5 files changed, 36 insertions, 10 deletions

diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
index 35b4ec3f7618..3592d04c33b2 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.c
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -1441,7 +1441,7 @@ static inline int cmd_address_audit(struct parser_exec_state *s,
 	}
 
 	if (index_mode)	{
-		if (guest_gma >= I915_GTT_PAGE_SIZE / sizeof(u64)) {
+		if (guest_gma >= I915_GTT_PAGE_SIZE) {
 			ret = -EFAULT;
 			goto err;
 		}
diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c
index c7103dd2d8d5..d7052ab7908c 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.c
+++ b/drivers/gpu/drm/i915/gvt/gtt.c
@@ -1882,7 +1882,11 @@ struct intel_vgpu_mm *intel_vgpu_create_ppgtt_mm(struct intel_vgpu *vgpu,
 	}
 
 	list_add_tail(&mm->ppgtt_mm.list, &vgpu->gtt.ppgtt_mm_list_head);
+
+	mutex_lock(&gvt->gtt.ppgtt_mm_lock);
 	list_add_tail(&mm->ppgtt_mm.lru_list, &gvt->gtt.ppgtt_mm_lru_list_head);
+	mutex_unlock(&gvt->gtt.ppgtt_mm_lock);
+
 	return mm;
 }
 
@@ -1967,9 +1971,10 @@ int intel_vgpu_pin_mm(struct intel_vgpu_mm *mm)
 		if (ret)
 			return ret;
 
+		mutex_lock(&mm->vgpu->gvt->gtt.ppgtt_mm_lock);
 		list_move_tail(&mm->ppgtt_mm.lru_list,
 			       &mm->vgpu->gvt->gtt.ppgtt_mm_lru_list_head);
-
+		mutex_unlock(&mm->vgpu->gvt->gtt.ppgtt_mm_lock);
 	}
 
 	return 0;
@@ -1980,6 +1985,8 @@ static int reclaim_one_ppgtt_mm(struct intel_gvt *gvt)
 	struct intel_vgpu_mm *mm;
 	struct list_head *pos, *n;
 
+	mutex_lock(&gvt->gtt.ppgtt_mm_lock);
+
 	list_for_each_safe(pos, n, &gvt->gtt.ppgtt_mm_lru_list_head) {
 		mm = container_of(pos, struct intel_vgpu_mm, ppgtt_mm.lru_list);
 
@@ -1987,9 +1994,11 @@ static int reclaim_one_ppgtt_mm(struct intel_gvt *gvt)
 			continue;
 
 		list_del_init(&mm->ppgtt_mm.lru_list);
+		mutex_unlock(&gvt->gtt.ppgtt_mm_lock);
 		invalidate_ppgtt_mm(mm);
 		return 1;
 	}
+	mutex_unlock(&gvt->gtt.ppgtt_mm_lock);
 	return 0;
 }
 
@@ -2659,6 +2668,7 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt)
 		}
 	}
 	INIT_LIST_HEAD(&gvt->gtt.ppgtt_mm_lru_list_head);
+	mutex_init(&gvt->gtt.ppgtt_mm_lock);
 	return 0;
 }
 
@@ -2699,7 +2709,9 @@ void intel_vgpu_invalidate_ppgtt(struct intel_vgpu *vgpu)
 	list_for_each_safe(pos, n, &vgpu->gtt.ppgtt_mm_list_head) {
 		mm = container_of(pos, struct intel_vgpu_mm, ppgtt_mm.list);
 		if (mm->type == INTEL_GVT_MM_PPGTT) {
+			mutex_lock(&vgpu->gvt->gtt.ppgtt_mm_lock);
 			list_del_init(&mm->ppgtt_mm.lru_list);
+			mutex_unlock(&vgpu->gvt->gtt.ppgtt_mm_lock);
 			if (mm->ppgtt_mm.shadowed)
 				invalidate_ppgtt_mm(mm);
 		}
diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h
index d8cb04cc946d..edb610dc5d86 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.h
+++ b/drivers/gpu/drm/i915/gvt/gtt.h
@@ -88,6 +88,7 @@ struct intel_gvt_gtt {
 	void (*mm_free_page_table)(struct intel_vgpu_mm *mm);
 	struct list_head oos_page_use_list_head;
 	struct list_head oos_page_free_list_head;
+	struct mutex ppgtt_mm_lock;
 	struct list_head ppgtt_mm_lru_list_head;
 
 	struct page *scratch_page;
diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c
index 7d84cfb9051a..7902fb162d09 100644
--- a/drivers/gpu/drm/i915/gvt/mmio_context.c
+++ b/drivers/gpu/drm/i915/gvt/mmio_context.c
@@ -132,6 +132,7 @@ static struct engine_mmio gen9_engine_mmio_list[] __cacheline_aligned = {
 
 	{RCS, GEN9_GAMT_ECO_REG_RW_IA, 0x0, false}, /* 0x4ab0 */
 	{RCS, GEN9_CSFE_CHICKEN1_RCS, 0xffff, false}, /* 0x20d4 */
+	{RCS, _MMIO(0x20D8), 0xffff, true}, /* 0x20d8 */
 
 	{RCS, GEN8_GARBCNTL, 0x0, false}, /* 0xb004 */
 	{RCS, GEN7_FF_THREAD_MODE, 0x0, false}, /* 0x20a0 */
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 1bb8f936fdaa..159192c097cc 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -346,7 +346,7 @@ static int set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload,
 	int i = 0;
 
 	if (mm->type != INTEL_GVT_MM_PPGTT || !mm->ppgtt_mm.shadowed)
-		return -1;
+		return -EINVAL;
 
 	if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
 		px_dma(&ppgtt->pml4) = mm->ppgtt_mm.shadow_pdps[0];
@@ -410,12 +410,6 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload)
 	if (workload->shadow)
 		return 0;
 
-	ret = set_context_ppgtt_from_shadow(workload, shadow_ctx);
-	if (ret < 0) {
-		gvt_vgpu_err("workload shadow ppgtt isn't ready\n");
-		return ret;
-	}
-
 	/* 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
@@ -678,6 +672,9 @@ static int dispatch_workload(struct intel_vgpu_workload *workload)
 {
 	struct intel_vgpu *vgpu = workload->vgpu;
 	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	struct intel_vgpu_submission *s = &vgpu->submission;
+	struct i915_gem_context *shadow_ctx = s->shadow_ctx;
+	struct i915_request *rq;
 	int ring_id = workload->ring_id;
 	int ret;
 
@@ -687,6 +684,12 @@ static int dispatch_workload(struct intel_vgpu_workload *workload)
 	mutex_lock(&vgpu->vgpu_lock);
 	mutex_lock(&dev_priv->drm.struct_mutex);
 
+	ret = set_context_ppgtt_from_shadow(workload, shadow_ctx);
+	if (ret < 0) {
+		gvt_vgpu_err("workload shadow ppgtt isn't ready\n");
+		goto err_req;
+	}
+
 	ret = intel_gvt_workload_req_alloc(workload);
 	if (ret)
 		goto err_req;
@@ -703,6 +706,14 @@ static int dispatch_workload(struct intel_vgpu_workload *workload)
 
 	ret = prepare_workload(workload);
 out:
+	if (ret) {
+		/* We might still need to add request with
+		 * clean ctx to retire it properly..
+		 */
+		rq = fetch_and_zero(&workload->req);
+		i915_request_put(rq);
+	}
+
 	if (!IS_ERR_OR_NULL(workload->req)) {
 		gvt_dbg_sched("ring id %d submit workload to i915 %p\n",
 				ring_id, workload->req);
@@ -739,7 +750,8 @@ static struct intel_vgpu_workload *pick_next_workload(
 		goto out;
 	}
 
-	if (list_empty(workload_q_head(scheduler->current_vgpu, ring_id)))
+	if (!scheduler->current_vgpu->active ||
+	    list_empty(workload_q_head(scheduler->current_vgpu, ring_id)))
 		goto out;
 
 	/*