Re: [RFC PATCH 3/9] drm/rockchip: Convert to support atomic API

From: Daniel Stone
Date: Tue Dec 01 2015 - 03:18:12 EST


Hi Mark,

On 1 December 2015 at 03:26, Mark Yao <mark.yao@xxxxxxxxxxxxxx> wrote:
> +static void rockchip_atomic_wait_for_complete(struct drm_atomic_state *state)
> +{
> + struct drm_crtc_state *crtc_state;
> + struct drm_crtc *crtc;
> + int i;
> +
> + for_each_crtc_in_state(state, crtc, crtc_state, i) {
> + if (!crtc->state->active)
> + continue;
> +
> + WARN_ON(drm_crtc_vblank_get(crtc));
> + }
> +
> + for_each_crtc_in_state(state, crtc, crtc_state, i) {
> + if (!crtc->state->active)
> + continue;
> +
> + rockchip_crtc_wait_for_update(crtc);
> + }

I'd be much more comfortable if this passed in an explicit pointer to
state, or an address to wait for, rather than have wait_for_complete
dig out state with no locking. The latter is potentially racy for
async operations.

> +struct vop_plane_state {
> + struct drm_plane_state base;
> + dma_addr_t dma_addr[ROCKCHIP_MAX_FB_BUFFER];

Can you get rid of this by just using the pointer to a
rockchip_gem_object already provided?

> -static int vop_update_plane_event(struct drm_plane *plane,
> - struct drm_crtc *crtc,
> - struct drm_framebuffer *fb, int crtc_x,
> - int crtc_y, unsigned int crtc_w,
> - unsigned int crtc_h, uint32_t src_x,
> - uint32_t src_y, uint32_t src_w,
> - uint32_t src_h,
> - struct drm_pending_vblank_event *event)
> +static int vop_plane_atomic_check(struct drm_plane *plane,
> + struct drm_plane_state *state)
> {
> + struct drm_crtc *crtc = state->crtc;
> + struct drm_framebuffer *fb = state->fb;
> struct vop_win *vop_win = to_vop_win(plane);
> + struct vop_plane_state *vop_plane_state = to_vop_plane_state(state);
> const struct vop_win_data *win = vop_win->data;
> - struct vop *vop = to_vop(crtc);
> struct drm_gem_object *obj;
> struct rockchip_gem_object *rk_obj;
> - struct drm_gem_object *uv_obj;
> - struct rockchip_gem_object *rk_uv_obj;
> unsigned long offset;
> - unsigned int actual_w;
> - unsigned int actual_h;
> - unsigned int dsp_stx;
> - unsigned int dsp_sty;
> - unsigned int y_vir_stride;
> - unsigned int uv_vir_stride = 0;
> - dma_addr_t yrgb_mst;
> - dma_addr_t uv_mst = 0;
> - enum vop_data_format format;
> - uint32_t val;
> - bool is_alpha;
> - bool rb_swap;
> bool is_yuv;
> bool visible;
> int ret;
> - struct drm_rect dest = {
> - .x1 = crtc_x,
> - .y1 = crtc_y,
> - .x2 = crtc_x + crtc_w,
> - .y2 = crtc_y + crtc_h,
> - };
> - struct drm_rect src = {
> - /* 16.16 fixed point */
> - .x1 = src_x,
> - .y1 = src_y,
> - .x2 = src_x + src_w,
> - .y2 = src_y + src_h,
> - };
> - const struct drm_rect clip = {
> - .x2 = crtc->mode.hdisplay,
> - .y2 = crtc->mode.vdisplay,
> - };
> - bool can_position = plane->type != DRM_PLANE_TYPE_PRIMARY;
> + struct drm_rect *dest = &vop_plane_state->dest;
> + struct drm_rect *src = &vop_plane_state->src;
> + struct drm_rect clip;
> int min_scale = win->phy->scl ? FRAC_16_16(1, 8) :
> DRM_PLANE_HELPER_NO_SCALING;
> int max_scale = win->phy->scl ? FRAC_16_16(8, 1) :
> DRM_PLANE_HELPER_NO_SCALING;
>
> - ret = drm_plane_helper_check_update(plane, crtc, fb,
> - &src, &dest, &clip,
> + crtc = crtc ? crtc : plane->state->crtc;
> + /*
> + * Both crtc or plane->state->crtc can be null.
> + */
> + if (!crtc || !fb)
> + goto out_disable;
> + src->x1 = state->src_x;
> + src->y1 = state->src_y;
> + src->x2 = state->src_x + state->src_w;
> + src->y2 = state->src_y + state->src_h;
> + dest->x1 = state->crtc_x;
> + dest->y1 = state->crtc_y;
> + dest->x2 = state->crtc_x + state->crtc_w;
> + dest->y2 = state->crtc_y + state->crtc_h;
> +
> + clip.x1 = 0;
> + clip.y1 = 0;
> + clip.x2 = crtc->mode.hdisplay;
> + clip.y2 = crtc->mode.vdisplay;
> +
> + ret = drm_plane_helper_check_update(plane, crtc, state->fb,
> + src, dest, &clip,
> min_scale,
> max_scale,
> - can_position, false, &visible);
> + true, true, &visible);
> if (ret)
> return ret;
>
> if (!visible)
> - return 0;
> -
> - is_alpha = is_alpha_support(fb->pixel_format);
> - rb_swap = has_rb_swapped(fb->pixel_format);
> - is_yuv = is_yuv_support(fb->pixel_format);
> + goto out_disable;
>
> - format = vop_convert_format(fb->pixel_format);
> - if (format < 0)
> - return format;
> + vop_plane_state->format = vop_convert_format(fb->pixel_format);
> + if (vop_plane_state->format < 0)
> + return vop_plane_state->format;
>
> - obj = rockchip_fb_get_gem_obj(fb, 0);
> - if (!obj) {
> - DRM_ERROR("fail to get rockchip gem object from framebuffer\n");
> - return -EINVAL;
> - }
> -
> - rk_obj = to_rockchip_obj(obj);
> + is_yuv = is_yuv_support(fb->pixel_format);
>
> if (is_yuv) {
> /*
> * Src.x1 can be odd when do clip, but yuv plane start point
> * need align with 2 pixel.
> */
> - val = (src.x1 >> 16) % 2;
> - src.x1 += val << 16;
> - src.x2 += val << 16;
> + uint32_t temp = (src->x1 >> 16) % 2;
> +
> + src->x1 += temp << 16;
> + src->x2 += temp << 16;
> }

I know this isn't new, but moving the plane around is bad. If the user
gives you a pixel boundary that you can't actually use, please reject
the configuration rather than silently trying to fix it up.

> -static void vop_plane_destroy(struct drm_plane *plane)
> +static void vop_atomic_plane_destroy_state(struct drm_plane *plane,
> + struct drm_plane_state *state)
> {
> - vop_disable_plane(plane);
> - drm_plane_cleanup(plane);
> + struct vop_plane_state *vop_state = to_vop_plane_state(state);
> +
> + if (state->fb)
> + drm_framebuffer_unreference(state->fb);
> +
> + kfree(vop_state);
> }

You can replace this hook with drm_atomic_helper_plane_destroy_state.

> -static void vop_win_state_complete(struct vop_win *vop_win,
> - struct vop_win_state *state)
> -{
> - struct vop *vop = vop_win->vop;
> - struct drm_crtc *crtc = &vop->crtc;
> - struct drm_device *drm = crtc->dev;
> - unsigned long flags;
> -
> - if (state->event) {
> - spin_lock_irqsave(&drm->event_lock, flags);
> - drm_crtc_send_vblank_event(crtc, state->event);
> - spin_unlock_irqrestore(&drm->event_lock, flags);
> - }

Ah, I see this is based on top of the patches I sent to fix pageflips?
If they have been merged, I would like to send you some others to
merge as well, which fix an OOPS when you close Weston. I didn't send
them yet as I didn't get a response to the previous patchset, so did
not know you had merged it. There are a lot of other correctness fixes
I had in there (e.g. using the CRTC vblank functions, some locking
fixes), but if this code is going to be thrown away then I will just
discard most of them as well.

Still, I would like to prepare another series for you to merge through
drm-fixes, to be able to run Weston (the same-fb-flip patch I sent
along with the drm_crtc_send_vblank_event conversion, also adding a
preclose hook to discard events when clients exit).

Cheers,
Daniel
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/