... apart from SHM.
EGL and dmabuf buffers already have a gl_buffer_state created for them
when we first attach the weston_buffer. By turning
gl_surface_state::buffer into a pointer, we can just reference rather
than inline the gl_buffer_state.
SHM buffers are special, in that we don't keep individual copies of them
within the GL renderer. Instead, the GL surface has a texture allocated
with a shadow copy of the most up-to-date surface content. Handle this
by allocating and destroying gl_buffer_state every time we need to
respecify textures or somehow meaningfully change the parameters.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Similarly to EGL buffers, store the gl_buffer_state for a dmabuf buffer
inside weston_buffer, rather than on the linux_dmabuf_buffer. This
slightly simplifies our gl_buffer_state handling, and will be used later
to eliminate the egl_image refcounting.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Introduce a renderer_private hook for weston_buffer, and use this to
store a copy of the gl_buffer_state for EGL buffers (i.e. non-dmabuf, via
EGL_WL_bind_wayland_display).
As part of this, we create the EGLImage along with the weston_buffer
information, and just take a reference to it each time it is attached.
If you have bisected a failure to update surface content to this commit,
it very likely means that your EGL implementation requires images to be
recreated rather than only rebound in order to have their content
updated, which is contrary to specification.
Signed-off-by: Daniel Stone <daniels@collabora.com>
At the moment, attach_shm() will modify the gl_buffer_state in place,
then compare it and see if it differs enough to require a new one. That
rather mixes up the old and new worlds, so quite explicitly build up a
shadow gl_buffer_state with variables first before we change the one
which already exists.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Now that we can reliably access buffer dimensions from weston_buffer,
and gl-renderer isn't doing strange things with buffer widths, just use
that. The renderer interface is now unused and can be deleted.
Signed-off-by: Daniel Stone <daniels@collabora.com>
This was only used for what was presumably an attempt at an
optimisation, to force the texture's pitch in pixels to match the SHM
buffer. This is really unlikely to have ever made a difference, given
the alignments GPUs demand.
Signed-off-by: Daniel Stone <daniels@collabora.com>
It's just a shadow of weston_buffer.buffer_origin, which also has a
slightly more descriptive name.
Signed-off-by: Daniel Stone <daniels@collabora.com>
If we can reuse the textures we already have, just return early, rather
than putting all the work in a large indented body.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Allow the various attach handlers to access the existing buffer, only
referencing the new buffer when they have successfully attached.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Deduplicate the no-buffer and the import-fail case, and try to fall
through where we can. This will make it easier to shift the buffer
reference change later, so the attach subhandlers can reference the old
buffer when checking for compatibility.
Signed-off-by: Daniel Stone <daniels@collabora.com>
It's good to know if we succeeded or failed to import our buffers. This
will also later make for a more smooth transition when we start
returning a gl_buffer_state from them.
Signed-off-by: Daniel Stone <daniels@collabora.com>
This only happens for the legacy renderer, but still, might as well
clean up after ourselves when we can't import a secondary plane.
Signed-off-by: Daniel Stone <daniels@collabora.com>
gl_surface_state has a bunch of members which are tied to the input
buffer, rather than the surface per se.
Split them out into a separate gl_buffer_state member as a first step
towards sanitising its use and lifetime.
Signed-off-by: Daniel Stone <daniels@collabora.com>
No-one should be implementing an external renderer, so move the
interface out of the public header.
Signed-off-by: Daniel Stone <daniels@collabora.com>
This value is always `NULL`.
Silences:
`warning: ‘%s’ directive argument is null [-Wformat-overflow=]`
Signed-off-by: Robert Mader <robert.mader@collabora.com>
This fixes the following weston_view leak at compositor shutdown:
#0 0x7f4250247987 in __interceptor_calloc ../../../../src/libsanitizer/asan/asan_malloc_linux.cpp:154
#1 0x7f424fd37aa3 in zalloc ../include/libweston/zalloc.h:38
#2 0x7f424fd3a05f in weston_view_create ../libweston/compositor.c:386
#3 0x7f423be7be6a in weston_desktop_surface_create_desktop_view ../libweston-desktop/surface.c:364
#4 0x7f423be7c0a8 in weston_desktop_surface_create_view ../libweston-desktop/surface.c:404
#5 0x7f423beae91d in desktop_surface_added ../desktop-shell/shell.c:2273
#6 0x7f423be77db1 in weston_desktop_api_surface_added ../libweston-desktop/libweston-desktop.c:138
#7 0x7f423be80c73 in weston_desktop_xdg_toplevel_ensure_added ../libweston-desktop/xdg-shell.c:362
#8 0x7f423be8207a in weston_desktop_xdg_toplevel_committed ../libweston-desktop/xdg-shell.c:697
#9 0x7f423be84d45 in weston_desktop_xdg_surface_committed ../libweston-desktop/xdg-shell.c:1374
#10 0x7f423be7b382 in weston_desktop_surface_surface_committed ../libweston-desktop/surface.c:174
#11 0x7f424fd378a6 in wl_signal_emit /home/mvlad/install-amd64/include/wayland-server-core.h:481
#12 0x7f424fd510e2 in weston_surface_commit_state ../libweston/compositor.c:4062
#13 0x7f424fd51161 in weston_surface_commit ../libweston/compositor.c:4068
#14 0x7f424fd516ef in surface_commit ../libweston/compositor.c:4146
#15 0x7f424fb597e9 (/usr/lib/x86_64-linux-gnu/libffi.so.8+0x77e9)
With commit 'libweston, desktop-shell: Add a wrapper for weston_surface
reference' we've removed an explicit weston_view_destroy() call due to a
UAF which would've happen if we had close animations enabled, upon
terminating a client. In that patch I've incorrectly wrote this happened
when animations are off, but in fact it happened when they're on, see the
following trace:
READ of size 8 at 0x616000026498 thread T0
#0 0x7f757fba8797 in weston_signal_emit_mutable ../shared/signal.c:52
#1 0x7f757fb4bba1 in weston_view_destroy ../libweston/compositor.c:2269
#2 0x7f756bca89c0 in desktop_shell_destroy_surface ../desktop-shell/shell.c:275
#3 0x7f756bcb379e in fade_out_done_idle_cb ../desktop-shell/shell.c:2228
#4 0x7f757faec1da in wl_event_loop_dispatch_idle ../src/event-loop.c:969
#5 0x7f757faec31d in wl_event_loop_dispatch ../src/event-loop.c:1032
#6 0x7f757faea114 in wl_display_run ../src/wayland-server.c:1408
#7 0x7f757ff777ba in wet_main ../compositor/main.c:3589
#8 0x55f765c8d17d in main ../compositor/executable.c:33
#9 0x7f757fd997fc in __libc_start_main ../csu/libc-start.c:332
#10 0x55f765c8d099 in _start (/home/mvlad/install-amd64/bin/weston+0x1099)
0x616000026498 is located 24 bytes inside of 608-byte region [0x616000026480,0x6160000266e0)
freed by thread T0 here:
#0 0x7f758004c4d7 in __interceptor_free ../../../../src/libsanitizer/asan/asan_malloc_linux.cpp:127
#1 0x7f757fb4bdc8 in weston_view_destroy ../libweston/compositor.c:2295
#2 0x7f757fb4c14d in weston_surface_unref ../libweston/compositor.c:2334
#3 0x7f756bca898b in desktop_shell_destroy_surface ../desktop-shell/shell.c:273
#4 0x7f756bcb379e in fade_out_done_idle_cb ../desktop-shell/shell.c:2228
#5 0x7f757faec1da in wl_event_loop_dispatch_idle ../src/event-loop.c:969
This patch re-introduces it to avoid leaking the view upon compositor
shutdown, but it does it in tandem with weston_desktop_surface_unlink_view(),
(which was added in a later patch) and before weston_surface_unref() call.
This way we should be safe to terminate/close clients with additional views
created by libweston-desktop (pop-ups), but also in other different situations.
Verified it in the following circumstances:
- terminating a client with close animation on
- terminating a client with close animations off
- shutting down the compositor with clients running, with and
without close animations
- terminating top-level clients with additional pop-ups with both with
and without close animations
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
clang-13 complains about bitwise xor assigments like the following:
../libweston/noop-renderer.c:62:25: warning: variable 'unused' set but
not used [-Wunused-but-set-variable] volatile unsigned char unused = 0;
Use the __attribute__((unused)) instead.
Suggested-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Fix an apparent copy and paste error in resize code. I'm not sure anything
sets the relevant callback that would lead to height being different than
width, so there's no easy way to demonstrate a bug, but this change
appears to rectify the intent of the code.
Reported-by: Hideyuki Nagase <hideyukn@microsoft.com>
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
This documents the fact that other views are handled implictly by
libweston-desktop, and we shouldn't attempt to destroy indiscriminately
views that aren't created by desktop-shell.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
This patch fixes the following trace:
#0 0x7f07d1bcecfa in weston_desktop_surface_get_surface ../libweston-desktop/surface.c:585
#1 0x7f07d1bfc9b8 in move_grab_motion ../desktop-shell/shell.c:1499
#2 0x7f07e539f841 in notify_motion ../libweston/input.c:1794
#3 0x7f07e1e8ace4 in handle_pointer_motion ../libweston/libinput-device.c:132
#4 0x7f07e1e8cad5 in evdev_device_process_event ../libweston/libinput-device.c:535
#5 0x7f07e1e89311 in udev_input_process_event ../libweston/libinput-seat.c:208
#6 0x7f07e1e8932f in process_event ../libweston/libinput-seat.c:218
#7 0x7f07e1e8935f in process_events ../libweston/libinput-seat.c:228
#8 0x7f07e1e8940a in udev_input_dispatch ../libweston/libinput-seat.c:239
#9 0x7f07e1e89437 in libinput_source_dispatch ../libweston/libinput-seat.c:249
#10 0x7f07e53122b1 in wl_event_loop_dispatch ../src/event-loop.c:1027
#11 0x7f07e5310114 in wl_display_run ../src/wayland-server.c:1408
#12 0x7f07e579c7ba in wet_main ../compositor/main.c:3589
#13 0x555611d6b17d in main ../compositor/executable.c:33
#14 0x7f07e55be7fc in __libc_start_main ../csu/libc-start.c:332
#15 0x555611d6b099 in _start (/home/mvlad/install-amd64/bin/weston+0x1099)
A highly unlikely, but still valid operation, is to close/destroy the
window while still having it grabbed and moved around, basically having
an in-flight destruction of grabbed moving window. Another situation
would be that the client terminates abruptly (crashing for instance),
while being dragged which might take down the compositor.
This could happen for both touch/pointer grab operations and could cause
a NULL pointer access while accessing desktop_surface when being used
to retrieve the underlying weston_surface.
With this patch we check for a valid desktop_surface and return early
if that's not the case.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Moving weston_desktop_surface_unlink_view() to
desktop_shell_destroy_surface() makes sure we don't leak the underlying
weston_desktop_view when tearing/shutting down the compositor.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Creates a distinct view, separated from the one created by
libweston-desktop, in order to avoid a potential ownership fight with
libweston-desktop upon destroying the view. Upon weston_desktop_surface
destruction, libweston-desktop inflicts damage on the view it creates,
so we need the view to be alive at that time.
This wasn't such an issue before because we had different destruction paths but
with commit 'desktop-shell: Do not leave views in layers upon shell
destruction' all of the destruction paths now land in the same spot
+ handle compositor tear down.
Note as we still use the same weston_surface we'll keep the previous
construct where we were taking a reference to keep it alive.
The original view is destroyed when releasing the ownership, while for
the view created in this patch we handle the destruction directly upon
compositor tear down.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Similar to how we do it with drm_fb ref counts, increase a reference
count and return the same object.
Plug-in in desktop-shell when we map up the view in order to survive a
weston_surface destruction.
Astute readers will notice that this patch removes weston_view_destroy()
while keeping the balance between removing and adding a
weston_surface_unref() call in desktop_shell_destroy_surface().
The reason is to let weston_surface_unref() handle destruction on its
own. If multiple references are taken, then weston_surface_unref()
doesn't destroy the view, it just decreases the reference, with
a latter call to weston_surface_unref() to determine if the view
should be destroyed as well. This situation happens if we have
close animation enabled, were we have more than one reference taken: one
when mapping the view/surface and when when the surface itself was created,
(what we call, a weak reference).
If only a single reference is taken (for instance if we don't have close
animations enabled) then this weston_surface_unref()
call is inert as that reference is not set-up, leaving libweston to
handle the view destruction.
Following that with a weston_view_destroy() explicit call would cause a
UAF as the view was previous destroyed by a weston_surface_unref() call.
A side-effect of not keeping the weston_view_destroy() call would
happen when tearing down the compositor. If close animations are enabled,
weston_surface_unref() would not destroy the view, and because
weston_view_destroy() no longer exists, we would still have the
view in the other layers by the time we check-up if layers
have views present.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Calling weston_surface_unref() one time too many could be a sign we
haven't correctly increased the ref count for it.
Also, if we don't have a surface being passed, do no attempt to
use it.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Make it obvious that weston_surface has a reference counting happening
and destruction of the weston_surface happens when the last
weston_surface reference has been accounted for.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Unconditionally creating a surface feedback for each surface
creates unnecessary overhead and noise in the logs. Thus
create it when the first surface feedback resource for a
surface is requested and delete it again once all those
resources have been destroyed.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
Weston uses a timeout of 2 seconds before it sends scanout
tranches to clients in order to not trigger excessive buffer
reallocations in clients.
`simple-dmabuf-feedback` in turn counts redraws (200) before
exiting. That doesn't work on e.g. 144Hz screens, thus use a
timer here as well.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
For a surface that is already fullscreen making it maximized means to
exit fullscreen then set to it maximized. Instead of doing it, refuse to
do anything until the user explicitly performs that operation.
With this approach we follow other DE (desktop environments) which would
not perform any operation until the user exits fullscreen state.
Fixes#321
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
To me, the use of setup_search_param() makes the code harder to
understand than it needs to be. Replacing that function with open-coding
the struct cmlcms_color_transform_search_param initialization makes it
more clear that:
- get_surface_color_transform is the only one that actually uses a
surface to initialize it
- get_blend_to_output does not use an input profile at all
- get_sRGB_to_output and get_sRGB_to_blend hardcode the sRGB profile
like they should
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
I am going to need to add yet another output property to be set by a
color manager: HDR Static Metadata Type 1. With the old color manager
API design, I would have needed to add the fourth function pointer to be
called always in the same group as the previous three. This seemed more
convoluted than it needs to be.
Therefore collapse the three existing function pointers in the API into
just one that is resposible for setting up all three things.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This new struct collects all the things that a color manager needs to
set up when any colorimetry aspect of an output changes. The intention
is to make the color manager API less verbose.
In this first step, the new struct is added and replaces the fields in
weston_output.
The intention is for the following color manager API changes to
dynamically allocate this structure. Unfortunately, until that actually
happens, we need a temporary way to allocate it. That is
weston_output::colorout_, which will be removed in the next patch. This
keeps the patches more palatable for review at the cost of some
back-and-forth in code changes.
This is a pure refactoring, no functional changes.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
- Use more consistent style, e.g. the tree structure uses
the same indentation level throughout
- Swap format name and code for consistency with modifiers
- Use constants for ASCII art (taken from drm_info)
Signed-off-by: Simon Ser <contact@emersion.fr>
Trying to do HDR with XRGB8888 is a bit like using RGB565 on SDR: you
get visible color quantization and banding in gradients (without dithering
which Weston does not implement yet, and might not work too well for HDR
anyway).
Therefore, on any HDR mode, default output framebuffer format to 10 bpc
instead of 8 bpc.
Ideally we'd also optionally try 16F or 16 bpc formats, but automatic
fallbacks for those are more complicated to arrange. You can still
configure 16F or 16 bpc manually.
This patch also moves the default format setting from
drm_output_set_gbm_format() to drm_output_enable(), because setting the
default now requires eotf_mode. Frontends may call set_gbm_format()
first and set eotf_mode next. This does create an awkward situation for
outputs that a frontend disables and re-enables. This patch here makes
sure that the old output configuration remains, but changing eotf_mode
may not change the default format. One needs to call
set_gbm_format(NULL) to re-evaluate the default format. Resetting the
format on drm_output_deinit() would lose the current setting.
DRM_FORMAT_INVALID was introduced in libdrm 2.4.95 which we already
hard-depend on.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Program the connector property HDR_OUTPUT_METADATA based on the EOTF
mode of the output.
For now, this changes only the EOTF. The colorimetry and luminance are
left undefined, to be filled in by later patches. This should still be
enough to put a video sink into HDR mode, albeit the response is
probably unknown.
drm_output keeps track of the currently existing blob id. If the blob
contents need to be re-created, this blob would be destroyed and the
field set to zero. In this patch, there is no provision for runtime
changing of HDR metadata, so there is no code doing that.
Destroying the blob at arbitrary times is not a problem, because the
kernel keeps a reference to the data as long as the blob id remains with
KMS.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Daniel Stone
Robert Mader
Marius Vlad
Derek Foreman
Erik Faye-Lund
Hideyuki Nagase
Pekka Paalanen
Simon Ser