Merge branch 'drm-patches' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6

* 'drm-patches' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6: (21 commits)
  drm: only trust core drm ioctls - driver ioctls are a mess.
  drm/i915: add support for Intel series 4 chipsets.
  drm/radeon: add hier-z registers for r300 and r500 chipsets
  drm/radeon: use DSTCACHE_CTLSTAT rather than RB2D_DSTCACHE_CTLSTAT
  drm/radeon: switch IGP gart to use radeon_write_agp_base()
  drm/radeon: Restore sw interrupt on resume
  drm/r500: add support for AGP based cards.
  drm/radeon: fix texture uploads with large 3d textures (bug 13980)
  drm/radeon: add initial r500 support.
  drm/radeon: init pipe setup in kernel code.
  drm/radeon: fixup radeon_do_engine_reset
  drm/radeon: fix pixcache and purge/cache flushing registers
  drm/radeon: write AGP_BASE_2 on chips that support it.
  drm/radeon: merge IGP chip setup and fixup RS400 vs RS480 support
  drm/radeon: IGP clean up register and magic numbers.
  drm/rs690: set base 2 to 0.
  drm/rs690: set all of gart base address.
  radeon: add production microcode from AMD
  drm: pcigart use proper pci map interfaces.
  drm: the sg alloc ioctl should write back the handle to userspace
  ...

drivers/char/drm/ati_pcigart.c

@@ -76,7 +76,7 @@ int drm_ati_pcigart_cleanup(struct drm_device *dev, struct drm_ati_pcigart_info
 		for (i = 0; i < pages; i++) {
 			if (!entry->busaddr[i])
 				break;
-			pci_unmap_single(dev->pdev, entry->busaddr[i],
+			pci_unmap_page(dev->pdev, entry->busaddr[i],
 					 PAGE_SIZE, PCI_DMA_TODEVICE);
 		}
 
@@ -137,10 +137,8 @@ int drm_ati_pcigart_init(struct drm_device *dev, struct drm_ati_pcigart_info *ga
 
 	for (i = 0; i < pages; i++) {
 		/* we need to support large memory configurations */
-		entry->busaddr[i] = pci_map_single(dev->pdev,
-						   page_address(entry->
-								pagelist[i]),
-						   PAGE_SIZE, PCI_DMA_TODEVICE);
+		entry->busaddr[i] = pci_map_page(dev->pdev, entry->pagelist[i],
+						 0, PAGE_SIZE, PCI_DMA_TODEVICE);
 		if (entry->busaddr[i] == 0) {
 			DRM_ERROR("unable to map PCIGART pages!\n");
 			drm_ati_pcigart_cleanup(dev, gart_info);

drivers/char/drm/drm.h

@@ -628,7 +628,7 @@ struct drm_set_version {
 #define DRM_IOCTL_AGP_BIND		DRM_IOW( 0x36, struct drm_agp_binding)
 #define DRM_IOCTL_AGP_UNBIND		DRM_IOW( 0x37, struct drm_agp_binding)
 
-#define DRM_IOCTL_SG_ALLOC		DRM_IOW( 0x38, struct drm_scatter_gather)
+#define DRM_IOCTL_SG_ALLOC		DRM_IOWR(0x38, struct drm_scatter_gather)
 #define DRM_IOCTL_SG_FREE		DRM_IOW( 0x39, struct drm_scatter_gather)
 
 #define DRM_IOCTL_WAIT_VBLANK		DRM_IOWR(0x3a, union drm_wait_vblank)

drivers/char/drm/drm_drv.c

@@ -470,17 +470,18 @@ int drm_ioctl(struct inode *inode, struct file *filp,
 	if ((nr >= DRM_COMMAND_BASE) && (nr < DRM_COMMAND_END) &&
 	    (nr < DRM_COMMAND_BASE + dev->driver->num_ioctls))
 		ioctl = &dev->driver->ioctls[nr - DRM_COMMAND_BASE];
-	else if ((nr >= DRM_COMMAND_END) || (nr < DRM_COMMAND_BASE))
+	else if ((nr >= DRM_COMMAND_END) || (nr < DRM_COMMAND_BASE)) {
 		ioctl = &drm_ioctls[nr];
-	else
+		cmd = ioctl->cmd;
+	} else
 		goto err_i1;
 
+	/* Do not trust userspace, use our own definition */
 	func = ioctl->func;
 	/* is there a local override? */
 	if ((nr == DRM_IOCTL_NR(DRM_IOCTL_DMA)) && dev->driver->dma_ioctl)
 		func = dev->driver->dma_ioctl;
 
-
 	if (!func) {
 		DRM_DEBUG("no function\n");
 		retcode = -EINVAL;

drivers/char/drm/drm_pciids.h

@@ -103,20 +103,18 @@
 	{0x1002, 0x5653, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV410|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP}, \
 	{0x1002, 0x5834, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS300|RADEON_IS_IGP}, \
 	{0x1002, 0x5835, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS300|RADEON_IS_IGP|RADEON_IS_MOBILITY}, \
-	{0x1002, 0x5954, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
-	{0x1002, 0x5955, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
-	{0x1002, 0x5974, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
-	{0x1002, 0x5975, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
+	{0x1002, 0x5954, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
+	{0x1002, 0x5955, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
+	{0x1002, 0x5974, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
+	{0x1002, 0x5975, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
 	{0x1002, 0x5960, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV280}, \
 	{0x1002, 0x5961, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV280}, \
 	{0x1002, 0x5962, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV280}, \
 	{0x1002, 0x5964, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV280}, \
 	{0x1002, 0x5965, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV280}, \
 	{0x1002, 0x5969, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV100}, \
-	{0x1002, 0x5a41, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
-	{0x1002, 0x5a42, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
-	{0x1002, 0x5a61, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
-	{0x1002, 0x5a62, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
+	{0x1002, 0x5a61, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
+	{0x1002, 0x5a62, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS480|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
 	{0x1002, 0x5b60, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV380|RADEON_NEW_MEMMAP}, \
 	{0x1002, 0x5b62, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV380|RADEON_NEW_MEMMAP}, \
 	{0x1002, 0x5b63, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV380|RADEON_NEW_MEMMAP}, \
@@ -411,4 +409,7 @@
 	{0x8086, 0x2a02, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0x8086, 0x2a12, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0x8086, 0x2a42, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
+	{0x8086, 0x2e02, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
+	{0x8086, 0x2e12, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
+	{0x8086, 0x2e22, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0, 0, 0}

drivers/char/drm/i915_drv.h

@@ -1112,12 +1112,19 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller);
 		       (dev)->pci_device == 0x29A2 || \
 		       (dev)->pci_device == 0x2A02 || \
 		       (dev)->pci_device == 0x2A12 || \
-		       (dev)->pci_device == 0x2A42)
+		       (dev)->pci_device == 0x2A42 || \
+		       (dev)->pci_device == 0x2E02 || \
+		       (dev)->pci_device == 0x2E12 || \
+		       (dev)->pci_device == 0x2E22)
 
 #define IS_I965GM(dev) ((dev)->pci_device == 0x2A02)
 
 #define IS_IGD_GM(dev) ((dev)->pci_device == 0x2A42)
 
+#define IS_G4X(dev) ((dev)->pci_device == 0x2E02 || \
+		     (dev)->pci_device == 0x2E12 || \
+		     (dev)->pci_device == 0x2E22)
+
 #define IS_G33(dev)    ((dev)->pci_device == 0x29C2 ||	\
 			(dev)->pci_device == 0x29B2 ||	\
 			(dev)->pci_device == 0x29D2)
@@ -1128,7 +1135,7 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller);
 #define IS_MOBILE(dev) (IS_I830(dev) || IS_I85X(dev) || IS_I915GM(dev) || \
 			IS_I945GM(dev) || IS_I965GM(dev) || IS_IGD_GM(dev))
 
-#define I915_NEED_GFX_HWS(dev) (IS_G33(dev) || IS_IGD_GM(dev))
+#define I915_NEED_GFX_HWS(dev) (IS_G33(dev) || IS_IGD_GM(dev) || IS_G4X(dev))
 
 #define PRIMARY_RINGBUFFER_SIZE         (128*1024)
 

drivers/char/drm/r300_cmdbuf.c

@@ -189,18 +189,12 @@ void r300_init_reg_flags(struct drm_device *dev)
 	ADD_RANGE(R300_RE_CULL_CNTL, 1);
 	ADD_RANGE(0x42C0, 2);
 	ADD_RANGE(R300_RS_CNTL_0, 2);
-	ADD_RANGE(R300_RS_INTERP_0, 8);
-	ADD_RANGE(R300_RS_ROUTE_0, 8);
-	ADD_RANGE(0x43A4, 2);
+
+	ADD_RANGE(R300_SC_HYPERZ, 2);
 	ADD_RANGE(0x43E8, 1);
-	ADD_RANGE(R300_PFS_CNTL_0, 3);
-	ADD_RANGE(R300_PFS_NODE_0, 4);
-	ADD_RANGE(R300_PFS_TEXI_0, 64);
+
 	ADD_RANGE(0x46A4, 5);
-	ADD_RANGE(R300_PFS_INSTR0_0, 64);
-	ADD_RANGE(R300_PFS_INSTR1_0, 64);
-	ADD_RANGE(R300_PFS_INSTR2_0, 64);
-	ADD_RANGE(R300_PFS_INSTR3_0, 64);
+
 	ADD_RANGE(R300_RE_FOG_STATE, 1);
 	ADD_RANGE(R300_FOG_COLOR_R, 3);
 	ADD_RANGE(R300_PP_ALPHA_TEST, 2);
@@ -215,14 +209,12 @@ void r300_init_reg_flags(struct drm_device *dev)
 	ADD_RANGE(0x4E50, 9);
 	ADD_RANGE(0x4E88, 1);
 	ADD_RANGE(0x4EA0, 2);
-	ADD_RANGE(R300_RB3D_ZSTENCIL_CNTL_0, 3);
-	ADD_RANGE(R300_RB3D_ZSTENCIL_FORMAT, 4);
-	ADD_RANGE_MARK(R300_RB3D_DEPTHOFFSET, 1, MARK_CHECK_OFFSET);	/* check offset */
-	ADD_RANGE(R300_RB3D_DEPTHPITCH, 1);
-	ADD_RANGE(0x4F28, 1);
-	ADD_RANGE(0x4F30, 2);
-	ADD_RANGE(0x4F44, 1);
-	ADD_RANGE(0x4F54, 1);
+	ADD_RANGE(R300_ZB_CNTL, 3);
+	ADD_RANGE(R300_ZB_FORMAT, 4);
+	ADD_RANGE_MARK(R300_ZB_DEPTHOFFSET, 1, MARK_CHECK_OFFSET);	/* check offset */
+	ADD_RANGE(R300_ZB_DEPTHPITCH, 1);
+	ADD_RANGE(R300_ZB_DEPTHCLEARVALUE, 1);
+	ADD_RANGE(R300_ZB_ZMASK_OFFSET, 13);
 
 	ADD_RANGE(R300_TX_FILTER_0, 16);
 	ADD_RANGE(R300_TX_FILTER1_0, 16);
@@ -235,13 +227,32 @@ void r300_init_reg_flags(struct drm_device *dev)
 	ADD_RANGE(R300_TX_BORDER_COLOR_0, 16);
 
 	/* Sporadic registers used as primitives are emitted */
-	ADD_RANGE(R300_RB3D_ZCACHE_CTLSTAT, 1);
+	ADD_RANGE(R300_ZB_ZCACHE_CTLSTAT, 1);
 	ADD_RANGE(R300_RB3D_DSTCACHE_CTLSTAT, 1);
 	ADD_RANGE(R300_VAP_INPUT_ROUTE_0_0, 8);
 	ADD_RANGE(R300_VAP_INPUT_ROUTE_1_0, 8);
 
 	if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV515) {
-		ADD_RANGE(0x4074, 16);
+		ADD_RANGE(R500_VAP_INDEX_OFFSET, 1);
+		ADD_RANGE(R500_US_CONFIG, 2);
+		ADD_RANGE(R500_US_CODE_ADDR, 3);
+		ADD_RANGE(R500_US_FC_CTRL, 1);
+		ADD_RANGE(R500_RS_IP_0, 16);
+		ADD_RANGE(R500_RS_INST_0, 16);
+		ADD_RANGE(R500_RB3D_COLOR_CLEAR_VALUE_AR, 2);
+		ADD_RANGE(R500_RB3D_CONSTANT_COLOR_AR, 2);
+		ADD_RANGE(R500_ZB_FIFO_SIZE, 2);
+	} else {
+		ADD_RANGE(R300_PFS_CNTL_0, 3);
+		ADD_RANGE(R300_PFS_NODE_0, 4);
+		ADD_RANGE(R300_PFS_TEXI_0, 64);
+		ADD_RANGE(R300_PFS_INSTR0_0, 64);
+		ADD_RANGE(R300_PFS_INSTR1_0, 64);
+		ADD_RANGE(R300_PFS_INSTR2_0, 64);
+		ADD_RANGE(R300_PFS_INSTR3_0, 64);
+		ADD_RANGE(R300_RS_INTERP_0, 8);
+		ADD_RANGE(R300_RS_ROUTE_0, 8);
+
 	}
 }
 
@@ -707,8 +718,9 @@ static __inline__ void r300_pacify(drm_radeon_private_t *dev_priv)
 	BEGIN_RING(6);
 	OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
 	OUT_RING(R300_RB3D_DSTCACHE_UNKNOWN_0A);
-	OUT_RING(CP_PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
-	OUT_RING(R300_RB3D_ZCACHE_UNKNOWN_03);
+	OUT_RING(CP_PACKET0(R300_ZB_ZCACHE_CTLSTAT, 0));
+	OUT_RING(R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE|
+		 R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
 	OUT_RING(CP_PACKET3(RADEON_CP_NOP, 0));
 	OUT_RING(0x0);
 	ADVANCE_RING();
@@ -828,6 +840,54 @@ static int r300_scratch(drm_radeon_private_t *dev_priv,
 	return 0;
 }
 
+/**
+ * Uploads user-supplied vertex program instructions or parameters onto
+ * the graphics card.
+ * Called by r300_do_cp_cmdbuf.
+ */
+static inline int r300_emit_r500fp(drm_radeon_private_t *dev_priv,
+				       drm_radeon_kcmd_buffer_t *cmdbuf,
+				       drm_r300_cmd_header_t header)
+{
+	int sz;
+	int addr;
+	int type;
+	int clamp;
+	int stride;
+	RING_LOCALS;
+
+	sz = header.r500fp.count;
+	/* address is 9 bits 0 - 8, bit 1 of flags is part of address */
+	addr = ((header.r500fp.adrhi_flags & 1) << 8) | header.r500fp.adrlo;
+
+	type = !!(header.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE);
+	clamp = !!(header.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP);
+
+	addr |= (type << 16);
+	addr |= (clamp << 17);
+
+	stride = type ? 4 : 6;
+
+	DRM_DEBUG("r500fp %d %d type: %d\n", sz, addr, type);
+	if (!sz)
+		return 0;
+	if (sz * stride * 4 > cmdbuf->bufsz)
+		return -EINVAL;
+
+	BEGIN_RING(3 + sz * stride);
+	OUT_RING_REG(R500_GA_US_VECTOR_INDEX, addr);
+	OUT_RING(CP_PACKET0_TABLE(R500_GA_US_VECTOR_DATA, sz * stride - 1));
+	OUT_RING_TABLE((int *)cmdbuf->buf, sz * stride);
+
+	ADVANCE_RING();
+
+	cmdbuf->buf += sz * stride * 4;
+	cmdbuf->bufsz -= sz * stride * 4;
+
+	return 0;
+}
+
+
 /**
  * Parses and validates a user-supplied command buffer and emits appropriate
  * commands on the DMA ring buffer.
@@ -963,6 +1023,19 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 			}
 			break;
 
+		case R300_CMD_R500FP:
+			if ((dev_priv->flags & RADEON_FAMILY_MASK) < CHIP_RV515) {
+				DRM_ERROR("Calling r500 command on r300 card\n");
+				ret = -EINVAL;
+				goto cleanup;
+			}
+			DRM_DEBUG("R300_CMD_R500FP\n");
+			ret = r300_emit_r500fp(dev_priv, cmdbuf, header);
+			if (ret) {
+				DRM_ERROR("r300_emit_r500fp failed\n");
+				goto cleanup;
+			}
+			break;
 		default:
 			DRM_ERROR("bad cmd_type %i at %p\n",
 				  header.header.cmd_type,

drivers/char/drm/radeon_drm.h

@@ -240,6 +240,7 @@ typedef union {
 #	define R300_NEW_WAIT_2D_2D_CLEAN_3D_3D_CLEAN	0x8
 
 #define R300_CMD_SCRATCH		8
+#define R300_CMD_R500FP                 9
 
 typedef union {
 	unsigned int u;
@@ -268,6 +269,9 @@ typedef union {
 	struct {
 		unsigned char cmd_type, reg, n_bufs, flags;
 	} scratch;
+	struct {
+		unsigned char cmd_type, count, adrlo, adrhi_flags;
+	} r500fp;
 } drm_r300_cmd_header_t;
 
 #define RADEON_FRONT			0x1
@@ -278,6 +282,9 @@ typedef union {
 #define RADEON_USE_HIERZ		0x40000000
 #define RADEON_USE_COMP_ZBUF		0x20000000
 
+#define R500FP_CONSTANT_TYPE  (1 << 1)
+#define R500FP_CONSTANT_CLAMP (1 << 2)
+
 /* Primitive types
  */
 #define RADEON_POINTS			0x1
@@ -669,6 +676,7 @@ typedef struct drm_radeon_indirect {
 #define RADEON_PARAM_CARD_TYPE             12
 #define RADEON_PARAM_VBLANK_CRTC           13   /* VBLANK CRTC */
 #define RADEON_PARAM_FB_LOCATION           14   /* FB location */
+#define RADEON_PARAM_NUM_GB_PIPES          15   /* num GB pipes */
 
 typedef struct drm_radeon_getparam {
 	int param;

drivers/char/drm/radeon_irq.c

@@ -234,7 +234,7 @@ int radeon_irq_wait(struct drm_device *dev, void *data, struct drm_file *file_pr
 	return radeon_wait_irq(dev, irqwait->irq_seq);
 }
 
-static void radeon_enable_interrupt(struct drm_device *dev)
+void radeon_enable_interrupt(struct drm_device *dev)
 {
 	drm_radeon_private_t *dev_priv = (drm_radeon_private_t *) dev->dev_private;
 

drivers/char/drm/radeon_state.c

@@ -1662,7 +1662,7 @@ static int radeon_cp_dispatch_texture(struct drm_device * dev,
 	u32 height;
 	int i;
 	u32 texpitch, microtile;
-	u32 offset;
+	u32 offset, byte_offset;
 	RING_LOCALS;
 
 	if (radeon_check_and_fixup_offset(dev_priv, file_priv, &tex->offset)) {
@@ -1727,6 +1727,13 @@ static int radeon_cp_dispatch_texture(struct drm_device * dev,
 	} else
 		microtile = 0;
 
+	/* this might fail for zero-sized uploads - are those illegal? */
+	if (!radeon_check_offset(dev_priv, tex->offset + image->height *
+				blit_width - 1)) {
+		DRM_ERROR("Invalid final destination offset\n");
+		return -EINVAL;
+	}
+
 	DRM_DEBUG("tex=%dx%d blit=%d\n", tex_width, tex->height, blit_width);
 
 	do {
@@ -1840,6 +1847,7 @@ static int radeon_cp_dispatch_texture(struct drm_device * dev,
 		}
 
 #undef RADEON_COPY_MT
+		byte_offset = (image->y & ~2047) * blit_width;
 		buf->file_priv = file_priv;
 		buf->used = size;
 		offset = dev_priv->gart_buffers_offset + buf->offset;
@@ -1854,9 +1862,9 @@ static int radeon_cp_dispatch_texture(struct drm_device * dev,
 			 RADEON_DP_SRC_SOURCE_MEMORY |
 			 RADEON_GMC_CLR_CMP_CNTL_DIS | RADEON_GMC_WR_MSK_DIS);
 		OUT_RING((spitch << 22) | (offset >> 10));
-		OUT_RING((texpitch << 22) | (tex->offset >> 10));
+		OUT_RING((texpitch << 22) | ((tex->offset >> 10) + (byte_offset >> 10)));
 		OUT_RING(0);
-		OUT_RING((image->x << 16) | image->y);
+		OUT_RING((image->x << 16) | (image->y % 2048));
 		OUT_RING((image->width << 16) | height);
 		RADEON_WAIT_UNTIL_2D_IDLE();
 		ADVANCE_RING();
@@ -3037,6 +3045,9 @@ static int radeon_cp_getparam(struct drm_device *dev, void *data, struct drm_fil
 	case RADEON_PARAM_FB_LOCATION:
 		value = radeon_read_fb_location(dev_priv);
 		break;
+	case RADEON_PARAM_NUM_GB_PIPES:
+		value = dev_priv->num_gb_pipes;
+		break;
 	default:
 		DRM_DEBUG("Invalid parameter %d\n", param->param);
 		return -EINVAL;