Re: [PATCH 1/6] drm/i915/cursor: Check joiner cursor commit status

["Borah, Chaitanya Kumar" <chaitanya.kumar.borah@intel.com>]

Hello Nemesa,

On 6/8/2026 11:56 AM, Nemesa Garg wrote:
> In joiner mode, secondary cursor commits may still be running
> even when the primary cursor commit is done. Walking the secondary
> pipes also requires holding the secondary planes modeset locks.
> Add intel_cursor_lock_joined_planes() to acquire modeset locks
> for all secondary cursor planes. Check all joined cursor commit
> status before taking the fast path. If any commit is still pending,
> fallback to slow path.
> 
> v2: Use intel_crtc_joined_pipe_mask(). [Ville]
> v3: Lock secondary cursor CRTCs and planes. [sashiko]
> 
> Assisted-by: Claude:claude-sonnet-4.6
> Signed-off-by: Nemesa Garg <nemesa.garg@intel.com>
> ---
>   drivers/gpu/drm/i915/display/intel_cursor.c | 63 +++++++++++++++++++++
>   1 file changed, 63 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/display/intel_cursor.c b/drivers/gpu/drm/i915/display/intel_cursor.c
> index 88384dea868b..f8b24865c93a 100644
> --- a/drivers/gpu/drm/i915/display/intel_cursor.c
> +++ b/drivers/gpu/drm/i915/display/intel_cursor.c
> @@ -13,6 +13,7 @@
>   #include <drm/drm_vblank.h>
>   
>   #include "intel_atomic.h"
> +#include "intel_crtc.h"
>   #include "intel_cursor.h"
>   #include "intel_cursor_regs.h"
>   #include "intel_de.h"
> @@ -796,6 +797,58 @@ void intel_cursor_unpin_work(struct kthread_work *base)
>   	intel_plane_destroy_state(&plane->base, &plane_state->uapi);
>   }
>   
> +static int intel_cursor_lock_joined_planes(struct intel_display *display,
> +					   const struct intel_crtc_state *crtc_state,
> +					   struct intel_plane *primary_plane,
> +					   struct drm_modeset_acquire_ctx *ctx)
> +{
> +	struct intel_crtc *pipe_crtc;
> +	int ret;
> +
> +	for_each_intel_crtc_in_pipe_mask(display, pipe_crtc,
> +					 intel_crtc_joined_pipe_mask(crtc_state)) {
> +		struct intel_plane *pipe_plane =
> +			intel_crtc_get_plane(pipe_crtc, PLANE_CURSOR);
> +
> +		if (pipe_plane == primary_plane)
> +			continue;

...

> +
> +		ret = drm_modeset_lock(&pipe_crtc->base.mutex, ctx);
> +		if (ret)
> +			return ret;
> +
> +		ret = drm_modeset_lock(&pipe_plane->base.mutex, ctx);
> +		if (ret)
> +			return ret;
> +	}
> +	return 0;
> +}
> +
> +static bool
> +intel_cursor_joiner_commits_idle(struct intel_display *display,
> +				 const struct intel_crtc_state *crtc_state)
> +{
> +	struct intel_crtc *pipe_crtc;
> +
> +	for_each_intel_crtc_in_pipe_mask(display, pipe_crtc,
> +					 intel_crtc_joined_pipe_mask(crtc_state)) {
> +		struct intel_plane *pipe_plane;
> +		struct intel_plane_state *pipe_plane_state;
> +
> +		if (pipe_crtc == to_intel_crtc(crtc_state->uapi.crtc))
> +			continue;
> +

Both the helpers have two different ways to skip the primary. I would 
suggest, pick one and use it in both.

> +		pipe_plane = intel_crtc_get_plane(pipe_crtc, PLANE_CURSOR);
> +		pipe_plane_state = to_intel_plane_state(pipe_plane->base.state);
> +
> +		if (pipe_plane_state->uapi.commit &&
> +		    !try_wait_for_completion(&pipe_plane_state->uapi.commit->hw_done))
> +			return false;
> +	}
> +
> +	return true;
> +}
> +
>   static int
>   intel_legacy_cursor_update(struct drm_plane *_plane,
>   			   struct drm_crtc *_crtc,
> @@ -842,6 +895,16 @@ intel_legacy_cursor_update(struct drm_plane *_plane,
>   	    !try_wait_for_completion(&old_plane_state->uapi.commit->hw_done))
>   		goto slow;
>   
> +	ret = intel_cursor_lock_joined_planes(display, crtc_state, plane, ctx);
> +	if (ret == -EDEADLK)
> +		return ret;
> +	if (ret)
> +		goto slow;
> +
> +	/* Check all joined pipes for pending commits */
> +	if (!intel_cursor_joiner_commits_idle(display, crtc_state))
> +		goto slow;
> +

The param check right below can be a quick early bail out to the slow 
path. Better to move it above the lock so we don't grab every secondary 
crtc/plane lock just to drop straight to the slow path.

Rather than special-casing the primary pipe in both helpers (and in two
different ways), can we just iterate the full joined mask uniformly?
Obviously, taking care of any special treatment that the primary pipe needs.

==
Chaitanya

>   	/*
>   	 * If any parameters change that may affect watermarks,
>   	 * take the slowpath. Only changing fb or position should be