r128_state.c

linux-2.6.15.6

	drm_r128_buf_priv_t *buf_priv = buf->dev_private;
	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
	int format = sarea_priv->vc_format;
	int offset = buf->bus_address;
	int size = buf->used;
	int prim = buf_priv->prim;
	int i = 0;
	RING_LOCALS;
	DRM_DEBUG("buf=%d nbox=%d\n", buf->idx, sarea_priv->nbox);

	if (0)
		r128_print_dirty("dispatch_vertex", sarea_priv->dirty);

	if (buf->used) {
		buf_priv->dispatched = 1;

		if (sarea_priv->dirty & ~R128_UPLOAD_CLIPRECTS) {
			r128_emit_state(dev_priv);
		}

		do {
			/* Emit the next set of up to three cliprects */
			if (i < sarea_priv->nbox) {
				r128_emit_clip_rects(dev_priv,
						     &sarea_priv->boxes[i],
						     sarea_priv->nbox - i);
			}

			/* Emit the vertex buffer rendering commands */
			BEGIN_RING(5);

			OUT_RING(CCE_PACKET3(R128_3D_RNDR_GEN_INDX_PRIM, 3));
			OUT_RING(offset);
			OUT_RING(size);
			OUT_RING(format);
			OUT_RING(prim | R128_CCE_VC_CNTL_PRIM_WALK_LIST |
				 (size << R128_CCE_VC_CNTL_NUM_SHIFT));

			ADVANCE_RING();

			i += 3;
		} while (i < sarea_priv->nbox);
	}

	if (buf_priv->discard) {
		buf_priv->age = dev_priv->sarea_priv->last_dispatch;

		/* Emit the vertex buffer age */
		BEGIN_RING(2);

		OUT_RING(CCE_PACKET0(R128_LAST_DISPATCH_REG, 0));
		OUT_RING(buf_priv->age);

		ADVANCE_RING();

		buf->pending = 1;
		buf->used = 0;
		/* FIXME: Check dispatched field */
		buf_priv->dispatched = 0;
	}

	dev_priv->sarea_priv->last_dispatch++;

	sarea_priv->dirty &= ~R128_UPLOAD_CLIPRECTS;
	sarea_priv->nbox = 0;
}

static void r128_cce_dispatch_indirect(drm_device_t * dev,
				       drm_buf_t * buf, int start, int end)
{
	drm_r128_private_t *dev_priv = dev->dev_private;
	drm_r128_buf_priv_t *buf_priv = buf->dev_private;
	RING_LOCALS;
	DRM_DEBUG("indirect: buf=%d s=0x%x e=0x%x\n", buf->idx, start, end);

	if (start != end) {
		int offset = buf->bus_address + start;
		int dwords = (end - start + 3) / sizeof(u32);

		/* Indirect buffer data must be an even number of
		 * dwords, so if we've been given an odd number we must
		 * pad the data with a Type-2 CCE packet.
		 */
		if (dwords & 1) {
			u32 *data = (u32 *)
			    ((char *)dev->agp_buffer_map->handle
			     + buf->offset + start);
			data[dwords++] = cpu_to_le32(R128_CCE_PACKET2);
		}

		buf_priv->dispatched = 1;

		/* Fire off the indirect buffer */
		BEGIN_RING(3);

		OUT_RING(CCE_PACKET0(R128_PM4_IW_INDOFF, 1));
		OUT_RING(offset);
		OUT_RING(dwords);

		ADVANCE_RING();
	}

	if (buf_priv->discard) {
		buf_priv->age = dev_priv->sarea_priv->last_dispatch;

		/* Emit the indirect buffer age */
		BEGIN_RING(2);

		OUT_RING(CCE_PACKET0(R128_LAST_DISPATCH_REG, 0));
		OUT_RING(buf_priv->age);

		ADVANCE_RING();

		buf->pending = 1;
		buf->used = 0;
		/* FIXME: Check dispatched field */
		buf_priv->dispatched = 0;
	}

	dev_priv->sarea_priv->last_dispatch++;
}

static void r128_cce_dispatch_indices(drm_device_t * dev,
				      drm_buf_t * buf,
				      int start, int end, int count)
{
	drm_r128_private_t *dev_priv = dev->dev_private;
	drm_r128_buf_priv_t *buf_priv = buf->dev_private;
	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
	int format = sarea_priv->vc_format;
	int offset = dev->agp_buffer_map->offset - dev_priv->cce_buffers_offset;
	int prim = buf_priv->prim;
	u32 *data;
	int dwords;
	int i = 0;
	RING_LOCALS;
	DRM_DEBUG("indices: s=%d e=%d c=%d\n", start, end, count);

	if (0)
		r128_print_dirty("dispatch_indices", sarea_priv->dirty);

	if (start != end) {
		buf_priv->dispatched = 1;

		if (sarea_priv->dirty & ~R128_UPLOAD_CLIPRECTS) {
			r128_emit_state(dev_priv);
		}

		dwords = (end - start + 3) / sizeof(u32);

		data = (u32 *) ((char *)dev->agp_buffer_map->handle
				+ buf->offset + start);

		data[0] = cpu_to_le32(CCE_PACKET3(R128_3D_RNDR_GEN_INDX_PRIM,
						  dwords - 2));

		data[1] = cpu_to_le32(offset);
		data[2] = cpu_to_le32(R128_MAX_VB_VERTS);
		data[3] = cpu_to_le32(format);
		data[4] = cpu_to_le32((prim | R128_CCE_VC_CNTL_PRIM_WALK_IND |
				       (count << 16)));

		if (count & 0x1) {
#ifdef __LITTLE_ENDIAN
			data[dwords - 1] &= 0x0000ffff;
#else
			data[dwords - 1] &= 0xffff0000;
#endif
		}

		do {
			/* Emit the next set of up to three cliprects */
			if (i < sarea_priv->nbox) {
				r128_emit_clip_rects(dev_priv,
						     &sarea_priv->boxes[i],
						     sarea_priv->nbox - i);
			}

			r128_cce_dispatch_indirect(dev, buf, start, end);

			i += 3;
		} while (i < sarea_priv->nbox);
	}

	if (buf_priv->discard) {
		buf_priv->age = dev_priv->sarea_priv->last_dispatch;

		/* Emit the vertex buffer age */
		BEGIN_RING(2);

		OUT_RING(CCE_PACKET0(R128_LAST_DISPATCH_REG, 0));
		OUT_RING(buf_priv->age);

		ADVANCE_RING();

		buf->pending = 1;
		/* FIXME: Check dispatched field */
		buf_priv->dispatched = 0;
	}

	dev_priv->sarea_priv->last_dispatch++;

	sarea_priv->dirty &= ~R128_UPLOAD_CLIPRECTS;
	sarea_priv->nbox = 0;
}

static int r128_cce_dispatch_blit(DRMFILE filp,
				  drm_device_t * dev, drm_r128_blit_t * blit)
{
	drm_r128_private_t *dev_priv = dev->dev_private;
	drm_device_dma_t *dma = dev->dma;
	drm_buf_t *buf;
	drm_r128_buf_priv_t *buf_priv;
	u32 *data;
	int dword_shift, dwords;
	RING_LOCALS;
	DRM_DEBUG("\n");

	/* The compiler won't optimize away a division by a variable,
	 * even if the only legal values are powers of two.  Thus, we'll
	 * use a shift instead.
	 */
	switch (blit->format) {
	case R128_DATATYPE_ARGB8888:
		dword_shift = 0;
		break;
	case R128_DATATYPE_ARGB1555:
	case R128_DATATYPE_RGB565:
	case R128_DATATYPE_ARGB4444:
	case R128_DATATYPE_YVYU422:
	case R128_DATATYPE_VYUY422:
		dword_shift = 1;
		break;
	case R128_DATATYPE_CI8:
	case R128_DATATYPE_RGB8:
		dword_shift = 2;
		break;
	default:
		DRM_ERROR("invalid blit format %d\n", blit->format);
		return DRM_ERR(EINVAL);
	}

	/* Flush the pixel cache, and mark the contents as Read Invalid.
	 * This ensures no pixel data gets mixed up with the texture
	 * data from the host data blit, otherwise part of the texture
	 * image may be corrupted.
	 */
	BEGIN_RING(2);

	OUT_RING(CCE_PACKET0(R128_PC_GUI_CTLSTAT, 0));
	OUT_RING(R128_PC_RI_GUI | R128_PC_FLUSH_GUI);

	ADVANCE_RING();

	/* Dispatch the indirect buffer.
	 */
	buf = dma->buflist[blit->idx];
	buf_priv = buf->dev_private;

	if (buf->filp != filp) {
		DRM_ERROR("process %d using buffer owned by %p\n",
			  DRM_CURRENTPID, buf->filp);
		return DRM_ERR(EINVAL);
	}
	if (buf->pending) {
		DRM_ERROR("sending pending buffer %d\n", blit->idx);
		return DRM_ERR(EINVAL);
	}

	buf_priv->discard = 1;

	dwords = (blit->width * blit->height) >> dword_shift;

	data = (u32 *) ((char *)dev->agp_buffer_map->handle + buf->offset);

	data[0] = cpu_to_le32(CCE_PACKET3(R128_CNTL_HOSTDATA_BLT, dwords + 6));
	data[1] = cpu_to_le32((R128_GMC_DST_PITCH_OFFSET_CNTL |
			       R128_GMC_BRUSH_NONE |
			       (blit->format << 8) |
			       R128_GMC_SRC_DATATYPE_COLOR |
			       R128_ROP3_S |
			       R128_DP_SRC_SOURCE_HOST_DATA |
			       R128_GMC_CLR_CMP_CNTL_DIS |
			       R128_GMC_AUX_CLIP_DIS | R128_GMC_WR_MSK_DIS));

	data[2] = cpu_to_le32((blit->pitch << 21) | (blit->offset >> 5));
	data[3] = cpu_to_le32(0xffffffff);
	data[4] = cpu_to_le32(0xffffffff);
	data[5] = cpu_to_le32((blit->y << 16) | blit->x);
	data[6] = cpu_to_le32((blit->height << 16) | blit->width);
	data[7] = cpu_to_le32(dwords);

	buf->used = (dwords + 8) * sizeof(u32);

	r128_cce_dispatch_indirect(dev, buf, 0, buf->used);

	/* Flush the pixel cache after the blit completes.  This ensures
	 * the texture data is written out to memory before rendering
	 * continues.
	 */
	BEGIN_RING(2);

	OUT_RING(CCE_PACKET0(R128_PC_GUI_CTLSTAT, 0));
	OUT_RING(R128_PC_FLUSH_GUI);

	ADVANCE_RING();

	return 0;
}

/* ================================================================
 * Tiled depth buffer management
 *
 * FIXME: These should all set the destination write mask for when we
 * have hardware stencil support.
 */

static int r128_cce_dispatch_write_span(drm_device_t * dev,
					drm_r128_depth_t * depth)
{
	drm_r128_private_t *dev_priv = dev->dev_private;
	int count, x, y;
	u32 *buffer;
	u8 *mask;
	int i, buffer_size, mask_size;
	RING_LOCALS;
	DRM_DEBUG("\n");

	count = depth->n;
	if (count > 4096 || count <= 0)
		return DRM_ERR(EMSGSIZE);

	if (DRM_COPY_FROM_USER(&x, depth->x, sizeof(x))) {
		return DRM_ERR(EFAULT);
	}
	if (DRM_COPY_FROM_USER(&y, depth->y, sizeof(y))) {
		return DRM_ERR(EFAULT);
	}

	buffer_size = depth->n * sizeof(u32);
	buffer = drm_alloc(buffer_size, DRM_MEM_BUFS);
	if (buffer == NULL)
		return DRM_ERR(ENOMEM);
	if (DRM_COPY_FROM_USER(buffer, depth->buffer, buffer_size)) {
		drm_free(buffer, buffer_size, DRM_MEM_BUFS);
		return DRM_ERR(EFAULT);
	}

	mask_size = depth->n * sizeof(u8);
	if (depth->mask) {
		mask = drm_alloc(mask_size, DRM_MEM_BUFS);
		if (mask == NULL) {
			drm_free(buffer, buffer_size, DRM_MEM_BUFS);
			return DRM_ERR(ENOMEM);
		}
		if (DRM_COPY_FROM_USER(mask, depth->mask, mask_size)) {
			drm_free(buffer, buffer_size, DRM_MEM_BUFS);
			drm_free(mask, mask_size, DRM_MEM_BUFS);
			return DRM_ERR(EFAULT);
		}

		for (i = 0; i < count; i++, x++) {
			if (mask[i]) {
				BEGIN_RING(6);

				OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
				OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
					 R128_GMC_BRUSH_SOLID_COLOR |
					 (dev_priv->depth_fmt << 8) |
					 R128_GMC_SRC_DATATYPE_COLOR |
					 R128_ROP3_P |
					 R128_GMC_CLR_CMP_CNTL_DIS |
					 R128_GMC_WR_MSK_DIS);

				OUT_RING(dev_priv->depth_pitch_offset_c);
				OUT_RING(buffer[i]);

				OUT_RING((x << 16) | y);
				OUT_RING((1 << 16) | 1);

				ADVANCE_RING();
			}
		}

		drm_free(mask, mask_size, DRM_MEM_BUFS);
	} else {
		for (i = 0; i < count; i++, x++) {
			BEGIN_RING(6);

			OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
			OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
				 R128_GMC_BRUSH_SOLID_COLOR |
				 (dev_priv->depth_fmt << 8) |
				 R128_GMC_SRC_DATATYPE_COLOR |
				 R128_ROP3_P |
				 R128_GMC_CLR_CMP_CNTL_DIS |
				 R128_GMC_WR_MSK_DIS);

			OUT_RING(dev_priv->depth_pitch_offset_c);
			OUT_RING(buffer[i]);

			OUT_RING((x << 16) | y);
			OUT_RING((1 << 16) | 1);

			ADVANCE_RING();
		}
	}

	drm_free(buffer, buffer_size, DRM_MEM_BUFS);

	return 0;
}

static int r128_cce_dispatch_write_pixels(drm_device_t * dev,
					  drm_r128_depth_t * depth)
{
	drm_r128_private_t *dev_priv = dev->dev_private;
	int count, *x, *y;
	u32 *buffer;
	u8 *mask;
	int i, xbuf_size, ybuf_size, buffer_size, mask_size;
	RING_LOCALS;
	DRM_DEBUG("\n");

	count = depth->n;
	if (count > 4096 || count <= 0)
		return DRM_ERR(EMSGSIZE);

	xbuf_size = count * sizeof(*x);
	ybuf_size = count * sizeof(*y);
	x = drm_alloc(xbuf_size, DRM_MEM_BUFS);
	if (x == NULL) {
		return DRM_ERR(ENOMEM);
	}
	y = drm_alloc(ybuf_size, DRM_MEM_BUFS);
	if (y == NULL) {
		drm_free(x, xbuf_size, DRM_MEM_BUFS);
		return DRM_ERR(ENOMEM);
	}
	if (DRM_COPY_FROM_USER(x, depth->x, xbuf_size)) {
		drm_free(x, xbuf_size, DRM_MEM_BUFS);
		drm_free(y, ybuf_size, DRM_MEM_BUFS);
		return DRM_ERR(EFAULT);
	}
	if (DRM_COPY_FROM_USER(y, depth->y, xbuf_size)) {
		drm_free(x, xbuf_size, DRM_MEM_BUFS);
		drm_free(y, ybuf_size, DRM_MEM_BUFS);
		return DRM_ERR(EFAULT);
	}

	buffer_size = depth->n * sizeof(u32);
	buffer = drm_alloc(buffer_size, DRM_MEM_BUFS);
	if (buffer == NULL) {
		drm_free(x, xbuf_size, DRM_MEM_BUFS);
		drm_free(y, ybuf_size, DRM_MEM_BUFS);
		return DRM_ERR(ENOMEM);
	}
	if (DRM_COPY_FROM_USER(buffer, depth->buffer, buffer_size)) {
		drm_free(x, xbuf_size, DRM_MEM_BUFS);
		drm_free(y, ybuf_size, DRM_MEM_BUFS);
		drm_free(buffer, buffer_size, DRM_MEM_BUFS);
		return DRM_ERR(EFAULT);
	}

	if (depth->mask) {
		mask_size = depth->n * sizeof(u8);
		mask = drm_alloc(mask_size, DRM_MEM_BUFS);
		if (mask == NULL) {
			drm_free(x, xbuf_size, DRM_MEM_BUFS);
			drm_free(y, ybuf_size, DRM_MEM_BUFS);
			drm_free(buffer, buffer_size, DRM_MEM_BUFS);
			return DRM_ERR(ENOMEM);
		}
		if (DRM_COPY_FROM_USER(mask, depth->mask, mask_size)) {
			drm_free(x, xbuf_size, DRM_MEM_BUFS);
			drm_free(y, ybuf_size, DRM_MEM_BUFS);
			drm_free(buffer, buffer_size, DRM_MEM_BUFS);
			drm_free(mask, mask_size, DRM_MEM_BUFS);
			return DRM_ERR(EFAULT);
		}

		for (i = 0; i < count; i++) {
			if (mask[i]) {
				BEGIN_RING(6);

				OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
				OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
					 R128_GMC_BRUSH_SOLID_COLOR |
					 (dev_priv->depth_fmt << 8) |
					 R128_GMC_SRC_DATATYPE_COLOR |
					 R128_ROP3_P |
					 R128_GMC_CLR_CMP_CNTL_DIS |
					 R128_GMC_WR_MSK_DIS);

				OUT_RING(dev_priv->depth_pitch_offset_c);
				OUT_RING(buffer[i]);

				OUT_RING((x[i] << 16) | y[i]);
				OUT_RING((1 << 16) | 1);

				ADVANCE_RING();
			}
		}

		drm_free(mask, mask_size, DRM_MEM_BUFS);
	} else {
		for (i = 0; i < count; i++) {
			BEGIN_RING(6);

			OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
			OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
				 R128_GMC_BRUSH_SOLID_COLOR |
				 (dev_priv->depth_fmt << 8) |
				 R128_GMC_SRC_DATATYPE_COLOR |
				 R128_ROP3_P |
				 R128_GMC_CLR_CMP_CNTL_DIS |
				 R128_GMC_WR_MSK_DIS);

			OUT_RING(dev_priv->depth_pitch_offset_c);
			OUT_RING(buffer[i]);

			OUT_RING((x[i] << 16) | y[i]);
			OUT_RING((1 << 16) | 1);

			ADVANCE_RING();
		}
	}

	drm_free(x, xbuf_size, DRM_MEM_BUFS);
	drm_free(y, ybuf_size, DRM_MEM_BUFS);
	drm_free(buffer, buffer_size, DRM_MEM_BUFS);

	return 0;
}

static int r128_cce_dispatch_read_span(drm_device_t * dev,
				       drm_r128_depth_t * depth)
{
	drm_r128_private_t *dev_priv = dev->dev_private;
	int count, x, y;
	RING_LOCALS;
	DRM_DEBUG("\n");

	count = depth->n;
	if (count > 4096 || count <= 0)
		return DRM_ERR(EMSGSIZE);

	if (DRM_COPY_FROM_USER(&x, depth->x, sizeof(x))) {
		return DRM_ERR(EFAULT);
	}
	if (DRM_COPY_FROM_USER(&y, depth->y, sizeof(y))) {
		return DRM_ERR(EFAULT);
	}

	BEGIN_RING(7);

	OUT_RING(CCE_PACKET3(R128_CNTL_BITBLT_MULTI, 5));
	OUT_RING(R128_GMC_SRC_PITCH_OFFSET_CNTL |
		 R128_GMC_DST_PITCH_OFFSET_CNTL |
		 R128_GMC_BRUSH_NONE |
		 (dev_priv->depth_fmt << 8) |
		 R128_GMC_SRC_DATATYPE_COLOR |
		 R128_ROP3_S |
		 R128_DP_SRC_SOURCE_MEMORY |
		 R128_GMC_CLR_CMP_CNTL_DIS | R128_GMC_WR_MSK_DIS);

	OUT_RING(dev_priv->depth_pitch_offset_c);
	OUT_RING(dev_priv->span_pitch_offset_c);

	OUT_RING((x << 16) | y);
	OUT_RING((0 << 16) | 0);