Merge tag 'drm-for-v4.16' of git://people.freedesktop.org/~airlied/linux

Pull drm updates from Dave Airlie:
 "This seems to have been a comparatively quieter merge window, I assume
  due to holidays etc. The "biggest" change is AMD header cleanups, which
  merge/remove a bunch of them. The AMD gpu scheduler is now being made generic
  with the etnaviv driver wanting to reuse the code, hopefully other drivers
  can go in the same direction.

  Otherwise it's the usual lots of stuff in i915/amdgpu, not so much stuff
  elsewhere.

  Core:
   - Add .last_close and .output_poll_changed helpers to reduce driver footprints
   - Fix plane clipping
   - Improved debug printing support
   - Add panel orientation property
   - Update edid derived properties at edid setting
   - Reduction in fbdev driver footprint
   - Move amdgpu scheduler into core for other drivers to use.

  i915:
   - Selftest and IGT improvements
   - Fast boot prep work on IPS, pipe config
   - HW workarounds for Cannonlake, Geminilake
   - Cannonlake clock and HDMI2.0 fixes
   - GPU cache invalidation and context switch improvements
   - Display planes cleanup
   - New PMU interface for perf queries
   - New firmware support for KBL/SKL
   - Geminilake HW workaround for perforamce
   - Coffeelake stolen memory improvements
   - GPU reset robustness work
   - Cannonlake horizontal plane flipping
   - GVT work

  amdgpu/radeon:
   - RV and Vega header file cleanups (lots of lines gone!)
   - TTM operation context support
   - 48-bit GPUVM support for Vega/RV
   - ECC support for Vega
   - Resizeable BAR support
   - Multi-display sync support
   - Enable swapout for reserved BOs during allocation
   - S3 fixes on Raven
   - GPU reset cleanup and fixes
   - 2+1 level GPU page table

  amdkfd:
   - GFX7/8 SDMA user queues support
   - Hardware scheduling for multiple processes
   - dGPU prep work

  rcar:
   - Added R8A7743/5 support
   - System suspend/resume support

  sun4i:
   - Multi-plane support for YUV formats
   - A83T and LVDS support

  msm:
   - Devfreq support for GPU

  tegra:
   - Prep work for adding Tegra186 support
   - Tegra186 HDMI support
   - HDMI2.0 and zpos support by using generic helpers

  tilcdc:
   - Misc fixes

  omapdrm:
   - Support memory bandwidth limits
   - DSI command mode panel cleanups
   - DMM error handling

  exynos:
   - drop the old IPP subdriver.

  etnaviv:
   - Occlusion query fixes
   - Job handling fixes
   - Prep work for hooking in gpu scheduler

  armada:
   - Move closer to atomic modesetting
   - Allow disabling primary plane if overlay is full screen

  imx:
   - Format modifier support
   - Add tile prefetch to PRE
   - Runtime PM support for PRG

  ast:
   - fix LUT loading"

* tag 'drm-for-v4.16' of git://people.freedesktop.org/~airlied/linux: (1471 commits)
  drm/ast: Load lut in crtc_commit
  drm: Check for lessee in DROP_MASTER ioctl
  drm: fix gpu scheduler link order
  drm/amd/display: Demote error print to debug print when ATOM impl missing
  dma-buf: fix reservation_object_wait_timeout_rcu once more v2
  drm/amdgpu: Avoid leaking PM domain on driver unbind (v2)
  drm/amd/amdgpu: Add Polaris version check
  drm/amdgpu: Reenable manual GPU reset from sysfs
  drm/amdgpu: disable MMHUB power gating on raven
  drm/ttm: Don't unreserve swapped BOs that were previously reserved
  drm/ttm: Don't add swapped BOs to swap-LRU list
  drm/amdgpu: only check for ECC on Vega10
  drm/amd/powerplay: Fix smu_table_entry.handle type
  drm/ttm: add VADDR_FLAG_UPDATED_COUNT to correctly update dma_page global count
  drm: Fix PANEL_ORIENTATION_QUIRKS breaking the Kconfig DRM menuconfig
  drm/radeon: fill in rb backend map on evergreen/ni.
  drm/amdgpu/gfx9: fix ngg enablement to clear gds reserved memory (v2)
  drm/ttm: only free pages rather than update global memory count together
  drm/amdgpu: fix CPU based VM updates
  drm/amdgpu: fix typo in amdgpu_vce_validate_bo
  ...

36 files changed, 4380 insertions, 1987 deletions

diff --git a/drivers/gpu/drm/i915/gvt/Makefile b/drivers/gpu/drm/i915/gvt/Makefile
index 2641ba510a61..347116faa558 100644
--- a/drivers/gpu/drm/i915/gvt/Makefile
+++ b/drivers/gpu/drm/i915/gvt/Makefile
@@ -2,7 +2,8 @@
 GVT_DIR := gvt
 GVT_SOURCE := gvt.o aperture_gm.o handlers.o vgpu.o trace_points.o firmware.o \
 	interrupt.o gtt.o cfg_space.o opregion.o mmio.o display.o edid.o \
-	execlist.o scheduler.o sched_policy.o render.o cmd_parser.o
+	execlist.o scheduler.o sched_policy.o mmio_context.o cmd_parser.o debugfs.o \
+	fb_decoder.o dmabuf.o
 
 ccflags-y				+= -I$(src) -I$(src)/$(GVT_DIR)
 i915-y					+= $(addprefix $(GVT_DIR)/, $(GVT_SOURCE))
diff --git a/drivers/gpu/drm/i915/gvt/cfg_space.c b/drivers/gpu/drm/i915/gvt/cfg_space.c
index 4ce2e6bd0680..97bfc00d2a82 100644
--- a/drivers/gpu/drm/i915/gvt/cfg_space.c
+++ b/drivers/gpu/drm/i915/gvt/cfg_space.c
@@ -335,7 +335,8 @@ int intel_vgpu_emulate_cfg_write(struct intel_vgpu *vgpu, unsigned int offset,
 	case INTEL_GVT_PCI_OPREGION:
 		if (WARN_ON(!IS_ALIGNED(offset, 4)))
 			return -EINVAL;
-		ret = intel_vgpu_init_opregion(vgpu, *(u32 *)p_data);
+		ret = intel_vgpu_opregion_base_write_handler(vgpu,
+						   *(u32 *)p_data);
 		if (ret)
 			return ret;
 
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
index 49af94627c8a..c8454ac43fae 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.c
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -709,18 +709,13 @@ static void parser_exec_state_dump(struct parser_exec_state *s)
 
 	print_opcode(cmd_val(s, 0), s->ring_id);
 
-	/* print the whole page to trace */
-	pr_err("    ip_va=%p: %08x %08x %08x %08x\n",
-			s->ip_va, cmd_val(s, 0), cmd_val(s, 1),
-			cmd_val(s, 2), cmd_val(s, 3));
-
 	s->ip_va = (u32 *)((((u64)s->ip_va) >> 12) << 12);
 
 	while (cnt < 1024) {
-		pr_err("ip_va=%p: ", s->ip_va);
+		gvt_dbg_cmd("ip_va=%p: ", s->ip_va);
 		for (i = 0; i < 8; i++)
-			pr_err("%08x ", cmd_val(s, i));
-		pr_err("\n");
+			gvt_dbg_cmd("%08x ", cmd_val(s, i));
+		gvt_dbg_cmd("\n");
 
 		s->ip_va += 8 * sizeof(u32);
 		cnt += 8;
@@ -825,11 +820,26 @@ static int force_nonpriv_reg_handler(struct parser_exec_state *s,
 	if (!intel_gvt_in_force_nonpriv_whitelist(gvt, data)) {
 		gvt_err("Unexpected forcenonpriv 0x%x LRI write, value=0x%x\n",
 			offset, data);
-		return -EINVAL;
+		return -EPERM;
 	}
 	return 0;
 }
 
+static inline bool is_mocs_mmio(unsigned int offset)
+{
+	return ((offset >= 0xc800) && (offset <= 0xcff8)) ||
+		((offset >= 0xb020) && (offset <= 0xb0a0));
+}
+
+static int mocs_cmd_reg_handler(struct parser_exec_state *s,
+				unsigned int offset, unsigned int index)
+{
+	if (!is_mocs_mmio(offset))
+		return -EINVAL;
+	vgpu_vreg(s->vgpu, offset) = cmd_val(s, index + 1);
+	return 0;
+}
+
 static int cmd_reg_handler(struct parser_exec_state *s,
 	unsigned int offset, unsigned int index, char *cmd)
 {
@@ -839,7 +849,7 @@ static int cmd_reg_handler(struct parser_exec_state *s,
 	if (offset + 4 > gvt->device_info.mmio_size) {
 		gvt_vgpu_err("%s access to (%x) outside of MMIO range\n",
 				cmd, offset);
-		return -EINVAL;
+		return -EFAULT;
 	}
 
 	if (!intel_gvt_mmio_is_cmd_access(gvt, offset)) {
@@ -853,10 +863,14 @@ static int cmd_reg_handler(struct parser_exec_state *s,
 		return 0;
 	}
 
-	if (is_force_nonpriv_mmio(offset) &&
-	    force_nonpriv_reg_handler(s, offset, index))
+	if (is_mocs_mmio(offset) &&
+	    mocs_cmd_reg_handler(s, offset, index))
 		return -EINVAL;
 
+	if (is_force_nonpriv_mmio(offset) &&
+		force_nonpriv_reg_handler(s, offset, index))
+		return -EPERM;
+
 	if (offset == i915_mmio_reg_offset(DERRMR) ||
 		offset == i915_mmio_reg_offset(FORCEWAKE_MT)) {
 		/* Writing to HW VGT_PVINFO_PAGE offset will be discarded */
@@ -894,11 +908,14 @@ static int cmd_handler_lri(struct parser_exec_state *s)
 					i915_mmio_reg_offset(DERRMR))
 				ret |= 0;
 			else
-				ret |= (cmd_reg_inhibit(s, i)) ? -EINVAL : 0;
+				ret |= (cmd_reg_inhibit(s, i)) ?
+					-EBADRQC : 0;
 		}
 		if (ret)
 			break;
 		ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lri");
+		if (ret)
+			break;
 	}
 	return ret;
 }
@@ -912,11 +929,15 @@ static int cmd_handler_lrr(struct parser_exec_state *s)
 		if (IS_BROADWELL(s->vgpu->gvt->dev_priv))
 			ret |= ((cmd_reg_inhibit(s, i) ||
 					(cmd_reg_inhibit(s, i + 1)))) ?
-				-EINVAL : 0;
+				-EBADRQC : 0;
 		if (ret)
 			break;
 		ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrr-src");
+		if (ret)
+			break;
 		ret |= cmd_reg_handler(s, cmd_reg(s, i + 1), i, "lrr-dst");
+		if (ret)
+			break;
 	}
 	return ret;
 }
@@ -934,15 +955,19 @@ static int cmd_handler_lrm(struct parser_exec_state *s)
 
 	for (i = 1; i < cmd_len;) {
 		if (IS_BROADWELL(gvt->dev_priv))
-			ret |= (cmd_reg_inhibit(s, i)) ? -EINVAL : 0;
+			ret |= (cmd_reg_inhibit(s, i)) ? -EBADRQC : 0;
 		if (ret)
 			break;
 		ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrm");
+		if (ret)
+			break;
 		if (cmd_val(s, 0) & (1 << 22)) {
 			gma = cmd_gma(s, i + 1);
 			if (gmadr_bytes == 8)
 				gma |= (cmd_gma_hi(s, i + 2)) << 32;
 			ret |= cmd_address_audit(s, gma, sizeof(u32), false);
+			if (ret)
+				break;
 		}
 		i += gmadr_dw_number(s) + 1;
 	}
@@ -958,11 +983,15 @@ static int cmd_handler_srm(struct parser_exec_state *s)
 
 	for (i = 1; i < cmd_len;) {
 		ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "srm");
+		if (ret)
+			break;
 		if (cmd_val(s, 0) & (1 << 22)) {
 			gma = cmd_gma(s, i + 1);
 			if (gmadr_bytes == 8)
 				gma |= (cmd_gma_hi(s, i + 2)) << 32;
 			ret |= cmd_address_audit(s, gma, sizeof(u32), false);
+			if (ret)
+				break;
 		}
 		i += gmadr_dw_number(s) + 1;
 	}
@@ -1116,7 +1145,7 @@ static int gen8_decode_mi_display_flip(struct parser_exec_state *s,
 
 	v = (dword0 & GENMASK(21, 19)) >> 19;
 	if (WARN_ON(v >= ARRAY_SIZE(gen8_plane_code)))
-		return -EINVAL;
+		return -EBADRQC;
 
 	info->pipe = gen8_plane_code[v].pipe;
 	info->plane = gen8_plane_code[v].plane;
@@ -1136,7 +1165,7 @@ static int gen8_decode_mi_display_flip(struct parser_exec_state *s,
 		info->surf_reg = SPRSURF(info->pipe);
 	} else {
 		WARN_ON(1);
-		return -EINVAL;
+		return -EBADRQC;
 	}
 	return 0;
 }
@@ -1185,7 +1214,7 @@ static int skl_decode_mi_display_flip(struct parser_exec_state *s,
 
 	default:
 		gvt_vgpu_err("unknown plane code %d\n", plane);
-		return -EINVAL;
+		return -EBADRQC;
 	}
 
 	info->stride_val = (dword1 & GENMASK(15, 6)) >> 6;
@@ -1210,13 +1239,13 @@ static int gen8_check_mi_display_flip(struct parser_exec_state *s,
 		return 0;
 
 	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
-		stride = vgpu_vreg(s->vgpu, info->stride_reg) & GENMASK(9, 0);
-		tile = (vgpu_vreg(s->vgpu, info->ctrl_reg) &
+		stride = vgpu_vreg_t(s->vgpu, info->stride_reg) & GENMASK(9, 0);
+		tile = (vgpu_vreg_t(s->vgpu, info->ctrl_reg) &
 				GENMASK(12, 10)) >> 10;
 	} else {
-		stride = (vgpu_vreg(s->vgpu, info->stride_reg) &
+		stride = (vgpu_vreg_t(s->vgpu, info->stride_reg) &
 				GENMASK(15, 6)) >> 6;
-		tile = (vgpu_vreg(s->vgpu, info->ctrl_reg) & (1 << 10)) >> 10;
+		tile = (vgpu_vreg_t(s->vgpu, info->ctrl_reg) & (1 << 10)) >> 10;
 	}
 
 	if (stride != info->stride_val)
@@ -1235,21 +1264,21 @@ static int gen8_update_plane_mmio_from_mi_display_flip(
 	struct drm_i915_private *dev_priv = s->vgpu->gvt->dev_priv;
 	struct intel_vgpu *vgpu = s->vgpu;
 
-	set_mask_bits(&vgpu_vreg(vgpu, info->surf_reg), GENMASK(31, 12),
+	set_mask_bits(&vgpu_vreg_t(vgpu, info->surf_reg), GENMASK(31, 12),
 		      info->surf_val << 12);
 	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
-		set_mask_bits(&vgpu_vreg(vgpu, info->stride_reg), GENMASK(9, 0),
+		set_mask_bits(&vgpu_vreg_t(vgpu, info->stride_reg), GENMASK(9, 0),
 			      info->stride_val);
-		set_mask_bits(&vgpu_vreg(vgpu, info->ctrl_reg), GENMASK(12, 10),
+		set_mask_bits(&vgpu_vreg_t(vgpu, info->ctrl_reg), GENMASK(12, 10),
 			      info->tile_val << 10);
 	} else {
-		set_mask_bits(&vgpu_vreg(vgpu, info->stride_reg), GENMASK(15, 6),
+		set_mask_bits(&vgpu_vreg_t(vgpu, info->stride_reg), GENMASK(15, 6),
 			      info->stride_val << 6);
-		set_mask_bits(&vgpu_vreg(vgpu, info->ctrl_reg), GENMASK(10, 10),
+		set_mask_bits(&vgpu_vreg_t(vgpu, info->ctrl_reg), GENMASK(10, 10),
 			      info->tile_val << 10);
 	}
 
-	vgpu_vreg(vgpu, PIPE_FRMCOUNT_G4X(info->pipe))++;
+	vgpu_vreg_t(vgpu, PIPE_FRMCOUNT_G4X(info->pipe))++;
 	intel_vgpu_trigger_virtual_event(vgpu, info->event);
 	return 0;
 }
@@ -1348,10 +1377,13 @@ static unsigned long get_gma_bb_from_cmd(struct parser_exec_state *s, int index)
 {
 	unsigned long addr;
 	unsigned long gma_high, gma_low;
-	int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
+	struct intel_vgpu *vgpu = s->vgpu;
+	int gmadr_bytes = vgpu->gvt->device_info.gmadr_bytes_in_cmd;
 
-	if (WARN_ON(gmadr_bytes != 4 && gmadr_bytes != 8))
+	if (WARN_ON(gmadr_bytes != 4 && gmadr_bytes != 8)) {
+		gvt_vgpu_err("invalid gma bytes %d\n", gmadr_bytes);
 		return INTEL_GVT_INVALID_ADDR;
+	}
 
 	gma_low = cmd_val(s, index) & BATCH_BUFFER_ADDR_MASK;
 	if (gmadr_bytes == 4) {
@@ -1374,16 +1406,16 @@ static inline int cmd_address_audit(struct parser_exec_state *s,
 	if (op_size > max_surface_size) {
 		gvt_vgpu_err("command address audit fail name %s\n",
 			s->info->name);
-		return -EINVAL;
+		return -EFAULT;
 	}
 
 	if (index_mode)	{
-		if (guest_gma >= GTT_PAGE_SIZE / sizeof(u64)) {
-			ret = -EINVAL;
+		if (guest_gma >= I915_GTT_PAGE_SIZE / sizeof(u64)) {
+			ret = -EFAULT;
 			goto err;
 		}
 	} else if (!intel_gvt_ggtt_validate_range(vgpu, guest_gma, op_size)) {
-		ret = -EINVAL;
+		ret = -EFAULT;
 		goto err;
 	}
 
@@ -1439,7 +1471,7 @@ static inline int unexpected_cmd(struct parser_exec_state *s)
 
 	gvt_vgpu_err("Unexpected %s in command buffer!\n", s->info->name);
 
-	return -EINVAL;
+	return -EBADRQC;
 }
 
 static int cmd_handler_mi_semaphore_wait(struct parser_exec_state *s)
@@ -1545,10 +1577,10 @@ static int copy_gma_to_hva(struct intel_vgpu *vgpu, struct intel_vgpu_mm *mm,
 			return -EFAULT;
 		}
 
-		offset = gma & (GTT_PAGE_SIZE - 1);
+		offset = gma & (I915_GTT_PAGE_SIZE - 1);
 
-		copy_len = (end_gma - gma) >= (GTT_PAGE_SIZE - offset) ?
-			GTT_PAGE_SIZE - offset : end_gma - gma;
+		copy_len = (end_gma - gma) >= (I915_GTT_PAGE_SIZE - offset) ?
+			I915_GTT_PAGE_SIZE - offset : end_gma - gma;
 
 		intel_gvt_hypervisor_read_gpa(vgpu, gpa, va + len, copy_len);
 
@@ -1576,110 +1608,113 @@ static int batch_buffer_needs_scan(struct parser_exec_state *s)
 	return 1;
 }
 
-static int find_bb_size(struct parser_exec_state *s)
+static int find_bb_size(struct parser_exec_state *s, unsigned long *bb_size)
 {
 	unsigned long gma = 0;
 	struct cmd_info *info;
-	int bb_size = 0;
 	uint32_t cmd_len = 0;
-	bool met_bb_end = false;
+	bool bb_end = false;
 	struct intel_vgpu *vgpu = s->vgpu;
 	u32 cmd;
 
+	*bb_size = 0;
+
 	/* get the start gm address of the batch buffer */
 	gma = get_gma_bb_from_cmd(s, 1);
-	cmd = cmd_val(s, 0);
+	if (gma == INTEL_GVT_INVALID_ADDR)
+		return -EFAULT;
 
+	cmd = cmd_val(s, 0);
 	info = get_cmd_info(s->vgpu->gvt, cmd, s->ring_id);
 	if (info == NULL) {
 		gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x\n",
 				cmd, get_opcode(cmd, s->ring_id));
-		return -EINVAL;
+		return -EBADRQC;
 	}
 	do {
-		copy_gma_to_hva(s->vgpu, s->vgpu->gtt.ggtt_mm,
-				gma, gma + 4, &cmd);
+		if (copy_gma_to_hva(s->vgpu, s->vgpu->gtt.ggtt_mm,
+				gma, gma + 4, &cmd) < 0)
+			return -EFAULT;
 		info = get_cmd_info(s->vgpu->gvt, cmd, s->ring_id);
 		if (info == NULL) {
 			gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x\n",
 				cmd, get_opcode(cmd, s->ring_id));
-			return -EINVAL;
+			return -EBADRQC;
 		}
 
 		if (info->opcode == OP_MI_BATCH_BUFFER_END) {
-			met_bb_end = true;
+			bb_end = true;
 		} else if (info->opcode == OP_MI_BATCH_BUFFER_START) {
-			if (BATCH_BUFFER_2ND_LEVEL_BIT(cmd) == 0) {
+			if (BATCH_BUFFER_2ND_LEVEL_BIT(cmd) == 0)
 				/* chained batch buffer */
-				met_bb_end = true;
-			}
+				bb_end = true;
 		}
 		cmd_len = get_cmd_length(info, cmd) << 2;
-		bb_size += cmd_len;
+		*bb_size += cmd_len;
 		gma += cmd_len;
+	} while (!bb_end);
 
-	} while (!met_bb_end);
-
-	return bb_size;
+	return 0;
 }
 
 static int perform_bb_shadow(struct parser_exec_state *s)
 {
-	struct intel_shadow_bb_entry *entry_obj;
 	struct intel_vgpu *vgpu = s->vgpu;
+	struct intel_vgpu_shadow_bb *bb;
 	unsigned long gma = 0;
-	int bb_size;
-	void *dst = NULL;
+	unsigned long bb_size;
 	int ret = 0;
 
 	/* get the start gm address of the batch buffer */
 	gma = get_gma_bb_from_cmd(s, 1);
+	if (gma == INTEL_GVT_INVALID_ADDR)
+		return -EFAULT;
 
-	/* get the size of the batch buffer */
-	bb_size = find_bb_size(s);
-	if (bb_size < 0)
-		return -EINVAL;
+	ret = find_bb_size(s, &bb_size);
+	if (ret)
+		return ret;
 
-	/* allocate shadow batch buffer */
-	entry_obj = kmalloc(sizeof(*entry_obj), GFP_KERNEL);
-	if (entry_obj == NULL)
+	bb = kzalloc(sizeof(*bb), GFP_KERNEL);
+	if (!bb)
 		return -ENOMEM;
 
-	entry_obj->obj =
-		i915_gem_object_create(s->vgpu->gvt->dev_priv,
-				       roundup(bb_size, PAGE_SIZE));
-	if (IS_ERR(entry_obj->obj)) {
-		ret = PTR_ERR(entry_obj->obj);
-		goto free_entry;
+	bb->obj = i915_gem_object_create(s->vgpu->gvt->dev_priv,
+					 roundup(bb_size, PAGE_SIZE));
+	if (IS_ERR(bb->obj)) {
+		ret = PTR_ERR(bb->obj);
+		goto err_free_bb;
 	}
-	entry_obj->len = bb_size;
-	INIT_LIST_HEAD(&entry_obj->list);
 
-	dst = i915_gem_object_pin_map(entry_obj->obj, I915_MAP_WB);
-	if (IS_ERR(dst)) {
-		ret = PTR_ERR(dst);
-		goto put_obj;
-	}
+	ret = i915_gem_obj_prepare_shmem_write(bb->obj, &bb->clflush);
+	if (ret)
+		goto err_free_obj;
 
-	ret = i915_gem_object_set_to_cpu_domain(entry_obj->obj, false);
-	if (ret) {
-		gvt_vgpu_err("failed to set shadow batch to CPU\n");
-		goto unmap_src;
+	bb->va = i915_gem_object_pin_map(bb->obj, I915_MAP_WB);
+	if (IS_ERR(bb->va)) {
+		ret = PTR_ERR(bb->va);
+		goto err_finish_shmem_access;
 	}
 
-	entry_obj->va = dst;
-	entry_obj->bb_start_cmd_va = s->ip_va;
+	if (bb->clflush & CLFLUSH_BEFORE) {
+		drm_clflush_virt_range(bb->va, bb->obj->base.size);
+		bb->clflush &= ~CLFLUSH_BEFORE;
+	}
 
-	/* copy batch buffer to shadow batch buffer*/
 	ret = copy_gma_to_hva(s->vgpu, s->vgpu->gtt.ggtt_mm,
 			      gma, gma + bb_size,
-			      dst);
+			      bb->va);
 	if (ret < 0) {
 		gvt_vgpu_err("fail to copy guest ring buffer\n");
-		goto unmap_src;
+		ret = -EFAULT;
+		goto err_unmap;
 	}
 
-	list_add(&entry_obj->list, &s->workload->shadow_bb);
+	INIT_LIST_HEAD(&bb->list);
+	list_add(&bb->list, &s->workload->shadow_bb);
+
+	bb->accessing = true;
+	bb->bb_start_cmd_va = s->ip_va;
+
 	/*
 	 * ip_va saves the virtual address of the shadow batch buffer, while
 	 * ip_gma saves the graphics address of the original batch buffer.
@@ -1688,17 +1723,17 @@ static int perform_bb_shadow(struct parser_exec_state *s)
 	 * buffer's gma in pair. After all, we don't want to pin the shadow
 	 * buffer here (too early).
 	 */
-	s->ip_va = dst;
+	s->ip_va = bb->va;
 	s->ip_gma = gma;
-
 	return 0;
-
-unmap_src:
-	i915_gem_object_unpin_map(entry_obj->obj);
-put_obj:
-	i915_gem_object_put(entry_obj->obj);
-free_entry:
-	kfree(entry_obj);
+err_unmap:
+	i915_gem_object_unpin_map(bb->obj);
+err_finish_shmem_access:
+	i915_gem_obj_finish_shmem_access(bb->obj);
+err_free_obj:
+	i915_gem_object_put(bb->obj);
+err_free_bb:
+	kfree(bb);
 	return ret;
 }
 
@@ -1710,13 +1745,13 @@ static int cmd_handler_mi_batch_buffer_start(struct parser_exec_state *s)
 
 	if (s->buf_type == BATCH_BUFFER_2ND_LEVEL) {
 		gvt_vgpu_err("Found MI_BATCH_BUFFER_START in 2nd level BB\n");
-		return -EINVAL;
+		return -EFAULT;
 	}
 
 	second_level = BATCH_BUFFER_2ND_LEVEL_BIT(cmd_val(s, 0)) == 1;
 	if (second_level && (s->buf_type != BATCH_BUFFER_INSTRUCTION)) {
 		gvt_vgpu_err("Jumping to 2nd level BB from RB is not allowed\n");
-		return -EINVAL;
+		return -EFAULT;
 	}
 
 	s->saved_buf_addr_type = s->buf_addr_type;
@@ -1740,7 +1775,6 @@ static int cmd_handler_mi_batch_buffer_start(struct parser_exec_state *s)
 		if (ret < 0)
 			return ret;
 	}
-
 	return ret;
 }
 
@@ -2430,7 +2464,7 @@ static int cmd_parser_exec(struct parser_exec_state *s)
 	if (info == NULL) {
 		gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x\n",
 				cmd, get_opcode(cmd, s->ring_id));
-		return -EINVAL;
+		return -EBADRQC;
 	}
 
 	s->info = info;
@@ -2465,6 +2499,10 @@ static inline bool gma_out_of_range(unsigned long gma,
 		return (gma > gma_tail) && (gma < gma_head);
 }
 
+/* Keep the consistent return type, e.g EBADRQC for unknown
+ * cmd, EFAULT for invalid address, EPERM for nonpriv. later
+ * works as the input of VM healthy status.
+ */
 static int command_scan(struct parser_exec_state *s,
 		unsigned long rb_head, unsigned long rb_tail,
 		unsigned long rb_start, unsigned long rb_len)
@@ -2487,7 +2525,7 @@ static int command_scan(struct parser_exec_state *s,
 					s->ip_gma, rb_start,
 					gma_bottom);
 				parser_exec_state_dump(s);
-				return -EINVAL;
+				return -EFAULT;
 			}
 			if (gma_out_of_range(s->ip_gma, gma_head, gma_tail)) {
 				gvt_vgpu_err("ip_gma %lx out of range."
@@ -2516,7 +2554,7 @@ static int scan_workload(struct intel_vgpu_workload *workload)
 	int ret = 0;
 
 	/* ring base is page aligned */
-	if (WARN_ON(!IS_ALIGNED(workload->rb_start, GTT_PAGE_SIZE)))
+	if (WARN_ON(!IS_ALIGNED(workload->rb_start, I915_GTT_PAGE_SIZE)))
 		return -EINVAL;
 
 	gma_head = workload->rb_start + workload->rb_head;
@@ -2565,7 +2603,8 @@ static int scan_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
 				wa_ctx);
 
 	/* ring base is page aligned */
-	if (WARN_ON(!IS_ALIGNED(wa_ctx->indirect_ctx.guest_gma, GTT_PAGE_SIZE)))
+	if (WARN_ON(!IS_ALIGNED(wa_ctx->indirect_ctx.guest_gma,
+					I915_GTT_PAGE_SIZE)))
 		return -EINVAL;
 
 	ring_tail = wa_ctx->indirect_ctx.size + 3 * sizeof(uint32_t);
@@ -2604,6 +2643,7 @@ out:
 static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload)
 {
 	struct intel_vgpu *vgpu = workload->vgpu;
+	struct intel_vgpu_submission *s = &vgpu->submission;
 	unsigned long gma_head, gma_tail, gma_top, guest_rb_size;
 	void *shadow_ring_buffer_va;
 	int ring_id = workload->ring_id;
@@ -2619,19 +2659,21 @@ static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload)
 	gma_tail = workload->rb_start + workload->rb_tail;
 	gma_top = workload->rb_start + guest_rb_size;
 
-	if (workload->rb_len > vgpu->reserve_ring_buffer_size[ring_id]) {
-		void *va = vgpu->reserve_ring_buffer_va[ring_id];
+	if (workload->rb_len > s->ring_scan_buffer_size[ring_id]) {
+		void *p;
+
 		/* realloc the new ring buffer if needed */
-		vgpu->reserve_ring_buffer_va[ring_id] =
-			krealloc(va, workload->rb_len, GFP_KERNEL);
-		if (!vgpu->reserve_ring_buffer_va[ring_id]) {
-			gvt_vgpu_err("fail to alloc reserve ring buffer\n");
+		p = krealloc(s->ring_scan_buffer[ring_id], workload->rb_len,
+				GFP_KERNEL);
+		if (!p) {
+			gvt_vgpu_err("fail to re-alloc ring scan buffer\n");
 			return -ENOMEM;
 		}
-		vgpu->reserve_ring_buffer_size[ring_id] = workload->rb_len;
+		s->ring_scan_buffer[ring_id] = p;
+		s->ring_scan_buffer_size[ring_id] = workload->rb_len;
 	}
 
-	shadow_ring_buffer_va = vgpu->reserve_ring_buffer_va[ring_id];
+	shadow_ring_buffer_va = s->ring_scan_buffer[ring_id];
 
 	/* get shadow ring buffer va */
 	workload->shadow_ring_buffer_va = shadow_ring_buffer_va;
diff --git a/drivers/gpu/drm/i915/gvt/debug.h b/drivers/gpu/drm/i915/gvt/debug.h
index b0cff4dc2684..c6027125c1ec 100644
--- a/drivers/gpu/drm/i915/gvt/debug.h
+++ b/drivers/gpu/drm/i915/gvt/debug.h
@@ -25,41 +25,41 @@
 #define __GVT_DEBUG_H__
 
 #define gvt_err(fmt, args...) \
-	DRM_ERROR("gvt: "fmt, ##args)
+	pr_err("gvt: "fmt, ##args)
 
 #define gvt_vgpu_err(fmt, args...)					\
 do {									\
 	if (IS_ERR_OR_NULL(vgpu))					\
-		DRM_DEBUG_DRIVER("gvt: "fmt, ##args);			\
+		pr_err("gvt: "fmt, ##args);			\
 	else								\
-		DRM_DEBUG_DRIVER("gvt: vgpu %d: "fmt, vgpu->id, ##args);\
+		pr_err("gvt: vgpu %d: "fmt, vgpu->id, ##args);\
 } while (0)
 
 #define gvt_dbg_core(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: core: "fmt, ##args)
+	pr_debug("gvt: core: "fmt, ##args)
 
 #define gvt_dbg_irq(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: irq: "fmt, ##args)
+	pr_debug("gvt: irq: "fmt, ##args)
 
 #define gvt_dbg_mm(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: mm: "fmt, ##args)
+	pr_debug("gvt: mm: "fmt, ##args)
 
 #define gvt_dbg_mmio(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: mmio: "fmt, ##args)
+	pr_debug("gvt: mmio: "fmt, ##args)
 
 #define gvt_dbg_dpy(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: dpy: "fmt, ##args)
+	pr_debug("gvt: dpy: "fmt, ##args)
 
 #define gvt_dbg_el(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: el: "fmt, ##args)
+	pr_debug("gvt: el: "fmt, ##args)
 
 #define gvt_dbg_sched(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: sched: "fmt, ##args)
+	pr_debug("gvt: sched: "fmt, ##args)
 
 #define gvt_dbg_render(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: render: "fmt, ##args)
+	pr_debug("gvt: render: "fmt, ##args)
 
 #define gvt_dbg_cmd(fmt, args...) \
-	DRM_DEBUG_DRIVER("gvt: cmd: "fmt, ##args)
+	pr_debug("gvt: cmd: "fmt, ##args)
 
 #endif
diff --git a/drivers/gpu/drm/i915/gvt/debugfs.c b/drivers/gpu/drm/i915/gvt/debugfs.c
new file mode 100644
index 000000000000..32a66dfdf112
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/debugfs.c
@@ -0,0 +1,212 @@
+/*
+ * Copyright(c) 2011-2017 Intel Corporation. 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, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include <linux/debugfs.h>
+#include <linux/list_sort.h>
+#include "i915_drv.h"
+#include "gvt.h"
+
+struct mmio_diff_param {
+	struct intel_vgpu *vgpu;
+	int total;
+	int diff;
+	struct list_head diff_mmio_list;
+};
+
+struct diff_mmio {
+	struct list_head node;
+	u32 offset;
+	u32 preg;
+	u32 vreg;
+};
+
+/* Compare two diff_mmio items. */
+static int mmio_offset_compare(void *priv,
+	struct list_head *a, struct list_head *b)
+{
+	struct diff_mmio *ma;
+	struct diff_mmio *mb;
+
+	ma = container_of(a, struct diff_mmio, node);
+	mb = container_of(b, struct diff_mmio, node);
+	if (ma->offset < mb->offset)
+		return -1;
+	else if (ma->offset > mb->offset)
+		return 1;
+	return 0;
+}
+
+static inline int mmio_diff_handler(struct intel_gvt *gvt,
+				    u32 offset, void *data)
+{
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	struct mmio_diff_param *param = data;
+	struct diff_mmio *node;
+	u32 preg, vreg;
+
+	preg = I915_READ_NOTRACE(_MMIO(offset));
+	vreg = vgpu_vreg(param->vgpu, offset);
+
+	if (preg != vreg) {
+		node = kmalloc(sizeof(*node), GFP_KERNEL);
+		if (!node)
+			return -ENOMEM;
+
+		node->offset = offset;
+		node->preg = preg;
+		node->vreg = vreg;
+		list_add(&node->node, &param->diff_mmio_list);
+		param->diff++;
+	}
+	param->total++;
+	return 0;
+}
+
+/* Show the all the different values of tracked mmio. */
+static int vgpu_mmio_diff_show(struct seq_file *s, void *unused)
+{
+	struct intel_vgpu *vgpu = s->private;
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct mmio_diff_param param = {
+		.vgpu = vgpu,
+		.total = 0,
+		.diff = 0,
+	};
+	struct diff_mmio *node, *next;
+
+	INIT_LIST_HEAD(&param.diff_mmio_list);
+
+	mutex_lock(&gvt->lock);
+	spin_lock_bh(&gvt->scheduler.mmio_context_lock);
+
+	mmio_hw_access_pre(gvt->dev_priv);
+	/* Recognize all the diff mmios to list. */
+	intel_gvt_for_each_tracked_mmio(gvt, mmio_diff_handler, &param);
+	mmio_hw_access_post(gvt->dev_priv);
+
+	spin_unlock_bh(&gvt->scheduler.mmio_context_lock);
+	mutex_unlock(&gvt->lock);
+
+	/* In an ascending order by mmio offset. */
+	list_sort(NULL, &param.diff_mmio_list, mmio_offset_compare);
+
+	seq_printf(s, "%-8s %-8s %-8s %-8s\n", "Offset", "HW", "vGPU", "Diff");
+	list_for_each_entry_safe(node, next, &param.diff_mmio_list, node) {
+		u32 diff = node->preg ^ node->vreg;
+
+		seq_printf(s, "%08x %08x %08x %*pbl\n",
+			   node->offset, node->preg, node->vreg,
+			   32, &diff);
+		list_del(&node->node);
+		kfree(node);
+	}
+	seq_printf(s, "Total: %d, Diff: %d\n", param.total, param.diff);
+	return 0;
+}
+
+static int vgpu_mmio_diff_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, vgpu_mmio_diff_show, inode->i_private);
+}
+
+static const struct file_operations vgpu_mmio_diff_fops = {
+	.open		= vgpu_mmio_diff_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+/**
+ * intel_gvt_debugfs_add_vgpu - register debugfs entries for a vGPU
+ * @vgpu: a vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu)
+{
+	struct dentry *ent;
+	char name[10] = "";
+
+	sprintf(name, "vgpu%d", vgpu->id);
+	vgpu->debugfs = debugfs_create_dir(name, vgpu->gvt->debugfs_root);
+	if (!vgpu->debugfs)
+		return -ENOMEM;
+
+	ent = debugfs_create_bool("active", 0444, vgpu->debugfs,
+				  &vgpu->active);
+	if (!ent)
+		return -ENOMEM;
+
+	ent = debugfs_create_file("mmio_diff", 0444, vgpu->debugfs,
+				  vgpu, &vgpu_mmio_diff_fops);
+	if (!ent)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/**
+ * intel_gvt_debugfs_remove_vgpu - remove debugfs entries of a vGPU
+ * @vgpu: a vGPU
+ */
+void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu)
+{
+	debugfs_remove_recursive(vgpu->debugfs);
+	vgpu->debugfs = NULL;
+}
+
+/**
+ * intel_gvt_debugfs_init - register gvt debugfs root entry
+ * @gvt: GVT device
+ *
+ * Returns:
+ * zero on success, negative if failed.
+ */
+int intel_gvt_debugfs_init(struct intel_gvt *gvt)
+{
+	struct drm_minor *minor = gvt->dev_priv->drm.primary;
+	struct dentry *ent;
+
+	gvt->debugfs_root = debugfs_create_dir("gvt", minor->debugfs_root);
+	if (!gvt->debugfs_root) {
+		gvt_err("Cannot create debugfs dir\n");
+		return -ENOMEM;
+	}
+
+	ent = debugfs_create_ulong("num_tracked_mmio", 0444, gvt->debugfs_root,
+				   &gvt->mmio.num_tracked_mmio);
+	if (!ent)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/**
+ * intel_gvt_debugfs_clean - remove debugfs entries
+ * @gvt: GVT device
+ */
+void intel_gvt_debugfs_clean(struct intel_gvt *gvt)
+{
+	debugfs_remove_recursive(gvt->debugfs_root);
+	gvt->debugfs_root = NULL;
+}
diff --git a/drivers/gpu/drm/i915/gvt/display.c b/drivers/gpu/drm/i915/gvt/display.c
index 309f3fa6794a..dd96ffc878ac 100644
--- a/drivers/gpu/drm/i915/gvt/display.c
+++ b/drivers/gpu/drm/i915/gvt/display.c
@@ -59,7 +59,7 @@ static int edp_pipe_is_enabled(struct intel_vgpu *vgpu)
 {
 	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
 
-	if (!(vgpu_vreg(vgpu, PIPECONF(_PIPE_EDP)) & PIPECONF_ENABLE))
+	if (!(vgpu_vreg_t(vgpu, PIPECONF(_PIPE_EDP)) & PIPECONF_ENABLE))
 		return 0;
 
 	if (!(vgpu_vreg(vgpu, _TRANS_DDI_FUNC_CTL_EDP) & TRANS_DDI_FUNC_ENABLE))
@@ -67,14 +67,14 @@ static int edp_pipe_is_enabled(struct intel_vgpu *vgpu)
 	return 1;
 }
 
-static int pipe_is_enabled(struct intel_vgpu *vgpu, int pipe)
+int pipe_is_enabled(struct intel_vgpu *vgpu, int pipe)
 {
 	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
 
 	if (WARN_ON(pipe < PIPE_A || pipe >= I915_MAX_PIPES))
 		return -EINVAL;
 
-	if (vgpu_vreg(vgpu, PIPECONF(pipe)) & PIPECONF_ENABLE)
+	if (vgpu_vreg_t(vgpu, PIPECONF(pipe)) & PIPECONF_ENABLE)
 		return 1;
 
 	if (edp_pipe_is_enabled(vgpu) &&
@@ -169,105 +169,105 @@ static u8 dpcd_fix_data[DPCD_HEADER_SIZE] = {
 static void emulate_monitor_status_change(struct intel_vgpu *vgpu)
 {
 	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	vgpu_vreg(vgpu, SDEISR) &= ~(SDE_PORTB_HOTPLUG_CPT |
+	vgpu_vreg_t(vgpu, SDEISR) &= ~(SDE_PORTB_HOTPLUG_CPT |
 			SDE_PORTC_HOTPLUG_CPT |
 			SDE_PORTD_HOTPLUG_CPT);
 
 	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
-		vgpu_vreg(vgpu, SDEISR) &= ~(SDE_PORTA_HOTPLUG_SPT |
+		vgpu_vreg_t(vgpu, SDEISR) &= ~(SDE_PORTA_HOTPLUG_SPT |
 				SDE_PORTE_HOTPLUG_SPT);
-		vgpu_vreg(vgpu, SKL_FUSE_STATUS) |=
+		vgpu_vreg_t(vgpu, SKL_FUSE_STATUS) |=
 				SKL_FUSE_DOWNLOAD_STATUS |
 				SKL_FUSE_PG_DIST_STATUS(SKL_PG0) |
 				SKL_FUSE_PG_DIST_STATUS(SKL_PG1) |
 				SKL_FUSE_PG_DIST_STATUS(SKL_PG2);
-		vgpu_vreg(vgpu, LCPLL1_CTL) |=
+		vgpu_vreg_t(vgpu, LCPLL1_CTL) |=
 				LCPLL_PLL_ENABLE |
 				LCPLL_PLL_LOCK;
-		vgpu_vreg(vgpu, LCPLL2_CTL) |= LCPLL_PLL_ENABLE;
+		vgpu_vreg_t(vgpu, LCPLL2_CTL) |= LCPLL_PLL_ENABLE;
 
 	}
 
 	if (intel_vgpu_has_monitor_on_port(vgpu, PORT_B)) {
-		vgpu_vreg(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIB_DETECTED;
-		vgpu_vreg(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
+		vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIB_DETECTED;
+		vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
 			~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
 			TRANS_DDI_PORT_MASK);
-		vgpu_vreg(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
+		vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
 			(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
 			(PORT_B << TRANS_DDI_PORT_SHIFT) |
 			TRANS_DDI_FUNC_ENABLE);
 		if (IS_BROADWELL(dev_priv)) {
-			vgpu_vreg(vgpu, PORT_CLK_SEL(PORT_B)) &=
+			vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_B)) &=
 				~PORT_CLK_SEL_MASK;
-			vgpu_vreg(vgpu, PORT_CLK_SEL(PORT_B)) |=
+			vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_B)) |=
 				PORT_CLK_SEL_LCPLL_810;
 		}
-		vgpu_vreg(vgpu, DDI_BUF_CTL(PORT_B)) |= DDI_BUF_CTL_ENABLE;
-		vgpu_vreg(vgpu, DDI_BUF_CTL(PORT_B)) &= ~DDI_BUF_IS_IDLE;
-		vgpu_vreg(vgpu, SDEISR) |= SDE_PORTB_HOTPLUG_CPT;
+		vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_B)) |= DDI_BUF_CTL_ENABLE;
+		vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_B)) &= ~DDI_BUF_IS_IDLE;
+		vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTB_HOTPLUG_CPT;
 	}
 
 	if (intel_vgpu_has_monitor_on_port(vgpu, PORT_C)) {
-		vgpu_vreg(vgpu, SDEISR) |= SDE_PORTC_HOTPLUG_CPT;
-		vgpu_vreg(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
+		vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTC_HOTPLUG_CPT;
+		vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
 			~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
 			TRANS_DDI_PORT_MASK);
-		vgpu_vreg(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
+		vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
 			(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
 			(PORT_C << TRANS_DDI_PORT_SHIFT) |
 			TRANS_DDI_FUNC_ENABLE);
 		if (IS_BROADWELL(dev_priv)) {
-			vgpu_vreg(vgpu, PORT_CLK_SEL(PORT_C)) &=
+			vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_C)) &=
 				~PORT_CLK_SEL_MASK;
-			vgpu_vreg(vgpu, PORT_CLK_SEL(PORT_C)) |=
+			vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_C)) |=
 				PORT_CLK_SEL_LCPLL_810;
 		}
-		vgpu_vreg(vgpu, DDI_BUF_CTL(PORT_C)) |= DDI_BUF_CTL_ENABLE;
-		vgpu_vreg(vgpu, DDI_BUF_CTL(PORT_C)) &= ~DDI_BUF_IS_IDLE;
-		vgpu_vreg(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIC_DETECTED;
+		vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_C)) |= DDI_BUF_CTL_ENABLE;
+		vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_C)) &= ~DDI_BUF_IS_IDLE;
+		vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDIC_DETECTED;
 	}
 
 	if (intel_vgpu_has_monitor_on_port(vgpu, PORT_D)) {
-		vgpu_vreg(vgpu, SDEISR) |= SDE_PORTD_HOTPLUG_CPT;
-		vgpu_vreg(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
+		vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTD_HOTPLUG_CPT;
+		vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) &=
 			~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
 			TRANS_DDI_PORT_MASK);
-		vgpu_vreg(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
+		vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
 			(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
 			(PORT_D << TRANS_DDI_PORT_SHIFT) |
 			TRANS_DDI_FUNC_ENABLE);
 		if (IS_BROADWELL(dev_priv)) {
-			vgpu_vreg(vgpu, PORT_CLK_SEL(PORT_D)) &=
+			vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_D)) &=
 				~PORT_CLK_SEL_MASK;
-			vgpu_vreg(vgpu, PORT_CLK_SEL(PORT_D)) |=
+			vgpu_vreg_t(vgpu, PORT_CLK_SEL(PORT_D)) |=
 				PORT_CLK_SEL_LCPLL_810;
 		}
-		vgpu_vreg(vgpu, DDI_BUF_CTL(PORT_D)) |= DDI_BUF_CTL_ENABLE;
-		vgpu_vreg(vgpu, DDI_BUF_CTL(PORT_D)) &= ~DDI_BUF_IS_IDLE;
-		vgpu_vreg(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDID_DETECTED;
+		vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_D)) |= DDI_BUF_CTL_ENABLE;
+		vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_D)) &= ~DDI_BUF_IS_IDLE;
+		vgpu_vreg_t(vgpu, SFUSE_STRAP) |= SFUSE_STRAP_DDID_DETECTED;
 	}
 
 	if ((IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) &&
 			intel_vgpu_has_monitor_on_port(vgpu, PORT_E)) {
-		vgpu_vreg(vgpu, SDEISR) |= SDE_PORTE_HOTPLUG_SPT;
+		vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTE_HOTPLUG_SPT;
 	}
 
 	if (intel_vgpu_has_monitor_on_port(vgpu, PORT_A)) {
 		if (IS_BROADWELL(dev_priv))
-			vgpu_vreg(vgpu, GEN8_DE_PORT_ISR) |=
+			vgpu_vreg_t(vgpu, GEN8_DE_PORT_ISR) |=
 				GEN8_PORT_DP_A_HOTPLUG;
 		else
-			vgpu_vreg(vgpu, SDEISR) |= SDE_PORTA_HOTPLUG_SPT;
+			vgpu_vreg_t(vgpu, SDEISR) |= SDE_PORTA_HOTPLUG_SPT;
 
-		vgpu_vreg(vgpu, DDI_BUF_CTL(PORT_A)) |= DDI_INIT_DISPLAY_DETECTED;
+		vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_A)) |= DDI_INIT_DISPLAY_DETECTED;
 	}
 
 	/* Clear host CRT status, so guest couldn't detect this host CRT. */
 	if (IS_BROADWELL(dev_priv))
-		vgpu_vreg(vgpu, PCH_ADPA) &= ~ADPA_CRT_HOTPLUG_MONITOR_MASK;
+		vgpu_vreg_t(vgpu, PCH_ADPA) &= ~ADPA_CRT_HOTPLUG_MONITOR_MASK;
 
-	vgpu_vreg(vgpu, PIPECONF(PIPE_A)) |= PIPECONF_ENABLE;
+	vgpu_vreg_t(vgpu, PIPECONF(PIPE_A)) |= PIPECONF_ENABLE;
 }
 
 static void clean_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num)
@@ -369,12 +369,12 @@ static void emulate_vblank_on_pipe(struct intel_vgpu *vgpu, int pipe)
 		if (!pipe_is_enabled(vgpu, pipe))
 			continue;
 
-		vgpu_vreg(vgpu, PIPE_FLIPCOUNT_G4X(pipe))++;
+		vgpu_vreg_t(vgpu, PIPE_FLIPCOUNT_G4X(pipe))++;
 		intel_vgpu_trigger_virtual_event(vgpu, event);
 	}
 
 	if (pipe_is_enabled(vgpu, pipe)) {
-		vgpu_vreg(vgpu, PIPE_FRMCOUNT_G4X(pipe))++;
+		vgpu_vreg_t(vgpu, PIPE_FRMCOUNT_G4X(pipe))++;
 		intel_vgpu_trigger_virtual_event(vgpu, vblank_event[pipe]);
 	}
 }
diff --git a/drivers/gpu/drm/i915/gvt/display.h b/drivers/gpu/drm/i915/gvt/display.h
index d73de22102e2..b46b86892d58 100644
--- a/drivers/gpu/drm/i915/gvt/display.h
+++ b/drivers/gpu/drm/i915/gvt/display.h
@@ -179,4 +179,6 @@ int intel_vgpu_init_display(struct intel_vgpu *vgpu, u64 resolution);
 void intel_vgpu_reset_display(struct intel_vgpu *vgpu);
 void intel_vgpu_clean_display(struct intel_vgpu *vgpu);
 
+int pipe_is_enabled(struct intel_vgpu *vgpu, int pipe);
+
 #endif
diff --git a/drivers/gpu/drm/i915/gvt/dmabuf.c b/drivers/gpu/drm/i915/gvt/dmabuf.c
new file mode 100644
index 000000000000..2ab584f97dfb
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/dmabuf.c
@@ -0,0 +1,537 @@
+/*
+ * Copyright 2017 Intel Corporation. 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, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ *    Xiaoguang Chen
+ *    Tina Zhang <tina.zhang@intel.com>
+ */
+
+#include <linux/dma-buf.h>
+#include <drm/drmP.h>
+#include <linux/vfio.h>
+
+#include "i915_drv.h"
+#include "gvt.h"
+
+#define GEN8_DECODE_PTE(pte) (pte & GENMASK_ULL(63, 12))
+
+static int vgpu_gem_get_pages(
+		struct drm_i915_gem_object *obj)
+{
+	struct drm_i915_private *dev_priv = to_i915(obj->base.dev);
+	struct sg_table *st;
+	struct scatterlist *sg;
+	int i, ret;
+	gen8_pte_t __iomem *gtt_entries;
+	struct intel_vgpu_fb_info *fb_info;
+
+	fb_info = (struct intel_vgpu_fb_info *)obj->gvt_info;
+	if (WARN_ON(!fb_info))
+		return -ENODEV;
+
+	st = kmalloc(sizeof(*st), GFP_KERNEL);
+	if (unlikely(!st))
+		return -ENOMEM;
+
+	ret = sg_alloc_table(st, fb_info->size, GFP_KERNEL);
+	if (ret) {
+		kfree(st);
+		return ret;
+	}
+	gtt_entries = (gen8_pte_t __iomem *)dev_priv->ggtt.gsm +
+		(fb_info->start >> PAGE_SHIFT);
+	for_each_sg(st->sgl, sg, fb_info->size, i) {
+		sg->offset = 0;
+		sg->length = PAGE_SIZE;
+		sg_dma_address(sg) =
+			GEN8_DECODE_PTE(readq(&gtt_entries[i]));
+		sg_dma_len(sg) = PAGE_SIZE;
+	}
+
+	__i915_gem_object_set_pages(obj, st, PAGE_SIZE);
+
+	return 0;
+}
+
+static void vgpu_gem_put_pages(struct drm_i915_gem_object *obj,
+		struct sg_table *pages)
+{
+	sg_free_table(pages);
+	kfree(pages);
+}
+
+static void dmabuf_gem_object_free(struct kref *kref)
+{
+	struct intel_vgpu_dmabuf_obj *obj =
+		container_of(kref, struct intel_vgpu_dmabuf_obj, kref);
+	struct intel_vgpu *vgpu = obj->vgpu;
+	struct list_head *pos;
+	struct intel_vgpu_dmabuf_obj *dmabuf_obj;
+
+	if (vgpu && vgpu->active && !list_empty(&vgpu->dmabuf_obj_list_head)) {
+		list_for_each(pos, &vgpu->dmabuf_obj_list_head) {
+			dmabuf_obj = container_of(pos,
+					struct intel_vgpu_dmabuf_obj, list);
+			if (dmabuf_obj == obj) {
+				intel_gvt_hypervisor_put_vfio_device(vgpu);
+				idr_remove(&vgpu->object_idr,
+					   dmabuf_obj->dmabuf_id);
+				kfree(dmabuf_obj->info);
+				kfree(dmabuf_obj);
+				list_del(pos);
+				break;
+			}
+		}
+	} else {
+		/* Free the orphan dmabuf_objs here */
+		kfree(obj->info);
+		kfree(obj);
+	}
+}
+
+
+static inline void dmabuf_obj_get(struct intel_vgpu_dmabuf_obj *obj)
+{
+	kref_get(&obj->kref);
+}
+
+static inline void dmabuf_obj_put(struct intel_vgpu_dmabuf_obj *obj)
+{
+	kref_put(&obj->kref, dmabuf_gem_object_free);
+}
+
+static void vgpu_gem_release(struct drm_i915_gem_object *gem_obj)
+{
+
+	struct intel_vgpu_fb_info *fb_info = gem_obj->gvt_info;
+	struct intel_vgpu_dmabuf_obj *obj = fb_info->obj;
+	struct intel_vgpu *vgpu = obj->vgpu;
+
+	if (vgpu) {
+		mutex_lock(&vgpu->dmabuf_lock);
+		gem_obj->base.dma_buf = NULL;
+		dmabuf_obj_put(obj);
+		mutex_unlock(&vgpu->dmabuf_lock);
+	} else {
+		/* vgpu is NULL, as it has been removed already */
+		gem_obj->base.dma_buf = NULL;
+		dmabuf_obj_put(obj);
+	}
+}
+
+static const struct drm_i915_gem_object_ops intel_vgpu_gem_ops = {
+	.flags = I915_GEM_OBJECT_IS_PROXY,
+	.get_pages = vgpu_gem_get_pages,
+	.put_pages = vgpu_gem_put_pages,
+	.release = vgpu_gem_release,
+};
+
+static struct drm_i915_gem_object *vgpu_create_gem(struct drm_device *dev,
+		struct intel_vgpu_fb_info *info)
+{
+	struct drm_i915_private *dev_priv = to_i915(dev);
+	struct drm_i915_gem_object *obj;
+
+	obj = i915_gem_object_alloc(dev_priv);
+	if (obj == NULL)
+		return NULL;
+
+	drm_gem_private_object_init(dev, &obj->base,
+		info->size << PAGE_SHIFT);
+	i915_gem_object_init(obj, &intel_vgpu_gem_ops);
+
+	obj->base.read_domains = I915_GEM_DOMAIN_GTT;
+	obj->base.write_domain = 0;
+	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
+		unsigned int tiling_mode = 0;
+		unsigned int stride = 0;
+
+		switch (info->drm_format_mod << 10) {
+		case PLANE_CTL_TILED_LINEAR:
+			tiling_mode = I915_TILING_NONE;
+			break;
+		case PLANE_CTL_TILED_X:
+			tiling_mode = I915_TILING_X;
+			stride = info->stride;
+			break;
+		case PLANE_CTL_TILED_Y:
+			tiling_mode = I915_TILING_Y;
+			stride = info->stride;
+			break;
+		default:
+			gvt_dbg_core("not supported tiling mode\n");
+		}
+		obj->tiling_and_stride = tiling_mode | stride;
+	} else {
+		obj->tiling_and_stride = info->drm_format_mod ?
+					I915_TILING_X : 0;
+	}
+
+	return obj;
+}
+
+static int vgpu_get_plane_info(struct drm_device *dev,
+		struct intel_vgpu *vgpu,
+		struct intel_vgpu_fb_info *info,
+		int plane_id)
+{
+	struct drm_i915_private *dev_priv = to_i915(dev);
+	struct intel_vgpu_primary_plane_format p;
+	struct intel_vgpu_cursor_plane_format c;
+	int ret;
+
+	if (plane_id == DRM_PLANE_TYPE_PRIMARY) {
+		ret = intel_vgpu_decode_primary_plane(vgpu, &p);
+		if (ret)
+			return ret;
+		info->start = p.base;
+		info->start_gpa = p.base_gpa;
+		info->width = p.width;
+		info->height = p.height;
+		info->stride = p.stride;
+		info->drm_format = p.drm_format;
+		info->drm_format_mod = p.tiled;
+		info->size = (((p.stride * p.height * p.bpp) / 8) +
+				(PAGE_SIZE - 1)) >> PAGE_SHIFT;
+	} else if (plane_id == DRM_PLANE_TYPE_CURSOR) {
+		ret = intel_vgpu_decode_cursor_plane(vgpu, &c);
+		if (ret)
+			return ret;
+		info->start = c.base;
+		info->start_gpa = c.base_gpa;
+		info->width = c.width;
+		info->height = c.height;
+		info->stride = c.width * (c.bpp / 8);
+		info->drm_format = c.drm_format;
+		info->drm_format_mod = 0;
+		info->x_pos = c.x_pos;
+		info->y_pos = c.y_pos;
+
+		/* The invalid cursor hotspot value is delivered to host
+		 * until we find a way to get the cursor hotspot info of
+		 * guest OS.
+		 */
+		info->x_hot = UINT_MAX;
+		info->y_hot = UINT_MAX;
+		info->size = (((info->stride * c.height * c.bpp) / 8)
+				+ (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+	} else {
+		gvt_vgpu_err("invalid plane id:%d\n", plane_id);
+		return -EINVAL;
+	}
+
+	if (info->size == 0) {
+		gvt_vgpu_err("fb size is zero\n");
+		return -EINVAL;
+	}
+
+	if (info->start & (PAGE_SIZE - 1)) {
+		gvt_vgpu_err("Not aligned fb address:0x%llx\n", info->start);
+		return -EFAULT;
+	}
+	if (((info->start >> PAGE_SHIFT) + info->size) >
+		ggtt_total_entries(&dev_priv->ggtt)) {
+		gvt_vgpu_err("Invalid GTT offset or size\n");
+		return -EFAULT;
+	}
+
+	if (!intel_gvt_ggtt_validate_range(vgpu, info->start, info->size)) {
+		gvt_vgpu_err("invalid gma addr\n");
+		return -EFAULT;
+	}
+
+	return 0;
+}
+
+static struct intel_vgpu_dmabuf_obj *
+pick_dmabuf_by_info(struct intel_vgpu *vgpu,
+		    struct intel_vgpu_fb_info *latest_info)
+{
+	struct list_head *pos;
+	struct intel_vgpu_fb_info *fb_info;
+	struct intel_vgpu_dmabuf_obj *dmabuf_obj = NULL;
+	struct intel_vgpu_dmabuf_obj *ret = NULL;
+
+	list_for_each(pos, &vgpu->dmabuf_obj_list_head) {
+		dmabuf_obj = container_of(pos, struct intel_vgpu_dmabuf_obj,
+						list);
+		if ((dmabuf_obj == NULL) ||
+		    (dmabuf_obj->info == NULL))
+			continue;
+
+		fb_info = (struct intel_vgpu_fb_info *)dmabuf_obj->info;
+		if ((fb_info->start == latest_info->start) &&
+		    (fb_info->start_gpa == latest_info->start_gpa) &&
+		    (fb_info->size == latest_info->size) &&
+		    (fb_info->drm_format_mod == latest_info->drm_format_mod) &&
+		    (fb_info->drm_format == latest_info->drm_format) &&
+		    (fb_info->width == latest_info->width) &&
+		    (fb_info->height == latest_info->height)) {
+			ret = dmabuf_obj;
+			break;
+		}
+	}
+
+	return ret;
+}
+
+static struct intel_vgpu_dmabuf_obj *
+pick_dmabuf_by_num(struct intel_vgpu *vgpu, u32 id)
+{
+	struct list_head *pos;
+	struct intel_vgpu_dmabuf_obj *dmabuf_obj = NULL;
+	struct intel_vgpu_dmabuf_obj *ret = NULL;
+
+	list_for_each(pos, &vgpu->dmabuf_obj_list_head) {
+		dmabuf_obj = container_of(pos, struct intel_vgpu_dmabuf_obj,
+						list);
+		if (!dmabuf_obj)
+			continue;
+
+		if (dmabuf_obj->dmabuf_id == id) {
+			ret = dmabuf_obj;
+			break;
+		}
+	}
+
+	return ret;
+}
+
+static void update_fb_info(struct vfio_device_gfx_plane_info *gvt_dmabuf,
+		      struct intel_vgpu_fb_info *fb_info)
+{
+	gvt_dmabuf->drm_format = fb_info->drm_format;
+	gvt_dmabuf->width = fb_info->width;
+	gvt_dmabuf->height = fb_info->height;
+	gvt_dmabuf->stride = fb_info->stride;
+	gvt_dmabuf->size = fb_info->size;
+	gvt_dmabuf->x_pos = fb_info->x_pos;
+	gvt_dmabuf->y_pos = fb_info->y_pos;
+	gvt_dmabuf->x_hot = fb_info->x_hot;
+	gvt_dmabuf->y_hot = fb_info->y_hot;
+}
+
+int intel_vgpu_query_plane(struct intel_vgpu *vgpu, void *args)
+{
+	struct drm_device *dev = &vgpu->gvt->dev_priv->drm;
+	struct vfio_device_gfx_plane_info *gfx_plane_info = args;
+	struct intel_vgpu_dmabuf_obj *dmabuf_obj;
+	struct intel_vgpu_fb_info fb_info;
+	int ret = 0;
+
+	if (gfx_plane_info->flags == (VFIO_GFX_PLANE_TYPE_DMABUF |
+				       VFIO_GFX_PLANE_TYPE_PROBE))
+		return ret;
+	else if ((gfx_plane_info->flags & ~VFIO_GFX_PLANE_TYPE_DMABUF) ||
+			(!gfx_plane_info->flags))
+		return -EINVAL;
+
+	ret = vgpu_get_plane_info(dev, vgpu, &fb_info,
+					gfx_plane_info->drm_plane_type);
+	if (ret != 0)
+		goto out;
+
+	mutex_lock(&vgpu->dmabuf_lock);
+	/* If exists, pick up the exposed dmabuf_obj */
+	dmabuf_obj = pick_dmabuf_by_info(vgpu, &fb_info);
+	if (dmabuf_obj) {
+		update_fb_info(gfx_plane_info, &fb_info);
+		gfx_plane_info->dmabuf_id = dmabuf_obj->dmabuf_id;
+
+		/* This buffer may be released between query_plane ioctl and
+		 * get_dmabuf ioctl. Add the refcount to make sure it won't
+		 * be released between the two ioctls.
+		 */
+		if (!dmabuf_obj->initref) {
+			dmabuf_obj->initref = true;
+			dmabuf_obj_get(dmabuf_obj);
+		}
+		ret = 0;
+		gvt_dbg_dpy("vgpu%d: re-use dmabuf_obj ref %d, id %d\n",
+			    vgpu->id, kref_read(&dmabuf_obj->kref),
+			    gfx_plane_info->dmabuf_id);
+		mutex_unlock(&vgpu->dmabuf_lock);
+		goto out;
+	}
+
+	mutex_unlock(&vgpu->dmabuf_lock);
+
+	/* Need to allocate a new one*/
+	dmabuf_obj = kmalloc(sizeof(struct intel_vgpu_dmabuf_obj), GFP_KERNEL);
+	if (unlikely(!dmabuf_obj)) {
+		gvt_vgpu_err("alloc dmabuf_obj failed\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	dmabuf_obj->info = kmalloc(sizeof(struct intel_vgpu_fb_info),
+				   GFP_KERNEL);
+	if (unlikely(!dmabuf_obj->info)) {
+		gvt_vgpu_err("allocate intel vgpu fb info failed\n");
+		ret = -ENOMEM;
+		goto out_free_dmabuf;
+	}
+	memcpy(dmabuf_obj->info, &fb_info, sizeof(struct intel_vgpu_fb_info));
+
+	((struct intel_vgpu_fb_info *)dmabuf_obj->info)->obj = dmabuf_obj;
+
+	dmabuf_obj->vgpu = vgpu;
+
+	ret = idr_alloc(&vgpu->object_idr, dmabuf_obj, 1, 0, GFP_NOWAIT);
+	if (ret < 0)
+		goto out_free_info;
+	gfx_plane_info->dmabuf_id = ret;
+	dmabuf_obj->dmabuf_id = ret;
+
+	dmabuf_obj->initref = true;
+
+	kref_init(&dmabuf_obj->kref);
+
+	mutex_lock(&vgpu->dmabuf_lock);
+	if (intel_gvt_hypervisor_get_vfio_device(vgpu)) {
+		gvt_vgpu_err("get vfio device failed\n");
+		mutex_unlock(&vgpu->dmabuf_lock);
+		goto out_free_info;
+	}
+	mutex_unlock(&vgpu->dmabuf_lock);
+
+	update_fb_info(gfx_plane_info, &fb_info);
+
+	INIT_LIST_HEAD(&dmabuf_obj->list);
+	mutex_lock(&vgpu->dmabuf_lock);
+	list_add_tail(&dmabuf_obj->list, &vgpu->dmabuf_obj_list_head);
+	mutex_unlock(&vgpu->dmabuf_lock);
+
+	gvt_dbg_dpy("vgpu%d: %s new dmabuf_obj ref %d, id %d\n", vgpu->id,
+		    __func__, kref_read(&dmabuf_obj->kref), ret);
+
+	return 0;
+
+out_free_info:
+	kfree(dmabuf_obj->info);
+out_free_dmabuf:
+	kfree(dmabuf_obj);
+out:
+	/* ENODEV means plane isn't ready, which might be a normal case. */
+	return (ret == -ENODEV) ? 0 : ret;
+}
+
+/* To associate an exposed dmabuf with the dmabuf_obj */
+int intel_vgpu_get_dmabuf(struct intel_vgpu *vgpu, unsigned int dmabuf_id)
+{
+	struct drm_device *dev = &vgpu->gvt->dev_priv->drm;
+	struct intel_vgpu_dmabuf_obj *dmabuf_obj;
+	struct drm_i915_gem_object *obj;
+	struct dma_buf *dmabuf;
+	int dmabuf_fd;
+	int ret = 0;
+
+	mutex_lock(&vgpu->dmabuf_lock);
+
+	dmabuf_obj = pick_dmabuf_by_num(vgpu, dmabuf_id);
+	if (dmabuf_obj == NULL) {
+		gvt_vgpu_err("invalid dmabuf id:%d\n", dmabuf_id);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	obj = vgpu_create_gem(dev, dmabuf_obj->info);
+	if (obj == NULL) {
+		gvt_vgpu_err("create gvt gem obj failed:%d\n", vgpu->id);
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	obj->gvt_info = dmabuf_obj->info;
+
+	dmabuf = i915_gem_prime_export(dev, &obj->base, DRM_CLOEXEC | DRM_RDWR);
+	if (IS_ERR(dmabuf)) {
+		gvt_vgpu_err("export dma-buf failed\n");
+		ret = PTR_ERR(dmabuf);
+		goto out_free_gem;
+	}
+	obj->base.dma_buf = dmabuf;
+
+	i915_gem_object_put(obj);
+
+	ret = dma_buf_fd(dmabuf, DRM_CLOEXEC | DRM_RDWR);
+	if (ret < 0) {
+		gvt_vgpu_err("create dma-buf fd failed ret:%d\n", ret);
+		goto out_free_dmabuf;
+	}
+	dmabuf_fd = ret;
+
+	dmabuf_obj_get(dmabuf_obj);
+
+	if (dmabuf_obj->initref) {
+		dmabuf_obj->initref = false;
+		dmabuf_obj_put(dmabuf_obj);
+	}
+
+	mutex_unlock(&vgpu->dmabuf_lock);
+
+	gvt_dbg_dpy("vgpu%d: dmabuf:%d, dmabuf ref %d, fd:%d\n"
+		    "        file count: %ld, GEM ref: %d\n",
+		    vgpu->id, dmabuf_obj->dmabuf_id,
+		    kref_read(&dmabuf_obj->kref),
+		    dmabuf_fd,
+		    file_count(dmabuf->file),
+		    kref_read(&obj->base.refcount));
+
+	return dmabuf_fd;
+
+out_free_dmabuf:
+	dma_buf_put(dmabuf);
+out_free_gem:
+	i915_gem_object_put(obj);
+out:
+	mutex_unlock(&vgpu->dmabuf_lock);
+	return ret;
+}
+
+void intel_vgpu_dmabuf_cleanup(struct intel_vgpu *vgpu)
+{
+	struct list_head *pos, *n;
+	struct intel_vgpu_dmabuf_obj *dmabuf_obj;
+
+	mutex_lock(&vgpu->dmabuf_lock);
+	list_for_each_safe(pos, n, &vgpu->dmabuf_obj_list_head) {
+		dmabuf_obj = container_of(pos, struct intel_vgpu_dmabuf_obj,
+						list);
+		dmabuf_obj->vgpu = NULL;
+
+		idr_remove(&vgpu->object_idr, dmabuf_obj->dmabuf_id);
+		intel_gvt_hypervisor_put_vfio_device(vgpu);
+		list_del(pos);
+
+		/* dmabuf_obj might be freed in dmabuf_obj_put */
+		if (dmabuf_obj->initref) {
+			dmabuf_obj->initref = false;
+			dmabuf_obj_put(dmabuf_obj);
+		}
+
+	}
+	mutex_unlock(&vgpu->dmabuf_lock);
+}
diff --git a/drivers/gpu/drm/i915/gvt/dmabuf.h b/drivers/gpu/drm/i915/gvt/dmabuf.h
new file mode 100644
index 000000000000..5f8f03fb1d1b
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/dmabuf.h
@@ -0,0 +1,67 @@
+/*
+ * Copyright(c) 2017 Intel Corporation. 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, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Authors:
+ *    Zhiyuan Lv <zhiyuan.lv@intel.com>
+ *
+ * Contributors:
+ *    Xiaoguang Chen
+ *    Tina Zhang <tina.zhang@intel.com>
+ */
+
+#ifndef _GVT_DMABUF_H_
+#define _GVT_DMABUF_H_
+#include <linux/vfio.h>
+
+struct intel_vgpu_fb_info {
+	__u64 start;
+	__u64 start_gpa;
+	__u64 drm_format_mod;
+	__u32 drm_format;	/* drm format of plane */
+	__u32 width;	/* width of plane */
+	__u32 height;	/* height of plane */
+	__u32 stride;	/* stride of plane */
+	__u32 size;	/* size of plane in bytes, align on page */
+	__u32 x_pos;	/* horizontal position of cursor plane */
+	__u32 y_pos;	/* vertical position of cursor plane */
+	__u32 x_hot;    /* horizontal position of cursor hotspot */
+	__u32 y_hot;    /* vertical position of cursor hotspot */
+	struct intel_vgpu_dmabuf_obj *obj;
+};
+
+/**
+ * struct intel_vgpu_dmabuf_obj- Intel vGPU device buffer object
+ */
+struct intel_vgpu_dmabuf_obj {
+	struct intel_vgpu *vgpu;
+	struct intel_vgpu_fb_info *info;
+	__u32 dmabuf_id;
+	struct kref kref;
+	bool initref;
+	struct list_head list;
+};
+
+int intel_vgpu_query_plane(struct intel_vgpu *vgpu, void *args);
+int intel_vgpu_get_dmabuf(struct intel_vgpu *vgpu, unsigned int dmabuf_id);
+void intel_vgpu_dmabuf_cleanup(struct intel_vgpu *vgpu);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/edid.c b/drivers/gpu/drm/i915/gvt/edid.c
index 42cd09ec63fa..f61337632969 100644
--- a/drivers/gpu/drm/i915/gvt/edid.c
+++ b/drivers/gpu/drm/i915/gvt/edid.c
@@ -95,9 +95,9 @@ static inline int get_port_from_gmbus0(u32 gmbus0)
 
 static void reset_gmbus_controller(struct intel_vgpu *vgpu)
 {
-	vgpu_vreg(vgpu, PCH_GMBUS2) = GMBUS_HW_RDY;
+	vgpu_vreg_t(vgpu, PCH_GMBUS2) = GMBUS_HW_RDY;
 	if (!vgpu->display.i2c_edid.edid_available)
-		vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_SATOER;
+		vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_SATOER;
 	vgpu->display.i2c_edid.gmbus.phase = GMBUS_IDLE_PHASE;
 }
 
@@ -123,16 +123,16 @@ static int gmbus0_mmio_write(struct intel_vgpu *vgpu,
 	vgpu->display.i2c_edid.state = I2C_GMBUS;
 	vgpu->display.i2c_edid.gmbus.phase = GMBUS_IDLE_PHASE;
 
-	vgpu_vreg(vgpu, PCH_GMBUS2) &= ~GMBUS_ACTIVE;
-	vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_HW_RDY | GMBUS_HW_WAIT_PHASE;
+	vgpu_vreg_t(vgpu, PCH_GMBUS2) &= ~GMBUS_ACTIVE;
+	vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_HW_RDY | GMBUS_HW_WAIT_PHASE;
 
 	if (intel_vgpu_has_monitor_on_port(vgpu, port) &&
 			!intel_vgpu_port_is_dp(vgpu, port)) {
 		vgpu->display.i2c_edid.port = port;
 		vgpu->display.i2c_edid.edid_available = true;
-		vgpu_vreg(vgpu, PCH_GMBUS2) &= ~GMBUS_SATOER;
+		vgpu_vreg_t(vgpu, PCH_GMBUS2) &= ~GMBUS_SATOER;
 	} else
-		vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_SATOER;
+		vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_SATOER;
 	return 0;
 }
 
@@ -159,8 +159,8 @@ static int gmbus1_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 		 * 2) HW_RDY bit asserted
 		 */
 		if (wvalue & GMBUS_SW_CLR_INT) {
-			vgpu_vreg(vgpu, PCH_GMBUS2) &= ~GMBUS_INT;
-			vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_HW_RDY;
+			vgpu_vreg_t(vgpu, PCH_GMBUS2) &= ~GMBUS_INT;
+			vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_HW_RDY;
 		}
 
 		/* For virtualization, we suppose that HW is always ready,
@@ -208,7 +208,7 @@ static int gmbus1_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 				 * visible in gmbus interface)
 				 */
 				i2c_edid->gmbus.phase = GMBUS_IDLE_PHASE;
-				vgpu_vreg(vgpu, PCH_GMBUS2) &= ~GMBUS_ACTIVE;
+				vgpu_vreg_t(vgpu, PCH_GMBUS2) &= ~GMBUS_ACTIVE;
 			}
 			break;
 		case NIDX_NS_W:
@@ -220,7 +220,7 @@ static int gmbus1_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 			 * START (-->INDEX) -->DATA
 			 */
 			i2c_edid->gmbus.phase = GMBUS_DATA_PHASE;
-			vgpu_vreg(vgpu, PCH_GMBUS2) |= GMBUS_ACTIVE;
+			vgpu_vreg_t(vgpu, PCH_GMBUS2) |= GMBUS_ACTIVE;
 			break;
 		default:
 			gvt_vgpu_err("Unknown/reserved GMBUS cycle detected!\n");
@@ -256,7 +256,7 @@ static int gmbus3_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
 	u32 reg_data = 0;
 
 	/* Data can only be recevied if previous settings correct */
-	if (vgpu_vreg(vgpu, PCH_GMBUS1) & GMBUS_SLAVE_READ) {
+	if (vgpu_vreg_t(vgpu, PCH_GMBUS1) & GMBUS_SLAVE_READ) {
 		if (byte_left <= 0) {
 			memcpy(p_data, &vgpu_vreg(vgpu, offset), bytes);
 			return 0;
diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c
index 940cdaaa3f24..769c1c24ae75 100644
--- a/drivers/gpu/drm/i915/gvt/execlist.c
+++ b/drivers/gpu/drm/i915/gvt/execlist.c
@@ -46,8 +46,6 @@
 #define same_context(a, b) (((a)->context_id == (b)->context_id) && \
 		((a)->lrca == (b)->lrca))
 
-static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask);
-
 static int context_switch_events[] = {
 	[RCS] = RCS_AS_CONTEXT_SWITCH,
 	[BCS] = BCS_AS_CONTEXT_SWITCH,
@@ -135,6 +133,8 @@ static void emulate_csb_update(struct intel_vgpu_execlist *execlist,
 	struct execlist_context_status_pointer_format ctx_status_ptr;
 	u32 write_pointer;
 	u32 ctx_status_ptr_reg, ctx_status_buf_reg, offset;
+	unsigned long hwsp_gpa;
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
 
 	ctx_status_ptr_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
 			_EL_OFFSET_STATUS_PTR);
@@ -160,6 +160,20 @@ static void emulate_csb_update(struct intel_vgpu_execlist *execlist,
 	ctx_status_ptr.write_ptr = write_pointer;
 	vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw;
 
+	/* Update the CSB and CSB write pointer in HWSP */
+	hwsp_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
+					 vgpu->hws_pga[ring_id]);
+	if (hwsp_gpa != INTEL_GVT_INVALID_ADDR) {
+		intel_gvt_hypervisor_write_gpa(vgpu,
+			hwsp_gpa + I915_HWS_CSB_BUF0_INDEX * 4 +
+			write_pointer * 8,
+			status, 8);
+		intel_gvt_hypervisor_write_gpa(vgpu,
+			hwsp_gpa +
+			intel_hws_csb_write_index(dev_priv) * 4,
+			&write_pointer, 4);
+	}
+
 	gvt_dbg_el("vgpu%d: w pointer %u reg %x csb l %x csb h %x\n",
 		vgpu->id, write_pointer, offset, status->ldw, status->udw);
 
@@ -358,218 +372,47 @@ static int emulate_execlist_schedule_in(struct intel_vgpu_execlist *execlist,
 	return 0;
 }
 
-static void free_workload(struct intel_vgpu_workload *workload)
-{
-	intel_vgpu_unpin_mm(workload->shadow_mm);
-	intel_gvt_mm_unreference(workload->shadow_mm);
-	kmem_cache_free(workload->vgpu->workloads, workload);
-}
-
 #define get_desc_from_elsp_dwords(ed, i) \
 	((struct execlist_ctx_descriptor_format *)&((ed)->data[i * 2]))
 
-static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
-{
-	const int gmadr_bytes = workload->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
-	struct intel_shadow_bb_entry *entry_obj;
-
-	/* pin the gem object to ggtt */
-	list_for_each_entry(entry_obj, &workload->shadow_bb, list) {
-		struct i915_vma *vma;
-
-		vma = i915_gem_object_ggtt_pin(entry_obj->obj, NULL, 0, 4, 0);
-		if (IS_ERR(vma)) {
-			return PTR_ERR(vma);
-		}
-
-		/* FIXME: we are not tracking our pinned VMA leaving it
-		 * up to the core to fix up the stray pin_count upon
-		 * free.
-		 */
-
-		/* update the relocate gma with shadow batch buffer*/
-		entry_obj->bb_start_cmd_va[1] = i915_ggtt_offset(vma);
-		if (gmadr_bytes == 8)
-			entry_obj->bb_start_cmd_va[2] = 0;
-	}
-	return 0;
-}
-
-static int update_wa_ctx_2_shadow_ctx(struct intel_shadow_wa_ctx *wa_ctx)
-{
-	struct intel_vgpu_workload *workload = container_of(wa_ctx,
-					struct intel_vgpu_workload,
-					wa_ctx);
-	int ring_id = workload->ring_id;
-	struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
-	struct drm_i915_gem_object *ctx_obj =
-		shadow_ctx->engine[ring_id].state->obj;
-	struct execlist_ring_context *shadow_ring_context;
-	struct page *page;
-
-	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
-	shadow_ring_context = kmap_atomic(page);
-
-	shadow_ring_context->bb_per_ctx_ptr.val =
-		(shadow_ring_context->bb_per_ctx_ptr.val &
-		(~PER_CTX_ADDR_MASK)) | wa_ctx->per_ctx.shadow_gma;
-	shadow_ring_context->rcs_indirect_ctx.val =
-		(shadow_ring_context->rcs_indirect_ctx.val &
-		(~INDIRECT_CTX_ADDR_MASK)) | wa_ctx->indirect_ctx.shadow_gma;
-
-	kunmap_atomic(shadow_ring_context);
-	return 0;
-}
-
-static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
-{
-	struct i915_vma *vma;
-	unsigned char *per_ctx_va =
-		(unsigned char *)wa_ctx->indirect_ctx.shadow_va +
-		wa_ctx->indirect_ctx.size;
-
-	if (wa_ctx->indirect_ctx.size == 0)
-		return 0;
-
-	vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL,
-				       0, CACHELINE_BYTES, 0);
-	if (IS_ERR(vma)) {
-		return PTR_ERR(vma);
-	}
-
-	/* FIXME: we are not tracking our pinned VMA leaving it
-	 * up to the core to fix up the stray pin_count upon
-	 * free.
-	 */
-
-	wa_ctx->indirect_ctx.shadow_gma = i915_ggtt_offset(vma);
-
-	wa_ctx->per_ctx.shadow_gma = *((unsigned int *)per_ctx_va + 1);
-	memset(per_ctx_va, 0, CACHELINE_BYTES);
-
-	update_wa_ctx_2_shadow_ctx(wa_ctx);
-	return 0;
-}
-
-static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
-{
-	/* release all the shadow batch buffer */
-	if (!list_empty(&workload->shadow_bb)) {
-		struct intel_shadow_bb_entry *entry_obj =
-			list_first_entry(&workload->shadow_bb,
-					 struct intel_shadow_bb_entry,
-					 list);
-		struct intel_shadow_bb_entry *temp;
-
-		list_for_each_entry_safe(entry_obj, temp, &workload->shadow_bb,
-					 list) {
-			i915_gem_object_unpin_map(entry_obj->obj);
-			i915_gem_object_put(entry_obj->obj);
-			list_del(&entry_obj->list);
-			kfree(entry_obj);
-		}
-	}
-}
-
 static int prepare_execlist_workload(struct intel_vgpu_workload *workload)
 {
 	struct intel_vgpu *vgpu = workload->vgpu;
+	struct intel_vgpu_submission *s = &vgpu->submission;
 	struct execlist_ctx_descriptor_format ctx[2];
 	int ring_id = workload->ring_id;
 	int ret;
 
-	ret = intel_vgpu_pin_mm(workload->shadow_mm);
-	if (ret) {
-		gvt_vgpu_err("fail to vgpu pin mm\n");
-		goto out;
-	}
-
-	ret = intel_vgpu_sync_oos_pages(workload->vgpu);
-	if (ret) {
-		gvt_vgpu_err("fail to vgpu sync oos pages\n");
-		goto err_unpin_mm;
-	}
-
-	ret = intel_vgpu_flush_post_shadow(workload->vgpu);
-	if (ret) {
-		gvt_vgpu_err("fail to flush post shadow\n");
-		goto err_unpin_mm;
-	}
-
-	ret = intel_gvt_generate_request(workload);
-	if (ret) {
-		gvt_vgpu_err("fail to generate request\n");
-		goto err_unpin_mm;
-	}
-
-	ret = prepare_shadow_batch_buffer(workload);
-	if (ret) {
-		gvt_vgpu_err("fail to prepare_shadow_batch_buffer\n");
-		goto err_unpin_mm;
-	}
-
-	ret = prepare_shadow_wa_ctx(&workload->wa_ctx);
-	if (ret) {
-		gvt_vgpu_err("fail to prepare_shadow_wa_ctx\n");
-		goto err_shadow_batch;
-	}
-
 	if (!workload->emulate_schedule_in)
 		return 0;
 
-	ctx[0] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1);
-	ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 0);
+	ctx[0] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 0);
+	ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1);
 
-	ret = emulate_execlist_schedule_in(&vgpu->execlist[ring_id], ctx);
-	if (!ret)
-		goto out;
-	else
+	ret = emulate_execlist_schedule_in(&s->execlist[ring_id], ctx);
+	if (ret) {
 		gvt_vgpu_err("fail to emulate execlist schedule in\n");
-
-	release_shadow_wa_ctx(&workload->wa_ctx);
-err_shadow_batch:
-	release_shadow_batch_buffer(workload);
-err_unpin_mm:
-	intel_vgpu_unpin_mm(workload->shadow_mm);
-out:
-	return ret;
+		return ret;
+	}
+	return 0;
 }
 
 static int complete_execlist_workload(struct intel_vgpu_workload *workload)
 {
 	struct intel_vgpu *vgpu = workload->vgpu;
 	int ring_id = workload->ring_id;
-	struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id];
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct intel_vgpu_execlist *execlist = &s->execlist[ring_id];
 	struct intel_vgpu_workload *next_workload;
 	struct list_head *next = workload_q_head(vgpu, ring_id)->next;
 	bool lite_restore = false;
-	int ret;
+	int ret = 0;
 
 	gvt_dbg_el("complete workload %p status %d\n", workload,
 			workload->status);
 
-	if (!workload->status) {
-		release_shadow_batch_buffer(workload);
-		release_shadow_wa_ctx(&workload->wa_ctx);
-	}
-
-	if (workload->status || (vgpu->resetting_eng & ENGINE_MASK(ring_id))) {
-		/* if workload->status is not successful means HW GPU
-		 * has occurred GPU hang or something wrong with i915/GVT,
-		 * and GVT won't inject context switch interrupt to guest.
-		 * So this error is a vGPU hang actually to the guest.
-		 * According to this we should emunlate a vGPU hang. If
-		 * there are pending workloads which are already submitted
-		 * from guest, we should clean them up like HW GPU does.
-		 *
-		 * if it is in middle of engine resetting, the pending
-		 * workloads won't be submitted to HW GPU and will be
-		 * cleaned up during the resetting process later, so doing
-		 * the workload clean up here doesn't have any impact.
-		 **/
-		clean_workloads(vgpu, ENGINE_MASK(ring_id));
+	if (workload->status || (vgpu->resetting_eng & ENGINE_MASK(ring_id)))
 		goto out;
-	}
 
 	if (!list_empty(workload_q_head(vgpu, ring_id))) {
 		struct execlist_ctx_descriptor_format *this_desc, *next_desc;
@@ -584,213 +427,60 @@ static int complete_execlist_workload(struct intel_vgpu_workload *workload)
 
 	if (lite_restore) {
 		gvt_dbg_el("next context == current - no schedule-out\n");
-		free_workload(workload);
-		return 0;
+		goto out;
 	}
 
 	ret = emulate_execlist_ctx_schedule_out(execlist, &workload->ctx_desc);
-	if (ret)
-		goto err;
 out:
-	free_workload(workload);
-	return 0;
-err:
-	free_workload(workload);
+	intel_vgpu_unpin_mm(workload->shadow_mm);
+	intel_vgpu_destroy_workload(workload);
 	return ret;
 }
 
-#define RING_CTX_OFF(x) \
-	offsetof(struct execlist_ring_context, x)
-
-static void read_guest_pdps(struct intel_vgpu *vgpu,
-		u64 ring_context_gpa, u32 pdp[8])
-{
-	u64 gpa;
-	int i;
-
-	gpa = ring_context_gpa + RING_CTX_OFF(pdp3_UDW.val);
-
-	for (i = 0; i < 8; i++)
-		intel_gvt_hypervisor_read_gpa(vgpu,
-				gpa + i * 8, &pdp[7 - i], 4);
-}
-
-static int prepare_mm(struct intel_vgpu_workload *workload)
-{
-	struct execlist_ctx_descriptor_format *desc = &workload->ctx_desc;
-	struct intel_vgpu_mm *mm;
-	struct intel_vgpu *vgpu = workload->vgpu;
-	int page_table_level;
-	u32 pdp[8];
-
-	if (desc->addressing_mode == 1) { /* legacy 32-bit */
-		page_table_level = 3;
-	} else if (desc->addressing_mode == 3) { /* legacy 64 bit */
-		page_table_level = 4;
-	} else {
-		gvt_vgpu_err("Advanced Context mode(SVM) is not supported!\n");
-		return -EINVAL;
-	}
-
-	read_guest_pdps(workload->vgpu, workload->ring_context_gpa, pdp);
-
-	mm = intel_vgpu_find_ppgtt_mm(workload->vgpu, page_table_level, pdp);
-	if (mm) {
-		intel_gvt_mm_reference(mm);
-	} else {
-
-		mm = intel_vgpu_create_mm(workload->vgpu, INTEL_GVT_MM_PPGTT,
-				pdp, page_table_level, 0);
-		if (IS_ERR(mm)) {
-			gvt_vgpu_err("fail to create mm object.\n");
-			return PTR_ERR(mm);
-		}
-	}
-	workload->shadow_mm = mm;
-	return 0;
-}
-
-#define get_last_workload(q) \
-	(list_empty(q) ? NULL : container_of(q->prev, \
-	struct intel_vgpu_workload, list))
-
 static int submit_context(struct intel_vgpu *vgpu, int ring_id,
 		struct execlist_ctx_descriptor_format *desc,
 		bool emulate_schedule_in)
 {
-	struct list_head *q = workload_q_head(vgpu, ring_id);
-	struct intel_vgpu_workload *last_workload = get_last_workload(q);
+	struct intel_vgpu_submission *s = &vgpu->submission;
 	struct intel_vgpu_workload *workload = NULL;
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	u64 ring_context_gpa;
-	u32 head, tail, start, ctl, ctx_ctl, per_ctx, indirect_ctx;
-	int ret;
-
-	ring_context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
-			(u32)((desc->lrca + 1) << GTT_PAGE_SHIFT));
-	if (ring_context_gpa == INTEL_GVT_INVALID_ADDR) {
-		gvt_vgpu_err("invalid guest context LRCA: %x\n", desc->lrca);
-		return -EINVAL;
-	}
-
-	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
-			RING_CTX_OFF(ring_header.val), &head, 4);
 
-	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
-			RING_CTX_OFF(ring_tail.val), &tail, 4);
+	workload = intel_vgpu_create_workload(vgpu, ring_id, desc);
+	if (IS_ERR(workload))
+		return PTR_ERR(workload);
 
-	head &= RB_HEAD_OFF_MASK;
-	tail &= RB_TAIL_OFF_MASK;
-
-	if (last_workload && same_context(&last_workload->ctx_desc, desc)) {
-		gvt_dbg_el("ring id %d cur workload == last\n", ring_id);
-		gvt_dbg_el("ctx head %x real head %lx\n", head,
-				last_workload->rb_tail);
-		/*
-		 * cannot use guest context head pointer here,
-		 * as it might not be updated at this time
-		 */
-		head = last_workload->rb_tail;
-	}
-
-	gvt_dbg_el("ring id %d begin a new workload\n", ring_id);
-
-	workload = kmem_cache_zalloc(vgpu->workloads, GFP_KERNEL);
-	if (!workload)
-		return -ENOMEM;
-
-	/* record some ring buffer register values for scan and shadow */
-	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
-			RING_CTX_OFF(rb_start.val), &start, 4);
-	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
-			RING_CTX_OFF(rb_ctrl.val), &ctl, 4);
-	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
-			RING_CTX_OFF(ctx_ctrl.val), &ctx_ctl, 4);
-
-	INIT_LIST_HEAD(&workload->list);
-	INIT_LIST_HEAD(&workload->shadow_bb);
-
-	init_waitqueue_head(&workload->shadow_ctx_status_wq);
-	atomic_set(&workload->shadow_ctx_active, 0);
-
-	workload->vgpu = vgpu;
-	workload->ring_id = ring_id;
-	workload->ctx_desc = *desc;
-	workload->ring_context_gpa = ring_context_gpa;
-	workload->rb_head = head;
-	workload->rb_tail = tail;
-	workload->rb_start = start;
-	workload->rb_ctl = ctl;
 	workload->prepare = prepare_execlist_workload;
 	workload->complete = complete_execlist_workload;
-	workload->status = -EINPROGRESS;
 	workload->emulate_schedule_in = emulate_schedule_in;
-	workload->shadowed = false;
-
-	if (ring_id == RCS) {
-		intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
-			RING_CTX_OFF(bb_per_ctx_ptr.val), &per_ctx, 4);
-		intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
-			RING_CTX_OFF(rcs_indirect_ctx.val), &indirect_ctx, 4);
-
-		workload->wa_ctx.indirect_ctx.guest_gma =
-			indirect_ctx & INDIRECT_CTX_ADDR_MASK;
-		workload->wa_ctx.indirect_ctx.size =
-			(indirect_ctx & INDIRECT_CTX_SIZE_MASK) *
-			CACHELINE_BYTES;
-		workload->wa_ctx.per_ctx.guest_gma =
-			per_ctx & PER_CTX_ADDR_MASK;
-		workload->wa_ctx.per_ctx.valid = per_ctx & 1;
-	}
 
 	if (emulate_schedule_in)
-		workload->elsp_dwords = vgpu->execlist[ring_id].elsp_dwords;
-
-	gvt_dbg_el("workload %p ring id %d head %x tail %x start %x ctl %x\n",
-			workload, ring_id, head, tail, start, ctl);
+		workload->elsp_dwords = s->execlist[ring_id].elsp_dwords;
 
 	gvt_dbg_el("workload %p emulate schedule_in %d\n", workload,
 			emulate_schedule_in);
 
-	ret = prepare_mm(workload);
-	if (ret) {
-		kmem_cache_free(vgpu->workloads, workload);
-		return ret;
-	}
-
-	/* Only scan and shadow the first workload in the queue
-	 * as there is only one pre-allocated buf-obj for shadow.
-	 */
-	if (list_empty(workload_q_head(vgpu, ring_id))) {
-		intel_runtime_pm_get(dev_priv);
-		mutex_lock(&dev_priv->drm.struct_mutex);
-		intel_gvt_scan_and_shadow_workload(workload);
-		mutex_unlock(&dev_priv->drm.struct_mutex);
-		intel_runtime_pm_put(dev_priv);
-	}
-
-	queue_workload(workload);
+	intel_vgpu_queue_workload(workload);
 	return 0;
 }
 
 int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id)
 {
-	struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id];
-	struct execlist_ctx_descriptor_format desc[2];
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct intel_vgpu_execlist *execlist = &s->execlist[ring_id];
+	struct execlist_ctx_descriptor_format *desc[2];
 	int i, ret;
 
-	desc[0] = *get_desc_from_elsp_dwords(&execlist->elsp_dwords, 1);
-	desc[1] = *get_desc_from_elsp_dwords(&execlist->elsp_dwords, 0);
+	desc[0] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 0);
+	desc[1] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 1);
 
-	if (!desc[0].valid) {
+	if (!desc[0]->valid) {
 		gvt_vgpu_err("invalid elsp submission, desc0 is invalid\n");
 		goto inv_desc;
 	}
 
 	for (i = 0; i < ARRAY_SIZE(desc); i++) {
-		if (!desc[i].valid)
+		if (!desc[i]->valid)
 			continue;
-		if (!desc[i].privilege_access) {
+		if (!desc[i]->privilege_access) {
 			gvt_vgpu_err("unexpected GGTT elsp submission\n");
 			goto inv_desc;
 		}
@@ -798,9 +488,9 @@ int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id)
 
 	/* submit workload */
 	for (i = 0; i < ARRAY_SIZE(desc); i++) {
-		if (!desc[i].valid)
+		if (!desc[i]->valid)
 			continue;
-		ret = submit_context(vgpu, ring_id, &desc[i], i == 0);
+		ret = submit_context(vgpu, ring_id, desc[i], i == 0);
 		if (ret) {
 			gvt_vgpu_err("failed to submit desc %d\n", i);
 			return ret;
@@ -811,13 +501,14 @@ int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id)
 
 inv_desc:
 	gvt_vgpu_err("descriptors content: desc0 %08x %08x desc1 %08x %08x\n",
-		     desc[0].udw, desc[0].ldw, desc[1].udw, desc[1].ldw);
+		     desc[0]->udw, desc[0]->ldw, desc[1]->udw, desc[1]->ldw);
 	return -EINVAL;
 }
 
 static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id)
 {
-	struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id];
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct intel_vgpu_execlist *execlist = &s->execlist[ring_id];
 	struct execlist_context_status_pointer_format ctx_status_ptr;
 	u32 ctx_status_ptr_reg;
 
@@ -837,91 +528,40 @@ static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id)
 	vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw;
 }
 
-static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask)
-{
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	struct intel_engine_cs *engine;
-	struct intel_vgpu_workload *pos, *n;
-	unsigned int tmp;
-
-	/* free the unsubmited workloads in the queues. */
-	for_each_engine_masked(engine, dev_priv, engine_mask, tmp) {
-		list_for_each_entry_safe(pos, n,
-			&vgpu->workload_q_head[engine->id], list) {
-			list_del_init(&pos->list);
-			free_workload(pos);
-		}
-
-		clear_bit(engine->id, vgpu->shadow_ctx_desc_updated);
-	}
-}
-
-void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu)
+static void clean_execlist(struct intel_vgpu *vgpu)
 {
 	enum intel_engine_id i;
 	struct intel_engine_cs *engine;
 
-	clean_workloads(vgpu, ALL_ENGINES);
-	kmem_cache_destroy(vgpu->workloads);
-
 	for_each_engine(engine, vgpu->gvt->dev_priv, i) {
-		kfree(vgpu->reserve_ring_buffer_va[i]);
-		vgpu->reserve_ring_buffer_va[i] = NULL;
-		vgpu->reserve_ring_buffer_size[i] = 0;
-	}
+		struct intel_vgpu_submission *s = &vgpu->submission;
 
-}
-
-#define RESERVE_RING_BUFFER_SIZE		((1 * PAGE_SIZE)/8)
-int intel_vgpu_init_execlist(struct intel_vgpu *vgpu)
-{
-	enum intel_engine_id i;
-	struct intel_engine_cs *engine;
-
-	/* each ring has a virtual execlist engine */
-	for_each_engine(engine, vgpu->gvt->dev_priv, i) {
-		init_vgpu_execlist(vgpu, i);
-		INIT_LIST_HEAD(&vgpu->workload_q_head[i]);
-	}
-
-	vgpu->workloads = kmem_cache_create("gvt-g_vgpu_workload",
-			sizeof(struct intel_vgpu_workload), 0,
-			SLAB_HWCACHE_ALIGN,
-			NULL);
-
-	if (!vgpu->workloads)
-		return -ENOMEM;
-
-	/* each ring has a shadow ring buffer until vgpu destroyed */
-	for_each_engine(engine, vgpu->gvt->dev_priv, i) {
-		vgpu->reserve_ring_buffer_va[i] =
-			kmalloc(RESERVE_RING_BUFFER_SIZE, GFP_KERNEL);
-		if (!vgpu->reserve_ring_buffer_va[i]) {
-			gvt_vgpu_err("fail to alloc reserve ring buffer\n");
-			goto out;
-		}
-		vgpu->reserve_ring_buffer_size[i] = RESERVE_RING_BUFFER_SIZE;
+		kfree(s->ring_scan_buffer[i]);
+		s->ring_scan_buffer[i] = NULL;
+		s->ring_scan_buffer_size[i] = 0;
 	}
-	return 0;
-out:
-	for_each_engine(engine, vgpu->gvt->dev_priv, i) {
-		if (vgpu->reserve_ring_buffer_size[i]) {
-			kfree(vgpu->reserve_ring_buffer_va[i]);
-			vgpu->reserve_ring_buffer_va[i] = NULL;
-			vgpu->reserve_ring_buffer_size[i] = 0;
-		}
-	}
-	return -ENOMEM;
 }
 
-void intel_vgpu_reset_execlist(struct intel_vgpu *vgpu,
+static void reset_execlist(struct intel_vgpu *vgpu,
 		unsigned long engine_mask)
 {
 	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
 	struct intel_engine_cs *engine;
 	unsigned int tmp;
 
-	clean_workloads(vgpu, engine_mask);
 	for_each_engine_masked(engine, dev_priv, engine_mask, tmp)
 		init_vgpu_execlist(vgpu, engine->id);
 }
+
+static int init_execlist(struct intel_vgpu *vgpu)
+{
+	reset_execlist(vgpu, ALL_ENGINES);
+	return 0;
+}
+
+const struct intel_vgpu_submission_ops intel_vgpu_execlist_submission_ops = {
+	.name = "execlist",
+	.init = init_execlist,
+	.reset = reset_execlist,
+	.clean = clean_execlist,
+};
diff --git a/drivers/gpu/drm/i915/gvt/execlist.h b/drivers/gpu/drm/i915/gvt/execlist.h
index 7eced40a1e30..427e40e64d41 100644
--- a/drivers/gpu/drm/i915/gvt/execlist.h
+++ b/drivers/gpu/drm/i915/gvt/execlist.h
@@ -37,10 +37,6 @@
 
 struct execlist_ctx_descriptor_format {
 	union {
-		u32 udw;
-		u32 context_id;
-	};
-	union {
 		u32 ldw;
 		struct {
 			u32 valid                  : 1;
@@ -54,6 +50,10 @@ struct execlist_ctx_descriptor_format {
 			u32 lrca                   : 20;
 		};
 	};
+	union {
+		u32 udw;
+		u32 context_id;
+	};
 };
 
 struct execlist_status_format {
diff --git a/drivers/gpu/drm/i915/gvt/fb_decoder.c b/drivers/gpu/drm/i915/gvt/fb_decoder.c
new file mode 100644
index 000000000000..6b50fe78dc1b
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/fb_decoder.c
@@ -0,0 +1,514 @@
+/*
+ * Copyright(c) 2011-2016 Intel Corporation. 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, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Authors:
+ *    Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ *    Bing Niu <bing.niu@intel.com>
+ *    Xu Han <xu.han@intel.com>
+ *    Ping Gao <ping.a.gao@intel.com>
+ *    Xiaoguang Chen <xiaoguang.chen@intel.com>
+ *    Yang Liu <yang2.liu@intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *
+ */
+
+#include <uapi/drm/drm_fourcc.h>
+#include "i915_drv.h"
+#include "gvt.h"
+
+#define PRIMARY_FORMAT_NUM	16
+struct pixel_format {
+	int	drm_format;	/* Pixel format in DRM definition */
+	int	bpp;		/* Bits per pixel, 0 indicates invalid */
+	char	*desc;		/* The description */
+};
+
+static struct pixel_format bdw_pixel_formats[] = {
+	{DRM_FORMAT_C8, 8, "8-bit Indexed"},
+	{DRM_FORMAT_RGB565, 16, "16-bit BGRX (5:6:5 MSB-R:G:B)"},
+	{DRM_FORMAT_XRGB8888, 32, "32-bit BGRX (8:8:8:8 MSB-X:R:G:B)"},
+	{DRM_FORMAT_XBGR2101010, 32, "32-bit RGBX (2:10:10:10 MSB-X:B:G:R)"},
+
+	{DRM_FORMAT_XRGB2101010, 32, "32-bit BGRX (2:10:10:10 MSB-X:R:G:B)"},
+	{DRM_FORMAT_XBGR8888, 32, "32-bit RGBX (8:8:8:8 MSB-X:B:G:R)"},
+
+	/* non-supported format has bpp default to 0 */
+	{0, 0, NULL},
+};
+
+static struct pixel_format skl_pixel_formats[] = {
+	{DRM_FORMAT_YUYV, 16, "16-bit packed YUYV (8:8:8:8 MSB-V:Y2:U:Y1)"},
+	{DRM_FORMAT_UYVY, 16, "16-bit packed UYVY (8:8:8:8 MSB-Y2:V:Y1:U)"},
+	{DRM_FORMAT_YVYU, 16, "16-bit packed YVYU (8:8:8:8 MSB-U:Y2:V:Y1)"},
+	{DRM_FORMAT_VYUY, 16, "16-bit packed VYUY (8:8:8:8 MSB-Y2:U:Y1:V)"},
+
+	{DRM_FORMAT_C8, 8, "8-bit Indexed"},
+	{DRM_FORMAT_RGB565, 16, "16-bit BGRX (5:6:5 MSB-R:G:B)"},
+	{DRM_FORMAT_ABGR8888, 32, "32-bit RGBA (8:8:8:8 MSB-A:B:G:R)"},
+	{DRM_FORMAT_XBGR8888, 32, "32-bit RGBX (8:8:8:8 MSB-X:B:G:R)"},
+
+	{DRM_FORMAT_ARGB8888, 32, "32-bit BGRA (8:8:8:8 MSB-A:R:G:B)"},
+	{DRM_FORMAT_XRGB8888, 32, "32-bit BGRX (8:8:8:8 MSB-X:R:G:B)"},
+	{DRM_FORMAT_XBGR2101010, 32, "32-bit RGBX (2:10:10:10 MSB-X:B:G:R)"},
+	{DRM_FORMAT_XRGB2101010, 32, "32-bit BGRX (2:10:10:10 MSB-X:R:G:B)"},
+
+	/* non-supported format has bpp default to 0 */
+	{0, 0, NULL},
+};
+
+static int bdw_format_to_drm(int format)
+{
+	int bdw_pixel_formats_index = 6;
+
+	switch (format) {
+	case DISPPLANE_8BPP:
+		bdw_pixel_formats_index = 0;
+		break;
+	case DISPPLANE_BGRX565:
+		bdw_pixel_formats_index = 1;
+		break;
+	case DISPPLANE_BGRX888:
+		bdw_pixel_formats_index = 2;
+		break;
+	case DISPPLANE_RGBX101010:
+		bdw_pixel_formats_index = 3;
+		break;
+	case DISPPLANE_BGRX101010:
+		bdw_pixel_formats_index = 4;
+		break;
+	case DISPPLANE_RGBX888:
+		bdw_pixel_formats_index = 5;
+		break;
+
+	default:
+		break;
+	}
+
+	return bdw_pixel_formats_index;
+}
+
+static int skl_format_to_drm(int format, bool rgb_order, bool alpha,
+	int yuv_order)
+{
+	int skl_pixel_formats_index = 12;
+
+	switch (format) {
+	case PLANE_CTL_FORMAT_INDEXED:
+		skl_pixel_formats_index = 4;
+		break;
+	case PLANE_CTL_FORMAT_RGB_565:
+		skl_pixel_formats_index = 5;
+		break;
+	case PLANE_CTL_FORMAT_XRGB_8888:
+		if (rgb_order)
+			skl_pixel_formats_index = alpha ? 6 : 7;
+		else
+			skl_pixel_formats_index = alpha ? 8 : 9;
+		break;
+	case PLANE_CTL_FORMAT_XRGB_2101010:
+		skl_pixel_formats_index = rgb_order ? 10 : 11;
+		break;
+	case PLANE_CTL_FORMAT_YUV422:
+		skl_pixel_formats_index = yuv_order >> 16;
+		if (skl_pixel_formats_index > 3)
+			return -EINVAL;
+		break;
+
+	default:
+		break;
+	}
+
+	return skl_pixel_formats_index;
+}
+
+static u32 intel_vgpu_get_stride(struct intel_vgpu *vgpu, int pipe,
+	u32 tiled, int stride_mask, int bpp)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+
+	u32 stride_reg = vgpu_vreg_t(vgpu, DSPSTRIDE(pipe)) & stride_mask;
+	u32 stride = stride_reg;
+
+	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
+		switch (tiled) {
+		case PLANE_CTL_TILED_LINEAR:
+			stride = stride_reg * 64;
+			break;
+		case PLANE_CTL_TILED_X:
+			stride = stride_reg * 512;
+			break;
+		case PLANE_CTL_TILED_Y:
+			stride = stride_reg * 128;
+			break;
+		case PLANE_CTL_TILED_YF:
+			if (bpp == 8)
+				stride = stride_reg * 64;
+			else if (bpp == 16 || bpp == 32 || bpp == 64)
+				stride = stride_reg * 128;
+			else
+				gvt_dbg_core("skl: unsupported bpp:%d\n", bpp);
+			break;
+		default:
+			gvt_dbg_core("skl: unsupported tile format:%x\n",
+				tiled);
+		}
+	}
+
+	return stride;
+}
+
+static int get_active_pipe(struct intel_vgpu *vgpu)
+{
+	int i;
+
+	for (i = 0; i < I915_MAX_PIPES; i++)
+		if (pipe_is_enabled(vgpu, i))
+			break;
+
+	return i;
+}
+
+/**
+ * intel_vgpu_decode_primary_plane - Decode primary plane
+ * @vgpu: input vgpu
+ * @plane: primary plane to save decoded info
+ * This function is called for decoding plane
+ *
+ * Returns:
+ * 0 on success, non-zero if failed.
+ */
+int intel_vgpu_decode_primary_plane(struct intel_vgpu *vgpu,
+	struct intel_vgpu_primary_plane_format *plane)
+{
+	u32 val, fmt;
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	int pipe;
+
+	pipe = get_active_pipe(vgpu);
+	if (pipe >= I915_MAX_PIPES)
+		return -ENODEV;
+
+	val = vgpu_vreg_t(vgpu, DSPCNTR(pipe));
+	plane->enabled = !!(val & DISPLAY_PLANE_ENABLE);
+	if (!plane->enabled)
+		return -ENODEV;
+
+	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
+		plane->tiled = (val & PLANE_CTL_TILED_MASK) >>
+		_PLANE_CTL_TILED_SHIFT;
+		fmt = skl_format_to_drm(
+			val & PLANE_CTL_FORMAT_MASK,
+			val & PLANE_CTL_ORDER_RGBX,
+			val & PLANE_CTL_ALPHA_MASK,
+			val & PLANE_CTL_YUV422_ORDER_MASK);
+
+		if (fmt >= ARRAY_SIZE(skl_pixel_formats)) {
+			gvt_vgpu_err("Out-of-bounds pixel format index\n");
+			return -EINVAL;
+		}
+
+		plane->bpp = skl_pixel_formats[fmt].bpp;
+		plane->drm_format = skl_pixel_formats[fmt].drm_format;
+	} else {
+		plane->tiled = !!(val & DISPPLANE_TILED);
+		fmt = bdw_format_to_drm(val & DISPPLANE_PIXFORMAT_MASK);
+		plane->bpp = bdw_pixel_formats[fmt].bpp;
+		plane->drm_format = bdw_pixel_formats[fmt].drm_format;
+	}
+
+	if (!plane->bpp) {
+		gvt_vgpu_err("Non-supported pixel format (0x%x)\n", fmt);
+		return -EINVAL;
+	}
+
+	plane->hw_format = fmt;
+
+	plane->base = vgpu_vreg_t(vgpu, DSPSURF(pipe)) & I915_GTT_PAGE_MASK;
+	if (!intel_gvt_ggtt_validate_range(vgpu, plane->base, 0)) {
+		gvt_vgpu_err("invalid gma address: %lx\n",
+			     (unsigned long)plane->base);
+		return  -EINVAL;
+	}
+
+	plane->base_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, plane->base);
+	if (plane->base_gpa == INTEL_GVT_INVALID_ADDR) {
+		gvt_vgpu_err("invalid gma address: %lx\n",
+				(unsigned long)plane->base);
+		return  -EINVAL;
+	}
+
+	plane->stride = intel_vgpu_get_stride(vgpu, pipe, (plane->tiled << 10),
+		(IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) ?
+			(_PRI_PLANE_STRIDE_MASK >> 6) :
+				_PRI_PLANE_STRIDE_MASK, plane->bpp);
+
+	plane->width = (vgpu_vreg_t(vgpu, PIPESRC(pipe)) & _PIPE_H_SRCSZ_MASK) >>
+		_PIPE_H_SRCSZ_SHIFT;
+	plane->width += 1;
+	plane->height = (vgpu_vreg_t(vgpu, PIPESRC(pipe)) &
+			_PIPE_V_SRCSZ_MASK) >> _PIPE_V_SRCSZ_SHIFT;
+	plane->height += 1;	/* raw height is one minus the real value */
+
+	val = vgpu_vreg_t(vgpu, DSPTILEOFF(pipe));
+	plane->x_offset = (val & _PRI_PLANE_X_OFF_MASK) >>
+		_PRI_PLANE_X_OFF_SHIFT;
+	plane->y_offset = (val & _PRI_PLANE_Y_OFF_MASK) >>
+		_PRI_PLANE_Y_OFF_SHIFT;
+
+	return 0;
+}
+
+#define CURSOR_FORMAT_NUM	(1 << 6)
+struct cursor_mode_format {
+	int	drm_format;	/* Pixel format in DRM definition */
+	u8	bpp;		/* Bits per pixel; 0 indicates invalid */
+	u32	width;		/* In pixel */
+	u32	height;		/* In lines */
+	char	*desc;		/* The description */
+};
+
+static struct cursor_mode_format cursor_pixel_formats[] = {
+	{DRM_FORMAT_ARGB8888, 32, 128, 128, "128x128 32bpp ARGB"},
+	{DRM_FORMAT_ARGB8888, 32, 256, 256, "256x256 32bpp ARGB"},
+	{DRM_FORMAT_ARGB8888, 32, 64, 64, "64x64 32bpp ARGB"},
+	{DRM_FORMAT_ARGB8888, 32, 64, 64, "64x64 32bpp ARGB"},
+
+	/* non-supported format has bpp default to 0 */
+	{0, 0, 0, 0, NULL},
+};
+
+static int cursor_mode_to_drm(int mode)
+{
+	int cursor_pixel_formats_index = 4;
+
+	switch (mode) {
+	case CURSOR_MODE_128_ARGB_AX:
+		cursor_pixel_formats_index = 0;
+		break;
+	case CURSOR_MODE_256_ARGB_AX:
+		cursor_pixel_formats_index = 1;
+		break;
+	case CURSOR_MODE_64_ARGB_AX:
+		cursor_pixel_formats_index = 2;
+		break;
+	case CURSOR_MODE_64_32B_AX:
+		cursor_pixel_formats_index = 3;
+		break;
+
+	default:
+		break;
+	}
+
+	return cursor_pixel_formats_index;
+}
+
+/**
+ * intel_vgpu_decode_cursor_plane - Decode sprite plane
+ * @vgpu: input vgpu
+ * @plane: cursor plane to save decoded info
+ * This function is called for decoding plane
+ *
+ * Returns:
+ * 0 on success, non-zero if failed.
+ */
+int intel_vgpu_decode_cursor_plane(struct intel_vgpu *vgpu,
+	struct intel_vgpu_cursor_plane_format *plane)
+{
+	u32 val, mode, index;
+	u32 alpha_plane, alpha_force;
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	int pipe;
+
+	pipe = get_active_pipe(vgpu);
+	if (pipe >= I915_MAX_PIPES)
+		return -ENODEV;
+
+	val = vgpu_vreg_t(vgpu, CURCNTR(pipe));
+	mode = val & CURSOR_MODE;
+	plane->enabled = (mode != CURSOR_MODE_DISABLE);
+	if (!plane->enabled)
+		return -ENODEV;
+
+	index = cursor_mode_to_drm(mode);
+
+	if (!cursor_pixel_formats[index].bpp) {
+		gvt_vgpu_err("Non-supported cursor mode (0x%x)\n", mode);
+		return -EINVAL;
+	}
+	plane->mode = mode;
+	plane->bpp = cursor_pixel_formats[index].bpp;
+	plane->drm_format = cursor_pixel_formats[index].drm_format;
+	plane->width = cursor_pixel_formats[index].width;
+	plane->height = cursor_pixel_formats[index].height;
+
+	alpha_plane = (val & _CURSOR_ALPHA_PLANE_MASK) >>
+				_CURSOR_ALPHA_PLANE_SHIFT;
+	alpha_force = (val & _CURSOR_ALPHA_FORCE_MASK) >>
+				_CURSOR_ALPHA_FORCE_SHIFT;
+	if (alpha_plane || alpha_force)
+		gvt_dbg_core("alpha_plane=0x%x, alpha_force=0x%x\n",
+			alpha_plane, alpha_force);
+
+	plane->base = vgpu_vreg_t(vgpu, CURBASE(pipe)) & I915_GTT_PAGE_MASK;
+	if (!intel_gvt_ggtt_validate_range(vgpu, plane->base, 0)) {
+		gvt_vgpu_err("invalid gma address: %lx\n",
+			     (unsigned long)plane->base);
+		return  -EINVAL;
+	}
+
+	plane->base_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, plane->base);
+	if (plane->base_gpa == INTEL_GVT_INVALID_ADDR) {
+		gvt_vgpu_err("invalid gma address: %lx\n",
+				(unsigned long)plane->base);
+		return  -EINVAL;
+	}
+
+	val = vgpu_vreg_t(vgpu, CURPOS(pipe));
+	plane->x_pos = (val & _CURSOR_POS_X_MASK) >> _CURSOR_POS_X_SHIFT;
+	plane->x_sign = (val & _CURSOR_SIGN_X_MASK) >> _CURSOR_SIGN_X_SHIFT;
+	plane->y_pos = (val & _CURSOR_POS_Y_MASK) >> _CURSOR_POS_Y_SHIFT;
+	plane->y_sign = (val & _CURSOR_SIGN_Y_MASK) >> _CURSOR_SIGN_Y_SHIFT;
+
+	return 0;
+}
+
+#define SPRITE_FORMAT_NUM	(1 << 3)
+
+static struct pixel_format sprite_pixel_formats[SPRITE_FORMAT_NUM] = {
+	[0x0] = {DRM_FORMAT_YUV422, 16, "YUV 16-bit 4:2:2 packed"},
+	[0x1] = {DRM_FORMAT_XRGB2101010, 32, "RGB 32-bit 2:10:10:10"},
+	[0x2] = {DRM_FORMAT_XRGB8888, 32, "RGB 32-bit 8:8:8:8"},
+	[0x4] = {DRM_FORMAT_AYUV, 32,
+		"YUV 32-bit 4:4:4 packed (8:8:8:8 MSB-X:Y:U:V)"},
+};
+
+/**
+ * intel_vgpu_decode_sprite_plane - Decode sprite plane
+ * @vgpu: input vgpu
+ * @plane: sprite plane to save decoded info
+ * This function is called for decoding plane
+ *
+ * Returns:
+ * 0 on success, non-zero if failed.
+ */
+int intel_vgpu_decode_sprite_plane(struct intel_vgpu *vgpu,
+	struct intel_vgpu_sprite_plane_format *plane)
+{
+	u32 val, fmt;
+	u32 color_order, yuv_order;
+	int drm_format;
+	int pipe;
+
+	pipe = get_active_pipe(vgpu);
+	if (pipe >= I915_MAX_PIPES)
+		return -ENODEV;
+
+	val = vgpu_vreg_t(vgpu, SPRCTL(pipe));
+	plane->enabled = !!(val & SPRITE_ENABLE);
+	if (!plane->enabled)
+		return -ENODEV;
+
+	plane->tiled = !!(val & SPRITE_TILED);
+	color_order = !!(val & SPRITE_RGB_ORDER_RGBX);
+	yuv_order = (val & SPRITE_YUV_BYTE_ORDER_MASK) >>
+				_SPRITE_YUV_ORDER_SHIFT;
+
+	fmt = (val & SPRITE_PIXFORMAT_MASK) >> _SPRITE_FMT_SHIFT;
+	if (!sprite_pixel_formats[fmt].bpp) {
+		gvt_vgpu_err("Non-supported pixel format (0x%x)\n", fmt);
+		return -EINVAL;
+	}
+	plane->hw_format = fmt;
+	plane->bpp = sprite_pixel_formats[fmt].bpp;
+	drm_format = sprite_pixel_formats[fmt].drm_format;
+
+	/* Order of RGB values in an RGBxxx buffer may be ordered RGB or
+	 * BGR depending on the state of the color_order field
+	 */
+	if (!color_order) {
+		if (drm_format == DRM_FORMAT_XRGB2101010)
+			drm_format = DRM_FORMAT_XBGR2101010;
+		else if (drm_format == DRM_FORMAT_XRGB8888)
+			drm_format = DRM_FORMAT_XBGR8888;
+	}
+
+	if (drm_format == DRM_FORMAT_YUV422) {
+		switch (yuv_order) {
+		case 0:
+			drm_format = DRM_FORMAT_YUYV;
+			break;
+		case 1:
+			drm_format = DRM_FORMAT_UYVY;
+			break;
+		case 2:
+			drm_format = DRM_FORMAT_YVYU;
+			break;
+		case 3:
+			drm_format = DRM_FORMAT_VYUY;
+			break;
+		default:
+			/* yuv_order has only 2 bits */
+			break;
+		}
+	}
+
+	plane->drm_format = drm_format;
+
+	plane->base = vgpu_vreg_t(vgpu, SPRSURF(pipe)) & I915_GTT_PAGE_MASK;
+	if (!intel_gvt_ggtt_validate_range(vgpu, plane->base, 0)) {
+		gvt_vgpu_err("invalid gma address: %lx\n",
+			     (unsigned long)plane->base);
+		return  -EINVAL;
+	}
+
+	plane->base_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, plane->base);
+	if (plane->base_gpa == INTEL_GVT_INVALID_ADDR) {
+		gvt_vgpu_err("invalid gma address: %lx\n",
+				(unsigned long)plane->base);
+		return  -EINVAL;
+	}
+
+	plane->stride = vgpu_vreg_t(vgpu, SPRSTRIDE(pipe)) &
+				_SPRITE_STRIDE_MASK;
+
+	val = vgpu_vreg_t(vgpu, SPRSIZE(pipe));
+	plane->height = (val & _SPRITE_SIZE_HEIGHT_MASK) >>
+		_SPRITE_SIZE_HEIGHT_SHIFT;
+	plane->width = (val & _SPRITE_SIZE_WIDTH_MASK) >>
+		_SPRITE_SIZE_WIDTH_SHIFT;
+	plane->height += 1;	/* raw height is one minus the real value */
+	plane->width += 1;	/* raw width is one minus the real value */
+
+	val = vgpu_vreg_t(vgpu, SPRPOS(pipe));
+	plane->x_pos = (val & _SPRITE_POS_X_MASK) >> _SPRITE_POS_X_SHIFT;
+	plane->y_pos = (val & _SPRITE_POS_Y_MASK) >> _SPRITE_POS_Y_SHIFT;
+
+	val = vgpu_vreg_t(vgpu, SPROFFSET(pipe));
+	plane->x_offset = (val & _SPRITE_OFFSET_START_X_MASK) >>
+			   _SPRITE_OFFSET_START_X_SHIFT;
+	plane->y_offset = (val & _SPRITE_OFFSET_START_Y_MASK) >>
+			   _SPRITE_OFFSET_START_Y_SHIFT;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/i915/gvt/fb_decoder.h b/drivers/gpu/drm/i915/gvt/fb_decoder.h
new file mode 100644
index 000000000000..cb055f3c81a2
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/fb_decoder.h
@@ -0,0 +1,169 @@
+/*
+ * Copyright(c) 2011-2016 Intel Corporation. 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, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Authors:
+ *    Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ *    Bing Niu <bing.niu@intel.com>
+ *    Xu Han <xu.han@intel.com>
+ *    Ping Gao <ping.a.gao@intel.com>
+ *    Xiaoguang Chen <xiaoguang.chen@intel.com>
+ *    Yang Liu <yang2.liu@intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *
+ */
+
+#ifndef _GVT_FB_DECODER_H_
+#define _GVT_FB_DECODER_H_
+
+#define _PLANE_CTL_FORMAT_SHIFT		24
+#define _PLANE_CTL_TILED_SHIFT		10
+#define _PIPE_V_SRCSZ_SHIFT		0
+#define _PIPE_V_SRCSZ_MASK		(0xfff << _PIPE_V_SRCSZ_SHIFT)
+#define _PIPE_H_SRCSZ_SHIFT		16
+#define _PIPE_H_SRCSZ_MASK		(0x1fff << _PIPE_H_SRCSZ_SHIFT)
+
+#define _PRI_PLANE_FMT_SHIFT		26
+#define _PRI_PLANE_STRIDE_MASK		(0x3ff << 6)
+#define _PRI_PLANE_X_OFF_SHIFT		0
+#define _PRI_PLANE_X_OFF_MASK		(0x1fff << _PRI_PLANE_X_OFF_SHIFT)
+#define _PRI_PLANE_Y_OFF_SHIFT		16
+#define _PRI_PLANE_Y_OFF_MASK		(0xfff << _PRI_PLANE_Y_OFF_SHIFT)
+
+#define _CURSOR_MODE			0x3f
+#define _CURSOR_ALPHA_FORCE_SHIFT	8
+#define _CURSOR_ALPHA_FORCE_MASK	(0x3 << _CURSOR_ALPHA_FORCE_SHIFT)
+#define _CURSOR_ALPHA_PLANE_SHIFT	10
+#define _CURSOR_ALPHA_PLANE_MASK	(0x3 << _CURSOR_ALPHA_PLANE_SHIFT)
+#define _CURSOR_POS_X_SHIFT		0
+#define _CURSOR_POS_X_MASK		(0x1fff << _CURSOR_POS_X_SHIFT)
+#define _CURSOR_SIGN_X_SHIFT		15
+#define _CURSOR_SIGN_X_MASK		(1 << _CURSOR_SIGN_X_SHIFT)
+#define _CURSOR_POS_Y_SHIFT		16
+#define _CURSOR_POS_Y_MASK		(0xfff << _CURSOR_POS_Y_SHIFT)
+#define _CURSOR_SIGN_Y_SHIFT		31
+#define _CURSOR_SIGN_Y_MASK		(1 << _CURSOR_SIGN_Y_SHIFT)
+
+#define _SPRITE_FMT_SHIFT		25
+#define _SPRITE_COLOR_ORDER_SHIFT	20
+#define _SPRITE_YUV_ORDER_SHIFT		16
+#define _SPRITE_STRIDE_SHIFT		6
+#define _SPRITE_STRIDE_MASK		(0x1ff << _SPRITE_STRIDE_SHIFT)
+#define _SPRITE_SIZE_WIDTH_SHIFT	0
+#define _SPRITE_SIZE_HEIGHT_SHIFT	16
+#define _SPRITE_SIZE_WIDTH_MASK		(0x1fff << _SPRITE_SIZE_WIDTH_SHIFT)
+#define _SPRITE_SIZE_HEIGHT_MASK	(0xfff << _SPRITE_SIZE_HEIGHT_SHIFT)
+#define _SPRITE_POS_X_SHIFT		0
+#define _SPRITE_POS_Y_SHIFT		16
+#define _SPRITE_POS_X_MASK		(0x1fff << _SPRITE_POS_X_SHIFT)
+#define _SPRITE_POS_Y_MASK		(0xfff << _SPRITE_POS_Y_SHIFT)
+#define _SPRITE_OFFSET_START_X_SHIFT	0
+#define _SPRITE_OFFSET_START_Y_SHIFT	16
+#define _SPRITE_OFFSET_START_X_MASK	(0x1fff << _SPRITE_OFFSET_START_X_SHIFT)
+#define _SPRITE_OFFSET_START_Y_MASK	(0xfff << _SPRITE_OFFSET_START_Y_SHIFT)
+
+enum GVT_FB_EVENT {
+	FB_MODE_SET_START = 1,
+	FB_MODE_SET_END,
+	FB_DISPLAY_FLIP,
+};
+
+enum DDI_PORT {
+	DDI_PORT_NONE	= 0,
+	DDI_PORT_B	= 1,
+	DDI_PORT_C	= 2,
+	DDI_PORT_D	= 3,
+	DDI_PORT_E	= 4
+};
+
+struct intel_gvt;
+
+/* color space conversion and gamma correction are not included */
+struct intel_vgpu_primary_plane_format {
+	u8	enabled;	/* plane is enabled */
+	u8	tiled;		/* X-tiled */
+	u8	bpp;		/* bits per pixel */
+	u32	hw_format;	/* format field in the PRI_CTL register */
+	u32	drm_format;	/* format in DRM definition */
+	u32	base;		/* framebuffer base in graphics memory */
+	u64     base_gpa;
+	u32	x_offset;	/* in pixels */
+	u32	y_offset;	/* in lines */
+	u32	width;		/* in pixels */
+	u32	height;		/* in lines */
+	u32	stride;		/* in bytes */
+};
+
+struct intel_vgpu_sprite_plane_format {
+	u8	enabled;	/* plane is enabled */
+	u8	tiled;		/* X-tiled */
+	u8	bpp;		/* bits per pixel */
+	u32	hw_format;	/* format field in the SPR_CTL register */
+	u32	drm_format;	/* format in DRM definition */
+	u32	base;		/* sprite base in graphics memory */
+	u64     base_gpa;
+	u32	x_pos;		/* in pixels */
+	u32	y_pos;		/* in lines */
+	u32	x_offset;	/* in pixels */
+	u32	y_offset;	/* in lines */
+	u32	width;		/* in pixels */
+	u32	height;		/* in lines */
+	u32	stride;		/* in bytes */
+};
+
+struct intel_vgpu_cursor_plane_format {
+	u8	enabled;
+	u8	mode;		/* cursor mode select */
+	u8	bpp;		/* bits per pixel */
+	u32	drm_format;	/* format in DRM definition */
+	u32	base;		/* cursor base in graphics memory */
+	u64     base_gpa;
+	u32	x_pos;		/* in pixels */
+	u32	y_pos;		/* in lines */
+	u8	x_sign;		/* X Position Sign */
+	u8	y_sign;		/* Y Position Sign */
+	u32	width;		/* in pixels */
+	u32	height;		/* in lines */
+	u32	x_hot;		/* in pixels */
+	u32	y_hot;		/* in pixels */
+};
+
+struct intel_vgpu_pipe_format {
+	struct intel_vgpu_primary_plane_format	primary;
+	struct intel_vgpu_sprite_plane_format	sprite;
+	struct intel_vgpu_cursor_plane_format	cursor;
+	enum DDI_PORT ddi_port;  /* the DDI port that pipe is connected to */
+};
+
+struct intel_vgpu_fb_format {
+	struct intel_vgpu_pipe_format	pipes[I915_MAX_PIPES];
+};
+
+int intel_vgpu_decode_primary_plane(struct intel_vgpu *vgpu,
+	struct intel_vgpu_primary_plane_format *plane);
+int intel_vgpu_decode_cursor_plane(struct intel_vgpu *vgpu,
+	struct intel_vgpu_cursor_plane_format *plane);
+int intel_vgpu_decode_sprite_plane(struct intel_vgpu *vgpu,
+	struct intel_vgpu_sprite_plane_format *plane);
+
+#endif
diff --git a/drivers/gpu/drm/i915/gvt/firmware.c b/drivers/gpu/drm/i915/gvt/firmware.c
index a26c1705430e..a73e1d418c22 100644
--- a/drivers/gpu/drm/i915/gvt/firmware.c
+++ b/drivers/gpu/drm/i915/gvt/firmware.c
@@ -66,20 +66,23 @@ static struct bin_attribute firmware_attr = {
 	.mmap = NULL,
 };
 
-static int expose_firmware_sysfs(struct intel_gvt *gvt)
+static int mmio_snapshot_handler(struct intel_gvt *gvt, u32 offset, void *data)
 {
 	struct drm_i915_private *dev_priv = gvt->dev_priv;
+
+	*(u32 *)(data + offset) = I915_READ_NOTRACE(_MMIO(offset));
+	return 0;
+}
+
+static int expose_firmware_sysfs(struct intel_gvt *gvt)
+{
 	struct intel_gvt_device_info *info = &gvt->device_info;
 	struct pci_dev *pdev = gvt->dev_priv->drm.pdev;
-	struct intel_gvt_mmio_info *e;
-	struct gvt_mmio_block *block = gvt->mmio.mmio_block;
-	int num = gvt->mmio.num_mmio_block;
 	struct gvt_firmware_header *h;
 	void *firmware;
 	void *p;
 	unsigned long size, crc32_start;
-	int i, j;
-	int ret;
+	int i, ret;
 
 	size = sizeof(*h) + info->mmio_size + info->cfg_space_size;
 	firmware = vzalloc(size);
@@ -104,15 +107,8 @@ static int expose_firmware_sysfs(struct intel_gvt *gvt)
 
 	p = firmware + h->mmio_offset;
 
-	hash_for_each(gvt->mmio.mmio_info_table, i, e, node)
-		*(u32 *)(p + e->offset) = I915_READ_NOTRACE(_MMIO(e->offset));
-
-	for (i = 0; i < num; i++, block++) {
-		for (j = 0; j < block->size; j += 4)
-			*(u32 *)(p + INTEL_GVT_MMIO_OFFSET(block->offset) + j) =
-				I915_READ_NOTRACE(_MMIO(INTEL_GVT_MMIO_OFFSET(
-							block->offset) + j));
-	}
+	/* Take a snapshot of hw mmio registers. */
+	intel_gvt_for_each_tracked_mmio(gvt, mmio_snapshot_handler, p);
 
 	memcpy(gvt->firmware.mmio, p, info->mmio_size);
 
diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c
index 64d67ff9bf08..a529d2bd393c 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.c
+++ b/drivers/gpu/drm/i915/gvt/gtt.c
@@ -94,12 +94,12 @@ int intel_gvt_ggtt_index_g2h(struct intel_vgpu *vgpu, unsigned long g_index,
 	u64 h_addr;
 	int ret;
 
-	ret = intel_gvt_ggtt_gmadr_g2h(vgpu, g_index << GTT_PAGE_SHIFT,
+	ret = intel_gvt_ggtt_gmadr_g2h(vgpu, g_index << I915_GTT_PAGE_SHIFT,
 				       &h_addr);
 	if (ret)
 		return ret;
 
-	*h_index = h_addr >> GTT_PAGE_SHIFT;
+	*h_index = h_addr >> I915_GTT_PAGE_SHIFT;
 	return 0;
 }
 
@@ -109,12 +109,12 @@ int intel_gvt_ggtt_h2g_index(struct intel_vgpu *vgpu, unsigned long h_index,
 	u64 g_addr;
 	int ret;
 
-	ret = intel_gvt_ggtt_gmadr_h2g(vgpu, h_index << GTT_PAGE_SHIFT,
+	ret = intel_gvt_ggtt_gmadr_h2g(vgpu, h_index << I915_GTT_PAGE_SHIFT,
 				       &g_addr);
 	if (ret)
 		return ret;
 
-	*g_index = g_addr >> GTT_PAGE_SHIFT;
+	*g_index = g_addr >> I915_GTT_PAGE_SHIFT;
 	return 0;
 }
 
@@ -156,13 +156,15 @@ int intel_gvt_ggtt_h2g_index(struct intel_vgpu *vgpu, unsigned long h_index,
 
 struct gtt_type_table_entry {
 	int entry_type;
+	int pt_type;
 	int next_pt_type;
 	int pse_entry_type;
 };
 
-#define GTT_TYPE_TABLE_ENTRY(type, e_type, npt_type, pse_type) \
+#define GTT_TYPE_TABLE_ENTRY(type, e_type, cpt_type, npt_type, pse_type) \
 	[type] = { \
 		.entry_type = e_type, \
+		.pt_type = cpt_type, \
 		.next_pt_type = npt_type, \
 		.pse_entry_type = pse_type, \
 	}
@@ -170,55 +172,68 @@ struct gtt_type_table_entry {
 static struct gtt_type_table_entry gtt_type_table[] = {
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_ROOT_L4_ENTRY,
 			GTT_TYPE_PPGTT_ROOT_L4_ENTRY,
+			GTT_TYPE_INVALID,
 			GTT_TYPE_PPGTT_PML4_PT,
 			GTT_TYPE_INVALID),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PML4_PT,
 			GTT_TYPE_PPGTT_PML4_ENTRY,
+			GTT_TYPE_PPGTT_PML4_PT,
 			GTT_TYPE_PPGTT_PDP_PT,
 			GTT_TYPE_INVALID),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PML4_ENTRY,
 			GTT_TYPE_PPGTT_PML4_ENTRY,
+			GTT_TYPE_PPGTT_PML4_PT,
 			GTT_TYPE_PPGTT_PDP_PT,
 			GTT_TYPE_INVALID),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDP_PT,
 			GTT_TYPE_PPGTT_PDP_ENTRY,
+			GTT_TYPE_PPGTT_PDP_PT,
 			GTT_TYPE_PPGTT_PDE_PT,
 			GTT_TYPE_PPGTT_PTE_1G_ENTRY),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_ROOT_L3_ENTRY,
 			GTT_TYPE_PPGTT_ROOT_L3_ENTRY,
+			GTT_TYPE_INVALID,
 			GTT_TYPE_PPGTT_PDE_PT,
 			GTT_TYPE_PPGTT_PTE_1G_ENTRY),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDP_ENTRY,
 			GTT_TYPE_PPGTT_PDP_ENTRY,
+			GTT_TYPE_PPGTT_PDP_PT,
 			GTT_TYPE_PPGTT_PDE_PT,
 			GTT_TYPE_PPGTT_PTE_1G_ENTRY),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDE_PT,
 			GTT_TYPE_PPGTT_PDE_ENTRY,
+			GTT_TYPE_PPGTT_PDE_PT,
 			GTT_TYPE_PPGTT_PTE_PT,
 			GTT_TYPE_PPGTT_PTE_2M_ENTRY),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDE_ENTRY,
 			GTT_TYPE_PPGTT_PDE_ENTRY,
+			GTT_TYPE_PPGTT_PDE_PT,
 			GTT_TYPE_PPGTT_PTE_PT,
 			GTT_TYPE_PPGTT_PTE_2M_ENTRY),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_PT,
 			GTT_TYPE_PPGTT_PTE_4K_ENTRY,
+			GTT_TYPE_PPGTT_PTE_PT,
 			GTT_TYPE_INVALID,
 			GTT_TYPE_INVALID),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_4K_ENTRY,
 			GTT_TYPE_PPGTT_PTE_4K_ENTRY,
+			GTT_TYPE_PPGTT_PTE_PT,
 			GTT_TYPE_INVALID,
 			GTT_TYPE_INVALID),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_2M_ENTRY,
 			GTT_TYPE_PPGTT_PDE_ENTRY,
+			GTT_TYPE_PPGTT_PDE_PT,
 			GTT_TYPE_INVALID,
 			GTT_TYPE_PPGTT_PTE_2M_ENTRY),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_1G_ENTRY,
 			GTT_TYPE_PPGTT_PDP_ENTRY,
+			GTT_TYPE_PPGTT_PDP_PT,
 			GTT_TYPE_INVALID,
 			GTT_TYPE_PPGTT_PTE_1G_ENTRY),
 	GTT_TYPE_TABLE_ENTRY(GTT_TYPE_GGTT_PTE,
 			GTT_TYPE_GGTT_PTE,
 			GTT_TYPE_INVALID,
+			GTT_TYPE_INVALID,
 			GTT_TYPE_INVALID),
 };
 
@@ -227,6 +242,11 @@ static inline int get_next_pt_type(int type)
 	return gtt_type_table[type].next_pt_type;
 }
 
+static inline int get_pt_type(int type)
+{
+	return gtt_type_table[type].pt_type;
+}
+
 static inline int get_entry_type(int type)
 {
 	return gtt_type_table[type].entry_type;
@@ -351,7 +371,7 @@ static bool gen8_gtt_test_pse(struct intel_gvt_gtt_entry *e)
 		return false;
 
 	e->type = get_entry_type(e->type);
-	if (!(e->val64 & (1 << 7)))
+	if (!(e->val64 & BIT(7)))
 		return false;
 
 	e->type = get_pse_type(e->type);
@@ -369,12 +389,17 @@ static bool gen8_gtt_test_present(struct intel_gvt_gtt_entry *e)
 			|| e->type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY)
 		return (e->val64 != 0);
 	else
-		return (e->val64 & (1 << 0));
+		return (e->val64 & BIT(0));
 }
 
 static void gtt_entry_clear_present(struct intel_gvt_gtt_entry *e)
 {
-	e->val64 &= ~(1 << 0);
+	e->val64 &= ~BIT(0);
+}
+
+static void gtt_entry_set_present(struct intel_gvt_gtt_entry *e)
+{
+	e->val64 |= BIT(0);
 }
 
 /*
@@ -382,7 +407,7 @@ static void gtt_entry_clear_present(struct intel_gvt_gtt_entry *e)
  */
 static unsigned long gma_to_ggtt_pte_index(unsigned long gma)
 {
-	unsigned long x = (gma >> GTT_PAGE_SHIFT);
+	unsigned long x = (gma >> I915_GTT_PAGE_SHIFT);
 
 	trace_gma_index(__func__, gma, x);
 	return x;
@@ -406,6 +431,7 @@ static struct intel_gvt_gtt_pte_ops gen8_gtt_pte_ops = {
 	.get_entry = gtt_get_entry64,
 	.set_entry = gtt_set_entry64,
 	.clear_present = gtt_entry_clear_present,
+	.set_present = gtt_entry_set_present,
 	.test_present = gen8_gtt_test_present,
 	.test_pse = gen8_gtt_test_pse,
 	.get_pfn = gen8_gtt_get_pfn,
@@ -494,7 +520,7 @@ static inline int ppgtt_spt_get_entry(
 		return -EINVAL;
 
 	ret = ops->get_entry(page_table, e, index, guest,
-			spt->guest_page.gfn << GTT_PAGE_SHIFT,
+			spt->guest_page.track.gfn << I915_GTT_PAGE_SHIFT,
 			spt->vgpu);
 	if (ret)
 		return ret;
@@ -516,7 +542,7 @@ static inline int ppgtt_spt_set_entry(
 		return -EINVAL;
 
 	return ops->set_entry(page_table, e, index, guest,
-			spt->guest_page.gfn << GTT_PAGE_SHIFT,
+			spt->guest_page.track.gfn << I915_GTT_PAGE_SHIFT,
 			spt->vgpu);
 }
 
@@ -537,88 +563,103 @@ static inline int ppgtt_spt_set_entry(
 		spt->shadow_page.type, e, index, false)
 
 /**
- * intel_vgpu_init_guest_page - init a guest page data structure
+ * intel_vgpu_init_page_track - init a page track data structure
  * @vgpu: a vGPU
- * @p: a guest page data structure
+ * @t: a page track data structure
  * @gfn: guest memory page frame number
- * @handler: function will be called when target guest memory page has
+ * @handler: the function will be called when target guest memory page has
  * been modified.
  *
- * This function is called when user wants to track a guest memory page.
+ * This function is called when a user wants to prepare a page track data
+ * structure to track a guest memory page.
  *
  * Returns:
  * Zero on success, negative error code if failed.
  */
-int intel_vgpu_init_guest_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *p,
+int intel_vgpu_init_page_track(struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t,
 		unsigned long gfn,
 		int (*handler)(void *, u64, void *, int),
 		void *data)
 {
-	INIT_HLIST_NODE(&p->node);
+	INIT_HLIST_NODE(&t->node);
 
-	p->writeprotection = false;
-	p->gfn = gfn;
-	p->handler = handler;
-	p->data = data;
-	p->oos_page = NULL;
-	p->write_cnt = 0;
+	t->tracked = false;
+	t->gfn = gfn;
+	t->handler = handler;
+	t->data = data;
 
-	hash_add(vgpu->gtt.guest_page_hash_table, &p->node, p->gfn);
+	hash_add(vgpu->gtt.tracked_guest_page_hash_table, &t->node, t->gfn);
 	return 0;
 }
 
-static int detach_oos_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_oos_page *oos_page);
-
 /**
- * intel_vgpu_clean_guest_page - release the resource owned by guest page data
- * structure
+ * intel_vgpu_clean_page_track - release a page track data structure
  * @vgpu: a vGPU
- * @p: a tracked guest page
+ * @t: a page track data structure
  *
- * This function is called when user tries to stop tracking a guest memory
- * page.
+ * This function is called before a user frees a page track data structure.
  */
-void intel_vgpu_clean_guest_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *p)
+void intel_vgpu_clean_page_track(struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t)
 {
-	if (!hlist_unhashed(&p->node))
-		hash_del(&p->node);
-
-	if (p->oos_page)
-		detach_oos_page(vgpu, p->oos_page);
+	if (!hlist_unhashed(&t->node))
+		hash_del(&t->node);
 
-	if (p->writeprotection)
-		intel_gvt_hypervisor_unset_wp_page(vgpu, p);
+	if (t->tracked)
+		intel_gvt_hypervisor_disable_page_track(vgpu, t);
 }
 
 /**
- * intel_vgpu_find_guest_page - find a guest page data structure by GFN.
+ * intel_vgpu_find_tracked_page - find a tracked guest page
  * @vgpu: a vGPU
  * @gfn: guest memory page frame number
  *
- * This function is called when emulation logic wants to know if a trapped GFN
- * is a tracked guest page.
+ * This function is called when the emulation layer wants to figure out if a
+ * trapped GFN is a tracked guest page.
  *
  * Returns:
- * Pointer to guest page data structure, NULL if failed.
+ * Pointer to page track data structure, NULL if not found.
  */
-struct intel_vgpu_guest_page *intel_vgpu_find_guest_page(
+struct intel_vgpu_page_track *intel_vgpu_find_tracked_page(
 		struct intel_vgpu *vgpu, unsigned long gfn)
 {
-	struct intel_vgpu_guest_page *p;
+	struct intel_vgpu_page_track *t;
 
-	hash_for_each_possible(vgpu->gtt.guest_page_hash_table,
-		p, node, gfn) {
-		if (p->gfn == gfn)
-			return p;
+	hash_for_each_possible(vgpu->gtt.tracked_guest_page_hash_table,
+			t, node, gfn) {
+		if (t->gfn == gfn)
+			return t;
 	}
 	return NULL;
 }
 
+static int init_guest_page(struct intel_vgpu *vgpu,
+		struct intel_vgpu_guest_page *p,
+		unsigned long gfn,
+		int (*handler)(void *, u64, void *, int),
+		void *data)
+{
+	p->oos_page = NULL;
+	p->write_cnt = 0;
+
+	return intel_vgpu_init_page_track(vgpu, &p->track, gfn, handler, data);
+}
+
+static int detach_oos_page(struct intel_vgpu *vgpu,
+		struct intel_vgpu_oos_page *oos_page);
+
+static void clean_guest_page(struct intel_vgpu *vgpu,
+		struct intel_vgpu_guest_page *p)
+{
+	if (p->oos_page)
+		detach_oos_page(vgpu, p->oos_page);
+
+	intel_vgpu_clean_page_track(vgpu, &p->track);
+}
+
 static inline int init_shadow_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_shadow_page *p, int type)
+		struct intel_vgpu_shadow_page *p, int type, bool hash)
 {
 	struct device *kdev = &vgpu->gvt->dev_priv->drm.pdev->dev;
 	dma_addr_t daddr;
@@ -634,8 +675,9 @@ static inline int init_shadow_page(struct intel_vgpu *vgpu,
 
 	INIT_HLIST_NODE(&p->node);
 
-	p->mfn = daddr >> GTT_PAGE_SHIFT;
-	hash_add(vgpu->gtt.shadow_page_hash_table, &p->node, p->mfn);
+	p->mfn = daddr >> I915_GTT_PAGE_SHIFT;
+	if (hash)
+		hash_add(vgpu->gtt.shadow_page_hash_table, &p->node, p->mfn);
 	return 0;
 }
 
@@ -644,7 +686,7 @@ static inline void clean_shadow_page(struct intel_vgpu *vgpu,
 {
 	struct device *kdev = &vgpu->gvt->dev_priv->drm.pdev->dev;
 
-	dma_unmap_page(kdev, p->mfn << GTT_PAGE_SHIFT, 4096,
+	dma_unmap_page(kdev, p->mfn << I915_GTT_PAGE_SHIFT, 4096,
 			PCI_DMA_BIDIRECTIONAL);
 
 	if (!hlist_unhashed(&p->node))
@@ -664,6 +706,9 @@ static inline struct intel_vgpu_shadow_page *find_shadow_page(
 	return NULL;
 }
 
+#define page_track_to_guest_page(ptr) \
+	container_of(ptr, struct intel_vgpu_guest_page, track)
+
 #define guest_page_to_ppgtt_spt(ptr) \
 	container_of(ptr, struct intel_vgpu_ppgtt_spt, guest_page)
 
@@ -697,7 +742,7 @@ static void ppgtt_free_shadow_page(struct intel_vgpu_ppgtt_spt *spt)
 	trace_spt_free(spt->vgpu->id, spt, spt->shadow_page.type);
 
 	clean_shadow_page(spt->vgpu, &spt->shadow_page);
-	intel_vgpu_clean_guest_page(spt->vgpu, &spt->guest_page);
+	clean_guest_page(spt->vgpu, &spt->guest_page);
 	list_del_init(&spt->post_shadow_list);
 
 	free_spt(spt);
@@ -713,22 +758,24 @@ static void ppgtt_free_all_shadow_page(struct intel_vgpu *vgpu)
 		ppgtt_free_shadow_page(shadow_page_to_ppgtt_spt(sp));
 }
 
-static int ppgtt_handle_guest_write_page_table_bytes(void *gp,
+static int ppgtt_handle_guest_write_page_table_bytes(
+		struct intel_vgpu_guest_page *gpt,
 		u64 pa, void *p_data, int bytes);
 
-static int ppgtt_write_protection_handler(void *gp, u64 pa,
+static int ppgtt_write_protection_handler(void *data, u64 pa,
 		void *p_data, int bytes)
 {
-	struct intel_vgpu_guest_page *gpt = (struct intel_vgpu_guest_page *)gp;
+	struct intel_vgpu_page_track *t = data;
+	struct intel_vgpu_guest_page *p = page_track_to_guest_page(t);
 	int ret;
 
 	if (bytes != 4 && bytes != 8)
 		return -EINVAL;
 
-	if (!gpt->writeprotection)
+	if (!t->tracked)
 		return -EINVAL;
 
-	ret = ppgtt_handle_guest_write_page_table_bytes(gp,
+	ret = ppgtt_handle_guest_write_page_table_bytes(p,
 		pa, p_data, bytes);
 	if (ret)
 		return ret;
@@ -762,13 +809,13 @@ retry:
 	 * TODO: guest page type may be different with shadow page type,
 	 *	 when we support PSE page in future.
 	 */
-	ret = init_shadow_page(vgpu, &spt->shadow_page, type);
+	ret = init_shadow_page(vgpu, &spt->shadow_page, type, true);
 	if (ret) {
 		gvt_vgpu_err("fail to initialize shadow page for spt\n");
 		goto err;
 	}
 
-	ret = intel_vgpu_init_guest_page(vgpu, &spt->guest_page,
+	ret = init_guest_page(vgpu, &spt->guest_page,
 			gfn, ppgtt_write_protection_handler, NULL);
 	if (ret) {
 		gvt_vgpu_err("fail to initialize guest page for spt\n");
@@ -798,7 +845,7 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_find_shadow_page(
 	((spt)->vgpu->gvt->device_info.gtt_entry_size_shift)
 
 #define pt_entries(spt) \
-	(GTT_PAGE_SIZE >> pt_entry_size_shift(spt))
+	(I915_GTT_PAGE_SIZE >> pt_entry_size_shift(spt))
 
 #define for_each_present_guest_entry(spt, e, i) \
 	for (i = 0; i < pt_entries(spt); i++) \
@@ -856,7 +903,7 @@ static int ppgtt_invalidate_shadow_page(struct intel_vgpu_ppgtt_spt *spt)
 	int v = atomic_read(&spt->refcount);
 
 	trace_spt_change(spt->vgpu->id, "die", spt,
-			spt->guest_page.gfn, spt->shadow_page.type);
+			spt->guest_page.track.gfn, spt->shadow_page.type);
 
 	trace_spt_refcount(spt->vgpu->id, "dec", spt, v, (v - 1));
 
@@ -878,7 +925,7 @@ static int ppgtt_invalidate_shadow_page(struct intel_vgpu_ppgtt_spt *spt)
 	}
 release:
 	trace_spt_change(spt->vgpu->id, "release", spt,
-			spt->guest_page.gfn, spt->shadow_page.type);
+			spt->guest_page.track.gfn, spt->shadow_page.type);
 	ppgtt_free_shadow_page(spt);
 	return 0;
 fail:
@@ -895,6 +942,7 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_populate_shadow_page_by_guest_entry(
 	struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
 	struct intel_vgpu_ppgtt_spt *s = NULL;
 	struct intel_vgpu_guest_page *g;
+	struct intel_vgpu_page_track *t;
 	int ret;
 
 	if (WARN_ON(!gtt_type_is_pt(get_next_pt_type(we->type)))) {
@@ -902,8 +950,9 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_populate_shadow_page_by_guest_entry(
 		goto fail;
 	}
 
-	g = intel_vgpu_find_guest_page(vgpu, ops->get_pfn(we));
-	if (g) {
+	t = intel_vgpu_find_tracked_page(vgpu, ops->get_pfn(we));
+	if (t) {
+		g = page_track_to_guest_page(t);
 		s = guest_page_to_ppgtt_spt(g);
 		ppgtt_get_shadow_page(s);
 	} else {
@@ -915,7 +964,8 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_populate_shadow_page_by_guest_entry(
 			goto fail;
 		}
 
-		ret = intel_gvt_hypervisor_set_wp_page(vgpu, &s->guest_page);
+		ret = intel_gvt_hypervisor_enable_page_track(vgpu,
+				&s->guest_page.track);
 		if (ret)
 			goto fail;
 
@@ -923,7 +973,7 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_populate_shadow_page_by_guest_entry(
 		if (ret)
 			goto fail;
 
-		trace_spt_change(vgpu->id, "new", s, s->guest_page.gfn,
+		trace_spt_change(vgpu->id, "new", s, s->guest_page.track.gfn,
 			s->shadow_page.type);
 	}
 	return s;
@@ -953,7 +1003,7 @@ static int ppgtt_populate_shadow_page(struct intel_vgpu_ppgtt_spt *spt)
 	int ret;
 
 	trace_spt_change(spt->vgpu->id, "born", spt,
-			spt->guest_page.gfn, spt->shadow_page.type);
+			spt->guest_page.track.gfn, spt->shadow_page.type);
 
 	if (gtt_type_is_pte_pt(spt->shadow_page.type)) {
 		for_each_present_guest_entry(spt, &ge, i) {
@@ -1078,11 +1128,11 @@ static int sync_oos_page(struct intel_vgpu *vgpu,
 	old.type = new.type = get_entry_type(spt->guest_page_type);
 	old.val64 = new.val64 = 0;
 
-	for (index = 0; index < (GTT_PAGE_SIZE >> info->gtt_entry_size_shift);
-		index++) {
+	for (index = 0; index < (I915_GTT_PAGE_SIZE >>
+				info->gtt_entry_size_shift); index++) {
 		ops->get_entry(oos_page->mem, &old, index, false, 0, vgpu);
 		ops->get_entry(NULL, &new, index, true,
-			oos_page->guest_page->gfn << PAGE_SHIFT, vgpu);
+			oos_page->guest_page->track.gfn << PAGE_SHIFT, vgpu);
 
 		if (old.val64 == new.val64
 			&& !test_and_clear_bit(index, spt->post_shadow_bitmap))
@@ -1132,8 +1182,9 @@ static int attach_oos_page(struct intel_vgpu *vgpu,
 	struct intel_gvt *gvt = vgpu->gvt;
 	int ret;
 
-	ret = intel_gvt_hypervisor_read_gpa(vgpu, gpt->gfn << GTT_PAGE_SHIFT,
-		oos_page->mem, GTT_PAGE_SIZE);
+	ret = intel_gvt_hypervisor_read_gpa(vgpu,
+			gpt->track.gfn << I915_GTT_PAGE_SHIFT,
+			oos_page->mem, I915_GTT_PAGE_SIZE);
 	if (ret)
 		return ret;
 
@@ -1152,7 +1203,7 @@ static int ppgtt_set_guest_page_sync(struct intel_vgpu *vgpu,
 {
 	int ret;
 
-	ret = intel_gvt_hypervisor_set_wp_page(vgpu, gpt);
+	ret = intel_gvt_hypervisor_enable_page_track(vgpu, &gpt->track);
 	if (ret)
 		return ret;
 
@@ -1200,7 +1251,7 @@ static int ppgtt_set_guest_page_oos(struct intel_vgpu *vgpu,
 			gpt, guest_page_to_ppgtt_spt(gpt)->guest_page_type);
 
 	list_add_tail(&oos_page->vm_list, &vgpu->gtt.oos_page_list_head);
-	return intel_gvt_hypervisor_unset_wp_page(vgpu, gpt);
+	return intel_gvt_hypervisor_disable_page_track(vgpu, &gpt->track);
 }
 
 /**
@@ -1335,10 +1386,10 @@ int intel_vgpu_flush_post_shadow(struct intel_vgpu *vgpu)
 	return 0;
 }
 
-static int ppgtt_handle_guest_write_page_table_bytes(void *gp,
+static int ppgtt_handle_guest_write_page_table_bytes(
+		struct intel_vgpu_guest_page *gpt,
 		u64 pa, void *p_data, int bytes)
 {
-	struct intel_vgpu_guest_page *gpt = (struct intel_vgpu_guest_page *)gp;
 	struct intel_vgpu_ppgtt_spt *spt = guest_page_to_ppgtt_spt(gpt);
 	struct intel_vgpu *vgpu = spt->vgpu;
 	struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
@@ -1418,7 +1469,7 @@ static int gen8_mm_alloc_page_table(struct intel_vgpu_mm *mm)
 		mm->shadow_page_table = mem + mm->page_table_entry_size;
 	} else if (mm->type == INTEL_GVT_MM_GGTT) {
 		mm->page_table_entry_cnt =
-			(gvt_ggtt_gm_sz(gvt) >> GTT_PAGE_SHIFT);
+			(gvt_ggtt_gm_sz(gvt) >> I915_GTT_PAGE_SHIFT);
 		mm->page_table_entry_size = mm->page_table_entry_cnt *
 			info->gtt_entry_size;
 		mem = vzalloc(mm->page_table_entry_size);
@@ -1740,8 +1791,8 @@ unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm, unsigned long gma)
 				gma_ops->gma_to_ggtt_pte_index(gma));
 		if (ret)
 			goto err;
-		gpa = (pte_ops->get_pfn(&e) << GTT_PAGE_SHIFT)
-			+ (gma & ~GTT_PAGE_MASK);
+		gpa = (pte_ops->get_pfn(&e) << I915_GTT_PAGE_SHIFT)
+			+ (gma & ~I915_GTT_PAGE_MASK);
 
 		trace_gma_translate(vgpu->id, "ggtt", 0, 0, gma, gpa);
 		return gpa;
@@ -1793,8 +1844,8 @@ unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm, unsigned long gma)
 		}
 	}
 
-	gpa = (pte_ops->get_pfn(&e) << GTT_PAGE_SHIFT)
-		+ (gma & ~GTT_PAGE_MASK);
+	gpa = (pte_ops->get_pfn(&e) << I915_GTT_PAGE_SHIFT)
+		+ (gma & ~I915_GTT_PAGE_MASK);
 
 	trace_gma_translate(vgpu->id, "ppgtt", 0,
 			mm->page_table_level, gma, gpa);
@@ -1862,7 +1913,7 @@ static int emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
 	if (bytes != 4 && bytes != 8)
 		return -EINVAL;
 
-	gma = g_gtt_index << GTT_PAGE_SHIFT;
+	gma = g_gtt_index << I915_GTT_PAGE_SHIFT;
 
 	/* the VM may configure the whole GM space when ballooning is used */
 	if (!vgpu_gmadr_is_valid(vgpu, gma))
@@ -1881,11 +1932,11 @@ static int emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
 			 * update the entry in this situation p2m will fail
 			 * settting the shadow entry to point to a scratch page
 			 */
-			ops->set_pfn(&m, gvt->gtt.scratch_ggtt_mfn);
+			ops->set_pfn(&m, gvt->gtt.scratch_mfn);
 		}
 	} else {
 		m = e;
-		ops->set_pfn(&m, gvt->gtt.scratch_ggtt_mfn);
+		ops->set_pfn(&m, gvt->gtt.scratch_mfn);
 	}
 
 	ggtt_set_shadow_entry(ggtt_mm, &m, g_gtt_index);
@@ -1920,12 +1971,45 @@ int intel_vgpu_emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
 	return ret;
 }
 
+int intel_vgpu_write_protect_handler(struct intel_vgpu *vgpu, u64 pa,
+				     void *p_data, unsigned int bytes)
+{
+	struct intel_gvt *gvt = vgpu->gvt;
+	int ret = 0;
+
+	if (atomic_read(&vgpu->gtt.n_tracked_guest_page)) {
+		struct intel_vgpu_page_track *t;
+
+		mutex_lock(&gvt->lock);
+
+		t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT);
+		if (t) {
+			if (unlikely(vgpu->failsafe)) {
+				/* remove write protection to prevent furture traps */
+				intel_vgpu_clean_page_track(vgpu, t);
+			} else {
+				ret = t->handler(t, pa, p_data, bytes);
+				if (ret) {
+					gvt_err("guest page write error %d, "
+						"gfn 0x%lx, pa 0x%llx, "
+						"var 0x%x, len %d\n",
+						ret, t->gfn, pa,
+						*(u32 *)p_data, bytes);
+				}
+			}
+		}
+		mutex_unlock(&gvt->lock);
+	}
+	return ret;
+}
+
+
 static int alloc_scratch_pages(struct intel_vgpu *vgpu,
 		intel_gvt_gtt_type_t type)
 {
 	struct intel_vgpu_gtt *gtt = &vgpu->gtt;
 	struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
-	int page_entry_num = GTT_PAGE_SIZE >>
+	int page_entry_num = I915_GTT_PAGE_SIZE >>
 				vgpu->gvt->device_info.gtt_entry_size_shift;
 	void *scratch_pt;
 	int i;
@@ -1949,7 +2033,7 @@ static int alloc_scratch_pages(struct intel_vgpu *vgpu,
 		return -ENOMEM;
 	}
 	gtt->scratch_pt[type].page_mfn =
-		(unsigned long)(daddr >> GTT_PAGE_SHIFT);
+		(unsigned long)(daddr >> I915_GTT_PAGE_SHIFT);
 	gtt->scratch_pt[type].page = virt_to_page(scratch_pt);
 	gvt_dbg_mm("vgpu%d create scratch_pt: type %d mfn=0x%lx\n",
 			vgpu->id, type, gtt->scratch_pt[type].page_mfn);
@@ -1992,7 +2076,7 @@ static int release_scratch_page_tree(struct intel_vgpu *vgpu)
 	for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) {
 		if (vgpu->gtt.scratch_pt[i].page != NULL) {
 			daddr = (dma_addr_t)(vgpu->gtt.scratch_pt[i].page_mfn <<
-					GTT_PAGE_SHIFT);
+					I915_GTT_PAGE_SHIFT);
 			dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL);
 			__free_page(vgpu->gtt.scratch_pt[i].page);
 			vgpu->gtt.scratch_pt[i].page = NULL;
@@ -2035,7 +2119,7 @@ int intel_vgpu_init_gtt(struct intel_vgpu *vgpu)
 	struct intel_vgpu_gtt *gtt = &vgpu->gtt;
 	struct intel_vgpu_mm *ggtt_mm;
 
-	hash_init(gtt->guest_page_hash_table);
+	hash_init(gtt->tracked_guest_page_hash_table);
 	hash_init(gtt->shadow_page_hash_table);
 
 	INIT_LIST_HEAD(&gtt->mm_list_head);
@@ -2196,7 +2280,7 @@ struct intel_vgpu_mm *intel_vgpu_find_ppgtt_mm(struct intel_vgpu *vgpu,
 int intel_vgpu_g2v_create_ppgtt_mm(struct intel_vgpu *vgpu,
 		int page_table_level)
 {
-	u64 *pdp = (u64 *)&vgpu_vreg64(vgpu, vgtif_reg(pdp[0]));
+	u64 *pdp = (u64 *)&vgpu_vreg64_t(vgpu, vgtif_reg(pdp[0]));
 	struct intel_vgpu_mm *mm;
 
 	if (WARN_ON((page_table_level != 4) && (page_table_level != 3)))
@@ -2231,7 +2315,7 @@ int intel_vgpu_g2v_create_ppgtt_mm(struct intel_vgpu *vgpu,
 int intel_vgpu_g2v_destroy_ppgtt_mm(struct intel_vgpu *vgpu,
 		int page_table_level)
 {
-	u64 *pdp = (u64 *)&vgpu_vreg64(vgpu, vgtif_reg(pdp[0]));
+	u64 *pdp = (u64 *)&vgpu_vreg64_t(vgpu, vgtif_reg(pdp[0]));
 	struct intel_vgpu_mm *mm;
 
 	if (WARN_ON((page_table_level != 4) && (page_table_level != 3)))
@@ -2288,15 +2372,16 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt)
 		__free_page(virt_to_page(page));
 		return -ENOMEM;
 	}
-	gvt->gtt.scratch_ggtt_page = virt_to_page(page);
-	gvt->gtt.scratch_ggtt_mfn = (unsigned long)(daddr >> GTT_PAGE_SHIFT);
+
+	gvt->gtt.scratch_page = virt_to_page(page);
+	gvt->gtt.scratch_mfn = (unsigned long)(daddr >> I915_GTT_PAGE_SHIFT);
 
 	if (enable_out_of_sync) {
 		ret = setup_spt_oos(gvt);
 		if (ret) {
 			gvt_err("fail to initialize SPT oos\n");
 			dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL);
-			__free_page(gvt->gtt.scratch_ggtt_page);
+			__free_page(gvt->gtt.scratch_page);
 			return ret;
 		}
 	}
@@ -2315,12 +2400,12 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt)
 void intel_gvt_clean_gtt(struct intel_gvt *gvt)
 {
 	struct device *dev = &gvt->dev_priv->drm.pdev->dev;
-	dma_addr_t daddr = (dma_addr_t)(gvt->gtt.scratch_ggtt_mfn <<
-					GTT_PAGE_SHIFT);
+	dma_addr_t daddr = (dma_addr_t)(gvt->gtt.scratch_mfn <<
+					I915_GTT_PAGE_SHIFT);
 
 	dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL);
 
-	__free_page(gvt->gtt.scratch_ggtt_page);
+	__free_page(gvt->gtt.scratch_page);
 
 	if (enable_out_of_sync)
 		clean_spt_oos(gvt);
@@ -2346,7 +2431,7 @@ void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu)
 
 	memset(&e, 0, sizeof(struct intel_gvt_gtt_entry));
 	e.type = GTT_TYPE_GGTT_PTE;
-	ops->set_pfn(&e, gvt->gtt.scratch_ggtt_mfn);
+	ops->set_pfn(&e, gvt->gtt.scratch_mfn);
 	e.val64 |= _PAGE_PRESENT;
 
 	index = vgpu_aperture_gmadr_base(vgpu) >> PAGE_SHIFT;
@@ -2372,8 +2457,6 @@ void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu)
  */
 void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu)
 {
-	int i;
-
 	ppgtt_free_all_shadow_page(vgpu);
 
 	/* Shadow pages are only created when there is no page
@@ -2383,11 +2466,4 @@ void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu)
 	intel_vgpu_free_mm(vgpu, INTEL_GVT_MM_PPGTT);
 
 	intel_vgpu_reset_ggtt(vgpu);
-
-	/* clear scratch page for security */
-	for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) {
-		if (vgpu->gtt.scratch_pt[i].page != NULL)
-			memset(page_address(vgpu->gtt.scratch_pt[i].page),
-				0, PAGE_SIZE);
-	}
 }
diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h
index 30a4c8d16026..4cc13b5934f1 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.h
+++ b/drivers/gpu/drm/i915/gvt/gtt.h
@@ -34,9 +34,8 @@
 #ifndef _GVT_GTT_H_
 #define _GVT_GTT_H_
 
-#define GTT_PAGE_SHIFT		12
-#define GTT_PAGE_SIZE		(1UL << GTT_PAGE_SHIFT)
-#define GTT_PAGE_MASK		(~(GTT_PAGE_SIZE-1))
+#define I915_GTT_PAGE_SHIFT         12
+#define I915_GTT_PAGE_MASK		(~(I915_GTT_PAGE_SIZE - 1))
 
 struct intel_vgpu_mm;
 
@@ -63,6 +62,7 @@ struct intel_gvt_gtt_pte_ops {
 			 struct intel_vgpu *vgpu);
 	bool (*test_present)(struct intel_gvt_gtt_entry *e);
 	void (*clear_present)(struct intel_gvt_gtt_entry *e);
+	void (*set_present)(struct intel_gvt_gtt_entry *e);
 	bool (*test_pse)(struct intel_gvt_gtt_entry *e);
 	void (*set_pfn)(struct intel_gvt_gtt_entry *e, unsigned long pfn);
 	unsigned long (*get_pfn)(struct intel_gvt_gtt_entry *e);
@@ -86,8 +86,8 @@ struct intel_gvt_gtt {
 	struct list_head oos_page_free_list_head;
 	struct list_head mm_lru_list_head;
 
-	struct page *scratch_ggtt_page;
-	unsigned long scratch_ggtt_mfn;
+	struct page *scratch_page;
+	unsigned long scratch_mfn;
 };
 
 enum {
@@ -193,18 +193,16 @@ struct intel_vgpu_scratch_pt {
 	unsigned long page_mfn;
 };
 
-
 struct intel_vgpu_gtt {
 	struct intel_vgpu_mm *ggtt_mm;
 	unsigned long active_ppgtt_mm_bitmap;
 	struct list_head mm_list_head;
 	DECLARE_HASHTABLE(shadow_page_hash_table, INTEL_GVT_GTT_HASH_BITS);
-	DECLARE_HASHTABLE(guest_page_hash_table, INTEL_GVT_GTT_HASH_BITS);
-	atomic_t n_write_protected_guest_page;
+	DECLARE_HASHTABLE(tracked_guest_page_hash_table, INTEL_GVT_GTT_HASH_BITS);
+	atomic_t n_tracked_guest_page;
 	struct list_head oos_page_list_head;
 	struct list_head post_shadow_list_head;
 	struct intel_vgpu_scratch_pt scratch_pt[GTT_TYPE_MAX];
-
 };
 
 extern int intel_vgpu_init_gtt(struct intel_vgpu *vgpu);
@@ -228,12 +226,16 @@ struct intel_vgpu_shadow_page {
 	unsigned long mfn;
 };
 
-struct intel_vgpu_guest_page {
+struct intel_vgpu_page_track {
 	struct hlist_node node;
-	bool writeprotection;
+	bool tracked;
 	unsigned long gfn;
 	int (*handler)(void *, u64, void *, int);
 	void *data;
+};
+
+struct intel_vgpu_guest_page {
+	struct intel_vgpu_page_track track;
 	unsigned long write_cnt;
 	struct intel_vgpu_oos_page *oos_page;
 };
@@ -243,7 +245,7 @@ struct intel_vgpu_oos_page {
 	struct list_head list;
 	struct list_head vm_list;
 	int id;
-	unsigned char mem[GTT_PAGE_SIZE];
+	unsigned char mem[I915_GTT_PAGE_SIZE];
 };
 
 #define GTT_ENTRY_NUM_IN_ONE_PAGE 512
@@ -258,22 +260,16 @@ struct intel_vgpu_ppgtt_spt {
 	struct list_head post_shadow_list;
 };
 
-int intel_vgpu_init_guest_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *guest_page,
+int intel_vgpu_init_page_track(struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t,
 		unsigned long gfn,
 		int (*handler)(void *gp, u64, void *, int),
 		void *data);
 
-void intel_vgpu_clean_guest_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *guest_page);
-
-int intel_vgpu_set_guest_page_writeprotection(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *guest_page);
+void intel_vgpu_clean_page_track(struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t);
 
-void intel_vgpu_clear_guest_page_writeprotection(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *guest_page);
-
-struct intel_vgpu_guest_page *intel_vgpu_find_guest_page(
+struct intel_vgpu_page_track *intel_vgpu_find_tracked_page(
 		struct intel_vgpu *vgpu, unsigned long gfn);
 
 int intel_vgpu_sync_oos_pages(struct intel_vgpu *vgpu);
@@ -312,4 +308,7 @@ int intel_vgpu_emulate_gtt_mmio_read(struct intel_vgpu *vgpu,
 int intel_vgpu_emulate_gtt_mmio_write(struct intel_vgpu *vgpu,
 	unsigned int off, void *p_data, unsigned int bytes);
 
+int intel_vgpu_write_protect_handler(struct intel_vgpu *vgpu, u64 pa,
+				     void *p_data, unsigned int bytes);
+
 #endif /* _GVT_GTT_H_ */
diff --git a/drivers/gpu/drm/i915/gvt/gvt.c b/drivers/gpu/drm/i915/gvt/gvt.c
index aaa347f8620c..fac54f32d33f 100644
--- a/drivers/gpu/drm/i915/gvt/gvt.c
+++ b/drivers/gpu/drm/i915/gvt/gvt.c
@@ -36,6 +36,8 @@
 
 #include "i915_drv.h"
 #include "gvt.h"
+#include <linux/vfio.h>
+#include <linux/mdev.h>
 
 struct intel_gvt_host intel_gvt_host;
 
@@ -44,6 +46,129 @@ static const char * const supported_hypervisors[] = {
 	[INTEL_GVT_HYPERVISOR_KVM] = "KVM",
 };
 
+static struct intel_vgpu_type *intel_gvt_find_vgpu_type(struct intel_gvt *gvt,
+		const char *name)
+{
+	int i;
+	struct intel_vgpu_type *t;
+	const char *driver_name = dev_driver_string(
+			&gvt->dev_priv->drm.pdev->dev);
+
+	for (i = 0; i < gvt->num_types; i++) {
+		t = &gvt->types[i];
+		if (!strncmp(t->name, name + strlen(driver_name) + 1,
+			sizeof(t->name)))
+			return t;
+	}
+
+	return NULL;
+}
+
+static ssize_t available_instances_show(struct kobject *kobj,
+					struct device *dev, char *buf)
+{
+	struct intel_vgpu_type *type;
+	unsigned int num = 0;
+	void *gvt = kdev_to_i915(dev)->gvt;
+
+	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
+	if (!type)
+		num = 0;
+	else
+		num = type->avail_instance;
+
+	return sprintf(buf, "%u\n", num);
+}
+
+static ssize_t device_api_show(struct kobject *kobj, struct device *dev,
+		char *buf)
+{
+	return sprintf(buf, "%s\n", VFIO_DEVICE_API_PCI_STRING);
+}
+
+static ssize_t description_show(struct kobject *kobj, struct device *dev,
+		char *buf)
+{
+	struct intel_vgpu_type *type;
+	void *gvt = kdev_to_i915(dev)->gvt;
+
+	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
+	if (!type)
+		return 0;
+
+	return sprintf(buf, "low_gm_size: %dMB\nhigh_gm_size: %dMB\n"
+		       "fence: %d\nresolution: %s\n"
+		       "weight: %d\n",
+		       BYTES_TO_MB(type->low_gm_size),
+		       BYTES_TO_MB(type->high_gm_size),
+		       type->fence, vgpu_edid_str(type->resolution),
+		       type->weight);
+}
+
+static MDEV_TYPE_ATTR_RO(available_instances);
+static MDEV_TYPE_ATTR_RO(device_api);
+static MDEV_TYPE_ATTR_RO(description);
+
+static struct attribute *gvt_type_attrs[] = {
+	&mdev_type_attr_available_instances.attr,
+	&mdev_type_attr_device_api.attr,
+	&mdev_type_attr_description.attr,
+	NULL,
+};
+
+static struct attribute_group *gvt_vgpu_type_groups[] = {
+	[0 ... NR_MAX_INTEL_VGPU_TYPES - 1] = NULL,
+};
+
+static bool intel_get_gvt_attrs(struct attribute ***type_attrs,
+		struct attribute_group ***intel_vgpu_type_groups)
+{
+	*type_attrs = gvt_type_attrs;
+	*intel_vgpu_type_groups = gvt_vgpu_type_groups;
+	return true;
+}
+
+static bool intel_gvt_init_vgpu_type_groups(struct intel_gvt *gvt)
+{
+	int i, j;
+	struct intel_vgpu_type *type;
+	struct attribute_group *group;
+
+	for (i = 0; i < gvt->num_types; i++) {
+		type = &gvt->types[i];
+
+		group = kzalloc(sizeof(struct attribute_group), GFP_KERNEL);
+		if (WARN_ON(!group))
+			goto unwind;
+
+		group->name = type->name;
+		group->attrs = gvt_type_attrs;
+		gvt_vgpu_type_groups[i] = group;
+	}
+
+	return true;
+
+unwind:
+	for (j = 0; j < i; j++) {
+		group = gvt_vgpu_type_groups[j];
+		kfree(group);
+	}
+
+	return false;
+}
+
+static void intel_gvt_cleanup_vgpu_type_groups(struct intel_gvt *gvt)
+{
+	int i;
+	struct attribute_group *group;
+
+	for (i = 0; i < gvt->num_types; i++) {
+		group = gvt_vgpu_type_groups[i];
+		gvt_vgpu_type_groups[i] = NULL;
+		kfree(group);
+	}
+}
+
 static const struct intel_gvt_ops intel_gvt_ops = {
 	.emulate_cfg_read = intel_vgpu_emulate_cfg_read,
 	.emulate_cfg_write = intel_vgpu_emulate_cfg_write,
@@ -54,6 +179,11 @@ static const struct intel_gvt_ops intel_gvt_ops = {
 	.vgpu_reset = intel_gvt_reset_vgpu,
 	.vgpu_activate = intel_gvt_activate_vgpu,
 	.vgpu_deactivate = intel_gvt_deactivate_vgpu,
+	.gvt_find_vgpu_type = intel_gvt_find_vgpu_type,
+	.get_gvt_attrs = intel_get_gvt_attrs,
+	.vgpu_query_plane = intel_vgpu_query_plane,
+	.vgpu_get_dmabuf = intel_vgpu_get_dmabuf,
+	.write_protect_handler = intel_vgpu_write_protect_handler,
 };
 
 /**
@@ -191,17 +321,18 @@ void intel_gvt_clean_device(struct drm_i915_private *dev_priv)
 	if (WARN_ON(!gvt))
 		return;
 
+	intel_gvt_debugfs_clean(gvt);
 	clean_service_thread(gvt);
 	intel_gvt_clean_cmd_parser(gvt);
 	intel_gvt_clean_sched_policy(gvt);
 	intel_gvt_clean_workload_scheduler(gvt);
-	intel_gvt_clean_opregion(gvt);
 	intel_gvt_clean_gtt(gvt);
 	intel_gvt_clean_irq(gvt);
 	intel_gvt_clean_mmio_info(gvt);
 	intel_gvt_free_firmware(gvt);
 
 	intel_gvt_hypervisor_host_exit(&dev_priv->drm.pdev->dev, gvt);
+	intel_gvt_cleanup_vgpu_type_groups(gvt);
 	intel_gvt_clean_vgpu_types(gvt);
 
 	idr_destroy(&gvt->vgpu_idr);
@@ -256,6 +387,8 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 	if (ret)
 		goto out_clean_idr;
 
+	intel_gvt_init_engine_mmio_context(gvt);
+
 	ret = intel_gvt_load_firmware(gvt);
 	if (ret)
 		goto out_clean_mmio_info;
@@ -268,13 +401,9 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 	if (ret)
 		goto out_clean_irq;
 
-	ret = intel_gvt_init_opregion(gvt);
-	if (ret)
-		goto out_clean_gtt;
-
 	ret = intel_gvt_init_workload_scheduler(gvt);
 	if (ret)
-		goto out_clean_opregion;
+		goto out_clean_gtt;
 
 	ret = intel_gvt_init_sched_policy(gvt);
 	if (ret)
@@ -292,6 +421,12 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 	if (ret)
 		goto out_clean_thread;
 
+	ret = intel_gvt_init_vgpu_type_groups(gvt);
+	if (ret == false) {
+		gvt_err("failed to init vgpu type groups: %d\n", ret);
+		goto out_clean_types;
+	}
+
 	ret = intel_gvt_hypervisor_host_init(&dev_priv->drm.pdev->dev, gvt,
 				&intel_gvt_ops);
 	if (ret) {
@@ -307,6 +442,10 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv)
 	}
 	gvt->idle_vgpu = vgpu;
 
+	ret = intel_gvt_debugfs_init(gvt);
+	if (ret)
+		gvt_err("debugfs registeration failed, go on.\n");
+
 	gvt_dbg_core("gvt device initialization is done\n");
 	dev_priv->gvt = gvt;
 	return 0;
@@ -321,8 +460,6 @@ out_clean_sched_policy:
 	intel_gvt_clean_sched_policy(gvt);
 out_clean_workload_scheduler:
 	intel_gvt_clean_workload_scheduler(gvt);
-out_clean_opregion:
-	intel_gvt_clean_opregion(gvt);
 out_clean_gtt:
 	intel_gvt_clean_gtt(gvt);
 out_clean_irq:
diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h
index 9c2e7c0aa38f..7dc7a80213a8 100644
--- a/drivers/gpu/drm/i915/gvt/gvt.h
+++ b/drivers/gpu/drm/i915/gvt/gvt.h
@@ -44,8 +44,10 @@
 #include "execlist.h"
 #include "scheduler.h"
 #include "sched_policy.h"
-#include "render.h"
+#include "mmio_context.h"
 #include "cmd_parser.h"
+#include "fb_decoder.h"
+#include "dmabuf.h"
 
 #define GVT_MAX_VGPU 8
 
@@ -123,9 +125,10 @@ struct intel_vgpu_irq {
 };
 
 struct intel_vgpu_opregion {
+	bool mapped;
 	void *va;
+	void *va_gopregion;
 	u32 gfn[INTEL_GVT_OPREGION_PAGES];
-	struct page *pages[INTEL_GVT_OPREGION_PAGES];
 };
 
 #define vgpu_opregion(vgpu) (&(vgpu->opregion))
@@ -142,6 +145,33 @@ struct vgpu_sched_ctl {
 	int weight;
 };
 
+enum {
+	INTEL_VGPU_EXECLIST_SUBMISSION = 1,
+	INTEL_VGPU_GUC_SUBMISSION,
+};
+
+struct intel_vgpu_submission_ops {
+	const char *name;
+	int (*init)(struct intel_vgpu *vgpu);
+	void (*clean)(struct intel_vgpu *vgpu);
+	void (*reset)(struct intel_vgpu *vgpu, unsigned long engine_mask);
+};
+
+struct intel_vgpu_submission {
+	struct intel_vgpu_execlist execlist[I915_NUM_ENGINES];
+	struct list_head workload_q_head[I915_NUM_ENGINES];
+	struct kmem_cache *workloads;
+	atomic_t running_workload_num;
+	struct i915_gem_context *shadow_ctx;
+	DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES);
+	DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
+	void *ring_scan_buffer[I915_NUM_ENGINES];
+	int ring_scan_buffer_size[I915_NUM_ENGINES];
+	const struct intel_vgpu_submission_ops *ops;
+	int virtual_submission_interface;
+	bool active;
+};
+
 struct intel_vgpu {
 	struct intel_gvt *gvt;
 	int id;
@@ -161,16 +191,10 @@ struct intel_vgpu {
 	struct intel_vgpu_gtt gtt;
 	struct intel_vgpu_opregion opregion;
 	struct intel_vgpu_display display;
-	struct intel_vgpu_execlist execlist[I915_NUM_ENGINES];
-	struct list_head workload_q_head[I915_NUM_ENGINES];
-	struct kmem_cache *workloads;
-	atomic_t running_workload_num;
-	/* 1/2K for each reserve ring buffer */
-	void *reserve_ring_buffer_va[I915_NUM_ENGINES];
-	int reserve_ring_buffer_size[I915_NUM_ENGINES];
-	DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
-	struct i915_gem_context *shadow_ctx;
-	DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES);
+	struct intel_vgpu_submission submission;
+	u32 hws_pga[I915_NUM_ENGINES];
+
+	struct dentry *debugfs;
 
 #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT)
 	struct {
@@ -186,10 +210,22 @@ struct intel_vgpu {
 		struct kvm *kvm;
 		struct work_struct release_work;
 		atomic_t released;
+		struct vfio_device *vfio_device;
 	} vdev;
 #endif
+
+	struct list_head dmabuf_obj_list_head;
+	struct mutex dmabuf_lock;
+	struct idr object_idr;
+
+	struct completion vblank_done;
+
 };
 
+/* validating GM healthy status*/
+#define vgpu_is_vm_unhealthy(ret_val) \
+	(((ret_val) == -EBADRQC) || ((ret_val) == -EFAULT))
+
 struct intel_gvt_gm {
 	unsigned long vgpu_allocated_low_gm_size;
 	unsigned long vgpu_allocated_high_gm_size;
@@ -231,7 +267,7 @@ struct intel_gvt_mmio {
 	unsigned int num_mmio_block;
 
 	DECLARE_HASHTABLE(mmio_info_table, INTEL_GVT_MMIO_HASH_BITS);
-	unsigned int num_tracked_mmio;
+	unsigned long num_tracked_mmio;
 };
 
 struct intel_gvt_firmware {
@@ -240,11 +276,6 @@ struct intel_gvt_firmware {
 	bool firmware_loaded;
 };
 
-struct intel_gvt_opregion {
-	void *opregion_va;
-	u32 opregion_pa;
-};
-
 #define NR_MAX_INTEL_VGPU_TYPES 20
 struct intel_vgpu_type {
 	char name[16];
@@ -268,7 +299,6 @@ struct intel_gvt {
 	struct intel_gvt_firmware firmware;
 	struct intel_gvt_irq irq;
 	struct intel_gvt_gtt gtt;
-	struct intel_gvt_opregion opregion;
 	struct intel_gvt_workload_scheduler scheduler;
 	struct notifier_block shadow_ctx_notifier_block[I915_NUM_ENGINES];
 	DECLARE_HASHTABLE(cmd_table, GVT_CMD_HASH_BITS);
@@ -279,6 +309,10 @@ struct intel_gvt {
 	struct task_struct *service_thread;
 	wait_queue_head_t service_thread_wq;
 	unsigned long service_request;
+
+	struct engine_mmio *engine_mmio_list;
+
+	struct dentry *debugfs_root;
 };
 
 static inline struct intel_gvt *to_gvt(struct drm_i915_private *i915)
@@ -316,7 +350,7 @@ int intel_gvt_load_firmware(struct intel_gvt *gvt);
 
 /* Aperture/GM space definitions for GVT device */
 #define gvt_aperture_sz(gvt)	  (gvt->dev_priv->ggtt.mappable_end)
-#define gvt_aperture_pa_base(gvt) (gvt->dev_priv->ggtt.mappable_base)
+#define gvt_aperture_pa_base(gvt) (gvt->dev_priv->ggtt.gmadr.start)
 
 #define gvt_ggtt_gm_sz(gvt)	  (gvt->dev_priv->ggtt.base.total)
 #define gvt_ggtt_sz(gvt) \
@@ -378,23 +412,20 @@ void intel_vgpu_free_resource(struct intel_vgpu *vgpu);
 void intel_vgpu_write_fence(struct intel_vgpu *vgpu,
 	u32 fence, u64 value);
 
-/* Macros for easily accessing vGPU virtual/shadow register */
-#define vgpu_vreg(vgpu, reg) \
-	(*(u32 *)(vgpu->mmio.vreg + INTEL_GVT_MMIO_OFFSET(reg)))
-#define vgpu_vreg8(vgpu, reg) \
-	(*(u8 *)(vgpu->mmio.vreg + INTEL_GVT_MMIO_OFFSET(reg)))
-#define vgpu_vreg16(vgpu, reg) \
-	(*(u16 *)(vgpu->mmio.vreg + INTEL_GVT_MMIO_OFFSET(reg)))
-#define vgpu_vreg64(vgpu, reg) \
-	(*(u64 *)(vgpu->mmio.vreg + INTEL_GVT_MMIO_OFFSET(reg)))
-#define vgpu_sreg(vgpu, reg) \
-	(*(u32 *)(vgpu->mmio.sreg + INTEL_GVT_MMIO_OFFSET(reg)))
-#define vgpu_sreg8(vgpu, reg) \
-	(*(u8 *)(vgpu->mmio.sreg + INTEL_GVT_MMIO_OFFSET(reg)))
-#define vgpu_sreg16(vgpu, reg) \
-	(*(u16 *)(vgpu->mmio.sreg + INTEL_GVT_MMIO_OFFSET(reg)))
-#define vgpu_sreg64(vgpu, reg) \
-	(*(u64 *)(vgpu->mmio.sreg + INTEL_GVT_MMIO_OFFSET(reg)))
+/* Macros for easily accessing vGPU virtual/shadow register.
+   Explicitly seperate use for typed MMIO reg or real offset.*/
+#define vgpu_vreg_t(vgpu, reg) \
+	(*(u32 *)(vgpu->mmio.vreg + i915_mmio_reg_offset(reg)))
+#define vgpu_vreg(vgpu, offset) \
+	(*(u32 *)(vgpu->mmio.vreg + (offset)))
+#define vgpu_vreg64_t(vgpu, reg) \
+	(*(u64 *)(vgpu->mmio.vreg + i915_mmio_reg_offset(reg)))
+#define vgpu_vreg64(vgpu, offset) \
+	(*(u64 *)(vgpu->mmio.vreg + (offset)))
+#define vgpu_sreg_t(vgpu, reg) \
+	(*(u32 *)(vgpu->mmio.sreg + i915_mmio_reg_offset(reg)))
+#define vgpu_sreg(vgpu, offset) \
+	(*(u32 *)(vgpu->mmio.sreg + (offset)))
 
 #define for_each_active_vgpu(gvt, vgpu, id) \
 	idr_for_each_entry((&(gvt)->vgpu_idr), (vgpu), (id)) \
@@ -484,16 +515,15 @@ static inline u64 intel_vgpu_get_bar_gpa(struct intel_vgpu *vgpu, int bar)
 			PCI_BASE_ADDRESS_MEM_MASK;
 }
 
-void intel_gvt_clean_opregion(struct intel_gvt *gvt);
-int intel_gvt_init_opregion(struct intel_gvt *gvt);
-
 void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu);
-int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa);
+int intel_vgpu_init_opregion(struct intel_vgpu *vgpu);
+int intel_vgpu_opregion_base_write_handler(struct intel_vgpu *vgpu, u32 gpa);
 
 int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci);
 void populate_pvinfo_page(struct intel_vgpu *vgpu);
 
 int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload);
+void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason);
 
 struct intel_gvt_ops {
 	int (*emulate_cfg_read)(struct intel_vgpu *, unsigned int, void *,
@@ -510,12 +540,21 @@ struct intel_gvt_ops {
 	void (*vgpu_reset)(struct intel_vgpu *);
 	void (*vgpu_activate)(struct intel_vgpu *);
 	void (*vgpu_deactivate)(struct intel_vgpu *);
+	struct intel_vgpu_type *(*gvt_find_vgpu_type)(struct intel_gvt *gvt,
+			const char *name);
+	bool (*get_gvt_attrs)(struct attribute ***type_attrs,
+			struct attribute_group ***intel_vgpu_type_groups);
+	int (*vgpu_query_plane)(struct intel_vgpu *vgpu, void *);
+	int (*vgpu_get_dmabuf)(struct intel_vgpu *vgpu, unsigned int);
+	int (*write_protect_handler)(struct intel_vgpu *, u64, void *,
+				     unsigned int);
 };
 
 
 enum {
 	GVT_FAILSAFE_UNSUPPORTED_GUEST,
 	GVT_FAILSAFE_INSUFFICIENT_RESOURCE,
+	GVT_FAILSAFE_GUEST_ERR,
 };
 
 static inline void mmio_hw_access_pre(struct drm_i915_private *dev_priv)
@@ -591,6 +630,12 @@ static inline bool intel_gvt_mmio_has_mode_mask(
 	return gvt->mmio.mmio_attribute[offset >> 2] & F_MODE_MASK;
 }
 
+int intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu);
+void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu);
+int intel_gvt_debugfs_init(struct intel_gvt *gvt);
+void intel_gvt_debugfs_clean(struct intel_gvt *gvt);
+
+
 #include "trace.h"
 #include "mpt.h"
 
diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c
index 1f840f6b81bb..92d6468daeee 100644
--- a/drivers/gpu/drm/i915/gvt/handlers.c
+++ b/drivers/gpu/drm/i915/gvt/handlers.c
@@ -166,7 +166,7 @@ int intel_gvt_render_mmio_to_ring_id(struct intel_gvt *gvt,
 	(num * 8 + i915_mmio_reg_offset(FENCE_REG_GEN6_LO(0)))
 
 
-static void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason)
+void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason)
 {
 	switch (reason) {
 	case GVT_FAILSAFE_UNSUPPORTED_GUEST:
@@ -174,6 +174,10 @@ static void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason)
 		break;
 	case GVT_FAILSAFE_INSUFFICIENT_RESOURCE:
 		pr_err("Graphics resource is not enough for the guest\n");
+		break;
+	case GVT_FAILSAFE_GUEST_ERR:
+		pr_err("GVT Internal error  for the guest\n");
+		break;
 	default:
 		break;
 	}
@@ -339,13 +343,13 @@ static int pch_pp_control_mmio_write(struct intel_vgpu *vgpu,
 	write_vreg(vgpu, offset, p_data, bytes);
 
 	if (vgpu_vreg(vgpu, offset) & PANEL_POWER_ON) {
-		vgpu_vreg(vgpu, PCH_PP_STATUS) |= PP_ON;
-		vgpu_vreg(vgpu, PCH_PP_STATUS) |= PP_SEQUENCE_STATE_ON_IDLE;
-		vgpu_vreg(vgpu, PCH_PP_STATUS) &= ~PP_SEQUENCE_POWER_DOWN;
-		vgpu_vreg(vgpu, PCH_PP_STATUS) &= ~PP_CYCLE_DELAY_ACTIVE;
+		vgpu_vreg_t(vgpu, PCH_PP_STATUS) |= PP_ON;
+		vgpu_vreg_t(vgpu, PCH_PP_STATUS) |= PP_SEQUENCE_STATE_ON_IDLE;
+		vgpu_vreg_t(vgpu, PCH_PP_STATUS) &= ~PP_SEQUENCE_POWER_DOWN;
+		vgpu_vreg_t(vgpu, PCH_PP_STATUS) &= ~PP_CYCLE_DELAY_ACTIVE;
 
 	} else
-		vgpu_vreg(vgpu, PCH_PP_STATUS) &=
+		vgpu_vreg_t(vgpu, PCH_PP_STATUS) &=
 			~(PP_ON | PP_SEQUENCE_POWER_DOWN
 					| PP_CYCLE_DELAY_ACTIVE);
 	return 0;
@@ -499,7 +503,7 @@ static int ddi_buf_ctl_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 	} else {
 		vgpu_vreg(vgpu, offset) |= DDI_BUF_IS_IDLE;
 		if (offset == i915_mmio_reg_offset(DDI_BUF_CTL(PORT_E)))
-			vgpu_vreg(vgpu, DP_TP_STATUS(PORT_E))
+			vgpu_vreg_t(vgpu, DP_TP_STATUS(PORT_E))
 				&= ~DP_TP_STATUS_AUTOTRAIN_DONE;
 	}
 	return 0;
@@ -517,9 +521,9 @@ static int fdi_rx_iir_mmio_write(struct intel_vgpu *vgpu,
 
 static int fdi_auto_training_started(struct intel_vgpu *vgpu)
 {
-	u32 ddi_buf_ctl = vgpu_vreg(vgpu, DDI_BUF_CTL(PORT_E));
+	u32 ddi_buf_ctl = vgpu_vreg_t(vgpu, DDI_BUF_CTL(PORT_E));
 	u32 rx_ctl = vgpu_vreg(vgpu, _FDI_RXA_CTL);
-	u32 tx_ctl = vgpu_vreg(vgpu, DP_TP_CTL(PORT_E));
+	u32 tx_ctl = vgpu_vreg_t(vgpu, DP_TP_CTL(PORT_E));
 
 	if ((ddi_buf_ctl & DDI_BUF_CTL_ENABLE) &&
 			(rx_ctl & FDI_RX_ENABLE) &&
@@ -560,12 +564,12 @@ static int check_fdi_rx_train_status(struct intel_vgpu *vgpu,
 	fdi_tx_check_bits = FDI_TX_ENABLE | fdi_tx_train_bits;
 
 	/* If imr bit has been masked */
-	if (vgpu_vreg(vgpu, fdi_rx_imr) & fdi_iir_check_bits)
+	if (vgpu_vreg_t(vgpu, fdi_rx_imr) & fdi_iir_check_bits)
 		return 0;
 
-	if (((vgpu_vreg(vgpu, fdi_tx_ctl) & fdi_tx_check_bits)
+	if (((vgpu_vreg_t(vgpu, fdi_tx_ctl) & fdi_tx_check_bits)
 			== fdi_tx_check_bits)
-		&& ((vgpu_vreg(vgpu, fdi_rx_ctl) & fdi_rx_check_bits)
+		&& ((vgpu_vreg_t(vgpu, fdi_rx_ctl) & fdi_rx_check_bits)
 			== fdi_rx_check_bits))
 		return 1;
 	else
@@ -622,17 +626,17 @@ static int update_fdi_rx_iir_status(struct intel_vgpu *vgpu,
 	if (ret < 0)
 		return ret;
 	if (ret)
-		vgpu_vreg(vgpu, fdi_rx_iir) |= FDI_RX_BIT_LOCK;
+		vgpu_vreg_t(vgpu, fdi_rx_iir) |= FDI_RX_BIT_LOCK;
 
 	ret = check_fdi_rx_train_status(vgpu, index, FDI_LINK_TRAIN_PATTERN2);
 	if (ret < 0)
 		return ret;
 	if (ret)
-		vgpu_vreg(vgpu, fdi_rx_iir) |= FDI_RX_SYMBOL_LOCK;
+		vgpu_vreg_t(vgpu, fdi_rx_iir) |= FDI_RX_SYMBOL_LOCK;
 
 	if (offset == _FDI_RXA_CTL)
 		if (fdi_auto_training_started(vgpu))
-			vgpu_vreg(vgpu, DP_TP_STATUS(PORT_E)) |=
+			vgpu_vreg_t(vgpu, DP_TP_STATUS(PORT_E)) |=
 				DP_TP_STATUS_AUTOTRAIN_DONE;
 	return 0;
 }
@@ -653,7 +657,7 @@ static int dp_tp_ctl_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 	data = (vgpu_vreg(vgpu, offset) & GENMASK(10, 8)) >> 8;
 	if (data == 0x2) {
 		status_reg = DP_TP_STATUS(index);
-		vgpu_vreg(vgpu, status_reg) |= (1 << 25);
+		vgpu_vreg_t(vgpu, status_reg) |= (1 << 25);
 	}
 	return 0;
 }
@@ -717,7 +721,7 @@ static int pri_surf_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 	};
 
 	write_vreg(vgpu, offset, p_data, bytes);
-	vgpu_vreg(vgpu, surflive_reg) = vgpu_vreg(vgpu, offset);
+	vgpu_vreg_t(vgpu, surflive_reg) = vgpu_vreg(vgpu, offset);
 
 	set_bit(flip_event[index], vgpu->irq.flip_done_event[index]);
 	return 0;
@@ -738,7 +742,7 @@ static int spr_surf_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 	};
 
 	write_vreg(vgpu, offset, p_data, bytes);
-	vgpu_vreg(vgpu, surflive_reg) = vgpu_vreg(vgpu, offset);
+	vgpu_vreg_t(vgpu, surflive_reg) = vgpu_vreg(vgpu, offset);
 
 	set_bit(flip_event[index], vgpu->irq.flip_done_event[index]);
 	return 0;
@@ -1060,9 +1064,9 @@ static void write_virtual_sbi_register(struct intel_vgpu *vgpu,
 static int sbi_data_mmio_read(struct intel_vgpu *vgpu, unsigned int offset,
 		void *p_data, unsigned int bytes)
 {
-	if (((vgpu_vreg(vgpu, SBI_CTL_STAT) & SBI_OPCODE_MASK) >>
+	if (((vgpu_vreg_t(vgpu, SBI_CTL_STAT) & SBI_OPCODE_MASK) >>
 				SBI_OPCODE_SHIFT) == SBI_CMD_CRRD) {
-		unsigned int sbi_offset = (vgpu_vreg(vgpu, SBI_ADDR) &
+		unsigned int sbi_offset = (vgpu_vreg_t(vgpu, SBI_ADDR) &
 				SBI_ADDR_OFFSET_MASK) >> SBI_ADDR_OFFSET_SHIFT;
 		vgpu_vreg(vgpu, offset) = read_virtual_sbi_register(vgpu,
 				sbi_offset);
@@ -1087,13 +1091,13 @@ static int sbi_ctl_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 
 	vgpu_vreg(vgpu, offset) = data;
 
-	if (((vgpu_vreg(vgpu, SBI_CTL_STAT) & SBI_OPCODE_MASK) >>
+	if (((vgpu_vreg_t(vgpu, SBI_CTL_STAT) & SBI_OPCODE_MASK) >>
 				SBI_OPCODE_SHIFT) == SBI_CMD_CRWR) {
-		unsigned int sbi_offset = (vgpu_vreg(vgpu, SBI_ADDR) &
+		unsigned int sbi_offset = (vgpu_vreg_t(vgpu, SBI_ADDR) &
 				SBI_ADDR_OFFSET_MASK) >> SBI_ADDR_OFFSET_SHIFT;
 
 		write_virtual_sbi_register(vgpu, sbi_offset,
-				vgpu_vreg(vgpu, SBI_DATA));
+					   vgpu_vreg_t(vgpu, SBI_DATA));
 	}
 	return 0;
 }
@@ -1339,7 +1343,7 @@ static int mailbox_write(struct intel_vgpu *vgpu, unsigned int offset,
 {
 	u32 value = *(u32 *)p_data;
 	u32 cmd = value & 0xff;
-	u32 *data0 = &vgpu_vreg(vgpu, GEN6_PCODE_DATA);
+	u32 *data0 = &vgpu_vreg_t(vgpu, GEN6_PCODE_DATA);
 
 	switch (cmd) {
 	case GEN9_PCODE_READ_MEM_LATENCY:
@@ -1378,6 +1382,34 @@ static int mailbox_write(struct intel_vgpu *vgpu, unsigned int offset,
 	return intel_vgpu_default_mmio_write(vgpu, offset, &value, bytes);
 }
 
+static int hws_pga_write(struct intel_vgpu *vgpu, unsigned int offset,
+		void *p_data, unsigned int bytes)
+{
+	u32 value = *(u32 *)p_data;
+	int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset);
+
+	if (!intel_gvt_ggtt_validate_range(vgpu, value, I915_GTT_PAGE_SIZE)) {
+		gvt_vgpu_err("VM(%d) write invalid HWSP address, reg:0x%x, value:0x%x\n",
+			      vgpu->id, offset, value);
+		return -EINVAL;
+	}
+	/*
+	 * Need to emulate all the HWSP register write to ensure host can
+	 * update the VM CSB status correctly. Here listed registers can
+	 * support BDW, SKL or other platforms with same HWSP registers.
+	 */
+	if (unlikely(ring_id < 0 || ring_id >= I915_NUM_ENGINES)) {
+		gvt_vgpu_err("VM(%d) access unknown hardware status page register:0x%x\n",
+			     vgpu->id, offset);
+		return -EINVAL;
+	}
+	vgpu->hws_pga[ring_id] = value;
+	gvt_dbg_mmio("VM(%d) write: 0x%x to HWSP: 0x%x\n",
+		     vgpu->id, value, offset);
+
+	return intel_vgpu_default_mmio_write(vgpu, offset, &value, bytes);
+}
+
 static int skl_power_well_ctl_write(struct intel_vgpu *vgpu,
 		unsigned int offset, void *p_data, unsigned int bytes)
 {
@@ -1441,12 +1473,12 @@ static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 	u32 data = *(u32 *)p_data;
 	int ret = 0;
 
-	if (WARN_ON(ring_id < 0 || ring_id > I915_NUM_ENGINES - 1))
+	if (WARN_ON(ring_id < 0 || ring_id >= I915_NUM_ENGINES))
 		return -EINVAL;
 
-	execlist = &vgpu->execlist[ring_id];
+	execlist = &vgpu->submission.execlist[ring_id];
 
-	execlist->elsp_dwords.data[execlist->elsp_dwords.index] = data;
+	execlist->elsp_dwords.data[3 - execlist->elsp_dwords.index] = data;
 	if (execlist->elsp_dwords.index == 3) {
 		ret = intel_vgpu_submit_execlist(vgpu, ring_id);
 		if(ret)
@@ -1462,9 +1494,11 @@ static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 		void *p_data, unsigned int bytes)
 {
+	struct intel_vgpu_submission *s = &vgpu->submission;
 	u32 data = *(u32 *)p_data;
 	int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset);
 	bool enable_execlist;
+	int ret;
 
 	write_vreg(vgpu, offset, p_data, bytes);
 
@@ -1486,8 +1520,18 @@ static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 				(enable_execlist ? "enabling" : "disabling"),
 				ring_id);
 
-		if (enable_execlist)
-			intel_vgpu_start_schedule(vgpu);
+		if (!enable_execlist)
+			return 0;
+
+		if (s->active)
+			return 0;
+
+		ret = intel_vgpu_select_submission_ops(vgpu,
+				INTEL_VGPU_EXECLIST_SUBMISSION);
+		if (ret)
+			return ret;
+
+		intel_vgpu_start_schedule(vgpu);
 	}
 	return 0;
 }
@@ -1519,7 +1563,7 @@ static int gvt_reg_tlb_control_handler(struct intel_vgpu *vgpu,
 	default:
 		return -EINVAL;
 	}
-	set_bit(id, (void *)vgpu->tlb_handle_pending);
+	set_bit(id, (void *)vgpu->submission.tlb_handle_pending);
 
 	return 0;
 }
@@ -1542,7 +1586,7 @@ static int ring_reset_ctl_write(struct intel_vgpu *vgpu,
 }
 
 #define MMIO_F(reg, s, f, am, rm, d, r, w) do { \
-	ret = new_mmio_info(gvt, INTEL_GVT_MMIO_OFFSET(reg), \
+	ret = new_mmio_info(gvt, i915_mmio_reg_offset(reg), \
 		f, s, am, rm, d, r, w); \
 	if (ret) \
 		return ret; \
@@ -1610,22 +1654,22 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_GM_RDR(BLT_HWS_PGA_GEN7, D_ALL, NULL, NULL);
 	MMIO_GM_RDR(VEBOX_HWS_PGA_GEN7, D_ALL, NULL, NULL);
 
-#define RING_REG(base) (base + 0x28)
+#define RING_REG(base) _MMIO((base) + 0x28)
 	MMIO_RING_DFH(RING_REG, D_ALL, F_CMD_ACCESS, NULL, NULL);
 #undef RING_REG
 
-#define RING_REG(base) (base + 0x134)
+#define RING_REG(base) _MMIO((base) + 0x134)
 	MMIO_RING_DFH(RING_REG, D_ALL, F_CMD_ACCESS, NULL, NULL);
 #undef RING_REG
 
-#define RING_REG(base) (base + 0x6c)
+#define RING_REG(base) _MMIO((base) + 0x6c)
 	MMIO_RING_DFH(RING_REG, D_ALL, 0, mmio_read_from_hw, NULL);
 #undef RING_REG
 	MMIO_DH(GEN7_SC_INSTDONE, D_BDW_PLUS, mmio_read_from_hw, NULL);
 
-	MMIO_GM_RDR(0x2148, D_ALL, NULL, NULL);
+	MMIO_GM_RDR(_MMIO(0x2148), D_ALL, NULL, NULL);
 	MMIO_GM_RDR(CCID, D_ALL, NULL, NULL);
-	MMIO_GM_RDR(0x12198, D_ALL, NULL, NULL);
+	MMIO_GM_RDR(_MMIO(0x12198), D_ALL, NULL, NULL);
 	MMIO_D(GEN7_CXT_SIZE, D_ALL);
 
 	MMIO_RING_DFH(RING_TAIL, D_ALL, F_CMD_ACCESS, NULL, NULL);
@@ -1635,7 +1679,7 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_RING_GM_RDR(RING_START, D_ALL, NULL, NULL);
 
 	/* RING MODE */
-#define RING_REG(base) (base + 0x29c)
+#define RING_REG(base) _MMIO((base) + 0x29c)
 	MMIO_RING_DFH(RING_REG, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL,
 		ring_mode_mmio_write);
 #undef RING_REG
@@ -1654,37 +1698,37 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 		NULL, NULL);
 	MMIO_DFH(CACHE_MODE_1, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(CACHE_MODE_0, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2124, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2124), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 
-	MMIO_DFH(0x20dc, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x20dc), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(_3D_CHICKEN3, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2088, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x20e4, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2470, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2088), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x20e4), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2470), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(GAM_ECOCHK, D_ALL, F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(GEN7_COMMON_SLICE_CHICKEN1, D_ALL, F_MODE_MASK | F_CMD_ACCESS,
 		NULL, NULL);
 	MMIO_DFH(COMMON_SLICE_CHICKEN2, D_ALL, F_MODE_MASK | F_CMD_ACCESS,
 		 NULL, NULL);
-	MMIO_DFH(0x9030, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x20a0, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2420, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2430, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2434, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2438, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x243c, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x7018, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x9030), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x20a0), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2420), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2430), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2434), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2438), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x243c), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x7018), D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(HALF_SLICE_CHICKEN3, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(GEN7_HALF_SLICE_CHICKEN1, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 
 	/* display */
-	MMIO_F(0x60220, 0x20, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_D(0x602a0, D_ALL);
+	MMIO_F(_MMIO(0x60220), 0x20, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_D(_MMIO(0x602a0), D_ALL);
 
-	MMIO_D(0x65050, D_ALL);
-	MMIO_D(0x650b4, D_ALL);
+	MMIO_D(_MMIO(0x65050), D_ALL);
+	MMIO_D(_MMIO(0x650b4), D_ALL);
 
-	MMIO_D(0xc4040, D_ALL);
+	MMIO_D(_MMIO(0xc4040), D_ALL);
 	MMIO_D(DERRMR, D_ALL);
 
 	MMIO_D(PIPEDSL(PIPE_A), D_ALL);
@@ -1724,14 +1768,14 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(CURBASE(PIPE_B), D_ALL);
 	MMIO_D(CURBASE(PIPE_C), D_ALL);
 
-	MMIO_D(0x700ac, D_ALL);
-	MMIO_D(0x710ac, D_ALL);
-	MMIO_D(0x720ac, D_ALL);
+	MMIO_D(_MMIO(0x700ac), D_ALL);
+	MMIO_D(_MMIO(0x710ac), D_ALL);
+	MMIO_D(_MMIO(0x720ac), D_ALL);
 
-	MMIO_D(0x70090, D_ALL);
-	MMIO_D(0x70094, D_ALL);
-	MMIO_D(0x70098, D_ALL);
-	MMIO_D(0x7009c, D_ALL);
+	MMIO_D(_MMIO(0x70090), D_ALL);
+	MMIO_D(_MMIO(0x70094), D_ALL);
+	MMIO_D(_MMIO(0x70098), D_ALL);
+	MMIO_D(_MMIO(0x7009c), D_ALL);
 
 	MMIO_D(DSPCNTR(PIPE_A), D_ALL);
 	MMIO_D(DSPADDR(PIPE_A), D_ALL);
@@ -1907,24 +1951,24 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(BLC_PWM_PCH_CTL1, D_ALL);
 	MMIO_D(BLC_PWM_PCH_CTL2, D_ALL);
 
-	MMIO_D(0x48268, D_ALL);
+	MMIO_D(_MMIO(0x48268), D_ALL);
 
 	MMIO_F(PCH_GMBUS0, 4 * 4, 0, 0, 0, D_ALL, gmbus_mmio_read,
 		gmbus_mmio_write);
 	MMIO_F(PCH_GPIOA, 6 * 4, F_UNALIGN, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0xe4f00, 0x28, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0xe4f00), 0x28, 0, 0, 0, D_ALL, NULL, NULL);
 
-	MMIO_F(_PCH_DPB_AUX_CH_CTL, 6 * 4, 0, 0, 0, D_PRE_SKL, NULL,
+	MMIO_F(_MMIO(_PCH_DPB_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_PRE_SKL, NULL,
 		dp_aux_ch_ctl_mmio_write);
-	MMIO_F(_PCH_DPC_AUX_CH_CTL, 6 * 4, 0, 0, 0, D_PRE_SKL, NULL,
+	MMIO_F(_MMIO(_PCH_DPC_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_PRE_SKL, NULL,
 		dp_aux_ch_ctl_mmio_write);
-	MMIO_F(_PCH_DPD_AUX_CH_CTL, 6 * 4, 0, 0, 0, D_PRE_SKL, NULL,
+	MMIO_F(_MMIO(_PCH_DPD_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_PRE_SKL, NULL,
 		dp_aux_ch_ctl_mmio_write);
 
 	MMIO_DH(PCH_ADPA, D_PRE_SKL, NULL, pch_adpa_mmio_write);
 
-	MMIO_DH(_PCH_TRANSACONF, D_ALL, NULL, transconf_mmio_write);
-	MMIO_DH(_PCH_TRANSBCONF, D_ALL, NULL, transconf_mmio_write);
+	MMIO_DH(_MMIO(_PCH_TRANSACONF), D_ALL, NULL, transconf_mmio_write);
+	MMIO_DH(_MMIO(_PCH_TRANSBCONF), D_ALL, NULL, transconf_mmio_write);
 
 	MMIO_DH(FDI_RX_IIR(PIPE_A), D_ALL, NULL, fdi_rx_iir_mmio_write);
 	MMIO_DH(FDI_RX_IIR(PIPE_B), D_ALL, NULL, fdi_rx_iir_mmio_write);
@@ -1936,30 +1980,30 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_DH(FDI_RX_CTL(PIPE_B), D_ALL, NULL, update_fdi_rx_iir_status);
 	MMIO_DH(FDI_RX_CTL(PIPE_C), D_ALL, NULL, update_fdi_rx_iir_status);
 
-	MMIO_D(_PCH_TRANS_HTOTAL_A, D_ALL);
-	MMIO_D(_PCH_TRANS_HBLANK_A, D_ALL);
-	MMIO_D(_PCH_TRANS_HSYNC_A, D_ALL);
-	MMIO_D(_PCH_TRANS_VTOTAL_A, D_ALL);
-	MMIO_D(_PCH_TRANS_VBLANK_A, D_ALL);
-	MMIO_D(_PCH_TRANS_VSYNC_A, D_ALL);
-	MMIO_D(_PCH_TRANS_VSYNCSHIFT_A, D_ALL);
-
-	MMIO_D(_PCH_TRANS_HTOTAL_B, D_ALL);
-	MMIO_D(_PCH_TRANS_HBLANK_B, D_ALL);
-	MMIO_D(_PCH_TRANS_HSYNC_B, D_ALL);
-	MMIO_D(_PCH_TRANS_VTOTAL_B, D_ALL);
-	MMIO_D(_PCH_TRANS_VBLANK_B, D_ALL);
-	MMIO_D(_PCH_TRANS_VSYNC_B, D_ALL);
-	MMIO_D(_PCH_TRANS_VSYNCSHIFT_B, D_ALL);
-
-	MMIO_D(_PCH_TRANSA_DATA_M1, D_ALL);
-	MMIO_D(_PCH_TRANSA_DATA_N1, D_ALL);
-	MMIO_D(_PCH_TRANSA_DATA_M2, D_ALL);
-	MMIO_D(_PCH_TRANSA_DATA_N2, D_ALL);
-	MMIO_D(_PCH_TRANSA_LINK_M1, D_ALL);
-	MMIO_D(_PCH_TRANSA_LINK_N1, D_ALL);
-	MMIO_D(_PCH_TRANSA_LINK_M2, D_ALL);
-	MMIO_D(_PCH_TRANSA_LINK_N2, D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_HTOTAL_A), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_HBLANK_A), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_HSYNC_A), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_VTOTAL_A), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_VBLANK_A), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_VSYNC_A), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_VSYNCSHIFT_A), D_ALL);
+
+	MMIO_D(_MMIO(_PCH_TRANS_HTOTAL_B), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_HBLANK_B), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_HSYNC_B), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_VTOTAL_B), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_VBLANK_B), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_VSYNC_B), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANS_VSYNCSHIFT_B), D_ALL);
+
+	MMIO_D(_MMIO(_PCH_TRANSA_DATA_M1), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANSA_DATA_N1), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANSA_DATA_M2), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANSA_DATA_N2), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANSA_LINK_M1), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANSA_LINK_N1), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANSA_LINK_M2), D_ALL);
+	MMIO_D(_MMIO(_PCH_TRANSA_LINK_N2), D_ALL);
 
 	MMIO_D(TRANS_DP_CTL(PIPE_A), D_ALL);
 	MMIO_D(TRANS_DP_CTL(PIPE_B), D_ALL);
@@ -1977,38 +2021,38 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(TVIDEO_DIP_DATA(PIPE_C), D_ALL);
 	MMIO_D(TVIDEO_DIP_GCP(PIPE_C), D_ALL);
 
-	MMIO_D(_FDI_RXA_MISC, D_ALL);
-	MMIO_D(_FDI_RXB_MISC, D_ALL);
-	MMIO_D(_FDI_RXA_TUSIZE1, D_ALL);
-	MMIO_D(_FDI_RXA_TUSIZE2, D_ALL);
-	MMIO_D(_FDI_RXB_TUSIZE1, D_ALL);
-	MMIO_D(_FDI_RXB_TUSIZE2, D_ALL);
+	MMIO_D(_MMIO(_FDI_RXA_MISC), D_ALL);
+	MMIO_D(_MMIO(_FDI_RXB_MISC), D_ALL);
+	MMIO_D(_MMIO(_FDI_RXA_TUSIZE1), D_ALL);
+	MMIO_D(_MMIO(_FDI_RXA_TUSIZE2), D_ALL);
+	MMIO_D(_MMIO(_FDI_RXB_TUSIZE1), D_ALL);
+	MMIO_D(_MMIO(_FDI_RXB_TUSIZE2), D_ALL);
 
 	MMIO_DH(PCH_PP_CONTROL, D_ALL, NULL, pch_pp_control_mmio_write);
 	MMIO_D(PCH_PP_DIVISOR, D_ALL);
 	MMIO_D(PCH_PP_STATUS,  D_ALL);
 	MMIO_D(PCH_LVDS, D_ALL);
-	MMIO_D(_PCH_DPLL_A, D_ALL);
-	MMIO_D(_PCH_DPLL_B, D_ALL);
-	MMIO_D(_PCH_FPA0, D_ALL);
-	MMIO_D(_PCH_FPA1, D_ALL);
-	MMIO_D(_PCH_FPB0, D_ALL);
-	MMIO_D(_PCH_FPB1, D_ALL);
+	MMIO_D(_MMIO(_PCH_DPLL_A), D_ALL);
+	MMIO_D(_MMIO(_PCH_DPLL_B), D_ALL);
+	MMIO_D(_MMIO(_PCH_FPA0), D_ALL);
+	MMIO_D(_MMIO(_PCH_FPA1), D_ALL);
+	MMIO_D(_MMIO(_PCH_FPB0), D_ALL);
+	MMIO_D(_MMIO(_PCH_FPB1), D_ALL);
 	MMIO_D(PCH_DREF_CONTROL, D_ALL);
 	MMIO_D(PCH_RAWCLK_FREQ, D_ALL);
 	MMIO_D(PCH_DPLL_SEL, D_ALL);
 
-	MMIO_D(0x61208, D_ALL);
-	MMIO_D(0x6120c, D_ALL);
+	MMIO_D(_MMIO(0x61208), D_ALL);
+	MMIO_D(_MMIO(0x6120c), D_ALL);
 	MMIO_D(PCH_PP_ON_DELAYS, D_ALL);
 	MMIO_D(PCH_PP_OFF_DELAYS, D_ALL);
 
-	MMIO_DH(0xe651c, D_ALL, dpy_reg_mmio_read, NULL);
-	MMIO_DH(0xe661c, D_ALL, dpy_reg_mmio_read, NULL);
-	MMIO_DH(0xe671c, D_ALL, dpy_reg_mmio_read, NULL);
-	MMIO_DH(0xe681c, D_ALL, dpy_reg_mmio_read, NULL);
-	MMIO_DH(0xe6c04, D_ALL, dpy_reg_mmio_read, NULL);
-	MMIO_DH(0xe6e1c, D_ALL, dpy_reg_mmio_read, NULL);
+	MMIO_DH(_MMIO(0xe651c), D_ALL, dpy_reg_mmio_read, NULL);
+	MMIO_DH(_MMIO(0xe661c), D_ALL, dpy_reg_mmio_read, NULL);
+	MMIO_DH(_MMIO(0xe671c), D_ALL, dpy_reg_mmio_read, NULL);
+	MMIO_DH(_MMIO(0xe681c), D_ALL, dpy_reg_mmio_read, NULL);
+	MMIO_DH(_MMIO(0xe6c04), D_ALL, dpy_reg_mmio_read, NULL);
+	MMIO_DH(_MMIO(0xe6e1c), D_ALL, dpy_reg_mmio_read, NULL);
 
 	MMIO_RO(PCH_PORT_HOTPLUG, D_ALL, 0,
 		PORTA_HOTPLUG_STATUS_MASK
@@ -2030,11 +2074,11 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 
 	MMIO_D(SOUTH_CHICKEN1, D_ALL);
 	MMIO_DH(SOUTH_CHICKEN2, D_ALL, NULL, south_chicken2_mmio_write);
-	MMIO_D(_TRANSA_CHICKEN1, D_ALL);
-	MMIO_D(_TRANSB_CHICKEN1, D_ALL);
+	MMIO_D(_MMIO(_TRANSA_CHICKEN1), D_ALL);
+	MMIO_D(_MMIO(_TRANSB_CHICKEN1), D_ALL);
 	MMIO_D(SOUTH_DSPCLK_GATE_D, D_ALL);
-	MMIO_D(_TRANSA_CHICKEN2, D_ALL);
-	MMIO_D(_TRANSB_CHICKEN2, D_ALL);
+	MMIO_D(_MMIO(_TRANSA_CHICKEN2), D_ALL);
+	MMIO_D(_MMIO(_TRANSB_CHICKEN2), D_ALL);
 
 	MMIO_D(ILK_DPFC_CB_BASE, D_ALL);
 	MMIO_D(ILK_DPFC_CONTROL, D_ALL);
@@ -2100,24 +2144,24 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(PREC_PAL_DATA(PIPE_C), D_ALL);
 	MMIO_F(PREC_PAL_GC_MAX(PIPE_C, 0), 4 * 3, 0, 0, 0, D_ALL, NULL, NULL);
 
-	MMIO_D(0x60110, D_ALL);
-	MMIO_D(0x61110, D_ALL);
-	MMIO_F(0x70400, 0x40, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x71400, 0x40, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x72400, 0x40, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x70440, 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
-	MMIO_F(0x71440, 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
-	MMIO_F(0x72440, 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
-	MMIO_F(0x7044c, 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
-	MMIO_F(0x7144c, 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
-	MMIO_F(0x7244c, 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
+	MMIO_D(_MMIO(0x60110), D_ALL);
+	MMIO_D(_MMIO(0x61110), D_ALL);
+	MMIO_F(_MMIO(0x70400), 0x40, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x71400), 0x40, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x72400), 0x40, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x70440), 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
+	MMIO_F(_MMIO(0x71440), 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
+	MMIO_F(_MMIO(0x72440), 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
+	MMIO_F(_MMIO(0x7044c), 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
+	MMIO_F(_MMIO(0x7144c), 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
+	MMIO_F(_MMIO(0x7244c), 0xc, 0, 0, 0, D_PRE_SKL, NULL, NULL);
 
 	MMIO_D(PIPE_WM_LINETIME(PIPE_A), D_ALL);
 	MMIO_D(PIPE_WM_LINETIME(PIPE_B), D_ALL);
 	MMIO_D(PIPE_WM_LINETIME(PIPE_C), D_ALL);
 	MMIO_D(SPLL_CTL, D_ALL);
-	MMIO_D(_WRPLL_CTL1, D_ALL);
-	MMIO_D(_WRPLL_CTL2, D_ALL);
+	MMIO_D(_MMIO(_WRPLL_CTL1), D_ALL);
+	MMIO_D(_MMIO(_WRPLL_CTL2), D_ALL);
 	MMIO_D(PORT_CLK_SEL(PORT_A), D_ALL);
 	MMIO_D(PORT_CLK_SEL(PORT_B), D_ALL);
 	MMIO_D(PORT_CLK_SEL(PORT_C), D_ALL);
@@ -2128,15 +2172,15 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(TRANS_CLK_SEL(TRANSCODER_C), D_ALL);
 
 	MMIO_D(HSW_NDE_RSTWRN_OPT, D_ALL);
-	MMIO_D(0x46508, D_ALL);
+	MMIO_D(_MMIO(0x46508), D_ALL);
 
-	MMIO_D(0x49080, D_ALL);
-	MMIO_D(0x49180, D_ALL);
-	MMIO_D(0x49280, D_ALL);
+	MMIO_D(_MMIO(0x49080), D_ALL);
+	MMIO_D(_MMIO(0x49180), D_ALL);
+	MMIO_D(_MMIO(0x49280), D_ALL);
 
-	MMIO_F(0x49090, 0x14, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x49190, 0x14, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x49290, 0x14, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x49090), 0x14, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x49190), 0x14, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x49290), 0x14, 0, 0, 0, D_ALL, NULL, NULL);
 
 	MMIO_D(GAMMA_MODE(PIPE_A), D_ALL);
 	MMIO_D(GAMMA_MODE(PIPE_B), D_ALL);
@@ -2156,7 +2200,7 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_DH(SBI_CTL_STAT, D_ALL, NULL, sbi_ctl_mmio_write);
 	MMIO_D(PIXCLK_GATE, D_ALL);
 
-	MMIO_F(_DPA_AUX_CH_CTL, 6 * 4, 0, 0, 0, D_ALL, NULL,
+	MMIO_F(_MMIO(_DPA_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_ALL, NULL,
 		dp_aux_ch_ctl_mmio_write);
 
 	MMIO_DH(DDI_BUF_CTL(PORT_A), D_ALL, NULL, ddi_buf_ctl_mmio_write);
@@ -2177,24 +2221,24 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_DH(DP_TP_STATUS(PORT_D), D_ALL, NULL, dp_tp_status_mmio_write);
 	MMIO_DH(DP_TP_STATUS(PORT_E), D_ALL, NULL, NULL);
 
-	MMIO_F(_DDI_BUF_TRANS_A, 0x50, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x64e60, 0x50, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x64eC0, 0x50, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x64f20, 0x50, 0, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x64f80, 0x50, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(_DDI_BUF_TRANS_A), 0x50, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x64e60), 0x50, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x64eC0), 0x50, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x64f20), 0x50, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x64f80), 0x50, 0, 0, 0, D_ALL, NULL, NULL);
 
 	MMIO_D(HSW_AUD_CFG(PIPE_A), D_ALL);
 	MMIO_D(HSW_AUD_PIN_ELD_CP_VLD, D_ALL);
 
-	MMIO_DH(_TRANS_DDI_FUNC_CTL_A, D_ALL, NULL, NULL);
-	MMIO_DH(_TRANS_DDI_FUNC_CTL_B, D_ALL, NULL, NULL);
-	MMIO_DH(_TRANS_DDI_FUNC_CTL_C, D_ALL, NULL, NULL);
-	MMIO_DH(_TRANS_DDI_FUNC_CTL_EDP, D_ALL, NULL, NULL);
+	MMIO_DH(_MMIO(_TRANS_DDI_FUNC_CTL_A), D_ALL, NULL, NULL);
+	MMIO_DH(_MMIO(_TRANS_DDI_FUNC_CTL_B), D_ALL, NULL, NULL);
+	MMIO_DH(_MMIO(_TRANS_DDI_FUNC_CTL_C), D_ALL, NULL, NULL);
+	MMIO_DH(_MMIO(_TRANS_DDI_FUNC_CTL_EDP), D_ALL, NULL, NULL);
 
-	MMIO_D(_TRANSA_MSA_MISC, D_ALL);
-	MMIO_D(_TRANSB_MSA_MISC, D_ALL);
-	MMIO_D(_TRANSC_MSA_MISC, D_ALL);
-	MMIO_D(_TRANS_EDP_MSA_MISC, D_ALL);
+	MMIO_D(_MMIO(_TRANSA_MSA_MISC), D_ALL);
+	MMIO_D(_MMIO(_TRANSB_MSA_MISC), D_ALL);
+	MMIO_D(_MMIO(_TRANSC_MSA_MISC), D_ALL);
+	MMIO_D(_MMIO(_TRANS_EDP_MSA_MISC), D_ALL);
 
 	MMIO_DH(FORCEWAKE, D_ALL, NULL, NULL);
 	MMIO_D(FORCEWAKE_ACK, D_ALL);
@@ -2260,101 +2304,101 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(GEN6_UCGCTL1, D_ALL);
 	MMIO_D(GEN6_UCGCTL2, D_ALL);
 
-	MMIO_F(0x4f000, 0x90, 0, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x4f000), 0x90, 0, 0, 0, D_ALL, NULL, NULL);
 
 	MMIO_D(GEN6_PCODE_DATA, D_ALL);
-	MMIO_D(0x13812c, D_ALL);
+	MMIO_D(_MMIO(0x13812c), D_ALL);
 	MMIO_DH(GEN7_ERR_INT, D_ALL, NULL, NULL);
 	MMIO_D(HSW_EDRAM_CAP, D_ALL);
 	MMIO_D(HSW_IDICR, D_ALL);
 	MMIO_DH(GFX_FLSH_CNTL_GEN6, D_ALL, NULL, NULL);
 
-	MMIO_D(0x3c, D_ALL);
-	MMIO_D(0x860, D_ALL);
+	MMIO_D(_MMIO(0x3c), D_ALL);
+	MMIO_D(_MMIO(0x860), D_ALL);
 	MMIO_D(ECOSKPD, D_ALL);
-	MMIO_D(0x121d0, D_ALL);
+	MMIO_D(_MMIO(0x121d0), D_ALL);
 	MMIO_D(GEN6_BLITTER_ECOSKPD, D_ALL);
-	MMIO_D(0x41d0, D_ALL);
+	MMIO_D(_MMIO(0x41d0), D_ALL);
 	MMIO_D(GAC_ECO_BITS, D_ALL);
-	MMIO_D(0x6200, D_ALL);
-	MMIO_D(0x6204, D_ALL);
-	MMIO_D(0x6208, D_ALL);
-	MMIO_D(0x7118, D_ALL);
-	MMIO_D(0x7180, D_ALL);
-	MMIO_D(0x7408, D_ALL);
-	MMIO_D(0x7c00, D_ALL);
+	MMIO_D(_MMIO(0x6200), D_ALL);
+	MMIO_D(_MMIO(0x6204), D_ALL);
+	MMIO_D(_MMIO(0x6208), D_ALL);
+	MMIO_D(_MMIO(0x7118), D_ALL);
+	MMIO_D(_MMIO(0x7180), D_ALL);
+	MMIO_D(_MMIO(0x7408), D_ALL);
+	MMIO_D(_MMIO(0x7c00), D_ALL);
 	MMIO_DH(GEN6_MBCTL, D_ALL, NULL, mbctl_write);
-	MMIO_D(0x911c, D_ALL);
-	MMIO_D(0x9120, D_ALL);
+	MMIO_D(_MMIO(0x911c), D_ALL);
+	MMIO_D(_MMIO(0x9120), D_ALL);
 	MMIO_DFH(GEN7_UCGCTL4, D_ALL, F_CMD_ACCESS, NULL, NULL);
 
 	MMIO_D(GAB_CTL, D_ALL);
-	MMIO_D(0x48800, D_ALL);
-	MMIO_D(0xce044, D_ALL);
-	MMIO_D(0xe6500, D_ALL);
-	MMIO_D(0xe6504, D_ALL);
-	MMIO_D(0xe6600, D_ALL);
-	MMIO_D(0xe6604, D_ALL);
-	MMIO_D(0xe6700, D_ALL);
-	MMIO_D(0xe6704, D_ALL);
-	MMIO_D(0xe6800, D_ALL);
-	MMIO_D(0xe6804, D_ALL);
+	MMIO_D(_MMIO(0x48800), D_ALL);
+	MMIO_D(_MMIO(0xce044), D_ALL);
+	MMIO_D(_MMIO(0xe6500), D_ALL);
+	MMIO_D(_MMIO(0xe6504), D_ALL);
+	MMIO_D(_MMIO(0xe6600), D_ALL);
+	MMIO_D(_MMIO(0xe6604), D_ALL);
+	MMIO_D(_MMIO(0xe6700), D_ALL);
+	MMIO_D(_MMIO(0xe6704), D_ALL);
+	MMIO_D(_MMIO(0xe6800), D_ALL);
+	MMIO_D(_MMIO(0xe6804), D_ALL);
 	MMIO_D(PCH_GMBUS4, D_ALL);
 	MMIO_D(PCH_GMBUS5, D_ALL);
 
-	MMIO_D(0x902c, D_ALL);
-	MMIO_D(0xec008, D_ALL);
-	MMIO_D(0xec00c, D_ALL);
-	MMIO_D(0xec008 + 0x18, D_ALL);
-	MMIO_D(0xec00c + 0x18, D_ALL);
-	MMIO_D(0xec008 + 0x18 * 2, D_ALL);
-	MMIO_D(0xec00c + 0x18 * 2, D_ALL);
-	MMIO_D(0xec008 + 0x18 * 3, D_ALL);
-	MMIO_D(0xec00c + 0x18 * 3, D_ALL);
-	MMIO_D(0xec408, D_ALL);
-	MMIO_D(0xec40c, D_ALL);
-	MMIO_D(0xec408 + 0x18, D_ALL);
-	MMIO_D(0xec40c + 0x18, D_ALL);
-	MMIO_D(0xec408 + 0x18 * 2, D_ALL);
-	MMIO_D(0xec40c + 0x18 * 2, D_ALL);
-	MMIO_D(0xec408 + 0x18 * 3, D_ALL);
-	MMIO_D(0xec40c + 0x18 * 3, D_ALL);
-	MMIO_D(0xfc810, D_ALL);
-	MMIO_D(0xfc81c, D_ALL);
-	MMIO_D(0xfc828, D_ALL);
-	MMIO_D(0xfc834, D_ALL);
-	MMIO_D(0xfcc00, D_ALL);
-	MMIO_D(0xfcc0c, D_ALL);
-	MMIO_D(0xfcc18, D_ALL);
-	MMIO_D(0xfcc24, D_ALL);
-	MMIO_D(0xfd000, D_ALL);
-	MMIO_D(0xfd00c, D_ALL);
-	MMIO_D(0xfd018, D_ALL);
-	MMIO_D(0xfd024, D_ALL);
-	MMIO_D(0xfd034, D_ALL);
+	MMIO_D(_MMIO(0x902c), D_ALL);
+	MMIO_D(_MMIO(0xec008), D_ALL);
+	MMIO_D(_MMIO(0xec00c), D_ALL);
+	MMIO_D(_MMIO(0xec008 + 0x18), D_ALL);
+	MMIO_D(_MMIO(0xec00c + 0x18), D_ALL);
+	MMIO_D(_MMIO(0xec008 + 0x18 * 2), D_ALL);
+	MMIO_D(_MMIO(0xec00c + 0x18 * 2), D_ALL);
+	MMIO_D(_MMIO(0xec008 + 0x18 * 3), D_ALL);
+	MMIO_D(_MMIO(0xec00c + 0x18 * 3), D_ALL);
+	MMIO_D(_MMIO(0xec408), D_ALL);
+	MMIO_D(_MMIO(0xec40c), D_ALL);
+	MMIO_D(_MMIO(0xec408 + 0x18), D_ALL);
+	MMIO_D(_MMIO(0xec40c + 0x18), D_ALL);
+	MMIO_D(_MMIO(0xec408 + 0x18 * 2), D_ALL);
+	MMIO_D(_MMIO(0xec40c + 0x18 * 2), D_ALL);
+	MMIO_D(_MMIO(0xec408 + 0x18 * 3), D_ALL);
+	MMIO_D(_MMIO(0xec40c + 0x18 * 3), D_ALL);
+	MMIO_D(_MMIO(0xfc810), D_ALL);
+	MMIO_D(_MMIO(0xfc81c), D_ALL);
+	MMIO_D(_MMIO(0xfc828), D_ALL);
+	MMIO_D(_MMIO(0xfc834), D_ALL);
+	MMIO_D(_MMIO(0xfcc00), D_ALL);
+	MMIO_D(_MMIO(0xfcc0c), D_ALL);
+	MMIO_D(_MMIO(0xfcc18), D_ALL);
+	MMIO_D(_MMIO(0xfcc24), D_ALL);
+	MMIO_D(_MMIO(0xfd000), D_ALL);
+	MMIO_D(_MMIO(0xfd00c), D_ALL);
+	MMIO_D(_MMIO(0xfd018), D_ALL);
+	MMIO_D(_MMIO(0xfd024), D_ALL);
+	MMIO_D(_MMIO(0xfd034), D_ALL);
 
 	MMIO_DH(FPGA_DBG, D_ALL, NULL, fpga_dbg_mmio_write);
-	MMIO_D(0x2054, D_ALL);
-	MMIO_D(0x12054, D_ALL);
-	MMIO_D(0x22054, D_ALL);
-	MMIO_D(0x1a054, D_ALL);
-
-	MMIO_D(0x44070, D_ALL);
-	MMIO_DFH(0x215c, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2178, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x217c, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x12178, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x1217c, D_ALL, F_CMD_ACCESS, NULL, NULL);
-
-	MMIO_F(0x2290, 8, F_CMD_ACCESS, 0, 0, D_BDW_PLUS, NULL, NULL);
-	MMIO_D(0x2b00, D_BDW_PLUS);
-	MMIO_D(0x2360, D_BDW_PLUS);
-	MMIO_F(0x5200, 32, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x5240, 32, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
-	MMIO_F(0x5280, 16, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
-
-	MMIO_DFH(0x1c17c, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x1c178, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_D(_MMIO(0x2054), D_ALL);
+	MMIO_D(_MMIO(0x12054), D_ALL);
+	MMIO_D(_MMIO(0x22054), D_ALL);
+	MMIO_D(_MMIO(0x1a054), D_ALL);
+
+	MMIO_D(_MMIO(0x44070), D_ALL);
+	MMIO_DFH(_MMIO(0x215c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2178), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x217c), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x12178), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x1217c), D_ALL, F_CMD_ACCESS, NULL, NULL);
+
+	MMIO_F(_MMIO(0x2290), 8, F_CMD_ACCESS, 0, 0, D_BDW_PLUS, NULL, NULL);
+	MMIO_D(_MMIO(0x2b00), D_BDW_PLUS);
+	MMIO_D(_MMIO(0x2360), D_BDW_PLUS);
+	MMIO_F(_MMIO(0x5200), 32, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x5240), 32, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+	MMIO_F(_MMIO(0x5280), 16, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
+
+	MMIO_DFH(_MMIO(0x1c17c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x1c178), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(BCS_SWCTRL, D_ALL, F_CMD_ACCESS, NULL, NULL);
 
 	MMIO_F(HS_INVOCATION_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
@@ -2368,24 +2412,24 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_F(CL_PRIMITIVES_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
 	MMIO_F(PS_INVOCATION_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
 	MMIO_F(PS_DEPTH_COUNT, 8, F_CMD_ACCESS, 0, 0, D_ALL, NULL, NULL);
-	MMIO_DH(0x4260, D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
-	MMIO_DH(0x4264, D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
-	MMIO_DH(0x4268, D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
-	MMIO_DH(0x426c, D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
-	MMIO_DH(0x4270, D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
-	MMIO_DFH(0x4094, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DH(_MMIO(0x4260), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+	MMIO_DH(_MMIO(0x4264), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+	MMIO_DH(_MMIO(0x4268), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+	MMIO_DH(_MMIO(0x426c), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+	MMIO_DH(_MMIO(0x4270), D_BDW_PLUS, NULL, gvt_reg_tlb_control_handler);
+	MMIO_DFH(_MMIO(0x4094), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
 
 	MMIO_DFH(ARB_MODE, D_ALL, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 	MMIO_RING_GM_RDR(RING_BBADDR, D_ALL, NULL, NULL);
-	MMIO_DFH(0x2220, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x12220, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x22220, D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2220), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x12220), D_ALL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x22220), D_ALL, F_CMD_ACCESS, NULL, NULL);
 	MMIO_RING_DFH(RING_SYNC_1, D_ALL, F_CMD_ACCESS, NULL, NULL);
 	MMIO_RING_DFH(RING_SYNC_0, D_ALL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x22178, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x1a178, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x1a17c, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2217c, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x22178), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x1a178), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x1a17c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2217c), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
 	return 0;
 }
 
@@ -2459,40 +2503,40 @@ static int init_broadwell_mmio_info(struct intel_gvt *gvt)
 	MMIO_RING_DFH(RING_ACTHD_UDW, D_BDW_PLUS, F_CMD_ACCESS,
 		mmio_read_from_hw, NULL);
 
-#define RING_REG(base) (base + 0xd0)
+#define RING_REG(base) _MMIO((base) + 0xd0)
 	MMIO_RING_F(RING_REG, 4, F_RO, 0,
 		~_MASKED_BIT_ENABLE(RESET_CTL_REQUEST_RESET), D_BDW_PLUS, NULL,
 		ring_reset_ctl_write);
 #undef RING_REG
 
-#define RING_REG(base) (base + 0x230)
+#define RING_REG(base) _MMIO((base) + 0x230)
 	MMIO_RING_DFH(RING_REG, D_BDW_PLUS, 0, NULL, elsp_mmio_write);
 #undef RING_REG
 
-#define RING_REG(base) (base + 0x234)
+#define RING_REG(base) _MMIO((base) + 0x234)
 	MMIO_RING_F(RING_REG, 8, F_RO | F_CMD_ACCESS, 0, ~0, D_BDW_PLUS,
 		NULL, NULL);
 #undef RING_REG
 
-#define RING_REG(base) (base + 0x244)
+#define RING_REG(base) _MMIO((base) + 0x244)
 	MMIO_RING_DFH(RING_REG, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
 #undef RING_REG
 
-#define RING_REG(base) (base + 0x370)
+#define RING_REG(base) _MMIO((base) + 0x370)
 	MMIO_RING_F(RING_REG, 48, F_RO, 0, ~0, D_BDW_PLUS, NULL, NULL);
 #undef RING_REG
 
-#define RING_REG(base) (base + 0x3a0)
+#define RING_REG(base) _MMIO((base) + 0x3a0)
 	MMIO_RING_DFH(RING_REG, D_BDW_PLUS, F_MODE_MASK, NULL, NULL);
 #undef RING_REG
 
 	MMIO_D(PIPEMISC(PIPE_A), D_BDW_PLUS);
 	MMIO_D(PIPEMISC(PIPE_B), D_BDW_PLUS);
 	MMIO_D(PIPEMISC(PIPE_C), D_BDW_PLUS);
-	MMIO_D(0x1c1d0, D_BDW_PLUS);
+	MMIO_D(_MMIO(0x1c1d0), D_BDW_PLUS);
 	MMIO_D(GEN6_MBCUNIT_SNPCR, D_BDW_PLUS);
 	MMIO_D(GEN7_MISCCPCTL, D_BDW_PLUS);
-	MMIO_D(0x1c054, D_BDW_PLUS);
+	MMIO_D(_MMIO(0x1c054), D_BDW_PLUS);
 
 	MMIO_DH(GEN6_PCODE_MAILBOX, D_BDW_PLUS, NULL, mailbox_write);
 
@@ -2501,11 +2545,11 @@ static int init_broadwell_mmio_info(struct intel_gvt *gvt)
 
 	MMIO_D(GAMTARBMODE, D_BDW_PLUS);
 
-#define RING_REG(base) (base + 0x270)
+#define RING_REG(base) _MMIO((base) + 0x270)
 	MMIO_RING_F(RING_REG, 32, 0, 0, 0, D_BDW_PLUS, NULL, NULL);
 #undef RING_REG
 
-	MMIO_RING_GM_RDR(RING_HWS_PGA, D_BDW_PLUS, NULL, NULL);
+	MMIO_RING_GM_RDR(RING_HWS_PGA, D_BDW_PLUS, NULL, hws_pga_write);
 
 	MMIO_DFH(HDC_CHICKEN0, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 
@@ -2514,10 +2558,10 @@ static int init_broadwell_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(CHICKEN_PIPESL_1(PIPE_C), D_BDW_PLUS);
 
 	MMIO_D(WM_MISC, D_BDW);
-	MMIO_D(BDW_EDP_PSR_BASE, D_BDW);
+	MMIO_D(_MMIO(BDW_EDP_PSR_BASE), D_BDW);
 
-	MMIO_D(0x66c00, D_BDW_PLUS);
-	MMIO_D(0x66c04, D_BDW_PLUS);
+	MMIO_D(_MMIO(0x66c00), D_BDW_PLUS);
+	MMIO_D(_MMIO(0x66c04), D_BDW_PLUS);
 
 	MMIO_D(HSW_GTT_CACHE_EN, D_BDW_PLUS);
 
@@ -2525,54 +2569,54 @@ static int init_broadwell_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(GEN8_EU_DISABLE1, D_BDW_PLUS);
 	MMIO_D(GEN8_EU_DISABLE2, D_BDW_PLUS);
 
-	MMIO_D(0xfdc, D_BDW_PLUS);
+	MMIO_D(_MMIO(0xfdc), D_BDW_PLUS);
 	MMIO_DFH(GEN8_ROW_CHICKEN, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS,
 		NULL, NULL);
 	MMIO_DFH(GEN7_ROW_CHICKEN2, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS,
 		NULL, NULL);
 	MMIO_DFH(GEN8_UCGCTL6, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
 
-	MMIO_DFH(0xb1f0, D_BDW, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xb1c0, D_BDW, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xb1f0), D_BDW, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xb1c0), D_BDW, F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(GEN8_L3SQCREG4, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xb100, D_BDW, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xb10c, D_BDW, F_CMD_ACCESS, NULL, NULL);
-	MMIO_D(0xb110, D_BDW);
+	MMIO_DFH(_MMIO(0xb100), D_BDW, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xb10c), D_BDW, F_CMD_ACCESS, NULL, NULL);
+	MMIO_D(_MMIO(0xb110), D_BDW);
 
-	MMIO_F(0x24d0, 48, F_CMD_ACCESS, 0, 0, D_BDW_PLUS,
+	MMIO_F(_MMIO(0x24d0), 48, F_CMD_ACCESS, 0, 0, D_BDW_PLUS,
 		NULL, force_nonpriv_write);
 
-	MMIO_D(0x44484, D_BDW_PLUS);
-	MMIO_D(0x4448c, D_BDW_PLUS);
+	MMIO_D(_MMIO(0x44484), D_BDW_PLUS);
+	MMIO_D(_MMIO(0x4448c), D_BDW_PLUS);
 
-	MMIO_DFH(0x83a4, D_BDW, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x83a4), D_BDW, F_CMD_ACCESS, NULL, NULL);
 	MMIO_D(GEN8_L3_LRA_1_GPGPU, D_BDW_PLUS);
 
-	MMIO_DFH(0x8430, D_BDW, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x8430), D_BDW, F_CMD_ACCESS, NULL, NULL);
 
-	MMIO_D(0x110000, D_BDW_PLUS);
+	MMIO_D(_MMIO(0x110000), D_BDW_PLUS);
 
-	MMIO_D(0x48400, D_BDW_PLUS);
+	MMIO_D(_MMIO(0x48400), D_BDW_PLUS);
 
-	MMIO_D(0x6e570, D_BDW_PLUS);
-	MMIO_D(0x65f10, D_BDW_PLUS);
+	MMIO_D(_MMIO(0x6e570), D_BDW_PLUS);
+	MMIO_D(_MMIO(0x65f10), D_BDW_PLUS);
 
-	MMIO_DFH(0xe194, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xe188, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe194), D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe188), D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(HALF_SLICE_CHICKEN2, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x2580, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
-
-	MMIO_DFH(0x2248, D_BDW, F_CMD_ACCESS, NULL, NULL);
-
-	MMIO_DFH(0xe220, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xe230, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xe240, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xe260, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xe270, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xe280, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xe2a0, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xe2b0, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0xe2c0, D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x2580), D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+
+	MMIO_DFH(_MMIO(0x2248), D_BDW, F_CMD_ACCESS, NULL, NULL);
+
+	MMIO_DFH(_MMIO(0xe220), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe230), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe240), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe260), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe270), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe280), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe2a0), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe2b0), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0xe2c0), D_BDW_PLUS, F_CMD_ACCESS, NULL, NULL);
 	return 0;
 }
 
@@ -2588,11 +2632,11 @@ static int init_skl_mmio_info(struct intel_gvt *gvt)
 	MMIO_DH(FORCEWAKE_MEDIA_GEN9, D_SKL_PLUS, NULL, mul_force_wake_write);
 	MMIO_DH(FORCEWAKE_ACK_MEDIA_GEN9, D_SKL_PLUS, NULL, NULL);
 
-	MMIO_F(_DPB_AUX_CH_CTL, 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
+	MMIO_F(_MMIO(_DPB_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
 						dp_aux_ch_ctl_mmio_write);
-	MMIO_F(_DPC_AUX_CH_CTL, 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
+	MMIO_F(_MMIO(_DPC_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
 						dp_aux_ch_ctl_mmio_write);
-	MMIO_F(_DPD_AUX_CH_CTL, 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
+	MMIO_F(_MMIO(_DPD_AUX_CH_CTL), 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
 						dp_aux_ch_ctl_mmio_write);
 
 	/*
@@ -2603,26 +2647,26 @@ static int init_skl_mmio_info(struct intel_gvt *gvt)
 	MMIO_DH(HSW_PWR_WELL_CTL_DRIVER(SKL_DISP_PW_MISC_IO), D_SKL_PLUS, NULL,
 		skl_power_well_ctl_write);
 
-	MMIO_D(0xa210, D_SKL_PLUS);
+	MMIO_D(_MMIO(0xa210), D_SKL_PLUS);
 	MMIO_D(GEN9_MEDIA_PG_IDLE_HYSTERESIS, D_SKL_PLUS);
 	MMIO_D(GEN9_RENDER_PG_IDLE_HYSTERESIS, D_SKL_PLUS);
 	MMIO_DFH(GEN9_GAMT_ECO_REG_RW_IA, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DH(0x4ddc, D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(0x42080, D_SKL_PLUS, NULL, NULL);
-	MMIO_D(0x45504, D_SKL_PLUS);
-	MMIO_D(0x45520, D_SKL_PLUS);
-	MMIO_D(0x46000, D_SKL_PLUS);
-	MMIO_DH(0x46010, D_SKL | D_KBL, NULL, skl_lcpll_write);
-	MMIO_DH(0x46014, D_SKL | D_KBL, NULL, skl_lcpll_write);
-	MMIO_D(0x6C040, D_SKL | D_KBL);
-	MMIO_D(0x6C048, D_SKL | D_KBL);
-	MMIO_D(0x6C050, D_SKL | D_KBL);
-	MMIO_D(0x6C044, D_SKL | D_KBL);
-	MMIO_D(0x6C04C, D_SKL | D_KBL);
-	MMIO_D(0x6C054, D_SKL | D_KBL);
-	MMIO_D(0x6c058, D_SKL | D_KBL);
-	MMIO_D(0x6c05c, D_SKL | D_KBL);
-	MMIO_DH(0X6c060, D_SKL | D_KBL, dpll_status_read, NULL);
+	MMIO_DH(_MMIO(0x4ddc), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(0x42080), D_SKL_PLUS, NULL, NULL);
+	MMIO_D(_MMIO(0x45504), D_SKL_PLUS);
+	MMIO_D(_MMIO(0x45520), D_SKL_PLUS);
+	MMIO_D(_MMIO(0x46000), D_SKL_PLUS);
+	MMIO_DH(_MMIO(0x46010), D_SKL | D_KBL, NULL, skl_lcpll_write);
+	MMIO_DH(_MMIO(0x46014), D_SKL | D_KBL, NULL, skl_lcpll_write);
+	MMIO_D(_MMIO(0x6C040), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6C048), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6C050), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6C044), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6C04C), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6C054), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6c058), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6c05c), D_SKL | D_KBL);
+	MMIO_DH(_MMIO(0x6c060), D_SKL | D_KBL, dpll_status_read, NULL);
 
 	MMIO_DH(SKL_PS_WIN_POS(PIPE_A, 0), D_SKL_PLUS, NULL, pf_write);
 	MMIO_DH(SKL_PS_WIN_POS(PIPE_A, 1), D_SKL_PLUS, NULL, pf_write);
@@ -2711,105 +2755,105 @@ static int init_skl_mmio_info(struct intel_gvt *gvt)
 	MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_C, 2), D_SKL_PLUS, NULL, NULL);
 	MMIO_DH(PLANE_NV12_BUF_CFG(PIPE_C, 3), D_SKL_PLUS, NULL, NULL);
 
-	MMIO_DH(_REG_701C0(PIPE_A, 1), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C0(PIPE_A, 2), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C0(PIPE_A, 3), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C0(PIPE_A, 4), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_A, 1)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_A, 2)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_A, 3)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_A, 4)), D_SKL_PLUS, NULL, NULL);
 
-	MMIO_DH(_REG_701C0(PIPE_B, 1), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C0(PIPE_B, 2), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C0(PIPE_B, 3), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C0(PIPE_B, 4), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_B, 1)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_B, 2)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_B, 3)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_B, 4)), D_SKL_PLUS, NULL, NULL);
 
-	MMIO_DH(_REG_701C0(PIPE_C, 1), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C0(PIPE_C, 2), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C0(PIPE_C, 3), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C0(PIPE_C, 4), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_C, 1)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_C, 2)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_C, 3)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C0(PIPE_C, 4)), D_SKL_PLUS, NULL, NULL);
 
-	MMIO_DH(_REG_701C4(PIPE_A, 1), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C4(PIPE_A, 2), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C4(PIPE_A, 3), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C4(PIPE_A, 4), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_A, 1)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_A, 2)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_A, 3)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_A, 4)), D_SKL_PLUS, NULL, NULL);
 
-	MMIO_DH(_REG_701C4(PIPE_B, 1), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C4(PIPE_B, 2), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C4(PIPE_B, 3), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C4(PIPE_B, 4), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_B, 1)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_B, 2)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_B, 3)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_B, 4)), D_SKL_PLUS, NULL, NULL);
 
-	MMIO_DH(_REG_701C4(PIPE_C, 1), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C4(PIPE_C, 2), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C4(PIPE_C, 3), D_SKL_PLUS, NULL, NULL);
-	MMIO_DH(_REG_701C4(PIPE_C, 4), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_C, 1)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_C, 2)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_C, 3)), D_SKL_PLUS, NULL, NULL);
+	MMIO_DH(_MMIO(_REG_701C4(PIPE_C, 4)), D_SKL_PLUS, NULL, NULL);
 
-	MMIO_D(0x70380, D_SKL_PLUS);
-	MMIO_D(0x71380, D_SKL_PLUS);
-	MMIO_D(0x72380, D_SKL_PLUS);
-	MMIO_D(0x7039c, D_SKL_PLUS);
+	MMIO_D(_MMIO(0x70380), D_SKL_PLUS);
+	MMIO_D(_MMIO(0x71380), D_SKL_PLUS);
+	MMIO_D(_MMIO(0x72380), D_SKL_PLUS);
+	MMIO_D(_MMIO(0x7039c), D_SKL_PLUS);
 
-	MMIO_D(0x8f074, D_SKL | D_KBL);
-	MMIO_D(0x8f004, D_SKL | D_KBL);
-	MMIO_D(0x8f034, D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x8f074), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x8f004), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x8f034), D_SKL | D_KBL);
 
-	MMIO_D(0xb11c, D_SKL | D_KBL);
+	MMIO_D(_MMIO(0xb11c), D_SKL | D_KBL);
 
-	MMIO_D(0x51000, D_SKL | D_KBL);
-	MMIO_D(0x6c00c, D_SKL_PLUS);
+	MMIO_D(_MMIO(0x51000), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6c00c), D_SKL_PLUS);
 
-	MMIO_F(0xc800, 0x7f8, F_CMD_ACCESS, 0, 0, D_SKL | D_KBL, NULL, NULL);
-	MMIO_F(0xb020, 0x80, F_CMD_ACCESS, 0, 0, D_SKL | D_KBL, NULL, NULL);
+	MMIO_F(_MMIO(0xc800), 0x7f8, F_CMD_ACCESS, 0, 0, D_SKL | D_KBL, NULL, NULL);
+	MMIO_F(_MMIO(0xb020), 0x80, F_CMD_ACCESS, 0, 0, D_SKL | D_KBL, NULL, NULL);
 
-	MMIO_D(0xd08, D_SKL_PLUS);
-	MMIO_DFH(0x20e0, D_SKL_PLUS, F_MODE_MASK, NULL, NULL);
-	MMIO_DFH(0x20ec, D_SKL_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
+	MMIO_D(_MMIO(0xd08), D_SKL_PLUS);
+	MMIO_DFH(_MMIO(0x20e0), D_SKL_PLUS, F_MODE_MASK, NULL, NULL);
+	MMIO_DFH(_MMIO(0x20ec), D_SKL_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL);
 
 	/* TRTT */
-	MMIO_DFH(0x4de0, D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x4de4, D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x4de8, D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x4dec, D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x4df0, D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(0x4df4, D_SKL | D_KBL, F_CMD_ACCESS, NULL, gen9_trtte_write);
-	MMIO_DH(0x4dfc, D_SKL | D_KBL, NULL, gen9_trtt_chicken_write);
+	MMIO_DFH(_MMIO(0x4de0), D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x4de4), D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x4de8), D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x4dec), D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x4df0), D_SKL | D_KBL, F_CMD_ACCESS, NULL, NULL);
+	MMIO_DFH(_MMIO(0x4df4), D_SKL | D_KBL, F_CMD_ACCESS, NULL, gen9_trtte_write);
+	MMIO_DH(_MMIO(0x4dfc), D_SKL | D_KBL, NULL, gen9_trtt_chicken_write);
 
-	MMIO_D(0x45008, D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x45008), D_SKL | D_KBL);
 
-	MMIO_D(0x46430, D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x46430), D_SKL | D_KBL);
 
-	MMIO_D(0x46520, D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x46520), D_SKL | D_KBL);
 
-	MMIO_D(0xc403c, D_SKL | D_KBL);
-	MMIO_D(0xb004, D_SKL_PLUS);
+	MMIO_D(_MMIO(0xc403c), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0xb004), D_SKL_PLUS);
 	MMIO_DH(DMA_CTRL, D_SKL_PLUS, NULL, dma_ctrl_write);
 
-	MMIO_D(0x65900, D_SKL_PLUS);
-	MMIO_D(0x1082c0, D_SKL | D_KBL);
-	MMIO_D(0x4068, D_SKL | D_KBL);
-	MMIO_D(0x67054, D_SKL | D_KBL);
-	MMIO_D(0x6e560, D_SKL | D_KBL);
-	MMIO_D(0x6e554, D_SKL | D_KBL);
-	MMIO_D(0x2b20, D_SKL | D_KBL);
-	MMIO_D(0x65f00, D_SKL | D_KBL);
-	MMIO_D(0x65f08, D_SKL | D_KBL);
-	MMIO_D(0x320f0, D_SKL | D_KBL);
-
-	MMIO_D(0x70034, D_SKL_PLUS);
-	MMIO_D(0x71034, D_SKL_PLUS);
-	MMIO_D(0x72034, D_SKL_PLUS);
-
-	MMIO_D(_PLANE_KEYVAL_1(PIPE_A), D_SKL_PLUS);
-	MMIO_D(_PLANE_KEYVAL_1(PIPE_B), D_SKL_PLUS);
-	MMIO_D(_PLANE_KEYVAL_1(PIPE_C), D_SKL_PLUS);
-	MMIO_D(_PLANE_KEYMSK_1(PIPE_A), D_SKL_PLUS);
-	MMIO_D(_PLANE_KEYMSK_1(PIPE_B), D_SKL_PLUS);
-	MMIO_D(_PLANE_KEYMSK_1(PIPE_C), D_SKL_PLUS);
-
-	MMIO_D(0x44500, D_SKL_PLUS);
+	MMIO_D(_MMIO(0x65900), D_SKL_PLUS);
+	MMIO_D(_MMIO(0x1082c0), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x4068), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x67054), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6e560), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x6e554), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x2b20), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x65f00), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x65f08), D_SKL | D_KBL);
+	MMIO_D(_MMIO(0x320f0), D_SKL | D_KBL);
+
+	MMIO_D(_MMIO(0x70034), D_SKL_PLUS);
+	MMIO_D(_MMIO(0x71034), D_SKL_PLUS);
+	MMIO_D(_MMIO(0x72034), D_SKL_PLUS);
+
+	MMIO_D(_MMIO(_PLANE_KEYVAL_1(PIPE_A)), D_SKL_PLUS);
+	MMIO_D(_MMIO(_PLANE_KEYVAL_1(PIPE_B)), D_SKL_PLUS);
+	MMIO_D(_MMIO(_PLANE_KEYVAL_1(PIPE_C)), D_SKL_PLUS);
+	MMIO_D(_MMIO(_PLANE_KEYMSK_1(PIPE_A)), D_SKL_PLUS);
+	MMIO_D(_MMIO(_PLANE_KEYMSK_1(PIPE_B)), D_SKL_PLUS);
+	MMIO_D(_MMIO(_PLANE_KEYMSK_1(PIPE_C)), D_SKL_PLUS);
+
+	MMIO_D(_MMIO(0x44500), D_SKL_PLUS);
 	MMIO_DFH(GEN9_CSFE_CHICKEN1_RCS, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
 	MMIO_DFH(GEN8_HDC_CHICKEN1, D_SKL | D_KBL, F_MODE_MASK | F_CMD_ACCESS,
 		NULL, NULL);
 
-	MMIO_D(0x4ab8, D_KBL);
-	MMIO_D(0x2248, D_SKL_PLUS | D_KBL);
+	MMIO_D(_MMIO(0x4ab8), D_KBL);
+	MMIO_D(_MMIO(0x2248), D_SKL_PLUS | D_KBL);
 
 	return 0;
 }
@@ -2825,8 +2869,8 @@ static struct gvt_mmio_block *find_mmio_block(struct intel_gvt *gvt,
 	for (i = 0; i < num; i++, block++) {
 		if (!(device & block->device))
 			continue;
-		if (offset >= INTEL_GVT_MMIO_OFFSET(block->offset) &&
-		    offset < INTEL_GVT_MMIO_OFFSET(block->offset) + block->size)
+		if (offset >= i915_mmio_reg_offset(block->offset) &&
+		    offset < i915_mmio_reg_offset(block->offset) + block->size)
 			return block;
 	}
 	return NULL;
@@ -2906,14 +2950,46 @@ int intel_gvt_setup_mmio_info(struct intel_gvt *gvt)
 	gvt->mmio.mmio_block = mmio_blocks;
 	gvt->mmio.num_mmio_block = ARRAY_SIZE(mmio_blocks);
 
-	gvt_dbg_mmio("traced %u virtual mmio registers\n",
-		     gvt->mmio.num_tracked_mmio);
 	return 0;
 err:
 	intel_gvt_clean_mmio_info(gvt);
 	return ret;
 }
 
+/**
+ * intel_gvt_for_each_tracked_mmio - iterate each tracked mmio
+ * @gvt: a GVT device
+ * @handler: the handler
+ * @data: private data given to handler
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt,
+	int (*handler)(struct intel_gvt *gvt, u32 offset, void *data),
+	void *data)
+{
+	struct gvt_mmio_block *block = gvt->mmio.mmio_block;
+	struct intel_gvt_mmio_info *e;
+	int i, j, ret;
+
+	hash_for_each(gvt->mmio.mmio_info_table, i, e, node) {
+		ret = handler(gvt, e->offset, data);
+		if (ret)
+			return ret;
+	}
+
+	for (i = 0; i < gvt->mmio.num_mmio_block; i++, block++) {
+		for (j = 0; j < block->size; j += 4) {
+			ret = handler(gvt,
+				      i915_mmio_reg_offset(block->offset) + j,
+				      data);
+			if (ret)
+				return ret;
+		}
+	}
+	return 0;
+}
 
 /**
  * intel_vgpu_default_mmio_read - default MMIO read handler
diff --git a/drivers/gpu/drm/i915/gvt/hypercall.h b/drivers/gpu/drm/i915/gvt/hypercall.h
index df7f33abd393..a1bd82feb827 100644
--- a/drivers/gpu/drm/i915/gvt/hypercall.h
+++ b/drivers/gpu/drm/i915/gvt/hypercall.h
@@ -55,6 +55,9 @@ struct intel_gvt_mpt {
 			      unsigned long mfn, unsigned int nr, bool map);
 	int (*set_trap_area)(unsigned long handle, u64 start, u64 end,
 			     bool map);
+	int (*set_opregion)(void *vgpu);
+	int (*get_vfio_device)(void *vgpu);
+	void (*put_vfio_device)(void *vgpu);
 };
 
 extern struct intel_gvt_mpt xengt_mpt;
diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
index 0a7d084da1a2..554d1db1f3c8 100644
--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -53,11 +53,23 @@ static const struct intel_gvt_ops *intel_gvt_ops;
 #define VFIO_PCI_INDEX_TO_OFFSET(index) ((u64)(index) << VFIO_PCI_OFFSET_SHIFT)
 #define VFIO_PCI_OFFSET_MASK    (((u64)(1) << VFIO_PCI_OFFSET_SHIFT) - 1)
 
+#define OPREGION_SIGNATURE "IntelGraphicsMem"
+
+struct vfio_region;
+struct intel_vgpu_regops {
+	size_t (*rw)(struct intel_vgpu *vgpu, char *buf,
+			size_t count, loff_t *ppos, bool iswrite);
+	void (*release)(struct intel_vgpu *vgpu,
+			struct vfio_region *region);
+};
+
 struct vfio_region {
 	u32				type;
 	u32				subtype;
 	size_t				size;
 	u32				flags;
+	const struct intel_vgpu_regops	*ops;
+	void				*data;
 };
 
 struct kvmgt_pgfn {
@@ -248,120 +260,6 @@ static void gvt_cache_destroy(struct intel_vgpu *vgpu)
 	}
 }
 
-static struct intel_vgpu_type *intel_gvt_find_vgpu_type(struct intel_gvt *gvt,
-		const char *name)
-{
-	int i;
-	struct intel_vgpu_type *t;
-	const char *driver_name = dev_driver_string(
-			&gvt->dev_priv->drm.pdev->dev);
-
-	for (i = 0; i < gvt->num_types; i++) {
-		t = &gvt->types[i];
-		if (!strncmp(t->name, name + strlen(driver_name) + 1,
-			sizeof(t->name)))
-			return t;
-	}
-
-	return NULL;
-}
-
-static ssize_t available_instances_show(struct kobject *kobj,
-					struct device *dev, char *buf)
-{
-	struct intel_vgpu_type *type;
-	unsigned int num = 0;
-	void *gvt = kdev_to_i915(dev)->gvt;
-
-	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
-	if (!type)
-		num = 0;
-	else
-		num = type->avail_instance;
-
-	return sprintf(buf, "%u\n", num);
-}
-
-static ssize_t device_api_show(struct kobject *kobj, struct device *dev,
-		char *buf)
-{
-	return sprintf(buf, "%s\n", VFIO_DEVICE_API_PCI_STRING);
-}
-
-static ssize_t description_show(struct kobject *kobj, struct device *dev,
-		char *buf)
-{
-	struct intel_vgpu_type *type;
-	void *gvt = kdev_to_i915(dev)->gvt;
-
-	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
-	if (!type)
-		return 0;
-
-	return sprintf(buf, "low_gm_size: %dMB\nhigh_gm_size: %dMB\n"
-		       "fence: %d\nresolution: %s\n"
-		       "weight: %d\n",
-		       BYTES_TO_MB(type->low_gm_size),
-		       BYTES_TO_MB(type->high_gm_size),
-		       type->fence, vgpu_edid_str(type->resolution),
-		       type->weight);
-}
-
-static MDEV_TYPE_ATTR_RO(available_instances);
-static MDEV_TYPE_ATTR_RO(device_api);
-static MDEV_TYPE_ATTR_RO(description);
-
-static struct attribute *type_attrs[] = {
-	&mdev_type_attr_available_instances.attr,
-	&mdev_type_attr_device_api.attr,
-	&mdev_type_attr_description.attr,
-	NULL,
-};
-
-static struct attribute_group *intel_vgpu_type_groups[] = {
-	[0 ... NR_MAX_INTEL_VGPU_TYPES - 1] = NULL,
-};
-
-static bool intel_gvt_init_vgpu_type_groups(struct intel_gvt *gvt)
-{
-	int i, j;
-	struct intel_vgpu_type *type;
-	struct attribute_group *group;
-
-	for (i = 0; i < gvt->num_types; i++) {
-		type = &gvt->types[i];
-
-		group = kzalloc(sizeof(struct attribute_group), GFP_KERNEL);
-		if (WARN_ON(!group))
-			goto unwind;
-
-		group->name = type->name;
-		group->attrs = type_attrs;
-		intel_vgpu_type_groups[i] = group;
-	}
-
-	return true;
-
-unwind:
-	for (j = 0; j < i; j++) {
-		group = intel_vgpu_type_groups[j];
-		kfree(group);
-	}
-
-	return false;
-}
-
-static void intel_gvt_cleanup_vgpu_type_groups(struct intel_gvt *gvt)
-{
-	int i;
-	struct attribute_group *group;
-
-	for (i = 0; i < gvt->num_types; i++) {
-		group = intel_vgpu_type_groups[i];
-		kfree(group);
-	}
-}
-
 static void kvmgt_protect_table_init(struct kvmgt_guest_info *info)
 {
 	hash_init(info->ptable);
@@ -430,6 +328,108 @@ static void kvmgt_protect_table_del(struct kvmgt_guest_info *info,
 	}
 }
 
+static size_t intel_vgpu_reg_rw_opregion(struct intel_vgpu *vgpu, char *buf,
+		size_t count, loff_t *ppos, bool iswrite)
+{
+	unsigned int i = VFIO_PCI_OFFSET_TO_INDEX(*ppos) -
+			VFIO_PCI_NUM_REGIONS;
+	void *base = vgpu->vdev.region[i].data;
+	loff_t pos = *ppos & VFIO_PCI_OFFSET_MASK;
+
+	if (pos >= vgpu->vdev.region[i].size || iswrite) {
+		gvt_vgpu_err("invalid op or offset for Intel vgpu OpRegion\n");
+		return -EINVAL;
+	}
+	count = min(count, (size_t)(vgpu->vdev.region[i].size - pos));
+	memcpy(buf, base + pos, count);
+
+	return count;
+}
+
+static void intel_vgpu_reg_release_opregion(struct intel_vgpu *vgpu,
+		struct vfio_region *region)
+{
+}
+
+static const struct intel_vgpu_regops intel_vgpu_regops_opregion = {
+	.rw = intel_vgpu_reg_rw_opregion,
+	.release = intel_vgpu_reg_release_opregion,
+};
+
+static int intel_vgpu_register_reg(struct intel_vgpu *vgpu,
+		unsigned int type, unsigned int subtype,
+		const struct intel_vgpu_regops *ops,
+		size_t size, u32 flags, void *data)
+{
+	struct vfio_region *region;
+
+	region = krealloc(vgpu->vdev.region,
+			(vgpu->vdev.num_regions + 1) * sizeof(*region),
+			GFP_KERNEL);
+	if (!region)
+		return -ENOMEM;
+
+	vgpu->vdev.region = region;
+	vgpu->vdev.region[vgpu->vdev.num_regions].type = type;
+	vgpu->vdev.region[vgpu->vdev.num_regions].subtype = subtype;
+	vgpu->vdev.region[vgpu->vdev.num_regions].ops = ops;
+	vgpu->vdev.region[vgpu->vdev.num_regions].size = size;
+	vgpu->vdev.region[vgpu->vdev.num_regions].flags = flags;
+	vgpu->vdev.region[vgpu->vdev.num_regions].data = data;
+	vgpu->vdev.num_regions++;
+	return 0;
+}
+
+static int kvmgt_get_vfio_device(void *p_vgpu)
+{
+	struct intel_vgpu *vgpu = (struct intel_vgpu *)p_vgpu;
+
+	vgpu->vdev.vfio_device = vfio_device_get_from_dev(
+		mdev_dev(vgpu->vdev.mdev));
+	if (!vgpu->vdev.vfio_device) {
+		gvt_vgpu_err("failed to get vfio device\n");
+		return -ENODEV;
+	}
+	return 0;
+}
+
+
+static int kvmgt_set_opregion(void *p_vgpu)
+{
+	struct intel_vgpu *vgpu = (struct intel_vgpu *)p_vgpu;
+	void *base;
+	int ret;
+
+	/* Each vgpu has its own opregion, although VFIO would create another
+	 * one later. This one is used to expose opregion to VFIO. And the
+	 * other one created by VFIO later, is used by guest actually.
+	 */
+	base = vgpu_opregion(vgpu)->va;
+	if (!base)
+		return -ENOMEM;
+
+	if (memcmp(base, OPREGION_SIGNATURE, 16)) {
+		memunmap(base);
+		return -EINVAL;
+	}
+
+	ret = intel_vgpu_register_reg(vgpu,
+			PCI_VENDOR_ID_INTEL | VFIO_REGION_TYPE_PCI_VENDOR_TYPE,
+			VFIO_REGION_SUBTYPE_INTEL_IGD_OPREGION,
+			&intel_vgpu_regops_opregion, OPREGION_SIZE,
+			VFIO_REGION_INFO_FLAG_READ, base);
+
+	return ret;
+}
+
+static void kvmgt_put_vfio_device(void *vgpu)
+{
+	if (WARN_ON(!((struct intel_vgpu *)vgpu)->vdev.vfio_device))
+		return;
+
+	vfio_device_put(((struct intel_vgpu *)vgpu)->vdev.vfio_device);
+}
+
 static int intel_vgpu_create(struct kobject *kobj, struct mdev_device *mdev)
 {
 	struct intel_vgpu *vgpu = NULL;
@@ -441,7 +441,7 @@ static int intel_vgpu_create(struct kobject *kobj, struct mdev_device *mdev)
 	pdev = mdev_parent_dev(mdev);
 	gvt = kdev_to_i915(pdev)->gvt;
 
-	type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj));
+	type = intel_gvt_ops->gvt_find_vgpu_type(gvt, kobject_name(kobj));
 	if (!type) {
 		gvt_vgpu_err("failed to find type %s to create\n",
 						kobject_name(kobj));
@@ -660,7 +660,7 @@ static ssize_t intel_vgpu_rw(struct mdev_device *mdev, char *buf,
 	int ret = -EINVAL;
 
 
-	if (index >= VFIO_PCI_NUM_REGIONS) {
+	if (index >= VFIO_PCI_NUM_REGIONS + vgpu->vdev.num_regions) {
 		gvt_vgpu_err("invalid index: %u\n", index);
 		return -EINVAL;
 	}
@@ -688,8 +688,14 @@ static ssize_t intel_vgpu_rw(struct mdev_device *mdev, char *buf,
 	case VFIO_PCI_BAR5_REGION_INDEX:
 	case VFIO_PCI_VGA_REGION_INDEX:
 	case VFIO_PCI_ROM_REGION_INDEX:
+		break;
 	default:
-		gvt_vgpu_err("unsupported region: %u\n", index);
+		if (index >= VFIO_PCI_NUM_REGIONS + vgpu->vdev.num_regions)
+			return -EINVAL;
+
+		index -= VFIO_PCI_NUM_REGIONS;
+		return vgpu->vdev.region[index].ops->rw(vgpu, buf, count,
+				ppos, is_write);
 	}
 
 	return ret == 0 ? count : ret;
@@ -952,7 +958,8 @@ static long intel_vgpu_ioctl(struct mdev_device *mdev, unsigned int cmd,
 
 		info.flags = VFIO_DEVICE_FLAGS_PCI;
 		info.flags |= VFIO_DEVICE_FLAGS_RESET;
-		info.num_regions = VFIO_PCI_NUM_REGIONS;
+		info.num_regions = VFIO_PCI_NUM_REGIONS +
+				vgpu->vdev.num_regions;
 		info.num_irqs = VFIO_PCI_NUM_IRQS;
 
 		return copy_to_user((void __user *)arg, &info, minsz) ?
@@ -1024,13 +1031,17 @@ static long intel_vgpu_ioctl(struct mdev_device *mdev, unsigned int cmd,
 		case VFIO_PCI_BAR3_REGION_INDEX ... VFIO_PCI_BAR5_REGION_INDEX:
 			info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index);
 			info.size = 0;
-
 			info.flags = 0;
+
 			gvt_dbg_core("get region info bar:%d\n", info.index);
 			break;
 
 		case VFIO_PCI_ROM_REGION_INDEX:
 		case VFIO_PCI_VGA_REGION_INDEX:
+			info.offset = VFIO_PCI_INDEX_TO_OFFSET(info.index);
+			info.size = 0;
+			info.flags = 0;
+
 			gvt_dbg_core("get region info index:%d\n", info.index);
 			break;
 		default:
@@ -1078,6 +1089,7 @@ static long intel_vgpu_ioctl(struct mdev_device *mdev, unsigned int cmd,
 		}
 
 		if (caps.size) {
+			info.flags |= VFIO_REGION_INFO_FLAG_CAPS;
 			if (info.argsz < sizeof(info) + caps.size) {
 				info.argsz = sizeof(info) + caps.size;
 				info.cap_offset = 0;
@@ -1164,6 +1176,33 @@ static long intel_vgpu_ioctl(struct mdev_device *mdev, unsigned int cmd,
 	} else if (cmd == VFIO_DEVICE_RESET) {
 		intel_gvt_ops->vgpu_reset(vgpu);
 		return 0;
+	} else if (cmd == VFIO_DEVICE_QUERY_GFX_PLANE) {
+		struct vfio_device_gfx_plane_info dmabuf;
+		int ret = 0;
+
+		minsz = offsetofend(struct vfio_device_gfx_plane_info,
+				    dmabuf_id);
+		if (copy_from_user(&dmabuf, (void __user *)arg, minsz))
+			return -EFAULT;
+		if (dmabuf.argsz < minsz)
+			return -EINVAL;
+
+		ret = intel_gvt_ops->vgpu_query_plane(vgpu, &dmabuf);
+		if (ret != 0)
+			return ret;
+
+		return copy_to_user((void __user *)arg, &dmabuf, minsz) ?
+								-EFAULT : 0;
+	} else if (cmd == VFIO_DEVICE_GET_GFX_DMABUF) {
+		__u32 dmabuf_id;
+		__s32 dmabuf_fd;
+
+		if (get_user(dmabuf_id, (__u32 __user *)arg))
+			return -EFAULT;
+
+		dmabuf_fd = intel_gvt_ops->vgpu_get_dmabuf(vgpu, dmabuf_id);
+		return dmabuf_fd;
+
 	}
 
 	return 0;
@@ -1193,7 +1232,7 @@ hw_id_show(struct device *dev, struct device_attribute *attr,
 		struct intel_vgpu *vgpu = (struct intel_vgpu *)
 			mdev_get_drvdata(mdev);
 		return sprintf(buf, "%u\n",
-			       vgpu->shadow_ctx->hw_id);
+			       vgpu->submission.shadow_ctx->hw_id);
 	}
 	return sprintf(buf, "\n");
 }
@@ -1217,8 +1256,7 @@ static const struct attribute_group *intel_vgpu_groups[] = {
 	NULL,
 };
 
-static const struct mdev_parent_ops intel_vgpu_ops = {
-	.supported_type_groups	= intel_vgpu_type_groups,
+static struct mdev_parent_ops intel_vgpu_ops = {
 	.mdev_attr_groups       = intel_vgpu_groups,
 	.create			= intel_vgpu_create,
 	.remove			= intel_vgpu_remove,
@@ -1234,17 +1272,20 @@ static const struct mdev_parent_ops intel_vgpu_ops = {
 
 static int kvmgt_host_init(struct device *dev, void *gvt, const void *ops)
 {
-	if (!intel_gvt_init_vgpu_type_groups(gvt))
-		return -EFAULT;
+	struct attribute **kvm_type_attrs;
+	struct attribute_group **kvm_vgpu_type_groups;
 
 	intel_gvt_ops = ops;
+	if (!intel_gvt_ops->get_gvt_attrs(&kvm_type_attrs,
+			&kvm_vgpu_type_groups))
+		return -EFAULT;
+	intel_vgpu_ops.supported_type_groups = kvm_vgpu_type_groups;
 
 	return mdev_register_device(dev, &intel_vgpu_ops);
 }
 
 static void kvmgt_host_exit(struct device *dev, void *gvt)
 {
-	intel_gvt_cleanup_vgpu_type_groups(gvt);
 	mdev_unregister_device(dev);
 }
 
@@ -1324,8 +1365,8 @@ static void kvmgt_page_track_write(struct kvm_vcpu *vcpu, gpa_t gpa,
 					struct kvmgt_guest_info, track_node);
 
 	if (kvmgt_gfn_is_write_protected(info, gpa_to_gfn(gpa)))
-		intel_gvt_ops->emulate_mmio_write(info->vgpu, gpa,
-					(void *)val, len);
+		intel_gvt_ops->write_protect_handler(info->vgpu, gpa,
+						     (void *)val, len);
 }
 
 static void kvmgt_page_track_flush_slot(struct kvm *kvm,
@@ -1403,6 +1444,9 @@ static int kvmgt_guest_init(struct mdev_device *mdev)
 	kvmgt_protect_table_init(info);
 	gvt_cache_init(vgpu);
 
+	mutex_init(&vgpu->dmabuf_lock);
+	init_completion(&vgpu->vblank_done);
+
 	info->track_node.track_write = kvmgt_page_track_write;
 	info->track_node.track_flush_slot = kvmgt_page_track_flush_slot;
 	kvm_page_track_register_notifier(kvm, &info->track_node);
@@ -1543,6 +1587,9 @@ struct intel_gvt_mpt kvmgt_mpt = {
 	.read_gpa = kvmgt_read_gpa,
 	.write_gpa = kvmgt_write_gpa,
 	.gfn_to_mfn = kvmgt_gfn_to_pfn,
+	.set_opregion = kvmgt_set_opregion,
+	.get_vfio_device = kvmgt_get_vfio_device,
+	.put_vfio_device = kvmgt_put_vfio_device,
 };
 EXPORT_SYMBOL_GPL(kvmgt_mpt);
 
diff --git a/drivers/gpu/drm/i915/gvt/mmio.c b/drivers/gpu/drm/i915/gvt/mmio.c
index 1e1310f50289..562b5ad857a4 100644
--- a/drivers/gpu/drm/i915/gvt/mmio.c
+++ b/drivers/gpu/drm/i915/gvt/mmio.c
@@ -117,25 +117,6 @@ static void failsafe_emulate_mmio_rw(struct intel_vgpu *vgpu, uint64_t pa,
 		else
 			memcpy(pt, p_data, bytes);
 
-	} else if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
-		struct intel_vgpu_guest_page *gp;
-
-		/* Since we enter the failsafe mode early during guest boot,
-		 * guest may not have chance to set up its ppgtt table, so
-		 * there should not be any wp pages for guest. Keep the wp
-		 * related code here in case we need to handle it in furture.
-		 */
-		gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT);
-		if (gp) {
-			/* remove write protection to prevent furture traps */
-			intel_vgpu_clean_guest_page(vgpu, gp);
-			if (read)
-				intel_gvt_hypervisor_read_gpa(vgpu, pa,
-						p_data, bytes);
-			else
-				intel_gvt_hypervisor_write_gpa(vgpu, pa,
-						p_data, bytes);
-		}
 	}
 	mutex_unlock(&gvt->lock);
 }
@@ -157,7 +138,6 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa,
 	unsigned int offset = 0;
 	int ret = -EINVAL;
 
-
 	if (vgpu->failsafe) {
 		failsafe_emulate_mmio_rw(vgpu, pa, p_data, bytes, true);
 		return 0;
@@ -166,26 +146,7 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa,
 
 	if (vgpu_gpa_is_aperture(vgpu, pa)) {
 		ret = vgpu_aperture_rw(vgpu, pa, p_data, bytes, true);
-		mutex_unlock(&gvt->lock);
-		return ret;
-	}
-
-	if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
-		struct intel_vgpu_guest_page *gp;
-
-		gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT);
-		if (gp) {
-			ret = intel_gvt_hypervisor_read_gpa(vgpu, pa,
-					p_data, bytes);
-			if (ret) {
-				gvt_vgpu_err("guest page read error %d, "
-					"gfn 0x%lx, pa 0x%llx, var 0x%x, len %d\n",
-					ret, gp->gfn, pa, *(u32 *)p_data,
-					bytes);
-			}
-			mutex_unlock(&gvt->lock);
-			return ret;
-		}
+		goto out;
 	}
 
 	offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
@@ -205,14 +166,12 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa,
 				p_data, bytes);
 		if (ret)
 			goto err;
-		mutex_unlock(&gvt->lock);
-		return ret;
+		goto out;
 	}
 
 	if (WARN_ON_ONCE(!reg_is_mmio(gvt, offset))) {
 		ret = intel_gvt_hypervisor_read_gpa(vgpu, pa, p_data, bytes);
-		mutex_unlock(&gvt->lock);
-		return ret;
+		goto out;
 	}
 
 	if (WARN_ON(!reg_is_mmio(gvt, offset + bytes - 1)))
@@ -228,11 +187,13 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa,
 		goto err;
 
 	intel_gvt_mmio_set_accessed(gvt, offset);
-	mutex_unlock(&gvt->lock);
-	return 0;
+	ret = 0;
+	goto out;
+
 err:
 	gvt_vgpu_err("fail to emulate MMIO read %08x len %d\n",
 			offset, bytes);
+out:
 	mutex_unlock(&gvt->lock);
 	return ret;
 }
@@ -263,26 +224,7 @@ int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, uint64_t pa,
 
 	if (vgpu_gpa_is_aperture(vgpu, pa)) {
 		ret = vgpu_aperture_rw(vgpu, pa, p_data, bytes, false);
-		mutex_unlock(&gvt->lock);
-		return ret;
-	}
-
-	if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) {
-		struct intel_vgpu_guest_page *gp;
-
-		gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT);
-		if (gp) {
-			ret = gp->handler(gp, pa, p_data, bytes);
-			if (ret) {
-				gvt_err("guest page write error %d, "
-					"gfn 0x%lx, pa 0x%llx, "
-					"var 0x%x, len %d\n",
-					ret, gp->gfn, pa,
-					*(u32 *)p_data, bytes);
-			}
-			mutex_unlock(&gvt->lock);
-			return ret;
-		}
+		goto out;
 	}
 
 	offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
@@ -302,14 +244,12 @@ int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, uint64_t pa,
 				p_data, bytes);
 		if (ret)
 			goto err;
-		mutex_unlock(&gvt->lock);
-		return ret;
+		goto out;
 	}
 
 	if (WARN_ON_ONCE(!reg_is_mmio(gvt, offset))) {
 		ret = intel_gvt_hypervisor_write_gpa(vgpu, pa, p_data, bytes);
-		mutex_unlock(&gvt->lock);
-		return ret;
+		goto out;
 	}
 
 	ret = intel_vgpu_mmio_reg_rw(vgpu, offset, p_data, bytes, false);
@@ -317,11 +257,12 @@ int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, uint64_t pa,
 		goto err;
 
 	intel_gvt_mmio_set_accessed(gvt, offset);
-	mutex_unlock(&gvt->lock);
-	return 0;
+	ret = 0;
+	goto out;
 err:
 	gvt_vgpu_err("fail to emulate MMIO write %08x len %d\n", offset,
 		     bytes);
+out:
 	mutex_unlock(&gvt->lock);
 	return ret;
 }
@@ -342,10 +283,10 @@ void intel_vgpu_reset_mmio(struct intel_vgpu *vgpu, bool dmlr)
 		memcpy(vgpu->mmio.vreg, mmio, info->mmio_size);
 		memcpy(vgpu->mmio.sreg, mmio, info->mmio_size);
 
-		vgpu_vreg(vgpu, GEN6_GT_THREAD_STATUS_REG) = 0;
+		vgpu_vreg_t(vgpu, GEN6_GT_THREAD_STATUS_REG) = 0;
 
 		/* set the bit 0:2(Core C-State ) to C0 */
-		vgpu_vreg(vgpu, GEN6_GT_CORE_STATUS) = 0;
+		vgpu_vreg_t(vgpu, GEN6_GT_CORE_STATUS) = 0;
 
 		vgpu->mmio.disable_warn_untrack = false;
 	} else {
diff --git a/drivers/gpu/drm/i915/gvt/mmio.h b/drivers/gpu/drm/i915/gvt/mmio.h
index dbc04ad2c7a1..71b620875943 100644
--- a/drivers/gpu/drm/i915/gvt/mmio.h
+++ b/drivers/gpu/drm/i915/gvt/mmio.h
@@ -72,12 +72,9 @@ bool intel_gvt_match_device(struct intel_gvt *gvt, unsigned long device);
 
 int intel_gvt_setup_mmio_info(struct intel_gvt *gvt);
 void intel_gvt_clean_mmio_info(struct intel_gvt *gvt);
-
-#define INTEL_GVT_MMIO_OFFSET(reg) ({ \
-	typeof(reg) __reg = reg; \
-	u32 *offset = (u32 *)&__reg; \
-	*offset; \
-})
+int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt,
+	int (*handler)(struct intel_gvt *gvt, u32 offset, void *data),
+	void *data);
 
 int intel_vgpu_init_mmio(struct intel_vgpu *vgpu);
 void intel_vgpu_reset_mmio(struct intel_vgpu *vgpu, bool dmlr);
diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c
new file mode 100644
index 000000000000..74834395dd89
--- /dev/null
+++ b/drivers/gpu/drm/i915/gvt/mmio_context.c
@@ -0,0 +1,412 @@
+/*
+ * Copyright(c) 2011-2016 Intel Corporation. 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, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Authors:
+ *    Eddie Dong <eddie.dong@intel.com>
+ *    Kevin Tian <kevin.tian@intel.com>
+ *
+ * Contributors:
+ *    Zhi Wang <zhi.a.wang@intel.com>
+ *    Changbin Du <changbin.du@intel.com>
+ *    Zhenyu Wang <zhenyuw@linux.intel.com>
+ *    Tina Zhang <tina.zhang@intel.com>
+ *    Bing Niu <bing.niu@intel.com>
+ *
+ */
+
+#include "i915_drv.h"
+#include "gvt.h"
+#include "trace.h"
+
+/**
+ * Defined in Intel Open Source PRM.
+ * Ref: https://01.org/linuxgraphics/documentation/hardware-specification-prms
+ */
+#define TRVATTL3PTRDW(i)	_MMIO(0x4de0 + (i)*4)
+#define TRNULLDETCT		_MMIO(0x4de8)
+#define TRINVTILEDETCT		_MMIO(0x4dec)
+#define TRVADR			_MMIO(0x4df0)
+#define TRTTE			_MMIO(0x4df4)
+#define RING_EXCC(base)		_MMIO((base) + 0x28)
+#define RING_GFX_MODE(base)	_MMIO((base) + 0x29c)
+#define VF_GUARDBAND		_MMIO(0x83a4)
+
+/* Raw offset is appened to each line for convenience. */
+static struct engine_mmio gen8_engine_mmio_list[] __cacheline_aligned = {
+	{RCS, GFX_MODE_GEN7, 0xffff, false}, /* 0x229c */
+	{RCS, GEN9_CTX_PREEMPT_REG, 0x0, false}, /* 0x2248 */
+	{RCS, HWSTAM, 0x0, false}, /* 0x2098 */
+	{RCS, INSTPM, 0xffff, true}, /* 0x20c0 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 0), 0, false}, /* 0x24d0 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 1), 0, false}, /* 0x24d4 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 2), 0, false}, /* 0x24d8 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 3), 0, false}, /* 0x24dc */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 4), 0, false}, /* 0x24e0 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 5), 0, false}, /* 0x24e4 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 6), 0, false}, /* 0x24e8 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 7), 0, false}, /* 0x24ec */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 8), 0, false}, /* 0x24f0 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 9), 0, false}, /* 0x24f4 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 10), 0, false}, /* 0x24f8 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 11), 0, false}, /* 0x24fc */
+	{RCS, CACHE_MODE_1, 0xffff, true}, /* 0x7004 */
+	{RCS, GEN7_GT_MODE, 0xffff, true}, /* 0x7008 */
+	{RCS, CACHE_MODE_0_GEN7, 0xffff, true}, /* 0x7000 */
+	{RCS, GEN7_COMMON_SLICE_CHICKEN1, 0xffff, true}, /* 0x7010 */
+	{RCS, HDC_CHICKEN0, 0xffff, true}, /* 0x7300 */
+	{RCS, VF_GUARDBAND, 0xffff, true}, /* 0x83a4 */
+
+	{BCS, RING_GFX_MODE(BLT_RING_BASE), 0xffff, false}, /* 0x2229c */
+	{BCS, RING_MI_MODE(BLT_RING_BASE), 0xffff, false}, /* 0x2209c */
+	{BCS, RING_INSTPM(BLT_RING_BASE), 0xffff, false}, /* 0x220c0 */
+	{BCS, RING_HWSTAM(BLT_RING_BASE), 0x0, false}, /* 0x22098 */
+	{BCS, RING_EXCC(BLT_RING_BASE), 0x0, false}, /* 0x22028 */
+	{ /* Terminated */ }
+};
+
+static struct engine_mmio gen9_engine_mmio_list[] __cacheline_aligned = {
+	{RCS, GFX_MODE_GEN7, 0xffff, false}, /* 0x229c */
+	{RCS, GEN9_CTX_PREEMPT_REG, 0x0, false}, /* 0x2248 */
+	{RCS, HWSTAM, 0x0, false}, /* 0x2098 */
+	{RCS, INSTPM, 0xffff, true}, /* 0x20c0 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 0), 0, false}, /* 0x24d0 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 1), 0, false}, /* 0x24d4 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 2), 0, false}, /* 0x24d8 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 3), 0, false}, /* 0x24dc */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 4), 0, false}, /* 0x24e0 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 5), 0, false}, /* 0x24e4 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 6), 0, false}, /* 0x24e8 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 7), 0, false}, /* 0x24ec */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 8), 0, false}, /* 0x24f0 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 9), 0, false}, /* 0x24f4 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 10), 0, false}, /* 0x24f8 */
+	{RCS, RING_FORCE_TO_NONPRIV(RENDER_RING_BASE, 11), 0, false}, /* 0x24fc */
+	{RCS, CACHE_MODE_1, 0xffff, true}, /* 0x7004 */
+	{RCS, GEN7_GT_MODE, 0xffff, true}, /* 0x7008 */
+	{RCS, CACHE_MODE_0_GEN7, 0xffff, true}, /* 0x7000 */
+	{RCS, GEN7_COMMON_SLICE_CHICKEN1, 0xffff, true}, /* 0x7010 */
+	{RCS, HDC_CHICKEN0, 0xffff, true}, /* 0x7300 */
+	{RCS, VF_GUARDBAND, 0xffff, true}, /* 0x83a4 */
+
+	{RCS, GEN8_PRIVATE_PAT_LO, 0, false}, /* 0x40e0 */
+	{RCS, GEN8_PRIVATE_PAT_HI, 0, false}, /* 0x40e4 */
+	{RCS, GEN8_CS_CHICKEN1, 0xffff, true}, /* 0x2580 */
+	{RCS, COMMON_SLICE_CHICKEN2, 0xffff, true}, /* 0x7014 */
+	{RCS, GEN9_CS_DEBUG_MODE1, 0xffff, false}, /* 0x20ec */
+	{RCS, GEN8_L3SQCREG4, 0, false}, /* 0xb118 */
+	{RCS, GEN7_HALF_SLICE_CHICKEN1, 0xffff, true}, /* 0xe100 */
+	{RCS, HALF_SLICE_CHICKEN2, 0xffff, true}, /* 0xe180 */
+	{RCS, HALF_SLICE_CHICKEN3, 0xffff, true}, /* 0xe184 */
+	{RCS, GEN9_HALF_SLICE_CHICKEN5, 0xffff, true}, /* 0xe188 */
+	{RCS, GEN9_HALF_SLICE_CHICKEN7, 0xffff, true}, /* 0xe194 */
+	{RCS, TRVATTL3PTRDW(0), 0, false}, /* 0x4de0 */
+	{RCS, TRVATTL3PTRDW(1), 0, false}, /* 0x4de4 */
+	{RCS, TRNULLDETCT, 0, false}, /* 0x4de8 */
+	{RCS, TRINVTILEDETCT, 0, false}, /* 0x4dec */
+	{RCS, TRVADR, 0, false}, /* 0x4df0 */
+	{RCS, TRTTE, 0, false}, /* 0x4df4 */
+
+	{BCS, RING_GFX_MODE(BLT_RING_BASE), 0xffff, false}, /* 0x2229c */
+	{BCS, RING_MI_MODE(BLT_RING_BASE), 0xffff, false}, /* 0x2209c */
+	{BCS, RING_INSTPM(BLT_RING_BASE), 0xffff, false}, /* 0x220c0 */
+	{BCS, RING_HWSTAM(BLT_RING_BASE), 0x0, false}, /* 0x22098 */
+	{BCS, RING_EXCC(BLT_RING_BASE), 0x0, false}, /* 0x22028 */
+
+	{VCS2, RING_EXCC(GEN8_BSD2_RING_BASE), 0xffff, false}, /* 0x1c028 */
+
+	{VECS, RING_EXCC(VEBOX_RING_BASE), 0xffff, false}, /* 0x1a028 */
+
+	{RCS, GEN8_HDC_CHICKEN1, 0xffff, true}, /* 0x7304 */
+	{RCS, GEN9_CTX_PREEMPT_REG, 0x0, false}, /* 0x2248 */
+	{RCS, GEN7_UCGCTL4, 0x0, false}, /* 0x940c */
+	{RCS, GAMT_CHKN_BIT_REG, 0x0, false}, /* 0x4ab8 */
+
+	{RCS, GEN9_GAMT_ECO_REG_RW_IA, 0x0, false}, /* 0x4ab0 */
+	{RCS, GEN9_CSFE_CHICKEN1_RCS, 0x0, false}, /* 0x20d4 */
+
+	{RCS, GEN8_GARBCNTL, 0x0, false}, /* 0xb004 */
+	{RCS, GEN7_FF_THREAD_MODE, 0x0, false}, /* 0x20a0 */
+	{RCS, FF_SLICE_CS_CHICKEN2, 0xffff, false}, /* 0x20e4 */
+	{ /* Terminated */ }
+};
+
+static struct {
+	bool initialized;
+	u32 control_table[I915_NUM_ENGINES][64];
+	u32 l3cc_table[32];
+} gen9_render_mocs;
+
+static void load_render_mocs(struct drm_i915_private *dev_priv)
+{
+	i915_reg_t offset;
+	u32 regs[] = {
+		[RCS] = 0xc800,
+		[VCS] = 0xc900,
+		[VCS2] = 0xca00,
+		[BCS] = 0xcc00,
+		[VECS] = 0xcb00,
+	};
+	int ring_id, i;
+
+	for (ring_id = 0; ring_id < I915_NUM_ENGINES; ring_id++) {
+		offset.reg = regs[ring_id];
+		for (i = 0; i < 64; i++) {
+			gen9_render_mocs.control_table[ring_id][i] =
+				I915_READ_FW(offset);
+			offset.reg += 4;
+		}
+	}
+
+	offset.reg = 0xb020;
+	for (i = 0; i < 32; i++) {
+		gen9_render_mocs.l3cc_table[i] =
+			I915_READ_FW(offset);
+		offset.reg += 4;
+	}
+	gen9_render_mocs.initialized = true;
+}
+
+static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	enum forcewake_domains fw;
+	i915_reg_t reg;
+	u32 regs[] = {
+		[RCS] = 0x4260,
+		[VCS] = 0x4264,
+		[VCS2] = 0x4268,
+		[BCS] = 0x426c,
+		[VECS] = 0x4270,
+	};
+
+	if (WARN_ON(ring_id >= ARRAY_SIZE(regs)))
+		return;
+
+	if (!test_and_clear_bit(ring_id, (void *)s->tlb_handle_pending))
+		return;
+
+	reg = _MMIO(regs[ring_id]);
+
+	/* WaForceWakeRenderDuringMmioTLBInvalidate:skl
+	 * we need to put a forcewake when invalidating RCS TLB caches,
+	 * otherwise device can go to RC6 state and interrupt invalidation
+	 * process
+	 */
+	fw = intel_uncore_forcewake_for_reg(dev_priv, reg,
+					    FW_REG_READ | FW_REG_WRITE);
+	if (ring_id == RCS && (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)))
+		fw |= FORCEWAKE_RENDER;
+
+	intel_uncore_forcewake_get(dev_priv, fw);
+
+	I915_WRITE_FW(reg, 0x1);
+
+	if (wait_for_atomic((I915_READ_FW(reg) == 0), 50))
+		gvt_vgpu_err("timeout in invalidate ring (%d) tlb\n", ring_id);
+	else
+		vgpu_vreg_t(vgpu, reg) = 0;
+
+	intel_uncore_forcewake_put(dev_priv, fw);
+
+	gvt_dbg_core("invalidate TLB for ring %d\n", ring_id);
+}
+
+static void switch_mocs(struct intel_vgpu *pre, struct intel_vgpu *next,
+			int ring_id)
+{
+	struct drm_i915_private *dev_priv;
+	i915_reg_t offset, l3_offset;
+	u32 old_v, new_v;
+
+	u32 regs[] = {
+		[RCS] = 0xc800,
+		[VCS] = 0xc900,
+		[VCS2] = 0xca00,
+		[BCS] = 0xcc00,
+		[VECS] = 0xcb00,
+	};
+	int i;
+
+	dev_priv = pre ? pre->gvt->dev_priv : next->gvt->dev_priv;
+	if (WARN_ON(ring_id >= ARRAY_SIZE(regs)))
+		return;
+
+	if (!pre && !gen9_render_mocs.initialized)
+		load_render_mocs(dev_priv);
+
+	offset.reg = regs[ring_id];
+	for (i = 0; i < 64; i++) {
+		if (pre)
+			old_v = vgpu_vreg_t(pre, offset);
+		else
+			old_v = gen9_render_mocs.control_table[ring_id][i];
+		if (next)
+			new_v = vgpu_vreg_t(next, offset);
+		else
+			new_v = gen9_render_mocs.control_table[ring_id][i];
+
+		if (old_v != new_v)
+			I915_WRITE_FW(offset, new_v);
+
+		offset.reg += 4;
+	}
+
+	if (ring_id == RCS) {
+		l3_offset.reg = 0xb020;
+		for (i = 0; i < 32; i++) {
+			if (pre)
+				old_v = vgpu_vreg_t(pre, l3_offset);
+			else
+				old_v = gen9_render_mocs.l3cc_table[i];
+			if (next)
+				new_v = vgpu_vreg_t(next, l3_offset);
+			else
+				new_v = gen9_render_mocs.l3cc_table[i];
+
+			if (old_v != new_v)
+				I915_WRITE_FW(l3_offset, new_v);
+
+			l3_offset.reg += 4;
+		}
+	}
+}
+
+#define CTX_CONTEXT_CONTROL_VAL	0x03
+
+/* Switch ring mmio values (context). */
+static void switch_mmio(struct intel_vgpu *pre,
+			struct intel_vgpu *next,
+			int ring_id)
+{
+	struct drm_i915_private *dev_priv;
+	struct intel_vgpu_submission *s;
+	u32 *reg_state, ctx_ctrl;
+	u32 inhibit_mask =
+		_MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT);
+	struct engine_mmio *mmio;
+	u32 old_v, new_v;
+
+	dev_priv = pre ? pre->gvt->dev_priv : next->gvt->dev_priv;
+	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv))
+		switch_mocs(pre, next, ring_id);
+
+	mmio = dev_priv->gvt->engine_mmio_list;
+	while (i915_mmio_reg_offset((mmio++)->reg)) {
+		if (mmio->ring_id != ring_id)
+			continue;
+		// save
+		if (pre) {
+			vgpu_vreg_t(pre, mmio->reg) = I915_READ_FW(mmio->reg);
+			if (mmio->mask)
+				vgpu_vreg_t(pre, mmio->reg) &=
+						~(mmio->mask << 16);
+			old_v = vgpu_vreg_t(pre, mmio->reg);
+		} else
+			old_v = mmio->value = I915_READ_FW(mmio->reg);
+
+		// restore
+		if (next) {
+			s = &next->submission;
+			reg_state =
+				s->shadow_ctx->engine[ring_id].lrc_reg_state;
+			ctx_ctrl = reg_state[CTX_CONTEXT_CONTROL_VAL];
+			/*
+			 * if it is an inhibit context, load in_context mmio
+			 * into HW by mmio write. If it is not, skip this mmio
+			 * write.
+			 */
+			if (mmio->in_context &&
+			    (ctx_ctrl & inhibit_mask) != inhibit_mask)
+				continue;
+
+			if (mmio->mask)
+				new_v = vgpu_vreg_t(next, mmio->reg) |
+							(mmio->mask << 16);
+			else
+				new_v = vgpu_vreg_t(next, mmio->reg);
+		} else {
+			if (mmio->in_context)
+				continue;
+			if (mmio->mask)
+				new_v = mmio->value | (mmio->mask << 16);
+			else
+				new_v = mmio->value;
+		}
+
+		I915_WRITE_FW(mmio->reg, new_v);
+
+		trace_render_mmio(pre ? pre->id : 0,
+				  next ? next->id : 0,
+				  "switch",
+				  i915_mmio_reg_offset(mmio->reg),
+				  old_v, new_v);
+	}
+
+	if (next)
+		handle_tlb_pending_event(next, ring_id);
+}
+
+/**
+ * intel_gvt_switch_render_mmio - switch mmio context of specific engine
+ * @pre: the last vGPU that own the engine
+ * @next: the vGPU to switch to
+ * @ring_id: specify the engine
+ *
+ * If pre is null indicates that host own the engine. If next is null
+ * indicates that we are switching to host workload.
+ */
+void intel_gvt_switch_mmio(struct intel_vgpu *pre,
+			   struct intel_vgpu *next, int ring_id)
+{
+	struct drm_i915_private *dev_priv;
+
+	if (WARN_ON(!pre && !next))
+		return;
+
+	gvt_dbg_render("switch ring %d from %s to %s\n", ring_id,
+		       pre ? "vGPU" : "host", next ? "vGPU" : "HOST");
+
+	dev_priv = pre ? pre->gvt->dev_priv : next->gvt->dev_priv;
+
+	/**
+	 * We are using raw mmio access wrapper to improve the
+	 * performace for batch mmio read/write, so we need
+	 * handle forcewake mannually.
+	 */
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
+	switch_mmio(pre, next, ring_id);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
+}
+
+/**
+ * intel_gvt_init_engine_mmio_context - Initiate the engine mmio list
+ * @gvt: GVT device
+ *
+ */
+void intel_gvt_init_engine_mmio_context(struct intel_gvt *gvt)
+{
+	if (IS_SKYLAKE(gvt->dev_priv) || IS_KABYLAKE(gvt->dev_priv))
+		gvt->engine_mmio_list = gen9_engine_mmio_list;
+	else
+		gvt->engine_mmio_list = gen8_engine_mmio_list;
+}
diff --git a/drivers/gpu/drm/i915/gvt/render.h b/drivers/gpu/drm/i915/gvt/mmio_context.h
index 91db1d39d28f..ca2c6a745673 100644
--- a/drivers/gpu/drm/i915/gvt/render.h
+++ b/drivers/gpu/drm/i915/gvt/mmio_context.h
@@ -36,8 +36,17 @@
 #ifndef __GVT_RENDER_H__
 #define __GVT_RENDER_H__
 
+struct engine_mmio {
+	int ring_id;
+	i915_reg_t reg;
+	u32 mask;
+	bool in_context;
+	u32 value;
+};
+
 void intel_gvt_switch_mmio(struct intel_vgpu *pre,
 			   struct intel_vgpu *next, int ring_id);
 
+void intel_gvt_init_engine_mmio_context(struct intel_gvt *gvt);
 
 #endif
diff --git a/drivers/gpu/drm/i915/gvt/mpt.h b/drivers/gpu/drm/i915/gvt/mpt.h
index f0e5487e6688..ca8005a6d5fa 100644
--- a/drivers/gpu/drm/i915/gvt/mpt.h
+++ b/drivers/gpu/drm/i915/gvt/mpt.h
@@ -154,51 +154,53 @@ static inline unsigned long intel_gvt_hypervisor_virt_to_mfn(void *p)
 }
 
 /**
- * intel_gvt_hypervisor_set_wp_page - set a guest page to write-protected
+ * intel_gvt_hypervisor_enable - set a guest page to write-protected
  * @vgpu: a vGPU
- * @p: intel_vgpu_guest_page
+ * @t: page track data structure
  *
  * Returns:
  * Zero on success, negative error code if failed.
  */
-static inline int intel_gvt_hypervisor_set_wp_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *p)
+static inline int intel_gvt_hypervisor_enable_page_track(
+		struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t)
 {
 	int ret;
 
-	if (p->writeprotection)
+	if (t->tracked)
 		return 0;
 
-	ret = intel_gvt_host.mpt->set_wp_page(vgpu->handle, p->gfn);
+	ret = intel_gvt_host.mpt->set_wp_page(vgpu->handle, t->gfn);
 	if (ret)
 		return ret;
-	p->writeprotection = true;
-	atomic_inc(&vgpu->gtt.n_write_protected_guest_page);
+	t->tracked = true;
+	atomic_inc(&vgpu->gtt.n_tracked_guest_page);
 	return 0;
 }
 
 /**
- * intel_gvt_hypervisor_unset_wp_page - remove the write-protection of a
+ * intel_gvt_hypervisor_disable_page_track - remove the write-protection of a
  * guest page
  * @vgpu: a vGPU
- * @p: intel_vgpu_guest_page
+ * @t: page track data structure
  *
  * Returns:
  * Zero on success, negative error code if failed.
  */
-static inline int intel_gvt_hypervisor_unset_wp_page(struct intel_vgpu *vgpu,
-		struct intel_vgpu_guest_page *p)
+static inline int intel_gvt_hypervisor_disable_page_track(
+		struct intel_vgpu *vgpu,
+		struct intel_vgpu_page_track *t)
 {
 	int ret;
 
-	if (!p->writeprotection)
+	if (!t->tracked)
 		return 0;
 
-	ret = intel_gvt_host.mpt->unset_wp_page(vgpu->handle, p->gfn);
+	ret = intel_gvt_host.mpt->unset_wp_page(vgpu->handle, t->gfn);
 	if (ret)
 		return ret;
-	p->writeprotection = false;
-	atomic_dec(&vgpu->gtt.n_write_protected_guest_page);
+	t->tracked = false;
+	atomic_dec(&vgpu->gtt.n_tracked_guest_page);
 	return 0;
 }
 
@@ -292,4 +294,49 @@ static inline int intel_gvt_hypervisor_set_trap_area(
 	return intel_gvt_host.mpt->set_trap_area(vgpu->handle, start, end, map);
 }
 
+/**
+ * intel_gvt_hypervisor_set_opregion - Set opregion for guest
+ * @vgpu: a vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_hypervisor_set_opregion(struct intel_vgpu *vgpu)
+{
+	if (!intel_gvt_host.mpt->set_opregion)
+		return 0;
+
+	return intel_gvt_host.mpt->set_opregion(vgpu);
+}
+
+/**
+ * intel_gvt_hypervisor_get_vfio_device - increase vfio device ref count
+ * @vgpu: a vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline int intel_gvt_hypervisor_get_vfio_device(struct intel_vgpu *vgpu)
+{
+	if (!intel_gvt_host.mpt->get_vfio_device)
+		return 0;
+
+	return intel_gvt_host.mpt->get_vfio_device(vgpu);
+}
+
+/**
+ * intel_gvt_hypervisor_put_vfio_device - decrease vfio device ref count
+ * @vgpu: a vGPU
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+static inline void intel_gvt_hypervisor_put_vfio_device(struct intel_vgpu *vgpu)
+{
+	if (!intel_gvt_host.mpt->put_vfio_device)
+		return;
+
+	intel_gvt_host.mpt->put_vfio_device(vgpu);
+}
+
 #endif /* _GVT_MPT_H_ */
diff --git a/drivers/gpu/drm/i915/gvt/opregion.c b/drivers/gpu/drm/i915/gvt/opregion.c
index 311799136d7f..8420d1fc3ddb 100644
--- a/drivers/gpu/drm/i915/gvt/opregion.c
+++ b/drivers/gpu/drm/i915/gvt/opregion.c
@@ -25,36 +25,237 @@
 #include "i915_drv.h"
 #include "gvt.h"
 
-static int init_vgpu_opregion(struct intel_vgpu *vgpu, u32 gpa)
+/*
+ * Note: Only for GVT-g virtual VBT generation, other usage must
+ * not do like this.
+ */
+#define _INTEL_BIOS_PRIVATE
+#include "intel_vbt_defs.h"
+
+#define OPREGION_SIGNATURE "IntelGraphicsMem"
+#define MBOX_VBT      (1<<3)
+
+/* device handle */
+#define DEVICE_TYPE_CRT    0x01
+#define DEVICE_TYPE_EFP1   0x04
+#define DEVICE_TYPE_EFP2   0x40
+#define DEVICE_TYPE_EFP3   0x20
+#define DEVICE_TYPE_EFP4   0x10
+
+#define DEV_SIZE	38
+
+struct opregion_header {
+	u8 signature[16];
+	u32 size;
+	u32 opregion_ver;
+	u8 bios_ver[32];
+	u8 vbios_ver[16];
+	u8 driver_ver[16];
+	u32 mboxes;
+	u32 driver_model;
+	u32 pcon;
+	u8 dver[32];
+	u8 rsvd[124];
+} __packed;
+
+struct bdb_data_header {
+	u8 id;
+	u16 size; /* data size */
+} __packed;
+
+struct efp_child_device_config {
+	u16 handle;
+	u16 device_type;
+	u16 device_class;
+	u8 i2c_speed;
+	u8 dp_onboard_redriver; /* 158 */
+	u8 dp_ondock_redriver; /* 158 */
+	u8 hdmi_level_shifter_value:4; /* 169 */
+	u8 hdmi_max_data_rate:4; /* 204 */
+	u16 dtd_buf_ptr; /* 161 */
+	u8 edidless_efp:1; /* 161 */
+	u8 compression_enable:1; /* 198 */
+	u8 compression_method:1; /* 198 */
+	u8 ganged_edp:1; /* 202 */
+	u8 skip0:4;
+	u8 compression_structure_index:4; /* 198 */
+	u8 skip1:4;
+	u8 slave_port; /*  202 */
+	u8 skip2;
+	u8 dvo_port;
+	u8 i2c_pin; /* for add-in card */
+	u8 slave_addr; /* for add-in card */
+	u8 ddc_pin;
+	u16 edid_ptr;
+	u8 dvo_config;
+	u8 efp_docked_port:1; /* 158 */
+	u8 lane_reversal:1; /* 184 */
+	u8 onboard_lspcon:1; /* 192 */
+	u8 iboost_enable:1; /* 196 */
+	u8 hpd_invert:1; /* BXT 196 */
+	u8 slip3:3;
+	u8 hdmi_compat:1;
+	u8 dp_compat:1;
+	u8 tmds_compat:1;
+	u8 skip4:5;
+	u8 aux_channel;
+	u8 dongle_detect;
+	u8 pipe_cap:2;
+	u8 sdvo_stall:1; /* 158 */
+	u8 hpd_status:2;
+	u8 integrated_encoder:1;
+	u8 skip5:2;
+	u8 dvo_wiring;
+	u8 mipi_bridge_type; /* 171 */
+	u16 device_class_ext;
+	u8 dvo_function;
+	u8 dp_usb_type_c:1; /* 195 */
+	u8 skip6:7;
+	u8 dp_usb_type_c_2x_gpio_index; /* 195 */
+	u16 dp_usb_type_c_2x_gpio_pin; /* 195 */
+	u8 iboost_dp:4; /* 196 */
+	u8 iboost_hdmi:4; /* 196 */
+} __packed;
+
+struct vbt {
+	/* header->bdb_offset point to bdb_header offset */
+	struct vbt_header header;
+	struct bdb_header bdb_header;
+
+	struct bdb_data_header general_features_header;
+	struct bdb_general_features general_features;
+
+	struct bdb_data_header general_definitions_header;
+	struct bdb_general_definitions general_definitions;
+
+	struct efp_child_device_config child0;
+	struct efp_child_device_config child1;
+	struct efp_child_device_config child2;
+	struct efp_child_device_config child3;
+
+	struct bdb_data_header driver_features_header;
+	struct bdb_driver_features driver_features;
+};
+
+static void virt_vbt_generation(struct vbt *v)
 {
-	u8 *buf;
-	int i;
+	int num_child;
+
+	memset(v, 0, sizeof(struct vbt));
+
+	v->header.signature[0] = '$';
+	v->header.signature[1] = 'V';
+	v->header.signature[2] = 'B';
+	v->header.signature[3] = 'T';
+
+	/* there's features depending on version! */
+	v->header.version = 155;
+	v->header.header_size = sizeof(v->header);
+	v->header.vbt_size = sizeof(struct vbt) - sizeof(v->header);
+	v->header.bdb_offset = offsetof(struct vbt, bdb_header);
+
+	strcpy(&v->bdb_header.signature[0], "BIOS_DATA_BLOCK");
+	v->bdb_header.version = 186; /* child_dev_size = 38 */
+	v->bdb_header.header_size = sizeof(v->bdb_header);
+
+	v->bdb_header.bdb_size = sizeof(struct vbt) - sizeof(struct vbt_header)
+		- sizeof(struct bdb_header);
+
+	/* general features */
+	v->general_features_header.id = BDB_GENERAL_FEATURES;
+	v->general_features_header.size = sizeof(struct bdb_general_features);
+	v->general_features.int_crt_support = 0;
+	v->general_features.int_tv_support = 0;
+
+	/* child device */
+	num_child = 4; /* each port has one child */
+	v->general_definitions_header.id = BDB_GENERAL_DEFINITIONS;
+	/* size will include child devices */
+	v->general_definitions_header.size =
+		sizeof(struct bdb_general_definitions) + num_child * DEV_SIZE;
+	v->general_definitions.child_dev_size = DEV_SIZE;
+
+	/* portA */
+	v->child0.handle = DEVICE_TYPE_EFP1;
+	v->child0.device_type = DEVICE_TYPE_DP;
+	v->child0.dvo_port = DVO_PORT_DPA;
+	v->child0.aux_channel = DP_AUX_A;
+	v->child0.dp_compat = true;
+	v->child0.integrated_encoder = true;
+
+	/* portB */
+	v->child1.handle = DEVICE_TYPE_EFP2;
+	v->child1.device_type = DEVICE_TYPE_DP;
+	v->child1.dvo_port = DVO_PORT_DPB;
+	v->child1.aux_channel = DP_AUX_B;
+	v->child1.dp_compat = true;
+	v->child1.integrated_encoder = true;
+
+	/* portC */
+	v->child2.handle = DEVICE_TYPE_EFP3;
+	v->child2.device_type = DEVICE_TYPE_DP;
+	v->child2.dvo_port = DVO_PORT_DPC;
+	v->child2.aux_channel = DP_AUX_C;
+	v->child2.dp_compat = true;
+	v->child2.integrated_encoder = true;
+
+	/* portD */
+	v->child3.handle = DEVICE_TYPE_EFP4;
+	v->child3.device_type = DEVICE_TYPE_DP;
+	v->child3.dvo_port = DVO_PORT_DPD;
+	v->child3.aux_channel = DP_AUX_D;
+	v->child3.dp_compat = true;
+	v->child3.integrated_encoder = true;
+
+	/* driver features */
+	v->driver_features_header.id = BDB_DRIVER_FEATURES;
+	v->driver_features_header.size = sizeof(struct bdb_driver_features);
+	v->driver_features.lvds_config = BDB_DRIVER_FEATURE_NO_LVDS;
+}
 
-	if (WARN((vgpu_opregion(vgpu)->va),
-			"vgpu%d: opregion has been initialized already.\n",
-			vgpu->id))
-		return -EINVAL;
+/**
+ * intel_vgpu_init_opregion - initialize the stuff used to emulate opregion
+ * @vgpu: a vGPU
+ * @gpa: guest physical address of opregion
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_init_opregion(struct intel_vgpu *vgpu)
+{
+	u8 *buf;
+	struct opregion_header *header;
+	struct vbt v;
+	const char opregion_signature[16] = OPREGION_SIGNATURE;
 
+	gvt_dbg_core("init vgpu%d opregion\n", vgpu->id);
 	vgpu_opregion(vgpu)->va = (void *)__get_free_pages(GFP_KERNEL |
 			__GFP_ZERO,
 			get_order(INTEL_GVT_OPREGION_SIZE));
-
-	if (!vgpu_opregion(vgpu)->va)
+	if (!vgpu_opregion(vgpu)->va) {
+		gvt_err("fail to get memory for vgpu virt opregion\n");
 		return -ENOMEM;
+	}
 
-	memcpy(vgpu_opregion(vgpu)->va, vgpu->gvt->opregion.opregion_va,
-	       INTEL_GVT_OPREGION_SIZE);
-
-	for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++)
-		vgpu_opregion(vgpu)->gfn[i] = (gpa >> PAGE_SHIFT) + i;
+	/* emulated opregion with VBT mailbox only */
+	buf = (u8 *)vgpu_opregion(vgpu)->va;
+	header = (struct opregion_header *)buf;
+	memcpy(header->signature, opregion_signature,
+	       sizeof(opregion_signature));
+	header->size = 0x8;
+	header->opregion_ver = 0x02000000;
+	header->mboxes = MBOX_VBT;
 
 	/* for unknown reason, the value in LID field is incorrect
 	 * which block the windows guest, so workaround it by force
 	 * setting it to "OPEN"
 	 */
-	buf = (u8 *)vgpu_opregion(vgpu)->va;
 	buf[INTEL_GVT_OPREGION_CLID] = 0x3;
 
+	/* emulated vbt from virt vbt generation */
+	virt_vbt_generation(&v);
+	memcpy(buf + INTEL_GVT_OPREGION_VBT_OFFSET, &v, sizeof(struct vbt));
+
 	return 0;
 }
 
@@ -79,93 +280,91 @@ static int map_vgpu_opregion(struct intel_vgpu *vgpu, bool map)
 			return ret;
 		}
 	}
-	return 0;
-}
 
-/**
- * intel_vgpu_clean_opregion - clean the stuff used to emulate opregion
- * @vgpu: a vGPU
- *
- */
-void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu)
-{
-	gvt_dbg_core("vgpu%d: clean vgpu opregion\n", vgpu->id);
-
-	if (!vgpu_opregion(vgpu)->va)
-		return;
-
-	if (intel_gvt_host.hypervisor_type == INTEL_GVT_HYPERVISOR_XEN) {
-		map_vgpu_opregion(vgpu, false);
-		free_pages((unsigned long)vgpu_opregion(vgpu)->va,
-				get_order(INTEL_GVT_OPREGION_SIZE));
+	vgpu_opregion(vgpu)->mapped = map;
 
-		vgpu_opregion(vgpu)->va = NULL;
-	}
+	return 0;
 }
 
 /**
- * intel_vgpu_init_opregion - initialize the stuff used to emulate opregion
+ * intel_vgpu_opregion_base_write_handler - Opregion base register write handler
+ *
  * @vgpu: a vGPU
  * @gpa: guest physical address of opregion
  *
  * Returns:
  * Zero on success, negative error code if failed.
  */
-int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa)
+int intel_vgpu_opregion_base_write_handler(struct intel_vgpu *vgpu, u32 gpa)
 {
-	int ret;
 
-	gvt_dbg_core("vgpu%d: init vgpu opregion\n", vgpu->id);
+	int i, ret = 0;
+	unsigned long pfn;
 
-	if (intel_gvt_host.hypervisor_type == INTEL_GVT_HYPERVISOR_XEN) {
-		gvt_dbg_core("emulate opregion from kernel\n");
+	gvt_dbg_core("emulate opregion from kernel\n");
 
-		ret = init_vgpu_opregion(vgpu, gpa);
-		if (ret)
-			return ret;
+	switch (intel_gvt_host.hypervisor_type) {
+	case INTEL_GVT_HYPERVISOR_KVM:
+		pfn = intel_gvt_hypervisor_gfn_to_mfn(vgpu, gpa >> PAGE_SHIFT);
+		vgpu_opregion(vgpu)->va_gopregion = memremap(pfn << PAGE_SHIFT,
+						INTEL_GVT_OPREGION_SIZE,
+						MEMREMAP_WB);
+		if (!vgpu_opregion(vgpu)->va_gopregion) {
+			gvt_vgpu_err("failed to map guest opregion\n");
+			ret = -EFAULT;
+		}
+		vgpu_opregion(vgpu)->mapped = true;
+		break;
+	case INTEL_GVT_HYPERVISOR_XEN:
+		/**
+		 * Wins guest on Xengt will write this register twice: xen
+		 * hvmloader and windows graphic driver.
+		 */
+		if (vgpu_opregion(vgpu)->mapped)
+			map_vgpu_opregion(vgpu, false);
+
+		for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++)
+			vgpu_opregion(vgpu)->gfn[i] = (gpa >> PAGE_SHIFT) + i;
 
 		ret = map_vgpu_opregion(vgpu, true);
-		if (ret)
-			return ret;
+		break;
+	default:
+		ret = -EINVAL;
+		gvt_vgpu_err("not supported hypervisor\n");
 	}
 
-	return 0;
-}
-
-/**
- * intel_gvt_clean_opregion - clean host opergion related stuffs
- * @gvt: a GVT device
- *
- */
-void intel_gvt_clean_opregion(struct intel_gvt *gvt)
-{
-	memunmap(gvt->opregion.opregion_va);
-	gvt->opregion.opregion_va = NULL;
+	return ret;
 }
 
 /**
- * intel_gvt_init_opregion - initialize host opergion related stuffs
- * @gvt: a GVT device
+ * intel_vgpu_clean_opregion - clean the stuff used to emulate opregion
+ * @vgpu: a vGPU
  *
- * Returns:
- * Zero on success, negative error code if failed.
  */
-int intel_gvt_init_opregion(struct intel_gvt *gvt)
+void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu)
 {
-	gvt_dbg_core("init host opregion\n");
+	gvt_dbg_core("vgpu%d: clean vgpu opregion\n", vgpu->id);
 
-	pci_read_config_dword(gvt->dev_priv->drm.pdev, INTEL_GVT_PCI_OPREGION,
-			&gvt->opregion.opregion_pa);
+	if (!vgpu_opregion(vgpu)->va)
+		return;
 
-	gvt->opregion.opregion_va = memremap(gvt->opregion.opregion_pa,
-					     INTEL_GVT_OPREGION_SIZE, MEMREMAP_WB);
-	if (!gvt->opregion.opregion_va) {
-		gvt_err("fail to map host opregion\n");
-		return -EFAULT;
+	if (intel_gvt_host.hypervisor_type == INTEL_GVT_HYPERVISOR_XEN) {
+		if (vgpu_opregion(vgpu)->mapped)
+			map_vgpu_opregion(vgpu, false);
+	} else if (intel_gvt_host.hypervisor_type == INTEL_GVT_HYPERVISOR_KVM) {
+		if (vgpu_opregion(vgpu)->mapped) {
+			memunmap(vgpu_opregion(vgpu)->va_gopregion);
+			vgpu_opregion(vgpu)->va_gopregion = NULL;
+		}
 	}
-	return 0;
+	free_pages((unsigned long)vgpu_opregion(vgpu)->va,
+		   get_order(INTEL_GVT_OPREGION_SIZE));
+
+	vgpu_opregion(vgpu)->va = NULL;
+
 }
 
+
 #define GVT_OPREGION_FUNC(scic)					\
 	({							\
 	 u32 __ret;						\
@@ -284,8 +483,21 @@ int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci)
 	u32 *scic, *parm;
 	u32 func, subfunc;
 
-	scic = vgpu_opregion(vgpu)->va + INTEL_GVT_OPREGION_SCIC;
-	parm = vgpu_opregion(vgpu)->va + INTEL_GVT_OPREGION_PARM;
+	switch (intel_gvt_host.hypervisor_type) {
+	case INTEL_GVT_HYPERVISOR_XEN:
+		scic = vgpu_opregion(vgpu)->va + INTEL_GVT_OPREGION_SCIC;
+		parm = vgpu_opregion(vgpu)->va + INTEL_GVT_OPREGION_PARM;
+		break;
+	case INTEL_GVT_HYPERVISOR_KVM:
+		scic = vgpu_opregion(vgpu)->va_gopregion +
+						INTEL_GVT_OPREGION_SCIC;
+		parm = vgpu_opregion(vgpu)->va_gopregion +
+						INTEL_GVT_OPREGION_PARM;
+		break;
+	default:
+		gvt_vgpu_err("not supported hypervisor\n");
+		return -EINVAL;
+	}
 
 	if (!(swsci & SWSCI_SCI_SELECT)) {
 		gvt_vgpu_err("requesting SMI service\n");
diff --git a/drivers/gpu/drm/i915/gvt/reg.h b/drivers/gpu/drm/i915/gvt/reg.h
index 7d01c77a0f7a..d4f7ce6dc1d7 100644
--- a/drivers/gpu/drm/i915/gvt/reg.h
+++ b/drivers/gpu/drm/i915/gvt/reg.h
@@ -51,6 +51,9 @@
 
 #define INTEL_GVT_OPREGION_PAGES	2
 #define INTEL_GVT_OPREGION_SIZE		(INTEL_GVT_OPREGION_PAGES * PAGE_SIZE)
+#define INTEL_GVT_OPREGION_VBT_OFFSET	0x400
+#define INTEL_GVT_OPREGION_VBT_SIZE	\
+		(INTEL_GVT_OPREGION_SIZE - INTEL_GVT_OPREGION_VBT_OFFSET)
 
 #define VGT_SPRSTRIDE(pipe)	_PIPE(pipe, _SPRA_STRIDE, _PLANE_STRIDE_2_B)
 
@@ -71,6 +74,7 @@
 #define RB_HEAD_OFF_MASK	((1U << 21) - (1U << 2))
 #define RB_TAIL_OFF_MASK	((1U << 21) - (1U << 3))
 #define RB_TAIL_SIZE_MASK	((1U << 21) - (1U << 12))
-#define _RING_CTL_BUF_SIZE(ctl) (((ctl) & RB_TAIL_SIZE_MASK) + GTT_PAGE_SIZE)
+#define _RING_CTL_BUF_SIZE(ctl) (((ctl) & RB_TAIL_SIZE_MASK) + \
+		I915_GTT_PAGE_SIZE)
 
 #endif
diff --git a/drivers/gpu/drm/i915/gvt/render.c b/drivers/gpu/drm/i915/gvt/render.c
deleted file mode 100644
index 6d066cf35478..000000000000
--- a/drivers/gpu/drm/i915/gvt/render.c
+++ /dev/null
@@ -1,405 +0,0 @@
-/*
- * Copyright(c) 2011-2016 Intel Corporation. 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, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- *
- * Authors:
- *    Eddie Dong <eddie.dong@intel.com>
- *    Kevin Tian <kevin.tian@intel.com>
- *
- * Contributors:
- *    Zhi Wang <zhi.a.wang@intel.com>
- *    Changbin Du <changbin.du@intel.com>
- *    Zhenyu Wang <zhenyuw@linux.intel.com>
- *    Tina Zhang <tina.zhang@intel.com>
- *    Bing Niu <bing.niu@intel.com>
- *
- */
-
-#include "i915_drv.h"
-#include "gvt.h"
-#include "trace.h"
-
-struct render_mmio {
-	int ring_id;
-	i915_reg_t reg;
-	u32 mask;
-	bool in_context;
-	u32 value;
-};
-
-static struct render_mmio gen8_render_mmio_list[] __cacheline_aligned = {
-	{RCS, _MMIO(0x229c), 0xffff, false},
-	{RCS, _MMIO(0x2248), 0x0, false},
-	{RCS, _MMIO(0x2098), 0x0, false},
-	{RCS, _MMIO(0x20c0), 0xffff, true},
-	{RCS, _MMIO(0x24d0), 0, false},
-	{RCS, _MMIO(0x24d4), 0, false},
-	{RCS, _MMIO(0x24d8), 0, false},
-	{RCS, _MMIO(0x24dc), 0, false},
-	{RCS, _MMIO(0x24e0), 0, false},
-	{RCS, _MMIO(0x24e4), 0, false},
-	{RCS, _MMIO(0x24e8), 0, false},
-	{RCS, _MMIO(0x24ec), 0, false},
-	{RCS, _MMIO(0x24f0), 0, false},
-	{RCS, _MMIO(0x24f4), 0, false},
-	{RCS, _MMIO(0x24f8), 0, false},
-	{RCS, _MMIO(0x24fc), 0, false},
-	{RCS, _MMIO(0x7004), 0xffff, true},
-	{RCS, _MMIO(0x7008), 0xffff, true},
-	{RCS, _MMIO(0x7000), 0xffff, true},
-	{RCS, _MMIO(0x7010), 0xffff, true},
-	{RCS, _MMIO(0x7300), 0xffff, true},
-	{RCS, _MMIO(0x83a4), 0xffff, true},
-
-	{BCS, _MMIO(0x2229c), 0xffff, false},
-	{BCS, _MMIO(0x2209c), 0xffff, false},
-	{BCS, _MMIO(0x220c0), 0xffff, false},
-	{BCS, _MMIO(0x22098), 0x0, false},
-	{BCS, _MMIO(0x22028), 0x0, false},
-};
-
-static struct render_mmio gen9_render_mmio_list[] __cacheline_aligned = {
-	{RCS, _MMIO(0x229c), 0xffff, false},
-	{RCS, _MMIO(0x2248), 0x0, false},
-	{RCS, _MMIO(0x2098), 0x0, false},
-	{RCS, _MMIO(0x20c0), 0xffff, true},
-	{RCS, _MMIO(0x24d0), 0, false},
-	{RCS, _MMIO(0x24d4), 0, false},
-	{RCS, _MMIO(0x24d8), 0, false},
-	{RCS, _MMIO(0x24dc), 0, false},
-	{RCS, _MMIO(0x24e0), 0, false},
-	{RCS, _MMIO(0x24e4), 0, false},
-	{RCS, _MMIO(0x24e8), 0, false},
-	{RCS, _MMIO(0x24ec), 0, false},
-	{RCS, _MMIO(0x24f0), 0, false},
-	{RCS, _MMIO(0x24f4), 0, false},
-	{RCS, _MMIO(0x24f8), 0, false},
-	{RCS, _MMIO(0x24fc), 0, false},
-	{RCS, _MMIO(0x7004), 0xffff, true},
-	{RCS, _MMIO(0x7008), 0xffff, true},
-	{RCS, _MMIO(0x7000), 0xffff, true},
-	{RCS, _MMIO(0x7010), 0xffff, true},
-	{RCS, _MMIO(0x7300), 0xffff, true},
-	{RCS, _MMIO(0x83a4), 0xffff, true},
-
-	{RCS, _MMIO(0x40e0), 0, false},
-	{RCS, _MMIO(0x40e4), 0, false},
-	{RCS, _MMIO(0x2580), 0xffff, true},
-	{RCS, _MMIO(0x7014), 0xffff, true},
-	{RCS, _MMIO(0x20ec), 0xffff, false},
-	{RCS, _MMIO(0xb118), 0, false},
-	{RCS, _MMIO(0xe100), 0xffff, true},
-	{RCS, _MMIO(0xe180), 0xffff, true},
-	{RCS, _MMIO(0xe184), 0xffff, true},
-	{RCS, _MMIO(0xe188), 0xffff, true},
-	{RCS, _MMIO(0xe194), 0xffff, true},
-	{RCS, _MMIO(0x4de0), 0, false},
-	{RCS, _MMIO(0x4de4), 0, false},
-	{RCS, _MMIO(0x4de8), 0, false},
-	{RCS, _MMIO(0x4dec), 0, false},
-	{RCS, _MMIO(0x4df0), 0, false},
-	{RCS, _MMIO(0x4df4), 0, false},
-
-	{BCS, _MMIO(0x2229c), 0xffff, false},
-	{BCS, _MMIO(0x2209c), 0xffff, false},
-	{BCS, _MMIO(0x220c0), 0xffff, false},
-	{BCS, _MMIO(0x22098), 0x0, false},
-	{BCS, _MMIO(0x22028), 0x0, false},
-
-	{VCS2, _MMIO(0x1c028), 0xffff, false},
-
-	{VECS, _MMIO(0x1a028), 0xffff, false},
-
-	{RCS, _MMIO(0x7304), 0xffff, true},
-	{RCS, _MMIO(0x2248), 0x0, false},
-	{RCS, _MMIO(0x940c), 0x0, false},
-	{RCS, _MMIO(0x4ab8), 0x0, false},
-
-	{RCS, _MMIO(0x4ab0), 0x0, false},
-	{RCS, _MMIO(0x20d4), 0x0, false},
-
-	{RCS, _MMIO(0xb004), 0x0, false},
-	{RCS, _MMIO(0x20a0), 0x0, false},
-	{RCS, _MMIO(0x20e4), 0xffff, false},
-};
-
-static u32 gen9_render_mocs[I915_NUM_ENGINES][64];
-static u32 gen9_render_mocs_L3[32];
-
-static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id)
-{
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	enum forcewake_domains fw;
-	i915_reg_t reg;
-	u32 regs[] = {
-		[RCS] = 0x4260,
-		[VCS] = 0x4264,
-		[VCS2] = 0x4268,
-		[BCS] = 0x426c,
-		[VECS] = 0x4270,
-	};
-
-	if (WARN_ON(ring_id >= ARRAY_SIZE(regs)))
-		return;
-
-	if (!test_and_clear_bit(ring_id, (void *)vgpu->tlb_handle_pending))
-		return;
-
-	reg = _MMIO(regs[ring_id]);
-
-	/* WaForceWakeRenderDuringMmioTLBInvalidate:skl
-	 * we need to put a forcewake when invalidating RCS TLB caches,
-	 * otherwise device can go to RC6 state and interrupt invalidation
-	 * process
-	 */
-	fw = intel_uncore_forcewake_for_reg(dev_priv, reg,
-					    FW_REG_READ | FW_REG_WRITE);
-	if (ring_id == RCS && (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)))
-		fw |= FORCEWAKE_RENDER;
-
-	intel_uncore_forcewake_get(dev_priv, fw);
-
-	I915_WRITE_FW(reg, 0x1);
-
-	if (wait_for_atomic((I915_READ_FW(reg) == 0), 50))
-		gvt_vgpu_err("timeout in invalidate ring (%d) tlb\n", ring_id);
-	else
-		vgpu_vreg(vgpu, regs[ring_id]) = 0;
-
-	intel_uncore_forcewake_put(dev_priv, fw);
-
-	gvt_dbg_core("invalidate TLB for ring %d\n", ring_id);
-}
-
-static void load_mocs(struct intel_vgpu *vgpu, int ring_id)
-{
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	i915_reg_t offset, l3_offset;
-	u32 regs[] = {
-		[RCS] = 0xc800,
-		[VCS] = 0xc900,
-		[VCS2] = 0xca00,
-		[BCS] = 0xcc00,
-		[VECS] = 0xcb00,
-	};
-	int i;
-
-	if (WARN_ON(ring_id >= ARRAY_SIZE(regs)))
-		return;
-
-	offset.reg = regs[ring_id];
-	for (i = 0; i < 64; i++) {
-		gen9_render_mocs[ring_id][i] = I915_READ_FW(offset);
-		I915_WRITE(offset, vgpu_vreg(vgpu, offset));
-		offset.reg += 4;
-	}
-
-	if (ring_id == RCS) {
-		l3_offset.reg = 0xb020;
-		for (i = 0; i < 32; i++) {
-			gen9_render_mocs_L3[i] = I915_READ_FW(l3_offset);
-			I915_WRITE_FW(l3_offset, vgpu_vreg(vgpu, l3_offset));
-			l3_offset.reg += 4;
-		}
-	}
-}
-
-static void restore_mocs(struct intel_vgpu *vgpu, int ring_id)
-{
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	i915_reg_t offset, l3_offset;
-	u32 regs[] = {
-		[RCS] = 0xc800,
-		[VCS] = 0xc900,
-		[VCS2] = 0xca00,
-		[BCS] = 0xcc00,
-		[VECS] = 0xcb00,
-	};
-	int i;
-
-	if (WARN_ON(ring_id >= ARRAY_SIZE(regs)))
-		return;
-
-	offset.reg = regs[ring_id];
-	for (i = 0; i < 64; i++) {
-		vgpu_vreg(vgpu, offset) = I915_READ_FW(offset);
-		I915_WRITE_FW(offset, gen9_render_mocs[ring_id][i]);
-		offset.reg += 4;
-	}
-
-	if (ring_id == RCS) {
-		l3_offset.reg = 0xb020;
-		for (i = 0; i < 32; i++) {
-			vgpu_vreg(vgpu, l3_offset) = I915_READ_FW(l3_offset);
-			I915_WRITE_FW(l3_offset, gen9_render_mocs_L3[i]);
-			l3_offset.reg += 4;
-		}
-	}
-}
-
-#define CTX_CONTEXT_CONTROL_VAL	0x03
-
-/* Switch ring mmio values (context) from host to a vgpu. */
-static void switch_mmio_to_vgpu(struct intel_vgpu *vgpu, int ring_id)
-{
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	struct render_mmio *mmio;
-	u32 v;
-	int i, array_size;
-	u32 *reg_state = vgpu->shadow_ctx->engine[ring_id].lrc_reg_state;
-	u32 ctx_ctrl = reg_state[CTX_CONTEXT_CONTROL_VAL];
-	u32 inhibit_mask =
-		_MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT);
-	i915_reg_t last_reg = _MMIO(0);
-
-	if (IS_SKYLAKE(vgpu->gvt->dev_priv)
-		|| IS_KABYLAKE(vgpu->gvt->dev_priv)) {
-		mmio = gen9_render_mmio_list;
-		array_size = ARRAY_SIZE(gen9_render_mmio_list);
-		load_mocs(vgpu, ring_id);
-	} else {
-		mmio = gen8_render_mmio_list;
-		array_size = ARRAY_SIZE(gen8_render_mmio_list);
-	}
-
-	for (i = 0; i < array_size; i++, mmio++) {
-		if (mmio->ring_id != ring_id)
-			continue;
-
-		mmio->value = I915_READ_FW(mmio->reg);
-
-		/*
-		 * if it is an inhibit context, load in_context mmio
-		 * into HW by mmio write. If it is not, skip this mmio
-		 * write.
-		 */
-		if (mmio->in_context &&
-				((ctx_ctrl & inhibit_mask) != inhibit_mask) &&
-				i915_modparams.enable_execlists)
-			continue;
-
-		if (mmio->mask)
-			v = vgpu_vreg(vgpu, mmio->reg) | (mmio->mask << 16);
-		else
-			v = vgpu_vreg(vgpu, mmio->reg);
-
-		I915_WRITE_FW(mmio->reg, v);
-		last_reg = mmio->reg;
-
-		trace_render_mmio(vgpu->id, "load",
-				  i915_mmio_reg_offset(mmio->reg),
-				  mmio->value, v);
-	}
-
-	/* Make sure the swiched MMIOs has taken effect. */
-	if (likely(INTEL_GVT_MMIO_OFFSET(last_reg)))
-		I915_READ_FW(last_reg);
-
-	handle_tlb_pending_event(vgpu, ring_id);
-}
-
-/* Switch ring mmio values (context) from vgpu to host. */
-static void switch_mmio_to_host(struct intel_vgpu *vgpu, int ring_id)
-{
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	struct render_mmio *mmio;
-	i915_reg_t last_reg = _MMIO(0);
-	u32 v;
-	int i, array_size;
-
-	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
-		mmio = gen9_render_mmio_list;
-		array_size = ARRAY_SIZE(gen9_render_mmio_list);
-		restore_mocs(vgpu, ring_id);
-	} else {
-		mmio = gen8_render_mmio_list;
-		array_size = ARRAY_SIZE(gen8_render_mmio_list);
-	}
-
-	for (i = 0; i < array_size; i++, mmio++) {
-		if (mmio->ring_id != ring_id)
-			continue;
-
-		vgpu_vreg(vgpu, mmio->reg) = I915_READ_FW(mmio->reg);
-
-		if (mmio->mask) {
-			vgpu_vreg(vgpu, mmio->reg) &= ~(mmio->mask << 16);
-			v = mmio->value | (mmio->mask << 16);
-		} else
-			v = mmio->value;
-
-		if (mmio->in_context)
-			continue;
-
-		I915_WRITE_FW(mmio->reg, v);
-		last_reg = mmio->reg;
-
-		trace_render_mmio(vgpu->id, "restore",
-				  i915_mmio_reg_offset(mmio->reg),
-				  mmio->value, v);
-	}
-
-	/* Make sure the swiched MMIOs has taken effect. */
-	if (likely(INTEL_GVT_MMIO_OFFSET(last_reg)))
-		I915_READ_FW(last_reg);
-}
-
-/**
- * intel_gvt_switch_render_mmio - switch mmio context of specific engine
- * @pre: the last vGPU that own the engine
- * @next: the vGPU to switch to
- * @ring_id: specify the engine
- *
- * If pre is null indicates that host own the engine. If next is null
- * indicates that we are switching to host workload.
- */
-void intel_gvt_switch_mmio(struct intel_vgpu *pre,
-			   struct intel_vgpu *next, int ring_id)
-{
-	struct drm_i915_private *dev_priv;
-
-	if (WARN_ON(!pre && !next))
-		return;
-
-	gvt_dbg_render("switch ring %d from %s to %s\n", ring_id,
-		       pre ? "vGPU" : "host", next ? "vGPU" : "HOST");
-
-	dev_priv = pre ? pre->gvt->dev_priv : next->gvt->dev_priv;
-
-	/**
-	 * We are using raw mmio access wrapper to improve the
-	 * performace for batch mmio read/write, so we need
-	 * handle forcewake mannually.
-	 */
-	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
-
-	/**
-	 * TODO: Optimize for vGPU to vGPU switch by merging
-	 * switch_mmio_to_host() and switch_mmio_to_vgpu().
-	 */
-	if (pre)
-		switch_mmio_to_host(pre, ring_id);
-
-	if (next)
-		switch_mmio_to_vgpu(next, ring_id);
-
-	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
-}
diff --git a/drivers/gpu/drm/i915/gvt/sched_policy.c b/drivers/gpu/drm/i915/gvt/sched_policy.c
index 03532dfc0cd5..eea1a2f92099 100644
--- a/drivers/gpu/drm/i915/gvt/sched_policy.c
+++ b/drivers/gpu/drm/i915/gvt/sched_policy.c
@@ -372,6 +372,11 @@ void intel_vgpu_start_schedule(struct intel_vgpu *vgpu)
 	vgpu->gvt->scheduler.sched_ops->start_schedule(vgpu);
 }
 
+void intel_gvt_kick_schedule(struct intel_gvt *gvt)
+{
+	intel_gvt_request_service(gvt, INTEL_GVT_REQUEST_EVENT_SCHED);
+}
+
 void intel_vgpu_stop_schedule(struct intel_vgpu *vgpu)
 {
 	struct intel_gvt_workload_scheduler *scheduler =
diff --git a/drivers/gpu/drm/i915/gvt/sched_policy.h b/drivers/gpu/drm/i915/gvt/sched_policy.h
index ba00a5f7455f..7b59e3e88b8b 100644
--- a/drivers/gpu/drm/i915/gvt/sched_policy.h
+++ b/drivers/gpu/drm/i915/gvt/sched_policy.h
@@ -57,4 +57,6 @@ void intel_vgpu_start_schedule(struct intel_vgpu *vgpu);
 
 void intel_vgpu_stop_schedule(struct intel_vgpu *vgpu);
 
+void intel_gvt_kick_schedule(struct intel_gvt *gvt);
+
 #endif
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 69f8f0d155b9..0056638b0c16 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -57,7 +57,7 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload)
 	struct intel_vgpu *vgpu = workload->vgpu;
 	struct intel_gvt *gvt = vgpu->gvt;
 	int ring_id = workload->ring_id;
-	struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
+	struct i915_gem_context *shadow_ctx = vgpu->submission.shadow_ctx;
 	struct drm_i915_gem_object *ctx_obj =
 		shadow_ctx->engine[ring_id].state->obj;
 	struct execlist_ring_context *shadow_ring_context;
@@ -81,16 +81,16 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload)
 	while (i < context_page_num) {
 		context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
 				(u32)((workload->ctx_desc.lrca + i) <<
-				GTT_PAGE_SHIFT));
+				I915_GTT_PAGE_SHIFT));
 		if (context_gpa == INTEL_GVT_INVALID_ADDR) {
 			gvt_vgpu_err("Invalid guest context descriptor\n");
-			return -EINVAL;
+			return -EFAULT;
 		}
 
 		page = i915_gem_object_get_page(ctx_obj, LRC_HEADER_PAGES + i);
 		dst = kmap(page);
 		intel_gvt_hypervisor_read_gpa(vgpu, context_gpa, dst,
-				GTT_PAGE_SIZE);
+				I915_GTT_PAGE_SIZE);
 		kunmap(page);
 		i++;
 	}
@@ -120,7 +120,7 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload)
 			sizeof(*shadow_ring_context),
 			(void *)shadow_ring_context +
 			sizeof(*shadow_ring_context),
-			GTT_PAGE_SIZE - sizeof(*shadow_ring_context));
+			I915_GTT_PAGE_SIZE - sizeof(*shadow_ring_context));
 
 	kunmap(page);
 	return 0;
@@ -248,7 +248,7 @@ static int copy_workload_to_ring_buffer(struct intel_vgpu_workload *workload)
 	return 0;
 }
 
-void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
 {
 	if (!wa_ctx->indirect_ctx.obj)
 		return;
@@ -267,11 +267,12 @@ void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
  */
 int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload)
 {
+	struct intel_vgpu *vgpu = workload->vgpu;
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct i915_gem_context *shadow_ctx = s->shadow_ctx;
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
 	int ring_id = workload->ring_id;
-	struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
-	struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv;
 	struct intel_engine_cs *engine = dev_priv->engine[ring_id];
-	struct intel_vgpu *vgpu = workload->vgpu;
 	struct intel_ring *ring;
 	int ret;
 
@@ -284,7 +285,7 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload)
 	shadow_ctx->desc_template |= workload->ctx_desc.addressing_mode <<
 				    GEN8_CTX_ADDRESSING_MODE_SHIFT;
 
-	if (!test_and_set_bit(ring_id, vgpu->shadow_ctx_desc_updated))
+	if (!test_and_set_bit(ring_id, s->shadow_ctx_desc_updated))
 		shadow_context_descriptor_update(shadow_ctx,
 					dev_priv->engine[ring_id]);
 
@@ -327,14 +328,15 @@ err_scan:
 	return ret;
 }
 
-int intel_gvt_generate_request(struct intel_vgpu_workload *workload)
+static int intel_gvt_generate_request(struct intel_vgpu_workload *workload)
 {
 	int ring_id = workload->ring_id;
 	struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv;
 	struct intel_engine_cs *engine = dev_priv->engine[ring_id];
 	struct drm_i915_gem_request *rq;
 	struct intel_vgpu *vgpu = workload->vgpu;
-	struct i915_gem_context *shadow_ctx = vgpu->shadow_ctx;
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct i915_gem_context *shadow_ctx = s->shadow_ctx;
 	int ret;
 
 	rq = i915_gem_request_alloc(dev_priv->engine[ring_id], shadow_ctx);
@@ -358,11 +360,203 @@ err_unpin:
 	return ret;
 }
 
+static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload);
+
+static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
+{
+	struct intel_gvt *gvt = workload->vgpu->gvt;
+	const int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
+	struct intel_vgpu_shadow_bb *bb;
+	int ret;
+
+	list_for_each_entry(bb, &workload->shadow_bb, list) {
+		bb->vma = i915_gem_object_ggtt_pin(bb->obj, NULL, 0, 0, 0);
+		if (IS_ERR(bb->vma)) {
+			ret = PTR_ERR(bb->vma);
+			goto err;
+		}
+
+		/* relocate shadow batch buffer */
+		bb->bb_start_cmd_va[1] = i915_ggtt_offset(bb->vma);
+		if (gmadr_bytes == 8)
+			bb->bb_start_cmd_va[2] = 0;
+
+		/* No one is going to touch shadow bb from now on. */
+		if (bb->clflush & CLFLUSH_AFTER) {
+			drm_clflush_virt_range(bb->va, bb->obj->base.size);
+			bb->clflush &= ~CLFLUSH_AFTER;
+		}
+
+		ret = i915_gem_object_set_to_gtt_domain(bb->obj, false);
+		if (ret)
+			goto err;
+
+		i915_gem_obj_finish_shmem_access(bb->obj);
+		bb->accessing = false;
+
+		i915_vma_move_to_active(bb->vma, workload->req, 0);
+	}
+	return 0;
+err:
+	release_shadow_batch_buffer(workload);
+	return ret;
+}
+
+static int update_wa_ctx_2_shadow_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+	struct intel_vgpu_workload *workload = container_of(wa_ctx,
+					struct intel_vgpu_workload,
+					wa_ctx);
+	int ring_id = workload->ring_id;
+	struct intel_vgpu_submission *s = &workload->vgpu->submission;
+	struct i915_gem_context *shadow_ctx = s->shadow_ctx;
+	struct drm_i915_gem_object *ctx_obj =
+		shadow_ctx->engine[ring_id].state->obj;
+	struct execlist_ring_context *shadow_ring_context;
+	struct page *page;
+
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
+	shadow_ring_context = kmap_atomic(page);
+
+	shadow_ring_context->bb_per_ctx_ptr.val =
+		(shadow_ring_context->bb_per_ctx_ptr.val &
+		(~PER_CTX_ADDR_MASK)) | wa_ctx->per_ctx.shadow_gma;
+	shadow_ring_context->rcs_indirect_ctx.val =
+		(shadow_ring_context->rcs_indirect_ctx.val &
+		(~INDIRECT_CTX_ADDR_MASK)) | wa_ctx->indirect_ctx.shadow_gma;
+
+	kunmap_atomic(shadow_ring_context);
+	return 0;
+}
+
+static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
+{
+	struct i915_vma *vma;
+	unsigned char *per_ctx_va =
+		(unsigned char *)wa_ctx->indirect_ctx.shadow_va +
+		wa_ctx->indirect_ctx.size;
+
+	if (wa_ctx->indirect_ctx.size == 0)
+		return 0;
+
+	vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL,
+				       0, CACHELINE_BYTES, 0);
+	if (IS_ERR(vma))
+		return PTR_ERR(vma);
+
+	/* FIXME: we are not tracking our pinned VMA leaving it
+	 * up to the core to fix up the stray pin_count upon
+	 * free.
+	 */
+
+	wa_ctx->indirect_ctx.shadow_gma = i915_ggtt_offset(vma);
+
+	wa_ctx->per_ctx.shadow_gma = *((unsigned int *)per_ctx_va + 1);
+	memset(per_ctx_va, 0, CACHELINE_BYTES);
+
+	update_wa_ctx_2_shadow_ctx(wa_ctx);
+	return 0;
+}
+
+static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
+{
+	struct intel_vgpu *vgpu = workload->vgpu;
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	struct intel_vgpu_shadow_bb *bb, *pos;
+
+	if (list_empty(&workload->shadow_bb))
+		return;
+
+	bb = list_first_entry(&workload->shadow_bb,
+			struct intel_vgpu_shadow_bb, list);
+
+	mutex_lock(&dev_priv->drm.struct_mutex);
+
+	list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) {
+		if (bb->obj) {
+			if (bb->accessing)
+				i915_gem_obj_finish_shmem_access(bb->obj);
+
+			if (bb->va && !IS_ERR(bb->va))
+				i915_gem_object_unpin_map(bb->obj);
+
+			if (bb->vma && !IS_ERR(bb->vma)) {
+				i915_vma_unpin(bb->vma);
+				i915_vma_close(bb->vma);
+			}
+			__i915_gem_object_release_unless_active(bb->obj);
+		}
+		list_del(&bb->list);
+		kfree(bb);
+	}
+
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+}
+
+static int prepare_workload(struct intel_vgpu_workload *workload)
+{
+	struct intel_vgpu *vgpu = workload->vgpu;
+	int ret = 0;
+
+	ret = intel_vgpu_pin_mm(workload->shadow_mm);
+	if (ret) {
+		gvt_vgpu_err("fail to vgpu pin mm\n");
+		return ret;
+	}
+
+	ret = intel_vgpu_sync_oos_pages(workload->vgpu);
+	if (ret) {
+		gvt_vgpu_err("fail to vgpu sync oos pages\n");
+		goto err_unpin_mm;
+	}
+
+	ret = intel_vgpu_flush_post_shadow(workload->vgpu);
+	if (ret) {
+		gvt_vgpu_err("fail to flush post shadow\n");
+		goto err_unpin_mm;
+	}
+
+	ret = intel_gvt_generate_request(workload);
+	if (ret) {
+		gvt_vgpu_err("fail to generate request\n");
+		goto err_unpin_mm;
+	}
+
+	ret = prepare_shadow_batch_buffer(workload);
+	if (ret) {
+		gvt_vgpu_err("fail to prepare_shadow_batch_buffer\n");
+		goto err_unpin_mm;
+	}
+
+	ret = prepare_shadow_wa_ctx(&workload->wa_ctx);
+	if (ret) {
+		gvt_vgpu_err("fail to prepare_shadow_wa_ctx\n");
+		goto err_shadow_batch;
+	}
+
+	if (workload->prepare) {
+		ret = workload->prepare(workload);
+		if (ret)
+			goto err_shadow_wa_ctx;
+	}
+
+	return 0;
+err_shadow_wa_ctx:
+	release_shadow_wa_ctx(&workload->wa_ctx);
+err_shadow_batch:
+	release_shadow_batch_buffer(workload);
+err_unpin_mm:
+	intel_vgpu_unpin_mm(workload->shadow_mm);
+	return ret;
+}
+
 static int dispatch_workload(struct intel_vgpu_workload *workload)
 {
+	struct intel_vgpu *vgpu = workload->vgpu;
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct i915_gem_context *shadow_ctx = s->shadow_ctx;
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
 	int ring_id = workload->ring_id;
-	struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
-	struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv;
 	struct intel_engine_cs *engine = dev_priv->engine[ring_id];
 	int ret = 0;
 
@@ -375,12 +569,10 @@ static int dispatch_workload(struct intel_vgpu_workload *workload)
 	if (ret)
 		goto out;
 
-	if (workload->prepare) {
-		ret = workload->prepare(workload);
-		if (ret) {
-			engine->context_unpin(engine, shadow_ctx);
-			goto out;
-		}
+	ret = prepare_workload(workload);
+	if (ret) {
+		engine->context_unpin(engine, shadow_ctx);
+		goto out;
 	}
 
 out:
@@ -448,7 +640,7 @@ static struct intel_vgpu_workload *pick_next_workload(
 
 	gvt_dbg_sched("ring id %d pick new workload %p\n", ring_id, workload);
 
-	atomic_inc(&workload->vgpu->running_workload_num);
+	atomic_inc(&workload->vgpu->submission.running_workload_num);
 out:
 	mutex_unlock(&gvt->lock);
 	return workload;
@@ -458,8 +650,9 @@ static void update_guest_context(struct intel_vgpu_workload *workload)
 {
 	struct intel_vgpu *vgpu = workload->vgpu;
 	struct intel_gvt *gvt = vgpu->gvt;
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct i915_gem_context *shadow_ctx = s->shadow_ctx;
 	int ring_id = workload->ring_id;
-	struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
 	struct drm_i915_gem_object *ctx_obj =
 		shadow_ctx->engine[ring_id].state->obj;
 	struct execlist_ring_context *shadow_ring_context;
@@ -483,7 +676,7 @@ static void update_guest_context(struct intel_vgpu_workload *workload)
 	while (i < context_page_num) {
 		context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
 				(u32)((workload->ctx_desc.lrca + i) <<
-					GTT_PAGE_SHIFT));
+					I915_GTT_PAGE_SHIFT));
 		if (context_gpa == INTEL_GVT_INVALID_ADDR) {
 			gvt_vgpu_err("invalid guest context descriptor\n");
 			return;
@@ -492,7 +685,7 @@ static void update_guest_context(struct intel_vgpu_workload *workload)
 		page = i915_gem_object_get_page(ctx_obj, LRC_HEADER_PAGES + i);
 		src = kmap(page);
 		intel_gvt_hypervisor_write_gpa(vgpu, context_gpa, src,
-				GTT_PAGE_SIZE);
+				I915_GTT_PAGE_SIZE);
 		kunmap(page);
 		i++;
 	}
@@ -517,23 +710,41 @@ static void update_guest_context(struct intel_vgpu_workload *workload)
 			sizeof(*shadow_ring_context),
 			(void *)shadow_ring_context +
 			sizeof(*shadow_ring_context),
-			GTT_PAGE_SIZE - sizeof(*shadow_ring_context));
+			I915_GTT_PAGE_SIZE - sizeof(*shadow_ring_context));
 
 	kunmap(page);
 }
 
+static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask)
+{
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	struct intel_engine_cs *engine;
+	struct intel_vgpu_workload *pos, *n;
+	unsigned int tmp;
+
+	/* free the unsubmited workloads in the queues. */
+	for_each_engine_masked(engine, dev_priv, engine_mask, tmp) {
+		list_for_each_entry_safe(pos, n,
+			&s->workload_q_head[engine->id], list) {
+			list_del_init(&pos->list);
+			intel_vgpu_destroy_workload(pos);
+		}
+		clear_bit(engine->id, s->shadow_ctx_desc_updated);
+	}
+}
+
 static void complete_current_workload(struct intel_gvt *gvt, int ring_id)
 {
 	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
-	struct intel_vgpu_workload *workload;
-	struct intel_vgpu *vgpu;
+	struct intel_vgpu_workload *workload =
+		scheduler->current_workload[ring_id];
+	struct intel_vgpu *vgpu = workload->vgpu;
+	struct intel_vgpu_submission *s = &vgpu->submission;
 	int event;
 
 	mutex_lock(&gvt->lock);
 
-	workload = scheduler->current_workload[ring_id];
-	vgpu = workload->vgpu;
-
 	/* For the workload w/ request, needs to wait for the context
 	 * switch to make sure request is completed.
 	 * For the workload w/o request, directly complete the workload.
@@ -570,7 +781,7 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id)
 		}
 		mutex_lock(&dev_priv->drm.struct_mutex);
 		/* unpin shadow ctx as the shadow_ctx update is done */
-		engine->context_unpin(engine, workload->vgpu->shadow_ctx);
+		engine->context_unpin(engine, s->shadow_ctx);
 		mutex_unlock(&dev_priv->drm.struct_mutex);
 	}
 
@@ -580,9 +791,32 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id)
 	scheduler->current_workload[ring_id] = NULL;
 
 	list_del_init(&workload->list);
+
+	if (!workload->status) {
+		release_shadow_batch_buffer(workload);
+		release_shadow_wa_ctx(&workload->wa_ctx);
+	}
+
+	if (workload->status || (vgpu->resetting_eng & ENGINE_MASK(ring_id))) {
+		/* if workload->status is not successful means HW GPU
+		 * has occurred GPU hang or something wrong with i915/GVT,
+		 * and GVT won't inject context switch interrupt to guest.
+		 * So this error is a vGPU hang actually to the guest.
+		 * According to this we should emunlate a vGPU hang. If
+		 * there are pending workloads which are already submitted
+		 * from guest, we should clean them up like HW GPU does.
+		 *
+		 * if it is in middle of engine resetting, the pending
+		 * workloads won't be submitted to HW GPU and will be
+		 * cleaned up during the resetting process later, so doing
+		 * the workload clean up here doesn't have any impact.
+		 **/
+		clean_workloads(vgpu, ENGINE_MASK(ring_id));
+	}
+
 	workload->complete(workload);
 
-	atomic_dec(&vgpu->running_workload_num);
+	atomic_dec(&s->running_workload_num);
 	wake_up(&scheduler->workload_complete_wq);
 
 	if (gvt->scheduler.need_reschedule)
@@ -665,20 +899,23 @@ complete:
 					FORCEWAKE_ALL);
 
 		intel_runtime_pm_put(gvt->dev_priv);
+		if (ret && (vgpu_is_vm_unhealthy(ret)))
+			enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR);
 	}
 	return 0;
 }
 
 void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu)
 {
+	struct intel_vgpu_submission *s = &vgpu->submission;
 	struct intel_gvt *gvt = vgpu->gvt;
 	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
 
-	if (atomic_read(&vgpu->running_workload_num)) {
+	if (atomic_read(&s->running_workload_num)) {
 		gvt_dbg_sched("wait vgpu idle\n");
 
 		wait_event(scheduler->workload_complete_wq,
-				!atomic_read(&vgpu->running_workload_num));
+				!atomic_read(&s->running_workload_num));
 	}
 }
 
@@ -743,26 +980,369 @@ err:
 	return ret;
 }
 
-void intel_vgpu_clean_gvt_context(struct intel_vgpu *vgpu)
+/**
+ * intel_vgpu_clean_submission - free submission-related resource for vGPU
+ * @vgpu: a vGPU
+ *
+ * This function is called when a vGPU is being destroyed.
+ *
+ */
+void intel_vgpu_clean_submission(struct intel_vgpu *vgpu)
 {
-	i915_gem_context_put(vgpu->shadow_ctx);
+	struct intel_vgpu_submission *s = &vgpu->submission;
+
+	intel_vgpu_select_submission_ops(vgpu, 0);
+	i915_gem_context_put(s->shadow_ctx);
+	kmem_cache_destroy(s->workloads);
 }
 
-int intel_vgpu_init_gvt_context(struct intel_vgpu *vgpu)
+
+/**
+ * intel_vgpu_reset_submission - reset submission-related resource for vGPU
+ * @vgpu: a vGPU
+ * @engine_mask: engines expected to be reset
+ *
+ * This function is called when a vGPU is being destroyed.
+ *
+ */
+void intel_vgpu_reset_submission(struct intel_vgpu *vgpu,
+		unsigned long engine_mask)
 {
-	atomic_set(&vgpu->running_workload_num, 0);
+	struct intel_vgpu_submission *s = &vgpu->submission;
 
-	vgpu->shadow_ctx = i915_gem_context_create_gvt(
+	if (!s->active)
+		return;
+
+	clean_workloads(vgpu, engine_mask);
+	s->ops->reset(vgpu, engine_mask);
+}
+
+/**
+ * intel_vgpu_setup_submission - setup submission-related resource for vGPU
+ * @vgpu: a vGPU
+ *
+ * This function is called when a vGPU is being created.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
+{
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	enum intel_engine_id i;
+	struct intel_engine_cs *engine;
+	int ret;
+
+	s->shadow_ctx = i915_gem_context_create_gvt(
 			&vgpu->gvt->dev_priv->drm);
-	if (IS_ERR(vgpu->shadow_ctx))
-		return PTR_ERR(vgpu->shadow_ctx);
+	if (IS_ERR(s->shadow_ctx))
+		return PTR_ERR(s->shadow_ctx);
+
+	if (HAS_LOGICAL_RING_PREEMPTION(vgpu->gvt->dev_priv))
+		s->shadow_ctx->priority = INT_MAX;
+
+	bitmap_zero(s->shadow_ctx_desc_updated, I915_NUM_ENGINES);
 
-	if (INTEL_INFO(vgpu->gvt->dev_priv)->has_logical_ring_preemption)
-		vgpu->shadow_ctx->priority = INT_MAX;
+	s->workloads = kmem_cache_create("gvt-g_vgpu_workload",
+			sizeof(struct intel_vgpu_workload), 0,
+			SLAB_HWCACHE_ALIGN,
+			NULL);
 
-	vgpu->shadow_ctx->engine[RCS].initialised = true;
+	if (!s->workloads) {
+		ret = -ENOMEM;
+		goto out_shadow_ctx;
+	}
+
+	for_each_engine(engine, vgpu->gvt->dev_priv, i)
+		INIT_LIST_HEAD(&s->workload_q_head[i]);
 
-	bitmap_zero(vgpu->shadow_ctx_desc_updated, I915_NUM_ENGINES);
+	atomic_set(&s->running_workload_num, 0);
+	bitmap_zero(s->tlb_handle_pending, I915_NUM_ENGINES);
 
 	return 0;
+
+out_shadow_ctx:
+	i915_gem_context_put(s->shadow_ctx);
+	return ret;
+}
+
+/**
+ * intel_vgpu_select_submission_ops - select virtual submission interface
+ * @vgpu: a vGPU
+ * @interface: expected vGPU virtual submission interface
+ *
+ * This function is called when guest configures submission interface.
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ *
+ */
+int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu,
+				     unsigned int interface)
+{
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	const struct intel_vgpu_submission_ops *ops[] = {
+		[INTEL_VGPU_EXECLIST_SUBMISSION] =
+			&intel_vgpu_execlist_submission_ops,
+	};
+	int ret;
+
+	if (WARN_ON(interface >= ARRAY_SIZE(ops)))
+		return -EINVAL;
+
+	if (s->active) {
+		s->ops->clean(vgpu);
+		s->active = false;
+		gvt_dbg_core("vgpu%d: de-select ops [ %s ] \n",
+				vgpu->id, s->ops->name);
+	}
+
+	if (interface == 0) {
+		s->ops = NULL;
+		s->virtual_submission_interface = 0;
+		gvt_dbg_core("vgpu%d: no submission ops\n", vgpu->id);
+		return 0;
+	}
+
+	ret = ops[interface]->init(vgpu);
+	if (ret)
+		return ret;
+
+	s->ops = ops[interface];
+	s->virtual_submission_interface = interface;
+	s->active = true;
+
+	gvt_dbg_core("vgpu%d: activate ops [ %s ]\n",
+			vgpu->id, s->ops->name);
+
+	return 0;
+}
+
+/**
+ * intel_vgpu_destroy_workload - destroy a vGPU workload
+ * @vgpu: a vGPU
+ *
+ * This function is called when destroy a vGPU workload.
+ *
+ */
+void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload)
+{
+	struct intel_vgpu_submission *s = &workload->vgpu->submission;
+
+	if (workload->shadow_mm)
+		intel_gvt_mm_unreference(workload->shadow_mm);
+
+	kmem_cache_free(s->workloads, workload);
+}
+
+static struct intel_vgpu_workload *
+alloc_workload(struct intel_vgpu *vgpu)
+{
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct intel_vgpu_workload *workload;
+
+	workload = kmem_cache_zalloc(s->workloads, GFP_KERNEL);
+	if (!workload)
+		return ERR_PTR(-ENOMEM);
+
+	INIT_LIST_HEAD(&workload->list);
+	INIT_LIST_HEAD(&workload->shadow_bb);
+
+	init_waitqueue_head(&workload->shadow_ctx_status_wq);
+	atomic_set(&workload->shadow_ctx_active, 0);
+
+	workload->status = -EINPROGRESS;
+	workload->shadowed = false;
+	workload->vgpu = vgpu;
+
+	return workload;
+}
+
+#define RING_CTX_OFF(x) \
+	offsetof(struct execlist_ring_context, x)
+
+static void read_guest_pdps(struct intel_vgpu *vgpu,
+		u64 ring_context_gpa, u32 pdp[8])
+{
+	u64 gpa;
+	int i;
+
+	gpa = ring_context_gpa + RING_CTX_OFF(pdp3_UDW.val);
+
+	for (i = 0; i < 8; i++)
+		intel_gvt_hypervisor_read_gpa(vgpu,
+				gpa + i * 8, &pdp[7 - i], 4);
+}
+
+static int prepare_mm(struct intel_vgpu_workload *workload)
+{
+	struct execlist_ctx_descriptor_format *desc = &workload->ctx_desc;
+	struct intel_vgpu_mm *mm;
+	struct intel_vgpu *vgpu = workload->vgpu;
+	int page_table_level;
+	u32 pdp[8];
+
+	if (desc->addressing_mode == 1) { /* legacy 32-bit */
+		page_table_level = 3;
+	} else if (desc->addressing_mode == 3) { /* legacy 64 bit */
+		page_table_level = 4;
+	} else {
+		gvt_vgpu_err("Advanced Context mode(SVM) is not supported!\n");
+		return -EINVAL;
+	}
+
+	read_guest_pdps(workload->vgpu, workload->ring_context_gpa, pdp);
+
+	mm = intel_vgpu_find_ppgtt_mm(workload->vgpu, page_table_level, pdp);
+	if (mm) {
+		intel_gvt_mm_reference(mm);
+	} else {
+
+		mm = intel_vgpu_create_mm(workload->vgpu, INTEL_GVT_MM_PPGTT,
+				pdp, page_table_level, 0);
+		if (IS_ERR(mm)) {
+			gvt_vgpu_err("fail to create mm object.\n");
+			return PTR_ERR(mm);
+		}
+	}
+	workload->shadow_mm = mm;
+	return 0;
+}
+
+#define same_context(a, b) (((a)->context_id == (b)->context_id) && \
+		((a)->lrca == (b)->lrca))
+
+#define get_last_workload(q) \
+	(list_empty(q) ? NULL : container_of(q->prev, \
+	struct intel_vgpu_workload, list))
+/**
+ * intel_vgpu_create_workload - create a vGPU workload
+ * @vgpu: a vGPU
+ * @desc: a guest context descriptor
+ *
+ * This function is called when creating a vGPU workload.
+ *
+ * Returns:
+ * struct intel_vgpu_workload * on success, negative error code in
+ * pointer if failed.
+ *
+ */
+struct intel_vgpu_workload *
+intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id,
+			   struct execlist_ctx_descriptor_format *desc)
+{
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct list_head *q = workload_q_head(vgpu, ring_id);
+	struct intel_vgpu_workload *last_workload = get_last_workload(q);
+	struct intel_vgpu_workload *workload = NULL;
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	u64 ring_context_gpa;
+	u32 head, tail, start, ctl, ctx_ctl, per_ctx, indirect_ctx;
+	int ret;
+
+	ring_context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
+			(u32)((desc->lrca + 1) << I915_GTT_PAGE_SHIFT));
+	if (ring_context_gpa == INTEL_GVT_INVALID_ADDR) {
+		gvt_vgpu_err("invalid guest context LRCA: %x\n", desc->lrca);
+		return ERR_PTR(-EINVAL);
+	}
+
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(ring_header.val), &head, 4);
+
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(ring_tail.val), &tail, 4);
+
+	head &= RB_HEAD_OFF_MASK;
+	tail &= RB_TAIL_OFF_MASK;
+
+	if (last_workload && same_context(&last_workload->ctx_desc, desc)) {
+		gvt_dbg_el("ring id %d cur workload == last\n", ring_id);
+		gvt_dbg_el("ctx head %x real head %lx\n", head,
+				last_workload->rb_tail);
+		/*
+		 * cannot use guest context head pointer here,
+		 * as it might not be updated at this time
+		 */
+		head = last_workload->rb_tail;
+	}
+
+	gvt_dbg_el("ring id %d begin a new workload\n", ring_id);
+
+	/* record some ring buffer register values for scan and shadow */
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(rb_start.val), &start, 4);
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(rb_ctrl.val), &ctl, 4);
+	intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(ctx_ctrl.val), &ctx_ctl, 4);
+
+	workload = alloc_workload(vgpu);
+	if (IS_ERR(workload))
+		return workload;
+
+	workload->ring_id = ring_id;
+	workload->ctx_desc = *desc;
+	workload->ring_context_gpa = ring_context_gpa;
+	workload->rb_head = head;
+	workload->rb_tail = tail;
+	workload->rb_start = start;
+	workload->rb_ctl = ctl;
+
+	if (ring_id == RCS) {
+		intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(bb_per_ctx_ptr.val), &per_ctx, 4);
+		intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa +
+			RING_CTX_OFF(rcs_indirect_ctx.val), &indirect_ctx, 4);
+
+		workload->wa_ctx.indirect_ctx.guest_gma =
+			indirect_ctx & INDIRECT_CTX_ADDR_MASK;
+		workload->wa_ctx.indirect_ctx.size =
+			(indirect_ctx & INDIRECT_CTX_SIZE_MASK) *
+			CACHELINE_BYTES;
+		workload->wa_ctx.per_ctx.guest_gma =
+			per_ctx & PER_CTX_ADDR_MASK;
+		workload->wa_ctx.per_ctx.valid = per_ctx & 1;
+	}
+
+	gvt_dbg_el("workload %p ring id %d head %x tail %x start %x ctl %x\n",
+			workload, ring_id, head, tail, start, ctl);
+
+	ret = prepare_mm(workload);
+	if (ret) {
+		kmem_cache_free(s->workloads, workload);
+		return ERR_PTR(ret);
+	}
+
+	/* Only scan and shadow the first workload in the queue
+	 * as there is only one pre-allocated buf-obj for shadow.
+	 */
+	if (list_empty(workload_q_head(vgpu, ring_id))) {
+		intel_runtime_pm_get(dev_priv);
+		mutex_lock(&dev_priv->drm.struct_mutex);
+		ret = intel_gvt_scan_and_shadow_workload(workload);
+		mutex_unlock(&dev_priv->drm.struct_mutex);
+		intel_runtime_pm_put(dev_priv);
+	}
+
+	if (ret && (vgpu_is_vm_unhealthy(ret))) {
+		enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR);
+		intel_vgpu_destroy_workload(workload);
+		return ERR_PTR(ret);
+	}
+
+	return workload;
+}
+
+/**
+ * intel_vgpu_queue_workload - Qeue a vGPU workload
+ * @workload: the workload to queue in
+ */
+void intel_vgpu_queue_workload(struct intel_vgpu_workload *workload)
+{
+	list_add_tail(&workload->list,
+		workload_q_head(workload->vgpu, workload->ring_id));
+	intel_gvt_kick_schedule(workload->vgpu->gvt);
+	wake_up(&workload->vgpu->gvt->scheduler.waitq[workload->ring_id]);
 }
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h
index b9f872204d7e..3de77dfa7c59 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.h
+++ b/drivers/gpu/drm/i915/gvt/scheduler.h
@@ -112,24 +112,20 @@ struct intel_vgpu_workload {
 	struct intel_shadow_wa_ctx wa_ctx;
 };
 
-/* Intel shadow batch buffer is a i915 gem object */
-struct intel_shadow_bb_entry {
+struct intel_vgpu_shadow_bb {
 	struct list_head list;
 	struct drm_i915_gem_object *obj;
+	struct i915_vma *vma;
 	void *va;
-	unsigned long len;
 	u32 *bb_start_cmd_va;
+	unsigned int clflush;
+	bool accessing;
 };
 
 #define workload_q_head(vgpu, ring_id) \
-	(&(vgpu->workload_q_head[ring_id]))
+	(&(vgpu->submission.workload_q_head[ring_id]))
 
-#define queue_workload(workload) do { \
-	list_add_tail(&workload->list, \
-	workload_q_head(workload->vgpu, workload->ring_id)); \
-	wake_up(&workload->vgpu->gvt-> \
-	scheduler.waitq[workload->ring_id]); \
-} while (0)
+void intel_vgpu_queue_workload(struct intel_vgpu_workload *workload);
 
 int intel_gvt_init_workload_scheduler(struct intel_gvt *gvt);
 
@@ -137,12 +133,23 @@ void intel_gvt_clean_workload_scheduler(struct intel_gvt *gvt);
 
 void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu);
 
-int intel_vgpu_init_gvt_context(struct intel_vgpu *vgpu);
+int intel_vgpu_setup_submission(struct intel_vgpu *vgpu);
 
-void intel_vgpu_clean_gvt_context(struct intel_vgpu *vgpu);
+void intel_vgpu_reset_submission(struct intel_vgpu *vgpu,
+				 unsigned long engine_mask);
 
-void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx);
+void intel_vgpu_clean_submission(struct intel_vgpu *vgpu);
 
-int intel_gvt_generate_request(struct intel_vgpu_workload *workload);
+int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu,
+				     unsigned int interface);
+
+extern const struct intel_vgpu_submission_ops
+intel_vgpu_execlist_submission_ops;
+
+struct intel_vgpu_workload *
+intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id,
+			   struct execlist_ctx_descriptor_format *desc);
+
+void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload);
 
 #endif
diff --git a/drivers/gpu/drm/i915/gvt/trace.h b/drivers/gpu/drm/i915/gvt/trace.h
index 8c150381d9a4..7a2511538f34 100644
--- a/drivers/gpu/drm/i915/gvt/trace.h
+++ b/drivers/gpu/drm/i915/gvt/trace.h
@@ -330,13 +330,14 @@ TRACE_EVENT(inject_msi,
 );
 
 TRACE_EVENT(render_mmio,
-	TP_PROTO(int id, char *action, unsigned int reg,
+	TP_PROTO(int old_id, int new_id, char *action, unsigned int reg,
 		 unsigned int old_val, unsigned int new_val),
 
-	TP_ARGS(id, action, reg, new_val, old_val),
+	TP_ARGS(old_id, new_id, action, reg, new_val, old_val),
 
 	TP_STRUCT__entry(
-		__field(int, id)
+		__field(int, old_id)
+		__field(int, new_id)
 		__array(char, buf, GVT_TEMP_STR_LEN)
 		__field(unsigned int, reg)
 		__field(unsigned int, old_val)
@@ -344,15 +345,17 @@ TRACE_EVENT(render_mmio,
 	),
 
 	TP_fast_assign(
-		__entry->id = id;
+		__entry->old_id = old_id;
+		__entry->new_id = new_id;
 		snprintf(__entry->buf, GVT_TEMP_STR_LEN, "%s", action);
 		__entry->reg = reg;
 		__entry->old_val = old_val;
 		__entry->new_val = new_val;
 	),
 
-	TP_printk("VM%u %s reg %x, old %08x new %08x\n",
-		  __entry->id, __entry->buf, __entry->reg,
+	TP_printk("VM%u -> VM%u %s reg %x, old %08x new %08x\n",
+		  __entry->old_id, __entry->new_id,
+		  __entry->buf, __entry->reg,
 		  __entry->old_val, __entry->new_val)
 );
 
diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c
index 02c61a1ad56a..4688619f6a1c 100644
--- a/drivers/gpu/drm/i915/gvt/vgpu.c
+++ b/drivers/gpu/drm/i915/gvt/vgpu.c
@@ -38,22 +38,25 @@
 void populate_pvinfo_page(struct intel_vgpu *vgpu)
 {
 	/* setup the ballooning information */
-	vgpu_vreg64(vgpu, vgtif_reg(magic)) = VGT_MAGIC;
-	vgpu_vreg(vgpu, vgtif_reg(version_major)) = 1;
-	vgpu_vreg(vgpu, vgtif_reg(version_minor)) = 0;
-	vgpu_vreg(vgpu, vgtif_reg(display_ready)) = 0;
-	vgpu_vreg(vgpu, vgtif_reg(vgt_id)) = vgpu->id;
-	vgpu_vreg(vgpu, vgtif_reg(vgt_caps)) = VGT_CAPS_FULL_48BIT_PPGTT;
-	vgpu_vreg(vgpu, vgtif_reg(avail_rs.mappable_gmadr.base)) =
+	vgpu_vreg64_t(vgpu, vgtif_reg(magic)) = VGT_MAGIC;
+	vgpu_vreg_t(vgpu, vgtif_reg(version_major)) = 1;
+	vgpu_vreg_t(vgpu, vgtif_reg(version_minor)) = 0;
+	vgpu_vreg_t(vgpu, vgtif_reg(display_ready)) = 0;
+	vgpu_vreg_t(vgpu, vgtif_reg(vgt_id)) = vgpu->id;
+
+	vgpu_vreg_t(vgpu, vgtif_reg(vgt_caps)) = VGT_CAPS_FULL_48BIT_PPGTT;
+	vgpu_vreg_t(vgpu, vgtif_reg(vgt_caps)) |= VGT_CAPS_HWSP_EMULATION;
+
+	vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.mappable_gmadr.base)) =
 		vgpu_aperture_gmadr_base(vgpu);
-	vgpu_vreg(vgpu, vgtif_reg(avail_rs.mappable_gmadr.size)) =
+	vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.mappable_gmadr.size)) =
 		vgpu_aperture_sz(vgpu);
-	vgpu_vreg(vgpu, vgtif_reg(avail_rs.nonmappable_gmadr.base)) =
+	vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.nonmappable_gmadr.base)) =
 		vgpu_hidden_gmadr_base(vgpu);
-	vgpu_vreg(vgpu, vgtif_reg(avail_rs.nonmappable_gmadr.size)) =
+	vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.nonmappable_gmadr.size)) =
 		vgpu_hidden_sz(vgpu);
 
-	vgpu_vreg(vgpu, vgtif_reg(avail_rs.fence_num)) = vgpu_fence_sz(vgpu);
+	vgpu_vreg_t(vgpu, vgtif_reg(avail_rs.fence_num)) = vgpu_fence_sz(vgpu);
 
 	gvt_dbg_core("Populate PVINFO PAGE for vGPU %d\n", vgpu->id);
 	gvt_dbg_core("aperture base [GMADR] 0x%llx size 0x%llx\n",
@@ -226,13 +229,14 @@ void intel_gvt_deactivate_vgpu(struct intel_vgpu *vgpu)
 
 	vgpu->active = false;
 
-	if (atomic_read(&vgpu->running_workload_num)) {
+	if (atomic_read(&vgpu->submission.running_workload_num)) {
 		mutex_unlock(&gvt->lock);
 		intel_gvt_wait_vgpu_idle(vgpu);
 		mutex_lock(&gvt->lock);
 	}
 
 	intel_vgpu_stop_schedule(vgpu);
+	intel_vgpu_dmabuf_cleanup(vgpu);
 
 	mutex_unlock(&gvt->lock);
 }
@@ -252,16 +256,17 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu)
 
 	WARN(vgpu->active, "vGPU is still active!\n");
 
+	intel_gvt_debugfs_remove_vgpu(vgpu);
 	idr_remove(&gvt->vgpu_idr, vgpu->id);
 	intel_vgpu_clean_sched_policy(vgpu);
-	intel_vgpu_clean_gvt_context(vgpu);
-	intel_vgpu_clean_execlist(vgpu);
+	intel_vgpu_clean_submission(vgpu);
 	intel_vgpu_clean_display(vgpu);
 	intel_vgpu_clean_opregion(vgpu);
 	intel_vgpu_clean_gtt(vgpu);
 	intel_gvt_hypervisor_detach_vgpu(vgpu);
 	intel_vgpu_free_resource(vgpu);
 	intel_vgpu_clean_mmio(vgpu);
+	intel_vgpu_dmabuf_cleanup(vgpu);
 	vfree(vgpu);
 
 	intel_gvt_update_vgpu_types(gvt);
@@ -293,7 +298,7 @@ struct intel_vgpu *intel_gvt_create_idle_vgpu(struct intel_gvt *gvt)
 	vgpu->gvt = gvt;
 
 	for (i = 0; i < I915_NUM_ENGINES; i++)
-		INIT_LIST_HEAD(&vgpu->workload_q_head[i]);
+		INIT_LIST_HEAD(&vgpu->submission.workload_q_head[i]);
 
 	ret = intel_vgpu_init_sched_policy(vgpu);
 	if (ret)
@@ -346,8 +351,8 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	vgpu->handle = param->handle;
 	vgpu->gvt = gvt;
 	vgpu->sched_ctl.weight = param->weight;
-	bitmap_zero(vgpu->tlb_handle_pending, I915_NUM_ENGINES);
-
+	INIT_LIST_HEAD(&vgpu->dmabuf_obj_list_head);
+	idr_init(&vgpu->object_idr);
 	intel_vgpu_init_cfg_space(vgpu, param->primary);
 
 	ret = intel_vgpu_init_mmio(vgpu);
@@ -368,32 +373,42 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	if (ret)
 		goto out_detach_hypervisor_vgpu;
 
-	ret = intel_vgpu_init_display(vgpu, param->resolution);
+	ret = intel_vgpu_init_opregion(vgpu);
 	if (ret)
 		goto out_clean_gtt;
 
-	ret = intel_vgpu_init_execlist(vgpu);
+	ret = intel_vgpu_init_display(vgpu, param->resolution);
 	if (ret)
-		goto out_clean_display;
+		goto out_clean_opregion;
 
-	ret = intel_vgpu_init_gvt_context(vgpu);
+	ret = intel_vgpu_setup_submission(vgpu);
 	if (ret)
-		goto out_clean_execlist;
+		goto out_clean_display;
 
 	ret = intel_vgpu_init_sched_policy(vgpu);
 	if (ret)
-		goto out_clean_shadow_ctx;
+		goto out_clean_submission;
+
+	ret = intel_gvt_debugfs_add_vgpu(vgpu);
+	if (ret)
+		goto out_clean_sched_policy;
+
+	ret = intel_gvt_hypervisor_set_opregion(vgpu);
+	if (ret)
+		goto out_clean_sched_policy;
 
 	mutex_unlock(&gvt->lock);
 
 	return vgpu;
 
-out_clean_shadow_ctx:
-	intel_vgpu_clean_gvt_context(vgpu);
-out_clean_execlist:
-	intel_vgpu_clean_execlist(vgpu);
+out_clean_sched_policy:
+	intel_vgpu_clean_sched_policy(vgpu);
+out_clean_submission:
+	intel_vgpu_clean_submission(vgpu);
 out_clean_display:
 	intel_vgpu_clean_display(vgpu);
+out_clean_opregion:
+	intel_vgpu_clean_opregion(vgpu);
 out_clean_gtt:
 	intel_vgpu_clean_gtt(vgpu);
 out_detach_hypervisor_vgpu:
@@ -500,10 +515,10 @@ void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr,
 		mutex_lock(&gvt->lock);
 	}
 
-	intel_vgpu_reset_execlist(vgpu, resetting_eng);
-
+	intel_vgpu_reset_submission(vgpu, resetting_eng);
 	/* full GPU reset or device model level reset */
 	if (engine_mask == ALL_ENGINES || dmlr) {
+		intel_vgpu_select_submission_ops(vgpu, 0);
 
 		/*fence will not be reset during virtual reset */
 		if (dmlr) {