Studying how Linux boots up its timekeeping subsystems...
It inits a couple spinlocks & some properties via a helper, possibly (via another helper) init each entry of an array in the same way, invokes (via a helper) CPU-specific code in one way or another to retrieve the realtime clock, validates (via helpers) results, sums initial monotonic time, clears out NTP state & starts a timer for it via a couple helpers, retrieves a "jiffies" clocksource (compiled specially on s390 machines), ...
1/?
... invokes its `enable()` method, invokes to populate (and compute, some of which is conditional) numerous properties on it, populates some global properties, & invokes another suite of helpers to update more properties under lock.
This followed up by initializing a CPU-specific "time" subsystem, which generally involves invoking a callback that was chosen earlier.
1.5/1.5 Fin for today! Tomorrow: Randomness!