radeon_state.c

linux-2.6.15.6

}

static void radeon_apply_surface_regs(int surf_index,
				      drm_radeon_private_t * dev_priv)
{
	if (!dev_priv->mmio)
		return;

	radeon_do_cp_idle(dev_priv);

	RADEON_WRITE(RADEON_SURFACE0_INFO + 16 * surf_index,
		     dev_priv->surfaces[surf_index].flags);
	RADEON_WRITE(RADEON_SURFACE0_LOWER_BOUND + 16 * surf_index,
		     dev_priv->surfaces[surf_index].lower);
	RADEON_WRITE(RADEON_SURFACE0_UPPER_BOUND + 16 * surf_index,
		     dev_priv->surfaces[surf_index].upper);
}

/* Allocates a virtual surface
 * doesn't always allocate a real surface, will stretch an existing
 * surface when possible.
 *
 * Note that refcount can be at most 2, since during a free refcount=3
 * might mean we have to allocate a new surface which might not always
 * be available.
 * For example : we allocate three contigous surfaces ABC. If B is
 * freed, we suddenly need two surfaces to store A and C, which might
 * not always be available.
 */
static int alloc_surface(drm_radeon_surface_alloc_t * new,
			 drm_radeon_private_t * dev_priv, DRMFILE filp)
{
	struct radeon_virt_surface *s;
	int i;
	int virt_surface_index;
	uint32_t new_upper, new_lower;

	new_lower = new->address;
	new_upper = new_lower + new->size - 1;

	/* sanity check */
	if ((new_lower >= new_upper) || (new->flags == 0) || (new->size == 0) ||
	    ((new_upper & RADEON_SURF_ADDRESS_FIXED_MASK) !=
	     RADEON_SURF_ADDRESS_FIXED_MASK)
	    || ((new_lower & RADEON_SURF_ADDRESS_FIXED_MASK) != 0))
		return -1;

	/* make sure there is no overlap with existing surfaces */
	for (i = 0; i < RADEON_MAX_SURFACES; i++) {
		if ((dev_priv->surfaces[i].refcount != 0) &&
		    (((new_lower >= dev_priv->surfaces[i].lower) &&
		      (new_lower < dev_priv->surfaces[i].upper)) ||
		     ((new_lower < dev_priv->surfaces[i].lower) &&
		      (new_upper > dev_priv->surfaces[i].lower)))) {
			return -1;
		}
	}

	/* find a virtual surface */
	for (i = 0; i < 2 * RADEON_MAX_SURFACES; i++)
		if (dev_priv->virt_surfaces[i].filp == 0)
			break;
	if (i == 2 * RADEON_MAX_SURFACES) {
		return -1;
	}
	virt_surface_index = i;

	/* try to reuse an existing surface */
	for (i = 0; i < RADEON_MAX_SURFACES; i++) {
		/* extend before */
		if ((dev_priv->surfaces[i].refcount == 1) &&
		    (new->flags == dev_priv->surfaces[i].flags) &&
		    (new_upper + 1 == dev_priv->surfaces[i].lower)) {
			s = &(dev_priv->virt_surfaces[virt_surface_index]);
			s->surface_index = i;
			s->lower = new_lower;
			s->upper = new_upper;
			s->flags = new->flags;
			s->filp = filp;
			dev_priv->surfaces[i].refcount++;
			dev_priv->surfaces[i].lower = s->lower;
			radeon_apply_surface_regs(s->surface_index, dev_priv);
			return virt_surface_index;
		}

		/* extend after */
		if ((dev_priv->surfaces[i].refcount == 1) &&
		    (new->flags == dev_priv->surfaces[i].flags) &&
		    (new_lower == dev_priv->surfaces[i].upper + 1)) {
			s = &(dev_priv->virt_surfaces[virt_surface_index]);
			s->surface_index = i;
			s->lower = new_lower;
			s->upper = new_upper;
			s->flags = new->flags;
			s->filp = filp;
			dev_priv->surfaces[i].refcount++;
			dev_priv->surfaces[i].upper = s->upper;
			radeon_apply_surface_regs(s->surface_index, dev_priv);
			return virt_surface_index;
		}
	}

	/* okay, we need a new one */
	for (i = 0; i < RADEON_MAX_SURFACES; i++) {
		if (dev_priv->surfaces[i].refcount == 0) {
			s = &(dev_priv->virt_surfaces[virt_surface_index]);
			s->surface_index = i;
			s->lower = new_lower;
			s->upper = new_upper;
			s->flags = new->flags;
			s->filp = filp;
			dev_priv->surfaces[i].refcount = 1;
			dev_priv->surfaces[i].lower = s->lower;
			dev_priv->surfaces[i].upper = s->upper;
			dev_priv->surfaces[i].flags = s->flags;
			radeon_apply_surface_regs(s->surface_index, dev_priv);
			return virt_surface_index;
		}
	}

	/* we didn't find anything */
	return -1;
}

static int free_surface(DRMFILE filp, drm_radeon_private_t * dev_priv,
			int lower)
{
	struct radeon_virt_surface *s;
	int i;
	/* find the virtual surface */
	for (i = 0; i < 2 * RADEON_MAX_SURFACES; i++) {
		s = &(dev_priv->virt_surfaces[i]);
		if (s->filp) {
			if ((lower == s->lower) && (filp == s->filp)) {
				if (dev_priv->surfaces[s->surface_index].
				    lower == s->lower)
					dev_priv->surfaces[s->surface_index].
					    lower = s->upper;

				if (dev_priv->surfaces[s->surface_index].
				    upper == s->upper)
					dev_priv->surfaces[s->surface_index].
					    upper = s->lower;

				dev_priv->surfaces[s->surface_index].refcount--;
				if (dev_priv->surfaces[s->surface_index].
				    refcount == 0)
					dev_priv->surfaces[s->surface_index].
					    flags = 0;
				s->filp = NULL;
				radeon_apply_surface_regs(s->surface_index,
							  dev_priv);
				return 0;
			}
		}
	}
	return 1;
}

static void radeon_surfaces_release(DRMFILE filp,
				    drm_radeon_private_t * dev_priv)
{
	int i;
	for (i = 0; i < 2 * RADEON_MAX_SURFACES; i++) {
		if (dev_priv->virt_surfaces[i].filp == filp)
			free_surface(filp, dev_priv,
				     dev_priv->virt_surfaces[i].lower);
	}
}

/* ================================================================
 * IOCTL functions
 */
static int radeon_surface_alloc(DRM_IOCTL_ARGS)
{
	DRM_DEVICE;
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_radeon_surface_alloc_t alloc;

	if (!dev_priv) {
		DRM_ERROR("%s called with no initialization\n", __FUNCTION__);
		return DRM_ERR(EINVAL);
	}

	DRM_COPY_FROM_USER_IOCTL(alloc,
				 (drm_radeon_surface_alloc_t __user *) data,
				 sizeof(alloc));

	if (alloc_surface(&alloc, dev_priv, filp) == -1)
		return DRM_ERR(EINVAL);
	else
		return 0;
}

static int radeon_surface_free(DRM_IOCTL_ARGS)
{
	DRM_DEVICE;
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_radeon_surface_free_t memfree;

	if (!dev_priv) {
		DRM_ERROR("%s called with no initialization\n", __FUNCTION__);
		return DRM_ERR(EINVAL);
	}

	DRM_COPY_FROM_USER_IOCTL(memfree, (drm_radeon_mem_free_t __user *) data,
				 sizeof(memfree));

	if (free_surface(filp, dev_priv, memfree.address))
		return DRM_ERR(EINVAL);
	else
		return 0;
}

static int radeon_cp_clear(DRM_IOCTL_ARGS)
{
	DRM_DEVICE;
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
	drm_radeon_clear_t clear;
	drm_radeon_clear_rect_t depth_boxes[RADEON_NR_SAREA_CLIPRECTS];
	DRM_DEBUG("\n");

	LOCK_TEST_WITH_RETURN(dev, filp);

	DRM_COPY_FROM_USER_IOCTL(clear, (drm_radeon_clear_t __user *) data,
				 sizeof(clear));

	RING_SPACE_TEST_WITH_RETURN(dev_priv);

	if (sarea_priv->nbox > RADEON_NR_SAREA_CLIPRECTS)
		sarea_priv->nbox = RADEON_NR_SAREA_CLIPRECTS;

	if (DRM_COPY_FROM_USER(&depth_boxes, clear.depth_boxes,
			       sarea_priv->nbox * sizeof(depth_boxes[0])))
		return DRM_ERR(EFAULT);

	radeon_cp_dispatch_clear(dev, &clear, depth_boxes);

	COMMIT_RING();
	return 0;
}

/* Not sure why this isn't set all the time:
 */
static int radeon_do_init_pageflip(drm_device_t * dev)
{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	RING_LOCALS;

	DRM_DEBUG("\n");

	BEGIN_RING(6);
	RADEON_WAIT_UNTIL_3D_IDLE();
	OUT_RING(CP_PACKET0(RADEON_CRTC_OFFSET_CNTL, 0));
	OUT_RING(RADEON_READ(RADEON_CRTC_OFFSET_CNTL) |
		 RADEON_CRTC_OFFSET_FLIP_CNTL);
	OUT_RING(CP_PACKET0(RADEON_CRTC2_OFFSET_CNTL, 0));
	OUT_RING(RADEON_READ(RADEON_CRTC2_OFFSET_CNTL) |
		 RADEON_CRTC_OFFSET_FLIP_CNTL);
	ADVANCE_RING();

	dev_priv->page_flipping = 1;
	dev_priv->current_page = 0;
	dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page;

	return 0;
}

/* Called whenever a client dies, from drm_release.
 * NOTE:  Lock isn't necessarily held when this is called!
 */
static int radeon_do_cleanup_pageflip(drm_device_t * dev)
{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	DRM_DEBUG("\n");

	if (dev_priv->current_page != 0)
		radeon_cp_dispatch_flip(dev);

	dev_priv->page_flipping = 0;
	return 0;
}

/* Swapping and flipping are different operations, need different ioctls.
 * They can & should be intermixed to support multiple 3d windows.
 */
static int radeon_cp_flip(DRM_IOCTL_ARGS)
{
	DRM_DEVICE;
	drm_radeon_private_t *dev_priv = dev->dev_private;
	DRM_DEBUG("\n");

	LOCK_TEST_WITH_RETURN(dev, filp);

	RING_SPACE_TEST_WITH_RETURN(dev_priv);

	if (!dev_priv->page_flipping)
		radeon_do_init_pageflip(dev);

	radeon_cp_dispatch_flip(dev);

	COMMIT_RING();
	return 0;
}

static int radeon_cp_swap(DRM_IOCTL_ARGS)
{
	DRM_DEVICE;
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
	DRM_DEBUG("\n");

	LOCK_TEST_WITH_RETURN(dev, filp);

	RING_SPACE_TEST_WITH_RETURN(dev_priv);

	if (sarea_priv->nbox > RADEON_NR_SAREA_CLIPRECTS)
		sarea_priv->nbox = RADEON_NR_SAREA_CLIPRECTS;

	radeon_cp_dispatch_swap(dev);
	dev_priv->sarea_priv->ctx_owner = 0;

	COMMIT_RING();
	return 0;
}

static int radeon_cp_vertex(DRM_IOCTL_ARGS)
{
	DRM_DEVICE;
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_file_t *filp_priv;
	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
	drm_device_dma_t *dma = dev->dma;
	drm_buf_t *buf;
	drm_radeon_vertex_t vertex;
	drm_radeon_tcl_prim_t prim;

	LOCK_TEST_WITH_RETURN(dev, filp);

	DRM_GET_PRIV_WITH_RETURN(filp_priv, filp);

	DRM_COPY_FROM_USER_IOCTL(vertex, (drm_radeon_vertex_t __user *) data,
				 sizeof(vertex));

	DRM_DEBUG("pid=%d index=%d count=%d discard=%d\n",
		  DRM_CURRENTPID, vertex.idx, vertex.count, vertex.discard);

	if (vertex.idx < 0 || vertex.idx >= dma->buf_count) {
		DRM_ERROR("buffer index %d (of %d max)\n",
			  vertex.idx, dma->buf_count - 1);
		return DRM_ERR(EINVAL);
	}
	if (vertex.prim < 0 || vertex.prim > RADEON_PRIM_TYPE_3VRT_LINE_LIST) {
		DRM_ERROR("buffer prim %d\n", vertex.prim);
		return DRM_ERR(EINVAL);
	}

	RING_SPACE_TEST_WITH_RETURN(dev_priv);
	VB_AGE_TEST_WITH_RETURN(dev_priv);

	buf = dma->buflist[vertex.idx];

	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", vertex.idx);
		return DRM_ERR(EINVAL);
	}

	/* Build up a prim_t record:
	 */
	if (vertex.count) {
		buf->used = vertex.count;	/* not used? */

		if (sarea_priv->dirty & ~RADEON_UPLOAD_CLIPRECTS) {
			if (radeon_emit_state(dev_priv, filp_priv,
					      &sarea_priv->context_state,
					      sarea_priv->tex_state,
					      sarea_priv->dirty)) {
				DRM_ERROR("radeon_emit_state failed\n");
				return DRM_ERR(EINVAL);
			}

			sarea_priv->dirty &= ~(RADEON_UPLOAD_TEX0IMAGES |
					       RADEON_UPLOAD_TEX1IMAGES |
					       RADEON_UPLOAD_TEX2IMAGES |
					       RADEON_REQUIRE_QUIESCENCE);
		}

		prim.start = 0;
		prim.finish = vertex.count;	/* unused */
		prim.prim = vertex.prim;
		prim.numverts = vertex.count;
		prim.vc_format = dev_priv->sarea_priv->vc_format;

		radeon_cp_dispatch_vertex(dev, buf, &prim);
	}

	if (vertex.discard) {
		radeon_cp_discard_buffer(dev, buf);
	}

	COMMIT_RING();
	return 0;
}

static int radeon_cp_indices(DRM_IOCTL_ARGS)
{
	DRM_DEVICE;
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_file_t *filp_priv;
	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
	drm_device_dma_t *dma = dev->dma;
	drm_buf_t *buf;
	drm_radeon_indices_t elts;
	drm_radeon_tcl_prim_t prim;
	int count;

	LOCK_TEST_WITH_RETURN(dev, filp);

	if (!dev_priv) {
		DRM_ERROR("%s called with no initialization\n", __FUNCTION__);
		return DRM_ERR(EINVAL);
	}

	DRM_GET_PRIV_WITH_RETURN(filp_priv, filp);

	DRM_COPY_FROM_USER_IOCTL(elts, (drm_radeon_indices_t __user *) data,
				 sizeof(elts));

	DRM_DEBUG("pid=%d index=%d start=%d end=%d discard=%d\n",
		  DRM_CURRENTPID, elts.idx, elts.start, elts.end, elts.discard);

	if (elts.idx < 0 || elts.idx >= dma->buf_count) {
		DRM_ERROR("buffer index %d (of %d max)\n",
			  elts.idx, dma->buf_count - 1);
		return DRM_ERR(EINVAL);
	}
	if (elts.prim < 0 || elts.prim > RADEON_PRIM_TYPE_3VRT_LINE_LIST) {
		DRM_ERROR("buffer prim %d\n", elts.prim);
		return DRM_ERR(EINVAL);
	}

	RING_SPACE_TEST_WITH_RETURN(dev_priv);
	VB_AGE_TEST_WITH_RETURN(dev_priv);

	buf = dma->buflist[elts.idx];

	if (buf->fi