Le Potato - writing armbian to eMMC

[sbc_chrisb]

Yep, thanks for the reminder. I'm familiar with creating those scr files.

 

I've been waiting for a loooooong time for the nand-sata-install command on the stock download you mentioned, I think it's stuck or something. Just sitting at "Counting files.....few seconds" for, oh maybe two hours now. I'm pretty sure it's not supposed to take that long. About to abort and see what command line options that command has to see if I can get a log file or something. 

 

I was originally using an older armbian image and had a similar experience with it, which was why I had tried balbes images, which worked great for that command. Not really sure what's going on with the stock Armbian command. Doing more investigating. 

[TonyMac32]

Hmm ok.  I have not been tracking the development/maintenance of the nand-sata-install, it took about 45 minutes for mine to get to the "reboot now" screen.  I was using a 4.18 Stretch build.

[sbc_chrisb]

I found a log in /var/log/nand-sata-install.log, it looks like it was looking for open files still when I aborted? I'm not sure what files might be missing.

 

Tue Sep 18 17:28:10 UTC 2018: Start nand-sata-install.


Old UUID:  UUID=763b6c65-4dfe-4e7b-8180-426b7298fddf
SD UUID:   UUID=dc031793-8691-4641-96a0-9520084fe3b0
UUID=763b6c65-4dfe-4e7b-8180-426b7298fddf
SATA UUID: UUID=dc031793-8691-4641-96a0-9520084fe3b0
eMMC UUID: UUID=dc031793-8691-4641-96a0-9520084fe3b0 ext4
Boot: $1 /dev/mmcblk0p1 ext4
Root: $2 /dev/mmcblk0p1 
Usage: 1007
Dest: 6781

/etc/fstab:
UUID=763b6c65-4dfe-4e7b-8180-426b7298fddf / ext4 defaults,noatime,nodiratime,commit=600,errors=remount-ro 0 1
tmpfs /tmp tmpfs defaults,nosuid 0 0

/etc/mtab:
/dev/mmcblk0p1 /mnt/nand-sata-install.jY1gTX/bootfs ext4 rw,relatime,data=ordered 0 0
/dev/mmcblk0p1 /mnt/nand-sata-install.jY1gTX/rootfs ext4 rw,relatime,data=ordered 0 0
/dev/mmcblk1p1 / ext4 rw,noatime,nodiratime,errors=remount-ro,commit=600 0 0
/dev/mmcblk1p1 /var/log.hdd ext4 rw,noatime,nodiratime,errors=remount-ro,commit=600 0 0
/dev/zram0 /var/log ext4 rw,relatime,block_validity,discard,delalloc,nojournal_checksum,barrier,user_xattr,acl 0 0

Files currently open for writing:

Trying to stop running services to minimize open files:

Checking again for open files:

Copying 37009 files to /dev/mmcblk0p1. Finishing full install to eMMC.

Checking again for open files:

Tue Sep 18 18:34:28 UTC 2018: Finished

 

This stock image appears to be using kernel 4.14, that was one difference with the balbes images in that they used the 4.19rc kernel. I'm not finding any man page nor any options calling the nand-sata-install command via -h, so I don't know if there's any way to turn up the debugging on it. 

 

I'm going to try getting internet to the device and doing apt upgrades on the sd card that I'm installing with, then re-running the nand-sata-install command, and seeing if that helps. 

[TonyMac32]

Yes, the Next images are 4.18, I will be turning my attention to 4.19 and it will eventually become the default image. I don't mess with the building if official images (honestly don't know if I can/how), so that will be Igor's decision.

Sent from my Pixel using Tapatalk

[sbc_chrisb]

Okay, so I left it running for a while, almost 4 hours. It was stuck in the same "checking for open files" spot. I think at this point I'm going to look up some docs and build a newer kernel for this before I proceed any further, that was one factor that seemed to have success at least with this. Also, my USB wifi dongle is an ath9k_htc device and there doesn't appear to be a driver pre-built with these images (I enabled non-free and installed the firmware-atheros package and there's still no kernel module for the device), so I'm going to try building that as a module as well.  

 

 

Side note, these emmc modules appear to have two extra 4MB devices glued on, showing in lsblk. I copied them off with DD to inspect with a hex editor to try to figure out what they are, but does anyone happen to know? I didn't see that in any docs. 

[TonyMac32]

4 minutes ago, sbc_chrisb said:

Also, my USB wifi dongle is an ath9k_htc device and there doesn't appear to be a driver pre-built with these images

Hmm, I see it is not enabled.  I will enable as module/build it to test for future, our other kernels have it.

 

5 minutes ago, sbc_chrisb said:

Side note, these emmc modules appear to have two extra 4MB devices glued on, showing in lsblk. I copied them off with DD to inspect with a hex editor to try to figure out what they are, but does anyone happen to know? I didn't see that in any docs. 

eMMC has a partition format unlike SD, so there is a boot and an environment partition, if I remember correctly.

[sbc_chrisb]

Whew. Okay, so that was a fun ride. 

 

My suspicions about the kernel were correct. I followed the docs to do a new kernel build for armbian, 4.18.19. I hadn't built a kernel or anything large like that in a while, really stress tested my AMD 8350. Either way, I got the new packages over onto the Potato, installed them (with a --force-all to get uboot to update, since it refuses initially). Then after another reboot, nand-sata-install completed successfully.

 

First boot without an SD card failed hard. Went back in with the SD card, mounted the emmc and mv'd the boot.cmd and scr files out of the way and put in the one Tony provided me above. Rebuilt the scr with mkimage. That did the trick, I'm now running on emmc now. Thanks for the help, Tony. 

 

To recap, the stock image is fine to install to emmc *if* you:

 

1. build a new 4.18 kernel, hopefully the next armbian release will be fine

2. run nand-sata-install

3. mount the emmc and modify the boot.cmd command *in the emmc boot dir* with the updated file linked above, then rebuild the scr with mkimage. NOTE: don't use the command in the bottom of the file directly, as it gives absolute paths for the files and since you're on the SD card the path is different. Use the following while in the boot dir of the emmc:

mkimage -C none -A arm -T script -d boot.cmd boot.scr

 

 

[sbc_chrisb]

Follow up post, in case anyone ends up here in searches on this topic. 

 

Once i had this working, I then repeated it using my original SD card install with what started out as an older version of Armbian, upgraded via apt. I had a few minor bumps I'm going to explain here.

 

1. The new kernel is a must. The stock kernel did not work, I had to reuse the 4.18.19 kernel packages I built before.

2. The updated boot script I keep mentioning has to be tweaked. It will try to use mmcblk1p1 for the root device, but on my board the mmc is mmcblk0. The device would not boot from mmc until I modified the boot.cmd to set the rootdev and rebuilt the scr.  I'm not sure if this is expected behaviour or not.

  

[TonyMac32]

Not expected, root dev should be modified in the armbianEnv.txt file.

Sent from my Pixel using Tapatalk

[Igor]

Just a quick note here. Nand-sata-install utility works if you don't mess up with a system. On official images and on supported boards. For 3rd party I don't know and on Lepotato AFAIK we didn't get this far yet? (I could easily miss this)

 

@TonyMac32 Shell I rebuild all Lepotato images? Or later when bumped to 4.19.y?

[TonyMac32]

I think wait until 4.19, I was working with @sbc_chrisb to get an image with the typical Armbian package support (kernel/etc).  The updated boot script and a fresh 4.18 build can identify and install armbian to the eMMC without boot issues.  I assume the issue being seen by Chris has to do with the older image being updated, as you mentioned.

 

Just for sanity I booted my Le Potato up with only the eMMC present again to make sure.  ;-)  Rebuilt images should work, if you would want to test (assuming you have an eMMC...)

[sbc_chrisb]

58 minutes ago, TonyMac32 said:

Not expected, root dev should be modified in the armbianEnv.txt file.

Sent from my Pixel using Tapatalk
 

 

Okay, I might have goofed on something here. I'll fire it back up in a bit, revert my boot.cmd modification, and retest. 

 

I can confirm, with both an older armbian image that was upgraded via apt, and with a brand new image downloaded from the main site, that nand-sata-install does not work. The only way I got past that was rebuilding a newer 4.18.19 kernel, and uboot. After that, I required the use of Tony's updated boot.cmd in order to boot from it, it would not boot by default at all. 

[TonyMac32]

2 minutes ago, sbc_chrisb said:

After that, I required the use of Tony's updated boot.cmd in order to boot from it, it would not boot by default at all.  

 

Wouldn't boot from eMMC or from SD?  It should still work from SD with old boot script, just not from eMMC.

[sbc_chrisb]

1 minute ago, TonyMac32 said:

 

Wouldn't boot from eMMC or from SD?  It should still work from SD with old boot script, just not from eMMC.

 

Yes, sorry that statement is correct, it would still boot from SD which allowed me to go back in and fix it. I don't have the exact logs of that point but I believe that was when it was hanging at trying to remount the rootfs and failing, so it just sat there hung until I inserted the SD card. 

[TonyMac32]

OK, so, for you/anyone doing this before it's officially baked in:

 

Download 4.18 image

Build newest 4.18 kernel/u-boot

update boot script with the one currently in our repo

reboot (because of superstition)

run nand-sata-install

 

I would recommend being on the moderately advanced side before attempting it, unless you really want to experiment.

 

 

[sbc_chrisb]

1 hour ago, TonyMac32 said:

Not expected, root dev should be modified in the armbianEnv.txt file.

Sent from my Pixel using Tapatalk
 

 

 

Okay, so I reverted boot.cmd to your version, ie put the rootdev back to /dev/mmcblk1p1, then created armbianEnv.txt with the following line:

 

setenv rootdev "/dev/mmcblk0p1"

 

 

Rebuilt the scr. Rebooted. Fails to boot, this is the captured output:

Begin: Running /scripts/local-block ... done.
done.
Gave up waiting for suspend/resume device
done.
Begin: Will now check root file system ... fsck from util-linux 2.29.2
[/sbin/fsck.ext4 (1) -- /dev/mmcblk1p1] fsck.ext4 -a -C0 /dev/mmcblk1p1 
LPotato_SD: clean, 13/1954064 files, 166731/7814912 blocks
done.
done.
Begin: Running /scripts/local-bottom ... done.
Begin: Running /scripts/init-bottom ... mount: No such file or directory
mount: invalid option --
done.
mount: No such file or directory
run-init: opening console: No such file or directory
Target filesystem doesn't have requested /sbin/init.
run-init: opening console: No such file or directory
run-init: opening console: No such file or directory
run-init: opening console: No such file or directory
run-init: opening console: No such file or directory
run-init: opening console: No such file or directory
No init found. Try passing init= bootarg.
(initramfs) 

 

So, while on a fresh Armbian image the boot.cmd you posted works completely 100% as is, when using an older image that was updated via apt and etc seems to require me to directly change the cmd and rebuild scr. 

 

Unless I'm doing something wrong with the ArmbianEnv.txt, which is entirely possible. I'm not really familiar with that file. 

[TonyMac32]

The nand-sata-install script does that adjustment for you, that's the issue I think.  https://github.com/armbian/build/blob/master/packages/bsp/common/usr/sbin/nand-sata-install#L205

 

I'm building a model train speed controller to replace the 1970's one that just failed on my son, a bit distracted.  :-)

[sbc_chrisb]

No worries, I just wanted to confirm what you asked. I have it working with the tweak so I'm happy at this point. I'm also working on restoring an old TRS-80 Color 1 right now so I understand the distraction  

[TonyMac32]

20 minutes ago, sbc_chrisb said:

TRS-80 Color 1 right now

I have a CoCo 2 and a Model 100 Portable. 

[Igor]

I was trying to use this eMMC today (SD cards boots 4.14.y and 4.19.y OOB) and I am getting a bit confused ... Is DDing from X to Y prevent starting Android and we can boot this way from SD card ... and then installing Armbian to eMMC? Is that not working or what?

[TonyMac32]

Ah, I did "the wrong thing" very early on and dd'd garbage over the boot sector to kill Android.  Then you can boot SD.  There is also an Amlogic download tool I've never used...

[sbc_chrisb]

Agreed, I was pulling my hair out trying to install to mmc with any image as it would not flush the partitions, until I finally just tried dd'ing an armbian image directly to the mmc module, which flushed out the Android stuff, and then from there I did the stuff I mentioned in my posts. I could not get any installation to work while the Android stuff was present. Even with balbes' instructions, none of it worked until I used dd.

[TonyMac32]

The Amlogic flash tools will be relevant when discussing La Frite as well, should we support it (I think the S805X is to the S905X what the H2+ is to the H3). I'll need to review their function (I actually plugged the eMMC in during boot of the SD to get it in a position where I could work with it the first time, not a fun task)

Sent from my Pixel using Tapatalk

[balbes150]

I do not follow this topic, so I just saw that there are problems. Because LePotato does not have a default u-boot on the eMMC, it is sufficient to use the u-boot-2018 installation mode (described in the instructions) to install my images on the eMMC.

 

Forgot to add, since LePotato has its own version of u-boot-2018, you need to use the file from it. I know that's the file @Neil Armstrong has.

[Neil Armstrong]

We plan to use https://github.com/superna9999/pyamlboot for LaFrite, but it works on LePotato aswell, and it will be supported by the next U-boot version (in master right now https://github.com/u-boot/u-boot/commit/d96a782d09dbdc4a28ece3d18dc17a572e39d4f2)

We still need to generate a generic initramfs+kernel image to flash anything.

[WZ9V]

I've read the whole thread and only understand about 50% of it.  To summarize is there a relatively easy way to get the current stable Armbian Stretch (4.14y) to boot from eMMC on the Le Potato?

 

I have it working on a Tinkerboard S but the darn thing only goes about two minutes at 100% before throttling in the Pi-Top V2 case I'm using.  The Le Potato I've stressed for several hours in the same case and it remained full speed but I'd rather it run off of the eMMC if possible.  If not I'll just run off of SD for now since it sounds like the next bump will have a way to do it that doesn't require super kernel kung fu powers.

[TonyMac32]

22 hours ago, WZ9V said:

I have it working on a Tinkerboard S but the darn thing only goes about two minutes at 100% before throttling in the Pi-Top V2 case I'm using.

 

That stays cool forever, but won't fit in the pi-top I think.  An alternative could be https://www.aliexpress.com/item/3D-Printer-Cooler-Fan-DC-12V-24V-Connector-Brushless-Cooling-Turbo-Blower-Fan-50x15mm-3D-Printer/32944813107.html

Remember the Tinker Board is one of the highest performing boards you can buy, the RK3399 devices will do better due to way nicer interfaces (USB3, PCIe, etc), but the mainline support is not as good, and the Rockchip kernel is a bit of a mess, thankfully for 64-bit targets Ayufan has a nice fork.  The Le Potato has 64-bits on it's side, and an active team working on the mainline support.

 

Given the mobile nature (batteries, yes?) I'd look into a small circuit and a script to toggle the fan based on cooling state, so it's not just "on".  If you are really adventurous, moving the debug UART to UART3 frees up 2 PWM pins.  That reminds me I need to double check the vendor kernel and make sure we're matching it, there was some chaos this year...

 

For Le Potato, 4.14.y from armbian download page does not have an easy way to be flashed to eMMC.  4.19.y works with the out of the box tools, but screen results may vary.  I have not tried the amlogic downloader, my guess is it requires Android-looking images, which we do not create.  I will be starting to test @Neil Armstrong's u-boot update as soon as I have a moment, look for it in January/February since that will be U-boot 2019.1. 

 

BTW, Neil, I love this README: 

 

 

[WZ9V]

Thanks, I suspected the eMMC setup wasn't quite there yet based on what I did catch in the discussion.  I was hoping someone would say

1) Boot board without eMMC (to force it to boot from SD)

2) Install eMMC

3) Run the utility in armbian_config

 

Is there a way to run the eMMC copy in armbian_cofig from the commend shell via ssh?  Because of the way Libre suggests installing after boot I have to remove it from the Pi-Top and don't have a handy HDMI monitor at this time.

 

 

I'll check the Tinkerboard cooling you recommended since I suspect that even outside of the Pi-Top it will need more than the supplied passive heatsink.  Yes the fan setup wouldn't fit unless I redid the "patented" bridge that only works if the CPU is in EXACTLY the same place as on the rPi.  Where did you get the heatsink and fan in teh picture?

[TonyMac32]

Amazon was where I got them, however last I checked the heatsinks were no longer available for that seller. If you search a geometry that fits it shouldn't be hard to find something.

Sent from my Pixel using Tapatalk

[WZ9V]

I ordered some 20x20x2.5 copper spacers, a 25x50x10 heatsink, and a 25mm 5V fan.  That seemed similar to the picture.  I'll try that in the CEED again since it has more space but then I'll probably run into the hub V1 problems I still owe you debug output for.  HDMI worked with 4.4 but fails with 4.14.