Russell King wrote:
>
> ferret@phonewave.net writes:
> > I installed BeOS 5 PE on the system, and that OS also drops the card to
> > 10bT during configuration.
>
> I've upgraded to a Cyclone card, and recently changed the hub to a 10/100
> switching version. One thing that I did notice is that the Cyclone card
> will not automatically re-negotiate its speed once its dropped into 10base
> mode.
hmm.. It Works For Me. Please send me some details (platform, kernel
version, etc).
> Unfortunately, the driver doesn't accept the ifconfig media command
> (SIOCSIFMAP, dev->set_config),
My, aren't there are lot of goodies in struct net_device!
There are a lot of these management things which we should be doing and
aren't:
- Ability to query interface capabilities
- Ability to vary media type/speed/duplex
- API to query media type/speed/duplex
- select()able ability to detect loss and
reestablishment of carrier.
- etc.
I'll be putting together some thoughts on this, discuss with the
high-availability projects, etc. But I expect it'll end up bigger than
Ben Hur and right now the emphasis is on stabilising the current stuff.
> so if the driver is a module, you need
> to remove/reload it. If its in the kernel, you need to reboot the
> machine.
Sounds familiar :)
Have you tried setting the media with Don's mii-diag tool?
http://www.scyld.com/diag/
mii-diag uses the generic device-private MII ioctl().
> It would be nice to be able to configure the media type with ifconfig.
> (maybe this is a feature addition in 2.3 that got overlooked by a lot
> of the net drivers and no one knows about it?)
That'll be correct. Alexey snuck a bunch of good stuff into the
netdevice layer and we're only slowly discovering it...
The more general problem is that there has been a maintenance diaspora
with many of the net drivers. The most popular drivers are cared for,
but there are many less popular ones which are a bit moldy. The
intersection between those who have the hardware and those who care to
maintain the driver is a rather sparse set. Jeff is working this stuff
quite hard.
For 2.5 we should look at filling out the driver features and making
them more uniform. Of course, by then everyone will be using gigE :(
-- -akpm-- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon May 15 2000 - 21:00:20 EST