diff options
Diffstat (limited to '')
29 files changed, 1491 insertions, 164 deletions
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt index 8a66aa8ea..c3b688c5d 100644 --- a/src/core/CMakeLists.txt +++ b/src/core/CMakeLists.txt @@ -285,6 +285,7 @@ add_library(core STATIC hle/kernel/kernel.cpp hle/kernel/kernel.h hle/kernel/memory_types.h + hle/kernel/message_buffer.h hle/kernel/physical_core.cpp hle/kernel/physical_core.h hle/kernel/physical_memory.h diff --git a/src/core/arm/arm_interface.cpp b/src/core/arm/arm_interface.cpp index beaea64b3..aa0eb9791 100644 --- a/src/core/arm/arm_interface.cpp +++ b/src/core/arm/arm_interface.cpp @@ -185,7 +185,7 @@ void ARM_Interface::Run() { // Notify the debugger and go to sleep if a breakpoint was hit, // or if the thread is unable to continue for any reason. if (True(hr & HaltReason::InstructionBreakpoint) || True(hr & HaltReason::PrefetchAbort)) { - if (!True(hr & HaltReason::InstructionBreakpoint)) { + if (!True(hr & HaltReason::PrefetchAbort)) { RewindBreakpointInstruction(); } if (system.DebuggerEnabled()) { diff --git a/src/core/core.cpp b/src/core/core.cpp index b74fd0a58..9e3eb3795 100644 --- a/src/core/core.cpp +++ b/src/core/core.cpp @@ -27,6 +27,7 @@ #include "core/file_sys/savedata_factory.h" #include "core/file_sys/vfs_concat.h" #include "core/file_sys/vfs_real.h" +#include "core/gpu_dirty_memory_manager.h" #include "core/hid/hid_core.h" #include "core/hle/kernel/k_memory_manager.h" #include "core/hle/kernel/k_process.h" @@ -130,7 +131,10 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs, struct System::Impl { explicit Impl(System& system) : kernel{system}, fs_controller{system}, memory{system}, hid_core{}, room_network{}, - cpu_manager{system}, reporter{system}, applet_manager{system}, time_manager{system} {} + cpu_manager{system}, reporter{system}, applet_manager{system}, time_manager{system}, + gpu_dirty_memory_write_manager{} { + memory.SetGPUDirtyManagers(gpu_dirty_memory_write_manager); + } void Initialize(System& system) { device_memory = std::make_unique<Core::DeviceMemory>(); @@ -234,6 +238,8 @@ struct System::Impl { // Setting changes may require a full system reinitialization (e.g., disabling multicore). ReinitializeIfNecessary(system); + memory.SetGPUDirtyManagers(gpu_dirty_memory_write_manager); + kernel.Initialize(); cpu_manager.Initialize(); @@ -540,6 +546,9 @@ struct System::Impl { std::array<u64, Core::Hardware::NUM_CPU_CORES> dynarmic_ticks{}; std::array<MicroProfileToken, Core::Hardware::NUM_CPU_CORES> microprofile_cpu{}; + + std::array<Core::GPUDirtyMemoryManager, Core::Hardware::NUM_CPU_CORES> + gpu_dirty_memory_write_manager{}; }; System::System() : impl{std::make_unique<Impl>(*this)} {} @@ -629,10 +638,31 @@ void System::PrepareReschedule(const u32 core_index) { impl->kernel.PrepareReschedule(core_index); } +Core::GPUDirtyMemoryManager& System::CurrentGPUDirtyMemoryManager() { + const std::size_t core = impl->kernel.GetCurrentHostThreadID(); + return impl->gpu_dirty_memory_write_manager[core < Core::Hardware::NUM_CPU_CORES + ? core + : Core::Hardware::NUM_CPU_CORES - 1]; +} + +/// Provides a constant reference to the current gou dirty memory manager. +const Core::GPUDirtyMemoryManager& System::CurrentGPUDirtyMemoryManager() const { + const std::size_t core = impl->kernel.GetCurrentHostThreadID(); + return impl->gpu_dirty_memory_write_manager[core < Core::Hardware::NUM_CPU_CORES + ? core + : Core::Hardware::NUM_CPU_CORES - 1]; +} + size_t System::GetCurrentHostThreadID() const { return impl->kernel.GetCurrentHostThreadID(); } +void System::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) { + for (auto& manager : impl->gpu_dirty_memory_write_manager) { + manager.Gather(callback); + } +} + PerfStatsResults System::GetAndResetPerfStats() { return impl->GetAndResetPerfStats(); } diff --git a/src/core/core.h b/src/core/core.h index 93afc9303..14b2f7785 100644 --- a/src/core/core.h +++ b/src/core/core.h @@ -108,9 +108,10 @@ class CpuManager; class Debugger; class DeviceMemory; class ExclusiveMonitor; -class SpeedLimiter; +class GPUDirtyMemoryManager; class PerfStats; class Reporter; +class SpeedLimiter; class TelemetrySession; struct PerfStatsResults; @@ -225,6 +226,14 @@ public: /// Prepare the core emulation for a reschedule void PrepareReschedule(u32 core_index); + /// Provides a reference to the gou dirty memory manager. + [[nodiscard]] Core::GPUDirtyMemoryManager& CurrentGPUDirtyMemoryManager(); + + /// Provides a constant reference to the current gou dirty memory manager. + [[nodiscard]] const Core::GPUDirtyMemoryManager& CurrentGPUDirtyMemoryManager() const; + + void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback); + [[nodiscard]] size_t GetCurrentHostThreadID() const; /// Gets and resets core performance statistics diff --git a/src/core/core_timing.cpp b/src/core/core_timing.cpp index 4f0a3f8ea..b98a0cb33 100644 --- a/src/core/core_timing.cpp +++ b/src/core/core_timing.cpp @@ -70,7 +70,7 @@ void CoreTiming::Initialize(std::function<void()>&& on_thread_init_) { -> std::optional<std::chrono::nanoseconds> { return std::nullopt; }; ev_lost = CreateEvent("_lost_event", empty_timed_callback); if (is_multicore) { - timer_thread = std::make_unique<std::thread>(ThreadEntry, std::ref(*this)); + timer_thread = std::make_unique<std::jthread>(ThreadEntry, std::ref(*this)); } } @@ -253,12 +253,8 @@ void CoreTiming::ThreadLoop() { auto wait_time = *next_time - GetGlobalTimeNs().count(); if (wait_time > 0) { #ifdef _WIN32 - const auto timer_resolution_ns = - Common::Windows::GetCurrentTimerResolution().count(); - while (!paused && !event.IsSet() && wait_time > 0) { wait_time = *next_time - GetGlobalTimeNs().count(); - if (wait_time >= timer_resolution_ns) { Common::Windows::SleepForOneTick(); } else { @@ -316,4 +312,10 @@ std::chrono::microseconds CoreTiming::GetGlobalTimeUs() const { return std::chrono::microseconds{Common::WallClock::CPUTickToUS(cpu_ticks)}; } +#ifdef _WIN32 +void CoreTiming::SetTimerResolutionNs(std::chrono::nanoseconds ns) { + timer_resolution_ns = ns.count(); +} +#endif + } // namespace Core::Timing diff --git a/src/core/core_timing.h b/src/core/core_timing.h index 10db1de55..c20e906fb 100644 --- a/src/core/core_timing.h +++ b/src/core/core_timing.h @@ -131,6 +131,10 @@ public: /// Checks for events manually and returns time in nanoseconds for next event, threadsafe. std::optional<s64> Advance(); +#ifdef _WIN32 + void SetTimerResolutionNs(std::chrono::nanoseconds ns); +#endif + private: struct Event; @@ -143,6 +147,10 @@ private: s64 global_timer = 0; +#ifdef _WIN32 + s64 timer_resolution_ns; +#endif + // The queue is a min-heap using std::make_heap/push_heap/pop_heap. // We don't use std::priority_queue because we need to be able to serialize, unserialize and // erase arbitrary events (RemoveEvent()) regardless of the queue order. These aren't @@ -155,7 +163,7 @@ private: Common::Event pause_event{}; std::mutex basic_lock; std::mutex advance_lock; - std::unique_ptr<std::thread> timer_thread; + std::unique_ptr<std::jthread> timer_thread; std::atomic<bool> paused{}; std::atomic<bool> paused_set{}; std::atomic<bool> wait_set{}; diff --git a/src/core/file_sys/content_archive.cpp b/src/core/file_sys/content_archive.cpp index 50303fe42..06efab46d 100644 --- a/src/core/file_sys/content_archive.cpp +++ b/src/core/file_sys/content_archive.cpp @@ -57,11 +57,34 @@ struct NCASectionHeaderBlock { }; static_assert(sizeof(NCASectionHeaderBlock) == 0x8, "NCASectionHeaderBlock has incorrect size."); +struct NCABucketInfo { + u64 table_offset; + u64 table_size; + std::array<u8, 0x10> table_header; +}; +static_assert(sizeof(NCABucketInfo) == 0x20, "NCABucketInfo has incorrect size."); + +struct NCASparseInfo { + NCABucketInfo bucket; + u64 physical_offset; + u16 generation; + INSERT_PADDING_BYTES_NOINIT(0x6); +}; +static_assert(sizeof(NCASparseInfo) == 0x30, "NCASparseInfo has incorrect size."); + +struct NCACompressionInfo { + NCABucketInfo bucket; + INSERT_PADDING_BYTES_NOINIT(0x8); +}; +static_assert(sizeof(NCACompressionInfo) == 0x28, "NCACompressionInfo has incorrect size."); + struct NCASectionRaw { NCASectionHeaderBlock header; std::array<u8, 0x138> block_data; std::array<u8, 0x8> section_ctr; - INSERT_PADDING_BYTES_NOINIT(0xB8); + NCASparseInfo sparse_info; + NCACompressionInfo compression_info; + INSERT_PADDING_BYTES_NOINIT(0x60); }; static_assert(sizeof(NCASectionRaw) == 0x200, "NCASectionRaw has incorrect size."); @@ -225,6 +248,20 @@ bool NCA::ReadSections(const std::vector<NCASectionHeader>& sections, u64 bktr_b for (std::size_t i = 0; i < sections.size(); ++i) { const auto& section = sections[i]; + if (section.raw.sparse_info.bucket.table_offset != 0 && + section.raw.sparse_info.bucket.table_size != 0) { + LOG_ERROR(Loader, "Sparse NCAs are not supported."); + status = Loader::ResultStatus::ErrorSparseNCA; + return false; + } + + if (section.raw.compression_info.bucket.table_offset != 0 && + section.raw.compression_info.bucket.table_size != 0) { + LOG_ERROR(Loader, "Compressed NCAs are not supported."); + status = Loader::ResultStatus::ErrorCompressedNCA; + return false; + } + if (section.raw.header.filesystem_type == NCASectionFilesystemType::ROMFS) { if (!ReadRomFSSection(section, header.section_tables[i], bktr_base_ivfc_offset)) { return false; diff --git a/src/core/file_sys/fsmitm_romfsbuild.cpp b/src/core/file_sys/fsmitm_romfsbuild.cpp index 1ff83c08c..e39c7b62b 100644 --- a/src/core/file_sys/fsmitm_romfsbuild.cpp +++ b/src/core/file_sys/fsmitm_romfsbuild.cpp @@ -105,19 +105,11 @@ static u64 romfs_get_hash_table_count(u64 num_entries) { return count; } -void RomFSBuildContext::VisitDirectory(VirtualDir root_romfs, VirtualDir ext_dir, +void RomFSBuildContext::VisitDirectory(VirtualDir romfs_dir, VirtualDir ext_dir, std::shared_ptr<RomFSBuildDirectoryContext> parent) { std::vector<std::shared_ptr<RomFSBuildDirectoryContext>> child_dirs; - VirtualDir dir; - - if (parent->path_len == 0) { - dir = root_romfs; - } else { - dir = root_romfs->GetDirectoryRelative(parent->path); - } - - const auto entries = dir->GetEntries(); + const auto entries = romfs_dir->GetEntries(); for (const auto& kv : entries) { if (kv.second == VfsEntryType::Directory) { @@ -127,7 +119,7 @@ void RomFSBuildContext::VisitDirectory(VirtualDir root_romfs, VirtualDir ext_dir child->path_len = child->cur_path_ofs + static_cast<u32>(kv.first.size()); child->path = parent->path + "/" + kv.first; - if (ext_dir != nullptr && ext_dir->GetFileRelative(child->path + ".stub") != nullptr) { + if (ext_dir != nullptr && ext_dir->GetFile(kv.first + ".stub") != nullptr) { continue; } @@ -144,17 +136,17 @@ void RomFSBuildContext::VisitDirectory(VirtualDir root_romfs, VirtualDir ext_dir child->path_len = child->cur_path_ofs + static_cast<u32>(kv.first.size()); child->path = parent->path + "/" + kv.first; - if (ext_dir != nullptr && ext_dir->GetFileRelative(child->path + ".stub") != nullptr) { + if (ext_dir != nullptr && ext_dir->GetFile(kv.first + ".stub") != nullptr) { continue; } // Sanity check on path_len ASSERT(child->path_len < FS_MAX_PATH); - child->source = root_romfs->GetFileRelative(child->path); + child->source = romfs_dir->GetFile(kv.first); if (ext_dir != nullptr) { - if (const auto ips = ext_dir->GetFileRelative(child->path + ".ips")) { + if (const auto ips = ext_dir->GetFile(kv.first + ".ips")) { if (auto patched = PatchIPS(child->source, ips)) { child->source = std::move(patched); } @@ -168,23 +160,27 @@ void RomFSBuildContext::VisitDirectory(VirtualDir root_romfs, VirtualDir ext_dir } for (auto& child : child_dirs) { - this->VisitDirectory(root_romfs, ext_dir, child); + auto subdir_name = std::string_view(child->path).substr(child->cur_path_ofs); + auto child_romfs_dir = romfs_dir->GetSubdirectory(subdir_name); + auto child_ext_dir = ext_dir != nullptr ? ext_dir->GetSubdirectory(subdir_name) : nullptr; + this->VisitDirectory(child_romfs_dir, child_ext_dir, child); } } bool RomFSBuildContext::AddDirectory(std::shared_ptr<RomFSBuildDirectoryContext> parent_dir_ctx, std::shared_ptr<RomFSBuildDirectoryContext> dir_ctx) { // Check whether it's already in the known directories. - const auto existing = directories.find(dir_ctx->path); - if (existing != directories.end()) + const auto [it, is_new] = directories.emplace(dir_ctx->path, nullptr); + if (!is_new) { return false; + } // Add a new directory. num_dirs++; dir_table_size += sizeof(RomFSDirectoryEntry) + Common::AlignUp(dir_ctx->path_len - dir_ctx->cur_path_ofs, 4); dir_ctx->parent = parent_dir_ctx; - directories.emplace(dir_ctx->path, dir_ctx); + it->second = dir_ctx; return true; } @@ -192,8 +188,8 @@ bool RomFSBuildContext::AddDirectory(std::shared_ptr<RomFSBuildDirectoryContext> bool RomFSBuildContext::AddFile(std::shared_ptr<RomFSBuildDirectoryContext> parent_dir_ctx, std::shared_ptr<RomFSBuildFileContext> file_ctx) { // Check whether it's already in the known files. - const auto existing = files.find(file_ctx->path); - if (existing != files.end()) { + const auto [it, is_new] = files.emplace(file_ctx->path, nullptr); + if (!is_new) { return false; } @@ -202,7 +198,7 @@ bool RomFSBuildContext::AddFile(std::shared_ptr<RomFSBuildDirectoryContext> pare file_table_size += sizeof(RomFSFileEntry) + Common::AlignUp(file_ctx->path_len - file_ctx->cur_path_ofs, 4); file_ctx->parent = parent_dir_ctx; - files.emplace(file_ctx->path, file_ctx); + it->second = file_ctx; return true; } diff --git a/src/core/file_sys/vfs_real.cpp b/src/core/file_sys/vfs_real.cpp index b0515ec05..1c706e4d8 100644 --- a/src/core/file_sys/vfs_real.cpp +++ b/src/core/file_sys/vfs_real.cpp @@ -283,7 +283,8 @@ std::size_t RealVfsFile::GetSize() const { if (size) { return *size; } - return FS::GetSize(path); + auto lk = base.RefreshReference(path, perms, *reference); + return reference->file ? reference->file->GetSize() : 0; } bool RealVfsFile::Resize(std::size_t new_size) { diff --git a/src/core/gpu_dirty_memory_manager.h b/src/core/gpu_dirty_memory_manager.h new file mode 100644 index 000000000..9687531e8 --- /dev/null +++ b/src/core/gpu_dirty_memory_manager.h @@ -0,0 +1,122 @@ +// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-3.0-or-later + +#pragma once + +#include <atomic> +#include <bit> +#include <functional> +#include <mutex> +#include <utility> +#include <vector> + +#include "core/memory.h" + +namespace Core { + +class GPUDirtyMemoryManager { +public: + GPUDirtyMemoryManager() : current{default_transform} { + back_buffer.reserve(256); + front_buffer.reserve(256); + } + + ~GPUDirtyMemoryManager() = default; + + void Collect(VAddr address, size_t size) { + TransformAddress t = BuildTransform(address, size); + TransformAddress tmp, original; + do { + tmp = current.load(std::memory_order_acquire); + original = tmp; + if (tmp.address != t.address) { + if (IsValid(tmp.address)) { + std::scoped_lock lk(guard); + back_buffer.emplace_back(tmp); + current.exchange(t, std::memory_order_relaxed); + return; + } + tmp.address = t.address; + tmp.mask = 0; + } + if ((tmp.mask | t.mask) == tmp.mask) { + return; + } + tmp.mask |= t.mask; + } while (!current.compare_exchange_weak(original, tmp, std::memory_order_release, + std::memory_order_relaxed)); + } + + void Gather(std::function<void(VAddr, size_t)>& callback) { + { + std::scoped_lock lk(guard); + TransformAddress t = current.exchange(default_transform, std::memory_order_relaxed); + front_buffer.swap(back_buffer); + if (IsValid(t.address)) { + front_buffer.emplace_back(t); + } + } + for (auto& transform : front_buffer) { + size_t offset = 0; + u64 mask = transform.mask; + while (mask != 0) { + const size_t empty_bits = std::countr_zero(mask); + offset += empty_bits << align_bits; + mask = mask >> empty_bits; + + const size_t continuous_bits = std::countr_one(mask); + callback((static_cast<VAddr>(transform.address) << page_bits) + offset, + continuous_bits << align_bits); + mask = continuous_bits < align_size ? (mask >> continuous_bits) : 0; + offset += continuous_bits << align_bits; + } + } + front_buffer.clear(); + } + +private: + struct alignas(8) TransformAddress { + u32 address; + u32 mask; + }; + + constexpr static size_t page_bits = Memory::YUZU_PAGEBITS - 1; + constexpr static size_t page_size = 1ULL << page_bits; + constexpr static size_t page_mask = page_size - 1; + + constexpr static size_t align_bits = 6U; + constexpr static size_t align_size = 1U << align_bits; + constexpr static size_t align_mask = align_size - 1; + constexpr static TransformAddress default_transform = {.address = ~0U, .mask = 0U}; + + bool IsValid(VAddr address) { + return address < (1ULL << 39); + } + + template <typename T> + T CreateMask(size_t top_bit, size_t minor_bit) { + T mask = ~T(0); + mask <<= (sizeof(T) * 8 - top_bit); + mask >>= (sizeof(T) * 8 - top_bit); + mask >>= minor_bit; + mask <<= minor_bit; + return mask; + } + + TransformAddress BuildTransform(VAddr address, size_t size) { + const size_t minor_address = address & page_mask; + const size_t minor_bit = minor_address >> align_bits; + const size_t top_bit = (minor_address + size + align_mask) >> align_bits; + TransformAddress result{}; + result.address = static_cast<u32>(address >> page_bits); + result.mask = CreateMask<u32>(top_bit, minor_bit); + return result; + } + + std::atomic<TransformAddress> current{}; + std::mutex guard; + std::vector<TransformAddress> back_buffer; + std::vector<TransformAddress> front_buffer; +}; + +} // namespace Core diff --git a/src/core/hid/emulated_controller.cpp b/src/core/hid/emulated_controller.cpp index 1ebc32c1e..94bd656fe 100644 --- a/src/core/hid/emulated_controller.cpp +++ b/src/core/hid/emulated_controller.cpp @@ -1243,10 +1243,12 @@ Common::Input::DriverResult EmulatedController::SetPollingMode( auto& nfc_output_device = output_devices[3]; if (device_index == EmulatedDeviceIndex::LeftIndex) { + controller.left_polling_mode = polling_mode; return left_output_device->SetPollingMode(polling_mode); } if (device_index == EmulatedDeviceIndex::RightIndex) { + controller.right_polling_mode = polling_mode; const auto virtual_nfc_result = nfc_output_device->SetPollingMode(polling_mode); const auto mapped_nfc_result = right_output_device->SetPollingMode(polling_mode); @@ -1261,12 +1263,22 @@ Common::Input::DriverResult EmulatedController::SetPollingMode( return mapped_nfc_result; } + controller.left_polling_mode = polling_mode; + controller.right_polling_mode = polling_mode; left_output_device->SetPollingMode(polling_mode); right_output_device->SetPollingMode(polling_mode); nfc_output_device->SetPollingMode(polling_mode); return Common::Input::DriverResult::Success; } +Common::Input::PollingMode EmulatedController::GetPollingMode( + EmulatedDeviceIndex device_index) const { + if (device_index == EmulatedDeviceIndex::LeftIndex) { + return controller.left_polling_mode; + } + return controller.right_polling_mode; +} + bool EmulatedController::SetCameraFormat( Core::IrSensor::ImageTransferProcessorFormat camera_format) { LOG_INFO(Service_HID, "Set camera format {}", camera_format); diff --git a/src/core/hid/emulated_controller.h b/src/core/hid/emulated_controller.h index d511e5fac..88d77db8d 100644 --- a/src/core/hid/emulated_controller.h +++ b/src/core/hid/emulated_controller.h @@ -143,6 +143,8 @@ struct ControllerStatus { CameraState camera_state{}; RingSensorForce ring_analog_state{}; NfcState nfc_state{}; + Common::Input::PollingMode left_polling_mode{}; + Common::Input::PollingMode right_polling_mode{}; }; enum class ControllerTriggerType { @@ -370,6 +372,12 @@ public: */ Common::Input::DriverResult SetPollingMode(EmulatedDeviceIndex device_index, Common::Input::PollingMode polling_mode); + /** + * Get the current polling mode from a controller + * @param device_index index of the controller to set the polling mode + * @return current polling mode + */ + Common::Input::PollingMode GetPollingMode(EmulatedDeviceIndex device_index) const; /** * Sets the desired camera format to be polled from a controller diff --git a/src/core/hle/kernel/k_server_session.cpp b/src/core/hle/kernel/k_server_session.cpp index c66aff501..c64ceb530 100644 --- a/src/core/hle/kernel/k_server_session.cpp +++ b/src/core/hle/kernel/k_server_session.cpp @@ -20,12 +20,132 @@ #include "core/hle/kernel/k_thread.h" #include "core/hle/kernel/k_thread_queue.h" #include "core/hle/kernel/kernel.h" +#include "core/hle/kernel/message_buffer.h" #include "core/hle/service/hle_ipc.h" #include "core/hle/service/ipc_helpers.h" #include "core/memory.h" namespace Kernel { +namespace { + +template <bool MoveHandleAllowed> +Result ProcessMessageSpecialData(KProcess& dst_process, KProcess& src_process, KThread& src_thread, + MessageBuffer& dst_msg, const MessageBuffer& src_msg, + MessageBuffer::SpecialHeader& src_special_header) { + // Copy the special header to the destination. + s32 offset = dst_msg.Set(src_special_header); + + // Copy the process ID. + if (src_special_header.GetHasProcessId()) { + offset = dst_msg.SetProcessId(offset, src_process.GetProcessId()); + } + + // Prepare to process handles. + auto& dst_handle_table = dst_process.GetHandleTable(); + auto& src_handle_table = src_process.GetHandleTable(); + Result result = ResultSuccess; + + // Process copy handles. + for (auto i = 0; i < src_special_header.GetCopyHandleCount(); ++i) { + // Get the handles. + const Handle src_handle = src_msg.GetHandle(offset); + Handle dst_handle = Svc::InvalidHandle; + + // If we're in a success state, try to move the handle to the new table. + if (R_SUCCEEDED(result) && src_handle != Svc::InvalidHandle) { + KScopedAutoObject obj = + src_handle_table.GetObjectForIpc(src_handle, std::addressof(src_thread)); + if (obj.IsNotNull()) { + Result add_result = + dst_handle_table.Add(std::addressof(dst_handle), obj.GetPointerUnsafe()); + if (R_FAILED(add_result)) { + result = add_result; + dst_handle = Svc::InvalidHandle; + } + } else { + result = ResultInvalidHandle; + } + } + + // Set the handle. + offset = dst_msg.SetHandle(offset, dst_handle); + } + + // Process move handles. + if constexpr (MoveHandleAllowed) { + for (auto i = 0; i < src_special_header.GetMoveHandleCount(); ++i) { + // Get the handles. + const Handle src_handle = src_msg.GetHandle(offset); + Handle dst_handle = Svc::InvalidHandle; + + // Whether or not we've succeeded, we need to remove the handles from the source table. + if (src_handle != Svc::InvalidHandle) { + if (R_SUCCEEDED(result)) { + KScopedAutoObject obj = + src_handle_table.GetObjectForIpcWithoutPseudoHandle(src_handle); + if (obj.IsNotNull()) { + Result add_result = dst_handle_table.Add(std::addressof(dst_handle), + obj.GetPointerUnsafe()); + + src_handle_table.Remove(src_handle); + + if (R_FAILED(add_result)) { + result = add_result; + dst_handle = Svc::InvalidHandle; + } + } else { + result = ResultInvalidHandle; + } + } else { + src_handle_table.Remove(src_handle); + } + } + + // Set the handle. + offset = dst_msg.SetHandle(offset, dst_handle); + } + } + + R_RETURN(result); +} + +void CleanupSpecialData(KProcess& dst_process, u32* dst_msg_ptr, size_t dst_buffer_size) { + // Parse the message. + const MessageBuffer dst_msg(dst_msg_ptr, dst_buffer_size); + const MessageBuffer::MessageHeader dst_header(dst_msg); + const MessageBuffer::SpecialHeader dst_special_header(dst_msg, dst_header); + + // Check that the size is big enough. + if (MessageBuffer::GetMessageBufferSize(dst_header, dst_special_header) > dst_buffer_size) { + return; + } + + // Set the special header. + int offset = dst_msg.Set(dst_special_header); + + // Clear the process id, if needed. + if (dst_special_header.GetHasProcessId()) { + offset = dst_msg.SetProcessId(offset, 0); + } + + // Clear handles, as relevant. + auto& dst_handle_table = dst_process.GetHandleTable(); + for (auto i = 0; + i < (dst_special_header.GetCopyHandleCount() + dst_special_header.GetMoveHandleCount()); + ++i) { + const Handle handle = dst_msg.GetHandle(offset); + + if (handle != Svc::InvalidHandle) { + dst_handle_table.Remove(handle); + } + + offset = dst_msg.SetHandle(offset, Svc::InvalidHandle); + } +} + +} // namespace + using ThreadQueueImplForKServerSessionRequest = KThreadQueue; KServerSession::KServerSession(KernelCore& kernel) @@ -223,12 +343,27 @@ Result KServerSession::SendReply(bool is_hle) { // the reply has already been written in this case. } else { Core::Memory::Memory& memory{client_thread->GetOwnerProcess()->GetMemory()}; - KThread* server_thread{GetCurrentThreadPointer(m_kernel)}; + KThread* server_thread = GetCurrentThreadPointer(m_kernel); + KProcess& src_process = *client_thread->GetOwnerProcess(); + KProcess& dst_process = *server_thread->GetOwnerProcess(); UNIMPLEMENTED_IF(server_thread->GetOwnerProcess() != client_thread->GetOwnerProcess()); - auto* src_msg_buffer = memory.GetPointer(server_thread->GetTlsAddress()); - auto* dst_msg_buffer = memory.GetPointer(client_message); + auto* src_msg_buffer = memory.GetPointer<u32>(server_thread->GetTlsAddress()); + auto* dst_msg_buffer = memory.GetPointer<u32>(client_message); std::memcpy(dst_msg_buffer, src_msg_buffer, client_buffer_size); + + // Translate special header ad-hoc. + MessageBuffer src_msg(src_msg_buffer, client_buffer_size); + MessageBuffer::MessageHeader src_header(src_msg); + MessageBuffer::SpecialHeader src_special_header(src_msg, src_header); + if (src_header.GetHasSpecialHeader()) { + MessageBuffer dst_msg(dst_msg_buffer, client_buffer_size); + result = ProcessMessageSpecialData<true>(dst_process, src_process, *server_thread, + dst_msg, src_msg, src_special_header); + if (R_FAILED(result)) { + CleanupSpecialData(dst_process, dst_msg_buffer, client_buffer_size); + } + } } } else { result = ResultSessionClosed; @@ -330,12 +465,28 @@ Result KServerSession::ReceiveRequest(std::shared_ptr<Service::HLERequestContext ->PopulateFromIncomingCommandBuffer(client_thread->GetOwnerProcess()->GetHandleTable(), cmd_buf); } else { - KThread* server_thread{GetCurrentThreadPointer(m_kernel)}; - UNIMPLEMENTED_IF(server_thread->GetOwnerProcess() != client_thread->GetOwnerProcess()); + KThread* server_thread = GetCurrentThreadPointer(m_kernel); + KProcess& src_process = *client_thread->GetOwnerProcess(); + KProcess& dst_process = *server_thread->GetOwnerProcess(); + UNIMPLEMENTED_IF(client_thread->GetOwnerProcess() != server_thread->GetOwnerProcess()); - auto* src_msg_buffer = memory.GetPointer(client_message); - auto* dst_msg_buffer = memory.GetPointer(server_thread->GetTlsAddress()); + auto* src_msg_buffer = memory.GetPointer<u32>(client_message); + auto* dst_msg_buffer = memory.GetPointer<u32>(server_thread->GetTlsAddress()); std::memcpy(dst_msg_buffer, src_msg_buffer, client_buffer_size); + + // Translate special header ad-hoc. + // TODO: fix this mess + MessageBuffer src_msg(src_msg_buffer, client_buffer_size); + MessageBuffer::MessageHeader src_header(src_msg); + MessageBuffer::SpecialHeader src_special_header(src_msg, src_header); + if (src_header.GetHasSpecialHeader()) { + MessageBuffer dst_msg(dst_msg_buffer, client_buffer_size); + Result res = ProcessMessageSpecialData<false>(dst_process, src_process, *client_thread, + dst_msg, src_msg, src_special_header); + if (R_FAILED(res)) { + CleanupSpecialData(dst_process, dst_msg_buffer, client_buffer_size); + } + } } // We succeeded. diff --git a/src/core/hle/kernel/k_thread.h b/src/core/hle/kernel/k_thread.h index dd662b3f8..d178c2453 100644 --- a/src/core/hle/kernel/k_thread.h +++ b/src/core/hle/kernel/k_thread.h @@ -338,6 +338,15 @@ public: return m_parent != nullptr; } + std::span<KSynchronizationObject*> GetSynchronizationObjectBuffer() { + return m_sync_object_buffer.sync_objects; + } + + std::span<Handle> GetHandleBuffer() { + return {m_sync_object_buffer.handles.data() + Svc::ArgumentHandleCountMax, + Svc::ArgumentHandleCountMax}; + } + u16 GetUserDisableCount() const; void SetInterruptFlag(); void ClearInterruptFlag(); @@ -855,6 +864,7 @@ private: u32* m_light_ipc_data{}; KProcessAddress m_tls_address{}; KLightLock m_activity_pause_lock; + SyncObjectBuffer m_sync_object_buffer{}; s64 m_schedule_count{}; s64 m_last_scheduled_tick{}; std::array<QueueEntry, Core::Hardware::NUM_CPU_CORES> m_per_core_priority_queue_entry{}; diff --git a/src/core/hle/kernel/message_buffer.h b/src/core/hle/kernel/message_buffer.h new file mode 100644 index 000000000..75b275310 --- /dev/null +++ b/src/core/hle/kernel/message_buffer.h @@ -0,0 +1,612 @@ +// SPDX-FileCopyrightText: 2023 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include "common/alignment.h" +#include "common/bit_field.h" +#include "core/hle/kernel/k_thread.h" + +namespace Kernel { + +constexpr inline size_t MessageBufferSize = 0x100; + +class MessageBuffer { +public: + class MessageHeader { + private: + static constexpr inline u64 NullTag = 0; + + public: + enum class ReceiveListCountType : u32 { + None = 0, + ToMessageBuffer = 1, + ToSingleBuffer = 2, + + CountOffset = 2, + CountMax = 13, + }; + + private: + union { + std::array<u32, 2> raw; + + struct { + // Define fields for the first header word. + union { + BitField<0, 16, u16> tag; + BitField<16, 4, u32> pointer_count; + BitField<20, 4, u32> send_count; + BitField<24, 4, u32> receive_count; + BitField<28, 4, u32> exchange_count; + }; + + // Define fields for the second header word. + union { + BitField<0, 10, u32> raw_count; + BitField<10, 4, ReceiveListCountType> receive_list_count; + BitField<14, 6, u32> reserved0; + BitField<20, 11, u32> receive_list_offset; + BitField<31, 1, u32> has_special_header; + }; + }; + } m_header; + + public: + constexpr MessageHeader() : m_header{} {} + + constexpr MessageHeader(u16 tag, bool special, s32 ptr, s32 send, s32 recv, s32 exch, + s32 raw, ReceiveListCountType recv_list) + : m_header{} { + m_header.raw[0] = 0; + m_header.raw[1] = 0; + + m_header.tag.Assign(tag); + m_header.pointer_count.Assign(ptr); + m_header.send_count.Assign(send); + m_header.receive_count.Assign(recv); + m_header.exchange_count.Assign(exch); + + m_header.raw_count.Assign(raw); + m_header.receive_list_count.Assign(recv_list); + m_header.has_special_header.Assign(special); + } + + explicit MessageHeader(const MessageBuffer& buf) : m_header{} { + buf.Get(0, m_header.raw.data(), 2); + } + + explicit MessageHeader(const u32* msg) : m_header{{msg[0], msg[1]}} {} + + constexpr u16 GetTag() const { + return m_header.tag; + } + + constexpr s32 GetPointerCount() const { + return m_header.pointer_count; + } + + constexpr s32 GetSendCount() const { + return m_header.send_count; + } + + constexpr s32 GetReceiveCount() const { + return m_header.receive_count; + } + + constexpr s32 GetExchangeCount() const { + return m_header.exchange_count; + } + + constexpr s32 GetMapAliasCount() const { + return this->GetSendCount() + this->GetReceiveCount() + this->GetExchangeCount(); + } + + constexpr s32 GetRawCount() const { + return m_header.raw_count; + } + + constexpr ReceiveListCountType GetReceiveListCount() const { + return m_header.receive_list_count; + } + + constexpr s32 GetReceiveListOffset() const { + return m_header.receive_list_offset; + } + + constexpr bool GetHasSpecialHeader() const { + return m_header.has_special_header.Value() != 0; + } + + constexpr void SetReceiveListCount(ReceiveListCountType recv_list) { + m_header.receive_list_count.Assign(recv_list); + } + + constexpr const u32* GetData() const { + return m_header.raw.data(); + } + + static constexpr size_t GetDataSize() { + return sizeof(m_header); + } + }; + + class SpecialHeader { + private: + union { + std::array<u32, 1> raw; + + // Define fields for the header word. + BitField<0, 1, u32> has_process_id; + BitField<1, 4, u32> copy_handle_count; + BitField<5, 4, u32> move_handle_count; + } m_header; + bool m_has_header; + + public: + constexpr explicit SpecialHeader(bool pid, s32 copy, s32 move) + : m_header{}, m_has_header(true) { + m_header.has_process_id.Assign(pid); + m_header.copy_handle_count.Assign(copy); + m_header.move_handle_count.Assign(move); + } + + constexpr explicit SpecialHeader(bool pid, s32 copy, s32 move, bool _has_header) + : m_header{}, m_has_header(_has_header) { + m_header.has_process_id.Assign(pid); + m_header.copy_handle_count.Assign(copy); + m_header.move_handle_count.Assign(move); + } + + explicit SpecialHeader(const MessageBuffer& buf, const MessageHeader& hdr) + : m_header{}, m_has_header(hdr.GetHasSpecialHeader()) { + if (m_has_header) { + buf.Get(static_cast<s32>(MessageHeader::GetDataSize() / sizeof(u32)), + m_header.raw.data(), sizeof(m_header) / sizeof(u32)); + } + } + + constexpr bool GetHasProcessId() const { + return m_header.has_process_id.Value() != 0; + } + + constexpr s32 GetCopyHandleCount() const { + return m_header.copy_handle_count; + } + + constexpr s32 GetMoveHandleCount() const { + return m_header.move_handle_count; + } + + constexpr const u32* GetHeader() const { + return m_header.raw.data(); + } + + constexpr size_t GetHeaderSize() const { + if (m_has_header) { + return sizeof(m_header); + } else { + return 0; + } + } + + constexpr size_t GetDataSize() const { + if (m_has_header) { + return (this->GetHasProcessId() ? sizeof(u64) : 0) + + (this->GetCopyHandleCount() * sizeof(Handle)) + + (this->GetMoveHandleCount() * sizeof(Handle)); + } else { + return 0; + } + } + }; + + class MapAliasDescriptor { + public: + enum class Attribute : u32 { + Ipc = 0, + NonSecureIpc = 1, + NonDeviceIpc = 3, + }; + + private: + static constexpr u32 SizeLowCount = 32; + static constexpr u32 SizeHighCount = 4; + static constexpr u32 AddressLowCount = 32; + static constexpr u32 AddressMidCount = 4; + + constexpr u32 GetAddressMid(u64 address) { + return static_cast<u32>(address >> AddressLowCount) & ((1U << AddressMidCount) - 1); + } + + constexpr u32 GetAddressHigh(u64 address) { + return static_cast<u32>(address >> (AddressLowCount + AddressMidCount)); + } + + private: + union { + std::array<u32, 3> raw; + + struct { + // Define fields for the first two words. + u32 size_low; + u32 address_low; + + // Define fields for the packed descriptor word. + union { + BitField<0, 2, Attribute> attributes; + BitField<2, 3, u32> address_high; + BitField<5, 19, u32> reserved; + BitField<24, 4, u32> size_high; + BitField<28, 4, u32> address_mid; + }; + }; + } m_data; + + public: + constexpr MapAliasDescriptor() : m_data{} {} + + MapAliasDescriptor(const void* buffer, size_t _size, Attribute attr = Attribute::Ipc) + : m_data{} { + const u64 address = reinterpret_cast<u64>(buffer); + const u64 size = static_cast<u64>(_size); + m_data.size_low = static_cast<u32>(size); + m_data.address_low = static_cast<u32>(address); + m_data.attributes.Assign(attr); + m_data.address_mid.Assign(GetAddressMid(address)); + m_data.size_high.Assign(static_cast<u32>(size >> SizeLowCount)); + m_data.address_high.Assign(GetAddressHigh(address)); + } + + MapAliasDescriptor(const MessageBuffer& buf, s32 index) : m_data{} { + buf.Get(index, m_data.raw.data(), 3); + } + + constexpr uintptr_t GetAddress() const { + return (static_cast<u64>((m_data.address_high << AddressMidCount) | m_data.address_mid) + << AddressLowCount) | + m_data.address_low; + } + + constexpr uintptr_t GetSize() const { + return (static_cast<u64>(m_data.size_high) << SizeLowCount) | m_data.size_low; + } + + constexpr Attribute GetAttribute() const { + return m_data.attributes; + } + + constexpr const u32* GetData() const { + return m_data.raw.data(); + } + + static constexpr size_t GetDataSize() { + return sizeof(m_data); + } + }; + + class PointerDescriptor { + private: + static constexpr u32 AddressLowCount = 32; + static constexpr u32 AddressMidCount = 4; + + constexpr u32 GetAddressMid(u64 address) { + return static_cast<u32>(address >> AddressLowCount) & ((1u << AddressMidCount) - 1); + } + + constexpr u32 GetAddressHigh(u64 address) { + return static_cast<u32>(address >> (AddressLowCount + AddressMidCount)); + } + + private: + union { + std::array<u32, 2> raw; + + struct { + // Define fields for the packed descriptor word. + union { + BitField<0, 4, u32> index; + BitField<4, 2, u32> reserved0; + BitField<6, 3, u32> address_high; + BitField<9, 3, u32> reserved1; + BitField<12, 4, u32> address_mid; + BitField<16, 16, u32> size; + }; + + // Define fields for the second word. + u32 address_low; + }; + } m_data; + + public: + constexpr PointerDescriptor() : m_data{} {} + + PointerDescriptor(const void* buffer, size_t size, s32 index) : m_data{} { + const u64 address = reinterpret_cast<u64>(buffer); + + m_data.index.Assign(index); + m_data.address_high.Assign(GetAddressHigh(address)); + m_data.address_mid.Assign(GetAddressMid(address)); + m_data.size.Assign(static_cast<u32>(size)); + + m_data.address_low = static_cast<u32>(address); + } + + PointerDescriptor(const MessageBuffer& buf, s32 index) : m_data{} { + buf.Get(index, m_data.raw.data(), 2); + } + + constexpr s32 GetIndex() const { + return m_data.index; + } + + constexpr uintptr_t GetAddress() const { + return (static_cast<u64>((m_data.address_high << AddressMidCount) | m_data.address_mid) + << AddressLowCount) | + m_data.address_low; + } + + constexpr size_t GetSize() const { + return m_data.size; + } + + constexpr const u32* GetData() const { + return m_data.raw.data(); + } + + static constexpr size_t GetDataSize() { + return sizeof(m_data); + } + }; + + class ReceiveListEntry { + private: + static constexpr u32 AddressLowCount = 32; + + constexpr u32 GetAddressHigh(u64 address) { + return static_cast<u32>(address >> (AddressLowCount)); + } + + private: + union { + std::array<u32, 2> raw; + + struct { + // Define fields for the first word. + u32 address_low; + + // Define fields for the packed descriptor word. + union { + BitField<0, 7, u32> address_high; + BitField<7, 9, u32> reserved; + BitField<16, 16, u32> size; + }; + }; + } m_data; + + public: + constexpr ReceiveListEntry() : m_data{} {} + + ReceiveListEntry(const void* buffer, size_t size) : m_data{} { + const u64 address = reinterpret_cast<u64>(buffer); + + m_data.address_low = static_cast<u32>(address); + + m_data.address_high.Assign(GetAddressHigh(address)); + m_data.size.Assign(static_cast<u32>(size)); + } + + ReceiveListEntry(u32 a, u32 b) : m_data{{a, b}} {} + + constexpr uintptr_t GetAddress() const { + return (static_cast<u64>(m_data.address_high) << AddressLowCount) | m_data.address_low; + } + + constexpr size_t GetSize() const { + return m_data.size; + } + + constexpr const u32* GetData() const { + return m_data.raw.data(); + } + + static constexpr size_t GetDataSize() { + return sizeof(m_data); + } + }; + +private: + u32* m_buffer; + size_t m_size; + +public: + constexpr MessageBuffer(u32* b, size_t sz) : m_buffer(b), m_size(sz) {} + constexpr explicit MessageBuffer(u32* b) : m_buffer(b), m_size(MessageBufferSize) {} + + constexpr void* GetBufferForDebug() const { + return m_buffer; + } + + constexpr size_t GetBufferSize() const { + return m_size; + } + + void Get(s32 index, u32* dst, size_t count) const { + // Ensure that this doesn't get re-ordered. + std::atomic_thread_fence(std::memory_order_seq_cst); + + // Get the words. + static_assert(sizeof(*dst) == sizeof(*m_buffer)); + + memcpy(dst, m_buffer + index, count * sizeof(*dst)); + } + + s32 Set(s32 index, u32* src, size_t count) const { + // Ensure that this doesn't get re-ordered. + std::atomic_thread_fence(std::memory_order_seq_cst); + + // Set the words. + memcpy(m_buffer + index, src, count * sizeof(*src)); + + // Ensure that this doesn't get re-ordered. + std::atomic_thread_fence(std::memory_order_seq_cst); + + return static_cast<s32>(index + count); + } + + template <typename T> + const T& GetRaw(s32 index) const { + return *reinterpret_cast<const T*>(m_buffer + index); + } + + template <typename T> + s32 SetRaw(s32 index, const T& val) const { + *reinterpret_cast<const T*>(m_buffer + index) = val; + return index + (Common::AlignUp(sizeof(val), sizeof(*m_buffer)) / sizeof(*m_buffer)); + } + + void GetRawArray(s32 index, void* dst, size_t len) const { + memcpy(dst, m_buffer + index, len); + } + + void SetRawArray(s32 index, const void* src, size_t len) const { + memcpy(m_buffer + index, src, len); + } + + void SetNull() const { + this->Set(MessageHeader()); + } + + s32 Set(const MessageHeader& hdr) const { + memcpy(m_buffer, hdr.GetData(), hdr.GetDataSize()); + return static_cast<s32>(hdr.GetDataSize() / sizeof(*m_buffer)); + } + + s32 Set(const SpecialHeader& spc) const { + const s32 index = static_cast<s32>(MessageHeader::GetDataSize() / sizeof(*m_buffer)); + memcpy(m_buffer + index, spc.GetHeader(), spc.GetHeaderSize()); + return static_cast<s32>(index + (spc.GetHeaderSize() / sizeof(*m_buffer))); + } + + s32 SetHandle(s32 index, const Handle& hnd) const { + memcpy(m_buffer + index, std::addressof(hnd), sizeof(hnd)); + return static_cast<s32>(index + (sizeof(hnd) / sizeof(*m_buffer))); + } + + s32 SetProcessId(s32 index, const u64 pid) const { + memcpy(m_buffer + index, std::addressof(pid), sizeof(pid)); + return static_cast<s32>(index + (sizeof(pid) / sizeof(*m_buffer))); + } + + s32 Set(s32 index, const MapAliasDescriptor& desc) const { + memcpy(m_buffer + index, desc.GetData(), desc.GetDataSize()); + return static_cast<s32>(index + (desc.GetDataSize() / sizeof(*m_buffer))); + } + + s32 Set(s32 index, const PointerDescriptor& desc) const { + memcpy(m_buffer + index, desc.GetData(), desc.GetDataSize()); + return static_cast<s32>(index + (desc.GetDataSize() / sizeof(*m_buffer))); + } + + s32 Set(s32 index, const ReceiveListEntry& desc) const { + memcpy(m_buffer + index, desc.GetData(), desc.GetDataSize()); + return static_cast<s32>(index + (desc.GetDataSize() / sizeof(*m_buffer))); + } + + s32 Set(s32 index, const u32 val) const { + memcpy(m_buffer + index, std::addressof(val), sizeof(val)); + return static_cast<s32>(index + (sizeof(val) / sizeof(*m_buffer))); + } + + Result GetAsyncResult() const { + MessageHeader hdr(m_buffer); + MessageHeader null{}; + if (memcmp(hdr.GetData(), null.GetData(), MessageHeader::GetDataSize()) != 0) [[unlikely]] { + R_SUCCEED(); + } + return Result(m_buffer[MessageHeader::GetDataSize() / sizeof(*m_buffer)]); + } + + void SetAsyncResult(Result res) const { + const s32 index = this->Set(MessageHeader()); + const auto value = res.raw; + memcpy(m_buffer + index, std::addressof(value), sizeof(value)); + } + + u32 Get32(s32 index) const { + return m_buffer[index]; + } + + u64 Get64(s32 index) const { + u64 value; + memcpy(std::addressof(value), m_buffer + index, sizeof(value)); + return value; + } + + u64 GetProcessId(s32 index) const { + return this->Get64(index); + } + + Handle GetHandle(s32 index) const { + static_assert(sizeof(Handle) == sizeof(*m_buffer)); + return Handle(m_buffer[index]); + } + + static constexpr s32 GetSpecialDataIndex(const MessageHeader& hdr, const SpecialHeader& spc) { + return static_cast<s32>((MessageHeader::GetDataSize() / sizeof(u32)) + + (spc.GetHeaderSize() / sizeof(u32))); + } + + static constexpr s32 GetPointerDescriptorIndex(const MessageHeader& hdr, + const SpecialHeader& spc) { + return static_cast<s32>(GetSpecialDataIndex(hdr, spc) + (spc.GetDataSize() / sizeof(u32))); + } + + static constexpr s32 GetMapAliasDescriptorIndex(const MessageHeader& hdr, + const SpecialHeader& spc) { + return GetPointerDescriptorIndex(hdr, spc) + + static_cast<s32>(hdr.GetPointerCount() * PointerDescriptor::GetDataSize() / + sizeof(u32)); + } + + static constexpr s32 GetRawDataIndex(const MessageHeader& hdr, const SpecialHeader& spc) { + return GetMapAliasDescriptorIndex(hdr, spc) + + static_cast<s32>(hdr.GetMapAliasCount() * MapAliasDescriptor::GetDataSize() / + sizeof(u32)); + } + + static constexpr s32 GetReceiveListIndex(const MessageHeader& hdr, const SpecialHeader& spc) { + if (const s32 recv_list_index = hdr.GetReceiveListOffset()) { + return recv_list_index; + } else { + return GetRawDataIndex(hdr, spc) + hdr.GetRawCount(); + } + } + + static constexpr size_t GetMessageBufferSize(const MessageHeader& hdr, + const SpecialHeader& spc) { + // Get the size of the plain message. + size_t msg_size = GetReceiveListIndex(hdr, spc) * sizeof(u32); + + // Add the size of the receive list. + const auto count = hdr.GetReceiveListCount(); + switch (count) { + case MessageHeader::ReceiveListCountType::None: + break; + case MessageHeader::ReceiveListCountType::ToMessageBuffer: + break; + case MessageHeader::ReceiveListCountType::ToSingleBuffer: + msg_size += ReceiveListEntry::GetDataSize(); + break; + default: + msg_size += (static_cast<s32>(count) - + static_cast<s32>(MessageHeader::ReceiveListCountType::CountOffset)) * + ReceiveListEntry::GetDataSize(); + break; + } + + return msg_size; + } +}; + +} // namespace Kernel diff --git a/src/core/hle/kernel/svc/svc_ipc.cpp b/src/core/hle/kernel/svc/svc_ipc.cpp index 60247df2e..bb94f6934 100644 --- a/src/core/hle/kernel/svc/svc_ipc.cpp +++ b/src/core/hle/kernel/svc/svc_ipc.cpp @@ -38,22 +38,31 @@ Result SendAsyncRequestWithUserBuffer(Core::System& system, Handle* out_event_ha Result ReplyAndReceive(Core::System& system, s32* out_index, uint64_t handles_addr, s32 num_handles, Handle reply_target, s64 timeout_ns) { + // Ensure number of handles is valid. + R_UNLESS(0 <= num_handles && num_handles <= ArgumentHandleCountMax, ResultOutOfRange); + + // Get the synchronization context. auto& kernel = system.Kernel(); auto& handle_table = GetCurrentProcess(kernel).GetHandleTable(); - - R_UNLESS(0 <= num_handles && num_handles <= ArgumentHandleCountMax, ResultOutOfRange); - R_UNLESS(GetCurrentMemory(kernel).IsValidVirtualAddressRange( - handles_addr, static_cast<u64>(sizeof(Handle) * num_handles)), - ResultInvalidPointer); - - std::array<Handle, Svc::ArgumentHandleCountMax> handles; - GetCurrentMemory(kernel).ReadBlock(handles_addr, handles.data(), sizeof(Handle) * num_handles); - - // Convert handle list to object table. - std::array<KSynchronizationObject*, Svc::ArgumentHandleCountMax> objs; - R_UNLESS(handle_table.GetMultipleObjects<KSynchronizationObject>(objs.data(), handles.data(), - num_handles), - ResultInvalidHandle); + auto objs = GetCurrentThread(kernel).GetSynchronizationObjectBuffer(); + auto handles = GetCurrentThread(kernel).GetHandleBuffer(); + + // Copy user handles. + if (num_handles > 0) { + // Ensure we can try to get the handles. + R_UNLESS(GetCurrentMemory(kernel).IsValidVirtualAddressRange( + handles_addr, static_cast<u64>(sizeof(Handle) * num_handles)), + ResultInvalidPointer); + + // Get the handles. + GetCurrentMemory(kernel).ReadBlock(handles_addr, handles.data(), + sizeof(Handle) * num_handles); + + // Convert the handles to objects. + R_UNLESS(handle_table.GetMultipleObjects<KSynchronizationObject>( + objs.data(), handles.data(), num_handles), + ResultInvalidHandle); + } // Ensure handles are closed when we're done. SCOPE_EXIT({ diff --git a/src/core/hle/kernel/svc/svc_synchronization.cpp b/src/core/hle/kernel/svc/svc_synchronization.cpp index 53df5bcd8..f02d03f30 100644 --- a/src/core/hle/kernel/svc/svc_synchronization.cpp +++ b/src/core/hle/kernel/svc/svc_synchronization.cpp @@ -47,21 +47,35 @@ Result ResetSignal(Core::System& system, Handle handle) { R_THROW(ResultInvalidHandle); } -static Result WaitSynchronization(Core::System& system, int32_t* out_index, const Handle* handles, - int32_t num_handles, int64_t timeout_ns) { +/// Wait for the given handles to synchronize, timeout after the specified nanoseconds +Result WaitSynchronization(Core::System& system, int32_t* out_index, u64 user_handles, + int32_t num_handles, int64_t timeout_ns) { + LOG_TRACE(Kernel_SVC, "called user_handles={:#x}, num_handles={}, timeout_ns={}", user_handles, + num_handles, timeout_ns); + // Ensure number of handles is valid. R_UNLESS(0 <= num_handles && num_handles <= Svc::ArgumentHandleCountMax, ResultOutOfRange); // Get the synchronization context. auto& kernel = system.Kernel(); auto& handle_table = GetCurrentProcess(kernel).GetHandleTable(); - std::array<KSynchronizationObject*, Svc::ArgumentHandleCountMax> objs; + auto objs = GetCurrentThread(kernel).GetSynchronizationObjectBuffer(); + auto handles = GetCurrentThread(kernel).GetHandleBuffer(); // Copy user handles. if (num_handles > 0) { + // Ensure we can try to get the handles. + R_UNLESS(GetCurrentMemory(kernel).IsValidVirtualAddressRange( + user_handles, static_cast<u64>(sizeof(Handle) * num_handles)), + ResultInvalidPointer); + + // Get the handles. + GetCurrentMemory(kernel).ReadBlock(user_handles, handles.data(), + sizeof(Handle) * num_handles); + // Convert the handles to objects. - R_UNLESS(handle_table.GetMultipleObjects<KSynchronizationObject>(objs.data(), handles, - num_handles), + R_UNLESS(handle_table.GetMultipleObjects<KSynchronizationObject>( + objs.data(), handles.data(), num_handles), ResultInvalidHandle); } @@ -80,23 +94,6 @@ static Result WaitSynchronization(Core::System& system, int32_t* out_index, cons R_RETURN(res); } -/// Wait for the given handles to synchronize, timeout after the specified nanoseconds -Result WaitSynchronization(Core::System& system, int32_t* out_index, u64 user_handles, - int32_t num_handles, int64_t timeout_ns) { - LOG_TRACE(Kernel_SVC, "called user_handles={:#x}, num_handles={}, timeout_ns={}", user_handles, - num_handles, timeout_ns); - - // Ensure number of handles is valid. - R_UNLESS(0 <= num_handles && num_handles <= Svc::ArgumentHandleCountMax, ResultOutOfRange); - std::array<Handle, Svc::ArgumentHandleCountMax> handles; - if (num_handles > 0) { - GetCurrentMemory(system.Kernel()) - .ReadBlock(user_handles, handles.data(), num_handles * sizeof(Handle)); - } - - R_RETURN(WaitSynchronization(system, out_index, handles.data(), num_handles, timeout_ns)); -} - /// Resumes a thread waiting on WaitSynchronization Result CancelSynchronization(Core::System& system, Handle handle) { LOG_TRACE(Kernel_SVC, "called handle=0x{:X}", handle); diff --git a/src/core/hle/service/audio/audin_u.cpp b/src/core/hle/service/audio/audin_u.cpp index c8d574993..526a39130 100644 --- a/src/core/hle/service/audio/audin_u.cpp +++ b/src/core/hle/service/audio/audin_u.cpp @@ -5,7 +5,7 @@ #include "audio_core/renderer/audio_device.h" #include "common/common_funcs.h" #include "common/logging/log.h" -#include "common/settings.h" +#include "common/scratch_buffer.h" #include "common/string_util.h" #include "core/core.h" #include "core/hle/kernel/k_event.h" @@ -124,12 +124,15 @@ private: void GetReleasedAudioInBuffer(HLERequestContext& ctx) { const auto write_buffer_size = ctx.GetWriteBufferNumElements<u64>(); - tmp_buffer.resize_destructive(write_buffer_size); - tmp_buffer[0] = 0; + released_buffer.resize_destructive(write_buffer_size); + released_buffer[0] = 0; - const auto count = impl->GetReleasedBuffers(tmp_buffer); + const auto count = impl->GetReleasedBuffers(released_buffer); - ctx.WriteBuffer(tmp_buffer); + LOG_TRACE(Service_Audio, "called. Session {} released {} buffers", + impl->GetSystem().GetSessionId(), count); + + ctx.WriteBuffer(released_buffer); IPC::ResponseBuilder rb{ctx, 3}; rb.Push(ResultSuccess); @@ -155,7 +158,6 @@ private: LOG_DEBUG(Service_Audio, "called. Buffer count={}", buffer_count); IPC::ResponseBuilder rb{ctx, 3}; - rb.Push(ResultSuccess); rb.Push(buffer_count); } @@ -195,7 +197,7 @@ private: KernelHelpers::ServiceContext service_context; Kernel::KEvent* event; std::shared_ptr<AudioCore::AudioIn::In> impl; - Common::ScratchBuffer<u64> tmp_buffer; + Common::ScratchBuffer<u64> released_buffer; }; AudInU::AudInU(Core::System& system_) diff --git a/src/core/hle/service/audio/audout_u.cpp b/src/core/hle/service/audio/audout_u.cpp index 032c8c11f..23f84a29f 100644 --- a/src/core/hle/service/audio/audout_u.cpp +++ b/src/core/hle/service/audio/audout_u.cpp @@ -9,6 +9,7 @@ #include "audio_core/renderer/audio_device.h" #include "common/common_funcs.h" #include "common/logging/log.h" +#include "common/scratch_buffer.h" #include "common/string_util.h" #include "common/swap.h" #include "core/core.h" @@ -102,8 +103,8 @@ private: AudioOutBuffer buffer{}; std::memcpy(&buffer, in_buffer.data(), sizeof(AudioOutBuffer)); - [[maybe_unused]] auto sessionid{impl->GetSystem().GetSessionId()}; - LOG_TRACE(Service_Audio, "called. Session {} Appending buffer {:08X}", sessionid, tag); + LOG_TRACE(Service_Audio, "called. Session {} Appending buffer {:08X}", + impl->GetSystem().GetSessionId(), tag); auto result = impl->AppendBuffer(buffer, tag); @@ -123,12 +124,15 @@ private: void GetReleasedAudioOutBuffers(HLERequestContext& ctx) { const auto write_buffer_size = ctx.GetWriteBufferNumElements<u64>(); - tmp_buffer.resize_destructive(write_buffer_size); - tmp_buffer[0] = 0; + released_buffer.resize_destructive(write_buffer_size); + released_buffer[0] = 0; - const auto count = impl->GetReleasedBuffers(tmp_buffer); + const auto count = impl->GetReleasedBuffers(released_buffer); - ctx.WriteBuffer(tmp_buffer); + ctx.WriteBuffer(released_buffer); + + LOG_TRACE(Service_Audio, "called. Session {} released {} buffers", + impl->GetSystem().GetSessionId(), count); IPC::ResponseBuilder rb{ctx, 3}; rb.Push(ResultSuccess); @@ -154,7 +158,6 @@ private: LOG_DEBUG(Service_Audio, "called. Buffer count={}", buffer_count); IPC::ResponseBuilder rb{ctx, 3}; - rb.Push(ResultSuccess); rb.Push(buffer_count); } @@ -165,7 +168,6 @@ private: LOG_DEBUG(Service_Audio, "called. Played samples={}", samples_played); IPC::ResponseBuilder rb{ctx, 4}; - rb.Push(ResultSuccess); rb.Push(samples_played); } @@ -205,7 +207,7 @@ private: KernelHelpers::ServiceContext service_context; Kernel::KEvent* event; std::shared_ptr<AudioCore::AudioOut::Out> impl; - Common::ScratchBuffer<u64> tmp_buffer; + Common::ScratchBuffer<u64> released_buffer; }; AudOutU::AudOutU(Core::System& system_) diff --git a/src/core/hle/service/audio/audren_u.cpp b/src/core/hle/service/audio/audren_u.cpp index 12845c23a..003870176 100644 --- a/src/core/hle/service/audio/audren_u.cpp +++ b/src/core/hle/service/audio/audren_u.cpp @@ -15,6 +15,7 @@ #include "common/common_funcs.h" #include "common/logging/log.h" #include "common/polyfill_ranges.h" +#include "common/scratch_buffer.h" #include "common/string_util.h" #include "core/core.h" #include "core/hle/kernel/k_event.h" @@ -119,23 +120,23 @@ private: auto is_buffer_b{ctx.BufferDescriptorB()[0].Size() != 0}; if (is_buffer_b) { const auto buffersB{ctx.BufferDescriptorB()}; - tmp_output.resize_destructive(buffersB[0].Size()); - tmp_performance.resize_destructive(buffersB[1].Size()); + output_buffer.resize_destructive(buffersB[0].Size()); + performance_buffer.resize_destructive(buffersB[1].Size()); } else { const auto buffersC{ctx.BufferDescriptorC()}; - tmp_output.resize_destructive(buffersC[0].Size()); - tmp_performance.resize_destructive(buffersC[1].Size()); + output_buffer.resize_destructive(buffersC[0].Size()); + performance_buffer.resize_destructive(buffersC[1].Size()); } - auto result = impl->RequestUpdate(input, tmp_performance, tmp_output); + auto result = impl->RequestUpdate(input, performance_buffer, output_buffer); if (result.IsSuccess()) { if (is_buffer_b) { - ctx.WriteBufferB(tmp_output.data(), tmp_output.size(), 0); - ctx.WriteBufferB(tmp_performance.data(), tmp_performance.size(), 1); + ctx.WriteBufferB(output_buffer.data(), output_buffer.size(), 0); + ctx.WriteBufferB(performance_buffer.data(), performance_buffer.size(), 1); } else { - ctx.WriteBufferC(tmp_output.data(), tmp_output.size(), 0); - ctx.WriteBufferC(tmp_performance.data(), tmp_performance.size(), 1); + ctx.WriteBufferC(output_buffer.data(), output_buffer.size(), 0); + ctx.WriteBufferC(performance_buffer.data(), performance_buffer.size(), 1); } } else { LOG_ERROR(Service_Audio, "RequestUpdate failed error 0x{:02X}!", result.description); @@ -233,8 +234,8 @@ private: Kernel::KEvent* rendered_event; Manager& manager; std::unique_ptr<Renderer> impl; - Common::ScratchBuffer<u8> tmp_output; - Common::ScratchBuffer<u8> tmp_performance; + Common::ScratchBuffer<u8> output_buffer; + Common::ScratchBuffer<u8> performance_buffer; }; class IAudioDevice final : public ServiceFramework<IAudioDevice> { diff --git a/src/core/hle/service/audio/hwopus.cpp b/src/core/hle/service/audio/hwopus.cpp index c835f6cb7..fa77007f3 100644 --- a/src/core/hle/service/audio/hwopus.cpp +++ b/src/core/hle/service/audio/hwopus.cpp @@ -11,6 +11,7 @@ #include "common/assert.h" #include "common/logging/log.h" +#include "common/scratch_buffer.h" #include "core/hle/service/audio/hwopus.h" #include "core/hle/service/ipc_helpers.h" @@ -68,13 +69,13 @@ private: ExtraBehavior extra_behavior) { u32 consumed = 0; u32 sample_count = 0; - tmp_samples.resize_destructive(ctx.GetWriteBufferNumElements<opus_int16>()); + samples.resize_destructive(ctx.GetWriteBufferNumElements<opus_int16>()); if (extra_behavior == ExtraBehavior::ResetContext) { ResetDecoderContext(); } - if (!DecodeOpusData(consumed, sample_count, ctx.ReadBuffer(), tmp_samples, performance)) { + if (!DecodeOpusData(consumed, sample_count, ctx.ReadBuffer(), samples, performance)) { LOG_ERROR(Audio, "Failed to decode opus data"); IPC::ResponseBuilder rb{ctx, 2}; // TODO(ogniK): Use correct error code @@ -90,7 +91,7 @@ private: if (performance) { rb.Push<u64>(*performance); } - ctx.WriteBuffer(tmp_samples); + ctx.WriteBuffer(samples); } bool DecodeOpusData(u32& consumed, u32& sample_count, std::span<const u8> input, @@ -154,7 +155,7 @@ private: OpusDecoderPtr decoder; u32 sample_rate; u32 channel_count; - Common::ScratchBuffer<opus_int16> tmp_samples; + Common::ScratchBuffer<opus_int16> samples; }; class IHardwareOpusDecoderManager final : public ServiceFramework<IHardwareOpusDecoderManager> { diff --git a/src/core/hle/service/hle_ipc.cpp b/src/core/hle/service/hle_ipc.cpp index 2290df705..f6a1e54f2 100644 --- a/src/core/hle/service/hle_ipc.cpp +++ b/src/core/hle/service/hle_ipc.cpp @@ -329,8 +329,22 @@ std::vector<u8> HLERequestContext::ReadBufferCopy(std::size_t buffer_index) cons } std::span<const u8> HLERequestContext::ReadBuffer(std::size_t buffer_index) const { - static thread_local std::array<Common::ScratchBuffer<u8>, 2> read_buffer_a; - static thread_local std::array<Common::ScratchBuffer<u8>, 2> read_buffer_x; + static thread_local std::array read_buffer_a{ + Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0), + Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0), + }; + static thread_local std::array read_buffer_data_a{ + Common::ScratchBuffer<u8>(), + Common::ScratchBuffer<u8>(), + }; + static thread_local std::array read_buffer_x{ + Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0), + Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0), + }; + static thread_local std::array read_buffer_data_x{ + Common::ScratchBuffer<u8>(), + Common::ScratchBuffer<u8>(), + }; const bool is_buffer_a{BufferDescriptorA().size() > buffer_index && BufferDescriptorA()[buffer_index].Size()}; @@ -339,19 +353,17 @@ std::span<const u8> HLERequestContext::ReadBuffer(std::size_t buffer_index) cons BufferDescriptorA().size() > buffer_index, { return {}; }, "BufferDescriptorA invalid buffer_index {}", buffer_index); auto& read_buffer = read_buffer_a[buffer_index]; - read_buffer.resize_destructive(BufferDescriptorA()[buffer_index].Size()); - memory.ReadBlock(BufferDescriptorA()[buffer_index].Address(), read_buffer.data(), - read_buffer.size()); - return read_buffer; + return read_buffer.Read(BufferDescriptorA()[buffer_index].Address(), + BufferDescriptorA()[buffer_index].Size(), + &read_buffer_data_a[buffer_index]); } else { ASSERT_OR_EXECUTE_MSG( BufferDescriptorX().size() > buffer_index, { return {}; }, "BufferDescriptorX invalid buffer_index {}", buffer_index); auto& read_buffer = read_buffer_x[buffer_index]; - read_buffer.resize_destructive(BufferDescriptorX()[buffer_index].Size()); - memory.ReadBlock(BufferDescriptorX()[buffer_index].Address(), read_buffer.data(), - read_buffer.size()); - return read_buffer; + return read_buffer.Read(BufferDescriptorX()[buffer_index].Address(), + BufferDescriptorX()[buffer_index].Size(), + &read_buffer_data_x[buffer_index]); } } diff --git a/src/core/hle/service/nfc/common/device.cpp b/src/core/hle/service/nfc/common/device.cpp index 5bf289818..2d633b03f 100644 --- a/src/core/hle/service/nfc/common/device.cpp +++ b/src/core/hle/service/nfc/common/device.cpp @@ -66,10 +66,6 @@ NfcDevice::~NfcDevice() { }; void NfcDevice::NpadUpdate(Core::HID::ControllerTriggerType type) { - if (!is_initalized) { - return; - } - if (type == Core::HID::ControllerTriggerType::Connected) { Initialize(); availability_change_event->Signal(); @@ -77,12 +73,12 @@ void NfcDevice::NpadUpdate(Core::HID::ControllerTriggerType type) { } if (type == Core::HID::ControllerTriggerType::Disconnected) { - device_state = DeviceState::Unavailable; + Finalize(); availability_change_event->Signal(); return; } - if (type != Core::HID::ControllerTriggerType::Nfc) { + if (!is_initalized) { return; } @@ -90,6 +86,17 @@ void NfcDevice::NpadUpdate(Core::HID::ControllerTriggerType type) { return; } + // Ensure nfc mode is always active + if (npad_device->GetPollingMode(Core::HID::EmulatedDeviceIndex::RightIndex) == + Common::Input::PollingMode::Active) { + npad_device->SetPollingMode(Core::HID::EmulatedDeviceIndex::RightIndex, + Common::Input::PollingMode::NFC); + } + + if (type != Core::HID::ControllerTriggerType::Nfc) { + return; + } + const auto nfc_status = npad_device->GetNfc(); switch (nfc_status.state) { case Common::Input::NfcState::NewAmiibo: @@ -207,11 +214,14 @@ void NfcDevice::Initialize() { } void NfcDevice::Finalize() { - if (device_state == DeviceState::TagMounted) { - Unmount(); - } - if (device_state == DeviceState::SearchingForTag || device_state == DeviceState::TagRemoved) { - StopDetection(); + if (npad_device->IsConnected()) { + if (device_state == DeviceState::TagMounted) { + Unmount(); + } + if (device_state == DeviceState::SearchingForTag || + device_state == DeviceState::TagRemoved) { + StopDetection(); + } } if (device_state != DeviceState::Unavailable) { diff --git a/src/core/hle/service/nvdrv/nvdrv_interface.cpp b/src/core/hle/service/nvdrv/nvdrv_interface.cpp index 348207e25..c8a880e84 100644 --- a/src/core/hle/service/nvdrv/nvdrv_interface.cpp +++ b/src/core/hle/service/nvdrv/nvdrv_interface.cpp @@ -2,7 +2,6 @@ // SPDX-FileCopyrightText: 2021 Skyline Team and Contributors // SPDX-License-Identifier: GPL-3.0-or-later -#include <cinttypes> #include "common/logging/log.h" #include "core/core.h" #include "core/hle/kernel/k_event.h" @@ -63,12 +62,12 @@ void NVDRV::Ioctl1(HLERequestContext& ctx) { } // Check device - tmp_output.resize_destructive(ctx.GetWriteBufferSize(0)); + output_buffer.resize_destructive(ctx.GetWriteBufferSize(0)); const auto input_buffer = ctx.ReadBuffer(0); - const auto nv_result = nvdrv->Ioctl1(fd, command, input_buffer, tmp_output); + const auto nv_result = nvdrv->Ioctl1(fd, command, input_buffer, output_buffer); if (command.is_out != 0) { - ctx.WriteBuffer(tmp_output); + ctx.WriteBuffer(output_buffer); } IPC::ResponseBuilder rb{ctx, 3}; @@ -90,12 +89,12 @@ void NVDRV::Ioctl2(HLERequestContext& ctx) { const auto input_buffer = ctx.ReadBuffer(0); const auto input_inlined_buffer = ctx.ReadBuffer(1); - tmp_output.resize_destructive(ctx.GetWriteBufferSize(0)); + output_buffer.resize_destructive(ctx.GetWriteBufferSize(0)); const auto nv_result = - nvdrv->Ioctl2(fd, command, input_buffer, input_inlined_buffer, tmp_output); + nvdrv->Ioctl2(fd, command, input_buffer, input_inlined_buffer, output_buffer); if (command.is_out != 0) { - ctx.WriteBuffer(tmp_output); + ctx.WriteBuffer(output_buffer); } IPC::ResponseBuilder rb{ctx, 3}; @@ -116,12 +115,14 @@ void NVDRV::Ioctl3(HLERequestContext& ctx) { } const auto input_buffer = ctx.ReadBuffer(0); - tmp_output.resize_destructive(ctx.GetWriteBufferSize(0)); - tmp_output_inline.resize_destructive(ctx.GetWriteBufferSize(1)); - const auto nv_result = nvdrv->Ioctl3(fd, command, input_buffer, tmp_output, tmp_output_inline); + output_buffer.resize_destructive(ctx.GetWriteBufferSize(0)); + inline_output_buffer.resize_destructive(ctx.GetWriteBufferSize(1)); + + const auto nv_result = + nvdrv->Ioctl3(fd, command, input_buffer, output_buffer, inline_output_buffer); if (command.is_out != 0) { - ctx.WriteBuffer(tmp_output, 0); - ctx.WriteBuffer(tmp_output_inline, 1); + ctx.WriteBuffer(output_buffer, 0); + ctx.WriteBuffer(inline_output_buffer, 1); } IPC::ResponseBuilder rb{ctx, 3}; diff --git a/src/core/hle/service/nvdrv/nvdrv_interface.h b/src/core/hle/service/nvdrv/nvdrv_interface.h index 4b593ff90..6e98115dc 100644 --- a/src/core/hle/service/nvdrv/nvdrv_interface.h +++ b/src/core/hle/service/nvdrv/nvdrv_interface.h @@ -4,6 +4,7 @@ #pragma once #include <memory> + #include "common/scratch_buffer.h" #include "core/hle/service/nvdrv/nvdrv.h" #include "core/hle/service/service.h" @@ -34,8 +35,8 @@ private: u64 pid{}; bool is_initialized{}; - Common::ScratchBuffer<u8> tmp_output; - Common::ScratchBuffer<u8> tmp_output_inline; + Common::ScratchBuffer<u8> output_buffer; + Common::ScratchBuffer<u8> inline_output_buffer; }; } // namespace Service::Nvidia diff --git a/src/core/hle/service/nvnflinger/parcel.h b/src/core/hle/service/nvnflinger/parcel.h index 23ba315a0..e2c9bbd50 100644 --- a/src/core/hle/service/nvnflinger/parcel.h +++ b/src/core/hle/service/nvnflinger/parcel.h @@ -6,6 +6,7 @@ #include <memory> #include <span> #include <vector> + #include <boost/container/small_vector.hpp> #include "common/alignment.h" @@ -148,9 +149,9 @@ public: this->WriteImpl(0U, m_object_buffer); } - std::vector<u8> Serialize() const { - std::vector<u8> output_buffer(sizeof(ParcelHeader) + m_data_buffer.size() + - m_object_buffer.size()); + std::span<u8> Serialize() { + m_output_buffer.resize(sizeof(ParcelHeader) + m_data_buffer.size() + + m_object_buffer.size()); ParcelHeader header{}; header.data_size = static_cast<u32>(m_data_buffer.size()); @@ -158,17 +159,17 @@ public: header.objects_size = static_cast<u32>(m_object_buffer.size()); header.objects_offset = header.data_offset + header.data_size; - std::memcpy(output_buffer.data(), &header, sizeof(header)); - std::ranges::copy(m_data_buffer, output_buffer.data() + header.data_offset); - std::ranges::copy(m_object_buffer, output_buffer.data() + header.objects_offset); + std::memcpy(m_output_buffer.data(), &header, sizeof(ParcelHeader)); + std::ranges::copy(m_data_buffer, m_output_buffer.data() + header.data_offset); + std::ranges::copy(m_object_buffer, m_output_buffer.data() + header.objects_offset); - return output_buffer; + return m_output_buffer; } private: - template <typename T> + template <typename T, size_t BufferSize> requires(std::is_trivially_copyable_v<T>) - void WriteImpl(const T& val, boost::container::small_vector<u8, 0x200>& buffer) { + void WriteImpl(const T& val, boost::container::small_vector<u8, BufferSize>& buffer) { const size_t aligned_size = Common::AlignUp(sizeof(T), 4); const size_t old_size = buffer.size(); buffer.resize(old_size + aligned_size); @@ -177,8 +178,9 @@ private: } private: - boost::container::small_vector<u8, 0x200> m_data_buffer; - boost::container::small_vector<u8, 0x200> m_object_buffer; + boost::container::small_vector<u8, 0x1B0> m_data_buffer; + boost::container::small_vector<u8, 0x40> m_object_buffer; + boost::container::small_vector<u8, 0x200> m_output_buffer; }; } // namespace Service::android diff --git a/src/core/loader/loader.h b/src/core/loader/loader.h index 7b43f70ed..7a2a52fd4 100644 --- a/src/core/loader/loader.h +++ b/src/core/loader/loader.h @@ -79,6 +79,8 @@ enum class ResultStatus : u16 { ErrorBadPFSHeader, ErrorIncorrectPFSFileSize, ErrorBadNCAHeader, + ErrorCompressedNCA, + ErrorSparseNCA, ErrorMissingProductionKeyFile, ErrorMissingHeaderKey, ErrorIncorrectHeaderKey, diff --git a/src/core/memory.cpp b/src/core/memory.cpp index 514ba0d66..805963178 100644 --- a/src/core/memory.cpp +++ b/src/core/memory.cpp @@ -3,6 +3,7 @@ #include <algorithm> #include <cstring> +#include <span> #include "common/assert.h" #include "common/atomic_ops.h" @@ -13,6 +14,7 @@ #include "common/swap.h" #include "core/core.h" #include "core/device_memory.h" +#include "core/gpu_dirty_memory_manager.h" #include "core/hardware_properties.h" #include "core/hle/kernel/k_page_table.h" #include "core/hle/kernel/k_process.h" @@ -264,6 +266,22 @@ struct Memory::Impl { ReadBlockImpl<true>(*system.ApplicationProcess(), src_addr, dest_buffer, size); } + const u8* GetSpan(const VAddr src_addr, const std::size_t size) const { + if (current_page_table->blocks[src_addr >> YUZU_PAGEBITS] == + current_page_table->blocks[(src_addr + size) >> YUZU_PAGEBITS]) { + return GetPointerSilent(src_addr); + } + return nullptr; + } + + u8* GetSpan(const VAddr src_addr, const std::size_t size) { + if (current_page_table->blocks[src_addr >> YUZU_PAGEBITS] == + current_page_table->blocks[(src_addr + size) >> YUZU_PAGEBITS]) { + return GetPointerSilent(src_addr); + } + return nullptr; + } + template <bool UNSAFE> void WriteBlockImpl(const Kernel::KProcess& process, const Common::ProcessAddress dest_addr, const void* src_buffer, const std::size_t size) { @@ -557,7 +575,7 @@ struct Memory::Impl { } } - const Common::ProcessAddress end = base + size; + const auto end = base + size; ASSERT_MSG(end <= page_table.pointers.size(), "out of range mapping at {:016X}", base + page_table.pointers.size()); @@ -568,14 +586,18 @@ struct Memory::Impl { while (base != end) { page_table.pointers[base].Store(nullptr, type); page_table.backing_addr[base] = 0; - + page_table.blocks[base] = 0; base += 1; } } else { + auto orig_base = base; while (base != end) { - page_table.pointers[base].Store( - system.DeviceMemory().GetPointer<u8>(target) - (base << YUZU_PAGEBITS), type); - page_table.backing_addr[base] = GetInteger(target) - (base << YUZU_PAGEBITS); + auto host_ptr = + system.DeviceMemory().GetPointer<u8>(target) - (base << YUZU_PAGEBITS); + auto backing = GetInteger(target) - (base << YUZU_PAGEBITS); + page_table.pointers[base].Store(host_ptr, type); + page_table.backing_addr[base] = backing; + page_table.blocks[base] = orig_base << YUZU_PAGEBITS; ASSERT_MSG(page_table.pointers[base].Pointer(), "memory mapping base yield a nullptr within the table"); @@ -678,7 +700,7 @@ struct Memory::Impl { LOG_ERROR(HW_Memory, "Unmapped Write{} @ 0x{:016X} = 0x{:016X}", sizeof(T) * 8, GetInteger(vaddr), static_cast<u64>(data)); }, - [&]() { system.GPU().InvalidateRegion(GetInteger(vaddr), sizeof(T)); }); + [&]() { HandleRasterizerWrite(GetInteger(vaddr), sizeof(T)); }); if (ptr) { std::memcpy(ptr, &data, sizeof(T)); } @@ -692,7 +714,7 @@ struct Memory::Impl { LOG_ERROR(HW_Memory, "Unmapped WriteExclusive{} @ 0x{:016X} = 0x{:016X}", sizeof(T) * 8, GetInteger(vaddr), static_cast<u64>(data)); }, - [&]() { system.GPU().InvalidateRegion(GetInteger(vaddr), sizeof(T)); }); + [&]() { HandleRasterizerWrite(GetInteger(vaddr), sizeof(T)); }); if (ptr) { const auto volatile_pointer = reinterpret_cast<volatile T*>(ptr); return Common::AtomicCompareAndSwap(volatile_pointer, data, expected); @@ -707,7 +729,7 @@ struct Memory::Impl { LOG_ERROR(HW_Memory, "Unmapped WriteExclusive128 @ 0x{:016X} = 0x{:016X}{:016X}", GetInteger(vaddr), static_cast<u64>(data[1]), static_cast<u64>(data[0])); }, - [&]() { system.GPU().InvalidateRegion(GetInteger(vaddr), sizeof(u128)); }); + [&]() { HandleRasterizerWrite(GetInteger(vaddr), sizeof(u128)); }); if (ptr) { const auto volatile_pointer = reinterpret_cast<volatile u64*>(ptr); return Common::AtomicCompareAndSwap(volatile_pointer, data, expected); @@ -717,7 +739,7 @@ struct Memory::Impl { void HandleRasterizerDownload(VAddr address, size_t size) { const size_t core = system.GetCurrentHostThreadID(); - auto& current_area = rasterizer_areas[core]; + auto& current_area = rasterizer_read_areas[core]; const VAddr end_address = address + size; if (current_area.start_address <= address && end_address <= current_area.end_address) [[likely]] { @@ -726,9 +748,39 @@ struct Memory::Impl { current_area = system.GPU().OnCPURead(address, size); } - Common::PageTable* current_page_table = nullptr; - std::array<VideoCore::RasterizerDownloadArea, Core::Hardware::NUM_CPU_CORES> rasterizer_areas{}; + void HandleRasterizerWrite(VAddr address, size_t size) { + const size_t core = system.GetCurrentHostThreadID(); + auto& current_area = rasterizer_write_areas[core]; + VAddr subaddress = address >> YUZU_PAGEBITS; + bool do_collection = current_area.last_address == subaddress; + if (!do_collection) [[unlikely]] { + do_collection = system.GPU().OnCPUWrite(address, size); + if (!do_collection) { + return; + } + current_area.last_address = subaddress; + } + gpu_dirty_managers[core].Collect(address, size); + } + + struct GPUDirtyState { + VAddr last_address; + }; + + void InvalidateRegion(Common::ProcessAddress dest_addr, size_t size) { + system.GPU().InvalidateRegion(GetInteger(dest_addr), size); + } + + void FlushRegion(Common::ProcessAddress dest_addr, size_t size) { + system.GPU().FlushRegion(GetInteger(dest_addr), size); + } + Core::System& system; + Common::PageTable* current_page_table = nullptr; + std::array<VideoCore::RasterizerDownloadArea, Core::Hardware::NUM_CPU_CORES> + rasterizer_read_areas{}; + std::array<GPUDirtyState, Core::Hardware::NUM_CPU_CORES> rasterizer_write_areas{}; + std::span<Core::GPUDirtyMemoryManager> gpu_dirty_managers; }; Memory::Memory(Core::System& system_) : system{system_} { @@ -857,6 +909,14 @@ void Memory::ReadBlockUnsafe(const Common::ProcessAddress src_addr, void* dest_b impl->ReadBlockUnsafe(src_addr, dest_buffer, size); } +const u8* Memory::GetSpan(const VAddr src_addr, const std::size_t size) const { + return impl->GetSpan(src_addr, size); +} + +u8* Memory::GetSpan(const VAddr src_addr, const std::size_t size) { + return impl->GetSpan(src_addr, size); +} + void Memory::WriteBlock(const Common::ProcessAddress dest_addr, const void* src_buffer, const std::size_t size) { impl->WriteBlock(dest_addr, src_buffer, size); @@ -876,6 +936,10 @@ void Memory::ZeroBlock(Common::ProcessAddress dest_addr, const std::size_t size) impl->ZeroBlock(*system.ApplicationProcess(), dest_addr, size); } +void Memory::SetGPUDirtyManagers(std::span<Core::GPUDirtyMemoryManager> managers) { + impl->gpu_dirty_managers = managers; +} + Result Memory::InvalidateDataCache(Common::ProcessAddress dest_addr, const std::size_t size) { return impl->InvalidateDataCache(*system.ApplicationProcess(), dest_addr, size); } @@ -896,4 +960,12 @@ void Memory::MarkRegionDebug(Common::ProcessAddress vaddr, u64 size, bool debug) impl->MarkRegionDebug(GetInteger(vaddr), size, debug); } +void Memory::InvalidateRegion(Common::ProcessAddress dest_addr, size_t size) { + impl->InvalidateRegion(dest_addr, size); +} + +void Memory::FlushRegion(Common::ProcessAddress dest_addr, size_t size) { + impl->FlushRegion(dest_addr, size); +} + } // namespace Core::Memory diff --git a/src/core/memory.h b/src/core/memory.h index 72a0be813..ea33c769c 100644 --- a/src/core/memory.h +++ b/src/core/memory.h @@ -5,7 +5,12 @@ #include <cstddef> #include <memory> +#include <optional> +#include <span> #include <string> +#include <vector> + +#include "common/scratch_buffer.h" #include "common/typed_address.h" #include "core/hle/result.h" @@ -15,13 +20,18 @@ struct PageTable; namespace Core { class System; -} +class GPUDirtyMemoryManager; +} // namespace Core namespace Kernel { class PhysicalMemory; class KProcess; } // namespace Kernel +namespace Tegra { +class MemoryManager; +} + namespace Core::Memory { /** @@ -341,6 +351,9 @@ public: */ void ReadBlockUnsafe(Common::ProcessAddress src_addr, void* dest_buffer, std::size_t size); + const u8* GetSpan(const VAddr src_addr, const std::size_t size) const; + u8* GetSpan(const VAddr src_addr, const std::size_t size); + /** * Writes a range of bytes into the current process' address space at the specified * virtual address. @@ -458,6 +471,10 @@ public: */ void MarkRegionDebug(Common::ProcessAddress vaddr, u64 size, bool debug); + void SetGPUDirtyManagers(std::span<Core::GPUDirtyMemoryManager> managers); + void InvalidateRegion(Common::ProcessAddress dest_addr, size_t size); + void FlushRegion(Common::ProcessAddress dest_addr, size_t size); + private: Core::System& system; @@ -465,4 +482,203 @@ private: std::unique_ptr<Impl> impl; }; +enum GuestMemoryFlags : u32 { + Read = 1 << 0, + Write = 1 << 1, + Safe = 1 << 2, + Cached = 1 << 3, + + SafeRead = Read | Safe, + SafeWrite = Write | Safe, + SafeReadWrite = SafeRead | SafeWrite, + SafeReadCachedWrite = SafeReadWrite | Cached, + + UnsafeRead = Read, + UnsafeWrite = Write, + UnsafeReadWrite = UnsafeRead | UnsafeWrite, + UnsafeReadCachedWrite = UnsafeReadWrite | Cached, +}; + +namespace { +template <typename M, typename T, GuestMemoryFlags FLAGS> +class GuestMemory { + using iterator = T*; + using const_iterator = const T*; + using value_type = T; + using element_type = T; + using iterator_category = std::contiguous_iterator_tag; + +public: + GuestMemory() = delete; + explicit GuestMemory(M& memory_, u64 addr_, std::size_t size_, + Common::ScratchBuffer<T>* backup = nullptr) + : memory{memory_}, addr{addr_}, size{size_} { + static_assert(FLAGS & GuestMemoryFlags::Read || FLAGS & GuestMemoryFlags::Write); + if constexpr (FLAGS & GuestMemoryFlags::Read) { + Read(addr, size, backup); + } + } + + ~GuestMemory() = default; + + T* data() noexcept { + return data_span.data(); + } + + const T* data() const noexcept { + return data_span.data(); + } + + [[nodiscard]] T* begin() noexcept { + return data(); + } + + [[nodiscard]] const T* begin() const noexcept { + return data(); + } + + [[nodiscard]] T* end() noexcept { + return data() + size; + } + + [[nodiscard]] const T* end() const noexcept { + return data() + size; + } + + T& operator[](size_t index) noexcept { + return data_span[index]; + } + + const T& operator[](size_t index) const noexcept { + return data_span[index]; + } + + void SetAddressAndSize(u64 addr_, std::size_t size_) noexcept { + addr = addr_; + size = size_; + addr_changed = true; + } + + std::span<T> Read(u64 addr_, std::size_t size_, + Common::ScratchBuffer<T>* backup = nullptr) noexcept { + addr = addr_; + size = size_; + if (size == 0) { + is_data_copy = true; + return {}; + } + + if (TrySetSpan()) { + if constexpr (FLAGS & GuestMemoryFlags::Safe) { + memory.FlushRegion(addr, size * sizeof(T)); + } + } else { + if (backup) { + backup->resize_destructive(size); + data_span = *backup; + } else { + data_copy.resize(size); + data_span = std::span(data_copy); + } + is_data_copy = true; + span_valid = true; + if constexpr (FLAGS & GuestMemoryFlags::Safe) { + memory.ReadBlock(addr, data_span.data(), size * sizeof(T)); + } else { + memory.ReadBlockUnsafe(addr, data_span.data(), size * sizeof(T)); + } + } + return data_span; + } + + void Write(std::span<T> write_data) noexcept { + if constexpr (FLAGS & GuestMemoryFlags::Cached) { + memory.WriteBlockCached(addr, write_data.data(), size * sizeof(T)); + } else if constexpr (FLAGS & GuestMemoryFlags::Safe) { + memory.WriteBlock(addr, write_data.data(), size * sizeof(T)); + } else { + memory.WriteBlockUnsafe(addr, write_data.data(), size * sizeof(T)); + } + } + + bool TrySetSpan() noexcept { + if (u8* ptr = memory.GetSpan(addr, size * sizeof(T)); ptr) { + data_span = {reinterpret_cast<T*>(ptr), size}; + span_valid = true; + return true; + } + return false; + } + +protected: + bool IsDataCopy() const noexcept { + return is_data_copy; + } + + bool AddressChanged() const noexcept { + return addr_changed; + } + + M& memory; + u64 addr; + size_t size; + std::span<T> data_span{}; + std::vector<T> data_copy; + bool span_valid{false}; + bool is_data_copy{false}; + bool addr_changed{false}; +}; + +template <typename M, typename T, GuestMemoryFlags FLAGS> +class GuestMemoryScoped : public GuestMemory<M, T, FLAGS> { +public: + GuestMemoryScoped() = delete; + explicit GuestMemoryScoped(M& memory_, u64 addr_, std::size_t size_, + Common::ScratchBuffer<T>* backup = nullptr) + : GuestMemory<M, T, FLAGS>(memory_, addr_, size_, backup) { + if constexpr (!(FLAGS & GuestMemoryFlags::Read)) { + if (!this->TrySetSpan()) { + if (backup) { + this->data_span = *backup; + this->span_valid = true; + this->is_data_copy = true; + } + } + } + } + + ~GuestMemoryScoped() { + if constexpr (FLAGS & GuestMemoryFlags::Write) { + if (this->size == 0) [[unlikely]] { + return; + } + + if (this->AddressChanged() || this->IsDataCopy()) { + ASSERT(this->span_valid); + if constexpr (FLAGS & GuestMemoryFlags::Cached) { + this->memory.WriteBlockCached(this->addr, this->data_span.data(), + this->size * sizeof(T)); + } else if constexpr (FLAGS & GuestMemoryFlags::Safe) { + this->memory.WriteBlock(this->addr, this->data_span.data(), + this->size * sizeof(T)); + } else { + this->memory.WriteBlockUnsafe(this->addr, this->data_span.data(), + this->size * sizeof(T)); + } + } else if constexpr (FLAGS & GuestMemoryFlags::Safe) { + this->memory.InvalidateRegion(this->addr, this->size * sizeof(T)); + } + } + } +}; +} // namespace + +template <typename T, GuestMemoryFlags FLAGS> +using CpuGuestMemory = GuestMemory<Memory, T, FLAGS>; +template <typename T, GuestMemoryFlags FLAGS> +using CpuGuestMemoryScoped = GuestMemoryScoped<Memory, T, FLAGS>; +template <typename T, GuestMemoryFlags FLAGS> +using GpuGuestMemory = GuestMemory<Tegra::MemoryManager, T, FLAGS>; +template <typename T, GuestMemoryFlags FLAGS> +using GpuGuestMemoryScoped = GuestMemoryScoped<Tegra::MemoryManager, T, FLAGS>; } // namespace Core::Memory |