System hang when login ubuntu desktop with NVMe SSD plugged

[chiang404]

Good day,

When I login Ubuntu desktop, the OS hanged and the PCIe slot power seemed turned off (NVMe SSD LED was off ).
If I unplug the NVMe SSD from the PCIe slot and boot into OS, it was successfully login into desktop and all was normal.

The NVMe SSD can be recognized and normally operated under Ubuntu OS without desktop.
The system would hang during booting or login after installing desktop (I have tried gnome、mate、lxde、xfce...).
 
I want to use NVMe SSD as my storage under desktop environment (GUI).
If any setting is necessary for installing desktop environment I lost ?
Any one has the same issue or can help with your experience ?

PS. The OS image is the latest release in wiki. The board is ROCKPro64 V2.1.

Thanks!

[dukla2000]

Strange - I have no problem with Openbox. It could be that I start from a minimal install - in other places seems the LXDE build has caused problems.

Also, did you follow Bullet64 instructions to set up the NVMe? No magic but may mention something you missed.

[chiang404]

Yes, I have followed Bullet64 instructions to set up the NVMe, but it still failed.
If I remove NVMe SSD and rootfs on SD, it can successfully boot and login into gnome desktop.
If NVMe SSD is plugged and rootfs on SD or SSD, it would fail booting into desktop environment, but booting into command line mode would work normally.

Update: Slackware Aarch64 XFCE (ROCKPro64) works.

[dukla2000]

Update: Slackware Aarch64 XFCE (ROCKPro64) works.

Good that you have some progress. Sad that it indicates another problem starting from the LXDE build.

[chiang404]

I used Bionic minimal and installed kubuntu-desktop. It can successfully login and recognized NVMe SSD.
But install ubuntu-desktop or mate-desktop, the pcie slot power would be off and system hang when login to desktop.

Any one has the same issue ?

[grimace]

I have the same issue.
I am running:

Linux version 4.4.132-1077-rockchip-ayufan-gbaf35a9343cb (root@runner-f725ff63-project-5943294-concurrent-0) (gcc version 7.3.0 (Ubuntu/Linaro 7.3.0-16ubuntu3) ) #1 SMP Mon Jul 30 14:06:57 UTC 2018

Ubuntu 18.04.2 LTS \n \l

if I don't login to the xfce4 desktop then it runs just fine.

I found if I unplug the PCIe card, I can login to xfce4 desktop and everything works fine ( but no M.2 drive available - obviously ).
If I plug in the PCIe card but don't log in to the GUI Desktop, I can boot the rockpro64 and can ssh to it from my laptop.

It doesn't matter if the drive is mounted or not.   

This only happens when the PCIe card is plugged in and I login to the GUI Desktop.   If I don't login to the GUI Desktop,  the machine continues to run, no hang.

I've looked through syslog but can't find any reason for this.

[grimace]

I have the same issue.
I am running:

Linux version 4.4.132-1077-rockchip-ayufan-gbaf35a9343cb (root@runner-f725ff63-project-5943294-concurrent-0) (gcc version 7.3.0 (Ubuntu/Linaro 7.3.0-16ubuntu3) ) #1 SMP Mon Jul 30 14:06:57 UTC 2018

Ubuntu 18.04.2 LTS \n \l

if I don't login to the xfce4 desktop then it runs just fine.

I found if I unplug the PCIe card, I can login to xfce4 desktop and everything works fine ( but no M.2 drive available - obviously ).
If I plug in the PCIe card but don't log in to the GUI Desktop, I can boot the rockpro64 and can ssh to it from my laptop.

It doesn't matter if the drive is mounted or not.   

This only happens when the PCIe card is plugged in and I login to the GUI Desktop.   If I don't login to the GUI Desktop,  the machine continues to run, no hang.

I've looked through syslog but can't find any reason for this.

I'm still having problems with this. Even after update, upgrade, many removals and reinstalls.  Is this a faulty board?  I'm wondering if I should ship it back to manufacturer? I might add that I have 2 completely identical systems, one works, the other doesn't.

[axelf]

Hi, I have discovered exactly the same issue .......   I just posted under the Hardware section !
I am running the RockPro64 v2.1 Board with 64EMMC + PCIe NVMe Board EVO 970 500GB NVMe M.2 

I am using Armbian  rockpro64 4.4.174-rockchip64  which is based on Ayufan's kernel I believe.

Did you ever solve it ?   What kernel are you running now ?

I am not sure if its a physical power design issue or a driver issue .... the HW locks up and you get no debug information in which to troubleshoot.

If it was fixed by a kernel upgrade, the problem seems to have been re introduced !

[Luke]

Hi, I have discovered exactly the same issue .......   I just posted under the Hardware section !
I am running the RockPro64 v2.1 Board with 64EMMC + PCIe NVMe Board EVO 970 500GB NVMe M.2 

I am using Armbian  rockpro64 4.4.174-rockchip64  which is based on Ayufan's kernel I believe.

Did you ever solve it ?   What kernel are you running now ?

I am not sure if its a physical power design issue or a driver issue .... the HW locks up and you get no debug information in which to troubleshoot.

If it was fixed by a kernel upgrade, the problem seems to have been re introduced !

Currently working on the issue. It seems - as odd is its sounds - that the problem is somehow linked to pulseaudio. If you uninstall pulseaudio, and use alsa instead, the issue will just vanish. We have tried blacklisting PCIe for pulse in udev, and it prevents the issue from happening, but it also returns a segmentation error (SATA card / other adapter not accessible). Its very very strange.

[axelf]

Good to know  I have been pulling my hair out troubleshooting the issue as I have a 10 node cluster that is unusable at the moment.

let me know if I can do anything to help .. I have a basic electronics lab.

Are there any Gerber  Files I can look at ? Is there a JTAG interface exposed on the PCB ? ... There are gaps in the documentation a Gerber file would help.


You can see my investigations here on the Armbian forum;

https://forum.armbian.com/topic/10036-si...-crashing/


I also have the wifi and Bluetooth dongle available to test as well.

I do see a PCIe bus collision warning in the kernel and a voltage regulator warning on the vcc12v_pcie and vcc3v3_sys rails. (although vcc3v3_pcie child off the vcc3v3_sys rail does report its ability to invoke a voltage regulator).

I have been monitoring the current and power draws when the hangs happen ... and I don't think it is an issue with the power supply.

let me know how I can help.

the strange thing is .... if I boot in as root and use the armbian desktop instead of XFCE in user space  ... i can avoid the problem by using Firefox instead of chromium !!!!!!!!!! but if I boot in user space and launch XFCE ... it locks up immediately before I even have chance to launch a browser !!... I can eventually cause the board to grind to a halt as root using firefox but I have to go to web sites like lastminute.com and  skyscanner.com but its not a complete HW Freeze ... the mouse still responds. However I can 100% recreate the issue as root by launching chromium and loading the Armbian forum!! We have tried disabling HW Acceleration and chromium is still able to lock the board 100% reproducibility. What are Chrome and XFCE doing that could cause the issues to be more reproducible ?