http://linux-hotplug.sourceforge.net/?selected=net
says that for 'NET' events, 'register' and 'unregister' are the actions.
Starting ppp, I get this:
Jun 25 10:50:22 turing-police /etc/hotplug/net.agent: NET add event not supported
'NET add'?? WTF? ;)
(Fortunately, '/sbin/ifup ppp0' gets invoked anyhow, so it's not THAT crucial)
/Valdis (who still needs to fix hotplug not being called at all for the wireless card)
On Wed, 25 Jun 2003 16:28:28 -0400
[email protected] wrote:
>
> http://linux-hotplug.sourceforge.net/?selected=net
> says that for 'NET' events, 'register' and 'unregister' are the actions.
>
> Starting ppp, I get this:
>
> Jun 25 10:50:22 turing-police /etc/hotplug/net.agent: NET add event not supported
>
> 'NET add'?? WTF? ;)
>
> (Fortunately, '/sbin/ifup ppp0' gets invoked anyhow, so it's not THAT crucial)
>
> /Valdis (who still needs to fix hotplug not being called at all for the wireless card)
>
Look in the mailing list archives:
> rom: Stephen Hemminger <[email protected]>
> To: David S. Miller <[email protected]>, Greg KH <[email protected]>
> Cc: [email protected], [email protected]
> Subject: [PATCH] network hotplug via class_device/kobject
> Date: Fri, 13 Jun 2003 16:41:19 -0700
> Organization: Open Source Development Lab
> X-Mailer: Sylpheed version 0.8.11 (GTK+ 1.2.10; i686-pc-linux-gnu)
>
> This patch changes network devices to run hotplug out of the kobject/class_device
> infrastructure rather than calling it from the network core. The code gets simpler
> and there is only one place for Greg to fix when he changes the API ;-)
>
> All hotplug now happens off the chain:
> rtnl_unlock -> netdev_run_todo -> netdev_sysfs_{un}register
>
> The state flag "deadbeaf" was convertied to a state enumeration to handle the
> necessary book keeping, and adds some defense against drivers that have unexpected
> semantics. Paranoid about some driver doing something like:
> rtnl_lock(); register_netdevice(); unregister_netdevice(); rtnl_unlock() BOOM
>
> This patch causes an external script API change.
> Because network device go through the standard path, the action passed to the script
> is no longer register or unregister but is now "add" or "remove" like other devices.
> This is a good thing. When testing (at least on RHAT) just change /etc/hotplug/net.agent
> case statement:
>
> case $ACTION in
> add|register)
> # Don't do anything if the network is stopped
> if [ ! -f /var/lock/subsys/network ]; then
> exit 0
> fi
On Wed, 25 Jun 2003 14:58:35 PDT, Stephen Hemminger said:
> > This patch causes an external script API change.
> > Because network device go through the standard path, the action passed to t
he script
> > is no longer register or unregister but is now "add" or "remove" like other
devices.
> > This is a good thing. When testing (at least on RHAT) just change /etc/hot
plug/net.agent
> > case statement:
> >
> > case $ACTION in
> > add|register)
Yeah, that was the fairly obvious fix... now who's job is it to fix it on the
hotplug website, and when? Also, who gets to update the currently non-existent
entry in Documentation/Changes (i've cc'ed Chris Ricker)?
Or should this just get tacked on the "to do before 2.6.0 comes out" list for
now?
On Wed, Jun 25, 2003 at 11:07:43PM -0400, [email protected] wrote:
> > > case $ACTION in
> > > add|register)
>
> Yeah, that was the fairly obvious fix... now who's job is it to fix it on the
> hotplug website, and when?
That would probably be me :)
I'll try to get to it on Monday and get out a new linux-hotplug script
package at the same time.
thanks,
greg k-h