IPv6 WiFi Troubleshooting

Some users have encountered the problem that after enabling IPv6 on their optical network terminal (ONT) or router, they find that the wireless connection to wifi on their computers and mobile phones is slow, and sometimes even web pages cannot be opened. Here are some solutions.

It has been found that recently, many people have frequently mentioned the issue of slow IPv6 connection speed. At present, there are indeed insufficient servers and CDN nodes that support IPv6 in China, and the international bandwidth of IPv6 is smaller than that of IPv6. However, it is not to the extent that all domestic websites lag when opened. Under normal circumstances, encountering this problem indicates that there is a PMTU black hole on your link to the target server.

Regarding the PMTU black hole

MTU (Maximum transmission unit) is the maximum size of the three-layer data packets (including IP headers) that can pass through a link. The default MTU on Ethernet is 1500 bytes, but there may be links on the path between you and the target server that are smaller than the MTU 1500. The smallest MTU value on this path is the PMTU value of the entire path. When a router forwards packets, packets larger than the MTU size will be fragmented (Transmission), that is, a large packet will be split into multiple small packets not exceeding the MTU for transportation, and the transmission efficiency will decrease.

When the terminal device sends packets, it can also set the DF(Don't Fragment) flag to tell the router not to fragment. At this point, the intermediate router will discard packets exceeding the MTU and reply with an ICMP Fragmentation Needed message. After the sender receives this packet, they will send a smaller packet next time. This process is called PMTU Discovery. In reality, it can be seen that most of the HTTPS traffic is marked with "DF".

However, there are a large number of intermediate devices on the Internet that do not respond to ICMP for so-called "security" reasons or due to incorrect configuration

Fragmentation Needed packets, which cause a packet to be silently discarded when accessing certain websites if its size exceeds PMTU, until the TCP protocol detects the timeout packet loss and retransmits it, which is very slow. In such a situation, we can say that there is a PMTU black hole on the path between you and the target server.

In addition, IPV6 does not support sharding. In other words, it can be understood that all packets under IPv6 are marked with DF. When the intermediate router encounters a situation where the Packet size is larger than the MTU, it should respond to the ICMPv6 Packet Too Big message. Similarly, for various reasons, some intermediate devices may directly lose packets without responding to the ICMPV6 Packet Too Big message until the TCP protocol detects the timeout packet loss and retransmits it.

Why doesn't IPv4 have this problem?

In fact, IPv4 also has this problem. I have seen netizens say more than once that the soft router they set up was very slow to access certain websites, but it worked fine when they switched back to the hard router. This is because most home routers have MSS (maximum segment size) Clamping enabled by default for TCP under IPv4 (friends using OpenWRT software routers can find the MSS Clamping switch in the firewall Settings). MSS Clamping is a Workround for the PMTU black hole. Simply put, there is an MSS field during the TCP handshake that determines the maximum size of a single TCP packet. The router can reduce the MSS value by sniffing the TCP handshake packet, so that the size of the final three-layer IP packet (MSS+TCP header size +IP header size) does not exceed a specific value.

Nowadays, most domestic ISPs establish WAN port connections through PPPoE virtual dial-up. The default MTU of Ethernet is 1500, but the PPPoE tunnel has an overhead of 8 bytes. Therefore, the MTU of the PPPoE virtual connection is 1500-8=1492, minus IPv4

From the packet header (20 bytes) and the TCP packet header (20 bytes), it can be known that under IPv4, the MSS needs to be set below 1452.

The header of IPv6 is 40 bytes, so under IPv6, the MSS needs to be set below 1432.

At this point, the problem arises. Currently, many optical MOdems and home routers have poor optimization for IPv6 and do not support MSS Clamping on TCP packets under IPv6. This results in a very slow opening process when accessing IPv6 websites if there is a PMTU black hole in the path.