OpenWrt 21.02.0 first release candidate

Agree too,
What gets me most is that GUI's in general should be used to create an abstraction layer between the underlying tuning, tweak and config settings and the function it should accomplish, like switching between router and bridge function.
Currently the guy is almost a 1 to 1 mapping of the items in the various /etc/config files and the guy web pages, no abstraction what so ever. This is fine for techies but not for an average person.

I think this thread about testing a release candidate is not the right place to discuss new features.

The key change in 21.02 is DSA support.

4 Likes

Thanks again. I don't want to offend anyone with my opinions.
We say in my country ... Man does not live on bread alone.
DSA is important, other topics too.

1 Like

Requesting and implementing new features is not offending. The feature set of a release candidate is more or less complete and should focus on stabilizing the already existing RC code.

I suggest to continue working on the quick configuration page in the master branch, not in 21.02 branch.

DSA is important, other topics too

I didn’t comment the importance of your feature request. The focus in this thread is the already existing 21.02 release candidate.

3 Likes

Sorry this was an unintended example of user confusion.

The helper functions work as designed -- as long as the raw and helper kernel modules are installed. I confused myself with obsolete references to modules which were enabled by default -- the current installation does not install ANY which would have been just what I wanted if it were not for the obsession with making sure the defaults were disabled.

The error logging for missing helpers in fw3 is better than the silent behavior for missing kmod-ipt-raw. In both cases it would be handy to see fw3 output in gui but I can live with examining results to see if I got what I expected now that I know that is a good idea.

Wifi performance seems improved in 21.02.0-rc1 relative to 19.07.7

I'd run some quick tests a while back when overclocking my Archer C7 v4 with ASSIA's WiFi Sweetspots app, to measure if any performance improvements were real and not just a cosmetic change on reported CPU speed.

5GHz on a 2x2 phone at ~2m, direct LOS to the router
Stock (775MHz): ~530 Mbps
1000MHz: ~620 Mbps

This same test, at the same distance, at 1GHz, but 21.02.0-rc1 running on the same device, nets me ~730 Mbps, with 762 Mbps peaks. Link rate as shown in LuCI is 866Mbps as expected, 80MHz channel width.

That's a ~100Mbps improvement on an old single core MIPS device with an old radio (QCA9880). This of course is an isolated, synthetic test, but impressive nonetheless.

Lovely work!

As far as simplified interfaces go, there is also this: https://eko.one.pl/?p=easyconfig

Use Google translate if needed. Not sure if the UI itself has been translated at all. Looks pretty nice!

1 Like

Is this release coming out for fritzbox 4040?

It took me a short search to come to:
https://firmware-selector.openwrt.org/?version=21.02.0-rc1&target=ipq40xx%2Fgeneric&id=avm_fritzbox-4040

1 Like

Not now (release candidates), not anytime soon, or not ever? Because I explicitly bought a router (ER-X) where hardware nat offloading works fine, and if that won't work with DSA then upgrading to 21 is a non-option until that's solved.

Hardware offloading does work fine under DSA, as long as you don't use PPPoE. Hardware offloading with PPPoE is possible as well, but this only works under the master branch.

1 Like

Just flashed rc1 on redmi ac2100.
Wifi wireless device -> wired device is limited to about 300 mbit/s. Other direction works fine and reaches about 630 mbit/s, which indeed is improvement, older snapshots peaked around 560 mbit/s.
In both cases cpu is very loaded by kworker/napi_workq and mt76-tx kernel threads. Normally dump ap setups should not have any cpu load at all. Some older builds worked in that way.

Thanks
It took me a while because I have been testing the "easyconfig" interface. Not bad, it's an interesting idea, but it's only in Polish and you can't change the language. Also the weak point is that the configuration is "too" simple and it seems that some configurations are going to be left out.
The most interesting thing I have discovered is that the website that hosts all this is of Cezary Jackiewicz, one of the Gargoyle developers, and they have a fairly extensive list of routers supporting the experimental version 1.13.0, which is not yet available in the official website.
Take a look :wink:

Negative real time speed rate not yet fixed

Yes, unfortunately, the wire speed on my FritzBox 4040 with OpenWrt 21.02 dropped to about 650 Mbps, and with versions 19.07 it was over 900 Mbps.

3 Likes

Excuse the offtopic but that rate menu does not appear either on my x86 or on a wdr3500.
Is there a specific package for this functionality?
Been using bwmon for time being, but that screenshot looks more promising.

On topic, been using the 21.02 snapshots for a bit, haven't encountered any irregularities so far with my config.

Great work as always.

In RC1. If I change the timezone in Luci, then Luci deletes NTP server and client settings from system config file. But the checkbox in luci are afterward still set as usual with client=on and server=off. But the system config file only has the ntp servers left?

Are you sure the “negative download rate” isn’t actually upload speed without definition as it say so in the headline?

I believe this is normal behavior. Luci is programmed to remove some config options when they are set to the default value. It is simply that the default system config file has extra lines that are not necessary.

However in the future....I would like to see the config files and LuCI be better in sync, so that LuCI is not removing things every time you save config....

Comments come to mind. I stopped using the GUI years ago as I like to have multiple possible configurations by commenting out the stanzas for configurations not currently in use.

2 Likes