Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Announcements

Welcome to Cisco Support Community. We would love to have your feedback.

For an introduction to the new site, click here. And see here for current known issues.

Silver

OSPF Hello Parameters- SM Question

I observed a interesting behaviour while I was setting up a LAB yesterday- OSPF on serial links/Point to Point links ignore the mismatched subnet mask and form a full adj while the same config on a ethernet/broadcast links throws out an error saying mismatched Hello paremeters-

Can someone explain me the reason for this behaviour

Serial link

R1-s0/0-10.0.0.1 255.255.255.252---10.0.0.2 255.255.255.0-s0/0-R2-----Forms a Fulll adj

Ethernet Link

R1-F0/0-10.0.0.1 255.255.255.252---10.0.0.2 255.255.255.0-F0/0-R2----Doesnt reach 2way state because of Hello parameter(Subnet Mask) mismatch

Sidd

Siddhartha
4 ACCEPTED SOLUTIONS

Accepted Solutions
Hall of Fame Super Silver

Re: OSPF Hello Parameters- SM Question

Siddhartha

But on point to point serial link (which is what the RFC is oriented to) there is not any MAC address.

My understanding of the reason for the exception for point to point links is that some types of OSPF messages are sent to the unicast address (acknowledgements, retransmission of LSA, etc) of the neighbor. For broadcast (Ethernet) networks if the subnet mask does not match it is quite possible that arp will not work correctly and that neighbors might not be able to get to the neighbor unicast address. So there was a requirement for the subnet mask to match to be sure that unicast communication would work. But with point to point the subnet mask makes no difference. With point to point you just send the packet out the interface and it always gets to the right destination. So there is not a need for subnet mask to match on point to point interfaces.

HTH

Rick

Hall of Fame Super Blue

Re: OSPF Hello Parameters- SM Question

Lets say you have an ethernet segment with 3 routers attached R1, R2 and R3.

R1 = 192.168.5.10 255.255.255.240

R2 = 192.168.5.11 255.255.255.240

R3 = 192.168.5.17 255.255.255.240 

note that R3 is actually in a different network (192.168.5.16/28) and the administrator has incorrectly assigned the IP .17

If the network mask was not checked then these 3 routers would form adjacencies because each would see the multicast packets.  But R1 and R2 will both think R3 is on a different network and R3 will think R1 and R2 are on different networks. They will not be able to send unicast packets to R3 because they will think they have to route them.

In addition what would happen if R3 was elected DR. The DR generates network LSAs (type 2) and it would generate one for the 192.168.5.16/28 network but it is actually connected to the 192.168.5.0/28 network.  I don't know how this would affect the SPF calculation but there would be clearly be incorrect information being used.

Jon

Hall of Fame Super Blue

OSPF Hello Parameters- SM Question

From R3's persepctive R1 and R2 are on the same subnet. From R1 and R2s perspective R3 is on a different subnet. So you would get similiar issues in terms of unicasting etc.

Note that the bit about a DR being elected and bad LSA information is a best guess in that i can't actually test it (obviously as the adjacencies won't form) but i could see how it might cause problems.

Certainly as botth Rick and myself pointed out unicasting of LSA etc. would be an issue.

Jon

Hall of Fame Super Silver

OSPF Hello Parameters- SM Question

In this case R1 and R2 will believe that R3 is in a different subnet. So they will not send an arp request for R3 but will want some routing entry for how to get to that subnet.

In the scenario suggested by Jon R1 and R2 think that R3 is remote and R3 believes that R1 and R2 are remote. Your scenario seems to address the issue from the perspective of R3 which will now believe that R1 and R2 are local and will arp for them. However it is highly likely that R1 and R2 will reject the arp request from R3 because it came from a remote IP (and the Cisco behavior is to reject an arp request from a device whose source address is not on the local subnet). So the result in your scenario is still that the routers do not communicate.

HTH

Rick

21 REPLIES
VIP Super Bronze

OSPF Hello Parameters- SM Question

Silver

OSPF Hello Parameters- SM Question

Thanks Reza, I already read that post- that didn't explain the reason for the exception-- Does anyone why the point to point links have an exception?

Siddhartha

Siddhartha
Silver

OSPF Hello Parameters- SM Question

anyone?

Sidd

Siddhartha

Re: OSPF Hello Parameters- SM Question

Hello Sid,

What do you mean by it did not explain the reason for the execption??

I mean Reza already told u bud (some kudos to reza)

The RFC clearly states:

However, there is one exception to the above rule: on point-to-point networks and on virtual links, the Network Mask in the received Hello Packet should be ignored.

And back to you are scenario, what's the default encapsulation for a serial interface PPP cause it's a point to point link.

That being said the expection goes there

Rate all of the helpful posts!!!

Regards,

Jcarvaja

Follow me on http://laguiadelnetworking.com

Julio Carvajal
Senior Network Security and Core Specialist
CCIE #42930, 2xCCNP, JNCIP-SEC
Hall of Fame Super Blue

Re: OSPF Hello Parameters- SM Question

edited      

Hall of Fame Super Blue

Re: OSPF Hello Parameters- SM Question

Sidd

Why doesn't the link provide the explanation ?  It states that the RFC for OSPFv2 makes a specific exception for virtual and point-to-point networks and that exception is that the subnet mask does not need to match. It also says that Cisco strictly adhere to the RFC so they have implemented OSPF that way.

Am i misunderstanding your question ?

Jon

Silver

Re: OSPF Hello Parameters- SM Question

Hi Jon, I meant to ask what is the technical reason behind that exception.

Sidd

Siddhartha
Hall of Fame Super Blue

Re: OSPF Hello Parameters- SM Question

Sidd

Sorry, i was posting as you replied -

I think the technical reason is that on a point-to-point link if a router sends a packet, and providing the link is up, the only place that packet can end up is at the router at the other end. So the subnet mask is not really an issue.

Compare that to multi-access networks where the subnet mask very much matters in terms of which adjacencies are formed.

Jon

Silver

Re: OSPF Hello Parameters- SM Question

Thanks Jon, I am having hard time in understanding. Whther its a P to P link or multi access link the LSAs will always be sent to the layer 2 mac address calculated from either 224.0.0.5 or 224.0.0.6 address right? why does the layer3 address or subnet mask matter( as long as they have reachability)

Please correct me if I am wrong

Siddhartha

Siddhartha
Hall of Fame Super Silver

Re: OSPF Hello Parameters- SM Question

Siddhartha

But on point to point serial link (which is what the RFC is oriented to) there is not any MAC address.

My understanding of the reason for the exception for point to point links is that some types of OSPF messages are sent to the unicast address (acknowledgements, retransmission of LSA, etc) of the neighbor. For broadcast (Ethernet) networks if the subnet mask does not match it is quite possible that arp will not work correctly and that neighbors might not be able to get to the neighbor unicast address. So there was a requirement for the subnet mask to match to be sure that unicast communication would work. But with point to point the subnet mask makes no difference. With point to point you just send the packet out the interface and it always gets to the right destination. So there is not a need for subnet mask to match on point to point interfaces.

HTH

Rick

Silver

Re: OSPF Hello Parameters- SM Question

make sense- thanks Richard and Jon

Siddhartha

Siddhartha
Hall of Fame Super Blue

Re: OSPF Hello Parameters- SM Question

edited

Hall of Fame Super Blue

Re: OSPF Hello Parameters- SM Question

Lets say you have an ethernet segment with 3 routers attached R1, R2 and R3.

R1 = 192.168.5.10 255.255.255.240

R2 = 192.168.5.11 255.255.255.240

R3 = 192.168.5.17 255.255.255.240 

note that R3 is actually in a different network (192.168.5.16/28) and the administrator has incorrectly assigned the IP .17

If the network mask was not checked then these 3 routers would form adjacencies because each would see the multicast packets.  But R1 and R2 will both think R3 is on a different network and R3 will think R1 and R2 are on different networks. They will not be able to send unicast packets to R3 because they will think they have to route them.

In addition what would happen if R3 was elected DR. The DR generates network LSAs (type 2) and it would generate one for the 192.168.5.16/28 network but it is actually connected to the 192.168.5.0/28 network.  I don't know how this would affect the SPF calculation but there would be clearly be incorrect information being used.

Jon

Silver

Re: OSPF Hello Parameters- SM Question

what happens if

R1 = 192.168.5.10 255.255.255.240

R2 = 192.168.5.11 255.255.255.240

R3 = 192.168.5.17 255.255.255.0

Siddhartha

Siddhartha
Hall of Fame Super Blue

OSPF Hello Parameters- SM Question

From R3's persepctive R1 and R2 are on the same subnet. From R1 and R2s perspective R3 is on a different subnet. So you would get similiar issues in terms of unicasting etc.

Note that the bit about a DR being elected and bad LSA information is a best guess in that i can't actually test it (obviously as the adjacencies won't form) but i could see how it might cause problems.

Certainly as botth Rick and myself pointed out unicasting of LSA etc. would be an issue.

Jon

Silver

Re: OSPF Hello Parameters- SM Question

Thank you both !

Siddhartha

Siddhartha
Hall of Fame Super Silver

OSPF Hello Parameters- SM Question

In this case R1 and R2 will believe that R3 is in a different subnet. So they will not send an arp request for R3 but will want some routing entry for how to get to that subnet.

In the scenario suggested by Jon R1 and R2 think that R3 is remote and R3 believes that R1 and R2 are remote. Your scenario seems to address the issue from the perspective of R3 which will now believe that R1 and R2 are local and will arp for them. However it is highly likely that R1 and R2 will reject the arp request from R3 because it came from a remote IP (and the Cisco behavior is to reject an arp request from a device whose source address is not on the local subnet). So the result in your scenario is still that the routers do not communicate.

HTH

Rick

Hall of Fame Super Blue

Re: OSPF Hello Parameters- SM Question

Rick

To save me having to dig out my TCP/IP routing books do you think in the example i gave that R3 could end up as the DR and that this would indeed cause problems with the LSAs propogated and the subsequent SPF calculation ?

I agree on the unicasting bit but i'm just wondering whether it would actually create a corrupt LSA database among those three routers.

Jon

Hall of Fame Super Silver

OSPF Hello Parameters- SM Question

Jon

It is indeed an interesting scenario and not clear exactly what the outcome would be. But I believe that the bottom line is that forming adjacencies would fail and the OSPF database would not be populated. I believe that the scenario might go like this:

- R3 becomes DR.

- R3 sends multicast hello message announcing that it is DR.

- multicast message reaches R1 and R2.

- R1 and R2 send hello messages and begin the process of attempting to form adjacencies with R3.

- R3 would try to send data base updates to R1 and R2 (and I am pretty sure that the database update would be unicast). Assuming that the database updates are unicast here is where things break because I am pretty sure that R3 arp for R1 and R2 did not work.

- if the database update did get to R1 and R2 then they would send acknowledgements to R3. And things pretty certainly break here because R1 and R2 believe that R3 is in a remote subnet and they have no route to that subnet. So they can not send unicast acknowledgement.

This has certainly been an interesting thought process

HTH

Rick

Hall of Fame Super Blue

Re: OSPF Hello Parameters- SM Question

Rick

So i guess it would never get to the SPF calculation ie. either database updates or acks for those would be unicast and so would never get there. Thanks for clarifying and saving me having to pour through my books.

It has indeed been interesting and it's one of the things that makes CSC so worthwhile.

Jon

Hall of Fame Super Silver

OSPF Hello Parameters- SM Question

Jon

Agree with you 100%

HTH

Rick

323
Views
5
Helpful
21
Replies
CreatePlease login to create content