[PATCH 06/23] drm/i915: introduce page_size members

[Matthew Auld <matthew.auld@intel.com>]

In preparation for supporting huge gtt pages for the ppgtt, we introduce
page size members for gem objects.  We fill in the page sizes by
scanning the sg table.

v2: pass the sg_mask to set_pages

v3: calculate the sg_mask inline with populating the sg_table where
possible, and pass to set_pages along with the pages.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel@ffwll.ch>
---
 drivers/gpu/drm/i915/i915_drv.h                  |  5 +++-
 drivers/gpu/drm/i915/i915_gem.c                  | 35 ++++++++++++++++++++----
 drivers/gpu/drm/i915/i915_gem_dmabuf.c           |  8 +++++-
 drivers/gpu/drm/i915/i915_gem_internal.c         |  5 +++-
 drivers/gpu/drm/i915/i915_gem_object.h           | 17 ++++++++++++
 drivers/gpu/drm/i915/i915_gem_stolen.c           |  2 +-
 drivers/gpu/drm/i915/i915_gem_userptr.c          | 13 +++++++--
 drivers/gpu/drm/i915/selftests/huge_gem_object.c |  2 +-
 drivers/gpu/drm/i915/selftests/i915_gem_gtt.c    |  4 ++-
 9 files changed, 77 insertions(+), 14 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 224f8a8ae317..202adc992e08 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3019,6 +3019,8 @@ intel_info(const struct drm_i915_private *dev_priv)
 #define USES_PPGTT(dev_priv)		(i915.enable_ppgtt)
 #define USES_FULL_PPGTT(dev_priv)	(i915.enable_ppgtt >= 2)
 #define USES_FULL_48BIT_PPGTT(dev_priv)	(i915.enable_ppgtt == 3)
+#define HAS_PAGE_SIZE(dev_priv, page_size) \
+	((dev_priv)->info.page_size_mask & (page_size))
 
 #define HAS_OVERLAY(dev_priv)		 ((dev_priv)->info.has_overlay)
 #define OVERLAY_NEEDS_PHYSICAL(dev_priv) \
@@ -3428,7 +3430,8 @@ i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj,
 				unsigned long n);
 
 void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
-				 struct sg_table *pages);
+				 struct sg_table *pages,
+				 unsigned int sg_mask);
 int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj);
 
 static inline int __must_check
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 1c61d1b5eaa5..af3b9cfcc1fe 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -228,7 +228,7 @@ static int i915_gem_object_get_pages_phys(struct drm_i915_gem_object *obj)
 
 	obj->phys_handle = phys;
 
-	__i915_gem_object_set_pages(obj, st);
+	__i915_gem_object_set_pages(obj, st, sg->length);
 
 	return 0;
 
@@ -2264,6 +2264,8 @@ void __i915_gem_object_put_pages(struct drm_i915_gem_object *obj,
 	if (!IS_ERR(pages))
 		obj->ops->put_pages(obj, pages);
 
+	obj->mm.page_sizes.phys = obj->mm.page_sizes.sg = 0;
+
 unlock:
 	mutex_unlock(&obj->mm.lock);
 }
@@ -2306,6 +2308,7 @@ static int i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
 	struct page *page;
 	unsigned long last_pfn = 0;	/* suppress gcc warning */
 	unsigned int max_segment;
+	unsigned int sg_mask;
 	gfp_t noreclaim;
 	int ret;
 
@@ -2341,6 +2344,7 @@ static int i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
 
 	sg = st->sgl;
 	st->nents = 0;
+	sg_mask = 0;
 	for (i = 0; i < page_count; i++) {
 		const unsigned int shrink[] = {
 			I915_SHRINK_BOUND | I915_SHRINK_UNBOUND | I915_SHRINK_PURGEABLE,
@@ -2393,8 +2397,10 @@ static int i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
 		if (!i ||
 		    sg->length >= max_segment ||
 		    page_to_pfn(page) != last_pfn + 1) {
-			if (i)
+			if (i) {
+				sg_mask |= sg->length;
 				sg = sg_next(sg);
+			}
 			st->nents++;
 			sg_set_page(sg, page, PAGE_SIZE, 0);
 		} else {
@@ -2405,8 +2411,10 @@ static int i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
 		/* Check that the i965g/gm workaround works. */
 		WARN_ON((gfp & __GFP_DMA32) && (last_pfn >= 0x00100000UL));
 	}
-	if (sg) /* loop terminated early; short sg table */
+	if (sg) { /* loop terminated early; short sg table */
+		sg_mask |= sg->length;
 		sg_mark_end(sg);
+	}
 
 	/* Trim unused sg entries to avoid wasting memory. */
 	i915_sg_trim(st);
@@ -2435,7 +2443,7 @@ static int 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, st);
 
-	__i915_gem_object_set_pages(obj, st);
+	__i915_gem_object_set_pages(obj, st, sg_mask);
 
 	return 0;
 
@@ -2462,8 +2470,13 @@ static int i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
 }
 
 void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
-				 struct sg_table *pages)
+				 struct sg_table *pages,
+				 unsigned int sg_mask)
 {
+	struct drm_i915_private *i915 = to_i915(obj->base.dev);
+	unsigned long supported_page_sizes = INTEL_INFO(i915)->page_size_mask;
+	int bit;
+
 	lockdep_assert_held(&obj->mm.lock);
 
 	obj->mm.get_page.sg_pos = pages->sgl;
@@ -2477,6 +2490,18 @@ void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
 		__i915_gem_object_pin_pages(obj);
 		obj->mm.quirked = true;
 	}
+
+	GEM_BUG_ON(!sg_mask);
+
+	obj->mm.page_sizes.phys = sg_mask;
+
+	obj->mm.page_sizes.sg = 0;
+	for_each_set_bit(bit, &supported_page_sizes, BITS_PER_LONG) {
+		if (obj->mm.page_sizes.phys & ~0u << bit)
+			obj->mm.page_sizes.sg |= BIT(bit);
+	}
+
+	GEM_BUG_ON(!HAS_PAGE_SIZE(i915, obj->mm.page_sizes.sg));
 }
 
 static int ____i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
diff --git a/drivers/gpu/drm/i915/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/i915_gem_dmabuf.c
index 4c4dc85159fb..409747446854 100644
--- a/drivers/gpu/drm/i915/i915_gem_dmabuf.c
+++ b/drivers/gpu/drm/i915/i915_gem_dmabuf.c
@@ -259,13 +259,19 @@ struct dma_buf *i915_gem_prime_export(struct drm_device *dev,
 static int i915_gem_object_get_pages_dmabuf(struct drm_i915_gem_object *obj)
 {
 	struct sg_table *pages;
+	struct scatterlist *sg;
+	unsigned int sg_mask = 0;
+	int n;
 
 	pages = dma_buf_map_attachment(obj->base.import_attach,
 				       DMA_BIDIRECTIONAL);
 	if (IS_ERR(pages))
 		return PTR_ERR(pages);
 
-	__i915_gem_object_set_pages(obj, pages);
+	for_each_sg(pages->sgl, sg, pages->nents, n)
+		sg_mask |= sg->length;
+
+	__i915_gem_object_set_pages(obj, pages, sg_mask);
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/i915/i915_gem_internal.c b/drivers/gpu/drm/i915/i915_gem_internal.c
index f59764da4254..34ddfa70abf7 100644
--- a/drivers/gpu/drm/i915/i915_gem_internal.c
+++ b/drivers/gpu/drm/i915/i915_gem_internal.c
@@ -49,6 +49,7 @@ static int i915_gem_object_get_pages_internal(struct drm_i915_gem_object *obj)
 	struct drm_i915_private *i915 = to_i915(obj->base.dev);
 	struct sg_table *st;
 	struct scatterlist *sg;
+	unsigned int sg_mask;
 	unsigned int npages;
 	int max_order;
 	gfp_t gfp;
@@ -75,6 +76,7 @@ static int i915_gem_object_get_pages_internal(struct drm_i915_gem_object *obj)
 	}
 
 create_st:
+	sg_mask = 0;
 	st = kmalloc(sizeof(*st), GFP_KERNEL);
 	if (!st)
 		return -ENOMEM;
@@ -104,6 +106,7 @@ static int i915_gem_object_get_pages_internal(struct drm_i915_gem_object *obj)
 		} while (1);
 
 		sg_set_page(sg, page, PAGE_SIZE << order, 0);
+		sg_mask |= PAGE_SIZE << order;
 		st->nents++;
 
 		npages -= 1 << order;
@@ -132,7 +135,7 @@ static int i915_gem_object_get_pages_internal(struct drm_i915_gem_object *obj)
 	 */
 	obj->mm.madv = I915_MADV_DONTNEED;
 
-	__i915_gem_object_set_pages(obj, st);
+	__i915_gem_object_set_pages(obj, st, sg_mask);
 
 	return 0;
 
diff --git a/drivers/gpu/drm/i915/i915_gem_object.h b/drivers/gpu/drm/i915/i915_gem_object.h
index 036e847b27f0..110672952a1c 100644
--- a/drivers/gpu/drm/i915/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/i915_gem_object.h
@@ -169,6 +169,23 @@ struct drm_i915_gem_object {
 		struct sg_table *pages;
 		void *mapping;
 
+		struct i915_page_sizes {
+			/**
+			 * The sg mask of the pages sg_table. i.e the mask of
+			 * of the lengths for each sg entry.
+			 */
+			unsigned int phys;
+
+			/**
+			 * The gtt page sizes we are allowed to use given the
+			 * sg mask and the supported page sizes. This will
+			 * express the smallest unit we can use for the whole
+			 * object, as well as the larger sizes we may be able
+			 * to use opportunistically.
+			 */
+			unsigned int sg;
+		} page_sizes;
+
 		struct i915_gem_object_page_iter {
 			struct scatterlist *sg_pos;
 			unsigned int sg_idx; /* in pages, but 32bit eek! */
diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c
index 537ecb224db0..54fd4cfa9d07 100644
--- a/drivers/gpu/drm/i915/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/i915_gem_stolen.c
@@ -548,7 +548,7 @@ static int i915_gem_object_get_pages_stolen(struct drm_i915_gem_object *obj)
 	if (IS_ERR(pages))
 		return PTR_ERR(pages);
 
-	__i915_gem_object_set_pages(obj, pages);
+	__i915_gem_object_set_pages(obj, pages, obj->stolen->size);
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c
index 25da3c60e961..115dfff1d484 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/i915_gem_userptr.c
@@ -406,7 +406,8 @@ struct get_pages_work {
 #endif
 
 static int
-st_set_pages(struct sg_table **st, struct page **pvec, int num_pages)
+st_set_pages(struct sg_table **st, struct page **pvec, int num_pages,
+	     unsigned int *sg_mask)
 {
 	struct scatterlist *sg;
 	int ret, n;
@@ -422,12 +423,17 @@ st_set_pages(struct sg_table **st, struct page **pvec, int num_pages)
 
 		for_each_sg((*st)->sgl, sg, num_pages, n)
 			sg_set_page(sg, pvec[n], PAGE_SIZE, 0);
+
+		*sg_mask = PAGE_SIZE;
 	} else {
 		ret = sg_alloc_table_from_pages(*st, pvec, num_pages,
 						0, num_pages << PAGE_SHIFT,
 						GFP_KERNEL);
 		if (ret)
 			goto err;
+
+		for_each_sg((*st)->sgl, sg, num_pages, n)
+			*sg_mask |= sg->length;
 	}
 
 	return 0;
@@ -443,9 +449,10 @@ __i915_gem_userptr_set_pages(struct drm_i915_gem_object *obj,
 			     struct page **pvec, int num_pages)
 {
 	struct sg_table *pages;
+	unsigned int sg_mask = 0;
 	int ret;
 
-	ret = st_set_pages(&pages, pvec, num_pages);
+	ret = st_set_pages(&pages, pvec, num_pages, &sg_mask);
 	if (ret)
 		return ERR_PTR(ret);
 
@@ -456,7 +463,7 @@ __i915_gem_userptr_set_pages(struct drm_i915_gem_object *obj,
 		return ERR_PTR(ret);
 	}
 
-	__i915_gem_object_set_pages(obj, pages);
+	__i915_gem_object_set_pages(obj, pages, sg_mask);
 
 	return pages;
 }
diff --git a/drivers/gpu/drm/i915/selftests/huge_gem_object.c b/drivers/gpu/drm/i915/selftests/huge_gem_object.c
index 41c15f3aa467..a2632df39173 100644
--- a/drivers/gpu/drm/i915/selftests/huge_gem_object.c
+++ b/drivers/gpu/drm/i915/selftests/huge_gem_object.c
@@ -80,7 +80,7 @@ static int huge_get_pages(struct drm_i915_gem_object *obj)
 	if (i915_gem_gtt_prepare_pages(obj, pages))
 		goto err;
 
-	__i915_gem_object_set_pages(obj, pages);
+	__i915_gem_object_set_pages(obj, pages, PAGE_SIZE);
 
 	return 0;
 
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
index aa1db375d59a..2dd66f866a5d 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
@@ -45,6 +45,7 @@ static int fake_get_pages(struct drm_i915_gem_object *obj)
 #define PFN_BIAS 0x1000
 	struct sg_table *pages;
 	struct scatterlist *sg;
+	unsigned int sg_mask = 0;
 	typeof(obj->base.size) rem;
 
 	pages = kmalloc(sizeof(*pages), GFP);
@@ -65,6 +66,7 @@ static int fake_get_pages(struct drm_i915_gem_object *obj)
 		sg_set_page(sg, pfn_to_page(PFN_BIAS), len, 0);
 		sg_dma_address(sg) = page_to_phys(sg_page(sg));
 		sg_dma_len(sg) = len;
+		sg_mask |= len;
 
 		rem -= len;
 	}
@@ -72,7 +74,7 @@ static int fake_get_pages(struct drm_i915_gem_object *obj)
 
 	obj->mm.madv = I915_MADV_DONTNEED;
 
-	__i915_gem_object_set_pages(obj, pages);
+	__i915_gem_object_set_pages(obj, pages, sg_mask);
 
 	return 0;
 #undef GFP
-- 
2.13.5

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx