Refactor the code constructing the connector name into a new function.
This makes create_output_for_connector() slightly easier to read.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
Refactor the code for choosing the initial mode for an output from
create_output_for_connector() to drm_output_choose_initial_mode().
This makes create_output_for_connector() slightly easier to read.
v2: Document everything.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
If a stopped repaint loop gets restarted due to posting of new
damage, and this restart of the repaint loop happens late in the
video refresh cycle, ie. already inside the repaint-window and
thereby after the composition deadline for the current frame,
then defer the actual output repaint to the composition deadline
of the next video refresh cycle by setting the repaint timer
accordingly.
This tries to make sure that:
a) Client(s) posting damage timely before the composition deadline
(video refresh duration - "repaint-window" duration) of the
current refresh cycle will trigger a repaint within the current
refresh cycle, thereby avoiding one extra frame of compositor
lag due to the needed restart of the repaint loop if the loop
was stopped. This allows them to benefit from the earlier
"instant repaint restart" commit to keep latency low.
b) Late clients which post damage close to the end of a refresh
cycle can't race other clients if the repaint loop is restarted.
Instead they will get deferred to the next compositor cycle,
just as if the repaint loop would have been already running -
the semantic of the "repaint-window" parameter is preserved.
This is especially important to prevent a very late client
from triggering a repaint very close to the vblank, which
would cause the compositor to certainly miss the vblank and
skip one frame and then cause a delay of another frame for
other clients which posted their damage in time for the
following frame. Iow. this provides clients with a more
predictable compositor timing and makes it easier for them
to latch onto the compositors repaint cycle.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
drm_output_start_repaint_loop() incurred a delay of
one refresh cycle by using a no-op page-flip to get
an accurate vblank timestamp as reference. This causes
unwanted lag whenever Weston exited its repaint loop, e.g.,
whenever an application wants to repaint with less than
full video refresh rate but still minimum lag.
Try to use the drmWaitVblank ioctl to get a proper
timestamp instantaneously without lag. If that does
not work, fall back to the old method of idle page-flip.
This optimization will work on any drm/kms driver
which supports high precision vblank timestamping.
As of Linux 4.0 these would be intel, radeon and
nouveau on all their supported gpu's.
On kms drivers without instant high precision timestamping
support, the kernel is supposed to return a timestamp
of zero when calling drmWaitVblank() to query the current
vblank count and time iff vblank irqs are currently
disabled, because the only way to get a valid timestamp
on such kms drivers is to enable vblank interrupts and
then wait a bit for the next vblank irq to take a new valid
timestamp. The caller is supposed to poll until at next
vblank irq it gets a valid non-zero timestamp if it needs
a timestamp.
This zero-timestamp signalling works up to Linux 3.17, but
got broken due to a regression in Linux 3.18 and later. On
Linux 3.18+ with kms drivers that don't have high precision
timestamping, the kernel erroneously returns a stale timestamp
from an earlier vblank, ie. the vblank count and timestamp are
mismatched. A patch is under way to fix this, but to deal with
broken kernels, we also check non-zero timestamps if they are
more than one refresh duration in the past, as this indicates
a stale/invalid timestamp, so we need to take the page-flip
fallback for restarting the repaint loop.
v2: Implement review suggestions by Pekka Paalanen, especially
extend the commit message to describe when and why the
instant restart won't work due to missing Linux kernel
functionality or a Linux kernel regression.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
v3: Fix timespec_to_nsec() which was computing picoseconds,
use the new timespec-util.h helpers.
v4: Rebased to master, split long lines.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
A helper to improbe readability.
Cc: Daniel Stone <daniels@collabora.com>
Cc: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
The wl_list_for_each operation on the free_buffers list should use
free_link not link, which is a different list.
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
When an output is being destroyed reassign the output of the views
that were in it, to be sure not to keep a dangling pointer which could
be used later on by calling weston_surface_assign_output() on the
view's surface.
Also make sure we send wl_surface.leave events to the surfaces that
were in that output.
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
The wl_list_for_each operation on the free_buffers list should use
free_link not link, which is a different list.
This fixes a crash when entering fullscreen mode when using the pixman
renderer on the wayland back-end.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
This commits starts to separate the libweston code from the weston
specific code. As such, the main() is moved, together with signals
handling and configuration handling.
The definition of DEFAULT_REPAINT_WINDOW is left in compositor.c, so the
config loading of repaint_msec is slightly modified to account that.
Acked-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
This commit adds three new exported functions:
- weston_compositor_create() returns a new weston_compositor instance,
initializing it as the now removed weston_compositor_init() did.
- weston_compositor_exit(compositor) asks the compositor to tear
down by calling the compositor's exit vfunc which is set by the
libweston application.
- weston_compositor_destroy(compositor) is called by the libweston
application when tearing down the compositor. The compositor is destroyed
and the memory freed.
Reviewed-by: Jon A. Cruz <jonc@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
error(1, ...) already will exit, per man page: "If status has a nonzero
value, then error() calls exit(3) to terminate the program using the
given value as the exit status." So exit(EXIT_FAILURE) is never
reached.
The EXIT_FAILURE macro is guaranteed to be non-zero. Typically it's
just 1, but on some systems (e.g. OpenVMS apparently) exit(1) means
success so EXIT_FAILURE there is defined to some other non-zero value.
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Marek Chalupa <mchqwerty@gmail.com>
This is a preliminary change for libweston, with no functional modifications.
Separate the backends and the core weston_compositor struct, by creating
the weston_compositor in the main(), and having the various backends extend
the weston_backend struct, an instance of which is returned by the backend
entry point.
This enable us to logically separate the compositor core from the backend,
allowing the core to be extended without messing with the backends.
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Allow proper handling of output->pipe > 1 to support
triple-head graphics cards etc. by using the "high-crtc"
support introduced in Linux 2.6.39 and libdrm 2.4.25
around May 2011.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
Initialize output->native_mode with the initially chosen
mode for an output, so weston_output_mode_switch_to_native()
has something to work with and can switch back from temporary
selected modes to the outputs native mode. Before, this was a
no-op.
This allows an output to switch back to its default mode if
a former toplevel fullscreen shell surface created via method
WL_SHELL_SURFACE_FULLSCREEN_METHOD_DRIVER gets destroyed, or
it gets demoted to non-fullscreen, or if modesetting on the
output failed for some reason.
v2: Modified and split into a separate patch from original
patch "Allow restore_output_mode() to work properly.",
as suggested by Derek Foreman.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Cc: Derek Foreman <derekf@osg.samsung.com>
Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
The matching logic in choose_mode() compared refresh rate
of a drm_mode candidate mode expressed in Hz against the
requested refresh rate of the target weston_mode expressed
in milliHz, so the match always failed and the logic always
ended up the mode with the highest refresh rate for a given
resolution, instead of the one matching the requested rate.
Match proper fields to fix this.
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
We used to rely on the order in which the
weston_compositor::destroy_signal callbacks happened, to not access
freed memory. Don't know when, but this broke at least with ivi-shell,
which caused crashes in random places on compositor shutdown.
Valgrind found the following:
Invalid write of size 8
at 0xC2EDC69: unbind_input_panel (input-panel-ivi.c:340)
by 0x4E3B6BB: destroy_resource (wayland-server.c:537)
by 0x4E3E085: for_each_helper.isra.0 (wayland-util.c:359)
by 0x4E3E60D: wl_map_for_each (wayland-util.c:365)
by 0x4E3BEC7: wl_client_destroy (wayland-server.c:675)
by 0x4182F2: text_backend_notifier_destroy (text-backend.c:1047)
by 0x4084FB: wl_signal_emit (wayland-server-core.h:264)
by 0x4084FB: main (compositor.c:5465)
Address 0x67ea360 is 208 bytes inside a block of size 232 free'd
at 0x4C2A6BC: free (vg_replace_malloc.c:473)
by 0x4084FB: wl_signal_emit (wayland-server-core.h:264)
by 0x4084FB: main (compositor.c:5465)
Invalid write of size 8
at 0x4E3E0D7: wl_list_remove (wayland-util.c:57)
by 0xC2EDEE9: destroy_input_panel_surface (input-panel-ivi.c:191)
by 0x4E3B6BB: destroy_resource (wayland-server.c:537)
by 0x4E3BC7B: wl_resource_destroy (wayland-server.c:550)
by 0x40DB8B: wl_signal_emit (wayland-server-core.h:264)
by 0x40DB8B: weston_surface_destroy (compositor.c:1883)
by 0x40DB8B: weston_surface_destroy (compositor.c:1873)
by 0x4E3B6BB: destroy_resource (wayland-server.c:537)
by 0x4E3E085: for_each_helper.isra.0 (wayland-util.c:359)
by 0x4E3E60D: wl_map_for_each (wayland-util.c:365)
by 0x4E3BEC7: wl_client_destroy (wayland-server.c:675)
by 0x4182F2: text_backend_notifier_destroy (text-backend.c:1047)
by 0x4084FB: wl_signal_emit (wayland-server-core.h:264)
by 0x4084FB: main (compositor.c:5465)
Address 0x67ea370 is 224 bytes inside a block of size 232 free'd
at 0x4C2A6BC: free (vg_replace_malloc.c:473)
by 0x4084FB: wl_signal_emit (wayland-server-core.h:264)
by 0x4084FB: main (compositor.c:5465)
Invalid write of size 8
at 0x4E3E0E7: wl_list_remove (wayland-util.c:58)
by 0xC2EDEE9: destroy_input_panel_surface (input-panel-ivi.c:191)
by 0x4E3B6BB: destroy_resource (wayland-server.c:537)
by 0x4E3BC7B: wl_resource_destroy (wayland-server.c:550)
by 0x40DB8B: wl_signal_emit (wayland-server-core.h:264)
by 0x40DB8B: weston_surface_destroy (compositor.c:1883)
by 0x40DB8B: weston_surface_destroy (compositor.c:1873)
by 0x4E3B6BB: destroy_resource (wayland-server.c:537)
by 0x4E3E085: for_each_helper.isra.0 (wayland-util.c:359)
by 0x4E3E60D: wl_map_for_each (wayland-util.c:365)
by 0x4E3BEC7: wl_client_destroy (wayland-server.c:675)
by 0x4182F2: text_backend_notifier_destroy (text-backend.c:1047)
by 0x4084FB: wl_signal_emit (wayland-server-core.h:264)
by 0x4084FB: main (compositor.c:5465)
Address 0x67ea368 is 216 bytes inside a block of size 232 free'd
at 0x4C2A6BC: free (vg_replace_malloc.c:473)
by 0x4084FB: wl_signal_emit (wayland-server-core.h:264)
by 0x4084FB: main (compositor.c:5465)
Looking at the first of these, unbind_input_panel() gets called when the
text-backend destroys its helper client which has bound to input_panel
interface. This happens after the shell's destroy_signal callback has
been called, so the shell has already been freed.
The other two errors come from
wl_list_remove(&input_panel_surface->link);
which has gone stale when the shell was destroyed
(shell->input_panel.surfaces list).
Rather than creating even more destroy listeners and hooking them up in
spaghetti, modify text-backend to not hook up to the compositor destroy
signal. Instead, make it the text_backend_init() callers' responsibility
to also call text_backend_destroy() appropriately, before the shell goes
away.
This fixed all the above Valgrind errors, and avoid a crash with
ivi-shell when exiting Weston.
Also using desktop-shell exhibited similar Valgrind errors which are
fixed by this patch, but those didn't happen to cause any crashes AFAIK.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
The headless-backend.so was missing in available backend list
Signed-off-by: JoonCheol Park <jooncheol@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Help messages were missing for some command line options.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
EGLGetDisplay() doesn't generate a GL error, so we shouldn't print one.
I've also renamed the goto labels so it's a little clearer when to use them.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Removed duplicate definitions of the container_of() macro and
refactored sources to use the single implementation.
Signed-off-by: Jon A. Cruz <jonc@osg.samsung.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
Removed multiple definitions of the MIN() macro from existing
locations and unified with a single definition. Updated sources
to use the shared version.
Signed-off-by: Jon A. Cruz <jonc@osg.samsung.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
To help reduce code duplication and also 'kitchen-sink' includes
the ARRAY_LENGTH macro was moved to a stand-alone file and
referenced from the sources consuming it. Other macros will be
added in subsequent passes.
Signed-off-by: Jon A. Cruz <jonc@osg.samsung.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
Using the parent '../' path component in #include statements makes
the codebase more rigid and is redundant due to proper -I use.
Signed-off-by: Jon A. Cruz <jonc@osg.samsung.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
This file was provided under both the Expat and X11 variants of the MIT
license. We don't need the latter, so remove it and leave just Expat.
And reformat the Expat license so it matches our standard boilerplate.
This allows a user to explicitly disable the input
method by setting path to blank;
Signed-off-by: Murray Calavera <murray.calavera@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Whether a input method is used should be the responsibility
of the shell because some shells may not want to implement
an input method at all
Signed-off-by: Murray Calavera <murray.calavera@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
a following patch will be moving text init call into shell
modules, which will be called much later than in current code
Signed-off-by: Murray Calavera <murray.calavera@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
We already have a pointer to the compositor so change seat->compositor to ec
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
We already have a pointer to the keyboard, so we can change all
seat->keyboard to keyboard.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
We already have a pointer to the keyboard, so we can change all
seat->keyboard to keyboard.
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
We already have a pointer to the keyboard, so we can change all
seat->keyboard to keyboard
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
The RDP compositor is usable without certificates and key in a very limited
number of cases (local usage using xfreerdp), so let's force the presence of
keys and certificates.
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
When a compositor window is closed, remove the output instead of just exiting.
(The "if (!input->output)" checks are kind of ugly - but I couldn't find
a better way to handle the output going away.)
Signed-off-by: Dima Ryazanov <dima@gmail.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
With the recent universal plane and atomic modeset / nuclear pageflip
development in the kernel, cursor content updates on Intel are currently causing
an extra wait for vblank. This drops Weston's framerate to a fraction by
2 when cursor contents update. This combined with the damage tracking
bug in Weston which causes cursor content updates on every frame the
cursor moves makes using hw cursors really bad.
It is possible that the Intel DRM driver will get fixed and cursor
updates there revert to their old behaviour on the contemporary KMS API.
However, it is hardware dependant whether cursor updates can happen
immediately. Some other hardware, especially ARM-related, may not be
able to do immediate updates. Therefore it is better to just not even
try - we should rely only on the lowest common denominator behaviour
between hardware and drivers as there is no and will not be any way to
reliably detect it.
Note, that while having different drivers do different things (immediate
update vs. update that gets latched on the next vblank), we cannot
rearrange the contemporary KMS API calls such that it would always work
fine. Either some hardware would update the cursor too early, or other
hardware would update the cursor too late and perhaps cause the
framerate decimation.
Mark hardware cursors broken by default. This avoids using them, and
works around the immediate problem of framerate issues in Weston. This
follows the same reasoning why hardware overlay planes have been
disabled by default for a long time.
This disablement will be removed once the current code for hardware
planes and cursors is replaced with code using the atomic KMS API.
The Intel driver change that exposed this problem is
38f3ce3af5
which is first included in Linux 4.0-rc1.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Cc: nerdopolis <bluescreen_avenger@verizon.net>
Cc: Daniel Stone <daniel@fooishbar.org>
Cc: Giulio Camuffo <giuliocamuffo@gmail.com>
Reviewed-By: David FORT <contact@hardening-consulting.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
If the GL implementation doesn't provide an XRGB visual we may still be
able to proceed with an ARGB one. Since we're not changing the scanout
buffer format, and our current rendering loop always results in saturated
alpha in the frame buffer, it should be Just Fine(tm) - and probably better
than just exiting.
This is a workaround for https://bugs.freedesktop.org/show_bug.cgi?id=89689
Reviewed-By: Bryce Harrington <bryce@osg.samsung.com>
Tested-By: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-By: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Pekka Paalanen
Mario Kleiner
Bryce Harrington
Giulio Camuffo
Derek Foreman
JoonCheol Park
Murray Calavera
Jon Cruz
Daniel Stone
David FORT
Dima Ryazanov