Le 15/08/2025 à 05:50, Alex Hung a écrit :
The functions are to clean up color pipeline when a device driver
fails to create its color pipeline.
Signed-off-by: Alex Hung <alex.h...@amd.com>
Reviewed-by: Daniel Stone <dani...@collabora.com>
Reviewed-by: Simon Ser <cont...@emersion.fr>
Reviewed-by: Melissa Wen <m...@igalia.com>
---
v11:
- destroy function takes drm_device *dev instead of drm_plane *plane
(Nícolas Prado)
v9:
- Move from from latest commit to here, and drm_colorop_pipeline_destroy
is called in respective commits.
drivers/gpu/drm/drm_colorop.c | 35 +++++++++++++++++++++++++++++++++++
include/drm/drm_colorop.h | 2 ++
2 files changed, 37 insertions(+)
diff --git a/drivers/gpu/drm/drm_colorop.c b/drivers/gpu/drm/drm_colorop.c
index 7b3ecf7ddd11..6930d39c8ad3 100644
--- a/drivers/gpu/drm/drm_colorop.c
+++ b/drivers/gpu/drm/drm_colorop.c
@@ -135,6 +135,41 @@ static int drm_plane_colorop_init(struct drm_device *dev,
struct drm_colorop *co
return ret;
}
+/**
+ * drm_colorop_cleanup - Cleanup a drm_colorop object in color_pipeline
+ *
+ * @colorop: The drm_colorop object to be cleaned
+ */
+static void drm_colorop_cleanup(struct drm_colorop *colorop)
+{
+ struct drm_device *dev = colorop->dev;
+ struct drm_mode_config *config = &dev->mode_config;
+
+ list_del(&colorop->head);
+ config->num_colorop--;
+
+ kfree(colorop->state);
+}
+
+/**
+ * drm_colorop_pipeline_destroy - Helper for color pipeline destruction
+ *
+ * @dev: - The drm_device containing the drm_planes with the color_pipelines
+ *
+ * Provides a default color pipeline destroy handler for drm_device.
+ */
+void drm_colorop_pipeline_destroy(struct drm_device *dev)
+{
+ struct drm_mode_config *config = &dev->mode_config;
+ struct drm_colorop *colorop, *next;
+
+ list_for_each_entry_safe(colorop, next, &config->colorop_list, head) {
+ drm_colorop_cleanup(colorop);
+ kfree(colorop);
This free here seems a bit strange. I don't see any requirement on the
colorop pointer in the init function, so we can expect it to be embedded
in a bigger structure, so this kfree may affect a non-allocated pointer.
I would expect one of:
- a clear documentation in drm_plane_colorop_*_init documentation that
explicitly says that you need to pass a kzalloc pointer => very error
prone, if the user don't read carefully the documentation it may lead to
undefined behavior
- that drm_plane_colorop_*_init will do the kzalloc itself (so we
garantee that the pointer is always a kzalloc pointer) => it will forbid
to embed colorop structure in bigger structure. I don't think this is
the case today, but I don't know if this can become a limitation for
other drivers.
- that drm_colorop_pipeline_destroy does not free anything, this must be
done by the driver itself => "more flexible" solution, but also require
that you can attach a destroy function to each drm_colorop structure
(like drm_crtc_funcs->destroy) or postpone it to complete device removal
(using drmm_kzalloc or manually in device destroy)
In addition this function leave config->colorop_list in an uncertain
state, I would be nice to add call list_del(colorop->head) just to avoid
dangling pointers.
+ }
+}
+EXPORT_SYMBOL(drm_colorop_pipeline_destroy);
+
/**
* drm_plane_colorop_curve_1d_init - Initialize a DRM_COLOROP_1D_CURVE
*
diff --git a/include/drm/drm_colorop.h b/include/drm/drm_colorop.h
index 791ab940c158..0ea1a01ca04e 100644
--- a/include/drm/drm_colorop.h
+++ b/include/drm/drm_colorop.h
@@ -239,6 +239,8 @@ static inline struct drm_colorop *drm_colorop_find(struct
drm_device *dev,
return mo ? obj_to_colorop(mo) : NULL;
}
+void drm_colorop_pipeline_destroy(struct drm_device *dev);
+
int drm_plane_colorop_curve_1d_init(struct drm_device *dev, struct
drm_colorop *colorop,
struct drm_plane *plane, u64 supported_tfs);
--
Louis Chauvet, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
