FTTH

Qinn

Hi there my ISP uses

• PPPoE via VLAN 6 (802.1q).
• PPPoE authentication (PAP) using a username and password (e.g., internet / internet).
• Maximum packet size (MTU) 1500 bytes (RFC 4638)

So my reason was with VLAN tag overhead (4 byes). I have the physical interface (WAN) set to 1512, VLAN 6 to 1508, and PPPoE to 1500. This is with AON fiber.

but I cannot set VLAN in pfS, can someone give me some pointers

btw my ISP is KPN (Dutch)

KOM

@Qinn What do you mean in that you can't set a vlan? You can't find where to configure it, or you configured it and it didn't work?

Qinn

@KOM No can set all vlan an pppoe.

You have 3 interfaces igc2 vlan trunk interface
igc2.6 vlan inertface
ppoe1 tunnel interface

all three have MTU, For the "inner" -the PPPoE tunnel interface- you want an MTU of 1500, so that packets coming from your LAN do not need to be fragmented.
Because encapsulation is involved, the MTU for the underlying interfaces must be larger to accommodate that encapsulation:

igc2.6 = PPPoE MTU + 8 bytes PPP header = 1508
icg2 = PPPoE MTU + 8 bytes PPP header + 802.1q header = 1512

but in pfSense you can only see igc2.6

KOM

@Qinn Have you tried it without changing the mtu just to see if it performs as expected? I have a fibre link via PPP and i didnt play with mtu/mss at all, and i get about 960Mbps up/down.

Regardless, i think you have to assign igc2 to an interface and then set the mtu there.

Edit: My understanding is that the vlan tag is part of the frame header and not the payload so changing the mtu wouldn't have any effect?

stephenw10

If you set the parent NIC the other sub interfaces should inherit the MTU from it. You might need to resave them to apply it.

JKnott

@Qinn said in FTTH:

So my reason was with VLAN tag overhead (4 byes).

The VLAN tag is outside of the area that the MTU applies to, so no need to adjust it.

KOM

@Qinn Was there any resolution to this problem?

Qinn

@KOM Yes at current I am at 1500

Speeds are good, paying 1000Mb/s symmetric and getting:

netblues

@Qinn Apart from the fun of it, even without an mtu of 1500 you would still get the same throughput.

I have done the same on my pppoe link (because I can) but in practice I see no difference.

Having said that, a few months ago isp was doing some maintenance and traffic was rerouted to a different bras (which also forced all connections behind cgnat.)
And guess what, rfc4638 wasn't configured leading to connectivity issues.

So its nice, but rather non-standard and can result into issues and then you won't even remember to check this.

Qinn

@netblues said in FTTH:

Thanks for pointing that one out to me, will take it into consideration.

cgnat is it StarLink?

stephenw10

Unfortunately many ISPs use CGNAT. Increasingly so. Grr.

netblues

@Qinn I guess Starlink isn't using pppoe to begin with.

Now days I even see private address space everywhere and ipv4 public ip's in a 1-1 nat, carrier side, for static ipv4 requests.