Android #173

kernel: use simple mutex for object list container

README.md

@@ -1,3 +1,13 @@
+| Pull Request | Commit | Title | Author | Merged? |
+|----|----|----|----|----|
+| [12466](https://github.com/yuzu-emu/yuzu//pull/12466) | [`ddda76f9b`](https://github.com/yuzu-emu/yuzu//pull/12466/files) |  core: track separate heap allocation for linux  | [liamwhite](https://github.com/liamwhite/) | Yes |
+| [12479](https://github.com/yuzu-emu/yuzu//pull/12479) | [`20e040723`](https://github.com/yuzu-emu/yuzu//pull/12479/files) | video_core: Fix buffer_row_length for linear compressed textures | [GPUCode](https://github.com/GPUCode/) | Yes |
+
+
+End of merge log. You can find the original README.md below the break.
+
+-----
+
 <!--
 SPDX-FileCopyrightText: 2018 yuzu Emulator Project
 SPDX-License-Identifier: GPL-2.0-or-later

src/common/CMakeLists.txt

@@ -64,6 +64,8 @@ add_library(common STATIC
     fs/path_util.cpp
     fs/path_util.h
     hash.h
+    heap_tracker.cpp
+    heap_tracker.h
     hex_util.cpp
     hex_util.h
     host_memory.cpp

src/common/assert.cpp

@@ -3,16 +3,19 @@
 
 #include "common/assert.h"
 #include "common/common_funcs.h"
+#include "common/logging/backend.h"
 
 #include "common/settings.h"
 
 void assert_fail_impl() {
     if (Settings::values.use_debug_asserts) {
+        Common::Log::Stop();
         Crash();
     }
 }
 
 [[noreturn]] void unreachable_impl() {
+    Common::Log::Stop();
     Crash();
     throw std::runtime_error("Unreachable code");
 }

src/common/heap_tracker.cpp

@@ -0,0 +1,263 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <algorithm>
+#include <vector>
+
+#include "common/heap_tracker.h"
+#include "common/logging/log.h"
+
+namespace Common {
+
+namespace {
+
+constexpr s64 MaxResidentMapCount = 0x8000;
+
+} // namespace
+
+HeapTracker::HeapTracker(Common::HostMemory& buffer) : m_buffer(buffer) {}
+HeapTracker::~HeapTracker() = default;
+
+void HeapTracker::Map(size_t virtual_offset, size_t host_offset, size_t length,
+                      MemoryPermission perm, bool is_separate_heap) {
+    // When mapping other memory, map pages immediately.
+    if (!is_separate_heap) {
+        m_buffer.Map(virtual_offset, host_offset, length, perm, false);
+        return;
+    }
+
+    {
+        // We are mapping part of a separate heap.
+        std::scoped_lock lk{m_lock};
+
+        auto* const map = new SeparateHeapMap{
+            .vaddr = virtual_offset,
+            .paddr = host_offset,
+            .size = length,
+            .tick = m_tick++,
+            .perm = perm,
+            .is_resident = false,
+        };
+
+        // Insert into mappings.
+        m_map_count++;
+        m_mappings.insert(*map);
+    }
+
+    // Finally, map.
+    this->DeferredMapSeparateHeap(virtual_offset);
+}
+
+void HeapTracker::Unmap(size_t virtual_offset, size_t size, bool is_separate_heap) {
+    // If this is a separate heap...
+    if (is_separate_heap) {
+        std::scoped_lock lk{m_lock};
+
+        const SeparateHeapMap key{
+            .vaddr = virtual_offset,
+        };
+
+        // Split at the boundaries of the region we are removing.
+        this->SplitHeapMapLocked(virtual_offset);
+        this->SplitHeapMapLocked(virtual_offset + size);
+
+        // Erase all mappings in range.
+        auto it = m_mappings.find(key);
+        while (it != m_mappings.end() && it->vaddr < virtual_offset + size) {
+            // Get underlying item.
+            auto* const item = std::addressof(*it);
+
+            // If resident, erase from resident map.
+            if (item->is_resident) {
+                ASSERT(--m_resident_map_count >= 0);
+                m_resident_mappings.erase(m_resident_mappings.iterator_to(*item));
+            }
+
+            // Erase from map.
+            it = m_mappings.erase(it);
+            ASSERT(--m_map_count >= 0);
+
+            // Free the item.
+            delete item;
+        }
+    }
+
+    // Unmap pages.
+    m_buffer.Unmap(virtual_offset, size, false);
+}
+
+void HeapTracker::Protect(size_t virtual_offset, size_t size, MemoryPermission perm) {
+    // Ensure no rebuild occurs while reprotecting.
+    std::shared_lock lk{m_rebuild_lock};
+
+    // Split at the boundaries of the region we are reprotecting.
+    this->SplitHeapMap(virtual_offset, size);
+
+    // Declare tracking variables.
+    VAddr cur = virtual_offset;
+    VAddr end = virtual_offset + size;
+
+    while (cur < end) {
+        VAddr next = cur;
+        bool should_protect = false;
+
+        {
+            std::scoped_lock lk2{m_lock};
+
+            const SeparateHeapMap key{
+                .vaddr = next,
+            };
+
+            // Try to get the next mapping corresponding to this address.
+            const auto it = m_mappings.nfind(key);
+
+            if (it == m_mappings.end()) {
+                // There are no separate heap mappings remaining.
+                next = end;
+                should_protect = true;
+            } else if (it->vaddr == cur) {
+                // We are in range.
+                // Update permission bits.
+                it->perm = perm;
+
+                // Determine next address and whether we should protect.
+                next = cur + it->size;
+                should_protect = it->is_resident;
+            } else /* if (it->vaddr > cur) */ {
+                // We weren't in range, but there is a block coming up that will be.
+                next = it->vaddr;
+                should_protect = true;
+            }
+        }
+
+        // Clamp to end.
+        next = std::min(next, end);
+
+        // Reprotect, if we need to.
+        if (should_protect) {
+            m_buffer.Protect(cur, next - cur, perm);
+        }
+
+        // Advance.
+        cur = next;
+    }
+}
+
+bool HeapTracker::DeferredMapSeparateHeap(u8* fault_address) {
+    if (m_buffer.IsInVirtualRange(fault_address)) {
+        return this->DeferredMapSeparateHeap(fault_address - m_buffer.VirtualBasePointer());
+    }
+
+    return false;
+}
+
+bool HeapTracker::DeferredMapSeparateHeap(size_t virtual_offset) {
+    bool rebuild_required = false;
+
+    {
+        std::scoped_lock lk{m_lock};
+
+        // Check to ensure this was a non-resident separate heap mapping.
+        const auto it = this->GetNearestHeapMapLocked(virtual_offset);
+        if (it == m_mappings.end() || it->is_resident) {
+            return false;
+        }
+
+        // Update tick before possible rebuild.
+        it->tick = m_tick++;
+
+        // Check if we need to rebuild.
+        if (m_resident_map_count > MaxResidentMapCount) {
+            rebuild_required = true;
+        }
+
+        // Map the area.
+        m_buffer.Map(it->vaddr, it->paddr, it->size, it->perm, false);
+
+        // This map is now resident.
+        it->is_resident = true;
+        m_resident_map_count++;
+        m_resident_mappings.insert(*it);
+    }
+
+    if (rebuild_required) {
+        // A rebuild was required, so perform it now.
+        this->RebuildSeparateHeapAddressSpace();
+    }
+
+    return true;
+}
+
+void HeapTracker::RebuildSeparateHeapAddressSpace() {
+    std::scoped_lock lk{m_rebuild_lock, m_lock};
+
+    ASSERT(!m_resident_mappings.empty());
+
+    // Unmap so we have at least 4 maps available.
+    const size_t desired_count = std::min(m_resident_map_count, MaxResidentMapCount - 4);
+    const size_t evict_count = m_resident_map_count - desired_count;
+    auto it = m_resident_mappings.begin();
+
+    for (size_t i = 0; i < evict_count && it != m_resident_mappings.end(); i++) {
+        // Unmark and unmap.
+        it->is_resident = false;
+        m_buffer.Unmap(it->vaddr, it->size, false);
+
+        // Advance.
+        ASSERT(--m_resident_map_count >= 0);
+        it = m_resident_mappings.erase(it);
+    }
+}
+
+void HeapTracker::SplitHeapMap(VAddr offset, size_t size) {
+    std::scoped_lock lk{m_lock};
+
+    this->SplitHeapMapLocked(offset);
+    this->SplitHeapMapLocked(offset + size);
+}
+
+void HeapTracker::SplitHeapMapLocked(VAddr offset) {
+    const auto it = this->GetNearestHeapMapLocked(offset);
+    if (it == m_mappings.end() || it->vaddr == offset) {
+        // Not contained or no split required.
+        return;
+    }
+
+    // Cache the original values.
+    auto* const left = std::addressof(*it);
+    const size_t orig_size = left->size;
+
+    // Adjust the left map.
+    const size_t left_size = offset - left->vaddr;
+    left->size = left_size;
+
+    // Create the new right map.
+    auto* const right = new SeparateHeapMap{
+        .vaddr = left->vaddr + left_size,
+        .paddr = left->paddr + left_size,
+        .size = orig_size - left_size,
+        .tick = left->tick,
+        .perm = left->perm,
+        .is_resident = left->is_resident,
+    };
+
+    // Insert the new right map.
+    m_map_count++;
+    m_mappings.insert(*right);
+
+    // If resident, also insert into resident map.
+    if (right->is_resident) {
+        m_resident_mappings.insert(*right);
+        m_resident_map_count++;
+    }
+}
+
+HeapTracker::AddrTree::iterator HeapTracker::GetNearestHeapMapLocked(VAddr offset) {
+    const SeparateHeapMap key{
+        .vaddr = offset,
+    };
+
+    return m_mappings.find(key);
+}
+
+} // namespace Common

src/common/heap_tracker.h

@@ -0,0 +1,97 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <atomic>
+#include <mutex>
+#include <set>
+#include <shared_mutex>
+
+#include "common/host_memory.h"
+#include "common/intrusive_red_black_tree.h"
+
+namespace Common {
+
+struct SeparateHeapMap {
+    Common::IntrusiveRedBlackTreeNode addr_node{};
+    Common::IntrusiveRedBlackTreeNode tick_node{};
+    VAddr vaddr{};
+    PAddr paddr{};
+    size_t size{};
+    size_t tick{};
+    MemoryPermission perm{};
+    bool is_resident{};
+};
+
+struct SeparateHeapMapAddrComparator {
+    static constexpr int Compare(const SeparateHeapMap& lhs, const SeparateHeapMap& rhs) {
+        if (lhs.vaddr < rhs.vaddr) {
+            return -1;
+        } else if (lhs.vaddr <= (rhs.vaddr + rhs.size - 1)) {
+            return 0;
+        } else {
+            return 1;
+        }
+    }
+};
+
+struct SeparateHeapMapTickComparator {
+    static constexpr int Compare(const SeparateHeapMap& lhs, const SeparateHeapMap& rhs) {
+        if (lhs.tick < rhs.tick) {
+            return -1;
+        } else if (lhs.tick > rhs.tick) {
+            return 1;
+        } else {
+            return SeparateHeapMapAddrComparator::Compare(lhs, rhs);
+        }
+    }
+};
+
+class HeapTracker {
+public:
+    explicit HeapTracker(Common::HostMemory& buffer);
+    ~HeapTracker();
+
+    void Map(size_t virtual_offset, size_t host_offset, size_t length, MemoryPermission perm,
+             bool is_separate_heap);
+    void Unmap(size_t virtual_offset, size_t size, bool is_separate_heap);
+    void Protect(size_t virtual_offset, size_t length, MemoryPermission perm);
+    u8* VirtualBasePointer() {
+        return m_buffer.VirtualBasePointer();
+    }
+
+    bool DeferredMapSeparateHeap(u8* fault_address);
+    bool DeferredMapSeparateHeap(size_t virtual_offset);
+
+private:
+    using AddrTreeTraits =
+        Common::IntrusiveRedBlackTreeMemberTraitsDeferredAssert<&SeparateHeapMap::addr_node>;
+    using AddrTree = AddrTreeTraits::TreeType<SeparateHeapMapAddrComparator>;
+
+    using TickTreeTraits =
+        Common::IntrusiveRedBlackTreeMemberTraitsDeferredAssert<&SeparateHeapMap::tick_node>;
+    using TickTree = TickTreeTraits::TreeType<SeparateHeapMapTickComparator>;
+
+    AddrTree m_mappings{};
+    TickTree m_resident_mappings{};
+
+private:
+    void SplitHeapMap(VAddr offset, size_t size);
+    void SplitHeapMapLocked(VAddr offset);
+
+    AddrTree::iterator GetNearestHeapMapLocked(VAddr offset);
+
+    void RebuildSeparateHeapAddressSpace();
+
+private:
+    Common::HostMemory& m_buffer;
+
+    std::shared_mutex m_rebuild_lock{};
+    std::mutex m_lock{};
+    s64 m_map_count{};
+    s64 m_resident_map_count{};
+    size_t m_tick{};
+};
+
+} // namespace Common

src/common/host_memory.cpp

@@ -679,7 +679,7 @@ HostMemory::HostMemory(HostMemory&&) noexcept = default;
 HostMemory& HostMemory::operator=(HostMemory&&) noexcept = default;
 
 void HostMemory::Map(size_t virtual_offset, size_t host_offset, size_t length,
-                     MemoryPermission perms) {
+                     MemoryPermission perms, bool separate_heap) {
     ASSERT(virtual_offset % PageAlignment == 0);
     ASSERT(host_offset % PageAlignment == 0);
     ASSERT(length % PageAlignment == 0);
@@ -691,7 +691,7 @@ void HostMemory::Map(size_t virtual_offset, size_t host_offset, size_t length,
     impl->Map(virtual_offset + virtual_base_offset, host_offset, length, perms);
 }
 
-void HostMemory::Unmap(size_t virtual_offset, size_t length) {
+void HostMemory::Unmap(size_t virtual_offset, size_t length, bool separate_heap) {
     ASSERT(virtual_offset % PageAlignment == 0);
     ASSERT(length % PageAlignment == 0);
     ASSERT(virtual_offset + length <= virtual_size);
@@ -701,14 +701,16 @@ void HostMemory::Unmap(size_t virtual_offset, size_t length) {
     impl->Unmap(virtual_offset + virtual_base_offset, length);
 }
 
-void HostMemory::Protect(size_t virtual_offset, size_t length, bool read, bool write,
-                         bool execute) {
+void HostMemory::Protect(size_t virtual_offset, size_t length, MemoryPermission perm) {
     ASSERT(virtual_offset % PageAlignment == 0);
     ASSERT(length % PageAlignment == 0);
     ASSERT(virtual_offset + length <= virtual_size);
     if (length == 0 || !virtual_base || !impl) {
         return;
     }
+    const bool read = True(perm & MemoryPermission::Read);
+    const bool write = True(perm & MemoryPermission::Write);
+    const bool execute = True(perm & MemoryPermission::Execute);
     impl->Protect(virtual_offset + virtual_base_offset, length, read, write, execute);
 }
 

src/common/host_memory.h

@@ -40,11 +40,12 @@ public:
     HostMemory(HostMemory&& other) noexcept;
     HostMemory& operator=(HostMemory&& other) noexcept;
 
-    void Map(size_t virtual_offset, size_t host_offset, size_t length, MemoryPermission perms);
+    void Map(size_t virtual_offset, size_t host_offset, size_t length, MemoryPermission perms,
+             bool separate_heap);
 
-    void Unmap(size_t virtual_offset, size_t length);
+    void Unmap(size_t virtual_offset, size_t length, bool separate_heap);
 
-    void Protect(size_t virtual_offset, size_t length, bool read, bool write, bool execute = false);
+    void Protect(size_t virtual_offset, size_t length, MemoryPermission perms);
 
     void EnableDirectMappedAddress();
 
@@ -64,6 +65,10 @@ public:
         return virtual_base;
     }
 
+    bool IsInVirtualRange(void* address) const noexcept {
+        return address >= virtual_base && address < virtual_base + virtual_size;
+    }
+
 private:
     size_t backing_size{};
     size_t virtual_size{};

src/common/logging/backend.cpp

@@ -208,6 +208,10 @@ public:
         instance->StartBackendThread();
     }
 
+    static void Stop() {
+        instance->StopBackendThread();
+    }
+
     Impl(const Impl&) = delete;
     Impl& operator=(const Impl&) = delete;
 
@@ -259,6 +263,15 @@ private:
         });
     }
 
+    void StopBackendThread() {
+        backend_thread.request_stop();
+        if (backend_thread.joinable()) {
+            backend_thread.join();
+        }
+
+        ForEachBackend([](Backend& backend) { backend.Flush(); });
+    }
+
     Entry CreateEntry(Class log_class, Level log_level, const char* filename, unsigned int line_nr,
                       const char* function, std::string&& message) const {
         using std::chrono::duration_cast;
@@ -313,6 +326,10 @@ void Start() {
     Impl::Start();
 }
 
+void Stop() {
+    Impl::Stop();
+}
+
 void DisableLoggingInTests() {
     initialization_in_progress_suppress_logging = true;
 }

src/common/logging/backend.h

@@ -14,6 +14,9 @@ void Initialize();
 
 void Start();
 
+/// Explicitly stops the logger thread and flushes the buffers
+void Stop();
+
 void DisableLoggingInTests();
 
 /**

src/common/ring_buffer.h

@@ -103,7 +103,7 @@ private:
     // Having them on the same cache-line would result in false-sharing between them.
     // TODO: Remove this ifdef whenever clang and GCC support
     //       std::hardware_destructive_interference_size.
-#if defined(_MSC_VER) && _MSC_VER >= 1911
+#ifdef __cpp_lib_hardware_interference_size
     alignas(std::hardware_destructive_interference_size) std::atomic_size_t m_read_index{0};
     alignas(std::hardware_destructive_interference_size) std::atomic_size_t m_write_index{0};
 #else

src/core/CMakeLists.txt

@@ -978,6 +978,7 @@ endif()
 
 if (ARCHITECTURE_x86_64 OR ARCHITECTURE_arm64)
     target_sources(core PRIVATE
+        arm/dynarmic/arm_dynarmic.cpp
         arm/dynarmic/arm_dynarmic.h
         arm/dynarmic/arm_dynarmic_64.cpp
         arm/dynarmic/arm_dynarmic_64.h

src/core/arm/arm_interface.cpp

@@ -9,7 +9,7 @@
 
 namespace Core {
 
-void ArmInterface::LogBacktrace(const Kernel::KProcess* process) const {
+void ArmInterface::LogBacktrace(Kernel::KProcess* process) const {
     Kernel::Svc::ThreadContext ctx;
     this->GetContext(ctx);
 

src/core/arm/arm_interface.h

@@ -95,7 +95,7 @@ public:
     virtual void SignalInterrupt(Kernel::KThread* thread) = 0;
 
     // Stack trace generation.
-    void LogBacktrace(const Kernel::KProcess* process) const;
+    void LogBacktrace(Kernel::KProcess* process) const;
 
     // Debug functionality.
     virtual const Kernel::DebugWatchpoint* HaltedWatchpoint() const = 0;

src/core/arm/debug.cpp

@@ -79,7 +79,7 @@ constexpr std::array<u64, 2> SegmentBases{
     0x7100000000ULL,
 };
 
-void SymbolicateBacktrace(const Kernel::KProcess* process, std::vector<BacktraceEntry>& out) {
+void SymbolicateBacktrace(Kernel::KProcess* process, std::vector<BacktraceEntry>& out) {
     auto modules = FindModules(process);
 
     const bool is_64 = process->Is64Bit();
@@ -118,7 +118,7 @@ void SymbolicateBacktrace(const Kernel::KProcess* process, std::vector<Backtrace
     }
 }
 
-std::vector<BacktraceEntry> GetAArch64Backtrace(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetAArch64Backtrace(Kernel::KProcess* process,
                                                 const Kernel::Svc::ThreadContext& ctx) {
     std::vector<BacktraceEntry> out;
     auto& memory = process->GetMemory();
@@ -144,7 +144,7 @@ std::vector<BacktraceEntry> GetAArch64Backtrace(const Kernel::KProcess* process,
     return out;
 }
 
-std::vector<BacktraceEntry> GetAArch32Backtrace(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetAArch32Backtrace(Kernel::KProcess* process,
                                                 const Kernel::Svc::ThreadContext& ctx) {
     std::vector<BacktraceEntry> out;
     auto& memory = process->GetMemory();
@@ -173,7 +173,7 @@ std::vector<BacktraceEntry> GetAArch32Backtrace(const Kernel::KProcess* process,
 } // namespace
 
 std::optional<std::string> GetThreadName(const Kernel::KThread* thread) {
-    const auto* process = thread->GetOwnerProcess();
+    auto* process = thread->GetOwnerProcess();
     if (process->Is64Bit()) {
         return GetNameFromThreadType64(process->GetMemory(), *thread);
     } else {
@@ -248,7 +248,7 @@ Kernel::KProcessAddress GetModuleEnd(const Kernel::KProcess* process,
     return cur_addr - 1;
 }
 
-Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process) {
+Loader::AppLoader::Modules FindModules(Kernel::KProcess* process) {
     Loader::AppLoader::Modules modules;
 
     auto& page_table = process->GetPageTable();
@@ -312,7 +312,7 @@ Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process) {
     return modules;
 }
 
-Kernel::KProcessAddress FindMainModuleEntrypoint(const Kernel::KProcess* process) {
+Kernel::KProcessAddress FindMainModuleEntrypoint(Kernel::KProcess* process) {
     // Do we have any loaded executable sections?
     auto modules = FindModules(process);
 
@@ -337,7 +337,7 @@ void InvalidateInstructionCacheRange(const Kernel::KProcess* process, u64 addres
     }
 }
 
-std::vector<BacktraceEntry> GetBacktraceFromContext(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetBacktraceFromContext(Kernel::KProcess* process,
                                                     const Kernel::Svc::ThreadContext& ctx) {
     if (process->Is64Bit()) {
         return GetAArch64Backtrace(process, ctx);

src/core/arm/debug.h

@@ -14,9 +14,9 @@ std::optional<std::string> GetThreadName(const Kernel::KThread* thread);
 std::string_view GetThreadWaitReason(const Kernel::KThread* thread);
 std::string GetThreadState(const Kernel::KThread* thread);
 
-Loader::AppLoader::Modules FindModules(const Kernel::KProcess* process);
+Loader::AppLoader::Modules FindModules(Kernel::KProcess* process);
 Kernel::KProcessAddress GetModuleEnd(const Kernel::KProcess* process, Kernel::KProcessAddress base);
-Kernel::KProcessAddress FindMainModuleEntrypoint(const Kernel::KProcess* process);
+Kernel::KProcessAddress FindMainModuleEntrypoint(Kernel::KProcess* process);
 
 void InvalidateInstructionCacheRange(const Kernel::KProcess* process, u64 address, u64 size);
 
@@ -28,7 +28,7 @@ struct BacktraceEntry {
     std::string name;
 };
 
-std::vector<BacktraceEntry> GetBacktraceFromContext(const Kernel::KProcess* process,
+std::vector<BacktraceEntry> GetBacktraceFromContext(Kernel::KProcess* process,
                                                     const Kernel::Svc::ThreadContext& ctx);
 std::vector<BacktraceEntry> GetBacktrace(const Kernel::KThread* thread);
 

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -0,0 +1,49 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#ifdef __linux__
+
+#include "common/signal_chain.h"
+
+#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/hle/kernel/k_process.h"
+#include "core/memory.h"
+
+namespace Core {
+
+namespace {
+
+thread_local Core::Memory::Memory* g_current_memory{};
+std::once_flag g_registered{};
+struct sigaction g_old_segv {};
+
+void HandleSigSegv(int sig, siginfo_t* info, void* ctx) {
+    if (g_current_memory && g_current_memory->InvalidateSeparateHeap(info->si_addr)) {
+        return;
+    }
+
+    return g_old_segv.sa_sigaction(sig, info, ctx);
+}
+
+} // namespace
+
+ScopedJitExecution::ScopedJitExecution(Kernel::KProcess* process) {
+    g_current_memory = std::addressof(process->GetMemory());
+}
+
+ScopedJitExecution::~ScopedJitExecution() {
+    g_current_memory = nullptr;
+}
+
+void ScopedJitExecution::RegisterHandler() {
+    std::call_once(g_registered, [] {
+        struct sigaction sa {};
+        sa.sa_sigaction = &HandleSigSegv;
+        sa.sa_flags = SA_SIGINFO | SA_ONSTACK;
+        Common::SigAction(SIGSEGV, std::addressof(sa), std::addressof(g_old_segv));
+    });
+}
+
+} // namespace Core
+
+#endif

src/core/arm/dynarmic/arm_dynarmic.h

@@ -26,4 +26,24 @@ constexpr HaltReason TranslateHaltReason(Dynarmic::HaltReason hr) {
     return static_cast<HaltReason>(hr);
 }
 
+#ifdef __linux__
+
+class ScopedJitExecution {
+public:
+    explicit ScopedJitExecution(Kernel::KProcess* process);
+    ~ScopedJitExecution();
+    static void RegisterHandler();
+};
+
+#else
+
+class ScopedJitExecution {
+public:
+    explicit ScopedJitExecution(Kernel::KProcess* process) {}
+    ~ScopedJitExecution() {}
+    static void RegisterHandler() {}
+};
+
+#endif
+
 } // namespace Core

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -15,7 +15,7 @@ using namespace Common::Literals;
 
 class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
 public:
-    explicit DynarmicCallbacks32(ArmDynarmic32& parent, const Kernel::KProcess* process)
+    explicit DynarmicCallbacks32(ArmDynarmic32& parent, Kernel::KProcess* process)
         : m_parent{parent}, m_memory(process->GetMemory()),
           m_process(process), m_debugger_enabled{parent.m_system.DebuggerEnabled()},
           m_check_memory_access{m_debugger_enabled ||
@@ -169,7 +169,7 @@ public:
 
     ArmDynarmic32& m_parent;
     Core::Memory::Memory& m_memory;
-    const Kernel::KProcess* m_process{};
+    Kernel::KProcess* m_process{};
     const bool m_debugger_enabled{};
     const bool m_check_memory_access{};
     static constexpr u64 MinimumRunCycles = 10000U;
@@ -331,11 +331,15 @@ bool ArmDynarmic32::IsInThumbMode() const {
 }
 
 HaltReason ArmDynarmic32::RunThread(Kernel::KThread* thread) {
+    ScopedJitExecution sj(thread->GetOwnerProcess());
+
     m_jit->ClearExclusiveState();
     return TranslateHaltReason(m_jit->Run());
 }
 
 HaltReason ArmDynarmic32::StepThread(Kernel::KThread* thread) {
+    ScopedJitExecution sj(thread->GetOwnerProcess());
+
     m_jit->ClearExclusiveState();
     return TranslateHaltReason(m_jit->Step());
 }
@@ -370,13 +374,14 @@ void ArmDynarmic32::RewindBreakpointInstruction() {
     this->SetContext(m_breakpoint_context);
 }
 
-ArmDynarmic32::ArmDynarmic32(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+ArmDynarmic32::ArmDynarmic32(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                              DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index)
     : ArmInterface{uses_wall_clock}, m_system{system}, m_exclusive_monitor{exclusive_monitor},
       m_cb(std::make_unique<DynarmicCallbacks32>(*this, process)),
       m_cp15(std::make_shared<DynarmicCP15>(*this)), m_core_index{core_index} {
     auto& page_table_impl = process->GetPageTable().GetBasePageTable().GetImpl();
     m_jit = MakeJit(&page_table_impl);
+    ScopedJitExecution::RegisterHandler();
 }
 
 ArmDynarmic32::~ArmDynarmic32() = default;

src/core/arm/dynarmic/arm_dynarmic_32.h

@@ -20,7 +20,7 @@ class System;
 
 class ArmDynarmic32 final : public ArmInterface {
 public:
-    ArmDynarmic32(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+    ArmDynarmic32(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                   DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index);
     ~ArmDynarmic32() override;
 

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -15,7 +15,7 @@ using namespace Common::Literals;
 
 class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
-    explicit DynarmicCallbacks64(ArmDynarmic64& parent, const Kernel::KProcess* process)
+    explicit DynarmicCallbacks64(ArmDynarmic64& parent, Kernel::KProcess* process)
         : m_parent{parent}, m_memory(process->GetMemory()),
           m_process(process), m_debugger_enabled{parent.m_system.DebuggerEnabled()},
           m_check_memory_access{m_debugger_enabled ||
@@ -216,7 +216,7 @@ public:
     Core::Memory::Memory& m_memory;
     u64 m_tpidrro_el0{};
     u64 m_tpidr_el0{};
-    const Kernel::KProcess* m_process{};
+    Kernel::KProcess* m_process{};
     const bool m_debugger_enabled{};
     const bool m_check_memory_access{};
     static constexpr u64 MinimumRunCycles = 10000U;
@@ -362,11 +362,15 @@ std::shared_ptr<Dynarmic::A64::Jit> ArmDynarmic64::MakeJit(Common::PageTable* pa
 }
 
 HaltReason ArmDynarmic64::RunThread(Kernel::KThread* thread) {
+    ScopedJitExecution sj(thread->GetOwnerProcess());
+
     m_jit->ClearExclusiveState();
     return TranslateHaltReason(m_jit->Run());
 }
 
 HaltReason ArmDynarmic64::StepThread(Kernel::KThread* thread) {
+    ScopedJitExecution sj(thread->GetOwnerProcess());
+
     m_jit->ClearExclusiveState();
     return TranslateHaltReason(m_jit->Step());
 }
@@ -399,13 +403,14 @@ void ArmDynarmic64::RewindBreakpointInstruction() {
     this->SetContext(m_breakpoint_context);
 }
 
-ArmDynarmic64::ArmDynarmic64(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+ArmDynarmic64::ArmDynarmic64(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                              DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index)
     : ArmInterface{uses_wall_clock}, m_system{system}, m_exclusive_monitor{exclusive_monitor},
       m_cb(std::make_unique<DynarmicCallbacks64>(*this, process)), m_core_index{core_index} {
     auto& page_table = process->GetPageTable().GetBasePageTable();
     auto& page_table_impl = page_table.GetImpl();
     m_jit = MakeJit(&page_table_impl, page_table.GetAddressSpaceWidth());
+    ScopedJitExecution::RegisterHandler();
 }
 
 ArmDynarmic64::~ArmDynarmic64() = default;

src/core/arm/dynarmic/arm_dynarmic_64.h

@@ -25,7 +25,7 @@ class System;
 
 class ArmDynarmic64 final : public ArmInterface {
 public:
-    ArmDynarmic64(System& system, bool uses_wall_clock, const Kernel::KProcess* process,
+    ArmDynarmic64(System& system, bool uses_wall_clock, Kernel::KProcess* process,
                   DynarmicExclusiveMonitor& exclusive_monitor, std::size_t core_index);
     ~ArmDynarmic64() override;
 

src/core/core.cpp

@@ -28,7 +28,6 @@
 #include "core/file_sys/savedata_factory.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_real.h"
-#include "core/gpu_dirty_memory_manager.h"
 #include "core/hid/hid_core.h"
 #include "core/hle/kernel/k_memory_manager.h"
 #include "core/hle/kernel/k_process.h"
@@ -130,11 +129,8 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
 
 struct System::Impl {
     explicit Impl(System& system)
-        : kernel{system}, fs_controller{system}, memory{system}, hid_core{}, room_network{},
-          cpu_manager{system}, reporter{system}, applet_manager{system}, profile_manager{},
-          time_manager{system}, gpu_dirty_memory_write_manager{} {
-        memory.SetGPUDirtyManagers(gpu_dirty_memory_write_manager);
-    }
+        : kernel{system}, fs_controller{system}, hid_core{}, room_network{}, cpu_manager{system},
+          reporter{system}, applet_manager{system}, profile_manager{}, time_manager{system} {}
 
     void Initialize(System& system) {
         device_memory = std::make_unique<Core::DeviceMemory>();
@@ -241,17 +237,17 @@ struct System::Impl {
         debugger = std::make_unique<Debugger>(system, port);
     }
 
-    SystemResultStatus SetupForApplicationProcess(System& system, Frontend::EmuWindow& emu_window) {
+    void InitializeKernel(System& system) {
         LOG_DEBUG(Core, "initialized OK");
 
         // Setting changes may require a full system reinitialization (e.g., disabling multicore).
         ReinitializeIfNecessary(system);
 
-        memory.SetGPUDirtyManagers(gpu_dirty_memory_write_manager);
-
         kernel.Initialize();
         cpu_manager.Initialize();
+    }
 
+    SystemResultStatus SetupForApplicationProcess(System& system, Frontend::EmuWindow& emu_window) {
         /// Reset all glue registrations
         arp_manager.ResetAll();
 
@@ -300,17 +296,9 @@ struct System::Impl {
             return SystemResultStatus::ErrorGetLoader;
         }
 
-        SystemResultStatus init_result{SetupForApplicationProcess(system, emu_window)};
-        if (init_result != SystemResultStatus::Success) {
-            LOG_CRITICAL(Core, "Failed to initialize system (Error {})!",
-                         static_cast<int>(init_result));
-            ShutdownMainProcess();
-            return init_result;
-        }
+        InitializeKernel(system);
 
-        telemetry_session->AddInitialInfo(*app_loader, fs_controller, *content_provider);
-
-        // Create the process.
+        // Create the application process.
         auto main_process = Kernel::KProcess::Create(system.Kernel());
         Kernel::KProcess::Register(system.Kernel(), main_process);
         kernel.AppendNewProcess(main_process);
@@ -323,7 +311,18 @@ struct System::Impl {
             return static_cast<SystemResultStatus>(
                 static_cast<u32>(SystemResultStatus::ErrorLoader) + static_cast<u32>(load_result));
         }
+
+        // Set up the rest of the system.
+        SystemResultStatus init_result{SetupForApplicationProcess(system, emu_window)};
+        if (init_result != SystemResultStatus::Success) {
+            LOG_CRITICAL(Core, "Failed to initialize system (Error {})!",
+                         static_cast<int>(init_result));
+            ShutdownMainProcess();
+            return init_result;
+        }
+
         AddGlueRegistrationForProcess(*app_loader, *main_process);
+        telemetry_session->AddInitialInfo(*app_loader, fs_controller, *content_provider);
 
         // Initialize cheat engine
         if (cheat_engine) {
@@ -426,7 +425,6 @@ struct System::Impl {
         cpu_manager.Shutdown();
         debugger.reset();
         kernel.Shutdown();
-        memory.Reset();
         Network::RestartSocketOperations();
 
         if (auto room_member = room_network.GetRoomMember().lock()) {
@@ -507,7 +505,6 @@ struct System::Impl {
     std::unique_ptr<Tegra::Host1x::Host1x> host1x_core;
     std::unique_ptr<Core::DeviceMemory> device_memory;
     std::unique_ptr<AudioCore::AudioCore> audio_core;
-    Core::Memory::Memory memory;
     Core::HID::HIDCore hid_core;
     Network::RoomNetwork room_network;
 
@@ -567,9 +564,6 @@ struct System::Impl {
     std::array<u64, Core::Hardware::NUM_CPU_CORES> dynarmic_ticks{};
     std::array<MicroProfileToken, Core::Hardware::NUM_CPU_CORES> microprofile_cpu{};
 
-    std::array<Core::GPUDirtyMemoryManager, Core::Hardware::NUM_CPU_CORES>
-        gpu_dirty_memory_write_manager{};
-
     std::deque<std::vector<u8>> user_channel;
 };
 
@@ -652,29 +646,12 @@ void System::PrepareReschedule(const u32 core_index) {
     impl->kernel.PrepareReschedule(core_index);
 }
 
-Core::GPUDirtyMemoryManager& System::CurrentGPUDirtyMemoryManager() {
-    const std::size_t core = impl->kernel.GetCurrentHostThreadID();
-    return impl->gpu_dirty_memory_write_manager[core < Core::Hardware::NUM_CPU_CORES
-                                                    ? core
-                                                    : Core::Hardware::NUM_CPU_CORES - 1];
-}
-
-/// Provides a constant reference to the current gou dirty memory manager.
-const Core::GPUDirtyMemoryManager& System::CurrentGPUDirtyMemoryManager() const {
-    const std::size_t core = impl->kernel.GetCurrentHostThreadID();
-    return impl->gpu_dirty_memory_write_manager[core < Core::Hardware::NUM_CPU_CORES
-                                                    ? core
-                                                    : Core::Hardware::NUM_CPU_CORES - 1];
-}
-
 size_t System::GetCurrentHostThreadID() const {
     return impl->kernel.GetCurrentHostThreadID();
 }
 
 void System::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) {
-    for (auto& manager : impl->gpu_dirty_memory_write_manager) {
-        manager.Gather(callback);
-    }
+    return this->ApplicationProcess()->GatherGPUDirtyMemory(callback);
 }
 
 PerfStatsResults System::GetAndResetPerfStats() {
@@ -723,20 +700,12 @@ const Kernel::KProcess* System::ApplicationProcess() const {
     return impl->kernel.ApplicationProcess();
 }
 
-ExclusiveMonitor& System::Monitor() {
-    return impl->kernel.GetExclusiveMonitor();
-}
-
-const ExclusiveMonitor& System::Monitor() const {
-    return impl->kernel.GetExclusiveMonitor();
-}
-
 Memory::Memory& System::ApplicationMemory() {
-    return impl->memory;
+    return impl->kernel.ApplicationProcess()->GetMemory();
 }
 
 const Core::Memory::Memory& System::ApplicationMemory() const {
-    return impl->memory;
+    return impl->kernel.ApplicationProcess()->GetMemory();
 }
 
 Tegra::GPU& System::GPU() {

src/core/core.h

@@ -116,7 +116,6 @@ class CpuManager;
 class Debugger;
 class DeviceMemory;
 class ExclusiveMonitor;
-class GPUDirtyMemoryManager;
 class PerfStats;
 class Reporter;
 class SpeedLimiter;
@@ -225,12 +224,6 @@ public:
     /// Prepare the core emulation for a reschedule
     void PrepareReschedule(u32 core_index);
 
-    /// Provides a reference to the gou dirty memory manager.
-    [[nodiscard]] Core::GPUDirtyMemoryManager& CurrentGPUDirtyMemoryManager();
-
-    /// Provides a constant reference to the current gou dirty memory manager.
-    [[nodiscard]] const Core::GPUDirtyMemoryManager& CurrentGPUDirtyMemoryManager() const;
-
     void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback);
 
     [[nodiscard]] size_t GetCurrentHostThreadID() const;
@@ -250,12 +243,6 @@ public:
     /// Gets a const reference to the underlying CPU manager
     [[nodiscard]] const CpuManager& GetCpuManager() const;
 
-    /// Gets a reference to the exclusive monitor
-    [[nodiscard]] ExclusiveMonitor& Monitor();
-
-    /// Gets a constant reference to the exclusive monitor
-    [[nodiscard]] const ExclusiveMonitor& Monitor() const;
-
     /// Gets a mutable reference to the system memory instance.
     [[nodiscard]] Core::Memory::Memory& ApplicationMemory();
 

src/core/file_sys/program_metadata.cpp

@@ -166,6 +166,10 @@ u32 ProgramMetadata::GetSystemResourceSize() const {
     return npdm_header.system_resource_size;
 }
 
+PoolPartition ProgramMetadata::GetPoolPartition() const {
+    return acid_header.pool_partition;
+}
+
 const ProgramMetadata::KernelCapabilityDescriptors& ProgramMetadata::GetKernelCapabilities() const {
     return aci_kernel_capabilities;
 }
@@ -201,7 +205,7 @@ void ProgramMetadata::Print() const {
     // Begin ACID printing (potential perms, signed)
     LOG_DEBUG(Service_FS, "Magic:                  {:.4}", acid_header.magic.data());
     LOG_DEBUG(Service_FS, "Flags:                  0x{:02X}", acid_header.flags);
-    LOG_DEBUG(Service_FS, " > Is Retail:           {}", acid_header.is_retail ? "YES" : "NO");
+    LOG_DEBUG(Service_FS, " > Is Retail:           {}", acid_header.production_flag ? "YES" : "NO");
     LOG_DEBUG(Service_FS, "Title ID Min:           0x{:016X}", acid_header.title_id_min);
     LOG_DEBUG(Service_FS, "Title ID Max:           0x{:016X}", acid_header.title_id_max);
     LOG_DEBUG(Service_FS, "Filesystem Access:      0x{:016X}\n", acid_file_access.permissions);

src/core/file_sys/program_metadata.h

@@ -34,6 +34,13 @@ enum class ProgramFilePermission : u64 {
     Everything = 1ULL << 63,
 };
 
+enum class PoolPartition : u32 {
+    Application = 0,
+    Applet = 1,
+    System = 2,
+    SystemNonSecure = 3,
+};
+
 /**
  * Helper which implements an interface to parse Program Description Metadata (NPDM)
  * Data can either be loaded from a file path or with data and an offset into it.
@@ -72,6 +79,7 @@ public:
     u64 GetTitleID() const;
     u64 GetFilesystemPermissions() const;
     u32 GetSystemResourceSize() const;
+    PoolPartition GetPoolPartition() const;
     const KernelCapabilityDescriptors& GetKernelCapabilities() const;
     const std::array<u8, 0x10>& GetName() const {
         return npdm_header.application_name;
@@ -116,8 +124,9 @@ private:
         union {
             u32 flags;
 
-            BitField<0, 1, u32> is_retail;
-            BitField<1, 31, u32> flags_unk;
+            BitField<0, 1, u32> production_flag;
+            BitField<1, 1, u32> unqualified_approval;
+            BitField<2, 4, PoolPartition> pool_partition;
         };
         u64_le title_id_min;
         u64_le title_id_max;

src/core/hle/kernel/k_address_arbiter.cpp

@@ -4,6 +4,7 @@
 #include "core/arm/exclusive_monitor.h"
 #include "core/core.h"
 #include "core/hle/kernel/k_address_arbiter.h"
+#include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_scheduler.h"
 #include "core/hle/kernel/k_scoped_scheduler_lock_and_sleep.h"
 #include "core/hle/kernel/k_thread.h"
@@ -26,9 +27,9 @@ bool ReadFromUser(KernelCore& kernel, s32* out, KProcessAddress address) {
     return true;
 }
 
-bool DecrementIfLessThan(Core::System& system, s32* out, KProcessAddress address, s32 value) {
-    auto& monitor = system.Monitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+bool DecrementIfLessThan(KernelCore& kernel, s32* out, KProcessAddress address, s32 value) {
+    auto& monitor = GetCurrentProcess(kernel).GetExclusiveMonitor();
+    const auto current_core = kernel.CurrentPhysicalCoreIndex();
 
     // NOTE: If scheduler lock is not held here, interrupt disable is required.
     // KScopedInterruptDisable di;
@@ -66,10 +67,10 @@ bool DecrementIfLessThan(Core::System& system, s32* out, KProcessAddress address
     return true;
 }
 
-bool UpdateIfEqual(Core::System& system, s32* out, KProcessAddress address, s32 value,
+bool UpdateIfEqual(KernelCore& kernel, s32* out, KProcessAddress address, s32 value,
                    s32 new_value) {
-    auto& monitor = system.Monitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    auto& monitor = GetCurrentProcess(kernel).GetExclusiveMonitor();
+    const auto current_core = kernel.CurrentPhysicalCoreIndex();
 
     // NOTE: If scheduler lock is not held here, interrupt disable is required.
     // KScopedInterruptDisable di;
@@ -159,7 +160,7 @@ Result KAddressArbiter::SignalAndIncrementIfEqual(uint64_t addr, s32 value, s32
 
         // Check the userspace value.
         s32 user_value{};
-        R_UNLESS(UpdateIfEqual(m_system, std::addressof(user_value), addr, value, value + 1),
+        R_UNLESS(UpdateIfEqual(m_kernel, std::addressof(user_value), addr, value, value + 1),
                  ResultInvalidCurrentMemory);
         R_UNLESS(user_value == value, ResultInvalidState);
 
@@ -219,7 +220,7 @@ Result KAddressArbiter::SignalAndModifyByWaitingCountIfEqual(uint64_t addr, s32
         s32 user_value{};
         bool succeeded{};
         if (value != new_value) {
-            succeeded = UpdateIfEqual(m_system, std::addressof(user_value), addr, value, new_value);
+            succeeded = UpdateIfEqual(m_kernel, std::addressof(user_value), addr, value, new_value);
         } else {
             succeeded = ReadFromUser(m_kernel, std::addressof(user_value), addr);
         }
@@ -262,7 +263,7 @@ Result KAddressArbiter::WaitIfLessThan(uint64_t addr, s32 value, bool decrement,
         s32 user_value{};
         bool succeeded{};
         if (decrement) {
-            succeeded = DecrementIfLessThan(m_system, std::addressof(user_value), addr, value);
+            succeeded = DecrementIfLessThan(m_kernel, std::addressof(user_value), addr, value);
         } else {
             succeeded = ReadFromUser(m_kernel, std::addressof(user_value), addr);
         }

src/core/hle/kernel/k_auto_object_container.cpp

@@ -8,19 +8,22 @@
 namespace Kernel {
 
 void KAutoObjectWithListContainer::Register(KAutoObjectWithList* obj) {
-    KScopedLightLock lk(m_lock);
+    // KScopedInterruptDisable di;
+    KScopedSpinLock lk(m_lock);
 
     m_object_list.insert_unique(*obj);
 }
 
 void KAutoObjectWithListContainer::Unregister(KAutoObjectWithList* obj) {
-    KScopedLightLock lk(m_lock);
+    // KScopedInterruptDisable di;
+    KScopedSpinLock lk(m_lock);
 
     m_object_list.erase(*obj);
 }
 
 size_t KAutoObjectWithListContainer::GetOwnedCount(KProcess* owner) {
-    KScopedLightLock lk(m_lock);
+    // KScopedInterruptDisable di;
+    KScopedSpinLock lk(m_lock);
 
     return std::count_if(m_object_list.begin(), m_object_list.end(),
                          [&](const auto& obj) { return obj.GetOwner() == owner; });

src/core/hle/kernel/k_auto_object_container.h

@@ -7,7 +7,7 @@
 
 #include "common/common_funcs.h"
 #include "core/hle/kernel/k_auto_object.h"
-#include "core/hle/kernel/k_light_lock.h"
+#include "core/hle/kernel/k_spin_lock.h"
 
 namespace Kernel {
 
@@ -21,32 +21,7 @@ public:
 
     using ListType = boost::intrusive::rbtree<KAutoObjectWithList>;
 
-    class ListAccessor : public KScopedLightLock {
-    public:
-        explicit ListAccessor(KAutoObjectWithListContainer* container)
-            : KScopedLightLock(container->m_lock), m_list(container->m_object_list) {}
-        explicit ListAccessor(KAutoObjectWithListContainer& container)
-            : KScopedLightLock(container.m_lock), m_list(container.m_object_list) {}
-
-        typename ListType::iterator begin() const {
-            return m_list.begin();
-        }
-
-        typename ListType::iterator end() const {
-            return m_list.end();
-        }
-
-        typename ListType::iterator find(typename ListType::const_reference ref) const {
-            return m_list.find(ref);
-        }
-
-    private:
-        ListType& m_list;
-    };
-
-    friend class ListAccessor;
-
-    KAutoObjectWithListContainer(KernelCore& kernel) : m_lock(kernel), m_object_list() {}
+    KAutoObjectWithListContainer(KernelCore& kernel) : m_lock(), m_object_list() {}
 
     void Initialize() {}
     void Finalize() {}
@@ -56,7 +31,7 @@ public:
     size_t GetOwnedCount(KProcess* owner);
 
 private:
-    KLightLock m_lock;
+    KSpinLock m_lock;
     ListType m_object_list;
 };
 

src/core/hle/kernel/k_client_port.cpp

@@ -58,9 +58,8 @@ Result KClientPort::CreateSession(KClientSession** out) {
     KSession* session{};
 
     // Reserve a new session from the resource limit.
-    //! FIXME: we are reserving this from the wrong resource limit!
-    KScopedResourceReservation session_reservation(
-        m_kernel.ApplicationProcess()->GetResourceLimit(), LimitableResource::SessionCountMax);
+    KScopedResourceReservation session_reservation(GetCurrentProcessPointer(m_kernel),
+                                                   LimitableResource::SessionCountMax);
     R_UNLESS(session_reservation.Succeeded(), ResultLimitReached);
 
     // Allocate a session normally.

src/core/hle/kernel/k_condition_variable.cpp

@@ -28,10 +28,10 @@ bool WriteToUser(KernelCore& kernel, KProcessAddress address, const u32* p) {
     return true;
 }
 
-bool UpdateLockAtomic(Core::System& system, u32* out, KProcessAddress address, u32 if_zero,
+bool UpdateLockAtomic(KernelCore& kernel, u32* out, KProcessAddress address, u32 if_zero,
                       u32 new_orr_mask) {
-    auto& monitor = system.Monitor();
-    const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
+    auto& monitor = GetCurrentProcess(kernel).GetExclusiveMonitor();
+    const auto current_core = kernel.CurrentPhysicalCoreIndex();
 
     u32 expected{};
 
@@ -208,7 +208,7 @@ void KConditionVariable::SignalImpl(KThread* thread) {
         // TODO(bunnei): We should call CanAccessAtomic(..) here.
         can_access = true;
         if (can_access) [[likely]] {
-            UpdateLockAtomic(m_system, std::addressof(prev_tag), address, own_tag,
+            UpdateLockAtomic(m_kernel, std::addressof(prev_tag), address, own_tag,
                              Svc::HandleWaitMask);
         }
     }

src/core/hle/kernel/k_handle_table.h

@@ -90,8 +90,7 @@ public:
         // Handle pseudo-handles.
         if constexpr (std::derived_from<KProcess, T>) {
             if (handle == Svc::PseudoHandle::CurrentProcess) {
-                //! FIXME: this is the wrong process!
-                auto* const cur_process = m_kernel.ApplicationProcess();
+                auto* const cur_process = GetCurrentProcessPointer(m_kernel);
                 ASSERT(cur_process != nullptr);
                 return cur_process;
             }

src/core/hle/kernel/k_page_table_base.cpp

@@ -434,7 +434,7 @@ Result KPageTableBase::InitializeForProcess(Svc::CreateProcessFlag as_type, bool
 void KPageTableBase::Finalize() {
     auto HostUnmapCallback = [&](KProcessAddress addr, u64 size) {
         if (Settings::IsFastmemEnabled()) {
-            m_system.DeviceMemory().buffer.Unmap(GetInteger(addr), size);
+            m_system.DeviceMemory().buffer.Unmap(GetInteger(addr), size, false);
         }
     };
 
@@ -5243,7 +5243,7 @@ Result KPageTableBase::MapPhysicalMemory(KProcessAddress address, size_t size) {
                                 // Unmap.
                                 R_ASSERT(this->Operate(updater.GetPageList(), cur_address,
                                                        cur_pages, 0, false, unmap_properties,
-                                                       OperationType::Unmap, true));
+                                                       OperationType::UnmapPhysical, true));
                             }
 
                             // Check if we're done.
@@ -5326,7 +5326,7 @@ Result KPageTableBase::MapPhysicalMemory(KProcessAddress address, size_t size) {
                             // Map the papges.
                             R_TRY(this->Operate(updater.GetPageList(), cur_address, map_pages,
                                                 cur_pg, map_properties,
-                                                OperationType::MapFirstGroup, false));
+                                                OperationType::MapFirstGroupPhysical, false));
                         }
                     }
 
@@ -5480,7 +5480,7 @@ Result KPageTableBase::UnmapPhysicalMemory(KProcessAddress address, size_t size)
 
             // Unmap.
             R_ASSERT(this->Operate(updater.GetPageList(), cur_address, cur_pages, 0, false,
-                                   unmap_properties, OperationType::Unmap, false));
+                                   unmap_properties, OperationType::UnmapPhysical, false));
         }
 
         // Check if we're done.
@@ -5655,7 +5655,10 @@ Result KPageTableBase::Operate(PageLinkedList* page_list, KProcessAddress virt_a
     // or free them to the page list, and so it goes unused (along with page properties).
 
     switch (operation) {
-    case OperationType::Unmap: {
+    case OperationType::Unmap:
+    case OperationType::UnmapPhysical: {
+        const bool separate_heap = operation == OperationType::UnmapPhysical;
+
         // Ensure that any pages we track are closed on exit.
         KPageGroup pages_to_close(m_kernel, this->GetBlockInfoManager());
         SCOPE_EXIT({ pages_to_close.CloseAndReset(); });
@@ -5664,7 +5667,7 @@ Result KPageTableBase::Operate(PageLinkedList* page_list, KProcessAddress virt_a
         this->MakePageGroup(pages_to_close, virt_addr, num_pages);
 
         // Unmap.
-        m_memory->UnmapRegion(*m_impl, virt_addr, num_pages * PageSize);
+        m_memory->UnmapRegion(*m_impl, virt_addr, num_pages * PageSize, separate_heap);
 
         R_SUCCEED();
     }
@@ -5672,7 +5675,7 @@ Result KPageTableBase::Operate(PageLinkedList* page_list, KProcessAddress virt_a
         ASSERT(virt_addr != 0);
         ASSERT(Common::IsAligned(GetInteger(virt_addr), PageSize));
         m_memory->MapMemoryRegion(*m_impl, virt_addr, num_pages * PageSize, phys_addr,
-                                  ConvertToMemoryPermission(properties.perm));
+                                  ConvertToMemoryPermission(properties.perm), false);
 
         // Open references to pages, if we should.
         if (this->IsHeapPhysicalAddress(phys_addr)) {
@@ -5711,16 +5714,19 @@ Result KPageTableBase::Operate(PageLinkedList* page_list, KProcessAddress virt_a
 
     switch (operation) {
     case OperationType::MapGroup:
-    case OperationType::MapFirstGroup: {
+    case OperationType::MapFirstGroup:
+    case OperationType::MapFirstGroupPhysical: {
+        const bool separate_heap = operation == OperationType::MapFirstGroupPhysical;
+
         // We want to maintain a new reference to every page in the group.
-        KScopedPageGroup spg(page_group, operation != OperationType::MapFirstGroup);
+        KScopedPageGroup spg(page_group, operation == OperationType::MapGroup);
 
         for (const auto& node : page_group) {
             const size_t size{node.GetNumPages() * PageSize};
 
             // Map the pages.
             m_memory->MapMemoryRegion(*m_impl, virt_addr, size, node.GetAddress(),
-                                      ConvertToMemoryPermission(properties.perm));
+                                      ConvertToMemoryPermission(properties.perm), separate_heap);
 
             virt_addr += size;
         }

src/core/hle/kernel/k_page_table_base.h

@@ -104,6 +104,9 @@ protected:
         ChangePermissionsAndRefresh = 5,
         ChangePermissionsAndRefreshAndFlush = 6,
         Separate = 7,
+
+        MapFirstGroupPhysical = 65000,
+        UnmapPhysical = 65001,
     };
 
     static constexpr size_t MaxPhysicalMapAlignment = 1_GiB;

src/core/hle/kernel/k_process.cpp

@@ -306,12 +306,16 @@ Result KProcess::Initialize(const Svc::CreateProcessParameter& params, const KPa
             False(params.flags & Svc::CreateProcessFlag::DisableDeviceAddressSpaceMerge);
         R_TRY(m_page_table.Initialize(as_type, enable_aslr, enable_das_merge, !enable_aslr, pool,
                                       params.code_address, params.code_num_pages * PageSize,
-                                      m_system_resource, res_limit, this->GetMemory(), 0));
+                                      m_system_resource, res_limit, m_memory, 0));
     }
     ON_RESULT_FAILURE_2 {
         m_page_table.Finalize();
     };
 
+    // Ensure our memory is initialized.
+    m_memory.SetCurrentPageTable(*this);
+    m_memory.SetGPUDirtyManagers(m_dirty_memory_managers);
+
     // Ensure we can insert the code region.
     R_UNLESS(m_page_table.CanContain(params.code_address, params.code_num_pages * PageSize,
                                      KMemoryState::Code),
@@ -399,12 +403,16 @@ Result KProcess::Initialize(const Svc::CreateProcessParameter& params,
             False(params.flags & Svc::CreateProcessFlag::DisableDeviceAddressSpaceMerge);
         R_TRY(m_page_table.Initialize(as_type, enable_aslr, enable_das_merge, !enable_aslr, pool,
                                       params.code_address, code_size, m_system_resource, res_limit,
-                                      this->GetMemory(), aslr_space_start));
+                                      m_memory, aslr_space_start));
     }
     ON_RESULT_FAILURE_2 {
         m_page_table.Finalize();
     };
 
+    // Ensure our memory is initialized.
+    m_memory.SetCurrentPageTable(*this);
+    m_memory.SetGPUDirtyManagers(m_dirty_memory_managers);
+
     // Ensure we can insert the code region.
     R_UNLESS(m_page_table.CanContain(params.code_address, code_size, KMemoryState::Code),
              ResultInvalidMemoryRegion);
@@ -1094,8 +1102,7 @@ void KProcess::UnpinThread(KThread* thread) {
 
 Result KProcess::GetThreadList(s32* out_num_threads, KProcessAddress out_thread_ids,
                                s32 max_out_count) {
-    // TODO: use current memory reference
-    auto& memory = m_kernel.System().ApplicationMemory();
+    auto& memory = this->GetMemory();
 
     // Lock the list.
     KScopedLightLock lk(m_list_lock);
@@ -1128,14 +1135,15 @@ void KProcess::Switch(KProcess* cur_process, KProcess* next_process) {}
 KProcess::KProcess(KernelCore& kernel)
     : KAutoObjectWithSlabHeapAndContainer(kernel), m_page_table{kernel}, m_state_lock{kernel},
       m_list_lock{kernel}, m_cond_var{kernel.System()}, m_address_arbiter{kernel.System()},
-      m_handle_table{kernel} {}
+      m_handle_table{kernel}, m_dirty_memory_managers{},
+      m_exclusive_monitor{}, m_memory{kernel.System()} {}
 KProcess::~KProcess() = default;
 
 Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std::size_t code_size,
                                   KProcessAddress aslr_space_start, bool is_hbl) {
     // Create a resource limit for the process.
-    const auto physical_memory_size =
-        m_kernel.MemoryManager().GetSize(Kernel::KMemoryManager::Pool::Application);
+    const auto pool = static_cast<KMemoryManager::Pool>(metadata.GetPoolPartition());
+    const auto physical_memory_size = m_kernel.MemoryManager().GetSize(pool);
     auto* res_limit =
         Kernel::CreateResourceLimitForProcess(m_kernel.System(), physical_memory_size);
 
@@ -1146,8 +1154,10 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     Svc::CreateProcessFlag flag{};
     u64 code_address{};
 
-    // We are an application.
-    flag |= Svc::CreateProcessFlag::IsApplication;
+    // Determine if we are an application.
+    if (pool == KMemoryManager::Pool::Application) {
+        flag |= Svc::CreateProcessFlag::IsApplication;
+    }
 
     // If we are 64-bit, create as such.
     if (metadata.Is64BitProgram()) {
@@ -1196,8 +1206,8 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
     std::memcpy(params.name.data(), name.data(), sizeof(params.name));
 
     // Initialize for application process.
-    R_TRY(this->Initialize(params, metadata.GetKernelCapabilities(), res_limit,
-                           KMemoryManager::Pool::Application, aslr_space_start));
+    R_TRY(this->Initialize(params, metadata.GetKernelCapabilities(), res_limit, pool,
+                           aslr_space_start));
 
     // Assign remaining properties.
     m_is_hbl = is_hbl;
@@ -1223,22 +1233,25 @@ void KProcess::LoadModule(CodeSet code_set, KProcessAddress base_addr) {
     ReprotectSegment(code_set.DataSegment(), Svc::MemoryPermission::ReadWrite);
 
 #ifdef HAS_NCE
-    if (Settings::IsNceEnabled()) {
+    if (this->IsApplication() && Settings::IsNceEnabled()) {
         auto& buffer = m_kernel.System().DeviceMemory().buffer;
         const auto& code = code_set.CodeSegment();
         const auto& patch = code_set.PatchSegment();
-        buffer.Protect(GetInteger(base_addr + code.addr), code.size, true, true, true);
-        buffer.Protect(GetInteger(base_addr + patch.addr), patch.size, true, true, true);
+        buffer.Protect(GetInteger(base_addr + code.addr), code.size,
+                       Common::MemoryPermission::Read | Common::MemoryPermission::Execute);
+        buffer.Protect(GetInteger(base_addr + patch.addr), patch.size,
+                       Common::MemoryPermission::Read | Common::MemoryPermission::Execute);
         ReprotectSegment(code_set.PatchSegment(), Svc::MemoryPermission::None);
     }
 #endif
 }
 
 void KProcess::InitializeInterfaces() {
-    this->GetMemory().SetCurrentPageTable(*this);
+    m_exclusive_monitor =
+        Core::MakeExclusiveMonitor(this->GetMemory(), Core::Hardware::NUM_CPU_CORES);
 
 #ifdef HAS_NCE
-    if (this->Is64Bit() && Settings::IsNceEnabled()) {
+    if (this->IsApplication() && Settings::IsNceEnabled()) {
         for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             m_arm_interfaces[i] = std::make_unique<Core::ArmNce>(m_kernel.System(), true, i);
         }
@@ -1248,13 +1261,13 @@ void KProcess::InitializeInterfaces() {
         for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             m_arm_interfaces[i] = std::make_unique<Core::ArmDynarmic64>(
                 m_kernel.System(), m_kernel.IsMulticore(), this,
-                static_cast<Core::DynarmicExclusiveMonitor&>(m_kernel.GetExclusiveMonitor()), i);
+                static_cast<Core::DynarmicExclusiveMonitor&>(*m_exclusive_monitor), i);
         }
     } else {
         for (size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             m_arm_interfaces[i] = std::make_unique<Core::ArmDynarmic32>(
                 m_kernel.System(), m_kernel.IsMulticore(), this,
-                static_cast<Core::DynarmicExclusiveMonitor&>(m_kernel.GetExclusiveMonitor()), i);
+                static_cast<Core::DynarmicExclusiveMonitor&>(*m_exclusive_monitor), i);
         }
     }
 }
@@ -1305,9 +1318,10 @@ bool KProcess::RemoveWatchpoint(KProcessAddress addr, u64 size, DebugWatchpointT
     return true;
 }
 
-Core::Memory::Memory& KProcess::GetMemory() const {
-    // TODO: per-process memory
-    return m_kernel.System().ApplicationMemory();
+void KProcess::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) {
+    for (auto& manager : m_dirty_memory_managers) {
+        manager.Gather(callback);
+    }
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_process.h

@@ -7,6 +7,7 @@
 
 #include "core/arm/arm_interface.h"
 #include "core/file_sys/program_metadata.h"
+#include "core/gpu_dirty_memory_manager.h"
 #include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/k_address_arbiter.h"
 #include "core/hle/kernel/k_capabilities.h"
@@ -17,6 +18,7 @@
 #include "core/hle/kernel/k_system_resource.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/k_thread_local_page.h"
+#include "core/memory.h"
 
 namespace Kernel {
 
@@ -126,6 +128,9 @@ private:
 #ifdef HAS_NCE
     std::unordered_map<u64, u64> m_post_handlers{};
 #endif
+    std::array<Core::GPUDirtyMemoryManager, Core::Hardware::NUM_CPU_CORES> m_dirty_memory_managers;
+    std::unique_ptr<Core::ExclusiveMonitor> m_exclusive_monitor;
+    Core::Memory::Memory m_memory;
 
 private:
     Result StartTermination();
@@ -502,7 +507,15 @@ public:
 
     void InitializeInterfaces();
 
-    Core::Memory::Memory& GetMemory() const;
+    Core::Memory::Memory& GetMemory() {
+        return m_memory;
+    }
+
+    void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback);
+
+    Core::ExclusiveMonitor& GetExclusiveMonitor() const {
+        return *m_exclusive_monitor;
+    }
 
 public:
     // Overridden parent functions.

src/core/hle/kernel/k_server_session.h

@@ -49,14 +49,21 @@ public:
     bool IsSignaled() const override;
     void OnClientClosed();
 
-    /// TODO: flesh these out to match the real kernel
     Result OnRequest(KSessionRequest* request);
-    Result SendReply(bool is_hle = false);
-    Result ReceiveRequest(std::shared_ptr<Service::HLERequestContext>* out_context = nullptr,
+    Result SendReply(uintptr_t server_message, uintptr_t server_buffer_size,
+                     KPhysicalAddress server_message_paddr, bool is_hle = false);
+    Result ReceiveRequest(uintptr_t server_message, uintptr_t server_buffer_size,
+                          KPhysicalAddress server_message_paddr,
+                          std::shared_ptr<Service::HLERequestContext>* out_context = nullptr,
                           std::weak_ptr<Service::SessionRequestManager> manager = {});
 
     Result SendReplyHLE() {
-        return SendReply(true);
+        R_RETURN(this->SendReply(0, 0, 0, true));
+    }
+
+    Result ReceiveRequestHLE(std::shared_ptr<Service::HLERequestContext>* out_context,
+                             std::weak_ptr<Service::SessionRequestManager> manager) {
+        R_RETURN(this->ReceiveRequest(0, 0, 0, out_context, manager));
     }
 
 private:

src/core/hle/kernel/k_session.cpp

@@ -33,8 +33,7 @@ void KSession::Initialize(KClientPort* client_port, uintptr_t name) {
     m_name = name;
 
     // Set our owner process.
-    //! FIXME: this is the wrong process!
-    m_process = m_kernel.ApplicationProcess();
+    m_process = GetCurrentProcessPointer(m_kernel);
     m_process->Open();
 
     // Set our port.

src/core/hle/kernel/k_thread.cpp

@@ -1422,8 +1422,7 @@ s32 GetCurrentCoreId(KernelCore& kernel) {
 }
 
 Core::Memory::Memory& GetCurrentMemory(KernelCore& kernel) {
-    // TODO: per-process memory
-    return kernel.System().ApplicationMemory();
+    return GetCurrentProcess(kernel).GetMemory();
 }
 
 KScopedDisableDispatch::~KScopedDisableDispatch() {

src/core/hle/kernel/k_thread.h

@@ -314,11 +314,7 @@ public:
         m_current_core_id = core;
     }
 
-    KProcess* GetOwnerProcess() {
-        return m_parent;
-    }
-
-    const KProcess* GetOwnerProcess() const {
+    KProcess* GetOwnerProcess() const {
         return m_parent;
     }
 

src/core/hle/kernel/k_transfer_memory.h

@@ -5,6 +5,7 @@
 
 #include <optional>
 
+#include "core/hle/kernel/k_light_lock.h"
 #include "core/hle/kernel/k_page_group.h"
 #include "core/hle/kernel/slab_helpers.h"
 #include "core/hle/kernel/svc_types.h"

src/core/hle/kernel/kernel.cpp

@@ -68,8 +68,6 @@ struct KernelCore::Impl {
 
         global_object_list_container = std::make_unique<KAutoObjectWithListContainer>(kernel);
         global_scheduler_context = std::make_unique<Kernel::GlobalSchedulerContext>(kernel);
-        global_handle_table = std::make_unique<Kernel::KHandleTable>(kernel);
-        global_handle_table->Initialize(KHandleTable::MaxTableSize);
 
         is_phantom_mode_for_singlecore = false;
 
@@ -121,13 +119,8 @@ struct KernelCore::Impl {
         next_user_process_id = KProcess::ProcessIdMin;
         next_thread_id = 1;
 
-        global_handle_table->Finalize();
-        global_handle_table.reset();
-
         preemption_event = nullptr;
 
-        exclusive_monitor.reset();
-
         // Cleanup persistent kernel objects
         auto CleanupObject = [](KAutoObject* obj) {
             if (obj) {
@@ -191,8 +184,6 @@ struct KernelCore::Impl {
     }
 
     void InitializePhysicalCores() {
-        exclusive_monitor =
-            Core::MakeExclusiveMonitor(system.ApplicationMemory(), Core::Hardware::NUM_CPU_CORES);
         for (u32 i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             const s32 core{static_cast<s32>(i)};
 
@@ -791,10 +782,6 @@ struct KernelCore::Impl {
 
     std::shared_ptr<Core::Timing::EventType> preemption_event;
 
-    // This is the kernel's handle table or supervisor handle table which
-    // stores all the objects in place.
-    std::unique_ptr<KHandleTable> global_handle_table;
-
     std::unique_ptr<KAutoObjectWithListContainer> global_object_list_container;
 
     std::unique_ptr<KObjectNameGlobalData> object_name_global_data;
@@ -805,7 +792,6 @@ struct KernelCore::Impl {
     std::mutex server_lock;
     std::vector<std::unique_ptr<Service::ServerManager>> server_managers;
 
-    std::unique_ptr<Core::ExclusiveMonitor> exclusive_monitor;
     std::array<std::unique_ptr<Kernel::PhysicalCore>, Core::Hardware::NUM_CPU_CORES> cores;
 
     // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
@@ -882,10 +868,6 @@ KResourceLimit* KernelCore::GetSystemResourceLimit() {
     return impl->system_resource_limit;
 }
 
-KScopedAutoObject<KThread> KernelCore::RetrieveThreadFromGlobalHandleTable(Handle handle) const {
-    return impl->global_handle_table->GetObject<KThread>(handle);
-}
-
 void KernelCore::AppendNewProcess(KProcess* process) {
     impl->process_list.push_back(process);
 }
@@ -959,14 +941,6 @@ Kernel::KHardwareTimer& KernelCore::HardwareTimer() {
     return *impl->hardware_timer;
 }
 
-Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() {
-    return *impl->exclusive_monitor;
-}
-
-const Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() const {
-    return *impl->exclusive_monitor;
-}
-
 KAutoObjectWithListContainer& KernelCore::ObjectListContainer() {
     return *impl->global_object_list_container;
 }
@@ -1030,14 +1004,6 @@ u64 KernelCore::CreateNewUserProcessID() {
     return impl->next_user_process_id++;
 }
 
-KHandleTable& KernelCore::GlobalHandleTable() {
-    return *impl->global_handle_table;
-}
-
-const KHandleTable& KernelCore::GlobalHandleTable() const {
-    return *impl->global_handle_table;
-}
-
 void KernelCore::RegisterCoreThread(std::size_t core_id) {
     impl->RegisterCoreThread(core_id);
 }

src/core/hle/kernel/kernel.h

@@ -116,9 +116,6 @@ public:
     /// Retrieves a shared pointer to the system resource limit instance.
     KResourceLimit* GetSystemResourceLimit();
 
-    /// Retrieves a shared pointer to a Thread instance within the thread wakeup handle table.
-    KScopedAutoObject<KThread> RetrieveThreadFromGlobalHandleTable(Handle handle) const;
-
     /// Adds the given shared pointer to an internal list of active processes.
     void AppendNewProcess(KProcess* process);
 
@@ -170,10 +167,6 @@ public:
     /// Stops execution of 'id' core, in order to reschedule a new thread.
     void PrepareReschedule(std::size_t id);
 
-    Core::ExclusiveMonitor& GetExclusiveMonitor();
-
-    const Core::ExclusiveMonitor& GetExclusiveMonitor() const;
-
     KAutoObjectWithListContainer& ObjectListContainer();
 
     const KAutoObjectWithListContainer& ObjectListContainer() const;

src/core/hle/kernel/message_buffer.h

@@ -18,13 +18,13 @@ public:
         static constexpr inline u64 NullTag = 0;
 
     public:
-        enum class ReceiveListCountType : u32 {
-            None = 0,
-            ToMessageBuffer = 1,
-            ToSingleBuffer = 2,
+        enum ReceiveListCountType : u32 {
+            ReceiveListCountType_None = 0,
+            ReceiveListCountType_ToMessageBuffer = 1,
+            ReceiveListCountType_ToSingleBuffer = 2,
 
-            CountOffset = 2,
-            CountMax = 13,
+            ReceiveListCountType_CountOffset = 2,
+            ReceiveListCountType_CountMax = 13,
         };
 
     private:
@@ -591,16 +591,16 @@ public:
         // Add the size of the receive list.
         const auto count = hdr.GetReceiveListCount();
         switch (count) {
-        case MessageHeader::ReceiveListCountType::None:
+        case MessageHeader::ReceiveListCountType_None:
             break;
-        case MessageHeader::ReceiveListCountType::ToMessageBuffer:
+        case MessageHeader::ReceiveListCountType_ToMessageBuffer:
             break;
-        case MessageHeader::ReceiveListCountType::ToSingleBuffer:
+        case MessageHeader::ReceiveListCountType_ToSingleBuffer:
             msg_size += ReceiveListEntry::GetDataSize();
             break;
         default:
             msg_size += (static_cast<s32>(count) -
-                         static_cast<s32>(MessageHeader::ReceiveListCountType::CountOffset)) *
+                         static_cast<s32>(MessageHeader::ReceiveListCountType_CountOffset)) *
                         ReceiveListEntry::GetDataSize();
             break;
         }

src/core/hle/kernel/svc/svc_info.cpp

@@ -118,7 +118,6 @@ Result GetInfo(Core::System& system, u64* result, InfoType info_id_type, Handle
             R_SUCCEED();
 
         case InfoType::IsApplication:
-            LOG_WARNING(Kernel_SVC, "(STUBBED) Assuming process is application");
             *result = process->IsApplication();
             R_SUCCEED();
 

src/core/hle/kernel/svc/svc_ipc.cpp

@@ -48,8 +48,7 @@ Result ReplyAndReceiveImpl(KernelCore& kernel, int32_t* out_index, uintptr_t mes
         };
 
         // Send the reply.
-        R_TRY(session->SendReply());
-        // R_TRY(session->SendReply(message, buffer_size, message_paddr));
+        R_TRY(session->SendReply(message, buffer_size, message_paddr));
     }
 
     // Receive a message.
@@ -85,8 +84,7 @@ Result ReplyAndReceiveImpl(KernelCore& kernel, int32_t* out_index, uintptr_t mes
             if (R_SUCCEEDED(result)) {
                 KServerSession* session = objs[index]->DynamicCast<KServerSession*>();
                 if (session != nullptr) {
-                    // result = session->ReceiveRequest(message, buffer_size, message_paddr);
-                    result = session->ReceiveRequest();
+                    result = session->ReceiveRequest(message, buffer_size, message_paddr);
                     if (ResultNotFound == result) {
                         continue;
                     }

src/core/hle/kernel/svc_results.h

@@ -38,7 +38,9 @@ constexpr Result ResultInvalidState{ErrorModule::Kernel, 125};
 constexpr Result ResultReservedUsed{ErrorModule::Kernel, 126};
 constexpr Result ResultPortClosed{ErrorModule::Kernel, 131};
 constexpr Result ResultLimitReached{ErrorModule::Kernel, 132};
+constexpr Result ResultReceiveListBroken{ErrorModule::Kernel, 258};
 constexpr Result ResultOutOfAddressSpace{ErrorModule::Kernel, 259};
+constexpr Result ResultMessageTooLarge{ErrorModule::Kernel, 260};
 constexpr Result ResultInvalidId{ErrorModule::Kernel, 519};
 
 } // namespace Kernel

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

@@ -1513,8 +1513,7 @@ void ILibraryAppletCreator::CreateTransferMemoryStorage(HLERequestContext& ctx)
         return;
     }
 
-    auto transfer_mem =
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(handle);
+    auto transfer_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(handle);
 
     if (transfer_mem.IsNull()) {
         LOG_ERROR(Service_AM, "transfer_mem is a nullptr for handle={:08X}", handle);
@@ -1524,8 +1523,7 @@ void ILibraryAppletCreator::CreateTransferMemoryStorage(HLERequestContext& ctx)
     }
 
     std::vector<u8> memory(transfer_mem->GetSize());
-    system.ApplicationMemory().ReadBlock(transfer_mem->GetSourceAddress(), memory.data(),
-                                         memory.size());
+    ctx.GetMemory().ReadBlock(transfer_mem->GetSourceAddress(), memory.data(), memory.size());
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(ResultSuccess);
@@ -1547,8 +1545,7 @@ void ILibraryAppletCreator::CreateHandleStorage(HLERequestContext& ctx) {
         return;
     }
 
-    auto transfer_mem =
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(handle);
+    auto transfer_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(handle);
 
     if (transfer_mem.IsNull()) {
         LOG_ERROR(Service_AM, "transfer_mem is a nullptr for handle={:08X}", handle);
@@ -1558,8 +1555,7 @@ void ILibraryAppletCreator::CreateHandleStorage(HLERequestContext& ctx) {
     }
 
     std::vector<u8> memory(transfer_mem->GetSize());
-    system.ApplicationMemory().ReadBlock(transfer_mem->GetSourceAddress(), memory.data(),
-                                         memory.size());
+    ctx.GetMemory().ReadBlock(transfer_mem->GetSourceAddress(), memory.data(), memory.size());
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(ResultSuccess);

src/core/hle/service/audio/audren_u.cpp

@@ -454,10 +454,8 @@ void AudRenU::OpenAudioRenderer(HLERequestContext& ctx) {
         return;
     }
 
-    const auto& handle_table{system.ApplicationProcess()->GetHandleTable()};
-    auto process{handle_table.GetObject<Kernel::KProcess>(process_handle)};
-    auto transfer_memory{
-        process->GetHandleTable().GetObject<Kernel::KTransferMemory>(transfer_memory_handle)};
+    auto process{ctx.GetObjectFromHandle<Kernel::KProcess>(process_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     const auto session_id{impl->GetSessionId()};
     if (session_id == -1) {

src/core/hle/service/audio/hwopus.cpp

@@ -278,9 +278,7 @@ void HwOpus::OpenHardwareOpusDecoder(HLERequestContext& ctx) {
     auto params = rp.PopRaw<OpusParameters>();
     auto transfer_memory_size{rp.Pop<u32>()};
     auto transfer_memory_handle{ctx.GetCopyHandle(0)};
-    auto transfer_memory{
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-            transfer_memory_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     LOG_DEBUG(Service_Audio, "sample_rate {} channel_count {} transfer_memory_size 0x{:X}",
               params.sample_rate, params.channel_count, transfer_memory_size);
@@ -323,9 +321,7 @@ void HwOpus::OpenHardwareOpusDecoderForMultiStream(HLERequestContext& ctx) {
 
     auto transfer_memory_size{rp.Pop<u32>()};
     auto transfer_memory_handle{ctx.GetCopyHandle(0)};
-    auto transfer_memory{
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-            transfer_memory_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     LOG_DEBUG(Service_Audio,
               "sample_rate {} channel_count {} total_stream_count {} stereo_stream_count {} "
@@ -374,9 +370,7 @@ void HwOpus::OpenHardwareOpusDecoderEx(HLERequestContext& ctx) {
     auto params = rp.PopRaw<OpusParametersEx>();
     auto transfer_memory_size{rp.Pop<u32>()};
     auto transfer_memory_handle{ctx.GetCopyHandle(0)};
-    auto transfer_memory{
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-            transfer_memory_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     LOG_DEBUG(Service_Audio, "sample_rate {} channel_count {} transfer_memory_size 0x{:X}",
               params.sample_rate, params.channel_count, transfer_memory_size);
@@ -414,9 +408,7 @@ void HwOpus::OpenHardwareOpusDecoderForMultiStreamEx(HLERequestContext& ctx) {
 
     auto transfer_memory_size{rp.Pop<u32>()};
     auto transfer_memory_handle{ctx.GetCopyHandle(0)};
-    auto transfer_memory{
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-            transfer_memory_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     LOG_DEBUG(Service_Audio,
               "sample_rate {} channel_count {} total_stream_count {} stereo_stream_count {} "

src/core/hle/service/fatal/fatal.cpp

@@ -89,7 +89,7 @@ static void GenerateErrorReport(Core::System& system, Result error_code, const F
         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++) {
+        for (u32 i = 0; i < std::min<u32>(info.backtrace_size, 32); i++) {
             crash_report +=
                 fmt::format("    Backtrace[{:02d}]:               {:016x}\n", i, info.backtrace[i]);
         }

src/core/hle/service/hid/hid_server.cpp

@@ -1850,8 +1850,7 @@ void IHidServer::InitializeSevenSixAxisSensor(HLERequestContext& ctx) {
     ASSERT_MSG(t_mem_1_size == 0x1000, "t_mem_1_size is not 0x1000 bytes");
     ASSERT_MSG(t_mem_2_size == 0x7F000, "t_mem_2_size is not 0x7F000 bytes");
 
-    auto t_mem_1 = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_1_handle);
+    auto t_mem_1 = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_1_handle);
 
     if (t_mem_1.IsNull()) {
         LOG_ERROR(Service_HID, "t_mem_1 is a nullptr for handle=0x{:08X}", t_mem_1_handle);
@@ -1860,8 +1859,7 @@ void IHidServer::InitializeSevenSixAxisSensor(HLERequestContext& ctx) {
         return;
     }
 
-    auto t_mem_2 = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_2_handle);
+    auto t_mem_2 = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_2_handle);
 
     if (t_mem_2.IsNull()) {
         LOG_ERROR(Service_HID, "t_mem_2 is a nullptr for handle=0x{:08X}", t_mem_2_handle);
@@ -2142,8 +2140,7 @@ void IHidServer::WritePalmaWaveEntry(HLERequestContext& ctx) {
 
     ASSERT_MSG(t_mem_size == 0x3000, "t_mem_size is not 0x3000 bytes");
 
-    auto t_mem = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_handle);
+    auto t_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_handle);
 
     if (t_mem.IsNull()) {
         LOG_ERROR(Service_HID, "t_mem is a nullptr for handle=0x{:08X}", t_mem_handle);

src/core/hle/service/hid/hidbus.cpp

@@ -448,8 +448,7 @@ void HidBus::EnableJoyPollingReceiveMode(HLERequestContext& ctx) {
 
     ASSERT_MSG(t_mem_size == 0x1000, "t_mem_size is not 0x1000 bytes");
 
-    auto t_mem = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_handle);
+    auto t_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_handle);
 
     if (t_mem.IsNull()) {
         LOG_ERROR(Service_HID, "t_mem is a nullptr for handle=0x{:08X}", t_mem_handle);

src/core/hle/service/hid/irs.cpp

@@ -197,8 +197,7 @@ void IRS::RunImageTransferProcessor(HLERequestContext& ctx) {
     const auto parameters{rp.PopRaw<Parameters>()};
     const auto t_mem_handle{ctx.GetCopyHandle(0)};
 
-    auto t_mem = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_handle);
+    auto t_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_handle);
 
     if (t_mem.IsNull()) {
         LOG_ERROR(Service_IRS, "t_mem is a nullptr for handle=0x{:08X}", t_mem_handle);
@@ -444,8 +443,7 @@ void IRS::RunImageTransferExProcessor(HLERequestContext& ctx) {
     const auto parameters{rp.PopRaw<Parameters>()};
     const auto t_mem_handle{ctx.GetCopyHandle(0)};
 
-    auto t_mem = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_handle);
+    auto t_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_handle);
 
     LOG_INFO(Service_IRS,
              "called, npad_type={}, npad_id={}, transfer_memory_size={}, "

src/core/hle/service/hle_ipc.cpp

@@ -146,10 +146,7 @@ HLERequestContext::HLERequestContext(Kernel::KernelCore& kernel_, Core::Memory::
 
 HLERequestContext::~HLERequestContext() = default;
 
-void HLERequestContext::ParseCommandBuffer(Kernel::KProcess& process, u32_le* src_cmdbuf,
-                                           bool incoming) {
-    client_handle_table = &process.GetHandleTable();
-
+void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
     IPC::RequestParser rp(src_cmdbuf);
     command_header = rp.PopRaw<IPC::CommandHeader>();
 
@@ -162,7 +159,7 @@ void HLERequestContext::ParseCommandBuffer(Kernel::KProcess& process, u32_le* sr
     if (command_header->enable_handle_descriptor) {
         handle_descriptor_header = rp.PopRaw<IPC::HandleDescriptorHeader>();
         if (handle_descriptor_header->send_current_pid) {
-            pid = process.GetProcessId();
+            pid = thread->GetOwnerProcess()->GetProcessId();
             rp.Skip(2, false);
         }
         if (incoming) {
@@ -270,9 +267,10 @@ void HLERequestContext::ParseCommandBuffer(Kernel::KProcess& process, u32_le* sr
     rp.Skip(1, false); // The command is actually an u64, but we don't use the high part.
 }
 
-Result HLERequestContext::PopulateFromIncomingCommandBuffer(Kernel::KProcess& process,
-                                                            u32_le* src_cmdbuf) {
-    ParseCommandBuffer(process, src_cmdbuf, true);
+Result HLERequestContext::PopulateFromIncomingCommandBuffer(u32_le* src_cmdbuf) {
+    client_handle_table = &thread->GetOwnerProcess()->GetHandleTable();
+
+    ParseCommandBuffer(src_cmdbuf, true);
 
     if (command_header->IsCloseCommand()) {
         // Close does not populate the rest of the IPC header
@@ -284,9 +282,9 @@ Result HLERequestContext::PopulateFromIncomingCommandBuffer(Kernel::KProcess& pr
     return ResultSuccess;
 }
 
-Result HLERequestContext::WriteToOutgoingCommandBuffer(Kernel::KThread& requesting_thread) {
+Result HLERequestContext::WriteToOutgoingCommandBuffer() {
     auto current_offset = handles_offset;
-    auto& owner_process = *requesting_thread.GetOwnerProcess();
+    auto& owner_process = *thread->GetOwnerProcess();
     auto& handle_table = owner_process.GetHandleTable();
 
     for (auto& object : outgoing_copy_objects) {
@@ -319,7 +317,7 @@ Result HLERequestContext::WriteToOutgoingCommandBuffer(Kernel::KThread& requesti
     }
 
     // Copy the translated command buffer back into the thread's command buffer area.
-    memory.WriteBlock(requesting_thread.GetTlsAddress(), cmd_buf.data(), write_size * sizeof(u32));
+    memory.WriteBlock(thread->GetTlsAddress(), cmd_buf.data(), write_size * sizeof(u32));
 
     return ResultSuccess;
 }

src/core/hle/service/hle_ipc.h

@@ -17,6 +17,7 @@
 #include "common/concepts.h"
 #include "common/swap.h"
 #include "core/hle/ipc.h"
+#include "core/hle/kernel/k_handle_table.h"
 #include "core/hle/kernel/svc_common.h"
 
 union Result;
@@ -196,10 +197,10 @@ public:
     }
 
     /// Populates this context with data from the requesting process/thread.
-    Result PopulateFromIncomingCommandBuffer(Kernel::KProcess& process, u32_le* src_cmdbuf);
+    Result PopulateFromIncomingCommandBuffer(u32_le* src_cmdbuf);
 
     /// Writes data from this context back to the requesting process/thread.
-    Result WriteToOutgoingCommandBuffer(Kernel::KThread& requesting_thread);
+    Result WriteToOutgoingCommandBuffer();
 
     [[nodiscard]] u32_le GetHipcCommand() const {
         return command;
@@ -359,8 +360,17 @@ public:
         return *thread;
     }
 
-    Kernel::KHandleTable& GetClientHandleTable() {
-        return *client_handle_table;
+    [[nodiscard]] Core::Memory::Memory& GetMemory() const {
+        return memory;
+    }
+
+    template <typename T>
+    Kernel::KScopedAutoObject<T> GetObjectFromHandle(u32 handle) {
+        auto obj = client_handle_table->GetObjectForIpc(handle, thread);
+        if (obj.IsNotNull()) {
+            return obj->DynamicCast<T*>();
+        }
+        return nullptr;
     }
 
     [[nodiscard]] std::shared_ptr<SessionRequestManager> GetManager() const {
@@ -378,7 +388,7 @@ public:
 private:
     friend class IPC::ResponseBuilder;
 
-    void ParseCommandBuffer(Kernel::KProcess& process, u32_le* src_cmdbuf, bool incoming);
+    void ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming);
 
     std::array<u32, IPC::COMMAND_BUFFER_LENGTH> cmd_buf;
     Kernel::KServerSession* server_session{};

src/core/hle/service/ipc_helpers.h

@@ -151,8 +151,8 @@ public:
         if (manager->IsDomain()) {
             context->AddDomainObject(std::move(iface));
         } else {
-            kernel.ApplicationProcess()->GetResourceLimit()->Reserve(
-                Kernel::LimitableResource::SessionCountMax, 1);
+            ASSERT(Kernel::GetCurrentProcess(kernel).GetResourceLimit()->Reserve(
+                Kernel::LimitableResource::SessionCountMax, 1));
 
             auto* session = Kernel::KSession::Create(kernel);
             session->Initialize(nullptr, 0);

src/core/hle/service/jit/jit.cpp

@@ -26,7 +26,7 @@ public:
     explicit IJitEnvironment(Core::System& system_, Kernel::KProcess& process_, CodeRange user_rx,
                              CodeRange user_ro)
         : ServiceFramework{system_, "IJitEnvironment"}, process{&process_},
-          context{system_.ApplicationMemory()} {
+          context{process->GetMemory()} {
         // clang-format off
         static const FunctionInfo functions[] = {
             {0, &IJitEnvironment::GenerateCode, "GenerateCode"},
@@ -188,7 +188,7 @@ public:
             return;
         }
 
-        auto tmem{process->GetHandleTable().GetObject<Kernel::KTransferMemory>(tmem_handle)};
+        auto tmem{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(tmem_handle)};
         if (tmem.IsNull()) {
             LOG_ERROR(Service_JIT, "attempted to load plugin with invalid transfer memory handle");
             IPC::ResponseBuilder rb{ctx, 2};
@@ -356,11 +356,7 @@ public:
             return;
         }
 
-        // Fetch using the handle table for the application process here,
-        // since we are not multiprocess yet.
-        const auto& handle_table{system.ApplicationProcess()->GetHandleTable()};
-
-        auto process{handle_table.GetObject<Kernel::KProcess>(process_handle)};
+        auto process{ctx.GetObjectFromHandle<Kernel::KProcess>(process_handle)};
         if (process.IsNull()) {
             LOG_ERROR(Service_JIT, "process is null for handle=0x{:08X}", process_handle);
             IPC::ResponseBuilder rb{ctx, 2};
@@ -368,7 +364,7 @@ public:
             return;
         }
 
-        auto rx_mem{handle_table.GetObject<Kernel::KCodeMemory>(rx_mem_handle)};
+        auto rx_mem{ctx.GetObjectFromHandle<Kernel::KCodeMemory>(rx_mem_handle)};
         if (rx_mem.IsNull()) {
             LOG_ERROR(Service_JIT, "rx_mem is null for handle=0x{:08X}", rx_mem_handle);
             IPC::ResponseBuilder rb{ctx, 2};
@@ -376,7 +372,7 @@ public:
             return;
         }
 
-        auto ro_mem{handle_table.GetObject<Kernel::KCodeMemory>(ro_mem_handle)};
+        auto ro_mem{ctx.GetObjectFromHandle<Kernel::KCodeMemory>(ro_mem_handle)};
         if (ro_mem.IsNull()) {
             LOG_ERROR(Service_JIT, "ro_mem is null for handle=0x{:08X}", ro_mem_handle);
             IPC::ResponseBuilder rb{ctx, 2};

src/core/hle/service/ro/ro.cpp

@@ -651,10 +651,9 @@ private:
     void RegisterProcessHandle(HLERequestContext& ctx) {
         LOG_DEBUG(Service_LDR, "(called)");
 
-        auto process_h = ctx.GetClientHandleTable().GetObject(ctx.GetCopyHandle(0));
+        auto process = ctx.GetObjectFromHandle<Kernel::KProcess>(ctx.GetCopyHandle(0));
         auto client_pid = ctx.GetPID();
-        auto result = interface.RegisterProcessHandle(client_pid,
-                                                      process_h->DynamicCast<Kernel::KProcess*>());
+        auto result = interface.RegisterProcessHandle(client_pid, process.GetPointerUnsafe());
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);
@@ -671,12 +670,11 @@ private:
 
         IPC::RequestParser rp{ctx};
         auto params = rp.PopRaw<InputParameters>();
-        auto process_h = ctx.GetClientHandleTable().GetObject(ctx.GetCopyHandle(0));
+        auto process = ctx.GetObjectFromHandle<Kernel::KProcess>(ctx.GetCopyHandle(0));
 
         auto client_pid = ctx.GetPID();
-        auto result =
-            interface.RegisterProcessModuleInfo(client_pid, params.nrr_address, params.nrr_size,
-                                                process_h->DynamicCast<Kernel::KProcess*>());
+        auto result = interface.RegisterProcessModuleInfo(
+            client_pid, params.nrr_address, params.nrr_size, process.GetPointerUnsafe());
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);

src/core/hle/service/server_manager.cpp

@@ -47,7 +47,7 @@ ServerManager::~ServerManager() {
     m_stopped.Wait();
     m_threads.clear();
 
-    // Clean up ports.
+    // Clean up server ports.
     for (const auto& [port, handler] : m_ports) {
         port->Close();
     }
@@ -97,22 +97,15 @@ Result ServerManager::RegisterNamedService(const std::string& service_name,
                                            u32 max_sessions) {
     ASSERT(m_sessions.size() + m_ports.size() < MaximumWaitObjects);
 
-    // Add the new server to sm:.
-    ASSERT(R_SUCCEEDED(
-        m_system.ServiceManager().RegisterService(service_name, max_sessions, handler_factory)));
-
-    // Get the registered port.
-    Kernel::KPort* port{};
-    ASSERT(
-        R_SUCCEEDED(m_system.ServiceManager().GetServicePort(std::addressof(port), service_name)));
-
-    // Open a new reference to the server port.
-    port->GetServerPort().Open();
+    // Add the new server to sm: and get the moved server port.
+    Kernel::KServerPort* server_port{};
+    R_ASSERT(m_system.ServiceManager().RegisterService(std::addressof(server_port), service_name,
+                                                       max_sessions, handler_factory));
 
     // Begin tracking the server port.
     {
         std::scoped_lock ll{m_list_mutex};
-        m_ports.emplace(std::addressof(port->GetServerPort()), std::move(handler_factory));
+        m_ports.emplace(server_port, std::move(handler_factory));
     }
 
     // Signal the wakeup event.
@@ -372,7 +365,7 @@ Result ServerManager::OnSessionEvent(Kernel::KServerSession* session,
 
     // Try to receive a message.
     std::shared_ptr<HLERequestContext> context;
-    rc = session->ReceiveRequest(&context, manager);
+    rc = session->ReceiveRequestHLE(&context, manager);
 
     // If the session has been closed, we're done.
     if (rc == Kernel::ResultSessionClosed) {

src/core/hle/service/service.cpp

@@ -203,7 +203,7 @@ Result ServiceFrameworkBase::HandleSyncRequest(Kernel::KServerSession& session,
     // If emulation was shutdown, we are closing service threads, do not write the response back to
     // memory that may be shutting down as well.
     if (system.IsPoweredOn()) {
-        ctx.WriteToOutgoingCommandBuffer(ctx.GetThread());
+        ctx.WriteToOutgoingCommandBuffer();
     }
 
     return result;

src/core/hle/service/set/set_sys.cpp

@@ -507,6 +507,14 @@ void SET_SYS::SetTvSettings(HLERequestContext& ctx) {
     rb.Push(ResultSuccess);
 }
 
+void SET_SYS::GetDebugModeFlag(HLERequestContext& ctx) {
+    LOG_DEBUG(Service_SET, "called");
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(ResultSuccess);
+    rb.Push<u32>(0);
+}
+
 void SET_SYS::GetQuestFlag(HLERequestContext& ctx) {
     LOG_WARNING(Service_SET, "(STUBBED) called");
 
@@ -926,7 +934,7 @@ SET_SYS::SET_SYS(Core::System& system_) : ServiceFramework{system_, "set:sys"},
         {59, &SET_SYS::SetNetworkSystemClockContext, "SetNetworkSystemClockContext"},
         {60, &SET_SYS::IsUserSystemClockAutomaticCorrectionEnabled, "IsUserSystemClockAutomaticCorrectionEnabled"},
         {61, &SET_SYS::SetUserSystemClockAutomaticCorrectionEnabled, "SetUserSystemClockAutomaticCorrectionEnabled"},
-        {62, nullptr, "GetDebugModeFlag"},
+        {62, &SET_SYS::GetDebugModeFlag, "GetDebugModeFlag"},
         {63, &SET_SYS::GetPrimaryAlbumStorage, "GetPrimaryAlbumStorage"},
         {64, nullptr, "SetPrimaryAlbumStorage"},
         {65, nullptr, "GetUsb30EnableFlag"},
@@ -1143,6 +1151,8 @@ void SET_SYS::StoreSettings() {
 }
 
 void SET_SYS::StoreSettingsThreadFunc(std::stop_token stop_token) {
+    Common::SetCurrentThreadName("SettingsStore");
+
     while (Common::StoppableTimedWait(stop_token, std::chrono::minutes(1))) {
         std::scoped_lock l{m_save_needed_mutex};
         if (!std::exchange(m_save_needed, false)) {

src/core/hle/service/set/set_sys.h

@@ -98,6 +98,7 @@ private:
     void GetSettingsItemValue(HLERequestContext& ctx);
     void GetTvSettings(HLERequestContext& ctx);
     void SetTvSettings(HLERequestContext& ctx);
+    void GetDebugModeFlag(HLERequestContext& ctx);
     void GetQuestFlag(HLERequestContext& ctx);
     void GetDeviceTimeZoneLocationName(HLERequestContext& ctx);
     void SetDeviceTimeZoneLocationName(HLERequestContext& ctx);

src/core/hle/service/sm/sm.cpp

@@ -29,8 +29,7 @@ ServiceManager::ServiceManager(Kernel::KernelCore& kernel_) : kernel{kernel_} {
 
 ServiceManager::~ServiceManager() {
     for (auto& [name, port] : service_ports) {
-        port->GetClientPort().Close();
-        port->GetServerPort().Close();
+        port->Close();
     }
 
     if (deferral_event) {
@@ -50,8 +49,8 @@ static Result ValidateServiceName(const std::string& name) {
     return ResultSuccess;
 }
 
-Result ServiceManager::RegisterService(std::string name, u32 max_sessions,
-                                       SessionRequestHandlerFactory handler) {
+Result ServiceManager::RegisterService(Kernel::KServerPort** out_server_port, std::string name,
+                                       u32 max_sessions, SessionRequestHandlerFactory handler) {
     R_TRY(ValidateServiceName(name));
 
     std::scoped_lock lk{lock};
@@ -66,13 +65,17 @@ Result ServiceManager::RegisterService(std::string name, u32 max_sessions,
     // Register the port.
     Kernel::KPort::Register(kernel, port);
 
-    service_ports.emplace(name, port);
+    service_ports.emplace(name, std::addressof(port->GetClientPort()));
     registered_services.emplace(name, handler);
     if (deferral_event) {
         deferral_event->Signal();
     }
 
-    return ResultSuccess;
+    // Set our output.
+    *out_server_port = std::addressof(port->GetServerPort());
+
+    // We succeeded.
+    R_SUCCEED();
 }
 
 Result ServiceManager::UnregisterService(const std::string& name) {
@@ -91,7 +94,8 @@ Result ServiceManager::UnregisterService(const std::string& name) {
     return ResultSuccess;
 }
 
-Result ServiceManager::GetServicePort(Kernel::KPort** out_port, const std::string& name) {
+Result ServiceManager::GetServicePort(Kernel::KClientPort** out_client_port,
+                                      const std::string& name) {
     R_TRY(ValidateServiceName(name));
 
     std::scoped_lock lk{lock};
@@ -101,7 +105,7 @@ Result ServiceManager::GetServicePort(Kernel::KPort** out_port, const std::strin
         return Service::SM::ResultNotRegistered;
     }
 
-    *out_port = it->second;
+    *out_client_port = it->second;
     return ResultSuccess;
 }
 
@@ -172,8 +176,8 @@ Result SM::GetServiceImpl(Kernel::KClientSession** out_client_session, HLEReques
     std::string name(PopServiceName(rp));
 
     // Find the named port.
-    Kernel::KPort* port{};
-    auto port_result = service_manager.GetServicePort(&port, name);
+    Kernel::KClientPort* client_port{};
+    auto port_result = service_manager.GetServicePort(&client_port, name);
     if (port_result == Service::SM::ResultInvalidServiceName) {
         LOG_ERROR(Service_SM, "Invalid service name '{}'", name);
         return Service::SM::ResultInvalidServiceName;
@@ -187,7 +191,7 @@ Result SM::GetServiceImpl(Kernel::KClientSession** out_client_session, HLEReques
 
     // Create a new session.
     Kernel::KClientSession* session{};
-    if (const auto result = port->GetClientPort().CreateSession(&session); result.IsError()) {
+    if (const auto result = client_port->CreateSession(&session); result.IsError()) {
         LOG_ERROR(Service_SM, "called service={} -> error 0x{:08X}", name, result.raw);
         return result;
     }
@@ -221,7 +225,9 @@ void SM::RegisterServiceImpl(HLERequestContext& ctx, std::string name, u32 max_s
     LOG_DEBUG(Service_SM, "called with name={}, max_session_count={}, is_light={}", name,
               max_session_count, is_light);
 
-    if (const auto result = service_manager.RegisterService(name, max_session_count, nullptr);
+    Kernel::KServerPort* server_port{};
+    if (const auto result = service_manager.RegisterService(std::addressof(server_port), name,
+                                                            max_session_count, nullptr);
         result.IsError()) {
         LOG_ERROR(Service_SM, "failed to register service with error_code={:08X}", result.raw);
         IPC::ResponseBuilder rb{ctx, 2};
@@ -229,13 +235,9 @@ void SM::RegisterServiceImpl(HLERequestContext& ctx, std::string name, u32 max_s
         return;
     }
 
-    auto* port = Kernel::KPort::Create(kernel);
-    port->Initialize(ServerSessionCountMax, is_light, 0);
-    SCOPE_EXIT({ port->GetClientPort().Close(); });
-
     IPC::ResponseBuilder rb{ctx, 2, 0, 1, IPC::ResponseBuilder::Flags::AlwaysMoveHandles};
     rb.Push(ResultSuccess);
-    rb.PushMoveObjects(port->GetServerPort());
+    rb.PushMoveObjects(server_port);
 }
 
 void SM::UnregisterService(HLERequestContext& ctx) {

src/core/hle/service/sm/sm.h

@@ -56,10 +56,10 @@ public:
     explicit ServiceManager(Kernel::KernelCore& kernel_);
     ~ServiceManager();
 
-    Result RegisterService(std::string name, u32 max_sessions,
-                           SessionRequestHandlerFactory handler_factory);
+    Result RegisterService(Kernel::KServerPort** out_server_port, std::string name,
+                           u32 max_sessions, SessionRequestHandlerFactory handler_factory);
     Result UnregisterService(const std::string& name);
-    Result GetServicePort(Kernel::KPort** out_port, const std::string& name);
+    Result GetServicePort(Kernel::KClientPort** out_client_port, const std::string& name);
 
     template <Common::DerivedFrom<SessionRequestHandler> T>
     std::shared_ptr<T> GetService(const std::string& service_name) const {
@@ -84,7 +84,7 @@ private:
     /// Map of registered services, retrieved using GetServicePort.
     std::mutex lock;
     std::unordered_map<std::string, SessionRequestHandlerFactory> registered_services;
-    std::unordered_map<std::string, Kernel::KPort*> service_ports;
+    std::unordered_map<std::string, Kernel::KClientPort*> service_ports;
 
     /// Kernel context
     Kernel::KernelCore& kernel;

src/core/hle/service/sm/sm_controller.cpp

@@ -28,7 +28,6 @@ void Controller::ConvertCurrentObjectToDomain(HLERequestContext& ctx) {
 void Controller::CloneCurrentObject(HLERequestContext& ctx) {
     LOG_DEBUG(Service, "called");
 
-    auto& process = *ctx.GetThread().GetOwnerProcess();
     auto session_manager = ctx.GetManager();
 
     // FIXME: this is duplicated from the SVC, it should just call it instead
@@ -36,11 +35,11 @@ void Controller::CloneCurrentObject(HLERequestContext& ctx) {
 
     // Reserve a new session from the process resource limit.
     Kernel::KScopedResourceReservation session_reservation(
-        &process, Kernel::LimitableResource::SessionCountMax);
+        Kernel::GetCurrentProcessPointer(kernel), Kernel::LimitableResource::SessionCountMax);
     ASSERT(session_reservation.Succeeded());
 
     // Create the session.
-    Kernel::KSession* session = Kernel::KSession::Create(system.Kernel());
+    Kernel::KSession* session = Kernel::KSession::Create(kernel);
     ASSERT(session != nullptr);
 
     // Initialize the session.
@@ -50,7 +49,7 @@ void Controller::CloneCurrentObject(HLERequestContext& ctx) {
     session_reservation.Commit();
 
     // Register the session.
-    Kernel::KSession::Register(system.Kernel(), session);
+    Kernel::KSession::Register(kernel, session);
 
     // Register with server manager.
     session_manager->GetServerManager().RegisterSession(&session->GetServerSession(),

src/core/loader/deconstructed_rom_directory.cpp

@@ -129,9 +129,10 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
     }
     metadata.Print();
 
-    // Enable NCE only for programs with 39-bit address space.
+    // Enable NCE only for applications with 39-bit address space.
     const bool is_39bit =
         metadata.GetAddressSpaceType() == FileSys::ProgramAddressSpaceType::Is39Bit;
+    const bool is_application = metadata.GetPoolPartition() == FileSys::PoolPartition::Application;
     Settings::SetNceEnabled(is_39bit);
 
     const std::array static_modules = {"rtld",    "main",    "subsdk0", "subsdk1", "subsdk2",
@@ -147,7 +148,7 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
 
     const auto GetPatcher = [&](size_t i) -> Core::NCE::Patcher* {
 #ifdef HAS_NCE
-        if (Settings::IsNceEnabled()) {
+        if (is_application && Settings::IsNceEnabled()) {
             return &module_patchers[i];
         }
 #endif
@@ -175,7 +176,7 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
 
     // Enable direct memory mapping in case of NCE.
     const u64 fastmem_base = [&]() -> size_t {
-        if (Settings::IsNceEnabled()) {
+        if (is_application && Settings::IsNceEnabled()) {
             auto& buffer = system.DeviceMemory().buffer;
             buffer.EnableDirectMappedAddress();
             return reinterpret_cast<u64>(buffer.VirtualBasePointer());

src/core/memory.cpp

@@ -10,6 +10,7 @@
 #include "common/assert.h"
 #include "common/atomic_ops.h"
 #include "common/common_types.h"
+#include "common/heap_tracker.h"
 #include "common/logging/log.h"
 #include "common/page_table.h"
 #include "common/scope_exit.h"
@@ -45,11 +46,25 @@ struct Memory::Impl {
 
     void SetCurrentPageTable(Kernel::KProcess& process) {
         current_page_table = &process.GetPageTable().GetImpl();
-        current_page_table->fastmem_arena = system.DeviceMemory().buffer.VirtualBasePointer();
+
+        if (std::addressof(process) == system.ApplicationProcess() &&
+            Settings::IsFastmemEnabled()) {
+            current_page_table->fastmem_arena = system.DeviceMemory().buffer.VirtualBasePointer();
+        } else {
+            current_page_table->fastmem_arena = nullptr;
+        }
+
+#ifdef __linux__
+        heap_tracker.emplace(system.DeviceMemory().buffer);
+        buffer = std::addressof(*heap_tracker);
+#else
+        buffer = std::addressof(system.DeviceMemory().buffer);
+#endif
     }
 
     void MapMemoryRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size,
-                         Common::PhysicalAddress target, Common::MemoryPermission perms) {
+                         Common::PhysicalAddress target, Common::MemoryPermission perms,
+                         bool separate_heap) {
         ASSERT_MSG((size & YUZU_PAGEMASK) == 0, "non-page aligned size: {:016X}", size);
         ASSERT_MSG((base & YUZU_PAGEMASK) == 0, "non-page aligned base: {:016X}", GetInteger(base));
         ASSERT_MSG(target >= DramMemoryMap::Base, "Out of bounds target: {:016X}",
@@ -57,20 +72,21 @@ struct Memory::Impl {
         MapPages(page_table, base / YUZU_PAGESIZE, size / YUZU_PAGESIZE, target,
                  Common::PageType::Memory);
 
-        if (Settings::IsFastmemEnabled()) {
-            system.DeviceMemory().buffer.Map(GetInteger(base),
-                                             GetInteger(target) - DramMemoryMap::Base, size, perms);
+        if (current_page_table->fastmem_arena) {
+            buffer->Map(GetInteger(base), GetInteger(target) - DramMemoryMap::Base, size, perms,
+                        separate_heap);
         }
     }
 
-    void UnmapRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size) {
+    void UnmapRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size,
+                     bool separate_heap) {
         ASSERT_MSG((size & YUZU_PAGEMASK) == 0, "non-page aligned size: {:016X}", size);
         ASSERT_MSG((base & YUZU_PAGEMASK) == 0, "non-page aligned base: {:016X}", GetInteger(base));
         MapPages(page_table, base / YUZU_PAGESIZE, size / YUZU_PAGESIZE, 0,
                  Common::PageType::Unmapped);
 
-        if (Settings::IsFastmemEnabled()) {
-            system.DeviceMemory().buffer.Unmap(GetInteger(base), size);
+        if (current_page_table->fastmem_arena) {
+            buffer->Unmap(GetInteger(base), size, separate_heap);
         }
     }
 
@@ -79,17 +95,7 @@ struct Memory::Impl {
         ASSERT_MSG((size & YUZU_PAGEMASK) == 0, "non-page aligned size: {:016X}", size);
         ASSERT_MSG((vaddr & YUZU_PAGEMASK) == 0, "non-page aligned base: {:016X}", vaddr);
 
-        if (!Settings::IsFastmemEnabled()) {
-            return;
-        }
-
-        const bool is_r = True(perms & Common::MemoryPermission::Read);
-        const bool is_w = True(perms & Common::MemoryPermission::Write);
-        const bool is_x =
-            True(perms & Common::MemoryPermission::Execute) && Settings::IsNceEnabled();
-
-        if (!current_page_table) {
-            system.DeviceMemory().buffer.Protect(vaddr, size, is_r, is_w, is_x);
+        if (!current_page_table->fastmem_arena) {
             return;
         }
 
@@ -101,8 +107,7 @@ struct Memory::Impl {
             switch (page_type) {
             case Common::PageType::RasterizerCachedMemory:
                 if (protect_bytes > 0) {
-                    system.DeviceMemory().buffer.Protect(protect_begin, protect_bytes, is_r, is_w,
-                                                         is_x);
+                    buffer->Protect(protect_begin, protect_bytes, perms);
                     protect_bytes = 0;
                 }
                 break;
@@ -115,7 +120,7 @@ struct Memory::Impl {
         }
 
         if (protect_bytes > 0) {
-            system.DeviceMemory().buffer.Protect(protect_begin, protect_bytes, is_r, is_w, is_x);
+            buffer->Protect(protect_begin, protect_bytes, perms);
         }
     }
 
@@ -239,7 +244,7 @@ struct Memory::Impl {
 
     bool WalkBlock(const Common::ProcessAddress addr, const std::size_t size, auto on_unmapped,
                    auto on_memory, auto on_rasterizer, auto increment) {
-        const auto& page_table = system.ApplicationProcess()->GetPageTable().GetImpl();
+        const auto& page_table = *current_page_table;
         std::size_t remaining_size = size;
         std::size_t page_index = addr >> YUZU_PAGEBITS;
         std::size_t page_offset = addr & YUZU_PAGEMASK;
@@ -484,8 +489,10 @@ struct Memory::Impl {
             return;
         }
 
-        if (Settings::IsFastmemEnabled()) {
-            system.DeviceMemory().buffer.Protect(vaddr, size, !debug, !debug);
+        if (current_page_table->fastmem_arena) {
+            const auto perm{debug ? Common::MemoryPermission{}
+                                  : Common::MemoryPermission::ReadWrite};
+            buffer->Protect(vaddr, size, perm);
         }
 
         // Iterate over a contiguous CPU address space, marking/unmarking the region.
@@ -541,10 +548,15 @@ struct Memory::Impl {
             return;
         }
 
-        if (Settings::IsFastmemEnabled()) {
-            const bool is_read_enable =
-                !Settings::values.use_reactive_flushing.GetValue() || !cached;
-            system.DeviceMemory().buffer.Protect(vaddr, size, is_read_enable, !cached);
+        if (current_page_table->fastmem_arena) {
+            Common::MemoryPermission perm{};
+            if (!Settings::values.use_reactive_flushing.GetValue() || !cached) {
+                perm |= Common::MemoryPermission::Read;
+            }
+            if (!cached) {
+                perm |= Common::MemoryPermission::Write;
+            }
+            buffer->Protect(vaddr, size, perm);
         }
 
         // Iterate over a contiguous CPU address space, which corresponds to the specified GPU
@@ -855,6 +867,13 @@ struct Memory::Impl {
     std::array<GPUDirtyState, Core::Hardware::NUM_CPU_CORES> rasterizer_write_areas{};
     std::span<Core::GPUDirtyMemoryManager> gpu_dirty_managers;
     std::mutex sys_core_guard;
+
+    std::optional<Common::HeapTracker> heap_tracker;
+#ifdef __linux__
+    Common::HeapTracker* buffer{};
+#else
+    Common::HostMemory* buffer{};
+#endif
 };
 
 Memory::Memory(Core::System& system_) : system{system_} {
@@ -872,12 +891,14 @@ void Memory::SetCurrentPageTable(Kernel::KProcess& process) {
 }
 
 void Memory::MapMemoryRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size,
-                             Common::PhysicalAddress target, Common::MemoryPermission perms) {
-    impl->MapMemoryRegion(page_table, base, size, target, perms);
+                             Common::PhysicalAddress target, Common::MemoryPermission perms,
+                             bool separate_heap) {
+    impl->MapMemoryRegion(page_table, base, size, target, perms, separate_heap);
 }
 
-void Memory::UnmapRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size) {
-    impl->UnmapRegion(page_table, base, size);
+void Memory::UnmapRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size,
+                         bool separate_heap) {
+    impl->UnmapRegion(page_table, base, size, separate_heap);
 }
 
 void Memory::ProtectRegion(Common::PageTable& page_table, Common::ProcessAddress vaddr, u64 size,
@@ -886,8 +907,7 @@ void Memory::ProtectRegion(Common::PageTable& page_table, Common::ProcessAddress
 }
 
 bool Memory::IsValidVirtualAddress(const Common::ProcessAddress vaddr) const {
-    const Kernel::KProcess& process = *system.ApplicationProcess();
-    const auto& page_table = process.GetPageTable().GetImpl();
+    const auto& page_table = *impl->current_page_table;
     const size_t page = vaddr >> YUZU_PAGEBITS;
     if (page >= page_table.pointers.size()) {
         return false;
@@ -1048,7 +1068,9 @@ void Memory::FlushRegion(Common::ProcessAddress dest_addr, size_t size) {
 }
 
 bool Memory::InvalidateNCE(Common::ProcessAddress vaddr, size_t size) {
-    bool mapped = true;
+    [[maybe_unused]] bool mapped = true;
+    [[maybe_unused]] bool rasterizer = false;
+
     u8* const ptr = impl->GetPointerImpl(
         GetInteger(vaddr),
         [&] {
@@ -1056,8 +1078,26 @@ bool Memory::InvalidateNCE(Common::ProcessAddress vaddr, size_t size) {
                       GetInteger(vaddr));
             mapped = false;
         },
-        [&] { impl->system.GPU().InvalidateRegion(GetInteger(vaddr), size); });
+        [&] {
+            impl->system.GPU().InvalidateRegion(GetInteger(vaddr), size);
+            rasterizer = true;
+        });
+
+#ifdef __linux__
+    if (!rasterizer && mapped) {
+        impl->buffer->DeferredMapSeparateHeap(GetInteger(vaddr));
+    }
+#endif
+
     return mapped && ptr != nullptr;
 }
 
+bool Memory::InvalidateSeparateHeap(void* fault_address) {
+#ifdef __linux__
+    return impl->buffer->DeferredMapSeparateHeap(static_cast<u8*>(fault_address));
+#else
+    return false;
+#endif
+}
+
 } // namespace Core::Memory

src/core/memory.h

@@ -86,7 +86,8 @@ public:
      * @param perms      The permissions to map the memory with.
      */
     void MapMemoryRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size,
-                         Common::PhysicalAddress target, Common::MemoryPermission perms);
+                         Common::PhysicalAddress target, Common::MemoryPermission perms,
+                         bool separate_heap);
 
     /**
      * Unmaps a region of the emulated process address space.
@@ -95,7 +96,8 @@ public:
      * @param base       The address to begin unmapping at.
      * @param size       The amount of bytes to unmap.
      */
-    void UnmapRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size);
+    void UnmapRegion(Common::PageTable& page_table, Common::ProcessAddress base, u64 size,
+                     bool separate_heap);
 
     /**
      * Protects a region of the emulated process address space with the new permissions.
@@ -486,6 +488,7 @@ public:
     void SetGPUDirtyManagers(std::span<Core::GPUDirtyMemoryManager> managers);
     void InvalidateRegion(Common::ProcessAddress dest_addr, size_t size);
     bool InvalidateNCE(Common::ProcessAddress vaddr, size_t size);
+    bool InvalidateSeparateHeap(void* fault_address);
     void FlushRegion(Common::ProcessAddress dest_addr, size_t size);
 
 private:

src/shader_recompiler/backend/spirv/emit_spirv_image.cpp

@@ -214,16 +214,16 @@ Id TextureImage(EmitContext& ctx, IR::TextureInstInfo info, const IR::Value& ind
     }
 }
 
-Id Image(EmitContext& ctx, const IR::Value& index, IR::TextureInstInfo info) {
+std::pair<Id, bool> Image(EmitContext& ctx, const IR::Value& index, IR::TextureInstInfo info) {
     if (!index.IsImmediate() || index.U32() != 0) {
         throw NotImplementedException("Indirect image indexing");
     }
     if (info.type == TextureType::Buffer) {
         const ImageBufferDefinition def{ctx.image_buffers.at(info.descriptor_index)};
-        return ctx.OpLoad(def.image_type, def.id);
+        return {ctx.OpLoad(def.image_type, def.id), def.is_integer};
     } else {
         const ImageDefinition def{ctx.images.at(info.descriptor_index)};
-        return ctx.OpLoad(def.image_type, def.id);
+        return {ctx.OpLoad(def.image_type, def.id), def.is_integer};
     }
 }
 
@@ -566,13 +566,23 @@ Id EmitImageRead(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id co
         LOG_WARNING(Shader_SPIRV, "Typeless image read not supported by host");
         return ctx.ConstantNull(ctx.U32[4]);
     }
-    return Emit(&EmitContext::OpImageSparseRead, &EmitContext::OpImageRead, ctx, inst, ctx.U32[4],
-                Image(ctx, index, info), coords, std::nullopt, std::span<const Id>{});
+    const auto [image, is_integer] = Image(ctx, index, info);
+    const Id result_type{is_integer ? ctx.U32[4] : ctx.F32[4]};
+    Id color{Emit(&EmitContext::OpImageSparseRead, &EmitContext::OpImageRead, ctx, inst,
+                  result_type, image, coords, std::nullopt, std::span<const Id>{})};
+    if (!is_integer) {
+        color = ctx.OpBitcast(ctx.U32[4], color);
+    }
+    return color;
 }
 
 void EmitImageWrite(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id coords, Id color) {
     const auto info{inst->Flags<IR::TextureInstInfo>()};
-    ctx.OpImageWrite(Image(ctx, index, info), coords, color);
+    const auto [image, is_integer] = Image(ctx, index, info);
+    if (!is_integer) {
+        color = ctx.OpBitcast(ctx.F32[4], color);
+    }
+    ctx.OpImageWrite(image, coords, color);
 }
 
 Id EmitIsTextureScaled(EmitContext& ctx, const IR::Value& index) {

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -74,20 +74,19 @@ spv::ImageFormat GetImageFormat(ImageFormat format) {
     throw InvalidArgument("Invalid image format {}", format);
 }
 
-Id ImageType(EmitContext& ctx, const ImageDescriptor& desc) {
+Id ImageType(EmitContext& ctx, const ImageDescriptor& desc, Id sampled_type) {
     const spv::ImageFormat format{GetImageFormat(desc.format)};
-    const Id type{ctx.U32[1]};
     switch (desc.type) {
     case TextureType::Color1D:
-        return ctx.TypeImage(type, spv::Dim::Dim1D, false, false, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim1D, false, false, false, 2, format);
     case TextureType::ColorArray1D:
-        return ctx.TypeImage(type, spv::Dim::Dim1D, false, true, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim1D, false, true, false, 2, format);
     case TextureType::Color2D:
-        return ctx.TypeImage(type, spv::Dim::Dim2D, false, false, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim2D, false, false, false, 2, format);
     case TextureType::ColorArray2D:
-        return ctx.TypeImage(type, spv::Dim::Dim2D, false, true, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim2D, false, true, false, 2, format);
     case TextureType::Color3D:
-        return ctx.TypeImage(type, spv::Dim::Dim3D, false, false, false, 2, format);
+        return ctx.TypeImage(sampled_type, spv::Dim::Dim3D, false, false, false, 2, format);
     case TextureType::Buffer:
         throw NotImplementedException("Image buffer");
     default:
@@ -1273,7 +1272,9 @@ void EmitContext::DefineImageBuffers(const Info& info, u32& binding) {
             throw NotImplementedException("Array of image buffers");
         }
         const spv::ImageFormat format{GetImageFormat(desc.format)};
-        const Id image_type{TypeImage(U32[1], spv::Dim::Buffer, false, false, false, 2, format)};
+        const Id sampled_type{desc.is_integer ? U32[1] : F32[1]};
+        const Id image_type{
+            TypeImage(sampled_type, spv::Dim::Buffer, false, false, false, 2, format)};
         const Id pointer_type{TypePointer(spv::StorageClass::UniformConstant, image_type)};
         const Id id{AddGlobalVariable(pointer_type, spv::StorageClass::UniformConstant)};
         Decorate(id, spv::Decoration::Binding, binding);
@@ -1283,6 +1284,7 @@ void EmitContext::DefineImageBuffers(const Info& info, u32& binding) {
             .id = id,
             .image_type = image_type,
             .count = desc.count,
+            .is_integer = desc.is_integer,
         });
         if (profile.supported_spirv >= 0x00010400) {
             interfaces.push_back(id);
@@ -1327,7 +1329,8 @@ void EmitContext::DefineImages(const Info& info, u32& binding, u32& scaling_inde
         if (desc.count != 1) {
             throw NotImplementedException("Array of images");
         }
-        const Id image_type{ImageType(*this, desc)};
+        const Id sampled_type{desc.is_integer ? U32[1] : F32[1]};
+        const Id image_type{ImageType(*this, desc, sampled_type)};
         const Id pointer_type{TypePointer(spv::StorageClass::UniformConstant, image_type)};
         const Id id{AddGlobalVariable(pointer_type, spv::StorageClass::UniformConstant)};
         Decorate(id, spv::Decoration::Binding, binding);
@@ -1337,6 +1340,7 @@ void EmitContext::DefineImages(const Info& info, u32& binding, u32& scaling_inde
             .id = id,
             .image_type = image_type,
             .count = desc.count,
+            .is_integer = desc.is_integer,
         });
         if (profile.supported_spirv >= 0x00010400) {
             interfaces.push_back(id);

src/shader_recompiler/backend/spirv/spirv_emit_context.h

@@ -47,12 +47,14 @@ struct ImageBufferDefinition {
     Id id;
     Id image_type;
     u32 count;
+    bool is_integer;
 };
 
 struct ImageDefinition {
     Id id;
     Id image_type;
     u32 count;
+    bool is_integer;
 };
 
 struct UniformDefinitions {

src/shader_recompiler/environment.h

@@ -24,6 +24,8 @@ public:
 
     [[nodiscard]] virtual TexturePixelFormat ReadTexturePixelFormat(u32 raw_handle) = 0;
 
+    [[nodiscard]] virtual bool IsTexturePixelFormatInteger(u32 raw_handle) = 0;
+
     [[nodiscard]] virtual u32 ReadViewportTransformState() = 0;
 
     [[nodiscard]] virtual u32 TextureBoundBuffer() const = 0;

src/shader_recompiler/ir_opt/constant_propagation_pass.cpp

@@ -815,6 +815,15 @@ bool FindGradient3DDerivatives(std::array<IR::Value, 3>& results, IR::Value coor
     return true;
 }
 
+void ConvertDerivatives(std::array<IR::Value, 3>& results, IR::IREmitter& ir) {
+    for (size_t i = 0; i < 3; i++) {
+        if (results[i].Type() == IR::Type::U32) {
+            results[i] = results[i].IsImmediate() ? ir.Imm32(Common::BitCast<f32>(results[i].U32()))
+                                                  : ir.BitCast<IR::F32>(IR::U32(results[i]));
+        }
+    }
+}
+
 void FoldImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
     IR::TextureInstInfo info = inst.Flags<IR::TextureInstInfo>();
     auto orig_opcode = inst.GetOpcode();
@@ -831,12 +840,14 @@ void FoldImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
     if (!offset.IsImmediate()) {
         return;
     }
+    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::Inst* const inst2 = coords.InstRecursive();
     std::array<std::array<IR::Value, 3>, 3> results_matrix;
     for (size_t i = 0; i < 3; i++) {
         if (!FindGradient3DDerivatives(results_matrix[i], inst2->Arg(i).Resolve())) {
             return;
         }
+        ConvertDerivatives(results_matrix[i], ir);
     }
     IR::F32 lod_clamp{};
     if (info.has_lod_clamp != 0) {
@@ -846,7 +857,6 @@ void FoldImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
             lod_clamp = IR::F32{bias_lc};
         }
     }
-    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::Value new_coords =
         ir.CompositeConstruct(results_matrix[0][0], results_matrix[1][0], results_matrix[2][0]);
     IR::Value derivatives_1 = ir.CompositeConstruct(results_matrix[0][1], results_matrix[0][2],

src/shader_recompiler/ir_opt/texture_pass.cpp

@@ -372,6 +372,10 @@ TexturePixelFormat ReadTexturePixelFormat(Environment& env, const ConstBufferAdd
     return env.ReadTexturePixelFormat(GetTextureHandle(env, cbuf));
 }
 
+bool IsTexturePixelFormatInteger(Environment& env, const ConstBufferAddr& cbuf) {
+    return env.IsTexturePixelFormatInteger(GetTextureHandle(env, cbuf));
+}
+
 class Descriptors {
 public:
     explicit Descriptors(TextureBufferDescriptors& texture_buffer_descriptors_,
@@ -403,6 +407,7 @@ public:
         })};
         image_buffer_descriptors[index].is_written |= desc.is_written;
         image_buffer_descriptors[index].is_read |= desc.is_read;
+        image_buffer_descriptors[index].is_integer |= desc.is_integer;
         return index;
     }
 
@@ -432,6 +437,7 @@ public:
         })};
         image_descriptors[index].is_written |= desc.is_written;
         image_descriptors[index].is_read |= desc.is_read;
+        image_descriptors[index].is_integer |= desc.is_integer;
         return index;
     }
 
@@ -469,6 +475,20 @@ void PatchImageSampleImplicitLod(IR::Block& block, IR::Inst& inst) {
                         ir.FPRecip(ir.ConvertUToF(32, 32, ir.CompositeExtract(texture_size, 1))))));
 }
 
+bool IsPixelFormatSNorm(TexturePixelFormat pixel_format) {
+    switch (pixel_format) {
+    case TexturePixelFormat::A8B8G8R8_SNORM:
+    case TexturePixelFormat::R8G8_SNORM:
+    case TexturePixelFormat::R8_SNORM:
+    case TexturePixelFormat::R16G16B16A16_SNORM:
+    case TexturePixelFormat::R16G16_SNORM:
+    case TexturePixelFormat::R16_SNORM:
+        return true;
+    default:
+        return false;
+    }
+}
+
 void PatchTexelFetch(IR::Block& block, IR::Inst& inst, TexturePixelFormat pixel_format) {
     const auto it{IR::Block::InstructionList::s_iterator_to(inst)};
     IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
@@ -587,11 +607,13 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
             }
             const bool is_written{inst->GetOpcode() != IR::Opcode::ImageRead};
             const bool is_read{inst->GetOpcode() != IR::Opcode::ImageWrite};
+            const bool is_integer{IsTexturePixelFormatInteger(env, cbuf)};
             if (flags.type == TextureType::Buffer) {
                 index = descriptors.Add(ImageBufferDescriptor{
                     .format = flags.image_format,
                     .is_written = is_written,
                     .is_read = is_read,
+                    .is_integer = is_integer,
                     .cbuf_index = cbuf.index,
                     .cbuf_offset = cbuf.offset,
                     .count = cbuf.count,
@@ -603,6 +625,7 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
                     .format = flags.image_format,
                     .is_written = is_written,
                     .is_read = is_read,
+                    .is_integer = is_integer,
                     .cbuf_index = cbuf.index,
                     .cbuf_offset = cbuf.offset,
                     .count = cbuf.count,
@@ -658,7 +681,7 @@ void TexturePass(Environment& env, IR::Program& program, const HostTranslateInfo
         if (!host_info.support_snorm_render_buffer && inst->GetOpcode() == IR::Opcode::ImageFetch &&
             flags.type == TextureType::Buffer) {
             const auto pixel_format = ReadTexturePixelFormat(env, cbuf);
-            if (pixel_format != TexturePixelFormat::OTHER) {
+            if (IsPixelFormatSNorm(pixel_format)) {
                 PatchTexelFetch(*texture_inst.block, *texture_inst.inst, pixel_format);
             }
         }

src/shader_recompiler/shader_info.h

@@ -35,14 +35,109 @@ enum class TextureType : u32 {
 };
 constexpr u32 NUM_TEXTURE_TYPES = 9;
 
-enum class TexturePixelFormat : u32 {
+enum class TexturePixelFormat {
+    A8B8G8R8_UNORM,
     A8B8G8R8_SNORM,
+    A8B8G8R8_SINT,
+    A8B8G8R8_UINT,
+    R5G6B5_UNORM,
+    B5G6R5_UNORM,
+    A1R5G5B5_UNORM,
+    A2B10G10R10_UNORM,
+    A2B10G10R10_UINT,
+    A2R10G10B10_UNORM,
+    A1B5G5R5_UNORM,
+    A5B5G5R1_UNORM,
+    R8_UNORM,
     R8_SNORM,
-    R8G8_SNORM,
+    R8_SINT,
+    R8_UINT,
+    R16G16B16A16_FLOAT,
+    R16G16B16A16_UNORM,
     R16G16B16A16_SNORM,
-    R16G16_SNORM,
+    R16G16B16A16_SINT,
+    R16G16B16A16_UINT,
+    B10G11R11_FLOAT,
+    R32G32B32A32_UINT,
+    BC1_RGBA_UNORM,
+    BC2_UNORM,
+    BC3_UNORM,
+    BC4_UNORM,
+    BC4_SNORM,
+    BC5_UNORM,
+    BC5_SNORM,
+    BC7_UNORM,
+    BC6H_UFLOAT,
+    BC6H_SFLOAT,
+    ASTC_2D_4X4_UNORM,
+    B8G8R8A8_UNORM,
+    R32G32B32A32_FLOAT,
+    R32G32B32A32_SINT,
+    R32G32_FLOAT,
+    R32G32_SINT,
+    R32_FLOAT,
+    R16_FLOAT,
+    R16_UNORM,
     R16_SNORM,
-    OTHER
+    R16_UINT,
+    R16_SINT,
+    R16G16_UNORM,
+    R16G16_FLOAT,
+    R16G16_UINT,
+    R16G16_SINT,
+    R16G16_SNORM,
+    R32G32B32_FLOAT,
+    A8B8G8R8_SRGB,
+    R8G8_UNORM,
+    R8G8_SNORM,
+    R8G8_SINT,
+    R8G8_UINT,
+    R32G32_UINT,
+    R16G16B16X16_FLOAT,
+    R32_UINT,
+    R32_SINT,
+    ASTC_2D_8X8_UNORM,
+    ASTC_2D_8X5_UNORM,
+    ASTC_2D_5X4_UNORM,
+    B8G8R8A8_SRGB,
+    BC1_RGBA_SRGB,
+    BC2_SRGB,
+    BC3_SRGB,
+    BC7_SRGB,
+    A4B4G4R4_UNORM,
+    G4R4_UNORM,
+    ASTC_2D_4X4_SRGB,
+    ASTC_2D_8X8_SRGB,
+    ASTC_2D_8X5_SRGB,
+    ASTC_2D_5X4_SRGB,
+    ASTC_2D_5X5_UNORM,
+    ASTC_2D_5X5_SRGB,
+    ASTC_2D_10X8_UNORM,
+    ASTC_2D_10X8_SRGB,
+    ASTC_2D_6X6_UNORM,
+    ASTC_2D_6X6_SRGB,
+    ASTC_2D_10X6_UNORM,
+    ASTC_2D_10X6_SRGB,
+    ASTC_2D_10X5_UNORM,
+    ASTC_2D_10X5_SRGB,
+    ASTC_2D_10X10_UNORM,
+    ASTC_2D_10X10_SRGB,
+    ASTC_2D_12X10_UNORM,
+    ASTC_2D_12X10_SRGB,
+    ASTC_2D_12X12_UNORM,
+    ASTC_2D_12X12_SRGB,
+    ASTC_2D_8X6_UNORM,
+    ASTC_2D_8X6_SRGB,
+    ASTC_2D_6X5_UNORM,
+    ASTC_2D_6X5_SRGB,
+    E5B9G9R9_FLOAT,
+    D32_FLOAT,
+    D16_UNORM,
+    X8_D24_UNORM,
+    S8_UINT,
+    D24_UNORM_S8_UINT,
+    S8_UINT_D24_UNORM,
+    D32_FLOAT_S8_UINT,
 };
 
 enum class ImageFormat : u32 {
@@ -97,6 +192,7 @@ struct ImageBufferDescriptor {
     ImageFormat format;
     bool is_written;
     bool is_read;
+    bool is_integer;
     u32 cbuf_index;
     u32 cbuf_offset;
     u32 count;
@@ -129,6 +225,7 @@ struct ImageDescriptor {
     ImageFormat format;
     bool is_written;
     bool is_read;
+    bool is_integer;
     u32 cbuf_index;
     u32 cbuf_offset;
     u32 count;

src/tests/common/host_memory.cpp

@@ -12,6 +12,7 @@ using namespace Common::Literals;
 static constexpr size_t VIRTUAL_SIZE = 1ULL << 39;
 static constexpr size_t BACKING_SIZE = 4_GiB;
 static constexpr auto PERMS = Common::MemoryPermission::ReadWrite;
+static constexpr auto HEAP = false;
 
 TEST_CASE("HostMemory: Initialize and deinitialize", "[common]") {
     { HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE); }
@@ -20,7 +21,7 @@ TEST_CASE("HostMemory: Initialize and deinitialize", "[common]") {
 
 TEST_CASE("HostMemory: Simple map", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x5000, 0x8000, 0x1000, PERMS);
+    mem.Map(0x5000, 0x8000, 0x1000, PERMS, HEAP);
 
     volatile u8* const data = mem.VirtualBasePointer() + 0x5000;
     data[0] = 50;
@@ -29,8 +30,8 @@ TEST_CASE("HostMemory: Simple map", "[common]") {
 
 TEST_CASE("HostMemory: Simple mirror map", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x5000, 0x3000, 0x2000, PERMS);
-    mem.Map(0x8000, 0x4000, 0x1000, PERMS);
+    mem.Map(0x5000, 0x3000, 0x2000, PERMS, HEAP);
+    mem.Map(0x8000, 0x4000, 0x1000, PERMS, HEAP);
 
     volatile u8* const mirror_a = mem.VirtualBasePointer() + 0x5000;
     volatile u8* const mirror_b = mem.VirtualBasePointer() + 0x8000;
@@ -40,116 +41,116 @@ TEST_CASE("HostMemory: Simple mirror map", "[common]") {
 
 TEST_CASE("HostMemory: Simple unmap", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x5000, 0x3000, 0x2000, PERMS);
+    mem.Map(0x5000, 0x3000, 0x2000, PERMS, HEAP);
 
     volatile u8* const data = mem.VirtualBasePointer() + 0x5000;
     data[75] = 50;
     REQUIRE(data[75] == 50);
 
-    mem.Unmap(0x5000, 0x2000);
+    mem.Unmap(0x5000, 0x2000, HEAP);
 }
 
 TEST_CASE("HostMemory: Simple unmap and remap", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x5000, 0x3000, 0x2000, PERMS);
+    mem.Map(0x5000, 0x3000, 0x2000, PERMS, HEAP);
 
     volatile u8* const data = mem.VirtualBasePointer() + 0x5000;
     data[0] = 50;
     REQUIRE(data[0] == 50);
 
-    mem.Unmap(0x5000, 0x2000);
+    mem.Unmap(0x5000, 0x2000, HEAP);
 
-    mem.Map(0x5000, 0x3000, 0x2000, PERMS);
+    mem.Map(0x5000, 0x3000, 0x2000, PERMS, HEAP);
     REQUIRE(data[0] == 50);
 
-    mem.Map(0x7000, 0x2000, 0x5000, PERMS);
+    mem.Map(0x7000, 0x2000, 0x5000, PERMS, HEAP);
     REQUIRE(data[0x3000] == 50);
 }
 
 TEST_CASE("HostMemory: Nieche allocation", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x0000, 0, 0x20000, PERMS);
-    mem.Unmap(0x0000, 0x4000);
-    mem.Map(0x1000, 0, 0x2000, PERMS);
-    mem.Map(0x3000, 0, 0x1000, PERMS);
-    mem.Map(0, 0, 0x1000, PERMS);
+    mem.Map(0x0000, 0, 0x20000, PERMS, HEAP);
+    mem.Unmap(0x0000, 0x4000, HEAP);
+    mem.Map(0x1000, 0, 0x2000, PERMS, HEAP);
+    mem.Map(0x3000, 0, 0x1000, PERMS, HEAP);
+    mem.Map(0, 0, 0x1000, PERMS, HEAP);
 }
 
 TEST_CASE("HostMemory: Full unmap", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x8000, 0, 0x4000, PERMS);
-    mem.Unmap(0x8000, 0x4000);
-    mem.Map(0x6000, 0, 0x16000, PERMS);
+    mem.Map(0x8000, 0, 0x4000, PERMS, HEAP);
+    mem.Unmap(0x8000, 0x4000, HEAP);
+    mem.Map(0x6000, 0, 0x16000, PERMS, HEAP);
 }
 
 TEST_CASE("HostMemory: Right out of bounds unmap", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x0000, 0, 0x4000, PERMS);
-    mem.Unmap(0x2000, 0x4000);
-    mem.Map(0x2000, 0x80000, 0x4000, PERMS);
+    mem.Map(0x0000, 0, 0x4000, PERMS, HEAP);
+    mem.Unmap(0x2000, 0x4000, HEAP);
+    mem.Map(0x2000, 0x80000, 0x4000, PERMS, HEAP);
 }
 
 TEST_CASE("HostMemory: Left out of bounds unmap", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x8000, 0, 0x4000, PERMS);
-    mem.Unmap(0x6000, 0x4000);
-    mem.Map(0x8000, 0, 0x2000, PERMS);
+    mem.Map(0x8000, 0, 0x4000, PERMS, HEAP);
+    mem.Unmap(0x6000, 0x4000, HEAP);
+    mem.Map(0x8000, 0, 0x2000, PERMS, HEAP);
 }
 
 TEST_CASE("HostMemory: Multiple placeholder unmap", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x0000, 0, 0x4000, PERMS);
-    mem.Map(0x4000, 0, 0x1b000, PERMS);
-    mem.Unmap(0x3000, 0x1c000);
-    mem.Map(0x3000, 0, 0x20000, PERMS);
+    mem.Map(0x0000, 0, 0x4000, PERMS, HEAP);
+    mem.Map(0x4000, 0, 0x1b000, PERMS, HEAP);
+    mem.Unmap(0x3000, 0x1c000, HEAP);
+    mem.Map(0x3000, 0, 0x20000, PERMS, HEAP);
 }
 
 TEST_CASE("HostMemory: Unmap between placeholders", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x0000, 0, 0x4000, PERMS);
-    mem.Map(0x4000, 0, 0x4000, PERMS);
-    mem.Unmap(0x2000, 0x4000);
-    mem.Map(0x2000, 0, 0x4000, PERMS);
+    mem.Map(0x0000, 0, 0x4000, PERMS, HEAP);
+    mem.Map(0x4000, 0, 0x4000, PERMS, HEAP);
+    mem.Unmap(0x2000, 0x4000, HEAP);
+    mem.Map(0x2000, 0, 0x4000, PERMS, HEAP);
 }
 
 TEST_CASE("HostMemory: Unmap to origin", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x4000, 0, 0x4000, PERMS);
-    mem.Map(0x8000, 0, 0x4000, PERMS);
-    mem.Unmap(0x4000, 0x4000);
-    mem.Map(0, 0, 0x4000, PERMS);
-    mem.Map(0x4000, 0, 0x4000, PERMS);
+    mem.Map(0x4000, 0, 0x4000, PERMS, HEAP);
+    mem.Map(0x8000, 0, 0x4000, PERMS, HEAP);
+    mem.Unmap(0x4000, 0x4000, HEAP);
+    mem.Map(0, 0, 0x4000, PERMS, HEAP);
+    mem.Map(0x4000, 0, 0x4000, PERMS, HEAP);
 }
 
 TEST_CASE("HostMemory: Unmap to right", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x4000, 0, 0x4000, PERMS);
-    mem.Map(0x8000, 0, 0x4000, PERMS);
-    mem.Unmap(0x8000, 0x4000);
-    mem.Map(0x8000, 0, 0x4000, PERMS);
+    mem.Map(0x4000, 0, 0x4000, PERMS, HEAP);
+    mem.Map(0x8000, 0, 0x4000, PERMS, HEAP);
+    mem.Unmap(0x8000, 0x4000, HEAP);
+    mem.Map(0x8000, 0, 0x4000, PERMS, HEAP);
 }
 
 TEST_CASE("HostMemory: Partial right unmap check bindings", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x4000, 0x10000, 0x4000, PERMS);
+    mem.Map(0x4000, 0x10000, 0x4000, PERMS, HEAP);
 
     volatile u8* const ptr = mem.VirtualBasePointer() + 0x4000;
     ptr[0x1000] = 17;
 
-    mem.Unmap(0x6000, 0x2000);
+    mem.Unmap(0x6000, 0x2000, HEAP);
 
     REQUIRE(ptr[0x1000] == 17);
 }
 
 TEST_CASE("HostMemory: Partial left unmap check bindings", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x4000, 0x10000, 0x4000, PERMS);
+    mem.Map(0x4000, 0x10000, 0x4000, PERMS, HEAP);
 
     volatile u8* const ptr = mem.VirtualBasePointer() + 0x4000;
     ptr[0x3000] = 19;
     ptr[0x3fff] = 12;
 
-    mem.Unmap(0x4000, 0x2000);
+    mem.Unmap(0x4000, 0x2000, HEAP);
 
     REQUIRE(ptr[0x3000] == 19);
     REQUIRE(ptr[0x3fff] == 12);
@@ -157,13 +158,13 @@ TEST_CASE("HostMemory: Partial left unmap check bindings", "[common]") {
 
 TEST_CASE("HostMemory: Partial middle unmap check bindings", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x4000, 0x10000, 0x4000, PERMS);
+    mem.Map(0x4000, 0x10000, 0x4000, PERMS, HEAP);
 
     volatile u8* const ptr = mem.VirtualBasePointer() + 0x4000;
     ptr[0x0000] = 19;
     ptr[0x3fff] = 12;
 
-    mem.Unmap(0x1000, 0x2000);
+    mem.Unmap(0x1000, 0x2000, HEAP);
 
     REQUIRE(ptr[0x0000] == 19);
     REQUIRE(ptr[0x3fff] == 12);
@@ -171,14 +172,14 @@ TEST_CASE("HostMemory: Partial middle unmap check bindings", "[common]") {
 
 TEST_CASE("HostMemory: Partial sparse middle unmap and check bindings", "[common]") {
     HostMemory mem(BACKING_SIZE, VIRTUAL_SIZE);
-    mem.Map(0x4000, 0x10000, 0x2000, PERMS);
-    mem.Map(0x6000, 0x20000, 0x2000, PERMS);
+    mem.Map(0x4000, 0x10000, 0x2000, PERMS, HEAP);
+    mem.Map(0x6000, 0x20000, 0x2000, PERMS, HEAP);
 
     volatile u8* const ptr = mem.VirtualBasePointer() + 0x4000;
     ptr[0x0000] = 19;
     ptr[0x3fff] = 12;
 
-    mem.Unmap(0x5000, 0x2000);
+    mem.Unmap(0x5000, 0x2000, HEAP);
 
     REQUIRE(ptr[0x0000] == 19);
     REQUIRE(ptr[0x3fff] == 12);

src/video_core/engines/maxwell_3d.cpp

@@ -586,14 +586,22 @@ void Maxwell3D::ProcessQueryCondition() {
 }
 
 void Maxwell3D::ProcessCounterReset() {
-    switch (regs.clear_report_value) {
-    case Regs::ClearReport::ZPassPixelCount:
-        rasterizer->ResetCounter(VideoCommon::QueryType::ZPassPixelCount64);
-        break;
-    default:
-        LOG_DEBUG(Render_OpenGL, "Unimplemented counter reset={}", regs.clear_report_value);
-        break;
-    }
+    const auto query_type = [clear_report = regs.clear_report_value]() {
+        switch (clear_report) {
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::ZPassPixelCount:
+            return VideoCommon::QueryType::ZPassPixelCount64;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::StreamingPrimitivesSucceeded:
+            return VideoCommon::QueryType::StreamingPrimitivesSucceeded;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::PrimitivesGenerated:
+            return VideoCommon::QueryType::PrimitivesGenerated;
+        case Tegra::Engines::Maxwell3D::Regs::ClearReport::VtgPrimitivesOut:
+            return VideoCommon::QueryType::VtgPrimitivesOut;
+        default:
+            LOG_DEBUG(HW_GPU, "Unimplemented counter reset={}", clear_report);
+            return VideoCommon::QueryType::Payload;
+        }
+    }();
+    rasterizer->ResetCounter(query_type);
 }
 
 void Maxwell3D::ProcessSyncPoint() {

src/video_core/macro/macro_hle.cpp

@@ -327,12 +327,13 @@ public:
     explicit HLE_DrawIndirectByteCount(Maxwell3D& maxwell3d_) : HLEMacroImpl(maxwell3d_) {}
 
     void Execute(const std::vector<u32>& parameters, [[maybe_unused]] u32 method) override {
+        const bool force = maxwell3d.Rasterizer().HasDrawTransformFeedback();
+
         auto topology = static_cast<Maxwell3D::Regs::PrimitiveTopology>(parameters[0] & 0xFFFFU);
-        if (!maxwell3d.AnyParametersDirty() || !IsTopologySafe(topology)) {
+        if (!force && (!maxwell3d.AnyParametersDirty() || !IsTopologySafe(topology))) {
             Fallback(parameters);
             return;
         }
-
         auto& params = maxwell3d.draw_manager->GetIndirectParams();
         params.is_byte_count = true;
         params.is_indexed = false;
@@ -503,6 +504,8 @@ public:
         maxwell3d.CallMethod(static_cast<size_t>(MAXWELL3D_REG_INDEX(launch_dma)), 0x1011, true);
         maxwell3d.CallMethod(static_cast<size_t>(MAXWELL3D_REG_INDEX(inline_data)),
                              regs.transform_feedback.controls[0].stride, true);
+
+        maxwell3d.Rasterizer().RegisterTransformFeedback(regs.upload.dest.Address());
     }
 };
 

src/video_core/query_cache.h

@@ -28,8 +28,11 @@
 namespace VideoCore {
 enum class QueryType {
     SamplesPassed,
+    PrimitivesGenerated,
+    TfbPrimitivesWritten,
+    Count,
 };
-constexpr std::size_t NumQueryTypes = 1;
+constexpr std::size_t NumQueryTypes = static_cast<size_t>(QueryType::Count);
 } // namespace VideoCore
 
 namespace VideoCommon {
@@ -44,15 +47,6 @@ public:
     explicit CounterStreamBase(QueryCache& cache_, VideoCore::QueryType type_)
         : cache{cache_}, type{type_} {}
 
-    /// Updates the state of the stream, enabling or disabling as needed.
-    void Update(bool enabled) {
-        if (enabled) {
-            Enable();
-        } else {
-            Disable();
-        }
-    }
-
     /// Resets the stream to zero. It doesn't disable the query after resetting.
     void Reset() {
         if (current) {
@@ -80,7 +74,6 @@ public:
         return current != nullptr;
     }
 
-private:
     /// Enables the stream.
     void Enable() {
         if (current) {
@@ -97,6 +90,7 @@ private:
         last = std::exchange(current, nullptr);
     }
 
+private:
     QueryCache& cache;
     const VideoCore::QueryType type;
 
@@ -112,8 +106,14 @@ public:
         : rasterizer{rasterizer_},
           // Use reinterpret_cast instead of static_cast as workaround for
           // UBSan bug (https://github.com/llvm/llvm-project/issues/59060)
-          cpu_memory{cpu_memory_}, streams{{CounterStream{reinterpret_cast<QueryCache&>(*this),
-                                                          VideoCore::QueryType::SamplesPassed}}} {
+          cpu_memory{cpu_memory_}, streams{{
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::SamplesPassed}},
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::PrimitivesGenerated}},
+                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::TfbPrimitivesWritten}},
+                                   }} {
         (void)slot_async_jobs.insert(); // Null value
     }
 
@@ -157,12 +157,11 @@ public:
         AsyncFlushQuery(query, timestamp, lock);
     }
 
-    /// Updates counters from GPU state. Expected to be called once per draw, clear or dispatch.
-    void UpdateCounters() {
+    /// Enables all available GPU counters
+    void EnableCounters() {
         std::unique_lock lock{mutex};
-        if (maxwell3d) {
-            const auto& regs = maxwell3d->regs;
-            Stream(VideoCore::QueryType::SamplesPassed).Update(regs.zpass_pixel_count_enable);
+        for (auto& stream : streams) {
+            stream.Enable();
         }
     }
 
@@ -176,7 +175,7 @@ public:
     void DisableStreams() {
         std::unique_lock lock{mutex};
         for (auto& stream : streams) {
-            stream.Update(false);
+            stream.Disable();
         }
     }
 
@@ -353,7 +352,7 @@ private:
 
     std::shared_ptr<std::vector<AsyncJobId>> uncommitted_flushes{};
     std::list<std::shared_ptr<std::vector<AsyncJobId>>> committed_flushes;
-};
+}; // namespace VideoCommon
 
 template <class QueryCache, class HostCounter>
 class HostCounterBase {

src/video_core/rasterizer_interface.h

@@ -173,5 +173,13 @@ public:
     virtual void BindChannel(Tegra::Control::ChannelState& channel) {}
 
     virtual void ReleaseChannel(s32 channel_id) {}
+
+    /// Register the address as a Transform Feedback Object
+    virtual void RegisterTransformFeedback(GPUVAddr tfb_object_addr) {}
+
+    /// Returns true when the rasterizer has Draw Transform Feedback capabilities
+    virtual bool HasDrawTransformFeedback() {
+        return false;
+    }
 };
 } // namespace VideoCore

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -376,4 +376,15 @@ void BufferCacheRuntime::BindImageBuffer(Buffer& buffer, u32 offset, u32 size, P
     *image_handles++ = buffer.View(offset, size, format);
 }
 
+void BufferCacheRuntime::BindTransformFeedbackObject(GPUVAddr tfb_object_addr) {
+    OGLTransformFeedback& tfb_object = tfb_objects[tfb_object_addr];
+    tfb_object.Create();
+    glBindTransformFeedback(GL_TRANSFORM_FEEDBACK, tfb_object.handle);
+}
+
+GLuint BufferCacheRuntime::GetTransformFeedbackObject(GPUVAddr tfb_object_addr) {
+    ASSERT(tfb_objects.contains(tfb_object_addr));
+    return tfb_objects[tfb_object_addr].handle;
+}
+
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -5,6 +5,7 @@
 
 #include <array>
 #include <span>
+#include <unordered_map>
 
 #include "common/common_types.h"
 #include "video_core/buffer_cache/buffer_cache_base.h"
@@ -121,6 +122,9 @@ public:
     void BindImageBuffer(Buffer& buffer, u32 offset, u32 size,
                          VideoCore::Surface::PixelFormat format);
 
+    void BindTransformFeedbackObject(GPUVAddr tfb_object_addr);
+    GLuint GetTransformFeedbackObject(GPUVAddr tfb_object_addr);
+
     u64 GetDeviceMemoryUsage() const;
 
     void BindFastUniformBuffer(size_t stage, u32 binding_index, u32 size) {
@@ -233,6 +237,7 @@ private:
     u32 index_buffer_offset = 0;
 
     u64 device_access_memory;
+    std::unordered_map<GPUVAddr, OGLTransformFeedback> tfb_objects;
 };
 
 struct BufferCacheParams {

src/video_core/renderer_opengl/gl_query_cache.cpp

@@ -18,16 +18,27 @@ namespace OpenGL {
 
 namespace {
 
-constexpr std::array<GLenum, VideoCore::NumQueryTypes> QueryTargets = {GL_SAMPLES_PASSED};
-
 constexpr GLenum GetTarget(VideoCore::QueryType type) {
-    return QueryTargets[static_cast<std::size_t>(type)];
+    switch (type) {
+    case VideoCore::QueryType::SamplesPassed:
+        return GL_SAMPLES_PASSED;
+    case VideoCore::QueryType::PrimitivesGenerated:
+        return GL_PRIMITIVES_GENERATED;
+    case VideoCore::QueryType::TfbPrimitivesWritten:
+        return GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN;
+    default:
+        break;
+    }
+    UNIMPLEMENTED_MSG("Query type {}", type);
+    return 0;
 }
 
 } // Anonymous namespace
 
 QueryCache::QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_)
-    : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {}
+    : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {
+    EnableCounters();
+}
 
 QueryCache::~QueryCache() = default;
 
@@ -103,13 +114,13 @@ u64 CachedQuery::Flush([[maybe_unused]] bool async) {
     auto& stream = cache->Stream(type);
     const bool slice_counter = WaitPending() && stream.IsEnabled();
     if (slice_counter) {
-        stream.Update(false);
+        stream.Disable();
     }
 
     auto result = VideoCommon::CachedQueryBase<HostCounter>::Flush();
 
     if (slice_counter) {
-        stream.Update(true);
+        stream.Enable();
     }
 
     return result;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -51,6 +51,22 @@ constexpr size_t NUM_SUPPORTED_VERTEX_ATTRIBUTES = 16;
 void oglEnable(GLenum cap, bool state) {
     (state ? glEnable : glDisable)(cap);
 }
+
+std::optional<VideoCore::QueryType> MaxwellToVideoCoreQuery(VideoCommon::QueryType type) {
+    switch (type) {
+    case VideoCommon::QueryType::PrimitivesGenerated:
+    case VideoCommon::QueryType::VtgPrimitivesOut:
+        return VideoCore::QueryType::PrimitivesGenerated;
+    case VideoCommon::QueryType::ZPassPixelCount64:
+        return VideoCore::QueryType::SamplesPassed;
+    case VideoCommon::QueryType::StreamingPrimitivesSucceeded:
+        // case VideoCommon::QueryType::StreamingByteCount:
+        // TODO: StreamingByteCount = StreamingPrimitivesSucceeded * num_verts * vert_stride
+        return VideoCore::QueryType::TfbPrimitivesWritten;
+    default:
+        return std::nullopt;
+    }
+}
 } // Anonymous namespace
 
 RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
@@ -216,7 +232,6 @@ void RasterizerOpenGL::PrepareDraw(bool is_indexed, Func&& draw_func) {
 
     SCOPE_EXIT({ gpu.TickWork(); });
     gpu_memory->FlushCaching();
-    query_cache.UpdateCounters();
 
     GraphicsPipeline* const pipeline{shader_cache.CurrentGraphicsPipeline()};
     if (!pipeline) {
@@ -294,6 +309,13 @@ void RasterizerOpenGL::DrawIndirect() {
     const auto& params = maxwell3d->draw_manager->GetIndirectParams();
     buffer_cache.SetDrawIndirect(&params);
     PrepareDraw(params.is_indexed, [this, &params](GLenum primitive_mode) {
+        if (params.is_byte_count) {
+            const GPUVAddr tfb_object_base_addr = params.indirect_start_address - 4U;
+            const GLuint tfb_object =
+                buffer_cache_runtime.GetTransformFeedbackObject(tfb_object_base_addr);
+            glDrawTransformFeedback(primitive_mode, tfb_object);
+            return;
+        }
         const auto [buffer, offset] = buffer_cache.GetDrawIndirectBuffer();
         const GLvoid* const gl_offset =
             reinterpret_cast<const GLvoid*>(static_cast<uintptr_t>(offset));
@@ -334,7 +356,6 @@ void RasterizerOpenGL::DrawTexture() {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
-    query_cache.UpdateCounters();
 
     texture_cache.SynchronizeGraphicsDescriptors();
     texture_cache.UpdateRenderTargets(false);
@@ -401,21 +422,28 @@ void RasterizerOpenGL::DispatchCompute() {
 }
 
 void RasterizerOpenGL::ResetCounter(VideoCommon::QueryType type) {
-    if (type == VideoCommon::QueryType::ZPassPixelCount64) {
-        query_cache.ResetCounter(VideoCore::QueryType::SamplesPassed);
+    const auto query_cache_type = MaxwellToVideoCoreQuery(type);
+    if (!query_cache_type.has_value()) {
+        UNIMPLEMENTED_IF_MSG(type != VideoCommon::QueryType::Payload, "Reset query type: {}", type);
+        return;
     }
+    query_cache.ResetCounter(*query_cache_type);
 }
 
 void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCommon::QueryType type,
                              VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) {
-    if (type == VideoCommon::QueryType::ZPassPixelCount64) {
-        if (True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout)) {
-            query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, {gpu.GetTicks()});
-        } else {
-            query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, std::nullopt);
-        }
-        return;
+    const auto query_cache_type = MaxwellToVideoCoreQuery(type);
+    if (!query_cache_type.has_value()) {
+        return QueryFallback(gpu_addr, type, flags, payload, subreport);
     }
+    const bool has_timeout = True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout);
+    const auto timestamp = has_timeout ? std::optional<u64>{gpu.GetTicks()} : std::nullopt;
+    query_cache.Query(gpu_addr, *query_cache_type, timestamp);
+}
+
+void RasterizerOpenGL::QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
+                                     VideoCommon::QueryPropertiesFlags flags, u32 payload,
+                                     u32 subreport) {
     if (type != VideoCommon::QueryType::Payload) {
         payload = 1u;
     }
@@ -1350,6 +1378,10 @@ void RasterizerOpenGL::ReleaseChannel(s32 channel_id) {
     query_cache.EraseChannel(channel_id);
 }
 
+void RasterizerOpenGL::RegisterTransformFeedback(GPUVAddr tfb_object_addr) {
+    buffer_cache_runtime.BindTransformFeedbackObject(tfb_object_addr);
+}
+
 AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_, TextureCache& texture_cache_)
     : buffer_cache{buffer_cache_}, texture_cache{texture_cache_} {}
 

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -139,6 +139,12 @@ public:
 
     void ReleaseChannel(s32 channel_id) override;
 
+    void RegisterTransformFeedback(GPUVAddr tfb_object_addr) override;
+
+    bool HasDrawTransformFeedback() override {
+        return true;
+    }
+
 private:
     static constexpr size_t MAX_TEXTURES = 192;
     static constexpr size_t MAX_IMAGES = 48;
@@ -225,6 +231,9 @@ private:
     /// End a transform feedback
     void EndTransformFeedback();
 
+    void QueryFallback(GPUVAddr gpu_addr, VideoCommon::QueryType type,
+                       VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport);
+
     Tegra::GPU& gpu;
 
     const Device& device;

src/video_core/renderer_opengl/gl_resource_manager.cpp

@@ -207,4 +207,21 @@ void OGLQuery::Release() {
     handle = 0;
 }
 
+void OGLTransformFeedback::Create() {
+    if (handle != 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
+    glCreateTransformFeedbacks(1, &handle);
+}
+
+void OGLTransformFeedback::Release() {
+    if (handle == 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
+    glDeleteTransformFeedbacks(1, &handle);
+    handle = 0;
+}
+
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_resource_manager.h

@@ -323,4 +323,31 @@ public:
     GLuint handle = 0;
 };
 
+class OGLTransformFeedback final {
+public:
+    YUZU_NON_COPYABLE(OGLTransformFeedback);
+
+    OGLTransformFeedback() = default;
+
+    OGLTransformFeedback(OGLTransformFeedback&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+
+    ~OGLTransformFeedback() {
+        Release();
+    }
+
+    OGLTransformFeedback& operator=(OGLTransformFeedback&& o) noexcept {
+        Release();
+        handle = std::exchange(o.handle, 0);
+        return *this;
+    }
+
+    /// Creates a new internal OpenGL resource and stores the handle
+    void Create();
+
+    /// Deletes the internal OpenGL resource
+    void Release();
+
+    GLuint handle = 0;
+};
+
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -51,7 +51,7 @@ using VideoCommon::LoadPipelines;
 using VideoCommon::SerializePipeline;
 using Context = ShaderContext::Context;
 
-constexpr u32 CACHE_VERSION = 9;
+constexpr u32 CACHE_VERSION = 10;
 
 template <typename Container>
 auto MakeSpan(Container& container) {

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -78,8 +78,15 @@ vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allo
 }
 } // Anonymous namespace
 
-Buffer::Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params)
-    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params), tracker{4096} {}
+Buffer::Buffer(BufferCacheRuntime& runtime, VideoCommon::NullBufferParams null_params)
+    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params), tracker{4096} {
+    if (runtime.device.HasNullDescriptor()) {
+        return;
+    }
+    device = &runtime.device;
+    buffer = runtime.CreateNullBuffer();
+    is_null = true;
+}
 
 Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rasterizer_,
                VAddr cpu_addr_, u64 size_bytes_)
@@ -93,8 +100,12 @@ Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rast
 
 VkBufferView Buffer::View(u32 offset, u32 size, VideoCore::Surface::PixelFormat format) {
     if (!device) {
-        // Null buffer, return a null descriptor
+        // Null buffer supported, return a null descriptor
         return VK_NULL_HANDLE;
+    } else if (is_null) {
+        // Null buffer not supported, adjust offset and size
+        offset = 0;
+        size = 0;
     }
     const auto it{std::ranges::find_if(views, [offset, size, format](const BufferView& view) {
         return offset == view.offset && size == view.size && format == view.format;
@@ -627,9 +638,12 @@ void BufferCacheRuntime::BindTransformFeedbackBuffers(VideoCommon::HostBindings<
 }
 
 void BufferCacheRuntime::ReserveNullBuffer() {
-    if (null_buffer) {
-        return;
+    if (!null_buffer) {
+        null_buffer = CreateNullBuffer();
     }
+}
+
+vk::Buffer BufferCacheRuntime::CreateNullBuffer() {
     VkBufferCreateInfo create_info{
         .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
         .pNext = nullptr,
@@ -644,15 +658,17 @@ void BufferCacheRuntime::ReserveNullBuffer() {
     if (device.IsExtTransformFeedbackSupported()) {
         create_info.usage |= VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT;
     }
-    null_buffer = memory_allocator.CreateBuffer(create_info, MemoryUsage::DeviceLocal);
+    vk::Buffer ret = memory_allocator.CreateBuffer(create_info, MemoryUsage::DeviceLocal);
     if (device.HasDebuggingToolAttached()) {
-        null_buffer.SetObjectNameEXT("Null buffer");
+        ret.SetObjectNameEXT("Null buffer");
     }
 
     scheduler.RequestOutsideRenderPassOperationContext();
-    scheduler.Record([buffer = *null_buffer](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([buffer = *ret](vk::CommandBuffer cmdbuf) {
         cmdbuf.FillBuffer(buffer, 0, VK_WHOLE_SIZE, 0);
     });
+
+    return ret;
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -63,6 +63,7 @@ private:
     vk::Buffer buffer;
     std::vector<BufferView> views;
     VideoCommon::UsageTracker tracker;
+    bool is_null{};
 };
 
 class QuadArrayIndexBuffer;
@@ -151,6 +152,7 @@ private:
     }
 
     void ReserveNullBuffer();
+    vk::Buffer CreateNullBuffer();
 
     const Device& device;
     MemoryAllocator& memory_allocator;

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -54,7 +54,7 @@ using VideoCommon::FileEnvironment;
 using VideoCommon::GenericEnvironment;
 using VideoCommon::GraphicsEnvironment;
 
-constexpr u32 CACHE_VERSION = 10;
+constexpr u32 CACHE_VERSION = 11;
 constexpr std::array<char, 8> VULKAN_CACHE_MAGIC_NUMBER{'y', 'u', 'z', 'u', 'v', 'k', 'c', 'h'};
 
 template <typename Container>

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -485,6 +485,10 @@ void RasterizerVulkan::DispatchCompute() {
 }
 
 void RasterizerVulkan::ResetCounter(VideoCommon::QueryType type) {
+    if (type != VideoCommon::QueryType::ZPassPixelCount64) {
+        LOG_DEBUG(Render_Vulkan, "Unimplemented counter reset={}", type);
+        return;
+    }
     query_cache.CounterReset(type);
 }
 

src/video_core/shader_environment.cpp

@@ -62,23 +62,9 @@ static Shader::TextureType ConvertTextureType(const Tegra::Texture::TICEntry& en
 }
 
 static Shader::TexturePixelFormat ConvertTexturePixelFormat(const Tegra::Texture::TICEntry& entry) {
-    switch (PixelFormatFromTextureInfo(entry.format, entry.r_type, entry.g_type, entry.b_type,
-                                       entry.a_type, entry.srgb_conversion)) {
-    case VideoCore::Surface::PixelFormat::A8B8G8R8_SNORM:
-        return Shader::TexturePixelFormat::A8B8G8R8_SNORM;
-    case VideoCore::Surface::PixelFormat::R8_SNORM:
-        return Shader::TexturePixelFormat::R8_SNORM;
-    case VideoCore::Surface::PixelFormat::R8G8_SNORM:
-        return Shader::TexturePixelFormat::R8G8_SNORM;
-    case VideoCore::Surface::PixelFormat::R16G16B16A16_SNORM:
-        return Shader::TexturePixelFormat::R16G16B16A16_SNORM;
-    case VideoCore::Surface::PixelFormat::R16G16_SNORM:
-        return Shader::TexturePixelFormat::R16G16_SNORM;
-    case VideoCore::Surface::PixelFormat::R16_SNORM:
-        return Shader::TexturePixelFormat::R16_SNORM;
-    default:
-        return Shader::TexturePixelFormat::OTHER;
-    }
+    return static_cast<Shader::TexturePixelFormat>(
+        PixelFormatFromTextureInfo(entry.format, entry.r_type, entry.g_type, entry.b_type,
+                                   entry.a_type, entry.srgb_conversion));
 }
 
 static std::string_view StageToPrefix(Shader::Stage stage) {
@@ -398,6 +384,11 @@ Shader::TexturePixelFormat GraphicsEnvironment::ReadTexturePixelFormat(u32 handl
     return result;
 }
 
+bool GraphicsEnvironment::IsTexturePixelFormatInteger(u32 handle) {
+    return VideoCore::Surface::IsPixelFormatInteger(
+        static_cast<VideoCore::Surface::PixelFormat>(ReadTexturePixelFormat(handle)));
+}
+
 u32 GraphicsEnvironment::ReadViewportTransformState() {
     const auto& regs{maxwell3d->regs};
     viewport_transform_state = regs.viewport_scale_offset_enabled;
@@ -448,6 +439,11 @@ Shader::TexturePixelFormat ComputeEnvironment::ReadTexturePixelFormat(u32 handle
     return result;
 }
 
+bool ComputeEnvironment::IsTexturePixelFormatInteger(u32 handle) {
+    return VideoCore::Surface::IsPixelFormatInteger(
+        static_cast<VideoCore::Surface::PixelFormat>(ReadTexturePixelFormat(handle)));
+}
+
 u32 ComputeEnvironment::ReadViewportTransformState() {
     return viewport_transform_state;
 }
@@ -551,6 +547,11 @@ Shader::TexturePixelFormat FileEnvironment::ReadTexturePixelFormat(u32 handle) {
     return it->second;
 }
 
+bool FileEnvironment::IsTexturePixelFormatInteger(u32 handle) {
+    return VideoCore::Surface::IsPixelFormatInteger(
+        static_cast<VideoCore::Surface::PixelFormat>(ReadTexturePixelFormat(handle)));
+}
+
 u32 FileEnvironment::ReadViewportTransformState() {
     return viewport_transform_state;
 }

src/video_core/shader_environment.h

@@ -115,6 +115,8 @@ public:
 
     Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) override;
 
+    bool IsTexturePixelFormatInteger(u32 handle) override;
+
     u32 ReadViewportTransformState() override;
 
     std::optional<Shader::ReplaceConstant> GetReplaceConstBuffer(u32 bank, u32 offset) override;
@@ -139,6 +141,8 @@ public:
 
     Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) override;
 
+    bool IsTexturePixelFormatInteger(u32 handle) override;
+
     u32 ReadViewportTransformState() override;
 
     std::optional<Shader::ReplaceConstant> GetReplaceConstBuffer(
@@ -171,6 +175,8 @@ public:
 
     [[nodiscard]] Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) override;
 
+    [[nodiscard]] bool IsTexturePixelFormatInteger(u32 handle) override;
+
     [[nodiscard]] u32 ReadViewportTransformState() override;
 
     [[nodiscard]] u32 LocalMemorySize() const override;

src/video_core/texture_cache/decode_bc.cpp

@@ -60,66 +60,72 @@ u32 ConvertedBytesPerBlock(VideoCore::Surface::PixelFormat pixel_format) {
 }
 
 template <auto decompress, PixelFormat pixel_format>
-void DecompressBlocks(std::span<const u8> input, std::span<u8> output, Extent3D extent,
+void DecompressBlocks(std::span<const u8> input, std::span<u8> output, BufferImageCopy& copy,
                       bool is_signed = false) {
     const u32 out_bpp = ConvertedBytesPerBlock(pixel_format);
-    const u32 block_width = std::min(extent.width, BLOCK_SIZE);
-    const u32 block_height = std::min(extent.height, BLOCK_SIZE);
-    const u32 pitch = extent.width * out_bpp;
+    const u32 block_size = BlockSize(pixel_format);
+    const u32 width = copy.image_extent.width;
+    const u32 height = copy.image_extent.height * copy.image_subresource.num_layers;
+    const u32 depth = copy.image_extent.depth;
+    const u32 block_width = std::min(width, BLOCK_SIZE);
+    const u32 block_height = std::min(height, BLOCK_SIZE);
+    const u32 pitch = width * out_bpp;
     size_t input_offset = 0;
     size_t output_offset = 0;
-    for (u32 slice = 0; slice < extent.depth; ++slice) {
-        for (u32 y = 0; y < extent.height; y += block_height) {
-            size_t row_offset = 0;
-            for (u32 x = 0; x < extent.width;
-                 x += block_width, row_offset += block_width * out_bpp) {
-                const u8* src = input.data() + input_offset;
-                u8* const dst = output.data() + output_offset + row_offset;
+    for (u32 slice = 0; slice < depth; ++slice) {
+        for (u32 y = 0; y < height; y += block_height) {
+            size_t src_offset = input_offset;
+            size_t dst_offset = output_offset;
+            for (u32 x = 0; x < width; x += block_width) {
+                const u8* src = input.data() + src_offset;
+                u8* const dst = output.data() + dst_offset;
                 if constexpr (IsSigned(pixel_format)) {
-                    decompress(src, dst, x, y, extent.width, extent.height, is_signed);
+                    decompress(src, dst, x, y, width, height, is_signed);
                 } else {
-                    decompress(src, dst, x, y, extent.width, extent.height);
+                    decompress(src, dst, x, y, width, height);
                 }
-                input_offset += BlockSize(pixel_format);
+                src_offset += block_size;
+                dst_offset += block_width * out_bpp;
             }
+            input_offset += copy.buffer_row_length * block_size / block_width;
             output_offset += block_height * pitch;
         }
     }
 }
 
-void DecompressBCn(std::span<const u8> input, std::span<u8> output, Extent3D extent,
+void DecompressBCn(std::span<const u8> input, std::span<u8> output, BufferImageCopy& copy,
                    VideoCore::Surface::PixelFormat pixel_format) {
     switch (pixel_format) {
     case PixelFormat::BC1_RGBA_UNORM:
     case PixelFormat::BC1_RGBA_SRGB:
-        DecompressBlocks<bcn::DecodeBc1, PixelFormat::BC1_RGBA_UNORM>(input, output, extent);
+        DecompressBlocks<bcn::DecodeBc1, PixelFormat::BC1_RGBA_UNORM>(input, output, copy);
         break;
     case PixelFormat::BC2_UNORM:
     case PixelFormat::BC2_SRGB:
-        DecompressBlocks<bcn::DecodeBc2, PixelFormat::BC2_UNORM>(input, output, extent);
+        DecompressBlocks<bcn::DecodeBc2, PixelFormat::BC2_UNORM>(input, output, copy);
         break;
     case PixelFormat::BC3_UNORM:
     case PixelFormat::BC3_SRGB:
-        DecompressBlocks<bcn::DecodeBc3, PixelFormat::BC3_UNORM>(input, output, extent);
+        DecompressBlocks<bcn::DecodeBc3, PixelFormat::BC3_UNORM>(input, output, copy);
         break;
     case PixelFormat::BC4_SNORM:
     case PixelFormat::BC4_UNORM:
         DecompressBlocks<bcn::DecodeBc4, PixelFormat::BC4_UNORM>(
-            input, output, extent, pixel_format == PixelFormat::BC4_SNORM);
+            input, output, copy, pixel_format == PixelFormat::BC4_SNORM);
         break;
     case PixelFormat::BC5_SNORM:
     case PixelFormat::BC5_UNORM:
         DecompressBlocks<bcn::DecodeBc5, PixelFormat::BC5_UNORM>(
-            input, output, extent, pixel_format == PixelFormat::BC5_SNORM);
+            input, output, copy, pixel_format == PixelFormat::BC5_SNORM);
         break;
     case PixelFormat::BC6H_SFLOAT:
     case PixelFormat::BC6H_UFLOAT:
         DecompressBlocks<bcn::DecodeBc6, PixelFormat::BC6H_UFLOAT>(
-            input, output, extent, pixel_format == PixelFormat::BC6H_SFLOAT);
+            input, output, copy, pixel_format == PixelFormat::BC6H_SFLOAT);
         break;
     case PixelFormat::BC7_SRGB:
     case PixelFormat::BC7_UNORM:
-        DecompressBlocks<bcn::DecodeBc7, PixelFormat::BC7_UNORM>(input, output, extent);
+        DecompressBlocks<bcn::DecodeBc7, PixelFormat::BC7_UNORM>(input, output, copy);
         break;
     default:
         LOG_WARNING(HW_GPU, "Unimplemented BCn decompression {}", pixel_format);

src/video_core/texture_cache/decode_bc.h

@@ -13,7 +13,7 @@ namespace VideoCommon {
 
 [[nodiscard]] u32 ConvertedBytesPerBlock(VideoCore::Surface::PixelFormat pixel_format);
 
-void DecompressBCn(std::span<const u8> input, std::span<u8> output, Extent3D extent,
+void DecompressBCn(std::span<const u8> input, std::span<u8> output, BufferImageCopy& copy,
                    VideoCore::Surface::PixelFormat pixel_format);
 
 } // namespace VideoCommon

src/video_core/texture_cache/util.cpp

@@ -837,6 +837,7 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
                                                                    std::span<u8> output) {
     const size_t guest_size_bytes = input.size_bytes();
     const u32 bpp_log2 = BytesPerBlockLog2(info.format);
+    const Extent2D tile_size = DefaultBlockSize(info.format);
     const Extent3D size = info.size;
 
     if (info.type == ImageType::Linear) {
@@ -847,7 +848,7 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
         return {{
             .buffer_offset = 0,
             .buffer_size = guest_size_bytes,
-            .buffer_row_length = info.pitch >> bpp_log2,
+            .buffer_row_length = info.pitch * tile_size.width >> bpp_log2,
             .buffer_image_height = size.height,
             .image_subresource =
                 {
@@ -862,7 +863,6 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
     const LevelInfo level_info = MakeLevelInfo(info);
     const s32 num_layers = info.resources.layers;
     const s32 num_levels = info.resources.levels;
-    const Extent2D tile_size = DefaultBlockSize(info.format);
     const std::array level_sizes = CalculateLevelSizes(level_info, num_levels);
     const Extent2D gob = GobSize(bpp_log2, info.block.height, info.tile_width_spacing);
     const u32 layer_size = CalculateLevelBytes(level_sizes, num_levels);
@@ -926,8 +926,6 @@ void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8
 
         const auto input_offset = input.subspan(copy.buffer_offset);
         copy.buffer_offset = output_offset;
-        copy.buffer_row_length = mip_size.width;
-        copy.buffer_image_height = mip_size.height;
 
         const auto recompression_setting = Settings::values.astc_recompression.GetValue();
         const bool astc = IsPixelFormatASTC(info.format);
@@ -972,16 +970,14 @@ void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8
                 bpp_div;
             output_offset += static_cast<u32>(copy.buffer_size);
         } else {
-            const Extent3D image_extent{
-                .width = copy.image_extent.width,
-                .height = copy.image_extent.height * copy.image_subresource.num_layers,
-                .depth = copy.image_extent.depth,
-            };
-            DecompressBCn(input_offset, output.subspan(output_offset), image_extent, info.format);
+            DecompressBCn(input_offset, output.subspan(output_offset), copy, info.format);
             output_offset += copy.image_extent.width * copy.image_extent.height *
                              copy.image_subresource.num_layers *
                              ConvertedBytesPerBlock(info.format);
         }
+
+        copy.buffer_row_length = mip_size.width;
+        copy.buffer_image_height = mip_size.height;
     }
 }
 

src/video_core/vulkan_common/vulkan_device.cpp

@@ -755,10 +755,10 @@ VkFormat Device::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags
     // The wanted format is not supported by hardware, search for alternatives
     const VkFormat* alternatives = GetFormatAlternatives(wanted_format);
     if (alternatives == nullptr) {
-        ASSERT_MSG(false,
-                   "Format={} with usage={} and type={} has no defined alternatives and host "
-                   "hardware does not support it",
-                   wanted_format, wanted_usage, format_type);
+        LOG_ERROR(Render_Vulkan,
+                  "Format={} with usage={} and type={} has no defined alternatives and host "
+                  "hardware does not support it",
+                  wanted_format, wanted_usage, format_type);
         return wanted_format;
     }
 
@@ -774,10 +774,10 @@ VkFormat Device::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags
     }
 
     // No alternatives found, panic
-    ASSERT_MSG(false,
-               "Format={} with usage={} and type={} is not supported by the host hardware and "
-               "doesn't support any of the alternatives",
-               wanted_format, wanted_usage, format_type);
+    LOG_ERROR(Render_Vulkan,
+              "Format={} with usage={} and type={} is not supported by the host hardware and "
+              "doesn't support any of the alternatives",
+              wanted_format, wanted_usage, format_type);
     return wanted_format;
 }
 

src/video_core/vulkan_common/vulkan_wrapper.cpp

@@ -246,7 +246,9 @@ void SetObjectName(const DeviceDispatch* dld, VkDevice device, T handle, VkObjec
         .objectHandle = reinterpret_cast<u64>(handle),
         .pObjectName = name,
     };
-    Check(dld->vkSetDebugUtilsObjectNameEXT(device, &name_info));
+    if (dld->vkSetDebugUtilsObjectNameEXT) {
+        Check(dld->vkSetDebugUtilsObjectNameEXT(device, &name_info));
+    }
 }
 
 } // Anonymous namespace