You can configure your guest and the watchdog vdev in its VM to dump the guest's
state in the event of a crash.
This dump will provide you with the guest's state at the time that the watchdog
intervened: the guest memory and its CPU registers (i.e., the vCPU registers in the
hosting qvm process instance's vCPU threads).
To cause the guest to dump:
- In the guest, add a watchdog kicker (e.g., wdtkick) and
configure it to kick the virtual watchdog in the VM at the appropriate intervals
(see wdtkick in the QNX Neutrino OS
Utilities Reference).
- Add the appropriate watchdog vdev shared object to your hypervisor host's
buildfile, and add the watchdog vdev to the hosting VM's configuration. QNX
hypervisors include two watchdog vdevs: vdev wdt-sp805 (ARM) and vdev wdt-ib700 (x86).
- Configure the watchdog vdev to dump if the guest fails to kick it within the
specified time interval (see “Watchdogs” in the “QNX Hypervisor for Safety”
chapter).
- In the configuration for the hosting VM, specify the dump
option so that it writes the hosted guest's dump to a file (see dump in the “VM Configuration Reference” chapter).
Alternatively, you can send a SIGUSR2 signal to the
qvm process instance to cause the guest to dump. As with the
watchdog method, to have the hosted guest's dump written to a file you need the
dump option specified appropriately in the hosting VM.