If there is a pending frame, the RenderLoop will delay all schedule
repaint requests to the next vblank event. This means that the render
loop needs to be notified when a frame has been presented or failed.
At the moment, the RenderLoop is notified only about successfully
presented frames. If some frame fails, no repaints will be scheduled
on that output.
In order to make frame scheduling robust, the RenderLoop has to be
notified if a frame has failed.
We want to get notified when the next page flip occurs. The problem is
that kwin will avoid queueing a page flip if nothing has been changed on
the screen. From performance point of view, that is expected behavior,
but for frame scheduling and some wayland clients that create frame
callbacks to get notified about the next vblank, it's not suitable.
The compositing timing algorithm assumes that glXSwapBuffers() and
eglSwapBuffers() block. While this was true long time ago with NVIDIA
drivers, nowadays, it's not the case. The NVIDIA driver queues
several buffers in advance and if the application runs out of them,
it will block. With Mesa driver, swapping buffer was never blocking.
This change makes the render backends swap buffers right after ending
a compositing cycle. This may potentially block, but it shouldn't be
an issue with modern drivers. In case it gets proven, we can move
glXSwapBuffers() and eglSwapBuffers() in a separate thread.
Note that this change breaks the compositing timing algorithm, but
it's already sort of broken with Mesa drivers.
In case hardware transforms can be used again, the shadow framebuffer
must be destroyed; otherwise rendered results will be distorted due to a
mismatch between the dimensions of the shadow framebuffer and the mode
size.
Currently, the OpenGLBackend and the QPainterBackend have hooks to
indicate the start and the end of compositing cycle, but in both cases,
the hooks have different names. This change fixes that inconsistency.
In order to allow per screen rendering, we need the Compositor to be
able to drive rendering on each screen. Currently, it's not possible
because Scene::paint() paints all screen.
With this change, the Compositor will be able to ask the Scene to paint
only a screen with the specific id.
Currently, flip output transformations in the software fallback code
path are equivalent to normal rotate output transformations.
This change implements flip output transformations according to the
wl_output spec.
Currently, every time compositing is restarted, both the gbm and the egl
streams backend will re-obtain the EGLDisplay object.
This is wrong because the core assumption is that the EGL display doesn't
change once it has been obtained.
Summary:
Notify the driver about the parts of the screen that will be repainted.
In some cases this can be benefitial. This is especially useful on lima
and panfrost devices (e.g. pinephone, pinebook, pinebook pro).
Test Plan:
Tested on a pinebook pro with a late mesa version.
Basically I implemented it, then it didn't work and I fixed it.
Maybe next step we want to look into our damage algorithm.
The main advantage of SPDX license identifiers over the traditional
license headers is that it's more difficult to overlook inappropriate
licenses for kwin, for example GPL 3. We also don't have to copy a
lot of boilerplate text.
In order to create this change, I ran licensedigger -r -c from the
toplevel source directory.
Summary:
The previous code uses a fixed version of GSLS which does not match our suported GL targets. We need both, using kwin's existing MapTexture framework we can achieve the same thing in a better way with less code.
Test Plan:
Had some slight issues with the rotation, but everything appeared the same
as when I applied the patchset with the custom shader.
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D26371
Summary:
In case the hardware is not able to rotate the output for the configured
rotation value do this rotation in a post-process step.
For that rendering the current view into a separate framebuffer bound to a
texture that then gets sampled to the default framebuffer in an additional
rendering pass through a simple shader rotating it.
This allows us to leave the Effects system and internal model-view-projection
matrix untouched. The rotation in the post-processing step is isolated.
BUG: 389665
FIXED-IN: 5.18
Test Plan: With KScreen all rotations work.
Reviewers: #kwin
Subscribers: davidedmundson, PureTryOut, z3ntu, zzag, univerz, kwin
Tags: #kwin
Maniphest Tasks: T6106
Differential Revision: https://phabricator.kde.org/D25907
Summary:
This is a small overhaul of the EglGbmBackend. Functions are restructured,
white space added and variables renamed for better readability and style
aligned with Frameworks Coding Style.
Test Plan: Session starts.
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D25830
Summary:
Selecting not to vsync does not make sense for an X11 compositor. In the end
we want clients to be able to present async if they want to but the compositor
is supposed to send swaps with vsync to the XServer in order to not generate
tearing artifacts.
There was also a detection logic which did some questionable things in case
vsync was not available. I don't think this is necessary at all since we can
just always run a timer to present with or without vsync.
Test Plan: kwin_x11 tested on i915.
Reviewers: #kwin, zzag
Subscribers: zzag, kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D23511
Summary:
In order to separate high-level properties of individual outputs from
hardware-specific ones and access these, introduce a new generic class Output.
Also make the DrmOutput class directly a child class of this generic class.
The long-term goal is to get rid of the Screens global object on Wayland and
instead directly work with Output objects on compositing level.
This should enable us long-term to do direct scanout to hardware planes, what
I predict needs this generic output representation at one point.
Test Plan: Manually.
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D11781
Summary:
Do not blindly select first EGL config from returned list, but choose the one that matches GBM surfaces, that will be created later.
GBM surfaces are created with GBM_FORMAT_XRGB8888 format, so choose the config that matches it.
With wrong format EglGbmBackend::resetOutput() will later fail with error EGL_BAD_MATCH.
Test Plan: Compile, run startplasmacompositor. Verify that OpenGL compositing is used, either by kwin debug console, or by kwin support information.
Reviewers: graesslin, davidedmundson, #kwin, #plasma_on_wayland, bshah
Reviewed By: davidedmundson
Subscribers: zzag, kwin, #kwin
Tags: #kwin, #plasma_on_wayland
Differential Revision: https://phabricator.kde.org/D11758
Summary:
Implements a KWayland protocol to pass GBM fd from KWin to KRfb and
addictions to relevant projects from both sides.
Note that this patch does not affect default behaviour of mentioned projects. It can be used
only with KWIN_REMOTE=1 in env from KWin side and with preferredFrameBufferPlugin=gbm in krfbrc from
KRfb side. In all other aspects app behaviour remains unchanged.
Test Plan: Launched KWin in Wayland mode, launched KRfb in it, launched KRDC on a laptop, connected in read-only mode, observed a correctly retrieved desktop with Krfb window
Reviewers: graesslin, davidedmundson, romangg, #kwin
Reviewed By: davidedmundson, romangg, #kwin
Subscribers: kossebau, jgrulich, romangg, ngraham, alexeymin, aacid, kwin, #kwin, davidedmundson, plasma-devel
Tags: #plasma_on_wayland, #kwin
Maniphest Tasks: T5653, T7785
Differential Revision: https://phabricator.kde.org/D1230
Summary:
some drivers, like Mali have EGL_KHR_platform_gbm
but not EGL_MESA_platform_gbm
Test Plan: pending a test on rock64 board
Reviewers: #kwin, #plasma, davidedmundson, graesslin
Reviewed By: #kwin, #plasma, graesslin
Subscribers: graesslin, garg, davidedmundson, plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D10346
Summary:
The AMS code accesses elements in a vector which might not be valid. This
change refactors the code to be more robust, especially the DrmPlane,
which started to crash after adding transformation support.
BUG: 386490
Reviewers: #kwin, #plasma, fvogt, subdiff
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D8752
Summary:
This change hooks up the bits for changing the mode when a request to
change mode from kscreen is received. This includes:
* setting new mode in DrmOutput
* creating new buffers in SceneQPainter backend
* creating new GbmSurface in SceneOpenGL backend
Room for improvement:
* detect whether the new mode works and go back to old one if it fails
Test Plan:
Started both a QPainter/OpenGL session. Used kcmshell5 kscreen
to modify the resolution. Saw that it got applied, maximized a window to
verify it's correct. Changed back to original resolution. Verified it looked
good by maximizing a window.
Reviewers: #kwin, #plasma, subdiff, sebas
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D8398
Summary:
We want to translate by the monitor position, so that needs to be
the negative of the position.
But Kwin/KScreen treats 0 as the top of all monitors. GL treats 0 as
bottom, so that all needs inverting.
Hence this should be a positive y value for the viewport.
BUG: 386099
BUG: 385655
Test Plan:
Had two monitors
Side by side was - fine
Stacked vertically - still fine
Modded X code to extend in y instead of x.
3 monitors worked fine.
Nested wayland only seems to support one screen?
Reviewers: #plasma
Subscribers: plasma-devel, kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D8479
Summary:
In 47343fb we made GBM buffer shared.
What we wanted to do was:
Unbox the shared_pointer<GBMSurface> to give us a GBMSurface* object
Call the gbm_surface*() on that operator
Then cast that to a void* for eglCreatePlatformWindowSurfaceEXT
What we did:
Cast the std::shared_ptr<GBMSurface> to a gbm_surface* then cast that
to void*.
This is just a garbage value and it crashes in Mesa when we do our first
paint.
I've replaced that with an explicit method then we can use shared_ptr's
-> operator rather than get() which does the right thing in a readable
way.
Test Plan:
It crashed after rebasing to master (for Aleix too)
No longer crashes
Reviewers: #plasma
Subscribers: plasma-devel, kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D8251
Summary:
In 47343fb we made GBM buffer shared.
What we wanted to do was:
Unbox the shared_pointer<GBMSurface> to give us a GBMSurface* object
Call the gbm_surface*() on that operator
Then cast that to a void* for eglCreatePlatformWindowSurfaceEXT
What we did:
Cast the std::shared_ptr<GBMSurface> to a gbm_surface* then cast that
to void*.
This is just a garbage value and it crashes in Mesa when we do our first
paint.
I've replaced that with an explicit method then we can use shared_ptr's
-> operator rather than get() which does the right thing in a readable
way.
Test Plan:
It crashed after rebasing to master (for Aleix too)
No longer crashes
Reviewers: #plasma
Subscribers: plasma-devel, kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D8251
Summary:
The gbm_surface is owned by the EglGbmBackend, but it's not the only one
using it. The DrmSurfaceBuffer is also using it and needs it to destroy
the gbm_bo. Now this can become a problem in the following situation:
* a page flip is still pending
* the EglGbmBackend destroys the gbm_surface
-> when the page flip happens the DrmSurfaceBuffer will try to destroy
the gbm_bo and crash as the gbm_surface is no longer valid. This
situation can happen when switching screens or when switching compositing
backend (OpenGL 2 -> OpenGL 3).
To address this problem a class GbmSurface is added which wrapps the
gbm_surface pointer. The EglGbmBackend creates and holds a shared pointer
to the GbmSurface and passes that one to the DrmSurfaceBuffer. So when
cleaning up the gbm_surface only the shared pointer is reset and in case
the DrmSurfaceBuffer still needs it, it can access it without problems.
BUG: 385372
FIXED-IN: 5.11.0
Test Plan: Not yet
Reviewers: #kwin, #plasma, subdiff
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D8152
Summary:
Unfortunately a rather large change which required more refactoring than
initially expected. The main problem was that some parts needed to go
into platformsupport so that the platform plugins can link them. Due to
the rather monolithic nature of scene_opengl.h a few changes were
required:
* SceneOpenGL::Texture -> SceneOpenGLTexture
* SceneOpenGL::TexturePrivate -> SceneOpenGLTexturePrivate
* texture based code into dedicated files
* SwapProfiler code into dedicated files
* SwapProfiler only used in x11 variants
* Safety checks for OpenGL scene moved into the new plugin
* signal declared in SceneOpenGL moved to Scene, so that we don't need
to include SceneOpenGL in composite
Test Plan: Nested OpenGL compositor works
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D7740
Summary:
Several of the subclasses are already derived from QObject.
The main reason is that the class should be moved out of KWin core in
order to move the OpenGL scene into a plugin. As Compositor calls into
the AbstractEglBackend to unbind the wayland display this creates a
problem which is easily solved by turning the AbstractEglBackend into a
QObject and connect to the signal emitted by the Compositor.
Test Plan: Compiles
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D7669
To get an image from KWin to the screen in the DRM pipeline we combine a CRTC,
an encoder and a connector. These objects are static in the sense, that they
represent real hardware on the graphics card, which doesn't change in a
session. See here for more details:
https://01.org/linuxgraphics/gfx-docs/drm/gpu/drm-kms.html
Until now we used DrmOutput as the main representation for such an active
rendering pipeline. I.e. it gets created and destroyed on hot plug events of
displays. On the other side we had no fixed representation of the static kernel
objects throughout the lifetime of KWin. This has several disadvantages:
* We always need to query all available static objects on an hot plug event.
* We can't manipulate the frame buffer of a CRTC after an output has been
disconnected
* Adding functionality for driving multiple displays on a single CRTC (i.e.
cloning) would be difficult
* We can't destroy the last frame buffer on display disconnect because the CRTC
still accesses it and have therefore a memory leak on every display disconnect
This patch solves these issues by storing representations of all available CRTC
and Connector objects in DrmBackend on init via DrmCrtc and DrmConnector
instances. On an hotplug event these vectors are looped for a fitting CRTC and
Connector combinations. Buffer handling is moved to the respective CRTC
instance. All changes in overview:
* Query all available CRTCs and Connectors and save for subsequent hotplug
events
* Fix logic errors in `queryResources()`
* Move framebuffers, buffer flip and blank logic in DrmCrtc
* Remove `restoreSaved()`. It isn't necessary and is dangerous if the old
framebuffer was deleted in the meantime. Also could reveal sensitive user
info from old session.
Test Plan:
Login, logout, VT switching, connect and disconnect external monitor, energy
saving mode.
Reviewers: #kwin
Subscribers: kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D5118
Summary: We need to set the viewport so that we scale from device pixels to global compositor space.
Test Plan:
Ran kwin_wayland properly on my laptop without setting KWIN_COMPOSE.
Most things worked.
Reviewers: #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D3504
The logic on page flips was inconsistent in the sense, that it should never
release a GBM buffer before it's not used (i.e. viewable on the display)
anymore. This happened unintended until now (on page flip we never really
did something useful in the else statement, because if we're here it's not
a GBM buffer anyway).
With this patch we:
* delete on page flips always only the old buffer, which is not used anymore
* release the buffer earlier on interrupts like display disconnect in order
to be able to destroy the EGL context
Test Plan:
Tested in non atomic mode (atomic mode is obsolete right now until my other
WIP patch D4539 is ready to be merged): Login, logout, suspend, normal use.
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D5060
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
