diff options
Diffstat (limited to '')
42 files changed, 896 insertions, 938 deletions
diff --git a/CMakeLists.txt b/CMakeLists.txt index 7a49318aa..3282ae9d4 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -159,15 +159,15 @@ macro(yuzu_find_packages) # Capitalization matters here. We need the naming to match the generated paths from Conan set(REQUIRED_LIBS # Cmake Pkg Prefix Version Conan Pkg - "Boost 1.71 boost/1.72.0" - "Catch2 2.11 catch2/2.11.0" + "Boost 1.73 boost/1.73.0" + "Catch2 2.13 catch2/2.13.0" "fmt 7.0 fmt/7.0.1" # can't use until https://github.com/bincrafters/community/issues/1173 #"libzip 1.5 libzip/1.5.2@bincrafters/stable" "lz4 1.8 lz4/1.9.2" - "nlohmann_json 3.7 nlohmann_json/3.7.3" + "nlohmann_json 3.8 nlohmann_json/3.8.0" "ZLIB 1.2 zlib/1.2.11" - "zstd 1.4 zstd/1.4.4" + "zstd 1.4 zstd/1.4.5" ) foreach(PACKAGE ${REQUIRED_LIBS}) diff --git a/src/core/file_sys/mode.h b/src/core/file_sys/mode.h index c95205668..2b4f21073 100644 --- a/src/core/file_sys/mode.h +++ b/src/core/file_sys/mode.h @@ -4,6 +4,7 @@ #pragma once +#include "common/common_funcs.h" #include "common/common_types.h" namespace FileSys { @@ -11,13 +12,11 @@ namespace FileSys { enum class Mode : u32 { Read = 1, Write = 2, - ReadWrite = 3, + ReadWrite = Read | Write, Append = 4, - WriteAppend = 6, + WriteAppend = Write | Append, }; -inline u32 operator&(Mode lhs, Mode rhs) { - return static_cast<u32>(lhs) & static_cast<u32>(rhs); -} +DECLARE_ENUM_FLAG_OPERATORS(Mode) } // namespace FileSys diff --git a/src/core/file_sys/patch_manager.cpp b/src/core/file_sys/patch_manager.cpp index c47ff863e..729dbb5f4 100644 --- a/src/core/file_sys/patch_manager.cpp +++ b/src/core/file_sys/patch_manager.cpp @@ -288,8 +288,8 @@ std::optional<std::vector<Core::Memory::CheatEntry>> ReadCheatFileFromFolder( } Core::Memory::TextCheatParser parser; - return parser.Parse( - system, std::string_view(reinterpret_cast<const char* const>(data.data()), data.size())); + return parser.Parse(system, + std::string_view(reinterpret_cast<const char*>(data.data()), data.size())); } } // Anonymous namespace diff --git a/src/core/file_sys/registered_cache.cpp b/src/core/file_sys/registered_cache.cpp index e94eed3b6..f831487dd 100644 --- a/src/core/file_sys/registered_cache.cpp +++ b/src/core/file_sys/registered_cache.cpp @@ -344,15 +344,18 @@ VirtualFile RegisteredCache::GetFileAtID(NcaID id) const { static std::optional<NcaID> CheckMapForContentRecord(const std::map<u64, CNMT>& map, u64 title_id, ContentRecordType type) { - if (map.find(title_id) == map.end()) - return {}; - - const auto& cnmt = map.at(title_id); + const auto cmnt_iter = map.find(title_id); + if (cmnt_iter == map.cend()) { + return std::nullopt; + } - const auto iter = std::find_if(cnmt.GetContentRecords().begin(), cnmt.GetContentRecords().end(), + const auto& cnmt = cmnt_iter->second; + const auto& content_records = cnmt.GetContentRecords(); + const auto iter = std::find_if(content_records.cbegin(), content_records.cend(), [type](const ContentRecord& rec) { return rec.type == type; }); - if (iter == cnmt.GetContentRecords().end()) - return {}; + if (iter == content_records.cend()) { + return std::nullopt; + } return std::make_optional(iter->nca_id); } @@ -467,14 +470,16 @@ VirtualFile RegisteredCache::GetEntryUnparsed(u64 title_id, ContentRecordType ty std::optional<u32> RegisteredCache::GetEntryVersion(u64 title_id) const { const auto meta_iter = meta.find(title_id); - if (meta_iter != meta.end()) + if (meta_iter != meta.cend()) { return meta_iter->second.GetTitleVersion(); + } const auto yuzu_meta_iter = yuzu_meta.find(title_id); - if (yuzu_meta_iter != yuzu_meta.end()) + if (yuzu_meta_iter != yuzu_meta.cend()) { return yuzu_meta_iter->second.GetTitleVersion(); + } - return {}; + return std::nullopt; } VirtualFile RegisteredCache::GetEntryRaw(u64 title_id, ContentRecordType type) const { diff --git a/src/core/file_sys/savedata_factory.cpp b/src/core/file_sys/savedata_factory.cpp index adfd2c1a4..ba4efee3a 100644 --- a/src/core/file_sys/savedata_factory.cpp +++ b/src/core/file_sys/savedata_factory.cpp @@ -17,23 +17,23 @@ constexpr char SAVE_DATA_SIZE_FILENAME[] = ".yuzu_save_size"; namespace { -void PrintSaveDataDescriptorWarnings(SaveDataDescriptor meta) { +void PrintSaveDataAttributeWarnings(SaveDataAttribute meta) { if (meta.type == SaveDataType::SystemSaveData || meta.type == SaveDataType::SaveData) { if (meta.zero_1 != 0) { LOG_WARNING(Service_FS, - "Possibly incorrect SaveDataDescriptor, type is " + "Possibly incorrect SaveDataAttribute, type is " "SystemSaveData||SaveData but offset 0x28 is non-zero ({:016X}).", meta.zero_1); } if (meta.zero_2 != 0) { LOG_WARNING(Service_FS, - "Possibly incorrect SaveDataDescriptor, type is " + "Possibly incorrect SaveDataAttribute, type is " "SystemSaveData||SaveData but offset 0x30 is non-zero ({:016X}).", meta.zero_2); } if (meta.zero_3 != 0) { LOG_WARNING(Service_FS, - "Possibly incorrect SaveDataDescriptor, type is " + "Possibly incorrect SaveDataAttribute, type is " "SystemSaveData||SaveData but offset 0x38 is non-zero ({:016X}).", meta.zero_3); } @@ -41,33 +41,32 @@ void PrintSaveDataDescriptorWarnings(SaveDataDescriptor meta) { if (meta.type == SaveDataType::SystemSaveData && meta.title_id != 0) { LOG_WARNING(Service_FS, - "Possibly incorrect SaveDataDescriptor, type is SystemSaveData but title_id is " + "Possibly incorrect SaveDataAttribute, type is SystemSaveData but title_id is " "non-zero ({:016X}).", meta.title_id); } if (meta.type == SaveDataType::DeviceSaveData && meta.user_id != u128{0, 0}) { LOG_WARNING(Service_FS, - "Possibly incorrect SaveDataDescriptor, type is DeviceSaveData but user_id is " + "Possibly incorrect SaveDataAttribute, type is DeviceSaveData but user_id is " "non-zero ({:016X}{:016X})", meta.user_id[1], meta.user_id[0]); } } -bool ShouldSaveDataBeAutomaticallyCreated(SaveDataSpaceId space, const SaveDataDescriptor& desc) { - return desc.type == SaveDataType::CacheStorage || desc.type == SaveDataType::TemporaryStorage || +bool ShouldSaveDataBeAutomaticallyCreated(SaveDataSpaceId space, const SaveDataAttribute& attr) { + return attr.type == SaveDataType::CacheStorage || attr.type == SaveDataType::TemporaryStorage || (space == SaveDataSpaceId::NandUser && ///< Normal Save Data -- Current Title & User - (desc.type == SaveDataType::SaveData || desc.type == SaveDataType::DeviceSaveData) && - desc.title_id == 0 && desc.save_id == 0); + (attr.type == SaveDataType::SaveData || attr.type == SaveDataType::DeviceSaveData) && + attr.title_id == 0 && attr.save_id == 0); } } // Anonymous namespace -std::string SaveDataDescriptor::DebugInfo() const { - return fmt::format("[type={:02X}, title_id={:016X}, user_id={:016X}{:016X}, " - "save_id={:016X}, " +std::string SaveDataAttribute::DebugInfo() const { + return fmt::format("[title_id={:016X}, user_id={:016X}{:016X}, save_id={:016X}, type={:02X}, " "rank={}, index={}]", - static_cast<u8>(type), title_id, user_id[1], user_id[0], save_id, + title_id, user_id[1], user_id[0], save_id, static_cast<u8>(type), static_cast<u8>(rank), index); } @@ -80,8 +79,8 @@ SaveDataFactory::SaveDataFactory(VirtualDir save_directory) : dir(std::move(save SaveDataFactory::~SaveDataFactory() = default; ResultVal<VirtualDir> SaveDataFactory::Create(SaveDataSpaceId space, - const SaveDataDescriptor& meta) const { - PrintSaveDataDescriptorWarnings(meta); + const SaveDataAttribute& meta) const { + PrintSaveDataAttributeWarnings(meta); const auto save_directory = GetFullPath(space, meta.type, meta.title_id, meta.user_id, meta.save_id); @@ -98,7 +97,7 @@ ResultVal<VirtualDir> SaveDataFactory::Create(SaveDataSpaceId space, } ResultVal<VirtualDir> SaveDataFactory::Open(SaveDataSpaceId space, - const SaveDataDescriptor& meta) const { + const SaveDataAttribute& meta) const { const auto save_directory = GetFullPath(space, meta.type, meta.title_id, meta.user_id, meta.save_id); diff --git a/src/core/file_sys/savedata_factory.h b/src/core/file_sys/savedata_factory.h index 991e57aa1..6625bbbd8 100644 --- a/src/core/file_sys/savedata_factory.h +++ b/src/core/file_sys/savedata_factory.h @@ -21,6 +21,7 @@ enum class SaveDataSpaceId : u8 { TemporaryStorage = 3, SdCardUser = 4, ProperSystem = 100, + SafeMode = 101, }; enum class SaveDataType : u8 { @@ -30,28 +31,50 @@ enum class SaveDataType : u8 { DeviceSaveData = 3, TemporaryStorage = 4, CacheStorage = 5, + SystemBcat = 6, }; enum class SaveDataRank : u8 { - Primary, - Secondary, + Primary = 0, + Secondary = 1, }; -struct SaveDataDescriptor { - u64_le title_id; +enum class SaveDataFlags : u32 { + None = (0 << 0), + KeepAfterResettingSystemSaveData = (1 << 0), + KeepAfterRefurbishment = (1 << 1), + KeepAfterResettingSystemSaveDataWithoutUserSaveData = (1 << 2), + NeedsSecureDelete = (1 << 3), +}; + +struct SaveDataAttribute { + u64 title_id; u128 user_id; - u64_le save_id; + u64 save_id; SaveDataType type; SaveDataRank rank; - u16_le index; + u16 index; INSERT_PADDING_BYTES(4); - u64_le zero_1; - u64_le zero_2; - u64_le zero_3; + u64 zero_1; + u64 zero_2; + u64 zero_3; std::string DebugInfo() const; }; -static_assert(sizeof(SaveDataDescriptor) == 0x40, "SaveDataDescriptor has incorrect size."); +static_assert(sizeof(SaveDataAttribute) == 0x40, "SaveDataAttribute has incorrect size."); + +struct SaveDataExtraData { + SaveDataAttribute attr; + u64 owner_id; + s64 timestamp; + SaveDataFlags flags; + INSERT_PADDING_BYTES(4); + s64 available_size; + s64 journal_size; + s64 commit_id; + std::array<u8, 0x190> unused; +}; +static_assert(sizeof(SaveDataExtraData) == 0x200, "SaveDataExtraData has incorrect size."); struct SaveDataSize { u64 normal; @@ -64,8 +87,8 @@ public: explicit SaveDataFactory(VirtualDir dir); ~SaveDataFactory(); - ResultVal<VirtualDir> Create(SaveDataSpaceId space, const SaveDataDescriptor& meta) const; - ResultVal<VirtualDir> Open(SaveDataSpaceId space, const SaveDataDescriptor& meta) const; + ResultVal<VirtualDir> Create(SaveDataSpaceId space, const SaveDataAttribute& meta) const; + ResultVal<VirtualDir> Open(SaveDataSpaceId space, const SaveDataAttribute& meta) const; VirtualDir GetSaveDataSpaceDirectory(SaveDataSpaceId space) const; diff --git a/src/core/file_sys/vfs_real.cpp b/src/core/file_sys/vfs_real.cpp index 96ce5957c..0db0091f6 100644 --- a/src/core/file_sys/vfs_real.cpp +++ b/src/core/file_sys/vfs_real.cpp @@ -18,20 +18,22 @@ static std::string ModeFlagsToString(Mode mode) { std::string mode_str; // Calculate the correct open mode for the file. - if (mode & Mode::Read && mode & Mode::Write) { - if (mode & Mode::Append) + if (True(mode & Mode::Read) && True(mode & Mode::Write)) { + if (True(mode & Mode::Append)) { mode_str = "a+"; - else + } else { mode_str = "r+"; + } } else { - if (mode & Mode::Read) + if (True(mode & Mode::Read)) { mode_str = "r"; - else if (mode & Mode::Append) + } else if (True(mode & Mode::Append)) { mode_str = "a"; - else if (mode & Mode::Write) + } else if (True(mode & Mode::Write)) { mode_str = "w"; - else + } else { UNREACHABLE_MSG("Invalid file open mode: {:02X}", static_cast<u8>(mode)); + } } mode_str += "b"; @@ -73,8 +75,9 @@ VirtualFile RealVfsFilesystem::OpenFile(std::string_view path_, Mode perms) { } } - if (!FileUtil::Exists(path) && (perms & Mode::WriteAppend) != 0) + if (!FileUtil::Exists(path) && True(perms & Mode::WriteAppend)) { FileUtil::CreateEmptyFile(path); + } auto backing = std::make_shared<FileUtil::IOFile>(path, ModeFlagsToString(perms).c_str()); cache[path] = backing; @@ -247,11 +250,11 @@ std::shared_ptr<VfsDirectory> RealVfsFile::GetContainingDirectory() const { } bool RealVfsFile::IsWritable() const { - return (perms & Mode::WriteAppend) != 0; + return True(perms & Mode::WriteAppend); } bool RealVfsFile::IsReadable() const { - return (perms & Mode::ReadWrite) != 0; + return True(perms & Mode::ReadWrite); } std::size_t RealVfsFile::Read(u8* data, std::size_t length, std::size_t offset) const { @@ -319,8 +322,9 @@ RealVfsDirectory::RealVfsDirectory(RealVfsFilesystem& base_, const std::string& path_components(FileUtil::SplitPathComponents(path)), parent_components(FileUtil::SliceVector(path_components, 0, path_components.size() - 1)), perms(perms_) { - if (!FileUtil::Exists(path) && perms & Mode::WriteAppend) + if (!FileUtil::Exists(path) && True(perms & Mode::WriteAppend)) { FileUtil::CreateDir(path); + } } RealVfsDirectory::~RealVfsDirectory() = default; @@ -371,11 +375,11 @@ std::vector<std::shared_ptr<VfsDirectory>> RealVfsDirectory::GetSubdirectories() } bool RealVfsDirectory::IsWritable() const { - return (perms & Mode::WriteAppend) != 0; + return True(perms & Mode::WriteAppend); } bool RealVfsDirectory::IsReadable() const { - return (perms & Mode::ReadWrite) != 0; + return True(perms & Mode::ReadWrite); } std::string RealVfsDirectory::GetName() const { diff --git a/src/core/hle/service/acc/profile_manager.cpp b/src/core/hle/service/acc/profile_manager.cpp index eb8c81645..a98d57b5c 100644 --- a/src/core/hle/service/acc/profile_manager.cpp +++ b/src/core/hle/service/acc/profile_manager.cpp @@ -58,7 +58,7 @@ ProfileManager::~ProfileManager() { /// internal management of the users profiles std::optional<std::size_t> ProfileManager::AddToProfiles(const ProfileInfo& profile) { if (user_count >= MAX_USERS) { - return {}; + return std::nullopt; } profiles[user_count] = profile; return user_count++; @@ -101,13 +101,14 @@ ResultCode ProfileManager::CreateNewUser(UUID uuid, const ProfileUsername& usern [&uuid](const ProfileInfo& profile) { return uuid == profile.user_uuid; })) { return ERROR_USER_ALREADY_EXISTS; } - ProfileInfo profile; - profile.user_uuid = uuid; - profile.username = username; - profile.data = {}; - profile.creation_time = 0x0; - profile.is_open = false; - return AddUser(profile); + + return AddUser({ + .user_uuid = uuid, + .username = username, + .creation_time = 0, + .data = {}, + .is_open = false, + }); } /// Creates a new user on the system. This function allows a much simpler method of registration @@ -126,7 +127,7 @@ ResultCode ProfileManager::CreateNewUser(UUID uuid, const std::string& username) std::optional<UUID> ProfileManager::GetUser(std::size_t index) const { if (index >= MAX_USERS) { - return {}; + return std::nullopt; } return profiles[index].user_uuid; @@ -135,13 +136,13 @@ std::optional<UUID> ProfileManager::GetUser(std::size_t index) const { /// Returns a users profile index based on their user id. std::optional<std::size_t> ProfileManager::GetUserIndex(const UUID& uuid) const { if (!uuid) { - return {}; + return std::nullopt; } const auto iter = std::find_if(profiles.begin(), profiles.end(), [&uuid](const ProfileInfo& p) { return p.user_uuid == uuid; }); if (iter == profiles.end()) { - return {}; + return std::nullopt; } return static_cast<std::size_t>(std::distance(profiles.begin(), iter)); @@ -339,7 +340,13 @@ void ProfileManager::ParseUserSaveFile() { continue; } - AddUser({user.uuid, user.username, user.timestamp, user.extra_data, false}); + AddUser({ + .user_uuid = user.uuid, + .username = user.username, + .creation_time = user.timestamp, + .data = user.extra_data, + .is_open = false, + }); } std::stable_partition(profiles.begin(), profiles.end(), @@ -350,11 +357,13 @@ void ProfileManager::WriteUserSaveFile() { ProfileDataRaw raw{}; for (std::size_t i = 0; i < MAX_USERS; ++i) { - raw.users[i].username = profiles[i].username; - raw.users[i].uuid2 = profiles[i].user_uuid; - raw.users[i].uuid = profiles[i].user_uuid; - raw.users[i].timestamp = profiles[i].creation_time; - raw.users[i].extra_data = profiles[i].data; + raw.users[i] = { + .uuid = profiles[i].user_uuid, + .uuid2 = profiles[i].user_uuid, + .timestamp = profiles[i].creation_time, + .username = profiles[i].username, + .extra_data = profiles[i].data, + }; } const auto raw_path = diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp index ceed20609..55a1edf1a 100644 --- a/src/core/hle/service/am/am.cpp +++ b/src/core/hle/service/am/am.cpp @@ -1342,12 +1342,12 @@ void IApplicationFunctions::EnsureSaveData(Kernel::HLERequestContext& ctx) { LOG_DEBUG(Service_AM, "called, uid={:016X}{:016X}", user_id[1], user_id[0]); - FileSys::SaveDataDescriptor descriptor{}; - descriptor.title_id = system.CurrentProcess()->GetTitleID(); - descriptor.user_id = user_id; - descriptor.type = FileSys::SaveDataType::SaveData; + FileSys::SaveDataAttribute attribute{}; + attribute.title_id = system.CurrentProcess()->GetTitleID(); + attribute.user_id = user_id; + attribute.type = FileSys::SaveDataType::SaveData; const auto res = system.GetFileSystemController().CreateSaveData( - FileSys::SaveDataSpaceId::NandUser, descriptor); + FileSys::SaveDataSpaceId::NandUser, attribute); IPC::ResponseBuilder rb{ctx, 4}; rb.Push(res.Code()); diff --git a/src/core/hle/service/audio/audout_u.cpp b/src/core/hle/service/audio/audout_u.cpp index 106e89743..dd80dd1dc 100644 --- a/src/core/hle/service/audio/audout_u.cpp +++ b/src/core/hle/service/audio/audout_u.cpp @@ -71,7 +71,7 @@ public: stream = audio_core.OpenStream(system.CoreTiming(), audio_params.sample_rate, audio_params.channel_count, std::move(unique_name), - [=]() { buffer_event.writable->Signal(); }); + [this] { buffer_event.writable->Signal(); }); } private: diff --git a/src/core/hle/service/filesystem/filesystem.cpp b/src/core/hle/service/filesystem/filesystem.cpp index c66124998..4490f8e4c 100644 --- a/src/core/hle/service/filesystem/filesystem.cpp +++ b/src/core/hle/service/filesystem/filesystem.cpp @@ -311,7 +311,7 @@ ResultVal<FileSys::VirtualFile> FileSystemController::OpenRomFS( } ResultVal<FileSys::VirtualDir> FileSystemController::CreateSaveData( - FileSys::SaveDataSpaceId space, const FileSys::SaveDataDescriptor& save_struct) const { + FileSys::SaveDataSpaceId space, const FileSys::SaveDataAttribute& save_struct) const { LOG_TRACE(Service_FS, "Creating Save Data for space_id={:01X}, save_struct={}", static_cast<u8>(space), save_struct.DebugInfo()); @@ -323,15 +323,15 @@ ResultVal<FileSys::VirtualDir> FileSystemController::CreateSaveData( } ResultVal<FileSys::VirtualDir> FileSystemController::OpenSaveData( - FileSys::SaveDataSpaceId space, const FileSys::SaveDataDescriptor& descriptor) const { + FileSys::SaveDataSpaceId space, const FileSys::SaveDataAttribute& attribute) const { LOG_TRACE(Service_FS, "Opening Save Data for space_id={:01X}, save_struct={}", - static_cast<u8>(space), descriptor.DebugInfo()); + static_cast<u8>(space), attribute.DebugInfo()); if (save_data_factory == nullptr) { return FileSys::ERROR_ENTITY_NOT_FOUND; } - return save_data_factory->Open(space, descriptor); + return save_data_factory->Open(space, attribute); } ResultVal<FileSys::VirtualDir> FileSystemController::OpenSaveDataSpace( diff --git a/src/core/hle/service/filesystem/filesystem.h b/src/core/hle/service/filesystem/filesystem.h index 1b0a6a949..6dbbf0b2b 100644 --- a/src/core/hle/service/filesystem/filesystem.h +++ b/src/core/hle/service/filesystem/filesystem.h @@ -31,7 +31,7 @@ enum class SaveDataSpaceId : u8; enum class SaveDataType : u8; enum class StorageId : u8; -struct SaveDataDescriptor; +struct SaveDataAttribute; struct SaveDataSize; } // namespace FileSys @@ -69,9 +69,9 @@ public: ResultVal<FileSys::VirtualFile> OpenRomFS(u64 title_id, FileSys::StorageId storage_id, FileSys::ContentRecordType type) const; ResultVal<FileSys::VirtualDir> CreateSaveData( - FileSys::SaveDataSpaceId space, const FileSys::SaveDataDescriptor& save_struct) const; + FileSys::SaveDataSpaceId space, const FileSys::SaveDataAttribute& save_struct) const; ResultVal<FileSys::VirtualDir> OpenSaveData( - FileSys::SaveDataSpaceId space, const FileSys::SaveDataDescriptor& save_struct) const; + FileSys::SaveDataSpaceId space, const FileSys::SaveDataAttribute& save_struct) const; ResultVal<FileSys::VirtualDir> OpenSaveDataSpace(FileSys::SaveDataSpaceId space) const; ResultVal<FileSys::VirtualDir> OpenSDMC() const; ResultVal<FileSys::VirtualDir> OpenBISPartition(FileSys::BisPartitionId id) const; diff --git a/src/core/hle/service/filesystem/fsp_srv.cpp b/src/core/hle/service/filesystem/fsp_srv.cpp index 20c331b77..26fd87f58 100644 --- a/src/core/hle/service/filesystem/fsp_srv.cpp +++ b/src/core/hle/service/filesystem/fsp_srv.cpp @@ -696,8 +696,8 @@ FSP_SRV::FSP_SRV(FileSystemController& fsc, const Core::Reporter& reporter) {67, nullptr, "FindSaveDataWithFilter"}, {68, nullptr, "OpenSaveDataInfoReaderBySaveDataFilter"}, {69, nullptr, "ReadSaveDataFileSystemExtraDataBySaveDataAttribute"}, - {70, nullptr, "WriteSaveDataFileSystemExtraDataBySaveDataAttribute"}, - {71, nullptr, "ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute"}, + {70, &FSP_SRV::WriteSaveDataFileSystemExtraDataBySaveDataAttribute, "WriteSaveDataFileSystemExtraDataBySaveDataAttribute"}, + {71, &FSP_SRV::ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute, "ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute"}, {80, nullptr, "OpenSaveDataMetaFile"}, {81, nullptr, "OpenSaveDataTransferManager"}, {82, nullptr, "OpenSaveDataTransferManagerVersion2"}, @@ -812,7 +812,7 @@ void FSP_SRV::OpenSdCardFileSystem(Kernel::HLERequestContext& ctx) { void FSP_SRV::CreateSaveDataFileSystem(Kernel::HLERequestContext& ctx) { IPC::RequestParser rp{ctx}; - auto save_struct = rp.PopRaw<FileSys::SaveDataDescriptor>(); + auto save_struct = rp.PopRaw<FileSys::SaveDataAttribute>(); [[maybe_unused]] auto save_create_struct = rp.PopRaw<std::array<u8, 0x40>>(); u128 uid = rp.PopRaw<u128>(); @@ -826,17 +826,18 @@ void FSP_SRV::CreateSaveDataFileSystem(Kernel::HLERequestContext& ctx) { } void FSP_SRV::OpenSaveDataFileSystem(Kernel::HLERequestContext& ctx) { - LOG_INFO(Service_FS, "called."); + IPC::RequestParser rp{ctx}; struct Parameters { - FileSys::SaveDataSpaceId save_data_space_id; - FileSys::SaveDataDescriptor descriptor; + FileSys::SaveDataSpaceId space_id; + FileSys::SaveDataAttribute attribute; }; - IPC::RequestParser rp{ctx}; const auto parameters = rp.PopRaw<Parameters>(); - auto dir = fsc.OpenSaveData(parameters.save_data_space_id, parameters.descriptor); + LOG_INFO(Service_FS, "called."); + + auto dir = fsc.OpenSaveData(parameters.space_id, parameters.attribute); if (dir.Failed()) { IPC::ResponseBuilder rb{ctx, 2, 0, 0}; rb.Push(FileSys::ERROR_ENTITY_NOT_FOUND); @@ -844,13 +845,18 @@ void FSP_SRV::OpenSaveDataFileSystem(Kernel::HLERequestContext& ctx) { } FileSys::StorageId id; - if (parameters.save_data_space_id == FileSys::SaveDataSpaceId::NandUser) { + + switch (parameters.space_id) { + case FileSys::SaveDataSpaceId::NandUser: id = FileSys::StorageId::NandUser; - } else if (parameters.save_data_space_id == FileSys::SaveDataSpaceId::SdCardSystem || - parameters.save_data_space_id == FileSys::SaveDataSpaceId::SdCardUser) { + break; + case FileSys::SaveDataSpaceId::SdCardSystem: + case FileSys::SaveDataSpaceId::SdCardUser: id = FileSys::StorageId::SdCard; - } else { + break; + case FileSys::SaveDataSpaceId::NandSystem: id = FileSys::StorageId::NandSystem; + break; } auto filesystem = @@ -876,22 +882,31 @@ void FSP_SRV::OpenSaveDataInfoReaderBySaveDataSpaceId(Kernel::HLERequestContext& rb.PushIpcInterface<ISaveDataInfoReader>(std::make_shared<ISaveDataInfoReader>(space, fsc)); } -void FSP_SRV::SetGlobalAccessLogMode(Kernel::HLERequestContext& ctx) { - IPC::RequestParser rp{ctx}; - log_mode = rp.PopEnum<LogMode>(); - - LOG_DEBUG(Service_FS, "called, log_mode={:08X}", static_cast<u32>(log_mode)); +void FSP_SRV::WriteSaveDataFileSystemExtraDataBySaveDataAttribute(Kernel::HLERequestContext& ctx) { + LOG_WARNING(Service_FS, "(STUBBED) called."); IPC::ResponseBuilder rb{ctx, 2}; rb.Push(RESULT_SUCCESS); } -void FSP_SRV::GetGlobalAccessLogMode(Kernel::HLERequestContext& ctx) { - LOG_DEBUG(Service_FS, "called"); +void FSP_SRV::ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute( + Kernel::HLERequestContext& ctx) { + IPC::RequestParser rp{ctx}; + + struct Parameters { + FileSys::SaveDataSpaceId space_id; + FileSys::SaveDataAttribute attribute; + }; + + const auto parameters = rp.PopRaw<Parameters>(); + // Stub this to None for now, backend needs an impl to read/write the SaveDataExtraData + constexpr auto flags = static_cast<u32>(FileSys::SaveDataFlags::None); + + LOG_WARNING(Service_FS, "(STUBBED) called, flags={}", flags); IPC::ResponseBuilder rb{ctx, 3}; rb.Push(RESULT_SUCCESS); - rb.PushEnum(log_mode); + rb.Push(flags); } void FSP_SRV::OpenDataStorageByCurrentProcess(Kernel::HLERequestContext& ctx) { @@ -966,6 +981,24 @@ void FSP_SRV::OpenPatchDataStorageByCurrentProcess(Kernel::HLERequestContext& ct rb.Push(FileSys::ERROR_ENTITY_NOT_FOUND); } +void FSP_SRV::SetGlobalAccessLogMode(Kernel::HLERequestContext& ctx) { + IPC::RequestParser rp{ctx}; + log_mode = rp.PopEnum<LogMode>(); + + LOG_DEBUG(Service_FS, "called, log_mode={:08X}", static_cast<u32>(log_mode)); + + IPC::ResponseBuilder rb{ctx, 2}; + rb.Push(RESULT_SUCCESS); +} + +void FSP_SRV::GetGlobalAccessLogMode(Kernel::HLERequestContext& ctx) { + LOG_DEBUG(Service_FS, "called"); + + IPC::ResponseBuilder rb{ctx, 3}; + rb.Push(RESULT_SUCCESS); + rb.PushEnum(log_mode); +} + void FSP_SRV::OutputAccessLogToSdCard(Kernel::HLERequestContext& ctx) { const auto raw = ctx.ReadBuffer(); auto log = Common::StringFromFixedZeroTerminatedBuffer( diff --git a/src/core/hle/service/filesystem/fsp_srv.h b/src/core/hle/service/filesystem/fsp_srv.h index dfb3e395b..4964e874e 100644 --- a/src/core/hle/service/filesystem/fsp_srv.h +++ b/src/core/hle/service/filesystem/fsp_srv.h @@ -43,11 +43,13 @@ private: void OpenSaveDataFileSystem(Kernel::HLERequestContext& ctx); void OpenReadOnlySaveDataFileSystem(Kernel::HLERequestContext& ctx); void OpenSaveDataInfoReaderBySaveDataSpaceId(Kernel::HLERequestContext& ctx); - void SetGlobalAccessLogMode(Kernel::HLERequestContext& ctx); - void GetGlobalAccessLogMode(Kernel::HLERequestContext& ctx); + void WriteSaveDataFileSystemExtraDataBySaveDataAttribute(Kernel::HLERequestContext& ctx); + void ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute(Kernel::HLERequestContext& ctx); void OpenDataStorageByCurrentProcess(Kernel::HLERequestContext& ctx); void OpenDataStorageByDataId(Kernel::HLERequestContext& ctx); void OpenPatchDataStorageByCurrentProcess(Kernel::HLERequestContext& ctx); + void SetGlobalAccessLogMode(Kernel::HLERequestContext& ctx); + void GetGlobalAccessLogMode(Kernel::HLERequestContext& ctx); void OutputAccessLogToSdCard(Kernel::HLERequestContext& ctx); void GetAccessLogVersionInfo(Kernel::HLERequestContext& ctx); void OpenMultiCommitManager(Kernel::HLERequestContext& ctx); diff --git a/src/core/hle/service/ldr/ldr.cpp b/src/core/hle/service/ldr/ldr.cpp index 64a526b9e..d8cd10e31 100644 --- a/src/core/hle/service/ldr/ldr.cpp +++ b/src/core/hle/service/ldr/ldr.cpp @@ -310,7 +310,7 @@ public: ResultVal<VAddr> MapProcessCodeMemory(Kernel::Process* process, VAddr baseAddress, u64 size) const { - for (int retry{}; retry < MAXIMUM_MAP_RETRIES; retry++) { + for (std::size_t retry = 0; retry < MAXIMUM_MAP_RETRIES; retry++) { auto& page_table{process->PageTable()}; const VAddr addr{GetRandomMapRegion(page_table, size)}; const ResultCode result{page_table.MapProcessCodeMemory(addr, baseAddress, size)}; @@ -331,8 +331,7 @@ public: ResultVal<VAddr> MapNro(Kernel::Process* process, VAddr nro_addr, std::size_t nro_size, VAddr bss_addr, std::size_t bss_size, std::size_t size) const { - - for (int retry{}; retry < MAXIMUM_MAP_RETRIES; retry++) { + for (std::size_t retry = 0; retry < MAXIMUM_MAP_RETRIES; retry++) { auto& page_table{process->PageTable()}; VAddr addr{}; diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp index 195421cc0..d4ba88147 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp @@ -16,11 +16,12 @@ #include "video_core/renderer_base.h" namespace Service::Nvidia::Devices { + namespace NvErrCodes { -enum { - InvalidNmapHandle = -22, -}; -} +constexpr u32 Success{}; +constexpr u32 OutOfMemory{static_cast<u32>(-12)}; +constexpr u32 InvalidInput{static_cast<u32>(-22)}; +} // namespace NvErrCodes nvhost_as_gpu::nvhost_as_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev) : nvdevice(system), nvmap_dev(std::move(nvmap_dev)) {} @@ -49,8 +50,9 @@ u32 nvhost_as_gpu::ioctl(Ioctl command, const std::vector<u8>& input, const std: break; } - if (static_cast<IoctlCommand>(command.cmd.Value()) == IoctlCommand::IocRemapCommand) + if (static_cast<IoctlCommand>(command.cmd.Value()) == IoctlCommand::IocRemapCommand) { return Remap(input, output); + } UNIMPLEMENTED_MSG("Unimplemented ioctl command"); return 0; @@ -59,6 +61,7 @@ u32 nvhost_as_gpu::ioctl(Ioctl command, const std::vector<u8>& input, const std: u32 nvhost_as_gpu::InitalizeEx(const std::vector<u8>& input, std::vector<u8>& output) { IoctlInitalizeEx params{}; std::memcpy(¶ms, input.data(), input.size()); + LOG_WARNING(Service_NVDRV, "(STUBBED) called, big_page_size=0x{:X}", params.big_page_size); return 0; @@ -67,53 +70,61 @@ u32 nvhost_as_gpu::InitalizeEx(const std::vector<u8>& input, std::vector<u8>& ou u32 nvhost_as_gpu::AllocateSpace(const std::vector<u8>& input, std::vector<u8>& output) { IoctlAllocSpace params{}; std::memcpy(¶ms, input.data(), input.size()); + LOG_DEBUG(Service_NVDRV, "called, pages={:X}, page_size={:X}, flags={:X}", params.pages, params.page_size, params.flags); - auto& gpu = system.GPU(); - const u64 size{static_cast<u64>(params.pages) * static_cast<u64>(params.page_size)}; - if (params.flags & 1) { - params.offset = gpu.MemoryManager().AllocateSpace(params.offset, size, 1); + const auto size{static_cast<u64>(params.pages) * static_cast<u64>(params.page_size)}; + if ((params.flags & AddressSpaceFlags::FixedOffset) != AddressSpaceFlags::None) { + params.offset = *system.GPU().MemoryManager().AllocateFixed(params.offset, size); } else { - params.offset = gpu.MemoryManager().AllocateSpace(size, params.align); + params.offset = system.GPU().MemoryManager().Allocate(size, params.align); + } + + auto result{NvErrCodes::Success}; + if (!params.offset) { + LOG_CRITICAL(Service_NVDRV, "allocation failed for size {}", size); + result = NvErrCodes::OutOfMemory; } std::memcpy(output.data(), ¶ms, output.size()); - return 0; + return result; } u32 nvhost_as_gpu::Remap(const std::vector<u8>& input, std::vector<u8>& output) { - std::size_t num_entries = input.size() / sizeof(IoctlRemapEntry); + const auto num_entries = input.size() / sizeof(IoctlRemapEntry); - LOG_WARNING(Service_NVDRV, "(STUBBED) called, num_entries=0x{:X}", num_entries); + LOG_DEBUG(Service_NVDRV, "called, num_entries=0x{:X}", num_entries); + auto result{NvErrCodes::Success}; std::vector<IoctlRemapEntry> entries(num_entries); std::memcpy(entries.data(), input.data(), input.size()); - auto& gpu = system.GPU(); for (const auto& entry : entries) { - LOG_WARNING(Service_NVDRV, "remap entry, offset=0x{:X} handle=0x{:X} pages=0x{:X}", - entry.offset, entry.nvmap_handle, entry.pages); - GPUVAddr offset = static_cast<GPUVAddr>(entry.offset) << 0x10; - auto object = nvmap_dev->GetObject(entry.nvmap_handle); + LOG_DEBUG(Service_NVDRV, "remap entry, offset=0x{:X} handle=0x{:X} pages=0x{:X}", + entry.offset, entry.nvmap_handle, entry.pages); + + const auto object{nvmap_dev->GetObject(entry.nvmap_handle)}; if (!object) { - LOG_CRITICAL(Service_NVDRV, "nvmap {} is an invalid handle!", entry.nvmap_handle); - std::memcpy(output.data(), entries.data(), output.size()); - return static_cast<u32>(NvErrCodes::InvalidNmapHandle); + LOG_CRITICAL(Service_NVDRV, "invalid nvmap_handle={:X}", entry.nvmap_handle); + result = NvErrCodes::InvalidInput; + break; } - ASSERT(object->status == nvmap::Object::Status::Allocated); + const auto offset{static_cast<GPUVAddr>(entry.offset) << 0x10}; + const auto size{static_cast<u64>(entry.pages) << 0x10}; + const auto map_offset{static_cast<u64>(entry.map_offset) << 0x10}; + const auto addr{system.GPU().MemoryManager().Map(object->addr + map_offset, offset, size)}; - const u64 size = static_cast<u64>(entry.pages) << 0x10; - ASSERT(size <= object->size); - const u64 map_offset = static_cast<u64>(entry.map_offset) << 0x10; - - const GPUVAddr returned = - gpu.MemoryManager().MapBufferEx(object->addr + map_offset, offset, size); - ASSERT(returned == offset); + if (!addr) { + LOG_CRITICAL(Service_NVDRV, "map returned an invalid address!"); + result = NvErrCodes::InvalidInput; + break; + } } + std::memcpy(output.data(), entries.data(), output.size()); - return 0; + return result; } u32 nvhost_as_gpu::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& output) { @@ -126,44 +137,76 @@ u32 nvhost_as_gpu::MapBufferEx(const std::vector<u8>& input, std::vector<u8>& ou params.flags, params.nvmap_handle, params.buffer_offset, params.mapping_size, params.offset); - if (!params.nvmap_handle) { - return 0; + const auto object{nvmap_dev->GetObject(params.nvmap_handle)}; + if (!object) { + LOG_CRITICAL(Service_NVDRV, "invalid nvmap_handle={:X}", params.nvmap_handle); + std::memcpy(output.data(), ¶ms, output.size()); + return NvErrCodes::InvalidInput; } - auto object = nvmap_dev->GetObject(params.nvmap_handle); - ASSERT(object); - - // We can only map objects that have already been assigned a CPU address. - ASSERT(object->status == nvmap::Object::Status::Allocated); - - ASSERT(params.buffer_offset == 0); - // The real nvservices doesn't make a distinction between handles and ids, and // object can only have one handle and it will be the same as its id. Assert that this is the // case to prevent unexpected behavior. ASSERT(object->id == params.nvmap_handle); - auto& gpu = system.GPU(); - if (params.flags & 1) { - params.offset = gpu.MemoryManager().MapBufferEx(object->addr, params.offset, object->size); - } else { - params.offset = gpu.MemoryManager().MapBufferEx(object->addr, object->size); + u64 page_size{params.page_size}; + if (!page_size) { + page_size = object->align; + } + + if ((params.flags & AddressSpaceFlags::Remap) != AddressSpaceFlags::None) { + if (const auto buffer_map{FindBufferMap(params.offset)}; buffer_map) { + const auto cpu_addr{static_cast<VAddr>(buffer_map->CpuAddr() + params.buffer_offset)}; + const auto gpu_addr{static_cast<GPUVAddr>(params.offset + params.buffer_offset)}; + + if (!gpu.MemoryManager().Map(cpu_addr, gpu_addr, params.mapping_size)) { + LOG_CRITICAL(Service_NVDRV, + "remap failed, flags={:X}, nvmap_handle={:X}, buffer_offset={}, " + "mapping_size = {}, offset={}", + params.flags, params.nvmap_handle, params.buffer_offset, + params.mapping_size, params.offset); + + std::memcpy(output.data(), ¶ms, output.size()); + return NvErrCodes::InvalidInput; + } + + std::memcpy(output.data(), ¶ms, output.size()); + return NvErrCodes::Success; + } else { + LOG_CRITICAL(Service_NVDRV, "address not mapped offset={}", params.offset); + + std::memcpy(output.data(), ¶ms, output.size()); + return NvErrCodes::InvalidInput; + } } - // Create a new mapping entry for this operation. - ASSERT_MSG(buffer_mappings.find(params.offset) == buffer_mappings.end(), - "Offset is already mapped"); + // We can only map objects that have already been assigned a CPU address. + ASSERT(object->status == nvmap::Object::Status::Allocated); + + const auto physical_address{object->addr + params.buffer_offset}; + u64 size{params.mapping_size}; + if (!size) { + size = object->size; + } - BufferMapping mapping{}; - mapping.nvmap_handle = params.nvmap_handle; - mapping.offset = params.offset; - mapping.size = object->size; + const bool is_alloc{(params.flags & AddressSpaceFlags::FixedOffset) == AddressSpaceFlags::None}; + if (is_alloc) { + params.offset = gpu.MemoryManager().MapAllocate(physical_address, size, page_size); + } else { + params.offset = gpu.MemoryManager().Map(physical_address, params.offset, size); + } - buffer_mappings[params.offset] = mapping; + auto result{NvErrCodes::Success}; + if (!params.offset) { + LOG_CRITICAL(Service_NVDRV, "failed to map size={}", size); + result = NvErrCodes::InvalidInput; + } else { + AddBufferMap(params.offset, size, physical_address, is_alloc); + } std::memcpy(output.data(), ¶ms, output.size()); - return 0; + return result; } u32 nvhost_as_gpu::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& output) { @@ -172,24 +215,20 @@ u32 nvhost_as_gpu::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& ou LOG_DEBUG(Service_NVDRV, "called, offset=0x{:X}", params.offset); - const auto itr = buffer_mappings.find(params.offset); - if (itr == buffer_mappings.end()) { - LOG_WARNING(Service_NVDRV, "Tried to unmap an invalid offset 0x{:X}", params.offset); - // Hardware tests shows that unmapping an already unmapped buffer always returns successful - // and doesn't fail. - return 0; + if (const auto size{RemoveBufferMap(params.offset)}; size) { + system.GPU().MemoryManager().Unmap(params.offset, *size); + } else { + LOG_ERROR(Service_NVDRV, "invalid offset=0x{:X}", params.offset); } - params.offset = system.GPU().MemoryManager().UnmapBuffer(params.offset, itr->second.size); - buffer_mappings.erase(itr->second.offset); - std::memcpy(output.data(), ¶ms, output.size()); - return 0; + return NvErrCodes::Success; } u32 nvhost_as_gpu::BindChannel(const std::vector<u8>& input, std::vector<u8>& output) { IoctlBindChannel params{}; std::memcpy(¶ms, input.data(), input.size()); + LOG_DEBUG(Service_NVDRV, "called, fd={:X}", params.fd); channel = params.fd; @@ -199,6 +238,7 @@ u32 nvhost_as_gpu::BindChannel(const std::vector<u8>& input, std::vector<u8>& ou u32 nvhost_as_gpu::GetVARegions(const std::vector<u8>& input, std::vector<u8>& output) { IoctlGetVaRegions params{}; std::memcpy(¶ms, input.data(), input.size()); + LOG_WARNING(Service_NVDRV, "(STUBBED) called, buf_addr={:X}, buf_size={:X}", params.buf_addr, params.buf_size); @@ -210,9 +250,43 @@ u32 nvhost_as_gpu::GetVARegions(const std::vector<u8>& input, std::vector<u8>& o params.regions[1].offset = 0x04000000; params.regions[1].page_size = 0x10000; params.regions[1].pages = 0x1bffff; + // TODO(ogniK): This probably can stay stubbed but should add support way way later + std::memcpy(output.data(), ¶ms, output.size()); return 0; } +std::optional<nvhost_as_gpu::BufferMap> nvhost_as_gpu::FindBufferMap(GPUVAddr gpu_addr) const { + const auto end{buffer_mappings.upper_bound(gpu_addr)}; + for (auto iter{buffer_mappings.begin()}; iter != end; ++iter) { + if (gpu_addr >= iter->second.StartAddr() && gpu_addr < iter->second.EndAddr()) { + return iter->second; + } + } + + return {}; +} + +void nvhost_as_gpu::AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr, + bool is_allocated) { + buffer_mappings[gpu_addr] = {gpu_addr, size, cpu_addr, is_allocated}; +} + +std::optional<std::size_t> nvhost_as_gpu::RemoveBufferMap(GPUVAddr gpu_addr) { + if (const auto iter{buffer_mappings.find(gpu_addr)}; iter != buffer_mappings.end()) { + std::size_t size{}; + + if (iter->second.IsAllocated()) { + size = iter->second.Size(); + } + + buffer_mappings.erase(iter); + + return size; + } + + return {}; +} + } // namespace Service::Nvidia::Devices diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h index f79fcc065..9a0cdff0c 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h +++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h @@ -4,9 +4,12 @@ #pragma once +#include <map> #include <memory> -#include <unordered_map> +#include <optional> #include <vector> + +#include "common/common_funcs.h" #include "common/common_types.h" #include "common/swap.h" #include "core/hle/service/nvdrv/devices/nvdevice.h" @@ -15,6 +18,13 @@ namespace Service::Nvidia::Devices { class nvmap; +enum class AddressSpaceFlags : u32 { + None = 0x0, + FixedOffset = 0x1, + Remap = 0x100, +}; +DECLARE_ENUM_FLAG_OPERATORS(AddressSpaceFlags); + class nvhost_as_gpu final : public nvdevice { public: explicit nvhost_as_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev); @@ -25,6 +35,45 @@ public: IoctlVersion version) override; private: + class BufferMap final { + public: + constexpr BufferMap() = default; + + constexpr BufferMap(GPUVAddr start_addr, std::size_t size) + : start_addr{start_addr}, end_addr{start_addr + size} {} + + constexpr BufferMap(GPUVAddr start_addr, std::size_t size, VAddr cpu_addr, + bool is_allocated) + : start_addr{start_addr}, end_addr{start_addr + size}, cpu_addr{cpu_addr}, + is_allocated{is_allocated} {} + + constexpr VAddr StartAddr() const { + return start_addr; + } + + constexpr VAddr EndAddr() const { + return end_addr; + } + + constexpr std::size_t Size() const { + return end_addr - start_addr; + } + + constexpr VAddr CpuAddr() const { + return cpu_addr; + } + + constexpr bool IsAllocated() const { + return is_allocated; + } + + private: + GPUVAddr start_addr{}; + GPUVAddr end_addr{}; + VAddr cpu_addr{}; + bool is_allocated{}; + }; + enum class IoctlCommand : u32_le { IocInitalizeExCommand = 0x40284109, IocAllocateSpaceCommand = 0xC0184102, @@ -49,7 +98,7 @@ private: struct IoctlAllocSpace { u32_le pages; u32_le page_size; - u32_le flags; + AddressSpaceFlags flags; INSERT_PADDING_WORDS(1); union { u64_le offset; @@ -69,18 +118,18 @@ private: static_assert(sizeof(IoctlRemapEntry) == 20, "IoctlRemapEntry is incorrect size"); struct IoctlMapBufferEx { - u32_le flags; // bit0: fixed_offset, bit2: cacheable - u32_le kind; // -1 is default + AddressSpaceFlags flags; // bit0: fixed_offset, bit2: cacheable + u32_le kind; // -1 is default u32_le nvmap_handle; u32_le page_size; // 0 means don't care - u64_le buffer_offset; + s64_le buffer_offset; u64_le mapping_size; - u64_le offset; + s64_le offset; }; static_assert(sizeof(IoctlMapBufferEx) == 40, "IoctlMapBufferEx is incorrect size"); struct IoctlUnmapBuffer { - u64_le offset; + s64_le offset; }; static_assert(sizeof(IoctlUnmapBuffer) == 8, "IoctlUnmapBuffer is incorrect size"); @@ -106,15 +155,6 @@ private: static_assert(sizeof(IoctlGetVaRegions) == 16 + sizeof(IoctlVaRegion) * 2, "IoctlGetVaRegions is incorrect size"); - struct BufferMapping { - u64 offset; - u64 size; - u32 nvmap_handle; - }; - - /// Map containing the nvmap object mappings in GPU memory. - std::unordered_map<u64, BufferMapping> buffer_mappings; - u32 channel{}; u32 InitalizeEx(const std::vector<u8>& input, std::vector<u8>& output); @@ -125,7 +165,14 @@ private: u32 BindChannel(const std::vector<u8>& input, std::vector<u8>& output); u32 GetVARegions(const std::vector<u8>& input, std::vector<u8>& output); + std::optional<BufferMap> FindBufferMap(GPUVAddr gpu_addr) const; + void AddBufferMap(GPUVAddr gpu_addr, std::size_t size, VAddr cpu_addr, bool is_allocated); + std::optional<std::size_t> RemoveBufferMap(GPUVAddr gpu_addr); + std::shared_ptr<nvmap> nvmap_dev; + + // This is expected to be ordered, therefore we must use a map, not unordered_map + std::map<GPUVAddr, BufferMap> buffer_mappings; }; } // namespace Service::Nvidia::Devices diff --git a/src/core/hle/service/nvdrv/devices/nvmap.cpp b/src/core/hle/service/nvdrv/devices/nvmap.cpp index 8c742316c..9436e16ad 100644 --- a/src/core/hle/service/nvdrv/devices/nvmap.cpp +++ b/src/core/hle/service/nvdrv/devices/nvmap.cpp @@ -18,7 +18,12 @@ enum { }; } -nvmap::nvmap(Core::System& system) : nvdevice(system) {} +nvmap::nvmap(Core::System& system) : nvdevice(system) { + // Handle 0 appears to be used when remapping, so we create a placeholder empty nvmap object to + // represent this. + CreateObject(0); +} + nvmap::~nvmap() = default; VAddr nvmap::GetObjectAddress(u32 handle) const { @@ -50,6 +55,21 @@ u32 nvmap::ioctl(Ioctl command, const std::vector<u8>& input, const std::vector< return 0; } +u32 nvmap::CreateObject(u32 size) { + // Create a new nvmap object and obtain a handle to it. + auto object = std::make_shared<Object>(); + object->id = next_id++; + object->size = size; + object->status = Object::Status::Created; + object->refcount = 1; + + const u32 handle = next_handle++; + + handles.insert_or_assign(handle, std::move(object)); + + return handle; +} + u32 nvmap::IocCreate(const std::vector<u8>& input, std::vector<u8>& output) { IocCreateParams params; std::memcpy(¶ms, input.data(), sizeof(params)); @@ -59,17 +79,8 @@ u32 nvmap::IocCreate(const std::vector<u8>& input, std::vector<u8>& output) { LOG_ERROR(Service_NVDRV, "Size is 0"); return static_cast<u32>(NvErrCodes::InvalidValue); } - // Create a new nvmap object and obtain a handle to it. - auto object = std::make_shared<Object>(); - object->id = next_id++; - object->size = params.size; - object->status = Object::Status::Created; - object->refcount = 1; - - u32 handle = next_handle++; - handles[handle] = std::move(object); - params.handle = handle; + params.handle = CreateObject(params.size); std::memcpy(output.data(), ¶ms, sizeof(params)); return 0; diff --git a/src/core/hle/service/nvdrv/devices/nvmap.h b/src/core/hle/service/nvdrv/devices/nvmap.h index 73c2e8809..84624be00 100644 --- a/src/core/hle/service/nvdrv/devices/nvmap.h +++ b/src/core/hle/service/nvdrv/devices/nvmap.h @@ -49,10 +49,10 @@ public: private: /// Id to use for the next handle that is created. - u32 next_handle = 1; + u32 next_handle = 0; /// Id to use for the next object that is created. - u32 next_id = 1; + u32 next_id = 0; /// Mapping of currently allocated handles to the objects they represent. std::unordered_map<u32, std::shared_ptr<Object>> handles; @@ -119,6 +119,8 @@ private: }; static_assert(sizeof(IocGetIdParams) == 8, "IocGetIdParams has wrong size"); + u32 CreateObject(u32 size); + u32 IocCreate(const std::vector<u8>& input, std::vector<u8>& output); u32 IocAlloc(const std::vector<u8>& input, std::vector<u8>& output); u32 IocGetId(const std::vector<u8>& input, std::vector<u8>& output); diff --git a/src/core/hle/service/nvdrv/interface.cpp b/src/core/hle/service/nvdrv/interface.cpp index deaf0808b..88fbfa9b0 100644 --- a/src/core/hle/service/nvdrv/interface.cpp +++ b/src/core/hle/service/nvdrv/interface.cpp @@ -60,24 +60,24 @@ void NVDRV::IoctlBase(Kernel::HLERequestContext& ctx, IoctlVersion version) { if (ctrl.must_delay) { ctrl.fresh_call = false; - ctx.SleepClientThread("NVServices::DelayedResponse", ctrl.timeout, - [=](std::shared_ptr<Kernel::Thread> thread, - Kernel::HLERequestContext& ctx, - Kernel::ThreadWakeupReason reason) { - IoctlCtrl ctrl2{ctrl}; - std::vector<u8> tmp_output = output; - std::vector<u8> tmp_output2 = output2; - u32 result = nvdrv->Ioctl(fd, command, input, input2, tmp_output, - tmp_output2, ctrl2, version); - ctx.WriteBuffer(tmp_output, 0); - if (version == IoctlVersion::Version3) { - ctx.WriteBuffer(tmp_output2, 1); - } - IPC::ResponseBuilder rb{ctx, 3}; - rb.Push(RESULT_SUCCESS); - rb.Push(result); - }, - nvdrv->GetEventWriteable(ctrl.event_id)); + ctx.SleepClientThread( + "NVServices::DelayedResponse", ctrl.timeout, + [=, this](std::shared_ptr<Kernel::Thread> thread, Kernel::HLERequestContext& ctx_, + Kernel::ThreadWakeupReason reason) { + IoctlCtrl ctrl2{ctrl}; + std::vector<u8> tmp_output = output; + std::vector<u8> tmp_output2 = output2; + const u32 ioctl_result = nvdrv->Ioctl(fd, command, input, input2, tmp_output, + tmp_output2, ctrl2, version); + ctx_.WriteBuffer(tmp_output, 0); + if (version == IoctlVersion::Version3) { + ctx_.WriteBuffer(tmp_output2, 1); + } + IPC::ResponseBuilder rb{ctx_, 3}; + rb.Push(RESULT_SUCCESS); + rb.Push(ioctl_result); + }, + nvdrv->GetEventWriteable(ctrl.event_id)); } else { ctx.WriteBuffer(output); if (version == IoctlVersion::Version3) { diff --git a/src/core/hle/service/sm/sm.h b/src/core/hle/service/sm/sm.h index b06d2f103..b526a94fe 100644 --- a/src/core/hle/service/sm/sm.h +++ b/src/core/hle/service/sm/sm.h @@ -9,6 +9,7 @@ #include <type_traits> #include <unordered_map> +#include "common/concepts.h" #include "core/hle/kernel/client_port.h" #include "core/hle/kernel/object.h" #include "core/hle/kernel/server_port.h" @@ -56,10 +57,8 @@ public: ResultVal<std::shared_ptr<Kernel::ClientPort>> GetServicePort(const std::string& name); ResultVal<std::shared_ptr<Kernel::ClientSession>> ConnectToService(const std::string& name); - template <typename T> + template <Common::IsBaseOf<Kernel::SessionRequestHandler> T> std::shared_ptr<T> GetService(const std::string& service_name) const { - static_assert(std::is_base_of_v<Kernel::SessionRequestHandler, T>, - "Not a base of ServiceFrameworkBase"); auto service = registered_services.find(service_name); if (service == registered_services.end()) { LOG_DEBUG(Service, "Can't find service: {}", service_name); diff --git a/src/core/hle/service/vi/vi.cpp b/src/core/hle/service/vi/vi.cpp index 825d11a3f..480d34725 100644 --- a/src/core/hle/service/vi/vi.cpp +++ b/src/core/hle/service/vi/vi.cpp @@ -548,8 +548,8 @@ private: // Wait the current thread until a buffer becomes available ctx.SleepClientThread( "IHOSBinderDriver::DequeueBuffer", UINT64_MAX, - [=](std::shared_ptr<Kernel::Thread> thread, Kernel::HLERequestContext& ctx, - Kernel::ThreadWakeupReason reason) { + [=, this](std::shared_ptr<Kernel::Thread> thread, + Kernel::HLERequestContext& ctx, Kernel::ThreadWakeupReason reason) { // Repeat TransactParcel DequeueBuffer when a buffer is available const auto guard = nv_flinger->Lock(); auto& buffer_queue = nv_flinger->FindBufferQueue(id); @@ -1199,6 +1199,23 @@ private: } } + void GetIndirectLayerImageRequiredMemoryInfo(Kernel::HLERequestContext& ctx) { + IPC::RequestParser rp{ctx}; + const auto width = rp.Pop<u64>(); + const auto height = rp.Pop<u64>(); + LOG_DEBUG(Service_VI, "called width={}, height={}", width, height); + + constexpr std::size_t base_size = 0x20000; + constexpr std::size_t alignment = 0x1000; + const auto texture_size = width * height * 4; + const auto out_size = (texture_size + base_size - 1) / base_size * base_size; + + IPC::ResponseBuilder rb{ctx, 6}; + rb.Push(RESULT_SUCCESS); + rb.Push(out_size); + rb.Push(alignment); + } + static ResultVal<ConvertedScaleMode> ConvertScalingModeImpl(NintendoScaleMode mode) { switch (mode) { case NintendoScaleMode::None: @@ -1243,7 +1260,8 @@ IApplicationDisplayService::IApplicationDisplayService( {2102, &IApplicationDisplayService::ConvertScalingMode, "ConvertScalingMode"}, {2450, nullptr, "GetIndirectLayerImageMap"}, {2451, nullptr, "GetIndirectLayerImageCropMap"}, - {2460, nullptr, "GetIndirectLayerImageRequiredMemoryInfo"}, + {2460, &IApplicationDisplayService::GetIndirectLayerImageRequiredMemoryInfo, + "GetIndirectLayerImageRequiredMemoryInfo"}, {5202, &IApplicationDisplayService::GetDisplayVsyncEvent, "GetDisplayVsyncEvent"}, {5203, nullptr, "GetDisplayVsyncEventForDebug"}, }; diff --git a/src/core/memory/dmnt_cheat_vm.cpp b/src/core/memory/dmnt_cheat_vm.cpp index 2e7da23fe..48be80c12 100644 --- a/src/core/memory/dmnt_cheat_vm.cpp +++ b/src/core/memory/dmnt_cheat_vm.cpp @@ -313,30 +313,32 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { switch (opcode_type) { case CheatVmOpcodeType::StoreStatic: { - StoreStaticOpcode store_static{}; // 0TMR00AA AAAAAAAA YYYYYYYY (YYYYYYYY) // Read additional words. const u32 second_dword = GetNextDword(); - store_static.bit_width = (first_dword >> 24) & 0xF; - store_static.mem_type = static_cast<MemoryAccessType>((first_dword >> 20) & 0xF); - store_static.offset_register = ((first_dword >> 16) & 0xF); - store_static.rel_address = - (static_cast<u64>(first_dword & 0xFF) << 32ul) | static_cast<u64>(second_dword); - store_static.value = GetNextVmInt(store_static.bit_width); - opcode.opcode = store_static; + const u32 bit_width = (first_dword >> 24) & 0xF; + + opcode.opcode = StoreStaticOpcode{ + .bit_width = bit_width, + .mem_type = static_cast<MemoryAccessType>((first_dword >> 20) & 0xF), + .offset_register = (first_dword >> 16) & 0xF, + .rel_address = (static_cast<u64>(first_dword & 0xFF) << 32) | second_dword, + .value = GetNextVmInt(bit_width), + }; } break; case CheatVmOpcodeType::BeginConditionalBlock: { - BeginConditionalOpcode begin_cond{}; // 1TMC00AA AAAAAAAA YYYYYYYY (YYYYYYYY) // Read additional words. const u32 second_dword = GetNextDword(); - begin_cond.bit_width = (first_dword >> 24) & 0xF; - begin_cond.mem_type = static_cast<MemoryAccessType>((first_dword >> 20) & 0xF); - begin_cond.cond_type = static_cast<ConditionalComparisonType>((first_dword >> 16) & 0xF); - begin_cond.rel_address = - (static_cast<u64>(first_dword & 0xFF) << 32ul) | static_cast<u64>(second_dword); - begin_cond.value = GetNextVmInt(begin_cond.bit_width); - opcode.opcode = begin_cond; + const u32 bit_width = (first_dword >> 24) & 0xF; + + opcode.opcode = BeginConditionalOpcode{ + .bit_width = bit_width, + .mem_type = static_cast<MemoryAccessType>((first_dword >> 20) & 0xF), + .cond_type = static_cast<ConditionalComparisonType>((first_dword >> 16) & 0xF), + .rel_address = (static_cast<u64>(first_dword & 0xFF) << 32) | second_dword, + .value = GetNextVmInt(bit_width), + }; } break; case CheatVmOpcodeType::EndConditionalBlock: { // 20000000 @@ -344,12 +346,14 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { opcode.opcode = EndConditionalOpcode{}; } break; case CheatVmOpcodeType::ControlLoop: { - ControlLoopOpcode ctrl_loop{}; // 300R0000 VVVVVVVV // 310R0000 // Parse register, whether loop start or loop end. - ctrl_loop.start_loop = ((first_dword >> 24) & 0xF) == 0; - ctrl_loop.reg_index = ((first_dword >> 20) & 0xF); + ControlLoopOpcode ctrl_loop{ + .start_loop = ((first_dword >> 24) & 0xF) == 0, + .reg_index = (first_dword >> 20) & 0xF, + .num_iters = 0, + }; // Read number of iters if loop start. if (ctrl_loop.start_loop) { @@ -358,66 +362,65 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { opcode.opcode = ctrl_loop; } break; case CheatVmOpcodeType::LoadRegisterStatic: { - LoadRegisterStaticOpcode ldr_static{}; // 400R0000 VVVVVVVV VVVVVVVV // Read additional words. - ldr_static.reg_index = ((first_dword >> 16) & 0xF); - ldr_static.value = - (static_cast<u64>(GetNextDword()) << 32ul) | static_cast<u64>(GetNextDword()); - opcode.opcode = ldr_static; + opcode.opcode = LoadRegisterStaticOpcode{ + .reg_index = (first_dword >> 16) & 0xF, + .value = (static_cast<u64>(GetNextDword()) << 32) | GetNextDword(), + }; } break; case CheatVmOpcodeType::LoadRegisterMemory: { - LoadRegisterMemoryOpcode ldr_memory{}; // 5TMRI0AA AAAAAAAA // Read additional words. const u32 second_dword = GetNextDword(); - ldr_memory.bit_width = (first_dword >> 24) & 0xF; - ldr_memory.mem_type = static_cast<MemoryAccessType>((first_dword >> 20) & 0xF); - ldr_memory.reg_index = ((first_dword >> 16) & 0xF); - ldr_memory.load_from_reg = ((first_dword >> 12) & 0xF) != 0; - ldr_memory.rel_address = - (static_cast<u64>(first_dword & 0xFF) << 32ul) | static_cast<u64>(second_dword); - opcode.opcode = ldr_memory; + opcode.opcode = LoadRegisterMemoryOpcode{ + .bit_width = (first_dword >> 24) & 0xF, + .mem_type = static_cast<MemoryAccessType>((first_dword >> 20) & 0xF), + .reg_index = ((first_dword >> 16) & 0xF), + .load_from_reg = ((first_dword >> 12) & 0xF) != 0, + .rel_address = (static_cast<u64>(first_dword & 0xFF) << 32) | second_dword, + }; } break; case CheatVmOpcodeType::StoreStaticToAddress: { - StoreStaticToAddressOpcode str_static{}; // 6T0RIor0 VVVVVVVV VVVVVVVV // Read additional words. - str_static.bit_width = (first_dword >> 24) & 0xF; - str_static.reg_index = ((first_dword >> 16) & 0xF); - str_static.increment_reg = ((first_dword >> 12) & 0xF) != 0; - str_static.add_offset_reg = ((first_dword >> 8) & 0xF) != 0; - str_static.offset_reg_index = ((first_dword >> 4) & 0xF); - str_static.value = - (static_cast<u64>(GetNextDword()) << 32ul) | static_cast<u64>(GetNextDword()); - opcode.opcode = str_static; + opcode.opcode = StoreStaticToAddressOpcode{ + .bit_width = (first_dword >> 24) & 0xF, + .reg_index = (first_dword >> 16) & 0xF, + .increment_reg = ((first_dword >> 12) & 0xF) != 0, + .add_offset_reg = ((first_dword >> 8) & 0xF) != 0, + .offset_reg_index = (first_dword >> 4) & 0xF, + .value = (static_cast<u64>(GetNextDword()) << 32) | GetNextDword(), + }; } break; case CheatVmOpcodeType::PerformArithmeticStatic: { - PerformArithmeticStaticOpcode perform_math_static{}; // 7T0RC000 VVVVVVVV // Read additional words. - perform_math_static.bit_width = (first_dword >> 24) & 0xF; - perform_math_static.reg_index = ((first_dword >> 16) & 0xF); - perform_math_static.math_type = - static_cast<RegisterArithmeticType>((first_dword >> 12) & 0xF); - perform_math_static.value = GetNextDword(); - opcode.opcode = perform_math_static; + opcode.opcode = PerformArithmeticStaticOpcode{ + .bit_width = (first_dword >> 24) & 0xF, + .reg_index = ((first_dword >> 16) & 0xF), + .math_type = static_cast<RegisterArithmeticType>((first_dword >> 12) & 0xF), + .value = GetNextDword(), + }; } break; case CheatVmOpcodeType::BeginKeypressConditionalBlock: { - BeginKeypressConditionalOpcode begin_keypress_cond{}; // 8kkkkkkk // Just parse the mask. - begin_keypress_cond.key_mask = first_dword & 0x0FFFFFFF; - opcode.opcode = begin_keypress_cond; + opcode.opcode = BeginKeypressConditionalOpcode{ + .key_mask = first_dword & 0x0FFFFFFF, + }; } break; case CheatVmOpcodeType::PerformArithmeticRegister: { - PerformArithmeticRegisterOpcode perform_math_reg{}; // 9TCRSIs0 (VVVVVVVV (VVVVVVVV)) - perform_math_reg.bit_width = (first_dword >> 24) & 0xF; - perform_math_reg.math_type = static_cast<RegisterArithmeticType>((first_dword >> 20) & 0xF); - perform_math_reg.dst_reg_index = ((first_dword >> 16) & 0xF); - perform_math_reg.src_reg_1_index = ((first_dword >> 12) & 0xF); - perform_math_reg.has_immediate = ((first_dword >> 8) & 0xF) != 0; + PerformArithmeticRegisterOpcode perform_math_reg{ + .bit_width = (first_dword >> 24) & 0xF, + .math_type = static_cast<RegisterArithmeticType>((first_dword >> 20) & 0xF), + .dst_reg_index = (first_dword >> 16) & 0xF, + .src_reg_1_index = (first_dword >> 12) & 0xF, + .src_reg_2_index = 0, + .has_immediate = ((first_dword >> 8) & 0xF) != 0, + .value = {}, + }; if (perform_math_reg.has_immediate) { perform_math_reg.src_reg_2_index = 0; perform_math_reg.value = GetNextVmInt(perform_math_reg.bit_width); @@ -427,7 +430,6 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { opcode.opcode = perform_math_reg; } break; case CheatVmOpcodeType::StoreRegisterToAddress: { - StoreRegisterToAddressOpcode str_register{}; // ATSRIOxa (aaaaaaaa) // A = opcode 10 // T = bit width @@ -439,20 +441,23 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { // Relative Address // x = offset register (for offset type 1), memory type (for offset type 3) // a = relative address (for offset type 2+3) - str_register.bit_width = (first_dword >> 24) & 0xF; - str_register.str_reg_index = ((first_dword >> 20) & 0xF); - str_register.addr_reg_index = ((first_dword >> 16) & 0xF); - str_register.increment_reg = ((first_dword >> 12) & 0xF) != 0; - str_register.ofs_type = static_cast<StoreRegisterOffsetType>(((first_dword >> 8) & 0xF)); - str_register.ofs_reg_index = ((first_dword >> 4) & 0xF); + StoreRegisterToAddressOpcode str_register{ + .bit_width = (first_dword >> 24) & 0xF, + .str_reg_index = (first_dword >> 20) & 0xF, + .addr_reg_index = (first_dword >> 16) & 0xF, + .increment_reg = ((first_dword >> 12) & 0xF) != 0, + .ofs_type = static_cast<StoreRegisterOffsetType>(((first_dword >> 8) & 0xF)), + .mem_type = MemoryAccessType::MainNso, + .ofs_reg_index = (first_dword >> 4) & 0xF, + .rel_address = 0, + }; switch (str_register.ofs_type) { case StoreRegisterOffsetType::None: case StoreRegisterOffsetType::Reg: // Nothing more to do break; case StoreRegisterOffsetType::Imm: - str_register.rel_address = - ((static_cast<u64>(first_dword & 0xF) << 32ul) | static_cast<u64>(GetNextDword())); + str_register.rel_address = (static_cast<u64>(first_dword & 0xF) << 32) | GetNextDword(); break; case StoreRegisterOffsetType::MemReg: str_register.mem_type = static_cast<MemoryAccessType>((first_dword >> 4) & 0xF); @@ -460,8 +465,7 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { case StoreRegisterOffsetType::MemImm: case StoreRegisterOffsetType::MemImmReg: str_register.mem_type = static_cast<MemoryAccessType>((first_dword >> 4) & 0xF); - str_register.rel_address = - ((static_cast<u64>(first_dword & 0xF) << 32ul) | static_cast<u64>(GetNextDword())); + str_register.rel_address = (static_cast<u64>(first_dword & 0xF) << 32) | GetNextDword(); break; default: str_register.ofs_type = StoreRegisterOffsetType::None; @@ -470,7 +474,6 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { opcode.opcode = str_register; } break; case CheatVmOpcodeType::BeginRegisterConditionalBlock: { - BeginRegisterConditionalOpcode begin_reg_cond{}; // C0TcSX## // C0TcS0Ma aaaaaaaa // C0TcS1Mr @@ -492,11 +495,19 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { // r = offset register. // X = other register. // V = value. - begin_reg_cond.bit_width = (first_dword >> 20) & 0xF; - begin_reg_cond.cond_type = - static_cast<ConditionalComparisonType>((first_dword >> 16) & 0xF); - begin_reg_cond.val_reg_index = ((first_dword >> 12) & 0xF); - begin_reg_cond.comp_type = static_cast<CompareRegisterValueType>((first_dword >> 8) & 0xF); + + BeginRegisterConditionalOpcode begin_reg_cond{ + .bit_width = (first_dword >> 20) & 0xF, + .cond_type = static_cast<ConditionalComparisonType>((first_dword >> 16) & 0xF), + .val_reg_index = (first_dword >> 12) & 0xF, + .comp_type = static_cast<CompareRegisterValueType>((first_dword >> 8) & 0xF), + .mem_type = MemoryAccessType::MainNso, + .addr_reg_index = 0, + .other_reg_index = 0, + .ofs_reg_index = 0, + .rel_address = 0, + .value = {}, + }; switch (begin_reg_cond.comp_type) { case CompareRegisterValueType::StaticValue: @@ -508,26 +519,25 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { case CompareRegisterValueType::MemoryRelAddr: begin_reg_cond.mem_type = static_cast<MemoryAccessType>((first_dword >> 4) & 0xF); begin_reg_cond.rel_address = - ((static_cast<u64>(first_dword & 0xF) << 32ul) | static_cast<u64>(GetNextDword())); + (static_cast<u64>(first_dword & 0xF) << 32) | GetNextDword(); break; case CompareRegisterValueType::MemoryOfsReg: begin_reg_cond.mem_type = static_cast<MemoryAccessType>((first_dword >> 4) & 0xF); begin_reg_cond.ofs_reg_index = (first_dword & 0xF); break; case CompareRegisterValueType::RegisterRelAddr: - begin_reg_cond.addr_reg_index = ((first_dword >> 4) & 0xF); + begin_reg_cond.addr_reg_index = (first_dword >> 4) & 0xF; begin_reg_cond.rel_address = - ((static_cast<u64>(first_dword & 0xF) << 32ul) | static_cast<u64>(GetNextDword())); + (static_cast<u64>(first_dword & 0xF) << 32) | GetNextDword(); break; case CompareRegisterValueType::RegisterOfsReg: - begin_reg_cond.addr_reg_index = ((first_dword >> 4) & 0xF); - begin_reg_cond.ofs_reg_index = (first_dword & 0xF); + begin_reg_cond.addr_reg_index = (first_dword >> 4) & 0xF; + begin_reg_cond.ofs_reg_index = first_dword & 0xF; break; } opcode.opcode = begin_reg_cond; } break; case CheatVmOpcodeType::SaveRestoreRegister: { - SaveRestoreRegisterOpcode save_restore_reg{}; // C10D0Sx0 // C1 = opcode 0xC1 // D = destination index. @@ -535,36 +545,37 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { // x = 3 if clearing reg, 2 if clearing saved value, 1 if saving a register, 0 if restoring // a register. // NOTE: If we add more save slots later, current encoding is backwards compatible. - save_restore_reg.dst_index = (first_dword >> 16) & 0xF; - save_restore_reg.src_index = (first_dword >> 8) & 0xF; - save_restore_reg.op_type = static_cast<SaveRestoreRegisterOpType>((first_dword >> 4) & 0xF); - opcode.opcode = save_restore_reg; + opcode.opcode = SaveRestoreRegisterOpcode{ + .dst_index = (first_dword >> 16) & 0xF, + .src_index = (first_dword >> 8) & 0xF, + .op_type = static_cast<SaveRestoreRegisterOpType>((first_dword >> 4) & 0xF), + }; } break; case CheatVmOpcodeType::SaveRestoreRegisterMask: { - SaveRestoreRegisterMaskOpcode save_restore_regmask{}; // C2x0XXXX // C2 = opcode 0xC2 // x = 3 if clearing reg, 2 if clearing saved value, 1 if saving, 0 if restoring. // X = 16-bit bitmask, bit i --> save or restore register i. - save_restore_regmask.op_type = - static_cast<SaveRestoreRegisterOpType>((first_dword >> 20) & 0xF); + SaveRestoreRegisterMaskOpcode save_restore_regmask{ + .op_type = static_cast<SaveRestoreRegisterOpType>((first_dword >> 20) & 0xF), + .should_operate = {}, + }; for (std::size_t i = 0; i < NumRegisters; i++) { - save_restore_regmask.should_operate[i] = (first_dword & (1u << i)) != 0; + save_restore_regmask.should_operate[i] = (first_dword & (1U << i)) != 0; } opcode.opcode = save_restore_regmask; } break; case CheatVmOpcodeType::ReadWriteStaticRegister: { - ReadWriteStaticRegisterOpcode rw_static_reg{}; // C3000XXx // C3 = opcode 0xC3. // XX = static register index. // x = register index. - rw_static_reg.static_idx = ((first_dword >> 4) & 0xFF); - rw_static_reg.idx = (first_dword & 0xF); - opcode.opcode = rw_static_reg; + opcode.opcode = ReadWriteStaticRegisterOpcode{ + .static_idx = (first_dword >> 4) & 0xFF, + .idx = first_dword & 0xF, + }; } break; case CheatVmOpcodeType::DebugLog: { - DebugLogOpcode debug_log{}; // FFFTIX## // FFFTI0Ma aaaaaaaa // FFFTI1Mr @@ -583,31 +594,36 @@ bool DmntCheatVm::DecodeNextOpcode(CheatVmOpcode& out) { // a = relative address. // r = offset register. // X = value register. - debug_log.bit_width = (first_dword >> 16) & 0xF; - debug_log.log_id = ((first_dword >> 12) & 0xF); - debug_log.val_type = static_cast<DebugLogValueType>((first_dword >> 8) & 0xF); + DebugLogOpcode debug_log{ + .bit_width = (first_dword >> 16) & 0xF, + .log_id = (first_dword >> 12) & 0xF, + .val_type = static_cast<DebugLogValueType>((first_dword >> 8) & 0xF), + .mem_type = MemoryAccessType::MainNso, + .addr_reg_index = 0, + .val_reg_index = 0, + .ofs_reg_index = 0, + .rel_address = 0, + }; switch (debug_log.val_type) { case DebugLogValueType::RegisterValue: - debug_log.val_reg_index = ((first_dword >> 4) & 0xF); + debug_log.val_reg_index = (first_dword >> 4) & 0xF; break; case DebugLogValueType::MemoryRelAddr: debug_log.mem_type = static_cast<MemoryAccessType>((first_dword >> 4) & 0xF); - debug_log.rel_address = - ((static_cast<u64>(first_dword & 0xF) << 32ul) | static_cast<u64>(GetNextDword())); + debug_log.rel_address = (static_cast<u64>(first_dword & 0xF) << 32) | GetNextDword(); break; case DebugLogValueType::MemoryOfsReg: debug_log.mem_type = static_cast<MemoryAccessType>((first_dword >> 4) & 0xF); - debug_log.ofs_reg_index = (first_dword & 0xF); + debug_log.ofs_reg_index = first_dword & 0xF; break; case DebugLogValueType::RegisterRelAddr: - debug_log.addr_reg_index = ((first_dword >> 4) & 0xF); - debug_log.rel_address = - ((static_cast<u64>(first_dword & 0xF) << 32ul) | static_cast<u64>(GetNextDword())); + debug_log.addr_reg_index = (first_dword >> 4) & 0xF; + debug_log.rel_address = (static_cast<u64>(first_dword & 0xF) << 32) | GetNextDword(); break; case DebugLogValueType::RegisterOfsReg: - debug_log.addr_reg_index = ((first_dword >> 4) & 0xF); - debug_log.ofs_reg_index = (first_dword & 0xF); + debug_log.addr_reg_index = (first_dword >> 4) & 0xF; + debug_log.ofs_reg_index = first_dword & 0xF; break; } opcode.opcode = debug_log; diff --git a/src/core/perf_stats.cpp b/src/core/perf_stats.cpp index 29339ead7..b899ac884 100644 --- a/src/core/perf_stats.cpp +++ b/src/core/perf_stats.cpp @@ -74,15 +74,16 @@ void PerfStats::EndGameFrame() { game_frames += 1; } -double PerfStats::GetMeanFrametime() { +double PerfStats::GetMeanFrametime() const { std::lock_guard lock{object_mutex}; if (current_index <= IgnoreFrames) { return 0; } + const double sum = std::accumulate(perf_history.begin() + IgnoreFrames, perf_history.begin() + current_index, 0.0); - return sum / (current_index - IgnoreFrames); + return sum / static_cast<double>(current_index - IgnoreFrames); } PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us) { @@ -94,12 +95,13 @@ PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us const auto system_us_per_second = (current_system_time_us - reset_point_system_us) / interval; - 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() / - static_cast<double>(system_frames); - results.emulation_speed = system_us_per_second.count() / 1'000'000.0; + const PerfStatsResults results{ + .system_fps = static_cast<double>(system_frames) / interval, + .game_fps = static_cast<double>(game_frames) / interval, + .frametime = duration_cast<DoubleSecs>(accumulated_frametime).count() / + static_cast<double>(system_frames), + .emulation_speed = system_us_per_second.count() / 1'000'000.0, + }; // Reset counters reset_point = now; @@ -111,7 +113,7 @@ PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us return results; } -double PerfStats::GetLastFrameTimeScale() { +double PerfStats::GetLastFrameTimeScale() const { std::lock_guard lock{object_mutex}; constexpr double FRAME_LENGTH = 1.0 / 60; diff --git a/src/core/perf_stats.h b/src/core/perf_stats.h index d9a64f072..69256b960 100644 --- a/src/core/perf_stats.h +++ b/src/core/perf_stats.h @@ -30,7 +30,6 @@ struct PerfStatsResults { class PerfStats { public: explicit PerfStats(u64 title_id); - ~PerfStats(); using Clock = std::chrono::high_resolution_clock; @@ -42,18 +41,18 @@ public: PerfStatsResults GetAndResetStats(std::chrono::microseconds current_system_time_us); /** - * Returns the Arthimetic Mean of all frametime values stored in the performance history. + * Returns the arithmetic mean of all frametime values stored in the performance history. */ - double GetMeanFrametime(); + double GetMeanFrametime() const; /** * Gets the ratio between walltime and the emulated time of the previous system frame. This is * useful for scaling inputs or outputs moving between the two time domains. */ - double GetLastFrameTimeScale(); + double GetLastFrameTimeScale() const; private: - std::mutex object_mutex{}; + mutable std::mutex object_mutex; /// Title ID for the game that is running. 0 if there is no game running yet u64 title_id{0}; @@ -61,7 +60,7 @@ private: std::size_t current_index{0}; /// Stores an hour of historical frametime data useful for processing and tracking performance /// regressions with code changes. - std::array<double, 216000> perf_history = {}; + std::array<double, 216000> perf_history{}; /// Point when the cumulative counters were reset Clock::time_point reset_point = Clock::now(); diff --git a/src/tests/core/core_timing.cpp b/src/tests/core/core_timing.cpp index 022b26e6d..b35459152 100644 --- a/src/tests/core/core_timing.cpp +++ b/src/tests/core/core_timing.cpp @@ -46,20 +46,16 @@ struct ScopeInit final { Core::Timing::CoreTiming core_timing; }; -#pragma optimize("", off) - u64 TestTimerSpeed(Core::Timing::CoreTiming& core_timing) { - u64 start = core_timing.GetGlobalTimeNs().count(); - u64 placebo = 0; + const u64 start = core_timing.GetGlobalTimeNs().count(); + volatile u64 placebo = 0; for (std::size_t i = 0; i < 1000; i++) { - placebo += core_timing.GetGlobalTimeNs().count(); + placebo = placebo + core_timing.GetGlobalTimeNs().count(); } - u64 end = core_timing.GetGlobalTimeNs().count(); - return (end - start); + const u64 end = core_timing.GetGlobalTimeNs().count(); + return end - start; } -#pragma optimize("", on) - } // Anonymous namespace TEST_CASE("CoreTiming[BasicOrder]", "[core]") { diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 8e19c3373..512578c8b 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -81,7 +81,7 @@ void GPU::WaitFence(u32 syncpoint_id, u32 value) { } MICROPROFILE_SCOPE(GPU_wait); std::unique_lock lock{sync_mutex}; - sync_cv.wait(lock, [=]() { return syncpoints[syncpoint_id].load() >= value; }); + sync_cv.wait(lock, [=, this] { return syncpoints[syncpoint_id].load() >= value; }); } void GPU::IncrementSyncPoint(const u32 syncpoint_id) { diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index ff5505d12..844164645 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -4,7 +4,6 @@ #include "common/alignment.h" #include "common/assert.h" -#include "common/logging/log.h" #include "core/core.h" #include "core/hle/kernel/memory/page_table.h" #include "core/hle/kernel/process.h" @@ -16,121 +15,137 @@ namespace Tegra { MemoryManager::MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer) - : rasterizer{rasterizer}, system{system} { - page_table.Resize(address_space_width, page_bits, false); - - // Initialize the map with a single free region covering the entire managed space. - VirtualMemoryArea initial_vma; - initial_vma.size = address_space_end; - vma_map.emplace(initial_vma.base, initial_vma); - - UpdatePageTableForVMA(initial_vma); -} + : system{system}, rasterizer{rasterizer}, page_table(page_table_size) {} MemoryManager::~MemoryManager() = default; -GPUVAddr MemoryManager::AllocateSpace(u64 size, u64 align) { - const u64 aligned_size{Common::AlignUp(size, page_size)}; - const GPUVAddr gpu_addr{FindFreeRegion(address_space_base, aligned_size)}; - - AllocateMemory(gpu_addr, 0, aligned_size); - +GPUVAddr MemoryManager::UpdateRange(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size) { + u64 remaining_size{size}; + for (u64 offset{}; offset < size; offset += page_size) { + if (remaining_size < page_size) { + SetPageEntry(gpu_addr + offset, page_entry + offset, remaining_size); + } else { + SetPageEntry(gpu_addr + offset, page_entry + offset); + } + remaining_size -= page_size; + } return gpu_addr; } -GPUVAddr MemoryManager::AllocateSpace(GPUVAddr gpu_addr, u64 size, u64 align) { - const u64 aligned_size{Common::AlignUp(size, page_size)}; - - AllocateMemory(gpu_addr, 0, aligned_size); +GPUVAddr MemoryManager::Map(VAddr cpu_addr, GPUVAddr gpu_addr, std::size_t size) { + return UpdateRange(gpu_addr, cpu_addr, size); +} - return gpu_addr; +GPUVAddr MemoryManager::MapAllocate(VAddr cpu_addr, std::size_t size, std::size_t align) { + return Map(cpu_addr, *FindFreeRange(size, align), size); } -GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, u64 size) { - const u64 aligned_size{Common::AlignUp(size, page_size)}; - const GPUVAddr gpu_addr{FindFreeRegion(address_space_base, aligned_size)}; +void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) { + if (!size) { + return; + } - MapBackingMemory(gpu_addr, system.Memory().GetPointer(cpu_addr), aligned_size, cpu_addr); - ASSERT( - system.CurrentProcess()->PageTable().LockForDeviceAddressSpace(cpu_addr, size).IsSuccess()); + // Flush and invalidate through the GPU interface, to be asynchronous if possible. + system.GPU().FlushAndInvalidateRegion(*GpuToCpuAddress(gpu_addr), size); - return gpu_addr; + UpdateRange(gpu_addr, PageEntry::State::Unmapped, size); } -GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size) { - ASSERT((gpu_addr & page_mask) == 0); +std::optional<GPUVAddr> MemoryManager::AllocateFixed(GPUVAddr gpu_addr, std::size_t size) { + for (u64 offset{}; offset < size; offset += page_size) { + if (!GetPageEntry(gpu_addr + offset).IsUnmapped()) { + return {}; + } + } - const u64 aligned_size{Common::AlignUp(size, page_size)}; + return UpdateRange(gpu_addr, PageEntry::State::Allocated, size); +} - MapBackingMemory(gpu_addr, system.Memory().GetPointer(cpu_addr), aligned_size, cpu_addr); - ASSERT( - system.CurrentProcess()->PageTable().LockForDeviceAddressSpace(cpu_addr, size).IsSuccess()); - return gpu_addr; +GPUVAddr MemoryManager::Allocate(std::size_t size, std::size_t align) { + return *AllocateFixed(*FindFreeRange(size, align), size); } -GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) { - ASSERT((gpu_addr & page_mask) == 0); +void MemoryManager::TryLockPage(PageEntry page_entry, std::size_t size) { + if (!page_entry.IsValid()) { + return; + } - const u64 aligned_size{Common::AlignUp(size, page_size)}; - const auto cpu_addr = GpuToCpuAddress(gpu_addr); - ASSERT(cpu_addr); + ASSERT(system.CurrentProcess() + ->PageTable() + .LockForDeviceAddressSpace(page_entry.ToAddress(), size) + .IsSuccess()); +} - // Flush and invalidate through the GPU interface, to be asynchronous if possible. - system.GPU().FlushAndInvalidateRegion(*cpu_addr, aligned_size); +void MemoryManager::TryUnlockPage(PageEntry page_entry, std::size_t size) { + if (!page_entry.IsValid()) { + return; + } - UnmapRange(gpu_addr, aligned_size); ASSERT(system.CurrentProcess() ->PageTable() - .UnlockForDeviceAddressSpace(cpu_addr.value(), size) + .UnlockForDeviceAddressSpace(page_entry.ToAddress(), size) .IsSuccess()); - - return gpu_addr; } -GPUVAddr MemoryManager::FindFreeRegion(GPUVAddr region_start, u64 size) const { - // Find the first Free VMA. - const VMAHandle vma_handle{ - std::find_if(vma_map.begin(), vma_map.end(), [region_start, size](const auto& vma) { - if (vma.second.type != VirtualMemoryArea::Type::Unmapped) { - return false; - } +PageEntry MemoryManager::GetPageEntry(GPUVAddr gpu_addr) const { + return page_table[PageEntryIndex(gpu_addr)]; +} - const VAddr vma_end{vma.second.base + vma.second.size}; - return vma_end > region_start && vma_end >= region_start + size; - })}; +void MemoryManager::SetPageEntry(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size) { + // TODO(bunnei): We should lock/unlock device regions. This currently causes issues due to + // improper tracking, but should be fixed in the future. - if (vma_handle == vma_map.end()) { - return {}; - } + //// Unlock the old page + // TryUnlockPage(page_table[PageEntryIndex(gpu_addr)], size); - return std::max(region_start, vma_handle->second.base); -} + //// Lock the new page + // TryLockPage(page_entry, size); -bool MemoryManager::IsAddressValid(GPUVAddr addr) const { - return (addr >> page_bits) < page_table.pointers.size(); + page_table[PageEntryIndex(gpu_addr)] = page_entry; } -std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr addr) const { - if (!IsAddressValid(addr)) { - return {}; +std::optional<GPUVAddr> MemoryManager::FindFreeRange(std::size_t size, std::size_t align) const { + if (!align) { + align = page_size; + } else { + align = Common::AlignUp(align, page_size); } - const VAddr cpu_addr{page_table.backing_addr[addr >> page_bits]}; - if (cpu_addr) { - return cpu_addr + (addr & page_mask); + u64 available_size{}; + GPUVAddr gpu_addr{address_space_start}; + while (gpu_addr + available_size < address_space_size) { + if (GetPageEntry(gpu_addr + available_size).IsUnmapped()) { + available_size += page_size; + + if (available_size >= size) { + return gpu_addr; + } + } else { + gpu_addr += available_size + page_size; + available_size = 0; + + const auto remainder{gpu_addr % align}; + if (remainder) { + gpu_addr = (gpu_addr - remainder) + align; + } + } } return {}; } -template <typename T> -T MemoryManager::Read(GPUVAddr addr) const { - if (!IsAddressValid(addr)) { +std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) const { + const auto page_entry{GetPageEntry(gpu_addr)}; + if (!page_entry.IsValid()) { return {}; } - const u8* page_pointer{GetPointer(addr)}; - if (page_pointer) { + return page_entry.ToAddress() + (gpu_addr & page_mask); +} + +template <typename T> +T MemoryManager::Read(GPUVAddr addr) const { + if (auto page_pointer{GetPointer(addr)}; page_pointer) { // NOTE: Avoid adding any extra logic to this fast-path block T value; std::memcpy(&value, page_pointer, sizeof(T)); @@ -144,12 +159,7 @@ T MemoryManager::Read(GPUVAddr addr) const { template <typename T> void MemoryManager::Write(GPUVAddr addr, T data) { - if (!IsAddressValid(addr)) { - return; - } - - u8* page_pointer{GetPointer(addr)}; - if (page_pointer) { + if (auto page_pointer{GetPointer(addr)}; page_pointer) { // NOTE: Avoid adding any extra logic to this fast-path block std::memcpy(page_pointer, &data, sizeof(T)); return; @@ -167,66 +177,49 @@ template void MemoryManager::Write<u16>(GPUVAddr addr, u16 data); template void MemoryManager::Write<u32>(GPUVAddr addr, u32 data); template void MemoryManager::Write<u64>(GPUVAddr addr, u64 data); -u8* MemoryManager::GetPointer(GPUVAddr addr) { - if (!IsAddressValid(addr)) { +u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) { + if (!GetPageEntry(gpu_addr).IsValid()) { return {}; } - auto& memory = system.Memory(); - - const VAddr page_addr{page_table.backing_addr[addr >> page_bits]}; - - if (page_addr != 0) { - return memory.GetPointer(page_addr + (addr & page_mask)); + const auto address{GpuToCpuAddress(gpu_addr)}; + if (!address) { + return {}; } - LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr); - return {}; + return system.Memory().GetPointer(*address); } -const u8* MemoryManager::GetPointer(GPUVAddr addr) const { - if (!IsAddressValid(addr)) { +const u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) const { + if (!GetPageEntry(gpu_addr).IsValid()) { return {}; } - const auto& memory = system.Memory(); - - const VAddr page_addr{page_table.backing_addr[addr >> page_bits]}; - - if (page_addr != 0) { - return memory.GetPointer(page_addr + (addr & page_mask)); + const auto address{GpuToCpuAddress(gpu_addr)}; + if (!address) { + return {}; } - LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr); - return {}; -} - -bool MemoryManager::IsBlockContinuous(const GPUVAddr start, const std::size_t size) const { - const std::size_t inner_size = size - 1; - const GPUVAddr end = start + inner_size; - const auto host_ptr_start = reinterpret_cast<std::uintptr_t>(GetPointer(start)); - const auto host_ptr_end = reinterpret_cast<std::uintptr_t>(GetPointer(end)); - const auto range = static_cast<std::size_t>(host_ptr_end - host_ptr_start); - return range == inner_size; + return system.Memory().GetPointer(*address); } -void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, - const std::size_t size) const { +void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size) const { std::size_t remaining_size{size}; std::size_t page_index{gpu_src_addr >> page_bits}; std::size_t page_offset{gpu_src_addr & page_mask}; - auto& memory = system.Memory(); - while (remaining_size > 0) { const std::size_t copy_amount{ std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)}; - const VAddr src_addr{page_table.backing_addr[page_index] + page_offset}; - // Flush must happen on the rasterizer interface, such that memory is always synchronous - // when it is read (even when in asynchronous GPU mode). Fixes Dead Cells title menu. - rasterizer.FlushRegion(src_addr, copy_amount); - memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); + if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) { + const auto src_addr{*page_addr + page_offset}; + + // Flush must happen on the rasterizer interface, such that memory is always synchronous + // when it is read (even when in asynchronous GPU mode). Fixes Dead Cells title menu. + rasterizer.FlushRegion(src_addr, copy_amount); + system.Memory().ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); + } page_index++; page_offset = 0; @@ -241,18 +234,17 @@ void MemoryManager::ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t page_index{gpu_src_addr >> page_bits}; std::size_t page_offset{gpu_src_addr & page_mask}; - auto& memory = system.Memory(); - while (remaining_size > 0) { const std::size_t copy_amount{ std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)}; - const u8* page_pointer = page_table.pointers[page_index]; - if (page_pointer) { - const VAddr src_addr{page_table.backing_addr[page_index] + page_offset}; - memory.ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); + + if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) { + const auto src_addr{*page_addr + page_offset}; + system.Memory().ReadBlockUnsafe(src_addr, dest_buffer, copy_amount); } else { std::memset(dest_buffer, 0, copy_amount); } + page_index++; page_offset = 0; dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount; @@ -260,23 +252,23 @@ void MemoryManager::ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer, } } -void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, - const std::size_t size) { +void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size) { std::size_t remaining_size{size}; std::size_t page_index{gpu_dest_addr >> page_bits}; std::size_t page_offset{gpu_dest_addr & page_mask}; - auto& memory = system.Memory(); - while (remaining_size > 0) { const std::size_t copy_amount{ std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)}; - const VAddr dest_addr{page_table.backing_addr[page_index] + page_offset}; - // Invalidate must happen on the rasterizer interface, such that memory is always - // synchronous when it is written (even when in asynchronous GPU mode). - rasterizer.InvalidateRegion(dest_addr, copy_amount); - memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); + if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) { + const auto dest_addr{*page_addr + page_offset}; + + // Invalidate must happen on the rasterizer interface, such that memory is always + // synchronous when it is written (even when in asynchronous GPU mode). + rasterizer.InvalidateRegion(dest_addr, copy_amount); + system.Memory().WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); + } page_index++; page_offset = 0; @@ -286,21 +278,20 @@ void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, } void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buffer, - const std::size_t size) { + std::size_t size) { std::size_t remaining_size{size}; std::size_t page_index{gpu_dest_addr >> page_bits}; std::size_t page_offset{gpu_dest_addr & page_mask}; - auto& memory = system.Memory(); - while (remaining_size > 0) { const std::size_t copy_amount{ std::min(static_cast<std::size_t>(page_size) - page_offset, remaining_size)}; - u8* page_pointer = page_table.pointers[page_index]; - if (page_pointer) { - const VAddr dest_addr{page_table.backing_addr[page_index] + page_offset}; - memory.WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); + + if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) { + const auto dest_addr{*page_addr + page_offset}; + system.Memory().WriteBlockUnsafe(dest_addr, src_buffer, copy_amount); } + page_index++; page_offset = 0; src_buffer = static_cast<const u8*>(src_buffer) + copy_amount; @@ -308,273 +299,26 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buf } } -void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, - const std::size_t size) { +void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size) { std::vector<u8> tmp_buffer(size); ReadBlock(gpu_src_addr, tmp_buffer.data(), size); WriteBlock(gpu_dest_addr, tmp_buffer.data(), size); } void MemoryManager::CopyBlockUnsafe(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, - const std::size_t size) { + std::size_t size) { std::vector<u8> tmp_buffer(size); ReadBlockUnsafe(gpu_src_addr, tmp_buffer.data(), size); WriteBlockUnsafe(gpu_dest_addr, tmp_buffer.data(), size); } bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) { - const VAddr addr = page_table.backing_addr[gpu_addr >> page_bits]; - const std::size_t page = (addr & Core::Memory::PAGE_MASK) + size; - return page <= Core::Memory::PAGE_SIZE; -} - -void MemoryManager::MapPages(GPUVAddr base, u64 size, u8* memory, Common::PageType type, - VAddr backing_addr) { - LOG_DEBUG(HW_GPU, "Mapping {} onto {:016X}-{:016X}", fmt::ptr(memory), base * page_size, - (base + size) * page_size); - - const VAddr end{base + size}; - ASSERT_MSG(end <= page_table.pointers.size(), "out of range mapping at {:016X}", - base + page_table.pointers.size()); - - if (memory == nullptr) { - while (base != end) { - page_table.pointers[base] = nullptr; - page_table.backing_addr[base] = 0; - - base += 1; - } - } else { - while (base != end) { - page_table.pointers[base] = memory; - page_table.backing_addr[base] = backing_addr; - - base += 1; - memory += page_size; - backing_addr += page_size; - } - } -} - -void MemoryManager::MapMemoryRegion(GPUVAddr base, u64 size, u8* target, VAddr backing_addr) { - ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: {:016X}", size); - ASSERT_MSG((base & page_mask) == 0, "non-page aligned base: {:016X}", base); - MapPages(base / page_size, size / page_size, target, Common::PageType::Memory, backing_addr); -} - -void MemoryManager::UnmapRegion(GPUVAddr base, u64 size) { - ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: {:016X}", size); - ASSERT_MSG((base & page_mask) == 0, "non-page aligned base: {:016X}", base); - MapPages(base / page_size, size / page_size, nullptr, Common::PageType::Unmapped); -} - -bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const { - ASSERT(base + size == next.base); - if (type != next.type) { - return {}; - } - if (type == VirtualMemoryArea::Type::Allocated && (offset + size != next.offset)) { - return {}; - } - if (type == VirtualMemoryArea::Type::Mapped && backing_memory + size != next.backing_memory) { - return {}; - } - return true; -} - -MemoryManager::VMAHandle MemoryManager::FindVMA(GPUVAddr target) const { - if (target >= address_space_end) { - return vma_map.end(); - } else { - return std::prev(vma_map.upper_bound(target)); - } -} - -MemoryManager::VMAIter MemoryManager::Allocate(VMAIter vma_handle) { - VirtualMemoryArea& vma{vma_handle->second}; - - vma.type = VirtualMemoryArea::Type::Allocated; - vma.backing_addr = 0; - vma.backing_memory = {}; - UpdatePageTableForVMA(vma); - - return MergeAdjacent(vma_handle); -} - -MemoryManager::VMAHandle MemoryManager::AllocateMemory(GPUVAddr target, std::size_t offset, - u64 size) { - - // This is the appropriately sized VMA that will turn into our allocation. - VMAIter vma_handle{CarveVMA(target, size)}; - VirtualMemoryArea& vma{vma_handle->second}; - - ASSERT(vma.size == size); - - vma.offset = offset; - - return Allocate(vma_handle); -} - -MemoryManager::VMAHandle MemoryManager::MapBackingMemory(GPUVAddr target, u8* memory, u64 size, - VAddr backing_addr) { - // This is the appropriately sized VMA that will turn into our allocation. - VMAIter vma_handle{CarveVMA(target, size)}; - VirtualMemoryArea& vma{vma_handle->second}; - - ASSERT(vma.size == size); - - vma.type = VirtualMemoryArea::Type::Mapped; - vma.backing_memory = memory; - vma.backing_addr = backing_addr; - UpdatePageTableForVMA(vma); - - return MergeAdjacent(vma_handle); -} - -void MemoryManager::UnmapRange(GPUVAddr target, u64 size) { - VMAIter vma{CarveVMARange(target, size)}; - const VAddr target_end{target + size}; - const VMAIter end{vma_map.end()}; - - // The comparison against the end of the range must be done using addresses since VMAs can be - // merged during this process, causing invalidation of the iterators. - while (vma != end && vma->second.base < target_end) { - // Unmapped ranges return to allocated state and can be reused - // This behavior is used by Super Mario Odyssey, Sonic Forces, and likely other games - vma = std::next(Allocate(vma)); - } - - ASSERT(FindVMA(target)->second.size >= size); -} - -MemoryManager::VMAIter MemoryManager::StripIterConstness(const VMAHandle& iter) { - // This uses a neat C++ trick to convert a const_iterator to a regular iterator, given - // non-const access to its container. - return vma_map.erase(iter, iter); // Erases an empty range of elements -} - -MemoryManager::VMAIter MemoryManager::CarveVMA(GPUVAddr base, u64 size) { - ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: 0x{:016X}", size); - ASSERT_MSG((base & page_mask) == 0, "non-page aligned base: 0x{:016X}", base); - - VMAIter vma_handle{StripIterConstness(FindVMA(base))}; - if (vma_handle == vma_map.end()) { - // Target address is outside the managed range - return {}; - } - - const VirtualMemoryArea& vma{vma_handle->second}; - if (vma.type == VirtualMemoryArea::Type::Mapped) { - // Region is already allocated - return vma_handle; - } - - const VAddr start_in_vma{base - vma.base}; - const VAddr end_in_vma{start_in_vma + size}; - - ASSERT_MSG(end_in_vma <= vma.size, "region size 0x{:016X} is less than required size 0x{:016X}", - vma.size, end_in_vma); - - if (end_in_vma < vma.size) { - // Split VMA at the end of the allocated region - SplitVMA(vma_handle, end_in_vma); - } - if (start_in_vma != 0) { - // Split VMA at the start of the allocated region - vma_handle = SplitVMA(vma_handle, start_in_vma); - } - - return vma_handle; -} - -MemoryManager::VMAIter MemoryManager::CarveVMARange(GPUVAddr target, u64 size) { - ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: 0x{:016X}", size); - ASSERT_MSG((target & page_mask) == 0, "non-page aligned base: 0x{:016X}", target); - - const VAddr target_end{target + size}; - ASSERT(target_end >= target); - ASSERT(size > 0); - - VMAIter begin_vma{StripIterConstness(FindVMA(target))}; - const VMAIter i_end{vma_map.lower_bound(target_end)}; - if (std::any_of(begin_vma, i_end, [](const auto& entry) { - return entry.second.type == VirtualMemoryArea::Type::Unmapped; - })) { + const auto cpu_addr{GpuToCpuAddress(gpu_addr)}; + if (!cpu_addr) { return {}; } - - if (target != begin_vma->second.base) { - begin_vma = SplitVMA(begin_vma, target - begin_vma->second.base); - } - - VMAIter end_vma{StripIterConstness(FindVMA(target_end))}; - if (end_vma != vma_map.end() && target_end != end_vma->second.base) { - end_vma = SplitVMA(end_vma, target_end - end_vma->second.base); - } - - return begin_vma; -} - -MemoryManager::VMAIter MemoryManager::SplitVMA(VMAIter vma_handle, u64 offset_in_vma) { - VirtualMemoryArea& old_vma{vma_handle->second}; - VirtualMemoryArea new_vma{old_vma}; // Make a copy of the VMA - - // For now, don't allow no-op VMA splits (trying to split at a boundary) because it's probably - // a bug. This restriction might be removed later. - ASSERT(offset_in_vma < old_vma.size); - ASSERT(offset_in_vma > 0); - - old_vma.size = offset_in_vma; - new_vma.base += offset_in_vma; - new_vma.size -= offset_in_vma; - - switch (new_vma.type) { - case VirtualMemoryArea::Type::Unmapped: - break; - case VirtualMemoryArea::Type::Allocated: - new_vma.offset += offset_in_vma; - break; - case VirtualMemoryArea::Type::Mapped: - new_vma.backing_memory += offset_in_vma; - break; - } - - ASSERT(old_vma.CanBeMergedWith(new_vma)); - - return vma_map.emplace_hint(std::next(vma_handle), new_vma.base, new_vma); -} - -MemoryManager::VMAIter MemoryManager::MergeAdjacent(VMAIter iter) { - const VMAIter next_vma{std::next(iter)}; - if (next_vma != vma_map.end() && iter->second.CanBeMergedWith(next_vma->second)) { - iter->second.size += next_vma->second.size; - vma_map.erase(next_vma); - } - - if (iter != vma_map.begin()) { - VMAIter prev_vma{std::prev(iter)}; - if (prev_vma->second.CanBeMergedWith(iter->second)) { - prev_vma->second.size += iter->second.size; - vma_map.erase(iter); - iter = prev_vma; - } - } - - return iter; -} - -void MemoryManager::UpdatePageTableForVMA(const VirtualMemoryArea& vma) { - switch (vma.type) { - case VirtualMemoryArea::Type::Unmapped: - UnmapRegion(vma.base, vma.size); - break; - case VirtualMemoryArea::Type::Allocated: - MapMemoryRegion(vma.base, vma.size, nullptr, vma.backing_addr); - break; - case VirtualMemoryArea::Type::Mapped: - MapMemoryRegion(vma.base, vma.size, vma.backing_memory, vma.backing_addr); - break; - } + const std::size_t page{(*cpu_addr & Core::Memory::PAGE_MASK) + size}; + return page <= Core::Memory::PAGE_SIZE; } } // namespace Tegra diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index 87658e87a..681bd9588 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -6,9 +6,9 @@ #include <map> #include <optional> +#include <vector> #include "common/common_types.h" -#include "common/page_table.h" namespace VideoCore { class RasterizerInterface; @@ -20,45 +20,57 @@ class System; namespace Tegra { -/** - * Represents a VMA in an address space. A VMA is a contiguous region of virtual addressing space - * with homogeneous attributes across its extents. In this particular implementation each VMA is - * also backed by a single host memory allocation. - */ -struct VirtualMemoryArea { - enum class Type : u8 { - Unmapped, - Allocated, - Mapped, +class PageEntry final { +public: + enum class State : u32 { + Unmapped = static_cast<u32>(-1), + Allocated = static_cast<u32>(-2), }; - /// Virtual base address of the region. - GPUVAddr base{}; - /// Size of the region. - u64 size{}; - /// Memory area mapping type. - Type type{Type::Unmapped}; - /// CPU memory mapped address corresponding to this memory area. - VAddr backing_addr{}; - /// Offset into the backing_memory the mapping starts from. - std::size_t offset{}; - /// Pointer backing this VMA. - u8* backing_memory{}; - - /// Tests if this area can be merged to the right with `next`. - bool CanBeMergedWith(const VirtualMemoryArea& next) const; + constexpr PageEntry() = default; + constexpr PageEntry(State state) : state{state} {} + constexpr PageEntry(VAddr addr) : state{static_cast<State>(addr >> ShiftBits)} {} + + constexpr bool IsUnmapped() const { + return state == State::Unmapped; + } + + constexpr bool IsAllocated() const { + return state == State::Allocated; + } + + constexpr bool IsValid() const { + return !IsUnmapped() && !IsAllocated(); + } + + constexpr VAddr ToAddress() const { + if (!IsValid()) { + return {}; + } + + return static_cast<VAddr>(state) << ShiftBits; + } + + constexpr PageEntry operator+(u64 offset) { + // If this is a reserved value, offsets do not apply + if (!IsValid()) { + return *this; + } + return PageEntry{(static_cast<VAddr>(state) << ShiftBits) + offset}; + } + +private: + static constexpr std::size_t ShiftBits{12}; + + State state{State::Unmapped}; }; +static_assert(sizeof(PageEntry) == 4, "PageEntry is too large"); class MemoryManager final { public: explicit MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer); ~MemoryManager(); - GPUVAddr AllocateSpace(u64 size, u64 align); - GPUVAddr AllocateSpace(GPUVAddr addr, u64 size, u64 align); - GPUVAddr MapBufferEx(VAddr cpu_addr, u64 size); - GPUVAddr MapBufferEx(VAddr cpu_addr, GPUVAddr addr, u64 size); - GPUVAddr UnmapBuffer(GPUVAddr addr, u64 size); std::optional<VAddr> GpuToCpuAddress(GPUVAddr addr) const; template <typename T> @@ -70,9 +82,6 @@ public: u8* GetPointer(GPUVAddr addr); const u8* GetPointer(GPUVAddr addr) const; - /// Returns true if the block is continuous in host memory, false otherwise - bool IsBlockContinuous(GPUVAddr start, std::size_t size) const; - /** * ReadBlock and WriteBlock are full read and write operations over virtual * GPU Memory. It's important to use these when GPU memory may not be continuous @@ -98,92 +107,43 @@ public: void CopyBlockUnsafe(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size); /** - * IsGranularRange checks if a gpu region can be simply read with a pointer + * IsGranularRange checks if a gpu region can be simply read with a pointer. */ bool IsGranularRange(GPUVAddr gpu_addr, std::size_t size); -private: - using VMAMap = std::map<GPUVAddr, VirtualMemoryArea>; - using VMAHandle = VMAMap::const_iterator; - using VMAIter = VMAMap::iterator; - - bool IsAddressValid(GPUVAddr addr) const; - void MapPages(GPUVAddr base, u64 size, u8* memory, Common::PageType type, - VAddr backing_addr = 0); - void MapMemoryRegion(GPUVAddr base, u64 size, u8* target, VAddr backing_addr); - void UnmapRegion(GPUVAddr base, u64 size); - - /// Finds the VMA in which the given address is included in, or `vma_map.end()`. - VMAHandle FindVMA(GPUVAddr target) const; - - VMAHandle AllocateMemory(GPUVAddr target, std::size_t offset, u64 size); - - /** - * Maps an unmanaged host memory pointer at a given address. - * - * @param target The guest address to start the mapping at. - * @param memory The memory to be mapped. - * @param size Size of the mapping in bytes. - * @param backing_addr The base address of the range to back this mapping. - */ - VMAHandle MapBackingMemory(GPUVAddr target, u8* memory, u64 size, VAddr backing_addr); - - /// Unmaps a range of addresses, splitting VMAs as necessary. - void UnmapRange(GPUVAddr target, u64 size); - - /// Converts a VMAHandle to a mutable VMAIter. - VMAIter StripIterConstness(const VMAHandle& iter); - - /// Marks as the specified VMA as allocated. - VMAIter Allocate(VMAIter vma); - - /** - * Carves a VMA of a specific size at the specified address by splitting Free VMAs while doing - * the appropriate error checking. - */ - VMAIter CarveVMA(GPUVAddr base, u64 size); - - /** - * Splits the edges of the given range of non-Free VMAs so that there is a VMA split at each - * end of the range. - */ - VMAIter CarveVMARange(GPUVAddr base, u64 size); - - /** - * Splits a VMA in two, at the specified offset. - * @returns the right side of the split, with the original iterator becoming the left side. - */ - VMAIter SplitVMA(VMAIter vma, u64 offset_in_vma); + GPUVAddr Map(VAddr cpu_addr, GPUVAddr gpu_addr, std::size_t size); + GPUVAddr MapAllocate(VAddr cpu_addr, std::size_t size, std::size_t align); + std::optional<GPUVAddr> AllocateFixed(GPUVAddr gpu_addr, std::size_t size); + GPUVAddr Allocate(std::size_t size, std::size_t align); + void Unmap(GPUVAddr gpu_addr, std::size_t size); - /** - * Checks for and merges the specified VMA with adjacent ones if possible. - * @returns the merged VMA or the original if no merging was possible. - */ - VMAIter MergeAdjacent(VMAIter vma); +private: + PageEntry GetPageEntry(GPUVAddr gpu_addr) const; + void SetPageEntry(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size = page_size); + GPUVAddr UpdateRange(GPUVAddr gpu_addr, PageEntry page_entry, std::size_t size); + std::optional<GPUVAddr> FindFreeRange(std::size_t size, std::size_t align) const; - /// Updates the pages corresponding to this VMA so they match the VMA's attributes. - void UpdatePageTableForVMA(const VirtualMemoryArea& vma); + void TryLockPage(PageEntry page_entry, std::size_t size); + void TryUnlockPage(PageEntry page_entry, std::size_t size); - /// Finds a free (unmapped region) of the specified size starting at the specified address. - GPUVAddr FindFreeRegion(GPUVAddr region_start, u64 size) const; + static constexpr std::size_t PageEntryIndex(GPUVAddr gpu_addr) { + return (gpu_addr >> page_bits) & page_table_mask; + } -private: + static constexpr u64 address_space_size = 1ULL << 40; + static constexpr u64 address_space_start = 1ULL << 32; static constexpr u64 page_bits{16}; static constexpr u64 page_size{1 << page_bits}; static constexpr u64 page_mask{page_size - 1}; + static constexpr u64 page_table_bits{24}; + static constexpr u64 page_table_size{1 << page_table_bits}; + static constexpr u64 page_table_mask{page_table_size - 1}; - /// Address space in bits, according to Tegra X1 TRM - static constexpr u32 address_space_width{40}; - /// Start address for mapping, this is fairly arbitrary but must be non-zero. - static constexpr GPUVAddr address_space_base{0x100000}; - /// End of address space, based on address space in bits. - static constexpr GPUVAddr address_space_end{1ULL << address_space_width}; + Core::System& system; - Common::PageTable page_table; - VMAMap vma_map; VideoCore::RasterizerInterface& rasterizer; - Core::System& system; + std::vector<PageEntry> page_table; }; } // namespace Tegra diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 03e82c599..cb284db77 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -178,16 +178,11 @@ RasterizerOpenGL::RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWind if (device.UseAsynchronousShaders()) { // Max worker threads we should allow - constexpr auto MAX_THREADS = 2u; - // Amount of threads we should reserve for other parts of yuzu - constexpr auto RESERVED_THREADS = 6u; - // Get the amount of threads we can use(this can return zero) - const auto cpu_thread_count = - std::max(RESERVED_THREADS, std::thread::hardware_concurrency()); - // Deduce how many "extra" threads we have to use. - const auto max_threads_unused = cpu_thread_count - RESERVED_THREADS; + constexpr u32 MAX_THREADS = 4; + // Deduce how many threads we can use + const u32 threads_used = std::thread::hardware_concurrency() / 4; // Always allow at least 1 thread regardless of our settings - const auto max_worker_count = std::max(1u, max_threads_unused); + const auto max_worker_count = std::max(1U, threads_used); // Don't use more than MAX_THREADS const auto worker_count = std::min(max_worker_count, MAX_THREADS); async_shaders.AllocateWorkers(worker_count); diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp index ce53e5a6b..a551e3de8 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp +++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp @@ -696,6 +696,7 @@ void VKBlitScreen::CreateFramebuffers() { .flags = 0, .renderPass = *renderpass, .attachmentCount = 1, + .pAttachments = nullptr, .width = size.width, .height = size.height, .layers = 1, diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp index 6245e0d78..0c03e4d83 100644 --- a/src/video_core/renderer_vulkan/vk_device.cpp +++ b/src/video_core/renderer_vulkan/vk_device.cpp @@ -771,8 +771,9 @@ std::vector<VkDeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const .pNext = nullptr, .flags = 0, .queueFamilyIndex = queue_family, + .queueCount = 1, + .pQueuePriorities = nullptr, }); - ci.queueCount = 1; ci.pQueuePriorities = &QUEUE_PRIORITY; } diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp index 42b3a744c..418c62bc4 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp @@ -261,8 +261,13 @@ VKComputePipeline& VKPipelineCache::GetComputePipeline(const ComputePipelineCach } const Specialization specialization{ + .base_binding = 0, .workgroup_size = key.workgroup_size, .shared_memory_size = key.shared_memory_size, + .point_size = std::nullopt, + .enabled_attributes = {}, + .attribute_types = {}, + .ndc_minus_one_to_one = false, }; const SPIRVShader spirv_shader{Decompile(device, shader->GetIR(), ShaderType::Compute, shader->GetRegistry(), specialization), diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index 2ed2004f0..7500e8244 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -815,8 +815,13 @@ bool RasterizerVulkan::WalkAttachmentOverlaps(const CachedSurfaceView& attachmen std::tuple<VkFramebuffer, VkExtent2D> RasterizerVulkan::ConfigureFramebuffers( VkRenderPass renderpass) { - FramebufferCacheKey key{renderpass, std::numeric_limits<u32>::max(), - std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max()}; + FramebufferCacheKey key{ + .renderpass = renderpass, + .width = std::numeric_limits<u32>::max(), + .height = std::numeric_limits<u32>::max(), + .layers = std::numeric_limits<u32>::max(), + .views = {}, + }; const auto try_push = [&key](const View& view) { if (!view) { diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp index 2d5460776..b068888f9 100644 --- a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp @@ -47,6 +47,7 @@ vk::Sampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) c VkSamplerCustomBorderColorCreateInfoEXT border{ .sType = VK_STRUCTURE_TYPE_SAMPLER_CUSTOM_BORDER_COLOR_CREATE_INFO_EXT, .pNext = nullptr, + .customBorderColor = {}, .format = VK_FORMAT_UNDEFINED, }; std::memcpy(&border.customBorderColor, color.data(), sizeof(color)); diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index efd4bb13b..2c6f54101 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -473,6 +473,8 @@ VkImageView CachedSurfaceView::GetAttachment() { .aspectMask = aspect_mask, .baseMipLevel = base_level, .levelCount = num_levels, + .baseArrayLayer = 0, + .layerCount = 0, }, }; if (image_view_type == VK_IMAGE_VIEW_TYPE_3D) { diff --git a/src/yuzu/bootmanager.cpp b/src/yuzu/bootmanager.cpp index 5738787ac..8fc322b30 100644 --- a/src/yuzu/bootmanager.cpp +++ b/src/yuzu/bootmanager.cpp @@ -567,7 +567,7 @@ void GRenderWindow::CaptureScreenshot(u32 res_scale, const QString& screenshot_p screenshot_image = QImage(QSize(layout.width, layout.height), QImage::Format_RGB32); renderer.RequestScreenshot( screenshot_image.bits(), - [=] { + [=, this] { const std::string std_screenshot_path = screenshot_path.toStdString(); if (screenshot_image.mirrored(false, true).save(screenshot_path)) { LOG_INFO(Frontend, "Screenshot saved to \"{}\"", std_screenshot_path); diff --git a/src/yuzu/configuration/configure_input_player.cpp b/src/yuzu/configuration/configure_input_player.cpp index 00433926d..b1850bc95 100644 --- a/src/yuzu/configuration/configure_input_player.cpp +++ b/src/yuzu/configuration/configure_input_player.cpp @@ -280,9 +280,9 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i } button->setContextMenuPolicy(Qt::CustomContextMenu); - connect(button, &QPushButton::clicked, [=] { + connect(button, &QPushButton::clicked, [=, this] { HandleClick(button_map[button_id], - [=](Common::ParamPackage params) { + [=, this](Common::ParamPackage params) { // Workaround for ZL & ZR for analog triggers like on XBOX controllors. // Analog triggers (from controllers like the XBOX controller) would not // work due to a different range of their signals (from 0 to 255 on @@ -300,19 +300,20 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i }, InputCommon::Polling::DeviceType::Button); }); - connect(button, &QPushButton::customContextMenuRequested, [=](const QPoint& menu_location) { - QMenu context_menu; - context_menu.addAction(tr("Clear"), [&] { - buttons_param[button_id].Clear(); - button_map[button_id]->setText(tr("[not set]")); - }); - context_menu.addAction(tr("Restore Default"), [&] { - buttons_param[button_id] = Common::ParamPackage{ - InputCommon::GenerateKeyboardParam(Config::default_buttons[button_id])}; - button_map[button_id]->setText(ButtonToText(buttons_param[button_id])); - }); - context_menu.exec(button_map[button_id]->mapToGlobal(menu_location)); - }); + connect(button, &QPushButton::customContextMenuRequested, + [=, this](const QPoint& menu_location) { + QMenu context_menu; + context_menu.addAction(tr("Clear"), [&] { + buttons_param[button_id].Clear(); + button_map[button_id]->setText(tr("[not set]")); + }); + context_menu.addAction(tr("Restore Default"), [&] { + buttons_param[button_id] = Common::ParamPackage{ + InputCommon::GenerateKeyboardParam(Config::default_buttons[button_id])}; + button_map[button_id]->setText(ButtonToText(buttons_param[button_id])); + }); + context_menu.exec(button_map[button_id]->mapToGlobal(menu_location)); + }); } for (int analog_id = 0; analog_id < Settings::NativeAnalog::NumAnalogs; analog_id++) { @@ -323,16 +324,16 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i } analog_button->setContextMenuPolicy(Qt::CustomContextMenu); - connect(analog_button, &QPushButton::clicked, [=]() { + connect(analog_button, &QPushButton::clicked, [=, this] { HandleClick(analog_map_buttons[analog_id][sub_button_id], - [=](const Common::ParamPackage& params) { + [=, this](const Common::ParamPackage& params) { SetAnalogButton(params, analogs_param[analog_id], analog_sub_buttons[sub_button_id]); }, InputCommon::Polling::DeviceType::Button); }); connect(analog_button, &QPushButton::customContextMenuRequested, - [=](const QPoint& menu_location) { + [=, this](const QPoint& menu_location) { QMenu context_menu; context_menu.addAction(tr("Clear"), [&] { analogs_param[analog_id].Erase(analog_sub_buttons[sub_button_id]); @@ -350,32 +351,35 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i menu_location)); }); } - connect(analog_map_stick[analog_id], &QPushButton::clicked, [=] { + connect(analog_map_stick[analog_id], &QPushButton::clicked, [=, this] { if (QMessageBox::information( this, tr("Information"), tr("After pressing OK, first move your joystick horizontally, " "and then vertically."), QMessageBox::Ok | QMessageBox::Cancel) == QMessageBox::Ok) { - HandleClick( - analog_map_stick[analog_id], - [=](const Common::ParamPackage& params) { analogs_param[analog_id] = params; }, - InputCommon::Polling::DeviceType::Analog); + HandleClick(analog_map_stick[analog_id], + [=, this](const Common::ParamPackage& params) { + analogs_param[analog_id] = params; + }, + InputCommon::Polling::DeviceType::Analog); } }); - connect(analog_map_deadzone_and_modifier_slider[analog_id], &QSlider::valueChanged, [=] { - const float slider_value = analog_map_deadzone_and_modifier_slider[analog_id]->value(); - if (analogs_param[analog_id].Get("engine", "") == "sdl" || - analogs_param[analog_id].Get("engine", "") == "gcpad") { - analog_map_deadzone_and_modifier_slider_label[analog_id]->setText( - tr("Deadzone: %1%").arg(slider_value)); - analogs_param[analog_id].Set("deadzone", slider_value / 100.0f); - } else { - analog_map_deadzone_and_modifier_slider_label[analog_id]->setText( - tr("Modifier Scale: %1%").arg(slider_value)); - analogs_param[analog_id].Set("modifier_scale", slider_value / 100.0f); - } - }); + connect(analog_map_deadzone_and_modifier_slider[analog_id], &QSlider::valueChanged, + [=, this] { + const float slider_value = + analog_map_deadzone_and_modifier_slider[analog_id]->value(); + if (analogs_param[analog_id].Get("engine", "") == "sdl" || + analogs_param[analog_id].Get("engine", "") == "gcpad") { + analog_map_deadzone_and_modifier_slider_label[analog_id]->setText( + tr("Deadzone: %1%").arg(slider_value)); + analogs_param[analog_id].Set("deadzone", slider_value / 100.0f); + } else { + analog_map_deadzone_and_modifier_slider_label[analog_id]->setText( + tr("Modifier Scale: %1%").arg(slider_value)); + analogs_param[analog_id].Set("modifier_scale", slider_value / 100.0f); + } + }); } connect(ui->buttonClearAll, &QPushButton::clicked, [this] { ClearAll(); }); diff --git a/src/yuzu/configuration/configure_mouse_advanced.cpp b/src/yuzu/configuration/configure_mouse_advanced.cpp index e0647ea5b..ea2549363 100644 --- a/src/yuzu/configuration/configure_mouse_advanced.cpp +++ b/src/yuzu/configuration/configure_mouse_advanced.cpp @@ -83,25 +83,28 @@ ConfigureMouseAdvanced::ConfigureMouseAdvanced(QWidget* parent) } button->setContextMenuPolicy(Qt::CustomContextMenu); - connect(button, &QPushButton::clicked, [=] { - HandleClick( - button_map[button_id], - [=](const Common::ParamPackage& params) { buttons_param[button_id] = params; }, - InputCommon::Polling::DeviceType::Button); - }); - connect(button, &QPushButton::customContextMenuRequested, [=](const QPoint& menu_location) { - QMenu context_menu; - context_menu.addAction(tr("Clear"), [&] { - buttons_param[button_id].Clear(); - button_map[button_id]->setText(tr("[not set]")); - }); - context_menu.addAction(tr("Restore Default"), [&] { - buttons_param[button_id] = Common::ParamPackage{ - InputCommon::GenerateKeyboardParam(Config::default_mouse_buttons[button_id])}; - button_map[button_id]->setText(ButtonToText(buttons_param[button_id])); - }); - context_menu.exec(button_map[button_id]->mapToGlobal(menu_location)); + connect(button, &QPushButton::clicked, [=, this] { + HandleClick(button_map[button_id], + [=, this](const Common::ParamPackage& params) { + buttons_param[button_id] = params; + }, + InputCommon::Polling::DeviceType::Button); }); + connect(button, &QPushButton::customContextMenuRequested, + [=, this](const QPoint& menu_location) { + QMenu context_menu; + context_menu.addAction(tr("Clear"), [&] { + buttons_param[button_id].Clear(); + button_map[button_id]->setText(tr("[not set]")); + }); + context_menu.addAction(tr("Restore Default"), [&] { + buttons_param[button_id] = + Common::ParamPackage{InputCommon::GenerateKeyboardParam( + Config::default_mouse_buttons[button_id])}; + button_map[button_id]->setText(ButtonToText(buttons_param[button_id])); + }); + context_menu.exec(button_map[button_id]->mapToGlobal(menu_location)); + }); } connect(ui->buttonClearAll, &QPushButton::clicked, [this] { ClearAll(); }); diff --git a/src/yuzu/configuration/configure_ui.cpp b/src/yuzu/configuration/configure_ui.cpp index 91c21c572..2c20b68d0 100644 --- a/src/yuzu/configuration/configure_ui.cpp +++ b/src/yuzu/configuration/configure_ui.cpp @@ -54,9 +54,9 @@ ConfigureUi::ConfigureUi(QWidget* parent) : QWidget(parent), ui(new Ui::Configur // Update text ComboBoxes after user interaction. connect(ui->row_1_text_combobox, QOverload<int>::of(&QComboBox::activated), - [=]() { ConfigureUi::UpdateSecondRowComboBox(); }); + [this] { ConfigureUi::UpdateSecondRowComboBox(); }); connect(ui->row_2_text_combobox, QOverload<int>::of(&QComboBox::activated), - [=]() { ConfigureUi::UpdateFirstRowComboBox(); }); + [this] { ConfigureUi::UpdateFirstRowComboBox(); }); // Set screenshot path to user specification. connect(ui->screenshot_path_button, &QToolButton::pressed, this, [this] { diff --git a/src/yuzu/game_list_worker.cpp b/src/yuzu/game_list_worker.cpp index 2018150db..239016b94 100644 --- a/src/yuzu/game_list_worker.cpp +++ b/src/yuzu/game_list_worker.cpp @@ -369,8 +369,8 @@ void GameListWorker::run() { auto* const game_list_dir = new GameListDir(game_dir); emit DirEntryReady(game_list_dir); provider->ClearAllEntries(); - ScanFileSystem(ScanTarget::FillManualContentProvider, game_dir.path.toStdString(), 2, - game_list_dir); + ScanFileSystem(ScanTarget::FillManualContentProvider, game_dir.path.toStdString(), + game_dir.deep_scan ? 256 : 0, game_list_dir); ScanFileSystem(ScanTarget::PopulateGameList, game_dir.path.toStdString(), game_dir.deep_scan ? 256 : 0, game_list_dir); } diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp index e26cec78c..592993c36 100644 --- a/src/yuzu/main.cpp +++ b/src/yuzu/main.cpp @@ -583,7 +583,7 @@ void GMainWindow::InitializeWidgets() { renderer_status_button->setObjectName(QStringLiteral("RendererStatusBarButton")); renderer_status_button->setCheckable(true); renderer_status_button->setFocusPolicy(Qt::NoFocus); - connect(renderer_status_button, &QPushButton::toggled, [=](bool checked) { + connect(renderer_status_button, &QPushButton::toggled, [this](bool checked) { renderer_status_button->setText(checked ? tr("VULKAN") : tr("OPENGL")); }); renderer_status_button->toggle(); @@ -595,7 +595,7 @@ void GMainWindow::InitializeWidgets() { #else renderer_status_button->setChecked(Settings::values.renderer_backend.GetValue() == Settings::RendererBackend::Vulkan); - connect(renderer_status_button, &QPushButton::clicked, [=] { + connect(renderer_status_button, &QPushButton::clicked, [this] { if (emulation_running) { return; } |