I noticed that the 2017 pro version is also has SD slot as 2015 version. Maybe the pro can works.
How to boot Ubuntu 16.04 on the Shield TV with X1 GPU drivers
|
It may boot, but DO NOT FLASH IT. You will very likely irreversibly brick it.
Booting the rootfs from SD or USB should be safe though. 05-25-2018, 05:42 PM
If I just flash it into the recovery partition, it maybe not brick it, I guess.
In the ubuntu we installed by just copying the rootfs file, I found that the nvidia driver was not installed, which means the shield
can not run some programs relying on GPUs. The nvidia_drivers.tbz2 is locates in the Linux_for_Tegra/nv_tegra folder. I think it's the nvidia GPU driver. When you run lsmod or lspci, you will see nothing. And in /dev/ you will not find the nvidia0 or nvidia1 node. If we can't use the GPUs, I think it not good for study. Maybe we can install the ubuntu by using Jetpack? The Jatpack will install the right drivers and some other things like opencv. But I failed install in the Jetpack way. So, how to get the nvidia driver run? 05-26-2018, 03:02 PM
I can't really say for sure. This documentation was written for the original shield TV. While I highly doubt there would be any issues between the original and the refresh, without the refresh in my hands and I get around to looking at this project again, I can't say for certain.
Honestly I'm more inclined to believe that you simply installed the rootfs incorrectly. Perhaps looking at the documentation on XDA in the links in my post would help. 05-26-2018, 04:15 PM
(05-26-2018, 03:02 PM)Helios Wrote: Honestly I'm more inclined to believe that you simply installed the rootfs incorrectly. Perhaps looking at the documentation on XDA in the links in my post would help. Well, I am using shield 2015 non-pro version(rebuy one) now and I perfectly follow the XDA steps as it says. I successfully install the ubuntu on partiotion 29 using mmcblk0p29.img, but seems with no nvidia driver. Have your shield 2015 installed the nvidia driver? I also use some commands: To prevent the nvidia drivers from being overwritten using: sudo apt-mark hold xserver-xorg-core And check the integrity of the NVIDIA files run sha1sum -c /etc/nv_tegra_release. It's shows all is ok. But when I run CUDA examples it says that can't find the nvidia GPU. 05-26-2018, 07:53 PM
Wait,
Are you running 'apt-get upgrade' or 'apt-get dist-upgrade' at all? That will definitely break things. Don't do that. Just specify specific packages you want upgraded, skipping any core system packages (like the kernel, anything xorg related) (05-26-2018, 07:53 PM)Helios Wrote: Are you running 'apt-get upgrade' or 'apt-get dist-upgrade' at all? I firstly run '[color=#000000]sudo apt-mark hold xserver-xorg-core[/color]', then I run 'apt-get update' and 'apt-get upgrade'. Because I think this command has already prevent the NVIDIA drivers from being overwritten. Do you mean I can use the 'update' command and can not use 'upgrade' command? In my situation, should I reinstall the linux and run again? And what is your output if you run 'lspci'? Does it shows nvidia driver? 05-28-2018, 03:05 PM
re-extract your rootfs.
I'm surprised upgrading didn't straight up cause the system to not boot. That's what it did to me. Again, doing a full system upgrade, even to packages within the 16.04 repos, will break things. This wouldn't break on a Jetson dev board, but this isn't a dev board. |
« Next Oldest | Next Newest »
|
Users browsing this thread: 1 Guest(s)