From mboxrd@z Thu Jan  1 00:00:00 1970
From: Daniel Borkmann <daniel@iogearbox.net>
Subject: Re: [PATCH iproute2 -next] tc, bpf: finalize eBPF support for cls
 and act front-end
Date: Thu, 02 Apr 2015 20:38:54 +0200
Message-ID: <551D8CBE.1030609@iogearbox.net>
References: <d9a577418790f99130c1cb20dc01cf50c2d6a7da.1427754287.git.daniel@iogearbox.net> <1427933636.1888890.248325033.0A76BE0D@webmail.messagingengine.com> <551C8C2D.5060504@iogearbox.net> <1427934556.1892386.248333013.431CD73B@webmail.messagingengine.com> <551D17B4.2090903@iogearbox.net> <1427974255.2093319.248499373.05D36231@webmail.messagingengine.com> <551D3156.9030407@iogearbox.net> <551D6ACF.3090101@plumgrid.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 7bit
Cc: jiri@resnulli.us, tgraf@suug.ch, netdev@vger.kernel.org,
	Jamal Hadi Salim <jhs@mojatatu.com>
To: Alexei Starovoitov <ast@plumgrid.com>,
	Hannes Frederic Sowa <hannes@stressinduktion.org>,
	stephen@networkplumber.org
Return-path: <netdev-owner@vger.kernel.org>
Received: from www62.your-server.de ([213.133.104.62]:38885 "EHLO
	www62.your-server.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1753109AbbDBSjQ (ORCPT
	<rfc822;netdev@vger.kernel.org>); Thu, 2 Apr 2015 14:39:16 -0400
In-Reply-To: <551D6ACF.3090101@plumgrid.com>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

On 04/02/2015 06:14 PM, Alexei Starovoitov wrote:
> On 4/2/2015 5:08 AM, Daniel Borkmann wrote:
>> On 04/02/2015 01:30 PM, Hannes Frederic Sowa wrote:
>>>> ... perhaps such a built-in fake device for retrieving bpf map fds
>>>> might be interesting, e.g. exec 4<>/dev/bpf/<obj-file>/<map-name> if
>>>> that has been given to bash?
>>>>
>>>> Anyway, I think to have some utility for shell scripts, as you
>>>> suggest, certainly sounds interesting!
>>>
>>> All file descriptors will be inherited by exec as long as the O_CLOEXEC
>>> flag wasn't specified on them. So you can retrieve the fds via af_unix
>>> and just exec a new shell. The file descriptors will stay open and you
>>> can pass the numbers of the fds via environment. This wouldn't need
>>> changes to bash or kernel.
>>
>> Okay, I will give it a try. Thanks!
>
> I think even if it's little awkward to do it in user space, it's better
> than adding stuff to kernel. This /dev/bpf/obj-file/map-name or sysfs
> might start hitting scalability issues when we try to load thousands
> of bpf programs.

So I was just hacking a prototype together to execve(2) and pass those
map fds over env to the bash instance, so you have them there next to
stdin, stdout, etc. Bit funny, but works as well for bpf(2). It might
not work on a larger-scale, but perhaps for experimenting with eBPF
maps, it would be good as a start. I'll try to carve something out for
a follow-up to make it useful in shell scripts, thanks!