Merge tag 'gvt-next-2017-12-22' of https://github.com/intel/gvt-linux into drm-intel-next-queued

gvt-next-2017-12-22:

- more mmio switch optimization (Weinan)
- cleanup i915_reg_t vs. offset usage (Zhenyu)
- move write protect handler out of mmio handler (Zhenyu)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171222085141.vgewlvvni37dljdt@zhen-hp.sh.intel.com

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

@@ -825,6 +825,21 @@ static int force_nonpriv_reg_handler(struct parser_exec_state *s,
 	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)
 {
@@ -848,6 +863,10 @@ static int cmd_reg_handler(struct parser_exec_state *s,
 		return 0;
 	}
 
+	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;
@@ -1220,13 +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)
@@ -1245,21 +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;
 }

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))
@@ -74,7 +74,7 @@ int pipe_is_enabled(struct intel_vgpu *vgpu, int pipe)
 	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) &&
@@ -169,103 +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_t(vgpu, PIPECONF(PIPE_A)) |= PIPECONF_ENABLE;
 }
 
 static void clean_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num)
@@ -282,7 +284,6 @@ static void clean_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num)
 static int setup_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num,
 				    int type, unsigned int resolution)
 {
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
 	struct intel_vgpu_port *port = intel_vgpu_port(vgpu, port_num);
 
 	if (WARN_ON(resolution >= GVT_EDID_NUM))
@@ -308,7 +309,7 @@ static int setup_virtual_dp_monitor(struct intel_vgpu *vgpu, int port_num,
 	port->type = type;
 
 	emulate_monitor_status_change(vgpu);
-	vgpu_vreg(vgpu, PIPECONF(PIPE_A)) |= PIPECONF_ENABLE;
+
 	return 0;
 }
 
@@ -368,12 +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]);
 	}
 }

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;

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

@@ -147,7 +147,7 @@ static u32 intel_vgpu_get_stride(struct intel_vgpu *vgpu, int pipe,
 {
 	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
 
-	u32 stride_reg = vgpu_vreg(vgpu, DSPSTRIDE(pipe)) & stride_mask;
+	u32 stride_reg = vgpu_vreg_t(vgpu, DSPSTRIDE(pipe)) & stride_mask;
 	u32 stride = stride_reg;
 
 	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
@@ -209,7 +209,7 @@ int intel_vgpu_decode_primary_plane(struct intel_vgpu *vgpu,
 	if (pipe >= I915_MAX_PIPES)
 		return -ENODEV;
 
-	val = vgpu_vreg(vgpu, DSPCNTR(pipe));
+	val = vgpu_vreg_t(vgpu, DSPCNTR(pipe));
 	plane->enabled = !!(val & DISPLAY_PLANE_ENABLE);
 	if (!plane->enabled)
 		return -ENODEV;
@@ -244,7 +244,7 @@ int intel_vgpu_decode_primary_plane(struct intel_vgpu *vgpu,
 
 	plane->hw_format = fmt;
 
-	plane->base = vgpu_vreg(vgpu, DSPSURF(pipe)) & I915_GTT_PAGE_MASK;
+	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);
@@ -263,14 +263,14 @@ int intel_vgpu_decode_primary_plane(struct intel_vgpu *vgpu,
 			(_PRI_PLANE_STRIDE_MASK >> 6) :
 				_PRI_PLANE_STRIDE_MASK, plane->bpp);
 
-	plane->width = (vgpu_vreg(vgpu, PIPESRC(pipe)) & _PIPE_H_SRCSZ_MASK) >>
+	plane->width = (vgpu_vreg_t(vgpu, PIPESRC(pipe)) & _PIPE_H_SRCSZ_MASK) >>
 		_PIPE_H_SRCSZ_SHIFT;
 	plane->width += 1;
-	plane->height = (vgpu_vreg(vgpu, PIPESRC(pipe)) &
+	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(vgpu, DSPTILEOFF(pipe));
+	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) >>
@@ -344,7 +344,7 @@ int intel_vgpu_decode_cursor_plane(struct intel_vgpu *vgpu,
 	if (pipe >= I915_MAX_PIPES)
 		return -ENODEV;
 
-	val = vgpu_vreg(vgpu, CURCNTR(pipe));
+	val = vgpu_vreg_t(vgpu, CURCNTR(pipe));
 	mode = val & CURSOR_MODE;
 	plane->enabled = (mode != CURSOR_MODE_DISABLE);
 	if (!plane->enabled)
@@ -370,7 +370,7 @@ int intel_vgpu_decode_cursor_plane(struct intel_vgpu *vgpu,
 		gvt_dbg_core("alpha_plane=0x%x, alpha_force=0x%x\n",
 			alpha_plane, alpha_force);
 
-	plane->base = vgpu_vreg(vgpu, CURBASE(pipe)) & I915_GTT_PAGE_MASK;
+	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);
@@ -384,7 +384,7 @@ int intel_vgpu_decode_cursor_plane(struct intel_vgpu *vgpu,
 		return  -EINVAL;
 	}
 
-	val = vgpu_vreg(vgpu, CURPOS(pipe));
+	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;
@@ -424,7 +424,7 @@ int intel_vgpu_decode_sprite_plane(struct intel_vgpu *vgpu,
 	if (pipe >= I915_MAX_PIPES)
 		return -ENODEV;
 
-	val = vgpu_vreg(vgpu, SPRCTL(pipe));
+	val = vgpu_vreg_t(vgpu, SPRCTL(pipe));
 	plane->enabled = !!(val & SPRITE_ENABLE);
 	if (!plane->enabled)
 		return -ENODEV;
@@ -475,7 +475,7 @@ int intel_vgpu_decode_sprite_plane(struct intel_vgpu *vgpu,
 
 	plane->drm_format = drm_format;
 
-	plane->base = vgpu_vreg(vgpu, SPRSURF(pipe)) & I915_GTT_PAGE_MASK;
+	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);
@@ -489,10 +489,10 @@ int intel_vgpu_decode_sprite_plane(struct intel_vgpu *vgpu,
 		return  -EINVAL;
 	}
 
-	plane->stride = vgpu_vreg(vgpu, SPRSTRIDE(pipe)) &
+	plane->stride = vgpu_vreg_t(vgpu, SPRSTRIDE(pipe)) &
 				_SPRITE_STRIDE_MASK;
 
-	val = vgpu_vreg(vgpu, SPRSIZE(pipe));
+	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) >>
@@ -500,11 +500,11 @@ int intel_vgpu_decode_sprite_plane(struct intel_vgpu *vgpu,
 	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(vgpu, SPRPOS(pipe));
+	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(vgpu, SPROFFSET(pipe));
+	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) >>

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

@@ -1968,6 +1968,39 @@ 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)
 {
@@ -2244,7 +2277,7 @@ 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)))
@@ -2279,7 +2312,7 @@ 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)))

drivers/gpu/drm/i915/gvt/gtt.h

@@ -308,4 +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_ */

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

@@ -183,6 +183,7 @@ static const struct intel_gvt_ops intel_gvt_ops = {
 	.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,
 };
 
 /**

drivers/gpu/drm/i915/gvt/gvt.h

@@ -412,23 +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)) \
@@ -549,6 +546,8 @@ struct intel_gvt_ops {
 			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);
 };
 
 

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

@@ -1360,8 +1360,8 @@ 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,

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_tracked_guest_page)) {
-		struct intel_vgpu_page_track *t;
-
-		/* 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.
-		 */
-		t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT);
-		if (t) {
-			/* remove write protection to prevent furture traps */
-			intel_vgpu_clean_page_track(vgpu, t);
-			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);
 }
@@ -168,23 +149,6 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa,
 		goto out;
 	}
 
-	if (atomic_read(&vgpu->gtt.n_tracked_guest_page)) {
-		struct intel_vgpu_page_track *t;
-
-		t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT);
-		if (t) {
-			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, t->gfn, pa, *(u32 *)p_data,
-					bytes);
-			}
-			goto out;
-		}
-	}
-
 	offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
 
 	if (WARN_ON(bytes > 8))
@@ -263,23 +227,6 @@ int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, uint64_t pa,
 		goto out;
 	}
 
-	if (atomic_read(&vgpu->gtt.n_tracked_guest_page)) {
-		struct intel_vgpu_page_track *t;
-
-		t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT);
-		if (t) {
-			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);
-			}
-			goto out;
-		}
-	}
-
 	offset = intel_vgpu_gpa_to_mmio_offset(vgpu, pa);
 
 	if (WARN_ON(bytes > 8))
@@ -336,10 +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 {

drivers/gpu/drm/i915/gvt/mmio.h

@@ -76,13 +76,6 @@ int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt,
 	int (*handler)(struct intel_gvt *gvt, u32 offset, void *data),
 	void *data);
 
-
-#define INTEL_GVT_MMIO_OFFSET(reg) ({ \
-	typeof(reg) __reg = reg; \
-	u32 *offset = (u32 *)&__reg; \
-	*offset; \
-})
-
 int intel_vgpu_init_mmio(struct intel_vgpu *vgpu);
 void intel_vgpu_reset_mmio(struct intel_vgpu *vgpu, bool dmlr);
 void intel_vgpu_clean_mmio(struct intel_vgpu *vgpu);

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

@@ -149,8 +149,41 @@ static struct engine_mmio gen9_engine_mmio_list[] __cacheline_aligned = {
 	{ /* Terminated */ }
 };
 
-static u32 gen9_render_mocs[I915_NUM_ENGINES][64];
-static u32 gen9_render_mocs_L3[32];
+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)
 {
@@ -191,17 +224,20 @@ static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id)
 	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;
+		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 load_mocs(struct intel_vgpu *vgpu, int ring_id)
+static void switch_mocs(struct intel_vgpu *pre, struct intel_vgpu *next,
+			int ring_id)
 {
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
+	struct drm_i915_private *dev_priv;
 	i915_reg_t offset, l3_offset;
+	u32 old_v, new_v;
+
 	u32 regs[] = {
 		[RCS] = 0xc800,
 		[VCS] = 0xc900,
@@ -211,54 +247,45 @@ static void load_mocs(struct intel_vgpu *vgpu, int ring_id)
 	};
 	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++) {
-		gen9_render_mocs[ring_id][i] = I915_READ_FW(offset);
-		I915_WRITE_FW(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;
-		}
-	}
-}
+		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];
 
-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 (old_v != new_v)
+			I915_WRITE_FW(offset, new_v);
 
-	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]);
+			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;
 		}
 	}
@@ -266,84 +293,77 @@ static void restore_mocs(struct intel_vgpu *vgpu, int ring_id)
 
 #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)
+/* 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 = vgpu->gvt->dev_priv;
-	struct intel_vgpu_submission *s = &vgpu->submission;
-	u32 *reg_state = s->shadow_ctx->engine[ring_id].lrc_reg_state;
-	u32 ctx_ctrl = reg_state[CTX_CONTEXT_CONTROL_VAL];
+	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 v;
-
-	if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv))
-		load_mocs(vgpu, ring_id);
-
-	mmio = vgpu->gvt->engine_mmio_list;
-	while (i915_mmio_reg_offset((mmio++)->reg)) {
-		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)
-			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);
-
-		trace_render_mmio(vgpu->id, "load",
-				  i915_mmio_reg_offset(mmio->reg),
-				  mmio->value, v);
-	}
-
-	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 engine_mmio *mmio;
-	u32 v;
+	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))
-		restore_mocs(vgpu, ring_id);
+		switch_mocs(pre, next, ring_id);
 
-	mmio = vgpu->gvt->engine_mmio_list;
+	mmio = dev_priv->gvt->engine_mmio_list;
 	while (i915_mmio_reg_offset((mmio++)->reg)) {
 		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);
+		// 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
-			v = mmio->value;
-
-		if (mmio->in_context)
-			continue;
+			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, v);
+		I915_WRITE_FW(mmio->reg, new_v);
 
-		trace_render_mmio(vgpu->id, "restore",
+		trace_render_mmio(pre ? pre->id : 0,
+				  next ? next->id : 0,
+				  "switch",
 				  i915_mmio_reg_offset(mmio->reg),
-				  mmio->value, v);
+				  old_v, new_v);
 	}
+
+	if (next)
+		handle_tlb_pending_event(next, ring_id);
 }
 
 /**
@@ -374,17 +394,7 @@ void intel_gvt_switch_mmio(struct intel_vgpu *pre,
 	 * 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);
-
+	switch_mmio(pre, next, ring_id);
 	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 }
 

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)
 );
 

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

@@ -38,25 +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_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(vgpu, vgtif_reg(vgt_caps)) = VGT_CAPS_FULL_48BIT_PPGTT;
-	vgpu_vreg(vgpu, vgtif_reg(vgt_caps)) |= VGT_CAPS_HWSP_EMULATION;
+	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(vgpu, vgtif_reg(avail_rs.mappable_gmadr.base)) =
+	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",