Fw: DHCPv6 PD Duplicate Allocation issue

[Jonas Bergler]

Hi List,

Before I report a bug, I thought I'd ask if anyone has had similar issues. We’re running some testing of IPoE DHCPv6 subscribers and have discovered some strange behavior with the ISC DHCP implementation.

It appears that when we make concurrent requests to binds subscribers in a pool the same PD prefix is selected for multiple clients. I have built 4.2.3-P2 and tested against this.

Two subscribers try to bind to the DHCPv6 Relay server simultaneously.

Relay-forward message from 2200:931::2 port 547, link address 2200:931::2, peer address fe80::6:8
Picking pool prefix 2200:931:ffff:ffff::/64
Sending Relay-reply to 2200:931::2 port 547

Relay-forward message from 2200:931::2 port 547, link address 2200:931::2, peer address fe80::6:9
Picking pool prefix 2200:931:ffff:ffff::/64
Sending Relay-reply to 2200:931::2 port 547

Relay-forward message from 2200:931::2 port 547, link address 2200:931::2, peer address fe80::6:8
Sending Relay-reply to 2200:931::2 port 547

One Client has successfully bound – but the DHCPv6 relay server doesn’t like the same prefix being allocated to both subscribers.
If we retry the second client again, we see that a different prefix is offered and all is well.

Relay-forward message from 2200:931::2 port 547, link address 2200:931::2, peer address fe80::6:9
Picking pool prefix 2200:931:ffff:fffe::/64
Sending Relay-reply to 2200:931::2 port 547

Relay-forward message from 2200:931::2 port 547, link address 2200:931::2, peer address fe80::6:9
Sending Relay-reply to 2200:931::2 port 547


Cheers, Jonas

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/dhcp-users/attachments/20120206/20d7c284/attachment.html>