damiandixon/llvm-dataflow-analysis

These programs are demonstrations of how LLVM can be used for (very simple) static dataflow analyses (both inter- and intraprocedural). The presentation is illustrative and does not demonstrate how to implement scalable analyses.

The provided filepolicy analysis identifies simple errors in using fread, fwrite, and fclose where they may potentially be called on files that have already been closed.

The provided constant-propagation analysis identifies simple constant values that can be determined at compile time. It then prints out the computable constant arguments to all function calls in the module.

The provided futurefunctions analysis uses backward dataflow analysis to identify the functions that may be called in the future at all call sites in a program.

Building with CMake

1. Clone the repository.

    git clone https://github.com/nsumner/llvm-dataflow-analysis.git
   

2. Create a new directory for building.

    mkdir dfbuild
   

3. Change into the new directory.

    cd dfbuild
   

4. Run CMake with the path to the LLVM source.

    cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=True \
        -DLLVM_DIR=</path/to/LLVM/build>/lib/cmake/llvm/ ../llvm-dataflow-analysis
   

5. Run make inside the build directory:

    make
   

This produces tools called bin/filepolicy, bin/constant-propagation, and bin/futurefunctions.

Note, building with a tool like ninja can be done by adding -G Ninja to the cmake invocation and running ninja instead of make.

Running

First suppose that you have a program compiled to bitcode:

clang -g -c -O1 -emit-llvm ../llvm-dataflow-analysis/test/filepolicy/c/01straightCorrect.c -o 01.bc

Running the file policy analyzer:

bin/filepolicy 01.bc

The tests in the tests directory can be run using the provided Makefiles