Stack overflow when analyzing

asesidaa commented 1 year ago

[x] I have searched open and closed issues for duplicates
[x] I made sure that I am not using an old project version (DO: pull Phasar, update git submodules, rebuild the project and check if the bug is still there)

Bug description

When analyzing mpi libraries, stackoverflow will happen (resulting in segmentation fault)

Steps to reproduce

Compile OpenMPI-3.1.6 with gllvm, add -O0 and -g in flags, use options --disable-mpi-fortran --disable-dlopen, extract bitcode.
Run phasar-cli --data-flow-analysis=ide-solvertest --module=./libmpi.so.40.10.4.bc --entry-points=__ALL__

Actual result: Get segmentation fault (when using ASAN, it shows stackoverflow) Expected result: Analysis finishes normally

Context (Environment)

phasar: [v0323]

Operating System:

[x] Linux
[ ] Windows
[ ] macOS

Build Type:

[x] cmake (via bootstarp.sh)
[ ] custom build

Possible solution

Currently, for OpenMPI, I have identified that the following functions will crash Phasar:

mca_coll_base_comm_unselect
ompi_datatype_init
mca_coll_base_comm_select The crash happens when processing normal edge functions. However, if I use llvm-extract to extract these functions and analyze separately, no crash happens. Currently as a workaround I just skip these functions and the analysis can proceed normally.

MMory commented 1 year ago

Hi @asesidaa ,

I am pretty sure that you did not increase the stack size. For larger programs you need to increase the maximum stack size using the ulimit command line tool prior to running phasar-cli.

Cheers Martin

MMory commented 1 year ago

Hi @asesidaa,

did my reply help? If there are no updates within 24 hours I will consider this issue as resolved and close it.

Cheers Martin

secure-software-engineering / phasar