On Wed, Aug 12, 2015 at 03:52:12PM -0500, Rob Herring wrote:
I'm working on a DRM driver for PXA1928. Other than a stride alignment requirement of 16 bytes, I have no other reason not to use fbdev_cma. While I can adjust the stride for drm_gem_cma_dumb_create, I cannot do the same for drm_fbdev_cma_create without duplicating a bunch of code. This series allows fbdev_cma users to override the fb_probe function, so the stride can be adjusted.
It appears to me that rcar-du has a bug that it doesn't handle alignment requirements for this case as well. Probably just getting lucky with tested resolutions/bpp.
Also, AFAICT the Rockchip driver has no real reason to use a custom GEM allocator instead of the CMA one. It sets the DMA_ATTR_NO_KERNEL_MAPPING DMA attr, but that could easily be supported by the CMA allocator.
I think it'd be easier to review this with the driver at hand. That's also generally the requirement for merging new code - it needs an in-kernel user. -Daniel
Rob
Rob Herring (2): drm/cma: allow custom fb helper functions drm/cma: allow adjusting the pitch for CMA fbdev
drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_dc.c | 2 +- drivers/gpu/drm/drm_fb_cma_helper.c | 13 ++++++++++--- drivers/gpu/drm/imx/imx-drm-core.c | 2 +- drivers/gpu/drm/rcar-du/rcar_du_kms.c | 3 ++- drivers/gpu/drm/sti/sti_drm_drv.c | 2 +- drivers/gpu/drm/tilcdc/tilcdc_drv.c | 2 +- include/drm/drm_fb_cma_helper.h | 7 +++++++ include/drm/drm_fb_helper.h | 1 + 8 files changed, 24 insertions(+), 8 deletions(-)
-- 2.1.4
dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel