Re: [PATCH] drm/mediatek: Add spinlock for setting vblank event in atomic_begin

[AngeloGioacchino Del Regno]

Il 22/08/23 15:26, Jason-JH.Lin ha scritto:
> Add spinlock protection to avoid race condition on vblank event
> between mtk_drm_crtc_atomic_begin() and mtk_drm_finish_page_flip().

Hello Jason,

Can you please provide more information about this race condition?
(check below)

> Fixes: 119f5173628a ("drm/mediatek: Add DRM Driver for Mediatek SoC MT8173.")
> Signed-off-by: Jason-JH.Lin <jason-jh.lin@xxxxxxxxxxxx>
> ---
>   drivers/gpu/drm/mediatek/mtk_drm_crtc.c | 5 +++++
>   1 file changed, 5 insertions(+)
>
> diff --git a/drivers/gpu/drm/mediatek/mtk_drm_crtc.c b/drivers/gpu/drm/mediatek/mtk_drm_crtc.c
> index d40142842f85..128a672fe3c9 100644
> --- a/drivers/gpu/drm/mediatek/mtk_drm_crtc.c
> +++ b/drivers/gpu/drm/mediatek/mtk_drm_crtc.c
> @@ -746,6 +746,9 @@ static void mtk_drm_crtc_atomic_begin(struct drm_crtc *crtc,
>   									  crtc);
>   	struct mtk_crtc_state *mtk_crtc_state = to_mtk_crtc_state(crtc_state);
>   	struct mtk_drm_crtc *mtk_crtc = to_mtk_crtc(crtc);
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&crtc->dev->event_lock, flags);
>   
>   	if (mtk_crtc->event && mtk_crtc_state->base.event)
>   		DRM_ERROR("new event while there is still a pending event\n");
> @@ -756,6 +759,8 @@ static void mtk_drm_crtc_atomic_begin(struct drm_crtc *crtc,

...because my suspect is that what creates the race condition in this function is
the unlocked *assignment* to mtk_crtc->event, not the rest.

If I'm right, you don't need to unconditionally spinlock at the beginning of this
function hence ever-so-slightly improving performance compared to this version.

Can you please try this one and check if this *also* solves the issue?

	if (mtk_crtc_state->base.event) {
		mtk_crtc_state->base.event->pipe = drm_crtc_index(crtc);
		WARN_ON(drm_crtc_vblank_get(crtc) != 0);

		spin_lock_irqsave(&crtc->dev->event_lock, flags);
		mtk_crtc->event = mtk_crtc_state->base.event;
		spin_lock_irqrestore(&crtc->dev->event_lock, flags);

		mtk_crtc_state->base.event = NULL;
	}

P.S.: I'd try that myself, but I can't seem to reproduce the issue.

Regards,
Angelo