netdev - Re: [PATCH bpf-next v2 7/8] bpf, x86: emit patchable direct jump as tail call

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 22 Nov 2019 18:28:42 -0800
From:   Alexei Starovoitov <alexei.starovoitov@...il.com>
To:     Daniel Borkmann <daniel@...earbox.net>
Cc:     Andrii Nakryiko <andrii.nakryiko@...il.com>,
        Alexei Starovoitov <ast@...nel.org>,
        john fastabend <john.fastabend@...il.com>,
        Networking <netdev@...r.kernel.org>, bpf <bpf@...r.kernel.org>
Subject: Re: [PATCH bpf-next v2 7/8] bpf, x86: emit patchable direct jump as
 tail call

On Fri, Nov 22, 2019 at 3:25 PM Daniel Borkmann <daniel@...earbox.net> wrote:
> >> +       case BPF_MOD_CALL_TO_NOP:
> >> +       case BPF_MOD_JUMP_TO_NOP:
> >> +               if (old_addr && !new_addr) {
> >> +                       memcpy(new_insn, nop_insn, X86_PATCH_SIZE);
> >> +
> >> +                       prog = old_insn;
> >> +                       ret = emit_patch_fn(&prog, old_addr, ip);
> >> +                       if (ret)
> >> +                               return ret;
> >> +                       break;
> >> +               }
> >> +               return -ENXIO;
> >> +       default:
> >
> > There is this redundancy between BPF_MOD_xxx enums and
> > old_addr+new_addr (both encode what kind of transition it is), which
> > leads to this cumbersome logic. Would it be simpler to have
> > old_addr/new_addr determine whether it's X-to-NOP, NOP-to-Y, or X-to-Y
> > transition, while separate bool or simple BPF_MOD_CALL/BPF_MOD_JUMP
> > enum determining whether it's a call or a jump that we want to update.
> > Seems like that should be a simpler interface overall and cleaner
> > implementation?
>
> Right we can probably simplify it further, I kept preserving the original
> switch from Alexei's code where my assumption was that having the transition
> explicitly spelled out was preferred in here and then based on that doing
> the sanity checks to make sure we don't get bad input from any call-site
> since we're modifying kernel text, e.g. in the bpf_trampoline_update() as
> one example the BPF_MOD_* is a fixed constant input there.

I guess we can try adding one more argument
bpf_arch_text_poke(ip, BPF_MOD_NOP, old_addr, BPF_MOD_INTO_CALL, new_addr);
Not sure whether it's gonna be any cleaner.
Intuitively doesn't feel so.