[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <578A4498.9030307@mojatatu.com>
Date: Sat, 16 Jul 2016 10:28:40 -0400
From: Jamal Hadi Salim <jhs@...atatu.com>
To: Cong Wang <xiyou.wangcong@...il.com>
Cc: David Miller <davem@...emloft.net>,
Daniel Borkmann <daniel@...earbox.net>,
nikolay@...ulusnetworks.com,
Linux Kernel Network Developers <netdev@...r.kernel.org>
Subject: Re: [RFC PATCH net-next 1/1] Introduce skbmod action
On 16-07-12 01:10 PM, Cong Wang wrote:
> On Mon, Jul 11, 2016 at 5:12 AM, Jamal Hadi Salim <jhs@...atatu.com> wrote:
>> From: Jamal Hadi Salim <jhs@...atatu.com>
>>
>> This action is intended to be an upgrade from a usability perspective
>> from pedit. Compare this:
>>
>
> Definitely agree we need a more user-friendly interface.
>
>>
>> pedit is a good starting point - but once you start going to a large
>> number of policies then from a programmability, ops and usability point
>> of view you need something with more succint params.
>>
>
> But it is still unclear why we can't just build something on top of
> pedit? Since pedit accepts keys like u32, user-space is free to
> introduce any wrapper on top of it.
>
That hasnt served us well so far for reasons i described earlier
especially in the dump case.
It hasnt served u32 as a classifier well either (yes, those dump
wrappers exist on the tc user code); i am almost willing to bet
you didnt even know such a feature existed ;->
> It should not be a problem for libnl3 to build anything on top too.
And I dont use libnl3 either.
Again, as i stated earlier: pedit is a good first starting point.
At some point usability gets in the way and this is a natural next step.
I am going to work on the patch and will add src mac and ethertype for
completion.
cheers,
jamal
Powered by blists - more mailing lists