On Thu, 2 Dec 2004 14:13:33 +0100
ncrfgs <ncrfgs@xxxxxx> wrote:
> Hi,
>
> I'm using the great 802.1d Ethernet Bridging support via
> brctl along with Universal TUN/TAP devices created with
> openvpn on a GNU system with a vanilla 2.6.8.1 Linux
> kernel.
>
>
> To start it:
>
> /usr/sbin/brctl addbr br${X}
> /sbin/ifconfig brX xxx.xxx.xxx.xxx netmask yyy.yyy.yyy.yyy
> /sbin/route add default gw zzz.zzz.zzz.zzz
>
> /usr/sbin/brctl addif br${X} eth${Y}
> /sbin/ifconfig eth${Y} 0.0.0.0 promisc up
>
> [1] /usr/sbin/openvpn --mktun --dev tap${Z}
> /usr/sbin/brctl addif br${X} tap${Z}
> /sbin/ifconfig tap${Z} 0.0.0.0 promisc up
> [2] /usr/sbin/openvpn --config config.${Z}
>
>
> To stop it:
>
> [3] kill `cat /var/run/openvpn.${Z}.pid`
> /sbin/ifconfig tap${Z} down
> /usr/sbin/brctl delif br${X} tap${Z}
> /usr/sbin/openvpn --rmtun --dev tap${Z}
>
> /usr/sbin/brctl delif br${X} eth${Y}
> /sbin/ifconfig eth${Y} down
>
> /sbin/ifconfig br${X} down
> /usr/sbin/brctl delbr br${X}
>
>
> As far as I understood, if I create a tap device [1]
> before I start the openvpn instance that will use it [2],
> after I kill that instance [3], the tap device is still up
> and everything seems to go fine.
>
> In general, if I don't create the tap device explicitely
> [1] and I let the openvpn instance that will use it to
> create it implicitely [2], if I understood it well, the
> device will die along with the openvpn instance that
> created it [3].
>
> If this happen, and if the device was attached to a bridge
> device, once I kill the openvpn instance that implicitely
> created the tap device [3], the kernel seems to hang and
> the only thing I get is the following message on the
> current terminal.
>
> Debug: sleeping function called from invalid context at include/asm/semaphore.h:119
>
>
> Is it a bug?
Yes, but it could be in either the bridge or the tun/tap device.
What was teh backtrace.
