Libre Computer Tritium H5 does not boot from eMMC

[RussianNeuroMancer]

Hello!

 

I have three Libre Computer Tritium H5 boards and three eMMC modules on hands. All three eMMC modules is flashed with exactly same Armbian image and exactly same uboot. I also use exactly same PSU and cable for all three boards (however I reproduced this issue with four PSU and three microUSB cables, no difference in behavior was found).

 

1. All three boards start just fine from microSD (so it's not exactly "Your board doesn't work?" case).
2. One board never boot from eMMC on first power up, but almost always boot on second power up. That reproducible with all three eMMC modules.
3. Two other boards just never boot from any eMMC module.

 

When board can not boot from eMMC it just stuck there:

 

U-Boot SPL 2021.04-armbian (May 06 2021 - 19:52:02 +0000)
DRAM: 2048 MiB
Trying to boot from MMC2

U-Boot SPL 2021.04-armbian (May 06 2021 - 19:52:02 +0000)
DRAM: 2048 MiB
Trying to boot from MMC2

U-Boot SPL 2021.04-armbian (May 06 2021 - 19:52:02 +0000)
DRAM: 2048 MiB
Trying to boot from MMC2

U-Boot SPL 2021.04-armbian (May 06 2021 - 19:52:02 +0000)
DRAM: 2048 MiB
Trying to boot from MMC2

U-Boot SPL 2021.04-armbian (May 06 2021 - 19:52:02 +0000)
DRAM: 2048 MiB
Trying to boot from MMC2

 

That five power up attempt as you can see.

 

I tried to rollback uboot to 2020.04, but it behave in exactly same way.`

All three boards was ordered at the same time, in one package.

 

Is anyone can reproduce this?

Is there anything else to try, or I should just give up and stick with microSD for now? (Although I probably could left uboot and /boot on microSD and then put rootfs on eMMC, but that another story.)

[lanefu]

is a emmc v4 module?   Allwinner H3 and h5 doesn't' support v5

[RussianNeuroMancer]

3 minutes ago, lanefu said:

is a emmc v4 module?   Allwinner H3 and h5 doesn't' support v5

Yes.

 

My idea with uboot and /boot on microSD, and rootfs on eMMC works out on all three boards, by the way (which probably means that there is no hardware issues regarding eMMC).

[lanefu]

yeah honestly I doubt any of us have tried emmc on the H5....  

[jernej]

I have troubles booting Tritium H5 myself, but from SD card. I just power cycle it until it works. Sometimes it seems that it helps if HDMI is unplugged, but that might be just red herring. However, once it's booted, all reboots, suspend/resume cycles via Crust work. I have no idea why it's behaving that way.

[martinayotte]

2 hours ago, lanefu said:

yeah honestly I doubt any of us have tried emmc on the H5...

Personnally, I'm running on eMMC with other boards such OPi-ZeroPlus2-H5 since years ...

[lanefu]

56 minutes ago, martinayotte said:

Personnally, I'm running on eMMC with other boards such OPi-ZeroPlus2-H5 since years ...

 

ah yeah.. true.... I meant the pluggable emmc modules on the Tritium-Hx boards

[RussianNeuroMancer]

5 hours ago, jernej said:

Sometimes it seems that it helps if HDMI is unplugged

 

I had this idea but unplugging everything besides power doesn't help a bit.

[Da Xue]

@RussianNeuroMancer I didn't see this thread or I would have checked it out. This has to do with the delay in mmc initialization. Sometimes it's too fast and something hasn't stabilized. One method to delay it is to enable SPL debug. I have SD cards that just refuse to work but work flawlessly when I turn on debug. There's a delay somewhere that's missing in the u-boot code.

 

In sunxi_mmc.c sunxi_mmc_core_init, you need to raise the udelay to 5ms or 10ms. I didn't do extensive testing but it fixes nearly all mmc problems.

Edited July 20, 2022 by Da Xue

[RussianNeuroMancer]

4 часа назад, Da Xue сказал:

@RussianNeuroMancer I didn't see this thread or I would have checked it out. This has to do with the delay in mmc initialization. Sometimes it's too fast and something hasn't stabilized. One method to delay it is to enable SPL debug. I have SD cards that just refuse to work but work flawlessly when I turn on debug. There's a delay somewhere that's missing in the u-boot code.

 

In sunxi_mmc.c sunxi_mmc_core_init, you need to raise the udelay to 5ms or 10ms. I didn't do extensive testing but it fixes nearly all mmc problems.

Hi! Thanks for advice, but unfortunately I don't have any Tritium H5 left on hands anymore, so I can't test this. If it works for you - maybe it makes sense to submit this change to upstream or at least ask for tests on mail-list?

[Da Xue]

I sent a patch upstream for this issue. Hopefully it solves this issue for anybody running across this on H3/H5 boards.