Internals: crash dumps

kernel/core/core_crash.c + the exception path in arch_exceptions.c.

The panic sequence

1. The exception stub builds the frame (vector, RIP, registers, CR2/CR3 where relevant).
2. The dumper prints it to serial - the exact lines you need for a report.
3. A crashlog record goes to /etc - survives the reboot.
4. The machine halts (or the hardware watchdog resets it if one was armed - see watchdog.md).

After reboot

oc> crashlog      (the record list)
oc> dumpinfo      (the decoded dump: vector, fault address, registers)

Why IST matters here

#DF (double fault) and #MC run on IST1 - their own stack - so a fault during a fault handler still reaches the dumper. The exception-exit path never context-switches on IST2 (BUG-0139): a context switch on an IST would corrupt the exception frame of the next nested fault.

Testing it

crashtest = deliberate panic, then dumpinfo proves the round trip. The regression battery checks crashlog on every run.