drm/radeon/bo: add some fallback placements for VRAM only objects. (v2)

PATCH In review Dave Airlie 2 participants 6 messages Apr 27, 2010 – Apr 28, 2010

Subsystems (2)

DRM drivers Radeon and amdgpu DRM drivers

lore

Share & tools

lore

Dave Airlie

author maintainer

2010-04-27 02:34 UTC #0 lore raw

PATCH 1/1 drm/radeon/bo: add some fallback placements for VRAM only objects. (v2)

From: Dave Airlie <[email protected]>

On constrained r100 systems compiz would fail to start due to a lack of memory, we can just fallback place the objects rather than completely failing it works a lot better.

v2:
fixes issue identified by Michel when pinning could happen in a busy placement domain.

Signed-off-by Dave Airlie <[email protected]>

---
 drivers/gpu/drm/radeon/radeon.h        |    4 +++-
 drivers/gpu/drm/radeon/radeon_object.c |   22 +++++++++++++---------
 drivers/gpu/drm/radeon/radeon_ttm.c    |    6 +++---
 3 files changed, 19 insertions(+), 13 deletions(-)

drivers/gpu/drm/radeon/radeon.h +3 -1

index e912098..a23a6ab 100644
@@ -240,7 +240,9 @@ struct radeon_bo {
 	struct list_head		list;
 	/* Protected by tbo.reserved */
 	u32				placements[3];
+	u32				busy_placements[3];
 	struct ttm_placement		placement;
+	struct ttm_placement		busy_placement;
 	struct ttm_buffer_object	tbo;
 	struct ttm_bo_kmap_obj		kmap;
 	unsigned			pin_count;

@@ -1227,7 +1229,7 @@ extern void radeon_surface_init(struct radeon_device *rdev);
 extern int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data);
 extern void radeon_legacy_set_clock_gating(struct radeon_device *rdev, int enable);
 extern void radeon_atom_set_clock_gating(struct radeon_device *rdev, int enable);
-extern void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain);
+extern void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain, bool pinned);
 extern bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object *bo);
 extern void radeon_vram_location(struct radeon_device *rdev, struct radeon_mc *mc, u64 base);
 extern void radeon_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc);

drivers/gpu/drm/radeon/radeon_object.c +13 -9

index 6a8617b..ab3bc7b 100644
@@ -64,17 +64,21 @@ bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object *bo)
 	return false;
 }
 
-void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
+void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain, bool pinned)
 {
-	u32 c = 0;
+	u32 c = 0, b = 0;
 
 	rbo->placement.fpfn = 0;
 	rbo->placement.lpfn = 0;
 	rbo->placement.placement = rbo->placements;
-	rbo->placement.busy_placement = rbo->placements;
+	rbo->placement.busy_placement = rbo->busy_placements;
 	if (domain & RADEON_GEM_DOMAIN_VRAM)
 		rbo->placements[c++] = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED |
 					TTM_PL_FLAG_VRAM;
+	/* add busy placement to TTM if VRAM is only option */
+	if (domain == RADEON_GEM_DOMAIN_VRAM && pinned == false) {
+		rbo->busy_placements[b++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
+	}
 	if (domain & RADEON_GEM_DOMAIN_GTT)
 		rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
 	if (domain & RADEON_GEM_DOMAIN_CPU)

@@ -82,7 +86,7 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
 	if (!c)
 		rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM;
 	rbo->placement.num_placement = c;
-	rbo->placement.num_busy_placement = c;
+	rbo->placement.num_busy_placement = b;
 }
 
 int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj,

@@ -110,7 +114,7 @@ int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj,
 	bo->surface_reg = -1;
 	INIT_LIST_HEAD(&bo->list);
 
-	radeon_ttm_placement_from_domain(bo, domain);
+	radeon_ttm_placement_from_domain(bo, domain, false);
 	/* Kernel allocation are uninterruptible */
 	r = ttm_bo_init(&rdev->mman.bdev, &bo->tbo, size, type,
 			&bo->placement, 0, 0, !kernel, NULL, size,

@@ -185,7 +189,7 @@ int radeon_bo_pin(struct radeon_bo *bo, u32 domain, u64 *gpu_addr)
 			*gpu_addr = radeon_bo_gpu_offset(bo);
 		return 0;
 	}
-	radeon_ttm_placement_from_domain(bo, domain);
+	radeon_ttm_placement_from_domain(bo, domain, true);
 	if (domain == RADEON_GEM_DOMAIN_VRAM) {
 		/* force to pin into visible video ram */
 		bo->placement.lpfn = bo->rdev->mc.visible_vram_size >> PAGE_SHIFT;

@@ -325,10 +329,10 @@ int radeon_bo_list_validate(struct list_head *head)
 		if (!bo->pin_count) {
 			if (lobj->wdomain) {
 				radeon_ttm_placement_from_domain(bo,
-								lobj->wdomain);
+								 lobj->wdomain, false);
 			} else {
 				radeon_ttm_placement_from_domain(bo,
-								lobj->rdomain);
+								 lobj->rdomain, false);
 			}
 			r = ttm_bo_validate(&bo->tbo, &bo->placement,
 						true, false, false);

@@ -516,7 +520,7 @@ int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 		offset = bo->mem.mm_node->start << PAGE_SHIFT;
 		if ((offset + size) > rdev->mc.visible_vram_size) {
 			/* hurrah the memory is not visible ! */
-			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM);
+			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM, false);
 			rbo->placement.lpfn = rdev->mc.visible_vram_size >> PAGE_SHIFT;
 			r = ttm_bo_validate(bo, &rbo->placement, false, true, false);
 			if (unlikely(r != 0))

drivers/gpu/drm/radeon/radeon_ttm.c +3 -3

index af98f45..6557b44 100644
@@ -205,13 +205,13 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo,
 	switch (bo->mem.mem_type) {
 	case TTM_PL_VRAM:
 		if (rbo->rdev->cp.ready == false)
-			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_CPU);
+			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_CPU, false);
 		else
-			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_GTT);
+			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_GTT, false);
 		break;
 	case TTM_PL_TT:
 	default:
-		radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_CPU);
+		radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_CPU, false);
 	}
 	*placement = rbo->placement;
 }

Signature · 5 lines

1.6.6.1


------------------------------------------------------------------------------
--

Michel Dänzer

2010-04-27 08:31 UTC #1 lore raw

Quoted 8 lines · Dave Airlie

From: Dave Airlie <[email protected]>
On constrained r100 systems compiz would fail to start due to a lack of memory, we can just fallback place the objects rather than completely failing it works a lot better.
v2:
fixes issue identified by Michel when pinning could happen in a busy placement domain.

[...]

drivers/gpu/drm/radeon/radeon_object.c int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj, · line 120

Quoted 13 lines

drivers/gpu/drm/radeon/radeon_object.c +1 -1

index 6a8617b..ab3bc7b 100644
@@ -110,7 +114,7 @@ int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj,
 	bo->surface_reg = -1;
 	INIT_LIST_HEAD(&bo->list);
 
-	radeon_ttm_placement_from_domain(bo, domain);
+	radeon_ttm_placement_from_domain(bo, domain, false);
 	/* Kernel allocation are uninterruptible */
 	r = ttm_bo_init(&rdev->mman.bdev, &bo->tbo, size, type,
 			&bo->placement, 0, 0, !kernel, NULL, size,

Not sure it's a good idea to unconditionally allow BOs to be created in GTT if userspace requested VRAM. E.g. this would at least defeat the purpose of UploadToScreen if not turn it into pure overhead.

drivers/gpu/drm/radeon/radeon_object.c int radeon_bo_list_validate(struct list_head *head) · line 338

Quoted 13 lines

drivers/gpu/drm/radeon/radeon_object.c +2 -2

@@ -325,10 +329,10 @@ int radeon_bo_list_validate(struct list_head *head)
 		if (!bo->pin_count) {
 			if (lobj->wdomain) {
 				radeon_ttm_placement_from_domain(bo,
-								lobj->wdomain);
+								 lobj->wdomain, false);
 			} else {
 				radeon_ttm_placement_from_domain(bo,
-								lobj->rdomain);
+								 lobj->rdomain, false);
 			}
 			r = ttm_bo_validate(&bo->tbo, &bo->placement,
 						true, false, false);

How about something like the below (completely untested) instead? This should try to respect the userspace request first and only allow falling back from VRAM to GTT if that failed. (Maybe the new bool parameter shouldn't be called 'pinned' then but something like 'allow_vram_fallback' and its sense inverted)

drivers/gpu/drm/radeon/radeon_object.c +11 -3

index d10f246..b6f5177 100644
@@ -322,17 +322,25 @@ int radeon_bo_list_validate(struct list_head *head)
 	list_for_each_entry(lobj, head, list) {
 		bo = lobj->bo;
 		if (!bo->pin_count) {
+			bool pinned = true;
+
+retry:
 			if (lobj->wdomain) {
 				radeon_ttm_placement_from_domain(bo,
-								lobj->wdomain);
+								 lobj->wdomain, pinned);
 			} else {
 				radeon_ttm_placement_from_domain(bo,
-								lobj->rdomain);
+								 lobj->rdomain, pinned);
 			}
 			r = ttm_bo_validate(&bo->tbo, &bo->placement,
 						true, false);
-			if (unlikely(r))
+			if (unlikely(r)) {
+				if (pinned) {
+					pinned = false;
+					goto retry;
+				}
 				return r;
+			}
 		}
 		lobj->gpu_offset = radeon_bo_gpu_offset(bo);
 		lobj->tiling_flags = bo->tiling_flags;

drivers/gpu/drm/radeon/radeon_object.c int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo) · line 526

Quoted 9 lines

drivers/gpu/drm/radeon/radeon_object.c +1 -1

@@ -516,7 +520,7 @@ int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 		offset = bo->mem.mm_node->start << PAGE_SHIFT;
 		if ((offset + size) > rdev->mc.visible_vram_size) {
 			/* hurrah the memory is not visible ! */
-			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM);
+			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM, false);
 			rbo->placement.lpfn = rdev->mc.visible_vram_size >> PAGE_SHIFT;
 			r = ttm_bo_validate(bo, &rbo->placement, false, true, false);
 			if (unlikely(r != 0))

This will impose rbo->placement.lpfn for GTT as well, but it's only required for VRAM.

Signature · 9 lines

Earthling Michel Dänzer           |                http://www.vmware.com
Libre software enthusiast         |          Debian, X and DRI developer

------------------------------------------------------------------------------
--
_______________________________________________
Dri-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dri-devel

Michel Dänzer

2010-04-28 12:11 UTC #2 lore raw

Quoted 10 lines · Michel Dänzer

Quoted 8 lines · Dave Airlie
From: Dave Airlie <[email protected]>
On constrained r100 systems compiz would fail to start due to a lack of memory, we can just fallback place the objects rather than completely failing it works a lot better.
v2:
fixes issue identified by Michel when pinning could happen in a busy placement domain.

Quoted 7 lines

[...]
How about something like the below (completely untested) instead? This should try to respect the userspace request first and only allow falling back from VRAM to GTT if that failed. (Maybe the new bool parameter shouldn't be called 'pinned' then but something like 'allow_vram_fallback' and its sense inverted)

The alternative patch below is tested to not regress for me, but of course I'm not affected by the failure to put things in VRAM. Uwe / 'Chicken Shack', does this fix your problem with compiz?

This doesn't cover radeon_bo_fault_reserve_notify() because my tree doesn't have the invisible VRAM stuff yet, but it'll be easy to extend this approach to there as well, and it shouldn't affect the immediate problem we're trying to solve.

From c3009dd4df63a65fd15a485265fbdabfb1360c97 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Michel=20D=C3=A4nzer?= <[email protected]>
Date: Wed, 28 Apr 2010 09:24:38 +0200
Subject: [PATCH] drm/radeon/kms: Fall back to GTT if BO creation in / validation to VRAM failed.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Signed-off-by Michel Dänzer <[email protected]>

---
 drivers/gpu/drm/radeon/radeon_object.c |   31 ++++++++++++++++++++-----------
 1 files changed, 20 insertions(+), 11 deletions(-)

drivers/gpu/drm/radeon/radeon_object.c +20 ~~-11~~

index d10f246..1dfb056 100644
@@ -109,16 +109,22 @@ int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj,
 	bo->surface_reg = -1;
 	INIT_LIST_HEAD(&bo->list);
 
+retry:
 	radeon_ttm_placement_from_domain(bo, domain);
 	/* Kernel allocation are uninterruptible */
 	r = ttm_bo_init(&rdev->mman.bdev, &bo->tbo, size, type,
 			&bo->placement, 0, 0, !kernel, NULL, size,
 			&radeon_ttm_bo_destroy);
 	if (unlikely(r != 0)) {
-		if (r != -ERESTARTSYS)
+		if (r != -ERESTARTSYS) {
+			if (domain == RADEON_GEM_DOMAIN_VRAM) {
+				domain |= RADEON_GEM_DOMAIN_GTT;
+				goto retry;
+			}
 			dev_err(rdev->dev,
 				"object_init failed for (%lu, 0x%08X)\n",
 				size, domain);
+		}
 		return r;
 	}
 	*bo_ptr = bo;

@@ -313,6 +319,7 @@ int radeon_bo_list_validate(struct list_head *head)
 {
 	struct radeon_bo_list *lobj;
 	struct radeon_bo *bo;
+	u32 domain;
 	int r;
 
 	r = radeon_bo_list_reserve(head);

@@ -322,17 +329,19 @@ int radeon_bo_list_validate(struct list_head *head)
 	list_for_each_entry(lobj, head, list) {
 		bo = lobj->bo;
 		if (!bo->pin_count) {
-			if (lobj->wdomain) {
-				radeon_ttm_placement_from_domain(bo,
-								lobj->wdomain);
-			} else {
-				radeon_ttm_placement_from_domain(bo,
-								lobj->rdomain);
-			}
-			r = ttm_bo_validate(&bo->tbo, &bo->placement,
-						true, false);
-			if (unlikely(r))
+			domain = lobj->wdomain ? lobj->wdomain : lobj->rdomain;
+
+retry:
+			radeon_ttm_placement_from_domain(bo, domain);
+			r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false);
+			if (unlikely(r)) {
+				if (r != -ERESTARTSYS &&
+				    domain == RADEON_GEM_DOMAIN_VRAM) {
+					domain |= RADEON_GEM_DOMAIN_GTT;
+					goto retry;
+				}
 				return r;
+			}
 		}
 		lobj->gpu_offset = radeon_bo_gpu_offset(bo);
 		lobj->tiling_flags = bo->tiling_flags;

Signature · 11 lines

1.7.0.5



-- 
Earthling Michel Dänzer           |                http://www.vmware.com
Libre software enthusiast         |          Debian, X and DRI developer
_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel

Michel Dänzer

2010-04-27 10:11 UTC #3 lore raw

In reply to Dave Airlie

[ Moving to the new list ]

Quoted 8 lines · Dave Airlie

From: Dave Airlie <[email protected]>
On constrained r100 systems compiz would fail to start due to a lack of memory, we can just fallback place the objects rather than completely failing it works a lot better.
v2:
fixes issue identified by Michel when pinning could happen in a busy placement domain.

[...]

Quoted 30 lines

drivers/gpu/drm/radeon/radeon_object.c +7 -3

index 6a8617b..ab3bc7b 100644
@@ -64,17 +64,21 @@ bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object *bo)
 	return false;
 }
 
-void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
+void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain, bool pinned)
 {
-	u32 c = 0;
+	u32 c = 0, b = 0;
 
 	rbo->placement.fpfn = 0;
 	rbo->placement.lpfn = 0;
 	rbo->placement.placement = rbo->placements;
-	rbo->placement.busy_placement = rbo->placements;
+	rbo->placement.busy_placement = rbo->busy_placements;
 	if (domain & RADEON_GEM_DOMAIN_VRAM)
 		rbo->placements[c++] = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED |
 					TTM_PL_FLAG_VRAM;
+	/* add busy placement to TTM if VRAM is only option */
+	if (domain == RADEON_GEM_DOMAIN_VRAM && pinned == false) {
+		rbo->busy_placements[b++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
+	}
 	if (domain & RADEON_GEM_DOMAIN_GTT)
 		rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
 	if (domain & RADEON_GEM_DOMAIN_CPU)

drivers/gpu/drm/radeon/radeon_object.c void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) · line 92

Quoted 8 lines

drivers/gpu/drm/radeon/radeon_object.c +1 -1

@@ -82,7 +86,7 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
 	if (!c)
 		rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM;
 	rbo->placement.num_placement = c;
-	rbo->placement.num_busy_placement = c;
+	rbo->placement.num_busy_placement = b;
 }
 
 int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj,

BTW, this means there won't be any busy placements if (domain != RADEON_GEM_DOMAIN_VRAM || pinned). Not sure if that's a problem.

Signature · 6 lines

Earthling Michel Dänzer           |                http://www.vmware.com
Libre software enthusiast         |          Debian, X and DRI developer
_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel

Dave Airlie

author maintainer

2010-04-27 10:14 UTC #4 lore raw

2010/4/27 Michel Dänzer <[email protected]>:

Quoted 52 lines

[ Moving to the new list ]
Quoted 8 lines · Dave Airlie
From: Dave Airlie <[email protected]>
On constrained r100 systems compiz would fail to start due to a lack of memory, we can just fallback place the objects rather than completely failing it works a lot better.
v2:
fixes issue identified by Michel when pinning could happen in a busy placement domain.
[...]
Quoted 37 lines
drivers/gpu/drm/radeon/radeon_object.c +3 -3
index 6a8617b..ab3bc7b 100644
@@ -64,17 +64,21 @@ bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object *bo)
       return false;
  }
 
-void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
+void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain, bool pinned)
  {
-     u32 c = 0;
+     u32 c = 0, b = 0;
 
       rbo->placement.fpfn = 0;
       rbo->placement.lpfn = 0;
       rbo->placement.placement = rbo->placements;
-     rbo->placement.busy_placement = rbo->placements;
+     rbo->placement.busy_placement = rbo->busy_placements;
       if (domain & RADEON_GEM_DOMAIN_VRAM)
rbo->placements[c++] = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_VRAM;
+ /* add busy placement to TTM if VRAM is only option */
+ if (domain == RADEON_GEM_DOMAIN_VRAM && pinned == false) {
+ rbo->busy_placements[b++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
+ }
if (domain & RADEON_GEM_DOMAIN_GTT)
rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT; if (domain & RADEON_GEM_DOMAIN_CPU)
drivers/gpu/drm/radeon/radeon_object.c +1 -1
@@ -82,7 +86,7 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
       if (!c)
               rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM;
       rbo->placement.num_placement = c;
-     rbo->placement.num_busy_placement = c;
+     rbo->placement.num_busy_placement = b;
  }
 

Quoted 5 lines

drivers/gpu/drm/radeon/radeon_object.c void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) · line 92
Quoted 1 line
drivers/gpu/drm/radeon/radeon_object.c +0 -0
  int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj,
BTW, this means there won't be any busy placements if (domain != RADEON_GEM_DOMAIN_VRAM || pinned). Not sure if that's a problem.

Well there isn't really anywhere else to put things, I don't think falling GTT back to VRAM will ever be a case we have to worry about as much, its not like should be pinning anything major in GTT especially in the middle.

Falling back to system placement would sort of suck since the GPU can't access it.

Dave.

Michel Dänzer

2010-04-27 10:25 UTC #5 lore raw

Quoted 59 lines · Dave Airlie

2010/4/27 Michel Dänzer <[email protected]>:

Quoted 52 lines

[ Moving to the new list ]

Quoted 8 lines · Dave Airlie

From: Dave Airlie <[email protected]>
On constrained r100 systems compiz would fail to start due to a lack of memory, we can just fallback place the objects rather than completely failing it works a lot better.
v2:
fixes issue identified by Michel when pinning could happen in a busy placement domain.

[...]

Quoted 37 lines

drivers/gpu/drm/radeon/radeon_object.c +8 -4

index 6a8617b..ab3bc7b 100644
@@ -64,17 +64,21 @@ bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object *bo)
      return false;
 }
 
-void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
+void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain, bool pinned)
 {
-     u32 c = 0;
+     u32 c = 0, b = 0;
 
      rbo->placement.fpfn = 0;
      rbo->placement.lpfn = 0;
      rbo->placement.placement = rbo->placements;
-     rbo->placement.busy_placement = rbo->placements;
+     rbo->placement.busy_placement = rbo->busy_placements;
      if (domain & RADEON_GEM_DOMAIN_VRAM)
              rbo->placements[c++] = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED |
                                      TTM_PL_FLAG_VRAM;
+     /* add busy placement to TTM if VRAM is only option */
+     if (domain == RADEON_GEM_DOMAIN_VRAM && pinned == false) {
+             rbo->busy_placements[b++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
+     }
      if (domain & RADEON_GEM_DOMAIN_GTT)
              rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
      if (domain & RADEON_GEM_DOMAIN_CPU)

@@ -82,7 +86,7 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
      if (!c)
              rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM;
      rbo->placement.num_placement = c;
-     rbo->placement.num_busy_placement = c;
+     rbo->placement.num_busy_placement = b;
 }

Quoted 5 lines

drivers/gpu/drm/radeon/radeon_object.c void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) · line 92
Quoted 1 line
drivers/gpu/drm/radeon/radeon_object.c +0 -0
 int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj,
BTW, this means there won't be any busy placements if (domain != RADEON_GEM_DOMAIN_VRAM || pinned). Not sure if that's a problem.

Quoted 7 lines

Well there isn't really anywhere else to put things, I don't think falling GTT back to VRAM will ever be a case we have to worry about as much, its not like should be pinning anything major in GTT especially in the middle.
Falling back to system placement would sort of suck since the GPU can't access it.

Sure, but then the busy placements in the existing code would seem redundant as well, and I was wondering if there might have been a reason for them to be there anyway.

Signature · 6 lines

Earthling Michel Dänzer           |                http://www.vmware.com
Libre software enthusiast         |          Debian, X and DRI developer
_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel