05:36 <epoll> okay i've tried a few combinations

05:37 <epoll> u-boot 2022.10 with bl31 v1.32 hangs at the spl after checking the atf hashes

05:37 <epoll> u-boot 2022.10 with bl31 v1.28 gets to the kernel but the kernel hangs on "Starting kernel ..."

05:38 <epoll> u-boot 2023.01-rc4 with bl31 v1.28 hangs after "BL31: Initializing BL32" so it doesn't get past the spl

05:41 <epoll> and u-boot 2023.01-c4 with bl31 v1.32 hangs at atf invocation at the end of the spl (so it technically gets farther than 2022.10 with v1.32)

05:41 <epoll> i've tried building atf with this patch too https://review.trustedfirmware.org/c/TF-A/trusted-firmware-a/+/16952

05:41 <epoll> it gets to the kernel but it won't start (same symptoms as u-boot 2022.10 with v1.28)

05:54 <epoll> i've also tried a more strict reordering of the atf entries in u-boot.its (by copying it from the bsp) and the symptoms for 2023.01-rc4 swap (between the bl31 version), while with 2022.10 both bl31 versions hang at starting the kernel

05:55 <epoll> what combination of 1) bl31 and 2) u-boot were people able to get this working on? are there any pointers to trees?

05:55 <epoll> does the kernel version matter too?

08:50 <epoll> fwiw the bsp u-boot can't boot the kernel either with the atf from that gerrit

08:51 <epoll> although it does get past the "Starting kernel ..." line

08:52 <epoll> it just hangs later in the boot

08:54 <epoll> oh? i got a dump

08:54 <epoll> https://termbin.com/3fd3

08:54 <pinchartl> seems expected

08:56 <pinchartl> what puzzles me is why you can't get mainline u-boot + binary BL31 to work

08:56 <pinchartl> as multiple people here reported success

08:58 <pinchartl> can anyone share a working git branch for u-boot ? :-)

08:58 <epoll> yeah that's what im puzzled by too

08:58 <pinchartl> (and maybe any build instructions that would be specific to this)

09:53 <naoki> on rock 3a, u-boot 2023.01 with bl31 v1.36 hangs after "Trying to boot from MMC1"

09:54 <naoki> u-boot 2023.01 with bl31 v1.28 with 6.1.6 kernel hangs after "[ 3.349304] hub 8-0:1.0: 1 port detected"

09:54 <naoki> hmmm.....

09:55 <naoki> radxa cm3 sodimm & e25 are working with u-boot 2023.01, bl31 v1.36, kernel 6.1.6

09:56 <epoll> for 2023.01-rc4 u-boot im on this branch https://git.uk.ideasonboard.com/epaul/u-boot/src/branch/rk3568

11:36 <CounterPillow> pinchartl: I'm currently using https://github.com/CounterPillow/u-boot-quartz64 (based on 2022.04) with some old bl31

11:36 <CounterPillow> this bl31 and ram init in fact: https://github.com/Plebian-Linux/quartz64-images/blob/main/.github/workflows/build-image.yml#L52-L53

11:37 <CounterPillow> I don't think mainline proper works for RK3568 at the moment, macromorgan was working on that I believe

11:47 <pinchartl> CounterPillow: thank you

11:47 <pinchartl> epoll: ^^

11:47 <pinchartl> can you test that ?

11:48 <pinchartl> does anyone have gmac1 ethernet working in mainline (or mainline-ish, with extra patches) u-boot ? the compatible string used in rk356x.dtsi is not listed in the driver, I assume it requires custom ops/data

11:51 <CounterPillow> I know mine does not support it, unfortunately

13:42 <naoki> ah, kernel 6.1.6 is running... just uSD card is not detected...

13:46 <naoki> https://github.com/RadxaNaoki/u-boot/commits/rock-3

13:46 <naoki> mainline linux can be booted (when uSD (rootfs) is detected properly...)

13:49 <pinchartl> naoki: thank you!

13:50 <pinchartl> which version of BL31 do you use with that ?

13:52 <naoki> https://gitlab.com/rk3588_linux/rk/rkbin/-/blob/master/bin/rk35/rk3568_bl31_v1.36.elf

13:52 <pinchartl> thanks

13:52 <pinchartl> naoki: is ethernet working in u-boot in that branch ?

13:52 <naoki> well

13:53 <CounterPillow> naoki: try booting it a few times, I have an issue with my mainline-ish u-boot where the SD card isn't detected 1 out of every 4 boots

13:53 <naoki> I commented out most part of dts... only uart2 and mmc are enabled

13:54 <naoki> (let's try to use full dts...)

13:55 <pinchartl> naoki: I'm looking at the device tree sources in your branch, and gmac1 has compatible = "rockchip,rk3568-gmac", "snps,dwmac-4.20a";

13:56 <pinchartl> drivers/net/dwc_eth_qos.c has 3 matches entries

13:56 <pinchartl> "nvidia,tegra186-eqos", "st,stm32mp1-dwmac" and "nxp,imx8mp-dwmac-eqos"

13:56 <pinchartl> so I assume it won't work out-of-the-box

14:16 <naoki> grep compatible drivers/net/gmac_rockchip.c

14:16 <naoki> hmm

14:30 <pinchartl> oh, I hadn't noticed that one

14:33 <pinchartl> in the mainline kernel the compatible string matches with ethernet/stmicro/stmmac/dwmac-rk.c

14:33 <pinchartl> selected by CONFIG_DWMAC_ROCKCHIP

14:34 <pinchartl> which is a sub-option of CONFIG_STMMAC_ETH

14:34 <pinchartl> (if any)

14:43 <pinchartl> it looks like gmac_rockchip.c is for older SoCs, but I'm not entirely sure

14:47 <Tenkawa> pinchartl: what are you trying to work on? (I missed the beginning part)

14:54 <pinchartl> rock 3a, based on a rk3568

14:55 <pinchartl> I'd like to get mainline u-boot working on that, with ethernet support

14:55 <pinchartl> (and ideally mainline tf-a)

14:55 <pinchartl> it's epoll doing most of the work, I'm just supporting him

14:55 <Tenkawa> Let me check our build and see which u-boot we are using on our 3568

14:56 <Tenkawa> We build for the Odroid-M1 which is a 3568 so I'm going to see which u-boot we are using for it

14:59 <pinchartl> thanks

15:02 <Tenkawa> looks like we are still using a vendor u-boot still

15:03 <Tenkawa> We tried mainline but it was very flaky

15:03 <Tenkawa> The PWM, sensors, and something else (can't remember) was not reliable

15:05 <pinchartl> we tried the vendor u-boot too (from radxa) but it has no ethernet support

15:05 <Tenkawa> I have some RK3588's that are still waiting on a useable mainline update

15:05 <Tenkawa> Interesting..

15:05 <pinchartl> (well, it has ethernet, but no support for the ethernet controller on that SoC)

15:06 <Tenkawa> Nod

19:20 <macromorgan> sorry, was busy so I missed the previous conversation. Here's what I have for a branch that boots consistently for me on a 3566 device (tested on 4 distinct Anbernic based devices): https://github.com/macromorgan/u-boot-rk3566/tree/anbernic-v2022.10

19:21 <macromorgan> You either need to use the BL31 version 1.28 or prior, OR re-order the A-TF modules and rebuild your u-boot.itb if you use a BL31 greater than 1.28.

19:22 <macromorgan> I was slowly in the process of getting these patches mainlined, I started with the first 5 or so but still have more to go.

19:23 <macromorgan> The things that helped me with SD card detection are 1) setting the drive strength like so: https://github.com/u-boot/u-boot/commit/0f68095f176d4c5fcee4882a3960f87e07dc86ae and 2) dropping fifo mode like so: https://github.com/u-boot/u-boot/commit/918ffcf988fc89abb5fb3d74b6ffa9b8c0d692f8

19:24 <macromorgan> note that I have been using that branch for roughly 3 months now and in that time have had 0 issues detecting an SD card at all whether it be via cold boot or warm boot in U-Boot

19:26 <macromorgan> OHH, also make sure you set your clock speed for the MMC card... look at what I have for the "assigned-clocks" and "assigned-clock-rates". If the rates aren't set detection is intermittant too.

19:26 <macromorgan> so in summary for SD detection 1) set your drive strength (might be optional), 2) drop the fifo mode, and 3) set your clocks

19:27 <macromorgan> to be clear, the "assigned-clocks" and "assigned-clock-rates" I have set on the &cru node

19:29 <macromorgan> but yes... mainline U-Boot and mainline Linux work perfect for me now on the Anbernic RG353 devices, and all that's missing on the RG503 is the panel (and all of them are still missing wifi and bluetooth and suspend, but I'm working on that). U-Boot itself for me works perfectly fine.