On Tue, Jul 18, 2017 at 03:19:00PM -0700, Bart Van Assche wrote:

Hello Peter,

Sorry for being late, I'm trying to recover from a few weeks of leave
and the inbox is in shambles.

In a test I ran myself with kernel v4.12-rc1 I also noticed that a
WARN_ON_ONCE() statement triggered an oops in report_bug() and killed the
kernel thread of the caller instead of letting the caller continue. What I
ran into is probably the same oops as in the above call trace. For the test
I ran myself the disassembly is as follows:

(gdb) list *(report_bug+0x94)
0xffffffff812ba024 is in report_bug (lib/bug.c:177).
172                                     return BUG_TRAP_TYPE_WARN;
173
174                             /*
175                              * Since this is the only store, concurrency is not an issue.
176                              */
177                             bug->flags |= BUGFLAG_DONE;
178                     }
179             }
180
181             if (warning) {

Could this be related to patch "debug: Add _ONCE() logic to report_bug()"?
Is this a known issue? If so, is a fix perhaps already available?

Yep..  please see:

  https://marc.info/?l=linux-kernel&m=150055119925677