Re: [NetBSD-7] Panic in nfsd

[BERTRAND Joël <joel.bertrand%systella.fr@localhost>]

Christos Zoulas a écrit :
> On Jan 8,  2:37pm, joel.bertrand%systella.fr@localhost (=?UTF-8?Q?BERTRAND_Jo=c3=abl?=) wrote:
> -- Subject: Re: [NetBSD-7] Panic in nfsd
>
> | Christos Zoulas a écrit :
> | > On Jan 8,  2:11pm, joel.bertrand%systella.fr@localhost (=?UTF-8?Q?BERTRAND_Jo=c3=abl?=) wrote:
> | > -- Subject: Re: [NetBSD-7] Panic in nfsd
> | >
> | > | 	I don't know. A lot of lines have been printed after lines you are
> | > | waiting for and usable information was not available on console...
> | >
> | > Can you put a Debugger() line if the altq releases cause errors after the
> | > prints?
> | >
> | > christos
> | >
> |
> | 	I can test but with some difficulties as I don't have a workstation to
> | test. I can only do that on a running server...
>
> Can you provide a sample configuration that causes the crash?

	I can.

	I have configured an OpenVPN server (over UDP, tap interface, not tun) 
on a Linux server (debian/testing up to date). This server is connected 
to internet by a VDSL2 link.

	My NetBSD server (netbsd-7 CVS up to date) is connected to internet by 
a WIMAX link and runs OpenVPN client built from pkgsrc. On both sides, I 
have installed QoS (iptables + tc on Linux and altqd on NetBSD) as I 
have to route VoIP over this VPN. Since a few days, my VDSL2 link is 
unstable and tap0 is detached on NetBSD (green line on console). When 
tap0 is detached, NetBSD kernel can panic.

	My altqd.conf is :

legendre$ cat /etc/altq.conf
interface wm1 bandwidth 10M priq

class priq wm1 high_class NULL priority 1
class priq wm1 low_class NULL priority 0 default

filter wm1 high_class 0 1194 0 0 17
filter wm1 high_class 0 0 0 1194 17

interface tap0 bandwidth 2M priq

class priq tap0 high_class_vpn NULL priority 1
class priq tap0 low_class_vpn NULL priority 0 default

filter tap0 high_class_vpn 192.168.10.250 0 0 0 17
filter tap0 high_class_vpn 0 0 192.168.10.250 0 17

interface wm0 bandwidth 1000M priq

class priq wm0 high_class_lan NULL priority 1
class priq wm0 low_class_lan NULL priority 0 default

filter wm0 high_class_lan 192.168.10.250 0 0 0 17
filter wm0 high_class_lan 0 0 192.168.10.250 0 17

	altqd is not started by rc.conf but is controlled by OpenVPN.	I have 
written two scripts :

legendre$ cat /usr/pkg/etc/openvpn/route-up.sh
#!/bin/sh
/sbin/route add -net 192.168.0.0/24 192.168.1.1
/sbin/route add -net 192.168.253.0/24 192.168.1.1
/sbin/route add -net 192.168.254.0/24 192.168.1.1
/sbin/ifconfig tap0 inet6 2001:7a8:a8ed:1::2/64
/sbin/route add -inet6 default 2001:7a8:a8ed:1::1
/etc/rc.d/altqd onestart
exit 0

and

legendre$ cat /usr/pkg/etc/openvpn/route-down.sh
#!/bin/sh
/etc/rc.d/altqd onestop
/sbin/ifconfig tap0 inet6 delete 2001:7a8:a8ed:1::2/64
/sbin/route delete 192.168.0.0/24
/sbin/route delete 192.168.253.0/24
/sbin/route delete 192.168.254.0/24
/sbin/route delete -inet6 default
exit 0

	Thus, I suppose you have to install an OpenVPN server somewhere and to 
run altqd on OpenVPN client interface. If you stop OpenVPN server, 
client should panic.

	Best regards,

	JKB