RFC: auto eth0 and auto eth1 in netbase

RFC: auto eth0 and auto eth1 in netbase

[Rod Whitby]

Some machines don't have consoles, so the only way to access them on the
first boot after a reflash is via the network.

Of course, an 'auto' keyword is required for this, and netbase currently
has none.

Is there any reason why at least eth0 should not have an 'auto' keyword?

Should this be OE-wide, only for Angstrom (koen thinks eth0 and eth1
should be auto), or only for specific machines (in which case, does
anyone object to ixp4xx having auto eth0)?

-- Rod

Re: RFC: auto eth0 and auto eth1 in netbase

[Graeme Gregory]

On Thu, 29 Nov 2007 22:23:54 +1030
Rod Whitby <rod@whitby.id.au> wrote:

> Some machines don't have consoles, so the only way to access them on
> the first boot after a reflash is via the network.
> 
> Of course, an 'auto' keyword is required for this, and netbase
> currently has none.
> 
> Is there any reason why at least eth0 should not have an 'auto'
> keyword?
> 
> Should this be OE-wide, only for Angstrom (koen thinks eth0 and eth1
> should be auto), or only for specific machines (in which case, does
> anyone object to ixp4xx having auto eth0)?
> 

It is quiet common to find commercial equipment doing this in the real
world. You just tell your dhcp server the MAC and you can find it on
first boot easilly.

So I think we should probably allow this as well.

+1 for me on eth0

G

Re: RFC: auto eth0 and auto eth1 in netbase

[mwester]

> Some machines don't have consoles, so the only way to access them on the
> first boot after a reflash is via the network.
> 
> Of course, an 'auto' keyword is required for this, and netbase currently
> has none.
> 
> Is there any reason why at least eth0 should not have an 'auto' keyword?
> 
> Should this be OE-wide, only for Angstrom (koen thinks eth0 and eth1
> should be auto), or only for specific machines (in which case, does
> anyone object to ixp4xx having auto eth0)?

I think the "principle of least surprise" would favor "auto eth0".  Most users expect that if it has a network port, regardless of what type of equipment "it" might be, it should make an attempt to join the network.

+1 for me.

Mike (mwester)

Re: RFC: auto eth0 and auto eth1 in netbase

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Rod Whitby schreef:
> Some machines don't have consoles, so the only way to access them on the
> first boot after a reflash is via the network.
> 
> Of course, an 'auto' keyword is required for this, and netbase currently
> has none.
> 
> Is there any reason why at least eth0 should not have an 'auto' keyword?
> 
> Should this be OE-wide, only for Angstrom (koen thinks eth0 and eth1
> should be auto), or only for specific machines (in which case, does
> anyone object to ixp4xx having auto eth0)?

+1 for OE wide


- --
koen@dominion.kabel.utwente.nl will go go away in december 2007, please
use k.kooi@student.utwente.nl instead.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFHTs+aMkyGM64RGpERAtsvAJ9C+VMaervzVWZLtxXU6jKp7BTPyACcDAdM
fjkhfc+Tflx0X5v42JKwjxY=
=bBvk
-----END PGP SIGNATURE-----

Re: RFC: auto eth0 and auto eth1 in netbase

[Chris Larson]

On Nov 29, 2007 7:41 AM, Koen Kooi <k.kooi@student.utwente.nl> wrote:
> Rod Whitby schreef:
> > Some machines don't have consoles, so the only way to access them on the
> > first boot after a reflash is via the network.
> >
> > Of course, an 'auto' keyword is required for this, and netbase currently
> > has none.
> >
> > Is there any reason why at least eth0 should not have an 'auto' keyword?
> >
> > Should this be OE-wide, only for Angstrom (koen thinks eth0 and eth1
> > should be auto), or only for specific machines (in which case, does
> > anyone object to ixp4xx having auto eth0)?
>
> +1 for OE wide

Afaik, the execution of the hotplug/coldplug scripts on startup should
generate the hotplug events for the ethernet interfaces and bring them
up that way.  In debian, the hotplug scripts bring them up with ifup
--allow=hotplug [iface], such that you need an 'allow-hotplug eth0' in
interfaces to have them brought up that way.  I'm too rusty with OE to
say how we do it today, but it's not likely that auto is needed.
-- 
Chris Larson - clarson at kergoth dot com
Dedicated Engineer - MontaVista - clarson at mvista dot com
Core Developer/Architect - TSLib, BitBake, OpenEmbedded, OpenZaurus

Re: RFC: auto eth0 and auto eth1 in netbase

[Rod Whitby]

Chris Larson wrote:
> Afaik, the execution of the hotplug/coldplug scripts on startup should
> generate the hotplug events for the ethernet interfaces and bring them
> up that way.  In debian, the hotplug scripts bring them up with ifup
> --allow=hotplug [iface], such that you need an 'allow-hotplug eth0' in
> interfaces to have them brought up that way.  I'm too rusty with OE to
> say how we do it today, but it's not likely that auto is needed.

Unfortunately, the busybox ifup applet does not support the
allow-hotplug keyword.

Based on the feedback, I'll add 'auto eth0'.

-- Rod

Re: RFC: auto eth0 and auto eth1 in netbase

[Chris Larson]

On Nov 30, 2007 4:48 AM, Rod Whitby <rod@whitby.id.au> wrote:
> Chris Larson wrote:
> > Afaik, the execution of the hotplug/coldplug scripts on startup should
> > generate the hotplug events for the ethernet interfaces and bring them
> > up that way.  In debian, the hotplug scripts bring them up with ifup
> > --allow=hotplug [iface], such that you need an 'allow-hotplug eth0' in
> > interfaces to have them brought up that way.  I'm too rusty with OE to
> > say how we do it today, but it's not likely that auto is needed.
>
> Unfortunately, the busybox ifup applet does not support the
> allow-hotplug keyword.
>
> Based on the feedback, I'll add 'auto eth0'.

Network interfaces should still be brought up when hotplugged.
Lacking the keyword just means there's no way to prevent it from
happening via interfaces(5).
-- 
Chris Larson - clarson at kergoth dot com
Dedicated Engineer - MontaVista - clarson at mvista dot com
Core Developer/Architect - TSLib, BitBake, OpenEmbedded, OpenZaurus

Re: RFC: auto eth0 and auto eth1 in netbase

[Khem Raj]

On Nov 30, 2007 8:59 AM, Chris Larson <clarson@kergoth.com> wrote:
> On Nov 30, 2007 4:48 AM, Rod Whitby <rod@whitby.id.au> wrote:
> > Chris Larson wrote:
> > > Afaik, the execution of the hotplug/coldplug scripts on startup should
> > > generate the hotplug events for the ethernet interfaces and bring them
> > > up that way.  In debian, the hotplug scripts bring them up with ifup
> > > --allow=hotplug [iface], such that you need an 'allow-hotplug eth0' in
> > > interfaces to have them brought up that way.  I'm too rusty with OE to
> > > say how we do it today, but it's not likely that auto is needed.
> >
> > Unfortunately, the busybox ifup applet does not support the
> > allow-hotplug keyword.
> >
> > Based on the feedback, I'll add 'auto eth0'.
>

fwiw with this change my nfsrooted rfs does not boot. I think its a
nice to have nfsroot option for dev work.

Re: RFC: auto eth0 and auto eth1 in netbase

[Rod Whitby]

Khem Raj wrote:
>> On Nov 30, 2007 4:48 AM, Rod Whitby <rod@whitby.id.au> wrote:
>>> Based on the feedback, I'll add 'auto eth0'.
> 
> fwiw with this change my nfsrooted rfs does not boot. I think its a
> nice to have nfsroot option for dev work.

Do you have any more details on why it doesn't work, and do you have a
recommendation on how to meet both requirements?

(I don't use nfsroot with OE myself, so you'll need to give more details
and help debug why it doesn't work)

-- Rod

Re: RFC: auto eth0 and auto eth1 in netbase

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Khem Raj schreef:
> On Nov 30, 2007 8:59 AM, Chris Larson <clarson@kergoth.com> wrote:
>> On Nov 30, 2007 4:48 AM, Rod Whitby <rod@whitby.id.au> wrote:
>>> Chris Larson wrote:
>>>> Afaik, the execution of the hotplug/coldplug scripts on startup should
>>>> generate the hotplug events for the ethernet interfaces and bring them
>>>> up that way.  In debian, the hotplug scripts bring them up with ifup
>>>> --allow=hotplug [iface], such that you need an 'allow-hotplug eth0' in
>>>> interfaces to have them brought up that way.  I'm too rusty with OE to
>>>> say how we do it today, but it's not likely that auto is needed.
>>> Unfortunately, the busybox ifup applet does not support the
>>> allow-hotplug keyword.
>>>
>>> Based on the feedback, I'll add 'auto eth0'.
> 
> fwiw with this change my nfsrooted rfs does not boot. I think its a
> nice to have nfsroot option for dev work.

doing any form of 'ifup' on the interface that connects to you nfsroot
with shutdown the interface before trying to bring it up. I have the
same problem on my PVR, and as a workaround I edited
/path/to/nfs/etc/network/interface on the nfs server and removed the auto.
Volunteers to teach ifup not to not close the link when using nfsroot?

regards,

Koen

- --
koen@dominion.kabel.utwente.nl will go go away in december 2007, please
use k.kooi@student.utwente.nl instead.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFHWcGiMkyGM64RGpERAiAnAJ0ZHNyCyLjmRlRjXnIi9bWxfWIUGACfTwip
qPwF61BdfukK/IxyHKWNJ5o=
=lxi4
-----END PGP SIGNATURE-----

Re: RFC: auto eth0 and auto eth1 in netbase

[Paul Sokolovsky]

Hello Koen,

Friday, December 7, 2007, 11:56:50 PM, you wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1

> Khem Raj schreef:
>> On Nov 30, 2007 8:59 AM, Chris Larson <clarson@kergoth.com> wrote:
>>> On Nov 30, 2007 4:48 AM, Rod Whitby <rod@whitby.id.au> wrote:
>>>> Chris Larson wrote:
>>>>> Afaik, the execution of the hotplug/coldplug scripts on startup should
>>>>> generate the hotplug events for the ethernet interfaces and bring them
>>>>> up that way.  In debian, the hotplug scripts bring them up with ifup
>>>>> --allow=hotplug [iface], such that you need an 'allow-hotplug eth0' in
>>>>> interfaces to have them brought up that way.  I'm too rusty with OE to
>>>>> say how we do it today, but it's not likely that auto is needed.
>>>> Unfortunately, the busybox ifup applet does not support the
>>>> allow-hotplug keyword.
>>>>
>>>> Based on the feedback, I'll add 'auto eth0'.
>> 
>> fwiw with this change my nfsrooted rfs does not boot. I think its a
>> nice to have nfsroot option for dev work.

> doing any form of 'ifup' on the interface that connects to you nfsroot
> with shutdown the interface before trying to bring it up. I have the
> same problem on my PVR, and as a workaround I edited
> /path/to/nfs/etc/network/interface on the nfs server and removed the auto.
> Volunteers to teach ifup not to not close the link when using nfsroot?

  I've been having this issue in my queue. There're few problems with
current OE init scripts and nfsroot. I'm going to get to this one
sweet day, but if anyone gets to this earlier, even better ;-).

> regards,

> Koen




-- 
Best regards,
 Paul                            mailto:pmiscml@gmail.com