gl_buffer_cache: Default initialize member variables

Ensures that the cache always has a deterministic initial state.

.travis.yml

@@ -20,6 +20,7 @@ matrix:
       install: "./.travis/linux/deps.sh"
       script: "./.travis/linux/build.sh"
       after_success: "./.travis/linux/upload.sh"
+      cache: ccache
     - os: osx
       env: NAME="macos build"
       sudo: false
@@ -27,6 +28,7 @@ matrix:
       install: "./.travis/macos/deps.sh"
       script: "./.travis/macos/build.sh"
       after_success: "./.travis/macos/upload.sh"
+      cache: ccache
 
 deploy:
   provider: releases
@@ -42,7 +44,3 @@ notifications:
   webhooks:
     urls:
       - https://api.yuzu-emu.org/code/travis/notify
-
-cache:
-  directories:
-    - $HOME/.ccache

.travis/linux/build.sh

@@ -1,3 +1,4 @@
 #!/bin/bash -ex
 
-docker run -e CCACHE_DIR=/ccache -v $HOME/.ccache:/ccache --env-file .travis/common/travis-ci.env -v $(pwd):/yuzu ubuntu:18.04 /bin/bash /yuzu/.travis/linux/docker.sh
+mkdir -p "$HOME/.ccache"
+docker run --env-file .travis/common/travis-ci.env -v $(pwd):/yuzu -v "$HOME/.ccache":/root/.ccache ubuntu:18.04 /bin/bash /yuzu/.travis/linux/docker.sh

.travis/linux/docker.sh

@@ -5,14 +5,8 @@ apt-get install --no-install-recommends -y build-essential git libqt5opengl5-dev
 
 cd /yuzu
 
-export PATH=/usr/lib/ccache:$PATH
-ln -sf /usr/bin/ccache /usr/lib/ccache/cc
-ln -sf /usr/bin/ccache /usr/lib/ccache/c++
 mkdir build && cd build
-ccache --show-stats > ccache_before
-cmake .. -DYUZU_BUILD_UNICORN=ON -DCMAKE_BUILD_TYPE=Release -G Ninja
+cmake .. -DYUZU_BUILD_UNICORN=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=/usr/lib/ccache/gcc -DCMAKE_CXX_COMPILER=/usr/lib/ccache/g++ -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -G Ninja
 ninja
-ccache --show-stats > ccache_after
-diff -U100 ccache_before ccache_after || true
 
 ctest -VV -C Release

.travis/macos/build.sh

@@ -5,14 +5,11 @@ set -o pipefail
 export MACOSX_DEPLOYMENT_TARGET=10.12
 export Qt5_DIR=$(brew --prefix)/opt/qt5
 export UNICORNDIR=$(pwd)/externals/unicorn
+export PATH="/usr/local/opt/ccache/libexec:$PATH"
 
 mkdir build && cd build
-export PATH=/usr/local/opt/ccache/libexec:$PATH
-ccache --show-stats > ccache_before
 cmake --version
-cmake .. -DYUZU_BUILD_UNICORN=ON -DCMAKE_BUILD_TYPE=Release
+cmake .. -DYUZU_BUILD_UNICORN=ON -DCMAKE_BUILD_TYPE=Release -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON
 make -j4
-ccache --show-stats > ccache_after
-diff -U100 ccache_before ccache_after || true
 
 ctest -VV -C Release

CMakeLists.txt

@@ -41,6 +41,19 @@ function(check_submodules_present)
 endfunction()
 check_submodules_present()
 
+configure_file(${CMAKE_SOURCE_DIR}/dist/compatibility_list/compatibility_list.qrc
+               ${CMAKE_BINARY_DIR}/dist/compatibility_list/compatibility_list.qrc
+               COPYONLY)
+if (ENABLE_COMPATIBILITY_LIST_DOWNLOAD AND NOT EXISTS ${CMAKE_BINARY_DIR}/dist/compatibility_list/compatibility_list.json)
+    message(STATUS "Downloading compatibility list for yuzu...")
+    file(DOWNLOAD
+        https://api.yuzu-emu.org/gamedb/
+        "${CMAKE_BINARY_DIR}/dist/compatibility_list/compatibility_list.json" SHOW_PROGRESS)
+endif()
+if (NOT EXISTS ${CMAKE_BINARY_DIR}/dist/compatibility_list/compatibility_list.json)
+    file(WRITE ${CMAKE_BINARY_DIR}/dist/compatibility_list/compatibility_list.json "")
+endif()
+
 # Detect current compilation architecture and create standard definitions
 # =======================================================================
 

appveyor.yml

@@ -41,9 +41,9 @@ before_build:
   - ps: |
         if ($env:BUILD_TYPE -eq 'msvc') {
           # redirect stderr and change the exit code to prevent powershell from cancelling the build if cmake prints a warning
-          cmd /C 'cmake -G "Visual Studio 15 2017 Win64" -DYUZU_USE_BUNDLED_QT=1 -DYUZU_USE_BUNDLED_SDL2=1 -DYUZU_USE_BUNDLED_UNICORN=1 .. 2>&1 && exit 0'
+          cmd /C 'cmake -G "Visual Studio 15 2017 Win64" -DYUZU_USE_BUNDLED_QT=1 -DYUZU_USE_BUNDLED_SDL2=1 -DYUZU_USE_BUNDLED_UNICORN=1 -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON .. 2>&1 && exit 0'
         } else {
-          C:\msys64\usr\bin\bash.exe -lc "cmake -G 'MSYS Makefiles' -DYUZU_BUILD_UNICORN=1 -DCMAKE_BUILD_TYPE=Release .. 2>&1"
+          C:\msys64\usr\bin\bash.exe -lc "cmake -G 'MSYS Makefiles' -DYUZU_BUILD_UNICORN=1 -DCMAKE_BUILD_TYPE=Release -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON .. 2>&1"
         }
   - cd ..
 
@@ -162,10 +162,6 @@ artifacts:
   - path: $(BUILD_ZIP)
     name: build
     type: zip
-  - path: $(BUILD_SYMBOLS)
-    name: debugsymbols
-  - path: $(BUILD_UPDATE)
-    name: update
 
 deploy:
   provider: GitHub

dist/compatibility_list/compatibility_list.qrc

@@ -0,0 +1,5 @@
+<RCC>
+  <qresource prefix="compatibility_list">
+      <file>compatibility_list.json</file>
+  </qresource>
+</RCC>

src/audio_core/stream.cpp

@@ -10,6 +10,7 @@
 #include "audio_core/stream.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "common/microprofile.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/settings.h"
@@ -94,7 +95,10 @@ void Stream::PlayNextBuffer() {
     CoreTiming::ScheduleEventThreadsafe(GetBufferReleaseCycles(*active_buffer), release_event, {});
 }
 
+MICROPROFILE_DEFINE(AudioOutput, "Audio", "ReleaseActiveBuffer", MP_RGB(100, 100, 255));
+
 void Stream::ReleaseActiveBuffer() {
+    MICROPROFILE_SCOPE(AudioOutput);
     ASSERT(active_buffer);
     released_buffers.push(std::move(active_buffer));
     release_callback();

src/common/logging/filter.h

@@ -19,7 +19,7 @@ namespace Log {
 class Filter {
 public:
     /// Initializes the filter with all classes having `default_level` as the minimum level.
-    Filter(Level default_level = Level::Info);
+    explicit Filter(Level default_level = Level::Info);
 
     /// Resets the filter so that all classes have `level` as the minimum displayed level.
     void ResetAll(Level level);
@@ -49,6 +49,6 @@ public:
     bool IsDebug() const;
 
 private:
-    std::array<Level, (size_t)Class::Count> class_levels;
+    std::array<Level, static_cast<size_t>(Class::Count)> class_levels;
 };
 } // namespace Log

src/common/logging/log.h

@@ -12,14 +12,14 @@ namespace Log {
 /// Specifies the severity or level of detail of the log message.
 enum class Level : u8 {
     Trace,    ///< Extremely detailed and repetitive debugging information that is likely to
-              ///  pollute logs.
+              ///< pollute logs.
     Debug,    ///< Less detailed debugging information.
     Info,     ///< Status information from important points during execution.
     Warning,  ///< Minor or potential problems found during execution of a task.
     Error,    ///< Major problems found during execution of a task that prevent it from being
-              ///  completed.
-    Critical, ///< Major problems during execution that threathen the stability of the entire
-              ///  application.
+              ///< completed.
+    Critical, ///< Major problems during execution that threaten the stability of the entire
+              ///< application.
 
     Count ///< Total number of logging levels
 };
@@ -49,7 +49,7 @@ enum class Class : ClassType {
     Kernel,            ///< The HLE implementation of the CTR kernel
     Kernel_SVC,        ///< Kernel system calls
     Service,           ///< HLE implementation of system services. Each major service
-                       ///  should have its own subclass.
+                       ///< should have its own subclass.
     Service_ACC,       ///< The ACC (Accounts) service
     Service_AM,        ///< The AM (Applet manager) service
     Service_AOC,       ///< The AOC (AddOn Content) service

src/common/logging/text_formatter.h

@@ -15,6 +15,6 @@ struct Entry;
 std::string FormatLogMessage(const Entry& entry);
 /// Formats and prints a log entry to stderr.
 void PrintMessage(const Entry& entry);
-/// Prints the same message as `PrintMessage`, but colored acoording to the severity level.
+/// Prints the same message as `PrintMessage`, but colored according to the severity level.
 void PrintColoredMessage(const Entry& entry);
 } // namespace Log

src/core/CMakeLists.txt

@@ -35,8 +35,12 @@ add_library(core STATIC
     file_sys/mode.h
     file_sys/nca_metadata.cpp
     file_sys/nca_metadata.h
+    file_sys/nca_patch.cpp
+    file_sys/nca_patch.h
     file_sys/partition_filesystem.cpp
     file_sys/partition_filesystem.h
+    file_sys/patch_manager.cpp
+    file_sys/patch_manager.h
     file_sys/program_metadata.cpp
     file_sys/program_metadata.h
     file_sys/registered_cache.cpp
@@ -49,6 +53,8 @@ add_library(core STATIC
     file_sys/savedata_factory.h
     file_sys/sdmc_factory.cpp
     file_sys/sdmc_factory.h
+    file_sys/submission_package.cpp
+    file_sys/submission_package.h
     file_sys/vfs.cpp
     file_sys/vfs.h
     file_sys/vfs_concat.cpp
@@ -359,6 +365,8 @@ add_library(core STATIC
     loader/nro.h
     loader/nso.cpp
     loader/nso.h
+    loader/nsp.cpp
+    loader/nsp.h
     loader/xci.cpp
     loader/xci.h
     memory.cpp

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -7,8 +7,10 @@
 #include <dynarmic/A64/a64.h>
 #include <dynarmic/A64/config.h>
 #include "common/logging/log.h"
+#include "common/microprofile.h"
 #include "core/arm/dynarmic/arm_dynarmic.h"
 #include "core/core.h"
+#include "core/core_cpu.h"
 #include "core/core_timing.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/svc.h"
@@ -142,7 +144,10 @@ std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit() const {
     return std::make_unique<Dynarmic::A64::Jit>(config);
 }
 
+MICROPROFILE_DEFINE(ARM_Jit_Dynarmic, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
+
 void ARM_Dynarmic::Run() {
+    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic);
     ASSERT(Memory::GetCurrentPageTable() == current_page_table);
 
     jit->Run();

src/core/arm/unicorn/arm_unicorn.cpp

@@ -193,10 +193,10 @@ void ARM_Unicorn::Step() {
     ExecuteInstructions(1);
 }
 
-MICROPROFILE_DEFINE(ARM_Jit, "ARM JIT", "ARM JIT", MP_RGB(255, 64, 64));
+MICROPROFILE_DEFINE(ARM_Jit_Unicorn, "ARM JIT", "Unicorn", MP_RGB(255, 64, 64));
 
 void ARM_Unicorn::ExecuteInstructions(int num_instructions) {
-    MICROPROFILE_SCOPE(ARM_Jit);
+    MICROPROFILE_SCOPE(ARM_Jit_Unicorn);
     CHECKED(uc_emu_start(uc, GetPC(), 1ULL << 63, 0, num_instructions));
     CoreTiming::AddTicks(num_instructions);
     if (GDBStub::IsServerEnabled()) {

src/core/core.cpp

@@ -2,24 +2,36 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <array>
+#include <map>
 #include <memory>
+#include <thread>
 #include <utility>
+
 #include "common/logging/log.h"
 #include "common/string_util.h"
+#include "core/arm/exclusive_monitor.h"
 #include "core/core.h"
+#include "core/core_cpu.h"
 #include "core/core_timing.h"
+#include "core/file_sys/mode.h"
+#include "core/file_sys/vfs_concat.h"
+#include "core/file_sys/vfs_real.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
+#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/service/service.h"
 #include "core/hle/service/sm/controller.h"
 #include "core/hle/service/sm/sm.h"
 #include "core/loader/loader.h"
+#include "core/perf_stats.h"
 #include "core/settings.h"
-#include "file_sys/vfs_concat.h"
-#include "file_sys/vfs_real.h"
+#include "core/telemetry_session.h"
+#include "video_core/debug_utils/debug_utils.h"
+#include "video_core/gpu.h"
 #include "video_core/renderer_base.h"
 #include "video_core/video_core.h"
 
@@ -258,7 +270,7 @@ struct System::Impl {
         }
     }
 
-    PerfStats::Results GetAndResetPerfStats() {
+    PerfStatsResults GetAndResetPerfStats() {
         return perf_stats.GetAndResetStats(CoreTiming::GetGlobalTimeUs());
     }
 
@@ -326,7 +338,7 @@ void System::PrepareReschedule() {
     CurrentCpuCore().PrepareReschedule();
 }
 
-PerfStats::Results System::GetAndResetPerfStats() {
+PerfStatsResults System::GetAndResetPerfStats() {
     return impl->GetAndResetPerfStats();
 }
 
@@ -429,15 +441,15 @@ void System::SetGPUDebugContext(std::shared_ptr<Tegra::DebugContext> context) {
     impl->debug_context = std::move(context);
 }
 
-std::shared_ptr<Tegra::DebugContext> System::GetGPUDebugContext() const {
-    return impl->debug_context;
+Tegra::DebugContext* System::GetGPUDebugContext() const {
+    return impl->debug_context.get();
 }
 
-void System::SetFilesystem(FileSys::VirtualFilesystem vfs) {
+void System::SetFilesystem(std::shared_ptr<FileSys::VfsFilesystem> vfs) {
     impl->virtual_filesystem = std::move(vfs);
 }
 
-FileSys::VirtualFilesystem System::GetFilesystem() const {
+std::shared_ptr<FileSys::VfsFilesystem> System::GetFilesystem() const {
     return impl->virtual_filesystem;
 }
 

src/core/core.h

@@ -4,41 +4,55 @@
 
 #pragma once
 
-#include <array>
-#include <map>
+#include <cstddef>
 #include <memory>
 #include <string>
-#include <thread>
+
 #include "common/common_types.h"
-#include "core/arm/exclusive_monitor.h"
-#include "core/core_cpu.h"
-#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/object.h"
-#include "core/hle/kernel/scheduler.h"
-#include "core/loader/loader.h"
-#include "core/memory.h"
-#include "core/perf_stats.h"
-#include "core/telemetry_session.h"
-#include "file_sys/vfs_real.h"
-#include "hle/service/filesystem/filesystem.h"
-#include "video_core/debug_utils/debug_utils.h"
-#include "video_core/gpu.h"
 
 namespace Core::Frontend {
 class EmuWindow;
-}
+} // namespace Core::Frontend
+
+namespace FileSys {
+class VfsFilesystem;
+} // namespace FileSys
+
+namespace Kernel {
+class KernelCore;
+class Process;
+class Scheduler;
+} // namespace Kernel
+
+namespace Loader {
+class AppLoader;
+enum class ResultStatus : u16;
+} // namespace Loader
 
 namespace Service::SM {
 class ServiceManager;
-}
+} // namespace Service::SM
+
+namespace Tegra {
+class DebugContext;
+class GPU;
+} // namespace Tegra
 
 namespace VideoCore {
 class RendererBase;
-}
+} // namespace VideoCore
 
 namespace Core {
 
 class ARM_Interface;
+class Cpu;
+class ExclusiveMonitor;
+class FrameLimiter;
+class PerfStats;
+class TelemetrySession;
+
+struct PerfStatsResults;
 
 class System {
 public:
@@ -125,7 +139,7 @@ public:
     void PrepareReschedule();
 
     /// Gets and resets core performance statistics
-    PerfStats::Results GetAndResetPerfStats();
+    PerfStatsResults GetAndResetPerfStats();
 
     /// Gets an ARM interface to the CPU core that is currently running
     ARM_Interface& CurrentArmInterface();
@@ -195,11 +209,11 @@ public:
 
     void SetGPUDebugContext(std::shared_ptr<Tegra::DebugContext> context);
 
-    std::shared_ptr<Tegra::DebugContext> GetGPUDebugContext() const;
+    Tegra::DebugContext* GetGPUDebugContext() const;
 
-    void SetFilesystem(FileSys::VirtualFilesystem vfs);
+    void SetFilesystem(std::shared_ptr<FileSys::VfsFilesystem> vfs);
 
-    FileSys::VirtualFilesystem GetFilesystem() const;
+    std::shared_ptr<FileSys::VfsFilesystem> GetFilesystem() const;
 
 private:
     System();

src/core/crypto/aes_util.cpp

@@ -82,11 +82,25 @@ void AESCipher<Key, KeySize>::Transcode(const u8* src, size_t size, u8* dest, Op
         }
     } else {
         const auto block_size = mbedtls_cipher_get_block_size(context);
+        if (size < block_size) {
+            std::vector<u8> block(block_size);
+            std::memcpy(block.data(), src, size);
+            Transcode(block.data(), block.size(), block.data(), op);
+            std::memcpy(dest, block.data(), size);
+            return;
+        }
 
         for (size_t offset = 0; offset < size; offset += block_size) {
             auto length = std::min<size_t>(block_size, size - offset);
             mbedtls_cipher_update(context, src + offset, length, dest + offset, &written);
             if (written != length) {
+                if (length < block_size) {
+                    std::vector<u8> block(block_size);
+                    std::memcpy(block.data(), src + offset, length);
+                    Transcode(block.data(), block.size(), block.data(), op);
+                    std::memcpy(dest + offset, block.data(), length);
+                    return;
+                }
                 LOG_WARNING(Crypto, "Not all data was decrypted requested={:016X}, actual={:016X}.",
                             length, written);
             }

src/core/crypto/ctr_encryption_layer.cpp

@@ -21,7 +21,7 @@ size_t CTREncryptionLayer::Read(u8* data, size_t length, size_t offset) const {
         UpdateIV(base_offset + offset);
         std::vector<u8> raw = base->ReadBytes(length, offset);
         cipher.Transcode(raw.data(), raw.size(), data, Op::Decrypt);
-        return raw.size();
+        return length;
     }
 
     // offset does not fall on block boundary (0x10)

src/core/crypto/key_manager.cpp

@@ -8,12 +8,15 @@
 #include <locale>
 #include <sstream>
 #include <string_view>
+#include <tuple>
+#include <vector>
 #include "common/common_paths.h"
 #include "common/file_util.h"
 #include "common/hex_util.h"
 #include "common/logging/log.h"
 #include "core/crypto/aes_util.h"
 #include "core/crypto/key_manager.h"
+#include "core/loader/loader.h"
 #include "core/settings.h"
 
 namespace Core::Crypto {
@@ -228,18 +231,28 @@ void KeyManager::WriteKeyToFile(bool title_key, std::string_view keyname,
 }
 
 void KeyManager::SetKey(S128KeyType id, Key128 key, u64 field1, u64 field2) {
-    const auto iter = std::find_if(
+    if (s128_keys.find({id, field1, field2}) != s128_keys.end())
+        return;
+    if (id == S128KeyType::Titlekey) {
+        Key128 rights_id;
+        std::memcpy(rights_id.data(), &field2, sizeof(u64));
+        std::memcpy(rights_id.data() + sizeof(u64), &field1, sizeof(u64));
+        WriteKeyToFile(true, Common::HexArrayToString(rights_id), key);
+    }
+    const auto iter2 = std::find_if(
         s128_file_id.begin(), s128_file_id.end(),
         [&id, &field1, &field2](const std::pair<std::string, KeyIndex<S128KeyType>> elem) {
             return std::tie(elem.second.type, elem.second.field1, elem.second.field2) ==
                    std::tie(id, field1, field2);
         });
-    if (iter != s128_file_id.end())
-        WriteKeyToFile(id == S128KeyType::Titlekey, iter->first, key);
+    if (iter2 != s128_file_id.end())
+        WriteKeyToFile(false, iter2->first, key);
     s128_keys[{id, field1, field2}] = key;
 }
 
 void KeyManager::SetKey(S256KeyType id, Key256 key, u64 field1, u64 field2) {
+    if (s256_keys.find({id, field1, field2}) != s256_keys.end())
+        return;
     const auto iter = std::find_if(
         s256_file_id.begin(), s256_file_id.end(),
         [&id, &field1, &field2](const std::pair<std::string, KeyIndex<S256KeyType>> elem) {

src/core/crypto/key_manager.h

@@ -6,16 +6,19 @@
 
 #include <array>
 #include <string>
-#include <string_view>
-#include <type_traits>
-#include <vector>
 #include <boost/container/flat_map.hpp>
+#include <boost/optional.hpp>
 #include <fmt/format.h>
 #include "common/common_types.h"
-#include "core/loader/loader.h"
+
+namespace Loader {
+enum class ResultStatus : u16;
+}
 
 namespace Core::Crypto {
 
+constexpr u64 TICKET_FILE_TITLEKEY_OFFSET = 0x180;
+
 using Key128 = std::array<u8, 0x10>;
 using Key256 = std::array<u8, 0x20>;
 using SHA256Hash = std::array<u8, 0x20>;

src/core/file_sys/bis_factory.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include "core/file_sys/bis_factory.h"
+#include "core/file_sys/registered_cache.h"
 
 namespace FileSys {
 
@@ -13,6 +14,8 @@ BISFactory::BISFactory(VirtualDir nand_root_)
       usrnand_cache(std::make_shared<RegisteredCache>(
           GetOrCreateDirectoryRelative(nand_root, "/user/Contents/registered"))) {}
 
+BISFactory::~BISFactory() = default;
+
 std::shared_ptr<RegisteredCache> BISFactory::GetSystemNANDContents() const {
     return sysnand_cache;
 }

src/core/file_sys/bis_factory.h

@@ -5,17 +5,20 @@
 #pragma once
 
 #include <memory>
-#include "core/loader/loader.h"
-#include "registered_cache.h"
+
+#include "core/file_sys/vfs.h"
 
 namespace FileSys {
 
+class RegisteredCache;
+
 /// File system interface to the Built-In Storage
 /// This is currently missing accessors to BIS partitions, but seemed like a good place for the NAND
 /// registered caches.
 class BISFactory {
 public:
     explicit BISFactory(VirtualDir nand_root);
+    ~BISFactory();
 
     std::shared_ptr<RegisteredCache> GetSystemNANDContents() const;
     std::shared_ptr<RegisteredCache> GetUserNANDContents() const;

src/core/file_sys/card_image.cpp

@@ -9,7 +9,10 @@
 
 #include "common/logging/log.h"
 #include "core/file_sys/card_image.h"
+#include "core/file_sys/content_archive.h"
+#include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/partition_filesystem.h"
+#include "core/file_sys/submission_package.h"
 #include "core/file_sys/vfs_offset.h"
 #include "core/loader/loader.h"
 
@@ -43,15 +46,19 @@ XCI::XCI(VirtualFile file_) : file(std::move(file_)), partitions(0x4) {
             partitions[static_cast<size_t>(partition)] = std::make_shared<PartitionFilesystem>(raw);
     }
 
-    program_nca_status = Loader::ResultStatus::ErrorXCIMissingProgramNCA;
+    secure_partition = std::make_shared<NSP>(
+        main_hfs.GetFile(partition_names[static_cast<size_t>(XCIPartition::Secure)]));
 
-    auto result = AddNCAFromPartition(XCIPartition::Secure);
-    if (result != Loader::ResultStatus::Success) {
-        status = result;
-        return;
-    }
+    const auto secure_ncas = secure_partition->GetNCAsCollapsed();
+    std::copy(secure_ncas.begin(), secure_ncas.end(), std::back_inserter(ncas));
 
-    result = AddNCAFromPartition(XCIPartition::Update);
+    program =
+        secure_partition->GetNCA(secure_partition->GetProgramTitleID(), ContentRecordType::Program);
+    program_nca_status = secure_partition->GetProgramStatus(secure_partition->GetProgramTitleID());
+    if (program_nca_status == Loader::ResultStatus::ErrorNSPMissingProgramNCA)
+        program_nca_status = Loader::ResultStatus::ErrorXCIMissingProgramNCA;
+
+    auto result = AddNCAFromPartition(XCIPartition::Update);
     if (result != Loader::ResultStatus::Success) {
         status = result;
         return;
@@ -74,6 +81,8 @@ XCI::XCI(VirtualFile file_) : file(std::move(file_)), partitions(0x4) {
     status = Loader::ResultStatus::Success;
 }
 
+XCI::~XCI() = default;
+
 Loader::ResultStatus XCI::GetStatus() const {
     return status;
 }
@@ -86,6 +95,10 @@ VirtualDir XCI::GetPartition(XCIPartition partition) const {
     return partitions[static_cast<size_t>(partition)];
 }
 
+std::shared_ptr<NSP> XCI::GetSecurePartitionNSP() const {
+    return secure_partition;
+}
+
 VirtualDir XCI::GetSecurePartition() const {
     return GetPartition(XCIPartition::Secure);
 }
@@ -102,6 +115,20 @@ VirtualDir XCI::GetLogoPartition() const {
     return GetPartition(XCIPartition::Logo);
 }
 
+u64 XCI::GetProgramTitleID() const {
+    return secure_partition->GetProgramTitleID();
+}
+
+std::shared_ptr<NCA> XCI::GetProgramNCA() const {
+    return program;
+}
+
+VirtualFile XCI::GetProgramNCAFile() const {
+    if (GetProgramNCA() == nullptr)
+        return nullptr;
+    return GetProgramNCA()->GetBaseFile();
+}
+
 const std::vector<std::shared_ptr<NCA>>& XCI::GetNCAs() const {
     return ncas;
 }

src/core/file_sys/card_image.h

@@ -5,15 +5,22 @@
 #pragma once
 
 #include <array>
+#include <memory>
 #include <vector>
 #include "common/common_types.h"
 #include "common/swap.h"
-#include "core/file_sys/content_archive.h"
 #include "core/file_sys/vfs.h"
-#include "core/loader/loader.h"
+
+namespace Loader {
+enum class ResultStatus : u16;
+}
 
 namespace FileSys {
 
+class NCA;
+enum class NCAContentType : u8;
+class NSP;
+
 enum class GamecardSize : u8 {
     S_1GB = 0xFA,
     S_2GB = 0xF8,
@@ -57,6 +64,7 @@ enum class XCIPartition : u8 { Update, Normal, Secure, Logo };
 class XCI : public ReadOnlyVfsDirectory {
 public:
     explicit XCI(VirtualFile file);
+    ~XCI() override;
 
     Loader::ResultStatus GetStatus() const;
     Loader::ResultStatus GetProgramNCAStatus() const;
@@ -64,11 +72,16 @@ public:
     u8 GetFormatVersion() const;
 
     VirtualDir GetPartition(XCIPartition partition) const;
+    std::shared_ptr<NSP> GetSecurePartitionNSP() const;
     VirtualDir GetSecurePartition() const;
     VirtualDir GetNormalPartition() const;
     VirtualDir GetUpdatePartition() const;
     VirtualDir GetLogoPartition() const;
 
+    u64 GetProgramTitleID() const;
+
+    std::shared_ptr<NCA> GetProgramNCA() const;
+    VirtualFile GetProgramNCAFile() const;
     const std::vector<std::shared_ptr<NCA>>& GetNCAs() const;
     std::shared_ptr<NCA> GetNCAByType(NCAContentType type) const;
     VirtualFile GetNCAFileByType(NCAContentType type) const;
@@ -94,6 +107,8 @@ private:
     Loader::ResultStatus program_nca_status;
 
     std::vector<VirtualDir> partitions;
+    std::shared_ptr<NSP> secure_partition;
+    std::shared_ptr<NCA> program;
     std::vector<std::shared_ptr<NCA>> ncas;
 };
 } // namespace FileSys

src/core/file_sys/content_archive.cpp

@@ -3,12 +3,17 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
+#include <cstring>
 #include <utility>
+
 #include <boost/optional.hpp>
+
 #include "common/logging/log.h"
 #include "core/crypto/aes_util.h"
 #include "core/crypto/ctr_encryption_layer.h"
 #include "core/file_sys/content_archive.h"
+#include "core/file_sys/nca_patch.h"
+#include "core/file_sys/partition_filesystem.h"
 #include "core/file_sys/romfs.h"
 #include "core/file_sys/vfs_offset.h"
 #include "core/loader/loader.h"
@@ -64,10 +69,31 @@ struct RomFSSuperblock {
 };
 static_assert(sizeof(RomFSSuperblock) == 0x200, "RomFSSuperblock has incorrect size.");
 
+struct BKTRHeader {
+    u64_le offset;
+    u64_le size;
+    u32_le magic;
+    INSERT_PADDING_BYTES(0x4);
+    u32_le number_entries;
+    INSERT_PADDING_BYTES(0x4);
+};
+static_assert(sizeof(BKTRHeader) == 0x20, "BKTRHeader has incorrect size.");
+
+struct BKTRSuperblock {
+    NCASectionHeaderBlock header_block;
+    IVFCHeader ivfc;
+    INSERT_PADDING_BYTES(0x18);
+    BKTRHeader relocation;
+    BKTRHeader subsection;
+    INSERT_PADDING_BYTES(0xC0);
+};
+static_assert(sizeof(BKTRSuperblock) == 0x200, "BKTRSuperblock has incorrect size.");
+
 union NCASectionHeader {
     NCASectionRaw raw;
     PFS0Superblock pfs0;
     RomFSSuperblock romfs;
+    BKTRSuperblock bktr;
 };
 static_assert(sizeof(NCASectionHeader) == 0x200, "NCASectionHeader has incorrect size.");
 
@@ -100,7 +126,7 @@ boost::optional<Core::Crypto::Key128> NCA::GetKeyAreaKey(NCASectionCryptoType ty
     Core::Crypto::Key128 out;
     if (type == NCASectionCryptoType::XTS)
         std::copy(key_area.begin(), key_area.begin() + 0x10, out.begin());
-    else if (type == NCASectionCryptoType::CTR)
+    else if (type == NCASectionCryptoType::CTR || type == NCASectionCryptoType::BKTR)
         std::copy(key_area.begin() + 0x20, key_area.begin() + 0x30, out.begin());
     else
         LOG_CRITICAL(Crypto, "Called GetKeyAreaKey on invalid NCASectionCryptoType type={:02X}",
@@ -150,6 +176,9 @@ VirtualFile NCA::Decrypt(NCASectionHeader s_header, VirtualFile in, u64 starting
         LOG_DEBUG(Crypto, "called with mode=NONE");
         return in;
     case NCASectionCryptoType::CTR:
+    // During normal BKTR decryption, this entire function is skipped. This is for the metadata,
+    // which uses the same CTR as usual.
+    case NCASectionCryptoType::BKTR:
         LOG_DEBUG(Crypto, "called with mode=CTR, starting_offset={:016X}", starting_offset);
         {
             boost::optional<Core::Crypto::Key128> key = boost::none;
@@ -186,7 +215,9 @@ VirtualFile NCA::Decrypt(NCASectionHeader s_header, VirtualFile in, u64 starting
     }
 }
 
-NCA::NCA(VirtualFile file_) : file(std::move(file_)) {
+NCA::NCA(VirtualFile file_, VirtualFile bktr_base_romfs_, u64 bktr_base_ivfc_offset)
+    : file(std::move(file_)),
+      bktr_base_romfs(bktr_base_romfs_ ? std::move(bktr_base_romfs_) : nullptr) {
     status = Loader::ResultStatus::Success;
 
     if (file == nullptr) {
@@ -261,22 +292,21 @@ NCA::NCA(VirtualFile file_) : file(std::move(file_)) {
     is_update = std::find_if(sections.begin(), sections.end(), [](const NCASectionHeader& header) {
                     return header.raw.header.crypto_type == NCASectionCryptoType::BKTR;
                 }) != sections.end();
+    ivfc_offset = 0;
 
     for (std::ptrdiff_t i = 0; i < number_sections; ++i) {
         auto section = sections[i];
 
         if (section.raw.header.filesystem_type == NCASectionFilesystemType::ROMFS) {
-            const size_t romfs_offset =
-                header.section_tables[i].media_offset * MEDIA_OFFSET_MULTIPLIER +
-                section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset;
+            const size_t base_offset =
+                header.section_tables[i].media_offset * MEDIA_OFFSET_MULTIPLIER;
+            ivfc_offset = section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset;
+            const size_t romfs_offset = base_offset + ivfc_offset;
             const size_t romfs_size = section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].size;
-            auto dec =
-                Decrypt(section, std::make_shared<OffsetVfsFile>(file, romfs_size, romfs_offset),
-                        romfs_offset);
-            if (dec != nullptr) {
-                files.push_back(std::move(dec));
-                romfs = files.back();
-            } else {
+            auto raw = std::make_shared<OffsetVfsFile>(file, romfs_size, romfs_offset);
+            auto dec = Decrypt(section, raw, romfs_offset);
+
+            if (dec == nullptr) {
                 if (status != Loader::ResultStatus::Success)
                     return;
                 if (has_rights_id)
@@ -285,6 +315,117 @@ NCA::NCA(VirtualFile file_) : file(std::move(file_)) {
                     status = Loader::ResultStatus::ErrorIncorrectKeyAreaKey;
                 return;
             }
+
+            if (section.raw.header.crypto_type == NCASectionCryptoType::BKTR) {
+                if (section.bktr.relocation.magic != Common::MakeMagic('B', 'K', 'T', 'R') ||
+                    section.bktr.subsection.magic != Common::MakeMagic('B', 'K', 'T', 'R')) {
+                    status = Loader::ResultStatus::ErrorBadBKTRHeader;
+                    return;
+                }
+
+                if (section.bktr.relocation.offset + section.bktr.relocation.size !=
+                    section.bktr.subsection.offset) {
+                    status = Loader::ResultStatus::ErrorBKTRSubsectionNotAfterRelocation;
+                    return;
+                }
+
+                const u64 size =
+                    MEDIA_OFFSET_MULTIPLIER * (header.section_tables[i].media_end_offset -
+                                               header.section_tables[i].media_offset);
+                if (section.bktr.subsection.offset + section.bktr.subsection.size != size) {
+                    status = Loader::ResultStatus::ErrorBKTRSubsectionNotAtEnd;
+                    return;
+                }
+
+                const u64 offset = section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset;
+                RelocationBlock relocation_block{};
+                if (dec->ReadObject(&relocation_block, section.bktr.relocation.offset - offset) !=
+                    sizeof(RelocationBlock)) {
+                    status = Loader::ResultStatus::ErrorBadRelocationBlock;
+                    return;
+                }
+                SubsectionBlock subsection_block{};
+                if (dec->ReadObject(&subsection_block, section.bktr.subsection.offset - offset) !=
+                    sizeof(RelocationBlock)) {
+                    status = Loader::ResultStatus::ErrorBadSubsectionBlock;
+                    return;
+                }
+
+                std::vector<RelocationBucketRaw> relocation_buckets_raw(
+                    (section.bktr.relocation.size - sizeof(RelocationBlock)) /
+                    sizeof(RelocationBucketRaw));
+                if (dec->ReadBytes(relocation_buckets_raw.data(),
+                                   section.bktr.relocation.size - sizeof(RelocationBlock),
+                                   section.bktr.relocation.offset + sizeof(RelocationBlock) -
+                                       offset) !=
+                    section.bktr.relocation.size - sizeof(RelocationBlock)) {
+                    status = Loader::ResultStatus::ErrorBadRelocationBuckets;
+                    return;
+                }
+
+                std::vector<SubsectionBucketRaw> subsection_buckets_raw(
+                    (section.bktr.subsection.size - sizeof(SubsectionBlock)) /
+                    sizeof(SubsectionBucketRaw));
+                if (dec->ReadBytes(subsection_buckets_raw.data(),
+                                   section.bktr.subsection.size - sizeof(SubsectionBlock),
+                                   section.bktr.subsection.offset + sizeof(SubsectionBlock) -
+                                       offset) !=
+                    section.bktr.subsection.size - sizeof(SubsectionBlock)) {
+                    status = Loader::ResultStatus::ErrorBadSubsectionBuckets;
+                    return;
+                }
+
+                std::vector<RelocationBucket> relocation_buckets(relocation_buckets_raw.size());
+                std::transform(relocation_buckets_raw.begin(), relocation_buckets_raw.end(),
+                               relocation_buckets.begin(), &ConvertRelocationBucketRaw);
+                std::vector<SubsectionBucket> subsection_buckets(subsection_buckets_raw.size());
+                std::transform(subsection_buckets_raw.begin(), subsection_buckets_raw.end(),
+                               subsection_buckets.begin(), &ConvertSubsectionBucketRaw);
+
+                u32 ctr_low;
+                std::memcpy(&ctr_low, section.raw.section_ctr.data(), sizeof(ctr_low));
+                subsection_buckets.back().entries.push_back(
+                    {section.bktr.relocation.offset, {0}, ctr_low});
+                subsection_buckets.back().entries.push_back({size, {0}, 0});
+
+                boost::optional<Core::Crypto::Key128> key = boost::none;
+                if (encrypted) {
+                    if (has_rights_id) {
+                        status = Loader::ResultStatus::Success;
+                        key = GetTitlekey();
+                        if (key == boost::none) {
+                            status = Loader::ResultStatus::ErrorMissingTitlekey;
+                            return;
+                        }
+                    } else {
+                        key = GetKeyAreaKey(NCASectionCryptoType::BKTR);
+                        if (key == boost::none) {
+                            status = Loader::ResultStatus::ErrorMissingKeyAreaKey;
+                            return;
+                        }
+                    }
+                }
+
+                if (bktr_base_romfs == nullptr) {
+                    status = Loader::ResultStatus::ErrorMissingBKTRBaseRomFS;
+                    return;
+                }
+
+                auto bktr = std::make_shared<BKTR>(
+                    bktr_base_romfs, std::make_shared<OffsetVfsFile>(file, romfs_size, base_offset),
+                    relocation_block, relocation_buckets, subsection_block, subsection_buckets,
+                    encrypted, encrypted ? key.get() : Core::Crypto::Key128{}, base_offset,
+                    bktr_base_ivfc_offset, section.raw.section_ctr);
+
+                // BKTR applies to entire IVFC, so make an offset version to level 6
+
+                files.push_back(std::make_shared<OffsetVfsFile>(
+                    bktr, romfs_size, section.romfs.ivfc.levels[IVFC_MAX_LEVEL - 1].offset));
+                romfs = files.back();
+            } else {
+                files.push_back(std::move(dec));
+                romfs = files.back();
+            }
         } else if (section.raw.header.filesystem_type == NCASectionFilesystemType::PFS0) {
             u64 offset = (static_cast<u64>(header.section_tables[i].media_offset) *
                           MEDIA_OFFSET_MULTIPLIER) +
@@ -300,6 +441,12 @@ NCA::NCA(VirtualFile file_) : file(std::move(file_)) {
                     dirs.push_back(std::move(npfs));
                     if (IsDirectoryExeFS(dirs.back()))
                         exefs = dirs.back();
+                } else {
+                    if (has_rights_id)
+                        status = Loader::ResultStatus::ErrorIncorrectTitlekeyOrTitlekek;
+                    else
+                        status = Loader::ResultStatus::ErrorIncorrectKeyAreaKey;
+                    return;
                 }
             } else {
                 if (status != Loader::ResultStatus::Success)
@@ -345,11 +492,15 @@ NCAContentType NCA::GetType() const {
 }
 
 u64 NCA::GetTitleId() const {
-    if (status != Loader::ResultStatus::Success)
-        return {};
+    if (is_update || status == Loader::ResultStatus::ErrorMissingBKTRBaseRomFS)
+        return header.title_id | 0x800;
     return header.title_id;
 }
 
+bool NCA::IsUpdate() const {
+    return is_update;
+}
+
 VirtualFile NCA::GetRomFS() const {
     return romfs;
 }
@@ -362,8 +513,8 @@ VirtualFile NCA::GetBaseFile() const {
     return file;
 }
 
-bool NCA::IsUpdate() const {
-    return is_update;
+u64 NCA::GetBaseIVFCOffset() const {
+    return ivfc_offset;
 }
 
 bool NCA::ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) {

src/core/file_sys/content_archive.h

@@ -12,10 +12,12 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/swap.h"
-#include "control_metadata.h"
 #include "core/crypto/key_manager.h"
-#include "core/file_sys/partition_filesystem.h"
-#include "core/loader/loader.h"
+#include "core/file_sys/vfs.h"
+
+namespace Loader {
+enum class ResultStatus : u16;
+}
 
 namespace FileSys {
 
@@ -77,7 +79,8 @@ bool IsValidNCA(const NCAHeader& header);
 // After construction, use GetStatus to determine if the file is valid and ready to be used.
 class NCA : public ReadOnlyVfsDirectory {
 public:
-    explicit NCA(VirtualFile file);
+    explicit NCA(VirtualFile file, VirtualFile bktr_base_romfs = nullptr,
+                 u64 bktr_base_ivfc_offset = 0);
     Loader::ResultStatus GetStatus() const;
 
     std::vector<std::shared_ptr<VfsFile>> GetFiles() const override;
@@ -87,13 +90,15 @@ public:
 
     NCAContentType GetType() const;
     u64 GetTitleId() const;
+    bool IsUpdate() const;
 
     VirtualFile GetRomFS() const;
     VirtualDir GetExeFS() const;
 
     VirtualFile GetBaseFile() const;
 
-    bool IsUpdate() const;
+    // Returns the base ivfc offset used in BKTR patching.
+    u64 GetBaseIVFCOffset() const;
 
 protected:
     bool ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) override;
@@ -110,14 +115,16 @@ private:
     VirtualFile romfs = nullptr;
     VirtualDir exefs = nullptr;
     VirtualFile file;
+    VirtualFile bktr_base_romfs;
+    u64 ivfc_offset;
 
     NCAHeader header{};
     bool has_rights_id{};
-    bool is_update{};
 
     Loader::ResultStatus status{};
 
     bool encrypted;
+    bool is_update;
 
     Core::Crypto::KeyManager keys;
 };

src/core/file_sys/control_metadata.cpp

@@ -21,7 +21,17 @@ NACP::NACP(VirtualFile file) : raw(std::make_unique<RawNACP>()) {
 }
 
 const LanguageEntry& NACP::GetLanguageEntry(Language language) const {
-    return raw->language_entries.at(static_cast<u8>(language));
+    if (language != Language::Default) {
+        return raw->language_entries.at(static_cast<u8>(language));
+    } else {
+        for (const auto& language_entry : raw->language_entries) {
+            if (!language_entry.GetApplicationName().empty())
+                return language_entry;
+        }
+
+        // Fallback to English
+        return GetLanguageEntry(Language::AmericanEnglish);
+    }
 }
 
 std::string NACP::GetApplicationName(Language language) const {

src/core/file_sys/control_metadata.h

@@ -8,6 +8,8 @@
 #include <memory>
 #include <string>
 #include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "common/swap.h"
 #include "core/file_sys/vfs.h"
 
 namespace FileSys {
@@ -60,6 +62,8 @@ enum class Language : u8 {
     Korean = 12,
     Taiwanese = 13,
     Chinese = 14,
+
+    Default = 255,
 };
 
 static constexpr std::array<const char*, 15> LANGUAGE_NAMES = {
@@ -74,9 +78,9 @@ static constexpr std::array<const char*, 15> LANGUAGE_NAMES = {
 class NACP {
 public:
     explicit NACP(VirtualFile file);
-    const LanguageEntry& GetLanguageEntry(Language language = Language::AmericanEnglish) const;
-    std::string GetApplicationName(Language language = Language::AmericanEnglish) const;
-    std::string GetDeveloperName(Language language = Language::AmericanEnglish) const;
+    const LanguageEntry& GetLanguageEntry(Language language = Language::Default) const;
+    std::string GetApplicationName(Language language = Language::Default) const;
+    std::string GetDeveloperName(Language language = Language::Default) const;
     u64 GetTitleId() const;
     std::string GetVersionString() const;
 

src/core/file_sys/nca_metadata.cpp

@@ -3,10 +3,9 @@
 // Refer to the license.txt file included.
 
 #include <cstring>
-#include "common/common_funcs.h"
+#include "common/common_types.h"
 #include "common/logging/log.h"
 #include "common/swap.h"
-#include "content_archive.h"
 #include "core/file_sys/nca_metadata.h"
 
 namespace FileSys {

src/core/file_sys/nca_metadata.h

@@ -4,7 +4,6 @@
 
 #pragma once
 
-#include <cstring>
 #include <memory>
 #include <vector>
 #include "common/common_funcs.h"

src/core/file_sys/nca_patch.cpp

@@ -0,0 +1,206 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/crypto/aes_util.h"
+#include "core/file_sys/nca_patch.h"
+
+namespace FileSys {
+
+BKTR::BKTR(VirtualFile base_romfs_, VirtualFile bktr_romfs_, RelocationBlock relocation_,
+           std::vector<RelocationBucket> relocation_buckets_, SubsectionBlock subsection_,
+           std::vector<SubsectionBucket> subsection_buckets_, bool is_encrypted_,
+           Core::Crypto::Key128 key_, u64 base_offset_, u64 ivfc_offset_,
+           std::array<u8, 8> section_ctr_)
+    : base_romfs(std::move(base_romfs_)), bktr_romfs(std::move(bktr_romfs_)),
+      relocation(relocation_), relocation_buckets(std::move(relocation_buckets_)),
+      subsection(subsection_), subsection_buckets(std::move(subsection_buckets_)),
+      encrypted(is_encrypted_), key(key_), base_offset(base_offset_), ivfc_offset(ivfc_offset_),
+      section_ctr(section_ctr_) {
+    for (size_t i = 0; i < relocation.number_buckets - 1; ++i) {
+        relocation_buckets[i].entries.push_back({relocation.base_offsets[i + 1], 0, 0});
+    }
+
+    for (size_t i = 0; i < subsection.number_buckets - 1; ++i) {
+        subsection_buckets[i].entries.push_back({subsection_buckets[i + 1].entries[0].address_patch,
+                                                 {0},
+                                                 subsection_buckets[i + 1].entries[0].ctr});
+    }
+
+    relocation_buckets.back().entries.push_back({relocation.size, 0, 0});
+}
+
+BKTR::~BKTR() = default;
+
+size_t BKTR::Read(u8* data, size_t length, size_t offset) const {
+    // Read out of bounds.
+    if (offset >= relocation.size)
+        return 0;
+    const auto relocation = GetRelocationEntry(offset);
+    const auto section_offset = offset - relocation.address_patch + relocation.address_source;
+    const auto bktr_read = relocation.from_patch;
+
+    const auto next_relocation = GetNextRelocationEntry(offset);
+
+    if (offset + length > next_relocation.address_patch) {
+        const u64 partition = next_relocation.address_patch - offset;
+        return Read(data, partition, offset) +
+               Read(data + partition, length - partition, offset + partition);
+    }
+
+    if (!bktr_read) {
+        ASSERT_MSG(section_offset >= ivfc_offset, "Offset calculation negative.");
+        return base_romfs->Read(data, length, section_offset - ivfc_offset);
+    }
+
+    if (!encrypted) {
+        return bktr_romfs->Read(data, length, section_offset);
+    }
+
+    const auto subsection = GetSubsectionEntry(section_offset);
+    Core::Crypto::AESCipher<Core::Crypto::Key128> cipher(key, Core::Crypto::Mode::CTR);
+
+    // Calculate AES IV
+    std::vector<u8> iv(16);
+    auto subsection_ctr = subsection.ctr;
+    auto offset_iv = section_offset + base_offset;
+    for (size_t i = 0; i < section_ctr.size(); ++i)
+        iv[i] = section_ctr[0x8 - i - 1];
+    offset_iv >>= 4;
+    for (size_t i = 0; i < sizeof(u64); ++i) {
+        iv[0xF - i] = static_cast<u8>(offset_iv & 0xFF);
+        offset_iv >>= 8;
+    }
+    for (size_t i = 0; i < sizeof(u32); ++i) {
+        iv[0x7 - i] = static_cast<u8>(subsection_ctr & 0xFF);
+        subsection_ctr >>= 8;
+    }
+    cipher.SetIV(iv);
+
+    const auto next_subsection = GetNextSubsectionEntry(section_offset);
+
+    if (section_offset + length > next_subsection.address_patch) {
+        const u64 partition = next_subsection.address_patch - section_offset;
+        return Read(data, partition, offset) +
+               Read(data + partition, length - partition, offset + partition);
+    }
+
+    const auto block_offset = section_offset & 0xF;
+    if (block_offset != 0) {
+        auto block = bktr_romfs->ReadBytes(0x10, section_offset & ~0xF);
+        cipher.Transcode(block.data(), block.size(), block.data(), Core::Crypto::Op::Decrypt);
+        if (length + block_offset < 0x10) {
+            std::memcpy(data, block.data() + block_offset, std::min(length, block.size()));
+            return std::min(length, block.size());
+        }
+
+        const auto read = 0x10 - block_offset;
+        std::memcpy(data, block.data() + block_offset, read);
+        return read + Read(data + read, length - read, offset + read);
+    }
+
+    const auto raw_read = bktr_romfs->Read(data, length, section_offset);
+    cipher.Transcode(data, raw_read, data, Core::Crypto::Op::Decrypt);
+    return raw_read;
+}
+
+template <bool Subsection, typename BlockType, typename BucketType>
+std::pair<size_t, size_t> BKTR::SearchBucketEntry(u64 offset, BlockType block,
+                                                  BucketType buckets) const {
+    if constexpr (Subsection) {
+        const auto last_bucket = buckets[block.number_buckets - 1];
+        if (offset >= last_bucket.entries[last_bucket.number_entries].address_patch)
+            return {block.number_buckets - 1, last_bucket.number_entries};
+    } else {
+        ASSERT_MSG(offset <= block.size, "Offset is out of bounds in BKTR relocation block.");
+    }
+
+    size_t bucket_id = std::count_if(block.base_offsets.begin() + 1,
+                                     block.base_offsets.begin() + block.number_buckets,
+                                     [&offset](u64 base_offset) { return base_offset <= offset; });
+
+    const auto bucket = buckets[bucket_id];
+
+    if (bucket.number_entries == 1)
+        return {bucket_id, 0};
+
+    size_t low = 0;
+    size_t mid = 0;
+    size_t high = bucket.number_entries - 1;
+    while (low <= high) {
+        mid = (low + high) / 2;
+        if (bucket.entries[mid].address_patch > offset) {
+            high = mid - 1;
+        } else {
+            if (mid == bucket.number_entries - 1 ||
+                bucket.entries[mid + 1].address_patch > offset) {
+                return {bucket_id, mid};
+            }
+
+            low = mid + 1;
+        }
+    }
+
+    UNREACHABLE_MSG("Offset could not be found in BKTR block.");
+}
+
+RelocationEntry BKTR::GetRelocationEntry(u64 offset) const {
+    const auto res = SearchBucketEntry<false>(offset, relocation, relocation_buckets);
+    return relocation_buckets[res.first].entries[res.second];
+}
+
+RelocationEntry BKTR::GetNextRelocationEntry(u64 offset) const {
+    const auto res = SearchBucketEntry<false>(offset, relocation, relocation_buckets);
+    const auto bucket = relocation_buckets[res.first];
+    if (res.second + 1 < bucket.entries.size())
+        return bucket.entries[res.second + 1];
+    return relocation_buckets[res.first + 1].entries[0];
+}
+
+SubsectionEntry BKTR::GetSubsectionEntry(u64 offset) const {
+    const auto res = SearchBucketEntry<true>(offset, subsection, subsection_buckets);
+    return subsection_buckets[res.first].entries[res.second];
+}
+
+SubsectionEntry BKTR::GetNextSubsectionEntry(u64 offset) const {
+    const auto res = SearchBucketEntry<true>(offset, subsection, subsection_buckets);
+    const auto bucket = subsection_buckets[res.first];
+    if (res.second + 1 < bucket.entries.size())
+        return bucket.entries[res.second + 1];
+    return subsection_buckets[res.first + 1].entries[0];
+}
+
+std::string BKTR::GetName() const {
+    return base_romfs->GetName();
+}
+
+size_t BKTR::GetSize() const {
+    return relocation.size;
+}
+
+bool BKTR::Resize(size_t new_size) {
+    return false;
+}
+
+std::shared_ptr<VfsDirectory> BKTR::GetContainingDirectory() const {
+    return base_romfs->GetContainingDirectory();
+}
+
+bool BKTR::IsWritable() const {
+    return false;
+}
+
+bool BKTR::IsReadable() const {
+    return true;
+}
+
+size_t BKTR::Write(const u8* data, size_t length, size_t offset) {
+    return 0;
+}
+
+bool BKTR::Rename(std::string_view name) {
+    return base_romfs->Rename(name);
+}
+
+} // namespace FileSys

src/core/file_sys/nca_patch.h

@@ -0,0 +1,147 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <vector>
+#include <common/common_funcs.h>
+#include "core/crypto/key_manager.h"
+#include "core/file_sys/romfs.h"
+
+namespace FileSys {
+
+#pragma pack(push, 1)
+struct RelocationEntry {
+    u64_le address_patch;
+    u64_le address_source;
+    u32 from_patch;
+};
+#pragma pack(pop)
+static_assert(sizeof(RelocationEntry) == 0x14, "RelocationEntry has incorrect size.");
+
+struct RelocationBucketRaw {
+    INSERT_PADDING_BYTES(4);
+    u32_le number_entries;
+    u64_le end_offset;
+    std::array<RelocationEntry, 0x332> relocation_entries;
+    INSERT_PADDING_BYTES(8);
+};
+static_assert(sizeof(RelocationBucketRaw) == 0x4000, "RelocationBucketRaw has incorrect size.");
+
+// Vector version of RelocationBucketRaw
+struct RelocationBucket {
+    u32 number_entries;
+    u64 end_offset;
+    std::vector<RelocationEntry> entries;
+};
+
+struct RelocationBlock {
+    INSERT_PADDING_BYTES(4);
+    u32_le number_buckets;
+    u64_le size;
+    std::array<u64, 0x7FE> base_offsets;
+};
+static_assert(sizeof(RelocationBlock) == 0x4000, "RelocationBlock has incorrect size.");
+
+struct SubsectionEntry {
+    u64_le address_patch;
+    INSERT_PADDING_BYTES(0x4);
+    u32_le ctr;
+};
+static_assert(sizeof(SubsectionEntry) == 0x10, "SubsectionEntry has incorrect size.");
+
+struct SubsectionBucketRaw {
+    INSERT_PADDING_BYTES(4);
+    u32_le number_entries;
+    u64_le end_offset;
+    std::array<SubsectionEntry, 0x3FF> subsection_entries;
+};
+static_assert(sizeof(SubsectionBucketRaw) == 0x4000, "SubsectionBucketRaw has incorrect size.");
+
+// Vector version of SubsectionBucketRaw
+struct SubsectionBucket {
+    u32 number_entries;
+    u64 end_offset;
+    std::vector<SubsectionEntry> entries;
+};
+
+struct SubsectionBlock {
+    INSERT_PADDING_BYTES(4);
+    u32_le number_buckets;
+    u64_le size;
+    std::array<u64, 0x7FE> base_offsets;
+};
+static_assert(sizeof(SubsectionBlock) == 0x4000, "SubsectionBlock has incorrect size.");
+
+inline RelocationBucket ConvertRelocationBucketRaw(RelocationBucketRaw raw) {
+    return {raw.number_entries,
+            raw.end_offset,
+            {raw.relocation_entries.begin(), raw.relocation_entries.begin() + raw.number_entries}};
+}
+
+inline SubsectionBucket ConvertSubsectionBucketRaw(SubsectionBucketRaw raw) {
+    return {raw.number_entries,
+            raw.end_offset,
+            {raw.subsection_entries.begin(), raw.subsection_entries.begin() + raw.number_entries}};
+}
+
+class BKTR : public VfsFile {
+public:
+    BKTR(VirtualFile base_romfs, VirtualFile bktr_romfs, RelocationBlock relocation,
+         std::vector<RelocationBucket> relocation_buckets, SubsectionBlock subsection,
+         std::vector<SubsectionBucket> subsection_buckets, bool is_encrypted,
+         Core::Crypto::Key128 key, u64 base_offset, u64 ivfc_offset, std::array<u8, 8> section_ctr);
+    ~BKTR() override;
+
+    size_t Read(u8* data, size_t length, size_t offset) const override;
+
+    std::string GetName() const override;
+
+    size_t GetSize() const override;
+
+    bool Resize(size_t new_size) override;
+
+    std::shared_ptr<VfsDirectory> GetContainingDirectory() const override;
+
+    bool IsWritable() const override;
+
+    bool IsReadable() const override;
+
+    size_t Write(const u8* data, size_t length, size_t offset) override;
+
+    bool Rename(std::string_view name) override;
+
+private:
+    template <bool Subsection, typename BlockType, typename BucketType>
+    std::pair<size_t, size_t> SearchBucketEntry(u64 offset, BlockType block,
+                                                BucketType buckets) const;
+
+    RelocationEntry GetRelocationEntry(u64 offset) const;
+    RelocationEntry GetNextRelocationEntry(u64 offset) const;
+
+    SubsectionEntry GetSubsectionEntry(u64 offset) const;
+    SubsectionEntry GetNextSubsectionEntry(u64 offset) const;
+
+    RelocationBlock relocation;
+    std::vector<RelocationBucket> relocation_buckets;
+    SubsectionBlock subsection;
+    std::vector<SubsectionBucket> subsection_buckets;
+
+    // Should be the raw base romfs, decrypted.
+    VirtualFile base_romfs;
+    // Should be the raw BKTR romfs, (located at media_offset with size media_size).
+    VirtualFile bktr_romfs;
+
+    bool encrypted;
+    Core::Crypto::Key128 key;
+
+    // Base offset into NCA, used for IV calculation.
+    u64 base_offset;
+    // Distance between IVFC start and RomFS start, used for base reads
+    u64 ivfc_offset;
+    std::array<u8, 8> section_ctr;
+};
+
+} // namespace FileSys

src/core/file_sys/patch_manager.cpp

@@ -0,0 +1,153 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#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/hle/service/filesystem/filesystem.h"
+#include "core/loader/loader.h"
+
+namespace FileSys {
+
+constexpr u64 SINGLE_BYTE_MODULUS = 0x100;
+
+std::string FormatTitleVersion(u32 version, TitleVersionFormat format) {
+    std::array<u8, sizeof(u32)> bytes{};
+    bytes[0] = version % SINGLE_BYTE_MODULUS;
+    for (size_t i = 1; i < bytes.size(); ++i) {
+        version /= SINGLE_BYTE_MODULUS;
+        bytes[i] = version % SINGLE_BYTE_MODULUS;
+    }
+
+    if (format == TitleVersionFormat::FourElements)
+        return fmt::format("v{}.{}.{}.{}", bytes[3], bytes[2], bytes[1], bytes[0]);
+    return fmt::format("v{}.{}.{}", bytes[3], bytes[2], bytes[1]);
+}
+
+constexpr std::array<const char*, 1> PATCH_TYPE_NAMES{
+    "Update",
+};
+
+std::string FormatPatchTypeName(PatchType type) {
+    return PATCH_TYPE_NAMES.at(static_cast<size_t>(type));
+}
+
+PatchManager::PatchManager(u64 title_id) : title_id(title_id) {}
+
+VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
+    LOG_INFO(Loader, "Patching ExeFS for title_id={:016X}", title_id);
+
+    if (exefs == nullptr)
+        return exefs;
+
+    const auto installed = Service::FileSystem::GetUnionContents();
+
+    // Game Updates
+    const auto update_tid = GetUpdateTitleID(title_id);
+    const auto update = installed->GetEntry(update_tid, ContentRecordType::Program);
+    if (update != nullptr) {
+        if (update->GetStatus() == Loader::ResultStatus::ErrorMissingBKTRBaseRomFS &&
+            update->GetExeFS() != nullptr) {
+            LOG_INFO(Loader, "    ExeFS: Update ({}) applied successfully",
+                     FormatTitleVersion(installed->GetEntryVersion(update_tid).get_value_or(0)));
+            exefs = update->GetExeFS();
+        }
+    }
+
+    return exefs;
+}
+
+VirtualFile PatchManager::PatchRomFS(VirtualFile romfs, u64 ivfc_offset,
+                                     ContentRecordType type) const {
+    LOG_INFO(Loader, "Patching RomFS for title_id={:016X}, type={:02X}", title_id,
+             static_cast<u8>(type));
+
+    if (romfs == nullptr)
+        return romfs;
+
+    const auto installed = Service::FileSystem::GetUnionContents();
+
+    // Game Updates
+    const auto update_tid = GetUpdateTitleID(title_id);
+    const auto update = installed->GetEntryRaw(update_tid, type);
+    if (update != nullptr) {
+        const auto new_nca = std::make_shared<NCA>(update, romfs, ivfc_offset);
+        if (new_nca->GetStatus() == Loader::ResultStatus::Success &&
+            new_nca->GetRomFS() != nullptr) {
+            LOG_INFO(Loader, "    RomFS: Update ({}) applied successfully",
+                     FormatTitleVersion(installed->GetEntryVersion(update_tid).get_value_or(0)));
+            romfs = new_nca->GetRomFS();
+        }
+    }
+
+    return romfs;
+}
+
+std::map<PatchType, std::string> PatchManager::GetPatchVersionNames() const {
+    std::map<PatchType, std::string> out;
+    const auto installed = Service::FileSystem::GetUnionContents();
+
+    const auto update_tid = GetUpdateTitleID(title_id);
+    PatchManager update{update_tid};
+    auto [nacp, discard_icon_file] = update.GetControlMetadata();
+
+    if (nacp != nullptr) {
+        out[PatchType::Update] = nacp->GetVersionString();
+    } else {
+        if (installed->HasEntry(update_tid, ContentRecordType::Program)) {
+            const auto meta_ver = installed->GetEntryVersion(update_tid);
+            if (meta_ver == boost::none || meta_ver.get() == 0) {
+                out[PatchType::Update] = "";
+            } else {
+                out[PatchType::Update] =
+                    FormatTitleVersion(meta_ver.get(), TitleVersionFormat::ThreeElements);
+            }
+        }
+    }
+
+    return out;
+}
+
+std::pair<std::shared_ptr<NACP>, VirtualFile> PatchManager::GetControlMetadata() const {
+    const auto& installed{Service::FileSystem::GetUnionContents()};
+
+    const auto base_control_nca = installed->GetEntry(title_id, ContentRecordType::Control);
+    if (base_control_nca == nullptr)
+        return {};
+
+    return ParseControlNCA(base_control_nca);
+}
+
+std::pair<std::shared_ptr<NACP>, VirtualFile> PatchManager::ParseControlNCA(
+    const std::shared_ptr<NCA>& nca) const {
+    const auto base_romfs = nca->GetRomFS();
+    if (base_romfs == nullptr)
+        return {};
+
+    const auto romfs = PatchRomFS(base_romfs, nca->GetBaseIVFCOffset(), ContentRecordType::Control);
+    if (romfs == nullptr)
+        return {};
+
+    const auto extracted = ExtractRomFS(romfs);
+    if (extracted == nullptr)
+        return {};
+
+    auto nacp_file = extracted->GetFile("control.nacp");
+    if (nacp_file == nullptr)
+        nacp_file = extracted->GetFile("Control.nacp");
+
+    const auto nacp = nacp_file == nullptr ? nullptr : std::make_shared<NACP>(nacp_file);
+
+    VirtualFile icon_file;
+    for (const auto& language : FileSys::LANGUAGE_NAMES) {
+        icon_file = extracted->GetFile("icon_" + std::string(language) + ".dat");
+        if (icon_file != nullptr)
+            break;
+    }
+
+    return {nacp, icon_file};
+}
+} // namespace FileSys

src/core/file_sys/patch_manager.h

@@ -0,0 +1,62 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <map>
+#include <string>
+#include "common/common_types.h"
+#include "core/file_sys/nca_metadata.h"
+#include "core/file_sys/vfs.h"
+
+namespace FileSys {
+
+class NCA;
+class NACP;
+
+enum class TitleVersionFormat : u8 {
+    ThreeElements, ///< vX.Y.Z
+    FourElements,  ///< vX.Y.Z.W
+};
+
+std::string FormatTitleVersion(u32 version,
+                               TitleVersionFormat format = TitleVersionFormat::ThreeElements);
+
+enum class PatchType {
+    Update,
+};
+
+std::string FormatPatchTypeName(PatchType type);
+
+// A centralized class to manage patches to games.
+class PatchManager {
+public:
+    explicit PatchManager(u64 title_id);
+
+    // Currently tracked ExeFS patches:
+    // - Game Updates
+    VirtualDir PatchExeFS(VirtualDir exefs) const;
+
+    // Currently tracked RomFS patches:
+    // - Game Updates
+    VirtualFile PatchRomFS(VirtualFile base, u64 ivfc_offset,
+                           ContentRecordType type = ContentRecordType::Program) const;
+
+    // Returns a vector of pairs between patch names and patch versions.
+    // i.e. Update v80 will return {Update, 80}
+    std::map<PatchType, std::string> GetPatchVersionNames() const;
+
+    // Given title_id of the program, attempts to get the control data of the update and parse it,
+    // falling back to the base control data.
+    std::pair<std::shared_ptr<NACP>, VirtualFile> GetControlMetadata() const;
+
+    // Version of GetControlMetadata that takes an arbitrary NCA
+    std::pair<std::shared_ptr<NACP>, VirtualFile> ParseControlNCA(
+        const std::shared_ptr<NCA>& nca) const;
+
+private:
+    u64 title_id;
+};
+
+} // namespace FileSys

src/core/file_sys/program_metadata.cpp

@@ -2,7 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "common/file_util.h"
+#include <cstddef>
+#include <cstring>
+#include <vector>
+
 #include "common/logging/log.h"
 #include "core/file_sys/program_metadata.h"
 #include "core/loader/loader.h"

src/core/file_sys/program_metadata.h

@@ -5,12 +5,10 @@
 #pragma once
 
 #include <array>
-#include <string>
-#include <vector>
 #include "common/bit_field.h"
 #include "common/common_types.h"
 #include "common/swap.h"
-#include "partition_filesystem.h"
+#include "core/file_sys/vfs.h"
 
 namespace Loader {
 enum class ResultStatus : u16;

src/core/file_sys/registered_cache.cpp

@@ -5,13 +5,17 @@
 #include <regex>
 #include <mbedtls/sha256.h>
 #include "common/assert.h"
+#include "common/file_util.h"
 #include "common/hex_util.h"
 #include "common/logging/log.h"
-#include "core/crypto/encryption_layer.h"
+#include "core/crypto/key_manager.h"
 #include "core/file_sys/card_image.h"
+#include "core/file_sys/content_archive.h"
 #include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/registered_cache.h"
+#include "core/file_sys/submission_package.h"
 #include "core/file_sys/vfs_concat.h"
+#include "core/loader/loader.h"
 
 namespace FileSys {
 std::string RegisteredCacheEntry::DebugInfo() const {
@@ -276,6 +280,18 @@ VirtualFile RegisteredCache::GetEntryUnparsed(RegisteredCacheEntry entry) const
     return GetEntryUnparsed(entry.title_id, entry.type);
 }
 
+boost::optional<u32> RegisteredCache::GetEntryVersion(u64 title_id) const {
+    const auto meta_iter = meta.find(title_id);
+    if (meta_iter != meta.end())
+        return meta_iter->second.GetTitleVersion();
+
+    const auto yuzu_meta_iter = yuzu_meta.find(title_id);
+    if (yuzu_meta_iter != yuzu_meta.end())
+        return yuzu_meta_iter->second.GetTitleVersion();
+
+    return boost::none;
+}
+
 VirtualFile RegisteredCache::GetEntryRaw(u64 title_id, ContentRecordType type) const {
     const auto id = GetNcaIDFromMetadata(title_id, type);
     if (id == boost::none)
@@ -355,17 +371,21 @@ std::vector<RegisteredCacheEntry> RegisteredCache::ListEntriesFilter(
     return out;
 }
 
-static std::shared_ptr<NCA> GetNCAFromXCIForID(std::shared_ptr<XCI> xci, const NcaID& id) {
-    const auto filename = fmt::format("{}.nca", Common::HexArrayToString(id, false));
-    const auto iter =
-        std::find_if(xci->GetNCAs().begin(), xci->GetNCAs().end(),
-                     [&filename](std::shared_ptr<NCA> nca) { return nca->GetName() == filename; });
-    return iter == xci->GetNCAs().end() ? nullptr : *iter;
+static std::shared_ptr<NCA> GetNCAFromNSPForID(std::shared_ptr<NSP> nsp, const NcaID& id) {
+    const auto file = nsp->GetFile(fmt::format("{}.nca", Common::HexArrayToString(id, false)));
+    if (file == nullptr)
+        return nullptr;
+    return std::make_shared<NCA>(file);
 }
 
 InstallResult RegisteredCache::InstallEntry(std::shared_ptr<XCI> xci, bool overwrite_if_exists,
                                             const VfsCopyFunction& copy) {
-    const auto& ncas = xci->GetNCAs();
+    return InstallEntry(xci->GetSecurePartitionNSP(), overwrite_if_exists, copy);
+}
+
+InstallResult RegisteredCache::InstallEntry(std::shared_ptr<NSP> nsp, bool overwrite_if_exists,
+                                            const VfsCopyFunction& copy) {
+    const auto& ncas = nsp->GetNCAsCollapsed();
     const auto& meta_iter = std::find_if(ncas.begin(), ncas.end(), [](std::shared_ptr<NCA> nca) {
         return nca->GetType() == NCAContentType::Meta;
     });
@@ -389,7 +409,7 @@ InstallResult RegisteredCache::InstallEntry(std::shared_ptr<XCI> xci, bool overw
     const auto cnmt_file = section0->GetFiles()[0];
     const CNMT cnmt(cnmt_file);
     for (const auto& record : cnmt.GetContentRecords()) {
-        const auto nca = GetNCAFromXCIForID(xci, record.nca_id);
+        const auto nca = GetNCAFromNSPForID(nsp, record.nca_id);
         if (nca == nullptr)
             return InstallResult::ErrorCopyFailed;
         const auto res2 = RawInstallNCA(nca, copy, overwrite_if_exists, record.nca_id);
@@ -490,4 +510,107 @@ bool RegisteredCache::RawInstallYuzuMeta(const CNMT& cnmt) {
                                    kv.second.GetTitleID() == cnmt.GetTitleID();
                         }) != yuzu_meta.end();
 }
+
+RegisteredCacheUnion::RegisteredCacheUnion(std::vector<std::shared_ptr<RegisteredCache>> caches)
+    : caches(std::move(caches)) {}
+
+void RegisteredCacheUnion::Refresh() {
+    for (const auto& c : caches)
+        c->Refresh();
+}
+
+bool RegisteredCacheUnion::HasEntry(u64 title_id, ContentRecordType type) const {
+    return std::any_of(caches.begin(), caches.end(), [title_id, type](const auto& cache) {
+        return cache->HasEntry(title_id, type);
+    });
+}
+
+bool RegisteredCacheUnion::HasEntry(RegisteredCacheEntry entry) const {
+    return HasEntry(entry.title_id, entry.type);
+}
+
+boost::optional<u32> RegisteredCacheUnion::GetEntryVersion(u64 title_id) const {
+    for (const auto& c : caches) {
+        const auto res = c->GetEntryVersion(title_id);
+        if (res != boost::none)
+            return res;
+    }
+
+    return boost::none;
+}
+
+VirtualFile RegisteredCacheUnion::GetEntryUnparsed(u64 title_id, ContentRecordType type) const {
+    for (const auto& c : caches) {
+        const auto res = c->GetEntryUnparsed(title_id, type);
+        if (res != nullptr)
+            return res;
+    }
+
+    return nullptr;
+}
+
+VirtualFile RegisteredCacheUnion::GetEntryUnparsed(RegisteredCacheEntry entry) const {
+    return GetEntryUnparsed(entry.title_id, entry.type);
+}
+
+VirtualFile RegisteredCacheUnion::GetEntryRaw(u64 title_id, ContentRecordType type) const {
+    for (const auto& c : caches) {
+        const auto res = c->GetEntryRaw(title_id, type);
+        if (res != nullptr)
+            return res;
+    }
+
+    return nullptr;
+}
+
+VirtualFile RegisteredCacheUnion::GetEntryRaw(RegisteredCacheEntry entry) const {
+    return GetEntryRaw(entry.title_id, entry.type);
+}
+
+std::shared_ptr<NCA> RegisteredCacheUnion::GetEntry(u64 title_id, ContentRecordType type) const {
+    const auto raw = GetEntryRaw(title_id, type);
+    if (raw == nullptr)
+        return nullptr;
+    return std::make_shared<NCA>(raw);
+}
+
+std::shared_ptr<NCA> RegisteredCacheUnion::GetEntry(RegisteredCacheEntry entry) const {
+    return GetEntry(entry.title_id, entry.type);
+}
+
+std::vector<RegisteredCacheEntry> RegisteredCacheUnion::ListEntries() const {
+    std::vector<RegisteredCacheEntry> out;
+    for (const auto& c : caches) {
+        c->IterateAllMetadata<RegisteredCacheEntry>(
+            out,
+            [](const CNMT& c, const ContentRecord& r) {
+                return RegisteredCacheEntry{c.GetTitleID(), r.type};
+            },
+            [](const CNMT& c, const ContentRecord& r) { return true; });
+    }
+    return out;
+}
+
+std::vector<RegisteredCacheEntry> RegisteredCacheUnion::ListEntriesFilter(
+    boost::optional<TitleType> title_type, boost::optional<ContentRecordType> record_type,
+    boost::optional<u64> title_id) const {
+    std::vector<RegisteredCacheEntry> out;
+    for (const auto& c : caches) {
+        c->IterateAllMetadata<RegisteredCacheEntry>(
+            out,
+            [](const CNMT& c, const ContentRecord& r) {
+                return RegisteredCacheEntry{c.GetTitleID(), r.type};
+            },
+            [&title_type, &record_type, &title_id](const CNMT& c, const ContentRecord& r) {
+                if (title_type != boost::none && title_type.get() != c.GetType())
+                    return false;
+                if (record_type != boost::none && record_type.get() != r.type)
+                    return false;
+                if (title_id != boost::none && title_id.get() != c.GetTitleID())
+                    return false;
+                return true;
+            });
+    }
+    return out;
+}
 } // namespace FileSys

src/core/file_sys/registered_cache.h

@@ -11,15 +11,19 @@
 #include <string>
 #include <vector>
 #include <boost/container/flat_map.hpp>
-#include "common/common_funcs.h"
 #include "common/common_types.h"
-#include "content_archive.h"
-#include "core/file_sys/nca_metadata.h"
 #include "core/file_sys/vfs.h"
 
 namespace FileSys {
-class XCI;
 class CNMT;
+class NCA;
+class NSP;
+class XCI;
+
+enum class ContentRecordType : u8;
+enum class TitleType : u8;
+
+struct ContentRecord;
 
 using NcaID = std::array<u8, 0x10>;
 using RegisteredCacheParsingFunction = std::function<VirtualFile(const VirtualFile&, const NcaID&)>;
@@ -39,6 +43,10 @@ struct RegisteredCacheEntry {
     std::string DebugInfo() const;
 };
 
+constexpr u64 GetUpdateTitleID(u64 base_title_id) {
+    return base_title_id | 0x800;
+}
+
 // boost flat_map requires operator< for O(log(n)) lookups.
 bool operator<(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs);
 
@@ -56,6 +64,8 @@ bool operator<(const RegisteredCacheEntry& lhs, const RegisteredCacheEntry& rhs)
  * 4GB splitting can be ignored.)
  */
 class RegisteredCache {
+    friend class RegisteredCacheUnion;
+
 public:
     // Parsing function defines the conversion from raw file to NCA. If there are other steps
     // besides creating the NCA from the file (e.g. NAX0 on SD Card), that should go in a custom
@@ -70,6 +80,8 @@ public:
     bool HasEntry(u64 title_id, ContentRecordType type) const;
     bool HasEntry(RegisteredCacheEntry entry) const;
 
+    boost::optional<u32> GetEntryVersion(u64 title_id) const;
+
     VirtualFile GetEntryUnparsed(u64 title_id, ContentRecordType type) const;
     VirtualFile GetEntryUnparsed(RegisteredCacheEntry entry) const;
 
@@ -86,10 +98,12 @@ public:
         boost::optional<ContentRecordType> record_type = boost::none,
         boost::optional<u64> title_id = boost::none) const;
 
-    // Raw copies all the ncas from the xci to the csache. Does some quick checks to make sure there
-    // is a meta NCA and all of them are accessible.
+    // Raw copies all the ncas from the xci/nsp to the csache. Does some quick checks to make sure
+    // there is a meta NCA and all of them are accessible.
     InstallResult InstallEntry(std::shared_ptr<XCI> xci, bool overwrite_if_exists = false,
                                const VfsCopyFunction& copy = &VfsRawCopy);
+    InstallResult InstallEntry(std::shared_ptr<NSP> nsp, bool overwrite_if_exists = false,
+                               const VfsCopyFunction& copy = &VfsRawCopy);
 
     // Due to the fact that we must use Meta-type NCAs to determine the existance of files, this
     // poses quite a challenge. Instead of creating a new meta NCA for this file, yuzu will create a
@@ -125,4 +139,36 @@ private:
     boost::container::flat_map<u64, CNMT> yuzu_meta;
 };
 
+// Combines multiple RegisteredCaches (i.e. SysNAND, UserNAND, SDMC) into one interface.
+class RegisteredCacheUnion {
+public:
+    explicit RegisteredCacheUnion(std::vector<std::shared_ptr<RegisteredCache>> caches);
+
+    void Refresh();
+
+    bool HasEntry(u64 title_id, ContentRecordType type) const;
+    bool HasEntry(RegisteredCacheEntry entry) const;
+
+    boost::optional<u32> GetEntryVersion(u64 title_id) const;
+
+    VirtualFile GetEntryUnparsed(u64 title_id, ContentRecordType type) const;
+    VirtualFile GetEntryUnparsed(RegisteredCacheEntry entry) const;
+
+    VirtualFile GetEntryRaw(u64 title_id, ContentRecordType type) const;
+    VirtualFile GetEntryRaw(RegisteredCacheEntry entry) const;
+
+    std::shared_ptr<NCA> GetEntry(u64 title_id, ContentRecordType type) const;
+    std::shared_ptr<NCA> GetEntry(RegisteredCacheEntry entry) const;
+
+    std::vector<RegisteredCacheEntry> ListEntries() const;
+    // If a parameter is not boost::none, it will be filtered for from all entries.
+    std::vector<RegisteredCacheEntry> ListEntriesFilter(
+        boost::optional<TitleType> title_type = boost::none,
+        boost::optional<ContentRecordType> record_type = boost::none,
+        boost::optional<u64> title_id = boost::none) const;
+
+private:
+    std::vector<std::shared_ptr<RegisteredCache>> caches;
+};
+
 } // namespace FileSys

src/core/file_sys/romfs.h

@@ -6,6 +6,7 @@
 
 #include <array>
 #include "common/common_funcs.h"
+#include "common/common_types.h"
 #include "common/swap.h"
 #include "core/file_sys/vfs.h"
 

src/core/file_sys/romfs_factory.cpp

@@ -2,11 +2,14 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <algorithm>
 #include <memory>
+#include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
+#include "core/core.h"
+#include "core/file_sys/content_archive.h"
 #include "core/file_sys/nca_metadata.h"
+#include "core/file_sys/patch_manager.h"
 #include "core/file_sys/registered_cache.h"
 #include "core/file_sys/romfs_factory.h"
 #include "core/hle/kernel/process.h"
@@ -20,10 +23,17 @@ RomFSFactory::RomFSFactory(Loader::AppLoader& app_loader) {
     if (app_loader.ReadRomFS(file) != Loader::ResultStatus::Success) {
         LOG_ERROR(Service_FS, "Unable to read RomFS!");
     }
+
+    updatable = app_loader.IsRomFSUpdatable();
+    ivfc_offset = app_loader.ReadRomFSIVFCOffset();
 }
 
 ResultVal<VirtualFile> RomFSFactory::OpenCurrentProcess() {
-    return MakeResult<VirtualFile>(file);
+    if (!updatable)
+        return MakeResult<VirtualFile>(file);
+
+    const PatchManager patch_manager(Core::CurrentProcess()->program_id);
+    return MakeResult<VirtualFile>(patch_manager.PatchRomFS(file, ivfc_offset));
 }
 
 ResultVal<VirtualFile> RomFSFactory::Open(u64 title_id, StorageId storage, ContentRecordType type) {

src/core/file_sys/romfs_factory.h

@@ -36,6 +36,8 @@ public:
 
 private:
     VirtualFile file;
+    bool updatable;
+    u64 ivfc_offset;
 };
 
 } // namespace FileSys

src/core/file_sys/savedata_factory.cpp

@@ -3,10 +3,12 @@
 // Refer to the license.txt file included.
 
 #include <memory>
+#include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/file_sys/savedata_factory.h"
+#include "core/file_sys/vfs.h"
 #include "core/hle/kernel/process.h"
 
 namespace FileSys {

src/core/file_sys/savedata_factory.h

@@ -6,8 +6,10 @@
 
 #include <memory>
 #include <string>
+#include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/swap.h"
+#include "core/file_sys/vfs.h"
 #include "core/hle/result.h"
 
 namespace FileSys {

src/core/file_sys/submission_package.cpp

@@ -0,0 +1,245 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstring>
+#include <string_view>
+
+#include <fmt/ostream.h>
+
+#include "common/hex_util.h"
+#include "common/logging/log.h"
+#include "core/crypto/key_manager.h"
+#include "core/file_sys/content_archive.h"
+#include "core/file_sys/nca_metadata.h"
+#include "core/file_sys/partition_filesystem.h"
+#include "core/file_sys/submission_package.h"
+#include "core/loader/loader.h"
+
+namespace FileSys {
+NSP::NSP(VirtualFile file_)
+    : file(std::move(file_)), status{Loader::ResultStatus::Success},
+      pfs(std::make_shared<PartitionFilesystem>(file)) {
+    if (pfs->GetStatus() != Loader::ResultStatus::Success) {
+        status = pfs->GetStatus();
+        return;
+    }
+
+    if (IsDirectoryExeFS(pfs)) {
+        extracted = true;
+        exefs = pfs;
+
+        const auto& files = pfs->GetFiles();
+        const auto romfs_iter =
+            std::find_if(files.begin(), files.end(), [](const FileSys::VirtualFile& file) {
+                return file->GetName().find(".romfs") != std::string::npos;
+            });
+        if (romfs_iter != files.end())
+            romfs = *romfs_iter;
+        return;
+    }
+
+    extracted = false;
+    const auto files = pfs->GetFiles();
+
+    Core::Crypto::KeyManager keys;
+    for (const auto& ticket_file : files) {
+        if (ticket_file->GetExtension() == "tik") {
+            if (ticket_file == nullptr ||
+                ticket_file->GetSize() <
+                    Core::Crypto::TICKET_FILE_TITLEKEY_OFFSET + sizeof(Core::Crypto::Key128)) {
+                continue;
+            }
+
+            Core::Crypto::Key128 key{};
+            ticket_file->Read(key.data(), key.size(), Core::Crypto::TICKET_FILE_TITLEKEY_OFFSET);
+            std::string_view name_only(ticket_file->GetName());
+            name_only.remove_suffix(4);
+            const auto rights_id_raw = Common::HexStringToArray<16>(name_only);
+            u128 rights_id;
+            std::memcpy(rights_id.data(), rights_id_raw.data(), sizeof(u128));
+            keys.SetKey(Core::Crypto::S128KeyType::Titlekey, key, rights_id[1], rights_id[0]);
+        }
+    }
+
+    for (const auto& outer_file : files) {
+        if (outer_file->GetName().substr(outer_file->GetName().size() - 9) == ".cnmt.nca") {
+            const auto nca = std::make_shared<NCA>(outer_file);
+            if (nca->GetStatus() != Loader::ResultStatus::Success) {
+                program_status[nca->GetTitleId()] = nca->GetStatus();
+                continue;
+            }
+
+            const auto section0 = nca->GetSubdirectories()[0];
+
+            for (const auto& inner_file : section0->GetFiles()) {
+                if (inner_file->GetExtension() != "cnmt")
+                    continue;
+
+                const CNMT cnmt(inner_file);
+                auto& ncas_title = ncas[cnmt.GetTitleID()];
+
+                ncas_title[ContentRecordType::Meta] = nca;
+                for (const auto& rec : cnmt.GetContentRecords()) {
+                    const auto id_string = Common::HexArrayToString(rec.nca_id, false);
+                    const auto next_file = pfs->GetFile(fmt::format("{}.nca", id_string));
+                    if (next_file == nullptr) {
+                        LOG_WARNING(Service_FS,
+                                    "NCA with ID {}.nca is listed in content metadata, but cannot "
+                                    "be found in PFS. NSP appears to be corrupted.",
+                                    id_string);
+                        continue;
+                    }
+
+                    auto next_nca = std::make_shared<NCA>(next_file);
+                    if (next_nca->GetType() == NCAContentType::Program)
+                        program_status[cnmt.GetTitleID()] = next_nca->GetStatus();
+                    if (next_nca->GetStatus() == Loader::ResultStatus::Success)
+                        ncas_title[rec.type] = std::move(next_nca);
+                }
+
+                break;
+            }
+        }
+    }
+}
+
+NSP::~NSP() = default;
+
+Loader::ResultStatus NSP::GetStatus() const {
+    return status;
+}
+
+Loader::ResultStatus NSP::GetProgramStatus(u64 title_id) const {
+    const auto iter = program_status.find(title_id);
+    if (iter == program_status.end())
+        return Loader::ResultStatus::ErrorNSPMissingProgramNCA;
+    return iter->second;
+}
+
+u64 NSP::GetFirstTitleID() const {
+    if (program_status.empty())
+        return 0;
+    return program_status.begin()->first;
+}
+
+u64 NSP::GetProgramTitleID() const {
+    const auto out = GetFirstTitleID();
+    if ((out & 0x800) == 0)
+        return out;
+
+    const auto ids = GetTitleIDs();
+    const auto iter =
+        std::find_if(ids.begin(), ids.end(), [](u64 tid) { return (tid & 0x800) == 0; });
+    return iter == ids.end() ? out : *iter;
+}
+
+std::vector<u64> NSP::GetTitleIDs() const {
+    std::vector<u64> out;
+    out.reserve(ncas.size());
+    for (const auto& kv : ncas)
+        out.push_back(kv.first);
+    return out;
+}
+
+bool NSP::IsExtractedType() const {
+    return extracted;
+}
+
+VirtualFile NSP::GetRomFS() const {
+    return romfs;
+}
+
+VirtualDir NSP::GetExeFS() const {
+    return exefs;
+}
+
+std::vector<std::shared_ptr<NCA>> NSP::GetNCAsCollapsed() const {
+    if (extracted)
+        LOG_WARNING(Service_FS, "called on an NSP that is of type extracted.");
+    std::vector<std::shared_ptr<NCA>> out;
+    for (const auto& map : ncas) {
+        for (const auto& inner_map : map.second)
+            out.push_back(inner_map.second);
+    }
+    return out;
+}
+
+std::multimap<u64, std::shared_ptr<NCA>> NSP::GetNCAsByTitleID() const {
+    if (extracted)
+        LOG_WARNING(Service_FS, "called on an NSP that is of type extracted.");
+    std::multimap<u64, std::shared_ptr<NCA>> out;
+    for (const auto& map : ncas) {
+        for (const auto& inner_map : map.second)
+            out.emplace(map.first, inner_map.second);
+    }
+    return out;
+}
+
+std::map<u64, std::map<ContentRecordType, std::shared_ptr<NCA>>> NSP::GetNCAs() const {
+    return ncas;
+}
+
+std::shared_ptr<NCA> NSP::GetNCA(u64 title_id, ContentRecordType type) const {
+    if (extracted)
+        LOG_WARNING(Service_FS, "called on an NSP that is of type extracted.");
+
+    const auto title_id_iter = ncas.find(title_id);
+    if (title_id_iter == ncas.end())
+        return nullptr;
+
+    const auto type_iter = title_id_iter->second.find(type);
+    if (type_iter == title_id_iter->second.end())
+        return nullptr;
+
+    return type_iter->second;
+}
+
+VirtualFile NSP::GetNCAFile(u64 title_id, ContentRecordType type) const {
+    if (extracted)
+        LOG_WARNING(Service_FS, "called on an NSP that is of type extracted.");
+    const auto nca = GetNCA(title_id, type);
+    if (nca != nullptr)
+        return nca->GetBaseFile();
+    return nullptr;
+}
+
+std::vector<Core::Crypto::Key128> NSP::GetTitlekey() const {
+    if (extracted)
+        LOG_WARNING(Service_FS, "called on an NSP that is of type extracted.");
+    std::vector<Core::Crypto::Key128> out;
+    for (const auto& ticket_file : ticket_files) {
+        if (ticket_file == nullptr ||
+            ticket_file->GetSize() <
+                Core::Crypto::TICKET_FILE_TITLEKEY_OFFSET + sizeof(Core::Crypto::Key128)) {
+            continue;
+        }
+
+        out.emplace_back();
+        ticket_file->Read(out.back().data(), out.back().size(),
+                          Core::Crypto::TICKET_FILE_TITLEKEY_OFFSET);
+    }
+    return out;
+}
+
+std::vector<VirtualFile> NSP::GetFiles() const {
+    return pfs->GetFiles();
+}
+
+std::vector<VirtualDir> NSP::GetSubdirectories() const {
+    return pfs->GetSubdirectories();
+}
+
+std::string NSP::GetName() const {
+    return file->GetName();
+}
+
+VirtualDir NSP::GetParentDirectory() const {
+    return file->GetContainingDirectory();
+}
+
+bool NSP::ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) {
+    return false;
+}
+} // namespace FileSys

src/core/file_sys/submission_package.h

@@ -0,0 +1,76 @@
+// 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 <vector>
+#include "common/common_types.h"
+#include "core/file_sys/vfs.h"
+
+namespace Loader {
+enum class ResultStatus : u16;
+}
+
+namespace FileSys {
+
+class NCA;
+class PartitionFilesystem;
+
+enum class ContentRecordType : u8;
+
+class NSP : public ReadOnlyVfsDirectory {
+public:
+    explicit NSP(VirtualFile file);
+    ~NSP();
+
+    Loader::ResultStatus GetStatus() const;
+    Loader::ResultStatus GetProgramStatus(u64 title_id) const;
+    // Should only be used when one title id can be assured.
+    u64 GetFirstTitleID() const;
+    u64 GetProgramTitleID() const;
+    std::vector<u64> GetTitleIDs() const;
+
+    bool IsExtractedType() const;
+
+    // Common (Can be safely called on both types)
+    VirtualFile GetRomFS() const;
+    VirtualDir GetExeFS() const;
+
+    // Type 0 Only (Collection of NCAs + Certificate + Ticket + Meta XML)
+    std::vector<std::shared_ptr<NCA>> GetNCAsCollapsed() const;
+    std::multimap<u64, std::shared_ptr<NCA>> GetNCAsByTitleID() const;
+    std::map<u64, std::map<ContentRecordType, std::shared_ptr<NCA>>> GetNCAs() const;
+    std::shared_ptr<NCA> GetNCA(u64 title_id, ContentRecordType type) const;
+    VirtualFile GetNCAFile(u64 title_id, ContentRecordType type) const;
+    std::vector<Core::Crypto::Key128> GetTitlekey() const;
+
+    std::vector<VirtualFile> GetFiles() const override;
+
+    std::vector<VirtualDir> GetSubdirectories() const override;
+
+    std::string GetName() const override;
+
+    VirtualDir GetParentDirectory() const override;
+
+protected:
+    bool ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) override;
+
+private:
+    VirtualFile file;
+
+    bool extracted;
+    Loader::ResultStatus status;
+    std::map<u64, Loader::ResultStatus> program_status;
+
+    std::shared_ptr<PartitionFilesystem> pfs;
+    // Map title id -> {map type -> NCA}
+    std::map<u64, std::map<ContentRecordType, std::shared_ptr<NCA>>> ncas;
+    std::vector<VirtualFile> ticket_files;
+
+    VirtualFile romfs;
+    VirtualDir exefs;
+};
+} // namespace FileSys

src/core/file_sys/vfs_real.cpp

@@ -8,6 +8,7 @@
 #include <utility>
 #include "common/assert.h"
 #include "common/common_paths.h"
+#include "common/file_util.h"
 #include "common/logging/log.h"
 #include "core/file_sys/vfs_real.h"
 
@@ -39,6 +40,7 @@ static std::string ModeFlagsToString(Mode mode) {
 }
 
 RealVfsFilesystem::RealVfsFilesystem() : VfsFilesystem(nullptr) {}
+RealVfsFilesystem::~RealVfsFilesystem() = default;
 
 std::string RealVfsFilesystem::GetName() const {
     return "Real";
@@ -219,6 +221,8 @@ RealVfsFile::RealVfsFile(RealVfsFilesystem& base_, std::shared_ptr<FileUtil::IOF
       parent_components(FileUtil::SliceVector(path_components, 0, path_components.size() - 1)),
       perms(perms_) {}
 
+RealVfsFile::~RealVfsFile() = default;
+
 std::string RealVfsFile::GetName() const {
     return path_components.back();
 }
@@ -312,6 +316,8 @@ RealVfsDirectory::RealVfsDirectory(RealVfsFilesystem& base_, const std::string&
         FileUtil::CreateDir(path);
 }
 
+RealVfsDirectory::~RealVfsDirectory() = default;
+
 std::shared_ptr<VfsFile> RealVfsDirectory::GetFileRelative(std::string_view path) const {
     const auto full_path = FileUtil::SanitizePath(this->path + DIR_SEP + std::string(path));
     if (!FileUtil::Exists(full_path) || FileUtil::IsDirectory(full_path))

src/core/file_sys/vfs_real.h

@@ -6,15 +6,19 @@
 
 #include <string_view>
 #include <boost/container/flat_map.hpp>
-#include "common/file_util.h"
 #include "core/file_sys/mode.h"
 #include "core/file_sys/vfs.h"
 
+namespace FileUtil {
+class IOFile;
+}
+
 namespace FileSys {
 
 class RealVfsFilesystem : public VfsFilesystem {
 public:
     RealVfsFilesystem();
+    ~RealVfsFilesystem() override;
 
     std::string GetName() const override;
     bool IsReadable() const override;
@@ -40,10 +44,9 @@ class RealVfsFile : public VfsFile {
     friend class RealVfsDirectory;
     friend class RealVfsFilesystem;
 
-    RealVfsFile(RealVfsFilesystem& base, std::shared_ptr<FileUtil::IOFile> backing,
-                const std::string& path, Mode perms = Mode::Read);
-
 public:
+    ~RealVfsFile() override;
+
     std::string GetName() const override;
     size_t GetSize() const override;
     bool Resize(size_t new_size) override;
@@ -55,6 +58,9 @@ public:
     bool Rename(std::string_view name) override;
 
 private:
+    RealVfsFile(RealVfsFilesystem& base, std::shared_ptr<FileUtil::IOFile> backing,
+                const std::string& path, Mode perms = Mode::Read);
+
     bool Close();
 
     RealVfsFilesystem& base;
@@ -70,9 +76,9 @@ private:
 class RealVfsDirectory : public VfsDirectory {
     friend class RealVfsFilesystem;
 
-    RealVfsDirectory(RealVfsFilesystem& base, const std::string& path, Mode perms = Mode::Read);
-
 public:
+    ~RealVfsDirectory() override;
+
     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;
@@ -97,6 +103,8 @@ protected:
     bool ReplaceFileWithSubdirectory(VirtualFile file, VirtualDir dir) override;
 
 private:
+    RealVfsDirectory(RealVfsFilesystem& base, const std::string& path, Mode perms = Mode::Read);
+
     template <typename T, typename R>
     std::vector<std::shared_ptr<R>> IterateEntries() const;
 

src/core/file_sys/xts_archive.cpp

@@ -10,6 +10,7 @@
 #include <mbedtls/md.h>
 #include <mbedtls/sha256.h>
 #include "common/assert.h"
+#include "common/file_util.h"
 #include "common/hex_util.h"
 #include "common/logging/log.h"
 #include "core/crypto/aes_util.h"

src/core/hle/kernel/address_arbiter.cpp

@@ -8,9 +8,11 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "core/core.h"
+#include "core/core_cpu.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
+#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/result.h"
 #include "core/memory.h"

src/core/hle/kernel/hle_ipc.cpp

@@ -18,6 +18,7 @@
 #include "core/hle/kernel/event.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/server_session.h"

src/core/hle/kernel/kernel.cpp

@@ -13,6 +13,7 @@
 
 #include "core/core.h"
 #include "core/core_timing.h"
+#include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
@@ -124,6 +125,8 @@ struct KernelCore::Impl {
 
         timer_callback_handle_table.Clear();
         timer_callback_event_type = nullptr;
+
+        named_ports.clear();
     }
 
     void InitializeResourceLimits(KernelCore& kernel) {
@@ -217,6 +220,10 @@ struct KernelCore::Impl {
     // TODO(yuriks): This can be removed if Thread objects are explicitly pooled in the future,
     // allowing us to simply use a pool index or similar.
     Kernel::HandleTable thread_wakeup_callback_handle_table;
+
+    /// Map of named ports managed by the kernel, which can be retrieved using
+    /// the ConnectToPort SVC.
+    NamedPortTable named_ports;
 };
 
 KernelCore::KernelCore() : impl{std::make_unique<Impl>()} {}
@@ -257,6 +264,23 @@ void KernelCore::AppendNewProcess(SharedPtr<Process> process) {
     impl->process_list.push_back(std::move(process));
 }
 
+void KernelCore::AddNamedPort(std::string name, SharedPtr<ClientPort> port) {
+    impl->named_ports.emplace(std::move(name), std::move(port));
+}
+
+KernelCore::NamedPortTable::iterator KernelCore::FindNamedPort(const std::string& name) {
+    return impl->named_ports.find(name);
+}
+
+KernelCore::NamedPortTable::const_iterator KernelCore::FindNamedPort(
+    const std::string& name) const {
+    return impl->named_ports.find(name);
+}
+
+bool KernelCore::IsValidNamedPort(NamedPortTable::const_iterator port) const {
+    return port != impl->named_ports.cend();
+}
+
 u32 KernelCore::CreateNewObjectID() {
     return impl->next_object_id++;
 }

src/core/hle/kernel/kernel.h

@@ -4,6 +4,8 @@
 
 #pragma once
 
+#include <string>
+#include <unordered_map>
 #include "core/hle/kernel/object.h"
 
 template <typename T>
@@ -15,6 +17,7 @@ struct EventType;
 
 namespace Kernel {
 
+class ClientPort;
 class HandleTable;
 class Process;
 class ResourceLimit;
@@ -25,6 +28,9 @@ enum class ResourceLimitCategory : u8;
 
 /// Represents a single instance of the kernel.
 class KernelCore {
+private:
+    using NamedPortTable = std::unordered_map<std::string, SharedPtr<ClientPort>>;
+
 public:
     KernelCore();
     ~KernelCore();
@@ -59,6 +65,18 @@ public:
     /// Adds the given shared pointer to an internal list of active processes.
     void AppendNewProcess(SharedPtr<Process> process);
 
+    /// Adds a port to the named port table
+    void AddNamedPort(std::string name, SharedPtr<ClientPort> port);
+
+    /// Finds a port within the named port table with the given name.
+    NamedPortTable::iterator FindNamedPort(const std::string& name);
+
+    /// Finds a port within the named port table with the given name.
+    NamedPortTable::const_iterator FindNamedPort(const std::string& name) const;
+
+    /// Determines whether or not the given port is a valid named port.
+    bool IsValidNamedPort(NamedPortTable::const_iterator port) const;
+
 private:
     friend class Object;
     friend class Process;

src/core/hle/kernel/server_session.cpp

@@ -13,6 +13,7 @@
 #include "core/hle/kernel/client_session.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/server_session.h"
 #include "core/hle/kernel/session.h"
@@ -104,11 +105,10 @@ ResultCode ServerSession::HandleSyncRequest(SharedPtr<Thread> thread) {
     // The ServerSession received a sync request, this means that there's new data available
     // from its ClientSession, so wake up any threads that may be waiting on a svcReplyAndReceive or
     // similar.
-
-    auto& handle_table = Core::System::GetInstance().Kernel().HandleTable();
     Kernel::HLERequestContext context(this);
     u32* cmd_buf = (u32*)Memory::GetPointer(thread->GetTLSAddress());
-    context.PopulateFromIncomingCommandBuffer(cmd_buf, *Core::CurrentProcess(), handle_table);
+    context.PopulateFromIncomingCommandBuffer(cmd_buf, *Core::CurrentProcess(),
+                                              kernel.HandleTable());
 
     ResultCode result = RESULT_SUCCESS;
     // If the session has been converted to a domain, handle the domain request

src/core/hle/kernel/svc.cpp

@@ -12,16 +12,20 @@
 #include "common/logging/log.h"
 #include "common/microprofile.h"
 #include "common/string_util.h"
+#include "core/arm/exclusive_monitor.h"
 #include "core/core.h"
+#include "core/core_cpu.h"
 #include "core/core_timing.h"
 #include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/client_session.h"
 #include "core/hle/kernel/event.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/mutex.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
+#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/shared_memory.h"
 #include "core/hle/kernel/svc.h"
 #include "core/hle/kernel/svc_wrap.h"
@@ -64,19 +68,22 @@ static ResultCode UnmapMemory(VAddr dst_addr, VAddr src_addr, u64 size) {
 
 /// Connect to an OS service given the port name, returns the handle to the port to out
 static ResultCode ConnectToNamedPort(Handle* out_handle, VAddr port_name_address) {
-    if (!Memory::IsValidVirtualAddress(port_name_address))
+    if (!Memory::IsValidVirtualAddress(port_name_address)) {
         return ERR_NOT_FOUND;
+    }
 
     static constexpr std::size_t PortNameMaxLength = 11;
     // Read 1 char beyond the max allowed port name to detect names that are too long.
     std::string port_name = Memory::ReadCString(port_name_address, PortNameMaxLength + 1);
-    if (port_name.size() > PortNameMaxLength)
+    if (port_name.size() > PortNameMaxLength) {
         return ERR_PORT_NAME_TOO_LONG;
+    }
 
     LOG_TRACE(Kernel_SVC, "called port_name={}", port_name);
 
-    auto it = Service::g_kernel_named_ports.find(port_name);
-    if (it == Service::g_kernel_named_ports.end()) {
+    auto& kernel = Core::System::GetInstance().Kernel();
+    auto it = kernel.FindNamedPort(port_name);
+    if (!kernel.IsValidNamedPort(it)) {
         LOG_WARNING(Kernel_SVC, "tried to connect to unknown port: {}", port_name);
         return ERR_NOT_FOUND;
     }
@@ -87,7 +94,6 @@ static ResultCode ConnectToNamedPort(Handle* out_handle, VAddr port_name_address
     CASCADE_RESULT(client_session, client_port->Connect());
 
     // Return the client session
-    auto& kernel = Core::System::GetInstance().Kernel();
     CASCADE_RESULT(*out_handle, kernel.HandleTable().Create(client_session));
     return RESULT_SUCCESS;
 }

src/core/hle/kernel/thread.cpp

@@ -16,6 +16,7 @@
 #include "common/thread_queue_list.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
+#include "core/core_cpu.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/hle/kernel/errors.h"
@@ -23,8 +24,8 @@
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
+#include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
-#include "core/hle/lock.h"
 #include "core/hle/result.h"
 #include "core/memory.h"
 

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

@@ -10,6 +10,7 @@
 #include "core/file_sys/bis_factory.h"
 #include "core/file_sys/errors.h"
 #include "core/file_sys/mode.h"
+#include "core/file_sys/registered_cache.h"
 #include "core/file_sys/romfs_factory.h"
 #include "core/file_sys/savedata_factory.h"
 #include "core/file_sys/sdmc_factory.h"
@@ -60,17 +61,20 @@ ResultCode VfsDirectoryServiceWrapper::CreateFile(const std::string& path_, u64
 
 ResultCode VfsDirectoryServiceWrapper::DeleteFile(const std::string& path_) const {
     std::string path(FileUtil::SanitizePath(path_));
-    auto dir = GetDirectoryRelativeWrapped(backing, FileUtil::GetParentPath(path));
     if (path.empty()) {
         // TODO(DarkLordZach): Why do games call this and what should it do? Works as is but...
         return RESULT_SUCCESS;
     }
-    if (dir->GetFile(FileUtil::GetFilename(path)) == nullptr)
+
+    auto dir = GetDirectoryRelativeWrapped(backing, FileUtil::GetParentPath(path));
+    if (dir->GetFile(FileUtil::GetFilename(path)) == nullptr) {
         return FileSys::ERROR_PATH_NOT_FOUND;
+    }
     if (!dir->DeleteFile(FileUtil::GetFilename(path))) {
         // TODO(DarkLordZach): Find a better error code for this
         return ResultCode(-1);
     }
+
     return RESULT_SUCCESS;
 }
 
@@ -304,6 +308,12 @@ ResultVal<FileSys::VirtualDir> OpenSDMC() {
     return sdmc_factory->Open();
 }
 
+std::shared_ptr<FileSys::RegisteredCacheUnion> GetUnionContents() {
+    return std::make_shared<FileSys::RegisteredCacheUnion>(
+        std::vector<std::shared_ptr<FileSys::RegisteredCache>>{
+            GetSystemNANDContents(), GetUserNANDContents(), GetSDMCContents()});
+}
+
 std::shared_ptr<FileSys::RegisteredCache> GetSystemNANDContents() {
     LOG_TRACE(Service_FS, "Opening System NAND Contents");
 

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

@@ -7,11 +7,13 @@
 #include <memory>
 #include "common/common_types.h"
 #include "core/file_sys/directory.h"
+#include "core/file_sys/vfs.h"
 #include "core/hle/result.h"
 
 namespace FileSys {
 class BISFactory;
 class RegisteredCache;
+class RegisteredCacheUnion;
 class RomFSFactory;
 class SaveDataFactory;
 class SDMCFactory;
@@ -44,6 +46,8 @@ ResultVal<FileSys::VirtualDir> OpenSaveData(FileSys::SaveDataSpaceId space,
                                             FileSys::SaveDataDescriptor save_struct);
 ResultVal<FileSys::VirtualDir> OpenSDMC();
 
+std::shared_ptr<FileSys::RegisteredCacheUnion> GetUnionContents();
+
 std::shared_ptr<FileSys::RegisteredCache> GetSystemNANDContents();
 std::shared_ptr<FileSys::RegisteredCache> GetUserNANDContents();
 std::shared_ptr<FileSys::RegisteredCache> GetSDMCContents();

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

@@ -26,6 +26,17 @@
 
 namespace Service::FileSystem {
 
+enum class FileSystemType : u8 {
+    Invalid0 = 0,
+    Invalid1 = 1,
+    Logo = 2,
+    ContentControl = 3,
+    ContentManual = 4,
+    ContentMeta = 5,
+    ContentData = 6,
+    ApplicationPackage = 7,
+};
+
 class IStorage final : public ServiceFramework<IStorage> {
 public:
     explicit IStorage(FileSys::VirtualFile backend_)
@@ -420,7 +431,7 @@ FSP_SRV::FSP_SRV() : ServiceFramework("fsp-srv") {
         {0, nullptr, "MountContent"},
         {1, &FSP_SRV::Initialize, "Initialize"},
         {2, nullptr, "OpenDataFileSystemByCurrentProcess"},
-        {7, nullptr, "OpenFileSystemWithPatch"},
+        {7, &FSP_SRV::OpenFileSystemWithPatch, "OpenFileSystemWithPatch"},
         {8, nullptr, "OpenFileSystemWithId"},
         {9, nullptr, "OpenDataFileSystemByApplicationId"},
         {11, nullptr, "OpenBisFileSystem"},
@@ -444,7 +455,7 @@ FSP_SRV::FSP_SRV() : ServiceFramework("fsp-srv") {
         {34, nullptr, "GetCacheStorageSize"},
         {51, &FSP_SRV::MountSaveData, "MountSaveData"},
         {52, nullptr, "OpenSaveDataFileSystemBySystemSaveDataId"},
-        {53, nullptr, "OpenReadOnlySaveDataFileSystem"},
+        {53, &FSP_SRV::OpenReadOnlySaveDataFileSystem, "OpenReadOnlySaveDataFileSystem"},
         {57, nullptr, "ReadSaveDataFileSystemExtraDataBySaveDataSpaceId"},
         {58, nullptr, "ReadSaveDataFileSystemExtraData"},
         {59, nullptr, "WriteSaveDataFileSystemExtraData"},
@@ -516,6 +527,16 @@ void FSP_SRV::Initialize(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
+void FSP_SRV::OpenFileSystemWithPatch(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+
+    const auto type = rp.PopRaw<FileSystemType>();
+    const auto title_id = rp.PopRaw<u64>();
+
+    IPC::ResponseBuilder rb{ctx, 2, 0, 0};
+    rb.Push(ResultCode(-1));
+}
+
 void FSP_SRV::MountSdCard(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_FS, "called");
 
@@ -563,6 +584,11 @@ void FSP_SRV::MountSaveData(Kernel::HLERequestContext& ctx) {
     rb.PushIpcInterface<IFileSystem>(std::move(filesystem));
 }
 
+void FSP_SRV::OpenReadOnlySaveDataFileSystem(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service_FS, "(STUBBED) called, delegating to 51 OpenSaveDataFilesystem");
+    MountSaveData(ctx);
+}
+
 void FSP_SRV::GetGlobalAccessLogMode(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_FS, "(STUBBED) called");
 

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

@@ -20,9 +20,11 @@ public:
 
 private:
     void Initialize(Kernel::HLERequestContext& ctx);
+    void OpenFileSystemWithPatch(Kernel::HLERequestContext& ctx);
     void MountSdCard(Kernel::HLERequestContext& ctx);
     void CreateSaveData(Kernel::HLERequestContext& ctx);
     void MountSaveData(Kernel::HLERequestContext& ctx);
+    void OpenReadOnlySaveDataFileSystem(Kernel::HLERequestContext& ctx);
     void GetGlobalAccessLogMode(Kernel::HLERequestContext& ctx);
     void OpenDataStorageByCurrentProcess(Kernel::HLERequestContext& ctx);
     void OpenDataStorageByDataId(Kernel::HLERequestContext& ctx);

src/core/hle/service/ns/pl_u.cpp

@@ -5,7 +5,9 @@
 #include "common/common_paths.h"
 #include "common/file_util.h"
 #include "core/core.h"
-#include "core/file_sys/bis_factory.h"
+#include "core/file_sys/content_archive.h"
+#include "core/file_sys/nca_metadata.h"
+#include "core/file_sys/registered_cache.h"
 #include "core/file_sys/romfs.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/service/filesystem/filesystem.h"

src/core/hle/service/nvdrv/devices/nvhost_gpu.h

@@ -10,6 +10,7 @@
 #include "common/common_types.h"
 #include "common/swap.h"
 #include "core/hle/service/nvdrv/devices/nvdevice.h"
+#include "video_core/memory_manager.h"
 
 namespace Service::Nvidia::Devices {
 

src/core/hle/service/service.cpp

@@ -12,6 +12,7 @@
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/server_port.h"
 #include "core/hle/kernel/thread.h"
@@ -114,7 +115,7 @@ void ServiceFrameworkBase::InstallAsNamedPort() {
     std::tie(server_port, client_port) =
         ServerPort::CreatePortPair(kernel, max_sessions, service_name);
     server_port->SetHleHandler(shared_from_this());
-    AddNamedPort(service_name, std::move(client_port));
+    kernel.AddNamedPort(service_name, std::move(client_port));
 }
 
 Kernel::SharedPtr<Kernel::ClientPort> ServiceFrameworkBase::CreatePort() {
@@ -197,11 +198,6 @@ ResultCode ServiceFrameworkBase::HandleSyncRequest(Kernel::HLERequestContext& co
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Module interface
 
-// TODO(yuriks): Move to kernel
-void AddNamedPort(std::string name, SharedPtr<ClientPort> port) {
-    g_kernel_named_ports.emplace(std::move(name), std::move(port));
-}
-
 /// Initialize ServiceManager
 void Init(std::shared_ptr<SM::ServiceManager>& sm, const FileSys::VirtualFilesystem& rfs) {
     // NVFlinger needs to be accessed by several services like Vi and AppletOE so we instantiate it
@@ -264,7 +260,6 @@ void Init(std::shared_ptr<SM::ServiceManager>& sm, const FileSys::VirtualFilesys
 
 /// Shutdown ServiceManager
 void Shutdown() {
-    g_kernel_named_ports.clear();
     LOG_DEBUG(Service, "shutdown OK");
 }
 } // namespace Service

src/core/hle/service/service.h

@@ -6,7 +6,6 @@
 
 #include <cstddef>
 #include <string>
-#include <unordered_map>
 #include <boost/container/flat_map.hpp>
 #include "common/common_types.h"
 #include "core/hle/kernel/hle_ipc.h"
@@ -187,10 +186,4 @@ void Init(std::shared_ptr<SM::ServiceManager>& sm,
 /// Shutdown ServiceManager
 void Shutdown();
 
-/// Map of named ports managed by the kernel, which can be retrieved using the ConnectToPort SVC.
-extern std::unordered_map<std::string, Kernel::SharedPtr<Kernel::ClientPort>> g_kernel_named_ports;
-
-/// Adds a port to the named port table
-void AddNamedPort(std::string name, Kernel::SharedPtr<Kernel::ClientPort> port);
-
 } // namespace Service

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

@@ -3,6 +3,9 @@
 // Refer to the license.txt file included.
 
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/hle_ipc.h"
+#include "core/hle/service/service.h"
+#include "core/hle/service/sm/sm.h"
 #include "core/hle/service/ssl/ssl.h"
 
 namespace Service::SSL {
@@ -81,36 +84,43 @@ private:
     }
 };
 
-void SSL::CreateContext(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_SSL, "(STUBBED) called");
+class SSL final : public ServiceFramework<SSL> {
+public:
+    explicit SSL() : ServiceFramework{"ssl"} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0, &SSL::CreateContext, "CreateContext"},
+            {1, nullptr, "GetContextCount"},
+            {2, nullptr, "GetCertificates"},
+            {3, nullptr, "GetCertificateBufSize"},
+            {4, nullptr, "DebugIoctl"},
+            {5, &SSL::SetInterfaceVersion, "SetInterfaceVersion"},
+            {6, nullptr, "FlushSessionCache"},
+        };
+        // clang-format on
 
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISslContext>();
-}
+        RegisterHandlers(functions);
+    }
 
-SSL::SSL() : ServiceFramework("ssl") {
-    static const FunctionInfo functions[] = {
-        {0, &SSL::CreateContext, "CreateContext"},
-        {1, nullptr, "GetContextCount"},
-        {2, nullptr, "GetCertificates"},
-        {3, nullptr, "GetCertificateBufSize"},
-        {4, nullptr, "DebugIoctl"},
-        {5, &SSL::SetInterfaceVersion, "SetInterfaceVersion"},
-        {6, nullptr, "FlushSessionCache"},
-    };
-    RegisterHandlers(functions);
-}
+private:
+    void CreateContext(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_SSL, "(STUBBED) called");
 
-void SSL::SetInterfaceVersion(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_SSL, "(STUBBED) 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
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+        rb.Push(RESULT_SUCCESS);
+        rb.PushIpcInterface<ISslContext>();
+    }
 
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(RESULT_SUCCESS);
-}
+    void SetInterfaceVersion(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_SSL, "(STUBBED) 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
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+};
 
 void InstallInterfaces(SM::ServiceManager& service_manager) {
     std::make_shared<SSL>()->InstallAsService(service_manager);

src/core/hle/service/ssl/ssl.h

@@ -4,20 +4,12 @@
 
 #pragma once
 
-#include "core/hle/service/service.h"
+namespace Service::SM {
+class ServiceManager;
+}
 
 namespace Service::SSL {
 
-class SSL final : public ServiceFramework<SSL> {
-public:
-    explicit SSL();
-    ~SSL() = default;
-
-private:
-    void CreateContext(Kernel::HLERequestContext& ctx);
-    void SetInterfaceVersion(Kernel::HLERequestContext& ctx);
-};
-
 /// Registers all SSL services with the specified service manager.
 void InstallInterfaces(SM::ServiceManager& service_manager);
 

src/core/loader/deconstructed_rom_directory.cpp

@@ -9,8 +9,10 @@
 #include "core/core.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/romfs_factory.h"
 #include "core/gdbstub/gdbstub.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
 #include "core/hle/service/filesystem/filesystem.h"
@@ -20,10 +22,19 @@
 
 namespace Loader {
 
-AppLoader_DeconstructedRomDirectory::AppLoader_DeconstructedRomDirectory(FileSys::VirtualFile file_)
-    : AppLoader(std::move(file_)) {
+AppLoader_DeconstructedRomDirectory::AppLoader_DeconstructedRomDirectory(FileSys::VirtualFile file_,
+                                                                         bool override_update)
+    : AppLoader(std::move(file_)), override_update(override_update) {
     const auto dir = file->GetContainingDirectory();
 
+    // Title ID
+    const auto npdm = dir->GetFile("main.npdm");
+    if (npdm != nullptr) {
+        const auto res = metadata.Load(npdm);
+        if (res == ResultStatus::Success)
+            title_id = metadata.GetTitleID();
+    }
+
     // Icon
     FileSys::VirtualFile icon_file = nullptr;
     for (const auto& language : FileSys::LANGUAGE_NAMES) {
@@ -60,14 +71,14 @@ AppLoader_DeconstructedRomDirectory::AppLoader_DeconstructedRomDirectory(FileSys
 
     if (nacp_file != nullptr) {
         FileSys::NACP nacp(nacp_file);
-        title_id = nacp.GetTitleId();
         name = nacp.GetApplicationName();
     }
 }
 
 AppLoader_DeconstructedRomDirectory::AppLoader_DeconstructedRomDirectory(
-    FileSys::VirtualDir directory)
-    : AppLoader(directory->GetFile("main")), dir(std::move(directory)) {}
+    FileSys::VirtualDir directory, bool override_update)
+    : AppLoader(directory->GetFile("main")), dir(std::move(directory)),
+      override_update(override_update) {}
 
 FileType AppLoader_DeconstructedRomDirectory::IdentifyType(const FileSys::VirtualFile& file) {
     if (FileSys::IsDirectoryExeFS(file->GetContainingDirectory())) {
@@ -89,7 +100,8 @@ ResultStatus AppLoader_DeconstructedRomDirectory::Load(
         dir = file->GetContainingDirectory();
     }
 
-    const FileSys::VirtualFile npdm = dir->GetFile("main.npdm");
+    // Read meta to determine title ID
+    FileSys::VirtualFile npdm = dir->GetFile("main.npdm");
     if (npdm == nullptr)
         return ResultStatus::ErrorMissingNPDM;
 
@@ -97,6 +109,21 @@ ResultStatus AppLoader_DeconstructedRomDirectory::Load(
     if (result != ResultStatus::Success) {
         return result;
     }
+
+    if (override_update) {
+        const FileSys::PatchManager patch_manager(metadata.GetTitleID());
+        dir = patch_manager.PatchExeFS(dir);
+    }
+
+    // Reread in case PatchExeFS affected the main.npdm
+    npdm = dir->GetFile("main.npdm");
+    if (npdm == nullptr)
+        return ResultStatus::ErrorMissingNPDM;
+
+    ResultStatus result2 = metadata.Load(npdm);
+    if (result2 != ResultStatus::Success) {
+        return result2;
+    }
     metadata.Print();
 
     const FileSys::ProgramAddressSpaceType arch_bits{metadata.GetAddressSpaceType()};
@@ -158,8 +185,6 @@ ResultStatus AppLoader_DeconstructedRomDirectory::ReadIcon(std::vector<u8>& buff
 }
 
 ResultStatus AppLoader_DeconstructedRomDirectory::ReadProgramId(u64& out_program_id) {
-    if (name.empty())
-        return ResultStatus::ErrorNoControl;
     out_program_id = title_id;
     return ResultStatus::Success;
 }
@@ -171,4 +196,8 @@ ResultStatus AppLoader_DeconstructedRomDirectory::ReadTitle(std::string& title)
     return ResultStatus::Success;
 }
 
+bool AppLoader_DeconstructedRomDirectory::IsRomFSUpdatable() const {
+    return false;
+}
+
 } // namespace Loader

src/core/loader/deconstructed_rom_directory.h

@@ -20,10 +20,12 @@ namespace Loader {
  */
 class AppLoader_DeconstructedRomDirectory final : public AppLoader {
 public:
-    explicit AppLoader_DeconstructedRomDirectory(FileSys::VirtualFile main_file);
+    explicit AppLoader_DeconstructedRomDirectory(FileSys::VirtualFile main_file,
+                                                 bool override_update = false);
 
     // Overload to accept exefs directory. Must contain 'main' and 'main.npdm'
-    explicit AppLoader_DeconstructedRomDirectory(FileSys::VirtualDir directory);
+    explicit AppLoader_DeconstructedRomDirectory(FileSys::VirtualDir directory,
+                                                 bool override_update = false);
 
     /**
      * Returns the type of the file
@@ -42,6 +44,7 @@ public:
     ResultStatus ReadIcon(std::vector<u8>& buffer) override;
     ResultStatus ReadProgramId(u64& out_program_id) override;
     ResultStatus ReadTitle(std::string& title) override;
+    bool IsRomFSUpdatable() const override;
 
 private:
     FileSys::ProgramMetadata metadata;
@@ -51,6 +54,7 @@ private:
     std::vector<u8> icon_data;
     std::string name;
     u64 title_id{};
+    bool override_update;
 };
 
 } // namespace Loader

src/core/loader/elf.cpp

@@ -10,6 +10,7 @@
 #include "common/file_util.h"
 #include "common/logging/log.h"
 #include "core/core.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
 #include "core/loader/elf.h"

src/core/loader/loader.cpp

@@ -5,9 +5,9 @@
 #include <memory>
 #include <ostream>
 #include <string>
+#include "common/file_util.h"
 #include "common/logging/log.h"
 #include "common/string_util.h"
-#include "core/file_sys/vfs_real.h"
 #include "core/hle/kernel/process.h"
 #include "core/loader/deconstructed_rom_directory.h"
 #include "core/loader/elf.h"
@@ -15,6 +15,7 @@
 #include "core/loader/nca.h"
 #include "core/loader/nro.h"
 #include "core/loader/nso.h"
+#include "core/loader/nsp.h"
 #include "core/loader/xci.h"
 
 namespace Loader {
@@ -34,6 +35,7 @@ FileType IdentifyFile(FileSys::VirtualFile file) {
     CHECK_TYPE(NCA)
     CHECK_TYPE(XCI)
     CHECK_TYPE(NAX)
+    CHECK_TYPE(NSP)
 
 #undef CHECK_TYPE
 
@@ -59,6 +61,8 @@ FileType GuessFromFilename(const std::string& name) {
         return FileType::NCA;
     if (extension == "xci")
         return FileType::XCI;
+    if (extension == "nsp")
+        return FileType::NSP;
 
     return FileType::Unknown;
 }
@@ -77,6 +81,8 @@ std::string GetFileTypeString(FileType type) {
         return "XCI";
     case FileType::NAX:
         return "NAX";
+    case FileType::NSP:
+        return "NSP";
     case FileType::DeconstructedRomDirectory:
         return "Directory";
     case FileType::Error:
@@ -87,7 +93,7 @@ std::string GetFileTypeString(FileType type) {
     return "unknown";
 }
 
-constexpr std::array<const char*, 49> RESULT_MESSAGES{
+constexpr std::array<const char*, 58> RESULT_MESSAGES{
     "The operation completed successfully.",
     "The loader requested to load is already loaded.",
     "The operation is not implemented.",
@@ -137,6 +143,15 @@ constexpr std::array<const char*, 49> RESULT_MESSAGES{
     "The AES Key Generation Source could not be found.",
     "The SD Save Key Source could not be found.",
     "The SD NCA Key Source could not be found.",
+    "The NSP file is missing a Program-type NCA.",
+    "The BKTR-type NCA has a bad BKTR header.",
+    "The BKTR Subsection entry is not located immediately after the Relocation entry.",
+    "The BKTR Subsection entry is not at the end of the media block.",
+    "The BKTR-type NCA has a bad Relocation block.",
+    "The BKTR-type NCA has a bad Subsection block.",
+    "The BKTR-type NCA has a bad Relocation bucket.",
+    "The BKTR-type NCA has a bad Subsection bucket.",
+    "The BKTR-type NCA is missing the base RomFS.",
 };
 
 std::ostream& operator<<(std::ostream& os, ResultStatus status) {
@@ -144,6 +159,9 @@ std::ostream& operator<<(std::ostream& os, ResultStatus status) {
     return os;
 }
 
+AppLoader::AppLoader(FileSys::VirtualFile file) : file(std::move(file)) {}
+AppLoader::~AppLoader() = default;
+
 /**
  * Get a loader for a file with a specific type
  * @param file The file to load
@@ -179,6 +197,10 @@ static std::unique_ptr<AppLoader> GetFileLoader(FileSys::VirtualFile file, FileT
     case FileType::NAX:
         return std::make_unique<AppLoader_NAX>(std::move(file));
 
+    // NX NSP (Nintendo Submission Package) file format
+    case FileType::NSP:
+        return std::make_unique<AppLoader_NSP>(std::move(file));
+
     // NX deconstructed ROM directory.
     case FileType::DeconstructedRomDirectory:
         return std::make_unique<AppLoader_DeconstructedRomDirectory>(std::move(file));

src/core/loader/loader.h

@@ -4,7 +4,6 @@
 
 #pragma once
 
-#include <algorithm>
 #include <iosfwd>
 #include <memory>
 #include <string>
@@ -12,7 +11,6 @@
 #include <vector>
 #include <boost/optional.hpp>
 #include "common/common_types.h"
-#include "common/file_util.h"
 #include "core/file_sys/vfs.h"
 #include "core/hle/kernel/object.h"
 
@@ -31,6 +29,7 @@ enum class FileType {
     NSO,
     NRO,
     NCA,
+    NSP,
     XCI,
     NAX,
     DeconstructedRomDirectory,
@@ -107,6 +106,15 @@ enum class ResultStatus : u16 {
     ErrorMissingAESKeyGenerationSource,
     ErrorMissingSDSaveKeySource,
     ErrorMissingSDNCAKeySource,
+    ErrorNSPMissingProgramNCA,
+    ErrorBadBKTRHeader,
+    ErrorBKTRSubsectionNotAfterRelocation,
+    ErrorBKTRSubsectionNotAtEnd,
+    ErrorBadRelocationBlock,
+    ErrorBadSubsectionBlock,
+    ErrorBadRelocationBuckets,
+    ErrorBadSubsectionBuckets,
+    ErrorMissingBKTRBaseRomFS,
 };
 
 std::ostream& operator<<(std::ostream& os, ResultStatus status);
@@ -114,8 +122,8 @@ std::ostream& operator<<(std::ostream& os, ResultStatus status);
 /// Interface for loading an application
 class AppLoader : NonCopyable {
 public:
-    explicit AppLoader(FileSys::VirtualFile file) : file(std::move(file)) {}
-    virtual ~AppLoader() {}
+    explicit AppLoader(FileSys::VirtualFile file);
+    virtual ~AppLoader();
 
     /**
      * Returns the type of this file
@@ -197,13 +205,22 @@ public:
     }
 
     /**
-     * Get the update RomFS of the application
-     * Since the RomFS can be huge, we return a file reference instead of copying to a buffer
-     * @param file The file containing the RomFS
-     * @return ResultStatus result of function
+     * Get whether or not updates can be applied to the RomFS.
+     * By default, this is true, however for formats where it cannot be guaranteed that the RomFS is
+     * the base game it should be set to false.
+     * @return bool whether or not updatable.
      */
-    virtual ResultStatus ReadUpdateRomFS(FileSys::VirtualFile& file) {
-        return ResultStatus::ErrorNotImplemented;
+    virtual bool IsRomFSUpdatable() const {
+        return true;
+    }
+
+    /**
+     * Gets the difference between the start of the IVFC header and the start of level 6 (RomFS)
+     * data. Needed for bktr patching.
+     * @return IVFC offset for romfs.
+     */
+    virtual u64 ReadRomFSIVFCOffset() const {
+        return 0;
     }
 
     /**

src/core/loader/nca.cpp

@@ -48,7 +48,7 @@ ResultStatus AppLoader_NCA::Load(Kernel::SharedPtr<Kernel::Process>& process) {
     if (exefs == nullptr)
         return ResultStatus::ErrorNoExeFS;
 
-    directory_loader = std::make_unique<AppLoader_DeconstructedRomDirectory>(exefs);
+    directory_loader = std::make_unique<AppLoader_DeconstructedRomDirectory>(exefs, true);
 
     const auto load_result = directory_loader->Load(process);
     if (load_result != ResultStatus::Success)
@@ -71,6 +71,12 @@ ResultStatus AppLoader_NCA::ReadRomFS(FileSys::VirtualFile& dir) {
     return ResultStatus::Success;
 }
 
+u64 AppLoader_NCA::ReadRomFSIVFCOffset() const {
+    if (nca == nullptr)
+        return 0;
+    return nca->GetBaseIVFCOffset();
+}
+
 ResultStatus AppLoader_NCA::ReadProgramId(u64& out_program_id) {
     if (nca == nullptr || nca->GetStatus() != ResultStatus::Success)
         return ResultStatus::ErrorNotInitialized;

src/core/loader/nca.h

@@ -37,6 +37,7 @@ public:
     ResultStatus Load(Kernel::SharedPtr<Kernel::Process>& process) override;
 
     ResultStatus ReadRomFS(FileSys::VirtualFile& dir) override;
+    u64 ReadRomFSIVFCOffset() const override;
     ResultStatus ReadProgramId(u64& out_program_id) override;
 
 private:

src/core/loader/nro.cpp

@@ -14,6 +14,7 @@
 #include "core/file_sys/control_metadata.h"
 #include "core/file_sys/vfs_offset.h"
 #include "core/gdbstub/gdbstub.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
 #include "core/loader/nro.h"
@@ -231,4 +232,9 @@ ResultStatus AppLoader_NRO::ReadTitle(std::string& title) {
     title = nacp->GetApplicationName();
     return ResultStatus::Success;
 }
+
+bool AppLoader_NRO::IsRomFSUpdatable() const {
+    return false;
+}
+
 } // namespace Loader

src/core/loader/nro.h

@@ -39,6 +39,7 @@ public:
     ResultStatus ReadProgramId(u64& out_program_id) override;
     ResultStatus ReadRomFS(FileSys::VirtualFile& dir) override;
     ResultStatus ReadTitle(std::string& title) override;
+    bool IsRomFSUpdatable() const override;
 
 private:
     bool LoadNro(FileSys::VirtualFile file, VAddr load_base);

src/core/loader/nso.cpp

@@ -11,6 +11,7 @@
 #include "common/swap.h"
 #include "core/core.h"
 #include "core/gdbstub/gdbstub.h"
+#include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
 #include "core/loader/nso.h"

src/core/loader/nsp.cpp

@@ -0,0 +1,125 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <vector>
+
+#include "common/common_types.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/nca_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/submission_package.h"
+#include "core/hle/kernel/process.h"
+#include "core/loader/deconstructed_rom_directory.h"
+#include "core/loader/nca.h"
+#include "core/loader/nsp.h"
+
+namespace Loader {
+
+AppLoader_NSP::AppLoader_NSP(FileSys::VirtualFile file)
+    : AppLoader(file), nsp(std::make_unique<FileSys::NSP>(file)),
+      title_id(nsp->GetProgramTitleID()) {
+
+    if (nsp->GetStatus() != ResultStatus::Success)
+        return;
+    if (nsp->IsExtractedType())
+        return;
+
+    const auto control_nca =
+        nsp->GetNCA(nsp->GetProgramTitleID(), FileSys::ContentRecordType::Control);
+    if (control_nca == nullptr || control_nca->GetStatus() != ResultStatus::Success)
+        return;
+
+    std::tie(nacp_file, icon_file) =
+        FileSys::PatchManager(nsp->GetProgramTitleID()).ParseControlNCA(control_nca);
+}
+
+AppLoader_NSP::~AppLoader_NSP() = default;
+
+FileType AppLoader_NSP::IdentifyType(const FileSys::VirtualFile& file) {
+    FileSys::NSP nsp(file);
+
+    if (nsp.GetStatus() == ResultStatus::Success) {
+        // Extracted Type case
+        if (nsp.IsExtractedType() && nsp.GetExeFS() != nullptr &&
+            FileSys::IsDirectoryExeFS(nsp.GetExeFS()) && nsp.GetRomFS() != nullptr) {
+            return FileType::NSP;
+        }
+
+        // Non-Ectracted Type case
+        if (!nsp.IsExtractedType() &&
+            nsp.GetNCA(nsp.GetFirstTitleID(), FileSys::ContentRecordType::Program) != nullptr &&
+            AppLoader_NCA::IdentifyType(nsp.GetNCAFile(
+                nsp.GetFirstTitleID(), FileSys::ContentRecordType::Program)) == FileType::NCA) {
+            return FileType::NSP;
+        }
+    }
+
+    return FileType::Error;
+}
+
+ResultStatus AppLoader_NSP::Load(Kernel::SharedPtr<Kernel::Process>& process) {
+    if (is_loaded) {
+        return ResultStatus::ErrorAlreadyLoaded;
+    }
+
+    if (nsp->IsExtractedType()) {
+        secondary_loader = std::make_unique<AppLoader_DeconstructedRomDirectory>(nsp->GetExeFS());
+    } else {
+        if (title_id == 0)
+            return ResultStatus::ErrorNSPMissingProgramNCA;
+
+        secondary_loader = std::make_unique<AppLoader_NCA>(
+            nsp->GetNCAFile(title_id, FileSys::ContentRecordType::Program));
+
+        if (nsp->GetStatus() != ResultStatus::Success)
+            return nsp->GetStatus();
+
+        if (nsp->GetProgramStatus(title_id) != ResultStatus::Success)
+            return nsp->GetProgramStatus(title_id);
+
+        if (nsp->GetNCA(title_id, FileSys::ContentRecordType::Program) == nullptr) {
+            if (!Core::Crypto::KeyManager::KeyFileExists(false))
+                return ResultStatus::ErrorMissingProductionKeyFile;
+            return ResultStatus::ErrorNSPMissingProgramNCA;
+        }
+    }
+
+    const auto result = secondary_loader->Load(process);
+    if (result != ResultStatus::Success)
+        return result;
+
+    is_loaded = true;
+
+    return ResultStatus::Success;
+}
+
+ResultStatus AppLoader_NSP::ReadRomFS(FileSys::VirtualFile& dir) {
+    return secondary_loader->ReadRomFS(dir);
+}
+
+ResultStatus AppLoader_NSP::ReadProgramId(u64& out_program_id) {
+    if (title_id == 0)
+        return ResultStatus::ErrorNotInitialized;
+    out_program_id = title_id;
+    return ResultStatus::Success;
+}
+
+ResultStatus AppLoader_NSP::ReadIcon(std::vector<u8>& buffer) {
+    if (icon_file == nullptr)
+        return ResultStatus::ErrorNoControl;
+    buffer = icon_file->ReadAllBytes();
+    return ResultStatus::Success;
+}
+
+ResultStatus AppLoader_NSP::ReadTitle(std::string& title) {
+    if (nacp_file == nullptr)
+        return ResultStatus::ErrorNoControl;
+    title = nacp_file->GetApplicationName();
+    return ResultStatus::Success;
+}
+} // namespace Loader

src/core/loader/nsp.h

@@ -0,0 +1,54 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include "common/common_types.h"
+#include "core/file_sys/vfs.h"
+#include "core/loader/loader.h"
+
+namespace FileSys {
+class NACP;
+class NSP;
+} // namespace FileSys
+
+namespace Loader {
+
+class AppLoader_NCA;
+
+/// Loads an XCI file
+class AppLoader_NSP final : public AppLoader {
+public:
+    explicit AppLoader_NSP(FileSys::VirtualFile file);
+    ~AppLoader_NSP() override;
+
+    /**
+     * Returns the type of the file
+     * @param file std::shared_ptr<VfsFile> open file
+     * @return FileType found, or FileType::Error if this loader doesn't know it
+     */
+    static FileType IdentifyType(const FileSys::VirtualFile& file);
+
+    FileType GetFileType() override {
+        return IdentifyType(file);
+    }
+
+    ResultStatus Load(Kernel::SharedPtr<Kernel::Process>& process) override;
+
+    ResultStatus ReadRomFS(FileSys::VirtualFile& dir) override;
+    ResultStatus ReadProgramId(u64& out_program_id) override;
+    ResultStatus ReadIcon(std::vector<u8>& buffer) override;
+    ResultStatus ReadTitle(std::string& title) override;
+
+private:
+    std::unique_ptr<FileSys::NSP> nsp;
+    std::unique_ptr<AppLoader> secondary_loader;
+
+    FileSys::VirtualFile icon_file;
+    std::shared_ptr<FileSys::NACP> nacp_file;
+    u64 title_id;
+};
+
+} // namespace Loader

src/core/loader/xci.cpp

@@ -8,7 +8,9 @@
 #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/romfs.h"
+#include "core/file_sys/submission_package.h"
 #include "core/hle/kernel/process.h"
 #include "core/loader/nca.h"
 #include "core/loader/xci.h"
@@ -17,25 +19,16 @@ namespace Loader {
 
 AppLoader_XCI::AppLoader_XCI(FileSys::VirtualFile file)
     : AppLoader(file), xci(std::make_unique<FileSys::XCI>(file)),
-      nca_loader(std::make_unique<AppLoader_NCA>(
-          xci->GetNCAFileByType(FileSys::NCAContentType::Program))) {
+      nca_loader(std::make_unique<AppLoader_NCA>(xci->GetProgramNCAFile())) {
     if (xci->GetStatus() != ResultStatus::Success)
         return;
+
     const auto control_nca = xci->GetNCAByType(FileSys::NCAContentType::Control);
     if (control_nca == nullptr || control_nca->GetStatus() != ResultStatus::Success)
         return;
-    const auto romfs = FileSys::ExtractRomFS(control_nca->GetRomFS());
-    if (romfs == nullptr)
-        return;
-    for (const auto& language : FileSys::LANGUAGE_NAMES) {
-        icon_file = romfs->GetFile("icon_" + std::string(language) + ".dat");
-        if (icon_file != nullptr)
-            break;
-    }
-    const auto nacp_raw = romfs->GetFile("control.nacp");
-    if (nacp_raw == nullptr)
-        return;
-    nacp_file = std::make_shared<FileSys::NACP>(nacp_raw);
+
+    std::tie(nacp_file, icon_file) =
+        FileSys::PatchManager(xci->GetProgramTitleID()).ParseControlNCA(control_nca);
 }
 
 AppLoader_XCI::~AppLoader_XCI() = default;
@@ -64,11 +57,11 @@ ResultStatus AppLoader_XCI::Load(Kernel::SharedPtr<Kernel::Process>& process) {
     if (xci->GetProgramNCAStatus() != ResultStatus::Success)
         return xci->GetProgramNCAStatus();
 
-    const auto nca = xci->GetNCAFileByType(FileSys::NCAContentType::Program);
+    const auto nca = xci->GetProgramNCA();
     if (nca == nullptr && !Core::Crypto::KeyManager::KeyFileExists(false))
         return ResultStatus::ErrorMissingProductionKeyFile;
 
-    auto result = nca_loader->Load(process);
+    const auto result = nca_loader->Load(process);
     if (result != ResultStatus::Success)
         return result;
 

src/core/memory.cpp

@@ -251,8 +251,8 @@ std::string ReadCString(VAddr vaddr, std::size_t max_length) {
     return string;
 }
 
-void RasterizerMarkRegionCached(Tegra::GPUVAddr gpu_addr, u64 size, bool cached) {
-    if (gpu_addr == 0) {
+void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
+    if (vaddr == 0) {
         return;
     }
 
@@ -261,19 +261,8 @@ void RasterizerMarkRegionCached(Tegra::GPUVAddr gpu_addr, u64 size, bool cached)
     // CPU pages, hence why we iterate on a CPU page basis (note: GPU page size is different). This
     // assumes the specified GPU address region is contiguous as well.
 
-    u64 num_pages = ((gpu_addr + size - 1) >> PAGE_BITS) - (gpu_addr >> PAGE_BITS) + 1;
-    for (unsigned i = 0; i < num_pages; ++i, gpu_addr += PAGE_SIZE) {
-        boost::optional<VAddr> maybe_vaddr =
-            Core::System::GetInstance().GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
-        // The GPU <-> CPU virtual memory mapping is not 1:1
-        if (!maybe_vaddr) {
-            LOG_ERROR(HW_Memory,
-                      "Trying to flush a cached region to an invalid physical address {:016X}",
-                      gpu_addr);
-            continue;
-        }
-        VAddr vaddr = *maybe_vaddr;
-
+    u64 num_pages = ((vaddr + size - 1) >> PAGE_BITS) - (vaddr >> PAGE_BITS) + 1;
+    for (unsigned i = 0; i < num_pages; ++i, vaddr += PAGE_SIZE) {
         PageType& page_type = current_page_table->attributes[vaddr >> PAGE_BITS];
 
         if (cached) {
@@ -344,29 +333,19 @@ void RasterizerFlushVirtualRegion(VAddr start, u64 size, FlushMode mode) {
 
         const VAddr overlap_start = std::max(start, region_start);
         const VAddr overlap_end = std::min(end, region_end);
-
-        const std::vector<Tegra::GPUVAddr> gpu_addresses =
-            system_instance.GPU().MemoryManager().CpuToGpuAddress(overlap_start);
-
-        if (gpu_addresses.empty()) {
-            return;
-        }
-
         const u64 overlap_size = overlap_end - overlap_start;
 
-        for (const auto& gpu_address : gpu_addresses) {
-            auto& rasterizer = system_instance.Renderer().Rasterizer();
-            switch (mode) {
-            case FlushMode::Flush:
-                rasterizer.FlushRegion(gpu_address, overlap_size);
-                break;
-            case FlushMode::Invalidate:
-                rasterizer.InvalidateRegion(gpu_address, overlap_size);
-                break;
-            case FlushMode::FlushAndInvalidate:
-                rasterizer.FlushAndInvalidateRegion(gpu_address, overlap_size);
-                break;
-            }
+        auto& rasterizer = system_instance.Renderer().Rasterizer();
+        switch (mode) {
+        case FlushMode::Flush:
+            rasterizer.FlushRegion(overlap_start, overlap_size);
+            break;
+        case FlushMode::Invalidate:
+            rasterizer.InvalidateRegion(overlap_start, overlap_size);
+            break;
+        case FlushMode::FlushAndInvalidate:
+            rasterizer.FlushAndInvalidateRegion(overlap_start, overlap_size);
+            break;
         }
     };
 

src/core/memory.h

@@ -11,7 +11,6 @@
 #include <boost/icl/interval_map.hpp>
 #include "common/common_types.h"
 #include "core/memory_hook.h"
-#include "video_core/memory_manager.h"
 
 namespace Kernel {
 class Process;
@@ -179,7 +178,7 @@ enum class FlushMode {
 /**
  * Mark each page touching the region as cached.
  */
-void RasterizerMarkRegionCached(Tegra::GPUVAddr gpu_addr, u64 size, bool cached);
+void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached);
 
 /**
  * Flushes and invalidates any externally cached rasterizer resources touching the given virtual

src/core/perf_stats.cpp

@@ -40,7 +40,7 @@ void PerfStats::EndGameFrame() {
     game_frames += 1;
 }
 
-PerfStats::Results PerfStats::GetAndResetStats(microseconds current_system_time_us) {
+PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us) {
     std::lock_guard<std::mutex> lock(object_mutex);
 
     const auto now = Clock::now();
@@ -49,7 +49,7 @@ PerfStats::Results PerfStats::GetAndResetStats(microseconds current_system_time_
 
     const auto system_us_per_second = (current_system_time_us - reset_point_system_us) / interval;
 
-    Results results{};
+    PerfStatsResults results{};
     results.system_fps = static_cast<double>(system_frames) / interval;
     results.game_fps = static_cast<double>(game_frames) / interval;
     results.frametime = duration_cast<DoubleSecs>(accumulated_frametime).count() /

src/core/perf_stats.h

@@ -10,6 +10,17 @@
 
 namespace Core {
 
+struct PerfStatsResults {
+    /// System FPS (LCD VBlanks) in Hz
+    double system_fps;
+    /// Game FPS (GSP frame submissions) in Hz
+    double game_fps;
+    /// Walltime per system frame, in seconds, excluding any waits
+    double frametime;
+    /// Ratio of walltime / emulated time elapsed
+    double emulation_speed;
+};
+
 /**
  * Class to manage and query performance/timing statistics. All public functions of this class are
  * thread-safe unless stated otherwise.
@@ -18,22 +29,11 @@ class PerfStats {
 public:
     using Clock = std::chrono::high_resolution_clock;
 
-    struct Results {
-        /// System FPS (LCD VBlanks) in Hz
-        double system_fps;
-        /// Game FPS (GSP frame submissions) in Hz
-        double game_fps;
-        /// Walltime per system frame, in seconds, excluding any waits
-        double frametime;
-        /// Ratio of walltime / emulated time elapsed
-        double emulation_speed;
-    };
-
     void BeginSystemFrame();
     void EndSystemFrame();
     void EndGameFrame();
 
-    Results GetAndResetStats(std::chrono::microseconds current_system_time_us);
+    PerfStatsResults GetAndResetStats(std::chrono::microseconds current_system_time_us);
 
     /**
      * Gets the ratio between walltime and the emulated time of the previous system frame. This is

src/core/settings.h

@@ -127,6 +127,8 @@ struct Values {
 
     // Data Storage
     bool use_virtual_sd;
+    std::string nand_dir;
+    std::string sdmc_dir;
 
     // Renderer
     float resolution_factor;

src/core/telemetry_session.cpp

@@ -7,6 +7,9 @@
 #include "common/file_util.h"
 
 #include "core/core.h"
+#include "core/file_sys/control_metadata.h"
+#include "core/file_sys/patch_manager.h"
+#include "core/loader/loader.h"
 #include "core/settings.h"
 #include "core/telemetry_session.h"
 
@@ -87,12 +90,28 @@ TelemetrySession::TelemetrySession() {
                             std::chrono::system_clock::now().time_since_epoch())
                             .count()};
     AddField(Telemetry::FieldType::Session, "Init_Time", init_time);
-    std::string program_name;
-    const Loader::ResultStatus res{System::GetInstance().GetAppLoader().ReadTitle(program_name)};
+
+    u64 program_id{};
+    const Loader::ResultStatus res{System::GetInstance().GetAppLoader().ReadProgramId(program_id)};
     if (res == Loader::ResultStatus::Success) {
-        AddField(Telemetry::FieldType::Session, "ProgramName", program_name);
+        AddField(Telemetry::FieldType::Session, "ProgramId", program_id);
+
+        std::string name;
+        System::GetInstance().GetAppLoader().ReadTitle(name);
+
+        if (name.empty()) {
+            auto [nacp, icon_file] = FileSys::PatchManager(program_id).GetControlMetadata();
+            if (nacp != nullptr)
+                name = nacp->GetApplicationName();
+        }
+
+        if (!name.empty())
+            AddField(Telemetry::FieldType::Session, "ProgramName", name);
     }
 
+    AddField(Telemetry::FieldType::Session, "ProgramFormat",
+             static_cast<u8>(System::GetInstance().GetAppLoader().GetFileType()));
+
     // Log application information
     Telemetry::AppendBuildInfo(field_collection);
 

src/video_core/CMakeLists.txt

@@ -22,6 +22,7 @@ add_library(video_core STATIC
     rasterizer_interface.h
     renderer_base.cpp
     renderer_base.h
+    renderer_opengl/gl_buffer_cache.cpp
     renderer_opengl/gl_rasterizer.cpp
     renderer_opengl/gl_rasterizer.h
     renderer_opengl/gl_rasterizer_cache.cpp

src/video_core/command_processor.cpp

@@ -34,6 +34,8 @@ void GPU::WriteReg(u32 method, u32 subchannel, u32 value, u32 remaining_params)
               "{:08X} remaining params {}",
               method, subchannel, value, remaining_params);
 
+    ASSERT(subchannel < bound_engines.size());
+
     if (method == static_cast<u32>(BufferMethods::BindObject)) {
         // Bind the current subchannel to the desired engine id.
         LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {}", subchannel, value);
@@ -47,8 +49,6 @@ void GPU::WriteReg(u32 method, u32 subchannel, u32 value, u32 remaining_params)
         return;
     }
 
-    ASSERT(bound_engines.find(subchannel) != bound_engines.end());
-
     const EngineID engine = bound_engines[subchannel];
 
     switch (engine) {

src/video_core/engines/maxwell_3d.cpp

@@ -5,13 +5,13 @@
 #include <cinttypes>
 #include "common/assert.h"
 #include "core/core.h"
+#include "core/core_timing.h"
+#include "core/memory.h"
 #include "video_core/debug_utils/debug_utils.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
-#include "video_core/textures/decoders.h"
 #include "video_core/textures/texture.h"
-#include "video_core/video_core.h"
 
 namespace Tegra {
 namespace Engines {
@@ -195,8 +195,8 @@ void Maxwell3D::ProcessQueryGet() {
             // wait queues.
             LongQueryResult query_result{};
             query_result.value = result;
-            // TODO(Subv): Generate a real GPU timestamp and write it here instead of 0
-            query_result.timestamp = 0;
+            // TODO(Subv): Generate a real GPU timestamp and write it here instead of CoreTiming
+            query_result.timestamp = CoreTiming::GetTicks();
             Memory::WriteBlock(*address, &query_result, sizeof(query_result));
         }
         break;

src/video_core/engines/maxwell_3d.h

@@ -127,6 +127,7 @@ public:
                 BitField<21, 6, Size> size;
                 BitField<27, 3, Type> type;
                 BitField<31, 1, u32> bgra;
+                u32 hex;
             };
 
             u32 ComponentCount() const {
@@ -262,6 +263,10 @@ public:
             bool IsValid() const {
                 return size != Size::Invalid;
             }
+
+            bool operator<(const VertexAttribute& other) const {
+                return hex < other.hex;
+            }
         };
 
         enum class PrimitiveTopology : u32 {
@@ -545,7 +550,7 @@ public:
 
                 INSERT_PADDING_WORDS(0x5B);
 
-                VertexAttribute vertex_attrib_format[NumVertexAttributes];
+                std::array<VertexAttribute, NumVertexAttributes> vertex_attrib_format;
 
                 INSERT_PADDING_WORDS(0xF);
 
@@ -964,7 +969,7 @@ ASSERT_REG_POSITION(stencil_back_func_ref, 0x3D5);
 ASSERT_REG_POSITION(stencil_back_mask, 0x3D6);
 ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7);
 ASSERT_REG_POSITION(zeta, 0x3F8);
-ASSERT_REG_POSITION(vertex_attrib_format[0], 0x458);
+ASSERT_REG_POSITION(vertex_attrib_format, 0x458);
 ASSERT_REG_POSITION(rt_control, 0x487);
 ASSERT_REG_POSITION(zeta_width, 0x48a);
 ASSERT_REG_POSITION(zeta_height, 0x48b);

src/video_core/engines/shader_bytecode.h

@@ -76,6 +76,7 @@ union Attribute {
         Position = 7,
         Attribute_0 = 8,
         Attribute_31 = 39,
+        PointCoord = 46,
         // This attribute contains a tuple of (~, ~, InstanceId, VertexId) when inside a vertex
         // shader, and a tuple of (TessCoord.x, TessCoord.y, TessCoord.z, ~) when inside a Tess Eval
         // shader.
@@ -147,6 +148,7 @@ enum class PredCondition : u64 {
     LessThanWithNan = 9,
     GreaterThanWithNan = 12,
     NotEqualWithNan = 13,
+    GreaterEqualWithNan = 14,
     // TODO(Subv): Other condition types
 };
 
@@ -242,6 +244,20 @@ enum class TextureType : u64 {
     TextureCube = 3,
 };
 
+enum class IpaInterpMode : u64 { Linear = 0, Perspective = 1, Flat = 2, Sc = 3 };
+enum class IpaSampleMode : u64 { Default = 0, Centroid = 1, Offset = 2 };
+
+struct IpaMode {
+    IpaInterpMode interpolation_mode;
+    IpaSampleMode sampling_mode;
+    inline bool operator==(const IpaMode& a) {
+        return (a.interpolation_mode == interpolation_mode) && (a.sampling_mode == sampling_mode);
+    }
+    inline bool operator!=(const IpaMode& a) {
+        return !((*this) == a);
+    }
+};
+
 union Instruction {
     Instruction& operator=(const Instruction& instr) {
         value = instr.value;
@@ -325,6 +341,16 @@ union Instruction {
     } alu;
 
     union {
+        BitField<51, 1, u64> saturate;
+        BitField<52, 2, IpaSampleMode> sample_mode;
+        BitField<54, 2, IpaInterpMode> interp_mode;
+    } ipa;
+
+    union {
+        BitField<39, 2, u64> tab5cb8_2;
+        BitField<41, 3, u64> tab5c68_1;
+        BitField<44, 2, u64> tab5c68_0;
+        BitField<47, 1, u64> cc;
         BitField<48, 1, u64> negate_b;
     } fmul;
 
@@ -338,6 +364,10 @@ union Instruction {
         BitField<49, 1, u64> negate_a;
     } alu_integer;
 
+    union {
+        BitField<40, 1, u64> invert;
+    } popc;
+
     union {
         BitField<39, 3, u64> pred;
         BitField<42, 1, u64> neg_pred;
@@ -388,8 +418,11 @@ union Instruction {
     } flow;
 
     union {
+        BitField<47, 1, u64> cc;
         BitField<48, 1, u64> negate_b;
         BitField<49, 1, u64> negate_c;
+        BitField<51, 2, u64> tab5980_1;
+        BitField<53, 2, u64> tab5980_0;
     } ffma;
 
     union {
@@ -498,6 +531,7 @@ union Instruction {
     union {
         BitField<0, 8, Register> gpr0;
         BitField<28, 8, Register> gpr28;
+        BitField<49, 1, u64> nodep;
         BitField<50, 3, u64> component_mask_selector;
         BitField<53, 4, u64> texture_info;
 
@@ -665,6 +699,9 @@ public:
         ISCADD_C, // Scale and Add
         ISCADD_R,
         ISCADD_IMM,
+        POPC_C,
+        POPC_R,
+        POPC_IMM,
         SEL_C,
         SEL_R,
         SEL_IMM,
@@ -886,6 +923,9 @@ private:
             INST("0100110000011---", Id::ISCADD_C, Type::ArithmeticInteger, "ISCADD_C"),
             INST("0101110000011---", Id::ISCADD_R, Type::ArithmeticInteger, "ISCADD_R"),
             INST("0011100-00011---", Id::ISCADD_IMM, Type::ArithmeticInteger, "ISCADD_IMM"),
+            INST("0100110000001---", Id::POPC_C, Type::ArithmeticInteger, "POPC_C"),
+            INST("0101110000001---", Id::POPC_R, Type::ArithmeticInteger, "POPC_R"),
+            INST("0011100-00001---", Id::POPC_IMM, Type::ArithmeticInteger, "POPC_IMM"),
             INST("0100110010100---", Id::SEL_C, Type::ArithmeticInteger, "SEL_C"),
             INST("0101110010100---", Id::SEL_R, Type::ArithmeticInteger, "SEL_R"),
             INST("0011100-10100---", Id::SEL_IMM, Type::ArithmeticInteger, "SEL_IMM"),

src/video_core/gpu.h

@@ -4,8 +4,8 @@
 
 #pragma once
 
+#include <array>
 #include <memory>
-#include <unordered_map>
 #include "common/common_types.h"
 #include "core/hle/service/nvflinger/buffer_queue.h"
 #include "video_core/memory_manager.h"
@@ -136,7 +136,7 @@ private:
     std::unique_ptr<Tegra::MemoryManager> memory_manager;
 
     /// Mapping of command subchannels to their bound engine ids.
-    std::unordered_map<u32, EngineID> bound_engines;
+    std::array<EngineID, 8> bound_engines = {};
 
     /// 3D engine
     std::unique_ptr<Engines::Maxwell3D> maxwell_3d;

src/video_core/rasterizer_cache.h

@@ -4,113 +4,86 @@
 
 #pragma once
 
-#include <unordered_map>
+#include <set>
+
 #include <boost/icl/interval_map.hpp>
-#include <boost/range/iterator_range.hpp>
+#include <boost/range/iterator_range_core.hpp>
 
 #include "common/common_types.h"
-#include "core/memory.h"
-#include "video_core/memory_manager.h"
+#include "core/core.h"
+#include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_base.h"
 
 template <class T>
 class RasterizerCache : NonCopyable {
 public:
     /// Mark the specified region as being invalidated
-    void InvalidateRegion(Tegra::GPUVAddr region_addr, size_t region_size) {
-        for (auto iter = cached_objects.cbegin(); iter != cached_objects.cend();) {
-            const auto& object{iter->second};
+    void InvalidateRegion(VAddr addr, u64 size) {
+        if (size == 0)
+            return;
 
-            ++iter;
+        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;
 
-            if (object->GetAddr() <= (region_addr + region_size) &&
-                region_addr <= (object->GetAddr() + object->GetSizeInBytes())) {
-                // Regions overlap, so invalidate
-                Unregister(object);
+                remove_objects.emplace(cached_object);
             }
         }
+
+        for (auto& remove_object : remove_objects) {
+            Unregister(remove_object);
+        }
+
+        remove_objects.clear();
+    }
+
+    /// Invalidates everything in the cache
+    void InvalidateAll() {
+        while (object_cache.begin() != object_cache.end()) {
+            Unregister(*object_cache.begin()->second.begin());
+        }
     }
 
 protected:
     /// Tries to get an object from the cache with the specified address
-    T TryGet(Tegra::GPUVAddr addr) const {
-        const auto& search{cached_objects.find(addr)};
-        if (search != cached_objects.end()) {
-            return search->second;
+    T TryGet(VAddr addr) const {
+        const ObjectInterval interval{addr};
+        for (auto& pair : boost::make_iterator_range(object_cache.equal_range(interval))) {
+            for (auto& cached_object : pair.second) {
+                if (cached_object->GetAddr() == addr) {
+                    return cached_object;
+                }
+            }
         }
-
         return nullptr;
     }
 
-    /// Gets a reference to the cache
-    const std::unordered_map<Tegra::GPUVAddr, T>& GetCache() const {
-        return cached_objects;
-    }
-
     /// Register an object into the cache
     void Register(const T& object) {
-        const auto& search{cached_objects.find(object->GetAddr())};
-        if (search != cached_objects.end()) {
-            // Registered already
-            return;
-        }
-
-        cached_objects[object->GetAddr()] = object;
-        UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), 1);
+        object_cache.add({GetInterval(object), ObjectSet{object}});
+        auto& rasterizer = Core::System::GetInstance().Renderer().Rasterizer();
+        rasterizer.UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), 1);
     }
 
     /// Unregisters an object from the cache
     void Unregister(const T& object) {
-        const auto& search{cached_objects.find(object->GetAddr())};
-        if (search == cached_objects.end()) {
-            // Unregistered already
-            return;
-        }
-
-        UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), -1);
-        cached_objects.erase(search);
+        auto& rasterizer = Core::System::GetInstance().Renderer().Rasterizer();
+        rasterizer.UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), -1);
+        object_cache.subtract({GetInterval(object), ObjectSet{object}});
     }
 
 private:
-    using PageMap = boost::icl::interval_map<u64, int>;
+    using ObjectSet = std::set<T>;
+    using ObjectCache = boost::icl::interval_map<VAddr, ObjectSet>;
+    using ObjectInterval = typename ObjectCache::interval_type;
 
-    template <typename Map, typename Interval>
-    constexpr auto RangeFromInterval(Map& map, const Interval& interval) {
-        return boost::make_iterator_range(map.equal_range(interval));
+    static auto GetInterval(const T& object) {
+        return ObjectInterval::right_open(object->GetAddr(),
+                                          object->GetAddr() + object->GetSizeInBytes());
     }
 
-    /// Increase/decrease the number of object in pages touching the specified region
-    void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta) {
-        const u64 page_start{addr >> Tegra::MemoryManager::PAGE_BITS};
-        const u64 page_end{(addr + size) >> Tegra::MemoryManager::PAGE_BITS};
-
-        // Interval maps will erase segments if count reaches 0, so if delta is negative we have to
-        // subtract after iterating
-        const auto pages_interval = PageMap::interval_type::right_open(page_start, page_end);
-        if (delta > 0)
-            cached_pages.add({pages_interval, delta});
-
-        for (const auto& pair : RangeFromInterval(cached_pages, pages_interval)) {
-            const auto interval = pair.first & pages_interval;
-            const int count = pair.second;
-
-            const Tegra::GPUVAddr interval_start_addr = boost::icl::first(interval)
-                                                        << Tegra::MemoryManager::PAGE_BITS;
-            const Tegra::GPUVAddr interval_end_addr = boost::icl::last_next(interval)
-                                                      << Tegra::MemoryManager::PAGE_BITS;
-            const u64 interval_size = interval_end_addr - interval_start_addr;
-
-            if (delta > 0 && count == delta)
-                Memory::RasterizerMarkRegionCached(interval_start_addr, interval_size, true);
-            else if (delta < 0 && count == -delta)
-                Memory::RasterizerMarkRegionCached(interval_start_addr, interval_size, false);
-            else
-                ASSERT(count >= 0);
-        }
-
-        if (delta < 0)
-            cached_pages.add({pages_interval, delta});
-    }
-
-    std::unordered_map<Tegra::GPUVAddr, T> cached_objects;
-    PageMap cached_pages;
+    ObjectCache object_cache;
+    ObjectSet remove_objects;
 };

src/video_core/rasterizer_interface.h

@@ -27,14 +27,14 @@ public:
     virtual void FlushAll() = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    virtual void FlushRegion(Tegra::GPUVAddr addr, u64 size) = 0;
+    virtual void FlushRegion(VAddr addr, u64 size) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be invalidated
-    virtual void InvalidateRegion(Tegra::GPUVAddr addr, u64 size) = 0;
+    virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     /// and invalidated
-    virtual void FlushAndInvalidateRegion(Tegra::GPUVAddr addr, u64 size) = 0;
+    virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
 
     /// Attempt to use a faster method to perform a display transfer with is_texture_copy = 0
     virtual bool AccelerateDisplayTransfer(const void* config) {
@@ -60,5 +60,8 @@ public:
     virtual bool AccelerateDrawBatch(bool is_indexed) {
         return false;
     }
+
+    /// Increase/decrease the number of object in pages touching the specified region
+    virtual void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta) {}
 };
 } // namespace VideoCore

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -0,0 +1,92 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include <memory>
+
+#include "common/alignment.h"
+#include "core/core.h"
+#include "core/memory.h"
+#include "video_core/renderer_opengl/gl_buffer_cache.h"
+
+namespace OpenGL {
+
+OGLBufferCache::OGLBufferCache(size_t size) : stream_buffer(GL_ARRAY_BUFFER, size) {}
+
+GLintptr OGLBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, size_t size, size_t alignment,
+                                      bool cache) {
+    auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
+    const boost::optional<VAddr> cpu_addr{memory_manager.GpuToCpuAddress(gpu_addr)};
+
+    // Cache management is a big overhead, so only cache entries with a given size.
+    // TODO: Figure out which size is the best for given games.
+    cache &= size >= 2048;
+
+    if (cache) {
+        auto entry = TryGet(*cpu_addr);
+        if (entry) {
+            if (entry->size >= size && entry->alignment == alignment) {
+                return entry->offset;
+            }
+            Unregister(entry);
+        }
+    }
+
+    AlignBuffer(alignment);
+    GLintptr uploaded_offset = buffer_offset;
+
+    Memory::ReadBlock(*cpu_addr, buffer_ptr, size);
+
+    buffer_ptr += size;
+    buffer_offset += size;
+
+    if (cache) {
+        auto entry = std::make_shared<CachedBufferEntry>();
+        entry->offset = uploaded_offset;
+        entry->size = size;
+        entry->alignment = alignment;
+        entry->addr = *cpu_addr;
+        Register(entry);
+    }
+
+    return uploaded_offset;
+}
+
+GLintptr OGLBufferCache::UploadHostMemory(const void* raw_pointer, size_t size, size_t alignment) {
+    AlignBuffer(alignment);
+    std::memcpy(buffer_ptr, raw_pointer, size);
+    GLintptr uploaded_offset = buffer_offset;
+
+    buffer_ptr += size;
+    buffer_offset += size;
+    return uploaded_offset;
+}
+
+void OGLBufferCache::Map(size_t max_size) {
+    bool invalidate;
+    std::tie(buffer_ptr, buffer_offset_base, invalidate) =
+        stream_buffer.Map(static_cast<GLsizeiptr>(max_size), 4);
+    buffer_offset = buffer_offset_base;
+
+    if (invalidate) {
+        InvalidateAll();
+    }
+}
+void OGLBufferCache::Unmap() {
+    stream_buffer.Unmap(buffer_offset - buffer_offset_base);
+}
+
+GLuint OGLBufferCache::GetHandle() const {
+    return stream_buffer.GetHandle();
+}
+
+void OGLBufferCache::AlignBuffer(size_t alignment) {
+    // Align the offset, not the mapped pointer
+    GLintptr offset_aligned =
+        static_cast<GLintptr>(Common::AlignUp(static_cast<size_t>(buffer_offset), alignment));
+    buffer_ptr += offset_aligned - buffer_offset;
+    buffer_offset = offset_aligned;
+}
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -0,0 +1,57 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstddef>
+#include <memory>
+
+#include "common/common_types.h"
+#include "video_core/rasterizer_cache.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
+#include "video_core/renderer_opengl/gl_stream_buffer.h"
+
+namespace OpenGL {
+
+struct CachedBufferEntry final {
+    VAddr GetAddr() const {
+        return addr;
+    }
+
+    size_t GetSizeInBytes() const {
+        return size;
+    }
+
+    VAddr addr;
+    size_t size;
+    GLintptr offset;
+    size_t alignment;
+};
+
+class OGLBufferCache final : public RasterizerCache<std::shared_ptr<CachedBufferEntry>> {
+public:
+    explicit OGLBufferCache(size_t size);
+
+    GLintptr UploadMemory(Tegra::GPUVAddr gpu_addr, size_t size, size_t alignment = 4,
+                          bool cache = true);
+
+    GLintptr UploadHostMemory(const void* raw_pointer, size_t size, size_t alignment = 4);
+
+    void Map(size_t max_size);
+    void Unmap();
+
+    GLuint GetHandle() const;
+
+protected:
+    void AlignBuffer(size_t alignment);
+
+private:
+    OGLStreamBuffer stream_buffer;
+
+    u8* buffer_ptr = nullptr;
+    GLintptr buffer_offset = 0;
+    GLintptr buffer_offset_base = 0;
+};
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -33,14 +33,17 @@ using PixelFormat = SurfaceParams::PixelFormat;
 using SurfaceType = SurfaceParams::SurfaceType;
 
 MICROPROFILE_DEFINE(OpenGL_VAO, "OpenGL", "Vertex Array Setup", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_VS, "OpenGL", "Vertex Shader Setup", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_FS, "OpenGL", "Fragment Shader Setup", MP_RGB(128, 128, 192));
+MICROPROFILE_DEFINE(OpenGL_Shader, "OpenGL", "Shader Setup", MP_RGB(128, 128, 192));
+MICROPROFILE_DEFINE(OpenGL_UBO, "OpenGL", "Const Buffer Setup", MP_RGB(128, 128, 192));
+MICROPROFILE_DEFINE(OpenGL_Index, "OpenGL", "Index Buffer Setup", MP_RGB(128, 128, 192));
+MICROPROFILE_DEFINE(OpenGL_Texture, "OpenGL", "Texture Setup", MP_RGB(128, 128, 192));
+MICROPROFILE_DEFINE(OpenGL_Framebuffer, "OpenGL", "Framebuffer Setup", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_Drawing, "OpenGL", "Drawing", MP_RGB(128, 128, 192));
-MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(100, 100, 255));
+MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100));
 
 RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo& info)
-    : emu_window{window}, screen_info{info}, stream_buffer(GL_ARRAY_BUFFER, STREAM_BUFFER_SIZE) {
+    : emu_window{window}, screen_info{info}, buffer_cache(STREAM_BUFFER_SIZE) {
     // Create sampler objects
     for (size_t i = 0; i < texture_samplers.size(); ++i) {
         texture_samplers[i].Create();
@@ -67,28 +70,13 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo
     // Clipping plane 0 is always enabled for PICA fixed clip plane z <= 0
     state.clip_distance[0] = true;
 
-    // Generate VAO and UBO
-    sw_vao.Create();
-    uniform_buffer.Create();
-
-    state.draw.vertex_array = sw_vao.handle;
-    state.draw.uniform_buffer = uniform_buffer.handle;
-    state.Apply();
-
     // Create render framebuffer
     framebuffer.Create();
 
-    hw_vao.Create();
-
-    state.draw.vertex_buffer = stream_buffer.GetHandle();
-
     shader_program_manager = std::make_unique<GLShader::ProgramManager>();
     state.draw.shader_program = 0;
-    state.draw.vertex_array = hw_vao.handle;
     state.Apply();
 
-    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, stream_buffer.GetHandle());
-
     glEnable(GL_BLEND);
 
     glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &uniform_buffer_alignment);
@@ -98,14 +86,60 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo
 
 RasterizerOpenGL::~RasterizerOpenGL() {}
 
-std::pair<u8*, GLintptr> RasterizerOpenGL::SetupVertexArrays(u8* array_ptr,
-                                                             GLintptr buffer_offset) {
+void RasterizerOpenGL::SetupVertexArrays() {
     MICROPROFILE_SCOPE(OpenGL_VAO);
     const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
     const auto& regs = gpu.regs;
 
-    state.draw.vertex_array = hw_vao.handle;
-    state.draw.vertex_buffer = stream_buffer.GetHandle();
+    auto [iter, is_cache_miss] = vertex_array_cache.try_emplace(regs.vertex_attrib_format);
+    auto& VAO = iter->second;
+
+    if (is_cache_miss) {
+        VAO.Create();
+        state.draw.vertex_array = VAO.handle;
+        state.Apply();
+
+        // The index buffer binding is stored within the VAO. Stupid OpenGL, but easy to work
+        // around.
+        glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer_cache.GetHandle());
+
+        // Use the vertex array as-is, assumes that the data is formatted correctly for OpenGL.
+        // Enables the first 16 vertex attributes always, as we don't know which ones are actually
+        // used until shader time. Note, Tegra technically supports 32, but we're capping this to 16
+        // for now to avoid OpenGL errors.
+        // TODO(Subv): Analyze the shader to identify which attributes are actually used and don't
+        // assume every shader uses them all.
+        for (unsigned index = 0; index < 16; ++index) {
+            const auto& attrib = regs.vertex_attrib_format[index];
+
+            // Ignore invalid attributes.
+            if (!attrib.IsValid())
+                continue;
+
+            const auto& buffer = regs.vertex_array[attrib.buffer];
+            LOG_TRACE(HW_GPU,
+                      "vertex attrib {}, count={}, size={}, type={}, offset={}, normalize={}",
+                      index, attrib.ComponentCount(), attrib.SizeString(), attrib.TypeString(),
+                      attrib.offset.Value(), attrib.IsNormalized());
+
+            ASSERT(buffer.IsEnabled());
+
+            glEnableVertexAttribArray(index);
+            if (attrib.type == Tegra::Engines::Maxwell3D::Regs::VertexAttribute::Type::SignedInt ||
+                attrib.type ==
+                    Tegra::Engines::Maxwell3D::Regs::VertexAttribute::Type::UnsignedInt) {
+                glVertexAttribIFormat(index, attrib.ComponentCount(),
+                                      MaxwellToGL::VertexType(attrib), attrib.offset);
+            } else {
+                glVertexAttribFormat(index, attrib.ComponentCount(),
+                                     MaxwellToGL::VertexType(attrib),
+                                     attrib.IsNormalized() ? GL_TRUE : GL_FALSE, attrib.offset);
+            }
+            glVertexAttribBinding(index, attrib.buffer);
+        }
+    }
+    state.draw.vertex_array = VAO.handle;
+    state.draw.vertex_buffer = buffer_cache.GetHandle();
     state.Apply();
 
     // Upload all guest vertex arrays sequentially to our buffer
@@ -124,12 +158,10 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupVertexArrays(u8* array_ptr,
         ASSERT(end > start);
         u64 size = end - start + 1;
 
-        GLintptr vertex_buffer_offset;
-        std::tie(array_ptr, buffer_offset, vertex_buffer_offset) =
-            UploadMemory(array_ptr, buffer_offset, start, size);
+        GLintptr vertex_buffer_offset = buffer_cache.UploadMemory(start, size);
 
         // Bind the vertex array to the buffer at the current offset.
-        glBindVertexBuffer(index, stream_buffer.GetHandle(), vertex_buffer_offset,
+        glBindVertexBuffer(index, buffer_cache.GetHandle(), vertex_buffer_offset,
                            vertex_array.stride);
 
         if (regs.instanced_arrays.IsInstancingEnabled(index) && vertex_array.divisor != 0) {
@@ -142,43 +174,10 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupVertexArrays(u8* array_ptr,
             glVertexBindingDivisor(index, 0);
         }
     }
-
-    // Use the vertex array as-is, assumes that the data is formatted correctly for OpenGL.
-    // Enables the first 16 vertex attributes always, as we don't know which ones are actually used
-    // until shader time. Note, Tegra technically supports 32, but we're capping this to 16 for now
-    // to avoid OpenGL errors.
-    // TODO(Subv): Analyze the shader to identify which attributes are actually used and don't
-    // assume every shader uses them all.
-    for (unsigned index = 0; index < 16; ++index) {
-        auto& attrib = regs.vertex_attrib_format[index];
-
-        // Ignore invalid attributes.
-        if (!attrib.IsValid())
-            continue;
-
-        auto& buffer = regs.vertex_array[attrib.buffer];
-        LOG_TRACE(HW_GPU, "vertex attrib {}, count={}, size={}, type={}, offset={}, normalize={}",
-                  index, attrib.ComponentCount(), attrib.SizeString(), attrib.TypeString(),
-                  attrib.offset.Value(), attrib.IsNormalized());
-
-        ASSERT(buffer.IsEnabled());
-
-        glEnableVertexAttribArray(index);
-        if (attrib.type == Tegra::Engines::Maxwell3D::Regs::VertexAttribute::Type::SignedInt ||
-            attrib.type == Tegra::Engines::Maxwell3D::Regs::VertexAttribute::Type::UnsignedInt) {
-            glVertexAttribIFormat(index, attrib.ComponentCount(), MaxwellToGL::VertexType(attrib),
-                                  attrib.offset);
-        } else {
-            glVertexAttribFormat(index, attrib.ComponentCount(), MaxwellToGL::VertexType(attrib),
-                                 attrib.IsNormalized() ? GL_TRUE : GL_FALSE, attrib.offset);
-        }
-        glVertexAttribBinding(index, attrib.buffer);
-    }
-
-    return {array_ptr, buffer_offset};
 }
 
-std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr buffer_offset) {
+void RasterizerOpenGL::SetupShaders() {
+    MICROPROFILE_SCOPE(OpenGL_Shader);
     auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
 
     // Next available bindpoints to use when uploading the const buffers and textures to the GLSL
@@ -195,21 +194,15 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr
             continue;
         }
 
-        std::tie(buffer_ptr, buffer_offset) =
-            AlignBuffer(buffer_ptr, buffer_offset, static_cast<size_t>(uniform_buffer_alignment));
-
         const size_t stage{index == 0 ? 0 : index - 1}; // Stage indices are 0 - 5
 
         GLShader::MaxwellUniformData ubo{};
         ubo.SetFromRegs(gpu.state.shader_stages[stage]);
-        std::memcpy(buffer_ptr, &ubo, sizeof(ubo));
+        GLintptr offset = buffer_cache.UploadHostMemory(
+            &ubo, sizeof(ubo), static_cast<size_t>(uniform_buffer_alignment));
 
         // Bind the buffer
-        glBindBufferRange(GL_UNIFORM_BUFFER, stage, stream_buffer.GetHandle(), buffer_offset,
-                          sizeof(ubo));
-
-        buffer_ptr += sizeof(ubo);
-        buffer_offset += sizeof(ubo);
+        glBindBufferRange(GL_UNIFORM_BUFFER, stage, buffer_cache.GetHandle(), offset, sizeof(ubo));
 
         Shader shader{shader_cache.GetStageProgram(program)};
 
@@ -230,9 +223,8 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr
         }
 
         // Configure the const buffers for this shader stage.
-        std::tie(buffer_ptr, buffer_offset, current_constbuffer_bindpoint) =
-            SetupConstBuffers(buffer_ptr, buffer_offset, static_cast<Maxwell::ShaderStage>(stage),
-                              shader, current_constbuffer_bindpoint);
+        current_constbuffer_bindpoint = SetupConstBuffers(static_cast<Maxwell::ShaderStage>(stage),
+                                                          shader, current_constbuffer_bindpoint);
 
         // Configure the textures for this shader stage.
         current_texture_bindpoint = SetupTextures(static_cast<Maxwell::ShaderStage>(stage), shader,
@@ -246,8 +238,6 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupShaders(u8* buffer_ptr, GLintptr
     }
 
     shader_program_manager->UseTrivialGeometryShader();
-
-    return {buffer_ptr, buffer_offset};
 }
 
 size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
@@ -274,9 +264,45 @@ bool RasterizerOpenGL::AccelerateDrawBatch(bool is_indexed) {
     return true;
 }
 
+template <typename Map, typename Interval>
+static constexpr auto RangeFromInterval(Map& map, const Interval& interval) {
+    return boost::make_iterator_range(map.equal_range(interval));
+}
+
+void RasterizerOpenGL::UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {
+    const u64 page_start{addr >> Memory::PAGE_BITS};
+    const u64 page_end{(addr + size + Memory::PAGE_SIZE - 1) >> Memory::PAGE_BITS};
+
+    // Interval maps will erase segments if count reaches 0, so if delta is negative we have to
+    // subtract after iterating
+    const auto pages_interval = CachedPageMap::interval_type::right_open(page_start, page_end);
+    if (delta > 0)
+        cached_pages.add({pages_interval, delta});
+
+    for (const auto& pair : RangeFromInterval(cached_pages, pages_interval)) {
+        const auto interval = pair.first & pages_interval;
+        const int count = pair.second;
+
+        const VAddr interval_start_addr = boost::icl::first(interval) << Memory::PAGE_BITS;
+        const VAddr interval_end_addr = boost::icl::last_next(interval) << Memory::PAGE_BITS;
+        const u64 interval_size = interval_end_addr - interval_start_addr;
+
+        if (delta > 0 && count == delta)
+            Memory::RasterizerMarkRegionCached(interval_start_addr, interval_size, true);
+        else if (delta < 0 && count == -delta)
+            Memory::RasterizerMarkRegionCached(interval_start_addr, interval_size, false);
+        else
+            ASSERT(count >= 0);
+    }
+
+    if (delta < 0)
+        cached_pages.add({pages_interval, delta});
+}
+
 std::pair<Surface, Surface> RasterizerOpenGL::ConfigureFramebuffers(bool using_color_fb,
                                                                     bool using_depth_fb,
                                                                     bool preserve_contents) {
+    MICROPROFILE_SCOPE(OpenGL_Framebuffer);
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
 
     if (regs.rt[0].format == Tegra::RenderTargetFormat::NONE) {
@@ -397,41 +423,6 @@ void RasterizerOpenGL::Clear() {
     glClearStencil(regs.clear_stencil);
 
     glClear(clear_mask);
-
-    // Mark framebuffer surfaces as dirty
-    if (Settings::values.use_accurate_framebuffers) {
-        if (dirty_color_surface != nullptr) {
-            res_cache.FlushSurface(dirty_color_surface);
-        }
-        if (dirty_depth_surface != nullptr) {
-            res_cache.FlushSurface(dirty_depth_surface);
-        }
-    }
-}
-
-std::pair<u8*, GLintptr> RasterizerOpenGL::AlignBuffer(u8* buffer_ptr, GLintptr buffer_offset,
-                                                       size_t alignment) {
-    // Align the offset, not the mapped pointer
-    GLintptr offset_aligned =
-        static_cast<GLintptr>(Common::AlignUp(static_cast<size_t>(buffer_offset), alignment));
-    return {buffer_ptr + (offset_aligned - buffer_offset), offset_aligned};
-}
-
-std::tuple<u8*, GLintptr, GLintptr> RasterizerOpenGL::UploadMemory(u8* buffer_ptr,
-                                                                   GLintptr buffer_offset,
-                                                                   Tegra::GPUVAddr gpu_addr,
-                                                                   size_t size, size_t alignment) {
-    std::tie(buffer_ptr, buffer_offset) = AlignBuffer(buffer_ptr, buffer_offset, alignment);
-    GLintptr uploaded_offset = buffer_offset;
-
-    auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
-    const boost::optional<VAddr> cpu_addr{memory_manager.GpuToCpuAddress(gpu_addr)};
-    Memory::ReadBlock(*cpu_addr, buffer_ptr, size);
-
-    buffer_ptr += size;
-    buffer_offset += size;
-
-    return {buffer_ptr, buffer_offset, uploaded_offset};
 }
 
 void RasterizerOpenGL::DrawArrays() {
@@ -459,7 +450,7 @@ void RasterizerOpenGL::DrawArrays() {
     const bool is_indexed = accelerate_draw == AccelDraw::Indexed;
     const u64 index_buffer_size{regs.index_array.count * regs.index_array.FormatSizeInBytes()};
 
-    state.draw.vertex_buffer = stream_buffer.GetHandle();
+    state.draw.vertex_buffer = buffer_cache.GetHandle();
     state.Apply();
 
     size_t buffer_size = CalculateVertexArraysSize();
@@ -476,24 +467,21 @@ void RasterizerOpenGL::DrawArrays() {
     // Add space for at least 18 constant buffers
     buffer_size += Maxwell::MaxConstBuffers * (MaxConstbufferSize + uniform_buffer_alignment);
 
-    u8* buffer_ptr;
-    GLintptr buffer_offset;
-    std::tie(buffer_ptr, buffer_offset, std::ignore) =
-        stream_buffer.Map(static_cast<GLsizeiptr>(buffer_size), 4);
-    u8* buffer_ptr_base = buffer_ptr;
+    buffer_cache.Map(buffer_size);
 
-    std::tie(buffer_ptr, buffer_offset) = SetupVertexArrays(buffer_ptr, buffer_offset);
+    SetupVertexArrays();
 
     // If indexed mode, copy the index buffer
     GLintptr index_buffer_offset = 0;
     if (is_indexed) {
-        std::tie(buffer_ptr, buffer_offset, index_buffer_offset) = UploadMemory(
-            buffer_ptr, buffer_offset, regs.index_array.StartAddress(), index_buffer_size);
+        MICROPROFILE_SCOPE(OpenGL_Index);
+        index_buffer_offset =
+            buffer_cache.UploadMemory(regs.index_array.StartAddress(), index_buffer_size);
     }
 
-    std::tie(buffer_ptr, buffer_offset) = SetupShaders(buffer_ptr, buffer_offset);
+    SetupShaders();
 
-    stream_buffer.Unmap(buffer_ptr - buffer_ptr_base);
+    buffer_cache.Unmap();
 
     shader_program_manager->ApplyTo(state);
     state.Apply();
@@ -522,16 +510,6 @@ void RasterizerOpenGL::DrawArrays() {
         texture_unit.Unbind();
     }
     state.Apply();
-
-    // Mark framebuffer surfaces as dirty
-    if (Settings::values.use_accurate_framebuffers) {
-        if (dirty_color_surface != nullptr) {
-            res_cache.FlushSurface(dirty_color_surface);
-        }
-        if (dirty_depth_surface != nullptr) {
-            res_cache.FlushSurface(dirty_depth_surface);
-        }
-    }
 }
 
 void RasterizerOpenGL::NotifyMaxwellRegisterChanged(u32 method) {}
@@ -540,17 +518,18 @@ void RasterizerOpenGL::FlushAll() {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
 }
 
-void RasterizerOpenGL::FlushRegion(Tegra::GPUVAddr addr, u64 size) {
+void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
 }
 
-void RasterizerOpenGL::InvalidateRegion(Tegra::GPUVAddr addr, u64 size) {
+void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     res_cache.InvalidateRegion(addr, size);
     shader_cache.InvalidateRegion(addr, size);
+    buffer_cache.InvalidateRegion(addr, size);
 }
 
-void RasterizerOpenGL::FlushAndInvalidateRegion(Tegra::GPUVAddr addr, u64 size) {
+void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     InvalidateRegion(addr, size);
 }
@@ -637,11 +616,9 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr
     }
 }
 
-std::tuple<u8*, GLintptr, u32> RasterizerOpenGL::SetupConstBuffers(u8* buffer_ptr,
-                                                                   GLintptr buffer_offset,
-                                                                   Maxwell::ShaderStage stage,
-                                                                   Shader& shader,
-                                                                   u32 current_bindpoint) {
+u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, Shader& shader,
+                                        u32 current_bindpoint) {
+    MICROPROFILE_SCOPE(OpenGL_UBO);
     const auto& gpu = Core::System::GetInstance().GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
     const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<size_t>(stage)];
@@ -677,26 +654,25 @@ std::tuple<u8*, GLintptr, u32> RasterizerOpenGL::SetupConstBuffers(u8* buffer_pt
         size = Common::AlignUp(size, sizeof(GLvec4));
         ASSERT_MSG(size <= MaxConstbufferSize, "Constbuffer too big");
 
-        GLintptr const_buffer_offset;
-        std::tie(buffer_ptr, buffer_offset, const_buffer_offset) =
-            UploadMemory(buffer_ptr, buffer_offset, buffer.address, size,
-                         static_cast<size_t>(uniform_buffer_alignment));
+        GLintptr const_buffer_offset = buffer_cache.UploadMemory(
+            buffer.address, size, static_cast<size_t>(uniform_buffer_alignment));
 
         glBindBufferRange(GL_UNIFORM_BUFFER, current_bindpoint + bindpoint,
-                          stream_buffer.GetHandle(), const_buffer_offset, size);
+                          buffer_cache.GetHandle(), const_buffer_offset, size);
 
         // Now configure the bindpoint of the buffer inside the shader
         glUniformBlockBinding(shader->GetProgramHandle(),
-                              shader->GetProgramResourceIndex(used_buffer.GetName()),
+                              shader->GetProgramResourceIndex(used_buffer),
                               current_bindpoint + bindpoint);
     }
 
     state.Apply();
 
-    return {buffer_ptr, buffer_offset, current_bindpoint + static_cast<u32>(entries.size())};
+    return current_bindpoint + static_cast<u32>(entries.size());
 }
 
 u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader, u32 current_unit) {
+    MICROPROFILE_SCOPE(OpenGL_Texture);
     const auto& gpu = Core::System::GetInstance().GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
     const auto& entries = shader->GetShaderEntries().texture_samplers;
@@ -710,7 +686,7 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader,
 
         // Bind the uniform to the sampler.
 
-        glProgramUniform1i(shader->GetProgramHandle(), shader->GetUniformLocation(entry.GetName()),
+        glProgramUniform1i(shader->GetProgramHandle(), shader->GetUniformLocation(entry),
                            current_bindpoint);
 
         const auto texture = maxwell3d.GetStageTexture(entry.GetStage(), entry.GetOffset());

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -6,15 +6,22 @@
 
 #include <array>
 #include <cstddef>
+#include <map>
 #include <memory>
 #include <tuple>
 #include <utility>
 #include <vector>
+
+#include <boost/icl/interval_map.hpp>
+#include <boost/range/iterator_range.hpp>
 #include <glad/glad.h>
+
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/memory_manager.h"
+#include "video_core/rasterizer_cache.h"
 #include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_opengl/gl_buffer_cache.h"
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_cache.h"
@@ -40,15 +47,16 @@ public:
     void Clear() override;
     void NotifyMaxwellRegisterChanged(u32 method) override;
     void FlushAll() override;
-    void FlushRegion(Tegra::GPUVAddr addr, u64 size) override;
-    void InvalidateRegion(Tegra::GPUVAddr addr, u64 size) override;
-    void FlushAndInvalidateRegion(Tegra::GPUVAddr addr, u64 size) override;
+    void FlushRegion(VAddr addr, u64 size) override;
+    void InvalidateRegion(VAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     bool AccelerateDisplayTransfer(const void* config) override;
     bool AccelerateTextureCopy(const void* config) override;
     bool AccelerateFill(const void* config) override;
     bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
                            u32 pixel_stride) override;
     bool AccelerateDrawBatch(bool is_indexed) override;
+    void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta) override;
 
     /// OpenGL shader generated for a given Maxwell register state
     struct MaxwellShader {
@@ -104,9 +112,8 @@ private:
      * @param current_bindpoint The offset at which to start counting new buffer bindpoints.
      * @returns The next available bindpoint for use in the next shader stage.
      */
-    std::tuple<u8*, GLintptr, u32> SetupConstBuffers(
-        u8* buffer_ptr, GLintptr buffer_offset, Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
-        Shader& shader, u32 current_bindpoint);
+    u32 SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, Shader& shader,
+                          u32 current_bindpoint);
 
     /*
      * Configures the current textures to use for the draw command.
@@ -162,31 +169,29 @@ private:
     ScreenInfo& screen_info;
 
     std::unique_ptr<GLShader::ProgramManager> shader_program_manager;
-    OGLVertexArray sw_vao;
-    OGLVertexArray hw_vao;
+    std::map<std::array<Tegra::Engines::Maxwell3D::Regs::VertexAttribute,
+                        Tegra::Engines::Maxwell3D::Regs::NumVertexAttributes>,
+             OGLVertexArray>
+        vertex_array_cache;
 
     std::array<SamplerInfo, GLShader::NumTextureSamplers> texture_samplers;
 
     static constexpr size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;
-    OGLStreamBuffer stream_buffer;
-    OGLBuffer uniform_buffer;
+    OGLBufferCache buffer_cache;
     OGLFramebuffer framebuffer;
     GLint uniform_buffer_alignment;
 
     size_t CalculateVertexArraysSize() const;
 
-    std::pair<u8*, GLintptr> SetupVertexArrays(u8* array_ptr, GLintptr buffer_offset);
+    void SetupVertexArrays();
 
-    std::pair<u8*, GLintptr> SetupShaders(u8* buffer_ptr, GLintptr buffer_offset);
-
-    std::pair<u8*, GLintptr> AlignBuffer(u8* buffer_ptr, GLintptr buffer_offset, size_t alignment);
-
-    std::tuple<u8*, GLintptr, GLintptr> UploadMemory(u8* buffer_ptr, GLintptr buffer_offset,
-                                                     Tegra::GPUVAddr gpu_addr, size_t size,
-                                                     size_t alignment = 4);
+    void SetupShaders();
 
     enum class AccelDraw { Disabled, Arrays, Indexed };
     AccelDraw accelerate_draw = AccelDraw::Disabled;
+
+    using CachedPageMap = boost::icl::interval_map<u64, int>;
+    CachedPageMap cached_pages;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -33,11 +33,16 @@ struct FormatTuple {
     bool compressed;
 };
 
+static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
+    auto& gpu{Core::System::GetInstance().GPU()};
+    const auto cpu_addr{gpu.MemoryManager().GpuToCpuAddress(gpu_addr)};
+    return cpu_addr ? *cpu_addr : 0;
+}
+
 /*static*/ SurfaceParams SurfaceParams::CreateForTexture(
     const Tegra::Texture::FullTextureInfo& config) {
-
     SurfaceParams params{};
-    params.addr = config.tic.Address();
+    params.addr = TryGetCpuAddr(config.tic.Address());
     params.is_tiled = config.tic.IsTiled();
     params.block_height = params.is_tiled ? config.tic.BlockHeight() : 0,
     params.pixel_format =
@@ -55,9 +60,8 @@ struct FormatTuple {
 
 /*static*/ SurfaceParams SurfaceParams::CreateForFramebuffer(
     const Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig& config) {
-
     SurfaceParams params{};
-    params.addr = config.Address();
+    params.addr = TryGetCpuAddr(config.Address());
     params.is_tiled = true;
     params.block_height = Tegra::Texture::TICEntry::DefaultBlockHeight;
     params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
@@ -75,9 +79,8 @@ struct FormatTuple {
 /*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer(u32 zeta_width, u32 zeta_height,
                                                              Tegra::GPUVAddr zeta_address,
                                                              Tegra::DepthFormat format) {
-
     SurfaceParams params{};
-    params.addr = zeta_address;
+    params.addr = TryGetCpuAddr(zeta_address);
     params.is_tiled = true;
     params.block_height = Tegra::Texture::TICEntry::DefaultBlockHeight;
     params.pixel_format = PixelFormatFromDepthFormat(format);
@@ -120,7 +123,11 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
      true},                                                                     // DXN2UNORM
     {GL_COMPRESSED_SIGNED_RG_RGTC2, GL_RG, GL_INT, ComponentType::SNorm, true}, // DXN2SNORM
     {GL_COMPRESSED_RGBA_BPTC_UNORM_ARB, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
-     true},                                                                    // BC7U
+     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,
+     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
     {GL_RG8, GL_RG, GL_BYTE, ComponentType::SNorm, false},                     // G8R8S
@@ -167,11 +174,6 @@ static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType
     return format;
 }
 
-VAddr SurfaceParams::GetCpuAddr() const {
-    auto& gpu = Core::System::GetInstance().GPU();
-    return *gpu.MemoryManager().GpuToCpuAddress(addr);
-}
-
 static bool IsPixelFormatASTC(PixelFormat format) {
     switch (format) {
     case PixelFormat::ASTC_2D_4X4:
@@ -210,39 +212,36 @@ static bool IsFormatBCn(PixelFormat format) {
     case PixelFormat::DXN2SNORM:
     case PixelFormat::DXN2UNORM:
     case PixelFormat::BC7U:
+    case PixelFormat::BC6H_UF16:
+    case PixelFormat::BC6H_SF16:
         return true;
     }
     return false;
 }
 
 template <bool morton_to_gl, PixelFormat format>
-void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_buffer,
-                Tegra::GPUVAddr addr) {
+void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_buffer, VAddr addr) {
     constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
     constexpr u32 gl_bytes_per_pixel = CachedSurface::GetGLBytesPerPixel(format);
-    auto& gpu = Core::System::GetInstance().GPU();
 
     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(*gpu.MemoryManager().GpuToCpuAddress(addr), tile_size,
-                                             bytes_per_pixel, stride, height, block_height);
+        const std::vector<u8> data = Tegra::Texture::UnswizzleTexture(
+            addr, tile_size, bytes_per_pixel, stride, height, block_height);
         const size_t size_to_copy{std::min(gl_buffer.size(), data.size())};
         gl_buffer.assign(data.begin(), data.begin() + 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(*gpu.MemoryManager().GpuToCpuAddress(addr)), gl_buffer.data(),
-            morton_to_gl);
+        VideoCore::MortonCopyPixels128(stride, height, bytes_per_pixel, gl_bytes_per_pixel,
+                                       Memory::GetPointer(addr), gl_buffer.data(), morton_to_gl);
     }
 }
 
-static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPUVAddr),
+static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, VAddr),
                             SurfaceParams::MaxPixelFormat>
     morton_to_gl_fns = {
         // clang-format off
@@ -266,6 +265,8 @@ static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPU
         MortonCopy<true, PixelFormat::DXN2UNORM>,
         MortonCopy<true, PixelFormat::DXN2SNORM>,
         MortonCopy<true, PixelFormat::BC7U>,
+        MortonCopy<true, PixelFormat::BC6H_UF16>,
+        MortonCopy<true, PixelFormat::BC6H_SF16>,
         MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
         MortonCopy<true, PixelFormat::G8R8U>,
         MortonCopy<true, PixelFormat::G8R8S>,
@@ -297,7 +298,7 @@ static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPU
         // clang-format on
 };
 
-static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPUVAddr),
+static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, VAddr),
                             SurfaceParams::MaxPixelFormat>
     gl_to_morton_fns = {
         // clang-format off
@@ -314,8 +315,10 @@ static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPU
         MortonCopy<false, PixelFormat::RGBA16UI>,
         MortonCopy<false, PixelFormat::R11FG11FB10F>,
         MortonCopy<false, PixelFormat::RGBA32UI>,
-        // TODO(Subv): Swizzling DXT1/DXT23/DXT45/DXN1/DXN2/BC7U/ASTC_2D_4X4 formats is not
-        // supported
+        // 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,
@@ -532,7 +535,7 @@ MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 64
 void CachedSurface::LoadGLBuffer() {
     ASSERT(params.type != SurfaceType::Fill);
 
-    const u8* const texture_src_data = Memory::GetPointer(params.GetCpuAddr());
+    const u8* const texture_src_data = Memory::GetPointer(params.addr);
 
     ASSERT(texture_src_data);
 
@@ -557,7 +560,7 @@ void CachedSurface::LoadGLBuffer() {
 
 MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64));
 void CachedSurface::FlushGLBuffer() {
-    u8* const dst_buffer = Memory::GetPointer(params.GetCpuAddr());
+    u8* const dst_buffer = Memory::GetPointer(params.addr);
 
     ASSERT(dst_buffer);
     ASSERT(gl_buffer.size() ==
@@ -687,7 +690,7 @@ SurfaceSurfaceRect_Tuple RasterizerCacheOpenGL::GetFramebufferSurfaces(bool usin
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
 
     // TODO(bunnei): This is hard corded to use just the first render buffer
-    LOG_WARNING(Render_OpenGL, "hard-coded for render target 0!");
+    LOG_TRACE(Render_OpenGL, "hard-coded for render target 0!");
 
     // get color and depth surfaces
     SurfaceParams color_params{};
@@ -754,19 +757,10 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool pres
         return {};
     }
 
-    auto& gpu = Core::System::GetInstance().GPU();
-    // Don't try to create any entries in the cache if the address of the texture is invalid.
-    if (gpu.MemoryManager().GpuToCpuAddress(params.addr) == boost::none)
-        return {};
-
     // Look up surface in the cache based on address
     Surface surface{TryGet(params.addr)};
     if (surface) {
-        if (Settings::values.use_accurate_framebuffers) {
-            // If use_accurate_framebuffers is enabled, always load from memory
-            FlushSurface(surface);
-            Unregister(surface);
-        } else if (surface->GetSurfaceParams().IsCompatibleSurface(params)) {
+        if (surface->GetSurfaceParams().IsCompatibleSurface(params)) {
             // Use the cached surface as-is
             return surface;
         } else if (preserve_contents) {
@@ -782,15 +776,9 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool pres
         }
     }
 
-    // Try to get a previously reserved surface
-    surface = TryGetReservedSurface(params);
-
-    // No surface found - create a new one
-    if (!surface) {
-        surface = std::make_shared<CachedSurface>(params);
-        ReserveSurface(surface);
-        Register(surface);
-    }
+    // No cached surface found - get a new one
+    surface = GetUncachedSurface(params);
+    Register(surface);
 
     // Only load surface from memory if we care about the contents
     if (preserve_contents) {
@@ -800,13 +788,23 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool pres
     return surface;
 }
 
+Surface RasterizerCacheOpenGL::GetUncachedSurface(const SurfaceParams& params) {
+    Surface surface{TryGetReservedSurface(params)};
+    if (!surface) {
+        // No reserved surface available, create a new one and reserve it
+        surface = std::make_shared<CachedSurface>(params);
+        ReserveSurface(surface);
+    }
+    return surface;
+}
+
 Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
                                                const SurfaceParams& new_params) {
     // Verify surface is compatible for blitting
     const auto& params{surface->GetSurfaceParams()};
 
-    // Create a new surface with the new parameters, and blit the previous surface to it
-    Surface new_surface{std::make_shared<CachedSurface>(new_params)};
+    // Get a new surface with the new parameters, and blit the previous surface to it
+    Surface new_surface{GetUncachedSurface(new_params)};
 
     // If format is unchanged, we can do a faster blit without reinterpreting pixel data
     if (params.pixel_format == new_params.pixel_format) {
@@ -816,92 +814,73 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& surface,
         return new_surface;
     }
 
-    auto source_format = GetFormatTuple(params.pixel_format, params.component_type);
-    auto dest_format = GetFormatTuple(new_params.pixel_format, new_params.component_type);
+    // When using accurate framebuffers, always copy old data to new surface, regardless of format
+    if (Settings::values.use_accurate_framebuffers) {
+        auto source_format = GetFormatTuple(params.pixel_format, params.component_type);
+        auto dest_format = GetFormatTuple(new_params.pixel_format, new_params.component_type);
 
-    size_t buffer_size = std::max(params.SizeInBytes(), new_params.SizeInBytes());
+        size_t buffer_size = std::max(params.SizeInBytes(), new_params.SizeInBytes());
 
-    // Use a Pixel Buffer Object to download the previous texture and then upload it to the new one
-    // using the new format.
-    OGLBuffer pbo;
-    pbo.Create();
+        // Use a Pixel Buffer Object to download the previous texture and then upload it to the new
+        // one using the new format.
+        OGLBuffer pbo;
+        pbo.Create();
 
-    glBindBuffer(GL_PIXEL_PACK_BUFFER, pbo.handle);
-    glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW_ARB);
-    if (source_format.compressed) {
-        glGetCompressedTextureImage(surface->Texture().handle, 0,
-                                    static_cast<GLsizei>(params.SizeInBytes()), nullptr);
-    } else {
-        glGetTextureImage(surface->Texture().handle, 0, source_format.format, source_format.type,
-                          static_cast<GLsizei>(params.SizeInBytes()), nullptr);
-    }
-    // If the new texture is bigger than the previous one, we need to fill in the rest with data
-    // from the CPU.
-    if (params.SizeInBytes() < new_params.SizeInBytes()) {
-        // Upload the rest of the memory.
-        if (new_params.is_tiled) {
-            // TODO(Subv): We might have to de-tile the subtexture and re-tile it with the rest of
-            // the data in this case. Games like Super Mario Odyssey seem to hit this case when
-            // drawing, it re-uses the memory of a previous texture as a bigger framebuffer but it
-            // doesn't clear it beforehand, the texture is already full of zeros.
-            LOG_CRITICAL(HW_GPU, "Trying to upload extra texture data from the CPU during "
-                                 "reinterpretation but the texture is tiled.");
+        glBindBuffer(GL_PIXEL_PACK_BUFFER, pbo.handle);
+        glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW_ARB);
+        if (source_format.compressed) {
+            glGetCompressedTextureImage(surface->Texture().handle, 0,
+                                        static_cast<GLsizei>(params.SizeInBytes()), nullptr);
+        } else {
+            glGetTextureImage(surface->Texture().handle, 0, source_format.format,
+                              source_format.type, static_cast<GLsizei>(params.SizeInBytes()),
+                              nullptr);
         }
-        size_t remaining_size = new_params.SizeInBytes() - params.SizeInBytes();
-        auto address = Core::System::GetInstance().GPU().MemoryManager().GpuToCpuAddress(
-            new_params.addr + params.SizeInBytes());
-        std::vector<u8> data(remaining_size);
-        Memory::ReadBlock(*address, data.data(), data.size());
-        glBufferSubData(GL_PIXEL_PACK_BUFFER, params.SizeInBytes(), remaining_size, data.data());
+        // If the new texture is bigger than the previous one, we need to fill in the rest with data
+        // from the CPU.
+        if (params.SizeInBytes() < new_params.SizeInBytes()) {
+            // Upload the rest of the memory.
+            if (new_params.is_tiled) {
+                // TODO(Subv): We might have to de-tile the subtexture and re-tile it with the rest
+                // of the data in this case. Games like Super Mario Odyssey seem to hit this case
+                // when drawing, it re-uses the memory of a previous texture as a bigger framebuffer
+                // but it doesn't clear it beforehand, the texture is already full of zeros.
+                LOG_CRITICAL(HW_GPU, "Trying to upload extra texture data from the CPU during "
+                                     "reinterpretation but the texture is tiled.");
+            }
+            size_t remaining_size = new_params.SizeInBytes() - params.SizeInBytes();
+            std::vector<u8> data(remaining_size);
+            Memory::ReadBlock(new_params.addr + params.SizeInBytes(), data.data(), data.size());
+            glBufferSubData(GL_PIXEL_PACK_BUFFER, params.SizeInBytes(), remaining_size,
+                            data.data());
+        }
+
+        glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
+
+        const auto& dest_rect{new_params.GetRect()};
+
+        glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pbo.handle);
+        if (dest_format.compressed) {
+            glCompressedTexSubImage2D(
+                GL_TEXTURE_2D, 0, 0, 0, static_cast<GLsizei>(dest_rect.GetWidth()),
+                static_cast<GLsizei>(dest_rect.GetHeight()), dest_format.format,
+                static_cast<GLsizei>(new_params.SizeInBytes()), nullptr);
+        } else {
+            glTextureSubImage2D(new_surface->Texture().handle, 0, 0, 0,
+                                static_cast<GLsizei>(dest_rect.GetWidth()),
+                                static_cast<GLsizei>(dest_rect.GetHeight()), dest_format.format,
+                                dest_format.type, nullptr);
+        }
+        glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
+
+        pbo.Release();
     }
 
-    glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
-
-    const auto& dest_rect{new_params.GetRect()};
-
-    glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pbo.handle);
-    if (dest_format.compressed) {
-        glCompressedTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0,
-                                  static_cast<GLsizei>(dest_rect.GetWidth()),
-                                  static_cast<GLsizei>(dest_rect.GetHeight()), dest_format.format,
-                                  static_cast<GLsizei>(new_params.SizeInBytes()), nullptr);
-    } else {
-        glTextureSubImage2D(new_surface->Texture().handle, 0, 0, 0,
-                            static_cast<GLsizei>(dest_rect.GetWidth()),
-                            static_cast<GLsizei>(dest_rect.GetHeight()), dest_format.format,
-                            dest_format.type, nullptr);
-    }
-    glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
-
-    pbo.Release();
-
     return new_surface;
 }
 
-Surface RasterizerCacheOpenGL::TryFindFramebufferSurface(VAddr cpu_addr) const {
-    // Tries to find the GPU address of a framebuffer based on the CPU address. This is because
-    // final output framebuffers are specified by CPU address, but internally our GPU cache uses
-    // GPU addresses. We iterate through all cached framebuffers, and compare their starting CPU
-    // address to the one provided. This is obviously not great, and won't work if the
-    // framebuffer overlaps surfaces.
-
-    std::vector<Surface> surfaces;
-    for (const auto& surface : GetCache()) {
-        const auto& params = surface.second->GetSurfaceParams();
-        const VAddr surface_cpu_addr = params.GetCpuAddr();
-        if (cpu_addr >= surface_cpu_addr && cpu_addr < (surface_cpu_addr + params.size_in_bytes)) {
-            ASSERT_MSG(cpu_addr == surface_cpu_addr, "overlapping surfaces are unsupported");
-            surfaces.push_back(surface.second);
-        }
-    }
-
-    if (surfaces.empty()) {
-        return {};
-    }
-
-    ASSERT_MSG(surfaces.size() == 1, ">1 surface is unsupported");
-
-    return surfaces[0];
+Surface RasterizerCacheOpenGL::TryFindFramebufferSurface(VAddr addr) const {
+    return TryGet(addr);
 }
 
 void RasterizerCacheOpenGL::ReserveSurface(const Surface& surface) {
@@ -913,7 +892,6 @@ Surface RasterizerCacheOpenGL::TryGetReservedSurface(const SurfaceParams& params
     const auto& surface_reserve_key{SurfaceReserveKey::Create(params)};
     auto search{surface_reserve.find(surface_reserve_key)};
     if (search != surface_reserve.end()) {
-        Register(search->second);
         return search->second;
     }
     return {};

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -45,42 +45,44 @@ struct SurfaceParams {
         DXN2UNORM = 17,
         DXN2SNORM = 18,
         BC7U = 19,
-        ASTC_2D_4X4 = 20,
-        G8R8U = 21,
-        G8R8S = 22,
-        BGRA8 = 23,
-        RGBA32F = 24,
-        RG32F = 25,
-        R32F = 26,
-        R16F = 27,
-        R16U = 28,
-        R16S = 29,
-        R16UI = 30,
-        R16I = 31,
-        RG16 = 32,
-        RG16F = 33,
-        RG16UI = 34,
-        RG16I = 35,
-        RG16S = 36,
-        RGB32F = 37,
-        SRGBA8 = 38,
-        RG8U = 39,
-        RG8S = 40,
-        RG32UI = 41,
-        R32UI = 42,
+        BC6H_UF16 = 20,
+        BC6H_SF16 = 21,
+        ASTC_2D_4X4 = 22,
+        G8R8U = 23,
+        G8R8S = 24,
+        BGRA8 = 25,
+        RGBA32F = 26,
+        RG32F = 27,
+        R32F = 28,
+        R16F = 29,
+        R16U = 30,
+        R16S = 31,
+        R16UI = 32,
+        R16I = 33,
+        RG16 = 34,
+        RG16F = 35,
+        RG16UI = 36,
+        RG16I = 37,
+        RG16S = 38,
+        RGB32F = 39,
+        SRGBA8 = 40,
+        RG8U = 41,
+        RG8S = 42,
+        RG32UI = 43,
+        R32UI = 44,
 
         MaxColorFormat,
 
         // Depth formats
-        Z32F = 43,
-        Z16 = 44,
+        Z32F = 45,
+        Z16 = 46,
 
         MaxDepthFormat,
 
         // DepthStencil formats
-        Z24S8 = 45,
-        S8Z24 = 46,
-        Z32FS8 = 47,
+        Z24S8 = 47,
+        S8Z24 = 48,
+        Z32FS8 = 49,
 
         MaxDepthStencilFormat,
 
@@ -138,6 +140,8 @@ struct SurfaceParams {
             4, // DXN2UNORM
             4, // DXN2SNORM
             4, // BC7U
+            4, // BC6H_UF16
+            4, // BC6H_SF16
             4, // ASTC_2D_4X4
             1, // G8R8U
             1, // G8R8S
@@ -197,6 +201,8 @@ struct SurfaceParams {
             128, // DXN2UNORM
             128, // DXN2SNORM
             128, // BC7U
+            128, // BC6H_UF16
+            128, // BC6H_SF16
             32,  // ASTC_2D_4X4
             16,  // G8R8U
             16,  // G8R8S
@@ -482,6 +488,10 @@ struct SurfaceParams {
             UNREACHABLE();
         case Tegra::Texture::TextureFormat::BC7U:
             return PixelFormat::BC7U;
+        case Tegra::Texture::TextureFormat::BC6H_UF16:
+            return PixelFormat::BC6H_UF16;
+        case Tegra::Texture::TextureFormat::BC6H_SF16:
+            return PixelFormat::BC6H_SF16;
         case Tegra::Texture::TextureFormat::ASTC_2D_4X4:
             return PixelFormat::ASTC_2D_4X4;
         case Tegra::Texture::TextureFormat::R16_G16:
@@ -628,9 +638,6 @@ struct SurfaceParams {
                GetFormatBpp(pixel_format) / CHAR_BIT;
     }
 
-    /// Returns the CPU virtual address for this surface
-    VAddr GetCpuAddr() const;
-
     /// Creates SurfaceParams from a texture configuration
     static SurfaceParams CreateForTexture(const Tegra::Texture::FullTextureInfo& config);
 
@@ -643,25 +650,13 @@ struct SurfaceParams {
                                               Tegra::GPUVAddr zeta_address,
                                               Tegra::DepthFormat format);
 
-    bool operator==(const SurfaceParams& other) const {
-        return std::tie(addr, is_tiled, block_height, pixel_format, component_type, type, width,
-                        height, unaligned_height, size_in_bytes) ==
-               std::tie(other.addr, other.is_tiled, other.block_height, other.pixel_format,
-                        other.component_type, other.type, other.width, other.height,
-                        other.unaligned_height, other.size_in_bytes);
-    }
-
-    bool operator!=(const SurfaceParams& other) const {
-        return !operator==(other);
-    }
-
     /// Checks if surfaces are compatible for caching
     bool IsCompatibleSurface(const SurfaceParams& other) const {
         return std::tie(pixel_format, type, cache_width, cache_height) ==
                std::tie(other.pixel_format, other.type, other.cache_width, other.cache_height);
     }
 
-    Tegra::GPUVAddr addr;
+    VAddr addr;
     bool is_tiled;
     u32 block_height;
     PixelFormat pixel_format;
@@ -702,7 +697,7 @@ class CachedSurface final {
 public:
     CachedSurface(const SurfaceParams& params);
 
-    Tegra::GPUVAddr GetAddr() const {
+    VAddr GetAddr() const {
         return params.addr;
     }
 
@@ -753,13 +748,16 @@ public:
     /// Flushes the surface to Switch memory
     void FlushSurface(const Surface& surface);
 
-    /// Tries to find a framebuffer GPU address based on the provided CPU address
-    Surface TryFindFramebufferSurface(VAddr cpu_addr) const;
+    /// Tries to find a framebuffer using on the provided CPU address
+    Surface TryFindFramebufferSurface(VAddr addr) const;
 
 private:
     void LoadSurface(const Surface& surface);
     Surface GetSurface(const SurfaceParams& params, bool preserve_contents = true);
 
+    /// Gets an uncached surface, creating it if need be
+    Surface GetUncachedSurface(const SurfaceParams& params);
+
     /// Recreates a surface with new parameters
     Surface RecreateSurface(const Surface& surface, const SurfaceParams& new_params);
 

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -12,21 +12,17 @@
 namespace OpenGL {
 
 /// Gets the address for the specified shader stage program
-static Tegra::GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) {
+static VAddr GetShaderAddress(Maxwell::ShaderProgram program) {
     auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
     auto& shader_config = gpu.regs.shader_config[static_cast<size_t>(program)];
-
-    return gpu.regs.code_address.CodeAddress() + shader_config.offset;
+    return *gpu.memory_manager.GpuToCpuAddress(gpu.regs.code_address.CodeAddress() +
+                                               shader_config.offset);
 }
 
 /// Gets the shader program code from memory for the specified address
-static GLShader::ProgramCode GetShaderCode(Tegra::GPUVAddr addr) {
-    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
-
+static GLShader::ProgramCode GetShaderCode(VAddr addr) {
     GLShader::ProgramCode program_code(GLShader::MAX_PROGRAM_CODE_LENGTH);
-    const boost::optional<VAddr> cpu_address{gpu.memory_manager.GpuToCpuAddress(addr)};
-    Memory::ReadBlock(*cpu_address, program_code.data(), program_code.size() * sizeof(u64));
-
+    Memory::ReadBlock(addr, program_code.data(), program_code.size() * sizeof(u64));
     return program_code;
 }
 
@@ -55,7 +51,7 @@ static void SetShaderUniformBlockBindings(GLuint shader) {
                                  sizeof(GLShader::MaxwellUniformData));
 }
 
-CachedShader::CachedShader(Tegra::GPUVAddr addr, Maxwell::ShaderProgram program_type)
+CachedShader::CachedShader(VAddr addr, Maxwell::ShaderProgram program_type)
     : addr{addr}, program_type{program_type}, setup{GetShaderCode(addr)} {
 
     GLShader::ProgramResult program_result;
@@ -89,23 +85,23 @@ CachedShader::CachedShader(Tegra::GPUVAddr addr, Maxwell::ShaderProgram program_
     SetShaderUniformBlockBindings(program.handle);
 }
 
-GLuint CachedShader::GetProgramResourceIndex(const std::string& name) {
-    auto search{resource_cache.find(name)};
+GLuint CachedShader::GetProgramResourceIndex(const GLShader::ConstBufferEntry& buffer) {
+    auto search{resource_cache.find(buffer.GetHash())};
     if (search == resource_cache.end()) {
         const GLuint index{
-            glGetProgramResourceIndex(program.handle, GL_UNIFORM_BLOCK, name.c_str())};
-        resource_cache[name] = index;
+            glGetProgramResourceIndex(program.handle, GL_UNIFORM_BLOCK, buffer.GetName().c_str())};
+        resource_cache[buffer.GetHash()] = index;
         return index;
     }
 
     return search->second;
 }
 
-GLint CachedShader::GetUniformLocation(const std::string& name) {
-    auto search{uniform_cache.find(name)};
+GLint CachedShader::GetUniformLocation(const GLShader::SamplerEntry& sampler) {
+    auto search{uniform_cache.find(sampler.GetHash())};
     if (search == uniform_cache.end()) {
-        const GLint index{glGetUniformLocation(program.handle, name.c_str())};
-        uniform_cache[name] = index;
+        const GLint index{glGetUniformLocation(program.handle, sampler.GetName().c_str())};
+        uniform_cache[sampler.GetHash()] = index;
         return index;
     }
 
@@ -113,7 +109,7 @@ GLint CachedShader::GetUniformLocation(const std::string& name) {
 }
 
 Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
-    const Tegra::GPUVAddr program_addr{GetShaderAddress(program)};
+    const VAddr program_addr{GetShaderAddress(program)};
 
     // Look up shader in the cache based on address
     Shader shader{TryGet(program_addr)};