separating malloc and large physical ranges

[Jonathan S. Shapiro]

After all that, I sat down and wrote down a list of what really needs
to be dynamically allocated.

With the sole exception of mbuf clusters, it's all a lot smaller than
a page.  mbuf clusters are 1/2 page.

I'm therefore going to use a fairly simple expedient: allocate page
frames for use by malloc, use a per-frame free list, and doubly link
all frames of similar type so they can be thrown back into the free
pool.

I'll solve the more serious range management problem later; this
approach will get things moving quickly.


shap