Another alternative to log the kernel messages on oops or panic is to configure the function tracer to dump its buffer contents to the console so that the events leading up to the crash can be analyzed. Use the following command:

$ echo 1 > /proc/sys/kernel/ftrace_dump_on_oops  

The sysrq-z combination will also dump the contents of the tracing buffer to the console, as does calling ftrace_dump() from the kernel code.