11:39 <chewitt> I'm looking around to see what patches people are using at the moment

11:39 <chewitt> this is a dumb patch: https://github.com/armbian/build/blob/main/patch/kernel/archive/meson64-6.14/board-bananapi-cm4-cm4io.patch#L109

11:39 <chewitt> the /delete-property/ amlogic,dram-access-quirk; will indeed speed up wifi performance

11:40 <chewitt> but at the expense of having sdio issues due to silicon bugs in the mmc controller

11:40 <chewitt> which is, err, why the quirk was added in the first place

11:43 <chewitt> also, not sure who "Signed-off-by: Patrick Yavitz <pyavitz@xxxxx.com>" is

11:43 <chewitt> but

11:43 <chewitt> https://github.com/armbian/build/blob/main/patch/kernel/archive/meson64-6.14/board-bananapim5-002-add-wifi-bt-support.patch

11:44 <chewitt> here's the original patch with my attribution:

11:44 <chewitt> https://github.com/chewitt/linux/commit/43a534f7b59e2364221f972df637f5cc8e8ab153

11:44 <chewitt> no issues at all to copy stuff from my trees .. it's the reason things are there

11:44 <chewitt> but it would be appreciated if attribution was retained

11:47 <chewitt> I believe this is no longer required https://github.com/armbian/build/blob/main/patch/kernel/archive/meson64-6.14/board-odroidc2-usb-hub-disable-autosuspend-for-Genesys-Logic-.patch

11:51 <chewitt> also no longer required https://github.com/armbian/build/blob/main/patch/kernel/archive/meson64-6.14/general-memory-marked-nomap.patch

12:16 <c0rnelius> chewitt: The original PR https://github.com/armbian/build/pull/7193 I have since seen issue with that bit of the patch concerning the SDIO using one of my modules with the BPI baseboard but not when using the Waveshare.

12:17 <c0rnelius> I have personally stopped using the patch and left it in place on the Waveshare, but since no one has complained I left it in place in Armbian.

12:19 <c0rnelius> Are you seeing SDIO errors when making that adjustment?

14:18 <chewitt> I don't see errors because I don't have that patch

14:19 <chewitt> the vendor codebase has a rather invasive/egregious hack to work around the silicon bug (all boards from gxl to sm1)

14:19 <chewitt> the dram quirk in mainline kernels is a compromise between stability and speed

14:20 <chewitt> the hack that the vendor kernel uses would never be accepted upstream

14:25 <chewitt> see this https://patchwork.kernel.org/project/linux-amlogic/cover/20190527124307.32075-1-narmstrong@baylibre.com/

14:25 <chewitt> without the quirk the sdio has stability issues

14:27 <chewitt> ahh, I'm confusing things - there's an alignment issue that sometimes triggers on gxl/gxm, then the mmc issue on g12a/b and sm1 boards

14:28 <chewitt> the driver handles them both .. unless you patch out the quick :(

15:03 <c0rnelius> If you read the forum discussion part of the PR, you can see I'm skeptical. I saw instability, but only in some cases. At the time and with two people testing, I thought it was worth a try stepping outside of the box here. This is why the patch was created.

15:06 <c0rnelius> I'm sorry you find it `dumb`.

15:10 <c0rnelius> Your name sake has also been corrected on the M5 patch.