Files
compute-runtime/shared/source/page_fault_manager/tbx_page_fault_manager.h
Jack Myers 7f9fadc314 fix: regression caused by tbx fault mngr
Addresses regressions from the reverted merge
of the tbx fault manager for host memory.

Recursive locking of mutex caused deadlock.

To fix, separate tbx fault data from base
cpu fault data, allowing separate mutexes
for each, eliminating recursive locks on
the same mutex.

By separating, we also help ensure that tbx-related
changes don't affect the original cpu fault manager code
paths.

As an added safe guard preventing critical regressions
and avoiding another auto-revert, the tbx fault manager
is hidden behind a new debug flag which is disabled by default.

Related-To: NEO-12268
Signed-off-by: Jack Myers <jack.myers@intel.com>
2025-01-09 07:48:53 +01:00

45 lines
1.2 KiB
C++

/*
* Copyright (C) 2024-2025 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#pragma once
#include "shared/source/page_fault_manager/cpu_page_fault_manager.h"
namespace NEO {
class CommandStreamReceiver;
class GraphicsAllocation;
class TbxPageFaultManager : public virtual CpuPageFaultManager {
public:
struct PageFaultDataTbx {
size_t size;
bool hasBeenDownloaded = false;
GraphicsAllocation *gfxAllocation = nullptr;
uint32_t bank = 0;
CommandStreamReceiver *csr = nullptr;
};
static std::unique_ptr<TbxPageFaultManager> create();
TbxPageFaultManager() = default;
using CpuPageFaultManager::insertAllocation;
using CpuPageFaultManager::removeAllocation;
void insertAllocation(CommandStreamReceiver *csr, GraphicsAllocation *alloc, uint32_t bank, void *ptr, size_t size) override;
void removeAllocation(GraphicsAllocation *alloc) override;
using CpuPageFaultManager::checkFaultHandlerFromPageFaultManager;
bool verifyAndHandlePageFault(void *ptr, bool handlePageFault) override;
protected:
void handlePageFault(void *ptr, PageFaultDataTbx &faultData);
std::unordered_map<void *, PageFaultDataTbx> memoryDataTbx;
SpinLock mtxTbx;
};
} // namespace NEO