On Tue, 4 Sep 2012 21:02:53 +0100
Chris Wilson <chris at chris-wilson.co.uk> wrote:
> In order to specialise functions depending upon the type of object, we
> can attach vfuncs to each object via a new ->ops pointer.
>
> For instance, this will be used in future patches to only bind pages from
> a dma-buf for the duration that the object is used by the GPU - and so
> prevent them from pinning those pages for the entire of the object.
>
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_drv.h | 12 +++++-
> drivers/gpu/drm/i915/i915_gem.c | 71 +++++++++++++++++++++-----------
> drivers/gpu/drm/i915/i915_gem_dmabuf.c | 4 +-
> 3 files changed, 60 insertions(+), 27 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index f16ab5e..f180874 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -898,9 +898,16 @@ enum i915_cache_level {
> I915_CACHE_LLC_MLC, /* gen6+, in docs at least! */
> };
>
> +struct drm_i915_gem_object_ops {
> + int (*get_pages)(struct drm_i915_gem_object *);
> + void (*put_pages)(struct drm_i915_gem_object *);
> +};
> +
> struct drm_i915_gem_object {
> struct drm_gem_object base;
>
> + const struct drm_i915_gem_object_ops *ops;
> +
> /** Current space allocated to this object in the GTT, if any. */
> struct drm_mm_node *gtt_space;
> struct list_head gtt_list;
> @@ -1305,7 +1312,8 @@ int i915_gem_wait_ioctl(struct drm_device *dev, void *data,
> struct drm_file *file_priv);
> void i915_gem_load(struct drm_device *dev);
> int i915_gem_init_object(struct drm_gem_object *obj);
> -void i915_gem_object_init(struct drm_i915_gem_object *obj);
> +void i915_gem_object_init(struct drm_i915_gem_object *obj,
> + const struct drm_i915_gem_object_ops *ops);
> struct drm_i915_gem_object *i915_gem_alloc_object(struct drm_device *dev,
> size_t size);
> void i915_gem_free_object(struct drm_gem_object *obj);
> @@ -1318,7 +1326,7 @@ int __must_check i915_gem_object_unbind(struct drm_i915_gem_object *obj);
> void i915_gem_release_mmap(struct drm_i915_gem_object *obj);
> void i915_gem_lastclose(struct drm_device *dev);
>
> -int __must_check i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj);
> +int __must_check i915_gem_object_get_pages(struct drm_i915_gem_object *obj);
> int __must_check i915_mutex_lock_interruptible(struct drm_device *dev);
> int i915_gem_object_sync(struct drm_i915_gem_object *obj,
> struct intel_ring_buffer *to);
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 87a64e5..66fbd9f 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1650,18 +1650,12 @@ i915_gem_object_is_purgeable(struct drm_i915_gem_object *obj)
> return obj->madv == I915_MADV_DONTNEED;
> }
>
> -static int
> +static void
> i915_gem_object_put_pages_gtt(struct drm_i915_gem_object *obj)
> {
> int page_count = obj->base.size / PAGE_SIZE;
> int ret, i;
>
> - BUG_ON(obj->gtt_space);
> -
> - if (obj->pages == NULL)
> - return 0;
> -
> - BUG_ON(obj->gtt_space);
> BUG_ON(obj->madv == __I915_MADV_PURGED);
>
> ret = i915_gem_object_set_to_cpu_domain(obj, true);
> @@ -1693,9 +1687,21 @@ i915_gem_object_put_pages_gtt(struct drm_i915_gem_object *obj)
>
> drm_free_large(obj->pages);
> obj->pages = NULL;
> +}
>
> - list_del(&obj->gtt_list);
> +static int
> +i915_gem_object_put_pages(struct drm_i915_gem_object *obj)
> +{
> + const struct drm_i915_gem_object_ops *ops = obj->ops;
> +
> + if (obj->sg_table || obj->pages == NULL)
> + return 0;
> +
> + BUG_ON(obj->gtt_space);
>
> + ops->put_pages(obj);
> +
> + list_del(&obj->gtt_list);
> if (i915_gem_object_is_purgeable(obj))
> i915_gem_object_truncate(obj);
>
> @@ -1712,7 +1718,7 @@ i915_gem_purge(struct drm_i915_private *dev_priv, long target)
> &dev_priv->mm.unbound_list,
> gtt_list) {
> if (i915_gem_object_is_purgeable(obj) &&
> - i915_gem_object_put_pages_gtt(obj) == 0) {
> + i915_gem_object_put_pages(obj) == 0) {
> count += obj->base.size >> PAGE_SHIFT;
> if (count >= target)
> return count;
> @@ -1724,7 +1730,7 @@ i915_gem_purge(struct drm_i915_private *dev_priv, long target)
> mm_list) {
> if (i915_gem_object_is_purgeable(obj) &&
> i915_gem_object_unbind(obj) == 0 &&
> - i915_gem_object_put_pages_gtt(obj) == 0) {
> + i915_gem_object_put_pages(obj) == 0) {
> count += obj->base.size >> PAGE_SHIFT;
> if (count >= target)
> return count;
> @@ -1742,10 +1748,10 @@ i915_gem_shrink_all(struct drm_i915_private *dev_priv)
> i915_gem_evict_everything(dev_priv->dev);
>
> list_for_each_entry_safe(obj, next, &dev_priv->mm.unbound_list, gtt_list)
> - i915_gem_object_put_pages_gtt(obj);
> + i915_gem_object_put_pages(obj);
> }
>
> -int
> +static int
> i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
> {
> struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
> @@ -1754,9 +1760,6 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
> struct page *page;
> gfp_t gfp;
>
> - if (obj->pages || obj->sg_table)
> - return 0;
> -
> /* Assert that the object is not currently in any GPU domain. As it
> * wasn't in the GTT, there shouldn't be any way it could have been in
> * a GPU cache
> @@ -1806,7 +1809,6 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
> if (i915_gem_object_needs_bit17_swizzle(obj))
> i915_gem_object_do_bit_17_swizzle(obj);
>
> - list_add_tail(&obj->gtt_list, &dev_priv->mm.unbound_list);
> return 0;
>
> err_pages:
> @@ -1818,6 +1820,24 @@ err_pages:
> return PTR_ERR(page);
> }
>
> +int
> +i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
> +{
> + struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
> + const struct drm_i915_gem_object_ops *ops = obj->ops;
> + int ret;
> +
> + if (obj->sg_table || obj->pages)
> + return 0;
> +
> + ret = ops->get_pages(obj);
> + if (ret)
> + return ret;
> +
> + list_add_tail(&obj->gtt_list, &dev_priv->mm.unbound_list);
> + return 0;
> +}
> +
> void
> i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,
> struct intel_ring_buffer *ring,
> @@ -2071,7 +2091,6 @@ void i915_gem_reset(struct drm_device *dev)
> obj->base.read_domains &= ~I915_GEM_GPU_DOMAINS;
> }
>
> -
> /* The fence registers are invalidated so clear them out */
> i915_gem_reset_fences(dev);
> }
> @@ -2871,7 +2890,7 @@ i915_gem_object_bind_to_gtt(struct drm_i915_gem_object *obj,
> return -E2BIG;
> }
>
> - ret = i915_gem_object_get_pages_gtt(obj);
> + ret = i915_gem_object_get_pages(obj);
> if (ret)
> return ret;
>
> @@ -3610,15 +3629,16 @@ unlock:
> return ret;
> }
>
> -void i915_gem_object_init(struct drm_i915_gem_object *obj)
> +void i915_gem_object_init(struct drm_i915_gem_object *obj,
> + const struct drm_i915_gem_object_ops *ops)
> {
> - obj->base.driver_private = NULL;
> -
> INIT_LIST_HEAD(&obj->mm_list);
> INIT_LIST_HEAD(&obj->gtt_list);
> INIT_LIST_HEAD(&obj->ring_list);
> INIT_LIST_HEAD(&obj->exec_list);
>
> + obj->ops = ops;
> +
> obj->fence_reg = I915_FENCE_REG_NONE;
> obj->madv = I915_MADV_WILLNEED;
> /* Avoid an unnecessary call to unbind on the first bind. */
> @@ -3627,6 +3647,11 @@ void i915_gem_object_init(struct drm_i915_gem_object *obj)
> i915_gem_info_add_obj(obj->base.dev->dev_private, obj->base.size);
> }
>
> +static const struct drm_i915_gem_object_ops i915_gem_object_ops = {
> + .get_pages = i915_gem_object_get_pages_gtt,
> + .put_pages = i915_gem_object_put_pages_gtt,
> +};
> +
> struct drm_i915_gem_object *i915_gem_alloc_object(struct drm_device *dev,
> size_t size)
> {
> @@ -3653,7 +3678,7 @@ struct drm_i915_gem_object *i915_gem_alloc_object(struct drm_device *dev,
> mapping = obj->base.filp->f_path.dentry->d_inode->i_mapping;
> mapping_set_gfp_mask(mapping, mask);
>
> - i915_gem_object_init(obj);
> + i915_gem_object_init(obj, &i915_gem_object_ops);
>
> obj->base.write_domain = I915_GEM_DOMAIN_CPU;
> obj->base.read_domains = I915_GEM_DOMAIN_CPU;
> @@ -3711,7 +3736,7 @@ void i915_gem_free_object(struct drm_gem_object *gem_obj)
> dev_priv->mm.interruptible = was_interruptible;
> }
>
> - i915_gem_object_put_pages_gtt(obj);
> + i915_gem_object_put_pages(obj);
> i915_gem_object_free_mmap_offset(obj);
>
> drm_gem_object_release(&obj->base);
> diff --git a/drivers/gpu/drm/i915/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/i915_gem_dmabuf.c
> index 43c9530..e4f1141 100644
> --- a/drivers/gpu/drm/i915/i915_gem_dmabuf.c
> +++ b/drivers/gpu/drm/i915/i915_gem_dmabuf.c
> @@ -41,7 +41,7 @@ static struct sg_table *i915_gem_map_dma_buf(struct dma_buf_attachment *attachme
> if (ret)
> return ERR_PTR(ret);
>
> - ret = i915_gem_object_get_pages_gtt(obj);
> + ret = i915_gem_object_get_pages(obj);
> if (ret) {
> sg = ERR_PTR(ret);
> goto out;
> @@ -89,7 +89,7 @@ static void *i915_gem_dmabuf_vmap(struct dma_buf *dma_buf)
> goto out_unlock;
> }
>
> - ret = i915_gem_object_get_pages_gtt(obj);
> + ret = i915_gem_object_get_pages(obj);
> if (ret) {
> mutex_unlock(&dev->struct_mutex);
> return ERR_PTR(ret);
Ben's comments are good ones, some kdoc would be nice. Anyway:
Reviewed-by: Jesse Barnes <jbarnes at virtuousgeek.org>
--
Jesse Barnes, Intel Open Source Technology Center
