Re: [Idr] recap my questions and issues raised during IDR Thurs session for draft-ietf-idr-tunnel-encaps-12

[Keyur Patel <[e-mail]>]

Opps apologies. One more time. Comments inlined #Keyur3

From: Keyur Patel <[e-mail]>
Date: Wednesday, June 12, 2019 at 10:01 AM
To: Robert Raszuk <[e-mail]>
Cc: John Scudder <[e-mail]>, Linda Dunbar <[e-mail]>, "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>
Subject: Re: [Idr] recap my questions and issues raised during IDR Thurs session for draft-ietf-idr-tunnel-encaps-12

My comment are inlined #Keyur2

From: Robert Raszuk <[e-mail]>
Date: Wednesday, June 12, 2019 at 9:50 AM
To: Keyur Patel <[e-mail]>
Cc: John Scudder <[e-mail]>, Linda Dunbar <[e-mail]>, "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>
Subject: Re: [Idr] recap my questions and issues raised during IDR Thurs session for draft-ietf-idr-tunnel-encaps-12
Resent-From: <[e-mail]>

Hi,

So when A is creating the update to be send to B for A the meaning of the sub-TLV is really a "local tunnel endpoint".

When the update is received at B the same very field in the sub-TLV now becomes the tunnel destination so it indeed is a  "remote tunnel endpoint".

And then when such update is injected by say third party (for example controller) neither of this applies when the update is being created.

#Keyur3: It is still a remote end-point 😊  for B receiving a route from a controller. It just turns out that the endpoint is not located on the controller. It is located on router A (still an endpoint).

Regards,
Keyur

Therefor why not to simply call the field as "tunnel endpoint address" ? The "remote" word seems to be only generating confusion here.

To add to this let's observe that the tunnel will be bi-direction in vast majority of cases. So the same field now when stored in the OS means different things depending if this is a locally configured/originated value to be send to a peer(s) or remote value learned from the peer(s).

And it is a bit unfortunate but the word "remote" is heavily used in the document today. IMO use of remote should always be associated with the reference of from who's perspective it is to be "remote"

Thx,
R.


On Wed, Jun 12, 2019 at 6:34 PM Keyur Patel <[e-mail]<mailto:[e-mail]>> wrote:
My comments are inlined #Keyur1

From: Robert Raszuk <[e-mail]<mailto:[e-mail]>>
Date: Wednesday, June 12, 2019 at 3:07 AM
To: Keyur Patel <[e-mail]<mailto:[e-mail]>>
Cc: John Scudder <[e-mail]<mailto:[e-mail]>>, Linda Dunbar <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: Re: [Idr] recap my questions and issues raised during IDR Thurs session for draft-ietf-idr-tunnel-encaps-12
Resent-From: <[e-mail]<mailto:[e-mail]>>

Hey,

> #Keyur: Completely agree. Remote Endpoint is clearly not constrained to represent only the BGP speaker.

That is not the question here.

Draft is really not clear in its current state to define what or who the "remote endpoint" is. See there are two endpoints and each is the remote for the other :).

See I have two BGP speakers: A----B

If A sends an update to B with field of "remote endpoint" does this field apply to A or B ? Is it remote from sender perspective or receiver perspective of BGP update. That is the dilemma here.

#Keyur1: In this specific example assuming A originated the route and remote endpoint, then it is from B’s perspective. It is the endpoint address to which the tunnel is established.

Regards,
Keyur

So I agree with Linda that it is in our common interest to define it well before final publication.

And I do not share John's point that this falls into "rough consensus" category if he, Linda, Jun and me and I am sure bunch of other folks reading it while trying to implement will also be unclear. To me it rather falls into - "insufficient review" bucket.

Thx,
R.


On Wed, Jun 12, 2019 at 6:29 AM Keyur Patel <[e-mail]<mailto:[e-mail]>> wrote:
My comments are inlined #Keyur

From: John Scudder <[e-mail]<mailto:[e-mail]>>
Date: Tuesday, June 11, 2019 at 2:07 PM
To: Linda Dunbar <[e-mail]<mailto:[e-mail]>>
Cc: Keyur Patel <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: Re: [Idr] recap my questions and issues raised during IDR Thurs session for draft-ietf-idr-tunnel-encaps-12
Resent-From: <[e-mail]<mailto:[e-mail]>>

(Still as a WG contributor…)

Hi Linda,

On Jun 11, 2019, at 4:41 PM, Linda Dunbar <[e-mail]<mailto:[e-mail]>> wrote:

John,

Before I elaborate for more scenarios, I would like to get the answer to the following question:


  *   Does the “Remote Endpoint” in draft-ietf-idr-tunnel-encaps-12 represent the BGP speaker that originates the update? Or the remote end point that the “Tunnel” is established to?

     *   I have been told two different versions of the answers. I need confirmation from the authors.

I’m not an author of course but I’ll provide my own reading for the sake of continuing the discussion. I just reviewed section 3.1 of the draft, and based on that review it’s my opinion that the Remote Endpoint is clearly not constrained to represent only the BGP speaker that originates the update. Reasons for thinking this include:

- There is no text stating that it’s so constrained. “Everything is permitted except that which is forbidden.”
- There is a detailed set of validation procedures. If the authors had wanted this constraint, it seems unlikely they would have just forgotten to put it in the set of procedures.
- There is special case text that says if the AF subfield is 0, the remote endpoint is inferred from the NH. “The exception proves the rule."

#Keyur: Completely agree. Remote Endpoint is clearly not constrained to represent only the BGP speaker.

Regards,
Keyur


Furthermore, the first two paragraphs of section 9, while they don’t explicitly say a third-party route is being used, only make sense in that context, so that’s yet more evidence that this is the correct reading. Finally, the first paragraph of the Security section has the same implication.


     *
Can a node R send Tunnel-Encap update with “remote endpoint” being A?

Based on the above, my answer would be “yes”. (Subject to the various validation procedures listed in the draft, of course.)

The Section 13 suggests “BGP Origin Validation [RFC6811] can be used”.  But “BGP Origin Validation” is only to validate the Speaker, correct?

No. The TL;DR of RFC 6811 is that

- There’s a database (the RPKI) that lists (in a cryptographically secure way) what ASes are allowed to originate what prefixes
- A router can check a route against that database, considering the prefix and origin AS:
- If they’re found in the RPKI, the route is considered valid
- If no assertion about the prefix is found in the RPKI, the route is considered not found.
- If an assertion is found in the RPKI but it doesn’t match what was found in the update, the route is considered invalid.

You will note this has nothing to do with the speaker that provided the update, it relates only to the contents of the route.

Naturally this is only a quick summary, from memory. If you want authoritative detail you should refer to RFC 6811 itself. It's short and (IMHO :-) readable.

doesn’t seem to address the security scenario being described.

Maybe my description of RFC 6811 helps?

Regards,

—John


Linda

From: John Scudder <[e-mail]<mailto:[e-mail]>>
Sent: Monday, June 10, 2019 5:26 PM
To: Linda Dunbar <[e-mail]<mailto:[e-mail]>>
Cc: Keyur Patel <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Subject: Re: [Idr] recap my questions and issues raised during IDR Thurs session for draft-ietf-idr-tunnel-encaps-12

(Still as a WG contributor)

A little bit more on this.

Your comments led me to go look back at the draft’s Security section again. The first paragraph reads


   The Tunnel Encapsulation attribute can cause traffic to be diverted

   from its normal path, especially when the Remote Endpoint sub-TLV is

   used.  This can have serious consequences if the attribute is added

   or modified illegitimately, as it enables traffic to be "hijacked".

This seems like an explicit acknowledgement of the general class of attack you describe.

It might be helpful if you provide a worked example of a specific attack that would succeed against a tunnel-encaps implementation, but would not succeed against a 5512 implementation. I can’t think of one off the top of my head, but you clearly have something in mind.

Thanks,

—John


On Jun 10, 2019, at 6:16 PM, John Scudder <jgs=[e-mail]<mailto:jgs=[e-mail]>> wrote:

(As a WG contributor)

Hi Linda,

I have a question for you — when you say RFC5512 doesn’t allow a third party to inject routes on behalf of a legitimate router, what do you think would prevent it? You mention the endpoint address in the NLRI, but what would prevent the malicious entity you mention for from falsifying it?

Thanks,

—John

On Jun 10, 2019, at 4:47 PM, Linda Dunbar <[e-mail]<mailto:[e-mail]>> wrote:

Keyur,

Thank for the email.
One more question:

  *   Does the “Remote Endpoint” in draft-ietf-idr-tunnel-encaps-12 represent the BGP speaker that originates the update? Or the remote end point that the “Tunnel” is established to?

     *   I have been told two different versions of the answers. I need confirmation from the authors.


Reading through the Section 13 Security Consideration, I don’t think the following questions have been addressed:


  1.  In RFC5512, the BGP speaker indicates the originating Interface address in the NLRI (section 3):

<image001.png>

Questions:


  *   draft-ietf-idr-tunnel-encaps-12  no longer has the BGP speaker originating the update. Is it intended?


If Yes, does it mean that it allows a third party (which could be malicious entity) to inject routes on behalf of a legitimate router (but RFC5512 doesn’t)?


  *   Why add this scenario? If it is a conscious decision, should have some text to explain why and how to mitigate the security threats introduced.


  *   Section 13 suggests using BGP Origin Validation to obtain the additional assurances of the origin AS is valid. But being valid origin AS doesn’t mean the specific flow is supposed to go/come from there.


#Keyur: Section 13 of the draft version 12 describes Security Considerations that should address your security questions. The option is to provide flexibility.


Thank you,

Linda Dunbar

From: Keyur Patel <[e-mail]<mailto:[e-mail]>>
Sent: Saturday, June 08, 2019 3:45 AM
To: Linda Dunbar <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Cc: John Scudder <[e-mail]<mailto:[e-mail]>>
Subject: Re: recap my questions and issues raised during IDR Thurs session for draft-ietf-idr-tunnel-encaps-11

Hi Linda,

Apologies for the delayed response. Responses are inline. #Keyur

From: Linda Dunbar <[e-mail]<mailto:[e-mail]>>
Date: Thursday, March 28, 2019 at 6:52 AM
To: idr wg <[e-mail]<mailto:[e-mail]>>, "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: recap my questions and issues raised during IDR Thurs session for draft-ietf-idr-tunnel-encaps-11
Resent-From: <[e-mail]<mailto:[e-mail]>>
Resent-To: <[e-mail]<mailto:[e-mail]>>, <[e-mail]<mailto:[e-mail]>>, <[e-mail]<mailto:[e-mail]>>
Resent-Date: Thursday, March 28, 2019 at 6:52 AM

Just want to reiterate my questions and issues I raised during IDR Thurs session for draft-ietf-idr-tunnel-encaps-11, to make it easier for the authors to address them in the next revision (I have sent the questions multiple times on the IDR mailing list, but no one responded):


  1.  When a client route can egress multiple egress ports (each with different IP addresses), does the Tunnel-Encap allow multiple “Remote-endpoint” SubTLV to be attached one UPDATE?

#Keyur: Yes. Section 5 of the draft version 12 has a following  text:

<snip>
A Tunnel Encapsulation attribute may contain several TLVs that all
   specify the same tunnel type.  Each TLV should be considered as
   specifying a different tunnel.  Two tunnels of the same type may have
   different Remote Endpoint sub-TLVs, different Encapsulation sub-TLVs,
   etc.  Choosing between two such tunnels is a matter of local policy.
</snip>



  1.  Section 3.1 Page 10: The last paragraph states that if “Remote-Endpoint sub-TLV contains address is valid but not reachable, and the containing TLV is NOT be malformed ..”. Why a address not reachable is considered as “Not Malformed”?

#Keyur: That is because the Remote-Endpoint could become reachable at the later time. Making it malformed would mean that the Remote-Endpoint has to be dropped upon a receipt of the update message (and could never be used).



  1.  In RFC5512, the BGP speaker indicates the originating Interface address in the NLRI (section 3):

<image001.png>

draft-ietf-idr-tunnel-encaps-11  no longer has the BGP speaker originating the update. Is it intended? If Yes, does it mean that it allows a third party (which could be malicious entity) to inject routes on behalf of a legitimate router (but RFC5512 doesn’t)?  Why add this scenario? How to address the security threats introduced? If it is a conscious decision, should have some text to explain why and how to mitigate the security threats introduced.

#Keyur: Section 13 of the draft version 12 describes Security Considerations that should address your security questions. The option is to provide flexibility.

Regards,
Keyur



Thanks, Linda Dunbar

_______________________________________________
Idr mailing list
[e-mail]<mailto:[e-mail]>
https://urldefense.proofpoint.com/v2/url?u=https-3A__www.ietf.org_mailman_listinfo_idr&d=DwICAg&c=HAkYuh63rsuhr6Scbfh0UjBXeMK-ndb3voDTXcWzoCI&r=hLt5iDJpw7ukqICc0hoT7A&m=ss5r9k8Dew3MPUscEO1GEzNXkyfGpMlZLZUarUnS3Ys&s=_MC-6zg-U8xFzoGxuXaXYA3GVjY7anTiHQ-cuyr9Pfw&e=

_______________________________________________
Idr mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/idr