return plane >= fb->format->num_planes / 2;
 }
 
+static bool is_gen12_ccs_modifier(u64 modifier)
+{
+       return modifier == I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS;
+}
+
+static bool is_gen12_ccs_plane(const struct drm_framebuffer *fb, int plane)
+{
+       return is_gen12_ccs_modifier(fb->modifier) && is_ccs_plane(fb, plane);
+}
+
 static bool is_aux_plane(const struct drm_framebuffer *fb, int plane)
 {
        if (is_ccs_modifier(fb->modifier))
        return fb->format->num_planes / 2 + main_plane;
 }
 
+static int ccs_to_main_plane(const struct drm_framebuffer *fb, int ccs_plane)
+{
+       WARN_ON(!is_ccs_modifier(fb->modifier) ||
+               ccs_plane < fb->format->num_planes / 2);
+
+       return ccs_plane - fb->format->num_planes / 2;
+}
+
 /* Return either the main plane's CCS or - if not a CCS FB - UV plane */
 static int
 intel_main_to_aux_plane(const struct drm_framebuffer *fb, int main_plane)
                if (is_ccs_plane(fb, color_plane))
                        return 128;
                /* fall through */
+       case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
+               if (is_ccs_plane(fb, color_plane))
+                       return 64;
+               /* fall through */
        case I915_FORMAT_MOD_Y_TILED:
                if (IS_GEN(dev_priv, 2) || HAS_128_BYTE_Y_TILING(dev_priv))
                        return 128;
 static unsigned int
 intel_tile_height(const struct drm_framebuffer *fb, int color_plane)
 {
+       if (is_gen12_ccs_plane(fb, color_plane))
+               return 1;
+
        return intel_tile_size(to_i915(fb->dev)) /
                intel_tile_width_bytes(fb, color_plane);
 }
                if (INTEL_GEN(dev_priv) >= 9)
                        return 256 * 1024;
                return 0;
+       case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
+               return 16 * 1024;
        case I915_FORMAT_MOD_Y_TILED_CCS:
        case I915_FORMAT_MOD_Yf_TILED_CCS:
        case I915_FORMAT_MOD_Y_TILED:
        return new_offset;
 }
 
-static bool is_surface_linear(u64 modifier, int color_plane)
+static bool is_surface_linear(const struct drm_framebuffer *fb, int color_plane)
 {
-       return modifier == DRM_FORMAT_MOD_LINEAR;
+       return fb->modifier == DRM_FORMAT_MOD_LINEAR ||
+              is_gen12_ccs_plane(fb, color_plane);
 }
 
 static u32 intel_adjust_aligned_offset(int *x, int *y,
 
        WARN_ON(new_offset > old_offset);
 
-       if (!is_surface_linear(fb->modifier, color_plane)) {
+       if (!is_surface_linear(fb, color_plane)) {
                unsigned int tile_size, tile_width, tile_height;
                unsigned int pitch_tiles;
 
        if (alignment)
                alignment--;
 
-       if (!is_surface_linear(fb->modifier, color_plane)) {
+       if (!is_surface_linear(fb, color_plane)) {
                unsigned int tile_size, tile_width, tile_height;
                unsigned int tile_rows, tiles, pitch_tiles;
 
                return I915_TILING_X;
        case I915_FORMAT_MOD_Y_TILED:
        case I915_FORMAT_MOD_Y_TILED_CCS:
+       case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
                return I915_TILING_Y;
        default:
                return I915_TILING_NONE;
  * us a ratio of one byte in the CCS for each 8x16 pixels in the
  * main surface.
  */
-static const struct drm_format_info ccs_formats[] = {
+static const struct drm_format_info skl_ccs_formats[] = {
        { .format = DRM_FORMAT_XRGB8888, .depth = 24, .num_planes = 2,
          .cpp = { 4, 1, }, .hsub = 8, .vsub = 16, },
        { .format = DRM_FORMAT_XBGR8888, .depth = 24, .num_planes = 2,
          .cpp = { 4, 1, }, .hsub = 8, .vsub = 16, .has_alpha = true, },
 };
 
+/*
+ * Gen-12 compression uses 4 bits of CCS data for each cache line pair in the
+ * main surface. And each 64B CCS cache line represents an area of 4x1 Y-tiles
+ * in the main surface. With 4 byte pixels and each Y-tile having dimensions of
+ * 32x32 pixels, the ratio turns out to 1B in the CCS for every 2x32 pixels in
+ * the main surface.
+ */
+static const struct drm_format_info gen12_ccs_formats[] = {
+       { .format = DRM_FORMAT_XRGB8888, .depth = 24, .num_planes = 2,
+         .char_per_block = { 4, 1 }, .block_w = { 1, 2 }, .block_h = { 1, 1 },
+         .hsub = 1, .vsub = 1, },
+       { .format = DRM_FORMAT_XBGR8888, .depth = 24, .num_planes = 2,
+         .char_per_block = { 4, 1 }, .block_w = { 1, 2 }, .block_h = { 1, 1 },
+         .hsub = 1, .vsub = 1, },
+       { .format = DRM_FORMAT_ARGB8888, .depth = 32, .num_planes = 2,
+         .char_per_block = { 4, 1 }, .block_w = { 1, 2 }, .block_h = { 1, 1 },
+         .hsub = 1, .vsub = 1, .has_alpha = true },
+       { .format = DRM_FORMAT_ABGR8888, .depth = 32, .num_planes = 2,
+         .char_per_block = { 4, 1 }, .block_w = { 1, 2 }, .block_h = { 1, 1 },
+         .hsub = 1, .vsub = 1, .has_alpha = true },
+};
+
 static const struct drm_format_info *
 lookup_format_info(const struct drm_format_info formats[],
                   int num_formats, u32 format)
        switch (cmd->modifier[0]) {
        case I915_FORMAT_MOD_Y_TILED_CCS:
        case I915_FORMAT_MOD_Yf_TILED_CCS:
-               return lookup_format_info(ccs_formats,
-                                         ARRAY_SIZE(ccs_formats),
+               return lookup_format_info(skl_ccs_formats,
+                                         ARRAY_SIZE(skl_ccs_formats),
+                                         cmd->pixel_format);
+       case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
+               return lookup_format_info(gen12_ccs_formats,
+                                         ARRAY_SIZE(gen12_ccs_formats),
                                          cmd->pixel_format);
        default:
                return NULL;
 
 bool is_ccs_modifier(u64 modifier)
 {
-       return modifier == I915_FORMAT_MOD_Y_TILED_CCS ||
+       return modifier == I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS ||
+              modifier == I915_FORMAT_MOD_Y_TILED_CCS ||
               modifier == I915_FORMAT_MOD_Yf_TILED_CCS;
 }
 
 intel_fb_stride_alignment(const struct drm_framebuffer *fb, int color_plane)
 {
        struct drm_i915_private *dev_priv = to_i915(fb->dev);
+       u32 tile_width;
 
-       if (fb->modifier == DRM_FORMAT_MOD_LINEAR) {
+       if (is_surface_linear(fb, color_plane)) {
                u32 max_stride = intel_plane_fb_max_stride(dev_priv,
                                                           fb->format->format,
                                                           fb->modifier);
                 * To make remapping with linear generally feasible
                 * we need the stride to be page aligned.
                 */
-               if (fb->pitches[color_plane] > max_stride)
+               if (fb->pitches[color_plane] > max_stride &&
+                   !is_ccs_modifier(fb->modifier))
                        return intel_tile_size(dev_priv);
                else
                        return 64;
-       } else {
-               u32 tile_width = intel_tile_width_bytes(fb, color_plane);
+       }
 
+       tile_width = intel_tile_width_bytes(fb, color_plane);
+       if (is_ccs_modifier(fb->modifier) && color_plane == 0) {
                /*
                 * Display WA #0531: skl,bxt,kbl,glk
                 *
                 * require the entire fb to accommodate that to avoid
                 * potential runtime errors at plane configuration time.
                 */
-               if (IS_GEN(dev_priv, 9) && is_ccs_modifier(fb->modifier) &&
-                   color_plane == 0 && fb->width > 3840)
+               if (IS_GEN(dev_priv, 9) && fb->width > 3840)
+                       tile_width *= 4;
+               /*
+                * The main surface pitch must be padded to a multiple of four
+                * tile widths.
+                */
+               else if (INTEL_GEN(dev_priv) >= 12)
                        tile_width *= 4;
-
-               return tile_width;
        }
+       return tile_width;
 }
 
 bool intel_plane_can_remap(const struct intel_plane_state *plane_state)
        return stride > max_stride;
 }
 
+static void
+intel_fb_plane_get_subsampling(int *hsub, int *vsub,
+                              const struct drm_framebuffer *fb,
+                              int color_plane)
+{
+       int main_plane;
+
+       if (color_plane == 0) {
+               *hsub = 1;
+               *vsub = 1;
+
+               return;
+       }
+
+       /*
+        * TODO: Deduct the subsampling from the char block for all CCS
+        * formats and planes.
+        */
+       if (!is_gen12_ccs_plane(fb, color_plane)) {
+               *hsub = fb->format->hsub;
+               *vsub = fb->format->vsub;
+
+               return;
+       }
+
+       main_plane = ccs_to_main_plane(fb, color_plane);
+       *hsub = drm_format_info_block_width(fb->format, color_plane) /
+               drm_format_info_block_width(fb->format, main_plane);
+
+       /*
+        * The min stride check in the core framebuffer_check() function
+        * assumes that format->hsub applies to every plane except for the
+        * first plane. That's incorrect for the CCS AUX plane of the first
+        * plane, but for the above check to pass we must define the block
+        * width with that subsampling applied to it. Adjust the width here
+        * accordingly, so we can calculate the actual subsampling factor.
+        */
+       if (main_plane == 0)
+               *hsub *= fb->format->hsub;
+
+       *vsub = 32;
+}
 static int
 intel_fb_check_ccs_xy(struct drm_framebuffer *fb, int ccs_plane, int x, int y)
 {
        struct intel_framebuffer *intel_fb = to_intel_framebuffer(fb);
-       int hsub = fb->format->hsub;
-       int vsub = fb->format->vsub;
+       int main_plane;
+       int hsub, vsub;
        int tile_width, tile_height;
        int ccs_x, ccs_y;
        int main_x, main_y;
 
-       if (!is_ccs_modifier(fb->modifier) || ccs_plane != 1)
+       if (!is_ccs_plane(fb, ccs_plane))
                return 0;
 
-       intel_tile_dims(fb, 1, &tile_width, &tile_height);
+       intel_tile_dims(fb, ccs_plane, &tile_width, &tile_height);
+       intel_fb_plane_get_subsampling(&hsub, &vsub, fb, ccs_plane);
 
        tile_width *= hsub;
        tile_height *= vsub;
 
        ccs_x = (x * hsub) % tile_width;
        ccs_y = (y * vsub) % tile_height;
-       main_x = intel_fb->normal[0].x % tile_width;
-       main_y = intel_fb->normal[0].y % tile_height;
+
+       main_plane = ccs_to_main_plane(fb, ccs_plane);
+       main_x = intel_fb->normal[main_plane].x % tile_width;
+       main_y = intel_fb->normal[main_plane].y % tile_height;
 
        /*
         * CCS doesn't have its own x/y offset register, so the intra CCS tile
                DRM_DEBUG_KMS("Bad CCS x/y (main %d,%d ccs %d,%d) full (main %d,%d ccs %d,%d)\n",
                              main_x, main_y,
                              ccs_x, ccs_y,
-                             intel_fb->normal[0].x,
-                             intel_fb->normal[0].y,
+                             intel_fb->normal[main_plane].x,
+                             intel_fb->normal[main_plane].y,
                              x, y);
                return -EINVAL;
        }
        return 0;
 }
 
+static void
+intel_fb_plane_dims(int *w, int *h, struct drm_framebuffer *fb, int color_plane)
+{
+       int hsub, vsub;
+
+       intel_fb_plane_get_subsampling(&hsub, &vsub, fb, color_plane);
+       *w = fb->width / hsub;
+       *h = fb->height / vsub;
+}
+
 static int
 intel_fill_fb_info(struct drm_i915_private *dev_priv,
                   struct drm_framebuffer *fb)
                int ret;
 
                cpp = fb->format->cpp[i];
-               width = drm_framebuffer_plane_width(fb->width, fb, i);
-               height = drm_framebuffer_plane_height(fb->height, fb, i);
+               intel_fb_plane_dims(&width, &height, fb, i);
 
                ret = intel_fb_offset_to_xy(&x, &y, fb, i);
                if (ret) {
                                                      tile_size);
                offset /= tile_size;
 
-               if (!is_surface_linear(fb->modifier, i)) {
+               if (!is_surface_linear(fb, i)) {
                        unsigned int tile_width, tile_height;
                        unsigned int pitch_tiles;
                        struct drm_rect r;
                                           int main_x, int main_y, u32 main_offset)
 {
        const struct drm_framebuffer *fb = plane_state->hw.fb;
-       int hsub = fb->format->hsub;
-       int vsub = fb->format->vsub;
        int ccs_plane = main_to_ccs_plane(fb, 0);
        int aux_x = plane_state->color_plane[ccs_plane].x;
        int aux_y = plane_state->color_plane[ccs_plane].y;
        u32 aux_offset = plane_state->color_plane[ccs_plane].offset;
        u32 alignment = intel_surf_alignment(fb, ccs_plane);
+       int hsub;
+       int vsub;
 
+       intel_fb_plane_get_subsampling(&hsub, &vsub, fb, ccs_plane);
        while (aux_offset >= main_offset && aux_y <= main_y) {
                int x, y;
 
        const struct drm_framebuffer *fb = plane_state->hw.fb;
        int src_x = plane_state->uapi.src.x1 >> 16;
        int src_y = plane_state->uapi.src.y1 >> 16;
-       int hsub = fb->format->hsub;
-       int vsub = fb->format->vsub;
-       int x = src_x / hsub;
-       int y = src_y / vsub;
+       int hsub;
+       int vsub;
+       int x;
+       int y;
        u32 offset;
 
+       intel_fb_plane_get_subsampling(&hsub, &vsub, fb, 1);
+       x = src_x / hsub;
+       y = src_y / vsub;
        intel_add_fb_offsets(&x, &y, plane_state, 1);
        offset = intel_plane_compute_aligned_offset(&x, &y, plane_state, 1);
 
         * The stride is either expressed as a multiple of 64 bytes chunks for
         * linear buffers or in number of tiles for tiled buffers.
         */
-       if (fb->modifier == DRM_FORMAT_MOD_LINEAR)
+       if (is_surface_linear(fb, color_plane))
                return 64;
        else if (drm_rotation_90_or_270(rotation))
                return intel_tile_height(fb, color_plane);
                return PLANE_CTL_TILED_Y;
        case I915_FORMAT_MOD_Y_TILED_CCS:
                return PLANE_CTL_TILED_Y | PLANE_CTL_RENDER_DECOMPRESSION_ENABLE;
+       case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
+               return PLANE_CTL_TILED_Y |
+                      PLANE_CTL_RENDER_DECOMPRESSION_ENABLE |
+                      PLANE_CTL_CLEAR_COLOR_DISABLE;
        case I915_FORMAT_MOD_Yf_TILED:
                return PLANE_CTL_TILED_YF;
        case I915_FORMAT_MOD_Yf_TILED_CCS:
        case PLANE_CTL_TILED_Y:
                plane_config->tiling = I915_TILING_Y;
                if (val & PLANE_CTL_RENDER_DECOMPRESSION_ENABLE)
-                       fb->modifier = I915_FORMAT_MOD_Y_TILED_CCS;
+                       fb->modifier = INTEL_GEN(dev_priv) >= 12 ?
+                               I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS :
+                               I915_FORMAT_MOD_Y_TILED_CCS;
                else
                        fb->modifier = I915_FORMAT_MOD_Y_TILED;
                break;