CaptManiac Posted September 24, 2017 Posted September 24, 2017 I got the Le Potato Ubuntu server image up and running. Then installed the Mate desktop environment. Most everything is working fine, but the system freezes hard at times, mostly when I try to run several large programs concurrently. Problem with memory? Also, some resolutions have the refresh rate set at 68.5 Hz that don't work on most of my monitors. Anyway, just wanted to say thanks for the work you guys put into it, and I look forward to the continuing progress!
TonyMac32 Posted September 24, 2017 Posted September 24, 2017 I have not tried Mate, I built an image with the standard xfce desktop, I'll give it a run and see if I can make it hang. Do you have any logs or debug output from the serial port?
CaptManiac Posted September 24, 2017 Author Posted September 24, 2017 I do not have those logs. How do I get them? I don't think it's a Mate problem as it happens even when it's running headless. So here's an easy way I can get it to hang. I'm rsyncing a 5 GB directory from another computer to the Le Potato. After maybe a few hundred MBs of transfer I'll lose both an SSH connection and any keyboard/mouse response. I can still see the time/seconds ticking away in the desktop bar, so it's not completely dead.
TonyMac32 Posted September 25, 2017 Posted September 25, 2017 I replicated the failure pretty easy once you gave me what to look for, I decided to try and transfer a 5 GB file over the network, I locked it up about 800 MB in. I've got to call it a night soon, so far I haven't found anything jumping up and down screaming in the logs.
Tido Posted September 25, 2017 Posted September 25, 2017 11 hours ago, CaptManiac said: I'll lose both an SSH connection and any keyboard/mouse response Well, the mouse shouldn't be an issue on headless. Once you lost SSH connection, does it still deliver data to the device? So you would have to pack 5GB on the SDcard and copy it the otherway around to test that.
CaptManiac Posted September 26, 2017 Author Posted September 26, 2017 OK, I'll test that. I'm also getting errors when I do a fairly large-ish compilation. Attached is a dmesg showing clang bringing the system down. dmesg_20170925.txt
TonyMac32 Posted September 26, 2017 Posted September 26, 2017 Do you have the means of measuring the temperature on the board?
CaptManiac Posted September 26, 2017 Author Posted September 26, 2017 No, the board does not appear to have a temperature sensor.
TonyMac32 Posted September 26, 2017 Posted September 26, 2017 Well, that was my reason for asking, because it does, but Mainline does not appear to have the drivers. That's a bit surprising to me, you would think thermals on a processor would be among the "important things to do" list. I am hoping to get a look at the thermal situation and see if the drivers can be forward ported easily enough.
Neil Armstrong Posted September 27, 2017 Posted September 27, 2017 Hi, Mainline Linux has support for the SCPI Temperature Sensor, please check in /sys/class/hwmon/hwmon0/temp1_input For the hard crash, looking at the logs seems to indicate a memory issue, we will investigate further ! Thanks, Neil 2
TonyMac32 Posted September 27, 2017 Posted September 27, 2017 Thank you Neil, that is set up considerably differently than the Amlogic kernels, I missed it.
Igor Posted September 28, 2017 Posted September 28, 2017 Any better way to boot SD with attached eMMC than nuking its boot part?
TonyMac32 Posted September 28, 2017 Posted September 28, 2017 Honestly I pulled the eMMC and put it in a safe place, if the bootloader on eMMC is the same revision level as the one I'm working with you should be able to follow Balbes150's instructions, but tell it to load from ext4 instead of FAT. I am willing to risk "nuking" mine for science. It would mean the eMMC is the boot device for u-boot but it boots the SD when available.
V10lator Posted October 7, 2017 Posted October 7, 2017 On 26.9.2017 at 6:49 AM, CaptManiac said: getting errors when I do a fairly large-ish compilation. Attached is a dmesg showing clang bringing the system down. dmesg_20170925.txt Same here (make -j4 with gcc/g++) but it doesn't always bring the system down. On 27.9.2017 at 8:18 AM, Neil Armstrong said: please check in /sys/class/hwmon/hwmon0/temp1_input It showed 45000 at the time the log messages happened. A pretty good way to reproduce seems to be doing "make -j4" on http://distfiles.macports.org/openfst/openfst-1.3.4.tar.gz - you'll either see internal "compiler error: Bus error" or "g++: internal compiler error: Segmentation fault (program cc1plus)" while dmesg shows the messages from @CaptManiac. I observed this on the 2GB Kit Special (first batch of production-ready batch of 2GB boards, push pin heatsink, custom active cooling case with fan, HDMI cable, power supply, and 32GB UHS MicroSD) in headless mode.
TonyMac32 Posted October 12, 2017 Posted October 12, 2017 A patch I found on kernel.org coming out for the 4.14 [edit] 4.15 kernel, missing the LTS [/edit] is addressing an issue where Amlogic changed the reserved memory mapping for the Arm trusted firmware. Unfortunately this wasn't reflected in the kernel: http://lists.infradead.org/pipermail/linux-amlogic/2017-October/004860.html This addresses the "External Abort" error, and in my testing so far appears to have cleared it up. I had been able to recreate this failure 100% by moving a 2 GB file from my file server to the SBC, it would lock up around 1 GB like clockwork. I've moved several movie files to and from network storage, while playing a file, no problems. This has seemingly fixed my kernel crashes on the next 4.13 kernel, I have added the patch to the build system. I will add it to dev as well, for anyone who builds their own. Thank you @Neil Armstrong
Ian Murray Posted October 12, 2017 Posted October 12, 2017 My board showed up today, and I'm having a little weirdness with it. I get a brief flash of green on boot, then just a black screen. If I disconnect & reconnect the HDMI once I see the network lights go on, I get a display, and everything works, but it's squashed into the top half of the screen. It looks like every other line of the display is missing (text is hard to read with horizontal lines missing). I tried the official Ubuntu image, and the display is fine on that one, but it won't take input from either of the keyboards that I have available, so it's not super useful right now. I just tried again with a monitor instead of a TV, and it actually said no input and went into standby during boot. Replugging the HDMI was the same result, working but squashed.
LarryM Posted October 12, 2017 Posted October 12, 2017 I have a similar problem, but discovered this: a) if the hdmi port is connected to the monitor, there will be no display when booting completes. b if the hdmi port is disconnected on *restart* but plugged in when the blue light starts to flash, the monitor will be squashed to only the top half of the screen c) if the hdmi port is disconnected and power is shutdown, then booted, and when the blue light flashes then plug the monitor in, then display will be normal and everything is fine. I assume there's a timing problem someplace, but I'm not savvy enough for that, some help ??
TonyMac32 Posted October 12, 2017 Posted October 12, 2017 Yes, these are still experimental builds, there are some scripts that need run at boot on the legacy kernel, however, they do not seem to be having the desired effect, I need to go through and more properly understand them. If you go to a terminal, login as root and execute /usr/lib/armbian/c2_init.sh it will fix the screen. This is next on my list since mainline crashes (that we know of) are fixed. Presently I get proper behavior with my Samsung HDMI monitor, but not with my HDMI/DVI adapter on an Acer. The issue is more serious than I thought if none of your monitors are working properly. Thank you for the feedback/info.
LarryM Posted October 12, 2017 Posted October 12, 2017 I have another (sort of) boot issue - I have a Windows PC, a Ubuntu PC, a security system, and now LePotato that I'm trying to share the keyboard/mouse/video thru an IoGear KVM switch. LePotato seems to have a boot issue with the keyboard and mouse coming in on 1 USB connection. If I plug in independent mouse and keyboard and go thru the boot, then I can unplug them and plug in the combined from the switch its fine. However, if I boot with the combined plugged in, then it seems to have neither mouse nor keyboard. Not the end of the world, but if there's any thing you can think of, or at least put it on someone's list to look at, that'd be swell. Oh, one other thing - is there a way to turn audio off going out the HDMI and make it come out the audio plug? thanks, Larry
TonyMac32 Posted October 12, 2017 Posted October 12, 2017 Can you SSH into the board after booting it attached to your switch? Or use the UART. Then post the link provided by issuing the command armbianmonitor -u As far as audio goes, I have not tried anything yet, with the other issues. Thank you for your observations on HDMI behavior.
LarryM Posted October 12, 2017 Posted October 12, 2017 I'm not sure what you're asking, I'm assuming you want me to reboot with the KVM plugged into LePotato, then ssh into it from another PC, and run that command? I didn't get the part about "or use the uart" FYI, if I run it now (with the KVM plugged in and working) I get /var/log/armhwinfo.log had been uploaded to http://sprunge.us/fjQM I assume there's no cut-n-paste in this terminal emulator?
LarryM Posted October 12, 2017 Posted October 12, 2017 Oh, and kinda cool - conky works. There doesn't seem to be a conky-manager, but cool to have this handy tool.
Bret Barnes Posted October 13, 2017 Posted October 13, 2017 /usr/lib/armbian/c2_init.sh works on my DVI monitor. I still have to plug/unplug even if I run it first. Shows up squished on the 1280x1024 screen obviously.
TonyMac32 Posted October 13, 2017 Posted October 13, 2017 Yes, that script runs at boot, but for some reason the fb is not taking the hint if the monitor is present during boot. The fact it is fine if you don't have the monitor plugged in and then do is a hint, but one I'll have to explore tomorrow. Tomorrow there should be a mainline desktop build available, there are no screen issues with it.
balbes150 Posted October 13, 2017 Posted October 13, 2017 1 hour ago, TonyMac32 said: Yes, that script runs at boot, but for some reason the fb is not taking the hint if the monitor is present during boot. The fact it is fine if you don't have the monitor plugged in and then do is a hint, but one I'll have to explore tomorrow. Tomorrow there should be a mainline desktop build available, there are no screen issues with it. Hi Tony. You activate the monitor runs once (when the system is already running) ?
TonyMac32 Posted October 13, 2017 Posted October 13, 2017 I am currently just reusing the script that is used on the Odroid-C2 with Armbian, https://github.com/armbian/build/tree/master/packages/bsp/odroid https://github.com/armbian/build/blob/cdb4161ada4f0de2e92cb43748c5c32d0548f893/config/sources/meson64.conf#L59 It is run once during system init.
balbes150 Posted October 13, 2017 Posted October 13, 2017 46 minutes ago, TonyMac32 said: It is run once during system init. If you use a double run of the script (e.g., add one in the initrd), in the second run, all variables will be set, and the script will correctly include the monitor. Maybe at the first start (with some monitors) not all variables ready and the result is not as clear. I use two scripts with different parameters (first with 720 and two start with 1080), then the monitor is normally activated.
TonyMac32 Posted October 13, 2017 Posted October 13, 2017 29 minutes ago, balbes150 said: If you use a double run of the script I will try this out then, thank you for the input.
rinukkusu Posted October 13, 2017 Posted October 13, 2017 (edited) Hi, piggybacking on this thread. I downloaded Armbian_5.33.171013_Lepotato_Ubuntu_xenial_next_4.13.5.img and flashed it onto the SD card that came with the board. I'm not seeing any output on the HDMI port and also can't ssh into it - it's just giving me a timeout. EDIT: it works now and I can ssh into it!! Edited October 13, 2017 by rinukkusu
CaptManiac Posted October 13, 2017 Author Posted October 13, 2017 Tony, you did it! Today's apt-get update makes the system SO much more stable. I was able to do an entire 5 GB rsync transfer between machines, and complete a large compile with all four CPUs pegged at 100% for 20 minutes. I really appreciate your hard work.
Recommended Posts