[PATCH 1/3] sunxi: board: simplify early PMIC setup conditions

[Leon Anavi]

Hi Andre,

On 14.12.24 г. 4:19 ч., Andre Przywara wrote:
> On Thu, 12 Dec 2024 11:19:06 +0200
> Leon Anavi<leon.anavi at konsulko.com> wrote:
>
> Hi Leon,
>
>> On 11.12.24 г. 23:53 ч., Andre Przywara wrote:
>>> On Mon, 9 Dec 2024 23:08:19 +0200
>>> Leon Anavi<leon.anavi at konsulko.com> wrote:
>>>
>>> Hi Leon,
>>>
>>> thanks for the report!
>>>   
>>>> Commit ffb0294 from 12 November 2023 that simplifies early PMIC setup
>>>> conditions causes issues on Cubieboard 4 and Merrii A80 Optimus with
>>>> Allwinner A80 SoC (sun9i). The commit was introduced with U-Boot 2024.01
>>>> (rc3) and remains as of today. Because of it both of these boards hang at:
>>>>
>>>> Starting kernel ...
>>> That's odd, how do you boot the kernel, exactly?
>>> I just tried mainline U-Boot (via FEL), with:
>>> => setenv bootargs "console=ttyS0,115200n8 earlycon"
>>> => bootz $kernel_addr_r $ramdisk_addr_r:300000 $fdtcontroladdr
>>>
>>> and it booted fine to the prompt, on a Cubieboard 4 (CC-A80 v1.2).
>>> Kernel was some 6.11-rc6 I just had lying around.
>>>
>>> I also compared the code before and after that patch, the only
>>> difference is the order at which DCDC5 gets programmed: before it's
>>> after DCDC4, with the patch it's right after DCDC1.
>>> The rest looked the same.
>>> Booting ffb0294~1 and ffb0294~0 also worked for me, without issues.
>>> So can you please describe how you test that, exactly?
>> I built core-image-base using the Yocto Project and OpenEmbedded as well Starting kernel ...
>> as the meta-sunxi BSP layer. The U-Boot version depends on the Yocto
>> release. For Scarthgap it is with U-Boot 2024.01 and for the ongoing
>> development in the master branches the U-Boot version is 2024.10. The
>> Linux kernel version is 6.6.28. I experienced the same issue with both
>> U-Boot versions on Merrii A80 Optimus and Lazar (his email is CC)
>> confirmed the same results on Cubieboard 4. The boot sequence for the
>> boards in meta-sunxi is through boot.scr generated from the following
>> boot.cmd:
>> https://github.com/linux-sunxi/meta-sunxi/blob/master/recipes-bsp/u-boot/files/arm/boot.cmd
>>
>> In a nutshell, the issue can be reproduced with the following scenario
>> in U-Boot:
>>
>> setenv bootargs console=ttyS0,115200 console=tty1 root=/dev/mmcblk0p2
>> rootwait panic=10
> So I tried with an SD card now, and I think I see what you mean now.
> I was a bit confused first why it would hang for you after "Starting
> kernel ...", but I guess it actually doesn't: can you add "earlycon" to your
> command line and confirm that the kernel boots, but hangs while waiting
> for the secondary cores?
> I am still scratching my head how this commit would affect this
> behaviour, but this change here seems to fix it for me:
>
> diff --git a/board/sunxi/board.c b/board/sunxi/board.c
> index 29a329f41a2..d7c46eec615 100644
> --- a/board/sunxi/board.c
> +++ b/board/sunxi/board.c
> @@ -582,7 +582,6 @@ void sunxi_board_init(void)
>   
>   #ifdef CONFIG_AXP_DCDC1_VOLT
>   	power_failed |= axp_set_dcdc1(CONFIG_AXP_DCDC1_VOLT);
> -	power_failed |= axp_set_dcdc5(CONFIG_AXP_DCDC5_VOLT);
>   #endif
>   #ifdef CONFIG_AXP_DCDC2_VOLT
>   	power_failed |= axp_set_dcdc2(CONFIG_AXP_DCDC2_VOLT);
> @@ -591,6 +590,9 @@ void sunxi_board_init(void)
>   #ifdef CONFIG_AXP_DCDC4_VOLT
>   	power_failed |= axp_set_dcdc4(CONFIG_AXP_DCDC4_VOLT);
>   #endif
> +#ifdef CONFIG_AXP_DCDC5_VOLT
> +	power_failed |= axp_set_dcdc5(CONFIG_AXP_DCDC5_VOLT);
> +#endif
>   
>   #ifdef CONFIG_AXP_ALDO1_VOLT
>   	power_failed |= axp_set_aldo1(CONFIG_AXP_ALDO1_VOLT);
>
> Can you confirm this?


Yes, thank you. Excellent work! I confirm this solution works. I have 
just tested it on Merrii A80 Optimus board. I applied your patch on top 
of U-Boot 2024.10.

> This fix looks easily mainline-able, as it just restores the old
> initialisation order, but I still would like to know why the other
> order hangs. Will try to do some experiments.

OK. In the mean time I will contribute your patch to meta-sunxi BSP 
layer so that it can be used for the existing U-Boot versions in images 
for sun9i machines built using the Yocto project and OpenEmbedded. This 
is obviously a far better solution than reverting commit ffb0294.
Thanks,
Leon

>
> Cheers,
> Andre
>
>
>> load mmc 0:1 ${fdt_addr_r} ${fdtfile}
>> load mmc 0:1 ${kernel_addr_r} zImage
>> bootz ${kernel_addr_r} - ${fdt_addr_r}
>>
>> Merrii A80 Optimus board boots fine if I revert commit ffb0294.
>> However, if commit ffb0294 is present in U-Boot the board hangs. Here
>> is the log: U-Boot 2024.10-dirty (Oct 07 2024 - 14:54:35 +0000)
>> Allwinner Technology CPU: Allwinner A80 (SUN9I) Model: Merrii A80
>> Optimus Board DRAM: 2 GiB Core: 62 devices, 18 uclasses, devicetree:
>> separate WDT: Not starting watchdog at 6000ca0 WDT: Not starting
>> watchdog at 8001000 MMC: sunxi_set_reset: (RST#0) unhandled
>> sunxi_set_reset: (RST#2) unhandled sunxi_set_reset: (RST#1) unhandled
>> mmc at 1c0f000: 0, mmc at 1c10000: 2, mmc at 1c11000: 1 Loading Environment
>> from FAT... Unable to read "uboot.env" frommmc0:1... In:
>> serial at 7000000 Out: serial at 7000000 Err: serial at 7000000 Net: No
>> ethernet found. Hit any key to stop autoboot: 0 => setenv bootargs
>> console=ttyS0,115200 console=tty1 root=/dev/mmcblk0p2 rootwait
>> panic=10 => load mmc 0:1 ${fdt_addr_r} ${fdtfile} 24443 bytes read in
>> 3 ms (7.8 MiB/s) => load mmc 0:1 ${kernel_addr_r} zImage 5397992
>> bytes read in 236 ms (21.8 MiB/s) => bootz ${kernel_addr_r} -
>> ${fdt_addr_r} Kernel image @ 0x22000000 [ 0x000000 - 0x525de8 ] ##
>> Flattened Device Tree blob at 23000000 Booting using the fdt blob at
>> 0x23000000 Working FDT set to 23000000 Loading Device Tree to
>> 29ff7000, end 29ffff7a ... OK Working FDT set to 29ff7000 Starting
>> kernel ... Please note that U-Boot is marked with suffix "-dirty"
>> because of the patches applied by u-boot_%.bbappend from Yocto/OE
>> layer meta-sunxi but these patches are for other boards and don't
>> touch the file board/sunxi/board.c. Am I missing something as a
>> configuration that can make the board boot the same way without
>> hanging when commit ffb0294 is present?Best regards, Leon
>>
>>> Please also note we fixed d75fa8c80dcfa in U-Boot (DCDC4/5 typo),
>>> and dd36ad71ad6 in the kernel (DCDC5 constraints in the DT).
>>>
>>> Cheers,
>>> Andre
>>>   
>>>> Older U-Boot versions without this commit work fine. As a temporary
>>>> solution I reverted commit ffb0294 and this way the boards boot
>>>> successfully. I tested this work around on Merrii A80 Optimus with
>>>> several U-Boot versions, including with U-Boot 2024.10.
>>>   
>>>> Lazar, a friend who owns Cubieboard 4, also tested and confirmed
>>>> his board boots with U-Boot 2024.10 if this commit has been
>>>> reverted.
>>>>
>>>> How to fix this? Is there a known configuration that can be added
>>>> to Merrii_A80_Optimus_defconfig and Cubieboard4_defconfig to avoid
>>>> hanging with the existing source code from commit ffb0294 ?
>>>>
>>>> Best regards,
>>>> Leon