I’m having the same phone. I will try out 160 MHz tomorrow. How often do you have the disconnects? Is there any error message in the WiFi menu of android?
This also happens to me using a P6P with my Flint 2 using 80MHz running the 4.5.2 it shipped with. I have to set it to auto reconnect on my P6P as a temporary fix. I haven’t tried 160MHz and have only had it set to auto WIFI channel.
Hmm i noticed today something strange.
I got one flint 2 as isp modem with a OpenWrt snapshot, and one as dumbap with oem software, ive left wan alone and lan aswell, only for wifi i created my own vlan dhcp clients.
I noticed alot of strange dhcp issues variating to constantly renewing a lease on wan (spamming, which i confirmed on my isp modem when other devices dont behave like this) to not getting a lease after a dhcpnak happened and then i got very slow and halted connectivity, it resolved when the dumbap router restarted, this only happened when i was watching a very long youtube stream of a couple of hours over wifi 5ghz.
I wonder though is dnsmasq still fully up to date ?, it looks like some type of regression or hang within dnsmasq 
This was on firmware 4.5.6 release 1.
On wifi i had roaming active.
For the spamming this does not always happen, but it can after a while go into this state.
It could also be something else though, roaming (related) not flushing the clients, since i saw them increase in nlbwmon.
Model: GL-MT6000
Note: There is nightly snapshot build, it’s not stable, using on your own risk.
Version: 4.5.7
This file is for local upgrade in web Admin Panel and Uboot.
Date Compiled: 2024-01-26 20:21:48 (UTC+02:00)
SHA256: 7ff4abf29bced7e7784f0e003989fb9583844d3d324ddf0bd4a1d47259042dfc
4.5.7
- This is snapshot firmware, generated by an automatic compiler, please use caution
4 Likes
If it is this version to correct 2.4, unfortunately it was not corrected, it is still limited to 100mb with ping and latency at the top…
4.5.7-snapshot
used the mt3000 version.
EDIT: oh, the s21+ and s10 don’t exceed 100mb on wifi, the problem is on the cell phone… other people should test it, other than a Samsung device, which I think 2.4 must have solved…
Edit 2: Because it’s a snapshot version, I can’t give much feedback, but they removed fullconenat, removed the option to change the wifi region (this can be returned as it is used on the mt3000), updated the ddns package to correct the problems that were happening on another topic.
improved speed in wireguard connection.
on the other hand, I have this kernel bug, it seems to be related to wifi 2.4 because I’m using channel 11. I noticed that this new version is using a lot more ram than 23.05 - 4.5.6
Edit 2: After restarting I no longer have logs in the kernel
3 Likes
Using 4.5.7 now so far working ‘ok’, though ive not tested speeds.
But i did found a other issue here:
These ssids spawned here, i upgraded with keeping the settings.
This shows that they are working on the driver.
Unfortunately there is no way to turn those SSIDs off - they are spawned by the driver itself, IIRC
1 Like
Finally 2.4ghz have a good speed on Android (although high latency but it’s a huge improvement) on 4.5.7
Now wifi 5ghz seem to be a bit lower signal
1 Like
The speed on 2.4 is much better for me to on 4.5.7
Is this FW using the closed-source drivers?
Yep it seems it’s closed source
Google Pixel7 Smartphone disconnected once again, when using the 5Ghz at 160MHz bandwidth.
While the network is on 80MHz, it keeps connected without any problem.
The log is always the same:
A lot of DHCPRequests and DCHPack and then it’s disconnected by inactivity, as reported here:
The transmit power for 5G WiFi is 11dbm. For 2,4 G its 14dbM. (according to Wifiman)
This is below the maximum value (20dbm for 2,4 and 30dbm for 5GHz). I tried to change the country code in the config but it has no effect. I’m sure this will be fixed soon.
On my AX210 WiFi card iam also getting around 200MBit download on the 2,4 GHz WiFi.
Seem to having issue brining up 5Ghz with 160Mhz bandwidth on 11 ac/ax only.
IPv6 is hit and miss with pass though, but I can get it to work.
2.4Ghz seems much faster on my clients and running 20Mhz 11 n/ax. Previously I have N clients that would not connect using this mode, but had no issues connecting on N with an another Broadcom AX 6000 router and a different Mediatek chipset AX 3000 router.
Can’t do the range test unit I get the 5Ghz up again. I maybe have to drop out of the DFS range and see if that’s the issue. I have it up with channel 128 and 60, but 124 and 120 seem to not want to come up despite………UPDATE:Correction, it just came up. It only took 24 minutes so something is still not correct with that. It shouldn’t take that long to do a DFS scan when I am out in the country and have WiFi scanners running on two other devices so I know those channels are free.
I would say for the most part it seems to be an improvement with 2.4. IPv6 mighty be unrelated to this snapshot firmware. My big concern is the amount of time to 5Ghz on DFS to come up.
I’ll do a range test for distance and compare to what I have documented on previous firmware for the flint2 and the Broadcom chipset AX router I have.
Almost forgot to mention I had two clients (one apple iPad mini 6th gen and one google pixel 7a) that would hardly get any speed on 2.4Ghz if the Bluetooth was turned on at the same time. Both of these clients now seem to work as expected with this snapshot.
So this was fixed with those two clients and 2.4Ghz 11 n/ax client that used to fail to connect are now able to connect under that mode.
Please keep in mind that snapshots (also called nightly builds) are not beta versions. It is best to be patient until there is a new official firmware from a GL employee.
Posting all the bugs from the 4.5.7 snapshot here now makes absolutely no sense.
So it doesn’t make sense to post it so GL employees see what we are seeing and they can possibly correct it before releasing the Beta or official resales that they talked about releasing Monday?
Personally I’m just glad to see 2.4 working better and reaching speeds I have never seen off it before.
1 Like
Seems they have logging set to debug on the snapshots with an increase buffer size to 10240, This is probably the increase of ram usage you are seeing.
1 Like
Yup, makes no sense. This release was not checked by a human. It’s just a bunch of the current code commits.
1 Like
To use the proprietary MediaTek WiFi driver they’ve downgraded from OpenWrt 23.05.0-rc1 all the way to OpenWrt 21.02.0-rc1. But OpenWrt 21.02.x reached it’s end of life in May 2023 
4.5.6 stable release:
OpenWrt 23.05-SNAPSHOT, r23001+721-38c150612c
4.5.7 snapshot:
OpenWrt 21.02-SNAPSHOT, r15812+1067-46b6ee7ffc
Technically the builds are newer than 23.05.0-rc1 and 21.02.0-rc1, but the commit ID (e.g. 38c150612c) shows us where GL.iNet started their copy of OpenWrt. And of course they’ve then applied their own fixes and cherry picked some changes from newer OpenWrt snapshots too.
Stability with this approach is questionable though, since they forked OpenWrt 23.05 before it was stable and they haven’t kept their OpenWrt fork in sync with the OpenWrt 23.05 repository, so it’s missing dozens of fixes. For example, with firmware 4.5.6 if you open up /lib/netifd/wireless/mac80211.sh and go to line 148 you’ll notice that it’s missing this change from January 5th. But if you go to line 330 you’ll find this change from January 9th.
Have you tried using official OpenWrt snapshots, just to see how that compares? Because if it works with official OpenWrt snapshots then GL.iNet could fix the problem by rebasing their fork on a newer OpenWrt snapshot.
Rebasing the GL.iNet firmware on a newer OpenWrt snapshot might improve the overall stability, but it won’t do anything to fix the 2.4GHz issue. Although, now that we’ve got a firmware with the proprietary MediaTek WiFi driver to compare against someone could file an issue report to the mt76 repository.
8 Likes
I know it’s hard to do, but the best world will be the open source wifi driver for 5GHz and the closed source one for 2.4GHz