Re: [RFC PATCH bpf-next 4/5] bpf, x64: rework pro/epilogue and tailcall handling in JIT

[Alexei Starovoitov]

On Fri, Jul 10, 2020 at 8:20 PM Alexei Starovoitov
<alexei.starovoi...@gmail.com> wrote:
>
> Of course you are right.
> pop+nop+push is incorrect.
>
> How about the following instead:
> - during JIT:
> emit_jump(to_skip_below)  <- poke->tailcall_bypass
> pop_callee_regs
> emit_jump(to_tailcall_target) <- poke->tailcall_target
>
> - Transition from one target to another:
> text_poke(poke->tailcall_target, MOD_JMP, old_jmp, new_jmp)
> if (new_jmp != NULL)
>   text_poke(poke->tailcall_bypass, MOD jmp into nop);
> else
>   text_poke(poke->tailcall_bypass, MOD nop into jmp);

One more correction. I meant:

if (new_jmp != NULL) {
  text_poke(poke->tailcall_target, MOD_JMP, old_jmp, new_jmp)
  text_poke(poke->tailcall_bypass, MOD jmp into nop);
} else {
  text_poke(poke->tailcall_bypass, MOD nop into jmp);
}