Flint 4 BE-14000 Discussion Thread

Router VPN, DNS, Leaks
170

Rookie mistake. I set up port forwarding on my gpon ont to flint 3 oo and fit was different than Flint 4 so it could work. Now it's fine.

171

Nice that you let us know what the problem was.
In your log I saw that at the end it said server was active so it had to be something else.

172

I 've installed dockerman but it shows only a shortcut on Luci 'services' menu not whole menu like I had on flint 3. Link lead to docker menu so it works but I wonder why it looks different now.

obraz_2026-10-17_214602708

It looked like this:

image
image287×321 10.1 KB

173

My guess is that the one on the Flint 3 uses an older OpenWrt base which mean they compiled older version.

Especially if you use the newer betas on Flint 4 the OpenWrt base version is more updated.

Though, I also noticed some other missing parts on luci aswell like the dhcp options being very stripped down, from what you can actually get from luci even on older OpenWrt bases.

So some things can be also be vendor sdk specific too.

174

THis is what I see on my F4 B4.

image
image1629×1203 59.4 KB

175

Yup. dockerman work but shortcut on luci menu not. I think it is snapshot 25 related.

How is Your ram utilisation?

I have 95% ram used after samba transfer to hdd or other ram heavy task like docker instalation.

Ram in the GUI shows normal usage but elevated cache. 56% screen is from RouterPilot and and it report normal cache. Is it a bug or normal behavior?

Zrzut ekranu 2026-10-18 143730
Zrzut ekranu 2026-10-18 1437301178×258 4.66 KB

Zrzut ekranu 2026-10-18 142401
Zrzut ekranu 2026-10-18 142401246×181 9.38 KB

Zrzut ekranu 2026-10-18 142348
Zrzut ekranu 2026-10-18 142348519×206 7.03 KB
Edit: I have 1gb cache after reboot and only wireguard remote access.

I think it started when I've plugged in external storage.

176

It would be nice to see a router that you can add your own memory modules and you would only need DDR4. Imagine that, GL.iNet could be the first to implement it and start a new wave of super routers.

Heaven.......

177

and whatever NVME/SSD storage you want as well.

178

Even Turris Omnia by CZ.nic does not have the possibility to add RAM (but you can add SSD or change wifi card).

But it would be great to have such router with possibility to add RAM, SSD or change wifi cards etc.

179

They have this - you can install router software on a minipc and do everything you are asking for lol and might even save a little money by doing it yourself.

180

Hello everyone,

I am experiencing persistent wired network disconnections on my GL-BE14000 (Flint 4) running the latest beta firmware based on OpenWrt 25.12.

The Problem:
Every few minutes, the wired LAN connection completely drops for a few seconds and then reconnects. During these drops, dmesg shows constant PHY link state changes (link down / link up). Wi-Fi, however, works flawlessly without any disconnections.

Troubleshooting I've already tried:

  • Disabled Hardware Acceleration (set Network Acceleration to Software/None).

  • Disabled all proxy/VPN software (Tachyon).

  • Tested different Ethernet cables (known good Cat6).

  • Tested different LAN ports on the router.

None of these steps solved the issue. The drops persist on the wired connection.

Relevant logs from dmesg:

kern.info kernel: [ 2491.944302] yt92xx yt92xx.0: eth1.1(port 2, lan3) link down
kern.info kernel: [ 2494.024730] yt92xx yt92xx.0: eth1.1(port 2, lan3) link up, speed 2.5Gbps
kern.info kernel: [ 2528.344233] yt92xx yt92xx.0: eth1.1(port 2, lan3) link down
kern.info kernel: [ 2531.464673] yt92xx yt92xx.0: eth1.1(port 2, lan3) link up, speed 2.5Gbps

(Additionally, in earlier logs, the system showed NETDEV WATCHDOG: transmit queue ... timed out followed by warm reset failed).

My Firmware Version: 4.11 beta 4

It seems like there might be a driver regression in the mtk_soc_eth driver in the OpenWrt 25.12 branch. Is anyone else facing this wired connection issue? Are the developers aware of this, and is there a fix planned in the next beta release?

Thanks.

181

I also get the best performance with Parallel requests. In addition, I enable Optimistic caching.

Screenshot_20261021_181506_Chrome
Screenshot_20261021_181506_Chrome2767×881 160 KB

182

I am using ADGH Beta, load balancing, with the average upstream responce time for the past 24 hours being 6ms. That works for me.

183

Hi,

Thank you for sharing the details of your troubleshooting.

To better understand your situation, could you please provide the following information?

  1. Is your wired device connected directly to the Flint 4, or through a switch?
  2. Could you please connect another device using the same Ethernet cable and LAN port to check whether it experiences the same disconnections?
  3. When the issue occurs again, could you please log in to the router via SSH and execute the following commands?
ip -s link

Please also export the logs and send them to us via private message.

Thank you for your assistance.

184

It was fixed by changing WiFi region in openwrt via Luci

185

New firmware version released:

GL.iNet GL-BE14000· 4.11.0 beta5 build 1114 (2026-10-23 19:09:02)·OpenWrt 25.12-SNAPSHOT r32295+917-c25265953b·Linux 6.12.94

186

On GL-BE14000 Beta 5, the Wi-Fi driver generates significantly fewer errors than Beta 4.11, but the issue is still not fully fixed.
For example, ra0 now shows 167,039 RX packets with 32,664 errors/dropped, and I still see repeated kernel log messages:
WiFi@ERROR.PROTO,eht_dsc_bitmap_remap_non_eht_bw() 180: hc_radio_query_by_wdev failed!
This looks like a remaining MT7990/MT7996 driver issue.