There's no reason to suspect it's in any way changed, it's the same (closed source, shitty) driver.
It's not an issue with the mwlwifi driver, there is a bug in 21.02 related to the firmware blob or something else that causes wifi to stall. The old 19.07.8 branch and the current Master branch with 5.10 kernel seem to be unaffected by this issue.
Thanks for the update!
So the latest snapshot should work fine?
I'm still on David's wonderful build but its kinda old now....
Yea snapshots based off Master work fine. 21.02 not so much the wifi stalls a lot. There is another thread with 21.02-rc4 with the wifi firmware replaced with an older version that seems to fix it (I've been testing it a couple days), but not enough testing has been done yet to really confirm it's the fw.
Overall
So similar to RC3 and includes few developments, thanks for contribution.
Need to show real performance of the VR200
Looks like drops problem fixed substantially.
Connection LEDs
- [ ] LEDs are dimmed except USB and Refresh. Similar to RC3-RC2 as I mentioned
-
Also have good news, it's configured well according to the connections than RC3. Thank you
Wi-Fi
The connection quality varies by every encryption of Wi-Fi and the best one is no encryption
It's similar to RC3, getting better for every release from RC2 for each device
WPA3-SAE
802.11w Management Frame Protection:
Required
-
Speed well
-
Couldn't detect any drop
-
Good surfing comfort
-
Wire comfort
- Rare device support
- [ ] Looks range affect so much. Not only speed aspect also surf comfort
https://www.speedtest.net/result/11891365013.png
https://www.speedtest.net/result/11891521227.png [* ]
[*] Only this test done away form router. Just done at next room. Doors open
WPA2-PSK
802.11w Management Frame Protection:
Disabled
- [ ] Could't take stable speed
- [ ] drops barley,
-
present similar result in every device. So stable on all devices
https://www.speedtest.net/result/11891341753.png
https://www.speedtest.net/result/11891284825.png
WPA2-PSK/WPA3-SAE mixed mode(strong security)
802.11w Management Frame Protection:Disabled
- [ ] Speed not stable, but it becomes stable with use
-
No drops, especially in the latest tests
- [ ] Looks efficient for all devices [phones and laptops]
- [ ] None of the speed test parameter not stable first test stage but geting stable
-
The smoothest Wi-Fi between all of these. [Included No encryption]
- [ ] Low speed
-
Distance not affect the performance badly
https://www.speedtest.net/result/11889367869.png
https://www.speedtest.net/result/11891563685.png
WPA-PSK/WPA2-PSK mixed mode(strong security)
- [ ] Speed not so bad
-
Almost never drop, feels good web experience but not good as cable.
-
Some device performance low but many wells
- [ ] Better surfing comfort than every security
https://www.speedtest.net/result/11890287759.png
WPA-PSK
- [ ] Not usable speed, can hardly surf on OpenWrt interface
No Encryption (open network)
-
Great, same as cable speed
-
Perfect surfing comfort
- [ ] Ping little higher than cable
https://www.speedtest.net/result/11890314121.png
Cable connection
-
So good DSL connection
-
Downstream SNR offset usable but of course not good as native performance. Thanks good work
https://www.speedtest.net/result/11890216704.png
Test system
Those all test result done one device, next to router! Wi-Fi & Ethernet
But interestingly, some different devices give worse than those results. Selected a Linux laptop which I've gotten the best result overall
I've to say that some results have changed with use of the same device. So got worse result first time usage, I think router get stable itself by time.
I've to say that some results have changed with use of the same device. So got worse result first time usage, I think router get stable itself by time.
My experiences follow yours. Especially the part about a device may be having issues but then corrects itself. However, failover to mobile data can be slow. So slow that the Android device will show "connected without internet". Reverting back to an earlier WiFI FW version helps but the wireless performance is not the same as 19.7.08.
There is a WRT3200 thread where reverting back to an earlier FW revision helps tremendously.
I'm willing to bet the LED issue is because the physical "pins" aren't defined as outputs, but are "floating". Or something like that.
21.02 RC4 successfully installed on a Netgear WNDR 3700v4 (keeping config from RC3)
Still have the 8MB image size issue.
You probably meant WRT3200, right?
Yep! Fat finger syndrome!
Thanks for clarifying!
Thanks for your replies, unfortunately irqbalance made no difference.
Pinging the device i had 0 timeouts, pinging the modem i had 0 timeouts, pinging any wan IP gave me frequent timeouts.
How can I manage it?
To clarify further, I've now been running a master snapshot for a few days and the wifi stalls have not occured on my WRT32X, so it's clearly a 21.02 related but that hasn't been fixed.
I'm just using the 5.4 kernel right now, not sure how to easily upgrade to the 5.10 kernel without compiling myself although I know many mvebu users that run master branch use it.
Peoples findings seem to vary
Interesting, that testing is more involved than my browsing around on a phone, laptop, and a MS teams meeting. WiFi with that stuff is definitely better on master than 21.02-rc. I’ll have to try that too though and prepare to be disappointed.
edit: after 2 more days of testing I can confirm WiFi is much better on a Master snapshot. 21.02 is a disaster, stalls out frequently. 21.02 was branched what, 6 months ago? At this point there must be a ton of improvements on master. Either way it's definitely a bug in 21.02.
Tested with TP-Link TL-MR6400 v5 and it's working fine for a few days now. The LTE modem on that device must use QMI and the end user experience would be complete if the Luci QMI module (luci-proto-qmi) was preinstalled too. So, would it be possible to get that in the official images?
21.02 release appears to be waiting for ipv6 flow offload problems to be fixed:
http://lists.openwrt.org/pipermail/openwrt-devel/2021-August/036023.html
Anyone know if this is reasonably fixable in 21.02, or does it require a major evolution like moving all targets to the 5.10 kernel or something? My curiosity is whether or not there are near term prospects for a final 21.02 release. If not, I'll stop waiting and flash 19.07.8 on some routers I plan to gift to someone. The routers in question are not DSA targets in 21.02, so upgrade won't be too confusing for the recipient.
To be clear, I'm not criticizing. I think the 21.02 release candidates have been an inspired step toward getting the bugs shaken out of DSA implementation especially, but also other less tested changes that have built up in the main snapshot tree since 19.07. And I'm looking forward to a final 21.02 release - it's just a timing thing for me.
So its older cheap non-MT76 devices anyway, so HW offloading isnt an option anyway. Did you tests, whether soft offloading for the devices in question has its reasonable performance gain at all, worth these thoughts?
Or why not give them away as 21.02 with offloading unchecked?
In case it gets fixed, the receipient could then checkmark a clickbox and gain a few percent of performance. The recipient would need to do occasional updates anyway, effectively become active as OpenWRT admin.
Ouch. I wouldn't call an EA8500 found used on ebay for $50 shipped "cheap." A screaming bargain maybe, and it's a bit long in the tooth I'll grant you, but still quite capable hardware: 1.4 GHz dual core ipq8064 ARMv7; 512 MB memory, 128MB flash, Wave2/WiFi5, 4x4 MU-MIMO. It's a good friend. Were they not, I'd have found them a used $20 Archer C7 .
But your point is taken. Yes, I could install 21.02rc4 and just uncheck the offload boxes I suppose.