Re: Interface dependencies

On Sat, Jun 27, 2009 at 07:48:54PM -0800, Mel Flynn wrote:
> Hi,
> 
> maybe I'm overlooking something, so I thought I'd ask.
> As far as I can tell, there is no way to specify interface dependencies, so I 
> have an issue I cannot seem to solve:
> - Create a lagg0 that has em and wlan0 at boot time, because wlan0 takes too 
> long to be configured - and the default network_interfaces=AUTO sorts 
> alphabetically which is not making matters easier.

The interfaces should be in the order they are probed/created.  No
sorting should be performed beyond moving lo0 to the front in the
default case.

> I've been trying to use hacks, but I think interfaces really need 
> dependencies. Like ifconfig_lagg0_require="wlan0 em0", which would first 
> configure wlan0, wait for it to be availabe, then em0 and finally lagg0.
> 
> Is there something available, is it a known issue and ENOTIME to fix or am I 
> missing something else?

There isn't a feature to add dependencies, but there probably should be.
I'm not sure that's really what the problem is there though.

> At present, my rc.conf entries are:
> # Need to do this manually to prevent alphabetical sorting.
> network_interfaces="wpi0 lo0 em0"
> cloned_interfaces="lagg0"
> wlans_wpi0="wlan0"
> ifconfig_wpi0="ether 00:16:36:f2:3b:84"
> ifconfig_wlan0="WPA"
> ifconfig_em0="up"
> ifconfig_lagg0="laggproto failover laggport em0"
> ifconfig_lagg0_alias0="laggport wlan0"
> ifconfig_lagg0_alias1="inet 192.168.2.50 netmask 255.255.255.0"
> 
> And this gives me a lagg0 at boottime without wlan0, since the interface don't 
> exist. I also cannot add inet commands to laggport commands, thus the alias 
> trick is already needed, yet the delay caused by running separate commands 
> does not seem to be enough to have wlan0 available.

wlan0 should exist by the time lagg0 is created because it's created and
configured synchronously when wpi0 is configured.   I know other people
are using lagg this way so I'm a bit confused as to what's wrong.
Enabling verbose start up and examining the output might be telling.

-- Brooks

> -- 
> Mel
> _______________________________________________
> freebsd-current_at_freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
>