Flint 3 Issues / Feature Requests

[fhteagle]

Greetings all, new account and first time poster. I have had qty 2 Flint 3 for about 2 months now, running in AP mode. Longtime user of linux, dd-wrt, and built up custom SOHO router/firewall/AP using off the shelf hardware in the past. But relatively new to the OpenWRT ecosystem.

So I have a pile of bugs that I would like to write up here. Some have been sent into gl.inet support over email, but I do not think they truly understood the issues based on their responses.

Bugs and issues

1. The ancient kernel (5.4.213) in FW 4.7 through 4.8.2 … First of all how many security issues have been closed in the linux kernel since 5.4 was current? But the bigger issue is how it contributes to #2. I get this is sort of forced by the Qualcomm SDK… but this is pretty insane that a kernel this old is being pushed as current and secure.

2. NO BTRFS SUPPORT?!?!? I have btrfs-progs and kmod-fs-btrfs add-ons installed per GUI and per opkg on command line. Rebooted since install of these. But a known good btrfs external hard drive is incorrectly recognized as an NTFS drive (it is not). Tried with a newly formatted USB drive, with first MSDOS then re-done with GPT partition tables, same results.

3. MLO OPERATIONS. These are improved in 4.8.x, but still has issues with interrupted sessions, not actually allowing MLO connections on multiple bands, etc. Testing primarily against a Pixel 9 Pro XL as well as a Framework Laptop 13 Intel + Arch Linux + [ Intel BE200 || Mediatek MT7925 m.2 cards ]. Same results with all, usually have only 1 connected stream, and a second connected band immediately cuts stability. Seems worse with 6Ghz allowed in the MLO mix, but deselecting this does not completely solve the issues either. Not completely unusuable, but definitely not as reliable as I would like.

4. IPERF3 is basically unusuable with the Flint 3 as the sending or receiving endpoint. Mid 300mbps seems to be the most the CPU can create / ingest for this speed test via either WLAN or ethernet connection. When I iperf3 past the AP to another faster machine, I often get the expected ~2.3gbps over ethernet instead.

5. Throughput drops for no reason to ~200mbps past this AP, as if it was being traffic shaped to exactly that value. PHY rates are well north of 1gbps on 5ghz band according to wavemon, but iperf3 through to the faster machine on my LAN shows a very stubbornly capped 222 +/- 1mpbs. Reboot may improve throughput initially but it seems to fail down or get choked to this value with tens of minutes of uptime.

6. One of my two units is advertising itself as an SMB server, even with this deselected in GUI Applications > Network Storage > File Services tab

Feature requests:

A. In AP network mode, allow for VPN client connection + guest WLAN network that tunnels all guest WLAN traffic through the VPN interface. This is possible using GL.Inet GUI in router mode, but is not implemented in AP mode. Probably possible to do this in luci as well, I just have not spent enough time trying.

B. In Router network mode, allow a VPN connection be a failover WAN source with fail back to primary wired WAN connection. Yes this is a weird goofball edge case that is not likely to be applicable to too many topologies.

Apologies if these have been reported/discussed elsewhere here. I did some basic site searching but did not stumble on actual solutions for most of these. But if anyone has solutions/recipes, please let me know.

Also, @bruce , this discourse instance threw an issue when I was signing up for an account this morning. It said, “no more registrations allowed from this IP address”, even though I am on my own public static IP that serves my household only. Confirmed I was not being routed through a VPN by doing ``curl ifconfig.me`` , which returned my ISP assigned public static. So there should not have been any past registrations from this IPv4 address.

[bruce]

Hi,

Thank you very much for your attention and love.
We will continue to improve firmware functions and remain safe.

Bugs and issues:
A1, as we know that the kernel version of Flint3 is ancient, but in fact there are no risk vulnerabilities in the overall firmware. We will add patches for the firmware or SDK to fix vulnerabilities, so the firmware is safe.

A2, we will check this issue.

A3, you can try to choose 5GHz + 6GHz for MLO. Currently, MLO compatibility is slightly lacking, especially cross chip brands. Please connect 6GHz or 5GHz to fixedly use at single frequency.

A4, please install an OpenSpeedtest server to test network speed. In some wireless testing environments, iperf3 may not be able to fully run the bandwidth, since the compatible problem.

Script: OpenSpeedTest Installer for NGINX on GL.iNet Routers

Easily deploy OpenSpeedTest with NGINX on OpenWRT-based routers (GL.iNet, etc.) Features Installs and configures NGINX to run OpenSpeedTest Custom NGINX configuration that avoids conflicts with the GL.iNet web UI Installs to /www2, with automatic detection of available storage space Supports symlinking to external drives (e.g. SD cards or USB) if internal space is insufficient Creates startup and kill scripts for b…

A5, this question may need to be read in depth. You can send this question individually and contact [email protected]. We will have dedicated technical experts to follow up.

A6, SSH to router, check whether the samba sever instance is still running:

/etc/init.d/samba4 status
ps | grep -E "smbd|nmbd"
netstat -tulnp | grep -E "139|445"

If the samba server instance is not running/inactive, it should not advertise. Is there another samba server?

Feature requests:
A, AP mode belongs to layer 2 (bridge), while applications such as VPN feature require layer 3 (routing), so in AP mode, it is impossible to enable VPN.
If you need to use a VPN feature, only in router mode.

B, this request doesn't make sense.
When the WAN connection fails, the VPN tunnel will also be disconnected, and even if fallback the Internet will also not work.
If you have any other needs, please clarify again.

About account sign up, please PM me about your Internet export IP, I would like to check if the discourse server configuration is too strict.

[9b9e69c2-4b75-4420]

I'm also 'stung' with the stale kernels even on the current beta firmware for my Slate AX. Even if proper btrfs-progs were in the GL.iNet repos BTRFS requires kernel 5.5+ for all checksum algo support. The sick irony is that my EOL'd Certa (GL-AR750) from 2017 has an approp. one but is built on OWRT 22:

root@certa:~# cat /etc/glversion && grep 'PRETTY_NAME' /etc/os-release && uname -a && nowutc
4.3.25
PRETTY_NAME="OpenWrt 22.03.4"
Linux certa 5.10.176 #0 Sun Apr 9 12:27:46 2023 mips GNU/Linux
20250902T090334UTC

root@slate-ax:~# cat /etc/glversion && grep 'PRETTY_NAME' /etc/os-release && uname -a && nowutc
4.8.0
PRETTY_NAME="OpenWrt 23.05-SNAPSHOT"
Linux slate-ax 5.4.164 #0 SMP PREEMPT Thu Jul 24 01:15:41 2025 aarch64 GNU/Linux
20250902T090410UTC

... & I'm still ignoring OWRT 23.05 is EOL as of last month.

[bruce]

Hi,

About file system btrfs, update:

R&D team has conducted depth investigations:
If the firmware enables to support the file system btrfs, flash will be used as memory when the memory is full.
Fast reading and writing will lead to the flash reduce its life.

In view of this situation, I think we would not support this in the subsequent firmware.

If you need to use this file system, please install this attached firmware:

be9300-4.8.2-799-0908-1757323238.bin

Truly sorry.

[9b9e69c2-4b75-4420]

To clarify I'm not asking for btrfs to be supported by GL. What I am asking for is a Linux kernel 5.5+ for (in my case Slate AX if not others) & the accompanying kmod-fs-btrfs & btrfs-progs + related dependencies to be added to the GL repos. The intent is to have access to a fully modern file system available to install to use on an attached SSD/NVME via USB as would be expected by running a current stable release of OWRT 23.05.x. I would not expect such an advanced file system to be on the device OOTB.

As I've said here, elsewhere, the stale kernels & outdated repos GL builds on is very much a limiting factor to these devices. Eg: OWRT 23.05.6 is the last release of the 23.05 series but GL is still building on 23.05-SNAPSHOT — not even a stable 23.05.0 which was released 2023 October. OWRT's 23.05 series is fully EOL as of 2025 August.

GL.iNet needs to close the gap between whatever they're releasing & Upstream's releases. A window of one or two services releases behind is one thing but whole version numbers and near two (2) years out of date isn't doing the credibility of a ODM in business some 15 years any favors.

Pass that along to your PMs, please, if I haven't asked already.

Celebrate 15 Years of Innovation with GL.iNet!

"Meaningful connections" ... do you mean like git clone from Upstream so we're not all stuck with GL.iNet's stale repos & EOL'd OWRT? root@slate-ax:~# cat /etc/glversion && grep -m 1 'PRETTY_NAME' /etc/os-release && uname -a && nowutc 4.8.0 PRETTY_NAME="OpenWrt 23.05-SNAPSHOT" Linux slate-ax 5.4.164 #0 SMP PREEMPT Thu Jul 24 01:15:41 2025 aarch64 GNU/Linux 20250911T104558UTC