Jump to content

Odroid HC1 does not boot with kernel 5.4.6

jshc1

By jshc1
in Odroid XU4

 Share

Recommended Posts

jshc1

jshc1

Posted
Posted

My HC1 after changing from 4.14.157 to 5.4.6 using armbian-config is not able to get up ... zero response, no boot
Does HC1 work with kernel 5 and boot, can anyone confirm?

Igor

Igor

Posted
Posted
  On 1/2/2020 at 10:06 AM, jshc1 said:

My HC1 after changing from 4.14.157 to 5.4.6 using armbian-config is not able to get up ... zero response, no boot

Expand  


It must boot, but this change was not tested many times. Perhaps you need to power cycle, worse case installation is broken somehow - boot scripts were not successfully installed.

jshc1

jshc1

Posted
  • Author
Posted

This situation has been around for quite a long time with 5.x and HC1.
Using armbian-config theoretically the installation is the same as in the case of 4.14.x
At the end of the installation, automatic reboot is made and the HC1 does not get up again. Hard power on / off also does nothing. The blue idode will light up for a moment and afterwards go off and nothing happens as if something was wrong with the boot. I have performed the installation 5.x a dozen times always the same so a random error during installation I rather rule out. What's worse, this behavior has always been with 5.x, earlier I just treated it as an error with the early version 5.x but now with 5.4.6 it's a bit too far to ignore it.

Any ideas?

Igor

Igor

Posted
Posted
  On 1/2/2020 at 10:28 AM, jshc1 said:

Any ideas?

Expand  


Check if everything is in place:


- dtb exits and its in correct location

- UUID is correct
- did you only boot from SD but have root on SSD. Did you leave /boot mounted before upgrading

 

I assume you don't have serial console? How can I replicate this - which image to start, what to do to make this happen?

jshc1

jshc1

Posted
  • Author
Posted
  On 1/2/2020 at 11:26 AM, Igor said:

Check if everything is in place:


- dtb exits and its in correct location

- UUID is correct
- did you only boot from SD but have root on SSD. Did you leave /boot mounted before upgrading

 

I assume you don't have serial console? How can I replicate this - which image to start, what to do to make this happen?

Expand  

Everything looks similar between 5.4 and 4.14 ...

Everything is on SD. Nothing modified.
4.14 works normally without problems I do uptime 20-30 days. Only there is always this problem with boot at 5.x
As if something was wrong with armbian-config and the installation procedure or with the 5.x kernel for HC1.

I log in. I run armbian-config and System / Other / I change the kernel ... the typical 5.4 installation procedure begins and the old kernel is removed, after which the ssh connection is broken and after a while it gets reboot and HC1 does not get up anymore. Only the blue LED is lit for a while as if it was looking for a boot and goes off, nothing happens. Red and green behave normally.

dtb exists and points to the right place.

I did the test, restored working 4.14.157 boot partition from backup and manually copied 5.4.6 and changed the links ... the effect is the same no boot and just change the links back to 4.14.157 and it is boot.
It seems that specifically with 5.x for hc there is something wrong ...

UUID is also ok.

Unfortunately, I do not have a cable that would fit the odroid, I only have a regular uart.

5.4.png

4.14.png

arm.png

4a5.png

Igor

Igor

Posted
Posted
  On 1/3/2020 at 4:02 AM, jshc1 said:

I can send you the whole boot partition if it helps...

Expand  

 

It helps best if you do something you know but its project related, not necessarily this problem - perhaps solve some easy script bug or help with developing and maintaining https://github.com/armbian/config or diagnostics tools https://github.com/armbian/autotests? Its in our interest that images works, but when we have to deal with everything but the problem ...

My HC1 with freshly current (not DEV) build 5.4.y kernel boots but I noticed slight stability issue - sometimes restarts at loading kernel, but it always come up. Perhaps kernel is just clocked too optimistically? Out of my quick stress test one could conclude that the problem only appears when the board / PSU are still cold :) 

  Reveal hidden contents

Boot log: http://ix.io/26ge


Regarding kernel switching ... that features might need some hardening and DEV kernels are as is. They work or not.
 

jshc1

jshc1

Posted
  • Author
Posted
  On 1/3/2020 at 7:39 AM, Igor said:

My HC1 with freshly current (not DEV) build 5.4.y kernel boots but I noticed slight stability issue - sometimes restarts at loading kernel, but it always come up. Perhaps kernel is just clocked too optimistically? Out of my quick stress test one could conclude that the problem only appears when the board / PSU are still cold :) 

Expand  

Strange, since your hc is booting hmmm ... If I find the time, I'll try a clean installation.
Whether it's warm or cold I can't see the difference. ;) I have a maximum clock frequency of 2Ghz and 1.5Ghz, boot problems or stability I didn't have until 5.x
The same thing happened at 5.0 but then I thought it was the initial problem of underdevelopment and I hoped that it would be solved over time but remained with me to this day. ;)
But if you say that there are problems with stability, then probably I will stay for now at 4.14. Any plans for 4.14.161 in the near future? Because .157 is a bit old ...

Currently, arm-conf only shows me DEV when it comes to 5.x

 

 

PS
in the attachment both boot partitions ... Working correctly 4.14.157 and problematic with 5.4.6

Kernels.4_14.157-5.4.6.tarFetching info...

nihilista

nihilista

Posted
Posted

Hey, just a little question: On my other boards i can update u-boot within armbian-config, but not on odroid HC1 (system on ssd). Is that normal behavior or did i miss something?

Not a big thing as i can do that with dd...btw. Maybe i have to test with 5.4.y kernel, don't know i need newer u-boot...

Igor

Igor

Posted
Posted

There are no changes in u-boot since year(s). I was just about to check how far we can get with a pure mainline.

 

  On 1/3/2020 at 12:53 PM, nihilista said:

On my other boards i can update u-boot within armbian-config, but not on odroid HC1

Expand  


It's a small bug which is waiting for PR. I have no time and it doesn't bothers me.

Igor

Igor

Posted
Posted

I found last known working configuration ... it something build script related, change from old 5.98 -> v19.11. 

Compile with:

LIB_TAG="v19.08" = stable (default branch)

LIB_TAG="master" = unstable

 

Moving on later or tomorrow.

jshc1

jshc1

Posted
  • Author
Posted

I saw this behavior on one of 5.4.x
I changed to nightly and I probably used current and I had the same symptom as TK, the blue diode is lit all the time and no boot, on DEV, however, as I wrote above, the diode stops lighting and there is no boot.
Of course, I'm talking about an update, maybe with a clean installation something is better and works just like yours ;)

jshc1

jshc1

Posted
  • Author
Posted

So... .161 works ok. ;)


I checked again 5.4 and still no boot. I checked the clean Buster 5.4 installation and the same, 30 seconds heartbeat, and darkness. There is no ssh, hc dead ...
I checked clean 4.14 stretch installations and change to 5.4 again darkness.
My HC1 just has 5.x allergies :( I don't know how your HC1 works with 5.x ;)

Igor

Igor

Posted
Posted
  On 1/6/2020 at 12:26 PM, nihilista said:

I can confirm, no boot with latest 5.4.7 kernel....

Expand  


Me too. I had few successful boots but now it also crashed. I will try few things to see if I can fix this.
 

  Reveal hidden contents

 

nihilista

nihilista

Posted
Posted

Ah, will be fine if that get fixed. No really need for using 5.x kernel, but as it fully supports my dual dvb tuner (4.14 did not) it would be fine to have a working 5.x  kernel. Many thx for all the good work !

Igor

Igor

Posted
Posted
  On 1/6/2020 at 7:48 PM, nihilista said:

Ah, will be fine if that get fixed. No really need for using 5.x kernel, but as it fully supports my dual dvb tuner (4.14 did not) it would be fine to have a working 5.x  kernel. Many thx for all the good work !

Expand  


Sadly this was not a solution :( No quick fix but serious debugging ... when possible.

Igor

Igor

Posted
Posted
  On 1/5/2020 at 11:16 PM, jshc1 said:

I don't know how your HC1 works with 5.x

Expand  


Actually I only does sometimes. But when it does, it works, doesn't hang on numerous boots ... can't reproduce it easy, but I just did. No 5.4.y in production for now.

jshc1

jshc1

Posted
  • Author
Posted
  On 1/6/2020 at 8:09 PM, Igor said:

No 5.4.y in production for now.

Expand  

Can you tell me such a thing ... where is 162 you wrote about, because arm-conf only gives me a maximum of 161. I'm probably blind. :)

It would also be good to delete 5.4 for now, so that no one accidentally gets into a problem and complains about it as with TK recently ;)

Igor

Igor

Posted
Posted
  On 1/7/2020 at 8:42 AM, jshc1 said:

where is 162 you wrote about

Expand  


Self-build.

 

  On 1/7/2020 at 8:42 AM, jshc1 said:

It would also be good to delete 5.4 for now, so that no one accidentally gets into a problem and complains about it as with TK recently

Expand  


That was a different case. TK had issues with a stable 4.14.y kernel build. 5.4.y is noted as testing/experimental.

unixispower

unixispower

Posted
Posted

I can confirm a similar issue using the newest Buster minimal image (Armbian_19.11.7_Odroidxu4_buster_current_5.4.3_minimal) on the HC2. I am able to boot into a freshly written image, but cannot reboot successfully. When running `reboot now` or rebooting when prompted by `armbian-config` selections I hear the installed HDD spin down, see the heartbeat LED go solid, then start blinking as if a reboot happened -- but the system doesn't seem to come back up. The attached HDD never spins back up and I cannot connect via SSH until I pull the power and replug. After needing to manually replug twice (I have tested this a few times now) the HC2 fails to boot altogether with a solid blue heartbeat LED. I have both the HC1 and HC2, but I won't have access to my HC1 until tonight. I would be happy to provide any logs or try to troubleshoot with some direction.

Igor

Igor

Posted
Posted
  On 1/7/2020 at 3:27 PM, unixispower said:

I would be happy to provide any logs or try to troubleshoot with some direction.

Expand  


We have logs, "just" no time / resources to debug since we have to deal with numerous other things. Help there is better then providing us what we already have :) We need to move on with this: https://github.com/armbian/autotests just there is nobody who will lead the project - on a long run it will help us with the problem we have right now. Then https://github.com/armbian/config receive virtually no help with maintaining from community which means its still a security problematic but highly useful and used tool. When there will be more help on general project needs, there will be more time for things like this.


Use 4.14.y and try if you can do something elsewhere.

cyagon

cyagon

Posted
Posted

Can also confirm the problem with the newest 5.4x Kernel. The prior 5.4x kernel worked without problems for me (im using it because of the newest btrfs features).
It boots up normally, SSH is up for a few seconds until the connection fails and my HC1 and becomes unreachable.

It also has current problems when you apply the HC1 DTB via armbian-config, then it will not boot either.
 

nihilista

nihilista

Posted
Posted

Anyone tested new kernel 5.4.19 ? Changelog says "Odroid XU4 current with kernel 5.4.y seems unstable", but what does than mean? Till now it wasn't unstable, it didn't work at all ;-)

And why did same changelog say "Odroid XU4 legacy kernel images instability" ? Is 5.4.x now more stable (because it "seems") than 4.14.x?

I'm bit confused... Btw, its no problem to make an image/backup and test and i know, developers couldnt test all....but little information will be fine :-)

Igor

Igor

Posted
Posted
  On 2/19/2020 at 10:18 AM, nihilista said:

Odroid XU4 legacy kernel images instability

Expand  

 

Default kernel is more or less stock kernel. If Hardkernel breaks it, its broken ... We had troubles elsewhere - a compiler version definition fell out when we move to new branches: default -> legacy, next -> current. Since we didn't compile the kernel with some particular (old) compiler, kernel was unstable. I need to kill a day or two before finding out what was the problem ...


Kernel 5.4.y boots for me on Odroid XU4 with eMMC or from SD card. Also on HC1 ... but people report troubles and we were not able to sucesfully fix them yet. Which is why we label them as "Odroid XU4 current with kernel 5.4.y seems unstable"

Guest
This topic is now closed to further replies.
 Share
Go to topic listing
×
×
  • Create New...

Important Information

Terms of Use - Privacy Policy - Guidelines

  I accept