collisions on tun interfaces ...

[YongHyeon PYUN]

On Fri, Apr 29, 2011 at 11:28:07AM -0700, YongHyeon PYUN wrote:
> On Fri, Apr 29, 2011 at 12:52:31PM +0300, Zeus V Panchenko wrote:
> > Hi,
> > 
> > may somebody epxplain it for me, what can cause collisions on tun
> > interfaces created by ppp(8) and openvpn?
> > 
> > > uname -a
> > FreeBSD 8.2-STABLE #0 i386
> > 
> > > netstat -i
> > Name    Mtu Network       Address              Ipkts Ierrs Idrop    Opkts Oerrs  Coll
> > tun0   1492 <Link#7>                        18940349     0     0 15737760     0 45668
> > tun0   1492 A.B.C.D A-B-C-D.domain          15623965     -     - 12429351     -     -
> > tun1   1500 <Link#8>                        12721670     0     0  9957662     0 11161
> > tun1   1500 E.F.G.H E-F-G-H.vpn     	        6454     -     -   445751     -     -
> > 
> > 
> > rl0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
> >      options=3808<VLAN_MTU,WOL_UCAST,WOL_MCAST,WOL_MAGIC>
> >      ether 00:30:4f:67:cf:81
> >      media: Ethernet autoselect (100baseTX <full-duplex>)
> >      status: active
> > tun0: flags=8051<UP,POINTOPOINT,RUNNING,MULTICAST> metric 0 mtu 1492
> >       options=80000<LINKSTATE>
> >       inet A.B.C.D --> A1.B1.C1.D1 netmask 0xffffffff 
> >       Opened by PID 3943
> > tun1: flags=8051<UP,POINTOPOINT,RUNNING,MULTICAST> metric 0 mtu 1500
> >       options=80000<LINKSTATE>
> >       inet E.F.G.H --> E1.F1.G1.H1 netmask 0xffffffff 
> >       Opened by PID 1387
> > 
> > tun0 is created by ppp(8)
> > 
> > in /etc/ppp.conf is:
> > default:
> >  set log Phase Chat LCP IPCP CCP tun command
> >  set server /var/run/ppp/internet "" 0177
> >  set device PPPoE:rl0
> >  set speed sync
> >  enable lqr echo
> >  set lqrperiod 30
> >  set login
> >  set ctsrts off
> >  add default HISADDR
> >  set timeout 0
> >  set redial 0 0
> >  set cd 5
> > 
> > tun1 is created by OpenVPN with configuration:
> > client
> > dev tun1
> > proto udp
> > ...
> > 
> > so, what can cause the collisions and can i fix them?
> > 
> 
> It seems tun(4) just increments collision counter whenever it
> can't enqueue packet. Because it's not collision at all I think
> it's a bug that had been there from day 1. Just nuking updating the
> counter will address the issue. You still can get the previous
> collision counter of tun(4) with d option of netstat which shows
> number of packets dropped due to send queue full.

If there is no objection, I'll commit it in this week.