On Fri, Jan 5, 2024 at 6:33 PM Leon Hwang <hffilwlqm@xxxxxxxxx> wrote: > > > > We allow BPF_MAX_LOOPS = 8 * 1024 * 1024 in bpf_loop, > > so many calls to subprog(skb); is not an issue > > as long as they don't stall cpu and don't increase stack size. > > What if there are BPF_MAX_LOOPS subprog(skb) and there are BPF_MAX_LOOPS > loops in the tail-callee bpf prog? It's fine. Every bpf_loop is capped individually. We're working on generic "cancelling" of bpf progs when they consume too much cpu time. There is no way to do run-time counters. PS pls trim your replies.
