On Mon, Mar 09, 2015 at 11:27:06AM +0530, Archit Taneja wrote:
On 03/05/2015 09:14 PM, Daniel Vetter wrote:
On Thu, Mar 05, 2015 at 07:10:44AM -0500, Rob Clark wrote:
On Thu, Mar 5, 2015 at 5:06 AM, Archit Taneja <architt@xxxxxxxxxxxxxx> wrote:
On 02/23/2015 09:09 PM, Daniel Vetter wrote:
Rather than creating fb helper stub functions, maybe we could help each drm
On Mon, Feb 23, 2015 at 10:03:21AM -0500, Rob Clark wrote:
driver create two variants of functions needed by drm core(like
output_poll_changed and dev_lastclose), one variant supporting legacy fbdev,
and the other not?
So one quick thought.. building without fbdev would ideally/eventually
be a distro level decision, not a driver level decision.. so I think
it is *eventually* not a problem for it being a global drm level
decision. The only problem is right now some drivers support no-fbdev
config, and some do not. Maybe it is worth fixing that?
Yeah, if we get fbdev emulation Kconfig option then I think i915 and msm
should remove their own options and just use that. There's really no need
to have this per-driver, this is a question of what userspace expects and
so per-distro, independant of the driver.
Okay. If I understand right. We need to do something like this:
- Create a new Kconfig option that lets us emulate fbdev
- For drivers that already have a config for fbdev emulation, replace it
with our new emulation config.
- For drivers that assume fbdev emulation by default, select our new
emulation config in their respective Kconfigs.
Does this sound okay?
I suppose this could be the first step. Later we'd need to work on each
driver to work with and without the fbdev emulation Kconfig option.
See Rob's idea quoted above: Imo you should just do the conditional code
in the fbdev emulation helper in drm_fb_helper.c, not in drivers. There's
some additional code in i915 (and maybe also msm) to compile out (mostly
around taking/dropping console_lock) and we want to keep that. But that
should also just be using the same new config option. So:
- Add new config option DRM_FBDEV_EMULATION (default y for backwards
compat). Use that to stub out helpers in drm_fb_helper.c.
- Replace the msm/i915 specific options with the new, i.e. remove it from
Kconfig and use DRM_FBDEV_EMULATION in the code instead of the msm/i915
fbdev-less support for all other drivers will just magically happen
without the need for driver-specific code (except for the special locking
in i915 and similar things maybe).