On Wed, Apr 05, 2017 at 01:23:15PM +0800, Icenowy Zheng wrote:
2017年4月5日 10:27于 Chen-Yu Tsai wens@csie.org写道：
On Wed, Apr 5, 2017 at 3:53 AM, Icenowy Zheng icenowy@aosc.io wrote:
在 2017年04月05日 03:28, Sean Paul 写道:
On Thu, Mar 30, 2017 at 03:46:06AM +0800, Icenowy Zheng wrote:
As we are going to add support for the Allwinner DE2 Mixer in sun4i-drm driver, we will finally have two types of layer.
Abstract the layer type to void * and a ops struct, which contains the only function used by crtc -- get the drm_plane struct of the layer.
Signed-off-by: Icenowy Zheng icenowy@aosc.io
Refactored patch in v3.
drivers/gpu/drm/sun4i/sun4i_crtc.c | 19 +++++++++++-------- drivers/gpu/drm/sun4i/sun4i_crtc.h | 3 ++- drivers/gpu/drm/sun4i/sun4i_layer.c | 19 ++++++++++++++++++- drivers/gpu/drm/sun4i/sun4i_layer.h | 2 +- drivers/gpu/drm/sun4i/sunxi_layer.h | 17 +++++++++++++++++ 5 files changed, 49 insertions(+), 11 deletions(-) create mode 100644 drivers/gpu/drm/sun4i/sunxi_layer.h
diff --git a/drivers/gpu/drm/sun4i/sun4i_crtc.c b/drivers/gpu/drm/sun4i/sun4i_crtc.c index 3c876c3a356a..33854ee7f636 100644 --- a/drivers/gpu/drm/sun4i/sun4i_crtc.c +++ b/drivers/gpu/drm/sun4i/sun4i_crtc.c @@ -29,6 +29,7 @@ #include "sun4i_crtc.h" #include "sun4i_drv.h" #include "sun4i_layer.h" +#include "sunxi_layer.h" #include "sun4i_tcon.h"
static void sun4i_crtc_atomic_begin(struct drm_crtc *crtc, @@ -149,7 +150,7 @@ struct sun4i_crtc *sun4i_crtc_init(struct drm_device *drm, scrtc->tcon = tcon;
/* Create our layers */ - scrtc->layers = sun4i_layers_init(drm, scrtc->backend); + scrtc->layers = (void **)sun4i_layers_init(drm, scrtc); if (IS_ERR(scrtc->layers)) { dev_err(drm->dev, "Couldn't create the planes\n"); return NULL; @@ -157,14 +158,15 @@ struct sun4i_crtc *sun4i_crtc_init(struct drm_device *drm,
/* find primary and cursor planes for drm_crtc_init_with_planes */ for (i = 0; scrtc->layers[i]; i++) { - struct sun4i_layer *layer = scrtc->layers[i]; + void *layer = scrtc->layers[i]; + struct drm_plane *plane = scrtc->layer_ops->get_plane(layer);
- switch (layer->plane.type) { + switch (plane->type) { case DRM_PLANE_TYPE_PRIMARY: - primary = &layer->plane; + primary = plane; break; case DRM_PLANE_TYPE_CURSOR: - cursor = &layer->plane; + cursor = plane; break; default: break; @@ -190,10 +192,11 @@ struct sun4i_crtc *sun4i_crtc_init(struct drm_device *drm, /* Set possible_crtcs to this crtc for overlay planes */ for (i = 0; scrtc->layers[i]; i++) { uint32_t possible_crtcs = BIT(drm_crtc_index(&scrtc->crtc)); - struct sun4i_layer *layer = scrtc->layers[i]; + void *layer = scrtc->layers[i]; + struct drm_plane *plane = scrtc->layer_ops->get_plane(layer);
- if (layer->plane.type == DRM_PLANE_TYPE_OVERLAY) - layer->plane.possible_crtcs = possible_crtcs; + if (plane->type == DRM_PLANE_TYPE_OVERLAY) + plane->possible_crtcs = possible_crtcs; }
return scrtc; diff --git a/drivers/gpu/drm/sun4i/sun4i_crtc.h b/drivers/gpu/drm/sun4i/sun4i_crtc.h index 230cb8f0d601..a4036ee44cf8 100644 --- a/drivers/gpu/drm/sun4i/sun4i_crtc.h +++ b/drivers/gpu/drm/sun4i/sun4i_crtc.h @@ -19,7 +19,8 @@ struct sun4i_crtc {
struct sun4i_backend *backend; struct sun4i_tcon *tcon; - struct sun4i_layer **layers; + void **layers; + const struct sunxi_layer_ops *layer_ops;
I think you should probably take a different approach to abstract the layer type. How about creating
struct sunxi_layer { struct drm_plane plane; }
base and then subclassing that for sun4i and sun8i? By doing this you can avoid the nasty casting and you can also get rid of the get_plane() hook and layer_ops.
For the situation that using ** things are easily to get weird.
That code could be reworked, by initializing the layers directly within the crtc init code. If you look at rockchip's drm driver, you'll see they do this. There is a good reason to do it this way, as you need to first create the primary and cursor layers, pass them in when you create the crtc, then initialize any additional layers with the possible_crtcs bitmap.
But furthurly maybe more layers will be created for DE2 mixer, and may even depends on mixer type (On A83T/H3/A64/H5 mixer1 has fewer channel than mixer0).
You'll always have one primary and one cursor plane, no matter how much planes you support.
Maxime