intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-22 10:17:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: prevent underflow in per thread data offset calculation

Browse Source 
Related-To: NEO-14719

Signed-off-by: Naklicki, Mateusz <mateusz.naklicki@intel.com>
This commit is contained in:
Naklicki, Mateusz
2025-06-16 12:44:39 +00:00
committed by Compute-Runtime-Automation
parent 6368f43df8
commit b462f990b6
4 changed files with 88 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
shared/source/compiler_interface/linker.cpp
View File

@@ -450,8 +450,7 @@ void Linker::patchInstructionsSegments(const std::vector<PatchableSegment> &inst
auto relocAddress = ptrOffset(segment.hostPointer, static_cast<uintptr_t>(relocation.offset));
if (relocation.type == LinkerInput::RelocationInfo::Type::perThreadPayloadOffset) {
uint32_t crossThreadDataSize = kernelDescriptors.at(segId)->kernelAttributes.crossThreadDataSize - kernelDescriptors.at(segId)->kernelAttributes.inlineDataPayloadSize;
*reinterpret_cast<uint32_t *>(relocAddress) = crossThreadDataSize;
*reinterpret_cast<uint32_t *>(relocAddress) = kernelDescriptors.at(segId)->getPerThreadDataOffset();
} else if (relocation.symbolName == implicitArgsRelocationSymbolName) {
pImplicitArgsRelocationAddresses[static_cast<uint32_t>(segId)].push_back(std::pair<void *, RelocationInfo::Type>(relocAddress, relocation.type));
} else if (relocation.symbolName.empty()) {
@@ -701,8 +700,7 @@ void Linker::resolveBuiltins(Device *pDevice, UnresolvedExternals &outUnresolved
auto kernelDescriptor = std::find_if(kernelDescriptors.begin(), kernelDescriptors.end(), [&kernelName](const KernelDescriptor *obj) { return obj->kernelMetadata.kernelName == kernelName; });
if (kernelDescriptor != std::end(kernelDescriptors)) {
uint64_t crossThreadDataSize = (*kernelDescriptor)->kernelAttributes.crossThreadDataSize - (*kernelDescriptor)->kernelAttributes.inlineDataPayloadSize;
symbol.gpuAddress = crossThreadDataSize;
symbol.gpuAddress = (*kernelDescriptor)->getPerThreadDataOffset();
auto relocAddress = ptrOffset(instructionsSegments[outUnresolvedExternals[vecIndex].instructionsSegmentId].hostPointer,
static_cast<uintptr_t>(outUnresolvedExternals[vecIndex].unresolvedRelocation.offset));