linux-kernel - Re: [bpf-next 1/2] tcp: replace SOCK_DEBUG() with tcp

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 linux-cve-announce PHC
Open Source and information security mailing list archives

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

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

Message-ID: <CALOAHbA4PrgV8tem3cMt+22Vm2GQw=oaRsXMcAr2XBgX8hhjKw@mail.gmail.com>
Date:   Wed, 13 Feb 2019 11:04:21 +0800
From:   Yafang Shao <laoar.shao@...il.com>
To:     Alexei Starovoitov <alexei.starovoitov@...il.com>
Cc:     Eric Dumazet <edumazet@...gle.com>,
        Eric Dumazet <eric.dumazet@...il.com>,
        Daniel Borkmann <daniel@...earbox.net>,
        Alexei Starovoitov <ast@...nel.org>,
        Yonghong Song <yhs@...com>, Lawrence Brakmo <brakmo@...com>,
        David Miller <davem@...emloft.net>,
        netdev <netdev@...r.kernel.org>,
        LKML <linux-kernel@...r.kernel.org>, shaoyafang@...iglobal.com
Subject: Re: [bpf-next 1/2] tcp: replace SOCK_DEBUG() with tcp_stats()

On Wed, Feb 13, 2019 at 10:49 AM Alexei Starovoitov
<alexei.starovoitov@...il.com> wrote:
>
> On Tue, Feb 12, 2019 at 6:15 PM Eric Dumazet <edumazet@...gle.com> wrote:
> >
> > Do not add more debugging stuff unless you can demonstrate
> > they actually allowed you to find a real bug and that you sent a
> > public fix for it.
> >
> > Just adding "cool stuff" in TCP stack does not please me, it is only
> > more complexity for unproven gain.
>
> I agree.
> I don't see why this debugging of 'abnormal TCP' cannot be done
> with kprobes and tracepoints.

kprobe is not suitable, because we have to use the line number, that's
a trouble cross all kernel versions.
Regarding  tracepoints, I don't think it is a good idea to introduce
more and more tcp tracepoints and make them crowed as well.

> Instrumenting every tcp counter increment is overkill.