Bug #2744
closedARP related problem since upgrading
0%
Description
I upgraded, by request of another issue, to the latest snapshot yesterday. We've had a strange issue arise.
The setup is such that the WAN interface is bridged with a "Firewall" interface such that devices on the "Firewall" interface can have firewall rules controlling their access. The subnet on WAN is 31.24.0.195/26. .195 is the assigned address. A device on the "Firewall" interface has 31.24.0.211/26. On resetting the networking for the .211 device, the pfsense router responds to an arp request for .195 with 00:90:7f:3f:cd:8f. This is the MAC address of the "Firewall" interface.
Some time later, the pfsense router requests the IP address of .211 from the MAC address of its WAN interface. This changes the MAC address of .195 in the arp table of .211 to 00:90:7f:3f:cd:91. This is the MAc address of the "WAN" interface.
This greatly confuses .211 and breaks internet connectivity.
Having discussed internally, I might indeed be doing it wrong in such that the IP should be assigned to an interface created from the BRIDGE0 "Network Port" but this is a setup that's worked without issue on stable and previous 2.1 snapshots so I wonder if there's a wider issue. Previous snapshot was compiled circa the end of October 2012.
I've attached a Wireshark compatible .cap of the ARPs. Suggested Wireshark filter is arp.dst.proto_ipv4 31.24.0.195 || arp.src.proto_ipv4 31.24.0.195
Files
Updated by Jim Pingle almost 12 years ago
- Status changed from New to Rejected
Most likely a duplicate of #729
Updated by Phil Lavin almost 12 years ago
That's annoying :P
Let me know if there's any debug you need from my side. I'll watch the original.
Updated by Phil Lavin almost 12 years ago
Actually, having reviewed that, it's quite old. Why would this issue just come to light now when the setup has worked for ~2 years and only after an upgrade to the latest snapshot?
Is there a workaround - I don't see a definitive one in the original bug.
Updated by Jim Pingle almost 12 years ago
Not sure, but it's a known issue. Continue discussion on the other bug report for best results.