drm/radeon/kms: add pageflip ioctl support (v3)

This adds support for dri2 pageflipping.

v2: precision updates from Mario Kleiner.
v3: Multihead fixes from Mario Kleiner; missing crtc offset
    add note about update pending bit on pre-avivo chips
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drivers/gpu/drm/radeon/evergreen.c

@@ -40,6 +40,61 @@
 static void evergreen_gpu_init(struct radeon_device *rdev);
 void evergreen_fini(struct radeon_device *rdev);
 
+void evergreen_pre_page_flip(struct radeon_device *rdev, int crtc)
+{
+	struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc];
+	u32 tmp;
+
+	/* make sure flip is at vb rather than hb */
+	tmp = RREG32(EVERGREEN_GRPH_FLIP_CONTROL + radeon_crtc->crtc_offset);
+	tmp &= ~EVERGREEN_GRPH_SURFACE_UPDATE_H_RETRACE_EN;
+	WREG32(EVERGREEN_GRPH_FLIP_CONTROL + radeon_crtc->crtc_offset, tmp);
+
+	/* set pageflip to happen anywhere in vblank interval */
+	WREG32(EVERGREEN_MASTER_UPDATE_MODE + radeon_crtc->crtc_offset, 0);
+
+	/* enable the pflip int */
+	radeon_irq_kms_pflip_irq_get(rdev, crtc);
+}
+
+void evergreen_post_page_flip(struct radeon_device *rdev, int crtc)
+{
+	/* disable the pflip int */
+	radeon_irq_kms_pflip_irq_put(rdev, crtc);
+}
+
+u32 evergreen_page_flip(struct radeon_device *rdev, int crtc_id, u64 crtc_base)
+{
+	struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc_id];
+	u32 tmp = RREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset);
+
+	/* Lock the graphics update lock */
+	tmp |= EVERGREEN_GRPH_UPDATE_LOCK;
+	WREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset, tmp);
+
+	/* update the scanout addresses */
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + radeon_crtc->crtc_offset,
+	       upper_32_bits(crtc_base));
+	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
+	       (u32)crtc_base);
+
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + radeon_crtc->crtc_offset,
+	       upper_32_bits(crtc_base));
+	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
+	       (u32)crtc_base);
+
+	/* Wait for update_pending to go high. */
+	while (!(RREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset) & EVERGREEN_GRPH_SURFACE_UPDATE_PENDING));
+	DRM_DEBUG("Update pending now high. Unlocking vupdate_lock.\n");
+
+	/* Unlock the lock, so double-buffering can take place inside vblank */
+	tmp &= ~EVERGREEN_GRPH_UPDATE_LOCK;
+	WREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset, tmp);
+
+	/* Return current update_pending status: */
+	return RREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset) & EVERGREEN_GRPH_SURFACE_UPDATE_PENDING;
+}
+
 /* get temperature in millidegrees */
 u32 evergreen_get_temp(struct radeon_device *rdev)
 {
@@ -2060,6 +2115,7 @@ int evergreen_irq_set(struct radeon_device *rdev)
 	u32 crtc1 = 0, crtc2 = 0, crtc3 = 0, crtc4 = 0, crtc5 = 0, crtc6 = 0;
 	u32 hpd1, hpd2, hpd3, hpd4, hpd5, hpd6;
 	u32 grbm_int_cntl = 0;
+	u32 grph1 = 0, grph2 = 0, grph3 = 0, grph4 = 0, grph5 = 0, grph6 = 0;
 
 	if (!rdev->irq.installed) {
 		WARN(1, "Can't enable IRQ/MSI because no handler is installed\n");
@@ -2085,27 +2141,33 @@ int evergreen_irq_set(struct radeon_device *rdev)
 		cp_int_cntl |= RB_INT_ENABLE;
 		cp_int_cntl |= TIME_STAMP_INT_ENABLE;
 	}
-	if (rdev->irq.crtc_vblank_int[0]) {
+	if (rdev->irq.crtc_vblank_int[0] ||
+	    rdev->irq.pflip[0]) {
 		DRM_DEBUG("evergreen_irq_set: vblank 0\n");
 		crtc1 |= VBLANK_INT_MASK;
 	}
-	if (rdev->irq.crtc_vblank_int[1]) {
+	if (rdev->irq.crtc_vblank_int[1] ||
+	    rdev->irq.pflip[1]) {
 		DRM_DEBUG("evergreen_irq_set: vblank 1\n");
 		crtc2 |= VBLANK_INT_MASK;
 	}
-	if (rdev->irq.crtc_vblank_int[2]) {
+	if (rdev->irq.crtc_vblank_int[2] ||
+	    rdev->irq.pflip[2]) {
 		DRM_DEBUG("evergreen_irq_set: vblank 2\n");
 		crtc3 |= VBLANK_INT_MASK;
 	}
-	if (rdev->irq.crtc_vblank_int[3]) {
+	if (rdev->irq.crtc_vblank_int[3] ||
+	    rdev->irq.pflip[3]) {
 		DRM_DEBUG("evergreen_irq_set: vblank 3\n");
 		crtc4 |= VBLANK_INT_MASK;
 	}
-	if (rdev->irq.crtc_vblank_int[4]) {
+	if (rdev->irq.crtc_vblank_int[4] ||
+	    rdev->irq.pflip[4]) {
 		DRM_DEBUG("evergreen_irq_set: vblank 4\n");
 		crtc5 |= VBLANK_INT_MASK;
 	}
-	if (rdev->irq.crtc_vblank_int[5]) {
+	if (rdev->irq.crtc_vblank_int[5] ||
+	    rdev->irq.pflip[5]) {
 		DRM_DEBUG("evergreen_irq_set: vblank 5\n");
 		crtc6 |= VBLANK_INT_MASK;
 	}
@@ -2148,6 +2210,13 @@ int evergreen_irq_set(struct radeon_device *rdev)
 	WREG32(INT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, crtc5);
 	WREG32(INT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, crtc6);
 
+	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, grph1);
+	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, grph2);
+	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, grph3);
+	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, grph4);
+	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, grph5);
+	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, grph6);
+
 	WREG32(DC_HPD1_INT_CONTROL, hpd1);
 	WREG32(DC_HPD2_INT_CONTROL, hpd2);
 	WREG32(DC_HPD3_INT_CONTROL, hpd3);
@@ -2158,79 +2227,92 @@ int evergreen_irq_set(struct radeon_device *rdev)
 	return 0;
 }
 
-static inline void evergreen_irq_ack(struct radeon_device *rdev,
-				     u32 *disp_int,
-				     u32 *disp_int_cont,
-				     u32 *disp_int_cont2,
-				     u32 *disp_int_cont3,
-				     u32 *disp_int_cont4,
-				     u32 *disp_int_cont5)
+static inline void evergreen_irq_ack(struct radeon_device *rdev)
 {
 	u32 tmp;
 
-	*disp_int = RREG32(DISP_INTERRUPT_STATUS);
-	*disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE);
-	*disp_int_cont2 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE2);
-	*disp_int_cont3 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE3);
-	*disp_int_cont4 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE4);
-	*disp_int_cont5 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE5);
-
-	if (*disp_int & LB_D1_VBLANK_INTERRUPT)
+	rdev->irq.stat_regs.evergreen.disp_int = RREG32(DISP_INTERRUPT_STATUS);
+	rdev->irq.stat_regs.evergreen.disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE);
+	rdev->irq.stat_regs.evergreen.disp_int_cont2 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE2);
+	rdev->irq.stat_regs.evergreen.disp_int_cont3 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE3);
+	rdev->irq.stat_regs.evergreen.disp_int_cont4 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE4);
+	rdev->irq.stat_regs.evergreen.disp_int_cont5 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE5);
+	rdev->irq.stat_regs.evergreen.d1grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET);
+	rdev->irq.stat_regs.evergreen.d2grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET);
+	rdev->irq.stat_regs.evergreen.d3grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET);
+	rdev->irq.stat_regs.evergreen.d4grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET);
+	rdev->irq.stat_regs.evergreen.d5grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET);
+	rdev->irq.stat_regs.evergreen.d6grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET);
+
+	if (rdev->irq.stat_regs.evergreen.d1grph_int & GRPH_PFLIP_INT_OCCURRED)
+		WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
+	if (rdev->irq.stat_regs.evergreen.d2grph_int & GRPH_PFLIP_INT_OCCURRED)
+		WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
+	if (rdev->irq.stat_regs.evergreen.d3grph_int & GRPH_PFLIP_INT_OCCURRED)
+		WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
+	if (rdev->irq.stat_regs.evergreen.d4grph_int & GRPH_PFLIP_INT_OCCURRED)
+		WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
+	if (rdev->irq.stat_regs.evergreen.d5grph_int & GRPH_PFLIP_INT_OCCURRED)
+		WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
+	if (rdev->irq.stat_regs.evergreen.d6grph_int & GRPH_PFLIP_INT_OCCURRED)
+		WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
+
+	if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VBLANK_INTERRUPT)
 		WREG32(VBLANK_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, VBLANK_ACK);
-	if (*disp_int & LB_D1_VLINE_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VLINE_INTERRUPT)
 		WREG32(VLINE_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, VLINE_ACK);
 
-	if (*disp_int_cont & LB_D2_VBLANK_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VBLANK_INTERRUPT)
 		WREG32(VBLANK_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, VBLANK_ACK);
-	if (*disp_int_cont & LB_D2_VLINE_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VLINE_INTERRUPT)
 		WREG32(VLINE_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, VLINE_ACK);
 
-	if (*disp_int_cont2 & LB_D3_VBLANK_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VBLANK_INTERRUPT)
 		WREG32(VBLANK_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, VBLANK_ACK);
-	if (*disp_int_cont2 & LB_D3_VLINE_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VLINE_INTERRUPT)
 		WREG32(VLINE_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, VLINE_ACK);
 
-	if (*disp_int_cont3 & LB_D4_VBLANK_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VBLANK_INTERRUPT)
 		WREG32(VBLANK_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, VBLANK_ACK);
-	if (*disp_int_cont3 & LB_D4_VLINE_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VLINE_INTERRUPT)
 		WREG32(VLINE_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, VLINE_ACK);
 
-	if (*disp_int_cont4 & LB_D5_VBLANK_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VBLANK_INTERRUPT)
 		WREG32(VBLANK_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, VBLANK_ACK);
-	if (*disp_int_cont4 & LB_D5_VLINE_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VLINE_INTERRUPT)
 		WREG32(VLINE_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, VLINE_ACK);
 
-	if (*disp_int_cont5 & LB_D6_VBLANK_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VBLANK_INTERRUPT)
 		WREG32(VBLANK_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, VBLANK_ACK);
-	if (*disp_int_cont5 & LB_D6_VLINE_INTERRUPT)
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VLINE_INTERRUPT)
 		WREG32(VLINE_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, VLINE_ACK);
 
-	if (*disp_int & DC_HPD1_INTERRUPT) {
+	if (rdev->irq.stat_regs.evergreen.disp_int & DC_HPD1_INTERRUPT) {
 		tmp = RREG32(DC_HPD1_INT_CONTROL);
 		tmp |= DC_HPDx_INT_ACK;
 		WREG32(DC_HPD1_INT_CONTROL, tmp);
 	}
-	if (*disp_int_cont & DC_HPD2_INTERRUPT) {
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont & DC_HPD2_INTERRUPT) {
 		tmp = RREG32(DC_HPD2_INT_CONTROL);
 		tmp |= DC_HPDx_INT_ACK;
 		WREG32(DC_HPD2_INT_CONTROL, tmp);
 	}
-	if (*disp_int_cont2 & DC_HPD3_INTERRUPT) {
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & DC_HPD3_INTERRUPT) {
 		tmp = RREG32(DC_HPD3_INT_CONTROL);
 		tmp |= DC_HPDx_INT_ACK;
 		WREG32(DC_HPD3_INT_CONTROL, tmp);
 	}
-	if (*disp_int_cont3 & DC_HPD4_INTERRUPT) {
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & DC_HPD4_INTERRUPT) {
 		tmp = RREG32(DC_HPD4_INT_CONTROL);
 		tmp |= DC_HPDx_INT_ACK;
 		WREG32(DC_HPD4_INT_CONTROL, tmp);
 	}
-	if (*disp_int_cont4 & DC_HPD5_INTERRUPT) {
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & DC_HPD5_INTERRUPT) {
 		tmp = RREG32(DC_HPD5_INT_CONTROL);
 		tmp |= DC_HPDx_INT_ACK;
 		WREG32(DC_HPD5_INT_CONTROL, tmp);
 	}
-	if (*disp_int_cont5 & DC_HPD6_INTERRUPT) {
+	if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & DC_HPD6_INTERRUPT) {
 		tmp = RREG32(DC_HPD5_INT_CONTROL);
 		tmp |= DC_HPDx_INT_ACK;
 		WREG32(DC_HPD6_INT_CONTROL, tmp);
@@ -2239,14 +2321,10 @@ static inline void evergreen_irq_ack(struct radeon_device *rdev,
 
 void evergreen_irq_disable(struct radeon_device *rdev)
 {
-	u32 disp_int, disp_int_cont, disp_int_cont2;
-	u32 disp_int_cont3, disp_int_cont4, disp_int_cont5;
-
 	r600_disable_interrupts(rdev);
 	/* Wait and acknowledge irq */
 	mdelay(1);
-	evergreen_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2,
-			  &disp_int_cont3, &disp_int_cont4, &disp_int_cont5);
+	evergreen_irq_ack(rdev);
 	evergreen_disable_interrupt_state(rdev);
 }
 
@@ -2286,8 +2364,6 @@ int evergreen_irq_process(struct radeon_device *rdev)
 	u32 rptr = rdev->ih.rptr;
 	u32 src_id, src_data;
 	u32 ring_index;
-	u32 disp_int, disp_int_cont, disp_int_cont2;
-	u32 disp_int_cont3, disp_int_cont4, disp_int_cont5;
 	unsigned long flags;
 	bool queue_hotplug = false;
 
@@ -2308,8 +2384,7 @@ int evergreen_irq_process(struct radeon_device *rdev)
 
 restart_ih:
 	/* display interrupts */
-	evergreen_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2,
-			  &disp_int_cont3, &disp_int_cont4, &disp_int_cont5);
+	evergreen_irq_ack(rdev);
 
 	rdev->ih.wptr = wptr;
 	while (rptr != wptr) {
@@ -2322,17 +2397,21 @@ int evergreen_irq_process(struct radeon_device *rdev)
 		case 1: /* D1 vblank/vline */
 			switch (src_data) {
 			case 0: /* D1 vblank */
-				if (disp_int & LB_D1_VBLANK_INTERRUPT) {
+				if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VBLANK_INTERRUPT) {
+					if (rdev->irq.pflip[0])
+						radeon_crtc_handle_flip(rdev, 0);
+					if (rdev->irq.crtc_vblank_int[0]) {
 						drm_handle_vblank(rdev->ddev, 0);
 						rdev->pm.vblank_sync = true;
 						wake_up(&rdev->irq.vblank_queue);
-					disp_int &= ~LB_D1_VBLANK_INTERRUPT;
+					}
+					rdev->irq.stat_regs.evergreen.disp_int &= ~LB_D1_VBLANK_INTERRUPT;
 					DRM_DEBUG("IH: D1 vblank\n");
 				}
 				break;
 			case 1: /* D1 vline */
-				if (disp_int & LB_D1_VLINE_INTERRUPT) {
-					disp_int &= ~LB_D1_VLINE_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VLINE_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int &= ~LB_D1_VLINE_INTERRUPT;
 					DRM_DEBUG("IH: D1 vline\n");
 				}
 				break;
@@ -2344,17 +2423,21 @@ int evergreen_irq_process(struct radeon_device *rdev)
 		case 2: /* D2 vblank/vline */
 			switch (src_data) {
 			case 0: /* D2 vblank */
-				if (disp_int_cont & LB_D2_VBLANK_INTERRUPT) {
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VBLANK_INTERRUPT) {
+					if (rdev->irq.pflip[1])
+						radeon_crtc_handle_flip(rdev, 1);
+					if (rdev->irq.crtc_vblank_int[1]) {
 						drm_handle_vblank(rdev->ddev, 1);
 						rdev->pm.vblank_sync = true;
 						wake_up(&rdev->irq.vblank_queue);
-					disp_int_cont &= ~LB_D2_VBLANK_INTERRUPT;
+					}
+					rdev->irq.stat_regs.evergreen.disp_int_cont &= ~LB_D2_VBLANK_INTERRUPT;
 					DRM_DEBUG("IH: D2 vblank\n");
 				}
 				break;
 			case 1: /* D2 vline */
-				if (disp_int_cont & LB_D2_VLINE_INTERRUPT) {
-					disp_int_cont &= ~LB_D2_VLINE_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VLINE_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont &= ~LB_D2_VLINE_INTERRUPT;
 					DRM_DEBUG("IH: D2 vline\n");
 				}
 				break;
@@ -2366,17 +2449,21 @@ int evergreen_irq_process(struct radeon_device *rdev)
 		case 3: /* D3 vblank/vline */
 			switch (src_data) {
 			case 0: /* D3 vblank */
-				if (disp_int_cont2 & LB_D3_VBLANK_INTERRUPT) {
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VBLANK_INTERRUPT) {
+					if (rdev->irq.crtc_vblank_int[2]) {
 						drm_handle_vblank(rdev->ddev, 2);
 						rdev->pm.vblank_sync = true;
 						wake_up(&rdev->irq.vblank_queue);
-					disp_int_cont2 &= ~LB_D3_VBLANK_INTERRUPT;
+					}
+					if (rdev->irq.pflip[2])
+						radeon_crtc_handle_flip(rdev, 2);
+					rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~LB_D3_VBLANK_INTERRUPT;
 					DRM_DEBUG("IH: D3 vblank\n");
 				}
 				break;
 			case 1: /* D3 vline */
-				if (disp_int_cont2 & LB_D3_VLINE_INTERRUPT) {
-					disp_int_cont2 &= ~LB_D3_VLINE_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VLINE_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~LB_D3_VLINE_INTERRUPT;
 					DRM_DEBUG("IH: D3 vline\n");
 				}
 				break;
@@ -2388,17 +2475,21 @@ int evergreen_irq_process(struct radeon_device *rdev)
 		case 4: /* D4 vblank/vline */
 			switch (src_data) {
 			case 0: /* D4 vblank */
-				if (disp_int_cont3 & LB_D4_VBLANK_INTERRUPT) {
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VBLANK_INTERRUPT) {
+					if (rdev->irq.crtc_vblank_int[3]) {
 						drm_handle_vblank(rdev->ddev, 3);
 						rdev->pm.vblank_sync = true;
 						wake_up(&rdev->irq.vblank_queue);
-					disp_int_cont3 &= ~LB_D4_VBLANK_INTERRUPT;
+					}
+					if (rdev->irq.pflip[3])
+						radeon_crtc_handle_flip(rdev, 3);
+					rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~LB_D4_VBLANK_INTERRUPT;
 					DRM_DEBUG("IH: D4 vblank\n");
 				}
 				break;
 			case 1: /* D4 vline */
-				if (disp_int_cont3 & LB_D4_VLINE_INTERRUPT) {
-					disp_int_cont3 &= ~LB_D4_VLINE_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VLINE_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~LB_D4_VLINE_INTERRUPT;
 					DRM_DEBUG("IH: D4 vline\n");
 				}
 				break;
@@ -2410,17 +2501,21 @@ int evergreen_irq_process(struct radeon_device *rdev)
 		case 5: /* D5 vblank/vline */
 			switch (src_data) {
 			case 0: /* D5 vblank */
-				if (disp_int_cont4 & LB_D5_VBLANK_INTERRUPT) {
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VBLANK_INTERRUPT) {
+					if (rdev->irq.crtc_vblank_int[4]) {
 						drm_handle_vblank(rdev->ddev, 4);
 						rdev->pm.vblank_sync = true;
 						wake_up(&rdev->irq.vblank_queue);
-					disp_int_cont4 &= ~LB_D5_VBLANK_INTERRUPT;
+					}
+					if (rdev->irq.pflip[4])
+						radeon_crtc_handle_flip(rdev, 4);
+					rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~LB_D5_VBLANK_INTERRUPT;
 					DRM_DEBUG("IH: D5 vblank\n");
 				}
 				break;
 			case 1: /* D5 vline */
-				if (disp_int_cont4 & LB_D5_VLINE_INTERRUPT) {
-					disp_int_cont4 &= ~LB_D5_VLINE_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VLINE_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~LB_D5_VLINE_INTERRUPT;
 					DRM_DEBUG("IH: D5 vline\n");
 				}
 				break;
@@ -2432,17 +2527,21 @@ int evergreen_irq_process(struct radeon_device *rdev)
 		case 6: /* D6 vblank/vline */
 			switch (src_data) {
 			case 0: /* D6 vblank */
-				if (disp_int_cont5 & LB_D6_VBLANK_INTERRUPT) {
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VBLANK_INTERRUPT) {
+					if (rdev->irq.crtc_vblank_int[5]) {
 						drm_handle_vblank(rdev->ddev, 5);
 						rdev->pm.vblank_sync = true;
 						wake_up(&rdev->irq.vblank_queue);
-					disp_int_cont5 &= ~LB_D6_VBLANK_INTERRUPT;
+					}
+					if (rdev->irq.pflip[5])
+						radeon_crtc_handle_flip(rdev, 5);
+					rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~LB_D6_VBLANK_INTERRUPT;
 					DRM_DEBUG("IH: D6 vblank\n");
 				}
 				break;
 			case 1: /* D6 vline */
-				if (disp_int_cont5 & LB_D6_VLINE_INTERRUPT) {
-					disp_int_cont5 &= ~LB_D6_VLINE_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VLINE_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~LB_D6_VLINE_INTERRUPT;
 					DRM_DEBUG("IH: D6 vline\n");
 				}
 				break;
@@ -2454,43 +2553,43 @@ int evergreen_irq_process(struct radeon_device *rdev)
 		case 42: /* HPD hotplug */
 			switch (src_data) {
 			case 0:
-				if (disp_int & DC_HPD1_INTERRUPT) {
-					disp_int &= ~DC_HPD1_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int & DC_HPD1_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int &= ~DC_HPD1_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD1\n");
 				}
 				break;
 			case 1:
-				if (disp_int_cont & DC_HPD2_INTERRUPT) {
-					disp_int_cont &= ~DC_HPD2_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont & DC_HPD2_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont &= ~DC_HPD2_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD2\n");
 				}
 				break;
 			case 2:
-				if (disp_int_cont2 & DC_HPD3_INTERRUPT) {
-					disp_int_cont2 &= ~DC_HPD3_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & DC_HPD3_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~DC_HPD3_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD3\n");
 				}
 				break;
 			case 3:
-				if (disp_int_cont3 & DC_HPD4_INTERRUPT) {
-					disp_int_cont3 &= ~DC_HPD4_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & DC_HPD4_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~DC_HPD4_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD4\n");
 				}
 				break;
 			case 4:
-				if (disp_int_cont4 & DC_HPD5_INTERRUPT) {
-					disp_int_cont4 &= ~DC_HPD5_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & DC_HPD5_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~DC_HPD5_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD5\n");
 				}
 				break;
 			case 5:
-				if (disp_int_cont5 & DC_HPD6_INTERRUPT) {
-					disp_int_cont5 &= ~DC_HPD6_INTERRUPT;
+				if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & DC_HPD6_INTERRUPT) {
+					rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~DC_HPD6_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD6\n");
 				}

drivers/gpu/drm/radeon/evergreen_reg.h

@@ -105,6 +105,11 @@
 #define EVERGREEN_GRPH_Y_START                          0x6830
 #define EVERGREEN_GRPH_X_END                            0x6834
 #define EVERGREEN_GRPH_Y_END                            0x6838
+#define EVERGREEN_GRPH_UPDATE                           0x6844
+#       define EVERGREEN_GRPH_SURFACE_UPDATE_PENDING    (1 << 2)
+#       define EVERGREEN_GRPH_UPDATE_LOCK               (1 << 16)
+#define EVERGREEN_GRPH_FLIP_CONTROL                     0x6848
+#       define EVERGREEN_GRPH_SURFACE_UPDATE_H_RETRACE_EN (1 << 0)
 
 /* CUR blocks at 0x6998, 0x7598, 0x10198, 0x10d98, 0x11998, 0x12598 */
 #define EVERGREEN_CUR_CONTROL                           0x6998
@@ -178,6 +183,7 @@
 #       define EVERGREEN_CRTC_DISP_READ_REQUEST_DISABLE (1 << 24)
 #define EVERGREEN_CRTC_STATUS                           0x6e8c
 #define EVERGREEN_CRTC_STATUS_POSITION                  0x6e90
+#define EVERGREEN_MASTER_UPDATE_MODE                    0x6ef8
 #define EVERGREEN_CRTC_UPDATE_LOCK                      0x6ed4
 
 #define EVERGREEN_DC_GPIO_HPD_MASK                      0x64b0

drivers/gpu/drm/radeon/r100.c

@@ -68,6 +68,54 @@ MODULE_FIRMWARE(FIRMWARE_R520);
  * r100,rv100,rs100,rv200,rs200,r200,rv250,rs300,rv280
  */
 
+void r100_pre_page_flip(struct radeon_device *rdev, int crtc)
+{
+	struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc];
+	u32 tmp;
+
+	/* make sure flip is at vb rather than hb */
+	tmp = RREG32(RADEON_CRTC_OFFSET_CNTL + radeon_crtc->crtc_offset);
+	tmp &= ~RADEON_CRTC_OFFSET_FLIP_CNTL;
+	WREG32(RADEON_CRTC_OFFSET_CNTL + radeon_crtc->crtc_offset, tmp);
+
+	/* set pageflip to happen as late as possible in the vblank interval.
+	 * same field for crtc1/2
+	 */
+	tmp = RREG32(RADEON_CRTC_GEN_CNTL);
+	tmp &= ~RADEON_CRTC_VSTAT_MODE_MASK;
+	WREG32(RADEON_CRTC_GEN_CNTL, tmp);
+
+	/* enable the pflip int */
+	radeon_irq_kms_pflip_irq_get(rdev, crtc);
+}
+
+void r100_post_page_flip(struct radeon_device *rdev, int crtc)
+{
+	/* disable the pflip int */
+	radeon_irq_kms_pflip_irq_put(rdev, crtc);
+}
+
+u32 r100_page_flip(struct radeon_device *rdev, int crtc_id, u64 crtc_base)
+{
+	struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc_id];
+	u32 tmp = ((u32)crtc_base) | RADEON_CRTC_OFFSET__OFFSET_LOCK;
+
+	/* Lock the graphics update lock */
+	/* update the scanout addresses */
+	WREG32(RADEON_CRTC_OFFSET + radeon_crtc->crtc_offset, tmp);
+
+	/* Note: We don't wait for update_pending to assert, as this never
+	 * happens for some reason on R1xx - R4xx. Adds a bit of imprecision.
+	 */
+
+	/* Unlock the lock, so double-buffering can take place inside vblank */
+	tmp &= ~RADEON_CRTC_OFFSET__OFFSET_LOCK;
+	WREG32(RADEON_CRTC_OFFSET + radeon_crtc->crtc_offset, tmp);
+
+	/* Return current update_pending status: */
+	return RREG32(RADEON_CRTC_OFFSET + radeon_crtc->crtc_offset) & RADEON_CRTC_OFFSET__GUI_TRIG_OFFSET;
+}
+
 void r100_pm_get_dynpm_state(struct radeon_device *rdev)
 {
 	int i;
@@ -526,10 +574,12 @@ int r100_irq_set(struct radeon_device *rdev)
 	if (rdev->irq.gui_idle) {
 		tmp |= RADEON_GUI_IDLE_MASK;
 	}
-	if (rdev->irq.crtc_vblank_int[0]) {
+	if (rdev->irq.crtc_vblank_int[0] ||
+	    rdev->irq.pflip[0]) {
 		tmp |= RADEON_CRTC_VBLANK_MASK;
 	}
-	if (rdev->irq.crtc_vblank_int[1]) {
+	if (rdev->irq.crtc_vblank_int[1] ||
+	    rdev->irq.pflip[1]) {
 		tmp |= RADEON_CRTC2_VBLANK_MASK;
 	}
 	if (rdev->irq.hpd[0]) {
@@ -600,15 +650,23 @@ int r100_irq_process(struct radeon_device *rdev)
 		}
 		/* Vertical blank interrupts */
 		if (status & RADEON_CRTC_VBLANK_STAT) {
+			if (rdev->irq.pflip[0])
+				radeon_crtc_handle_flip(rdev, 0);
+			if (rdev->irq.crtc_vblank_int[0]) {
 				drm_handle_vblank(rdev->ddev, 0);
 				rdev->pm.vblank_sync = true;
 				wake_up(&rdev->irq.vblank_queue);
 			}
+		}
 		if (status & RADEON_CRTC2_VBLANK_STAT) {
+			if (rdev->irq.pflip[1])
+				radeon_crtc_handle_flip(rdev, 1);
+			if (rdev->irq.crtc_vblank_int[1]) {
 				drm_handle_vblank(rdev->ddev, 1);
 				rdev->pm.vblank_sync = true;
 				wake_up(&rdev->irq.vblank_queue);
 			}
+		}
 		if (status & RADEON_FP_DETECT_STAT) {
 			queue_hotplug = true;
 			DRM_DEBUG("HPD1\n");

drivers/gpu/drm/radeon/r500_reg.h

@@ -355,6 +355,8 @@
 #define AVIVO_D1CRTC_FRAME_COUNT                                0x60a4
 #define AVIVO_D1CRTC_STEREO_CONTROL                             0x60c4
 
+#define AVIVO_D1MODE_MASTER_UPDATE_MODE                         0x60e4
+
 /* master controls */
 #define AVIVO_DC_CRTC_MASTER_EN                                 0x60f8
 #define AVIVO_DC_CRTC_TV_CONTROL                                0x60fc
@@ -409,8 +411,10 @@
 #define AVIVO_D1GRPH_X_END                                      0x6134
 #define AVIVO_D1GRPH_Y_END                                      0x6138
 #define AVIVO_D1GRPH_UPDATE                                     0x6144
+#       define AVIVO_D1GRPH_SURFACE_UPDATE_PENDING              (1 << 2)
 #       define AVIVO_D1GRPH_UPDATE_LOCK                         (1 << 16)
 #define AVIVO_D1GRPH_FLIP_CONTROL                               0x6148
+#       define AVIVO_D1GRPH_SURFACE_UPDATE_H_RETRACE_EN         (1 << 0)
 
 #define AVIVO_D1CUR_CONTROL                     0x6400
 #       define AVIVO_D1CURSOR_EN                (1 << 0)

drivers/gpu/drm/radeon/r600.c

@@ -2863,6 +2863,8 @@ static void r600_disable_interrupt_state(struct radeon_device *rdev)
 	WREG32(CP_INT_CNTL, CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE);
 	WREG32(GRBM_INT_CNTL, 0);
 	WREG32(DxMODE_INT_MASK, 0);
+	WREG32(D1GRPH_INTERRUPT_CONTROL, 0);
+	WREG32(D2GRPH_INTERRUPT_CONTROL, 0);
 	if (ASIC_IS_DCE3(rdev)) {
 		WREG32(DCE3_DACA_AUTODETECT_INT_CONTROL, 0);
 		WREG32(DCE3_DACB_AUTODETECT_INT_CONTROL, 0);
@@ -2987,6 +2989,7 @@ int r600_irq_set(struct radeon_device *rdev)
 	u32 hpd1, hpd2, hpd3, hpd4 = 0, hpd5 = 0, hpd6 = 0;
 	u32 grbm_int_cntl = 0;
 	u32 hdmi1, hdmi2;
+	u32 d1grph = 0, d2grph = 0;
 
 	if (!rdev->irq.installed) {
 		WARN(1, "Can't enable IRQ/MSI because no handler is installed\n");
@@ -3023,11 +3026,13 @@ int r600_irq_set(struct radeon_device *rdev)
 		cp_int_cntl |= RB_INT_ENABLE;
 		cp_int_cntl |= TIME_STAMP_INT_ENABLE;
 	}
-	if (rdev->irq.crtc_vblank_int[0]) {
+	if (rdev->irq.crtc_vblank_int[0] ||
+	    rdev->irq.pflip[0]) {
 		DRM_DEBUG("r600_irq_set: vblank 0\n");
 		mode_int |= D1MODE_VBLANK_INT_MASK;
 	}
-	if (rdev->irq.crtc_vblank_int[1]) {
+	if (rdev->irq.crtc_vblank_int[1] ||
+	    rdev->irq.pflip[1]) {
 		DRM_DEBUG("r600_irq_set: vblank 1\n");
 		mode_int |= D2MODE_VBLANK_INT_MASK;
 	}
@@ -3070,6 +3075,8 @@ int r600_irq_set(struct radeon_device *rdev)
 
 	WREG32(CP_INT_CNTL, cp_int_cntl);
 	WREG32(DxMODE_INT_MASK, mode_int);
+	WREG32(D1GRPH_INTERRUPT_CONTROL, d1grph);
+	WREG32(D2GRPH_INTERRUPT_CONTROL, d2grph);
 	WREG32(GRBM_INT_CNTL, grbm_int_cntl);
 	WREG32(R600_HDMI_BLOCK1 + R600_HDMI_CNTL, hdmi1);
 	if (ASIC_IS_DCE3(rdev)) {
@@ -3092,32 +3099,35 @@ int r600_irq_set(struct radeon_device *rdev)
 	return 0;
 }
 
-static inline void r600_irq_ack(struct radeon_device *rdev,
-				u32 *disp_int,
-				u32 *disp_int_cont,
-				u32 *disp_int_cont2)
+static inline void r600_irq_ack(struct radeon_device *rdev)
 {
 	u32 tmp;
 
 	if (ASIC_IS_DCE3(rdev)) {
-		*disp_int = RREG32(DCE3_DISP_INTERRUPT_STATUS);
-		*disp_int_cont = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE);
-		*disp_int_cont2 = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE2);
+		rdev->irq.stat_regs.r600.disp_int = RREG32(DCE3_DISP_INTERRUPT_STATUS);
+		rdev->irq.stat_regs.r600.disp_int_cont = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE);
+		rdev->irq.stat_regs.r600.disp_int_cont2 = RREG32(DCE3_DISP_INTERRUPT_STATUS_CONTINUE2);
 	} else {
-		*disp_int = RREG32(DISP_INTERRUPT_STATUS);
-		*disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE);
-		*disp_int_cont2 = 0;
-	}
-
-	if (*disp_int & LB_D1_VBLANK_INTERRUPT)
+		rdev->irq.stat_regs.r600.disp_int = RREG32(DISP_INTERRUPT_STATUS);
+		rdev->irq.stat_regs.r600.disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE);
+		rdev->irq.stat_regs.r600.disp_int_cont2 = 0;
+	}
+	rdev->irq.stat_regs.r600.d1grph_int = RREG32(D1GRPH_INTERRUPT_STATUS);
+	rdev->irq.stat_regs.r600.d2grph_int = RREG32(D2GRPH_INTERRUPT_STATUS);
+
+	if (rdev->irq.stat_regs.r600.d1grph_int & DxGRPH_PFLIP_INT_OCCURRED)
+		WREG32(D1GRPH_INTERRUPT_STATUS, DxGRPH_PFLIP_INT_CLEAR);
+	if (rdev->irq.stat_regs.r600.d2grph_int & DxGRPH_PFLIP_INT_OCCURRED)
+		WREG32(D2GRPH_INTERRUPT_STATUS, DxGRPH_PFLIP_INT_CLEAR);
+	if (rdev->irq.stat_regs.r600.disp_int & LB_D1_VBLANK_INTERRUPT)
 		WREG32(D1MODE_VBLANK_STATUS, DxMODE_VBLANK_ACK);
-	if (*disp_int & LB_D1_VLINE_INTERRUPT)
+	if (rdev->irq.stat_regs.r600.disp_int & LB_D1_VLINE_INTERRUPT)
 		WREG32(D1MODE_VLINE_STATUS, DxMODE_VLINE_ACK);
-	if (*disp_int & LB_D2_VBLANK_INTERRUPT)
+	if (rdev->irq.stat_regs.r600.disp_int & LB_D2_VBLANK_INTERRUPT)
 		WREG32(D2MODE_VBLANK_STATUS, DxMODE_VBLANK_ACK);
-	if (*disp_int & LB_D2_VLINE_INTERRUPT)
+	if (rdev->irq.stat_regs.r600.disp_int & LB_D2_VLINE_INTERRUPT)
 		WREG32(D2MODE_VLINE_STATUS, DxMODE_VLINE_ACK);
-	if (*disp_int & DC_HPD1_INTERRUPT) {
+	if (rdev->irq.stat_regs.r600.disp_int & DC_HPD1_INTERRUPT) {
 		if (ASIC_IS_DCE3(rdev)) {
 			tmp = RREG32(DC_HPD1_INT_CONTROL);
 			tmp |= DC_HPDx_INT_ACK;
@@ -3128,7 +3138,7 @@ static inline void r600_irq_ack(struct radeon_device *rdev,
 			WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp);
 		}
 	}
-	if (*disp_int & DC_HPD2_INTERRUPT) {
+	if (rdev->irq.stat_regs.r600.disp_int & DC_HPD2_INTERRUPT) {
 		if (ASIC_IS_DCE3(rdev)) {
 			tmp = RREG32(DC_HPD2_INT_CONTROL);
 			tmp |= DC_HPDx_INT_ACK;
@@ -3139,7 +3149,7 @@ static inline void r600_irq_ack(struct radeon_device *rdev,
 			WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp);
 		}
 	}
-	if (*disp_int_cont & DC_HPD3_INTERRUPT) {
+	if (rdev->irq.stat_regs.r600.disp_int_cont & DC_HPD3_INTERRUPT) {
 		if (ASIC_IS_DCE3(rdev)) {
 			tmp = RREG32(DC_HPD3_INT_CONTROL);
 			tmp |= DC_HPDx_INT_ACK;
@@ -3150,18 +3160,18 @@ static inline void r600_irq_ack(struct radeon_device *rdev,
 			WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp);
 		}
 	}
-	if (*disp_int_cont & DC_HPD4_INTERRUPT) {
+	if (rdev->irq.stat_regs.r600.disp_int_cont & DC_HPD4_INTERRUPT) {
 		tmp = RREG32(DC_HPD4_INT_CONTROL);
 		tmp |= DC_HPDx_INT_ACK;
 		WREG32(DC_HPD4_INT_CONTROL, tmp);
 	}
 	if (ASIC_IS_DCE32(rdev)) {
-		if (*disp_int_cont2 & DC_HPD5_INTERRUPT) {
+		if (rdev->irq.stat_regs.r600.disp_int_cont2 & DC_HPD5_INTERRUPT) {
 			tmp = RREG32(DC_HPD5_INT_CONTROL);
 			tmp |= DC_HPDx_INT_ACK;
 			WREG32(DC_HPD5_INT_CONTROL, tmp);
 		}
-		if (*disp_int_cont2 & DC_HPD6_INTERRUPT) {
+		if (rdev->irq.stat_regs.r600.disp_int_cont2 & DC_HPD6_INTERRUPT) {
 			tmp = RREG32(DC_HPD5_INT_CONTROL);
 			tmp |= DC_HPDx_INT_ACK;
 			WREG32(DC_HPD6_INT_CONTROL, tmp);
@@ -3183,12 +3193,10 @@ static inline void r600_irq_ack(struct radeon_device *rdev,
 
 void r600_irq_disable(struct radeon_device *rdev)
 {
-	u32 disp_int, disp_int_cont, disp_int_cont2;
-
 	r600_disable_interrupts(rdev);
 	/* Wait and acknowledge irq */
 	mdelay(1);
-	r600_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2);
+	r600_irq_ack(rdev);
 	r600_disable_interrupt_state(rdev);
 }
 
@@ -3251,7 +3259,7 @@ int r600_irq_process(struct radeon_device *rdev)
 	u32 wptr = r600_get_ih_wptr(rdev);
 	u32 rptr = rdev->ih.rptr;
 	u32 src_id, src_data;
-	u32 ring_index, disp_int, disp_int_cont, disp_int_cont2;
+	u32 ring_index;
 	unsigned long flags;
 	bool queue_hotplug = false;
 
@@ -3272,7 +3280,7 @@ int r600_irq_process(struct radeon_device *rdev)
 
 restart_ih:
 	/* display interrupts */
-	r600_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2);
+	r600_irq_ack(rdev);
 
 	rdev->ih.wptr = wptr;
 	while (rptr != wptr) {
@@ -3285,17 +3293,21 @@ int r600_irq_process(struct radeon_device *rdev)
 		case 1: /* D1 vblank/vline */
 			switch (src_data) {
 			case 0: /* D1 vblank */
-				if (disp_int & LB_D1_VBLANK_INTERRUPT) {
+				if (rdev->irq.stat_regs.r600.disp_int & LB_D1_VBLANK_INTERRUPT) {
+					if (rdev->irq.pflip[0])
+						radeon_crtc_handle_flip(rdev, 0);
+					if (rdev->irq.crtc_vblank_int[0]) {
 						drm_handle_vblank(rdev->ddev, 0);
 						rdev->pm.vblank_sync = true;
 						wake_up(&rdev->irq.vblank_queue);
-					disp_int &= ~LB_D1_VBLANK_INTERRUPT;
+					}
+					rdev->irq.stat_regs.r600.disp_int &= ~LB_D1_VBLANK_INTERRUPT;
 					DRM_DEBUG("IH: D1 vblank\n");
 				}
 				break;
 			case 1: /* D1 vline */
-				if (disp_int & LB_D1_VLINE_INTERRUPT) {
-					disp_int &= ~LB_D1_VLINE_INTERRUPT;
+				if (rdev->irq.stat_regs.r600.disp_int & LB_D1_VLINE_INTERRUPT) {
+					rdev->irq.stat_regs.r600.disp_int &= ~LB_D1_VLINE_INTERRUPT;
 					DRM_DEBUG("IH: D1 vline\n");
 				}
 				break;
@@ -3307,17 +3319,21 @@ int r600_irq_process(struct radeon_device *rdev)
 		case 5: /* D2 vblank/vline */
 			switch (src_data) {
 			case 0: /* D2 vblank */
-				if (disp_int & LB_D2_VBLANK_INTERRUPT) {
+				if (rdev->irq.stat_regs.r600.disp_int & LB_D2_VBLANK_INTERRUPT) {
+					if (rdev->irq.pflip[1])
+						radeon_crtc_handle_flip(rdev, 1);
+					if (rdev->irq.crtc_vblank_int[1]) {
 						drm_handle_vblank(rdev->ddev, 1);
 						rdev->pm.vblank_sync = true;
 						wake_up(&rdev->irq.vblank_queue);
-					disp_int &= ~LB_D2_VBLANK_INTERRUPT;
+					}
+					rdev->irq.stat_regs.r600.disp_int &= ~LB_D2_VBLANK_INTERRUPT;
 					DRM_DEBUG("IH: D2 vblank\n");
 				}
 				break;
 			case 1: /* D1 vline */
-				if (disp_int & LB_D2_VLINE_INTERRUPT) {
-					disp_int &= ~LB_D2_VLINE_INTERRUPT;
+				if (rdev->irq.stat_regs.r600.disp_int & LB_D2_VLINE_INTERRUPT) {
+					rdev->irq.stat_regs.r600.disp_int &= ~LB_D2_VLINE_INTERRUPT;
 					DRM_DEBUG("IH: D2 vline\n");
 				}
 				break;
@@ -3329,43 +3345,43 @@ int r600_irq_process(struct radeon_device *rdev)
 		case 19: /* HPD/DAC hotplug */
 			switch (src_data) {
 			case 0:
-				if (disp_int & DC_HPD1_INTERRUPT) {
-					disp_int &= ~DC_HPD1_INTERRUPT;
+				if (rdev->irq.stat_regs.r600.disp_int & DC_HPD1_INTERRUPT) {
+					rdev->irq.stat_regs.r600.disp_int &= ~DC_HPD1_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD1\n");
 				}
 				break;
 			case 1:
-				if (disp_int & DC_HPD2_INTERRUPT) {
-					disp_int &= ~DC_HPD2_INTERRUPT;
+				if (rdev->irq.stat_regs.r600.disp_int & DC_HPD2_INTERRUPT) {
+					rdev->irq.stat_regs.r600.disp_int &= ~DC_HPD2_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD2\n");
 				}
 				break;
 			case 4:
-				if (disp_int_cont & DC_HPD3_INTERRUPT) {
-					disp_int_cont &= ~DC_HPD3_INTERRUPT;
+				if (rdev->irq.stat_regs.r600.disp_int_cont & DC_HPD3_INTERRUPT) {
+					rdev->irq.stat_regs.r600.disp_int_cont &= ~DC_HPD3_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD3\n");
 				}
 				break;
 			case 5:
-				if (disp_int_cont & DC_HPD4_INTERRUPT) {
-					disp_int_cont &= ~DC_HPD4_INTERRUPT;
+				if (rdev->irq.stat_regs.r600.disp_int_cont & DC_HPD4_INTERRUPT) {
+					rdev->irq.stat_regs.r600.disp_int_cont &= ~DC_HPD4_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD4\n");
 				}
 				break;
 			case 10:
-				if (disp_int_cont2 & DC_HPD5_INTERRUPT) {
-					disp_int_cont2 &= ~DC_HPD5_INTERRUPT;
+				if (rdev->irq.stat_regs.r600.disp_int_cont2 & DC_HPD5_INTERRUPT) {
+					rdev->irq.stat_regs.r600.disp_int_cont2 &= ~DC_HPD5_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD5\n");
 				}
 				break;
 			case 12:
-				if (disp_int_cont2 & DC_HPD6_INTERRUPT) {
-					disp_int_cont2 &= ~DC_HPD6_INTERRUPT;
+				if (rdev->irq.stat_regs.r600.disp_int_cont2 & DC_HPD6_INTERRUPT) {
+					rdev->irq.stat_regs.r600.disp_int_cont2 &= ~DC_HPD6_INTERRUPT;
 					queue_hotplug = true;
 					DRM_DEBUG("IH: HPD6\n");
 				}

drivers/gpu/drm/radeon/r600d.h

@@ -728,6 +728,15 @@
 /* DCE 3.2 */
 #       define DC_HPDx_EN                                 (1 << 28)
 
+#define D1GRPH_INTERRUPT_STATUS                           0x6158
+#define D2GRPH_INTERRUPT_STATUS                           0x6958
+#       define DxGRPH_PFLIP_INT_OCCURRED                  (1 << 0)
+#       define DxGRPH_PFLIP_INT_CLEAR                     (1 << 8)
+#define D1GRPH_INTERRUPT_CONTROL                          0x615c
+#define D2GRPH_INTERRUPT_CONTROL                          0x695c
+#       define DxGRPH_PFLIP_INT_MASK                      (1 << 0)
+#       define DxGRPH_PFLIP_INT_TYPE                      (1 << 8)
+
 /*
  * PM4
  */

drivers/gpu/drm/radeon/radeon.h

@@ -377,11 +377,56 @@ void radeon_scratch_free(struct radeon_device *rdev, uint32_t reg);
 /*
  * IRQS.
  */
+
+struct radeon_unpin_work {
+	struct work_struct work;
+	struct radeon_device *rdev;
+	int crtc_id;
+	struct radeon_fence *fence;
+	struct drm_pending_vblank_event *event;
+	struct radeon_bo *old_rbo;
+	u64 new_crtc_base;
+};
+
+struct r500_irq_stat_regs {
+	u32 disp_int;
+};
+
+struct r600_irq_stat_regs {
+	u32 disp_int;
+	u32 disp_int_cont;
+	u32 disp_int_cont2;
+	u32 d1grph_int;
+	u32 d2grph_int;
+};
+
+struct evergreen_irq_stat_regs {
+	u32 disp_int;
+	u32 disp_int_cont;
+	u32 disp_int_cont2;
+	u32 disp_int_cont3;
+	u32 disp_int_cont4;
+	u32 disp_int_cont5;
+	u32 d1grph_int;
+	u32 d2grph_int;
+	u32 d3grph_int;
+	u32 d4grph_int;
+	u32 d5grph_int;
+	u32 d6grph_int;
+};
+
+union radeon_irq_stat_regs {
+	struct r500_irq_stat_regs r500;
+	struct r600_irq_stat_regs r600;
+	struct evergreen_irq_stat_regs evergreen;
+};
+
 struct radeon_irq {
 	bool		installed;
 	bool		sw_int;
 	/* FIXME: use a define max crtc rather than hardcode it */
 	bool		crtc_vblank_int[6];
+	bool		pflip[6];
 	wait_queue_head_t	vblank_queue;
 	/* FIXME: use defines for max hpd/dacs */
 	bool            hpd[6];
@@ -392,12 +437,17 @@ struct radeon_irq {
 	bool		hdmi[2];
 	spinlock_t sw_lock;
 	int sw_refcount;
+	union radeon_irq_stat_regs stat_regs;
+	spinlock_t pflip_lock[6];
+	int pflip_refcount[6];
 };
 
 int radeon_irq_kms_init(struct radeon_device *rdev);
 void radeon_irq_kms_fini(struct radeon_device *rdev);
 void radeon_irq_kms_sw_irq_get(struct radeon_device *rdev);
 void radeon_irq_kms_sw_irq_put(struct radeon_device *rdev);
+void radeon_irq_kms_pflip_irq_get(struct radeon_device *rdev, int crtc);
+void radeon_irq_kms_pflip_irq_put(struct radeon_device *rdev, int crtc);
 
 /*
  * CP & ring.
@@ -881,6 +931,10 @@ struct radeon_asic {
 	void (*pm_finish)(struct radeon_device *rdev);
 	void (*pm_init_profile)(struct radeon_device *rdev);
 	void (*pm_get_dynpm_state)(struct radeon_device *rdev);
+	/* pageflipping */
+	void (*pre_page_flip)(struct radeon_device *rdev, int crtc);
+	u32 (*page_flip)(struct radeon_device *rdev, int crtc, u64 crtc_base);
+	void (*post_page_flip)(struct radeon_device *rdev, int crtc);
 };
 
 /*
@@ -1344,6 +1398,9 @@ static inline void radeon_ring_write(struct radeon_device *rdev, uint32_t v)
 #define radeon_pm_finish(rdev) (rdev)->asic->pm_finish((rdev))
 #define radeon_pm_init_profile(rdev) (rdev)->asic->pm_init_profile((rdev))
 #define radeon_pm_get_dynpm_state(rdev) (rdev)->asic->pm_get_dynpm_state((rdev))
+#define radeon_pre_page_flip(rdev, crtc) rdev->asic->pre_page_flip((rdev), (crtc))
+#define radeon_page_flip(rdev, crtc, base) rdev->asic->page_flip((rdev), (crtc), (base))
+#define radeon_post_page_flip(rdev, crtc) rdev->asic->post_page_flip((rdev), (crtc))
 
 /* Common functions */
 /* AGP */

drivers/gpu/drm/radeon/radeon_asic.c

@@ -171,6 +171,9 @@ static struct radeon_asic r100_asic = {
 	.pm_finish = &r100_pm_finish,
 	.pm_init_profile = &r100_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &r100_pre_page_flip,
+	.page_flip = &r100_page_flip,
+	.post_page_flip = &r100_post_page_flip,
 };
 
 static struct radeon_asic r200_asic = {
@@ -215,6 +218,9 @@ static struct radeon_asic r200_asic = {
 	.pm_finish = &r100_pm_finish,
 	.pm_init_profile = &r100_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &r100_pre_page_flip,
+	.page_flip = &r100_page_flip,
+	.post_page_flip = &r100_post_page_flip,
 };
 
 static struct radeon_asic r300_asic = {
@@ -260,6 +266,9 @@ static struct radeon_asic r300_asic = {
 	.pm_finish = &r100_pm_finish,
 	.pm_init_profile = &r100_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &r100_pre_page_flip,
+	.page_flip = &r100_page_flip,
+	.post_page_flip = &r100_post_page_flip,
 };
 
 static struct radeon_asic r300_asic_pcie = {
@@ -304,6 +313,9 @@ static struct radeon_asic r300_asic_pcie = {
 	.pm_finish = &r100_pm_finish,
 	.pm_init_profile = &r100_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &r100_pre_page_flip,
+	.page_flip = &r100_page_flip,
+	.post_page_flip = &r100_post_page_flip,
 };
 
 static struct radeon_asic r420_asic = {
@@ -349,6 +361,9 @@ static struct radeon_asic r420_asic = {
 	.pm_finish = &r100_pm_finish,
 	.pm_init_profile = &r420_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &r100_pre_page_flip,
+	.page_flip = &r100_page_flip,
+	.post_page_flip = &r100_post_page_flip,
 };
 
 static struct radeon_asic rs400_asic = {
@@ -394,6 +409,9 @@ static struct radeon_asic rs400_asic = {
 	.pm_finish = &r100_pm_finish,
 	.pm_init_profile = &r100_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &r100_pre_page_flip,
+	.page_flip = &r100_page_flip,
+	.post_page_flip = &r100_post_page_flip,
 };
 
 static struct radeon_asic rs600_asic = {
@@ -439,6 +457,9 @@ static struct radeon_asic rs600_asic = {
 	.pm_finish = &rs600_pm_finish,
 	.pm_init_profile = &r420_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &rs600_pre_page_flip,
+	.page_flip = &rs600_page_flip,
+	.post_page_flip = &rs600_post_page_flip,
 };
 
 static struct radeon_asic rs690_asic = {
@@ -484,6 +505,9 @@ static struct radeon_asic rs690_asic = {
 	.pm_finish = &rs600_pm_finish,
 	.pm_init_profile = &r420_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &rs600_pre_page_flip,
+	.page_flip = &rs600_page_flip,
+	.post_page_flip = &rs600_post_page_flip,
 };
 
 static struct radeon_asic rv515_asic = {
@@ -529,6 +553,9 @@ static struct radeon_asic rv515_asic = {
 	.pm_finish = &rs600_pm_finish,
 	.pm_init_profile = &r420_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &rs600_pre_page_flip,
+	.page_flip = &rs600_page_flip,
+	.post_page_flip = &rs600_post_page_flip,
 };
 
 static struct radeon_asic r520_asic = {
@@ -574,6 +601,9 @@ static struct radeon_asic r520_asic = {
 	.pm_finish = &rs600_pm_finish,
 	.pm_init_profile = &r420_pm_init_profile,
 	.pm_get_dynpm_state = &r100_pm_get_dynpm_state,
+	.pre_page_flip = &rs600_pre_page_flip,
+	.page_flip = &rs600_page_flip,
+	.post_page_flip = &rs600_post_page_flip,
 };
 
 static struct radeon_asic r600_asic = {
@@ -618,6 +648,9 @@ static struct radeon_asic r600_asic = {
 	.pm_finish = &rs600_pm_finish,
 	.pm_init_profile = &r600_pm_init_profile,
 	.pm_get_dynpm_state = &r600_pm_get_dynpm_state,
+	.pre_page_flip = &rs600_pre_page_flip,
+	.page_flip = &rs600_page_flip,
+	.post_page_flip = &rs600_post_page_flip,
 };
 
 static struct radeon_asic rs780_asic = {
@@ -662,6 +695,9 @@ static struct radeon_asic rs780_asic = {
 	.pm_finish = &rs600_pm_finish,
 	.pm_init_profile = &rs780_pm_init_profile,
 	.pm_get_dynpm_state = &r600_pm_get_dynpm_state,
+	.pre_page_flip = &rs600_pre_page_flip,
+	.page_flip = &rs600_page_flip,
+	.post_page_flip = &rs600_post_page_flip,
 };
 
 static struct radeon_asic rv770_asic = {
@@ -706,6 +742,9 @@ static struct radeon_asic rv770_asic = {
 	.pm_finish = &rs600_pm_finish,
 	.pm_init_profile = &r600_pm_init_profile,
 	.pm_get_dynpm_state = &r600_pm_get_dynpm_state,
+	.pre_page_flip = &rs600_pre_page_flip,
+	.page_flip = &rv770_page_flip,
+	.post_page_flip = &rs600_post_page_flip,
 };
 
 static struct radeon_asic evergreen_asic = {
@@ -749,6 +788,9 @@ static struct radeon_asic evergreen_asic = {
 	.pm_finish = &evergreen_pm_finish,
 	.pm_init_profile = &r600_pm_init_profile,
 	.pm_get_dynpm_state = &r600_pm_get_dynpm_state,
+	.pre_page_flip = &evergreen_pre_page_flip,
+	.page_flip = &evergreen_page_flip,
+	.post_page_flip = &evergreen_post_page_flip,
 };
 
 int radeon_asic_init(struct radeon_device *rdev)

drivers/gpu/drm/radeon/radeon_asic.h

@@ -130,6 +130,9 @@ extern void r100_pm_prepare(struct radeon_device *rdev);
 extern void r100_pm_finish(struct radeon_device *rdev);
 extern void r100_pm_init_profile(struct radeon_device *rdev);
 extern void r100_pm_get_dynpm_state(struct radeon_device *rdev);
+extern void r100_pre_page_flip(struct radeon_device *rdev, int crtc);
+extern u32 r100_page_flip(struct radeon_device *rdev, int crtc, u64 crtc_base);
+extern void r100_post_page_flip(struct radeon_device *rdev, int crtc);
 
 /*
  * r200,rv250,rs300,rv280
@@ -205,6 +208,9 @@ void rs600_hpd_set_polarity(struct radeon_device *rdev,
 extern void rs600_pm_misc(struct radeon_device *rdev);
 extern void rs600_pm_prepare(struct radeon_device *rdev);
 extern void rs600_pm_finish(struct radeon_device *rdev);
+extern void rs600_pre_page_flip(struct radeon_device *rdev, int crtc);
+extern u32 rs600_page_flip(struct radeon_device *rdev, int crtc, u64 crtc_base);
+extern void rs600_post_page_flip(struct radeon_device *rdev, int crtc);
 
 /*
  * rs690,rs740
@@ -287,6 +293,7 @@ void rv770_fini(struct radeon_device *rdev);
 int rv770_suspend(struct radeon_device *rdev);
 int rv770_resume(struct radeon_device *rdev);
 extern void rv770_pm_misc(struct radeon_device *rdev);
+extern u32 rv770_page_flip(struct radeon_device *rdev, int crtc, u64 crtc_base);
 
 /*
  * evergreen
@@ -314,5 +321,8 @@ extern int evergreen_cs_parse(struct radeon_cs_parser *p);
 extern void evergreen_pm_misc(struct radeon_device *rdev);
 extern void evergreen_pm_prepare(struct radeon_device *rdev);
 extern void evergreen_pm_finish(struct radeon_device *rdev);
+extern void evergreen_pre_page_flip(struct radeon_device *rdev, int crtc);
+extern u32 evergreen_page_flip(struct radeon_device *rdev, int crtc, u64 crtc_base);
+extern void evergreen_post_page_flip(struct radeon_device *rdev, int crtc);
 
 #endif

drivers/gpu/drm/radeon/radeon_display.c

@@ -183,12 +183,273 @@ static void radeon_crtc_destroy(struct drm_crtc *crtc)
 	kfree(radeon_crtc);
 }
 
+/*
+ * Handle unpin events outside the interrupt handler proper.
+ */
+static void radeon_unpin_work_func(struct work_struct *__work)
+{
+	struct radeon_unpin_work *work =
+		container_of(__work, struct radeon_unpin_work, work);
+	int r;
+
+	/* unpin of the old buffer */
+	r = radeon_bo_reserve(work->old_rbo, false);
+	if (likely(r == 0)) {
+		r = radeon_bo_unpin(work->old_rbo);
+		if (unlikely(r != 0)) {
+			DRM_ERROR("failed to unpin buffer after flip\n");
+		}
+		radeon_bo_unreserve(work->old_rbo);
+	} else
+		DRM_ERROR("failed to reserve buffer after flip\n");
+	kfree(work);
+}
+
+void radeon_crtc_handle_flip(struct radeon_device *rdev, int crtc_id)
+{
+	struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc_id];
+	struct radeon_unpin_work *work;
+	struct drm_pending_vblank_event *e;
+	struct timeval now;
+	unsigned long flags;
+	u32 update_pending;
+	int vpos, hpos;
+
+	spin_lock_irqsave(&rdev->ddev->event_lock, flags);
+	work = radeon_crtc->unpin_work;
+	if (work == NULL ||
+	    !radeon_fence_signaled(work->fence)) {
+		spin_unlock_irqrestore(&rdev->ddev->event_lock, flags);
+		return;
+	}
+	/* New pageflip, or just completion of a previous one? */
+	if (!radeon_crtc->deferred_flip_completion) {
+		/* do the flip (mmio) */
+		update_pending = radeon_page_flip(rdev, crtc_id, work->new_crtc_base);
+	} else {
+		/* This is just a completion of a flip queued in crtc
+		 * at last invocation. Make sure we go directly to
+		 * completion routine.
+		 */
+		update_pending = 0;
+		radeon_crtc->deferred_flip_completion = 0;
+	}
+
+	/* Has the pageflip already completed in crtc, or is it certain
+	 * to complete in this vblank?
+	 */
+	if (update_pending &&
+	    (DRM_SCANOUTPOS_VALID & radeon_get_crtc_scanoutpos(rdev->ddev, crtc_id,
+							       &vpos, &hpos)) &&
+	    (vpos >=0) &&
+	    (vpos < (99 * rdev->mode_info.crtcs[crtc_id]->base.hwmode.crtc_vdisplay)/100)) {
+		/* crtc didn't flip in this target vblank interval,
+		 * but flip is pending in crtc. It will complete it
+		 * in next vblank interval, so complete the flip at
+		 * next vblank irq.
+		 */
+		radeon_crtc->deferred_flip_completion = 1;
+		spin_unlock_irqrestore(&rdev->ddev->event_lock, flags);
+		return;
+	}
+
+	/* Pageflip (will be) certainly completed in this vblank. Clean up. */
+	radeon_crtc->unpin_work = NULL;
+
+	/* wakeup userspace */
+	if (work->event) {
+		e = work->event;
+		do_gettimeofday(&now);
+		e->event.sequence = drm_vblank_count(rdev->ddev, radeon_crtc->crtc_id);
+		e->event.tv_sec = now.tv_sec;
+		e->event.tv_usec = now.tv_usec;
+		list_add_tail(&e->base.link, &e->base.file_priv->event_list);
+		wake_up_interruptible(&e->base.file_priv->event_wait);
+	}
+	spin_unlock_irqrestore(&rdev->ddev->event_lock, flags);
+
+	drm_vblank_put(rdev->ddev, radeon_crtc->crtc_id);
+	radeon_fence_unref(&work->fence);
+	radeon_post_page_flip(work->rdev, work->crtc_id);
+	schedule_work(&work->work);
+}
+
+static int radeon_crtc_page_flip(struct drm_crtc *crtc,
+				 struct drm_framebuffer *fb,
+				 struct drm_pending_vblank_event *event)
+{
+	struct drm_device *dev = crtc->dev;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc);
+	struct radeon_framebuffer *old_radeon_fb;
+	struct radeon_framebuffer *new_radeon_fb;
+	struct drm_gem_object *obj;
+	struct radeon_bo *rbo;
+	struct radeon_fence *fence;
+	struct radeon_unpin_work *work;
+	unsigned long flags;
+	u32 tiling_flags, pitch_pixels;
+	u64 base;
+	int r;
+
+	work = kzalloc(sizeof *work, GFP_KERNEL);
+	if (work == NULL)
+		return -ENOMEM;
+
+	r = radeon_fence_create(rdev, &fence);
+	if (unlikely(r != 0)) {
+		kfree(work);
+		DRM_ERROR("flip queue: failed to create fence.\n");
+		return -ENOMEM;
+	}
+	work->event = event;
+	work->rdev = rdev;
+	work->crtc_id = radeon_crtc->crtc_id;
+	work->fence = radeon_fence_ref(fence);
+	old_radeon_fb = to_radeon_framebuffer(crtc->fb);
+	new_radeon_fb = to_radeon_framebuffer(fb);
+	/* schedule unpin of the old buffer */
+	obj = old_radeon_fb->obj;
+	rbo = obj->driver_private;
+	work->old_rbo = rbo;
+	INIT_WORK(&work->work, radeon_unpin_work_func);
+
+	/* We borrow the event spin lock for protecting unpin_work */
+	spin_lock_irqsave(&dev->event_lock, flags);
+	if (radeon_crtc->unpin_work) {
+		spin_unlock_irqrestore(&dev->event_lock, flags);
+		kfree(work);
+		radeon_fence_unref(&fence);
+
+		DRM_DEBUG_DRIVER("flip queue: crtc already busy\n");
+		return -EBUSY;
+	}
+	radeon_crtc->unpin_work = work;
+	radeon_crtc->deferred_flip_completion = 0;
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+
+	/* pin the new buffer */
+	obj = new_radeon_fb->obj;
+	rbo = obj->driver_private;
+
+	DRM_DEBUG_DRIVER("flip-ioctl() cur_fbo = %p, cur_bbo = %p\n",
+			 work->old_rbo, rbo);
+
+	r = radeon_bo_reserve(rbo, false);
+	if (unlikely(r != 0)) {
+		DRM_ERROR("failed to reserve new rbo buffer before flip\n");
+		goto pflip_cleanup;
+	}
+	r = radeon_bo_pin(rbo, RADEON_GEM_DOMAIN_VRAM, &base);
+	if (unlikely(r != 0)) {
+		radeon_bo_unreserve(rbo);
+		r = -EINVAL;
+		DRM_ERROR("failed to pin new rbo buffer before flip\n");
+		goto pflip_cleanup;
+	}
+	radeon_bo_get_tiling_flags(rbo, &tiling_flags, NULL);
+	radeon_bo_unreserve(rbo);
+
+	if (!ASIC_IS_AVIVO(rdev)) {
+		/* crtc offset is from display base addr not FB location */
+		base -= radeon_crtc->legacy_display_base_addr;
+		pitch_pixels = fb->pitch / (fb->bits_per_pixel / 8);
+
+		if (tiling_flags & RADEON_TILING_MACRO) {
+			if (ASIC_IS_R300(rdev)) {
+				base &= ~0x7ff;
+			} else {
+				int byteshift = fb->bits_per_pixel >> 4;
+				int tile_addr = (((crtc->y >> 3) * pitch_pixels +  crtc->x) >> (8 - byteshift)) << 11;
+				base += tile_addr + ((crtc->x << byteshift) % 256) + ((crtc->y % 8) << 8);
+			}
+		} else {
+			int offset = crtc->y * pitch_pixels + crtc->x;
+			switch (fb->bits_per_pixel) {
+			case 8:
+			default:
+				offset *= 1;
+				break;
+			case 15:
+			case 16:
+				offset *= 2;
+				break;
+			case 24:
+				offset *= 3;
+				break;
+			case 32:
+				offset *= 4;
+				break;
+			}
+			base += offset;
+		}
+		base &= ~7;
+	}
+
+	spin_lock_irqsave(&dev->event_lock, flags);
+	work->new_crtc_base = base;
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+
+	/* update crtc fb */
+	crtc->fb = fb;
+
+	r = drm_vblank_get(dev, radeon_crtc->crtc_id);
+	if (r) {
+		DRM_ERROR("failed to get vblank before flip\n");
+		goto pflip_cleanup1;
+	}
+
+	/* 32 ought to cover us */
+	r = radeon_ring_lock(rdev, 32);
+	if (r) {
+		DRM_ERROR("failed to lock the ring before flip\n");
+		goto pflip_cleanup2;
+	}
+
+	/* emit the fence */
+	radeon_fence_emit(rdev, fence);
+	/* set the proper interrupt */
+	radeon_pre_page_flip(rdev, radeon_crtc->crtc_id);
+	/* fire the ring */
+	radeon_ring_unlock_commit(rdev);
+
+	return 0;
+
+pflip_cleanup2:
+	drm_vblank_put(dev, radeon_crtc->crtc_id);
+
+pflip_cleanup1:
+	r = radeon_bo_reserve(rbo, false);
+	if (unlikely(r != 0)) {
+		DRM_ERROR("failed to reserve new rbo in error path\n");
+		goto pflip_cleanup;
+	}
+	r = radeon_bo_unpin(rbo);
+	if (unlikely(r != 0)) {
+		radeon_bo_unreserve(rbo);
+		r = -EINVAL;
+		DRM_ERROR("failed to unpin new rbo in error path\n");
+		goto pflip_cleanup;
+	}
+	radeon_bo_unreserve(rbo);
+
+pflip_cleanup:
+	spin_lock_irqsave(&dev->event_lock, flags);
+	radeon_crtc->unpin_work = NULL;
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+	radeon_fence_unref(&fence);
+	kfree(work);
+
+	return r;
+}
+
 static const struct drm_crtc_funcs radeon_crtc_funcs = {
 	.cursor_set = radeon_crtc_cursor_set,
 	.cursor_move = radeon_crtc_cursor_move,
 	.gamma_set = radeon_crtc_gamma_set,
 	.set_config = drm_crtc_helper_set_config,
 	.destroy = radeon_crtc_destroy,
+	.page_flip = radeon_crtc_page_flip,
 };
 
 static void radeon_crtc_init(struct drm_device *dev, int index)

drivers/gpu/drm/radeon/radeon_drv.c

@@ -48,9 +48,10 @@
  * - 2.5.0 - add get accel 2 to work around ddx breakage for evergreen
  * - 2.6.0 - add tiling config query (r6xx+), add initial HiZ support (r300->r500)
  *   2.7.0 - fixups for r600 2D tiling support. (no external ABI change), add eg dyn gpr regs
+ *   2.8.0 - pageflip support
  */
 #define KMS_DRIVER_MAJOR	2
-#define KMS_DRIVER_MINOR	7
+#define KMS_DRIVER_MINOR	8
 #define KMS_DRIVER_PATCHLEVEL	0
 int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags);
 int radeon_driver_unload_kms(struct drm_device *dev);

drivers/gpu/drm/radeon/radeon_irq_kms.c

@@ -71,8 +71,10 @@ void radeon_driver_irq_preinstall_kms(struct drm_device *dev)
 	rdev->irq.gui_idle = false;
 	for (i = 0; i < rdev->num_crtc; i++)
 		rdev->irq.crtc_vblank_int[i] = false;
-	for (i = 0; i < 6; i++)
+	for (i = 0; i < 6; i++) {
 		rdev->irq.hpd[i] = false;
+		rdev->irq.pflip[i] = false;
+	}
 	radeon_irq_set(rdev);
 	/* Clear bits */
 	radeon_irq_process(rdev);
@@ -101,8 +103,10 @@ void radeon_driver_irq_uninstall_kms(struct drm_device *dev)
 	rdev->irq.gui_idle = false;
 	for (i = 0; i < rdev->num_crtc; i++)
 		rdev->irq.crtc_vblank_int[i] = false;
-	for (i = 0; i < 6; i++)
+	for (i = 0; i < 6; i++) {
 		rdev->irq.hpd[i] = false;
+		rdev->irq.pflip[i] = false;
+	}
 	radeon_irq_set(rdev);
 }
 
@@ -175,3 +179,34 @@ void radeon_irq_kms_sw_irq_put(struct radeon_device *rdev)
 	spin_unlock_irqrestore(&rdev->irq.sw_lock, irqflags);
 }
 
+void radeon_irq_kms_pflip_irq_get(struct radeon_device *rdev, int crtc)
+{
+	unsigned long irqflags;
+
+	if (crtc < 0 || crtc >= rdev->num_crtc)
+		return;
+
+	spin_lock_irqsave(&rdev->irq.pflip_lock[crtc], irqflags);
+	if (rdev->ddev->irq_enabled && (++rdev->irq.pflip_refcount[crtc] == 1)) {
+		rdev->irq.pflip[crtc] = true;
+		radeon_irq_set(rdev);
+	}
+	spin_unlock_irqrestore(&rdev->irq.pflip_lock[crtc], irqflags);
+}
+
+void radeon_irq_kms_pflip_irq_put(struct radeon_device *rdev, int crtc)
+{
+	unsigned long irqflags;
+
+	if (crtc < 0 || crtc >= rdev->num_crtc)
+		return;
+
+	spin_lock_irqsave(&rdev->irq.pflip_lock[crtc], irqflags);
+	BUG_ON(rdev->ddev->irq_enabled && rdev->irq.pflip_refcount[crtc] <= 0);
+	if (rdev->ddev->irq_enabled && (--rdev->irq.pflip_refcount[crtc] == 0)) {
+		rdev->irq.pflip[crtc] = false;
+		radeon_irq_set(rdev);
+	}
+	spin_unlock_irqrestore(&rdev->irq.pflip_lock[crtc], irqflags);
+}
+

drivers/gpu/drm/radeon/radeon_mode.h

@@ -277,6 +277,9 @@ struct radeon_crtc {
 	fixed20_12 hsc;
 	struct drm_display_mode native_mode;
 	int pll_id;
+	/* page flipping */
+	struct radeon_unpin_work *unpin_work;
+	int deferred_flip_completion;
 };
 
 struct radeon_encoder_primary_dac {
@@ -659,4 +662,7 @@ int radeon_fbdev_total_size(struct radeon_device *rdev);
 bool radeon_fbdev_robj_is_fb(struct radeon_device *rdev, struct radeon_bo *robj);
 
 void radeon_fb_output_poll_changed(struct radeon_device *rdev);
+
+void radeon_crtc_handle_flip(struct radeon_device *rdev, int crtc_id);
+
 #endif

drivers/gpu/drm/radeon/radeon_reg.h

@@ -422,6 +422,7 @@
 #       define RADEON_CRTC_CSYNC_EN         (1 <<  4)
 #       define RADEON_CRTC_ICON_EN          (1 << 15)
 #       define RADEON_CRTC_CUR_EN           (1 << 16)
+#       define RADEON_CRTC_VSTAT_MODE_MASK  (3 << 17)
 #       define RADEON_CRTC_CUR_MODE_MASK    (7 << 20)
 #       define RADEON_CRTC_CUR_MODE_SHIFT   20
 #       define RADEON_CRTC_CUR_MODE_MONO    0

drivers/gpu/drm/radeon/rs600.c

@@ -46,6 +46,56 @@
 void rs600_gpu_init(struct radeon_device *rdev);
 int rs600_mc_wait_for_idle(struct radeon_device *rdev);
 
+void rs600_pre_page_flip(struct radeon_device *rdev, int crtc)
+{
+	struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc];
+	u32 tmp;
+
+	/* make sure flip is at vb rather than hb */
+	tmp = RREG32(AVIVO_D1GRPH_FLIP_CONTROL + radeon_crtc->crtc_offset);
+	tmp &= ~AVIVO_D1GRPH_SURFACE_UPDATE_H_RETRACE_EN;
+	WREG32(AVIVO_D1GRPH_FLIP_CONTROL + radeon_crtc->crtc_offset, tmp);
+
+	/* set pageflip to happen anywhere in vblank interval */
+	WREG32(AVIVO_D1MODE_MASTER_UPDATE_MODE + radeon_crtc->crtc_offset, 0);
+
+	/* enable the pflip int */
+	radeon_irq_kms_pflip_irq_get(rdev, crtc);
+}
+
+void rs600_post_page_flip(struct radeon_device *rdev, int crtc)
+{
+	/* disable the pflip int */
+	radeon_irq_kms_pflip_irq_put(rdev, crtc);
+}
+
+u32 rs600_page_flip(struct radeon_device *rdev, int crtc_id, u64 crtc_base)
+{
+	struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc_id];
+	u32 tmp = RREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset);
+
+	/* Lock the graphics update lock */
+	tmp |= AVIVO_D1GRPH_UPDATE_LOCK;
+	WREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset, tmp);
+
+	/* update the scanout addresses */
+	WREG32(AVIVO_D1GRPH_SECONDARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
+	       (u32)crtc_base);
+	WREG32(AVIVO_D1GRPH_PRIMARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
+	       (u32)crtc_base);
+
+	/* Wait for update_pending to go high. */
+	while (!(RREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset) & AVIVO_D1GRPH_SURFACE_UPDATE_PENDING));
+	DRM_DEBUG("Update pending now high. Unlocking vupdate_lock.\n");
+
+	/* Unlock the lock, so double-buffering can take place inside vblank */
+	tmp &= ~AVIVO_D1GRPH_UPDATE_LOCK;
+	WREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset, tmp);
+
+	/* Return current update_pending status: */
+	return RREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset) & AVIVO_D1GRPH_SURFACE_UPDATE_PENDING;
+}
+
 void rs600_pm_misc(struct radeon_device *rdev)
 {
 	int requested_index = rdev->pm.requested_power_state_index;
@@ -515,10 +565,12 @@ int rs600_irq_set(struct radeon_device *rdev)
 	if (rdev->irq.gui_idle) {
 		tmp |= S_000040_GUI_IDLE(1);
 	}
-	if (rdev->irq.crtc_vblank_int[0]) {
+	if (rdev->irq.crtc_vblank_int[0] ||
+	    rdev->irq.pflip[0]) {
 		mode_int |= S_006540_D1MODE_VBLANK_INT_MASK(1);
 	}
-	if (rdev->irq.crtc_vblank_int[1]) {
+	if (rdev->irq.crtc_vblank_int[1] ||
+	    rdev->irq.pflip[1]) {
 		mode_int |= S_006540_D2MODE_VBLANK_INT_MASK(1);
 	}
 	if (rdev->irq.hpd[0]) {
@@ -534,7 +586,7 @@ int rs600_irq_set(struct radeon_device *rdev)
 	return 0;
 }
 
-static inline uint32_t rs600_irq_ack(struct radeon_device *rdev, u32 *r500_disp_int)
+static inline u32 rs600_irq_ack(struct radeon_device *rdev)
 {
 	uint32_t irqs = RREG32(R_000044_GEN_INT_STATUS);
 	uint32_t irq_mask = S_000044_SW_INT(1);
@@ -547,27 +599,27 @@ static inline uint32_t rs600_irq_ack(struct radeon_device *rdev, u32 *r500_disp_
 	}
 
 	if (G_000044_DISPLAY_INT_STAT(irqs)) {
-		*r500_disp_int = RREG32(R_007EDC_DISP_INTERRUPT_STATUS);
-		if (G_007EDC_LB_D1_VBLANK_INTERRUPT(*r500_disp_int)) {
+		rdev->irq.stat_regs.r500.disp_int = RREG32(R_007EDC_DISP_INTERRUPT_STATUS);
+		if (G_007EDC_LB_D1_VBLANK_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) {
 			WREG32(R_006534_D1MODE_VBLANK_STATUS,
 				S_006534_D1MODE_VBLANK_ACK(1));
 		}
-		if (G_007EDC_LB_D2_VBLANK_INTERRUPT(*r500_disp_int)) {
+		if (G_007EDC_LB_D2_VBLANK_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) {
 			WREG32(R_006D34_D2MODE_VBLANK_STATUS,
 				S_006D34_D2MODE_VBLANK_ACK(1));
 		}
-		if (G_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT(*r500_disp_int)) {
+		if (G_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) {
 			tmp = RREG32(R_007D08_DC_HOT_PLUG_DETECT1_INT_CONTROL);
 			tmp |= S_007D08_DC_HOT_PLUG_DETECT1_INT_ACK(1);
 			WREG32(R_007D08_DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp);
 		}
-		if (G_007EDC_DC_HOT_PLUG_DETECT2_INTERRUPT(*r500_disp_int)) {
+		if (G_007EDC_DC_HOT_PLUG_DETECT2_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) {
 			tmp = RREG32(R_007D18_DC_HOT_PLUG_DETECT2_INT_CONTROL);
 			tmp |= S_007D18_DC_HOT_PLUG_DETECT2_INT_ACK(1);
 			WREG32(R_007D18_DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp);
 		}
 	} else {
-		*r500_disp_int = 0;
+		rdev->irq.stat_regs.r500.disp_int = 0;
 	}
 
 	if (irqs) {
@@ -578,32 +630,30 @@ static inline uint32_t rs600_irq_ack(struct radeon_device *rdev, u32 *r500_disp_
 
 void rs600_irq_disable(struct radeon_device *rdev)
 {
-	u32 tmp;
-
 	WREG32(R_000040_GEN_INT_CNTL, 0);
 	WREG32(R_006540_DxMODE_INT_MASK, 0);
 	/* Wait and acknowledge irq */
 	mdelay(1);
-	rs600_irq_ack(rdev, &tmp);
+	rs600_irq_ack(rdev);
 }
 
 int rs600_irq_process(struct radeon_device *rdev)
 {
-	uint32_t status, msi_rearm;
-	uint32_t r500_disp_int;
+	u32 status, msi_rearm;
 	bool queue_hotplug = false;
 
 	/* reset gui idle ack.  the status bit is broken */
 	rdev->irq.gui_idle_acked = false;
 
-	status = rs600_irq_ack(rdev, &r500_disp_int);
-	if (!status && !r500_disp_int) {
+	status = rs600_irq_ack(rdev);
+	if (!status && !rdev->irq.stat_regs.r500.disp_int) {
 		return IRQ_NONE;
 	}
-	while (status || r500_disp_int) {
+	while (status || rdev->irq.stat_regs.r500.disp_int) {
 		/* SW interrupt */
-		if (G_000044_SW_INT(status))
+		if (G_000044_SW_INT(status)) {
 			radeon_fence_process(rdev);
+		}
 		/* GUI idle */
 		if (G_000040_GUI_IDLE(status)) {
 			rdev->irq.gui_idle_acked = true;
@@ -611,25 +661,33 @@ int rs600_irq_process(struct radeon_device *rdev)
 			wake_up(&rdev->irq.idle_queue);
 		}
 		/* Vertical blank interrupts */
-		if (G_007EDC_LB_D1_VBLANK_INTERRUPT(r500_disp_int)) {
+		if (G_007EDC_LB_D1_VBLANK_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) {
+			if (rdev->irq.pflip[0])
+				radeon_crtc_handle_flip(rdev, 0);
+			if (rdev->irq.crtc_vblank_int[0]) {
 				drm_handle_vblank(rdev->ddev, 0);
 				rdev->pm.vblank_sync = true;
 				wake_up(&rdev->irq.vblank_queue);
 			}
-		if (G_007EDC_LB_D2_VBLANK_INTERRUPT(r500_disp_int)) {
+		}
+		if (G_007EDC_LB_D2_VBLANK_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) {
+			if (rdev->irq.pflip[1])
+				radeon_crtc_handle_flip(rdev, 1);
+			if (rdev->irq.crtc_vblank_int[1]) {
 				drm_handle_vblank(rdev->ddev, 1);
 				rdev->pm.vblank_sync = true;
 				wake_up(&rdev->irq.vblank_queue);
 			}
-		if (G_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT(r500_disp_int)) {
+		}
+		if (G_007EDC_DC_HOT_PLUG_DETECT1_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) {
 			queue_hotplug = true;
 			DRM_DEBUG("HPD1\n");
 		}
-		if (G_007EDC_DC_HOT_PLUG_DETECT2_INTERRUPT(r500_disp_int)) {
+		if (G_007EDC_DC_HOT_PLUG_DETECT2_INTERRUPT(rdev->irq.stat_regs.r500.disp_int)) {
 			queue_hotplug = true;
 			DRM_DEBUG("HPD2\n");
 		}
-		status = rs600_irq_ack(rdev, &r500_disp_int);
+		status = rs600_irq_ack(rdev);
 	}
 	/* reset gui idle ack.  the status bit is broken */
 	rdev->irq.gui_idle_acked = false;

drivers/gpu/drm/radeon/rv770.c

@@ -42,6 +42,40 @@
 static void rv770_gpu_init(struct radeon_device *rdev);
 void rv770_fini(struct radeon_device *rdev);
 
+u32 rv770_page_flip(struct radeon_device *rdev, int crtc_id, u64 crtc_base)
+{
+	struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc_id];
+	u32 tmp = RREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset);
+
+	/* Lock the graphics update lock */
+	tmp |= AVIVO_D1GRPH_UPDATE_LOCK;
+	WREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset, tmp);
+
+	/* update the scanout addresses */
+	if (radeon_crtc->crtc_id) {
+		WREG32(D2GRPH_SECONDARY_SURFACE_ADDRESS_HIGH, upper_32_bits(crtc_base));
+		WREG32(D2GRPH_PRIMARY_SURFACE_ADDRESS_HIGH, upper_32_bits(crtc_base));
+	} else {
+		WREG32(D1GRPH_SECONDARY_SURFACE_ADDRESS_HIGH, upper_32_bits(crtc_base));
+		WREG32(D1GRPH_PRIMARY_SURFACE_ADDRESS_HIGH, upper_32_bits(crtc_base));
+	}
+	WREG32(D1GRPH_SECONDARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
+	       (u32)crtc_base);
+	WREG32(D1GRPH_PRIMARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
+	       (u32)crtc_base);
+
+	/* Wait for update_pending to go high. */
+	while (!(RREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset) & AVIVO_D1GRPH_SURFACE_UPDATE_PENDING));
+	DRM_DEBUG("Update pending now high. Unlocking vupdate_lock.\n");
+
+	/* Unlock the lock, so double-buffering can take place inside vblank */
+	tmp &= ~AVIVO_D1GRPH_UPDATE_LOCK;
+	WREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset, tmp);
+
+	/* Return current update_pending status: */
+	return RREG32(AVIVO_D1GRPH_UPDATE + radeon_crtc->crtc_offset) & AVIVO_D1GRPH_SURFACE_UPDATE_PENDING;
+}
+
 /* get temperature in millidegrees */
 u32 rv770_get_temp(struct radeon_device *rdev)
 {

drivers/gpu/drm/radeon/rv770d.h

@@ -351,4 +351,11 @@
 
 #define	SRBM_STATUS				        0x0E50
 
+#define D1GRPH_PRIMARY_SURFACE_ADDRESS                    0x6110
+#define D1GRPH_PRIMARY_SURFACE_ADDRESS_HIGH               0x6914
+#define D2GRPH_PRIMARY_SURFACE_ADDRESS_HIGH               0x6114
+#define D1GRPH_SECONDARY_SURFACE_ADDRESS                  0x6118
+#define D1GRPH_SECONDARY_SURFACE_ADDRESS_HIGH             0x691c
+#define D2GRPH_SECONDARY_SURFACE_ADDRESS_HIGH             0x611c
+
 #endif