Arch Linux

The_Moves

Member

Registered: 2015-01-06

Posts: 59

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

so I suppose I can't use ovmf...

Have you looked at a different rom?

http://www.techpowerup.com/vgabios/inde … =&memSize=

Last edited by The_Moves (2015-05-14 00:15:53)

corna

Member

Registered: 2015-05-13

Posts: 12

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

Yesterday I ran QEMU without specifing any GPU ROM and it worked!
Will I experience any performance loss/problem without a GPU ROM?
Thanks for your help btw :-)

Last edited by corna (2015-05-14 07:55:28)

Bronek

Member

From: London

Registered: 2014-02-14

Posts: 123

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

Yesterday I ran QEMU without specifing any GPU ROM and it worked!
Will I experience any performance loss/problem without a GPU ROM?
Thanks for your help btw :-)

No at all; using GPU ROM only makes sense as a workaround in certain situations.

Schlunze

Member

Registered: 2013-10-03

Posts: 53

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

Hello again,

i know that this topic is about vga-passthrough in the first place, but it´s the most experienced community in virtualisation topics out there!

I have absolutly no idea how to install Windows 7 UEFI correctly as VM or on bare metal.

I had some success passing through the intel sata controller to VM.

Installing Windows 7 VM with UEFI using ovmf/libvirt/qemu on the passed through sata controller with attached ssd works fine, until installer reboots.
I´ve not configured any hd´s in vmmanager, just the sata controller, loaded all virtio drivers from iso (hd was visible without driver).
After installation reboot, Windows 7 UEFI VM end up in ovmf EFI-Shell. Partitioning scheme and data is written to disk, just Windows Boot Manager fails to boot in ovmf.

I´ve read some older posts about the fat32 uefi partition and i think i´m really close to get the VM to work without the option to dual boot on bare metal. Only missing a little detail about configuring ovmf/qemu/libvirt to boot from passed through SATA controller with attached disk that contains the uefi fat32 partiton.

Does qemu / libvirt / ovmf need a configured disk to boot?
I use ovmf 16229-1. Do i need a special ovmf svn package from AUR?

A bunch of another off topic questions is Windows 7 UEFI bare metal installation in this case (dual boot part).
Looks like nowadays UEFI Windows installations are not as simple as they should be, or me missing very much basic knowledge about UEFI Systems.

My MB is Asrock Z97Extreme6.

I´ve found no useful article that describes my usecase for basic hd UEFI configuration in main bios.
- 1 linux gpt disk with syslinux on SATA A0
- 3 gpt lvm disks (storage for home network) on SATA A1-3
- 2 lvm disks without partition table (unplugged for testing) SATA1-2
- 1 windows7 uefi gpt disk (for win7 installation unformatted, or like later described with fully installed win7 uefi) on SATA0
- 1 DVDRW on SATA A4

SATA connectors without A are the main ones from the Intel sata-controller.

In the good old non UEFI times just the disk configuration in main BIOS decided which disk is started, after Windows 7 UEFI installation i´ve got this strange Windows Boot Manager entry on main bios.

Do UEFI-Systems read data from disks for general hw initiation / OS initiation?
Why there is this new Windows entry in my BIOS instead of the disk description in main bios?
Is gummiboot a replacement for Windows Boot Manager or does it have to be the main bootloader that "chainloads" other OSses (WBM; syslinux...)?

These disks are my requirements to make use of UEFI on bare metal Asrock Bios. But im not able to install Win7 in this - i think - pretty common configuration.

It only installs and works, if i unplug all other disks. If i plug in the disks back on this previous working installation:
Windows Boot Manager fails to boot,
syslinux from linux gpt disk is second option in boot legacy, syslinux works normally from this point on.

I have absolutly no clue why a normal Windows 7 UEFI installation on bare metal does not work with all my disks attached.
All disks have gpt like described, except two lvm disks, they have no partition table. I´ve unpluged them for this test. I´ve tested with various disks unplugged, no luck.

Maybe someone of u have some experiences with this kinda UEFI stuff.

Ssd for Win7 is connected to sata0 on intel sata controller. Linux is on another disk (A0) with syslinux as another boot option in main bios.
With almost all necessary disks, exept the two disks without partition table (unplugged) the installation can´t be completed after reboot on bare metal.
Maybe its a wrong main bios configuration?

Maybe gummiboot is needed on bare metal to "chainload?" uefi OS and normal syslinux as another boot option, as main bootloader not the Asrock one.

Back to the VM - part.

Can somebody share a configuration example who managed to get a sata controller passthrough to work (with win7 uefi install, ovmf, libvirt xml)?

Thank you in advance.

Last edited by Schlunze (2015-05-14 10:57:19)

Duelist

Member

Registered: 2014-09-22

Posts: 358

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

The output of rom-parser (http://vfio.blogspot.it/2014/08/does-my … t-efi.html) is this one

        Valid ROM signature found @0h, PCIR offset 190h
        PCIR: type 0, vendor: 10de, device: 1183, class: 030000
        PCIR: revision 0, vendor revision: 1
        Last image

so I suppose I can't use ovmf...

What about libvirt? Can you link me the blog?
QEMU's version is 2.1, a recent one

Try checking your manufacturer's site for VBIOS updates - maybe there'll be an UEFI-compatible(with GOP) one. Or lurk through the techpowerup database.
Because my gigabyte GT610 supports GOP.

---

The forum rules prohibit requesting support for distributions other than arch.
I gave up. It was too late.
What I was trying to do.
The reference about VFIO and KVM VGA passthrough.

protake

Member

Registered: 2014-09-09

Posts: 9

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

The iommu group of my second card contains a lot of stuff, for example my network card.

ls /sys/bus/pci/devices/0000:02:00.0/iommu_group/devices/
    0000:00:1c.0  0000:00:1c.5  0000:00:1c.6  0000:02:00.0  0000:02:00.1  0000:03:00.0  0000:04:00.0

Unless I bind all these devices, qemu won't start. Is there anything I can do about this? This prevents my network connection from working, for example. Is my motherboard at fault here?

Last edited by protake (2015-05-14 11:38:28)

corna

Member

Registered: 2015-05-13

Posts: 12

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

The output of rom-parser (http://vfio.blogspot.it/2014/08/does-my … t-efi.html) is this one

        Valid ROM signature found @0h, PCIR offset 190h
        PCIR: type 0, vendor: 10de, device: 1183, class: 030000
        PCIR: revision 0, vendor revision: 1
        Last image

so I suppose I can't use ovmf...

What about libvirt? Can you link me the blog?
QEMU's version is 2.1, a recent one

Try checking your manufacturer's site for VBIOS updates - maybe there'll be an UEFI-compatible(with GOP) one. Or lurk through the techpowerup database.
Because my gigabyte GT610 supports GOP.

I've tried rom-parser with this Gigabyte BIOS and this is the output

Valid ROM signature found @400h, PCIR offset 190h
    PCIR: type 0, vendor: 10de, device: 1183, class: 030000
    PCIR: revision 0, vendor revision: 1
    Last image

but the BIOS description says

5. UEFI BIOS

so I suppose that the rom-parser script is broken. Later I'll try if it works with OVMF

slis

Member

Registered: 2014-06-02

Posts: 127

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

The iommu group of my second card contains a lot of stuff, for example my network card.

ls /sys/bus/pci/devices/0000:02:00.0/iommu_group/devices/
    0000:00:1c.0  0000:00:1c.5  0000:00:1c.6  0000:02:00.0  0000:02:00.1  0000:03:00.0  0000:04:00.0

Unless I bind all these devices, qemu won't start. Is there anything I can do about this? This prevents my network connection from working, for example. Is my motherboard at fault here?

You need ACS override patch for kernel.

protake

Member

Registered: 2014-09-09

Posts: 9

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

The iommu group of my second card contains a lot of stuff, for example my network card.

ls /sys/bus/pci/devices/0000:02:00.0/iommu_group/devices/
    0000:00:1c.0  0000:00:1c.5  0000:00:1c.6  0000:02:00.0  0000:02:00.1  0000:03:00.0  0000:04:00.0

Unless I bind all these devices, qemu won't start. Is there anything I can do about this? This prevents my network connection from working, for example. Is my motherboard at fault here?

You need ACS override patch for kernel.

Hm, strange. I was using the linux-vfio kernel yesterday, and I'm pretty sure it was the same. I don't mind trying it again if it works, though...

aw

Member

Registered: 2013-10-04

Posts: 921

Website

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

The iommu group of my second card contains a lot of stuff, for example my network card.

ls /sys/bus/pci/devices/0000:02:00.0/iommu_group/devices/
    0000:00:1c.0  0000:00:1c.5  0000:00:1c.6  0000:02:00.0  0000:02:00.1  0000:03:00.0  0000:04:00.0

Unless I bind all these devices, qemu won't start. Is there anything I can do about this? This prevents my network connection from working, for example. Is my motherboard at fault here?

You need ACS override patch for kernel.

Hm, strange. I was using the linux-vfio kernel yesterday, and I'm pretty sure it was the same. I don't mind trying it again if it works, though...

There are kernel commandline options required to enable the acs override patch.  What chipset is this?  We have quirks in place for nearly all Intel PCH root ports, 9-series is coming in kernel 4.2

---

http://vfio.blogspot.com
Looking for a more open forum to discuss vfio related uses?  Try https://www.redhat.com/mailman/listinfo/vfio-users

protake

Member

Registered: 2014-09-09

Posts: 9

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

There are kernel commandline options required to enable the acs override patch.  What chipset is this?  We have quirks in place for nearly all Intel PCH root ports, 9-series is coming in kernel 4.2

Intel H97.

Denso

Member

Registered: 2014-08-30

Posts: 179

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

Intel H97.

9-series is coming in kernel 4.2

protake

Member

Registered: 2014-09-09

Posts: 9

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

Intel H97.

9-series is coming in kernel 4.2

Yeah, I understood as much. I just wanted to mention it. Thanks for all the replies!

edit: Is there anything I could do until 4.2 hits?

Last edited by protake (2015-05-14 16:43:29)

apex8

Member

Registered: 2014-03-29

Posts: 60

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

I have the problem that google play movies doesn't work anymore on my arch host since around February this year. I didn't find a solution yet, also tried to use the windows version of firefox in wine, but it won't work.
Now I'm wondering why it also doesn't work in my Windows 7 VM. I havn't installed much there. Basically it's a Firefox, flash player, Steam and drivers. Are there any host requirements or some qemu configuration I need to respect for that?

BR
apex

Duelist

Member

Registered: 2014-09-22

Posts: 358

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

I have the problem that google play movies doesn't work anymore on my arch host since around February this year. I didn't find a solution yet, also tried to use the windows version of firefox in wine, but it won't work.
Now I'm wondering why it also doesn't work in my Windows 7 VM. I havn't installed much there. Basically it's a Firefox, flash player, Steam and drivers. Are there any host requirements or some qemu configuration I need to respect for that?

BR
apex

First thing that comes to my mind - some weird DRM system. You suuure that everything else works fine?

---

The forum rules prohibit requesting support for distributions other than arch.
I gave up. It was too late.
What I was trying to do.
The reference about VFIO and KVM VGA passthrough.

apex8

Member

Registered: 2014-03-29

Posts: 60

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

I have the problem that google play movies doesn't work anymore on my arch host since around February this year. I didn't find a solution yet, also tried to use the windows version of firefox in wine, but it won't work.
Now I'm wondering why it also doesn't work in my Windows 7 VM. I havn't installed much there. Basically it's a Firefox, flash player, Steam and drivers. Are there any host requirements or some qemu configuration I need to respect for that?

BR
apex

First thing that comes to my mind - some weird DRM system. You suuure that everything else works fine?

You mean in the VM? Gaming is totally fine. Couldn't think of anything in particular, which is missing..

Duelist

Member

Registered: 2014-09-22

Posts: 358

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

I have the problem that google play movies doesn't work anymore on my arch host since around February this year. I didn't find a solution yet, also tried to use the windows version of firefox in wine, but it won't work.
Now I'm wondering why it also doesn't work in my Windows 7 VM. I havn't installed much there. Basically it's a Firefox, flash player, Steam and drivers. Are there any host requirements or some qemu configuration I need to respect for that?

BR
apex

First thing that comes to my mind - some weird DRM system. You suuure that everything else works fine?

You mean in the VM? Gaming is totally fine. Couldn't think of anything in particular, which is missing..

Oh, it doesn't work on the host...
What hardware do you have on the host? Maybe it's just some messed up codecs(google likes using "new" codecs like vp8/9 or something)?

Slightly OT:
guys, please, save me from going insane - i'm considering buying a third asus HD7750 silent GPU. It's located ~180km away, one of the cheapest 7750s, and i don't even have space in the case to fit it yet. But something shivers inside me when i look at it. Just imagine running a tri-way crossfire(however, amd support once said that crossfire requires an even number of cards) in a VM. Even if crossfire won't work, i'll just set up another VM... Maybe running it will require some hardcore case modding butchering and fitting a riser-card to run it from x1 pci-e, or maybe i'll need to switch to kaveri C/APUs for pci-e3.0... But daaamn, i want that card, while understanding that the idea is bizarre.

Last edited by Duelist (2015-05-14 19:20:39)

---

The forum rules prohibit requesting support for distributions other than arch.
I gave up. It was too late.
What I was trying to do.
The reference about VFIO and KVM VGA passthrough.

aw

Member

Registered: 2013-10-04

Posts: 921

Website

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

Slightly OT:
guys, please, save me from going insane - i'm considering buying a third asus HD7750 silent GPU...

That's certainly a strange obsession you've got for a relatively mediocre card.  I've got my own, so I won't rib you on it too hard.  It's at least silent, but if you're already got two of these power sucking cards in your system, do you really want to add 50% more?  Personally I don't see why CF/SLI is even on the radar for most people, single card performance on a recent card should handily beat multi-card performance from a few years ago.  I don't know about AMD, but Nvidia is doing some pretty impressive work on power efficiency with Maxwell.

---

http://vfio.blogspot.com
Looking for a more open forum to discuss vfio related uses?  Try https://www.redhat.com/mailman/listinfo/vfio-users

Denso

Member

Registered: 2014-08-30

Posts: 179

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

edit: Is there anything I could do until 4.2 hits?

Sure , compile a kernel with ACS patch , add :

pcie_acs_override=downstream

to your boot parameters .

Cheers !

paperinick

Member

Registered: 2015-04-14

Posts: 8

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

I am trying to use OVMF but after reboot the VM hangs with a black screen (even if I don't have VGA passthrough enabled). This happens even with some Linux distros as guest.
My fear is that I might be hitting some random daily OVMF build bug. If you have a successful setup with OVMF would you mind sharing the OVMF image you are using? Thanks!!

Duelist

Member

Registered: 2014-09-22

Posts: 358

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

Slightly OT:
guys, please, save me from going insane - i'm considering buying a third asus HD7750 silent GPU...

That's certainly a strange obsession you've got for a relatively mediocre card.  I've got my own, so I won't rib you on it too hard.  It's at least silent, but if you're already got two of these power sucking cards in your system, do you really want to add 50% more?  Personally I don't see why CF/SLI is even on the radar for most people, single card performance on a recent card should handily beat multi-card performance from a few years ago.  I don't know about AMD, but Nvidia is doing some pretty impressive work on power efficiency with Maxwell.

Each card consumes only 55W full-throttle.Yes, that's almost twice as less than CPU's TDP. Considering ULPS and zerocore and simple declocking - when half-loaded or idle they should consume even less.
50% more? Not 50%. Second card gave me 1.3x-1.9x performance(unigine benchmark makes a good use of CF, heh). The reason i'm obsessed with this - i can't sell them due to... multiple reasons.
But it's not enough performance-wise(and as my mind hints me - it'll never be), and having a somewhat unique and bizarre config makes me feel funny. Oh, and 2gb of video ram isn't quite awesome for my taste, however i don't like that tendency to bloat video ram sizes(12GB?! 512mb was enough-for-everyone some years ago)

Oh, and that HD7750 is still good on performance/power/money chart. Of course, GTX750 overtakes it, but still.
Due to some magic, i could've continued mining litecoins and still have profit(because 55w of electricity was equal to zero cents per month. I'm serious) four months ago. That's how weird these cards are.

I am trying to use OVMF but after reboot the VM hangs with a black screen (even if I don't have VGA passthrough enabled). This happens even with some Linux distros as guest.
My fear is that I might be hitting some random daily OVMF build bug. If you have a successful setup with OVMF would you mind sharing the OVMF image you are using? Thanks!!

Weeeeeeird. I have libvirt and kraxel-org-based jenkins-git images, and it works fine there.

Last edited by Duelist (2015-05-15 21:20:55)

---

The forum rules prohibit requesting support for distributions other than arch.
I gave up. It was too late.
What I was trying to do.
The reference about VFIO and KVM VGA passthrough.

aw

Member

Registered: 2013-10-04

Posts: 921

Website

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

...because 55w of electricity was equal to zero cents per month....

How do you figure that?  Free electric?  Surplus production?  55 watts * 24 hours = 1.32kW, which is 10 cents per day from my utility even at the lowest tier

---

http://vfio.blogspot.com
Looking for a more open forum to discuss vfio related uses?  Try https://www.redhat.com/mailman/listinfo/vfio-users

sweetsuicide

Member

Registered: 2015-05-15

Posts: 1

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

Hello everybody here,
this is some awesome work done here. I built a computer specifically for running windows inside a vm with secondary GPU passthough. I followed instructions from the main page here mixed with the similar https://wiki.archlinux.org/index.php/PC … h_via_OVMF guide. This means that I have the builds from Hoffmann's repository. I have not stepped up to libvirt yet, mainly because I haven't installed any DE or WM.
So here's my problem: everything looks fine (qemu boots up, secondary gpu fires up with qxl), but I cannot install windows. Windows8 iso is not even browsable through the efi shell. Windows 7 is readable but it stops immediately with an error (boot related, 0xc000000d code something).
I have no I idea what's happening.
Setup is following:
i7-4770
ASRock Z87 Extreme4
16GB RAM
two ssd's
R9 280x

radeon module was blacklisted (even it doesn't seem like it's necessary, as video card powers up when I start qemu). Qemu command:

qemu-system-x86_64\
 -enable-kvm\
 -cpu host,kvm=off\
 -nographic -m 8192 -vga qxl\
 -device vfio-pci,host=01:00.0 -device vfio-pci,host=01:00.1 -device vfio-pci,host=00:1b.0\
 -drive if=pflash,format=raw,readonly,file=/usr/share/edk2.git/ovmf-x64/OVMF_CODE-pure-efi.fd\
 -drive if=pflash,format=raw,file=/home/luca/this.fd\ 
 -device virtio-scsi-pci,id=scsi -drive file=/mnt/SAND/win.cow,id=disk,if=none,cache=writeback\
 -device scsi-hd,drive=disk -drive file=/home/user/Windows81.iso,id=isocd,if=none -device scsi-cd,drive=isocd\
 -drive file=/home/luca/virtio-win-0.1.96.iso,id=virtiocd,if=none -device ide-cd,bus=ide.1,drive=virtiocd\
 -usb -usbdevice host:045e:07f8 -usbdevice host:046d:c408

Command given as root (not much choice here)

aw

Member

Registered: 2013-10-04

Posts: 921

Website

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

secondary gpu fires up with qxl

What does that even mean?

qemu-system-x86_64\
 -enable-kvm\
 -cpu host,kvm=off\
 -nographic -m 8192 -vga qxl\

Huh?  QXL is an emulated graphics card.  It might make sense to have it since you claim to be wanting to setup a secondary graphics configuration, but what's supposed to happen with -nographic?  Where is the QXL graphics supposed to display?  Your choices are SDL, VNC, or Spice, but if you leave -nographic, you might as well dump the QXL device with -vga none.

Others have had trouble with the ISO format, that may be what you're running into with the install image, something about UDF vs ISO9660.  I've never had a problem with images from MSDN.

Your setup seems textbook enough that I'd follow the howto guide in my blog (link in sig) once you get your ISO format straightened out (libvirt does not require a DE/WM.  Even the graphical tools like virt-manager can be run remotely if your system is headless).

BTW, you don't need kvm=off for an AMD GPU.

---

http://vfio.blogspot.com
Looking for a more open forum to discuss vfio related uses?  Try https://www.redhat.com/mailman/listinfo/vfio-users

apex8

Member

Registered: 2014-03-29

Posts: 60

Re: KVM VGA-Passthrough using the new vfio-vga support in kernel =>3.9

What would be a good choice for a motherboard at the moment, given that I still want to use it for a gaming VM and having at least two pcie x16 slots?