> In addition, we have no universally accepted definition of [**what is part**] of the operating system. 

We could rephrase it thus:

> * "What kinds of modules should be included in an operating system? What kind of a module should be considered **part of an operating system**? 
* "We don't know! There is no universally accepted definition of **what is part** of the typical operating system (and **what is not part** of the typical operating system)." 

Imagine you have a list of 1000 modules: I/O drivers, text editors, graphics procedures etc. And someone asks you: 

> A great collection! But **which are part** of the operating system? 

In your example, the word **what** is used instead of **which** because the number of components that could be included in an operating system is well nigh limitless nowadays.

----
Regarding your idea:

> What I think this is trying to say is probably that there is no agreed-upon definition of what does the term "part of the operation" system mean. If that's the case, then why not say it like this: *In addition, we have no universally accepted definition of what part of the operating system is*.  

Had we wanted to say this, we would've added the indefinite article **a**, because "part" in this sense is a singular count noun:

> We have no universally accepted definition of what is **a** part of the operating system. 

Compare: 

> We have no universally accepted definition of what **a** planet is. (or " [...] of what is **a** planet") 

----
(Kudos to *Stephie* for her comment on the implied clause "What is not part of the OS".)