intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-06 02:18:05 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: regression caused by tbx fault mngr

Browse Source 
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>
This commit is contained in:
Jack Myers
2024-12-23 21:42:43 +00:00
committed by Compute-Runtime-Automation
parent b8157a2547
commit 7f9fadc314
36 changed files with 874 additions and 169 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
level_zero/core/source/device/device_imp.h
View File

@@ -1,5 +1,5 @@
/*
* Copyright (C) 2020-2024 Intel Corporation
* Copyright (C) 2020-2025 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
@@ -196,6 +196,6 @@ struct DeviceImp : public Device, NEO::NonCopyableOrMovableClass {
std::unique_ptr<DebugSession> debugSession;
};
void transferAndUnprotectMemoryWithHints(NEO::PageFaultManager *pageFaultHandler, void *allocPtr, NEO::PageFaultManager::PageFaultData &pageFaultData);
void transferAndUnprotectMemoryWithHints(NEO::CpuPageFaultManager *pageFaultHandler, void *allocPtr, NEO::CpuPageFaultManager::PageFaultData &pageFaultData);
} // namespace L0

14
level_zero/core/source/memory/cpu_page_fault_memory_manager.cpp
View File

@@ -1,5 +1,5 @@
/*
* Copyright (C) 2020-2024 Intel Corporation
* Copyright (C) 2020-2025 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
@@ -16,7 +16,7 @@
#include "level_zero/core/source/driver/driver_handle_imp.h"
namespace NEO {
void PageFaultManager::transferToCpu(void *ptr, size_t size, void *device) {
void CpuPageFaultManager::transferToCpu(void *ptr, size_t size, void *device) {
L0::DeviceImp *deviceImp = static_cast<L0::DeviceImp *>(device);
deviceImp->getNEODevice()->stopDirectSubmissionForCopyEngine();
@@ -29,7 +29,7 @@ void PageFaultManager::transferToCpu(void *ptr, size_t size, void *device) {
allocData->size, true);
UNRECOVERABLE_IF(ret);
}
void PageFaultManager::transferToGpu(void *ptr, void *device) {
void CpuPageFaultManager::transferToGpu(void *ptr, void *device) {
L0::DeviceImp *deviceImp = static_cast<L0::DeviceImp *>(device);
deviceImp->getNEODevice()->stopDirectSubmissionForCopyEngine();
@@ -44,7 +44,7 @@ void PageFaultManager::transferToGpu(void *ptr, void *device) {
this->evictMemoryAfterImplCopy(allocData->cpuAllocation, deviceImp->getNEODevice());
}
void PageFaultManager::allowCPUMemoryEviction(bool evict, void *ptr, PageFaultData &pageFaultData) {
void CpuPageFaultManager::allowCPUMemoryEviction(bool evict, void *ptr, PageFaultData &pageFaultData) {
L0::DeviceImp *deviceImp = static_cast<L0::DeviceImp *>(pageFaultData.cmdQ);
CommandStreamReceiver *csr = nullptr;
@@ -61,9 +61,9 @@ void PageFaultManager::allowCPUMemoryEviction(bool evict, void *ptr, PageFaultDa
} // namespace NEO
namespace L0 {
void transferAndUnprotectMemoryWithHints(NEO::PageFaultManager *pageFaultHandler, void *allocPtr, NEO::PageFaultManager::PageFaultData &pageFaultData) {
void transferAndUnprotectMemoryWithHints(NEO::CpuPageFaultManager *pageFaultHandler, void *allocPtr, NEO::CpuPageFaultManager::PageFaultData &pageFaultData) {
bool migration = true;
if (pageFaultData.domain == NEO::PageFaultManager::AllocationDomain::gpu) {
if (pageFaultData.domain == NEO::CpuPageFaultManager::AllocationDomain::gpu) {
L0::DeviceImp *deviceImp = static_cast<L0::DeviceImp *>(pageFaultData.cmdQ);
NEO::SvmAllocationData *allocData = deviceImp->getDriverHandle()->getSvmAllocsManager()->getSVMAlloc(allocPtr);
@@ -87,7 +87,7 @@ void transferAndUnprotectMemoryWithHints(NEO::PageFaultManager *pageFaultHandler
}
}
if (migration) {
pageFaultData.domain = NEO::PageFaultManager::AllocationDomain::cpu;
pageFaultData.domain = NEO::CpuPageFaultManager::AllocationDomain::cpu;
}
pageFaultHandler->allowCPUMemoryAccess(allocPtr, pageFaultData.size);
}

28
level_zero/core/test/unit_tests/sources/cmdlist/test_cmdlist_1.cpp
View File

@@ -1,5 +1,5 @@
/*
* Copyright (C) 2020-2024 Intel Corporation
* Copyright (C) 2020-2025 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
@@ -545,9 +545,9 @@ TEST_F(CommandListMemAdvisePageFault, givenValidPtrAndPageFaultHandlerAndGpuDoma
EXPECT_EQ(handlerWithHints, reinterpret_cast<void *>(mockPageFaultManager->gpuDomainHandler));
NEO::PageFaultManager::PageFaultData pageData;
NEO::CpuPageFaultManager::PageFaultData pageData;
pageData.cmdQ = deviceImp;
pageData.domain = NEO::PageFaultManager::AllocationDomain::gpu;
pageData.domain = NEO::CpuPageFaultManager::AllocationDomain::gpu;
mockPageFaultManager->gpuDomainHandler(mockPageFaultManager, ptr, pageData);
flags = deviceImp->memAdviseSharedAllocations[allocData];
EXPECT_EQ(1, flags.cpuMigrationBlocked);
@@ -586,9 +586,9 @@ TEST_F(CommandListMemAdvisePageFault, givenValidPtrAndPageFaultHandlerAndGpuDoma
EXPECT_EQ(handlerWithHints, reinterpret_cast<void *>(mockPageFaultManager->gpuDomainHandler));
NEO::PageFaultManager::PageFaultData pageData;
NEO::CpuPageFaultManager::PageFaultData pageData;
pageData.cmdQ = deviceImp;
pageData.domain = NEO::PageFaultManager::AllocationDomain::gpu;
pageData.domain = NEO::CpuPageFaultManager::AllocationDomain::gpu;
pageData.unifiedMemoryManager = device->getDriverHandle()->getSvmAllocsManager();
EXPECT_EQ(0u, device->getDriverHandle()->getSvmAllocsManager()->nonGpuDomainAllocs.size());
mockPageFaultManager->gpuDomainHandler(mockPageFaultManager, ptr, pageData);
@@ -661,9 +661,9 @@ TEST_F(CommandListMemAdvisePageFault, givenValidPtrAndPageFaultHandlerAndGpuDoma
testing::internal::CaptureStdout(); // start capturing
NEO::PageFaultManager::PageFaultData pageData;
NEO::CpuPageFaultManager::PageFaultData pageData;
pageData.cmdQ = deviceImp;
pageData.domain = NEO::PageFaultManager::AllocationDomain::gpu;
pageData.domain = NEO::CpuPageFaultManager::AllocationDomain::gpu;
pageData.unifiedMemoryManager = device->getDriverHandle()->getSvmAllocsManager();
mockPageFaultManager->gpuDomainHandler(mockPageFaultManager, ptr, pageData);
flags = deviceImp->memAdviseSharedAllocations[allocData];
@@ -715,9 +715,9 @@ TEST_F(CommandListMemAdvisePageFault, givenValidPtrAndPageFaultHandlerAndGpuDoma
EXPECT_EQ(handlerWithHints, reinterpret_cast<void *>(mockPageFaultManager->gpuDomainHandler));
NEO::PageFaultManager::PageFaultData pageData;
NEO::CpuPageFaultManager::PageFaultData pageData;
pageData.cmdQ = deviceImp;
pageData.domain = NEO::PageFaultManager::AllocationDomain::gpu;
pageData.domain = NEO::CpuPageFaultManager::AllocationDomain::gpu;
pageData.unifiedMemoryManager = device->getDriverHandle()->getSvmAllocsManager();
mockPageFaultManager->gpuDomainHandler(mockPageFaultManager, ptr, pageData);
flags = deviceImp->memAdviseSharedAllocations[allocData];
@@ -762,9 +762,9 @@ TEST_F(CommandListMemAdvisePageFault, givenValidPtrAndPageFaultHandlerAndGpuDoma
EXPECT_EQ(handlerWithHints, reinterpret_cast<void *>(mockPageFaultManager->gpuDomainHandler));
NEO::PageFaultManager::PageFaultData pageData;
NEO::CpuPageFaultManager::PageFaultData pageData;
pageData.cmdQ = deviceImp;
pageData.domain = NEO::PageFaultManager::AllocationDomain::cpu;
pageData.domain = NEO::CpuPageFaultManager::AllocationDomain::cpu;
pageData.unifiedMemoryManager = device->getDriverHandle()->getSvmAllocsManager();
mockPageFaultManager->gpuDomainHandler(mockPageFaultManager, ptr, pageData);
flags = deviceImp->memAdviseSharedAllocations[allocData];
@@ -809,9 +809,9 @@ TEST_F(CommandListMemAdvisePageFault, givenInvalidPtrAndPageFaultHandlerAndGpuDo
EXPECT_EQ(handlerWithHints, reinterpret_cast<void *>(mockPageFaultManager->gpuDomainHandler));
NEO::PageFaultManager::PageFaultData pageData;
NEO::CpuPageFaultManager::PageFaultData pageData;
pageData.cmdQ = deviceImp;
pageData.domain = NEO::PageFaultManager::AllocationDomain::gpu;
pageData.domain = NEO::CpuPageFaultManager::AllocationDomain::gpu;
pageData.unifiedMemoryManager = device->getDriverHandle()->getSvmAllocsManager();
void *alloc = reinterpret_cast<void *>(0x1);
mockPageFaultManager->gpuDomainHandler(mockPageFaultManager, alloc, pageData);
@@ -838,7 +838,7 @@ TEST_F(CommandListMemAdvisePageFault, givenUnifiedMemoryAllocWhenAllowCPUMemoryE
L0::DeviceImp *deviceImp = static_cast<L0::DeviceImp *>((L0::Device::fromHandle(device)));
NEO::PageFaultManager::PageFaultData pageData;
NEO::CpuPageFaultManager::PageFaultData pageData;
pageData.cmdQ = deviceImp;
mockPageFaultManager->baseAllowCPUMemoryEviction(true, ptr, pageData);