diff options
Diffstat (limited to 'src')
107 files changed, 2005 insertions, 746 deletions
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt index f69d00a2b..6c99dd5e2 100644 --- a/src/CMakeLists.txt +++ b/src/CMakeLists.txt @@ -1,18 +1,79 @@ # Enable modules to include each other's files include_directories(.) +# CMake seems to only define _DEBUG on Windows +set_property(DIRECTORY APPEND PROPERTY + COMPILE_DEFINITIONS $<$<CONFIG:Debug>:_DEBUG> $<$<NOT:$<CONFIG:Debug>>:NDEBUG>) + +# Set compilation flags +if (MSVC) + set(CMAKE_CONFIGURATION_TYPES Debug Release CACHE STRING "" FORCE) + + # Silence "deprecation" warnings + add_definitions(-D_CRT_SECURE_NO_WARNINGS -D_CRT_NONSTDC_NO_DEPRECATE -D_SCL_SECURE_NO_WARNINGS) + + # Avoid windows.h junk + add_definitions(-DNOMINMAX) + + # Avoid windows.h from including some usually unused libs like winsocks.h, since this might cause some redefinition errors. + add_definitions(-DWIN32_LEAN_AND_MEAN) + + # /W3 - Level 3 warnings + # /MP - Multi-threaded compilation + # /Zi - Output debugging information + # /Zo - enhanced debug info for optimized builds + # /permissive- - enables stricter C++ standards conformance checks + # /EHsc - C++-only exception handling semantics + # /Zc:throwingNew - let codegen assume `operator new` will never return null + # /Zc:inline - let codegen omit inline functions in object files + add_compile_options(/W3 /MP /Zi /Zo /permissive- /EHsc /std:c++latest /Zc:throwingNew,inline) + + # /GS- - No stack buffer overflow checks + add_compile_options("$<$<CONFIG:Release>:/GS->") + + set(CMAKE_EXE_LINKER_FLAGS_DEBUG "/DEBUG /MANIFEST:NO" CACHE STRING "" FORCE) + set(CMAKE_EXE_LINKER_FLAGS_RELEASE "/DEBUG /MANIFEST:NO /INCREMENTAL:NO /OPT:REF,ICF" CACHE STRING "" FORCE) +else() + add_compile_options("-Wno-attributes") + + if (APPLE AND CMAKE_CXX_COMPILER_ID STREQUAL Clang) + add_compile_options("-stdlib=libc++") + endif() + + # Set file offset size to 64 bits. + # + # On modern Unixes, this is typically already the case. The lone exception is + # glibc, which may default to 32 bits. glibc allows this to be configured + # by setting _FILE_OFFSET_BITS. + if(CMAKE_SYSTEM_NAME STREQUAL "Linux" OR MINGW) + add_definitions(-D_FILE_OFFSET_BITS=64) + endif() + + if (MINGW) + add_definitions(-DMINGW_HAS_SECURE_API) + + if (MINGW_STATIC_BUILD) + add_definitions(-DQT_STATICPLUGIN) + add_compile_options("-static") + endif() + endif() +endif() + add_subdirectory(common) add_subdirectory(core) add_subdirectory(audio_core) add_subdirectory(video_core) add_subdirectory(input_common) add_subdirectory(tests) + if (ENABLE_SDL2) add_subdirectory(yuzu_cmd) endif() + if (ENABLE_QT) add_subdirectory(yuzu) endif() + if (ENABLE_WEB_SERVICE) add_subdirectory(web_service) endif() diff --git a/src/audio_core/stream.cpp b/src/audio_core/stream.cpp index 4b66a6786..22a3f8c84 100644 --- a/src/audio_core/stream.cpp +++ b/src/audio_core/stream.cpp @@ -38,7 +38,7 @@ Stream::Stream(Core::Timing::CoreTiming& core_timing, u32 sample_rate, Format fo sink_stream{sink_stream}, core_timing{core_timing}, name{std::move(name_)} { release_event = core_timing.RegisterEvent( - name, [this](u64 userdata, int cycles_late) { ReleaseActiveBuffer(); }); + name, [this](u64 userdata, s64 cycles_late) { ReleaseActiveBuffer(); }); } void Stream::Play() { diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt index 43ae8a9e7..850ce8006 100644 --- a/src/common/CMakeLists.txt +++ b/src/common/CMakeLists.txt @@ -98,6 +98,7 @@ add_library(common STATIC microprofile.h microprofileui.h misc.cpp + multi_level_queue.h page_table.cpp page_table.h param_package.cpp diff --git a/src/common/bit_util.h b/src/common/bit_util.h index 1eea17ba1..a4f9ed4aa 100644 --- a/src/common/bit_util.h +++ b/src/common/bit_util.h @@ -58,4 +58,43 @@ inline u64 CountLeadingZeroes64(u64 value) { return __builtin_clzll(value); } #endif + +#ifdef _MSC_VER +inline u32 CountTrailingZeroes32(u32 value) { + unsigned long trailing_zero = 0; + + if (_BitScanForward(&trailing_zero, value) != 0) { + return trailing_zero; + } + + return 32; +} + +inline u64 CountTrailingZeroes64(u64 value) { + unsigned long trailing_zero = 0; + + if (_BitScanForward64(&trailing_zero, value) != 0) { + return trailing_zero; + } + + return 64; +} +#else +inline u32 CountTrailingZeroes32(u32 value) { + if (value == 0) { + return 32; + } + + return __builtin_ctz(value); +} + +inline u64 CountTrailingZeroes64(u64 value) { + if (value == 0) { + return 64; + } + + return __builtin_ctzll(value); +} +#endif + } // namespace Common diff --git a/src/common/common_types.h b/src/common/common_types.h index 6b1766dca..4cec89fbd 100644 --- a/src/common/common_types.h +++ b/src/common/common_types.h @@ -40,10 +40,9 @@ using s64 = std::int64_t; ///< 64-bit signed int using f32 = float; ///< 32-bit floating point using f64 = double; ///< 64-bit floating point -// TODO: It would be nice to eventually replace these with strong types that prevent accidental -// conversion between each other. -using VAddr = u64; ///< Represents a pointer in the userspace virtual address space. -using PAddr = u64; ///< Represents a pointer in the ARM11 physical address space. +using VAddr = u64; ///< Represents a pointer in the userspace virtual address space. +using PAddr = u64; ///< Represents a pointer in the ARM11 physical address space. +using GPUVAddr = u64; ///< Represents a pointer in the GPU virtual address space. using u128 = std::array<std::uint64_t, 2>; static_assert(sizeof(u128) == 16, "u128 must be 128 bits wide"); diff --git a/src/common/detached_tasks.cpp b/src/common/detached_tasks.cpp index a347d9e02..f268d6021 100644 --- a/src/common/detached_tasks.cpp +++ b/src/common/detached_tasks.cpp @@ -16,22 +16,22 @@ DetachedTasks::DetachedTasks() { } void DetachedTasks::WaitForAllTasks() { - std::unique_lock<std::mutex> lock(mutex); + std::unique_lock lock{mutex}; cv.wait(lock, [this]() { return count == 0; }); } DetachedTasks::~DetachedTasks() { - std::unique_lock<std::mutex> lock(mutex); + std::unique_lock lock{mutex}; ASSERT(count == 0); instance = nullptr; } void DetachedTasks::AddTask(std::function<void()> task) { - std::unique_lock<std::mutex> lock(instance->mutex); + std::unique_lock lock{instance->mutex}; ++instance->count; std::thread([task{std::move(task)}]() { task(); - std::unique_lock<std::mutex> lock(instance->mutex); + std::unique_lock lock{instance->mutex}; --instance->count; std::notify_all_at_thread_exit(instance->cv, std::move(lock)); }) diff --git a/src/common/logging/backend.cpp b/src/common/logging/backend.cpp index 4462ff3fb..a03179520 100644 --- a/src/common/logging/backend.cpp +++ b/src/common/logging/backend.cpp @@ -46,12 +46,12 @@ public: } void AddBackend(std::unique_ptr<Backend> backend) { - std::lock_guard<std::mutex> lock(writing_mutex); + std::lock_guard lock{writing_mutex}; backends.push_back(std::move(backend)); } void RemoveBackend(std::string_view backend_name) { - std::lock_guard<std::mutex> lock(writing_mutex); + std::lock_guard lock{writing_mutex}; const auto it = std::remove_if(backends.begin(), backends.end(), [&backend_name](const auto& i) { return backend_name == i->GetName(); }); @@ -80,7 +80,7 @@ private: backend_thread = std::thread([&] { Entry entry; auto write_logs = [&](Entry& e) { - std::lock_guard<std::mutex> lock(writing_mutex); + std::lock_guard lock{writing_mutex}; for (const auto& backend : backends) { backend->Write(e); } diff --git a/src/common/multi_level_queue.h b/src/common/multi_level_queue.h new file mode 100644 index 000000000..2b61b91e0 --- /dev/null +++ b/src/common/multi_level_queue.h @@ -0,0 +1,337 @@ +// Copyright 2019 TuxSH +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include <array> +#include <iterator> +#include <list> +#include <utility> + +#include "common/bit_util.h" +#include "common/common_types.h" + +namespace Common { + +/** + * A MultiLevelQueue is a type of priority queue which has the following characteristics: + * - iteratable through each of its elements. + * - back can be obtained. + * - O(1) add, lookup (both front and back) + * - discrete priorities and a max of 64 priorities (limited domain) + * This type of priority queue is normaly used for managing threads within an scheduler + */ +template <typename T, std::size_t Depth> +class MultiLevelQueue { +public: + using value_type = T; + using reference = value_type&; + using const_reference = const value_type&; + using pointer = value_type*; + using const_pointer = const value_type*; + + using difference_type = typename std::pointer_traits<pointer>::difference_type; + using size_type = std::size_t; + + template <bool is_constant> + class iterator_impl { + public: + using iterator_category = std::bidirectional_iterator_tag; + using value_type = T; + using pointer = std::conditional_t<is_constant, T*, const T*>; + using reference = std::conditional_t<is_constant, const T&, T&>; + using difference_type = typename std::pointer_traits<pointer>::difference_type; + + friend bool operator==(const iterator_impl& lhs, const iterator_impl& rhs) { + if (lhs.IsEnd() && rhs.IsEnd()) + return true; + return std::tie(lhs.current_priority, lhs.it) == std::tie(rhs.current_priority, rhs.it); + } + + friend bool operator!=(const iterator_impl& lhs, const iterator_impl& rhs) { + return !operator==(lhs, rhs); + } + + reference operator*() const { + return *it; + } + + pointer operator->() const { + return it.operator->(); + } + + iterator_impl& operator++() { + if (IsEnd()) { + return *this; + } + + ++it; + + if (it == GetEndItForPrio()) { + u64 prios = mlq.used_priorities; + prios &= ~((1ULL << (current_priority + 1)) - 1); + if (prios == 0) { + current_priority = mlq.depth(); + } else { + current_priority = CountTrailingZeroes64(prios); + it = GetBeginItForPrio(); + } + } + return *this; + } + + iterator_impl& operator--() { + if (IsEnd()) { + if (mlq.used_priorities != 0) { + current_priority = 63 - CountLeadingZeroes64(mlq.used_priorities); + it = GetEndItForPrio(); + --it; + } + } else if (it == GetBeginItForPrio()) { + u64 prios = mlq.used_priorities; + prios &= (1ULL << current_priority) - 1; + if (prios != 0) { + current_priority = CountTrailingZeroes64(prios); + it = GetEndItForPrio(); + --it; + } + } else { + --it; + } + return *this; + } + + iterator_impl operator++(int) { + const iterator_impl v{*this}; + ++(*this); + return v; + } + + iterator_impl operator--(int) { + const iterator_impl v{*this}; + --(*this); + return v; + } + + // allow implicit const->non-const + iterator_impl(const iterator_impl<false>& other) + : mlq(other.mlq), it(other.it), current_priority(other.current_priority) {} + + iterator_impl(const iterator_impl<true>& other) + : mlq(other.mlq), it(other.it), current_priority(other.current_priority) {} + + iterator_impl& operator=(const iterator_impl<false>& other) { + mlq = other.mlq; + it = other.it; + current_priority = other.current_priority; + return *this; + } + + friend class iterator_impl<true>; + iterator_impl() = default; + + private: + friend class MultiLevelQueue; + using container_ref = + std::conditional_t<is_constant, const MultiLevelQueue&, MultiLevelQueue&>; + using list_iterator = std::conditional_t<is_constant, typename std::list<T>::const_iterator, + typename std::list<T>::iterator>; + + explicit iterator_impl(container_ref mlq, list_iterator it, u32 current_priority) + : mlq(mlq), it(it), current_priority(current_priority) {} + explicit iterator_impl(container_ref mlq, u32 current_priority) + : mlq(mlq), it(), current_priority(current_priority) {} + + bool IsEnd() const { + return current_priority == mlq.depth(); + } + + list_iterator GetBeginItForPrio() const { + return mlq.levels[current_priority].begin(); + } + + list_iterator GetEndItForPrio() const { + return mlq.levels[current_priority].end(); + } + + container_ref mlq; + list_iterator it; + u32 current_priority; + }; + + using iterator = iterator_impl<false>; + using const_iterator = iterator_impl<true>; + + void add(const T& element, u32 priority, bool send_back = true) { + if (send_back) + levels[priority].push_back(element); + else + levels[priority].push_front(element); + used_priorities |= 1ULL << priority; + } + + void remove(const T& element, u32 priority) { + auto it = ListIterateTo(levels[priority], element); + if (it == levels[priority].end()) + return; + levels[priority].erase(it); + if (levels[priority].empty()) { + used_priorities &= ~(1ULL << priority); + } + } + + void adjust(const T& element, u32 old_priority, u32 new_priority, bool adjust_front = false) { + remove(element, old_priority); + add(element, new_priority, !adjust_front); + } + void adjust(const_iterator it, u32 old_priority, u32 new_priority, bool adjust_front = false) { + adjust(*it, old_priority, new_priority, adjust_front); + } + + void transfer_to_front(const T& element, u32 priority, MultiLevelQueue& other) { + ListSplice(other.levels[priority], other.levels[priority].begin(), levels[priority], + ListIterateTo(levels[priority], element)); + + other.used_priorities |= 1ULL << priority; + + if (levels[priority].empty()) { + used_priorities &= ~(1ULL << priority); + } + } + + void transfer_to_front(const_iterator it, u32 priority, MultiLevelQueue& other) { + transfer_to_front(*it, priority, other); + } + + void transfer_to_back(const T& element, u32 priority, MultiLevelQueue& other) { + ListSplice(other.levels[priority], other.levels[priority].end(), levels[priority], + ListIterateTo(levels[priority], element)); + + other.used_priorities |= 1ULL << priority; + + if (levels[priority].empty()) { + used_priorities &= ~(1ULL << priority); + } + } + + void transfer_to_back(const_iterator it, u32 priority, MultiLevelQueue& other) { + transfer_to_back(*it, priority, other); + } + + void yield(u32 priority, std::size_t n = 1) { + ListShiftForward(levels[priority], n); + } + + std::size_t depth() const { + return Depth; + } + + std::size_t size(u32 priority) const { + return levels[priority].size(); + } + + std::size_t size() const { + u64 priorities = used_priorities; + std::size_t size = 0; + while (priorities != 0) { + const u64 current_priority = CountTrailingZeroes64(priorities); + size += levels[current_priority].size(); + priorities &= ~(1ULL << current_priority); + } + return size; + } + + bool empty() const { + return used_priorities == 0; + } + + bool empty(u32 priority) const { + return (used_priorities & (1ULL << priority)) == 0; + } + + u32 highest_priority_set(u32 max_priority = 0) const { + const u64 priorities = + max_priority == 0 ? used_priorities : (used_priorities & ~((1ULL << max_priority) - 1)); + return priorities == 0 ? Depth : static_cast<u32>(CountTrailingZeroes64(priorities)); + } + + u32 lowest_priority_set(u32 min_priority = Depth - 1) const { + const u64 priorities = min_priority >= Depth - 1 + ? used_priorities + : (used_priorities & ((1ULL << (min_priority + 1)) - 1)); + return priorities == 0 ? Depth : 63 - CountLeadingZeroes64(priorities); + } + + const_iterator cbegin(u32 max_prio = 0) const { + const u32 priority = highest_priority_set(max_prio); + return priority == Depth ? cend() + : const_iterator{*this, levels[priority].cbegin(), priority}; + } + const_iterator begin(u32 max_prio = 0) const { + return cbegin(max_prio); + } + iterator begin(u32 max_prio = 0) { + const u32 priority = highest_priority_set(max_prio); + return priority == Depth ? end() : iterator{*this, levels[priority].begin(), priority}; + } + + const_iterator cend(u32 min_prio = Depth - 1) const { + return min_prio == Depth - 1 ? const_iterator{*this, Depth} : cbegin(min_prio + 1); + } + const_iterator end(u32 min_prio = Depth - 1) const { + return cend(min_prio); + } + iterator end(u32 min_prio = Depth - 1) { + return min_prio == Depth - 1 ? iterator{*this, Depth} : begin(min_prio + 1); + } + + T& front(u32 max_priority = 0) { + const u32 priority = highest_priority_set(max_priority); + return levels[priority == Depth ? 0 : priority].front(); + } + const T& front(u32 max_priority = 0) const { + const u32 priority = highest_priority_set(max_priority); + return levels[priority == Depth ? 0 : priority].front(); + } + + T back(u32 min_priority = Depth - 1) { + const u32 priority = lowest_priority_set(min_priority); // intended + return levels[priority == Depth ? 63 : priority].back(); + } + const T& back(u32 min_priority = Depth - 1) const { + const u32 priority = lowest_priority_set(min_priority); // intended + return levels[priority == Depth ? 63 : priority].back(); + } + +private: + using const_list_iterator = typename std::list<T>::const_iterator; + + static void ListShiftForward(std::list<T>& list, const std::size_t shift = 1) { + if (shift >= list.size()) { + return; + } + + const auto begin_range = list.begin(); + const auto end_range = std::next(begin_range, shift); + list.splice(list.end(), list, begin_range, end_range); + } + + static void ListSplice(std::list<T>& in_list, const_list_iterator position, + std::list<T>& out_list, const_list_iterator element) { + in_list.splice(position, out_list, element); + } + + static const_list_iterator ListIterateTo(const std::list<T>& list, const T& element) { + auto it = list.cbegin(); + while (it != list.cend() && *it != element) { + ++it; + } + return it; + } + + std::array<std::list<T>, Depth> levels; + u64 used_priorities = 0; +}; + +} // namespace Common diff --git a/src/common/page_table.cpp b/src/common/page_table.cpp index 8eba1c3f1..69b7abc54 100644 --- a/src/common/page_table.cpp +++ b/src/common/page_table.cpp @@ -16,6 +16,7 @@ void PageTable::Resize(std::size_t address_space_width_in_bits) { pointers.resize(num_page_table_entries); attributes.resize(num_page_table_entries); + backing_addr.resize(num_page_table_entries); // The default is a 39-bit address space, which causes an initial 1GB allocation size. If the // vector size is subsequently decreased (via resize), the vector might not automatically @@ -24,6 +25,7 @@ void PageTable::Resize(std::size_t address_space_width_in_bits) { pointers.shrink_to_fit(); attributes.shrink_to_fit(); + backing_addr.shrink_to_fit(); } } // namespace Common diff --git a/src/common/page_table.h b/src/common/page_table.h index 8339f2890..8b8ff0bb8 100644 --- a/src/common/page_table.h +++ b/src/common/page_table.h @@ -21,6 +21,8 @@ enum class PageType : u8 { RasterizerCachedMemory, /// Page is mapped to a I/O region. Writing and reading to this page is handled by functions. Special, + /// Page is allocated for use. + Allocated, }; struct SpecialRegion { @@ -66,7 +68,7 @@ struct PageTable { * Contains MMIO handlers that back memory regions whose entries in the `attribute` vector is * of type `Special`. */ - boost::icl::interval_map<VAddr, std::set<SpecialRegion>> special_regions; + boost::icl::interval_map<u64, std::set<SpecialRegion>> special_regions; /** * Vector of fine grained page attributes. If it is set to any value other than `Memory`, then @@ -74,6 +76,8 @@ struct PageTable { */ std::vector<PageType> attributes; + std::vector<u64> backing_addr; + const std::size_t page_size_in_bits{}; }; diff --git a/src/common/thread.cpp b/src/common/thread.cpp index 5144c0d9f..fe7a420cc 100644 --- a/src/common/thread.cpp +++ b/src/common/thread.cpp @@ -27,18 +27,6 @@ namespace Common { #ifdef _MSC_VER -void SetThreadAffinity(std::thread::native_handle_type thread, u32 mask) { - SetThreadAffinityMask(thread, mask); -} - -void SetCurrentThreadAffinity(u32 mask) { - SetThreadAffinityMask(GetCurrentThread(), mask); -} - -void SwitchCurrentThread() { - SwitchToThread(); -} - // Sets the debugger-visible name of the current thread. // Uses undocumented (actually, it is now documented) trick. // http://msdn.microsoft.com/library/default.asp?url=/library/en-us/vsdebug/html/vxtsksettingthreadname.asp @@ -70,31 +58,6 @@ void SetCurrentThreadName(const char* name) { #else // !MSVC_VER, so must be POSIX threads -void SetThreadAffinity(std::thread::native_handle_type thread, u32 mask) { -#ifdef __APPLE__ - thread_policy_set(pthread_mach_thread_np(thread), THREAD_AFFINITY_POLICY, (integer_t*)&mask, 1); -#elif (defined __linux__ || defined __FreeBSD__) && !(defined ANDROID) - cpu_set_t cpu_set; - CPU_ZERO(&cpu_set); - - for (int i = 0; i != sizeof(mask) * 8; ++i) - if ((mask >> i) & 1) - CPU_SET(i, &cpu_set); - - pthread_setaffinity_np(thread, sizeof(cpu_set), &cpu_set); -#endif -} - -void SetCurrentThreadAffinity(u32 mask) { - SetThreadAffinity(pthread_self(), mask); -} - -#ifndef _WIN32 -void SwitchCurrentThread() { - usleep(1000 * 1); -} -#endif - // MinGW with the POSIX threading model does not support pthread_setname_np #if !defined(_WIN32) || defined(_MSC_VER) void SetCurrentThreadName(const char* name) { diff --git a/src/common/thread.h b/src/common/thread.h index 2cf74452d..0cfd98be6 100644 --- a/src/common/thread.h +++ b/src/common/thread.h @@ -9,14 +9,13 @@ #include <cstddef> #include <mutex> #include <thread> -#include "common/common_types.h" namespace Common { class Event { public: void Set() { - std::lock_guard<std::mutex> lk(mutex); + std::lock_guard lk{mutex}; if (!is_set) { is_set = true; condvar.notify_one(); @@ -24,14 +23,14 @@ public: } void Wait() { - std::unique_lock<std::mutex> lk(mutex); + std::unique_lock lk{mutex}; condvar.wait(lk, [&] { return is_set; }); is_set = false; } template <class Clock, class Duration> bool WaitUntil(const std::chrono::time_point<Clock, Duration>& time) { - std::unique_lock<std::mutex> lk(mutex); + std::unique_lock lk{mutex}; if (!condvar.wait_until(lk, time, [this] { return is_set; })) return false; is_set = false; @@ -39,7 +38,7 @@ public: } void Reset() { - std::unique_lock<std::mutex> lk(mutex); + std::unique_lock lk{mutex}; // no other action required, since wait loops on the predicate and any lingering signal will // get cleared on the first iteration is_set = false; @@ -57,7 +56,7 @@ public: /// Blocks until all "count" threads have called Sync() void Sync() { - std::unique_lock<std::mutex> lk(mutex); + std::unique_lock lk{mutex}; const std::size_t current_generation = generation; if (++waiting == count) { @@ -78,9 +77,6 @@ private: std::size_t generation = 0; // Incremented once each time the barrier is used }; -void SetThreadAffinity(std::thread::native_handle_type thread, u32 mask); -void SetCurrentThreadAffinity(u32 mask); -void SwitchCurrentThread(); // On Linux, this is equal to sleep 1ms void SetCurrentThreadName(const char* name); } // namespace Common diff --git a/src/common/threadsafe_queue.h b/src/common/threadsafe_queue.h index 821e8536a..e714ba5b3 100644 --- a/src/common/threadsafe_queue.h +++ b/src/common/threadsafe_queue.h @@ -78,7 +78,7 @@ public: T PopWait() { if (Empty()) { - std::unique_lock<std::mutex> lock(cv_mutex); + std::unique_lock lock{cv_mutex}; cv.wait(lock, [this]() { return !Empty(); }); } T t; @@ -137,7 +137,7 @@ public: template <typename Arg> void Push(Arg&& t) { - std::lock_guard<std::mutex> lock(write_lock); + std::lock_guard lock{write_lock}; spsc_queue.Push(t); } diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt index bbbe60896..9e23afe85 100644 --- a/src/core/CMakeLists.txt +++ b/src/core/CMakeLists.txt @@ -70,6 +70,8 @@ add_library(core STATIC file_sys/system_archive/ng_word.h file_sys/system_archive/system_archive.cpp file_sys/system_archive/system_archive.h + file_sys/system_archive/system_version.cpp + file_sys/system_archive/system_version.h file_sys/vfs.cpp file_sys/vfs.h file_sys/vfs_concat.cpp @@ -144,6 +146,8 @@ add_library(core STATIC hle/kernel/svc_wrap.h hle/kernel/thread.cpp hle/kernel/thread.h + hle/kernel/transfer_memory.cpp + hle/kernel/transfer_memory.h hle/kernel/vm_manager.cpp hle/kernel/vm_manager.h hle/kernel/wait_object.cpp diff --git a/src/core/core.cpp b/src/core/core.cpp index a88e332be..4fe77c25b 100644 --- a/src/core/core.cpp +++ b/src/core/core.cpp @@ -460,8 +460,8 @@ Tegra::DebugContext* System::GetGPUDebugContext() const { void System::RegisterCheatList(const std::vector<FileSys::CheatList>& list, const std::string& build_id, VAddr code_region_start, VAddr code_region_end) { - impl->cheat_engine = - std::make_unique<FileSys::CheatEngine>(list, build_id, code_region_start, code_region_end); + impl->cheat_engine = std::make_unique<FileSys::CheatEngine>(*this, list, build_id, + code_region_start, code_region_end); } void System::SetFilesystem(std::shared_ptr<FileSys::VfsFilesystem> vfs) { diff --git a/src/core/core_cpu.cpp b/src/core/core_cpu.cpp index 1eefed6d0..e75741db0 100644 --- a/src/core/core_cpu.cpp +++ b/src/core/core_cpu.cpp @@ -22,7 +22,7 @@ namespace Core { void CpuBarrier::NotifyEnd() { - std::unique_lock<std::mutex> lock(mutex); + std::unique_lock lock{mutex}; end = true; condition.notify_all(); } @@ -34,7 +34,7 @@ bool CpuBarrier::Rendezvous() { } if (!end) { - std::unique_lock<std::mutex> lock(mutex); + std::unique_lock lock{mutex}; --cores_waiting; if (!cores_waiting) { @@ -131,7 +131,7 @@ void Cpu::Reschedule() { reschedule_pending = false; // Lock the global kernel mutex when we manipulate the HLE state - std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock); + std::lock_guard lock{HLE::g_hle_lock}; scheduler->Reschedule(); } diff --git a/src/core/core_timing.cpp b/src/core/core_timing.cpp index a0dd5db24..41adb2302 100644 --- a/src/core/core_timing.cpp +++ b/src/core/core_timing.cpp @@ -186,7 +186,7 @@ void CoreTiming::Advance() { Event evt = std::move(event_queue.front()); std::pop_heap(event_queue.begin(), event_queue.end(), std::greater<>()); event_queue.pop_back(); - evt.type->callback(evt.userdata, static_cast<int>(global_timer - evt.time)); + evt.type->callback(evt.userdata, global_timer - evt.time); } is_global_timer_sane = false; diff --git a/src/core/core_timing.h b/src/core/core_timing.h index 59163bae1..9d2efde37 100644 --- a/src/core/core_timing.h +++ b/src/core/core_timing.h @@ -15,7 +15,7 @@ namespace Core::Timing { /// A callback that may be scheduled for a particular core timing event. -using TimedCallback = std::function<void(u64 userdata, int cycles_late)>; +using TimedCallback = std::function<void(u64 userdata, s64 cycles_late)>; /// Contains the characteristics of a particular event. struct EventType { diff --git a/src/core/file_sys/cheat_engine.cpp b/src/core/file_sys/cheat_engine.cpp index 09ca9d705..b06c2f20a 100644 --- a/src/core/file_sys/cheat_engine.cpp +++ b/src/core/file_sys/cheat_engine.cpp @@ -11,14 +11,13 @@ #include "core/core_timing_util.h" #include "core/file_sys/cheat_engine.h" #include "core/hle/kernel/process.h" -#include "core/hle/service/hid/controllers/controller_base.h" #include "core/hle/service/hid/controllers/npad.h" #include "core/hle/service/hid/hid.h" #include "core/hle/service/sm/sm.h" namespace FileSys { -constexpr u64 CHEAT_ENGINE_TICKS = Core::Timing::BASE_CLOCK_RATE / 60; +constexpr s64 CHEAT_ENGINE_TICKS = static_cast<s64>(Core::Timing::BASE_CLOCK_RATE / 60); constexpr u32 KEYPAD_BITMASK = 0x3FFFFFF; u64 Cheat::Address() const { @@ -77,8 +76,8 @@ void CheatList::Execute() { } } -CheatList::CheatList(ProgramSegment master, ProgramSegment standard) - : master_list(master), standard_list(standard) {} +CheatList::CheatList(const Core::System& system_, ProgramSegment master, ProgramSegment standard) + : master_list{std::move(master)}, standard_list{std::move(standard)}, system{&system_} {} bool CheatList::EvaluateConditional(const Cheat& cheat) const { using ComparisonFunction = bool (*)(u64, u64); @@ -89,10 +88,8 @@ bool CheatList::EvaluateConditional(const Cheat& cheat) const { }; if (cheat.type == CodeType::ConditionalInput) { - const auto applet_resource = Core::System::GetInstance() - .ServiceManager() - .GetService<Service::HID::Hid>("hid") - ->GetAppletResource(); + const auto applet_resource = + system->ServiceManager().GetService<Service::HID::Hid>("hid")->GetAppletResource(); if (applet_resource == nullptr) { LOG_WARNING( Common_Filesystem, @@ -188,8 +185,9 @@ void CheatList::Loop(const Cheat& cheat) { ASSERT(iter != block_pairs.end()); ASSERT(iter->first < iter->second); - for (int i = cheat.Value(4, 4); i >= 0; --i) { - register_3 = i; + const s32 initial_value = static_cast<s32>(cheat.Value(4, sizeof(s32))); + for (s32 i = initial_value; i >= 0; --i) { + register_3 = static_cast<u64>(i); for (std::size_t c = iter->first + 1; c < iter->second; ++c) { current_index = c; ExecuteSingleCheat( @@ -320,14 +318,14 @@ void CheatList::ExecuteBlock(const Block& block) { CheatParser::~CheatParser() = default; -CheatList CheatParser::MakeCheatList(CheatList::ProgramSegment master, +CheatList CheatParser::MakeCheatList(const Core::System& system, CheatList::ProgramSegment master, CheatList::ProgramSegment standard) const { - return {master, standard}; + return {system, std::move(master), std::move(standard)}; } TextCheatParser::~TextCheatParser() = default; -CheatList TextCheatParser::Parse(const std::vector<u8>& data) const { +CheatList TextCheatParser::Parse(const Core::System& system, const std::vector<u8>& data) const { std::stringstream ss; ss.write(reinterpret_cast<const char*>(data.data()), data.size()); @@ -375,7 +373,7 @@ CheatList TextCheatParser::Parse(const std::vector<u8>& data) const { } } - return MakeCheatList(master_list, standard_list); + return MakeCheatList(system, master_list, standard_list); } std::array<u8, 16> TextCheatParser::ParseSingleLineCheat(const std::string& line) const { @@ -425,6 +423,7 @@ std::array<u8, 16> TextCheatParser::ParseSingleLineCheat(const std::string& line return out; } +namespace { u64 MemoryReadImpl(u32 width, VAddr addr) { switch (width) { case 1: @@ -459,17 +458,18 @@ void MemoryWriteImpl(u32 width, VAddr addr, u64 value) { UNREACHABLE(); } } +} // Anonymous namespace -CheatEngine::CheatEngine(std::vector<CheatList> cheats, const std::string& build_id, - VAddr code_region_start, VAddr code_region_end) - : cheats(std::move(cheats)) { - auto& core_timing{Core::System::GetInstance().CoreTiming()}; +CheatEngine::CheatEngine(Core::System& system, std::vector<CheatList> cheats_, + const std::string& build_id, VAddr code_region_start, + VAddr code_region_end) + : cheats{std::move(cheats_)}, core_timing{system.CoreTiming()} { event = core_timing.RegisterEvent( "CheatEngine::FrameCallback::" + build_id, [this](u64 userdata, s64 cycles_late) { FrameCallback(userdata, cycles_late); }); core_timing.ScheduleEvent(CHEAT_ENGINE_TICKS, event); - const auto& vm_manager = Core::System::GetInstance().CurrentProcess()->VMManager(); + const auto& vm_manager = system.CurrentProcess()->VMManager(); for (auto& list : this->cheats) { list.SetMemoryParameters(code_region_start, vm_manager.GetHeapRegionBaseAddress(), code_region_end, vm_manager.GetHeapRegionEndAddress(), @@ -478,15 +478,14 @@ CheatEngine::CheatEngine(std::vector<CheatList> cheats, const std::string& build } CheatEngine::~CheatEngine() { - auto& core_timing{Core::System::GetInstance().CoreTiming()}; core_timing.UnscheduleEvent(event, 0); } -void CheatEngine::FrameCallback(u64 userdata, int cycles_late) { - for (auto& list : cheats) +void CheatEngine::FrameCallback(u64 userdata, s64 cycles_late) { + for (auto& list : cheats) { list.Execute(); + } - auto& core_timing{Core::System::GetInstance().CoreTiming()}; core_timing.ScheduleEvent(CHEAT_ENGINE_TICKS - cycles_late, event); } diff --git a/src/core/file_sys/cheat_engine.h b/src/core/file_sys/cheat_engine.h index 7ed69a2c8..ac22a82cb 100644 --- a/src/core/file_sys/cheat_engine.h +++ b/src/core/file_sys/cheat_engine.h @@ -7,13 +7,17 @@ #include <map> #include <set> #include <vector> -#include <queue> #include "common/bit_field.h" #include "common/common_types.h" +namespace Core { +class System; +} + namespace Core::Timing { +class CoreTiming; struct EventType; -} +} // namespace Core::Timing namespace FileSys { @@ -133,7 +137,7 @@ public: void Execute(); private: - CheatList(ProgramSegment master, ProgramSegment standard); + CheatList(const Core::System& system_, ProgramSegment master, ProgramSegment standard); void ProcessBlockPairs(const Block& block); void ExecuteSingleCheat(const Cheat& cheat); @@ -183,6 +187,8 @@ private: std::map<u64, u64> block_pairs; std::set<u64> encountered_loops; + + const Core::System* system; }; // Intermediary class that parses a text file or other disk format for storing cheats into a @@ -191,10 +197,10 @@ class CheatParser { public: virtual ~CheatParser(); - virtual CheatList Parse(const std::vector<u8>& data) const = 0; + virtual CheatList Parse(const Core::System& system, const std::vector<u8>& data) const = 0; protected: - CheatList MakeCheatList(CheatList::ProgramSegment master, + CheatList MakeCheatList(const Core::System& system_, CheatList::ProgramSegment master, CheatList::ProgramSegment standard) const; }; @@ -203,7 +209,7 @@ class TextCheatParser final : public CheatParser { public: ~TextCheatParser() override; - CheatList Parse(const std::vector<u8>& data) const override; + CheatList Parse(const Core::System& system, const std::vector<u8>& data) const override; private: std::array<u8, 16> ParseSingleLineCheat(const std::string& line) const; @@ -212,16 +218,17 @@ private: // Class that encapsulates a CheatList and manages its interaction with memory and CoreTiming class CheatEngine final { public: - CheatEngine(std::vector<CheatList> cheats, const std::string& build_id, VAddr code_region_start, - VAddr code_region_end); + CheatEngine(Core::System& system_, std::vector<CheatList> cheats_, const std::string& build_id, + VAddr code_region_start, VAddr code_region_end); ~CheatEngine(); private: - void FrameCallback(u64 userdata, int cycles_late); - - Core::Timing::EventType* event; + void FrameCallback(u64 userdata, s64 cycles_late); std::vector<CheatList> cheats; + + Core::Timing::EventType* event; + Core::Timing::CoreTiming& core_timing; }; } // namespace FileSys diff --git a/src/core/file_sys/errors.h b/src/core/file_sys/errors.h index e4a4ee4ab..bb4654366 100644 --- a/src/core/file_sys/errors.h +++ b/src/core/file_sys/errors.h @@ -11,6 +11,9 @@ namespace FileSys { constexpr ResultCode ERROR_PATH_NOT_FOUND{ErrorModule::FS, 1}; constexpr ResultCode ERROR_ENTITY_NOT_FOUND{ErrorModule::FS, 1002}; constexpr ResultCode ERROR_SD_CARD_NOT_FOUND{ErrorModule::FS, 2001}; +constexpr ResultCode ERROR_OUT_OF_BOUNDS{ErrorModule::FS, 3005}; +constexpr ResultCode ERROR_FAILED_MOUNT_ARCHIVE{ErrorModule::FS, 3223}; +constexpr ResultCode ERROR_INVALID_ARGUMENT{ErrorModule::FS, 6001}; constexpr ResultCode ERROR_INVALID_OFFSET{ErrorModule::FS, 6061}; constexpr ResultCode ERROR_INVALID_SIZE{ErrorModule::FS, 6062}; diff --git a/src/core/file_sys/patch_manager.cpp b/src/core/file_sys/patch_manager.cpp index 2b09e5d35..e11217708 100644 --- a/src/core/file_sys/patch_manager.cpp +++ b/src/core/file_sys/patch_manager.cpp @@ -20,6 +20,7 @@ #include "core/file_sys/vfs_vector.h" #include "core/hle/service/filesystem/filesystem.h" #include "core/loader/loader.h" +#include "core/loader/nso.h" #include "core/settings.h" namespace FileSys { @@ -32,14 +33,6 @@ constexpr std::array<const char*, 14> EXEFS_FILE_NAMES{ "subsdk3", "subsdk4", "subsdk5", "subsdk6", "subsdk7", "subsdk8", "subsdk9", }; -struct NSOBuildHeader { - u32_le magic; - INSERT_PADDING_BYTES(0x3C); - std::array<u8, 0x20> build_id; - INSERT_PADDING_BYTES(0xA0); -}; -static_assert(sizeof(NSOBuildHeader) == 0x100, "NSOBuildHeader has incorrect size."); - std::string FormatTitleVersion(u32 version, TitleVersionFormat format) { std::array<u8, sizeof(u32)> bytes{}; bytes[0] = version % SINGLE_BYTE_MODULUS; @@ -163,14 +156,16 @@ std::vector<VirtualFile> PatchManager::CollectPatches(const std::vector<VirtualD } std::vector<u8> PatchManager::PatchNSO(const std::vector<u8>& nso) const { - if (nso.size() < 0x100) + if (nso.size() < sizeof(Loader::NSOHeader)) { return nso; + } - NSOBuildHeader header; - std::memcpy(&header, nso.data(), sizeof(NSOBuildHeader)); + Loader::NSOHeader header; + std::memcpy(&header, nso.data(), sizeof(header)); - if (header.magic != Common::MakeMagic('N', 'S', 'O', '0')) + if (header.magic != Common::MakeMagic('N', 'S', 'O', '0')) { return nso; + } const auto build_id_raw = Common::HexArrayToString(header.build_id); const auto build_id = build_id_raw.substr(0, build_id_raw.find_last_not_of('0') + 1); @@ -213,9 +208,11 @@ std::vector<u8> PatchManager::PatchNSO(const std::vector<u8>& nso) const { } } - if (out.size() < 0x100) + if (out.size() < sizeof(Loader::NSOHeader)) { return nso; - std::memcpy(out.data(), &header, sizeof(NSOBuildHeader)); + } + + std::memcpy(out.data(), &header, sizeof(header)); return out; } @@ -233,7 +230,7 @@ bool PatchManager::HasNSOPatch(const std::array<u8, 32>& build_id_) const { return !CollectPatches(patch_dirs, build_id).empty(); } -static std::optional<CheatList> ReadCheatFileFromFolder(u64 title_id, +static std::optional<CheatList> ReadCheatFileFromFolder(const Core::System& system, u64 title_id, const std::array<u8, 0x20>& build_id_, const VirtualDir& base_path, bool upper) { const auto build_id_raw = Common::HexArrayToString(build_id_, upper); @@ -254,28 +251,28 @@ static std::optional<CheatList> ReadCheatFileFromFolder(u64 title_id, } TextCheatParser parser; - return parser.Parse(data); + return parser.Parse(system, data); } -std::vector<CheatList> PatchManager::CreateCheatList(const std::array<u8, 32>& build_id_) const { - std::vector<CheatList> out; - +std::vector<CheatList> PatchManager::CreateCheatList(const Core::System& system, + const std::array<u8, 32>& build_id_) const { const auto load_dir = Service::FileSystem::GetModificationLoadRoot(title_id); auto patch_dirs = load_dir->GetSubdirectories(); std::sort(patch_dirs.begin(), patch_dirs.end(), [](const VirtualDir& l, const VirtualDir& r) { return l->GetName() < r->GetName(); }); + std::vector<CheatList> out; out.reserve(patch_dirs.size()); for (const auto& subdir : patch_dirs) { auto cheats_dir = subdir->GetSubdirectory("cheats"); if (cheats_dir != nullptr) { - auto res = ReadCheatFileFromFolder(title_id, build_id_, cheats_dir, true); + auto res = ReadCheatFileFromFolder(system, title_id, build_id_, cheats_dir, true); if (res.has_value()) { out.push_back(std::move(*res)); continue; } - res = ReadCheatFileFromFolder(title_id, build_id_, cheats_dir, false); + res = ReadCheatFileFromFolder(system, title_id, build_id_, cheats_dir, false); if (res.has_value()) out.push_back(std::move(*res)); } diff --git a/src/core/file_sys/patch_manager.h b/src/core/file_sys/patch_manager.h index 3e3ac6aca..de2672c76 100644 --- a/src/core/file_sys/patch_manager.h +++ b/src/core/file_sys/patch_manager.h @@ -12,6 +12,10 @@ #include "core/file_sys/nca_metadata.h" #include "core/file_sys/vfs.h" +namespace Core { +class System; +} + namespace FileSys { class NCA; @@ -47,7 +51,8 @@ public: bool HasNSOPatch(const std::array<u8, 0x20>& build_id) const; // Creates a CheatList object with all - std::vector<CheatList> CreateCheatList(const std::array<u8, 0x20>& build_id) const; + std::vector<CheatList> CreateCheatList(const Core::System& system, + const std::array<u8, 0x20>& build_id) const; // Currently tracked RomFS patches: // - Game Updates diff --git a/src/core/file_sys/system_archive/system_archive.cpp b/src/core/file_sys/system_archive/system_archive.cpp index e3e79f40a..c9722ed77 100644 --- a/src/core/file_sys/system_archive/system_archive.cpp +++ b/src/core/file_sys/system_archive/system_archive.cpp @@ -6,6 +6,7 @@ #include "core/file_sys/romfs.h" #include "core/file_sys/system_archive/ng_word.h" #include "core/file_sys/system_archive/system_archive.h" +#include "core/file_sys/system_archive/system_version.h" namespace FileSys::SystemArchive { @@ -30,7 +31,7 @@ constexpr std::array<SystemArchiveDescriptor, SYSTEM_ARCHIVE_COUNT> SYSTEM_ARCHI {0x0100000000000806, "NgWord", &NgWord1}, {0x0100000000000807, "SsidList", nullptr}, {0x0100000000000808, "Dictionary", nullptr}, - {0x0100000000000809, "SystemVersion", nullptr}, + {0x0100000000000809, "SystemVersion", &SystemVersion}, {0x010000000000080A, "AvatarImage", nullptr}, {0x010000000000080B, "LocalNews", nullptr}, {0x010000000000080C, "Eula", nullptr}, diff --git a/src/core/file_sys/system_archive/system_version.cpp b/src/core/file_sys/system_archive/system_version.cpp new file mode 100644 index 000000000..6e22f97b0 --- /dev/null +++ b/src/core/file_sys/system_archive/system_version.cpp @@ -0,0 +1,52 @@ +// Copyright 2019 yuzu emulator team +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include "core/file_sys/system_archive/system_version.h" +#include "core/file_sys/vfs_vector.h" + +namespace FileSys::SystemArchive { + +namespace SystemVersionData { + +// This section should reflect the best system version to describe yuzu's HLE api. +// TODO(DarkLordZach): Update when HLE gets better. + +constexpr u8 VERSION_MAJOR = 5; +constexpr u8 VERSION_MINOR = 1; +constexpr u8 VERSION_MICRO = 0; + +constexpr u8 REVISION_MAJOR = 3; +constexpr u8 REVISION_MINOR = 0; + +constexpr char PLATFORM_STRING[] = "NX"; +constexpr char VERSION_HASH[] = "23f9df53e25709d756e0c76effcb2473bd3447dd"; +constexpr char DISPLAY_VERSION[] = "5.1.0"; +constexpr char DISPLAY_TITLE[] = "NintendoSDK Firmware for NX 5.1.0-3.0"; + +} // namespace SystemVersionData + +std::string GetLongDisplayVersion() { + return SystemVersionData::DISPLAY_TITLE; +} + +VirtualDir SystemVersion() { + VirtualFile file = std::make_shared<VectorVfsFile>(std::vector<u8>(0x100), "file"); + file->WriteObject(SystemVersionData::VERSION_MAJOR, 0); + file->WriteObject(SystemVersionData::VERSION_MINOR, 1); + file->WriteObject(SystemVersionData::VERSION_MICRO, 2); + file->WriteObject(SystemVersionData::REVISION_MAJOR, 4); + file->WriteObject(SystemVersionData::REVISION_MINOR, 5); + file->WriteArray(SystemVersionData::PLATFORM_STRING, + std::min<u64>(sizeof(SystemVersionData::PLATFORM_STRING), 0x20ULL), 0x8); + file->WriteArray(SystemVersionData::VERSION_HASH, + std::min<u64>(sizeof(SystemVersionData::VERSION_HASH), 0x40ULL), 0x28); + file->WriteArray(SystemVersionData::DISPLAY_VERSION, + std::min<u64>(sizeof(SystemVersionData::DISPLAY_VERSION), 0x18ULL), 0x68); + file->WriteArray(SystemVersionData::DISPLAY_TITLE, + std::min<u64>(sizeof(SystemVersionData::DISPLAY_TITLE), 0x80ULL), 0x80); + return std::make_shared<VectorVfsDirectory>(std::vector<VirtualFile>{file}, + std::vector<VirtualDir>{}, "data"); +} + +} // namespace FileSys::SystemArchive diff --git a/src/core/file_sys/system_archive/system_version.h b/src/core/file_sys/system_archive/system_version.h new file mode 100644 index 000000000..deed79b26 --- /dev/null +++ b/src/core/file_sys/system_archive/system_version.h @@ -0,0 +1,16 @@ +// Copyright 2019 yuzu emulator team +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include <string> +#include "core/file_sys/vfs_types.h" + +namespace FileSys::SystemArchive { + +std::string GetLongDisplayVersion(); + +VirtualDir SystemVersion(); + +} // namespace FileSys::SystemArchive diff --git a/src/core/frontend/emu_window.cpp b/src/core/frontend/emu_window.cpp index e29afd630..1320bbe77 100644 --- a/src/core/frontend/emu_window.cpp +++ b/src/core/frontend/emu_window.cpp @@ -30,7 +30,7 @@ private: explicit Device(std::weak_ptr<TouchState>&& touch_state) : touch_state(touch_state) {} std::tuple<float, float, bool> GetStatus() const override { if (auto state = touch_state.lock()) { - std::lock_guard<std::mutex> guard(state->mutex); + std::lock_guard guard{state->mutex}; return std::make_tuple(state->touch_x, state->touch_y, state->touch_pressed); } return std::make_tuple(0.0f, 0.0f, false); @@ -81,7 +81,7 @@ void EmuWindow::TouchPressed(unsigned framebuffer_x, unsigned framebuffer_y) { if (!IsWithinTouchscreen(framebuffer_layout, framebuffer_x, framebuffer_y)) return; - std::lock_guard<std::mutex> guard(touch_state->mutex); + std::lock_guard guard{touch_state->mutex}; touch_state->touch_x = static_cast<float>(framebuffer_x - framebuffer_layout.screen.left) / (framebuffer_layout.screen.right - framebuffer_layout.screen.left); touch_state->touch_y = static_cast<float>(framebuffer_y - framebuffer_layout.screen.top) / @@ -91,7 +91,7 @@ void EmuWindow::TouchPressed(unsigned framebuffer_x, unsigned framebuffer_y) { } void EmuWindow::TouchReleased() { - std::lock_guard<std::mutex> guard(touch_state->mutex); + std::lock_guard guard{touch_state->mutex}; touch_state->touch_pressed = false; touch_state->touch_x = 0; touch_state->touch_y = 0; diff --git a/src/core/hle/kernel/address_arbiter.cpp b/src/core/hle/kernel/address_arbiter.cpp index 352190da8..c8842410b 100644 --- a/src/core/hle/kernel/address_arbiter.cpp +++ b/src/core/hle/kernel/address_arbiter.cpp @@ -26,7 +26,7 @@ void WakeThreads(const std::vector<SharedPtr<Thread>>& waiting_threads, s32 num_ // them all. std::size_t last = waiting_threads.size(); if (num_to_wake > 0) { - last = num_to_wake; + last = std::min(last, static_cast<std::size_t>(num_to_wake)); } // Signal the waiting threads. @@ -90,9 +90,9 @@ ResultCode AddressArbiter::ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr a // Determine the modified value depending on the waiting count. s32 updated_value; if (waiting_threads.empty()) { - updated_value = value - 1; - } else if (num_to_wake <= 0 || waiting_threads.size() <= static_cast<u32>(num_to_wake)) { updated_value = value + 1; + } else if (num_to_wake <= 0 || waiting_threads.size() <= static_cast<u32>(num_to_wake)) { + updated_value = value - 1; } else { updated_value = value; } diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp index 4d224d01d..6baeb3494 100644 --- a/src/core/hle/kernel/kernel.cpp +++ b/src/core/hle/kernel/kernel.cpp @@ -29,12 +29,12 @@ namespace Kernel { * @param thread_handle The handle of the thread that's been awoken * @param cycles_late The number of CPU cycles that have passed since the desired wakeup time */ -static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] int cycles_late) { +static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] s64 cycles_late) { const auto proper_handle = static_cast<Handle>(thread_handle); const auto& system = Core::System::GetInstance(); // Lock the global kernel mutex when we enter the kernel HLE. - std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock); + std::lock_guard lock{HLE::g_hle_lock}; SharedPtr<Thread> thread = system.Kernel().RetrieveThreadFromWakeupCallbackHandleTable(proper_handle); @@ -62,7 +62,8 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] int cycles_ if (thread->GetMutexWaitAddress() != 0 || thread->GetCondVarWaitAddress() != 0 || thread->GetWaitHandle() != 0) { - ASSERT(thread->GetStatus() == ThreadStatus::WaitMutex); + ASSERT(thread->GetStatus() == ThreadStatus::WaitMutex || + thread->GetStatus() == ThreadStatus::WaitCondVar); thread->SetMutexWaitAddress(0); thread->SetCondVarWaitAddress(0); thread->SetWaitHandle(0); diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h index ff17ff865..03ea5b659 100644 --- a/src/core/hle/kernel/kernel.h +++ b/src/core/hle/kernel/kernel.h @@ -8,9 +8,6 @@ #include <unordered_map> #include "core/hle/kernel/object.h" -template <typename T> -class ResultVal; - namespace Core { class System; } diff --git a/src/core/hle/kernel/object.cpp b/src/core/hle/kernel/object.cpp index 8870463d0..217144efc 100644 --- a/src/core/hle/kernel/object.cpp +++ b/src/core/hle/kernel/object.cpp @@ -23,6 +23,7 @@ bool Object::IsWaitable() const { case HandleType::Unknown: case HandleType::WritableEvent: case HandleType::SharedMemory: + case HandleType::TransferMemory: case HandleType::AddressArbiter: case HandleType::ResourceLimit: case HandleType::ClientPort: diff --git a/src/core/hle/kernel/object.h b/src/core/hle/kernel/object.h index 4c2505908..3f6baa094 100644 --- a/src/core/hle/kernel/object.h +++ b/src/core/hle/kernel/object.h @@ -22,6 +22,7 @@ enum class HandleType : u32 { WritableEvent, ReadableEvent, SharedMemory, + TransferMemory, Thread, Process, AddressArbiter, diff --git a/src/core/hle/kernel/process.cpp b/src/core/hle/kernel/process.cpp index 87779a71c..a5144b8ad 100644 --- a/src/core/hle/kernel/process.cpp +++ b/src/core/hle/kernel/process.cpp @@ -5,6 +5,7 @@ #include <algorithm> #include <memory> #include <random> +#include "common/alignment.h" #include "common/assert.h" #include "common/logging/log.h" #include "core/core.h" @@ -75,6 +76,10 @@ SharedPtr<ResourceLimit> Process::GetResourceLimit() const { return resource_limit; } +u64 Process::GetTotalPhysicalMemoryUsed() const { + return vm_manager.GetCurrentHeapSize() + main_thread_stack_size + code_memory_size; +} + ResultCode Process::ClearSignalState() { if (status == ProcessStatus::Exited) { LOG_ERROR(Kernel, "called on a terminated process instance."); @@ -107,14 +112,17 @@ ResultCode Process::LoadFromMetadata(const FileSys::ProgramMetadata& metadata) { return handle_table.SetSize(capabilities.GetHandleTableSize()); } -void Process::Run(VAddr entry_point, s32 main_thread_priority, u32 stack_size) { +void Process::Run(VAddr entry_point, s32 main_thread_priority, u64 stack_size) { + // The kernel always ensures that the given stack size is page aligned. + main_thread_stack_size = Common::AlignUp(stack_size, Memory::PAGE_SIZE); + // Allocate and map the main thread stack // TODO(bunnei): This is heap area that should be allocated by the kernel and not mapped as part // of the user address space. + const VAddr mapping_address = vm_manager.GetTLSIORegionEndAddress() - main_thread_stack_size; vm_manager - .MapMemoryBlock(vm_manager.GetTLSIORegionEndAddress() - stack_size, - std::make_shared<std::vector<u8>>(stack_size, 0), 0, stack_size, - MemoryState::Stack) + .MapMemoryBlock(mapping_address, std::make_shared<std::vector<u8>>(main_thread_stack_size), + 0, main_thread_stack_size, MemoryState::Stack) .Unwrap(); vm_manager.LogLayout(); @@ -226,6 +234,8 @@ void Process::LoadModule(CodeSet module_, VAddr base_addr) { MapSegment(module_.RODataSegment(), VMAPermission::Read, MemoryState::CodeData); MapSegment(module_.DataSegment(), VMAPermission::ReadWrite, MemoryState::CodeData); + code_memory_size += module_.memory->size(); + // Clear instruction cache in CPU JIT system.InvalidateCpuInstructionCaches(); } diff --git a/src/core/hle/kernel/process.h b/src/core/hle/kernel/process.h index 1bd7bf5c1..732d12170 100644 --- a/src/core/hle/kernel/process.h +++ b/src/core/hle/kernel/process.h @@ -35,14 +35,6 @@ class Thread; struct CodeSet; -struct AddressMapping { - // Address and size must be page-aligned - VAddr address; - u64 size; - bool read_only; - bool unk_flag; -}; - enum class MemoryRegion : u16 { APPLICATION = 1, SYSTEM = 2, @@ -194,6 +186,9 @@ public: return random_entropy.at(index); } + /// Retrieves the total physical memory used by this process in bytes. + u64 GetTotalPhysicalMemoryUsed() const; + /// Clears the signaled state of the process if and only if it's signaled. /// /// @pre The process must not be already terminated. If this is called on a @@ -218,7 +213,7 @@ public: /** * Applies address space changes and launches the process main thread. */ - void Run(VAddr entry_point, s32 main_thread_priority, u32 stack_size); + void Run(VAddr entry_point, s32 main_thread_priority, u64 stack_size); /** * Prepares a process for termination by stopping all of its threads @@ -255,6 +250,12 @@ private: /// Memory manager for this process. Kernel::VMManager vm_manager; + /// Size of the main thread's stack in bytes. + u64 main_thread_stack_size = 0; + + /// Size of the loaded code memory in bytes. + u64 code_memory_size = 0; + /// Current status of the process ProcessStatus status; diff --git a/src/core/hle/kernel/scheduler.cpp b/src/core/hle/kernel/scheduler.cpp index cc189cc64..ac501bf7f 100644 --- a/src/core/hle/kernel/scheduler.cpp +++ b/src/core/hle/kernel/scheduler.cpp @@ -29,8 +29,8 @@ Scheduler::~Scheduler() { } bool Scheduler::HaveReadyThreads() const { - std::lock_guard<std::mutex> lock(scheduler_mutex); - return ready_queue.get_first() != nullptr; + std::lock_guard lock{scheduler_mutex}; + return !ready_queue.empty(); } Thread* Scheduler::GetCurrentThread() const { @@ -46,22 +46,27 @@ Thread* Scheduler::PopNextReadyThread() { Thread* thread = GetCurrentThread(); if (thread && thread->GetStatus() == ThreadStatus::Running) { + if (ready_queue.empty()) { + return thread; + } // We have to do better than the current thread. // This call returns null when that's not possible. - next = ready_queue.pop_first_better(thread->GetPriority()); - if (!next) { - // Otherwise just keep going with the current thread + next = ready_queue.front(); + if (next == nullptr || next->GetPriority() >= thread->GetPriority()) { next = thread; } } else { - next = ready_queue.pop_first(); + if (ready_queue.empty()) { + return nullptr; + } + next = ready_queue.front(); } return next; } void Scheduler::SwitchContext(Thread* new_thread) { - Thread* const previous_thread = GetCurrentThread(); + Thread* previous_thread = GetCurrentThread(); Process* const previous_process = system.Kernel().CurrentProcess(); UpdateLastContextSwitchTime(previous_thread, previous_process); @@ -75,7 +80,7 @@ void Scheduler::SwitchContext(Thread* new_thread) { if (previous_thread->GetStatus() == ThreadStatus::Running) { // This is only the case when a reschedule is triggered without the current thread // yielding execution (i.e. an event triggered, system core time-sliced, etc) - ready_queue.push_front(previous_thread->GetPriority(), previous_thread); + ready_queue.add(previous_thread, previous_thread->GetPriority(), false); previous_thread->SetStatus(ThreadStatus::Ready); } } @@ -90,7 +95,7 @@ void Scheduler::SwitchContext(Thread* new_thread) { current_thread = new_thread; - ready_queue.remove(new_thread->GetPriority(), new_thread); + ready_queue.remove(new_thread, new_thread->GetPriority()); new_thread->SetStatus(ThreadStatus::Running); auto* const thread_owner_process = current_thread->GetOwnerProcess(); @@ -127,7 +132,7 @@ void Scheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) { } void Scheduler::Reschedule() { - std::lock_guard<std::mutex> lock(scheduler_mutex); + std::lock_guard lock{scheduler_mutex}; Thread* cur = GetCurrentThread(); Thread* next = PopNextReadyThread(); @@ -143,51 +148,54 @@ void Scheduler::Reschedule() { SwitchContext(next); } -void Scheduler::AddThread(SharedPtr<Thread> thread, u32 priority) { - std::lock_guard<std::mutex> lock(scheduler_mutex); +void Scheduler::AddThread(SharedPtr<Thread> thread) { + std::lock_guard lock{scheduler_mutex}; thread_list.push_back(std::move(thread)); - ready_queue.prepare(priority); } void Scheduler::RemoveThread(Thread* thread) { - std::lock_guard<std::mutex> lock(scheduler_mutex); + std::lock_guard lock{scheduler_mutex}; thread_list.erase(std::remove(thread_list.begin(), thread_list.end(), thread), thread_list.end()); } void Scheduler::ScheduleThread(Thread* thread, u32 priority) { - std::lock_guard<std::mutex> lock(scheduler_mutex); + std::lock_guard lock{scheduler_mutex}; ASSERT(thread->GetStatus() == ThreadStatus::Ready); - ready_queue.push_back(priority, thread); + ready_queue.add(thread, priority); } void Scheduler::UnscheduleThread(Thread* thread, u32 priority) { - std::lock_guard<std::mutex> lock(scheduler_mutex); + std::lock_guard lock{scheduler_mutex}; ASSERT(thread->GetStatus() == ThreadStatus::Ready); - ready_queue.remove(priority, thread); + ready_queue.remove(thread, priority); } void Scheduler::SetThreadPriority(Thread* thread, u32 priority) { - std::lock_guard<std::mutex> lock(scheduler_mutex); + std::lock_guard lock{scheduler_mutex}; + if (thread->GetPriority() == priority) { + return; + } // If thread was ready, adjust queues if (thread->GetStatus() == ThreadStatus::Ready) - ready_queue.move(thread, thread->GetPriority(), priority); - else - ready_queue.prepare(priority); + ready_queue.adjust(thread, thread->GetPriority(), priority); } Thread* Scheduler::GetNextSuggestedThread(u32 core, u32 maximum_priority) const { - std::lock_guard<std::mutex> lock(scheduler_mutex); + std::lock_guard lock{scheduler_mutex}; const u32 mask = 1U << core; - return ready_queue.get_first_filter([mask, maximum_priority](Thread const* thread) { - return (thread->GetAffinityMask() & mask) != 0 && thread->GetPriority() < maximum_priority; - }); + for (auto* thread : ready_queue) { + if ((thread->GetAffinityMask() & mask) != 0 && thread->GetPriority() < maximum_priority) { + return thread; + } + } + return nullptr; } void Scheduler::YieldWithoutLoadBalancing(Thread* thread) { diff --git a/src/core/hle/kernel/scheduler.h b/src/core/hle/kernel/scheduler.h index 1c5bf57d9..b29bf7be8 100644 --- a/src/core/hle/kernel/scheduler.h +++ b/src/core/hle/kernel/scheduler.h @@ -7,7 +7,7 @@ #include <mutex> #include <vector> #include "common/common_types.h" -#include "common/thread_queue_list.h" +#include "common/multi_level_queue.h" #include "core/hle/kernel/object.h" #include "core/hle/kernel/thread.h" @@ -38,7 +38,7 @@ public: u64 GetLastContextSwitchTicks() const; /// Adds a new thread to the scheduler - void AddThread(SharedPtr<Thread> thread, u32 priority); + void AddThread(SharedPtr<Thread> thread); /// Removes a thread from the scheduler void RemoveThread(Thread* thread); @@ -156,7 +156,7 @@ private: std::vector<SharedPtr<Thread>> thread_list; /// Lists only ready thread ids. - Common::ThreadQueueList<Thread*, THREADPRIO_LOWEST + 1> ready_queue; + Common::MultiLevelQueue<Thread*, THREADPRIO_LOWEST + 1> ready_queue; SharedPtr<Thread> current_thread = nullptr; diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp index a6a17efe7..76a8b0191 100644 --- a/src/core/hle/kernel/svc.cpp +++ b/src/core/hle/kernel/svc.cpp @@ -32,6 +32,7 @@ #include "core/hle/kernel/svc.h" #include "core/hle/kernel/svc_wrap.h" #include "core/hle/kernel/thread.h" +#include "core/hle/kernel/transfer_memory.h" #include "core/hle/kernel/writable_event.h" #include "core/hle/lock.h" #include "core/hle/result.h" @@ -174,11 +175,8 @@ static ResultCode SetHeapSize(VAddr* heap_addr, u64 heap_size) { return ERR_INVALID_SIZE; } - auto& vm_manager = Core::CurrentProcess()->VMManager(); - const VAddr heap_base = vm_manager.GetHeapRegionBaseAddress(); - const auto alloc_result = - vm_manager.HeapAllocate(heap_base, heap_size, VMAPermission::ReadWrite); - + auto& vm_manager = Core::System::GetInstance().Kernel().CurrentProcess()->VMManager(); + const auto alloc_result = vm_manager.SetHeapSize(heap_size); if (alloc_result.Failed()) { return alloc_result.Code(); } @@ -711,7 +709,7 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id) HeapRegionBaseAddr = 4, HeapRegionSize = 5, TotalMemoryUsage = 6, - TotalHeapUsage = 7, + TotalPhysicalMemoryUsed = 7, IsCurrentProcessBeingDebugged = 8, RegisterResourceLimit = 9, IdleTickCount = 10, @@ -747,7 +745,7 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id) case GetInfoType::NewMapRegionBaseAddr: case GetInfoType::NewMapRegionSize: case GetInfoType::TotalMemoryUsage: - case GetInfoType::TotalHeapUsage: + case GetInfoType::TotalPhysicalMemoryUsed: case GetInfoType::IsVirtualAddressMemoryEnabled: case GetInfoType::PersonalMmHeapUsage: case GetInfoType::TitleId: @@ -807,8 +805,8 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id) *result = process->VMManager().GetTotalMemoryUsage(); return RESULT_SUCCESS; - case GetInfoType::TotalHeapUsage: - *result = process->VMManager().GetTotalHeapUsage(); + case GetInfoType::TotalPhysicalMemoryUsed: + *result = process->GetTotalPhysicalMemoryUsed(); return RESULT_SUCCESS; case GetInfoType::IsVirtualAddressMemoryEnabled: @@ -1355,7 +1353,7 @@ static ResultCode WaitProcessWideKeyAtomic(VAddr mutex_addr, VAddr condition_var current_thread->SetCondVarWaitAddress(condition_variable_addr); current_thread->SetMutexWaitAddress(mutex_addr); current_thread->SetWaitHandle(thread_handle); - current_thread->SetStatus(ThreadStatus::WaitMutex); + current_thread->SetStatus(ThreadStatus::WaitCondVar); current_thread->InvalidateWakeupCallback(); current_thread->WakeAfterDelay(nano_seconds); @@ -1399,10 +1397,10 @@ static ResultCode SignalProcessWideKey(VAddr condition_variable_addr, s32 target // them all. std::size_t last = waiting_threads.size(); if (target != -1) - last = target; + last = std::min(waiting_threads.size(), static_cast<std::size_t>(target)); // If there are no threads waiting on this condition variable, just exit - if (last > waiting_threads.size()) + if (last == 0) return RESULT_SUCCESS; for (std::size_t index = 0; index < last; ++index) { @@ -1410,6 +1408,9 @@ static ResultCode SignalProcessWideKey(VAddr condition_variable_addr, s32 target ASSERT(thread->GetCondVarWaitAddress() == condition_variable_addr); + // liberate Cond Var Thread. + thread->SetCondVarWaitAddress(0); + std::size_t current_core = Core::System::GetInstance().CurrentCoreIndex(); auto& monitor = Core::System::GetInstance().Monitor(); @@ -1428,10 +1429,9 @@ static ResultCode SignalProcessWideKey(VAddr condition_variable_addr, s32 target } } while (!monitor.ExclusiveWrite32(current_core, thread->GetMutexWaitAddress(), thread->GetWaitHandle())); - if (mutex_val == 0) { // We were able to acquire the mutex, resume this thread. - ASSERT(thread->GetStatus() == ThreadStatus::WaitMutex); + ASSERT(thread->GetStatus() == ThreadStatus::WaitCondVar); thread->ResumeFromWait(); auto* const lock_owner = thread->GetLockOwner(); @@ -1441,8 +1441,8 @@ static ResultCode SignalProcessWideKey(VAddr condition_variable_addr, s32 target thread->SetLockOwner(nullptr); thread->SetMutexWaitAddress(0); - thread->SetCondVarWaitAddress(0); thread->SetWaitHandle(0); + Core::System::GetInstance().CpuCore(thread->GetProcessorID()).PrepareReschedule(); } else { // Atomically signal that the mutex now has a waiting thread. do { @@ -1461,12 +1461,11 @@ static ResultCode SignalProcessWideKey(VAddr condition_variable_addr, s32 target const auto& handle_table = Core::CurrentProcess()->GetHandleTable(); auto owner = handle_table.Get<Thread>(owner_handle); ASSERT(owner); - ASSERT(thread->GetStatus() == ThreadStatus::WaitMutex); + ASSERT(thread->GetStatus() == ThreadStatus::WaitCondVar); thread->InvalidateWakeupCallback(); + thread->SetStatus(ThreadStatus::WaitMutex); owner->AddMutexWaiter(thread); - - Core::System::GetInstance().CpuCore(thread->GetProcessorID()).PrepareReschedule(); } } @@ -1586,14 +1585,121 @@ static ResultCode CreateTransferMemory(Handle* handle, VAddr addr, u64 size, u32 } auto& kernel = Core::System::GetInstance().Kernel(); - auto process = kernel.CurrentProcess(); - auto& handle_table = process->GetHandleTable(); - const auto shared_mem_handle = SharedMemory::Create(kernel, process, size, perms, perms, addr); + auto transfer_mem_handle = TransferMemory::Create(kernel, addr, size, perms); - CASCADE_RESULT(*handle, handle_table.Create(shared_mem_handle)); + auto& handle_table = kernel.CurrentProcess()->GetHandleTable(); + const auto result = handle_table.Create(std::move(transfer_mem_handle)); + if (result.Failed()) { + return result.Code(); + } + + *handle = *result; return RESULT_SUCCESS; } +static ResultCode MapTransferMemory(Handle handle, VAddr address, u64 size, u32 permission_raw) { + LOG_DEBUG(Kernel_SVC, + "called. handle=0x{:08X}, address=0x{:016X}, size=0x{:016X}, permissions=0x{:08X}", + handle, address, size, permission_raw); + + if (!Common::Is4KBAligned(address)) { + LOG_ERROR(Kernel_SVC, "Transfer memory addresses must be 4KB aligned (size=0x{:016X}).", + address); + return ERR_INVALID_ADDRESS; + } + + if (size == 0 || !Common::Is4KBAligned(size)) { + LOG_ERROR(Kernel_SVC, + "Transfer memory sizes must be 4KB aligned and not be zero (size=0x{:016X}).", + size); + return ERR_INVALID_SIZE; + } + + if (!IsValidAddressRange(address, size)) { + LOG_ERROR(Kernel_SVC, + "Given address and size overflows the 64-bit range (address=0x{:016X}, " + "size=0x{:016X}).", + address, size); + return ERR_INVALID_ADDRESS_STATE; + } + + const auto permissions = static_cast<MemoryPermission>(permission_raw); + if (permissions != MemoryPermission::None && permissions != MemoryPermission::Read && + permissions != MemoryPermission::ReadWrite) { + LOG_ERROR(Kernel_SVC, "Invalid transfer memory permissions given (permissions=0x{:08X}).", + permission_raw); + return ERR_INVALID_STATE; + } + + const auto& kernel = Core::System::GetInstance().Kernel(); + const auto* const current_process = kernel.CurrentProcess(); + const auto& handle_table = current_process->GetHandleTable(); + + auto transfer_memory = handle_table.Get<TransferMemory>(handle); + if (!transfer_memory) { + LOG_ERROR(Kernel_SVC, "Nonexistent transfer memory handle given (handle=0x{:08X}).", + handle); + return ERR_INVALID_HANDLE; + } + + if (!current_process->VMManager().IsWithinASLRRegion(address, size)) { + LOG_ERROR(Kernel_SVC, + "Given address and size don't fully fit within the ASLR region " + "(address=0x{:016X}, size=0x{:016X}).", + address, size); + return ERR_INVALID_MEMORY_RANGE; + } + + return transfer_memory->MapMemory(address, size, permissions); +} + +static ResultCode UnmapTransferMemory(Handle handle, VAddr address, u64 size) { + LOG_DEBUG(Kernel_SVC, "called. handle=0x{:08X}, address=0x{:016X}, size=0x{:016X}", handle, + address, size); + + if (!Common::Is4KBAligned(address)) { + LOG_ERROR(Kernel_SVC, "Transfer memory addresses must be 4KB aligned (size=0x{:016X}).", + address); + return ERR_INVALID_ADDRESS; + } + + if (size == 0 || !Common::Is4KBAligned(size)) { + LOG_ERROR(Kernel_SVC, + "Transfer memory sizes must be 4KB aligned and not be zero (size=0x{:016X}).", + size); + return ERR_INVALID_SIZE; + } + + if (!IsValidAddressRange(address, size)) { + LOG_ERROR(Kernel_SVC, + "Given address and size overflows the 64-bit range (address=0x{:016X}, " + "size=0x{:016X}).", + address, size); + return ERR_INVALID_ADDRESS_STATE; + } + + const auto& kernel = Core::System::GetInstance().Kernel(); + const auto* const current_process = kernel.CurrentProcess(); + const auto& handle_table = current_process->GetHandleTable(); + + auto transfer_memory = handle_table.Get<TransferMemory>(handle); + if (!transfer_memory) { + LOG_ERROR(Kernel_SVC, "Nonexistent transfer memory handle given (handle=0x{:08X}).", + handle); + return ERR_INVALID_HANDLE; + } + + if (!current_process->VMManager().IsWithinASLRRegion(address, size)) { + LOG_ERROR(Kernel_SVC, + "Given address and size don't fully fit within the ASLR region " + "(address=0x{:016X}, size=0x{:016X}).", + address, size); + return ERR_INVALID_MEMORY_RANGE; + } + + return transfer_memory->UnmapMemory(address, size); +} + static ResultCode GetThreadCoreMask(Handle thread_handle, u32* core, u64* mask) { LOG_TRACE(Kernel_SVC, "called, handle=0x{:08X}", thread_handle); @@ -1969,8 +2075,8 @@ static const FunctionDef SVC_Table[] = { {0x4E, nullptr, "ReadWriteRegister"}, {0x4F, nullptr, "SetProcessActivity"}, {0x50, SvcWrap<CreateSharedMemory>, "CreateSharedMemory"}, - {0x51, nullptr, "MapTransferMemory"}, - {0x52, nullptr, "UnmapTransferMemory"}, + {0x51, SvcWrap<MapTransferMemory>, "MapTransferMemory"}, + {0x52, SvcWrap<UnmapTransferMemory>, "UnmapTransferMemory"}, {0x53, nullptr, "CreateInterruptEvent"}, {0x54, nullptr, "QueryPhysicalAddress"}, {0x55, nullptr, "QueryIoMapping"}, @@ -2032,7 +2138,7 @@ void CallSVC(u32 immediate) { MICROPROFILE_SCOPE(Kernel_SVC); // Lock the global kernel mutex when we enter the kernel HLE. - std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock); + std::lock_guard lock{HLE::g_hle_lock}; const FunctionDef* info = GetSVCInfo(immediate); if (info) { diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp index 3b22e8e0d..fa3ac3abc 100644 --- a/src/core/hle/kernel/thread.cpp +++ b/src/core/hle/kernel/thread.cpp @@ -105,6 +105,7 @@ void Thread::ResumeFromWait() { case ThreadStatus::WaitSleep: case ThreadStatus::WaitIPC: case ThreadStatus::WaitMutex: + case ThreadStatus::WaitCondVar: case ThreadStatus::WaitArb: break; @@ -198,7 +199,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name thread->callback_handle = kernel.ThreadWakeupCallbackHandleTable().Create(thread).Unwrap(); thread->owner_process = &owner_process; thread->scheduler = &system.Scheduler(processor_id); - thread->scheduler->AddThread(thread, priority); + thread->scheduler->AddThread(thread); thread->tls_address = thread->owner_process->MarkNextAvailableTLSSlotAsUsed(*thread); // TODO(peachum): move to ScheduleThread() when scheduler is added so selected core is used @@ -351,7 +352,7 @@ void Thread::ChangeScheduler() { if (*new_processor_id != processor_id) { // Remove thread from previous core's scheduler scheduler->RemoveThread(this); - next_scheduler.AddThread(this, current_priority); + next_scheduler.AddThread(this); } processor_id = *new_processor_id; diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h index faad5f391..9c684758c 100644 --- a/src/core/hle/kernel/thread.h +++ b/src/core/hle/kernel/thread.h @@ -51,7 +51,8 @@ enum class ThreadStatus { WaitIPC, ///< Waiting for the reply from an IPC request WaitSynchAny, ///< Waiting due to WaitSynch1 or WaitSynchN with wait_all = false WaitSynchAll, ///< Waiting due to WaitSynchronizationN with wait_all = true - WaitMutex, ///< Waiting due to an ArbitrateLock/WaitProcessWideKey svc + WaitMutex, ///< Waiting due to an ArbitrateLock svc + WaitCondVar, ///< Waiting due to an WaitProcessWideKey svc WaitArb, ///< Waiting due to a SignalToAddress/WaitForAddress svc Dormant, ///< Created but not yet made ready Dead ///< Run to completion, or forcefully terminated diff --git a/src/core/hle/kernel/transfer_memory.cpp b/src/core/hle/kernel/transfer_memory.cpp new file mode 100644 index 000000000..23228e1b5 --- /dev/null +++ b/src/core/hle/kernel/transfer_memory.cpp @@ -0,0 +1,73 @@ +// Copyright 2019 yuzu emulator team +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include "core/hle/kernel/errors.h" +#include "core/hle/kernel/kernel.h" +#include "core/hle/kernel/process.h" +#include "core/hle/kernel/shared_memory.h" +#include "core/hle/kernel/transfer_memory.h" +#include "core/hle/result.h" + +namespace Kernel { + +TransferMemory::TransferMemory(KernelCore& kernel) : Object{kernel} {} +TransferMemory::~TransferMemory() = default; + +SharedPtr<TransferMemory> TransferMemory::Create(KernelCore& kernel, VAddr base_address, + size_t size, MemoryPermission permissions) { + SharedPtr<TransferMemory> transfer_memory{new TransferMemory(kernel)}; + + transfer_memory->base_address = base_address; + transfer_memory->memory_size = size; + transfer_memory->owner_permissions = permissions; + transfer_memory->owner_process = kernel.CurrentProcess(); + + return transfer_memory; +} + +ResultCode TransferMemory::MapMemory(VAddr address, size_t size, MemoryPermission permissions) { + if (memory_size != size) { + return ERR_INVALID_SIZE; + } + + if (owner_permissions != permissions) { + return ERR_INVALID_STATE; + } + + if (is_mapped) { + return ERR_INVALID_STATE; + } + + const auto map_state = owner_permissions == MemoryPermission::None + ? MemoryState::TransferMemoryIsolated + : MemoryState::TransferMemory; + auto& vm_manager = owner_process->VMManager(); + const auto map_result = vm_manager.MapMemoryBlock( + address, std::make_shared<std::vector<u8>>(size), 0, size, map_state); + + if (map_result.Failed()) { + return map_result.Code(); + } + + is_mapped = true; + return RESULT_SUCCESS; +} + +ResultCode TransferMemory::UnmapMemory(VAddr address, size_t size) { + if (memory_size != size) { + return ERR_INVALID_SIZE; + } + + auto& vm_manager = owner_process->VMManager(); + const auto result = vm_manager.UnmapRange(address, size); + + if (result.IsError()) { + return result; + } + + is_mapped = false; + return RESULT_SUCCESS; +} + +} // namespace Kernel diff --git a/src/core/hle/kernel/transfer_memory.h b/src/core/hle/kernel/transfer_memory.h new file mode 100644 index 000000000..ec294951e --- /dev/null +++ b/src/core/hle/kernel/transfer_memory.h @@ -0,0 +1,91 @@ +// Copyright 2019 yuzu emulator team +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include "core/hle/kernel/object.h" + +union ResultCode; + +namespace Kernel { + +class KernelCore; +class Process; + +enum class MemoryPermission : u32; + +/// Defines the interface for transfer memory objects. +/// +/// Transfer memory is typically used for the purpose of +/// transferring memory between separate process instances, +/// thus the name. +/// +class TransferMemory final : public Object { +public: + static constexpr HandleType HANDLE_TYPE = HandleType::TransferMemory; + + static SharedPtr<TransferMemory> Create(KernelCore& kernel, VAddr base_address, size_t size, + MemoryPermission permissions); + + TransferMemory(const TransferMemory&) = delete; + TransferMemory& operator=(const TransferMemory&) = delete; + + TransferMemory(TransferMemory&&) = delete; + TransferMemory& operator=(TransferMemory&&) = delete; + + std::string GetTypeName() const override { + return "TransferMemory"; + } + + std::string GetName() const override { + return GetTypeName(); + } + + HandleType GetHandleType() const override { + return HANDLE_TYPE; + } + + /// Attempts to map transfer memory with the given range and memory permissions. + /// + /// @param address The base address to being mapping memory at. + /// @param size The size of the memory to map, in bytes. + /// @param permissions The memory permissions to check against when mapping memory. + /// + /// @pre The given address, size, and memory permissions must all match + /// the same values that were given when creating the transfer memory + /// instance. + /// + ResultCode MapMemory(VAddr address, size_t size, MemoryPermission permissions); + + /// Unmaps the transfer memory with the given range + /// + /// @param address The base address to begin unmapping memory at. + /// @param size The size of the memory to unmap, in bytes. + /// + /// @pre The given address and size must be the same as the ones used + /// to create the transfer memory instance. + /// + ResultCode UnmapMemory(VAddr address, size_t size); + +private: + explicit TransferMemory(KernelCore& kernel); + ~TransferMemory() override; + + /// The base address for the memory managed by this instance. + VAddr base_address = 0; + + /// Size of the memory, in bytes, that this instance manages. + size_t memory_size = 0; + + /// The memory permissions that are applied to this instance. + MemoryPermission owner_permissions{}; + + /// The process that this transfer memory instance was created under. + Process* owner_process = nullptr; + + /// Whether or not this transfer memory instance has mapped memory. + bool is_mapped = false; +}; + +} // namespace Kernel diff --git a/src/core/hle/kernel/vm_manager.cpp b/src/core/hle/kernel/vm_manager.cpp index 22bf55ce7..ec0a480ce 100644 --- a/src/core/hle/kernel/vm_manager.cpp +++ b/src/core/hle/kernel/vm_manager.cpp @@ -256,57 +256,50 @@ ResultCode VMManager::ReprotectRange(VAddr target, u64 size, VMAPermission new_p return RESULT_SUCCESS; } -ResultVal<VAddr> VMManager::HeapAllocate(VAddr target, u64 size, VMAPermission perms) { - if (!IsWithinHeapRegion(target, size)) { - return ERR_INVALID_ADDRESS; +ResultVal<VAddr> VMManager::SetHeapSize(u64 size) { + if (size > GetHeapRegionSize()) { + return ERR_OUT_OF_MEMORY; + } + + // No need to do any additional work if the heap is already the given size. + if (size == GetCurrentHeapSize()) { + return MakeResult(heap_region_base); } if (heap_memory == nullptr) { // Initialize heap - heap_memory = std::make_shared<std::vector<u8>>(); - heap_start = heap_end = target; + heap_memory = std::make_shared<std::vector<u8>>(size); + heap_end = heap_region_base + size; } else { - UnmapRange(heap_start, heap_end - heap_start); - } - - // If necessary, expand backing vector to cover new heap extents. - if (target < heap_start) { - heap_memory->insert(begin(*heap_memory), heap_start - target, 0); - heap_start = target; - RefreshMemoryBlockMappings(heap_memory.get()); - } - if (target + size > heap_end) { - heap_memory->insert(end(*heap_memory), (target + size) - heap_end, 0); - heap_end = target + size; - RefreshMemoryBlockMappings(heap_memory.get()); + UnmapRange(heap_region_base, GetCurrentHeapSize()); } - ASSERT(heap_end - heap_start == heap_memory->size()); - CASCADE_RESULT(auto vma, MapMemoryBlock(target, heap_memory, target - heap_start, size, - MemoryState::Heap)); - Reprotect(vma, perms); + // If necessary, expand backing vector to cover new heap extents in + // the case of allocating. Otherwise, shrink the backing memory, + // if a smaller heap has been requested. + const u64 old_heap_size = GetCurrentHeapSize(); + if (size > old_heap_size) { + const u64 alloc_size = size - old_heap_size; - heap_used = size; - - return MakeResult<VAddr>(heap_end - size); -} + heap_memory->insert(heap_memory->end(), alloc_size, 0); + RefreshMemoryBlockMappings(heap_memory.get()); + } else if (size < old_heap_size) { + heap_memory->resize(size); + heap_memory->shrink_to_fit(); -ResultCode VMManager::HeapFree(VAddr target, u64 size) { - if (!IsWithinHeapRegion(target, size)) { - return ERR_INVALID_ADDRESS; + RefreshMemoryBlockMappings(heap_memory.get()); } - if (size == 0) { - return RESULT_SUCCESS; - } + heap_end = heap_region_base + size; + ASSERT(GetCurrentHeapSize() == heap_memory->size()); - const ResultCode result = UnmapRange(target, size); - if (result.IsError()) { - return result; + const auto mapping_result = + MapMemoryBlock(heap_region_base, heap_memory, 0, size, MemoryState::Heap); + if (mapping_result.Failed()) { + return mapping_result.Code(); } - heap_used -= size; - return RESULT_SUCCESS; + return MakeResult<VAddr>(heap_region_base); } MemoryInfo VMManager::QueryMemory(VAddr address) const { @@ -598,6 +591,7 @@ void VMManager::InitializeMemoryRegionRanges(FileSys::ProgramAddressSpaceType ty heap_region_base = map_region_end; heap_region_end = heap_region_base + heap_region_size; + heap_end = heap_region_base; new_map_region_base = heap_region_end; new_map_region_end = new_map_region_base + new_map_region_size; @@ -692,10 +686,6 @@ u64 VMManager::GetTotalMemoryUsage() const { return 0xF8000000; } -u64 VMManager::GetTotalHeapUsage() const { - return heap_used; -} - VAddr VMManager::GetAddressSpaceBaseAddress() const { return address_space_base; } @@ -778,6 +768,10 @@ u64 VMManager::GetHeapRegionSize() const { return heap_region_end - heap_region_base; } +u64 VMManager::GetCurrentHeapSize() const { + return heap_end - heap_region_base; +} + bool VMManager::IsWithinHeapRegion(VAddr address, u64 size) const { return IsInsideAddressRange(address, size, GetHeapRegionBaseAddress(), GetHeapRegionEndAddress()); diff --git a/src/core/hle/kernel/vm_manager.h b/src/core/hle/kernel/vm_manager.h index 7cdff6094..6f484b7bf 100644 --- a/src/core/hle/kernel/vm_manager.h +++ b/src/core/hle/kernel/vm_manager.h @@ -380,11 +380,41 @@ public: /// Changes the permissions of a range of addresses, splitting VMAs as necessary. ResultCode ReprotectRange(VAddr target, u64 size, VMAPermission new_perms); - ResultVal<VAddr> HeapAllocate(VAddr target, u64 size, VMAPermission perms); - ResultCode HeapFree(VAddr target, u64 size); - ResultCode MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size, MemoryState state); + /// Attempts to allocate a heap with the given size. + /// + /// @param size The size of the heap to allocate in bytes. + /// + /// @note If a heap is currently allocated, and this is called + /// with a size that is equal to the size of the current heap, + /// then this function will do nothing and return the current + /// heap's starting address, as there's no need to perform + /// any additional heap allocation work. + /// + /// @note If a heap is currently allocated, and this is called + /// with a size less than the current heap's size, then + /// this function will attempt to shrink the heap. + /// + /// @note If a heap is currently allocated, and this is called + /// with a size larger than the current heap's size, then + /// this function will attempt to extend the size of the heap. + /// + /// @returns A result indicating either success or failure. + /// <p> + /// If successful, this function will return a result + /// containing the starting address to the allocated heap. + /// <p> + /// If unsuccessful, this function will return a result + /// containing an error code. + /// + /// @pre The given size must lie within the allowable heap + /// memory region managed by this VMManager instance. + /// Failure to abide by this will result in ERR_OUT_OF_MEMORY + /// being returned as the result. + /// + ResultVal<VAddr> SetHeapSize(u64 size); + /// Queries the memory manager for information about the given address. /// /// @param address The address to query the memory manager about for information. @@ -418,9 +448,6 @@ public: /// Gets the total memory usage, used by svcGetInfo u64 GetTotalMemoryUsage() const; - /// Gets the total heap usage, used by svcGetInfo - u64 GetTotalHeapUsage() const; - /// Gets the address space base address VAddr GetAddressSpaceBaseAddress() const; @@ -469,6 +496,13 @@ public: /// Gets the total size of the heap region in bytes. u64 GetHeapRegionSize() const; + /// Gets the total size of the current heap in bytes. + /// + /// @note This is the current allocated heap size, not the size + /// of the region it's allowed to exist within. + /// + u64 GetCurrentHeapSize() const; + /// Determines whether or not the specified range is within the heap region. bool IsWithinHeapRegion(VAddr address, u64 size) const; @@ -617,9 +651,6 @@ private: VAddr new_map_region_base = 0; VAddr new_map_region_end = 0; - VAddr main_code_region_base = 0; - VAddr main_code_region_end = 0; - VAddr tls_io_region_base = 0; VAddr tls_io_region_end = 0; @@ -628,9 +659,9 @@ private: // This makes deallocation and reallocation of holes fast and keeps process memory contiguous // in the emulator address space, allowing Memory::GetPointer to be reasonably safe. std::shared_ptr<std::vector<u8>> heap_memory; - // The left/right bounds of the address space covered by heap_memory. - VAddr heap_start = 0; + + // The end of the currently allocated heap. This is not an inclusive + // end of the range. This is essentially 'base_address + current_size'. VAddr heap_end = 0; - u64 heap_used = 0; }; } // namespace Kernel diff --git a/src/core/hle/service/fatal/fatal.cpp b/src/core/hle/service/fatal/fatal.cpp index 770590d0b..2c229bcad 100644 --- a/src/core/hle/service/fatal/fatal.cpp +++ b/src/core/hle/service/fatal/fatal.cpp @@ -25,21 +25,34 @@ Module::Interface::Interface(std::shared_ptr<Module> module, const char* name) Module::Interface::~Interface() = default; struct FatalInfo { - std::array<u64_le, 31> registers{}; // TODO(ogniK): See if this actually is registers or - // not(find a game which has non zero valeus) - u64_le unk0{}; - u64_le unk1{}; - u64_le unk2{}; - u64_le unk3{}; - u64_le unk4{}; - u64_le unk5{}; - u64_le unk6{}; + enum class Architecture : s32 { + AArch64, + AArch32, + }; + + const char* ArchAsString() const { + return arch == Architecture::AArch64 ? "AArch64" : "AArch32"; + } + + std::array<u64_le, 31> registers{}; + u64_le sp{}; + u64_le pc{}; + u64_le pstate{}; + u64_le afsr0{}; + u64_le afsr1{}; + u64_le esr{}; + u64_le far{}; std::array<u64_le, 32> backtrace{}; - u64_le unk7{}; - u64_le unk8{}; + u64_le program_entry_point{}; + + // Bit flags that indicate which registers have been set with values + // for this context. The service itself uses these to determine which + // registers to specifically print out. + u64_le set_flags{}; + u32_le backtrace_size{}; - u32_le unk9{}; + Architecture arch{}; u32_le unk10{}; // TODO(ogniK): Is this even used or is it just padding? }; static_assert(sizeof(FatalInfo) == 0x250, "FatalInfo is an invalid size"); @@ -52,36 +65,36 @@ enum class FatalType : u32 { static void GenerateErrorReport(ResultCode error_code, const FatalInfo& info) { const auto title_id = Core::CurrentProcess()->GetTitleID(); - std::string crash_report = - fmt::format("Yuzu {}-{} crash report\n" - "Title ID: {:016x}\n" - "Result: 0x{:X} ({:04}-{:04d})\n" - "\n", - Common::g_scm_branch, Common::g_scm_desc, title_id, error_code.raw, - 2000 + static_cast<u32>(error_code.module.Value()), - static_cast<u32>(error_code.description.Value()), info.unk8, info.unk7); + std::string crash_report = fmt::format( + "Yuzu {}-{} crash report\n" + "Title ID: {:016x}\n" + "Result: 0x{:X} ({:04}-{:04d})\n" + "Set flags: 0x{:16X}\n" + "Program entry point: 0x{:16X}\n" + "\n", + Common::g_scm_branch, Common::g_scm_desc, title_id, error_code.raw, + 2000 + static_cast<u32>(error_code.module.Value()), + static_cast<u32>(error_code.description.Value()), info.set_flags, info.program_entry_point); if (info.backtrace_size != 0x0) { crash_report += "Registers:\n"; - // TODO(ogniK): This is just a guess, find a game which actually has non zero values for (size_t i = 0; i < info.registers.size(); i++) { crash_report += fmt::format(" X[{:02d}]: {:016x}\n", i, info.registers[i]); } - crash_report += fmt::format(" Unknown 0: {:016x}\n", info.unk0); - crash_report += fmt::format(" Unknown 1: {:016x}\n", info.unk1); - crash_report += fmt::format(" Unknown 2: {:016x}\n", info.unk2); - crash_report += fmt::format(" Unknown 3: {:016x}\n", info.unk3); - crash_report += fmt::format(" Unknown 4: {:016x}\n", info.unk4); - crash_report += fmt::format(" Unknown 5: {:016x}\n", info.unk5); - crash_report += fmt::format(" Unknown 6: {:016x}\n", info.unk6); + crash_report += fmt::format(" SP: {:016x}\n", info.sp); + crash_report += fmt::format(" PC: {:016x}\n", info.pc); + crash_report += fmt::format(" PSTATE: {:016x}\n", info.pstate); + crash_report += fmt::format(" AFSR0: {:016x}\n", info.afsr0); + crash_report += fmt::format(" AFSR1: {:016x}\n", info.afsr1); + crash_report += fmt::format(" ESR: {:016x}\n", info.esr); + crash_report += fmt::format(" FAR: {:016x}\n", info.far); crash_report += "\nBacktrace:\n"; for (size_t i = 0; i < info.backtrace_size; i++) { crash_report += fmt::format(" Backtrace[{:02d}]: {:016x}\n", i, info.backtrace[i]); } - crash_report += fmt::format("\nUnknown 7: 0x{:016x}\n", info.unk7); - crash_report += fmt::format("Unknown 8: 0x{:016x}\n", info.unk8); - crash_report += fmt::format("Unknown 9: 0x{:016x}\n", info.unk9); + + crash_report += fmt::format("Architecture: {}\n", info.ArchAsString()); crash_report += fmt::format("Unknown 10: 0x{:016x}\n", info.unk10); } @@ -125,13 +138,13 @@ static void ThrowFatalError(ResultCode error_code, FatalType fatal_type, const F case FatalType::ErrorReport: GenerateErrorReport(error_code, info); break; - }; + } } void Module::Interface::ThrowFatal(Kernel::HLERequestContext& ctx) { LOG_ERROR(Service_Fatal, "called"); IPC::RequestParser rp{ctx}; - auto error_code = rp.Pop<ResultCode>(); + const auto error_code = rp.Pop<ResultCode>(); ThrowFatalError(error_code, FatalType::ErrorScreen, {}); IPC::ResponseBuilder rb{ctx, 2}; @@ -141,8 +154,8 @@ void Module::Interface::ThrowFatal(Kernel::HLERequestContext& ctx) { void Module::Interface::ThrowFatalWithPolicy(Kernel::HLERequestContext& ctx) { LOG_ERROR(Service_Fatal, "called"); IPC::RequestParser rp(ctx); - auto error_code = rp.Pop<ResultCode>(); - auto fatal_type = rp.PopEnum<FatalType>(); + const auto error_code = rp.Pop<ResultCode>(); + const auto fatal_type = rp.PopEnum<FatalType>(); ThrowFatalError(error_code, fatal_type, {}); // No info is passed with ThrowFatalWithPolicy IPC::ResponseBuilder rb{ctx, 2}; @@ -152,9 +165,9 @@ void Module::Interface::ThrowFatalWithPolicy(Kernel::HLERequestContext& ctx) { void Module::Interface::ThrowFatalWithCpuContext(Kernel::HLERequestContext& ctx) { LOG_ERROR(Service_Fatal, "called"); IPC::RequestParser rp(ctx); - auto error_code = rp.Pop<ResultCode>(); - auto fatal_type = rp.PopEnum<FatalType>(); - auto fatal_info = ctx.ReadBuffer(); + const auto error_code = rp.Pop<ResultCode>(); + const auto fatal_type = rp.PopEnum<FatalType>(); + const auto fatal_info = ctx.ReadBuffer(); FatalInfo info{}; ASSERT_MSG(fatal_info.size() == sizeof(FatalInfo), "Invalid fatal info buffer size!"); diff --git a/src/core/hle/service/hid/hid.cpp b/src/core/hle/service/hid/hid.cpp index 8a6de83a2..63b55758b 100644 --- a/src/core/hle/service/hid/hid.cpp +++ b/src/core/hle/service/hid/hid.cpp @@ -36,9 +36,9 @@ namespace Service::HID { // Updating period for each HID device. // TODO(ogniK): Find actual polling rate of hid -constexpr u64 pad_update_ticks = Core::Timing::BASE_CLOCK_RATE / 66; -constexpr u64 accelerometer_update_ticks = Core::Timing::BASE_CLOCK_RATE / 100; -constexpr u64 gyroscope_update_ticks = Core::Timing::BASE_CLOCK_RATE / 100; +constexpr s64 pad_update_ticks = static_cast<s64>(Core::Timing::BASE_CLOCK_RATE / 66); +constexpr s64 accelerometer_update_ticks = static_cast<s64>(Core::Timing::BASE_CLOCK_RATE / 100); +constexpr s64 gyroscope_update_ticks = static_cast<s64>(Core::Timing::BASE_CLOCK_RATE / 100); constexpr std::size_t SHARED_MEMORY_SIZE = 0x40000; IAppletResource::IAppletResource() : ServiceFramework("IAppletResource") { @@ -75,7 +75,7 @@ IAppletResource::IAppletResource() : ServiceFramework("IAppletResource") { // Register update callbacks auto& core_timing = Core::System::GetInstance().CoreTiming(); pad_update_event = - core_timing.RegisterEvent("HID::UpdatePadCallback", [this](u64 userdata, int cycles_late) { + core_timing.RegisterEvent("HID::UpdatePadCallback", [this](u64 userdata, s64 cycles_late) { UpdateControllers(userdata, cycles_late); }); @@ -106,7 +106,7 @@ void IAppletResource::GetSharedMemoryHandle(Kernel::HLERequestContext& ctx) { rb.PushCopyObjects(shared_mem); } -void IAppletResource::UpdateControllers(u64 userdata, int cycles_late) { +void IAppletResource::UpdateControllers(u64 userdata, s64 cycles_late) { auto& core_timing = Core::System::GetInstance().CoreTiming(); const bool should_reload = Settings::values.is_device_reload_pending.exchange(false); diff --git a/src/core/hle/service/hid/hid.h b/src/core/hle/service/hid/hid.h index 498602de5..d3660cad2 100644 --- a/src/core/hle/service/hid/hid.h +++ b/src/core/hle/service/hid/hid.h @@ -65,7 +65,7 @@ private: } void GetSharedMemoryHandle(Kernel::HLERequestContext& ctx); - void UpdateControllers(u64 userdata, int cycles_late); + void UpdateControllers(u64 userdata, s64 cycles_late); Kernel::SharedPtr<Kernel::SharedMemory> shared_mem; diff --git a/src/core/hle/service/nfc/nfc.cpp b/src/core/hle/service/nfc/nfc.cpp index 5c62d42ba..ca88bf97f 100644 --- a/src/core/hle/service/nfc/nfc.cpp +++ b/src/core/hle/service/nfc/nfc.cpp @@ -150,7 +150,7 @@ private: IPC::ResponseBuilder rb{ctx, 3}; rb.Push(RESULT_SUCCESS); - rb.PushRaw<u8>(Settings::values.enable_nfc); + rb.PushRaw<u8>(true); } void GetStateOld(Kernel::HLERequestContext& ctx) { diff --git a/src/core/hle/service/nfp/nfp.cpp b/src/core/hle/service/nfp/nfp.cpp index 1c4482e47..c6babdd4d 100644 --- a/src/core/hle/service/nfp/nfp.cpp +++ b/src/core/hle/service/nfp/nfp.cpp @@ -335,7 +335,7 @@ void Module::Interface::CreateUserInterface(Kernel::HLERequestContext& ctx) { } bool Module::Interface::LoadAmiibo(const std::vector<u8>& buffer) { - std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock); + std::lock_guard lock{HLE::g_hle_lock}; if (buffer.size() < sizeof(AmiiboFile)) { return false; } diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp index b031ebc66..af62d33d2 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp @@ -89,7 +89,7 @@ u32 nvhost_as_gpu::Remap(const std::vector<u8>& input, std::vector<u8>& output) for (const auto& entry : entries) { LOG_WARNING(Service_NVDRV, "remap entry, offset=0x{:X} handle=0x{:X} pages=0x{:X}", entry.offset, entry.nvmap_handle, entry.pages); - Tegra::GPUVAddr offset = static_cast<Tegra::GPUVAddr>(entry.offset) << 0x10; + GPUVAddr offset = static_cast<GPUVAddr>(entry.offset) << 0x10; auto object = nvmap_dev->GetObject(entry.nvmap_handle); if (!object) { LOG_CRITICAL(Service_NVDRV, "nvmap {} is an invalid handle!", entry.nvmap_handle); @@ -102,7 +102,7 @@ u32 nvhost_as_gpu::Remap(const std::vector<u8>& input, std::vector<u8>& output) u64 size = static_cast<u64>(entry.pages) << 0x10; ASSERT(size <= object->size); - Tegra::GPUVAddr returned = gpu.MemoryManager().MapBufferEx(object->addr, offset, size); + GPUVAddr returned = gpu.MemoryManager().MapBufferEx(object->addr, offset, size); ASSERT(returned == offset); } std::memcpy(output.data(), entries.data(), output.size()); @@ -173,16 +173,8 @@ u32 nvhost_as_gpu::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& ou return 0; } - auto& system_instance = Core::System::GetInstance(); - - // Remove this memory region from the rasterizer cache. - auto& gpu = system_instance.GPU(); - auto cpu_addr = gpu.MemoryManager().GpuToCpuAddress(params.offset); - ASSERT(cpu_addr); - gpu.FlushAndInvalidateRegion(ToCacheAddr(Memory::GetPointer(*cpu_addr)), itr->second.size); - - params.offset = gpu.MemoryManager().UnmapBuffer(params.offset, itr->second.size); - + params.offset = Core::System::GetInstance().GPU().MemoryManager().UnmapBuffer(params.offset, + itr->second.size); buffer_mappings.erase(itr->second.offset); std::memcpy(output.data(), ¶ms, output.size()); diff --git a/src/core/hle/service/nvflinger/nvflinger.cpp b/src/core/hle/service/nvflinger/nvflinger.cpp index fc496b654..c7f5bbf28 100644 --- a/src/core/hle/service/nvflinger/nvflinger.cpp +++ b/src/core/hle/service/nvflinger/nvflinger.cpp @@ -26,7 +26,7 @@ namespace Service::NVFlinger { constexpr std::size_t SCREEN_REFRESH_RATE = 60; -constexpr u64 frame_ticks = static_cast<u64>(Core::Timing::BASE_CLOCK_RATE / SCREEN_REFRESH_RATE); +constexpr s64 frame_ticks = static_cast<s64>(Core::Timing::BASE_CLOCK_RATE / SCREEN_REFRESH_RATE); NVFlinger::NVFlinger(Core::Timing::CoreTiming& core_timing) : core_timing{core_timing} { displays.emplace_back(0, "Default"); @@ -37,7 +37,7 @@ NVFlinger::NVFlinger(Core::Timing::CoreTiming& core_timing) : core_timing{core_t // Schedule the screen composition events composition_event = - core_timing.RegisterEvent("ScreenComposition", [this](u64 userdata, int cycles_late) { + core_timing.RegisterEvent("ScreenComposition", [this](u64 userdata, s64 cycles_late) { Compose(); this->core_timing.ScheduleEvent(frame_ticks - cycles_late, composition_event); }); diff --git a/src/core/hle/service/set/set_sys.cpp b/src/core/hle/service/set/set_sys.cpp index c9b4da5b0..ecee554bf 100644 --- a/src/core/hle/service/set/set_sys.cpp +++ b/src/core/hle/service/set/set_sys.cpp @@ -2,13 +2,88 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include "common/assert.h" #include "common/logging/log.h" +#include "core/file_sys/errors.h" +#include "core/file_sys/system_archive/system_version.h" #include "core/hle/ipc_helpers.h" #include "core/hle/kernel/client_port.h" +#include "core/hle/service/filesystem/filesystem.h" #include "core/hle/service/set/set_sys.h" namespace Service::Set { +namespace { +constexpr u64 SYSTEM_VERSION_FILE_MINOR_REVISION_OFFSET = 0x05; + +enum class GetFirmwareVersionType { + Version1, + Version2, +}; + +void GetFirmwareVersionImpl(Kernel::HLERequestContext& ctx, GetFirmwareVersionType type) { + LOG_WARNING(Service_SET, "called - Using hardcoded firmware version '{}'", + FileSys::SystemArchive::GetLongDisplayVersion()); + + ASSERT_MSG(ctx.GetWriteBufferSize() == 0x100, + "FirmwareVersion output buffer must be 0x100 bytes in size!"); + + // Instead of using the normal procedure of checking for the real system archive and if it + // doesn't exist, synthesizing one, I feel that that would lead to strange bugs because a + // used is using a really old or really new SystemVersion title. The synthesized one ensures + // consistence (currently reports as 5.1.0-0.0) + const auto archive = FileSys::SystemArchive::SystemVersion(); + + const auto early_exit_failure = [&ctx](const std::string& desc, ResultCode code) { + LOG_ERROR(Service_SET, "General failure while attempting to resolve firmware version ({}).", + desc.c_str()); + IPC::ResponseBuilder rb{ctx, 2}; + rb.Push(code); + }; + + if (archive == nullptr) { + early_exit_failure("The system version archive couldn't be synthesized.", + FileSys::ERROR_FAILED_MOUNT_ARCHIVE); + return; + } + + const auto ver_file = archive->GetFile("file"); + if (ver_file == nullptr) { + early_exit_failure("The system version archive didn't contain the file 'file'.", + FileSys::ERROR_INVALID_ARGUMENT); + return; + } + + auto data = ver_file->ReadAllBytes(); + if (data.size() != 0x100) { + early_exit_failure("The system version file 'file' was not the correct size.", + FileSys::ERROR_OUT_OF_BOUNDS); + return; + } + + // If the command is GetFirmwareVersion (as opposed to GetFirmwareVersion2), hardware will + // zero out the REVISION_MINOR field. + if (type == GetFirmwareVersionType::Version1) { + data[SYSTEM_VERSION_FILE_MINOR_REVISION_OFFSET] = 0; + } + + ctx.WriteBuffer(data); + + IPC::ResponseBuilder rb{ctx, 2}; + rb.Push(RESULT_SUCCESS); +} +} // Anonymous namespace + +void SET_SYS::GetFirmwareVersion(Kernel::HLERequestContext& ctx) { + LOG_DEBUG(Service_SET, "called"); + GetFirmwareVersionImpl(ctx, GetFirmwareVersionType::Version1); +} + +void SET_SYS::GetFirmwareVersion2(Kernel::HLERequestContext& ctx) { + LOG_DEBUG(Service_SET, "called"); + GetFirmwareVersionImpl(ctx, GetFirmwareVersionType::Version2); +} + void SET_SYS::GetColorSetId(Kernel::HLERequestContext& ctx) { LOG_DEBUG(Service_SET, "called"); @@ -33,8 +108,8 @@ SET_SYS::SET_SYS() : ServiceFramework("set:sys") { {0, nullptr, "SetLanguageCode"}, {1, nullptr, "SetNetworkSettings"}, {2, nullptr, "GetNetworkSettings"}, - {3, nullptr, "GetFirmwareVersion"}, - {4, nullptr, "GetFirmwareVersion2"}, + {3, &SET_SYS::GetFirmwareVersion, "GetFirmwareVersion"}, + {4, &SET_SYS::GetFirmwareVersion2, "GetFirmwareVersion2"}, {5, nullptr, "GetFirmwareVersionDigest"}, {7, nullptr, "GetLockScreenFlag"}, {8, nullptr, "SetLockScreenFlag"}, diff --git a/src/core/hle/service/set/set_sys.h b/src/core/hle/service/set/set_sys.h index f602f3c77..13ee2cf46 100644 --- a/src/core/hle/service/set/set_sys.h +++ b/src/core/hle/service/set/set_sys.h @@ -20,6 +20,8 @@ private: BasicBlack = 1, }; + void GetFirmwareVersion(Kernel::HLERequestContext& ctx); + void GetFirmwareVersion2(Kernel::HLERequestContext& ctx); void GetColorSetId(Kernel::HLERequestContext& ctx); void SetColorSetId(Kernel::HLERequestContext& ctx); diff --git a/src/core/loader/nso.cpp b/src/core/loader/nso.cpp index 5caceb267..babc7e646 100644 --- a/src/core/loader/nso.cpp +++ b/src/core/loader/nso.cpp @@ -21,36 +21,8 @@ #include "core/settings.h" namespace Loader { - -struct NsoSegmentHeader { - u32_le offset; - u32_le location; - u32_le size; - union { - u32_le alignment; - u32_le bss_size; - }; -}; -static_assert(sizeof(NsoSegmentHeader) == 0x10, "NsoSegmentHeader has incorrect size."); - -struct NsoHeader { - u32_le magic; - u32_le version; - INSERT_PADDING_WORDS(1); - u8 flags; - std::array<NsoSegmentHeader, 3> segments; // Text, RoData, Data (in that order) - std::array<u8, 0x20> build_id; - std::array<u32_le, 3> segments_compressed_size; - - bool IsSegmentCompressed(size_t segment_num) const { - ASSERT_MSG(segment_num < 3, "Invalid segment {}", segment_num); - return ((flags >> segment_num) & 1); - } -}; -static_assert(sizeof(NsoHeader) == 0x6c, "NsoHeader has incorrect size."); -static_assert(std::is_trivially_copyable_v<NsoHeader>, "NsoHeader isn't trivially copyable."); - -struct ModHeader { +namespace { +struct MODHeader { u32_le magic; u32_le dynamic_offset; u32_le bss_start_offset; @@ -59,25 +31,10 @@ struct ModHeader { u32_le eh_frame_hdr_end_offset; u32_le module_offset; // Offset to runtime-generated module object. typically equal to .bss base }; -static_assert(sizeof(ModHeader) == 0x1c, "ModHeader has incorrect size."); - -AppLoader_NSO::AppLoader_NSO(FileSys::VirtualFile file) : AppLoader(std::move(file)) {} - -FileType AppLoader_NSO::IdentifyType(const FileSys::VirtualFile& file) { - u32 magic = 0; - if (file->ReadObject(&magic) != sizeof(magic)) { - return FileType::Error; - } - - if (Common::MakeMagic('N', 'S', 'O', '0') != magic) { - return FileType::Error; - } - - return FileType::NSO; -} +static_assert(sizeof(MODHeader) == 0x1c, "MODHeader has incorrect size."); -static std::vector<u8> DecompressSegment(const std::vector<u8>& compressed_data, - const NsoSegmentHeader& header) { +std::vector<u8> DecompressSegment(const std::vector<u8>& compressed_data, + const NSOSegmentHeader& header) { std::vector<u8> uncompressed_data(header.size); const int bytes_uncompressed = LZ4_decompress_safe(reinterpret_cast<const char*>(compressed_data.data()), @@ -91,23 +48,47 @@ static std::vector<u8> DecompressSegment(const std::vector<u8>& compressed_data, return uncompressed_data; } -static constexpr u32 PageAlignSize(u32 size) { +constexpr u32 PageAlignSize(u32 size) { return (size + Memory::PAGE_MASK) & ~Memory::PAGE_MASK; } +} // Anonymous namespace + +bool NSOHeader::IsSegmentCompressed(size_t segment_num) const { + ASSERT_MSG(segment_num < 3, "Invalid segment {}", segment_num); + return ((flags >> segment_num) & 1) != 0; +} + +AppLoader_NSO::AppLoader_NSO(FileSys::VirtualFile file) : AppLoader(std::move(file)) {} + +FileType AppLoader_NSO::IdentifyType(const FileSys::VirtualFile& file) { + u32 magic = 0; + if (file->ReadObject(&magic) != sizeof(magic)) { + return FileType::Error; + } + + if (Common::MakeMagic('N', 'S', 'O', '0') != magic) { + return FileType::Error; + } + + return FileType::NSO; +} std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::Process& process, const FileSys::VfsFile& file, VAddr load_base, bool should_pass_arguments, std::optional<FileSys::PatchManager> pm) { - if (file.GetSize() < sizeof(NsoHeader)) + if (file.GetSize() < sizeof(NSOHeader)) { return {}; + } - NsoHeader nso_header{}; - if (sizeof(NsoHeader) != file.ReadObject(&nso_header)) + NSOHeader nso_header{}; + if (sizeof(NSOHeader) != file.ReadObject(&nso_header)) { return {}; + } - if (nso_header.magic != Common::MakeMagic('N', 'S', 'O', '0')) + if (nso_header.magic != Common::MakeMagic('N', 'S', 'O', '0')) { return {}; + } // Build program image Kernel::CodeSet codeset; @@ -143,10 +124,10 @@ std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::Process& process, std::memcpy(&module_offset, program_image.data() + 4, sizeof(u32)); // Read MOD header - ModHeader mod_header{}; + MODHeader mod_header{}; // Default .bss to size in segment header if MOD0 section doesn't exist u32 bss_size{PageAlignSize(nso_header.segments[2].bss_size)}; - std::memcpy(&mod_header, program_image.data() + module_offset, sizeof(ModHeader)); + std::memcpy(&mod_header, program_image.data() + module_offset, sizeof(MODHeader)); const bool has_mod_header{mod_header.magic == Common::MakeMagic('M', 'O', 'D', '0')}; if (has_mod_header) { // Resize program image to include .bss section and page align each section @@ -158,22 +139,24 @@ std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::Process& process, // Apply patches if necessary if (pm && (pm->HasNSOPatch(nso_header.build_id) || Settings::values.dump_nso)) { - std::vector<u8> pi_header(program_image.size() + 0x100); - std::memcpy(pi_header.data(), &nso_header, sizeof(NsoHeader)); - std::memcpy(pi_header.data() + 0x100, program_image.data(), program_image.size()); + std::vector<u8> pi_header(sizeof(NSOHeader) + program_image.size()); + pi_header.insert(pi_header.begin(), reinterpret_cast<u8*>(&nso_header), + reinterpret_cast<u8*>(&nso_header) + sizeof(NSOHeader)); + pi_header.insert(pi_header.begin() + sizeof(NSOHeader), program_image.begin(), + program_image.end()); pi_header = pm->PatchNSO(pi_header); - std::memcpy(program_image.data(), pi_header.data() + 0x100, program_image.size()); + std::copy(pi_header.begin() + sizeof(NSOHeader), pi_header.end(), program_image.begin()); } // Apply cheats if they exist and the program has a valid title ID if (pm) { - const auto cheats = pm->CreateCheatList(nso_header.build_id); + auto& system = Core::System::GetInstance(); + const auto cheats = pm->CreateCheatList(system, nso_header.build_id); if (!cheats.empty()) { - Core::System::GetInstance().RegisterCheatList( - cheats, Common::HexArrayToString(nso_header.build_id), load_base, - load_base + program_image.size()); + system.RegisterCheatList(cheats, Common::HexArrayToString(nso_header.build_id), + load_base, load_base + program_image.size()); } } diff --git a/src/core/loader/nso.h b/src/core/loader/nso.h index 167c8a694..4674c3724 100644 --- a/src/core/loader/nso.h +++ b/src/core/loader/nso.h @@ -4,7 +4,9 @@ #pragma once +#include <array> #include <optional> +#include <type_traits> #include "common/common_types.h" #include "common/swap.h" #include "core/file_sys/patch_manager.h" @@ -16,6 +18,43 @@ class Process; namespace Loader { +struct NSOSegmentHeader { + u32_le offset; + u32_le location; + u32_le size; + union { + u32_le alignment; + u32_le bss_size; + }; +}; +static_assert(sizeof(NSOSegmentHeader) == 0x10, "NsoSegmentHeader has incorrect size."); + +struct NSOHeader { + using SHA256Hash = std::array<u8, 0x20>; + + struct RODataRelativeExtent { + u32_le data_offset; + u32_le size; + }; + + u32_le magic; + u32_le version; + u32 reserved; + u32_le flags; + std::array<NSOSegmentHeader, 3> segments; // Text, RoData, Data (in that order) + std::array<u8, 0x20> build_id; + std::array<u32_le, 3> segments_compressed_size; + std::array<u8, 0x1C> padding; + RODataRelativeExtent api_info_extent; + RODataRelativeExtent dynstr_extent; + RODataRelativeExtent dynsyn_extent; + std::array<SHA256Hash, 3> segment_hashes; + + bool IsSegmentCompressed(size_t segment_num) const; +}; +static_assert(sizeof(NSOHeader) == 0x100, "NSOHeader has incorrect size."); +static_assert(std::is_trivially_copyable_v<NSOHeader>, "NSOHeader must be trivially copyable."); + constexpr u64 NSO_ARGUMENT_DATA_ALLOCATION_SIZE = 0x9000; struct NSOArgumentHeader { diff --git a/src/core/memory.cpp b/src/core/memory.cpp index 365ac82b4..332c1037c 100644 --- a/src/core/memory.cpp +++ b/src/core/memory.cpp @@ -48,7 +48,7 @@ static void MapPages(Common::PageTable& page_table, VAddr base, u64 size, u8* me (base + size) * PAGE_SIZE); // During boot, current_page_table might not be set yet, in which case we need not flush - if (current_page_table) { + if (Core::System::GetInstance().IsPoweredOn()) { Core::System::GetInstance().GPU().FlushAndInvalidateRegion(base << PAGE_BITS, size * PAGE_SIZE); } diff --git a/src/core/perf_stats.cpp b/src/core/perf_stats.cpp index c716a462b..4afd6c8a3 100644 --- a/src/core/perf_stats.cpp +++ b/src/core/perf_stats.cpp @@ -18,13 +18,13 @@ using std::chrono::microseconds; namespace Core { void PerfStats::BeginSystemFrame() { - std::lock_guard<std::mutex> lock(object_mutex); + std::lock_guard lock{object_mutex}; frame_begin = Clock::now(); } void PerfStats::EndSystemFrame() { - std::lock_guard<std::mutex> lock(object_mutex); + std::lock_guard lock{object_mutex}; auto frame_end = Clock::now(); accumulated_frametime += frame_end - frame_begin; @@ -35,13 +35,13 @@ void PerfStats::EndSystemFrame() { } void PerfStats::EndGameFrame() { - std::lock_guard<std::mutex> lock(object_mutex); + std::lock_guard lock{object_mutex}; game_frames += 1; } PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us) { - std::lock_guard<std::mutex> lock(object_mutex); + std::lock_guard lock{object_mutex}; const auto now = Clock::now(); // Walltime elapsed since stats were reset @@ -67,7 +67,7 @@ PerfStatsResults PerfStats::GetAndResetStats(microseconds current_system_time_us } double PerfStats::GetLastFrameTimeScale() { - std::lock_guard<std::mutex> lock(object_mutex); + std::lock_guard lock{object_mutex}; constexpr double FRAME_LENGTH = 1.0 / 60; return duration_cast<DoubleSecs>(previous_frame_length).count() / FRAME_LENGTH; diff --git a/src/core/settings.cpp b/src/core/settings.cpp index 6dd3139cc..6d32ebea3 100644 --- a/src/core/settings.cpp +++ b/src/core/settings.cpp @@ -82,7 +82,6 @@ void LogSetting(const std::string& name, const T& value) { void LogSettings() { LOG_INFO(Config, "yuzu Configuration:"); LogSetting("System_UseDockedMode", Settings::values.use_docked_mode); - LogSetting("System_EnableNfc", Settings::values.enable_nfc); LogSetting("System_RngSeed", Settings::values.rng_seed.value_or(0)); LogSetting("System_CurrentUser", Settings::values.current_user); LogSetting("System_LanguageIndex", Settings::values.language_index); diff --git a/src/core/settings.h b/src/core/settings.h index cdfb2f742..d543eb32f 100644 --- a/src/core/settings.h +++ b/src/core/settings.h @@ -349,7 +349,6 @@ struct TouchscreenInput { struct Values { // System bool use_docked_mode; - bool enable_nfc; std::optional<u32> rng_seed; // Measured in seconds since epoch std::optional<std::chrono::seconds> custom_rtc; diff --git a/src/input_common/keyboard.cpp b/src/input_common/keyboard.cpp index 525fe6abc..078374be5 100644 --- a/src/input_common/keyboard.cpp +++ b/src/input_common/keyboard.cpp @@ -36,18 +36,18 @@ struct KeyButtonPair { class KeyButtonList { public: void AddKeyButton(int key_code, KeyButton* key_button) { - std::lock_guard<std::mutex> guard(mutex); + std::lock_guard guard{mutex}; list.push_back(KeyButtonPair{key_code, key_button}); } void RemoveKeyButton(const KeyButton* key_button) { - std::lock_guard<std::mutex> guard(mutex); + std::lock_guard guard{mutex}; list.remove_if( [key_button](const KeyButtonPair& pair) { return pair.key_button == key_button; }); } void ChangeKeyStatus(int key_code, bool pressed) { - std::lock_guard<std::mutex> guard(mutex); + std::lock_guard guard{mutex}; for (const KeyButtonPair& pair : list) { if (pair.key_code == key_code) pair.key_button->status.store(pressed); @@ -55,7 +55,7 @@ public: } void ChangeAllKeyStatus(bool pressed) { - std::lock_guard<std::mutex> guard(mutex); + std::lock_guard guard{mutex}; for (const KeyButtonPair& pair : list) { pair.key_button->status.store(pressed); } diff --git a/src/input_common/motion_emu.cpp b/src/input_common/motion_emu.cpp index 6d96d4019..868251628 100644 --- a/src/input_common/motion_emu.cpp +++ b/src/input_common/motion_emu.cpp @@ -39,7 +39,7 @@ public: void Tilt(int x, int y) { auto mouse_move = Common::MakeVec(x, y) - mouse_origin; if (is_tilting) { - std::lock_guard<std::mutex> guard(tilt_mutex); + std::lock_guard guard{tilt_mutex}; if (mouse_move.x == 0 && mouse_move.y == 0) { tilt_angle = 0; } else { @@ -51,13 +51,13 @@ public: } void EndTilt() { - std::lock_guard<std::mutex> guard(tilt_mutex); + std::lock_guard guard{tilt_mutex}; tilt_angle = 0; is_tilting = false; } std::tuple<Common::Vec3<float>, Common::Vec3<float>> GetStatus() { - std::lock_guard<std::mutex> guard(status_mutex); + std::lock_guard guard{status_mutex}; return status; } @@ -93,7 +93,7 @@ private: old_q = q; { - std::lock_guard<std::mutex> guard(tilt_mutex); + std::lock_guard guard{tilt_mutex}; // Find the quaternion describing current 3DS tilting q = Common::MakeQuaternion( @@ -115,7 +115,7 @@ private: // Update the sensor state { - std::lock_guard<std::mutex> guard(status_mutex); + std::lock_guard guard{status_mutex}; status = std::make_tuple(gravity, angular_rate); } } diff --git a/src/input_common/sdl/sdl_impl.cpp b/src/input_common/sdl/sdl_impl.cpp index b132d77f5..5949ecbae 100644 --- a/src/input_common/sdl/sdl_impl.cpp +++ b/src/input_common/sdl/sdl_impl.cpp @@ -55,22 +55,22 @@ public: : guid{std::move(guid_)}, port{port_}, sdl_joystick{joystick, deleter} {} void SetButton(int button, bool value) { - std::lock_guard<std::mutex> lock(mutex); + std::lock_guard lock{mutex}; state.buttons[button] = value; } bool GetButton(int button) const { - std::lock_guard<std::mutex> lock(mutex); + std::lock_guard lock{mutex}; return state.buttons.at(button); } void SetAxis(int axis, Sint16 value) { - std::lock_guard<std::mutex> lock(mutex); + std::lock_guard lock{mutex}; state.axes[axis] = value; } float GetAxis(int axis) const { - std::lock_guard<std::mutex> lock(mutex); + std::lock_guard lock{mutex}; return state.axes.at(axis) / 32767.0f; } @@ -92,12 +92,12 @@ public: } void SetHat(int hat, Uint8 direction) { - std::lock_guard<std::mutex> lock(mutex); + std::lock_guard lock{mutex}; state.hats[hat] = direction; } bool GetHatDirection(int hat, Uint8 direction) const { - std::lock_guard<std::mutex> lock(mutex); + std::lock_guard lock{mutex}; return (state.hats.at(hat) & direction) != 0; } /** @@ -140,7 +140,7 @@ private: * Get the nth joystick with the corresponding GUID */ std::shared_ptr<SDLJoystick> SDLState::GetSDLJoystickByGUID(const std::string& guid, int port) { - std::lock_guard<std::mutex> lock(joystick_map_mutex); + std::lock_guard lock{joystick_map_mutex}; const auto it = joystick_map.find(guid); if (it != joystick_map.end()) { while (it->second.size() <= port) { @@ -161,7 +161,8 @@ std::shared_ptr<SDLJoystick> SDLState::GetSDLJoystickByGUID(const std::string& g std::shared_ptr<SDLJoystick> SDLState::GetSDLJoystickBySDLID(SDL_JoystickID sdl_id) { auto sdl_joystick = SDL_JoystickFromInstanceID(sdl_id); const std::string guid = GetGUID(sdl_joystick); - std::lock_guard<std::mutex> lock(joystick_map_mutex); + + std::lock_guard lock{joystick_map_mutex}; auto map_it = joystick_map.find(guid); if (map_it != joystick_map.end()) { auto vec_it = std::find_if(map_it->second.begin(), map_it->second.end(), @@ -198,8 +199,9 @@ void SDLState::InitJoystick(int joystick_index) { LOG_ERROR(Input, "failed to open joystick {}", joystick_index); return; } - std::string guid = GetGUID(sdl_joystick); - std::lock_guard<std::mutex> lock(joystick_map_mutex); + const std::string guid = GetGUID(sdl_joystick); + + std::lock_guard lock{joystick_map_mutex}; if (joystick_map.find(guid) == joystick_map.end()) { auto joystick = std::make_shared<SDLJoystick>(guid, 0, sdl_joystick); joystick_map[guid].emplace_back(std::move(joystick)); @@ -221,7 +223,7 @@ void SDLState::CloseJoystick(SDL_Joystick* sdl_joystick) { std::string guid = GetGUID(sdl_joystick); std::shared_ptr<SDLJoystick> joystick; { - std::lock_guard<std::mutex> lock(joystick_map_mutex); + std::lock_guard lock{joystick_map_mutex}; // This call to guid is safe since the joystick is guaranteed to be in the map auto& joystick_guid_list = joystick_map[guid]; const auto joystick_it = @@ -274,7 +276,7 @@ void SDLState::HandleGameControllerEvent(const SDL_Event& event) { } void SDLState::CloseJoysticks() { - std::lock_guard<std::mutex> lock(joystick_map_mutex); + std::lock_guard lock{joystick_map_mutex}; joystick_map.clear(); } diff --git a/src/tests/CMakeLists.txt b/src/tests/CMakeLists.txt index d0284bdf4..c7038b217 100644 --- a/src/tests/CMakeLists.txt +++ b/src/tests/CMakeLists.txt @@ -1,5 +1,7 @@ add_executable(tests common/bit_field.cpp + common/bit_utils.cpp + common/multi_level_queue.cpp common/param_package.cpp common/ring_buffer.cpp core/arm/arm_test_common.cpp diff --git a/src/tests/common/bit_utils.cpp b/src/tests/common/bit_utils.cpp new file mode 100644 index 000000000..479b5995a --- /dev/null +++ b/src/tests/common/bit_utils.cpp @@ -0,0 +1,23 @@ +// Copyright 2017 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <catch2/catch.hpp> +#include <math.h> +#include "common/bit_util.h" + +namespace Common { + +TEST_CASE("BitUtils::CountTrailingZeroes", "[common]") { + REQUIRE(Common::CountTrailingZeroes32(0) == 32); + REQUIRE(Common::CountTrailingZeroes64(0) == 64); + REQUIRE(Common::CountTrailingZeroes32(9) == 0); + REQUIRE(Common::CountTrailingZeroes32(8) == 3); + REQUIRE(Common::CountTrailingZeroes32(0x801000) == 12); + REQUIRE(Common::CountTrailingZeroes64(9) == 0); + REQUIRE(Common::CountTrailingZeroes64(8) == 3); + REQUIRE(Common::CountTrailingZeroes64(0x801000) == 12); + REQUIRE(Common::CountTrailingZeroes64(0x801000000000UL) == 36); +} + +} // namespace Common diff --git a/src/tests/common/multi_level_queue.cpp b/src/tests/common/multi_level_queue.cpp new file mode 100644 index 000000000..cca7ec7da --- /dev/null +++ b/src/tests/common/multi_level_queue.cpp @@ -0,0 +1,55 @@ +// Copyright 2019 Yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <catch2/catch.hpp> +#include <math.h> +#include "common/common_types.h" +#include "common/multi_level_queue.h" + +namespace Common { + +TEST_CASE("MultiLevelQueue", "[common]") { + std::array<f32, 8> values = {0.0, 5.0, 1.0, 9.0, 8.0, 2.0, 6.0, 7.0}; + Common::MultiLevelQueue<f32, 64> mlq; + REQUIRE(mlq.empty()); + mlq.add(values[2], 2); + mlq.add(values[7], 7); + mlq.add(values[3], 3); + mlq.add(values[4], 4); + mlq.add(values[0], 0); + mlq.add(values[5], 5); + mlq.add(values[6], 6); + mlq.add(values[1], 1); + u32 index = 0; + bool all_set = true; + for (auto& f : mlq) { + all_set &= (f == values[index]); + index++; + } + REQUIRE(all_set); + REQUIRE(!mlq.empty()); + f32 v = 8.0; + mlq.add(v, 2); + v = -7.0; + mlq.add(v, 2, false); + REQUIRE(mlq.front(2) == -7.0); + mlq.yield(2); + REQUIRE(mlq.front(2) == values[2]); + REQUIRE(mlq.back(2) == -7.0); + REQUIRE(mlq.empty(8)); + v = 10.0; + mlq.add(v, 8); + mlq.adjust(v, 8, 9); + REQUIRE(mlq.front(9) == v); + REQUIRE(mlq.empty(8)); + REQUIRE(!mlq.empty(9)); + mlq.adjust(values[0], 0, 9); + REQUIRE(mlq.highest_priority_set() == 1); + REQUIRE(mlq.lowest_priority_set() == 9); + mlq.remove(values[1], 1); + REQUIRE(mlq.highest_priority_set() == 2); + REQUIRE(mlq.empty(1)); +} + +} // namespace Common diff --git a/src/video_core/debug_utils/debug_utils.cpp b/src/video_core/debug_utils/debug_utils.cpp index 5ffb492ea..f0ef67535 100644 --- a/src/video_core/debug_utils/debug_utils.cpp +++ b/src/video_core/debug_utils/debug_utils.cpp @@ -10,7 +10,7 @@ namespace Tegra { void DebugContext::DoOnEvent(Event event, void* data) { { - std::unique_lock<std::mutex> lock(breakpoint_mutex); + std::unique_lock lock{breakpoint_mutex}; // TODO(Subv): Commit the rasterizer's caches so framebuffers, render targets, etc. will // show on debug widgets @@ -32,7 +32,7 @@ void DebugContext::DoOnEvent(Event event, void* data) { void DebugContext::Resume() { { - std::lock_guard<std::mutex> lock(breakpoint_mutex); + std::lock_guard lock{breakpoint_mutex}; // Tell all observers that we are about to resume for (auto& breakpoint_observer : breakpoint_observers) { diff --git a/src/video_core/debug_utils/debug_utils.h b/src/video_core/debug_utils/debug_utils.h index c235faf46..ac3a2eb01 100644 --- a/src/video_core/debug_utils/debug_utils.h +++ b/src/video_core/debug_utils/debug_utils.h @@ -40,7 +40,7 @@ public: /// Constructs the object such that it observes events of the given DebugContext. explicit BreakPointObserver(std::shared_ptr<DebugContext> debug_context) : context_weak(debug_context) { - std::unique_lock<std::mutex> lock(debug_context->breakpoint_mutex); + std::unique_lock lock{debug_context->breakpoint_mutex}; debug_context->breakpoint_observers.push_back(this); } @@ -48,7 +48,7 @@ public: auto context = context_weak.lock(); if (context) { { - std::unique_lock<std::mutex> lock(context->breakpoint_mutex); + std::unique_lock lock{context->breakpoint_mutex}; context->breakpoint_observers.remove(this); } diff --git a/src/video_core/dma_pusher.h b/src/video_core/dma_pusher.h index 27a36348c..6ab06518f 100644 --- a/src/video_core/dma_pusher.h +++ b/src/video_core/dma_pusher.h @@ -9,7 +9,6 @@ #include "common/bit_field.h" #include "common/common_types.h" -#include "video_core/memory_manager.h" namespace Tegra { diff --git a/src/video_core/engines/kepler_memory.cpp b/src/video_core/engines/kepler_memory.cpp index 0931b9626..e259bf46b 100644 --- a/src/video_core/engines/kepler_memory.cpp +++ b/src/video_core/engines/kepler_memory.cpp @@ -46,7 +46,7 @@ void KeplerMemory::ProcessData(u32 data) { // contain a dirty surface that will have to be written back to memory. const GPUVAddr address{regs.dest.Address() + state.write_offset * sizeof(u32)}; rasterizer.InvalidateRegion(ToCacheAddr(memory_manager.GetPointer(address)), sizeof(u32)); - memory_manager.Write32(address, data); + memory_manager.Write<u32>(address, data); system.GPU().Maxwell3D().dirty_flags.OnMemoryWrite(); diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp index c5d5be4ef..defcfbd3f 100644 --- a/src/video_core/engines/maxwell_3d.cpp +++ b/src/video_core/engines/maxwell_3d.cpp @@ -307,7 +307,7 @@ void Maxwell3D::ProcessQueryGet() { // Write the current query sequence to the sequence address. // TODO(Subv): Find out what happens if you use a long query type but mark it as a short // query. - memory_manager.Write32(sequence_address, sequence); + memory_manager.Write<u32>(sequence_address, sequence); } else { // Write the 128-bit result structure in long mode. Note: We emulate an infinitely fast // GPU, this command may actually take a while to complete in real hardware due to GPU @@ -395,7 +395,7 @@ void Maxwell3D::ProcessCBData(u32 value) { u8* ptr{memory_manager.GetPointer(address)}; rasterizer.InvalidateRegion(ToCacheAddr(ptr), sizeof(u32)); - memory_manager.Write32(address, value); + memory_manager.Write<u32>(address, value); dirty_flags.OnMemoryWrite(); @@ -447,7 +447,7 @@ std::vector<Texture::FullTextureInfo> Maxwell3D::GetStageTextures(Regs::ShaderSt for (GPUVAddr current_texture = tex_info_buffer.address + TextureInfoOffset; current_texture < tex_info_buffer_end; current_texture += sizeof(Texture::TextureHandle)) { - const Texture::TextureHandle tex_handle{memory_manager.Read32(current_texture)}; + const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(current_texture)}; Texture::FullTextureInfo tex_info{}; // TODO(Subv): Use the shader to determine which textures are actually accessed. @@ -482,7 +482,7 @@ Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage, ASSERT(tex_info_address < tex_info_buffer.address + tex_info_buffer.size); - const Texture::TextureHandle tex_handle{memory_manager.Read32(tex_info_address)}; + const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)}; Texture::FullTextureInfo tex_info{}; tex_info.index = static_cast<u32>(offset); diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp index a0ded4c25..5cca5c29a 100644 --- a/src/video_core/engines/maxwell_dma.cpp +++ b/src/video_core/engines/maxwell_dma.cpp @@ -88,6 +88,16 @@ void MaxwellDMA::HandleCopy() { auto source_ptr{memory_manager.GetPointer(source)}; auto dst_ptr{memory_manager.GetPointer(dest)}; + if (!source_ptr) { + LOG_ERROR(HW_GPU, "source_ptr is invalid"); + return; + } + + if (!dst_ptr) { + LOG_ERROR(HW_GPU, "dst_ptr is invalid"); + return; + } + const auto FlushAndInvalidate = [&](u32 src_size, u64 dst_size) { // TODO(Subv): For now, manually flush the regions until we implement GPU-accelerated // copying. diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 66c690494..30b29e14d 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -12,6 +12,7 @@ #include "video_core/engines/maxwell_3d.h" #include "video_core/engines/maxwell_dma.h" #include "video_core/gpu.h" +#include "video_core/memory_manager.h" #include "video_core/renderer_base.h" namespace Tegra { @@ -285,9 +286,10 @@ void GPU::ProcessSemaphoreTriggerMethod() { // TODO(Kmather73): Generate a real GPU timestamp and write it here instead of // CoreTiming block.timestamp = Core::System::GetInstance().CoreTiming().GetTicks(); - memory_manager->WriteBlock(regs.smaphore_address.SmaphoreAddress(), &block, sizeof(block)); + memory_manager->WriteBlock(regs.semaphore_address.SemaphoreAddress(), &block, + sizeof(block)); } else { - const u32 word{memory_manager->Read32(regs.smaphore_address.SmaphoreAddress())}; + const u32 word{memory_manager->Read<u32>(regs.semaphore_address.SemaphoreAddress())}; if ((op == GpuSemaphoreOperation::AcquireEqual && word == regs.semaphore_sequence) || (op == GpuSemaphoreOperation::AcquireGequal && static_cast<s32>(word - regs.semaphore_sequence) > 0) || @@ -314,11 +316,11 @@ void GPU::ProcessSemaphoreTriggerMethod() { } void GPU::ProcessSemaphoreRelease() { - memory_manager->Write32(regs.smaphore_address.SmaphoreAddress(), regs.semaphore_release); + memory_manager->Write<u32>(regs.semaphore_address.SemaphoreAddress(), regs.semaphore_release); } void GPU::ProcessSemaphoreAcquire() { - const u32 word = memory_manager->Read32(regs.smaphore_address.SmaphoreAddress()); + const u32 word = memory_manager->Read<u32>(regs.semaphore_address.SemaphoreAddress()); const auto value = regs.semaphore_acquire; if (word != value) { regs.acquire_active = true; diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h index a14b95c30..de30ea354 100644 --- a/src/video_core/gpu.h +++ b/src/video_core/gpu.h @@ -9,7 +9,6 @@ #include "common/common_types.h" #include "core/hle/service/nvflinger/buffer_queue.h" #include "video_core/dma_pusher.h" -#include "video_core/memory_manager.h" using CacheAddr = std::uintptr_t; inline CacheAddr ToCacheAddr(const void* host_ptr) { @@ -124,6 +123,8 @@ enum class EngineID { MAXWELL_DMA_COPY_A = 0xB0B5, }; +class MemoryManager; + class GPU { public: explicit GPU(Core::System& system, VideoCore::RendererBase& renderer); @@ -176,11 +177,11 @@ public: u32 address_high; u32 address_low; - GPUVAddr SmaphoreAddress() const { + GPUVAddr SemaphoreAddress() const { return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) | address_low); } - } smaphore_address; + } semaphore_address; u32 semaphore_sequence; u32 semaphore_trigger; @@ -244,9 +245,8 @@ protected: private: std::unique_ptr<Tegra::MemoryManager> memory_manager; - /// Mapping of command subchannels to their bound engine ids. + /// Mapping of command subchannels to their bound engine ids std::array<EngineID, 8> bound_engines = {}; - /// 3D engine std::unique_ptr<Engines::Maxwell3D> maxwell_3d; /// 2D engine @@ -263,7 +263,7 @@ private: static_assert(offsetof(GPU::Regs, field_name) == position * 4, \ "Field " #field_name " has invalid position") -ASSERT_REG_POSITION(smaphore_address, 0x4); +ASSERT_REG_POSITION(semaphore_address, 0x4); ASSERT_REG_POSITION(semaphore_sequence, 0x6); ASSERT_REG_POSITION(semaphore_trigger, 0x7); ASSERT_REG_POSITION(reference_count, 0x14); diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index 086b2f625..c5dc199c5 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp @@ -52,8 +52,8 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p } ThreadManager::ThreadManager(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher) - : renderer{renderer}, dma_pusher{dma_pusher}, thread{RunThread, std::ref(renderer), - std::ref(dma_pusher), std::ref(state)} {} + : renderer{renderer}, thread{RunThread, std::ref(renderer), std::ref(dma_pusher), + std::ref(state)} {} ThreadManager::~ThreadManager() { // Notify GPU thread that a shutdown is pending diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h index 8cd7db1c6..70acb2e79 100644 --- a/src/video_core/gpu_thread.h +++ b/src/video_core/gpu_thread.h @@ -4,10 +4,8 @@ #pragma once -#include <array> #include <atomic> #include <condition_variable> -#include <memory> #include <mutex> #include <optional> #include <thread> @@ -97,13 +95,13 @@ struct SynchState final { std::condition_variable frames_condition; void IncrementFramesCounter() { - std::lock_guard<std::mutex> lock{frames_mutex}; + std::lock_guard lock{frames_mutex}; ++queued_frame_count; } void DecrementFramesCounter() { { - std::lock_guard<std::mutex> lock{frames_mutex}; + std::lock_guard lock{frames_mutex}; --queued_frame_count; if (queued_frame_count) { @@ -115,7 +113,7 @@ struct SynchState final { void WaitForFrames() { { - std::lock_guard<std::mutex> lock{frames_mutex}; + std::lock_guard lock{frames_mutex}; if (!queued_frame_count) { return; } @@ -123,14 +121,14 @@ struct SynchState final { // Wait for the GPU to be idle (all commands to be executed) { - std::unique_lock<std::mutex> lock{frames_mutex}; + std::unique_lock lock{frames_mutex}; frames_condition.wait(lock, [this] { return !queued_frame_count; }); } } void SignalCommands() { { - std::unique_lock<std::mutex> lock{commands_mutex}; + std::unique_lock lock{commands_mutex}; if (queue.Empty()) { return; } @@ -140,7 +138,7 @@ struct SynchState final { } void WaitForCommands() { - std::unique_lock<std::mutex> lock{commands_mutex}; + std::unique_lock lock{commands_mutex}; commands_condition.wait(lock, [this] { return !queue.Empty(); }); } @@ -177,7 +175,6 @@ private: private: SynchState state; VideoCore::RendererBase& renderer; - Tegra::DmaPusher& dma_pusher; std::thread thread; std::thread::id thread_id; }; diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp index 8e8f36f28..e76b59842 100644 --- a/src/video_core/memory_manager.cpp +++ b/src/video_core/memory_manager.cpp @@ -5,218 +5,446 @@ #include "common/alignment.h" #include "common/assert.h" #include "common/logging/log.h" +#include "core/core.h" #include "core/memory.h" +#include "video_core/gpu.h" #include "video_core/memory_manager.h" +#include "video_core/rasterizer_interface.h" +#include "video_core/renderer_base.h" namespace Tegra { MemoryManager::MemoryManager() { - // Mark the first page as reserved, so that 0 is not a valid GPUVAddr. Otherwise, games might - // try to use 0 as a valid address, which is also used to mean nullptr. This fixes a bug with - // Undertale using 0 for a render target. - PageSlot(0) = static_cast<u64>(PageStatus::Reserved); -} - -GPUVAddr MemoryManager::AllocateSpace(u64 size, u64 align) { - const std::optional<GPUVAddr> gpu_addr{FindFreeBlock(0, size, align, PageStatus::Unmapped)}; + std::fill(page_table.pointers.begin(), page_table.pointers.end(), nullptr); + std::fill(page_table.attributes.begin(), page_table.attributes.end(), + Common::PageType::Unmapped); + page_table.Resize(address_space_width); - ASSERT_MSG(gpu_addr, "unable to find available GPU memory"); + // Initialize the map with a single free region covering the entire managed space. + VirtualMemoryArea initial_vma; + initial_vma.size = address_space_end; + vma_map.emplace(initial_vma.base, initial_vma); - for (u64 offset{}; offset < size; offset += PAGE_SIZE) { - VAddr& slot{PageSlot(*gpu_addr + offset)}; + UpdatePageTableForVMA(initial_vma); +} - ASSERT(slot == static_cast<u64>(PageStatus::Unmapped)); +GPUVAddr MemoryManager::AllocateSpace(u64 size, u64 align) { + const u64 aligned_size{Common::AlignUp(size, page_size)}; + const GPUVAddr gpu_addr{FindFreeRegion(address_space_base, aligned_size)}; - slot = static_cast<u64>(PageStatus::Allocated); - } + AllocateMemory(gpu_addr, 0, aligned_size); - return *gpu_addr; + return gpu_addr; } GPUVAddr MemoryManager::AllocateSpace(GPUVAddr gpu_addr, u64 size, u64 align) { - for (u64 offset{}; offset < size; offset += PAGE_SIZE) { - VAddr& slot{PageSlot(gpu_addr + offset)}; + const u64 aligned_size{Common::AlignUp(size, page_size)}; - ASSERT(slot == static_cast<u64>(PageStatus::Unmapped)); - - slot = static_cast<u64>(PageStatus::Allocated); - } + AllocateMemory(gpu_addr, 0, aligned_size); return gpu_addr; } GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, u64 size) { - const std::optional<GPUVAddr> gpu_addr{FindFreeBlock(0, size, PAGE_SIZE, PageStatus::Unmapped)}; + const u64 aligned_size{Common::AlignUp(size, page_size)}; + const GPUVAddr gpu_addr{FindFreeRegion(address_space_base, aligned_size)}; - ASSERT_MSG(gpu_addr, "unable to find available GPU memory"); + MapBackingMemory(gpu_addr, Memory::GetPointer(cpu_addr), aligned_size, cpu_addr); - for (u64 offset{}; offset < size; offset += PAGE_SIZE) { - VAddr& slot{PageSlot(*gpu_addr + offset)}; + return gpu_addr; +} - ASSERT(slot == static_cast<u64>(PageStatus::Unmapped)); +GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size) { + ASSERT((gpu_addr & page_mask) == 0); - slot = cpu_addr + offset; - } + const u64 aligned_size{Common::AlignUp(size, page_size)}; - const MappedRegion region{cpu_addr, *gpu_addr, size}; - mapped_regions.push_back(region); + MapBackingMemory(gpu_addr, Memory::GetPointer(cpu_addr), aligned_size, cpu_addr); - return *gpu_addr; + return gpu_addr; } -GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size) { - ASSERT((gpu_addr & PAGE_MASK) == 0); +GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) { + ASSERT((gpu_addr & page_mask) == 0); - if (PageSlot(gpu_addr) != static_cast<u64>(PageStatus::Allocated)) { - // Page has been already mapped. In this case, we must find a new area of memory to use that - // is different than the specified one. Super Mario Odyssey hits this scenario when changing - // areas, but we do not want to overwrite the old pages. - // TODO(bunnei): We need to write a hardware test to confirm this behavior. + const u64 aligned_size{Common::AlignUp(size, page_size)}; + const CacheAddr cache_addr{ToCacheAddr(GetPointer(gpu_addr))}; - LOG_ERROR(HW_GPU, "attempting to map addr 0x{:016X}, which is not available!", gpu_addr); + Core::System::GetInstance().Renderer().Rasterizer().FlushAndInvalidateRegion(cache_addr, + aligned_size); + UnmapRange(gpu_addr, aligned_size); - const std::optional<GPUVAddr> new_gpu_addr{ - FindFreeBlock(gpu_addr, size, PAGE_SIZE, PageStatus::Allocated)}; + return gpu_addr; +} - ASSERT_MSG(new_gpu_addr, "unable to find available GPU memory"); +GPUVAddr MemoryManager::FindFreeRegion(GPUVAddr region_start, u64 size) { + // Find the first Free VMA. + const VMAHandle vma_handle{std::find_if(vma_map.begin(), vma_map.end(), [&](const auto& vma) { + if (vma.second.type != VirtualMemoryArea::Type::Unmapped) { + return false; + } - gpu_addr = *new_gpu_addr; + const VAddr vma_end{vma.second.base + vma.second.size}; + return vma_end > region_start && vma_end >= region_start + size; + })}; + + if (vma_handle == vma_map.end()) { + return {}; } - for (u64 offset{}; offset < size; offset += PAGE_SIZE) { - VAddr& slot{PageSlot(gpu_addr + offset)}; + return std::max(region_start, vma_handle->second.base); +} - ASSERT(slot == static_cast<u64>(PageStatus::Allocated)); +bool MemoryManager::IsAddressValid(GPUVAddr addr) const { + return (addr >> page_bits) < page_table.pointers.size(); +} - slot = cpu_addr + offset; +std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr addr) { + if (!IsAddressValid(addr)) { + return {}; } - const MappedRegion region{cpu_addr, gpu_addr, size}; - mapped_regions.push_back(region); + VAddr cpu_addr{page_table.backing_addr[addr >> page_bits]}; + if (cpu_addr) { + return cpu_addr + (addr & page_mask); + } - return gpu_addr; + return {}; } -GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) { - ASSERT((gpu_addr & PAGE_MASK) == 0); +template <typename T> +T MemoryManager::Read(GPUVAddr addr) { + if (!IsAddressValid(addr)) { + return {}; + } - for (u64 offset{}; offset < size; offset += PAGE_SIZE) { - VAddr& slot{PageSlot(gpu_addr + offset)}; + const u8* page_pointer{page_table.pointers[addr >> page_bits]}; + if (page_pointer) { + // NOTE: Avoid adding any extra logic to this fast-path block + T value; + std::memcpy(&value, &page_pointer[addr & page_mask], sizeof(T)); + return value; + } - ASSERT(slot != static_cast<u64>(PageStatus::Allocated) && - slot != static_cast<u64>(PageStatus::Unmapped)); + switch (page_table.attributes[addr >> page_bits]) { + case Common::PageType::Unmapped: + LOG_ERROR(HW_GPU, "Unmapped Read{} @ 0x{:08X}", sizeof(T) * 8, addr); + return 0; + case Common::PageType::Memory: + ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", addr); + break; + default: + UNREACHABLE(); + } + return {}; +} - slot = static_cast<u64>(PageStatus::Unmapped); +template <typename T> +void MemoryManager::Write(GPUVAddr addr, T data) { + if (!IsAddressValid(addr)) { + return; } - // Delete the region mappings that are contained within the unmapped region - mapped_regions.erase(std::remove_if(mapped_regions.begin(), mapped_regions.end(), - [&](const MappedRegion& region) { - return region.gpu_addr <= gpu_addr && - region.gpu_addr + region.size < gpu_addr + size; - }), - mapped_regions.end()); - return gpu_addr; -} + u8* page_pointer{page_table.pointers[addr >> page_bits]}; + if (page_pointer) { + // NOTE: Avoid adding any extra logic to this fast-path block + std::memcpy(&page_pointer[addr & page_mask], &data, sizeof(T)); + return; + } -GPUVAddr MemoryManager::GetRegionEnd(GPUVAddr region_start) const { - for (const auto& region : mapped_regions) { - const GPUVAddr region_end{region.gpu_addr + region.size}; - if (region_start >= region.gpu_addr && region_start < region_end) { - return region_end; - } + switch (page_table.attributes[addr >> page_bits]) { + case Common::PageType::Unmapped: + LOG_ERROR(HW_GPU, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8, + static_cast<u32>(data), addr); + return; + case Common::PageType::Memory: + ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", addr); + break; + default: + UNREACHABLE(); } - return {}; } -std::optional<GPUVAddr> MemoryManager::FindFreeBlock(GPUVAddr region_start, u64 size, u64 align, - PageStatus status) { - GPUVAddr gpu_addr{region_start}; - u64 free_space{}; - align = (align + PAGE_MASK) & ~PAGE_MASK; - - while (gpu_addr + free_space < MAX_ADDRESS) { - if (PageSlot(gpu_addr + free_space) == static_cast<u64>(status)) { - free_space += PAGE_SIZE; - if (free_space >= size) { - return gpu_addr; - } - } else { - gpu_addr += free_space + PAGE_SIZE; - free_space = 0; - gpu_addr = Common::AlignUp(gpu_addr, align); - } +template u8 MemoryManager::Read<u8>(GPUVAddr addr); +template u16 MemoryManager::Read<u16>(GPUVAddr addr); +template u32 MemoryManager::Read<u32>(GPUVAddr addr); +template u64 MemoryManager::Read<u64>(GPUVAddr addr); +template void MemoryManager::Write<u8>(GPUVAddr addr, u8 data); +template void MemoryManager::Write<u16>(GPUVAddr addr, u16 data); +template void MemoryManager::Write<u32>(GPUVAddr addr, u32 data); +template void MemoryManager::Write<u64>(GPUVAddr addr, u64 data); + +u8* MemoryManager::GetPointer(GPUVAddr addr) { + if (!IsAddressValid(addr)) { + return {}; } + u8* page_pointer{page_table.pointers[addr >> page_bits]}; + if (page_pointer) { + return page_pointer + (addr & page_mask); + } + + LOG_ERROR(HW_GPU, "Unknown GetPointer @ 0x{:016X}", addr); return {}; } -std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) { - const VAddr base_addr{PageSlot(gpu_addr)}; +void MemoryManager::ReadBlock(GPUVAddr src_addr, void* dest_buffer, std::size_t size) { + std::memcpy(dest_buffer, GetPointer(src_addr), size); +} +void MemoryManager::WriteBlock(GPUVAddr dest_addr, const void* src_buffer, std::size_t size) { + std::memcpy(GetPointer(dest_addr), src_buffer, size); +} + +void MemoryManager::CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size) { + std::memcpy(GetPointer(dest_addr), GetPointer(src_addr), size); +} - if (base_addr == static_cast<u64>(PageStatus::Allocated) || - base_addr == static_cast<u64>(PageStatus::Unmapped) || - base_addr == static_cast<u64>(PageStatus::Reserved)) { - return {}; +void MemoryManager::MapPages(GPUVAddr base, u64 size, u8* memory, Common::PageType type, + VAddr backing_addr) { + LOG_DEBUG(HW_GPU, "Mapping {} onto {:016X}-{:016X}", fmt::ptr(memory), base * page_size, + (base + size) * page_size); + + const VAddr end{base + size}; + ASSERT_MSG(end <= page_table.pointers.size(), "out of range mapping at {:016X}", + base + page_table.pointers.size()); + + std::fill(page_table.attributes.begin() + base, page_table.attributes.begin() + end, type); + + if (memory == nullptr) { + std::fill(page_table.pointers.begin() + base, page_table.pointers.begin() + end, memory); + std::fill(page_table.backing_addr.begin() + base, page_table.backing_addr.begin() + end, + backing_addr); + } else { + while (base != end) { + page_table.pointers[base] = memory; + page_table.backing_addr[base] = backing_addr; + + base += 1; + memory += page_size; + backing_addr += page_size; + } } +} - return base_addr + (gpu_addr & PAGE_MASK); +void MemoryManager::MapMemoryRegion(GPUVAddr base, u64 size, u8* target, VAddr backing_addr) { + ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: {:016X}", size); + ASSERT_MSG((base & page_mask) == 0, "non-page aligned base: {:016X}", base); + MapPages(base / page_size, size / page_size, target, Common::PageType::Memory, backing_addr); } -u8 MemoryManager::Read8(GPUVAddr addr) { - return Memory::Read8(*GpuToCpuAddress(addr)); +void MemoryManager::UnmapRegion(GPUVAddr base, u64 size) { + ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: {:016X}", size); + ASSERT_MSG((base & page_mask) == 0, "non-page aligned base: {:016X}", base); + MapPages(base / page_size, size / page_size, nullptr, Common::PageType::Unmapped); } -u16 MemoryManager::Read16(GPUVAddr addr) { - return Memory::Read16(*GpuToCpuAddress(addr)); +bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const { + ASSERT(base + size == next.base); + if (type != next.type) { + return {}; + } + if (type == VirtualMemoryArea::Type::Allocated && (offset + size != next.offset)) { + return {}; + } + if (type == VirtualMemoryArea::Type::Mapped && backing_memory + size != next.backing_memory) { + return {}; + } + return true; } -u32 MemoryManager::Read32(GPUVAddr addr) { - return Memory::Read32(*GpuToCpuAddress(addr)); +MemoryManager::VMAHandle MemoryManager::FindVMA(GPUVAddr target) const { + if (target >= address_space_end) { + return vma_map.end(); + } else { + return std::prev(vma_map.upper_bound(target)); + } } -u64 MemoryManager::Read64(GPUVAddr addr) { - return Memory::Read64(*GpuToCpuAddress(addr)); +MemoryManager::VMAIter MemoryManager::Allocate(VMAIter vma_handle) { + VirtualMemoryArea& vma{vma_handle->second}; + + vma.type = VirtualMemoryArea::Type::Allocated; + vma.backing_addr = 0; + vma.backing_memory = {}; + UpdatePageTableForVMA(vma); + + return MergeAdjacent(vma_handle); } -void MemoryManager::Write8(GPUVAddr addr, u8 data) { - Memory::Write8(*GpuToCpuAddress(addr), data); +MemoryManager::VMAHandle MemoryManager::AllocateMemory(GPUVAddr target, std::size_t offset, + u64 size) { + + // This is the appropriately sized VMA that will turn into our allocation. + VMAIter vma_handle{CarveVMA(target, size)}; + VirtualMemoryArea& vma{vma_handle->second}; + + ASSERT(vma.size == size); + + vma.offset = offset; + + return Allocate(vma_handle); } -void MemoryManager::Write16(GPUVAddr addr, u16 data) { - Memory::Write16(*GpuToCpuAddress(addr), data); +MemoryManager::VMAHandle MemoryManager::MapBackingMemory(GPUVAddr target, u8* memory, u64 size, + VAddr backing_addr) { + // This is the appropriately sized VMA that will turn into our allocation. + VMAIter vma_handle{CarveVMA(target, size)}; + VirtualMemoryArea& vma{vma_handle->second}; + + ASSERT(vma.size == size); + + vma.type = VirtualMemoryArea::Type::Mapped; + vma.backing_memory = memory; + vma.backing_addr = backing_addr; + UpdatePageTableForVMA(vma); + + return MergeAdjacent(vma_handle); } -void MemoryManager::Write32(GPUVAddr addr, u32 data) { - Memory::Write32(*GpuToCpuAddress(addr), data); +void MemoryManager::UnmapRange(GPUVAddr target, u64 size) { + VMAIter vma{CarveVMARange(target, size)}; + const VAddr target_end{target + size}; + const VMAIter end{vma_map.end()}; + + // The comparison against the end of the range must be done using addresses since VMAs can be + // merged during this process, causing invalidation of the iterators. + while (vma != end && vma->second.base < target_end) { + // Unmapped ranges return to allocated state and can be reused + // This behavior is used by Super Mario Odyssey, Sonic Forces, and likely other games + vma = std::next(Allocate(vma)); + } + + ASSERT(FindVMA(target)->second.size >= size); } -void MemoryManager::Write64(GPUVAddr addr, u64 data) { - Memory::Write64(*GpuToCpuAddress(addr), data); +MemoryManager::VMAIter MemoryManager::StripIterConstness(const VMAHandle& iter) { + // This uses a neat C++ trick to convert a const_iterator to a regular iterator, given + // non-const access to its container. + return vma_map.erase(iter, iter); // Erases an empty range of elements } -u8* MemoryManager::GetPointer(GPUVAddr addr) { - return Memory::GetPointer(*GpuToCpuAddress(addr)); +MemoryManager::VMAIter MemoryManager::CarveVMA(GPUVAddr base, u64 size) { + ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: 0x{:016X}", size); + ASSERT_MSG((base & page_mask) == 0, "non-page aligned base: 0x{:016X}", base); + + VMAIter vma_handle{StripIterConstness(FindVMA(base))}; + if (vma_handle == vma_map.end()) { + // Target address is outside the managed range + return {}; + } + + const VirtualMemoryArea& vma{vma_handle->second}; + if (vma.type == VirtualMemoryArea::Type::Mapped) { + // Region is already allocated + return {}; + } + + const VAddr start_in_vma{base - vma.base}; + const VAddr end_in_vma{start_in_vma + size}; + + ASSERT_MSG(end_in_vma <= vma.size, "region size 0x{:016X} is less than required size 0x{:016X}", + vma.size, end_in_vma); + + if (end_in_vma < vma.size) { + // Split VMA at the end of the allocated region + SplitVMA(vma_handle, end_in_vma); + } + if (start_in_vma != 0) { + // Split VMA at the start of the allocated region + vma_handle = SplitVMA(vma_handle, start_in_vma); + } + + return vma_handle; } -void MemoryManager::ReadBlock(GPUVAddr src_addr, void* dest_buffer, std::size_t size) { - std::memcpy(dest_buffer, GetPointer(src_addr), size); +MemoryManager::VMAIter MemoryManager::CarveVMARange(GPUVAddr target, u64 size) { + ASSERT_MSG((size & page_mask) == 0, "non-page aligned size: 0x{:016X}", size); + ASSERT_MSG((target & page_mask) == 0, "non-page aligned base: 0x{:016X}", target); + + const VAddr target_end{target + size}; + ASSERT(target_end >= target); + ASSERT(size > 0); + + VMAIter begin_vma{StripIterConstness(FindVMA(target))}; + const VMAIter i_end{vma_map.lower_bound(target_end)}; + if (std::any_of(begin_vma, i_end, [](const auto& entry) { + return entry.second.type == VirtualMemoryArea::Type::Unmapped; + })) { + return {}; + } + + if (target != begin_vma->second.base) { + begin_vma = SplitVMA(begin_vma, target - begin_vma->second.base); + } + + VMAIter end_vma{StripIterConstness(FindVMA(target_end))}; + if (end_vma != vma_map.end() && target_end != end_vma->second.base) { + end_vma = SplitVMA(end_vma, target_end - end_vma->second.base); + } + + return begin_vma; } -void MemoryManager::WriteBlock(GPUVAddr dest_addr, const void* src_buffer, std::size_t size) { - std::memcpy(GetPointer(dest_addr), src_buffer, size); + +MemoryManager::VMAIter MemoryManager::SplitVMA(VMAIter vma_handle, u64 offset_in_vma) { + VirtualMemoryArea& old_vma{vma_handle->second}; + VirtualMemoryArea new_vma{old_vma}; // Make a copy of the VMA + + // For now, don't allow no-op VMA splits (trying to split at a boundary) because it's probably + // a bug. This restriction might be removed later. + ASSERT(offset_in_vma < old_vma.size); + ASSERT(offset_in_vma > 0); + + old_vma.size = offset_in_vma; + new_vma.base += offset_in_vma; + new_vma.size -= offset_in_vma; + + switch (new_vma.type) { + case VirtualMemoryArea::Type::Unmapped: + break; + case VirtualMemoryArea::Type::Allocated: + new_vma.offset += offset_in_vma; + break; + case VirtualMemoryArea::Type::Mapped: + new_vma.backing_memory += offset_in_vma; + break; + } + + ASSERT(old_vma.CanBeMergedWith(new_vma)); + + return vma_map.emplace_hint(std::next(vma_handle), new_vma.base, new_vma); } -void MemoryManager::CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size) { - std::memcpy(GetPointer(dest_addr), GetPointer(src_addr), size); +MemoryManager::VMAIter MemoryManager::MergeAdjacent(VMAIter iter) { + const VMAIter next_vma{std::next(iter)}; + if (next_vma != vma_map.end() && iter->second.CanBeMergedWith(next_vma->second)) { + iter->second.size += next_vma->second.size; + vma_map.erase(next_vma); + } + + if (iter != vma_map.begin()) { + VMAIter prev_vma{std::prev(iter)}; + if (prev_vma->second.CanBeMergedWith(iter->second)) { + prev_vma->second.size += iter->second.size; + vma_map.erase(iter); + iter = prev_vma; + } + } + + return iter; } -VAddr& MemoryManager::PageSlot(GPUVAddr gpu_addr) { - auto& block{page_table[(gpu_addr >> (PAGE_BITS + PAGE_TABLE_BITS)) & PAGE_TABLE_MASK]}; - if (!block) { - block = std::make_unique<PageBlock>(); - block->fill(static_cast<VAddr>(PageStatus::Unmapped)); +void MemoryManager::UpdatePageTableForVMA(const VirtualMemoryArea& vma) { + switch (vma.type) { + case VirtualMemoryArea::Type::Unmapped: + UnmapRegion(vma.base, vma.size); + break; + case VirtualMemoryArea::Type::Allocated: + MapMemoryRegion(vma.base, vma.size, nullptr, vma.backing_addr); + break; + case VirtualMemoryArea::Type::Mapped: + MapMemoryRegion(vma.base, vma.size, vma.backing_memory, vma.backing_addr); + break; } - return (*block)[(gpu_addr >> PAGE_BITS) & PAGE_BLOCK_MASK]; } } // namespace Tegra diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h index 425e2f31c..34744bb27 100644 --- a/src/video_core/memory_manager.h +++ b/src/video_core/memory_manager.h @@ -1,82 +1,148 @@ -// Copyright 2018 yuzu emulator team +// Copyright 2018 yuzu emulator team // Licensed under GPLv2 or any later version // Refer to the license.txt file included. #pragma once -#include <array> -#include <memory> +#include <map> #include <optional> -#include <vector> #include "common/common_types.h" +#include "common/page_table.h" namespace Tegra { -/// Virtual addresses in the GPU's memory map are 64 bit. -using GPUVAddr = u64; +/** + * Represents a VMA in an address space. A VMA is a contiguous region of virtual addressing space + * with homogeneous attributes across its extents. In this particular implementation each VMA is + * also backed by a single host memory allocation. + */ +struct VirtualMemoryArea { + enum class Type : u8 { + Unmapped, + Allocated, + Mapped, + }; + + /// Virtual base address of the region. + GPUVAddr base{}; + /// Size of the region. + u64 size{}; + /// Memory area mapping type. + Type type{Type::Unmapped}; + /// CPU memory mapped address corresponding to this memory area. + VAddr backing_addr{}; + /// Offset into the backing_memory the mapping starts from. + std::size_t offset{}; + /// Pointer backing this VMA. + u8* backing_memory{}; + + /// Tests if this area can be merged to the right with `next`. + bool CanBeMergedWith(const VirtualMemoryArea& next) const; +}; class MemoryManager final { public: MemoryManager(); GPUVAddr AllocateSpace(u64 size, u64 align); - GPUVAddr AllocateSpace(GPUVAddr gpu_addr, u64 size, u64 align); + GPUVAddr AllocateSpace(GPUVAddr addr, u64 size, u64 align); GPUVAddr MapBufferEx(VAddr cpu_addr, u64 size); - GPUVAddr MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size); - GPUVAddr UnmapBuffer(GPUVAddr gpu_addr, u64 size); - GPUVAddr GetRegionEnd(GPUVAddr region_start) const; - std::optional<VAddr> GpuToCpuAddress(GPUVAddr gpu_addr); - - static constexpr u64 PAGE_BITS = 16; - static constexpr u64 PAGE_SIZE = 1 << PAGE_BITS; - static constexpr u64 PAGE_MASK = PAGE_SIZE - 1; + GPUVAddr MapBufferEx(VAddr cpu_addr, GPUVAddr addr, u64 size); + GPUVAddr UnmapBuffer(GPUVAddr addr, u64 size); + std::optional<VAddr> GpuToCpuAddress(GPUVAddr addr); - u8 Read8(GPUVAddr addr); - u16 Read16(GPUVAddr addr); - u32 Read32(GPUVAddr addr); - u64 Read64(GPUVAddr addr); + template <typename T> + T Read(GPUVAddr addr); - void Write8(GPUVAddr addr, u8 data); - void Write16(GPUVAddr addr, u16 data); - void Write32(GPUVAddr addr, u32 data); - void Write64(GPUVAddr addr, u64 data); + template <typename T> + void Write(GPUVAddr addr, T data); - u8* GetPointer(GPUVAddr vaddr); + u8* GetPointer(GPUVAddr addr); void ReadBlock(GPUVAddr src_addr, void* dest_buffer, std::size_t size); void WriteBlock(GPUVAddr dest_addr, const void* src_buffer, std::size_t size); - void CopyBlock(VAddr dest_addr, VAddr src_addr, std::size_t size); + void CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size); private: - enum class PageStatus : u64 { - Unmapped = 0xFFFFFFFFFFFFFFFFULL, - Allocated = 0xFFFFFFFFFFFFFFFEULL, - Reserved = 0xFFFFFFFFFFFFFFFDULL, - }; - - std::optional<GPUVAddr> FindFreeBlock(GPUVAddr region_start, u64 size, u64 align, - PageStatus status); - VAddr& PageSlot(GPUVAddr gpu_addr); - - static constexpr u64 MAX_ADDRESS{0x10000000000ULL}; - static constexpr u64 PAGE_TABLE_BITS{10}; - static constexpr u64 PAGE_TABLE_SIZE{1 << PAGE_TABLE_BITS}; - static constexpr u64 PAGE_TABLE_MASK{PAGE_TABLE_SIZE - 1}; - static constexpr u64 PAGE_BLOCK_BITS{14}; - static constexpr u64 PAGE_BLOCK_SIZE{1 << PAGE_BLOCK_BITS}; - static constexpr u64 PAGE_BLOCK_MASK{PAGE_BLOCK_SIZE - 1}; - - using PageBlock = std::array<VAddr, PAGE_BLOCK_SIZE>; - std::array<std::unique_ptr<PageBlock>, PAGE_TABLE_SIZE> page_table{}; - - struct MappedRegion { - VAddr cpu_addr; - GPUVAddr gpu_addr; - u64 size; - }; + using VMAMap = std::map<GPUVAddr, VirtualMemoryArea>; + using VMAHandle = VMAMap::const_iterator; + using VMAIter = VMAMap::iterator; + + bool IsAddressValid(GPUVAddr addr) const; + void MapPages(GPUVAddr base, u64 size, u8* memory, Common::PageType type, + VAddr backing_addr = 0); + void MapMemoryRegion(GPUVAddr base, u64 size, u8* target, VAddr backing_addr); + void UnmapRegion(GPUVAddr base, u64 size); + + /// Finds the VMA in which the given address is included in, or `vma_map.end()`. + VMAHandle FindVMA(GPUVAddr target) const; + + VMAHandle AllocateMemory(GPUVAddr target, std::size_t offset, u64 size); + + /** + * Maps an unmanaged host memory pointer at a given address. + * + * @param target The guest address to start the mapping at. + * @param memory The memory to be mapped. + * @param size Size of the mapping. + * @param state MemoryState tag to attach to the VMA. + */ + VMAHandle MapBackingMemory(GPUVAddr target, u8* memory, u64 size, VAddr backing_addr); + + /// Unmaps a range of addresses, splitting VMAs as necessary. + void UnmapRange(GPUVAddr target, u64 size); + + /// Converts a VMAHandle to a mutable VMAIter. + VMAIter StripIterConstness(const VMAHandle& iter); + + /// Marks as the specfied VMA as allocated. + VMAIter Allocate(VMAIter vma); + + /** + * Carves a VMA of a specific size at the specified address by splitting Free VMAs while doing + * the appropriate error checking. + */ + VMAIter CarveVMA(GPUVAddr base, u64 size); + + /** + * Splits the edges of the given range of non-Free VMAs so that there is a VMA split at each + * end of the range. + */ + VMAIter CarveVMARange(GPUVAddr base, u64 size); + + /** + * Splits a VMA in two, at the specified offset. + * @returns the right side of the split, with the original iterator becoming the left side. + */ + VMAIter SplitVMA(VMAIter vma, u64 offset_in_vma); + + /** + * Checks for and merges the specified VMA with adjacent ones if possible. + * @returns the merged VMA or the original if no merging was possible. + */ + VMAIter MergeAdjacent(VMAIter vma); + + /// Updates the pages corresponding to this VMA so they match the VMA's attributes. + void UpdatePageTableForVMA(const VirtualMemoryArea& vma); + + /// Finds a free (unmapped region) of the specified size starting at the specified address. + GPUVAddr FindFreeRegion(GPUVAddr region_start, u64 size); - std::vector<MappedRegion> mapped_regions; +private: + static constexpr u64 page_bits{16}; + static constexpr u64 page_size{1 << page_bits}; + static constexpr u64 page_mask{page_size - 1}; + + /// Address space in bits, this is fairly arbitrary but sufficiently large. + static constexpr u32 address_space_width{39}; + /// Start address for mapping, this is fairly arbitrary but must be non-zero. + static constexpr GPUVAddr address_space_base{0x100000}; + /// End of address space, based on address space in bits. + static constexpr GPUVAddr address_space_end{1ULL << address_space_width}; + + Common::PageTable page_table{page_bits}; + VMAMap vma_map; }; } // namespace Tegra diff --git a/src/video_core/rasterizer_cache.h b/src/video_core/rasterizer_cache.h index ecd9986a0..291772186 100644 --- a/src/video_core/rasterizer_cache.h +++ b/src/video_core/rasterizer_cache.h @@ -71,8 +71,8 @@ private: bool is_registered{}; ///< Whether the object is currently registered with the cache bool is_dirty{}; ///< Whether the object is dirty (out of sync with guest memory) u64 last_modified_ticks{}; ///< When the object was last modified, used for in-order flushing - CacheAddr cache_addr{}; ///< Cache address memory, unique from emulated virtual address space const u8* host_ptr{}; ///< Pointer to the memory backing this cached region + CacheAddr cache_addr{}; ///< Cache address memory, unique from emulated virtual address space }; template <class T> @@ -84,7 +84,7 @@ public: /// Write any cached resources overlapping the specified region back to memory void FlushRegion(CacheAddr addr, std::size_t size) { - std::lock_guard<std::recursive_mutex> lock{mutex}; + std::lock_guard lock{mutex}; const auto& objects{GetSortedObjectsFromRegion(addr, size)}; for (auto& object : objects) { @@ -94,7 +94,7 @@ public: /// Mark the specified region as being invalidated void InvalidateRegion(CacheAddr addr, u64 size) { - std::lock_guard<std::recursive_mutex> lock{mutex}; + std::lock_guard lock{mutex}; const auto& objects{GetSortedObjectsFromRegion(addr, size)}; for (auto& object : objects) { @@ -108,7 +108,7 @@ public: /// Invalidates everything in the cache void InvalidateAll() { - std::lock_guard<std::recursive_mutex> lock{mutex}; + std::lock_guard lock{mutex}; while (interval_cache.begin() != interval_cache.end()) { Unregister(*interval_cache.begin()->second.begin()); @@ -132,8 +132,8 @@ protected: } /// Register an object into the cache - void Register(const T& object) { - std::lock_guard<std::recursive_mutex> lock{mutex}; + virtual void Register(const T& object) { + std::lock_guard lock{mutex}; object->SetIsRegistered(true); interval_cache.add({GetInterval(object), ObjectSet{object}}); @@ -142,8 +142,8 @@ protected: } /// Unregisters an object from the cache - void Unregister(const T& object) { - std::lock_guard<std::recursive_mutex> lock{mutex}; + virtual void Unregister(const T& object) { + std::lock_guard lock{mutex}; object->SetIsRegistered(false); rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), -1); @@ -153,14 +153,14 @@ protected: /// Returns a ticks counter used for tracking when cached objects were last modified u64 GetModifiedTicks() { - std::lock_guard<std::recursive_mutex> lock{mutex}; + std::lock_guard lock{mutex}; return ++modified_ticks; } /// Flushes the specified object, updating appropriate cache state as needed void FlushObject(const T& object) { - std::lock_guard<std::recursive_mutex> lock{mutex}; + std::lock_guard lock{mutex}; if (!object->IsDirty()) { return; diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h index 76e292e87..d7b86df38 100644 --- a/src/video_core/rasterizer_interface.h +++ b/src/video_core/rasterizer_interface.h @@ -9,7 +9,6 @@ #include "common/common_types.h" #include "video_core/engines/fermi_2d.h" #include "video_core/gpu.h" -#include "video_core/memory_manager.h" namespace VideoCore { diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp index 5048ed6ce..fd091c84c 100644 --- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp @@ -15,14 +15,14 @@ namespace OpenGL { CachedBufferEntry::CachedBufferEntry(VAddr cpu_addr, std::size_t size, GLintptr offset, std::size_t alignment, u8* host_ptr) - : cpu_addr{cpu_addr}, size{size}, offset{offset}, alignment{alignment}, RasterizerCacheObject{ - host_ptr} {} + : RasterizerCacheObject{host_ptr}, cpu_addr{cpu_addr}, size{size}, offset{offset}, + alignment{alignment} {} OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, std::size_t size) : RasterizerCache{rasterizer}, stream_buffer(size, true) {} -GLintptr OGLBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, - std::size_t alignment, bool cache) { +GLintptr OGLBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::size_t alignment, + bool cache) { auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager(); // Cache management is a big overhead, so only cache entries with a given size. diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h index 1de1f84ae..fc33aa433 100644 --- a/src/video_core/renderer_opengl/gl_buffer_cache.h +++ b/src/video_core/renderer_opengl/gl_buffer_cache.h @@ -58,7 +58,7 @@ public: /// Uploads data from a guest GPU address. Returns host's buffer offset where it's been /// allocated. - GLintptr UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, std::size_t alignment = 4, + GLintptr UploadMemory(GPUVAddr gpu_addr, std::size_t size, std::size_t alignment = 4, bool cache = true); /// Uploads from a host memory. Returns host's buffer offset where it's been allocated. diff --git a/src/video_core/renderer_opengl/gl_global_cache.cpp b/src/video_core/renderer_opengl/gl_global_cache.cpp index c8dbcacbd..da9326253 100644 --- a/src/video_core/renderer_opengl/gl_global_cache.cpp +++ b/src/video_core/renderer_opengl/gl_global_cache.cpp @@ -15,7 +15,7 @@ namespace OpenGL { CachedGlobalRegion::CachedGlobalRegion(VAddr cpu_addr, u32 size, u8* host_ptr) - : cpu_addr{cpu_addr}, size{size}, RasterizerCacheObject{host_ptr} { + : RasterizerCacheObject{host_ptr}, cpu_addr{cpu_addr}, size{size} { buffer.Create(); // Bind and unbind the buffer so it gets allocated by the driver glBindBuffer(GL_SHADER_STORAGE_BUFFER, buffer.handle); @@ -46,7 +46,7 @@ GlobalRegion GlobalRegionCacheOpenGL::TryGetReservedGlobalRegion(CacheAddr addr, return search->second; } -GlobalRegion GlobalRegionCacheOpenGL::GetUncachedGlobalRegion(Tegra::GPUVAddr addr, u32 size, +GlobalRegion GlobalRegionCacheOpenGL::GetUncachedGlobalRegion(GPUVAddr addr, u32 size, u8* host_ptr) { GlobalRegion region{TryGetReservedGlobalRegion(ToCacheAddr(host_ptr), size)}; if (!region) { @@ -76,8 +76,8 @@ GlobalRegion GlobalRegionCacheOpenGL::GetGlobalRegion( const auto cbufs{gpu.Maxwell3D().state.shader_stages[static_cast<u64>(stage)]}; const auto addr{cbufs.const_buffers[global_region.GetCbufIndex()].address + global_region.GetCbufOffset()}; - const auto actual_addr{memory_manager.Read64(addr)}; - const auto size{memory_manager.Read32(addr + 8)}; + const auto actual_addr{memory_manager.Read<u64>(addr)}; + const auto size{memory_manager.Read<u32>(addr + 8)}; // Look up global region in the cache based on address const auto& host_ptr{memory_manager.GetPointer(actual_addr)}; diff --git a/src/video_core/renderer_opengl/gl_global_cache.h b/src/video_core/renderer_opengl/gl_global_cache.h index a840491f7..5a21ab66f 100644 --- a/src/video_core/renderer_opengl/gl_global_cache.h +++ b/src/video_core/renderer_opengl/gl_global_cache.h @@ -66,7 +66,7 @@ public: private: GlobalRegion TryGetReservedGlobalRegion(CacheAddr addr, u32 size) const; - GlobalRegion GetUncachedGlobalRegion(Tegra::GPUVAddr addr, u32 size, u8* host_ptr); + GlobalRegion GetUncachedGlobalRegion(GPUVAddr addr, u32 size, u8* host_ptr); void ReserveGlobalRegion(GlobalRegion region); std::unordered_map<CacheAddr, GlobalRegion> reserve; diff --git a/src/video_core/renderer_opengl/gl_primitive_assembler.cpp b/src/video_core/renderer_opengl/gl_primitive_assembler.cpp index 75d816795..2bcbd3da2 100644 --- a/src/video_core/renderer_opengl/gl_primitive_assembler.cpp +++ b/src/video_core/renderer_opengl/gl_primitive_assembler.cpp @@ -40,8 +40,7 @@ GLintptr PrimitiveAssembler::MakeQuadArray(u32 first, u32 count) { return index_offset; } -GLintptr PrimitiveAssembler::MakeQuadIndexed(Tegra::GPUVAddr gpu_addr, std::size_t index_size, - u32 count) { +GLintptr PrimitiveAssembler::MakeQuadIndexed(GPUVAddr gpu_addr, std::size_t index_size, u32 count) { const std::size_t map_size{CalculateQuadSize(count)}; auto [dst_pointer, index_offset] = buffer_cache.ReserveMemory(map_size); diff --git a/src/video_core/renderer_opengl/gl_primitive_assembler.h b/src/video_core/renderer_opengl/gl_primitive_assembler.h index a8cb88eb5..0e2e7dc36 100644 --- a/src/video_core/renderer_opengl/gl_primitive_assembler.h +++ b/src/video_core/renderer_opengl/gl_primitive_assembler.h @@ -24,7 +24,7 @@ public: GLintptr MakeQuadArray(u32 first, u32 count); - GLintptr MakeQuadIndexed(Tegra::GPUVAddr gpu_addr, std::size_t index_size, u32 count); + GLintptr MakeQuadIndexed(GPUVAddr gpu_addr, std::size_t index_size, u32 count); private: OGLBufferCache& buffer_cache; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 198c54872..046fc935b 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -100,11 +100,9 @@ struct FramebufferCacheKey { } }; -RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::System& system, - ScreenInfo& info) - : res_cache{*this}, shader_cache{*this, system}, global_cache{*this}, - emu_window{window}, system{system}, screen_info{info}, - buffer_cache(*this, STREAM_BUFFER_SIZE) { +RasterizerOpenGL::RasterizerOpenGL(Core::System& system, ScreenInfo& info) + : res_cache{*this}, shader_cache{*this, system}, global_cache{*this}, system{system}, + screen_info{info}, buffer_cache(*this, STREAM_BUFFER_SIZE) { // Create sampler objects for (std::size_t i = 0; i < texture_samplers.size(); ++i) { texture_samplers[i].Create(); @@ -225,8 +223,8 @@ void RasterizerOpenGL::SetupVertexBuffer(GLuint vao) { if (!vertex_array.IsEnabled()) continue; - const Tegra::GPUVAddr start = vertex_array.StartAddress(); - const Tegra::GPUVAddr end = regs.vertex_array_limit[index].LimitAddress(); + const GPUVAddr start = vertex_array.StartAddress(); + const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress(); ASSERT(end > start); const u64 size = end - start + 1; @@ -421,8 +419,8 @@ std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const { if (!regs.vertex_array[index].IsEnabled()) continue; - const Tegra::GPUVAddr start = regs.vertex_array[index].StartAddress(); - const Tegra::GPUVAddr end = regs.vertex_array_limit[index].LimitAddress(); + const GPUVAddr start = regs.vertex_array[index].StartAddress(); + const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress(); ASSERT(end > start); size += end - start + 1; diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index 30f3e8acb..4de565321 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -50,8 +50,7 @@ struct FramebufferCacheKey; class RasterizerOpenGL : public VideoCore::RasterizerInterface { public: - explicit RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::System& system, - ScreenInfo& info); + explicit RasterizerOpenGL(Core::System& system, ScreenInfo& info); ~RasterizerOpenGL() override; void DrawArrays() override; @@ -214,7 +213,6 @@ private: ShaderCacheOpenGL shader_cache; GlobalRegionCacheOpenGL global_cache; - Core::Frontend::EmuWindow& emu_window; Core::System& system; ScreenInfo& screen_info; diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp index 57329cd61..aba6ce731 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp @@ -55,7 +55,7 @@ static void ApplyTextureDefaults(GLuint texture, u32 max_mip_level) { } } -void SurfaceParams::InitCacheParameters(Tegra::GPUVAddr gpu_addr_) { +void SurfaceParams::InitCacheParameters(GPUVAddr gpu_addr_) { auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()}; gpu_addr = gpu_addr_; @@ -222,7 +222,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only, } /*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer( - u32 zeta_width, u32 zeta_height, Tegra::GPUVAddr zeta_address, Tegra::DepthFormat format, + u32 zeta_width, u32 zeta_height, GPUVAddr zeta_address, Tegra::DepthFormat format, u32 block_width, u32 block_height, u32 block_depth, Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type) { SurfaceParams params{}; @@ -562,8 +562,14 @@ void RasterizerCacheOpenGL::CopySurface(const Surface& src_surface, const Surfac } CachedSurface::CachedSurface(const SurfaceParams& params) - : params{params}, gl_target{SurfaceTargetToGL(params.target)}, - cached_size_in_bytes{params.size_in_bytes}, RasterizerCacheObject{params.host_ptr} { + : RasterizerCacheObject{params.host_ptr}, params{params}, + gl_target{SurfaceTargetToGL(params.target)}, cached_size_in_bytes{params.size_in_bytes} { + + const auto optional_cpu_addr{ + Core::System::GetInstance().GPU().MemoryManager().GpuToCpuAddress(params.gpu_addr)}; + ASSERT_MSG(optional_cpu_addr, "optional_cpu_addr is invalid"); + cpu_addr = *optional_cpu_addr; + texture.Create(gl_target); // TODO(Rodrigo): Using params.GetRect() returns a different size than using its Mip*(0) @@ -603,20 +609,6 @@ CachedSurface::CachedSurface(const SurfaceParams& params) ApplyTextureDefaults(texture.handle, params.max_mip_level); OpenGL::LabelGLObject(GL_TEXTURE, texture.handle, params.gpu_addr, params.IdentityString()); - - // Clamp size to mapped GPU memory region - // TODO(bunnei): Super Mario Odyssey maps a 0x40000 byte region and then uses it for a 0x80000 - // R32F render buffer. We do not yet know if this is a game bug or something else, but this - // check is necessary to prevent flushing from overwriting unmapped memory. - - auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()}; - const u64 max_size{memory_manager.GetRegionEnd(params.gpu_addr) - params.gpu_addr}; - if (cached_size_in_bytes > max_size) { - LOG_ERROR(HW_GPU, "Surface size {} exceeds region size {}", params.size_in_bytes, max_size); - cached_size_in_bytes = max_size; - } - - cpu_addr = *memory_manager.GpuToCpuAddress(params.gpu_addr); } MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 192, 64)); @@ -925,7 +917,7 @@ void RasterizerCacheOpenGL::LoadSurface(const Surface& surface) { } Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool preserve_contents) { - if (params.gpu_addr == 0 || params.height * params.width == 0) { + if (!params.IsValid()) { return {}; } @@ -941,7 +933,7 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool pres // If surface parameters changed and we care about keeping the previous data, recreate // the surface from the old one Surface new_surface{RecreateSurface(surface, params)}; - UnregisterSurface(surface); + Unregister(surface); Register(new_surface); if (new_surface->IsUploaded()) { RegisterReinterpretSurface(new_surface); @@ -949,7 +941,7 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool pres return new_surface; } else { // Delete the old surface before creating a new one to prevent collisions. - UnregisterSurface(surface); + Unregister(surface); } } @@ -980,11 +972,11 @@ void RasterizerCacheOpenGL::FastLayeredCopySurface(const Surface& src_surface, const auto& init_params{src_surface->GetSurfaceParams()}; const auto& dst_params{dst_surface->GetSurfaceParams()}; auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()}; - Tegra::GPUVAddr address{init_params.gpu_addr}; + GPUVAddr address{init_params.gpu_addr}; const std::size_t layer_size{dst_params.LayerMemorySize()}; for (u32 layer = 0; layer < dst_params.depth; layer++) { for (u32 mipmap = 0; mipmap < dst_params.max_mip_level; mipmap++) { - const Tegra::GPUVAddr sub_address{address + dst_params.GetMipmapLevelOffset(mipmap)}; + const GPUVAddr sub_address{address + dst_params.GetMipmapLevelOffset(mipmap)}; const Surface& copy{TryGet(memory_manager.GetPointer(sub_address))}; if (!copy) { continue; @@ -1244,10 +1236,9 @@ static std::optional<u32> TryFindBestMipMap(std::size_t memory, const SurfacePar return {}; } -static std::optional<u32> TryFindBestLayer(Tegra::GPUVAddr addr, const SurfaceParams params, - u32 mipmap) { +static std::optional<u32> TryFindBestLayer(GPUVAddr addr, const SurfaceParams params, u32 mipmap) { const std::size_t size{params.LayerMemorySize()}; - Tegra::GPUVAddr start{params.gpu_addr + params.GetMipmapLevelOffset(mipmap)}; + GPUVAddr start{params.gpu_addr + params.GetMipmapLevelOffset(mipmap)}; for (u32 i = 0; i < params.depth; i++) { if (start == addr) { return {i}; @@ -1304,12 +1295,12 @@ static bool IsReinterpretInvalidSecond(const Surface render_surface, bool RasterizerCacheOpenGL::PartialReinterpretSurface(Surface triggering_surface, Surface intersect) { if (IsReinterpretInvalid(triggering_surface, intersect)) { - UnregisterSurface(intersect); + Unregister(intersect); return false; } if (!LayerFitReinterpretSurface(*this, triggering_surface, intersect)) { if (IsReinterpretInvalidSecond(triggering_surface, intersect)) { - UnregisterSurface(intersect); + Unregister(intersect); return false; } FlushObject(intersect); diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h index 9366f47f2..e8073579f 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h @@ -109,6 +109,11 @@ struct SurfaceParams { return size; } + /// Returns true if the parameters constitute a valid rasterizer surface. + bool IsValid() const { + return gpu_addr && host_ptr && height && width; + } + /// Returns the exact size of the memory occupied by a layer in a texture in VRAM, including /// mipmaps. std::size_t LayerMemorySize() const { @@ -210,7 +215,7 @@ struct SurfaceParams { /// Creates SurfaceParams for a depth buffer configuration static SurfaceParams CreateForDepthBuffer( - u32 zeta_width, u32 zeta_height, Tegra::GPUVAddr zeta_address, Tegra::DepthFormat format, + u32 zeta_width, u32 zeta_height, GPUVAddr zeta_address, Tegra::DepthFormat format, u32 block_width, u32 block_height, u32 block_depth, Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type); @@ -232,7 +237,7 @@ struct SurfaceParams { } /// Initializes parameters for caching, should be called after everything has been initialized - void InitCacheParameters(Tegra::GPUVAddr gpu_addr); + void InitCacheParameters(GPUVAddr gpu_addr); std::string TargetName() const { switch (target) { @@ -297,7 +302,7 @@ struct SurfaceParams { bool srgb_conversion; // Parameters used for caching u8* host_ptr; - Tegra::GPUVAddr gpu_addr; + GPUVAddr gpu_addr; std::size_t size_in_bytes; std::size_t size_in_bytes_gl; @@ -533,13 +538,17 @@ private: return nullptr; } + void Register(const Surface& object) override { + RasterizerCache<Surface>::Register(object); + } + /// Unregisters an object from the cache - void UnregisterSurface(const Surface& object) { + void Unregister(const Surface& object) override { if (object->IsReinterpreted()) { auto interval = GetReinterpretInterval(object); reinterpreted_surfaces.erase(interval); } - Unregister(object); + RasterizerCache<Surface>::Unregister(object); } }; diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index 1ed740877..290e654bc 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -32,7 +32,7 @@ struct UnspecializedShader { namespace { /// Gets the address for the specified shader stage program -Tegra::GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) { +GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) { const auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()}; const auto& shader_config{gpu.regs.shader_config[static_cast<std::size_t>(program)]}; return gpu.regs.code_address.CodeAddress() + shader_config.offset; @@ -215,9 +215,9 @@ CachedShader::CachedShader(VAddr cpu_addr, u64 unique_identifier, Maxwell::ShaderProgram program_type, ShaderDiskCacheOpenGL& disk_cache, const PrecompiledPrograms& precompiled_programs, ProgramCode&& program_code, ProgramCode&& program_code_b, u8* host_ptr) - : host_ptr{host_ptr}, cpu_addr{cpu_addr}, unique_identifier{unique_identifier}, - program_type{program_type}, disk_cache{disk_cache}, - precompiled_programs{precompiled_programs}, RasterizerCacheObject{host_ptr} { + : RasterizerCacheObject{host_ptr}, host_ptr{host_ptr}, cpu_addr{cpu_addr}, + unique_identifier{unique_identifier}, program_type{program_type}, disk_cache{disk_cache}, + precompiled_programs{precompiled_programs} { const std::size_t code_size = CalculateProgramSize(program_code); const std::size_t code_size_b = @@ -245,9 +245,9 @@ CachedShader::CachedShader(VAddr cpu_addr, u64 unique_identifier, Maxwell::ShaderProgram program_type, ShaderDiskCacheOpenGL& disk_cache, const PrecompiledPrograms& precompiled_programs, GLShader::ProgramResult result, u8* host_ptr) - : cpu_addr{cpu_addr}, unique_identifier{unique_identifier}, program_type{program_type}, - disk_cache{disk_cache}, precompiled_programs{precompiled_programs}, RasterizerCacheObject{ - host_ptr} { + : RasterizerCacheObject{host_ptr}, cpu_addr{cpu_addr}, unique_identifier{unique_identifier}, + program_type{program_type}, disk_cache{disk_cache}, precompiled_programs{ + precompiled_programs} { code = std::move(result.first); entries = result.second; @@ -486,7 +486,7 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) { } auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()}; - const Tegra::GPUVAddr program_addr{GetShaderAddress(program)}; + const GPUVAddr program_addr{GetShaderAddress(program)}; // Look up shader in the cache based on address const auto& host_ptr{memory_manager.GetPointer(program_addr)}; diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index 5e3d862c6..a01efeb05 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -266,7 +266,7 @@ void RendererOpenGL::CreateRasterizer() { } // Initialize sRGB Usage OpenGLState::ClearsRGBUsed(); - rasterizer = std::make_unique<RasterizerOpenGL>(render_window, system, screen_info); + rasterizer = std::make_unique<RasterizerOpenGL>(system, screen_info); } void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture, diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp index 95eab3fec..388b5ffd5 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp @@ -19,8 +19,8 @@ namespace Vulkan { CachedBufferEntry::CachedBufferEntry(VAddr cpu_addr, std::size_t size, u64 offset, std::size_t alignment, u8* host_ptr) - : cpu_addr{cpu_addr}, size{size}, offset{offset}, alignment{alignment}, RasterizerCacheObject{ - host_ptr} {} + : RasterizerCacheObject{host_ptr}, cpu_addr{cpu_addr}, size{size}, offset{offset}, + alignment{alignment} {} VKBufferCache::VKBufferCache(Tegra::MemoryManager& tegra_memory_manager, VideoCore::RasterizerInterface& rasterizer, const VKDevice& device, @@ -39,8 +39,7 @@ VKBufferCache::VKBufferCache(Tegra::MemoryManager& tegra_memory_manager, VKBufferCache::~VKBufferCache() = default; -u64 VKBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, u64 alignment, - bool cache) { +u64 VKBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size, u64 alignment, bool cache) { const auto cpu_addr{tegra_memory_manager.GpuToCpuAddress(gpu_addr)}; ASSERT_MSG(cpu_addr, "Invalid GPU address"); diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h index 8b415744b..08b786aad 100644 --- a/src/video_core/renderer_vulkan/vk_buffer_cache.h +++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h @@ -68,8 +68,7 @@ public: /// Uploads data from a guest GPU address. Returns host's buffer offset where it's been /// allocated. - u64 UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, u64 alignment = 4, - bool cache = true); + u64 UploadMemory(GPUVAddr gpu_addr, std::size_t size, u64 alignment = 4, bool cache = true); /// Uploads from a host memory. Returns host's buffer offset where it's been allocated. u64 UploadHostMemory(const u8* raw_pointer, std::size_t size, u64 alignment = 4); diff --git a/src/video_core/renderer_vulkan/vk_resource_manager.cpp b/src/video_core/renderer_vulkan/vk_resource_manager.cpp index a1e117443..13c46e5b8 100644 --- a/src/video_core/renderer_vulkan/vk_resource_manager.cpp +++ b/src/video_core/renderer_vulkan/vk_resource_manager.cpp @@ -21,7 +21,7 @@ public: CommandBufferPool(const VKDevice& device) : VKFencedPool(COMMAND_BUFFER_POOL_SIZE), device{device} {} - void Allocate(std::size_t begin, std::size_t end) { + void Allocate(std::size_t begin, std::size_t end) override { const auto dev = device.GetLogical(); const auto& dld = device.GetDispatchLoader(); const u32 graphics_family = device.GetGraphicsFamily(); diff --git a/src/video_core/renderer_vulkan/vk_resource_manager.h b/src/video_core/renderer_vulkan/vk_resource_manager.h index 5bfe4cead..08ee86fa6 100644 --- a/src/video_core/renderer_vulkan/vk_resource_manager.h +++ b/src/video_core/renderer_vulkan/vk_resource_manager.h @@ -97,7 +97,7 @@ private: class VKFenceWatch final : public VKResource { public: explicit VKFenceWatch(); - ~VKFenceWatch(); + ~VKFenceWatch() override; /// Waits for the fence to be released. void Wait(); diff --git a/src/web_service/web_backend.cpp b/src/web_service/web_backend.cpp index 40da1a4e2..dc149d2ed 100644 --- a/src/web_service/web_backend.cpp +++ b/src/web_service/web_backend.cpp @@ -24,7 +24,7 @@ constexpr u32 TIMEOUT_SECONDS = 30; struct Client::Impl { Impl(std::string host, std::string username, std::string token) : host{std::move(host)}, username{std::move(username)}, token{std::move(token)} { - std::lock_guard<std::mutex> lock(jwt_cache.mutex); + std::lock_guard lock{jwt_cache.mutex}; if (this->username == jwt_cache.username && this->token == jwt_cache.token) { jwt = jwt_cache.jwt; } @@ -151,7 +151,7 @@ struct Client::Impl { if (result.result_code != Common::WebResult::Code::Success) { LOG_ERROR(WebService, "UpdateJWT failed"); } else { - std::lock_guard<std::mutex> lock(jwt_cache.mutex); + std::lock_guard lock{jwt_cache.mutex}; jwt_cache.username = username; jwt_cache.token = token; jwt_cache.jwt = jwt = result.returned_data; diff --git a/src/yuzu/applets/profile_select.cpp b/src/yuzu/applets/profile_select.cpp index 5c1b65a2c..730426c16 100644 --- a/src/yuzu/applets/profile_select.cpp +++ b/src/yuzu/applets/profile_select.cpp @@ -163,6 +163,6 @@ void QtProfileSelector::SelectProfile( void QtProfileSelector::MainWindowFinishedSelection(std::optional<Service::Account::UUID> uuid) { // Acquire the HLE mutex - std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock); + std::lock_guard lock{HLE::g_hle_lock}; callback(uuid); } diff --git a/src/yuzu/applets/software_keyboard.cpp b/src/yuzu/applets/software_keyboard.cpp index 8a26fdff1..eddc9c941 100644 --- a/src/yuzu/applets/software_keyboard.cpp +++ b/src/yuzu/applets/software_keyboard.cpp @@ -141,12 +141,12 @@ void QtSoftwareKeyboard::SendTextCheckDialog(std::u16string error_message, void QtSoftwareKeyboard::MainWindowFinishedText(std::optional<std::u16string> text) { // Acquire the HLE mutex - std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock); + std::lock_guard lock{HLE::g_hle_lock}; text_output(text); } void QtSoftwareKeyboard::MainWindowFinishedCheckDialog() { // Acquire the HLE mutex - std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock); + std::lock_guard lock{HLE::g_hle_lock}; finished_check(); } diff --git a/src/yuzu/applets/web_browser.cpp b/src/yuzu/applets/web_browser.cpp index 979b9ec14..ac80b2fa2 100644 --- a/src/yuzu/applets/web_browser.cpp +++ b/src/yuzu/applets/web_browser.cpp @@ -104,12 +104,12 @@ void QtWebBrowser::OpenPage(std::string_view url, std::function<void()> unpack_r void QtWebBrowser::MainWindowUnpackRomFS() { // Acquire the HLE mutex - std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock); + std::lock_guard lock{HLE::g_hle_lock}; unpack_romfs_callback(); } void QtWebBrowser::MainWindowFinishedBrowsing() { // Acquire the HLE mutex - std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock); + std::lock_guard lock{HLE::g_hle_lock}; finished_callback(); } diff --git a/src/yuzu/bootmanager.cpp b/src/yuzu/bootmanager.cpp index 05ad19e1d..7438fbc0a 100644 --- a/src/yuzu/bootmanager.cpp +++ b/src/yuzu/bootmanager.cpp @@ -67,7 +67,7 @@ void EmuThread::run() { was_active = false; } else { - std::unique_lock<std::mutex> lock(running_mutex); + std::unique_lock lock{running_mutex}; running_cv.wait(lock, [this] { return IsRunning() || exec_step || stop_run; }); } } diff --git a/src/yuzu/bootmanager.h b/src/yuzu/bootmanager.h index 7226e690e..3183621bc 100644 --- a/src/yuzu/bootmanager.h +++ b/src/yuzu/bootmanager.h @@ -53,7 +53,7 @@ public: * @note This function is thread-safe */ void SetRunning(bool running) { - std::unique_lock<std::mutex> lock(running_mutex); + std::unique_lock lock{running_mutex}; this->running = running; lock.unlock(); running_cv.notify_all(); diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp index 4650f96a3..dead9f807 100644 --- a/src/yuzu/configuration/config.cpp +++ b/src/yuzu/configuration/config.cpp @@ -413,7 +413,6 @@ void Config::ReadValues() { qt_config->beginGroup("System"); Settings::values.use_docked_mode = ReadSetting("use_docked_mode", false).toBool(); - Settings::values.enable_nfc = ReadSetting("enable_nfc", true).toBool(); Settings::values.current_user = std::clamp<int>(ReadSetting("current_user", 0).toInt(), 0, Service::Account::MAX_USERS - 1); @@ -675,7 +674,6 @@ void Config::SaveValues() { qt_config->beginGroup("System"); WriteSetting("use_docked_mode", Settings::values.use_docked_mode, false); - WriteSetting("enable_nfc", Settings::values.enable_nfc, true); WriteSetting("current_user", Settings::values.current_user, 0); WriteSetting("language_index", Settings::values.language_index, 1); diff --git a/src/yuzu/configuration/configure_general.cpp b/src/yuzu/configuration/configure_general.cpp index 4116b6cd7..389fcf667 100644 --- a/src/yuzu/configuration/configure_general.cpp +++ b/src/yuzu/configuration/configure_general.cpp @@ -33,7 +33,6 @@ void ConfigureGeneral::setConfiguration() { ui->toggle_user_on_boot->setChecked(UISettings::values.select_user_on_boot); ui->theme_combobox->setCurrentIndex(ui->theme_combobox->findData(UISettings::values.theme)); ui->use_cpu_jit->setChecked(Settings::values.use_cpu_jit); - ui->enable_nfc->setChecked(Settings::values.enable_nfc); } void ConfigureGeneral::PopulateHotkeyList(const HotkeyRegistry& registry) { @@ -48,5 +47,4 @@ void ConfigureGeneral::applyConfiguration() { ui->theme_combobox->itemData(ui->theme_combobox->currentIndex()).toString(); Settings::values.use_cpu_jit = ui->use_cpu_jit->isChecked(); - Settings::values.enable_nfc = ui->enable_nfc->isChecked(); } diff --git a/src/yuzu/configuration/configure_general.ui b/src/yuzu/configuration/configure_general.ui index dff0ad5d0..01d1c0b8e 100644 --- a/src/yuzu/configuration/configure_general.ui +++ b/src/yuzu/configuration/configure_general.ui @@ -71,26 +71,6 @@ </widget> </item> <item> - <widget class="QGroupBox" name="EmulationGroupBox"> - <property name="title"> - <string>Emulation</string> - </property> - <layout class="QHBoxLayout" name="EmulationHorizontalLayout"> - <item> - <layout class="QVBoxLayout" name="EmulationVerticalLayout"> - <item> - <widget class="QCheckBox" name="enable_nfc"> - <property name="text"> - <string>Enable NFC</string> - </property> - </widget> - </item> - </layout> - </item> - </layout> - </widget> - </item> - <item> <widget class="QGroupBox" name="theme_group_box"> <property name="title"> <string>Theme</string> diff --git a/src/yuzu/debugger/graphics/graphics_surface.cpp b/src/yuzu/debugger/graphics/graphics_surface.cpp index 29f01dfb2..11023ed63 100644 --- a/src/yuzu/debugger/graphics/graphics_surface.cpp +++ b/src/yuzu/debugger/graphics/graphics_surface.cpp @@ -261,7 +261,7 @@ void GraphicsSurfaceWidget::OnSurfaceSourceChanged(int new_value) { void GraphicsSurfaceWidget::OnSurfaceAddressChanged(qint64 new_value) { if (surface_address != new_value) { - surface_address = static_cast<Tegra::GPUVAddr>(new_value); + surface_address = static_cast<GPUVAddr>(new_value); surface_source_list->setCurrentIndex(static_cast<int>(Source::Custom)); emit Update(); diff --git a/src/yuzu/debugger/graphics/graphics_surface.h b/src/yuzu/debugger/graphics/graphics_surface.h index 323e39d94..89445b18f 100644 --- a/src/yuzu/debugger/graphics/graphics_surface.h +++ b/src/yuzu/debugger/graphics/graphics_surface.h @@ -87,7 +87,7 @@ private: QPushButton* save_surface; Source surface_source; - Tegra::GPUVAddr surface_address; + GPUVAddr surface_address; unsigned surface_width; unsigned surface_height; Tegra::Texture::TextureFormat surface_format; diff --git a/src/yuzu/debugger/wait_tree.cpp b/src/yuzu/debugger/wait_tree.cpp index 06ad74ffe..593bb681f 100644 --- a/src/yuzu/debugger/wait_tree.cpp +++ b/src/yuzu/debugger/wait_tree.cpp @@ -234,6 +234,9 @@ QString WaitTreeThread::GetText() const { case Kernel::ThreadStatus::WaitMutex: status = tr("waiting for mutex"); break; + case Kernel::ThreadStatus::WaitCondVar: + status = tr("waiting for condition variable"); + break; case Kernel::ThreadStatus::WaitArb: status = tr("waiting for address arbiter"); break; @@ -269,6 +272,7 @@ QColor WaitTreeThread::GetColor() const { case Kernel::ThreadStatus::WaitSynchAll: case Kernel::ThreadStatus::WaitSynchAny: case Kernel::ThreadStatus::WaitMutex: + case Kernel::ThreadStatus::WaitCondVar: case Kernel::ThreadStatus::WaitArb: return QColor(Qt::GlobalColor::red); case Kernel::ThreadStatus::Dormant: diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp index 32e78049c..f24cc77fe 100644 --- a/src/yuzu_cmd/config.cpp +++ b/src/yuzu_cmd/config.cpp @@ -319,7 +319,6 @@ void Config::ReadValues() { // System Settings::values.use_docked_mode = sdl2_config->GetBoolean("System", "use_docked_mode", false); - Settings::values.enable_nfc = sdl2_config->GetBoolean("System", "enable_nfc", true); const auto size = sdl2_config->GetInteger("System", "users_size", 0); Settings::values.current_user = std::clamp<int>( |