HVM pci passthrough impossible, need help

[mkru...@gmail.com]

Hello

I am having issues to do a successful PCI passthrough on a HVM VM i tried a lot of combinations of:
- host computers ( three different notebook PCs )
- host OS ( Windows 7 , Ubuntu )
- qubes versions ( Qubes R2 , Qubes R3.0 installation ISO with and without latest updates, Qubes R3.1 live USB )
- xen versions ( from R2 xen 4.1, from R3 4.4.2 and 4.4.3 )
- pci devices ( eSATA PCI, wifi pci, ethernet pci, usb hubs pci, audio pci )

I am sure i had IOMMU/VT-d enabled on all my computers that i tried.
I tried to do xl pci-attach/pci-detach
I followed any disscussions that had place on qubes forum regarding PCI passthrough and tried every thing that i could try in any of the discussions that i've seen,i tried rd.qubes.hide_pci, or different parameters on GRUB for xen.
I also checked dmesg logs, dm logs, libxl logs for any errors that would indicate this problem.
Sadly i couldn't find any errors related to any problem regarding ACPI, IOMMU, or PCI passthrough.

Also for any PCI passthrough that i tried it looks that for any PV VM works fine for any of the combinations that i tried. So no issues on PV VM.

I concluded that this problem is not OS related because i get the same issues on Windows or on Ubuntu installation so i continued to do extensive tests with a live Ubuntu USB memory stick.
To check if pci passthrough was successful i typed lspci inside ubuntu HVM, i also checked dmesg of ubuntu for any errors(found nothing)

One thing that i noticed on any combinations that i tried is that when i do a passthrough of USB bus of any of my notebooks tried i can see them with lspci inside Ubuntu. Any other devices that i tried does not show on lspci on any versions/computers that i tried.
The USB looks like is passing through however it does not work. When i tried lsusb, it hangs and no usb devices are displayed ever.

So logs are not complaining about everything, but pci devices are not seen or not working after passthrough.

Now i wonder if am i missing something? Are there any things that could be checked to find where the problem is and why the device won't reach the HVM VM?
Are there any tests that i can do?

[Marek Marczykowski-Górecki]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On Tue, Nov 17, 2015 at 12:33:50AM -0800, mkru...@gmail.com wrote:
> Hello
>
> I am having issues to do a successful PCI passthrough on a HVM VM i tried a lot of combinations of:
> - host computers ( three different notebook PCs )
> - host OS ( Windows 7 , Ubuntu )
> - qubes versions ( Qubes R2 , Qubes R3.0 installation ISO with and without latest updates, Qubes R3.1 live USB )
> - xen versions ( from R2 xen 4.1, from R3 4.4.2 and 4.4.3 )
> - pci devices ( eSATA PCI, wifi pci, ethernet pci, usb hubs pci, audio pci )
>
> I am sure i had IOMMU/VT-d enabled on all my computers that i tried.
> I tried to do xl pci-attach/pci-detach

Have you tried to use qvm-pci (or "Devices" tab in Qubes Manager),
_before_ starting the VM?

> I followed any disscussions that had place on qubes forum regarding PCI passthrough and tried every thing that i could try in any of the discussions that i've seen,i tried rd.qubes.hide_pci, or different parameters on GRUB for xen.
> I also checked dmesg logs, dm logs, libxl logs for any errors that would indicate this problem.
> Sadly i couldn't find any errors related to any problem regarding ACPI, IOMMU, or PCI passthrough.
>
> Also for any PCI passthrough that i tried it looks that for any PV VM works fine for any of the combinations that i tried. So no issues on PV VM.
>
> I concluded that this problem is not OS related because i get the same issues on Windows or on Ubuntu installation so i continued to do extensive tests with a live Ubuntu USB memory stick.
> To check if pci passthrough was successful i typed lspci inside ubuntu HVM, i also checked dmesg of ubuntu for any errors(found nothing)
>
> One thing that i noticed on any combinations that i tried is that when i do a passthrough of USB bus of any of my notebooks tried i can see them with lspci inside Ubuntu. Any other devices that i tried does not show on lspci on any versions/computers that i tried.
> The USB looks like is passing through however it does not work. When i tried lsusb, it hangs and no usb devices are displayed ever.
>
> So logs are not complaining about everything, but pci devices are not seen or not working after passthrough.
>
> Now i wonder if am i missing something? Are there any things that could be checked to find where the problem is and why the device won't reach the HVM VM?
> Are there any tests that i can do?
>

- --
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBCAAGBQJWSvOPAAoJENuP0xzK19csDc4H/As93NtVVdSJhKIW+PjQErNr
dXSiVUEEjjLHItkzIfOCDCC0dkLW1DZCN7IbCo2kPjqtWZtZCr+qd3m59an2tR/W
gN+5I+Wm9X37q+P67F5VBcaMp9bl7ZjvW+jI/ivA7Zv8l0fejfFv3hSMGfKbBhMn
8ynI94Rr4mw1KzxBcCsNvBCx8bd7RoiF5JylIIMWpy8pKO+B1emTEZd1dqujcfit
1f8U/BCiwWzNdL3/33Vzf7AVC84s1iDfqzX92qjcDb3Nr1skjYrlqzZTfHd26E86
kCaGv6yTZOk+h6SrOP1QUoKYou0P5f2bBUSZizUXQ/XEPLQ2uQdJ2DRgESgvxUM=
=+jha
-----END PGP SIGNATURE-----

[mkru...@gmail.com]

>
> Have you tried to use qvm-pci (or "Devices" tab in Qubes Manager),
> _before_ starting the VM?
>

I tried that. Same behaviour.

Can you confirm that you got running pci passthrough on any HVM VM by simply using "qvm-pci (or "Devices" tab in Qubes Manager)" on any of your machines?

[mkru...@gmail.com]

Good morning everyone.

Can anyone report a successfull PCI passthrough of any device on HVM VM in Qubes?

[Marek Marczykowski-Górecki]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

I'll try later today. I think I had it working on R3, but not so sure
now...

- --
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBCAAGBQJWTESpAAoJENuP0xzK19csOSMH/i7Ch6pr0tXe7fkfxHVs6K0V
268zk7YGj3Lnumrk7DCWUWeTsQFuDhQdHT8oVmuTq+yYB0ypN2mfX5awjtMkpa6H
b2iEQFdO6fKUOVimQSOMfDrmkteOc/NJfADGwE5zondtP+k9zKkrbNHH9SzcE6pi
XdecnyyZJ3HhWm+XU/2qFzmSmAvoesLYmbW+2aE3/15Fe56oKPkTCITxsShNcdZg
4U7QHDaS2uXdeHPOWmjiiLO8T/7CQerpHKqKi14CAwOwLmZgCoLYlLj0ULBS6oSF
lA2S7bYj/Cu5jeURsvHMNwBrKgx/IZtnt/v+BgMCAhK6OdIUWOK8sX45F+WMhz8=
=mR5u
-----END PGP SIGNATURE-----

[Eric Shelton]

On Wednesday, November 18, 2015 at 4:28:19 AM UTC-5, Marek Marczykowski-Górecki wrote:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On Tue, Nov 17, 2015 at 03:26:12AM -0800, mkru...@gmail.com wrote:
> >
> > Have you tried to use qvm-pci (or "Devices" tab in Qubes Manager),
> > _before_ starting the VM?
> >
> I tried that. Same behaviour.
>
> Can you confirm that you got running pci passthrough on any HVM VM by simply using "qvm-pci (or "Devices" tab in Qubes Manager)" on any of your machines?

I'll try later today. I think I had it working on R3, but not so sure
now...

I do not have firsthand experiences to report, but I believe the following provide some examples:

(1) https://groups.google.com/forum/#!msg/qubes-users/cmPRMOkxkdA/gIV68O0-CQAJ

discusses, under R3 rc2, successful PCI passthrough of a GPU, an audio device, and an ethernet controller to Windows.  However, the author reported being unable to get it to work using libvirt, and instead having to use xl to bring up the domain.

(2) https://groups.google.com/d/msg/qubes-users/IRgJkFpJe4g/GfVw-JurGCQJ

Appears to indicate successful passthrough of a USB2 controller to Windows.

Eric

[mkru...@gmail.com]

>
> I do not have firsthand experiences to report, but I believe the following provide some examples:
>
>
> (1) https://groups.google.com/forum/#!msg/qubes-users/cmPRMOkxkdA/gIV68O0-CQAJ
>
>
> discusses, under R3 rc2, successful PCI passthrough of a GPU, an audio device, and an ethernet controller to Windows.  However, the author reported being unable to get it to work using libvirt, and instead having to use xl to bring up the domain.

I'm aware of this thread. And i already tried exactly as in that post. Bringing up domain trough xl only leaded to same behaviour as in Qubes VM Manager

>
>
> (2) https://groups.google.com/d/msg/qubes-users/IRgJkFpJe4g/GfVw-JurGCQJ
>
>
> Appears to indicate successful passthrough of a USB2 controller to Windows.
>
>
>
>
> Eric

This second post says that it both failed and succeded in some case. Indeed the USB is seen on HVM both on Windown 7 and Ubuntu but it cannot list the devices. So i think hits the same issue as i did now.

I'm stuck on this with no errors in my logs that could lead to a solution on all the machines that i tried.

[mkru...@gmail.com]

>
>
> (1) https://groups.google.com/forum/#!msg/qubes-users/cmPRMOkxkdA/gIV68O0-CQAJ
>
>
> discusses, under R3 rc2, successful PCI passthrough of a GPU, an audio device, and an ethernet controller to Windows.  However, the author reported being unable to get it to work using libvirt, and instead having to use xl to bring up the domain.
>
>

I see this guy succeded on a desktop computer. All that i tried where laptops. I wonder if there is something about the hardware on the laptopts that makes them different in such a way that they don't work correctly on Xen. At least anything i tried are relatively new laptops.
I will try to grab some desktop computer from some friend or test on an older laptop model.
Sadly. If it still works it still doesn't solve the current issue.

[Marek Marczykowski-Górecki]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On Wed, Nov 18, 2015 at 10:28:09AM +0100, Marek Marczykowski-Górecki wrote:
> On Tue, Nov 17, 2015 at 03:26:12AM -0800, mkru...@gmail.com wrote:
> > >
> > > Have you tried to use qvm-pci (or "Devices" tab in Qubes Manager),
> > > _before_ starting the VM?
> > >
> > I tried that. Same behaviour.
> >
> > Can you confirm that you got running pci passthrough on any HVM VM by simply using "qvm-pci (or "Devices" tab in Qubes Manager)" on any of your machines?
>
> I'll try later today. I think I had it working on R3, but not so sure
> now...

Ok, I've tried, received no errors. And no device in the HVM...

- --
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBCAAGBQJWTOHZAAoJENuP0xzK19cs+KgH/0/SnX+8O93SPrLKQTQdSfbV
s9loN431AUbATeOgne75cBRsLW9HROTPxgem+fleemOwxEidZ1u+bJOQdcsssuhI
oYwO9TGu+kkW4brpXCJKs04wSDuOYA3+FxoPe9z9LxNfAO1RdC6uOk2T4huvh2W8
fQv7B+GKzVi0U9AP0+Cus0y6iPkp1I8aHQVG7l8ND67e7VPD55yejEz63LQuITI5
QVDuOguSf6iVmDYztBnR4HilPjIjhp+zgobfWXVXl6RfAYSDT6g9st2T+As5cl/1
elAAgIRFMiO1M7l43ZYce3ahHFAdwiVNRFXE4exf1Zlhl2nLL+iHMXDix4EMOzA=
=Obds
-----END PGP SIGNATURE-----

[mkru...@gmail.com]

On Wednesday, November 18, 2015 at 10:39:04 PM UTC+2, Marek Marczykowski-Górecki wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
>
> On Wed, Nov 18, 2015 at 10:28:09AM +0100, Marek Marczykowski-Górecki wrote:

> > On Tue, Nov 17, 2015 at 03:26:12AM -0800, mkr....@gmail.com wrote:
> > > >
> > > > Have you tried to use qvm-pci (or "Devices" tab in Qubes Manager),
> > > > _before_ starting the VM?
> > > >
> > > I tried that. Same behaviour.
> > >
> > > Can you confirm that you got running pci passthrough on any HVM VM by simply using "qvm-pci (or "Devices" tab in Qubes Manager)" on any of your machines?
> >
> > I'll try later today. I think I had it working on R3, but not so sure
> > now...
>
> Ok, I've tried, received no errors. And no device in the HVM...
>
> - --
> Best Regards,
> Marek Marczykowski-Górecki
> Invisible Things Lab
> A: Because it messes up the order in which people normally read text.
> Q: Why is top-posting such a bad thing?
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1
>
> iQEcBAEBCAAGBQJWTOHZAAoJENuP0xzK19cs+KgH/0/SnX+8O93SPrLKQTQdSfbV
> s9loN431AUbATeOgne75cBRsLW9HROTPxgem+fleemOwxEidZ1u+bJOQdcsssuhI
> oYwO9TGu+kkW4brpXCJKs04wSDuOYA3+FxoPe9z9LxNfAO1RdC6uOk2T4huvh2W8
> fQv7B+GKzVi0U9AP0+Cus0y6iPkp1I8aHQVG7l8ND67e7VPD55yejEz63LQuITI5
> QVDuOguSf6iVmDYztBnR4HilPjIjhp+zgobfWXVXl6RfAYSDT6g9st2T+As5cl/1
> elAAgIRFMiO1M7l43ZYce3ahHFAdwiVNRFXE4exf1Zlhl2nLL+iHMXDix4EMOzA=
> =Obds
> -----END PGP SIGNATURE-----

Hi

Are there any places to turn on extra debugging?
Or any ideas?

[Marek Marczykowski-Górecki]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On Wed, Nov 18, 2015 at 11:18:16PM -0800, mkru...@gmail.com wrote:
> On Wednesday, November 18, 2015 at 10:39:04 PM UTC+2, Marek Marczykowski-Górecki wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA256
> >
> > On Wed, Nov 18, 2015 at 10:28:09AM +0100, Marek Marczykowski-Górecki wrote:
> > > On Tue, Nov 17, 2015 at 03:26:12AM -0800, mkr....@gmail.com wrote:
> > > > >
> > > > > Have you tried to use qvm-pci (or "Devices" tab in Qubes Manager),
> > > > > _before_ starting the VM?
> > > > >
> > > > I tried that. Same behaviour.
> > > >
> > > > Can you confirm that you got running pci passthrough on any HVM VM by simply using "qvm-pci (or "Devices" tab in Qubes Manager)" on any of your machines?
> > >
> > > I'll try later today. I think I had it working on R3, but not so sure
> > > now...
> >
> > Ok, I've tried, received no errors. And no device in the HVM...
>

> Hi
>
> Are there any places to turn on extra debugging?
> Or any ideas?

Basic things looks ok (backend state, frontend in stubdomain, no errors
from qemu etc). Also dynamic attach (xl pci-attach) returns no errors,
but still no device in the VM. I've tried only Windows HVM. I guess it
is some problem with qemu in stubdomain, but that would be strange,
because qemu is mostly the same as in R2 - where it was working...

Will look more closely somehow tomorrow.

- --
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBCAAGBQJWTZYwAAoJENuP0xzK19csAE4H/0TrztoNi0ieLtqe+9QIg1tF
VizPfV3qhxC75uUKB5HxIcQUVGVdUoO0rlUpe30/WNLgOWOELC8YTb0ZXLTBVlyP
NoPW30qWTnMB5H03rSB8dcF3ZzDcP1/yauTmRzNON9FvJpoSs6JKqimJl9VFMTBx
W7LVgFhJ1CQHnQzlamBrZgNRTc/0HC0G4j21oH67NzC8CpXEYDdIkAc/HfIKG85o
oWMm0uP5O06U22N1SwP0rHcTnPyJX33sSoSnZ9fkmZ3T7kwG5aJYPX8ELuMCYy/t
S9Z8rqgdnW7QFqnAy/3F75/iE59EFGduUjmev0//tnKvDhW5KH+fRyMV8ABaSiU=
=zT1A
-----END PGP SIGNATURE-----

[Eric Shelton]

I don't know if this could be coming into play here, but a couple of months ago I came an item where another team appears to have found a PCI interrupt remapping bug that could be lurking in qemu-traditional:

https://groups.google.com/d/msg/qubes-users/cmPRMOkxkdA/2fHESzR1KAAJ

Eric

[Marek Marczykowski-Górecki]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On Thu, Nov 19, 2015 at 01:45:53PM -0800, Eric Shelton wrote:
> On Thursday, November 19, 2015 at 4:28:26 AM UTC-5, Marek
> Marczykowski-Górecki wrote:
> >
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA256
> >

> > On Wed, Nov 18, 2015 at 11:18:16PM -0800, mkru...@gmail.com <javascript:>

I don't see this error in logs, but the patch attached there isn't
applied to our qemu version. Will try that and report back :)

- --
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBCAAGBQJWTk92AAoJENuP0xzK19csYcIH/jnreJqsnEnu/5HU+wFOYrAg
7b5awwCZ7nT8JDGDdcWkgnVqyQ8rv39t6k3cldUprFiV8J3pP7SUHjWAa+Ft2Em0
o4C+SSbii7XuB2NfvjXwS9BWaLyJCUQuB+tQ5znMXiWXjePW2Wl67yCUhlpIMts1
aSUQG8EBuFLieRzMGyQah6BnsWDQtCNCb4ky6Y4+SYCKX4W2bYNs27LDguzsEp76
wcAytfcnFNNKwGE1BkxluopTl8sLpz+rSADwdeUAVjm7cBrVeHwZujGr8tDfMqrs
yXkMYz2mYupPmksEmpVbe7n+ZOaH01vly/sSZcbKwwOKIr8/jIVDSM1Ocq5JtSk=
=JaBI
-----END PGP SIGNATURE-----

[mkru...@gmail.com]

On Thursday, November 19, 2015 at 11:45:53 PM UTC+2, Eric Shelton wrote:
> I don't know if this could be coming into play here, but a couple of months ago I came an item where another team appears to have found a PCI interrupt remapping bug that could be lurking in qemu-traditional:
>
>
> https://groups.google.com/d/msg/qubes-users/cmPRMOkxkdA/2fHESzR1KAAJ
>
>
>
> Eric

Hi everyone

I don't know if this has something to do or not with the issue but it might help.

What i noticed is that when running Ubuntu HVM on Qubes VM Manager with USB device attached is that when i do "xl list" i notice two domains "ubuntu" and "ubuntu-dm"(stub domain). Searching through the ubuntu-dm.log i couldn't find any IRQ error messages.
Also calling lsusb inside VM hangs.

In one of the another test that i did to find a trace of this problem was running my Ubuntu HVM installation from xl running the following commands:

xl pci-assignable-add 0000:00:14.0
xl mem-set dom0 1024
xl create ubuntu2.hvm

Here is the .hvm file i used:

builder = "hvm"
name = "ubuntu2"
memory = 1024
acpi = 1
apic = 1
on_xend_stop = 'shutdown'
on_poweroff = 'destroy'
on_reboot = 'destroy'
on_crash = 'destroy'
vif = [ 'mac=00:16:3e:5e:6d:11,ip=10.137.2.22,script=vif-route-qubes,backend=sys-firewall' ]
disk = [ '/var/lib/qubes/appvms/ubuntu/root.img,raw,hda,rw' ]
pci = [ '00:14.0' ]

device_model_version = 'qemu-xen-traditional'
boot = 'dc'
sdl = 1
keymap = 'en-us'
stdvga = 0

serial='pty'
tsc_mode = 'default'
usb = 1
usbdevice = 'tablet'
localtime = 1
xen_platform_pci = 1
pci_power_mgmt = 1
pci_msitranslate = 1
monitor = 1
hpet = 1

In my ubuntu2.hvm file as you see i put "device_model_version = 'qemu-xen-traditional'" which yields on "xl list" only one domain called ubuntu2.
Which tells me that in this way, stubdomain doesn't run as a separate VM as it did on Qubes VM Manager (i know it's not safe to run it like this but this is just for test purposes only)
It also didn't create any ubuntu2-dm.log file this way.
Running this way all my tests where the same: USB could be seen but not working, the other devices simply doesn't show at all in HVM VM

So this test might tell that this is not a stubdomain problem. I might be right or wrong because i don't know how these things works.

[Eric Shelton]

Marek,

A couple of related questions: 

(1) Is PCI passthrough to HVM not working on R3?  That seems to be my experience trying to pass a device through to a HVM running Linux, but I just want to confirm this is what seems to be going on.

(2) Were you able to work out what is going on since the 19th?  I would guess it is libvirt related, just based on the various holes it has in its Xen support.

Eric

[marmarek]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On Sat, Nov 28, 2015 at 12:06:35PM -0800, Eric Shelton wrote:
> On Thursday, November 19, 2015 at 4:28:26 AM UTC-5, Marek
> Marczykowski-Górecki wrote:
> >
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA256
> >

> > On Wed, Nov 18, 2015 at 11:18:16PM -0800, mkru...@gmail.com <javascript:>

It looks like so.

> (2) Were you able to work out what is going on since the 19th? I would
> guess it is libvirt related, just based on the various holes it has in its
> Xen support.

Sorry, haven't time for that...

- --
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
-----BEGIN PGP SIGNATURE-----

Version: GnuPG v2

iQEcBAEBCAAGBQJWW3yLAAoJENuP0xzK19csEvgH/3jFg89kpB4DE895IkuAZFFs
+PRE3UWLXWh0z4yTaX1xWJt+Xo+k+kBL4GQ/YP4XAeAWbrxMM1n6UVGViAOXq0pQ
boy5H5oJEDph87qxhQVs/fZ6xEYB5Vb4fBRsKT0xcgKOMKZRPRL2ZRqCRWG7kuSp
J9QWXlR461qtQ2IG1h+d1DJdXgIkmefec140ECNI15B3MjWsFzFkUsvTdn8q5ugO
Iu3vFPyrCvKuBSYz76LZtkyov9KJN7PDqhkBamhkRsfFlOOHZTzG83S8vKaup1nl
xboUZiWDhkau/BzD/Ncnc6+Dh3drs+Z6IrfPDnuDhA3LoPRfQZ5Fg8Z/q9EB5Qk=
=Cl6Q
-----END PGP SIGNATURE-----

[df.g...@gmail.com]

Same problem here. Windows HVM is up and running but can't get a USB License to activate my software.

I looked around for possible fixes for pci passthrough but none of them seem to work. The best I could do for external media was attach a block device but this doesn't give any access to the license stored on the USB dongle.

Qubes R3.0 seems to have some issues with this.

When attempting to add Devices:

qvm-prefs -s HVM-name pci_strictreset true
1.) Crashes the VM Manager
2.) Crashes the entire system
3.) Error messages
"Error starting VM: internal error: Unable to reset PCI device XXXXXXXXX: no FLR, PM reset or bus reset available"
4.) Can attach block devices

qvm-prefs -s HVM-name pci_strictreset false
1.) Able to boot HVM but no access to device
2.) Lost ability to attach block devices

[Shilu Pilu]

any news on pci-passthrough in hvm win7 machines?

[mkru...@gmail.com]

As i noticed on the other threads you guys got Xen 4.6 working Qubes. Can anyone check to passthrough any pci device into a HVM to see if is working?

[tra...@gmail.com]

On Wednesday, 2 December 2015 11:05:01 UTC+2, mkru...@gmail.com wrote:
> As i noticed on the other threads you guys got Xen 4.6 working Qubes. Can anyone check to passthrough any pci device into a HVM to see if is working?

Qubes R3.1, Xen 4.6

I can assign a PCI device (USB controller) to an AppVM just fine, but it does not show up when I do it to a debian-based HVM. No errors starting up the VMs.