Summary:
With nouveau driver it can happen that KWin gets frozen when first trying
to render with OpenGL. This results in a freeze of the complete desktop
as the compositor is non functional.
Our OpenGL breakage detection is only able to detect crashes, but not
freezes. This change improves it by also added a freeze protection.
In the PreInit stage a thread is started with a QTimer of 15 sec. If the
timer fires, qFatal is triggered to terminate KWin. This can only happen
if the creation of the OpenGL compositor takes longer than said 15 sec.
In the PostInit stage the timer gets deleted and the thread stopeed
again.
Thus if a freeze is detected the OpenGL unsafe protection is written into
the config. KWin aborts and gets restarted by DrKonqui. The new KWin
instance will no longer try to activate the freezing OpenGL as the
protection is set.
If KWin doesn't freeze the protection is removed from the config as
we are used to.
Check for freezes for the first n frames, not just the first
This patch changes the freeze detection code to detect freezes in the
first 30 frames (by default, users can change that with the
KWIN_MAX_FRAMES_TESTED environment variable). This detects
successfully the freezes associated to nouveau drivers
in https://bugzilla.suse.com/show_bug.cgi?id=1005323
Reviewers: davidedmundson, #plasma, #kwin, graesslin
Reviewed By: #plasma, #kwin, graesslin
Subscribers: luebking, graesslin, kwin, plasma-devel, davidedmundson
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D3132
Summary:
The modifier-only-shortcuts break as soon as KWin grabs the keyboard
(e.g. alt+tab, present windows, etc.). The investigation shows that in
that case KWin does not get any raw key events any more and thus gets
confused about the state of the hold modifiers. E.g. alt+tab has the
alt key pressed, but we miss the release as the keyboard is grabbed.
This change addresses the problem by installing an additional event
filter for key press and release event which only filters for key events
on the root window. That way we can be sure that it only operates when
KWin grabbed the keyboard on the root window.
Note: the problem only exists when grabbing on the root window. If the
grab is on another window (e.g. moving a window) we still do get all
events.
The problem also seems to not happen if another application grabbed
keys on the root window. E.g. for key combinations grabbed by
kglobalaccel the correct sequence of key press/release as raw events
are reported. Also while the screen is locked the evemts are reported
and kscreenlocker grabs the keyboard on the root window.
Test Plan:
Used Alt+Tab and Present Windows and tried to activate launcher
afterwards.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D2980
This is Milestone 1 of full support of Atomic Mode Setting (AMS) and
Universal Planes in the KWin DRM backend.
With Milestone 1 we can use the primary plane of a DRM output and do an
AMS commit (this means mode setting aswell as page flipping), if the
driver supports it. Until now the functionality is only tested on Intel
graphics. You need the drm-next kernel for most recent DRM kernel
developments. As boot option set "i915.nuclear_pageflip". Additionally
at the moment AMS is still hidden behind the environment variable
KWIN_DRM_AMS. Set it, if you want to try out AMS.
What needs to be done next: Make it possible to transfer EGL buffers
directly to planes and implement logic for deciding about using a plane
or not for a specific buffer.
You can read more about it on LWN:
https://lwn.net/Articles/653071
And on Martin's blog:
https://blog.martin-graesslin.com/blog/2015/08/layered-compositing/
I used as model previous work by Daniel Stone for Weston:
https://git.collabora.com/cgit/user/daniels/weston.git
Reviewed-by: mgraesslin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D2370
Summary:
SharingPlatformContext::swapBuffers needs to make the context current.
For this is calls makeCurrent. Later on this tries to bind a fbo, which
checks calls into the current QOpenGLContext. As we called makeCurrent on
our own platform context Qt doesn't know that the context is current and
returns in the worst case a nullptr and crashes.
This change calls makeCurrent on the QOpenGLContext so that Qt also knows
that this context is current now. The QOpenGLContext calls makeCurrent on
the platform context, so what our code tried to do is still being done.
In addition the method context() in AbstractPlatformContext is renamed to
eglContext to no longer shadow the method in QPlatformContext.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2582
Summary:
The raw pointer button events intercepted in the XInput2 input filter
get sent through the Platform to the PointerInputRedirection. This
makes the PointerInputRedirection track the pointer button state and
emit the signals for button changed and axis changed.
These signals are used by the modifier-only shortcut detection to
determine whether the shortcut should trigger.
On X11 the "normal" input handling doesn't use the InputRedirection
and the emitted signals are not consumed by anything else. As
PointerInputRedirection is not inited the events are not forwarded
to the input filter, thus won't be processed by other parts and
won't interfere with the normal event processing on X11.
Given that it also doesn't matter that the input filter does not
apply the left-handed setting. The internal tracking will have a
wrong mouse button, but nothing is going to do decisions based on
the value of the pressed mouse button. For the moment all we are
interested in is that a button is pressed.
Test Plan:
Pressed meta, clicked, scrolled, released meta: launcher
did not open. Pressed meta, released meta: launcher opened
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D2506
BUG: 367730
Just like 4e7392b907:
the ownership of the gbm_device must be passed to the Platform as
the ownership of the EGLDisplay is also passed to the Platform and
we may not destroy the gbm_device for an EGLDisplay we are still using.
With this change I could restart the OpenGL compositor successfully
and switch from OpenGL 3.1 to OpenGL 2 without a crash or rendering
issues.
The Egl backend opens the drm device and creates a gbm_device from it.
From that it creates the EGL device. The EGL device outlives the EGL
backend (e.g. for compositor restart). So far the EGL backend destroyed
the gbm_device when going down which mesa did not like at all when a
new context got created on that EGL device.
Thus the ownership of gbm_device is also passed to the Platform so that
the compositor can be restarted.
This makes the TestSceneOpenGL no longer crash during the restart
compositor test.
At the same time the xinput2 integration is split out of X11Cursor
and made a standalone part of the platform plugin. XInput integration
is nowadays not only used by the cursor position polling, but also
for modifier only shortcuts.
By splitting it out the modifier shortcuts start to work also when
one doesn't have anything requesting a mouse position polling.
This also simplifies the conditional builds: xinput integration is
only included if we have support for it at compile time without having
to have many ifdefs in the cursor implementation. For the inclusion of
cursor in the kcmkwin this also removes all the ifdefs.
The key events are only requested if we have xinput 2.1. Otherwise we
would not get all raw events if the input device gets grabbed.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2473
By default the InputRedirectionCursor is created and only the X11
standalone platform creates the X11 specific cursor.
This is a preparation step for moving the X11 specific cursor
implementation into the x11standalone platform plugin.
Summary:
Now powerdevil can adjust brightness using leds subsystem, however, kwin
as well sets brightness to 0 for turning off screen and 0xff when
turning screen back on. This resets the brightness set by the powerdevil
to 100%.
As a solution now kwin listens to brightnessChanged dbus signal of
brightnesscontrol and book-keeps the changed brightness, if screen is
turned off it sets brightness to 0 and when turning screen on, it resets
to old brightness.
If powermanagement service doesn't appear on dbus by default it restores
100% brightness.
Test Plan:
Appearantly this still doesn't work as-it-is on phone, because powerdevil
doesn't emit brightnessChanged dbus signal because of bug in the driver of
backlight control, driver doesn't seem to trigger uevents for changes in
backlight. But with hack in powerdevil to emit brightnessChanged when setting
brightness, this works
Reviewers: broulik, #plasma_on_wayland, graesslin
Reviewed By: #plasma_on_wayland, graesslin
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2468
Summary:
With this change KWin can create window decorations for internal windows.
Thus it's also possible to move internal windows and resize them which is
especially important for the debug console.
Reviewers: #kwin, #plasma_on_wayland, sebas
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2371
Summary:
When closing a nested window kwin_wayland only terminates if the closed
window is the last output window. In a simulated multi-screen setup
closing a window results in the output being removed.
Test Plan: Closed windows, verified with xrandr
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2234
Screens::init sets up a Qt::QueuedConnection connect from countChanged
to changed. This means that when first setting up the connection a
changed signal will be delivered in the next event cycle. This can mess
with other code. E.g. WaylandServer recreating outputs which can make
Xwayland crash or KWin internally crash.
To solve this problem we just need to make sure to set the output count
to the initial value before performing the init.
Reviewed-By: bshah
Summary:
Rational: unredirect fullscreen windows is a weird beast. It's intended
to make fullscreen windows "faster" by not compositing that screen. But
that doesn't really work as KWin jumps out of that condition pretty
quickly. E.g. whenever a tooltip window is shown. KWin itself has a
better functionality by supporting to block compositing completely.
The complete code was full of hacks around it to try to ensure that
things don't break.
Overall unredirect fullscreen has always been the odd one. We had it
because a compositor needs to have it, but it never got truly integrated.
E.g. effects don't interact with it properly so that some things randomly
work, others don't. Will it trigger the screenedge, probably yes, but
will it show the highlight: properly no.
By removing the functionality we finally acknowledge that this mode is
not maintained and has not been maintained for years and that we do not
intend to support it better in future. Over the years we tried to make
it more and more hidden: it's disabled for Intel GPUs, because it used
to crash KWin. It's marked as an "expert" option, etc.
It's clearly something we tried to hide from the user that it exists.
For Wayland the whole unredirect infrastructure doesn't make sense
either. There is no such thing as "unredirecting". We might make use
of passing buffers directly to the underlying stack, but that will be
done automatically when we know it can be done, not by some magic is
this a window of specific size.
Test Plan:
Compiles, cannot really test as I am an Intel user who never
had that working.
Reviewers: #kwin, #plasma, #vdg
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D2180
If the event processing is started before everything is fully started
it can happen that we have "evil" events which may crash either Xwayland
or KWin itself. E.g. if the nested window is too large, the window
manager on the host X-Server will request a resize. If the backend does
that it can happen that the wl_output gets destroyed while Xwayland tries
to bind to it and will crash. Thus let's delay all events till we are
ready to process them.
Summary:
The egl implementation for the virtual platform tries to use a render
node if available. If there is no render node it looks for a virtual
(kernel driver vgem) device, which unfortunately does not create a
render node in mainline kernel (there are patches in ChromiumOS).
For this the Udev wrapper is extended to search for renderNode devices
and for virtual dri devices.
If either render node or vgem dri device is found, it is tried to be
opened (without logind escalation) and on success a gbm device is
created on it. If any step of this fails the so far default behavior
of default device is tried for creating the EGLDisplay.
All of this is compile optional, so that the virtual platform does not
hard depend on udev and/or gbm.
Test Plan:
Auto tests which need OpenGL executed and verified that they
use the render node or vgem device.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2216
Summary:
The position of the output being resized was adjusted instead of only
the other outputs.
Test Plan: Resized windows, verified xrandr
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2235
Summary:
So far SharingPlatformContext was only used if the OpenGL context
supports EGL_KHR_surfaceless_context. If not supported, KWin tried to
create a context through the Wayland API. Unfortunately on hwcomposer
platform this results in a crash as libhybris only supports the init
of EGLDisplay for one native platform.
This change tries to also use the SharingPlatformContext if there is
an OpenGL context in general. It reuses the native EGLSurface created
for the compositing scene and makes its own OpenGL context current on
that surface, too. As KWin creates an FBO, it never renders to it, so
it shouldn't matter at all.
In order to prevent EGL_BAD_MATCH errors when making Qt's OpenGL context
current also the EGLConfig from the scene is used to create the context.
Test Plan: Tested on Nexus5 with qtvirtualkeyboard in KWin
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2231
Summary:
KWin needs to support restarting the OpenGL compositor in case of a
graphics reset event.
On Wayland the tricky part is that the applications should not notice
this. Most importantly KWin cannot just destroy the EGLDisplay and create
a new one. But this is how a restart works: the complete compositor gets
torn down and recreated - including the EGLDisplay.
This change moves ownership of the EGLDisplay to the Platform.
The AbstractEglBackend subclasses query the Platform whether there is
already an EGLDisplay. Only if there is no EGLDisplay the EGLDisplay is
created and only if no EGLDisplay is registered with Wayland the bind
is performed.
Another change is regarding the destruction: the AbstractEglDisplay does
no longer unbind the Wayland display and does no longer destroy the
EGLDisplay. The EGLDisplay is destroyed by the Platform - so very late
on application exit. The Wayland display is unbound when the Compositor
terminates.
Test Plan:
Limited testing with the added auto-test. This one needs to
be extended to fully verify that OpenGL applications continue to work.
But this requires build.kde.org to support OpenGL on Wayland.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2202
Summary:
Most platforms like the nested and virtual do not handle the outputs
themselves and WaylandServer announces the Outputs to Wayland.
So far this was static: at startup it got announced once to Wayland
and any changes were not catched.
This change makes WaylandServer listen to changes to the Screens and
sync them to Wayland.
Unfortunately KWin's internal Screen information is not sufficient to
properly synchronize this to Wayland and also Wayland by not supporting
adding/removing modes does not help.
Thus the solution implemented here is to add new outputs reflecting the
changes and then removing the old ones. This creates situations with more
outputs being present than actually there, but prevents that there are
no outputs at all.
Test Plan: Auto test added which verifies this for the virtual platform
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2233
Summary:
When sharing surface and config used by the compositor scene with
QtQuick's OpenGL context an RGBA surface makes Qt way more happy.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2232
Summary:
With this change the Wayland platform plugin uses the XdgShell to
create the window if available. This allows to close the window and
thus quit the nested kwin_wayland properly.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2201
Some error conditions did not have a warning, so all we got is
"Could not initialize rendering context". Which is not helpful to
figure out what is going wrong.
For virtual machines with Cirrus device the drmModeAddFB fails. So
far there was no error message at all, we only saw it failed.
With this change the drmModeAddFB is checked for error and a warning
is shown. Also further debug messages are added for other cases where
things can fail.
Reviewed-By: Eike Hein
Summary:
So far if the framebuffer platform run into an error on initialization
it did not continue and caused the system to freeze. With this change
it properly emits the initFailed signal in all error conditions which
causes kwin_wayland to terminate. This is a much better situation than
just staying in a running, but frozen state.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2149
Summary:
If it's not possible to create a buffer to blank the output, KWin won't
be able to render to the output. In that case the output should not be
added to the list of outputs.
To support this DrmOutput::blank and DrmOutput::init return bool to
indicate whether they succeeded. DrmBackend handles this situation and
doesn't add the output to the list of outputs if init failed.
If after init there are no outputs KWin is in a state where it won't
be functional. Thus the platform emits the initFailed signal to
terminate.
BUG: 365242
FIXED-IN: 5.7.2
Test Plan: No hardware to reproduce the condition
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2135
Summary:
If the format of the framebuffer is BGR we cannot create an RGB image
format from it - the rendering is incorrect. Unfortunately QImage does
not support a BGR image format.
To solve this problem we still use an RGB image format but on rendering
the front buffer is rgbSwapped to convert the RGB image to a BGR image.
BUG: 365243
FIXED-IN: 5.7.2
Test Plan: Tested on a neon kvm
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2134
Summary:
No need to delegate the painting of the software cursor into the backend.
The core has enough information to perform the rendering itself.
This change means less code duplication and all platforms which might use
a software cursor in QPainter compositor gain support for it without any
further adjustments.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2028
Summary:
So far the drm platform did not verify whether creating and mapping a
DrmBuffer for a cursor works. This could result in a crash in the worst
case.
This change verfies whether mapping the two cursor buffers works, if
not software cursor is enabled. The code is adjusted to ensure that
none of the cursor buffers is accessed in case software cursor are
enabled.
Please note that right now the drm platform's rendering does not
support software cursors. Thus currently this change results in no
cursor at all. This will be addressed by following patches.
BUG: 364740
FIXED-IN: 5.7
Test Plan:
Verfied that it properly falls back to software cursor,
but could not verify that the crash is actually fixed.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2026
Summary:
Platform::setSoftWareCursor creates connections to the Cursor in order
to trigger repaints whenever the cursor position changes. The Cursor is
created before Platform::init is called, but after the Platform is
created. Thus the call needs to happen in init, otherwise the cursor
is not rendered correctly.
BUG: 356328
FIXED-IN: 5.7.0
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D2025
Summary:
When we get a configuration request, we also need to update the global
position on the DrmOutput, otherwise Wayland will know about the new
position, but KWin internally does not.
In addition we also need to trigger the changed signal on KWin::Screens,
so that internally code can react on the change.
With this change configuring layout of connected screens in KScreen
does work also in KWin.
Reviewers: #kwin, #plasma_on_wayland, sebas
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D1858
Summary:
To use eglCreateImageKhr for an X11 pixmap we need an EGLDisplay created
for the same XDisplay as the X11 pixmap. This means if we created an
EGLDisplay for a GBM device, we are not allowed to load a texture from
the X11 pixmap and can result in a crash in the driver.
Similar in the nested X11 setup the EGLDisplay is created for the
rendering window, but the X11 pixmaps are from the Xwayland server KWin
started. They don't belong to the same windowing system.
This change addresses this problem by moving the loading of X11 pixmaps
from AbstractEglTexture to EglTexture of the EglOnX11Backend. Thus for
any usage on a non X11 platform we cannot hit the code path any more.
In addition the nested X11 platform can indicate that it doesn't support
it and thus also doesn't go through the code path.
Test Plan: Tested standalone and nested X11 platform
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D1857
