[Qemu-devel] Regarding Intel IGD passthru support for QEMU/KVM

[Qemu-devel] Regarding Intel IGD passthru support for QEMU/KVM

[Raghavan Santhanam]

[-- Attachment #1: Type: text/plain, Size: 355 bytes --]

Hi,

Based on the Intel IGD passthru support that has been added to Qemu/Xen
code base,
is there any way to use/reuse the same logic currently to have a successful
passthru of an Intel
IGD with Qemu/KVM on a Linux host(Ubuntu x86_64) or will that require some
more
work in addition to what Xen code based already has for the IGD passthru?

Best,
Raghavan

[-- Attachment #2: Type: text/html, Size: 495 bytes --]

Re: [Qemu-devel] Regarding Intel IGD passthru support for QEMU/KVM

[Gerd Hoffmann]

On So, 2016-01-17 at 18:15 -0800, Raghavan Santhanam wrote:
> Hi,
> 
> 
> Based on the Intel IGD passthru support that has been added to
> Qemu/Xen code base,
> is there any way to use/reuse the same logic currently to have a
> successful passthru of an Intel
> IGD with Qemu/KVM on a Linux host(Ubuntu x86_64) or will that require
> some more
> work in addition to what Xen code based already has for the IGD
> passthru?

It's in progress.  The xen code in the code base seems to have problems
though, recent testing showed that it'll only work with recent linux
kernels.  Older linux kernels and windows guest drivers do not work
correctly.  So this is not just about enabling kvm support, but also
about figuring why things are not working correctly on xen ...

cheers,
  Gerd

Re: [Qemu-devel] Regarding Intel IGD passthru support for QEMU/KVM

[Raghavan Santhanam]

[-- Attachment #1: Type: text/plain, Size: 1144 bytes --]

Hi Gerd,

Okay, thanks for the update.

By the way, it's great to see the commitment of Qemu developers in the
open-source community. Appreciate your entire team's effort right from
the the inception of QEMU by Fabrice Bellard!

Best,
Raghavan

On Mon, Jan 18, 2016 at 3:18 AM, Gerd Hoffmann <kraxel@redhat.com> wrote:

> On So, 2016-01-17 at 18:15 -0800, Raghavan Santhanam wrote:
> > Hi,
> >
> >
> > Based on the Intel IGD passthru support that has been added to
> > Qemu/Xen code base,
> > is there any way to use/reuse the same logic currently to have a
> > successful passthru of an Intel
> > IGD with Qemu/KVM on a Linux host(Ubuntu x86_64) or will that require
> > some more
> > work in addition to what Xen code based already has for the IGD
> > passthru?
>
> It's in progress.  The xen code in the code base seems to have problems
> though, recent testing showed that it'll only work with recent linux
> kernels.  Older linux kernels and windows guest drivers do not work
> correctly.  So this is not just about enabling kvm support, but also
> about figuring why things are not working correctly on xen ...
>
> cheers,
>   Gerd
>
>
>

[-- Attachment #2: Type: text/html, Size: 1695 bytes --]

Re: [Qemu-devel] Regarding Intel IGD passthru support for QEMU/KVM

[Gerd Hoffmann]

  Hi,

> Based on the Intel IGD passthru support that has been added to
> Qemu/Xen code base, is there any way to use/reuse the same logic
> currently to have a successful passthru of an Intel IGD with Qemu/KVM
> on a Linux host(Ubuntu x86_64) or will that require some more
> work in addition to what Xen code based already has for the IGD
> passthru?

If you feel a bit adventurous you can try whenever current git master
(or 4.5-rc1 when Linus releases it next weekend) as guest kernel works
for you.

Recent linux kernels seem to not require the host bridge tweaks any
more.  A patch to recognize the qemu q35 south bridge has been added in
the merge window.

So using the q35 machine type + vfio passthrough without any special igd
tweaks has a fair chance to work.  opregion support isn't there yet, so
don't expect laptop panels work correctly.

cheers,
  Gerd

Re: [Qemu-devel] Regarding Intel IGD passthru support for QEMU/KVM

[Raghavan Santhanam]

[-- Attachment #1: Type: text/plain, Size: 1727 bytes --]

Hi Gerd,

Thanks for the update.

Actually, I have already got Intel IGD passthru to work couple of days back
with git commit being
at f02ccf53693758b65843264e077f90cf295e7d98(disas/libvixl: Really suppress
gcc 4.6.3 sign-compare warnings).

As far as Qemu is concerned, I am using default machine. I just had to
modify the creation of PCI
devices code a little. And also the ISA bridge code. Host uses 3.*
kernel(yes, it's quite old!). Guest is
using 4.* kernel. Everything is working without issues. By the way, in my
case, Intel IGD is the primary
VGA. So, it has been challenging to accomplish this task since a very long
period of time.

And guest I am using is Android-x86(5.1/Lollipop). Updates to follow . . .

Best,
Raghavan





On Wed, Jan 20, 2016 at 2:30 AM, Gerd Hoffmann <kraxel@redhat.com> wrote:

>   Hi,
>
> > Based on the Intel IGD passthru support that has been added to
> > Qemu/Xen code base, is there any way to use/reuse the same logic
> > currently to have a successful passthru of an Intel IGD with Qemu/KVM
> > on a Linux host(Ubuntu x86_64) or will that require some more
> > work in addition to what Xen code based already has for the IGD
> > passthru?
>
> If you feel a bit adventurous you can try whenever current git master
> (or 4.5-rc1 when Linus releases it next weekend) as guest kernel works
> for you.
>
> Recent linux kernels seem to not require the host bridge tweaks any
> more.  A patch to recognize the qemu q35 south bridge has been added in
> the merge window.
>
> So using the q35 machine type + vfio passthrough without any special igd
> tweaks has a fair chance to work.  opregion support isn't there yet, so
> don't expect laptop panels work correctly.
>
> cheers,
>   Gerd
>
>

[-- Attachment #2: Type: text/html, Size: 2429 bytes --]

Re: [Qemu-devel] Regarding Intel IGD passthru support for QEMU/KVM

[Raghavan Santhanam]

[-- Attachment #1: Type: text/plain, Size: 2209 bytes --]

Hi All,

As I mentioned, here is the Intel IGD Passthru in action, that's fully
working without any issues, with QEMU/KVM running hardware-accelerated
Android-x86 5.1/Lollipop : https://www.youtube.com/watch?v=PBN1DmarJ7k

Thank you all once again for the wonderful work on QEMU/KVM/XEN since many
years!

Best,
Raghavan


On Wed, Jan 20, 2016 at 9:31 AM, Raghavan Santhanam <raghavanil4m@gmail.com>
wrote:

> Hi Gerd,
>
> Thanks for the update.
>
> Actually, I have already got Intel IGD passthru to work couple of days
> back with git commit being
> at f02ccf53693758b65843264e077f90cf295e7d98(disas/libvixl: Really suppress
> gcc 4.6.3 sign-compare warnings).
>
> As far as Qemu is concerned, I am using default machine. I just had to
> modify the creation of PCI
> devices code a little. And also the ISA bridge code. Host uses 3.*
> kernel(yes, it's quite old!). Guest is
> using 4.* kernel. Everything is working without issues. By the way, in my
> case, Intel IGD is the primary
> VGA. So, it has been challenging to accomplish this task since a very long
> period of time.
>
> And guest I am using is Android-x86(5.1/Lollipop). Updates to follow . . .
>
> Best,
> Raghavan
>
>
>
>
>
> On Wed, Jan 20, 2016 at 2:30 AM, Gerd Hoffmann <kraxel@redhat.com> wrote:
>
>>   Hi,
>>
>> > Based on the Intel IGD passthru support that has been added to
>> > Qemu/Xen code base, is there any way to use/reuse the same logic
>> > currently to have a successful passthru of an Intel IGD with Qemu/KVM
>> > on a Linux host(Ubuntu x86_64) or will that require some more
>> > work in addition to what Xen code based already has for the IGD
>> > passthru?
>>
>> If you feel a bit adventurous you can try whenever current git master
>> (or 4.5-rc1 when Linus releases it next weekend) as guest kernel works
>> for you.
>>
>> Recent linux kernels seem to not require the host bridge tweaks any
>> more.  A patch to recognize the qemu q35 south bridge has been added in
>> the merge window.
>>
>> So using the q35 machine type + vfio passthrough without any special igd
>> tweaks has a fair chance to work.  opregion support isn't there yet, so
>> don't expect laptop panels work correctly.
>>
>> cheers,
>>   Gerd
>>
>>
>

[-- Attachment #2: Type: text/html, Size: 3331 bytes --]

Re: [Qemu-devel] Regarding Intel IGD passthru support for QEMU/KVM

[Raghavan Santhanam]

[-- Attachment #1: Type: text/plain, Size: 2828 bytes --]

Hi All,

In an effort to *validate* my successful Intel IGD passthru with
QEMU/KVM, more completely, I attempted
two of the other popular available releases as of now : Ubuntu 15.10 64-bit
and Windows 10 LTSB 64-bit and
they work smoothly.

Here are their demo videos :
*Ubuntu 15.10 64-bit as the guest* :
https://www.youtube.com/watch?v=1_VXTDgqnHg
*Windows 10 LTSB 64-bit as guest* :
https://www.youtube.com/watch?v=WmKzRGT2cjI

Cheers,
Raghavan

On Sat, Jan 23, 2016 at 3:08 PM, Raghavan Santhanam <raghavanil4m@gmail.com>
wrote:

> Hi All,
>
> As I mentioned, here is the Intel IGD Passthru in action, that's fully
> working without any issues, with QEMU/KVM running hardware-accelerated
> Android-x86 5.1/Lollipop : https://www.youtube.com/watch?v=PBN1DmarJ7k
>
> Thank you all once again for the wonderful work on QEMU/KVM/XEN since many
> years!
>
> Best,
> Raghavan
>
>
> On Wed, Jan 20, 2016 at 9:31 AM, Raghavan Santhanam <
> raghavanil4m@gmail.com> wrote:
>
>> Hi Gerd,
>>
>> Thanks for the update.
>>
>> Actually, I have already got Intel IGD passthru to work couple of days
>> back with git commit being
>> at f02ccf53693758b65843264e077f90cf295e7d98(disas/libvixl: Really suppress
>> gcc 4.6.3 sign-compare warnings).
>>
>> As far as Qemu is concerned, I am using default machine. I just had to
>> modify the creation of PCI
>> devices code a little. And also the ISA bridge code. Host uses 3.*
>> kernel(yes, it's quite old!). Guest is
>> using 4.* kernel. Everything is working without issues. By the way, in my
>> case, Intel IGD is the primary
>> VGA. So, it has been challenging to accomplish this task since a very
>> long period of time.
>>
>> And guest I am using is Android-x86(5.1/Lollipop). Updates to follow . . .
>>
>> Best,
>> Raghavan
>>
>>
>>
>>
>>
>> On Wed, Jan 20, 2016 at 2:30 AM, Gerd Hoffmann <kraxel@redhat.com> wrote:
>>
>>>   Hi,
>>>
>>> > Based on the Intel IGD passthru support that has been added to
>>> > Qemu/Xen code base, is there any way to use/reuse the same logic
>>> > currently to have a successful passthru of an Intel IGD with Qemu/KVM
>>> > on a Linux host(Ubuntu x86_64) or will that require some more
>>> > work in addition to what Xen code based already has for the IGD
>>> > passthru?
>>>
>>> If you feel a bit adventurous you can try whenever current git master
>>> (or 4.5-rc1 when Linus releases it next weekend) as guest kernel works
>>> for you.
>>>
>>> Recent linux kernels seem to not require the host bridge tweaks any
>>> more.  A patch to recognize the qemu q35 south bridge has been added in
>>> the merge window.
>>>
>>> So using the q35 machine type + vfio passthrough without any special igd
>>> tweaks has a fair chance to work.  opregion support isn't there yet, so
>>> don't expect laptop panels work correctly.
>>>
>>> cheers,
>>>   Gerd
>>>
>>>
>>
>

[-- Attachment #2: Type: text/html, Size: 4647 bytes --]

Re: [Qemu-devel] Regarding Intel IGD passthru support for QEMU/KVM

[Raghavan Santhanam]

[-- Attachment #1: Type: text/plain, Size: 5038 bytes --]

Hi All,

Here is my most recent successful and more complete Android Sensor
Emulation(Intel-IGD-passsthru ~ QEMU/KVM) update by me.

Watch it in action : https://www.youtube.com/watch?v=vdcOS247F-4

**== Android Sensor Emulation, fully UNLEASHED! ==**

 - Fueled by GoogleNexus5
 - Powered by Intel IGD
 - Run on QEMU/KVM

**== Android Sensor Emulation, fully UNLEASHED!! ==**

 - Turn your Android Smartphone into a JOYSTICK!



 - Play. Every. Single. Motion-sensor game. On. Any. Sensor-less Android
   device.
 - Run. Every. Single. sensor-based application. On. Any. Sensor-less
   Android device.



 - Let your games and other applications be "unique selves". No need to
   alter your "favorite" game or "useful" other sensor-based application
   to make use of my Android Sensor Emulation support.

**== DETAILS ==**

My current work as shown by this demo emulates the below 27 sensors(all the
sensors of Google Nexus 5). Watch the video for more and precise details.

---------------------------------------------------------------------------------------------------------
1. Accelerometer
2. Light
3. Proximity
4. Magnetometer
5. Gyroscope
6. Gravity
7. Linear Acceleration
8. Orientation
9. Rotation Vector

---------------------------------------------------------------------------------------------------------
10. Game Rotation Vector
11. Barometer
12. Step Detector
13. Step Counter
14. Significant Motion Detector
15. Tilt Detector
16. Absolute Motion Detector(AMD)
17. Relative Motion Detector(RMD)
18. Basic Gestures

---------------------------------------------------------------------------------------------------------
19. Tap Detector
20. Facing Detector
21. (QTI) Tilt Detector
22. Pedometer
23. Pedestrian Activity Monitor
24. Geomagnetic Rotation Vector
25. (InvenSense) Game Rotation Vector
26. Magnetometer (Uncalibrated)
27. Gyroscope (Uncalibrated)

---------------------------------------------------------------------------------------------------------

Cheers,
Raghavan

On Tue, Feb 9, 2016 at 7:00 AM, Raghavan Santhanam <raghavanil4m@gmail.com>
wrote:

> Hi All,
>
> In an effort to *validate* my successful Intel IGD passthru with
> QEMU/KVM, more completely, I attempted
> two of the other popular available releases as of now : Ubuntu 15.10
> 64-bit and Windows 10 LTSB 64-bit and
> they work smoothly.
>
> Here are their demo videos :
> *Ubuntu 15.10 64-bit as the guest* :
> https://www.youtube.com/watch?v=1_VXTDgqnHg
> *Windows 10 LTSB 64-bit as guest* :
> https://www.youtube.com/watch?v=WmKzRGT2cjI
>
> Cheers,
> Raghavan
>
> On Sat, Jan 23, 2016 at 3:08 PM, Raghavan Santhanam <
> raghavanil4m@gmail.com> wrote:
>
>> Hi All,
>>
>> As I mentioned, here is the Intel IGD Passthru in action, that's fully
>> working without any issues, with QEMU/KVM running hardware-accelerated
>> Android-x86 5.1/Lollipop : https://www.youtube.com/watch?v=PBN1DmarJ7k
>>
>> Thank you all once again for the wonderful work on QEMU/KVM/XEN since
>> many years!
>>
>> Best,
>> Raghavan
>>
>>
>> On Wed, Jan 20, 2016 at 9:31 AM, Raghavan Santhanam <
>> raghavanil4m@gmail.com> wrote:
>>
>>> Hi Gerd,
>>>
>>> Thanks for the update.
>>>
>>> Actually, I have already got Intel IGD passthru to work couple of days
>>> back with git commit being
>>> at f02ccf53693758b65843264e077f90cf295e7d98(disas/libvixl: Really suppress
>>> gcc 4.6.3 sign-compare warnings).
>>>
>>> As far as Qemu is concerned, I am using default machine. I just had to
>>> modify the creation of PCI
>>> devices code a little. And also the ISA bridge code. Host uses 3.*
>>> kernel(yes, it's quite old!). Guest is
>>> using 4.* kernel. Everything is working without issues. By the way, in
>>> my case, Intel IGD is the primary
>>> VGA. So, it has been challenging to accomplish this task since a very
>>> long period of time.
>>>
>>> And guest I am using is Android-x86(5.1/Lollipop). Updates to follow . .
>>> .
>>>
>>> Best,
>>> Raghavan
>>>
>>>
>>>
>>>
>>>
>>> On Wed, Jan 20, 2016 at 2:30 AM, Gerd Hoffmann <kraxel@redhat.com>
>>> wrote:
>>>
>>>>   Hi,
>>>>
>>>> > Based on the Intel IGD passthru support that has been added to
>>>> > Qemu/Xen code base, is there any way to use/reuse the same logic
>>>> > currently to have a successful passthru of an Intel IGD with Qemu/KVM
>>>> > on a Linux host(Ubuntu x86_64) or will that require some more
>>>> > work in addition to what Xen code based already has for the IGD
>>>> > passthru?
>>>>
>>>> If you feel a bit adventurous you can try whenever current git master
>>>> (or 4.5-rc1 when Linus releases it next weekend) as guest kernel works
>>>> for you.
>>>>
>>>> Recent linux kernels seem to not require the host bridge tweaks any
>>>> more.  A patch to recognize the qemu q35 south bridge has been added in
>>>> the merge window.
>>>>
>>>> So using the q35 machine type + vfio passthrough without any special igd
>>>> tweaks has a fair chance to work.  opregion support isn't there yet, so
>>>> don't expect laptop panels work correctly.
>>>>
>>>> cheers,
>>>>   Gerd
>>>>
>>>>
>>>
>>
>

[-- Attachment #2: Type: text/html, Size: 7957 bytes --]