Merge tag 'drm-for-v4.14' of git://people.freedesktop.org/~airlied/linux

Pull drm updates from Dave Airlie:
 "This is the main drm pull request for 4.14 merge window.

  I'm sending this early, as my continuing journey into fatherhood is
  occurring really soon now, I'm going to be mostly useless for the next
  couple of weeks, though I may be able to read email, I doubt I'll be
  doing much patch applications or git sending. If anything urgent pops
  up I've asked Daniel/Jani/Alex/Sean to try and direct stuff towards
  you.

  Outside drm changes:

  Some rcar-du updates that touch the V4L tree, all acks should be in
  place. It adds one export to the radix tree code for new i915 use
  case. There are some minor AGP cleanups (don't see that too often).
  Changes to the vbox driver in staging to avoid breaking compilation.

  Summary:

  core:
   - Atomic helper fixes
   - Atomic UAPI fixes
   - Add YCBCR 4:2:0 support
   - Drop set_busid hook
   - Refactor fb_helper locking
   - Remove a bunch of internal APIs
   - Add a bunch of better default handlers
   - Format modifier/blob plane property added
   - More internal header refactoring
   - Make more internal API names consistent
   - Enhanced syncobj APIs (wait/signal/reset/create signalled)

  bridge:
   - Add Synopsys Designware MIPI DSI host bridge driver

  tiny:
   - Add Pervasive Displays RePaper displays
   - Add support for LEGO MINDSTORMS EV3 LCD

  i915:
   - Lots of GEN10/CNL  support patches
   - drm syncobj support
   - Skylake+ watermark refactoring
   - GVT vGPU 48-bit ppgtt support
   - GVT performance improvements
   - NOA change ioctl
   - CCS (color compression) scanout support
   - GPU reset improvements

  amdgpu:
   - Initial hugepage support
   - BO migration logic rework
   - Vega10 improvements
   - Powerplay fixes
   - Stop reprogramming the MC
   - Fixes for ACP audio on stoney
   - SR-IOV fixes/improvements
   - Command submission overhead improvements

  amdkfd:
   - Non-dGPU upstreaming patches
   - Scratch VA ioctl
   - Image tiling modes
   - Update PM4 headers for new firmware
   - Drop all BUG_ONs.

  nouveau:
   - GP108 modesetting support.
   - Disable MSI on big endian.

  vmwgfx:
   - Add fence fd support.

  msm:
   - Runtime PM improvements

  exynos:
   - NV12MT support
   - Refactor KMS drivers

  imx-drm:
   - Lock scanout channel to improve memory bw
   - Cleanups

  etnaviv:
   - GEM object population fixes

  tegra:
   - Prep work for Tegra186 support
   - PRIME mmap support

  sunxi:
   - HDMI support improvements
   - HDMI CEC support

  omapdrm:
   - HDMI hotplug IRQ support
   - Big driver cleanup
   - OMAP5 DSI support

  rcar-du:
   - vblank fixes
   - VSP1 updates

  arcgpu:
   - Minor fixes

  stm:
   - Add STM32 DSI controller driver

  dw_hdmi:
   - Add support for Rockchip RK3399
   - HDMI CEC support

  atmel-hlcdc:
   - Add 8-bit color support

  vc4:
   - Atomic fixes
   - New ioctl to attach a label to a buffer object
   - HDMI CEC support
   - Allow userspace to dictate rendering order on submit ioctl"

* tag 'drm-for-v4.14' of git://people.freedesktop.org/~airlied/linux: (1074 commits)
  drm/syncobj: Add a signal ioctl (v3)
  drm/syncobj: Add a reset ioctl (v3)
  drm/syncobj: Add a syncobj_array_find helper
  drm/syncobj: Allow wait for submit and signal behavior (v5)
  drm/syncobj: Add a CREATE_SIGNALED flag
  drm/syncobj: Add a callback mechanism for replace_fence (v3)
  drm/syncobj: add sync obj wait interface. (v8)
  i915: Use drm_syncobj_fence_get
  drm/syncobj: Add a race-free drm_syncobj_fence_get helper (v2)
  drm/syncobj: Rename fence_get to find_fence
  drm: kirin: Add mode_valid logic to avoid mode clocks we can't generate
  drm/vmwgfx: Bump the version for fence FD support
  drm/vmwgfx: Add export fence to file descriptor support
  drm/vmwgfx: Add support for imported Fence File Descriptor
  drm/vmwgfx: Prepare to support fence fd
  drm/vmwgfx: Fix incorrect command header offset at restart
  drm/vmwgfx: Support the NOP_ERROR command
  drm/vmwgfx: Restart command buffers after errors
  drm/vmwgfx: Move irq bottom half processing to threads
  drm/vmwgfx: Don't use drm_irq_[un]install
  ...

14 files changed, 295 insertions, 139 deletions

diff --git a/drivers/gpu/drm/i915/gvt/aperture_gm.c b/drivers/gpu/drm/i915/gvt/aperture_gm.c
index 325618d969fe..ca3d1925beda 100644
--- a/drivers/gpu/drm/i915/gvt/aperture_gm.c
+++ b/drivers/gpu/drm/i915/gvt/aperture_gm.c
@@ -285,8 +285,8 @@ static int alloc_resource(struct intel_vgpu *vgpu,
 	return 0;
 
 no_enough_resource:
-	gvt_vgpu_err("fail to allocate resource %s\n", item);
-	gvt_vgpu_err("request %luMB avail %luMB max %luMB taken %luMB\n",
+	gvt_err("fail to allocate resource %s\n", item);
+	gvt_err("request %luMB avail %luMB max %luMB taken %luMB\n",
 		BYTES_TO_MB(request), BYTES_TO_MB(avail),
 		BYTES_TO_MB(max), BYTES_TO_MB(taken));
 	return -ENOSPC;
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
index e556a46cd4c2..21c36e256884 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.c
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -1382,13 +1382,13 @@ static inline int cmd_address_audit(struct parser_exec_state *s,
 			ret = -EINVAL;
 			goto err;
 		}
-	} else if ((!vgpu_gmadr_is_valid(s->vgpu, guest_gma)) ||
-			(!vgpu_gmadr_is_valid(s->vgpu,
-					      guest_gma + op_size - 1))) {
+	} else if (!intel_gvt_ggtt_validate_range(vgpu, guest_gma, op_size)) {
 		ret = -EINVAL;
 		goto err;
 	}
+
 	return 0;
+
 err:
 	gvt_vgpu_err("cmd_parser: Malicious %s detected, addr=0x%lx, len=%d!\n",
 			s->info->name, guest_gma, op_size);
@@ -2647,7 +2647,7 @@ static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload)
 	return 0;
 }
 
-int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload)
+int intel_gvt_scan_and_shadow_ringbuffer(struct intel_vgpu_workload *workload)
 {
 	int ret;
 	struct intel_vgpu *vgpu = workload->vgpu;
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.h b/drivers/gpu/drm/i915/gvt/cmd_parser.h
index bed33514103c..286703643002 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.h
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.h
@@ -42,7 +42,7 @@ void intel_gvt_clean_cmd_parser(struct intel_gvt *gvt);
 
 int intel_gvt_init_cmd_parser(struct intel_gvt *gvt);
 
-int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload);
+int intel_gvt_scan_and_shadow_ringbuffer(struct intel_vgpu_workload *workload);
 
 int intel_gvt_scan_and_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx);
 
diff --git a/drivers/gpu/drm/i915/gvt/display.c b/drivers/gpu/drm/i915/gvt/display.c
index 7cb0818a13de..3c318439a659 100644
--- a/drivers/gpu/drm/i915/gvt/display.c
+++ b/drivers/gpu/drm/i915/gvt/display.c
@@ -178,9 +178,9 @@ static void emulate_monitor_status_change(struct intel_vgpu *vgpu)
 				SDE_PORTE_HOTPLUG_SPT);
 		vgpu_vreg(vgpu, SKL_FUSE_STATUS) |=
 				SKL_FUSE_DOWNLOAD_STATUS |
-				SKL_FUSE_PG0_DIST_STATUS |
-				SKL_FUSE_PG1_DIST_STATUS |
-				SKL_FUSE_PG2_DIST_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) |=
 				LCPLL_PLL_ENABLE |
 				LCPLL_PLL_LOCK;
diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c
index 1648887d3f55..91b4300f3b39 100644
--- a/drivers/gpu/drm/i915/gvt/execlist.c
+++ b/drivers/gpu/drm/i915/gvt/execlist.c
@@ -622,6 +622,7 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id,
 	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;
@@ -685,6 +686,7 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id,
 	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 +
@@ -718,6 +720,17 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id,
 		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);
 	return 0;
 }
@@ -800,6 +813,8 @@ static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask)
 			list_del_init(&pos->list);
 			free_workload(pos);
 		}
+
+		clear_bit(engine->id, vgpu->shadow_ctx_desc_updated);
 	}
 }
 
diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c
index 6166e34d892b..e6dfc3331f4b 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.c
+++ b/drivers/gpu/drm/i915/gvt/gtt.c
@@ -259,7 +259,7 @@ static void write_pte64(struct drm_i915_private *dev_priv,
 	writeq(pte, addr);
 }
 
-static inline struct intel_gvt_gtt_entry *gtt_get_entry64(void *pt,
+static inline int gtt_get_entry64(void *pt,
 		struct intel_gvt_gtt_entry *e,
 		unsigned long index, bool hypervisor_access, unsigned long gpa,
 		struct intel_vgpu *vgpu)
@@ -268,22 +268,23 @@ static inline struct intel_gvt_gtt_entry *gtt_get_entry64(void *pt,
 	int ret;
 
 	if (WARN_ON(info->gtt_entry_size != 8))
-		return e;
+		return -EINVAL;
 
 	if (hypervisor_access) {
 		ret = intel_gvt_hypervisor_read_gpa(vgpu, gpa +
 				(index << info->gtt_entry_size_shift),
 				&e->val64, 8);
-		WARN_ON(ret);
+		if (WARN_ON(ret))
+			return ret;
 	} else if (!pt) {
 		e->val64 = read_pte64(vgpu->gvt->dev_priv, index);
 	} else {
 		e->val64 = *((u64 *)pt + index);
 	}
-	return e;
+	return 0;
 }
 
-static inline struct intel_gvt_gtt_entry *gtt_set_entry64(void *pt,
+static inline int gtt_set_entry64(void *pt,
 		struct intel_gvt_gtt_entry *e,
 		unsigned long index, bool hypervisor_access, unsigned long gpa,
 		struct intel_vgpu *vgpu)
@@ -292,19 +293,20 @@ static inline struct intel_gvt_gtt_entry *gtt_set_entry64(void *pt,
 	int ret;
 
 	if (WARN_ON(info->gtt_entry_size != 8))
-		return e;
+		return -EINVAL;
 
 	if (hypervisor_access) {
 		ret = intel_gvt_hypervisor_write_gpa(vgpu, gpa +
 				(index << info->gtt_entry_size_shift),
 				&e->val64, 8);
-		WARN_ON(ret);
+		if (WARN_ON(ret))
+			return ret;
 	} else if (!pt) {
 		write_pte64(vgpu->gvt->dev_priv, index, e->val64);
 	} else {
 		*((u64 *)pt + index) = e->val64;
 	}
-	return e;
+	return 0;
 }
 
 #define GTT_HAW 46
@@ -445,21 +447,25 @@ static int gtt_entry_p2m(struct intel_vgpu *vgpu, struct intel_gvt_gtt_entry *p,
 /*
  * MM helpers.
  */
-struct intel_gvt_gtt_entry *intel_vgpu_mm_get_entry(struct intel_vgpu_mm *mm,
+int intel_vgpu_mm_get_entry(struct intel_vgpu_mm *mm,
 		void *page_table, struct intel_gvt_gtt_entry *e,
 		unsigned long index)
 {
 	struct intel_gvt *gvt = mm->vgpu->gvt;
 	struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
+	int ret;
 
 	e->type = mm->page_table_entry_type;
 
-	ops->get_entry(page_table, e, index, false, 0, mm->vgpu);
+	ret = ops->get_entry(page_table, e, index, false, 0, mm->vgpu);
+	if (ret)
+		return ret;
+
 	ops->test_pse(e);
-	return e;
+	return 0;
 }
 
-struct intel_gvt_gtt_entry *intel_vgpu_mm_set_entry(struct intel_vgpu_mm *mm,
+int intel_vgpu_mm_set_entry(struct intel_vgpu_mm *mm,
 		void *page_table, struct intel_gvt_gtt_entry *e,
 		unsigned long index)
 {
@@ -472,7 +478,7 @@ struct intel_gvt_gtt_entry *intel_vgpu_mm_set_entry(struct intel_vgpu_mm *mm,
 /*
  * PPGTT shadow page table helpers.
  */
-static inline struct intel_gvt_gtt_entry *ppgtt_spt_get_entry(
+static inline int ppgtt_spt_get_entry(
 		struct intel_vgpu_ppgtt_spt *spt,
 		void *page_table, int type,
 		struct intel_gvt_gtt_entry *e, unsigned long index,
@@ -480,20 +486,24 @@ static inline struct intel_gvt_gtt_entry *ppgtt_spt_get_entry(
 {
 	struct intel_gvt *gvt = spt->vgpu->gvt;
 	struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
+	int ret;
 
 	e->type = get_entry_type(type);
 
 	if (WARN(!gtt_type_is_entry(e->type), "invalid entry type\n"))
-		return e;
+		return -EINVAL;
 
-	ops->get_entry(page_table, e, index, guest,
+	ret = ops->get_entry(page_table, e, index, guest,
 			spt->guest_page.gfn << GTT_PAGE_SHIFT,
 			spt->vgpu);
+	if (ret)
+		return ret;
+
 	ops->test_pse(e);
-	return e;
+	return 0;
 }
 
-static inline struct intel_gvt_gtt_entry *ppgtt_spt_set_entry(
+static inline int ppgtt_spt_set_entry(
 		struct intel_vgpu_ppgtt_spt *spt,
 		void *page_table, int type,
 		struct intel_gvt_gtt_entry *e, unsigned long index,
@@ -503,7 +513,7 @@ static inline struct intel_gvt_gtt_entry *ppgtt_spt_set_entry(
 	struct intel_gvt_gtt_pte_ops *ops = gvt->gtt.pte_ops;
 
 	if (WARN(!gtt_type_is_entry(e->type), "invalid entry type\n"))
-		return e;
+		return -EINVAL;
 
 	return ops->set_entry(page_table, e, index, guest,
 			spt->guest_page.gfn << GTT_PAGE_SHIFT,
@@ -792,13 +802,13 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_find_shadow_page(
 
 #define for_each_present_guest_entry(spt, e, i) \
 	for (i = 0; i < pt_entries(spt); i++) \
-	if (spt->vgpu->gvt->gtt.pte_ops->test_present( \
-		ppgtt_get_guest_entry(spt, e, i)))
+		if (!ppgtt_get_guest_entry(spt, e, i) && \
+		    spt->vgpu->gvt->gtt.pte_ops->test_present(e))
 
 #define for_each_present_shadow_entry(spt, e, i) \
 	for (i = 0; i < pt_entries(spt); i++) \
-	if (spt->vgpu->gvt->gtt.pte_ops->test_present( \
-		ppgtt_get_shadow_entry(spt, e, i)))
+		if (!ppgtt_get_shadow_entry(spt, e, i) && \
+		    spt->vgpu->gvt->gtt.pte_ops->test_present(e))
 
 static void ppgtt_get_shadow_page(struct intel_vgpu_ppgtt_spt *spt)
 {
@@ -979,29 +989,26 @@ fail:
 }
 
 static int ppgtt_handle_guest_entry_removal(struct intel_vgpu_guest_page *gpt,
-		unsigned long index)
+		struct intel_gvt_gtt_entry *se, unsigned long index)
 {
 	struct intel_vgpu_ppgtt_spt *spt = guest_page_to_ppgtt_spt(gpt);
 	struct intel_vgpu_shadow_page *sp = &spt->shadow_page;
 	struct intel_vgpu *vgpu = spt->vgpu;
 	struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
-	struct intel_gvt_gtt_entry e;
 	int ret;
 
-	ppgtt_get_shadow_entry(spt, &e, index);
-
-	trace_gpt_change(spt->vgpu->id, "remove", spt, sp->type, e.val64,
+	trace_gpt_change(spt->vgpu->id, "remove", spt, sp->type, se->val64,
 			 index);
 
-	if (!ops->test_present(&e))
+	if (!ops->test_present(se))
 		return 0;
 
-	if (ops->get_pfn(&e) == vgpu->gtt.scratch_pt[sp->type].page_mfn)
+	if (ops->get_pfn(se) == vgpu->gtt.scratch_pt[sp->type].page_mfn)
 		return 0;
 
-	if (gtt_type_is_pt(get_next_pt_type(e.type))) {
+	if (gtt_type_is_pt(get_next_pt_type(se->type))) {
 		struct intel_vgpu_ppgtt_spt *s =
-			ppgtt_find_shadow_page(vgpu, ops->get_pfn(&e));
+			ppgtt_find_shadow_page(vgpu, ops->get_pfn(se));
 		if (!s) {
 			gvt_vgpu_err("fail to find guest page\n");
 			ret = -ENXIO;
@@ -1011,12 +1018,10 @@ static int ppgtt_handle_guest_entry_removal(struct intel_vgpu_guest_page *gpt,
 		if (ret)
 			goto fail;
 	}
-	ops->set_pfn(&e, vgpu->gtt.scratch_pt[sp->type].page_mfn);
-	ppgtt_set_shadow_entry(spt, &e, index);
 	return 0;
 fail:
 	gvt_vgpu_err("fail: shadow page %p guest entry 0x%llx type %d\n",
-			spt, e.val64, e.type);
+			spt, se->val64, se->type);
 	return ret;
 }
 
@@ -1236,22 +1241,37 @@ static int ppgtt_handle_guest_write_page_table(
 {
 	struct intel_vgpu_ppgtt_spt *spt = guest_page_to_ppgtt_spt(gpt);
 	struct intel_vgpu *vgpu = spt->vgpu;
+	int type = spt->shadow_page.type;
 	struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
+	struct intel_gvt_gtt_entry se;
 
 	int ret;
 	int new_present;
 
 	new_present = ops->test_present(we);
 
-	ret = ppgtt_handle_guest_entry_removal(gpt, index);
-	if (ret)
-		goto fail;
+	/*
+	 * Adding the new entry first and then removing the old one, that can
+	 * guarantee the ppgtt table is validated during the window between
+	 * adding and removal.
+	 */
+	ppgtt_get_shadow_entry(spt, &se, index);
 
 	if (new_present) {
 		ret = ppgtt_handle_guest_entry_add(gpt, we, index);
 		if (ret)
 			goto fail;
 	}
+
+	ret = ppgtt_handle_guest_entry_removal(gpt, &se, index);
+	if (ret)
+		goto fail;
+
+	if (!new_present) {
+		ops->set_pfn(&se, vgpu->gtt.scratch_pt[type].page_mfn);
+		ppgtt_set_shadow_entry(spt, &se, index);
+	}
+
 	return 0;
 fail:
 	gvt_vgpu_err("fail: shadow page %p guest entry 0x%llx type %d.\n",
@@ -1323,7 +1343,7 @@ static int ppgtt_handle_guest_write_page_table_bytes(void *gp,
 	struct intel_vgpu *vgpu = spt->vgpu;
 	struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops;
 	const struct intel_gvt_device_info *info = &vgpu->gvt->device_info;
-	struct intel_gvt_gtt_entry we;
+	struct intel_gvt_gtt_entry we, se;
 	unsigned long index;
 	int ret;
 
@@ -1339,7 +1359,8 @@ static int ppgtt_handle_guest_write_page_table_bytes(void *gp,
 			return ret;
 	} else {
 		if (!test_bit(index, spt->post_shadow_bitmap)) {
-			ret = ppgtt_handle_guest_entry_removal(gpt, index);
+			ppgtt_get_shadow_entry(spt, &se, index);
+			ret = ppgtt_handle_guest_entry_removal(gpt, &se, index);
 			if (ret)
 				return ret;
 		}
@@ -1713,8 +1734,10 @@ unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm, unsigned long gma)
 		if (!vgpu_gmadr_is_valid(vgpu, gma))
 			goto err;
 
-		ggtt_get_guest_entry(mm, &e,
-			gma_ops->gma_to_ggtt_pte_index(gma));
+		ret = ggtt_get_guest_entry(mm, &e,
+				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);
 
@@ -1724,7 +1747,9 @@ unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm, unsigned long gma)
 
 	switch (mm->page_table_level) {
 	case 4:
-		ppgtt_get_shadow_root_entry(mm, &e, 0);
+		ret = ppgtt_get_shadow_root_entry(mm, &e, 0);
+		if (ret)
+			goto err;
 		gma_index[0] = gma_ops->gma_to_pml4_index(gma);
 		gma_index[1] = gma_ops->gma_to_l4_pdp_index(gma);
 		gma_index[2] = gma_ops->gma_to_pde_index(gma);
@@ -1732,15 +1757,19 @@ unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm, unsigned long gma)
 		index = 4;
 		break;
 	case 3:
-		ppgtt_get_shadow_root_entry(mm, &e,
+		ret = ppgtt_get_shadow_root_entry(mm, &e,
 				gma_ops->gma_to_l3_pdp_index(gma));
+		if (ret)
+			goto err;
 		gma_index[0] = gma_ops->gma_to_pde_index(gma);
 		gma_index[1] = gma_ops->gma_to_pte_index(gma);
 		index = 2;
 		break;
 	case 2:
-		ppgtt_get_shadow_root_entry(mm, &e,
+		ret = ppgtt_get_shadow_root_entry(mm, &e,
 				gma_ops->gma_to_pde_index(gma));
+		if (ret)
+			goto err;
 		gma_index[0] = gma_ops->gma_to_pte_index(gma);
 		index = 1;
 		break;
@@ -1755,6 +1784,11 @@ unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm, unsigned long gma)
 			(i == index - 1));
 		if (ret)
 			goto err;
+
+		if (!pte_ops->test_present(&e)) {
+			gvt_dbg_core("GMA 0x%lx is not present\n", gma);
+			goto err;
+		}
 	}
 
 	gpa = (pte_ops->get_pfn(&e) << GTT_PAGE_SHIFT)
@@ -2329,13 +2363,12 @@ void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu)
 /**
  * intel_vgpu_reset_gtt - reset the all GTT related status
  * @vgpu: a vGPU
- * @dmlr: true for vGPU Device Model Level Reset, false for GT Reset
  *
  * This function is called from vfio core to reset reset all
  * GTT related status, including GGTT, PPGTT, scratch page.
  *
  */
-void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu, bool dmlr)
+void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu)
 {
 	int i;
 
@@ -2347,9 +2380,6 @@ void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu, bool dmlr)
 	 */
 	intel_vgpu_free_mm(vgpu, INTEL_GVT_MM_PPGTT);
 
-	if (!dmlr)
-		return;
-
 	intel_vgpu_reset_ggtt(vgpu);
 
 	/* clear scratch page for security */
diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h
index f88eb5e89bea..30a4c8d16026 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.h
+++ b/drivers/gpu/drm/i915/gvt/gtt.h
@@ -49,14 +49,18 @@ struct intel_gvt_gtt_entry {
 };
 
 struct intel_gvt_gtt_pte_ops {
-	struct intel_gvt_gtt_entry *(*get_entry)(void *pt,
-		struct intel_gvt_gtt_entry *e,
-		unsigned long index, bool hypervisor_access, unsigned long gpa,
-		struct intel_vgpu *vgpu);
-	struct intel_gvt_gtt_entry *(*set_entry)(void *pt,
-		struct intel_gvt_gtt_entry *e,
-		unsigned long index, bool hypervisor_access, unsigned long gpa,
-		struct intel_vgpu *vgpu);
+	int (*get_entry)(void *pt,
+			 struct intel_gvt_gtt_entry *e,
+			 unsigned long index,
+			 bool hypervisor_access,
+			 unsigned long gpa,
+			 struct intel_vgpu *vgpu);
+	int (*set_entry)(void *pt,
+			 struct intel_gvt_gtt_entry *e,
+			 unsigned long index,
+			 bool hypervisor_access,
+			 unsigned long gpa,
+			 struct intel_vgpu *vgpu);
 	bool (*test_present)(struct intel_gvt_gtt_entry *e);
 	void (*clear_present)(struct intel_gvt_gtt_entry *e);
 	bool (*test_pse)(struct intel_gvt_gtt_entry *e);
@@ -143,12 +147,12 @@ struct intel_vgpu_mm {
 	struct intel_vgpu *vgpu;
 };
 
-extern struct intel_gvt_gtt_entry *intel_vgpu_mm_get_entry(
+extern int intel_vgpu_mm_get_entry(
 		struct intel_vgpu_mm *mm,
 		void *page_table, struct intel_gvt_gtt_entry *e,
 		unsigned long index);
 
-extern struct intel_gvt_gtt_entry *intel_vgpu_mm_set_entry(
+extern int intel_vgpu_mm_set_entry(
 		struct intel_vgpu_mm *mm,
 		void *page_table, struct intel_gvt_gtt_entry *e,
 		unsigned long index);
@@ -208,7 +212,7 @@ extern void intel_vgpu_clean_gtt(struct intel_vgpu *vgpu);
 void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu);
 
 extern int intel_gvt_init_gtt(struct intel_gvt *gvt);
-extern void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu, bool dmlr);
+void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu);
 extern void intel_gvt_clean_gtt(struct intel_gvt *gvt);
 
 extern struct intel_vgpu_mm *intel_gvt_find_ppgtt_mm(struct intel_vgpu *vgpu,
diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h
index 2964a4d01a66..44b719eda8c4 100644
--- a/drivers/gpu/drm/i915/gvt/gvt.h
+++ b/drivers/gpu/drm/i915/gvt/gvt.h
@@ -167,6 +167,7 @@ struct intel_vgpu {
 	atomic_t running_workload_num;
 	DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
 	struct i915_gem_context *shadow_ctx;
+	DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES);
 
 #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT)
 	struct {
@@ -482,6 +483,8 @@ int intel_vgpu_init_opregion(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);
+
 struct intel_gvt_ops {
 	int (*emulate_cfg_read)(struct intel_vgpu *, unsigned int, void *,
 				unsigned int);
diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c
index feed9921b3b3..3502a59166ff 100644
--- a/drivers/gpu/drm/i915/gvt/handlers.c
+++ b/drivers/gpu/drm/i915/gvt/handlers.c
@@ -113,9 +113,17 @@ static int new_mmio_info(struct intel_gvt *gvt,
 
 		info->offset = i;
 		p = find_mmio_info(gvt, info->offset);
-		if (p)
-			gvt_err("dup mmio definition offset %x\n",
+		if (p) {
+			WARN(1, "dup mmio definition offset %x\n",
 				info->offset);
+			kfree(info);
+
+			/* We return -EEXIST here to make GVT-g load fail.
+			 * So duplicated MMIO can be found as soon as
+			 * possible.
+			 */
+			return -EEXIST;
+		}
 
 		info->ro_mask = ro_mask;
 		info->device = device;
@@ -1222,10 +1230,12 @@ static int power_well_ctl_mmio_write(struct intel_vgpu *vgpu,
 {
 	write_vreg(vgpu, offset, p_data, bytes);
 
-	if (vgpu_vreg(vgpu, offset) & HSW_PWR_WELL_ENABLE_REQUEST)
-		vgpu_vreg(vgpu, offset) |= HSW_PWR_WELL_STATE_ENABLED;
+	if (vgpu_vreg(vgpu, offset) & HSW_PWR_WELL_CTL_REQ(HSW_DISP_PW_GLOBAL))
+		vgpu_vreg(vgpu, offset) |=
+			HSW_PWR_WELL_CTL_STATE(HSW_DISP_PW_GLOBAL);
 	else
-		vgpu_vreg(vgpu, offset) &= ~HSW_PWR_WELL_STATE_ENABLED;
+		vgpu_vreg(vgpu, offset) &=
+			~HSW_PWR_WELL_CTL_STATE(HSW_DISP_PW_GLOBAL);
 	return 0;
 }
 
@@ -2242,10 +2252,17 @@ static int init_generic_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(GEN6_RC6p_THRESHOLD, D_ALL);
 	MMIO_D(GEN6_RC6pp_THRESHOLD, D_ALL);
 	MMIO_D(GEN6_PMINTRMSK, D_ALL);
-	MMIO_DH(HSW_PWR_WELL_BIOS, D_BDW, NULL, power_well_ctl_mmio_write);
-	MMIO_DH(HSW_PWR_WELL_DRIVER, D_BDW, NULL, power_well_ctl_mmio_write);
-	MMIO_DH(HSW_PWR_WELL_KVMR, D_BDW, NULL, power_well_ctl_mmio_write);
-	MMIO_DH(HSW_PWR_WELL_DEBUG, D_BDW, NULL, power_well_ctl_mmio_write);
+	/*
+	 * Use an arbitrary power well controlled by the PWR_WELL_CTL
+	 * register.
+	 */
+	MMIO_DH(HSW_PWR_WELL_CTL_BIOS(HSW_DISP_PW_GLOBAL), D_BDW, NULL,
+		power_well_ctl_mmio_write);
+	MMIO_DH(HSW_PWR_WELL_CTL_DRIVER(HSW_DISP_PW_GLOBAL), D_BDW, NULL,
+		power_well_ctl_mmio_write);
+	MMIO_DH(HSW_PWR_WELL_CTL_KVMR, D_BDW, NULL, power_well_ctl_mmio_write);
+	MMIO_DH(HSW_PWR_WELL_CTL_DEBUG(HSW_DISP_PW_GLOBAL), D_BDW, NULL,
+		power_well_ctl_mmio_write);
 	MMIO_DH(HSW_PWR_WELL_CTL5, D_BDW, NULL, power_well_ctl_mmio_write);
 	MMIO_DH(HSW_PWR_WELL_CTL6, D_BDW, NULL, power_well_ctl_mmio_write);
 
@@ -2581,7 +2598,6 @@ static int init_broadwell_mmio_info(struct intel_gvt *gvt)
 	MMIO_F(0x24d0, 48, F_CMD_ACCESS, 0, 0, D_BDW_PLUS,
 		NULL, force_nonpriv_write);
 
-	MMIO_D(0x22040, D_BDW_PLUS);
 	MMIO_D(0x44484, D_BDW_PLUS);
 	MMIO_D(0x4448c, D_BDW_PLUS);
 
@@ -2636,9 +2652,13 @@ static int init_skl_mmio_info(struct intel_gvt *gvt)
 	MMIO_F(_DPD_AUX_CH_CTL, 6 * 4, 0, 0, 0, D_SKL_PLUS, NULL,
 						dp_aux_ch_ctl_mmio_write);
 
-	MMIO_D(HSW_PWR_WELL_BIOS, D_SKL_PLUS);
-	MMIO_DH(HSW_PWR_WELL_DRIVER, D_SKL_PLUS, NULL,
-						skl_power_well_ctl_write);
+	/*
+	 * Use an arbitrary power well controlled by the PWR_WELL_CTL
+	 * register.
+	 */
+	MMIO_D(HSW_PWR_WELL_CTL_BIOS(SKL_DISP_PW_MISC_IO), D_SKL_PLUS);
+	MMIO_DH(HSW_PWR_WELL_CTL_DRIVER(SKL_DISP_PW_MISC_IO), D_SKL_PLUS, NULL,
+		skl_power_well_ctl_write);
 	MMIO_DH(GEN6_PCODE_MAILBOX, D_SKL_PLUS, NULL, mailbox_write);
 
 	MMIO_D(0xa210, D_SKL_PLUS);
@@ -2831,7 +2851,6 @@ static int init_skl_mmio_info(struct intel_gvt *gvt)
 	MMIO_D(0x320f0, D_SKL | D_KBL);
 
 	MMIO_DFH(_REG_VCS2_EXCC, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
-	MMIO_DFH(_REG_VECS_EXCC, D_SKL_PLUS, F_CMD_ACCESS, NULL, NULL);
 	MMIO_D(0x70034, D_SKL_PLUS);
 	MMIO_D(0x71034, D_SKL_PLUS);
 	MMIO_D(0x72034, D_SKL_PLUS);
@@ -2849,10 +2868,7 @@ static int init_skl_mmio_info(struct intel_gvt *gvt)
 		NULL, NULL);
 
 	MMIO_D(0x4ab8, D_KBL);
-	MMIO_D(0x940c, D_SKL_PLUS);
 	MMIO_D(0x2248, D_SKL_PLUS | D_KBL);
-	MMIO_D(0x4ab0, D_SKL | D_KBL);
-	MMIO_D(0x20d4, D_SKL | D_KBL);
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
index fd0c85f9ef3c..83e88c70272a 100644
--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -1170,10 +1170,27 @@ vgpu_id_show(struct device *dev, struct device_attribute *attr,
 	return sprintf(buf, "\n");
 }
 
+static ssize_t
+hw_id_show(struct device *dev, struct device_attribute *attr,
+	   char *buf)
+{
+	struct mdev_device *mdev = mdev_from_dev(dev);
+
+	if (mdev) {
+		struct intel_vgpu *vgpu = (struct intel_vgpu *)
+			mdev_get_drvdata(mdev);
+		return sprintf(buf, "%u\n",
+			       vgpu->shadow_ctx->hw_id);
+	}
+	return sprintf(buf, "\n");
+}
+
 static DEVICE_ATTR_RO(vgpu_id);
+static DEVICE_ATTR_RO(hw_id);
 
 static struct attribute *intel_vgpu_attrs[] = {
 	&dev_attr_vgpu_id.attr,
+	&dev_attr_hw_id.attr,
 	NULL
 };
 
diff --git a/drivers/gpu/drm/i915/gvt/render.c b/drivers/gpu/drm/i915/gvt/render.c
index 504e57c3bc23..2ea542257f03 100644
--- a/drivers/gpu/drm/i915/gvt/render.c
+++ b/drivers/gpu/drm/i915/gvt/render.c
@@ -207,18 +207,16 @@ static void load_mocs(struct intel_vgpu *vgpu, int ring_id)
 
 	offset.reg = regs[ring_id];
 	for (i = 0; i < 64; i++) {
-		gen9_render_mocs[ring_id][i] = I915_READ(offset);
+		gen9_render_mocs[ring_id][i] = I915_READ_FW(offset);
 		I915_WRITE(offset, vgpu_vreg(vgpu, offset));
-		POSTING_READ(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(l3_offset);
-			I915_WRITE(l3_offset, vgpu_vreg(vgpu, l3_offset));
-			POSTING_READ(l3_offset);
+			gen9_render_mocs_L3[i] = I915_READ_FW(l3_offset);
+			I915_WRITE_FW(l3_offset, vgpu_vreg(vgpu, l3_offset));
 			l3_offset.reg += 4;
 		}
 	}
@@ -242,18 +240,16 @@ static void restore_mocs(struct intel_vgpu *vgpu, int ring_id)
 
 	offset.reg = regs[ring_id];
 	for (i = 0; i < 64; i++) {
-		vgpu_vreg(vgpu, offset) = I915_READ(offset);
-		I915_WRITE(offset, gen9_render_mocs[ring_id][i]);
-		POSTING_READ(offset);
+		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(l3_offset);
-			I915_WRITE(l3_offset, gen9_render_mocs_L3[i]);
-			POSTING_READ(l3_offset);
+			vgpu_vreg(vgpu, l3_offset) = I915_READ_FW(l3_offset);
+			I915_WRITE_FW(l3_offset, gen9_render_mocs_L3[i]);
 			l3_offset.reg += 4;
 		}
 	}
@@ -272,6 +268,7 @@ static void switch_mmio_to_vgpu(struct intel_vgpu *vgpu, int ring_id)
 	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)) {
@@ -287,7 +284,7 @@ static void switch_mmio_to_vgpu(struct intel_vgpu *vgpu, int ring_id)
 		if (mmio->ring_id != ring_id)
 			continue;
 
-		mmio->value = I915_READ(mmio->reg);
+		mmio->value = I915_READ_FW(mmio->reg);
 
 		/*
 		 * if it is an inhibit context, load in_context mmio
@@ -304,13 +301,18 @@ static void switch_mmio_to_vgpu(struct intel_vgpu *vgpu, int ring_id)
 		else
 			v = vgpu_vreg(vgpu, mmio->reg);
 
-		I915_WRITE(mmio->reg, v);
-		POSTING_READ(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);
 }
 
@@ -319,6 +321,7 @@ 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;
 
@@ -335,7 +338,7 @@ static void switch_mmio_to_host(struct intel_vgpu *vgpu, int ring_id)
 		if (mmio->ring_id != ring_id)
 			continue;
 
-		vgpu_vreg(vgpu, mmio->reg) = I915_READ(mmio->reg);
+		vgpu_vreg(vgpu, mmio->reg) = I915_READ_FW(mmio->reg);
 
 		if (mmio->mask) {
 			vgpu_vreg(vgpu, mmio->reg) &= ~(mmio->mask << 16);
@@ -346,13 +349,17 @@ static void switch_mmio_to_host(struct intel_vgpu *vgpu, int ring_id)
 		if (mmio->in_context)
 			continue;
 
-		I915_WRITE(mmio->reg, v);
-		POSTING_READ(mmio->reg);
+		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);
 }
 
 /**
@@ -367,12 +374,23 @@ static void switch_mmio_to_host(struct intel_vgpu *vgpu, int ring_id)
 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().
@@ -382,4 +400,6 @@ void intel_gvt_switch_mmio(struct intel_vgpu *pre,
 
 	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/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 22e08eb2d0b7..391800d2067b 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -184,41 +184,52 @@ static int shadow_context_status_change(struct notifier_block *nb,
 	return NOTIFY_OK;
 }
 
-static int dispatch_workload(struct intel_vgpu_workload *workload)
+static void shadow_context_descriptor_update(struct i915_gem_context *ctx,
+		struct intel_engine_cs *engine)
+{
+	struct intel_context *ce = &ctx->engine[engine->id];
+	u64 desc = 0;
+
+	desc = ce->lrc_desc;
+
+	/* Update bits 0-11 of the context descriptor which includes flags
+	 * like GEN8_CTX_* cached in desc_template
+	 */
+	desc &= U64_MAX << 12;
+	desc |= ctx->desc_template & ((1ULL << 12) - 1);
+
+	ce->lrc_desc = desc;
+}
+
+/**
+ * intel_gvt_scan_and_shadow_workload - audit the workload by scanning and
+ * shadow it as well, include ringbuffer,wa_ctx and ctx.
+ * @workload: an abstract entity for each execlist submission.
+ *
+ * This function is called before the workload submitting to i915, to make
+ * sure the content of the workload is valid.
+ */
+int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload)
 {
 	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 drm_i915_gem_request *rq;
 	struct intel_vgpu *vgpu = workload->vgpu;
-	struct intel_ring *ring;
 	int ret;
 
-	gvt_dbg_sched("ring id %d prepare to dispatch workload %p\n",
-		ring_id, workload);
+	lockdep_assert_held(&dev_priv->drm.struct_mutex);
+
+	if (workload->shadowed)
+		return 0;
 
 	shadow_ctx->desc_template &= ~(0x3 << GEN8_CTX_ADDRESSING_MODE_SHIFT);
 	shadow_ctx->desc_template |= workload->ctx_desc.addressing_mode <<
 				    GEN8_CTX_ADDRESSING_MODE_SHIFT;
 
-	mutex_lock(&dev_priv->drm.struct_mutex);
-
-	/* pin shadow context by gvt even the shadow context will be pinned
-	 * when i915 alloc request. That is because gvt will update the guest
-	 * context from shadow context when workload is completed, and at that
-	 * moment, i915 may already unpined the shadow context to make the
-	 * shadow_ctx pages invalid. So gvt need to pin itself. After update
-	 * the guest context, gvt can unpin the shadow_ctx safely.
-	 */
-	ring = engine->context_pin(engine, shadow_ctx);
-	if (IS_ERR(ring)) {
-		ret = PTR_ERR(ring);
-		gvt_vgpu_err("fail to pin shadow context\n");
-		workload->status = ret;
-		mutex_unlock(&dev_priv->drm.struct_mutex);
-		return ret;
-	}
+	if (!test_and_set_bit(ring_id, vgpu->shadow_ctx_desc_updated))
+		shadow_context_descriptor_update(shadow_ctx,
+					dev_priv->engine[ring_id]);
 
 	rq = i915_gem_request_alloc(dev_priv->engine[ring_id], shadow_ctx);
 	if (IS_ERR(rq)) {
@@ -231,7 +242,7 @@ static int dispatch_workload(struct intel_vgpu_workload *workload)
 
 	workload->req = i915_gem_request_get(rq);
 
-	ret = intel_gvt_scan_and_shadow_workload(workload);
+	ret = intel_gvt_scan_and_shadow_ringbuffer(workload);
 	if (ret)
 		goto out;
 
@@ -246,25 +257,61 @@ static int dispatch_workload(struct intel_vgpu_workload *workload)
 	if (ret)
 		goto out;
 
+	workload->shadowed = true;
+
+out:
+	return ret;
+}
+
+static int dispatch_workload(struct intel_vgpu_workload *workload)
+{
+	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 = 0;
+
+	gvt_dbg_sched("ring id %d prepare to dispatch workload %p\n",
+		ring_id, workload);
+
+	mutex_lock(&dev_priv->drm.struct_mutex);
+
+	ret = intel_gvt_scan_and_shadow_workload(workload);
+	if (ret)
+		goto out;
+
 	if (workload->prepare) {
 		ret = workload->prepare(workload);
 		if (ret)
 			goto out;
 	}
 
-	gvt_dbg_sched("ring id %d submit workload to i915 %p\n",
-			ring_id, workload->req);
+	/* pin shadow context by gvt even the shadow context will be pinned
+	 * when i915 alloc request. That is because gvt will update the guest
+	 * context from shadow context when workload is completed, and at that
+	 * moment, i915 may already unpined the shadow context to make the
+	 * shadow_ctx pages invalid. So gvt need to pin itself. After update
+	 * the guest context, gvt can unpin the shadow_ctx safely.
+	 */
+	ring = engine->context_pin(engine, shadow_ctx);
+	if (IS_ERR(ring)) {
+		ret = PTR_ERR(ring);
+		gvt_vgpu_err("fail to pin shadow context\n");
+		goto out;
+	}
 
-	ret = 0;
-	workload->dispatched = true;
 out:
 	if (ret)
 		workload->status = ret;
 
-	if (!IS_ERR_OR_NULL(rq))
-		i915_add_request(rq);
-	else
-		engine->context_unpin(engine, shadow_ctx);
+	if (!IS_ERR_OR_NULL(workload->req)) {
+		gvt_dbg_sched("ring id %d submit workload to i915 %p\n",
+				ring_id, workload->req);
+		i915_add_request(workload->req);
+		workload->dispatched = true;
+	}
 
 	mutex_unlock(&dev_priv->drm.struct_mutex);
 	return ret;
@@ -617,7 +664,7 @@ err:
 
 void intel_vgpu_clean_gvt_context(struct intel_vgpu *vgpu)
 {
-	i915_gem_context_put_unlocked(vgpu->shadow_ctx);
+	i915_gem_context_put(vgpu->shadow_ctx);
 }
 
 int intel_vgpu_init_gvt_context(struct intel_vgpu *vgpu)
@@ -631,5 +678,7 @@ int intel_vgpu_init_gvt_context(struct intel_vgpu *vgpu)
 
 	vgpu->shadow_ctx->engine[RCS].initialised = true;
 
+	bitmap_zero(vgpu->shadow_ctx_desc_updated, I915_NUM_ENGINES);
+
 	return 0;
 }
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h
index 9b6bf51e9b9b..0d431a968a32 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.h
+++ b/drivers/gpu/drm/i915/gvt/scheduler.h
@@ -82,6 +82,7 @@ struct intel_vgpu_workload {
 	struct drm_i915_gem_request *req;
 	/* if this workload has been dispatched to i915? */
 	bool dispatched;
+	bool shadowed;
 	int status;
 
 	struct intel_vgpu_mm *shadow_mm;
diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c
index 3deadcbd5a24..02c61a1ad56a 100644
--- a/drivers/gpu/drm/i915/gvt/vgpu.c
+++ b/drivers/gpu/drm/i915/gvt/vgpu.c
@@ -43,6 +43,7 @@ void populate_pvinfo_page(struct intel_vgpu *vgpu)
 	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_aperture_gmadr_base(vgpu);
 	vgpu_vreg(vgpu, vgtif_reg(avail_rs.mappable_gmadr.size)) =
@@ -504,11 +505,11 @@ void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr,
 	/* full GPU reset or device model level reset */
 	if (engine_mask == ALL_ENGINES || dmlr) {
 
-		intel_vgpu_reset_gtt(vgpu, dmlr);
-
 		/*fence will not be reset during virtual reset */
-		if (dmlr)
+		if (dmlr) {
+			intel_vgpu_reset_gtt(vgpu);
 			intel_vgpu_reset_resource(vgpu);
+		}
 
 		intel_vgpu_reset_mmio(vgpu, dmlr);
 		populate_pvinfo_page(vgpu);