dziobak Posted June 4, 2019 Posted June 4, 2019 Ubuntu Bionic, crash on linux-image-dev-sunxi64 5.88.190602, 5.88.190603 and 5.88.190604 (all kernel 5.1.6?) Last worked was 5.88.190601 (kernel 5.1.5) Debug log before crashing (full log in attachment): [ OK ] Reached target Network (Pre). Starting pNFS block layout mapping daemon... Starting Create Static Device Nodes in /dev... Starting Apply Kernel Variables... Mounting Kernel Configuration File System... [ OK ] Started pNFS block layout mapping daemon. [ OK ] Started Create Static Device Nodes in /dev. [ OK ] Started udev Coldplug all Devices. [ OK ] Started Apply Kernel Variables. [ OK ] Mounted Kernel Configuration File System. Starting udev Kernel Device Manager... [ OK ] Started Monitoring of LVM2 mirrors,�…sing dmeventd or progress pollin . [ OK ] Reached target Local File Systems (Pre). [ OK ] Started udev Kernel Device Manager. [ 10.523607] Unable to handle kernel paging request at virtual address ffff00c arm.log 1
dziobak Posted June 10, 2019 Author Posted June 10, 2019 Allwinner H6 Orange Pi 3 kernel 5.88.190609 (5.1.7). Booting OK, but there is stability problem: 4 crashes in 2 hours after change to new kernel. Quote root@OrangePi:~# date Mon Jun 10 13:49:33 CEST 2019 root@OrangePi:~# grep -a "Startup finished in" /var/log/syslog |grep kernel Jun 10 11:50:28 OrangePi systemd[1]: Startup finished in 11.725s (kernel) + 14.851s (userspace) = 26.576s. Jun 10 12:01:06 OrangePi systemd[1]: Startup finished in 11.447s (kernel) + 11.827s (userspace) = 23.275s. Jun 10 12:15:41 OrangePi systemd[1]: Startup finished in 11.280s (kernel) + 11.967s (userspace) = 23.248s. Jun 10 12:21:17 OrangePi systemd[1]: Startup finished in 12.093s (kernel) + 11.463s (userspace) = 23.556s. #some early message in logs when kernel booting root@OrangePi:~# grep -a "Current system time" /var/log/syslog Jun 10 11:49:43 OrangePi fake-hwclock[447]: Current system time: 2019-06-10 09:49:35 Jun 10 11:50:24 OrangePi fake-hwclock[424]: Current system time: 2019-06-10 09:50:15 Jun 10 12:01:04 OrangePi fake-hwclock[431]: Current system time: 2019-06-10 10:00:56 Jun 10 12:15:39 OrangePi fake-hwclock[428]: Current system time: 2019-06-10 10:15:30 Jun 10 12:21:16 OrangePi fake-hwclock[425]: Current system time: 2019-06-10 10:21:07
Igor Posted June 10, 2019 Posted June 10, 2019 We know its better but not alright. But it is impossible(simply too expensive) to communicate this rapidly changing information, especially not one2one or in private (yes, many people sends me this repeated known info on the email/private comm) on daily base to users. When things become stable under our standards image is labelled "supported" (https://docs.armbian.com/#what-is-supported) and then we are happy for users input ... while now there are far too many known troubles that we can handle. If you can contribute in finding the perfect values to make this board stable or help in current project problems, you are more then welcome. Edit: you need to make 3 more (valuable) contributions, before getting write access here. We had to make a limit to development parts of the forum because pointless and repeating questions from newbies are piling up.
dziobak Posted June 13, 2019 Author Posted June 13, 2019 Igor, I want help with testing correct values of voltage. Give me more info, please. Is there way to change these values online (in /proc or /sys keys)? Is there more info about voltage than https://github.com/armbian/build/pull/1407? And what purpose is with changing (lowering?) these values?
dziobak Posted June 28, 2019 Author Posted June 28, 2019 I did stability tests of voltage for CPU frequency from 1080 to 1800MHz. My metodology of searching for good, stable values was run StabilityTester (with default loop of 10), then add 10mV to values that not pass and repeat. This give me following results: CPU MHz 1080 1320 1488 1640 1800 mV 880 880 930 980 1050 I used above as target and min values for DTS/DTB, leaving max on defaults.
Igor Posted June 29, 2019 Posted June 29, 2019 14 hours ago, dziobak said: I did stability tests of voltage for CPU frequency from 1080 to 1800MHz. Prepare PR with new values and lets see if values are also generally O.K. Current are anyway not perfect.
dziobak Posted June 29, 2019 Author Posted June 29, 2019 I've tested latest above values by StabilityTester running 100 times, this gives few errors for 1488 and 1640 MHz. After increase voltage by 10 for both, StabilityTester run 100 times without error. So, finaly: MHz 1080 1320 1488 1640 1800 mV 880 880 940 990 1050 I've made PR to master with this. 1
Recommended Posts