Adds linux-dmabuf support to the pixman renderer. The main reason is so
clients can use GPU-accelerated rendering even if the compositor is
using the pixman renderer for whatever reason. This also allows the
pixman renderer to import dmabufs (only RGB/ARGB formats, not YUV for
now).
Since after this change all renderers store a DRM FD, I move the drm_fd
from the individual implementations to the shared wlr_renderer base.
This means that renderer_autocreate() can set pixman's drm_fd for it and
the pixman renderer doesn't have to know about DRM at all.
Originally based on
961edfe44e
Allow direct access to the pixel data of linux_dmabuf_v1 buffers by
mmapping the FD. This causes a wait on any outstanding fences and also
triggers the DMA_BUF_SYNC mechanism to do any cache fiddling needed.
This doesn't support multi-planar formats (e.g. YUV420 from hardware
codecs)
I also fix the comment on wlr_renderer.render_buffer_caps (it's used for
textures, not the render target) and update
wlr_renderer_init_wl_display() and
wlr_linux_dmabuf_feedback_v1_init_with_options() to use the renderer's
own claimed buffer_caps instead of hardcoding DMABUF as required.
Loosely based on 46ef2cfa3c
The old approach of using a signal is fundamentally broken for a common
usecase: When the waiter is ready, it's common to immediately finish and
free any resources associated with it.
Because of the semantics of wl_signal_emit_mutable() this is UB.
wl_signal_emit_mutable() always excepts that the waiter hasn't been freed
until the signal has finished being emitted.
Instead of over engineering the solution, let's just add a callback required
by wlr_drm_syncobj_timeline_waiter_init(). In this callback, the implementation
is free to finish() or free() any resource it likes.
Instead of having separate getters for shm formats and DMA-BUF
formats, use the same pattern as wlr_output_impl.get_primary_formats
with a single function which takes buffer caps as input.
Sadly, the new API is not backwards compatible with the old API. Since
we have already switched all users in wlroots to the new API compositors
are already practically mandated to implement the new API. Let's get rid
of the old one since there is no point.
Based on five calls:
wlr_render_timer_create - creates a timer which can be reused across
frames on the same renderer
wlr_renderer_begin_buffer_pass - now takes a timer so that backends can
record when the rendering starts and finishes
wlr_render_timer_get_time - should be called as late as possible so that
queries can make their way back from the GPU
wlr_render_timer_destroy - self-explanatory
The timer is exposed as an opaque `struct wlr_render_timer` so that
backends can store whatever they want in there.
