diff options
Diffstat (limited to 'src')
70 files changed, 1730 insertions, 660 deletions
diff --git a/src/common/host_memory.cpp b/src/common/host_memory.cpp index 28949fe5e..c465cfc14 100644 --- a/src/common/host_memory.cpp +++ b/src/common/host_memory.cpp @@ -327,8 +327,8 @@ private: bool IsNiechePlaceholder(size_t virtual_offset, size_t length) const { const auto it = placeholders.upper_bound({virtual_offset, virtual_offset + length}); if (it != placeholders.end() && it->lower() == virtual_offset + length) { - const bool is_root = it == placeholders.begin() && virtual_offset == 0; - return is_root || std::prev(it)->upper() == virtual_offset; + return it == placeholders.begin() ? virtual_offset == 0 + : std::prev(it)->upper() == virtual_offset; } return false; } diff --git a/src/common/logging/backend.cpp b/src/common/logging/backend.cpp index c51c05b28..f1c9ed6c4 100644 --- a/src/common/logging/backend.cpp +++ b/src/common/logging/backend.cpp @@ -218,19 +218,17 @@ private: Impl(const std::filesystem::path& file_backend_filename, const Filter& filter_) : filter{filter_}, file_backend{file_backend_filename} {} - ~Impl() { - StopBackendThread(); - } + ~Impl() = default; void StartBackendThread() { - backend_thread = std::thread([this] { + backend_thread = std::jthread([this](std::stop_token stop_token) { Common::SetCurrentThreadName("yuzu:Log"); Entry entry; const auto write_logs = [this, &entry]() { ForEachBackend([&entry](Backend& backend) { backend.Write(entry); }); }; - while (!stop.stop_requested()) { - entry = message_queue.PopWait(stop.get_token()); + while (!stop_token.stop_requested()) { + entry = message_queue.PopWait(stop_token); if (entry.filename != nullptr) { write_logs(); } @@ -244,11 +242,6 @@ private: }); } - void StopBackendThread() { - stop.request_stop(); - backend_thread.join(); - } - Entry CreateEntry(Class log_class, Level log_level, const char* filename, unsigned int line_nr, const char* function, std::string&& message) const { using std::chrono::duration_cast; @@ -283,8 +276,7 @@ private: ColorConsoleBackend color_console_backend{}; FileBackend file_backend; - std::stop_source stop; - std::thread backend_thread; + std::jthread backend_thread; MPSCQueue<Entry, true> message_queue{}; std::chrono::steady_clock::time_point time_origin{std::chrono::steady_clock::now()}; }; diff --git a/src/common/page_table.cpp b/src/common/page_table.cpp index 9fffd816f..4817b09f9 100644 --- a/src/common/page_table.cpp +++ b/src/common/page_table.cpp @@ -10,11 +10,65 @@ PageTable::PageTable() = default; PageTable::~PageTable() noexcept = default; -void PageTable::Resize(size_t address_space_width_in_bits, size_t page_size_in_bits) { - const size_t num_page_table_entries{1ULL << (address_space_width_in_bits - page_size_in_bits)}; +bool PageTable::BeginTraversal(TraversalEntry& out_entry, TraversalContext& out_context, + u64 address) const { + // Setup invalid defaults. + out_entry.phys_addr = 0; + out_entry.block_size = page_size; + out_context.next_page = 0; + + // Validate that we can read the actual entry. + const auto page = address / page_size; + if (page >= backing_addr.size()) { + return false; + } + + // Validate that the entry is mapped. + const auto phys_addr = backing_addr[page]; + if (phys_addr == 0) { + return false; + } + + // Populate the results. + out_entry.phys_addr = phys_addr + address; + out_context.next_page = page + 1; + out_context.next_offset = address + page_size; + + return true; +} + +bool PageTable::ContinueTraversal(TraversalEntry& out_entry, TraversalContext& context) const { + // Setup invalid defaults. + out_entry.phys_addr = 0; + out_entry.block_size = page_size; + + // Validate that we can read the actual entry. + const auto page = context.next_page; + if (page >= backing_addr.size()) { + return false; + } + + // Validate that the entry is mapped. + const auto phys_addr = backing_addr[page]; + if (phys_addr == 0) { + return false; + } + + // Populate the results. + out_entry.phys_addr = phys_addr + context.next_offset; + context.next_page = page + 1; + context.next_offset += page_size; + + return true; +} + +void PageTable::Resize(std::size_t address_space_width_in_bits, std::size_t page_size_in_bits) { + const std::size_t num_page_table_entries{1ULL + << (address_space_width_in_bits - page_size_in_bits)}; pointers.resize(num_page_table_entries); backing_addr.resize(num_page_table_entries); current_address_space_width_in_bits = address_space_width_in_bits; + page_size = 1ULL << page_size_in_bits; } } // namespace Common diff --git a/src/common/page_table.h b/src/common/page_table.h index 8267e8b4d..82d91e9f3 100644 --- a/src/common/page_table.h +++ b/src/common/page_table.h @@ -27,6 +27,16 @@ enum class PageType : u8 { * mimics the way a real CPU page table works. */ struct PageTable { + struct TraversalEntry { + u64 phys_addr{}; + std::size_t block_size{}; + }; + + struct TraversalContext { + u64 next_page{}; + u64 next_offset{}; + }; + /// Number of bits reserved for attribute tagging. /// This can be at most the guaranteed alignment of the pointers in the page table. static constexpr int ATTRIBUTE_BITS = 2; @@ -89,6 +99,10 @@ struct PageTable { PageTable(PageTable&&) noexcept = default; PageTable& operator=(PageTable&&) noexcept = default; + bool BeginTraversal(TraversalEntry& out_entry, TraversalContext& out_context, + u64 address) const; + bool ContinueTraversal(TraversalEntry& out_entry, TraversalContext& context) const; + /** * Resizes the page table to be able to accommodate enough pages within * a given address space. @@ -96,9 +110,9 @@ struct PageTable { * @param address_space_width_in_bits The address size width in bits. * @param page_size_in_bits The page size in bits. */ - void Resize(size_t address_space_width_in_bits, size_t page_size_in_bits); + void Resize(std::size_t address_space_width_in_bits, std::size_t page_size_in_bits); - size_t GetAddressSpaceBits() const { + std::size_t GetAddressSpaceBits() const { return current_address_space_width_in_bits; } @@ -110,9 +124,11 @@ struct PageTable { VirtualBuffer<u64> backing_addr; - size_t current_address_space_width_in_bits; + std::size_t current_address_space_width_in_bits{}; + + u8* fastmem_arena{}; - u8* fastmem_arena; + std::size_t page_size{}; }; } // namespace Common diff --git a/src/common/settings.cpp b/src/common/settings.cpp index 6964a8273..877e0faa4 100644 --- a/src/common/settings.cpp +++ b/src/common/settings.cpp @@ -167,6 +167,7 @@ void RestoreGlobalState(bool is_powered_on) { // Core values.use_multi_core.SetGlobal(true); + values.use_extended_memory_layout.SetGlobal(true); // CPU values.cpu_accuracy.SetGlobal(true); @@ -175,6 +176,7 @@ void RestoreGlobalState(bool is_powered_on) { values.cpuopt_unsafe_ignore_standard_fpcr.SetGlobal(true); values.cpuopt_unsafe_inaccurate_nan.SetGlobal(true); values.cpuopt_unsafe_fastmem_check.SetGlobal(true); + values.cpuopt_unsafe_ignore_global_monitor.SetGlobal(true); // Renderer values.renderer_backend.SetGlobal(true); diff --git a/src/common/settings.h b/src/common/settings.h index 9bee6e10f..a37d83fb3 100644 --- a/src/common/settings.h +++ b/src/common/settings.h @@ -466,6 +466,7 @@ struct Values { // Core Setting<bool> use_multi_core{true, "use_multi_core"}; + Setting<bool> use_extended_memory_layout{false, "use_extended_memory_layout"}; // Cpu RangedSetting<CPUAccuracy> cpu_accuracy{CPUAccuracy::Auto, CPUAccuracy::Auto, @@ -483,12 +484,15 @@ struct Values { BasicSetting<bool> cpuopt_misc_ir{true, "cpuopt_misc_ir"}; BasicSetting<bool> cpuopt_reduce_misalign_checks{true, "cpuopt_reduce_misalign_checks"}; BasicSetting<bool> cpuopt_fastmem{true, "cpuopt_fastmem"}; + BasicSetting<bool> cpuopt_fastmem_exclusives{true, "cpuopt_fastmem_exclusives"}; + BasicSetting<bool> cpuopt_recompile_exclusives{true, "cpuopt_recompile_exclusives"}; Setting<bool> cpuopt_unsafe_unfuse_fma{true, "cpuopt_unsafe_unfuse_fma"}; Setting<bool> cpuopt_unsafe_reduce_fp_error{true, "cpuopt_unsafe_reduce_fp_error"}; Setting<bool> cpuopt_unsafe_ignore_standard_fpcr{true, "cpuopt_unsafe_ignore_standard_fpcr"}; Setting<bool> cpuopt_unsafe_inaccurate_nan{true, "cpuopt_unsafe_inaccurate_nan"}; Setting<bool> cpuopt_unsafe_fastmem_check{true, "cpuopt_unsafe_fastmem_check"}; + Setting<bool> cpuopt_unsafe_ignore_global_monitor{true, "cpuopt_unsafe_ignore_global_monitor"}; // Renderer RangedSetting<RendererBackend> renderer_backend{ diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt index 0c10cd019..5db6a1b3a 100644 --- a/src/core/CMakeLists.txt +++ b/src/core/CMakeLists.txt @@ -152,6 +152,7 @@ add_library(core STATIC hle/api_version.h hle/ipc.h hle/ipc_helpers.h + hle/kernel/board/nintendo/nx/k_memory_layout.h hle/kernel/board/nintendo/nx/k_system_control.cpp hle/kernel/board/nintendo/nx/k_system_control.h hle/kernel/board/nintendo/nx/secure_monitor.h @@ -164,6 +165,7 @@ add_library(core STATIC hle/kernel/hle_ipc.h hle/kernel/init/init_slab_setup.cpp hle/kernel/init/init_slab_setup.h + hle/kernel/initial_process.h hle/kernel/k_address_arbiter.cpp hle/kernel/k_address_arbiter.h hle/kernel/k_address_space_info.cpp diff --git a/src/core/arm/dynarmic/arm_dynarmic_32.cpp b/src/core/arm/dynarmic/arm_dynarmic_32.cpp index b0d89c539..286976623 100644 --- a/src/core/arm/dynarmic/arm_dynarmic_32.cpp +++ b/src/core/arm/dynarmic/arm_dynarmic_32.cpp @@ -137,6 +137,8 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable* config.page_table_pointer_mask_bits = Common::PageTable::ATTRIBUTE_BITS; config.detect_misaligned_access_via_page_table = 16 | 32 | 64 | 128; config.only_detect_misalignment_via_page_table_on_page_boundary = true; + config.fastmem_exclusive_access = true; + config.recompile_on_exclusive_fastmem_failure = true; // Multi-process state config.processor_id = core_index; @@ -178,6 +180,12 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable* if (!Settings::values.cpuopt_fastmem) { config.fastmem_pointer = nullptr; } + if (!Settings::values.cpuopt_fastmem_exclusives) { + config.fastmem_exclusive_access = false; + } + if (!Settings::values.cpuopt_recompile_exclusives) { + config.recompile_on_exclusive_fastmem_failure = false; + } } // Unsafe optimizations @@ -195,6 +203,9 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable* if (Settings::values.cpuopt_unsafe_inaccurate_nan) { config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN; } + if (Settings::values.cpuopt_unsafe_ignore_global_monitor) { + config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor; + } } // Curated optimizations @@ -203,6 +214,7 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable* config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA; config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreStandardFPCRValue; config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN; + config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor; } return std::make_unique<Dynarmic::A32::Jit>(config); diff --git a/src/core/arm/dynarmic/arm_dynarmic_64.cpp b/src/core/arm/dynarmic/arm_dynarmic_64.cpp index 56836bd05..d96226c41 100644 --- a/src/core/arm/dynarmic/arm_dynarmic_64.cpp +++ b/src/core/arm/dynarmic/arm_dynarmic_64.cpp @@ -185,6 +185,9 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable* config.fastmem_pointer = page_table->fastmem_arena; config.fastmem_address_space_bits = address_space_bits; config.silently_mirror_fastmem = false; + + config.fastmem_exclusive_access = true; + config.recompile_on_exclusive_fastmem_failure = true; } // Multi-process state @@ -237,6 +240,12 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable* if (!Settings::values.cpuopt_fastmem) { config.fastmem_pointer = nullptr; } + if (!Settings::values.cpuopt_fastmem_exclusives) { + config.fastmem_exclusive_access = false; + } + if (!Settings::values.cpuopt_recompile_exclusives) { + config.recompile_on_exclusive_fastmem_failure = false; + } } // Unsafe optimizations @@ -254,6 +263,9 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable* if (Settings::values.cpuopt_unsafe_fastmem_check) { config.fastmem_address_space_bits = 64; } + if (Settings::values.cpuopt_unsafe_ignore_global_monitor) { + config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor; + } } // Curated optimizations @@ -262,6 +274,7 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable* config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_UnfuseFMA; config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN; config.fastmem_address_space_bits = 64; + config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_IgnoreGlobalMonitor; } return std::make_shared<Dynarmic::A64::Jit>(config); diff --git a/src/core/arm/dynarmic/arm_exclusive_monitor.cpp b/src/core/arm/dynarmic/arm_exclusive_monitor.cpp index 397d054a8..ea6b224e0 100644 --- a/src/core/arm/dynarmic/arm_exclusive_monitor.cpp +++ b/src/core/arm/dynarmic/arm_exclusive_monitor.cpp @@ -37,8 +37,8 @@ u128 DynarmicExclusiveMonitor::ExclusiveRead128(std::size_t core_index, VAddr ad }); } -void DynarmicExclusiveMonitor::ClearExclusive() { - monitor.Clear(); +void DynarmicExclusiveMonitor::ClearExclusive(std::size_t core_index) { + monitor.ClearProcessor(core_index); } bool DynarmicExclusiveMonitor::ExclusiveWrite8(std::size_t core_index, VAddr vaddr, u8 value) { diff --git a/src/core/arm/dynarmic/arm_exclusive_monitor.h b/src/core/arm/dynarmic/arm_exclusive_monitor.h index 265c4ecef..5a15b43ef 100644 --- a/src/core/arm/dynarmic/arm_exclusive_monitor.h +++ b/src/core/arm/dynarmic/arm_exclusive_monitor.h @@ -29,7 +29,7 @@ public: u32 ExclusiveRead32(std::size_t core_index, VAddr addr) override; u64 ExclusiveRead64(std::size_t core_index, VAddr addr) override; u128 ExclusiveRead128(std::size_t core_index, VAddr addr) override; - void ClearExclusive() override; + void ClearExclusive(std::size_t core_index) override; bool ExclusiveWrite8(std::size_t core_index, VAddr vaddr, u8 value) override; bool ExclusiveWrite16(std::size_t core_index, VAddr vaddr, u16 value) override; diff --git a/src/core/arm/exclusive_monitor.h b/src/core/arm/exclusive_monitor.h index 62f6e6023..9914ca3da 100644 --- a/src/core/arm/exclusive_monitor.h +++ b/src/core/arm/exclusive_monitor.h @@ -23,7 +23,7 @@ public: virtual u32 ExclusiveRead32(std::size_t core_index, VAddr addr) = 0; virtual u64 ExclusiveRead64(std::size_t core_index, VAddr addr) = 0; virtual u128 ExclusiveRead128(std::size_t core_index, VAddr addr) = 0; - virtual void ClearExclusive() = 0; + virtual void ClearExclusive(std::size_t core_index) = 0; virtual bool ExclusiveWrite8(std::size_t core_index, VAddr vaddr, u8 value) = 0; virtual bool ExclusiveWrite16(std::size_t core_index, VAddr vaddr, u16 value) = 0; diff --git a/src/core/core.cpp b/src/core/core.cpp index 3f9a7f44b..b0cfee3ee 100644 --- a/src/core/core.cpp +++ b/src/core/core.cpp @@ -28,7 +28,9 @@ #include "core/file_sys/vfs_real.h" #include "core/hardware_interrupt_manager.h" #include "core/hid/hid_core.h" +#include "core/hle/kernel/k_memory_manager.h" #include "core/hle/kernel/k_process.h" +#include "core/hle/kernel/k_resource_limit.h" #include "core/hle/kernel/k_scheduler.h" #include "core/hle/kernel/kernel.h" #include "core/hle/kernel/physical_core.h" @@ -252,9 +254,16 @@ struct System::Impl { } telemetry_session->AddInitialInfo(*app_loader, fs_controller, *content_provider); + + // Create a resource limit for the process. + const auto physical_memory_size = + kernel.MemoryManager().GetSize(Kernel::KMemoryManager::Pool::Application); + auto* resource_limit = Kernel::CreateResourceLimitForProcess(system, physical_memory_size); + + // Create the process. auto main_process = Kernel::KProcess::Create(system.Kernel()); ASSERT(Kernel::KProcess::Initialize(main_process, system, "main", - Kernel::KProcess::ProcessType::Userland) + Kernel::KProcess::ProcessType::Userland, resource_limit) .IsSuccess()); const auto [load_result, load_parameters] = app_loader->Load(*main_process, system); if (load_result != Loader::ResultStatus::Success) { diff --git a/src/core/device_memory.cpp b/src/core/device_memory.cpp index f19c0515f..e6bc63086 100644 --- a/src/core/device_memory.cpp +++ b/src/core/device_memory.cpp @@ -3,10 +3,13 @@ // Refer to the license.txt file included. #include "core/device_memory.h" +#include "hle/kernel/board/nintendo/nx/k_system_control.h" namespace Core { -DeviceMemory::DeviceMemory() : buffer{DramMemoryMap::Size, 1ULL << 39} {} +DeviceMemory::DeviceMemory() + : buffer{Kernel::Board::Nintendo::Nx::KSystemControl::Init::GetIntendedMemorySize(), + 1ULL << 39} {} DeviceMemory::~DeviceMemory() = default; } // namespace Core diff --git a/src/core/device_memory.h b/src/core/device_memory.h index c4d17705f..daeb551ea 100644 --- a/src/core/device_memory.h +++ b/src/core/device_memory.h @@ -12,12 +12,8 @@ namespace Core { namespace DramMemoryMap { enum : u64 { Base = 0x80000000ULL, - Size = 0x100000000ULL, - End = Base + Size, KernelReserveBase = Base + 0x60000, SlabHeapBase = KernelReserveBase + 0x85000, - SlapHeapSize = 0xa21000, - SlabHeapEnd = SlabHeapBase + SlapHeapSize, }; }; // namespace DramMemoryMap diff --git a/src/core/hle/kernel/board/nintendo/nx/k_memory_layout.h b/src/core/hle/kernel/board/nintendo/nx/k_memory_layout.h new file mode 100644 index 000000000..01e225088 --- /dev/null +++ b/src/core/hle/kernel/board/nintendo/nx/k_memory_layout.h @@ -0,0 +1,13 @@ +// Copyright 2022 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include "common/common_types.h" + +namespace Kernel { + +constexpr inline PAddr MainMemoryAddress = 0x80000000; + +} // namespace Kernel diff --git a/src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp b/src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp index 6f335c251..8027bec00 100644 --- a/src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp +++ b/src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp @@ -5,6 +5,7 @@ #include <random> #include "common/literals.h" +#include "common/settings.h" #include "core/hle/kernel/board/nintendo/nx/k_system_control.h" #include "core/hle/kernel/board/nintendo/nx/secure_monitor.h" @@ -28,33 +29,20 @@ namespace { using namespace Common::Literals; -u32 GetMemoryModeForInit() { - return 0x01; -} - u32 GetMemorySizeForInit() { - return 0; + return Settings::values.use_extended_memory_layout ? Smc::MemorySize_6GB : Smc::MemorySize_4GB; } Smc::MemoryArrangement GetMemoryArrangeForInit() { - switch (GetMemoryModeForInit() & 0x3F) { - case 0x01: - default: - return Smc::MemoryArrangement_4GB; - case 0x02: - return Smc::MemoryArrangement_4GBForAppletDev; - case 0x03: - return Smc::MemoryArrangement_4GBForSystemDev; - case 0x11: - return Smc::MemoryArrangement_6GB; - case 0x12: - return Smc::MemoryArrangement_6GBForAppletDev; - case 0x21: - return Smc::MemoryArrangement_8GB; - } + return Settings::values.use_extended_memory_layout ? Smc::MemoryArrangement_6GB + : Smc::MemoryArrangement_4GB; } } // namespace +size_t KSystemControl::Init::GetRealMemorySize() { + return GetIntendedMemorySize(); +} + // Initialization. size_t KSystemControl::Init::GetIntendedMemorySize() { switch (GetMemorySizeForInit()) { @@ -69,7 +57,13 @@ size_t KSystemControl::Init::GetIntendedMemorySize() { } PAddr KSystemControl::Init::GetKernelPhysicalBaseAddress(u64 base_address) { - return base_address; + const size_t real_dram_size = KSystemControl::Init::GetRealMemorySize(); + const size_t intended_dram_size = KSystemControl::Init::GetIntendedMemorySize(); + if (intended_dram_size * 2 < real_dram_size) { + return base_address; + } else { + return base_address + ((real_dram_size - intended_dram_size) / 2); + } } bool KSystemControl::Init::ShouldIncreaseThreadResourceLimit() { diff --git a/src/core/hle/kernel/board/nintendo/nx/k_system_control.h b/src/core/hle/kernel/board/nintendo/nx/k_system_control.h index 52f230ced..df2a17f2a 100644 --- a/src/core/hle/kernel/board/nintendo/nx/k_system_control.h +++ b/src/core/hle/kernel/board/nintendo/nx/k_system_control.h @@ -13,6 +13,7 @@ public: class Init { public: // Initialization. + static std::size_t GetRealMemorySize(); static std::size_t GetIntendedMemorySize(); static PAddr GetKernelPhysicalBaseAddress(u64 base_address); static bool ShouldIncreaseThreadResourceLimit(); diff --git a/src/core/hle/kernel/initial_process.h b/src/core/hle/kernel/initial_process.h new file mode 100644 index 000000000..25b27909c --- /dev/null +++ b/src/core/hle/kernel/initial_process.h @@ -0,0 +1,23 @@ +// Copyright 2022 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include "common/common_types.h" +#include "common/literals.h" +#include "core/hle/kernel/board/nintendo/nx/k_memory_layout.h" +#include "core/hle/kernel/board/nintendo/nx/k_system_control.h" + +namespace Kernel { + +using namespace Common::Literals; + +constexpr std::size_t InitialProcessBinarySizeMax = 12_MiB; + +static inline PAddr GetInitialProcessBinaryPhysicalAddress() { + return Kernel::Board::Nintendo::Nx::KSystemControl::Init::GetKernelPhysicalBaseAddress( + MainMemoryAddress); +} + +} // namespace Kernel diff --git a/src/core/hle/kernel/k_address_arbiter.cpp b/src/core/hle/kernel/k_address_arbiter.cpp index 783c69858..1d1f5e5f8 100644 --- a/src/core/hle/kernel/k_address_arbiter.cpp +++ b/src/core/hle/kernel/k_address_arbiter.cpp @@ -49,7 +49,7 @@ bool DecrementIfLessThan(Core::System& system, s32* out, VAddr address, s32 valu } } else { // Otherwise, clear our exclusive hold and finish - monitor.ClearExclusive(); + monitor.ClearExclusive(current_core); } // We're done. @@ -78,7 +78,7 @@ bool UpdateIfEqual(Core::System& system, s32* out, VAddr address, s32 value, s32 } } else { // Otherwise, clear our exclusive hold and finish. - monitor.ClearExclusive(); + monitor.ClearExclusive(current_core); } // We're done. diff --git a/src/core/hle/kernel/k_event.cpp b/src/core/hle/kernel/k_event.cpp index 0720efece..2e0e8de80 100644 --- a/src/core/hle/kernel/k_event.cpp +++ b/src/core/hle/kernel/k_event.cpp @@ -14,7 +14,7 @@ KEvent::KEvent(KernelCore& kernel_) KEvent::~KEvent() = default; -void KEvent::Initialize(std::string&& name_) { +void KEvent::Initialize(std::string&& name_, KProcess* owner_) { // Increment reference count. // Because reference count is one on creation, this will result // in a reference count of two. Thus, when both readable and @@ -30,10 +30,8 @@ void KEvent::Initialize(std::string&& name_) { writable_event.Initialize(this, name_ + ":Writable"); // Set our owner process. - owner = kernel.CurrentProcess(); - if (owner) { - owner->Open(); - } + owner = owner_; + owner->Open(); // Mark initialized. name = std::move(name_); @@ -47,10 +45,8 @@ void KEvent::Finalize() { void KEvent::PostDestroy(uintptr_t arg) { // Release the event count resource the owner process holds. KProcess* owner = reinterpret_cast<KProcess*>(arg); - if (owner) { - owner->GetResourceLimit()->Release(LimitableResource::Events, 1); - owner->Close(); - } + owner->GetResourceLimit()->Release(LimitableResource::Events, 1); + owner->Close(); } } // namespace Kernel diff --git a/src/core/hle/kernel/k_event.h b/src/core/hle/kernel/k_event.h index 3d3ec99e2..de9732ddf 100644 --- a/src/core/hle/kernel/k_event.h +++ b/src/core/hle/kernel/k_event.h @@ -22,7 +22,7 @@ public: explicit KEvent(KernelCore& kernel_); ~KEvent() override; - void Initialize(std::string&& name); + void Initialize(std::string&& name, KProcess* owner_); void Finalize() override; diff --git a/src/core/hle/kernel/k_memory_layout.h b/src/core/hle/kernel/k_memory_layout.h index 57ff538cc..bcddb0d62 100644 --- a/src/core/hle/kernel/k_memory_layout.h +++ b/src/core/hle/kernel/k_memory_layout.h @@ -173,6 +173,10 @@ public: return Dereference(FindVirtualLinear(address)); } + const KMemoryRegion& GetPhysicalLinearRegion(PAddr address) const { + return Dereference(FindPhysicalLinear(address)); + } + const KMemoryRegion* GetPhysicalKernelTraceBufferRegion() const { return GetPhysicalMemoryRegionTree().FindFirstDerived(KMemoryRegionType_KernelTraceBuffer); } diff --git a/src/core/hle/kernel/k_memory_manager.cpp b/src/core/hle/kernel/k_memory_manager.cpp index 1b44541b1..a2f18f643 100644 --- a/src/core/hle/kernel/k_memory_manager.cpp +++ b/src/core/hle/kernel/k_memory_manager.cpp @@ -10,189 +10,412 @@ #include "common/scope_exit.h" #include "core/core.h" #include "core/device_memory.h" +#include "core/hle/kernel/initial_process.h" #include "core/hle/kernel/k_memory_manager.h" #include "core/hle/kernel/k_page_linked_list.h" +#include "core/hle/kernel/kernel.h" #include "core/hle/kernel/svc_results.h" +#include "core/memory.h" namespace Kernel { -KMemoryManager::KMemoryManager(Core::System& system_) : system{system_} {} +namespace { + +constexpr KMemoryManager::Pool GetPoolFromMemoryRegionType(u32 type) { + if ((type | KMemoryRegionType_DramApplicationPool) == type) { + return KMemoryManager::Pool::Application; + } else if ((type | KMemoryRegionType_DramAppletPool) == type) { + return KMemoryManager::Pool::Applet; + } else if ((type | KMemoryRegionType_DramSystemPool) == type) { + return KMemoryManager::Pool::System; + } else if ((type | KMemoryRegionType_DramSystemNonSecurePool) == type) { + return KMemoryManager::Pool::SystemNonSecure; + } else { + UNREACHABLE_MSG("InvalidMemoryRegionType for conversion to Pool"); + return {}; + } +} -std::size_t KMemoryManager::Impl::Initialize(Pool new_pool, u64 start_address, u64 end_address) { - const auto size{end_address - start_address}; +} // namespace + +KMemoryManager::KMemoryManager(Core::System& system_) + : system{system_}, pool_locks{ + KLightLock{system_.Kernel()}, + KLightLock{system_.Kernel()}, + KLightLock{system_.Kernel()}, + KLightLock{system_.Kernel()}, + } {} + +void KMemoryManager::Initialize(VAddr management_region, size_t management_region_size) { + + // Clear the management region to zero. + const VAddr management_region_end = management_region + management_region_size; + + // Reset our manager count. + num_managers = 0; + + // Traverse the virtual memory layout tree, initializing each manager as appropriate. + while (num_managers != MaxManagerCount) { + // Locate the region that should initialize the current manager. + PAddr region_address = 0; + size_t region_size = 0; + Pool region_pool = Pool::Count; + for (const auto& it : system.Kernel().MemoryLayout().GetPhysicalMemoryRegionTree()) { + // We only care about regions that we need to create managers for. + if (!it.IsDerivedFrom(KMemoryRegionType_DramUserPool)) { + continue; + } - // Calculate metadata sizes - const auto ref_count_size{(size / PageSize) * sizeof(u16)}; - const auto optimize_map_size{(Common::AlignUp((size / PageSize), 64) / 64) * sizeof(u64)}; - const auto manager_size{Common::AlignUp(optimize_map_size + ref_count_size, PageSize)}; - const auto page_heap_size{KPageHeap::CalculateManagementOverheadSize(size)}; - const auto total_metadata_size{manager_size + page_heap_size}; - ASSERT(manager_size <= total_metadata_size); - ASSERT(Common::IsAligned(total_metadata_size, PageSize)); + // We want to initialize the managers in order. + if (it.GetAttributes() != num_managers) { + continue; + } - // Setup region - pool = new_pool; + const PAddr cur_start = it.GetAddress(); + const PAddr cur_end = it.GetEndAddress(); + + // Validate the region. + ASSERT(cur_end != 0); + ASSERT(cur_start != 0); + ASSERT(it.GetSize() > 0); + + // Update the region's extents. + if (region_address == 0) { + region_address = cur_start; + region_size = it.GetSize(); + region_pool = GetPoolFromMemoryRegionType(it.GetType()); + } else { + ASSERT(cur_start == region_address + region_size); + + // Update the size. + region_size = cur_end - region_address; + ASSERT(GetPoolFromMemoryRegionType(it.GetType()) == region_pool); + } + } + + // If we didn't find a region, we're done. + if (region_size == 0) { + break; + } - // Initialize the manager's KPageHeap - heap.Initialize(start_address, size, page_heap_size); + // Initialize a new manager for the region. + Impl* manager = std::addressof(managers[num_managers++]); + ASSERT(num_managers <= managers.size()); + + const size_t cur_size = manager->Initialize(region_address, region_size, management_region, + management_region_end, region_pool); + management_region += cur_size; + ASSERT(management_region <= management_region_end); + + // Insert the manager into the pool list. + const auto region_pool_index = static_cast<u32>(region_pool); + if (pool_managers_tail[region_pool_index] == nullptr) { + pool_managers_head[region_pool_index] = manager; + } else { + pool_managers_tail[region_pool_index]->SetNext(manager); + manager->SetPrev(pool_managers_tail[region_pool_index]); + } + pool_managers_tail[region_pool_index] = manager; + } - // Free the memory to the heap - heap.Free(start_address, size / PageSize); + // Free each region to its corresponding heap. + size_t reserved_sizes[MaxManagerCount] = {}; + const PAddr ini_start = GetInitialProcessBinaryPhysicalAddress(); + const PAddr ini_end = ini_start + InitialProcessBinarySizeMax; + const PAddr ini_last = ini_end - 1; + for (const auto& it : system.Kernel().MemoryLayout().GetPhysicalMemoryRegionTree()) { + if (it.IsDerivedFrom(KMemoryRegionType_DramUserPool)) { + // Get the manager for the region. + auto index = it.GetAttributes(); + auto& manager = managers[index]; + + const PAddr cur_start = it.GetAddress(); + const PAddr cur_last = it.GetLastAddress(); + const PAddr cur_end = it.GetEndAddress(); + + if (cur_start <= ini_start && ini_last <= cur_last) { + // Free memory before the ini to the heap. + if (cur_start != ini_start) { + manager.Free(cur_start, (ini_start - cur_start) / PageSize); + } - // Update the heap's used size - heap.UpdateUsedSize(); + // Open/reserve the ini memory. + manager.OpenFirst(ini_start, InitialProcessBinarySizeMax / PageSize); + reserved_sizes[it.GetAttributes()] += InitialProcessBinarySizeMax; - return total_metadata_size; -} + // Free memory after the ini to the heap. + if (ini_last != cur_last) { + ASSERT(cur_end != 0); + manager.Free(ini_end, cur_end - ini_end); + } + } else { + // Ensure there's no partial overlap with the ini image. + if (cur_start <= ini_last) { + ASSERT(cur_last < ini_start); + } else { + // Otherwise, check the region for general validity. + ASSERT(cur_end != 0); + } -void KMemoryManager::InitializeManager(Pool pool, u64 start_address, u64 end_address) { - ASSERT(pool < Pool::Count); - managers[static_cast<std::size_t>(pool)].Initialize(pool, start_address, end_address); + // Free the memory to the heap. + manager.Free(cur_start, it.GetSize() / PageSize); + } + } + } + + // Update the used size for all managers. + for (size_t i = 0; i < num_managers; ++i) { + managers[i].SetInitialUsedHeapSize(reserved_sizes[i]); + } } -VAddr KMemoryManager::AllocateAndOpenContinuous(std::size_t num_pages, std::size_t align_pages, - u32 option) { - // Early return if we're allocating no pages +PAddr KMemoryManager::AllocateAndOpenContinuous(size_t num_pages, size_t align_pages, u32 option) { + // Early return if we're allocating no pages. if (num_pages == 0) { - return {}; + return 0; } - // Lock the pool that we're allocating from + // Lock the pool that we're allocating from. const auto [pool, dir] = DecodeOption(option); - const auto pool_index{static_cast<std::size_t>(pool)}; - std::lock_guard lock{pool_locks[pool_index]}; - - // Choose a heap based on our page size request - const s32 heap_index{KPageHeap::GetAlignedBlockIndex(num_pages, align_pages)}; - - // Loop, trying to iterate from each block - // TODO (bunnei): Support multiple managers - Impl& chosen_manager{managers[pool_index]}; - VAddr allocated_block{chosen_manager.AllocateBlock(heap_index, false)}; + KScopedLightLock lk(pool_locks[static_cast<std::size_t>(pool)]); + + // Choose a heap based on our page size request. + const s32 heap_index = KPageHeap::GetAlignedBlockIndex(num_pages, align_pages); + + // Loop, trying to iterate from each block. + Impl* chosen_manager = nullptr; + PAddr allocated_block = 0; + for (chosen_manager = this->GetFirstManager(pool, dir); chosen_manager != nullptr; + chosen_manager = this->GetNextManager(chosen_manager, dir)) { + allocated_block = chosen_manager->AllocateBlock(heap_index, true); + if (allocated_block != 0) { + break; + } + } - // If we failed to allocate, quit now - if (!allocated_block) { - return {}; + // If we failed to allocate, quit now. + if (allocated_block == 0) { + return 0; } - // If we allocated more than we need, free some - const auto allocated_pages{KPageHeap::GetBlockNumPages(heap_index)}; + // If we allocated more than we need, free some. + const size_t allocated_pages = KPageHeap::GetBlockNumPages(heap_index); if (allocated_pages > num_pages) { - chosen_manager.Free(allocated_block + num_pages * PageSize, allocated_pages - num_pages); + chosen_manager->Free(allocated_block + num_pages * PageSize, allocated_pages - num_pages); } + // Open the first reference to the pages. + chosen_manager->OpenFirst(allocated_block, num_pages); + return allocated_block; } -ResultCode KMemoryManager::Allocate(KPageLinkedList& page_list, std::size_t num_pages, Pool pool, - Direction dir, u32 heap_fill_value) { - ASSERT(page_list.GetNumPages() == 0); +ResultCode KMemoryManager::AllocatePageGroupImpl(KPageLinkedList* out, size_t num_pages, Pool pool, + Direction dir, bool random) { + // Choose a heap based on our page size request. + const s32 heap_index = KPageHeap::GetBlockIndex(num_pages); + R_UNLESS(0 <= heap_index, ResultOutOfMemory); + + // Ensure that we don't leave anything un-freed. + auto group_guard = SCOPE_GUARD({ + for (const auto& it : out->Nodes()) { + auto& manager = this->GetManager(system.Kernel().MemoryLayout(), it.GetAddress()); + const size_t num_pages_to_free = + std::min(it.GetNumPages(), (manager.GetEndAddress() - it.GetAddress()) / PageSize); + manager.Free(it.GetAddress(), num_pages_to_free); + } + }); - // Early return if we're allocating no pages - if (num_pages == 0) { - return ResultSuccess; - } + // Keep allocating until we've allocated all our pages. + for (s32 index = heap_index; index >= 0 && num_pages > 0; index--) { + const size_t pages_per_alloc = KPageHeap::GetBlockNumPages(index); + for (Impl* cur_manager = this->GetFirstManager(pool, dir); cur_manager != nullptr; + cur_manager = this->GetNextManager(cur_manager, dir)) { + while (num_pages >= pages_per_alloc) { + // Allocate a block. + PAddr allocated_block = cur_manager->AllocateBlock(index, random); + if (allocated_block == 0) { + break; + } - // Lock the pool that we're allocating from - const auto pool_index{static_cast<std::size_t>(pool)}; - std::lock_guard lock{pool_locks[pool_index]}; + // Safely add it to our group. + { + auto block_guard = + SCOPE_GUARD({ cur_manager->Free(allocated_block, pages_per_alloc); }); + R_TRY(out->AddBlock(allocated_block, pages_per_alloc)); + block_guard.Cancel(); + } - // Choose a heap based on our page size request - const s32 heap_index{KPageHeap::GetBlockIndex(num_pages)}; - if (heap_index < 0) { - return ResultOutOfMemory; + num_pages -= pages_per_alloc; + } + } } - // TODO (bunnei): Support multiple managers - Impl& chosen_manager{managers[pool_index]}; + // Only succeed if we allocated as many pages as we wanted. + R_UNLESS(num_pages == 0, ResultOutOfMemory); - // Ensure that we don't leave anything un-freed - auto group_guard = detail::ScopeExit([&] { - for (const auto& it : page_list.Nodes()) { - const auto min_num_pages{std::min<size_t>( - it.GetNumPages(), (chosen_manager.GetEndAddress() - it.GetAddress()) / PageSize)}; - chosen_manager.Free(it.GetAddress(), min_num_pages); - } - }); + // We succeeded! + group_guard.Cancel(); + return ResultSuccess; +} - // Keep allocating until we've allocated all our pages - for (s32 index{heap_index}; index >= 0 && num_pages > 0; index--) { - const auto pages_per_alloc{KPageHeap::GetBlockNumPages(index)}; +ResultCode KMemoryManager::AllocateAndOpen(KPageLinkedList* out, size_t num_pages, u32 option) { + ASSERT(out != nullptr); + ASSERT(out->GetNumPages() == 0); - while (num_pages >= pages_per_alloc) { - // Allocate a block - VAddr allocated_block{chosen_manager.AllocateBlock(index, false)}; - if (!allocated_block) { - break; - } + // Early return if we're allocating no pages. + R_SUCCEED_IF(num_pages == 0); - // Safely add it to our group - { - auto block_guard = detail::ScopeExit( - [&] { chosen_manager.Free(allocated_block, pages_per_alloc); }); + // Lock the pool that we're allocating from. + const auto [pool, dir] = DecodeOption(option); + KScopedLightLock lk(pool_locks[static_cast<size_t>(pool)]); + + // Allocate the page group. + R_TRY(this->AllocatePageGroupImpl(out, num_pages, pool, dir, false)); + + // Open the first reference to the pages. + for (const auto& block : out->Nodes()) { + PAddr cur_address = block.GetAddress(); + size_t remaining_pages = block.GetNumPages(); + while (remaining_pages > 0) { + // Get the manager for the current address. + auto& manager = this->GetManager(system.Kernel().MemoryLayout(), cur_address); + + // Process part or all of the block. + const size_t cur_pages = + std::min(remaining_pages, manager.GetPageOffsetToEnd(cur_address)); + manager.OpenFirst(cur_address, cur_pages); + + // Advance. + cur_address += cur_pages * PageSize; + remaining_pages -= cur_pages; + } + } - if (const ResultCode result{page_list.AddBlock(allocated_block, pages_per_alloc)}; - result.IsError()) { - return result; - } + return ResultSuccess; +} - block_guard.Cancel(); - } +ResultCode KMemoryManager::AllocateAndOpenForProcess(KPageLinkedList* out, size_t num_pages, + u32 option, u64 process_id, u8 fill_pattern) { + ASSERT(out != nullptr); + ASSERT(out->GetNumPages() == 0); - num_pages -= pages_per_alloc; - } - } + // Decode the option. + const auto [pool, dir] = DecodeOption(option); - // Clear allocated memory. - for (const auto& it : page_list.Nodes()) { - std::memset(system.DeviceMemory().GetPointer(it.GetAddress()), heap_fill_value, - it.GetSize()); + // Allocate the memory. + { + // Lock the pool that we're allocating from. + KScopedLightLock lk(pool_locks[static_cast<size_t>(pool)]); + + // Allocate the page group. + R_TRY(this->AllocatePageGroupImpl(out, num_pages, pool, dir, false)); + + // Open the first reference to the pages. + for (const auto& block : out->Nodes()) { + PAddr cur_address = block.GetAddress(); + size_t remaining_pages = block.GetNumPages(); + while (remaining_pages > 0) { + // Get the manager for the current address. + auto& manager = this->GetManager(system.Kernel().MemoryLayout(), cur_address); + + // Process part or all of the block. + const size_t cur_pages = + std::min(remaining_pages, manager.GetPageOffsetToEnd(cur_address)); + manager.OpenFirst(cur_address, cur_pages); + + // Advance. + cur_address += cur_pages * PageSize; + remaining_pages -= cur_pages; + } + } } - // Only succeed if we allocated as many pages as we wanted - if (num_pages) { - return ResultOutOfMemory; + // Set all the allocated memory. + for (const auto& block : out->Nodes()) { + std::memset(system.DeviceMemory().GetPointer(block.GetAddress()), fill_pattern, + block.GetSize()); } - // We succeeded! - group_guard.Cancel(); - return ResultSuccess; } -ResultCode KMemoryManager::Free(KPageLinkedList& page_list, std::size_t num_pages, Pool pool, - Direction dir, u32 heap_fill_value) { - // Early return if we're freeing no pages - if (!num_pages) { - return ResultSuccess; +void KMemoryManager::Open(PAddr address, size_t num_pages) { + // Repeatedly open references until we've done so for all pages. + while (num_pages) { + auto& manager = this->GetManager(system.Kernel().MemoryLayout(), address); + const size_t cur_pages = std::min(num_pages, manager.GetPageOffsetToEnd(address)); + + { + KScopedLightLock lk(pool_locks[static_cast<size_t>(manager.GetPool())]); + manager.Open(address, cur_pages); + } + + num_pages -= cur_pages; + address += cur_pages * PageSize; } +} - // Lock the pool that we're freeing from - const auto pool_index{static_cast<std::size_t>(pool)}; - std::lock_guard lock{pool_locks[pool_index]}; +void KMemoryManager::Close(PAddr address, size_t num_pages) { + // Repeatedly close references until we've done so for all pages. + while (num_pages) { + auto& manager = this->GetManager(system.Kernel().MemoryLayout(), address); + const size_t cur_pages = std::min(num_pages, manager.GetPageOffsetToEnd(address)); - // TODO (bunnei): Support multiple managers - Impl& chosen_manager{managers[pool_index]}; + { + KScopedLightLock lk(pool_locks[static_cast<size_t>(manager.GetPool())]); + manager.Close(address, cur_pages); + } - // Free all of the pages - for (const auto& it : page_list.Nodes()) { - const auto min_num_pages{std::min<size_t>( - it.GetNumPages(), (chosen_manager.GetEndAddress() - it.GetAddress()) / PageSize)}; - chosen_manager.Free(it.GetAddress(), min_num_pages); + num_pages -= cur_pages; + address += cur_pages * PageSize; } +} - return ResultSuccess; +void KMemoryManager::Close(const KPageLinkedList& pg) { + for (const auto& node : pg.Nodes()) { + Close(node.GetAddress(), node.GetNumPages()); + } +} +void KMemoryManager::Open(const KPageLinkedList& pg) { + for (const auto& node : pg.Nodes()) { + Open(node.GetAddress(), node.GetNumPages()); + } +} + +size_t KMemoryManager::Impl::Initialize(PAddr address, size_t size, VAddr management, + VAddr management_end, Pool p) { + // Calculate management sizes. + const size_t ref_count_size = (size / PageSize) * sizeof(u16); + const size_t optimize_map_size = CalculateOptimizedProcessOverheadSize(size); + const size_t manager_size = Common::AlignUp(optimize_map_size + ref_count_size, PageSize); + const size_t page_heap_size = KPageHeap::CalculateManagementOverheadSize(size); + const size_t total_management_size = manager_size + page_heap_size; + ASSERT(manager_size <= total_management_size); + ASSERT(management + total_management_size <= management_end); + ASSERT(Common::IsAligned(total_management_size, PageSize)); + + // Setup region. + pool = p; + management_region = management; + page_reference_counts.resize( + Kernel::Board::Nintendo::Nx::KSystemControl::Init::GetIntendedMemorySize() / PageSize); + ASSERT(Common::IsAligned(management_region, PageSize)); + + // Initialize the manager's KPageHeap. + heap.Initialize(address, size, management + manager_size, page_heap_size); + + return total_management_size; } -std::size_t KMemoryManager::Impl::CalculateManagementOverheadSize(std::size_t region_size) { - const std::size_t ref_count_size = (region_size / PageSize) * sizeof(u16); - const std::size_t optimize_map_size = +size_t KMemoryManager::Impl::CalculateManagementOverheadSize(size_t region_size) { + const size_t ref_count_size = (region_size / PageSize) * sizeof(u16); + const size_t optimize_map_size = (Common::AlignUp((region_size / PageSize), Common::BitSize<u64>()) / Common::BitSize<u64>()) * sizeof(u64); - const std::size_t manager_meta_size = - Common::AlignUp(optimize_map_size + ref_count_size, PageSize); - const std::size_t page_heap_size = KPageHeap::CalculateManagementOverheadSize(region_size); + const size_t manager_meta_size = Common::AlignUp(optimize_map_size + ref_count_size, PageSize); + const size_t page_heap_size = KPageHeap::CalculateManagementOverheadSize(region_size); return manager_meta_size + page_heap_size; } diff --git a/src/core/hle/kernel/k_memory_manager.h b/src/core/hle/kernel/k_memory_manager.h index 17c7690f1..18775b262 100644 --- a/src/core/hle/kernel/k_memory_manager.h +++ b/src/core/hle/kernel/k_memory_manager.h @@ -5,11 +5,12 @@ #pragma once #include <array> -#include <mutex> #include <tuple> #include "common/common_funcs.h" #include "common/common_types.h" +#include "core/hle/kernel/k_light_lock.h" +#include "core/hle/kernel/k_memory_layout.h" #include "core/hle/kernel/k_page_heap.h" #include "core/hle/result.h" @@ -52,22 +53,33 @@ public: explicit KMemoryManager(Core::System& system_); - constexpr std::size_t GetSize(Pool pool) const { - return managers[static_cast<std::size_t>(pool)].GetSize(); + void Initialize(VAddr management_region, size_t management_region_size); + + constexpr size_t GetSize(Pool pool) const { + constexpr Direction GetSizeDirection = Direction::FromFront; + size_t total = 0; + for (auto* manager = this->GetFirstManager(pool, GetSizeDirection); manager != nullptr; + manager = this->GetNextManager(manager, GetSizeDirection)) { + total += manager->GetSize(); + } + return total; } - void InitializeManager(Pool pool, u64 start_address, u64 end_address); + PAddr AllocateAndOpenContinuous(size_t num_pages, size_t align_pages, u32 option); + ResultCode AllocateAndOpen(KPageLinkedList* out, size_t num_pages, u32 option); + ResultCode AllocateAndOpenForProcess(KPageLinkedList* out, size_t num_pages, u32 option, + u64 process_id, u8 fill_pattern); + + static constexpr size_t MaxManagerCount = 10; - VAddr AllocateAndOpenContinuous(size_t num_pages, size_t align_pages, u32 option); - ResultCode Allocate(KPageLinkedList& page_list, std::size_t num_pages, Pool pool, Direction dir, - u32 heap_fill_value = 0); - ResultCode Free(KPageLinkedList& page_list, std::size_t num_pages, Pool pool, Direction dir, - u32 heap_fill_value = 0); + void Close(PAddr address, size_t num_pages); + void Close(const KPageLinkedList& pg); - static constexpr std::size_t MaxManagerCount = 10; + void Open(PAddr address, size_t num_pages); + void Open(const KPageLinkedList& pg); public: - static std::size_t CalculateManagementOverheadSize(std::size_t region_size) { + static size_t CalculateManagementOverheadSize(size_t region_size) { return Impl::CalculateManagementOverheadSize(region_size); } @@ -100,17 +112,26 @@ private: Impl() = default; ~Impl() = default; - std::size_t Initialize(Pool new_pool, u64 start_address, u64 end_address); + size_t Initialize(PAddr address, size_t size, VAddr management, VAddr management_end, + Pool p); VAddr AllocateBlock(s32 index, bool random) { return heap.AllocateBlock(index, random); } - void Free(VAddr addr, std::size_t num_pages) { + void Free(VAddr addr, size_t num_pages) { heap.Free(addr, num_pages); } - constexpr std::size_t GetSize() const { + void SetInitialUsedHeapSize(size_t reserved_size) { + heap.SetInitialUsedSize(reserved_size); + } + + constexpr Pool GetPool() const { + return pool; + } + + constexpr size_t GetSize() const { return heap.GetSize(); } @@ -122,10 +143,88 @@ private: return heap.GetEndAddress(); } - static std::size_t CalculateManagementOverheadSize(std::size_t region_size); + constexpr size_t GetPageOffset(PAddr address) const { + return heap.GetPageOffset(address); + } + + constexpr size_t GetPageOffsetToEnd(PAddr address) const { + return heap.GetPageOffsetToEnd(address); + } + + constexpr void SetNext(Impl* n) { + next = n; + } + + constexpr void SetPrev(Impl* n) { + prev = n; + } + + constexpr Impl* GetNext() const { + return next; + } + + constexpr Impl* GetPrev() const { + return prev; + } + + void OpenFirst(PAddr address, size_t num_pages) { + size_t index = this->GetPageOffset(address); + const size_t end = index + num_pages; + while (index < end) { + const RefCount ref_count = (++page_reference_counts[index]); + ASSERT(ref_count == 1); - static constexpr std::size_t CalculateOptimizedProcessOverheadSize( - std::size_t region_size) { + index++; + } + } + + void Open(PAddr address, size_t num_pages) { + size_t index = this->GetPageOffset(address); + const size_t end = index + num_pages; + while (index < end) { + const RefCount ref_count = (++page_reference_counts[index]); + ASSERT(ref_count > 1); + + index++; + } + } + + void Close(PAddr address, size_t num_pages) { + size_t index = this->GetPageOffset(address); + const size_t end = index + num_pages; + + size_t free_start = 0; + size_t free_count = 0; + while (index < end) { + ASSERT(page_reference_counts[index] > 0); + const RefCount ref_count = (--page_reference_counts[index]); + + // Keep track of how many zero refcounts we see in a row, to minimize calls to free. + if (ref_count == 0) { + if (free_count > 0) { + free_count++; + } else { + free_start = index; + free_count = 1; + } + } else { + if (free_count > 0) { + this->Free(heap.GetAddress() + free_start * PageSize, free_count); + free_count = 0; + } + } + + index++; + } + + if (free_count > 0) { + this->Free(heap.GetAddress() + free_start * PageSize, free_count); + } + } + + static size_t CalculateManagementOverheadSize(size_t region_size); + + static constexpr size_t CalculateOptimizedProcessOverheadSize(size_t region_size) { return (Common::AlignUp((region_size / PageSize), Common::BitSize<u64>()) / Common::BitSize<u64>()) * sizeof(u64); @@ -135,13 +234,45 @@ private: using RefCount = u16; KPageHeap heap; + std::vector<RefCount> page_reference_counts; + VAddr management_region{}; Pool pool{}; + Impl* next{}; + Impl* prev{}; }; private: + Impl& GetManager(const KMemoryLayout& memory_layout, PAddr address) { + return managers[memory_layout.GetPhysicalLinearRegion(address).GetAttributes()]; + } + + const Impl& GetManager(const KMemoryLayout& memory_layout, PAddr address) const { + return managers[memory_layout.GetPhysicalLinearRegion(address).GetAttributes()]; + } + + constexpr Impl* GetFirstManager(Pool pool, Direction dir) const { + return dir == Direction::FromBack ? pool_managers_tail[static_cast<size_t>(pool)] + : pool_managers_head[static_cast<size_t>(pool)]; + } + + constexpr Impl* GetNextManager(Impl* cur, Direction dir) const { + if (dir == Direction::FromBack) { + return cur->GetPrev(); + } else { + return cur->GetNext(); + } + } + + ResultCode AllocatePageGroupImpl(KPageLinkedList* out, size_t num_pages, Pool pool, + Direction dir, bool random); + +private: Core::System& system; - std::array<std::mutex, static_cast<std::size_t>(Pool::Count)> pool_locks; + std::array<KLightLock, static_cast<size_t>(Pool::Count)> pool_locks; + std::array<Impl*, MaxManagerCount> pool_managers_head{}; + std::array<Impl*, MaxManagerCount> pool_managers_tail{}; std::array<Impl, MaxManagerCount> managers; + size_t num_managers{}; }; } // namespace Kernel diff --git a/src/core/hle/kernel/k_memory_region_type.h b/src/core/hle/kernel/k_memory_region_type.h index a05e66677..0baeddf51 100644 --- a/src/core/hle/kernel/k_memory_region_type.h +++ b/src/core/hle/kernel/k_memory_region_type.h @@ -14,7 +14,8 @@ namespace Kernel { enum KMemoryRegionType : u32 { - KMemoryRegionAttr_CarveoutProtected = 0x04000000, + KMemoryRegionAttr_CarveoutProtected = 0x02000000, + KMemoryRegionAttr_Uncached = 0x04000000, KMemoryRegionAttr_DidKernelMap = 0x08000000, KMemoryRegionAttr_ShouldKernelMap = 0x10000000, KMemoryRegionAttr_UserReadOnly = 0x20000000, @@ -239,6 +240,11 @@ static_assert(KMemoryRegionType_VirtualDramHeapBase.GetValue() == 0x1A); static_assert(KMemoryRegionType_VirtualDramKernelPtHeap.GetValue() == 0x2A); static_assert(KMemoryRegionType_VirtualDramKernelTraceBuffer.GetValue() == 0x4A); +// UNUSED: .DeriveSparse(2, 2, 0); +constexpr auto KMemoryRegionType_VirtualDramUnknownDebug = + KMemoryRegionType_Dram.DeriveSparse(2, 2, 1); +static_assert(KMemoryRegionType_VirtualDramUnknownDebug.GetValue() == (0x52)); + constexpr auto KMemoryRegionType_VirtualDramKernelInitPt = KMemoryRegionType_VirtualDramHeapBase.Derive(3, 0); constexpr auto KMemoryRegionType_VirtualDramPoolManagement = @@ -330,6 +336,8 @@ constexpr KMemoryRegionType GetTypeForVirtualLinearMapping(u32 type_id) { return KMemoryRegionType_VirtualDramKernelTraceBuffer; } else if (KMemoryRegionType_DramKernelPtHeap.IsAncestorOf(type_id)) { return KMemoryRegionType_VirtualDramKernelPtHeap; + } else if ((type_id | KMemoryRegionAttr_ShouldKernelMap) == type_id) { + return KMemoryRegionType_VirtualDramUnknownDebug; } else { return KMemoryRegionType_Dram; } diff --git a/src/core/hle/kernel/k_page_heap.cpp b/src/core/hle/kernel/k_page_heap.cpp index 29d996d62..97a5890a0 100644 --- a/src/core/hle/kernel/k_page_heap.cpp +++ b/src/core/hle/kernel/k_page_heap.cpp @@ -7,35 +7,51 @@ namespace Kernel { -void KPageHeap::Initialize(VAddr address, std::size_t size, std::size_t metadata_size) { - // Check our assumptions - ASSERT(Common::IsAligned((address), PageSize)); +void KPageHeap::Initialize(PAddr address, size_t size, VAddr management_address, + size_t management_size, const size_t* block_shifts, + size_t num_block_shifts) { + // Check our assumptions. + ASSERT(Common::IsAligned(address, PageSize)); ASSERT(Common::IsAligned(size, PageSize)); + ASSERT(0 < num_block_shifts && num_block_shifts <= NumMemoryBlockPageShifts); + const VAddr management_end = management_address + management_size; - // Set our members - heap_address = address; - heap_size = size; - - // Setup bitmaps - metadata.resize(metadata_size / sizeof(u64)); - u64* cur_bitmap_storage{metadata.data()}; - for (std::size_t i = 0; i < MemoryBlockPageShifts.size(); i++) { - const std::size_t cur_block_shift{MemoryBlockPageShifts[i]}; - const std::size_t next_block_shift{ - (i != MemoryBlockPageShifts.size() - 1) ? MemoryBlockPageShifts[i + 1] : 0}; - cur_bitmap_storage = blocks[i].Initialize(heap_address, heap_size, cur_block_shift, - next_block_shift, cur_bitmap_storage); + // Set our members. + m_heap_address = address; + m_heap_size = size; + m_num_blocks = num_block_shifts; + + // Setup bitmaps. + m_management_data.resize(management_size / sizeof(u64)); + u64* cur_bitmap_storage{m_management_data.data()}; + for (size_t i = 0; i < num_block_shifts; i++) { + const size_t cur_block_shift = block_shifts[i]; + const size_t next_block_shift = (i != num_block_shifts - 1) ? block_shifts[i + 1] : 0; + cur_bitmap_storage = m_blocks[i].Initialize(m_heap_address, m_heap_size, cur_block_shift, + next_block_shift, cur_bitmap_storage); } + + // Ensure we didn't overextend our bounds. + ASSERT(VAddr(cur_bitmap_storage) <= management_end); +} + +size_t KPageHeap::GetNumFreePages() const { + size_t num_free = 0; + + for (size_t i = 0; i < m_num_blocks; i++) { + num_free += m_blocks[i].GetNumFreePages(); + } + + return num_free; } -VAddr KPageHeap::AllocateBlock(s32 index, bool random) { - const std::size_t needed_size{blocks[index].GetSize()}; +PAddr KPageHeap::AllocateBlock(s32 index, bool random) { + const size_t needed_size = m_blocks[index].GetSize(); - for (s32 i{index}; i < static_cast<s32>(MemoryBlockPageShifts.size()); i++) { - if (const VAddr addr{blocks[i].PopBlock(random)}; addr) { - if (const std::size_t allocated_size{blocks[i].GetSize()}; - allocated_size > needed_size) { - Free(addr + needed_size, (allocated_size - needed_size) / PageSize); + for (s32 i = index; i < static_cast<s32>(m_num_blocks); i++) { + if (const PAddr addr = m_blocks[i].PopBlock(random); addr != 0) { + if (const size_t allocated_size = m_blocks[i].GetSize(); allocated_size > needed_size) { + this->Free(addr + needed_size, (allocated_size - needed_size) / PageSize); } return addr; } @@ -44,34 +60,34 @@ VAddr KPageHeap::AllocateBlock(s32 index, bool random) { return 0; } -void KPageHeap::FreeBlock(VAddr block, s32 index) { +void KPageHeap::FreeBlock(PAddr block, s32 index) { do { - block = blocks[index++].PushBlock(block); + block = m_blocks[index++].PushBlock(block); } while (block != 0); } -void KPageHeap::Free(VAddr addr, std::size_t num_pages) { - // Freeing no pages is a no-op +void KPageHeap::Free(PAddr addr, size_t num_pages) { + // Freeing no pages is a no-op. if (num_pages == 0) { return; } - // Find the largest block size that we can free, and free as many as possible - s32 big_index{static_cast<s32>(MemoryBlockPageShifts.size()) - 1}; - const VAddr start{addr}; - const VAddr end{(num_pages * PageSize) + addr}; - VAddr before_start{start}; - VAddr before_end{start}; - VAddr after_start{end}; - VAddr after_end{end}; + // Find the largest block size that we can free, and free as many as possible. + s32 big_index = static_cast<s32>(m_num_blocks) - 1; + const PAddr start = addr; + const PAddr end = addr + num_pages * PageSize; + PAddr before_start = start; + PAddr before_end = start; + PAddr after_start = end; + PAddr after_end = end; while (big_index >= 0) { - const std::size_t block_size{blocks[big_index].GetSize()}; - const VAddr big_start{Common::AlignUp((start), block_size)}; - const VAddr big_end{Common::AlignDown((end), block_size)}; + const size_t block_size = m_blocks[big_index].GetSize(); + const PAddr big_start = Common::AlignUp(start, block_size); + const PAddr big_end = Common::AlignDown(end, block_size); if (big_start < big_end) { - // Free as many big blocks as we can - for (auto block{big_start}; block < big_end; block += block_size) { - FreeBlock(block, big_index); + // Free as many big blocks as we can. + for (auto block = big_start; block < big_end; block += block_size) { + this->FreeBlock(block, big_index); } before_end = big_start; after_start = big_end; @@ -81,31 +97,31 @@ void KPageHeap::Free(VAddr addr, std::size_t num_pages) { } ASSERT(big_index >= 0); - // Free space before the big blocks - for (s32 i{big_index - 1}; i >= 0; i--) { - const std::size_t block_size{blocks[i].GetSize()}; + // Free space before the big blocks. + for (s32 i = big_index - 1; i >= 0; i--) { + const size_t block_size = m_blocks[i].GetSize(); while (before_start + block_size <= before_end) { before_end -= block_size; - FreeBlock(before_end, i); + this->FreeBlock(before_end, i); } } - // Free space after the big blocks - for (s32 i{big_index - 1}; i >= 0; i--) { - const std::size_t block_size{blocks[i].GetSize()}; + // Free space after the big blocks. + for (s32 i = big_index - 1; i >= 0; i--) { + const size_t block_size = m_blocks[i].GetSize(); while (after_start + block_size <= after_end) { - FreeBlock(after_start, i); + this->FreeBlock(after_start, i); after_start += block_size; } } } -std::size_t KPageHeap::CalculateManagementOverheadSize(std::size_t region_size) { - std::size_t overhead_size = 0; - for (std::size_t i = 0; i < MemoryBlockPageShifts.size(); i++) { - const std::size_t cur_block_shift{MemoryBlockPageShifts[i]}; - const std::size_t next_block_shift{ - (i != MemoryBlockPageShifts.size() - 1) ? MemoryBlockPageShifts[i + 1] : 0}; +size_t KPageHeap::CalculateManagementOverheadSize(size_t region_size, const size_t* block_shifts, + size_t num_block_shifts) { + size_t overhead_size = 0; + for (size_t i = 0; i < num_block_shifts; i++) { + const size_t cur_block_shift = block_shifts[i]; + const size_t next_block_shift = (i != num_block_shifts - 1) ? block_shifts[i + 1] : 0; overhead_size += KPageHeap::Block::CalculateManagementOverheadSize( region_size, cur_block_shift, next_block_shift); } diff --git a/src/core/hle/kernel/k_page_heap.h b/src/core/hle/kernel/k_page_heap.h index a65aa28a0..60fff766b 100644 --- a/src/core/hle/kernel/k_page_heap.h +++ b/src/core/hle/kernel/k_page_heap.h @@ -23,54 +23,73 @@ public: KPageHeap() = default; ~KPageHeap() = default; - constexpr VAddr GetAddress() const { - return heap_address; + constexpr PAddr GetAddress() const { + return m_heap_address; } - constexpr std::size_t GetSize() const { - return heap_size; + constexpr size_t GetSize() const { + return m_heap_size; } - constexpr VAddr GetEndAddress() const { - return GetAddress() + GetSize(); + constexpr PAddr GetEndAddress() const { + return this->GetAddress() + this->GetSize(); } - constexpr std::size_t GetPageOffset(VAddr block) const { - return (block - GetAddress()) / PageSize; + constexpr size_t GetPageOffset(PAddr block) const { + return (block - this->GetAddress()) / PageSize; + } + constexpr size_t GetPageOffsetToEnd(PAddr block) const { + return (this->GetEndAddress() - block) / PageSize; + } + + void Initialize(PAddr heap_address, size_t heap_size, VAddr management_address, + size_t management_size) { + return this->Initialize(heap_address, heap_size, management_address, management_size, + MemoryBlockPageShifts.data(), NumMemoryBlockPageShifts); + } + + size_t GetFreeSize() const { + return this->GetNumFreePages() * PageSize; } - void Initialize(VAddr heap_address, std::size_t heap_size, std::size_t metadata_size); - VAddr AllocateBlock(s32 index, bool random); - void Free(VAddr addr, std::size_t num_pages); + void SetInitialUsedSize(size_t reserved_size) { + // Check that the reserved size is valid. + const size_t free_size = this->GetNumFreePages() * PageSize; + ASSERT(m_heap_size >= free_size + reserved_size); - void UpdateUsedSize() { - used_size = heap_size - (GetNumFreePages() * PageSize); + // Set the initial used size. + m_initial_used_size = m_heap_size - free_size - reserved_size; } - static std::size_t CalculateManagementOverheadSize(std::size_t region_size); + PAddr AllocateBlock(s32 index, bool random); + void Free(PAddr addr, size_t num_pages); + + static size_t CalculateManagementOverheadSize(size_t region_size) { + return CalculateManagementOverheadSize(region_size, MemoryBlockPageShifts.data(), + NumMemoryBlockPageShifts); + } - static constexpr s32 GetAlignedBlockIndex(std::size_t num_pages, std::size_t align_pages) { - const auto target_pages{std::max(num_pages, align_pages)}; - for (std::size_t i = 0; i < NumMemoryBlockPageShifts; i++) { - if (target_pages <= - (static_cast<std::size_t>(1) << MemoryBlockPageShifts[i]) / PageSize) { + static constexpr s32 GetAlignedBlockIndex(size_t num_pages, size_t align_pages) { + const size_t target_pages = std::max(num_pages, align_pages); + for (size_t i = 0; i < NumMemoryBlockPageShifts; i++) { + if (target_pages <= (size_t(1) << MemoryBlockPageShifts[i]) / PageSize) { return static_cast<s32>(i); } } return -1; } - static constexpr s32 GetBlockIndex(std::size_t num_pages) { - for (s32 i{static_cast<s32>(NumMemoryBlockPageShifts) - 1}; i >= 0; i--) { - if (num_pages >= (static_cast<std::size_t>(1) << MemoryBlockPageShifts[i]) / PageSize) { + static constexpr s32 GetBlockIndex(size_t num_pages) { + for (s32 i = static_cast<s32>(NumMemoryBlockPageShifts) - 1; i >= 0; i--) { + if (num_pages >= (size_t(1) << MemoryBlockPageShifts[i]) / PageSize) { return i; } } return -1; } - static constexpr std::size_t GetBlockSize(std::size_t index) { - return static_cast<std::size_t>(1) << MemoryBlockPageShifts[index]; + static constexpr size_t GetBlockSize(size_t index) { + return size_t(1) << MemoryBlockPageShifts[index]; } - static constexpr std::size_t GetBlockNumPages(std::size_t index) { + static constexpr size_t GetBlockNumPages(size_t index) { return GetBlockSize(index) / PageSize; } @@ -83,114 +102,116 @@ private: Block() = default; ~Block() = default; - constexpr std::size_t GetShift() const { - return block_shift; + constexpr size_t GetShift() const { + return m_block_shift; } - constexpr std::size_t GetNextShift() const { - return next_block_shift; + constexpr size_t GetNextShift() const { + return m_next_block_shift; } - constexpr std::size_t GetSize() const { - return static_cast<std::size_t>(1) << GetShift(); + constexpr size_t GetSize() const { + return u64(1) << this->GetShift(); } - constexpr std::size_t GetNumPages() const { - return GetSize() / PageSize; + constexpr size_t GetNumPages() const { + return this->GetSize() / PageSize; } - constexpr std::size_t GetNumFreeBlocks() const { - return bitmap.GetNumBits(); + constexpr size_t GetNumFreeBlocks() const { + return m_bitmap.GetNumBits(); } - constexpr std::size_t GetNumFreePages() const { - return GetNumFreeBlocks() * GetNumPages(); + constexpr size_t GetNumFreePages() const { + return this->GetNumFreeBlocks() * this->GetNumPages(); } - u64* Initialize(VAddr addr, std::size_t size, std::size_t bs, std::size_t nbs, - u64* bit_storage) { - // Set shifts - block_shift = bs; - next_block_shift = nbs; - - // Align up the address - VAddr end{addr + size}; - const auto align{(next_block_shift != 0) ? (1ULL << next_block_shift) - : (1ULL << block_shift)}; - addr = Common::AlignDown((addr), align); - end = Common::AlignUp((end), align); - - heap_address = addr; - end_offset = (end - addr) / (1ULL << block_shift); - return bitmap.Initialize(bit_storage, end_offset); + u64* Initialize(PAddr addr, size_t size, size_t bs, size_t nbs, u64* bit_storage) { + // Set shifts. + m_block_shift = bs; + m_next_block_shift = nbs; + + // Align up the address. + PAddr end = addr + size; + const size_t align = (m_next_block_shift != 0) ? (u64(1) << m_next_block_shift) + : (u64(1) << m_block_shift); + addr = Common::AlignDown(addr, align); + end = Common::AlignUp(end, align); + + m_heap_address = addr; + m_end_offset = (end - addr) / (u64(1) << m_block_shift); + return m_bitmap.Initialize(bit_storage, m_end_offset); } - VAddr PushBlock(VAddr address) { - // Set the bit for the free block - std::size_t offset{(address - heap_address) >> GetShift()}; - bitmap.SetBit(offset); + PAddr PushBlock(PAddr address) { + // Set the bit for the free block. + size_t offset = (address - m_heap_address) >> this->GetShift(); + m_bitmap.SetBit(offset); - // If we have a next shift, try to clear the blocks below and return the address - if (GetNextShift()) { - const auto diff{1ULL << (GetNextShift() - GetShift())}; + // If we have a next shift, try to clear the blocks below this one and return the new + // address. + if (this->GetNextShift()) { + const size_t diff = u64(1) << (this->GetNextShift() - this->GetShift()); offset = Common::AlignDown(offset, diff); - if (bitmap.ClearRange(offset, diff)) { - return heap_address + (offset << GetShift()); + if (m_bitmap.ClearRange(offset, diff)) { + return m_heap_address + (offset << this->GetShift()); } } - // We couldn't coalesce, or we're already as big as possible - return 0; + // We couldn't coalesce, or we're already as big as possible. + return {}; } - VAddr PopBlock(bool random) { - // Find a free block - const s64 soffset{bitmap.FindFreeBlock(random)}; + PAddr PopBlock(bool random) { + // Find a free block. + s64 soffset = m_bitmap.FindFreeBlock(random); if (soffset < 0) { - return 0; + return {}; } - const auto offset{static_cast<std::size_t>(soffset)}; + const size_t offset = static_cast<size_t>(soffset); - // Update our tracking and return it - bitmap.ClearBit(offset); - return heap_address + (offset << GetShift()); + // Update our tracking and return it. + m_bitmap.ClearBit(offset); + return m_heap_address + (offset << this->GetShift()); } - static constexpr std::size_t CalculateManagementOverheadSize(std::size_t region_size, - std::size_t cur_block_shift, - std::size_t next_block_shift) { - const auto cur_block_size{(1ULL << cur_block_shift)}; - const auto next_block_size{(1ULL << next_block_shift)}; - const auto align{(next_block_shift != 0) ? next_block_size : cur_block_size}; + public: + static constexpr size_t CalculateManagementOverheadSize(size_t region_size, + size_t cur_block_shift, + size_t next_block_shift) { + const size_t cur_block_size = (u64(1) << cur_block_shift); + const size_t next_block_size = (u64(1) << next_block_shift); + const size_t align = (next_block_shift != 0) ? next_block_size : cur_block_size; return KPageBitmap::CalculateManagementOverheadSize( (align * 2 + Common::AlignUp(region_size, align)) / cur_block_size); } private: - KPageBitmap bitmap; - VAddr heap_address{}; - uintptr_t end_offset{}; - std::size_t block_shift{}; - std::size_t next_block_shift{}; + KPageBitmap m_bitmap; + PAddr m_heap_address{}; + uintptr_t m_end_offset{}; + size_t m_block_shift{}; + size_t m_next_block_shift{}; }; - constexpr std::size_t GetNumFreePages() const { - std::size_t num_free{}; - - for (const auto& block : blocks) { - num_free += block.GetNumFreePages(); - } - - return num_free; - } +private: + void Initialize(PAddr heap_address, size_t heap_size, VAddr management_address, + size_t management_size, const size_t* block_shifts, size_t num_block_shifts); + size_t GetNumFreePages() const; - void FreeBlock(VAddr block, s32 index); + void FreeBlock(PAddr block, s32 index); - static constexpr std::size_t NumMemoryBlockPageShifts{7}; - static constexpr std::array<std::size_t, NumMemoryBlockPageShifts> MemoryBlockPageShifts{ + static constexpr size_t NumMemoryBlockPageShifts{7}; + static constexpr std::array<size_t, NumMemoryBlockPageShifts> MemoryBlockPageShifts{ 0xC, 0x10, 0x15, 0x16, 0x19, 0x1D, 0x1E, }; - VAddr heap_address{}; - std::size_t heap_size{}; - std::size_t used_size{}; - std::array<Block, NumMemoryBlockPageShifts> blocks{}; - std::vector<u64> metadata; +private: + static size_t CalculateManagementOverheadSize(size_t region_size, const size_t* block_shifts, + size_t num_block_shifts); + +private: + PAddr m_heap_address{}; + size_t m_heap_size{}; + size_t m_initial_used_size{}; + size_t m_num_blocks{}; + std::array<Block, NumMemoryBlockPageShifts> m_blocks{}; + std::vector<u64> m_management_data; }; } // namespace Kernel diff --git a/src/core/hle/kernel/k_page_table.cpp b/src/core/hle/kernel/k_page_table.cpp index 912853e5c..dfea0b6e2 100644 --- a/src/core/hle/kernel/k_page_table.cpp +++ b/src/core/hle/kernel/k_page_table.cpp @@ -41,24 +41,6 @@ constexpr std::size_t GetAddressSpaceWidthFromType(FileSys::ProgramAddressSpaceT } } -constexpr u64 GetAddressInRange(const KMemoryInfo& info, VAddr addr) { - if (info.GetAddress() < addr) { - return addr; - } - return info.GetAddress(); -} - -constexpr std::size_t GetSizeInRange(const KMemoryInfo& info, VAddr start, VAddr end) { - std::size_t size{info.GetSize()}; - if (info.GetAddress() < start) { - size -= start - info.GetAddress(); - } - if (info.GetEndAddress() > end) { - size -= info.GetEndAddress() - end; - } - return size; -} - } // namespace KPageTable::KPageTable(Core::System& system_) @@ -291,11 +273,12 @@ ResultCode KPageTable::MapProcessCode(VAddr addr, std::size_t num_pages, KMemory R_TRY(this->CheckMemoryState(addr, size, KMemoryState::All, KMemoryState::Free, KMemoryPermission::None, KMemoryPermission::None, KMemoryAttribute::None, KMemoryAttribute::None)); + KPageLinkedList pg; + R_TRY(system.Kernel().MemoryManager().AllocateAndOpen( + &pg, num_pages, + KMemoryManager::EncodeOption(KMemoryManager::Pool::Application, allocation_option))); - KPageLinkedList page_linked_list; - R_TRY(system.Kernel().MemoryManager().Allocate(page_linked_list, num_pages, memory_pool, - allocation_option)); - R_TRY(Operate(addr, num_pages, page_linked_list, OperationType::MapGroup)); + R_TRY(Operate(addr, num_pages, pg, OperationType::MapGroup)); block_manager->Update(addr, num_pages, state, perm); @@ -400,148 +383,476 @@ ResultCode KPageTable::UnmapProcessMemory(VAddr dst_addr, std::size_t size, return ResultSuccess; } -ResultCode KPageTable::MapPhysicalMemory(VAddr addr, std::size_t size) { +ResultCode KPageTable::MapPhysicalMemory(VAddr address, std::size_t size) { // Lock the physical memory lock. KScopedLightLock map_phys_mem_lk(map_physical_memory_lock); - // Lock the table. - KScopedLightLock lk(general_lock); + // Calculate the last address for convenience. + const VAddr last_address = address + size - 1; - std::size_t mapped_size{}; - const VAddr end_addr{addr + size}; + // Define iteration variables. + VAddr cur_address; + std::size_t mapped_size; - block_manager->IterateForRange(addr, end_addr, [&](const KMemoryInfo& info) { - if (info.state != KMemoryState::Free) { - mapped_size += GetSizeInRange(info, addr, end_addr); - } - }); - - if (mapped_size == size) { - return ResultSuccess; - } + // The entire mapping process can be retried. + while (true) { + // Check if the memory is already mapped. + { + // Lock the table. + KScopedLightLock lk(general_lock); + + // Iterate over the memory. + cur_address = address; + mapped_size = 0; + + auto it = block_manager->FindIterator(cur_address); + while (true) { + // Check that the iterator is valid. + ASSERT(it != block_manager->end()); + + // Get the memory info. + const KMemoryInfo info = it->GetMemoryInfo(); + + // Check if we're done. + if (last_address <= info.GetLastAddress()) { + if (info.GetState() != KMemoryState::Free) { + mapped_size += (last_address + 1 - cur_address); + } + break; + } + + // Track the memory if it's mapped. + if (info.GetState() != KMemoryState::Free) { + mapped_size += VAddr(info.GetEndAddress()) - cur_address; + } + + // Advance. + cur_address = info.GetEndAddress(); + ++it; + } - const std::size_t remaining_size{size - mapped_size}; - const std::size_t remaining_pages{remaining_size / PageSize}; + // If the size mapped is the size requested, we've nothing to do. + R_SUCCEED_IF(size == mapped_size); + } - // Reserve the memory from the process resource limit. - KScopedResourceReservation memory_reservation( - system.Kernel().CurrentProcess()->GetResourceLimit(), LimitableResource::PhysicalMemory, - remaining_size); - if (!memory_reservation.Succeeded()) { - LOG_ERROR(Kernel, "Could not reserve remaining {:X} bytes", remaining_size); - return ResultLimitReached; + // Allocate and map the memory. + { + // Reserve the memory from the process resource limit. + KScopedResourceReservation memory_reservation( + system.Kernel().CurrentProcess()->GetResourceLimit(), + LimitableResource::PhysicalMemory, size - mapped_size); + R_UNLESS(memory_reservation.Succeeded(), ResultLimitReached); + + // Allocate pages for the new memory. + KPageLinkedList pg; + R_TRY(system.Kernel().MemoryManager().AllocateAndOpenForProcess( + &pg, (size - mapped_size) / PageSize, + KMemoryManager::EncodeOption(memory_pool, allocation_option), 0, 0)); + + // Map the memory. + { + // Lock the table. + KScopedLightLock lk(general_lock); + + size_t num_allocator_blocks = 0; + + // Verify that nobody has mapped memory since we first checked. + { + // Iterate over the memory. + size_t checked_mapped_size = 0; + cur_address = address; + + auto it = block_manager->FindIterator(cur_address); + while (true) { + // Check that the iterator is valid. + ASSERT(it != block_manager->end()); + + // Get the memory info. + const KMemoryInfo info = it->GetMemoryInfo(); + + const bool is_free = info.GetState() == KMemoryState::Free; + if (is_free) { + if (info.GetAddress() < address) { + ++num_allocator_blocks; + } + if (last_address < info.GetLastAddress()) { + ++num_allocator_blocks; + } + } + + // Check if we're done. + if (last_address <= info.GetLastAddress()) { + if (!is_free) { + checked_mapped_size += (last_address + 1 - cur_address); + } + break; + } + + // Track the memory if it's mapped. + if (!is_free) { + checked_mapped_size += VAddr(info.GetEndAddress()) - cur_address; + } + + // Advance. + cur_address = info.GetEndAddress(); + ++it; + } + + // If the size now isn't what it was before, somebody mapped or unmapped + // concurrently. If this happened, retry. + if (mapped_size != checked_mapped_size) { + continue; + } + } + + // Reset the current tracking address, and make sure we clean up on failure. + cur_address = address; + auto unmap_guard = detail::ScopeExit([&] { + if (cur_address > address) { + const VAddr last_unmap_address = cur_address - 1; + + // Iterate, unmapping the pages. + cur_address = address; + + auto it = block_manager->FindIterator(cur_address); + while (true) { + // Check that the iterator is valid. + ASSERT(it != block_manager->end()); + + // Get the memory info. + const KMemoryInfo info = it->GetMemoryInfo(); + + // If the memory state is free, we mapped it and need to unmap it. + if (info.GetState() == KMemoryState::Free) { + // Determine the range to unmap. + const size_t cur_pages = + std::min(VAddr(info.GetEndAddress()) - cur_address, + last_unmap_address + 1 - cur_address) / + PageSize; + + // Unmap. + ASSERT(Operate(cur_address, cur_pages, KMemoryPermission::None, + OperationType::Unmap) + .IsSuccess()); + } + + // Check if we're done. + if (last_unmap_address <= info.GetLastAddress()) { + break; + } + + // Advance. + cur_address = info.GetEndAddress(); + ++it; + } + } + }); + + // Iterate over the memory. + auto pg_it = pg.Nodes().begin(); + PAddr pg_phys_addr = pg_it->GetAddress(); + size_t pg_pages = pg_it->GetNumPages(); + + auto it = block_manager->FindIterator(cur_address); + while (true) { + // Check that the iterator is valid. + ASSERT(it != block_manager->end()); + + // Get the memory info. + const KMemoryInfo info = it->GetMemoryInfo(); + + // If it's unmapped, we need to map it. + if (info.GetState() == KMemoryState::Free) { + // Determine the range to map. + size_t map_pages = std::min(VAddr(info.GetEndAddress()) - cur_address, + last_address + 1 - cur_address) / + PageSize; + + // While we have pages to map, map them. + while (map_pages > 0) { + // Check if we're at the end of the physical block. + if (pg_pages == 0) { + // Ensure there are more pages to map. + ASSERT(pg_it != pg.Nodes().end()); + + // Advance our physical block. + ++pg_it; + pg_phys_addr = pg_it->GetAddress(); + pg_pages = pg_it->GetNumPages(); + } + + // Map whatever we can. + const size_t cur_pages = std::min(pg_pages, map_pages); + R_TRY(Operate(cur_address, cur_pages, KMemoryPermission::UserReadWrite, + OperationType::Map, pg_phys_addr)); + + // Advance. + cur_address += cur_pages * PageSize; + map_pages -= cur_pages; + + pg_phys_addr += cur_pages * PageSize; + pg_pages -= cur_pages; + } + } + + // Check if we're done. + if (last_address <= info.GetLastAddress()) { + break; + } + + // Advance. + cur_address = info.GetEndAddress(); + ++it; + } + + // We succeeded, so commit the memory reservation. + memory_reservation.Commit(); + + // Increase our tracked mapped size. + mapped_physical_memory_size += (size - mapped_size); + + // Update the relevant memory blocks. + block_manager->Update(address, size / PageSize, KMemoryState::Free, + KMemoryPermission::None, KMemoryAttribute::None, + KMemoryState::Normal, KMemoryPermission::UserReadWrite, + KMemoryAttribute::None); + + // Cancel our guard. + unmap_guard.Cancel(); + + return ResultSuccess; + } + } } +} - KPageLinkedList page_linked_list; +ResultCode KPageTable::UnmapPhysicalMemory(VAddr address, std::size_t size) { + // Lock the physical memory lock. + KScopedLightLock map_phys_mem_lk(map_physical_memory_lock); - CASCADE_CODE(system.Kernel().MemoryManager().Allocate(page_linked_list, remaining_pages, - memory_pool, allocation_option)); + // Lock the table. + KScopedLightLock lk(general_lock); - // We succeeded, so commit the memory reservation. - memory_reservation.Commit(); + // Calculate the last address for convenience. + const VAddr last_address = address + size - 1; - // Map the memory. - auto node{page_linked_list.Nodes().begin()}; - PAddr map_addr{node->GetAddress()}; - std::size_t src_num_pages{node->GetNumPages()}; - block_manager->IterateForRange(addr, end_addr, [&](const KMemoryInfo& info) { - if (info.state != KMemoryState::Free) { - return; - } + // Define iteration variables. + VAddr cur_address = 0; + std::size_t mapped_size = 0; + std::size_t num_allocator_blocks = 0; - std::size_t dst_num_pages{GetSizeInRange(info, addr, end_addr) / PageSize}; - VAddr dst_addr{GetAddressInRange(info, addr)}; + // Check if the memory is mapped. + { + // Iterate over the memory. + cur_address = address; + mapped_size = 0; + + auto it = block_manager->FindIterator(cur_address); + while (true) { + // Check that the iterator is valid. + ASSERT(it != block_manager->end()); + + // Get the memory info. + const KMemoryInfo info = it->GetMemoryInfo(); + + // Verify the memory's state. + const bool is_normal = info.GetState() == KMemoryState::Normal && + info.GetAttribute() == KMemoryAttribute::None; + const bool is_free = info.GetState() == KMemoryState::Free; + R_UNLESS(is_normal || is_free, ResultInvalidCurrentMemory); + + if (is_normal) { + R_UNLESS(info.GetAttribute() == KMemoryAttribute::None, ResultInvalidCurrentMemory); + + if (info.GetAddress() < address) { + ++num_allocator_blocks; + } + if (last_address < info.GetLastAddress()) { + ++num_allocator_blocks; + } + } - while (dst_num_pages) { - if (!src_num_pages) { - node = std::next(node); - map_addr = node->GetAddress(); - src_num_pages = node->GetNumPages(); + // Check if we're done. + if (last_address <= info.GetLastAddress()) { + if (is_normal) { + mapped_size += (last_address + 1 - cur_address); + } + break; } - const std::size_t num_pages{std::min(src_num_pages, dst_num_pages)}; - Operate(dst_addr, num_pages, KMemoryPermission::UserReadWrite, OperationType::Map, - map_addr); + // Track the memory if it's mapped. + if (is_normal) { + mapped_size += VAddr(info.GetEndAddress()) - cur_address; + } - dst_addr += num_pages * PageSize; - map_addr += num_pages * PageSize; - src_num_pages -= num_pages; - dst_num_pages -= num_pages; + // Advance. + cur_address = info.GetEndAddress(); + ++it; } - }); - mapped_physical_memory_size += remaining_size; - - const std::size_t num_pages{size / PageSize}; - block_manager->Update(addr, num_pages, KMemoryState::Free, KMemoryPermission::None, - KMemoryAttribute::None, KMemoryState::Normal, - KMemoryPermission::UserReadWrite, KMemoryAttribute::None); - - return ResultSuccess; -} + // If there's nothing mapped, we've nothing to do. + R_SUCCEED_IF(mapped_size == 0); + } -ResultCode KPageTable::UnmapPhysicalMemory(VAddr addr, std::size_t size) { - // Lock the physical memory lock. - KScopedLightLock map_phys_mem_lk(map_physical_memory_lock); + // Make a page group for the unmap region. + KPageLinkedList pg; + { + auto& impl = this->PageTableImpl(); + + // Begin traversal. + Common::PageTable::TraversalContext context; + Common::PageTable::TraversalEntry cur_entry = {.phys_addr = 0, .block_size = 0}; + bool cur_valid = false; + Common::PageTable::TraversalEntry next_entry; + bool next_valid = false; + size_t tot_size = 0; + + cur_address = address; + next_valid = impl.BeginTraversal(next_entry, context, cur_address); + next_entry.block_size = + (next_entry.block_size - (next_entry.phys_addr & (next_entry.block_size - 1))); + + // Iterate, building the group. + while (true) { + if ((!next_valid && !cur_valid) || + (next_valid && cur_valid && + next_entry.phys_addr == cur_entry.phys_addr + cur_entry.block_size)) { + cur_entry.block_size += next_entry.block_size; + } else { + if (cur_valid) { + // ASSERT(IsHeapPhysicalAddress(cur_entry.phys_addr)); + R_TRY(pg.AddBlock(cur_entry.phys_addr, cur_entry.block_size / PageSize)); + } + + // Update tracking variables. + tot_size += cur_entry.block_size; + cur_entry = next_entry; + cur_valid = next_valid; + } - // Lock the table. - KScopedLightLock lk(general_lock); + if (cur_entry.block_size + tot_size >= size) { + break; + } - const VAddr end_addr{addr + size}; - ResultCode result{ResultSuccess}; - std::size_t mapped_size{}; + next_valid = impl.ContinueTraversal(next_entry, context); + } - // Verify that the region can be unmapped - block_manager->IterateForRange(addr, end_addr, [&](const KMemoryInfo& info) { - if (info.state == KMemoryState::Normal) { - if (info.attribute != KMemoryAttribute::None) { - result = ResultInvalidCurrentMemory; - return; + // Add the last block. + if (cur_valid) { + // ASSERT(IsHeapPhysicalAddress(cur_entry.phys_addr)); + R_TRY(pg.AddBlock(cur_entry.phys_addr, (size - tot_size) / PageSize)); + } + } + ASSERT(pg.GetNumPages() == mapped_size / PageSize); + + // Reset the current tracking address, and make sure we clean up on failure. + cur_address = address; + auto remap_guard = detail::ScopeExit([&] { + if (cur_address > address) { + const VAddr last_map_address = cur_address - 1; + cur_address = address; + + // Iterate over the memory we unmapped. + auto it = block_manager->FindIterator(cur_address); + auto pg_it = pg.Nodes().begin(); + PAddr pg_phys_addr = pg_it->GetAddress(); + size_t pg_pages = pg_it->GetNumPages(); + + while (true) { + // Get the memory info for the pages we unmapped, convert to property. + const KMemoryInfo info = it->GetMemoryInfo(); + + // If the memory is normal, we unmapped it and need to re-map it. + if (info.GetState() == KMemoryState::Normal) { + // Determine the range to map. + size_t map_pages = std::min(VAddr(info.GetEndAddress()) - cur_address, + last_map_address + 1 - cur_address) / + PageSize; + + // While we have pages to map, map them. + while (map_pages > 0) { + // Check if we're at the end of the physical block. + if (pg_pages == 0) { + // Ensure there are more pages to map. + ASSERT(pg_it != pg.Nodes().end()); + + // Advance our physical block. + ++pg_it; + pg_phys_addr = pg_it->GetAddress(); + pg_pages = pg_it->GetNumPages(); + } + + // Map whatever we can. + const size_t cur_pages = std::min(pg_pages, map_pages); + ASSERT(this->Operate(cur_address, cur_pages, info.GetPermission(), + OperationType::Map, pg_phys_addr) == ResultSuccess); + + // Advance. + cur_address += cur_pages * PageSize; + map_pages -= cur_pages; + + pg_phys_addr += cur_pages * PageSize; + pg_pages -= cur_pages; + } + } + + // Check if we're done. + if (last_map_address <= info.GetLastAddress()) { + break; + } + + // Advance. + ++it; } - mapped_size += GetSizeInRange(info, addr, end_addr); - } else if (info.state != KMemoryState::Free) { - result = ResultInvalidCurrentMemory; } }); - if (result.IsError()) { - return result; - } + // Iterate over the memory, unmapping as we go. + auto it = block_manager->FindIterator(cur_address); + while (true) { + // Check that the iterator is valid. + ASSERT(it != block_manager->end()); - if (!mapped_size) { - return ResultSuccess; - } + // Get the memory info. + const KMemoryInfo info = it->GetMemoryInfo(); - // Unmap each region within the range - KPageLinkedList page_linked_list; - block_manager->IterateForRange(addr, end_addr, [&](const KMemoryInfo& info) { - if (info.state == KMemoryState::Normal) { - const std::size_t block_size{GetSizeInRange(info, addr, end_addr)}; - const std::size_t block_num_pages{block_size / PageSize}; - const VAddr block_addr{GetAddressInRange(info, addr)}; - - AddRegionToPages(block_addr, block_size / PageSize, page_linked_list); - - if (result = Operate(block_addr, block_num_pages, KMemoryPermission::None, - OperationType::Unmap); - result.IsError()) { - return; - } + // If the memory state is normal, we need to unmap it. + if (info.GetState() == KMemoryState::Normal) { + // Determine the range to unmap. + const size_t cur_pages = std::min(VAddr(info.GetEndAddress()) - cur_address, + last_address + 1 - cur_address) / + PageSize; + + // Unmap. + R_TRY(Operate(cur_address, cur_pages, KMemoryPermission::None, OperationType::Unmap)); } - }); - if (result.IsError()) { - return result; - } - const std::size_t num_pages{size / PageSize}; - system.Kernel().MemoryManager().Free(page_linked_list, num_pages, memory_pool, - allocation_option); + // Check if we're done. + if (last_address <= info.GetLastAddress()) { + break; + } - block_manager->Update(addr, num_pages, KMemoryState::Free); + // Advance. + cur_address = info.GetEndAddress(); + ++it; + } + // Release the memory resource. + mapped_physical_memory_size -= mapped_size; auto process{system.Kernel().CurrentProcess()}; process->GetResourceLimit()->Release(LimitableResource::PhysicalMemory, mapped_size); - mapped_physical_memory_size -= mapped_size; + + // Update memory blocks. + block_manager->Update(address, size / PageSize, KMemoryState::Free, KMemoryPermission::None, + KMemoryAttribute::None); + + // TODO(bunnei): This is a workaround until the next set of changes, where we add reference + // counting for mapped pages. Until then, we must manually close the reference to the page + // group. + system.Kernel().MemoryManager().Close(pg); + + // We succeeded. + remap_guard.Cancel(); return ResultSuccess; } @@ -681,9 +992,8 @@ ResultCode KPageTable::UnmapPages(VAddr addr, const KPageLinkedList& page_linked VAddr cur_addr{addr}; for (const auto& node : page_linked_list.Nodes()) { - const std::size_t num_pages{(addr - cur_addr) / PageSize}; - if (const auto result{ - Operate(addr, num_pages, KMemoryPermission::None, OperationType::Unmap)}; + if (const auto result{Operate(cur_addr, node.GetNumPages(), KMemoryPermission::None, + OperationType::Unmap)}; result.IsError()) { return result; } @@ -966,9 +1276,16 @@ ResultCode KPageTable::SetHeapSize(VAddr* out, std::size_t size) { R_UNLESS(memory_reservation.Succeeded(), ResultLimitReached); // Allocate pages for the heap extension. - KPageLinkedList page_linked_list; - R_TRY(system.Kernel().MemoryManager().Allocate(page_linked_list, allocation_size / PageSize, - memory_pool, allocation_option)); + KPageLinkedList pg; + R_TRY(system.Kernel().MemoryManager().AllocateAndOpen( + &pg, allocation_size / PageSize, + KMemoryManager::EncodeOption(memory_pool, allocation_option))); + + // Clear all the newly allocated pages. + for (const auto& it : pg.Nodes()) { + std::memset(system.DeviceMemory().GetPointer(it.GetAddress()), heap_fill_value, + it.GetSize()); + } // Map the pages. { @@ -987,7 +1304,7 @@ ResultCode KPageTable::SetHeapSize(VAddr* out, std::size_t size) { // Map the pages. const auto num_pages = allocation_size / PageSize; - R_TRY(Operate(current_heap_end, num_pages, page_linked_list, OperationType::MapGroup)); + R_TRY(Operate(current_heap_end, num_pages, pg, OperationType::MapGroup)); // Clear all the newly allocated pages. for (std::size_t cur_page = 0; cur_page < num_pages; ++cur_page) { @@ -1035,8 +1352,9 @@ ResultVal<VAddr> KPageTable::AllocateAndMapMemory(std::size_t needed_num_pages, R_TRY(Operate(addr, needed_num_pages, perm, OperationType::Map, map_addr)); } else { KPageLinkedList page_group; - R_TRY(system.Kernel().MemoryManager().Allocate(page_group, needed_num_pages, memory_pool, - allocation_option)); + R_TRY(system.Kernel().MemoryManager().AllocateAndOpenForProcess( + &page_group, needed_num_pages, + KMemoryManager::EncodeOption(memory_pool, allocation_option), 0, 0)); R_TRY(Operate(addr, needed_num_pages, page_group, OperationType::MapGroup)); } @@ -1243,7 +1561,7 @@ ResultCode KPageTable::Operate(VAddr addr, std::size_t num_pages, KMemoryPermiss return ResultSuccess; } -constexpr VAddr KPageTable::GetRegionAddress(KMemoryState state) const { +VAddr KPageTable::GetRegionAddress(KMemoryState state) const { switch (state) { case KMemoryState::Free: case KMemoryState::Kernel: @@ -1279,7 +1597,7 @@ constexpr VAddr KPageTable::GetRegionAddress(KMemoryState state) const { } } -constexpr std::size_t KPageTable::GetRegionSize(KMemoryState state) const { +std::size_t KPageTable::GetRegionSize(KMemoryState state) const { switch (state) { case KMemoryState::Free: case KMemoryState::Kernel: diff --git a/src/core/hle/kernel/k_page_table.h b/src/core/hle/kernel/k_page_table.h index c98887d34..194177332 100644 --- a/src/core/hle/kernel/k_page_table.h +++ b/src/core/hle/kernel/k_page_table.h @@ -102,8 +102,8 @@ private: OperationType operation); ResultCode Operate(VAddr addr, std::size_t num_pages, KMemoryPermission perm, OperationType operation, PAddr map_addr = 0); - constexpr VAddr GetRegionAddress(KMemoryState state) const; - constexpr std::size_t GetRegionSize(KMemoryState state) const; + VAddr GetRegionAddress(KMemoryState state) const; + std::size_t GetRegionSize(KMemoryState state) const; ResultCode CheckMemoryStateContiguous(std::size_t* out_blocks_needed, VAddr addr, std::size_t size, KMemoryState state_mask, @@ -254,8 +254,7 @@ public: return !IsOutsideASLRRegion(address, size); } - PAddr GetPhysicalAddr(VAddr addr) { - ASSERT(IsLockedByCurrentThread()); + PAddr GetPhysicalAddr(VAddr addr) const { const auto backing_addr = page_table_impl.backing_addr[addr >> PageBits]; ASSERT(backing_addr); return backing_addr + addr; @@ -311,6 +310,8 @@ private: bool is_kernel{}; bool is_aslr_enabled{}; + u32 heap_fill_value{}; + KMemoryManager::Pool memory_pool{KMemoryManager::Pool::Application}; KMemoryManager::Direction allocation_option{KMemoryManager::Direction::FromFront}; diff --git a/src/core/hle/kernel/k_process.cpp b/src/core/hle/kernel/k_process.cpp index 85c506979..9233261cd 100644 --- a/src/core/hle/kernel/k_process.cpp +++ b/src/core/hle/kernel/k_process.cpp @@ -123,12 +123,11 @@ private: }; ResultCode KProcess::Initialize(KProcess* process, Core::System& system, std::string process_name, - ProcessType type) { + ProcessType type, KResourceLimit* res_limit) { auto& kernel = system.Kernel(); process->name = std::move(process_name); - - process->resource_limit = kernel.GetSystemResourceLimit(); + process->resource_limit = res_limit; process->status = ProcessStatus::Created; process->program_id = 0; process->process_id = type == ProcessType::KernelInternal ? kernel.CreateNewKernelProcessID() @@ -143,9 +142,6 @@ ResultCode KProcess::Initialize(KProcess* process, Core::System& system, std::st kernel.AppendNewProcess(process); - // Open a reference to the resource limit. - process->resource_limit->Open(); - // Clear remaining fields. process->num_running_threads = 0; process->is_signaled = false; @@ -153,6 +149,9 @@ ResultCode KProcess::Initialize(KProcess* process, Core::System& system, std::st process->is_suspended = false; process->schedule_count = 0; + // Open a reference to the resource limit. + process->resource_limit->Open(); + return ResultSuccess; } diff --git a/src/core/hle/kernel/k_process.h b/src/core/hle/kernel/k_process.h index 38b446350..cf1b67428 100644 --- a/src/core/hle/kernel/k_process.h +++ b/src/core/hle/kernel/k_process.h @@ -91,7 +91,7 @@ public: static constexpr std::size_t RANDOM_ENTROPY_SIZE = 4; static ResultCode Initialize(KProcess* process, Core::System& system, std::string process_name, - ProcessType type); + ProcessType type, KResourceLimit* res_limit); /// Gets a reference to the process' page table. KPageTable& PageTable() { diff --git a/src/core/hle/kernel/k_resource_limit.cpp b/src/core/hle/kernel/k_resource_limit.cpp index 0c4bba66b..a84977c68 100644 --- a/src/core/hle/kernel/k_resource_limit.cpp +++ b/src/core/hle/kernel/k_resource_limit.cpp @@ -3,6 +3,7 @@ // Refer to the license.txt file included. #include "common/assert.h" +#include "core/core.h" #include "core/core_timing.h" #include "core/hle/kernel/k_resource_limit.h" #include "core/hle/kernel/svc_results.h" @@ -151,4 +152,22 @@ void KResourceLimit::Release(LimitableResource which, s64 value, s64 hint) { } } +KResourceLimit* CreateResourceLimitForProcess(Core::System& system, s64 physical_memory_size) { + auto* resource_limit = KResourceLimit::Create(system.Kernel()); + resource_limit->Initialize(&system.CoreTiming()); + + // Initialize default resource limit values. + // TODO(bunnei): These values are the system defaults, the limits for service processes are + // lower. These should use the correct limit values. + + ASSERT(resource_limit->SetLimitValue(LimitableResource::PhysicalMemory, physical_memory_size) + .IsSuccess()); + ASSERT(resource_limit->SetLimitValue(LimitableResource::Threads, 800).IsSuccess()); + ASSERT(resource_limit->SetLimitValue(LimitableResource::Events, 900).IsSuccess()); + ASSERT(resource_limit->SetLimitValue(LimitableResource::TransferMemory, 200).IsSuccess()); + ASSERT(resource_limit->SetLimitValue(LimitableResource::Sessions, 1133).IsSuccess()); + + return resource_limit; +} + } // namespace Kernel diff --git a/src/core/hle/kernel/k_resource_limit.h b/src/core/hle/kernel/k_resource_limit.h index fab6005ff..d23d16aa4 100644 --- a/src/core/hle/kernel/k_resource_limit.h +++ b/src/core/hle/kernel/k_resource_limit.h @@ -67,4 +67,7 @@ private: KLightConditionVariable cond_var; const Core::Timing::CoreTiming* core_timing{}; }; + +KResourceLimit* CreateResourceLimitForProcess(Core::System& system, s64 physical_memory_size); + } // namespace Kernel diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp index 49c0714ed..71bd466cf 100644 --- a/src/core/hle/kernel/kernel.cpp +++ b/src/core/hle/kernel/kernel.cpp @@ -70,13 +70,12 @@ struct KernelCore::Impl { // Derive the initial memory layout from the emulated board Init::InitializeSlabResourceCounts(kernel); - KMemoryLayout memory_layout; - DeriveInitialMemoryLayout(memory_layout); - Init::InitializeSlabHeaps(system, memory_layout); + DeriveInitialMemoryLayout(); + Init::InitializeSlabHeaps(system, *memory_layout); // Initialize kernel memory and resources. - InitializeSystemResourceLimit(kernel, system.CoreTiming(), memory_layout); - InitializeMemoryLayout(memory_layout); + InitializeSystemResourceLimit(kernel, system.CoreTiming()); + InitializeMemoryLayout(); InitializePageSlab(); InitializeSchedulers(); InitializeSuspendThreads(); @@ -219,12 +218,11 @@ struct KernelCore::Impl { // Creates the default system resource limit void InitializeSystemResourceLimit(KernelCore& kernel, - const Core::Timing::CoreTiming& core_timing, - const KMemoryLayout& memory_layout) { + const Core::Timing::CoreTiming& core_timing) { system_resource_limit = KResourceLimit::Create(system.Kernel()); system_resource_limit->Initialize(&core_timing); - const auto [total_size, kernel_size] = memory_layout.GetTotalAndKernelMemorySizes(); + const auto [total_size, kernel_size] = memory_layout->GetTotalAndKernelMemorySizes(); // If setting the default system values fails, then something seriously wrong has occurred. ASSERT(system_resource_limit->SetLimitValue(LimitableResource::PhysicalMemory, total_size) @@ -240,13 +238,6 @@ struct KernelCore::Impl { constexpr u64 secure_applet_memory_size{4_MiB}; ASSERT(system_resource_limit->Reserve(LimitableResource::PhysicalMemory, secure_applet_memory_size)); - - // This memory seems to be reserved on hardware, but is not reserved/used by yuzu. - // Likely Horizon OS reserved memory - // TODO(ameerj): Derive the memory rather than hardcode it. - constexpr u64 unknown_reserved_memory{0x2f896000}; - ASSERT(system_resource_limit->Reserve(LimitableResource::PhysicalMemory, - unknown_reserved_memory)); } void InitializePreemption(KernelCore& kernel) { @@ -360,16 +351,18 @@ struct KernelCore::Impl { return schedulers[thread_id]->GetCurrentThread(); } - void DeriveInitialMemoryLayout(KMemoryLayout& memory_layout) { + void DeriveInitialMemoryLayout() { + memory_layout = std::make_unique<KMemoryLayout>(); + // Insert the root region for the virtual memory tree, from which all other regions will // derive. - memory_layout.GetVirtualMemoryRegionTree().InsertDirectly( + memory_layout->GetVirtualMemoryRegionTree().InsertDirectly( KernelVirtualAddressSpaceBase, KernelVirtualAddressSpaceBase + KernelVirtualAddressSpaceSize - 1); // Insert the root region for the physical memory tree, from which all other regions will // derive. - memory_layout.GetPhysicalMemoryRegionTree().InsertDirectly( + memory_layout->GetPhysicalMemoryRegionTree().InsertDirectly( KernelPhysicalAddressSpaceBase, KernelPhysicalAddressSpaceBase + KernelPhysicalAddressSpaceSize - 1); @@ -386,7 +379,7 @@ struct KernelCore::Impl { if (!(kernel_region_start + KernelRegionSize - 1 <= KernelVirtualAddressSpaceLast)) { kernel_region_size = KernelVirtualAddressSpaceEnd - kernel_region_start; } - ASSERT(memory_layout.GetVirtualMemoryRegionTree().Insert( + ASSERT(memory_layout->GetVirtualMemoryRegionTree().Insert( kernel_region_start, kernel_region_size, KMemoryRegionType_Kernel)); // Setup the code region. @@ -395,11 +388,11 @@ struct KernelCore::Impl { Common::AlignDown(code_start_virt_addr, CodeRegionAlign); constexpr VAddr code_region_end = Common::AlignUp(code_end_virt_addr, CodeRegionAlign); constexpr size_t code_region_size = code_region_end - code_region_start; - ASSERT(memory_layout.GetVirtualMemoryRegionTree().Insert( + ASSERT(memory_layout->GetVirtualMemoryRegionTree().Insert( code_region_start, code_region_size, KMemoryRegionType_KernelCode)); // Setup board-specific device physical regions. - Init::SetupDevicePhysicalMemoryRegions(memory_layout); + Init::SetupDevicePhysicalMemoryRegions(*memory_layout); // Determine the amount of space needed for the misc region. size_t misc_region_needed_size; @@ -408,7 +401,7 @@ struct KernelCore::Impl { misc_region_needed_size = Core::Hardware::NUM_CPU_CORES * (3 * (PageSize + PageSize)); // Account for each auto-map device. - for (const auto& region : memory_layout.GetPhysicalMemoryRegionTree()) { + for (const auto& region : memory_layout->GetPhysicalMemoryRegionTree()) { if (region.HasTypeAttribute(KMemoryRegionAttr_ShouldKernelMap)) { // Check that the region is valid. ASSERT(region.GetEndAddress() != 0); @@ -433,22 +426,22 @@ struct KernelCore::Impl { // Setup the misc region. const VAddr misc_region_start = - memory_layout.GetVirtualMemoryRegionTree().GetRandomAlignedRegion( + memory_layout->GetVirtualMemoryRegionTree().GetRandomAlignedRegion( misc_region_size, MiscRegionAlign, KMemoryRegionType_Kernel); - ASSERT(memory_layout.GetVirtualMemoryRegionTree().Insert( + ASSERT(memory_layout->GetVirtualMemoryRegionTree().Insert( misc_region_start, misc_region_size, KMemoryRegionType_KernelMisc)); // Setup the stack region. constexpr size_t StackRegionSize = 14_MiB; constexpr size_t StackRegionAlign = KernelAslrAlignment; const VAddr stack_region_start = - memory_layout.GetVirtualMemoryRegionTree().GetRandomAlignedRegion( + memory_layout->GetVirtualMemoryRegionTree().GetRandomAlignedRegion( StackRegionSize, StackRegionAlign, KMemoryRegionType_Kernel); - ASSERT(memory_layout.GetVirtualMemoryRegionTree().Insert( + ASSERT(memory_layout->GetVirtualMemoryRegionTree().Insert( stack_region_start, StackRegionSize, KMemoryRegionType_KernelStack)); // Determine the size of the resource region. - const size_t resource_region_size = memory_layout.GetResourceRegionSizeForInit(); + const size_t resource_region_size = memory_layout->GetResourceRegionSizeForInit(); // Determine the size of the slab region. const size_t slab_region_size = @@ -465,23 +458,23 @@ struct KernelCore::Impl { Common::AlignUp(code_end_phys_addr + slab_region_size, SlabRegionAlign) - Common::AlignDown(code_end_phys_addr, SlabRegionAlign); const VAddr slab_region_start = - memory_layout.GetVirtualMemoryRegionTree().GetRandomAlignedRegion( + memory_layout->GetVirtualMemoryRegionTree().GetRandomAlignedRegion( slab_region_needed_size, SlabRegionAlign, KMemoryRegionType_Kernel) + (code_end_phys_addr % SlabRegionAlign); - ASSERT(memory_layout.GetVirtualMemoryRegionTree().Insert( + ASSERT(memory_layout->GetVirtualMemoryRegionTree().Insert( slab_region_start, slab_region_size, KMemoryRegionType_KernelSlab)); // Setup the temp region. constexpr size_t TempRegionSize = 128_MiB; constexpr size_t TempRegionAlign = KernelAslrAlignment; const VAddr temp_region_start = - memory_layout.GetVirtualMemoryRegionTree().GetRandomAlignedRegion( + memory_layout->GetVirtualMemoryRegionTree().GetRandomAlignedRegion( TempRegionSize, TempRegionAlign, KMemoryRegionType_Kernel); - ASSERT(memory_layout.GetVirtualMemoryRegionTree().Insert(temp_region_start, TempRegionSize, - KMemoryRegionType_KernelTemp)); + ASSERT(memory_layout->GetVirtualMemoryRegionTree().Insert(temp_region_start, TempRegionSize, + KMemoryRegionType_KernelTemp)); // Automatically map in devices that have auto-map attributes. - for (auto& region : memory_layout.GetPhysicalMemoryRegionTree()) { + for (auto& region : memory_layout->GetPhysicalMemoryRegionTree()) { // We only care about kernel regions. if (!region.IsDerivedFrom(KMemoryRegionType_Kernel)) { continue; @@ -508,21 +501,21 @@ struct KernelCore::Impl { const size_t map_size = Common::AlignUp(region.GetEndAddress(), PageSize) - map_phys_addr; const VAddr map_virt_addr = - memory_layout.GetVirtualMemoryRegionTree().GetRandomAlignedRegionWithGuard( + memory_layout->GetVirtualMemoryRegionTree().GetRandomAlignedRegionWithGuard( map_size, PageSize, KMemoryRegionType_KernelMisc, PageSize); - ASSERT(memory_layout.GetVirtualMemoryRegionTree().Insert( + ASSERT(memory_layout->GetVirtualMemoryRegionTree().Insert( map_virt_addr, map_size, KMemoryRegionType_KernelMiscMappedDevice)); region.SetPairAddress(map_virt_addr + region.GetAddress() - map_phys_addr); } - Init::SetupDramPhysicalMemoryRegions(memory_layout); + Init::SetupDramPhysicalMemoryRegions(*memory_layout); // Insert a physical region for the kernel code region. - ASSERT(memory_layout.GetPhysicalMemoryRegionTree().Insert( + ASSERT(memory_layout->GetPhysicalMemoryRegionTree().Insert( code_start_phys_addr, code_region_size, KMemoryRegionType_DramKernelCode)); // Insert a physical region for the kernel slab region. - ASSERT(memory_layout.GetPhysicalMemoryRegionTree().Insert( + ASSERT(memory_layout->GetPhysicalMemoryRegionTree().Insert( slab_start_phys_addr, slab_region_size, KMemoryRegionType_DramKernelSlab)); // Determine size available for kernel page table heaps, requiring > 8 MB. @@ -531,12 +524,12 @@ struct KernelCore::Impl { ASSERT(page_table_heap_size / 4_MiB > 2); // Insert a physical region for the kernel page table heap region - ASSERT(memory_layout.GetPhysicalMemoryRegionTree().Insert( + ASSERT(memory_layout->GetPhysicalMemoryRegionTree().Insert( slab_end_phys_addr, page_table_heap_size, KMemoryRegionType_DramKernelPtHeap)); // All DRAM regions that we haven't tagged by this point will be mapped under the linear // mapping. Tag them. - for (auto& region : memory_layout.GetPhysicalMemoryRegionTree()) { + for (auto& region : memory_layout->GetPhysicalMemoryRegionTree()) { if (region.GetType() == KMemoryRegionType_Dram) { // Check that the region is valid. ASSERT(region.GetEndAddress() != 0); @@ -548,7 +541,7 @@ struct KernelCore::Impl { // Get the linear region extents. const auto linear_extents = - memory_layout.GetPhysicalMemoryRegionTree().GetDerivedRegionExtents( + memory_layout->GetPhysicalMemoryRegionTree().GetDerivedRegionExtents( KMemoryRegionAttr_LinearMapped); ASSERT(linear_extents.GetEndAddress() != 0); @@ -560,7 +553,7 @@ struct KernelCore::Impl { Common::AlignUp(linear_extents.GetEndAddress(), LinearRegionAlign) - aligned_linear_phys_start; const VAddr linear_region_start = - memory_layout.GetVirtualMemoryRegionTree().GetRandomAlignedRegionWithGuard( + memory_layout->GetVirtualMemoryRegionTree().GetRandomAlignedRegionWithGuard( linear_region_size, LinearRegionAlign, KMemoryRegionType_None, LinearRegionAlign); const u64 linear_region_phys_to_virt_diff = linear_region_start - aligned_linear_phys_start; @@ -569,7 +562,7 @@ struct KernelCore::Impl { { PAddr cur_phys_addr = 0; u64 cur_size = 0; - for (auto& region : memory_layout.GetPhysicalMemoryRegionTree()) { + for (auto& region : memory_layout->GetPhysicalMemoryRegionTree()) { if (!region.HasTypeAttribute(KMemoryRegionAttr_LinearMapped)) { continue; } @@ -588,55 +581,49 @@ struct KernelCore::Impl { const VAddr region_virt_addr = region.GetAddress() + linear_region_phys_to_virt_diff; - ASSERT(memory_layout.GetVirtualMemoryRegionTree().Insert( + ASSERT(memory_layout->GetVirtualMemoryRegionTree().Insert( region_virt_addr, region.GetSize(), GetTypeForVirtualLinearMapping(region.GetType()))); region.SetPairAddress(region_virt_addr); KMemoryRegion* virt_region = - memory_layout.GetVirtualMemoryRegionTree().FindModifiable(region_virt_addr); + memory_layout->GetVirtualMemoryRegionTree().FindModifiable(region_virt_addr); ASSERT(virt_region != nullptr); virt_region->SetPairAddress(region.GetAddress()); } } // Insert regions for the initial page table region. - ASSERT(memory_layout.GetPhysicalMemoryRegionTree().Insert( + ASSERT(memory_layout->GetPhysicalMemoryRegionTree().Insert( resource_end_phys_addr, KernelPageTableHeapSize, KMemoryRegionType_DramKernelInitPt)); - ASSERT(memory_layout.GetVirtualMemoryRegionTree().Insert( + ASSERT(memory_layout->GetVirtualMemoryRegionTree().Insert( resource_end_phys_addr + linear_region_phys_to_virt_diff, KernelPageTableHeapSize, KMemoryRegionType_VirtualDramKernelInitPt)); // All linear-mapped DRAM regions that we haven't tagged by this point will be allocated to // some pool partition. Tag them. - for (auto& region : memory_layout.GetPhysicalMemoryRegionTree()) { + for (auto& region : memory_layout->GetPhysicalMemoryRegionTree()) { if (region.GetType() == (KMemoryRegionType_Dram | KMemoryRegionAttr_LinearMapped)) { region.SetType(KMemoryRegionType_DramPoolPartition); } } // Setup all other memory regions needed to arrange the pool partitions. - Init::SetupPoolPartitionMemoryRegions(memory_layout); + Init::SetupPoolPartitionMemoryRegions(*memory_layout); // Cache all linear regions in their own trees for faster access, later. - memory_layout.InitializeLinearMemoryRegionTrees(aligned_linear_phys_start, - linear_region_start); + memory_layout->InitializeLinearMemoryRegionTrees(aligned_linear_phys_start, + linear_region_start); } - void InitializeMemoryLayout(const KMemoryLayout& memory_layout) { - const auto system_pool = memory_layout.GetKernelSystemPoolRegionPhysicalExtents(); - const auto applet_pool = memory_layout.GetKernelAppletPoolRegionPhysicalExtents(); - const auto application_pool = memory_layout.GetKernelApplicationPoolRegionPhysicalExtents(); + void InitializeMemoryLayout() { + const auto system_pool = memory_layout->GetKernelSystemPoolRegionPhysicalExtents(); - // Initialize memory managers + // Initialize the memory manager. memory_manager = std::make_unique<KMemoryManager>(system); - memory_manager->InitializeManager(KMemoryManager::Pool::Application, - application_pool.GetAddress(), - application_pool.GetEndAddress()); - memory_manager->InitializeManager(KMemoryManager::Pool::Applet, applet_pool.GetAddress(), - applet_pool.GetEndAddress()); - memory_manager->InitializeManager(KMemoryManager::Pool::System, system_pool.GetAddress(), - system_pool.GetEndAddress()); + const auto& management_region = memory_layout->GetPoolManagementRegion(); + ASSERT(management_region.GetEndAddress() != 0); + memory_manager->Initialize(management_region.GetAddress(), management_region.GetSize()); // Setup memory regions for emulated processes // TODO(bunnei): These should not be hardcoded regions initialized within the kernel @@ -777,6 +764,9 @@ struct KernelCore::Impl { Kernel::KSharedMemory* irs_shared_mem{}; Kernel::KSharedMemory* time_shared_mem{}; + // Memory layout + std::unique_ptr<KMemoryLayout> memory_layout; + // Threads used for services std::unordered_set<std::shared_ptr<Kernel::ServiceThread>> service_threads; Common::ThreadWorker service_threads_manager; @@ -1142,6 +1132,10 @@ const KWorkerTaskManager& KernelCore::WorkerTaskManager() const { return impl->worker_task_manager; } +const KMemoryLayout& KernelCore::MemoryLayout() const { + return *impl->memory_layout; +} + bool KernelCore::IsPhantomModeForSingleCore() const { return impl->IsPhantomModeForSingleCore(); } diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h index 0e04fc3bb..c1254b18d 100644 --- a/src/core/hle/kernel/kernel.h +++ b/src/core/hle/kernel/kernel.h @@ -41,6 +41,7 @@ class KClientSession; class KEvent; class KHandleTable; class KLinkedListNode; +class KMemoryLayout; class KMemoryManager; class KPort; class KProcess; @@ -350,6 +351,9 @@ public: /// Gets the current worker task manager, used for dispatching KThread/KProcess tasks. const KWorkerTaskManager& WorkerTaskManager() const; + /// Gets the memory layout. + const KMemoryLayout& MemoryLayout() const; + private: friend class KProcess; friend class KThread; diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp index 9836809f2..839171e85 100644 --- a/src/core/hle/kernel/svc.cpp +++ b/src/core/hle/kernel/svc.cpp @@ -2332,7 +2332,7 @@ static ResultCode CreateEvent(Core::System& system, Handle* out_write, Handle* o R_UNLESS(event != nullptr, ResultOutOfResource); // Initialize the event. - event->Initialize("CreateEvent"); + event->Initialize("CreateEvent", kernel.CurrentProcess()); // Commit the thread reservation. event_reservation.Commit(); diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp index 773dc9f29..2f8e21568 100644 --- a/src/core/hle/service/am/am.cpp +++ b/src/core/hle/service/am/am.cpp @@ -618,7 +618,7 @@ void AppletMessageQueue::PushMessage(AppletMessage msg) { AppletMessageQueue::AppletMessage AppletMessageQueue::PopMessage() { if (messages.empty()) { on_new_message->GetWritableEvent().Clear(); - return AppletMessage::NoMessage; + return AppletMessage::None; } auto msg = messages.front(); messages.pop(); @@ -633,7 +633,7 @@ std::size_t AppletMessageQueue::GetMessageCount() const { } void AppletMessageQueue::RequestExit() { - PushMessage(AppletMessage::ExitRequested); + PushMessage(AppletMessage::Exit); } void AppletMessageQueue::FocusStateChanged() { @@ -732,7 +732,7 @@ void ICommonStateGetter::ReceiveMessage(Kernel::HLERequestContext& ctx) { const auto message = msg_queue->PopMessage(); IPC::ResponseBuilder rb{ctx, 3}; - if (message == AppletMessageQueue::AppletMessage::NoMessage) { + if (message == AppletMessageQueue::AppletMessage::None) { LOG_ERROR(Service_AM, "Message queue is empty"); rb.Push(ERR_NO_MESSAGES); rb.PushEnum<AppletMessageQueue::AppletMessage>(message); diff --git a/src/core/hle/service/am/am.h b/src/core/hle/service/am/am.h index 2a578aea5..fdd937b82 100644 --- a/src/core/hle/service/am/am.h +++ b/src/core/hle/service/am/am.h @@ -22,6 +22,7 @@ class NVFlinger; namespace Service::AM { +// This is nn::settings::Language enum SystemLanguage { Japanese = 0, English = 1, // en-US @@ -41,16 +42,44 @@ enum SystemLanguage { // 4.0.0+ SimplifiedChinese = 15, TraditionalChinese = 16, + // 10.1.0+ + BrazilianPortuguese = 17, }; class AppletMessageQueue { public: + // This is nn::am::AppletMessage enum class AppletMessage : u32 { - NoMessage = 0, - ExitRequested = 4, + None = 0, + ChangeIntoForeground = 1, + ChangeIntoBackground = 2, + Exit = 4, + ApplicationExited = 6, FocusStateChanged = 15, + Resume = 16, + DetectShortPressingHomeButton = 20, + DetectLongPressingHomeButton = 21, + DetectShortPressingPowerButton = 22, + DetectMiddlePressingPowerButton = 23, + DetectLongPressingPowerButton = 24, + RequestToPrepareSleep = 25, + FinishedSleepSequence = 26, + SleepRequiredByHighTemperature = 27, + SleepRequiredByLowBattery = 28, + AutoPowerDown = 29, OperationModeChanged = 30, PerformanceModeChanged = 31, + DetectReceivingCecSystemStandby = 32, + SdCardRemoved = 33, + LaunchApplicationRequested = 50, + RequestToDisplay = 51, + ShowApplicationLogo = 55, + HideApplicationLogo = 56, + ForceHideApplicationLogo = 57, + FloatingApplicationDetected = 60, + DetectShortPressingCaptureButton = 90, + AlbumScreenShotTaken = 92, + AlbumRecordingSaved = 93, }; explicit AppletMessageQueue(Core::System& system); @@ -179,11 +208,14 @@ public: ~ICommonStateGetter() override; private: + // This is nn::oe::FocusState enum class FocusState : u8 { InFocus = 1, NotInFocus = 2, + Background = 3, }; + // This is nn::oe::OperationMode enum class OperationMode : u8 { Handheld = 0, Docked = 1, diff --git a/src/core/hle/service/apm/apm_controller.cpp b/src/core/hle/service/apm/apm_controller.cpp index 98839fe97..187fef2ad 100644 --- a/src/core/hle/service/apm/apm_controller.cpp +++ b/src/core/hle/service/apm/apm_controller.cpp @@ -17,8 +17,8 @@ constexpr auto DEFAULT_PERFORMANCE_CONFIGURATION = PerformanceConfiguration::Con Controller::Controller(Core::Timing::CoreTiming& core_timing_) : core_timing{core_timing_}, configs{ - {PerformanceMode::Handheld, DEFAULT_PERFORMANCE_CONFIGURATION}, - {PerformanceMode::Docked, DEFAULT_PERFORMANCE_CONFIGURATION}, + {PerformanceMode::Normal, DEFAULT_PERFORMANCE_CONFIGURATION}, + {PerformanceMode::Boost, DEFAULT_PERFORMANCE_CONFIGURATION}, } {} Controller::~Controller() = default; @@ -63,13 +63,13 @@ void Controller::SetFromCpuBoostMode(CpuBoostMode mode) { PerformanceConfiguration::Config15, }}; - SetPerformanceConfiguration(PerformanceMode::Docked, + SetPerformanceConfiguration(PerformanceMode::Boost, BOOST_MODE_TO_CONFIG_MAP.at(static_cast<u32>(mode))); } PerformanceMode Controller::GetCurrentPerformanceMode() const { - return Settings::values.use_docked_mode.GetValue() ? PerformanceMode::Docked - : PerformanceMode::Handheld; + return Settings::values.use_docked_mode.GetValue() ? PerformanceMode::Boost + : PerformanceMode::Normal; } PerformanceConfiguration Controller::GetCurrentPerformanceConfiguration(PerformanceMode mode) { diff --git a/src/core/hle/service/apm/apm_controller.h b/src/core/hle/service/apm/apm_controller.h index 8d48e0104..d6fbd2c0c 100644 --- a/src/core/hle/service/apm/apm_controller.h +++ b/src/core/hle/service/apm/apm_controller.h @@ -32,15 +32,18 @@ enum class PerformanceConfiguration : u32 { Config16 = 0x9222000C, }; +// This is nn::oe::CpuBoostMode enum class CpuBoostMode : u32 { - Disabled = 0, - Full = 1, // CPU + GPU -> Config 13, 14, 15, or 16 - Partial = 2, // GPU Only -> Config 15 or 16 + Normal = 0, // Boost mode disabled + FastLoad = 1, // CPU + GPU -> Config 13, 14, 15, or 16 + Partial = 2, // GPU Only -> Config 15 or 16 }; -enum class PerformanceMode : u8 { - Handheld = 0, - Docked = 1, +// This is nn::oe::PerformanceMode +enum class PerformanceMode : s32 { + Invalid = -1, + Normal = 0, + Boost = 1, }; // Class to manage the state and change of the emulated system performance. diff --git a/src/core/hle/service/kernel_helpers.cpp b/src/core/hle/service/kernel_helpers.cpp index 62f4cdfb2..b8c2c6e51 100644 --- a/src/core/hle/service/kernel_helpers.cpp +++ b/src/core/hle/service/kernel_helpers.cpp @@ -3,7 +3,9 @@ // Refer to the license.txt file included. #include "core/core.h" +#include "core/core_timing.h" #include "core/hle/kernel/k_event.h" +#include "core/hle/kernel/k_memory_manager.h" #include "core/hle/kernel/k_process.h" #include "core/hle/kernel/k_readable_event.h" #include "core/hle/kernel/k_resource_limit.h" @@ -15,10 +17,21 @@ namespace Service::KernelHelpers { ServiceContext::ServiceContext(Core::System& system_, std::string name_) : kernel(system_.Kernel()) { + + // Create a resource limit for the process. + const auto physical_memory_size = + kernel.MemoryManager().GetSize(Kernel::KMemoryManager::Pool::System); + auto* resource_limit = Kernel::CreateResourceLimitForProcess(system_, physical_memory_size); + + // Create the process. process = Kernel::KProcess::Create(kernel); ASSERT(Kernel::KProcess::Initialize(process, system_, std::move(name_), - Kernel::KProcess::ProcessType::Userland) + Kernel::KProcess::ProcessType::KernelInternal, + resource_limit) .IsSuccess()); + + // Close reference to our resource limit, as the process opens one. + resource_limit->Close(); } ServiceContext::~ServiceContext() { @@ -43,7 +56,7 @@ Kernel::KEvent* ServiceContext::CreateEvent(std::string&& name) { } // Initialize the event. - event->Initialize(std::move(name)); + event->Initialize(std::move(name), process); // Commit the thread reservation. event_reservation.Commit(); diff --git a/src/core/memory.cpp b/src/core/memory.cpp index 88d6ec908..28d30eee2 100644 --- a/src/core/memory.cpp +++ b/src/core/memory.cpp @@ -39,8 +39,7 @@ struct Memory::Impl { void MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size, PAddr target) { ASSERT_MSG((size & PAGE_MASK) == 0, "non-page aligned size: {:016X}", size); ASSERT_MSG((base & PAGE_MASK) == 0, "non-page aligned base: {:016X}", base); - ASSERT_MSG(target >= DramMemoryMap::Base && target < DramMemoryMap::End, - "Out of bounds target: {:016X}", target); + ASSERT_MSG(target >= DramMemoryMap::Base, "Out of bounds target: {:016X}", target); MapPages(page_table, base / PAGE_SIZE, size / PAGE_SIZE, target, Common::PageType::Memory); if (Settings::IsFastmemEnabled()) { diff --git a/src/input_common/drivers/sdl_driver.cpp b/src/input_common/drivers/sdl_driver.cpp index 5cf1987ad..c17ea305e 100644 --- a/src/input_common/drivers/sdl_driver.cpp +++ b/src/input_common/drivers/sdl_driver.cpp @@ -175,22 +175,23 @@ public: return false; } - BatteryLevel GetBatteryLevel() { + Common::Input::BatteryLevel GetBatteryLevel() { const auto level = SDL_JoystickCurrentPowerLevel(sdl_joystick.get()); switch (level) { case SDL_JOYSTICK_POWER_EMPTY: - return BatteryLevel::Empty; + return Common::Input::BatteryLevel::Empty; case SDL_JOYSTICK_POWER_LOW: - return BatteryLevel::Low; + return Common::Input::BatteryLevel::Low; case SDL_JOYSTICK_POWER_MEDIUM: - return BatteryLevel::Medium; + return Common::Input::BatteryLevel::Medium; case SDL_JOYSTICK_POWER_FULL: case SDL_JOYSTICK_POWER_MAX: - return BatteryLevel::Full; - case SDL_JOYSTICK_POWER_UNKNOWN: + return Common::Input::BatteryLevel::Full; case SDL_JOYSTICK_POWER_WIRED: + return Common::Input::BatteryLevel::Charging; + case SDL_JOYSTICK_POWER_UNKNOWN: default: - return BatteryLevel::Charging; + return Common::Input::BatteryLevel::None; } } @@ -351,6 +352,8 @@ void SDLDriver::HandleGameControllerEvent(const SDL_Event& event) { if (const auto joystick = GetSDLJoystickBySDLID(event.jbutton.which)) { const PadIdentifier identifier = joystick->GetPadIdentifier(); SetButton(identifier, event.jbutton.button, true); + // Battery doesn't trigger an event so just update every button press + SetBattery(identifier, joystick->GetBatteryLevel()); } break; } diff --git a/src/input_common/drivers/udp_client.cpp b/src/input_common/drivers/udp_client.cpp index 64162f431..9780ead10 100644 --- a/src/input_common/drivers/udp_client.cpp +++ b/src/input_common/drivers/udp_client.cpp @@ -192,22 +192,22 @@ std::size_t UDPClient::GetClientNumber(std::string_view host, u16 port) const { return MAX_UDP_CLIENTS; } -BatteryLevel UDPClient::GetBatteryLevel(Response::Battery battery) const { +Common::Input::BatteryLevel UDPClient::GetBatteryLevel(Response::Battery battery) const { switch (battery) { case Response::Battery::Dying: - return BatteryLevel::Empty; + return Common::Input::BatteryLevel::Empty; case Response::Battery::Low: - return BatteryLevel::Critical; + return Common::Input::BatteryLevel::Critical; case Response::Battery::Medium: - return BatteryLevel::Low; + return Common::Input::BatteryLevel::Low; case Response::Battery::High: - return BatteryLevel::Medium; + return Common::Input::BatteryLevel::Medium; case Response::Battery::Full: case Response::Battery::Charged: - return BatteryLevel::Full; + return Common::Input::BatteryLevel::Full; case Response::Battery::Charging: default: - return BatteryLevel::Charging; + return Common::Input::BatteryLevel::Charging; } } diff --git a/src/input_common/drivers/udp_client.h b/src/input_common/drivers/udp_client.h index 76e32bd04..c7cc7d846 100644 --- a/src/input_common/drivers/udp_client.h +++ b/src/input_common/drivers/udp_client.h @@ -141,7 +141,7 @@ private: std::size_t GetClientNumber(std::string_view host, u16 port) const; // Translates UDP battery level to input engine battery level - BatteryLevel GetBatteryLevel(Response::Battery battery) const; + Common::Input::BatteryLevel GetBatteryLevel(Response::Battery battery) const; void OnVersion(Response::Version); void OnPortInfo(Response::PortInfo); diff --git a/src/input_common/input_engine.cpp b/src/input_common/input_engine.cpp index 65ae1b848..7adf7e3d7 100644 --- a/src/input_common/input_engine.cpp +++ b/src/input_common/input_engine.cpp @@ -70,7 +70,7 @@ void InputEngine::SetAxis(const PadIdentifier& identifier, int axis, f32 value) TriggerOnAxisChange(identifier, axis, value); } -void InputEngine::SetBattery(const PadIdentifier& identifier, BatteryLevel value) { +void InputEngine::SetBattery(const PadIdentifier& identifier, Common::Input::BatteryLevel value) { { std::lock_guard lock{mutex}; ControllerData& controller = controller_list.at(identifier); @@ -143,13 +143,13 @@ f32 InputEngine::GetAxis(const PadIdentifier& identifier, int axis) const { return axis_iter->second; } -BatteryLevel InputEngine::GetBattery(const PadIdentifier& identifier) const { +Common::Input::BatteryLevel InputEngine::GetBattery(const PadIdentifier& identifier) const { std::lock_guard lock{mutex}; const auto controller_iter = controller_list.find(identifier); if (controller_iter == controller_list.cend()) { LOG_ERROR(Input, "Invalid identifier guid={}, pad={}, port={}", identifier.guid.RawString(), identifier.pad, identifier.port); - return BatteryLevel::Charging; + return Common::Input::BatteryLevel::Charging; } const ControllerData& controller = controller_iter->second; return controller.battery; @@ -270,7 +270,7 @@ void InputEngine::TriggerOnAxisChange(const PadIdentifier& identifier, int axis, } void InputEngine::TriggerOnBatteryChange(const PadIdentifier& identifier, - [[maybe_unused]] BatteryLevel value) { + [[maybe_unused]] Common::Input::BatteryLevel value) { std::lock_guard lock{mutex_callback}; for (const auto& poller_pair : callback_list) { const InputIdentifier& poller = poller_pair.second; diff --git a/src/input_common/input_engine.h b/src/input_common/input_engine.h index c6c027aef..f44e0799b 100644 --- a/src/input_common/input_engine.h +++ b/src/input_common/input_engine.h @@ -34,16 +34,6 @@ struct BasicMotion { u64 delta_timestamp{}; }; -// Stages of a battery charge -enum class BatteryLevel { - Empty, - Critical, - Low, - Medium, - Full, - Charging, -}; - // Types of input that are stored in the engine enum class EngineInputType { None, @@ -178,7 +168,7 @@ public: bool GetButton(const PadIdentifier& identifier, int button) const; bool GetHatButton(const PadIdentifier& identifier, int button, u8 direction) const; f32 GetAxis(const PadIdentifier& identifier, int axis) const; - BatteryLevel GetBattery(const PadIdentifier& identifier) const; + Common::Input::BatteryLevel GetBattery(const PadIdentifier& identifier) const; BasicMotion GetMotion(const PadIdentifier& identifier, int motion) const; int SetCallback(InputIdentifier input_identifier); @@ -189,7 +179,7 @@ protected: void SetButton(const PadIdentifier& identifier, int button, bool value); void SetHatButton(const PadIdentifier& identifier, int button, u8 value); void SetAxis(const PadIdentifier& identifier, int axis, f32 value); - void SetBattery(const PadIdentifier& identifier, BatteryLevel value); + void SetBattery(const PadIdentifier& identifier, Common::Input::BatteryLevel value); void SetMotion(const PadIdentifier& identifier, int motion, const BasicMotion& value); virtual std::string GetHatButtonName([[maybe_unused]] u8 direction_value) const { @@ -202,13 +192,13 @@ private: std::unordered_map<int, u8> hat_buttons; std::unordered_map<int, float> axes; std::unordered_map<int, BasicMotion> motions; - BatteryLevel battery{}; + Common::Input::BatteryLevel battery{}; }; void TriggerOnButtonChange(const PadIdentifier& identifier, int button, bool value); void TriggerOnHatButtonChange(const PadIdentifier& identifier, int button, u8 value); void TriggerOnAxisChange(const PadIdentifier& identifier, int axis, f32 value); - void TriggerOnBatteryChange(const PadIdentifier& identifier, BatteryLevel value); + void TriggerOnBatteryChange(const PadIdentifier& identifier, Common::Input::BatteryLevel value); void TriggerOnMotionChange(const PadIdentifier& identifier, int motion, const BasicMotion& value); diff --git a/src/input_common/input_poller.cpp b/src/input_common/input_poller.cpp index 7f3c08597..82b585ff2 100644 --- a/src/input_common/input_poller.cpp +++ b/src/input_common/input_poller.cpp @@ -470,7 +470,7 @@ public: } Common::Input::BatteryStatus GetStatus() const { - return static_cast<Common::Input::BatteryLevel>(input_engine->GetBattery(identifier)); + return input_engine->GetBattery(identifier); } void ForceUpdate() override { diff --git a/src/video_core/renderer_opengl/gl_fence_manager.cpp b/src/video_core/renderer_opengl/gl_fence_manager.cpp index 151290101..293ad7d59 100644 --- a/src/video_core/renderer_opengl/gl_fence_manager.cpp +++ b/src/video_core/renderer_opengl/gl_fence_manager.cpp @@ -31,9 +31,8 @@ bool GLInnerFence::IsSignaled() const { return true; } ASSERT(sync_object.handle != 0); - GLsizei length; GLint sync_status; - glGetSynciv(sync_object.handle, GL_SYNC_STATUS, sizeof(GLint), &length, &sync_status); + glGetSynciv(sync_object.handle, GL_SYNC_STATUS, 1, nullptr, &sync_status); return sync_status == GL_SIGNALED; } diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp index f8495896c..9e6732abd 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp @@ -243,10 +243,6 @@ GraphicsPipeline::GraphicsPipeline( case Settings::ShaderBackend::GLASM: if (!sources[stage].empty()) { assembly_programs[stage] = CompileProgram(sources[stage], AssemblyStage(stage)); - if (in_parallel) { - // Make sure program is built before continuing when building in parallel - glGetString(GL_PROGRAM_ERROR_STRING_NV); - } } break; case Settings::ShaderBackend::SPIRV: @@ -256,20 +252,18 @@ GraphicsPipeline::GraphicsPipeline( break; } } - if (in_parallel && backend != Settings::ShaderBackend::GLASM) { - // Make sure programs have built if we are building shaders in parallel - for (OGLProgram& program : source_programs) { - if (program.handle != 0) { - GLint status{}; - glGetProgramiv(program.handle, GL_LINK_STATUS, &status); - } - } + if (in_parallel) { + std::lock_guard lock{built_mutex}; + built_fence.Create(); + // Flush this context to ensure compilation commands and fence are in the GPU pipe. + glFlush(); + built_condvar.notify_one(); + } else { + is_built = true; } if (shader_notify) { shader_notify->MarkShaderComplete(); } - is_built = true; - built_condvar.notify_one(); }}; if (thread_worker) { thread_worker->QueueWork(std::move(func)); @@ -440,7 +434,7 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) { buffer_cache.UpdateGraphicsBuffers(is_indexed); buffer_cache.BindHostGeometryBuffers(is_indexed); - if (!is_built.load(std::memory_order::relaxed)) { + if (!IsBuilt()) { WaitForBuild(); } const bool use_assembly{assembly_programs[0].handle != 0}; @@ -585,8 +579,26 @@ void GraphicsPipeline::GenerateTransformFeedbackState() { } void GraphicsPipeline::WaitForBuild() { - std::unique_lock lock{built_mutex}; - built_condvar.wait(lock, [this] { return is_built.load(std::memory_order::relaxed); }); + if (built_fence.handle == 0) { + std::unique_lock lock{built_mutex}; + built_condvar.wait(lock, [this] { return built_fence.handle != 0; }); + } + ASSERT(glClientWaitSync(built_fence.handle, 0, GL_TIMEOUT_IGNORED) != GL_WAIT_FAILED); + is_built = true; +} + +bool GraphicsPipeline::IsBuilt() noexcept { + if (is_built) { + return true; + } + if (built_fence.handle == 0) { + return false; + } + // Timeout of zero means this is non-blocking + const auto sync_status = glClientWaitSync(built_fence.handle, 0, 0); + ASSERT(sync_status != GL_WAIT_FAILED); + is_built = sync_status != GL_TIMEOUT_EXPIRED; + return is_built; } } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.h b/src/video_core/renderer_opengl/gl_graphics_pipeline.h index 4e28d9a42..311d49f3f 100644 --- a/src/video_core/renderer_opengl/gl_graphics_pipeline.h +++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.h @@ -100,9 +100,7 @@ public: return writes_global_memory; } - [[nodiscard]] bool IsBuilt() const noexcept { - return is_built.load(std::memory_order::relaxed); - } + [[nodiscard]] bool IsBuilt() noexcept; template <typename Spec> static auto MakeConfigureSpecFunc() { @@ -154,7 +152,8 @@ private: std::mutex built_mutex; std::condition_variable built_condvar; - std::atomic_bool is_built{false}; + OGLSync built_fence{}; + bool is_built{false}; }; } // namespace OpenGL diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h index daba42ed9..db5bf1d30 100644 --- a/src/video_core/renderer_opengl/maxwell_to_gl.h +++ b/src/video_core/renderer_opengl/maxwell_to_gl.h @@ -184,6 +184,8 @@ inline GLenum VertexFormat(Maxwell::VertexAttribute attrib) { case Maxwell::VertexAttribute::Size::Size_32_32_32: case Maxwell::VertexAttribute::Size::Size_32_32_32_32: return GL_FLOAT; + case Maxwell::VertexAttribute::Size::Size_11_11_10: + return GL_UNSIGNED_INT_10F_11F_11F_REV; default: break; } diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp index 751e4792b..1c136c410 100644 --- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp +++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp @@ -495,6 +495,8 @@ VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttrib return VK_FORMAT_R32G32B32_SFLOAT; case Maxwell::VertexAttribute::Size::Size_32_32_32_32: return VK_FORMAT_R32G32B32A32_SFLOAT; + case Maxwell::VertexAttribute::Size::Size_11_11_10: + return VK_FORMAT_B10G11R11_UFLOAT_PACK32; default: break; } diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp index c71a1f44d..621a6a071 100644 --- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp +++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp @@ -100,6 +100,8 @@ VkFormat GetFormat(const Tegra::FramebufferConfig& framebuffer) { return VK_FORMAT_A8B8G8R8_UNORM_PACK32; case Tegra::FramebufferConfig::PixelFormat::RGB565_UNORM: return VK_FORMAT_R5G6B5_UNORM_PACK16; + case Tegra::FramebufferConfig::PixelFormat::B8G8R8A8_UNORM: + return VK_FORMAT_B8G8R8A8_UNORM; default: UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}", static_cast<u32>(framebuffer.pixel_format)); diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp index 0ba56ff1e..0f62779de 100644 --- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp @@ -554,10 +554,12 @@ void CopyBufferToImage(vk::CommandBuffer cmdbuf, VkBuffer src_buffer, VkImage im }; } -[[nodiscard]] bool IsFormatFlipped(PixelFormat format) { +[[nodiscard]] bool IsFormatFlipped(PixelFormat format, bool emulate_bgr565) { switch (format) { case PixelFormat::A1B5G5R5_UNORM: return true; + case PixelFormat::B5G6R5_UNORM: + return emulate_bgr565; default: return false; } @@ -1488,7 +1490,7 @@ ImageView::ImageView(TextureCacheRuntime& runtime, const VideoCommon::ImageViewI }; if (!info.IsRenderTarget()) { swizzle = info.Swizzle(); - if (IsFormatFlipped(format)) { + if (IsFormatFlipped(format, device->MustEmulateBGR565())) { std::ranges::transform(swizzle, swizzle.begin(), SwapBlueRed); } if ((aspect_mask & (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) != 0) { diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp index 153702c0b..effde73c9 100644 --- a/src/video_core/vulkan_common/vulkan_device.cpp +++ b/src/video_core/vulkan_common/vulkan_device.cpp @@ -39,6 +39,11 @@ constexpr std::array DEPTH16_UNORM_STENCIL8_UINT{ VK_FORMAT_D32_SFLOAT_S8_UINT, VK_FORMAT_UNDEFINED, }; + +constexpr std::array B5G6R5_UNORM_PACK16{ + VK_FORMAT_R5G6B5_UNORM_PACK16, + VK_FORMAT_UNDEFINED, +}; } // namespace Alternatives enum class NvidiaArchitecture { @@ -87,6 +92,8 @@ constexpr const VkFormat* GetFormatAlternatives(VkFormat format) { return Alternatives::DEPTH24_UNORM_STENCIL8_UINT.data(); case VK_FORMAT_D16_UNORM_S8_UINT: return Alternatives::DEPTH16_UNORM_STENCIL8_UINT.data(); + case VK_FORMAT_B5G6R5_UNORM_PACK16: + return Alternatives::B5G6R5_UNORM_PACK16.data(); default: return nullptr; } @@ -224,9 +231,14 @@ std::vector<std::string> GetSupportedExtensions(vk::PhysicalDevice physical) { return supported_extensions; } +bool IsExtensionSupported(std::span<const std::string> supported_extensions, + std::string_view extension) { + return std::ranges::find(supported_extensions, extension) != supported_extensions.end(); +} + NvidiaArchitecture GetNvidiaArchitecture(vk::PhysicalDevice physical, std::span<const std::string> exts) { - if (std::ranges::find(exts, VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME) != exts.end()) { + if (IsExtensionSupported(exts, VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME)) { VkPhysicalDeviceFragmentShadingRatePropertiesKHR shading_rate_props{}; shading_rate_props.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FRAGMENT_SHADING_RATE_PROPERTIES_KHR; @@ -239,7 +251,7 @@ NvidiaArchitecture GetNvidiaArchitecture(vk::PhysicalDevice physical, return NvidiaArchitecture::AmpereOrNewer; } } - if (std::ranges::find(exts, VK_NV_SHADING_RATE_IMAGE_EXTENSION_NAME) != exts.end()) { + if (IsExtensionSupported(exts, VK_NV_SHADING_RATE_IMAGE_EXTENSION_NAME)) { return NvidiaArchitecture::Turing; } return NvidiaArchitecture::VoltaOrOlder; @@ -604,7 +616,8 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR break; } } - if (ext_extended_dynamic_state && driver_id == VK_DRIVER_ID_MESA_RADV) { + const bool is_radv = driver_id == VK_DRIVER_ID_MESA_RADV; + if (ext_extended_dynamic_state && is_radv) { // Mask driver version variant const u32 version = (properties.driverVersion << 3) >> 3; if (version < VK_MAKE_API_VERSION(0, 21, 2, 0)) { @@ -613,6 +626,17 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR ext_extended_dynamic_state = false; } } + if (ext_vertex_input_dynamic_state && is_radv) { + // TODO(ameerj): Blacklist only offending driver versions + // TODO(ameerj): Confirm if RDNA1 is affected + const bool is_rdna2 = + IsExtensionSupported(supported_extensions, VK_KHR_FRAGMENT_SHADING_RATE_EXTENSION_NAME); + if (is_rdna2) { + LOG_WARNING(Render_Vulkan, + "RADV has broken VK_EXT_vertex_input_dynamic_state on RDNA2 hardware"); + ext_vertex_input_dynamic_state = false; + } + } sets_per_pool = 64; const bool is_amd = @@ -628,7 +652,7 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR has_broken_cube_compatibility = true; } } - const bool is_amd_or_radv = is_amd || driver_id == VK_DRIVER_ID_MESA_RADV; + const bool is_amd_or_radv = is_amd || is_radv; if (ext_sampler_filter_minmax && is_amd_or_radv) { // Disable ext_sampler_filter_minmax on AMD GCN4 and lower as it is broken. if (!is_float16_supported) { @@ -639,6 +663,7 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR } const bool is_intel_windows = driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS; + const bool is_intel_anv = driver_id == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA; if (ext_vertex_input_dynamic_state && is_intel_windows) { LOG_WARNING(Render_Vulkan, "Blacklisting Intel for VK_EXT_vertex_input_dynamic_state"); ext_vertex_input_dynamic_state = false; @@ -652,6 +677,10 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR LOG_WARNING(Render_Vulkan, "Intel proprietary drivers do not support MSAA image blits"); cant_blit_msaa = true; } + if (is_intel_anv) { + LOG_WARNING(Render_Vulkan, "ANV driver does not support native BGR format"); + must_emulate_bgr565 = true; + } supports_d24_depth = IsFormatSupported(VK_FORMAT_D24_UNORM_S8_UINT, diff --git a/src/video_core/vulkan_common/vulkan_device.h b/src/video_core/vulkan_common/vulkan_device.h index 37d140ebd..34b1add16 100644 --- a/src/video_core/vulkan_common/vulkan_device.h +++ b/src/video_core/vulkan_common/vulkan_device.h @@ -354,6 +354,10 @@ public: return cant_blit_msaa; } + bool MustEmulateBGR565() const { + return must_emulate_bgr565; + } + private: /// Checks if the physical device is suitable. void CheckSuitability(bool requires_swapchain) const; @@ -448,6 +452,7 @@ private: bool has_nsight_graphics{}; ///< Has Nsight Graphics attached bool supports_d24_depth{}; ///< Supports D24 depth buffers. bool cant_blit_msaa{}; ///< Does not support MSAA<->MSAA blitting. + bool must_emulate_bgr565{}; ///< Emulates BGR565 by swizzling RGB565 format. // Telemetry parameters std::string vendor_name; ///< Device's driver name. diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp index d5ba86c03..c2b66ff14 100644 --- a/src/yuzu/configuration/config.cpp +++ b/src/yuzu/configuration/config.cpp @@ -445,6 +445,7 @@ void Config::ReadCoreValues() { qt_config->beginGroup(QStringLiteral("Core")); ReadGlobalSetting(Settings::values.use_multi_core); + ReadGlobalSetting(Settings::values.use_extended_memory_layout); qt_config->endGroup(); } @@ -608,6 +609,7 @@ void Config::ReadCpuValues() { ReadGlobalSetting(Settings::values.cpuopt_unsafe_ignore_standard_fpcr); ReadGlobalSetting(Settings::values.cpuopt_unsafe_inaccurate_nan); ReadGlobalSetting(Settings::values.cpuopt_unsafe_fastmem_check); + ReadGlobalSetting(Settings::values.cpuopt_unsafe_ignore_global_monitor); if (global) { ReadBasicSetting(Settings::values.cpu_debug_mode); @@ -620,6 +622,8 @@ void Config::ReadCpuValues() { ReadBasicSetting(Settings::values.cpuopt_misc_ir); ReadBasicSetting(Settings::values.cpuopt_reduce_misalign_checks); ReadBasicSetting(Settings::values.cpuopt_fastmem); + ReadBasicSetting(Settings::values.cpuopt_fastmem_exclusives); + ReadBasicSetting(Settings::values.cpuopt_recompile_exclusives); } qt_config->endGroup(); @@ -1019,6 +1023,7 @@ void Config::SaveCoreValues() { qt_config->beginGroup(QStringLiteral("Core")); WriteGlobalSetting(Settings::values.use_multi_core); + WriteGlobalSetting(Settings::values.use_extended_memory_layout); qt_config->endGroup(); } @@ -1137,6 +1142,7 @@ void Config::SaveCpuValues() { WriteGlobalSetting(Settings::values.cpuopt_unsafe_ignore_standard_fpcr); WriteGlobalSetting(Settings::values.cpuopt_unsafe_inaccurate_nan); WriteGlobalSetting(Settings::values.cpuopt_unsafe_fastmem_check); + WriteGlobalSetting(Settings::values.cpuopt_unsafe_ignore_global_monitor); if (global) { WriteBasicSetting(Settings::values.cpu_debug_mode); diff --git a/src/yuzu/configuration/configure_cpu.cpp b/src/yuzu/configuration/configure_cpu.cpp index f66cab5d4..bf74ccc7c 100644 --- a/src/yuzu/configuration/configure_cpu.cpp +++ b/src/yuzu/configuration/configure_cpu.cpp @@ -36,6 +36,7 @@ void ConfigureCpu::SetConfiguration() { ui->cpuopt_unsafe_ignore_standard_fpcr->setEnabled(runtime_lock); ui->cpuopt_unsafe_inaccurate_nan->setEnabled(runtime_lock); ui->cpuopt_unsafe_fastmem_check->setEnabled(runtime_lock); + ui->cpuopt_unsafe_ignore_global_monitor->setEnabled(runtime_lock); ui->cpuopt_unsafe_unfuse_fma->setChecked(Settings::values.cpuopt_unsafe_unfuse_fma.GetValue()); ui->cpuopt_unsafe_reduce_fp_error->setChecked( @@ -46,6 +47,8 @@ void ConfigureCpu::SetConfiguration() { Settings::values.cpuopt_unsafe_inaccurate_nan.GetValue()); ui->cpuopt_unsafe_fastmem_check->setChecked( Settings::values.cpuopt_unsafe_fastmem_check.GetValue()); + ui->cpuopt_unsafe_ignore_global_monitor->setChecked( + Settings::values.cpuopt_unsafe_ignore_global_monitor.GetValue()); if (Settings::IsConfiguringGlobal()) { ui->accuracy->setCurrentIndex(static_cast<int>(Settings::values.cpu_accuracy.GetValue())); @@ -82,6 +85,9 @@ void ConfigureCpu::ApplyConfiguration() { ConfigurationShared::ApplyPerGameSetting(&Settings::values.cpuopt_unsafe_fastmem_check, ui->cpuopt_unsafe_fastmem_check, cpuopt_unsafe_fastmem_check); + ConfigurationShared::ApplyPerGameSetting(&Settings::values.cpuopt_unsafe_ignore_global_monitor, + ui->cpuopt_unsafe_ignore_global_monitor, + cpuopt_unsafe_ignore_global_monitor); } void ConfigureCpu::changeEvent(QEvent* event) { @@ -120,4 +126,7 @@ void ConfigureCpu::SetupPerGameUI() { ConfigurationShared::SetColoredTristate(ui->cpuopt_unsafe_fastmem_check, Settings::values.cpuopt_unsafe_fastmem_check, cpuopt_unsafe_fastmem_check); + ConfigurationShared::SetColoredTristate(ui->cpuopt_unsafe_ignore_global_monitor, + Settings::values.cpuopt_unsafe_ignore_global_monitor, + cpuopt_unsafe_ignore_global_monitor); } diff --git a/src/yuzu/configuration/configure_cpu.h b/src/yuzu/configuration/configure_cpu.h index ed9af0e9f..733e38be4 100644 --- a/src/yuzu/configuration/configure_cpu.h +++ b/src/yuzu/configuration/configure_cpu.h @@ -45,6 +45,7 @@ private: ConfigurationShared::CheckState cpuopt_unsafe_ignore_standard_fpcr; ConfigurationShared::CheckState cpuopt_unsafe_inaccurate_nan; ConfigurationShared::CheckState cpuopt_unsafe_fastmem_check; + ConfigurationShared::CheckState cpuopt_unsafe_ignore_global_monitor; const Core::System& system; }; diff --git a/src/yuzu/configuration/configure_cpu.ui b/src/yuzu/configuration/configure_cpu.ui index d8064db24..5d80a8c91 100644 --- a/src/yuzu/configuration/configure_cpu.ui +++ b/src/yuzu/configuration/configure_cpu.ui @@ -150,6 +150,18 @@ </property> </widget> </item> + <item> + <widget class="QCheckBox" name="cpuopt_unsafe_ignore_global_monitor"> + <property name="toolTip"> + <string> + <div>This option improves speed by relying only on the semantics of cmpxchg to ensure safety of exclusive access instructions. Please note this may result in deadlocks and other race conditions.</div> + </string> + </property> + <property name="text"> + <string>Ignore global monitor</string> + </property> + </widget> + </item> </layout> </widget> </item> diff --git a/src/yuzu/configuration/configure_cpu_debug.cpp b/src/yuzu/configuration/configure_cpu_debug.cpp index 05a90963d..616a0be75 100644 --- a/src/yuzu/configuration/configure_cpu_debug.cpp +++ b/src/yuzu/configuration/configure_cpu_debug.cpp @@ -44,6 +44,12 @@ void ConfigureCpuDebug::SetConfiguration() { Settings::values.cpuopt_reduce_misalign_checks.GetValue()); ui->cpuopt_fastmem->setEnabled(runtime_lock); ui->cpuopt_fastmem->setChecked(Settings::values.cpuopt_fastmem.GetValue()); + ui->cpuopt_fastmem_exclusives->setEnabled(runtime_lock); + ui->cpuopt_fastmem_exclusives->setChecked( + Settings::values.cpuopt_fastmem_exclusives.GetValue()); + ui->cpuopt_recompile_exclusives->setEnabled(runtime_lock); + ui->cpuopt_recompile_exclusives->setChecked( + Settings::values.cpuopt_recompile_exclusives.GetValue()); } void ConfigureCpuDebug::ApplyConfiguration() { @@ -56,6 +62,8 @@ void ConfigureCpuDebug::ApplyConfiguration() { Settings::values.cpuopt_misc_ir = ui->cpuopt_misc_ir->isChecked(); Settings::values.cpuopt_reduce_misalign_checks = ui->cpuopt_reduce_misalign_checks->isChecked(); Settings::values.cpuopt_fastmem = ui->cpuopt_fastmem->isChecked(); + Settings::values.cpuopt_fastmem_exclusives = ui->cpuopt_fastmem_exclusives->isChecked(); + Settings::values.cpuopt_recompile_exclusives = ui->cpuopt_recompile_exclusives->isChecked(); } void ConfigureCpuDebug::changeEvent(QEvent* event) { diff --git a/src/yuzu/configuration/configure_cpu_debug.ui b/src/yuzu/configuration/configure_cpu_debug.ui index 6e635bb2f..2bc268810 100644 --- a/src/yuzu/configuration/configure_cpu_debug.ui +++ b/src/yuzu/configuration/configure_cpu_debug.ui @@ -144,7 +144,34 @@ </string> </property> <property name="text"> - <string>Enable Host MMU Emulation</string> + <string>Enable Host MMU Emulation (general memory instructions)</string> + </property> + </widget> + </item> + <item> + <widget class="QCheckBox" name="cpuopt_fastmem_exclusives"> + <property name="toolTip"> + <string> + <div style="white-space: nowrap">This optimization speeds up exclusive memory accesses by the guest program.</div> + <div style="white-space: nowrap">Enabling it causes guest exclusive memory reads/writes to be done directly into memory and make use of Host's MMU.</div> + <div style="white-space: nowrap">Disabling this forces all exclusive memory accesses to use Software MMU Emulation.</div> + </string> + </property> + <property name="text"> + <string>Enable Host MMU Emulation (exclusive memory instructions)</string> + </property> + </widget> + </item> + <item> + <widget class="QCheckBox" name="cpuopt_recompile_exclusives"> + <property name="toolTip"> + <string> + <div style="white-space: nowrap">This optimization speeds up exclusive memory accesses by the guest program.</div> + <div style="white-space: nowrap">Enabling it reduces the overhead of fastmem failure of exclusive memory accesses.</div> + </string> + </property> + <property name="text"> + <string>Enable recompilation of exclusive memory instructions</string> </property> </widget> </item> diff --git a/src/yuzu/configuration/configure_general.cpp b/src/yuzu/configuration/configure_general.cpp index 978a29fe6..08d5444ec 100644 --- a/src/yuzu/configuration/configure_general.cpp +++ b/src/yuzu/configuration/configure_general.cpp @@ -42,6 +42,9 @@ void ConfigureGeneral::SetConfiguration() { ui->use_multi_core->setEnabled(runtime_lock); ui->use_multi_core->setChecked(Settings::values.use_multi_core.GetValue()); + ui->use_extended_memory_layout->setEnabled(runtime_lock); + ui->use_extended_memory_layout->setChecked( + Settings::values.use_extended_memory_layout.GetValue()); ui->toggle_check_exit->setChecked(UISettings::values.confirm_before_closing.GetValue()); ui->toggle_user_on_boot->setChecked(UISettings::values.select_user_on_boot.GetValue()); @@ -91,6 +94,9 @@ void ConfigureGeneral::ResetDefaults() { void ConfigureGeneral::ApplyConfiguration() { ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_multi_core, ui->use_multi_core, use_multi_core); + ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_extended_memory_layout, + ui->use_extended_memory_layout, + use_extended_memory_layout); if (Settings::IsConfiguringGlobal()) { UISettings::values.confirm_before_closing = ui->toggle_check_exit->isChecked(); @@ -160,6 +166,9 @@ void ConfigureGeneral::SetupPerGameUI() { Settings::values.use_speed_limit, use_speed_limit); ConfigurationShared::SetColoredTristate(ui->use_multi_core, Settings::values.use_multi_core, use_multi_core); + ConfigurationShared::SetColoredTristate(ui->use_extended_memory_layout, + Settings::values.use_extended_memory_layout, + use_extended_memory_layout); connect(ui->toggle_speed_limit, &QCheckBox::clicked, ui->speed_limit, [this]() { ui->speed_limit->setEnabled(ui->toggle_speed_limit->isChecked() && diff --git a/src/yuzu/configuration/configure_general.h b/src/yuzu/configuration/configure_general.h index 85c1dd4a8..b6f3bb5ed 100644 --- a/src/yuzu/configuration/configure_general.h +++ b/src/yuzu/configuration/configure_general.h @@ -48,6 +48,7 @@ private: ConfigurationShared::CheckState use_speed_limit; ConfigurationShared::CheckState use_multi_core; + ConfigurationShared::CheckState use_extended_memory_layout; const Core::System& system; }; diff --git a/src/yuzu/configuration/configure_general.ui b/src/yuzu/configuration/configure_general.ui index bfc771135..c6ef2ab70 100644 --- a/src/yuzu/configuration/configure_general.ui +++ b/src/yuzu/configuration/configure_general.ui @@ -143,6 +143,13 @@ </widget> </item> <item> + <widget class="QCheckBox" name="use_extended_memory_layout"> + <property name="text"> + <string>Extended memory layout (6GB DRAM)</string> + </property> + </widget> + </item> + <item> <widget class="QCheckBox" name="toggle_check_exit"> <property name="text"> <string>Confirm exit while emulation is running</string> diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp index e3fd38a02..b3a8da0ea 100644 --- a/src/yuzu/main.cpp +++ b/src/yuzu/main.cpp @@ -2937,6 +2937,7 @@ void GMainWindow::OnLoadAmiibo() { if (nfc_state == Service::NFP::DeviceState::TagFound || nfc_state == Service::NFP::DeviceState::TagMounted) { nfc->CloseAmiibo(); + QMessageBox::warning(this, tr("Amiibo"), tr("The current amiibo has been removed")); return; } @@ -2965,6 +2966,15 @@ void GMainWindow::LoadAmiibo(const QString& filename) { return; } + // Remove amiibo if one is connected + const auto nfc_state = nfc->GetCurrentState(); + if (nfc_state == Service::NFP::DeviceState::TagFound || + nfc_state == Service::NFP::DeviceState::TagMounted) { + nfc->CloseAmiibo(); + QMessageBox::warning(this, tr("Amiibo"), tr("The current amiibo has been removed")); + return; + } + QFile nfc_file{filename}; if (!nfc_file.open(QIODevice::ReadOnly)) { QMessageBox::warning(this, tr("Error opening Amiibo data file"), diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp index ff616da70..b74411c84 100644 --- a/src/yuzu_cmd/config.cpp +++ b/src/yuzu_cmd/config.cpp @@ -266,6 +266,7 @@ void Config::ReadValues() { // Core ReadSetting("Core", Settings::values.use_multi_core); + ReadSetting("Core", Settings::values.use_extended_memory_layout); // Cpu ReadSetting("Cpu", Settings::values.cpu_accuracy); @@ -279,11 +280,14 @@ void Config::ReadValues() { ReadSetting("Cpu", Settings::values.cpuopt_misc_ir); ReadSetting("Cpu", Settings::values.cpuopt_reduce_misalign_checks); ReadSetting("Cpu", Settings::values.cpuopt_fastmem); + ReadSetting("Cpu", Settings::values.cpuopt_fastmem_exclusives); + ReadSetting("Cpu", Settings::values.cpuopt_recompile_exclusives); ReadSetting("Cpu", Settings::values.cpuopt_unsafe_unfuse_fma); ReadSetting("Cpu", Settings::values.cpuopt_unsafe_reduce_fp_error); ReadSetting("Cpu", Settings::values.cpuopt_unsafe_ignore_standard_fpcr); ReadSetting("Cpu", Settings::values.cpuopt_unsafe_inaccurate_nan); ReadSetting("Cpu", Settings::values.cpuopt_unsafe_fastmem_check); + ReadSetting("Cpu", Settings::values.cpuopt_unsafe_ignore_global_monitor); // Renderer ReadSetting("Renderer", Settings::values.renderer_backend); diff --git a/src/yuzu_cmd/default_ini.h b/src/yuzu_cmd/default_ini.h index 6d613bf7a..3ac1440c9 100644 --- a/src/yuzu_cmd/default_ini.h +++ b/src/yuzu_cmd/default_ini.h @@ -174,6 +174,14 @@ cpuopt_reduce_misalign_checks = # 0: Disabled, 1 (default): Enabled cpuopt_fastmem = +# Enable Host MMU Emulation for exclusive memory instructions (faster guest memory access) +# 0: Disabled, 1 (default): Enabled +cpuopt_fastmem_exclusives = + +# Enable fallback on failure of fastmem of exclusive memory instructions (faster guest memory access) +# 0: Disabled, 1 (default): Enabled +cpuopt_recompile_exclusives = + # Enable unfuse FMA (improve performance on CPUs without FMA) # Only enabled if cpu_accuracy is set to Unsafe. Automatically chosen with cpu_accuracy = Auto-select. # 0: Disabled, 1 (default): Enabled @@ -199,6 +207,11 @@ cpuopt_unsafe_inaccurate_nan = # 0: Disabled, 1 (default): Enabled cpuopt_unsafe_fastmem_check = +# Enable faster exclusive instructions +# Only enabled if cpu_accuracy is set to Unsafe. Automatically chosen with cpu_accuracy = Auto-select. +# 0: Disabled, 1 (default): Enabled +cpuopt_unsafe_ignore_global_monitor = + [Renderer] # Which backend API to use. # 0 (default): OpenGL, 1: Vulkan |