[PATCH v2 0/3] Support for mapping an object page by page

[PATCH v2 0/3] Support for mapping an object page by page

[ankitprasad.r.sharma]

From: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>

It is possible that when we want to map an object to the aperture, either
we run out of aperture space or the size of the object is larger than
the mappable aperture. In such cases we might not be able to map the whole
object to the aperture. For cases as such, here we introduce insert_page()
which allows us to map a single page in to the mappable aperture space
(which has a higher probabilty of succeeding). This can be iterated
over to access the whole object by using space as meagre as page size.

Here we try to use insert_page() for pwrite_fast in case a nonblocking
pin for the whole object fails, which helps us to iterate over the whole
object and perform the pwrite without mapping the whole object to the
mappable aperture.

We also introduce i915_gem_object_get_dma_address() to perform fast
sequential lookup of the dma address associated with any page within the
object.

v2: Addressed Comments

Ankitprasad Sharma (1):
  drm/i915: Use insert_page for pwrite_fast

Chris Wilson (2):
  drm/i915: Add support for mapping an object page by page
  drm/i915: Introduce i915_gem_object_get_dma_address()

 drivers/char/agp/intel-gtt.c        |  9 +++++
 drivers/gpu/drm/i915/i915_drv.h     | 17 ++++++++++
 drivers/gpu/drm/i915/i915_gem.c     | 68 +++++++++++++++++++++++++++++--------
 drivers/gpu/drm/i915/i915_gem_gtt.c | 49 ++++++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_gem_gtt.h |  5 +++
 include/drm/intel-gtt.h             |  3 ++
 6 files changed, 136 insertions(+), 15 deletions(-)

-- 
1.9.1

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

[PATCH 1/3] drm/i915: Add support for mapping an object page by page

[ankitprasad.r.sharma]

From: Chris Wilson <chris@chris-wilson.co.uk>

Introduced a new vm specfic callback insert_page() to program a single pte in
ggtt or ppgtt. This allows us to map a single page in to the mappable aperture
space. This can be iterated over to access the whole object by using space as
meagre as page size.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>
---
 drivers/char/agp/intel-gtt.c        |  9 +++++++
 drivers/gpu/drm/i915/i915_gem_gtt.c | 49 +++++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_gem_gtt.h |  5 ++++
 include/drm/intel-gtt.h             |  3 +++
 4 files changed, 66 insertions(+)

diff --git a/drivers/char/agp/intel-gtt.c b/drivers/char/agp/intel-gtt.c
index 1341a94..7c68576 100644
--- a/drivers/char/agp/intel-gtt.c
+++ b/drivers/char/agp/intel-gtt.c
@@ -838,6 +838,15 @@ static bool i830_check_flags(unsigned int flags)
 	return false;
 }
 
+void intel_gtt_insert_page(dma_addr_t addr,
+			   unsigned int pg,
+			   unsigned int flags)
+{
+	intel_private.driver->write_entry(addr, pg, flags);
+	wmb();
+}
+EXPORT_SYMBOL(intel_gtt_insert_page);
+
 void intel_gtt_insert_sg_entries(struct sg_table *st,
 				 unsigned int pg_start,
 				 unsigned int flags)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 47344d0..f374360 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2340,6 +2340,23 @@ static void gen8_set_pte(void __iomem *addr, gen8_pte_t pte)
 #endif
 }
 
+static void gen8_ggtt_insert_page(struct i915_address_space *vm,
+				  dma_addr_t addr,
+				  uint64_t offset,
+				  enum i915_cache_level level,
+				  u32 unused)
+{
+	struct drm_i915_private *dev_priv = to_i915(vm->dev);
+	gen8_pte_t __iomem *pte =
+		(gen8_pte_t __iomem *)dev_priv->gtt.gsm +
+		(offset >> PAGE_SHIFT);
+
+	gen8_set_pte(pte, gen8_pte_encode(addr, level, true));
+	wmb();
+
+	I915_WRITE(GFX_FLSH_CNTL_GEN6, GFX_FLSH_CNTL_EN);
+}
+
 static void gen8_ggtt_insert_entries(struct i915_address_space *vm,
 				     struct sg_table *st,
 				     uint64_t start,
@@ -2380,6 +2397,23 @@ static void gen8_ggtt_insert_entries(struct i915_address_space *vm,
 	POSTING_READ(GFX_FLSH_CNTL_GEN6);
 }
 
+static void gen6_ggtt_insert_page(struct i915_address_space *vm,
+				  dma_addr_t addr,
+				  uint64_t offset,
+				  enum i915_cache_level level,
+				  u32 flags)
+{
+	struct drm_i915_private *dev_priv = to_i915(vm->dev);
+	gen6_pte_t __iomem *pte =
+		(gen6_pte_t __iomem *)dev_priv->gtt.gsm +
+		(offset >> PAGE_SHIFT);
+
+	iowrite32(vm->pte_encode(addr, level, true, flags), pte);
+	wmb();
+
+	I915_WRITE(GFX_FLSH_CNTL_GEN6, GFX_FLSH_CNTL_EN);
+}
+
 /*
  * Binds an object into the global gtt with the specified cache level. The object
  * will be accessible to the GPU via commands whose operands reference offsets
@@ -2476,6 +2510,18 @@ static void gen6_ggtt_clear_range(struct i915_address_space *vm,
 	readl(gtt_base);
 }
 
+static void i915_ggtt_insert_page(struct i915_address_space *vm,
+				  dma_addr_t addr,
+				  uint64_t offset,
+				  enum i915_cache_level cache_level,
+				  u32 unused)
+{
+	unsigned int flags = (cache_level == I915_CACHE_NONE) ?
+		AGP_USER_MEMORY : AGP_USER_CACHED_MEMORY;
+
+	intel_gtt_insert_page(addr, offset >> PAGE_SHIFT, flags);
+}
+
 static void i915_ggtt_insert_entries(struct i915_address_space *vm,
 				     struct sg_table *pages,
 				     uint64_t start,
@@ -2969,6 +3015,7 @@ static int gen8_gmch_probe(struct drm_device *dev,
 	ret = ggtt_probe_common(dev, gtt_size);
 
 	dev_priv->gtt.base.clear_range = gen8_ggtt_clear_range;
+	dev_priv->gtt.base.insert_page = gen8_ggtt_insert_page;
 	dev_priv->gtt.base.insert_entries = gen8_ggtt_insert_entries;
 	dev_priv->gtt.base.bind_vma = ggtt_bind_vma;
 	dev_priv->gtt.base.unbind_vma = ggtt_unbind_vma;
@@ -3011,6 +3058,7 @@ static int gen6_gmch_probe(struct drm_device *dev,
 	ret = ggtt_probe_common(dev, gtt_size);
 
 	dev_priv->gtt.base.clear_range = gen6_ggtt_clear_range;
+	dev_priv->gtt.base.insert_page = gen6_ggtt_insert_page;
 	dev_priv->gtt.base.insert_entries = gen6_ggtt_insert_entries;
 	dev_priv->gtt.base.bind_vma = ggtt_bind_vma;
 	dev_priv->gtt.base.unbind_vma = ggtt_unbind_vma;
@@ -3045,6 +3093,7 @@ static int i915_gmch_probe(struct drm_device *dev,
 	intel_gtt_get(gtt_total, stolen, mappable_base, mappable_end);
 
 	dev_priv->gtt.do_idle_maps = needs_idle_maps(dev_priv->dev);
+	dev_priv->gtt.base.insert_page = i915_ggtt_insert_page;
 	dev_priv->gtt.base.insert_entries = i915_ggtt_insert_entries;
 	dev_priv->gtt.base.clear_range = i915_ggtt_clear_range;
 	dev_priv->gtt.base.bind_vma = ggtt_bind_vma;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 9fbb07d..f7e3b45 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -320,6 +320,11 @@ struct i915_address_space {
 			    uint64_t start,
 			    uint64_t length,
 			    bool use_scratch);
+	void (*insert_page)(struct i915_address_space *vm,
+			    dma_addr_t addr,
+			    uint64_t offset,
+			    enum i915_cache_level cache_level,
+			    u32 flags);
 	void (*insert_entries)(struct i915_address_space *vm,
 			       struct sg_table *st,
 			       uint64_t start,
diff --git a/include/drm/intel-gtt.h b/include/drm/intel-gtt.h
index 9e9bddaa5..f49edec 100644
--- a/include/drm/intel-gtt.h
+++ b/include/drm/intel-gtt.h
@@ -13,6 +13,9 @@ void intel_gmch_remove(void);
 bool intel_enable_gtt(void);
 
 void intel_gtt_chipset_flush(void);
+void intel_gtt_insert_page(dma_addr_t addr,
+			   unsigned int pg,
+			   unsigned int flags);
 void intel_gtt_insert_sg_entries(struct sg_table *st,
 				 unsigned int pg_start,
 				 unsigned int flags);
-- 
1.9.1

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

[PATCH 2/3] drm/i915: Introduce i915_gem_object_get_dma_address()

[ankitprasad.r.sharma]

From: Chris Wilson <chris@chris-wilson.co.uk>

This utility function is a companion to i915_gem_object_get_page() that
uses the same cached iterator for the scatterlist to perform fast
sequential lookup of the dma address associated with any page within the
object.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>
---
 drivers/gpu/drm/i915/i915_drv.h | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 2b5d587..4baf252 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2854,6 +2854,23 @@ static inline int __sg_page_count(struct scatterlist *sg)
 	return sg->length >> PAGE_SHIFT;
 }
 
+static inline dma_addr_t
+i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj, int n)
+{
+	if (n < obj->get_page.last) {
+		obj->get_page.sg = obj->pages->sgl;
+		obj->get_page.last = 0;
+	}
+
+	while (obj->get_page.last + __sg_page_count(obj->get_page.sg) <= n) {
+		obj->get_page.last += __sg_page_count(obj->get_page.sg++);
+		if (unlikely(sg_is_chain(obj->get_page.sg)))
+			obj->get_page.sg = sg_chain_ptr(obj->get_page.sg);
+	}
+
+	return sg_dma_address(obj->get_page.sg) + ((n - obj->get_page.last) << PAGE_SHIFT);
+}
+
 static inline struct page *
 i915_gem_object_get_page(struct drm_i915_gem_object *obj, int n)
 {
-- 
1.9.1

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

[PATCH 3/3] drm/i915: Use insert_page for pwrite_fast

[ankitprasad.r.sharma]

From: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>

In pwrite_fast, map an object page by page if obj_ggtt_pin fails. First,
we try a nonblocking pin for the whole object (since that is fastest if
reused), then failing that we try to grab one page in the mappable
aperture. It also allows us to handle objects larger than the mappable
aperture (e.g. if we need to pwrite with vGPU restricting the aperture
to a measely 8MiB or something like that).

v2: Pin pages before starting pwrite, Combined duplicate loops (Chris)

Signed-off-by: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>
---
 drivers/gpu/drm/i915/i915_gem.c | 68 ++++++++++++++++++++++++++++++++---------
 1 file changed, 53 insertions(+), 15 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index bf5ef7a..2556936 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -766,14 +766,26 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev,
 			 struct drm_file *file)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_mm_node node;
 	ssize_t remain;
 	loff_t offset, page_base;
 	char __user *user_data;
-	int page_offset, page_length, ret;
+	int page_offset, page_length, ret, i = 0;
+	bool pinned = true;
 
 	ret = i915_gem_obj_ggtt_pin(obj, 0, PIN_MAPPABLE | PIN_NONBLOCK);
-	if (ret)
-		goto out;
+	if (ret) {
+		pinned = false;
+		memset(&node, 0, sizeof(node));
+		ret = drm_mm_insert_node_in_range_generic(&dev_priv->gtt.base.mm,
+							  &node, 4096, 0,
+							  I915_CACHE_NONE, 0,
+							  dev_priv->gtt.mappable_end,
+							  DRM_MM_SEARCH_DEFAULT,
+							  DRM_MM_CREATE_DEFAULT);
+		if (ret)
+			goto out;
+	}
 
 	ret = i915_gem_object_set_to_gtt_domain(obj, true);
 	if (ret)
@@ -783,26 +795,41 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev,
 	if (ret)
 		goto out_unpin;
 
+	i915_gem_object_pin_pages(obj);
+
 	user_data = to_user_ptr(args->data_ptr);
 	remain = args->size;
 
-	offset = i915_gem_obj_ggtt_offset(obj) + args->offset;
-
 	intel_fb_obj_invalidate(obj, ORIGIN_GTT);
 
-	while (remain > 0) {
-		/* Operation in this page
-		 *
-		 * page_base = page offset within aperture
-		 * page_offset = offset within page
-		 * page_length = bytes to copy for this page
-		 */
+	if (pinned) {
+		offset = i915_gem_obj_ggtt_offset(obj) + args->offset;
+		/* page_base = page offset within aperture */
 		page_base = offset & PAGE_MASK;
-		page_offset = offset_in_page(offset);
+	} else {
+		offset = args->offset;
+		page_base = node.start;
+		i = args->offset / PAGE_SIZE;
+	}
+
+	/* page_offset = offset within page */
+	page_offset = offset_in_page(offset);
+	while (remain > 0) {
+		/* page_length = bytes to copy for the page */
 		page_length = remain;
 		if ((page_offset + remain) > PAGE_SIZE)
 			page_length = PAGE_SIZE - page_offset;
 
+		if (!pinned) {
+			wmb();
+			dev_priv->gtt.base.insert_page(&dev_priv->gtt.base,
+					i915_gem_object_get_dma_address(obj, i),
+					node.start,
+					I915_CACHE_NONE,
+					0);
+			wmb();
+			i++;
+		}
 		/* If we get a fault while copying data, then (presumably) our
 		 * source page isn't available.  Return the error and we'll
 		 * retry in the slow path.
@@ -815,13 +842,24 @@ i915_gem_gtt_pwrite_fast(struct drm_device *dev,
 
 		remain -= page_length;
 		user_data += page_length;
-		offset += page_length;
+		page_offset = 0;
 	}
 
 out_flush:
+	if (!pinned) {
+		wmb();
+		dev_priv->gtt.base.clear_range(&dev_priv->gtt.base,
+				node.start, node.size,
+				true);
+	}
+
+	i915_gem_object_unpin_pages(obj);
 	intel_fb_obj_flush(obj, false, ORIGIN_GTT);
 out_unpin:
-	i915_gem_object_ggtt_unpin(obj);
+	if (pinned)
+		i915_gem_object_ggtt_unpin(obj);
+	else
+		drm_mm_remove_node(&node);
 out:
 	return ret;
 }
-- 
1.9.1

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

Re: [PATCH 3/3] drm/i915: Use insert_page for pwrite_fast

[Chris Wilson]

On Sat, Nov 07, 2015 at 01:32:15PM +0530, ankitprasad.r.sharma@intel.com wrote:
> From: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>
> 
> In pwrite_fast, map an object page by page if obj_ggtt_pin fails. First,
> we try a nonblocking pin for the whole object (since that is fastest if
> reused), then failing that we try to grab one page in the mappable
> aperture. It also allows us to handle objects larger than the mappable
> aperture (e.g. if we need to pwrite with vGPU restricting the aperture
> to a measely 8MiB or something like that).
> 
> v2: Pin pages before starting pwrite, Combined duplicate loops (Chris)

Please just use my code as you haven't combined the two paths as well
yet.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH 2/3] drm/i915: Introduce i915_gem_object_get_dma_address()

[ankitprasad.r.sharma]

From: Chris Wilson <chris@chris-wilson.co.uk>

This utility function is a companion to i915_gem_object_get_page() that
uses the same cached iterator for the scatterlist to perform fast
sequential lookup of the dma address associated with any page within the
object.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>
---
 drivers/gpu/drm/i915/i915_drv.h | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index d2a546a..548a0eb 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2862,6 +2862,23 @@ static inline int __sg_page_count(struct scatterlist *sg)
 	return sg->length >> PAGE_SHIFT;
 }
 
+static inline dma_addr_t
+i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj, int n)
+{
+	if (n < obj->get_page.last) {
+		obj->get_page.sg = obj->pages->sgl;
+		obj->get_page.last = 0;
+	}
+
+	while (obj->get_page.last + __sg_page_count(obj->get_page.sg) <= n) {
+		obj->get_page.last += __sg_page_count(obj->get_page.sg++);
+		if (unlikely(sg_is_chain(obj->get_page.sg)))
+			obj->get_page.sg = sg_chain_ptr(obj->get_page.sg);
+	}
+
+	return sg_dma_address(obj->get_page.sg) + ((n - obj->get_page.last) << PAGE_SHIFT);
+}
+
 static inline struct page *
 i915_gem_object_get_page(struct drm_i915_gem_object *obj, int n)
 {
-- 
1.9.1

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

[PATCH 2/3] drm/i915: Introduce i915_gem_object_get_dma_address()

[ankitprasad.r.sharma]

From: Chris Wilson <chris@chris-wilson.co.uk>

This utility function is a companion to i915_gem_object_get_page() that
uses the same cached iterator for the scatterlist to perform fast
sequential lookup of the dma address associated with any page within the
object.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com>
---
 drivers/gpu/drm/i915/i915_drv.h | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 2b5d587..4baf252 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2854,6 +2854,23 @@ static inline int __sg_page_count(struct scatterlist *sg)
 	return sg->length >> PAGE_SHIFT;
 }
 
+static inline dma_addr_t
+i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj, int n)
+{
+	if (n < obj->get_page.last) {
+		obj->get_page.sg = obj->pages->sgl;
+		obj->get_page.last = 0;
+	}
+
+	while (obj->get_page.last + __sg_page_count(obj->get_page.sg) <= n) {
+		obj->get_page.last += __sg_page_count(obj->get_page.sg++);
+		if (unlikely(sg_is_chain(obj->get_page.sg)))
+			obj->get_page.sg = sg_chain_ptr(obj->get_page.sg);
+	}
+
+	return sg_dma_address(obj->get_page.sg) + ((n - obj->get_page.last) << PAGE_SHIFT);
+}
+
 static inline struct page *
 i915_gem_object_get_page(struct drm_i915_gem_object *obj, int n)
 {
-- 
1.9.1

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