Arch Linux

Col_Panic

Member

Registered: 2015-07-19

Posts: 4

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

So I'm not quite ready to give up yet.
Since it looks to me like my [kvm/machine/driver] is trying to map and unmap the same memory region over and over again when it tries to attach the Radeon card to the host, is it possible that multiple devices are competing for the same memory space, or it's otherwise inaccessible? Maybe there's a way to tell VFIO to set the DMA mappings for this device at an early stage of the Linux boot process?
Or maybe I just have no idea what I'm talking about, and should give up already?

aw

Member

Registered: 2013-10-04

Posts: 921

Website

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

@Col_Panic

in fact, the majority of the IOMMU mappings for the VM are setup prior to an execution of the VM, it's only a small set of legacy ranges and memory mapped I/O regions that change run-time.

---

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

The_Moves

Member

Registered: 2015-01-06

Posts: 59

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

@impulse_255

Have you run into the KMODE_EXCEPTION_NOT_HANDLED error?

I've modified my GPU rom to have the Quadro K5000 (from a GTX 760) Device ID as well as the appropriate checksum. However, when I pass the romfile, all i get is a black screen... When I add a Virtual Console to the VM and view the boot through virt-manager, I see the above mentioned error. When I just change the device IDs and not pass a ROM, I am able to get Windows 10 setup to start, but when formatting the disk it shows the above error. Booting off a Windows 7 disk, all I get is a Windows 7 pulsating windows icon.

For some reason now, i can't even get my VM to boot with the previous config without hacking the device ID. I have power cycled my system, but I haven't pulled the GPU, maybe something bad is being stored on the GPU?

Denso

Member

Registered: 2014-08-30

Posts: 179

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

@The_Moves :

I had this exact BSOD with this exact error .

As soon as I switched to an older version of OVMF (15000~ something) , it was solved and I could install Windows 10 without a BSOD .

impulse_255

Member

Registered: 2015-06-29

Posts: 22

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

@The_Moves:

Did you modify your own rom or did you download it from somewhere? I tried a 3rd party rom for my card and got a black screen. Likewise a rom dump made in the guest from an initialised card would not work. The only way that worked was this incantation after a power cycle/suspend:

# echo 1 > /sys/bus/pci/devices/<pciid>/rom;
# cat /sys/bus/pci/devices/<pciid>/rom > vbios.rom;
# echo 0 > /sys/bus/pci/devices/<pciid>/rom

Also, changing IDs/roms without a power cycle gave me KMODE_EXCEPTION_NOT_HANDLED BSODs.

Did you have a working, installed guest before? Like said, the new OVMF builds throw the same BSOD at boot or installation, namely at the copying files step. Windows 7 also won't work with OVMF without some magic but I gave up with that.

The_Moves

Member

Registered: 2015-01-06

Posts: 59

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

@Denso and @impulse_255

I will have to look at downgrading my qemu to what I was at before, I believe 2.1 or so. Impulse, what version are you running currently? You guys described exactly what I was experiencing. I believe the ROM file I was using was a dump from my card, but i'll boot onto my FreeDOS USB and grab a new copy. I followed updating the Device ID and Checksum in the ROM, but maybe I should do the Subsystem ID as well - I only discovered late last night that the x-ss-vid was for subsystem...

I've had a working Windows 10 build with OVMF for months now , only got the issue when I was when I started messing with the IDs for the card and used the latest version of Qemu to do it. It was my gaming system and was working on performance tweaking, trying to get as many FPS as possible.

I've seen this before, but what does it do? How does echoing 1 to a file change things?

# echo 1 > /sys/bus/pci/devices/<pciid>/rom;
# cat /sys/bus/pci/devices/<pciid>/rom > vbios.rom;
# echo 0 > /sys/bus/pci/devices/<pciid>/rom

@impulse_255, by changing IDs you mean changing the x-vid and x-ss-vid in the qemu line?

Nazfellun

Member

Registered: 2012-10-22

Posts: 19

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

Well, I ended up flashing the updated vbios from gigabyte, which has a valid UEFI rom signature. This worked fine and still runs as normal in my seabios setup on the same system, but I get the same issue as before when trying to install catalyst -- it gets up to installing the driver proper, the system momentarily freezes up and then reboots during installing the actual driver. I do get (low-resolution) output on the physical card up until the VM freezes and reboots.

I've pasted my vm config here. Is there anything I might be missing on the host side of the setup for ovmf to work?

impulse_255

Member

Registered: 2015-06-29

Posts: 22

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

@Denso and @impulse_255

I will have to look at downgrading my qemu to what I was at before, I believe 2.1 or so. Impulse, what version are you running currently? You guys described exactly what I was experiencing. I believe the ROM file I was using was a dump from my card, but i'll boot onto my FreeDOS USB and grab a new copy. I followed updating the Device ID and Checksum in the ROM, but maybe I should do the Subsystem ID as well - I only discovered late last night that the x-ss-vid was for subsystem...

I've had a working Windows 10 build with OVMF for months now , only got the issue when I was when I started messing with the IDs for the card and used the latest version of Qemu to do it. It was my gaming system and was working on performance tweaking, trying to get as many FPS as possible.

Oh. :( I can't say what's wrong there. Power cycling has fixed all the black screen/BSOD messes I've caused so far, and there shouldn't be any lasting hardware changes involved. The installer issues sound exactly like the OVMF issue. I once accidentally used a fresh, non-working version with my already installed Windows and it crashed all the same, so the issue is not limited to just the installer. Are you still using the same OVMF as before?

I'm currently using Qemu 2.3.0 and the old sata patched OVMF build whose build date or number I don't know.

I've seen this before, but what does it do? How does echoing 1 to a file change things?

Frankly, I don't know what it exactly does. It's some kernel interface that presents itself as a file.

@impulse_255, by changing IDs you mean changing the x-vid and x-ss-vid in the qemu line?

Yes. To clarify: if I booted the guest with one set of IDs, shut it down and tried to boot it again without a power cycle (or S3 suspend, which I use) with another set of IDs, Windows would BSOD. This happened even if the IDs and rom would otherwise work when I stick to using one at a time. BTW, did you modify the subsystem ID in the rom as well? I searched & replaced all I could find: 2 instances of PCI ID and one of SS ID. edit: couldn't read

Last edited by impulse_255 (2015-07-22 17:19:08)

aw

Member

Registered: 2013-10-04

Posts: 921

Website

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

BTW, I reported the Win8.1/OVMF install issue to Laszlo who bisected and found the commit that introduced it.  There's already a patch on their list to fix it, but the ongoing sorceforge outage is delaying it being committed and new builds including it.  Stick with a pre-July build until things get moving again.

---

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

impulse_255

Member

Registered: 2015-06-29

Posts: 22

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

BTW, I reported the Win8.1/OVMF install issue to Laszlo who bisected and found the commit that introduced it.  There's already a patch on their list to fix it, but the ongoing sorceforge outage is delaying it being committed and new builds including it.  Stick with a pre-July build until things get moving again.

Great, thanks!

The_Moves

Member

Registered: 2015-01-06

Posts: 59

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

I installed stock Qemu 2.1 from Source RPMs and am still getting the 'KMODE_EXCEPTION_NOT_HANDLED' error. When I remove libvirt and qemu, are there other files I should remove manually? I'm almost to the point of rebuilding from stock Fedora 21 or Fedora 22. I had no issues until I upgraded from 21 to 22, but 21 is going out of support soon.

The_Moves

Member

Registered: 2015-01-06

Posts: 59

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

BTW, I reported the Win8.1/OVMF install issue to Laszlo who bisected and found the commit that introduced it.  There's already a patch on their list to fix it, but the ongoing sorceforge outage is delaying it being committed and new builds including it.  Stick with a pre-July build until things get moving again.

Oh, you mean OVMF itself. The OVMF firmware from EDK2 has an issue...so my Qemu/Libvirt reinstalling really wasn't the issue...ugh. I need to read closer.

Edit: I got finally got it working, but am still unable to pass a romfile or the multifunction attribute (not sure what it does). Windows does see the card as a K5000. In my game World of Tanks, normally when I would hover over different icons for tanks in the Garage, the screen would freeze for a split second. Overall though, there really isn't an improvement in FPS from what I can see. Loading into maps has improved by 50%. But it wouldn't hurt to do some actual benchmarks.

  <qemu:commandline>
    <qemu:arg value='-device'/>
    <qemu:arg value='vfio-pci,host=04:00.0,addr=0x0f,addr=0x0f,bus=pci.0,x-vid=0x10DE,x-did=0x11BA,x-ss-vid=0x10DE,x-ss-did=0x0965,x-vga=on'/>
    <qemu:arg value='-device'/>
    <qemu:arg value='vfio-pci,host=04:00.1,addr=0x10,bus=pci.0'/>
  </qemu:commandline>

Now for some reason, my Start Menu doesn't work, nor does Edge. It might just be a downlevel version of OVMF, libvirt, or qemu.

Edit 2: I was able to get my Start Menu and Edge to work with a reset of Windows itself.

Last edited by The_Moves (2015-07-23 23:15:38)

jack_boss

Member

Registered: 2014-09-09

Posts: 6

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

Hello, I need help; before I go buy a KM switch which are not cheap considering I have to import it.. etc.

I need to passthrough a USB Host controller for mouse and keyboard, becasue if I want to pass through a USB Host device it has to be attached. I went with the following and it has its own iommu group:

USB controller: Intel Corporation 9 Series Chipset Family USB EHCI Controller #1
/sys/kernel/iommu_groups/9/devices/0000:00:1d.0

lspci -n -s 00:1d gave me 8086:8ca6 which I added to my kernel command line.
The other vendor:product id's are for my GTX770

options		root=UUID=1aea1286-102b-444d-9f30-494305255937 rw elevator=noop intel_iommu=on iommu=pt pci-stub.ids=10de:1184,10de:0e0a,8086:8ca6 hugepages=2048

A USB Controller was already present (I've tried USB2 and USB3, same results)
Finally in virt-manager I clicked Add Hardware >PCI Host Device> USB controller: Intel Corporation 9 Series Chipset Family USB EHCI Controller #1

In the VM (Windows) I can see drivers being loaded and installed. The thing is, it doesnt matter in which port I plug in my mouse (removed from host/tried wireless usb mouse dongle, pen drive) nothing is listed or moves (just in case i tried every port)

Please help.

Just if somebody needs to see this:

00:1d.0 USB controller: Intel Corporation 9 Series Chipset Family USB EHCI Controller #1
	Subsystem: ASRock Incorporation Device 8ca6
	Kernel driver in use: vfio-pci
	Kernel modules: ehci_pci

lsmod | grep -E "virt|kvm|iommu"

vfio_iommu_type1       20480  1
vfio                   24576  8 vfio_iommu_type1,vfio_pci
kvm_intel             155648  4
kvm                   442368  1 kvm_intel
virtio_blk             20480  0
virtio_scsi            20480  0
scsi_mod              151552  5 uas,usb_storage,libata,sd_mod,virtio_scsi
virtio_net             32768  0
virtio_ring            20480  3 virtio_blk,virtio_net,virtio_scsi
virtio                 16384  3 virtio_blk,virtio_net,virtio_scsi

fld

Member

Registered: 2015-02-06

Posts: 9

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

Hello, I need help; before I go buy a KM switch which are not cheap considering I have to import it.. etc.

I need to passthrough a USB Host controller for mouse and keyboard, becasue if I want to pass through a USB Host device it has to be attached. I went with the following and it has its own iommu group:

USB controller: Intel Corporation 9 Series Chipset Family USB EHCI Controller #1
/sys/kernel/iommu_groups/9/devices/0000:00:1d.0

lspci -n -s 00:1d gave me 8086:8ca6 which I added to my kernel command line.
The other vendor:product id's are for my GTX770

options		root=UUID=1aea1286-102b-444d-9f30-494305255937 rw elevator=noop intel_iommu=on iommu=pt pci-stub.ids=10de:1184,10de:0e0a,8086:8ca6 hugepages=2048

A USB Controller was already present (I've tried USB2 and USB3, same results)
Finally in virt-manager I clicked Add Hardware >PCI Host Device> USB controller: Intel Corporation 9 Series Chipset Family USB EHCI Controller #1

In the VM (Windows) I can see drivers being loaded and installed. The thing is, it doesnt matter in which port I plug in my mouse (removed from host/tried wireless usb mouse dongle, pen drive) nothing is listed or moves (just in case i tried every port)

Please help.

Just if somebody needs to see this:

00:1d.0 USB controller: Intel Corporation 9 Series Chipset Family USB EHCI Controller #1
	Subsystem: ASRock Incorporation Device 8ca6
	Kernel driver in use: vfio-pci
	Kernel modules: ehci_pci

lsmod | grep -E "virt|kvm|iommu"

vfio_iommu_type1       20480  1
vfio                   24576  8 vfio_iommu_type1,vfio_pci
kvm_intel             155648  4
kvm                   442368  1 kvm_intel
virtio_blk             20480  0
virtio_scsi            20480  0
scsi_mod              151552  5 uas,usb_storage,libata,sd_mod,virtio_scsi
virtio_net             32768  0
virtio_ring            20480  3 virtio_blk,virtio_net,virtio_scsi
virtio                 16384  3 virtio_blk,virtio_net,virtio_scsi

I think that on some Intel chipsets, the kernel maps all your physical USB ports behind the Intel xHCI USB controller by default and leaves the EHCI ones unused. http://lxr.free-electrons.com/source/dr … rks.c#L850. I worked around it by manually changing the Port Routing Mask Register to map physical ports behind the EHCI controllers. The register can simply be changed on the fly.

jack_boss

Member

Registered: 2014-09-09

Posts: 6

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

I think that on some Intel chipsets, the kernel maps all your physical USB ports behind the Intel xHCI USB controller by default and leaves the EHCI ones unused. http://lxr.free-electrons.com/source/dr … rks.c#L850. I worked around it by manually changing the Port Routing Mask Register to map physical ports behind the EHCI controllers. The register can simply be changed on the fly.

Might be worth a try, if I just knew how to do it... first google attempt didn't really help either.

fld

Member

Registered: 2015-02-06

Posts: 9

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

I think that on some Intel chipsets, the kernel maps all your physical USB ports behind the Intel xHCI USB controller by default and leaves the EHCI ones unused. http://lxr.free-electrons.com/source/dr … rks.c#L850. I worked around it by manually changing the Port Routing Mask Register to map physical ports behind the EHCI controllers. The register can simply be changed on the fly.

Might be worth a try, if I just knew how to do it... first google attempt didn't really help either.

You want to access the port routing register in the xhci controllers pci config space. Personally I ended up using "setpci -s 00:14.0 0xd0.W=2037" in my /etc/rc.local file on the Intel 8 Series PCH. Intel has pretty decent documentation available for the chipsets.

Duelist

Member

Registered: 2014-09-22

Posts: 358

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

Someone 2 pages ago asked about APU and it's GPU. His questions remain unanswered ever since.
Well, he was right, but there's some caveats left:
1. disable hugepages. They trash everything.
2. do not try to p-t APU's GPU, p-t only a dedicated one. The integrated one is... integrated into the CPU, so there's a sophisticated power management system present, which doesn't work with only the GPU p-t'd.(i've learned that only by examining amd's BKDG, so no practice experience) As such, even though CrossFireX works, dual graphics will not.
3. x-vga generally doesn't work on these platforms. Use OVMF, which is quite unstable now.
4. some bandwidth issues present on some motherboards. For example, i am crazy enough to run four GPUs on an FM2 system(i have an athlon 750k), and the only pci-e slot routed fully 16x is the first one, the others are 8x(and if there are other devices present in 1x connectors, it works as x4).
5. this platform has good device isolation capabilities, no ACS patch needed. When stuffed with GPUs, there are three IOMMU groups: the first GPU, the second, and the other stuff you have on PCI bus, so you'll have to use xen's pci-assign to use those devices.
6. lspci shows that the integrated GPU is near the root, so there must be no problems with device isolation, you may use it as a host device.
7. the setup switch, AFAIK, routes only VGA. GOP(UEFI) gets mirrored to all available devices and GPUs, making that switch useless. Gigabyte motherboards can change the primary VGA device on any platform, which is a nice feature for our objectives.
8. you might want to try running proprietary drivers on your host GPU. There was some problems in the past, i haven't heard about them very long. Maybe Device ID stanza in xorg.conf helps to sort out that "autobinding" problem.

I hope i've answered most questions about FM2 platforms.
Conclusion: FM2 is not recommended for device pass-through, but you can do a dozen of work-arounds to make it work.

BTW: Win7+OVMF worked for me until july. OVMF whitepaper tells us to add an additional QXL emulated GPU to be able to see something until the GPU fires up properly. Win7 likes VBE, which is emulated only on QXL and not present on real UEFI systems.

---

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.

JHeaton

Member

From: United Kingdom

Registered: 2009-05-16

Posts: 158

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

Hi all. So I set this up over the past couple of days, thanks in huge part to the posters in this thread and the various linked resources. I'm running two GTX 970 cards in my PC and passing them both through to the virtual machine; these are recognised by Windows, I have them showing in Device Manager, the latest nVidia drivers are installed, however I seem unable to use SLI; the nVidia control panel simply does not offer me the option, whereas normally in Windows it does so straight away. Has anyone so far managed to get SLI working using this kind of setup?

Thanks!

The_Moves

Member

Registered: 2015-01-06

Posts: 59

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

@JHeaton - I believe people have gotten CrossFire to work, but not that I remember reading. Have you tried setting it up to there the SLI doesn't use the bridge, but instead the interconnect through the PCIe bus? What motherboard are you running? What generation PCIe?

JHeaton

Member

From: United Kingdom

Registered: 2009-05-16

Posts: 158

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

@JHeaton - I believe people have gotten CrossFire to work, but not that I remember reading. Have you tried setting it up to there the SLI doesn't use the bridge, but instead the interconnect through the PCIe bus? What motherboard are you running? What generation PCIe?

Thanks for the reply. I actually wasn't aware you could even do that, do you have information on how I would go about it? Motherboard is Asrock Z97 Extreme6 and PCIe 3.0.

The_Moves

Member

Registered: 2015-01-06

Posts: 59

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

I've only heard about it. Don't actually utilize SLI.

Last edited by The_Moves (2015-07-24 04:26:29)

SXX

Member

Registered: 2014-03-31

Posts: 15

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

Get second mouse for host (wireless for example) and passthrough main one
or
passthrough your mouse to VM and use Synergy (host as client, VM as server).

Sadly none of this work for me because I have multiple VMs at the same time where all of them must have mouse working in games. Still trying to figure out solution, may be going to try add some non-standard mouse support into QEMU.

Also I'm managed to pass GTX 260 successfully, but only without "x-vga" and active "-vga qxl". It's end up that my issue with it was caused by fact that GTX 260 won't reset properly on VM reset/poweroff so it's remain in some sort of unusable state after first usage. So I bypass it by booting VM with emulated GPU and then just disable/enable GTX 260 via Device Manager: it's end up working just fine.

Now I have other question. I heard that virtio driver must support "read only" mode with filesystem changes written into separate file and QCOW must support that. What I want is to share disk with Steam games between two VMs so both may access it without breakage. Is there anyone who successfully used read-only mode / snapshots with Windows guests?

Last edited by SXX (2015-07-24 14:03:11)

aw

Member

Registered: 2013-10-04

Posts: 921

Website

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

@SXX

It's not very read-only if the filesystem can be changed, is it?  qemu-img can create multiple qcow2 images, all using the same backing store file, so writes are added to each leaf image, not the base image.  I suspect you do not want a read-only image exposed to each VM, because games generally like to apply updates.  There's a QEMU -snapshot option which stores disk writes in memory and throws them away on VM shutdown, but I wouldn't recommend it for this case since any long running VM will quickly consume all your RAM.

---

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

Duelist

Member

Registered: 2014-09-22

Posts: 358

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

Hi all. So I set this up over the past couple of days, thanks in huge part to the posters in this thread and the various linked resources. I'm running two GTX 970 cards in my PC and passing them both through to the virtual machine; these are recognised by Windows, I have them showing in Device Manager, the latest nVidia drivers are installed, however I seem unable to use SLI; the nVidia control panel simply does not offer me the option, whereas normally in Windows it does so straight away. Has anyone so far managed to get SLI working using this kind of setup?

Thanks!

I am the one with the crossfire working. My cards use XDMA crossfire(nvidia doesn't have that technology), which utilizes PCI-e bandwidth instead of an external bridge.
So. Nvidia are some tight asses, and prohibit SLI running on non-SLI certified motherboards.
There was a project called HyperSLI to enable SLI on any motherboard. Generally it requires the machine to have virtualization support(meaning nested virtualization for us), BUT i've red a story where some sick folks ran two GeForce 6600 in SLI on a P3 motherboard(obviously lacking any virtualization), however they had a performance drawback instead of a boost.
I remember someone tried to use a more modern version of HyperSLI, but he failed. You can try too, but i doubt that it will work.

I don't know about 2-chip(SLI on one PCB) solutions working in a virtual machine. Maybe they do work.

---

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.

SXX

Member

Registered: 2014-03-31

Posts: 15

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

First of all thanks for an answer! This is super helpful because all information I can find on internet it's RHEL bugtracker where I suppose everyone talking about Linux guests and bugs related to VMs that's filesystem aware drive is read-only.

It's not very read-only if the filesystem can be changed, is it?

The point it's to have them read-only on host. Reason for this is fact I don't want to apply any changes to system drive because I don't really need any Windows or driver updates and image itself stored on small SSD in compressed state. When Windows using drive it's start to become bigger and bigger so I wish to keep system drive image completely unchanged until I decide otherwise.

I suspect you do not want a read-only image exposed to each VM, because games generally like to apply updates.

It's won't be running like that all the time because I only need it when some friends come to play with me.
So when I need to apply updates I'll just run it like normal VM with writetable drives.

qemu-img can create multiple qcow2 images, all using the same backing store file, so writes are added to each leaf image, not the base image.

It's not exactly what I need, but may be hanby for system drive. Is it possible to use that with existing QCOW image?
I already have Windows guest QCOW image so I want clone into two with little changes.

There's a QEMU -snapshot option which stores disk writes in memory and throws them away on VM shutdown, but I wouldn't recommend it for this case since any long running VM will quickly consume all your RAM.

It's actually exactly what I need. Any idea if it's going to work for physical HDD partition?

-drive file=/dev/disk/by-uuid/0AD949C94647B394,format=raw,if=virtio

Also according to wiki "any changes made to the virtual machine while it is running are written to temporary". Would be interesting if I can write temporary changes on my HDD RAID.

Currently my setup looks like that:
- Host system and two Windows QCOW images (13GB each) on small SSD (60GB).
- RAID of two 250GB HDDs for data.
- 1TB HDD with NTFS partition for VM games
So I with it's possible to use that NTFS partition in both VMs because it's would take several days to move it into QCOW and not sure what performance would be.

Last edited by SXX (2015-07-24 15:23:47)