In order to prevent any errors on connector being disabled, move the
state->crtc check upfront. This should fix the issues during suspend
when the writeback connector gets forcebly disabled.

Fixes: 71174f362d67 ("drm/msm/dpu: move writeback's atomic_check to dpu_writeback.c")
Cc: stable@vger.kernel.org
Reported-by: Leonard Lausen <leonard@lausen.nl>
Closes: https://gitlab.freedesktop.org/drm/msm/-/issues/57
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
 drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c
index 16f144cbc0c9..5c172bcf3419 100644
--- a/drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c
+++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c
@@ -39,6 +39,13 @@ static int dpu_wb_conn_atomic_check(struct drm_connector *connector,
 
 	DPU_DEBUG("[atomic_check:%d]\n", connector->base.id);
 
+	crtc = conn_state->crtc;
+	if (!crtc)
+		return 0;
+
+	if (!conn_state->writeback_job || !conn_state->writeback_job->fb)
+		return 0;
+
 	if (!conn_state || !conn_state->connector) {
 		DPU_ERROR("invalid connector state\n");
 		return -EINVAL;
@@ -47,13 +54,6 @@ static int dpu_wb_conn_atomic_check(struct drm_connector *connector,
 		return -EINVAL;
 	}
 
-	crtc = conn_state->crtc;
-	if (!crtc)
-		return 0;
-
-	if (!conn_state->writeback_job || !conn_state->writeback_job->fb)
-		return 0;
-
 	crtc_state = drm_atomic_get_crtc_state(state, crtc);
 	if (IS_ERR(crtc_state))
 		return PTR_ERR(crtc_state);

-- 
2.39.2

On 7/9/2024 6:48 AM, Dmitry Baryshkov wrote:
> In order to prevent any errors on connector being disabled, move the
> state->crtc check upfront. This should fix the issues during suspend
> when the writeback connector gets forcebly disabled.
> 
> Fixes: 71174f362d67 ("drm/msm/dpu: move writeback's atomic_check to dpu_writeback.c")
> Cc: stable@vger.kernel.org
> Reported-by: Leonard Lausen <leonard@lausen.nl>
> Closes: https://gitlab.freedesktop.org/drm/msm/-/issues/57
> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
> ---
>   drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c | 14 +++++++-------
>   1 file changed, 7 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c
> index 16f144cbc0c9..5c172bcf3419 100644
> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c
> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_writeback.c
> @@ -39,6 +39,13 @@ static int dpu_wb_conn_atomic_check(struct drm_connector *connector,
>   
>   	DPU_DEBUG("[atomic_check:%d]\n", connector->base.id);
>   
> +	crtc = conn_state->crtc;

We are checking for !conn_state a few lines below but we are 
dereferencing conn_state here.

This is bound to hit a smatch error and also does not look right.

If conn_state will always be valid, we should drop that check too rather 
than checking it later.

Coming to the issue itself, I tried checking the logs but it was not clear.

During force disable, were we hitting below check and hence the 
connector was not getting disabled?

else if (conn_state->connector->status != connector_status_connected) {
                 DPU_ERROR("connector not connected %d\n", 
conn_state->connector->status);
                 return -EINVAL;
         }


I did not see this error log there, so can you pls explain where we were 
bailing out? The check seems valid to me.

> +	if (!crtc)
> +		return 0;
> +
> +	if (!conn_state->writeback_job || !conn_state->writeback_job->fb)
> +		return 0;
> +
>   	if (!conn_state || !conn_state->connector) {
>   		DPU_ERROR("invalid connector state\n");
>   		return -EINVAL;
> @@ -47,13 +54,6 @@ static int dpu_wb_conn_atomic_check(struct drm_connector *connector,
>   		return -EINVAL;
>   	}
>   
> -	crtc = conn_state->crtc;
> -	if (!crtc)
> -		return 0;
> -
> -	if (!conn_state->writeback_job || !conn_state->writeback_job->fb)
> -		return 0;
> -
>   	crtc_state = drm_atomic_get_crtc_state(state, crtc);
>   	if (IS_ERR(crtc_state))
>   		return PTR_ERR(crtc_state);
>