Jump to content

Next LTS kernel 4.19.y Allwinner A10, A20, A64, H2+, H3, H5, H6 debugging party


Igor

Recommended Posts

Earlier today I pushed a fairy large patchset containing various functional improvements of many boards.  If you have Allwinner board and some spare time:

 

1. Build DEV image/kernel with https://github.com/armbian/build (you need to add EXPERT="yes" to the config to unlock)

2. Install DEV kernel from beta repository

 

  1. Optional Defreeze kernel updates
  2. Switch to nightly kernel (armbian-config -> system -> Nightly)
  3. Reboot
  4. Switch to other kernel (armbian-config -> system -> Other -> DEV)

 

When board came up, do some exploration. Most important information is to find out if there is a regression toward kernel 4.14.y! Then make a test report https://github.com/armbian/testings#how-to-create-a-test-report
 

If you know how to fix certain problems, you are more than welcome! Our resources are tiny and we can't possible fix all problems :( 

 

This topic is a place to discuss how certain problems/bugs can be solved. 

 

When reporting a bug, provide logs with: armbianmonitor -u

 

Bugs:

- serial gadget console is not working (anywhere?)

- Pinebook doesn't boot properly

- Mesa (OS Mali drivers are enabled by default) / WebGL works on Debian based Chromium, fails on Ubuntu

 

I checked those:

 

 

With problems:
 

  • Pinebook 

 

Confirmed working:
 

  • Neo2 v1.1 512MB
  • Neo2 v1.1 1GB
  • Pine64
  • Orange Pi Zero Plus2 H5
  • Nanopi Duo http://ix.io/1uVC
  • Orangepi R1 http://ix.io/1uaP

  • Nanopi Neo Core 2 LTS
  • Nano Pi Neo Plus2
  • Tritium H3 and H5
  • Orange Pi Zero Plus
  • Bananapi M1

 

For now.

P1110540.JPGDry2Q06X4AAKH2J.jpg

 

BTW.

Do you want to become a (Allwinner) board maintainer? Duties:
- responsible for content at the download page,

- running latests updates and managing bug list there.

Link to comment
Share on other sites

6 hours ago, Igor said:

 

BTW.

Do you want to become a (Allwinner) board maintainer? Duties:
- responsible for content at the download page,

- running latests updates and managing bug list there.

 

can you go into a little more detail about managing bug list?

 

I have an embarrassing quantity of allwinner boards.

Link to comment
Share on other sites

4 hours ago, lanefu said:

can you go into a little more detail about managing bug list?


https://linux-sunxi.org/Linux_mainlining_effort#Status_Matrix
Essentials which should work, one board at the time, focus into the diff - what worked in 4.14.y and it doesn't work anymore. We should find very little. One example is DVFS not working on Orangepi Prime. Regulator bits were change a bit ... while they work at Orangepi PC2 which is the same in this manner. If you understand how to fix this, you fix it, otherwise you place a note here and ... I hope someone else will do this, else I will.
 

Link to comment
Share on other sites

18 minutes ago, TonyMac32 said:

Tritium tested, fixed (DTS/config in u-boot back-ported from 2018.09), and complete with "testings" files merged.  I see NanoPi K1+ is done already, so my work here is done.  :-)


Great. Tritiums h3&h5 checked.

Can someone stress one of dumb regulator board too see if it doesn't result in a magic smoke ? :) Is there any heal for:

Spoiler

[    5.858164] core: _opp_supported_by_regulators: OPP minuV: 1320000 maxuV: 1320000, not supported by regulator
[    5.858177] cpu cpu0: _opp_add: OPP not supported by regulators (1056000000)
[    5.858268] core: _opp_supported_by_regulators: OPP minuV: 1320000 maxuV: 1320000, not supported by regulator
[    5.858274] cpu cpu0: _opp_add: OPP not supported by regulators (1104000000)
[    5.858376] core: _opp_supported_by_regulators: OPP minuV: 1320000 maxuV: 1320000, not supported by regulator
[    5.858382] cpu cpu0: _opp_add: OPP not supported by regulators (1152000000)
[    5.858454] core: _opp_supported_by_regulators: OPP minuV: 1320000 maxuV: 1320000, not supported by regulator
[    5.858460] cpu cpu0: _opp_add: OPP not supported by regulators (1200000000)
[    5.858563] core: _opp_supported_by_regulators: OPP minuV: 1340000 maxuV: 1340000, not supported by regulator
[    5.858570] cpu cpu0: _opp_add: OPP not supported by regulators (1224000000)
[    5.858667] core: _opp_supported_by_regulators: OPP minuV: 1340000 maxuV: 1340000, not supported by regulator
[    5.858673] cpu cpu0: _opp_add: OPP not supported by regulators (1248000000)
[    5.858785] core: _opp_supported_by_regulators: OPP minuV: 1340000 maxuV: 1340000, not supported by regulator
[    5.858796] cpu cpu0: _opp_add: OPP not supported by regulators (1296000000)
[    5.858917] core: _opp_supported_by_regulators: OPP minuV: 1400000 maxuV: 1400000, not supported by regulator
[    5.858925] cpu cpu0: _opp_add: OPP not supported by regulators (1344000000)
[    5.859059] core: _opp_supported_by_regulators: OPP minuV: 1400000 maxuV: 1400000, not supported by regulator
[    5.859070] cpu cpu0: _opp_add: OPP not supported by regulators (1368000000)
[    5.869186] input: gpio_keys as /devices/platform/gpio_keys/input/input9
[    6.164558] sun8i_ths 1c25000.thermal-sensor: no memory resources defined
[    6.164588] sun8i_ths: probe of 1c25000.thermal-sensor failed with error -22
[    6.212018] thermal thermal_zone0: failed to read out thermal zone (-110)

 


 

Link to comment
Share on other sites

On 11/15/2018 at 10:05 PM, Igor said:

2. Install DEV kernel from beta repository (currently in making -> available from November 16th)

  1. Optional Defreeze kernel updates
  2. Switch to nightly kernel (armbian-config -> system -> Nightly)
  3. Reboot
  4. Switch to other kernel (armbian-config -> system -> Other -> DEV)

 

 

Hi @Igor Could you give me a hint/command for installing the DEV-Kernel?

I did compile DEV for the NanoPi Nero Core2, switched to nightly but I cant switch to DEV-Kernel (Other) because I dont know how to install this kernel.

 

PS: I did create a report without switching - maybe you could delete this one? 

Link to comment
Share on other sites

2 hours ago, guidol said:

I did compile DEV for the NanoPi Nero Core2, switched to nightly but I cant switch to DEV-Kernel (Other) because I dont know how to install this kernel.

 

 

If you compiled on your own just transfer deb files (image and dtb at minimum) to your board and install with dpkg -i *.deb ... 

 

beta repository:

1. remove kernel and deb if it is not DEV already. apt purge linux-image* linux-dtb*

2. apt install linux-image-dev-sunxi64 linux-dtb-dev-sunxi64 (without 64 for 32 bit boards)

Link to comment
Share on other sites

34 minutes ago, rusatch said:

Hi. Is there overlays works on dev? I mean that as i know, overlays not works on dev and usb not available on OPi h5 with dev kernel. Or i missed something? )

Overlays are working on DEV since awhile, depending which SoC, I've recently added some overlays for H6 ...

What do you means by "usb not available on OPi h5" ? Which board did you try ? On some boards, you need to load usbhostX overlays ...

Link to comment
Share on other sites

2 hours ago, martinayotte said:

What do you means by "usb not available on OPi h5" ? Which board did you try ?

I mean that usb works only with overlays with mainline and next kernel, as i know. I was try dev kernel earler, and usb did not available.

I use OPi zero 2 plus with usb-lan dongle, and working usb for me is critical.

Link to comment
Share on other sites

3 minutes ago, rusatch said:

I mean that usb works only with overlays with mainline and next kernel, as i know. I was try dev kernel earler, and usb did not available.

I use OPi zero 2 plus with usb-lan dongle, and working usb for me is critical.

Do you have this "overlays=usbhost2 usbhost3" in /boot/armbianEnv.txt ?

If you don't want to use overlays, you can manually edit the Main DT and enable all "ehci" and "ohci" using "okay" statuses...

Link to comment
Share on other sites

42 minutes ago, rusatch said:

I was try dev kernel earler, and usb did not available.

 

Please provide armbianmonitor -u since we need to know exactly what you are talking about. 

 

42 minutes ago, rusatch said:

working usb for me is critical.


Likewise and this topic here is to make sure it works before we push this out to all users. If it doesn't work at this moment, this is expected. In some cases.

Edited by Igor
typo
Link to comment
Share on other sites

1 hour ago, rusatch said:

I mean that usb works only with overlays with mainline and next kernel, as i know. I was try dev kernel earler, and usb did not available.

I use OPi zero 2 plus with usb-lan dongle, and working usb for me is critical.

 

Hi @rusatch, if you are using the 1.3GHz CPU overclock overlay in your "overlays=" line for your H5 then this could be the cause of your problem (overlay "cpu-clock-1.3GHz").  (If this isn't the case for you, then it's odd because USB works fine for me on four different boards I tried with the new 4.19 kernel...)

 

This overlay needed to be updated for 4.19.y, and if you load it by default in the current build then the load will fail and the DT will reset to the board default.  I just pushed an update for this for the new kernel (see https://github.com/armbian/build/commit/06451d37a64a4fd2599207722c73c9c160b63100).

 

The new 4.19.y mainline includes a much more complete cpu opp frequency table (nice!), with frequency and voltage definitions all the way to 1.368GHz.  For boards that have real DVFS-capable regulators then the higher frequencies will work by default.  So I updated the 1.3GHz overclock overlay hack for boards that only provide a fixed 1.3v maximum, and the new name for the overlay is "cpu-clock-1.3GHz-1.3v".

 

(Side note:  I removed the previous "cpu-clock-1.4GHz" overlay as given that the maximum clock speed default at 1.4v is now 1.368GHz it's not really worth maintaining one that goes to 1.396GHz.)

 

Link to comment
Share on other sites

On 11/15/2018 at 10:10 PM, Igor said:

Can someone stress one of dumb regulator board too see if it doesn't result in a magic smoke ? :)

 

Hi @Igor, I've tried a top-of-tree H5 build so far on:

  • Neo2 v1.1 512MB
  • Neo2 v1.1 1GB
  • Orange Pi Zero Plus2 H5
  • Nano Pi Neo Plus2

...and all works well.  Very nice!!  :)  (These are all dumb regulator boards BTW)

 

Tests include USB (Wi-Fi and Ethernet dongles), H/W SPI interface (via spidev), multiple GPIOs (via libgpiod).  I've switched to 4.19.y for all of these now.

 

(I still have a few more boards that I'll dig up and try it on...)

 

Link to comment
Share on other sites

@Igor - in 4.19.y it looks like the CPU temperature now reported by "/sys/class/thermal/thermal_zone0/temp" is ~10C greater than it was in previous kernels for the H5.  The "sun4i-gpadc-iio.c" driver in 4.19.y has numerous changes compared to the previous kernels, and I'm not sure if the previous value reported was incorrect and the new value is correct, or vice-versa.  This is consistent across all my different H5 boards.  Previously my H5 boards would idle around 30C (reported), now they idle at around 40C (reported):

root@orangepizeroplus2:/sys/class/thermal/thermal_zone0# cat temp
40341
root@orangepizeroplus2:/sys/class/thermal/thermal_zone0#

Unfortunately I don't have a thermometer handy to check the case temperature of the CPU, but by finger test on the boards the CPU appears to physically be at the same temperature as with the previous kernel...  All boards are running idle with no activity (e.g., 240MHz/480MHz clock speed).

 

@Icenowy might you have any ideas regarding this?

Link to comment
Share on other sites

7 minutes ago, 5kft said:

it looks like the CPU temperature

 

Yes I did notice odd numbers but forgot to check. I do have some thermometer around but I am not close to the boards until tomorrow evening.

Link to comment
Share on other sites

I found a bug with Orange Pi Prime: rebooting works fine but shutting down with poweroff command and cycling the power makes booting freeze at various stages.

 

More info:

there are four packages that did not updated with armbian-config when I was switching to nightly builds and later to dev branch:

armbian-firmware/stretch,stretch 5.67.181116 all [upgradable from: 5.65]
armbian-tools-stretch/stretch 5.67.181116 arm64 [upgradable from: 5.65]
hostapd/stretch 3:2.6-99~armbian5.67.181116+1 arm64 [upgradable from: 3:2.6-99~armbian5.60+1]
sunxi-tools/stretch 1.4.2-2~armbian5.67.181116+1 arm64 [upgradable from: 1.4.2-2~armbian5.60+1]

I did fresh reinstalls two times: updating these packages after switching to nightly builds and after switching to dev branch. That made no difference, as soon as I power cycle the dev kernel it freezes at boot regardless of the versions of these packages.

 

Other than that the stability and functionality of the dev kernel seem to be great.

Edited by lomady
more info
Link to comment
Share on other sites

13 minutes ago, Icenowy said:

@5kft H5 is troublesome on thermal sensor, because the raw value - temperature function is not a single line, but two segments of line.

 

I don't know how is this dealed in sun4i-gpadc-iio.

 

Thanks @Icenowy!  I will try to dig further into this when I get time...!

Link to comment
Share on other sites

8 hours ago, lomady said:

shutting down with poweroff command and cycling the power makes booting freeze at various stages.

 

Power off on H3/H5 is still not implemented AFAIK and don't know if this can be hacked somehow.
https://forum.armbian.com/search/?q=arisc

https://github.com/Icenowy/h3-arisc-shutdown

 

It works on boards which have external PMU, like A64.

Link to comment
Share on other sites

4 hours ago, Igor said:

 

Power off on H3/H5 is still not implemented AFAIK and don't know if this can be hacked somehow.

I just wanted to make a clean shutdown before pulling the power plug.

The problem is that the Prime does not boot after I plug the power back in. As I said it freezes at various stages as seen in the boot log on the monitor connected via HDMI. I even saw a message "Unable to handle kernel paging request at address xxxxx" once.

Link to comment
Share on other sites

Quick note - the Nano Pi Neo Core2 works as well with the new kernel.  Very nice that it clocks up to 1.37GHz by default now!

root@nanopineocore2:~# cat /proc/version
Linux version 4.19.2-sunxi64 (root@gandalf) (gcc version 7.2.1 20171011 (Linaro GCC 7.2-2017.11)) #11 SMP Sun Nov 18 13:40:50 UTC 2018
root@nanopineocore2:~# cpufreq-info -c 0
cpufrequtils 008: cpufreq-info (C) Dominik Brodowski 2004-2009
Report errors and bugs to cpufreq@vger.kernel.org, please.
analyzing CPU 0:
  driver: cpufreq-dt
  CPUs which run at the same hardware frequency: 0 1 2 3
  CPUs which need to have their frequency coordinated by software: 0 1 2 3
  maximum transition latency: 2.04 ms.
  hardware limits: 120 MHz - 1.37 GHz
  available frequency steps: 120 MHz, 240 MHz, 480 MHz, 648 MHz, 816 MHz, 960 MHz, 1.01 GHz, 1.06 GHz, 1.10 GHz, 1.15 GHz, 1.20 GHz, 1.22 GHz, 1.25 GHz, 1.30 GHz, 1.34 GHz, 1.37 GHz
  available cpufreq governors: conservative, userspace, powersave, ondemand, performance, schedutil
  current policy: frequency should be within 408 MHz and 1.37 GHz.
                  The governor "ondemand" may decide which speed to use
                  within this range.
  current CPU frequency is 480 MHz (asserted by call to hardware).
  cpufreq stats: 120 MHz:0.01%, 240 MHz:0.04%, 480 MHz:73.59%, 648 MHz:0.10%, 816 MHz:0.01%, 960 MHz:0.02%, 1.01 GHz:0.00%, 1.06 GHz:0.01%, 1.10 GHz:0.00%, 1.15 GHz:0.01%, 1.20 GHz:0.01%, 1.22 GHz:0.00%, 1.25 GHz:0.02%, 1.30 GHz:0.04%, 1.34 GHz:0.00%, 1.37 GHz:26.15%  (55)
root@nanopineocore2:~#

 

Link to comment
Share on other sites

18 hours ago, Igor said:

Yes I did notice odd numbers but forgot to check. I do have some thermometer around but I am not close to the boards until tomorrow evening.

 

I found an old laser IR thermometer I have and when aiming measuring the heatsink on the Core2 it doesn't seem too far off (e.g. 32C measured vs 34 reported via "temp").  Perhaps the previous implementation was incorrect?  I'm curious what your results will be!

Link to comment
Share on other sites

All by "boards garden" is now migrated to 4.19.y, nothing special to mentioned since all my boards are running headless, so I've never tested HDMI or Audio.

 

One small thing I've seen that bother me a little, and I didn't figured out how to fix, I don't recall how :

even if I have "verbosity=7" in /boot/armbianEnv.txt, kernel start booting with that verbosity, but suddenly becomes quiet until the login prompt.

Doesn any one remember which settings, probably in systemd, which cause that quietness ?

 

Link to comment
Share on other sites

Quitness is due to changing default console to HDMI/serial console login. IMO its better that every day Joe see something on the hdmi screen. Both would be perfect ...

Wrote on mobile

Link to comment
Share on other sites

33 minutes ago, Igor said:

Quitness is due to changing default console to HDMI/serial console login

Where is that settings ? It wasn't like that in 4.18.12 ... Personally, I prefer keeping serial since all my boards are headless.

 

34 minutes ago, Igor said:

Joe see something on the hdmi screen.

But in many of my post, I've encouraging Joe to get USB-TTL Serial dongle in their toolbox ... :P

 

35 minutes ago, Igor said:

Both would be perfect ...

Is it possible ?

Link to comment
Share on other sites

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

Important Information

Terms of Use - Privacy Policy - Guidelines