10:45 <nomis> robmur01: I right now try to follow the instructions in board/rockchip/evb_rk3328/README to create an uboot-image with miniloader manually. The docs are outdated and there are wrong paths all over the place, but I finally fail in the "rkflashloader" step because that shell script does not have support for rk3328. *sigh*

10:46 <nomis> the git repository for these tools is basically ancient. I suspect that somebody did not push his local repository upstream... :-/

13:02 <nomis> robmur01: updating to u-boot 2004.04 has helped tremendously.

13:05 <robmur01> oh, I'm so accustomed to building master (of just about anything) that I didn't consider that a build system would probably pull down some older stable release :)

13:06 <robmur01> I think a lot of the work on the DDR code is only relatively recent

13:06 <nomis> robmur01: yeah. that happened basically immediately after the 2020.01-Release.

13:07 <nomis> the Kernel starts but then fails, but this is probably an issue of mismatches between the evb-rk3328 Devicetree and my own hardware.

13:08 <robmur01> yay progress! :D

13:09 <nomis> indeed. Quite a bumpy ride.

13:09 <robmur01> any particular errors stand out in the kernel log?

13:10 <nomis> right now it stops while probing stuff: [ 6.773755] phy phy-ff450000.syscon:usb2-phy@100.0: charger = USB_SDP_CHARGER

13:11 <nomis> dunno, possibly a nonexistant 2nd ethernet phy?

13:15 <nomis> ah no, that refers to some USB phy. It feels very timeout'y anyway, so it probably runs into some lockup trying to initialize a device not available on my hardware.

13:16 <nomis> But that feels way more manageable than this TPL/SPL-Stuff...

13:16 <robmur01> regulators and I/O domains are probably the next thing to look at

13:17 <nomis> robmur01: I have some other devicetree lying around, scraped from a different running kernel. First thing is to try that.

13:17 <robmur01> those are relatively straightforward to figure out from the original DTB

13:28 <robmur01> oh, also probably worth checking that the eMMC/SD/SDIO aren't wired up to different interfaces than expected - IIRC some boxes did that

13:32 <nomis> at least in U-Boot I can find the eMMC as well as a SD Card.

13:33 <nomis> so apparently nothing unexpected there.

13:34 <nomis> JPEW: can you give me a pointer on how to incorporate a custom dts-file into the build process?

13:40 <JPEW> Hmm

13:40 <JPEW> There might be a better way, but you could make a patch file that adds the dts and add it in SRC_URI

13:42 <nomis> JPEW: that would be in the linux-yocto recipe then, not in the machine conf, right?

13:45 <JPEW> nomis: correct

14:49 <nomis> hrm. Ok. That devicetree is worse than the other. bah.

14:50 <nomis> (probably because it is for the wrong kernel. ah well. Seems no shortcuts this time...)

15:06 <robmur01> cheeky trick for reverse-engineering TV box DTBs - build rk3328-box.dtb from the appropriate 3.10 or 4.4 branch of the rockchip kernel and dtdiff that against rk-kernel.dtb extracted from the box's firmware image

15:07 <robmur01> then you can see what if anything the vendor actually tweaked, and have readable source to correlate against

15:09 <robmur01> as long as they haven't made significant changes that renumber all the phandles, that usually works pretty well

16:37 <nomis> dtdiff is new to me. sounds good.