The code is split between .h and .c files

Rebalancing and a few more operations are library functions
There is a little more work involved in porting to user space
The implementation is very efficient, both in size and speed
The suggested traversal of the tree is iterative, not recursive

rudo$ nm --size-sort linux_rbtree.o
00000021 T rb_first
00000021 T rb_last
00000042 T rb_next
00000042 T rb_prev
0000005d T rb_replace_node
0000006d t __rb_rotate_left
0000006d t __rb_rotate_right
000000eb T rb_insert_color
000002c9 T rb_erase

A sorting program built with linux-rbtree is
Almost as fast as /usr/bin/sort
Comparable with trivial-tree on random data
Slower than qsort, but data is always sorted during operation