thank you for the share. Man I have tried this about a dozen times> simply can not get the Apex to work, I get error ls /dev/apex_0 ls: cannot access '/dev/apex_0': No such file or directory I don't even have the LXC container created yet. I deleted it since I am using Proxmox Helper-Scripts. I can easily add it again. That script uses Debian 11. All other steps have been completed to problem. What show I do?
Hi, can I ask if you can use “ls /dev/apex_0” in your Proxmox host shell? If there's an output, can you also do it in the LXC container? Its possible the driver didn't load, can you find the driver via “sudo lshw”? Also do you have secure boot enabled? If so, please disable it.
I forgot to add, that this can also happen if you did a downgrade to your kernel, earlier today I was tinkering with the Proxmox kernel and downgraded it to 6.5 and because of that the problem at step 6 occurred again. The fix was to repeat the process in the gasket driver. ruclips.net/video/HIvh_ZT2CFU/видео.html
@@JackofAllGates no i have no downgraded. This is whee I am root@c1u103-lab:~# dpkg -l | grep pve-kernel ii pve-firmware 3.14-1 all Binary firmware code for the pve-kernel
@@sikiriki The problem most likely is in your Proxmox host, the good news is that your system recognizes your TPU. I currently have the latest kernel 6.8.12-4-pve, I recommend upgrading your kernel and also build the gasket-dkms again, I think the issue mostly lies now on the gasket driver. Before you do build gasket-dkms, I recommend installing the pve headers: apt install pve-headers I encourage you to further read the troubleshooting of the driver in the proxmox forum, very informative: forum.proxmox.com/threads/update-error-with-coral-tpu-drivers.136888/ For example, In page 4 they tried using a different deb: apt remove gasket-dkms apt install dkms lsb-release sudo git dh-dkms devscripts pve-headers git clone github.com/KyleGospo/gasket-dkms cd gasket-dkms debuild -us -uc -tc -b cd .. dpkg -i gasket-dkms_1.0-18_all.deb
Hi, can I ask if you are able to do "ls /dev/apex_0" in your shell in your pve and if you see the coral tpu in lshw command? Also if you can send me pictures of any errors you encounter, it could narrow down the problem, thank you!
Just set up proxmox yesterday but cant get the coral working. I did all of the steps but the lspci -nn | grep 089a command does not return anything. I get an error installing gasket-dkms: Error! Bad return status for module build on kernel: 6.8.12-1-pve (x86_64) Consult /var/lib/dkms/gasket/1.0/build/make.log for more information. dpkg: error processing package gasket-dkms (--configure): installed gasket-dkms package post-installation script subprocess returned error exit status 10 Errors were encountered while processing: gasket-dkms E: Sub-process /usr/bin/dpkg returned an error code (1)
Hi, thank you for tuning in, let's trace a step back, reading the error it seems gasket-dkms is not installed, doing some research on the Proxmox forum I found a thread with someone having the same issue as you: forum.proxmox.com/threads/coral-tpu-not-working-after-upgrade-to-8-1-3.137282/ The solution leads to LiloBzH solution which we covered in 12:04, I recommend checking the logs if your logs are same as fcorrao (the user who started the thread): cat '/var/lib/dkms/gasket/1.0/build/make.log' Also to clarify when you did lspci -nn | grep 089a was this performed on shell or in your LXC container? Because if there is no output, my current guess is the driver has not loaded properly. Does the driver exist in "sudo lshw"?
@@JackofAllGates ok my bad i had removed the self built version of dkms and tried again with the normal one. There is no error from dkms now but still lspci -nn | grep 089a returns nothing, ls /dev/apex_0 says no such file, and i can not find the driver with lshw. I only see ethernet interfaces and the ssd. All of this is done in the shell
@@jere1023 Hi, if you cannot find the driver in lshw, it could mean 3 possible things: 1. It could be a bad seat, I suggest cleaning the slot with a brush and reseat the Coral TPU. 2. You might have Secure Boot enabled in the bios. I recommend disabling secure boot as this was a solution for some people. 3. Worst case scenario, MSI-X Incompatibility or a malfunctioned Coral TPU which is not our in our favor.
@@JackofAllGates Its a new computer and coral and it should be seated correctly but I will have to recheck. Secure boot is disabled. I have the A+E key coral which should work in the slot meant for a wifi card right? Looking through the bios i cant find an option to disable/enable wifi like some have and I also cant see the coral in the pcie configuration where i see nics and the ssd. Im not sure if it should show there but more reason to reseat
@@jere1023 Hello, Yes the A+E key is for the Wi-Fi card slot. After the reseat can you check if the driver is available if you use the command "lscpi -vvv" in your pve shell. Below is the output on mine. Can I also ask what is your motherboard and your linux kernel version? 02:00.0 System peripheral: Global Unichip Corp. Coral Edge TPU (prog-if ff) Subsystem: Global Unichip Corp. Coral Edge TPU Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx+ Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-
very helpful for me. thanks.
Im glad i could help!
I tried not installing the driver in the container, only the Proxmox host, and it still works for me.
thank you for the share. Man I have tried this about a dozen times> simply can not get the Apex to work, I get error
ls /dev/apex_0
ls: cannot access '/dev/apex_0': No such file or directory
I don't even have the LXC container created yet. I deleted it since I am using Proxmox Helper-Scripts. I can easily add it again. That script uses Debian 11.
All other steps have been completed to problem. What show I do?
Hi, can I ask if you can use “ls /dev/apex_0” in your Proxmox host shell? If there's an output, can you also do it in the LXC container?
Its possible the driver didn't load, can you find the driver via “sudo lshw”? Also do you have secure boot enabled? If so, please disable it.
I forgot to add, that this can also happen if you did a downgrade to your kernel, earlier today I was tinkering with the Proxmox kernel and downgraded it to 6.5 and because of that the problem at step 6 occurred again. The fix was to repeat the process in the gasket driver. ruclips.net/video/HIvh_ZT2CFU/видео.html
@@JackofAllGates Yes I have tried the command with reply
ls /dev/apex_0
ls: cannot access '/dev/apex_0': No such file or directory
sudo lshw returns
*-generic
description: Unassigned class
product: Coral Edge TPU
vendor: Global Unichip Corp.
physical id: 0
bus info: pci@0000:02:00.0
version: ff
width: 32 bits
clock: 66MHz
capabilities: bus_master vga_palette cap_list
configuration: driver=vfio-pci latency=255 maxlatency=255 mingnt=255
resources: irq:255 memory:a1100000-a1103fff memory:a1000000-a10fffff
@@JackofAllGates no i have no downgraded. This is whee I am
root@c1u103-lab:~# dpkg -l | grep pve-kernel
ii pve-firmware 3.14-1 all Binary firmware code for the pve-kernel
@@sikiriki The problem most likely is in your Proxmox host, the good news is that your system recognizes your TPU. I currently have the latest kernel 6.8.12-4-pve, I recommend upgrading your kernel and also build the gasket-dkms again, I think the issue mostly lies now on the gasket driver. Before you do build gasket-dkms, I recommend installing the pve headers: apt install pve-headers
I encourage you to further read the troubleshooting of the driver in the proxmox forum, very informative: forum.proxmox.com/threads/update-error-with-coral-tpu-drivers.136888/
For example, In page 4 they tried using a different deb:
apt remove gasket-dkms
apt install dkms lsb-release sudo git dh-dkms devscripts pve-headers
git clone github.com/KyleGospo/gasket-dkms
cd gasket-dkms
debuild -us -uc -tc -b
cd ..
dpkg -i gasket-dkms_1.0-18_all.deb
I made all steps and still don’t have apex drivers on pve.
Any ideas ?
Hi, can I ask if you are able to do "ls /dev/apex_0" in your shell in your pve and if you see the coral tpu in lshw command? Also if you can send me pictures of any errors you encounter, it could narrow down the problem, thank you!
🎉 Finally it’s woks thank you so much. Key points was /home
@@piotrtrabinski7983 Hello, I'm glad things worked out, enjoy! 🥳
Just set up proxmox yesterday but cant get the coral working. I did all of the steps but the lspci -nn | grep 089a command does not return anything.
I get an error installing gasket-dkms:
Error! Bad return status for module build on kernel: 6.8.12-1-pve (x86_64)
Consult /var/lib/dkms/gasket/1.0/build/make.log for more information.
dpkg: error processing package gasket-dkms (--configure):
installed gasket-dkms package post-installation script subprocess returned error exit status 10
Errors were encountered while processing:
gasket-dkms
E: Sub-process /usr/bin/dpkg returned an error code (1)
Hi, thank you for tuning in, let's trace a step back, reading the error it seems gasket-dkms is not installed, doing some research on the Proxmox forum I found a thread with someone having the same issue as you: forum.proxmox.com/threads/coral-tpu-not-working-after-upgrade-to-8-1-3.137282/
The solution leads to LiloBzH solution which we covered in 12:04, I recommend checking the logs if your logs are same as fcorrao (the user who started the thread):
cat '/var/lib/dkms/gasket/1.0/build/make.log'
Also to clarify when you did lspci -nn | grep 089a was this performed on shell or in your LXC container? Because if there is no output, my current guess is the driver has not loaded properly. Does the driver exist in "sudo lshw"?
@@JackofAllGates ok my bad i had removed the self built version of dkms and tried again with the normal one. There is no error from dkms now but still lspci -nn | grep 089a returns nothing, ls /dev/apex_0 says no such file, and i can not find the driver with lshw. I only see ethernet interfaces and the ssd. All of this is done in the shell
@@jere1023 Hi, if you cannot find the driver in lshw, it could mean 3 possible things:
1. It could be a bad seat, I suggest cleaning the slot with a brush and reseat the Coral TPU.
2. You might have Secure Boot enabled in the bios. I recommend disabling secure boot as this was a solution for some people.
3. Worst case scenario, MSI-X Incompatibility or a malfunctioned Coral TPU which is not our in our favor.
@@JackofAllGates Its a new computer and coral and it should be seated correctly but I will have to recheck. Secure boot is disabled. I have the A+E key coral which should work in the slot meant for a wifi card right? Looking through the bios i cant find an option to disable/enable wifi like some have and I also cant see the coral in the pcie configuration where i see nics and the ssd. Im not sure if it should show there but more reason to reseat
@@jere1023 Hello, Yes the A+E key is for the Wi-Fi card slot. After the reseat can you check if the driver is available if you use the command "lscpi -vvv" in your pve shell. Below is the output on mine. Can I also ask what is your motherboard and your linux kernel version?
02:00.0 System peripheral: Global Unichip Corp. Coral Edge TPU (prog-if ff)
Subsystem: Global Unichip Corp. Coral Edge TPU
Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx+
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-