Now, let's say that, unlike the previous simple case, when the device driver requests 128 KB, the order 5 list is null; thus, as per the page allocator algorithm, we go to the list on the next order, 6, and check it. Let's say it's non-null; the algorithm now dequeues a 256 KB chunk and splits (or cuts) it in half. Now, one half (of size 128 KB) goes to the requester, and the remaining half (again, of size 128 KB) is enqueued on to the order 5 list.
The really interesting property of the buddy system is what happens when the requester (the device driver), at some later point in time, frees the memory chunk. As expected, the algorithm calculates (via its order) that the just-freed chunk belongs on the order 5 list. But before blindly enqueuing it there, it looks for its buddy block, and in this case, it (possibly) finds it! It now merges the two buddy blocks into a single larger block (of...
