[Plugin] Intel-GVT-g

[hjaltioj]

51 minutes ago, ich777 said:

What mode are you using for the VM?

Does only the VM crash or the entire systen?

VM is on i915-GVTg_V5_4.

BIOS is OVMF

Machine i440fx-5.1

 

The VM crashes and it crashes the vm manager. i can ssh into my linux vm's but cant go under vm tab i unraid. and if i type "/etc/rc.d/rc.libvirt stop" in terminal on unraid it fails. if i try restart unraid server it fails because it cant stop libvirt service. only way i can restart or shutdown the unraid server is by holding shutdown button on server for 10seconds. 

[ich777]

43 minutes ago, hjaltioj said:

VM is on i915-GVTg_V5_4.

Is anything running in the VM in the background that uses the iGPU?

 

44 minutes ago, hjaltioj said:

The VM crashes and it crashes the vm manager

This is a common issue when more VRAM is used than allocated in the GVT-g mode because it tries to allocate memory where no memory is available.

[hjaltioj]

23 minutes ago, ich777 said:

Is anything running in the VM in the background that uses the iGPU?

 

This is a common issue when more VRAM is used than allocated in the GVT-g mode because it tries to allocate memory where no memory is available.

No not. only use the Firefox and microsoft office on that machine.
It just happened again and here is the log from unraid:

 

Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e00
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e10
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e20
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e30
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e40
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e50
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e60
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e70
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e80
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e90
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4ea0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4eb0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4ec0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4ed0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4ee0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4ef0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f00
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f10
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f20
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f30
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f40
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f50
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f60
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f70
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f80
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4f90
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4fa0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4fb0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4fc0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4fd0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4fe0
Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4ff0
Nov 24 13:56:03 unRAID kernel: gvt: vgpu 1: fail: shadow page 0000000000000000 guest entry 0xf0b82a75e5ff94f3 type 9
Nov 24 13:56:03 unRAID kernel: gvt: vgpu 1: fail: spt 00000000510a9478 guest entry 0xf0b82a75e5ff94f3 type 9
Nov 24 13:56:03 unRAID kernel: gvt: vgpu 1: fail: shadow page 00000000510a9478 guest entry 0xf0b82a75e5ff94f3 type 9.
Nov 24 13:56:03 unRAID kernel: gvt: vgpu 1: fail to flush post shadow
Nov 24 13:56:03 unRAID kernel: gvt: vgpu 1: fail to dispatch workload, skip
 

[ich777]

7 minutes ago, hjaltioj said:

No not. only use the Firefox and microsoft office on that machine.

7 minutes ago, hjaltioj said:

Nov 24 13:55:58 unRAID kernel: gvt: guest page write error, gpa 852d4e00

That is such a message when the VM try to allocate memory when no VRAM is available.

 

On Github there are many issues open with similar errors but no real solution to this.

The main issue is that are only 256/512MB available and this is by today standards really not that much.

[hjaltioj]

2 minutes ago, ich777 said:

That is such a message when the VM try to allocate memory when no VRAM is available.

 

On Github there are many issues open with similar errors but no real solution to this.

The main issue is that are only 256/512MB available and this is by today standards really not that much.

Okey, then there is nothing i can do?

Do you know why i only can select:

i915-GVTg_V5_4
i915-GVTg_V5_8

and not i915-GVTg_V5_1 ?
:)

[ich777]

6 minutes ago, hjaltioj said:

Okey, then there is nothing i can do?

Sadly enough no...

 

8 minutes ago, hjaltioj said:

Do you know why i only can select

This is because most motherboard manufacturers haven't implemented the aperture size or better speaking they haven't enabled so that a user can change the aperture size for the iGPU.

 

You can go to your BIOS and see what options you got for the iGPU.

[hjaltioj]

12 minutes ago, ich777 said:

Sadly enough no...

 

This is because most motherboard manufacturers haven't implemented the aperture size or better speaking they haven't enabled so that a user can change the aperture size for the iGPU.

 

You can go to your BIOS and see what options you got for the iGPU.

i can set memory size to 1024mb and then there is a setting called "Above 4G Decoding" ? dont know what that is?

[ich777]

1 hour ago, hjaltioj said:

"Above 4G Decoding"

That's manly for PCIe cards who need larger address space.

 

The think you want to change is aperture size, if you can't change it or if it's not there you can't get the other modes to work.

[mao]

When will I update the 11th generation Intel AGPU. After I upgraded the CPU, I found that it can't work anymore, which is very distressing. What is the specific reason why the 11th generation CPU is not supported, please tell me if there is still a chance to continue to use it on the 11th generation? Otherwise I will replace 9900k

[ich777]

17 minutes ago, mao said:

When will I update the 11th generation Intel AGPU. After I upgraded the CPU, I found that it can't work anymore, which is very distressing. What is the specific reason why the 11th generation CPU is not supported, please tell me if there is still a chance to continue to use it on the 11th generation? Otherwise I will replace 9900k

Please see this post:

 

[mao]

On 2021/11/25 at AM2点09分, ich777 said:

请看这个帖子：

Is there any relevant support provided? SR-IOV is a bit difficult for me

 

[ich777]

5 minutes ago, mao said:

Is there any relevant support provided? SR-IOV is a bit difficult for me

SR-IOV is currently not supported by KVM/QEMU so at the time of writing it is not possible to get it working on unRAID.

[kkkk]

I'm in trouble.

I noticed  that:
when I run Synology emby transcoding alone, there is 60fps , the power draw can reach 5W

 

when VM win10 is running too, the transcoding will fall to less than 20fps, the power draw is limited to 1W each

 

l also try to run vm win10 alone, the power draw is also  limited to 1W。

 

[ich777]

17 hours ago, kkkk said:

when VM win10 is running too, the transcoding will fall to less than 20fps, the power draw is limited to 1W each

Can you please share your Diagnostics, I've tried that now on my server and it works flawlessly.

[qy2009]

@ich777

I switched to a new cpu- coffeelake, and when I tried to use intel gvt and adding the grahpic card inside VM, I cannot update the VM settings, it said "XML error: 'model' attribute in <hostdev> is only supported when type='pci'". what does this mean? 

 

Here is some log I saw:

Dec  4 16:57:36 Tower root: ------------Intel GVT-g found locally, please wait...!-----------
Dec  4 16:57:38 Tower root: 
Dec  4 16:57:38 Tower root: -----------------Loading Kernel Module 'i915'------------------
Dec  4 16:57:38 Tower kernel: Linux agpgart interface v0.103
Dec  4 16:57:38 Tower kernel: Setting dangerous option force_probe - tainting kernel
Dec  4 16:57:38 Tower kernel: i915 0000:00:02.0: [drm] Found 128MB of eDRAM
Dec  4 16:57:38 Tower kernel: i915 0000:00:02.0: [drm] VT-d active for gfx access
Dec  4 16:57:38 Tower kernel: i915 0000:00:02.0: vgaarb: deactivate vga console
Dec  4 16:57:38 Tower kernel: Console: switching to colour dummy device 80x25
Dec  4 16:57:38 Tower kernel: i915 0000:00:02.0: Direct firmware load for i915/gvt/vid_0x8086_did_0x3ea5_rid_0x01.golden_hw_state failed with error -2
Dec  4 16:57:38 Tower kernel: i915 0000:00:02.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=io+mem:owns=io+mem
Dec  4 16:57:38 Tower kernel: i915 0000:00:02.0: [drm] Finished loading DMC firmware i915/kbl_dmc_ver1_04.bin (v1.4)
Dec  4 16:57:38 Tower kernel: i915 0000:00:02.0: [drm] failed to retrieve link info, disabling eDP
Dec  4 16:57:38 Tower kernel: [drm] Initialized i915 1.6.0 20200917 for 0000:00:02.0 on minor 0
Dec  4 16:57:38 Tower kernel: ACPI: Video Device [GFX0] (multi-head: yes  rom: no  post: no)
Dec  4 16:57:38 Tower kernel: acpi device:15: registered as cooling_device9
Dec  4 16:57:38 Tower kernel: input: Video Bus as /devices/LNXSYSTM:00/LNXSYBUS:00/PNP0A08:00/LNXVIDEO:00/input/input12
Dec  4 16:57:39 Tower kernel: fbcon: i915drmfb (fb0) is primary device
Dec  4 16:57:39 Tower kernel: Console: switching to colour frame buffer device 240x67
Dec  4 16:57:39 Tower kernel: i915 0000:00:02.0: [drm] fb0: i915drmfb frame buffer device
Dec  4 16:57:39 Tower kernel: i915 0000:00:02.0: MDEV: Registered
Dec  4 16:57:40 Tower root: 
Dec  4 16:57:40 Tower root: -------------Installation of Intel GVT-g successful--------------
Dec  4 16:57:40 Tower root: 
Dec  4 16:57:40 Tower root: 
Dec  4 16:57:40 Tower root: -----------------------------------------------------------------
Dec  4 16:57:40 Tower root: -----------------Intel GVT-g plugin installed--------------------
Dec  4 16:57:40 Tower root: -----------------------------------------------------------------
Dec  4 16:57:40 Tower root: plugin: intel-gvt-g.plg installed
Dec  4 16:57:40 Tower root: plugin: installing: /boot/config/plugins/libvirt.hotplug.usb.plg
Dec  4 16:57:40 Tower root: plugin: running: anonymous
Dec  4 16:57:40 Tower root: plugin: skipping: /boot/config/plugins/libvirt.hotplug.usb/libvirt.hotplug.usb-2021.09.17.tgz already exists
Dec  4 16:57:40 Tower root: plugin: running: anonymous
Dec  4 16:57:40 Tower root: plugin: running: anonymous

 

attached is the dialogics, thanks so much!

 

tower-diagnostics-20211204-1914.zip

[ich777]

4 hours ago, qy2009 said:

I switched to a new cpu- coffeelake, and when I tried to use intel gvt and adding the grahpic card inside VM, I cannot update the VM settings, it said "XML error: 'model' attribute in <hostdev> is only supported when type='pci'". what does this mean? 

I would recommend that you are deleting the assigned vGPU on the plugin page, try to start the VM again without a vGPU attached to see if it gives you an error.

If the VM starts fine, try to assign a new vGPU and then try to start the VM again.

[kkkk]

On 12/4/2021 at 1:24 PM, ich777 said:

Can you please share your Diagnostics, I've tried that now on my server and it works flawlessly.

sure！

Do you need an unraid diagnosis？

diagnostics-20211205-1719.zip

 

[qy2009]

8 hours ago, ich777 said:

I would recommend that you are deleting the assigned vGPU on the plugin page, try to start the VM again without a vGPU attached to see if it gives you an error.

If the VM starts fine, try to assign a new vGPU and then try to start the VM again.

 

without a vGPU settings, the VM can start up fine.

I tried removed the vGPU in intel-gvt-g page, and created a new one, but still got the same error message.

[ich777]

7 hours ago, kkkk said:

Do you need an unraid diagnosis？

I see nothing suspicious from the Diagnostics.

 

On 12/3/2021 at 1:13 PM, kkkk said:

when VM win10 is running too, the transcoding will fall to less than 20fps, the power draw is limited to 1W each

Have you enabled throttling in Emby?

Maybe try to disable it for testing purposes.

[ich777]

35 minutes ago, qy2009 said:

I tried removed the vGPU in intel-gvt-g page, and created a new one, but still got the same error message.

Which VM are you running with GVT-g?

[qy2009]

Just now, ich777 said:

Which VM are you running with GVT-g?

I have tried win7, win10, and win 11. all with same error message.

[ich777]

3 minutes ago, qy2009 said:

I have tried win7, win10, and win 11. all with same error message.

But I don't see a entry for GVT-g in any of the VM xml's from the Diagnostics that you've provided above.

[qy2009]

ah, I removed them a while ago. I just tried it again, and go the same message. can you check the attached log? 

 

thanks so much!

tower-diagnostics-20211205-0912.zip

[ich777]

Just now, qy2009 said:

thanks so much!

Now I get it, you are using a U processor.

A few pages back a user got a similar issue on a U processor and we weren't able to solve it.

If I'm not mistaken those are low power CPUs from Intel, maybe it's related to that and they stripped out the GVT-g feature, that my only/best guess.

 

In the newly attached Diagnostics is nothing either...

 

But I think that U processors are not supported, you can only create a issue on Github, wait I saw now that you've created an issue about that...

 

I would rather stick to the guide from Arch, can you try to do it manually and see if it is working?

[qy2009]

3 minutes ago, ich777 said:

Now I get it, you are using a U processor.

A few pages back a user got a similar issue on a U processor and we weren't able to solve it.

If I'm not mistaken those are low power CPUs from Intel, maybe it's related to that and they stripped out the GVT-g feature, that my only/best guess.

 

In the newly attached Diagnostics is nothing either...

 

But I think that U processors are not supported, you can only create a issue on Github, wait I saw now that you've created an issue about that...

 

I would rather stick to the guide from Arch, can you try to do it manually and see if it is working?

 

thanks @ich777, I guess I have to try it manually then. thanks so much for your reply