On 12-02-2025 13:02, Borah, Chaitanya
Kumar wrote:
Typo from my side, corrected!-----Original Message----- From: Murthy, Arun R <arun.r.murthy@xxxxxxxxx> Sent: Wednesday, February 5, 2025 3:57 PM To: dri-devel@xxxxxxxxxxxxxxxxxxxxx; intel-gfx@xxxxxxxxxxxxxxxxxxxxx; intel- xe@xxxxxxxxxxxxxxxxxxxxx Cc: Borah, Chaitanya Kumar <chaitanya.kumar.borah@xxxxxxxxx>; Syrjala, Ville <ville.syrjala@xxxxxxxxx>; Murthy, Arun R <arun.r.murthy@xxxxxxxxx> Subject: [PATCH v4 3/3] drm/i915/display: Add i915 hook for format_mod_supported_async Hook up the newly added plane function pointer format_mod_supported_async to populate the modifiers/formats supported by asynchronous flips. Signed-off-by: Arun R Murthy <arun.r.murthy@xxxxxxxxx> --- drivers/gpu/drm/i915/display/skl_universal_plane.c | 56 ++++++++++++++++- ----- 1 file changed, 41 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/i915/display/skl_universal_plane.c b/drivers/gpu/drm/i915/display/skl_universal_plane.c index ba5db553c374259f8f3246c1408b55d32c8794e5..96d53b22715cf397483acbdb 23b4dd60d6cbb604 100644 --- a/drivers/gpu/drm/i915/display/skl_universal_plane.c +++ b/drivers/gpu/drm/i915/display/skl_universal_plane.c @@ -511,6 +511,33 @@ skl_plane_max_stride(struct intel_plane *plane, modifier, rotation, max_pixels, max_bytes); } +static bool intel_plane_async_formats(struct intel_plane *plane, +uint32_t format) { + switch (format) { + case DRM_FORMAT_RGB565: + case DRM_FORMAT_XRGB8888: + case DRM_FORMAT_XBGR8888: + case DRM_FORMAT_ARGB8888: + case DRM_FORMAT_ABGR8888: + case DRM_FORMAT_XRGB2101010: + case DRM_FORMAT_XBGR2101010: + case DRM_FORMAT_XRGB16161616F: + case DRM_FORMAT_XBGR16161616F: + return true; + default: + return false; + } +} + +static bool intel_plane_format_mod_supported_async(struct drm_plane *plane, + uint32_t format, + uint64_t modifier) +{ + if (intel_plane_can_async_flip(to_intel_plane(plane), modifier)) + return false;This should be if (!intel_plane_can_async_flip(to_intel_plane(plane), modifier))
I am fine! I can switch back to my previous patchset from which achieving this switch case latter is much easier.I think we should eventually move to a switch case ladder that actually checks for compatibility among format-modifier pair rather than checking them individually. Though we don't have a good use of it today, we might end up having scenarios where some formats are supported only for certain modifiers and vice versa. But it can be taken up later.
I see some comments made on rev 2 [1] regarding re-using the format_mod_sync() hook for this. I have not yet formed a strong opinion about it and with the above suggested change, it at least seems to work.
The comment from rev 2 [1] is
> > intel_plane_format_mod_supported_async() > > { > > // some generic checks here (eg. reject planar formats) > > > > return plane->format_mod_supported() && > > plane->can_async_flip(); > > }
Here some generic checks to reject plannar formats is being done from intel_plane_async_formats() After this we have the result as to if the format is supported for not and with plane->can_async_flip we have a result of modifier being supported or not. That itself should be sufficient. If you still prefer to make use of the call plane->format_mod_supported() with some hacks can be done. Let me know!
[1] https://patchwork.freedesktop.org/patch/631264/?series=140935&rev=2+ + return intel_plane_async_formats(to_intel_plane(plane), format); } static bool tgl_plane_can_async_flip(u64 modifier) { @@ -2559,30 +2586,29 @@ static bool tgl_plane_format_mod_supported(struct drm_plane *_plane, } } +#define INTEL_PLANE_FUNCS \ + .update_plane = drm_atomic_helper_update_plane, \ + .disable_plane = drm_atomic_helper_disable_plane, \ + .destroy = intel_plane_destroy, \ + .atomic_duplicate_state = intel_plane_duplicate_state, \ + .atomic_destroy_state = intel_plane_destroy_state, \ + .format_mod_supported_async = intel_plane_format_mod_supported_async + static const struct drm_plane_funcs skl_plane_funcs = { - .update_plane = drm_atomic_helper_update_plane, - .disable_plane = drm_atomic_helper_disable_plane, - .destroy = intel_plane_destroy, - .atomic_duplicate_state = intel_plane_duplicate_state, - .atomic_destroy_state = intel_plane_destroy_state, + INTEL_PLANE_FUNCS, + .format_mod_supported = skl_plane_format_mod_supported, }; static const struct drm_plane_funcs icl_plane_funcs = { - .update_plane = drm_atomic_helper_update_plane, - .disable_plane = drm_atomic_helper_disable_plane, - .destroy = intel_plane_destroy, - .atomic_duplicate_state = intel_plane_duplicate_state, - .atomic_destroy_state = intel_plane_destroy_state, + INTEL_PLANE_FUNCS, + .format_mod_supported = icl_plane_format_mod_supported, }; static const struct drm_plane_funcs tgl_plane_funcs = { - .update_plane = drm_atomic_helper_update_plane, - .disable_plane = drm_atomic_helper_disable_plane, - .destroy = intel_plane_destroy, - .atomic_duplicate_state = intel_plane_duplicate_state, - .atomic_destroy_state = intel_plane_destroy_state, + INTEL_PLANE_FUNCS, + .format_mod_supported = tgl_plane_format_mod_supported, };This leads to exposing the property even on planes that do not support async flip. Functionally it should not be a problem as can_async is exposed only for primary plane.
Yes property will be created for all the planes but data will be populated only for primary plane.
Thanks and Regards,
Arun R Murthy
-------------------
May be we can live with it for now. Ville any thoughts? Regards Chaitanya-- 2.25.1
