Anurag S. Maskey wrote:
> Renee and I discussed this yesterday and came up with the solution that 
> if an NCU requests both v4 and v6 addresses, then both of them should 
> come up (ignoring v6 link-local address) before the NCU moves to the 
> online state.  We didn't think about static addresses.  What would be 
> the correct state if v4 static address is configured, but v4 dhcp 
> address is not and v6 is not configured?
> 
> I remember us discussing that an NCU will be considered online as long 
> as at least one address on it remains.  This ignored whether the last 
> remaining address is v4 or v6 or both.  I feel that the right thing to 
> do for the NCU to be online as long as at least one address remains for 
> each configure IP version.  So, if both v4 and v6 are configured, then 
> the NCU will be online as long as at least one v4 and one v6 address 
> (except link-local) remains.  If there is only one v4 address and no v6 
> address, then the NCU changes to offline*/waiting state.
> 
> Thoughts?
> 

My opinion here is that the NCU should be on-line as soon as one non 
link-local address is assigned.  Otherwise, NWAM fails to install a 
useful location (ie, not NoNet), so connectivity is broken even though a 
usable address is likely present.  Any NCU with a static address 
assigned is online, unless that address fails DAD.  If a the presence of 
a duplicate causes the NCU to go online, but then offline, that probably 
fine.

-John


> Anurag
> 
> 
> bugzilla at defect.opensolaris.org wrote:
>> http://defect.opensolaris.org/bz/show_bug.cgi?id=11103
>>
>>
>>
>> --- Comment #2 from Anurag S. Maskey <Anurag.Maskey at Sun.COM> 
>> 2009-09-03 16:09:53 UTC ---
>> I think the RTM_NEWADDR routing socket messages for link-local 
>> addresses should
>> be ignored and should not generate any state change events.  Setting 
>> of the
>> link-local addresses will always succeed and will stomp over the state 
>> of the
>> interface NCUs.
>>
>>   
> _______________________________________________
> nwam-dev mailing list
> nwam-dev at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/nwam-dev

Reply via email to