Hi, I have a couple of questions related to using bpf programs with tc. Apologies in advance because I am brand new to tc. 1. I've noticed that a lot of user space tools that orchestrate bpf programs rely on bpf_tail_call to chain bpf programs together. However, I see that tc has its own chaining mechanism. Is there any downside to using tc's chaining mechanism instead of bpf_tail_calls? For some context, I'm working on one such user space orchestration tool, and I would prefer that the bpf programs themselves don't need to have the logic to do the chaining. 2. Sorry if this question is really naive, but I'm having trouble finding documentation about tc chaining in general: How does chain modification work? Say I have a chain of three BPF programs handling traffic: A->B->C, and I want to reorder them to A->C->B. What would be the proper way to do that, and could it be done (nearly) atomically? My guess is that I would be able to build up a new class or qdisc that's not taking traffic, then add my bpf programs in the new order, and then replace my original class or qdisc with the new one. I wouldn't exactly know how to do that, but am I even close? Thanks so much for any advice; I think it will really help me get my bearings here. -Brian
