kernel: memory: Add PageHeap class, to manage a heap of pages.

src/core/CMakeLists.txt

@@ -159,6 +159,8 @@ add_library(core STATIC
     hle/kernel/memory/memory_block.h
     hle/kernel/memory/memory_types.h
     hle/kernel/memory/page_linked_list.h
+    hle/kernel/memory/page_heap.cpp
+    hle/kernel/memory/page_heap.h
     hle/kernel/memory/slab_heap.h
     hle/kernel/memory/system_control.cpp
     hle/kernel/memory/system_control.h

src/core/hle/kernel/memory/page_heap.cpp

@@ -0,0 +1,116 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/core.h"
+#include "core/hle/kernel/memory/page_heap.h"
+#include "core/memory.h"
+
+namespace Kernel::Memory {
+
+void PageHeap::Initialize(VAddr address, std::size_t size, std::size_t metadata_size) {
+    // Check our assumptions
+    ASSERT(Common::IsAligned((address), PageSize));
+    ASSERT(Common::IsAligned(size, PageSize));
+
+    // Set our members
+    heap_address = address;
+    heap_size = size;
+
+    // Setup bitmaps
+    metadata.resize(metadata_size / sizeof(u64));
+    u64* cur_bitmap_storage{metadata.data()};
+    for (std::size_t i = 0; i < MemoryBlockPageShifts.size(); i++) {
+        const std::size_t cur_block_shift{MemoryBlockPageShifts[i]};
+        const std::size_t next_block_shift{
+            (i != MemoryBlockPageShifts.size() - 1) ? MemoryBlockPageShifts[i + 1] : 0};
+        cur_bitmap_storage = blocks[i].Initialize(heap_address, heap_size, cur_block_shift,
+                                                  next_block_shift, cur_bitmap_storage);
+    }
+}
+
+VAddr PageHeap::AllocateBlock(s32 index) {
+    const std::size_t needed_size{blocks[index].GetSize()};
+
+    for (s32 i{index}; i < static_cast<s32>(MemoryBlockPageShifts.size()); i++) {
+        if (const VAddr addr{blocks[i].PopBlock()}; addr) {
+            if (const std::size_t allocated_size{blocks[i].GetSize()};
+                allocated_size > needed_size) {
+                Free(addr + needed_size, (allocated_size - needed_size) / PageSize);
+            }
+            return addr;
+        }
+    }
+
+    return 0;
+}
+
+void PageHeap::FreeBlock(VAddr block, s32 index) {
+    do {
+        block = blocks[index++].PushBlock(block);
+    } while (block != 0);
+}
+
+void PageHeap::Free(VAddr addr, std::size_t num_pages) {
+    // Freeing no pages is a no-op
+    if (num_pages == 0) {
+        return;
+    }
+
+    // Find the largest block size that we can free, and free as many as possible
+    s32 big_index{static_cast<s32>(MemoryBlockPageShifts.size()) - 1};
+    const VAddr start{addr};
+    const VAddr end{(num_pages * PageSize) + addr};
+    VAddr before_start{start};
+    VAddr before_end{start};
+    VAddr after_start{end};
+    VAddr after_end{end};
+    while (big_index >= 0) {
+        const std::size_t block_size{blocks[big_index].GetSize()};
+        const VAddr big_start{Common::AlignUp((start), block_size)};
+        const VAddr big_end{Common::AlignDown((end), block_size)};
+        if (big_start < big_end) {
+            // Free as many big blocks as we can
+            for (auto block{big_start}; block < big_end; block += block_size) {
+                FreeBlock(block, big_index);
+            }
+            before_end = big_start;
+            after_start = big_end;
+            break;
+        }
+        big_index--;
+    }
+    ASSERT(big_index >= 0);
+
+    // Free space before the big blocks
+    for (s32 i{big_index - 1}; i >= 0; i--) {
+        const std::size_t block_size{blocks[i].GetSize()};
+        while (before_start + block_size <= before_end) {
+            before_end -= block_size;
+            FreeBlock(before_end, i);
+        }
+    }
+
+    // Free space after the big blocks
+    for (s32 i{big_index - 1}; i >= 0; i--) {
+        const std::size_t block_size{blocks[i].GetSize()};
+        while (after_start + block_size <= after_end) {
+            FreeBlock(after_start, i);
+            after_start += block_size;
+        }
+    }
+}
+
+std::size_t PageHeap::CalculateMetadataOverheadSize(std::size_t region_size) {
+    std::size_t overhead_size = 0;
+    for (std::size_t i = 0; i < MemoryBlockPageShifts.size(); i++) {
+        const std::size_t cur_block_shift{MemoryBlockPageShifts[i]};
+        const std::size_t next_block_shift{
+            (i != MemoryBlockPageShifts.size() - 1) ? MemoryBlockPageShifts[i + 1] : 0};
+        overhead_size += PageHeap::Block::CalculateMetadataOverheadSize(
+            region_size, cur_block_shift, next_block_shift);
+    }
+    return Common::AlignUp(overhead_size, PageSize);
+}
+
+} // namespace Kernel::Memory

src/core/hle/kernel/memory/page_heap.h

@@ -0,0 +1,365 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <vector>
+
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/bit_util.h"
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+#include "core/hle/kernel/memory/memory_types.h"
+
+namespace Kernel::Memory {
+
+class PageHeap final : NonCopyable {
+public:
+    static constexpr s32 GetAlignedBlockIndex(std::size_t num_pages, std::size_t align_pages) {
+        const std::size_t target_pages{std::max(num_pages, align_pages)};
+        for (std::size_t i = 0; i < NumMemoryBlockPageShifts; i++) {
+            if (target_pages <= (std::size_t(1) << MemoryBlockPageShifts[i]) / PageSize) {
+                return static_cast<s32>(i);
+            }
+        }
+        return -1;
+    }
+
+    static constexpr s32 GetBlockIndex(std::size_t num_pages) {
+        for (s32 i{static_cast<s32>(NumMemoryBlockPageShifts) - 1}; i >= 0; i--) {
+            if (num_pages >= (std::size_t(1) << MemoryBlockPageShifts[i]) / PageSize) {
+                return i;
+            }
+        }
+        return -1;
+    }
+
+    static constexpr std::size_t GetBlockSize(std::size_t index) {
+        return std::size_t(1) << MemoryBlockPageShifts[index];
+    }
+
+    static constexpr std::size_t GetBlockNumPages(std::size_t index) {
+        return GetBlockSize(index) / PageSize;
+    }
+
+private:
+    static constexpr std::size_t NumMemoryBlockPageShifts{7};
+    static constexpr std::array<std::size_t, NumMemoryBlockPageShifts> MemoryBlockPageShifts{
+        0xC, 0x10, 0x15, 0x16, 0x19, 0x1D, 0x1E};
+
+    class Block final : NonCopyable {
+    private:
+        class Bitmap final : NonCopyable {
+        public:
+            static constexpr std::size_t MaxDepth{4};
+
+        private:
+            std::array<u64*, MaxDepth> bit_storages{};
+            std::size_t num_bits{};
+            std::size_t used_depths{};
+
+        public:
+            constexpr Bitmap() = default;
+
+            constexpr std::size_t GetNumBits() const {
+                return num_bits;
+            }
+            constexpr s32 GetHighestDepthIndex() const {
+                return static_cast<s32>(used_depths) - 1;
+            }
+
+            constexpr u64* Initialize(u64* storage, std::size_t size) {
+                //* Initially, everything is un-set
+                num_bits = 0;
+
+                // Calculate the needed bitmap depth
+                used_depths = static_cast<std::size_t>(GetRequiredDepth(size));
+                ASSERT(used_depths <= MaxDepth);
+
+                // Set the bitmap pointers
+                for (s32 depth{GetHighestDepthIndex()}; depth >= 0; depth--) {
+                    bit_storages[depth] = storage;
+                    size = Common::AlignUp(size, 64) / 64;
+                    storage += size;
+                }
+
+                return storage;
+            }
+
+            s64 FindFreeBlock() const {
+                uintptr_t offset{};
+                s32 depth{};
+
+                do {
+                    const u64 v{bit_storages[depth][offset]};
+                    if (v == 0) {
+                        // Non-zero depth indicates that a previous level had a free block
+                        ASSERT(depth == 0);
+                        return -1;
+                    }
+                    offset = offset * 64 + Common::CountTrailingZeroes64(v);
+                    ++depth;
+                } while (depth < static_cast<s32>(used_depths));
+
+                return static_cast<s64>(offset);
+            }
+
+            constexpr void SetBit(std::size_t offset) {
+                SetBit(GetHighestDepthIndex(), offset);
+                num_bits++;
+            }
+
+            constexpr void ClearBit(std::size_t offset) {
+                ClearBit(GetHighestDepthIndex(), offset);
+                num_bits--;
+            }
+
+            constexpr bool ClearRange(std::size_t offset, std::size_t count) {
+                const s32 depth{GetHighestDepthIndex()};
+                const std::size_t bit_ind{offset / 64};
+                u64* bits{bit_storages[depth]};
+                if (count < 64) {
+                    const std::size_t shift{offset % 64};
+                    ASSERT(shift + count <= 64);
+                    // Check that all the bits are set
+                    const u64 mask{((u64(1) << count) - 1) << shift};
+                    u64 v{bits[bit_ind]};
+                    if ((v & mask) != mask) {
+                        return false;
+                    }
+
+                    // Clear the bits
+                    v &= ~mask;
+                    bits[bit_ind] = v;
+                    if (v == 0) {
+                        ClearBit(depth - 1, bit_ind);
+                    }
+                } else {
+                    ASSERT(offset % 64 == 0);
+                    ASSERT(count % 64 == 0);
+                    // Check that all the bits are set
+                    std::size_t remaining{count};
+                    std::size_t i = 0;
+                    do {
+                        if (bits[bit_ind + i++] != ~u64(0)) {
+                            return false;
+                        }
+                        remaining -= 64;
+                    } while (remaining > 0);
+
+                    // Clear the bits
+                    remaining = count;
+                    i = 0;
+                    do {
+                        bits[bit_ind + i] = 0;
+                        ClearBit(depth - 1, bit_ind + i);
+                        i++;
+                        remaining -= 64;
+                    } while (remaining > 0);
+                }
+
+                num_bits -= count;
+                return true;
+            }
+
+        private:
+            constexpr void SetBit(s32 depth, std::size_t offset) {
+                while (depth >= 0) {
+                    const std::size_t ind{offset / 64};
+                    const std::size_t which{offset % 64};
+                    const u64 mask{u64(1) << which};
+
+                    u64* bit{std::addressof(bit_storages[depth][ind])};
+                    const u64 v{*bit};
+                    ASSERT((v & mask) == 0);
+                    *bit = v | mask;
+                    if (v) {
+                        break;
+                    }
+                    offset = ind;
+                    depth--;
+                }
+            }
+
+            constexpr void ClearBit(s32 depth, std::size_t offset) {
+                while (depth >= 0) {
+                    const std::size_t ind{offset / 64};
+                    const std::size_t which{offset % 64};
+                    const u64 mask{u64(1) << which};
+
+                    u64* bit{std::addressof(bit_storages[depth][ind])};
+                    u64 v{*bit};
+                    ASSERT((v & mask) != 0);
+                    v &= ~mask;
+                    *bit = v;
+                    if (v) {
+                        break;
+                    }
+                    offset = ind;
+                    depth--;
+                }
+            }
+
+        private:
+            static constexpr s32 GetRequiredDepth(std::size_t region_size) {
+                s32 depth = 0;
+                while (true) {
+                    region_size /= 64;
+                    depth++;
+                    if (region_size == 0) {
+                        return depth;
+                    }
+                }
+            }
+
+        public:
+            static constexpr std::size_t CalculateMetadataOverheadSize(std::size_t region_size) {
+                std::size_t overhead_bits = 0;
+                for (s32 depth{GetRequiredDepth(region_size) - 1}; depth >= 0; depth--) {
+                    region_size = Common::AlignUp(region_size, 64) / 64;
+                    overhead_bits += region_size;
+                }
+                return overhead_bits * sizeof(u64);
+            }
+        };
+
+    private:
+        Bitmap bitmap;
+        VAddr heap_address{};
+        uintptr_t end_offset{};
+        std::size_t block_shift{};
+        std::size_t next_block_shift{};
+
+    public:
+        constexpr Block() = default;
+
+        constexpr std::size_t GetShift() const {
+            return block_shift;
+        }
+        constexpr std::size_t GetNextShift() const {
+            return next_block_shift;
+        }
+        constexpr std::size_t GetSize() const {
+            return std::size_t(1) << GetShift();
+        }
+        constexpr std::size_t GetNumPages() const {
+            return GetSize() / PageSize;
+        }
+        constexpr std::size_t GetNumFreeBlocks() const {
+            return bitmap.GetNumBits();
+        }
+        constexpr std::size_t GetNumFreePages() const {
+            return GetNumFreeBlocks() * GetNumPages();
+        }
+
+        constexpr u64* Initialize(VAddr addr, std::size_t size, std::size_t bs, std::size_t nbs,
+                                  u64* bit_storage) {
+            // Set shifts
+            block_shift = bs;
+            next_block_shift = nbs;
+
+            // Align up the address
+            VAddr end{addr + size};
+            const std::size_t align{(next_block_shift != 0) ? (u64(1) << next_block_shift)
+                                                            : (u64(1) << block_shift)};
+            addr = Common::AlignDown((addr), align);
+            end = Common::AlignUp((end), align);
+
+            heap_address = addr;
+            end_offset = (end - addr) / (u64(1) << block_shift);
+            return bitmap.Initialize(bit_storage, end_offset);
+        }
+
+        constexpr VAddr PushBlock(VAddr address) {
+            // Set the bit for the free block
+            std::size_t offset{(address - heap_address) >> GetShift()};
+            bitmap.SetBit(offset);
+
+            // If we have a next shift, try to clear the blocks below and return the address
+            if (GetNextShift()) {
+                const std::size_t diff{u64(1) << (GetNextShift() - GetShift())};
+                offset = Common::AlignDown(offset, diff);
+                if (bitmap.ClearRange(offset, diff)) {
+                    return heap_address + (offset << GetShift());
+                }
+            }
+
+            // We couldn't coalesce, or we're already as big as possible
+            return 0;
+        }
+
+        VAddr PopBlock() {
+            // Find a free block
+            const s64 soffset{bitmap.FindFreeBlock()};
+            if (soffset < 0) {
+                return 0;
+            }
+            const std::size_t offset{static_cast<std::size_t>(soffset)};
+
+            // Update our tracking and return it
+            bitmap.ClearBit(offset);
+            return heap_address + (offset << GetShift());
+        }
+
+    public:
+        static constexpr std::size_t CalculateMetadataOverheadSize(std::size_t region_size,
+                                                                   std::size_t cur_block_shift,
+                                                                   std::size_t next_block_shift) {
+            const std::size_t cur_block_size{(u64(1) << cur_block_shift)};
+            const std::size_t next_block_size{(u64(1) << next_block_shift)};
+            const std::size_t align{(next_block_shift != 0) ? next_block_size : cur_block_size};
+            return Bitmap::CalculateMetadataOverheadSize(
+                (align * 2 + Common::AlignUp(region_size, align)) / cur_block_size);
+        }
+    };
+
+public:
+    PageHeap() = default;
+
+    constexpr VAddr GetAddress() const {
+        return heap_address;
+    }
+    constexpr std::size_t GetSize() const {
+        return heap_size;
+    }
+    constexpr VAddr GetEndAddress() const {
+        return GetAddress() + GetSize();
+    }
+    constexpr std::size_t GetPageOffset(VAddr block) const {
+        return (block - GetAddress()) / PageSize;
+    }
+
+    void Initialize(VAddr heap_address, std::size_t heap_size, std::size_t metadata_size);
+    VAddr AllocateBlock(s32 index);
+    void Free(VAddr addr, std::size_t num_pages);
+
+    void UpdateUsedSize() {
+        used_size = heap_size - (GetNumFreePages() * PageSize);
+    }
+
+    static std::size_t CalculateMetadataOverheadSize(std::size_t region_size);
+
+private:
+    constexpr std::size_t GetNumFreePages() const {
+        std::size_t num_free{};
+
+        for (const auto& block : blocks) {
+            num_free += block.GetNumFreePages();
+        }
+
+        return num_free;
+    }
+
+    void FreeBlock(VAddr block, s32 index);
+
+    VAddr heap_address{};
+    std::size_t heap_size{};
+    std::size_t used_size{};
+    std::array<Block, NumMemoryBlockPageShifts> blocks{};
+    std::vector<u64> metadata;
+};
+
+} // namespace Kernel::Memory