Does linux kernel make use of virtual memory for its data structures (page tables, descriptors, etc.)? More specifically:
Are kernel space addresses translated in the MMU (pagetable walking)?
Could kernel memory get swapped out?
Could a memory access to a kernel data structure cause a page fault?
Are there differences between linux and other unix in this respect?
Are kernel space addresses translated in the MMU (pagetable walking)?
Yes, all addresses are translated in the MMU; see Is the MMU inside of Unix/Linux kernel? or just in a hardware device with its own memory? for details.
Could kernel memory get swapped out?
A kernel could theoretically be designed so that it can be swapped out. In practice it’s difficult; the Linux kernel in particular can’t be swapped out. Some code paths in the kernel do have to deal with page-ins however; see Why are `copy_from_user()` and `copy_to_user()` needed, when the kernel is mapped into the same virtual address space as the process itself? for example.
Could a memory access to a kernel data structure cause a page fault?
In most if not all cases, if this were to happen, it would lead to a kernel panic. So yes, it could happen, but it would be a bug.
Are there differences between linux and other unix in this respect?
As far as I’m aware other (current) Unix-style implementations are similar. Early Unix didn’t support virtual-memory-based swapping (i.e. paging out arbitrary pages) anyway so it wasn’t a concern there.
