On Fri, Nov 15, 2019 at 02:04:01AM +0100, Daniel Borkmann wrote: > for later modifications. In ii) fixup_bpf_tail_call_direct() walks > over the progs poke_tab, locks the tail call maps poke_mutex to > prevent from parallel updates and patches in the right locations via ... > @@ -1610,6 +1671,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) > prog->bpf_func = (void *)image; > prog->jited = 1; > prog->jited_len = proglen; > + fixup_bpf_tail_call_direct(prog); Why not to move fixup_bpf_tail_call_direct() just before bpf_jit_binary_lock_ro() and use simple memcpy instead of text_poke ? imo this logic in patch 7: case BPF_JMP | BPF_TAIL_CALL: + if (imm32) + emit_bpf_tail_call_direct(&bpf_prog->aux->poke_tab[imm32 - 1], would have been easier to understand if patch 7 and 8 were swapped.
