Commit Graph

38 Commits

Author SHA1 Message Date
Emmanuel Gil Peyrot
11ae2a3036 compositor-drm: pageflip timeout implementation
Weston will not repaint until previous update has been acked by a
pageflip event coming from the drm driver. However, some buggy drivers
won’t return those events or will stop sending them at some point and
Weston output repaints will completely freeze. To ease developers’ task
in testing their drivers, this patch makes compositor-drm use a timer
to detect cases where those pageflip events stop coming.

This timeout implementation is software only and includes basic
features usually found in a watchdog. We simply exit Weston gracefully
with a log message and an exit code when the timout is reached.

The timeout value can be set via weston.ini by adding a
pageflip-timeout=<MILLISECONDS> entry under [core]
section. Setting it to 0 disables the timeout feature.

v2:
- Made sure we would get both the pageflip and the vblank events before
  stopping the timer.
- Reordered the error and success cases in
  drm_output_pageflip_timer_create() to be more in line with the rest
  of the code.

v3:
- Reordered (de)arming of the timer with the code around it to avoid it
  being rearmed before the current dearming.
- Return the proper value for the dispatcher in the pageflip_timeout
  callback.
- Also display the output name in case the timer fires.

v4:
- Reordered a forgotten timer rearming after its drmModePageFlip().

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=83884
Signed-off-by: Frederic Plourde <frederic.plourde at collabora.co.uk>
Signed-off-by: Emmanuel Gil Peyrot <emmanuel.peyrot@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2017-03-07 16:42:40 +02:00
Daniel Stone
906488790b compositor-drm: Make scanout view preparation more stringent
Don't import buffers which span multiple outputs, short-cut any attempt
to import SHM buffers, and ignore buffers with a global alpha set.

I'm not convinced all of these conditions entirely make sense, but this
at least makes them equally nonsensical.

Differential Revision: https://phabricator.freedesktop.org/D1414

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
2017-02-27 21:39:13 +00:00
Armin Krezović
445b41b9d5 compositor-drm: Construct mode list in create_output_for_connector
And properly deconstruct it in drm_output_destroy.

Might be useful for finding out which modes are supported
before even setting them, in case we want to extend the
modesetting API.

Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
2017-02-27 21:38:54 +00:00
Daniel Stone
75487c2560 compositor-drm: Try to preserve existing output routing
Previously in picking CRTC -> encoder -> connecting routing, we went for
the first triplet we found which claimed to work.

Preserving the existing routing means that startup will be faster: on a
multi-head system, changing the routing implies disabling both CRTCs,
then re-enabling them with a new configuration, which may involve
retraining links etc.

Furthermore, the existing routing may be set for a reason; each
CRTC/encoder is not necessarily as capable as the other, so the routing
may be configured to stay within such device limits.

Try where possible to respect the routing we pick up, rather than
blithely configuring our own.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2017-02-27 21:38:51 +00:00
Daniel Stone
efa504f4ea compositor-drm: Ignore non-KMS devices
Given that we can have render-only devices, or vgem in a class of its
own, ignore any non-KMS devices in compositor-drm's device selection.
For x86 platforms, this is mostly a non-issue since we look at the udev
boot_vga issue, but other architectures which lack this, and have
multiple KMS devices present, will hit this.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reported-by: Thierry Reding <treding@nvidia.com>
Reported-by: Daniel Vetter <daniel.vetter@intel.com>
2017-02-27 21:38:49 +00:00
Armin Krezović
9bcf4c55e0 compositor-drm: Mark eDP connection as internal
Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
2017-02-13 21:28:27 +00:00
Daniel Stone
efc2b1d4db compositor-drm: Remove connector_allocator
Remove the last usage of connector_allocator, which was to check for
displays which have been hot-unplugged, and replace it with an array
which doesn't rely on the connector IDs remaining below 32 (or 64).

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Reported-by: Peter Senna Tschudin <peter.senna@collabora.com>
2017-02-09 18:05:04 +00:00
Daniel Stone
c0ec75919a compositor-drm: Avoid connector_allocator for hotplugs
Rather than using connector_allocator to determine whether an output is
newly connected or not, use a list walk across all outputs instead.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Reported-by: Peter Senna Tschudin <peter.senna@collabora.com>
2017-02-09 17:39:17 +00:00
Daniel Stone
72c0e1b821 compositor-drm: Remove crtc_allocator
crtc_allocator was used as a bitmask of CRTC IDs, so we didn't try to
use the same CRTC for multiple outputs. Unfortunately, this only works
to the extent that CRTC object IDs fit within the bitmask; though they
were previously, they are not guaranteed to be under 32 or even 64.

Replace the only use of crtc_allocator with a list walk across outputs.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Reported-by: Peter Senna Tschudin <peter.senna@collabora.com>
2017-02-09 17:39:06 +00:00
Ucan, Emre (ADITG/SW1)
9a200d7559 compositor-drm: don't pass option_connector to create_outputs
The connector option is a part of drm_backend struct.
Therefore, it is not needed to pass it as an argument
to create_outputs function.

Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Acked-by: Daniel Stone <daniels@collabora.com>
2017-02-07 11:33:16 +02:00
Ucan, Emre (ADITG/SW1)
21e4944178 compositor-drm: update connectors with connector config
weston can be started with --connector option to be initialized
with a particular output. But in the update_outputs this option
is not considered and output is created for all the available
connectors. This patch fixes this issue by considering
the option for connectors in the update_outputs.

Signed-off-by: Emre Ucan <eucan@de.adit-jv.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Acked-by: Daniel Stone <daniels@collabora.com>
2017-02-07 11:33:16 +02:00
Micah Fedke
c8890125ad compositor-drm: don't put y-inverted / interlaced / bottom-first dmabufs on overlays
This patch checks the attribute flags on incoming dmabufs and refuses to
put them overlays if they have any of the flags set (currently:
ZWP_LINUX_BUFFER_PARAMS_V1_FLAGS_Y_INVERT,
ZWP_LINUX_BUFFER_PARAMS_V1_FLAGS_INTERLACED and
ZWP_LINUX_BUFFER_PARAMS_V1_FLAGS_BOTTOM_FIRST), instead defaulting to
the gl-renderer which can handle some of the flags.

This check should be superceded by buffer transforms, when they become
available.

Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2017-02-06 13:43:40 +02:00
Armin Krezović
605ac8e685 compositor-drm: Restore use-current-mode functionality
It got lost during the porting to the config API.

Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
2017-01-26 14:14:14 +00:00
Quentin Glidic
23e1d6f176
libweston: Properly namespace backends entrypoint
This prevents loading a backend as a simple module. This will avoid
messing up with backends when we will introduce libweston common
modules.

Signed-off-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Reviewed-by: Daniel Stone <daniels@collabora.com>
2017-01-17 18:24:49 +01:00
Fabien Dessenne
2d66a7d883 compositor-drm: allow mode frequency selection
As an option, allow to specify a mode (from the configuration file) by
its refresh rate.
Example of valid syntax:
- "mode=1920x1080"    Select a 1920x1080 mode, refresh rate undefined.
- "mode=1920x1080@60" Select the (or one of the) 1920x1080 60 Hz mode.

Signed-off-by: Fabien Dessenne <fabien.dessenne@st.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
2017-01-17 16:38:43 +00:00
Abdur Rehman
4dca0e1717 compositor-drm: fix a couple of typos
Signed-off-by: Abdur Rehman <arehmanq199@gmail.com>
Reviewed-by: Yong Bakos <ybakos@humanoriented.com>
2017-01-03 11:58:55 +00:00
Daniel Stone
296d7a92ad compositor-drm: Reshuffle and comment plane conditions
Try to harmonise the various plane-import paths a little bit, starting
with reshuffling and commenting the conditions to do so.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>

Differential Revision: https://phabricator.freedesktop.org/D1413
2016-12-12 20:49:29 +00:00
Daniel Stone
f33e104865 compositor-drm: Remove open-coded weston_compositor_wake
This always changes the state to ACTIVE when we enter the session,
whereas the previous implementation preserved the state (i.e. if state
was SLEEPING on exit, it would be restored to SLEEPING, but also with a
repaint). This seems more helpful behaviour, however: if you enter a
session, it's probably in order to interact with it.

Differential Revision: https://phabricator.freedesktop.org/D1482

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
2016-12-12 17:18:08 +00:00
Daniel Stone
47224cc931 compositor-drm: Delete drm_backend_set_modes
Even if we do have a framebuffer matching the mode, we immediately
schedule a repaint, meaning we either do work for no reason, or show
stale content before we bring up the new content.

Delete this and just let repaint deal with it.

Differential Revision: https://phabricator.freedesktop.org/D1481

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
2016-12-12 17:15:08 +00:00
Daniel Stone
c8c917cb61 compositor-drm: Store width and height inside drm_fb
This will be used so we can later determine the compatibility of drm_fbs
without needing to introspect external state.

Differential Revision: https://phabricator.freedesktop.org/D1487

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
2016-12-12 17:15:08 +00:00
Daniel Stone
f214fdca5a compositor-drm: Use signed int for width/height
This makes it sign-compatible with weston_output->{width,height}.

Differential Revision: https://phabricator.freedesktop.org/D1486

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
2016-12-12 17:15:08 +00:00
Daniel Stone
4e5eceb075 compositor-drm: Use fb->fd consistently
Everyone else uses fb->fd rather than pulling the FD back out of GBM.
Use that in the destroy callback too.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>

Differential Revision: https://phabricator.freedesktop.org/D1406
2016-12-12 17:15:08 +00:00
Daniel Stone
3e661f7b6c compositor-drm: Extract EGL destroy to helper
No functional change.

Differential Revision: https://phabricator.freedesktop.org/D1484

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
2016-12-12 17:15:08 +00:00
Daniel Stone
a3ae4767ad compositor-drm: Simplify drm_sprite_crtc_supported
No need to walk the CRTC list every time looking for CRTC indices, when we
already have the CRTC index stashed away. Taking the plane as an argument
also simplifies things a little for callers, and future-proofs for a
potential future KMS API which passes a list of supported CRTC IDs rather
than a bitmask of supported CRTC indices.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>

Differential Revision: https://phabricator.freedesktop.org/D1407
2016-12-12 17:15:08 +00:00
Daniel Stone
17339233a0 compositor-drm: Comment struct members
Clarify the difference between crtc_id (DRM object) and pipe (index into
drmModeRes->crtcs array, possible_crtcs bitmask).

Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Reviewed-by: Armin Krezović <krezovic.armin@gmail.com>
Differential Revision: https://phabricator.freedesktop.org/D1405
2016-12-12 16:52:22 +00:00
Miguel A. Vico
41700e355f gl-renderer: Add <platform_attribs> param to gl_renderer_display_create
This change adds <platform_attribs> parameter to
gl_renderer_display_create() in case we ever want to pass non-NULL.

Signed-off-by: Miguel A Vico Moya <mvicomoya@nvidia.com>
Reviewed-by: Andy Ritger <aritger@nvidia.com>
Reviewed-by: James Jones <jajones@nvidia.com>
[Pekka: removed notes about EGLOutput]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2016-11-07 17:06:44 +02:00
Miguel A. Vico
c095cde364 gl-renderer: Rename gl_renderer::output_create to output_window_create
No functional change. This patch renames gl_renderer_output_create() to
gl_renderer_output_window_create(), which is something more descriptive
of what the function does.

Signed-off-by: Miguel A Vico Moya <mvicomoya@nvidia.com>
Reviewed-by: Andy Ritger <aritger@nvidia.com>
Reviewed-by: James Jones <jajones@nvidia.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2016-11-07 17:06:01 +02:00
Miguel A. Vico
dddc670c04 gl-renderer: Rename gl_renderer_create to gl_renderer_display_create
No functional change. This patch only renames gl_renderer_create() to
gl_renderer_display_create(), which is something more descriptive of
what the function does.

Signed-off-by: Miguel A Vico Moya <mvicomoya@nvidia.com>
Reviewed-by: James Jones <jajones@nvidia.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2016-11-07 17:05:18 +02:00
Vincent Abriou
c950667e87 libweston: include weston-egl-ext.h in drm, x11 and wayland compositor
As to what is done for gl-renderer.c, weston-egl-ext.h should be
include in compositor-drm.c, compositor-x11.c and compositor-wayland.c.
This fix building issue with GPU that does not have EGL_PLATFORM_xxx_KHR
in their extension header file eglext.h.

Signed-off-by: Vincent Abriou <vincent.abriou@st.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
2016-10-05 16:47:43 +01:00
Armin Krezović
4008740d5e weston: Rename weston_output_init_pending() to weston_output_init()
v2:

 - Rebased for latest changes.

v3:

 - Rebased for changes in wayland backend.

Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-10-05 14:59:09 +03:00
Armin Krezović
083681325b weston: Port DRM backend to new output handling API
This is a complete port of the DRM backend that uses
the recently added output handling API for output
configuration.

Output can be configured at runtime by passing the
necessary configuration parameters, which can be
filled in manually or obtained from the configuration
file using previously added functionality. It is
required that the scale and transform values are set
using the previously added functionality.

After everything has been set, output needs to be
enabled manually using weston_output_enable().

v2:

 - Added missing drmModeFreeCrtc() to drm_output_enable()
   cleanup list in case of failure.
 - Split drm_backend_disable() into drm_backend_deinit()
   to accomodate for changes in the first patch in the
   series. Moved restoring original crtc to
   drm_output_destroy().

v3:

 - Moved origcrtc allocation to drm_output_set_mode().
 - Swapped connector_get_current_mode() and
   drm_output_add_mode() calls in drm_output_set_mode()
   to match current weston.
 - Moved crtc_allocator and connector_allocator update
   from drm_output_enable() to create_output_for_connector()
   to avoid problems when more than one monitor is connected
   at startup and crtc allocator wasn't updated before
   create_output_for_connector() was called second time,
   resulting in one screen being turned off.
 - Moved crtc_allocator and connector_allocator update from
   drm_output_deinit() to drm_output_destroy(), as it
   should not be called on drm_output_disable().
 - Use weston_compositor_add_pending_output().
 - Bump weston_drm_backend_config version to 2.

v4:

 - Reset output->original_crtc to NULL if drm_output_set_mode()
   fails.
 - Remove unneeded log message when disabling an output when a
   pageflip is pending.
 - Document that create_output_for_connector() takes ownership
   of the connector.
 - Free the connector if create output conditionals are not met
   in create_outputs() and update_outputs().

Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2016-10-03 14:22:50 +03:00
Yong Bakos
4b6321f49a
compositor-drm: Zero drmModeAddFB2 data
Initialize arrays of data passed to drmModeAddFB2, just as
drm_fb_get_from_bo does.

See https://lists.freedesktop.org/archives/wayland-devel/2016-August/030645.html

Signed-off-by: Yong Bakos <ybakos@humanoriented.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
2016-08-18 10:28:35 +02:00
Derek Foreman
482ffdf006
compositor-drm: Stop sending uninit data to the kernel
Valgrind noticed that we send uninit data to drmModeAddFB2.  While
the kernel should never read this (because of the plane format),
it's probably still nicer to zero the data before we send it.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Yong Bakos <ybakos@humanoriented.com>
2016-08-15 17:35:42 +02:00
Armin Krezović
545dba6fda compositor-drm: Use non-cast functions to retrieve backend and output objects
This uses container_of instead of explicit cast to retrieve
backend and output objects from generic weston_backend and
weston_output pointers.

v2:

- Remove unneeded cast
- Remove unneeded line breaks

Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2016-08-08 14:48:51 +03:00
Armin Krezović
984781f302 compositor-drm: Do not exit when there are no outputs left
When there are no outputs left after a hotplug event, weston
will terminate. This isn't desired when trying to get weston
to work with zero outputs.

Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2016-08-05 15:14:07 +03:00
Armin Krezović
b12a754259 compositor-drm: Do not return an error when no connectors are configured
Returning an error when there are no connectors results in
weston terminating after that. That's not expected when
trying to get weston to start with zero drm outputs.

Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2016-08-05 15:14:07 +03:00
Jussi Kukkonen
649bbce607 include stdint.h for int32_t/uint32_t
Signed-off-by: Jussi Kukkonen <jussi.kukkonen@intel.com>
Reviewed-by: Yong Bakos <ybakos@humanoriented.com>
2016-07-26 16:26:08 -07:00
Pekka Paalanen
b5e3ea218b Rename src/ to libweston/
This clarifies what is supposed to be the libweston code.

v2: screen-share.c is already in compositor/ instead.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Yong Bakos <ybakos@humanoriented.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Tested-by: Quentin Glidic <sardemff7+git@sardemff7.net>
Tested-by: Benoit Gschwind <gschwind@gnu-log.net>
Acked-by: Benoit Gschwind <gschwind@gnu-log.net>
[Pekka: rebased]
2016-06-23 17:44:54 +03:00