
 } // namespace FileSys

src/core/file_sys/vfs.cpp

@@ -399,6 +399,15 @@ bool VfsDirectory::Copy(std::string_view src, std::string_view dest) {
     return f2->WriteBytes(f1->ReadAllBytes()) == f1->GetSize();
 }
 
+std::map<std::string, VfsEntryType, std::less<>> VfsDirectory::GetEntries() const {
+    std::map<std::string, VfsEntryType, std::less<>> out;
+    for (const auto& dir : GetSubdirectories())
+        out.emplace(dir->GetName(), VfsEntryType::Directory);
+    for (const auto& file : GetFiles())
+        out.emplace(file->GetName(), VfsEntryType::File);
+    return out;
+}
+
 std::string VfsDirectory::GetFullPath() const {
     if (IsRoot())
         return GetName();
@@ -454,13 +463,41 @@ bool DeepEquals(const VirtualFile& file1, const VirtualFile& file2, std::size_t
     return true;
 }
 
-bool VfsRawCopy(VirtualFile src, VirtualFile dest) {
-    if (src == nullptr || dest == nullptr)
+bool VfsRawCopy(const VirtualFile& src, const VirtualFile& dest, std::size_t block_size) {
+    if (src == nullptr || dest == nullptr || !src->IsReadable() || !dest->IsWritable())
         return false;
     if (!dest->Resize(src->GetSize()))
         return false;
-    std::vector<u8> data = src->ReadAllBytes();
-    return dest->WriteBytes(data, 0) == data.size();
+
+    std::vector<u8> temp(std::min(block_size, src->GetSize()));
+    for (std::size_t i = 0; i < src->GetSize(); i += block_size) {
+        const auto read = std::min(block_size, src->GetSize() - i);
+        const auto block = src->Read(temp.data(), read, i);
+
+        if (dest->Write(temp.data(), read, i) != read)
+            return false;
+    }
+
+    return true;
+}
+
+bool VfsRawCopyD(const VirtualDir& src, const VirtualDir& dest, std::size_t block_size) {
+    if (src == nullptr || dest == nullptr || !src->IsReadable() || !dest->IsWritable())
+        return false;
+
+    for (const auto& file : src->GetFiles()) {
+        const auto out = dest->CreateFile(file->GetName());
+        if (!VfsRawCopy(file, out, block_size))
+            return false;
+    }
+
+    for (const auto& dir : src->GetSubdirectories()) {
+        const auto out = dest->CreateSubdirectory(dir->GetName());
+        if (!VfsRawCopyD(dir, out, block_size))
+            return false;
+    }
+
+    return true;
 }
 
 VirtualDir GetOrCreateDirectoryRelative(const VirtualDir& rel, std::string_view path) {

src/core/file_sys/vfs.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <map>
 #include <memory>
 #include <string>
 #include <string_view>
@@ -265,6 +266,10 @@ public:
     // dest.
     virtual bool Copy(std::string_view src, std::string_view dest);
 
+    // Gets all of the entries directly in the directory (files and dirs), returning a map between
+    // item name -> type.
+    virtual std::map<std::string, VfsEntryType, std::less<>> GetEntries() const;
+
     // Interprets the file with name file instead as a directory of type directory.
     // The directory must have a constructor that takes a single argument of type
     // std::shared_ptr<VfsFile>. Allows to reinterpret container files (i.e NCA, zip, XCI, etc) as a
@@ -310,13 +315,19 @@ public:
     bool Rename(std::string_view name) override;
 };
 
-// Compare the two files, byte-for-byte, in increments specificed by block_size
-bool DeepEquals(const VirtualFile& file1, const VirtualFile& file2, std::size_t block_size = 0x200);
+// Compare the two files, byte-for-byte, in increments specified by block_size
+bool DeepEquals(const VirtualFile& file1, const VirtualFile& file2,
+                std::size_t block_size = 0x1000);
 
 // A method that copies the raw data between two different implementations of VirtualFile. If you
 // are using the same implementation, it is probably better to use the Copy method in the parent
 // directory of src/dest.
-bool VfsRawCopy(VirtualFile src, VirtualFile dest);
+bool VfsRawCopy(const VirtualFile& src, const VirtualFile& dest, std::size_t block_size = 0x1000);
+
+// A method that performs a similar function to VfsRawCopy above, but instead copies entire
+// directories. It suffers the same performance penalties as above and an implementation-specific
+// Copy should always be preferred.
+bool VfsRawCopyD(const VirtualDir& src, const VirtualDir& dest, std::size_t block_size = 0x1000);
 
 // Checks if the directory at path relative to rel exists. If it does, returns that. If it does not
 // it attempts to create it and returns the new dir or nullptr on failure.

src/core/file_sys/vfs_concat.cpp

@@ -5,17 +5,22 @@
 #include <algorithm>
 #include <utility>
 
+#include "common/assert.h"
 #include "core/file_sys/vfs_concat.h"
+#include "core/file_sys/vfs_static.h"
 
 namespace FileSys {
 
-VirtualFile ConcatenateFiles(std::vector<VirtualFile> files, std::string name) {
-    if (files.empty())
-        return nullptr;
-    if (files.size() == 1)
-        return files[0];
+static bool VerifyConcatenationMapContinuity(const std::map<u64, VirtualFile>& map) {
+    const auto last_valid = --map.end();
+    for (auto iter = map.begin(); iter != last_valid;) {
+        const auto old = iter++;
+        if (old->first + old->second->GetSize() != iter->first) {
+            return false;
+        }
+    }
 
-    return std::shared_ptr<VfsFile>(new ConcatenatedVfsFile(std::move(files), std::move(name)));
+    return map.begin()->first == 0;
 }
 
 ConcatenatedVfsFile::ConcatenatedVfsFile(std::vector<VirtualFile> files_, std::string name)
@@ -27,8 +32,48 @@ ConcatenatedVfsFile::ConcatenatedVfsFile(std::vector<VirtualFile> files_, std::s
     }
 }
 
+ConcatenatedVfsFile::ConcatenatedVfsFile(std::map<u64, VirtualFile> files_, std::string name)
+    : files(std::move(files_)), name(std::move(name)) {
+    ASSERT(VerifyConcatenationMapContinuity(files));
+}
+
 ConcatenatedVfsFile::~ConcatenatedVfsFile() = default;
 
+VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(std::vector<VirtualFile> files,
+                                                      std::string name) {
+    if (files.empty())
+        return nullptr;
+    if (files.size() == 1)
+        return files[0];
+
+    return std::shared_ptr<VfsFile>(new ConcatenatedVfsFile(std::move(files), std::move(name)));
+}
+
+VirtualFile ConcatenatedVfsFile::MakeConcatenatedFile(u8 filler_byte,
+                                                      std::map<u64, VirtualFile> files,
+                                                      std::string name) {
+    if (files.empty())
+        return nullptr;
+    if (files.size() == 1)
+        return files.begin()->second;
+
+    const auto last_valid = --files.end();
+    for (auto iter = files.begin(); iter != last_valid;) {
+        const auto old = iter++;
+        if (old->first + old->second->GetSize() != iter->first) {
+            files.emplace(old->first + old->second->GetSize(),
+                          std::make_shared<StaticVfsFile>(filler_byte, iter->first - old->first -
+                                                                           old->second->GetSize()));
+        }
+    }
+
+    // Ensure the map starts at offset 0 (start of file), otherwise pad to fill.
+    if (files.begin()->first != 0)
+        files.emplace(0, std::make_shared<StaticVfsFile>(filler_byte, files.begin()->first));
+
+    return std::shared_ptr<VfsFile>(new ConcatenatedVfsFile(std::move(files), std::move(name)));
+}
+
 std::string ConcatenatedVfsFile::GetName() const {
     if (files.empty())
         return "";
@@ -62,7 +107,7 @@ bool ConcatenatedVfsFile::IsReadable() const {
 }
 
 std::size_t ConcatenatedVfsFile::Read(u8* data, std::size_t length, std::size_t offset) const {
-    auto entry = files.end();
+    auto entry = --files.end();
     for (auto iter = files.begin(); iter != files.end(); ++iter) {
         if (iter->first > offset) {
             entry = --iter;
@@ -70,20 +115,17 @@ std::size_t ConcatenatedVfsFile::Read(u8* data, std::size_t length, std::size_t
         }
     }
 
-    // Check if the entry should be the last one. The loop above will make it end().
-    if (entry == files.end() && offset < files.rbegin()->first + files.rbegin()->second->GetSize())
-        --entry;
-
-    if (entry == files.end())
+    if (entry->first + entry->second->GetSize() <= offset)
         return 0;
 
-    const auto remaining = entry->second->GetSize() + offset - entry->first;
-    if (length > remaining) {
-        return entry->second->Read(data, remaining, offset - entry->first) +
-               Read(data + remaining, length - remaining, offset + remaining);
+    const auto read_in =
+        std::min<u64>(entry->first + entry->second->GetSize() - offset, entry->second->GetSize());
+    if (length > read_in) {
+        return entry->second->Read(data, read_in, offset - entry->first) +
+               Read(data + read_in, length - read_in, offset + read_in);
     }
 
-    return entry->second->Read(data, length, offset - entry->first);
+    return entry->second->Read(data, std::min<u64>(read_in, length), offset - entry->first);
 }
 
 std::size_t ConcatenatedVfsFile::Write(const u8* data, std::size_t length, std::size_t offset) {
@@ -93,4 +135,5 @@ std::size_t ConcatenatedVfsFile::Write(const u8* data, std::size_t length, std::
 bool ConcatenatedVfsFile::Rename(std::string_view name) {
     return false;
 }
+
 } // namespace FileSys

src/core/file_sys/vfs_concat.h

@@ -4,26 +4,30 @@
 
 #pragma once
 
+#include <map>
 #include <memory>
 #include <string_view>
-#include <boost/container/flat_map.hpp>
 #include "core/file_sys/vfs.h"
 
 namespace FileSys {
 
-// Wrapper function to allow for more efficient handling of files.size() == 0, 1 cases.
-VirtualFile ConcatenateFiles(std::vector<VirtualFile> files, std::string name = "");
-
 // Class that wraps multiple vfs files and concatenates them, making reads seamless. Currently
 // read-only.
 class ConcatenatedVfsFile : public VfsFile {
-    friend VirtualFile ConcatenateFiles(std::vector<VirtualFile> files, std::string name);
-
     ConcatenatedVfsFile(std::vector<VirtualFile> files, std::string name);
+    ConcatenatedVfsFile(std::map<u64, VirtualFile> files, std::string name);
 
 public:
     ~ConcatenatedVfsFile() override;
 
+    /// Wrapper function to allow for more efficient handling of files.size() == 0, 1 cases.
+    static VirtualFile MakeConcatenatedFile(std::vector<VirtualFile> files, std::string name);
+
+    /// Convenience function that turns a map of offsets to files into a concatenated file, filling
+    /// gaps with a given filler byte.
+    static VirtualFile MakeConcatenatedFile(u8 filler_byte, std::map<u64, VirtualFile> files,
+                                            std::string name);
+
     std::string GetName() const override;
     std::size_t GetSize() const override;
     bool Resize(std::size_t new_size) override;
@@ -36,7 +40,7 @@ public:
 
 private:
     // Maps starting offset to file -- more efficient.
-    boost::container::flat_map<u64, VirtualFile> files;
+    std::map<u64, VirtualFile> files;
     std::string name;
 };
 

src/core/file_sys/vfs_layered.cpp

@@ -0,0 +1,132 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <utility>
+#include "core/file_sys/vfs_layered.h"
+
+namespace FileSys {
+
+LayeredVfsDirectory::LayeredVfsDirectory(std::vector<VirtualDir> dirs, std::string name)
+    : dirs(std::move(dirs)), name(std::move(name)) {}
+
+LayeredVfsDirectory::~LayeredVfsDirectory() = default;
+
+VirtualDir LayeredVfsDirectory::MakeLayeredDirectory(std::vector<VirtualDir> dirs,
+                                                     std::string name) {
+    if (dirs.empty())
+        return nullptr;
+    if (dirs.size() == 1)
+        return dirs[0];
+
+    return std::shared_ptr<VfsDirectory>(new LayeredVfsDirectory(std::move(dirs), std::move(name)));
+}
+
+std::shared_ptr<VfsFile> LayeredVfsDirectory::GetFileRelative(std::string_view path) const {
+    for (const auto& layer : dirs) {
+        const auto file = layer->GetFileRelative(path);
+        if (file != nullptr)
+            return file;
+    }
+
+    return nullptr;
+}
+
+std::shared_ptr<VfsDirectory> LayeredVfsDirectory::GetDirectoryRelative(
+    std::string_view path) const {
+    std::vector<VirtualDir> out;
+    for (const auto& layer : dirs) {
+        auto dir = layer->GetDirectoryRelative(path);
+        if (dir != nullptr)
+            out.push_back(std::move(dir));
+    }
+
+    return MakeLayeredDirectory(std::move(out));
+}
+
+std::shared_ptr<VfsFile> LayeredVfsDirectory::GetFile(std::string_view name) const {
+    return GetFileRelative(name);
+}
+
+std::shared_ptr<VfsDirectory> LayeredVfsDirectory::GetSubdirectory(std::string_view name) const {
+    return GetDirectoryRelative(name);
+}
+
+std::string LayeredVfsDirectory::GetFullPath() const {
+    return dirs[0]->GetFullPath();
+}
+
+std::vector<std::shared_ptr<VfsFile>> LayeredVfsDirectory::GetFiles() const {
+    std::vector<VirtualFile> out;
+    for (const auto& layer : dirs) {
+        for (const auto& file : layer->GetFiles()) {
+            if (std::find_if(out.begin(), out.end(), [&file](const VirtualFile& comp) {
+                    return comp->GetName() == file->GetName();
+                }) == out.end()) {
+                out.push_back(file);
+            }
+        }
+    }
+
+    return out;
+}
+
+std::vector<std::shared_ptr<VfsDirectory>> LayeredVfsDirectory::GetSubdirectories() const {
+    std::vector<std::string> names;
+    for (const auto& layer : dirs) {
+        for (const auto& sd : layer->GetSubdirectories()) {
+            if (std::find(names.begin(), names.end(), sd->GetName()) == names.end())
+                names.push_back(sd->GetName());
+        }
+    }
+
+    std::vector<VirtualDir> out;
+    out.reserve(names.size());
+    for (const auto& subdir : names)
+        out.push_back(GetSubdirectory(subdir));
+
+    return out;
+}
+
+bool LayeredVfsDirectory::IsWritable() const {
+    return false;
+}
+
+bool LayeredVfsDirectory::IsReadable() const {
+    return true;
+}
+
+std::string LayeredVfsDirectory::GetName() const {
+    return name.empty() ? dirs[0]->GetName() : name;
+}
+
+std::shared_ptr<VfsDirectory> LayeredVfsDirectory::GetParentDirectory() const {
+    return dirs[0]->GetParentDirectory();
+}
+
+std::shared_ptr<VfsDirectory> LayeredVfsDirectory::CreateSubdirectory(std::string_view name) {
+    return nullptr;
+}
+
+std::shared_ptr<VfsFile> LayeredVfsDirectory::CreateFile(std::string_view name) {
+    return nullptr;
+}
+
+bool LayeredVfsDirectory::DeleteSubdirectory(std::string_view name) {
+    return false;
+}
+
+bool LayeredVfsDirectory::DeleteFile(std::string_view name) {
+    return false;
+}
+
+bool LayeredVfsDirectory::Rename(std::string_view name_) {
+    name = name_;
+    return true;
+}
+
+bool LayeredVfsDirectory::ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) {
+    return false;
+}
+} // namespace FileSys

src/core/file_sys/vfs_layered.h

@@ -0,0 +1,50 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include "core/file_sys/vfs.h"
+
+namespace FileSys {
+
+// Class that stacks multiple VfsDirectories on top of each other, attempting to read from the first
+// one and falling back to the one after. The highest priority directory (overwrites all others)
+// should be element 0 in the dirs vector.
+class LayeredVfsDirectory : public VfsDirectory {
+    LayeredVfsDirectory(std::vector<VirtualDir> dirs, std::string name);
+
+public:
+    ~LayeredVfsDirectory() override;
+
+    /// Wrapper function to allow for more efficient handling of dirs.size() == 0, 1 cases.
+    static VirtualDir MakeLayeredDirectory(std::vector<VirtualDir> dirs, std::string name = "");
+
+    std::shared_ptr<VfsFile> GetFileRelative(std::string_view path) const override;
+    std::shared_ptr<VfsDirectory> GetDirectoryRelative(std::string_view path) const override;
+    std::shared_ptr<VfsFile> GetFile(std::string_view name) const override;
+    std::shared_ptr<VfsDirectory> GetSubdirectory(std::string_view name) const override;
+    std::string GetFullPath() const override;
+
+    std::vector<std::shared_ptr<VfsFile>> GetFiles() const override;
+    std::vector<std::shared_ptr<VfsDirectory>> GetSubdirectories() const override;
+    bool IsWritable() const override;
+    bool IsReadable() const override;
+    std::string GetName() const override;
+    std::shared_ptr<VfsDirectory> GetParentDirectory() const override;
+    std::shared_ptr<VfsDirectory> CreateSubdirectory(std::string_view name) override;
+    std::shared_ptr<VfsFile> CreateFile(std::string_view name) override;
+    bool DeleteSubdirectory(std::string_view name) override;
+    bool DeleteFile(std::string_view name) override;
+    bool Rename(std::string_view name) override;
+
+protected:
+    bool ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) override;
+
+private:
+    std::vector<VirtualDir> dirs;
+    std::string name;
+};
+
+} // namespace FileSys

src/core/file_sys/vfs_real.cpp

@@ -413,6 +413,23 @@ std::string RealVfsDirectory::GetFullPath() const {
     return out;
 }
 
+std::map<std::string, VfsEntryType, std::less<>> RealVfsDirectory::GetEntries() const {
+    if (perms == Mode::Append)
+        return {};
+
+    std::map<std::string, VfsEntryType, std::less<>> out;
+    FileUtil::ForeachDirectoryEntry(
+        nullptr, path,
+        [&out](u64* entries_out, const std::string& directory, const std::string& filename) {
+            const std::string full_path = directory + DIR_SEP + filename;
+            out.emplace(filename, FileUtil::IsDirectory(full_path) ? VfsEntryType::Directory
+                                                                   : VfsEntryType::File);
+            return true;
+        });
+
+    return out;
+}
+
 bool RealVfsDirectory::ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) {
     return false;
 }

src/core/file_sys/vfs_real.h

@@ -98,6 +98,7 @@ public:
     bool DeleteFile(std::string_view name) override;
     bool Rename(std::string_view name) override;
     std::string GetFullPath() const override;
+    std::map<std::string, VfsEntryType, std::less<>> GetEntries() const override;
 
 protected:
     bool ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) override;

src/core/file_sys/vfs_static.h

@@ -0,0 +1,79 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <algorithm>
+#include <memory>
+#include <string_view>
+
+#include "core/file_sys/vfs.h"
+
+namespace FileSys {
+
+class StaticVfsFile : public VfsFile {
+public:
+    explicit StaticVfsFile(u8 value, std::size_t size = 0, std::string name = "",
+                           VirtualDir parent = nullptr)
+        : value{value}, size{size}, name{std::move(name)}, parent{std::move(parent)} {}
+
+    std::string GetName() const override {
+        return name;
+    }
+
+    std::size_t GetSize() const override {
+        return size;
+    }
+
+    bool Resize(std::size_t new_size) override {
+        size = new_size;
+        return true;
+    }
+
+    std::shared_ptr<VfsDirectory> GetContainingDirectory() const override {
+        return parent;
+    }
+
+    bool IsWritable() const override {
+        return false;
+    }
+
+    bool IsReadable() const override {
+        return true;
+    }
+
+    std::size_t Read(u8* data, std::size_t length, std::size_t offset) const override {
+        const auto read = std::min(length, size - offset);
+        std::fill(data, data + read, value);
+        return read;
+    }
+
+    std::size_t Write(const u8* data, std::size_t length, std::size_t offset) override {
+        return 0;
+    }
+
+    boost::optional<u8> ReadByte(std::size_t offset) const override {
+        if (offset < size)
+            return value;
+        return boost::none;
+    }
+
+    std::vector<u8> ReadBytes(std::size_t length, std::size_t offset) const override {
+        const auto read = std::min(length, size - offset);
+        return std::vector<u8>(read, value);
+    }
+
+    bool Rename(std::string_view new_name) override {
+        name = new_name;
+        return true;
+    }
+
+private:
+    u8 value;
+    std::size_t size;
+    std::string name;
+    VirtualDir parent;
+};
+
+} // namespace FileSys

src/core/file_sys/vfs_vector.cpp

@@ -3,10 +3,64 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
+#include <cstring>
 #include <utility>
 #include "core/file_sys/vfs_vector.h"
 
 namespace FileSys {
+VectorVfsFile::VectorVfsFile(std::vector<u8> initial_data, std::string name, VirtualDir parent)
+    : data(std::move(initial_data)), parent(std::move(parent)), name(std::move(name)) {}
+
+VectorVfsFile::~VectorVfsFile() = default;
+
+std::string VectorVfsFile::GetName() const {
+    return name;
+}
+
+size_t VectorVfsFile::GetSize() const {
+    return data.size();
+}
+
+bool VectorVfsFile::Resize(size_t new_size) {
+    data.resize(new_size);
+    return true;
+}
+
+std::shared_ptr<VfsDirectory> VectorVfsFile::GetContainingDirectory() const {
+    return parent;
+}
+
+bool VectorVfsFile::IsWritable() const {
+    return true;
+}
+
+bool VectorVfsFile::IsReadable() const {
+    return true;
+}
+
+std::size_t VectorVfsFile::Read(u8* data_, std::size_t length, std::size_t offset) const {
+    const auto read = std::min(length, data.size() - offset);
+    std::memcpy(data_, data.data() + offset, read);
+    return read;
+}
+
+std::size_t VectorVfsFile::Write(const u8* data_, std::size_t length, std::size_t offset) {
+    if (offset + length > data.size())
+        data.resize(offset + length);
+    const auto write = std::min(length, data.size() - offset);
+    std::memcpy(data.data(), data_, write);
+    return write;
+}
+
+bool VectorVfsFile::Rename(std::string_view name_) {
+    name = name_;
+    return true;
+}
+
+void VectorVfsFile::Assign(std::vector<u8> new_data) {
+    data = std::move(new_data);
+}
+
 VectorVfsDirectory::VectorVfsDirectory(std::vector<VirtualFile> files_,
                                        std::vector<VirtualDir> dirs_, std::string name_,
                                        VirtualDir parent_)

src/core/file_sys/vfs_vector.h

@@ -8,6 +8,31 @@
 
 namespace FileSys {
 
+// An implementation of VfsFile that is backed by a vector optionally supplied upon construction
+class VectorVfsFile : public VfsFile {
+public:
+    explicit VectorVfsFile(std::vector<u8> initial_data = {}, std::string name = "",
+                           VirtualDir parent = nullptr);
+    ~VectorVfsFile() override;
+
+    std::string GetName() const override;
+    std::size_t GetSize() const override;
+    bool Resize(std::size_t new_size) override;
+    std::shared_ptr<VfsDirectory> GetContainingDirectory() const override;
+    bool IsWritable() const override;
+    bool IsReadable() const override;
+    std::size_t Read(u8* data, std::size_t length, std::size_t offset) const override;
+    std::size_t Write(const u8* data, std::size_t length, std::size_t offset) override;
+    bool Rename(std::string_view name) override;
+
+    virtual void Assign(std::vector<u8> new_data);
+
+private:
+    std::vector<u8> data;
+    VirtualDir parent;
+    std::string name;
+};
+
 // An implementation of VfsDirectory that maintains two vectors for subdirectories and files.
 // Vector data is supplied upon construction.
 class VectorVfsDirectory : public VfsDirectory {

src/core/hle/kernel/process.cpp

@@ -7,10 +7,12 @@
 #include "common/assert.h"
 #include "common/common_funcs.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
+#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/kernel/vm_manager.h"
 #include "core/memory.h"
@@ -128,6 +130,91 @@ void Process::Run(VAddr entry_point, s32 main_thread_priority, u32 stack_size) {
     Kernel::SetupMainThread(kernel, entry_point, main_thread_priority, *this);
 }
 
+void Process::PrepareForTermination() {
+    status = ProcessStatus::Exited;
+
+    const auto stop_threads = [this](const std::vector<SharedPtr<Thread>>& thread_list) {
+        for (auto& thread : thread_list) {
+            if (thread->owner_process != this)
+                continue;
+
+            if (thread == GetCurrentThread())
+                continue;
+
+            // TODO(Subv): When are the other running/ready threads terminated?
+            ASSERT_MSG(thread->status == ThreadStatus::WaitSynchAny ||
+                           thread->status == ThreadStatus::WaitSynchAll,
+                       "Exiting processes with non-waiting threads is currently unimplemented");
+
+            thread->Stop();
+        }
+    };
+
+    auto& system = Core::System::GetInstance();
+    stop_threads(system.Scheduler(0)->GetThreadList());
+    stop_threads(system.Scheduler(1)->GetThreadList());
+    stop_threads(system.Scheduler(2)->GetThreadList());
+    stop_threads(system.Scheduler(3)->GetThreadList());
+}
+
+/**
+ * Finds a free location for the TLS section of a thread.
+ * @param tls_slots The TLS page array of the thread's owner process.
+ * Returns a tuple of (page, slot, alloc_needed) where:
+ * page: The index of the first allocated TLS page that has free slots.
+ * slot: The index of the first free slot in the indicated page.
+ * alloc_needed: Whether there's a need to allocate a new TLS page (All pages are full).
+ */
+static std::tuple<std::size_t, std::size_t, bool> FindFreeThreadLocalSlot(
+    const std::vector<std::bitset<8>>& tls_slots) {
+    // Iterate over all the allocated pages, and try to find one where not all slots are used.
+    for (std::size_t page = 0; page < tls_slots.size(); ++page) {
+        const auto& page_tls_slots = tls_slots[page];
+        if (!page_tls_slots.all()) {
+            // We found a page with at least one free slot, find which slot it is
+            for (std::size_t slot = 0; slot < page_tls_slots.size(); ++slot) {
+                if (!page_tls_slots.test(slot)) {
+                    return std::make_tuple(page, slot, false);
+                }
+            }
+        }
+    }
+
+    return std::make_tuple(0, 0, true);
+}
+
+VAddr Process::MarkNextAvailableTLSSlotAsUsed(Thread& thread) {
+    auto [available_page, available_slot, needs_allocation] = FindFreeThreadLocalSlot(tls_slots);
+
+    if (needs_allocation) {
+        tls_slots.emplace_back(0); // The page is completely available at the start
+        available_page = tls_slots.size() - 1;
+        available_slot = 0; // Use the first slot in the new page
+
+        // Allocate some memory from the end of the linear heap for this region.
+        auto& tls_memory = thread.GetTLSMemory();
+        tls_memory->insert(tls_memory->end(), Memory::PAGE_SIZE, 0);
+
+        vm_manager.RefreshMemoryBlockMappings(tls_memory.get());
+
+        vm_manager.MapMemoryBlock(Memory::TLS_AREA_VADDR + available_page * Memory::PAGE_SIZE,
+                                  tls_memory, 0, Memory::PAGE_SIZE, MemoryState::ThreadLocal);
+    }
+
+    tls_slots[available_page].set(available_slot);
+
+    return Memory::TLS_AREA_VADDR + available_page * Memory::PAGE_SIZE +
+           available_slot * Memory::TLS_ENTRY_SIZE;
+}
+
+void Process::FreeTLSSlot(VAddr tls_address) {
+    const VAddr tls_base = tls_address - Memory::TLS_AREA_VADDR;
+    const VAddr tls_page = tls_base / Memory::PAGE_SIZE;
+    const VAddr tls_slot = (tls_base % Memory::PAGE_SIZE) / Memory::TLS_ENTRY_SIZE;
+
+    tls_slots[tls_page].reset(tls_slot);
+}
+
 void Process::LoadModule(SharedPtr<CodeSet> module_, VAddr base_addr) {
     const auto MapSegment = [&](CodeSet::Segment& segment, VMAPermission permissions,
                                 MemoryState memory_state) {

src/core/hle/kernel/process.h

@@ -131,6 +131,16 @@ public:
         return HANDLE_TYPE;
     }
 
+    /// Gets the current status of the process
+    ProcessStatus GetStatus() const {
+        return status;
+    }
+
+    /// Gets the unique ID that identifies this particular process.
+    u32 GetProcessID() const {
+        return process_id;
+    }
+
     /// Title ID corresponding to the process
     u64 program_id;
 
@@ -154,11 +164,6 @@ public:
     u32 allowed_processor_mask = THREADPROCESSORID_DEFAULT_MASK;
     u32 allowed_thread_priority_mask = 0xFFFFFFFF;
     u32 is_virtual_address_memory_enabled = 0;
-    /// Current status of the process
-    ProcessStatus status;
-
-    /// The ID of this process
-    u32 process_id = 0;
 
     /**
      * Parses a list of kernel capability descriptors (as found in the ExHeader) and applies them
@@ -171,13 +176,42 @@ public:
      */
     void Run(VAddr entry_point, s32 main_thread_priority, u32 stack_size);
 
+    /**
+     * Prepares a process for termination by stopping all of its threads
+     * and clearing any other resources.
+     */
+    void PrepareForTermination();
+
     void LoadModule(SharedPtr<CodeSet> module_, VAddr base_addr);
 
     ///////////////////////////////////////////////////////////////////////////////////////////////
     // Memory Management
 
+    // Marks the next available region as used and returns the address of the slot.
+    VAddr MarkNextAvailableTLSSlotAsUsed(Thread& thread);
+
+    // Frees a used TLS slot identified by the given address
+    void FreeTLSSlot(VAddr tls_address);
+
+    ResultVal<VAddr> HeapAllocate(VAddr target, u64 size, VMAPermission perms);
+    ResultCode HeapFree(VAddr target, u32 size);
+
+    ResultCode MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size);
+
+    ResultCode UnmapMemory(VAddr dst_addr, VAddr src_addr, u64 size);
+
     VMManager vm_manager;
 
+private:
+    explicit Process(KernelCore& kernel);
+    ~Process() override;
+
+    /// Current status of the process
+    ProcessStatus status;
+
+    /// The ID of this process
+    u32 process_id = 0;
+
     // Memory used to back the allocations in the regular heap. A single vector is used to cover
     // the entire virtual address space extents that bound the allocations, including any holes.
     // This makes deallocation and reallocation of holes fast and keeps process memory contiguous
@@ -197,17 +231,6 @@ public:
     std::vector<std::bitset<8>> tls_slots;
 
     std::string name;
-
-    ResultVal<VAddr> HeapAllocate(VAddr target, u64 size, VMAPermission perms);
-    ResultCode HeapFree(VAddr target, u32 size);
-
-    ResultCode MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size);
-
-    ResultCode UnmapMemory(VAddr dst_addr, VAddr src_addr, u64 size);
-
-private:
-    explicit Process(KernelCore& kernel);
-    ~Process() override;
 };
 
 } // namespace Kernel

src/core/hle/kernel/scheduler.cpp

@@ -17,7 +17,7 @@ namespace Kernel {
 
 std::mutex Scheduler::scheduler_mutex;
 
-Scheduler::Scheduler(Core::ARM_Interface* cpu_core) : cpu_core(cpu_core) {}
+Scheduler::Scheduler(Core::ARM_Interface& cpu_core) : cpu_core(cpu_core) {}
 
 Scheduler::~Scheduler() {
     for (auto& thread : thread_list) {
@@ -59,9 +59,9 @@ void Scheduler::SwitchContext(Thread* new_thread) {
     // Save context for previous thread
     if (previous_thread) {
         previous_thread->last_running_ticks = CoreTiming::GetTicks();
-        cpu_core->SaveContext(previous_thread->context);
+        cpu_core.SaveContext(previous_thread->context);
         // Save the TPIDR_EL0 system register in case it was modified.
-        previous_thread->tpidr_el0 = cpu_core->GetTPIDR_EL0();
+        previous_thread->tpidr_el0 = cpu_core.GetTPIDR_EL0();
 
         if (previous_thread->status == ThreadStatus::Running) {
             // This is only the case when a reschedule is triggered without the current thread
@@ -91,10 +91,10 @@ void Scheduler::SwitchContext(Thread* new_thread) {
             SetCurrentPageTable(&Core::CurrentProcess()->vm_manager.page_table);
         }
 
-        cpu_core->LoadContext(new_thread->context);
-        cpu_core->SetTlsAddress(new_thread->GetTLSAddress());
-        cpu_core->SetTPIDR_EL0(new_thread->GetTPIDR_EL0());
-        cpu_core->ClearExclusiveState();
+        cpu_core.LoadContext(new_thread->context);
+        cpu_core.SetTlsAddress(new_thread->GetTLSAddress());
+        cpu_core.SetTPIDR_EL0(new_thread->GetTPIDR_EL0());
+        cpu_core.ClearExclusiveState();
     } else {
         current_thread = nullptr;
         // Note: We do not reset the current process and current page table when idling because

src/core/hle/kernel/scheduler.h

@@ -19,7 +19,7 @@ namespace Kernel {
 
 class Scheduler final {
 public:
-    explicit Scheduler(Core::ARM_Interface* cpu_core);
+    explicit Scheduler(Core::ARM_Interface& cpu_core);
     ~Scheduler();
 
     /// Returns whether there are any threads that are ready to run.
@@ -72,7 +72,7 @@ private:
 
     SharedPtr<Thread> current_thread = nullptr;
 
-    Core::ARM_Interface* cpu_core;
+    Core::ARM_Interface& cpu_core;
 
     static std::mutex scheduler_mutex;
 };

src/core/hle/kernel/svc.cpp

@@ -169,7 +169,7 @@ static ResultCode GetProcessId(u32* process_id, Handle process_handle) {
         return ERR_INVALID_HANDLE;
     }
 
-    *process_id = process->process_id;
+    *process_id = process->GetProcessID();
     return RESULT_SUCCESS;
 }
 
@@ -530,35 +530,13 @@ static ResultCode QueryMemory(MemoryInfo* memory_info, PageInfo* page_info, VAdd
 
 /// Exits the current process
 static void ExitProcess() {
-    LOG_INFO(Kernel_SVC, "Process {} exiting", Core::CurrentProcess()->process_id);
+    auto& current_process = Core::CurrentProcess();
 
-    ASSERT_MSG(Core::CurrentProcess()->status == ProcessStatus::Running,
+    LOG_INFO(Kernel_SVC, "Process {} exiting", current_process->GetProcessID());
+    ASSERT_MSG(current_process->GetStatus() == ProcessStatus::Running,
                "Process has already exited");
 
-    Core::CurrentProcess()->status = ProcessStatus::Exited;
-
-    auto stop_threads = [](const std::vector<SharedPtr<Thread>>& thread_list) {
-        for (auto& thread : thread_list) {
-            if (thread->owner_process != Core::CurrentProcess())
-                continue;
-
-            if (thread == GetCurrentThread())
-                continue;
-
-            // TODO(Subv): When are the other running/ready threads terminated?
-            ASSERT_MSG(thread->status == ThreadStatus::WaitSynchAny ||
-                           thread->status == ThreadStatus::WaitSynchAll,
-                       "Exiting processes with non-waiting threads is currently unimplemented");
-
-            thread->Stop();
-        }
-    };
-
-    auto& system = Core::System::GetInstance();
-    stop_threads(system.Scheduler(0)->GetThreadList());
-    stop_threads(system.Scheduler(1)->GetThreadList());
-    stop_threads(system.Scheduler(2)->GetThreadList());
-    stop_threads(system.Scheduler(3)->GetThreadList());
+    current_process->PrepareForTermination();
 
     // Kill the current thread
     GetCurrentThread()->Stop();
@@ -1039,7 +1017,7 @@ static const FunctionDef SVC_Table[] = {
     {0x2B, nullptr, "FlushDataCache"},
     {0x2C, nullptr, "MapPhysicalMemory"},
     {0x2D, nullptr, "UnmapPhysicalMemory"},
-    {0x2E, nullptr, "GetNextThreadInfo"},
+    {0x2E, nullptr, "GetFutureThreadInfo"},
     {0x2F, nullptr, "GetLastThreadInfo"},
     {0x30, nullptr, "GetResourceLimitLimitValue"},
     {0x31, nullptr, "GetResourceLimitCurrentValue"},
@@ -1065,11 +1043,11 @@ static const FunctionDef SVC_Table[] = {
     {0x45, nullptr, "CreateEvent"},
     {0x46, nullptr, "Unknown"},
     {0x47, nullptr, "Unknown"},
-    {0x48, nullptr, "AllocateUnsafeMemory"},
-    {0x49, nullptr, "FreeUnsafeMemory"},
-    {0x4A, nullptr, "SetUnsafeAllocationLimit"},
-    {0x4B, nullptr, "CreateJitMemory"},
-    {0x4C, nullptr, "MapJitMemory"},
+    {0x48, nullptr, "MapPhysicalMemoryUnsafe"},
+    {0x49, nullptr, "UnmapPhysicalMemoryUnsafe"},
+    {0x4A, nullptr, "SetUnsafeLimit"},
+    {0x4B, nullptr, "CreateCodeMemory"},
+    {0x4C, nullptr, "ControlCodeMemory"},
     {0x4D, nullptr, "SleepSystem"},
     {0x4E, nullptr, "ReadWriteRegister"},
     {0x4F, nullptr, "SetProcessActivity"},
@@ -1104,7 +1082,7 @@ static const FunctionDef SVC_Table[] = {
     {0x6C, nullptr, "SetHardwareBreakPoint"},
     {0x6D, nullptr, "GetDebugThreadParam"},
     {0x6E, nullptr, "Unknown"},
-    {0x6F, nullptr, "GetMemoryInfo"},
+    {0x6F, nullptr, "GetSystemInfo"},
     {0x70, nullptr, "CreatePort"},
     {0x71, nullptr, "ManageNamedPort"},
     {0x72, nullptr, "ConnectToPort"},

src/core/hle/kernel/thread.cpp

@@ -65,10 +65,7 @@ void Thread::Stop() {
     wait_objects.clear();
 
     // Mark the TLS slot in the thread's page as free.
-    const u64 tls_page = (tls_address - Memory::TLS_AREA_VADDR) / Memory::PAGE_SIZE;
-    const u64 tls_slot =
-        ((tls_address - Memory::TLS_AREA_VADDR) % Memory::PAGE_SIZE) / Memory::TLS_ENTRY_SIZE;
-    Core::CurrentProcess()->tls_slots[tls_page].reset(tls_slot);
+    owner_process->FreeTLSSlot(tls_address);
 }
 
 void WaitCurrentThread_Sleep() {
@@ -177,32 +174,6 @@ void Thread::ResumeFromWait() {
     Core::System::GetInstance().CpuCore(processor_id).PrepareReschedule();
 }
 
-/**
- * Finds a free location for the TLS section of a thread.
- * @param tls_slots The TLS page array of the thread's owner process.
- * Returns a tuple of (page, slot, alloc_needed) where:
- * page: The index of the first allocated TLS page that has free slots.
- * slot: The index of the first free slot in the indicated page.
- * alloc_needed: Whether there's a need to allocate a new TLS page (All pages are full).
- */
-static std::tuple<std::size_t, std::size_t, bool> GetFreeThreadLocalSlot(
-    const std::vector<std::bitset<8>>& tls_slots) {
-    // Iterate over all the allocated pages, and try to find one where not all slots are used.
-    for (std::size_t page = 0; page < tls_slots.size(); ++page) {
-        const auto& page_tls_slots = tls_slots[page];
-        if (!page_tls_slots.all()) {
-            // We found a page with at least one free slot, find which slot it is
-            for (std::size_t slot = 0; slot < page_tls_slots.size(); ++slot) {
-                if (!page_tls_slots.test(slot)) {
-                    return std::make_tuple(page, slot, false);
-                }
-            }
-        }
-    }
-
-    return std::make_tuple(0, 0, true);
-}
-
 /**
  * Resets a thread context, making it ready to be scheduled and run by the CPU
  * @param context Thread context to reset
@@ -264,32 +235,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
     thread->owner_process = owner_process;
     thread->scheduler = Core::System::GetInstance().Scheduler(processor_id);
     thread->scheduler->AddThread(thread, priority);
-
-    // Find the next available TLS index, and mark it as used
-    auto& tls_slots = owner_process->tls_slots;
-
-    auto [available_page, available_slot, needs_allocation] = GetFreeThreadLocalSlot(tls_slots);
-    if (needs_allocation) {
-        tls_slots.emplace_back(0); // The page is completely available at the start
-        available_page = tls_slots.size() - 1;
-        available_slot = 0; // Use the first slot in the new page
-
-        // Allocate some memory from the end of the linear heap for this region.
-        const std::size_t offset = thread->tls_memory->size();
-        thread->tls_memory->insert(thread->tls_memory->end(), Memory::PAGE_SIZE, 0);
-
-        auto& vm_manager = owner_process->vm_manager;
-        vm_manager.RefreshMemoryBlockMappings(thread->tls_memory.get());
-
-        vm_manager.MapMemoryBlock(Memory::TLS_AREA_VADDR + available_page * Memory::PAGE_SIZE,
-                                  thread->tls_memory, 0, Memory::PAGE_SIZE,
-                                  MemoryState::ThreadLocal);
-    }
-
-    // Mark the slot as used
-    tls_slots[available_page].set(available_slot);
-    thread->tls_address = Memory::TLS_AREA_VADDR + available_page * Memory::PAGE_SIZE +
-                          available_slot * Memory::TLS_ENTRY_SIZE;
+    thread->tls_address = thread->owner_process->MarkNextAvailableTLSSlotAsUsed(*thread);
 
     // TODO(peachum): move to ScheduleThread() when scheduler is added so selected core is used
     // to initialize the context

src/core/hle/kernel/thread.h

@@ -62,6 +62,9 @@ enum class ThreadWakeupReason {
 
 class Thread final : public WaitObject {
 public:
+    using TLSMemory = std::vector<u8>;
+    using TLSMemoryPtr = std::shared_ptr<TLSMemory>;
+
     /**
      * Creates and returns a new thread. The new thread is immediately scheduled
      * @param kernel The kernel instance this thread will be created under.
@@ -134,6 +137,14 @@ public:
         return thread_id;
     }
 
+    TLSMemoryPtr& GetTLSMemory() {
+        return tls_memory;
+    }
+
+    const TLSMemoryPtr& GetTLSMemory() const {
+        return tls_memory;
+    }
+
     /**
      * Resumes a thread from waiting
      */
@@ -269,7 +280,7 @@ private:
     explicit Thread(KernelCore& kernel);
     ~Thread() override;
 
-    std::shared_ptr<std::vector<u8>> tls_memory = std::make_shared<std::vector<u8>>();
+    TLSMemoryPtr tls_memory = std::make_shared<TLSMemory>();
 };
 
 /**

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

@@ -10,6 +10,7 @@
 #include "common/alignment.h"
 #include "common/common_funcs.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/event.h"
 #include "core/hle/kernel/hle_ipc.h"
@@ -25,7 +26,7 @@ public:
             {0, &IAudioRenderer::GetAudioRendererSampleRate, "GetAudioRendererSampleRate"},
             {1, &IAudioRenderer::GetAudioRendererSampleCount, "GetAudioRendererSampleCount"},
             {2, &IAudioRenderer::GetAudioRendererMixBufferCount, "GetAudioRendererMixBufferCount"},
-            {3, nullptr, "GetAudioRendererState"},
+            {3, &IAudioRenderer::GetAudioRendererState, "GetAudioRendererState"},
             {4, &IAudioRenderer::RequestUpdateAudioRenderer, "RequestUpdateAudioRenderer"},
             {5, &IAudioRenderer::StartAudioRenderer, "StartAudioRenderer"},
             {6, &IAudioRenderer::StopAudioRenderer, "StopAudioRenderer"},
@@ -62,6 +63,13 @@ private:
         LOG_DEBUG(Service_Audio, "called");
     }
 
+    void GetAudioRendererState(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 3};
+        rb.Push(RESULT_SUCCESS);
+        rb.Push<u32>(renderer->GetState());
+        LOG_DEBUG(Service_Audio, "called");
+    }
+
     void GetAudioRendererMixBufferCount(Kernel::HLERequestContext& ctx) {
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);

src/core/hle/service/fatal/fatal.cpp

@@ -2,8 +2,17 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <array>
+#include <cstring>
+#include <ctime>
+#include <fmt/time.h>
+#include "common/file_util.h"
 #include "common/logging/log.h"
+#include "common/scm_rev.h"
+#include "common/swap.h"
+#include "core/core.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/process.h"
 #include "core/hle/service/fatal/fatal.h"
 #include "core/hle/service/fatal/fatal_p.h"
 #include "core/hle/service/fatal/fatal_u.h"
@@ -15,16 +24,142 @@ Module::Interface::Interface(std::shared_ptr<Module> module, const char* name)
 
 Module::Interface::~Interface() = default;
 
+struct FatalInfo {
+    std::array<u64_le, 31> registers{}; // TODO(ogniK): See if this actually is registers or
+                                        // not(find a game which has non zero valeus)
+    u64_le unk0{};
+    u64_le unk1{};
+    u64_le unk2{};
+    u64_le unk3{};
+    u64_le unk4{};
+    u64_le unk5{};
+    u64_le unk6{};
+
+    std::array<u64_le, 32> backtrace{};
+    u64_le unk7{};
+    u64_le unk8{};
+    u32_le backtrace_size{};
+    u32_le unk9{};
+    u32_le unk10{}; // TODO(ogniK): Is this even used or is it just padding?
+};
+static_assert(sizeof(FatalInfo) == 0x250, "FatalInfo is an invalid size");
+
+enum class FatalType : u32 {
+    ErrorReportAndScreen = 0,
+    ErrorReport = 1,
+    ErrorScreen = 2,
+};
+
+static void GenerateErrorReport(ResultCode error_code, const FatalInfo& info) {
+    const auto title_id = Core::CurrentProcess()->program_id;
+    std::string crash_report =
+        fmt::format("Yuzu {}-{} crash report\n"
+                    "Title ID:                        {:016x}\n"
+                    "Result:                          0x{:X} ({:04}-{:04d})\n"
+                    "\n",
+                    Common::g_scm_branch, Common::g_scm_desc, title_id, error_code.raw,
+                    2000 + static_cast<u32>(error_code.module.Value()),
+                    static_cast<u32>(error_code.description.Value()), info.unk8, info.unk7);
+    if (info.backtrace_size != 0x0) {
+        crash_report += "Registers:\n";
+        // TODO(ogniK): This is just a guess, find a game which actually has non zero values
+        for (size_t i = 0; i < info.registers.size(); i++) {
+            crash_report +=
+                fmt::format("    X[{:02d}]:                       {:016x}\n", i, info.registers[i]);
+        }
+        crash_report += fmt::format("    Unknown 0:                   {:016x}\n", info.unk0);
+        crash_report += fmt::format("    Unknown 1:                   {:016x}\n", info.unk1);
+        crash_report += fmt::format("    Unknown 2:                   {:016x}\n", info.unk2);
+        crash_report += fmt::format("    Unknown 3:                   {:016x}\n", info.unk3);
+        crash_report += fmt::format("    Unknown 4:                   {:016x}\n", info.unk4);
+        crash_report += fmt::format("    Unknown 5:                   {:016x}\n", info.unk5);
+        crash_report += fmt::format("    Unknown 6:                   {:016x}\n", info.unk6);
+        crash_report += "\nBacktrace:\n";
+        for (size_t i = 0; i < info.backtrace_size; i++) {
+            crash_report +=
+                fmt::format("    Backtrace[{:02d}]:               {:016x}\n", i, info.backtrace[i]);
+        }
+        crash_report += fmt::format("\nUnknown 7:                       0x{:016x}\n", info.unk7);
+        crash_report += fmt::format("Unknown 8:                       0x{:016x}\n", info.unk8);
+        crash_report += fmt::format("Unknown 9:                       0x{:016x}\n", info.unk9);
+        crash_report += fmt::format("Unknown 10:                      0x{:016x}\n", info.unk10);
+    }
+
+    LOG_ERROR(Service_Fatal, "{}", crash_report);
+
+    const std::string crashreport_dir =
+        FileUtil::GetUserPath(FileUtil::UserPath::LogDir) + "crash_logs";
+
+    if (!FileUtil::CreateFullPath(crashreport_dir)) {
+        LOG_ERROR(
+            Service_Fatal,
+            "Unable to create crash report directory. Possible log directory permissions issue.");
+        return;
+    }
+
+    const std::time_t t = std::time(nullptr);
+    const std::string crashreport_filename =
+        fmt::format("{}/{:016x}-{:%F-%H%M%S}.log", crashreport_dir, title_id, *std::localtime(&t));
+
+    auto file = FileUtil::IOFile(crashreport_filename, "wb");
+    if (file.IsOpen()) {
+        file.WriteString(crash_report);
+        LOG_ERROR(Service_Fatal, "Saving error report to {}", crashreport_filename);
+    } else {
+        LOG_ERROR(Service_Fatal, "Failed to save error report to {}", crashreport_filename);
+    }
+}
+
+static void ThrowFatalError(ResultCode error_code, FatalType fatal_type, const FatalInfo& info) {
+    LOG_ERROR(Service_Fatal, "Threw fatal error type {}", static_cast<u32>(fatal_type));
+    switch (fatal_type) {
+    case FatalType::ErrorReportAndScreen:
+        GenerateErrorReport(error_code, info);
+        [[fallthrough]];
+    case FatalType::ErrorScreen:
+        // Since we have no fatal:u error screen. We should just kill execution instead
+        ASSERT(false);
+        break;
+        // Should not throw a fatal screen but should generate an error report
+    case FatalType::ErrorReport:
+        GenerateErrorReport(error_code, info);
+        break;
+    };
+}
+
+void Module::Interface::ThrowFatal(Kernel::HLERequestContext& ctx) {
+    LOG_ERROR(Service_Fatal, "called");
+    IPC::RequestParser rp{ctx};
+    auto error_code = rp.Pop<ResultCode>();
+
+    ThrowFatalError(error_code, FatalType::ErrorScreen, {});
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 void Module::Interface::ThrowFatalWithPolicy(Kernel::HLERequestContext& ctx) {
+    LOG_ERROR(Service_Fatal, "called");
     IPC::RequestParser rp(ctx);
-    u32 error_code = rp.Pop<u32>();
-    LOG_WARNING(Service_Fatal, "(STUBBED) called, error_code=0x{:X}", error_code);
+    auto error_code = rp.Pop<ResultCode>();
+    auto fatal_type = rp.PopEnum<FatalType>();
+
+    ThrowFatalError(error_code, fatal_type, {}); // No info is passed with ThrowFatalWithPolicy
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(RESULT_SUCCESS);
 }
 
 void Module::Interface::ThrowFatalWithCpuContext(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_Fatal, "(STUBBED) called");
+    LOG_ERROR(Service_Fatal, "called");
+    IPC::RequestParser rp(ctx);
+    auto error_code = rp.Pop<ResultCode>();
+    auto fatal_type = rp.PopEnum<FatalType>();
+    auto fatal_info = ctx.ReadBuffer();
+    FatalInfo info{};
+
+    ASSERT_MSG(fatal_info.size() == sizeof(FatalInfo), "Invalid fatal info buffer size!");
+    std::memcpy(&info, fatal_info.data(), sizeof(FatalInfo));
+
+    ThrowFatalError(error_code, fatal_type, info);
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(RESULT_SUCCESS);
 }

src/core/hle/service/fatal/fatal.h

@@ -15,6 +15,7 @@ public:
         explicit Interface(std::shared_ptr<Module> module, const char* name);
         ~Interface() override;
 
+        void ThrowFatal(Kernel::HLERequestContext& ctx);
         void ThrowFatalWithPolicy(Kernel::HLERequestContext& ctx);
         void ThrowFatalWithCpuContext(Kernel::HLERequestContext& ctx);
 

src/core/hle/service/fatal/fatal_u.cpp

@@ -8,7 +8,7 @@ namespace Service::Fatal {
 
 Fatal_U::Fatal_U(std::shared_ptr<Module> module) : Module::Interface(std::move(module), "fatal:u") {
     static const FunctionInfo functions[] = {
-        {0, nullptr, "ThrowFatal"},
+        {0, &Fatal_U::ThrowFatal, "ThrowFatal"},
         {1, &Fatal_U::ThrowFatalWithPolicy, "ThrowFatalWithPolicy"},
         {2, &Fatal_U::ThrowFatalWithCpuContext, "ThrowFatalWithCpuContext"},
     };

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

@@ -343,6 +343,15 @@ std::shared_ptr<FileSys::RegisteredCache> GetSDMCContents() {
     return sdmc_factory->GetSDMCContents();
 }
 
+FileSys::VirtualDir GetModificationLoadRoot(u64 title_id) {
+    LOG_TRACE(Service_FS, "Opening mod load root for tid={:016X}", title_id);
+
+    if (bis_factory == nullptr)
+        return nullptr;
+
+    return bis_factory->GetModificationLoadRoot(title_id);
+}
+
 void CreateFactories(const FileSys::VirtualFilesystem& vfs, bool overwrite) {
     if (overwrite) {
         bis_factory = nullptr;
@@ -354,9 +363,11 @@ void CreateFactories(const FileSys::VirtualFilesystem& vfs, bool overwrite) {
                                              FileSys::Mode::ReadWrite);
     auto sd_directory = vfs->OpenDirectory(FileUtil::GetUserPath(FileUtil::UserPath::SDMCDir),
                                            FileSys::Mode::ReadWrite);
+    auto load_directory = vfs->OpenDirectory(FileUtil::GetUserPath(FileUtil::UserPath::LoadDir),
+                                             FileSys::Mode::ReadWrite);
 
     if (bis_factory == nullptr)
-        bis_factory = std::make_unique<FileSys::BISFactory>(nand_directory);
+        bis_factory = std::make_unique<FileSys::BISFactory>(nand_directory, load_directory);
     if (save_data_factory == nullptr)
         save_data_factory = std::make_unique<FileSys::SaveDataFactory>(std::move(nand_directory));
     if (sdmc_factory == nullptr)

src/core/hle/service/filesystem/filesystem.h

@@ -52,6 +52,8 @@ std::shared_ptr<FileSys::RegisteredCache> GetSystemNANDContents();
 std::shared_ptr<FileSys::RegisteredCache> GetUserNANDContents();
 std::shared_ptr<FileSys::RegisteredCache> GetSDMCContents();
 
+FileSys::VirtualDir GetModificationLoadRoot(u64 title_id);
+
 // Creates the SaveData, SDMC, and BIS Factories. Should be called once and before any function
 // above is called.
 void CreateFactories(const FileSys::VirtualFilesystem& vfs, bool overwrite = true);

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

@@ -2,6 +2,11 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/swap.h"
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/shared_memory.h"
 #include "core/hle/service/hid/irs.h"
 
 namespace Service::HID {
@@ -9,28 +14,145 @@ namespace Service::HID {
 IRS::IRS() : ServiceFramework{"irs"} {
     // clang-format off
     static const FunctionInfo functions[] = {
-        {302, nullptr, "ActivateIrsensor"},
-        {303, nullptr, "DeactivateIrsensor"},
-        {304, nullptr, "GetIrsensorSharedMemoryHandle"},
-        {305, nullptr, "StopImageProcessor"},
-        {306, nullptr, "RunMomentProcessor"},
-        {307, nullptr, "RunClusteringProcessor"},
-        {308, nullptr, "RunImageTransferProcessor"},
-        {309, nullptr, "GetImageTransferProcessorState"},
-        {310, nullptr, "RunTeraPluginProcessor"},
-        {311, nullptr, "GetNpadIrCameraHandle"},
-        {312, nullptr, "RunPointingProcessor"},
-        {313, nullptr, "SuspendImageProcessor"},
-        {314, nullptr, "CheckFirmwareVersion"},
-        {315, nullptr, "SetFunctionLevel"},
-        {316, nullptr, "RunImageTransferExProcessor"},
-        {317, nullptr, "RunIrLedProcessor"},
-        {318, nullptr, "StopImageProcessorAsync"},
-        {319, nullptr, "ActivateIrsensorWithFunctionLevel"},
+        {302, &IRS::ActivateIrsensor, "ActivateIrsensor"},
+        {303, &IRS::DeactivateIrsensor, "DeactivateIrsensor"},
+        {304, &IRS::GetIrsensorSharedMemoryHandle, "GetIrsensorSharedMemoryHandle"},
+        {305, &IRS::StopImageProcessor, "StopImageProcessor"},
+        {306, &IRS::RunMomentProcessor, "RunMomentProcessor"},
+        {307, &IRS::RunClusteringProcessor, "RunClusteringProcessor"},
+        {308, &IRS::RunImageTransferProcessor, "RunImageTransferProcessor"},
+        {309, &IRS::GetImageTransferProcessorState, "GetImageTransferProcessorState"},
+        {310, &IRS::RunTeraPluginProcessor, "RunTeraPluginProcessor"},
+        {311, &IRS::GetNpadIrCameraHandle, "GetNpadIrCameraHandle"},
+        {312, &IRS::RunPointingProcessor, "RunPointingProcessor"},
+        {313, &IRS::SuspendImageProcessor, "SuspendImageProcessor"},
+        {314, &IRS::CheckFirmwareVersion, "CheckFirmwareVersion"},
+        {315, &IRS::SetFunctionLevel, "SetFunctionLevel"},
+        {316, &IRS::RunImageTransferExProcessor, "RunImageTransferExProcessor"},
+        {317, &IRS::RunIrLedProcessor, "RunIrLedProcessor"},
+        {318, &IRS::StopImageProcessorAsync, "StopImageProcessorAsync"},
+        {319, &IRS::ActivateIrsensorWithFunctionLevel, "ActivateIrsensorWithFunctionLevel"},
     };
     // clang-format on
 
     RegisterHandlers(functions);
+
+    auto& kernel = Core::System::GetInstance().Kernel();
+    shared_mem = Kernel::SharedMemory::Create(
+        kernel, nullptr, 0x8000, Kernel::MemoryPermission::ReadWrite,
+        Kernel::MemoryPermission::Read, 0, Kernel::MemoryRegion::BASE, "IRS:SharedMemory");
+}
+
+void IRS::ActivateIrsensor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::DeactivateIrsensor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::GetIrsensorSharedMemoryHandle(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2, 1};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushCopyObjects(shared_mem);
+    LOG_DEBUG(Service_IRS, "called");
+}
+
+void IRS::StopImageProcessor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::RunMomentProcessor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::RunClusteringProcessor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::RunImageTransferProcessor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::GetImageTransferProcessorState(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 5};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushRaw<u64>(CoreTiming::GetTicks());
+    rb.PushRaw<u32>(0);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::RunTeraPluginProcessor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::GetNpadIrCameraHandle(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushRaw<u32>(device_handle);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::RunPointingProcessor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::SuspendImageProcessor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::CheckFirmwareVersion(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::SetFunctionLevel(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::RunImageTransferExProcessor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::RunIrLedProcessor(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::StopImageProcessorAsync(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
+}
+
+void IRS::ActivateIrsensorWithFunctionLevel(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_IRS, "(STUBBED) called");
 }
 
 IRS::~IRS() = default;

src/core/hle/service/hid/irs.h

@@ -4,14 +4,41 @@
 
 #pragma once
 
+#include "core/hle/kernel/object.h"
 #include "core/hle/service/service.h"
 
+namespace Kernel {
+class SharedMemory;
+}
+
 namespace Service::HID {
 
 class IRS final : public ServiceFramework<IRS> {
 public:
     explicit IRS();
     ~IRS() override;
+
+private:
+    void ActivateIrsensor(Kernel::HLERequestContext& ctx);
+    void DeactivateIrsensor(Kernel::HLERequestContext& ctx);
+    void GetIrsensorSharedMemoryHandle(Kernel::HLERequestContext& ctx);
+    void StopImageProcessor(Kernel::HLERequestContext& ctx);
+    void RunMomentProcessor(Kernel::HLERequestContext& ctx);
+    void RunClusteringProcessor(Kernel::HLERequestContext& ctx);
+    void RunImageTransferProcessor(Kernel::HLERequestContext& ctx);
+    void GetImageTransferProcessorState(Kernel::HLERequestContext& ctx);
+    void RunTeraPluginProcessor(Kernel::HLERequestContext& ctx);
+    void GetNpadIrCameraHandle(Kernel::HLERequestContext& ctx);
+    void RunPointingProcessor(Kernel::HLERequestContext& ctx);
+    void SuspendImageProcessor(Kernel::HLERequestContext& ctx);
+    void CheckFirmwareVersion(Kernel::HLERequestContext& ctx);
+    void SetFunctionLevel(Kernel::HLERequestContext& ctx);
+    void RunImageTransferExProcessor(Kernel::HLERequestContext& ctx);
+    void RunIrLedProcessor(Kernel::HLERequestContext& ctx);
+    void StopImageProcessorAsync(Kernel::HLERequestContext& ctx);
+    void ActivateIrsensorWithFunctionLevel(Kernel::HLERequestContext& ctx);
+    Kernel::SharedPtr<Kernel::SharedMemory> shared_mem;
+    const u32 device_handle{0xABCD};
 };
 
 class IRS_SYS final : public ServiceFramework<IRS_SYS> {

src/core/hle/service/nfp/nfp.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/event.h"
 #include "core/hle/service/hid/hid.h"

src/core/hle/service/nim/nim.cpp

@@ -4,6 +4,7 @@
 
 #include <chrono>
 #include <ctime>
+#include "core/core.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/event.h"
 #include "core/hle/service/nim/nim.h"

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

@@ -2,8 +2,11 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/server_session.h"
 #include "core/hle/kernel/session.h"
 #include "core/hle/service/sm/controller.h"
 

src/core/hle/service/ssl/ssl.cpp

@@ -103,6 +103,7 @@ public:
     }
 
 private:
+    u32 ssl_version{};
     void CreateContext(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_SSL, "(STUBBED) called");
 
@@ -112,10 +113,9 @@ private:
     }
 
     void SetInterfaceVersion(Kernel::HLERequestContext& ctx) {
-        LOG_WARNING(Service_SSL, "(STUBBED) called");
+        LOG_DEBUG(Service_SSL, "called");
         IPC::RequestParser rp{ctx};
-        u32 unk1 = rp.Pop<u32>(); // Probably minor/major?
-        u32 unk2 = rp.Pop<u32>(); // TODO(ogniK): Figure out what this does
+        ssl_version = rp.Pop<u32>();
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);

src/video_core/engines/fermi_2d.cpp

@@ -4,11 +4,13 @@
 
 #include "core/memory.h"
 #include "video_core/engines/fermi_2d.h"
+#include "video_core/rasterizer_interface.h"
 #include "video_core/textures/decoders.h"
 
 namespace Tegra::Engines {
 
-Fermi2D::Fermi2D(MemoryManager& memory_manager) : memory_manager(memory_manager) {}
+Fermi2D::Fermi2D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager)
+    : memory_manager(memory_manager), rasterizer{rasterizer} {}
 
 void Fermi2D::WriteReg(u32 method, u32 value) {
     ASSERT_MSG(method < Regs::NUM_REGS,
@@ -52,6 +54,12 @@ void Fermi2D::HandleSurfaceCopy() {
         return;
     }
 
+    rasterizer.FlushRegion(source_cpu, src_bytes_per_pixel * regs.src.width * regs.src.height);
+    // We have to invalidate the destination region to evict any outdated surfaces from the cache.
+    // We do this before actually writing the new data because the destination address might contain
+    // a dirty surface that will have to be written back to memory.
+    rasterizer.InvalidateRegion(dest_cpu, dst_bytes_per_pixel * regs.dst.width * regs.dst.height);
+
     u8* src_buffer = Memory::GetPointer(source_cpu);
     u8* dst_buffer = Memory::GetPointer(dest_cpu);
 

src/video_core/engines/fermi_2d.h

@@ -12,6 +12,10 @@
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 
+namespace VideoCore {
+class RasterizerInterface;
+}
+
 namespace Tegra::Engines {
 
 #define FERMI2D_REG_INDEX(field_name)                                                              \
@@ -19,7 +23,7 @@ namespace Tegra::Engines {
 
 class Fermi2D final {
 public:
-    explicit Fermi2D(MemoryManager& memory_manager);
+    explicit Fermi2D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager);
     ~Fermi2D() = default;
 
     /// Write the value to the register identified by method.
@@ -94,6 +98,8 @@ public:
     MemoryManager& memory_manager;
 
 private:
+    VideoCore::RasterizerInterface& rasterizer;
+
     /// Performs the copy from the source surface to the destination surface as configured in the
     /// registers.
     void HandleSurfaceCopy();

src/video_core/engines/kepler_memory.cpp

@@ -5,10 +5,14 @@
 #include "common/logging/log.h"
 #include "core/memory.h"
 #include "video_core/engines/kepler_memory.h"
+#include "video_core/rasterizer_interface.h"
 
 namespace Tegra::Engines {
 
-KeplerMemory::KeplerMemory(MemoryManager& memory_manager) : memory_manager(memory_manager) {}
+KeplerMemory::KeplerMemory(VideoCore::RasterizerInterface& rasterizer,
+                           MemoryManager& memory_manager)
+    : memory_manager(memory_manager), rasterizer{rasterizer} {}
+
 KeplerMemory::~KeplerMemory() = default;
 
 void KeplerMemory::WriteReg(u32 method, u32 value) {
@@ -37,6 +41,11 @@ void KeplerMemory::ProcessData(u32 data) {
     VAddr dest_address =
         *memory_manager.GpuToCpuAddress(address + state.write_offset * sizeof(u32));
 
+    // We have to invalidate the destination region to evict any outdated surfaces from the cache.
+    // We do this before actually writing the new data because the destination address might contain
+    // a dirty surface that will have to be written back to memory.
+    rasterizer.InvalidateRegion(dest_address, sizeof(u32));
+
     Memory::Write32(dest_address, data);
 
     state.write_offset++;

src/video_core/engines/kepler_memory.h

@@ -11,6 +11,10 @@
 #include "common/common_types.h"
 #include "video_core/memory_manager.h"
 
+namespace VideoCore {
+class RasterizerInterface;
+}
+
 namespace Tegra::Engines {
 
 #define KEPLERMEMORY_REG_INDEX(field_name)                                                         \
@@ -18,7 +22,7 @@ namespace Tegra::Engines {
 
 class KeplerMemory final {
 public:
-    KeplerMemory(MemoryManager& memory_manager);
+    KeplerMemory(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager);
     ~KeplerMemory();
 
     /// Write the value to the register identified by method.
@@ -72,6 +76,7 @@ public:
 
 private:
     MemoryManager& memory_manager;
+    VideoCore::RasterizerInterface& rasterizer;
 
     void ProcessData(u32 data);
 };

src/video_core/engines/maxwell_3d.h

@@ -461,7 +461,11 @@ public:
                     u32 entry;
                 } macros;
 
-                INSERT_PADDING_WORDS(0x1B8);
+                INSERT_PADDING_WORDS(0x189);
+
+                u32 tfb_enabled;
+
+                INSERT_PADDING_WORDS(0x2E);
 
                 RenderTargetConfig rt[NumRenderTargets];
 
@@ -594,7 +598,9 @@ public:
 
                 u32 depth_write_enabled;
 
-                INSERT_PADDING_WORDS(0x7);
+                u32 alpha_test_enabled;
+
+                INSERT_PADDING_WORDS(0x6);
 
                 u32 d3d_cull_mode;
 
@@ -977,6 +983,7 @@ private:
                   "Field " #field_name " has invalid position")
 
 ASSERT_REG_POSITION(macros, 0x45);
+ASSERT_REG_POSITION(tfb_enabled, 0x1D1);
 ASSERT_REG_POSITION(rt, 0x200);
 ASSERT_REG_POSITION(viewport_transform[0], 0x280);
 ASSERT_REG_POSITION(viewport, 0x300);
@@ -996,6 +1003,7 @@ ASSERT_REG_POSITION(zeta_height, 0x48b);
 ASSERT_REG_POSITION(depth_test_enable, 0x4B3);
 ASSERT_REG_POSITION(independent_blend_enable, 0x4B9);
 ASSERT_REG_POSITION(depth_write_enabled, 0x4BA);
+ASSERT_REG_POSITION(alpha_test_enabled, 0x4BB);
 ASSERT_REG_POSITION(d3d_cull_mode, 0x4C2);
 ASSERT_REG_POSITION(depth_test_func, 0x4C3);
 ASSERT_REG_POSITION(blend, 0x4CF);

src/video_core/engines/maxwell_compute.cpp

@@ -2,12 +2,29 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/logging/log.h"
+#include "core/core.h"
 #include "video_core/engines/maxwell_compute.h"
 
 namespace Tegra {
 namespace Engines {
 
-void MaxwellCompute::WriteReg(u32 method, u32 value) {}
+void MaxwellCompute::WriteReg(u32 method, u32 value) {
+    ASSERT_MSG(method < Regs::NUM_REGS,
+               "Invalid MaxwellCompute register, increase the size of the Regs structure");
+
+    regs.reg_array[method] = value;
+
+    switch (method) {
+    case MAXWELL_COMPUTE_REG_INDEX(compute): {
+        LOG_CRITICAL(HW_GPU, "Compute shaders are not implemented");
+        UNREACHABLE();
+        break;
+    }
+    default:
+        break;
+    }
+}
 
 } // namespace Engines
 } // namespace Tegra

src/video_core/engines/maxwell_compute.h

@@ -4,17 +4,53 @@
 
 #pragma once
 
+#include <array>
+#include "common/assert.h"
+#include "common/bit_field.h"
+#include "common/common_funcs.h"
 #include "common/common_types.h"
 
 namespace Tegra::Engines {
 
+#define MAXWELL_COMPUTE_REG_INDEX(field_name)                                                      \
+    (offsetof(Tegra::Engines::MaxwellCompute::Regs, field_name) / sizeof(u32))
+
 class MaxwellCompute final {
 public:
     MaxwellCompute() = default;
     ~MaxwellCompute() = default;
 
+    struct Regs {
+        static constexpr std::size_t NUM_REGS = 0xCF8;
+
+        union {
+            struct {
+                INSERT_PADDING_WORDS(0x281);
+
+                union {
+                    u32 compute_end;
+                    BitField<0, 1, u32> unknown;
+                } compute;
+
+                INSERT_PADDING_WORDS(0xA76);
+            };
+            std::array<u32, NUM_REGS> reg_array;
+        };
+    } regs{};
+
+    static_assert(sizeof(Regs) == Regs::NUM_REGS * sizeof(u32),
+                  "MaxwellCompute Regs has wrong size");
+
     /// Write the value to the register identified by method.
     void WriteReg(u32 method, u32 value);
 };
 
+#define ASSERT_REG_POSITION(field_name, position)                                                  \
+    static_assert(offsetof(MaxwellCompute::Regs, field_name) == position * 4,                      \
+                  "Field " #field_name " has invalid position")
+
+ASSERT_REG_POSITION(compute, 0x281);
+
+#undef ASSERT_REG_POSITION
+
 } // namespace Tegra::Engines

src/video_core/engines/maxwell_dma.cpp

@@ -4,12 +4,14 @@
 
 #include "core/memory.h"
 #include "video_core/engines/maxwell_dma.h"
+#include "video_core/rasterizer_interface.h"
 #include "video_core/textures/decoders.h"
 
 namespace Tegra {
 namespace Engines {
 
-MaxwellDMA::MaxwellDMA(MemoryManager& memory_manager) : memory_manager(memory_manager) {}
+MaxwellDMA::MaxwellDMA(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager)
+    : memory_manager(memory_manager), rasterizer{rasterizer} {}
 
 void MaxwellDMA::WriteReg(u32 method, u32 value) {
     ASSERT_MSG(method < Regs::NUM_REGS,
@@ -44,36 +46,79 @@ void MaxwellDMA::HandleCopy() {
     ASSERT(regs.exec.query_mode == Regs::QueryMode::None);
     ASSERT(regs.exec.query_intr == Regs::QueryIntr::None);
     ASSERT(regs.exec.copy_mode == Regs::CopyMode::Unk2);
-    ASSERT(regs.src_params.pos_x == 0);
-    ASSERT(regs.src_params.pos_y == 0);
     ASSERT(regs.dst_params.pos_x == 0);
     ASSERT(regs.dst_params.pos_y == 0);
 
-    if (regs.exec.is_dst_linear == regs.exec.is_src_linear) {
-        std::size_t copy_size = regs.x_count;
+    if (!regs.exec.is_dst_linear && !regs.exec.is_src_linear) {
+        // If both the source and the destination are in block layout, assert.
+        UNREACHABLE_MSG("Tiled->Tiled DMA transfers are not yet implemented");
+        return;
+    }
 
+    if (regs.exec.is_dst_linear && regs.exec.is_src_linear) {
         // When the enable_2d bit is disabled, the copy is performed as if we were copying a 1D
-        // buffer of length `x_count`, otherwise we copy a 2D buffer of size (x_count, y_count).
-        if (regs.exec.enable_2d) {
-            copy_size = copy_size * regs.y_count;
+        // buffer of length `x_count`, otherwise we copy a 2D image of dimensions (x_count,
+        // y_count).
+        if (!regs.exec.enable_2d) {
+            Memory::CopyBlock(dest_cpu, source_cpu, regs.x_count);
+            return;
         }
 
-        Memory::CopyBlock(dest_cpu, source_cpu, copy_size);
+        // If both the source and the destination are in linear layout, perform a line-by-line
+        // copy. We're going to take a subrect of size (x_count, y_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.y_count; ++line) {
+            const VAddr source_line = source_cpu + line * regs.src_pitch;
+            const VAddr dest_line = dest_cpu + line * regs.dst_pitch;
+            Memory::CopyBlock(dest_line, source_line, regs.x_count);
+        }
         return;
     }
 
     ASSERT(regs.exec.enable_2d == 1);
+
+    size_t copy_size = regs.x_count * regs.y_count;
+
+    const auto FlushAndInvalidate = [&](u32 src_size, u32 dst_size) {
+        // TODO(Subv): For now, manually flush the regions until we implement GPU-accelerated
+        // copying.
+        rasterizer.FlushRegion(source_cpu, src_size);
+
+        // We have to invalidate the destination region to evict any outdated surfaces from the
+        // cache. We do this before actually writing the new data because the destination address
+        // might contain a dirty surface that will have to be written back to memory.
+        rasterizer.InvalidateRegion(dest_cpu, dst_size);
+    };
+
     u8* src_buffer = Memory::GetPointer(source_cpu);
     u8* dst_buffer = Memory::GetPointer(dest_cpu);
 
     if (regs.exec.is_dst_linear && !regs.exec.is_src_linear) {
+        ASSERT(regs.src_params.size_z == 1);
         // If the input is tiled and the output is linear, deswizzle the input and copy it over.
-        Texture::CopySwizzledData(regs.src_params.size_x, regs.src_params.size_y, 1, 1, src_buffer,
-                                  dst_buffer, true, regs.src_params.BlockHeight());
+
+        u32 src_bytes_per_pixel = regs.src_pitch / regs.src_params.size_x;
+
+        FlushAndInvalidate(regs.src_pitch * regs.src_params.size_y,
+                           copy_size * src_bytes_per_pixel);
+
+        Texture::UnswizzleSubrect(regs.x_count, regs.y_count, regs.dst_pitch,
+                                  regs.src_params.size_x, src_bytes_per_pixel, source_cpu, dest_cpu,
+                                  regs.src_params.BlockHeight(), regs.src_params.pos_x,
+                                  regs.src_params.pos_y);
     } else {
+        ASSERT(regs.dst_params.size_z == 1);
+        ASSERT(regs.src_pitch == regs.x_count);
+
+        u32 src_bpp = regs.src_pitch / regs.x_count;
+
+        FlushAndInvalidate(regs.src_pitch * regs.y_count,
+                           regs.dst_params.size_x * regs.dst_params.size_y * src_bpp);
+
         // If the input is linear and the output is tiled, swizzle the input and copy it over.
-        Texture::CopySwizzledData(regs.dst_params.size_x, regs.dst_params.size_y, 1, 1, dst_buffer,
-                                  src_buffer, false, regs.dst_params.BlockHeight());
+        Texture::SwizzleSubrect(regs.x_count, regs.y_count, regs.src_pitch, regs.dst_params.size_x,
+                                src_bpp, dest_cpu, source_cpu, regs.dst_params.BlockHeight());
     }
 }
 

src/video_core/engines/maxwell_dma.h

@@ -12,11 +12,15 @@
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 
+namespace VideoCore {
+class RasterizerInterface;
+}
+
 namespace Tegra::Engines {
 
 class MaxwellDMA final {
 public:
-    explicit MaxwellDMA(MemoryManager& memory_manager);
+    explicit MaxwellDMA(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager);
     ~MaxwellDMA() = default;
 
     /// Write the value to the register identified by method.
@@ -129,6 +133,8 @@ public:
     MemoryManager& memory_manager;
 
 private:
+    VideoCore::RasterizerInterface& rasterizer;
+
     /// Performs the copy from the source buffer to the destination buffer as configured in the
     /// registers.
     void HandleCopy();

src/video_core/gpu.cpp

@@ -25,10 +25,10 @@ u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
 GPU::GPU(VideoCore::RasterizerInterface& rasterizer) {
     memory_manager = std::make_unique<Tegra::MemoryManager>();
     maxwell_3d = std::make_unique<Engines::Maxwell3D>(rasterizer, *memory_manager);
-    fermi_2d = std::make_unique<Engines::Fermi2D>(*memory_manager);
+    fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer, *memory_manager);
     maxwell_compute = std::make_unique<Engines::MaxwellCompute>();
-    maxwell_dma = std::make_unique<Engines::MaxwellDMA>(*memory_manager);
-    kepler_memory = std::make_unique<Engines::KeplerMemory>(*memory_manager);
+    maxwell_dma = std::make_unique<Engines::MaxwellDMA>(rasterizer, *memory_manager);
+    kepler_memory = std::make_unique<Engines::KeplerMemory>(rasterizer, *memory_manager);
 }
 
 GPU::~GPU() = default;

src/video_core/rasterizer_cache.h

@@ -17,6 +17,22 @@
 template <class T>
 class RasterizerCache : NonCopyable {
 public:
+    /// Write any cached resources overlapping the region back to memory (if dirty)
+    void FlushRegion(Tegra::GPUVAddr addr, size_t size) {
+        if (size == 0)
+            return;
+
+        const ObjectInterval interval{addr, addr + size};
+        for (auto& pair : boost::make_iterator_range(object_cache.equal_range(interval))) {
+            for (auto& cached_object : pair.second) {
+                if (!cached_object)
+                    continue;
+
+                cached_object->Flush();
+            }
+        }
+    }
+
     /// Mark the specified region as being invalidated
     void InvalidateRegion(VAddr addr, u64 size) {
         if (size == 0)
@@ -71,6 +87,7 @@ protected:
     void Unregister(const T& object) {
         auto& rasterizer = Core::System::GetInstance().Renderer().Rasterizer();
         rasterizer.UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), -1);
+        object->Flush();
         object_cache.subtract({GetInterval(object), ObjectSet{object}});
     }
 

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -23,6 +23,9 @@ struct CachedBufferEntry final {
         return size;
     }
 
+    // We do not have to flush this cache as things in it are never modified by us.
+    void Flush() {}
+
     VAddr addr;
     std::size_t size;
     GLintptr offset;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -322,6 +322,13 @@ void RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb, bool using_dep
             // Used when just a single color attachment is enabled, e.g. for clearing a color buffer
             Surface color_surface =
                 res_cache.GetColorBufferSurface(*single_color_target, preserve_contents);
+
+            if (color_surface) {
+                // Assume that a surface will be written to if it is used as a framebuffer, even if
+                // the shader doesn't actually write to it.
+                color_surface->MarkAsDirty();
+            }
+
             glFramebufferTexture2D(
                 GL_DRAW_FRAMEBUFFER,
                 GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(*single_color_target), GL_TEXTURE_2D,
@@ -332,6 +339,11 @@ void RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb, bool using_dep
             std::array<GLenum, Maxwell::NumRenderTargets> buffers;
             for (std::size_t index = 0; index < Maxwell::NumRenderTargets; ++index) {
                 Surface color_surface = res_cache.GetColorBufferSurface(index, preserve_contents);
+                if (color_surface) {
+                    // Assume that a surface will be written to if it is used as a framebuffer, even
+                    // if the shader doesn't actually write to it.
+                    color_surface->MarkAsDirty();
+                }
                 buffers[index] = GL_COLOR_ATTACHMENT0 + regs.rt_control.GetMap(index);
                 glFramebufferTexture2D(
                     GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + static_cast<GLenum>(index),
@@ -351,6 +363,10 @@ void RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb, bool using_dep
     }
 
     if (depth_surface) {
+        // Assume that a surface will be written to if it is used as a framebuffer, even if
+        // the shader doesn't actually write to it.
+        depth_surface->MarkAsDirty();
+
         if (regs.stencil_enable) {
             // Attach both depth and stencil
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
@@ -450,6 +466,8 @@ void RasterizerOpenGL::DrawArrays() {
     SyncBlendState();
     SyncLogicOpState();
     SyncCullMode();
+    SyncAlphaTest();
+    SyncTransformFeedback();
 
     // TODO(bunnei): Sync framebuffer_scale uniform here
     // TODO(bunnei): Sync scissorbox uniform(s) here
@@ -538,9 +556,19 @@ void RasterizerOpenGL::DrawArrays() {
     state.Apply();
 }
 
-void RasterizerOpenGL::FlushAll() {}
+void RasterizerOpenGL::FlushAll() {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    res_cache.FlushRegion(0, Kernel::VMManager::MAX_ADDRESS);
+    shader_cache.FlushRegion(0, Kernel::VMManager::MAX_ADDRESS);
+    buffer_cache.FlushRegion(0, Kernel::VMManager::MAX_ADDRESS);
+}
 
-void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {}
+void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    res_cache.FlushRegion(addr, size);
+    shader_cache.FlushRegion(addr, size);
+    buffer_cache.FlushRegion(addr, size);
+}
 
 void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
@@ -550,6 +578,7 @@ void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
 }
 
 void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+    FlushRegion(addr, size);
     InvalidateRegion(addr, size);
 }
 
@@ -883,4 +912,24 @@ void RasterizerOpenGL::SyncLogicOpState() {
     state.logic_op.operation = MaxwellToGL::LogicOp(regs.logic_op.operation);
 }
 
+void RasterizerOpenGL::SyncAlphaTest() {
+    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+
+    // TODO(Rodrigo): Alpha testing is a legacy OpenGL feature, but it can be
+    // implemented with a test+discard in fragment shaders.
+    if (regs.alpha_test_enabled != 0) {
+        LOG_CRITICAL(Render_OpenGL, "Alpha testing is not implemented");
+        UNREACHABLE();
+    }
+}
+
+void RasterizerOpenGL::SyncTransformFeedback() {
+    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+
+    if (regs.tfb_enabled != 0) {
+        LOG_CRITICAL(Render_OpenGL, "Transform feedbacks are not implemented");
+        UNREACHABLE();
+    }
+}
+
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -158,6 +158,12 @@ private:
     /// Syncs the LogicOp state to match the guest state
     void SyncLogicOpState();
 
+    /// Syncs the alpha test state to match the guest state
+    void SyncAlphaTest();
+
+    /// Syncs the transform feedback state to match the guest state
+    void SyncTransformFeedback();
+
     bool has_ARB_direct_state_access = false;
     bool has_ARB_multi_bind = false;
     bool has_ARB_separate_shader_objects = false;

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -141,8 +141,8 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
     {GL_COMPRESSED_RGBA_BPTC_UNORM_ARB, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
      true}, // BC7U
     {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT_ARB, GL_RGB, GL_UNSIGNED_INT_8_8_8_8,
-     ComponentType::UNorm, true}, // BC6H_UF16
-    {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT_ARB, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
+     ComponentType::Float, true}, // BC6H_UF16
+    {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT_ARB, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::Float,
      true},                                                                    // BC6H_SF16
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false},        // ASTC_2D_4X4
     {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, ComponentType::UNorm, false},            // G8R8U
@@ -265,20 +265,22 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, u8* gl_buffer, std::si
     constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
     constexpr u32 gl_bytes_per_pixel = CachedSurface::GetGLBytesPerPixel(format);
 
+    // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
+    // pixel values.
+    const u32 tile_size{IsFormatBCn(format) ? 4U : 1U};
+
     if (morton_to_gl) {
-        // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
-        // pixel values.
-        const u32 tile_size{IsFormatBCn(format) ? 4U : 1U};
         const std::vector<u8> data = Tegra::Texture::UnswizzleTexture(
             addr, tile_size, bytes_per_pixel, stride, height, block_height);
         const std::size_t size_to_copy{std::min(gl_buffer_size, data.size())};
         memcpy(gl_buffer, data.data(), size_to_copy);
     } else {
-        // TODO(bunnei): Assumes the default rendering GOB size of 16 (128 lines). We should
-        // check the configuration for this and perform more generic un/swizzle
-        LOG_WARNING(Render_OpenGL, "need to use correct swizzle/GOB parameters!");
-        VideoCore::MortonCopyPixels128(stride, height, bytes_per_pixel, gl_bytes_per_pixel,
-                                       Memory::GetPointer(addr), gl_buffer, morton_to_gl);
+        std::vector<u8> data(height * stride * bytes_per_pixel);
+        Tegra::Texture::CopySwizzledData(stride / tile_size, height / tile_size, bytes_per_pixel,
+                                         bytes_per_pixel, data.data(), gl_buffer, false,
+                                         block_height);
+        const std::size_t size_to_copy{std::min(gl_buffer_size, data.size())};
+        memcpy(Memory::GetPointer(addr), data.data(), size_to_copy);
     }
 }
 
@@ -357,17 +359,16 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, std::size_t, VAddr),
         MortonCopy<false, PixelFormat::RGBA16UI>,
         MortonCopy<false, PixelFormat::R11FG11FB10F>,
         MortonCopy<false, PixelFormat::RGBA32UI>,
-        // TODO(Subv): Swizzling DXT1/DXT23/DXT45/DXN1/DXN2/BC7U/BC6H_UF16/BC6H_SF16/ASTC_2D_4X4
-        // formats are not supported
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
+        MortonCopy<false, PixelFormat::DXT1>,
+        MortonCopy<false, PixelFormat::DXT23>,
+        MortonCopy<false, PixelFormat::DXT45>,
+        MortonCopy<false, PixelFormat::DXN1>,
+        MortonCopy<false, PixelFormat::DXN2UNORM>,
+        MortonCopy<false, PixelFormat::DXN2SNORM>,
+        MortonCopy<false, PixelFormat::BC7U>,
+        MortonCopy<false, PixelFormat::BC6H_UF16>,
+        MortonCopy<false, PixelFormat::BC6H_SF16>,
+        // TODO(Subv): Swizzling ASTC formats are not supported
         nullptr,
         MortonCopy<false, PixelFormat::G8R8U>,
         MortonCopy<false, PixelFormat::G8R8S>,
@@ -501,9 +502,12 @@ CachedSurface::CachedSurface(const SurfaceParams& params)
     glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MIN_FILTER, GL_LINEAR);
     glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
     glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+
+    VideoCore::LabelGLObject(GL_TEXTURE, texture.handle, params.addr,
+                             SurfaceParams::SurfaceTargetName(params.target));
 }
 
-static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height) {
+static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height, bool reverse) {
     union S8Z24 {
         BitField<0, 24, u32> z24;
         BitField<24, 8, u32> s8;
@@ -516,16 +520,23 @@ static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height) {
     };
     static_assert(sizeof(Z24S8) == 4, "Z24S8 is incorrect size");
 
-    S8Z24 input_pixel{};
-    Z24S8 output_pixel{};
+    S8Z24 s8z24_pixel{};
+    Z24S8 z24s8_pixel{};
     constexpr auto bpp{CachedSurface::GetGLBytesPerPixel(PixelFormat::S8Z24)};
     for (std::size_t y = 0; y < height; ++y) {
         for (std::size_t x = 0; x < width; ++x) {
             const std::size_t offset{bpp * (y * width + x)};
-            std::memcpy(&input_pixel, &data[offset], sizeof(S8Z24));
-            output_pixel.s8.Assign(input_pixel.s8);
-            output_pixel.z24.Assign(input_pixel.z24);
-            std::memcpy(&data[offset], &output_pixel, sizeof(Z24S8));
+            if (reverse) {
+                std::memcpy(&z24s8_pixel, &data[offset], sizeof(Z24S8));
+                s8z24_pixel.s8.Assign(z24s8_pixel.s8);
+                s8z24_pixel.z24.Assign(z24s8_pixel.z24);
+                std::memcpy(&data[offset], &s8z24_pixel, sizeof(S8Z24));
+            } else {
+                std::memcpy(&s8z24_pixel, &data[offset], sizeof(S8Z24));
+                z24s8_pixel.s8.Assign(s8z24_pixel.s8);
+                z24s8_pixel.z24.Assign(s8z24_pixel.z24);
+                std::memcpy(&data[offset], &z24s8_pixel, sizeof(Z24S8));
+            }
         }
     }
 }
@@ -561,7 +572,7 @@ static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelForma
     }
     case PixelFormat::S8Z24:
         // Convert the S8Z24 depth format to Z24S8, as OpenGL does not support S8Z24.
-        ConvertS8Z24ToZ24S8(data, width, height);
+        ConvertS8Z24ToZ24S8(data, width, height, false);
         break;
 
     case PixelFormat::G8R8U:
@@ -572,6 +583,30 @@ static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelForma
     }
 }
 
+/**
+ * Helper function to perform software conversion (as needed) when flushing a buffer from OpenGL to
+ * Switch memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or
+ * with typical desktop GPUs.
+ */
+static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& data, PixelFormat pixel_format,
+                                                u32 width, u32 height) {
+    switch (pixel_format) {
+    case PixelFormat::G8R8U:
+    case PixelFormat::G8R8S:
+    case PixelFormat::ASTC_2D_4X4:
+    case PixelFormat::ASTC_2D_8X8: {
+        LOG_CRITICAL(HW_GPU, "Conversion of format {} after texture flushing is not implemented",
+                     static_cast<u32>(pixel_format));
+        UNREACHABLE();
+        break;
+    }
+    case PixelFormat::S8Z24:
+        // Convert the Z24S8 depth format to S8Z24, as OpenGL does not support S8Z24.
+        ConvertS8Z24ToZ24S8(data, width, height, true);
+        break;
+    }
+}
+
 MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 64, 192));
 void CachedSurface::LoadGLBuffer() {
     ASSERT(params.type != SurfaceType::Fill);
@@ -609,11 +644,70 @@ void CachedSurface::LoadGLBuffer() {
     }
 
     ConvertFormatAsNeeded_LoadGLBuffer(gl_buffer, params.pixel_format, params.width, params.height);
+
+    dirty = false;
 }
 
 MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64));
 void CachedSurface::FlushGLBuffer() {
-    ASSERT_MSG(false, "Unimplemented");
+    MICROPROFILE_SCOPE(OpenGL_SurfaceFlush);
+
+    const auto& rect{params.GetRect()};
+
+    // Load data from memory to the surface
+    const GLint x0 = static_cast<GLint>(rect.left);
+    const GLint y0 = static_cast<GLint>(rect.bottom);
+    const size_t buffer_offset =
+        static_cast<size_t>(static_cast<size_t>(y0) * params.width + static_cast<size_t>(x0)) *
+        GetGLBytesPerPixel(params.pixel_format);
+
+    const u32 bytes_per_pixel = GetGLBytesPerPixel(params.pixel_format);
+    const u32 copy_size = params.width * params.height * bytes_per_pixel;
+    gl_buffer.resize(static_cast<size_t>(params.depth) * copy_size);
+
+    const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
+
+    // Ensure no bad interactions with GL_UNPACK_ALIGNMENT
+    ASSERT(params.width * GetGLBytesPerPixel(params.pixel_format) % 4 == 0);
+    glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(params.width));
+
+    ASSERT(!tuple.compressed);
+    ASSERT(x0 == 0 && y0 == 0);
+
+    glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
+    glGetTextureImage(texture.handle, 0, tuple.format, tuple.type, gl_buffer.size(),
+                      gl_buffer.data());
+    glPixelStorei(GL_PACK_ROW_LENGTH, 0);
+
+    ConvertFormatAsNeeded_FlushGLBuffer(gl_buffer, params.pixel_format, params.width,
+                                        params.height);
+
+    ASSERT(params.type != SurfaceType::Fill);
+
+    const u8* const texture_src_data = Memory::GetPointer(params.addr);
+
+    ASSERT(texture_src_data);
+
+    if (params.is_tiled) {
+        // TODO(bunnei): This only swizzles and copies a 2D texture - we do not yet know how to do
+        // this for 3D textures, etc.
+        switch (params.target) {
+        case SurfaceParams::SurfaceTarget::Texture2D:
+            // Pass impl. to the fallback code below
+            break;
+        default:
+            LOG_CRITICAL(HW_GPU, "Unimplemented tiled unload for target={}",
+                         static_cast<u32>(params.target));
+            UNREACHABLE();
+        }
+
+        gl_to_morton_fns[static_cast<size_t>(params.pixel_format)](
+            params.width, params.block_height, params.height, &gl_buffer[buffer_offset], copy_size,
+            params.addr + buffer_offset);
+    } else {
+        Memory::WriteBlock(params.addr + buffer_offset, &gl_buffer[buffer_offset],
+                           gl_buffer.size() - buffer_offset);
+    }
 }
 
 MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 64, 192));

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -137,6 +137,27 @@ struct SurfaceParams {
         }
     }
 
+    static std::string SurfaceTargetName(SurfaceTarget target) {
+        switch (target) {
+        case SurfaceTarget::Texture1D:
+            return "Texture1D";
+        case SurfaceTarget::Texture2D:
+            return "Texture2D";
+        case SurfaceTarget::Texture3D:
+            return "Texture3D";
+        case SurfaceTarget::Texture1DArray:
+            return "Texture1DArray";
+        case SurfaceTarget::Texture2DArray:
+            return "Texture2DArray";
+        case SurfaceTarget::TextureCubemap:
+            return "TextureCubemap";
+        default:
+            LOG_CRITICAL(HW_GPU, "Unimplemented surface_target={}", static_cast<u32>(target));
+            UNREACHABLE();
+            return fmt::format("TextureUnknown({})", static_cast<u32>(target));
+        }
+    }
+
     /**
      * Gets the compression factor for the specified PixelFormat. This applies to just the
      * "compressed width" and "compressed height", not the overall compression factor of a
@@ -741,6 +762,19 @@ public:
         return params.size_in_bytes;
     }
 
+    void Flush() {
+        // There is no need to flush the surface if it hasn't been modified by us.
+        if (!dirty)
+            return;
+
+        FlushGLBuffer();
+        dirty = false;
+    }
+
+    void MarkAsDirty() {
+        dirty = true;
+    }
+
     const OGLTexture& Texture() const {
         return texture;
     }
@@ -772,6 +806,7 @@ private:
     std::vector<u8> gl_buffer;
     SurfaceParams params;
     GLenum gl_target;
+    bool dirty = false;
 };
 
 class RasterizerCacheOpenGL final : public RasterizerCache<Surface> {

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -8,6 +8,7 @@
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_shader_cache.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
+#include "video_core/utils.h"
 
 namespace OpenGL {
 
@@ -83,6 +84,7 @@ CachedShader::CachedShader(VAddr addr, Maxwell::ShaderProgram program_type)
     shader.Create(program_result.first.c_str(), gl_type);
     program.Create(true, shader.handle);
     SetShaderUniformBlockBindings(program.handle);
+    VideoCore::LabelGLObject(GL_PROGRAM, program.handle, addr);
 }
 
 GLuint CachedShader::GetProgramResourceIndex(const GLShader::ConstBufferEntry& buffer) {

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -32,6 +32,9 @@ public:
         return GLShader::MAX_PROGRAM_CODE_LENGTH * sizeof(u64);
     }
 
+    // We do not have to flush this cache as things in it are never modified by us.
+    void Flush() {}
+
     /// Gets the shader entries for the shader
     const GLShader::ShaderEntries& GetShaderEntries() const {
         return entries;

src/video_core/textures/decoders.cpp

@@ -88,6 +88,38 @@ void FastSwizzleData(u32 width, u32 height, u32 bytes_per_pixel, u8* swizzled_da
     }
 }
 
+void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
+                    u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
+                    u32 block_height) {
+    for (u32 line = 0; line < subrect_height; ++line) {
+        for (u32 x = 0; x < subrect_width; ++x) {
+            u32 swizzled_offset =
+                GetSwizzleOffset(x, line, swizzled_width, bytes_per_pixel, block_height);
+
+            const VAddr source_line = unswizzled_data + line * source_pitch + x * bytes_per_pixel;
+            const VAddr dest_addr = swizzled_data + swizzled_offset;
+
+            Memory::CopyBlock(dest_addr, source_line, bytes_per_pixel);
+        }
+    }
+}
+
+void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 swizzled_width,
+                      u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
+                      u32 block_height, u32 offset_x, u32 offset_y) {
+    for (u32 line = 0; line < subrect_height; ++line) {
+        for (u32 x = 0; x < subrect_width; ++x) {
+            u32 swizzled_offset = GetSwizzleOffset(offset_x, line + offset_y, swizzled_width,
+                                                   bytes_per_pixel, block_height);
+
+            const VAddr dest_line = unswizzled_data + line * dest_pitch + x;
+            const VAddr source_addr = swizzled_data + swizzled_offset;
+
+            Memory::CopyBlock(dest_line, source_addr, bytes_per_pixel);
+        }
+    }
+}
+
 u32 BytesPerPixel(TextureFormat format) {
     switch (format) {
     case TextureFormat::DXT1:

src/video_core/textures/decoders.h

@@ -26,6 +26,16 @@ std::vector<u8> UnswizzleDepthTexture(VAddr address, DepthFormat format, u32 wid
 void CopySwizzledData(u32 width, u32 height, u32 bytes_per_pixel, u32 out_bytes_per_pixel,
                       u8* swizzled_data, u8* unswizzled_data, bool unswizzle, u32 block_height);
 
+/// Copies an untiled subrectangle into a tiled surface.
+void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
+                    u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
+                    u32 block_height);
+
+/// Copies a tiled subrectangle into a linear surface.
+void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 swizzled_width,
+                      u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
+                      u32 block_height, u32 offset_x, u32 offset_y);
+
 /**
  * Decodes an unswizzled texture into a A8R8G8B8 texture.
  */

src/video_core/utils.h

@@ -161,4 +161,26 @@ static inline void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixe
     }
 }
 
+static void LabelGLObject(GLenum identifier, GLuint handle, VAddr addr,
+                          std::string extra_info = "") {
+    if (!GLAD_GL_KHR_debug) {
+        return; // We don't need to throw an error as this is just for debugging
+    }
+    const std::string nice_addr = fmt::format("0x{:016x}", addr);
+    std::string object_label;
+
+    switch (identifier) {
+    case GL_TEXTURE:
+        object_label = extra_info + "@" + nice_addr;
+        break;
+    case GL_PROGRAM:
+        object_label = "ShaderProgram@" + nice_addr;
+        break;
+    default:
+        object_label = fmt::format("Object(0x{:x})@{}", identifier, nice_addr);
+        break;
+    }
+    glObjectLabel(identifier, handle, -1, static_cast<const GLchar*>(object_label.c_str()));
+}
+
 } // namespace VideoCore

src/yuzu/game_list.cpp

@@ -318,9 +318,14 @@ void GameList::PopupContextMenu(const QPoint& menu_location) {
     int row = item_model->itemFromIndex(item)->row();
     QStandardItem* child_file = item_model->invisibleRootItem()->child(row, COLUMN_NAME);
     u64 program_id = child_file->data(GameListItemPath::ProgramIdRole).toULongLong();
+    std::string path = child_file->data(GameListItemPath::FullPathRole).toString().toStdString();
 
     QMenu context_menu;
     QAction* open_save_location = context_menu.addAction(tr("Open Save Data Location"));
+    QAction* open_lfs_location = context_menu.addAction(tr("Open Mod Data Location"));
+    context_menu.addSeparator();
+    QAction* dump_romfs = context_menu.addAction(tr("Dump RomFS"));
+    QAction* copy_tid = context_menu.addAction(tr("Copy Title ID to Clipboard"));
     QAction* navigate_to_gamedb_entry = context_menu.addAction(tr("Navigate to GameDB entry"));
 
     open_save_location->setEnabled(program_id != 0);
@@ -329,6 +334,10 @@ void GameList::PopupContextMenu(const QPoint& menu_location) {
 
     connect(open_save_location, &QAction::triggered,
             [&]() { emit OpenFolderRequested(program_id, GameListOpenTarget::SaveData); });
+    connect(open_lfs_location, &QAction::triggered,
+            [&]() { emit OpenFolderRequested(program_id, GameListOpenTarget::ModData); });
+    connect(dump_romfs, &QAction::triggered, [&]() { emit DumpRomFSRequested(program_id, path); });
+    connect(copy_tid, &QAction::triggered, [&]() { emit CopyTIDRequested(program_id); });
     connect(navigate_to_gamedb_entry, &QAction::triggered,
             [&]() { emit NavigateToGamedbEntryRequested(program_id, compatibility_list); });
 

src/yuzu/game_list.h

@@ -28,7 +28,10 @@ namespace FileSys {
 class VfsFilesystem;
 }
 
-enum class GameListOpenTarget { SaveData };
+enum class GameListOpenTarget {
+    SaveData,
+    ModData,
+};
 
 class GameList : public QWidget {
     Q_OBJECT
@@ -89,6 +92,8 @@ signals:
     void GameChosen(QString game_path);
     void ShouldCancelWorker();
     void OpenFolderRequested(u64 program_id, GameListOpenTarget target);
+    void DumpRomFSRequested(u64 program_id, const std::string& game_path);
+    void CopyTIDRequested(u64 program_id);
     void NavigateToGamedbEntryRequested(u64 program_id,
                                         const CompatibilityList& compatibility_list);
 

src/yuzu/main.cpp

@@ -7,6 +7,22 @@
 #include <memory>
 #include <thread>
 
+// VFS includes must be before glad as they will conflict with Windows file api, which uses defines.
+#include "core/file_sys/vfs.h"
+#include "core/file_sys/vfs_real.h"
+
+// These are wrappers to avoid the calls to CreateDirectory and CreateFile becuase of the Windows
+// defines.
+static FileSys::VirtualDir VfsFilesystemCreateDirectoryWrapper(
+    const FileSys::VirtualFilesystem& vfs, const std::string& path, FileSys::Mode mode) {
+    return vfs->CreateDirectory(path, mode);
+}
+
+static FileSys::VirtualFile VfsDirectoryCreateFileWrapper(const FileSys::VirtualDir& dir,
+                                                          const std::string& path) {
+    return dir->CreateFile(path);
+}
+
 #include <fmt/ostream.h>
 #include <glad/glad.h>
 
@@ -30,16 +46,18 @@
 #include "common/telemetry.h"
 #include "core/core.h"
 #include "core/crypto/key_manager.h"
+#include "core/file_sys/bis_factory.h"
 #include "core/file_sys/card_image.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/control_metadata.h"
 #include "core/file_sys/patch_manager.h"
 #include "core/file_sys/registered_cache.h"
+#include "core/file_sys/romfs.h"
 #include "core/file_sys/savedata_factory.h"
 #include "core/file_sys/submission_package.h"
-#include "core/file_sys/vfs_real.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/service/filesystem/filesystem.h"
+#include "core/hle/service/filesystem/fsp_ldr.h"
 #include "core/loader/loader.h"
 #include "core/perf_stats.h"
 #include "core/settings.h"
@@ -362,6 +380,8 @@ void GMainWindow::RestoreUIState() {
 void GMainWindow::ConnectWidgetEvents() {
     connect(game_list, &GameList::GameChosen, this, &GMainWindow::OnGameListLoadFile);
     connect(game_list, &GameList::OpenFolderRequested, this, &GMainWindow::OnGameListOpenFolder);
+    connect(game_list, &GameList::DumpRomFSRequested, this, &GMainWindow::OnGameListDumpRomFS);
+    connect(game_list, &GameList::CopyTIDRequested, this, &GMainWindow::OnGameListCopyTID);
     connect(game_list, &GameList::NavigateToGamedbEntryRequested, this,
             &GMainWindow::OnGameListNavigateToGamedbEntry);
 
@@ -713,6 +733,12 @@ void GMainWindow::OnGameListOpenFolder(u64 program_id, GameListOpenTarget target
                                                                 program_id, user_id, 0);
         break;
     }
+    case GameListOpenTarget::ModData: {
+        open_target = "Mod Data";
+        const auto load_dir = FileUtil::GetUserPath(FileUtil::UserPath::LoadDir);
+        path = fmt::format("{}{:016X}", load_dir, program_id);
+        break;
+    }
     default:
         UNIMPLEMENTED();
     }
@@ -730,6 +756,120 @@ void GMainWindow::OnGameListOpenFolder(u64 program_id, GameListOpenTarget target
     QDesktopServices::openUrl(QUrl::fromLocalFile(qpath));
 }
 
+static std::size_t CalculateRomFSEntrySize(const FileSys::VirtualDir& dir, bool full) {
+    std::size_t out = 0;
+
+    for (const auto& subdir : dir->GetSubdirectories()) {
+        out += 1 + CalculateRomFSEntrySize(subdir, full);
+    }
+
+    return out + (full ? dir->GetFiles().size() : 0);
+}
+
+static bool RomFSRawCopy(QProgressDialog& dialog, const FileSys::VirtualDir& src,
+                         const FileSys::VirtualDir& dest, std::size_t block_size, bool full) {
+    if (src == nullptr || dest == nullptr || !src->IsReadable() || !dest->IsWritable())
+        return false;
+    if (dialog.wasCanceled())
+        return false;
+
+    if (full) {
+        for (const auto& file : src->GetFiles()) {
+            const auto out = VfsDirectoryCreateFileWrapper(dest, file->GetName());
+            if (!FileSys::VfsRawCopy(file, out, block_size))
+                return false;
+            dialog.setValue(dialog.value() + 1);
+            if (dialog.wasCanceled())
+                return false;
+        }
+    }
+
+    for (const auto& dir : src->GetSubdirectories()) {
+        const auto out = dest->CreateSubdirectory(dir->GetName());
+        if (!RomFSRawCopy(dialog, dir, out, block_size, full))
+            return false;
+        dialog.setValue(dialog.value() + 1);
+        if (dialog.wasCanceled())
+            return false;
+    }
+
+    return true;
+}
+
+void GMainWindow::OnGameListDumpRomFS(u64 program_id, const std::string& game_path) {
+    const auto path = fmt::format("{}{:016X}/romfs",
+                                  FileUtil::GetUserPath(FileUtil::UserPath::DumpDir), program_id);
+
+    const auto failed = [this, &path] {
+        QMessageBox::warning(this, tr("RomFS Extraction Failed!"),
+                             tr("There was an error copying the RomFS files or the user "
+                                "cancelled the operation."));
+        vfs->DeleteDirectory(path);
+    };
+
+    const auto loader = Loader::GetLoader(vfs->OpenFile(game_path, FileSys::Mode::Read));
+    if (loader == nullptr) {
+        failed();
+        return;
+    }
+
+    FileSys::VirtualFile file;
+    if (loader->ReadRomFS(file) != Loader::ResultStatus::Success) {
+        failed();
+        return;
+    }
+
+    const auto romfs =
+        loader->IsRomFSUpdatable()
+            ? FileSys::PatchManager(program_id).PatchRomFS(file, loader->ReadRomFSIVFCOffset())
+            : file;
+
+    const auto extracted = FileSys::ExtractRomFS(romfs, FileSys::RomFSExtractionType::Full);
+    if (extracted == nullptr) {
+        failed();
+        return;
+    }
+
+    const auto out = VfsFilesystemCreateDirectoryWrapper(vfs, path, FileSys::Mode::ReadWrite);
+
+    if (out == nullptr) {
+        failed();
+        return;
+    }
+
+    bool ok;
+    const auto res = QInputDialog::getItem(
+        this, tr("Select RomFS Dump Mode"),
+        tr("Please select the how you would like the RomFS dumped.<br>Full will copy all of the "
+           "files into the new directory while <br>skeleton will only create the directory "
+           "structure."),
+        {"Full", "Skeleton"}, 0, false, &ok);
+    if (!ok)
+        failed();
+
+    const auto full = res == "Full";
+    const auto entry_size = CalculateRomFSEntrySize(extracted, full);
+
+    QProgressDialog progress(tr("Extracting RomFS..."), tr("Cancel"), 0, entry_size, this);
+    progress.setWindowModality(Qt::WindowModal);
+    progress.setMinimumDuration(100);
+
+    if (RomFSRawCopy(progress, extracted, out, 0x400000, full)) {
+        progress.close();
+        QMessageBox::information(this, tr("RomFS Extraction Succeeded!"),
+                                 tr("The operation completed successfully."));
+        QDesktopServices::openUrl(QUrl::fromLocalFile(QString::fromStdString(path)));
+    } else {
+        progress.close();
+        failed();
+    }
+}
+
+void GMainWindow::OnGameListCopyTID(u64 program_id) {
+    QClipboard* clipboard = QGuiApplication::clipboard();
+    clipboard->setText(QString::fromStdString(fmt::format("{:016X}", program_id)));
+}
+
 void GMainWindow::OnGameListNavigateToGamedbEntry(u64 program_id,
                                                   const CompatibilityList& compatibility_list) {
     const auto it = FindMatchingCompatibilityEntry(compatibility_list, program_id);
@@ -790,7 +930,8 @@ void GMainWindow::OnMenuInstallToNAND() {
         return;
     }
 
-    const auto qt_raw_copy = [this](FileSys::VirtualFile src, FileSys::VirtualFile dest) {
+    const auto qt_raw_copy = [this](const FileSys::VirtualFile& src,
+                                    const FileSys::VirtualFile& dest, std::size_t block_size) {
         if (src == nullptr || dest == nullptr)
             return false;
         if (!dest->Resize(src->GetSize()))

src/yuzu/main.h

@@ -138,6 +138,8 @@ private slots:
     /// Called whenever a user selects a game in the game list widget.
     void OnGameListLoadFile(QString game_path);
     void OnGameListOpenFolder(u64 program_id, GameListOpenTarget target);
+    void OnGameListDumpRomFS(u64 program_id, const std::string& game_path);
+    void OnGameListCopyTID(u64 program_id);
     void OnGameListNavigateToGamedbEntry(u64 program_id,
                                          const CompatibilityList& compatibility_list);
     void OnMenuLoadFile();