google-coral / edgetpu

Coral issue tracker (and legacy Edge TPU API source)
https://coral.ai
Apache License 2.0
411 stars 124 forks source link

apex_0 not being created when installing drivers on proxmox #816

Closed JEngel98 closed 6 months ago

JEngel98 commented 6 months ago

Description

I am trying to pass a coral TPU through to frigate. The hardware is an A+E TPU installed in the wifi slot on a 6th gen intel processor. I am running frigate in a portainer in a container on proxmox.

I have run through the following guide 'https://github.com/Bytelake/Coral-in-LXC' and the getting started page 'https://coral.ai/docs/m2/get-started/#2a-on-linux' before spent the next few days following countless guides to get this working.

I have worked out that gasket and apex aren't installed properly as when I run the commands 'lsmod | grep gasket' or 'lsmod | grep gasket' I get nothing back, whereas my brother with an identical setup gets back the expected 40ish lines listing filename, author, etc... When I run 'ls /dev/apex_0' that returns 'ls: cannot access '/dev/apex_0': No such file or directory'

The only errors I get are promox complaining about packets being unsigned, but my brother has the same issue and it doesn't affect him his TPU has passed right through, he has a 7th gen processor.

Please help!

Issue Type

Build/Install

Operating System

Proxmox

Coral Device

M.2 Accelerator A+E

JEngel98 commented 6 months ago

Sorted, I had subscription proxmox repositories (removed) a 'buster' suite of debian with issues (removed) and secure boot enabled

google-coral-bot[bot] commented 6 months ago

Are you satisfied with the resolution of your issue? Yes No

figadore commented 5 months ago

Sorted, I had subscription proxmox repositories (removed) a 'buster' suite of debian with issues (removed) and secure boot enabled

Can you explain this a bit more? What did you do to solve it?

JEngel98 commented 5 months ago

So, the first thing I mentioned was the subscription repository I was using. This would throw up errors every time apt update and apt upgrade would run. if your get an 'E:' error, that will likely stop the kernal from updating. clear all your errors, 'W:' warnings seem not to matter.

For me the following was the cause of my 'E:; errors When you get proxmox it loads the subscription version of debian, I guess there might be a way of specifying otherwise, but by default it's subscription. Without a valid subscription, you're unable to change the kernel. So first thing is to enter the node in question, go to repositories under update, disable the subscription repo and add a non-subscription one that was there by default for me I believe? I also had a repo with the label of 'buster' under Suites tab, that needed disabling too as it was also causing issue when running the update and upgrade commands.

Once that was done, I ran all the code from these two links:

By this point running the following two commands generated a proper list which they handed before and I believe it is showing details on the kernal HAT? Apologies, some of this goes over my head from time to time.

If these two files are present, then in my case, secure boot was enabled and disabling that in the bios setting fitted it for me. I'd recommend googling 'device model number + disabling secure boot'

Hope this helps

jaxonstagecrew commented 5 months ago

So, the first thing I mentioned was the subscription repository I was using. This would throw up errors every time apt update and apt upgrade would run. if your get an 'E:' error, that will likely stop the kernal from updating. clear all your errors, 'W:' warnings seem not to matter.

For me the following was the cause of my 'E:; errors When you get proxmox it loads the subscription version of debian, I guess there might be a way of specifying otherwise, but by default it's subscription. Without a valid subscription, you're unable to change the kernel. So first thing is to enter the node in question, go to repositories under update, disable the subscription repo and add a non-subscription one that was there by default for me I believe? I also had a repo with the label of 'buster' under Suites tab, that needed disabling too as it was also causing issue when running the update and upgrade commands.

Once that was done, I ran all the code from these two links:

By this point running the following two commands generated a proper list which they handed before and I believe it is showing details on the kernal HAT? Apologies, some of this goes over my head from time to time.

  • sudo modinfo apex
  • sudo modinfo gasket

If these two files are present, then in my case, secure boot was enabled and disabling that in the bios setting fitted it for me. I'd recommend googling 'device model number + disabling secure boot'

Hope this helps

You just saved me hours of troubleshooting. Thanks!