Segfault at the end of reconstruction

[ngduchai]

I got hit by a Segmentation Fault when sirt_stream for 354 iterations. More details below:

• Testing environment: Ubuntu 24.04.1 LTS (Linux dev 6.8.0-51-generic #52-Ubuntu SMP PREEMPT_DYNAMIC Thu Dec 5 13:32:09 UTC 2024 aarch64 aarch64 aarch64 GNU/Linux) Virtual Machine with 4 vCPU 8GB Memory hypervised by VMWare Fusion Professional Version 13.6.2 (24409261). Host machine: Macbook Pro M3 Max Sonoma 14.6
• Execution command: ./build/bin/sirt_stream --write-freq 4 --dest-host 0.0.0.0 --dest-port 50010 --window-iter 1 --window-step 4 --window-length 4 -t 2 -c 1427 --pub-addr tcp://*:52000
• gdb output:
  [Switching to Thread 0xfffff140ef20 (LWP 262781)] 0x0000aaaaaaad53f0 in SIRTReconSpace::UpdateReconReplica ( this=0xaaaaaabc5850, simdata=0.0129386261, ray=-0.0399322398, curr_slice=0, indi=0xffffaca00010, leng2=0xffffafe00010, leng=0xffffb3200010, len=87) at /home/ndhai/diaspora/aps-mini-apps-original/src/sirt/sirt.cc:76 76 slice[index] += leng[i]*upd;

• Stacktrace:
  #0 0x0000aaaaaaad53f0 in SIRTReconSpace::UpdateReconReplica (this=0xaaaaaabc5850, simdata=0.0129386261, ray=-0.0399322398, curr_slice=0, indi=0xffffaca00010, leng2=0xffffafe00010, leng=0xffffb3200010, len=87) at /home/ndhai/diaspora/aps-mini-apps-original/src/sirt/sirt.cc:76
#1 0x0000aaaaaaad5bf4 in SIRTReconSpace::Reduce ( this=0xaaaaaabc5850, input=...) at /home/ndhai/diaspora/aps-mini-apps-original/src/sirt/sirt.cc:193
#2 0x0000aaaaaaac0ebc in AReductionSpaceBase<SIRTReconSpace, float>::Process (this=0xaaaaaabc5850, input=...) at /home/ndhai/diaspora/aps-mini-apps-original/include/tracelib/reduction_space_a.h:16 `` #3 0x0000aaaaaaabe65c in DISPEngineReduction<SIRTReconSpace, float>::ReductionWrapper (this=0xaaaaaac1c800, reduction_space=..., input_data=..., req_units=@0xffffffffd3dc: 2560) at /home/ndhai/diaspora/aps-mini-apps-original/include/tracelib/disp_engine_reduction.h:42
#4 0x0000aaaaaaac8e48 in std::__invoke_impl<void, void (DISPEngineReduction<SIRTReconSpace, float>::*)(AReductionSpaceBase<SIRTReconSpace, float>&, ADataRegion<float>&, int&), DISPEngineReduction<SIRTReconSpace, float>*, std::reference_wrapper<AReductionSpaceBase<SIRTReconSpace, float> >, std::reference_wrapper<ADataRegion<float> >, std::reference_wrapper<int> > ( __f=@0xffff9c000d78: &virtual DISPEngineReduction<SIRTReconSpace, float>::ReductionWrapper(AReductionSpaceBase<SIRTReconSpace, float>&, ADataRegion<float>&, int&), __t=@0xffff9c000d70: 0xaaaaaac1c800) at /usr/include/c++/13/bits/invoke.h:74
#5 0x0000aaaaaaac8ab0 in std::__invoke<void (DISPEngineReduction<SIRTReconSpace, float>::*)(AReductionSpaceBase<SIRTReconSpace, float>&, ADataRegion<float>&, int&), DISPEngineReduction<SIRTReconSpace, float>*, std::reference_wrapper<AReductionSpaceBase<SIRTReconSpace, float> >, std::reference_wrapper<ADataRegion<float> >, std::reference_wrapper<int> > ( __fn=@0xffff9c000d78: &virtual DISPEngineReduction<SIRTReconSpace, float>::ReductionWrapper(AReductionSpaceBase<SIRTReconSpace, float>&, ADataRegion<float>&, int&)) at /usr/include/c++/13/bits/invoke.h:96
#6 0x0000aaaaaaac8788 in std::thread::_Invoker<std::tuple<void (DISPEngineReduction<SIRTReconSpace, float>::*)(AReductionSpaceBase<SIRTReconSpace, float>&, ADataRegion<float>&,--Type <RET> fo --Type <RET> for more, q to quit, c to continue without paging-- int&), DISPEngineReduction<SIRTReconSpace, float>*, std::reference_wrapper<AReductionSpaceBase<SIRTReconSpace, float> >, std::reference_wrapper<ADataRegion<float> >, std::reference_wrapper<int> > >::_M_invoke<0ul, 1ul, 2ul, 3ul, 4ul> (this=0xffff9c000d58) at /usr/include/c++/13/bits/std_thread.h:292
#7 0x0000aaaaaaac8624 in std::thread::_Invoker<std::tuple<void (DISPEngineReduction<SIRTReconSpace, float>::*)(AReductionSpaceBase<SIRTReconSpace, float>&, ADataRegion<float>&, int&), DISPEngineReduction<SIRTReconSpace, float>*, std::reference_wrapper<AReductionSpaceBase<SIRTReconSpace, float> >, std::reference_wrapper<ADataRegion<float> >, std::reference_wrapper<int> > >::operator() ( this=0xffff9c000d58) at /usr/include/c++/13/bits/std_thread.h:299
#8 0x0000aaaaaaac85c4 in std::thread::_State_impl<std::thread::_Invoker<std::tuple<void (DISPEngineReduction<SIRTReconSpace, float>::*)(AReductionSpaceBase<SIRTReconSpace, float>&, ADataRegion<float>&, int&), DISPEngineReduction<SIRTReconSpace, float>*, std::reference_wrapper<AReductionSpaceBase<SIRTReconSpace, float> >, std::reference_wrapper<ADataRegion<float> >, std::reference_wrapper<int> > > >::_M_run (this=0xffff9c000d50) at /usr/include/c++/13/bits/std_thread.h:244
#9 0x0000fffff6ae1ae0 in ?? ()
from /home/ndhai/spack/var/spack/environments/APS_GDB/.spack-env/view/lib/libstdc++.so.6
#10 0x0000fffff68c597c [PAC] in start_thread (arg=0xfffff7fb9760) at ./nptl/pthread_create.c:447
#11 0x0000fffff692ba4c in thread_start () at ../sysdeps/unix/sysv/linux/aarch64/clone3.S:76

@tekinbicer @GueroudjiAmal

[ngduchai]

I also got segfault at the same line in the Mofka integration.

[GueroudjiAmal]

Thanks @ngduchai , it's great! that you can reproduce it with both code!

[tekinbicer]

We figured one of the precalculated voxel addresses (out of 1,5K2K2K) accesses out of allocated memory. We added an 'if' statement to address this, which is not ideal. The algorithm that performs voxel index calculations needs to be revised.

aps-mini-apps/src/sirt/sirt.cc

Lines 89 to 90 in 68206c6

	slice[index] += leng[i]*upd;
	slice[index+1] += leng[i];