misc/103761: some tun interfaces with a mtu of 1500 while i should
never exceed 1472 with ppp
dblais at interplex.ca
Thu Sep 28 07:50:48 PDT 2006
>Synopsis: some tun interfaces with a mtu of 1500 while i should never exceed 1472 with ppp
>Arrival-Date: Thu Sep 28 14:50:20 GMT 2006
>Originator: Dominic Blais
>Release: FreeBSD RELEASE-6.3-p7
FreeBSD chapdelaine.ipx.gw.interplex.ca 6.1-RELEASE-p7 FreeBSD 6.1-RELEASE-p7 #0: Thu Sep 21 16:11:35 EDT 2006 admin at chapdelaine.rt.interplex.ca:/usr/src/sys/i386/compile/CHAPDELAINE i386
We have multiple FreeBSD systems across our network acting as PPPoE server for Internet access to our customers. All these systems have this configuration included in ppp.conf:
set mru max 1472
set mtu max 1472
On every systems but one the mtu never exceeds 1472. It seems that the ppp processes associated with the tun interfaces with an mtu of 1500 are freezed. I mean, they never stop and are avoid limited IP addresses from being used again by new pppoe authentication.
tun87: flags=8051<UP,POINTOPOINT,RUNNING,MULTICAST> mtu 1472
inet 10.0.0.1 --> 10.0.0.254 netmask 0xffffffff
Opened by PID 57916
tun88: flags=8051<UP,POINTOPOINT,RUNNING,MULTICAST> mtu 1472
inet 10.0.0.1 --> 10.0.0.220 netmask 0xffffffff
Opened by PID 76102
tun89: flags=8051<UP,POINTOPOINT,RUNNING,MULTICAST> mtu 1472
inet 10.0.0.1 --> 10.0.0.215 netmask 0xffffffff
Opened by PID 76229
tun90: flags=8010<POINTOPOINT,MULTICAST> mtu 1500
tun91: flags=8051<UP,POINTOPOINT,RUNNING,MULTICAST> mtu 1500
inet 10.0.0.1 --> 10.0.0.232 netmask 0xffffffff
Opened by PID 58042
The interface tun91 is locked by the ppp process with the 58042 PID. It will never be freed if I'm not killing it.
Since it's happening on only one of our servers, my guess is that it is triggered by something received on this specific network.
I made a script which kills these processes so that it can free the locked IP addresses but that's an ugly workaround.
Start a pppoe server on my network ;) I'm kidding but as I said, it only happen on one of our networks even thought we're using FreeBSD 6.1-p7 on each of them.
More information about the freebsd-bugs