On Tue, Apr 13, 2021 at 7:43 PM Steven Rostedt [off-list ref] wrote:

On Tue, 13 Apr 2021 13:41:47 -0400
Steven Rostedt [off-list ref] wrote:

quoted

As the below splats look like it has nothing to do with this patch, and
this patch will add a WARN() if there's broken logic somewhere, I bet the
bisect got confused (if it is automated and does a panic_on_warning),
because it will panic for broken code that his patch detects.

That is, the bisect was confused because it was triggering on two different
issues. One that triggered the reported splat below, and another that this
commit detects and warns on.

Is it possible to update the the bisect to make sure that if it is failing
on warnings, to make sure the warnings are somewhat related, before decided
that its the same bug?

It does not seem to be feasible, bugs manifest differently in both
space and time. Also even if we somehow conclude the crash we see is
different, it says nothing about the original bug. For more data see:
https://groups.google.com/g/syzkaller/c/sR8aAXaWEF4/m/tTWYRgvmAwAJ