revert to 2018/04/20 10:00:00, before oops

9 files changed, 3231 insertions, 0 deletions

diff --git a/sys/dev/pci/drm/radeon/evergreen_blit_kms.c b/sys/dev/pci/drm/radeon/evergreen_blit_kms.c
index e69de29bb2d..149bb1de0f6 100644
--- a/sys/dev/pci/drm/radeon/evergreen_blit_kms.c
+++ b/sys/dev/pci/drm/radeon/evergreen_blit_kms.c
@@ -0,0 +1,730 @@
+/*	$OpenBSD: evergreen_blit_kms.c,v 1.5 2018/04/20 21:12:50 naddy Exp $	*/
+/*
+ * Copyright 2010 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 (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS 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:
+ *     Alex Deucher <alexander.deucher@amd.com>
+ */
+
+#include <dev/pci/drm/drmP.h>
+#include <dev/pci/drm/radeon_drm.h>
+#include "radeon.h"
+
+#include "evergreend.h"
+#include "evergreen_blit_shaders.h"
+#include "cayman_blit_shaders.h"
+#include "radeon_blit_common.h"
+
+/* emits 17 */
+static void
+set_render_target(struct radeon_device *rdev, int format,
+		  int w, int h, u64 gpu_addr)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u32 cb_color_info;
+	int pitch, slice;
+
+	h = roundup2(h, 8);
+	if (h < 8)
+		h = 8;
+
+	cb_color_info = CB_FORMAT(format) |
+		CB_SOURCE_FORMAT(CB_SF_EXPORT_NORM) |
+		CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+	pitch = (w / 8) - 1;
+	slice = ((w * h) / 64) - 1;
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 15));
+	radeon_ring_write(ring, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_START) >> 2);
+	radeon_ring_write(ring, gpu_addr >> 8);
+	radeon_ring_write(ring, pitch);
+	radeon_ring_write(ring, slice);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, cb_color_info);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, (w - 1) | ((h - 1) << 16));
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+}
+
+/* emits 5dw */
+static void
+cp_set_surface_sync(struct radeon_device *rdev,
+		    u32 sync_type, u32 size,
+		    u64 mc_addr)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u32 cp_coher_size;
+
+	if (size == 0xffffffff)
+		cp_coher_size = 0xffffffff;
+	else
+		cp_coher_size = ((size + 255) >> 8);
+
+	if (rdev->family >= CHIP_CAYMAN) {
+		/* CP_COHER_CNTL2 has to be set manually when submitting a surface_sync
+		 * to the RB directly. For IBs, the CP programs this as part of the
+		 * surface_sync packet.
+		 */
+		radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+		radeon_ring_write(ring, (0x85e8 - PACKET3_SET_CONFIG_REG_START) >> 2);
+		radeon_ring_write(ring, 0); /* CP_COHER_CNTL2 */
+	}
+	radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3));
+	radeon_ring_write(ring, sync_type);
+	radeon_ring_write(ring, cp_coher_size);
+	radeon_ring_write(ring, mc_addr >> 8);
+	radeon_ring_write(ring, 10); /* poll interval */
+}
+
+/* emits 11dw + 1 surface sync = 16dw */
+static void
+set_shaders(struct radeon_device *rdev)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u64 gpu_addr;
+
+	/* VS */
+	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 3));
+	radeon_ring_write(ring, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_START) >> 2);
+	radeon_ring_write(ring, gpu_addr >> 8);
+	radeon_ring_write(ring, 2);
+	radeon_ring_write(ring, 0);
+
+	/* PS */
+	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.ps_offset;
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 4));
+	radeon_ring_write(ring, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_START) >> 2);
+	radeon_ring_write(ring, gpu_addr >> 8);
+	radeon_ring_write(ring, 1);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 2);
+
+	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
+	cp_set_surface_sync(rdev, PACKET3_SH_ACTION_ENA, 512, gpu_addr);
+}
+
+/* emits 10 + 1 sync (5) = 15 */
+static void
+set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u32 sq_vtx_constant_word2, sq_vtx_constant_word3;
+
+	/* high addr, stride */
+	sq_vtx_constant_word2 = SQ_VTXC_BASE_ADDR_HI(upper_32_bits(gpu_addr) & 0xff) |
+		SQ_VTXC_STRIDE(16);
+#ifdef __BIG_ENDIAN
+	sq_vtx_constant_word2 |= SQ_VTXC_ENDIAN_SWAP(SQ_ENDIAN_8IN32);
+#endif
+	/* xyzw swizzles */
+	sq_vtx_constant_word3 = SQ_VTCX_SEL_X(SQ_SEL_X) |
+		SQ_VTCX_SEL_Y(SQ_SEL_Y) |
+		SQ_VTCX_SEL_Z(SQ_SEL_Z) |
+		SQ_VTCX_SEL_W(SQ_SEL_W);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_RESOURCE, 8));
+	radeon_ring_write(ring, 0x580);
+	radeon_ring_write(ring, gpu_addr & 0xffffffff);
+	radeon_ring_write(ring, 48 - 1); /* size */
+	radeon_ring_write(ring, sq_vtx_constant_word2);
+	radeon_ring_write(ring, sq_vtx_constant_word3);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, S__SQ_CONSTANT_TYPE(SQ_TEX_VTX_VALID_BUFFER));
+
+	if ((rdev->family == CHIP_CEDAR) ||
+	    (rdev->family == CHIP_PALM) ||
+	    (rdev->family == CHIP_SUMO) ||
+	    (rdev->family == CHIP_SUMO2) ||
+	    (rdev->family == CHIP_CAICOS))
+		cp_set_surface_sync(rdev,
+				    PACKET3_TC_ACTION_ENA, 48, gpu_addr);
+	else
+		cp_set_surface_sync(rdev,
+				    PACKET3_VC_ACTION_ENA, 48, gpu_addr);
+
+}
+
+/* emits 10 */
+static void
+set_tex_resource(struct radeon_device *rdev,
+		 int format, int w, int h, int pitch,
+		 u64 gpu_addr, u32 size)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u32 sq_tex_resource_word0, sq_tex_resource_word1;
+	u32 sq_tex_resource_word4, sq_tex_resource_word7;
+
+	if (h < 1)
+		h = 1;
+
+	sq_tex_resource_word0 = TEX_DIM(SQ_TEX_DIM_2D);
+	sq_tex_resource_word0 |= ((((pitch >> 3) - 1) << 6) |
+				  ((w - 1) << 18));
+	sq_tex_resource_word1 = ((h - 1) << 0) |
+				TEX_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+	/* xyzw swizzles */
+	sq_tex_resource_word4 = TEX_DST_SEL_X(SQ_SEL_X) |
+				TEX_DST_SEL_Y(SQ_SEL_Y) |
+				TEX_DST_SEL_Z(SQ_SEL_Z) |
+				TEX_DST_SEL_W(SQ_SEL_W);
+
+	sq_tex_resource_word7 = format |
+		S__SQ_CONSTANT_TYPE(SQ_TEX_VTX_VALID_TEXTURE);
+
+	cp_set_surface_sync(rdev,
+			    PACKET3_TC_ACTION_ENA, size, gpu_addr);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_RESOURCE, 8));
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, sq_tex_resource_word0);
+	radeon_ring_write(ring, sq_tex_resource_word1);
+	radeon_ring_write(ring, gpu_addr >> 8);
+	radeon_ring_write(ring, gpu_addr >> 8);
+	radeon_ring_write(ring, sq_tex_resource_word4);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, sq_tex_resource_word7);
+}
+
+/* emits 12 */
+static void
+set_scissors(struct radeon_device *rdev, int x1, int y1,
+	     int x2, int y2)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	/* workaround some hw bugs */
+	if (x2 == 0)
+		x1 = 1;
+	if (y2 == 0)
+		y1 = 1;
+	if (rdev->family >= CHIP_CAYMAN) {
+		if ((x2 == 1) && (y2 == 1))
+			x2 = 2;
+	}
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+	radeon_ring_write(ring, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
+	radeon_ring_write(ring, (x1 << 0) | (y1 << 16));
+	radeon_ring_write(ring, (x2 << 0) | (y2 << 16));
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+	radeon_ring_write(ring, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
+	radeon_ring_write(ring, (x1 << 0) | (y1 << 16) | (1 << 31));
+	radeon_ring_write(ring, (x2 << 0) | (y2 << 16));
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+	radeon_ring_write(ring, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2);
+	radeon_ring_write(ring, (x1 << 0) | (y1 << 16) | (1 << 31));
+	radeon_ring_write(ring, (x2 << 0) | (y2 << 16));
+}
+
+/* emits 10 */
+static void
+draw_auto(struct radeon_device *rdev)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+	radeon_ring_write(ring, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_START) >> 2);
+	radeon_ring_write(ring, DI_PT_RECTLIST);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_INDEX_TYPE, 0));
+	radeon_ring_write(ring,
+#ifdef __BIG_ENDIAN
+			  (2 << 2) |
+#endif
+			  DI_INDEX_SIZE_16_BIT);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_NUM_INSTANCES, 0));
+	radeon_ring_write(ring, 1);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1));
+	radeon_ring_write(ring, 3);
+	radeon_ring_write(ring, DI_SRC_SEL_AUTO_INDEX);
+
+}
+
+/* emits 39 */
+static void
+set_default_state(struct radeon_device *rdev)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2, sq_gpr_resource_mgmt_3;
+	u32 sq_thread_resource_mgmt, sq_thread_resource_mgmt_2;
+	u32 sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2, sq_stack_resource_mgmt_3;
+	int num_ps_gprs, num_vs_gprs, num_temp_gprs;
+	int num_gs_gprs, num_es_gprs, num_hs_gprs, num_ls_gprs;
+	int num_ps_threads, num_vs_threads, num_gs_threads, num_es_threads;
+	int num_hs_threads, num_ls_threads;
+	int num_ps_stack_entries, num_vs_stack_entries, num_gs_stack_entries, num_es_stack_entries;
+	int num_hs_stack_entries, num_ls_stack_entries;
+	u64 gpu_addr;
+	int dwords;
+
+	/* set clear context state */
+	radeon_ring_write(ring, PACKET3(PACKET3_CLEAR_STATE, 0));
+	radeon_ring_write(ring, 0);
+
+	if (rdev->family < CHIP_CAYMAN) {
+		switch (rdev->family) {
+		case CHIP_CEDAR:
+		default:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 96;
+			num_vs_threads = 16;
+			num_gs_threads = 16;
+			num_es_threads = 16;
+			num_hs_threads = 16;
+			num_ls_threads = 16;
+			num_ps_stack_entries = 42;
+			num_vs_stack_entries = 42;
+			num_gs_stack_entries = 42;
+			num_es_stack_entries = 42;
+			num_hs_stack_entries = 42;
+			num_ls_stack_entries = 42;
+			break;
+		case CHIP_REDWOOD:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 128;
+			num_vs_threads = 20;
+			num_gs_threads = 20;
+			num_es_threads = 20;
+			num_hs_threads = 20;
+			num_ls_threads = 20;
+			num_ps_stack_entries = 42;
+			num_vs_stack_entries = 42;
+			num_gs_stack_entries = 42;
+			num_es_stack_entries = 42;
+			num_hs_stack_entries = 42;
+			num_ls_stack_entries = 42;
+			break;
+		case CHIP_JUNIPER:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 128;
+			num_vs_threads = 20;
+			num_gs_threads = 20;
+			num_es_threads = 20;
+			num_hs_threads = 20;
+			num_ls_threads = 20;
+			num_ps_stack_entries = 85;
+			num_vs_stack_entries = 85;
+			num_gs_stack_entries = 85;
+			num_es_stack_entries = 85;
+			num_hs_stack_entries = 85;
+			num_ls_stack_entries = 85;
+			break;
+		case CHIP_CYPRESS:
+		case CHIP_HEMLOCK:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 128;
+			num_vs_threads = 20;
+			num_gs_threads = 20;
+			num_es_threads = 20;
+			num_hs_threads = 20;
+			num_ls_threads = 20;
+			num_ps_stack_entries = 85;
+			num_vs_stack_entries = 85;
+			num_gs_stack_entries = 85;
+			num_es_stack_entries = 85;
+			num_hs_stack_entries = 85;
+			num_ls_stack_entries = 85;
+			break;
+		case CHIP_PALM:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 96;
+			num_vs_threads = 16;
+			num_gs_threads = 16;
+			num_es_threads = 16;
+			num_hs_threads = 16;
+			num_ls_threads = 16;
+			num_ps_stack_entries = 42;
+			num_vs_stack_entries = 42;
+			num_gs_stack_entries = 42;
+			num_es_stack_entries = 42;
+			num_hs_stack_entries = 42;
+			num_ls_stack_entries = 42;
+			break;
+		case CHIP_SUMO:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 96;
+			num_vs_threads = 25;
+			num_gs_threads = 25;
+			num_es_threads = 25;
+			num_hs_threads = 25;
+			num_ls_threads = 25;
+			num_ps_stack_entries = 42;
+			num_vs_stack_entries = 42;
+			num_gs_stack_entries = 42;
+			num_es_stack_entries = 42;
+			num_hs_stack_entries = 42;
+			num_ls_stack_entries = 42;
+			break;
+		case CHIP_SUMO2:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 96;
+			num_vs_threads = 25;
+			num_gs_threads = 25;
+			num_es_threads = 25;
+			num_hs_threads = 25;
+			num_ls_threads = 25;
+			num_ps_stack_entries = 85;
+			num_vs_stack_entries = 85;
+			num_gs_stack_entries = 85;
+			num_es_stack_entries = 85;
+			num_hs_stack_entries = 85;
+			num_ls_stack_entries = 85;
+			break;
+		case CHIP_BARTS:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 128;
+			num_vs_threads = 20;
+			num_gs_threads = 20;
+			num_es_threads = 20;
+			num_hs_threads = 20;
+			num_ls_threads = 20;
+			num_ps_stack_entries = 85;
+			num_vs_stack_entries = 85;
+			num_gs_stack_entries = 85;
+			num_es_stack_entries = 85;
+			num_hs_stack_entries = 85;
+			num_ls_stack_entries = 85;
+			break;
+		case CHIP_TURKS:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 128;
+			num_vs_threads = 20;
+			num_gs_threads = 20;
+			num_es_threads = 20;
+			num_hs_threads = 20;
+			num_ls_threads = 20;
+			num_ps_stack_entries = 42;
+			num_vs_stack_entries = 42;
+			num_gs_stack_entries = 42;
+			num_es_stack_entries = 42;
+			num_hs_stack_entries = 42;
+			num_ls_stack_entries = 42;
+			break;
+		case CHIP_CAICOS:
+			num_ps_gprs = 93;
+			num_vs_gprs = 46;
+			num_temp_gprs = 4;
+			num_gs_gprs = 31;
+			num_es_gprs = 31;
+			num_hs_gprs = 23;
+			num_ls_gprs = 23;
+			num_ps_threads = 128;
+			num_vs_threads = 10;
+			num_gs_threads = 10;
+			num_es_threads = 10;
+			num_hs_threads = 10;
+			num_ls_threads = 10;
+			num_ps_stack_entries = 42;
+			num_vs_stack_entries = 42;
+			num_gs_stack_entries = 42;
+			num_es_stack_entries = 42;
+			num_hs_stack_entries = 42;
+			num_ls_stack_entries = 42;
+			break;
+		}
+
+		if ((rdev->family == CHIP_CEDAR) ||
+		    (rdev->family == CHIP_PALM) ||
+		    (rdev->family == CHIP_SUMO) ||
+		    (rdev->family == CHIP_SUMO2) ||
+		    (rdev->family == CHIP_CAICOS))
+			sq_config = 0;
+		else
+			sq_config = VC_ENABLE;
+
+		sq_config |= (EXPORT_SRC_C |
+			      CS_PRIO(0) |
+			      LS_PRIO(0) |
+			      HS_PRIO(0) |
+			      PS_PRIO(0) |
+			      VS_PRIO(1) |
+			      GS_PRIO(2) |
+			      ES_PRIO(3));
+
+		sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(num_ps_gprs) |
+					  NUM_VS_GPRS(num_vs_gprs) |
+					  NUM_CLAUSE_TEMP_GPRS(num_temp_gprs));
+		sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(num_gs_gprs) |
+					  NUM_ES_GPRS(num_es_gprs));
+		sq_gpr_resource_mgmt_3 = (NUM_HS_GPRS(num_hs_gprs) |
+					  NUM_LS_GPRS(num_ls_gprs));
+		sq_thread_resource_mgmt = (NUM_PS_THREADS(num_ps_threads) |
+					   NUM_VS_THREADS(num_vs_threads) |
+					   NUM_GS_THREADS(num_gs_threads) |
+					   NUM_ES_THREADS(num_es_threads));
+		sq_thread_resource_mgmt_2 = (NUM_HS_THREADS(num_hs_threads) |
+					     NUM_LS_THREADS(num_ls_threads));
+		sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(num_ps_stack_entries) |
+					    NUM_VS_STACK_ENTRIES(num_vs_stack_entries));
+		sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(num_gs_stack_entries) |
+					    NUM_ES_STACK_ENTRIES(num_es_stack_entries));
+		sq_stack_resource_mgmt_3 = (NUM_HS_STACK_ENTRIES(num_hs_stack_entries) |
+					    NUM_LS_STACK_ENTRIES(num_ls_stack_entries));
+
+		/* disable dyn gprs */
+		radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+		radeon_ring_write(ring, (SQ_DYN_GPR_CNTL_PS_FLUSH_REQ - PACKET3_SET_CONFIG_REG_START) >> 2);
+		radeon_ring_write(ring, 0);
+
+		/* setup LDS */
+		radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+		radeon_ring_write(ring, (SQ_LDS_RESOURCE_MGMT - PACKET3_SET_CONFIG_REG_START) >> 2);
+		radeon_ring_write(ring, 0x10001000);
+
+		/* SQ config */
+		radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 11));
+		radeon_ring_write(ring, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_START) >> 2);
+		radeon_ring_write(ring, sq_config);
+		radeon_ring_write(ring, sq_gpr_resource_mgmt_1);
+		radeon_ring_write(ring, sq_gpr_resource_mgmt_2);
+		radeon_ring_write(ring, sq_gpr_resource_mgmt_3);
+		radeon_ring_write(ring, 0);
+		radeon_ring_write(ring, 0);
+		radeon_ring_write(ring, sq_thread_resource_mgmt);
+		radeon_ring_write(ring, sq_thread_resource_mgmt_2);
+		radeon_ring_write(ring, sq_stack_resource_mgmt_1);
+		radeon_ring_write(ring, sq_stack_resource_mgmt_2);
+		radeon_ring_write(ring, sq_stack_resource_mgmt_3);
+	}
+
+	/* CONTEXT_CONTROL */
+	radeon_ring_write(ring, 0xc0012800);
+	radeon_ring_write(ring, 0x80000000);
+	radeon_ring_write(ring, 0x80000000);
+
+	/* SQ_VTX_BASE_VTX_LOC */
+	radeon_ring_write(ring, 0xc0026f00);
+	radeon_ring_write(ring, 0x00000000);
+	radeon_ring_write(ring, 0x00000000);
+	radeon_ring_write(ring, 0x00000000);
+
+	/* SET_SAMPLER */
+	radeon_ring_write(ring, 0xc0036e00);
+	radeon_ring_write(ring, 0x00000000);
+	radeon_ring_write(ring, 0x00000012);
+	radeon_ring_write(ring, 0x00000000);
+	radeon_ring_write(ring, 0x00000000);
+
+	/* set to DX10/11 mode */
+	radeon_ring_write(ring, PACKET3(PACKET3_MODE_CONTROL, 0));
+	radeon_ring_write(ring, 1);
+
+	/* emit an IB pointing at default state */
+	dwords = roundup2(rdev->r600_blit.state_len, 0x10);
+	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset;
+	radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
+	radeon_ring_write(ring, gpu_addr & 0xFFFFFFFC);
+	radeon_ring_write(ring, upper_32_bits(gpu_addr) & 0xFF);
+	radeon_ring_write(ring, dwords);
+
+}
+
+int evergreen_blit_init(struct radeon_device *rdev)
+{
+	u32 obj_size;
+	int i, r, dwords;
+	void *ptr;
+	u32 packet2s[16];
+	int num_packet2s = 0;
+
+	rdev->r600_blit.primitives.set_render_target = set_render_target;
+	rdev->r600_blit.primitives.cp_set_surface_sync = cp_set_surface_sync;
+	rdev->r600_blit.primitives.set_shaders = set_shaders;
+	rdev->r600_blit.primitives.set_vtx_resource = set_vtx_resource;
+	rdev->r600_blit.primitives.set_tex_resource = set_tex_resource;
+	rdev->r600_blit.primitives.set_scissors = set_scissors;
+	rdev->r600_blit.primitives.draw_auto = draw_auto;
+	rdev->r600_blit.primitives.set_default_state = set_default_state;
+
+	rdev->r600_blit.ring_size_common = 8; /* sync semaphore */
+	rdev->r600_blit.ring_size_common += 55; /* shaders + def state */
+	rdev->r600_blit.ring_size_common += 16; /* fence emit for VB IB */
+	rdev->r600_blit.ring_size_common += 5; /* done copy */
+	rdev->r600_blit.ring_size_common += 16; /* fence emit for done copy */
+
+	rdev->r600_blit.ring_size_per_loop = 74;
+	if (rdev->family >= CHIP_CAYMAN)
+		rdev->r600_blit.ring_size_per_loop += 9; /* additional DWs for surface sync */
+
+	rdev->r600_blit.max_dim = 16384;
+
+	rdev->r600_blit.state_offset = 0;
+
+	if (rdev->family < CHIP_CAYMAN)
+		rdev->r600_blit.state_len = evergreen_default_size;
+	else
+		rdev->r600_blit.state_len = cayman_default_size;
+
+	dwords = rdev->r600_blit.state_len;
+	while (dwords & 0xf) {
+		packet2s[num_packet2s++] = cpu_to_le32(PACKET2(0));
+		dwords++;
+	}
+
+	obj_size = dwords * 4;
+	obj_size = roundup2(obj_size, 256);
+
+	rdev->r600_blit.vs_offset = obj_size;
+	if (rdev->family < CHIP_CAYMAN)
+		obj_size += evergreen_vs_size * 4;
+	else
+		obj_size += cayman_vs_size * 4;
+	obj_size = roundup2(obj_size, 256);
+
+	rdev->r600_blit.ps_offset = obj_size;
+	if (rdev->family < CHIP_CAYMAN)
+		obj_size += evergreen_ps_size * 4;
+	else
+		obj_size += cayman_ps_size * 4;
+	obj_size = roundup2(obj_size, 256);
+
+	/* pin copy shader into vram if not already initialized */
+	if (!rdev->r600_blit.shader_obj) {
+		r = radeon_bo_create(rdev, obj_size, PAGE_SIZE, true,
+				     RADEON_GEM_DOMAIN_VRAM,
+				     NULL, &rdev->r600_blit.shader_obj);
+		if (r) {
+			DRM_ERROR("evergreen failed to allocate shader\n");
+			return r;
+		}
+
+		r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+		if (unlikely(r != 0))
+			return r;
+		r = radeon_bo_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM,
+				  &rdev->r600_blit.shader_gpu_addr);
+		radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+		if (r) {
+			dev_err(rdev->dev, "(%d) pin blit object failed\n", r);
+			return r;
+		}
+	}
+
+	DRM_DEBUG("evergreen blit allocated bo %08x vs %08x ps %08x\n",
+		  obj_size,
+		  rdev->r600_blit.vs_offset, rdev->r600_blit.ps_offset);
+
+	r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+	if (unlikely(r != 0))
+		return r;
+	r = radeon_bo_kmap(rdev->r600_blit.shader_obj, &ptr);
+	if (r) {
+		DRM_ERROR("failed to map blit object %d\n", r);
+		return r;
+	}
+
+	if (rdev->family < CHIP_CAYMAN) {
+		memcpy_toio(ptr + rdev->r600_blit.state_offset,
+			    evergreen_default_state, rdev->r600_blit.state_len * 4);
+
+		if (num_packet2s)
+			memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4),
+				    packet2s, num_packet2s * 4);
+		for (i = 0; i < evergreen_vs_size; i++)
+			*(u32 *)((unsigned long)ptr + rdev->r600_blit.vs_offset + i * 4) = cpu_to_le32(evergreen_vs[i]);
+		for (i = 0; i < evergreen_ps_size; i++)
+			*(u32 *)((unsigned long)ptr + rdev->r600_blit.ps_offset + i * 4) = cpu_to_le32(evergreen_ps[i]);
+	} else {
+		memcpy_toio(ptr + rdev->r600_blit.state_offset,
+			    cayman_default_state, rdev->r600_blit.state_len * 4);
+
+		if (num_packet2s)
+			memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4),
+				    packet2s, num_packet2s * 4);
+		for (i = 0; i < cayman_vs_size; i++)
+			*(u32 *)((unsigned long)ptr + rdev->r600_blit.vs_offset + i * 4) = cpu_to_le32(cayman_vs[i]);
+		for (i = 0; i < cayman_ps_size; i++)
+			*(u32 *)((unsigned long)ptr + rdev->r600_blit.ps_offset + i * 4) = cpu_to_le32(cayman_ps[i]);
+	}
+	radeon_bo_kunmap(rdev->r600_blit.shader_obj);
+	radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+
+	radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size);
+	return 0;
+}
diff --git a/sys/dev/pci/drm/radeon/r600_audio.c b/sys/dev/pci/drm/radeon/r600_audio.c
index e69de29bb2d..b7aecc05ea3 100644
--- a/sys/dev/pci/drm/radeon/r600_audio.c
+++ b/sys/dev/pci/drm/radeon/r600_audio.c
@@ -0,0 +1,255 @@
+/*	$OpenBSD: r600_audio.c,v 1.5 2018/04/20 21:12:50 naddy Exp $	*/
+/*
+ * Copyright 2008 Advanced Micro Devices, Inc.
+ * Copyright 2008 Red Hat Inc.
+ * Copyright 2009 Christian König.
+ *
+ * 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: Christian König
+ */
+#include <dev/pci/drm/drmP.h>
+#include "radeon.h"
+#include "radeon_reg.h"
+#include "radeon_asic.h"
+#include "atom.h"
+
+/*
+ * check if enc_priv stores radeon_encoder_atom_dig
+ */
+static bool radeon_dig_encoder(struct drm_encoder *encoder)
+{
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	switch (radeon_encoder->encoder_id) {
+	case ENCODER_OBJECT_ID_INTERNAL_LVDS:
+	case ENCODER_OBJECT_ID_INTERNAL_TMDS1:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1:
+	case ENCODER_OBJECT_ID_INTERNAL_LVTM1:
+	case ENCODER_OBJECT_ID_INTERNAL_DVO1:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1:
+	case ENCODER_OBJECT_ID_INTERNAL_DDI:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+		return true;
+	}
+	return false;
+}
+
+/*
+ * check if the chipset is supported
+ */
+static int r600_audio_chipset_supported(struct radeon_device *rdev)
+{
+	return (rdev->family >= CHIP_R600 && !ASIC_IS_DCE6(rdev))
+		|| rdev->family == CHIP_RS600
+		|| rdev->family == CHIP_RS690
+		|| rdev->family == CHIP_RS740;
+}
+
+struct r600_audio r600_audio_status(struct radeon_device *rdev)
+{
+	struct r600_audio status;
+	uint32_t value;
+
+	value = RREG32(R600_AUDIO_RATE_BPS_CHANNEL);
+
+	/* number of channels */
+	status.channels = (value & 0x7) + 1;
+
+	/* bits per sample */
+	switch ((value & 0xF0) >> 4) {
+	case 0x0:
+		status.bits_per_sample = 8;
+		break;
+	case 0x1:
+		status.bits_per_sample = 16;
+		break;
+	case 0x2:
+		status.bits_per_sample = 20;
+		break;
+	case 0x3:
+		status.bits_per_sample = 24;
+		break;
+	case 0x4:
+		status.bits_per_sample = 32;
+		break;
+	default:
+		dev_err(rdev->dev, "Unknown bits per sample 0x%x, using 16\n",
+			(int)value);
+		status.bits_per_sample = 16;
+	}
+
+	/* current sampling rate in HZ */
+	if (value & 0x4000)
+		status.rate = 44100;
+	else
+		status.rate = 48000;
+	status.rate *= ((value >> 11) & 0x7) + 1;
+	status.rate /= ((value >> 8) & 0x7) + 1;
+
+	value = RREG32(R600_AUDIO_STATUS_BITS);
+
+	/* iec 60958 status bits */
+	status.status_bits = value & 0xff;
+
+	/* iec 60958 category code */
+	status.category_code = (value >> 8) & 0xff;
+
+	return status;
+}
+
+/*
+ * update all hdmi interfaces with current audio parameters
+ */
+void r600_audio_update_hdmi(void *arg1)
+{
+	struct radeon_device *rdev = arg1;
+	struct drm_device *dev = rdev->ddev;
+	struct r600_audio audio_status = r600_audio_status(rdev);
+	struct drm_encoder *encoder;
+	bool changed = false;
+
+	if (rdev->audio_status.channels != audio_status.channels ||
+	    rdev->audio_status.rate != audio_status.rate ||
+	    rdev->audio_status.bits_per_sample != audio_status.bits_per_sample ||
+	    rdev->audio_status.status_bits != audio_status.status_bits ||
+	    rdev->audio_status.category_code != audio_status.category_code) {
+		rdev->audio_status = audio_status;
+		changed = true;
+	}
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		if (!radeon_dig_encoder(encoder))
+			continue;
+		if (changed || r600_hdmi_buffer_status_changed(encoder))
+			r600_hdmi_update_audio_settings(encoder);
+	}
+}
+
+/*
+ * turn on/off audio engine
+ */
+static void r600_audio_engine_enable(struct radeon_device *rdev, bool enable)
+{
+	u32 value = 0;
+	DRM_INFO("%s audio support\n", enable ? "Enabling" : "Disabling");
+	if (ASIC_IS_DCE4(rdev)) {
+		if (enable) {
+			value |= 0x81000000; /* Required to enable audio */
+			value |= 0x0e1000f0; /* fglrx sets that too */
+		}
+		WREG32(EVERGREEN_AUDIO_ENABLE, value);
+	} else {
+		WREG32_P(R600_AUDIO_ENABLE,
+			 enable ? 0x81000000 : 0x0, ~0x81000000);
+	}
+	rdev->audio_enabled = enable;
+}
+
+/*
+ * initialize the audio vars
+ */
+int r600_audio_init(struct radeon_device *rdev)
+{
+	if (!radeon_audio || !r600_audio_chipset_supported(rdev))
+		return 0;
+
+	r600_audio_engine_enable(rdev, true);
+
+	rdev->audio_status.channels = -1;
+	rdev->audio_status.rate = -1;
+	rdev->audio_status.bits_per_sample = -1;
+	rdev->audio_status.status_bits = 0;
+	rdev->audio_status.category_code = 0;
+
+	return 0;
+}
+
+/*
+ * atach the audio codec to the clock source of the encoder
+ */
+void r600_audio_set_clock(struct drm_encoder *encoder, int clock)
+{
+	struct drm_device *dev = encoder->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
+	struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv;
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc);
+	int base_rate = 48000;
+
+	switch (radeon_encoder->encoder_id) {
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1:
+	case ENCODER_OBJECT_ID_INTERNAL_LVTM1:
+		WREG32_P(R600_AUDIO_TIMING, 0, ~0x301);
+		break;
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA:
+		WREG32_P(R600_AUDIO_TIMING, 0x100, ~0x301);
+		break;
+	default:
+		dev_err(rdev->dev, "Unsupported encoder type 0x%02X\n",
+			  radeon_encoder->encoder_id);
+		return;
+	}
+
+	if (ASIC_IS_DCE4(rdev)) {
+		/* TODO: other PLLs? */
+		WREG32(EVERGREEN_AUDIO_PLL1_MUL, base_rate * 10);
+		WREG32(EVERGREEN_AUDIO_PLL1_DIV, clock * 10);
+		WREG32(EVERGREEN_AUDIO_PLL1_UNK, 0x00000071);
+
+		/* Select DTO source */
+		WREG32(0x5ac, radeon_crtc->crtc_id);
+	} else {
+		switch (dig->dig_encoder) {
+		case 0:
+			WREG32(R600_AUDIO_PLL1_MUL, base_rate * 50);
+			WREG32(R600_AUDIO_PLL1_DIV, clock * 100);
+			WREG32(R600_AUDIO_CLK_SRCSEL, 0);
+			break;
+
+		case 1:
+			WREG32(R600_AUDIO_PLL2_MUL, base_rate * 50);
+			WREG32(R600_AUDIO_PLL2_DIV, clock * 100);
+			WREG32(R600_AUDIO_CLK_SRCSEL, 1);
+			break;
+		default:
+			dev_err(rdev->dev,
+				"Unsupported DIG on encoder 0x%02X\n",
+				radeon_encoder->encoder_id);
+			return;
+		}
+	}
+}
+
+/*
+ * release the audio timer
+ * TODO: How to do this correctly on SMP systems?
+ */
+void r600_audio_fini(struct radeon_device *rdev)
+{
+	if (!rdev->audio_enabled)
+		return;
+
+	r600_audio_engine_enable(rdev, false);
+}
diff --git a/sys/dev/pci/drm/radeon/r600_blit_kms.c b/sys/dev/pci/drm/radeon/r600_blit_kms.c
index e69de29bb2d..47157d09a4d 100644
--- a/sys/dev/pci/drm/radeon/r600_blit_kms.c
+++ b/sys/dev/pci/drm/radeon/r600_blit_kms.c
@@ -0,0 +1,786 @@
+/*	$OpenBSD: r600_blit_kms.c,v 1.5 2018/04/20 21:12:50 naddy Exp $	*/
+/*
+ * Copyright 2009 Advanced Micro Devices, Inc.
+ * Copyright 2009 Red Hat 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 (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS 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.
+ *
+ */
+
+#include <dev/pci/drm/drmP.h>
+#include <dev/pci/drm/radeon_drm.h>
+#include "radeon.h"
+
+#include "r600d.h"
+#include "r600_blit_shaders.h"
+#include "radeon_blit_common.h"
+
+/* emits 21 on rv770+, 23 on r600 */
+static void
+set_render_target(struct radeon_device *rdev, int format,
+		  int w, int h, u64 gpu_addr)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u32 cb_color_info;
+	int pitch, slice;
+
+	h = roundup2(h, 8);
+	if (h < 8)
+		h = 8;
+
+	cb_color_info = CB_FORMAT(format) |
+		CB_SOURCE_FORMAT(CB_SF_EXPORT_NORM) |
+		CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
+	pitch = (w / 8) - 1;
+	slice = ((w * h) / 64) - 1;
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, gpu_addr >> 8);
+
+	if (rdev->family > CHIP_R600 && rdev->family < CHIP_RV770) {
+		radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_BASE_UPDATE, 0));
+		radeon_ring_write(ring, 2 << 0);
+	}
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (CB_COLOR0_SIZE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, (pitch << 0) | (slice << 10));
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (CB_COLOR0_VIEW - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, 0);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (CB_COLOR0_INFO - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, cb_color_info);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (CB_COLOR0_TILE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, 0);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (CB_COLOR0_FRAG - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, 0);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (CB_COLOR0_MASK - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, 0);
+}
+
+/* emits 5dw */
+static void
+cp_set_surface_sync(struct radeon_device *rdev,
+		    u32 sync_type, u32 size,
+		    u64 mc_addr)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u32 cp_coher_size;
+
+	if (size == 0xffffffff)
+		cp_coher_size = 0xffffffff;
+	else
+		cp_coher_size = ((size + 255) >> 8);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3));
+	radeon_ring_write(ring, sync_type);
+	radeon_ring_write(ring, cp_coher_size);
+	radeon_ring_write(ring, mc_addr >> 8);
+	radeon_ring_write(ring, 10); /* poll interval */
+}
+
+/* emits 21dw + 1 surface sync = 26dw */
+static void
+set_shaders(struct radeon_device *rdev)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u64 gpu_addr;
+	u32 sq_pgm_resources;
+
+	/* setup shader regs */
+	sq_pgm_resources = (1 << 0);
+
+	/* VS */
+	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, gpu_addr >> 8);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (SQ_PGM_RESOURCES_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, sq_pgm_resources);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (SQ_PGM_CF_OFFSET_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, 0);
+
+	/* PS */
+	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.ps_offset;
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, gpu_addr >> 8);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (SQ_PGM_RESOURCES_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, sq_pgm_resources | (1 << 28));
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (SQ_PGM_EXPORTS_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, 2);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1));
+	radeon_ring_write(ring, (SQ_PGM_CF_OFFSET_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, 0);
+
+	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset;
+	cp_set_surface_sync(rdev, PACKET3_SH_ACTION_ENA, 512, gpu_addr);
+}
+
+/* emits 9 + 1 sync (5) = 14*/
+static void
+set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u32 sq_vtx_constant_word2;
+
+	sq_vtx_constant_word2 = SQ_VTXC_BASE_ADDR_HI(upper_32_bits(gpu_addr) & 0xff) |
+		SQ_VTXC_STRIDE(16);
+#ifdef __BIG_ENDIAN
+	sq_vtx_constant_word2 |=  SQ_VTXC_ENDIAN_SWAP(SQ_ENDIAN_8IN32);
+#endif
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_RESOURCE, 7));
+	radeon_ring_write(ring, 0x460);
+	radeon_ring_write(ring, gpu_addr & 0xffffffff);
+	radeon_ring_write(ring, 48 - 1);
+	radeon_ring_write(ring, sq_vtx_constant_word2);
+	radeon_ring_write(ring, 1 << 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, SQ_TEX_VTX_VALID_BUFFER << 30);
+
+	if ((rdev->family == CHIP_RV610) ||
+	    (rdev->family == CHIP_RV620) ||
+	    (rdev->family == CHIP_RS780) ||
+	    (rdev->family == CHIP_RS880) ||
+	    (rdev->family == CHIP_RV710))
+		cp_set_surface_sync(rdev,
+				    PACKET3_TC_ACTION_ENA, 48, gpu_addr);
+	else
+		cp_set_surface_sync(rdev,
+				    PACKET3_VC_ACTION_ENA, 48, gpu_addr);
+}
+
+/* emits 9 */
+static void
+set_tex_resource(struct radeon_device *rdev,
+		 int format, int w, int h, int pitch,
+		 u64 gpu_addr, u32 size)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	uint32_t sq_tex_resource_word0, sq_tex_resource_word1, sq_tex_resource_word4;
+
+	if (h < 1)
+		h = 1;
+
+	sq_tex_resource_word0 = S_038000_DIM(V_038000_SQ_TEX_DIM_2D) |
+		S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1);
+	sq_tex_resource_word0 |= S_038000_PITCH((pitch >> 3) - 1) |
+		S_038000_TEX_WIDTH(w - 1);
+
+	sq_tex_resource_word1 = S_038004_DATA_FORMAT(format);
+	sq_tex_resource_word1 |= S_038004_TEX_HEIGHT(h - 1);
+
+	sq_tex_resource_word4 = S_038010_REQUEST_SIZE(1) |
+		S_038010_DST_SEL_X(SQ_SEL_X) |
+		S_038010_DST_SEL_Y(SQ_SEL_Y) |
+		S_038010_DST_SEL_Z(SQ_SEL_Z) |
+		S_038010_DST_SEL_W(SQ_SEL_W);
+
+	cp_set_surface_sync(rdev,
+			    PACKET3_TC_ACTION_ENA, size, gpu_addr);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_RESOURCE, 7));
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, sq_tex_resource_word0);
+	radeon_ring_write(ring, sq_tex_resource_word1);
+	radeon_ring_write(ring, gpu_addr >> 8);
+	radeon_ring_write(ring, gpu_addr >> 8);
+	radeon_ring_write(ring, sq_tex_resource_word4);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, SQ_TEX_VTX_VALID_TEXTURE << 30);
+}
+
+/* emits 12 */
+static void
+set_scissors(struct radeon_device *rdev, int x1, int y1,
+	     int x2, int y2)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+	radeon_ring_write(ring, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, (x1 << 0) | (y1 << 16));
+	radeon_ring_write(ring, (x2 << 0) | (y2 << 16));
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+	radeon_ring_write(ring, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, (x1 << 0) | (y1 << 16) | (1 << 31));
+	radeon_ring_write(ring, (x2 << 0) | (y2 << 16));
+
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+	radeon_ring_write(ring, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, (x1 << 0) | (y1 << 16) | (1 << 31));
+	radeon_ring_write(ring, (x2 << 0) | (y2 << 16));
+}
+
+/* emits 10 */
+static void
+draw_auto(struct radeon_device *rdev)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+	radeon_ring_write(ring, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, DI_PT_RECTLIST);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_INDEX_TYPE, 0));
+	radeon_ring_write(ring,
+#ifdef __BIG_ENDIAN
+			  (2 << 2) |
+#endif
+			  DI_INDEX_SIZE_16_BIT);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_NUM_INSTANCES, 0));
+	radeon_ring_write(ring, 1);
+
+	radeon_ring_write(ring, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1));
+	radeon_ring_write(ring, 3);
+	radeon_ring_write(ring, DI_SRC_SEL_AUTO_INDEX);
+
+}
+
+/* emits 14 */
+static void
+set_default_state(struct radeon_device *rdev)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2;
+	u32 sq_thread_resource_mgmt, sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2;
+	int num_ps_gprs, num_vs_gprs, num_temp_gprs, num_gs_gprs, num_es_gprs;
+	int num_ps_threads, num_vs_threads, num_gs_threads, num_es_threads;
+	int num_ps_stack_entries, num_vs_stack_entries, num_gs_stack_entries, num_es_stack_entries;
+	u64 gpu_addr;
+	int dwords;
+
+	switch (rdev->family) {
+	case CHIP_R600:
+		num_ps_gprs = 192;
+		num_vs_gprs = 56;
+		num_temp_gprs = 4;
+		num_gs_gprs = 0;
+		num_es_gprs = 0;
+		num_ps_threads = 136;
+		num_vs_threads = 48;
+		num_gs_threads = 4;
+		num_es_threads = 4;
+		num_ps_stack_entries = 128;
+		num_vs_stack_entries = 128;
+		num_gs_stack_entries = 0;
+		num_es_stack_entries = 0;
+		break;
+	case CHIP_RV630:
+	case CHIP_RV635:
+		num_ps_gprs = 84;
+		num_vs_gprs = 36;
+		num_temp_gprs = 4;
+		num_gs_gprs = 0;
+		num_es_gprs = 0;
+		num_ps_threads = 144;
+		num_vs_threads = 40;
+		num_gs_threads = 4;
+		num_es_threads = 4;
+		num_ps_stack_entries = 40;
+		num_vs_stack_entries = 40;
+		num_gs_stack_entries = 32;
+		num_es_stack_entries = 16;
+		break;
+	case CHIP_RV610:
+	case CHIP_RV620:
+	case CHIP_RS780:
+	case CHIP_RS880:
+	default:
+		num_ps_gprs = 84;
+		num_vs_gprs = 36;
+		num_temp_gprs = 4;
+		num_gs_gprs = 0;
+		num_es_gprs = 0;
+		num_ps_threads = 136;
+		num_vs_threads = 48;
+		num_gs_threads = 4;
+		num_es_threads = 4;
+		num_ps_stack_entries = 40;
+		num_vs_stack_entries = 40;
+		num_gs_stack_entries = 32;
+		num_es_stack_entries = 16;
+		break;
+	case CHIP_RV670:
+		num_ps_gprs = 144;
+		num_vs_gprs = 40;
+		num_temp_gprs = 4;
+		num_gs_gprs = 0;
+		num_es_gprs = 0;
+		num_ps_threads = 136;
+		num_vs_threads = 48;
+		num_gs_threads = 4;
+		num_es_threads = 4;
+		num_ps_stack_entries = 40;
+		num_vs_stack_entries = 40;
+		num_gs_stack_entries = 32;
+		num_es_stack_entries = 16;
+		break;
+	case CHIP_RV770:
+		num_ps_gprs = 192;
+		num_vs_gprs = 56;
+		num_temp_gprs = 4;
+		num_gs_gprs = 0;
+		num_es_gprs = 0;
+		num_ps_threads = 188;
+		num_vs_threads = 60;
+		num_gs_threads = 0;
+		num_es_threads = 0;
+		num_ps_stack_entries = 256;
+		num_vs_stack_entries = 256;
+		num_gs_stack_entries = 0;
+		num_es_stack_entries = 0;
+		break;
+	case CHIP_RV730:
+	case CHIP_RV740:
+		num_ps_gprs = 84;
+		num_vs_gprs = 36;
+		num_temp_gprs = 4;
+		num_gs_gprs = 0;
+		num_es_gprs = 0;
+		num_ps_threads = 188;
+		num_vs_threads = 60;
+		num_gs_threads = 0;
+		num_es_threads = 0;
+		num_ps_stack_entries = 128;
+		num_vs_stack_entries = 128;
+		num_gs_stack_entries = 0;
+		num_es_stack_entries = 0;
+		break;
+	case CHIP_RV710:
+		num_ps_gprs = 192;
+		num_vs_gprs = 56;
+		num_temp_gprs = 4;
+		num_gs_gprs = 0;
+		num_es_gprs = 0;
+		num_ps_threads = 144;
+		num_vs_threads = 48;
+		num_gs_threads = 0;
+		num_es_threads = 0;
+		num_ps_stack_entries = 128;
+		num_vs_stack_entries = 128;
+		num_gs_stack_entries = 0;
+		num_es_stack_entries = 0;
+		break;
+	}
+
+	if ((rdev->family == CHIP_RV610) ||
+	    (rdev->family == CHIP_RV620) ||
+	    (rdev->family == CHIP_RS780) ||
+	    (rdev->family == CHIP_RS880) ||
+	    (rdev->family == CHIP_RV710))
+		sq_config = 0;
+	else
+		sq_config = VC_ENABLE;
+
+	sq_config |= (DX9_CONSTS |
+		      ALU_INST_PREFER_VECTOR |
+		      PS_PRIO(0) |
+		      VS_PRIO(1) |
+		      GS_PRIO(2) |
+		      ES_PRIO(3));
+
+	sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(num_ps_gprs) |
+				  NUM_VS_GPRS(num_vs_gprs) |
+				  NUM_CLAUSE_TEMP_GPRS(num_temp_gprs));
+	sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(num_gs_gprs) |
+				  NUM_ES_GPRS(num_es_gprs));
+	sq_thread_resource_mgmt = (NUM_PS_THREADS(num_ps_threads) |
+				   NUM_VS_THREADS(num_vs_threads) |
+				   NUM_GS_THREADS(num_gs_threads) |
+				   NUM_ES_THREADS(num_es_threads));
+	sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(num_ps_stack_entries) |
+				    NUM_VS_STACK_ENTRIES(num_vs_stack_entries));
+	sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(num_gs_stack_entries) |
+				    NUM_ES_STACK_ENTRIES(num_es_stack_entries));
+
+	/* emit an IB pointing at default state */
+	dwords = roundup2(rdev->r600_blit.state_len, 0x10);
+	gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset;
+	radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
+	radeon_ring_write(ring,
+#ifdef __BIG_ENDIAN
+			  (2 << 0) |
+#endif
+			  (gpu_addr & 0xFFFFFFFC));
+	radeon_ring_write(ring, upper_32_bits(gpu_addr) & 0xFF);
+	radeon_ring_write(ring, dwords);
+
+	/* SQ config */
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 6));
+	radeon_ring_write(ring, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
+	radeon_ring_write(ring, sq_config);
+	radeon_ring_write(ring, sq_gpr_resource_mgmt_1);
+	radeon_ring_write(ring, sq_gpr_resource_mgmt_2);
+	radeon_ring_write(ring, sq_thread_resource_mgmt);
+	radeon_ring_write(ring, sq_stack_resource_mgmt_1);
+	radeon_ring_write(ring, sq_stack_resource_mgmt_2);
+}
+
+int r600_blit_init(struct radeon_device *rdev)
+{
+	u32 obj_size;
+	int i, r, dwords;
+	void *ptr;
+	u32 packet2s[16];
+	int num_packet2s = 0;
+
+	rdev->r600_blit.primitives.set_render_target = set_render_target;
+	rdev->r600_blit.primitives.cp_set_surface_sync = cp_set_surface_sync;
+	rdev->r600_blit.primitives.set_shaders = set_shaders;
+	rdev->r600_blit.primitives.set_vtx_resource = set_vtx_resource;
+	rdev->r600_blit.primitives.set_tex_resource = set_tex_resource;
+	rdev->r600_blit.primitives.set_scissors = set_scissors;
+	rdev->r600_blit.primitives.draw_auto = draw_auto;
+	rdev->r600_blit.primitives.set_default_state = set_default_state;
+
+	rdev->r600_blit.ring_size_common = 8; /* sync semaphore */
+	rdev->r600_blit.ring_size_common += 40; /* shaders + def state */
+	rdev->r600_blit.ring_size_common += 5; /* done copy */
+	rdev->r600_blit.ring_size_common += 16; /* fence emit for done copy */
+
+	rdev->r600_blit.ring_size_per_loop = 76;
+	/* set_render_target emits 2 extra dwords on rv6xx */
+	if (rdev->family > CHIP_R600 && rdev->family < CHIP_RV770)
+		rdev->r600_blit.ring_size_per_loop += 2;
+
+	rdev->r600_blit.max_dim = 8192;
+
+	rdev->r600_blit.state_offset = 0;
+
+	if (rdev->family >= CHIP_RV770)
+		rdev->r600_blit.state_len = r7xx_default_size;
+	else
+		rdev->r600_blit.state_len = r6xx_default_size;
+
+	dwords = rdev->r600_blit.state_len;
+	while (dwords & 0xf) {
+		packet2s[num_packet2s++] = cpu_to_le32(PACKET2(0));
+		dwords++;
+	}
+
+	obj_size = dwords * 4;
+	obj_size = roundup2(obj_size, 256);
+
+	rdev->r600_blit.vs_offset = obj_size;
+	obj_size += r6xx_vs_size * 4;
+	obj_size = roundup2(obj_size, 256);
+
+	rdev->r600_blit.ps_offset = obj_size;
+	obj_size += r6xx_ps_size * 4;
+	obj_size = roundup2(obj_size, 256);
+
+	/* pin copy shader into vram if not already initialized */
+	if (rdev->r600_blit.shader_obj == NULL) {
+		r = radeon_bo_create(rdev, obj_size, PAGE_SIZE, true,
+				     RADEON_GEM_DOMAIN_VRAM,
+				     NULL, &rdev->r600_blit.shader_obj);
+		if (r) {
+			DRM_ERROR("r600 failed to allocate shader\n");
+			return r;
+		}
+
+		r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+		if (unlikely(r != 0))
+			return r;
+		r = radeon_bo_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM,
+				  &rdev->r600_blit.shader_gpu_addr);
+		radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+		if (r) {
+			dev_err(rdev->dev, "(%d) pin blit object failed\n", r);
+			return r;
+		}
+	}
+
+	DRM_DEBUG("r6xx blit allocated bo %08x vs %08x ps %08x\n",
+		  obj_size,
+		  rdev->r600_blit.vs_offset, rdev->r600_blit.ps_offset);
+
+	r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+	if (unlikely(r != 0))
+		return r;
+	r = radeon_bo_kmap(rdev->r600_blit.shader_obj, &ptr);
+	if (r) {
+		DRM_ERROR("failed to map blit object %d\n", r);
+		return r;
+	}
+	if (rdev->family >= CHIP_RV770)
+		memcpy_toio(ptr + rdev->r600_blit.state_offset,
+			    r7xx_default_state, rdev->r600_blit.state_len * 4);
+	else
+		memcpy_toio(ptr + rdev->r600_blit.state_offset,
+			    r6xx_default_state, rdev->r600_blit.state_len * 4);
+	if (num_packet2s)
+		memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4),
+			    packet2s, num_packet2s * 4);
+	for (i = 0; i < r6xx_vs_size; i++)
+		*(u32 *)((unsigned long)ptr + rdev->r600_blit.vs_offset + i * 4) = cpu_to_le32(r6xx_vs[i]);
+	for (i = 0; i < r6xx_ps_size; i++)
+		*(u32 *)((unsigned long)ptr + rdev->r600_blit.ps_offset + i * 4) = cpu_to_le32(r6xx_ps[i]);
+	radeon_bo_kunmap(rdev->r600_blit.shader_obj);
+	radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+
+	radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size);
+	return 0;
+}
+
+void r600_blit_fini(struct radeon_device *rdev)
+{
+	int r;
+
+	radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size);
+	if (rdev->r600_blit.shader_obj == NULL)
+		return;
+	/* If we can't reserve the bo, unref should be enough to destroy
+	 * it when it becomes idle.
+	 */
+	r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false);
+	if (!r) {
+		radeon_bo_unpin(rdev->r600_blit.shader_obj);
+		radeon_bo_unreserve(rdev->r600_blit.shader_obj);
+	}
+	radeon_bo_unref(&rdev->r600_blit.shader_obj);
+}
+
+static unsigned r600_blit_create_rect(unsigned num_gpu_pages,
+				      int *width, int *height, int max_dim)
+{
+	unsigned max_pages;
+	unsigned pages = num_gpu_pages;
+	int w, h;
+
+	if (num_gpu_pages == 0) {
+		/* not supposed to be called with no pages, but just in case */
+		h = 0;
+		w = 0;
+		pages = 0;
+		WARN_ON(1);
+	} else {
+		int rect_order = 2;
+		h = RECT_UNIT_H;
+		while (num_gpu_pages / rect_order) {
+			h *= 2;
+			rect_order *= 4;
+			if (h >= max_dim) {
+				h = max_dim;
+				break;
+			}
+		}
+		max_pages = (max_dim * h) / (RECT_UNIT_W * RECT_UNIT_H);
+		if (pages > max_pages)
+			pages = max_pages;
+		w = (pages * RECT_UNIT_W * RECT_UNIT_H) / h;
+		w = (w / RECT_UNIT_W) * RECT_UNIT_W;
+		pages = (w * h) / (RECT_UNIT_W * RECT_UNIT_H);
+		BUG_ON(pages == 0);
+	}
+
+
+	DRM_DEBUG("blit_rectangle: h=%d, w=%d, pages=%d\n", h, w, pages);
+
+	/* return width and height only of the caller wants it */
+	if (height)
+		*height = h;
+	if (width)
+		*width = w;
+
+	return pages;
+}
+
+
+int r600_blit_prepare_copy(struct radeon_device *rdev, unsigned num_gpu_pages,
+			   struct radeon_fence **fence, struct radeon_sa_bo **vb,
+			   struct radeon_semaphore **sem)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	int r;
+	int ring_size;
+	int num_loops = 0;
+	int dwords_per_loop = rdev->r600_blit.ring_size_per_loop;
+
+	/* num loops */
+	while (num_gpu_pages) {
+		num_gpu_pages -=
+			r600_blit_create_rect(num_gpu_pages, NULL, NULL,
+					      rdev->r600_blit.max_dim);
+		num_loops++;
+	}
+
+	/* 48 bytes for vertex per loop */
+	r = radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo, vb,
+			     (num_loops*48)+256, 256, true);
+	if (r) {
+		return r;
+	}
+
+	r = radeon_semaphore_create(rdev, sem);
+	if (r) {
+		radeon_sa_bo_free(rdev, vb, NULL);
+		return r;
+	}
+
+	/* calculate number of loops correctly */
+	ring_size = num_loops * dwords_per_loop;
+	ring_size += rdev->r600_blit.ring_size_common;
+	r = radeon_ring_lock(rdev, ring, ring_size);
+	if (r) {
+		radeon_sa_bo_free(rdev, vb, NULL);
+		radeon_semaphore_free(rdev, sem, NULL);
+		return r;
+	}
+
+	if (radeon_fence_need_sync(*fence, RADEON_RING_TYPE_GFX_INDEX)) {
+		radeon_semaphore_sync_rings(rdev, *sem, (*fence)->ring,
+					    RADEON_RING_TYPE_GFX_INDEX);
+		radeon_fence_note_sync(*fence, RADEON_RING_TYPE_GFX_INDEX);
+	} else {
+		radeon_semaphore_free(rdev, sem, NULL);
+	}
+
+	rdev->r600_blit.primitives.set_default_state(rdev);
+	rdev->r600_blit.primitives.set_shaders(rdev);
+	return 0;
+}
+
+void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence **fence,
+			 struct radeon_sa_bo *vb, struct radeon_semaphore *sem)
+{
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+	int r;
+
+	r = radeon_fence_emit(rdev, fence, RADEON_RING_TYPE_GFX_INDEX);
+	if (r) {
+		radeon_ring_unlock_undo(rdev, ring);
+		return;
+	}
+
+	radeon_ring_unlock_commit(rdev, ring);
+	radeon_sa_bo_free(rdev, &vb, *fence);
+	radeon_semaphore_free(rdev, &sem, *fence);
+}
+
+void r600_kms_blit_copy(struct radeon_device *rdev,
+			u64 src_gpu_addr, u64 dst_gpu_addr,
+			unsigned num_gpu_pages,
+			struct radeon_sa_bo *vb)
+{
+	u64 vb_gpu_addr;
+	u32 *vb_cpu_addr;
+
+	DRM_DEBUG("emitting copy %16llx %16llx %d\n",
+		  src_gpu_addr, dst_gpu_addr, num_gpu_pages);
+	vb_cpu_addr = (u32 *)radeon_sa_bo_cpu_addr(vb);
+	vb_gpu_addr = radeon_sa_bo_gpu_addr(vb);
+
+	while (num_gpu_pages) {
+		int w, h;
+		unsigned size_in_bytes;
+		unsigned pages_per_loop =
+			r600_blit_create_rect(num_gpu_pages, &w, &h,
+					      rdev->r600_blit.max_dim);
+
+		size_in_bytes = pages_per_loop * RADEON_GPU_PAGE_SIZE;
+		DRM_DEBUG("rectangle w=%d h=%d\n", w, h);
+
+		vb_cpu_addr[0] = 0;
+		vb_cpu_addr[1] = 0;
+		vb_cpu_addr[2] = 0;
+		vb_cpu_addr[3] = 0;
+
+		vb_cpu_addr[4] = 0;
+		vb_cpu_addr[5] = int2float(h);
+		vb_cpu_addr[6] = 0;
+		vb_cpu_addr[7] = int2float(h);
+
+		vb_cpu_addr[8] = int2float(w);
+		vb_cpu_addr[9] = int2float(h);
+		vb_cpu_addr[10] = int2float(w);
+		vb_cpu_addr[11] = int2float(h);
+
+		rdev->r600_blit.primitives.set_tex_resource(rdev, FMT_8_8_8_8,
+							    w, h, w, src_gpu_addr, size_in_bytes);
+		rdev->r600_blit.primitives.set_render_target(rdev, COLOR_8_8_8_8,
+							     w, h, dst_gpu_addr);
+		rdev->r600_blit.primitives.set_scissors(rdev, 0, 0, w, h);
+		rdev->r600_blit.primitives.set_vtx_resource(rdev, vb_gpu_addr);
+		rdev->r600_blit.primitives.draw_auto(rdev);
+		rdev->r600_blit.primitives.cp_set_surface_sync(rdev,
+				    PACKET3_CB_ACTION_ENA | PACKET3_CB0_DEST_BASE_ENA,
+				    size_in_bytes, dst_gpu_addr);
+
+		vb_cpu_addr += 12;
+		vb_gpu_addr += 4*12;
+		src_gpu_addr += size_in_bytes;
+		dst_gpu_addr += size_in_bytes;
+		num_gpu_pages -= pages_per_loop;
+	}
+}
+
+/* 23 bits of float fractional data */
+#define I2F_FRAC_BITS  23
+#define I2F_MASK ((1 << I2F_FRAC_BITS) - 1)
+
+/*
+ * Converts unsigned integer into 32-bit IEEE floating point representation.
+ * Will be exact from 0 to 2^24.  Above that, we round towards zero
+ * as the fractional bits will not fit in a float.  (It would be better to
+ * round towards even as the fpu does, but that is slower.)
+ */
+__pure uint32_t int2float(uint32_t x)
+{
+	uint32_t msb, exponent, fraction;
+
+	/* Zero is special */
+	if (!x) return 0;
+
+	/* Get location of the most significant bit */
+	msb = fls(x);
+
+	/*
+	 * Use a rotate instead of a shift because that works both leftwards
+	 * and rightwards due to the mod(32) behaviour.  This means we don't
+	 * need to check to see if we are above 2^24 or not.
+	 */
+	fraction = ror32(x, (msb - I2F_FRAC_BITS) & 0x1f) & I2F_MASK;
+	exponent = (127 + msb) << I2F_FRAC_BITS;
+
+	return fraction + exponent;
+}
diff --git a/sys/dev/pci/drm/radeon/radeon_blit_common.h b/sys/dev/pci/drm/radeon/radeon_blit_common.h
index e69de29bb2d..3a18d4b22ee 100644
--- a/sys/dev/pci/drm/radeon/radeon_blit_common.h
+++ b/sys/dev/pci/drm/radeon/radeon_blit_common.h
@@ -0,0 +1,45 @@
+/*	$OpenBSD: radeon_blit_common.h,v 1.3 2018/04/20 21:12:50 naddy Exp $	*/
+/*
+ * Copyright 2009 Advanced Micro Devices, Inc.
+ * Copyright 2009 Red Hat Inc.
+ * Copyright 2012 Alcatel-Lucent, 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 (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS 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.
+ *
+ */
+
+#ifndef __RADEON_BLIT_COMMON_H__
+
+#define DI_PT_RECTLIST        0x11
+#define DI_INDEX_SIZE_16_BIT  0x0
+#define DI_SRC_SEL_AUTO_INDEX 0x2
+
+#define FMT_8                 0x1
+#define FMT_5_6_5             0x8
+#define FMT_8_8_8_8           0x1a
+#define COLOR_8               0x1
+#define COLOR_5_6_5           0x8
+#define COLOR_8_8_8_8         0x1a
+
+#define RECT_UNIT_H           32
+#define RECT_UNIT_W           (RADEON_GPU_PAGE_SIZE / 4 / RECT_UNIT_H)
+
+#define __RADEON_BLIT_COMMON_H__
+#endif
diff --git a/sys/dev/pci/drm/ttm/ttm_lock.c b/sys/dev/pci/drm/ttm/ttm_lock.c
index e69de29bb2d..5eafda1ba49 100644
--- a/sys/dev/pci/drm/ttm/ttm_lock.c
+++ b/sys/dev/pci/drm/ttm/ttm_lock.c
@@ -0,0 +1,342 @@
+/*	$OpenBSD: ttm_lock.c,v 1.7 2018/04/20 21:12:50 naddy Exp $	*/
+/**************************************************************************
+ *
+ * Copyright (c) 2007-2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * 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, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#include <dev/pci/drm/drmP.h>
+#include <dev/pci/drm/ttm/ttm_lock.h>
+#include <dev/pci/drm/ttm/ttm_module.h>
+
+#define TTM_WRITE_LOCK_PENDING    (1 << 0)
+#define TTM_VT_LOCK_PENDING       (1 << 1)
+#define TTM_SUSPEND_LOCK_PENDING  (1 << 2)
+#define TTM_VT_LOCK               (1 << 3)
+#define TTM_SUSPEND_LOCK          (1 << 4)
+
+void	 ttm_write_lock_downgrade(struct ttm_lock *);
+
+void ttm_lock_init(struct ttm_lock *lock)
+{
+	mtx_init(&lock->lock, IPL_NONE);
+	init_waitqueue_head(&lock->queue);
+	lock->rw = 0;
+	lock->flags = 0;
+	lock->kill_takers = false;
+	lock->signal = SIGKILL;
+}
+EXPORT_SYMBOL(ttm_lock_init);
+
+void ttm_read_unlock(struct ttm_lock *lock)
+{
+	spin_lock(&lock->lock);
+	if (--lock->rw == 0)
+		wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+}
+EXPORT_SYMBOL(ttm_read_unlock);
+
+#ifdef notyet
+static bool __ttm_read_lock(struct ttm_lock *lock)
+{
+	bool locked = false;
+
+	spin_lock(&lock->lock);
+	if (unlikely(lock->kill_takers)) {
+		send_sig(lock->signal, current, 0);
+		spin_unlock(&lock->lock);
+		return false;
+	}
+	if (lock->rw >= 0 && lock->flags == 0) {
+		++lock->rw;
+		locked = true;
+	}
+	spin_unlock(&lock->lock);
+	return locked;
+}
+#endif
+
+int ttm_read_lock(struct ttm_lock *lock, bool interruptible)
+{
+	printf("%s stub\n", __func__);
+	return -ENOSYS;
+#ifdef notyet
+	int ret = 0;
+
+	if (interruptible)
+		ret = wait_event_interruptible(lock->queue,
+					       __ttm_read_lock(lock));
+	else
+		wait_event(lock->queue, __ttm_read_lock(lock));
+	return ret;
+#endif
+}
+EXPORT_SYMBOL(ttm_read_lock);
+
+#ifdef notyet
+static bool __ttm_read_trylock(struct ttm_lock *lock, bool *locked)
+{
+	bool block = true;
+
+	*locked = false;
+
+	spin_lock(&lock->lock);
+	if (unlikely(lock->kill_takers)) {
+		send_sig(lock->signal, current, 0);
+		spin_unlock(&lock->lock);
+		return false;
+	}
+	if (lock->rw >= 0 && lock->flags == 0) {
+		++lock->rw;
+		block = false;
+		*locked = true;
+	} else if (lock->flags == 0) {
+		block = false;
+	}
+	spin_unlock(&lock->lock);
+
+	return !block;
+}
+#endif
+
+int ttm_read_trylock(struct ttm_lock *lock, bool interruptible)
+{
+	printf("%s stub\n", __func__);
+	return -ENOSYS;
+#ifdef notyet
+	int ret = 0;
+	bool locked;
+
+	if (interruptible)
+		ret = wait_event_interruptible
+			(lock->queue, __ttm_read_trylock(lock, &locked));
+	else
+		wait_event(lock->queue, __ttm_read_trylock(lock, &locked));
+
+	if (unlikely(ret != 0)) {
+		BUG_ON(locked);
+		return ret;
+	}
+
+	return (locked) ? 0 : -EBUSY;
+#endif
+}
+
+void ttm_write_unlock(struct ttm_lock *lock)
+{
+	spin_lock(&lock->lock);
+	lock->rw = 0;
+	wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+}
+EXPORT_SYMBOL(ttm_write_unlock);
+
+#ifdef notyet
+static bool __ttm_write_lock(struct ttm_lock *lock)
+{
+	bool locked = false;
+
+	spin_lock(&lock->lock);
+	if (unlikely(lock->kill_takers)) {
+		send_sig(lock->signal, current, 0);
+		spin_unlock(&lock->lock);
+		return false;
+	}
+	if (lock->rw == 0 && ((lock->flags & ~TTM_WRITE_LOCK_PENDING) == 0)) {
+		lock->rw = -1;
+		lock->flags &= ~TTM_WRITE_LOCK_PENDING;
+		locked = true;
+	} else {
+		lock->flags |= TTM_WRITE_LOCK_PENDING;
+	}
+	spin_unlock(&lock->lock);
+	return locked;
+}
+#endif
+
+int ttm_write_lock(struct ttm_lock *lock, bool interruptible)
+{
+	printf("%s stub\n", __func__);
+	return -ENOSYS;
+#ifdef notyet
+	int ret = 0;
+
+	if (interruptible) {
+		ret = wait_event_interruptible(lock->queue,
+					       __ttm_write_lock(lock));
+		if (unlikely(ret != 0)) {
+			spin_lock(&lock->lock);
+			lock->flags &= ~TTM_WRITE_LOCK_PENDING;
+			wake_up_all(&lock->queue);
+			spin_unlock(&lock->lock);
+		}
+	} else
+		wait_event(lock->queue, __ttm_read_lock(lock));
+
+	return ret;
+#endif
+}
+EXPORT_SYMBOL(ttm_write_lock);
+
+void ttm_write_lock_downgrade(struct ttm_lock *lock)
+{
+	spin_lock(&lock->lock);
+	lock->rw = 1;
+	wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+}
+
+#ifdef notyet
+static int __ttm_vt_unlock(struct ttm_lock *lock)
+{
+	int ret = 0;
+
+	spin_lock(&lock->lock);
+	if (unlikely(!(lock->flags & TTM_VT_LOCK)))
+		ret = -EINVAL;
+	lock->flags &= ~TTM_VT_LOCK;
+	wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+
+	return ret;
+}
+#endif
+
+#ifdef notyet
+static void ttm_vt_lock_remove(struct ttm_base_object **p_base)
+{
+	struct ttm_base_object *base = *p_base;
+	struct ttm_lock *lock = container_of(base, struct ttm_lock, base);
+	int ret;
+
+	*p_base = NULL;
+	ret = __ttm_vt_unlock(lock);
+	BUG_ON(ret != 0);
+}
+#endif
+
+#ifdef notyet
+static bool __ttm_vt_lock(struct ttm_lock *lock)
+{
+	bool locked = false;
+
+	spin_lock(&lock->lock);
+	if (lock->rw == 0) {
+		lock->flags &= ~TTM_VT_LOCK_PENDING;
+		lock->flags |= TTM_VT_LOCK;
+		locked = true;
+	} else {
+		lock->flags |= TTM_VT_LOCK_PENDING;
+	}
+	spin_unlock(&lock->lock);
+	return locked;
+}
+#endif
+
+int ttm_vt_lock(struct ttm_lock *lock,
+		bool interruptible,
+		struct ttm_object_file *tfile)
+{
+	printf("%s stub\n", __func__);
+	return -ENOSYS;
+#ifdef notyet
+	int ret = 0;
+
+	if (interruptible) {
+		ret = wait_event_interruptible(lock->queue,
+					       __ttm_vt_lock(lock));
+		if (unlikely(ret != 0)) {
+			spin_lock(&lock->lock);
+			lock->flags &= ~TTM_VT_LOCK_PENDING;
+			wake_up_all(&lock->queue);
+			spin_unlock(&lock->lock);
+			return ret;
+		}
+	} else
+		wait_event(lock->queue, __ttm_vt_lock(lock));
+
+	/*
+	 * Add a base-object, the destructor of which will
+	 * make sure the lock is released if the client dies
+	 * while holding it.
+	 */
+
+	ret = ttm_base_object_init(tfile, &lock->base, false,
+				   ttm_lock_type, &ttm_vt_lock_remove, NULL);
+	if (ret)
+		(void)__ttm_vt_unlock(lock);
+	else
+		lock->vt_holder = tfile;
+
+	return ret;
+#endif
+}
+EXPORT_SYMBOL(ttm_vt_lock);
+
+int ttm_vt_unlock(struct ttm_lock *lock)
+{
+	return ttm_ref_object_base_unref(lock->vt_holder,
+					 lock->base.hash.key, TTM_REF_USAGE);
+}
+EXPORT_SYMBOL(ttm_vt_unlock);
+
+void ttm_suspend_unlock(struct ttm_lock *lock)
+{
+	spin_lock(&lock->lock);
+	lock->flags &= ~TTM_SUSPEND_LOCK;
+	wake_up_all(&lock->queue);
+	spin_unlock(&lock->lock);
+}
+EXPORT_SYMBOL(ttm_suspend_unlock);
+
+#ifdef notyet
+static bool __ttm_suspend_lock(struct ttm_lock *lock)
+{
+	bool locked = false;
+
+	spin_lock(&lock->lock);
+	if (lock->rw == 0) {
+		lock->flags &= ~TTM_SUSPEND_LOCK_PENDING;
+		lock->flags |= TTM_SUSPEND_LOCK;
+		locked = true;
+	} else {
+		lock->flags |= TTM_SUSPEND_LOCK_PENDING;
+	}
+	spin_unlock(&lock->lock);
+	return locked;
+}
+#endif
+
+void ttm_suspend_lock(struct ttm_lock *lock)
+{
+	printf("%s stub\n", __func__);
+#ifdef notyet
+	wait_event(lock->queue, __ttm_suspend_lock(lock));
+#endif
+}
+EXPORT_SYMBOL(ttm_suspend_lock);
diff --git a/sys/dev/pci/drm/ttm/ttm_lock.h b/sys/dev/pci/drm/ttm/ttm_lock.h
index e69de29bb2d..e2ae0deab0b 100644
--- a/sys/dev/pci/drm/ttm/ttm_lock.h
+++ b/sys/dev/pci/drm/ttm/ttm_lock.h
@@ -0,0 +1,247 @@
+/*	$OpenBSD: ttm_lock.h,v 1.5 2018/04/20 21:12:50 naddy Exp $	*/
+/**************************************************************************
+ *
+ * Copyright (c) 2007-2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * 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, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+/** @file ttm_lock.h
+ * This file implements a simple replacement for the buffer manager use
+ * of the DRM heavyweight hardware lock.
+ * The lock is a read-write lock. Taking it in read mode and write mode
+ * is relatively fast, and intended for in-kernel use only.
+ *
+ * The vt mode is used only when there is a need to block all
+ * user-space processes from validating buffers.
+ * It's allowed to leave kernel space with the vt lock held.
+ * If a user-space process dies while having the vt-lock,
+ * it will be released during the file descriptor release. The vt lock
+ * excludes write lock and read lock.
+ *
+ * The suspend mode is used to lock out all TTM users when preparing for
+ * and executing suspend operations.
+ *
+ */
+
+#ifndef _TTM_LOCK_H_
+#define _TTM_LOCK_H_
+
+#include <sys/mutex.h>
+#include <dev/pci/drm/ttm/ttm_object.h>
+
+/**
+ * struct ttm_lock
+ *
+ * @base: ttm base object used solely to release the lock if the client
+ * holding the lock dies.
+ * @queue: Queue for processes waiting for lock change-of-status.
+ * @lock: Spinlock protecting some lock members.
+ * @rw: Read-write lock counter. Protected by @lock.
+ * @flags: Lock state. Protected by @lock.
+ * @kill_takers: Boolean whether to kill takers of the lock.
+ * @signal: Signal to send when kill_takers is true.
+ */
+
+struct ttm_lock {
+	struct ttm_base_object base;
+	wait_queue_head_t queue;
+	spinlock_t lock;
+	int32_t rw;
+	uint32_t flags;
+	bool kill_takers;
+	int signal;
+	struct ttm_object_file *vt_holder;
+};
+
+
+/**
+ * ttm_lock_init
+ *
+ * @lock: Pointer to a struct ttm_lock
+ * Initializes the lock.
+ */
+extern void ttm_lock_init(struct ttm_lock *lock);
+
+/**
+ * ttm_read_unlock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ *
+ * Releases a read lock.
+ */
+extern void ttm_read_unlock(struct ttm_lock *lock);
+
+/**
+ * ttm_read_lock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ * @interruptible: Interruptible sleeping while waiting for a lock.
+ *
+ * Takes the lock in read mode.
+ * Returns:
+ * -ERESTARTSYS If interrupted by a signal and interruptible is true.
+ */
+extern int ttm_read_lock(struct ttm_lock *lock, bool interruptible);
+
+/**
+ * ttm_read_trylock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ * @interruptible: Interruptible sleeping while waiting for a lock.
+ *
+ * Tries to take the lock in read mode. If the lock is already held
+ * in write mode, the function will return -EBUSY. If the lock is held
+ * in vt or suspend mode, the function will sleep until these modes
+ * are unlocked.
+ *
+ * Returns:
+ * -EBUSY The lock was already held in write mode.
+ * -ERESTARTSYS If interrupted by a signal and interruptible is true.
+ */
+extern int ttm_read_trylock(struct ttm_lock *lock, bool interruptible);
+
+/**
+ * ttm_write_unlock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ *
+ * Releases a write lock.
+ */
+extern void ttm_write_unlock(struct ttm_lock *lock);
+
+/**
+ * ttm_write_lock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ * @interruptible: Interruptible sleeping while waiting for a lock.
+ *
+ * Takes the lock in write mode.
+ * Returns:
+ * -ERESTARTSYS If interrupted by a signal and interruptible is true.
+ */
+extern int ttm_write_lock(struct ttm_lock *lock, bool interruptible);
+
+/**
+ * ttm_lock_downgrade
+ *
+ * @lock: Pointer to a struct ttm_lock
+ *
+ * Downgrades a write lock to a read lock.
+ */
+extern void ttm_lock_downgrade(struct ttm_lock *lock);
+
+/**
+ * ttm_suspend_lock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ *
+ * Takes the lock in suspend mode. Excludes read and write mode.
+ */
+extern void ttm_suspend_lock(struct ttm_lock *lock);
+
+/**
+ * ttm_suspend_unlock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ *
+ * Releases a suspend lock
+ */
+extern void ttm_suspend_unlock(struct ttm_lock *lock);
+
+/**
+ * ttm_vt_lock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ * @interruptible: Interruptible sleeping while waiting for a lock.
+ * @tfile: Pointer to a struct ttm_object_file to register the lock with.
+ *
+ * Takes the lock in vt mode.
+ * Returns:
+ * -ERESTARTSYS If interrupted by a signal and interruptible is true.
+ * -ENOMEM: Out of memory when locking.
+ */
+extern int ttm_vt_lock(struct ttm_lock *lock, bool interruptible,
+		       struct ttm_object_file *tfile);
+
+/**
+ * ttm_vt_unlock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ *
+ * Releases a vt lock.
+ * Returns:
+ * -EINVAL If the lock was not held.
+ */
+extern int ttm_vt_unlock(struct ttm_lock *lock);
+
+/**
+ * ttm_write_unlock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ *
+ * Releases a write lock.
+ */
+extern void ttm_write_unlock(struct ttm_lock *lock);
+
+/**
+ * ttm_write_lock
+ *
+ * @lock: Pointer to a struct ttm_lock
+ * @interruptible: Interruptible sleeping while waiting for a lock.
+ *
+ * Takes the lock in write mode.
+ * Returns:
+ * -ERESTARTSYS If interrupted by a signal and interruptible is true.
+ */
+extern int ttm_write_lock(struct ttm_lock *lock, bool interruptible);
+
+/**
+ * ttm_lock_set_kill
+ *
+ * @lock: Pointer to a struct ttm_lock
+ * @val: Boolean whether to kill processes taking the lock.
+ * @signal: Signal to send to the process taking the lock.
+ *
+ * The kill-when-taking-lock functionality is used to kill processes that keep
+ * on using the TTM functionality when its resources has been taken down, for
+ * example when the X server exits. A typical sequence would look like this:
+ * - X server takes lock in write mode.
+ * - ttm_lock_set_kill() is called with @val set to true.
+ * - As part of X server exit, TTM resources are taken down.
+ * - X server releases the lock on file release.
+ * - Another dri client wants to render, takes the lock and is killed.
+ *
+ */
+static inline void ttm_lock_set_kill(struct ttm_lock *lock, bool val,
+				     int signal)
+{
+	lock->kill_takers = val;
+	if (val)
+		lock->signal = signal;
+}
+
+#endif
diff --git a/sys/dev/pci/drm/ttm/ttm_module.c b/sys/dev/pci/drm/ttm/ttm_module.c
index e69de29bb2d..cd9037fecab 100644
--- a/sys/dev/pci/drm/ttm/ttm_module.c
+++ b/sys/dev/pci/drm/ttm/ttm_module.c
@@ -0,0 +1,100 @@
+/*	$OpenBSD: ttm_module.c,v 1.3 2018/04/20 21:12:50 naddy Exp $	*/
+/**************************************************************************
+ *
+ * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * 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, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ * 	    Jerome Glisse
+ */
+#include <dev/pci/drm/ttm/ttm_module.h>
+#include <dev/pci/drm/drm_sysfs.h>
+
+static DECLARE_WAIT_QUEUE_HEAD(exit_q);
+atomic_t device_released;
+
+static struct device_type ttm_drm_class_type = {
+	.name = "ttm",
+	/**
+	 * Add pm ops here.
+	 */
+};
+
+static void ttm_drm_class_device_release(struct device *dev)
+{
+	atomic_set(&device_released, 1);
+	wake_up_all(&exit_q);
+}
+
+static struct device ttm_drm_class_device = {
+	.type = &ttm_drm_class_type,
+	.release = &ttm_drm_class_device_release
+};
+
+struct kobject *ttm_get_kobj(void)
+{
+	struct kobject *kobj = &ttm_drm_class_device.kobj;
+	BUG_ON(kobj == NULL);
+	return kobj;
+}
+
+static int __init ttm_init(void)
+{
+	int ret;
+
+	ret = dev_set_name(&ttm_drm_class_device, "ttm");
+	if (unlikely(ret != 0))
+		return ret;
+
+	atomic_set(&device_released, 0);
+	ret = drm_class_device_register(&ttm_drm_class_device);
+	if (unlikely(ret != 0))
+		goto out_no_dev_reg;
+
+	return 0;
+out_no_dev_reg:
+	atomic_set(&device_released, 1);
+	wake_up_all(&exit_q);
+	return ret;
+}
+
+static void __exit ttm_exit(void)
+{
+	drm_class_device_unregister(&ttm_drm_class_device);
+
+	/**
+	 * Refuse to unload until the TTM device is released.
+	 * Not sure this is 100% needed.
+	 */
+
+	wait_event(exit_q, atomic_read(&device_released) == 1);
+}
+
+module_init(ttm_init);
+module_exit(ttm_exit);
+
+MODULE_AUTHOR("Thomas Hellstrom, Jerome Glisse");
+MODULE_DESCRIPTION("TTM memory manager subsystem (for DRM device)");
+MODULE_LICENSE("GPL and additional rights");
diff --git a/sys/dev/pci/drm/ttm/ttm_object.c b/sys/dev/pci/drm/ttm/ttm_object.c
index e69de29bb2d..d73f641988d 100644
--- a/sys/dev/pci/drm/ttm/ttm_object.c
+++ b/sys/dev/pci/drm/ttm/ttm_object.c
@@ -0,0 +1,451 @@
+/*	$OpenBSD: ttm_object.c,v 1.11 2018/04/20 21:12:50 naddy Exp $	*/
+/**************************************************************************
+ *
+ * Copyright (c) 2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * 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, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+/** @file ttm_ref_object.c
+ *
+ * Base- and reference object implementation for the various
+ * ttm objects. Implements reference counting, minimal security checks
+ * and release on file close.
+ */
+
+/**
+ * struct ttm_object_file
+ *
+ * @tdev: Pointer to the ttm_object_device.
+ *
+ * @lock: Lock that protects the ref_list list and the
+ * ref_hash hash tables.
+ *
+ * @ref_list: List of ttm_ref_objects to be destroyed at
+ * file release.
+ *
+ * @ref_hash: Hash tables of ref objects, one per ttm_ref_type,
+ * for fast lookup of ref objects given a base object.
+ */
+
+#define pr_fmt(fmt) "[TTM] " fmt
+
+#include <dev/pci/drm/drmP.h>
+#include <dev/pci/drm/ttm/ttm_object.h>
+#include <dev/pci/drm/ttm/ttm_module.h>
+
+struct ttm_object_file {
+	struct ttm_object_device *tdev;
+	rwlock_t lock;
+	struct list_head ref_list;
+	struct drm_open_hash ref_hash[TTM_REF_NUM];
+	struct kref refcount;
+};
+
+/**
+ * struct ttm_object_device
+ *
+ * @object_lock: lock that protects the object_hash hash table.
+ *
+ * @object_hash: hash table for fast lookup of object global names.
+ *
+ * @object_count: Per device object count.
+ *
+ * This is the per-device data structure needed for ttm object management.
+ */
+
+struct ttm_object_device {
+	spinlock_t object_lock;
+	struct drm_open_hash object_hash;
+	atomic_t object_count;
+	struct ttm_mem_global *mem_glob;
+};
+
+/**
+ * struct ttm_ref_object
+ *
+ * @hash: Hash entry for the per-file object reference hash.
+ *
+ * @head: List entry for the per-file list of ref-objects.
+ *
+ * @kref: Ref count.
+ *
+ * @obj: Base object this ref object is referencing.
+ *
+ * @ref_type: Type of ref object.
+ *
+ * This is similar to an idr object, but it also has a hash table entry
+ * that allows lookup with a pointer to the referenced object as a key. In
+ * that way, one can easily detect whether a base object is referenced by
+ * a particular ttm_object_file. It also carries a ref count to avoid creating
+ * multiple ref objects if a ttm_object_file references the same base
+ * object more than once.
+ */
+
+struct ttm_ref_object {
+	struct drm_hash_item hash;
+	struct list_head head;
+	struct kref kref;
+	enum ttm_ref_type ref_type;
+	struct ttm_base_object *obj;
+	struct ttm_object_file *tfile;
+};
+
+static inline struct ttm_object_file *
+ttm_object_file_ref(struct ttm_object_file *tfile)
+{
+	kref_get(&tfile->refcount);
+	return tfile;
+}
+
+static void ttm_object_file_destroy(struct kref *kref)
+{
+	struct ttm_object_file *tfile =
+		container_of(kref, struct ttm_object_file, refcount);
+
+	kfree(tfile);
+}
+
+
+static inline void ttm_object_file_unref(struct ttm_object_file **p_tfile)
+{
+	struct ttm_object_file *tfile = *p_tfile;
+
+	*p_tfile = NULL;
+	kref_put(&tfile->refcount, ttm_object_file_destroy);
+}
+
+
+int ttm_base_object_init(struct ttm_object_file *tfile,
+			 struct ttm_base_object *base,
+			 bool shareable,
+			 enum ttm_object_type object_type,
+			 void (*refcount_release) (struct ttm_base_object **),
+			 void (*ref_obj_release) (struct ttm_base_object *,
+						  enum ttm_ref_type ref_type))
+{
+	struct ttm_object_device *tdev = tfile->tdev;
+	int ret;
+
+	base->shareable = shareable;
+	base->tfile = ttm_object_file_ref(tfile);
+	base->refcount_release = refcount_release;
+	base->ref_obj_release = ref_obj_release;
+	base->object_type = object_type;
+	kref_init(&base->refcount);
+	spin_lock(&tdev->object_lock);
+	ret = drm_ht_just_insert_please_rcu(&tdev->object_hash,
+					    &base->hash,
+					    (unsigned long)base, 31, 0, 0);
+	spin_unlock(&tdev->object_lock);
+	if (unlikely(ret != 0))
+		goto out_err0;
+
+	ret = ttm_ref_object_add(tfile, base, TTM_REF_USAGE, NULL);
+	if (unlikely(ret != 0))
+		goto out_err1;
+
+	ttm_base_object_unref(&base);
+
+	return 0;
+out_err1:
+	spin_lock(&tdev->object_lock);
+	(void)drm_ht_remove_item_rcu(&tdev->object_hash, &base->hash);
+	spin_unlock(&tdev->object_lock);
+out_err0:
+	return ret;
+}
+EXPORT_SYMBOL(ttm_base_object_init);
+
+static void ttm_release_base(struct kref *kref)
+{
+	struct ttm_base_object *base =
+	    container_of(kref, struct ttm_base_object, refcount);
+	struct ttm_object_device *tdev = base->tfile->tdev;
+
+	spin_lock(&tdev->object_lock);
+	(void)drm_ht_remove_item_rcu(&tdev->object_hash, &base->hash);
+	spin_unlock(&tdev->object_lock);
+
+	/*
+	 * Note: We don't use synchronize_rcu() here because it's far
+	 * too slow. It's up to the user to free the object using
+	 * call_rcu() or ttm_base_object_kfree().
+	 */
+
+	if (base->refcount_release) {
+		ttm_object_file_unref(&base->tfile);
+		base->refcount_release(&base);
+	}
+}
+
+void ttm_base_object_unref(struct ttm_base_object **p_base)
+{
+	struct ttm_base_object *base = *p_base;
+
+	*p_base = NULL;
+
+	kref_put(&base->refcount, ttm_release_base);
+}
+EXPORT_SYMBOL(ttm_base_object_unref);
+
+struct ttm_base_object *ttm_base_object_lookup(struct ttm_object_file *tfile,
+					       uint32_t key)
+{
+	struct ttm_object_device *tdev = tfile->tdev;
+	struct ttm_base_object *base;
+	struct drm_hash_item *hash;
+	int ret;
+
+	mtx_enter(&tdev->object_lock);
+	ret = drm_ht_find_item(&tdev->object_hash, key, &hash);
+
+	if (likely(ret == 0)) {
+		base = drm_hash_entry(hash, struct ttm_base_object, hash);
+		ret = kref_get_unless_zero(&base->refcount) ? 0 : -EINVAL;
+	}
+	mtx_leave(&tdev->object_lock);
+
+	if (unlikely(ret != 0))
+		return NULL;
+
+	if (tfile != base->tfile && !base->shareable) {
+		pr_err("Attempted access of non-shareable object\n");
+		ttm_base_object_unref(&base);
+		return NULL;
+	}
+
+	return base;
+}
+EXPORT_SYMBOL(ttm_base_object_lookup);
+
+int ttm_ref_object_add(struct ttm_object_file *tfile,
+		       struct ttm_base_object *base,
+		       enum ttm_ref_type ref_type, bool *existed)
+{
+	struct drm_open_hash *ht = &tfile->ref_hash[ref_type];
+	struct ttm_ref_object *ref;
+	struct drm_hash_item *hash;
+	struct ttm_mem_global *mem_glob = tfile->tdev->mem_glob;
+	int ret = -EINVAL;
+
+	if (existed != NULL)
+		*existed = true;
+
+	while (ret == -EINVAL) {
+		read_lock(&tfile->lock);
+		ret = drm_ht_find_item(ht, base->hash.key, &hash);
+
+		if (ret == 0) {
+			ref = drm_hash_entry(hash, struct ttm_ref_object, hash);
+			kref_get(&ref->kref);
+			read_unlock(&tfile->lock);
+			break;
+		}
+
+		read_unlock(&tfile->lock);
+		ret = ttm_mem_global_alloc(mem_glob, sizeof(*ref),
+					   false, false);
+		if (unlikely(ret != 0))
+			return ret;
+		ref = kmalloc(sizeof(*ref), GFP_KERNEL);
+		if (unlikely(ref == NULL)) {
+			ttm_mem_global_free(mem_glob, sizeof(*ref));
+			return -ENOMEM;
+		}
+
+		ref->hash.key = base->hash.key;
+		ref->obj = base;
+		ref->tfile = tfile;
+		ref->ref_type = ref_type;
+		kref_init(&ref->kref);
+
+		write_lock(&tfile->lock);
+		ret = drm_ht_insert_item(ht, &ref->hash);
+
+		if (likely(ret == 0)) {
+			list_add_tail(&ref->head, &tfile->ref_list);
+			kref_get(&base->refcount);
+			write_unlock(&tfile->lock);
+			if (existed != NULL)
+				*existed = false;
+			break;
+		}
+
+		write_unlock(&tfile->lock);
+		BUG_ON(ret != -EINVAL);
+
+		ttm_mem_global_free(mem_glob, sizeof(*ref));
+		kfree(ref);
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(ttm_ref_object_add);
+
+static void ttm_ref_object_release(struct kref *kref)
+{
+	struct ttm_ref_object *ref =
+	    container_of(kref, struct ttm_ref_object, kref);
+	struct ttm_base_object *base = ref->obj;
+	struct ttm_object_file *tfile = ref->tfile;
+	struct drm_open_hash *ht;
+	struct ttm_mem_global *mem_glob = tfile->tdev->mem_glob;
+
+	ht = &tfile->ref_hash[ref->ref_type];
+	(void)drm_ht_remove_item(ht, &ref->hash);
+	list_del(&ref->head);
+	write_unlock(&tfile->lock);
+
+	if (ref->ref_type != TTM_REF_USAGE && base->ref_obj_release)
+		base->ref_obj_release(base, ref->ref_type);
+
+	ttm_base_object_unref(&ref->obj);
+	ttm_mem_global_free(mem_glob, sizeof(*ref));
+	kfree(ref);
+	write_lock(&tfile->lock);
+}
+
+int ttm_ref_object_base_unref(struct ttm_object_file *tfile,
+			      unsigned long key, enum ttm_ref_type ref_type)
+{
+	struct drm_open_hash *ht = &tfile->ref_hash[ref_type];
+	struct ttm_ref_object *ref;
+	struct drm_hash_item *hash;
+	int ret;
+
+	write_lock(&tfile->lock);
+	ret = drm_ht_find_item(ht, key, &hash);
+	if (unlikely(ret != 0)) {
+		write_unlock(&tfile->lock);
+		return -EINVAL;
+	}
+	ref = drm_hash_entry(hash, struct ttm_ref_object, hash);
+	kref_put(&ref->kref, ttm_ref_object_release);
+	write_unlock(&tfile->lock);
+	return 0;
+}
+EXPORT_SYMBOL(ttm_ref_object_base_unref);
+
+void ttm_object_file_release(struct ttm_object_file **p_tfile)
+{
+	struct ttm_ref_object *ref;
+	struct list_head *list;
+	unsigned int i;
+	struct ttm_object_file *tfile = *p_tfile;
+
+	*p_tfile = NULL;
+	write_lock(&tfile->lock);
+
+	/*
+	 * Since we release the lock within the loop, we have to
+	 * restart it from the beginning each time.
+	 */
+
+	while (!list_empty(&tfile->ref_list)) {
+		list = tfile->ref_list.next;
+		ref = list_entry(list, struct ttm_ref_object, head);
+		ttm_ref_object_release(&ref->kref);
+	}
+
+	for (i = 0; i < TTM_REF_NUM; ++i)
+		drm_ht_remove(&tfile->ref_hash[i]);
+
+	write_unlock(&tfile->lock);
+	ttm_object_file_unref(&tfile);
+}
+EXPORT_SYMBOL(ttm_object_file_release);
+
+struct ttm_object_file *ttm_object_file_init(struct ttm_object_device *tdev,
+					     unsigned int hash_order)
+{
+	struct ttm_object_file *tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
+	unsigned int i;
+	unsigned int j = 0;
+	int ret;
+
+	if (unlikely(tfile == NULL))
+		return NULL;
+
+	rw_init(&tfile->lock, "ttmfl");
+	tfile->tdev = tdev;
+	kref_init(&tfile->refcount);
+	INIT_LIST_HEAD(&tfile->ref_list);
+
+	for (i = 0; i < TTM_REF_NUM; ++i) {
+		ret = drm_ht_create(&tfile->ref_hash[i], hash_order);
+		if (ret) {
+			j = i;
+			goto out_err;
+		}
+	}
+
+	return tfile;
+out_err:
+	for (i = 0; i < j; ++i)
+		drm_ht_remove(&tfile->ref_hash[i]);
+
+	kfree(tfile);
+
+	return NULL;
+}
+EXPORT_SYMBOL(ttm_object_file_init);
+
+struct ttm_object_device *ttm_object_device_init(struct ttm_mem_global
+						 *mem_glob,
+						 unsigned int hash_order)
+{
+	struct ttm_object_device *tdev = kmalloc(sizeof(*tdev), GFP_KERNEL);
+	int ret;
+
+	if (unlikely(tdev == NULL))
+		return NULL;
+
+	tdev->mem_glob = mem_glob;
+	mtx_init(&tdev->object_lock, IPL_NONE);
+	atomic_set(&tdev->object_count, 0);
+	ret = drm_ht_create(&tdev->object_hash, hash_order);
+
+	if (likely(ret == 0))
+		return tdev;
+
+	kfree(tdev);
+	return NULL;
+}
+EXPORT_SYMBOL(ttm_object_device_init);
+
+void ttm_object_device_release(struct ttm_object_device **p_tdev)
+{
+	struct ttm_object_device *tdev = *p_tdev;
+
+	*p_tdev = NULL;
+
+	spin_lock(&tdev->object_lock);
+	drm_ht_remove(&tdev->object_hash);
+	spin_unlock(&tdev->object_lock);
+
+	kfree(tdev);
+}
+EXPORT_SYMBOL(ttm_object_device_release);
diff --git a/sys/dev/pci/drm/ttm/ttm_object.h b/sys/dev/pci/drm/ttm/ttm_object.h
index e69de29bb2d..b59da0c78c5 100644
--- a/sys/dev/pci/drm/ttm/ttm_object.h
+++ b/sys/dev/pci/drm/ttm/ttm_object.h
@@ -0,0 +1,275 @@
+/*	$OpenBSD: ttm_object.h,v 1.5 2018/04/20 21:12:50 naddy Exp $	*/
+/**************************************************************************
+ *
+ * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA
+ * All Rights Reserved.
+ *
+ * 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, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+/** @file ttm_object.h
+ *
+ * Base- and reference object implementation for the various
+ * ttm objects. Implements reference counting, minimal security checks
+ * and release on file close.
+ */
+
+#ifndef _TTM_OBJECT_H_
+#define _TTM_OBJECT_H_
+
+#include <dev/pci/drm/drm_hashtab.h>
+#include <dev/pci/drm/ttm/ttm_memory.h>
+
+/**
+ * enum ttm_ref_type
+ *
+ * Describes what type of reference a ref object holds.
+ *
+ * TTM_REF_USAGE is a simple refcount on a base object.
+ *
+ * TTM_REF_SYNCCPU_READ is a SYNCCPU_READ reference on a
+ * buffer object.
+ *
+ * TTM_REF_SYNCCPU_WRITE is a SYNCCPU_WRITE reference on a
+ * buffer object.
+ *
+ */
+
+enum ttm_ref_type {
+	TTM_REF_USAGE,
+	TTM_REF_SYNCCPU_READ,
+	TTM_REF_SYNCCPU_WRITE,
+	TTM_REF_NUM
+};
+
+/**
+ * enum ttm_object_type
+ *
+ * One entry per ttm object type.
+ * Device-specific types should use the
+ * ttm_driver_typex types.
+ */
+
+enum ttm_object_type {
+	ttm_fence_type,
+	ttm_buffer_type,
+	ttm_lock_type,
+	ttm_driver_type0 = 256,
+	ttm_driver_type1,
+	ttm_driver_type2,
+	ttm_driver_type3,
+	ttm_driver_type4,
+	ttm_driver_type5
+};
+
+struct ttm_object_file;
+struct ttm_object_device;
+
+/**
+ * struct ttm_base_object
+ *
+ * @hash: hash entry for the per-device object hash.
+ * @type: derived type this object is base class for.
+ * @shareable: Other ttm_object_files can access this object.
+ *
+ * @tfile: Pointer to ttm_object_file of the creator.
+ * NULL if the object was not created by a user request.
+ * (kernel object).
+ *
+ * @refcount: Number of references to this object, not
+ * including the hash entry. A reference to a base object can
+ * only be held by a ref object.
+ *
+ * @refcount_release: A function to be called when there are
+ * no more references to this object. This function should
+ * destroy the object (or make sure destruction eventually happens),
+ * and when it is called, the object has
+ * already been taken out of the per-device hash. The parameter
+ * "base" should be set to NULL by the function.
+ *
+ * @ref_obj_release: A function to be called when a reference object
+ * with another ttm_ref_type than TTM_REF_USAGE is deleted.
+ * This function may, for example, release a lock held by a user-space
+ * process.
+ *
+ * This struct is intended to be used as a base struct for objects that
+ * are visible to user-space. It provides a global name, race-safe
+ * access and refcounting, minimal access contol and hooks for unref actions.
+ */
+
+struct ttm_base_object {
+#ifdef notyet
+	struct rcu_head rhead;
+#endif
+	struct drm_hash_item hash;
+	enum ttm_object_type object_type;
+	bool shareable;
+	struct ttm_object_file *tfile;
+	struct kref refcount;
+	void (*refcount_release) (struct ttm_base_object **base);
+	void (*ref_obj_release) (struct ttm_base_object *base,
+				 enum ttm_ref_type ref_type);
+};
+
+/**
+ * ttm_base_object_init
+ *
+ * @tfile: Pointer to a struct ttm_object_file.
+ * @base: The struct ttm_base_object to initialize.
+ * @shareable: This object is shareable with other applcations.
+ * (different @tfile pointers.)
+ * @type: The object type.
+ * @refcount_release: See the struct ttm_base_object description.
+ * @ref_obj_release: See the struct ttm_base_object description.
+ *
+ * Initializes a struct ttm_base_object.
+ */
+
+extern int ttm_base_object_init(struct ttm_object_file *tfile,
+				struct ttm_base_object *base,
+				bool shareable,
+				enum ttm_object_type type,
+				void (*refcount_release) (struct ttm_base_object
+							  **),
+				void (*ref_obj_release) (struct ttm_base_object
+							 *,
+							 enum ttm_ref_type
+							 ref_type));
+
+/**
+ * ttm_base_object_lookup
+ *
+ * @tfile: Pointer to a struct ttm_object_file.
+ * @key: Hash key
+ *
+ * Looks up a struct ttm_base_object with the key @key.
+ * Also verifies that the object is visible to the application, by
+ * comparing the @tfile argument and checking the object shareable flag.
+ */
+
+extern struct ttm_base_object *ttm_base_object_lookup(struct ttm_object_file
+						      *tfile, uint32_t key);
+
+/**
+ * ttm_base_object_unref
+ *
+ * @p_base: Pointer to a pointer referencing a struct ttm_base_object.
+ *
+ * Decrements the base object refcount and clears the pointer pointed to by
+ * p_base.
+ */
+
+extern void ttm_base_object_unref(struct ttm_base_object **p_base);
+
+/**
+ * ttm_ref_object_add.
+ *
+ * @tfile: A struct ttm_object_file representing the application owning the
+ * ref_object.
+ * @base: The base object to reference.
+ * @ref_type: The type of reference.
+ * @existed: Upon completion, indicates that an identical reference object
+ * already existed, and the refcount was upped on that object instead.
+ *
+ * Adding a ref object to a base object is basically like referencing the
+ * base object, but a user-space application holds the reference. When the
+ * file corresponding to @tfile is closed, all its reference objects are
+ * deleted. A reference object can have different types depending on what
+ * it's intended for. It can be refcounting to prevent object destruction,
+ * When user-space takes a lock, it can add a ref object to that lock to
+ * make sure the lock is released if the application dies. A ref object
+ * will hold a single reference on a base object.
+ */
+extern int ttm_ref_object_add(struct ttm_object_file *tfile,
+			      struct ttm_base_object *base,
+			      enum ttm_ref_type ref_type, bool *existed);
+/**
+ * ttm_ref_object_base_unref
+ *
+ * @key: Key representing the base object.
+ * @ref_type: Ref type of the ref object to be dereferenced.
+ *
+ * Unreference a ref object with type @ref_type
+ * on the base object identified by @key. If there are no duplicate
+ * references, the ref object will be destroyed and the base object
+ * will be unreferenced.
+ */
+extern int ttm_ref_object_base_unref(struct ttm_object_file *tfile,
+				     unsigned long key,
+				     enum ttm_ref_type ref_type);
+
+/**
+ * ttm_object_file_init - initialize a struct ttm_object file
+ *
+ * @tdev: A struct ttm_object device this file is initialized on.
+ * @hash_order: Order of the hash table used to hold the reference objects.
+ *
+ * This is typically called by the file_ops::open function.
+ */
+
+extern struct ttm_object_file *ttm_object_file_init(struct ttm_object_device
+						    *tdev,
+						    unsigned int hash_order);
+
+/**
+ * ttm_object_file_release - release data held by a ttm_object_file
+ *
+ * @p_tfile: Pointer to pointer to the ttm_object_file object to release.
+ * *p_tfile will be set to NULL by this function.
+ *
+ * Releases all data associated by a ttm_object_file.
+ * Typically called from file_ops::release. The caller must
+ * ensure that there are no concurrent users of tfile.
+ */
+
+extern void ttm_object_file_release(struct ttm_object_file **p_tfile);
+
+/**
+ * ttm_object device init - initialize a struct ttm_object_device
+ *
+ * @hash_order: Order of hash table used to hash the base objects.
+ *
+ * This function is typically called on device initialization to prepare
+ * data structures needed for ttm base and ref objects.
+ */
+
+extern struct ttm_object_device *ttm_object_device_init
+    (struct ttm_mem_global *mem_glob, unsigned int hash_order);
+
+/**
+ * ttm_object_device_release - release data held by a ttm_object_device
+ *
+ * @p_tdev: Pointer to pointer to the ttm_object_device object to release.
+ * *p_tdev will be set to NULL by this function.
+ *
+ * Releases all data associated by a ttm_object_device.
+ * Typically called from driver::unload before the destruction of the
+ * device private data structure.
+ */
+
+extern void ttm_object_device_release(struct ttm_object_device **p_tdev);
+
+#define ttm_base_object_kfree(__object, __base)\
+	kfree_rcu(__object, __base.rhead)
+#endif