A Linux inode is the file system's representation of the base info of a file, from which the file's data blocks can be found. It also carries file metadata, but its real function is as the root of the file's block index tree. The format depends on the file system. The internal identity of a file is an inode number, not an inode. What you get from 'stat' is the file's metadata, which mostly comes from the inode data structure on disk, but isn't necessarily in the same format. The formats were the same back around UNIX V7, but there have been some changes in the last few decades as file systems improved.
If you need to count bits in a word, the first question is whether your CPU has hardware to do that. NSA always liked population count instructions, which are useful in cryptanalysis, and that's why most supercomputers had them since the 1960s. Now they're finally in Intel x86 CPUs with SSE4.2 (added around 2006), which has a "popcount" instruction.[1]
A MAC address for Ethernet is six bytes. There are other hardware layer systems, and Google probably uses some of them. Fiber Channel fabric uses only a 3-byte address, for example.
Hash tables are not O(1) lookup. It's an exponential as the table fills up. It's near O(1) only with a near-empty table. There's a space/time tradeoff on how full you let the table get before you expand it.
Quicksort is average O(N log N), but the worst case is much worse, which is why nobody uses pure Quicksort any more. You can beat O(N log N) with a distribution sort. The first sort to do that was SyncSort, the first patented algorithm. It's a distribution sort with self-adjusting buckets.
Who wrote this interviewer's answer sheet?