wayland on embedded

[Pekka Paalanen]

On Mon, 12 Mar 2012 14:51:15 -0700
L A <l.thee.a at gmail.com> wrote:

> I am working on a gles2/egl device (similar to Raspberry Pi) with a
> propriety display driver. What would it take to run Wayland on it? Is
> it possible to directly go to gles2/egl (no dri, no mesa)? Also do I
> really need xcb, glib (... etc) or would it be possible to live
> without them?

Hi,

the requirements are given here:
http://wayland.freedesktop.org/architecture.html#heading_toc_j_2

If that is not too clear, I have written a description of how it works
with Mesa currently, so you can get an idea of how it might work:
http://ppaalanen.blogspot.com/2012/03/what-does-egl-do-in-wayland-stack.html

To run a Wayland server, you need to be able to run GL ES 2 rendering
without X, i.e. on the framebuffer, somehow. When you get that, you
should already have software rendered Wayland clients working, too.
That is the shm buffer passing. You could write a new backend to Weston
to avoid implementing all of DRM and GBM APIs.

Any X dependencies are a Mesa artifact, and should be eliminated some
day. Weston does not use glib, only some demo clients do (did?).

For hardware accelerated Wayland clients, you need to implement Wayland
support in your EGL library. The major requirement here is the ability
to share graphics memory buffers between processes, preferrably
zero-copy.

DRI, Mesa, etc. things can be seen as the graphics stack implementation
details, you don't have to use or implement them as is. You do need
something that does similar things.


Hope this helps,
pq