netdev - Re: [Patch bpf-next v3 1/4] tcp: introduce tcp_read

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: <62a8e7c340baf_2f2a0208a2@john.notmuch>
Date:   Tue, 14 Jun 2022 12:55:47 -0700
From:   John Fastabend <john.fastabend@...il.com>
To:     Cong Wang <xiyou.wangcong@...il.com>,
        John Fastabend <john.fastabend@...il.com>
Cc:     Linux Kernel Network Developers <netdev@...r.kernel.org>,
        bpf <bpf@...r.kernel.org>, Cong Wang <cong.wang@...edance.com>,
        Eric Dumazet <edumazet@...gle.com>,
        Daniel Borkmann <daniel@...earbox.net>,
        Jakub Sitnicki <jakub@...udflare.com>
Subject: Re: [Patch bpf-next v3 1/4] tcp: introduce tcp_read_skb()

Cong Wang wrote:
> On Thu, Jun 9, 2022 at 12:12 PM John Fastabend <john.fastabend@...il.com> wrote:
> > Considering, the other case where we do kfree_skb when consume_skb()
> > is correct. We have logic in the Cilium tracing tools (tetragon) to
> > trace kfree_skb's and count them. So in the good case here
> > we end up tripping that logic even though its expected.
> >
> > The question is which is better noisy kfree_skb even when
> > expected or missing kfree_skb on the drops. I'm leaning
> > to consume_skb() is safer instead of noisy kfree_skb().
> 
> Oh, sure. As long as we all know neither of them is accurate,
> I am 100% fine with changing it to consume_skb() to reduce the noise
> for you.

Thanks that would be great.

> 
> Meanwhile, let me think about how to make it accurate, if possible at
> all. But clearly this deserves a separate patch.

Yep should be ok. We set the error code in desc->error in the verdict
recv handler maybe tracking through this.

> 
> Thanks.