[SOLVED] [ODROID-XU4] Spontaneous reboot under certain loads

This forum is for supported devices using an ARMv7 Samsung SoC.

[SOLVED] [ODROID-XU4] Spontaneous reboot under certain loads

Postby aphirst » Thu Sep 07, 2017 10:36 am

EDIT: ISSUE SEEMS TO HAVE BEEN POWER-SUPPLY RELATED, SEE MOST RECENT POST.

After some teething problems, I was able to get my new ODROID-XU4 booting fine under ALARM, along with setting up my network shares and other various services I run thereon. So far, so good.

However, I decided to try installing some extra stuff relevant to my mathematical/computational work, specifically the AUR package "openblas-lapack", which is a fairly long build even on desktop hardware.

From a mostly-stock installation of ALARM, the XU4 undergoes an instant reboot about 10-15 seconds into the build process. I spoke with some people on #archlinux-arm on Freenode last night, but nothing useful seemed to be showing up in journalctl or when watching the temps (though I assume there are many sensors I should monitor simultaneously).

I do suspect that it has something to do with load though, since the machine can otherwise idle fine for at least several hours, and other package builds (yet) haven't caused the issue. I think that as part of this package's build process it sets the j makeflag to the number of cores, but I'd have to source that somehow if that's relevant.

I decided to check which CPU governor was set and, to my surprise, it was "performance", which seems a bit overkill for devices like these. I haven't had chance to test "ondemand" or "conservative", but I used cpupower to set "powersave" after which I was able to perform a complete build of this package, albeit not much faster than my should-be-obsolete RPI2.

Short of purchasing the serial debug cable, what can I do to further diagnose this issue, or scrutinise my (various) assumptions?

As an aside, the kernel version I had was:
Code: Select all
core/linux-odroid-xu3 4.9.47-2 [installed]

but I notice just now that there's an update to 4.9.47-3 so I'll apply that and test again later.
Last edited by aphirst on Fri Sep 15, 2017 10:56 am, edited 1 time in total.
aphirst
 
Posts: 13
Joined: Wed Sep 06, 2017 10:09 am

Re: [ODROID-XU4] Spontaneous reboot under certain loads

Postby aphirst » Sat Sep 09, 2017 12:59 pm

I'm still having the issue - I verified again that the reboot occurs on all the CPU governors other than powersave, which I'm afraid is the case.

I got some messages last night in the #odroid Freenode channel which suggested another course of investigation.

Code: Select all
<memeka> aphirst: not sure what kernel arch uses but I remember similar issues a long time ago with mainline kernel - because there were some wrong voltages in mainline. Try and compile the kernel from the HK GitHub


After establishing that ALARM does indeed use hardkernel's, er, kernel:
Code: Select all
<memeka> Maybe try make sure you use xu4 dtb
<memeka> In any case it looks to me like an arch kernel issue


I tried to work out which dtb file is indeed being used (I suspect the potential issue here is that the XU4 actually uses the -xu3 package, and there are different hardware tweaks to be applied based on the actual device), but the boot.txt seems to script this, and there's no indication in e.g. /proc/cmdline. How can I check which .dtb file was loaded at boot?

Also, my UART cable arrived today, so hopefully later today or tomorrow morning I can use my RPI2 to debug the XU4 directly. Watch this space.
aphirst
 
Posts: 13
Joined: Wed Sep 06, 2017 10:09 am

Re: [ODROID-XU4] Spontaneous reboot under certain loads

Postby aphirst » Mon Sep 11, 2017 10:57 am

Ok, I have something from my attempts to debug using the UART cable.

I stopped some unnecessary surfaces, unmounted some disks, and set the "ondemand" governor ("powersave" is set in /etc/default/cpupower). I set up minicom on my RPI2 (which is connected to the XU4 via the UART-USB thing) and then ran `# minicom -w`.
On the XU4 (via a separate SSH session) I then ran makepkg for the package which triggers the reboot (openblas-lapack). The following is what was spat out into the log:

https://pastebin.com/BJaasuMn

Things to note:
  1. At the point of crashing, the only thing emitted before the reboot itself was `[638`. Is there a way to get a bit more info than this?
  2. After crashing and rebooting, the device actually rebooted several times before it stayed online.
  3. It seems that something is going wrong with cgroups (there are several sections of the log which mention this breaking).
  4. There's also an MMC CRC warning. Is this something I need to worry about, and if so, how should I proceed in fixing it?
  5. Once the device came online, it seems to have not bothered to remount my disks (set in fstab), and when manually rebooting,
    the device didn't come back online, necessitating me unplugging then replugging the power adapter (after waiting a few seconds).

Does this yet shed any more light on the situation? How should I proceed from here to get more info?
aphirst
 
Posts: 13
Joined: Wed Sep 06, 2017 10:09 am

Re: [ODROID-XU4] Spontaneous reboot under certain loads

Postby aphirst » Fri Sep 15, 2017 10:42 am

So, I got hold of the 6A adapter (I had previously been using the 4A adapter) and decided to test with that.

The device booted fine, I ensured that the CPU governor was set to "ondemand", then ran this culprit build task.

No reboot, no serial console errors. The build wasn't successful, which I at first thought was a problem, but I tried again after clearing the build directory after which it worked just fine.

So, it looks as if this actually was power related. I don't have the means to test whether it's due to there having not been enough current with the old adapter, or whether it was actually faulty in some way, but I'm regardless going to contact the supplier and see whether they're willing to accept a return in this case.
aphirst
 
Posts: 13
Joined: Wed Sep 06, 2017 10:09 am


Return to Samsung

Who is online

Users browsing this forum: No registered users and 2 guests