netdev - Re: [RFC patch net-next-2.6] net: allow multiple rx

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]

Date:	Fri, 1 Jul 2011 18:55:59 +0200
From:	David Lamparter <equinox@...c24.net>
To:	Jiri Pirko <jpirko@...hat.com>
Cc:	Nicolas de Pesloüan <nicolas.2p.debian@...il.com>,
	netdev@...r.kernel.org, davem@...emloft.net,
	shemminger@...ux-foundation.org, kaber@...sh.net, fubar@...ibm.com,
	eric.dumazet@...il.com, andy@...yhouse.net,
	Ben Greear <greearb@...delatech.com>
Subject: Re: [RFC patch net-next-2.6] net: allow multiple rx_handler
 registration

On Fri, Jul 01, 2011 at 07:45:38AM +0200, Jiri Pirko wrote:
> Thu, Jun 30, 2011 at 09:50:52PM CEST, nicolas.2p.debian@...il.com wrote:
> >Le 30/06/2011 17:16, Jiri Pirko a écrit :
> >>For some net topos it is necessary to have multiple "soft-net-devices"
> >>hooked on one netdev. For example very common is to have
> >>eth<->(br+vlan). Vlan is not using rh_handler (yet) but also for example
> >>macvlan would be useful to have hooked on same netdev as br.
> >>
> >>This patch introduces rx_handler list. size struct net_device stays
> >>intact. Measured performance regression on eth-br topo is ~1% (on received
> >>pkts generated by pktgen) and on eth-bond topo it is ~0.25%
> >>
> >>On br I think that the performance can be brought back maybe by using per-cpu
> >>variables to store port in rx_path (I must check this)
> >>
> >>Please comment.
> >>
> >>Signed-off-by: Jiri Pirko<jpirko@...hat.com>
> >
> >I like the idea of this patch.
> >
> >I didn't take time for a technical review yet, but I'm not sure
> >ordering should be static, depending on the kind of device (bond, br,
> >macvlan). Ordering is currently static, because it is hard coded.
> >
> >I think rx_handler_prio should be exposed to userspace, to allow user
> >setup to decide the exact order. Default order should be safe, but
> >user should be allowed to force a different order for special setups.
> 
> Hmm I didn't think about this. Not sure about the right way how to expose
> this.

I agree with Nicolas, static ordering might not be sufficient. I'm not
even sure if dynamic ordering is sufficient...

Currently, I can use ebtables to control which packets a bridge eats, so
I can implement e.g.
eth0
 -> eth0.123
 -> br0
    -> br0.234
by dropping vlan packets with ID 123 into ebtables "brouting", but other
devices do not have this granularity - if I have a macvlan and a 8021q
vlan, what happens?

Also, btw, the example eth<->(br+vlan) case you're citing works quite
nice if you have above ebtables setup... you're trying to make it work
without ebtables i assume?


-David

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html