Realtek R8188EU Driver Sees "Huge Cleanups" With Linux 6.0

[phoronix]

Phoronix: Realtek R8188EU Driver Sees "Huge Cleanups" With Linux 6.0

Along with his various other pull requests for areas of the kernel he oversees, Greg Kroah-Hartman submitted the Linux 6.0 staging changes this week...

https://www.phoronix.com/news/Linux-6.0-Staging

[Danny3]

It would be really nice if Realtek or somebody else would do something to add built-in support for Realtek RTL8812AU and RTL8812BU!

[milkylainen]

Cleaning up Realtek...

Is that the equivalent of watching a cringe episode of "How clean is your house"?

[arzeth]

Speaking of Realtek, I had 50 KiB/s, 500-700ms ping, 25% packet drop when torrenting with built-in Linux driver ("rtl8xxxu") on v5.18 on ~$10 "TP-Link TL-WN822N Version 4 RTL8192EU". I was thinking of going back to Ethernet until I discovered aur/8192eu-dkms (which is a fork of official Realtek's v4.4 driver with many user patches; there's also a fork of Realtek's v5 driver but without patches, which I didn't try). With this aur/8192eu-dkms I have 10 MiB/s max and 3.1ms avg ping 192.168.1.1; When torrenting: 17ms avg ping with frequent 50-70ms spikes.

[LoveRPi]

The sad state of Realtek's upstream. Consistently forsaken except for PCI-E.

[roviq]

I have a tp-link TL-WN725N USB Wifi adapter that uses this chipset, it works out of the box with the r8188eu module, but with the "from the staging directory, the quality is unknown, you have been warned" warning, works fine for light browsing (2MB/s direct line of sight to access point at 6 feet), but wouldn't use for heavy file transfers, it drops a lot of packets and connection dies constantly, hopefully this cleanup will help. This USB adapter works fine with *BSD too.

Also have another tp-link Archer T2U PLUS (RTL8821AU) with out of tree driver module that stops working with every new kernel release, even point-maintenance ones, and usually the fix come from another github project, I think I have used at least four different projects or forks for that one. That one gets 15MB/s. Just noticed that the dkms AUR package at Arch Linux recommends another github repository for a fix. Whee! Just build it, didn't work, but I don't need it atm, just wanted to check support.

I really hope at some point everything Realtek will be merged to something like Intel's. When possible I just pop in a PCIe slot anything between an Intel 7260 and AX210 that I have at hand, those are reliable, cheap, and get 60MB/s or more in Wifi5/ac. Before that (10+ years?), I used Atheros ath5k or ath9k, was very reliable in Linux, wonder how it is nowadays.

[timofonic]

Realtek, craptek...

[RejectModernity]

It would be really nice if Realtek or somebody else would do something to add built-in support for Realtek RTL8812AU and RTL8812BU!

Oh man, I have RTL8812AU and RTL8814AU, such a shame...

[tildearrow]

I really hope at some point everything Realtek will be merged to something like Intel's. When possible I just pop in a PCIe slot anything between an Intel 7260 and AX210 that I have at hand, those are reliable, cheap, and get 60MB/s or more in Wifi5/ac. Before that (10+ years?), I used Atheros ath5k or ath9k, was very reliable in Linux, wonder how it is nowadays.

The only problem with these Intel chips is LAR, and for hostapd it is a big problem...

[arQon]

it drops a lot of packets and connection dies constantly

Jep, that's "every realtek adapter ever" in a nutshell.

I've got an old one (8273) in one of the NUCs, and it was like that for multiple years until someone rewrote the driver. Since then it tends to run at around 70Mb/s, and I've seen it break 100Mb/s (VERY rarely, but it can happen), which is outright amazing for a b/g/n card going to a different floor through multiple walls too. It still drops out: exactly once per host power or suspend cycle, after about 2mins of uptime - but then stays up forever, so I can live with that.

It actually outperformed the Intel adapter in the last laptop I had in here, by about 40%, which was a shock. So AFAICT, the RTL hardware itself is actually fine: it's just that they put absolutely no effort into their Linux drivers, and they also don't provide any documentation for anyone else wanting to improve or replace their code.

Doesn't surprise me at all that the drivers for their newest chips are just as terrible as the ones for their old chips. That's just how this market is, unfortunately. Given the choice, I'll always go with the Intel wifi too, even when it's that much slower, because I at least know it's likely to work on day one and will probably be stable.