How does cache coherence work when there are multiple levels of private caches and then shared memory?

Question

Most of the examples online explain coherence with a configuration similar to this: each single core connected to single private cache (no multiple cache levels per core) an