I´m writing a slab allocator for my os and the allocation does work, but I have a design problem with the deallocation.

I don´t know how the slab allocators in Linux and Solaris are deallocating a slab. Do they go through the lists (free, full, partial) and find the right pg(s) for the slab? Or do they do it another way?

I thought of sorting the lists by the start address of the pg(s). This would be faster for deallocating, but lower for allocating.

Any tips?