drm/amd/display: Merge amdgpu_dm_types and amdgpu_dm
authorHarry Wentland <harry.wentland@amd.com>
Thu, 10 Aug 2017 17:29:07 +0000 (13:29 -0400)
committerAlex Deucher <alexander.deucher@amd.com>
Tue, 26 Sep 2017 22:15:52 +0000 (18:15 -0400)
This separation stopped making sense a long time ago.

Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/display/amdgpu_dm/Makefile
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c [deleted file]
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.h [deleted file]

index 698b1d4f83f522ef19f35a7071f057ba9200f9ee..4699e47aa76b00969bd0a7a7cc1f074df709563c 100644 (file)
@@ -4,7 +4,7 @@
 
 
 
-AMDGPUDM = amdgpu_dm_types.o amdgpu_dm.o amdgpu_dm_irq.o amdgpu_dm_mst_types.o
+AMDGPUDM = amdgpu_dm.o amdgpu_dm_irq.o amdgpu_dm_mst_types.o
 
 ifneq ($(CONFIG_DRM_AMD_DC),)
 AMDGPUDM += amdgpu_dm_services.o amdgpu_dm_helpers.o
index c802437f4858f6b1a55b185e5e2d42e49753aeba..646a946a2571f7fe06471e97cf4d8427c3bc010a 100644 (file)
 #include "amdgpu_display.h"
 #include "atom.h"
 #include "amdgpu_dm.h"
-#include "amdgpu_dm_types.h"
+#include "amdgpu_pm.h"
 
 #include "amd_shared.h"
 #include "amdgpu_dm_irq.h"
 #include "dm_helpers.h"
+#include "dm_services_types.h"
+#include "amdgpu_dm_mst_types.h"
 
 #include "ivsrcid/ivsrcid_vislands30.h"
 
 #include <linux/module.h>
 #include <linux/moduleparam.h>
 #include <linux/version.h>
+#include <linux/types.h>
 
+#include <drm/drmP.h>
 #include <drm/drm_atomic.h>
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_dp_mst_helper.h>
+#include <drm/drm_fb_helper.h>
+#include <drm/drm_edid.h>
 
 #include "modules/inc/mod_freesync.h"
 
 #include "soc15_common.h"
 #endif
 
+#include "modules/inc/mod_freesync.h"
+
+#include "i2caux_interface.h"
+
+
 static enum drm_plane_type dm_surfaces_type_default[AMDGPU_MAX_PLANES] = {
        DRM_PLANE_TYPE_PRIMARY,
        DRM_PLANE_TYPE_PRIMARY,
@@ -1552,8 +1563,3238 @@ bool amdgpu_dm_acquire_dal_lock(struct amdgpu_display_manager *dm)
 
 bool amdgpu_dm_release_dal_lock(struct amdgpu_display_manager *dm)
 {
-       /* TODO */
+       /* TODO */      return true;
+}
+
+
+struct dm_connector_state {
+       struct drm_connector_state base;
+
+       enum amdgpu_rmx_type scaling;
+       uint8_t underscan_vborder;
+       uint8_t underscan_hborder;
+       bool underscan_enable;
+};
+
+#define to_dm_connector_state(x)\
+       container_of((x), struct dm_connector_state, base)
+
+static bool modeset_required(struct drm_crtc_state *crtc_state)
+{
+       if (!drm_atomic_crtc_needs_modeset(crtc_state))
+               return false;
+
+       if (!crtc_state->enable)
+               return false;
+
+       return crtc_state->active;
+}
+
+static bool modereset_required(struct drm_crtc_state *crtc_state)
+{
+       if (!drm_atomic_crtc_needs_modeset(crtc_state))
+               return false;
+
+       return !crtc_state->enable || !crtc_state->active;
+}
+
+void amdgpu_dm_encoder_destroy(struct drm_encoder *encoder)
+{
+       drm_encoder_cleanup(encoder);
+       kfree(encoder);
+}
+
+static const struct drm_encoder_funcs amdgpu_dm_encoder_funcs = {
+       .destroy = amdgpu_dm_encoder_destroy,
+};
+
+static void dm_set_cursor(
+       struct amdgpu_crtc *amdgpu_crtc,
+       uint64_t gpu_addr,
+       uint32_t width,
+       uint32_t height)
+{
+       struct dc_cursor_attributes attributes;
+       struct dc_cursor_position position;
+       struct drm_crtc *crtc = &amdgpu_crtc->base;
+       int x, y;
+       int xorigin = 0, yorigin = 0;
+       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state);
+
+       amdgpu_crtc->cursor_width = width;
+       amdgpu_crtc->cursor_height = height;
+
+       attributes.address.high_part = upper_32_bits(gpu_addr);
+       attributes.address.low_part  = lower_32_bits(gpu_addr);
+       attributes.width             = width;
+       attributes.height            = height;
+       attributes.color_format      = CURSOR_MODE_COLOR_PRE_MULTIPLIED_ALPHA;
+       attributes.rotation_angle    = 0;
+       attributes.attribute_flags.value = 0;
+
+       attributes.pitch = attributes.width;
+
+       x = amdgpu_crtc->cursor_x;
+       y = amdgpu_crtc->cursor_y;
+
+       /* avivo cursor are offset into the total surface */
+       x += crtc->primary->state->src_x >> 16;
+       y += crtc->primary->state->src_y >> 16;
+
+       if (x < 0) {
+               xorigin = min(-x, amdgpu_crtc->max_cursor_width - 1);
+               x = 0;
+       }
+       if (y < 0) {
+               yorigin = min(-y, amdgpu_crtc->max_cursor_height - 1);
+               y = 0;
+       }
+
+       position.enable = true;
+       position.x = x;
+       position.y = y;
+
+       position.x_hotspot = xorigin;
+       position.y_hotspot = yorigin;
+
+       if (!dc_stream_set_cursor_attributes(
+                               acrtc_state->stream,
+                               &attributes)) {
+               DRM_ERROR("DC failed to set cursor attributes\n");
+       }
+
+       if (!dc_stream_set_cursor_position(
+                               acrtc_state->stream,
+                               &position)) {
+               DRM_ERROR("DC failed to set cursor position\n");
+       }
+}
+
+static int dm_crtc_cursor_set(
+       struct drm_crtc *crtc,
+       uint64_t address,
+       uint32_t width,
+       uint32_t height)
+{
+       struct dc_cursor_position position;
+       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state);
+
+       int ret;
+
+       struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+       ret             = EINVAL;
+
+       DRM_DEBUG_KMS(
+               "%s: crtc_id=%d with size %d to %d \n",
+               __func__,
+               amdgpu_crtc->crtc_id,
+               width,
+               height);
+
+       if (!address) {
+               /* turn off cursor */
+               position.enable = false;
+               position.x = 0;
+               position.y = 0;
+
+               if (acrtc_state->stream) {
+                       /*set cursor visible false*/
+                       dc_stream_set_cursor_position(
+                               acrtc_state->stream,
+                               &position);
+               }
+               goto release;
+
+       }
+
+       if ((width > amdgpu_crtc->max_cursor_width) ||
+               (height > amdgpu_crtc->max_cursor_height)) {
+               DRM_ERROR(
+                       "%s: bad cursor width or height %d x %d\n",
+                       __func__,
+                       width,
+                       height);
+               goto release;
+       }
+
+       /*program new cursor bo to hardware*/
+       dm_set_cursor(amdgpu_crtc, address, width, height);
+
+release:
+       return ret;
+
+}
+
+static int dm_crtc_cursor_move(struct drm_crtc *crtc,
+                                    int x, int y)
+{
+       struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+       int xorigin = 0, yorigin = 0;
+       struct dc_cursor_position position;
+       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state);
+
+       amdgpu_crtc->cursor_x = x;
+       amdgpu_crtc->cursor_y = y;
+
+       /* avivo cursor are offset into the total surface */
+       x += crtc->primary->state->src_x >> 16;
+       y += crtc->primary->state->src_y >> 16;
+
+       /*
+        * TODO: for cursor debugging unguard the following
+        */
+#if 0
+       DRM_DEBUG_KMS(
+               "%s: x %d y %d c->x %d c->y %d\n",
+               __func__,
+               x,
+               y,
+               crtc->x,
+               crtc->y);
+#endif
+
+       if (x < 0) {
+               xorigin = min(-x, amdgpu_crtc->max_cursor_width - 1);
+               x = 0;
+       }
+       if (y < 0) {
+               yorigin = min(-y, amdgpu_crtc->max_cursor_height - 1);
+               y = 0;
+       }
+
+       position.enable = true;
+       position.x = x;
+       position.y = y;
+
+       position.x_hotspot = xorigin;
+       position.y_hotspot = yorigin;
+
+       if (acrtc_state->stream) {
+               if (!dc_stream_set_cursor_position(
+                                       acrtc_state->stream,
+                                       &position)) {
+                       DRM_ERROR("DC failed to set cursor position\n");
+                       return -EINVAL;
+               }
+       }
+
+       return 0;
+}
+
+static bool fill_rects_from_plane_state(
+       const struct drm_plane_state *state,
+       struct dc_surface *surface)
+{
+       surface->src_rect.x = state->src_x >> 16;
+       surface->src_rect.y = state->src_y >> 16;
+       /*we ignore for now mantissa and do not to deal with floating pixels :(*/
+       surface->src_rect.width = state->src_w >> 16;
+
+       if (surface->src_rect.width == 0)
+               return false;
+
+       surface->src_rect.height = state->src_h >> 16;
+       if (surface->src_rect.height == 0)
+               return false;
+
+       surface->dst_rect.x = state->crtc_x;
+       surface->dst_rect.y = state->crtc_y;
+
+       if (state->crtc_w == 0)
+               return false;
+
+       surface->dst_rect.width = state->crtc_w;
+
+       if (state->crtc_h == 0)
+               return false;
+
+       surface->dst_rect.height = state->crtc_h;
+
+       surface->clip_rect = surface->dst_rect;
+
+       switch (state->rotation & DRM_MODE_ROTATE_MASK) {
+       case DRM_MODE_ROTATE_0:
+               surface->rotation = ROTATION_ANGLE_0;
+               break;
+       case DRM_MODE_ROTATE_90:
+               surface->rotation = ROTATION_ANGLE_90;
+               break;
+       case DRM_MODE_ROTATE_180:
+               surface->rotation = ROTATION_ANGLE_180;
+               break;
+       case DRM_MODE_ROTATE_270:
+               surface->rotation = ROTATION_ANGLE_270;
+               break;
+       default:
+               surface->rotation = ROTATION_ANGLE_0;
+               break;
+       }
+
        return true;
 }
+static int get_fb_info(
+       const struct amdgpu_framebuffer *amdgpu_fb,
+       uint64_t *tiling_flags,
+       uint64_t *fb_location)
+{
+       struct amdgpu_bo *rbo = gem_to_amdgpu_bo(amdgpu_fb->obj);
+       int r = amdgpu_bo_reserve(rbo, false);
+       if (unlikely(r)) {
+               DRM_ERROR("Unable to reserve buffer\n");
+               return r;
+       }
+
+       if (fb_location)
+               *fb_location = amdgpu_bo_gpu_offset(rbo);
+
+       if (tiling_flags)
+               amdgpu_bo_get_tiling_flags(rbo, tiling_flags);
+
+       amdgpu_bo_unreserve(rbo);
+
+       return r;
+}
+
+static int fill_plane_attributes_from_fb(
+       struct amdgpu_device *adev,
+       struct dc_surface *surface,
+       const struct amdgpu_framebuffer *amdgpu_fb, bool addReq)
+{
+       uint64_t tiling_flags;
+       uint64_t fb_location = 0;
+       unsigned int awidth;
+       const struct drm_framebuffer *fb = &amdgpu_fb->base;
+       int ret = 0;
+       struct drm_format_name_buf format_name;
+
+       ret = get_fb_info(
+               amdgpu_fb,
+               &tiling_flags,
+               addReq == true ? &fb_location:NULL);
+
+       if (ret)
+               return ret;
+
+       switch (fb->format->format) {
+       case DRM_FORMAT_C8:
+               surface->format = SURFACE_PIXEL_FORMAT_GRPH_PALETA_256_COLORS;
+               break;
+       case DRM_FORMAT_RGB565:
+               surface->format = SURFACE_PIXEL_FORMAT_GRPH_RGB565;
+               break;
+       case DRM_FORMAT_XRGB8888:
+       case DRM_FORMAT_ARGB8888:
+               surface->format = SURFACE_PIXEL_FORMAT_GRPH_ARGB8888;
+               break;
+       case DRM_FORMAT_XRGB2101010:
+       case DRM_FORMAT_ARGB2101010:
+               surface->format = SURFACE_PIXEL_FORMAT_GRPH_ARGB2101010;
+               break;
+       case DRM_FORMAT_XBGR2101010:
+       case DRM_FORMAT_ABGR2101010:
+               surface->format = SURFACE_PIXEL_FORMAT_GRPH_ABGR2101010;
+               break;
+       case DRM_FORMAT_NV21:
+               surface->format = SURFACE_PIXEL_FORMAT_VIDEO_420_YCbCr;
+               break;
+       case DRM_FORMAT_NV12:
+               surface->format = SURFACE_PIXEL_FORMAT_VIDEO_420_YCrCb;
+               break;
+       default:
+               DRM_ERROR("Unsupported screen format %s\n",
+                         drm_get_format_name(fb->format->format, &format_name));
+               return -EINVAL;
+       }
+
+       if (surface->format < SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) {
+               surface->address.type = PLN_ADDR_TYPE_GRAPHICS;
+               surface->address.grph.addr.low_part = lower_32_bits(fb_location);
+               surface->address.grph.addr.high_part = upper_32_bits(fb_location);
+               surface->plane_size.grph.surface_size.x = 0;
+               surface->plane_size.grph.surface_size.y = 0;
+               surface->plane_size.grph.surface_size.width = fb->width;
+               surface->plane_size.grph.surface_size.height = fb->height;
+               surface->plane_size.grph.surface_pitch =
+                               fb->pitches[0] / fb->format->cpp[0];
+               /* TODO: unhardcode */
+               surface->color_space = COLOR_SPACE_SRGB;
+
+       } else {
+               awidth = ALIGN(fb->width, 64);
+               surface->address.type = PLN_ADDR_TYPE_VIDEO_PROGRESSIVE;
+               surface->address.video_progressive.luma_addr.low_part
+                                               = lower_32_bits(fb_location);
+               surface->address.video_progressive.chroma_addr.low_part
+                                               = lower_32_bits(fb_location) +
+                                                       (awidth * fb->height);
+               surface->plane_size.video.luma_size.x = 0;
+               surface->plane_size.video.luma_size.y = 0;
+               surface->plane_size.video.luma_size.width = awidth;
+               surface->plane_size.video.luma_size.height = fb->height;
+               /* TODO: unhardcode */
+               surface->plane_size.video.luma_pitch = awidth;
+
+               surface->plane_size.video.chroma_size.x = 0;
+               surface->plane_size.video.chroma_size.y = 0;
+               surface->plane_size.video.chroma_size.width = awidth;
+               surface->plane_size.video.chroma_size.height = fb->height;
+               surface->plane_size.video.chroma_pitch = awidth / 2;
+
+               /* TODO: unhardcode */
+               surface->color_space = COLOR_SPACE_YCBCR709;
+       }
+
+       memset(&surface->tiling_info, 0, sizeof(surface->tiling_info));
+
+       /* Fill GFX params */
+       if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == DC_ARRAY_2D_TILED_THIN1)
+       {
+               unsigned bankw, bankh, mtaspect, tile_split, num_banks;
+
+               bankw = AMDGPU_TILING_GET(tiling_flags, BANK_WIDTH);
+               bankh = AMDGPU_TILING_GET(tiling_flags, BANK_HEIGHT);
+               mtaspect = AMDGPU_TILING_GET(tiling_flags, MACRO_TILE_ASPECT);
+               tile_split = AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT);
+               num_banks = AMDGPU_TILING_GET(tiling_flags, NUM_BANKS);
+
+               /* XXX fix me for VI */
+               surface->tiling_info.gfx8.num_banks = num_banks;
+               surface->tiling_info.gfx8.array_mode =
+                               DC_ARRAY_2D_TILED_THIN1;
+               surface->tiling_info.gfx8.tile_split = tile_split;
+               surface->tiling_info.gfx8.bank_width = bankw;
+               surface->tiling_info.gfx8.bank_height = bankh;
+               surface->tiling_info.gfx8.tile_aspect = mtaspect;
+               surface->tiling_info.gfx8.tile_mode =
+                               DC_ADDR_SURF_MICRO_TILING_DISPLAY;
+       } else if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE)
+                       == DC_ARRAY_1D_TILED_THIN1) {
+               surface->tiling_info.gfx8.array_mode = DC_ARRAY_1D_TILED_THIN1;
+       }
+
+       surface->tiling_info.gfx8.pipe_config =
+                       AMDGPU_TILING_GET(tiling_flags, PIPE_CONFIG);
+
+       if (adev->asic_type == CHIP_VEGA10 ||
+           adev->asic_type == CHIP_RAVEN) {
+               /* Fill GFX9 params */
+               surface->tiling_info.gfx9.num_pipes =
+                       adev->gfx.config.gb_addr_config_fields.num_pipes;
+               surface->tiling_info.gfx9.num_banks =
+                       adev->gfx.config.gb_addr_config_fields.num_banks;
+               surface->tiling_info.gfx9.pipe_interleave =
+                       adev->gfx.config.gb_addr_config_fields.pipe_interleave_size;
+               surface->tiling_info.gfx9.num_shader_engines =
+                       adev->gfx.config.gb_addr_config_fields.num_se;
+               surface->tiling_info.gfx9.max_compressed_frags =
+                       adev->gfx.config.gb_addr_config_fields.max_compress_frags;
+               surface->tiling_info.gfx9.num_rb_per_se =
+                       adev->gfx.config.gb_addr_config_fields.num_rb_per_se;
+               surface->tiling_info.gfx9.swizzle =
+                       AMDGPU_TILING_GET(tiling_flags, SWIZZLE_MODE);
+               surface->tiling_info.gfx9.shaderEnable = 1;
+       }
+
+       surface->visible = true;
+       surface->scaling_quality.h_taps_c = 0;
+       surface->scaling_quality.v_taps_c = 0;
+
+       /* is this needed? is surface zeroed at allocation? */
+       surface->scaling_quality.h_taps = 0;
+       surface->scaling_quality.v_taps = 0;
+       surface->stereo_format = PLANE_STEREO_FORMAT_NONE;
+
+       return ret;
+
+}
+
+#define NUM_OF_RAW_GAMMA_RAMP_RGB_256 256
+
+static void fill_gamma_from_crtc_state(
+       const struct drm_crtc_state *crtc_state,
+       struct dc_surface *dc_surface)
+{
+       int i;
+       struct dc_gamma *gamma;
+       struct drm_color_lut *lut = (struct drm_color_lut *) crtc_state->gamma_lut->data;
+
+       gamma = dc_create_gamma();
+
+       if (gamma == NULL) {
+               WARN_ON(1);
+               return;
+       }
+
+       for (i = 0; i < NUM_OF_RAW_GAMMA_RAMP_RGB_256; i++) {
+               gamma->red[i] = lut[i].red;
+               gamma->green[i] = lut[i].green;
+               gamma->blue[i] = lut[i].blue;
+       }
+
+       dc_surface->gamma_correction = gamma;
+}
+
+static int fill_plane_attributes(
+                       struct amdgpu_device *adev,
+                       struct dc_surface *surface,
+                       struct drm_plane_state *plane_state,
+                       struct drm_crtc_state *crtc_state,
+                       bool addrReq)
+{
+       const struct amdgpu_framebuffer *amdgpu_fb =
+               to_amdgpu_framebuffer(plane_state->fb);
+       const struct drm_crtc *crtc = plane_state->crtc;
+       struct dc_transfer_func *input_tf;
+       int ret = 0;
+
+       if (!fill_rects_from_plane_state(plane_state, surface))
+               return -EINVAL;
+
+       ret = fill_plane_attributes_from_fb(
+               crtc->dev->dev_private,
+               surface,
+               amdgpu_fb,
+               addrReq);
+
+       if (ret)
+               return ret;
+
+       input_tf = dc_create_transfer_func();
+
+       if (input_tf == NULL)
+               return -ENOMEM;
+
+       input_tf->type = TF_TYPE_PREDEFINED;
+       input_tf->tf = TRANSFER_FUNCTION_SRGB;
+
+       surface->in_transfer_func = input_tf;
+
+       /* In case of gamma set, update gamma value */
+       if (crtc_state->gamma_lut)
+               fill_gamma_from_crtc_state(crtc_state, surface);
+
+       return ret;
+}
+
+/*****************************************************************************/
+
+struct amdgpu_connector *aconnector_from_drm_crtc_id(
+               const struct drm_crtc *crtc)
+{
+       struct drm_device *dev = crtc->dev;
+       struct drm_connector *connector;
+       struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
+       struct amdgpu_connector *aconnector;
+
+       list_for_each_entry(connector,
+                       &dev->mode_config.connector_list, head) {
+
+               aconnector = to_amdgpu_connector(connector);
+
+               if (aconnector->base.state->crtc != &acrtc->base)
+                       continue;
+
+               /* Found the connector */
+               return aconnector;
+       }
+
+       /* If we get here, not found. */
+       return NULL;
+}
+
+static void update_stream_scaling_settings(
+               const struct drm_display_mode *mode,
+               const struct dm_connector_state *dm_state,
+               struct dc_stream *stream)
+{
+       enum amdgpu_rmx_type rmx_type;
+
+       struct rect src = { 0 }; /* viewport in composition space*/
+       struct rect dst = { 0 }; /* stream addressable area */
+
+       /* no mode. nothing to be done */
+       if (!mode)
+               return;
+
+       /* Full screen scaling by default */
+       src.width = mode->hdisplay;
+       src.height = mode->vdisplay;
+       dst.width = stream->timing.h_addressable;
+       dst.height = stream->timing.v_addressable;
+
+       rmx_type = dm_state->scaling;
+       if (rmx_type == RMX_ASPECT || rmx_type == RMX_OFF) {
+               if (src.width * dst.height <
+                               src.height * dst.width) {
+                       /* height needs less upscaling/more downscaling */
+                       dst.width = src.width *
+                                       dst.height / src.height;
+               } else {
+                       /* width needs less upscaling/more downscaling */
+                       dst.height = src.height *
+                                       dst.width / src.width;
+               }
+       } else if (rmx_type == RMX_CENTER) {
+               dst = src;
+       }
+
+       dst.x = (stream->timing.h_addressable - dst.width) / 2;
+       dst.y = (stream->timing.v_addressable - dst.height) / 2;
+
+       if (dm_state->underscan_enable) {
+               dst.x += dm_state->underscan_hborder / 2;
+               dst.y += dm_state->underscan_vborder / 2;
+               dst.width -= dm_state->underscan_hborder;
+               dst.height -= dm_state->underscan_vborder;
+       }
+
+       stream->src = src;
+       stream->dst = dst;
+
+       DRM_DEBUG_KMS("Destination Rectangle x:%d  y:%d  width:%d  height:%d\n",
+                       dst.x, dst.y, dst.width, dst.height);
+
+}
+
+static enum dc_color_depth convert_color_depth_from_display_info(
+               const struct drm_connector *connector)
+{
+       uint32_t bpc = connector->display_info.bpc;
+
+       /* Limited color depth to 8bit
+        * TODO: Still need to handle deep color*/
+       if (bpc > 8)
+               bpc = 8;
+
+       switch (bpc) {
+       case 0:
+               /* Temporary Work around, DRM don't parse color depth for
+                * EDID revision before 1.4
+                * TODO: Fix edid parsing
+                */
+               return COLOR_DEPTH_888;
+       case 6:
+               return COLOR_DEPTH_666;
+       case 8:
+               return COLOR_DEPTH_888;
+       case 10:
+               return COLOR_DEPTH_101010;
+       case 12:
+               return COLOR_DEPTH_121212;
+       case 14:
+               return COLOR_DEPTH_141414;
+       case 16:
+               return COLOR_DEPTH_161616;
+       default:
+               return COLOR_DEPTH_UNDEFINED;
+       }
+}
+
+static enum dc_aspect_ratio get_aspect_ratio(
+               const struct drm_display_mode *mode_in)
+{
+       int32_t width = mode_in->crtc_hdisplay * 9;
+       int32_t height = mode_in->crtc_vdisplay * 16;
+       if ((width - height) < 10 && (width - height) > -10)
+               return ASPECT_RATIO_16_9;
+       else
+               return ASPECT_RATIO_4_3;
+}
+
+static enum dc_color_space get_output_color_space(
+                               const struct dc_crtc_timing *dc_crtc_timing)
+{
+       enum dc_color_space color_space = COLOR_SPACE_SRGB;
+
+       switch (dc_crtc_timing->pixel_encoding) {
+       case PIXEL_ENCODING_YCBCR422:
+       case PIXEL_ENCODING_YCBCR444:
+       case PIXEL_ENCODING_YCBCR420:
+       {
+               /*
+                * 27030khz is the separation point between HDTV and SDTV
+                * according to HDMI spec, we use YCbCr709 and YCbCr601
+                * respectively
+                */
+               if (dc_crtc_timing->pix_clk_khz > 27030) {
+                       if (dc_crtc_timing->flags.Y_ONLY)
+                               color_space =
+                                       COLOR_SPACE_YCBCR709_LIMITED;
+                       else
+                               color_space = COLOR_SPACE_YCBCR709;
+               } else {
+                       if (dc_crtc_timing->flags.Y_ONLY)
+                               color_space =
+                                       COLOR_SPACE_YCBCR601_LIMITED;
+                       else
+                               color_space = COLOR_SPACE_YCBCR601;
+               }
+
+       }
+       break;
+       case PIXEL_ENCODING_RGB:
+               color_space = COLOR_SPACE_SRGB;
+               break;
+
+       default:
+               WARN_ON(1);
+               break;
+       }
+
+       return color_space;
+}
+
+/*****************************************************************************/
+
+static void fill_stream_properties_from_drm_display_mode(
+       struct dc_stream *stream,
+       const struct drm_display_mode *mode_in,
+       const struct drm_connector *connector)
+{
+       struct dc_crtc_timing *timing_out = &stream->timing;
+       memset(timing_out, 0, sizeof(struct dc_crtc_timing));
+
+       timing_out->h_border_left = 0;
+       timing_out->h_border_right = 0;
+       timing_out->v_border_top = 0;
+       timing_out->v_border_bottom = 0;
+       /* TODO: un-hardcode */
+
+       if ((connector->display_info.color_formats & DRM_COLOR_FORMAT_YCRCB444)
+                       && stream->sink->sink_signal == SIGNAL_TYPE_HDMI_TYPE_A)
+               timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR444;
+       else
+               timing_out->pixel_encoding = PIXEL_ENCODING_RGB;
+
+       timing_out->timing_3d_format = TIMING_3D_FORMAT_NONE;
+       timing_out->display_color_depth = convert_color_depth_from_display_info(
+                       connector);
+       timing_out->scan_type = SCANNING_TYPE_NODATA;
+       timing_out->hdmi_vic = 0;
+       timing_out->vic = drm_match_cea_mode(mode_in);
+
+       timing_out->h_addressable = mode_in->crtc_hdisplay;
+       timing_out->h_total = mode_in->crtc_htotal;
+       timing_out->h_sync_width =
+               mode_in->crtc_hsync_end - mode_in->crtc_hsync_start;
+       timing_out->h_front_porch =
+               mode_in->crtc_hsync_start - mode_in->crtc_hdisplay;
+       timing_out->v_total = mode_in->crtc_vtotal;
+       timing_out->v_addressable = mode_in->crtc_vdisplay;
+       timing_out->v_front_porch =
+               mode_in->crtc_vsync_start - mode_in->crtc_vdisplay;
+       timing_out->v_sync_width =
+               mode_in->crtc_vsync_end - mode_in->crtc_vsync_start;
+       timing_out->pix_clk_khz = mode_in->crtc_clock;
+       timing_out->aspect_ratio = get_aspect_ratio(mode_in);
+       if (mode_in->flags & DRM_MODE_FLAG_PHSYNC)
+               timing_out->flags.HSYNC_POSITIVE_POLARITY = 1;
+       if (mode_in->flags & DRM_MODE_FLAG_PVSYNC)
+               timing_out->flags.VSYNC_POSITIVE_POLARITY = 1;
+
+       stream->output_color_space = get_output_color_space(timing_out);
+
+       {
+               struct dc_transfer_func *tf = dc_create_transfer_func();
+               tf->type = TF_TYPE_PREDEFINED;
+               tf->tf = TRANSFER_FUNCTION_SRGB;
+               stream->out_transfer_func = tf;
+       }
+}
+
+static void fill_audio_info(
+       struct audio_info *audio_info,
+       const struct drm_connector *drm_connector,
+       const struct dc_sink *dc_sink)
+{
+       int i = 0;
+       int cea_revision = 0;
+       const struct dc_edid_caps *edid_caps = &dc_sink->edid_caps;
+
+       audio_info->manufacture_id = edid_caps->manufacturer_id;
+       audio_info->product_id = edid_caps->product_id;
+
+       cea_revision = drm_connector->display_info.cea_rev;
+
+       while (i < AUDIO_INFO_DISPLAY_NAME_SIZE_IN_CHARS &&
+               edid_caps->display_name[i]) {
+               audio_info->display_name[i] = edid_caps->display_name[i];
+               i++;
+       }
+
+       if(cea_revision >= 3) {
+               audio_info->mode_count = edid_caps->audio_mode_count;
+
+               for (i = 0; i < audio_info->mode_count; ++i) {
+                       audio_info->modes[i].format_code =
+                                       (enum audio_format_code)
+                                       (edid_caps->audio_modes[i].format_code);
+                       audio_info->modes[i].channel_count =
+                                       edid_caps->audio_modes[i].channel_count;
+                       audio_info->modes[i].sample_rates.all =
+                                       edid_caps->audio_modes[i].sample_rate;
+                       audio_info->modes[i].sample_size =
+                                       edid_caps->audio_modes[i].sample_size;
+               }
+       }
+
+       audio_info->flags.all = edid_caps->speaker_flags;
+
+       /* TODO: We only check for the progressive mode, check for interlace mode too */
+       if(drm_connector->latency_present[0]) {
+               audio_info->video_latency = drm_connector->video_latency[0];
+               audio_info->audio_latency = drm_connector->audio_latency[0];
+       }
+
+       /* TODO: For DP, video and audio latency should be calculated from DPCD caps */
+
+}
+
+static void copy_crtc_timing_for_drm_display_mode(
+               const struct drm_display_mode *src_mode,
+               struct drm_display_mode *dst_mode)
+{
+       dst_mode->crtc_hdisplay = src_mode->crtc_hdisplay;
+       dst_mode->crtc_vdisplay = src_mode->crtc_vdisplay;
+       dst_mode->crtc_clock = src_mode->crtc_clock;
+       dst_mode->crtc_hblank_start = src_mode->crtc_hblank_start;
+       dst_mode->crtc_hblank_end = src_mode->crtc_hblank_end;
+       dst_mode->crtc_hsync_start=  src_mode->crtc_hsync_start;
+       dst_mode->crtc_hsync_end = src_mode->crtc_hsync_end;
+       dst_mode->crtc_htotal = src_mode->crtc_htotal;
+       dst_mode->crtc_hskew = src_mode->crtc_hskew;
+       dst_mode->crtc_vblank_start = src_mode->crtc_vblank_start;
+       dst_mode->crtc_vblank_end = src_mode->crtc_vblank_end;
+       dst_mode->crtc_vsync_start = src_mode->crtc_vsync_start;
+       dst_mode->crtc_vsync_end = src_mode->crtc_vsync_end;
+       dst_mode->crtc_vtotal = src_mode->crtc_vtotal;
+}
+
+static void decide_crtc_timing_for_drm_display_mode(
+               struct drm_display_mode *drm_mode,
+               const struct drm_display_mode *native_mode,
+               bool scale_enabled)
+{
+       if (scale_enabled) {
+               copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
+       } else if (native_mode->clock == drm_mode->clock &&
+                       native_mode->htotal == drm_mode->htotal &&
+                       native_mode->vtotal == drm_mode->vtotal) {
+               copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
+       } else {
+               /* no scaling nor amdgpu inserted, no need to patch */
+       }
+}
+
+static struct dc_stream *create_stream_for_sink(
+               struct amdgpu_connector *aconnector,
+               const struct drm_display_mode *drm_mode,
+               const struct dm_connector_state *dm_state)
+{
+       struct drm_display_mode *preferred_mode = NULL;
+       const struct drm_connector *drm_connector;
+       struct dc_stream *stream = NULL;
+       struct drm_display_mode mode = *drm_mode;
+       bool native_mode_found = false;
+
+       if (NULL == aconnector) {
+               DRM_ERROR("aconnector is NULL!\n");
+               goto drm_connector_null;
+       }
+
+       if (NULL == dm_state) {
+               DRM_ERROR("dm_state is NULL!\n");
+               goto dm_state_null;
+       }
 
+       drm_connector = &aconnector->base;
+       stream = dc_create_stream_for_sink(aconnector->dc_sink);
 
+       if (NULL == stream) {
+               DRM_ERROR("Failed to create stream for sink!\n");
+               goto stream_create_fail;
+       }
+
+       list_for_each_entry(preferred_mode, &aconnector->base.modes, head) {
+               /* Search for preferred mode */
+               if (preferred_mode->type & DRM_MODE_TYPE_PREFERRED) {
+                       native_mode_found = true;
+                       break;
+               }
+       }
+       if (!native_mode_found)
+               preferred_mode = list_first_entry_or_null(
+                               &aconnector->base.modes,
+                               struct drm_display_mode,
+                               head);
+
+       if (NULL == preferred_mode) {
+               /* This may not be an error, the use case is when we we have no
+                * usermode calls to reset and set mode upon hotplug. In this
+                * case, we call set mode ourselves to restore the previous mode
+                * and the modelist may not be filled in in time.
+                */
+               DRM_INFO("No preferred mode found\n");
+       } else {
+               decide_crtc_timing_for_drm_display_mode(
+                               &mode, preferred_mode,
+                               dm_state->scaling != RMX_OFF);
+       }
+
+       fill_stream_properties_from_drm_display_mode(stream,
+                       &mode, &aconnector->base);
+       update_stream_scaling_settings(&mode, dm_state, stream);
+
+       fill_audio_info(
+               &stream->audio_info,
+               drm_connector,
+               aconnector->dc_sink);
+
+stream_create_fail:
+dm_state_null:
+drm_connector_null:
+       return stream;
+}
+
+void amdgpu_dm_crtc_destroy(struct drm_crtc *crtc)
+{
+       drm_crtc_cleanup(crtc);
+       kfree(crtc);
+}
+
+static void dm_crtc_destroy_state(struct drm_crtc *crtc,
+                                          struct drm_crtc_state *state)
+{
+       struct dm_crtc_state *cur = to_dm_crtc_state(state);
+
+       /* TODO Destroy dc_stream objects are stream object is flattened */
+       if (cur->stream)
+               dc_stream_release(cur->stream);
+
+
+       __drm_atomic_helper_crtc_destroy_state(state);
+
+
+       kfree(state);
+}
+
+static void dm_crtc_reset_state(struct drm_crtc *crtc)
+{
+       struct dm_crtc_state *state;
+
+       if (crtc->state)
+               dm_crtc_destroy_state(crtc, crtc->state);
+
+       state = kzalloc(sizeof(*state), GFP_KERNEL);
+       if (WARN_ON(!state))
+               return;
+
+       crtc->state = &state->base;
+       crtc->state->crtc = crtc;
+
+}
+
+static struct drm_crtc_state *
+dm_crtc_duplicate_state(struct drm_crtc *crtc)
+{
+       struct dm_crtc_state *state, *cur;
+
+       cur = to_dm_crtc_state(crtc->state);
+
+       if (WARN_ON(!crtc->state))
+               return NULL;
+
+       state = dm_alloc(sizeof(*state));
+
+       __drm_atomic_helper_crtc_duplicate_state(crtc, &state->base);
+
+       if (cur->stream) {
+               state->stream = cur->stream;
+               dc_stream_retain(state->stream);
+       }
+
+       /* TODO Duplicate dc_stream after objects are stream object is flattened */
+
+       return &state->base;
+}
+
+/* Implemented only the options currently availible for the driver */
+static const struct drm_crtc_funcs amdgpu_dm_crtc_funcs = {
+       .reset = dm_crtc_reset_state,
+       .destroy = amdgpu_dm_crtc_destroy,
+       .gamma_set = drm_atomic_helper_legacy_gamma_set,
+       .set_config = drm_atomic_helper_set_config,
+       .page_flip = drm_atomic_helper_page_flip,
+       .atomic_duplicate_state = dm_crtc_duplicate_state,
+       .atomic_destroy_state = dm_crtc_destroy_state,
+};
+
+static enum drm_connector_status
+amdgpu_dm_connector_detect(struct drm_connector *connector, bool force)
+{
+       bool connected;
+       struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
+
+       /* Notes:
+        * 1. This interface is NOT called in context of HPD irq.
+        * 2. This interface *is called* in context of user-mode ioctl. Which
+        * makes it a bad place for *any* MST-related activit. */
+
+       if (aconnector->base.force == DRM_FORCE_UNSPECIFIED)
+               connected = (aconnector->dc_sink != NULL);
+       else
+               connected = (aconnector->base.force == DRM_FORCE_ON);
+
+       return (connected ? connector_status_connected :
+                       connector_status_disconnected);
+}
+
+int amdgpu_dm_connector_atomic_set_property(
+       struct drm_connector *connector,
+       struct drm_connector_state *connector_state,
+       struct drm_property *property,
+       uint64_t val)
+{
+       struct drm_device *dev = connector->dev;
+       struct amdgpu_device *adev = dev->dev_private;
+       struct dm_connector_state *dm_old_state =
+               to_dm_connector_state(connector->state);
+       struct dm_connector_state *dm_new_state =
+               to_dm_connector_state(connector_state);
+
+       int ret = -EINVAL;
+
+       if (property == dev->mode_config.scaling_mode_property) {
+               enum amdgpu_rmx_type rmx_type;
+
+               switch (val) {
+               case DRM_MODE_SCALE_CENTER:
+                       rmx_type = RMX_CENTER;
+                       break;
+               case DRM_MODE_SCALE_ASPECT:
+                       rmx_type = RMX_ASPECT;
+                       break;
+               case DRM_MODE_SCALE_FULLSCREEN:
+                       rmx_type = RMX_FULL;
+                       break;
+               case DRM_MODE_SCALE_NONE:
+               default:
+                       rmx_type = RMX_OFF;
+                       break;
+               }
+
+               if (dm_old_state->scaling == rmx_type)
+                       return 0;
+
+               dm_new_state->scaling = rmx_type;
+               ret = 0;
+       } else if (property == adev->mode_info.underscan_hborder_property) {
+               dm_new_state->underscan_hborder = val;
+               ret = 0;
+       } else if (property == adev->mode_info.underscan_vborder_property) {
+               dm_new_state->underscan_vborder = val;
+               ret = 0;
+       } else if (property == adev->mode_info.underscan_property) {
+               dm_new_state->underscan_enable = val;
+               ret = 0;
+       }
+
+       return ret;
+}
+
+int amdgpu_dm_connector_atomic_get_property(
+       struct drm_connector *connector,
+       const struct drm_connector_state *state,
+       struct drm_property *property,
+       uint64_t *val)
+{
+       struct drm_device *dev = connector->dev;
+       struct amdgpu_device *adev = dev->dev_private;
+       struct dm_connector_state *dm_state =
+               to_dm_connector_state(state);
+       int ret = -EINVAL;
+
+       if (property == dev->mode_config.scaling_mode_property) {
+               switch (dm_state->scaling) {
+               case RMX_CENTER:
+                       *val = DRM_MODE_SCALE_CENTER;
+                       break;
+               case RMX_ASPECT:
+                       *val = DRM_MODE_SCALE_ASPECT;
+                       break;
+               case RMX_FULL:
+                       *val = DRM_MODE_SCALE_FULLSCREEN;
+                       break;
+               case RMX_OFF:
+               default:
+                       *val = DRM_MODE_SCALE_NONE;
+                       break;
+               }
+               ret = 0;
+       } else if (property == adev->mode_info.underscan_hborder_property) {
+               *val = dm_state->underscan_hborder;
+               ret = 0;
+       } else if (property == adev->mode_info.underscan_vborder_property) {
+               *val = dm_state->underscan_vborder;
+               ret = 0;
+       } else if (property == adev->mode_info.underscan_property) {
+               *val = dm_state->underscan_enable;
+               ret = 0;
+       }
+       return ret;
+}
+
+void amdgpu_dm_connector_destroy(struct drm_connector *connector)
+{
+       struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
+       const struct dc_link *link = aconnector->dc_link;
+       struct amdgpu_device *adev = connector->dev->dev_private;
+       struct amdgpu_display_manager *dm = &adev->dm;
+#if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) ||\
+       defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE)
+
+       if (link->connector_signal & (SIGNAL_TYPE_EDP | SIGNAL_TYPE_LVDS)) {
+               amdgpu_dm_register_backlight_device(dm);
+
+               if (dm->backlight_dev) {
+                       backlight_device_unregister(dm->backlight_dev);
+                       dm->backlight_dev = NULL;
+               }
+
+       }
+#endif
+       drm_connector_unregister(connector);
+       drm_connector_cleanup(connector);
+       kfree(connector);
+}
+
+void amdgpu_dm_connector_funcs_reset(struct drm_connector *connector)
+{
+       struct dm_connector_state *state =
+               to_dm_connector_state(connector->state);
+
+       kfree(state);
+
+       state = kzalloc(sizeof(*state), GFP_KERNEL);
+
+       if (state) {
+               state->scaling = RMX_OFF;
+               state->underscan_enable = false;
+               state->underscan_hborder = 0;
+               state->underscan_vborder = 0;
+
+               connector->state = &state->base;
+               connector->state->connector = connector;
+       }
+}
+
+struct drm_connector_state *amdgpu_dm_connector_atomic_duplicate_state(
+       struct drm_connector *connector)
+{
+       struct dm_connector_state *state =
+               to_dm_connector_state(connector->state);
+
+       struct dm_connector_state *new_state =
+                       kmemdup(state, sizeof(*state), GFP_KERNEL);
+
+       if (new_state) {
+               __drm_atomic_helper_connector_duplicate_state(connector,
+                                                                     &new_state->base);
+               return &new_state->base;
+       }
+
+       return NULL;
+}
+
+static const struct drm_connector_funcs amdgpu_dm_connector_funcs = {
+       .reset = amdgpu_dm_connector_funcs_reset,
+       .detect = amdgpu_dm_connector_detect,
+       .fill_modes = drm_helper_probe_single_connector_modes,
+       .destroy = amdgpu_dm_connector_destroy,
+       .atomic_duplicate_state = amdgpu_dm_connector_atomic_duplicate_state,
+       .atomic_destroy_state = drm_atomic_helper_connector_destroy_state,
+       .atomic_set_property = amdgpu_dm_connector_atomic_set_property,
+       .atomic_get_property = amdgpu_dm_connector_atomic_get_property
+};
+
+static struct drm_encoder *best_encoder(struct drm_connector *connector)
+{
+       int enc_id = connector->encoder_ids[0];
+       struct drm_mode_object *obj;
+       struct drm_encoder *encoder;
+
+       DRM_DEBUG_KMS("Finding the best encoder\n");
+
+       /* pick the encoder ids */
+       if (enc_id) {
+               obj = drm_mode_object_find(connector->dev, enc_id, DRM_MODE_OBJECT_ENCODER);
+               if (!obj) {
+                       DRM_ERROR("Couldn't find a matching encoder for our connector\n");
+                       return NULL;
+               }
+               encoder = obj_to_encoder(obj);
+               return encoder;
+       }
+       DRM_ERROR("No encoder id\n");
+       return NULL;
+}
+
+static int get_modes(struct drm_connector *connector)
+{
+       return amdgpu_dm_connector_get_modes(connector);
+}
+
+static void create_eml_sink(struct amdgpu_connector *aconnector)
+{
+       struct dc_sink_init_data init_params = {
+                       .link = aconnector->dc_link,
+                       .sink_signal = SIGNAL_TYPE_VIRTUAL
+       };
+       struct edid *edid = (struct edid *) aconnector->base.edid_blob_ptr->data;
+
+       if (!aconnector->base.edid_blob_ptr ||
+               !aconnector->base.edid_blob_ptr->data) {
+               DRM_ERROR("No EDID firmware found on connector: %s ,forcing to OFF!\n",
+                               aconnector->base.name);
+
+               aconnector->base.force = DRM_FORCE_OFF;
+               aconnector->base.override_edid = false;
+               return;
+       }
+
+       aconnector->edid = edid;
+
+       aconnector->dc_em_sink = dc_link_add_remote_sink(
+               aconnector->dc_link,
+               (uint8_t *)edid,
+               (edid->extensions + 1) * EDID_LENGTH,
+               &init_params);
+
+       if (aconnector->base.force
+                                       == DRM_FORCE_ON)
+               aconnector->dc_sink = aconnector->dc_link->local_sink ?
+               aconnector->dc_link->local_sink :
+               aconnector->dc_em_sink;
+}
+
+static void handle_edid_mgmt(struct amdgpu_connector *aconnector)
+{
+       struct dc_link *link = (struct dc_link *)aconnector->dc_link;
+
+       /* In case of headless boot with force on for DP managed connector
+        * Those settings have to be != 0 to get initial modeset
+        */
+       if (link->connector_signal == SIGNAL_TYPE_DISPLAY_PORT) {
+               link->verified_link_cap.lane_count = LANE_COUNT_FOUR;
+               link->verified_link_cap.link_rate = LINK_RATE_HIGH2;
+       }
+
+
+       aconnector->base.override_edid = true;
+       create_eml_sink(aconnector);
+}
+
+int amdgpu_dm_connector_mode_valid(
+               struct drm_connector *connector,
+               struct drm_display_mode *mode)
+{
+       int result = MODE_ERROR;
+       struct dc_sink *dc_sink;
+       struct amdgpu_device *adev = connector->dev->dev_private;
+       /* TODO: Unhardcode stream count */
+       struct dc_stream *stream;
+       struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
+
+       if ((mode->flags & DRM_MODE_FLAG_INTERLACE) ||
+                       (mode->flags & DRM_MODE_FLAG_DBLSCAN))
+               return result;
+
+       /* Only run this the first time mode_valid is called to initilialize
+        * EDID mgmt
+        */
+       if (aconnector->base.force != DRM_FORCE_UNSPECIFIED &&
+               !aconnector->dc_em_sink)
+               handle_edid_mgmt(aconnector);
+
+       dc_sink = to_amdgpu_connector(connector)->dc_sink;
+
+       if (NULL == dc_sink) {
+               DRM_ERROR("dc_sink is NULL!\n");
+               goto fail;
+       }
+
+       stream = dc_create_stream_for_sink(dc_sink);
+       if (NULL == stream) {
+               DRM_ERROR("Failed to create stream for sink!\n");
+               goto fail;
+       }
+
+       drm_mode_set_crtcinfo(mode, 0);
+       fill_stream_properties_from_drm_display_mode(stream, mode, connector);
+
+       stream->src.width = mode->hdisplay;
+       stream->src.height = mode->vdisplay;
+       stream->dst = stream->src;
+
+       if (dc_validate_stream(adev->dm.dc, stream))
+               result = MODE_OK;
+
+       dc_stream_release(stream);
+
+fail:
+       /* TODO: error handling*/
+       return result;
+}
+
+static const struct drm_connector_helper_funcs
+amdgpu_dm_connector_helper_funcs = {
+       /*
+       * If hotplug a second bigger display in FB Con mode, bigger resolution
+       * modes will be filtered by drm_mode_validate_size(), and those modes
+       * is missing after user start lightdm. So we need to renew modes list.
+       * in get_modes call back, not just return the modes count
+       */
+       .get_modes = get_modes,
+       .mode_valid = amdgpu_dm_connector_mode_valid,
+       .best_encoder = best_encoder
+};
+
+static void dm_crtc_helper_disable(struct drm_crtc *crtc)
+{
+}
+
+static int dm_crtc_helper_atomic_check(
+       struct drm_crtc *crtc,
+       struct drm_crtc_state *state)
+{
+       struct amdgpu_device *adev = crtc->dev->dev_private;
+       struct dc *dc = adev->dm.dc;
+       struct dm_crtc_state *dm_crtc_state = to_dm_crtc_state(state);
+       int ret = -EINVAL;
+
+       if (unlikely(!dm_crtc_state->stream && modeset_required(state))) {
+               WARN_ON(1);
+               return ret;
+       }
+
+       /* In some use cases, like reset, no stream  is attached */
+       if (!dm_crtc_state->stream)
+               return 0;
+
+       if (dc_validate_stream(dc, dm_crtc_state->stream))
+               return 0;
+
+       return ret;
+}
+
+static bool dm_crtc_helper_mode_fixup(
+       struct drm_crtc *crtc,
+       const struct drm_display_mode *mode,
+       struct drm_display_mode *adjusted_mode)
+{
+       return true;
+}
+
+static const struct drm_crtc_helper_funcs amdgpu_dm_crtc_helper_funcs = {
+       .disable = dm_crtc_helper_disable,
+       .atomic_check = dm_crtc_helper_atomic_check,
+       .mode_fixup = dm_crtc_helper_mode_fixup
+};
+
+static void dm_encoder_helper_disable(struct drm_encoder *encoder)
+{
+
+}
+
+static int dm_encoder_helper_atomic_check(
+       struct drm_encoder *encoder,
+       struct drm_crtc_state *crtc_state,
+       struct drm_connector_state *conn_state)
+{
+       return 0;
+}
+
+const struct drm_encoder_helper_funcs amdgpu_dm_encoder_helper_funcs = {
+       .disable = dm_encoder_helper_disable,
+       .atomic_check = dm_encoder_helper_atomic_check
+};
+
+static void dm_drm_plane_reset(struct drm_plane *plane)
+{
+       struct dm_plane_state *amdgpu_state = NULL;
+
+       if (plane->state)
+               plane->funcs->atomic_destroy_state(plane, plane->state);
+
+       amdgpu_state = kzalloc(sizeof(*amdgpu_state), GFP_KERNEL);
+
+       if (amdgpu_state) {
+               plane->state = &amdgpu_state->base;
+               plane->state->plane = plane;
+               plane->state->rotation = DRM_MODE_ROTATE_0;
+       } else
+               WARN_ON(1);
+}
+
+static struct drm_plane_state *
+dm_drm_plane_duplicate_state(struct drm_plane *plane)
+{
+       struct dm_plane_state *dm_plane_state, *old_dm_plane_state;
+
+       old_dm_plane_state = to_dm_plane_state(plane->state);
+       dm_plane_state = kzalloc(sizeof(*dm_plane_state), GFP_KERNEL);
+       if (!dm_plane_state)
+               return NULL;
+
+       __drm_atomic_helper_plane_duplicate_state(plane, &dm_plane_state->base);
+
+       if (old_dm_plane_state->surface) {
+               dm_plane_state->surface = old_dm_plane_state->surface;
+               dc_surface_retain(dm_plane_state->surface);
+       }
+
+       return &dm_plane_state->base;
+}
+
+void dm_drm_plane_destroy_state(struct drm_plane *plane,
+                                          struct drm_plane_state *state)
+{
+       struct dm_plane_state *dm_plane_state = to_dm_plane_state(state);
+
+       if (dm_plane_state->surface)
+               dc_surface_release(dm_plane_state->surface);
+
+       __drm_atomic_helper_plane_destroy_state(state);
+       kfree(dm_plane_state);
+}
+
+static const struct drm_plane_funcs dm_plane_funcs = {
+       .update_plane   = drm_atomic_helper_update_plane,
+       .disable_plane  = drm_atomic_helper_disable_plane,
+       .destroy        = drm_plane_cleanup,
+       .reset = dm_drm_plane_reset,
+       .atomic_duplicate_state = dm_drm_plane_duplicate_state,
+       .atomic_destroy_state = dm_drm_plane_destroy_state,
+};
+
+static int dm_plane_helper_prepare_fb(
+       struct drm_plane *plane,
+       struct drm_plane_state *new_state)
+{
+       struct amdgpu_framebuffer *afb;
+       struct drm_gem_object *obj;
+       struct amdgpu_bo *rbo;
+       int r;
+       struct dm_plane_state *dm_plane_state_new, *dm_plane_state_old;
+       unsigned int awidth;
+
+       dm_plane_state_old = to_dm_plane_state(plane->state);
+       dm_plane_state_new = to_dm_plane_state(new_state);
+
+       if (!new_state->fb) {
+               DRM_DEBUG_KMS("No FB bound\n");
+               return 0;
+       }
+
+       afb = to_amdgpu_framebuffer(new_state->fb);
+
+       obj = afb->obj;
+       rbo = gem_to_amdgpu_bo(obj);
+       r = amdgpu_bo_reserve(rbo, false);
+       if (unlikely(r != 0))
+               return r;
+
+       r = amdgpu_bo_pin(rbo, AMDGPU_GEM_DOMAIN_VRAM, &afb->address);
+
+
+       amdgpu_bo_unreserve(rbo);
+
+       if (unlikely(r != 0)) {
+               DRM_ERROR("Failed to pin framebuffer\n");
+               return r;
+       }
+
+       amdgpu_bo_ref(rbo);
+
+       if (dm_plane_state_new->surface &&
+                       dm_plane_state_old->surface != dm_plane_state_new->surface) {
+               struct dc_surface *surface = dm_plane_state_new->surface;
+
+               if (surface->format < SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) {
+                       surface->address.grph.addr.low_part = lower_32_bits(afb->address);
+                       surface->address.grph.addr.high_part = upper_32_bits(afb->address);
+               } else {
+                       awidth = ALIGN(new_state->fb->width, 64);
+                       surface->address.video_progressive.luma_addr.low_part
+                                                       = lower_32_bits(afb->address);
+                       surface->address.video_progressive.chroma_addr.low_part
+                                                       = lower_32_bits(afb->address) +
+                                                               (awidth * new_state->fb->height);
+               }
+       }
+
+       /* It's a hack for s3 since in 4.9 kernel filter out cursor buffer
+        * prepare and cleanup in drm_atomic_helper_prepare_planes
+        * and drm_atomic_helper_cleanup_planes because fb doens't in s3.
+        * IN 4.10 kernel this code should be removed and amdgpu_device_suspend
+        * code touching fram buffers should be avoided for DC.
+        */
+       if (plane->type == DRM_PLANE_TYPE_CURSOR) {
+               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(new_state->crtc);
+
+               acrtc->cursor_bo = obj;
+       }
+       return 0;
+}
+
+static void dm_plane_helper_cleanup_fb(
+       struct drm_plane *plane,
+       struct drm_plane_state *old_state)
+{
+       struct amdgpu_bo *rbo;
+       struct amdgpu_framebuffer *afb;
+       int r;
+
+       if (!old_state->fb)
+               return;
+
+       afb = to_amdgpu_framebuffer(old_state->fb);
+       rbo = gem_to_amdgpu_bo(afb->obj);
+       r = amdgpu_bo_reserve(rbo, false);
+       if (unlikely(r)) {
+               DRM_ERROR("failed to reserve rbo before unpin\n");
+               return;
+       } else {
+               amdgpu_bo_unpin(rbo);
+               amdgpu_bo_unreserve(rbo);
+               amdgpu_bo_unref(&rbo);
+       };
+}
+
+int dm_create_validation_set_for_connector(struct drm_connector *connector,
+               struct drm_display_mode *mode, struct dc_validation_set *val_set)
+{
+       int result = MODE_ERROR;
+       struct dc_sink *dc_sink =
+                       to_amdgpu_connector(connector)->dc_sink;
+       /* TODO: Unhardcode stream count */
+       struct dc_stream *stream;
+
+       if ((mode->flags & DRM_MODE_FLAG_INTERLACE) ||
+                       (mode->flags & DRM_MODE_FLAG_DBLSCAN))
+               return result;
+
+       if (NULL == dc_sink) {
+               DRM_ERROR("dc_sink is NULL!\n");
+               return result;
+       }
+
+       stream = dc_create_stream_for_sink(dc_sink);
+
+       if (NULL == stream) {
+               DRM_ERROR("Failed to create stream for sink!\n");
+               return result;
+       }
+
+       drm_mode_set_crtcinfo(mode, 0);
+
+       fill_stream_properties_from_drm_display_mode(stream, mode, connector);
+
+       val_set->stream = stream;
+
+       stream->src.width = mode->hdisplay;
+       stream->src.height = mode->vdisplay;
+       stream->dst = stream->src;
+
+       return MODE_OK;
+}
+
+static const struct drm_plane_helper_funcs dm_plane_helper_funcs = {
+       .prepare_fb = dm_plane_helper_prepare_fb,
+       .cleanup_fb = dm_plane_helper_cleanup_fb,
+};
+
+/*
+ * TODO: these are currently initialized to rgb formats only.
+ * For future use cases we should either initialize them dynamically based on
+ * plane capabilities, or initialize this array to all formats, so internal drm
+ * check will succeed, and let DC to implement proper check
+ */
+static uint32_t rgb_formats[] = {
+       DRM_FORMAT_RGB888,
+       DRM_FORMAT_XRGB8888,
+       DRM_FORMAT_ARGB8888,
+       DRM_FORMAT_RGBA8888,
+       DRM_FORMAT_XRGB2101010,
+       DRM_FORMAT_XBGR2101010,
+       DRM_FORMAT_ARGB2101010,
+       DRM_FORMAT_ABGR2101010,
+};
+
+static uint32_t yuv_formats[] = {
+       DRM_FORMAT_NV12,
+       DRM_FORMAT_NV21,
+};
+
+static const u32 cursor_formats[] = {
+       DRM_FORMAT_ARGB8888
+};
+
+int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
+                       struct amdgpu_plane *aplane,
+                       unsigned long possible_crtcs)
+{
+       int res = -EPERM;
+
+       switch (aplane->base.type) {
+       case DRM_PLANE_TYPE_PRIMARY:
+               aplane->base.format_default = true;
+
+               res = drm_universal_plane_init(
+                               dm->adev->ddev,
+                               &aplane->base,
+                               possible_crtcs,
+                               &dm_plane_funcs,
+                               rgb_formats,
+                               ARRAY_SIZE(rgb_formats),
+                               NULL, aplane->base.type, NULL);
+               break;
+       case DRM_PLANE_TYPE_OVERLAY:
+               res = drm_universal_plane_init(
+                               dm->adev->ddev,
+                               &aplane->base,
+                               possible_crtcs,
+                               &dm_plane_funcs,
+                               yuv_formats,
+                               ARRAY_SIZE(yuv_formats),
+                               NULL, aplane->base.type, NULL);
+               break;
+       case DRM_PLANE_TYPE_CURSOR:
+               res = drm_universal_plane_init(
+                               dm->adev->ddev,
+                               &aplane->base,
+                               possible_crtcs,
+                               &dm_plane_funcs,
+                               cursor_formats,
+                               ARRAY_SIZE(cursor_formats),
+                               NULL, aplane->base.type, NULL);
+               break;
+       }
+
+       drm_plane_helper_add(&aplane->base, &dm_plane_helper_funcs);
+
+       return res;
+}
+
+int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm,
+                       struct drm_plane *plane,
+                       uint32_t crtc_index)
+{
+       struct amdgpu_crtc *acrtc = NULL;
+       struct amdgpu_plane *cursor_plane;
+
+       int res = -ENOMEM;
+
+       cursor_plane = kzalloc(sizeof(*cursor_plane), GFP_KERNEL);
+       if (!cursor_plane)
+               goto fail;
+
+       cursor_plane->base.type = DRM_PLANE_TYPE_CURSOR;
+       res = amdgpu_dm_plane_init(dm, cursor_plane, 0);
+
+       acrtc = kzalloc(sizeof(struct amdgpu_crtc), GFP_KERNEL);
+       if (!acrtc)
+               goto fail;
+
+       res = drm_crtc_init_with_planes(
+                       dm->ddev,
+                       &acrtc->base,
+                       plane,
+                       &cursor_plane->base,
+                       &amdgpu_dm_crtc_funcs, NULL);
+
+       if (res)
+               goto fail;
+
+       drm_crtc_helper_add(&acrtc->base, &amdgpu_dm_crtc_helper_funcs);
+
+       acrtc->max_cursor_width = dm->adev->dm.dc->caps.max_cursor_size;
+       acrtc->max_cursor_height = dm->adev->dm.dc->caps.max_cursor_size;
+
+       acrtc->crtc_id = crtc_index;
+       acrtc->base.enabled = false;
+
+       dm->adev->mode_info.crtcs[crtc_index] = acrtc;
+       drm_mode_crtc_set_gamma_size(&acrtc->base, 256);
+
+       return 0;
+
+fail:
+       if (acrtc)
+               kfree(acrtc);
+       if (cursor_plane)
+               kfree(cursor_plane);
+       acrtc->crtc_id = -1;
+       return res;
+}
+
+
+static int to_drm_connector_type(enum signal_type st)
+{
+       switch (st) {
+       case SIGNAL_TYPE_HDMI_TYPE_A:
+               return DRM_MODE_CONNECTOR_HDMIA;
+       case SIGNAL_TYPE_EDP:
+               return DRM_MODE_CONNECTOR_eDP;
+       case SIGNAL_TYPE_RGB:
+               return DRM_MODE_CONNECTOR_VGA;
+       case SIGNAL_TYPE_DISPLAY_PORT:
+       case SIGNAL_TYPE_DISPLAY_PORT_MST:
+               return DRM_MODE_CONNECTOR_DisplayPort;
+       case SIGNAL_TYPE_DVI_DUAL_LINK:
+       case SIGNAL_TYPE_DVI_SINGLE_LINK:
+               return DRM_MODE_CONNECTOR_DVID;
+       case SIGNAL_TYPE_VIRTUAL:
+               return DRM_MODE_CONNECTOR_VIRTUAL;
+
+       default:
+               return DRM_MODE_CONNECTOR_Unknown;
+       }
+}
+
+static void amdgpu_dm_get_native_mode(struct drm_connector *connector)
+{
+       const struct drm_connector_helper_funcs *helper =
+               connector->helper_private;
+       struct drm_encoder *encoder;
+       struct amdgpu_encoder *amdgpu_encoder;
+
+       encoder = helper->best_encoder(connector);
+
+       if (encoder == NULL)
+               return;
+
+       amdgpu_encoder = to_amdgpu_encoder(encoder);
+
+       amdgpu_encoder->native_mode.clock = 0;
+
+       if (!list_empty(&connector->probed_modes)) {
+               struct drm_display_mode *preferred_mode = NULL;
+               list_for_each_entry(preferred_mode,
+                               &connector->probed_modes,
+                               head) {
+               if (preferred_mode->type & DRM_MODE_TYPE_PREFERRED) {
+                       amdgpu_encoder->native_mode = *preferred_mode;
+               }
+                       break;
+               }
+
+       }
+}
+
+static struct drm_display_mode *amdgpu_dm_create_common_mode(
+               struct drm_encoder *encoder, char *name,
+               int hdisplay, int vdisplay)
+{
+       struct drm_device *dev = encoder->dev;
+       struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+       struct drm_display_mode *mode = NULL;
+       struct drm_display_mode *native_mode = &amdgpu_encoder->native_mode;
+
+       mode = drm_mode_duplicate(dev, native_mode);
+
+       if(mode == NULL)
+               return NULL;
+
+       mode->hdisplay = hdisplay;
+       mode->vdisplay = vdisplay;
+       mode->type &= ~DRM_MODE_TYPE_PREFERRED;
+       strncpy(mode->name, name, DRM_DISPLAY_MODE_LEN);
+
+       return mode;
+
+}
+
+static void amdgpu_dm_connector_add_common_modes(struct drm_encoder *encoder,
+                                       struct drm_connector *connector)
+{
+       struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+       struct drm_display_mode *mode = NULL;
+       struct drm_display_mode *native_mode = &amdgpu_encoder->native_mode;
+       struct amdgpu_connector *amdgpu_connector =
+                               to_amdgpu_connector(connector);
+       int i;
+       int n;
+       struct mode_size {
+               char name[DRM_DISPLAY_MODE_LEN];
+               int w;
+               int h;
+       }common_modes[] = {
+               {  "640x480",  640,  480},
+               {  "800x600",  800,  600},
+               { "1024x768", 1024,  768},
+               { "1280x720", 1280,  720},
+               { "1280x800", 1280,  800},
+               {"1280x1024", 1280, 1024},
+               { "1440x900", 1440,  900},
+               {"1680x1050", 1680, 1050},
+               {"1600x1200", 1600, 1200},
+               {"1920x1080", 1920, 1080},
+               {"1920x1200", 1920, 1200}
+       };
+
+       n = sizeof(common_modes) / sizeof(common_modes[0]);
+
+       for (i = 0; i < n; i++) {
+               struct drm_display_mode *curmode = NULL;
+               bool mode_existed = false;
+
+               if (common_modes[i].w > native_mode->hdisplay ||
+                       common_modes[i].h > native_mode->vdisplay ||
+                       (common_modes[i].w == native_mode->hdisplay &&
+                       common_modes[i].h == native_mode->vdisplay))
+                               continue;
+
+               list_for_each_entry(curmode, &connector->probed_modes, head) {
+                       if (common_modes[i].w == curmode->hdisplay &&
+                               common_modes[i].h == curmode->vdisplay) {
+                               mode_existed = true;
+                               break;
+                       }
+               }
+
+               if (mode_existed)
+                       continue;
+
+               mode = amdgpu_dm_create_common_mode(encoder,
+                               common_modes[i].name, common_modes[i].w,
+                               common_modes[i].h);
+               drm_mode_probed_add(connector, mode);
+               amdgpu_connector->num_modes++;
+       }
+}
+
+static void amdgpu_dm_connector_ddc_get_modes(
+       struct drm_connector *connector,
+       struct edid *edid)
+{
+       struct amdgpu_connector *amdgpu_connector =
+                       to_amdgpu_connector(connector);
+
+       if (edid) {
+               /* empty probed_modes */
+               INIT_LIST_HEAD(&connector->probed_modes);
+               amdgpu_connector->num_modes =
+                               drm_add_edid_modes(connector, edid);
+
+               drm_edid_to_eld(connector, edid);
+
+               amdgpu_dm_get_native_mode(connector);
+       } else
+               amdgpu_connector->num_modes = 0;
+}
+
+int amdgpu_dm_connector_get_modes(struct drm_connector *connector)
+{
+       const struct drm_connector_helper_funcs *helper =
+                       connector->helper_private;
+       struct amdgpu_connector *amdgpu_connector =
+                       to_amdgpu_connector(connector);
+       struct drm_encoder *encoder;
+       struct edid *edid = amdgpu_connector->edid;
+
+       encoder = helper->best_encoder(connector);
+
+       amdgpu_dm_connector_ddc_get_modes(connector, edid);
+       amdgpu_dm_connector_add_common_modes(encoder, connector);
+       return amdgpu_connector->num_modes;
+}
+
+void amdgpu_dm_connector_init_helper(
+       struct amdgpu_display_manager *dm,
+       struct amdgpu_connector *aconnector,
+       int connector_type,
+       struct dc_link *link,
+       int link_index)
+{
+       struct amdgpu_device *adev = dm->ddev->dev_private;
+
+       aconnector->connector_id = link_index;
+       aconnector->dc_link = link;
+       aconnector->base.interlace_allowed = false;
+       aconnector->base.doublescan_allowed = false;
+       aconnector->base.stereo_allowed = false;
+       aconnector->base.dpms = DRM_MODE_DPMS_OFF;
+       aconnector->hpd.hpd = AMDGPU_HPD_NONE; /* not used */
+
+       mutex_init(&aconnector->hpd_lock);
+
+       /*configure suport HPD hot plug connector_>polled default value is 0
+        * which means HPD hot plug not supported*/
+       switch (connector_type) {
+       case DRM_MODE_CONNECTOR_HDMIA:
+               aconnector->base.polled = DRM_CONNECTOR_POLL_HPD;
+               break;
+       case DRM_MODE_CONNECTOR_DisplayPort:
+               aconnector->base.polled = DRM_CONNECTOR_POLL_HPD;
+               break;
+       case DRM_MODE_CONNECTOR_DVID:
+               aconnector->base.polled = DRM_CONNECTOR_POLL_HPD;
+               break;
+       default:
+               break;
+       }
+
+       drm_object_attach_property(&aconnector->base.base,
+                               dm->ddev->mode_config.scaling_mode_property,
+                               DRM_MODE_SCALE_NONE);
+
+       drm_object_attach_property(&aconnector->base.base,
+                               adev->mode_info.underscan_property,
+                               UNDERSCAN_OFF);
+       drm_object_attach_property(&aconnector->base.base,
+                               adev->mode_info.underscan_hborder_property,
+                               0);
+       drm_object_attach_property(&aconnector->base.base,
+                               adev->mode_info.underscan_vborder_property,
+                               0);
+
+}
+
+int amdgpu_dm_i2c_xfer(struct i2c_adapter *i2c_adap,
+                     struct i2c_msg *msgs, int num)
+{
+       struct amdgpu_i2c_adapter *i2c = i2c_get_adapdata(i2c_adap);
+       struct ddc_service *ddc_service = i2c->ddc_service;
+       struct i2c_command cmd;
+       int i;
+       int result = -EIO;
+
+       cmd.payloads = kzalloc(num * sizeof(struct i2c_payload), GFP_KERNEL);
+
+       if (!cmd.payloads)
+               return result;
+
+       cmd.number_of_payloads = num;
+       cmd.engine = I2C_COMMAND_ENGINE_DEFAULT;
+       cmd.speed = 100;
+
+       for (i = 0; i < num; i++) {
+               cmd.payloads[i].write = !(msgs[i].flags & I2C_M_RD);
+               cmd.payloads[i].address = msgs[i].addr;
+               cmd.payloads[i].length = msgs[i].len;
+               cmd.payloads[i].data = msgs[i].buf;
+       }
+
+       if (dal_i2caux_submit_i2c_command(
+                       ddc_service->ctx->i2caux,
+                       ddc_service->ddc_pin,
+                       &cmd))
+               result = num;
+
+       kfree(cmd.payloads);
+       return result;
+}
+
+u32 amdgpu_dm_i2c_func(struct i2c_adapter *adap)
+{
+       return I2C_FUNC_I2C | I2C_FUNC_SMBUS_EMUL;
+}
+
+static const struct i2c_algorithm amdgpu_dm_i2c_algo = {
+       .master_xfer = amdgpu_dm_i2c_xfer,
+       .functionality = amdgpu_dm_i2c_func,
+};
+
+static struct amdgpu_i2c_adapter *create_i2c(
+               struct ddc_service *ddc_service,
+               int link_index,
+               int *res)
+{
+       struct amdgpu_device *adev = ddc_service->ctx->driver_context;
+       struct amdgpu_i2c_adapter *i2c;
+
+       i2c = kzalloc(sizeof (struct amdgpu_i2c_adapter), GFP_KERNEL);
+       i2c->base.owner = THIS_MODULE;
+       i2c->base.class = I2C_CLASS_DDC;
+       i2c->base.dev.parent = &adev->pdev->dev;
+       i2c->base.algo = &amdgpu_dm_i2c_algo;
+       snprintf(i2c->base.name, sizeof (i2c->base.name), "AMDGPU DM i2c hw bus %d", link_index);
+       i2c_set_adapdata(&i2c->base, i2c);
+       i2c->ddc_service = ddc_service;
+
+       return i2c;
+}
+
+/* Note: this function assumes that dc_link_detect() was called for the
+ * dc_link which will be represented by this aconnector. */
+int amdgpu_dm_connector_init(
+       struct amdgpu_display_manager *dm,
+       struct amdgpu_connector *aconnector,
+       uint32_t link_index,
+       struct amdgpu_encoder *aencoder)
+{
+       int res = 0;
+       int connector_type;
+       struct dc *dc = dm->dc;
+       struct dc_link *link = dc_get_link_at_index(dc, link_index);
+       struct amdgpu_i2c_adapter *i2c;
+       ((struct dc_link *)link)->priv = aconnector;
+
+       DRM_DEBUG_KMS("%s()\n", __func__);
+
+       i2c = create_i2c(link->ddc, link->link_index, &res);
+       aconnector->i2c = i2c;
+       res = i2c_add_adapter(&i2c->base);
+
+       if (res) {
+               DRM_ERROR("Failed to register hw i2c %d\n", link->link_index);
+               goto out_free;
+       }
+
+       connector_type = to_drm_connector_type(link->connector_signal);
+
+       res = drm_connector_init(
+                       dm->ddev,
+                       &aconnector->base,
+                       &amdgpu_dm_connector_funcs,
+                       connector_type);
+
+       if (res) {
+               DRM_ERROR("connector_init failed\n");
+               aconnector->connector_id = -1;
+               goto out_free;
+       }
+
+       drm_connector_helper_add(
+                       &aconnector->base,
+                       &amdgpu_dm_connector_helper_funcs);
+
+       amdgpu_dm_connector_init_helper(
+               dm,
+               aconnector,
+               connector_type,
+               link,
+               link_index);
+
+       drm_mode_connector_attach_encoder(
+               &aconnector->base, &aencoder->base);
+
+       drm_connector_register(&aconnector->base);
+
+       if (connector_type == DRM_MODE_CONNECTOR_DisplayPort
+               || connector_type == DRM_MODE_CONNECTOR_eDP)
+               amdgpu_dm_initialize_dp_connector(dm, aconnector);
+
+#if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) ||\
+       defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE)
+
+       /* NOTE: this currently will create backlight device even if a panel
+        * is not connected to the eDP/LVDS connector.
+        *
+        * This is less than ideal but we don't have sink information at this
+        * stage since detection happens after. We can't do detection earlier
+        * since MST detection needs connectors to be created first.
+        */
+       if (link->connector_signal & (SIGNAL_TYPE_EDP | SIGNAL_TYPE_LVDS)) {
+               /* Event if registration failed, we should continue with
+                * DM initialization because not having a backlight control
+                * is better then a black screen. */
+               amdgpu_dm_register_backlight_device(dm);
+
+               if (dm->backlight_dev)
+                       dm->backlight_link = link;
+       }
+#endif
+
+out_free:
+       if (res) {
+               kfree(i2c);
+               aconnector->i2c = NULL;
+       }
+       return res;
+}
+
+int amdgpu_dm_get_encoder_crtc_mask(struct amdgpu_device *adev)
+{
+       switch (adev->mode_info.num_crtc) {
+       case 1:
+               return 0x1;
+       case 2:
+               return 0x3;
+       case 3:
+               return 0x7;
+       case 4:
+               return 0xf;
+       case 5:
+               return 0x1f;
+       case 6:
+       default:
+               return 0x3f;
+       }
+}
+
+int amdgpu_dm_encoder_init(
+       struct drm_device *dev,
+       struct amdgpu_encoder *aencoder,
+       uint32_t link_index)
+{
+       struct amdgpu_device *adev = dev->dev_private;
+
+       int res = drm_encoder_init(dev,
+                                  &aencoder->base,
+                                  &amdgpu_dm_encoder_funcs,
+                                  DRM_MODE_ENCODER_TMDS,
+                                  NULL);
+
+       aencoder->base.possible_crtcs = amdgpu_dm_get_encoder_crtc_mask(adev);
+
+       if (!res)
+               aencoder->encoder_id = link_index;
+       else
+               aencoder->encoder_id = -1;
+
+       drm_encoder_helper_add(&aencoder->base, &amdgpu_dm_encoder_helper_funcs);
+
+       return res;
+}
+
+static void manage_dm_interrupts(
+       struct amdgpu_device *adev,
+       struct amdgpu_crtc *acrtc,
+       bool enable)
+{
+       /*
+        * this is not correct translation but will work as soon as VBLANK
+        * constant is the same as PFLIP
+        */
+       int irq_type =
+               amdgpu_crtc_idx_to_irq_type(
+                       adev,
+                       acrtc->crtc_id);
+
+       if (enable) {
+               drm_crtc_vblank_on(&acrtc->base);
+               amdgpu_irq_get(
+                       adev,
+                       &adev->pageflip_irq,
+                       irq_type);
+       } else {
+
+               amdgpu_irq_put(
+                       adev,
+                       &adev->pageflip_irq,
+                       irq_type);
+               drm_crtc_vblank_off(&acrtc->base);
+       }
+}
+
+static bool is_scaling_state_different(
+               const struct dm_connector_state *dm_state,
+               const struct dm_connector_state *old_dm_state)
+{
+       if (dm_state->scaling != old_dm_state->scaling)
+               return true;
+       if (!dm_state->underscan_enable && old_dm_state->underscan_enable) {
+               if (old_dm_state->underscan_hborder != 0 && old_dm_state->underscan_vborder != 0)
+                       return true;
+       } else  if (dm_state->underscan_enable && !old_dm_state->underscan_enable) {
+               if (dm_state->underscan_hborder != 0 && dm_state->underscan_vborder != 0)
+                       return true;
+       } else if (dm_state->underscan_hborder != old_dm_state->underscan_hborder
+                               || dm_state->underscan_vborder != old_dm_state->underscan_vborder)
+                       return true;
+       return false;
+}
+
+static void remove_stream(
+               struct amdgpu_device *adev,
+               struct amdgpu_crtc *acrtc,
+               struct dc_stream *stream)
+{
+       /* this is the update mode case */
+       if (adev->dm.freesync_module)
+               mod_freesync_remove_stream(adev->dm.freesync_module, stream);
+
+       acrtc->otg_inst = -1;
+       acrtc->enabled = false;
+}
+
+static void handle_cursor_update(
+               struct drm_plane *plane,
+               struct drm_plane_state *old_plane_state)
+{
+       if (!plane->state->fb && !old_plane_state->fb)
+               return;
+
+       /* Check if it's a cursor on/off update or just cursor move*/
+       if (plane->state->fb == old_plane_state->fb)
+               dm_crtc_cursor_move(
+                               plane->state->crtc,
+                               plane->state->crtc_x,
+                               plane->state->crtc_y);
+       else {
+               struct amdgpu_framebuffer *afb =
+                               to_amdgpu_framebuffer(plane->state->fb);
+               dm_crtc_cursor_set(
+                               (!!plane->state->fb) ?
+                                               plane->state->crtc :
+                                               old_plane_state->crtc,
+                               (!!plane->state->fb) ?
+                                               afb->address :
+                                               0,
+                               plane->state->crtc_w,
+                               plane->state->crtc_h);
+       }
+}
+
+
+static void prepare_flip_isr(struct amdgpu_crtc *acrtc)
+{
+
+       assert_spin_locked(&acrtc->base.dev->event_lock);
+       WARN_ON(acrtc->event);
+
+       acrtc->event = acrtc->base.state->event;
+
+       /* Set the flip status */
+       acrtc->pflip_status = AMDGPU_FLIP_SUBMITTED;
+
+       /* Mark this event as consumed */
+       acrtc->base.state->event = NULL;
+
+       DRM_DEBUG_DRIVER("crtc:%d, pflip_stat:AMDGPU_FLIP_SUBMITTED\n",
+                                                acrtc->crtc_id);
+}
+
+/*
+ * Executes flip
+ *
+ * Waits on all BO's fences and for proper vblank count
+ */
+static void amdgpu_dm_do_flip(
+                               struct drm_crtc *crtc,
+                               struct drm_framebuffer *fb,
+                               uint32_t target)
+{
+       unsigned long flags;
+       uint32_t target_vblank;
+       int r, vpos, hpos;
+       struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
+       struct amdgpu_framebuffer *afb = to_amdgpu_framebuffer(fb);
+       struct amdgpu_bo *abo = gem_to_amdgpu_bo(afb->obj);
+       struct amdgpu_device *adev = crtc->dev->dev_private;
+       bool async_flip = (acrtc->flip_flags & DRM_MODE_PAGE_FLIP_ASYNC) != 0;
+       struct dc_flip_addrs addr = { {0} };
+       struct dc_surface_update surface_updates[1] = { {0} };
+       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state);
+
+
+       /* Prepare wait for target vblank early - before the fence-waits */
+       target_vblank = target - drm_crtc_vblank_count(crtc) +
+                       amdgpu_get_vblank_counter_kms(crtc->dev, acrtc->crtc_id);
+
+       /*TODO This might fail and hence better not used, wait
+        * explicitly on fences instead
+        * and in general should be called for
+        * blocking commit to as per framework helpers
+        * */
+       r = amdgpu_bo_reserve(abo, true);
+       if (unlikely(r != 0)) {
+               DRM_ERROR("failed to reserve buffer before flip\n");
+               WARN_ON(1);
+       }
+
+       /* Wait for all fences on this FB */
+       WARN_ON(reservation_object_wait_timeout_rcu(abo->tbo.resv, true, false,
+                                                                   MAX_SCHEDULE_TIMEOUT) < 0);
+
+       amdgpu_bo_unreserve(abo);
+
+       /* Wait until we're out of the vertical blank period before the one
+        * targeted by the flip
+        */
+       while ((acrtc->enabled &&
+               (amdgpu_get_crtc_scanoutpos(adev->ddev, acrtc->crtc_id, 0,
+                                       &vpos, &hpos, NULL, NULL,
+                                       &crtc->hwmode)
+                & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK)) ==
+               (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK) &&
+               (int)(target_vblank -
+                 amdgpu_get_vblank_counter_kms(adev->ddev, acrtc->crtc_id)) > 0)) {
+               usleep_range(1000, 1100);
+       }
+
+       /* Flip */
+       spin_lock_irqsave(&crtc->dev->event_lock, flags);
+       /* update crtc fb */
+       crtc->primary->fb = fb;
+
+       WARN_ON(acrtc->pflip_status != AMDGPU_FLIP_NONE);
+       WARN_ON(!acrtc_state->stream);
+
+       addr.address.grph.addr.low_part = lower_32_bits(afb->address);
+       addr.address.grph.addr.high_part = upper_32_bits(afb->address);
+       addr.flip_immediate = async_flip;
+
+
+       if (acrtc->base.state->event)
+               prepare_flip_isr(acrtc);
+
+       surface_updates->surface = dc_stream_get_status(acrtc_state->stream)->surfaces[0];
+       surface_updates->flip_addr = &addr;
+
+
+       dc_update_surfaces_and_stream(adev->dm.dc, surface_updates, 1, acrtc_state->stream, NULL);
+
+       DRM_DEBUG_DRIVER("%s Flipping to hi: 0x%x, low: 0x%x \n",
+                        __func__,
+                        addr.address.grph.addr.high_part,
+                        addr.address.grph.addr.low_part);
+
+
+       spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
+}
+
+static void amdgpu_dm_commit_surfaces(struct drm_atomic_state *state,
+                       struct drm_device *dev,
+                       struct amdgpu_display_manager *dm,
+                       struct drm_crtc *pcrtc,
+                       bool *wait_for_vblank)
+{
+       uint32_t i;
+       struct drm_plane *plane;
+       struct drm_plane_state *old_plane_state;
+       struct dc_stream *dc_stream_attach;
+       struct dc_surface *dc_surfaces_constructed[MAX_SURFACES];
+       struct amdgpu_crtc *acrtc_attach = to_amdgpu_crtc(pcrtc);
+       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(pcrtc->state);
+       int planes_count = 0;
+       unsigned long flags;
+
+       /* update planes when needed */
+       for_each_plane_in_state(state, plane, old_plane_state, i) {
+               struct drm_plane_state *plane_state = plane->state;
+               struct drm_crtc *crtc = plane_state->crtc;
+               struct drm_framebuffer *fb = plane_state->fb;
+               bool pflip_needed;
+               struct dm_plane_state *dm_plane_state = to_dm_plane_state(plane_state);
+
+               if (plane->type == DRM_PLANE_TYPE_CURSOR) {
+                       handle_cursor_update(plane, old_plane_state);
+                       continue;
+               }
+
+               if (!fb || !crtc || pcrtc != crtc || !crtc->state->active ||
+                               (!crtc->state->planes_changed &&
+                                               !pcrtc->state->color_mgmt_changed))
+                       continue;
+
+               pflip_needed = !state->allow_modeset;
+
+               spin_lock_irqsave(&crtc->dev->event_lock, flags);
+               if (acrtc_attach->pflip_status != AMDGPU_FLIP_NONE) {
+                       DRM_ERROR("add_surface: acrtc %d, already busy\n",
+                                       acrtc_attach->crtc_id);
+                       spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
+                       /* In comit tail framework this cannot happen */
+                       WARN_ON(1);
+               }
+               spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
+
+               if (!pflip_needed) {
+                       WARN_ON(!dm_plane_state->surface);
+
+                       dc_surfaces_constructed[planes_count] = dm_plane_state->surface;
+
+                       dc_stream_attach = acrtc_state->stream;
+                       planes_count++;
+
+               } else if (crtc->state->planes_changed) {
+                       /* Assume even ONE crtc with immediate flip means
+                        * entire can't wait for VBLANK
+                        * TODO Check if it's correct
+                        */
+                       *wait_for_vblank =
+                               acrtc_attach->flip_flags & DRM_MODE_PAGE_FLIP_ASYNC ?
+                               false : true;
+
+                       /* TODO: Needs rework for multiplane flip */
+                       if (plane->type == DRM_PLANE_TYPE_PRIMARY)
+                               drm_crtc_vblank_get(crtc);
+
+                       amdgpu_dm_do_flip(
+                               crtc,
+                               fb,
+                               drm_crtc_vblank_count(crtc) + *wait_for_vblank);
+
+                       /*TODO BUG remove ASAP in 4.12 to avoid race between worker and flip IOCTL */
+
+                       /*clean up the flags for next usage*/
+                       acrtc_attach->flip_flags = 0;
+               }
+
+       }
+
+       if (planes_count) {
+               unsigned long flags;
+
+               if (pcrtc->state->event) {
+
+                       drm_crtc_vblank_get(pcrtc);
+
+                       spin_lock_irqsave(&pcrtc->dev->event_lock, flags);
+                       prepare_flip_isr(acrtc_attach);
+                       spin_unlock_irqrestore(&pcrtc->dev->event_lock, flags);
+               }
+
+               if (false == dc_commit_surfaces_to_stream(dm->dc,
+                                                         dc_surfaces_constructed,
+                                                         planes_count,
+                                                         dc_stream_attach))
+                       dm_error("%s: Failed to attach surface!\n", __func__);
+       } else {
+               /*TODO BUG Here should go disable planes on CRTC. */
+       }
+}
+
+
+int amdgpu_dm_atomic_commit(
+               struct drm_device *dev,
+               struct drm_atomic_state *state,
+               bool nonblock)
+{
+       struct drm_crtc *crtc;
+       struct drm_crtc_state *new_state;
+       struct amdgpu_device *adev = dev->dev_private;
+       int i;
+
+       /*
+        * We evade vblanks and pflips on crtc that
+        * should be changed. We do it here to flush & disable
+        * interrupts before drm_swap_state is called in drm_atomic_helper_commit
+        * it will update crtc->dm_crtc_state->stream pointer which is used in
+        * the ISRs.
+        */
+       for_each_crtc_in_state(state, crtc, new_state, i) {
+               struct dm_crtc_state *old_acrtc_state = to_dm_crtc_state(crtc->state);
+               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
+
+               if (drm_atomic_crtc_needs_modeset(new_state) && old_acrtc_state->stream)
+                       manage_dm_interrupts(adev, acrtc, false);
+       }
+
+       return drm_atomic_helper_commit(dev, state, nonblock);
+
+       /*TODO Handle EINTR, reenable IRQ*/
+}
+
+void amdgpu_dm_atomic_commit_tail(
+       struct drm_atomic_state *state)
+{
+       struct drm_device *dev = state->dev;
+       struct amdgpu_device *adev = dev->dev_private;
+       struct amdgpu_display_manager *dm = &adev->dm;
+       struct dm_atomic_state *dm_state;
+       uint32_t i, j;
+       uint32_t new_crtcs_count = 0;
+       struct drm_crtc *crtc, *pcrtc;
+       struct drm_crtc_state *old_crtc_state;
+       struct amdgpu_crtc *new_crtcs[MAX_STREAMS];
+       struct dc_stream *new_stream = NULL;
+       unsigned long flags;
+       bool wait_for_vblank = true;
+       struct drm_connector *connector;
+       struct drm_connector_state *old_conn_state;
+       struct dm_crtc_state *old_acrtc_state, *new_acrtc_state;
+
+       drm_atomic_helper_update_legacy_modeset_state(dev, state);
+
+       dm_state = to_dm_atomic_state(state);
+
+       /* update changed items */
+       for_each_crtc_in_state(state, crtc, old_crtc_state, i) {
+               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
+               struct drm_crtc_state *new_state = crtc->state;
+               new_acrtc_state = to_dm_crtc_state(new_state);
+               old_acrtc_state = to_dm_crtc_state(old_crtc_state);
+
+               DRM_DEBUG_KMS(
+                       "amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, "
+                       "planes_changed:%d, mode_changed:%d,active_changed:%d,"
+                       "connectors_changed:%d\n",
+                       acrtc->crtc_id,
+                       new_state->enable,
+                       new_state->active,
+                       new_state->planes_changed,
+                       new_state->mode_changed,
+                       new_state->active_changed,
+                       new_state->connectors_changed);
+
+               /* handles headless hotplug case, updating new_state and
+                * aconnector as needed
+                */
+
+               if (modeset_required(new_state)) {
+
+                       DRM_INFO("Atomic commit: SET crtc id %d: [%p]\n", acrtc->crtc_id, acrtc);
+
+                       if (!new_acrtc_state->stream) {
+                               /*
+                               * this could happen because of issues with
+                               * userspace notifications delivery.
+                               * In this case userspace tries to set mode on
+                               * display which is disconnect in fact.
+                               * dc_sink in NULL in this case on aconnector.
+                               * We expect reset mode will come soon.
+                               *
+                               * This can also happen when unplug is done
+                               * during resume sequence ended
+                               *
+                               * In this case, we want to pretend we still
+                               * have a sink to keep the pipe running so that
+                               * hw state is consistent with the sw state
+                               */
+                               DRM_DEBUG_KMS("%s: Failed to create new stream for crtc %d\n",
+                                               __func__, acrtc->base.base.id);
+                               continue;
+                       }
+
+
+                       if (old_acrtc_state->stream)
+                               remove_stream(adev, acrtc, old_acrtc_state->stream);
+
+
+                       /*
+                        * this loop saves set mode crtcs
+                        * we needed to enable vblanks once all
+                        * resources acquired in dc after dc_commit_streams
+                        */
+
+                       /*TODO move all this into dm_crtc_state, get rid of
+                        * new_crtcs array and use old and new atomic states
+                        * instead
+                        */
+                       new_crtcs[new_crtcs_count] = acrtc;
+                       new_crtcs_count++;
+
+                       acrtc->enabled = true;
+                       acrtc->hw_mode = crtc->state->mode;
+                       crtc->hwmode = crtc->state->mode;
+               } else if (modereset_required(new_state)) {
+                       DRM_INFO("Atomic commit: RESET. crtc id %d:[%p]\n", acrtc->crtc_id, acrtc);
+
+                       /* i.e. reset mode */
+                       if (old_acrtc_state->stream)
+                               remove_stream(adev, acrtc, old_acrtc_state->stream);
+               }
+       } /* for_each_crtc_in_state() */
+
+       /*
+        * Add streams after required streams from new and replaced streams
+        * are removed from freesync module
+        */
+       if (adev->dm.freesync_module) {
+               for (i = 0; i < new_crtcs_count; i++) {
+                       struct amdgpu_connector *aconnector = NULL;
+                       new_acrtc_state = to_dm_crtc_state(new_crtcs[i]->base.state);
+
+                       new_stream = new_acrtc_state->stream;
+                       aconnector =
+                               amdgpu_dm_find_first_crct_matching_connector(
+                                       state,
+                                       &new_crtcs[i]->base,
+                                       false);
+                       if (!aconnector) {
+                               DRM_INFO(
+                                               "Atomic commit: Failed to find connector for acrtc id:%d "
+                                               "skipping freesync init\n",
+                                               new_crtcs[i]->crtc_id);
+                               continue;
+                       }
+
+                       mod_freesync_add_stream(adev->dm.freesync_module,
+                                               new_stream, &aconnector->caps);
+               }
+       }
+
+       if (dm_state->context)
+               WARN_ON(!dc_commit_context(dm->dc, dm_state->context));
+
+
+       list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
+               new_acrtc_state = to_dm_crtc_state(crtc->state);
+
+               if (new_acrtc_state->stream != NULL) {
+                       const struct dc_stream_status *status =
+                                       dc_stream_get_status(new_acrtc_state->stream);
+
+                       if (!status)
+                               DC_ERR("got no status for stream %p on acrtc%p\n", new_acrtc_state->stream, acrtc);
+                       else
+                               acrtc->otg_inst = status->primary_otg_inst;
+               }
+       }
+
+       /* Handle scaling and undersacn changes*/
+       for_each_connector_in_state(state, connector, old_conn_state, i) {
+               struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
+               struct dm_connector_state *con_new_state =
+                               to_dm_connector_state(aconnector->base.state);
+               struct dm_connector_state *con_old_state =
+                               to_dm_connector_state(old_conn_state);
+               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(con_new_state->base.crtc);
+               struct dc_stream_status *status = NULL;
+
+               /* Skip any modesets/resets */
+               if (!acrtc || drm_atomic_crtc_needs_modeset(acrtc->base.state))
+                       continue;
+
+               /* Skip any thing not scale or underscan changes */
+               if (!is_scaling_state_different(con_new_state, con_old_state))
+                       continue;
+
+               new_acrtc_state = to_dm_crtc_state(acrtc->base.state);
+
+               update_stream_scaling_settings(&con_new_state->base.crtc->mode,
+                               con_new_state, (struct dc_stream *)new_acrtc_state->stream);
+
+               status = dc_stream_get_status(new_acrtc_state->stream);
+               WARN_ON(!status);
+               WARN_ON(!status->surface_count);
+
+               if (!new_acrtc_state->stream)
+                       continue;
+
+               /*TODO How it works with MPO ?*/
+               if (!dc_commit_surfaces_to_stream(
+                               dm->dc,
+                               status->surfaces,
+                               status->surface_count,
+                               new_acrtc_state->stream))
+                       dm_error("%s: Failed to update stream scaling!\n", __func__);
+       }
+
+       for (i = 0; i < new_crtcs_count; i++) {
+               /*
+                * loop to enable interrupts on newly arrived crtc
+                */
+               struct amdgpu_crtc *acrtc = new_crtcs[i];
+               new_acrtc_state = to_dm_crtc_state(acrtc->base.state);
+
+               if (adev->dm.freesync_module)
+                       mod_freesync_notify_mode_change(
+                               adev->dm.freesync_module, &new_acrtc_state->stream, 1);
+
+               manage_dm_interrupts(adev, acrtc, true);
+       }
+
+       /* update planes when needed per crtc*/
+       for_each_crtc_in_state(state, pcrtc, old_crtc_state, j) {
+               new_acrtc_state = to_dm_crtc_state(pcrtc->state);
+
+               if (new_acrtc_state->stream)
+                       amdgpu_dm_commit_surfaces(state, dev, dm, pcrtc, &wait_for_vblank);
+       }
+
+
+       /*
+        * send vblank event on all events not handled in flip and
+        * mark consumed event for drm_atomic_helper_commit_hw_done
+        */
+       spin_lock_irqsave(&adev->ddev->event_lock, flags);
+       for_each_crtc_in_state(state, crtc, old_crtc_state, i) {
+               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
+
+               if (acrtc->base.state->event)
+                       drm_send_event_locked(dev, &crtc->state->event->base);
+
+               acrtc->base.state->event = NULL;
+       }
+       spin_unlock_irqrestore(&adev->ddev->event_lock, flags);
+
+       /* Signal HW programming completion */
+       drm_atomic_helper_commit_hw_done(state);
+
+       if (wait_for_vblank)
+               drm_atomic_helper_wait_for_vblanks(dev, state);
+
+       drm_atomic_helper_cleanup_planes(dev, state);
+}
+
+
+static int dm_force_atomic_commit(struct drm_connector *connector)
+{
+       int ret = 0;
+       struct drm_device *ddev = connector->dev;
+       struct drm_atomic_state *state = drm_atomic_state_alloc(ddev);
+       struct amdgpu_crtc *disconnected_acrtc = to_amdgpu_crtc(connector->encoder->crtc);
+       struct drm_plane *plane = disconnected_acrtc->base.primary;
+       struct drm_connector_state *conn_state;
+       struct drm_crtc_state *crtc_state;
+       struct drm_plane_state *plane_state;
+
+       if (!state)
+               return -ENOMEM;
+
+       state->acquire_ctx = ddev->mode_config.acquire_ctx;
+
+       /* Construct an atomic state to restore previous display setting */
+
+       /*
+        * Attach connectors to drm_atomic_state
+        */
+       conn_state = drm_atomic_get_connector_state(state, connector);
+
+       ret = PTR_ERR_OR_ZERO(conn_state);
+       if (ret)
+               goto err;
+
+       /* Attach crtc to drm_atomic_state*/
+       crtc_state = drm_atomic_get_crtc_state(state, &disconnected_acrtc->base);
+
+       ret = PTR_ERR_OR_ZERO(crtc_state);
+       if (ret)
+               goto err;
+
+       /* force a restore */
+       crtc_state->mode_changed = true;
+
+       /* Attach plane to drm_atomic_state */
+       plane_state = drm_atomic_get_plane_state(state, plane);
+
+       ret = PTR_ERR_OR_ZERO(plane_state);
+       if (ret)
+               goto err;
+
+
+       /* Call commit internally with the state we just constructed */
+       ret = drm_atomic_commit(state);
+       if (!ret)
+               return 0;
+
+err:
+       DRM_ERROR("Restoring old state failed with %i\n", ret);
+       drm_atomic_state_put(state);
+
+       return ret;
+}
+
+/*
+ * This functions handle all cases when set mode does not come upon hotplug.
+ * This include when the same display is unplugged then plugged back into the
+ * same port and when we are running without usermode desktop manager supprot
+ */
+void dm_restore_drm_connector_state(struct drm_device *dev, struct drm_connector *connector)
+{
+       struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
+       struct amdgpu_crtc *disconnected_acrtc;
+       struct dm_crtc_state *acrtc_state;
+
+       if (!aconnector->dc_sink || !connector->state || !connector->encoder)
+               return;
+
+       disconnected_acrtc = to_amdgpu_crtc(connector->encoder->crtc);
+       acrtc_state = to_dm_crtc_state(disconnected_acrtc->base.state);
+
+       if (!disconnected_acrtc || !acrtc_state->stream)
+               return;
+
+       /*
+        * If the previous sink is not released and different from the current,
+        * we deduce we are in a state where we can not rely on usermode call
+        * to turn on the display, so we do it here
+        */
+       if (acrtc_state->stream->sink != aconnector->dc_sink)
+               dm_force_atomic_commit(&aconnector->base);
+}
+
+static uint32_t add_val_sets_surface(
+       struct dc_validation_set *val_sets,
+       uint32_t set_count,
+       const struct dc_stream *stream,
+       struct dc_surface *surface)
+{
+       uint32_t i = 0, j = 0;
+
+       while (i < set_count) {
+               if (val_sets[i].stream == stream) {
+                       while (val_sets[i].surfaces[j])
+                               j++;
+                       break;
+               }
+               ++i;
+       }
+
+       val_sets[i].surfaces[j] = surface;
+       val_sets[i].surface_count++;
+
+       return val_sets[i].surface_count;
+}
+
+static uint32_t update_in_val_sets_stream(
+       struct dc_validation_set *val_sets,
+       uint32_t set_count,
+       struct dc_stream *old_stream,
+       struct dc_stream *new_stream,
+       struct drm_crtc *crtc)
+{
+       uint32_t i = 0;
+
+       while (i < set_count) {
+               if (val_sets[i].stream == old_stream)
+                       break;
+               ++i;
+       }
+
+       val_sets[i].stream = new_stream;
+
+       if (i == set_count)
+               /* nothing found. add new one to the end */
+               return set_count + 1;
+
+       return set_count;
+}
+
+static uint32_t remove_from_val_sets(
+       struct dc_validation_set *val_sets,
+       uint32_t set_count,
+       const struct dc_stream *stream)
+{
+       int i;
+
+       for (i = 0; i < set_count; i++)
+               if (val_sets[i].stream == stream)
+                       break;
+
+       if (i == set_count) {
+               /* nothing found */
+               return set_count;
+       }
+
+       set_count--;
+
+       for (; i < set_count; i++) {
+               val_sets[i] = val_sets[i + 1];
+       }
+
+       return set_count;
+}
+
+/*`
+ * Grabs all modesetting locks to serialize against any blocking commits,
+ * Waits for completion of all non blocking commits.
+ */
+static int do_aquire_global_lock(
+               struct drm_device *dev,
+               struct drm_atomic_state *state)
+{
+       struct drm_crtc *crtc;
+       struct drm_crtc_commit *commit;
+       long ret;
+
+       /* Adding all modeset locks to aquire_ctx will
+        * ensure that when the framework release it the
+        * extra locks we are locking here will get released to
+        */
+       ret = drm_modeset_lock_all_ctx(dev, state->acquire_ctx);
+       if (ret)
+               return ret;
+
+       list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+               spin_lock(&crtc->commit_lock);
+               commit = list_first_entry_or_null(&crtc->commit_list,
+                               struct drm_crtc_commit, commit_entry);
+               if (commit)
+                       drm_crtc_commit_get(commit);
+               spin_unlock(&crtc->commit_lock);
+
+               if (!commit)
+                       continue;
+
+               /* Make sure all pending HW programming completed and
+                * page flips done
+                */
+               ret = wait_for_completion_interruptible_timeout(&commit->hw_done, 10*HZ);
+
+               if (ret > 0)
+                       ret = wait_for_completion_interruptible_timeout(
+                                       &commit->flip_done, 10*HZ);
+
+               if (ret == 0)
+                       DRM_ERROR("[CRTC:%d:%s] hw_done or flip_done "
+                                       "timed out\n", crtc->base.id, crtc->name);
+
+               drm_crtc_commit_put(commit);
+       }
+
+       return ret < 0 ? ret : 0;
+}
+
+int amdgpu_dm_atomic_check(struct drm_device *dev,
+                       struct drm_atomic_state *state)
+{
+       struct dm_atomic_state *dm_state;
+       struct drm_crtc *crtc;
+       struct drm_crtc_state *crtc_state;
+       struct drm_plane *plane;
+       struct drm_plane_state *plane_state;
+       int i, j;
+       int ret;
+       struct amdgpu_device *adev = dev->dev_private;
+       struct dc *dc = adev->dm.dc;
+       struct drm_connector *connector;
+       struct drm_connector_state *conn_state;
+       int set_count;
+       struct dc_validation_set set[MAX_STREAMS] = { { 0 } };
+       struct dm_crtc_state *old_acrtc_state, *new_acrtc_state;
+
+       /*
+        * This bool will be set for true for any modeset/reset
+        * or surface update which implies non fast surface update.
+        */
+       bool lock_and_validation_needed = false;
+
+       ret = drm_atomic_helper_check_modeset(dev, state);
+
+       if (ret) {
+               DRM_ERROR("Atomic state validation failed with error :%d !\n", ret);
+               return ret;
+       }
+
+       dm_state = to_dm_atomic_state(state);
+
+       /* copy existing configuration */
+       set_count = 0;
+       list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
+
+               old_acrtc_state = to_dm_crtc_state(crtc->state);
+
+               if (old_acrtc_state->stream) {
+                       dc_stream_retain(old_acrtc_state->stream);
+                       set[set_count].stream = old_acrtc_state->stream;
+                       ++set_count;
+               }
+       }
+
+       /*TODO Move this code into dm_crtc_atomic_check once we get rid of dc_validation_set */
+       /* update changed items */
+       for_each_crtc_in_state(state, crtc, crtc_state, i) {
+               struct amdgpu_crtc *acrtc = NULL;
+               struct amdgpu_connector *aconnector = NULL;
+               old_acrtc_state = to_dm_crtc_state(crtc->state);
+               new_acrtc_state = to_dm_crtc_state(crtc_state);
+               acrtc = to_amdgpu_crtc(crtc);
+
+               aconnector = amdgpu_dm_find_first_crct_matching_connector(state, crtc, true);
+
+               DRM_DEBUG_KMS(
+                       "amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, "
+                       "planes_changed:%d, mode_changed:%d,active_changed:%d,"
+                       "connectors_changed:%d\n",
+                       acrtc->crtc_id,
+                       crtc_state->enable,
+                       crtc_state->active,
+                       crtc_state->planes_changed,
+                       crtc_state->mode_changed,
+                       crtc_state->active_changed,
+                       crtc_state->connectors_changed);
+
+               if (modeset_required(crtc_state)) {
+
+                       struct dc_stream *new_stream = NULL;
+                       struct drm_connector_state *conn_state = NULL;
+                       struct dm_connector_state *dm_conn_state = NULL;
+
+                       if (aconnector) {
+                               conn_state = drm_atomic_get_connector_state(state, &aconnector->base);
+                               if (IS_ERR(conn_state)) {
+                                       ret = PTR_ERR_OR_ZERO(conn_state);
+                                       goto fail;
+                               }
+
+                               dm_conn_state = to_dm_connector_state(conn_state);
+                       }
+
+                       new_stream = create_stream_for_sink(aconnector, &crtc_state->mode, dm_conn_state);
+
+                       /*
+                        * we can have no stream on ACTION_SET if a display
+                        * was disconnected during S3, in this case it not and
+                        * error, the OS will be updated after detection, and
+                        * do the right thing on next atomic commit
+                        */
+                       if (!new_stream) {
+                               DRM_DEBUG_KMS("%s: Failed to create new stream for crtc %d\n",
+                                               __func__, acrtc->base.base.id);
+                               break;
+                       }
+
+                       if (new_acrtc_state->stream)
+                               dc_stream_release(new_acrtc_state->stream);
+
+                       new_acrtc_state->stream = new_stream;
+
+                       set_count = update_in_val_sets_stream(
+                                       set,
+                                       set_count,
+                                       old_acrtc_state->stream,
+                                       new_acrtc_state->stream,
+                                       crtc);
+
+                       lock_and_validation_needed = true;
+
+               } else if (modereset_required(crtc_state)) {
+
+                       /* i.e. reset mode */
+                       if (new_acrtc_state->stream) {
+                               set_count = remove_from_val_sets(
+                                               set,
+                                               set_count,
+                                               new_acrtc_state->stream);
+
+                               dc_stream_release(new_acrtc_state->stream);
+                               new_acrtc_state->stream = NULL;
+
+                               lock_and_validation_needed = true;
+                       }
+               }
+
+
+               /*
+                * Hack: Commit needs planes right now, specifically for gamma
+                * TODO rework commit to check CRTC for gamma change
+                */
+               if (crtc_state->color_mgmt_changed) {
+
+                       ret = drm_atomic_add_affected_planes(state, crtc);
+                       if (ret)
+                               goto fail;
+               }
+       }
+
+       /* Check scaling and undersacn changes*/
+       /*TODO Removed scaling changes validation due to inability to commit
+        * new stream into context w\o causing full reset. Need to
+        * decide how to handle.
+        */
+       for_each_connector_in_state(state, connector, conn_state, i) {
+               struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
+               struct dm_connector_state *con_old_state =
+                               to_dm_connector_state(aconnector->base.state);
+               struct dm_connector_state *con_new_state =
+                                               to_dm_connector_state(conn_state);
+               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(con_new_state->base.crtc);
+
+               /* Skip any modesets/resets */
+               if (!acrtc || drm_atomic_crtc_needs_modeset(acrtc->base.state))
+                       continue;
+
+               /* Skip any thing not scale or underscan chnages */
+               if (!is_scaling_state_different(con_new_state, con_old_state))
+                       continue;
+
+               lock_and_validation_needed = true;
+       }
+
+       for_each_crtc_in_state(state, crtc, crtc_state, i) {
+               new_acrtc_state = to_dm_crtc_state(crtc_state);
+
+               for_each_plane_in_state(state, plane, plane_state, j) {
+                       struct drm_crtc *plane_crtc = plane_state->crtc;
+                       struct drm_framebuffer *fb = plane_state->fb;
+                       bool pflip_needed;
+                       struct dm_plane_state *dm_plane_state = to_dm_plane_state(plane_state);
+
+                       /*TODO Implement atomic check for cursor plane */
+                       if (plane->type == DRM_PLANE_TYPE_CURSOR)
+                               continue;
+
+                       if (!fb || !plane_crtc || crtc != plane_crtc || !crtc_state->active)
+                               continue;
+
+                       WARN_ON(!new_acrtc_state->stream);
+
+                       pflip_needed = !state->allow_modeset;
+                       if (!pflip_needed) {
+                               struct dc_surface *surface;
+
+                               surface = dc_create_surface(dc);
+
+                               ret = fill_plane_attributes(
+                                       plane_crtc->dev->dev_private,
+                                       surface,
+                                       plane_state,
+                                       crtc_state,
+                                       false);
+                               if (ret)
+                                       goto fail;
+
+
+                               if (dm_plane_state->surface)
+                                       dc_surface_release(dm_plane_state->surface);
+
+                               dm_plane_state->surface = surface;
+
+                               add_val_sets_surface(set,
+                                                    set_count,
+                                                    new_acrtc_state->stream,
+                                                    surface);
+
+                               lock_and_validation_needed = true;
+                       }
+               }
+       }
+
+       /* Run this here since we want to validate the streams we created */
+       ret = drm_atomic_helper_check_planes(dev, state);
+       if (ret)
+               goto fail;
+
+       /*
+        * For full updates case when
+        * removing/adding/updating  streams on once CRTC while flipping
+        * on another CRTC,
+        * acquiring global lock  will guarantee that any such full
+        * update commit
+        * will wait for completion of any outstanding flip using DRMs
+        * synchronization events.
+        */
+
+       if (lock_and_validation_needed) {
+
+               ret = do_aquire_global_lock(dev, state);
+               if (ret)
+                       goto fail;
+               WARN_ON(dm_state->context);
+               dm_state->context = dc_get_validate_context(dc, set, set_count);
+               if (!dm_state->context) {
+                       ret = -EINVAL;
+                       goto fail;
+               }
+       }
+
+       /* Must be success */
+       WARN_ON(ret);
+       return ret;
+
+fail:
+       if (ret == -EDEADLK)
+               DRM_DEBUG_KMS("Atomic check stopped due to to deadlock.\n");
+       else if (ret == -EINTR || ret == -EAGAIN || ret == -ERESTARTSYS)
+               DRM_DEBUG_KMS("Atomic check stopped due to to signal.\n");
+       else
+               DRM_ERROR("Atomic check failed with err: %d .\n", ret);
+
+       return ret;
+}
+
+static bool is_dp_capable_without_timing_msa(
+               struct dc *dc,
+               struct amdgpu_connector *amdgpu_connector)
+{
+       uint8_t dpcd_data;
+       bool capable = false;
+
+       if (amdgpu_connector->dc_link &&
+               dm_helpers_dp_read_dpcd(
+                               NULL,
+                               amdgpu_connector->dc_link,
+                               DP_DOWN_STREAM_PORT_COUNT,
+                               &dpcd_data,
+                               sizeof(dpcd_data))) {
+               capable = (dpcd_data & DP_MSA_TIMING_PAR_IGNORED) ? true:false;
+       }
+
+       return capable;
+}
+void amdgpu_dm_add_sink_to_freesync_module(
+               struct drm_connector *connector,
+               struct edid *edid)
+{
+       int i;
+       uint64_t val_capable;
+       bool edid_check_required;
+       struct detailed_timing *timing;
+       struct detailed_non_pixel *data;
+       struct detailed_data_monitor_range *range;
+       struct amdgpu_connector *amdgpu_connector =
+                       to_amdgpu_connector(connector);
+
+       struct drm_device *dev = connector->dev;
+       struct amdgpu_device *adev = dev->dev_private;
+       edid_check_required = false;
+       if (!amdgpu_connector->dc_sink) {
+               DRM_ERROR("dc_sink NULL, could not add free_sync module.\n");
+               return;
+       }
+       if (!adev->dm.freesync_module)
+               return;
+       /*
+        * if edid non zero restrict freesync only for dp and edp
+        */
+       if (edid) {
+               if (amdgpu_connector->dc_sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT
+                       || amdgpu_connector->dc_sink->sink_signal == SIGNAL_TYPE_EDP) {
+                       edid_check_required = is_dp_capable_without_timing_msa(
+                                               adev->dm.dc,
+                                               amdgpu_connector);
+               }
+       }
+       val_capable = 0;
+       if (edid_check_required == true && (edid->version > 1 ||
+          (edid->version == 1 && edid->revision > 1))) {
+               for (i = 0; i < 4; i++) {
+
+                       timing  = &edid->detailed_timings[i];
+                       data    = &timing->data.other_data;
+                       range   = &data->data.range;
+                       /*
+                        * Check if monitor has continuous frequency mode
+                        */
+                       if (data->type != EDID_DETAIL_MONITOR_RANGE)
+                               continue;
+                       /*
+                        * Check for flag range limits only. If flag == 1 then
+                        * no additional timing information provided.
+                        * Default GTF, GTF Secondary curve and CVT are not
+                        * supported
+                        */
+                       if (range->flags != 1)
+                               continue;
+
+                       amdgpu_connector->min_vfreq = range->min_vfreq;
+                       amdgpu_connector->max_vfreq = range->max_vfreq;
+                       amdgpu_connector->pixel_clock_mhz =
+                               range->pixel_clock_mhz * 10;
+                       break;
+               }
+
+               if (amdgpu_connector->max_vfreq -
+                               amdgpu_connector->min_vfreq > 10) {
+                       amdgpu_connector->caps.supported = true;
+                       amdgpu_connector->caps.min_refresh_in_micro_hz =
+                                       amdgpu_connector->min_vfreq * 1000000;
+                       amdgpu_connector->caps.max_refresh_in_micro_hz =
+                                       amdgpu_connector->max_vfreq * 1000000;
+                               val_capable = 1;
+               }
+       }
+
+       /*
+        * TODO figure out how to notify user-mode or DRM of freesync caps
+        * once we figure out how to deal with freesync in an upstreamable
+        * fashion
+        */
+
+}
+
+void amdgpu_dm_remove_sink_from_freesync_module(
+               struct drm_connector *connector)
+{
+       /*
+        * TODO fill in once we figure out how to deal with freesync in
+        * an upstreamable fashion
+        */
+}
index ee69179636a1f0efe7da3dc18957ec7cabc780d6..16b2e0849a6551a4ed0389129891bfb62831077d 100644 (file)
@@ -26,9 +26,9 @@
 #ifndef __AMDGPU_DM_H__
 #define __AMDGPU_DM_H__
 
-/*
-#include "linux/switch.h"
-*/
+#include <drm/drmP.h>
+#include <drm/drm_atomic.h>
+#include "dc.h"
 
 /*
  * This file contains the definition for amdgpu_display_manager
@@ -173,4 +173,106 @@ struct amdgpu_connector *amdgpu_dm_find_first_crct_matching_connector(
        struct drm_crtc *crtc,
        bool from_state_var);
 
+
+struct amdgpu_framebuffer;
+struct amdgpu_display_manager;
+struct dc_validation_set;
+struct dc_surface;
+/* TODO rename to dc_stream_state */
+struct  dc_stream;
+
+
+struct dm_plane_state {
+       struct drm_plane_state base;
+       struct dc_surface *surface;
+};
+
+struct dm_crtc_state {
+       struct drm_crtc_state base;
+       struct dc_stream *stream;
+};
+
+#define to_dm_crtc_state(x)    container_of(x, struct dm_crtc_state, base)
+
+struct dm_atomic_state {
+       struct drm_atomic_state base;
+
+       struct validate_context *context;
+};
+
+#define to_dm_atomic_state(x) container_of(x, struct dm_atomic_state, base)
+
+
+/*TODO Jodan Hersen use the one in amdgpu_dm*/
+int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
+                       struct amdgpu_plane *aplane,
+                       unsigned long possible_crtcs);
+int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm,
+                       struct drm_plane *plane,
+                       uint32_t link_index);
+int amdgpu_dm_connector_init(struct amdgpu_display_manager *dm,
+                       struct amdgpu_connector *amdgpu_connector,
+                       uint32_t link_index,
+                       struct amdgpu_encoder *amdgpu_encoder);
+int amdgpu_dm_encoder_init(
+       struct drm_device *dev,
+       struct amdgpu_encoder *aencoder,
+       uint32_t link_index);
+
+void amdgpu_dm_crtc_destroy(struct drm_crtc *crtc);
+void amdgpu_dm_connector_destroy(struct drm_connector *connector);
+void amdgpu_dm_encoder_destroy(struct drm_encoder *encoder);
+
+int amdgpu_dm_connector_get_modes(struct drm_connector *connector);
+
+int amdgpu_dm_atomic_commit(
+               struct drm_device *dev,
+               struct drm_atomic_state *state,
+               bool nonblock);
+
+void amdgpu_dm_atomic_commit_tail(
+       struct drm_atomic_state *state);
+
+int amdgpu_dm_atomic_check(struct drm_device *dev,
+                               struct drm_atomic_state *state);
+
+void amdgpu_dm_connector_funcs_reset(struct drm_connector *connector);
+struct drm_connector_state *amdgpu_dm_connector_atomic_duplicate_state(
+       struct drm_connector *connector);
+int amdgpu_dm_connector_atomic_set_property(
+       struct drm_connector *connector,
+       struct drm_connector_state *state,
+       struct drm_property *property,
+       uint64_t val);
+
+int amdgpu_dm_connector_atomic_get_property(
+       struct drm_connector *connector,
+       const struct drm_connector_state *state,
+       struct drm_property *property,
+       uint64_t *val);
+
+int amdgpu_dm_get_encoder_crtc_mask(struct amdgpu_device *adev);
+
+void amdgpu_dm_connector_init_helper(
+       struct amdgpu_display_manager *dm,
+       struct amdgpu_connector *aconnector,
+       int connector_type,
+       struct dc_link *link,
+       int link_index);
+
+int amdgpu_dm_connector_mode_valid(
+       struct drm_connector *connector,
+       struct drm_display_mode *mode);
+
+void dm_restore_drm_connector_state(struct drm_device *dev, struct drm_connector *connector);
+
+void amdgpu_dm_add_sink_to_freesync_module(
+               struct drm_connector *connector,
+               struct edid *edid);
+
+void amdgpu_dm_remove_sink_from_freesync_module(
+               struct drm_connector *connector);
+
+extern const struct drm_encoder_helper_funcs amdgpu_dm_encoder_helper_funcs;
+
 #endif /* __AMDGPU_DM_H__ */
index 6c19a07709dfe9ddf14551d335efca7ab11345b7..0a615583de63915d01c53f272edec630ed11acae 100644 (file)
@@ -38,7 +38,6 @@
 #include "dc.h"
 #include "amdgpu_dm.h"
 #include "amdgpu_dm_irq.h"
-#include "amdgpu_dm_types.h"
 
 #include "dm_helpers.h"
 
index ec91a52c002f956b2b83d8743eefea3fcb6daaa0..82e4d49fb1844f911579263c5e44535341aa2a56 100644 (file)
@@ -27,7 +27,7 @@
 #include <drm/drm_atomic_helper.h>
 #include "dm_services.h"
 #include "amdgpu.h"
-#include "amdgpu_dm_types.h"
+#include "amdgpu_dm.h"
 #include "amdgpu_dm_mst_types.h"
 
 #include "dc.h"
index a8a53b85905a151f91f37f4f39f33d57277dc236..3348e90a0a370cb103cd57d5ea20ef9ac929c8b4 100644 (file)
@@ -33,7 +33,6 @@
 #include "amdgpu.h"
 #include "amdgpu_dm.h"
 #include "amdgpu_dm_irq.h"
-#include "amdgpu_dm_types.h"
 #include "amdgpu_pm.h"
 
 /******************************************************************************
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.c
deleted file mode 100644 (file)
index 80d4e26..0000000
+++ /dev/null
@@ -1,3286 +0,0 @@
-/*
- * Copyright 2012-13 Advanced Micro Devices, Inc.
- *
- * 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 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 COPYRIGHT HOLDER(S) OR AUTHOR(S) 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.
- *
- * Authors: AMD
- *
- */
-
-#include <linux/types.h>
-#include <linux/version.h>
-
-#include <drm/drmP.h>
-#include <drm/drm_atomic_helper.h>
-#include <drm/drm_fb_helper.h>
-#include <drm/drm_atomic.h>
-#include <drm/drm_edid.h>
-
-
-#include "amdgpu.h"
-#include "amdgpu_pm.h"
-#include "dm_helpers.h"
-#include "dm_services_types.h"
-
-// We need to #undef FRAME_SIZE and DEPRECATED because they conflict
-// with ptrace-abi.h's #define's of them.
-#undef FRAME_SIZE
-#undef DEPRECATED
-
-#include "dc.h"
-
-#include "amdgpu_dm_types.h"
-#include "amdgpu_dm_mst_types.h"
-
-#include "modules/inc/mod_freesync.h"
-
-#include "i2caux_interface.h"
-
-struct dm_connector_state {
-       struct drm_connector_state base;
-
-       enum amdgpu_rmx_type scaling;
-       uint8_t underscan_vborder;
-       uint8_t underscan_hborder;
-       bool underscan_enable;
-};
-
-#define to_dm_connector_state(x)\
-       container_of((x), struct dm_connector_state, base)
-
-static bool modeset_required(struct drm_crtc_state *crtc_state)
-{
-       if (!drm_atomic_crtc_needs_modeset(crtc_state))
-               return false;
-
-       if (!crtc_state->enable)
-               return false;
-
-       return crtc_state->active;
-}
-
-static bool modereset_required(struct drm_crtc_state *crtc_state)
-{
-       if (!drm_atomic_crtc_needs_modeset(crtc_state))
-               return false;
-
-       return !crtc_state->enable || !crtc_state->active;
-}
-
-void amdgpu_dm_encoder_destroy(struct drm_encoder *encoder)
-{
-       drm_encoder_cleanup(encoder);
-       kfree(encoder);
-}
-
-static const struct drm_encoder_funcs amdgpu_dm_encoder_funcs = {
-       .destroy = amdgpu_dm_encoder_destroy,
-};
-
-static void dm_set_cursor(
-       struct amdgpu_crtc *amdgpu_crtc,
-       uint64_t gpu_addr,
-       uint32_t width,
-       uint32_t height)
-{
-       struct dc_cursor_attributes attributes;
-       struct dc_cursor_position position;
-       struct drm_crtc *crtc = &amdgpu_crtc->base;
-       int x, y;
-       int xorigin = 0, yorigin = 0;
-       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state);
-
-       amdgpu_crtc->cursor_width = width;
-       amdgpu_crtc->cursor_height = height;
-
-       attributes.address.high_part = upper_32_bits(gpu_addr);
-       attributes.address.low_part  = lower_32_bits(gpu_addr);
-       attributes.width             = width;
-       attributes.height            = height;
-       attributes.color_format      = CURSOR_MODE_COLOR_PRE_MULTIPLIED_ALPHA;
-       attributes.rotation_angle    = 0;
-       attributes.attribute_flags.value = 0;
-
-       attributes.pitch = attributes.width;
-
-       x = amdgpu_crtc->cursor_x;
-       y = amdgpu_crtc->cursor_y;
-
-       /* avivo cursor are offset into the total surface */
-       x += crtc->primary->state->src_x >> 16;
-       y += crtc->primary->state->src_y >> 16;
-
-       if (x < 0) {
-               xorigin = min(-x, amdgpu_crtc->max_cursor_width - 1);
-               x = 0;
-       }
-       if (y < 0) {
-               yorigin = min(-y, amdgpu_crtc->max_cursor_height - 1);
-               y = 0;
-       }
-
-       position.enable = true;
-       position.x = x;
-       position.y = y;
-
-       position.x_hotspot = xorigin;
-       position.y_hotspot = yorigin;
-
-       if (!dc_stream_set_cursor_attributes(
-                               acrtc_state->stream,
-                               &attributes)) {
-               DRM_ERROR("DC failed to set cursor attributes\n");
-       }
-
-       if (!dc_stream_set_cursor_position(
-                               acrtc_state->stream,
-                               &position)) {
-               DRM_ERROR("DC failed to set cursor position\n");
-       }
-}
-
-static int dm_crtc_cursor_set(
-       struct drm_crtc *crtc,
-       uint64_t address,
-       uint32_t width,
-       uint32_t height)
-{
-       struct dc_cursor_position position;
-       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state);
-
-       int ret;
-
-       struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-       ret             = EINVAL;
-
-       DRM_DEBUG_KMS(
-               "%s: crtc_id=%d with size %d to %d \n",
-               __func__,
-               amdgpu_crtc->crtc_id,
-               width,
-               height);
-
-       if (!address) {
-               /* turn off cursor */
-               position.enable = false;
-               position.x = 0;
-               position.y = 0;
-
-               if (acrtc_state->stream) {
-                       /*set cursor visible false*/
-                       dc_stream_set_cursor_position(
-                               acrtc_state->stream,
-                               &position);
-               }
-               goto release;
-
-       }
-
-       if ((width > amdgpu_crtc->max_cursor_width) ||
-               (height > amdgpu_crtc->max_cursor_height)) {
-               DRM_ERROR(
-                       "%s: bad cursor width or height %d x %d\n",
-                       __func__,
-                       width,
-                       height);
-               goto release;
-       }
-
-       /*program new cursor bo to hardware*/
-       dm_set_cursor(amdgpu_crtc, address, width, height);
-
-release:
-       return ret;
-
-}
-
-static int dm_crtc_cursor_move(struct drm_crtc *crtc,
-                                    int x, int y)
-{
-       struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-       int xorigin = 0, yorigin = 0;
-       struct dc_cursor_position position;
-       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state);
-
-       amdgpu_crtc->cursor_x = x;
-       amdgpu_crtc->cursor_y = y;
-
-       /* avivo cursor are offset into the total surface */
-       x += crtc->primary->state->src_x >> 16;
-       y += crtc->primary->state->src_y >> 16;
-
-       /*
-        * TODO: for cursor debugging unguard the following
-        */
-#if 0
-       DRM_DEBUG_KMS(
-               "%s: x %d y %d c->x %d c->y %d\n",
-               __func__,
-               x,
-               y,
-               crtc->x,
-               crtc->y);
-#endif
-
-       if (x < 0) {
-               xorigin = min(-x, amdgpu_crtc->max_cursor_width - 1);
-               x = 0;
-       }
-       if (y < 0) {
-               yorigin = min(-y, amdgpu_crtc->max_cursor_height - 1);
-               y = 0;
-       }
-
-       position.enable = true;
-       position.x = x;
-       position.y = y;
-
-       position.x_hotspot = xorigin;
-       position.y_hotspot = yorigin;
-
-       if (acrtc_state->stream) {
-               if (!dc_stream_set_cursor_position(
-                                       acrtc_state->stream,
-                                       &position)) {
-                       DRM_ERROR("DC failed to set cursor position\n");
-                       return -EINVAL;
-               }
-       }
-
-       return 0;
-}
-
-static bool fill_rects_from_plane_state(
-       const struct drm_plane_state *state,
-       struct dc_surface *surface)
-{
-       surface->src_rect.x = state->src_x >> 16;
-       surface->src_rect.y = state->src_y >> 16;
-       /*we ignore for now mantissa and do not to deal with floating pixels :(*/
-       surface->src_rect.width = state->src_w >> 16;
-
-       if (surface->src_rect.width == 0)
-               return false;
-
-       surface->src_rect.height = state->src_h >> 16;
-       if (surface->src_rect.height == 0)
-               return false;
-
-       surface->dst_rect.x = state->crtc_x;
-       surface->dst_rect.y = state->crtc_y;
-
-       if (state->crtc_w == 0)
-               return false;
-
-       surface->dst_rect.width = state->crtc_w;
-
-       if (state->crtc_h == 0)
-               return false;
-
-       surface->dst_rect.height = state->crtc_h;
-
-       surface->clip_rect = surface->dst_rect;
-
-       switch (state->rotation & DRM_MODE_ROTATE_MASK) {
-       case DRM_MODE_ROTATE_0:
-               surface->rotation = ROTATION_ANGLE_0;
-               break;
-       case DRM_MODE_ROTATE_90:
-               surface->rotation = ROTATION_ANGLE_90;
-               break;
-       case DRM_MODE_ROTATE_180:
-               surface->rotation = ROTATION_ANGLE_180;
-               break;
-       case DRM_MODE_ROTATE_270:
-               surface->rotation = ROTATION_ANGLE_270;
-               break;
-       default:
-               surface->rotation = ROTATION_ANGLE_0;
-               break;
-       }
-
-       return true;
-}
-static int get_fb_info(
-       const struct amdgpu_framebuffer *amdgpu_fb,
-       uint64_t *tiling_flags,
-       uint64_t *fb_location)
-{
-       struct amdgpu_bo *rbo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-       int r = amdgpu_bo_reserve(rbo, false);
-       if (unlikely(r)) {
-               DRM_ERROR("Unable to reserve buffer\n");
-               return r;
-       }
-
-       if (fb_location)
-               *fb_location = amdgpu_bo_gpu_offset(rbo);
-
-       if (tiling_flags)
-               amdgpu_bo_get_tiling_flags(rbo, tiling_flags);
-
-       amdgpu_bo_unreserve(rbo);
-
-       return r;
-}
-
-static int fill_plane_attributes_from_fb(
-       struct amdgpu_device *adev,
-       struct dc_surface *surface,
-       const struct amdgpu_framebuffer *amdgpu_fb, bool addReq)
-{
-       uint64_t tiling_flags;
-       uint64_t fb_location = 0;
-       unsigned int awidth;
-       const struct drm_framebuffer *fb = &amdgpu_fb->base;
-       int ret = 0;
-       struct drm_format_name_buf format_name;
-
-       ret = get_fb_info(
-               amdgpu_fb,
-               &tiling_flags,
-               addReq == true ? &fb_location:NULL);
-
-       if (ret)
-               return ret;
-
-       switch (fb->format->format) {
-       case DRM_FORMAT_C8:
-               surface->format = SURFACE_PIXEL_FORMAT_GRPH_PALETA_256_COLORS;
-               break;
-       case DRM_FORMAT_RGB565:
-               surface->format = SURFACE_PIXEL_FORMAT_GRPH_RGB565;
-               break;
-       case DRM_FORMAT_XRGB8888:
-       case DRM_FORMAT_ARGB8888:
-               surface->format = SURFACE_PIXEL_FORMAT_GRPH_ARGB8888;
-               break;
-       case DRM_FORMAT_XRGB2101010:
-       case DRM_FORMAT_ARGB2101010:
-               surface->format = SURFACE_PIXEL_FORMAT_GRPH_ARGB2101010;
-               break;
-       case DRM_FORMAT_XBGR2101010:
-       case DRM_FORMAT_ABGR2101010:
-               surface->format = SURFACE_PIXEL_FORMAT_GRPH_ABGR2101010;
-               break;
-       case DRM_FORMAT_NV21:
-               surface->format = SURFACE_PIXEL_FORMAT_VIDEO_420_YCbCr;
-               break;
-       case DRM_FORMAT_NV12:
-               surface->format = SURFACE_PIXEL_FORMAT_VIDEO_420_YCrCb;
-               break;
-       default:
-               DRM_ERROR("Unsupported screen format %s\n",
-                         drm_get_format_name(fb->format->format, &format_name));
-               return -EINVAL;
-       }
-
-       if (surface->format < SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) {
-               surface->address.type = PLN_ADDR_TYPE_GRAPHICS;
-               surface->address.grph.addr.low_part = lower_32_bits(fb_location);
-               surface->address.grph.addr.high_part = upper_32_bits(fb_location);
-               surface->plane_size.grph.surface_size.x = 0;
-               surface->plane_size.grph.surface_size.y = 0;
-               surface->plane_size.grph.surface_size.width = fb->width;
-               surface->plane_size.grph.surface_size.height = fb->height;
-               surface->plane_size.grph.surface_pitch =
-                               fb->pitches[0] / fb->format->cpp[0];
-               /* TODO: unhardcode */
-               surface->color_space = COLOR_SPACE_SRGB;
-
-       } else {
-               awidth = ALIGN(fb->width, 64);
-               surface->address.type = PLN_ADDR_TYPE_VIDEO_PROGRESSIVE;
-               surface->address.video_progressive.luma_addr.low_part
-                                               = lower_32_bits(fb_location);
-               surface->address.video_progressive.chroma_addr.low_part
-                                               = lower_32_bits(fb_location) +
-                                                       (awidth * fb->height);
-               surface->plane_size.video.luma_size.x = 0;
-               surface->plane_size.video.luma_size.y = 0;
-               surface->plane_size.video.luma_size.width = awidth;
-               surface->plane_size.video.luma_size.height = fb->height;
-               /* TODO: unhardcode */
-               surface->plane_size.video.luma_pitch = awidth;
-
-               surface->plane_size.video.chroma_size.x = 0;
-               surface->plane_size.video.chroma_size.y = 0;
-               surface->plane_size.video.chroma_size.width = awidth;
-               surface->plane_size.video.chroma_size.height = fb->height;
-               surface->plane_size.video.chroma_pitch = awidth / 2;
-
-               /* TODO: unhardcode */
-               surface->color_space = COLOR_SPACE_YCBCR709;
-       }
-
-       memset(&surface->tiling_info, 0, sizeof(surface->tiling_info));
-
-       /* Fill GFX params */
-       if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == DC_ARRAY_2D_TILED_THIN1)
-       {
-               unsigned bankw, bankh, mtaspect, tile_split, num_banks;
-
-               bankw = AMDGPU_TILING_GET(tiling_flags, BANK_WIDTH);
-               bankh = AMDGPU_TILING_GET(tiling_flags, BANK_HEIGHT);
-               mtaspect = AMDGPU_TILING_GET(tiling_flags, MACRO_TILE_ASPECT);
-               tile_split = AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT);
-               num_banks = AMDGPU_TILING_GET(tiling_flags, NUM_BANKS);
-
-               /* XXX fix me for VI */
-               surface->tiling_info.gfx8.num_banks = num_banks;
-               surface->tiling_info.gfx8.array_mode =
-                               DC_ARRAY_2D_TILED_THIN1;
-               surface->tiling_info.gfx8.tile_split = tile_split;
-               surface->tiling_info.gfx8.bank_width = bankw;
-               surface->tiling_info.gfx8.bank_height = bankh;
-               surface->tiling_info.gfx8.tile_aspect = mtaspect;
-               surface->tiling_info.gfx8.tile_mode =
-                               DC_ADDR_SURF_MICRO_TILING_DISPLAY;
-       } else if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE)
-                       == DC_ARRAY_1D_TILED_THIN1) {
-               surface->tiling_info.gfx8.array_mode = DC_ARRAY_1D_TILED_THIN1;
-       }
-
-       surface->tiling_info.gfx8.pipe_config =
-                       AMDGPU_TILING_GET(tiling_flags, PIPE_CONFIG);
-
-       if (adev->asic_type == CHIP_VEGA10 ||
-           adev->asic_type == CHIP_RAVEN) {
-               /* Fill GFX9 params */
-               surface->tiling_info.gfx9.num_pipes =
-                       adev->gfx.config.gb_addr_config_fields.num_pipes;
-               surface->tiling_info.gfx9.num_banks =
-                       adev->gfx.config.gb_addr_config_fields.num_banks;
-               surface->tiling_info.gfx9.pipe_interleave =
-                       adev->gfx.config.gb_addr_config_fields.pipe_interleave_size;
-               surface->tiling_info.gfx9.num_shader_engines =
-                       adev->gfx.config.gb_addr_config_fields.num_se;
-               surface->tiling_info.gfx9.max_compressed_frags =
-                       adev->gfx.config.gb_addr_config_fields.max_compress_frags;
-               surface->tiling_info.gfx9.num_rb_per_se =
-                       adev->gfx.config.gb_addr_config_fields.num_rb_per_se;
-               surface->tiling_info.gfx9.swizzle =
-                       AMDGPU_TILING_GET(tiling_flags, SWIZZLE_MODE);
-               surface->tiling_info.gfx9.shaderEnable = 1;
-       }
-
-       surface->visible = true;
-       surface->scaling_quality.h_taps_c = 0;
-       surface->scaling_quality.v_taps_c = 0;
-
-       /* is this needed? is surface zeroed at allocation? */
-       surface->scaling_quality.h_taps = 0;
-       surface->scaling_quality.v_taps = 0;
-       surface->stereo_format = PLANE_STEREO_FORMAT_NONE;
-
-       return ret;
-
-}
-
-#define NUM_OF_RAW_GAMMA_RAMP_RGB_256 256
-
-static void fill_gamma_from_crtc_state(
-       const struct drm_crtc_state *crtc_state,
-       struct dc_surface *dc_surface)
-{
-       int i;
-       struct dc_gamma *gamma;
-       struct drm_color_lut *lut = (struct drm_color_lut *) crtc_state->gamma_lut->data;
-
-       gamma = dc_create_gamma();
-
-       if (gamma == NULL) {
-               WARN_ON(1);
-               return;
-       }
-
-       for (i = 0; i < NUM_OF_RAW_GAMMA_RAMP_RGB_256; i++) {
-               gamma->red[i] = lut[i].red;
-               gamma->green[i] = lut[i].green;
-               gamma->blue[i] = lut[i].blue;
-       }
-
-       dc_surface->gamma_correction = gamma;
-}
-
-static int fill_plane_attributes(
-                       struct amdgpu_device *adev,
-                       struct dc_surface *surface,
-                       struct drm_plane_state *plane_state,
-                       struct drm_crtc_state *crtc_state,
-                       bool addrReq)
-{
-       const struct amdgpu_framebuffer *amdgpu_fb =
-               to_amdgpu_framebuffer(plane_state->fb);
-       const struct drm_crtc *crtc = plane_state->crtc;
-       struct dc_transfer_func *input_tf;
-       int ret = 0;
-
-       if (!fill_rects_from_plane_state(plane_state, surface))
-               return -EINVAL;
-
-       ret = fill_plane_attributes_from_fb(
-               crtc->dev->dev_private,
-               surface,
-               amdgpu_fb,
-               addrReq);
-
-       if (ret)
-               return ret;
-
-       input_tf = dc_create_transfer_func();
-
-       if (input_tf == NULL)
-               return -ENOMEM;
-
-       input_tf->type = TF_TYPE_PREDEFINED;
-       input_tf->tf = TRANSFER_FUNCTION_SRGB;
-
-       surface->in_transfer_func = input_tf;
-
-       /* In case of gamma set, update gamma value */
-       if (crtc_state->gamma_lut)
-               fill_gamma_from_crtc_state(crtc_state, surface);
-
-       return ret;
-}
-
-/*****************************************************************************/
-
-struct amdgpu_connector *aconnector_from_drm_crtc_id(
-               const struct drm_crtc *crtc)
-{
-       struct drm_device *dev = crtc->dev;
-       struct drm_connector *connector;
-       struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
-       struct amdgpu_connector *aconnector;
-
-       list_for_each_entry(connector,
-                       &dev->mode_config.connector_list, head) {
-
-               aconnector = to_amdgpu_connector(connector);
-
-               if (aconnector->base.state->crtc != &acrtc->base)
-                       continue;
-
-               /* Found the connector */
-               return aconnector;
-       }
-
-       /* If we get here, not found. */
-       return NULL;
-}
-
-static void update_stream_scaling_settings(
-               const struct drm_display_mode *mode,
-               const struct dm_connector_state *dm_state,
-               struct dc_stream *stream)
-{
-       enum amdgpu_rmx_type rmx_type;
-
-       struct rect src = { 0 }; /* viewport in composition space*/
-       struct rect dst = { 0 }; /* stream addressable area */
-
-       /* no mode. nothing to be done */
-       if (!mode)
-               return;
-
-       /* Full screen scaling by default */
-       src.width = mode->hdisplay;
-       src.height = mode->vdisplay;
-       dst.width = stream->timing.h_addressable;
-       dst.height = stream->timing.v_addressable;
-
-       rmx_type = dm_state->scaling;
-       if (rmx_type == RMX_ASPECT || rmx_type == RMX_OFF) {
-               if (src.width * dst.height <
-                               src.height * dst.width) {
-                       /* height needs less upscaling/more downscaling */
-                       dst.width = src.width *
-                                       dst.height / src.height;
-               } else {
-                       /* width needs less upscaling/more downscaling */
-                       dst.height = src.height *
-                                       dst.width / src.width;
-               }
-       } else if (rmx_type == RMX_CENTER) {
-               dst = src;
-       }
-
-       dst.x = (stream->timing.h_addressable - dst.width) / 2;
-       dst.y = (stream->timing.v_addressable - dst.height) / 2;
-
-       if (dm_state->underscan_enable) {
-               dst.x += dm_state->underscan_hborder / 2;
-               dst.y += dm_state->underscan_vborder / 2;
-               dst.width -= dm_state->underscan_hborder;
-               dst.height -= dm_state->underscan_vborder;
-       }
-
-       stream->src = src;
-       stream->dst = dst;
-
-       DRM_DEBUG_KMS("Destination Rectangle x:%d  y:%d  width:%d  height:%d\n",
-                       dst.x, dst.y, dst.width, dst.height);
-
-}
-
-static enum dc_color_depth convert_color_depth_from_display_info(
-               const struct drm_connector *connector)
-{
-       uint32_t bpc = connector->display_info.bpc;
-
-       /* Limited color depth to 8bit
-        * TODO: Still need to handle deep color*/
-       if (bpc > 8)
-               bpc = 8;
-
-       switch (bpc) {
-       case 0:
-               /* Temporary Work around, DRM don't parse color depth for
-                * EDID revision before 1.4
-                * TODO: Fix edid parsing
-                */
-               return COLOR_DEPTH_888;
-       case 6:
-               return COLOR_DEPTH_666;
-       case 8:
-               return COLOR_DEPTH_888;
-       case 10:
-               return COLOR_DEPTH_101010;
-       case 12:
-               return COLOR_DEPTH_121212;
-       case 14:
-               return COLOR_DEPTH_141414;
-       case 16:
-               return COLOR_DEPTH_161616;
-       default:
-               return COLOR_DEPTH_UNDEFINED;
-       }
-}
-
-static enum dc_aspect_ratio get_aspect_ratio(
-               const struct drm_display_mode *mode_in)
-{
-       int32_t width = mode_in->crtc_hdisplay * 9;
-       int32_t height = mode_in->crtc_vdisplay * 16;
-       if ((width - height) < 10 && (width - height) > -10)
-               return ASPECT_RATIO_16_9;
-       else
-               return ASPECT_RATIO_4_3;
-}
-
-static enum dc_color_space get_output_color_space(
-                               const struct dc_crtc_timing *dc_crtc_timing)
-{
-       enum dc_color_space color_space = COLOR_SPACE_SRGB;
-
-       switch (dc_crtc_timing->pixel_encoding) {
-       case PIXEL_ENCODING_YCBCR422:
-       case PIXEL_ENCODING_YCBCR444:
-       case PIXEL_ENCODING_YCBCR420:
-       {
-               /*
-                * 27030khz is the separation point between HDTV and SDTV
-                * according to HDMI spec, we use YCbCr709 and YCbCr601
-                * respectively
-                */
-               if (dc_crtc_timing->pix_clk_khz > 27030) {
-                       if (dc_crtc_timing->flags.Y_ONLY)
-                               color_space =
-                                       COLOR_SPACE_YCBCR709_LIMITED;
-                       else
-                               color_space = COLOR_SPACE_YCBCR709;
-               } else {
-                       if (dc_crtc_timing->flags.Y_ONLY)
-                               color_space =
-                                       COLOR_SPACE_YCBCR601_LIMITED;
-                       else
-                               color_space = COLOR_SPACE_YCBCR601;
-               }
-
-       }
-       break;
-       case PIXEL_ENCODING_RGB:
-               color_space = COLOR_SPACE_SRGB;
-               break;
-
-       default:
-               WARN_ON(1);
-               break;
-       }
-
-       return color_space;
-}
-
-/*****************************************************************************/
-
-static void fill_stream_properties_from_drm_display_mode(
-       struct dc_stream *stream,
-       const struct drm_display_mode *mode_in,
-       const struct drm_connector *connector)
-{
-       struct dc_crtc_timing *timing_out = &stream->timing;
-       memset(timing_out, 0, sizeof(struct dc_crtc_timing));
-
-       timing_out->h_border_left = 0;
-       timing_out->h_border_right = 0;
-       timing_out->v_border_top = 0;
-       timing_out->v_border_bottom = 0;
-       /* TODO: un-hardcode */
-
-       if ((connector->display_info.color_formats & DRM_COLOR_FORMAT_YCRCB444)
-                       && stream->sink->sink_signal == SIGNAL_TYPE_HDMI_TYPE_A)
-               timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR444;
-       else
-               timing_out->pixel_encoding = PIXEL_ENCODING_RGB;
-
-       timing_out->timing_3d_format = TIMING_3D_FORMAT_NONE;
-       timing_out->display_color_depth = convert_color_depth_from_display_info(
-                       connector);
-       timing_out->scan_type = SCANNING_TYPE_NODATA;
-       timing_out->hdmi_vic = 0;
-       timing_out->vic = drm_match_cea_mode(mode_in);
-
-       timing_out->h_addressable = mode_in->crtc_hdisplay;
-       timing_out->h_total = mode_in->crtc_htotal;
-       timing_out->h_sync_width =
-               mode_in->crtc_hsync_end - mode_in->crtc_hsync_start;
-       timing_out->h_front_porch =
-               mode_in->crtc_hsync_start - mode_in->crtc_hdisplay;
-       timing_out->v_total = mode_in->crtc_vtotal;
-       timing_out->v_addressable = mode_in->crtc_vdisplay;
-       timing_out->v_front_porch =
-               mode_in->crtc_vsync_start - mode_in->crtc_vdisplay;
-       timing_out->v_sync_width =
-               mode_in->crtc_vsync_end - mode_in->crtc_vsync_start;
-       timing_out->pix_clk_khz = mode_in->crtc_clock;
-       timing_out->aspect_ratio = get_aspect_ratio(mode_in);
-       if (mode_in->flags & DRM_MODE_FLAG_PHSYNC)
-               timing_out->flags.HSYNC_POSITIVE_POLARITY = 1;
-       if (mode_in->flags & DRM_MODE_FLAG_PVSYNC)
-               timing_out->flags.VSYNC_POSITIVE_POLARITY = 1;
-
-       stream->output_color_space = get_output_color_space(timing_out);
-
-       {
-               struct dc_transfer_func *tf = dc_create_transfer_func();
-               tf->type = TF_TYPE_PREDEFINED;
-               tf->tf = TRANSFER_FUNCTION_SRGB;
-               stream->out_transfer_func = tf;
-       }
-}
-
-static void fill_audio_info(
-       struct audio_info *audio_info,
-       const struct drm_connector *drm_connector,
-       const struct dc_sink *dc_sink)
-{
-       int i = 0;
-       int cea_revision = 0;
-       const struct dc_edid_caps *edid_caps = &dc_sink->edid_caps;
-
-       audio_info->manufacture_id = edid_caps->manufacturer_id;
-       audio_info->product_id = edid_caps->product_id;
-
-       cea_revision = drm_connector->display_info.cea_rev;
-
-       while (i < AUDIO_INFO_DISPLAY_NAME_SIZE_IN_CHARS &&
-               edid_caps->display_name[i]) {
-               audio_info->display_name[i] = edid_caps->display_name[i];
-               i++;
-       }
-
-       if(cea_revision >= 3) {
-               audio_info->mode_count = edid_caps->audio_mode_count;
-
-               for (i = 0; i < audio_info->mode_count; ++i) {
-                       audio_info->modes[i].format_code =
-                                       (enum audio_format_code)
-                                       (edid_caps->audio_modes[i].format_code);
-                       audio_info->modes[i].channel_count =
-                                       edid_caps->audio_modes[i].channel_count;
-                       audio_info->modes[i].sample_rates.all =
-                                       edid_caps->audio_modes[i].sample_rate;
-                       audio_info->modes[i].sample_size =
-                                       edid_caps->audio_modes[i].sample_size;
-               }
-       }
-
-       audio_info->flags.all = edid_caps->speaker_flags;
-
-       /* TODO: We only check for the progressive mode, check for interlace mode too */
-       if(drm_connector->latency_present[0]) {
-               audio_info->video_latency = drm_connector->video_latency[0];
-               audio_info->audio_latency = drm_connector->audio_latency[0];
-       }
-
-       /* TODO: For DP, video and audio latency should be calculated from DPCD caps */
-
-}
-
-static void copy_crtc_timing_for_drm_display_mode(
-               const struct drm_display_mode *src_mode,
-               struct drm_display_mode *dst_mode)
-{
-       dst_mode->crtc_hdisplay = src_mode->crtc_hdisplay;
-       dst_mode->crtc_vdisplay = src_mode->crtc_vdisplay;
-       dst_mode->crtc_clock = src_mode->crtc_clock;
-       dst_mode->crtc_hblank_start = src_mode->crtc_hblank_start;
-       dst_mode->crtc_hblank_end = src_mode->crtc_hblank_end;
-       dst_mode->crtc_hsync_start=  src_mode->crtc_hsync_start;
-       dst_mode->crtc_hsync_end = src_mode->crtc_hsync_end;
-       dst_mode->crtc_htotal = src_mode->crtc_htotal;
-       dst_mode->crtc_hskew = src_mode->crtc_hskew;
-       dst_mode->crtc_vblank_start = src_mode->crtc_vblank_start;
-       dst_mode->crtc_vblank_end = src_mode->crtc_vblank_end;
-       dst_mode->crtc_vsync_start = src_mode->crtc_vsync_start;
-       dst_mode->crtc_vsync_end = src_mode->crtc_vsync_end;
-       dst_mode->crtc_vtotal = src_mode->crtc_vtotal;
-}
-
-static void decide_crtc_timing_for_drm_display_mode(
-               struct drm_display_mode *drm_mode,
-               const struct drm_display_mode *native_mode,
-               bool scale_enabled)
-{
-       if (scale_enabled) {
-               copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
-       } else if (native_mode->clock == drm_mode->clock &&
-                       native_mode->htotal == drm_mode->htotal &&
-                       native_mode->vtotal == drm_mode->vtotal) {
-               copy_crtc_timing_for_drm_display_mode(native_mode, drm_mode);
-       } else {
-               /* no scaling nor amdgpu inserted, no need to patch */
-       }
-}
-
-static struct dc_stream *create_stream_for_sink(
-               struct amdgpu_connector *aconnector,
-               const struct drm_display_mode *drm_mode,
-               const struct dm_connector_state *dm_state)
-{
-       struct drm_display_mode *preferred_mode = NULL;
-       const struct drm_connector *drm_connector;
-       struct dc_stream *stream = NULL;
-       struct drm_display_mode mode = *drm_mode;
-       bool native_mode_found = false;
-
-       if (NULL == aconnector) {
-               DRM_ERROR("aconnector is NULL!\n");
-               goto drm_connector_null;
-       }
-
-       if (NULL == dm_state) {
-               DRM_ERROR("dm_state is NULL!\n");
-               goto dm_state_null;
-       }
-
-       drm_connector = &aconnector->base;
-       stream = dc_create_stream_for_sink(aconnector->dc_sink);
-
-       if (NULL == stream) {
-               DRM_ERROR("Failed to create stream for sink!\n");
-               goto stream_create_fail;
-       }
-
-       list_for_each_entry(preferred_mode, &aconnector->base.modes, head) {
-               /* Search for preferred mode */
-               if (preferred_mode->type & DRM_MODE_TYPE_PREFERRED) {
-                       native_mode_found = true;
-                       break;
-               }
-       }
-       if (!native_mode_found)
-               preferred_mode = list_first_entry_or_null(
-                               &aconnector->base.modes,
-                               struct drm_display_mode,
-                               head);
-
-       if (NULL == preferred_mode) {
-               /* This may not be an error, the use case is when we we have no
-                * usermode calls to reset and set mode upon hotplug. In this
-                * case, we call set mode ourselves to restore the previous mode
-                * and the modelist may not be filled in in time.
-                */
-               DRM_INFO("No preferred mode found\n");
-       } else {
-               decide_crtc_timing_for_drm_display_mode(
-                               &mode, preferred_mode,
-                               dm_state->scaling != RMX_OFF);
-       }
-
-       fill_stream_properties_from_drm_display_mode(stream,
-                       &mode, &aconnector->base);
-       update_stream_scaling_settings(&mode, dm_state, stream);
-
-       fill_audio_info(
-               &stream->audio_info,
-               drm_connector,
-               aconnector->dc_sink);
-
-stream_create_fail:
-dm_state_null:
-drm_connector_null:
-       return stream;
-}
-
-void amdgpu_dm_crtc_destroy(struct drm_crtc *crtc)
-{
-       drm_crtc_cleanup(crtc);
-       kfree(crtc);
-}
-
-static void dm_crtc_destroy_state(struct drm_crtc *crtc,
-                                          struct drm_crtc_state *state)
-{
-       struct dm_crtc_state *cur = to_dm_crtc_state(state);
-
-       /* TODO Destroy dc_stream objects are stream object is flattened */
-       if (cur->stream)
-               dc_stream_release(cur->stream);
-
-
-       __drm_atomic_helper_crtc_destroy_state(state);
-
-
-       kfree(state);
-}
-
-static void dm_crtc_reset_state(struct drm_crtc *crtc)
-{
-       struct dm_crtc_state *state;
-
-       if (crtc->state)
-               dm_crtc_destroy_state(crtc, crtc->state);
-
-       state = kzalloc(sizeof(*state), GFP_KERNEL);
-       if (WARN_ON(!state))
-               return;
-
-       crtc->state = &state->base;
-       crtc->state->crtc = crtc;
-
-}
-
-static struct drm_crtc_state *
-dm_crtc_duplicate_state(struct drm_crtc *crtc)
-{
-       struct dm_crtc_state *state, *cur;
-
-       cur = to_dm_crtc_state(crtc->state);
-
-       if (WARN_ON(!crtc->state))
-               return NULL;
-
-       state = dm_alloc(sizeof(*state));
-
-       __drm_atomic_helper_crtc_duplicate_state(crtc, &state->base);
-
-       if (cur->stream) {
-               state->stream = cur->stream;
-               dc_stream_retain(state->stream);
-       }
-
-       /* TODO Duplicate dc_stream after objects are stream object is flattened */
-
-       return &state->base;
-}
-
-/* Implemented only the options currently availible for the driver */
-static const struct drm_crtc_funcs amdgpu_dm_crtc_funcs = {
-       .reset = dm_crtc_reset_state,
-       .destroy = amdgpu_dm_crtc_destroy,
-       .gamma_set = drm_atomic_helper_legacy_gamma_set,
-       .set_config = drm_atomic_helper_set_config,
-       .page_flip = drm_atomic_helper_page_flip,
-       .atomic_duplicate_state = dm_crtc_duplicate_state,
-       .atomic_destroy_state = dm_crtc_destroy_state,
-};
-
-static enum drm_connector_status
-amdgpu_dm_connector_detect(struct drm_connector *connector, bool force)
-{
-       bool connected;
-       struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
-
-       /* Notes:
-        * 1. This interface is NOT called in context of HPD irq.
-        * 2. This interface *is called* in context of user-mode ioctl. Which
-        * makes it a bad place for *any* MST-related activit. */
-
-       if (aconnector->base.force == DRM_FORCE_UNSPECIFIED)
-               connected = (aconnector->dc_sink != NULL);
-       else
-               connected = (aconnector->base.force == DRM_FORCE_ON);
-
-       return (connected ? connector_status_connected :
-                       connector_status_disconnected);
-}
-
-int amdgpu_dm_connector_atomic_set_property(
-       struct drm_connector *connector,
-       struct drm_connector_state *connector_state,
-       struct drm_property *property,
-       uint64_t val)
-{
-       struct drm_device *dev = connector->dev;
-       struct amdgpu_device *adev = dev->dev_private;
-       struct dm_connector_state *dm_old_state =
-               to_dm_connector_state(connector->state);
-       struct dm_connector_state *dm_new_state =
-               to_dm_connector_state(connector_state);
-
-       int ret = -EINVAL;
-
-       if (property == dev->mode_config.scaling_mode_property) {
-               enum amdgpu_rmx_type rmx_type;
-
-               switch (val) {
-               case DRM_MODE_SCALE_CENTER:
-                       rmx_type = RMX_CENTER;
-                       break;
-               case DRM_MODE_SCALE_ASPECT:
-                       rmx_type = RMX_ASPECT;
-                       break;
-               case DRM_MODE_SCALE_FULLSCREEN:
-                       rmx_type = RMX_FULL;
-                       break;
-               case DRM_MODE_SCALE_NONE:
-               default:
-                       rmx_type = RMX_OFF;
-                       break;
-               }
-
-               if (dm_old_state->scaling == rmx_type)
-                       return 0;
-
-               dm_new_state->scaling = rmx_type;
-               ret = 0;
-       } else if (property == adev->mode_info.underscan_hborder_property) {
-               dm_new_state->underscan_hborder = val;
-               ret = 0;
-       } else if (property == adev->mode_info.underscan_vborder_property) {
-               dm_new_state->underscan_vborder = val;
-               ret = 0;
-       } else if (property == adev->mode_info.underscan_property) {
-               dm_new_state->underscan_enable = val;
-               ret = 0;
-       }
-
-       return ret;
-}
-
-int amdgpu_dm_connector_atomic_get_property(
-       struct drm_connector *connector,
-       const struct drm_connector_state *state,
-       struct drm_property *property,
-       uint64_t *val)
-{
-       struct drm_device *dev = connector->dev;
-       struct amdgpu_device *adev = dev->dev_private;
-       struct dm_connector_state *dm_state =
-               to_dm_connector_state(state);
-       int ret = -EINVAL;
-
-       if (property == dev->mode_config.scaling_mode_property) {
-               switch (dm_state->scaling) {
-               case RMX_CENTER:
-                       *val = DRM_MODE_SCALE_CENTER;
-                       break;
-               case RMX_ASPECT:
-                       *val = DRM_MODE_SCALE_ASPECT;
-                       break;
-               case RMX_FULL:
-                       *val = DRM_MODE_SCALE_FULLSCREEN;
-                       break;
-               case RMX_OFF:
-               default:
-                       *val = DRM_MODE_SCALE_NONE;
-                       break;
-               }
-               ret = 0;
-       } else if (property == adev->mode_info.underscan_hborder_property) {
-               *val = dm_state->underscan_hborder;
-               ret = 0;
-       } else if (property == adev->mode_info.underscan_vborder_property) {
-               *val = dm_state->underscan_vborder;
-               ret = 0;
-       } else if (property == adev->mode_info.underscan_property) {
-               *val = dm_state->underscan_enable;
-               ret = 0;
-       }
-       return ret;
-}
-
-void amdgpu_dm_connector_destroy(struct drm_connector *connector)
-{
-       struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
-       const struct dc_link *link = aconnector->dc_link;
-       struct amdgpu_device *adev = connector->dev->dev_private;
-       struct amdgpu_display_manager *dm = &adev->dm;
-#if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) ||\
-       defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE)
-
-       if (link->connector_signal & (SIGNAL_TYPE_EDP | SIGNAL_TYPE_LVDS)) {
-               amdgpu_dm_register_backlight_device(dm);
-
-               if (dm->backlight_dev) {
-                       backlight_device_unregister(dm->backlight_dev);
-                       dm->backlight_dev = NULL;
-               }
-
-       }
-#endif
-       drm_connector_unregister(connector);
-       drm_connector_cleanup(connector);
-       kfree(connector);
-}
-
-void amdgpu_dm_connector_funcs_reset(struct drm_connector *connector)
-{
-       struct dm_connector_state *state =
-               to_dm_connector_state(connector->state);
-
-       kfree(state);
-
-       state = kzalloc(sizeof(*state), GFP_KERNEL);
-
-       if (state) {
-               state->scaling = RMX_OFF;
-               state->underscan_enable = false;
-               state->underscan_hborder = 0;
-               state->underscan_vborder = 0;
-
-               connector->state = &state->base;
-               connector->state->connector = connector;
-       }
-}
-
-struct drm_connector_state *amdgpu_dm_connector_atomic_duplicate_state(
-       struct drm_connector *connector)
-{
-       struct dm_connector_state *state =
-               to_dm_connector_state(connector->state);
-
-       struct dm_connector_state *new_state =
-                       kmemdup(state, sizeof(*state), GFP_KERNEL);
-
-       if (new_state) {
-               __drm_atomic_helper_connector_duplicate_state(connector,
-                                                                     &new_state->base);
-               return &new_state->base;
-       }
-
-       return NULL;
-}
-
-static const struct drm_connector_funcs amdgpu_dm_connector_funcs = {
-       .reset = amdgpu_dm_connector_funcs_reset,
-       .detect = amdgpu_dm_connector_detect,
-       .fill_modes = drm_helper_probe_single_connector_modes,
-       .destroy = amdgpu_dm_connector_destroy,
-       .atomic_duplicate_state = amdgpu_dm_connector_atomic_duplicate_state,
-       .atomic_destroy_state = drm_atomic_helper_connector_destroy_state,
-       .atomic_set_property = amdgpu_dm_connector_atomic_set_property,
-       .atomic_get_property = amdgpu_dm_connector_atomic_get_property
-};
-
-static struct drm_encoder *best_encoder(struct drm_connector *connector)
-{
-       int enc_id = connector->encoder_ids[0];
-       struct drm_mode_object *obj;
-       struct drm_encoder *encoder;
-
-       DRM_DEBUG_KMS("Finding the best encoder\n");
-
-       /* pick the encoder ids */
-       if (enc_id) {
-               obj = drm_mode_object_find(connector->dev, enc_id, DRM_MODE_OBJECT_ENCODER);
-               if (!obj) {
-                       DRM_ERROR("Couldn't find a matching encoder for our connector\n");
-                       return NULL;
-               }
-               encoder = obj_to_encoder(obj);
-               return encoder;
-       }
-       DRM_ERROR("No encoder id\n");
-       return NULL;
-}
-
-static int get_modes(struct drm_connector *connector)
-{
-       return amdgpu_dm_connector_get_modes(connector);
-}
-
-static void create_eml_sink(struct amdgpu_connector *aconnector)
-{
-       struct dc_sink_init_data init_params = {
-                       .link = aconnector->dc_link,
-                       .sink_signal = SIGNAL_TYPE_VIRTUAL
-       };
-       struct edid *edid = (struct edid *) aconnector->base.edid_blob_ptr->data;
-
-       if (!aconnector->base.edid_blob_ptr ||
-               !aconnector->base.edid_blob_ptr->data) {
-               DRM_ERROR("No EDID firmware found on connector: %s ,forcing to OFF!\n",
-                               aconnector->base.name);
-
-               aconnector->base.force = DRM_FORCE_OFF;
-               aconnector->base.override_edid = false;
-               return;
-       }
-
-       aconnector->edid = edid;
-
-       aconnector->dc_em_sink = dc_link_add_remote_sink(
-               aconnector->dc_link,
-               (uint8_t *)edid,
-               (edid->extensions + 1) * EDID_LENGTH,
-               &init_params);
-
-       if (aconnector->base.force
-                                       == DRM_FORCE_ON)
-               aconnector->dc_sink = aconnector->dc_link->local_sink ?
-               aconnector->dc_link->local_sink :
-               aconnector->dc_em_sink;
-}
-
-static void handle_edid_mgmt(struct amdgpu_connector *aconnector)
-{
-       struct dc_link *link = (struct dc_link *)aconnector->dc_link;
-
-       /* In case of headless boot with force on for DP managed connector
-        * Those settings have to be != 0 to get initial modeset
-        */
-       if (link->connector_signal == SIGNAL_TYPE_DISPLAY_PORT) {
-               link->verified_link_cap.lane_count = LANE_COUNT_FOUR;
-               link->verified_link_cap.link_rate = LINK_RATE_HIGH2;
-       }
-
-
-       aconnector->base.override_edid = true;
-       create_eml_sink(aconnector);
-}
-
-int amdgpu_dm_connector_mode_valid(
-               struct drm_connector *connector,
-               struct drm_display_mode *mode)
-{
-       int result = MODE_ERROR;
-       struct dc_sink *dc_sink;
-       struct amdgpu_device *adev = connector->dev->dev_private;
-       /* TODO: Unhardcode stream count */
-       struct dc_stream *stream;
-       struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
-
-       if ((mode->flags & DRM_MODE_FLAG_INTERLACE) ||
-                       (mode->flags & DRM_MODE_FLAG_DBLSCAN))
-               return result;
-
-       /* Only run this the first time mode_valid is called to initilialize
-        * EDID mgmt
-        */
-       if (aconnector->base.force != DRM_FORCE_UNSPECIFIED &&
-               !aconnector->dc_em_sink)
-               handle_edid_mgmt(aconnector);
-
-       dc_sink = to_amdgpu_connector(connector)->dc_sink;
-
-       if (NULL == dc_sink) {
-               DRM_ERROR("dc_sink is NULL!\n");
-               goto fail;
-       }
-
-       stream = dc_create_stream_for_sink(dc_sink);
-       if (NULL == stream) {
-               DRM_ERROR("Failed to create stream for sink!\n");
-               goto fail;
-       }
-
-       drm_mode_set_crtcinfo(mode, 0);
-       fill_stream_properties_from_drm_display_mode(stream, mode, connector);
-
-       stream->src.width = mode->hdisplay;
-       stream->src.height = mode->vdisplay;
-       stream->dst = stream->src;
-
-       if (dc_validate_stream(adev->dm.dc, stream))
-               result = MODE_OK;
-
-       dc_stream_release(stream);
-
-fail:
-       /* TODO: error handling*/
-       return result;
-}
-
-static const struct drm_connector_helper_funcs
-amdgpu_dm_connector_helper_funcs = {
-       /*
-       * If hotplug a second bigger display in FB Con mode, bigger resolution
-       * modes will be filtered by drm_mode_validate_size(), and those modes
-       * is missing after user start lightdm. So we need to renew modes list.
-       * in get_modes call back, not just return the modes count
-       */
-       .get_modes = get_modes,
-       .mode_valid = amdgpu_dm_connector_mode_valid,
-       .best_encoder = best_encoder
-};
-
-static void dm_crtc_helper_disable(struct drm_crtc *crtc)
-{
-}
-
-static int dm_crtc_helper_atomic_check(
-       struct drm_crtc *crtc,
-       struct drm_crtc_state *state)
-{
-       struct amdgpu_device *adev = crtc->dev->dev_private;
-       struct dc *dc = adev->dm.dc;
-       struct dm_crtc_state *dm_crtc_state = to_dm_crtc_state(state);
-       int ret = -EINVAL;
-
-       if (unlikely(!dm_crtc_state->stream && modeset_required(state))) {
-               WARN_ON(1);
-               return ret;
-       }
-
-       /* In some use cases, like reset, no stream  is attached */
-       if (!dm_crtc_state->stream)
-               return 0;
-
-       if (dc_validate_stream(dc, dm_crtc_state->stream))
-               return 0;
-
-       return ret;
-}
-
-static bool dm_crtc_helper_mode_fixup(
-       struct drm_crtc *crtc,
-       const struct drm_display_mode *mode,
-       struct drm_display_mode *adjusted_mode)
-{
-       return true;
-}
-
-static const struct drm_crtc_helper_funcs amdgpu_dm_crtc_helper_funcs = {
-       .disable = dm_crtc_helper_disable,
-       .atomic_check = dm_crtc_helper_atomic_check,
-       .mode_fixup = dm_crtc_helper_mode_fixup
-};
-
-static void dm_encoder_helper_disable(struct drm_encoder *encoder)
-{
-
-}
-
-static int dm_encoder_helper_atomic_check(
-       struct drm_encoder *encoder,
-       struct drm_crtc_state *crtc_state,
-       struct drm_connector_state *conn_state)
-{
-       return 0;
-}
-
-const struct drm_encoder_helper_funcs amdgpu_dm_encoder_helper_funcs = {
-       .disable = dm_encoder_helper_disable,
-       .atomic_check = dm_encoder_helper_atomic_check
-};
-
-static void dm_drm_plane_reset(struct drm_plane *plane)
-{
-       struct dm_plane_state *amdgpu_state = NULL;
-
-       if (plane->state)
-               plane->funcs->atomic_destroy_state(plane, plane->state);
-
-       amdgpu_state = kzalloc(sizeof(*amdgpu_state), GFP_KERNEL);
-
-       if (amdgpu_state) {
-               plane->state = &amdgpu_state->base;
-               plane->state->plane = plane;
-               plane->state->rotation = DRM_MODE_ROTATE_0;
-       }
-       else
-               WARN_ON(1);
-}
-
-static struct drm_plane_state *
-dm_drm_plane_duplicate_state(struct drm_plane *plane)
-{
-       struct dm_plane_state *dm_plane_state, *old_dm_plane_state;
-
-       old_dm_plane_state = to_dm_plane_state(plane->state);
-       dm_plane_state = kzalloc(sizeof(*dm_plane_state), GFP_KERNEL);
-       if (!dm_plane_state)
-               return NULL;
-
-       __drm_atomic_helper_plane_duplicate_state(plane, &dm_plane_state->base);
-
-       if (old_dm_plane_state->surface) {
-               dm_plane_state->surface = old_dm_plane_state->surface;
-               dc_surface_retain(dm_plane_state->surface);
-       }
-
-       return &dm_plane_state->base;
-}
-
-void dm_drm_plane_destroy_state(struct drm_plane *plane,
-                                          struct drm_plane_state *state)
-{
-       struct dm_plane_state *dm_plane_state = to_dm_plane_state(state);
-
-       if (dm_plane_state->surface)
-               dc_surface_release(dm_plane_state->surface);
-
-       __drm_atomic_helper_plane_destroy_state(state);
-       kfree(dm_plane_state);
-}
-
-static const struct drm_plane_funcs dm_plane_funcs = {
-       .update_plane   = drm_atomic_helper_update_plane,
-       .disable_plane  = drm_atomic_helper_disable_plane,
-       .destroy        = drm_plane_cleanup,
-       .reset = dm_drm_plane_reset,
-       .atomic_duplicate_state = dm_drm_plane_duplicate_state,
-       .atomic_destroy_state = dm_drm_plane_destroy_state,
-};
-
-static int dm_plane_helper_prepare_fb(
-       struct drm_plane *plane,
-       struct drm_plane_state *new_state)
-{
-       struct amdgpu_framebuffer *afb;
-       struct drm_gem_object *obj;
-       struct amdgpu_bo *rbo;
-       int r;
-       struct dm_plane_state *dm_plane_state_new, *dm_plane_state_old;
-       unsigned int awidth;
-
-       dm_plane_state_old = to_dm_plane_state(plane->state);
-       dm_plane_state_new = to_dm_plane_state(new_state);
-
-       if (!new_state->fb) {
-               DRM_DEBUG_KMS("No FB bound\n");
-               return 0;
-       }
-
-       afb = to_amdgpu_framebuffer(new_state->fb);
-
-       obj = afb->obj;
-       rbo = gem_to_amdgpu_bo(obj);
-       r = amdgpu_bo_reserve(rbo, false);
-       if (unlikely(r != 0))
-               return r;
-
-       r = amdgpu_bo_pin(rbo, AMDGPU_GEM_DOMAIN_VRAM, &afb->address);
-
-
-       amdgpu_bo_unreserve(rbo);
-
-       if (unlikely(r != 0)) {
-               DRM_ERROR("Failed to pin framebuffer\n");
-               return r;
-       }
-
-       amdgpu_bo_ref(rbo);
-
-       if (dm_plane_state_new->surface &&
-                       dm_plane_state_old->surface != dm_plane_state_new->surface) {
-               struct dc_surface *surface = dm_plane_state_new->surface;
-
-               if (surface->format < SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) {
-                       surface->address.grph.addr.low_part = lower_32_bits(afb->address);
-                       surface->address.grph.addr.high_part = upper_32_bits(afb->address);
-               } else {
-                       awidth = ALIGN(new_state->fb->width, 64);
-                       surface->address.video_progressive.luma_addr.low_part
-                                                       = lower_32_bits(afb->address);
-                       surface->address.video_progressive.chroma_addr.low_part
-                                                       = lower_32_bits(afb->address) +
-                                                               (awidth * new_state->fb->height);
-               }
-       }
-
-       /* It's a hack for s3 since in 4.9 kernel filter out cursor buffer
-        * prepare and cleanup in drm_atomic_helper_prepare_planes
-        * and drm_atomic_helper_cleanup_planes because fb doens't in s3.
-        * IN 4.10 kernel this code should be removed and amdgpu_device_suspend
-        * code touching fram buffers should be avoided for DC.
-        */
-       if (plane->type == DRM_PLANE_TYPE_CURSOR) {
-               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(new_state->crtc);
-
-               acrtc->cursor_bo = obj;
-       }
-       return 0;
-}
-
-static void dm_plane_helper_cleanup_fb(
-       struct drm_plane *plane,
-       struct drm_plane_state *old_state)
-{
-       struct amdgpu_bo *rbo;
-       struct amdgpu_framebuffer *afb;
-       int r;
-
-       if (!old_state->fb)
-               return;
-
-       afb = to_amdgpu_framebuffer(old_state->fb);
-       rbo = gem_to_amdgpu_bo(afb->obj);
-       r = amdgpu_bo_reserve(rbo, false);
-       if (unlikely(r)) {
-               DRM_ERROR("failed to reserve rbo before unpin\n");
-               return;
-       } else {
-               amdgpu_bo_unpin(rbo);
-               amdgpu_bo_unreserve(rbo);
-               amdgpu_bo_unref(&rbo);
-       };
-}
-
-int dm_create_validation_set_for_connector(struct drm_connector *connector,
-               struct drm_display_mode *mode, struct dc_validation_set *val_set)
-{
-       int result = MODE_ERROR;
-       struct dc_sink *dc_sink =
-                       to_amdgpu_connector(connector)->dc_sink;
-       /* TODO: Unhardcode stream count */
-       struct dc_stream *stream;
-
-       if ((mode->flags & DRM_MODE_FLAG_INTERLACE) ||
-                       (mode->flags & DRM_MODE_FLAG_DBLSCAN))
-               return result;
-
-       if (NULL == dc_sink) {
-               DRM_ERROR("dc_sink is NULL!\n");
-               return result;
-       }
-
-       stream = dc_create_stream_for_sink(dc_sink);
-
-       if (NULL == stream) {
-               DRM_ERROR("Failed to create stream for sink!\n");
-               return result;
-       }
-
-       drm_mode_set_crtcinfo(mode, 0);
-
-       fill_stream_properties_from_drm_display_mode(stream, mode, connector);
-
-       val_set->stream = stream;
-
-       stream->src.width = mode->hdisplay;
-       stream->src.height = mode->vdisplay;
-       stream->dst = stream->src;
-
-       return MODE_OK;
-}
-
-static const struct drm_plane_helper_funcs dm_plane_helper_funcs = {
-       .prepare_fb = dm_plane_helper_prepare_fb,
-       .cleanup_fb = dm_plane_helper_cleanup_fb,
-};
-
-/*
- * TODO: these are currently initialized to rgb formats only.
- * For future use cases we should either initialize them dynamically based on
- * plane capabilities, or initialize this array to all formats, so internal drm
- * check will succeed, and let DC to implement proper check
- */
-static uint32_t rgb_formats[] = {
-       DRM_FORMAT_RGB888,
-       DRM_FORMAT_XRGB8888,
-       DRM_FORMAT_ARGB8888,
-       DRM_FORMAT_RGBA8888,
-       DRM_FORMAT_XRGB2101010,
-       DRM_FORMAT_XBGR2101010,
-       DRM_FORMAT_ARGB2101010,
-       DRM_FORMAT_ABGR2101010,
-};
-
-static uint32_t yuv_formats[] = {
-       DRM_FORMAT_NV12,
-       DRM_FORMAT_NV21,
-};
-
-static const u32 cursor_formats[] = {
-       DRM_FORMAT_ARGB8888
-};
-
-int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
-                       struct amdgpu_plane *aplane,
-                       unsigned long possible_crtcs)
-{
-       int res = -EPERM;
-
-       switch (aplane->base.type) {
-       case DRM_PLANE_TYPE_PRIMARY:
-               aplane->base.format_default = true;
-
-               res = drm_universal_plane_init(
-                               dm->adev->ddev,
-                               &aplane->base,
-                               possible_crtcs,
-                               &dm_plane_funcs,
-                               rgb_formats,
-                               ARRAY_SIZE(rgb_formats),
-                               NULL, aplane->base.type, NULL);
-               break;
-       case DRM_PLANE_TYPE_OVERLAY:
-               res = drm_universal_plane_init(
-                               dm->adev->ddev,
-                               &aplane->base,
-                               possible_crtcs,
-                               &dm_plane_funcs,
-                               yuv_formats,
-                               ARRAY_SIZE(yuv_formats),
-                               NULL, aplane->base.type, NULL);
-               break;
-       case DRM_PLANE_TYPE_CURSOR:
-               res = drm_universal_plane_init(
-                               dm->adev->ddev,
-                               &aplane->base,
-                               possible_crtcs,
-                               &dm_plane_funcs,
-                               cursor_formats,
-                               ARRAY_SIZE(cursor_formats),
-                               NULL, aplane->base.type, NULL);
-               break;
-       }
-
-       drm_plane_helper_add(&aplane->base, &dm_plane_helper_funcs);
-
-       return res;
-}
-
-int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm,
-                       struct drm_plane *plane,
-                       uint32_t crtc_index)
-{
-       struct amdgpu_crtc *acrtc = NULL;
-       struct amdgpu_plane *cursor_plane;
-
-       int res = -ENOMEM;
-
-       cursor_plane = kzalloc(sizeof(*cursor_plane), GFP_KERNEL);
-       if (!cursor_plane)
-               goto fail;
-
-       cursor_plane->base.type = DRM_PLANE_TYPE_CURSOR;
-       res = amdgpu_dm_plane_init(dm, cursor_plane, 0);
-
-       acrtc = kzalloc(sizeof(struct amdgpu_crtc), GFP_KERNEL);
-       if (!acrtc)
-               goto fail;
-
-       res = drm_crtc_init_with_planes(
-                       dm->ddev,
-                       &acrtc->base,
-                       plane,
-                       &cursor_plane->base,
-                       &amdgpu_dm_crtc_funcs, NULL);
-
-       if (res)
-               goto fail;
-
-       drm_crtc_helper_add(&acrtc->base, &amdgpu_dm_crtc_helper_funcs);
-
-       acrtc->max_cursor_width = dm->adev->dm.dc->caps.max_cursor_size;
-       acrtc->max_cursor_height = dm->adev->dm.dc->caps.max_cursor_size;
-
-       acrtc->crtc_id = crtc_index;
-       acrtc->base.enabled = false;
-
-       dm->adev->mode_info.crtcs[crtc_index] = acrtc;
-       drm_mode_crtc_set_gamma_size(&acrtc->base, 256);
-
-       return 0;
-
-fail:
-       if (acrtc)
-               kfree(acrtc);
-       if (cursor_plane)
-               kfree(cursor_plane);
-       acrtc->crtc_id = -1;
-       return res;
-}
-
-
-static int to_drm_connector_type(enum signal_type st)
-{
-       switch (st) {
-       case SIGNAL_TYPE_HDMI_TYPE_A:
-               return DRM_MODE_CONNECTOR_HDMIA;
-       case SIGNAL_TYPE_EDP:
-               return DRM_MODE_CONNECTOR_eDP;
-       case SIGNAL_TYPE_RGB:
-               return DRM_MODE_CONNECTOR_VGA;
-       case SIGNAL_TYPE_DISPLAY_PORT:
-       case SIGNAL_TYPE_DISPLAY_PORT_MST:
-               return DRM_MODE_CONNECTOR_DisplayPort;
-       case SIGNAL_TYPE_DVI_DUAL_LINK:
-       case SIGNAL_TYPE_DVI_SINGLE_LINK:
-               return DRM_MODE_CONNECTOR_DVID;
-       case SIGNAL_TYPE_VIRTUAL:
-               return DRM_MODE_CONNECTOR_VIRTUAL;
-
-       default:
-               return DRM_MODE_CONNECTOR_Unknown;
-       }
-}
-
-static void amdgpu_dm_get_native_mode(struct drm_connector *connector)
-{
-       const struct drm_connector_helper_funcs *helper =
-               connector->helper_private;
-       struct drm_encoder *encoder;
-       struct amdgpu_encoder *amdgpu_encoder;
-
-       encoder = helper->best_encoder(connector);
-
-       if (encoder == NULL)
-               return;
-
-       amdgpu_encoder = to_amdgpu_encoder(encoder);
-
-       amdgpu_encoder->native_mode.clock = 0;
-
-       if (!list_empty(&connector->probed_modes)) {
-               struct drm_display_mode *preferred_mode = NULL;
-               list_for_each_entry(preferred_mode,
-                               &connector->probed_modes,
-                               head) {
-               if (preferred_mode->type & DRM_MODE_TYPE_PREFERRED) {
-                       amdgpu_encoder->native_mode = *preferred_mode;
-               }
-                       break;
-               }
-
-       }
-}
-
-static struct drm_display_mode *amdgpu_dm_create_common_mode(
-               struct drm_encoder *encoder, char *name,
-               int hdisplay, int vdisplay)
-{
-       struct drm_device *dev = encoder->dev;
-       struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-       struct drm_display_mode *mode = NULL;
-       struct drm_display_mode *native_mode = &amdgpu_encoder->native_mode;
-
-       mode = drm_mode_duplicate(dev, native_mode);
-
-       if(mode == NULL)
-               return NULL;
-
-       mode->hdisplay = hdisplay;
-       mode->vdisplay = vdisplay;
-       mode->type &= ~DRM_MODE_TYPE_PREFERRED;
-       strncpy(mode->name, name, DRM_DISPLAY_MODE_LEN);
-
-       return mode;
-
-}
-
-static void amdgpu_dm_connector_add_common_modes(struct drm_encoder *encoder,
-                                       struct drm_connector *connector)
-{
-       struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-       struct drm_display_mode *mode = NULL;
-       struct drm_display_mode *native_mode = &amdgpu_encoder->native_mode;
-       struct amdgpu_connector *amdgpu_connector =
-                               to_amdgpu_connector(connector);
-       int i;
-       int n;
-       struct mode_size {
-               char name[DRM_DISPLAY_MODE_LEN];
-               int w;
-               int h;
-       }common_modes[] = {
-               {  "640x480",  640,  480},
-               {  "800x600",  800,  600},
-               { "1024x768", 1024,  768},
-               { "1280x720", 1280,  720},
-               { "1280x800", 1280,  800},
-               {"1280x1024", 1280, 1024},
-               { "1440x900", 1440,  900},
-               {"1680x1050", 1680, 1050},
-               {"1600x1200", 1600, 1200},
-               {"1920x1080", 1920, 1080},
-               {"1920x1200", 1920, 1200}
-       };
-
-       n = sizeof(common_modes) / sizeof(common_modes[0]);
-
-       for (i = 0; i < n; i++) {
-               struct drm_display_mode *curmode = NULL;
-               bool mode_existed = false;
-
-               if (common_modes[i].w > native_mode->hdisplay ||
-                       common_modes[i].h > native_mode->vdisplay ||
-                       (common_modes[i].w == native_mode->hdisplay &&
-                       common_modes[i].h == native_mode->vdisplay))
-                               continue;
-
-               list_for_each_entry(curmode, &connector->probed_modes, head) {
-                       if (common_modes[i].w == curmode->hdisplay &&
-                               common_modes[i].h == curmode->vdisplay) {
-                               mode_existed = true;
-                               break;
-                       }
-               }
-
-               if (mode_existed)
-                       continue;
-
-               mode = amdgpu_dm_create_common_mode(encoder,
-                               common_modes[i].name, common_modes[i].w,
-                               common_modes[i].h);
-               drm_mode_probed_add(connector, mode);
-               amdgpu_connector->num_modes++;
-       }
-}
-
-static void amdgpu_dm_connector_ddc_get_modes(
-       struct drm_connector *connector,
-       struct edid *edid)
-{
-       struct amdgpu_connector *amdgpu_connector =
-                       to_amdgpu_connector(connector);
-
-       if (edid) {
-               /* empty probed_modes */
-               INIT_LIST_HEAD(&connector->probed_modes);
-               amdgpu_connector->num_modes =
-                               drm_add_edid_modes(connector, edid);
-
-               drm_edid_to_eld(connector, edid);
-
-               amdgpu_dm_get_native_mode(connector);
-       } else
-               amdgpu_connector->num_modes = 0;
-}
-
-int amdgpu_dm_connector_get_modes(struct drm_connector *connector)
-{
-       const struct drm_connector_helper_funcs *helper =
-                       connector->helper_private;
-       struct amdgpu_connector *amdgpu_connector =
-                       to_amdgpu_connector(connector);
-       struct drm_encoder *encoder;
-       struct edid *edid = amdgpu_connector->edid;
-
-       encoder = helper->best_encoder(connector);
-
-       amdgpu_dm_connector_ddc_get_modes(connector, edid);
-       amdgpu_dm_connector_add_common_modes(encoder, connector);
-       return amdgpu_connector->num_modes;
-}
-
-void amdgpu_dm_connector_init_helper(
-       struct amdgpu_display_manager *dm,
-       struct amdgpu_connector *aconnector,
-       int connector_type,
-       struct dc_link *link,
-       int link_index)
-{
-       struct amdgpu_device *adev = dm->ddev->dev_private;
-
-       aconnector->connector_id = link_index;
-       aconnector->dc_link = link;
-       aconnector->base.interlace_allowed = false;
-       aconnector->base.doublescan_allowed = false;
-       aconnector->base.stereo_allowed = false;
-       aconnector->base.dpms = DRM_MODE_DPMS_OFF;
-       aconnector->hpd.hpd = AMDGPU_HPD_NONE; /* not used */
-
-       mutex_init(&aconnector->hpd_lock);
-
-       /*configure suport HPD hot plug connector_>polled default value is 0
-        * which means HPD hot plug not supported*/
-       switch (connector_type) {
-       case DRM_MODE_CONNECTOR_HDMIA:
-               aconnector->base.polled = DRM_CONNECTOR_POLL_HPD;
-               break;
-       case DRM_MODE_CONNECTOR_DisplayPort:
-               aconnector->base.polled = DRM_CONNECTOR_POLL_HPD;
-               break;
-       case DRM_MODE_CONNECTOR_DVID:
-               aconnector->base.polled = DRM_CONNECTOR_POLL_HPD;
-               break;
-       default:
-               break;
-       }
-
-       drm_object_attach_property(&aconnector->base.base,
-                               dm->ddev->mode_config.scaling_mode_property,
-                               DRM_MODE_SCALE_NONE);
-
-       drm_object_attach_property(&aconnector->base.base,
-                               adev->mode_info.underscan_property,
-                               UNDERSCAN_OFF);
-       drm_object_attach_property(&aconnector->base.base,
-                               adev->mode_info.underscan_hborder_property,
-                               0);
-       drm_object_attach_property(&aconnector->base.base,
-                               adev->mode_info.underscan_vborder_property,
-                               0);
-
-}
-
-int amdgpu_dm_i2c_xfer(struct i2c_adapter *i2c_adap,
-                     struct i2c_msg *msgs, int num)
-{
-       struct amdgpu_i2c_adapter *i2c = i2c_get_adapdata(i2c_adap);
-       struct ddc_service *ddc_service = i2c->ddc_service;
-       struct i2c_command cmd;
-       int i;
-       int result = -EIO;
-
-       cmd.payloads = kzalloc(num * sizeof(struct i2c_payload), GFP_KERNEL);
-
-       if (!cmd.payloads)
-               return result;
-
-       cmd.number_of_payloads = num;
-       cmd.engine = I2C_COMMAND_ENGINE_DEFAULT;
-       cmd.speed = 100;
-
-       for (i = 0; i < num; i++) {
-               cmd.payloads[i].write = !(msgs[i].flags & I2C_M_RD);
-               cmd.payloads[i].address = msgs[i].addr;
-               cmd.payloads[i].length = msgs[i].len;
-               cmd.payloads[i].data = msgs[i].buf;
-       }
-
-       if (dal_i2caux_submit_i2c_command(
-                       ddc_service->ctx->i2caux,
-                       ddc_service->ddc_pin,
-                       &cmd))
-               result = num;
-
-       kfree(cmd.payloads);
-       return result;
-}
-
-u32 amdgpu_dm_i2c_func(struct i2c_adapter *adap)
-{
-       return I2C_FUNC_I2C | I2C_FUNC_SMBUS_EMUL;
-}
-
-static const struct i2c_algorithm amdgpu_dm_i2c_algo = {
-       .master_xfer = amdgpu_dm_i2c_xfer,
-       .functionality = amdgpu_dm_i2c_func,
-};
-
-static struct amdgpu_i2c_adapter *create_i2c(
-               struct ddc_service *ddc_service,
-               int link_index,
-               int *res)
-{
-       struct amdgpu_device *adev = ddc_service->ctx->driver_context;
-       struct amdgpu_i2c_adapter *i2c;
-
-       i2c = kzalloc(sizeof (struct amdgpu_i2c_adapter), GFP_KERNEL);
-       i2c->base.owner = THIS_MODULE;
-       i2c->base.class = I2C_CLASS_DDC;
-       i2c->base.dev.parent = &adev->pdev->dev;
-       i2c->base.algo = &amdgpu_dm_i2c_algo;
-       snprintf(i2c->base.name, sizeof (i2c->base.name), "AMDGPU DM i2c hw bus %d", link_index);
-       i2c_set_adapdata(&i2c->base, i2c);
-       i2c->ddc_service = ddc_service;
-
-       return i2c;
-}
-
-/* Note: this function assumes that dc_link_detect() was called for the
- * dc_link which will be represented by this aconnector. */
-int amdgpu_dm_connector_init(
-       struct amdgpu_display_manager *dm,
-       struct amdgpu_connector *aconnector,
-       uint32_t link_index,
-       struct amdgpu_encoder *aencoder)
-{
-       int res = 0;
-       int connector_type;
-       struct dc *dc = dm->dc;
-       struct dc_link *link = dc_get_link_at_index(dc, link_index);
-       struct amdgpu_i2c_adapter *i2c;
-       ((struct dc_link *)link)->priv = aconnector;
-
-       DRM_DEBUG_KMS("%s()\n", __func__);
-
-       i2c = create_i2c(link->ddc, link->link_index, &res);
-       aconnector->i2c = i2c;
-       res = i2c_add_adapter(&i2c->base);
-
-       if (res) {
-               DRM_ERROR("Failed to register hw i2c %d\n", link->link_index);
-               goto out_free;
-       }
-
-       connector_type = to_drm_connector_type(link->connector_signal);
-
-       res = drm_connector_init(
-                       dm->ddev,
-                       &aconnector->base,
-                       &amdgpu_dm_connector_funcs,
-                       connector_type);
-
-       if (res) {
-               DRM_ERROR("connector_init failed\n");
-               aconnector->connector_id = -1;
-               goto out_free;
-       }
-
-       drm_connector_helper_add(
-                       &aconnector->base,
-                       &amdgpu_dm_connector_helper_funcs);
-
-       amdgpu_dm_connector_init_helper(
-               dm,
-               aconnector,
-               connector_type,
-               link,
-               link_index);
-
-       drm_mode_connector_attach_encoder(
-               &aconnector->base, &aencoder->base);
-
-       drm_connector_register(&aconnector->base);
-
-       if (connector_type == DRM_MODE_CONNECTOR_DisplayPort
-               || connector_type == DRM_MODE_CONNECTOR_eDP)
-               amdgpu_dm_initialize_dp_connector(dm, aconnector);
-
-#if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) ||\
-       defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE)
-
-       /* NOTE: this currently will create backlight device even if a panel
-        * is not connected to the eDP/LVDS connector.
-        *
-        * This is less than ideal but we don't have sink information at this
-        * stage since detection happens after. We can't do detection earlier
-        * since MST detection needs connectors to be created first.
-        */
-       if (link->connector_signal & (SIGNAL_TYPE_EDP | SIGNAL_TYPE_LVDS)) {
-               /* Event if registration failed, we should continue with
-                * DM initialization because not having a backlight control
-                * is better then a black screen. */
-               amdgpu_dm_register_backlight_device(dm);
-
-               if (dm->backlight_dev)
-                       dm->backlight_link = link;
-       }
-#endif
-
-out_free:
-       if (res) {
-               kfree(i2c);
-               aconnector->i2c = NULL;
-       }
-       return res;
-}
-
-int amdgpu_dm_get_encoder_crtc_mask(struct amdgpu_device *adev)
-{
-       switch (adev->mode_info.num_crtc) {
-       case 1:
-               return 0x1;
-       case 2:
-               return 0x3;
-       case 3:
-               return 0x7;
-       case 4:
-               return 0xf;
-       case 5:
-               return 0x1f;
-       case 6:
-       default:
-               return 0x3f;
-       }
-}
-
-int amdgpu_dm_encoder_init(
-       struct drm_device *dev,
-       struct amdgpu_encoder *aencoder,
-       uint32_t link_index)
-{
-       struct amdgpu_device *adev = dev->dev_private;
-
-       int res = drm_encoder_init(dev,
-                                  &aencoder->base,
-                                  &amdgpu_dm_encoder_funcs,
-                                  DRM_MODE_ENCODER_TMDS,
-                                  NULL);
-
-       aencoder->base.possible_crtcs = amdgpu_dm_get_encoder_crtc_mask(adev);
-
-       if (!res)
-               aencoder->encoder_id = link_index;
-       else
-               aencoder->encoder_id = -1;
-
-       drm_encoder_helper_add(&aencoder->base, &amdgpu_dm_encoder_helper_funcs);
-
-       return res;
-}
-
-static void manage_dm_interrupts(
-       struct amdgpu_device *adev,
-       struct amdgpu_crtc *acrtc,
-       bool enable)
-{
-       /*
-        * this is not correct translation but will work as soon as VBLANK
-        * constant is the same as PFLIP
-        */
-       int irq_type =
-               amdgpu_crtc_idx_to_irq_type(
-                       adev,
-                       acrtc->crtc_id);
-
-       if (enable) {
-               drm_crtc_vblank_on(&acrtc->base);
-               amdgpu_irq_get(
-                       adev,
-                       &adev->pageflip_irq,
-                       irq_type);
-       } else {
-
-               amdgpu_irq_put(
-                       adev,
-                       &adev->pageflip_irq,
-                       irq_type);
-               drm_crtc_vblank_off(&acrtc->base);
-       }
-}
-
-static bool is_scaling_state_different(
-               const struct dm_connector_state *dm_state,
-               const struct dm_connector_state *old_dm_state)
-{
-       if (dm_state->scaling != old_dm_state->scaling)
-               return true;
-       if (!dm_state->underscan_enable && old_dm_state->underscan_enable) {
-               if (old_dm_state->underscan_hborder != 0 && old_dm_state->underscan_vborder != 0)
-                       return true;
-       } else  if (dm_state->underscan_enable && !old_dm_state->underscan_enable) {
-               if (dm_state->underscan_hborder != 0 && dm_state->underscan_vborder != 0)
-                       return true;
-       } else if (dm_state->underscan_hborder != old_dm_state->underscan_hborder
-                               || dm_state->underscan_vborder != old_dm_state->underscan_vborder)
-                       return true;
-       return false;
-}
-
-static void remove_stream(
-               struct amdgpu_device *adev,
-               struct amdgpu_crtc *acrtc,
-               struct dc_stream *stream)
-{
-       /* this is the update mode case */
-       if (adev->dm.freesync_module)
-               mod_freesync_remove_stream(adev->dm.freesync_module, stream);
-
-       acrtc->otg_inst = -1;
-       acrtc->enabled = false;
-}
-
-static void handle_cursor_update(
-               struct drm_plane *plane,
-               struct drm_plane_state *old_plane_state)
-{
-       if (!plane->state->fb && !old_plane_state->fb)
-               return;
-
-       /* Check if it's a cursor on/off update or just cursor move*/
-       if (plane->state->fb == old_plane_state->fb)
-               dm_crtc_cursor_move(
-                               plane->state->crtc,
-                               plane->state->crtc_x,
-                               plane->state->crtc_y);
-       else {
-               struct amdgpu_framebuffer *afb =
-                               to_amdgpu_framebuffer(plane->state->fb);
-               dm_crtc_cursor_set(
-                               (!!plane->state->fb) ?
-                                               plane->state->crtc :
-                                               old_plane_state->crtc,
-                               (!!plane->state->fb) ?
-                                               afb->address :
-                                               0,
-                               plane->state->crtc_w,
-                               plane->state->crtc_h);
-       }
-}
-
-
-static void prepare_flip_isr(struct amdgpu_crtc *acrtc)
-{
-
-       assert_spin_locked(&acrtc->base.dev->event_lock);
-       WARN_ON(acrtc->event);
-
-       acrtc->event = acrtc->base.state->event;
-
-       /* Set the flip status */
-       acrtc->pflip_status = AMDGPU_FLIP_SUBMITTED;
-
-       /* Mark this event as consumed */
-       acrtc->base.state->event = NULL;
-
-       DRM_DEBUG_DRIVER("crtc:%d, pflip_stat:AMDGPU_FLIP_SUBMITTED\n",
-                                                acrtc->crtc_id);
-}
-
-/*
- * Executes flip
- *
- * Waits on all BO's fences and for proper vblank count
- */
-static void amdgpu_dm_do_flip(
-                               struct drm_crtc *crtc,
-                               struct drm_framebuffer *fb,
-                               uint32_t target)
-{
-       unsigned long flags;
-       uint32_t target_vblank;
-       int r, vpos, hpos;
-       struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
-       struct amdgpu_framebuffer *afb = to_amdgpu_framebuffer(fb);
-       struct amdgpu_bo *abo = gem_to_amdgpu_bo(afb->obj);
-       struct amdgpu_device *adev = crtc->dev->dev_private;
-       bool async_flip = (acrtc->flip_flags & DRM_MODE_PAGE_FLIP_ASYNC) != 0;
-       struct dc_flip_addrs addr = { {0} };
-       struct dc_surface_update surface_updates[1] = { {0} };
-       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(crtc->state);
-
-
-       /* Prepare wait for target vblank early - before the fence-waits */
-       target_vblank = target - drm_crtc_vblank_count(crtc) +
-                       amdgpu_get_vblank_counter_kms(crtc->dev, acrtc->crtc_id);
-
-       /*TODO This might fail and hence better not used, wait
-        * explicitly on fences instead
-        * and in general should be called for
-        * blocking commit to as per framework helpers
-        * */
-       r = amdgpu_bo_reserve(abo, true);
-       if (unlikely(r != 0)) {
-               DRM_ERROR("failed to reserve buffer before flip\n");
-               WARN_ON(1);
-       }
-
-       /* Wait for all fences on this FB */
-       WARN_ON(reservation_object_wait_timeout_rcu(abo->tbo.resv, true, false,
-                                                                   MAX_SCHEDULE_TIMEOUT) < 0);
-
-       amdgpu_bo_unreserve(abo);
-
-       /* Wait until we're out of the vertical blank period before the one
-        * targeted by the flip
-        */
-       while ((acrtc->enabled &&
-               (amdgpu_get_crtc_scanoutpos(adev->ddev, acrtc->crtc_id, 0,
-                                       &vpos, &hpos, NULL, NULL,
-                                       &crtc->hwmode)
-                & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK)) ==
-               (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK) &&
-               (int)(target_vblank -
-                 amdgpu_get_vblank_counter_kms(adev->ddev, acrtc->crtc_id)) > 0)) {
-               usleep_range(1000, 1100);
-       }
-
-       /* Flip */
-       spin_lock_irqsave(&crtc->dev->event_lock, flags);
-       /* update crtc fb */
-       crtc->primary->fb = fb;
-
-       WARN_ON(acrtc->pflip_status != AMDGPU_FLIP_NONE);
-       WARN_ON(!acrtc_state->stream);
-
-       addr.address.grph.addr.low_part = lower_32_bits(afb->address);
-       addr.address.grph.addr.high_part = upper_32_bits(afb->address);
-       addr.flip_immediate = async_flip;
-
-
-       if (acrtc->base.state->event)
-               prepare_flip_isr(acrtc);
-
-       surface_updates->surface = dc_stream_get_status(acrtc_state->stream)->surfaces[0];
-       surface_updates->flip_addr = &addr;
-
-
-       dc_update_surfaces_and_stream(adev->dm.dc, surface_updates, 1, acrtc_state->stream, NULL);
-
-       DRM_DEBUG_DRIVER("%s Flipping to hi: 0x%x, low: 0x%x \n",
-                        __func__,
-                        addr.address.grph.addr.high_part,
-                        addr.address.grph.addr.low_part);
-
-
-       spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
-}
-
-static void amdgpu_dm_commit_surfaces(struct drm_atomic_state *state,
-                       struct drm_device *dev,
-                       struct amdgpu_display_manager *dm,
-                       struct drm_crtc *pcrtc,
-                       bool *wait_for_vblank)
-{
-       uint32_t i;
-       struct drm_plane *plane;
-       struct drm_plane_state *old_plane_state;
-       struct dc_stream *dc_stream_attach;
-       struct dc_surface *dc_surfaces_constructed[MAX_SURFACES];
-       struct amdgpu_crtc *acrtc_attach = to_amdgpu_crtc(pcrtc);
-       struct dm_crtc_state *acrtc_state = to_dm_crtc_state(pcrtc->state);
-       int planes_count = 0;
-       unsigned long flags;
-
-       /* update planes when needed */
-       for_each_plane_in_state(state, plane, old_plane_state, i) {
-               struct drm_plane_state *plane_state = plane->state;
-               struct drm_crtc *crtc = plane_state->crtc;
-               struct drm_framebuffer *fb = plane_state->fb;
-               bool pflip_needed;
-               struct dm_plane_state *dm_plane_state = to_dm_plane_state(plane_state);
-
-               if (plane->type == DRM_PLANE_TYPE_CURSOR) {
-                       handle_cursor_update(plane, old_plane_state);
-                       continue;
-               }
-
-               if (!fb || !crtc || pcrtc != crtc || !crtc->state->active ||
-                               (!crtc->state->planes_changed &&
-                                               !pcrtc->state->color_mgmt_changed))
-                       continue;
-
-               pflip_needed = !state->allow_modeset;
-
-               spin_lock_irqsave(&crtc->dev->event_lock, flags);
-               if (acrtc_attach->pflip_status != AMDGPU_FLIP_NONE) {
-                       DRM_ERROR("add_surface: acrtc %d, already busy\n",
-                                       acrtc_attach->crtc_id);
-                       spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
-                       /* In comit tail framework this cannot happen */
-                       WARN_ON(1);
-               }
-               spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
-
-               if (!pflip_needed) {
-                       WARN_ON(!dm_plane_state->surface);
-
-                       dc_surfaces_constructed[planes_count] = dm_plane_state->surface;
-
-                       dc_stream_attach = acrtc_state->stream;
-                       planes_count++;
-
-               } else if (crtc->state->planes_changed) {
-                       /* Assume even ONE crtc with immediate flip means
-                        * entire can't wait for VBLANK
-                        * TODO Check if it's correct
-                        */
-                       *wait_for_vblank =
-                               acrtc_attach->flip_flags & DRM_MODE_PAGE_FLIP_ASYNC ?
-                               false : true;
-
-                       /* TODO: Needs rework for multiplane flip */
-                       if (plane->type == DRM_PLANE_TYPE_PRIMARY)
-                               drm_crtc_vblank_get(crtc);
-
-                       amdgpu_dm_do_flip(
-                               crtc,
-                               fb,
-                               drm_crtc_vblank_count(crtc) + *wait_for_vblank);
-
-                       /*TODO BUG remove ASAP in 4.12 to avoid race between worker and flip IOCTL */
-
-                       /*clean up the flags for next usage*/
-                       acrtc_attach->flip_flags = 0;
-               }
-
-       }
-
-       if (planes_count) {
-               unsigned long flags;
-
-               if (pcrtc->state->event) {
-
-                       drm_crtc_vblank_get(pcrtc);
-
-                       spin_lock_irqsave(&pcrtc->dev->event_lock, flags);
-                       prepare_flip_isr(acrtc_attach);
-                       spin_unlock_irqrestore(&pcrtc->dev->event_lock, flags);
-               }
-
-               if (false == dc_commit_surfaces_to_stream(dm->dc,
-                                                         dc_surfaces_constructed,
-                                                         planes_count,
-                                                         dc_stream_attach))
-                       dm_error("%s: Failed to attach surface!\n", __func__);
-       } else {
-               /*TODO BUG Here should go disable planes on CRTC. */
-       }
-}
-
-
-int amdgpu_dm_atomic_commit(
-               struct drm_device *dev,
-               struct drm_atomic_state *state,
-               bool nonblock)
-{
-       struct drm_crtc *crtc;
-       struct drm_crtc_state *new_state;
-       struct amdgpu_device *adev = dev->dev_private;
-       int i;
-
-       /*
-        * We evade vblanks and pflips on crtc that
-        * should be changed. We do it here to flush & disable
-        * interrupts before drm_swap_state is called in drm_atomic_helper_commit
-        * it will update crtc->dm_crtc_state->stream pointer which is used in
-        * the ISRs.
-        */
-       for_each_crtc_in_state(state, crtc, new_state, i) {
-               struct dm_crtc_state *old_acrtc_state = to_dm_crtc_state(crtc->state);
-               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
-
-               if (drm_atomic_crtc_needs_modeset(new_state) && old_acrtc_state->stream)
-                       manage_dm_interrupts(adev, acrtc, false);
-       }
-
-       return drm_atomic_helper_commit(dev, state, nonblock);
-
-       /*TODO Handle EINTR, reenable IRQ*/
-}
-
-void amdgpu_dm_atomic_commit_tail(
-       struct drm_atomic_state *state)
-{
-       struct drm_device *dev = state->dev;
-       struct amdgpu_device *adev = dev->dev_private;
-       struct amdgpu_display_manager *dm = &adev->dm;
-       struct dm_atomic_state *dm_state;
-       uint32_t i, j;
-       uint32_t new_crtcs_count = 0;
-       struct drm_crtc *crtc, *pcrtc;
-       struct drm_crtc_state *old_crtc_state;
-       struct amdgpu_crtc *new_crtcs[MAX_STREAMS];
-       struct dc_stream *new_stream = NULL;
-       unsigned long flags;
-       bool wait_for_vblank = true;
-       struct drm_connector *connector;
-       struct drm_connector_state *old_conn_state;
-       struct dm_crtc_state *old_acrtc_state, *new_acrtc_state;
-
-       drm_atomic_helper_update_legacy_modeset_state(dev, state);
-
-       dm_state = to_dm_atomic_state(state);
-
-       /* update changed items */
-       for_each_crtc_in_state(state, crtc, old_crtc_state, i) {
-               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
-               struct drm_crtc_state *new_state = crtc->state;
-               new_acrtc_state = to_dm_crtc_state(new_state);
-               old_acrtc_state = to_dm_crtc_state(old_crtc_state);
-
-               DRM_DEBUG_KMS(
-                       "amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, "
-                       "planes_changed:%d, mode_changed:%d,active_changed:%d,"
-                       "connectors_changed:%d\n",
-                       acrtc->crtc_id,
-                       new_state->enable,
-                       new_state->active,
-                       new_state->planes_changed,
-                       new_state->mode_changed,
-                       new_state->active_changed,
-                       new_state->connectors_changed);
-
-               /* handles headless hotplug case, updating new_state and
-                * aconnector as needed
-                */
-
-               if (modeset_required(new_state)) {
-
-                       DRM_INFO("Atomic commit: SET crtc id %d: [%p]\n", acrtc->crtc_id, acrtc);
-
-                       if (!new_acrtc_state->stream) {
-                               /*
-                               * this could happen because of issues with
-                               * userspace notifications delivery.
-                               * In this case userspace tries to set mode on
-                               * display which is disconnect in fact.
-                               * dc_sink in NULL in this case on aconnector.
-                               * We expect reset mode will come soon.
-                               *
-                               * This can also happen when unplug is done
-                               * during resume sequence ended
-                               *
-                               * In this case, we want to pretend we still
-                               * have a sink to keep the pipe running so that
-                               * hw state is consistent with the sw state
-                               */
-                               DRM_DEBUG_KMS("%s: Failed to create new stream for crtc %d\n",
-                                               __func__, acrtc->base.base.id);
-                               continue;
-                       }
-
-
-                       if (old_acrtc_state->stream)
-                               remove_stream(adev, acrtc, old_acrtc_state->stream);
-
-
-                       /*
-                        * this loop saves set mode crtcs
-                        * we needed to enable vblanks once all
-                        * resources acquired in dc after dc_commit_streams
-                        */
-
-                       /*TODO move all this into dm_crtc_state, get rid of
-                        * new_crtcs array and use old and new atomic states
-                        * instead
-                        */
-                       new_crtcs[new_crtcs_count] = acrtc;
-                       new_crtcs_count++;
-
-                       acrtc->enabled = true;
-                       acrtc->hw_mode = crtc->state->mode;
-                       crtc->hwmode = crtc->state->mode;
-               } else if (modereset_required(new_state)) {
-                       DRM_INFO("Atomic commit: RESET. crtc id %d:[%p]\n", acrtc->crtc_id, acrtc);
-
-                       /* i.e. reset mode */
-                       if (old_acrtc_state->stream)
-                               remove_stream(adev, acrtc, old_acrtc_state->stream);
-               }
-       } /* for_each_crtc_in_state() */
-
-       /*
-        * Add streams after required streams from new and replaced streams
-        * are removed from freesync module
-        */
-       if (adev->dm.freesync_module) {
-               for (i = 0; i < new_crtcs_count; i++) {
-                       struct amdgpu_connector *aconnector = NULL;
-                       new_acrtc_state = to_dm_crtc_state(new_crtcs[i]->base.state);
-
-                       new_stream = new_acrtc_state->stream;
-                       aconnector =
-                               amdgpu_dm_find_first_crct_matching_connector(
-                                       state,
-                                       &new_crtcs[i]->base,
-                                       false);
-                       if (!aconnector) {
-                               DRM_INFO(
-                                               "Atomic commit: Failed to find connector for acrtc id:%d "
-                                               "skipping freesync init\n",
-                                               new_crtcs[i]->crtc_id);
-                               continue;
-                       }
-
-                       mod_freesync_add_stream(adev->dm.freesync_module,
-                                               new_stream, &aconnector->caps);
-               }
-       }
-
-       if (dm_state->context)
-               WARN_ON(!dc_commit_context(dm->dc, dm_state->context));
-
-
-       list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
-               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
-               new_acrtc_state = to_dm_crtc_state(crtc->state);
-
-               if (new_acrtc_state->stream != NULL) {
-                       const struct dc_stream_status *status =
-                                       dc_stream_get_status(new_acrtc_state->stream);
-
-                       if (!status)
-                               DC_ERR("got no status for stream %p on acrtc%p\n", new_acrtc_state->stream, acrtc);
-                       else
-                               acrtc->otg_inst = status->primary_otg_inst;
-               }
-       }
-
-       /* Handle scaling and undersacn changes*/
-       for_each_connector_in_state(state, connector, old_conn_state, i) {
-               struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
-               struct dm_connector_state *con_new_state =
-                               to_dm_connector_state(aconnector->base.state);
-               struct dm_connector_state *con_old_state =
-                               to_dm_connector_state(old_conn_state);
-               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(con_new_state->base.crtc);
-               struct dc_stream_status *status = NULL;
-
-               /* Skip any modesets/resets */
-               if (!acrtc || drm_atomic_crtc_needs_modeset(acrtc->base.state))
-                       continue;
-
-               /* Skip any thing not scale or underscan changes */
-               if (!is_scaling_state_different(con_new_state, con_old_state))
-                       continue;
-
-               new_acrtc_state = to_dm_crtc_state(acrtc->base.state);
-
-               update_stream_scaling_settings(&con_new_state->base.crtc->mode,
-                               con_new_state, (struct dc_stream *)new_acrtc_state->stream);
-
-               status = dc_stream_get_status(new_acrtc_state->stream);
-               WARN_ON(!status);
-               WARN_ON(!status->surface_count);
-
-               if (!new_acrtc_state->stream)
-                       continue;
-
-               /*TODO How it works with MPO ?*/
-               if (!dc_commit_surfaces_to_stream(
-                               dm->dc,
-                               status->surfaces,
-                               status->surface_count,
-                               new_acrtc_state->stream))
-                       dm_error("%s: Failed to update stream scaling!\n", __func__);
-       }
-
-       for (i = 0; i < new_crtcs_count; i++) {
-               /*
-                * loop to enable interrupts on newly arrived crtc
-                */
-               struct amdgpu_crtc *acrtc = new_crtcs[i];
-               new_acrtc_state = to_dm_crtc_state(acrtc->base.state);
-
-               if (adev->dm.freesync_module)
-                       mod_freesync_notify_mode_change(
-                               adev->dm.freesync_module, &new_acrtc_state->stream, 1);
-
-               manage_dm_interrupts(adev, acrtc, true);
-       }
-
-       /* update planes when needed per crtc*/
-       for_each_crtc_in_state(state, pcrtc, old_crtc_state, j) {
-               new_acrtc_state = to_dm_crtc_state(pcrtc->state);
-
-               if (new_acrtc_state->stream)
-                       amdgpu_dm_commit_surfaces(state, dev, dm, pcrtc, &wait_for_vblank);
-       }
-
-
-       /*
-        * send vblank event on all events not handled in flip and
-        * mark consumed event for drm_atomic_helper_commit_hw_done
-        */
-       spin_lock_irqsave(&adev->ddev->event_lock, flags);
-       for_each_crtc_in_state(state, crtc, old_crtc_state, i) {
-               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc);
-
-               if (acrtc->base.state->event)
-                       drm_send_event_locked(dev, &crtc->state->event->base);
-
-               acrtc->base.state->event = NULL;
-       }
-       spin_unlock_irqrestore(&adev->ddev->event_lock, flags);
-
-       /* Signal HW programming completion */
-       drm_atomic_helper_commit_hw_done(state);
-
-       if (wait_for_vblank)
-               drm_atomic_helper_wait_for_vblanks(dev, state);
-
-       drm_atomic_helper_cleanup_planes(dev, state);
-}
-
-
-static int dm_force_atomic_commit(struct drm_connector *connector)
-{
-       int ret = 0;
-       struct drm_device *ddev = connector->dev;
-       struct drm_atomic_state *state = drm_atomic_state_alloc(ddev);
-       struct amdgpu_crtc *disconnected_acrtc = to_amdgpu_crtc(connector->encoder->crtc);
-       struct drm_plane *plane = disconnected_acrtc->base.primary;
-       struct drm_connector_state *conn_state;
-       struct drm_crtc_state *crtc_state;
-       struct drm_plane_state *plane_state;
-
-       if (!state)
-               return -ENOMEM;
-
-       state->acquire_ctx = ddev->mode_config.acquire_ctx;
-
-       /* Construct an atomic state to restore previous display setting */
-
-       /*
-        * Attach connectors to drm_atomic_state
-        */
-       conn_state = drm_atomic_get_connector_state(state, connector);
-
-       ret = PTR_ERR_OR_ZERO(conn_state);
-       if (ret)
-               goto err;
-
-       /* Attach crtc to drm_atomic_state*/
-       crtc_state = drm_atomic_get_crtc_state(state, &disconnected_acrtc->base);
-
-       ret = PTR_ERR_OR_ZERO(crtc_state);
-       if (ret)
-               goto err;
-
-       /* force a restore */
-       crtc_state->mode_changed = true;
-
-       /* Attach plane to drm_atomic_state */
-       plane_state = drm_atomic_get_plane_state(state, plane);
-
-       ret = PTR_ERR_OR_ZERO(plane_state);
-       if (ret)
-               goto err;
-
-
-       /* Call commit internally with the state we just constructed */
-       ret = drm_atomic_commit(state);
-       if (!ret)
-               return 0;
-
-err:
-       DRM_ERROR("Restoring old state failed with %i\n", ret);
-       drm_atomic_state_put(state);
-
-       return ret;
-}
-
-/*
- * This functions handle all cases when set mode does not come upon hotplug.
- * This include when the same display is unplugged then plugged back into the
- * same port and when we are running without usermode desktop manager supprot
- */
-void dm_restore_drm_connector_state(struct drm_device *dev, struct drm_connector *connector)
-{
-       struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
-       struct amdgpu_crtc *disconnected_acrtc;
-       struct dm_crtc_state *acrtc_state;
-
-       if (!aconnector->dc_sink || !connector->state || !connector->encoder)
-               return;
-
-       disconnected_acrtc = to_amdgpu_crtc(connector->encoder->crtc);
-       acrtc_state = to_dm_crtc_state(disconnected_acrtc->base.state);
-
-       if (!disconnected_acrtc || !acrtc_state->stream)
-               return;
-
-       /*
-        * If the previous sink is not released and different from the current,
-        * we deduce we are in a state where we can not rely on usermode call
-        * to turn on the display, so we do it here
-        */
-       if (acrtc_state->stream->sink != aconnector->dc_sink)
-               dm_force_atomic_commit(&aconnector->base);
-}
-
-static uint32_t add_val_sets_surface(
-       struct dc_validation_set *val_sets,
-       uint32_t set_count,
-       const struct dc_stream *stream,
-       struct dc_surface *surface)
-{
-       uint32_t i = 0, j = 0;
-
-       while (i < set_count) {
-               if (val_sets[i].stream == stream) {
-                       while (val_sets[i].surfaces[j])
-                               j++;
-                       break;
-               }
-               ++i;
-       }
-
-       val_sets[i].surfaces[j] = surface;
-       val_sets[i].surface_count++;
-
-       return val_sets[i].surface_count;
-}
-
-static uint32_t update_in_val_sets_stream(
-       struct dc_validation_set *val_sets,
-       uint32_t set_count,
-       struct dc_stream *old_stream,
-       struct dc_stream *new_stream,
-       struct drm_crtc *crtc)
-{
-       uint32_t i = 0;
-
-       while (i < set_count) {
-               if (val_sets[i].stream == old_stream)
-                       break;
-               ++i;
-       }
-
-       val_sets[i].stream = new_stream;
-
-       if (i == set_count)
-               /* nothing found. add new one to the end */
-               return set_count + 1;
-
-       return set_count;
-}
-
-static uint32_t remove_from_val_sets(
-       struct dc_validation_set *val_sets,
-       uint32_t set_count,
-       const struct dc_stream *stream)
-{
-       int i;
-
-       for (i = 0; i < set_count; i++)
-               if (val_sets[i].stream == stream)
-                       break;
-
-       if (i == set_count) {
-               /* nothing found */
-               return set_count;
-       }
-
-       set_count--;
-
-       for (; i < set_count; i++) {
-               val_sets[i] = val_sets[i + 1];
-       }
-
-       return set_count;
-}
-
-/*`
- * Grabs all modesetting locks to serialize against any blocking commits,
- * Waits for completion of all non blocking commits.
- */
-static int do_aquire_global_lock(
-               struct drm_device *dev,
-               struct drm_atomic_state *state)
-{
-       struct drm_crtc *crtc;
-       struct drm_crtc_commit *commit;
-       long ret;
-
-       /* Adding all modeset locks to aquire_ctx will
-        * ensure that when the framework release it the
-        * extra locks we are locking here will get released to
-        */
-       ret = drm_modeset_lock_all_ctx(dev, state->acquire_ctx);
-       if (ret)
-               return ret;
-
-       list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
-               spin_lock(&crtc->commit_lock);
-               commit = list_first_entry_or_null(&crtc->commit_list,
-                               struct drm_crtc_commit, commit_entry);
-               if (commit)
-                       drm_crtc_commit_get(commit);
-               spin_unlock(&crtc->commit_lock);
-
-               if (!commit)
-                       continue;
-
-               /* Make sure all pending HW programming completed and
-                * page flips done
-                */
-               ret = wait_for_completion_interruptible_timeout(&commit->hw_done, 10*HZ);
-
-               if (ret > 0)
-                       ret = wait_for_completion_interruptible_timeout(
-                                       &commit->flip_done, 10*HZ);
-
-               if (ret == 0)
-                       DRM_ERROR("[CRTC:%d:%s] hw_done or flip_done "
-                                       "timed out\n", crtc->base.id, crtc->name);
-
-               drm_crtc_commit_put(commit);
-       }
-
-       return ret < 0 ? ret : 0;
-}
-
-int amdgpu_dm_atomic_check(struct drm_device *dev,
-                       struct drm_atomic_state *state)
-{
-       struct dm_atomic_state *dm_state;
-       struct drm_crtc *crtc;
-       struct drm_crtc_state *crtc_state;
-       struct drm_plane *plane;
-       struct drm_plane_state *plane_state;
-       int i, j;
-       int ret;
-       struct amdgpu_device *adev = dev->dev_private;
-       struct dc *dc = adev->dm.dc;
-       struct drm_connector *connector;
-       struct drm_connector_state *conn_state;
-       int set_count;
-       struct dc_validation_set set[MAX_STREAMS] = { { 0 } };
-       struct dm_crtc_state *old_acrtc_state, *new_acrtc_state;
-
-       /*
-        * This bool will be set for true for any modeset/reset
-        * or surface update which implies non fast surface update.
-        */
-       bool lock_and_validation_needed = false;
-
-       ret = drm_atomic_helper_check_modeset(dev, state);
-
-       if (ret) {
-               DRM_ERROR("Atomic state validation failed with error :%d !\n", ret);
-               return ret;
-       }
-
-       dm_state = to_dm_atomic_state(state);
-
-       /* copy existing configuration */
-       set_count = 0;
-       list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
-
-               old_acrtc_state = to_dm_crtc_state(crtc->state);
-
-               if (old_acrtc_state->stream) {
-                       dc_stream_retain(old_acrtc_state->stream);
-                       set[set_count].stream = old_acrtc_state->stream;
-                       ++set_count;
-               }
-       }
-
-       /*TODO Move this code into dm_crtc_atomic_check once we get rid of dc_validation_set */
-       /* update changed items */
-       for_each_crtc_in_state(state, crtc, crtc_state, i) {
-               struct amdgpu_crtc *acrtc = NULL;
-               struct amdgpu_connector *aconnector = NULL;
-               old_acrtc_state = to_dm_crtc_state(crtc->state);
-               new_acrtc_state = to_dm_crtc_state(crtc_state);
-               acrtc = to_amdgpu_crtc(crtc);
-
-               aconnector = amdgpu_dm_find_first_crct_matching_connector(state, crtc, true);
-
-               DRM_DEBUG_KMS(
-                       "amdgpu_crtc id:%d crtc_state_flags: enable:%d, active:%d, "
-                       "planes_changed:%d, mode_changed:%d,active_changed:%d,"
-                       "connectors_changed:%d\n",
-                       acrtc->crtc_id,
-                       crtc_state->enable,
-                       crtc_state->active,
-                       crtc_state->planes_changed,
-                       crtc_state->mode_changed,
-                       crtc_state->active_changed,
-                       crtc_state->connectors_changed);
-
-               if (modeset_required(crtc_state)) {
-
-                       struct dc_stream *new_stream = NULL;
-                       struct drm_connector_state *conn_state = NULL;
-                       struct dm_connector_state *dm_conn_state = NULL;
-
-                       if (aconnector) {
-                               conn_state = drm_atomic_get_connector_state(state, &aconnector->base);
-                               if (IS_ERR(conn_state)) {
-                                       ret = PTR_ERR_OR_ZERO(conn_state);
-                                       goto fail;
-                               }
-
-                               dm_conn_state = to_dm_connector_state(conn_state);
-                       }
-
-                       new_stream = create_stream_for_sink(aconnector, &crtc_state->mode, dm_conn_state);
-
-                       /*
-                        * we can have no stream on ACTION_SET if a display
-                        * was disconnected during S3, in this case it not and
-                        * error, the OS will be updated after detection, and
-                        * do the right thing on next atomic commit
-                        */
-                       if (!new_stream) {
-                               DRM_DEBUG_KMS("%s: Failed to create new stream for crtc %d\n",
-                                               __func__, acrtc->base.base.id);
-                               break;
-                       }
-
-                       if (new_acrtc_state->stream)
-                               dc_stream_release(new_acrtc_state->stream);
-
-                       new_acrtc_state->stream = new_stream;
-
-                       set_count = update_in_val_sets_stream(
-                                       set,
-                                       set_count,
-                                       old_acrtc_state->stream,
-                                       new_acrtc_state->stream,
-                                       crtc);
-
-                       lock_and_validation_needed = true;
-
-               } else if (modereset_required(crtc_state)) {
-
-                       /* i.e. reset mode */
-                       if (new_acrtc_state->stream) {
-                               set_count = remove_from_val_sets(
-                                               set,
-                                               set_count,
-                                               new_acrtc_state->stream);
-
-                               dc_stream_release(new_acrtc_state->stream);
-                               new_acrtc_state->stream = NULL;
-
-                               lock_and_validation_needed = true;
-                       }
-               }
-
-
-               /*
-                * Hack: Commit needs planes right now, specifically for gamma
-                * TODO rework commit to check CRTC for gamma change
-                */
-               if (crtc_state->color_mgmt_changed) {
-
-                       ret = drm_atomic_add_affected_planes(state, crtc);
-                       if (ret)
-                               goto fail;
-               }
-       }
-
-       /* Check scaling and undersacn changes*/
-       /*TODO Removed scaling changes validation due to inability to commit
-        * new stream into context w\o causing full reset. Need to
-        * decide how to handle.
-        */
-       for_each_connector_in_state(state, connector, conn_state, i) {
-               struct amdgpu_connector *aconnector = to_amdgpu_connector(connector);
-               struct dm_connector_state *con_old_state =
-                               to_dm_connector_state(aconnector->base.state);
-               struct dm_connector_state *con_new_state =
-                                               to_dm_connector_state(conn_state);
-               struct amdgpu_crtc *acrtc = to_amdgpu_crtc(con_new_state->base.crtc);
-
-               /* Skip any modesets/resets */
-               if (!acrtc || drm_atomic_crtc_needs_modeset(acrtc->base.state))
-                       continue;
-
-               /* Skip any thing not scale or underscan chnages */
-               if (!is_scaling_state_different(con_new_state, con_old_state))
-                       continue;
-
-               lock_and_validation_needed = true;
-       }
-
-       for_each_crtc_in_state(state, crtc, crtc_state, i) {
-               new_acrtc_state = to_dm_crtc_state(crtc_state);
-
-               for_each_plane_in_state(state, plane, plane_state, j) {
-                       struct drm_crtc *plane_crtc = plane_state->crtc;
-                       struct drm_framebuffer *fb = plane_state->fb;
-                       bool pflip_needed;
-                       struct dm_plane_state *dm_plane_state = to_dm_plane_state(plane_state);
-
-                       /*TODO Implement atomic check for cursor plane */
-                       if (plane->type == DRM_PLANE_TYPE_CURSOR)
-                               continue;
-
-                       if (!fb || !plane_crtc || crtc != plane_crtc || !crtc_state->active)
-                               continue;
-
-                       WARN_ON(!new_acrtc_state->stream);
-
-                       pflip_needed = !state->allow_modeset;
-                       if (!pflip_needed) {
-                               struct dc_surface *surface;
-
-                               surface = dc_create_surface(dc);
-
-                               ret = fill_plane_attributes(
-                                       plane_crtc->dev->dev_private,
-                                       surface,
-                                       plane_state,
-                                       crtc_state,
-                                       false);
-                               if (ret)
-                                       goto fail;
-
-
-                               if (dm_plane_state->surface)
-                                       dc_surface_release(dm_plane_state->surface);
-
-                               dm_plane_state->surface = surface;
-
-                               add_val_sets_surface(set,
-                                                    set_count,
-                                                    new_acrtc_state->stream,
-                                                    surface);
-
-                               lock_and_validation_needed = true;
-                       }
-               }
-       }
-
-       /* Run this here since we want to validate the streams we created */
-       ret = drm_atomic_helper_check_planes(dev, state);
-       if (ret)
-               goto fail;
-
-       /*
-        * For full updates case when
-        * removing/adding/updating  streams on once CRTC while flipping
-        * on another CRTC,
-        * acquiring global lock  will guarantee that any such full
-        * update commit
-        * will wait for completion of any outstanding flip using DRMs
-        * synchronization events.
-        */
-
-       if (lock_and_validation_needed) {
-
-               ret = do_aquire_global_lock(dev, state);
-               if (ret)
-                       goto fail;
-               WARN_ON(dm_state->context);
-               dm_state->context = dc_get_validate_context(dc, set, set_count);
-               if (!dm_state->context) {
-                       ret = -EINVAL;
-                       goto fail;
-               }
-       }
-
-       /* Must be success */
-       WARN_ON(ret);
-       return ret;
-
-fail:
-       if (ret == -EDEADLK)
-               DRM_DEBUG_KMS("Atomic check stopped due to to deadlock.\n");
-       else if (ret == -EINTR || ret == -EAGAIN || ret == -ERESTARTSYS)
-               DRM_DEBUG_KMS("Atomic check stopped due to to signal.\n");
-       else
-               DRM_ERROR("Atomic check failed with err: %d .\n", ret);
-
-       return ret;
-}
-
-static bool is_dp_capable_without_timing_msa(
-               struct dc *dc,
-               struct amdgpu_connector *amdgpu_connector)
-{
-       uint8_t dpcd_data;
-       bool capable = false;
-
-       if (amdgpu_connector->dc_link &&
-               dm_helpers_dp_read_dpcd(
-                               NULL,
-                               amdgpu_connector->dc_link,
-                               DP_DOWN_STREAM_PORT_COUNT,
-                               &dpcd_data,
-                               sizeof(dpcd_data))) {
-               capable = (dpcd_data & DP_MSA_TIMING_PAR_IGNORED) ? true:false;
-       }
-
-       return capable;
-}
-void amdgpu_dm_add_sink_to_freesync_module(
-               struct drm_connector *connector,
-               struct edid *edid)
-{
-       int i;
-       uint64_t val_capable;
-       bool edid_check_required;
-       struct detailed_timing *timing;
-       struct detailed_non_pixel *data;
-       struct detailed_data_monitor_range *range;
-       struct amdgpu_connector *amdgpu_connector =
-                       to_amdgpu_connector(connector);
-
-       struct drm_device *dev = connector->dev;
-       struct amdgpu_device *adev = dev->dev_private;
-       edid_check_required = false;
-       if (!amdgpu_connector->dc_sink) {
-               DRM_ERROR("dc_sink NULL, could not add free_sync module.\n");
-               return;
-       }
-       if (!adev->dm.freesync_module)
-               return;
-       /*
-        * if edid non zero restrict freesync only for dp and edp
-        */
-       if (edid) {
-               if (amdgpu_connector->dc_sink->sink_signal == SIGNAL_TYPE_DISPLAY_PORT
-                       || amdgpu_connector->dc_sink->sink_signal == SIGNAL_TYPE_EDP) {
-                       edid_check_required = is_dp_capable_without_timing_msa(
-                                               adev->dm.dc,
-                                               amdgpu_connector);
-               }
-       }
-       val_capable = 0;
-       if (edid_check_required == true && (edid->version > 1 ||
-          (edid->version == 1 && edid->revision > 1))) {
-               for (i = 0; i < 4; i++) {
-
-                       timing  = &edid->detailed_timings[i];
-                       data    = &timing->data.other_data;
-                       range   = &data->data.range;
-                       /*
-                        * Check if monitor has continuous frequency mode
-                        */
-                       if (data->type != EDID_DETAIL_MONITOR_RANGE)
-                               continue;
-                       /*
-                        * Check for flag range limits only. If flag == 1 then
-                        * no additional timing information provided.
-                        * Default GTF, GTF Secondary curve and CVT are not
-                        * supported
-                        */
-                       if (range->flags != 1)
-                               continue;
-
-                       amdgpu_connector->min_vfreq = range->min_vfreq;
-                       amdgpu_connector->max_vfreq = range->max_vfreq;
-                       amdgpu_connector->pixel_clock_mhz =
-                               range->pixel_clock_mhz * 10;
-                       break;
-               }
-
-               if (amdgpu_connector->max_vfreq -
-                               amdgpu_connector->min_vfreq > 10) {
-                       amdgpu_connector->caps.supported = true;
-                       amdgpu_connector->caps.min_refresh_in_micro_hz =
-                                       amdgpu_connector->min_vfreq * 1000000;
-                       amdgpu_connector->caps.max_refresh_in_micro_hz =
-                                       amdgpu_connector->max_vfreq * 1000000;
-                               val_capable = 1;
-               }
-       }
-
-       /*
-        * TODO figure out how to notify user-mode or DRM of freesync caps
-        * once we figure out how to deal with freesync in an upstreamable
-        * fashion
-        */
-
-}
-
-void amdgpu_dm_remove_sink_from_freesync_module(
-               struct drm_connector *connector)
-{
-       /*
-        * TODO fill in once we figure out how to deal with freesync in
-        * an upstreamable fashion
-        */
-}
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_types.h
deleted file mode 100644 (file)
index 94de6a3..0000000
+++ /dev/null
@@ -1,135 +0,0 @@
-/*
- * Copyright 2012-13 Advanced Micro Devices, Inc.
- *
- * 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 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 COPYRIGHT HOLDER(S) OR AUTHOR(S) 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.
- *
- * Authors: AMD
- *
- */
-
-#ifndef __AMDGPU_DM_TYPES_H__
-#define __AMDGPU_DM_TYPES_H__
-
-#include <drm/drmP.h>
-#include <drm/drm_atomic.h>
-#include "dc.h"
-
-struct amdgpu_framebuffer;
-struct amdgpu_display_manager;
-struct dc_validation_set;
-struct dc_surface;
-/* TODO rename to dc_stream_state */
-struct  dc_stream;
-
-
-struct dm_plane_state {
-       struct drm_plane_state base;
-       struct dc_surface *surface;
-};
-
-struct dm_crtc_state {
-       struct drm_crtc_state base;
-       struct dc_stream *stream;
-};
-
-#define to_dm_crtc_state(x)    container_of(x, struct dm_crtc_state, base)
-
-struct dm_atomic_state {
-       struct drm_atomic_state base;
-
-       struct validate_context *context;
-};
-
-#define to_dm_atomic_state(x) container_of(x, struct dm_atomic_state, base)
-
-
-/*TODO Jodan Hersen use the one in amdgpu_dm*/
-int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm,
-                       struct amdgpu_plane *aplane,
-                       unsigned long possible_crtcs);
-int amdgpu_dm_crtc_init(struct amdgpu_display_manager *dm,
-                       struct drm_plane *plane,
-                       uint32_t link_index);
-int amdgpu_dm_connector_init(struct amdgpu_display_manager *dm,
-                       struct amdgpu_connector *amdgpu_connector,
-                       uint32_t link_index,
-                       struct amdgpu_encoder *amdgpu_encoder);
-int amdgpu_dm_encoder_init(
-       struct drm_device *dev,
-       struct amdgpu_encoder *aencoder,
-       uint32_t link_index);
-
-void amdgpu_dm_crtc_destroy(struct drm_crtc *crtc);
-void amdgpu_dm_connector_destroy(struct drm_connector *connector);
-void amdgpu_dm_encoder_destroy(struct drm_encoder *encoder);
-
-int amdgpu_dm_connector_get_modes(struct drm_connector *connector);
-
-int amdgpu_dm_atomic_commit(
-               struct drm_device *dev,
-               struct drm_atomic_state *state,
-               bool nonblock);
-
-void amdgpu_dm_atomic_commit_tail(
-       struct drm_atomic_state *state);
-
-int amdgpu_dm_atomic_check(struct drm_device *dev,
-                               struct drm_atomic_state *state);
-
-void amdgpu_dm_connector_funcs_reset(struct drm_connector *connector);
-struct drm_connector_state *amdgpu_dm_connector_atomic_duplicate_state(
-       struct drm_connector *connector);
-
-int amdgpu_dm_connector_atomic_set_property(
-       struct drm_connector *connector,
-       struct drm_connector_state *state,
-       struct drm_property *property,
-       uint64_t val);
-
-int amdgpu_dm_connector_atomic_get_property(
-       struct drm_connector *connector,
-       const struct drm_connector_state *state,
-       struct drm_property *property,
-       uint64_t *val);
-
-int amdgpu_dm_get_encoder_crtc_mask(struct amdgpu_device *adev);
-
-void amdgpu_dm_connector_init_helper(
-       struct amdgpu_display_manager *dm,
-       struct amdgpu_connector *aconnector,
-       int connector_type,
-       struct dc_link *link,
-       int link_index);
-
-int amdgpu_dm_connector_mode_valid(
-       struct drm_connector *connector,
-       struct drm_display_mode *mode);
-
-void dm_restore_drm_connector_state(struct drm_device *dev, struct drm_connector *connector);
-
-void amdgpu_dm_add_sink_to_freesync_module(
-               struct drm_connector *connector,
-               struct edid *edid);
-
-void amdgpu_dm_remove_sink_from_freesync_module(
-               struct drm_connector *connector);
-
-extern const struct drm_encoder_helper_funcs amdgpu_dm_encoder_helper_funcs;
-
-#endif         /* __AMDGPU_DM_TYPES_H__ */