Flint 2 - packet loss

Router
1

I have a Flint 2, connected to a Virgin Media docsis cable connection (on WAN1, 2.5gbps). The cable router is in bridge / modem mode and the Flint 2 is doing the routing.

I have been running vanilla Openwrt for the last 6 months and everything has been fine until I upgraded Openwrt from 24.10.6 to 25.10.4. I started getting packet loss, no trend, day or night and packet loss significant enough at times to drop the WAN.

I obviously thought let’s go back to 24.10.6 then. Issue still persisted.

I then decided to flash 4.8.3,4.8.4 GL-iNet firmware and Openwrt24 4.9.0.

The router is now running on 4.7.7 as I know that was very stable for me……….and it’s still dropping packets.

I then started to look towards the cable side. Firstly, I took the router out of modem mode, ran it in router mode for 24 hours and the cable router did not drop a single packet.

I then thought hmm, maybe the 2.5gbe WAN port is flapping on the cable router. I plugged in an Asus TUF gaming router and it didn’t drop a single packet too. Same ethernet cable from WAN to WAN.

I have left all my Ethernet devices unplugged apart from WAN1. Makes zero difference.

I have disabled the WAN failover interface detection for Ethernet 1 as I have read it can be overly aggressive and cause connection issues. With it enabled, it was detecting WAN1 dropping. With it disabled I have not seen that since but still see the packet loss.

I have tried setting the MTU to 1492.

I am starting to run out of ideas on what it could be, apart from a physical hardware port issue, which I really hope it isn’t.

One thing I have noticed is I cannot flash by Uboot any more since flashing 4.8.3 GL-iNet firmware. It fails to upload the GL-iNet bin files. I have tried two different PC’s, it just says FAILED and I may be using an incorrect file type.

Any help or tips on this packet loss issue would be great.

2

Monitoring the connection and router logs today the huge influx of packet loss which coincides with the connection blipping is related to…

Maximum number of concurrent dns queries reached (max 150).

That could be related to the WAN blipping or it could be a roque device. My question is now, how do I prove it?

3

Another update……

I have managed to flash via UBoot. It needed to be plugged into LAN 5 for some reason.

The Flint is now running 4.8.4.

1 Like
4

Did you recognise the device making all those dns requests, causing others to drop?

5

I wasn’t able to distinguish what device was causing it on 4.7.7 as DHCP and DNS logging was not turned on.

On firmware 4.8.4 it is now turned on so if a device is causing it, I should be able to workout what device it is.

If I cannot find a device causing it, it will most probably be the wan connection dropping?

6

This seem to look like a race condition going on, do you have adguardhome enabled and particularly also rebind protection ?

A) if true, disable the rebind protection please, it will confirm in the logs aswell, basically adguardhome blocked a domain and the routers dhcp server sees that as rebinding...

Then maybe there are also some other things...

lets begin with the devices connected to the Flint 2 on the wired segment.

Can you give us a more thorough topology how it is connected?

Q1: are there managed switches or switches involved?

Q2: what kind of devices have trouble here, what is the hardwares nic?

Here are some possible clues to my questions:

A1) managed switches need to have priority set, the closer to your router the lower the number the farer away the higher, this will ensure that daisy chained switches know who the 'main' switch is, it is important for STP too (loop detection).

another issue could be that you have mixed different negotiations, if all are 1gb negotiated except one device uses 100M, you need to disable flow control for this port if lan2 was 100M ethtool -A lan2 tx off rx off, this will ensure the port will not panic and sent pause frames to the full bridge when there is heavy congestion on the network, same counts for downstream switches, you need this command on every boot, best is to place it in /etc/rc.local.

A2): this is difficult, you may have to look into Windows device manager or in linux to turn off certain offload functions (you have to learn a little about your nic card what is stable and what not), you want to look for LRO, TSO and some other offload features some can be severely bugged or don't play well with the router, green ethernet is exceptionally one which is reported quite alot breaking connections.

There can be more reasons, these are the harder ones, but it could be as easily as a ip conflict, or having a second router daisy chained connected on lan (which often works on normal routers, this doesn't work by default on OpenWrt based routers only by manual user configuration).

1 Like
7

I have just had another WAN blip by the looks or the router has skipped a couple of pings and triggered the WAN monitor.

Tue Jun  2 22:51:18 2026 daemon.info gl-repeater[2751]: (repeater.lua:1776) interface wan status offline
Tue Jun  2 22:51:23 2026 daemon.info gl-repeater[2751]: (repeater.lua:1776) interface wan status online

Physical topology.

Virgin Media cable modem / router 2.5Gbe cat6 into WAN 1 on Flint2.

I have 2x Ethernet devices plugged directly into the Flint 2. A 1gbe power line adapter (end device currently off) and a 100Mbps CCTV recording base station. Both of these are currently set to auto negotiation.

Adguard is not enabled.

DNS server is still set to the ISP’s currently.

Redbind protection is currently off.

8

Seems like a problem on the upstream virgin media device then. Can it not be completely replaced either this flint2?

9

I replaced the Flint 2 with an ASUS TUF WiFi 7 router and it did not drop WAN once in 24 hours.

I plugged the Flint 2 back in and it immediately started dropping packets again.

What I don’t understand is if the Virgin connection was intermittent I would see it when the Virgin modem router was in router mode. It is absolutely rock solid, zero packet loss.

10

Are you behind double Nat? Is there an ip conflict?

11

Has repeater modus be enabled?

Make sure repeater modus is disabled, I understand between isp modem and flint 2 it is ethernet wan connected correct?

12

I have now disabled repeater modus.

No double NAT issues, the Flint 2 is doing NAT, DHCP and all routing. The Virgin Media hub is in modem mode and is acting as a bridge.

No IP clashes, VM hub is on 192.168.100.1 and the Flint is on 192.168.8.1.

Huge amounts of packet loss at 5am and 6:20am resulting in the connection dropping. VM normally do their maintenance at 2am and I highly doubt it will be utilisation at that time of the day.

I have exported the logs for this period so will see if I can see anything.

13

Hmm this make me think about this.

Is your central connection from outside to in house fiber ?

Maybe the optics are not clean, or damaged, then you can also have very unstable internet, often it is in a box you are not allowed to open by isp but then you need to ask isp to test it and monitor.

14

It is a fiber connection which converts back to coaxial.

I have had to disconnect the Flint 2 as it is not stable enough for me to work from home today. Using the Virgin Media supplied router, the connection appears to be faultless.

I have opened the log file and I cannot see anything at all related to the WAN dropping which makes me think the connection is hanging briefly enough for packet loss but it quickly recovers.

Also worth noting when using an Asus TUF 9400, the connection is also faultless with the cable router in Bridge Mode.

There is a lot going on in the logs though.

15

The logs seem to highlight issues……

ICMP to the router being handled in a “slow/marked” path

Example:

PROTO=ICMP TYPE=8 CODE=0 ID=56026 SEQ=58100 MARK=0x8000

Those are pings from multiple LAN clients (192.168.8.195, .215, .244) to the router (192.168.8.1) on br-lan, all tagged with MARK=0x8000.

Continuous ADDBA/Block Ack negotiation failures on the Wi‑Fi driver

You see this spammed for long stretches:

7986@C13L2,PeerAddBARspActionSanity() 248: ADDBA Resp Ba Policy[0] not support

That’s the MT7986 Wi‑Fi driver complaining that the peer’s Block Ack (BA) policy isn’t supported I think. That is a constant log.

At the moment it seems like the WAN port on my Flint 2 is temporarily flapping but it’s not highlighted in the logs. I ran a continuous ping on the router when it was running on OpenWrt and it seems like the WAN was dropping during the packet loss spell.

I have also ran ping plotter and that was highlighting HOP 2 as being the source of the packet loss. That is the Virgin Media cable modem termination server.

It’s all very confusion but to summarise…

Cable connection appears to be faultless in router mode and in modem mode connected to an Asus TUF 9400.

Connection has brief moments of instabillity connected to WAN1 on Flint 2. 90% of the time the connection is performing well but there are brief spells of packet loss throughout the day……not seen on the VM router or the Asus.

How do I prove without doubt the WAN connection is either dropping or erroring on the Flint 2 then?

16

Hmm, what model is this isp router?

Do you think it may contain a puma 6 chipset?

These are mainly very bad especially when there is constant load if you ping long enough you randomly see 1000ms latency.

But you also say that Asus does not cause problems, my guess point then to mtu, likewise you may want to lower it by 50, so 1500-50 and then see if this solves behaviour patterns, you can decrease until 1280 then ipv6 would fail, please do not use wireguard doing this, purely for internet testing.

It could be that the isp limits something on mtu here and this then drops packets.

17

It is a Virgin Media Super Hub 5 F3896LG-VMB.

It uses a broadcom chipset. Virgin moved away from the troublesome puma chipsets.

I have used my Flint 2 with this Superhub in modem mode for 2 years without a single issue. It also ran on Openwrt 24.10.6 perfectly for nearly 2 months.

The upgrade to 25.10.2 seems to have broken something and it appears going back to GL-iNet firmware hasn’t fixed it.

I have tried reducing the mtu to 1492 and 1472.

Packet loss is still present.

IPV6 is disabled as Virgin Media UK are IPV4 only.

1 Like
18

Did you cross upgrade also between op25 and gl firmware?

I think this can cause some issues it is better to do a clean upgrade without settings kept.

OpenWrt OP versions or even vanilla towards the Mediatek SDK versions (their default mainline firmware) and vice versa will come with incompatible configs.

The wireless drivers also have different names breaking configuration.

Other than that it sound like a OpenWrt issue or GL issue I do not know if you refer to vanilla OpenWrt or GL's version of OpenWrt (not the Mediatek one).

19

I went from Openwrt 24.10.6 to 25.10.2 using attended sysupgrade.

The packet loss started straight away.

I then went to GL-iNet Openwrt OP24 and did not retain settings. Then to GL-iNet 4.7.7 - did not retain settings.

I then flashed GL-iNet 4.8.4 via Uboot yesterday.

Packet loss present on every single version.

20

IMG_1252
IMG_12521241×453 86.3 KB

I have a broadband quality monitor running on the connection. It pings the WAN constantly.

The first big red bar was me upgrading the router from GL-iNet 4.7.7 to 4.8.4. You can see the packet loss at just after 5am and after 6am this morning.

Second red bar was me switching to the Virgin Media router.

As you can see, no packet loss at all since using the cable modem as a router.