On 07/22/2015 01:11 PM, Daniel Vetter wrote:
On Wed, Jul 22, 2015 at 12:29:36PM +0530, Archit Taneja wrote:
DRM drivers using drm_fb_helpers still call some fbdev core functions. This makes the driver depend on CONFIG_FB, resulting in complicated Kconfig options, and preventing us from creating a top level drm config option to enable/disable FBDEV emulation.
Create new drm_fb_helper functions that replace these fbdev functions.
In most cases, the new helper funcs simply wrap around the original fbdev functions. For a few (like framebufer_alloc), we actually do some work that is currently redundant across multiple drivers.
With these patches, the drivers don't call any fbdev functions directly. They are now called through functions in drm_fb_helper.c. We will later create a fbdev emulation config option to stub out the fb helpers.
The only exception is vmwgfx driver. This doesn't use drm_fb_helper. It creates a fb device how a driver in drivers/video/fbdev would. Maybe this needs to be converted to use drm_fb_helpers.
For more info, have a look at the threads: http://lists.freedesktop.org/archives/dri-devel/2015-March/078729.html http://lists.freedesktop.org/archives/dri-devel/2015-March/078975.html
v2:
- Remove if (info) checks in fb_sys functions
- Use 'if (fb_helper && fb_helper->fbdev)' checks where appropriate
- Add kerneldocs for the new helper funcs
- Follow drm way of aligning of arguments in func definitions
- Fixed build error in drm/virtio
- Fixed error cleanup path in drm/tegra
- Fixed build error in drm/rockchip
- Fixed return issue in drm/amdgpu
- Fixed 'unused variable' warnings in bochs, ast, udl and radeon
- Fixed stray goto labels in cirrus, exynos, mgag
When resending patches please also have changelogs per-patch, not just in the cover letter. Otherwise you have to constantly jump back&forth.
Also I'd drop the compile tested only from patches, just adds noise.
I'll remove the 'compile tested only' line from the patches.
About the changelogs per-patch, I'll add them for the drm_fb_helper.c changes, and for a few drm drivers where people found flaws in the changes. I'm going to skip for the ones which were build error/warning fixes found by kbots. I didn't intend the first version to be full proof, just wanted people to have a look.
Thanks, Archit