[PATCH 05/25] drm/i915/region: support continuous allocations
Matthew Auld
matthew.auld at intel.com
Fri Sep 20 22:31:58 UTC 2019
Some kernel internal objects may need to be allocated as a continuous
block, also thinking ahead the various kernel io_mapping interfaces seem
to expect it, although this is purely a limitation in the kernel
API...so perhaps something to be improved.
Signed-off-by: Matthew Auld <matthew.auld at intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
Cc: Abdiel Janulgue <abdiel.janulgue at linux.intel.com>
---
.../gpu/drm/i915/gem/i915_gem_object_types.h | 4 +
drivers/gpu/drm/i915/gem/i915_gem_region.c | 15 +-
drivers/gpu/drm/i915/gem/i915_gem_region.h | 3 +-
.../gpu/drm/i915/gem/selftests/huge_pages.c | 3 +-
drivers/gpu/drm/i915/intel_memory_region.c | 13 +-
drivers/gpu/drm/i915/intel_memory_region.h | 3 +-
.../drm/i915/selftests/intel_memory_region.c | 154 +++++++++++++++++-
drivers/gpu/drm/i915/selftests/mock_region.c | 2 +-
8 files changed, 182 insertions(+), 15 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
index 1745bd8b0e42..015cfff75e5d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -117,6 +117,10 @@ struct drm_i915_gem_object {
I915_SELFTEST_DECLARE(struct list_head st_link);
+ unsigned long flags;
+#define I915_BO_ALLOC_CONTIGUOUS BIT(0)
+#define I915_BO_ALLOC_FLAGS (I915_BO_ALLOC_CONTIGUOUS)
+
/*
* Is the object to be mapped as read-only to the GPU
* Only honoured if hardware has relevant pte bit
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_region.c b/drivers/gpu/drm/i915/gem/i915_gem_region.c
index cda1e1aef3c8..8a329958b3a4 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_region.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_region.c
@@ -23,10 +23,10 @@ i915_gem_object_get_pages_buddy(struct drm_i915_gem_object *obj)
{
struct intel_memory_region *mem = obj->mm.region;
struct list_head *blocks = &obj->mm.blocks;
- unsigned int flags = I915_ALLOC_MIN_PAGE_SIZE;
resource_size_t size = obj->base.size;
resource_size_t prev_end;
struct i915_buddy_block *block;
+ unsigned int flags;
struct sg_table *st;
struct scatterlist *sg;
unsigned int sg_page_sizes;
@@ -42,6 +42,10 @@ i915_gem_object_get_pages_buddy(struct drm_i915_gem_object *obj)
return -ENOMEM;
}
+ flags = I915_ALLOC_MIN_PAGE_SIZE;
+ if (obj->flags & I915_BO_ALLOC_CONTIGUOUS)
+ flags |= I915_ALLOC_CONTIGUOUS;
+
ret = __intel_memory_region_get_pages_buddy(mem, size, flags, blocks);
if (ret)
goto err_free_sg;
@@ -56,7 +60,8 @@ i915_gem_object_get_pages_buddy(struct drm_i915_gem_object *obj)
list_for_each_entry(block, blocks, link) {
u64 block_size, offset;
- block_size = i915_buddy_block_size(&mem->mm, block);
+ block_size = min_t(u64, size,
+ i915_buddy_block_size(&mem->mm, block));
offset = i915_buddy_block_offset(block);
GEM_BUG_ON(overflows_type(block_size, sg->length));
@@ -98,10 +103,12 @@ i915_gem_object_get_pages_buddy(struct drm_i915_gem_object *obj)
}
void i915_gem_object_init_memory_region(struct drm_i915_gem_object *obj,
- struct intel_memory_region *mem)
+ struct intel_memory_region *mem,
+ unsigned long flags)
{
INIT_LIST_HEAD(&obj->mm.blocks);
obj->mm.region= mem;
+ obj->flags = flags;
mutex_lock(&mem->obj_lock);
list_add(&obj->mm.region_link, &mem->objects);
@@ -122,6 +129,8 @@ i915_gem_object_create_region(struct intel_memory_region *mem,
{
struct drm_i915_gem_object *obj;
+ GEM_BUG_ON(flags & ~I915_BO_ALLOC_FLAGS);
+
if (!mem)
return ERR_PTR(-ENODEV);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_region.h b/drivers/gpu/drm/i915/gem/i915_gem_region.h
index ebddc86d78f7..f2ff6f8bff74 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_region.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_region.h
@@ -17,7 +17,8 @@ void i915_gem_object_put_pages_buddy(struct drm_i915_gem_object *obj,
struct sg_table *pages);
void i915_gem_object_init_memory_region(struct drm_i915_gem_object *obj,
- struct intel_memory_region *mem);
+ struct intel_memory_region *mem,
+ unsigned long flags);
void i915_gem_object_release_memory_region(struct drm_i915_gem_object *obj);
struct drm_i915_gem_object *
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index 4e1805aaeb99..f9fbf2865782 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -471,7 +471,8 @@ static int igt_mock_memory_region_huge_pages(void *arg)
unsigned int page_size = BIT(bit);
resource_size_t phys;
- obj = i915_gem_object_create_region(mem, page_size, 0);
+ obj = i915_gem_object_create_region(mem, page_size,
+ I915_BO_ALLOC_CONTIGUOUS);
if (IS_ERR(obj)) {
err = PTR_ERR(obj);
goto out_destroy_device;
diff --git a/drivers/gpu/drm/i915/intel_memory_region.c b/drivers/gpu/drm/i915/intel_memory_region.c
index ea3980bd2288..3f5e82d78d36 100644
--- a/drivers/gpu/drm/i915/intel_memory_region.c
+++ b/drivers/gpu/drm/i915/intel_memory_region.c
@@ -87,8 +87,8 @@ __intel_memory_region_get_pages_buddy(struct intel_memory_region *mem,
unsigned int flags,
struct list_head *blocks)
{
- unsigned long n_pages = size >> ilog2(mem->mm.chunk_size);
unsigned int min_order = 0;
+ unsigned long n_pages;
GEM_BUG_ON(!IS_ALIGNED(size, mem->mm.chunk_size));
GEM_BUG_ON(!list_empty(blocks));
@@ -98,6 +98,13 @@ __intel_memory_region_get_pages_buddy(struct intel_memory_region *mem,
ilog2(mem->mm.chunk_size);
}
+ if (flags & I915_ALLOC_CONTIGUOUS) {
+ size = roundup_pow_of_two(size);
+ min_order = ilog2(size) - ilog2(mem->mm.chunk_size);
+ }
+
+ n_pages = size >> ilog2(mem->mm.chunk_size);
+
mutex_lock(&mem->mm_lock);
do {
@@ -161,7 +168,9 @@ __intel_memory_region_get_block_buddy(struct intel_memory_region *mem,
int ret;
INIT_LIST_HEAD(&blocks);
- ret = __intel_memory_region_get_pages_buddy(mem, size, 0, &blocks);
+ ret = __intel_memory_region_get_pages_buddy(mem, size,
+ I915_ALLOC_CONTIGUOUS,
+ &blocks);
if (ret)
return ERR_PTR(ret);
diff --git a/drivers/gpu/drm/i915/intel_memory_region.h b/drivers/gpu/drm/i915/intel_memory_region.h
index a7d0819a3319..095f5a8b77af 100644
--- a/drivers/gpu/drm/i915/intel_memory_region.h
+++ b/drivers/gpu/drm/i915/intel_memory_region.h
@@ -17,7 +17,8 @@ struct drm_i915_gem_object;
struct intel_memory_region;
struct sg_table;
-#define I915_ALLOC_MIN_PAGE_SIZE BIT(0)
+#define I915_ALLOC_MIN_PAGE_SIZE BIT(0)
+#define I915_ALLOC_CONTIGUOUS BIT(1)
struct intel_memory_region_ops {
unsigned int flags;
diff --git a/drivers/gpu/drm/i915/selftests/intel_memory_region.c b/drivers/gpu/drm/i915/selftests/intel_memory_region.c
index 8947c2d9549b..19bb60d50680 100644
--- a/drivers/gpu/drm/i915/selftests/intel_memory_region.c
+++ b/drivers/gpu/drm/i915/selftests/intel_memory_region.c
@@ -81,17 +81,17 @@ static int igt_mock_fill(void *arg)
static void igt_mark_evictable(struct drm_i915_gem_object *obj)
{
- i915_gem_object_unpin_pages(obj);
+ if (i915_gem_object_has_pinned_pages(obj))
+ i915_gem_object_unpin_pages(obj);
obj->mm.madv = I915_MADV_DONTNEED;
list_move(&obj->mm.region_link, &obj->mm.region->purgeable);
}
-static int igt_mock_evict(void *arg)
+static int igt_frag_region(struct intel_memory_region *mem,
+ struct list_head *objects)
{
- struct intel_memory_region *mem = arg;
struct drm_i915_gem_object *obj;
unsigned long n_objects;
- LIST_HEAD(objects);
resource_size_t target;
resource_size_t total;
int err = 0;
@@ -109,7 +109,7 @@ static int igt_mock_evict(void *arg)
goto err_close_objects;
}
- list_add(&obj->st_link, &objects);
+ list_add(&obj->st_link, objects);
err = i915_gem_object_pin_pages(obj);
if (err)
@@ -123,7 +123,40 @@ static int igt_mock_evict(void *arg)
igt_mark_evictable(obj);
}
- while (target <= total / 2) {
+ return 0;
+
+err_close_objects:
+ close_objects(objects);
+ return err;
+}
+
+static void igt_defrag_region(struct list_head *objects)
+{
+ struct drm_i915_gem_object *obj;
+
+ list_for_each_entry(obj, objects, st_link) {
+ if (obj->mm.madv == I915_MADV_WILLNEED)
+ igt_mark_evictable(obj);
+ }
+}
+
+static int igt_mock_evict(void *arg)
+{
+ struct intel_memory_region *mem = arg;
+ struct drm_i915_gem_object *obj;
+ LIST_HEAD(objects);
+ resource_size_t target;
+ resource_size_t total;
+ int err;
+
+ err = igt_frag_region(mem, &objects);
+ if (err)
+ return err;
+
+ total = resource_size(&mem->region);
+ target = mem->mm.chunk_size;
+
+ while (target <= total >> 1) {
obj = i915_gem_object_create_region(mem, target, 0);
if (IS_ERR(obj)) {
err = PTR_ERR(obj);
@@ -153,11 +186,120 @@ static int igt_mock_evict(void *arg)
return err;
}
+static int igt_mock_continuous(void *arg)
+{
+ struct intel_memory_region *mem = arg;
+ struct drm_i915_gem_object *obj;
+ LIST_HEAD(objects);
+ resource_size_t target;
+ resource_size_t total;
+ int err;
+
+ err = igt_frag_region(mem, &objects);
+ if (err)
+ return err;
+
+ total = resource_size(&mem->region);
+ target = total / 2;
+
+ /*
+ * Sanity check that we can allocate all of the available fragmented
+ * space.
+ */
+ obj = i915_gem_object_create_region(mem, target, 0);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ goto err_close_objects;
+ }
+
+ list_add(&obj->st_link, &objects);
+
+ err = i915_gem_object_pin_pages(obj);
+ if (err) {
+ pr_err("failed to allocate available space\n");
+ goto err_close_objects;
+ }
+
+ igt_mark_evictable(obj);
+
+ /* Try the smallest possible size -- should succeed */
+ obj = i915_gem_object_create_region(mem, mem->mm.chunk_size,
+ I915_BO_ALLOC_CONTIGUOUS);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ goto err_close_objects;
+ }
+
+ list_add(&obj->st_link, &objects);
+
+ err = i915_gem_object_pin_pages(obj);
+ if (err) {
+ pr_err("failed to allocate smallest possible size\n");
+ goto err_close_objects;
+ }
+
+ igt_mark_evictable(obj);
+
+ if (obj->mm.pages->nents != 1) {
+ pr_err("[1]object spans multiple sg entries\n");
+ err = -EINVAL;
+ goto err_close_objects;
+ }
+
+ /*
+ * Even though there is enough free space for the allocation, we
+ * shouldn't be able to allocate it, given that it is fragmented, and
+ * non-continuous.
+ */
+ obj = i915_gem_object_create_region(mem, target, I915_BO_ALLOC_CONTIGUOUS);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ goto err_close_objects;
+ }
+
+ list_add(&obj->st_link, &objects);
+
+ err = i915_gem_object_pin_pages(obj);
+ if (!err) {
+ pr_err("expected allocation to fail\n");
+ err = -EINVAL;
+ goto err_close_objects;
+ }
+
+ igt_defrag_region(&objects);
+
+ /* Should now succeed */
+ obj = i915_gem_object_create_region(mem, target, I915_BO_ALLOC_CONTIGUOUS);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ goto err_close_objects;
+ }
+
+ list_add(&obj->st_link, &objects);
+
+ err = i915_gem_object_pin_pages(obj);
+ if (err) {
+ pr_err("failed to allocate from defraged area\n");
+ goto err_close_objects;
+ }
+
+ if (obj->mm.pages->nents != 1) {
+ pr_err("object spans multiple sg entries\n");
+ err = -EINVAL;
+ }
+
+err_close_objects:
+ close_objects(&objects);
+
+ return err;
+}
+
int intel_memory_region_mock_selftests(void)
{
static const struct i915_subtest tests[] = {
SUBTEST(igt_mock_fill),
SUBTEST(igt_mock_evict),
+ SUBTEST(igt_mock_continuous),
};
struct intel_memory_region *mem;
struct drm_i915_private *i915;
diff --git a/drivers/gpu/drm/i915/selftests/mock_region.c b/drivers/gpu/drm/i915/selftests/mock_region.c
index 0e9a575ede3b..7b0c99ddc2d5 100644
--- a/drivers/gpu/drm/i915/selftests/mock_region.c
+++ b/drivers/gpu/drm/i915/selftests/mock_region.c
@@ -36,7 +36,7 @@ mock_object_create(struct intel_memory_region *mem,
i915_gem_object_set_cache_coherency(obj, I915_CACHE_NONE);
- i915_gem_object_init_memory_region(obj, mem);
+ i915_gem_object_init_memory_region(obj, mem, flags);
return obj;
}
--
2.20.1
More information about the Intel-gfx-trybot
mailing list