drm/i915/gvt: Limit read hw reg to active vgpu

mmio_read_from_hw() let vgpu could read hw reg, if vgpu's workload
is running on hw, things is good. Otherwise vgpu will get other
vgpu's reg val, it is unsafe.

This patch limit such hw access to active vgpu. If vgpu isn't
running on hw, the reg read of this vgpu will get the last active
val which saved at schedule_out.

v2: ring timestamp is walking continuously even if the ring is idle.
    so read hw directly. (Zhenyu)
Signed-off-by: Xiong Zhang <xiong.y.zhang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
(cherry picked from commit 295764cd)

drivers/gpu/drm/i915/gvt/handlers.c

@@ -1407,11 +1407,29 @@ static int skl_lcpll_write(struct intel_vgpu *vgpu, unsigned int offset,
 static int mmio_read_from_hw(struct intel_vgpu *vgpu,
 		unsigned int offset, void *p_data, unsigned int bytes)
 {
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	struct intel_gvt *gvt = vgpu->gvt;
+	struct drm_i915_private *dev_priv = gvt->dev_priv;
+	int ring_id;
+	u32 ring_base;
+
+	ring_id = intel_gvt_render_mmio_to_ring_id(gvt, offset);
+	/**
+	 * Read HW reg in following case
+	 * a. the offset isn't a ring mmio
+	 * b. the offset's ring is running on hw.
+	 * c. the offset is ring time stamp mmio
+	 */
+	if (ring_id >= 0)
+		ring_base = dev_priv->engine[ring_id]->mmio_base;
+
+	if (ring_id < 0 || vgpu  == gvt->scheduler.engine_owner[ring_id] ||
+	    offset == i915_mmio_reg_offset(RING_TIMESTAMP(ring_base)) ||
+	    offset == i915_mmio_reg_offset(RING_TIMESTAMP_UDW(ring_base))) {
+		mmio_hw_access_pre(dev_priv);
+		vgpu_vreg(vgpu, offset) = I915_READ(_MMIO(offset));
+		mmio_hw_access_post(dev_priv);
+	}
 
-	mmio_hw_access_pre(dev_priv);
-	vgpu_vreg(vgpu, offset) = I915_READ(_MMIO(offset));
-	mmio_hw_access_post(dev_priv);
 	return intel_vgpu_default_mmio_read(vgpu, offset, p_data, bytes);
 }
 

drivers/gpu/drm/i915/gvt/scheduler.c

@@ -131,6 +131,20 @@ static inline bool is_gvt_request(struct drm_i915_gem_request *req)
 	return i915_gem_context_force_single_submission(req->ctx);
 }
 
+static void save_ring_hw_state(struct intel_vgpu *vgpu, int ring_id)
+{
+	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	u32 ring_base = dev_priv->engine[ring_id]->mmio_base;
+	i915_reg_t reg;
+
+	reg = RING_INSTDONE(ring_base);
+	vgpu_vreg(vgpu, i915_mmio_reg_offset(reg)) = I915_READ_FW(reg);
+	reg = RING_ACTHD(ring_base);
+	vgpu_vreg(vgpu, i915_mmio_reg_offset(reg)) = I915_READ_FW(reg);
+	reg = RING_ACTHD_UDW(ring_base);
+	vgpu_vreg(vgpu, i915_mmio_reg_offset(reg)) = I915_READ_FW(reg);
+}
+
 static int shadow_context_status_change(struct notifier_block *nb,
 		unsigned long action, void *data)
 {
@@ -176,6 +190,7 @@ static int shadow_context_status_change(struct notifier_block *nb,
 		break;
 	case INTEL_CONTEXT_SCHEDULE_OUT:
 	case INTEL_CONTEXT_SCHEDULE_PREEMPTED:
+		save_ring_hw_state(workload->vgpu, ring_id);
 		atomic_set(&workload->shadow_ctx_active, 0);
 		break;
 	default: