The compositor shouldn't write to client buffers if the client
attaches a DMA-BUF to a wl_surface, then attaches a shm buffer.
Make gles2_texture_write_pixels return an error to prevent this
from happening.
The get_drm_fd was made available in an internal header with a53ab146f. Move it
now to the public header so consumers opting in to the unstable interfaces can
make use of it.
PRIME support for buffer sharing has become mandatory since the renderer
rewrite. Make sure we check for the appropriate capabilities in backend,
allocator and renderer.
See also #2819.
All backends use the GBM platform. We can't use it to figure out
whether the DRM backend is used anymore.
Let's just try to always request a high-priority EGL context. Failing
to do so is not fatal.
Split render/display setups have two separate devices: one display-only
with a primary node, and one render-only with a render node. However
in these cases the EGL implementation and the Wayland compositor will
advertise the display device instead of the render device [1]. The EGL
implementation will magically open the render device when the display
device is passed in.
So just pass the display device as if it were a render device. Maybe in
the future Mesa will advertise the render device instead and we'll be
able to remove this workaround.
[1]: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4178
Compute only the transform matrix in the output. The projection matrix
will be calculated inside the gles2 renderer when we start rendering.
The goal is to help the pixman rendering process.
Mesa provides YUV shaders, and can import multi-planar YUV DMA-BUFs
as a single EGLImage. Remove the arbitrary limitation.
If the driver doesn't support importing YUV as a single EGLImage,
the import will fail and the result will be the same anyways.
Clamping texture coordinates prevents OpenGL from blending the left and
right edge (or top and bottom edge) when scaling textures with GL_LINEAR
filtering. This prevents visual artifacts like swaywm/sway#5809.
Per discussion on IRC, this behaviour is made default. Compositors that want
the wrapping behaviour (e.g. for tiled patterns) can override this by doing:
struct wlr_gles2_texture_attribs attribs;
wlr_gles2_texture_get_attribs(texture, &attribs);
glBindTexture(attribs.target, attribs.tex);
glTexParameteri(attribs.target, GL_TEXTURE_WRAP_S, GL_REPEAT);
glTexParameteri(attribs.target, GL_TEXTURE_WRAP_T, GL_REPEAT);
glBindTexture(attribs.target, 0);
As surfaces are no longer going to be used for wlr_egl, I may as well just go and add this check as it is needed for safety whenever surface-less rendering is being used.
The creation of `wlr_egl` will fail is the device extension
EGL_MESA_device_software is defined. The creation process is allowed to
continue only if the environment variable `WLR_RENDERER_ALLOW_SOFTWARE`
is defined to the value 1.
When gbm_bo_create is used and the GBM implementation does support
modifiers, gbm_bo_get_modifier may return something other than
DRM_FORMAT_MOD_INVALID. This can cause issues with the rest of the
stack (e.g. EGL or KMS) in case these don't support modifiers.
Instead, force the modifier to INVALID, to make sure no one uses
modifiers.
Instead of lazily exporting BOs to DMA-BUFs, do it on init. This is
the only way to use the buffer, so there's no point in deferring
that work.
This is preparatory work for the next commit.
The GBM device needs to be destroyed after the EGL display.
==50931==ERROR: AddressSanitizer: SEGV on unknown address 0x7fe40a000049 (pc 0x7fe446121d30 bp 0x60400001bbd0 sp 0x7ffc99c774d0 T0)
==50931==The signal is caused by a READ memory access.
#0 0x7fe446121d30 (/usr/lib/dri/radeonsi_dri.so+0x5f0d30)
#1 0x7fe4474717bd (/usr/lib/../lib/libEGL_mesa.so.0+0x177bd)
#2 0x7fe4474677d9 (/usr/lib/../lib/libEGL_mesa.so.0+0xd7d9)
#3 0x7fe44cca7b6f in wlr_egl_destroy ../subprojects/wlroots/render/egl.c:379
#4 0x7fe44ccc2626 in gles2_destroy ../subprojects/wlroots/render/gles2/renderer.c:705
#5 0x7fe44ccb5041 in wlr_renderer_destroy ../subprojects/wlroots/render/wlr_renderer.c:37
#6 0x7fe44cd17850 in backend_destroy ../subprojects/wlroots/backend/wayland/backend.c:296
#7 0x7fe44ccca4de in wlr_backend_destroy ../subprojects/wlroots/backend/backend.c:48
#8 0x7fe44cd11b21 in multi_backend_destroy ../subprojects/wlroots/backend/multi/backend.c:58
#9 0x7fe44cd125b0 in handle_display_destroy ../subprojects/wlroots/backend/multi/backend.c:125
#10 0x7fe44c315e0e (/usr/lib/libwayland-server.so.0+0x8e0e)
#11 0x7fe44c3165a6 in wl_display_destroy (/usr/lib/libwayland-server.so.0+0x95a6)
#12 0x55a2c8870683 in server_fini ../sway/server.c:203
#13 0x55a2c886cbf2 in main ../sway/main.c:436
#14 0x7fe44b77c151 in __libc_start_main (/usr/lib/libc.so.6+0x28151)
#15 0x55a2c883172d in _start (/home/simon/src/sway/build/sway/sway+0x33472d)
Instead of requiring callers to manually make the EGL context current
before binding a buffer and unsetting it after unbinding a buffer, do
it inside wlr_renderer_bind_buffer.
This hides renderer-specific implementation details inside the
wlr_renderer interface. Non-GLES2 renderers may not use EGL.
This removes all EGL dependencies from the backends.
References: https://github.com/swaywm/wlroots/issues/2618
References: https://github.com/swaywm/wlroots/pull/2615#issuecomment-756687006
It can be surprising and unexpected that texture operations (such as
texture upload and import) change the current EGL context, especially
when it's done under-the-hood by wlroots in response to wl_surface
requests.
To prevent surprises, save and restore the previous EGL context.
