post_run_reset in Executors about libafl HOT 6 OPEN

for me it is not yet intuitive if and when a wrapping Executor calls the post_run_reset functions of the Executors it wraps.

Yes we know it that these APIs are bad and we have a plan to refactor it in near future.
Originally the post_run_reset() was for TimeoutExecutor, but now I think wrapping wrapping executors to implement timeout is a bad choice of design.

Our plan now is to remove TimeoutExecutor and merge it to other executors. Then no post_run_reset is needed anymore.
@andreafioraldi said he'll work on in December.

from libafl.

Allright! Nevertheless, I think a post_run_reset would still be good to have, to allow the executors to clean up after running (e.g. running a reset script on the server).

from libafl.

The clean up can just happen on execute, right? Or would you specifically want to have that outside of the evaluation/after the evaluation/async?

from libafl.

For my usecase, I am combining several executors that have different tasks and each of the executors gets called to cleanup its specific things after the whole execution (consiting of several executors) is finished.

To be more specific, I have an executor which starts a network sniffing to record the network traffic of the actual network-based execution. This sniffing needs to be stoped after the actual execution is finished. Right now, I am solving this by stopping the network sniffing in the post_run_reset function, but there might be a different solution to this as well.

from libafl.

can you use observer to do it instead?
You should have something like NetworkSniffingObserver right? You could do the clean there too.

In my opinion, the post_run_reset function should be removed

from libafl.

Missing from my specification was that I am using the recorded network traffic to update the coverage map and then use a MapObserver to analyze this coverage map. So yes, I think it might also be possible to use an Observer for this but then I would need to stack observers.

Just to be clear, I am not trying to defend the post_run_reset function in general, just trying to explain why I used it in the first place ^^

from libafl.