Re: [patch 1/3] [PATCH] qeth: HiperSockets Network Traffic Analyzer

[Ursula Braun <ubraun@linux.vnet.ibm.com>]

On Tue, 2010-01-12 at 04:59 -0800, David Miller wrote:
> From: Ursula Braun <ubraun@linux.vnet.ibm.com>
> Date: Tue, 12 Jan 2010 13:25:31 +0100
> 
> > Sounds like a good idea, but there exists a qeth-specific problem if
> > sniffer configuration should be switched to ethtool: Ethtool requires
> > the existence of the network interface for the network device. The
> > network interface for s390 network devices is defined within an
> > s390-specific initialization step called "online setting". The sniffing
> > characteristic must be configured before the online setting of the
> > device, since NTA-specific initialization steps are necessary. At that
> > point of time we cannot make use of ethtool, because the network
> > interface name is not known yet. That means switching to your proposed
> > ethtool-based configuration mechanism for HS NTA would result in an
> > additional major rework of qeth device setup.
> 
> If you have a place to stick sysfs nodes, you've obviously
> instantiated the device already.

I should have been more precise: In the s390-world there are 2 steps for
network device setup - a ccwgroup-step, which establishes the "struct
device", and an online-step, which establishes the "struct net_device".
After the ccwgroup-step we have an instantiated device and can make use
of sysfs. After the online-step, we have an instantiated net_device
incl. network interface name and can make use of ethtool. HiperSockets
sniffer configuration is required before the online-step.