14:53 <tdebrouw> narmstrong: I see some CSI clocks for the g12a have been added in mainline / 5205628ab0bfe64952974d476ee001f6c7e0ef7f

14:54 <tdebrouw> I also see that CSI support is WiP on the A311D here: https://www.linux-meson.com/hardware.html

14:55 <tdebrouw> I checked your tree on codelinaro

14:55 <tdebrouw> but I couldn't find any branch related to the CSI inputs

14:55 <tdebrouw> Is someone working on that?

15:00 <tdebrouw> I read the irclogs, in case I'm not online.

15:49 <minute> is meson-g12b stuff gone from current u-boot dts Makefile or am i holding it wrong?

15:51 <minute> ok, need to understand how this works... where are the dts files coming from nowadays? https://github.com/u-boot/u-boot/commit/114df8b5339e503b50b6ac1a61bf1daac57c2e11

16:12 <tdebrouw> minute: https://git.kernel.org/pub/scm/linux/kernel/git/devicetree/devicetree-rebasing.git

16:14 <tdebrouw> CONFIG_ARCH is selected in defconfig

16:14 <tdebrouw> (example: khadas-vim3_defconfig has CONFIG_ARCH_MESON=y)

16:15 <tdebrouw> following this example:

16:15 <tdebrouw> this defconfig also contains: CONFIG_MESON_G12A=y

16:15 <tdebrouw> then in the following file:

16:16 <tdebrouw> arch/arm/mach-meson/Kconfig

16:16 <tdebrouw> you'll find a "config MESON_G12A"

16:17 <tdebrouw> which has a "imply OF_UPSTREAM"

16:18 <tdebrouw> "OF_UPSTREAM" can be found in dts/Kconfig

16:19 <tdebrouw> It was added to solve the device trees getting out of sync from the kernel

16:21 <xdarklight> minute: I was also wondering. tdebrouw gave a more technical explanation, the plain answer is: https://github.com/u-boot/u-boot/tree/master/dts/upstream/src/arm64/amlogic

16:40 <minute> thanks!

16:40 <minute> currently fighting with weird emmc clock issue in uboot on g12b

16:41 <minute> mmc_select_mode_and_width works maybe 50% of times

16:42 <minute> sometimes the freq is changed mysteriously after a line like > trying mode MMC High Speed (52MHz) width 8 (at 52 MHz)

16:42 <minute> for example, after this one comes > selecting mode MMC legacy (freq : 25 MHz)

16:42 <minute> but sometimes it does the right thing > selecting mode MMC High Speed (52MHz) (freq : 52 MHz)

16:58 <narmstrong> tdebrouw: there’s work on upstreaming the arm isp, but not the same version as the one in g12b, afaik nobody is working on adapting it and adding the missing phy and csi path bits. And I never had a csi camera module for a g12b board so I can’t help here

17:08 <minute> ah, __mmc_switch fails sometimes with -5

17:27 <minute> > mmc_send_cmd_retry ret: -5

17:38 <minute> hmm hmm https://patchwork.ozlabs.org/project/uboot/patch/20191126211206.4537-2-linux.amoon@gmail.com/#2313230

17:58 <minute> so sometimes we can't communicate with the emmc for a while

17:58 <minute> getting STATUS_ERR_MASK

17:58 <minute> but when it starts working, it's totally stable

17:58 <minute> so weird

18:32 <minute> as a workaround i have a loop in mmc_init that tries up to three times

19:23 <minute> not reliable

19:23 <minute> how is it possible that the rom can load uboot reliably from emmc every time but then uboot fails to talk to the emmc sometimes

19:28 <vagrantc> u-boot might try to use a faster but less stable mode?

19:29 <minute> vagrantc: maybe, but in the failure case, none of the modes it tries work

19:29 <minute> but if i wait a while after a failure and enter "mmc dev 1" again, it works

19:29 <vagrantc> sometimes things are on their own schedules ...

19:30 <minute> in the failure mode, even this fails: >

19:30 <minute> mmc trying mode MMC legacy width 1 (at 25 MHz)

19:31 <minute> i have a loop that does 10x retries with 1sec wait. this can fail, i.e. it exhausts the 10 tries. but when i wait a bit and enter "mmc dev 1" it most probably will work

19:31 <minute> i don't get it... i thought ok, maybe cache issue, but cache flush doesn't help

19:31 <minute> also, the emmc _always_ works in linux

19:35 <vagrantc> more complex or well tested initialization code in linux, probably

19:39 <minute> hm, uboot does not set CLK_RX_PHASE, only CLK_TX_PHASE

19:41 <minute> (doesn't make a difference, it would add 0)

20:08 <f_> I guess the ROM does only very minimal init - enough to load 48K BL2 and the rest

20:15 <minute> so none of you have emmc issues in uboot on g12b?

20:22 <f_> I did not experience eMMC issues lately, no.

20:23 <f_> But I'm on an old U-Boot.

20:23 <f_> dirty 2024.01

20:24 <minute> ok? aml_send_calibration_blocksaml_send_calibration_blocks

20:24 <minute> sorry, copypaste issue

20:24 <minute> i meant: https://github.com/Dangku/amlogic-u-boot/blob/g12b-v2015.01-m2s-cm4/drivers/mmc/aml_sd_emmc.c#L1085

20:33 <minute> interestingly, linux drives the card at hs200, but it doesn't look like meson_gx_mmc in uboot can do that?!