Fatal error: Crash requested by program

[Alfred-hhy]

Hi hidden
I forgot earlier that when I run poetry run make ring script=audit_fairness dataset=mnist_full_3party or cifar_alexnet_3party on my lab's server（104cpu,128G RAM） I always run into this problem but not with the ADULT dataset, I'm am not quite sure how to fix it.I just boost version installed is 1.83 (https://github.com/mkskeller/softspoken-implementation-cryptotools/blob/softspokenot/thirdparty/getBoost.py) due to your The link to the version that the previous code depended on is not working

Consistency check with type pc
Stopped timer 99 at 0.0424824 (3.45492 MB, 434 rounds)
Starting timer 98 at 0 (0 MB, 0 rounds) after 1.7e-06
Stopped timer 98 at 2.313e-06 (0 MB, 0 rounds)
Starting timer 99 at 0.0424824 (3.45492 MB, 434 rounds) after 2.2314e-05Stopped timer 99 at 0.0424842 (3.45492 MB, 434 rounds)
Starting timer 100 at 0 (0 MB, 0 rounds) after 1.046e-06
Creating 1024 audit trigger sample perturbations...
Performing inference on 1024 audit trigger sample perturbations...
overflowoverflow: : overflow: overoverflow: flowoveroverover2721overflow/: 28over31flow[//: overflowflow/: 16[19, 23//flow[: over[: 1625: 16flow[flow10/, : 1624over], [: 30/1610over20, over16flow/29, : [10/22/][[flow[/16
, 17[10:
1616]10, , []flow, over16flow1810//Fatal error: [1016Crash requested by program10: Fatal error: Crash requested by program, ]10]16][,

, 10]]

Fatal error: Fatal error: Crash requested by program
16/16, 1010]
]
Fatal error: Crash requested by program
Fatal error: Crash requested by program
=== Party 1

corrupted double-linked list
=== Party 2

Crash requested by programCrash requested by program

make: *** [Makefile:76: ring] Error 1

[hiddely]

This seems to be an issue with the MP-SPDZ program. Does it compile and run successfully when you run it locally?

[Alfred-hhy]

There are no errors in the compilation process, but as soon as it starts executing, an error occurs.

`Writing to Programs/Bytecode/audit_fairness-dataset__mnist_full_3party-audit_trigger_idx__1-emulate__True-debug__False-0.bc
Hash: c46e43a936674090be2f3e9a14a132b8ccc351c8579e495c3312021560466e28
Program requires at most:
431865 integer inputs from player 0
4622429364 bit triples
80980234 integer dabits
2364304396 integer triples
16282689 integer opens
20480 matrix multiplications (1x25 * 25x576)
16282688 integer bits
51200 matrix multiplications (1x500 * 500x64)
1024 matrix multiplications (1x800 * 800x500)
1024 matrix multiplications (1x500 * 500x10)
25663581 virtual machine rounds
Compilation finished, running program...
Running /home/ubuntu/arc/MP-SPDZ/Scripts/../replicated-ring-party.x 0 audit_fairness-dataset__mnist_full_3party-audit_trigger_idx__1-emulate__True-debug__False -pn 12984 -h localhost
Running /home/ubuntu/arc/MP-SPDZ/Scripts/../replicated-ring-party.x 1 audit_fairness-dataset__mnist_full_3party-audit_trigger_idx__1-emulate__True-debug__False -pn 12984 -h localhost
Running /home/ubuntu/arc/MP-SPDZ/Scripts/../replicated-ring-party.x 2 audit_fairness-dataset__mnist_full_3party-audit_trigger_idx__1-emulate__True-debug__False -pn 12984 -h localhost
Using statistical security parameter 40
Trying to run 64-bit computation
Starting timer 99 at 0 (0 MB, 0 rounds) after 1.3148e-05
loading 1 trigger mislabels...
loading 1 trigger samples...
loading model weights...
Using SGD
Consistency check with type pc
Stopped timer 99 at 0.0478629 (3.45492 MB, 434 rounds)
Starting timer 98 at 0 (0 MB, 0 rounds) after 1.643e-06
Stopped timer 98 at 2.159e-06 (0 MB, 0 rounds)
Starting timer 99 at 0.0478629 (3.45492 MB, 434 rounds) after 2.0927e-05Stopped timer 99 at 0.0478645 (3.45492 MB, 434 rounds)
Starting timer 100 at 0 (0 MB, 0 rounds) after 1.031e-06
Creating 1024 audit trigger sample perturbations...
Performing inference on 1024 audit trigger sample perturbations...
overoveroveroveroverflowflow: over: overoveroverflowoverflowover: flowflowoverflow: flow: : overoverflow: : overflow17: 22flowover29flow2819/[24flowflow/: /3016flow///31: [: 1616, /: : [flow1025[/]1627,
//21[[/: [10[[1616, [16/18, 1616, [10, ]]10
Fatal error: 10Crash requested by program/16: [16]10]]
Fatal error: , 10]

Fatal error: Crash requested by program
Fatal error: Crash requested by program
Fatal error: Crash requested by program
/home/ubuntu/arc/MP-SPDZ/Scripts/run-common.sh: line 90: 6309 Segmentation fault (core dumped) $my_prefix $SPDZROOT/$bin $i $params 2>&1
6310 Done | { if test "$BENCH"; then
if test $i = $front_player; then
tee -a $log;
else
cat >> $log;
fi;
else
if test $i = $front_player; then
tee $log;
else
cat > $log;
fi;
fi; }
/home/ubuntu/arc/MP-SPDZ/Scripts/run-common.sh: line 90: 6311 Segmentation fault (core dumped) $my_prefix $SPDZROOT/$bin $i $params 2>&1
6313 Done | { if test "$BENCH"; then
if test $i = $front_player; then
tee -a $log;
else
cat >> $log;
fi;
else
if test $i = $front_player; then
tee $log;
else
cat > $log;
fi;
fi; }
=== Party 1
Fatal error: Crash requested by program
Fatal error: Crash requested by program
Fatal error: Crash requested by program
=== Party 2
Crash requested by program
Crash requested by program
free(): corrupted unsorted chunks
make: *** [Makefile:76: ring] Error 1
`

[Alfred-hhy]

hi, hidde
When I run basicadult experiments with three 4 core 16g servers I always encounter 101 errors, when I run experiments with docker I always encounter 134 errors, have you encountered these problems before when you run it but the process of running it doesn't stop these errors automatically, Meanwhile I encounter the process of multiple error attempts, after which the task continues to run, which it should just check the status of whether the task is executed or not, so I did not stop the run because of this, but there are some problems when checking the results, I hope you can help me to see what is the problem if you want to

doe-suite-results-cloud.tar.gz

doe-suite-results-docker.tar.gz

[Alfred-hhy]

Command line: ./replicated-ring-party.x -OF yqji9iwdzrdm2ez1rooq -h 10.104.0.3 -pn 12300 -v 0 audit_robustness-audit_trigger_idx__0-batch_size__128-consistency_check__pc-dataset__adult_3p-debug__False-emulate__False-n_input_parties__3-n_samples__1-n_threads__4-sha3_approx_factor__1
Traceback (most recent call last):
File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/home/ubuntu/code/utils/python_utils/scripts/experiment_runner.py", line 618, in
cli()
File "/home/ubuntu/code/.venv/lib/python3.10/site-packages/click/core.py", line 1157, in call
return self.main(*args, **kwargs)
File "/home/ubuntu/code/.venv/lib/python3.10/site-packages/click/core.py", line 1078, in main
rv = self.invoke(ctx)
File "/home/ubuntu/code/.venv/lib/python3.10/site-packages/click/core.py", line 1434, in invoke
return ctx.invoke(self.callback, **ctx.params)
File "/home/ubuntu/code/.venv/lib/python3.10/site-packages/click/core.py", line 783, in invoke
return __callback(*args, **kwargs)
File "/home/ubuntu/code/utils/python_utils/scripts/experiment_runner.py", line 611, in cli
prove_commitment_opening(task_config=task_config,
File "/home/ubuntu/code/utils/python_utils/scripts/experiment_runner.py", line 178, in prove_commitment_opening
subprocess.run(
File "/usr/lib/python3.10/subprocess.py", line 526, in run
raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command 'target/release/gen_commitments_kzg --hosts /home/ubuntu/code/mpc-consistency/data/vpc --party 0 --player-input-binary-path /home/ubuntu/code/MP-SPDZ/Player-Data/Input-Binary-P0-0 --save --debug ' returned non-zero exit status 101.

[Alfred-hhy]

Command line: ./sy-rep-ring-party.x -OF hzruemf40ykylgcjtonp -h 453a4594a0e1 -pn 12300 -v 0 audit_fairness-audit_trigger_idx__0-batch_size__128-consistency_check__pc-dataset__adult_3p-debug__False-emulate__False-n_input_parties__3-n_samples__1-n_threads__20-sha3_approx_factor__1
Traceback (most recent call last):
File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/home/ubuntu/code/utils/python_utils/scripts/experiment_runner.py", line 618, in
cli()
File "/home/ubuntu/code/.venv/lib/python3.10/site-packages/click/core.py", line 1157, in call
return self.main(*args, **kwargs)
File "/home/ubuntu/code/.venv/lib/python3.10/site-packages/click/core.py", line 1078, in main
rv = self.invoke(ctx)
File "/home/ubuntu/code/.venv/lib/python3.10/site-packages/click/core.py", line 1434, in invoke
return ctx.invoke(self.callback, **ctx.params)
File "/home/ubuntu/code/.venv/lib/python3.10/site-packages/click/core.py", line 783, in invoke
return __callback(*args, **kwargs)
File "/home/ubuntu/code/utils/python_utils/scripts/experiment_runner.py", line 610, in cli
convert_shares(task_config=task_config, output_prefix=output_prefix)
File "/home/ubuntu/code/utils/python_utils/scripts/experiment_runner.py", line 437, in convert_shares
subprocess.run(
File "/usr/lib/python3.10/subprocess.py", line 526, in run
raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command './sy-rep-pe-party.x -p 0 -N 3 -h 453a4594a0e1 --n_shares 1 --start 0 --input_party_i 0' returned non-zero exit status 134.

this one is run with docker

[hiddely]

Hi, can you give me the suite ids and the runs you are referring to? The zip files you sent me contain a lot of successful runs

[Alfred-hhy]

hi,hidde
Thank you very much for your reply, sorry for the lack of details, you can check the latest run11 of fairness and robustness of cloud and docker (not only these two, others will also appear), this experiment always seems to be an error!
Then today when I tried again, I found that if the config.n and batchsize of the cifar and mnist datasets in the project scripts directory audit_fairness and audit_robustness are not the same for the auditfairness and robustness datasets, the overflow error
overoveroveroveroverflowflow: over: overoveroverflowoverflowover: flowflowoverflow: flow: : overoverflow: : overflow17: 22flowover29flow2819/[24flowflow/: /3016flow///31: [: 1616, /: : [flow1025[/]1627, //21[[/: [10[[1616, [16/18, 1616, [10, ]]10
I mentioned before may occur when I run poetry run make ring script=audit_fairness dataset=mnist_full_3party, but not for the adult dataset. The original batchsize is 128 and n is 1024, but here the eval method of mpspdz is called, which should work normally, so I'm not sure if it's a memory problem, but the memory usage isn't as high as it should be!
Since my configuration is a bit different from your original code, I changed the repository address to learn how to test it, and I'll send the docker container image docker pull 3555765403/mpspdz:latest I'm using (with mpspdz compiled and basically installed, as well as the -j parameter and the number of threads changed) to dockerhub later on, which should help you reproduce the problem!

[Alfred-hhy]

hi,hidde,
I'm sorry there were some configuration issues with the network in my previous docker image configuration, the new one is docker pull 3555765403/mpspdz:flattened , you may still need to delete the proxy settings.I’m encountering two issues after fixing my network configuration:

1. Signature mismatch error
   When running the adult suite under the MAL protocol pc commitment, the program always aborts with(no matter what suite with mal protocal):

   terminate called after throwing an instance of 'signature_mismatch'
     what():  Signature in Persistence/Transactions-P0-P251.data doesn't match protocol. Re-run preprocessing
   Aborted (core dumped)
   

   Comparing the logs, I see the actual signature:

   5350445a2d77697365206d616c6963696f7573207265706c69636174656420676670002000000012ab655e9a2ca55660b44d1e5c37b00159aa76fed00000010a11800000000001833f6ceebee9705971fac8620891e4f71c3694e689f161352cda9df193e6f202d5045a175602b7116c46852a100248635778ef0aee6527aa300c3a37c7e7d805
   

   but the expected one (from your example) ends at(howerve with the sh protocal, It's correct end with 00000010a11800000000001 ):

   …00000010a11800000000001
   

   I’m using mpspdz/ecdsa/share_utils.hpp for reading and writing these signatures, so I’m confused why they don’t match. I’ve tried re-running preprocessing and manually adjusting the code, but the error persists.

2. ValueError: Unknown type my_squant
   During inference on MNIST and CIFAR10, I get:

   ValueError: Unknown type my_squant
   

   It seems that my_squant is a class defined in MP-SPDZ, but I can’t find any handler for it in the reference code. How should I add support for this type?

Any guidance on resolving these errors would be greatly appreciated. Thank you!

inference_code.tar.gz
inference_3pc_mal_pc_wrong.tar.gz

I've only provided the log files for the error, the log files for the correct experiment are basically the same as yours

[Alfred-hhy]

Hi, hidde
If possible, could you share a working LAN‑based test‑environment image (for example, a Docker image) that successfully runs simple datasets such as Adult and MNIST? I’ve tried several times but still can’t get it to work.