+spring

Hi Joel, Ack on forward reference in 6.2, I’ll use the text I suggested.

However Prefix-H is an example for the following

<quote>
   SR Domain ingress routers permit traffic destined to select SIDs with
   segment lists.
</quote>

   local policy requiring HMAC TLV processing for those select SIDs,
   i.e. those SIDs provide a gateway to the SR Domain for a set of


Prefix-H is is a useful illustration, but it is only one way an administrator 
may choose to permit packets destined to SIDs requiring HMAC TLV processing to 
enter their domain.

Depending on the deployment she may decide to permit specific source address to 
use specific SIDs requiring HMAC processing on specific interfaces only.
In this case the fact that all those SIDs are allocated within a Prefix-H is 
not relevant, nor would she permit packets destined to prefix-H at all ingress 
nodes.

Darren


On Oct 22, 2018, at 8:39 PM, Joel M. Halpern 
<[email protected]<mailto:[email protected]>> wrote:

I am not sure I am putting the intended pieces together right.

A packet arrives at a domain edge that supports SRv6.  There seem now to be 
several cases (I had ignored prefix-H as fragile, but that is a different 
debate.

1) The packet is not addressed to any SID in the domain.  Then the packet can 
be processed, possibly encapsulated with an SRH, and sent inwards.

2) The packet is addressed to a SID that is not in prefix-H (i.e. not a SID 
known to be configured to check HMACs.)  Drop the packet.  This was the step I 
had not understood in by previous reading of 6.2.1, as it seems structurally to 
be part of an example, not a required piece of configuration behavior.

3) The packet is address to a SID in prefix-H.  If that SID is this node, check 
the HMAC and act based on it.  If it is some other SID within prefix-H, then 
send it on its way.

If that is the intended behavior, then I would suggest extra text in 6.2.  The 
text can forward reference prefix-H for SIDs known to check HMAC.  And should 
include the case where teh SID is this edge node (which is frankly the most 
likely and robust way to do the checking.)

If you do not intend the configuration and use of prefix-H to be normative, 
then you need to explain that magic knowledge of such checking is necessary, or 
that the ingress edge node must check the H-MAC even if it is not the current 
destination, when it does not have such configuration.

Yours,
Joel

On 10/22/18 8:28 PM, Darren Dukes (ddukes) wrote:
On Oct 22, 2018, at 7:52 PM, Joel M. Halpern 
<[email protected]<mailto:[email protected]> 
<mailto:[email protected]>> wrote:

The text in section 6.2.1 is fine, and clear.  It merely contradicts the text 
in section 6.2.  What I am asking is not for a behavioral change, but simply 
that the text in 6.2 acknoweldge the acception of 6.2.1 explicitly,
Are you suggesting text like the following in section 6.2 para 1?
 “...SHOULD discard packets destined to SIDs within the SR Domain *<new> which 
do not require HMAC verification </new>* (of the presence of an SRH) to avoid 
attacks on the SR Domain as described and referenced in [RFC5095]. *<new> SIDs 
requiring HMAC verification are discussed in Section 6.2.1..</new>*"
and as a side effect not state that the discard happens "regardless of the 
presence of an SRH", since in the presence of an SRH, the edge node has to (I 
would hope MUST) check the SRH for an HMAC, and attempt to verify the HMAC.
There is no need for an edge to check for an SRH when the destination is within 
Prefix-H (as defined in 6.2.1).

I suppose that if the edge node  knew that there were no valid HMACs, then it 
could skip checking the SRH since it can't be valid.  If you want that 
included, do so.

It would take a few extra sentence for 6.2 tyo properly set the ground for 
6.2.1.

I really dislike when the text says "X SHOULD do Y" and then in a later 
section, without even saying "this is an exception to the above" the text says 
"X SHOULD do Z under condition Q".  It confuses readers.

Yours,
Joel

On 10/22/18 7:45 PM, Darren Dukes (ddukes) wrote:
Thanks Joel, please see inline.
On Oct 22, 2018, at 6:59 PM, Joel M. Halpern 
<[email protected]<mailto:[email protected]> 
<mailto:[email protected]>> wrote:

I think this is a minor issue.  Sections 6.2 and 6.2.1 talk about processing of 
packets from outside the SR Domain.  Section 6.2, says that edge nodes "SHOULD 
discard packets destined to SIDs within the SR domain."  Sounds good.  It even 
includes the parenthetical "(regardless of the presence of an SRH)"  Which 
sounds even better.  But is contradicted by section 6.2.1.

Section 6.2.1 says that an trusted external node can provide an SRH. The 
trusted node protects that SRH with an HMAC TLV.  At the domain edge, I presume 
we know it is a trusted node by the HMAC TLV.  (Not by the IP Source address, 
since that may be false.)  Which means that in fact, the discard in section 6.2 
is not "regardless of the presence of an SRH" it is rather unless there is an 
SRH with a valid HMAC.
Please re-read page 20 (the rest of 6.2.1) and let me know if you still think 
this is not fully defined.  This is clear with Prefix-H and the definition of 
the SIDs assigned from that prefix.

If that is the intention for 6.2, could we please have the text say that.  And 
then upgrade the SHOULD for the edge to a MUST?
For the Internal nodes in section 6.2, the SHOULD ought to also reference the 
HMAC.  I can see leaving that as a SHOULD, although I would prefer to see it as 
a MUST.
If after re-reading you still have a concern in this area let me know and we 
can talk more.
Darren

Yours,
Joel

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]<mailto:[email protected]> <mailto:[email protected]>
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]<mailto:[email protected]>
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

_______________________________________________
spring mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/spring

Reply via email to