No behavior changes; this is a follow-up of "drm: handle client buffer
destroyed while writeback scheduled".
That commit protects against clients disconnecting while a writeback job
is scheduled, which could otherwise lead to crashes if the buffer is
destroyed before the wb job completes. However, output capture provides
the same functionality: it listens to the client buffer destroy event to
retire the capture task.
Instead of listening to the wl_buffer destroy event, simply listen to
the capture task destroy event. GL renderer already follows this
pattern, and now DRM aligns with it.
See also:
weston_capture_task_buffer_destroy_handler()
weston_capture_task_add_destroy_listener()
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
(cherry picked from commit 04a85ab71c)
No behavior change, just a refactor to make it more clear that the state
is freed after the capture task is retired.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
(cherry picked from commit 7444c29fd2)
Currently when a client buffer gets destroyed, the output capture task
gets destroyed with weston_capture_task_buffer_destroy_handler().
The problem is that we may have a writeback task scheduled, so
wb_state->ct would be pointing to a ct that has already been retired
and destroyed.
In this commit we start handling this case.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
(cherry picked from commit 1152c53e58)
When building without any of the x11 and wayland we still to have some
unused variables/functions.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
(cherry picked from commit 02aa53eb4b)
Fix compilation errors when compiling with x11 or wayland backends
disabled or not available.
Fixes: 8f56d03d ("libweston: Vulkan renderer")
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
(cherry picked from commit ba10e6d814)
yuv_coefficients and yuv_offsets should get optimized away by shader
compilers as the related code paths can never be reached. This seems to
work well on Mesa but not necessarily with other drivers.
While on it, assert that the uniforms *are* present, unless the
yuv-to-rgb conversion is handled by the driver.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
(cherry picked from commit 5c49563ef4)
Texture storage formats and parameters validity checks are done using
glGet*() functions which can be pretty slow depending on the OpenGL
implementation. Wrap these checks around a define disabled by default
in order to get the best OpenGL renderer performance in release
builds.
Signed-off-by: Loïc Molinari <loic.molinari@collabora.com>
This helps to see if the driver returned garbage that triggers an
assertion failure, or if the assertion failure in
weston_output_finish_frame() is more subtle.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
If we might tear, we cannot claim we don't tear.
If we read the software clock, we don't use measurements from the KMS
hardware.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Remove the explicit newline in the log message. Since the log may be wrapped by
the viewer, having the newline in there is not helpful.
While at it, rewrap the string to have shorter lines in the source code.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Check that the view and region completely cover the whole output.
If the view is (partly) semi-transparent, the background region should
cover the area below it - after the changes in the previous commits.
This allows us to remove the now redundant check in the plane selection.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
In the following commits we want to allow semi-transparent views to go
to the primary plane in more cases, e.g when two (or more) instances of
weston-simple-egl are visible on top of a black background.
In that scenario both/all instances would qualify, resulting in the
first candidate to be picked, accidentally making us use underlay
planes. While the visual result would still be correct wherever
supported by the HW (e.g. on the AMD system I mainly test on), it may
break if only overlay planes are supported. Generally it's an unexpected
and unintended change of the plane selection algorithm.
Thus let's be conservative and only consider the background region for
the last visible pnode (the one with the lowes zpos).
This leaves room for further optimizations in the future.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
The background region only contains visible parts of the region, excluding
any regions obscured by pnodes on top of it. If such (partly) opaque pnodes
have been assigned to overlay planes, we should consider them when checking
whether a view can go to the primary/scanout plane.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
To make use of recent optimizations for the is_fully_opaque
calculation, not only relying on the opaque region any more.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
When we switch to the deferred state, we want to complete all outstanding
repaints first - even ones started from the idle handler immediately
after we called weston_backend_set_deferred().
Fixes#1090
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
When we end up with invalid state, we'll be rebuilding the entire state from
scratch. This new rebuilt state would conflict with any outstanding flips,
so we need to wait until all flips are complete to begin the rebuild.
After we set the rebuilt state, we now have to be careful not to allow any
other outputs to attempt flips until the rebuilt state's flip completes,
or we could provoke an EBUSY from the drm device.
Add a recovery state machine to the drm backend to synchronize all of this.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Backends can end up in indeterminate/invalid state and be unable to
process updates.
Add a way for the backend to tell the core to stop scheduling repaints,
and a way for the backend to schedule all the repaints that should have
taken place during the invalid state period.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
The paint node is_fully_opaque and is_fully_blended properties are overriden
when the paint node is censored or used to clear a hole for an underlay plane,
in which case the paint node is forced to be a solid color.
Those properties need to be recomputed when the paint node solid state changes,
that is:
- when the view becomes uncensored
- before assigning planes; if the paint node was overriden late in the repaint
loop to clear a hole for an underlay plane, those properties need to be reset
early in the next repaint loop to properly compute the visibility of
underneath views.
Signed-off-by: Arnaud Vrac <avrac@freebox.fr>
We seem to initialize the uniform with -1 so check against as the test
branch will actually try to use the supplied matrix (although it is
zero in this case).
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
This seems to cause some misuse of buffer handling especially for cases
where the views are being promoted into HW planes. Vulkan clients seem
to crash at start-up while EGL clients seem to avoid being promoted to a
HW plane after changing window state (fullscreen to non-fullscreen).
Revert the now the change to better understand what is going on.
This reverts commit 119ce40714.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
CVD correction is composed by a linear transformation, so we should be
able to collapse it into a single matrix multiplication:
Y = M * X
Where Y is the result, X is the original color, and M is the CVD
correction matrix.
As we need to perform CVD correction for every pixel, this can be
beneficial for limited hardware.
In this patch we do that, updating the libweston core code and also the
GL-renderer and its fragment shader.
Suggested-by: Christopher Healy <healych@amazon.com>
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
CVD correction have the following format. These are linear
transformations in which M is a 3x3 matrix and X is a RGB vector.
F(X) = M * X
It is well known that CVD simulation/correction is done using straight
alpha values. So when we have alpha premultiplied values, we compute
straight alpha values first, apply the correction and premultiply by
alpha again. But this is not necessary, as we can see mathematically:
a * F(X) = a * (M * X) = M * (a * X) = F(a * X)
So do not convert premultiplied alpha values to straight alpha values
before applying CVD correction. This saves a few operations per pixel.
Note that the same does not apply to color inversion:
F(X) = 1 - X
a * F(X) = a - a * X
F(a * X) = 1 - a * X
=>
a * F(X) != F(a * X)
Suggested-by: Christopher Healy <healych@amazon.com>
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Commit 008884f28 ("compositor: Move dirty buffer handling to paint node
early update") makes the dirty paint node bit being
cleared in the update early function which makes t too late to check in
the paint node update late where we used to accumulate painted frames.
Just move it before the early update to make we still have a way to
print painted frames counter.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
In function ‘drm_fb_maybe_import’,
inlined from ‘drm_fb_addfb’ at ../../git/weston/libweston/backend-drm/fb.c:236:8:
../../git/weston/libweston/backend-drm/fb.c:210:31: error: ‘num_planes’ may be used uninitialized [-Werror=maybe-uninitialized]
210 | for (plane = 0; plane < num_planes; plane++) {
Direct-display is mutually exclusive with having GBM BO, because the
point of direct-display is to not import to GBM. We can use this to make
the logic more provable to the compiler, and be more obvious about the
expectation to the human reader.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
{width,height}_from_buffer include the transformations from set_buffer_transform
and set_buffer_scale.
Currently {width,height}_from_buffer are only updated when a commit contains a
new buffer but not for commits with only transformation changes.
If a transform/scale change is commited without new buffer, the old values
remain, which results in incorrect rendering or the client is disconnected
because weston_surface_is_pending_viewport_source_valid() fails.
Make sure to update {width,height}_from_buffer for transformation changes only
to avoid this.
Add a test to verify that the transformations are handled correctly. It is
identical to previous test, except that is spits the buffer attachment and
transformation changes into two commits. So it can reuse the existing images for
validation.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
In handle_touch_up, check whether the associated output is NULL before calling
notify_touch. This prevents unnecessary logging of "Unmatched touch up event
on seat" when a touch device is not bound to any output, as the absence of an
output is already handled gracefully in handle_touch_with_coords.
Signed-off-by: liang zhou <liang.zhou@gehealthcare.com>
We'll be rebuilding state assuming it's all broken, so we should probably
avoid the reuse path as a precaution.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
When repainting a paint node, if the buffer is a solid color buffer we
may skip the rendering pipeline and directly call glClear() with the
solid color, for optimization purposes.
Currently there's a bug where if the output has an output color effect
set, it is skipped when we have a solid color buffer (as we don't go
through the rendering pipeline, where the effects are implemented).
When we have color-management and the paint node contains a color xform,
we've opted for skipping this optimization and going through the
rendering pipeline for now. However, output color effects are simple,
so let's add the logic to apply them on CPU when we have a solid color
buffer.
Reported-by: Christopher Healy <healych@amazon.com>
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
When moving the views between two outputs, the view's paint node is
removed from an output's z order list, but no paint_node_changes are set.
The drm-backend uses paint_node_changes to determine if it can reuse state,
so could leave a stale image behind on a plane until something else
triggered new state generation.
Make z_order_link removal dirty the paint node changes to prevent stale
state reuse.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Users might start the compositor with the scene-graph debug scope which
requires having the clock set-up as implicitly one of the backends needs to be
loaded.
Just avoid doing that until we have one the backends loaded to prevent
a start-up crash.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
The protocol requires clients to perform an initial commit before they
receive configure events:
> After creating a role-specific object and setting it up (e.g. by sending
> the title, app ID, size constraints, parent, etc), the client must
> perform an initial commit without any buffer attached. The compositor
> will reply with initial wl_surface state such as
> wl_surface.preferred_buffer_scale followed by an xdg_surface.configure
> event. The client must acknowledge it and is then allowed to attach a
> buffer to map the surface.
Previous to this patch various calls such as set_fullscreen() or
set_maximized() would schedule configure events, resulting in clients
being able skip the initial commit. This again made it possible to write
clients that only work on Weston, in violation of the protocol.
For xdg-popups we already tracked the initial commit status. Move it
to xdg-surface, guard schedule_configure() on it and ensure to run the
later on the initial commit.
Incorporate tests that checks if we get configure events when calling
set_fullscreen/set_maximized and tests that uses the main xdg_surface as
a parent to a sub-surface (which initially triggered this issue).
Signed-off-by: Robert Mader <robert.mader@collabora.com>
The change in 65227cb7d4 is actually destructive, as it causes multiple
drm-backend heads to no longer be repainted at the exact same time at
startup.
This frequently leads to a failure to flip when only one head is properly
set up and others have stale state.
Force stampless finish_frame to skip any time adjustments to restore proper
behaviour.
Fixes: 65227cb7 (compositor: Round repaint times down to possible presentation times)
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Make sure we check if a paint node has been assigned to a view before
trying to dump its contents.
The scene graph dump from the repaint loop should always be ok, but if
we're using weston-debug we could race with the repaint for a new view
and try to print the scene graph before the node is created.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Using values from the color representation protocol. Only supported for
YCbCr formats in KMS drivers for now.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
`EGL_YUV_COLOR_SPACE_HINT_EXT` and `EGL_SAMPLE_RANGE_HINT_EXT` have
to be set at image import, which again happens on wl_buffer creation.
This is problematic because in Wayland (and Vulkan) those values are
surface attributes and thus only known once the buffer gets attached
to a particular surface.
We solve that by first importing the dmabuf with default values and,
when attaching a buffer to a surface, creating additional EGLImages
as needed.
Note that we ignore `EGL_YUV_CHROMA_HORIZONTAL_SITING_HINT_EXT` and
`EGL_YUV_CHROMA_VERTICAL_SITING_HINT_EXT` for now, which are neither
required by the Wayland color-representation protocol nor supported by
Mesa shaders (as of 25.3).
Signed-off-by: Robert Mader <robert.mader@collabora.com>
Unlike EGL image import, texture creation and binding is quite fast and
cheap. Thus, instead of doing it as early as possible in various places,
do it right before we actually need it in prepare_textured_draw().
This will follow-up changes easier and includes some preparations for
them.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
Add matrix-coefficients and range tracking to more gl-renderer structures
and get the relevant values from the color-representation implementation
instead of hard-coding them in the shader.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
For now limited to coefficients and ranges that are typically
supported by KMS drivers. We notably leave out alpha modes and
chroma locations for now.
The protocol initialization is guarded by the WESTON_CAP_COLOR_REP
backend capability and thus not enabled anywhere yet.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
Doesn't yet deal with work area (monitor area excluding panels etc), or
reconstraining while moving, but it's a start.
It's more or less based on the mutter implementation. While the mutter
implementation is GPL, it was authored by me, thus grant permission to
relicense to MIT.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
