[c-nsp] PPPoE on dot1q VLANs vs. sub-interfaces

[Kristofer Sigurdsson]

Hi list,

We just bumped into odd behavior in a Cisco 7301, running IOS 
12.3(11)T2.

We have been terminating PPPoE over 802.1q VLANs on this box.  Nothing
fancy, just local termination.  We have one GigabitEthernet interface
facing the customers.

We were using the following interface config on the interface facing the
customers:

interface GigabitEthernet0/1
 description xxx
 no ip address
 duplex full
 speed 1000
 media-type rj45
 no negotiation auto
 vlan-range dot1q 301 399
  pppoe enable
  exit-vlan-config
 !
!

There were no sub-interfaces under gig0/1.

However, this morning, I had to configure 2 point-to-point connections
on this router, for a "special" customer, I did the following:

interface GigabitEthernet0/1.60
 description xxx
 encapsulation dot1Q 60
 ip address xxx 255.255.255.252

interface GigabitEthernet0/1.61
 description xxx
 encapsulation dot1Q 61
 ip address xxx 255.255.255.252

When I did this, all of a sudden, new PPPoE sessions could not be
established, but the sessions already up continued to be up.  They went
to "LCP" and stopped there.  According to debug output, they went to 
"data path set to PPP" and stopped there.

I created subinterfaces for vlans 301-399, eg:

interface GigabitEthernet0/1.345
 encapsulation dot1q 345
 pppoe enable

This seems to have fixed the problem (we're still verifying), but this
is definately not something we want - crud in the config.

Is this a bug, behavior to be expected, or am I doing something wrong?  
Any pointers would be appreciated.

Thanks,
Kristofer