[Cerowrt-devel] Link layer adaptation VDSL2 - Basis for 8 byte overhead?

[Kevin Darbyshire-Bryant]

On 28/04/2015 16:32, Sebastian Moeller wrote:
> Hi Kevin,
>
> On Apr 28, 2015, at 15:20 , Kevin Darbyshire-Bryant <kevin at darbyshire-bryant.me.uk> wrote:
>
>> Hi,
>>
>> Enquiring mind here :-)  Reading http://www.bufferbloat.net/projects/cerowrt/wiki/Everything_you_wanted_to_know_about_Link_Layer_Adaptation and the linked to email discussion, I'm curious as to where the 8 byte overhead recommendation for VDSL2 comes from.  Is this assumption that PPPoE is being used?  
> 	Not exactly, for all we know using PPPoE and a VLAN on a VDSL2 link results in:
> VDSL2 header:
> VDSL (IEEE 802.3-2012 61.3 relevant for VDSL2): 2 Byte PPP + 6 Byte PPPoE + 4 Byte VLAN + 1 Byte Start of Frame (S), 1 Byte End of Frame (Ck), 2 Byte TC-CRC (PTM-FCS), = 16 Byte
>
> Or in other words, 8 byte either just reflect PPPoE or the real VDSL2 headers plus a VLAN (I am still unsure what to do with the ethernet FCS).
>
> But I think you are right that initially the 8 byte came as a recommendation jus to handle PPPoE overhead ;)
Ha!  Think to myself "See Kevin, *this* is what happens when you start digging and asking questions!" :-)

Hmm, I've a feeling that for most in the UK the value should be 16 to cope with PPP, PPPoE, VLAN, S, Ck, PTM-FCS.  Skyt VDSL2 customers can use 8.  Sky don't use the PPP/PPPoE (& CHAP for login) - rather it is DHCP option 60 used as the customer ID for login purposes.


>
>> Based on that assumption it raises further questions in my mind:
>>
>> My ISP supplier (Sky in the UK) provide straight ethernet over PTM, with DHCP to obtain a public IP address, so in theory no PPPoE overhead unlike other ISPs offering 'fibre' (ha!) broadband in the UK.  There appears to be a tagged VLAN on the WAN port, therefore I think the correct overhead in my case is 4 (VLAN) and for everyone else it should be 12 (VLAN + PPPoE)
> 	See above. BUT this might or might not be relevant; my ISP actually throttles my link to a speed below the VDSL2 link speed and accounts for 16 bytes overhead at the BRAS level, so ymmv… I had a nice way to figure out the per packet overhead on ATM links (actually only ATM links using AAL5, but that should be all of them ;) ), but for PTM I have no real idea...
>
>> Please correct my assumptions :-)
> 	The other thing I am uncertain of is the VLAN tag, if your router terminates it will the kernel account for it or not? Not that I can test this currently as my modem terminates the VLAN “silently"
I was looking at the modem (HG612) which appears to bridge an ethernet port to a PTM port and is set to tag the WAN bound packets with a VLAN tag.  Based on what I can see, the ethernet port is untagged so can only assume the vlan tag goes out on WAN side interface.  Therefore the kernel in OpenWrt doesn't know the tag is there or not so cannot automatically account for it.  The VLAN id is curious too.  It's VLAN 101 which is same as the ATM VPI/VCI (0.101) that Sky UK used for a while with their ADSL 'MER' implementation.  They used VC-Mux but did NOT want ethernet FCS included, if you did include the ethernet FCS the other end just went 'huh?'  Again I've a feeling but no real proof that passing on FCS is unpopular.
>
> Best Regards & hope that helps
> 	Sebastian
It helps prove what a complicated beast this is :-)  and yes it did help (I'm setting the overhead to 8, closing my eyes & shutting my ears!)

Kevin

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4791 bytes
Desc: S/MIME Cryptographic Signature
URL: <https://lists.bufferbloat.net/pipermail/cerowrt-devel/attachments/20150428/ba24cc7f/attachment-0002.bin>