drm/radeon: rework UVD writeback & [rw]ptr handling

The hardware just doesn't support this correctly.
Disable it before we accidentally write anywhere we shouldn't.

Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/radeon/cik.c

@@ -7705,8 +7705,7 @@ static int cik_startup(struct radeon_device *rdev)
 
 	ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX];
 	if (ring->ring_size) {
-		r = radeon_ring_init(rdev, ring, ring->ring_size,
-				     R600_WB_UVD_RPTR_OFFSET,
+		r = radeon_ring_init(rdev, ring, ring->ring_size, 0,
 				     UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR,
 				     0, 0xfffff, RADEON_CP_PACKET2);
 		if (!r)

drivers/gpu/drm/radeon/evergreen.c

@@ -5291,8 +5291,7 @@ static int evergreen_startup(struct radeon_device *rdev)
 
 	ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX];
 	if (ring->ring_size) {
-		r = radeon_ring_init(rdev, ring, ring->ring_size,
-				     R600_WB_UVD_RPTR_OFFSET,
+		r = radeon_ring_init(rdev, ring, ring->ring_size, 0,
 				     UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR,
 				     0, 0xfffff, RADEON_CP_PACKET2);
 		if (!r)

drivers/gpu/drm/radeon/ni.c

@@ -2225,8 +2225,7 @@ static int cayman_startup(struct radeon_device *rdev)
 
 	ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX];
 	if (ring->ring_size) {
-		r = radeon_ring_init(rdev, ring, ring->ring_size,
-				     R600_WB_UVD_RPTR_OFFSET,
+		r = radeon_ring_init(rdev, ring, ring->ring_size, 0,
 				     UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR,
 				     0, 0xfffff, RADEON_CP_PACKET2);
 		if (!r)

drivers/gpu/drm/radeon/r600.c

@@ -2623,31 +2623,38 @@ void r600_dma_fini(struct radeon_device *rdev)
 /*
  * UVD
  */
+uint32_t r600_uvd_get_rptr(struct radeon_device *rdev,
+			   struct radeon_ring *ring)
+{
+	return RREG32(UVD_RBC_RB_RPTR);
+}
+
+uint32_t r600_uvd_get_wptr(struct radeon_device *rdev,
+			   struct radeon_ring *ring)
+{
+	return RREG32(UVD_RBC_RB_WPTR);
+}
+
+void r600_uvd_set_wptr(struct radeon_device *rdev,
+		       struct radeon_ring *ring)
+{
+	WREG32(UVD_RBC_RB_WPTR, ring->wptr);
+}
+
 static int r600_uvd_rbc_start(struct radeon_device *rdev, bool ring_test)
 {
 	struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX];
-	uint64_t rptr_addr;
 	uint32_t rb_bufsz, tmp;
 	int r;
 
-	rptr_addr = rdev->wb.gpu_addr + R600_WB_UVD_RPTR_OFFSET;
-
-	if (upper_32_bits(rptr_addr) != upper_32_bits(ring->gpu_addr)) {
-		DRM_ERROR("UVD ring and rptr not in the same 4GB segment!\n");
-		return -EINVAL;
-	}
-
 	/* force RBC into idle state */
 	WREG32(UVD_RBC_RB_CNTL, 0x11010101);
 
 	/* Set the write pointer delay */
 	WREG32(UVD_RBC_RB_WPTR_CNTL, 0);
 
-	/* set the wb address */
-	WREG32(UVD_RBC_RB_RPTR_ADDR, rptr_addr >> 2);
-
 	/* programm the 4GB memory segment for rptr and ring buffer */
-	WREG32(UVD_LMI_EXT40_ADDR, upper_32_bits(rptr_addr) |
+	WREG32(UVD_LMI_EXT40_ADDR, upper_32_bits(ring->gpu_addr) |
 				   (0x7 << 16) | (0x1 << 31));
 
 	/* Initialize the ring buffer's read and write pointers */
@@ -2662,7 +2669,7 @@ static int r600_uvd_rbc_start(struct radeon_device *rdev, bool ring_test)
 	/* Set ring buffer size */
 	rb_bufsz = drm_order(ring->ring_size);
 	rb_bufsz = (0x1 << 8) | rb_bufsz;
-	WREG32(UVD_RBC_RB_CNTL, rb_bufsz);
+	WREG32_P(UVD_RBC_RB_CNTL, rb_bufsz, ~0x11f1f);
 
 	if (ring_test) {
 		ring->ready = true;

drivers/gpu/drm/radeon/radeon.h

@@ -1027,7 +1027,6 @@ struct radeon_wb {
 #define R600_WB_DMA_RPTR_OFFSET   1792
 #define R600_WB_IH_WPTR_OFFSET   2048
 #define CAYMAN_WB_DMA1_RPTR_OFFSET   2304
-#define R600_WB_UVD_RPTR_OFFSET  2560
 #define R600_WB_EVENT_OFFSET     3072
 #define CIK_WB_CP1_WPTR_OFFSET     3328
 #define CIK_WB_CP2_WPTR_OFFSET     3584

drivers/gpu/drm/radeon/radeon_asic.c

@@ -1157,9 +1157,9 @@ static struct radeon_asic_ring rv770_uvd_ring = {
 	.ring_test = &r600_uvd_ring_test,
 	.ib_test = &r600_uvd_ib_test,
 	.is_lockup = &radeon_ring_test_lockup,
-	.get_rptr = &radeon_ring_generic_get_rptr,
-	.get_wptr = &radeon_ring_generic_get_wptr,
-	.set_wptr = &radeon_ring_generic_set_wptr,
+	.get_rptr = &r600_uvd_get_rptr,
+	.get_wptr = &r600_uvd_get_wptr,
+	.set_wptr = &r600_uvd_set_wptr,
 };
 
 static struct radeon_asic rv770_asic = {
@@ -1593,9 +1593,9 @@ static struct radeon_asic_ring cayman_uvd_ring = {
 	.ring_test = &r600_uvd_ring_test,
 	.ib_test = &r600_uvd_ib_test,
 	.is_lockup = &radeon_ring_test_lockup,
-	.get_rptr = &radeon_ring_generic_get_rptr,
-	.get_wptr = &radeon_ring_generic_get_wptr,
-	.set_wptr = &radeon_ring_generic_set_wptr,
+	.get_rptr = &r600_uvd_get_rptr,
+	.get_wptr = &r600_uvd_get_wptr,
+	.set_wptr = &r600_uvd_set_wptr,
 };
 
 static struct radeon_asic cayman_asic = {

drivers/gpu/drm/radeon/radeon_asic.h

@@ -424,6 +424,12 @@ void rs780_dpm_debugfs_print_current_performance_level(struct radeon_device *rde
 						       struct seq_file *m);
 
 /* uvd */
+uint32_t r600_uvd_get_rptr(struct radeon_device *rdev,
+                           struct radeon_ring *ring);
+uint32_t r600_uvd_get_wptr(struct radeon_device *rdev,
+                           struct radeon_ring *ring);
+void r600_uvd_set_wptr(struct radeon_device *rdev,
+                       struct radeon_ring *ring);
 int r600_uvd_init(struct radeon_device *rdev, bool ring_test);
 void r600_uvd_stop(struct radeon_device *rdev);
 int r600_uvd_ib_test(struct radeon_device *rdev, struct radeon_ring *ring);

drivers/gpu/drm/radeon/radeon_ring.c

@@ -363,7 +363,7 @@ u32 radeon_ring_generic_get_rptr(struct radeon_device *rdev,
 {
 	u32 rptr;
 
-	if (rdev->wb.enabled && ring != &rdev->ring[R600_RING_TYPE_UVD_INDEX])
+	if (rdev->wb.enabled)
 		rptr = le32_to_cpu(rdev->wb.wb[ring->rptr_offs/4]);
 	else
 		rptr = RREG32(ring->rptr_reg);

drivers/gpu/drm/radeon/rv770.c

@@ -1923,8 +1923,7 @@ static int rv770_startup(struct radeon_device *rdev)
 
 	ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX];
 	if (ring->ring_size) {
-		r = radeon_ring_init(rdev, ring, ring->ring_size,
-				     R600_WB_UVD_RPTR_OFFSET,
+		r = radeon_ring_init(rdev, ring, ring->ring_size, 0,
 				     UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR,
 				     0, 0xfffff, RADEON_CP_PACKET2);
 		if (!r)

drivers/gpu/drm/radeon/si.c

@@ -6416,8 +6416,7 @@ static int si_startup(struct radeon_device *rdev)
 	if (rdev->has_uvd) {
 		ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX];
 		if (ring->ring_size) {
-			r = radeon_ring_init(rdev, ring, ring->ring_size,
-					     R600_WB_UVD_RPTR_OFFSET,
+			r = radeon_ring_init(rdev, ring, ring->ring_size, 0,
 					     UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR,
 					     0, 0xfffff, RADEON_CP_PACKET2);
 			if (!r)