09:43 <narmstrong_> warpme_: make sure you have amlogic,dram-access-quirk in the sdio node

09:43 <narmstrong_> warpme_: what base dtsi are you using ?

09:50 <warpme_> narmstrong_: i added dram-access-quirk. no better :-(. Here is DT i'm using: https://pastebin.com/yHAb6gn3

09:51 <warpme_> narmstrong_: btw: i received link to android kernel radxa uses for radza-zero. comparing with mainline linux dt for other boards with ap6256 (i.e. rockpi-4b) i see difference in clocking: mainline kernel: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/arch/arm64/boot/dts/rockchip/rk3399-rock-pi-4.dtsi?h=v5.13.5#n32 radxa android kernel:

09:51 <warpme_> https://gitlab.com/amlogic-android/p-amlogic/kernel/common/-/blob/b9254b71f3c32b22122243758854d2722c52b543/arch/arm/boot/dts/amlogic/g12a_s905y2_u221.dts#L202

09:54 <narmstrong_> the pwm_channel2_conf is weird

09:57 <warpme_> narmstrong_: just fyi (if you not get sch for radxa: https://dl.radxa.com/zero/docs/hw/ )

09:57 <narmstrong_> ok yeah they use 2 pwms to get a more precise pwm output

10:00 <warpme_> narmstrong_: clocking entries were suggested by radxa team (i used radxa work on initial armbian boot)....

10:00 <narmstrong_> I see nothing special on clocking

10:08 <warpme_> narmstrong_: in android dt is see pwr_reg_on approach while in mainlinux we have sdio pwrseq approach. Isn't difference that: req_on is 1 binary change (off->on) - like power supply turning on. pwrseq is however short activation of reset line. shouldn't we use reg_on (like in android)? At end there should be no difference but.... (btw: i already tested with change reset-gpios from GPIO_ACTIVE_LOW to GPIO_ACTIVE_HIGH:

10:08 <warpme_> with high driver not seeing chip at all)

10:19 <warpme_> narmstrong_: btw: i REALLY surprised how well this board works. h264 decoder works well (it fails for me only on 2160p h264; this is contrary to sm1 which fails an many samples). hevc however works partially (some samples makes it hang). If hevc will be fixed - i would say g12 is useable soc for linux media player...

11:49 <narmstrong_> warpme_: the pwrseq alway worked up to now, but the errors you encounter really reminds me what amlogic,dram-access-quirk solved

11:52 <narmstrong_> maybe you can try to revert https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/drivers/mmc/host/meson-gx-mmc.c?h=v5.13.4&id=103a5348c22c3fca8b96c735a9e353b8a0801842

12:13 <warpme_> narmstrong_: with above commit reverted:

12:13 <warpme_> https://www.irccloud.com/pastebin/ia60xKo3/

12:14 <warpme_> :-)

12:22 <narmstrong_> :-/

21:03 <Ballerburg9005> 4<<:q