DHCP v6 - no route to host

MrKoen · May 10, 2011, 5:20 PM

Might be work in progress, but works very well already :) I found out that deleting the existing /var/dhcpd/var/db/dhcpd6.leases file on my system and restarting the dhcpd service solved the problem of the dhcpd service only staying active for a few requests. It did give my quite some errors after restarting the dhcpd service the first time, but now it's up and running for two days already and still providing IPv6 addresses according to both the pool addresses and the configured reservations. So its almost perfect now.

I'm still wondering how I can solve the problem where the clients assign a /128 route to the assigned IPv6 address thus sending all IPv6 traffic via the pfSense firewall which is the default gateway. For IPv6 traffic on my local LAN this causes a lot of problems. Remote desktops from one machine to another for example will go over IPv6 by default but pfSense will drop a few packets every now and then causing the remote desktop to freeze. These dropped packets are logged by pfSense. When connecting to the same client using its local IPv4 address, it works well. If I could just get it to assign a /64 in the routing table on the client the problem would be solved. Anyone got a clue how to get this done?

databeestje · May 11, 2011, 8:48 AM

https://lists.isc.org/pipermail/dhcp-users/2010-January/010810.html

MrKoen · May 11, 2011, 8:52 AM

Sounds like exactly the problem I was facing. Same errors and outcome. As stated above, for me it was solved by deleting the leases file and having dhcpd create a new one. Its still happily providing IPv6 leases now after being up for 2,5 day already.

databeestje · May 11, 2011, 9:01 AM

Currently importing isc dhcpd 4.2.1-P1

MrKoen · May 11, 2011, 9:02 AM

Cool :) Do you have a clue how I could have the DHCPv6 leases set to /64 in the clients routing table?

databeestje · May 11, 2011, 4:01 PM

not yet, open for options. new snapshots on my site with the new dhcpd

MrKoen · May 11, 2011, 7:16 PM

Hmm.. isn't this a thing that rtadvd should provide? My knowledge about IPv6 isn't sufficient to know if router advertisements should provide this, but it sounds logical. I'll do some searches on the internet to find out more about this.

MrKoen · May 11, 2011, 7:49 PM

I found a statement in RFC4862 on page 8 which to me sounds like the router advertisement should indeed provide the cidr block by providing an address prefix. I quote:

Router Advertisements also contain zero or more Prefix Information
options that contain information used by stateless address
autoconfiguration to generate global addresses. It should be noted
that a host may use both stateless address autoconfiguration and
DHCPv6 simultaneously. One Prefix Information option field, the
"autonomous address-configuration flag", indicates whether or not the
option even applies to stateless autoconfiguration. If it does,
additional option fields contain a subnet prefix, together with
lifetime values, indicating how long addresses created from the
prefix remain preferred and valid.

I remember seeing the cidr notation in the rta config before. I'll check out the rtadvd man pages for more info.

MrKoen · May 11, 2011, 8:10 PM

I see you already got this stuff included in the /var/etc/rtadvd.conf file and it looks fine. Hmm. What could it be.

databeestje · May 11, 2011, 9:47 PM

it appears to be a issue with a combination of the dhcp server and the client.

I've upgraded the dhcp server to 4.2, that should fix the issue with the dhcp leases file. Maybe i'm missing a option I need to send that carries the prefix length.

More work soon, I have a family issue that I find far more important then IPv6 atm.

MrKoen · May 12, 2011, 6:39 AM

Without a doubt Databeestje. Good luck solving the family issue.

I'll play with it some more to see if I can perhaps find something. I will share any possible findings in this topic.

MrKoen · May 15, 2011, 8:01 AM

I just updated my pfSense to the latest AMD64 release and gitsynced with the latest Smos updates. I now notice that my interfaces having the DHCPv6 service enabled (router advertisements set to Assisted - if this matters) will first at boot have their static IPv6 addresses assigned, but seconds after done booting they will acquire an IPv6 lease from the DHCPv6 service themselves. Interfaces not having the DHCPv6 service enabled stay with their configured static leases. To correct this, I go go the interface configuration page, don't change anything, press Save and after that Apply changes and the static IPv6 address is back again. I guess this is a thing for the todo list whenever you find time for it.

I'll continue to see if I can find something to change the /128 cidr notation on provided IPv6 leases.

databeestje · May 15, 2011, 9:29 AM

if there is 1 interface that has dhcp enabled I toggle the allow router advertisements sysctl. Problem is that it is not a per interface, but a global.

Thus pfSense receives it's own advertisement. That wasn't supposed to happen. There should have been a few new arguments I could pass to ifconfig to enable or disable those.

Needs more work.

MrKoen · May 15, 2011, 1:41 PM

I think I might have found a solution to the /128 problem on DHCPv6 leases. If you remove the following parameter from the /var/etc/rtadvd.conf config file, it does work as expected and it does register a /64 route on the provided DHCPv6 address:

:pinfoflags#64

My working config:

common definitions.

default:
:raflags#0:rltime#3600:
:vltime#360000:pltime#360000:mtu#1500:
ether:
:mtu#1280:tc=default:

Generated for DHCPv6 Server lan

em0:
:addr="2001:470:1:1:0:0:0:0":
:prefixlen#64:
:raflags="mo":
:tc=ether:

Generated for DHCPv6 Server opt1

em0_vlan3:
:addr="2001:470:1:2:0:0:0:0":
:prefixlen#64:
:raflags="mo":
:tc=ether:

With this config, I do get assigned an address in the 2001:470:1:1/64 range on my em0 interface and an address in the 2001:470:1:2/64 range on my em0_vlan3 interface. Clients on both interfaces now do register a /64 route in the routing table.

/edit: some additional info

I test this by going to the console and executing the "killall -9 rtadvd" command. Next I edit the /var/etc/rtadvd.conf file taking out the :pinfoflags#64 and leave all else as is. Then on the console I run "rtadvd -d -D -c /var/etc/rtadvd.conf em0 em0_vlan3". On my Windows 7 client I run the following command "ipconfig /release6 | ipconfig /renew6" and after that I run "route print". I now see the default gateway being registered at its fe80 link local address and an On-link registration for 2001:470:1:1::/64.

:D

MrKoen · May 25, 2011, 1:50 PM

Just gitsynced with the new gitsync hub. I notice the DHCPv6 service is no longer working. In the DHCP logs I find many of these errors:

dhcpd: dhcpv6: send_packet6() sent -1 of 120 bytes
dhcpd: send_packet6: Operation not permitted

Any clues to what could be wrong this time?

wallabybob · May 25, 2011, 10:53 PM

The interface dhcpd is using has a link local (fe80::…) IPv6 address? http://forum.pfsense.org/index.php/topic,37079.0.html might be relevant.

MrKoen · May 26, 2011, 9:04 AM

Yes it is. This is my ifconfig output:

ifconfig em0
em0: flags=8843 <up,broadcast,running,simplex,multicast>metric 0 mtu 1500
options=1209b <rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum,wol_magic,vlan_hwfilter>ether 00:ff:ff:51:00:01
inet6 fe80::225:90ff:fe34:bfae%em0 prefixlen 64 scopeid 0x1
inet 192.168.xxx.1 netmask 0xffffff00 broadcast 192.168.51.255
inet6 2001:470:xxx:xxx::1 prefixlen 64
nd6 options=3 <performnud,accept_rtadv>media: Ethernet autoselect (1000baseT <full-duplex>)
status: active</full-duplex></performnud,accept_rtadv></rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum,wol_magic,vlan_hwfilter></up,broadcast,running,simplex,multicast>

I gitsynced against the latest version, so what you're describing in your topic should also be included with that version. I finally had the DHCPv6 thing to work and now its broken again. Humpf :) Any other clues to what it could be? The RTADVD announcements work well. When the machine times out querying for a DHCPv6 address, it will auto generate one in the proper subnet with the proper route and proper gateway.

wallabybob · May 26, 2011, 11:46 AM

@Koen:

Yes it is. This is my ifconfig output:

ifconfig em0
em0: flags=8843 <up,broadcast,running,simplex,multicast>metric 0 mtu 1500
options=1209b <rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum,wol_magic,vlan_hwfilter>ether 00:ff:ff:51:00:01
inet6 fe80::225:90ff:fe34:bfae%em0 prefixlen 64 scopeid 0x1
inet 192.168.xxx.1 netmask 0xffffff00 broadcast 192.168.51.255
inet6 2001:470:xxx:xxx::1 prefixlen 64
nd6 options=3 <performnud,accept_rtadv>media: Ethernet autoselect (1000baseT <full-duplex>)
status: active</full-duplex></performnud,accept_rtadv></rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum,wol_magic,vlan_hwfilter></up,broadcast,running,simplex,multicast>

I gitsynced against the latest version, so what you're describing in your topic should also be included with that version.

Yes, but that should have been relevant only on bridges because they don't automatically create an IPv6 link local address from their generated MAC address. (My LAN interface is bridge0).

I notice that your em0 interface has a link local IPv6 address that seems totally unrelated to the reported MAC address. This doesn't seem right but I don't know if it is relevant.

On my system vr0 is a member of bridge0, the LAN interface. Notice the relationship between MAC address and IPv6 link local address:

ifconfig vr0

vr0: flags=8943 <up,broadcast,running,promisc,simplex,multicast>metric 0 mtu 1500
options=82808 <vlan_mtu,wol_ucast,wol_magic,linkstate>ether 00:30:18:b0:19:48
inet6 fe80::230:18ff:feb0:1948%vr0 prefixlen 64 scopeid 0x3
nd6 options=3 <performnud,accept_rtadv>media: Ethernet autoselect (100baseTX <full-duplex>)
status: active
#</full-duplex></performnud,accept_rtadv></vlan_mtu,wol_ucast,wol_magic,linkstate></up,broadcast,running,promisc,simplex,multicast>

MrKoen · May 26, 2011, 12:20 PM

This might be because I have spoofed my WAN side MAC address. The "official" MAC address can be noted in the fe80 address. It has been like that for a while already, so don't think that will be the problem.

Does the DHCPv6 service work correctly at your installation? Would you mind comparing settings to see if something is different there?

My DHCPv6 settings are (/services_dhcpv6.php):

Router Advertisements: Assisted
Enable DHCPV6 server on LAN interface: Checked
Deny unknown clients: Not checked
Subnet: 2001:470:xxx:xxx:0:0:0:0
Subnet mask: 64 bits
Available range: 2001:470:xxx:xxx:0:0:0:1 - 2001:470:xxx:xxx:ffff:ffff:ffff:ffff
Range: 2001:470:xxx:xxx::1000 to 2001:470:xxx:xxx::1050
Prefix Delegation Range: <empty>to <empty>prefix length 48
DNS servers: 2001:470:xxx:xxx::20 and 2001:1af8:xxx:xxx:2::1001
Gateway: 2001:470:xxx:xxx::1
Domain name: home.networks.zomers.eu
Domain search list: <empty>Default lease time: <empty>Maximum lease time: <empty>Failover peer IP: <empty>All of the Advanced options are also emtpy.
No additional BOOTP/DHCP options.

I do use static DUID/IPv6 mappings.</empty></empty></empty></empty></empty></empty>

MrKoen · May 27, 2011, 11:35 AM

I have found the problem! I accidentally added the fe80::/16 range to the EasyRuleBlockHostsLAN alias to avoid my pfSense logs being swamped with local traffic. This caused these packets to be blocked thus the errors to be generated. I have removed the fe80::/16 range from the alias and the problem went away. DHCPv6 and static mappings are working smoothly now :D