destroyedlolo

Thanks for links. I'm not concerned by AR-305 as I have an HDMI attached (It's my home automation dashboard, images you can see on my Twitter profile). But i don't have a SATA disk on this machine. The only google hits I found about this issue is related to disk access : as the only one I have is the SD card, I'm a bit fearing about its duration. But nothing is clear about this particular kworker : what is its goal ?

Hello, I'm running Armbian on the original (LeMaker) A20 powered BananaPI : cat /etc/issue Armbian 21.08.6 Buster \l I did an apt-get upgrade few days back, and since, it's CPU load is always around 1.00 even if I don't run my custom application (obviously, it wasn't the case before). Nothing special in /var/logs/messages, but top show me a kernel process always on top : top - 15:03:43 up 1 day, 17:40, 1 user, load average: 0,98, 1,03, 1,12 Tasks: 112 total, 2 running, 110 sleeping, 0 stopped, 0 zombie %Cpu(s): 4,5 us, 6,3 sy, 0,0 ni, 89,0 id, 0,0 wa, 0,0 hi, 0,2 si, 0,0 st MiB Mem : 993,9 total, 583,3 free, 119,7 used, 291,0 buff/cache MiB Swap: 497,0 total, 497,0 free, 0,0 used. 840,4 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 26647 root 20 0 0 0 0 R 10,3 0,0 1:44.71 kworker/1:0+events_freezable_power_ 32375 laurent 20 0 103912 51264 5472 S 10,3 5,0 1:20.54 Selene 3664 laurent 20 0 7588 2372 2016 R 0,7 0,2 0:01.23 top 14 root 20 0 0 0 0 I 0,3 0,0 2:55.81 rcu_sched 1927 root 20 0 1408 336 292 S 0,3 0,0 10:37.20 BananaLEDd 31455 root 20 0 0 0 0 I 0,3 0,0 0:00.26 kworker/1:3-mm_percpu_wq I did some googling but it doesn't help to find out what is this events_freezable_power_ In another hand, it doesn't explain this 1.00 load as you can see it consums only 10 to 15% of the CPU as well as my own custom application. So any tip, help, whatever is welcome. Thanks

Hi, a test as been done on Raspbian Bullseye and it's not working as well with the same error. Options suggested by xwiggen are already included It seems the problem is related to something upstream. Any idea ?

Hello, I did a try on my Gentoo PC using GCC 10.3.0 (same code, same Makefile, ...). Libraries are correctly included : ldd Selene linux-vdso.so.1 (0x00007ffdcb866000) liblua5.3.so.0 => /usr/lib64/liblua5.3.so.0 (0x00007fa2a4efc000) libm.so.6 => /lib64/libm.so.6 (0x00007fa2a4dc8000) libdrm.so.2 => /usr/lib64/libdrm.so.2 (0x00007fa2a4db3000) libcairo.so.2 => /usr/lib64/libcairo.so.2 (0x00007fa2a4c7a000) libfreetype.so.6 => /usr/lib64/libfreetype.so.6 (0x00007fa2a4bb1000) libSelene.so => /usr/local/lib/libSelene.so (0x00007fa2a4b94000) libpaho-mqtt3c.so.1 => /usr/local/lib64/libpaho-mqtt3c.so.1 (0x00007fa2a4ada000) libdl.so.2 => /lib64/libdl.so.2 (0x00007fa2a4ad4000) libpthread.so.0 => /lib64/libpthread.so.0 (0x00007fa2a4ab4000) libc.so.6 => /lib64/libc.so.6 (0x00007fa2a48fb000) /lib64/ld-linux-x86-64.so.2 (0x00007fa2a4f5b000) libpixman-1.so.0 => /usr/lib64/libpixman-1.so.0 (0x00007fa2a4857000) libfontconfig.so.1 => /usr/lib64/libfontconfig.so.1 (0x00007fa2a4810000) ... So the problem is with the way Bullseye is configured (Gcc or the linker or whatever). Note : GCC is compiled with the default options, no trick, no cheat, only optimized for my processor. Bye Laurent

I'll try, thanks.

Hello, Returning back to this issue, hoping software update will help. I was able to make the board working for about 2 full days (which is better than previously)... but it hangs afterward. I make it twice : the issue is still here I can see on the forum that other people was able to make BananaPRO working with Armbian but it's unfortunately not my case : potentially something bad with my board (it's one of the earliest one). So I'm giving up : I switched back to a BananaPI still under ArmBian and it's solid as a rock : Too bad for the wifi. I will try my bPro with other distrib (Gentoo ) to see if I can use this board for other thing. Thanks anyway Laurent

Hi, thanks for your reply. The only package missing in Buster is libpaho-mqtt-dev so I installed it from source. Anyway, I'll investigate about this issue as one day or another, my Gentoo boxes will move to GCC 10 (and in case someone will use my tool on recent distribs).

I'm suspecting a change/optimisation within GCC 10. I tried to use -rdynamic which is supposed to add all symbols, including not used ones ... but it doesn't change anything. I don't find a solution and it's blocking : I'm return back to Buster for my dashboard but I'm still investigating on other board.

Hello, I migrated my dashboard project from a BananaPro under "Buster" to a BananaPI under "Bullseye", image took from the website, verified and up to date as per Armbian-config. The source code of my project is exactly the same as well a the compilation procedure, and the used Makefile is : cc=cc opts=-Wall -fPIC $(shell pkg-config --cflags lua5.1 ) $(shell pkg-config --libs lua5.1 ) -DDRMC_WITH_FB -DUSE_DRMCAIRO $(shell pkg-config --cflags libdrm cairo freetype2 ) -DKMS_MISSING $(shell pkg-config --libs libdrm cairo freetype2 ) -DPLUGIN_DIR='"/home/laurent/Projets/Selene"' -L/home/laurent/Projets/Selene -L/home/laurent/Projets/Selene -lSelene -lpaho-mqtt3c $(shell pkg-config --cflags lua5.1 ) -lm -ldl -Wl,--export-dynamic -lpthread selene.o : selene.c SeleneLibrary/libSelene.h version.h Makefile $(cc) -c -o selene.o selene.c $(opts) ../Selene : selene.o Makefile $(cc) -o ../Selene selene.o $(opts) all: ../Selene Note : DRM, Cairo and Freetype libraries are not used in selene.c itself but in .so which are dynamically loaded (full source code can be found on https://github.com/destroyedlolo/Selene The problem : my projet is perfectly working on my Gentoo development box, as well on BananaPro/Buster but on the BananaPI/Bullseye, it's falling complaining about missing libraries : laurent@TdB:~/Projets/Selene$ ./Selene SelenitesDRMCairo/Gfx.sel ./Selene: symbol lookup error: /home/laurent/Projets/Selene/SelDRMCairo.so: undefined symbol: FT_Init_FreeType The root cause can be found with ldd On working boxes (here gentoo one) : laurent@torchwood ~ $ ldd /usr/local/bin/Selene liblua.so.5 => /usr/lib/liblua.so.5 (0xb6f23000) libm.so.6 => /lib/libm.so.6 (0xb6eb6000) libdrm.so.2 => /usr/lib/libdrm.so.2 (0xb6e96000) libcairo.so.2 => /usr/lib/libcairo.so.2 (0xb6dbe000) libfreetype.so.6 => /usr/lib/libfreetype.so.6 (0xb6d1a000) libkms.so.1 => /usr/lib/libkms.so.1 (0xb6d07000) libpthread.so.0 => /lib/libpthread.so.0 (0xb6b7d000) libSelene.so => /usr/local/lib/libSelene.so (0xb6abc000) libpaho-mqtt3c.so.1 => /usr/local/lib/libpaho-mqtt3c.so.1 (0xb6a1a000) libdl.so.2 => /lib/libdl.so.2 (0xb6a07000) libc.so.6 => /lib/libc.so.6 (0xb68bf000) /lib/ld-linux-armhf.so.3 (0xb6f7c000) ... whereas drm cairo freetype2 and dependancies are missing on the Bullseye box laurent@TdB:~/Projets/Selene$ ldd Selene linux-vdso.so.1 (0xbef14000) liblua5.1.so.0 => /lib/arm-linux-gnueabihf/liblua5.1.so.0 (0xb6f38000) libSelene.so => /home/laurent/Projets/Selene/libSelene.so (0xb6f16000) libpaho-mqtt3c.so.1 => /lib/arm-linux-gnueabihf/libpaho-mqtt3c.so.1 (0xb6e70000) libdl.so.2 => /lib/arm-linux-gnueabihf/libdl.so.2 (0xb6e5d000) libc.so.6 => /lib/arm-linux-gnueabihf/libc.so.6 (0xb6d5e000) libm.so.6 => /lib/arm-linux-gnueabihf/libm.so.6 (0xb6cf6000) /lib/ld-linux-armhf.so.3 (0xb6f74000) libpthread.so.0 => /lib/arm-linux-gnueabihf/libpthread.so.0 (0xb6cd0000) Does someone having any tips why libraries are not referenced under Bullseye ? Thanks Laurent

Hello, didn't worked on the subjet for a while as too busy by other task and as I duno if I have or not to modify uboot as well. Anyway, it's still crashing on 2 days basis maximum. Now, I gather additional information : I'm not sure it's really linked with Cairo or any graphical stuff. In case where the PI is not totally stuck, the CPU load become over 2 but my graphical dashboard is still working with network connectivity, BUT avahi is not working anymore. So I'm suspecting (without further information) a problem with Avahi. Bye Laurent

Hi, During the last crash, I got in the stdout of my application some error coming form the kernel (like "kernel message :" + some hexa code). So I'll try to rebuild the kernel with the setting I'm using on Gentoo. But it's not clear to me the relationship with uboot. 1/ are any information shared b/w uboot and the loaded kernel ? 2/ especially, if I'm doing any changes in the DTS, do they have to be applied to the uboot side ? 3/ do I have to rebuild and reinstall uboot as well ? Thanks Laurent

Hello, Unfortunately changing the governor did'nt helped. Bye Laurent

Hi, Only thing needed is an HDMI screen, a working MQTT broker (Mosquitto, ...) and I can wrote some fake mqtt provider if needed. All the other thing is totally open-source. I can also provide in compiled form. A Lemaker's BananaPI or M1 is working as well. I duno about the DTS, but the FEX was exactly the same b/w BananaPI and BananaPro with only a small setting for a (normally unused) UART and obviously the WiFi adapter. It the first time I"m using this board and Armbian : previously, it was a BananaPI under Gentoo. I switched 2 weeks back as I lost my ADSL connection and BananaPro WiFi was useful to bridge with my Phone. So I duno what should be "stable image" or even the stable DTS (potentially the bPI one but I didn't tested yet). I may try with my bPI's Gentoo image to see if it's working (which will be a proof the DTS are almost the same but the WiFi) and takes it's DTS. Now I'm not sure it evolved since I installed my previous system. Ok, I'll try. Bye, have a nice W.E.

I sent all the information I can as even Armbian-monitor is hanging (as I said). About your last message, as you know, it was night b/w and on the morning I was ... working So just got the notification. I thought it was part of the information I copy/pasted. Anyway, the image is the one from download page, Armbian Buster for BananaPRO, with some apt-get updates to keep the system up to date but it doesn't improve (or degrade) the situation. The provided one : 5.8.16, no change, no module added (like i2c, 1-wire, ...) My bPRO was displaying its dashboard as usual. The framework related code is here : https://github.com/destroyedlolo/Selene with DRMCairo plugin. It's not linked to anything special (message arrival or such) as they are coming all day long without trouble. The same code is running on my test machine, a BananaPI (M1) under Gentoo with the same plugin and exactly the same code (by the way, I didn't check about libraries versions). As I said previously, it has been compiled exactly on the same way, w/o any CPU optimization. Otherwise, nothing else is running : stock system without X, only with network enabled. I did apt-get on : libnss-mdns libcairo2-dev libncurses (but was not used when it crashed) liblua5.1-dev libdrm-dev libfreertype6-dev In addition, I compiled locally Patho-c from official sources and obviously my framework. Not tested. 2 scenarii : - or the board is totally struck after displaying the same message as the kernel.log, and in this case the solution is a power cycle - or system is still heathy and fully usable (my dashboard is even running), without any other messages or process missing but with a CPU load of 2.0 and without stuck processes. I also trying to trace in my own application if it's linked to something special but it's happening randomly whatever is the message that triggered a GUI update. It's not a kernel problem as it's the same I'm using on Gentoo. As I found on other discuss related to the same kind of A20 Oops, it may be (conditional !) a compilation issue where some libraries has been compiled with options that are not compatible with A20. The pb is to identify with one and I don't have enough clue and/or knowledge to identify witch one with the only trace I got. There is no accusation, only fact : - my box is not stable enough for the expected usage - at the time of writing, I didn't got any reply. Now, I can spend some time to figure out the root cause of this issue, it's not because I'll probably switch I'll give it up with this solution. Cheers aussi.