MuonHitsBanksDecoder makes out-of-bounds memory access

added Muon bug labels

assigned to @satta

added Crash label

TDCMap[TDCMap_length - 1] = ( ( range_link_TDC[TDCMap_length - 1] & 0x0F ) << 4 );

is executed with values

TDCMap_length       7
TDCMap size         6
range_link_TDC size 7

@gunther I have a look at it in the next days

any news @satta? I consider this somewhat urgent, a bug in the decoding should not run in production

Could you please point me to a data/MC file where the bug was noted?

I can reproduce it when checking out !4458 (merged) and running Moore/run gaudirun.py Moore/Hlt/RecoConf/tests/qmtest/hlt2_light_reco_pr_kf_without_UT.qmt

@satta: any progress on fixing this out-of-bounds access?

Sorry but not yet managed to have time. I Will have time beginning of next week when U2 muon performance estimation can have a stop.

changed milestone to %Developments targeting restart after May MD

Pinging also @ahennequ and @raaij since this might also be relevant for the HLT1 decoding.

The way this is done in the HLT1 decoding avoids this issue. There's no intermediate storage of the hitmaps and TDC values.

mentioned in merge request !4458 (merged)

@gunther I tried to compile Moore from stack using branch 2024-patches only for LHCb I checkout the branch gsl-span-fixes. It doesn't compile with many error messages The first are copied below. Could you send me instruction to compile the code? thanks

/home/asatta/Moore/stack/Rec/Tr/TrackKernel/include/TrackKernel/PrimaryVertexUtils.h: In function 'auto LHCb::bestPVIndex(const Vertices&, FTYPE, FTYPE, FTYPE, FTYPE, FTYPE)': /home/asatta/Moore/stack/Rec/Tr/TrackKernel/include/TrackKernel/PrimaryVertexUtils.h:64:61: error: 'PVIndexInvalid' is not a member of 'LHCb::Event::PV' 64 | : LHCb::Event::PV::PVIndex( LHCb::Event::PV::PVIndexInvalid ); | ^~~~~~~~~~~~~~ /home/asatta/Moore/stack/Rec/Tr/TrackKernel/include/TrackKernel/PrimaryVertexUtils.h: At global scope: /home/asatta/Moore/stack/Rec/Tr/TrackKernel/include/TrackKernel/PrimaryVertexUtils.h:92:24: error: 'PrimaryVertex' was not declared in this scope; did you mean 'LHCb::Event::PV::PrimaryVertex'? 92 | inline std::optional unbiasedBestPV( const LHCb::Event::PV::PrimaryVertexContainer& pvdata, | ^~~~~~~~~~~~~ | LHCb::Event::PV::PrimaryVertex In file included from /home/asatta/Moore/stack/LHCb/InstallArea/x86_64_v2-el9-gcc13-opt/include/Event/Particle.h:15, from ../Tr/TrackKernel/include/TrackKernel/PrimaryVertexUtils.h:14: /home/asatta/Moore/stack/LHCb/InstallArea/x86_64_v2-el9-gcc13-opt/include/Event/PrimaryVertices.h:135:9: note: 'LHCb::Event::PV::PrimaryVertex' declared here 135 | class PrimaryVertex final : public LHCb::VertexBase { | ^~~~~~~~~~~~~ /home/asatta/Moore/stack/Rec/Tr/TrackKernel/include/TrackKernel/PrimaryVertexUtils.h:92:37: error: template argument 1 is invalid 92 | inline std::optional unbiasedBestPV( const LHCb::Event::PV::PrimaryVertexContainer& pvdata, | ^ /home/asatta/Moore/stack/Rec/Tr/TrackKernel/include/TrackKernel/PrimaryVertexUtils.h:92:15: error: '' in namespace 'std' does not name a type 92 | inline std::optional unbiasedBestPV( const LHCb::Event::PV::PrimaryVertexContainer& pvdata, | ^~~~~~~~~~~~~~~~~~~~~~~ /home/asatta/Moore/stack/Rec/Tr/TrackKernel/include/TrackKernel/PrimaryVertexUtils.h:101:24: error: 'PrimaryVertex' was not declared in this scope; did you mean 'LHCb::Event::PV::PrimaryVertex'? 101 | inline std::optional unbiasedBestPV( const LHCb::Event::PV::PrimaryVertexContainer& pvdata,

the gsl-span-fixes branch has not been touched as it was waiting for resolution of this problmem. And since it has been quite a while, it has of course now fallen 'behind' the target branch. So you'll have to rebase your checked-out version of it on top of the 'tip' of the regular branch you'd otherwise use.

So I've just rebased the MR on top of 2024-patches, so for now it's up to date (until something gets merged in that branch obviously).

But you can also just take the very small changes to the 3 lines in Muon/MuonDAQ/include/MuonDAQ/MuonHitsBankDecoder.h and to the 2 lines in Muon/MuonDAQ/src/components/MuonRawInUpgradeToHits.cpp and just apply those explicitly to the stack which you're using... (as those 5 lines are sufficient to expose the problem)

@graven thanks, your rebase solved the compilation issue. It there a simple way to change output level using the "options" class?

@gunther @graven I commit the patch

mentioned in merge request !4527 (merged)

somehow issues are not closed anymore automatically when the corresponding MR is merged

this seems to be a very long-standing issue https://gitlab.com/gitlab-org/gitlab/-/issues/14289 when not targeting the default branch

closed

MuonHitsBanksDecoder makes out-of-bounds memory access

Child items 0

Activity