Why not just use IDs from 0 through N? You can then do >> 5 to get the word number and & 31 to get the bit offset. That's really how the FDT should have represented it too, probably... logically it's just a big bit vector.

<< 5 but & 0xff is inconsistent and not what I said (which was & 31 ie 0x1f). Also is cells[1] the bit number or bit mask? Because it's being encoded as if it's the former, but used in the other functions as if it's the latter (no 1u << bit).