Use mainline rtl8xxxu driver

CGarces · December 10, 2017 at 7:12 PM

Currently we are using several out of tree kernel drivers for realteck wifi devices.

That cause additional work to keep the diver updated due kernel changes (usually done by milhouse ).

For example we are using rtl8192cu driver from this repository.

The author claims on the readme.md that the mainline rtl8xxxu driver is mature enough and the out of tree kernel is not needed anymore.

Maybe we should remove the RTL drivers (all or just some of them) from generic and RPI images, and keep the out of tree drivers frozen, just for devices that not use mainline kernel.

**milhouse** · December 10, 2017 at 7:28 PM

Yes I think that's likely. We've looked at switching to iwd and this may make it easier to switch to the in-kernel rtl8xxx driver.

The changes required in 4.15 are making it more likely (IMHO) that we will drop the out-of-tree WiFi drivers sooner than later, certainly for RPi/RPi2/Generic or any other project that builds with a mainline kernel. I don't have the time to spend on these drivers trying to fix them (particularly as I'd only be making "blind" changes) so if they're not fixed upstream by the community I guess they'll remain broken (I saw you had committed some changes, thanks for that).

Making the switch to the in-kernel driver and testing/getting feedback may be the only way we can determine whether the mainline driver is "good enough" to make the move permanently. Perhaps it's something we can trial with 4.15-rc considering all the RTL* drivers are broken (not entirely sure how to make the switch, haven't looked).

**lrusak** · December 10, 2017 at 11:27 PM Official Post

I'd rather wait till we move to iwd before going to the in kernel drivers.

**milhouse** · December 11, 2017 at 12:13 AM

Unless the situation improves with the out-of-tree WiFi drivers and 4.15 we may not have a choice!

The other option is to continue with 4.14.y in LE9 when we had planned on using 4.15.y because the WiFi drivers are still broken, but hopefully they'll be fixed eventually (still early days for 4.15-rc).

What's the state of iwd and could we use it in LE9 with in-kernel drivers? Or do you see it as an LE10 feature (possibly due to iwd immaturity?)

**chewitt** · December 11, 2017 at 8:02 AM Official Post

The hope was to have everything on a mainline-ish kernel before switching, but that isn't going to happen for AML devices for a while (not on the 9.0 schedule) so revised intent was to leave iwd until a mid-year 9.2 or similar release, or 10.0 if testing shows drivers need more baking time.

**lrusak** · December 11, 2017 at 8:27 AM Official Post

Quote from milhouse

Unless the situation improves with the out-of-tree WiFi drivers and 4.15 we may not have a choice!

The other option is to continue with 4.14.y in LE9 when we had planned on using 4.15.y because the WiFi drivers are still broken, but hopefully they'll be fixed eventually (still early days for 4.15-rc).

What's the state of iwd and could we use it in LE9 with in-kernel drivers? Or do you see it as an LE10 feature (possibly due to iwd immaturity?)

Quote from chewitt

The hope was to have everything on a mainline-ish kernel before switching, but that isn't going to happen for AML devices for a while (not on the 9.0 schedule) so revised intent was to leave iwd until a mid-year 9.2 or similar release, or 10.0 if testing shows drivers need more baking time.

I'd rather move to iwd right after we branch LE9.0 to get testing happening as it will most likely introduce some breakage and device incompatibilities.