While the page allocator serves as an interface for memory allocations (in multiples of page size), the buddy system operates at the back-end to administer physical page management. This algorithm manages all physical pages for each zone. It is optimized to accomplish allocations of large physically contiguous blocks (pages), by minimizing external fragmentation. Let's explore its operational details.
The zone descriptor structure contains an array of struct free_area, and the size of the array is defined through a kernel macro MAX_ORDER whose default value is 11:
struct zone { ... ... struct free_area[MAX_ORDER]; ... ... };
Each offset contains an instance of free_area structure. All free pages are split into 11 (MAX_ORDER ...
