In this section, we highlight some common problems users may face when running Magma or integrating with it. Make sure to check out the FAQ section in case your question is answered there.

• Campaigns terminate as soon as they start, regardless of the value of TIMEOUT.
  Check the log files in workdir/log and workdir/ar/fuzzer/target/program/run/log. It is likely that AFL (and AFL-based fuzzers) are terminating early due to the Pipe at the beginning of 'core_pattern' error. To fix that:

   sudo bash -c 'echo core > /proc/sys/kernel/core_pattern' 

• rm: cannot remove 'cache': Device or resource busy when deleting the work directory.
  The captain/run.sh script does not currently implement an exit handler for cleanup. When it is terminated prematurely (through Ctrl-C or a runtime error), it may not unmount the workdir/cache directory automatically. To fix that, just unmount it manually before attempting to remove it:

  sudo umount /path/to/workdir/cache
  rm -r /path/to/workdir/cache
  

• level=error msg="can't add file ... to tar" when building a Magma image.

  When building a Docker image, the context of the image is sent to the Docker daemon. In Magma, the context is the root directory of the project, and everything in it and its subdirectories. If your WORKDIR points somewhere within the context, and if your workdir is not empty, then it will be copied over to the Docker daemon. This may be the root of the issue.

  To fix it, move your working directory somewhere outside the project hierarchy, and re-assign your WORKDIR parameter to point to it.

• groupadd: GID '0' already exists when building a Magma image.
  You are likely running the captain/build.sh script as the root user. One reason to do that may be that your normal user account does not have the proper permissions to launch docker. To fix that, add your current user to the docker user group:

  sudo usermod -aG docker $USER
  newgrp docker # OR logout and log back in
  

• Seed corpus directory is empty inside container.
  Magma's magma/run.sh script performs a seed pruning phase before launching every campaign, to make sure that no seeds trigger any bugs or crash the target program. If the seed corpus is empty, it means the pruning stage detected all seeds as crashing, and it is more likely that the compiled program just crashes on launch, regardless of the input file. Verify that the compilation process is not flawed and test your target program manually inside the container.
• • can't find file to patch OR
  • Hunk #NNN FAILED at MMM
  You may be using the trunk version of Magma, in which targets and patches have not been verified. Magma targets are constantly being changed and updated, and Magma patches may no longer apply to the latest version of those targets. As such, you are encouraged to use one of the versioned Magma releases which anchor the versions of targets and are verified to compile.
• I want to debug a binary inside the container, but GDB shows <optimized out> even after I added -O0 -g -ggdb to the compile flags.

  This may be due to AFL's compiler implicitly optimizing the target. From AFL's environment variables document:

   By default, the wrapper appends -O3 to optimize builds. Very rarely, this will cause problems in programs built with -Werror, simply because -O3 enables more thorough code analysis and can spew out additional warnings. To disable optimizations, set AFL_DONT_OPTIMIZE. 

  So, export AFL_DONT_OPTIMIZE=1 before launching the compiler should do the trick.