netdev - Re: [PATCH RFC v3 0/8] xdp: Infrastructure to generalize XDP

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: <20170221.233123.1325495796860476497.davem@davemloft.net>
Date:   Tue, 21 Feb 2017 23:31:23 -0500 (EST)
From:   David Miller <davem@...emloft.net>
To:     tom@...bertland.com
Cc:     kubakici@...pl, netdev@...r.kernel.org, kernel-team@...com
Subject: Re: [PATCH RFC v3 0/8] xdp: Infrastructure to generalize XDP

From: Tom Herbert <tom@...bertland.com>
Date: Tue, 21 Feb 2017 20:02:07 -0800

> I took out the parts about allowing non-BPF hooks in the patchset as
> you requested.I believe code that is cleaner than what is currently
> there, and the fact that the API is extensible to allow other uses
> is a hallmark of good design.

For BPF only cases, there is no need for a chain of callbacks.  BPF
has tailcalls, which can be used to solve whatever problem chained
XDP stuff would be used for.

The only other possible use of chained XDP is non-BPF uses, thus my
and others's fundamental objection.

It's not a cleanup or a simplification for driver writers, you're
adding a new feature, and one whose need is not at all clear.

I also object to chaining on another level.  The whole idea of XDP
programs are that they exist in a very _precise_ and specific context
and do one thing.  If there are chains, there are dependencies, and
thus the context is variable.  We do not want this, trust me.

Is it such a huge deal to defer this chaining thing to after your
other simplifications?  All of these things you are trying to add are
blocking the less controversial and useful parts of your work.

Thanks.