For reference, we've also tried flipping the display on the Linux side of the system. The problem is that the hardware has a linked touch screen which takes the orientation from the Windows driver, so if the config on Linux doesn't hatch the Windows config then the touch screen isn't correctly aligned. This means we don't have any workaround for the issue without a QXL change. 

Thanks

Created attachment 1657831 [details]
Trial driver 2

Please try attached driver (comment #2)
Question: which mouse you use on VM? (command-line/profile)

Hi,
After a first test the Landscape/Portrait (flipped) configuration are available, but the mouse and the touchscreen are not mapped with the new flipped screen configuration.
I have tested either with a generic usb mouse and the touchpad. The mouse cursor is not flipped too.

Thanks,
Giuseppe

Is it correct that also with previous test where only 'Portrait' configuration present, the mouse/touch also do not function correctly in it?

Hi,
Yes in the previous test there were only the portrait and landscape configuration, but the mouse/touch worked correctly in accord with the used resolution.

In my experiments the pointing device never works correctly in any configuration except of plain Landscape.
Can you please specify which Spice client you use and which resolutions?

Hi,
These are the package version of the spice client/server:
spice-glib-0.35-7.el08.x86_64
spice-gtk3-0.35-7.el08.x86_64
spice-server-0.14.2-1.el08.x86_64
virt-viewer-7.0-3.el08.x86_64

My laptop has a 4K screen resolution but I have tested the VM using the 1024x768 screen resolution too and there is the same result.


Another test on the machine we need this functionality that has two displays there is another behavior: the touch on the second screen is acted on the first one.
To enable the two display we add two display adapters to the VM: 
    <video>
      <model type='qxl' ram='65536' vram='65536' vgamem='32768' heads='1' primary='yes'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x0'/>
    </video>
    <video>
      <model type='qxl' ram='65536' vram='65536' vgamem='32768' heads='1'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x03' function='0x0'/>
    </video>


is this a possible cause of this strange touchscreen behavior?

Created attachment 1658137 [details]
Image 1

Let's try to get synchronized. In comment #9 I attach typical picture of the screen (with qxldod driver) after the Portrait setting applied (using keyboard, as it is impossible to click the button with the mouse). Look at the pointer, its orientation (pointer shape is for 'Landscape', i.e. wrong) and highlighted 'Sound' captions (where the logical cursor is).
Do you have the same behavior on your system?

Yes, this is the current behavior on the laptop, but only if I use the touchpad, with the usb mouse the cursor is never shown. I'm using a full-screen kiosk mode config for the virt-viewer so the expected result is to do a click on the screen and act on the correct position.
On the other pc where there is a rotated screen the mouse cursor is never shown

Created attachment 1661663 [details]
Trial driver 3 (19003)

I've created additional build of the qxl-wddm-dod driver (not sure it will be useful).
It remaps the pointer for rotated screen (when the input device in VM is usb-mouse).
As far as I understand the case relevant for the product is when the input comes from the touch screen.
In this case the display driver is not in the game as it does not participate in the input path. In the best case (not always) it can only define where to draw the pointer.
Several questions (for this relevant case):
1. How the touch screen is defined on host side? (sudo evtest -> respective device)
2. How the input device is defined on VM side (command line / profile).

Created attachment 1663514 [details]
Touchscreen info

evtest logs and Windows system info

Hi,
The pointer is still with the original landscape orientation, so when I try to rotate the screen the cursor doesn't map the screen orientation, but this is not the priority for the moment.
In my laptop with a single monitor, the position of the touch seems to reflect the screen position, but not the click of the mouse: with landscape flipped configuration, a click on the bottom left is acted on the top right even if the mouse cursor is on the bottom left.
In general, there is a strange behavior with the touchscreen with some devices like my laptop: using the QXL drivers after a while the touchscreen stop to work, but using the Microsoft standard video drivers this never happens.


On the kiosk machine, the touch on the second monitor is acted on the first monitor, could be this related to video driver?

On Comment 15 I have attached the evtest logs, on Windows side what do you mean with "profile"? I have attached the system information

(In reply to Giuseppe from comment #16)
> Hi,
> The pointer is still with the original landscape orientation, so when I try
> to rotate the screen the cursor doesn't map the screen orientation, but this
> is not the priority for the moment.
> In my laptop with a single monitor, the position of the touch seems to
> reflect the screen position, but not the click of the mouse: with landscape
> flipped configuration, a click on the bottom left is acted on the top right
> even if the mouse cursor is on the bottom left.
> In general, there is a strange behavior with the touchscreen with some
> devices like my laptop: using the QXL drivers after a while the touchscreen
> stop to work, but using the Microsoft standard video drivers this never
> happens.
> 
> 
> On the kiosk machine, the touch on the second monitor is acted on the first
> monitor, could be this related to video driver?
> 

Please just confirm this is with Trial driver 2 or 3?

> On Comment 15 I have attached the evtest logs, on Windows side what do you
> mean with "profile"? I have attached the system information

The question was how the input device(s) appear in the machine XML or qemu command line.

Can you please also explain a little more about 2 displays configuration?
Are they duplicated or extended?
Additional question: do you use Spice agent (usually 'sc query vdservice') on the Guest?

> Please just confirm this is with Trial driver 2 or 3?

It's Trial driver 3 (Version 10.0.0.19003)


> The question was how the input device(s) appear in the machine XML or qemu command line.

The touchscreen is passed through to the VM like a normal USB device:
<hostdev mode='subsystem' type='usb'><source><vendor id='XXX'/><address bus='YYY' device='ZZZ'/><product id='TTT'/></source></hostdev>
I have added also this:
<input type='tablet' bus='usb'><address type='usb' bus='0' port='1'/></input>
to the domain xml definition


> Can you please also explain a little more about 2 displays configuration?
> Are they duplicated or extended?

They are extended. 
The VM is defined with two graphic cards:
    <video>
      <model type='qxl' ram='65536' vram='65536' vgamem='32768' heads='1' primary='yes'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x0'/>
    </video>
    <video>
      <model type='qxl' ram='65536' vram='65536' vgamem='32768' heads='1'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x03' function='0x0'/>
    </video>


> Additional question: do you use Spice agent (usually 'sc query vdservice') on the Guest?

No, I haven't used the spice agent on the VM

1. About USB mouse: I suggest to try <input type='mouse' ...> instead of 'tablet' - then the graphics driver is in the game.
2. About touch: If you just redirect the USB touch from the host, I'm afraid Spice components are not involved, this is IMO between the OS and the device.

Another note about touch screen: if you just redirect it from the host as USB device and have 2 displays, how can the guest OS decide to which display it is attached?
On the host this is probably described somehow in the ACPI or, for example, one screen is defined in the ACPI as integrated display panel.
If you change the 'main' screen or swap their placement - does it change?

Some news?

Hi,
Sorry for the delay, unfortunately changing tablet in mouse and switching the primary with secondary there is no change.
I have also tried to use two heads on the video device but the Windows VM can see only one monitor
    <video>
      <model type='qxl' ram='65536' vram='65536' vgamem='32768' heads='2' primary='yes'/>
      <address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x0'/>
    </video>

also using a type='vga' the VM can see only one monitor

Trying to summarize:
Notes: Multiple heads on single adapter are not supported by current driver under Windows.
AFAIK it is impossible to have multiple VGA adapters.
  
1. Problems with touch screen when 2 adapters defined: touch screen action is reflected on wrong display. The problem is clear, I do not see a simple solution, but I'm not an expert in this specific are (probably need to define somehow the binding between touch screen to specific device).
2. Problem with the mouse pointers: clicks in some of orientations take effect in wrong place. The problem is unclear, in case when the mouse is driven by host mouse and the VM input device is usb-mouse, the VM receives delta of mouse moves, sends the positions to the graphics driver and the driver 'Trial 3' should remap the mouse coordinates and request to draw the pointer at proper location (in case the VM input device is usb-tablet, the VM _receives the position_ of the pointer and the client draws the cursor where it thinks it is, the client is not aware of screen orientation). If solution for this problem required, we can cooperate and try to debug it.

Please correct me if I'm wrong.

Fixed in 0.20