netdev - Re: [PATCH bpf-next 01/10] tcp: Use a struct to represent a saved

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

Message-ID: <CANn89iKj-okNmJhUnOmzdGVbOHi8zdF+=oO4RUFOo2X6M1kZHw@mail.gmail.com>
Date:   Tue, 30 Jun 2020 16:35:20 -0700
From:   Eric Dumazet <edumazet@...gle.com>
To:     Martin KaFai Lau <kafai@...com>
Cc:     bpf <bpf@...r.kernel.org>, Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        kernel-team <kernel-team@...com>,
        Lawrence Brakmo <brakmo@...com>,
        Neal Cardwell <ncardwell@...gle.com>,
        netdev <netdev@...r.kernel.org>,
        Yuchung Cheng <ycheng@...gle.com>
Subject: Re: [PATCH bpf-next 01/10] tcp: Use a struct to represent a saved_syn

On Tue, Jun 30, 2020 at 4:24 PM Martin KaFai Lau <kafai@...com> wrote:

> Interesting idea.  I don't have an immediate use case in the mac header of
> the SYN but I think it may be useful going forward.
>
> Although bpf_hdr_opt_off may be no longer needed in v2,
> it will still be convenient for the bpf prog to be able to get the TCP header
> only instead of reparsing from the mac/ip[46] and also save some stack space
> of the bpf prog.  Thus, storing the length of each hdr would still be nice
> so that the bpf helper can directly offset to the start of the required
> header.
>
> Do you prefer to incorporate this "save_syn:2" idea in this set
> so that mac hdrlen can be stored in the "struct saved_syn"?

Sure, please go ahead.

>
> This "unused:2" bits have already been used by "fastopen_client_fail:2".

Oh that is possible, I have sent the patch as it was when we merged it
years ago.

> May be get 2 bits from repair_queue?

Hmm... this repair_queue is used in fast path, I would rather keep it
as u8 for better code generation.