Ryujinx/Ryujinx.HLE/HOS/Kernel/Memory/KMemoryRegionManager.cs
riperiperi 54ea2285f0
POWER - Performance Optimizations With Extensive Ramifications (#2286)
* Refactoring of KMemoryManager class

* Replace some trivial uses of DRAM address with VA

* Get rid of GetDramAddressFromVa

* Abstracting more operations on derived page table class

* Run auto-format on KPageTableBase

* Managed to make TryConvertVaToPa private, few uses remains now

* Implement guest physical pages ref counting, remove manual freeing

* Make DoMmuOperation private and call new abstract methods only from the base class

* Pass pages count rather than size on Map/UnmapMemory

* Change memory managers to take host pointers

* Fix a guest memory leak and simplify KPageTable

* Expose new methods for host range query and mapping

* Some refactoring of MapPagesFromClientProcess to allow proper page ref counting and mapping without KPageLists

* Remove more uses of AddVaRangeToPageList, now only one remains (shared memory page checking)

* Add a SharedMemoryStorage class, will be useful for host mapping

* Sayonara AddVaRangeToPageList, you served us well

* Start to implement host memory mapping (WIP)

* Support memory tracking through host exception handling

* Fix some access violations from HLE service guest memory access and CPU

* Fix memory tracking

* Fix mapping list bugs, including a race and a error adding mapping ranges

* Simple page table for memory tracking

* Simple "volatile" region handle mode

* Update UBOs directly (experimental, rough)

* Fix the overlap check

* Only set non-modified buffers as volatile

* Fix some memory tracking issues

* Fix possible race in MapBufferFromClientProcess (block list updates were not locked)

* Write uniform update to memory immediately, only defer the buffer set.

* Fix some memory tracking issues

* Pass correct pages count on shared memory unmap

* Armeilleure Signal Handler v1 + Unix changes

Unix currently behaves like windows, rather than remapping physical

* Actually check if the host platform is unix

* Fix decommit on linux.

* Implement windows 10 placeholder shared memory, fix a buffer issue.

* Make PTC version something that will never match with master

* Remove testing variable for block count

* Add reference count for memory manager, fix dispose

Can still deadlock with OpenAL

* Add address validation, use page table for mapped check, add docs

Might clean up the page table traversing routines.

* Implement batched mapping/tracking.

* Move documentation, fix tests.

* Cleanup uniform buffer update stuff.

* Remove unnecessary assignment.

* Add unsafe host mapped memory switch

On by default. Would be good to turn this off for untrusted code (homebrew, exefs mods) and give the user the option to turn it on manually, though that requires some UI work.

* Remove C# exception handlers

They have issues due to current .NET limitations, so the meilleure one fully replaces them for now.

* Fix MapPhysicalMemory on the software MemoryManager.

* Null check for GetHostAddress, docs

* Add configuration for setting memory manager mode (not in UI yet)

* Add config to UI

* Fix type mismatch on Unix signal handler code emit

* Fix 6GB DRAM mode.

The size can be greater than `uint.MaxValue` when the DRAM is >4GB.

* Address some feedback.

* More detailed error if backing memory cannot be mapped.

* SetLastError on all OS functions for consistency

* Force pages dirty with UBO update instead of setting them directly.

Seems to be much faster across a few games. Need retesting.

* Rebase, configuration rework, fix mem tracking regression

* Fix race in FreePages

* Set memory managers null after decrementing ref count

* Remove readonly keyword, as this is now modified.

* Use a local variable for the signal handler rather than a register.

* Fix bug with buffer resize, and index/uniform buffer binding.

Should fix flickering in games.

* Add InvalidAccessHandler to MemoryTracking

Doesn't do anything yet

* Call invalid access handler on unmapped read/write.

Same rules as the regular memory manager.

* Make unsafe mapped memory its own MemoryManagerType

* Move FlushUboDirty into UpdateState.

* Buffer dirty cache, rather than ubo cache

Much cleaner, may be reusable for Inline2Memory updates.

* This doesn't return anything anymore.

* Add sigaction remove methods, correct a few function signatures.

* Return empty list of physical regions for size 0.

* Also on AddressSpaceManager

Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2021-05-24 22:52:44 +02:00

559 lines
No EOL
18 KiB
C#

using Ryujinx.Common;
using Ryujinx.HLE.HOS.Kernel.Common;
using System.Diagnostics;
namespace Ryujinx.HLE.HOS.Kernel.Memory
{
class KMemoryRegionManager
{
private static readonly int[] BlockOrders = new int[] { 12, 16, 21, 22, 25, 29, 30 };
public ulong Address { get; private set; }
public ulong EndAddr { get; private set; }
public ulong Size { get; private set; }
private int _blockOrdersCount;
private readonly KMemoryRegionBlock[] _blocks;
private readonly ushort[] _pageReferenceCounts;
public KMemoryRegionManager(ulong address, ulong size, ulong endAddr)
{
_blocks = new KMemoryRegionBlock[BlockOrders.Length];
Address = address;
Size = size;
EndAddr = endAddr;
_blockOrdersCount = BlockOrders.Length;
for (int blockIndex = 0; blockIndex < _blockOrdersCount; blockIndex++)
{
_blocks[blockIndex] = new KMemoryRegionBlock();
_blocks[blockIndex].Order = BlockOrders[blockIndex];
int nextOrder = blockIndex == _blockOrdersCount - 1 ? 0 : BlockOrders[blockIndex + 1];
_blocks[blockIndex].NextOrder = nextOrder;
int currBlockSize = 1 << BlockOrders[blockIndex];
int nextBlockSize = currBlockSize;
if (nextOrder != 0)
{
nextBlockSize = 1 << nextOrder;
}
ulong startAligned = BitUtils.AlignDown(address, nextBlockSize);
ulong endAddrAligned = BitUtils.AlignDown(endAddr, currBlockSize);
ulong sizeInBlocksTruncated = (endAddrAligned - startAligned) >> BlockOrders[blockIndex];
ulong endAddrRounded = BitUtils.AlignUp(address + size, nextBlockSize);
ulong sizeInBlocksRounded = (endAddrRounded - startAligned) >> BlockOrders[blockIndex];
_blocks[blockIndex].StartAligned = startAligned;
_blocks[blockIndex].SizeInBlocksTruncated = sizeInBlocksTruncated;
_blocks[blockIndex].SizeInBlocksRounded = sizeInBlocksRounded;
ulong currSizeInBlocks = sizeInBlocksRounded;
int maxLevel = 0;
do
{
maxLevel++;
}
while ((currSizeInBlocks /= 64) != 0);
_blocks[blockIndex].MaxLevel = maxLevel;
_blocks[blockIndex].Masks = new long[maxLevel][];
currSizeInBlocks = sizeInBlocksRounded;
for (int level = maxLevel - 1; level >= 0; level--)
{
currSizeInBlocks = (currSizeInBlocks + 63) / 64;
_blocks[blockIndex].Masks[level] = new long[currSizeInBlocks];
}
}
_pageReferenceCounts = new ushort[size / KPageTableBase.PageSize];
if (size != 0)
{
FreePages(address, size / KPageTableBase.PageSize);
}
}
public KernelResult AllocatePages(ulong pagesCount, bool backwards, out KPageList pageList)
{
lock (_blocks)
{
KernelResult result = AllocatePagesImpl(pagesCount, backwards, out pageList);
if (result == KernelResult.Success)
{
foreach (var node in pageList)
{
IncrementPagesReferenceCount(node.Address, node.PagesCount);
}
}
return result;
}
}
public ulong AllocatePagesContiguous(KernelContext context, ulong pagesCount, bool backwards)
{
lock (_blocks)
{
ulong address = AllocatePagesContiguousImpl(pagesCount, backwards);
if (address != 0)
{
IncrementPagesReferenceCount(address, pagesCount);
context.Memory.Commit(address - DramMemoryMap.DramBase, pagesCount * KPageTableBase.PageSize);
}
return address;
}
}
private KernelResult AllocatePagesImpl(ulong pagesCount, bool backwards, out KPageList pageList)
{
pageList = new KPageList();
if (_blockOrdersCount > 0)
{
if (GetFreePagesImpl() < pagesCount)
{
return KernelResult.OutOfMemory;
}
}
else if (pagesCount != 0)
{
return KernelResult.OutOfMemory;
}
for (int blockIndex = _blockOrdersCount - 1; blockIndex >= 0; blockIndex--)
{
KMemoryRegionBlock block = _blocks[blockIndex];
ulong bestFitBlockSize = 1UL << block.Order;
ulong blockPagesCount = bestFitBlockSize / KPageTableBase.PageSize;
// Check if this is the best fit for this page size.
// If so, try allocating as much requested pages as possible.
while (blockPagesCount <= pagesCount)
{
ulong address = AllocatePagesForOrder(blockIndex, backwards, bestFitBlockSize);
// The address being zero means that no free space was found on that order,
// just give up and try with the next one.
if (address == 0)
{
break;
}
// Add new allocated page(s) to the pages list.
// If an error occurs, then free all allocated pages and fail.
KernelResult result = pageList.AddRange(address, blockPagesCount);
if (result != KernelResult.Success)
{
FreePages(address, blockPagesCount);
foreach (KPageNode pageNode in pageList)
{
FreePages(pageNode.Address, pageNode.PagesCount);
}
return result;
}
pagesCount -= blockPagesCount;
}
}
// Success case, all requested pages were allocated successfully.
if (pagesCount == 0)
{
return KernelResult.Success;
}
// Error case, free allocated pages and return out of memory.
foreach (KPageNode pageNode in pageList)
{
FreePages(pageNode.Address, pageNode.PagesCount);
}
pageList = null;
return KernelResult.OutOfMemory;
}
private ulong AllocatePagesContiguousImpl(ulong pagesCount, bool backwards)
{
if (pagesCount == 0 || _blocks.Length < 1)
{
return 0;
}
int blockIndex = 0;
while ((1UL << _blocks[blockIndex].Order) / KPageTableBase.PageSize < pagesCount)
{
if (++blockIndex >= _blocks.Length)
{
return 0;
}
}
ulong tightestFitBlockSize = 1UL << _blocks[blockIndex].Order;
ulong address = AllocatePagesForOrder(blockIndex, backwards, tightestFitBlockSize);
ulong requiredSize = pagesCount * KPageTableBase.PageSize;
if (address != 0 && tightestFitBlockSize > requiredSize)
{
FreePages(address + requiredSize, (tightestFitBlockSize - requiredSize) / KPageTableBase.PageSize);
}
return address;
}
private ulong AllocatePagesForOrder(int blockIndex, bool backwards, ulong bestFitBlockSize)
{
ulong address = 0;
KMemoryRegionBlock block = null;
for (int currBlockIndex = blockIndex;
currBlockIndex < _blockOrdersCount && address == 0;
currBlockIndex++)
{
block = _blocks[currBlockIndex];
int index = 0;
bool zeroMask = false;
for (int level = 0; level < block.MaxLevel; level++)
{
long mask = block.Masks[level][index];
if (mask == 0)
{
zeroMask = true;
break;
}
if (backwards)
{
index = (index * 64 + 63) - BitUtils.CountLeadingZeros64(mask);
}
else
{
index = index * 64 + BitUtils.CountLeadingZeros64(BitUtils.ReverseBits64(mask));
}
}
if (block.SizeInBlocksTruncated <= (ulong)index || zeroMask)
{
continue;
}
block.FreeCount--;
int tempIdx = index;
for (int level = block.MaxLevel - 1; level >= 0; level--, tempIdx /= 64)
{
block.Masks[level][tempIdx / 64] &= ~(1L << (tempIdx & 63));
if (block.Masks[level][tempIdx / 64] != 0)
{
break;
}
}
address = block.StartAligned + ((ulong)index << block.Order);
}
for (int currBlockIndex = blockIndex;
currBlockIndex < _blockOrdersCount && address == 0;
currBlockIndex++)
{
block = _blocks[currBlockIndex];
int index = 0;
bool zeroMask = false;
for (int level = 0; level < block.MaxLevel; level++)
{
long mask = block.Masks[level][index];
if (mask == 0)
{
zeroMask = true;
break;
}
if (backwards)
{
index = index * 64 + BitUtils.CountLeadingZeros64(BitUtils.ReverseBits64(mask));
}
else
{
index = (index * 64 + 63) - BitUtils.CountLeadingZeros64(mask);
}
}
if (block.SizeInBlocksTruncated <= (ulong)index || zeroMask)
{
continue;
}
block.FreeCount--;
int tempIdx = index;
for (int level = block.MaxLevel - 1; level >= 0; level--, tempIdx /= 64)
{
block.Masks[level][tempIdx / 64] &= ~(1L << (tempIdx & 63));
if (block.Masks[level][tempIdx / 64] != 0)
{
break;
}
}
address = block.StartAligned + ((ulong)index << block.Order);
}
if (address != 0)
{
// If we are using a larger order than best fit, then we should
// split it into smaller blocks.
ulong firstFreeBlockSize = 1UL << block.Order;
if (firstFreeBlockSize > bestFitBlockSize)
{
FreePages(address + bestFitBlockSize, (firstFreeBlockSize - bestFitBlockSize) / KPageTableBase.PageSize);
}
}
return address;
}
private void FreePages(ulong address, ulong pagesCount)
{
lock (_blocks)
{
ulong endAddr = address + pagesCount * KPageTableBase.PageSize;
int blockIndex = _blockOrdersCount - 1;
ulong addressRounded = 0;
ulong endAddrTruncated = 0;
for (; blockIndex >= 0; blockIndex--)
{
KMemoryRegionBlock allocInfo = _blocks[blockIndex];
int blockSize = 1 << allocInfo.Order;
addressRounded = BitUtils.AlignUp (address, blockSize);
endAddrTruncated = BitUtils.AlignDown(endAddr, blockSize);
if (addressRounded < endAddrTruncated)
{
break;
}
}
void FreeRegion(ulong currAddress)
{
for (int currBlockIndex = blockIndex;
currBlockIndex < _blockOrdersCount && currAddress != 0;
currBlockIndex++)
{
KMemoryRegionBlock block = _blocks[currBlockIndex];
block.FreeCount++;
ulong freedBlocks = (currAddress - block.StartAligned) >> block.Order;
int index = (int)freedBlocks;
for (int level = block.MaxLevel - 1; level >= 0; level--, index /= 64)
{
long mask = block.Masks[level][index / 64];
block.Masks[level][index / 64] = mask | (1L << (index & 63));
if (mask != 0)
{
break;
}
}
int blockSizeDelta = 1 << (block.NextOrder - block.Order);
int freedBlocksTruncated = BitUtils.AlignDown((int)freedBlocks, blockSizeDelta);
if (!block.TryCoalesce(freedBlocksTruncated, blockSizeDelta))
{
break;
}
currAddress = block.StartAligned + ((ulong)freedBlocksTruncated << block.Order);
}
}
// Free inside aligned region.
ulong baseAddress = addressRounded;
while (baseAddress < endAddrTruncated)
{
ulong blockSize = 1UL << _blocks[blockIndex].Order;
FreeRegion(baseAddress);
baseAddress += blockSize;
}
int nextBlockIndex = blockIndex - 1;
// Free region between Address and aligned region start.
baseAddress = addressRounded;
for (blockIndex = nextBlockIndex; blockIndex >= 0; blockIndex--)
{
ulong blockSize = 1UL << _blocks[blockIndex].Order;
while (baseAddress - blockSize >= address)
{
baseAddress -= blockSize;
FreeRegion(baseAddress);
}
}
// Free region between aligned region end and End Address.
baseAddress = endAddrTruncated;
for (blockIndex = nextBlockIndex; blockIndex >= 0; blockIndex--)
{
ulong blockSize = 1UL << _blocks[blockIndex].Order;
while (baseAddress + blockSize <= endAddr)
{
FreeRegion(baseAddress);
baseAddress += blockSize;
}
}
}
}
public ulong GetFreePages()
{
lock (_blocks)
{
return GetFreePagesImpl();
}
}
private ulong GetFreePagesImpl()
{
ulong availablePages = 0;
for (int blockIndex = 0; blockIndex < _blockOrdersCount; blockIndex++)
{
KMemoryRegionBlock block = _blocks[blockIndex];
ulong blockPagesCount = (1UL << block.Order) / KPageTableBase.PageSize;
availablePages += blockPagesCount * block.FreeCount;
}
return availablePages;
}
public void IncrementPagesReferenceCount(ulong address, ulong pagesCount)
{
ulong index = GetPageOffset(address);
ulong endIndex = index + pagesCount;
while (index < endIndex)
{
ushort referenceCount = ++_pageReferenceCounts[index];
Debug.Assert(referenceCount >= 1);
index++;
}
}
public void DecrementPagesReferenceCount(ulong address, ulong pagesCount)
{
ulong index = GetPageOffset(address);
ulong endIndex = index + pagesCount;
ulong freeBaseIndex = 0;
ulong freePagesCount = 0;
while (index < endIndex)
{
Debug.Assert(_pageReferenceCounts[index] > 0);
ushort referenceCount = --_pageReferenceCounts[index];
if (referenceCount == 0)
{
if (freePagesCount != 0)
{
freePagesCount++;
}
else
{
freeBaseIndex = index;
freePagesCount = 1;
}
}
else if (freePagesCount != 0)
{
FreePages(Address + freeBaseIndex * KPageTableBase.PageSize, freePagesCount);
freePagesCount = 0;
}
index++;
}
if (freePagesCount != 0)
{
FreePages(Address + freeBaseIndex * KPageTableBase.PageSize, freePagesCount);
}
}
public ulong GetPageOffset(ulong address)
{
return (address - Address) / KPageTableBase.PageSize;
}
public ulong GetPageOffsetFromEnd(ulong address)
{
return (EndAddr - address) / KPageTableBase.PageSize;
}
}
}