Adding OpenWrt Support for Netgear RAX120 (Nighthawk AX12)

Could you upload your version (sysupgrade) somewhere on the server and add a download link?

I uploaded it as a pre-release to github. Works for me but otherwise completely untested and minimal build.

1 Like

I work your bild but I think the wifi range is significantly worse than the original firmware. In addition, in the original firmware on 5GHz, 1 radio is visible and on openwrt we have 2 radios. Why is that?

WiFi is not my strong point, I hope someone else can chime in.

I assume you adjusted the power for the radio? According to the wikidev it got two 5GHz chips, not sure how stock represents them.

RAX120 has two 5Ghz Radios and ath11k represents that as two independent devices both capable of 4x4 streams.

Stock firmware use proprietary QCA-WIFI that is combining that two radios to one device with 8x8 streams.

Please note that each radio is working in a different frequency range, one in the lower 5Ghz band and one in the upper band.

1 Like

just new here, and have a couple of questions:

  1. does this solve the random rebooting issue in the official firmware?
  2. is openwrt on rax120v2 stable enough for daily use?
  3. looks like official firmware is also based on openwrt, are we using the same stock radio driver?

thanks!!!

Looks like no movement on github for weeks...

How's the stability @patient0 ? Should we work with @robimarko to upstream ?

The custom QCA firmware that combines both 5G modems into one is also an interesting point to explore, but it's probably out of scope given there is no upstream support for it.

I left comments and they need to be resolved, author has not been active so far.

@nschimme I don't use it as my daily device and don't know about stability.
@robimarko @jewest added me as an collaborator back in January. I already made the changes you requested in my fork of his (I DM's him three weeks ago to ask if it is ok if I push the changes, no answer). So I can push them, what is preferred a forced push or an additional commit? I'm a dummy, I'm on holiday and while I can apply and push the commit, I can't test it. Only my trusty APU2 got with me on holiday :slight_smile:

I wonder if anyone can confirm (e.g. @jewest @patient0) if the jffs2 persistence related problem (mentioned by @NastyNeightK and me) happens on your side with an old or a newer (e.g. rebased and compiled in 2023) branch.

I am interested in giving this problem another shot, but before that I'd like to confirm if it's a unsloved problem for everyone. (is it already fixed?) (or is it a problem happens only on my and @NastyNeightK side bc of mis configuration?)

I haven't tested. But the method itself is pretty generic, it should work regardless of branch version: either work directly or with very light work to port to new branch.

I does work for me (with the commit/repo the image mentioned in the post 3 three weeks ago is based on). But it's a regular UBIFS image not jffs2 anymore, which seems appropriate for a NAND based device.

(how do I reference a previous post, any way to shortcut like #185?)

select the text with mouse, then click the "quote" popup button. like in the screenshot:

Thanks for confirm. If it's UBIFS now then the original problem shouldn't exist anymore...

Looks like the comments that @robimarko is referring to are here and here: https://github.com/robimarko/openwrt/pull/43
https://github.com/robimarko/openwrt/pull/45#event-8170810551

@patient0 I guess it's on you to resolve them?

Hi @nschimme, no he's referring to

After @robimarko ipq807x work was merged in OpenWrt a new PR was opened directly in OpenWrt. I'll be back from holiday this week and will look into it then.

1 Like

Thanks for solving this mystery for me. I bought this router 2 years ago because of the 8x8 Mu-MIMO but returned it for subpar performance. There was only one other vendor using this "8x8" MU-MIMO, a ruckus that got paused for production/discontinued. It was weird -- why a enterprise product suddenly get unavailable when it was the most expensive product. So it was red flags on top of red flags, the main red flag being why signal didn't improve over 4x4 products. Now you solved it for me! --> that the 8x8 was a software hack stitching two 4x4, not a true 8x8.

A quick question:

I just bricked my RAX120v2 by using nand erase 0 7e00000 (should have used nande erase.spread after setting the offset; or the proper offset). No serial output of course, all LEDs are on but dead otherwise since I erased the boot partition+.

Looking at the datasheet of the Winbond W29N04GZBIBA (according to Wikidev) there seems no easy way to flash it for a semi-dummy like me. I got a flasher for motherboard BIOS' but nothing more, and I'm more of plumber then an electronics guy.

Any thought of someone with better knowledge about in that area?

Do you mean you were able to solder the W29N04GZBIBA NAND out of the board and trying to program it with the flasher?

No no, pretty much what I wrote: I'm really really bad a soldering, I even have a hard time soldering TTL (to another RAX120v2 I found used) let alone anything smaller.

My question was more is it possible for someone who is good at soldering?