QSYM results are inaccurate, does not work on modern kernel about fuzzbench HOT 9 OPEN

@jakkdu - We would really appreciate if you can help us fix QSYM integration and benchmark results from it.

from fuzzbench.

We have temporarily removed it since integration is not working. Once we get fixes in upstream QSYM code, then we want to bring it back.

from fuzzbench.

Hi.
This is related to sslab-gatech/qsym#4.
If QSYM is ready for the latest kernel, I will let you know.
Thank you.

from fuzzbench.

Thanks @jakkdu !

from fuzzbench.

Hi, @inferno-chromium. I recently checked recent PIN. Unfortunately, I think they still have issues in supporting c++ features, system calls, and external libraries (e.g., https://groups.io/g/pinheads/topic/65995449#12831). I want to dig into this issue more if I can shift QSYM to use PIN 3.x, which seems the most reasonable way to support latest kernel. If it fails, I need to use other DBT, which will require more time to modify :(, sadly. If I have any update, I will let you know. Thank you.

from fuzzbench.

Thanks @jakkdu , using PIND 3.x seems reasonable, thanks for working on this!

from fuzzbench.

@jakkdu I can highly recommend DynamoRIO. it is about 10x faster than PIN (e.g. see my afl-pin and afl-dynamorio implementations). It also works on ARM and AARCH64.

from fuzzbench.

@vanhauser-thc Actually, at the start of the QSYM project, I tried to use DynamoRIO because I love open-source tools for easier debugging. But at that moment, DynamoRIO fails to work with z3 (DynamoRIO/dynamorio#1881). I think it is worth to check it again whether this problem is still there. Thank you for your recommendation :).

from fuzzbench.

@vanhauser-thc It's off topic, though. I should merge AFL++ to QSYM. Please forgive my laziness + lack of time due to my job searching.

from fuzzbench.