[Support] ich777 - AMD Vendor Reset, CoralTPU, hpsahba,...


Recommended Posts

1 hour ago, BiiRO said:

Thanks ich777 for all your work...similar to above post about iScsi going wonky..i updated to RC2 and iscsi dropped out..Looking at the tab it only shows 1-IQN (Create NEW IQN). No details or the other tabs showing the info from my already configured iscsi target. I assume I need new files for the RC2 version ? How do I get them to get this working again ?

Have you clicked on save on the main page?

If not the iSCSI changes are not saved and now lost, but the images are still on your machine, you have to recreate the configuration and click on save on the main page...

Hope this solves your problem.

 

EDIT: try first to create a IQN and then look at the main screen if something of your settings survived.

Link to comment
1 hour ago, ich777 said:

Have you clicked on save on the main page?

If not the iSCSI changes are not saved and now lost, but the images are still on your machine, you have to recreate the configuration and click on save on the main page...

Hope this solves your problem.

 

EDIT: try first to create a IQN and then look at the main screen if something of your settings survived.

The main page of Unraid-Kernel-Helper Plugin ? no save there..If i click on create IQN...the page constantly reloads in a loop and does nothing..the tabs are like 'flashing' (thats in Edge) Firefox just throws up this message...click resend and the same message pops up

Unraid.PNG

Edited by BiiRO
Link to comment
11 minutes ago, ich777 said:

What version of the Unraid-Kernel-Helper plugin have you installed?

 

May I ask why you don't use the Plugin that is available in the CA App instead of creating custom images?

 

EDIT: It should look something like this:

grafik.thumb.png.038868b437834763ddabe9e8df9b586f.png

Im using 2021.01.04a of Unraid-Kernel-Helper plugin....I only installed this at the start of December last year following what was in the thread.. think I was on RC30 then upgraded to 6.9RC2 and this happened. Should I just removed everything and start again ?

 

Actually just noticed something in the Docker log for the Unraid-Kernel-Helper...where is the mentioned 'old kernel folder' ?

-----------------------------------------------
----------------A L L D O N E----------------
---Please copy the generated files from the----
----output folder to your Unraid USB Stick-----
-----------------------------------------------
----MAKE SURE TO BACKUP YOUR OLD FILES FROM----
----YOUR UNRAID USB STICK IN CASE SOMETHING----
------WENT WRONG WITH THE KERNEL COMPILING-----
-----------------------------------------------
---Setting umask to 000---
---Checking for optional scripts---
---No optional script found, continuing---
---Starting...---
---Setting compile cores to 24---
-------------------------------------------------
---Found old Kernel output folder v5.10.1---
----Please delte this folder and restart the-----
-------------container to continue!--------------
-------------------------------------------------

Edited by BiiRO
Link to comment
10 minutes ago, BiiRO said:

Im using 2021.01.04a of Unraid-Kernel-Helper plugin....I only installed this at the start of December last year following what was in the thread.. think I was on RC30 then upgraded to 6.9RC2 and this happened. Should I just removed everything and start again ?

You can just use the default 6.9.0rc2 and download the iSCSI plugin from the CA App since beta35 intruduced a change so that you can install everything now with plugins. :)

This thread or Docker container is for people who want a AiO solution and integrate everything into the bzroot.

 

10 minutes ago, BiiRO said:

---Found old Kernel output folder v5.10.1---
----Please delte this folder and restart the-----
-------------container to continue!--------------

Yes, before you can compile new images you have to actually delete the "old" output folder from the kernel directory that lives in your appdata folder.

 

 

EDIT: I would strongly recommend you to use the Stock build of Unraid and install the iSCSI Plugin from the CA App.

Link to comment

Hi all, hoping for a sanity check on the kernel driver docker; I'm attempting to use it to build a custom kernel with the addition of vendor-reset and hopefully get rid of the dreaded vfio reset issue on my AMD GPU; I started into this after I found the treyf711 youtube video a couple of other posters had mentioned and then this kernel-building docker to find that the docker in its current state has a simple option to apply the vendor reset (good stuff), now I'm able to run the docker and get the files (leaving all default settings other than vendor_reset=true). Everything seems to work fine except the the reset issue persists.  Hopefully I didn't miss a step, but this seems to be about all that is needed.

 

Quick info/steps to get the seemingly relevant info out there:

  • Ryzen 3950x -> Asus X470 Crosshair VII
  • Powercolor AMD Vega 56
  • Unraid 3.9.0-rc2
  • As I was troubleshooting this last night, I noticed my GPU was particularly hot to the touch as I replaced with a nvidia gpu to further test; maybe the kernel module is not playing nice with it?
  1. Run plugin, all defaults aside from vendor_reset=true
  2. Patiently wait, I get the 'ALLDONE please copy..' a short wait later
  3. Use Unraid root terminal window to copy the newly created output bz* files to /boot; timestamps update, all seems good
  4. nano edit /boot/config/go and add in 'modprobe vendor-reset' line
  5. reboot
  6. run 'lsmod | grep vendor' as a verification; line shows up (perhaps I'm unfamiliar with the kernel bootup routine but is it expected that this also shows up for me when rebooted WITHOUT 'modprobe vendor-reset'?)
  7. Start VM with the GPU, boots as expected, shut it down
  8. Start up that same vm again, I see my passed thru keyboard and mouse flash colors indicating the startup is rolling, but my display will never load the graphical output. Need to force stop VM
  9. As generally known, the full bare metal restart will get the VM working (once), any subsequent VM boots will not have display out from the GPU

Thank you for time and consideration

Edited by dakota_654
Link to comment
17 hours ago, ich777 said:

You can just use the default 6.9.0rc2 and download the iSCSI plugin from the CA App since beta35 intruduced a change so that you can install everything now with plugins. :)

This thread or Docker container is for people who want a AiO solution and integrate everything into the bzroot.

 

Yes, before you can compile new images you have to actually delete the "old" output folder from the kernel directory that lives in your appdata folder.

 

 

EDIT: I would strongly recommend you to use the Stock build of Unraid and install the iSCSI Plugin from the CA App.

I just noticed you did a specific iscsi plugin as well..the one I was using was for the container.. thanks. So yes nothing shows up in the new plugin page..but the fileio scsi file is still there..can that be manually edited to show up ?

 

 

Edited by BiiRO
Link to comment
5 hours ago, dakota_654 said:

nano edit /boot/config/go and add in 'modprobe vendor-reset' line

You don't have to do that because it's integrated, it is designed to be as easy as possible.

 

5 hours ago, dakota_654 said:

Start up that same vm again, I see my passed thru keyboard and mouse flash colors indicating the startup is rolling, but my display will never load the graphical output. Need to force stop VM

Is this a Windows VM? I don't have any kind of AMD hardware but I will contact @giganode about that eventually he can help out.

 

5 hours ago, BiiRO said:

I just noticed you did a specific iscsi plugin as well..the one I was using was for the container.. thanks. So yes nothing shows up in the new plugin page..but the fileio scsi file is still there..can that be manually edited to show up ?

 

 

Have you already created a IQN?

Please post a screenshot of what is sowing up, don't know exactly what you mean.

Link to comment

I installed the new iGPU plugin, if I try to run the intel_gpu_top command I get this:

Failed to detect engines! (No such file or directory)
(Kernel 4.16 or newer is required for i915 PMU support.)

 

I removed modprobe i915 from my GO file and rebooted, it didn't seem to fix anything.

Running 6.9RC2 if that helps.

Link to comment
16 minutes ago, ich777 said:

You don't have to do that because it's integrated, it is designed to be as easy as possible.

 

Is this a Windows VM? I don't have any kind of AMD hardware but I will contact @giganode about that eventually he can help out.

 

Have you already created a IQN?

Please post a screenshot of what is sowing up, don't know exactly what you mean.

Yes it recognised the 'original' IQN but didnt see the original Volume (fileio-2TB) ...I created a new one to test (fileio-2TB-2) but I want to know if I can manually re-add the original (fileio-2TB) ?

iscsi.png

iscsi-2.png

Link to comment
1 hour ago, Dimtar said:

I installed the new iGPU plugin, if I try to run the intel_gpu_top command I get this:

Failed to detect engines! (No such file or directory)
(Kernel 4.16 or newer is required for i915 PMU support.)

 

I removed modprobe i915 from my GO file and rebooted, it didn't seem to fix anything.

Running 6.9RC2 if that helps.

The acceleration from the iGPU should still work fine (transcoding) or am I wrong, are you having IOMMU enabled?

 

Was the Intel-GPU-Tools container working or haven't you tried it yet?

 

EDIT: What iGPU are you using - processor model?

 

1 hour ago, BiiRO said:

but I want to know if I can manually re-add the original (fileio-2TB) ?

Yes, simply create the volume with the path to the "old" volume and it should just work fine.

(For sefety reasons I recommend you to back up the "old" image.

 

iSCSI should create only the mapping inside of targetcli and should not change your existing volume (but as said above create for safety reasons a backup of the "old" 2TB image)

Link to comment
10 hours ago, dakota_654 said:
  1. nano edit /boot/config/go and add in 'modprobe vendor-reset' line

Please remove that entry. The vendor-reset has to be loaded as soon as possible on host boot. If this already has happend and you reload it again with your entry in the go file this might lead into issues. Please reboot unraid after that and try again.

 

The vendor-reset is a wip. Maybe it just doesn't work right now. While your vega is a vega10, there is a reported issue for vega20.

The vendor-reset for example breaks the audio on my 5700xt. Had to switch back to the old navi patch..

 

List of supported GPUs:

|AMD|Polaris 10|

|AMD|Polaris 11|

|AMD|Polaris 12|

|AMD|Vega 10| Vega 56/64 |

|AMD|Vega 10| Vega 56/64/FE |

|AMD|Vega 20| Radeon VII |

|AMD|Navi 10| 5600XT, 5700, 5700XT

|AMD|Navi 12| Pro 5600M |

 

I keep my fingers crossed for you :) 

Keep us updated!

Edited by giganode
  • Thanks 1
Link to comment
14 hours ago, ich777 said:

You don't have to do that because it's integrated, it is designed to be as easy as possible.

 

Is this a Windows VM? I don't have any kind of AMD hardware but I will contact @giganode about that eventually he can help out.

 

Got that line out of there, thank you both for the insight; Didn't do the trick though.  I've had been using an Ubuntu VM mostly, seems like I've had the most luck getting things to boot (initially with my AMD card) and consistently with my Nvidia card

11 hours ago, giganode said:

Please remove that entry. The vendor-reset has to be loaded as soon as possible on host boot. If this already has happend and you reload it again with your entry in the go file this might lead into issues. Please reboot unraid after that and try again.

 

The vendor-reset is a wip. Maybe it just doesn't work right now. While your vega is a vega10, there is a reported issue for vega20.

The vendor-reset for example breaks the audio on my 5700xt. Had to switch back to the old navi patch..

 

List of supported GPUs:

|AMD|Polaris 10|

|AMD|Polaris 11|

|AMD|Polaris 12|

|AMD|Vega 10| Vega 56/64 |

|AMD|Vega 10| Vega 56/64/FE |

|AMD|Vega 20| Radeon VII |

|AMD|Navi 10| 5600XT, 5700, 5700XT

|AMD|Navi 12| Pro 5600M |

 

I keep my fingers crossed for you :) 

Keep us updated!

Critical thing I forgot to mention in my last post is that I receive this same behavior even with the vanilla Unraid kernels and no vendor-reset things in play at all, so I'm not 100% certain if the reset code is being triggered in my setup; is there any logs or otherwise diagnostic I can dig up that would indicate or be helpful with what is going on?

 

Link to comment
7 minutes ago, dakota_654 said:

 

Got that line out of there, thank you both for the insight; Didn't do the trick though.  I've had been using an Ubuntu VM mostly, seems like I've had the most luck getting things to boot (initially with my AMD card) and consistently with my Nvidia card

Critical thing I forgot to mention in my last post is that I receive this same behavior even with the vanilla Unraid kernels and no vendor-reset things in play at all, so I'm not 100% certain if the reset code is being triggered in my setup; is there any logs or otherwise diagnostic I can dig up that would indicate or be helpful with what is going on?

 

Currently I don't think that the vendor-reset isn't working properly and it will even break compatibility if Unraid with higher Kernel version than 5.11 is released for now and I haven't got time to look into it yet, look also at this issue here for the Vega20 cards. It would be best to open a new issue for your card.

The system log should display everything relevant.

Link to comment
20 minutes ago, dakota_654 said:

 

Got that line out of there, thank you both for the insight; Didn't do the trick though.  I've had been using an Ubuntu VM mostly, seems like I've had the most luck getting things to boot (initially with my AMD card) and consistently with my Nvidia card

Critical thing I forgot to mention in my last post is that I receive this same behavior even with the vanilla Unraid kernels and no vendor-reset things in play at all, so I'm not 100% certain if the reset code is being triggered in my setup; is there any logs or otherwise diagnostic I can dig up that would indicate or be helpful with what is going on?

 

Open terminal and try:

 

dmesg | grep VEGA

Link to comment
On 1/10/2021 at 7:28 PM, dakota_654 said:

Hi all, hoping for a sanity check on the kernel driver docker; I'm attempting to use it to build a custom kernel with the addition of vendor-reset and hopefully get rid of the dreaded vfio reset issue on my AMD GPU; I started into this after I found the treyf711 youtube video a couple of other posters had mentioned and then this kernel-building docker to find that the docker in its current state has a simple option to apply the vendor reset (good stuff), now I'm able to run the docker and get the files (leaving all default settings other than vendor_reset=true). Everything seems to work fine except the the reset issue persists.  Hopefully I didn't miss a step, but this seems to be about all that is needed.

 

Quick info/steps to get the seemingly relevant info out there:

  • Ryzen 3950x -> Asus X470 Crosshair VII
  • Powercolor AMD Vega 56
  • Unraid 3.9.0-rc2
  • As I was troubleshooting this last night, I noticed my GPU was particularly hot to the touch as I replaced with a nvidia gpu to further test; maybe the kernel module is not playing nice with it?
  1. Run plugin, all defaults aside from vendor_reset=true
  2. Patiently wait, I get the 'ALLDONE please copy..' a short wait later
  3. Use Unraid root terminal window to copy the newly created output bz* files to /boot; timestamps update, all seems good
  4. nano edit /boot/config/go and add in 'modprobe vendor-reset' line
  5. reboot
  6. run 'lsmod | grep vendor' as a verification; line shows up (perhaps I'm unfamiliar with the kernel bootup routine but is it expected that this also shows up for me when rebooted WITHOUT 'modprobe vendor-reset'?)
  7. Start VM with the GPU, boots as expected, shut it down
  8. Start up that same vm again, I see my passed thru keyboard and mouse flash colors indicating the startup is rolling, but my display will never load the graphical output. Need to force stop VM
  9. As generally known, the full bare metal restart will get the VM working (once), any subsequent VM boots will not have display out from the GPU

Thank you for time and consideration

If you boot your Unraid in UEFI mode then try disabling it. Goto your bios and select the non-UEFI usb stick to boot from. This fixed ALL vm related issues i had. It is also suggested by Spaceinvaderone.

  • Thanks 1
Link to comment
42 minutes ago, xPliZit_xs said:

If you boot your Unraid in UEFI mode then try disabling it. Goto your bios and select the non-UEFI usb stick to boot from. This fixed ALL vm related issues i had. It is also suggested by Spaceinvaderone.

 

Actually had been booting bios in legacy mode this whole time, but spurred me to try out UEFI boot, not much luck getting ANY graphics boot though; there is a line in the attached dmesg output that kept repeating while attempting this.

 

5 hours ago, giganode said:

Open terminal and try:

 

dmesg | grep VEGA

 

5 hours ago, ich777 said:

Currently I don't think that the vendor-reset isn't working properly and it will even break compatibility if Unraid with higher Kernel version than 5.11 is released for now and I haven't got time to look into it yet, look also at this issue here for the Vega20 cards. It would be best to open a new issue for your card.

The system log should display everything relevant.

 

So it would appear vendor-reset is operating as expected; I attached some output for that in an attached file if you have any final thoughts there, otherwise I'll take further action to the vendor-reset github, looks like the the kernel helper implementation works just fine so won't clutter this thread with any more on this tangent.

 

Thanks all for the replies and suggestions (+++ to ich777 for this and a number of the other great dockers you have thrown to the community)

powercolor_vega_56_vendor-reset_20210111.txt

  • Like 1
Link to comment
4 hours ago, dakota_654 said:

That looks similar to the reported outputs for navi10..

 

So you've got the same problem like me. The vendor-reset does not work for you.

But...there is a possible workaround for you:

 

https://forum.level1techs.com/t/vega-10-and-12-reset-application/145666

 

If you can not patch the kernel on your own, aks @ich777 for help. He has been very helpful for me for integrating the old navi patch.

Link to comment

The vendor-reset got an update for navi users. I tested it on my system.

I have no broken audio anymore after resets. For me this is a real breakthrough!

I don't need the old navi patch anymore now.

 

I can now boot between Windows 10 20H2, macOS Big Sur 11.1 and Ubuntu 20.10.

 

For all navi user who wanna try it out:

 

  1. Force update the docker
  2. Edit the docker and add a variable like this:image.thumb.png.39429a29f1c092625e8a643ca9823ebd.png
  3. Try and hopefully enjoy!

Keep in mind, this only fixes the specific audio issue for navi users.

 

Please report your expierences here. :) 

 

Special Thanks to @ich777 for that fast edit.

 

 

Edited by giganode
  • Like 2
Link to comment
2 hours ago, giganode said:

The vendor-reset got an update for navi users. I tested it on my system.

I have no broken audio anymore after resets. For me this is a real breakthrough!

I don't need the old navi patch anymore now.

 

I can now boot between Windows 10 20H2, macOS Big Sur 11.1 and Ubuntu 20.10.

 

For all navi user who wanna try it out:

 

  1. Force update the docker
  2. Edit the docker and add a variable like this:image.thumb.png.39429a29f1c092625e8a643ca9823ebd.png
  3. Try and hopefully enjoy!

Keep in mind, this only fixes the specific audio issue for navi users.

 

Please report your expierences here. :) 

 

Special Thanks to @ich777 for that fast edit.

 

 

 

I have a 5600XT. This is my first time trying the gnif/vender-reset (in the past I used the navi reset patch), but when I tried it Unraid failed to startup, it just stopped at this point. I am pretty sure the compile completed correctly, got the usual all done message. More than willing to do more testing/troubleshooting.

 

20210113_145700.thumb.jpg.df9d31383f3b67e1108d8e8ac6dad3e3.jpg

Edited by ndetar
Link to comment
6 hours ago, ndetar said:

I have a 5600XT. This is my first time trying the gnif/vender-reset (in the past I used the navi reset patch), but when I tried it Unraid failed to startup, it just stopped at this point. I am pretty sure the compile completed correctly, got the usual all done message. More than willing to do more testing/troubleshooting.

Have you something appended to your syslinux.cfg?

Eventually @giganode can help you out.

 

What you can try in the meantime is delete the container entirely and the output folder that lives in the kernel directory in your appdata folder and redownload the Unraid-Kernel-Helper from the CA App the new Variable that @giganode mentioned should be already in there.

 

Then select to build it with the gnif/vendor-reset patch and also click on 'Show more settings...' and at gnif/vendor-reset Branch enter 'feature/audio_reset' and also set to option Save to Log to 'true'.

 

If that all doesn't work please post the Log file that it creates in the Kernel directory.

Link to comment
1 hour ago, ich777 said:

Have you something appended to your syslinux.cfg?

Eventually @giganode can help you out.

I can share my settings, but I feel like something went wrong with the build.. But I don't know for a fact :) 

 

kernel /bzimage
append video=vesafb:off,efifb:off vfio_iommu_type1.allow_unsafe_interrupts=1 isolcpus=4-15,20-31 pcie_acs_override=downstream,multifunction initrd=/bzroot

Link to comment
30 minutes ago, Spucoly said:

Is there a guide on how to install the custom image?

Have you read the first post:

 

Quote

This is how the build of the Images is working (simplyfied):

  1. The build process begins as soon as the docker starts (you will see the docker image is stopped when the process is finished)
    Please be sure to set the build options that you need.
  2. Use the logs or better open up a Console window and type: 'docker attach Unraid-Kernel-Helper' (without quotes) to also see the log (can be verry laggy in the browser depending on how many components you choose).
    The whole process status is outlined by watching the logs (the button on the right of the docker).
  3. The image is built into /mnt/cache/appdata/kernel/output-VERSION by default. You need to copy the output files to /boot on your USB key manually and you also need to delete it or move it for any subsequent builds.
  4. There is a backup copied to /mnt/cache/appdata/kernel/backup-version. Copy that to another drive external to your Unraid Server, that way you can easily copy it straight onto the Unraid USB if something goes wrong.

THIS CONTAINER WILL NOT CHANGE ANYTHING TO YOUR EXISTING INSTALLATION OR ON YOUR USB KEY/DRIVE, YOU HAVE TO MANUALLY PUT THE CREATED FILES IN THE OUTPUT FOLDER TO YOUR USB KEY/DRIVE AND REBOOT YOUR SERVER.

 

PLEASE BACKUP YOUR EXISTING USB DRIVE FILES TO YOUR LOCAL COMPUTER IN CASE SOMETHING GOES WRONG!
I AM NOT RESPONSIBLE IF YOU BREAK YOUR SERVER OR SOMETHING OTHER WITH THIS CONTAINER, THIS CONTAINER IS THERE TO HELP YOU EASILY BUILD A NEW IMAGE AND UNDERSTAND HOW THIS IS WORKING.

 

Step 3 is where it says what to do, you could also copy the files that are located in '/mnt/cache/appdata/kernel/output-VERSION' to your computer, shutdown the server, pull the USB boot device out of the server, put the files that you copied over to your computer in the root of your USB boot device and replace all bz* files (please be sure to also backup all bz* files that are on the USB boot device to your local computer in case something goes wrong).

 

Another way is to use WinSCP if you are on Windows, connect to your server with the IP 'root' and your password and copy the files from '/mnt/cache/appdata/kernel/output-VERSION' to your '/boot' folder.

 

This should be it.

(Please keep in mind that if you shared the appdata folder via samba you can go into the '.../appdata/kernel/output-VERSION' and grab the files there)

Link to comment
  • ich777 changed the title to [Support] ich777 - AMD Vendor Reset, CoralTPU, hpsahba,...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.