Re: [syzbot] WARNING in bpf_test_run

[Steven Rostedt]

On Tue, 13 Apr 2021 09:56:40 +0200
Dmitry Vyukov <dvyukov@xxxxxxxxxx> wrote:

> Thanks for looking into this.
> If this is not a kernel bug, then it must not use WARN_ON[_ONCE]. It
> makes the kernel untestable for both automated systems and humans:
> 
> https://lwn.net/Articles/769365/
> 
> <quote>
> Greg Kroah-Hartman raised the problem of core kernel API code that
> will use WARN_ON_ONCE() to complain about bad usage; that will not
> generate the desired result if WARN_ON_ONCE() is configured to crash
> the machine. He was told that the code should just call pr_warn()
> instead, and that the called function should return an error in such
> situations. It was generally agreed that any WARN_ON() or
> WARN_ON_ONCE() calls that can be triggered from user space need to be
> fixed.
> </quote>

I agree. WARN_ON(_ONCE) should be reserved for anomalies that should not
happen ever. Anything that the user could trigger, should not trigger a
WARN_ON.

A WARN_ON is perfectly fine for detecting an accounting error inside the
kernel. I have them scattered all over my code, but they should never be
hit, even if something in user space tries to hit it. (with an exception of
an interface I want to deprecate, where I want to know if it's still being
used ;-) Of course, that wouldn't help bots testing the code. And I haven't
done that in years)

Any anomaly that can be triggered by user space doing something it should
not be doing really needs a pr_warn().

Thanks,

-- Steve