[KERNEL]custom kernel build with treaks (2020.03.07 v6.8.3|5.5.8|4.19.108|NAVI|VEGA|NFSv4|R8125|Zen2)


Recommended Posts

Hi everyone.

 

Just for info - at last - I have tested the latest GIGABYTE Bios F50a for my x470 board - and (with ACS downstream enabled) I can use GPU Passthrough for all of my cards (two) without Error 127

 

It took over half a year (from F6 to F50a) - but they finally fixed the bug.

 

So I can now finally use Unraid without Kernel Patch AND a newer Bios version as F6...

 

Thanks again for the Kernel Patch - it did fulfill its purpose until now I don't need it anymore 🙂

 

Bye!

Link to comment
On 8/15/2019 at 2:42 AM, Pducharme said:

Just sad that Linux-Server.io doesn't incorporate this to the Unraid-NVIDIA plugins they create, that would make it so much easier.   I don't want to compile anything, and don't want to lose my Hardware Transcoding that is possible with the Unraid-NVIDIA build, so i'm stuck until LimeTech upgrade the Kernel on the regular Unraid I suppose.

I understand you don't want to compile yourself, but I don't particularly want to compile any more than the exisiting four/five builds I do with every release.  If these kernel modifications are needed then they should be pushed upstream to LimeTech.

  • Like 4
Link to comment
On 11/22/2019 at 7:25 AM, cobhc said:

Scored a really good deal on a 5700XT and had already updated to 6.8.0rc6. I need to downgrade to rc5 before using this kernel right?

Went ahead and reverted to rc5 with this kernel and everything is working great.

 

I initially had issues with restarting and shutting down VM's, much more than what I was aware of with the Navi reset bug. This fixed everything for me, so thanks!

Edited by cobhc
Link to comment

1.0.0.4B for the Gigabyte ax370 gaming has working passthrough. Still one of the best motherboards on the platform even after all these years. Now it ready for a 3000 series CPU. I went from F25 to F31 then flashed the ecw firmware tool using freedos then to F50a. All went smooth.

Passing through a rx 570 to my Hackintosh VMs.

Link to comment
  • 2 weeks later...

I was REALLY hoping that unRAID 6.8 would fix this issue. I have been running version 6.6.6 for almost a year as a daily driver with two windows VMs on a 2700x, with two 1070 ti, I have a ASUS Crosshair Hero VII (WiFi).
The latest BIOS that works without the error 127 bug is version 2008, I just upgraded to unRAID 6.8 without issues. I tried to update to the newest BIOS which is version 2901, ASUS website seems to state that it has AGESA version 1.0.0.3. This has failed GPU passthrough.

Going to either revert BIOS back to version 2008, or try to learn how to do this kernel hack.

EDIT:
I will just bios flashback and wait to see hopefully this kernel update for the stable unraid 5.8.
I don't want to learn how to roll-back unraid to a RC version.

Thanks for all the hard work!!!

Edited by tsawind
Link to comment

@tsawind I have an ASUS ROG STRIX X470-F GAMING. Until a week ago i was running it with a Ryzen 2700x for which i had had to flash back to bios Version 4207 from Dec 2018 as I was unable to passthrough any other way. 

 

Last week i replaced the ryzen for a new 3900x. so no option but to upgrade the bios to the latest; in my case  Version 5406 from nov 25 2019.

 

Rebooted while still with the ryzen 2700x and the bug was fixed. I was able to passthorugh without issues. Substituted with the new cpu (3900x) and same story. Pass through with no problem.

 

I am running unraid in legacy mode.

Edited by juan11perez
Link to comment
6 hours ago, juan11perez said:

@tsawind I have an ASUS ROG STRIX X470-F GAMING. Until a week ago i was running it with a Ryzen 2700x for which i had had to flash back to bios Version 4207 from Dec 2018 as I was unable to passthrough any other way. 

 

Last week i replaced the ryzen for a new 3900x. so no option but to upgrade the bios to the latest; in my case  Version 5406 from nov 25 2019.

 

Rebooted while still with the ryzen 2700x and the bug was fixed. I was able to passthorugh without issues. Substituted with the new cpu (3900x) and same story. Pass through with no problem.

 

I am running unraid in legacy mode.

Which version of unraid are you using? You arent using any special kernel?
I am unfamiliar with unraid legacy mode.

Link to comment

For all with AMD Navi and Error 127... I've compiled my first linux kernel with the updated NAVI patch (from 27-11-2019) for unraid 6.8.0 kernel version 4.19.88. I do not need other patches like pci-reset-quirk or HW support of the 5.x kernels.

 

It works for me (B450 Board with Ryzen 3600 + Sapphire 5700 XT Pulse) except the Navi audio. Some times I get the following error (after "Force Stop") and still need to reboot the rig:

qemu-system-x86_64: vfio: Cannot reset device 0000:09:00.1, no available reset mechanism.

But within Win10 ever reboot works like a charm... without the Navi patch I get always the D3 issue for every reboot.

 

Use at your own risk!!! My compile is based on the Unraid-DVB Kernel build script but without Highpoint/Rocketraid drivers (got errors there).

boot-4.19.88_navi_patch_20191214.zip

Link to comment
3 minutes ago, segator said:

Hello guys someone can give me instructions to build the kernel and package for unraid 6.8? I need to add some modules to kernel

Thanks!

It may sound good to do a custom build, but I strongly recommend you use the Unraid stock build and make a request for the missing parts (no guarantee they are added)

 

Custom builds are impossible to support. Next time you have an issue, nobody can help you.

Even more it will be extremely confusing. A custom build will not be mentioned in the error topic and people will be clueless about the differences.

Link to comment
4 hours ago, subivoodoo said:

For all with AMD Navi and Error 127... I've compiled my first linux kernel with the updated NAVI patch (from 27-11-2019) for unraid 6.8.0 kernel version 4.19.88. I do not need other patches like pci-reset-quirk or HW support of the 5.x kernels.

 

It works for me (B450 Board with Ryzen 3600 + Sapphire 5700 XT Pulse) except the Navi audio. Some times I get the following error (after "Force Stop") and still need to reboot the rig:


qemu-system-x86_64: vfio: Cannot reset device 0000:09:00.1, no available reset mechanism.

But within Win10 ever reboot works like a charm... without the Navi patch I get always the D3 issue for every reboot.

 

Use at your own risk!!! My compile is based on the Unraid-DVB Kernel build script but without Highpoint/Rocketraid drivers (got errors there).

boot-4.19.88_navi_patch_20191214.zip 16.34 MB · 2 downloads

 

Thank you for applying this patch. I was getting D3 on reboot of a Windows VM on a reference Asus 5700 XT but now it works!
It appears to work even on Force Stop but I didn't do enough testing to confirm this.

 

I also get the error for the audio. I think its because this patch is focused on the GPU part.

qemu-system-x86_64: vfio: Cannot reset device 0000:25:00.1, no available reset mechanism.


Am I right to assume that future Unraid updates will eventually overwrite the changes?

My system specs are:
MSI x570 Edge Wifi
3900x
Reference Asus 5700 XT

Edited by jonnyczi
Link to comment
19 hours ago, subivoodoo said:

For all with AMD Navi and Error 127... I've compiled my first linux kernel with the updated NAVI patch (from 27-11-2019) for unraid 6.8.0 kernel version 4.19.88. I do not need other patches like pci-reset-quirk or HW support of the 5.x kernels.

 

It works for me (B450 Board with Ryzen 3600 + Sapphire 5700 XT Pulse) except the Navi audio. Some times I get the following error (after "Force Stop") and still need to reboot the rig:


qemu-system-x86_64: vfio: Cannot reset device 0000:09:00.1, no available reset mechanism.

But within Win10 ever reboot works like a charm... without the Navi patch I get always the D3 issue for every reboot.

 

Use at your own risk!!! My compile is based on the Unraid-DVB Kernel build script but without Highpoint/Rocketraid drivers (got errors there).

boot-4.19.88_navi_patch_20191214.zip 16.34 MB · 5 downloads

Thanks for this, seems to be working fine for me. I'm curious if this fixes the issue I've been having with VM's locking up my whole server while passing through a Navi card (will need further testing). I haven't had a way to test the newer Navi patch as I'm not versed in compiling kernels!

Link to comment

So after reading into this, and researching more, it seems like if we just wait, for possible integration in unRAID 6.9, or just pick a unRAID version that has a linux kernel version of 5.4ish, this will fix the Error 127 bug.
This all seems to revolve around the newest generation AMD cpus, and the newer BIOS having conflicts with older linux kernel.

In my particular situation, I have an Asus Crosshair Hero VII (WiFi) x470 motherboard. I can't update past Version 2008 (2019/03/14) without getting the 127 bug, and according to this article: https://www.asus.com/News/EtaH71Hbjuio1arV I need at least BIOS version 2302 to use ryzen 3k cpus.
I believe that running some of the RC version of unRAID would fix this issue. I.E. unRAID version 6.8.0-RC1 has linux kernel 5.4

Link to comment
On 12/15/2019 at 11:51 AM, cobhc said:

Thanks for this, seems to be working fine for me. I'm curious if this fixes the issue I've been having with VM's locking up my whole server while passing through a Navi card (will need further testing). I haven't had a way to test the newer Navi patch as I'm not versed in compiling kernels!

Doesn't seem to fix this weird issue I'm having. I think I'm going to just convert my old machine back into an Unraid server, current machine back to a Windows PC and forget using VM's. Thanks again for the kernel though.

Link to comment
4 hours ago, cobhc said:

Doesn't seem to fix this weird issue I'm having. I think I'm going to just convert my old machine back into an Unraid server, current machine back to a Windows PC and forget using VM's. Thanks again for the kernel though.

Can you explain your problem? Is it locking up at boot, or randomly during use? With my 5700 XT, I would randomly crash unraid during gaming or even just using chrome in the win 10 vm. I made a ton of changes, but I finally got it stable. I settled on 6.8.0-rc5 with the kernel from the first thread. I added a second GPU in slot 3 I have set in my bios as the initial video device. With this setup I no longer have to pass a vbios to the vm. I also updated to Adrenalin 2020 drivers. With those changes, instead of locking up the entire host I would "only" lose signal where I was previously crashing. I could hear game audio continue, but the only recourse was the force stop the vm. The final fix was to DISABLE Radeon Anti-Lag and Radeon Enhanced Sync in Adrenalin. With that final change I am 100% stable and have no problem restarting my vm.

 

That's a long way of saying check if you have Radeon Anti-Lag and Radeon Enhanced Sync on and turn them OFF. They default to on, at least in Adrenalin 2020. The second gpu might not be necessary, it might have just changed the behavior from crashing all of unraid to only having for force stop the vm. Also, use q35-4.0.1 (or newer) if you want gen4 pcie speed without xml changes.

Edited by Skitals
  • Thanks 2
Link to comment
On 12/14/2019 at 2:03 PM, segator said:

Hey @bonienl you are completly right but first I want to ensure what I wants works then I will try to help admin's to add those modules.

What I'm trying to achieve is install kubernetes on unraid, but I have some problems with ipset and some  networking modules.

@segator , did you ever find decent instructions on a custom kernel for unraid? I'm in a similar kind of situation. I would like to add some modules to test something before making a better request.

Link to comment

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.