Any plans for Realtek SOC support?

Realtek are lazy.

Why bother adding your hw to the Linux kernel, if you don't have to ?

Vote with your wallet, and stay away.

2 Likes

No one prevents you from picking up the development, but after you have done the work on the SOC- and toolchain (gcc, binutils, musl) side, you'd be still left with the sorry state of Realtek wireless drivers for Linux.

If you're looking for a timeframe less than years-at-best-but-more-likely-never, there are options that work right now.

1 Like

OK, sorry to open a can of worms :pensive:

I agree with your stance regarding the Realtek Wireless HW/SOCs, and also do now want to 'reward' their behavior, but when it comes to Managed POE-Switches with Openwrt-support (and POE-Powermonitoring/-switching), they seem to be the only viable/available option right now (unfortunately).

I'll be honest here. This all started for me today from a few things:

  1. I keep seeing people asking can we support this router and that router, and I understand and fully support why there are many routers that are currently not supported and that the chip manufacturers play a large part in that. I was thinking the forum should have a sticky post with a long list of routers that are unlikely to be supported, and the reasons why, partly to stem the questions. I still strongly think that is a good idea. Perhaps it could even be part of some sort of PR push?

  2. The first point came to mind as I googled ax routers and found a bunch that seemed a step cheaper than the netgear wax202 that I think is currently one of the best OpenWRT options in this space for its price, at least in my part of the world (Australia). I noticed that prominent among the slightly cheaper AX routers brands here is Tenda, and I started ruling out models one by one because of the chipsets that I could see on the FCC site showing either Broadcom or Realtek. I then thought of the idea in point 1.

  3. I realised that there is at least some support for Realtek, even if minor, because I know my Netgear Wax206 has a Realtek PHY for the ethernet 2.5Gb WAN port. When I searched I found this thread and at first I thought it was just confirming what I thought was the OpenWRT/Realtek story, but then it seems that some amazing people made a lot of amazing progress, only to get stuck at a certain point. Amazing people were still active on this topic over 2.5 years until late last year. I was inspired by this and wondered why it stopped when such great progress had been made by these amazing people.

  4. In my googling, I realised that not only are brand like Tenda with their Realtek chips slightly cheaper in the west, but I kept seeing stuff about India and other third world countries. Perhaps some of this amazing work that is so close to bearing real fruit could also be a real win not just for a stingy westerner such as myself, but for a whole group of people who currently don't seem greatly represented by or within OpenWRT. Perhaps it might even stimulate people (who are not westerners) to jump on board because they could be a pay off for them with hardware that is available to them at a reasonable price?

  5. I fully agree that open source unfriendly manufacturers should not be rewarded. I just saw a possibility of a coming together of a bunch of things in a way that might lead to something good. Sometimes things like this can happen for the general good in spite of the companies and the economics and the politics. I mean who ever saw Rick Rolling relaunching a pop star's dead career?

  6. No, am not a programmer and no I don't know all the nuances of the toolchains etc, but I saw something really positive in this thread. I wish I could personally run with this a bit more. I meant no disrespect to the community and the current and previous devs and I certainly did not want to appear demanding, only amazed and encouraged by this thread. I deeply respect the devs that have brought and keep bringing us OpenWRT, and I deeply respect the blood sweat and tears that the people involved in the amazing story of this thread alone have put into something that almost got there for the greater good.

Peace and thanks.

1 Like

I would not exactly call Realtek "opensource unfriendly[0]", they tend to meet the license requirements. But -until recently- that didn't imply contributing their SOC/ (USB in particular-) wireless support upstream, but merely throwing a(n opensource) vendor driver over the fence and immediately forgetting about it and doing the same for the next hardware generation. Legally meeting the requirements, but neither helping their users, nor themselves.

But with Realtek, we were long plagued by their use of the lexra ISA (patent mine field, 'thanks' to Imagination Technologies, and zero linux/ libc/ toolchain support), the wireless vendor drivers (~2.6.13 era ieee80211softmac forks) without nl80211 support (and custom hostapd interfaces/ ancient forks) and typically very low-end hardware (SOC speed, flash, RAM). The combination of all these makes it very hard to work on it - and the low-end hardware successfully prevents anyone serious to want buying these devices.

Realtek having moved away from lexra to full mips (or ARM) and starting to care more about mainline contributions (e.g. rtw88/ rtw89) is promising, but after them being a non-option for one-and-a-half decade, they are far behind in gaining volunteer cooperation. Being cheap and plenty does not (always) offset unusable (but technically available) driver support and usually very low-end specs (volunteer developers are more willing to move away obstacles for new/ high-end/ challenging hardware that will remain relevant for longer).

From an enduser's perspective none of this really matters - the facts just are, these SOCs currently aren't supported, there is no one actively working on them either - and all realistic expectations imply that they never will be. From time to time miracles may happen (e.g. rtl838x/ rtl93xx), but betting on them before they are actually materializing as pre-merge PRs would be 'misguided' and a game prone to lose.

If you're looking to actually use your devices during their prospected life span, look at already fully supported ones only.

--
[0] that would be more fitting for Broadcom-, NXP- or Maxlinear wireless.

3 Likes

Thanks for your thoughtful response. I know you are with me when I wish it were otherwise and what you say makes sense.

I just got excited over this thread and took it as too hopeful a sign. Man, these guys put in some great work though!

I really want to do that sticky topic now, including some of the stuff I've learned.

How i can check it?
I see config of main branch and can't see RTL819xx

might need to enable experimental kernel

1 Like

Note:

That website is for switch SoCs of Realtek (RTL83xx, RTL93xx), not for router SoCs (RTL819x).

1 Like

yeah, link deleted.

Can you tell me how to enable the experimental kernel?

better How to compile with kernel 6.1? (noob question)

Unfortunately I don't see this option. I also don't see CONFIG_TESTING_KERNEL in .config

I'm using the main branch of the repository https://github.com/openwrt/openwrt. Maybe I was wrong here?

then try setting the param directly in the .config file, or master is already running as high as you can go.

No difference. They probably don't have RTL819x support from the post(
Ok, I'll try https://github.com/Alexey-Tsarev/openwrt-rtl819x

Current official source doesn't support Realtek router SoCs (RTL819x).
You need to clone 3rd party repository instead, like the following.

You may have to select the main/master snapshot branch before launching the configuration.