[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CABymUCMcKwrUFZ=KJdS852-KkWPaGGigHr2Jo8B17_oZa6HEeA@mail.gmail.com>
Date: Fri, 11 Oct 2024 16:12:56 +0800
From: Jun Nie <jun.nie@...aro.org>
To: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
Cc: Rob Clark <robdclark@...il.com>, Abhinav Kumar <quic_abhinavk@...cinc.com>,
Sean Paul <sean@...rly.run>, Marijn Suijten <marijn.suijten@...ainline.org>,
David Airlie <airlied@...il.com>, Daniel Vetter <daniel@...ll.ch>, linux-arm-msm@...r.kernel.org,
dri-devel@...ts.freedesktop.org, freedreno@...ts.freedesktop.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 12/14] drm/msm/dpu: support plane splitting in
quad-pipe case
Dmitry Baryshkov <dmitry.baryshkov@...aro.org> 于2024年10月10日周四 21:29写道:
>
> On Wed, Oct 09, 2024 at 04:50:25PM GMT, Jun Nie wrote:
> > Clip plane into pipes per left and right half screen ROI if topology
> > is quad pipe.
>
> Why? Please provide an explanation for the reviewers not knowing the
> details.
The content of every half of screen is sent out via one interface in
dual-DSI case. The content for every interface is blended by a LM
pair, thus no content of any pipe shall cross the LM pair. We need
to clip plane into pipes per left and right half screen ROI if topology
is quad pipe.
>
> > Then split the clipped rectangle by half if the rectangle
> > width still exceeds width limit.
> >
> > Signed-off-by: Jun Nie <jun.nie@...aro.org>
> > ---
> > drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c | 7 +++
> > drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h | 6 ++
> > drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c | 99 ++++++++++++++++++++++---------
> > 3 files changed, 84 insertions(+), 28 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
> > index 66f745399a602..d2aca0a9493d5 100644
> > --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
> > +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
> > @@ -1310,6 +1310,13 @@ int dpu_crtc_vblank(struct drm_crtc *crtc, bool en)
> > return 0;
> > }
> >
> > +unsigned int dpu_crtc_get_lm_num(const struct drm_crtc_state *state)
>
> I think the DPU driver uses num_foo rather than foo_num
dpu_crtc_get_num_lm()
>
> > +{
> > + struct dpu_crtc_state *cstate = to_dpu_crtc_state(state);
> > +
> > + return cstate->num_mixers;
> > +}
> > +
> > #ifdef CONFIG_DEBUG_FS
> > static int _dpu_debugfs_status_show(struct seq_file *s, void *data)
> > {
> > diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h
> > index 5260e2440f059..ee7cf71f89fc7 100644
> > --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h
> > +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h
> > @@ -304,4 +304,10 @@ static inline enum dpu_crtc_client_type dpu_crtc_get_client_type(
> >
> > void dpu_crtc_frame_event_cb(struct drm_crtc *crtc, u32 event);
> >
> > +/**
> > + * dpu_crtc_get_lm_num - Get mixer number in this CRTC pipeline
> > + * state: Pointer to drm crtc state object
> > + */
> > +unsigned int dpu_crtc_get_lm_num(const struct drm_crtc_state *state);
> > +
> > #endif /* _DPU_CRTC_H_ */
> > diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c
> > index 898fc2937954e..480a1b46aba72 100644
> > --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c
> > +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c
> > @@ -837,10 +837,12 @@ static int dpu_plane_atomic_check_nopipe(struct drm_plane *plane,
> > struct dpu_kms *kms = _dpu_plane_get_kms(&pdpu->base);
> > u64 max_mdp_clk_rate = kms->perf.max_core_clk_rate;
> > struct dpu_plane_state *pstate = to_dpu_plane_state(new_plane_state);
> > - struct dpu_sw_pipe_cfg *pipe_cfg;
> > - struct dpu_sw_pipe_cfg *r_pipe_cfg;
> > + struct dpu_sw_pipe_cfg pipe_cfg;
> > struct drm_rect fb_rect = { 0 };
> > + const struct drm_display_mode *mode = &crtc_state->adjusted_mode;
> > uint32_t max_linewidth;
> > + u32 lm_num;
> > + int lmcfg_id, lmcfg_num;
> >
> > min_scale = FRAC_16_16(1, MAX_UPSCALE_RATIO);
> > max_scale = MAX_DOWNSCALE_RATIO << 16;
> > @@ -863,13 +865,10 @@ static int dpu_plane_atomic_check_nopipe(struct drm_plane *plane,
> > return -EINVAL;
> > }
> >
> > - /* move the assignment here, to ease handling to another pairs later */
> > - pipe_cfg = &pstate->pipe_cfg[0];
> > - r_pipe_cfg = &pstate->pipe_cfg[1];
> > - /* state->src is 16.16, src_rect is not */
> > - drm_rect_fp_to_int(&pipe_cfg->src_rect, &new_plane_state->src);
> > + lm_num = dpu_crtc_get_lm_num(crtc_state);
> >
> > - pipe_cfg->dst_rect = new_plane_state->dst;
> > + /* state->src is 16.16, src_rect is not */
> > + drm_rect_fp_to_int(&pipe_cfg.src_rect, &new_plane_state->src);
> >
> > fb_rect.x2 = new_plane_state->fb->width;
> > fb_rect.y2 = new_plane_state->fb->height;
> > @@ -884,34 +883,78 @@ static int dpu_plane_atomic_check_nopipe(struct drm_plane *plane,
> >
> > max_linewidth = pdpu->catalog->caps->max_linewidth;
> >
> > - drm_rect_rotate(&pipe_cfg->src_rect,
> > + drm_rect_rotate(&pipe_cfg.src_rect,
> > new_plane_state->fb->width, new_plane_state->fb->height,
> > new_plane_state->rotation);
> >
> > - if ((drm_rect_width(&pipe_cfg->src_rect) > max_linewidth) ||
> > - _dpu_plane_calc_clk(&crtc_state->adjusted_mode, pipe_cfg) > max_mdp_clk_rate) {
> > - if (drm_rect_width(&pipe_cfg->src_rect) > 2 * max_linewidth) {
> > - DPU_DEBUG_PLANE(pdpu, "invalid src " DRM_RECT_FMT " line:%u\n",
> > - DRM_RECT_ARG(&pipe_cfg->src_rect), max_linewidth);
> > - return -E2BIG;
> > + /*
> > + * We have 1 mixer pair cfg for 1:1:1 and 2:2:1 topology, 2 mixer pair
> > + * configs for left and right half screen in case of 4:4:2 topology.
> > + * But we may have 2 rect to split plane with 1 config for 2:2:1.
> > + * So need to handle super wide plane splitting, and plane on right half
> > + * for quad-pipe case. Check dest rectangle left/right clipping
> > + * first, then check super wide rectangle splitting in every half next.
> > + */
> > + lmcfg_num = (lm_num + 1) / 2;
>
> num_stages?
OK. Then lmcfg_id -> stage_id
>
> > + /* iterate mixer configs for this plane, to separate left/right with the id */
> > + for (lmcfg_id = 0; lmcfg_id < lmcfg_num; lmcfg_id++) {
> > + struct drm_rect mixer_rect = {lmcfg_id * mode->hdisplay / lmcfg_num, 0,
> > + (lmcfg_id + 1) * mode->hdisplay / lmcfg_num, mode->vdisplay};
> > + int cfg_idx = lmcfg_id * PIPES_PER_LM_PAIR;
> > + struct dpu_sw_pipe_cfg *cur_pipecfg = &pstate->pipe_cfg[cfg_idx];
> > +
> > + drm_rect_fp_to_int(&cur_pipecfg->src_rect, &new_plane_state->src);
> > + cur_pipecfg->dst_rect = new_plane_state->dst;
> > +
> > + DPU_DEBUG_PLANE(pdpu, "checking src " DRM_RECT_FMT
> > + " vs clip window " DRM_RECT_FMT "\n",
> > + DRM_RECT_ARG(&cur_pipecfg->src_rect),
> > + DRM_RECT_ARG(&mixer_rect));
> > +
> > + /* If this plane does not fall into mixer rect, check next mixer rect */
> > + if (!drm_rect_clip_scaled(&cur_pipecfg->src_rect, &cur_pipecfg->dst_rect, &mixer_rect)) {
> > + memset(&pstate->pipe_cfg[cfg_idx], 0, 2 * sizeof(struct dpu_sw_pipe_cfg));
> > + memset(&pstate->pipe[cfg_idx], 0, 2 * sizeof(struct dpu_sw_pipe));
> > + continue;
> > }
> >
> > - *r_pipe_cfg = *pipe_cfg;
> > - pipe_cfg->src_rect.x2 = (pipe_cfg->src_rect.x1 + pipe_cfg->src_rect.x2) >> 1;
> > - pipe_cfg->dst_rect.x2 = (pipe_cfg->dst_rect.x1 + pipe_cfg->dst_rect.x2) >> 1;
> > - r_pipe_cfg->src_rect.x1 = pipe_cfg->src_rect.x2;
> > - r_pipe_cfg->dst_rect.x1 = pipe_cfg->dst_rect.x2;
> > - } else {
> > - memset(r_pipe_cfg, 0, sizeof(*r_pipe_cfg));
> > - }
> > + cur_pipecfg->valid = true;
>
> ... and checks have been broken up to now. This isn't good.
Will move this patch before the plane checking patch.
>
> > + cur_pipecfg->dst_rect.x1 -= mixer_rect.x1;
> > + cur_pipecfg->dst_rect.x2 -= mixer_rect.x1;
> > +
> > + DPU_DEBUG_PLANE(pdpu, "Got clip src:" DRM_RECT_FMT " dst: " DRM_RECT_FMT "\n",
> > + DRM_RECT_ARG(&cur_pipecfg->src_rect), DRM_RECT_ARG(&cur_pipecfg->dst_rect));
> > +
> > + /* Split super wide rect into 2 rect */
> > + if ((drm_rect_width(&cur_pipecfg->src_rect) > max_linewidth) ||
> > + _dpu_plane_calc_clk(mode, cur_pipecfg) > max_mdp_clk_rate) {
> > + struct dpu_sw_pipe_cfg *r_pipe_cfg = &pstate->pipe_cfg[cfg_idx + 1];
> > +
> > + if (drm_rect_width(&cur_pipecfg->src_rect) > 2 * max_linewidth) {
> > + DPU_DEBUG_PLANE(pdpu, "invalid src " DRM_RECT_FMT " line:%u\n",
> > + DRM_RECT_ARG(&cur_pipecfg->src_rect), max_linewidth);
> > + return -E2BIG;
> > + }
> > +
> > + memcpy(r_pipe_cfg, cur_pipecfg, sizeof(struct dpu_sw_pipe_cfg));
> > + cur_pipecfg->src_rect.x2 = (cur_pipecfg->src_rect.x1 + cur_pipecfg->src_rect.x2) >> 1;
> > + cur_pipecfg->dst_rect.x2 = (cur_pipecfg->dst_rect.x1 + cur_pipecfg->dst_rect.x2) >> 1;
>
> pipe_cfg. If you need, rename the topmost var name.
OK. pipe_cfg_cur ?
>
> > + r_pipe_cfg->src_rect.x1 = cur_pipecfg->src_rect.x2;
> > + r_pipe_cfg->dst_rect.x1 = cur_pipecfg->dst_rect.x2;
> > + r_pipe_cfg->valid = true;
> > + DPU_DEBUG_PLANE(pdpu, "Split super wide plane into:"
> > + DRM_RECT_FMT " and " DRM_RECT_FMT "\n",
> > + DRM_RECT_ARG(&cur_pipecfg->src_rect),
> > + DRM_RECT_ARG(&r_pipe_cfg->src_rect));
> > + } else {
> > + memset(&pstate->pipe_cfg[cfg_idx + 1], 0, sizeof(struct dpu_sw_pipe_cfg));
> > + memset(&pstate->pipe[cfg_idx + 1], 0, sizeof(struct dpu_sw_pipe));
>
> Please keep using r_pipe_cfg here.
OK, will make r_pipe_cfg a variable in function scope, not bracket scope.
>
> > + }
> >
> > - drm_rect_rotate_inv(&pipe_cfg->src_rect,
> > - new_plane_state->fb->width, new_plane_state->fb->height,
> > - new_plane_state->rotation);
> > - if (drm_rect_width(&r_pipe_cfg->src_rect) != 0)
> > - drm_rect_rotate_inv(&r_pipe_cfg->src_rect,
> > + drm_rect_rotate_inv(&cur_pipecfg->src_rect,
> > new_plane_state->fb->width, new_plane_state->fb->height,
> > new_plane_state->rotation);
> > + }
> >
> > pstate->needs_qos_remap = drm_atomic_crtc_needs_modeset(crtc_state);
> >
> >
> > --
> > 2.34.1
> >
>
> --
> With best wishes
> Dmitry
Powered by blists - more mailing lists