In case the driver supports VSync control feature, it
maintains timer for VSync interrupt indication.
In further commits this timer will be started upon
class driver request. The interrupt notification and
related DPC do not access device registers.
Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Do not clear interrupt mask upon interrupt.
Instead clear pending interrupt status and write
QXL_IO_UPDATE_IRQ register (this drops interrupt level).
There are 3 advantages:
1. We do not need to wake the host to enable interrupt
mask in DPC procedure (1 wake per interrupt instead of 2)
2. The driver is not sensitive to failure when queues DPC, as
already queued DPC will process this interrupt when executed.
3. When we implement VSync interrupt simulation, we do not
need to touch registers neither when notify the OS nor when
process DPC related to this notification.
Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Concentrate filling of signal info in single procedure.
Fill signal info with specific or default frequency data
according to the global flag of VSync support.
Note that the state of this flobal flag must be defined only
on driver startup (based on OS version or any other information
available at DriverEntry) and can't be changed later.
Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This is a follow up of 31ca5d035d
("Implements screen to screen move correctly").
As now we don't copy memory to implement moves there's no
reason to take into account these area for memory mappings
as memory is not accessed anymore.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Yuri Benditovich <yuri.benditovich@daynix.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1411340
Do not set SupportNonVGA field for device rev.4 and higher.
Then the class driver will not call miniport's procedure
StopDeviceAndReleasePostDisplayOwnership upon PnP stop.
QXL device modes are not compatible with VGA ones; the
driver in this procedure returns display information related
to QXL mode which later used by BasicDisplay driver and causes
it to show the screen incorrectly or fail to show it at all.
Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Doing some fast check on Windows 8.1 you could note that
moving windows you got a weird effect were windows were a
bit misaligned.
As documented in DXGKARG_PRESENT_DISPLAYONLY page
NumMoves/pMoves fields point to an array of screen-to-screen
moves while our code implemented them as an image drawing
(the same implementation of dirty rects) causing the weird
effect mentioned.
This patch implement the moves using QXL_COPY_BITS operation
instead of a QXL_DRAW_COPY fixing the issue and avoiding sending
image to the server making the move/scroll operations on the
guest faster (and taking less bandwidth).
It seems that Windows 10 doesn't send the move commands but
instead send only dirty rects so you can't note this problem
using Windows 10.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Windows disables hw pointer acceleration and draws the cursor
in frame buffer if the driver does not report pointer capabilities
in QueryAdapterInfo. In this mode OS usually does not send pointer
commands to the driver or does it with pointer set to invisible
state. This mode of operation consumes CPU time and device-to-host
traffic. Now we enable pointer hardware acceleration.
Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
A more permissive licence is more in line with what Microsoft requires
for its WHQL certification
Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
This make sure that these function goes into the non
paged section.
Also this change is much shorter than marking code if
can be paged as almost code can be paged.
It's also coherent with other source files.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Sameeh Jubran <sameeh@daynix.com>
When qxl revision is 3, the vga driver is the one that is running.
While installing the driver the function VgaDevice::HWInit
has displayInfo structure zeroed out. The displayInfo should be
initialized using DxgkCbAcquirePostDisplayOwnership and thus it
should be called before calling HWInit.
Please note that we can't just move the call to "DxgkCbAcquirePostDisplayOwnership"
before calling HWInit as the m_Id isn't initialized for QxlDevice until the call to HWinit is over.
This patch fixes a bug similar to the one found here:
https://bugzilla.redhat.com/show_bug.cgi?id=1202267
However this one occurs when installing the driver.
Signed-off-by: Sameeh Jubran <sameeh@daynix.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Part of source image mapped by PresentDisplayOnly
should be big enough to cover all rectangles being
transferred.
Signed-off-by: Sameeh Jubran <sameeh@daynix.com>
Signed-off-by: Dmitry Fleytman <dmitry@daynix.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Framebuffer should only be used in VGA mode, as WDDM DOD
driver doesn't use the frame buffer (bar0), so no
reason to map in into memory. However the mode is only known
at runtime therefore framebuffer logic should be active when the
driver is operating in vga mode only.
There were rare BSOD failures when the mapping failed.
Signed-off-by: Sameeh Jubran <sameeh@daynix.com>
Signed-off-by: Dmitry Fleytman <dmitry@daynix.com>
This patch adds support for arbitrary resolution and updating
monitor configurations using custom ioctls.
This patch doesn't make any restriction for minimal resolution at all
as Windows can handle low resolution just fine, moreover the old (XPDM)
driver behaves like this too.
Based on a patch by Sandy Stutsman <sstutsma@redhat.com>
Signed-off-by: Sameeh Jubran <sameeh@daynix.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
As documented in MSDN, "DxgkDdiPresentDisplayOnly" should copy
each "move rectangle" from a given source point in the source buffer
to the destination. The current implementation doesn't take into account
the shift of the given source point at all. This patch fixies the issue.
Signed-off-by: Sameeh Jubran <sameeh@daynix.com>
Signed-off-by: Dmitry Fleytman <dmitry@daynix.com>
Acked-by: Frediano Ziglio <fziglio@redhat.com>
Some files had different line ending (some lines DOS while some others
UNIX). Use the same line ending using the nearest ending.
Signed-off-by: Frediano Ziglio <fziglio@redhat.com>
Acked-by: Christophe Fergeau <cfergeau@redhat.com>
