 1 edit | [HELP] Usage of IP Verify Unicast cmdCurrently using the following config for the interface facing my ISP, and want to use URPF as follows, in particular to simplfy my inbound ACLs to drop RFC1918 addresses.
R1811#sh run int fa0
Building configuration...
Current configuration : 402 bytes
!
interface FastEthernet0
description "WAN - ISP"
ip address dhcp
ip access-group INBOUND_ACL in
ip verify unicast source reachable-via rx allow-default
no ip redirects
no ip unreachables
no ip proxy-arp
ip nat outside
ip inspect INBOUND_FW in
ip inspect OUTBOUND_FW out
ip virtual-reassembly max-fragments 3 max-reassemblies 64 timeout 1
duplex auto
speed auto
no cdp enable
end
I checked the Cisco docs, in particular this one, to make sure the router would accept a DHCP replies from my ISP. Since then, I've had a couple reloads that forced the router to request a DHCP transaction and in all cases, I can confirm that the router is accepting the 255.255.255.255 broadcasts, but until I remove the "ip verify unicast" command, the FastEthernet interface to the ISP never accepts the DHCP assignment.
Am I missing something here?
Regards |
|
 nosx join:2004-12-27 00000 kudos:5 | Several thoughts, first is the simplest: How are you learning the default? Do you have a 0/0 static --> modem or are you not learning default until DHCP is complete? Are the DHCP replies coming from the ISPs DHCP server? Can you slap a sniffer on the interface and wireshark the DHCP exchange to see what the deal is for me? Hit me up on skype if you have a PCAP file for me to peek at =) |
|
 cramer join:2007-04-10 Raleigh, NC kudos:5 Reviews:
·AT&T Southeast
| reply to HELLFIRE I've found uRPF to be rather useless on CPE routers. You've noticed the first problem... preventing dhcp from working. Ultimately, "allow-default" renders the technology moot as it will match anything that isn't local.
That said, to get the RFC1918 blocks filtered without putting them in an ACL, you have to have null routes to those blocks. uRPF would then block them inbound; the route entires themselves would block them outbound (which means you really don't need to filter them inbound.) And, of course, this will prevent DHCP from working with many ISPs since the reply often comes from the range you're blocking. I recall a configuration option to fix that, but I don't have it in any of my configs. (having long ago turned off uRPF) |
|
|
|
 aryobaPremium,MVM join:2002-08-22 kudos:1 | reply to HELLFIRE The ip verify unicast command basically verifies the routes according to routing table. As example, the command denies attempts to have router interfaces IP address to be 0.0.0.0 since it is considered invalid. Unfortunately with DHCP, such consideration is breaking DHCP mechanism since DHCP requires the router interface that acts as DHCP client to have 0.0.0.0 IP address before DHCP server approves the DHCP request to provide actual working IP address. In other words, DHCP client router and ip verify unicast command do not mix. |
|
 | reply to HELLFIRE
Thanks for the replies all. Since the last reboot, I've stripped the command out completely and am not planning on turning it back on as it seems to cause more issues than it fixes.
@nosx The 1811 fe0's interface is wired into the ISP's DOCSIS modem direct.
FE0 reports as up but in a "address will be assigned via DHCP" state, yet my syslogs do show inbound traffic to my (previously) assigned address still coming in.
I didn't get a sniffer onto the connection, but I did turn on DHCP debugging for chuckles, which I've attached. Other than stripping out a couple unique IP identifiers, all the data is there. Knock yourself out.
FYI 1728MST today was when I stripped out the URPF command on the FE0 int.
@cramer Already null routed the 10.x.x.x, 127.x.x.x, 169.254.x.x, and 192.168.x.x ranges. Figured I'd turn this option on to be doubly sure, but like I said, I'm not planning on turning back on the command
If you are able to dig up that config option, let me know.
@aryoba Think Cisco needs to add a really BIG disclaimer / caveat in the BOOTP / DHCP doc I was reading. Oh well, live and learn...
Regards |
|
 cramer join:2007-04-10 Raleigh, NC kudos:5 Reviews:
·AT&T Southeast
| You forgot 172.16.... (and multicast)
I could'a swore Cisco's docs did say it would interfere with DHCP, but I'm remembering decade old docs. That said, uRPF was designed for the ISP, full-BGP-table-having crowd. It really doesn't do much for a CPE, plus their needs are small enough that an ACL is a quick, simple solution. ('tho, it does mean you have to maintain the acl.) |
|
 aryobaPremium,MVM join:2002-08-22 kudos:1 | reply to HELLFIRE said by HELLFIRE:@aryoba Think Cisco needs to add a really BIG disclaimer / caveat in the BOOTP / DHCP doc I was reading. Oh well, live and learn...
One way to learn DHCP behavior is to do debug DHCP on your router and see what your router sends and what your router expects to receive  |
|
 | reply to HELLFIRE @cramer Guess what IP range I use for my internal LAN...
@aryoba already did the debug, and yeah, had a reload again last night after I posted the output, the interface to the ISP came up and grabbed an IP address again, no problems.
Thanks all.
Regards |
|
 | reply to HELLFIRE On a sidenote, is it worthwhile to pull a static null to the rest of the 172.16.0.0/12 network, but have an admin distance of 200+, even when I've only got the following subnets configured?
ie "ip route 172.16.0.0.0 0.0.15.255 200"
R1811#sh ip ro 172.16.0.0
Routing entry for 172.16.0.0/24, 4 known subnets
Attached (4 connections)
C 172.16.30.0 is directly connected, Vlan30
C 172.16.20.0 is directly connected, Vlan20
C 172.16.10.0 is directly connected, Vlan10
C 172.16.0.0 is directly connected, Vlan1
Again, it's about keeping RFC1918 addresses out.
Regards |
|
 nosx join:2004-12-27 00000 kudos:5 | Hey HF, i have a question. If you add a static route toward your modem interface for the /32 address of your ISPs DHCP server, does that resolve the DHCP RFP issue? Ping me on skype or something, i want to see if we can make it work. |
|
 cramer join:2007-04-10 Raleigh, NC kudos:5 Reviews:
·AT&T Southeast
| reply to HELLFIRE Yes, because anything not explicitly in the table matches the default route. And the correct syntax: ip route 172.16.0.0 255.240.0.0 Null0 254 permanent (metric and permanent are up to you.)
I don't have a 10/8 null route specifically because of uRPF issues, TWC uses them, I use them , and my employer uses them (ipsec). |
|
 | reply to HELLFIRE @nosx I'll see if I have some time this weekend upcoming to give that a shot.
@cramer Thanks for the input... and the correction. What happens when you add lack of sleep and not thinking things through 
Regards |
|