Dmitry Vyukov

The idea behind saving these crashes was to give at least some visibility into what was lost. If we are not showing them, then it's easier to just not save...

Sounds reasonable. Any help to enforce this would be useful. Function signature with different arguments for const/non-const part will foce anybody writing code to think about this at least. Without...

We already control it statically including indirect calls (jp.Logf): https://github.com/google/syzkaller/blob/bfb20202da73451f997b711447ee68bc3635868e/tools/syz-linter/linter.go#L299 If we don't check it statically, do you think such bugs won't be introduced? I would assume they will be...

I am thinking if it's better to do this, or just change fail to exitf in the event_set function... We have a long history of executor managing to corrupt itself...

Why would a kernel driver want to do this? This does not look like it's a serious security protection. Can this check be removed entirely? Or at least made configurable...

Dup of #604?

We don't have UML support in syzkaller. We run CI on our machines with our docker image, so theoretically we could use qemu there.

No, I don't have any concrete examples. I just remember that several times I saw that my local instances is reproducing something for 1+ days and eventually I had to...

> FWIW, I think we want to port the ORC unwinder to UML, I was planning to take a look at it eventually, but no promises. Is it a prerequisite...

What compiler are you using? And what warnings does it produce?