I don’t think your solution works. It will make sure the pointer is never null (assuming these back pointers are !Sync) sure, but rust guarantees far more that

I don’t think your solution works. It will make sure the pointer is never null (assuming these back pointers are !Sync) sure, but rust guarantees far more that that. Each ownership type (value, &, &mut) is a capability, it very important to rust guarantees that you can’t go up the chain. Each node have an owned pointer to the next means that it can mutate it at anytime, the back pointer would always alias the previous node meaning that it has to be a weakened “&” reference (allows aliasing, lifetime rules relaxed). Because all of the nodes are owned it’s possible to traverse the back pointer then move forward again and recover ownership or a &mut of the same node twice. You can however allocate the nodes in an arena and store your forward and back pointers in Cells. You can’t move the arena, but you’ll get nice back pointers. You could also make the linked list be backed by a slab and store indexes instead of node pointers. Both of these solutions will likely be more cache and allocation efficient than a traditional link list which does allocation for every node.