]> git.apps.os.sepia.ceph.com Git - ceph-client.git/commitdiff
drm/i915/plane: rename intel_atomic_plane.[ch] to intel_plane.[ch]
authorJani Nikula <jani.nikula@intel.com>
Tue, 17 Jun 2025 08:13:36 +0000 (11:13 +0300)
committerJani Nikula <jani.nikula@intel.com>
Thu, 19 Jun 2025 09:14:29 +0000 (12:14 +0300)
It's all atomic, no need to emphasize this.

v2: Also update Documentation/gpu/i915.rst (Gustavo)

Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Link: https://lore.kernel.org/r/ba5f304e9fe71723191d872e6828d461e1a572bd.1750147992.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
21 files changed:
Documentation/gpu/i915.rst
drivers/gpu/drm/i915/Makefile
drivers/gpu/drm/i915/display/i9xx_plane.c
drivers/gpu/drm/i915/display/intel_atomic.c
drivers/gpu/drm/i915/display/intel_atomic_plane.c [deleted file]
drivers/gpu/drm/i915/display/intel_atomic_plane.h [deleted file]
drivers/gpu/drm/i915/display/intel_cdclk.c
drivers/gpu/drm/i915/display/intel_crtc.c
drivers/gpu/drm/i915/display/intel_cursor.c
drivers/gpu/drm/i915/display/intel_display.c
drivers/gpu/drm/i915/display/intel_display_irq.c
drivers/gpu/drm/i915/display/intel_fb.c
drivers/gpu/drm/i915/display/intel_fb_pin.c
drivers/gpu/drm/i915/display/intel_plane.c [new file with mode: 0644]
drivers/gpu/drm/i915/display/intel_plane.h [new file with mode: 0644]
drivers/gpu/drm/i915/display/intel_plane_initial.c
drivers/gpu/drm/i915/display/intel_sprite.c
drivers/gpu/drm/i915/display/skl_universal_plane.c
drivers/gpu/drm/i915/display/skl_watermark.c
drivers/gpu/drm/xe/Makefile
drivers/gpu/drm/xe/display/xe_plane_initial.c

index 7a469df675d82b8f2dae7349bfd56ad5efb491bb..5c5436413d7a00304733e9427ac70941d73405ac 100644 (file)
@@ -112,10 +112,10 @@ panel self refresh.
 Atomic Plane Helpers
 --------------------
 
-.. kernel-doc:: drivers/gpu/drm/i915/display/intel_atomic_plane.c
+.. kernel-doc:: drivers/gpu/drm/i915/display/intel_plane.c
    :doc: atomic plane helpers
 
-.. kernel-doc:: drivers/gpu/drm/i915/display/intel_atomic_plane.c
+.. kernel-doc:: drivers/gpu/drm/i915/display/intel_plane.c
    :internal:
 
 Asynchronous Page Flip
index 7c6075bc483c0f27ee8df06a259edf177757e2e9..0c3fbd0d14e71333cba6af651c3b543abfdb6062 100644 (file)
@@ -218,12 +218,11 @@ i915-$(CONFIG_HWMON) += \
 # modesetting core code
 i915-y += \
        display/hsw_ips.o \
-       display/i9xx_plane.o \
        display/i9xx_display_sr.o \
+       display/i9xx_plane.o \
        display/i9xx_wm.o \
        display/intel_alpm.o \
        display/intel_atomic.o \
-       display/intel_atomic_plane.o \
        display/intel_audio.o \
        display/intel_bios.o \
        display/intel_bo.o \
@@ -283,6 +282,7 @@ i915-y += \
        display/intel_pch.o \
        display/intel_pch_display.o \
        display/intel_pch_refclk.o \
+       display/intel_plane.o \
        display/intel_plane_initial.o \
        display/intel_pmdemand.o \
        display/intel_psr.o \
index 8f15333a4b07ce21e960bfeee4c75aa3cd828f91..eba8eb745aa7929355dace168f09c92c538e3013 100644 (file)
@@ -15,7 +15,6 @@
 #include "i9xx_plane.h"
 #include "i9xx_plane_regs.h"
 #include "intel_atomic.h"
-#include "intel_atomic_plane.h"
 #include "intel_de.h"
 #include "intel_display_irq.h"
 #include "intel_display_regs.h"
@@ -23,6 +22,7 @@
 #include "intel_fb.h"
 #include "intel_fbc.h"
 #include "intel_frontbuffer.h"
+#include "intel_plane.h"
 #include "intel_sprite.h"
 
 /* Primary plane formats for gen <= 3 */
index f85edb374c9750bac881b0a6a10db0535df3b36d..348b1655435e1cc597d2adca3aa10555e670335d 100644 (file)
@@ -26,7 +26,7 @@
  *
  * The functions here implement the state management and hardware programming
  * dispatch required by the atomic modeset infrastructure.
- * See intel_atomic_plane.c for the plane-specific atomic functionality.
+ * See intel_plane.c for the plane-specific atomic functionality.
  */
 
 #include <drm/display/drm_dp_tunnel.h>
diff --git a/drivers/gpu/drm/i915/display/intel_atomic_plane.c b/drivers/gpu/drm/i915/display/intel_atomic_plane.c
deleted file mode 100644 (file)
index e0336c7..0000000
+++ /dev/null
@@ -1,1588 +0,0 @@
-/*
- * Copyright © 2014 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
- * DEALINGS IN THE SOFTWARE.
- */
-
-/**
- * DOC: atomic plane helpers
- *
- * The functions here are used by the atomic plane helper functions to
- * implement legacy plane updates (i.e., drm_plane->update_plane() and
- * drm_plane->disable_plane()).  This allows plane updates to use the
- * atomic state infrastructure and perform plane updates as separate
- * prepare/check/commit/cleanup steps.
- */
-
-#include <linux/dma-fence-chain.h>
-#include <linux/dma-resv.h>
-
-#include <drm/drm_atomic_helper.h>
-#include <drm/drm_blend.h>
-#include <drm/drm_damage_helper.h>
-#include <drm/drm_fourcc.h>
-#include <drm/drm_gem.h>
-#include <drm/drm_gem_atomic_helper.h>
-
-#include "gem/i915_gem_object.h"
-#include "i915_scheduler_types.h"
-#include "i915_vma.h"
-#include "i9xx_plane_regs.h"
-#include "intel_atomic_plane.h"
-#include "intel_cdclk.h"
-#include "intel_cursor.h"
-#include "intel_display_rps.h"
-#include "intel_display_trace.h"
-#include "intel_display_types.h"
-#include "intel_fb.h"
-#include "intel_fb_pin.h"
-#include "skl_scaler.h"
-#include "skl_universal_plane.h"
-#include "skl_watermark.h"
-
-static void intel_plane_state_reset(struct intel_plane_state *plane_state,
-                                   struct intel_plane *plane)
-{
-       memset(plane_state, 0, sizeof(*plane_state));
-
-       __drm_atomic_helper_plane_state_reset(&plane_state->uapi, &plane->base);
-
-       plane_state->scaler_id = -1;
-}
-
-struct intel_plane *intel_plane_alloc(void)
-{
-       struct intel_plane_state *plane_state;
-       struct intel_plane *plane;
-
-       plane = kzalloc(sizeof(*plane), GFP_KERNEL);
-       if (!plane)
-               return ERR_PTR(-ENOMEM);
-
-       plane_state = kzalloc(sizeof(*plane_state), GFP_KERNEL);
-       if (!plane_state) {
-               kfree(plane);
-               return ERR_PTR(-ENOMEM);
-       }
-
-       intel_plane_state_reset(plane_state, plane);
-
-       plane->base.state = &plane_state->uapi;
-
-       return plane;
-}
-
-void intel_plane_free(struct intel_plane *plane)
-{
-       intel_plane_destroy_state(&plane->base, plane->base.state);
-       kfree(plane);
-}
-
-/**
- * intel_plane_destroy - destroy a plane
- * @plane: plane to destroy
- *
- * Common destruction function for all types of planes (primary, cursor,
- * sprite).
- */
-void intel_plane_destroy(struct drm_plane *plane)
-{
-       drm_plane_cleanup(plane);
-       kfree(to_intel_plane(plane));
-}
-
-/**
- * intel_plane_duplicate_state - duplicate plane state
- * @plane: drm plane
- *
- * Allocates and returns a copy of the plane state (both common and
- * Intel-specific) for the specified plane.
- *
- * Returns: The newly allocated plane state, or NULL on failure.
- */
-struct drm_plane_state *
-intel_plane_duplicate_state(struct drm_plane *plane)
-{
-       struct intel_plane_state *intel_state;
-
-       intel_state = to_intel_plane_state(plane->state);
-       intel_state = kmemdup(intel_state, sizeof(*intel_state), GFP_KERNEL);
-
-       if (!intel_state)
-               return NULL;
-
-       __drm_atomic_helper_plane_duplicate_state(plane, &intel_state->uapi);
-
-       intel_state->ggtt_vma = NULL;
-       intel_state->dpt_vma = NULL;
-       intel_state->flags = 0;
-       intel_state->damage = DRM_RECT_INIT(0, 0, 0, 0);
-
-       /* add reference to fb */
-       if (intel_state->hw.fb)
-               drm_framebuffer_get(intel_state->hw.fb);
-
-       return &intel_state->uapi;
-}
-
-/**
- * intel_plane_destroy_state - destroy plane state
- * @plane: drm plane
- * @state: state object to destroy
- *
- * Destroys the plane state (both common and Intel-specific) for the
- * specified plane.
- */
-void
-intel_plane_destroy_state(struct drm_plane *plane,
-                         struct drm_plane_state *state)
-{
-       struct intel_plane_state *plane_state = to_intel_plane_state(state);
-
-       drm_WARN_ON(plane->dev, plane_state->ggtt_vma);
-       drm_WARN_ON(plane->dev, plane_state->dpt_vma);
-
-       __drm_atomic_helper_plane_destroy_state(&plane_state->uapi);
-       if (plane_state->hw.fb)
-               drm_framebuffer_put(plane_state->hw.fb);
-       kfree(plane_state);
-}
-
-bool intel_plane_needs_physical(struct intel_plane *plane)
-{
-       struct intel_display *display = to_intel_display(plane);
-
-       return plane->id == PLANE_CURSOR &&
-               DISPLAY_INFO(display)->cursor_needs_physical;
-}
-
-bool intel_plane_can_async_flip(struct intel_plane *plane, u32 format,
-                               u64 modifier)
-{
-       if (intel_format_info_is_yuv_semiplanar(drm_format_info(format), modifier) ||
-           format == DRM_FORMAT_C8)
-               return false;
-
-       return plane->can_async_flip && plane->can_async_flip(modifier);
-}
-
-bool intel_plane_format_mod_supported_async(struct drm_plane *plane,
-                                           u32 format,
-                                           u64 modifier)
-{
-       if (!plane->funcs->format_mod_supported(plane, format, modifier))
-               return false;
-
-       return intel_plane_can_async_flip(to_intel_plane(plane),
-                                       format, modifier);
-}
-
-unsigned int intel_adjusted_rate(const struct drm_rect *src,
-                                const struct drm_rect *dst,
-                                unsigned int rate)
-{
-       unsigned int src_w, src_h, dst_w, dst_h;
-
-       src_w = drm_rect_width(src) >> 16;
-       src_h = drm_rect_height(src) >> 16;
-       dst_w = drm_rect_width(dst);
-       dst_h = drm_rect_height(dst);
-
-       /* Downscaling limits the maximum pixel rate */
-       dst_w = min(src_w, dst_w);
-       dst_h = min(src_h, dst_h);
-
-       return DIV_ROUND_UP_ULL(mul_u32_u32(rate, src_w * src_h),
-                               dst_w * dst_h);
-}
-
-unsigned int intel_plane_pixel_rate(const struct intel_crtc_state *crtc_state,
-                                   const struct intel_plane_state *plane_state)
-{
-       /*
-        * Note we don't check for plane visibility here as
-        * we want to use this when calculating the cursor
-        * watermarks even if the cursor is fully offscreen.
-        * That depends on the src/dst rectangles being
-        * correctly populated whenever the watermark code
-        * considers the cursor to be visible, whether or not
-        * it is actually visible.
-        *
-        * See: intel_wm_plane_visible() and intel_check_cursor()
-        */
-
-       return intel_adjusted_rate(&plane_state->uapi.src,
-                                  &plane_state->uapi.dst,
-                                  crtc_state->pixel_rate);
-}
-
-unsigned int intel_plane_data_rate(const struct intel_crtc_state *crtc_state,
-                                  const struct intel_plane_state *plane_state,
-                                  int color_plane)
-{
-       const struct drm_framebuffer *fb = plane_state->hw.fb;
-
-       if (!plane_state->uapi.visible)
-               return 0;
-
-       return intel_plane_pixel_rate(crtc_state, plane_state) *
-               fb->format->cpp[color_plane];
-}
-
-static unsigned int
-intel_plane_relative_data_rate(const struct intel_crtc_state *crtc_state,
-                              const struct intel_plane_state *plane_state,
-                              int color_plane)
-{
-       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
-       const struct drm_framebuffer *fb = plane_state->hw.fb;
-       unsigned int rel_data_rate;
-       int width, height;
-
-       if (plane->id == PLANE_CURSOR)
-               return 0;
-
-       if (!plane_state->uapi.visible)
-               return 0;
-
-       /*
-        * Src coordinates are already rotated by 270 degrees for
-        * the 90/270 degree plane rotation cases (to match the
-        * GTT mapping), hence no need to account for rotation here.
-        */
-       width = drm_rect_width(&plane_state->uapi.src) >> 16;
-       height = drm_rect_height(&plane_state->uapi.src) >> 16;
-
-       /* UV plane does 1/2 pixel sub-sampling */
-       if (color_plane == 1) {
-               width /= 2;
-               height /= 2;
-       }
-
-       rel_data_rate =
-               skl_plane_relative_data_rate(crtc_state, plane, width, height,
-                                            fb->format->cpp[color_plane]);
-       if (!rel_data_rate)
-               return 0;
-
-       return intel_adjusted_rate(&plane_state->uapi.src,
-                                  &plane_state->uapi.dst,
-                                  rel_data_rate);
-}
-
-int intel_plane_calc_min_cdclk(struct intel_atomic_state *state,
-                              struct intel_plane *plane,
-                              bool *need_cdclk_calc)
-{
-       struct intel_display *display = to_intel_display(plane);
-       const struct intel_plane_state *plane_state =
-               intel_atomic_get_new_plane_state(state, plane);
-       struct intel_crtc *crtc = to_intel_crtc(plane_state->hw.crtc);
-       const struct intel_cdclk_state *cdclk_state;
-       const struct intel_crtc_state *old_crtc_state;
-       struct intel_crtc_state *new_crtc_state;
-
-       if (!plane_state->uapi.visible || !plane->min_cdclk)
-               return 0;
-
-       old_crtc_state = intel_atomic_get_old_crtc_state(state, crtc);
-       new_crtc_state = intel_atomic_get_new_crtc_state(state, crtc);
-
-       new_crtc_state->min_cdclk[plane->id] =
-               plane->min_cdclk(new_crtc_state, plane_state);
-
-       /*
-        * No need to check against the cdclk state if
-        * the min cdclk for the plane doesn't increase.
-        *
-        * Ie. we only ever increase the cdclk due to plane
-        * requirements. This can reduce back and forth
-        * display blinking due to constant cdclk changes.
-        */
-       if (new_crtc_state->min_cdclk[plane->id] <=
-           old_crtc_state->min_cdclk[plane->id])
-               return 0;
-
-       cdclk_state = intel_atomic_get_cdclk_state(state);
-       if (IS_ERR(cdclk_state))
-               return PTR_ERR(cdclk_state);
-
-       /*
-        * No need to recalculate the cdclk state if
-        * the min cdclk for the pipe doesn't increase.
-        *
-        * Ie. we only ever increase the cdclk due to plane
-        * requirements. This can reduce back and forth
-        * display blinking due to constant cdclk changes.
-        */
-       if (new_crtc_state->min_cdclk[plane->id] <=
-           cdclk_state->min_cdclk[crtc->pipe])
-               return 0;
-
-       drm_dbg_kms(display->drm,
-                   "[PLANE:%d:%s] min cdclk (%d kHz) > [CRTC:%d:%s] min cdclk (%d kHz)\n",
-                   plane->base.base.id, plane->base.name,
-                   new_crtc_state->min_cdclk[plane->id],
-                   crtc->base.base.id, crtc->base.name,
-                   cdclk_state->min_cdclk[crtc->pipe]);
-       *need_cdclk_calc = true;
-
-       return 0;
-}
-
-static void intel_plane_clear_hw_state(struct intel_plane_state *plane_state)
-{
-       if (plane_state->hw.fb)
-               drm_framebuffer_put(plane_state->hw.fb);
-
-       memset(&plane_state->hw, 0, sizeof(plane_state->hw));
-}
-
-static void
-intel_plane_copy_uapi_plane_damage(struct intel_plane_state *new_plane_state,
-                                  const struct intel_plane_state *old_uapi_plane_state,
-                                  const struct intel_plane_state *new_uapi_plane_state)
-{
-       struct intel_display *display = to_intel_display(new_plane_state);
-       struct drm_rect *damage = &new_plane_state->damage;
-
-       /* damage property tracking enabled from display version 12 onwards */
-       if (DISPLAY_VER(display) < 12)
-               return;
-
-       if (!drm_atomic_helper_damage_merged(&old_uapi_plane_state->uapi,
-                                            &new_uapi_plane_state->uapi,
-                                            damage))
-               /* Incase helper fails, mark whole plane region as damage */
-               *damage = drm_plane_state_src(&new_uapi_plane_state->uapi);
-}
-
-void intel_plane_copy_uapi_to_hw_state(struct intel_plane_state *plane_state,
-                                      const struct intel_plane_state *from_plane_state,
-                                      struct intel_crtc *crtc)
-{
-       intel_plane_clear_hw_state(plane_state);
-
-       /*
-        * For the joiner secondary uapi.crtc will point at
-        * the primary crtc. So we explicitly assign the right
-        * secondary crtc to hw.crtc. uapi.crtc!=NULL simply
-        * indicates the plane is logically enabled on the uapi level.
-        */
-       plane_state->hw.crtc = from_plane_state->uapi.crtc ? &crtc->base : NULL;
-
-       plane_state->hw.fb = from_plane_state->uapi.fb;
-       if (plane_state->hw.fb)
-               drm_framebuffer_get(plane_state->hw.fb);
-
-       plane_state->hw.alpha = from_plane_state->uapi.alpha;
-       plane_state->hw.pixel_blend_mode =
-               from_plane_state->uapi.pixel_blend_mode;
-       plane_state->hw.rotation = from_plane_state->uapi.rotation;
-       plane_state->hw.color_encoding = from_plane_state->uapi.color_encoding;
-       plane_state->hw.color_range = from_plane_state->uapi.color_range;
-       plane_state->hw.scaling_filter = from_plane_state->uapi.scaling_filter;
-
-       plane_state->uapi.src = drm_plane_state_src(&from_plane_state->uapi);
-       plane_state->uapi.dst = drm_plane_state_dest(&from_plane_state->uapi);
-}
-
-void intel_plane_copy_hw_state(struct intel_plane_state *plane_state,
-                              const struct intel_plane_state *from_plane_state)
-{
-       intel_plane_clear_hw_state(plane_state);
-
-       memcpy(&plane_state->hw, &from_plane_state->hw,
-              sizeof(plane_state->hw));
-
-       if (plane_state->hw.fb)
-               drm_framebuffer_get(plane_state->hw.fb);
-}
-
-void intel_plane_set_invisible(struct intel_crtc_state *crtc_state,
-                              struct intel_plane_state *plane_state)
-{
-       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
-
-       crtc_state->active_planes &= ~BIT(plane->id);
-       crtc_state->scaled_planes &= ~BIT(plane->id);
-       crtc_state->nv12_planes &= ~BIT(plane->id);
-       crtc_state->c8_planes &= ~BIT(plane->id);
-       crtc_state->async_flip_planes &= ~BIT(plane->id);
-       crtc_state->data_rate[plane->id] = 0;
-       crtc_state->data_rate_y[plane->id] = 0;
-       crtc_state->rel_data_rate[plane->id] = 0;
-       crtc_state->rel_data_rate_y[plane->id] = 0;
-       crtc_state->min_cdclk[plane->id] = 0;
-
-       plane_state->uapi.visible = false;
-}
-
-static bool intel_plane_is_scaled(const struct intel_plane_state *plane_state)
-{
-       int src_w = drm_rect_width(&plane_state->uapi.src) >> 16;
-       int src_h = drm_rect_height(&plane_state->uapi.src) >> 16;
-       int dst_w = drm_rect_width(&plane_state->uapi.dst);
-       int dst_h = drm_rect_height(&plane_state->uapi.dst);
-
-       return src_w != dst_w || src_h != dst_h;
-}
-
-static bool intel_plane_do_async_flip(struct intel_plane *plane,
-                                     const struct intel_crtc_state *old_crtc_state,
-                                     const struct intel_crtc_state *new_crtc_state)
-{
-       struct intel_display *display = to_intel_display(plane);
-
-       if (!plane->async_flip)
-               return false;
-
-       if (!new_crtc_state->uapi.async_flip)
-               return false;
-
-       /*
-        * In platforms after DISPLAY13, we might need to override
-        * first async flip in order to change watermark levels
-        * as part of optimization.
-        *
-        * And let's do this for all skl+ so that we can eg. change the
-        * modifier as well.
-        *
-        * TODO: For older platforms there is less reason to do this as
-        * only X-tile is supported with async flips, though we could
-        * extend this so other scanout parameters (stride/etc) could
-        * be changed as well...
-        */
-       return DISPLAY_VER(display) < 9 || old_crtc_state->uapi.async_flip;
-}
-
-static bool i9xx_must_disable_cxsr(const struct intel_crtc_state *new_crtc_state,
-                                  const struct intel_plane_state *old_plane_state,
-                                  const struct intel_plane_state *new_plane_state)
-{
-       struct intel_plane *plane = to_intel_plane(new_plane_state->uapi.plane);
-       bool old_visible = old_plane_state->uapi.visible;
-       bool new_visible = new_plane_state->uapi.visible;
-       u32 old_ctl = old_plane_state->ctl;
-       u32 new_ctl = new_plane_state->ctl;
-       bool modeset, turn_on, turn_off;
-
-       if (plane->id == PLANE_CURSOR)
-               return false;
-
-       modeset = intel_crtc_needs_modeset(new_crtc_state);
-       turn_off = old_visible && (!new_visible || modeset);
-       turn_on = new_visible && (!old_visible || modeset);
-
-       /* Must disable CxSR around plane enable/disable */
-       if (turn_on || turn_off)
-               return true;
-
-       if (!old_visible || !new_visible)
-               return false;
-
-       /*
-        * Most plane control register updates are blocked while in CxSR.
-        *
-        * Tiling mode is one exception where the primary plane can
-        * apparently handle it, whereas the sprites can not (the
-        * sprite issue being only relevant on VLV/CHV where CxSR
-        * is actually possible with a sprite enabled).
-        */
-       if (plane->id == PLANE_PRIMARY) {
-               old_ctl &= ~DISP_TILED;
-               new_ctl &= ~DISP_TILED;
-       }
-
-       return old_ctl != new_ctl;
-}
-
-static bool ilk_must_disable_cxsr(const struct intel_crtc_state *new_crtc_state,
-                                 const struct intel_plane_state *old_plane_state,
-                                 const struct intel_plane_state *new_plane_state)
-{
-       struct intel_plane *plane = to_intel_plane(new_plane_state->uapi.plane);
-       bool old_visible = old_plane_state->uapi.visible;
-       bool new_visible = new_plane_state->uapi.visible;
-       bool modeset, turn_on;
-
-       if (plane->id == PLANE_CURSOR)
-               return false;
-
-       modeset = intel_crtc_needs_modeset(new_crtc_state);
-       turn_on = new_visible && (!old_visible || modeset);
-
-       /*
-        * ILK/SNB DVSACNTR/Sprite Enable
-        * IVB SPR_CTL/Sprite Enable
-        * "When in Self Refresh Big FIFO mode, a write to enable the
-        *  plane will be internally buffered and delayed while Big FIFO
-        *  mode is exiting."
-        *
-        * Which means that enabling the sprite can take an extra frame
-        * when we start in big FIFO mode (LP1+). Thus we need to drop
-        * down to LP0 and wait for vblank in order to make sure the
-        * sprite gets enabled on the next vblank after the register write.
-        * Doing otherwise would risk enabling the sprite one frame after
-        * we've already signalled flip completion. We can resume LP1+
-        * once the sprite has been enabled.
-        *
-        * With experimental results seems this is needed also for primary
-        * plane, not only sprite plane.
-        */
-       if (turn_on)
-               return true;
-
-       /*
-        * WaCxSRDisabledForSpriteScaling:ivb
-        * IVB SPR_SCALE/Scaling Enable
-        * "Low Power watermarks must be disabled for at least one
-        *  frame before enabling sprite scaling, and kept disabled
-        *  until sprite scaling is disabled."
-        *
-        * ILK/SNB DVSASCALE/Scaling Enable
-        * "When in Self Refresh Big FIFO mode, scaling enable will be
-        *  masked off while Big FIFO mode is exiting."
-        *
-        * Despite the w/a only being listed for IVB we assume that
-        * the ILK/SNB note has similar ramifications, hence we apply
-        * the w/a on all three platforms.
-        */
-       return !intel_plane_is_scaled(old_plane_state) &&
-               intel_plane_is_scaled(new_plane_state);
-}
-
-static int intel_plane_atomic_calc_changes(const struct intel_crtc_state *old_crtc_state,
-                                          struct intel_crtc_state *new_crtc_state,
-                                          const struct intel_plane_state *old_plane_state,
-                                          struct intel_plane_state *new_plane_state)
-{
-       struct intel_display *display = to_intel_display(new_crtc_state);
-       struct intel_crtc *crtc = to_intel_crtc(new_crtc_state->uapi.crtc);
-       struct intel_plane *plane = to_intel_plane(new_plane_state->uapi.plane);
-       bool mode_changed = intel_crtc_needs_modeset(new_crtc_state);
-       bool was_crtc_enabled = old_crtc_state->hw.active;
-       bool is_crtc_enabled = new_crtc_state->hw.active;
-       bool turn_off, turn_on, visible, was_visible;
-       int ret;
-
-       if (DISPLAY_VER(display) >= 9 && plane->id != PLANE_CURSOR) {
-               ret = skl_update_scaler_plane(new_crtc_state, new_plane_state);
-               if (ret)
-                       return ret;
-       }
-
-       was_visible = old_plane_state->uapi.visible;
-       visible = new_plane_state->uapi.visible;
-
-       if (!was_crtc_enabled && drm_WARN_ON(display->drm, was_visible))
-               was_visible = false;
-
-       /*
-        * Visibility is calculated as if the crtc was on, but
-        * after scaler setup everything depends on it being off
-        * when the crtc isn't active.
-        *
-        * FIXME this is wrong for watermarks. Watermarks should also
-        * be computed as if the pipe would be active. Perhaps move
-        * per-plane wm computation to the .check_plane() hook, and
-        * only combine the results from all planes in the current place?
-        */
-       if (!is_crtc_enabled) {
-               intel_plane_set_invisible(new_crtc_state, new_plane_state);
-               visible = false;
-       }
-
-       if (!was_visible && !visible)
-               return 0;
-
-       turn_off = was_visible && (!visible || mode_changed);
-       turn_on = visible && (!was_visible || mode_changed);
-
-       drm_dbg_atomic(display->drm,
-                      "[CRTC:%d:%s] with [PLANE:%d:%s] visible %i -> %i, off %i, on %i, ms %i\n",
-                      crtc->base.base.id, crtc->base.name,
-                      plane->base.base.id, plane->base.name,
-                      was_visible, visible,
-                      turn_off, turn_on, mode_changed);
-
-       if (visible || was_visible)
-               new_crtc_state->fb_bits |= plane->frontbuffer_bit;
-
-       if (HAS_GMCH(display) &&
-           i9xx_must_disable_cxsr(new_crtc_state, old_plane_state, new_plane_state))
-               new_crtc_state->disable_cxsr = true;
-
-       if ((display->platform.ironlake || display->platform.sandybridge || display->platform.ivybridge) &&
-           ilk_must_disable_cxsr(new_crtc_state, old_plane_state, new_plane_state))
-               new_crtc_state->disable_cxsr = true;
-
-       if (intel_plane_do_async_flip(plane, old_crtc_state, new_crtc_state)) {
-               new_crtc_state->do_async_flip = true;
-               new_crtc_state->async_flip_planes |= BIT(plane->id);
-       } else if (plane->need_async_flip_toggle_wa &&
-                  new_crtc_state->uapi.async_flip) {
-               /*
-                * On platforms with double buffered async flip bit we
-                * set the bit already one frame early during the sync
-                * flip (see {i9xx,skl}_plane_update_arm()). The
-                * hardware will therefore be ready to perform a real
-                * async flip during the next commit, without having
-                * to wait yet another frame for the bit to latch.
-                */
-               new_crtc_state->async_flip_planes |= BIT(plane->id);
-       }
-
-       return 0;
-}
-
-int intel_plane_atomic_check_with_state(const struct intel_crtc_state *old_crtc_state,
-                                       struct intel_crtc_state *new_crtc_state,
-                                       const struct intel_plane_state *old_plane_state,
-                                       struct intel_plane_state *new_plane_state)
-{
-       struct intel_plane *plane = to_intel_plane(new_plane_state->uapi.plane);
-       const struct drm_framebuffer *fb = new_plane_state->hw.fb;
-       int ret;
-
-       intel_plane_set_invisible(new_crtc_state, new_plane_state);
-       new_crtc_state->enabled_planes &= ~BIT(plane->id);
-
-       if (!new_plane_state->hw.crtc && !old_plane_state->hw.crtc)
-               return 0;
-
-       ret = plane->check_plane(new_crtc_state, new_plane_state);
-       if (ret)
-               return ret;
-
-       if (fb)
-               new_crtc_state->enabled_planes |= BIT(plane->id);
-
-       /* FIXME pre-g4x don't work like this */
-       if (new_plane_state->uapi.visible)
-               new_crtc_state->active_planes |= BIT(plane->id);
-
-       if (new_plane_state->uapi.visible &&
-           intel_plane_is_scaled(new_plane_state))
-               new_crtc_state->scaled_planes |= BIT(plane->id);
-
-       if (new_plane_state->uapi.visible &&
-           intel_format_info_is_yuv_semiplanar(fb->format, fb->modifier))
-               new_crtc_state->nv12_planes |= BIT(plane->id);
-
-       if (new_plane_state->uapi.visible &&
-           fb->format->format == DRM_FORMAT_C8)
-               new_crtc_state->c8_planes |= BIT(plane->id);
-
-       if (new_plane_state->uapi.visible || old_plane_state->uapi.visible)
-               new_crtc_state->update_planes |= BIT(plane->id);
-
-       if (new_plane_state->uapi.visible &&
-           intel_format_info_is_yuv_semiplanar(fb->format, fb->modifier)) {
-               new_crtc_state->data_rate_y[plane->id] =
-                       intel_plane_data_rate(new_crtc_state, new_plane_state, 0);
-               new_crtc_state->data_rate[plane->id] =
-                       intel_plane_data_rate(new_crtc_state, new_plane_state, 1);
-
-               new_crtc_state->rel_data_rate_y[plane->id] =
-                       intel_plane_relative_data_rate(new_crtc_state,
-                                                      new_plane_state, 0);
-               new_crtc_state->rel_data_rate[plane->id] =
-                       intel_plane_relative_data_rate(new_crtc_state,
-                                                      new_plane_state, 1);
-       } else if (new_plane_state->uapi.visible) {
-               new_crtc_state->data_rate[plane->id] =
-                       intel_plane_data_rate(new_crtc_state, new_plane_state, 0);
-
-               new_crtc_state->rel_data_rate[plane->id] =
-                       intel_plane_relative_data_rate(new_crtc_state,
-                                                      new_plane_state, 0);
-       }
-
-       return intel_plane_atomic_calc_changes(old_crtc_state, new_crtc_state,
-                                              old_plane_state, new_plane_state);
-}
-
-struct intel_plane *
-intel_crtc_get_plane(struct intel_crtc *crtc, enum plane_id plane_id)
-{
-       struct intel_display *display = to_intel_display(crtc);
-       struct intel_plane *plane;
-
-       for_each_intel_plane_on_crtc(display->drm, crtc, plane) {
-               if (plane->id == plane_id)
-                       return plane;
-       }
-
-       return NULL;
-}
-
-int intel_plane_atomic_check(struct intel_atomic_state *state,
-                            struct intel_plane *plane)
-{
-       struct intel_display *display = to_intel_display(state);
-       struct intel_plane_state *new_plane_state =
-               intel_atomic_get_new_plane_state(state, plane);
-       const struct intel_plane_state *old_plane_state =
-               intel_atomic_get_old_plane_state(state, plane);
-       const struct intel_plane_state *new_primary_crtc_plane_state;
-       const struct intel_plane_state *old_primary_crtc_plane_state;
-       struct intel_crtc *crtc = intel_crtc_for_pipe(display, plane->pipe);
-       const struct intel_crtc_state *old_crtc_state =
-               intel_atomic_get_old_crtc_state(state, crtc);
-       struct intel_crtc_state *new_crtc_state =
-               intel_atomic_get_new_crtc_state(state, crtc);
-
-       if (new_crtc_state && intel_crtc_is_joiner_secondary(new_crtc_state)) {
-               struct intel_crtc *primary_crtc =
-                       intel_primary_crtc(new_crtc_state);
-               struct intel_plane *primary_crtc_plane =
-                       intel_crtc_get_plane(primary_crtc, plane->id);
-
-               new_primary_crtc_plane_state =
-                       intel_atomic_get_new_plane_state(state, primary_crtc_plane);
-               old_primary_crtc_plane_state =
-                       intel_atomic_get_old_plane_state(state, primary_crtc_plane);
-       } else {
-               new_primary_crtc_plane_state = new_plane_state;
-               old_primary_crtc_plane_state = old_plane_state;
-       }
-
-       intel_plane_copy_uapi_plane_damage(new_plane_state,
-                                          old_primary_crtc_plane_state,
-                                          new_primary_crtc_plane_state);
-
-       intel_plane_copy_uapi_to_hw_state(new_plane_state,
-                                         new_primary_crtc_plane_state,
-                                         crtc);
-
-       new_plane_state->uapi.visible = false;
-       if (!new_crtc_state)
-               return 0;
-
-       return intel_plane_atomic_check_with_state(old_crtc_state,
-                                                  new_crtc_state,
-                                                  old_plane_state,
-                                                  new_plane_state);
-}
-
-static struct intel_plane *
-skl_next_plane_to_commit(struct intel_atomic_state *state,
-                        struct intel_crtc *crtc,
-                        struct skl_ddb_entry ddb[I915_MAX_PLANES],
-                        struct skl_ddb_entry ddb_y[I915_MAX_PLANES],
-                        unsigned int *update_mask)
-{
-       struct intel_crtc_state *crtc_state =
-               intel_atomic_get_new_crtc_state(state, crtc);
-       struct intel_plane_state __maybe_unused *plane_state;
-       struct intel_plane *plane;
-       int i;
-
-       if (*update_mask == 0)
-               return NULL;
-
-       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
-               enum plane_id plane_id = plane->id;
-
-               if (crtc->pipe != plane->pipe ||
-                   !(*update_mask & BIT(plane_id)))
-                       continue;
-
-               if (skl_ddb_allocation_overlaps(&crtc_state->wm.skl.plane_ddb[plane_id],
-                                               ddb, I915_MAX_PLANES, plane_id) ||
-                   skl_ddb_allocation_overlaps(&crtc_state->wm.skl.plane_ddb_y[plane_id],
-                                               ddb_y, I915_MAX_PLANES, plane_id))
-                       continue;
-
-               *update_mask &= ~BIT(plane_id);
-               ddb[plane_id] = crtc_state->wm.skl.plane_ddb[plane_id];
-               ddb_y[plane_id] = crtc_state->wm.skl.plane_ddb_y[plane_id];
-
-               return plane;
-       }
-
-       /* should never happen */
-       drm_WARN_ON(state->base.dev, 1);
-
-       return NULL;
-}
-
-void intel_plane_update_noarm(struct intel_dsb *dsb,
-                             struct intel_plane *plane,
-                             const struct intel_crtc_state *crtc_state,
-                             const struct intel_plane_state *plane_state)
-{
-       struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
-
-       trace_intel_plane_update_noarm(plane_state, crtc);
-
-       if (plane->fbc)
-               intel_fbc_dirty_rect_update_noarm(dsb, plane);
-
-       if (plane->update_noarm)
-               plane->update_noarm(dsb, plane, crtc_state, plane_state);
-}
-
-void intel_plane_async_flip(struct intel_dsb *dsb,
-                           struct intel_plane *plane,
-                           const struct intel_crtc_state *crtc_state,
-                           const struct intel_plane_state *plane_state,
-                           bool async_flip)
-{
-       struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
-
-       trace_intel_plane_async_flip(plane, crtc, async_flip);
-       plane->async_flip(dsb, plane, crtc_state, plane_state, async_flip);
-}
-
-void intel_plane_update_arm(struct intel_dsb *dsb,
-                           struct intel_plane *plane,
-                           const struct intel_crtc_state *crtc_state,
-                           const struct intel_plane_state *plane_state)
-{
-       struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
-
-       if (crtc_state->do_async_flip && plane->async_flip) {
-               intel_plane_async_flip(dsb, plane, crtc_state, plane_state, true);
-               return;
-       }
-
-       trace_intel_plane_update_arm(plane_state, crtc);
-       plane->update_arm(dsb, plane, crtc_state, plane_state);
-}
-
-void intel_plane_disable_arm(struct intel_dsb *dsb,
-                            struct intel_plane *plane,
-                            const struct intel_crtc_state *crtc_state)
-{
-       struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
-
-       trace_intel_plane_disable_arm(plane, crtc);
-       plane->disable_arm(dsb, plane, crtc_state);
-}
-
-void intel_crtc_planes_update_noarm(struct intel_dsb *dsb,
-                                   struct intel_atomic_state *state,
-                                   struct intel_crtc *crtc)
-{
-       struct intel_crtc_state *new_crtc_state =
-               intel_atomic_get_new_crtc_state(state, crtc);
-       u32 update_mask = new_crtc_state->update_planes;
-       struct intel_plane_state *new_plane_state;
-       struct intel_plane *plane;
-       int i;
-
-       if (new_crtc_state->do_async_flip)
-               return;
-
-       /*
-        * Since we only write non-arming registers here,
-        * the order does not matter even for skl+.
-        */
-       for_each_new_intel_plane_in_state(state, plane, new_plane_state, i) {
-               if (crtc->pipe != plane->pipe ||
-                   !(update_mask & BIT(plane->id)))
-                       continue;
-
-               /* TODO: for mailbox updates this should be skipped */
-               if (new_plane_state->uapi.visible ||
-                   new_plane_state->is_y_plane)
-                       intel_plane_update_noarm(dsb, plane,
-                                                new_crtc_state, new_plane_state);
-       }
-}
-
-static void skl_crtc_planes_update_arm(struct intel_dsb *dsb,
-                                      struct intel_atomic_state *state,
-                                      struct intel_crtc *crtc)
-{
-       struct intel_crtc_state *old_crtc_state =
-               intel_atomic_get_old_crtc_state(state, crtc);
-       struct intel_crtc_state *new_crtc_state =
-               intel_atomic_get_new_crtc_state(state, crtc);
-       struct skl_ddb_entry ddb[I915_MAX_PLANES];
-       struct skl_ddb_entry ddb_y[I915_MAX_PLANES];
-       u32 update_mask = new_crtc_state->update_planes;
-       struct intel_plane *plane;
-
-       memcpy(ddb, old_crtc_state->wm.skl.plane_ddb,
-              sizeof(old_crtc_state->wm.skl.plane_ddb));
-       memcpy(ddb_y, old_crtc_state->wm.skl.plane_ddb_y,
-              sizeof(old_crtc_state->wm.skl.plane_ddb_y));
-
-       while ((plane = skl_next_plane_to_commit(state, crtc, ddb, ddb_y, &update_mask))) {
-               struct intel_plane_state *new_plane_state =
-                       intel_atomic_get_new_plane_state(state, plane);
-
-               /*
-                * TODO: for mailbox updates intel_plane_update_noarm()
-                * would have to be called here as well.
-                */
-               if (new_plane_state->uapi.visible ||
-                   new_plane_state->is_y_plane)
-                       intel_plane_update_arm(dsb, plane, new_crtc_state, new_plane_state);
-               else
-                       intel_plane_disable_arm(dsb, plane, new_crtc_state);
-       }
-}
-
-static void i9xx_crtc_planes_update_arm(struct intel_dsb *dsb,
-                                       struct intel_atomic_state *state,
-                                       struct intel_crtc *crtc)
-{
-       struct intel_crtc_state *new_crtc_state =
-               intel_atomic_get_new_crtc_state(state, crtc);
-       u32 update_mask = new_crtc_state->update_planes;
-       struct intel_plane_state *new_plane_state;
-       struct intel_plane *plane;
-       int i;
-
-       for_each_new_intel_plane_in_state(state, plane, new_plane_state, i) {
-               if (crtc->pipe != plane->pipe ||
-                   !(update_mask & BIT(plane->id)))
-                       continue;
-
-               /*
-                * TODO: for mailbox updates intel_plane_update_noarm()
-                * would have to be called here as well.
-                */
-               if (new_plane_state->uapi.visible)
-                       intel_plane_update_arm(dsb, plane, new_crtc_state, new_plane_state);
-               else
-                       intel_plane_disable_arm(dsb, plane, new_crtc_state);
-       }
-}
-
-void intel_crtc_planes_update_arm(struct intel_dsb *dsb,
-                                 struct intel_atomic_state *state,
-                                 struct intel_crtc *crtc)
-{
-       struct intel_display *display = to_intel_display(state);
-
-       if (DISPLAY_VER(display) >= 9)
-               skl_crtc_planes_update_arm(dsb, state, crtc);
-       else
-               i9xx_crtc_planes_update_arm(dsb, state, crtc);
-}
-
-int intel_atomic_plane_check_clipping(struct intel_plane_state *plane_state,
-                                     struct intel_crtc_state *crtc_state,
-                                     int min_scale, int max_scale,
-                                     bool can_position)
-{
-       struct intel_display *display = to_intel_display(plane_state);
-       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
-       struct drm_framebuffer *fb = plane_state->hw.fb;
-       struct drm_rect *src = &plane_state->uapi.src;
-       struct drm_rect *dst = &plane_state->uapi.dst;
-       const struct drm_rect *clip = &crtc_state->pipe_src;
-       unsigned int rotation = plane_state->hw.rotation;
-       int hscale, vscale;
-
-       if (!fb) {
-               plane_state->uapi.visible = false;
-               return 0;
-       }
-
-       drm_rect_rotate(src, fb->width << 16, fb->height << 16, rotation);
-
-       /* Check scaling */
-       hscale = drm_rect_calc_hscale(src, dst, min_scale, max_scale);
-       vscale = drm_rect_calc_vscale(src, dst, min_scale, max_scale);
-       if (hscale < 0 || vscale < 0) {
-               drm_dbg_kms(display->drm,
-                           "[PLANE:%d:%s] invalid scaling "DRM_RECT_FP_FMT " -> " DRM_RECT_FMT "\n",
-                           plane->base.base.id, plane->base.name,
-                           DRM_RECT_FP_ARG(src), DRM_RECT_ARG(dst));
-               return -ERANGE;
-       }
-
-       /*
-        * FIXME: This might need further adjustment for seamless scaling
-        * with phase information, for the 2p2 and 2p1 scenarios.
-        */
-       plane_state->uapi.visible = drm_rect_clip_scaled(src, dst, clip);
-
-       drm_rect_rotate_inv(src, fb->width << 16, fb->height << 16, rotation);
-
-       if (!can_position && plane_state->uapi.visible &&
-           !drm_rect_equals(dst, clip)) {
-               drm_dbg_kms(display->drm,
-                           "[PLANE:%d:%s] plane (" DRM_RECT_FMT ") must cover entire CRTC (" DRM_RECT_FMT ")\n",
-                           plane->base.base.id, plane->base.name,
-                           DRM_RECT_ARG(dst), DRM_RECT_ARG(clip));
-               return -EINVAL;
-       }
-
-       /* final plane coordinates will be relative to the plane's pipe */
-       drm_rect_translate(dst, -clip->x1, -clip->y1);
-
-       return 0;
-}
-
-int intel_plane_check_src_coordinates(struct intel_plane_state *plane_state)
-{
-       struct intel_display *display = to_intel_display(plane_state);
-       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
-       const struct drm_framebuffer *fb = plane_state->hw.fb;
-       struct drm_rect *src = &plane_state->uapi.src;
-       u32 src_x, src_y, src_w, src_h, hsub, vsub;
-       bool rotated = drm_rotation_90_or_270(plane_state->hw.rotation);
-
-       /*
-        * FIXME hsub/vsub vs. block size is a mess. Pre-tgl CCS
-        * abuses hsub/vsub so we can't use them here. But as they
-        * are limited to 32bpp RGB formats we don't actually need
-        * to check anything.
-        */
-       if (fb->modifier == I915_FORMAT_MOD_Y_TILED_CCS ||
-           fb->modifier == I915_FORMAT_MOD_Yf_TILED_CCS)
-               return 0;
-
-       /*
-        * Hardware doesn't handle subpixel coordinates.
-        * Adjust to (macro)pixel boundary, but be careful not to
-        * increase the source viewport size, because that could
-        * push the downscaling factor out of bounds.
-        */
-       src_x = src->x1 >> 16;
-       src_w = drm_rect_width(src) >> 16;
-       src_y = src->y1 >> 16;
-       src_h = drm_rect_height(src) >> 16;
-
-       drm_rect_init(src, src_x << 16, src_y << 16,
-                     src_w << 16, src_h << 16);
-
-       if (fb->format->format == DRM_FORMAT_RGB565 && rotated) {
-               hsub = 2;
-               vsub = 2;
-       } else if (DISPLAY_VER(display) >= 20 &&
-                  intel_format_info_is_yuv_semiplanar(fb->format, fb->modifier)) {
-               /*
-                * This allows NV12 and P0xx formats to have odd size and/or odd
-                * source coordinates on DISPLAY_VER(display) >= 20
-                */
-               hsub = 1;
-               vsub = 1;
-
-               /* Wa_16023981245 */
-               if ((DISPLAY_VERx100(display) == 2000 ||
-                    DISPLAY_VERx100(display) == 3000 ||
-                    DISPLAY_VERx100(display) == 3002) &&
-                    src_x % 2 != 0)
-                       hsub = 2;
-       } else {
-               hsub = fb->format->hsub;
-               vsub = fb->format->vsub;
-       }
-
-       if (rotated)
-               hsub = vsub = max(hsub, vsub);
-
-       if (src_x % hsub || src_w % hsub) {
-               drm_dbg_kms(display->drm,
-                           "[PLANE:%d:%s] src x/w (%u, %u) must be a multiple of %u (rotated: %s)\n",
-                           plane->base.base.id, plane->base.name,
-                           src_x, src_w, hsub, str_yes_no(rotated));
-               return -EINVAL;
-       }
-
-       if (src_y % vsub || src_h % vsub) {
-               drm_dbg_kms(display->drm,
-                           "[PLANE:%d:%s] src y/h (%u, %u) must be a multiple of %u (rotated: %s)\n",
-                           plane->base.base.id, plane->base.name,
-                           src_y, src_h, vsub, str_yes_no(rotated));
-               return -EINVAL;
-       }
-
-       return 0;
-}
-
-static int add_dma_resv_fences(struct dma_resv *resv,
-                              struct drm_plane_state *new_plane_state)
-{
-       struct dma_fence *fence = dma_fence_get(new_plane_state->fence);
-       struct dma_fence *new;
-       int ret;
-
-       ret = dma_resv_get_singleton(resv, dma_resv_usage_rw(false), &new);
-       if (ret)
-               goto error;
-
-       if (new && fence) {
-               struct dma_fence_chain *chain = dma_fence_chain_alloc();
-
-               if (!chain) {
-                       ret = -ENOMEM;
-                       goto error;
-               }
-
-               dma_fence_chain_init(chain, fence, new, 1);
-               fence = &chain->base;
-
-       } else if (new) {
-               fence = new;
-       }
-
-       dma_fence_put(new_plane_state->fence);
-       new_plane_state->fence = fence;
-       return 0;
-
-error:
-       dma_fence_put(fence);
-       return ret;
-}
-
-/**
- * intel_prepare_plane_fb - Prepare fb for usage on plane
- * @_plane: drm plane to prepare for
- * @_new_plane_state: the plane state being prepared
- *
- * Prepares a framebuffer for usage on a display plane.  Generally this
- * involves pinning the underlying object and updating the frontbuffer tracking
- * bits.  Some older platforms need special physical address handling for
- * cursor planes.
- *
- * Returns 0 on success, negative error code on failure.
- */
-static int
-intel_prepare_plane_fb(struct drm_plane *_plane,
-                      struct drm_plane_state *_new_plane_state)
-{
-       struct i915_sched_attr attr = { .priority = I915_PRIORITY_DISPLAY };
-       struct intel_plane *plane = to_intel_plane(_plane);
-       struct intel_display *display = to_intel_display(plane);
-       struct intel_plane_state *new_plane_state =
-               to_intel_plane_state(_new_plane_state);
-       struct intel_atomic_state *state =
-               to_intel_atomic_state(new_plane_state->uapi.state);
-       struct intel_plane_state *old_plane_state =
-               intel_atomic_get_old_plane_state(state, plane);
-       struct drm_gem_object *obj = intel_fb_bo(new_plane_state->hw.fb);
-       struct drm_gem_object *old_obj = intel_fb_bo(old_plane_state->hw.fb);
-       int ret;
-
-       if (old_obj) {
-               const struct intel_crtc_state *new_crtc_state =
-                       intel_atomic_get_new_crtc_state(state,
-                                                       to_intel_crtc(old_plane_state->hw.crtc));
-
-               /* Big Hammer, we also need to ensure that any pending
-                * MI_WAIT_FOR_EVENT inside a user batch buffer on the
-                * current scanout is retired before unpinning the old
-                * framebuffer. Note that we rely on userspace rendering
-                * into the buffer attached to the pipe they are waiting
-                * on. If not, userspace generates a GPU hang with IPEHR
-                * point to the MI_WAIT_FOR_EVENT.
-                *
-                * This should only fail upon a hung GPU, in which case we
-                * can safely continue.
-                */
-               if (intel_crtc_needs_modeset(new_crtc_state)) {
-                       ret = add_dma_resv_fences(old_obj->resv,
-                                                 &new_plane_state->uapi);
-                       if (ret < 0)
-                               return ret;
-               }
-       }
-
-       if (!obj)
-               return 0;
-
-       ret = intel_plane_pin_fb(new_plane_state, old_plane_state);
-       if (ret)
-               return ret;
-
-       ret = drm_gem_plane_helper_prepare_fb(&plane->base, &new_plane_state->uapi);
-       if (ret < 0)
-               goto unpin_fb;
-
-       if (new_plane_state->uapi.fence) {
-               i915_gem_fence_wait_priority(new_plane_state->uapi.fence,
-                                            &attr);
-
-               intel_display_rps_boost_after_vblank(new_plane_state->hw.crtc,
-                                                    new_plane_state->uapi.fence);
-       }
-
-       /*
-        * We declare pageflips to be interactive and so merit a small bias
-        * towards upclocking to deliver the frame on time. By only changing
-        * the RPS thresholds to sample more regularly and aim for higher
-        * clocks we can hopefully deliver low power workloads (like kodi)
-        * that are not quite steady state without resorting to forcing
-        * maximum clocks following a vblank miss (see do_rps_boost()).
-        */
-       intel_display_rps_mark_interactive(display, state, true);
-
-       return 0;
-
-unpin_fb:
-       intel_plane_unpin_fb(new_plane_state);
-
-       return ret;
-}
-
-/**
- * intel_cleanup_plane_fb - Cleans up an fb after plane use
- * @plane: drm plane to clean up for
- * @_old_plane_state: the state from the previous modeset
- *
- * Cleans up a framebuffer that has just been removed from a plane.
- */
-static void
-intel_cleanup_plane_fb(struct drm_plane *plane,
-                      struct drm_plane_state *_old_plane_state)
-{
-       struct intel_display *display = to_intel_display(plane->dev);
-       struct intel_plane_state *old_plane_state =
-               to_intel_plane_state(_old_plane_state);
-       struct intel_atomic_state *state =
-               to_intel_atomic_state(old_plane_state->uapi.state);
-       struct drm_gem_object *obj = intel_fb_bo(old_plane_state->hw.fb);
-
-       if (!obj)
-               return;
-
-       intel_display_rps_mark_interactive(display, state, false);
-
-       intel_plane_unpin_fb(old_plane_state);
-}
-
-static const struct drm_plane_helper_funcs intel_plane_helper_funcs = {
-       .prepare_fb = intel_prepare_plane_fb,
-       .cleanup_fb = intel_cleanup_plane_fb,
-};
-
-void intel_plane_helper_add(struct intel_plane *plane)
-{
-       drm_plane_helper_add(&plane->base, &intel_plane_helper_funcs);
-}
-
-void intel_plane_init_cursor_vblank_work(struct intel_plane_state *old_plane_state,
-                                        struct intel_plane_state *new_plane_state)
-{
-       if (!old_plane_state->ggtt_vma ||
-           old_plane_state->ggtt_vma == new_plane_state->ggtt_vma)
-               return;
-
-       drm_vblank_work_init(&old_plane_state->unpin_work, old_plane_state->uapi.crtc,
-                            intel_cursor_unpin_work);
-}
-
-static void link_nv12_planes(struct intel_crtc_state *crtc_state,
-                            struct intel_plane_state *uv_plane_state,
-                            struct intel_plane_state *y_plane_state)
-{
-       struct intel_display *display = to_intel_display(uv_plane_state);
-       struct intel_plane *uv_plane = to_intel_plane(uv_plane_state->uapi.plane);
-       struct intel_plane *y_plane = to_intel_plane(y_plane_state->uapi.plane);
-
-       drm_dbg_kms(display->drm, "UV plane [PLANE:%d:%s] using Y plane [PLANE:%d:%s]\n",
-                   uv_plane->base.base.id, uv_plane->base.name,
-                   y_plane->base.base.id, y_plane->base.name);
-
-       uv_plane_state->planar_linked_plane = y_plane;
-
-       y_plane_state->is_y_plane = true;
-       y_plane_state->planar_linked_plane = uv_plane;
-
-       crtc_state->enabled_planes |= BIT(y_plane->id);
-       crtc_state->active_planes |= BIT(y_plane->id);
-       crtc_state->update_planes |= BIT(y_plane->id);
-
-       crtc_state->data_rate[y_plane->id] = crtc_state->data_rate_y[uv_plane->id];
-       crtc_state->rel_data_rate[y_plane->id] = crtc_state->rel_data_rate_y[uv_plane->id];
-
-       /* Copy parameters to Y plane */
-       intel_plane_copy_hw_state(y_plane_state, uv_plane_state);
-       y_plane_state->uapi.src = uv_plane_state->uapi.src;
-       y_plane_state->uapi.dst = uv_plane_state->uapi.dst;
-
-       y_plane_state->ctl = uv_plane_state->ctl;
-       y_plane_state->color_ctl = uv_plane_state->color_ctl;
-       y_plane_state->view = uv_plane_state->view;
-       y_plane_state->decrypt = uv_plane_state->decrypt;
-
-       icl_link_nv12_planes(uv_plane_state, y_plane_state);
-}
-
-static void unlink_nv12_plane(struct intel_crtc_state *crtc_state,
-                             struct intel_plane_state *plane_state)
-{
-       struct intel_display *display = to_intel_display(plane_state);
-       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
-
-       plane_state->planar_linked_plane = NULL;
-
-       if (!plane_state->is_y_plane)
-               return;
-
-       drm_WARN_ON(display->drm, plane_state->uapi.visible);
-
-       plane_state->is_y_plane = false;
-
-       crtc_state->enabled_planes &= ~BIT(plane->id);
-       crtc_state->active_planes &= ~BIT(plane->id);
-       crtc_state->update_planes |= BIT(plane->id);
-       crtc_state->data_rate[plane->id] = 0;
-       crtc_state->rel_data_rate[plane->id] = 0;
-}
-
-static int icl_check_nv12_planes(struct intel_atomic_state *state,
-                                struct intel_crtc *crtc)
-{
-       struct intel_display *display = to_intel_display(state);
-       struct intel_crtc_state *crtc_state =
-               intel_atomic_get_new_crtc_state(state, crtc);
-       struct intel_plane_state *plane_state;
-       struct intel_plane *plane;
-       int i;
-
-       if (DISPLAY_VER(display) < 11)
-               return 0;
-
-       /*
-        * Destroy all old plane links and make the Y plane invisible
-        * in the crtc_state->active_planes mask.
-        */
-       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
-               if (plane->pipe != crtc->pipe)
-                       continue;
-
-               if (plane_state->planar_linked_plane)
-                       unlink_nv12_plane(crtc_state, plane_state);
-       }
-
-       if (!crtc_state->nv12_planes)
-               return 0;
-
-       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
-               struct intel_plane_state *y_plane_state = NULL;
-               struct intel_plane *y_plane;
-
-               if (plane->pipe != crtc->pipe)
-                       continue;
-
-               if ((crtc_state->nv12_planes & BIT(plane->id)) == 0)
-                       continue;
-
-               for_each_intel_plane_on_crtc(display->drm, crtc, y_plane) {
-                       if (!icl_is_nv12_y_plane(display, y_plane->id))
-                               continue;
-
-                       if (crtc_state->active_planes & BIT(y_plane->id))
-                               continue;
-
-                       y_plane_state = intel_atomic_get_plane_state(state, y_plane);
-                       if (IS_ERR(y_plane_state))
-                               return PTR_ERR(y_plane_state);
-
-                       break;
-               }
-
-               if (!y_plane_state) {
-                       drm_dbg_kms(display->drm,
-                                   "[CRTC:%d:%s] need %d free Y planes for planar YUV\n",
-                                   crtc->base.base.id, crtc->base.name,
-                                   hweight8(crtc_state->nv12_planes));
-                       return -EINVAL;
-               }
-
-               link_nv12_planes(crtc_state, plane_state, y_plane_state);
-       }
-
-       return 0;
-}
-
-static int intel_crtc_add_planes_to_state(struct intel_atomic_state *state,
-                                         struct intel_crtc *crtc,
-                                         u8 plane_ids_mask)
-{
-       struct intel_display *display = to_intel_display(state);
-       struct intel_plane *plane;
-
-       for_each_intel_plane_on_crtc(display->drm, crtc, plane) {
-               struct intel_plane_state *plane_state;
-
-               if ((plane_ids_mask & BIT(plane->id)) == 0)
-                       continue;
-
-               plane_state = intel_atomic_get_plane_state(state, plane);
-               if (IS_ERR(plane_state))
-                       return PTR_ERR(plane_state);
-       }
-
-       return 0;
-}
-
-int intel_atomic_add_affected_planes(struct intel_atomic_state *state,
-                                    struct intel_crtc *crtc)
-{
-       const struct intel_crtc_state *old_crtc_state =
-               intel_atomic_get_old_crtc_state(state, crtc);
-       const struct intel_crtc_state *new_crtc_state =
-               intel_atomic_get_new_crtc_state(state, crtc);
-
-       return intel_crtc_add_planes_to_state(state, crtc,
-                                             old_crtc_state->enabled_planes |
-                                             new_crtc_state->enabled_planes);
-}
-
-static bool active_planes_affects_min_cdclk(struct intel_display *display)
-{
-       /* See {hsw,vlv,ivb}_plane_ratio() */
-       return display->platform.broadwell || display->platform.haswell ||
-               display->platform.cherryview || display->platform.valleyview ||
-               display->platform.ivybridge;
-}
-
-static u8 intel_joiner_affected_planes(struct intel_atomic_state *state,
-                                      u8 joined_pipes)
-{
-       const struct intel_plane_state *plane_state;
-       struct intel_plane *plane;
-       u8 affected_planes = 0;
-       int i;
-
-       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
-               struct intel_plane *linked = plane_state->planar_linked_plane;
-
-               if ((joined_pipes & BIT(plane->pipe)) == 0)
-                       continue;
-
-               affected_planes |= BIT(plane->id);
-               if (linked)
-                       affected_planes |= BIT(linked->id);
-       }
-
-       return affected_planes;
-}
-
-static int intel_joiner_add_affected_planes(struct intel_atomic_state *state,
-                                           u8 joined_pipes)
-{
-       u8 prev_affected_planes, affected_planes = 0;
-
-       /*
-        * We want all the joined pipes to have the same
-        * set of planes in the atomic state, to make sure
-        * state copying always works correctly, and the
-        * UV<->Y plane linkage is always up to date.
-        * Keep pulling planes in until we've determined
-        * the full set of affected planes. A bit complicated
-        * on account of each pipe being capable of selecting
-        * their own Y planes independently of the other pipes,
-        * and the selection being done from the set of
-        * inactive planes.
-        */
-       do {
-               struct intel_crtc *crtc;
-
-               for_each_intel_crtc_in_pipe_mask(state->base.dev, crtc, joined_pipes) {
-                       int ret;
-
-                       ret = intel_crtc_add_planes_to_state(state, crtc, affected_planes);
-                       if (ret)
-                               return ret;
-               }
-
-               prev_affected_planes = affected_planes;
-               affected_planes = intel_joiner_affected_planes(state, joined_pipes);
-       } while (affected_planes != prev_affected_planes);
-
-       return 0;
-}
-
-static int intel_add_affected_planes(struct intel_atomic_state *state)
-{
-       const struct intel_crtc_state *crtc_state;
-       struct intel_crtc *crtc;
-       int i;
-
-       for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) {
-               int ret;
-
-               ret = intel_joiner_add_affected_planes(state, intel_crtc_joined_pipe_mask(crtc_state));
-               if (ret)
-                       return ret;
-       }
-
-       return 0;
-}
-
-int intel_atomic_check_planes(struct intel_atomic_state *state)
-{
-       struct intel_display *display = to_intel_display(state);
-       struct intel_crtc_state *old_crtc_state, *new_crtc_state;
-       struct intel_plane_state __maybe_unused *plane_state;
-       struct intel_plane *plane;
-       struct intel_crtc *crtc;
-       int i, ret;
-
-       ret = intel_add_affected_planes(state);
-       if (ret)
-               return ret;
-
-       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
-               ret = intel_plane_atomic_check(state, plane);
-               if (ret) {
-                       drm_dbg_atomic(display->drm,
-                                      "[PLANE:%d:%s] atomic driver check failed\n",
-                                      plane->base.base.id, plane->base.name);
-                       return ret;
-               }
-       }
-
-       for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state,
-                                           new_crtc_state, i) {
-               u8 old_active_planes, new_active_planes;
-
-               ret = icl_check_nv12_planes(state, crtc);
-               if (ret)
-                       return ret;
-
-               /*
-                * On some platforms the number of active planes affects
-                * the planes' minimum cdclk calculation. Add such planes
-                * to the state before we compute the minimum cdclk.
-                */
-               if (!active_planes_affects_min_cdclk(display))
-                       continue;
-
-               old_active_planes = old_crtc_state->active_planes & ~BIT(PLANE_CURSOR);
-               new_active_planes = new_crtc_state->active_planes & ~BIT(PLANE_CURSOR);
-
-               if (hweight8(old_active_planes) == hweight8(new_active_planes))
-                       continue;
-
-               ret = intel_crtc_add_planes_to_state(state, crtc, new_active_planes);
-               if (ret)
-                       return ret;
-       }
-
-       return 0;
-}
-
-u32 intel_plane_ggtt_offset(const struct intel_plane_state *plane_state)
-{
-       return i915_ggtt_offset(plane_state->ggtt_vma);
-}
diff --git a/drivers/gpu/drm/i915/display/intel_atomic_plane.h b/drivers/gpu/drm/i915/display/intel_atomic_plane.h
deleted file mode 100644 (file)
index 317320c..0000000
+++ /dev/null
@@ -1,97 +0,0 @@
-/* SPDX-License-Identifier: MIT */
-/*
- * Copyright © 2019 Intel Corporation
- */
-
-#ifndef __INTEL_ATOMIC_PLANE_H__
-#define __INTEL_ATOMIC_PLANE_H__
-
-#include <linux/types.h>
-
-struct drm_plane;
-struct drm_property;
-struct drm_rect;
-struct intel_atomic_state;
-struct intel_crtc;
-struct intel_crtc_state;
-struct intel_dsb;
-struct intel_plane;
-struct intel_plane_state;
-enum plane_id;
-
-struct intel_plane *
-intel_crtc_get_plane(struct intel_crtc *crtc, enum plane_id plane_id);
-bool intel_plane_can_async_flip(struct intel_plane *plane, u32 format,
-                               u64 modifier);
-unsigned int intel_adjusted_rate(const struct drm_rect *src,
-                                const struct drm_rect *dst,
-                                unsigned int rate);
-unsigned int intel_plane_pixel_rate(const struct intel_crtc_state *crtc_state,
-                                   const struct intel_plane_state *plane_state);
-
-unsigned int intel_plane_data_rate(const struct intel_crtc_state *crtc_state,
-                                  const struct intel_plane_state *plane_state,
-                                  int color_plane);
-void intel_plane_copy_uapi_to_hw_state(struct intel_plane_state *plane_state,
-                                      const struct intel_plane_state *from_plane_state,
-                                      struct intel_crtc *crtc);
-void intel_plane_copy_hw_state(struct intel_plane_state *plane_state,
-                              const struct intel_plane_state *from_plane_state);
-void intel_plane_async_flip(struct intel_dsb *dsb,
-                           struct intel_plane *plane,
-                           const struct intel_crtc_state *crtc_state,
-                           const struct intel_plane_state *plane_state,
-                           bool async_flip);
-void intel_plane_update_noarm(struct intel_dsb *dsb,
-                             struct intel_plane *plane,
-                             const struct intel_crtc_state *crtc_state,
-                             const struct intel_plane_state *plane_state);
-void intel_plane_update_arm(struct intel_dsb *dsb,
-                           struct intel_plane *plane,
-                           const struct intel_crtc_state *crtc_state,
-                           const struct intel_plane_state *plane_state);
-void intel_plane_disable_arm(struct intel_dsb *dsb,
-                            struct intel_plane *plane,
-                            const struct intel_crtc_state *crtc_state);
-struct intel_plane *intel_plane_alloc(void);
-void intel_plane_free(struct intel_plane *plane);
-void intel_plane_destroy(struct drm_plane *plane);
-struct drm_plane_state *intel_plane_duplicate_state(struct drm_plane *plane);
-void intel_plane_destroy_state(struct drm_plane *plane,
-                              struct drm_plane_state *state);
-void intel_crtc_planes_update_noarm(struct intel_dsb *dsb,
-                                   struct intel_atomic_state *state,
-                                   struct intel_crtc *crtc);
-void intel_crtc_planes_update_arm(struct intel_dsb *dsbx,
-                                 struct intel_atomic_state *state,
-                                 struct intel_crtc *crtc);
-int intel_plane_atomic_check_with_state(const struct intel_crtc_state *old_crtc_state,
-                                       struct intel_crtc_state *crtc_state,
-                                       const struct intel_plane_state *old_plane_state,
-                                       struct intel_plane_state *intel_state);
-int intel_plane_atomic_check(struct intel_atomic_state *state,
-                            struct intel_plane *plane);
-int intel_plane_calc_min_cdclk(struct intel_atomic_state *state,
-                              struct intel_plane *plane,
-                              bool *need_cdclk_calc);
-int intel_atomic_plane_check_clipping(struct intel_plane_state *plane_state,
-                                     struct intel_crtc_state *crtc_state,
-                                     int min_scale, int max_scale,
-                                     bool can_position);
-int intel_plane_check_src_coordinates(struct intel_plane_state *plane_state);
-void intel_plane_set_invisible(struct intel_crtc_state *crtc_state,
-                              struct intel_plane_state *plane_state);
-void intel_plane_helper_add(struct intel_plane *plane);
-bool intel_plane_needs_physical(struct intel_plane *plane);
-void intel_plane_init_cursor_vblank_work(struct intel_plane_state *old_plane_state,
-                                        struct intel_plane_state *new_plane_state);
-int intel_atomic_add_affected_planes(struct intel_atomic_state *state,
-                                    struct intel_crtc *crtc);
-int intel_atomic_check_planes(struct intel_atomic_state *state);
-
-u32 intel_plane_ggtt_offset(const struct intel_plane_state *plane_state);
-bool intel_plane_format_mod_supported_async(struct drm_plane *plane,
-                                           u32 format,
-                                           u64 modifier);
-
-#endif /* __INTEL_ATOMIC_PLANE_H__ */
index 1cbb1d526fe8706c075231409cd623fa8142afa5..3f68098f27c9a2566cb8a03e1589842d7901841b 100644 (file)
@@ -32,7 +32,6 @@
 #include "i915_drv.h"
 #include "i915_reg.h"
 #include "intel_atomic.h"
-#include "intel_atomic_plane.h"
 #include "intel_audio.h"
 #include "intel_bw.h"
 #include "intel_cdclk.h"
@@ -43,6 +42,7 @@
 #include "intel_mchbar_regs.h"
 #include "intel_pci_config.h"
 #include "intel_pcode.h"
+#include "intel_plane.h"
 #include "intel_psr.h"
 #include "intel_vdsc.h"
 #include "skl_watermark.h"
index a88317ea4e9c0e3002f5ca552b997b1c46e5505c..a187db6df2d36ec7dbbba56fe7aa4e9f9b92751c 100644 (file)
@@ -17,7 +17,6 @@
 #include "i9xx_plane.h"
 #include "icl_dsi.h"
 #include "intel_atomic.h"
-#include "intel_atomic_plane.h"
 #include "intel_color.h"
 #include "intel_crtc.h"
 #include "intel_cursor.h"
@@ -29,6 +28,7 @@
 #include "intel_dsi.h"
 #include "intel_fifo_underrun.h"
 #include "intel_pipe_crc.h"
+#include "intel_plane.h"
 #include "intel_psr.h"
 #include "intel_sprite.h"
 #include "intel_vblank.h"
index 6bd4f6a28cae33a29ff1a1e3adae41eb5ec329c0..27d86549af2c100285c70e6789dd50b2528c1793 100644 (file)
@@ -14,7 +14,6 @@
 
 #include "i915_utils.h"
 #include "intel_atomic.h"
-#include "intel_atomic_plane.h"
 #include "intel_cursor.h"
 #include "intel_cursor_regs.h"
 #include "intel_de.h"
@@ -23,6 +22,7 @@
 #include "intel_fb.h"
 #include "intel_fb_pin.h"
 #include "intel_frontbuffer.h"
+#include "intel_plane.h"
 #include "intel_psr.h"
 #include "intel_psr_regs.h"
 #include "intel_vblank.h"
index 6ec786198f43564129c7e6ac8929975f1b0f2abe..cbe96f21f3304262b8d190170387c7939b6fa142 100644 (file)
@@ -57,7 +57,6 @@
 #include "i9xx_wm.h"
 #include "intel_alpm.h"
 #include "intel_atomic.h"
-#include "intel_atomic_plane.h"
 #include "intel_audio.h"
 #include "intel_bo.h"
 #include "intel_bw.h"
 #include "intel_pch_refclk.h"
 #include "intel_pfit.h"
 #include "intel_pipe_crc.h"
+#include "intel_plane.h"
 #include "intel_plane_initial.h"
 #include "intel_pmdemand.h"
 #include "intel_pps.h"
index 8d0dcf252bed2625500cd3c3c9baca8049ae90ab..fb25ec8adae3ff91e804eaf15713fe90d53a3487 100644 (file)
@@ -9,7 +9,6 @@
 #include "i915_irq.h"
 #include "i915_reg.h"
 #include "icl_dsi_regs.h"
-#include "intel_atomic_plane.h"
 #include "intel_crtc.h"
 #include "intel_de.h"
 #include "intel_display_irq.h"
@@ -27,6 +26,7 @@
 #include "intel_gmbus.h"
 #include "intel_hotplug_irq.h"
 #include "intel_pipe_crc_regs.h"
+#include "intel_plane.h"
 #include "intel_pmdemand.h"
 #include "intel_psr.h"
 #include "intel_psr_regs.h"
index a5906cb4900c0be78c8f14468b4edc3f4c663902..763b36c4de1055f76f93d20d618444f24dbe0dfc 100644 (file)
@@ -11,7 +11,6 @@
 #include <drm/drm_modeset_helper.h>
 
 #include "i915_drv.h"
-#include "intel_atomic_plane.h"
 #include "intel_bo.h"
 #include "intel_display.h"
 #include "intel_display_core.h"
@@ -20,6 +19,7 @@
 #include "intel_fb.h"
 #include "intel_fb_bo.h"
 #include "intel_frontbuffer.h"
+#include "intel_plane.h"
 
 #define check_array_bounds(display, a, i) drm_WARN_ON((display)->drm, (i) >= ARRAY_SIZE(a))
 
index 98a61a7b0b9325cf23559a1b43d6a04bf8fb0987..fcc88e677e47037b0ae89e6d8eec25dcc7fdd7ae 100644 (file)
 #include "gem/i915_gem_object.h"
 
 #include "i915_drv.h"
-#include "intel_atomic_plane.h"
 #include "intel_display_core.h"
 #include "intel_display_rpm.h"
 #include "intel_display_types.h"
 #include "intel_dpt.h"
 #include "intel_fb.h"
 #include "intel_fb_pin.h"
+#include "intel_plane.h"
 
 static struct i915_vma *
 intel_fb_pin_to_dpt(const struct drm_framebuffer *fb,
diff --git a/drivers/gpu/drm/i915/display/intel_plane.c b/drivers/gpu/drm/i915/display/intel_plane.c
new file mode 100644 (file)
index 0000000..3d6f77a
--- /dev/null
@@ -0,0 +1,1588 @@
+/*
+ * Copyright © 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * DOC: atomic plane helpers
+ *
+ * The functions here are used by the atomic plane helper functions to
+ * implement legacy plane updates (i.e., drm_plane->update_plane() and
+ * drm_plane->disable_plane()).  This allows plane updates to use the
+ * atomic state infrastructure and perform plane updates as separate
+ * prepare/check/commit/cleanup steps.
+ */
+
+#include <linux/dma-fence-chain.h>
+#include <linux/dma-resv.h>
+
+#include <drm/drm_atomic_helper.h>
+#include <drm/drm_blend.h>
+#include <drm/drm_damage_helper.h>
+#include <drm/drm_fourcc.h>
+#include <drm/drm_gem.h>
+#include <drm/drm_gem_atomic_helper.h>
+
+#include "gem/i915_gem_object.h"
+#include "i915_scheduler_types.h"
+#include "i915_vma.h"
+#include "i9xx_plane_regs.h"
+#include "intel_cdclk.h"
+#include "intel_cursor.h"
+#include "intel_display_rps.h"
+#include "intel_display_trace.h"
+#include "intel_display_types.h"
+#include "intel_fb.h"
+#include "intel_fb_pin.h"
+#include "intel_plane.h"
+#include "skl_scaler.h"
+#include "skl_universal_plane.h"
+#include "skl_watermark.h"
+
+static void intel_plane_state_reset(struct intel_plane_state *plane_state,
+                                   struct intel_plane *plane)
+{
+       memset(plane_state, 0, sizeof(*plane_state));
+
+       __drm_atomic_helper_plane_state_reset(&plane_state->uapi, &plane->base);
+
+       plane_state->scaler_id = -1;
+}
+
+struct intel_plane *intel_plane_alloc(void)
+{
+       struct intel_plane_state *plane_state;
+       struct intel_plane *plane;
+
+       plane = kzalloc(sizeof(*plane), GFP_KERNEL);
+       if (!plane)
+               return ERR_PTR(-ENOMEM);
+
+       plane_state = kzalloc(sizeof(*plane_state), GFP_KERNEL);
+       if (!plane_state) {
+               kfree(plane);
+               return ERR_PTR(-ENOMEM);
+       }
+
+       intel_plane_state_reset(plane_state, plane);
+
+       plane->base.state = &plane_state->uapi;
+
+       return plane;
+}
+
+void intel_plane_free(struct intel_plane *plane)
+{
+       intel_plane_destroy_state(&plane->base, plane->base.state);
+       kfree(plane);
+}
+
+/**
+ * intel_plane_destroy - destroy a plane
+ * @plane: plane to destroy
+ *
+ * Common destruction function for all types of planes (primary, cursor,
+ * sprite).
+ */
+void intel_plane_destroy(struct drm_plane *plane)
+{
+       drm_plane_cleanup(plane);
+       kfree(to_intel_plane(plane));
+}
+
+/**
+ * intel_plane_duplicate_state - duplicate plane state
+ * @plane: drm plane
+ *
+ * Allocates and returns a copy of the plane state (both common and
+ * Intel-specific) for the specified plane.
+ *
+ * Returns: The newly allocated plane state, or NULL on failure.
+ */
+struct drm_plane_state *
+intel_plane_duplicate_state(struct drm_plane *plane)
+{
+       struct intel_plane_state *intel_state;
+
+       intel_state = to_intel_plane_state(plane->state);
+       intel_state = kmemdup(intel_state, sizeof(*intel_state), GFP_KERNEL);
+
+       if (!intel_state)
+               return NULL;
+
+       __drm_atomic_helper_plane_duplicate_state(plane, &intel_state->uapi);
+
+       intel_state->ggtt_vma = NULL;
+       intel_state->dpt_vma = NULL;
+       intel_state->flags = 0;
+       intel_state->damage = DRM_RECT_INIT(0, 0, 0, 0);
+
+       /* add reference to fb */
+       if (intel_state->hw.fb)
+               drm_framebuffer_get(intel_state->hw.fb);
+
+       return &intel_state->uapi;
+}
+
+/**
+ * intel_plane_destroy_state - destroy plane state
+ * @plane: drm plane
+ * @state: state object to destroy
+ *
+ * Destroys the plane state (both common and Intel-specific) for the
+ * specified plane.
+ */
+void
+intel_plane_destroy_state(struct drm_plane *plane,
+                         struct drm_plane_state *state)
+{
+       struct intel_plane_state *plane_state = to_intel_plane_state(state);
+
+       drm_WARN_ON(plane->dev, plane_state->ggtt_vma);
+       drm_WARN_ON(plane->dev, plane_state->dpt_vma);
+
+       __drm_atomic_helper_plane_destroy_state(&plane_state->uapi);
+       if (plane_state->hw.fb)
+               drm_framebuffer_put(plane_state->hw.fb);
+       kfree(plane_state);
+}
+
+bool intel_plane_needs_physical(struct intel_plane *plane)
+{
+       struct intel_display *display = to_intel_display(plane);
+
+       return plane->id == PLANE_CURSOR &&
+               DISPLAY_INFO(display)->cursor_needs_physical;
+}
+
+bool intel_plane_can_async_flip(struct intel_plane *plane, u32 format,
+                               u64 modifier)
+{
+       if (intel_format_info_is_yuv_semiplanar(drm_format_info(format), modifier) ||
+           format == DRM_FORMAT_C8)
+               return false;
+
+       return plane->can_async_flip && plane->can_async_flip(modifier);
+}
+
+bool intel_plane_format_mod_supported_async(struct drm_plane *plane,
+                                           u32 format,
+                                           u64 modifier)
+{
+       if (!plane->funcs->format_mod_supported(plane, format, modifier))
+               return false;
+
+       return intel_plane_can_async_flip(to_intel_plane(plane),
+                                       format, modifier);
+}
+
+unsigned int intel_adjusted_rate(const struct drm_rect *src,
+                                const struct drm_rect *dst,
+                                unsigned int rate)
+{
+       unsigned int src_w, src_h, dst_w, dst_h;
+
+       src_w = drm_rect_width(src) >> 16;
+       src_h = drm_rect_height(src) >> 16;
+       dst_w = drm_rect_width(dst);
+       dst_h = drm_rect_height(dst);
+
+       /* Downscaling limits the maximum pixel rate */
+       dst_w = min(src_w, dst_w);
+       dst_h = min(src_h, dst_h);
+
+       return DIV_ROUND_UP_ULL(mul_u32_u32(rate, src_w * src_h),
+                               dst_w * dst_h);
+}
+
+unsigned int intel_plane_pixel_rate(const struct intel_crtc_state *crtc_state,
+                                   const struct intel_plane_state *plane_state)
+{
+       /*
+        * Note we don't check for plane visibility here as
+        * we want to use this when calculating the cursor
+        * watermarks even if the cursor is fully offscreen.
+        * That depends on the src/dst rectangles being
+        * correctly populated whenever the watermark code
+        * considers the cursor to be visible, whether or not
+        * it is actually visible.
+        *
+        * See: intel_wm_plane_visible() and intel_check_cursor()
+        */
+
+       return intel_adjusted_rate(&plane_state->uapi.src,
+                                  &plane_state->uapi.dst,
+                                  crtc_state->pixel_rate);
+}
+
+unsigned int intel_plane_data_rate(const struct intel_crtc_state *crtc_state,
+                                  const struct intel_plane_state *plane_state,
+                                  int color_plane)
+{
+       const struct drm_framebuffer *fb = plane_state->hw.fb;
+
+       if (!plane_state->uapi.visible)
+               return 0;
+
+       return intel_plane_pixel_rate(crtc_state, plane_state) *
+               fb->format->cpp[color_plane];
+}
+
+static unsigned int
+intel_plane_relative_data_rate(const struct intel_crtc_state *crtc_state,
+                              const struct intel_plane_state *plane_state,
+                              int color_plane)
+{
+       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
+       const struct drm_framebuffer *fb = plane_state->hw.fb;
+       unsigned int rel_data_rate;
+       int width, height;
+
+       if (plane->id == PLANE_CURSOR)
+               return 0;
+
+       if (!plane_state->uapi.visible)
+               return 0;
+
+       /*
+        * Src coordinates are already rotated by 270 degrees for
+        * the 90/270 degree plane rotation cases (to match the
+        * GTT mapping), hence no need to account for rotation here.
+        */
+       width = drm_rect_width(&plane_state->uapi.src) >> 16;
+       height = drm_rect_height(&plane_state->uapi.src) >> 16;
+
+       /* UV plane does 1/2 pixel sub-sampling */
+       if (color_plane == 1) {
+               width /= 2;
+               height /= 2;
+       }
+
+       rel_data_rate =
+               skl_plane_relative_data_rate(crtc_state, plane, width, height,
+                                            fb->format->cpp[color_plane]);
+       if (!rel_data_rate)
+               return 0;
+
+       return intel_adjusted_rate(&plane_state->uapi.src,
+                                  &plane_state->uapi.dst,
+                                  rel_data_rate);
+}
+
+int intel_plane_calc_min_cdclk(struct intel_atomic_state *state,
+                              struct intel_plane *plane,
+                              bool *need_cdclk_calc)
+{
+       struct intel_display *display = to_intel_display(plane);
+       const struct intel_plane_state *plane_state =
+               intel_atomic_get_new_plane_state(state, plane);
+       struct intel_crtc *crtc = to_intel_crtc(plane_state->hw.crtc);
+       const struct intel_cdclk_state *cdclk_state;
+       const struct intel_crtc_state *old_crtc_state;
+       struct intel_crtc_state *new_crtc_state;
+
+       if (!plane_state->uapi.visible || !plane->min_cdclk)
+               return 0;
+
+       old_crtc_state = intel_atomic_get_old_crtc_state(state, crtc);
+       new_crtc_state = intel_atomic_get_new_crtc_state(state, crtc);
+
+       new_crtc_state->min_cdclk[plane->id] =
+               plane->min_cdclk(new_crtc_state, plane_state);
+
+       /*
+        * No need to check against the cdclk state if
+        * the min cdclk for the plane doesn't increase.
+        *
+        * Ie. we only ever increase the cdclk due to plane
+        * requirements. This can reduce back and forth
+        * display blinking due to constant cdclk changes.
+        */
+       if (new_crtc_state->min_cdclk[plane->id] <=
+           old_crtc_state->min_cdclk[plane->id])
+               return 0;
+
+       cdclk_state = intel_atomic_get_cdclk_state(state);
+       if (IS_ERR(cdclk_state))
+               return PTR_ERR(cdclk_state);
+
+       /*
+        * No need to recalculate the cdclk state if
+        * the min cdclk for the pipe doesn't increase.
+        *
+        * Ie. we only ever increase the cdclk due to plane
+        * requirements. This can reduce back and forth
+        * display blinking due to constant cdclk changes.
+        */
+       if (new_crtc_state->min_cdclk[plane->id] <=
+           cdclk_state->min_cdclk[crtc->pipe])
+               return 0;
+
+       drm_dbg_kms(display->drm,
+                   "[PLANE:%d:%s] min cdclk (%d kHz) > [CRTC:%d:%s] min cdclk (%d kHz)\n",
+                   plane->base.base.id, plane->base.name,
+                   new_crtc_state->min_cdclk[plane->id],
+                   crtc->base.base.id, crtc->base.name,
+                   cdclk_state->min_cdclk[crtc->pipe]);
+       *need_cdclk_calc = true;
+
+       return 0;
+}
+
+static void intel_plane_clear_hw_state(struct intel_plane_state *plane_state)
+{
+       if (plane_state->hw.fb)
+               drm_framebuffer_put(plane_state->hw.fb);
+
+       memset(&plane_state->hw, 0, sizeof(plane_state->hw));
+}
+
+static void
+intel_plane_copy_uapi_plane_damage(struct intel_plane_state *new_plane_state,
+                                  const struct intel_plane_state *old_uapi_plane_state,
+                                  const struct intel_plane_state *new_uapi_plane_state)
+{
+       struct intel_display *display = to_intel_display(new_plane_state);
+       struct drm_rect *damage = &new_plane_state->damage;
+
+       /* damage property tracking enabled from display version 12 onwards */
+       if (DISPLAY_VER(display) < 12)
+               return;
+
+       if (!drm_atomic_helper_damage_merged(&old_uapi_plane_state->uapi,
+                                            &new_uapi_plane_state->uapi,
+                                            damage))
+               /* Incase helper fails, mark whole plane region as damage */
+               *damage = drm_plane_state_src(&new_uapi_plane_state->uapi);
+}
+
+void intel_plane_copy_uapi_to_hw_state(struct intel_plane_state *plane_state,
+                                      const struct intel_plane_state *from_plane_state,
+                                      struct intel_crtc *crtc)
+{
+       intel_plane_clear_hw_state(plane_state);
+
+       /*
+        * For the joiner secondary uapi.crtc will point at
+        * the primary crtc. So we explicitly assign the right
+        * secondary crtc to hw.crtc. uapi.crtc!=NULL simply
+        * indicates the plane is logically enabled on the uapi level.
+        */
+       plane_state->hw.crtc = from_plane_state->uapi.crtc ? &crtc->base : NULL;
+
+       plane_state->hw.fb = from_plane_state->uapi.fb;
+       if (plane_state->hw.fb)
+               drm_framebuffer_get(plane_state->hw.fb);
+
+       plane_state->hw.alpha = from_plane_state->uapi.alpha;
+       plane_state->hw.pixel_blend_mode =
+               from_plane_state->uapi.pixel_blend_mode;
+       plane_state->hw.rotation = from_plane_state->uapi.rotation;
+       plane_state->hw.color_encoding = from_plane_state->uapi.color_encoding;
+       plane_state->hw.color_range = from_plane_state->uapi.color_range;
+       plane_state->hw.scaling_filter = from_plane_state->uapi.scaling_filter;
+
+       plane_state->uapi.src = drm_plane_state_src(&from_plane_state->uapi);
+       plane_state->uapi.dst = drm_plane_state_dest(&from_plane_state->uapi);
+}
+
+void intel_plane_copy_hw_state(struct intel_plane_state *plane_state,
+                              const struct intel_plane_state *from_plane_state)
+{
+       intel_plane_clear_hw_state(plane_state);
+
+       memcpy(&plane_state->hw, &from_plane_state->hw,
+              sizeof(plane_state->hw));
+
+       if (plane_state->hw.fb)
+               drm_framebuffer_get(plane_state->hw.fb);
+}
+
+void intel_plane_set_invisible(struct intel_crtc_state *crtc_state,
+                              struct intel_plane_state *plane_state)
+{
+       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
+
+       crtc_state->active_planes &= ~BIT(plane->id);
+       crtc_state->scaled_planes &= ~BIT(plane->id);
+       crtc_state->nv12_planes &= ~BIT(plane->id);
+       crtc_state->c8_planes &= ~BIT(plane->id);
+       crtc_state->async_flip_planes &= ~BIT(plane->id);
+       crtc_state->data_rate[plane->id] = 0;
+       crtc_state->data_rate_y[plane->id] = 0;
+       crtc_state->rel_data_rate[plane->id] = 0;
+       crtc_state->rel_data_rate_y[plane->id] = 0;
+       crtc_state->min_cdclk[plane->id] = 0;
+
+       plane_state->uapi.visible = false;
+}
+
+static bool intel_plane_is_scaled(const struct intel_plane_state *plane_state)
+{
+       int src_w = drm_rect_width(&plane_state->uapi.src) >> 16;
+       int src_h = drm_rect_height(&plane_state->uapi.src) >> 16;
+       int dst_w = drm_rect_width(&plane_state->uapi.dst);
+       int dst_h = drm_rect_height(&plane_state->uapi.dst);
+
+       return src_w != dst_w || src_h != dst_h;
+}
+
+static bool intel_plane_do_async_flip(struct intel_plane *plane,
+                                     const struct intel_crtc_state *old_crtc_state,
+                                     const struct intel_crtc_state *new_crtc_state)
+{
+       struct intel_display *display = to_intel_display(plane);
+
+       if (!plane->async_flip)
+               return false;
+
+       if (!new_crtc_state->uapi.async_flip)
+               return false;
+
+       /*
+        * In platforms after DISPLAY13, we might need to override
+        * first async flip in order to change watermark levels
+        * as part of optimization.
+        *
+        * And let's do this for all skl+ so that we can eg. change the
+        * modifier as well.
+        *
+        * TODO: For older platforms there is less reason to do this as
+        * only X-tile is supported with async flips, though we could
+        * extend this so other scanout parameters (stride/etc) could
+        * be changed as well...
+        */
+       return DISPLAY_VER(display) < 9 || old_crtc_state->uapi.async_flip;
+}
+
+static bool i9xx_must_disable_cxsr(const struct intel_crtc_state *new_crtc_state,
+                                  const struct intel_plane_state *old_plane_state,
+                                  const struct intel_plane_state *new_plane_state)
+{
+       struct intel_plane *plane = to_intel_plane(new_plane_state->uapi.plane);
+       bool old_visible = old_plane_state->uapi.visible;
+       bool new_visible = new_plane_state->uapi.visible;
+       u32 old_ctl = old_plane_state->ctl;
+       u32 new_ctl = new_plane_state->ctl;
+       bool modeset, turn_on, turn_off;
+
+       if (plane->id == PLANE_CURSOR)
+               return false;
+
+       modeset = intel_crtc_needs_modeset(new_crtc_state);
+       turn_off = old_visible && (!new_visible || modeset);
+       turn_on = new_visible && (!old_visible || modeset);
+
+       /* Must disable CxSR around plane enable/disable */
+       if (turn_on || turn_off)
+               return true;
+
+       if (!old_visible || !new_visible)
+               return false;
+
+       /*
+        * Most plane control register updates are blocked while in CxSR.
+        *
+        * Tiling mode is one exception where the primary plane can
+        * apparently handle it, whereas the sprites can not (the
+        * sprite issue being only relevant on VLV/CHV where CxSR
+        * is actually possible with a sprite enabled).
+        */
+       if (plane->id == PLANE_PRIMARY) {
+               old_ctl &= ~DISP_TILED;
+               new_ctl &= ~DISP_TILED;
+       }
+
+       return old_ctl != new_ctl;
+}
+
+static bool ilk_must_disable_cxsr(const struct intel_crtc_state *new_crtc_state,
+                                 const struct intel_plane_state *old_plane_state,
+                                 const struct intel_plane_state *new_plane_state)
+{
+       struct intel_plane *plane = to_intel_plane(new_plane_state->uapi.plane);
+       bool old_visible = old_plane_state->uapi.visible;
+       bool new_visible = new_plane_state->uapi.visible;
+       bool modeset, turn_on;
+
+       if (plane->id == PLANE_CURSOR)
+               return false;
+
+       modeset = intel_crtc_needs_modeset(new_crtc_state);
+       turn_on = new_visible && (!old_visible || modeset);
+
+       /*
+        * ILK/SNB DVSACNTR/Sprite Enable
+        * IVB SPR_CTL/Sprite Enable
+        * "When in Self Refresh Big FIFO mode, a write to enable the
+        *  plane will be internally buffered and delayed while Big FIFO
+        *  mode is exiting."
+        *
+        * Which means that enabling the sprite can take an extra frame
+        * when we start in big FIFO mode (LP1+). Thus we need to drop
+        * down to LP0 and wait for vblank in order to make sure the
+        * sprite gets enabled on the next vblank after the register write.
+        * Doing otherwise would risk enabling the sprite one frame after
+        * we've already signalled flip completion. We can resume LP1+
+        * once the sprite has been enabled.
+        *
+        * With experimental results seems this is needed also for primary
+        * plane, not only sprite plane.
+        */
+       if (turn_on)
+               return true;
+
+       /*
+        * WaCxSRDisabledForSpriteScaling:ivb
+        * IVB SPR_SCALE/Scaling Enable
+        * "Low Power watermarks must be disabled for at least one
+        *  frame before enabling sprite scaling, and kept disabled
+        *  until sprite scaling is disabled."
+        *
+        * ILK/SNB DVSASCALE/Scaling Enable
+        * "When in Self Refresh Big FIFO mode, scaling enable will be
+        *  masked off while Big FIFO mode is exiting."
+        *
+        * Despite the w/a only being listed for IVB we assume that
+        * the ILK/SNB note has similar ramifications, hence we apply
+        * the w/a on all three platforms.
+        */
+       return !intel_plane_is_scaled(old_plane_state) &&
+               intel_plane_is_scaled(new_plane_state);
+}
+
+static int intel_plane_atomic_calc_changes(const struct intel_crtc_state *old_crtc_state,
+                                          struct intel_crtc_state *new_crtc_state,
+                                          const struct intel_plane_state *old_plane_state,
+                                          struct intel_plane_state *new_plane_state)
+{
+       struct intel_display *display = to_intel_display(new_crtc_state);
+       struct intel_crtc *crtc = to_intel_crtc(new_crtc_state->uapi.crtc);
+       struct intel_plane *plane = to_intel_plane(new_plane_state->uapi.plane);
+       bool mode_changed = intel_crtc_needs_modeset(new_crtc_state);
+       bool was_crtc_enabled = old_crtc_state->hw.active;
+       bool is_crtc_enabled = new_crtc_state->hw.active;
+       bool turn_off, turn_on, visible, was_visible;
+       int ret;
+
+       if (DISPLAY_VER(display) >= 9 && plane->id != PLANE_CURSOR) {
+               ret = skl_update_scaler_plane(new_crtc_state, new_plane_state);
+               if (ret)
+                       return ret;
+       }
+
+       was_visible = old_plane_state->uapi.visible;
+       visible = new_plane_state->uapi.visible;
+
+       if (!was_crtc_enabled && drm_WARN_ON(display->drm, was_visible))
+               was_visible = false;
+
+       /*
+        * Visibility is calculated as if the crtc was on, but
+        * after scaler setup everything depends on it being off
+        * when the crtc isn't active.
+        *
+        * FIXME this is wrong for watermarks. Watermarks should also
+        * be computed as if the pipe would be active. Perhaps move
+        * per-plane wm computation to the .check_plane() hook, and
+        * only combine the results from all planes in the current place?
+        */
+       if (!is_crtc_enabled) {
+               intel_plane_set_invisible(new_crtc_state, new_plane_state);
+               visible = false;
+       }
+
+       if (!was_visible && !visible)
+               return 0;
+
+       turn_off = was_visible && (!visible || mode_changed);
+       turn_on = visible && (!was_visible || mode_changed);
+
+       drm_dbg_atomic(display->drm,
+                      "[CRTC:%d:%s] with [PLANE:%d:%s] visible %i -> %i, off %i, on %i, ms %i\n",
+                      crtc->base.base.id, crtc->base.name,
+                      plane->base.base.id, plane->base.name,
+                      was_visible, visible,
+                      turn_off, turn_on, mode_changed);
+
+       if (visible || was_visible)
+               new_crtc_state->fb_bits |= plane->frontbuffer_bit;
+
+       if (HAS_GMCH(display) &&
+           i9xx_must_disable_cxsr(new_crtc_state, old_plane_state, new_plane_state))
+               new_crtc_state->disable_cxsr = true;
+
+       if ((display->platform.ironlake || display->platform.sandybridge || display->platform.ivybridge) &&
+           ilk_must_disable_cxsr(new_crtc_state, old_plane_state, new_plane_state))
+               new_crtc_state->disable_cxsr = true;
+
+       if (intel_plane_do_async_flip(plane, old_crtc_state, new_crtc_state)) {
+               new_crtc_state->do_async_flip = true;
+               new_crtc_state->async_flip_planes |= BIT(plane->id);
+       } else if (plane->need_async_flip_toggle_wa &&
+                  new_crtc_state->uapi.async_flip) {
+               /*
+                * On platforms with double buffered async flip bit we
+                * set the bit already one frame early during the sync
+                * flip (see {i9xx,skl}_plane_update_arm()). The
+                * hardware will therefore be ready to perform a real
+                * async flip during the next commit, without having
+                * to wait yet another frame for the bit to latch.
+                */
+               new_crtc_state->async_flip_planes |= BIT(plane->id);
+       }
+
+       return 0;
+}
+
+int intel_plane_atomic_check_with_state(const struct intel_crtc_state *old_crtc_state,
+                                       struct intel_crtc_state *new_crtc_state,
+                                       const struct intel_plane_state *old_plane_state,
+                                       struct intel_plane_state *new_plane_state)
+{
+       struct intel_plane *plane = to_intel_plane(new_plane_state->uapi.plane);
+       const struct drm_framebuffer *fb = new_plane_state->hw.fb;
+       int ret;
+
+       intel_plane_set_invisible(new_crtc_state, new_plane_state);
+       new_crtc_state->enabled_planes &= ~BIT(plane->id);
+
+       if (!new_plane_state->hw.crtc && !old_plane_state->hw.crtc)
+               return 0;
+
+       ret = plane->check_plane(new_crtc_state, new_plane_state);
+       if (ret)
+               return ret;
+
+       if (fb)
+               new_crtc_state->enabled_planes |= BIT(plane->id);
+
+       /* FIXME pre-g4x don't work like this */
+       if (new_plane_state->uapi.visible)
+               new_crtc_state->active_planes |= BIT(plane->id);
+
+       if (new_plane_state->uapi.visible &&
+           intel_plane_is_scaled(new_plane_state))
+               new_crtc_state->scaled_planes |= BIT(plane->id);
+
+       if (new_plane_state->uapi.visible &&
+           intel_format_info_is_yuv_semiplanar(fb->format, fb->modifier))
+               new_crtc_state->nv12_planes |= BIT(plane->id);
+
+       if (new_plane_state->uapi.visible &&
+           fb->format->format == DRM_FORMAT_C8)
+               new_crtc_state->c8_planes |= BIT(plane->id);
+
+       if (new_plane_state->uapi.visible || old_plane_state->uapi.visible)
+               new_crtc_state->update_planes |= BIT(plane->id);
+
+       if (new_plane_state->uapi.visible &&
+           intel_format_info_is_yuv_semiplanar(fb->format, fb->modifier)) {
+               new_crtc_state->data_rate_y[plane->id] =
+                       intel_plane_data_rate(new_crtc_state, new_plane_state, 0);
+               new_crtc_state->data_rate[plane->id] =
+                       intel_plane_data_rate(new_crtc_state, new_plane_state, 1);
+
+               new_crtc_state->rel_data_rate_y[plane->id] =
+                       intel_plane_relative_data_rate(new_crtc_state,
+                                                      new_plane_state, 0);
+               new_crtc_state->rel_data_rate[plane->id] =
+                       intel_plane_relative_data_rate(new_crtc_state,
+                                                      new_plane_state, 1);
+       } else if (new_plane_state->uapi.visible) {
+               new_crtc_state->data_rate[plane->id] =
+                       intel_plane_data_rate(new_crtc_state, new_plane_state, 0);
+
+               new_crtc_state->rel_data_rate[plane->id] =
+                       intel_plane_relative_data_rate(new_crtc_state,
+                                                      new_plane_state, 0);
+       }
+
+       return intel_plane_atomic_calc_changes(old_crtc_state, new_crtc_state,
+                                              old_plane_state, new_plane_state);
+}
+
+struct intel_plane *
+intel_crtc_get_plane(struct intel_crtc *crtc, enum plane_id plane_id)
+{
+       struct intel_display *display = to_intel_display(crtc);
+       struct intel_plane *plane;
+
+       for_each_intel_plane_on_crtc(display->drm, crtc, plane) {
+               if (plane->id == plane_id)
+                       return plane;
+       }
+
+       return NULL;
+}
+
+int intel_plane_atomic_check(struct intel_atomic_state *state,
+                            struct intel_plane *plane)
+{
+       struct intel_display *display = to_intel_display(state);
+       struct intel_plane_state *new_plane_state =
+               intel_atomic_get_new_plane_state(state, plane);
+       const struct intel_plane_state *old_plane_state =
+               intel_atomic_get_old_plane_state(state, plane);
+       const struct intel_plane_state *new_primary_crtc_plane_state;
+       const struct intel_plane_state *old_primary_crtc_plane_state;
+       struct intel_crtc *crtc = intel_crtc_for_pipe(display, plane->pipe);
+       const struct intel_crtc_state *old_crtc_state =
+               intel_atomic_get_old_crtc_state(state, crtc);
+       struct intel_crtc_state *new_crtc_state =
+               intel_atomic_get_new_crtc_state(state, crtc);
+
+       if (new_crtc_state && intel_crtc_is_joiner_secondary(new_crtc_state)) {
+               struct intel_crtc *primary_crtc =
+                       intel_primary_crtc(new_crtc_state);
+               struct intel_plane *primary_crtc_plane =
+                       intel_crtc_get_plane(primary_crtc, plane->id);
+
+               new_primary_crtc_plane_state =
+                       intel_atomic_get_new_plane_state(state, primary_crtc_plane);
+               old_primary_crtc_plane_state =
+                       intel_atomic_get_old_plane_state(state, primary_crtc_plane);
+       } else {
+               new_primary_crtc_plane_state = new_plane_state;
+               old_primary_crtc_plane_state = old_plane_state;
+       }
+
+       intel_plane_copy_uapi_plane_damage(new_plane_state,
+                                          old_primary_crtc_plane_state,
+                                          new_primary_crtc_plane_state);
+
+       intel_plane_copy_uapi_to_hw_state(new_plane_state,
+                                         new_primary_crtc_plane_state,
+                                         crtc);
+
+       new_plane_state->uapi.visible = false;
+       if (!new_crtc_state)
+               return 0;
+
+       return intel_plane_atomic_check_with_state(old_crtc_state,
+                                                  new_crtc_state,
+                                                  old_plane_state,
+                                                  new_plane_state);
+}
+
+static struct intel_plane *
+skl_next_plane_to_commit(struct intel_atomic_state *state,
+                        struct intel_crtc *crtc,
+                        struct skl_ddb_entry ddb[I915_MAX_PLANES],
+                        struct skl_ddb_entry ddb_y[I915_MAX_PLANES],
+                        unsigned int *update_mask)
+{
+       struct intel_crtc_state *crtc_state =
+               intel_atomic_get_new_crtc_state(state, crtc);
+       struct intel_plane_state __maybe_unused *plane_state;
+       struct intel_plane *plane;
+       int i;
+
+       if (*update_mask == 0)
+               return NULL;
+
+       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
+               enum plane_id plane_id = plane->id;
+
+               if (crtc->pipe != plane->pipe ||
+                   !(*update_mask & BIT(plane_id)))
+                       continue;
+
+               if (skl_ddb_allocation_overlaps(&crtc_state->wm.skl.plane_ddb[plane_id],
+                                               ddb, I915_MAX_PLANES, plane_id) ||
+                   skl_ddb_allocation_overlaps(&crtc_state->wm.skl.plane_ddb_y[plane_id],
+                                               ddb_y, I915_MAX_PLANES, plane_id))
+                       continue;
+
+               *update_mask &= ~BIT(plane_id);
+               ddb[plane_id] = crtc_state->wm.skl.plane_ddb[plane_id];
+               ddb_y[plane_id] = crtc_state->wm.skl.plane_ddb_y[plane_id];
+
+               return plane;
+       }
+
+       /* should never happen */
+       drm_WARN_ON(state->base.dev, 1);
+
+       return NULL;
+}
+
+void intel_plane_update_noarm(struct intel_dsb *dsb,
+                             struct intel_plane *plane,
+                             const struct intel_crtc_state *crtc_state,
+                             const struct intel_plane_state *plane_state)
+{
+       struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
+
+       trace_intel_plane_update_noarm(plane_state, crtc);
+
+       if (plane->fbc)
+               intel_fbc_dirty_rect_update_noarm(dsb, plane);
+
+       if (plane->update_noarm)
+               plane->update_noarm(dsb, plane, crtc_state, plane_state);
+}
+
+void intel_plane_async_flip(struct intel_dsb *dsb,
+                           struct intel_plane *plane,
+                           const struct intel_crtc_state *crtc_state,
+                           const struct intel_plane_state *plane_state,
+                           bool async_flip)
+{
+       struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
+
+       trace_intel_plane_async_flip(plane, crtc, async_flip);
+       plane->async_flip(dsb, plane, crtc_state, plane_state, async_flip);
+}
+
+void intel_plane_update_arm(struct intel_dsb *dsb,
+                           struct intel_plane *plane,
+                           const struct intel_crtc_state *crtc_state,
+                           const struct intel_plane_state *plane_state)
+{
+       struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
+
+       if (crtc_state->do_async_flip && plane->async_flip) {
+               intel_plane_async_flip(dsb, plane, crtc_state, plane_state, true);
+               return;
+       }
+
+       trace_intel_plane_update_arm(plane_state, crtc);
+       plane->update_arm(dsb, plane, crtc_state, plane_state);
+}
+
+void intel_plane_disable_arm(struct intel_dsb *dsb,
+                            struct intel_plane *plane,
+                            const struct intel_crtc_state *crtc_state)
+{
+       struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
+
+       trace_intel_plane_disable_arm(plane, crtc);
+       plane->disable_arm(dsb, plane, crtc_state);
+}
+
+void intel_crtc_planes_update_noarm(struct intel_dsb *dsb,
+                                   struct intel_atomic_state *state,
+                                   struct intel_crtc *crtc)
+{
+       struct intel_crtc_state *new_crtc_state =
+               intel_atomic_get_new_crtc_state(state, crtc);
+       u32 update_mask = new_crtc_state->update_planes;
+       struct intel_plane_state *new_plane_state;
+       struct intel_plane *plane;
+       int i;
+
+       if (new_crtc_state->do_async_flip)
+               return;
+
+       /*
+        * Since we only write non-arming registers here,
+        * the order does not matter even for skl+.
+        */
+       for_each_new_intel_plane_in_state(state, plane, new_plane_state, i) {
+               if (crtc->pipe != plane->pipe ||
+                   !(update_mask & BIT(plane->id)))
+                       continue;
+
+               /* TODO: for mailbox updates this should be skipped */
+               if (new_plane_state->uapi.visible ||
+                   new_plane_state->is_y_plane)
+                       intel_plane_update_noarm(dsb, plane,
+                                                new_crtc_state, new_plane_state);
+       }
+}
+
+static void skl_crtc_planes_update_arm(struct intel_dsb *dsb,
+                                      struct intel_atomic_state *state,
+                                      struct intel_crtc *crtc)
+{
+       struct intel_crtc_state *old_crtc_state =
+               intel_atomic_get_old_crtc_state(state, crtc);
+       struct intel_crtc_state *new_crtc_state =
+               intel_atomic_get_new_crtc_state(state, crtc);
+       struct skl_ddb_entry ddb[I915_MAX_PLANES];
+       struct skl_ddb_entry ddb_y[I915_MAX_PLANES];
+       u32 update_mask = new_crtc_state->update_planes;
+       struct intel_plane *plane;
+
+       memcpy(ddb, old_crtc_state->wm.skl.plane_ddb,
+              sizeof(old_crtc_state->wm.skl.plane_ddb));
+       memcpy(ddb_y, old_crtc_state->wm.skl.plane_ddb_y,
+              sizeof(old_crtc_state->wm.skl.plane_ddb_y));
+
+       while ((plane = skl_next_plane_to_commit(state, crtc, ddb, ddb_y, &update_mask))) {
+               struct intel_plane_state *new_plane_state =
+                       intel_atomic_get_new_plane_state(state, plane);
+
+               /*
+                * TODO: for mailbox updates intel_plane_update_noarm()
+                * would have to be called here as well.
+                */
+               if (new_plane_state->uapi.visible ||
+                   new_plane_state->is_y_plane)
+                       intel_plane_update_arm(dsb, plane, new_crtc_state, new_plane_state);
+               else
+                       intel_plane_disable_arm(dsb, plane, new_crtc_state);
+       }
+}
+
+static void i9xx_crtc_planes_update_arm(struct intel_dsb *dsb,
+                                       struct intel_atomic_state *state,
+                                       struct intel_crtc *crtc)
+{
+       struct intel_crtc_state *new_crtc_state =
+               intel_atomic_get_new_crtc_state(state, crtc);
+       u32 update_mask = new_crtc_state->update_planes;
+       struct intel_plane_state *new_plane_state;
+       struct intel_plane *plane;
+       int i;
+
+       for_each_new_intel_plane_in_state(state, plane, new_plane_state, i) {
+               if (crtc->pipe != plane->pipe ||
+                   !(update_mask & BIT(plane->id)))
+                       continue;
+
+               /*
+                * TODO: for mailbox updates intel_plane_update_noarm()
+                * would have to be called here as well.
+                */
+               if (new_plane_state->uapi.visible)
+                       intel_plane_update_arm(dsb, plane, new_crtc_state, new_plane_state);
+               else
+                       intel_plane_disable_arm(dsb, plane, new_crtc_state);
+       }
+}
+
+void intel_crtc_planes_update_arm(struct intel_dsb *dsb,
+                                 struct intel_atomic_state *state,
+                                 struct intel_crtc *crtc)
+{
+       struct intel_display *display = to_intel_display(state);
+
+       if (DISPLAY_VER(display) >= 9)
+               skl_crtc_planes_update_arm(dsb, state, crtc);
+       else
+               i9xx_crtc_planes_update_arm(dsb, state, crtc);
+}
+
+int intel_atomic_plane_check_clipping(struct intel_plane_state *plane_state,
+                                     struct intel_crtc_state *crtc_state,
+                                     int min_scale, int max_scale,
+                                     bool can_position)
+{
+       struct intel_display *display = to_intel_display(plane_state);
+       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
+       struct drm_framebuffer *fb = plane_state->hw.fb;
+       struct drm_rect *src = &plane_state->uapi.src;
+       struct drm_rect *dst = &plane_state->uapi.dst;
+       const struct drm_rect *clip = &crtc_state->pipe_src;
+       unsigned int rotation = plane_state->hw.rotation;
+       int hscale, vscale;
+
+       if (!fb) {
+               plane_state->uapi.visible = false;
+               return 0;
+       }
+
+       drm_rect_rotate(src, fb->width << 16, fb->height << 16, rotation);
+
+       /* Check scaling */
+       hscale = drm_rect_calc_hscale(src, dst, min_scale, max_scale);
+       vscale = drm_rect_calc_vscale(src, dst, min_scale, max_scale);
+       if (hscale < 0 || vscale < 0) {
+               drm_dbg_kms(display->drm,
+                           "[PLANE:%d:%s] invalid scaling "DRM_RECT_FP_FMT " -> " DRM_RECT_FMT "\n",
+                           plane->base.base.id, plane->base.name,
+                           DRM_RECT_FP_ARG(src), DRM_RECT_ARG(dst));
+               return -ERANGE;
+       }
+
+       /*
+        * FIXME: This might need further adjustment for seamless scaling
+        * with phase information, for the 2p2 and 2p1 scenarios.
+        */
+       plane_state->uapi.visible = drm_rect_clip_scaled(src, dst, clip);
+
+       drm_rect_rotate_inv(src, fb->width << 16, fb->height << 16, rotation);
+
+       if (!can_position && plane_state->uapi.visible &&
+           !drm_rect_equals(dst, clip)) {
+               drm_dbg_kms(display->drm,
+                           "[PLANE:%d:%s] plane (" DRM_RECT_FMT ") must cover entire CRTC (" DRM_RECT_FMT ")\n",
+                           plane->base.base.id, plane->base.name,
+                           DRM_RECT_ARG(dst), DRM_RECT_ARG(clip));
+               return -EINVAL;
+       }
+
+       /* final plane coordinates will be relative to the plane's pipe */
+       drm_rect_translate(dst, -clip->x1, -clip->y1);
+
+       return 0;
+}
+
+int intel_plane_check_src_coordinates(struct intel_plane_state *plane_state)
+{
+       struct intel_display *display = to_intel_display(plane_state);
+       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
+       const struct drm_framebuffer *fb = plane_state->hw.fb;
+       struct drm_rect *src = &plane_state->uapi.src;
+       u32 src_x, src_y, src_w, src_h, hsub, vsub;
+       bool rotated = drm_rotation_90_or_270(plane_state->hw.rotation);
+
+       /*
+        * FIXME hsub/vsub vs. block size is a mess. Pre-tgl CCS
+        * abuses hsub/vsub so we can't use them here. But as they
+        * are limited to 32bpp RGB formats we don't actually need
+        * to check anything.
+        */
+       if (fb->modifier == I915_FORMAT_MOD_Y_TILED_CCS ||
+           fb->modifier == I915_FORMAT_MOD_Yf_TILED_CCS)
+               return 0;
+
+       /*
+        * Hardware doesn't handle subpixel coordinates.
+        * Adjust to (macro)pixel boundary, but be careful not to
+        * increase the source viewport size, because that could
+        * push the downscaling factor out of bounds.
+        */
+       src_x = src->x1 >> 16;
+       src_w = drm_rect_width(src) >> 16;
+       src_y = src->y1 >> 16;
+       src_h = drm_rect_height(src) >> 16;
+
+       drm_rect_init(src, src_x << 16, src_y << 16,
+                     src_w << 16, src_h << 16);
+
+       if (fb->format->format == DRM_FORMAT_RGB565 && rotated) {
+               hsub = 2;
+               vsub = 2;
+       } else if (DISPLAY_VER(display) >= 20 &&
+                  intel_format_info_is_yuv_semiplanar(fb->format, fb->modifier)) {
+               /*
+                * This allows NV12 and P0xx formats to have odd size and/or odd
+                * source coordinates on DISPLAY_VER(display) >= 20
+                */
+               hsub = 1;
+               vsub = 1;
+
+               /* Wa_16023981245 */
+               if ((DISPLAY_VERx100(display) == 2000 ||
+                    DISPLAY_VERx100(display) == 3000 ||
+                    DISPLAY_VERx100(display) == 3002) &&
+                    src_x % 2 != 0)
+                       hsub = 2;
+       } else {
+               hsub = fb->format->hsub;
+               vsub = fb->format->vsub;
+       }
+
+       if (rotated)
+               hsub = vsub = max(hsub, vsub);
+
+       if (src_x % hsub || src_w % hsub) {
+               drm_dbg_kms(display->drm,
+                           "[PLANE:%d:%s] src x/w (%u, %u) must be a multiple of %u (rotated: %s)\n",
+                           plane->base.base.id, plane->base.name,
+                           src_x, src_w, hsub, str_yes_no(rotated));
+               return -EINVAL;
+       }
+
+       if (src_y % vsub || src_h % vsub) {
+               drm_dbg_kms(display->drm,
+                           "[PLANE:%d:%s] src y/h (%u, %u) must be a multiple of %u (rotated: %s)\n",
+                           plane->base.base.id, plane->base.name,
+                           src_y, src_h, vsub, str_yes_no(rotated));
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+static int add_dma_resv_fences(struct dma_resv *resv,
+                              struct drm_plane_state *new_plane_state)
+{
+       struct dma_fence *fence = dma_fence_get(new_plane_state->fence);
+       struct dma_fence *new;
+       int ret;
+
+       ret = dma_resv_get_singleton(resv, dma_resv_usage_rw(false), &new);
+       if (ret)
+               goto error;
+
+       if (new && fence) {
+               struct dma_fence_chain *chain = dma_fence_chain_alloc();
+
+               if (!chain) {
+                       ret = -ENOMEM;
+                       goto error;
+               }
+
+               dma_fence_chain_init(chain, fence, new, 1);
+               fence = &chain->base;
+
+       } else if (new) {
+               fence = new;
+       }
+
+       dma_fence_put(new_plane_state->fence);
+       new_plane_state->fence = fence;
+       return 0;
+
+error:
+       dma_fence_put(fence);
+       return ret;
+}
+
+/**
+ * intel_prepare_plane_fb - Prepare fb for usage on plane
+ * @_plane: drm plane to prepare for
+ * @_new_plane_state: the plane state being prepared
+ *
+ * Prepares a framebuffer for usage on a display plane.  Generally this
+ * involves pinning the underlying object and updating the frontbuffer tracking
+ * bits.  Some older platforms need special physical address handling for
+ * cursor planes.
+ *
+ * Returns 0 on success, negative error code on failure.
+ */
+static int
+intel_prepare_plane_fb(struct drm_plane *_plane,
+                      struct drm_plane_state *_new_plane_state)
+{
+       struct i915_sched_attr attr = { .priority = I915_PRIORITY_DISPLAY };
+       struct intel_plane *plane = to_intel_plane(_plane);
+       struct intel_display *display = to_intel_display(plane);
+       struct intel_plane_state *new_plane_state =
+               to_intel_plane_state(_new_plane_state);
+       struct intel_atomic_state *state =
+               to_intel_atomic_state(new_plane_state->uapi.state);
+       struct intel_plane_state *old_plane_state =
+               intel_atomic_get_old_plane_state(state, plane);
+       struct drm_gem_object *obj = intel_fb_bo(new_plane_state->hw.fb);
+       struct drm_gem_object *old_obj = intel_fb_bo(old_plane_state->hw.fb);
+       int ret;
+
+       if (old_obj) {
+               const struct intel_crtc_state *new_crtc_state =
+                       intel_atomic_get_new_crtc_state(state,
+                                                       to_intel_crtc(old_plane_state->hw.crtc));
+
+               /* Big Hammer, we also need to ensure that any pending
+                * MI_WAIT_FOR_EVENT inside a user batch buffer on the
+                * current scanout is retired before unpinning the old
+                * framebuffer. Note that we rely on userspace rendering
+                * into the buffer attached to the pipe they are waiting
+                * on. If not, userspace generates a GPU hang with IPEHR
+                * point to the MI_WAIT_FOR_EVENT.
+                *
+                * This should only fail upon a hung GPU, in which case we
+                * can safely continue.
+                */
+               if (intel_crtc_needs_modeset(new_crtc_state)) {
+                       ret = add_dma_resv_fences(old_obj->resv,
+                                                 &new_plane_state->uapi);
+                       if (ret < 0)
+                               return ret;
+               }
+       }
+
+       if (!obj)
+               return 0;
+
+       ret = intel_plane_pin_fb(new_plane_state, old_plane_state);
+       if (ret)
+               return ret;
+
+       ret = drm_gem_plane_helper_prepare_fb(&plane->base, &new_plane_state->uapi);
+       if (ret < 0)
+               goto unpin_fb;
+
+       if (new_plane_state->uapi.fence) {
+               i915_gem_fence_wait_priority(new_plane_state->uapi.fence,
+                                            &attr);
+
+               intel_display_rps_boost_after_vblank(new_plane_state->hw.crtc,
+                                                    new_plane_state->uapi.fence);
+       }
+
+       /*
+        * We declare pageflips to be interactive and so merit a small bias
+        * towards upclocking to deliver the frame on time. By only changing
+        * the RPS thresholds to sample more regularly and aim for higher
+        * clocks we can hopefully deliver low power workloads (like kodi)
+        * that are not quite steady state without resorting to forcing
+        * maximum clocks following a vblank miss (see do_rps_boost()).
+        */
+       intel_display_rps_mark_interactive(display, state, true);
+
+       return 0;
+
+unpin_fb:
+       intel_plane_unpin_fb(new_plane_state);
+
+       return ret;
+}
+
+/**
+ * intel_cleanup_plane_fb - Cleans up an fb after plane use
+ * @plane: drm plane to clean up for
+ * @_old_plane_state: the state from the previous modeset
+ *
+ * Cleans up a framebuffer that has just been removed from a plane.
+ */
+static void
+intel_cleanup_plane_fb(struct drm_plane *plane,
+                      struct drm_plane_state *_old_plane_state)
+{
+       struct intel_display *display = to_intel_display(plane->dev);
+       struct intel_plane_state *old_plane_state =
+               to_intel_plane_state(_old_plane_state);
+       struct intel_atomic_state *state =
+               to_intel_atomic_state(old_plane_state->uapi.state);
+       struct drm_gem_object *obj = intel_fb_bo(old_plane_state->hw.fb);
+
+       if (!obj)
+               return;
+
+       intel_display_rps_mark_interactive(display, state, false);
+
+       intel_plane_unpin_fb(old_plane_state);
+}
+
+static const struct drm_plane_helper_funcs intel_plane_helper_funcs = {
+       .prepare_fb = intel_prepare_plane_fb,
+       .cleanup_fb = intel_cleanup_plane_fb,
+};
+
+void intel_plane_helper_add(struct intel_plane *plane)
+{
+       drm_plane_helper_add(&plane->base, &intel_plane_helper_funcs);
+}
+
+void intel_plane_init_cursor_vblank_work(struct intel_plane_state *old_plane_state,
+                                        struct intel_plane_state *new_plane_state)
+{
+       if (!old_plane_state->ggtt_vma ||
+           old_plane_state->ggtt_vma == new_plane_state->ggtt_vma)
+               return;
+
+       drm_vblank_work_init(&old_plane_state->unpin_work, old_plane_state->uapi.crtc,
+                            intel_cursor_unpin_work);
+}
+
+static void link_nv12_planes(struct intel_crtc_state *crtc_state,
+                            struct intel_plane_state *uv_plane_state,
+                            struct intel_plane_state *y_plane_state)
+{
+       struct intel_display *display = to_intel_display(uv_plane_state);
+       struct intel_plane *uv_plane = to_intel_plane(uv_plane_state->uapi.plane);
+       struct intel_plane *y_plane = to_intel_plane(y_plane_state->uapi.plane);
+
+       drm_dbg_kms(display->drm, "UV plane [PLANE:%d:%s] using Y plane [PLANE:%d:%s]\n",
+                   uv_plane->base.base.id, uv_plane->base.name,
+                   y_plane->base.base.id, y_plane->base.name);
+
+       uv_plane_state->planar_linked_plane = y_plane;
+
+       y_plane_state->is_y_plane = true;
+       y_plane_state->planar_linked_plane = uv_plane;
+
+       crtc_state->enabled_planes |= BIT(y_plane->id);
+       crtc_state->active_planes |= BIT(y_plane->id);
+       crtc_state->update_planes |= BIT(y_plane->id);
+
+       crtc_state->data_rate[y_plane->id] = crtc_state->data_rate_y[uv_plane->id];
+       crtc_state->rel_data_rate[y_plane->id] = crtc_state->rel_data_rate_y[uv_plane->id];
+
+       /* Copy parameters to Y plane */
+       intel_plane_copy_hw_state(y_plane_state, uv_plane_state);
+       y_plane_state->uapi.src = uv_plane_state->uapi.src;
+       y_plane_state->uapi.dst = uv_plane_state->uapi.dst;
+
+       y_plane_state->ctl = uv_plane_state->ctl;
+       y_plane_state->color_ctl = uv_plane_state->color_ctl;
+       y_plane_state->view = uv_plane_state->view;
+       y_plane_state->decrypt = uv_plane_state->decrypt;
+
+       icl_link_nv12_planes(uv_plane_state, y_plane_state);
+}
+
+static void unlink_nv12_plane(struct intel_crtc_state *crtc_state,
+                             struct intel_plane_state *plane_state)
+{
+       struct intel_display *display = to_intel_display(plane_state);
+       struct intel_plane *plane = to_intel_plane(plane_state->uapi.plane);
+
+       plane_state->planar_linked_plane = NULL;
+
+       if (!plane_state->is_y_plane)
+               return;
+
+       drm_WARN_ON(display->drm, plane_state->uapi.visible);
+
+       plane_state->is_y_plane = false;
+
+       crtc_state->enabled_planes &= ~BIT(plane->id);
+       crtc_state->active_planes &= ~BIT(plane->id);
+       crtc_state->update_planes |= BIT(plane->id);
+       crtc_state->data_rate[plane->id] = 0;
+       crtc_state->rel_data_rate[plane->id] = 0;
+}
+
+static int icl_check_nv12_planes(struct intel_atomic_state *state,
+                                struct intel_crtc *crtc)
+{
+       struct intel_display *display = to_intel_display(state);
+       struct intel_crtc_state *crtc_state =
+               intel_atomic_get_new_crtc_state(state, crtc);
+       struct intel_plane_state *plane_state;
+       struct intel_plane *plane;
+       int i;
+
+       if (DISPLAY_VER(display) < 11)
+               return 0;
+
+       /*
+        * Destroy all old plane links and make the Y plane invisible
+        * in the crtc_state->active_planes mask.
+        */
+       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
+               if (plane->pipe != crtc->pipe)
+                       continue;
+
+               if (plane_state->planar_linked_plane)
+                       unlink_nv12_plane(crtc_state, plane_state);
+       }
+
+       if (!crtc_state->nv12_planes)
+               return 0;
+
+       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
+               struct intel_plane_state *y_plane_state = NULL;
+               struct intel_plane *y_plane;
+
+               if (plane->pipe != crtc->pipe)
+                       continue;
+
+               if ((crtc_state->nv12_planes & BIT(plane->id)) == 0)
+                       continue;
+
+               for_each_intel_plane_on_crtc(display->drm, crtc, y_plane) {
+                       if (!icl_is_nv12_y_plane(display, y_plane->id))
+                               continue;
+
+                       if (crtc_state->active_planes & BIT(y_plane->id))
+                               continue;
+
+                       y_plane_state = intel_atomic_get_plane_state(state, y_plane);
+                       if (IS_ERR(y_plane_state))
+                               return PTR_ERR(y_plane_state);
+
+                       break;
+               }
+
+               if (!y_plane_state) {
+                       drm_dbg_kms(display->drm,
+                                   "[CRTC:%d:%s] need %d free Y planes for planar YUV\n",
+                                   crtc->base.base.id, crtc->base.name,
+                                   hweight8(crtc_state->nv12_planes));
+                       return -EINVAL;
+               }
+
+               link_nv12_planes(crtc_state, plane_state, y_plane_state);
+       }
+
+       return 0;
+}
+
+static int intel_crtc_add_planes_to_state(struct intel_atomic_state *state,
+                                         struct intel_crtc *crtc,
+                                         u8 plane_ids_mask)
+{
+       struct intel_display *display = to_intel_display(state);
+       struct intel_plane *plane;
+
+       for_each_intel_plane_on_crtc(display->drm, crtc, plane) {
+               struct intel_plane_state *plane_state;
+
+               if ((plane_ids_mask & BIT(plane->id)) == 0)
+                       continue;
+
+               plane_state = intel_atomic_get_plane_state(state, plane);
+               if (IS_ERR(plane_state))
+                       return PTR_ERR(plane_state);
+       }
+
+       return 0;
+}
+
+int intel_atomic_add_affected_planes(struct intel_atomic_state *state,
+                                    struct intel_crtc *crtc)
+{
+       const struct intel_crtc_state *old_crtc_state =
+               intel_atomic_get_old_crtc_state(state, crtc);
+       const struct intel_crtc_state *new_crtc_state =
+               intel_atomic_get_new_crtc_state(state, crtc);
+
+       return intel_crtc_add_planes_to_state(state, crtc,
+                                             old_crtc_state->enabled_planes |
+                                             new_crtc_state->enabled_planes);
+}
+
+static bool active_planes_affects_min_cdclk(struct intel_display *display)
+{
+       /* See {hsw,vlv,ivb}_plane_ratio() */
+       return display->platform.broadwell || display->platform.haswell ||
+               display->platform.cherryview || display->platform.valleyview ||
+               display->platform.ivybridge;
+}
+
+static u8 intel_joiner_affected_planes(struct intel_atomic_state *state,
+                                      u8 joined_pipes)
+{
+       const struct intel_plane_state *plane_state;
+       struct intel_plane *plane;
+       u8 affected_planes = 0;
+       int i;
+
+       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
+               struct intel_plane *linked = plane_state->planar_linked_plane;
+
+               if ((joined_pipes & BIT(plane->pipe)) == 0)
+                       continue;
+
+               affected_planes |= BIT(plane->id);
+               if (linked)
+                       affected_planes |= BIT(linked->id);
+       }
+
+       return affected_planes;
+}
+
+static int intel_joiner_add_affected_planes(struct intel_atomic_state *state,
+                                           u8 joined_pipes)
+{
+       u8 prev_affected_planes, affected_planes = 0;
+
+       /*
+        * We want all the joined pipes to have the same
+        * set of planes in the atomic state, to make sure
+        * state copying always works correctly, and the
+        * UV<->Y plane linkage is always up to date.
+        * Keep pulling planes in until we've determined
+        * the full set of affected planes. A bit complicated
+        * on account of each pipe being capable of selecting
+        * their own Y planes independently of the other pipes,
+        * and the selection being done from the set of
+        * inactive planes.
+        */
+       do {
+               struct intel_crtc *crtc;
+
+               for_each_intel_crtc_in_pipe_mask(state->base.dev, crtc, joined_pipes) {
+                       int ret;
+
+                       ret = intel_crtc_add_planes_to_state(state, crtc, affected_planes);
+                       if (ret)
+                               return ret;
+               }
+
+               prev_affected_planes = affected_planes;
+               affected_planes = intel_joiner_affected_planes(state, joined_pipes);
+       } while (affected_planes != prev_affected_planes);
+
+       return 0;
+}
+
+static int intel_add_affected_planes(struct intel_atomic_state *state)
+{
+       const struct intel_crtc_state *crtc_state;
+       struct intel_crtc *crtc;
+       int i;
+
+       for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) {
+               int ret;
+
+               ret = intel_joiner_add_affected_planes(state, intel_crtc_joined_pipe_mask(crtc_state));
+               if (ret)
+                       return ret;
+       }
+
+       return 0;
+}
+
+int intel_atomic_check_planes(struct intel_atomic_state *state)
+{
+       struct intel_display *display = to_intel_display(state);
+       struct intel_crtc_state *old_crtc_state, *new_crtc_state;
+       struct intel_plane_state __maybe_unused *plane_state;
+       struct intel_plane *plane;
+       struct intel_crtc *crtc;
+       int i, ret;
+
+       ret = intel_add_affected_planes(state);
+       if (ret)
+               return ret;
+
+       for_each_new_intel_plane_in_state(state, plane, plane_state, i) {
+               ret = intel_plane_atomic_check(state, plane);
+               if (ret) {
+                       drm_dbg_atomic(display->drm,
+                                      "[PLANE:%d:%s] atomic driver check failed\n",
+                                      plane->base.base.id, plane->base.name);
+                       return ret;
+               }
+       }
+
+       for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state,
+                                           new_crtc_state, i) {
+               u8 old_active_planes, new_active_planes;
+
+               ret = icl_check_nv12_planes(state, crtc);
+               if (ret)
+                       return ret;
+
+               /*
+                * On some platforms the number of active planes affects
+                * the planes' minimum cdclk calculation. Add such planes
+                * to the state before we compute the minimum cdclk.
+                */
+               if (!active_planes_affects_min_cdclk(display))
+                       continue;
+
+               old_active_planes = old_crtc_state->active_planes & ~BIT(PLANE_CURSOR);
+               new_active_planes = new_crtc_state->active_planes & ~BIT(PLANE_CURSOR);
+
+               if (hweight8(old_active_planes) == hweight8(new_active_planes))
+                       continue;
+
+               ret = intel_crtc_add_planes_to_state(state, crtc, new_active_planes);
+               if (ret)
+                       return ret;
+       }
+
+       return 0;
+}
+
+u32 intel_plane_ggtt_offset(const struct intel_plane_state *plane_state)
+{
+       return i915_ggtt_offset(plane_state->ggtt_vma);
+}
diff --git a/drivers/gpu/drm/i915/display/intel_plane.h b/drivers/gpu/drm/i915/display/intel_plane.h
new file mode 100644 (file)
index 0000000..1dd3258
--- /dev/null
@@ -0,0 +1,97 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2019 Intel Corporation
+ */
+
+#ifndef __INTEL_PLANE_H__
+#define __INTEL_PLANE_H__
+
+#include <linux/types.h>
+
+struct drm_plane;
+struct drm_property;
+struct drm_rect;
+struct intel_atomic_state;
+struct intel_crtc;
+struct intel_crtc_state;
+struct intel_dsb;
+struct intel_plane;
+struct intel_plane_state;
+enum plane_id;
+
+struct intel_plane *
+intel_crtc_get_plane(struct intel_crtc *crtc, enum plane_id plane_id);
+bool intel_plane_can_async_flip(struct intel_plane *plane, u32 format,
+                               u64 modifier);
+unsigned int intel_adjusted_rate(const struct drm_rect *src,
+                                const struct drm_rect *dst,
+                                unsigned int rate);
+unsigned int intel_plane_pixel_rate(const struct intel_crtc_state *crtc_state,
+                                   const struct intel_plane_state *plane_state);
+
+unsigned int intel_plane_data_rate(const struct intel_crtc_state *crtc_state,
+                                  const struct intel_plane_state *plane_state,
+                                  int color_plane);
+void intel_plane_copy_uapi_to_hw_state(struct intel_plane_state *plane_state,
+                                      const struct intel_plane_state *from_plane_state,
+                                      struct intel_crtc *crtc);
+void intel_plane_copy_hw_state(struct intel_plane_state *plane_state,
+                              const struct intel_plane_state *from_plane_state);
+void intel_plane_async_flip(struct intel_dsb *dsb,
+                           struct intel_plane *plane,
+                           const struct intel_crtc_state *crtc_state,
+                           const struct intel_plane_state *plane_state,
+                           bool async_flip);
+void intel_plane_update_noarm(struct intel_dsb *dsb,
+                             struct intel_plane *plane,
+                             const struct intel_crtc_state *crtc_state,
+                             const struct intel_plane_state *plane_state);
+void intel_plane_update_arm(struct intel_dsb *dsb,
+                           struct intel_plane *plane,
+                           const struct intel_crtc_state *crtc_state,
+                           const struct intel_plane_state *plane_state);
+void intel_plane_disable_arm(struct intel_dsb *dsb,
+                            struct intel_plane *plane,
+                            const struct intel_crtc_state *crtc_state);
+struct intel_plane *intel_plane_alloc(void);
+void intel_plane_free(struct intel_plane *plane);
+void intel_plane_destroy(struct drm_plane *plane);
+struct drm_plane_state *intel_plane_duplicate_state(struct drm_plane *plane);
+void intel_plane_destroy_state(struct drm_plane *plane,
+                              struct drm_plane_state *state);
+void intel_crtc_planes_update_noarm(struct intel_dsb *dsb,
+                                   struct intel_atomic_state *state,
+                                   struct intel_crtc *crtc);
+void intel_crtc_planes_update_arm(struct intel_dsb *dsbx,
+                                 struct intel_atomic_state *state,
+                                 struct intel_crtc *crtc);
+int intel_plane_atomic_check_with_state(const struct intel_crtc_state *old_crtc_state,
+                                       struct intel_crtc_state *crtc_state,
+                                       const struct intel_plane_state *old_plane_state,
+                                       struct intel_plane_state *intel_state);
+int intel_plane_atomic_check(struct intel_atomic_state *state,
+                            struct intel_plane *plane);
+int intel_plane_calc_min_cdclk(struct intel_atomic_state *state,
+                              struct intel_plane *plane,
+                              bool *need_cdclk_calc);
+int intel_atomic_plane_check_clipping(struct intel_plane_state *plane_state,
+                                     struct intel_crtc_state *crtc_state,
+                                     int min_scale, int max_scale,
+                                     bool can_position);
+int intel_plane_check_src_coordinates(struct intel_plane_state *plane_state);
+void intel_plane_set_invisible(struct intel_crtc_state *crtc_state,
+                              struct intel_plane_state *plane_state);
+void intel_plane_helper_add(struct intel_plane *plane);
+bool intel_plane_needs_physical(struct intel_plane *plane);
+void intel_plane_init_cursor_vblank_work(struct intel_plane_state *old_plane_state,
+                                        struct intel_plane_state *new_plane_state);
+int intel_atomic_add_affected_planes(struct intel_atomic_state *state,
+                                    struct intel_crtc *crtc);
+int intel_atomic_check_planes(struct intel_atomic_state *state);
+
+u32 intel_plane_ggtt_offset(const struct intel_plane_state *plane_state);
+bool intel_plane_format_mod_supported_async(struct drm_plane *plane,
+                                           u32 format,
+                                           u64 modifier);
+
+#endif /* __INTEL_PLANE_H__ */
index 8800a657cd21f67228d7333f88e8dfe07b5323ae..2194d39a5c9821eab93ca55dd1b854ca497e6a64 100644 (file)
@@ -6,13 +6,13 @@
 #include "gem/i915_gem_lmem.h"
 #include "gem/i915_gem_region.h"
 #include "i915_drv.h"
-#include "intel_atomic_plane.h"
 #include "intel_crtc.h"
 #include "intel_display.h"
 #include "intel_display_core.h"
 #include "intel_display_types.h"
 #include "intel_fb.h"
 #include "intel_frontbuffer.h"
+#include "intel_plane.h"
 #include "intel_plane_initial.h"
 
 void intel_plane_initial_vblank_wait(struct intel_crtc *crtc)
index fd92e6b89b431b0db8f9b09c4d881e76a6bb1116..db0c48d0c743db706079cd20e6b279980a588a8c 100644 (file)
 
 #include "i915_utils.h"
 #include "i9xx_plane.h"
-#include "intel_atomic_plane.h"
 #include "intel_de.h"
 #include "intel_display_types.h"
 #include "intel_fb.h"
 #include "intel_frontbuffer.h"
+#include "intel_plane.h"
 #include "intel_sprite.h"
 #include "intel_sprite_regs.h"
 
index 2aa64482d44b3c5fde6671d10338ccd96b0d6b16..9c8dac97cc40a5f39d0e05df9afd5424cb31bfc0 100644 (file)
@@ -9,9 +9,7 @@
 #include <drm/drm_fourcc.h>
 
 #include "pxp/intel_pxp.h"
-
 #include "i915_drv.h"
-#include "intel_atomic_plane.h"
 #include "intel_bo.h"
 #include "intel_de.h"
 #include "intel_display_irq.h"
@@ -21,6 +19,7 @@
 #include "intel_fb.h"
 #include "intel_fbc.h"
 #include "intel_frontbuffer.h"
+#include "intel_plane.h"
 #include "intel_psr.h"
 #include "intel_psr_regs.h"
 #include "skl_scaler.h"
index 2c2371574d6ff38d9ca68f308ee7f95fe55793c8..c2e11d9ee652cc4282579b0a659048451470c37b 100644 (file)
@@ -13,7 +13,6 @@
 #include "i915_reg.h"
 #include "i9xx_wm.h"
 #include "intel_atomic.h"
-#include "intel_atomic_plane.h"
 #include "intel_bw.h"
 #include "intel_cdclk.h"
 #include "intel_crtc.h"
@@ -27,6 +26,7 @@
 #include "intel_fb.h"
 #include "intel_fixed.h"
 #include "intel_pcode.h"
+#include "intel_plane.h"
 #include "intel_wm.h"
 #include "skl_universal_plane_regs.h"
 #include "skl_watermark.h"
index e4bf484d4121691f223da83a589bf08c28ebf191..47a7989c0b85994b91743b084386b1febb95931c 100644 (file)
@@ -204,7 +204,6 @@ xe-$(CONFIG_DRM_XE_DISPLAY) += \
        i915-display/icl_dsi.o \
        i915-display/intel_alpm.o \
        i915-display/intel_atomic.o \
-       i915-display/intel_atomic_plane.o \
        i915-display/intel_audio.o \
        i915-display/intel_backlight.o \
        i915-display/intel_bios.o \
@@ -270,6 +269,7 @@ xe-$(CONFIG_DRM_XE_DISPLAY) += \
        i915-display/intel_modeset_verify.o \
        i915-display/intel_panel.o \
        i915-display/intel_pfit.o \
+       i915-display/intel_plane.o \
        i915-display/intel_pmdemand.o \
        i915-display/intel_pch.o \
        i915-display/intel_pps.o \
index 413600a4fd2151c239d6cf622fec83b461dc9f52..2ad6dd8e923ce2e09c4e98f8b3a4f59760cde33d 100644 (file)
@@ -10,7 +10,6 @@
 #include "xe_ggtt.h"
 #include "xe_mmio.h"
 
-#include "intel_atomic_plane.h"
 #include "intel_crtc.h"
 #include "intel_display.h"
 #include "intel_display_core.h"
@@ -19,6 +18,7 @@
 #include "intel_fb.h"
 #include "intel_fb_pin.h"
 #include "intel_frontbuffer.h"
+#include "intel_plane.h"
 #include "intel_plane_initial.h"
 #include "xe_bo.h"
 #include "xe_wa.h"