OMAP: TILER: Optimize placement for NV12 buffers

This patch changes the behaviour when allocating 2D buffers.  If
the required alignment is less than 4KiB and a security token is set,
the allocation will be flagged specifically for this format.  All
subsequent buffers of the same format and size will be placed inside
this area until the area is exhausted.

For NV12 allocations, this means that all Y allocations will share
one area, and the UV allocations will share another.

Change-Id: I13d29257c0cbab627c2938d4623557918448ee86
Signed-off-by: Andy Gross <andy.gross@ti.com>

2 files changed, 32 insertions, 16 deletions

diff --git a/drivers/media/video/tiler/_tiler.h b/drivers/media/video/tiler/_tiler.h
index 382239d..d2fbcd8 100644
--- a/drivers/media/video/tiler/_tiler.h
+++ b/drivers/media/video/tiler/_tiler.h
@@ -45,6 +45,8 @@
 
 #define TILER_FORMATS		(TILFMT_MAX - TILFMT_MIN + 1)
 
+#define FLAGS_ALLOC_NO_COLOCATE	0x1
+
 enum secure_id_type {
 	SECURE_BY_PID = 0,
 	SECURE_BY_TOKEN
@@ -88,6 +90,8 @@ struct area_info {
 
 	struct tcm_area area;		/* area details */
 	struct gid_info *gi;		/* link to parent, if still alive */
+
+	u32 allowed_modes;
 };
 
 /* info for a block */
diff --git a/drivers/media/video/tiler/tiler-main.c b/drivers/media/video/tiler/tiler-main.c
index 85b657b..15897cf 100644
--- a/drivers/media/video/tiler/tiler-main.c
+++ b/drivers/media/video/tiler/tiler-main.c
@@ -424,8 +424,9 @@ static void release_gi(struct gid_info *gi)
 
 /* allocate an reserved area of size, alignment and link it to gi */
 /* leaves mutex locked to be able to add block to area */
-static struct area_info *area_new_m(u16 width, u16 height, u16 align,
-				  struct tcm *tcm, struct gid_info *gi)
+static struct area_info *area_new_m(enum tiler_fmt fmt, u16 width, u16 height,
+					u16 align, struct gid_info *gi,
+					u32 alloc_flags)
 {
 	struct area_info *ai = kmalloc(sizeof(*ai), GFP_KERNEL);
 	if (!ai)
@@ -436,12 +437,15 @@ static struct area_info *area_new_m(u16 width, u16 height, u16 align,
 	INIT_LIST_HEAD(&ai->blocks);
 
 	/* reserve an allocation area */
-	if (tcm_reserve_2d(tcm, width, height, align, &ai->area)) {
+	if (tcm_reserve_2d(tcm[fmt], width, height, align, &ai->area)) {
 		kfree(ai);
 		return NULL;
 	}
 
 	ai->gi = gi;
+	if (alloc_flags & FLAGS_ALLOC_NO_COLOCATE)
+		ai->allowed_modes |= 1 << fmt;
+
 	mutex_lock(&mtx);
 	list_add_tail(&ai->by_gid, &gi->areas);
 	return ai;
@@ -665,8 +669,9 @@ struct mem_info *_m_add2area(struct mem_info *mi, struct area_info *ai,
 	return mi;
 }
 
-static struct mem_info *get_2d_area(u16 w, u16 h, u16 align, u16 offs, u16 band,
-					struct gid_info *gi, struct tcm *tcm)
+static struct mem_info *get_2d_area(enum tiler_fmt fmt, u16 w, u16 h, u16 align,
+					u16 offs, u16 band, struct gid_info *gi,
+					u32 alloc_flags)
 {
 	struct area_info *ai = NULL;
 	struct mem_info *mi = NULL;
@@ -678,7 +683,7 @@ static struct mem_info *get_2d_area(u16 w, u16 h, u16 align, u16 offs, u16 band,
 	/* see if there is available prereserved space */
 	mutex_lock(&mtx);
 	list_for_each_entry(mi, &gi->reserved, global) {
-		if (mi->area.tcm == tcm &&
+		if (mi->area.tcm == tcm[fmt] &&
 		    tcm_aheight(mi->area) == h &&
 		    tcm_awidth(mi->area) == w &&
 		    (mi->area.p0.x & (align - 1)) == offs) {
@@ -710,8 +715,10 @@ static struct mem_info *get_2d_area(u16 w, u16 h, u16 align, u16 offs, u16 band,
 	/* this sets x, ai and before */
 	mutex_lock(&mtx);
 	list_for_each_entry(ai, &gi->areas, by_gid) {
-		if (ai->area.tcm == tcm &&
-		    tcm_aheight(ai->area) == h) {
+		if (ai->area.tcm == tcm[fmt] &&
+		    tcm_aheight(ai->area) == h &&
+		    (!ai->allowed_modes ||
+			(ai->allowed_modes & (1 << fmt)))) {
 			x = _m_blk_find_fit(w, align, offs, ai, &before);
 			if (x) {
 				_m_add2area(mi, ai, x - w, w, before);
@@ -732,8 +739,8 @@ static struct mem_info *get_2d_area(u16 w, u16 h, u16 align, u16 offs, u16 band,
 	mutex_unlock(&mtx);
 
 	/* if no area fit, reserve a new one */
-	ai = area_new_m(ALIGN(w + offs, max(band, align)), h,
-		      max(band, align), tcm, gi);
+	ai = area_new_m(fmt, ALIGN(w + offs, max(band, align)), h,
+		      max(band, align), gi, alloc_flags);
 	if (ai) {
 		_m_add2area(mi, ai, ai->area.p0.x + offs, w, &ai->blocks);
 		if (tiler_alloc_debug & 1)
@@ -768,8 +775,8 @@ static s32 lay_2d(enum tiler_fmt fmt, u16 n, u16 w, u16 h, u16 band,
 
 	/* calculate dimensions, band, offs, and alignment in slots */
 	/* reserve an area */
-	ai = area_new_m(ALIGN(w_res + offs, max(band, align)), h,
-			max(band, align), tcm[fmt], gi);
+	ai = area_new_m(fmt, ALIGN(w_res + offs, max(band, align)), h,
+			max(band, align), gi, 0);
 	if (!ai)
 		return -ENOMEM;
 
@@ -806,7 +813,7 @@ static s32 lay_nv12(int n, u16 w, u16 w1, u16 h, struct gid_info *gi, u8 *p)
 	struct list_head *pos;
 
 	/* reserve area */
-	ai = area_new_m(w, h, a, TILFMT_8BIT, gi);
+	ai = area_new_m(TILFMT_8BIT, w, h, a, gi, 0);
 	if (!ai)
 		return -ENOMEM;
 
@@ -1161,7 +1168,8 @@ static void fill_block_info(struct mem_info *i, struct tiler_block_info *blk)
  *  ==========================================================================
  */
 static struct mem_info *alloc_area(enum tiler_fmt fmt, u32 width, u32 height,
-				   struct gid_info *gi, u16 align, u16 offs)
+				   struct gid_info *gi, u16 align, u16 offs,
+					u32 alloc_flags)
 {
 	u16 x, y, band, remainder = 0;
 	struct mem_info *mi = NULL;
@@ -1193,7 +1201,7 @@ static struct mem_info *alloc_area(enum tiler_fmt fmt, u32 width, u32 height,
 		mi->parent = gi;
 		list_add(&mi->by_area, &gi->onedim);
 	} else {
-		mi = get_2d_area(x, y, align, offs, band, gi, tcm[fmt]);
+		mi = get_2d_area(fmt, x, y, align, offs, band, gi, alloc_flags);
 		if (!mi)
 			return NULL;
 
@@ -1218,6 +1226,7 @@ static struct mem_info *alloc_block_area(enum tiler_fmt fmt, u32 width,
 {
 	struct mem_info *mi = NULL;
 	struct gid_info *gi = NULL;
+	unsigned int alloc_flags = 0;
 
 	/* validate parameters */
 	if (!si || align > PAGE_SIZE || offs >= (align ? : PAGE_SIZE))
@@ -1231,8 +1240,11 @@ static struct mem_info *alloc_block_area(enum tiler_fmt fmt, u32 width,
 	if (!gi)
 		return ERR_PTR(-ENOMEM);
 
+	if (align && align < PAGE_SIZE)
+		alloc_flags |= FLAGS_ALLOC_NO_COLOCATE;
+
 	/* reserve area in tiler container */
-	mi = alloc_area(fmt, width, height, gi, align, offs);
+	mi = alloc_area(fmt, width, height, gi, align, offs, alloc_flags);
 	if (!mi) {
 		mutex_lock(&mtx);
 		gi->refs--;