kernel/process: Hook up the process capability parser to the process itself

While we're at it, we can also toss out the leftover capability parsing
from Citra.

src/audio_core/audio_out.cpp

@@ -30,8 +30,7 @@ static Stream::Format ChannelsToStreamFormat(u32 num_channels) {
 StreamPtr AudioOut::OpenStream(u32 sample_rate, u32 num_channels, std::string&& name,
                                Stream::ReleaseCallback&& release_callback) {
     if (!sink) {
-        const SinkDetails& sink_details = GetSinkDetails(Settings::values.sink_id);
-        sink = sink_details.factory(Settings::values.audio_device_id);
+        sink = CreateSinkFromID(Settings::values.sink_id, Settings::values.audio_device_id);
     }
 
     return std::make_shared<Stream>(

src/audio_core/cubeb_sink.cpp

@@ -107,7 +107,7 @@ private:
     static void StateCallback(cubeb_stream* stream, void* user_data, cubeb_state state);
 };
 
-CubebSink::CubebSink(std::string target_device_name) {
+CubebSink::CubebSink(std::string_view target_device_name) {
     if (cubeb_init(&ctx, "yuzu", nullptr) != CUBEB_OK) {
         LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
         return;

src/audio_core/cubeb_sink.h

@@ -15,7 +15,7 @@ namespace AudioCore {
 
 class CubebSink final : public Sink {
 public:
-    explicit CubebSink(std::string device_id);
+    explicit CubebSink(std::string_view device_id);
     ~CubebSink() override;
 
     SinkStream& AcquireSinkStream(u32 sample_rate, u32 num_channels,

src/audio_core/null_sink.h

@@ -10,7 +10,7 @@ namespace AudioCore {
 
 class NullSink final : public Sink {
 public:
-    explicit NullSink(std::string){};
+    explicit NullSink(std::string_view) {}
     ~NullSink() override = default;
 
     SinkStream& AcquireSinkStream(u32 /*sample_rate*/, u32 /*num_channels*/,

src/audio_core/sink_details.cpp

@@ -14,31 +14,68 @@
 #include "common/logging/log.h"
 
 namespace AudioCore {
+namespace {
+struct SinkDetails {
+    using FactoryFn = std::unique_ptr<Sink> (*)(std::string_view);
+    using ListDevicesFn = std::vector<std::string> (*)();
 
-// g_sink_details is ordered in terms of desirability, with the best choice at the top.
-const std::vector<SinkDetails> g_sink_details = {
+    /// Name for this sink.
+    const char* id;
+    /// A method to call to construct an instance of this type of sink.
+    FactoryFn factory;
+    /// A method to call to list available devices.
+    ListDevicesFn list_devices;
+};
+
+// sink_details is ordered in terms of desirability, with the best choice at the top.
+constexpr SinkDetails sink_details[] = {
 #ifdef HAVE_CUBEB
-    SinkDetails{"cubeb", &std::make_unique<CubebSink, std::string>, &ListCubebSinkDevices},
+    SinkDetails{"cubeb",
+                [](std::string_view device_id) -> std::unique_ptr<Sink> {
+                    return std::make_unique<CubebSink>(device_id);
+                },
+                &ListCubebSinkDevices},
 #endif
-    SinkDetails{"null", &std::make_unique<NullSink, std::string>,
+    SinkDetails{"null",
+                [](std::string_view device_id) -> std::unique_ptr<Sink> {
+                    return std::make_unique<NullSink>(device_id);
+                },
                 [] { return std::vector<std::string>{"null"}; }},
 };
 
 const SinkDetails& GetSinkDetails(std::string_view sink_id) {
     auto iter =
-        std::find_if(g_sink_details.begin(), g_sink_details.end(),
+        std::find_if(std::begin(sink_details), std::end(sink_details),
                      [sink_id](const auto& sink_detail) { return sink_detail.id == sink_id; });
 
-    if (sink_id == "auto" || iter == g_sink_details.end()) {
+    if (sink_id == "auto" || iter == std::end(sink_details)) {
         if (sink_id != "auto") {
             LOG_ERROR(Audio, "AudioCore::SelectSink given invalid sink_id {}", sink_id);
         }
         // Auto-select.
-        // g_sink_details is ordered in terms of desirability, with the best choice at the front.
-        iter = g_sink_details.begin();
+        // sink_details is ordered in terms of desirability, with the best choice at the front.
+        iter = std::begin(sink_details);
     }
 
     return *iter;
 }
+} // Anonymous namespace
+
+std::vector<const char*> GetSinkIDs() {
+    std::vector<const char*> sink_ids(std::size(sink_details));
+
+    std::transform(std::begin(sink_details), std::end(sink_details), std::begin(sink_ids),
+                   [](const auto& sink) { return sink.id; });
+
+    return sink_ids;
+}
+
+std::vector<std::string> GetDeviceListForSink(std::string_view sink_id) {
+    return GetSinkDetails(sink_id).list_devices();
+}
+
+std::unique_ptr<Sink> CreateSinkFromID(std::string_view sink_id, std::string_view device_id) {
+    return GetSinkDetails(sink_id).factory(device_id);
+}
 
 } // namespace AudioCore

src/audio_core/sink_details.h

@@ -4,34 +4,21 @@
 
 #pragma once
 
-#include <functional>
-#include <memory>
 #include <string>
 #include <string_view>
-#include <utility>
 #include <vector>
 
 namespace AudioCore {
 
 class Sink;
 
-struct SinkDetails {
-    using FactoryFn = std::function<std::unique_ptr<Sink>(std::string)>;
-    using ListDevicesFn = std::function<std::vector<std::string>()>;
+/// Retrieves the IDs for all available audio sinks.
+std::vector<const char*> GetSinkIDs();
 
-    SinkDetails(const char* id_, FactoryFn factory_, ListDevicesFn list_devices_)
-        : id(id_), factory(std::move(factory_)), list_devices(std::move(list_devices_)) {}
+/// Gets the list of devices for a particular sink identified by the given ID.
+std::vector<std::string> GetDeviceListForSink(std::string_view sink_id);
 
-    /// Name for this sink.
-    const char* id;
-    /// A method to call to construct an instance of this type of sink.
-    FactoryFn factory;
-    /// A method to call to list available devices.
-    ListDevicesFn list_devices;
-};
-
-extern const std::vector<SinkDetails> g_sink_details;
-
-const SinkDetails& GetSinkDetails(std::string_view sink_id);
+/// Creates an audio sink identified by the given device ID.
+std::unique_ptr<Sink> CreateSinkFromID(std::string_view sink_id, std::string_view device_id);
 
 } // namespace AudioCore

src/common/CMakeLists.txt

@@ -44,6 +44,7 @@ add_library(common STATIC
     detached_tasks.cpp
     detached_tasks.h
     bit_field.h
+    bit_util.h
     cityhash.cpp
     cityhash.h
     color.h

src/common/bit_util.h

@@ -0,0 +1,61 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <climits>
+#include <cstddef>
+
+#ifdef _MSC_VER
+#include <intrin.h>
+#endif
+
+#include "common/common_types.h"
+
+namespace Common {
+
+/// Gets the size of a specified type T in bits.
+template <typename T>
+constexpr std::size_t BitSize() {
+    return sizeof(T) * CHAR_BIT;
+}
+
+#ifdef _MSC_VER
+inline u32 CountLeadingZeroes32(u32 value) {
+    unsigned long leading_zero = 0;
+
+    if (_BitScanReverse(&leading_zero, value) != 0) {
+        return 31 - leading_zero;
+    }
+
+    return 32;
+}
+
+inline u64 CountLeadingZeroes64(u64 value) {
+    unsigned long leading_zero = 0;
+
+    if (_BitScanReverse64(&leading_zero, value) != 0) {
+        return 63 - leading_zero;
+    }
+
+    return 64;
+}
+#else
+inline u32 CountLeadingZeroes32(u32 value) {
+    if (value == 0) {
+        return 32;
+    }
+
+    return __builtin_clz(value);
+}
+
+inline u64 CountLeadingZeroes64(u64 value) {
+    if (value == 0) {
+        return 64;
+    }
+
+    return __builtin_clzll(value);
+}
+#endif
+} // namespace Common

src/common/thread_queue_list.h

@@ -49,6 +49,22 @@ struct ThreadQueueList {
         return T();
     }
 
+    template <typename UnaryPredicate>
+    T get_first_filter(UnaryPredicate filter) const {
+        const Queue* cur = first;
+        while (cur != nullptr) {
+            if (!cur->data.empty()) {
+                for (const auto& item : cur->data) {
+                    if (filter(item))
+                        return item;
+                }
+            }
+            cur = cur->next_nonempty;
+        }
+
+        return T();
+    }
+
     T pop_first() {
         Queue* cur = first;
         while (cur != nullptr) {

src/core/CMakeLists.txt

@@ -113,6 +113,8 @@ add_library(core STATIC
     hle/kernel/object.h
     hle/kernel/process.cpp
     hle/kernel/process.h
+    hle/kernel/process_capability.cpp
+    hle/kernel/process_capability.h
     hle/kernel/readable_event.cpp
     hle/kernel/readable_event.h
     hle/kernel/resource_limit.cpp

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -151,6 +151,7 @@ std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit() const {
     config.tpidr_el0 = &cb->tpidr_el0;
     config.dczid_el0 = 4;
     config.ctr_el0 = 0x8444c004;
+    config.cntfrq_el0 = 19200000; // Value from fusee.
 
     // Unpredictable instructions
     config.define_unpredictable_behaviour = true;

src/core/file_sys/program_metadata.cpp

@@ -40,6 +40,13 @@ Loader::ResultStatus ProgramMetadata::Load(VirtualFile file) {
     if (sizeof(FileAccessHeader) != file->ReadObject(&aci_file_access, aci_header.fah_offset))
         return Loader::ResultStatus::ErrorBadFileAccessHeader;
 
+    aci_kernel_capabilities.resize(aci_header.kac_size / sizeof(u32));
+    const u64 read_size = aci_header.kac_size;
+    const u64 read_offset = npdm_header.aci_offset + aci_header.kac_offset;
+    if (file->ReadBytes(aci_kernel_capabilities.data(), read_size, read_offset) != read_size) {
+        return Loader::ResultStatus::ErrorBadKernelCapabilityDescriptors;
+    }
+
     return Loader::ResultStatus::Success;
 }
 
@@ -71,6 +78,10 @@ u64 ProgramMetadata::GetFilesystemPermissions() const {
     return aci_file_access.permissions;
 }
 
+const ProgramMetadata::KernelCapabilityDescriptors& ProgramMetadata::GetKernelCapabilities() const {
+    return aci_kernel_capabilities;
+}
+
 void ProgramMetadata::Print() const {
     LOG_DEBUG(Service_FS, "Magic:                  {:.4}", npdm_header.magic.data());
     LOG_DEBUG(Service_FS, "Main thread priority:   0x{:02X}", npdm_header.main_thread_priority);

src/core/file_sys/program_metadata.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <vector>
 #include "common/bit_field.h"
 #include "common/common_types.h"
 #include "common/swap.h"
@@ -38,6 +39,8 @@ enum class ProgramFilePermission : u64 {
  */
 class ProgramMetadata {
 public:
+    using KernelCapabilityDescriptors = std::vector<u32>;
+
     ProgramMetadata();
     ~ProgramMetadata();
 
@@ -50,6 +53,7 @@ public:
     u32 GetMainThreadStackSize() const;
     u64 GetTitleID() const;
     u64 GetFilesystemPermissions() const;
+    const KernelCapabilityDescriptors& GetKernelCapabilities() const;
 
     void Print() const;
 
@@ -154,6 +158,8 @@ private:
 
     FileAccessControl acid_file_access;
     FileAccessHeader aci_file_access;
+
+    KernelCapabilityDescriptors aci_kernel_capabilities;
 };
 
 } // namespace FileSys

src/core/hle/kernel/errors.h

@@ -11,6 +11,7 @@ namespace Kernel {
 // Confirmed Switch kernel error codes
 
 constexpr ResultCode ERR_MAX_CONNECTIONS_REACHED{ErrorModule::Kernel, 7};
+constexpr ResultCode ERR_INVALID_CAPABILITY_DESCRIPTOR{ErrorModule::Kernel, 14};
 constexpr ResultCode ERR_INVALID_SIZE{ErrorModule::Kernel, 101};
 constexpr ResultCode ERR_INVALID_ADDRESS{ErrorModule::Kernel, 102};
 constexpr ResultCode ERR_HANDLE_TABLE_FULL{ErrorModule::Kernel, 105};
@@ -30,6 +31,7 @@ constexpr ResultCode ERR_NOT_FOUND{ErrorModule::Kernel, 121};
 constexpr ResultCode ERR_ALREADY_REGISTERED{ErrorModule::Kernel, 122};
 constexpr ResultCode ERR_SESSION_CLOSED_BY_REMOTE{ErrorModule::Kernel, 123};
 constexpr ResultCode ERR_INVALID_STATE{ErrorModule::Kernel, 125};
+constexpr ResultCode ERR_RESERVED_VALUE{ErrorModule::Kernel, 126};
 constexpr ResultCode ERR_RESOURCE_LIMIT_EXCEEDED{ErrorModule::Kernel, 132};
 
 } // namespace Kernel

src/core/hle/kernel/handle_table.h

@@ -43,6 +43,9 @@ enum KernelHandle : Handle {
  */
 class HandleTable final : NonCopyable {
 public:
+    /// This is the maximum limit of handles allowed per process in Horizon
+    static constexpr std::size_t MAX_COUNT = 1024;
+
     HandleTable();
     ~HandleTable();
 
@@ -91,9 +94,6 @@ public:
     void Clear();
 
 private:
-    /// This is the maximum limit of handles allowed per process in Horizon
-    static constexpr std::size_t MAX_COUNT = 1024;
-
     /// Stores the Object referenced by the handle or null if the slot is empty.
     std::array<SharedPtr<Object>, MAX_COUNT> objects;
 

src/core/hle/kernel/process.cpp

@@ -28,13 +28,11 @@ SharedPtr<Process> Process::Create(KernelCore& kernel, std::string&& name) {
     SharedPtr<Process> process(new Process(kernel));
 
     process->name = std::move(name);
-    process->flags.raw = 0;
-    process->flags.memory_region.Assign(MemoryRegion::APPLICATION);
     process->resource_limit = kernel.GetSystemResourceLimit();
     process->status = ProcessStatus::Created;
     process->program_id = 0;
     process->process_id = kernel.CreateNewProcessID();
-    process->svc_access_mask.set();
+    process->capabilities.InitializeForMetadatalessProcess();
 
     std::mt19937 rng(Settings::values.rng_seed.value_or(0));
     std::uniform_int_distribution<u64> distribution;
@@ -64,83 +62,15 @@ ResultCode Process::ClearSignalState() {
     return RESULT_SUCCESS;
 }
 
-void Process::LoadFromMetadata(const FileSys::ProgramMetadata& metadata) {
+ResultCode Process::LoadFromMetadata(const FileSys::ProgramMetadata& metadata) {
     program_id = metadata.GetTitleID();
     ideal_processor = metadata.GetMainThreadCore();
     is_64bit_process = metadata.Is64BitProgram();
+
     vm_manager.Reset(metadata.GetAddressSpaceType());
-}
 
-void Process::ParseKernelCaps(const u32* kernel_caps, std::size_t len) {
-    for (std::size_t i = 0; i < len; ++i) {
-        u32 descriptor = kernel_caps[i];
-        u32 type = descriptor >> 20;
-
-        if (descriptor == 0xFFFFFFFF) {
-            // Unused descriptor entry
-            continue;
-        } else if ((type & 0xF00) == 0xE00) { // 0x0FFF
-            // Allowed interrupts list
-            LOG_WARNING(Loader, "ExHeader allowed interrupts list ignored");
-        } else if ((type & 0xF80) == 0xF00) { // 0x07FF
-            // Allowed syscalls mask
-            unsigned int index = ((descriptor >> 24) & 7) * 24;
-            u32 bits = descriptor & 0xFFFFFF;
-
-            while (bits && index < svc_access_mask.size()) {
-                svc_access_mask.set(index, bits & 1);
-                ++index;
-                bits >>= 1;
-            }
-        } else if ((type & 0xFF0) == 0xFE0) { // 0x00FF
-            // Handle table size
-            handle_table_size = descriptor & 0x3FF;
-        } else if ((type & 0xFF8) == 0xFF0) { // 0x007F
-            // Misc. flags
-            flags.raw = descriptor & 0xFFFF;
-        } else if ((type & 0xFFE) == 0xFF8) { // 0x001F
-            // Mapped memory range
-            if (i + 1 >= len || ((kernel_caps[i + 1] >> 20) & 0xFFE) != 0xFF8) {
-                LOG_WARNING(Loader, "Incomplete exheader memory range descriptor ignored.");
-                continue;
-            }
-            u32 end_desc = kernel_caps[i + 1];
-            ++i; // Skip over the second descriptor on the next iteration
-
-            AddressMapping mapping;
-            mapping.address = descriptor << 12;
-            VAddr end_address = end_desc << 12;
-
-            if (mapping.address < end_address) {
-                mapping.size = end_address - mapping.address;
-            } else {
-                mapping.size = 0;
-            }
-
-            mapping.read_only = (descriptor & (1 << 20)) != 0;
-            mapping.unk_flag = (end_desc & (1 << 20)) != 0;
-
-            address_mappings.push_back(mapping);
-        } else if ((type & 0xFFF) == 0xFFE) { // 0x000F
-            // Mapped memory page
-            AddressMapping mapping;
-            mapping.address = descriptor << 12;
-            mapping.size = Memory::PAGE_SIZE;
-            mapping.read_only = false;
-            mapping.unk_flag = false;
-
-            address_mappings.push_back(mapping);
-        } else if ((type & 0xFE0) == 0xFC0) { // 0x01FF
-            // Kernel version
-            kernel_version = descriptor & 0xFFFF;
-
-            int minor = kernel_version & 0xFF;
-            int major = (kernel_version >> 8) & 0xFF;
-            LOG_INFO(Loader, "ExHeader kernel version: {}.{}", major, minor);
-        } else {
-            LOG_ERROR(Loader, "Unhandled kernel caps descriptor: 0x{:08X}", descriptor);
-        }
-    }
+    const auto& caps = metadata.GetKernelCapabilities();
+    return capabilities.InitializeForUserProcess(caps.data(), caps.size(), vm_manager);
 }
 
 void Process::Run(VAddr entry_point, s32 main_thread_priority, u32 stack_size) {

src/core/hle/kernel/process.h

@@ -11,9 +11,9 @@
 #include <string>
 #include <vector>
 #include <boost/container/static_vector.hpp>
-#include "common/bit_field.h"
 #include "common/common_types.h"
 #include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/process_capability.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/kernel/vm_manager.h"
 #include "core/hle/kernel/wait_object.h"
@@ -42,24 +42,6 @@ enum class MemoryRegion : u16 {
     BASE = 3,
 };
 
-union ProcessFlags {
-    u16 raw;
-
-    BitField<0, 1, u16>
-        allow_debug; ///< Allows other processes to attach to and debug this process.
-    BitField<1, 1, u16> force_debug; ///< Allows this process to attach to processes even if they
-                                     /// don't have allow_debug set.
-    BitField<2, 1, u16> allow_nonalphanum;
-    BitField<3, 1, u16> shared_page_writable; ///< Shared page is mapped with write permissions.
-    BitField<4, 1, u16> privileged_priority;  ///< Can use priority levels higher than 24.
-    BitField<5, 1, u16> allow_main_args;
-    BitField<6, 1, u16> shared_device_mem;
-    BitField<7, 1, u16> runnable_on_sleep;
-    BitField<8, 4, MemoryRegion>
-        memory_region;                ///< Default region for memory allocations for this process
-    BitField<12, 1, u16> loaded_high; ///< Application loaded high (not at 0x00100000).
-};
-
 /**
  * Indicates the status of a Process instance.
  *
@@ -180,13 +162,13 @@ public:
     }
 
     /// Gets the bitmask of allowed CPUs that this process' threads can run on.
-    u32 GetAllowedProcessorMask() const {
-        return allowed_processor_mask;
+    u64 GetAllowedProcessorMask() const {
+        return capabilities.GetCoreMask();
     }
 
     /// Gets the bitmask of allowed thread priorities.
-    u32 GetAllowedThreadPriorityMask() const {
-        return allowed_thread_priority_mask;
+    u64 GetAllowedThreadPriorityMask() const {
+        return capabilities.GetPriorityMask();
     }
 
     u32 IsVirtualMemoryEnabled() const {
@@ -227,15 +209,12 @@ public:
      * Loads process-specifics configuration info with metadata provided
      * by an executable.
      *
-     * @param metadata The provided metadata to load process specific info.
+     * @param metadata The provided metadata to load process specific info from.
+     *
+     * @returns RESULT_SUCCESS if all relevant metadata was able to be
+     *          loaded and parsed. Otherwise, an error code is returned.
      */
-    void LoadFromMetadata(const FileSys::ProgramMetadata& metadata);
-
-    /**
-     * Parses a list of kernel capability descriptors (as found in the ExHeader) and applies them
-     * to this process.
-     */
-    void ParseKernelCaps(const u32* kernel_caps, std::size_t len);
+    ResultCode LoadFromMetadata(const FileSys::ProgramMetadata& metadata);
 
     /**
      * Applies address space changes and launches the process main thread.
@@ -296,22 +275,8 @@ private:
     /// Resource limit descriptor for this process
     SharedPtr<ResourceLimit> resource_limit;
 
-    /// The process may only call SVCs which have the corresponding bit set.
-    std::bitset<0x80> svc_access_mask;
-    /// Maximum size of the handle table for the process.
-    u32 handle_table_size = 0x200;
-    /// Special memory ranges mapped into this processes address space. This is used to give
-    /// processes access to specific I/O regions and device memory.
-    boost::container::static_vector<AddressMapping, 8> address_mappings;
-    ProcessFlags flags;
-    /// Kernel compatibility version for this process
-    u16 kernel_version = 0;
     /// The default CPU for this process, threads are scheduled on this cpu by default.
     u8 ideal_processor = 0;
-    /// Bitmask of allowed CPUs that this process' threads can run on. TODO(Subv): Actually parse
-    /// this value from the process header.
-    u32 allowed_processor_mask = THREADPROCESSORID_DEFAULT_MASK;
-    u32 allowed_thread_priority_mask = 0xFFFFFFFF;
     u32 is_virtual_address_memory_enabled = 0;
 
     /// The Thread Local Storage area is allocated as processes create threads,
@@ -321,6 +286,9 @@ private:
     /// This vector will grow as more pages are allocated for new threads.
     std::vector<std::bitset<8>> tls_slots;
 
+    /// Contains the parsed process capability descriptors.
+    ProcessCapabilities capabilities;
+
     /// Whether or not this process is AArch64, or AArch32.
     /// By default, we currently assume this is true, unless otherwise
     /// specified by metadata provided to the process during loading.

src/core/hle/kernel/process_capability.cpp

@@ -0,0 +1,355 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/bit_util.h"
+#include "core/hle/kernel/errors.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/process_capability.h"
+#include "core/hle/kernel/vm_manager.h"
+
+namespace Kernel {
+namespace {
+
+// clang-format off
+
+// Shift offsets for kernel capability types.
+enum : u32 {
+    CapabilityOffset_PriorityAndCoreNum = 3,
+    CapabilityOffset_Syscall            = 4,
+    CapabilityOffset_MapPhysical        = 6,
+    CapabilityOffset_MapIO              = 7,
+    CapabilityOffset_Interrupt          = 11,
+    CapabilityOffset_ProgramType        = 13,
+    CapabilityOffset_KernelVersion      = 14,
+    CapabilityOffset_HandleTableSize    = 15,
+    CapabilityOffset_Debug              = 16,
+};
+
+// Combined mask of all parameters that may be initialized only once.
+constexpr u32 InitializeOnceMask = (1U << CapabilityOffset_PriorityAndCoreNum) |
+                                   (1U << CapabilityOffset_ProgramType) |
+                                   (1U << CapabilityOffset_KernelVersion) |
+                                   (1U << CapabilityOffset_HandleTableSize) |
+                                   (1U << CapabilityOffset_Debug);
+
+// Packed kernel version indicating 10.4.0
+constexpr u32 PackedKernelVersion = 0x520000;
+
+// Indicates possible types of capabilities that can be specified.
+enum class CapabilityType : u32 {
+    Unset              = 0U,
+    PriorityAndCoreNum = (1U << CapabilityOffset_PriorityAndCoreNum) - 1,
+    Syscall            = (1U << CapabilityOffset_Syscall) - 1,
+    MapPhysical        = (1U << CapabilityOffset_MapPhysical) - 1,
+    MapIO              = (1U << CapabilityOffset_MapIO) - 1,
+    Interrupt          = (1U << CapabilityOffset_Interrupt) - 1,
+    ProgramType        = (1U << CapabilityOffset_ProgramType) - 1,
+    KernelVersion      = (1U << CapabilityOffset_KernelVersion) - 1,
+    HandleTableSize    = (1U << CapabilityOffset_HandleTableSize) - 1,
+    Debug              = (1U << CapabilityOffset_Debug) - 1,
+    Ignorable          = 0xFFFFFFFFU,
+};
+
+// clang-format on
+
+constexpr CapabilityType GetCapabilityType(u32 value) {
+    return static_cast<CapabilityType>((~value & (value + 1)) - 1);
+}
+
+u32 GetFlagBitOffset(CapabilityType type) {
+    const auto value = static_cast<u32>(type);
+    return static_cast<u32>(Common::BitSize<u32>() - Common::CountLeadingZeroes32(value));
+}
+
+} // Anonymous namespace
+
+ResultCode ProcessCapabilities::InitializeForKernelProcess(const u32* capabilities,
+                                                           std::size_t num_capabilities,
+                                                           VMManager& vm_manager) {
+    Clear();
+
+    // Allow all cores and priorities.
+    core_mask = 0xF;
+    priority_mask = 0xFFFFFFFFFFFFFFFF;
+    kernel_version = PackedKernelVersion;
+
+    return ParseCapabilities(capabilities, num_capabilities, vm_manager);
+}
+
+ResultCode ProcessCapabilities::InitializeForUserProcess(const u32* capabilities,
+                                                         std::size_t num_capabilities,
+                                                         VMManager& vm_manager) {
+    Clear();
+
+    return ParseCapabilities(capabilities, num_capabilities, vm_manager);
+}
+
+void ProcessCapabilities::InitializeForMetadatalessProcess() {
+    // Allow all cores and priorities
+    core_mask = 0xF;
+    priority_mask = 0xFFFFFFFFFFFFFFFF;
+    kernel_version = PackedKernelVersion;
+
+    // Allow all system calls and interrupts.
+    svc_capabilities.set();
+    interrupt_capabilities.set();
+
+    // Allow using the maximum possible amount of handles
+    handle_table_size = static_cast<u32>(HandleTable::MAX_COUNT);
+
+    // Allow all debugging capabilities.
+    is_debuggable = true;
+    can_force_debug = true;
+}
+
+ResultCode ProcessCapabilities::ParseCapabilities(const u32* capabilities,
+                                                  std::size_t num_capabilities,
+                                                  VMManager& vm_manager) {
+    u32 set_flags = 0;
+    u32 set_svc_bits = 0;
+
+    for (std::size_t i = 0; i < num_capabilities; ++i) {
+        const u32 descriptor = capabilities[i];
+        const auto type = GetCapabilityType(descriptor);
+
+        if (type == CapabilityType::MapPhysical) {
+            i++;
+
+            // The MapPhysical type uses two descriptor flags for its parameters.
+            // If there's only one, then there's a problem.
+            if (i >= num_capabilities) {
+                return ERR_INVALID_COMBINATION;
+            }
+
+            const auto size_flags = capabilities[i];
+            if (GetCapabilityType(size_flags) != CapabilityType::MapPhysical) {
+                return ERR_INVALID_COMBINATION;
+            }
+
+            const auto result = HandleMapPhysicalFlags(descriptor, size_flags, vm_manager);
+            if (result.IsError()) {
+                return result;
+            }
+        } else {
+            const auto result =
+                ParseSingleFlagCapability(set_flags, set_svc_bits, descriptor, vm_manager);
+            if (result.IsError()) {
+                return result;
+            }
+        }
+    }
+
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::ParseSingleFlagCapability(u32& set_flags, u32& set_svc_bits,
+                                                          u32 flag, VMManager& vm_manager) {
+    const auto type = GetCapabilityType(flag);
+
+    if (type == CapabilityType::Unset) {
+        return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+    }
+
+    // Bail early on ignorable entries, as one would expect,
+    // ignorable descriptors can be ignored.
+    if (type == CapabilityType::Ignorable) {
+        return RESULT_SUCCESS;
+    }
+
+    // Ensure that the give flag hasn't already been initialized before.
+    // If it has been, then bail.
+    const u32 flag_length = GetFlagBitOffset(type);
+    const u32 set_flag = 1U << flag_length;
+    if ((set_flag & set_flags & InitializeOnceMask) != 0) {
+        return ERR_INVALID_COMBINATION;
+    }
+    set_flags |= set_flag;
+
+    switch (type) {
+    case CapabilityType::PriorityAndCoreNum:
+        return HandlePriorityCoreNumFlags(flag);
+    case CapabilityType::Syscall:
+        return HandleSyscallFlags(set_svc_bits, flag);
+    case CapabilityType::MapIO:
+        return HandleMapIOFlags(flag, vm_manager);
+    case CapabilityType::Interrupt:
+        return HandleInterruptFlags(flag);
+    case CapabilityType::ProgramType:
+        return HandleProgramTypeFlags(flag);
+    case CapabilityType::KernelVersion:
+        return HandleKernelVersionFlags(flag);
+    case CapabilityType::HandleTableSize:
+        return HandleHandleTableFlags(flag);
+    case CapabilityType::Debug:
+        return HandleDebugFlags(flag);
+    default:
+        break;
+    }
+
+    return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+}
+
+void ProcessCapabilities::Clear() {
+    svc_capabilities.reset();
+    interrupt_capabilities.reset();
+
+    core_mask = 0;
+    priority_mask = 0;
+
+    handle_table_size = 0;
+    kernel_version = 0;
+
+    program_type = ProgramType::SysModule;
+
+    is_debuggable = false;
+    can_force_debug = false;
+}
+
+ResultCode ProcessCapabilities::HandlePriorityCoreNumFlags(u32 flags) {
+    if (priority_mask != 0 || core_mask != 0) {
+        return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+    }
+
+    const u32 core_num_min = (flags >> 16) & 0xFF;
+    const u32 core_num_max = (flags >> 24) & 0xFF;
+    if (core_num_min > core_num_max) {
+        return ERR_INVALID_COMBINATION;
+    }
+
+    const u32 priority_min = (flags >> 10) & 0x3F;
+    const u32 priority_max = (flags >> 4) & 0x3F;
+    if (priority_min > priority_max) {
+        return ERR_INVALID_COMBINATION;
+    }
+
+    // The switch only has 4 usable cores.
+    if (core_num_max >= 4) {
+        return ERR_INVALID_PROCESSOR_ID;
+    }
+
+    const auto make_mask = [](u64 min, u64 max) {
+        const u64 range = max - min + 1;
+        const u64 mask = (1ULL << range) - 1;
+
+        return mask << min;
+    };
+
+    core_mask = make_mask(core_num_min, core_num_max);
+    priority_mask = make_mask(priority_min, priority_max);
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleSyscallFlags(u32& set_svc_bits, u32 flags) {
+    const u32 index = flags >> 29;
+    const u32 svc_bit = 1U << index;
+
+    // If we've already set this svc before, bail.
+    if ((set_svc_bits & svc_bit) != 0) {
+        return ERR_INVALID_COMBINATION;
+    }
+    set_svc_bits |= svc_bit;
+
+    const u32 svc_mask = (flags >> 5) & 0xFFFFFF;
+    for (u32 i = 0; i < 24; ++i) {
+        const u32 svc_number = index * 24 + i;
+
+        if ((svc_mask & (1U << i)) == 0) {
+            continue;
+        }
+
+        if (svc_number >= svc_capabilities.size()) {
+            return ERR_OUT_OF_RANGE;
+        }
+
+        svc_capabilities[svc_number] = true;
+    }
+
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleMapPhysicalFlags(u32 flags, u32 size_flags,
+                                                       VMManager& vm_manager) {
+    // TODO(Lioncache): Implement once the memory manager can handle this.
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleMapIOFlags(u32 flags, VMManager& vm_manager) {
+    // TODO(Lioncache): Implement once the memory manager can handle this.
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleInterruptFlags(u32 flags) {
+    constexpr u32 interrupt_ignore_value = 0x3FF;
+    const u32 interrupt0 = (flags >> 12) & 0x3FF;
+    const u32 interrupt1 = (flags >> 22) & 0x3FF;
+
+    for (u32 interrupt : {interrupt0, interrupt1}) {
+        if (interrupt == interrupt_ignore_value) {
+            continue;
+        }
+
+        // NOTE:
+        // This should be checking a generic interrupt controller value
+        // as part of the calculation, however, given we don't currently
+        // emulate that, it's sufficient to mark every interrupt as defined.
+
+        if (interrupt >= interrupt_capabilities.size()) {
+            return ERR_OUT_OF_RANGE;
+        }
+
+        interrupt_capabilities[interrupt] = true;
+    }
+
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleProgramTypeFlags(u32 flags) {
+    const u32 reserved = flags >> 17;
+    if (reserved != 0) {
+        return ERR_RESERVED_VALUE;
+    }
+
+    program_type = static_cast<ProgramType>((flags >> 14) & 0b111);
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleKernelVersionFlags(u32 flags) {
+    // Yes, the internal member variable is checked in the actual kernel here.
+    // This might look odd for options that are only allowed to be initialized
+    // just once, however the kernel has a separate initialization function for
+    // kernel processes and userland processes. The kernel variant sets this
+    // member variable ahead of time.
+
+    const u32 major_version = kernel_version >> 19;
+
+    if (major_version != 0 || flags < 0x80000) {
+        return ERR_INVALID_CAPABILITY_DESCRIPTOR;
+    }
+
+    kernel_version = flags;
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleHandleTableFlags(u32 flags) {
+    const u32 reserved = flags >> 26;
+    if (reserved != 0) {
+        return ERR_RESERVED_VALUE;
+    }
+
+    handle_table_size = (flags >> 16) & 0x3FF;
+    return RESULT_SUCCESS;
+}
+
+ResultCode ProcessCapabilities::HandleDebugFlags(u32 flags) {
+    const u32 reserved = flags >> 19;
+    if (reserved != 0) {
+        return ERR_RESERVED_VALUE;
+    }
+
+    is_debuggable = (flags & 0x20000) != 0;
+    can_force_debug = (flags & 0x40000) != 0;
+    return RESULT_SUCCESS;
+}
+
+} // namespace Kernel

src/core/hle/kernel/process_capability.h

@@ -0,0 +1,264 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <bitset>
+
+#include "common/common_types.h"
+
+union ResultCode;
+
+namespace Kernel {
+
+class VMManager;
+
+/// The possible types of programs that may be indicated
+/// by the program type capability descriptor.
+enum class ProgramType {
+    SysModule,
+    Application,
+    Applet,
+};
+
+/// Handles kernel capability descriptors that are provided by
+/// application metadata. These descriptors provide information
+/// that alters certain parameters for kernel process instance
+/// that will run said application (or applet).
+///
+/// Capabilities are a sequence of flag descriptors, that indicate various
+/// configurations and constraints for a particular process.
+///
+/// Flag types are indicated by a sequence of set low bits. E.g. the
+/// types are indicated with the low bits as follows (where x indicates "don't care"):
+///
+/// - Priority and core mask   : 0bxxxxxxxxxxxx0111
+/// - Allowed service call mask: 0bxxxxxxxxxxx01111
+/// - Map physical memory      : 0bxxxxxxxxx0111111
+/// - Map IO memory            : 0bxxxxxxxx01111111
+/// - Interrupts               : 0bxxxx011111111111
+/// - Application type         : 0bxx01111111111111
+/// - Kernel version           : 0bx011111111111111
+/// - Handle table size        : 0b0111111111111111
+/// - Debugger flags           : 0b1111111111111111
+///
+/// These are essentially a bit offset subtracted by 1 to create a mask.
+/// e.g. The first entry in the above list is simply bit 3 (value 8 -> 0b1000)
+///      subtracted by one (7 -> 0b0111)
+///
+/// An example of a bit layout (using the map physical layout):
+/// <example>
+///   The MapPhysical type indicates a sequence entry pair of:
+///
+///   [initial, memory_flags], where:
+///
+///   initial:
+///     bits:
+///       7-24: Starting page to map memory at.
+///       25  : Indicates if the memory should be mapped as read only.
+///
+///   memory_flags:
+///     bits:
+///       7-20 : Number of pages to map
+///       21-25: Seems to be reserved (still checked against though)
+///       26   : Whether or not the memory being mapped is IO memory, or physical memory
+/// </example>
+///
+class ProcessCapabilities {
+public:
+    using InterruptCapabilities = std::bitset<1024>;
+    using SyscallCapabilities = std::bitset<128>;
+
+    ProcessCapabilities() = default;
+    ProcessCapabilities(const ProcessCapabilities&) = delete;
+    ProcessCapabilities(ProcessCapabilities&&) = default;
+
+    ProcessCapabilities& operator=(const ProcessCapabilities&) = delete;
+    ProcessCapabilities& operator=(ProcessCapabilities&&) = default;
+
+    /// Initializes this process capabilities instance for a kernel process.
+    ///
+    /// @param capabilities     The capabilities to parse
+    /// @param num_capabilities The number of capabilities to parse.
+    /// @param vm_manager       The memory manager to use for handling any mapping-related
+    ///                         operations (such as mapping IO memory, etc).
+    ///
+    /// @returns RESULT_SUCCESS if this capabilities instance was able to be initialized,
+    ///          otherwise, an error code upon failure.
+    ///
+    ResultCode InitializeForKernelProcess(const u32* capabilities, std::size_t num_capabilities,
+                                          VMManager& vm_manager);
+
+    /// Initializes this process capabilities instance for a userland process.
+    ///
+    /// @param capabilities     The capabilities to parse.
+    /// @param num_capabilities The total number of capabilities to parse.
+    /// @param vm_manager       The memory manager to use for handling any mapping-related
+    ///                         operations (such as mapping IO memory, etc).
+    ///
+    /// @returns RESULT_SUCCESS if this capabilities instance was able to be initialized,
+    ///          otherwise, an error code upon failure.
+    ///
+    ResultCode InitializeForUserProcess(const u32* capabilities, std::size_t num_capabilities,
+                                        VMManager& vm_manager);
+
+    /// Initializes this process capabilities instance for a process that does not
+    /// have any metadata to parse.
+    ///
+    /// This is necessary, as we allow running raw executables, and the internal
+    /// kernel process capabilities also determine what CPU cores the process is
+    /// allowed to run on, and what priorities are allowed for  threads. It also
+    /// determines the max handle table size, what the program type is, whether or
+    /// not the process can be debugged, or whether it's possible for a process to
+    /// forcibly debug another process.
+    ///
+    /// Given the above, this essentially enables all capabilities across the board
+    /// for the process. It allows the process to:
+    ///
+    /// - Run on any core
+    /// - Use any thread priority
+    /// - Use the maximum amount of handles a process is allowed to.
+    /// - Be debuggable
+    /// - Forcibly debug other processes.
+    ///
+    /// Note that this is not a behavior that the kernel allows a process to do via
+    /// a single function like this. This is yuzu-specific behavior to handle
+    /// executables with no capability descriptors whatsoever to derive behavior from.
+    /// It being yuzu-specific is why this is also not the default behavior and not
+    /// done by default in the constructor.
+    ///
+    void InitializeForMetadatalessProcess();
+
+    /// Gets the allowable core mask
+    u64 GetCoreMask() const {
+        return core_mask;
+    }
+
+    /// Gets the allowable priority mask
+    u64 GetPriorityMask() const {
+        return priority_mask;
+    }
+
+    /// Gets the SVC access permission bits
+    const SyscallCapabilities& GetServiceCapabilities() const {
+        return svc_capabilities;
+    }
+
+    /// Gets the valid interrupt bits.
+    const InterruptCapabilities& GetInterruptCapabilities() const {
+        return interrupt_capabilities;
+    }
+
+    /// Gets the program type for this process.
+    ProgramType GetProgramType() const {
+        return program_type;
+    }
+
+    /// Gets the number of total allowable handles for the process' handle table.
+    u32 GetHandleTableSize() const {
+        return handle_table_size;
+    }
+
+    /// Gets the kernel version value.
+    u32 GetKernelVersion() const {
+        return kernel_version;
+    }
+
+    /// Whether or not this process can be debugged.
+    bool IsDebuggable() const {
+        return is_debuggable;
+    }
+
+    /// Whether or not this process can forcibly debug another
+    /// process, even if that process is not considered debuggable.
+    bool CanForceDebug() const {
+        return can_force_debug;
+    }
+
+private:
+    /// Attempts to parse a given sequence of capability descriptors.
+    ///
+    /// @param capabilities     The sequence of capability descriptors to parse.
+    /// @param num_capabilities The number of descriptors within the given sequence.
+    /// @param vm_manager       The memory manager that will perform any memory
+    ///                         mapping if necessary.
+    ///
+    /// @return RESULT_SUCCESS if no errors occur, otherwise an error code.
+    ///
+    ResultCode ParseCapabilities(const u32* capabilities, std::size_t num_capabilities,
+                                 VMManager& vm_manager);
+
+    /// Attempts to parse a capability descriptor that is only represented by a
+    /// single flag set.
+    ///
+    /// @param set_flags    Running set of flags that are used to catch
+    ///                     flags being initialized more than once when they shouldn't be.
+    /// @param set_svc_bits Running set of bits representing the allowed supervisor calls mask.
+    /// @param flag         The flag to attempt to parse.
+    /// @param vm_manager   The memory manager that will perform any memory
+    ///                     mapping if necessary.
+    ///
+    /// @return RESULT_SUCCESS if no errors occurred, otherwise an error code.
+    ///
+    ResultCode ParseSingleFlagCapability(u32& set_flags, u32& set_svc_bits, u32 flag,
+                                         VMManager& vm_manager);
+
+    /// Clears the internal state of this process capability instance. Necessary,
+    /// to have a sane starting point due to us allowing running executables without
+    /// configuration metadata. We assume a process is not going to have metadata,
+    /// and if it turns out that the process does, in fact, have metadata, then
+    /// we attempt to parse it. Thus, we need this to reset data members back to
+    /// a good state.
+    ///
+    /// DO NOT ever make this a public member function. This isn't an invariant
+    /// anything external should depend upon (and if anything comes to rely on it,
+    /// you should immediately be questioning the design of that thing, not this
+    /// class. If the kernel itself can run without depending on behavior like that,
+    /// then so can yuzu).
+    ///
+    void Clear();
+
+    /// Handles flags related to the priority and core number capability flags.
+    ResultCode HandlePriorityCoreNumFlags(u32 flags);
+
+    /// Handles flags related to determining the allowable SVC mask.
+    ResultCode HandleSyscallFlags(u32& set_svc_bits, u32 flags);
+
+    /// Handles flags related to mapping physical memory pages.
+    ResultCode HandleMapPhysicalFlags(u32 flags, u32 size_flags, VMManager& vm_manager);
+
+    /// Handles flags related to mapping IO pages.
+    ResultCode HandleMapIOFlags(u32 flags, VMManager& vm_manager);
+
+    /// Handles flags related to the interrupt capability flags.
+    ResultCode HandleInterruptFlags(u32 flags);
+
+    /// Handles flags related to the program type.
+    ResultCode HandleProgramTypeFlags(u32 flags);
+
+    /// Handles flags related to the handle table size.
+    ResultCode HandleHandleTableFlags(u32 flags);
+
+    /// Handles flags related to the kernel version capability flags.
+    ResultCode HandleKernelVersionFlags(u32 flags);
+
+    /// Handles flags related to debug-specific capabilities.
+    ResultCode HandleDebugFlags(u32 flags);
+
+    SyscallCapabilities svc_capabilities;
+    InterruptCapabilities interrupt_capabilities;
+
+    u64 core_mask = 0;
+    u64 priority_mask = 0;
+
+    u32 handle_table_size = 0;
+    u32 kernel_version = 0;
+
+    ProgramType program_type = ProgramType::SysModule;
+
+    bool is_debuggable = false;
+    bool can_force_debug = false;
+};
+
+} // namespace Kernel

src/core/hle/kernel/scheduler.cpp

@@ -9,6 +9,7 @@
 #include "common/logging/log.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
+#include "core/core_cpu.h"
 #include "core/core_timing.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
@@ -179,4 +180,69 @@ void Scheduler::SetThreadPriority(Thread* thread, u32 priority) {
         ready_queue.prepare(priority);
 }
 
+Thread* Scheduler::GetNextSuggestedThread(u32 core, u32 maximum_priority) const {
+    std::lock_guard<std::mutex> 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;
+    });
+}
+
+void Scheduler::YieldWithoutLoadBalancing(Thread* thread) {
+    ASSERT(thread != nullptr);
+    // Avoid yielding if the thread isn't even running.
+    ASSERT(thread->GetStatus() == ThreadStatus::Running);
+
+    // Sanity check that the priority is valid
+    ASSERT(thread->GetPriority() < THREADPRIO_COUNT);
+
+    // Yield this thread -- sleep for zero time and force reschedule to different thread
+    WaitCurrentThread_Sleep();
+    GetCurrentThread()->WakeAfterDelay(0);
+}
+
+void Scheduler::YieldWithLoadBalancing(Thread* thread) {
+    ASSERT(thread != nullptr);
+    const auto priority = thread->GetPriority();
+    const auto core = static_cast<u32>(thread->GetProcessorID());
+
+    // Avoid yielding if the thread isn't even running.
+    ASSERT(thread->GetStatus() == ThreadStatus::Running);
+
+    // Sanity check that the priority is valid
+    ASSERT(priority < THREADPRIO_COUNT);
+
+    // Sleep for zero time to be able to force reschedule to different thread
+    WaitCurrentThread_Sleep();
+    GetCurrentThread()->WakeAfterDelay(0);
+
+    Thread* suggested_thread = nullptr;
+
+    // Search through all of the cpu cores (except this one) for a suggested thread.
+    // Take the first non-nullptr one
+    for (unsigned cur_core = 0; cur_core < Core::NUM_CPU_CORES; ++cur_core) {
+        const auto res =
+            Core::System::GetInstance().CpuCore(cur_core).Scheduler().GetNextSuggestedThread(
+                core, priority);
+
+        // If scheduler provides a suggested thread
+        if (res != nullptr) {
+            // And its better than the current suggested thread (or is the first valid one)
+            if (suggested_thread == nullptr ||
+                suggested_thread->GetPriority() > res->GetPriority()) {
+                suggested_thread = res;
+            }
+        }
+    }
+
+    // If a suggested thread was found, queue that for this core
+    if (suggested_thread != nullptr)
+        suggested_thread->ChangeCore(core, suggested_thread->GetAffinityMask());
+}
+
+void Scheduler::YieldAndWaitForLoadBalancing(Thread* thread) {
+    UNIMPLEMENTED_MSG("Wait for load balancing thread yield type is not implemented!");
+}
+
 } // namespace Kernel

src/core/hle/kernel/scheduler.h

@@ -51,6 +51,75 @@ public:
     /// Sets the priority of a thread in the scheduler
     void SetThreadPriority(Thread* thread, u32 priority);
 
+    /// Gets the next suggested thread for load balancing
+    Thread* GetNextSuggestedThread(u32 core, u32 minimum_priority) const;
+
+    /**
+     * YieldWithoutLoadBalancing -- analogous to normal yield on a system
+     * Moves the thread to the end of the ready queue for its priority, and then reschedules the
+     * system to the new head of the queue.
+     *
+     * Example (Single Core -- but can be extrapolated to multi):
+     * ready_queue[prio=0]: ThreadA, ThreadB, ThreadC (->exec order->)
+     * Currently Running: ThreadR
+     *
+     * ThreadR calls YieldWithoutLoadBalancing
+     *
+     * ThreadR is moved to the end of ready_queue[prio=0]:
+     * ready_queue[prio=0]: ThreadA, ThreadB, ThreadC, ThreadR (->exec order->)
+     * Currently Running: Nothing
+     *
+     * System is rescheduled (ThreadA is popped off of queue):
+     * ready_queue[prio=0]: ThreadB, ThreadC, ThreadR (->exec order->)
+     * Currently Running: ThreadA
+     *
+     * If the queue is empty at time of call, no yielding occurs. This does not cross between cores
+     * or priorities at all.
+     */
+    void YieldWithoutLoadBalancing(Thread* thread);
+
+    /**
+     * YieldWithLoadBalancing -- yield but with better selection of the new running thread
+     * Moves the current thread to the end of the ready queue for its priority, then selects a
+     * 'suggested thread' (a thread on a different core that could run on this core) from the
+     * scheduler, changes its core, and reschedules the current core to that thread.
+     *
+     * Example (Dual Core -- can be extrapolated to Quad Core, this is just normal yield if it were
+     * single core):
+     * ready_queue[core=0][prio=0]: ThreadA, ThreadB (affinities not pictured as irrelevant
+     * ready_queue[core=1][prio=0]: ThreadC[affinity=both], ThreadD[affinity=core1only]
+     * Currently Running: ThreadQ on Core 0 || ThreadP on Core 1
+     *
+     * ThreadQ calls YieldWithLoadBalancing
+     *
+     * ThreadQ is moved to the end of ready_queue[core=0][prio=0]:
+     * ready_queue[core=0][prio=0]: ThreadA, ThreadB
+     * ready_queue[core=1][prio=0]: ThreadC[affinity=both], ThreadD[affinity=core1only]
+     * Currently Running: ThreadQ on Core 0 || ThreadP on Core 1
+     *
+     * A list of suggested threads for each core is compiled
+     * Suggested Threads: {ThreadC on Core 1}
+     * If this were quad core (as the switch is), there could be between 0 and 3 threads in this
+     * list. If there are more than one, the thread is selected by highest prio.
+     *
+     * ThreadC is core changed to Core 0:
+     * ready_queue[core=0][prio=0]: ThreadC, ThreadA, ThreadB, ThreadQ
+     * ready_queue[core=1][prio=0]: ThreadD
+     * Currently Running: None on Core 0 || ThreadP on Core 1
+     *
+     * System is rescheduled (ThreadC is popped off of queue):
+     * ready_queue[core=0][prio=0]: ThreadA, ThreadB, ThreadQ
+     * ready_queue[core=1][prio=0]: ThreadD
+     * Currently Running: ThreadC on Core 0 || ThreadP on Core 1
+     *
+     * If no suggested threads can be found this will behave just as normal yield. If there are
+     * multiple candidates for the suggested thread on a core, the highest prio is taken.
+     */
+    void YieldWithLoadBalancing(Thread* thread);
+
+    /// Currently unknown -- asserts as unimplemented on call
+    void YieldAndWaitForLoadBalancing(Thread* thread);
+
     /// Returns a list of all threads managed by the scheduler
     const std::vector<SharedPtr<Thread>>& GetThreadList() const {
         return thread_list;

src/core/hle/kernel/svc.cpp

@@ -254,11 +254,52 @@ static ResultCode SetMemoryPermission(VAddr addr, u64 size, u32 prot) {
     return vm_manager.ReprotectRange(addr, size, converted_permissions);
 }
 
-static ResultCode SetMemoryAttribute(VAddr addr, u64 size, u32 state0, u32 state1) {
-    LOG_WARNING(Kernel_SVC,
-                "(STUBBED) called, addr=0x{:X}, size=0x{:X}, state0=0x{:X}, state1=0x{:X}", addr,
-                size, state0, state1);
-    return RESULT_SUCCESS;
+static ResultCode SetMemoryAttribute(VAddr address, u64 size, u32 mask, u32 attribute) {
+    LOG_DEBUG(Kernel_SVC,
+              "called, address=0x{:016X}, size=0x{:X}, mask=0x{:08X}, attribute=0x{:08X}", address,
+              size, mask, attribute);
+
+    if (!Common::Is4KBAligned(address)) {
+        LOG_ERROR(Kernel_SVC, "Address not page aligned (0x{:016X})", address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (size == 0 || !Common::Is4KBAligned(size)) {
+        LOG_ERROR(Kernel_SVC, "Invalid size (0x{:X}). Size must be non-zero and page aligned.",
+                  size);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (!IsValidAddressRange(address, size)) {
+        LOG_ERROR(Kernel_SVC, "Address range overflowed (Address: 0x{:016X}, Size: 0x{:016X})",
+                  address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    const auto mem_attribute = static_cast<MemoryAttribute>(attribute);
+    const auto mem_mask = static_cast<MemoryAttribute>(mask);
+    const auto attribute_with_mask = mem_attribute | mem_mask;
+
+    if (attribute_with_mask != mem_mask) {
+        LOG_ERROR(Kernel_SVC,
+                  "Memory attribute doesn't match the given mask (Attribute: 0x{:X}, Mask: {:X}",
+                  attribute, mask);
+        return ERR_INVALID_COMBINATION;
+    }
+
+    if ((attribute_with_mask | MemoryAttribute::Uncached) != MemoryAttribute::Uncached) {
+        LOG_ERROR(Kernel_SVC, "Specified attribute isn't equal to MemoryAttributeUncached (8).");
+        return ERR_INVALID_COMBINATION;
+    }
+
+    auto& vm_manager = Core::CurrentProcess()->VMManager();
+    if (!IsInsideAddressSpace(vm_manager, address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Given address (0x{:016X}) is outside the bounds of the address space.", address);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    return vm_manager.SetMemoryAttribute(address, size, mem_mask, mem_attribute);
 }
 
 /// Maps a memory range into a different range.
@@ -1208,18 +1249,38 @@ static void ExitThread() {
 static void SleepThread(s64 nanoseconds) {
     LOG_TRACE(Kernel_SVC, "called nanoseconds={}", nanoseconds);
 
-    // Don't attempt to yield execution if there are no available threads to run,
-    // this way we avoid a useless reschedule to the idle thread.
-    if (nanoseconds == 0 && !Core::System::GetInstance().CurrentScheduler().HaveReadyThreads())
-        return;
+    enum class SleepType : s64 {
+        YieldWithoutLoadBalancing = 0,
+        YieldWithLoadBalancing = -1,
+        YieldAndWaitForLoadBalancing = -2,
+    };
 
-    // Sleep current thread and check for next thread to schedule
-    WaitCurrentThread_Sleep();
+    if (nanoseconds <= 0) {
+        auto& scheduler{Core::System::GetInstance().CurrentScheduler()};
+        switch (static_cast<SleepType>(nanoseconds)) {
+        case SleepType::YieldWithoutLoadBalancing:
+            scheduler.YieldWithoutLoadBalancing(GetCurrentThread());
+            break;
+        case SleepType::YieldWithLoadBalancing:
+            scheduler.YieldWithLoadBalancing(GetCurrentThread());
+            break;
+        case SleepType::YieldAndWaitForLoadBalancing:
+            scheduler.YieldAndWaitForLoadBalancing(GetCurrentThread());
+            break;
+        default:
+            UNREACHABLE_MSG("Unimplemented sleep yield type '{:016X}'!", nanoseconds);
+        }
+    } else {
+        // Sleep current thread and check for next thread to schedule
+        WaitCurrentThread_Sleep();
 
-    // Create an event to wake the thread up after the specified nanosecond delay has passed
-    GetCurrentThread()->WakeAfterDelay(nanoseconds);
+        // Create an event to wake the thread up after the specified nanosecond delay has passed
+        GetCurrentThread()->WakeAfterDelay(nanoseconds);
+    }
 
-    Core::System::GetInstance().PrepareReschedule();
+    // Reschedule all CPU cores
+    for (std::size_t i = 0; i < Core::NUM_CPU_CORES; ++i)
+        Core::System::GetInstance().CpuCore(i).PrepareReschedule();
 }
 
 /// Wait process wide key atomic

src/core/hle/kernel/thread.cpp

@@ -158,6 +158,9 @@ static void ResetThreadContext(Core::ARM_Interface::ThreadContext& context, VAdd
     context.cpu_registers[0] = arg;
     context.pc = entry_point;
     context.sp = stack_top;
+    // TODO(merry): Perform a hardware test to determine the below value.
+    // AHP = 0, DN = 1, FTZ = 1, RMode = Round towards zero
+    context.fpcr = 0x03C00000;
 }
 
 ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name, VAddr entry_point,

src/core/hle/kernel/thread.h

@@ -26,6 +26,7 @@ enum ThreadPriority : u32 {
     THREADPRIO_USERLAND_MAX = 24, ///< Highest thread priority for userland apps
     THREADPRIO_DEFAULT = 44,      ///< Default thread priority for userland apps
     THREADPRIO_LOWEST = 63,       ///< Lowest thread priority
+    THREADPRIO_COUNT = 64,        ///< Total number of possible thread priorities.
 };
 
 enum ThreadProcessorId : s32 {

src/core/hle/kernel/vm_manager.cpp

@@ -37,7 +37,7 @@ static const char* GetMemoryStateName(MemoryState state) {
 
 bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const {
     ASSERT(base + size == next.base);
-    if (permissions != next.permissions || meminfo_state != next.meminfo_state ||
+    if (permissions != next.permissions || state != next.state || attribute != next.attribute ||
         type != next.type) {
         return false;
     }
@@ -115,7 +115,7 @@ ResultVal<VMManager::VMAHandle> VMManager::MapMemoryBlock(VAddr target,
 
     final_vma.type = VMAType::AllocatedMemoryBlock;
     final_vma.permissions = VMAPermission::ReadWrite;
-    final_vma.meminfo_state = state;
+    final_vma.state = state;
     final_vma.backing_block = std::move(block);
     final_vma.offset = offset;
     UpdatePageTableForVMA(final_vma);
@@ -140,7 +140,7 @@ ResultVal<VMManager::VMAHandle> VMManager::MapBackingMemory(VAddr target, u8* me
 
     final_vma.type = VMAType::BackingMemory;
     final_vma.permissions = VMAPermission::ReadWrite;
-    final_vma.meminfo_state = state;
+    final_vma.state = state;
     final_vma.backing_memory = memory;
     UpdatePageTableForVMA(final_vma);
 
@@ -177,7 +177,7 @@ ResultVal<VMManager::VMAHandle> VMManager::MapMMIO(VAddr target, PAddr paddr, u6
 
     final_vma.type = VMAType::MMIO;
     final_vma.permissions = VMAPermission::ReadWrite;
-    final_vma.meminfo_state = state;
+    final_vma.state = state;
     final_vma.paddr = paddr;
     final_vma.mmio_handler = std::move(mmio_handler);
     UpdatePageTableForVMA(final_vma);
@@ -189,7 +189,7 @@ VMManager::VMAIter VMManager::Unmap(VMAIter vma_handle) {
     VirtualMemoryArea& vma = vma_handle->second;
     vma.type = VMAType::Free;
     vma.permissions = VMAPermission::None;
-    vma.meminfo_state = MemoryState::Unmapped;
+    vma.state = MemoryState::Unmapped;
 
     vma.backing_block = nullptr;
     vma.offset = 0;
@@ -308,9 +308,10 @@ MemoryInfo VMManager::QueryMemory(VAddr address) const {
 
     if (IsValidHandle(vma)) {
         memory_info.base_address = vma->second.base;
+        memory_info.attributes = ToSvcMemoryAttribute(vma->second.attribute);
         memory_info.permission = static_cast<u32>(vma->second.permissions);
         memory_info.size = vma->second.size;
-        memory_info.state = ToSvcMemoryState(vma->second.meminfo_state);
+        memory_info.state = ToSvcMemoryState(vma->second.state);
     } else {
         memory_info.base_address = address_space_end;
         memory_info.permission = static_cast<u32>(VMAPermission::None);
@@ -321,6 +322,34 @@ MemoryInfo VMManager::QueryMemory(VAddr address) const {
     return memory_info;
 }
 
+ResultCode VMManager::SetMemoryAttribute(VAddr address, u64 size, MemoryAttribute mask,
+                                         MemoryAttribute attribute) {
+    constexpr auto ignore_mask = MemoryAttribute::Uncached | MemoryAttribute::DeviceMapped;
+    constexpr auto attribute_mask = ~ignore_mask;
+
+    const auto result = CheckRangeState(
+        address, size, MemoryState::FlagUncached, MemoryState::FlagUncached, VMAPermission::None,
+        VMAPermission::None, attribute_mask, MemoryAttribute::None, ignore_mask);
+
+    if (result.Failed()) {
+        return result.Code();
+    }
+
+    const auto [prev_state, prev_permissions, prev_attributes] = *result;
+    const auto new_attribute = (prev_attributes & ~mask) | (mask & attribute);
+
+    const auto carve_result = CarveVMARange(address, size);
+    if (carve_result.Failed()) {
+        return carve_result.Code();
+    }
+
+    auto vma_iter = *carve_result;
+    vma_iter->second.attribute = new_attribute;
+
+    MergeAdjacent(vma_iter);
+    return RESULT_SUCCESS;
+}
+
 ResultCode VMManager::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size, MemoryState state) {
     const auto vma = FindVMA(src_addr);
 
@@ -364,7 +393,7 @@ void VMManager::LogLayout() const {
                   (u8)vma.permissions & (u8)VMAPermission::Read ? 'R' : '-',
                   (u8)vma.permissions & (u8)VMAPermission::Write ? 'W' : '-',
                   (u8)vma.permissions & (u8)VMAPermission::Execute ? 'X' : '-',
-                  GetMemoryStateName(vma.meminfo_state));
+                  GetMemoryStateName(vma.state));
     }
 }
 
@@ -591,6 +620,66 @@ void VMManager::ClearPageTable() {
               Memory::PageType::Unmapped);
 }
 
+VMManager::CheckResults VMManager::CheckRangeState(VAddr address, u64 size, MemoryState state_mask,
+                                                   MemoryState state, VMAPermission permission_mask,
+                                                   VMAPermission permissions,
+                                                   MemoryAttribute attribute_mask,
+                                                   MemoryAttribute attribute,
+                                                   MemoryAttribute ignore_mask) const {
+    auto iter = FindVMA(address);
+
+    // If we don't have a valid VMA handle at this point, then it means this is
+    // being called with an address outside of the address space, which is definitely
+    // indicative of a bug, as this function only operates on mapped memory regions.
+    DEBUG_ASSERT(IsValidHandle(iter));
+
+    const VAddr end_address = address + size - 1;
+    const MemoryAttribute initial_attributes = iter->second.attribute;
+    const VMAPermission initial_permissions = iter->second.permissions;
+    const MemoryState initial_state = iter->second.state;
+
+    while (true) {
+        // The iterator should be valid throughout the traversal. Hitting the end of
+        // the mapped VMA regions is unquestionably indicative of a bug.
+        DEBUG_ASSERT(IsValidHandle(iter));
+
+        const auto& vma = iter->second;
+
+        if (vma.state != initial_state) {
+            return ERR_INVALID_ADDRESS_STATE;
+        }
+
+        if ((vma.state & state_mask) != state) {
+            return ERR_INVALID_ADDRESS_STATE;
+        }
+
+        if (vma.permissions != initial_permissions) {
+            return ERR_INVALID_ADDRESS_STATE;
+        }
+
+        if ((vma.permissions & permission_mask) != permissions) {
+            return ERR_INVALID_ADDRESS_STATE;
+        }
+
+        if ((vma.attribute | ignore_mask) != (initial_attributes | ignore_mask)) {
+            return ERR_INVALID_ADDRESS_STATE;
+        }
+
+        if ((vma.attribute & attribute_mask) != attribute) {
+            return ERR_INVALID_ADDRESS_STATE;
+        }
+
+        if (end_address <= vma.EndAddress()) {
+            break;
+        }
+
+        ++iter;
+    }
+
+    return MakeResult(
+        std::make_tuple(initial_state, initial_permissions, initial_attributes & ~ignore_mask));
+}
+
 u64 VMManager::GetTotalMemoryUsage() const {
     LOG_WARNING(Kernel, "(STUBBED) called");
     return 0xF8000000;

src/core/hle/kernel/vm_manager.h

@@ -6,6 +6,7 @@
 
 #include <map>
 #include <memory>
+#include <tuple>
 #include <vector>
 #include "common/common_types.h"
 #include "core/hle/result.h"
@@ -43,6 +44,88 @@ enum class VMAPermission : u8 {
     ReadWriteExecute = Read | Write | Execute,
 };
 
+constexpr VMAPermission operator|(VMAPermission lhs, VMAPermission rhs) {
+    return static_cast<VMAPermission>(u32(lhs) | u32(rhs));
+}
+
+constexpr VMAPermission operator&(VMAPermission lhs, VMAPermission rhs) {
+    return static_cast<VMAPermission>(u32(lhs) & u32(rhs));
+}
+
+constexpr VMAPermission operator^(VMAPermission lhs, VMAPermission rhs) {
+    return static_cast<VMAPermission>(u32(lhs) ^ u32(rhs));
+}
+
+constexpr VMAPermission operator~(VMAPermission permission) {
+    return static_cast<VMAPermission>(~u32(permission));
+}
+
+constexpr VMAPermission& operator|=(VMAPermission& lhs, VMAPermission rhs) {
+    lhs = lhs | rhs;
+    return lhs;
+}
+
+constexpr VMAPermission& operator&=(VMAPermission& lhs, VMAPermission rhs) {
+    lhs = lhs & rhs;
+    return lhs;
+}
+
+constexpr VMAPermission& operator^=(VMAPermission& lhs, VMAPermission rhs) {
+    lhs = lhs ^ rhs;
+    return lhs;
+}
+
+/// Attribute flags that can be applied to a VMA
+enum class MemoryAttribute : u32 {
+    Mask = 0xFF,
+
+    /// No particular qualities
+    None = 0,
+    /// Memory locked/borrowed for use. e.g. This would be used by transfer memory.
+    Locked = 1,
+    /// Memory locked for use by IPC-related internals.
+    LockedForIPC = 2,
+    /// Mapped as part of the device address space.
+    DeviceMapped = 4,
+    /// Uncached memory
+    Uncached = 8,
+};
+
+constexpr MemoryAttribute operator|(MemoryAttribute lhs, MemoryAttribute rhs) {
+    return static_cast<MemoryAttribute>(u32(lhs) | u32(rhs));
+}
+
+constexpr MemoryAttribute operator&(MemoryAttribute lhs, MemoryAttribute rhs) {
+    return static_cast<MemoryAttribute>(u32(lhs) & u32(rhs));
+}
+
+constexpr MemoryAttribute operator^(MemoryAttribute lhs, MemoryAttribute rhs) {
+    return static_cast<MemoryAttribute>(u32(lhs) ^ u32(rhs));
+}
+
+constexpr MemoryAttribute operator~(MemoryAttribute attribute) {
+    return static_cast<MemoryAttribute>(~u32(attribute));
+}
+
+constexpr MemoryAttribute& operator|=(MemoryAttribute& lhs, MemoryAttribute rhs) {
+    lhs = lhs | rhs;
+    return lhs;
+}
+
+constexpr MemoryAttribute& operator&=(MemoryAttribute& lhs, MemoryAttribute rhs) {
+    lhs = lhs & rhs;
+    return lhs;
+}
+
+constexpr MemoryAttribute& operator^=(MemoryAttribute& lhs, MemoryAttribute rhs) {
+    lhs = lhs ^ rhs;
+    return lhs;
+}
+
+constexpr u32 ToSvcMemoryAttribute(MemoryAttribute attribute) {
+    return static_cast<u32>(attribute & MemoryAttribute::Mask);
+}
+
 // clang-format off
 /// Represents memory states and any relevant flags, as used by the kernel.
 /// svcQueryMemory interprets these by masking away all but the first eight
@@ -174,6 +257,16 @@ struct PageInfo {
  * also backed by a single host memory allocation.
  */
 struct VirtualMemoryArea {
+    /// Gets the starting (base) address of this VMA.
+    VAddr StartAddress() const {
+        return base;
+    }
+
+    /// Gets the ending address of this VMA.
+    VAddr EndAddress() const {
+        return base + size - 1;
+    }
+
     /// Virtual base address of the region.
     VAddr base = 0;
     /// Size of the region.
@@ -181,8 +274,8 @@ struct VirtualMemoryArea {
 
     VMAType type = VMAType::Free;
     VMAPermission permissions = VMAPermission::None;
-    /// Tag returned by svcQueryMemory. Not otherwise used.
-    MemoryState meminfo_state = MemoryState::Unmapped;
+    MemoryState state = MemoryState::Unmapped;
+    MemoryAttribute attribute = MemoryAttribute::None;
 
     // Settings for type = AllocatedMemoryBlock
     /// Memory block backing this VMA.
@@ -299,6 +392,19 @@ public:
     ///
     MemoryInfo QueryMemory(VAddr address) const;
 
+    /// Sets an attribute across the given address range.
+    ///
+    /// @param address   The starting address
+    /// @param size      The size of the range to set the attribute on.
+    /// @param mask      The attribute mask
+    /// @param attribute The attribute to set across the given address range
+    ///
+    /// @returns RESULT_SUCCESS if successful
+    /// @returns ERR_INVALID_ADDRESS_STATE if the attribute could not be set.
+    ///
+    ResultCode SetMemoryAttribute(VAddr address, u64 size, MemoryAttribute mask,
+                                  MemoryAttribute attribute);
+
     /**
      * Scans all VMAs and updates the page table range of any that use the given vector as backing
      * memory. This should be called after any operation that causes reallocation of the vector.
@@ -435,6 +541,35 @@ private:
     /// Clears out the page table
     void ClearPageTable();
 
+    using CheckResults = ResultVal<std::tuple<MemoryState, VMAPermission, MemoryAttribute>>;
+
+    /// Checks if an address range adheres to the specified states provided.
+    ///
+    /// @param address         The starting address of the address range.
+    /// @param size            The size of the address range.
+    /// @param state_mask      The memory state mask.
+    /// @param state           The state to compare the individual VMA states against,
+    ///                        which is done in the form of: (vma.state & state_mask) != state.
+    /// @param permission_mask The memory permissions mask.
+    /// @param permissions     The permission to compare the individual VMA permissions against,
+    ///                        which is done in the form of:
+    ///                        (vma.permission & permission_mask) != permission.
+    /// @param attribute_mask  The memory attribute mask.
+    /// @param attribute       The memory attributes to compare the individual VMA attributes
+    ///                        against, which is done in the form of:
+    ///                        (vma.attributes & attribute_mask) != attribute.
+    /// @param ignore_mask     The memory attributes to ignore during the check.
+    ///
+    /// @returns If successful, returns a tuple containing the memory attributes
+    ///          (with ignored bits specified by ignore_mask unset), memory permissions, and
+    ///          memory state across the memory range.
+    /// @returns If not successful, returns ERR_INVALID_ADDRESS_STATE.
+    ///
+    CheckResults CheckRangeState(VAddr address, u64 size, MemoryState state_mask, MemoryState state,
+                                 VMAPermission permission_mask, VMAPermission permissions,
+                                 MemoryAttribute attribute_mask, MemoryAttribute attribute,
+                                 MemoryAttribute ignore_mask) const;
+
     /**
      * A map covering the entirety of the managed address space, keyed by the `base` field of each
      * VMA. It must always be modified by splitting or merging VMAs, so that the invariant

src/core/hle/service/am/am.cpp

@@ -565,7 +565,6 @@ private:
     void GetAppletStateChangedEvent(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_AM, "called");
 
-        applet->GetBroker().SignalStateChanged();
         const auto event = applet->GetBroker().GetStateChangedEvent();
 
         IPC::ResponseBuilder rb{ctx, 2, 1};

src/core/hle/service/am/applets/applets.cpp

@@ -16,11 +16,11 @@ namespace Service::AM::Applets {
 AppletDataBroker::AppletDataBroker() {
     auto& kernel = Core::System::GetInstance().Kernel();
     state_changed_event = Kernel::WritableEvent::CreateEventPair(
-        kernel, Kernel::ResetType::OneShot, "ILibraryAppletAccessor:StateChangedEvent");
+        kernel, Kernel::ResetType::Sticky, "ILibraryAppletAccessor:StateChangedEvent");
     pop_out_data_event = Kernel::WritableEvent::CreateEventPair(
-        kernel, Kernel::ResetType::OneShot, "ILibraryAppletAccessor:PopDataOutEvent");
+        kernel, Kernel::ResetType::Sticky, "ILibraryAppletAccessor:PopDataOutEvent");
     pop_interactive_out_data_event = Kernel::WritableEvent::CreateEventPair(
-        kernel, Kernel::ResetType::OneShot, "ILibraryAppletAccessor:PopInteractiveDataOutEvent");
+        kernel, Kernel::ResetType::Sticky, "ILibraryAppletAccessor:PopInteractiveDataOutEvent");
 }
 
 AppletDataBroker::~AppletDataBroker() = default;

src/core/hle/service/am/applets/software_keyboard.cpp

@@ -146,11 +146,10 @@ void SoftwareKeyboard::WriteText(std::optional<std::u16string> text) {
 
         if (complete) {
             broker.PushNormalDataFromApplet(IStorage{output_main});
+            broker.SignalStateChanged();
         } else {
             broker.PushInteractiveDataFromApplet(IStorage{output_sub});
         }
-
-        broker.SignalStateChanged();
     } else {
         output_main[0] = 1;
         complete = true;

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

@@ -137,6 +137,10 @@ u32 nvhost_gpu::AllocateObjectContext(const std::vector<u8>& input, std::vector<
 }
 
 static void PushGPUEntries(Tegra::CommandList&& entries) {
+    if (entries.empty()) {
+        return;
+    }
+
     auto& dma_pusher{Core::System::GetInstance().GPU().DmaPusher()};
     dma_pusher.Push(std::move(entries));
     dma_pusher.DispatchCalls();

src/core/hle/service/service.cpp

@@ -97,29 +97,33 @@ ServiceFrameworkBase::ServiceFrameworkBase(const char* service_name, u32 max_ses
 ServiceFrameworkBase::~ServiceFrameworkBase() = default;
 
 void ServiceFrameworkBase::InstallAsService(SM::ServiceManager& service_manager) {
-    ASSERT(port == nullptr);
-    port = service_manager.RegisterService(service_name, max_sessions).Unwrap();
+    ASSERT(!port_installed);
+
+    auto port = service_manager.RegisterService(service_name, max_sessions).Unwrap();
     port->SetHleHandler(shared_from_this());
+    port_installed = true;
 }
 
 void ServiceFrameworkBase::InstallAsNamedPort() {
-    ASSERT(port == nullptr);
+    ASSERT(!port_installed);
 
     auto& kernel = Core::System::GetInstance().Kernel();
     auto [server_port, client_port] =
         Kernel::ServerPort::CreatePortPair(kernel, max_sessions, service_name);
     server_port->SetHleHandler(shared_from_this());
     kernel.AddNamedPort(service_name, std::move(client_port));
+    port_installed = true;
 }
 
 Kernel::SharedPtr<Kernel::ClientPort> ServiceFrameworkBase::CreatePort() {
-    ASSERT(port == nullptr);
+    ASSERT(!port_installed);
 
     auto& kernel = Core::System::GetInstance().Kernel();
     auto [server_port, client_port] =
         Kernel::ServerPort::CreatePortPair(kernel, max_sessions, service_name);
-    port = MakeResult(std::move(server_port)).Unwrap();
+    auto port = MakeResult(std::move(server_port)).Unwrap();
     port->SetHleHandler(shared_from_this());
+    port_installed = true;
     return client_port;
 }
 

src/core/hle/service/service.h

@@ -96,11 +96,9 @@ private:
     /// Maximum number of concurrent sessions that this service can handle.
     u32 max_sessions;
 
-    /**
-     * Port where incoming connections will be received. Only created when InstallAsService() or
-     * InstallAsNamedPort() are called.
-     */
-    Kernel::SharedPtr<Kernel::ServerPort> port;
+    /// Flag to store if a port was already create/installed to detect multiple install attempts,
+    /// which is not supported.
+    bool port_installed = false;
 
     /// Function used to safely up-cast pointers to the derived class before invoking a handler.
     InvokerFn* handler_invoker;

src/core/loader/deconstructed_rom_directory.cpp

@@ -129,7 +129,10 @@ ResultStatus AppLoader_DeconstructedRomDirectory::Load(Kernel::Process& process)
         return ResultStatus::Error32BitISA;
     }
 
-    process.LoadFromMetadata(metadata);
+    if (process.LoadFromMetadata(metadata).IsError()) {
+        return ResultStatus::ErrorUnableToParseKernelMetadata;
+    }
+
     const FileSys::PatchManager pm(metadata.GetTitleID());
 
     // Load NSO modules

src/core/loader/loader.cpp

@@ -93,7 +93,7 @@ std::string GetFileTypeString(FileType type) {
     return "unknown";
 }
 
-constexpr std::array<const char*, 60> RESULT_MESSAGES{
+constexpr std::array<const char*, 62> RESULT_MESSAGES{
     "The operation completed successfully.",
     "The loader requested to load is already loaded.",
     "The operation is not implemented.",
@@ -103,6 +103,7 @@ constexpr std::array<const char*, 60> RESULT_MESSAGES{
     "The NPDM has a bad ACI header,",
     "The NPDM file has a bad file access control.",
     "The NPDM has a bad file access header.",
+    "The NPDM has bad kernel capability descriptors.",
     "The PFS/HFS partition has a bad header.",
     "The PFS/HFS partition has incorrect size as determined by the header.",
     "The NCA file has a bad header.",
@@ -125,6 +126,7 @@ constexpr std::array<const char*, 60> RESULT_MESSAGES{
     "The file could not be found or does not exist.",
     "The game is missing a program metadata file (main.npdm).",
     "The game uses the currently-unimplemented 32-bit architecture.",
+    "Unable to completely parse the kernel metadata when loading the emulated process",
     "The RomFS could not be found.",
     "The ELF file has incorrect size as determined by the header.",
     "There was a general error loading the NRO into emulated memory.",

src/core/loader/loader.h

@@ -67,6 +67,7 @@ enum class ResultStatus : u16 {
     ErrorBadACIHeader,
     ErrorBadFileAccessControl,
     ErrorBadFileAccessHeader,
+    ErrorBadKernelCapabilityDescriptors,
     ErrorBadPFSHeader,
     ErrorIncorrectPFSFileSize,
     ErrorBadNCAHeader,
@@ -89,6 +90,7 @@ enum class ResultStatus : u16 {
     ErrorNullFile,
     ErrorMissingNPDM,
     Error32BitISA,
+    ErrorUnableToParseKernelMetadata,
     ErrorNoRomFS,
     ErrorIncorrectELFFileSize,
     ErrorLoadingNRO,

src/video_core/engines/shader_bytecode.h

@@ -609,7 +609,7 @@ union Instruction {
 
         BitField<31, 1, u64> negate_b;
         BitField<30, 1, u64> abs_b;
-        BitField<47, 2, HalfType> type_b;
+        BitField<28, 2, HalfType> type_b;
 
         BitField<35, 2, HalfType> type_c;
     } alu_half;

src/yuzu/configuration/configure_audio.cpp

@@ -17,8 +17,8 @@ ConfigureAudio::ConfigureAudio(QWidget* parent)
 
     ui->output_sink_combo_box->clear();
     ui->output_sink_combo_box->addItem("auto");
-    for (const auto& sink_detail : AudioCore::g_sink_details) {
-        ui->output_sink_combo_box->addItem(sink_detail.id);
+    for (const char* id : AudioCore::GetSinkIDs()) {
+        ui->output_sink_combo_box->addItem(id);
     }
 
     connect(ui->volume_slider, &QSlider::valueChanged, this,
@@ -97,8 +97,7 @@ void ConfigureAudio::updateAudioDevices(int sink_index) {
     ui->audio_device_combo_box->addItem(AudioCore::auto_device_name);
 
     const std::string sink_id = ui->output_sink_combo_box->itemText(sink_index).toStdString();
-    const std::vector<std::string> device_list = AudioCore::GetSinkDetails(sink_id).list_devices();
-    for (const auto& device : device_list) {
+    for (const auto& device : AudioCore::GetDeviceListForSink(sink_id)) {
         ui->audio_device_combo_box->addItem(QString::fromStdString(device));
     }
 }