This consolidates duplicate code in meta_drm_buffer_gbm_blit_to_framebuffer
to use the newly added meta_drm_buffer_gbm_create_native_blit_image, which
also has the side-effect of caching creation of the EGLImage per GBM BO.
(cherry picked from commit c866ca2206ab279f639fa452f43812abd9be243a)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4027>
Signed-off-by: Mingi Sung <sungmg@saltyming.net>
Creating an EGLImage is rather expensive and is taking the bulk of the
time the secondary GPU copy path is using for each frame. By caching
these per GBM BO we avoid this expensive recreation, which seems to
significantly improve FPS throughput in these scenarios, e.g. an
AMD or Intel iGPU with an NVIDIA dGPU.
(cherry picked from commit 0255c1f5c099bd0a761d5c42cfdbc74e4375882d)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4027>
Signed-off-by: Mingi Sung <sungmg@saltyming.net>
cogl_framebuffer_finish can result in a CPU-side stall because it waits for
the primary GPU to flush and execute all commands that were queued before
that. By using a GPU-side EGLSync we can let the primary GPU inform us when
it is done with the queued commands instead. We then create another EGLSync
on the secondary GPU using the same fd so the primary GPU effectively
signals the secondary GPU when it is done rendering, causing the latter
to wait for the former before copying part of the frames it needs for
monitors attached to it directly.
This solves the corruption that cogl_framebuffer_finish also solved, but
without needing a CPU-side stall.
(cherry picked from commit 3a40413f7cf036622da3c22c609b703c138c709c)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4015>
Signed-off-by: Mingi Sung <sungmg@saltyming.net>
This adds meta_egl_create_sync and meta_egl_destroy_sync to be able to
create and dispose EGLSync objects, respectively, as well as
meta_egl_wait_sync to be able to wait for an EGLSync on the GPU.
(cherry picked from commit d2e70c15ec1cf6b6f7cd8c8825a5d92a1f89b35b)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4015>
Signed-off-by: Mingi Sung <sungmg@saltyming.net>
The sync_fd represents only the primary GPU work.
Fixes: c2621eca15 ("onscreen/native: Set latest cogl sync_fd on KMS update")
(cherry picked from commit 45666111d44201affc6a0eb99d1995458d3bf1a4)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4060>
Signed-off-by: Mingi Sung <sungmg@saltyming.net>
This describes the condition it represents, rather than what it's
currently used for.
(cherry picked from commit 8d3afca12684d37ca935b7e1963d48433ee6d8df)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4060>
Signed-off-by: Mingi Sung <sungmg@saltyming.net>
Since we use the buffer age to create a union of damage rectangles from the
last n frames, we often end up with the same rectangle multiple times if an
application happens to be changing that is in the same place. This in turn
results in duplicate vertices being generated and uploaded to the GPU in
the blitting and painting paths.
Spend some time (seems to take around 15 microseconds on average on my
system) dropping duplicate rectangles using a GHashTable. This partially
pays itself back CPU-wise by having to copy fewer rectangles in
copy_shared_framebuffer_gpu, looping and calculating less in
paint_egl_image. GPU-wise it uploads less data to the GPU and makes the
shaders invoke less.
(cherry picked from commit 07e021b3a4facec18d1db64deb74b955f55321e9)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4073>
Signed-off-by: Mingi Sung <sungmg@saltyming.net>
This provides damage rectangles to the secondary GPU in the copy path so
it, much like the primary GPU, can avoid updating parts of the framebuffer
that haven't been changed since the last frame(s). This works in tandem
with EGL_EXT_buffer_age to determine how old the contents of the current
framebuffer are (i.e. to know if there is anything to reuse from the last
frame).
(cherry picked from commit bf4a767ac3a36eac7a2fc612550a218ebfb9480d)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4073>
Signed-off-by: Mingi Sung <sungmg@saltyming.net>
Add meta_egl_query_surface to allow querying EGL surfaces for properties
such as EGL_BUFFER_AGE_EXT.
(cherry picked from commit 9dc0ac92fec1ee98314b5a648f462b7e962e6e34)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4073>
Signed-off-by: Mingi Sung <sungmg@saltyming.net>
During an onscreen swap, the cogl_onscreen_swap_buffers_with_damage()
function ensures that the Cogl renderer creates a sync object every
frame. This sync object is later shared with the Wayland clients that
utilize the explicit sync protocol.
However, in headless mode, the function mentioned above is not called.
As a result, the sync object the Cogl renderer stores seems to be not
created. This causes cogl_context_get_latest_sync_fd() function to
return an invalid sync fd, causing Mutter to not be able to materialize
the sync timeline point that the clients wait for when the explicit sync
protocol is in use.
This change simply adds a call to the cogl_framebuffer_flush() function
to the offscreen swap path to make sure that there is a sync object that
can be shared with the clients, which will be signalled when all the
queued operations before the swap are completed.
Signed-off-by: Doğukan Korkmaztürk <dkorkmazturk@nvidia.com>
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4056>
(cherry picked from commit 6a0cc1371c)
The meta_input_device_get_wacom_device function may return NULL, so
we should check it before use. I'm not 100% sure that the code as
currently written could reach here with a NULL wacom_device, but
that could change in the future.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4070>
lease_disappeared() was calling meta_drm_lease_disappeared() and
set_connectors_as_available().
meta_drm_lease_disappeared() emits the "revoked" signal, connected to
on_lease_revoked(), which marks the leased connectors as available by
calling set_connectors_as_available().
Therefore, calling set_connectors_as_available() again from
lease_disappeared() is redundant.
Remove the redundant call to set_connectors_as_available() and drop two
unused functions.
Fixes: fc44437876 ("wayland/drm-lease: Withdrawn leased connectors")
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/3675
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4029>
(cherry picked from commit 3199d4bbb8)
In multi-GPU setups, when one of the GPUs is used to render the desktop
and the other one has all of their devices available for lease,
meta_kms_impl_device_list_lessees() can be called after the device fd
has been unhold.
Make sure to hold and unhold it in this function.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4030>
(cherry picked from commit 45bf41bf94)
Because for `COGL_DRIVER_GL3`, `_cogl_context_get_gl_extensions` needs to
know that the GL version *really* is >= 3.0 before it calls `glGetStringi`
which didn't exist prior to GL 3.0 or ES 3.0.
This was causing crashes on Xilinx Mali implementations that only support
ES 2.0 (hence `glGetStringi` == NULL), but were being forced to call
that function before the GL version check which should tell you the
function isn't supported.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4019>
The NULL check was inverted, meaning we'd grab with no leader device.
That meant updates coming from the what-should-have-been leader device
getting lost because incorrectly being classified as non-leader of
the grab.
Fix this by only allowing to grab if we have a device, and always mark
the current tool device as the grab leader.
Fixes: e4004a7c4f ("wayland: Use the tool's current_tablet device instead of caching it")
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4033>
GLES 2.0 does not have RGB8 and RGBA8 as sized internal formats. There
is OES_rgb8_rgba8 which adds RGB8 and RGBA8 but only for
RenderbufferStorageOES and not for TexImage2D which I wrongly assumed.
It seems like there is currently no GLES2 extension which adds RGB8 and
RGBA8 to TexImage2D so we have no choice but to fall back to unsized
internal formats in those cases as long as we don't want to drop GLES2
support.
This should be fine in practice and we should get our 8bpc textures.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/3680
Fixes: 7f943613a8 ("cogl: Use sized internal renderable formats")
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4036>
meta_wayland_color_management_dispose func is only called when the compositor
is shutting down, in that case the wl_globals are already automatically removed.
meta_wayland_color_management_dispose calls wl_global_remove again,
this makes a SIGSEGV when color_management is enabled and mutter is being
shut down.
Stop calling wl_global_remove to fix it.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4032>
XPending() will do a recvmsg() syscall if there are no items in the queue.
In most cases, this is unnecessary because we know that there is data to be
read of the connection or there are items already read which simply need
to be processed.
Discovering both of those conditions can be done without recvmsg() in the
hot paths.
Before this path, every iteration of the main loop had the potential to
submit a recvmsg() syscall. This reduces that overhead drastically.
XFlush() on the other-hand knows if it needs to write data or not and will
do no IO in the case the buffer is empty.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/3653
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4006>
We only need to wake up the other side of the GAsyncQueue if we transition
from 0 to 1 item in the queue. Otherwise, we can be certain that the other
side has received a wakeup and will eventually flush the queue.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4007>
The wrong type resulted in a crash when returning from the function,
because g_slist_free was called instead of g_free for the old_struts
list data pointers.
Fixes: e005d035c0 ("boxes: Define cleanup function for MetaStrut and use auto-pointers")
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4028>
Track toplevels being saved, and save state some time after. This
will make session state somewhat remembered on shell crashes, as
long as there was time to snapshot the data in disk.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3825>
Allow saving the session gvdb file in the background, with as little
overhead in the main thread as possible. We still need to serialize
all created/deserialized MetaSessionState to a GVDB hashtable there,
in order to avoid these being poked from the async task thread.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3825>
The xdg_session_manager_v1 global interface is the generator
of xdg_session_v1 objects for clients. These will notify of an
unique ID that can be used for future instantiations.
Once a xdg_session_v1 object is obtained, toplevels can be added
to be managed by it, and clients may get a hint about whether the
toplevel was restored to a saved state.
Changes by Carlos Garnacho: Integrate with MetaSessionManager core
object. Flesh out event emission of xdg_session_v1 and
xdg_toplevel_session_v1 objects, handle sessions being
replaced/deleted.
Changes by Sebastian Wick:
* make lifetimes of xdg_sessions entirely determined by the wayland and
handle its destruction via the signal
* fix session destruction vs deletion
* do not drop refcount of replaced session state temporarily to make
sure the replacing session keeps the state
* disconnect signals of destroyed and replaced sessions
* disconnect window-unmanaging signal handler for
MetaWaylandXdgToplevelSession
* call wl_resource_destroy in xdg_toplevel_session_remove to make it a
destructor
* handle session being destroyed before topevel-sessions
* handle the toplevel going away before the topevel-sessions
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3825>
This object is a windowing-specific implementation of MetaSessionState,
allowing to save window state for toplevel surfaces of a Wayland client
using the xdg_session_management_v1 protocol.
This object is detached from windowing logic itself, and will be
integrated in later commits.
Changes from Carlos Garnacho: Integrate state serialization with
MetaSessionState and MetaSessionManager.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3825>
Make this core object own the MetaSessionManager, for the window management
code to access.
At this level, we will be able to integrate with systemd notification
system, and use systemd fdstore to keep the mapped memory warm for
us for the case of soft reboot. This is at the moment not implemented
here.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3825>
This core object will be the manager of "client sessions", allowing
the windowing-specific paths to generate MetaSessionState objects to
track their clients.
This object is unused at the moment, and will be integrated in later
commits.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3825>
This is an abstract base class to implement a "client session",
carrying the accounting of the windows, and allowing to serialize/read
their state into a Gvdb table.
Since different windowing backends may require slightly different
data to be saved for each window, this is meant to have windowing-specific
implementations.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3825>
