Re: [PATCH RFC v6 08/10] drm/msm/dpu: Allow NULL FBs in atomic commit

[Jessica Zhang <quic_jesszhan@quicinc.com>]

On 9/24/2023 3:29 AM, Dmitry Baryshkov wrote:
> On 29/08/2023 03:05, Jessica Zhang wrote:
>> Since solid fill planes allow for a NULL framebuffer in a valid commit,
>> add NULL framebuffer checks to atomic commit calls within DPU.
>>
>> Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
>> Signed-off-by: Jessica Zhang <quic_jesszhan@quicinc.com>
>> ---
>>   drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c  |  9 ++++++-
>>   drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c | 41 
>> ++++++++++++++++++++-----------
>>   2 files changed, 34 insertions(+), 16 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c 
>> b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
>> index 8ce7586e2ddf..5e845510e8c1 100644
>> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
>> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
>> @@ -451,6 +451,7 @@ static void _dpu_crtc_blend_setup_mixer(struct 
>> drm_crtc *crtc,
>>       struct drm_plane_state *state;
>>       struct dpu_crtc_state *cstate = to_dpu_crtc_state(crtc->state);
>>       struct dpu_plane_state *pstate = NULL;
>> +    const struct msm_format *fmt;
>>       struct dpu_format *format;
>>       struct dpu_hw_ctl *ctl = mixer->lm_ctl;
>> @@ -470,7 +471,13 @@ static void _dpu_crtc_blend_setup_mixer(struct 
>> drm_crtc *crtc,
>>           pstate = to_dpu_plane_state(state);
>>           fb = state->fb;
>> -        format = to_dpu_format(msm_framebuffer_format(pstate->base.fb));
>> +        if (drm_plane_solid_fill_enabled(state))
>> +            fmt = dpu_get_msm_format(&_dpu_crtc_get_kms(crtc)->base,
>> +                    DRM_FORMAT_ABGR8888, 0);
>> +        else
>> +            fmt = msm_framebuffer_format(pstate->base.fb);
>> +
>> +        format = to_dpu_format(fmt);
>>           if (pstate->stage == DPU_STAGE_BASE && format->alpha_enable)
>>               bg_alpha_enable = true;
>> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c 
>> b/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c
>> index c2aaaded07ed..114c803ff99b 100644
>> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c
>> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c
>> @@ -837,8 +837,13 @@ static int dpu_plane_atomic_check(struct 
>> drm_plane *plane,
>>       pipe_cfg->dst_rect = new_plane_state->dst;
>> -    fb_rect.x2 = new_plane_state->fb->width;
>> -    fb_rect.y2 = new_plane_state->fb->height;
>> +    if (drm_plane_solid_fill_enabled(new_plane_state))
>> +        return 0;
> 
> This would skip all the width checks, dpu_plane_atomic_check_pipe(), 
> etc. Could you please confirm that all of those checks are irrelevant 
> for solid fill?

Hi Dmitry,

Good point -- I think it would be good to keep the rect and pipe checks 
for solid fill.

> 
>> +
>> +    if (new_plane_state->pixel_source == DRM_PLANE_PIXEL_SOURCE_FB && 
>> new_plane_state->fb) {
>> +        fb_rect.x2 = new_plane_state->fb->width;
>> +        fb_rect.y2 = new_plane_state->fb->height;
>> +    }
>>       /* Ensure fb size is supported */
>>       if (drm_rect_width(&fb_rect) > MAX_IMG_WIDTH ||
>> @@ -1082,21 +1087,32 @@ static void 
>> dpu_plane_sspp_atomic_update(struct drm_plane *plane)
>>       struct drm_crtc *crtc = state->crtc;
>>       struct drm_framebuffer *fb = state->fb;
>>       bool is_rt_pipe;
>> -    const struct dpu_format *fmt =
>> -        to_dpu_format(msm_framebuffer_format(fb));
>> +    const struct dpu_format *fmt;
>>       struct dpu_sw_pipe_cfg *pipe_cfg = &pstate->pipe_cfg;
>>       struct dpu_sw_pipe_cfg *r_pipe_cfg = &pstate->r_pipe_cfg;
>>       struct dpu_kms *kms = _dpu_plane_get_kms(&pdpu->base);
>>       struct msm_gem_address_space *aspace = kms->base.aspace;
>>       struct dpu_hw_fmt_layout layout;
>>       bool layout_valid = false;
>> -    int ret;
>> -    ret = dpu_format_populate_layout(aspace, fb, &layout);
>> -    if (ret)
>> -        DPU_ERROR_PLANE(pdpu, "failed to get format layout, %d\n", ret);
>> -    else
>> -        layout_valid = true;
>> +    if (state->pixel_source == DRM_PLANE_PIXEL_SOURCE_FB && fb) {
>> +        int ret;
>> +
>> +        fmt = to_dpu_format(msm_framebuffer_format(fb));
>> +
>> +        ret = dpu_format_populate_layout(aspace, fb, &layout);
>> +        if (ret)
>> +            DPU_ERROR_PLANE(pdpu, "failed to get format layout, 
>> %d\n", ret);
>> +        else
>> +            layout_valid = true;
>> +
>> +        DPU_DEBUG_PLANE(pdpu, "FB[%u] " DRM_RECT_FP_FMT "->crtc%u " 
>> DRM_RECT_FMT
>> +                ", %4.4s ubwc %d\n", fb->base.id, 
>> DRM_RECT_FP_ARG(&state->src),
>> +                crtc->base.id, DRM_RECT_ARG(&state->dst),
>> +                (char *)&fmt->base.pixel_format, 
>> DPU_FORMAT_IS_UBWC(fmt));
>> +    } else {
>> +        fmt = dpu_get_dpu_format(DRM_FORMAT_ABGR8888);
> 
> #define DPU_SOLID_FILL_FORMAT ?

Acked.

> 
> Also, I don't think that solid_fill planes consume bandwidth, so this 
> likely needs to be fixed too.

You're right -- I think we can actually return early here if solid fill 
is enabled.

Thanks,

Jessica Zhang

> 
> 
>> +    }
>>       pstate->pending = true;
>> @@ -1104,11 +1120,6 @@ static void dpu_plane_sspp_atomic_update(struct 
>> drm_plane *plane)
>>       pstate->needs_qos_remap |= (is_rt_pipe != pdpu->is_rt_pipe);
>>       pdpu->is_rt_pipe = is_rt_pipe;
>> -    DPU_DEBUG_PLANE(pdpu, "FB[%u] " DRM_RECT_FP_FMT "->crtc%u " 
>> DRM_RECT_FMT
>> -            ", %4.4s ubwc %d\n", fb->base.id, 
>> DRM_RECT_FP_ARG(&state->src),
>> -            crtc->base.id, DRM_RECT_ARG(&state->dst),
>> -            (char *)&fmt->base.pixel_format, DPU_FORMAT_IS_UBWC(fmt));
>> -
>>       dpu_plane_sspp_update_pipe(plane, pipe, pipe_cfg, fmt,
>>                      drm_mode_vrefresh(&crtc->mode),
>>                      layout_valid ? &layout : NULL);
>>
> 
> -- 
> With best wishes
> Dmitry
>