Replying to correct myself: > Since sockmap update can happen in a tracing prog, can you really expect a > socket to be always owned?
Ha, but I'm wrong! Of tracers only BPF_TRACE_ITER allows calling sock_map_update_elem(). And those are supposed to lock_sock(). Except some don't, e.g. sock_map_seq_show(). Or is there something else I'm missing?
