Clients like i915 needs to segregate cache domains within the GTT which
can lead to small amounts of fragmentation. By allocating the uncached
buffers from the bottom and the cacheable buffers from the top, we can
reduce the amount of wasted space and also optimize allocation of the
mappable portion of the GTT to only those buffers that require CPU
access through the GTT.
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
drivers/gpu/drm/drm_mm.c | 73 +++++++++++++++++++++++++--------------
drivers/gpu/drm/i915/i915_gem.c | 4 +--
include/drm/drm_mm.h | 42 ++++++++++++++--------
3 files changed, 77 insertions(+), 42 deletions(-)
diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
index b751b8e..59b21ec 100644
--- a/drivers/gpu/drm/drm_mm.c
+++ b/drivers/gpu/drm/drm_mm.c
@@ -49,7 +49,7 @@
#define MM_UNUSED_TARGET 4
-static struct drm_mm_node *drm_mm_kmalloc(struct drm_mm *mm, int atomic)
+static struct drm_mm_node *drm_mm_kmalloc(struct drm_mm *mm, bool atomic)
{
struct drm_mm_node *child;
@@ -105,7 +105,8 @@ EXPORT_SYMBOL(drm_mm_pre_get);
static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
struct drm_mm_node *node,
unsigned long size, unsigned alignment,
- unsigned long color)
+ unsigned long color,
+ unsigned flags)
{
struct drm_mm *mm = hole_node->mm;
unsigned long hole_start = drm_mm_hole_node_start(hole_node);
@@ -118,12 +119,22 @@ static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
if (mm->color_adjust)
mm->color_adjust(hole_node, color, &adj_start, &adj_end);
+ if (flags & DRM_MM_CREATE_TOP)
+ adj_start = adj_end - size;
+
if (alignment) {
unsigned tmp = adj_start % alignment;
- if (tmp)
- adj_start += alignment - tmp;
+ if (tmp) {
+ if (flags & DRM_MM_CREATE_TOP)
+ adj_start -= tmp;
+ else
+ adj_start += alignment - tmp;
+ }
}
+ BUG_ON(adj_start < hole_start);
+ BUG_ON(adj_end > hole_end);
+
if (adj_start == hole_start) {
hole_node->hole_follows = 0;
list_del(&hole_node->hole_stack);
@@ -150,7 +161,7 @@ static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
struct drm_mm_node *drm_mm_create_block(struct drm_mm *mm,
unsigned long start,
unsigned long size,
- bool atomic)
+ unsigned flags)
{
struct drm_mm_node *hole, *node;
unsigned long end = start + size;
@@ -161,7 +172,7 @@ struct drm_mm_node *drm_mm_create_block(struct drm_mm *mm,
if (hole_start > start || hole_end < end)
continue;
- node = drm_mm_kmalloc(mm, atomic);
+ node = drm_mm_kmalloc(mm, flags & DRM_MM_CREATE_ATOMIC);
if (unlikely(node == NULL))
return NULL;
@@ -196,15 +207,15 @@ struct drm_mm_node *drm_mm_get_block_generic(struct drm_mm_node *hole_node,
unsigned long size,
unsigned alignment,
unsigned long color,
- int atomic)
+ unsigned flags)
{
struct drm_mm_node *node;
- node = drm_mm_kmalloc(hole_node->mm, atomic);
+ node = drm_mm_kmalloc(hole_node->mm, flags & DRM_MM_CREATE_ATOMIC);
if (unlikely(node == NULL))
return NULL;
- drm_mm_insert_helper(hole_node, node, size, alignment, color);
+ drm_mm_insert_helper(hole_node, node, size, alignment, color, flags);
return node;
}
@@ -220,11 +231,11 @@ int drm_mm_insert_node(struct drm_mm *mm, struct drm_mm_node *node,
{
struct drm_mm_node *hole_node;
- hole_node = drm_mm_search_free(mm, size, alignment, false);
+ hole_node = drm_mm_search_free(mm, size, alignment, 0);
if (!hole_node)
return -ENOSPC;
- drm_mm_insert_helper(hole_node, node, size, alignment, 0);
+ drm_mm_insert_helper(hole_node, node, size, alignment, 0, 0);
return 0;
}
@@ -234,7 +245,8 @@ static void drm_mm_insert_helper_range(struct drm_mm_node *hole_node,
struct drm_mm_node *node,
unsigned long size, unsigned alignment,
unsigned long color,
- unsigned long start, unsigned long end)
+ unsigned long start, unsigned long end,
+ unsigned flags)
{
struct drm_mm *mm = hole_node->mm;
unsigned long hole_start = drm_mm_hole_node_start(hole_node);
@@ -249,11 +261,20 @@ static void drm_mm_insert_helper_range(struct drm_mm_node *hole_node,
if (adj_start < start)
adj_start = start;
+ if (adj_end > end)
+ adj_end = end;
+
+ if (flags & DRM_MM_CREATE_TOP)
+ adj_start = adj_end - size;
if (alignment) {
unsigned tmp = adj_start % alignment;
- if (tmp)
- adj_start += alignment - tmp;
+ if (tmp) {
+ if (flags & DRM_MM_CREATE_TOP)
+ adj_start -= tmp;
+ else
+ adj_start += alignment - tmp;
+ }
}
if (adj_start == hole_start) {
@@ -270,6 +291,8 @@ static void drm_mm_insert_helper_range(struct drm_mm_node *hole_node,
INIT_LIST_HEAD(&node->hole_stack);
list_add(&node->node_list, &hole_node->node_list);
+ BUG_ON(node->start < start);
+ BUG_ON(node->start < adj_start);
BUG_ON(node->start + node->size > adj_end);
BUG_ON(node->start + node->size > end);
@@ -286,16 +309,16 @@ struct drm_mm_node *drm_mm_get_block_range_generic(struct drm_mm_node *hole_node
unsigned long color,
unsigned long start,
unsigned long end,
- int atomic)
+ unsigned flags)
{
struct drm_mm_node *node;
- node = drm_mm_kmalloc(hole_node->mm, atomic);
+ node = drm_mm_kmalloc(hole_node->mm, flags & DRM_MM_CREATE_ATOMIC);
if (unlikely(node == NULL))
return NULL;
drm_mm_insert_helper_range(hole_node, node, size, alignment, color,
- start, end);
+ start, end, flags);
return node;
}
@@ -313,12 +336,12 @@ int drm_mm_insert_node_in_range(struct drm_mm *mm, struct drm_mm_node *node,
struct drm_mm_node *hole_node;
hole_node = drm_mm_search_free_in_range(mm, size, alignment,
- start, end, false);
+ start, end, 0);
if (!hole_node)
return -ENOSPC;
drm_mm_insert_helper_range(hole_node, node, size, alignment, 0,
- start, end);
+ start, end, 0);
return 0;
}
@@ -399,7 +422,7 @@ struct drm_mm_node *drm_mm_search_free_generic(const struct drm_mm *mm,
unsigned long size,
unsigned alignment,
unsigned long color,
- bool best_match)
+ unsigned flags)
{
struct drm_mm_node *entry;
struct drm_mm_node *best;
@@ -412,7 +435,7 @@ struct drm_mm_node *drm_mm_search_free_generic(const struct drm_mm *mm,
best = NULL;
best_size = ~0UL;
- drm_mm_for_each_hole(entry, mm, adj_start, adj_end) {
+ __drm_mm_for_each_hole(entry, mm, adj_start, adj_end, flags & DRM_MM_SEARCH_BELOW) {
if (mm->color_adjust) {
mm->color_adjust(entry, color, &adj_start, &adj_end);
if (adj_end <= adj_start)
@@ -422,7 +445,7 @@ struct drm_mm_node *drm_mm_search_free_generic(const struct drm_mm *mm,
if (!check_free_hole(adj_start, adj_end, size, alignment))
continue;
- if (!best_match)
+ if ((flags & DRM_MM_SEARCH_BEST) == 0)
return entry;
if (entry->size < best_size) {
@@ -441,7 +464,7 @@ struct drm_mm_node *drm_mm_search_free_in_range_generic(const struct drm_mm *mm,
unsigned long color,
unsigned long start,
unsigned long end,
- bool best_match)
+ unsigned flags)
{
struct drm_mm_node *entry;
struct drm_mm_node *best;
@@ -454,7 +477,7 @@ struct drm_mm_node *drm_mm_search_free_in_range_generic(const struct drm_mm *mm,
best = NULL;
best_size = ~0UL;
- drm_mm_for_each_hole(entry, mm, adj_start, adj_end) {
+ __drm_mm_for_each_hole(entry, mm, adj_start, adj_end, flags & DRM_MM_SEARCH_BELOW) {
if (adj_start < start)
adj_start = start;
if (adj_end > end)
@@ -469,7 +492,7 @@ struct drm_mm_node *drm_mm_search_free_in_range_generic(const struct drm_mm *mm,
if (!check_free_hole(adj_start, adj_end, size, alignment))
continue;
- if (!best_match)
+ if ((flags & DRM_MM_SEARCH_BEST) == 0)
return entry;
if (entry->size < best_size) {
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 89a1f03..8728ca2 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2972,12 +2972,12 @@ i915_gem_object_bind_to_gtt(struct drm_i915_gem_object *obj,
drm_mm_get_block_range_generic(free_space,
size, alignment, obj->cache_level,
0, dev_priv->mm.gtt_mappable_end,
- false);
+ 0);
else
free_space =
drm_mm_get_block_generic(free_space,
size, alignment, obj->cache_level,
- false);
+ 0);
}
if (free_space == NULL) {
ret = i915_gem_evict_something(dev, size, alignment,
diff --git a/include/drm/drm_mm.h b/include/drm/drm_mm.h
index cd45365..1f3f5f1 100644
--- a/include/drm/drm_mm.h
+++ b/include/drm/drm_mm.h
@@ -135,18 +135,26 @@ static inline unsigned long drm_mm_hole_node_end(struct drm_mm_node *hole_node)
1 : 0; \
entry = list_entry(entry->hole_stack.next, struct drm_mm_node, hole_stack))
+#define __drm_mm_for_each_hole(entry, mm, hole_start, hole_end, backwards) \
+ for (entry = list_entry((backwards) ? (mm)->hole_stack.prev : (mm)->hole_stack.next, struct drm_mm_node, hole_stack); \
+ &entry->hole_stack != &(mm)->hole_stack ? \
+ hole_start = drm_mm_hole_node_start(entry), \
+ hole_end = drm_mm_hole_node_end(entry), \
+ 1 : 0; \
+ entry = list_entry((backwards) ? entry->hole_stack.prev : entry->hole_stack.next, struct drm_mm_node, hole_stack))
+
/*
* Basic range manager support (drm_mm.c)
*/
extern struct drm_mm_node *drm_mm_create_block(struct drm_mm *mm,
unsigned long start,
unsigned long size,
- bool atomic);
+ unsigned flags);
extern struct drm_mm_node *drm_mm_get_block_generic(struct drm_mm_node *node,
unsigned long size,
unsigned alignment,
unsigned long color,
- int atomic);
+ unsigned flags);
extern struct drm_mm_node *drm_mm_get_block_range_generic(
struct drm_mm_node *node,
unsigned long size,
@@ -154,7 +162,9 @@ extern struct drm_mm_node *drm_mm_get_block_range_generic(
unsigned long color,
unsigned long start,
unsigned long end,
- int atomic);
+ unsigned flags);
+#define DRM_MM_CREATE_ATOMIC 0x1
+#define DRM_MM_CREATE_TOP 0x2
static inline struct drm_mm_node *drm_mm_get_block(struct drm_mm_node *parent,
unsigned long size,
unsigned alignment)
@@ -165,7 +175,7 @@ static inline struct drm_mm_node *drm_mm_get_block_atomic(struct drm_mm_node *pa
unsigned long size,
unsigned alignment)
{
- return drm_mm_get_block_generic(parent, size, alignment, 0, 1);
+ return drm_mm_get_block_generic(parent, size, alignment, 0, DRM_MM_CREATE_ATOMIC);
}
static inline struct drm_mm_node *drm_mm_get_block_range(
struct drm_mm_node *parent,
@@ -196,7 +206,7 @@ static inline struct drm_mm_node *drm_mm_get_block_atomic_range(
unsigned long end)
{
return drm_mm_get_block_range_generic(parent, size, alignment, 0,
- start, end, 1);
+ start, end, DRM_MM_CREATE_ATOMIC);
}
extern int drm_mm_insert_node(struct drm_mm *mm, struct drm_mm_node *node,
unsigned long size, unsigned alignment);
@@ -211,7 +221,7 @@ extern struct drm_mm_node *drm_mm_search_free_generic(const struct drm_mm *mm,
unsigned long size,
unsigned alignment,
unsigned long color,
- bool best_match);
+ unsigned flags);
extern struct drm_mm_node *drm_mm_search_free_in_range_generic(
const struct drm_mm *mm,
unsigned long size,
@@ -219,13 +229,15 @@ extern struct drm_mm_node *drm_mm_search_free_in_range_generic(
unsigned long color,
unsigned long start,
unsigned long end,
- bool best_match);
+ unsigned flags);
+#define DRM_MM_SEARCH_BEST 0x1
+#define DRM_MM_SEARCH_BELOW 0x2
static inline struct drm_mm_node *drm_mm_search_free(const struct drm_mm *mm,
unsigned long size,
unsigned alignment,
- bool best_match)
+ unsigned flags)
{
- return drm_mm_search_free_generic(mm,size, alignment, 0, best_match);
+ return drm_mm_search_free_generic(mm,size, alignment, 0, flags);
}
static inline struct drm_mm_node *drm_mm_search_free_in_range(
const struct drm_mm *mm,
@@ -233,18 +245,18 @@ static inline struct drm_mm_node *drm_mm_search_free_in_range(
unsigned alignment,
unsigned long start,
unsigned long end,
- bool best_match)
+ unsigned flags)
{
return drm_mm_search_free_in_range_generic(mm, size, alignment, 0,
- start, end, best_match);
+ start, end, flags);
}
static inline struct drm_mm_node *drm_mm_search_free_color(const struct drm_mm *mm,
unsigned long size,
unsigned alignment,
unsigned long color,
- bool best_match)
+ unsigned flags)
{
- return drm_mm_search_free_generic(mm,size, alignment, color, best_match);
+ return drm_mm_search_free_generic(mm,size, alignment, color, flags);
}
static inline struct drm_mm_node *drm_mm_search_free_in_range_color(
const struct drm_mm *mm,
@@ -253,10 +265,10 @@ static inline struct drm_mm_node *drm_mm_search_free_in_range_color(
unsigned long color,
unsigned long start,
unsigned long end,
- bool best_match)
+ unsigned flags)
{
return drm_mm_search_free_in_range_generic(mm, size, alignment, color,
- start, end, best_match);
+ start, end, flags);
}
extern int drm_mm_init(struct drm_mm *mm,
unsigned long start,
--
1.7.10.4
