The timestamp could be either NULL if there's no mode set, or 0 when output gets
awaken. It either crashes weston or we get vblanks at [0, 0] for that output.
Signed-off-by: Marius Vlad <marius-cristian.vlad@nxp.com>
CC: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
[Pekka: note, most start_repaint_loop pass in current time, not 0]
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Due to race conditions, it is (vanishingly unlikely but) possible to
receive a wl_pointer.enter event referring to a wl_surface we have just
destroyed. If this happens, wl_surface will be NULL. Detect this, clear
out our focus, and return.
Other pointer and keyboard events are robust against destroyed surfaces.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Cc: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
[Pekka: remove call to input_set_cursor()]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Do not attempt to set keyboard focus to a surface that has no
wl_resource. The destroy listener hangs off the wl_resource, so if that
is not present, nothing will clean up the pointer when the
weston_surface gets destroyed and it goes stale.
As keyboard_focus_resource_destroyed() sets the focus to NULL, this
patch should be enough to guarantee that the keyboard focus surface will
always have a wl_resource.
I have confirmed the added branch in weston_keyboard_set_focus() can be
hit, but doing so is hard.
My test case has weston/x11 with two outputs, and weston/wayland
--sprawl running on top of that, then closing the parent compositor
output windows one by one. Sometimes it hits, often it does not. Having
the window closing animation enabled may help to hit it.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Destroying an output (wl_surface) can race against the parent compositor
sending wl_keyboard.enter. When this race is lost, wayland-backend
receives wl_keyboard.enter with a NULL wl_surface for the surface it
just destroyed.
Handle this case by ignoring such enter events. Since it is
theoretically possible to follow enter with key events, drop those too.
The modifiers event is sent before enter, so we cannot drop that on the
same condition.
wl_keyboard.leave handler seems to already handle the NULL focus case,
but there is a question if the notify_keyboard_focus_out() call should
be avoided.
This patch fixes a hard to reproduce crash. I was running weston/x11
with two outputs, and weston/wayland --sprawl inside that, then closing
the parent compositor windows one by one. Sometimes it would trigger
this crash.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Daniel Stone <daniels@collabora.com>
EGL_IMG_context_priority allows the client to request that their
rendering be considered high priority. For ourselves, this is important
as we are interactive and any delay in our rendering causes input-output
jitter; a less than smooth user interactive. So if the driver supports
setting the context priority, try and create our EGLContext as high
priority. The driver may reject our request due to system restrictions,
in which case it will fallback to normal priority, but if successful it
will reschedule our rendering and all of its dependencies to execute
earlier, especially important when the GPU is being hogged by background
clients.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
If AddFB2 ever fails for any reason, we fall back to legacy AddFB, which
doesn't support the same swathe of formats, or multi-planar formats, or
modifiers.
This can happen with arbitrary client buffers, condemning us to the
fallback forever more. Remove this, at the cost of an unnecessary ioctl
for users on old kernels without AddFB2; unfortunately, we cannot detect
the complete absence of the ioctl, as the return here is -EINVAL rather
than -ENOTTY.
A check for whether or not the format is valid has been replaced with an
assert, as its callers either check that the format is non-zero, return
a FourCC format code from GBM, or use a static FourCC format.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Make it a bit more clear what the purpose of the variable is.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Nothing in this loop reorders views within the compositor's view_list.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
The direction of scrolling in the RDP compositor appears to be inverted.
When using Weston directly in X, sending X11 button 4 cuases window
contents to scroll up and button 4 to be reported to xwayland clients.
Conversely, when using Weston through RDP (xfreerdp client), sending
X11 button 4 causes window contents to scroll down and button 5 to be
reported to xwayland clients. The xfreerdp client does not seem to be
the cause of this since scrolling works correctly when connecting to
a Windows host.
Signed-off-by: Jason Gerecke <jason.gerecke@wacom.com>
Reviewed-by: David Fort <contact@hardening-consulting.com>
Implement the zwp_input_timestamps_manager_v1.get_touch_timestamps
request to subscribe to timestamp events for wl_touch resources. Ensure
that the request handling code can gracefully handle inert touch
resources.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Implement the zwp_input_timestamps_manager_v1.get_pointer_timestamps
request to subscribe to timestamp events for wl_pointer resources.
Ensure that the request handling code can gracefully handle inert
pointer resources.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Implement the zwp_input_timestamps_manager_v1.get_keyboard_timestamps
request to subscribe to timestamp events for wl_keyboard resources.
Ensure that the request handling code can gracefully handle inert
keyboard resources.
This commit introduces a few internal helper functions which will also
be useful in the implementation of the remaining
zwp_input_timestamps_manager_v1 requests.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Introduce code to support the implementation of the
input_timestamps_unstable_v1 protocol in libweston. This commit does not
implement the actual timestamp subscriptions, but sets up the
zwp_input_timestamps_manager_v1 object and introduces dummy request
handling functions for it, laying the foundation for timestamp
subscriptions for keyboard/pointer/touch to be added cleanly in upcoming
commits.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
... to get the user_data. Like everywhere else through weston.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Ensure the server can safely handle client requests for wl_seat resource
that have become inert due to weston_seat object release and subsequent
destruction.
The clean-up involves, among other things, unsetting the destroyed
weston_seat object from the user data of wl_seat resources, and handling
this NULL user data case where required.
The list of sites extracting and using weston_seat object from wl_seat
resources which were audited for this patch are:
Legend:
N/A = Not Applicable (not implemented by weston)
FIXED = Fixed in the commit
OK = Already works correctly
== keyboard_shortcuts_inhibit_unstable_v1 ==
[N/A] zwp_keyboard_shortcuts_inhibit_manager_v1.inhibit_shortcuts
== tablet_input_unstable_v{1,2} ==
[N/A] zwp_tablet_manager_v{1,2}.get_tablet_seat
== text_input_unstable_v1 ==
[FIXED] zwp_text_input_v1.activate
[FIXED] zwp_text_input_v1.deactivate
== wl_data_device ==
[FIXED] wl_data_device_manager.get_data_device
[OK] wl_data_device.start_drag
[FIXED] wl_data_device.set_selection
[OK] wl_data_device.release
== wl_shell ==
[FIXED] wl_shell_surface.move
[FIXED] wl_shell_surface.resize
[FIXED] wl_shell_surface.set_popup
== xdg_shell and xdg_shell_unstable_v6 ==
[FIXED] xdg_toplevel.show_window_menu
[FIXED] xdg_toplevel.move
[FIXED] xdg_toplevel.resize
[FIXED] xdg_popup.grab
== xdg_shell_unstable_v5 ==
[FIXED] xdg_shell.get_xdg_popup
[FIXED] xdg_surface.show_window_menu
[FIXED] xdg_surface.move
[FIXED] xdg_surface.resize
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Properly clean up all sub-objects (e.g., weston_pointer_client objects)
when a weston_pointer object is destroyed. The clean-up ensures that the
server is able to safely handle client requests to any associated
pointer resources, which, as a consenquence of a weston_pointer
destruction, have now become inert.
The clean-up involves, among other things, unsetting the destroyed
weston_pointer object from the user data of pointer resources, and
handling this NULL user data case where required. Note that in many
sites affected by this change the existing code already properly handles
NULL weston_pointer (e.g. in init_pointer_constraint), so there is no
need for additional updates there.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Fix init_pointer_constraint so that it creates a valid, but inert,
resource if a NULL weston_pointer value is passed in. In that case no
constraint object is associated with the resource, but this is not an
issue since affected code can already handle NULL constraint objects.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Leaks spotted by Valgrind.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Spotted by Valgrind.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Fixes the following Valgrind error:
==21607== Syscall param ioctl(generic) points to uninitialised byte(s)
==21607== at 0x5E8C787: ioctl (in /lib64/libc-2.25.so)
==21607== by 0x8220C17: drmIoctl (in /usr/lib64/libdrm.so.2.4.0)
==21607== by 0x82263CD: drmModeSetCrtc (in /usr/lib64/libdrm.so.2.4.0)
==21607== by 0x7B22095: drm_output_apply_state_legacy (compositor-drm.c:2107)
==21607== by 0x7B2335D: drm_pending_state_apply (compositor-drm.c:2539)
==21607== by 0x7B23AEB: drm_repaint_flush (compositor-drm.c:2773)
==21607== by 0x4E4A3E4: output_repaint_timer_handler (compositor.c:2500)
==21607== by 0x5081496: wl_event_source_timer_dispatch (event-loop.c:235)
==21607== by 0x5081B61: wl_event_loop_dispatch (event-loop.c:633)
==21607== by 0x50803A4: wl_display_run (wayland-server.c:1245)
==21607== by 0x409DD8: main (main.c:2644)
==21607== Address 0xffefff59a is on thread 1's stack
==21607== in frame #2, created by drmModeSetCrtc (???:)
==21607== Uninitialised value was created by a stack allocation
==21607== at 0x7B2782F: drm_output_choose_initial_mode (compositor-drm.c:4842)
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
[Pekka: switch to memset]
Reviewed-by: Daniel Stone <daniels@collabora.com>
Move it into to a new function. Following patches want to compute it
elsewhere as well.
No functional changes.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
This function needs to reset the structures to NULL, otherwise it is not
possible to re-use a once "freed" property info array.
Being able to re-use an array is useful when the memory allocation and
array lifetimes do not match. A specific example is drm_output that is
changed to allocate the CRTC on enable() and deallocate it on disable().
A drm_output might be enabled and disabled multiple times.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
I need to destroy the list from more places, so factor out the common
bits. No functional changes.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Factor out drm_output_init_crtc() and drm_output_fini_crtc(), so that
the call sites can later be moved easily.
On fini, reset scanout_plane and cursor_plane to NULL, so that in the
future when the drm_output is not longer destroyed immediately after, we
free the planes for other use.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
[Pekka: set crtc_id/pipe at top, reset both on error]
Reviewed-by: Daniel Stone <daniels@collabora.com>
Move this bit of code into its own function. The caller of this already
cluttered and origcrtc is not used for anything else.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Let drm_output_enable() remove the CRTC and the connector from the
unused id arrays.
In the future when a list of drm_heads supersedes unused_connectors
array, the usedness of a connector will be determined by the enabled
state of the output the connector (head) is attached to. The enabled
state is turned on by drm_output_enable(). If unused_crtcs array was
still updated in drm_output_repaint(), the CRTC and connector usedness
would be tracked in different places. Logically the two belong together.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Since formats is an out parameter, we need to copy to the alloc'ed
memory and not over the pointer address.
Signed-off-by: Philipp Kerling <pkerling@casix.org>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Setting state_invalid to true is moved together with the code adding new
unused CRTCs and connectors in drm_output_deinit(). Logically these two
operations belong together: state_invalid is required for the new unused
item to be turned off.
This does not hinder initial turning off of outputs, because on
compositor start-up, state_invalid is initialized to true, making calls
to drm_output_disable() for non-enabled outputs a no-op.
Previous changes already ensure that if a compositor does not explicitly
enable an output, the CRTC and connector will be turned off even without
an explicit disable (provided there is a at least one enabled output).
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Rather than smashing the state to disable a CRTC immediately, just
delegate that to the normal repaint cycle by setting state_invalid =
true. drm_pending_state_apply() will pick up the unused_crtcs.
A caveat here is that we have no enabled outputs at all, we will never
enter repaint, and so CRTCs do not actually get turned off until we get
at least one output to drive.
However, this should help the problem reported here:
https://lists.freedesktop.org/archives/wayland-devel/2018-January/036713.html
Arguably it is better to leave an output spuriously on in rare cases
rather than fail modeset completely in somewhat more common cases.
My personal motivation for this change is that it helps if we later move
CRTC allocation to output enable/deinit instead of create/destroy,
because then the CRTC will not be available here for initial turn-off as
the output has not been enabled to begin with.
Cc: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
This was used from the crash handlers, which do not exist anymore.
Nothing calls restore, so delete the dead code.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Instead of assuming the file prefix contains the path and filename
prefix, give these two items separately.
A NULL or empty string path may still be given to refer to the current
directory.
Signed-off-by: Aleksander Morgado <aleksander@aleksander.es>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Replace every use of DATADIR to create a filename with a call to the new
function that allows overriding DATADIR with an env var at runtime.
No attention is paid to asprintf failure.
This restores make distcheck to a passing state after commit 6b58ea
began checking cairo surfaces for validity and exchanged undefined
behaviour we shouldn't have been dependent on for consistent test failure.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
[Pekka: split if-branches into two lines]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Add support for using the atomic-modesetting API to apply output state.
Unlike previous series, this commit does not unflip sprites_are_broken,
until further work has been done with assign_planes to make it reliable.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Co-authored-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Co-authored-by: Louis-Francis Ratté-Boulianne <louis-francis.ratte-boulianne@collabora.com>
Co-authored-by: Derek Foreman <derek.foreman@collabora.co.uk>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
For atomic modesetting support, the mode is identified by a blob
property ID, rather than being passed inline. Add a blob_id member to
drm_mode to handle this, including refactoring mode destruction into a
helper function.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Set the atomic client cap, where it exists, and use this to discover the
plane/CRTC/connector properties we require for atomic modesetting.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Co-authored-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
When leaving Weston, don't attempt to restore the previous CRTC
settings. The framebuffer may well have disappeared, and in every
likelihood, whoever gets the KMS device afterwards will be repainting
anyway.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Rather than a smattering of error handlers, use consistent jump labels
for error paths in create_output_for_connector().
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reported-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Rather than open-coding it ourselves, use the new apply_state helper in
drm_output_start_repaint.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reported-by: Fabien DESSENNE <fabien.dessenne@st.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Split repaint into two stages, as implied by the grouped-repaint
interface: drm_output_repaint generates the repaint state only, and
drm_repaint_flush applies it.
This also moves DPMS into output state. Previously, the usual way to
DPMS off was that repaint would be called and apply its state, followed
by set_dpms being called afterwards to push the DPMS state separately.
As this happens before the repaint_flush hook, with no change to DPMS we
would set DPMS off, then immediately re-enable the output by posting the
repaint. Not ideal.
Moving DPMS application at the same time complicates this patch, but I
couldn't find a way to split it; if we keep set_dpms before begin_flush
then we break DPMS off, or if we try to move DPMS to output state before
using the repaint flush, we get stuck as the repaint hook generates an
asynchronous state update, followed immediately by set_dpms generating a
synchronous state update.
In drm_output_update_complete, the *_pending flags are cleared
before any of the pending actions are taken; this ensures that the
actions cannot recurse.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
If we have an unused CRTC or connector, explicitly disable it during the
end of the repaint cycle, or when we get VT-switched back in.
This commit moves state_invalid from an output property to a backend
property, as the unused CRTCs or connectors are likely not tracked by
drm_outputs. This matches the mechanics of later commits, where we move
to a global repaint-flush hook, applying the state for all outputs in
one go.
The output state_invalid flag originally provoked full changes on output
creation (via setting the flag at output enable time) and session enter.
For the new-output case, we will not have any FB in output->state_cur,
so we still take the same path in repaint as if state_invalid were set.
At session enter, we preserve the existing behaviour: as
start_repaint_loop will fail when state_invalid is set, all outputs will
be scheduled for repaint together, and state_invalid will not be cleared
until after all outputs have been repainted, inside repaint_flush.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Rather than a more piecemeal approach at backend creation, explicitly
track connectors and CRTCs we do not intend to use, so we can ensure
they are disabled where appropriate.
When we have an updated list of connector and CRTC IDs, we add any which
are not owned by an enabled drm_output to the list. We remove them from
the list when drm_output_repaint() is called for that output, and re-add
them when the output is disabled or destroyed.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Ensure the server can safely handle client requests for wl_touch
resources that have become inert due to a weston_touch object
destruction.
This change involves, among other things, setting the weston_touch
object, instead of the weston_seat object, as the user data for wl_touch
resources. Although this is not strictly required at the moment (since
no code is using the wl_touch user data), it makes the code safer:
* It makes more sense conceptually.
* It is consistent with how wl_pointer resources are handled.
* It allows us to clear the user data during weston_touch
destruction, so other code can check whether the resource is
inert.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Ensure the server can safely handle client requests for wl_keyboard
resources that have become inert due to a weston_keyboard object
destruction.
This change involves, among other things, setting the weston_keyboard
object, instead of the weston_seat object, as the user data for
wl_keyboard resources. Although this is not strictly required at the
moment (since no code is using the wl_keyboard user data), it makes the
code safer:
* It makes more sense conceptually.
* It is consistent with how wl_pointer resources are handled.
* It allows us to clear the user data during weston_keyboard
destruction, so other code can check whether the resource is inert.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
This fetches the _NET_WM_ICON property of the X11 window, and use the
first image found as the frame icon.
This has been tested with various X11 programs, and improves usability
and user-friendliness a bit.
Changes since v1:
- Changed frame_button_create() to use
frame_button_create_from_surface() internally.
- Removed a check that should never have been commited.
Changes since v2:
- Request UINT32_MAX items instead of 2048, to avoid cutting valid
icons.
- Strengthen checks against malformed input.
- Handle XCB_PROPERTY_DELETE to remove the icon.
- Schedule a repaint if the icon changed.
Changes since v3:
- Keep the previous Cairo surface until the new one has been
successfully loaded.
- Use uint32_t for cardinals. Unsigned is the same type except on
16-bit machines, but uint32_t is clearer.
- Declare length as uint32_t too, like in xcb_get_property_reply_t.
Signed-off-by: Emmanuel Gil Peyrot <linkmauve@linkmauve.fr>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Much like we already have to_drm_output and to_drm_backend.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Up to now we could set the transform only on output initialization.
However, on certain situations(like tablets and convertible laptops),
screen orientation can change while the compositor is running and thus
the need for change of the output transform arises.
When the transform changes, we must update the output geometry,
output->region and output->previous_damage, as well as send this change
to clients. We also have to check whether any of the pointers are inside
the output which is being rotated. If this is the case, they are moved
to the new center, because otherwise the pointer is stuck outside of the
screen ans "lost" to the user.
What is more, after calling this function compositors should check if
any view is now outside of the screen and move it according to their
wish.
Signed-off-by: Ilia Bozhinov <iliyabo@hotmail.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Acked-by: Daniel Stone <daniels@collabora.com>
If the EGL_EXT_image_dma_buf_import_modifiers extension is not
supported, let gl_renderer_query_dmabuf_formats return a hardcoded
fallback list. That list contains ARGB8888, XRGB8888, and if the
GL_EXT_texture_rg extension is supported, YUYV, NV12, YUV420, and
YUV444.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Although the format event is deprecated, some clients, especially the
GStreamer waylandsink, only support zwp_linux_dmabuf_v1 version 1 and
require the deprecated format event.
Send format events instead of the modifier event, if the client binds on
a protocol version before version 3, skipping formats that only support
non-linear modifiers.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Daniel Stone <daniels@collabora.com>
modifier_hi and modifier_lo are set to 0 by clients,
which are not supporting modifiers. Modifier attributes
of buffers of these clients set to 0 too in linux-dmabuf.c
import_simple_dmabuf function in gl-renderer.c compares
modifier attribute of the buffer with DRM_FORMAT_MOD_INVALID.
DRM_FORMAT_MOD_INVALID is equal to ((1ULL<<56) - 1).
Therefore, modifer 0 is accepted as valid. Then, the function
checks support for eglQueryDmaBufModifiersEXT.
If it is not supported import_simple_dmabuf function is returning
NULL.
This patch sets the modifier attribute to DRM_FORMAT_MOD_INVALID
for clients which are not supporting modifiers. Without this patch
linux-dmabuf protocol is not working for not supporting clients.
Fixes: b138d7afb3 ("gl-renderer: Ignore INVALID modifier")
Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Tested-by: Tomohito Esaki <etom@igel.co.jp>
The GL_EXT_unpack_subimage and GL_EXT_texture_rg are part of the core ES
3.0 specification, so also check the GL driver version in addition to
the extension string to determine if those features are supported.
This allows using those extensions on some GL drivers that do not expose
them in the extensions string, but still support OpenGLES3.
Signed-off-by: Arnaud Vrac <avrac@freebox.fr>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
This is a GL extension and not EGL, so it should be checked after the
EGL context has been created.
Signed-off-by: Arnaud Vrac <avrac@freebox.fr>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
GL drivers might allow using GLES3 features even in GLES2 contexts, but
that's not always the case. To make sure we can use GLES3, first try to
create a GLES3 context and then fallback to GLES2 on failure.
The reported GL version is used to determine which GLES version is
actually available.
Signed-off-by: Arnaud Vrac <avrac@freebox.fr>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
This will allow to make some assumptions in further patches when GLES3
is available.
Signed-off-by: Arnaud Vrac <avrac@freebox.fr>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
If we don't have any damage for the primary plane, then don't force a
repaint; simply reuse the old buffer we already have.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Calling switch_mode with no output or mode never makes any sense. Drop
the NULL checks.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Use a real drm_plane to back the scanout plane, displacing
output->fb_{last,cur,pending} to their plane-tracked equivalents.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Change the type of cursor_plane from a weston_plane (base tracking
structure) to a drm_plane (wrapper containing additional DRM-specific
details), and make it a dynamically-allocated pointer.
Using the standard drm_plane allows us to reuse code which already deals
with drm_planes, e.g. a common cleanup function.
This patch introduces a 'special plane' helper, creating a drm_plane
either from a real KMS plane when using universal planes, or a fake plane
otherwise. Without universal planes, the cursor and primary planes are
hidden from us; this helper allows us to pretend otherwise.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Helper for the pattern of checking whether or not a plane can be used on
an output during the current repaint cycle.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Track dynamic plane state (CRTC, FB, position) in separate structures,
rather than as part of the plane. This will make it easier to handle
state management later, and much more closely tracks what the kernel
does with atomic modesets.
The fb_last pointer previously used in drm_plane now becomes part of
output->state_last, and is not directly visible from the plane itself.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Currently this doesn't actually really do anything, but will be used in
the future to track the state for both modeset and repaint requests.
Completion of the request gives us a single request-completion path for
both pageflip and vblank events.
This merges the timing paths for scanout and plane-but-but-atomic-plane
content.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Does what it says on the box: is true when the compositor is shutting
down.
When we begin to use universal planes, we need divergent destruction
paths. With universal planes, the drm_planes are created at backend
initialisation time, and destroyed with the backend. However, without
universal planes, we create per-output drm_planes to hold the
primary/scanout and cursor planes, whose lifetime is tied to the output.
We will use the new shutting_down flag to determine if output
destruction is hot-unplug or compositor shutdown, and make a decision on
whether or not to destroy the special planes.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Developers with testing rigs having multiple graphics cards plugged in
often want to test things on a specific card. We have ways to choose a
card through seat assignments, but configuring that run by run is
awkward.
Add a new DRM backend option to try to open a specific device, and quit
if it fails.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Add helper function to convert from struct timespec values to tv_sec_hi,
tv_sec_lo, tv_nsec triplets used for sending high-resolution timestamp
data over the wayland protocol. Replace existing conversion code with
the helper function.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Nobody checks for the bool returned by these functions. At the same
time: a) the functions set the respective num_foo to zero on error and
b) callers honour that variable.
Just drop the return type - it's useless.
Note: this is an ABI break.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: Arnaud Vrac <rawoul@gmail.com>
Fixes: 00a03d2f72 ("gl-renderer: add support of WL_SHM_FORMAT_NV12")
Reviewed-by: Daniel Stone <daniels@collabora.com>
In glTexImage2D / glTexSubImage2D calls, the only pixel formats allowed
for the GL_R8 and GL_RG internal formats are respectively GL_RED and
GL_RG [1].
Make sure we match this requirement, as some drivers will fail with the
current code.
[1] https://www.khronos.org/registry/OpenGL-Refpages/es3.0/html/glTexImage2D.xhtml, Table 2
Signed-off-by: Arnaud Vrac <rawoul@gmail.com>
Fixes: 00a03d2f72 ("gl-renderer: add support of WL_SHM_FORMAT_NV12")
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
If we start a special (grabbing) client when Weston is unfocused, it
would lose focus when coming back to Weston.
Signed-off-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Reviewed-by: Daniel Stone <daniels@collabora.com>
As discussed in the following thread:
https://lists.freedesktop.org/archives/wayland-devel/2017-August/034755.html
the existing plane assignment in the DRM backend is vulnerable to
accidental masking of the intended fullscreen surface. This change
adds a simple stateful memory to the plane assignment algorithm
to prevent that.
Reviewed-by: Daniel Stone <daniels@collabora.com>
zwp_linux_dmabuf_v1 allows clients to pass an explicit pitch for each
plane, but wl_shm only takes a single pitch parameter; the pitch for
secondary planes must be inferred.
Multi-plane sub-sampled textures have partial width/height, e.g.
YUV420/I420 has a full-size Y plane, followed by a half-width/height U
plane, and a half-width/height V plane.
GStreamer's waylandsink - the only user of wl_shm YUV formats - expects
the implementation to follow the example of Xv and implicitly divide the
pitch for secondary planes by the subsampling factor. gl-renderer was
not doing this, and instead just using the (larger) stride provided by
the client for all planes in the buffer.
Fix gl-renderer to divide pitch by the subsampling factor when uploading
from subsampled SHM buffers into GL textures, also dividing co-ordinates
when doing offset partial uploads.
Tested with:
$ gst-launch-1.0 videotestsrc ! waylandsink
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reported-by: Fabien Lahoudere <fabien.lahoudere@collabora.co.uk>
Tested-by: Fabien Lahoudere <fabien.lahoudere@collabora.co.uk>
Reviewed-by: Arnaud Vrac <avrac@freebox.fr>
Acked-by: Vincent ABRIOU <vincent.abriou@st.com>
Acked-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Fixes: fdeefe4241 ("gl-renderer: add support of WL_SHM_FORMAT_YUV420")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103063
Stop suggesting to run Weston as root, it is only meant for debugging.
Instead, mention the two supported ways to run Weston on DRM and fbdev:
weston-launch helper and logind service.
Cc: "Ucan, Emre (ADITG/ESB)" <eucan@de.adit-jv.com>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
[Pekka: added forgotten "using" word.]
Explain that -u requires root and -t requires -u. Most importantly,
document in what format does -t expect the tty to be given.
It has been confusing, because Weston's --tty option takes an integer,
weston-launch takes a full device path.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Acked-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Fix an issue introduced in:
commit ab4999492c
Author: Kristian Høgsberg <krh@bitplanet.net>
Date: Fri Jul 19 21:26:24 2013 -0700
weston-launch: Drop sleep_fork option
where the option string accidentally became "t::". That causes
$ weston-lauch -t /dev/tty4
to be parsed incorrectly, as if -t option had no argument and the tty
path gets passed to weston which errors out because of it.
This patch fixes the above to work as expected.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Acked-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
setup_tty() function uses the tty argument for choosing the tty/VT only
if wl->new_user (the -u option) is given. If the tty option is given
without -u, it will only be used for misleading error messages.
To make it clear to the user that -t without -u does not work the way
one might think, let weston-launch exit with an error in that case.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Acked-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Change weston_compositor_get_time to return the current compositor time
as a struct timespec. Also, use clock_gettime (with CLOCK_REALTIME) to
get the time, since it's equivalent to the currently used gettimeofday
call, but returns the data directly in a struct timespec.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Change code related to touch motion events to use struct timespec to
represent time.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Change code related to touch up events to use struct timespec to represent
time.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Change code related to touch down events to use struct timespec to
represent time.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Change code related to key events to use struct timespec to represent
time.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Change code related to axis events to use struct timespec to represent
time.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Change code related to button events to use struct timespec to represent
time.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Change code related to motion events to use struct timespec to represent
time.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Store the output presentation timestamp as struct timespec.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Change code related to animations to use struct timespec to represent
time.
This commit is part of a larger effort to transition the Weston codebase
to struct timespec.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
This bumps the libweston major version due to breakage in the animation
ABI. The commits following this one break more ABI in other parts.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
-uninstalled.pc files are a pkg-config facility for working with
uninstalled libraries.
With pkg-config, foo-uninstalled.pc overrides foo.pc. foo-uninstalled.pc
should never be installed, and will be generated with references to the
build directory.
If you set up your environment so pkg-config looks for .pc files in your
build directories, you can use this to build and link against libraries
you haven't installed with "make install".
This can save time and space over installing with a prefix.
Signed-off-by: Reynaldo H. Verdejo Pinochet <reynaldo@osg.samsung.com>
Acked-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
It appears that wayland_shm_buffer::damage is in the global coordinate
space. Therefore initializing it to width x height at 0,0 is not correct
for any output not positioned at 0,0. That is, all outputs after the
first one get it wrong.
Initialize it from the output region, which is in the global coordinate
space.
While at it, add a comment to note that damage is in global coordinate
space. As I can see, this was the last confusion about it.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Acked-by: Daniel Stone <daniels@collabora.com>
Damage coordinates are in global coordinate space, and they need to
be translated to local coordinate space so multiple outputs can work.
This path now matches the similar path in the X11 backend.
This patch fixes the appearance of multiple windows in the parent
compositor. Previously, all windows except the one with nested output
position 0,0 would have their damage for the parent wl_surface always
fall outside of the wl_surface, save the decorations which were handled
separately. If the parent compositor was Weston/GL, this would lead to
the output area remaining black as partial GL texture uploads would
practically never update the texture. If the parent compositor was
Weston/pixman, the parent windows would not update on screen unless
something else caused the area to be repainted.
[Pekka: adjusted commit message]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Acked-by: Daniel Stone <daniels@collabora.com>
Clarify the error message to explicitly say one was trying to connect to
a parent Wayland compositor. This hopefully is a good enough hint on
what using the wayland-backend is trying to do.
Add the command line display option value and WAYLAND_DISPLAY values for
good measure. WAYLAND_SOCKET is not shown as libwayland-client removes
it.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Acked-by: Daniel Stone <daniels@collabora.com>
The set_windowed and set_fullscreen functions are only useful on a
desktop shell, and never called on fullscreen-shell.
Remove the confusing dead code, and ensure we notice if these functions
get called in the wrong environment.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Acked-by: Daniel Stone <daniels@collabora.com>
To be more symmetric with wayland_output_set_fullscreen(), implement the
xdg-shell path in wayland_output_set_windowed(). This should make it
possible to use the fullscreen key binding to toggle between a floating
window and fullscreen also under xdg-shell.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Acked-by: Daniel Stone <daniels@collabora.com>
I could not find anywhere where struct parent_output was freed, so
apparently we were leaking it.
Check against wayland_backend_register_output() and add the missing
clean-up: removal from the parent output list, and free().
registry_handle_global_remove() also needs fixing to use a safer loop,
because now we are actually removing the list item.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Acked-by: Daniel Stone <daniels@collabora.com>
The wayland-backend with --sprawl is one way to trigger
wayland_output_create_for_parent_output(), which intends to find a mode
from the parent mode list and use it. Calling wayland_output_set_size()
initialized an embedded struct weston_mode and inserts that into the
mode list. Then the assignment output->mode = *mode; corrupts the
mode_list by overwriting the link entry. This leads to an endless loop
in bind_output() in compositor.c.
Fix this by manually doing the setup that wayland_output_set_size() did
and do not call it.
As a side effect, it now relays the parent compositor's physical output
size to our own clients. It no longer smashes the refresh rate either.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Acked-by: Daniel Stone <daniels@collabora.com>
This member is only ever set and never read, therefore it is dead.
Delete dead code.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Acked-by: Daniel Stone <daniels@collabora.com>
Calling wl_display_roundtrip() from within a Wayland event handler means
we will re-enter event dispatch, that is, it will lead to recursive
dispatch. Even if libwayland-client was safe, this would lead to
unexpected code paths: the first event handler function has not returned
when other event handler functions may get called. In this particular
case it maybe didn't hurt, but it's still a fragile pattern to use.
Replace the wl_display_roundtrip() with a manual sync callback to do the
work.
This does not break the wayland-backend initialization sequence, because
sprawl_across_outputs was set only after the roundtrip to ensure
wl_registry globals have been received so the code would not have been
hit anyway, and weston_backend_init() also has a second roundtrip that
ensures the per wl_output events have been received before continuing.
For wayland-backend output hotplug the change is insignificant because
it will only delay the output creation a bit, and the parent outputs are
not processed anywhere in between.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Acked-by: Daniel Stone <daniels@collabora.com>
This patch fixes the wayland backend to not use two different
presentation methods when running on fullscreen-shell.
See also: https://patchwork.freedesktop.org/patch/114534/
v2:
- Add missing wayland_output_resize_surface() call
- Start repaint loop after initial frame has been drawn
v3:
- Redraw the initial frame if present for mode fails
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93514
Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
Tested-by: nerdopolis <bluescreen_avenger@verizon.net>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v2: Fix use after free spotted by Daniel Stone
Signed-off-by: Sergi Granell <xerpi.g.12@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
The opaque region is a few pixels off due to the rounded corners
of the frame decorations, and, therefore, the input region
matches the window's geometry more closely.
Signed-off-by: Sergi Granell <xerpi.g.12@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Add a test environment variable to allow disabling universal planes.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
If the user has passed an INVALID modifier, it's because there is no
applicable modifier, and the buffer layout should be determined by a
magic side-channel call (e.g. bo_get_tiling). If the modifier is
INVALID, don't try to pass it through to EGL, but just drop it.
On the other hand, if a modifier _is_ explicitly specified and we don't
have the modifiers extension, then refuse to import the buffer.
Signed-off-by: Daniel Stone <daniels@collabora.com>
The timer was never removed anywhere. Remove it in disable() to match
what happens in enable().
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
This is a more logical name for the function, matching the pattern used
in other backends and the hook names.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
If we pass the base->enabled test, then the renderer output is
guaranteed to be there, so we can just destroy it.
Destroying it before unmap makes the sequence match better the enable
path.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
Rename fbdev_frame_buffer_destroy() to fbdev_frame_buffer_unmap()
because that is what it does. Adding the destruction of hw_surface in it
makes it the perfect counterpart to fbdev_frame_buffer_map() which
simplifies the code.
fbdev_frame_buffer_map() can no longer call that, so just open-code the
munmap() there. It is an error path, we don't really care about
failures in an error path.
The error path of fbdev_output_enable() is converted to call
buffer_unmap() since that is exactly what it did.
fbdev_output_disable() became redundant, being identical to
fbdev_frame_buffer_unmap().
Invariant: output->hw_surface cannot be non-NULL without output->fb
being non-NULL. hw_surface wraps the mmapped memory so cannot exist
without the mmap.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
A few functions had argument 'output' which was not used at all. Remove
such unused arguments.
The coming migration to the head-based output API would have made it
awkward to come up with the output argument for these, but luckily they
are not actually needed.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Sergi Granell <xerpi.g.12@gmail.com>
Acked-by Daniel Stone <daniels@collabora.com>
Change all backends to set the core backend pointer early.
This is necessary for libweston core to be able to access the backend
vfuncs before the backend init function returns. Particularly,
weston_output_init() will be needing to inspect the backend vfuncs to
see if the backend has been converted to a new API. Backends that create
outputs as part of their init would fail without setting the pointer
earlier.
For consistency, all backends are modified instead of just those that
could hit an issue.
Libweston core will take care of resetting the backend pointer to NULL
in case of error since "libweston: ensure backend is not loaded twice".
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
Check and ensure that a compositor can only load one backend
successfully. If a backend fails to load, it is theoretically possible
to try another backend. Once loading succeeds, only destroying the
compositor would allow "unloading" a backend.
If backend init fail, ensure the backend pointer remains NULL to avoid
calling into a half-loaded backend on e.g. compositor destruction.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
'release' is a more appropriate name because the function does not free
the underlying memory. The main reason for this is that we need the name
weston_output_destroy() for new API that actually will free also the
underlying memory.
Since the function is only used in backends and external backends are
not a thing, this does not cause libweston major version bump, even
though it does change the ABI. There is no way external users could have
successfully used this function.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
A client may have bound the same wl_output multiple times, for who knows
what reason. As the server cannot know which wl_output resource to use
for which wl_surface, send enter/leave events for all of them.
This is a protocol correctness fix.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
Move the remaining scattered setup of the fixed properties into
create_output_for_connector(). All these are already known and they
cannot change.
This helps future refactoring.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
This fixes a regression where monitor make and model would always be
advertised as "unknown" to Wayland clients. The EDID strings were parsed
at create_output_for_connector() time, but the fallback "unknown" values
were set in weston_drm_output_api::set_mode vfunc later. This made the
correct monitor info be shown in the log, but not sent to clients.
The purpose of the "unknown" assignments is to give fallback values in
case EDID is not providing them.
Fix all that by moving all setting of the make, model and serial into
create_output_for_connector(). These values cannot change afterwards
anyway. While at it, document find_and_parse_output_edid().
The ugly casts in create_output_for_connector() are required to silence
compositor warnings from ignoring const attribute. This is temporary,
and a future refactoring will get rid of the casts.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
Move the weston_output_init() call into wayland_output_create_common().
This avoids passing the name twice to different functions, and follows
the precedent set in "libweston: weston_output_init(..., +name)" for
calling init before accessing fields.
Since the error paths in wayland_output_create_for_parent_output() and
wayland_output_create_fullscreen() are now guaranteed to have
weston_output init'd, call weston_output_destroy() appropriately. There
might be more to free than just the name.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
Add 'name' argument to weston_output_init(). This is much more obvious
than the assert inside weston_output_init() to ensure the caller has set
a field in weston_output first.
Now weston_output_init() will strdup() the name itself, which means we
can drop a whole bunch of strdup()s in the backends. This matches
weston_output_destroy() which was already calling free() on the name.
All backends are slightly reordered to call weston_output_init() before
accessing any fields of weston_output, except the Wayland backend which
would make it a little awkward to do it in this patch. Mind, that
weston_output_init() still does not reset the struct to zero - it is
presumed the caller has done it, since weston_output is embedded in the
backend output structs.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Reviewed-by: David Fort <contact@hardening-consulting.com>
[Daniel: document name copying]
Acked-by Daniel Stone <daniels@collabora.com>
Initialize the list in weston_output_init() instead of doing it
separately in each backend.
One would expect weston_output_init() to initialize all weston_output
members, at least those that are not NULL.
We rely on the set_size() functions to be called only once, as is
assert()'d. If set_size() becomes callable multiple times, this patch
will force them to be fixed to properly manage the mode list instead of
losing all members.
compositor-wayland.c is strange in
wayland_output_create_for_parent_output(): it first called
wayland_output_set_size() that initialized the mode list with a single
mode manufactured from width and height and set that mode as current.
Then it continued to reset the mode list and adding the list of modes
from the parent output, leaving the current mode left to point to a mode
struct that is no longer in the mode list and with a broken 'link'
element. This patch changes things such that the manufactured mode is
left in the list, and the parent mode list is added. This is probably
not quite right either.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Acked-by Daniel Stone <daniels@collabora.com>
With FreeRDP 2.0 the crypto needs to be initialized or we fail as soon as we try to
compute a md5. The API also changed for the suppress output callback.
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Use EGL fence sync objects to emit timepoints for the beginning and the
end of rendering on the GPU. The timepoints are emitted asynchronously
using the sync file fds associated with the fence sync objects. The sync
file fds are acquired using the facilities provided by the
EGL_ANDROID_native_fence_sync extension.
The asynchronous timepoint submissions are stored in a list in
gl_output_state until they are executed, and any pending submissions
that remain at output destruction time are cleaned up.
If timelining is inactive or the required EGL extensions are not
present, then GPU timepoint processing and emission are skipped.
Note that the GPU timestamps returned by sync files are in the
CLOCK_MONOTONIC clock domain, and are thus compatible with the
timeline timestamps (which also are in the CLOCK_MONOTONIC domain).
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
The sync file functionality is required by the upcoming GPU render
timeline work, but it's only available in relatively new linux kernel
versions (4.7 and above).
This commit provides a "sanitized" version of the required sync file
definitions. On systems that don't have the sync file header (due to
having an older kernel), we will be able to fall back to our own
definitions when building.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Check for the EGL_KHR_fence_sync and EGL_ANDROID_native_fence_sync
extensions and get pointers to required extension functions.
These extensions allow us to acquire GPU timestamp information
asynchronously, and are required by the upcoming work to add
rendering begin/end timepoints to the weston timeline.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
The purpose of this argument is to hold timestamp information about
events that occurred on the GPU. This argument allows us to include GPU
timestamps in timepoints such as the beginning and end of frame
rendering.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
This change replaces a queued emission of buffer-release events (which
is prone to starvation) with a regular event emission. This means that
client programs no longer need to secretly install surface frame
listeners just to guarantee that they get correctly notified of buffer
lifecycle events.
v2:
More information about the historical reasons why this change hadn't
happened yet, and the consensus to finally move ahead with it can be
found at the discussion terminating in this message:
https://lists.freedesktop.org/archives/wayland-devel/2017-September/035147.html
Signed-off-by: Matt Hoosier <matt.hoosier@gmail.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Not referenced anywhere ever, has been there since the introduction of
fbdev-backend.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Fix the assumption that MAP_FAILED would be equal to NULL. It is not.
Set 'fb' explicitly to NULL on mmap failure so that comparisons to NULL
would produce the expected result.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Simplifies the code, and makes moving weston_output_init() into
wayland_output_create_common() a little easier.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Gets rid of the constant size char array.
While here, document the function.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Ian Ray <ian.ray@ge.com>
Remove the option, because it is hard to use.
Drm connector ids are hard to reach for users,
and they can change when kernel or device tree
is modified.
Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
[Pekka: bump WESTON_DRM_BACKEND_CONFIG_VERSION]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
drm_pending_state is currently skeletal, but will be used to retain
data through begin_repaint -> assign_planes -> repaint -> repaint_flush.
The flush and cancel functions are currently identical, only freeing the
state, but they will be used for different purposes in later patches.
Specifically, the intent is to apply any pending output changes (through
PageFlip/SetCrtc, or the atomic ioctls) in flush, and only free the
state in cancel.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Retain drm_plane tracking objects for all actual DRM planes when using
universal planes, not just overlay planes. Rename uses of 'sprite' to
'plane' to make it clear that it can now be any kind of plane, not just
an overlay/sprite.
These are currently unused.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Add awareness of, rather than support for, universal planes. Activate
the client cap when we start if possible, and if this is activated,
studiously ignore non-overlay planes. For now.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Co-authored-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Co-authored-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Add a cache for DRM property IDs and values, and use it for the two
connector properties we currently update: DPMS and EDID.
As DRM property ID values are not stable, we need to do a name -> ID
lookup each run in order to discover the property IDs and enum values to
use for those properties. Rather than open-coding this, add a property
cache which we can use across multiple different object types.
This patch takes substantial work from the universal planes support
originally authored by Pekka Paalanen, though it has been heavily
reworked.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Co-authored-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
All planes being displayed have a framebuffer. What makes 'fb_plane'
special is that it's being displayed as the primary plane by KMS.
Previous patchsets renamed this to 'primary_plane' to match the KMS
terminology, namely the CRTC's base plane, which is controlled by
drmModeSetCrtc in the legacy API, and identified by PLANE_TYPE ==
"Primary" in the universal-plane API.
However, Weston uses 'primary_plane' internally to refer to the case
where client content is _not_ directly displayed on a plane, but
composited via the renderer, with the result of the compositing then
shown.
Rename to 'scanout_plane' as our least-ambiguous name, and document it a
bit.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Emre Ucan <eucan@de.adit-jv.com>
This moves the single sprite creation code from create_sprites() into a
new function. The readability clean-up is small, but my intention is to
write an alternate version of create_sprites(), and sharing the single
sprite creation code is useful.
The removal code now actually removes the plane from the list.
In doing this, the gymnastics required to exact the CRTC ID the plane
was last on when making a disabling drmModeSetPlane call have been
removed; specifying the CRTC is not necessary when disabling a plane.
(The atomic API goes a step further, mandating it be zero.)
[daniels: Genericised from drm_sprite to drm_plane, moving some of the
logic back into create_sprites(), also symmetrical
drm_plane_destroy.]
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Emre Ucan <eucan@de.adit-jv.com>
Fixes the failure to start with fbdev-backend:
weston: /home/pq/git/weston/libweston/compositor.c:4733: weston_compositor_add_pending_output: Assertion `output->disable' failed.
The disable hook was completely unimplemented, and the regression was
caused by e952a01c3b
"libweston: move asserts to add_pending_output()".
It used to work because Weston never tried to explicitly disable the
fbdev output, but now it is hitting the assert.
Fix it by tentatively implementing a disable hook. It has not been
tested to work for explicit disabling, but it does solve the regression.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=102208
Cc: bluescreen_avenger@verizon.net
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Tested-by: n3rdopolis <bluescreen_avenger@verizon.net>
It's been unused since the legacy (non-libinput) input backends have
been removed.
Signed-off-by: Arnaud Vrac <rawoul@gmail.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
This is easily noticed as a leaked fd on every VC switch.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Tested-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Fix a regression with VT-switching away from Weston and then back
causing drmModePageFlip() to fail with ENOSPC or EINVAL, leaving one or
more outputs not updated. The regression appeared in
47224cc931:
compositor-drm: Delete drm_backend_set_modes
Fix it by forcing a drmModeSetCrtc() on all outputs both initially
created and after VT-switch in.
Cc: Daniel Stone <daniels@collabora.com>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
v2: moved state_invalid=true from create_output_for_connector() to
drm_output_enable()
Reviewed-by: Daniel Stone <daniels@collabora.com>
We still need to close fds passed to us - or we leak quite a few fds
on VC switch.
Regression, originally fixed in 8f5acc2f3a
and re-broken in commit 72dea06d79
but only for the logind launcher.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
This is a simple wrapper for casting the user data of a wl_resource into
a struct weston_output pointer. Using the wrapper clearly marks all the
places where a wl_output protocol object is used.
Replace ALL wl_output related calls to wl_resource_get_user_data() with
a call to weston_output_from_resource().
v2: add type assert in weston_output_from_resource().
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
It's a little awkward to try to keep the weston_output::region and
weston_output::previous_damage allocate exactly only when the output is
enabled. There was also a leak: weston_output_move() was calling
weston_output_init_geometry() on an already allocated regions without
fini in between.
Fix both issues by allocating the regions in weston_output_init(),
always fini/init'ing them in weston_output_init_geometry(), and fini'ing
for good in weston_output_destroy().
This nicely gets rid of weston_output_enable_undo() so I do not need to
try to figure out what to do with it later.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Move the wl_output global management into weston_compositor_add_output()
and weston_compositor_remove_output().
If weston_output_enable() fails, there is no need to clean up the global
and the clients will not see a wl_output come and go.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Move the output id management into weston_compositor_add_output() and
weston_compositor_remove_output(). This is a more logical place, and
works towards assimilating weston_output_enable_undo().
The output id is no longer available to the backend enable() vfuncs, but
it was not used there to begin with.
v2: moved assert earlier in weston_compositor_add_output()
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Enabling an already enabled output is an error, at least with the
current implementation.
However, disabling an output that has not been enabled is ok.
Cope with the first and document the second.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
It was ambiguous what this flag meant - it did not mean whether the
backend is considering this output to be enabled, because
weston_output_destroy() unsets it while deliberately not calling the
backend disable() vfunc.
Perhaps the most clear definition is with respect to the output's
assignment in the pending vs. enabled output lists. There is also a whole
bunch of variables that are allocated only when enabled is true.
Since the flag is related to the list membership, set and clear the flag
only when manipulating the lists.
Assert that weston_compositor_add_output() and
weston_compositor_remove_output() are not called in a wrong state.
v2:
- talk about "list of enabled outputs"
- clear 'enabled' in weston_compositor_remove_output() earlier
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
weston_compositor_add_pending_output() is the point through which all
backends must go when creating a new output. The enable and disable
vfuns are essential for anything to be done with the output, so it makes
sense to check them here, rather than when actually enabling or
disabling.
Particularly the disable vfunc is rarely called, so this gets the check
better excercised.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Only used internally in core. Needs to happen automatically when
something changes, so there should no need to call it from outside.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Only used by weston_output_enable().
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Document two more functions of the weston_output API.
Exported functions marked internal are meant for backends only.
Exported functions not marked internal are meant for libweston users.
v2: talk about "list of enabled outputs".
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
A weston_output available to the compositor should always be either in
the list of pending outputs or the list of enabled outputs. Let
weston_compositor_add_output() and weston_compositor_remove_output()
handle the moves between the lists.
This way weston_output_enable() does not need to remove and
oops-it-failed-add-it-back. weston_output_disable() does not need to
manually re-add the output back to the pending list.
To make everything nicely symmetric and fix any unbalancing caused by
this:
- weston_output_destroy() explicitly wl_list_remove()s
- weston_compositor_add_pending_output() first removes then inserts, as
we have the assumption that the link is always valid, even if empty.
Update the documentations, too.
v2:
- talk about "list of enabled outputs"
- keep wl_list_remove in weston_compositor_remove_output in its old
place
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Trying to make it more readable. Things that happen in the same step are
kept in the same paragraph.
v2: talk about "list of enabled outputs"
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
We need to make sure the client bound dmabuf with a high enough
version to receive modifier events before sending them or the
client will crash.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
We shouldn't free &modifier_invalid because it wasn't allocated
with malloc()
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Some log messages weren't terminated with a newline.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
It looks like there are some code paths where this has been forgotten, so
it likely doesn't work as is. It's probable that nobody has actually
used this in a very long time, so it's not worth the maintenance burden
of keeping xkbcommon vs raw keyboard code anymore.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
EGL_EXT_image_dma_buf_import_modifiers supports importing upto four dmabuf
planes into an EGLImage.
v2: correct PLANE3_PITCH token (Daniel Stone)
Signed-off-by: Varad Gautam <varad.gautam@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
pass over the modifier attributes to EGL.
v2: ensure same modifier is passed for all planes (Daniel Stone)
Signed-off-by: Varad Gautam <varad.gautam@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
implement 'modifier' event to communicate available formats and modifiers
to the client and support zwp_linux_dmabuf_v1 interface version 3.
v2: handle zero modifiers case, deprecate 'format' event.
Signed-off-by: Varad Gautam <varad.gautam@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
EGL_EXT_image_dma_buf_import_modifiers allows querying the formats
and modifiers supported by the platform. expose these to the compositor.
v2:
- change calloc args (Daniel Stone)
- check for modifier support before querying formats (Daniel Stone)
Signed-off-by: Varad Gautam <varad.gautam@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
handle create_immed() dmabuf import requests and support
zwp_linux_dmabuf_v1_interface version 2.
v2: terminate client with INVALID_WL_BUFFER when reason
for create_immed failure is unknown.
[daniels: Bump wayland-protocols dependency.]
Signed-off-by: Varad Gautam <varad.gautam@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
I ran Weston on a Nexus 4 mobile phone, with a native GNU/Linux userland,
and the latest Android kernel for that device from LineageOS [1].
calculate_refresh_rate() returned 1 (mHz), which gets rounded to 0 Hz later
and results in nothing being drawn to the screen.
This patch makes sure, that there is at least a refresh rate of 1 Hz, because
it returns the default refresh rate of 60 Hz otherwise.
[1]: https://github.com/LineageOS/lge-kernel-mako
Signed-off-by: Oliver Smith <ollieparanoid@bitmessage.ch>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
wl_surface_damage_buffer landed ages ago, but in order for GL to
use it the client must bind a wl_compositor version >= 4 (the
version where damage_buffer was introduced).
This patch updates the bind version and allows
eglSwapBuffersWithDamage to actually use the provided damage
rectangles instead of performing full surface damage.
Reviewed-by: Daniel Stone <daniels@collabora.com>
At the bottom of weston_output_finish_frame(), code exists to account
for flips which have missed the repaint window, by shifting them to lock
on to the next repaint window rather than repainting immediately.
This code only accounted for flips which missed their target by one
repaint window. If they miss by multiples of the repaint window, adjust
them until the next repaint timestamp is in the future. This will only
happen in fairly extreme situations, such as Weston being scheduled out
for a punitively long period of time. Nevertheless, try to help recovery
by still aiming for more predictable timings.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
commit a7cba1d4cd changed the way
the cursor plane is setup. Previously it was pre-emptively set
disabled for the next frame, and that would be changed at next
frame time if the cursor plane was to be used. It was changed
to be disabled at plane assignment time.
We disable the use of planes entirely by setting disable_planes to
a non-zero value, which bypasses all calls to assign_planes - so
if the plane was set-up in the previous frame it will retain its
state post-disable.
This leads to desktop zoom leaving the cursor plane in place when
it sets disable_planes.
This patch clears any stale cursor plane state from the redraw
handler if disable_planes is set so drm_output_set_cursor()
will do the right thing.
Reviewed-by: Daniel Stone <daniels@collabora.com>
Reported-by: Emmanuel Gil Peyrot <emmanuel.peyrot@collabora.com>
Reorder some paragraphs to be more logically ordered. Rewrite the
description of the backend-specific disable function to explain the
semantics instead of the mechanics. Remove the paragraph about
pending_output_list as unnecessary details.
Add a big fat comment on why we call output->disable() always instead of
only for actually enabled outputs.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
We make the differentiation where planes are an abstract framebuffer
with a position within a CRTC/output, and sprites are special cases of
planes that are neither the primary (base/framebuffer) nor cursor plane.
drm_sprite, OTOH, contains nothing that's actually specific to sprites,
and we end up duplicating a lot of code to deal with them, especially
when we come to use an entirely plane-based interface with atomic
modesetting.
Rename drm_sprite to drm_plane, to reflect that it's actually generic.
No functional changes.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
[Pekka: dropped the removal of an unrelated comment]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
page_flip_pending is only be set when do a pageflip to a newly-rendered
buffer; if the flag is not set, we have landed in the start_repaint_loop
path where the vblank query fails, and thus we must pageflip to the same
buffer.
This test was not sufficient for what it was supposed to guard:
releasing framebuffers back. When using client-supplied framebuffers, it
is possible to reuse the same buffer multiple times, and we would send a
framebuffer-release event too early.
However, since we have a properly reference-counted drm_fb now, we can
just drop this test, and rely on the reference counting to prevent
too-early release of client framebuffers.
page_flip_pending now becomes exactly what the name suggests: a flag
which indicates whether or not we are expecting a pageflip event. Add
asserts here to verify that we never receive a pageflip event we weren't
expecting.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
vblank_pending is currently a bool, which is reset on every vblank
requests (i.e. sprite pageflip). This can occur more than once per
frame, so turn it into a callback, so we only fire frame-done when we've
collected all the events.
This fixes unexpected behaviour when multiple views per output have been
promoted to DRM planes.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Previously, framebuffers were stored as fb_current and fb_pending.
In this scheme, current was the last buffer that the kernel/hardware had
acknowledged displaying: a framebuffer would be created, set as
fb_pending, and Weston would request the kernel display it. When the
kernel signals that the request was completed and the hardware had made
the buffer current (i.e. page_flip_handler / vblank_handler), we would
unreference the old fb_current, and promote fb_pending to fb_current.
In other words, the view is 'which buffer has turned to light?'.
This patch changes them to a tristate of fb_last, fb_current and
fb_pending, based around the kernel's view of the current state.
fb_pending is used purely as a staging area for request construction;
when the kernel acknowledges a request (e.g. drmModePageFlip returns 0),
the previous buffer is moved to fb_last, and this new buffer to
fb_current. When the kernel signals that the request has completed and
the hardware has made the buffer current, we simply unreference and
clear fb_last, without touching fb_current/fb_pending.
The view here is now 'which state is current in the kernel?'.
As all state changes are incremental on the last state submitted to the
kernel, even if the hardware has not yet been able to make it current,
this simplifies state tracking: all state submissions will always be
relative to fb_current, rather than the previous
(fb_pending) ? fb_pending : fb_current.
The use of fb_pending is strictly bounded between a repaint cycle
(including a grouped set of repaints) beginning, and those repaints
being flushed to the kernel.
fb_current will always be valid between an output's first repaint
flush, and when a disable/destroy request has been processed. For a
plane, it will be valid when a repaint cycle enabling that plane has
been flushed, and when a repaint cycle disabling that plane has been
flushed.
fb_last is only present when a repaint request for the output/plane has
been submitted, but not yet completed by the hardware.
This is the same set of constructs which will be used for storing
plane/output state objects in future patches.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
We implement v2 so use that instead of the DRM_EVENT_CONTEXT_VERSION
macro.
The latter defines the version of the drmEventContext struct declared in
the header [used in the current build] and can be 2, 3 or even 1000.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Instead of setting state members directly in the drm_output_render
functions (to paint using Pixman or GL), just return a drm_fb, and let
the core function place it in state.
This brings damage handling in line with repaint state, so we do not
clear damage if repaint fails.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Call drm_output_render unconditionally, doing an early exit if we're
already rendering a client buffer on the primary plane.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
'next' is used as a framebuffer which has either been rendered but not
had a configuration request (pageflip or CRTC set) applied to it, or
when for a framebuffer that has had configuration requested but not
applied (delayed pageflip where the event has not been applied).
'current' is used as the last framebuffer for which we know
configuration has been fully applied, i.e. CRTC set executed or pageflip
requested and event received.
Rename these members to fb_current and fb_pending, doing some small
reordering of drm_output whilst in the vicinity.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Now that we have better types in drm_fb, use it for cursor buffers as
well. This gives us easier refcounting for our cursors, as well as a
unified buffer-destruction path.
Currently this makes no difference, as the KMS legacy cursor update API
uses GEM names directly, and never touches DRM FBs. However, the cursor
plane becomes a regular KMS plane under atomic, at which point we
require DRM FBs.
Take the opportunity to move to drm_fb ahead of time.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Sometimes we need to duplicate an existing drm_fb, e.g. when
pageflipping to the same buffer to kickstart the repaint loop. To handle
situations like these, and simplify resource management for dumb and
cursor buffers, refcount drm_fb.
drm_fb_get_from_bo has a path where it may reuse a drm_fb, if the BO has
been imported and not released yet. As drm_fb_unref now relies on actual
refcounting (backed up by asserts), we add a balancing drm_fb_ref() to
the path where we return a reused drm_fb.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
We only need it for the GBM surface the FB was originally created
against; a mismatch here is very bad indeed, so no reason to pass it in
explictly every time rather than store it.
Following patches change drm_fb to be explicitly reference counted; in
order to reduce churn, rename drm_output_release_fb to drm_fb_unref
whilst changing its call signature here, even though it does not yet
actually perform reference counting.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
The drm_fb destroy callback to mostly the same thing regardless of
whether the buffer is a dumb buffer or gbm buffer. This patch refactors
the common parts into a new function that can be called for both cases.
[daniels: Rebased on top of fb->fd changes, cosmetic changes.]
Signed-off-by: Tomohito Esaki <etom@igel.co.jp>
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
This uses the new pixel-format helpers, so we can also replace depth/bpp
with these.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalaneN@collabora.co.uk>
Rather than magically trying to infer what the buffer is and what we
should do with it when we go to destroy it, add an explicit type
instead.
In doing so, the test for dumb images (destroying them, but only if
they're not the 'live' ones) is removed. This was dead code, as the only
path which could cause us to shuffle images is drm_output_switch_mode.
This calls drm_output_release_fb before the images are reallocated in
drm_output_fini_pixman / drm_output_init_pixman, with the reallocation
unconditionally destroying the images, so can never be hit.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Make drm_output_set_cursor more deterministic, by calculating more state
and performing more plane manipulation, inside
drm_output_prepare_cursor_view.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Rather than duplicating knowledge of pixel formats across several
components, create a custom central repository.
Signed-off-by: Daniel Stone <daniels@collabora.com>
[Pekka: fix include paths and two copy-pastas]
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
HAVE_LIBDRM was used as a condition for the launcher infrastructure to
call libdrm.so functions. It was set by an independent test for libdrm,
which would silently continue if libdrm was not found. It was assumed
that if you enabled a feature that used libdrm at runtime, the test for
that feature would imply that HAVE_LIBDRM is also set. This was quite
subtle.
The only feature that actually uses libdrm.so at runtime is the DRM
backend. No other backend needs the libdrm calls in the launcher
infrastructure.
Therefore to simplify things, stop using HAVE_LIBDRM and use
BUILD_DRM_COMPOSITOR instead. If you enable the DRM compositor, you
automatically also get libdrm support in the launchers.
There are still things depending on LIBDRM_CFLAGS and LIBDRM_LIBS, so
the test cannot be removed completely.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
This way Wayland compositors will be aware of Weston's
"visible bounds" (and ignore its shadows).
Signed-off-by: Sergi Granell <xerpi.g.12@gmail.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
If wayland_output_create_common returns NULL, it means that
the output creation failed.
Signed-off-by: Sergi Granell <xerpi.g.12@gmail.com>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
struct wayland_output::name was used but never initialized.
Also zxdg_toplevel_v6_set_title was only called for windowed outputs,
and some compositors let you see the client's name even when it is
fullscreen (GNOME Shell's Activities menu for example).
So rename struct wayland_output::name to struct wayland_output::title and
precompute it on wayland_output_create_common(), so it can be later used
on xdg's set_title and frame_create.
v2: Move zxdg_toplevel_v6_set_title() before the wl_surface_commit()
as per Quentin Glidic's suggestion.
Signed-off-by: Sergi Granell <xerpi.g.12@gmail.com>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
When the wheel tilt source is present, gcc complains that we don't
handle all possible enumeration values. We already ensure this cannot
happen in its only caller (handle_pointer_axis), but gcc doesn't
recognise this. Give it a default value to quiet the warning.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Implement new repaint_begin and repaint_flush hooks inside
weston_backend, allowing backends to gang together repaints which
trigger at the same time.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
In preparation for grouping output repaint together where possible,
switch the per-output repaint timer, to a global timer which iterates
across all outputs.
This is implemented by storing the absolute time for the next repaint
for each output locally, and maintaining a global timer which iterates
all of them, scheduling the repaint for the first available time.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Cc: Mario Kleiner <mario.kleiner.de@gmail.com>
Cc: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
[Pekka: The comment about 1 ms delay.]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
glibc 2.25 produces a warning when sysmacros.h is not directly included
but major() is used, as it is intended to be moved to sysmacros.h and
only there. Include it to keep the build happy.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Emilio Pozuelo Monfort <emilio.pozuelo@collabora.co.uk>
repaint_scheduled is actually cleverly a quad-state, disguised as a
boolean. There are four possible conditions for the repaint loop to be
in at any time:
- loop idle; no repaint will occur until specifically requested, which
may be never (repaint_scheduled == 0)
- loop schedule to begin: the loop was previously idle, but due to a
repaint-schedule request, we will call the start_repaint_loop hook
in the next idle task
- repaint scheduled: the compositor has definitively scheduled a
repaint request for this output, which will occur in fixed time
- awaiting repaint completion: the backend has not yet signaled
completion of the last repaint request, and the compositor will not
schedule another until it does so
All but the first condition were previously conflated as
repaint_scheduled == 1, but break them out into separate conditions to
aid clarity, backed up by some asserts.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
