L1T: address issue 36647 with int overrun

[cecilecaillol]

PR description:

Fixes undefined behavior from int overruns. Addressing issue #36647

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-37487/29190

• This PR adds an extra 16KB to repository

[cmsbuild]

A new Pull Request was created by @cecilecaillol for master.

It involves the following packages:

• L1Trigger/GlobalCaloTrigger (l1)

@epalencia, @cmsbuild, @cecilecaillol, @rekovic can you please review it and eventually sign? Thanks.
@Martin-Grunewald, @missirol this is something you requested to watch as well.
@perrotta, @dpiparo, @qliphy you are the release manager for this.

cms-bot commands are listed here

[cecilecaillol]

please test

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-a56297/23727/summary.html
COMMIT: bcca82f
CMSSW: CMSSW_12_4_X_2022-04-07-1100/slc7_amd64_gcc10
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week1/cms-sw/cmssw/37487/23727/install.sh to create a dev area with all the needed externals and cmssw changes.

Comparison Summary

Summary:

• No significant changes to the logs found
• Reco comparison results: 0 differences found in the comparisons
• DQMHistoTests: Total files compared: 48
• DQMHistoTests: Total histograms compared: 3593039
• DQMHistoTests: Total failures: 2
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 3593015
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 47 files compared)
• Checked 200 log files, 45 edm output root files, 48 DQM output files
• TriggerResults: no differences found

[cecilecaillol]

+l1

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @perrotta, @dpiparo, @qliphy (and backports should be raised in the release meeting by the corresponding L2)

[perrotta]

@cecilecaillol did you check that this PR actually fixes the issue?
At a first sight, it just moves it to etComponentSum...

[cecilecaillol]

@perrotta How can I check if the PR fixes the issue?

[perrotta]

@perrotta How can I check if the PR fixes the issue?

@Dr15Jones opened #36647, perhaps he can provide some suggestion

[perrotta]

hold
(while waiting for the confirmation that the fix as implemented is effective)

[cmsbuild]

Pull request has been put on hold by @perrotta
They need to issue an unhold command to remove the hold state or L1 can unhold it for all

[perrotta]

@cecilecaillol any suggestion about how to proceed here?

[cecilecaillol]

@perrotta I can make etComponentSum a long int too, but I still dont know how to check if this will actually solve the problem.

[perrotta]

@Dr15Jones @smuzaffar @makortel : can any of you suggest us how to run UBSAN on the proposed fix, so that it can be verified its effectiveness (or further adjust, if needed) before merging it?

[smuzaffar]

please test for CMSSW_12_4_UBSAN_X
this is run relval for UBSAN and then one can check workflow 1000.0/step2 if this has fixed the ubsan issues

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-a56297/24219/summary.html
COMMIT: bcca82f
CMSSW: CMSSW_12_4_UBSAN_X_2022-04-25-1100/slc7_amd64_gcc11
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/37487/24219/install.sh to create a dev area with all the needed externals and cmssw changes.

[smuzaffar]

so looks like it does fix the L1Trigger/GlobalCaloTrigger/src/L1GctJetFinderBase.cc:580:19: runtime error: signed integer overflow: 268432611 * 157 cannot be represented in type 'int' issue. Though there are only issues e.g. DQM/L1TMonitor/src/L1TGMT.cc:205:40: runtime error: signed integer overflow: 15991514 * 3564 cannot be represented in type 'int' but workfow 1000.0/step2 is now clean of #36647

[Dr15Jones]

@perrotta as Shahzad has shown, if you create a build area from one of the UBSAN IBs

scram project CMSSW_12_4_UBSAN_X_2022-04-25-1100

then build and run the RelVal workflow

runTheMatrix -l 1000.0

you should get the full UBSAN report.

[perrotta]

Thank you @smuzaffar and @Dr15Jones!

So, this fixes the issue. My concern about etComponentSum also extending above the int size was then wrong, at least for this particular example.

What I overlooked is that at line

cmssw/L1Trigger/GlobalCaloTrigger/src/L1GctJetFinderBase.cc

Line 588 in bcca82f

etComponentSum = ((rotatedValue0 + rotatedValue1) + halfInputLsb) >> bitsToShift;

there is a right shift of bitsToShift bits, which is evidently enough to reenter in the int size

[perrotta]

+1

• Thank you @cecilecaillol

[perrotta]

unhold

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will be automatically merged.