A Vulkan renderer for weston, based on the GL renderer.
The goal is to impose the least requirements as possible on Vulkan
implementations, as to allow even Vulkan 1.0 (or early development)
drivers to run a Wayland compositor. Any additional features or
extensions are made optional if possible.
Currently supports drm, wayland, x11 and headless backends.
As of this implementation, this is still considered an experimental
renderer.
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
When weston tries to update drm head info, if the connector is disconnected, it cannot get the connector's property.
If get connector's property failed, drm_connector_assign_connector_info returns -1 and head->connector->conn, head->connector->props_drm remain to be nullptr, but they are used in update_head_from_connector, so the crash occurs.
So just return ret when drm_connector_assign_connector_info returns -1, and head->connector, head->connector will not be used later.
Signed-off-by: xufeng wang <550002860@gehealthcare.com>
drm_output_propose_state() is called during from drm_assign_planes(). At this
point, the output is enabled, but it may be disabled in the current state, that
is copied.
So explicitly enable the output by setting dpms to WESTON_DPMS_ON.
The same thing is already done in drm_output_repaint() but that's to late for
drm_output_propose_state(). Move that to the fallback path when planes are
disabled. It's no longer needed in general and changing the state between the
atomic test and commit is not a good idea anyways.
Without this, the atomic commit may fail, even though the corresponding test was
successful.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
when drm-backend debug scope is set.
This follows-up with the surface counter timer but because this is
buried in the drm-backend we make use the drm-backend scope if there's a
subscription to it and enable and arm the counter as well when that
happens.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
This patch counts the page flips (either in atomic or legacy page flips)
for DRM outputs and prints them using the drm-backend scope.
Similar to the frame callback timer counter this installs a counter that
periodically computes page flips per a pre-defined interval.
This also includes a perfetto counter to display these in perfetto.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Query the vrr_capable property for heads, and expose it to frontends.
Currently we only support game mode VRR, but expose VRR capabilities as
an enum in anticipation of supporting QMS VRR in the future.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Although backends can create renderbuffers of any size, they always
request the output's current mode size (including decorations).
Letting backends ask for a different size than the output has a few
read-back related design issues like for instance weston_renderer's
read_pixels() API users, currently assuming the output size and
without knowledge of renderbuffers, can retrieve cropped images if a
backend asks for a smaller size. Same issue for the output capture
subsystem.
This commit proposes to fix these issues by simply, albeit perhaps
radically, removing the width and height parameters from
create_renderbuffer(), enforcing the current mode's size of the
associated output.
The VNC and PipeWire backends now also access the output size via the
current mode, not through the width and height variables. This has the
benefit of unifying the backends, as well as the renderers, in their
use of output sizes.
Signed-off-by: Loïc Molinari <loic.molinari@collabora.com>
The Pixman renderer creates renderbuffers with the create_image() and
create_image_from_ptr() functions. The recent addition of the GL
renderer's create_fbo() function, which is pretty similar to the
Pixman ones, brings the opportunity to unify Pixman and GL renderers.
This commit proposes a common renderer function create_renderbuffer()
to create a renderbuffer of the specified format with an optional user
provided destination buffer.
Signed-off-by: Loïc Molinari <loic.molinari@collabora.com>
The output doesn't get finalised and initialised again on resizes
anymore with the Pixman renderer. Only the current dumb and render
buffers get reallocated.
Signed-off-by: Loïc Molinari <loic.molinari@collabora.com>
It's currently assumed by backends that renderers discard
renderbuffers on resize. This commit introduces a new
weston_renderbuffer_discarded_func callback that must be passed at
renderbuffer creation in order to be notified of discarded events from
the renderer. This discarded event could potentially be reused later
by renderers on other occasions without having to change backends once
they get proper support for that.
On output resize, once a discarded event handler fails (returns false)
on a renderbuffer, all the remaining renderbuffers in the output list
go stale and weston_renderer_resize_output() ultimately returns false
for backends to be notified of the failure.
Signed-off-by: Loïc Molinari <loic.molinari@collabora.com>
Renderbuffers currently have a libweston-internal base data structure
with a ref-counting system to handle their lifetime. The problem is
that renderers keep a ref to all renderbuffers in a list per output
(to deal with damages) and that it prevents backends from releasing
renderbuffer resources when not needed anymore. Renderbuffers are then
only released (last ref removed) when the output is destroyed or
resized. dma-buf renderbuffers even expose a dedicated function
remove_renderbuffer_dmabuf() to explictly request the release of their
resources.
This commit proposes to get rid of the ref-counting system by exposing
a single entry point to explicitly destroy all types of renderbuffers
from the renderer.
Instead of removing a renderbuffer from its output list and dropping
its ref when the output is resized, this commit also introduces the
concept of stale renderbuffers which consists in releasing the
resources of a renderbuffer when it's discarded by the renderer while
keeping it in the output list, with a stale state, until it's
explicitly destroyed.
Signed-off-by: Loïc Molinari <loic.molinari@collabora.com>
Move the responsibility for damaging the entire area of new
renderbuffers from backends to renderers.
There's one little drawback: VNC damage logging can't log the
accumulated renderbuffer damage anymore, but I guess this should
somehow be done as an option in the renderers.
Signed-off-by: Loïc Molinari <loic.molinari@collabora.com>
This is problematic as we don't have namespacing for these and some of
the macros can interfere with other defines.
This reverts commit 8634c7e349.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Previously, whether a KMS plane is underlay-able is determined by
whether it's zpos_max is < the primary plane's zpos_min. In other words,
a plane will only be considered underlay-able if its entire valid zpos
range is under the primary plane's lowest zpos.
This is too restrictive - it's possible for planes to have a valid zpos
range that spans below and above the primary's zpos range.
Therefore, allow planes to be used as underlays if their zpos_min is <
the primary plane's zpos_min.
In addition, force rendering on a view if it contains alpha, and is
occluded by a rendered view. If such a view is overlaid, it would render
with incorrect zorder. If it's underlaid, it would render with incorrect
alpha-blending due to hole-punching. Therefore, it must be rendered.
Force rendering prevents the view from going into
`drm_output_find_plane_for_view()`, which serves as an optimization, but
is also observed to prevent dmabuf feedback (derived from
`try_view_on_plane_failure_reasons`) from ping-ponging between two
values, causing some apps (like weston-simple-egl) to constantly
reallocate its buffers.
Because a plane can now - if supported - be used as an underlay, an
overlay, or both, add a `enum drm_plane_subtype` to differentiate
between them. Then, print it's subtype and underlay/overlay assignment
once a decision is made.
v2:
* Squash w/ patch to force rendering on alpha view occluded by rendered
view
* Bring back plane subtype enum to be more expressive about plane
capabilities
* Correct need_hole != false when a view's assignment changes from
underlay to overlay
Signed-off-by: Leo Li <sunpeng.li@amd.com>
As weston_windowed_output_get_api needs ARRAY_LENGTH() move helpers to a
libweston/ so other users can re-use that instead of re-defining these
all over. Easier for other front-ends to make use of them.
With this change windowed-output-api.h also includes the helpers header.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Update the field returned by weston_head_get_display_info(). This makes
EDID di_info available to frontends.
Since EDID data has changed, then head device_changed must be true as
well, because di_info may expose all EDID information and not just what
we track in the backend.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Add 'is_underlay' in drm_plane, which is used to indicate whether the
HW plane is below the primary plane at the Z position.
Add 'has_underlay' in drm_backend, which is used indicate that there
are underlay planes in drm backend.
Signed-off-by: Chao Guo <chao.guo@nxp.com>
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
This reverts f843ba34d1 ("drm-backend: limit
reset/restart to output of a failed commit") and actually solves the problem
correctly.
The pending_state is no longer valid at this point, so it cannot be used to
determine the outputs of the current commit. So only clear will_repaint when
starting to repaint, so it can be used to determine which outputs of a device
were actually repainted and use it to reset/restart those outputs.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
After c08a6ff8 moved attach to the render loop, we have a bad situation
when clients delete an attached shm buffer. We try to query the stride
at attach time, but the shm_buffer has been destroyed, and we crash.
Instead of carefully fixing that, I've instead stored the stride at
buffer creation time (as we already do with buffer width and height).
This lets attach succeed in the gl-renderer, keeping the old texture data
available for any upcoming rendering.
Fixes: #927
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Tested-by: Connor Abbott <cwabbott0@gmail.com>
This most likely is a driver fail, as we shouldn't be getting a zero
value for max_brightness. Later on, this will be used to compute a
normalized brightness which would trip a with a division by zero.
Rather than fixing that up, just don't enable backlight support and
reportat that to the user that we have an invalid max_brightness value
and not backlight support.
Fixes: #878
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
When a commit fails, then only the outputs that where part of this commit should
be reset or restarted.
Otherwise an unrelated output that has another successful pending commit may be
restarted incorrectly. Then the output is in an inconsistent state and will
trigger an assertion:
weston: ../libweston/backend-drm/state-propose.c:627: drm_output_propose_state: Assertion `!output->state_last' failed.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Currently only the return value of the last backend that implements
repaint_flush() is actually used. That only works because the drm backend is the
only backend that implements repaint_flush().
In the drm backend only the return value of the last device is used. And if a
failure is actually detected, then all repainted outputs of all backends are
reset. This can trigger asserts:
weston_output_schedule_repaint_reset() (or _restart()) will change the state of
the output but the backend that did not cause the failure will still call
weston_output_finish_frame().
The same thing can happen with multiple devices in the drm backend. Or outputs
get stuck if an error is dropped.
Since the drm backend is the only one that implements repaint_flush() anyways,
move the failure handling into the backend and make it device specific.
This way only the outputs that need it are reset.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
With multiple backends or devices repaint_begin/repaint_flush may be called even
if no outout of a device will be repainted. This results in an "empty" drm state
without any output.
The actual commit to the kernel is already skipped but the drm-backend log is
still filled with "Beginning repaint"/"repaint-flush" messages and the scene
graph.
Use the new prepare_repaint() callback to determine if a backend needs to be
repainted and only create the pending_state if necessary.
Exit early in repaint_flush()/repaint_cancel() when no pending_state was created.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Devices created via drm_device_create have this hash table, but those
created via drm_backend_create don't.
Signed-off-by: Ray Smith <rsmith@brightsign.biz>
We've forgotten to set this up in some backends, so let's just do it in
weston_compositor_init_renderer().
The headless backend used to fail out if linux_dmabuf_setup() failed, but
had no reason to do so, so just remove that to make the code common.
Suggested by cwabbott on irc.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Based on what is configured in weston_output, check and set the
colorimetry mode into KMS connector property "Colorspace".
This changes how video sinks interpret the pixels, and should allow
driving e.g. WCG monitors in BT.2020 mode.
This does not alter the pixel values themselves. That is the color
manager responsibility, and ultimately the responsibility of the
frontend and the end user to match the monitor driving mode with the
output color profile they chose.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Based on KMS "Colorspace" connector property, populate the mask of
supported colorimetry modes on a head.
EDID should be checked too, but it is currently ignored.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Avoid memory use-after-free when the trying to remove entries from an
already freed list later.
Also add missing removal in drm_output_detach_head() and drm_head_destroy().
Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
I want re-use this variable for printing the colorimetry mode list as
well in a future patch, so a generic name is less confusing.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Store the EDID data as-is, so that we can tell when the EDID blob has
changed. This is not too useful yet, because all the weston_head_set_*()
API raises the device_changed flag only if the information actually
changes. However, I want to expose the libdisplay-info di_info structure
through weston_head, and those cannot be (as) easily compared.
We need to know when the EDID blob changes, so we can call
weston_head_set_device_changed() appropriately when updating di_info.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Just like we already do for planes with proper zpos. Otherwise we'll
often end up choosing the primary plane instead of an overlay one
in `drm_output_find_plane_for_view()`.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
The `drm_output_deinit` should use the drm device passed by the
function, should not use the `b->drm` device.
Signed-off-by: Zhou Liang <174381115@qq.com>
If both the head and writeback are not found, then we should add
connectors to the drm device passed by the function, not the b->drm
device.
Signed-off-by: Zhou Liang <174381115@qq.com>
In a multi-GPU environment, different cards may contain connectors with the
same ID, and drm_head_find_by_connector just use the connector_id to find
the connector, it may find the wrong connector.
Fix this by find the connector based on the drm device and connector id.
Signed-off-by: Zhou Liang <174381115@qq.com>
Currently we flush damage for the "primary plane" every repaint, but this
is folly.
The drm backend may skip rendering entirely if using an all-planes
composition. This could leave the renderer plane in a messy state if a
surface on an overlay plane disappears.
Instead, let the backends flush the primary plane damage when they know
they need to render.
Fixes#864
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
We can't use the surface damage to determine when to upload new cursor
images because when heads overlap the first repainted head will accumulate
that damage as plane damage.
We can't easily use plane damage either because the plane isn't really
assigned until after an atomic test, which requires the cursor fb to be
current.
Untangle this mess a little by always testing with the first cursor fb,
which is identical to the second in all ways, then replace with the correct
fb in repaint.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
This reverts commit 55bf6b5046.
This accidentally removed things that should have stayed - libseat
can still use the logind API, even if weston doesn't directly use
it.
Note that the logind-launcher does not actually build anymore
because breaking changes landed before this revert.
Since we're removing it again right away, I've not taken care to
fix that.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Insert the backend into the weston_compositor::backend_list instead
of setting weston_compositor::backend. The compositor uses this to
determine that the backend is capable of being loaded simultaneously
with other backends.
The DRM backend can only be loaded as primary backend.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
This has been deprecated and non-default for a full release cycle, so
we're going to remove it now.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
The primary_plane is currently shared amongst all outputs, and is the last
barrier to having overlapping outputs.
Split it up and make it per output instead.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
We don't need to do this, we can just leave them in the plane list until
they're used.
Also, doing so helps for when we want to move the primary_plane from
the compositor to the outputs in the future.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Right now every backend clears output damage from the primary plane when
it repaints. Instead of having this same operation spread across all
the backends, just do it in the core instead.
In the future, we want to remove damage tracking from the primary plane
entirely, and this is a small step in that direction.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
A video mode change would be needed to change the underlying renderer
framebuffer. All other backends make uses of this so let's do it for the
DRM-backend as well.
This would also be needed for the output capture to function properly as
we need call weston_output_update_capture_info() when a new mode set has
set. Otherwise we'd run into mismatched dimensions for the current mode
versus the dimensions set-up initially in weston_output_capture_source_info.
Signed-off-by: marius vlad <marius.vlad@collabora.com>
If we want to support multiple backends, the compositor must take care
to call this once, at the appropriate moment, so stop letting the
backends handle compositor shutdown themselves.
Move the weston_compositor_shutdown() calls from the backend::destroy
callbacks into weston_compositor_destroy() and the calls in the backend
creation error paths into weston_compositor_load_backend().
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Add a weston_backend::shutdown callback to split out the part of
weston_backend::destroy that needs to be done before compositor
shutdown.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Before this patch, we would leak the drm_output if there was a pending
flip during shutdown.
Now we destroy the drm_output even if there's a pending flip (only
during shutdown, as we don't want to wait until flip completion to
destroy the output).
Also, it fixes a problem where weston_output_enable() is called right
after weston_output_enable() or weston_output_disable() and it could
fail to find available DRM objects (as they are only released after
the flip completion).
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
