Can a switch fragment a jumbo frame?

Question

In all resources seen that tackle the term of "jumbo frame", I understand that it is applied in a LAN, because:

• It's a Jumbo frame, and a frame is for Layer 2
• These resources talk a lot about switches (L2), and the scenario when they didn't support the jumbo frame.

But I came across this information:

If there is within the network, a switch incompatible with Jumbo frame, it will either drop the frame, or fragment it

The fragmentation is a layer 3 function, how come that a switch can do fragmentation ?

---

Update: adding the resources that cause confusion (to me) about the use of jumbo frame by the routers, and the application of fragmentation by the switches

Starting by the most informative resource, the answer of Alex Maison. As mentionned under the section "Key Point #1":

For a large frame to be transmitted intact from end to end, every component on the path must support that frame size. This means that the switch(es), router(s), and NIC(s) from one end to the other must all support the same size of jumbo frame transmission for a successful jumbo frame communication session.

As I know, the jumbo frame can be used only within the LAN, but he mentions the word "router" !! same thing later under section "Key Point #3"

For a jumbo packet to pass through a router, both the ingress and egress interfaces must support the larger packet size. Otherwise, the packets will be dropped or fragmented.

This lead me to wonder if a jumbo frame can go out of the LAN ?

Hereafter other resources regarding the misunderstanding / misuse of the word "fragmentation" with the switch:

• Ref 2, see "3. What are the drawbacks of jumbo frames?" section (the last paragraph)
• Ref 3, see the answer before the last one (andrew.burns' answer)

Thank you

Answer 1

Can a switch fragment a jumbo frame?

No. Ethernet has no fragmentation. Jumbo frames are non-standard, and different vendors, different switches by the same vendor, and different interfaces in the same switch model, may all have different jumbo frame sizes.

A jumbo frame needs to have every interface in its path be able to accept the frame size, or the frame will simply be dropped.

I have no idea where you found the quote about switches fragmenting jumbo frames, but it is incorrect because Ethernet has nothing in its header that facilitates fragmentation and reassembly. Even IPv6 has eliminated in-path fragmentation because it is resource intensive and slows packet forwarding.

---

Edit for your edit:

This lead me to wonder if a jumbo frame can go out of the LAN ?

No, frames (layer-2 PDUs) are confined to the layer-2 LAN. Routers, including layer-3 switches, strip the frame off the packet before forwarding the packet to the next interface, which may use a different data-link protocol that needs a different frame, and the router builds a new frame for the data-link protocol of the next interface.

If the payload of the received layer-2 frame (layer-3 packet) is an IPv4 packet that does not have the DF bit set, it may be fragmented, but that is layer-3 packet fragmentation, not layer-2 frame fragmentation. Ethernet frames are not fragmented.

---

I think you are not carefully parsing packets and frames.

• The Key Point #3 quote is about jumbo packets, not jumbo frames, so the mention of routers is appropriate.
• Your Ref 2 implies that a router gets the frame to forward the packet, and it must fragment the packet because the next interface MTU is smaller than the MTU of the receiving interface.
• Your Ref 3 is about fragmenting packets, not frames.

It is possible to fragment IPv4 packets that do not have the DF bit set, but you cannot fragment ethernet frames, nor can you fragment IPv6 packets. I do not know how many times we need to say that, but that is the way it is.

IPv4 has some fields in the IPv4 packet header that facilitate packet fragmentation. Neither ethernet nor IPv6 have any of that, and both are incapable of fragmentation. If an ethernet interface receives a frame larger than its MTU, the frame is silently dropped. If a router receives an IPv6 packet that is larger than the MTU of the next interface, it is dropped and an ICMP error is sent back to the source host. The source host can know if a packet is dropped because of the MTU, but has no knowledge that a frame is dropped because of the MTU.