09:41 <hexdump0815> do i understand it right that a h6 u-boot does not strictly require crust as long as i do not need crust related features like suspend/resume?

09:42 <hexdump0815> is there a prebuilt crust available somewhere for the h6?

09:43 <gamiee> hexdump0815: yes, crust is optional

10:31 <hexdump0815> jernej: hdmi is working now after making sure CMA and DMA_CMA are enabled ... now i'm struggling with eth - dmesg ist here: https://pastebin.com/raw/LZKHDB9k

10:32 <hexdump0815> i have built a fresh u-boot with fresh atf as well to be on the safe side

10:32 <hexdump0815> if i cold boot i get the dmesg from the pastebin and eth is not working

10:32 <gamiee> hexdump0815: I'm not pro, but does your DTS have everything correctly set for EMAC/GMAC controller?

10:33 <hexdump0815> if i reboot from 5.10 (where eth is working fine) into 5.15 then eth seems to work as well for some strange reason

10:33 <hexdump0815> gamiee: its set in a way in which it worked perfectly fine with 5.10

10:33 <jernej> what kind of ethernet is on eachlink mini? gigabit or fast ethernet?

10:33 <gamiee> ahh, then DTS is fine I guess

10:34 <hexdump0815> jernej: the box is a qplus one which requires your ac200 patches for eth

10:34 <hexdump0815> so its the internal (i think) phy

10:35 <jernej> then you have to make sure everything probes in right order

10:35 <jernej> I suggest you mark all ethernet related drivers as built in and only dwmac sun8i as module

10:35 <jernej> that should do it, but no promises

10:36 <hexdump0815> jernej: thanks a lot i'll give it a try

10:36 <jernej> there is nothing broken in 5.15, I'm using it

10:37 <jernej> hexdump0815: do you still need that hack for hdmi on eachlink?

10:38 <hexdump0815> i can re-test it during the next days when i got 5.15 working well on h6 again :)

10:40 <hexdump0815> jernej: btw. do you have any plans to upstream the ac200 support? iirc it was not easy to get it into mainline ready shape - right?

10:41 <jernej> try with this HDMI patch: https://github.com/LibreELEC/LibreELEC.tv/blob/master/projects/Allwinner/patches/linux/0065-wip-fix-H6-4k-60.patch

10:42 <jernej> it might solve more issues than only 4k@60

10:42 <jernej> not yet, first someone must figure out proper DT nodes relations

10:42 <jernej> currently it's impossible to satisfy them

10:42 <jernej> so some changes are needed, just I don't know which

10:43 <jernej> I got some response, but it was all code related, not DT

11:17 <hexdump0815> jernej: apritzel: just an observation - somehow the just built u-boot v2021.10 results in a less stable systems ten my old v2020.07 with the same v5.10 kernel

11:18 <hexdump0815> with the new u-boot it just hangs from time to time - looks a bit like memory timing issues ... i already lowered CONFIG_DRAM_CLK=840 to 660, but no change

11:21 <hexdump0815> i'll investigate further - maybe i can find something more concrete - one thing might be that i'm running the h6 at 2ghz which works absolutely stable with the old u-boot but maybe the timinig with the new one is slightly different so that its not stable anymore

11:21 <jernej> 2 ghz is overclocking anyway

11:21 <jernej> and best bet would be to bisect

11:22 <jernej> latest U-Boot works for me just fine on defaults

11:36 <hexdump0815> once u-boot even got the memory size wrong (4g in spl and 3g in u-boot vs 2/2 in reality), but after reboot it was fine and did never happen again

11:52 <jernej> wait, is this the board with most persisting DRAM size issue?

12:03 <hexdump0815> no - that was the eachlink - the qplus was ok so far regarding the dram detection

12:14 <hexdump0815> jernej: setting stmmac to =y did not help with the eth - now there is only left mdio-mux as module on which dwmac sun8i depends, but this is forced to =m by dwmac sun8i being =m

13:53 <jernej> idk, check LE kernel config for ideas

17:43 <hexdump0815> jernej: it is definitely not a probe order problem - i just moved dwmac_sun8i aside, rebuilt the initramfs and rebooted (so without it) and modprobed it when the system was up and i'm getting the same error - the trace output from the time stamp of probing is https://pastebin.com/raw/XAvsbQqb

17:44 <hexdump0815> to me it looks like "dwmac-sun8i 5020000.ethernet: EMAC reset timeout" results in the driver being removed and this results in the trace

17:52 <jernej> so then you have some other issue

17:52 <jernej> and I have no idea what

18:18 <hexdump0815> jernej: i checked the latest (v2021.10) u-boot+atf on the eachlink box now - it still detects the memory wrong as 4gb (spl) and 3gb (u-boot) instead of the real 2gb and as a result linux will crash later if not forced to 2gb

18:19 <hexdump0815> hdmi also still does not work with v5.15 out pf the box if not forced on - will test you hdmi patch on it next

18:20 <hexdump0815> the funny thing is: eth on this one works fine using the same sd card - so looks like the problem i see with the eth probing on the qplus box seems to be hardware related

18:21 <hexdump0815> unexpected is that eth works absolutely stable with 5.10 on this box - my 5.10 kernel was based on the megous tree, so maybe there are some fixes in which prevent the trouble

18:22 <hexdump0815> on the other side on the 5.15 megous tree i got the exact same problems with eth

18:27 <jernej> as I said, I don't have ethernet issues on TX6 on 5.15 with LE patches

18:27 <jernej> maybe you can test that kernel config + patches

18:58 <hexdump0815> jernej: i just tested your hdmi patch from above on the eachlink - still no hdmi out if not forced ... i guess this hardware is simply broken

22:35 <hexdump0815> jernej: found it - i had CONFIG_PHYLIB=m which pushed CONFIG_AC200_PHY to m on olddefconfig - with this fixed eth works now fine on this box too on v5.15 :)