* Rework TTMs busy handling
@ 2024-01-12 12:51 Christian König
2024-01-12 12:51 ` [PATCH 1/5] drm/vmwgfx: remove vmw_vram_gmr_placement Christian König
` (5 more replies)
0 siblings, 6 replies; 19+ messages in thread
From: Christian König @ 2024-01-12 12:51 UTC (permalink / raw)
To: dri-devel, intel-gfx, thomas.hellstrom, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
Hi guys,
same as the last time. Things I've changed:
Implemented the requirements from Zack to correctly fill in the busy
placements for VMWGFX.
Renamed the placement flags to desired and fallback as suggested by
Michel.
Rebased on drm-tip instead of drm-misc-next and fixed XE as well.
Please review and comment,
Christian.
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH 1/5] drm/vmwgfx: remove vmw_vram_gmr_placement
2024-01-12 12:51 Rework TTMs busy handling Christian König
@ 2024-01-12 12:51 ` Christian König
2024-01-12 12:51 ` [PATCH 2/5] drm/ttm: return ENOSPC from ttm_bo_mem_space Christian König
` (4 subsequent siblings)
5 siblings, 0 replies; 19+ messages in thread
From: Christian König @ 2024-01-12 12:51 UTC (permalink / raw)
To: dri-devel, intel-gfx, thomas.hellstrom, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
Seems to be unused.
Signed-off-by: Christian König <christian.koenig@amd.com>
---
drivers/gpu/drm/vmwgfx/vmwgfx_drv.h | 1 -
drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 28 ----------------------
2 files changed, 29 deletions(-)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
index 3cd5090dedfc..12efecc17df6 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
@@ -942,7 +942,6 @@ vmw_is_cursor_bypass3_enabled(const struct vmw_private *dev_priv)
extern const size_t vmw_tt_size;
extern struct ttm_placement vmw_vram_placement;
-extern struct ttm_placement vmw_vram_gmr_placement;
extern struct ttm_placement vmw_sys_placement;
extern struct ttm_device_funcs vmw_bo_driver;
extern const struct vmw_sg_table *
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index af8562c95cc3..a84fffcef8e1 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -43,13 +43,6 @@ static const struct ttm_place sys_placement_flags = {
.flags = 0
};
-static const struct ttm_place gmr_placement_flags = {
- .fpfn = 0,
- .lpfn = 0,
- .mem_type = VMW_PL_GMR,
- .flags = 0
-};
-
struct ttm_placement vmw_vram_placement = {
.num_placement = 1,
.placement = &vram_placement_flags,
@@ -57,27 +50,6 @@ struct ttm_placement vmw_vram_placement = {
.busy_placement = &vram_placement_flags
};
-static const struct ttm_place vram_gmr_placement_flags[] = {
- {
- .fpfn = 0,
- .lpfn = 0,
- .mem_type = TTM_PL_VRAM,
- .flags = 0
- }, {
- .fpfn = 0,
- .lpfn = 0,
- .mem_type = VMW_PL_GMR,
- .flags = 0
- }
-};
-
-struct ttm_placement vmw_vram_gmr_placement = {
- .num_placement = 2,
- .placement = vram_gmr_placement_flags,
- .num_busy_placement = 1,
- .busy_placement = &gmr_placement_flags
-};
-
struct ttm_placement vmw_sys_placement = {
.num_placement = 1,
.placement = &sys_placement_flags,
--
2.34.1
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH 2/5] drm/ttm: return ENOSPC from ttm_bo_mem_space
2024-01-12 12:51 Rework TTMs busy handling Christian König
2024-01-12 12:51 ` [PATCH 1/5] drm/vmwgfx: remove vmw_vram_gmr_placement Christian König
@ 2024-01-12 12:51 ` Christian König
2024-01-17 10:18 ` Thomas Hellström
2024-01-12 12:51 ` [PATCH 3/5] drm/ttm: replace busy placement with flags v6 Christian König
` (3 subsequent siblings)
5 siblings, 1 reply; 19+ messages in thread
From: Christian König @ 2024-01-12 12:51 UTC (permalink / raw)
To: dri-devel, intel-gfx, thomas.hellstrom, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
Only convert it to ENOMEM in ttm_bo_validate.
This allows ttm_bo_validate to distinct between an out of memory
situation and just out of space in a placement domain.
Signed-off-by: Christian König <christian.koenig@amd.com>
---
drivers/gpu/drm/ttm/ttm_bo.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index edf10618fe2b..8c1eaa74fa21 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -830,7 +830,7 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
goto error;
}
- ret = -ENOMEM;
+ ret = -ENOSPC;
if (!type_found) {
pr_err(TTM_PFX "No compatible memory type found\n");
ret = -EINVAL;
@@ -916,6 +916,9 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
return -EINVAL;
ret = ttm_bo_move_buffer(bo, placement, ctx);
+ /* For backward compatibility with userspace */
+ if (ret == -ENOSPC)
+ return -ENOMEM;
if (ret)
return ret;
--
2.34.1
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH 3/5] drm/ttm: replace busy placement with flags v6
2024-01-12 12:51 Rework TTMs busy handling Christian König
2024-01-12 12:51 ` [PATCH 1/5] drm/vmwgfx: remove vmw_vram_gmr_placement Christian König
2024-01-12 12:51 ` [PATCH 2/5] drm/ttm: return ENOSPC from ttm_bo_mem_space Christian König
@ 2024-01-12 12:51 ` Christian König
2024-01-17 8:01 ` Thomas Zimmermann
2024-01-17 10:47 ` Thomas Hellström
2024-01-12 12:51 ` [PATCH 4/5] drm/ttm: improve idle/busy handling v3 Christian König
` (2 subsequent siblings)
5 siblings, 2 replies; 19+ messages in thread
From: Christian König @ 2024-01-12 12:51 UTC (permalink / raw)
To: dri-devel, intel-gfx, thomas.hellstrom, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
From: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
Instead of a list of separate busy placement add flags which indicate
that a placement should only be used when there is room or if we need to
evict.
v2: add missing TTM_PL_FLAG_IDLE for i915
v3: fix auto build test ERROR on drm-tip/drm-tip
v4: fix some typos pointed out by checkpatch
v5: cleanup some rebase problems with VMWGFX
v6: implement some missing VMWGFX functionality pointed out by Zack,
rename the flags as suggested by Michel, rebase on drm-tip and
adjust XE as well
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 6 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 +---
drivers/gpu/drm/drm_gem_vram_helper.c | 2 -
drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 37 +++++------
drivers/gpu/drm/loongson/lsdc_ttm.c | 2 -
drivers/gpu/drm/nouveau/nouveau_bo.c | 59 +++++++----------
drivers/gpu/drm/nouveau/nouveau_bo.h | 1 -
drivers/gpu/drm/qxl/qxl_object.c | 2 -
drivers/gpu/drm/qxl/qxl_ttm.c | 2 -
drivers/gpu/drm/radeon/radeon_object.c | 2 -
drivers/gpu/drm/radeon/radeon_ttm.c | 8 +--
drivers/gpu/drm/radeon/radeon_uvd.c | 1 -
drivers/gpu/drm/ttm/ttm_bo.c | 21 ++++---
drivers/gpu/drm/ttm/ttm_resource.c | 73 +++++-----------------
drivers/gpu/drm/vmwgfx/vmwgfx_bo.c | 33 +++++++---
drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 4 --
drivers/gpu/drm/xe/xe_bo.c | 33 +++++-----
include/drm/ttm/ttm_placement.h | 10 +--
include/drm/ttm/ttm_resource.h | 8 +--
19 files changed, 118 insertions(+), 197 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index 425cebcc5cbf..b671b0665492 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -220,9 +220,6 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
placement->num_placement = c;
placement->placement = places;
-
- placement->num_busy_placement = c;
- placement->busy_placement = places;
}
/**
@@ -1397,8 +1394,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
AMDGPU_GEM_DOMAIN_GTT);
/* Avoid costly evictions; only set GTT as a busy placement */
- abo->placement.num_busy_placement = 1;
- abo->placement.busy_placement = &abo->placements[1];
+ abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
r = ttm_bo_validate(bo, &abo->placement, &ctx);
if (unlikely(r == -EBUSY || r == -ERESTARTSYS))
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 75c9fd2c6c2a..8722beba494e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -102,23 +102,19 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
/* Don't handle scatter gather BOs */
if (bo->type == ttm_bo_type_sg) {
placement->num_placement = 0;
- placement->num_busy_placement = 0;
return;
}
/* Object isn't an AMDGPU object so ignore */
if (!amdgpu_bo_is_amdgpu_bo(bo)) {
placement->placement = &placements;
- placement->busy_placement = &placements;
placement->num_placement = 1;
- placement->num_busy_placement = 1;
return;
}
abo = ttm_to_amdgpu_bo(bo);
if (abo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) {
placement->num_placement = 0;
- placement->num_busy_placement = 0;
return;
}
@@ -128,13 +124,13 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
case AMDGPU_PL_OA:
case AMDGPU_PL_DOORBELL:
placement->num_placement = 0;
- placement->num_busy_placement = 0;
return;
case TTM_PL_VRAM:
if (!adev->mman.buffer_funcs_enabled) {
/* Move to system memory */
amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU);
+
} else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
!(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) &&
amdgpu_bo_in_cpu_visible_vram(abo)) {
@@ -149,8 +145,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
AMDGPU_GEM_DOMAIN_CPU);
abo->placements[0].fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
abo->placements[0].lpfn = 0;
- abo->placement.busy_placement = &abo->placements[1];
- abo->placement.num_busy_placement = 1;
+ abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
} else {
/* Move to GTT memory */
amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT |
@@ -966,8 +961,6 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
/* allocate GART space */
placement.num_placement = 1;
placement.placement = &placements;
- placement.num_busy_placement = 1;
- placement.busy_placement = &placements;
placements.fpfn = 0;
placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
placements.mem_type = TTM_PL_TT;
diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
index b67eafa55715..75f2eaf0d5b6 100644
--- a/drivers/gpu/drm/drm_gem_vram_helper.c
+++ b/drivers/gpu/drm/drm_gem_vram_helper.c
@@ -147,7 +147,6 @@ static void drm_gem_vram_placement(struct drm_gem_vram_object *gbo,
invariant_flags = TTM_PL_FLAG_TOPDOWN;
gbo->placement.placement = gbo->placements;
- gbo->placement.busy_placement = gbo->placements;
if (pl_flag & DRM_GEM_VRAM_PL_FLAG_VRAM) {
gbo->placements[c].mem_type = TTM_PL_VRAM;
@@ -160,7 +159,6 @@ static void drm_gem_vram_placement(struct drm_gem_vram_object *gbo,
}
gbo->placement.num_placement = c;
- gbo->placement.num_busy_placement = c;
for (i = 0; i < c; ++i) {
gbo->placements[i].fpfn = 0;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
index 6b69ef0cdbb4..80c6cafc8887 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
@@ -65,8 +65,6 @@ static const struct ttm_place sys_placement_flags = {
static struct ttm_placement i915_sys_placement = {
.num_placement = 1,
.placement = &sys_placement_flags,
- .num_busy_placement = 1,
- .busy_placement = &sys_placement_flags,
};
/**
@@ -157,32 +155,28 @@ i915_ttm_place_from_region(const struct intel_memory_region *mr,
static void
i915_ttm_placement_from_obj(const struct drm_i915_gem_object *obj,
- struct ttm_place *requested,
- struct ttm_place *busy,
+ struct ttm_place *places,
struct ttm_placement *placement)
{
unsigned int num_allowed = obj->mm.n_placements;
unsigned int flags = obj->flags;
unsigned int i;
- placement->num_placement = 1;
+ places[0].flags |= TTM_PL_FLAG_DESIRED;
i915_ttm_place_from_region(num_allowed ? obj->mm.placements[0] :
- obj->mm.region, requested, obj->bo_offset,
+ obj->mm.region, &places[0], obj->bo_offset,
obj->base.size, flags);
/* Cache this on object? */
- placement->num_busy_placement = num_allowed;
- for (i = 0; i < placement->num_busy_placement; ++i)
- i915_ttm_place_from_region(obj->mm.placements[i], busy + i,
- obj->bo_offset, obj->base.size, flags);
-
- if (num_allowed == 0) {
- *busy = *requested;
- placement->num_busy_placement = 1;
+ for (i = 0; i < num_allowed; ++i) {
+ i915_ttm_place_from_region(obj->mm.placements[i],
+ &places[i + 1], obj->bo_offset,
+ obj->base.size, flags);
+ places[i + 1].flags |= TTM_PL_FLAG_FALLBACK;
}
- placement->placement = requested;
- placement->busy_placement = busy;
+ placement->num_placement = num_allowed + 1;
+ placement->placement = places;
}
static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev,
@@ -789,7 +783,8 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj,
int ret;
/* First try only the requested placement. No eviction. */
- real_num_busy = fetch_and_zero(&placement->num_busy_placement);
+ real_num_busy = placement->num_placement;
+ placement->num_placement = 1;
ret = ttm_bo_validate(bo, placement, &ctx);
if (ret) {
ret = i915_ttm_err_to_gem(ret);
@@ -805,7 +800,7 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj,
* If the initial attempt fails, allow all accepted placements,
* evicting if necessary.
*/
- placement->num_busy_placement = real_num_busy;
+ placement->num_placement = real_num_busy;
ret = ttm_bo_validate(bo, placement, &ctx);
if (ret)
return i915_ttm_err_to_gem(ret);
@@ -839,7 +834,7 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj,
static int i915_ttm_get_pages(struct drm_i915_gem_object *obj)
{
- struct ttm_place requested, busy[I915_TTM_MAX_PLACEMENTS];
+ struct ttm_place places[I915_TTM_MAX_PLACEMENTS + 1];
struct ttm_placement placement;
/* restricted by sg_alloc_table */
@@ -849,7 +844,7 @@ static int i915_ttm_get_pages(struct drm_i915_gem_object *obj)
GEM_BUG_ON(obj->mm.n_placements > I915_TTM_MAX_PLACEMENTS);
/* Move to the requested placement. */
- i915_ttm_placement_from_obj(obj, &requested, busy, &placement);
+ i915_ttm_placement_from_obj(obj, places, &placement);
return __i915_ttm_get_pages(obj, &placement);
}
@@ -879,9 +874,7 @@ static int __i915_ttm_migrate(struct drm_i915_gem_object *obj,
i915_ttm_place_from_region(mr, &requested, obj->bo_offset,
obj->base.size, flags);
placement.num_placement = 1;
- placement.num_busy_placement = 1;
placement.placement = &requested;
- placement.busy_placement = &requested;
ret = __i915_ttm_get_pages(obj, &placement);
if (ret)
diff --git a/drivers/gpu/drm/loongson/lsdc_ttm.c b/drivers/gpu/drm/loongson/lsdc_ttm.c
index bf79dc55afa4..465f622ac05d 100644
--- a/drivers/gpu/drm/loongson/lsdc_ttm.c
+++ b/drivers/gpu/drm/loongson/lsdc_ttm.c
@@ -54,7 +54,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo *lbo, u32 domain)
pflags |= TTM_PL_FLAG_TOPDOWN;
lbo->placement.placement = lbo->placements;
- lbo->placement.busy_placement = lbo->placements;
if (domain & LSDC_GEM_DOMAIN_VRAM) {
lbo->placements[c].mem_type = TTM_PL_VRAM;
@@ -77,7 +76,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo *lbo, u32 domain)
}
lbo->placement.num_placement = c;
- lbo->placement.num_busy_placement = c;
for (i = 0; i < c; ++i) {
lbo->placements[i].fpfn = 0;
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 00cc7d1abaa3..56dcd25db1ce 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -404,27 +404,6 @@ nouveau_bo_new(struct nouveau_cli *cli, u64 size, int align,
return 0;
}
-static void
-set_placement_list(struct ttm_place *pl, unsigned *n, uint32_t domain)
-{
- *n = 0;
-
- if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
- pl[*n].mem_type = TTM_PL_VRAM;
- pl[*n].flags = 0;
- (*n)++;
- }
- if (domain & NOUVEAU_GEM_DOMAIN_GART) {
- pl[*n].mem_type = TTM_PL_TT;
- pl[*n].flags = 0;
- (*n)++;
- }
- if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
- pl[*n].mem_type = TTM_PL_SYSTEM;
- pl[(*n)++].flags = 0;
- }
-}
-
static void
set_placement_range(struct nouveau_bo *nvbo, uint32_t domain)
{
@@ -452,10 +431,6 @@ set_placement_range(struct nouveau_bo *nvbo, uint32_t domain)
nvbo->placements[i].fpfn = fpfn;
nvbo->placements[i].lpfn = lpfn;
}
- for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
- nvbo->busy_placements[i].fpfn = fpfn;
- nvbo->busy_placements[i].lpfn = lpfn;
- }
}
}
@@ -463,15 +438,32 @@ void
nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t domain,
uint32_t busy)
{
- struct ttm_placement *pl = &nvbo->placement;
+ unsigned int *n = &nvbo->placement.num_placement;
+ struct ttm_place *pl = nvbo->placements;
- pl->placement = nvbo->placements;
- set_placement_list(nvbo->placements, &pl->num_placement, domain);
+ domain |= busy;
- pl->busy_placement = nvbo->busy_placements;
- set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
- domain | busy);
+ *n = 0;
+ if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
+ pl[*n].mem_type = TTM_PL_VRAM;
+ pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_VRAM ?
+ TTM_PL_FLAG_FALLBACK : 0;
+ (*n)++;
+ }
+ if (domain & NOUVEAU_GEM_DOMAIN_GART) {
+ pl[*n].mem_type = TTM_PL_TT;
+ pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_GART ?
+ TTM_PL_FLAG_FALLBACK : 0;
+ (*n)++;
+ }
+ if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
+ pl[*n].mem_type = TTM_PL_SYSTEM;
+ pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_CPU ?
+ TTM_PL_FLAG_FALLBACK : 0;
+ (*n)++;
+ }
+ nvbo->placement.placement = nvbo->placements;
set_placement_range(nvbo, domain);
}
@@ -1314,11 +1306,6 @@ vm_fault_t nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
nvbo->placements[i].lpfn = mappable;
}
- for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
- nvbo->busy_placements[i].fpfn = 0;
- nvbo->busy_placements[i].lpfn = mappable;
- }
-
nouveau_bo_placement_set(nvbo, NOUVEAU_GEM_DOMAIN_VRAM, 0);
}
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.h b/drivers/gpu/drm/nouveau/nouveau_bo.h
index 70c551921a9e..e9dfab6a8156 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.h
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.h
@@ -15,7 +15,6 @@ struct nouveau_bo {
struct ttm_placement placement;
u32 valid_domains;
struct ttm_place placements[3];
- struct ttm_place busy_placements[3];
bool force_coherent;
struct ttm_bo_kmap_obj kmap;
struct list_head head;
diff --git a/drivers/gpu/drm/qxl/qxl_object.c b/drivers/gpu/drm/qxl/qxl_object.c
index 06a58dad5f5c..1e46b0a6e478 100644
--- a/drivers/gpu/drm/qxl/qxl_object.c
+++ b/drivers/gpu/drm/qxl/qxl_object.c
@@ -66,7 +66,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo *qbo, u32 domain)
pflag |= TTM_PL_FLAG_TOPDOWN;
qbo->placement.placement = qbo->placements;
- qbo->placement.busy_placement = qbo->placements;
if (domain == QXL_GEM_DOMAIN_VRAM) {
qbo->placements[c].mem_type = TTM_PL_VRAM;
qbo->placements[c++].flags = pflag;
@@ -86,7 +85,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo *qbo, u32 domain)
qbo->placements[c++].flags = 0;
}
qbo->placement.num_placement = c;
- qbo->placement.num_busy_placement = c;
for (i = 0; i < c; ++i) {
qbo->placements[i].fpfn = 0;
qbo->placements[i].lpfn = 0;
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index 1a82629bce3f..765a144cea14 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -60,9 +60,7 @@ static void qxl_evict_flags(struct ttm_buffer_object *bo,
if (!qxl_ttm_bo_is_qxl_bo(bo)) {
placement->placement = &placements;
- placement->busy_placement = &placements;
placement->num_placement = 1;
- placement->num_busy_placement = 1;
return;
}
qbo = to_qxl_bo(bo);
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index 10c0fbd9d2b4..a955f8a2f7fe 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -78,7 +78,6 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
u32 c = 0, i;
rbo->placement.placement = rbo->placements;
- rbo->placement.busy_placement = rbo->placements;
if (domain & RADEON_GEM_DOMAIN_VRAM) {
/* Try placing BOs which don't need CPU access outside of the
* CPU accessible part of VRAM
@@ -114,7 +113,6 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
}
rbo->placement.num_placement = c;
- rbo->placement.num_busy_placement = c;
for (i = 0; i < c; ++i) {
if ((rbo->flags & RADEON_GEM_CPU_ACCESS) &&
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index de4e6d78f1e1..2078b0000e22 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -92,9 +92,7 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo,
if (!radeon_ttm_bo_is_radeon_bo(bo)) {
placement->placement = &placements;
- placement->busy_placement = &placements;
placement->num_placement = 1;
- placement->num_busy_placement = 1;
return;
}
rbo = container_of(bo, struct radeon_bo, tbo);
@@ -114,15 +112,11 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo,
*/
radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM |
RADEON_GEM_DOMAIN_GTT);
- rbo->placement.num_busy_placement = 0;
for (i = 0; i < rbo->placement.num_placement; i++) {
if (rbo->placements[i].mem_type == TTM_PL_VRAM) {
if (rbo->placements[i].fpfn < fpfn)
rbo->placements[i].fpfn = fpfn;
- } else {
- rbo->placement.busy_placement =
- &rbo->placements[i];
- rbo->placement.num_busy_placement = 1;
+ rbo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
}
}
} else
diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c b/drivers/gpu/drm/radeon/radeon_uvd.c
index a2cda184b2b2..058a1c8451b2 100644
--- a/drivers/gpu/drm/radeon/radeon_uvd.c
+++ b/drivers/gpu/drm/radeon/radeon_uvd.c
@@ -324,7 +324,6 @@ void radeon_uvd_force_into_uvd_segment(struct radeon_bo *rbo,
rbo->placements[1].fpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
rbo->placements[1].lpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
rbo->placement.num_placement++;
- rbo->placement.num_busy_placement++;
}
void radeon_uvd_free_handles(struct radeon_device *rdev, struct drm_file *filp)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 8c1eaa74fa21..a5e11a92e0b9 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -410,8 +410,8 @@ static int ttm_bo_bounce_temp_buffer(struct ttm_buffer_object *bo,
struct ttm_resource *hop_mem;
int ret;
- hop_placement.num_placement = hop_placement.num_busy_placement = 1;
- hop_placement.placement = hop_placement.busy_placement = hop;
+ hop_placement.num_placement = 1;
+ hop_placement.placement = hop;
/* find space in the bounce domain */
ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
@@ -440,10 +440,9 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo,
dma_resv_assert_held(bo->base.resv);
placement.num_placement = 0;
- placement.num_busy_placement = 0;
bdev->funcs->evict_flags(bo, &placement);
- if (!placement.num_placement && !placement.num_busy_placement) {
+ if (!placement.num_placement) {
ret = ttm_bo_wait_ctx(bo, ctx);
if (ret)
return ret;
@@ -791,6 +790,9 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
const struct ttm_place *place = &placement->placement[i];
struct ttm_resource_manager *man;
+ if (place->flags & TTM_PL_FLAG_FALLBACK)
+ continue;
+
man = ttm_manager_type(bdev, place->mem_type);
if (!man || !ttm_resource_manager_used(man))
continue;
@@ -813,10 +815,13 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
return 0;
}
- for (i = 0; i < placement->num_busy_placement; ++i) {
- const struct ttm_place *place = &placement->busy_placement[i];
+ for (i = 0; i < placement->num_placement; ++i) {
+ const struct ttm_place *place = &placement->placement[i];
struct ttm_resource_manager *man;
+ if (place->flags & TTM_PL_FLAG_DESIRED)
+ continue;
+
man = ttm_manager_type(bdev, place->mem_type);
if (!man || !ttm_resource_manager_used(man))
continue;
@@ -904,11 +909,11 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
/*
* Remove the backing store if no placement is given.
*/
- if (!placement->num_placement && !placement->num_busy_placement)
+ if (!placement->num_placement)
return ttm_bo_pipeline_gutting(bo);
/* Check whether we need to move buffer. */
- if (bo->resource && ttm_resource_compat(bo->resource, placement))
+ if (bo->resource && ttm_resource_compatible(bo->resource, placement))
return 0;
/* Moving of pinned BOs is forbidden */
diff --git a/drivers/gpu/drm/ttm/ttm_resource.c b/drivers/gpu/drm/ttm/ttm_resource.c
index 02b96d23fdb9..fb14f7716cf8 100644
--- a/drivers/gpu/drm/ttm/ttm_resource.c
+++ b/drivers/gpu/drm/ttm/ttm_resource.c
@@ -291,37 +291,15 @@ bool ttm_resource_intersects(struct ttm_device *bdev,
}
/**
- * ttm_resource_compatible - test for compatibility
+ * ttm_resource_compatible - check if resource is compatible with placement
*
- * @bdev: TTM device structure
- * @res: The resource to test
- * @place: The placement to test
- * @size: How many bytes the new allocation needs.
- *
- * Test if @res compatible with @place and @size.
+ * @res: the resource to check
+ * @placement: the placement to check against
*
- * Returns true if the res placement compatible with @place and @size.
+ * Returns true if the placement is compatible.
*/
-bool ttm_resource_compatible(struct ttm_device *bdev,
- struct ttm_resource *res,
- const struct ttm_place *place,
- size_t size)
-{
- struct ttm_resource_manager *man;
-
- if (!res || !place)
- return false;
-
- man = ttm_manager_type(bdev, res->mem_type);
- if (!man->func->compatible)
- return true;
-
- return man->func->compatible(man, res, place, size);
-}
-
-static bool ttm_resource_places_compat(struct ttm_resource *res,
- const struct ttm_place *places,
- unsigned num_placement)
+bool ttm_resource_compatible(struct ttm_resource *res,
+ struct ttm_placement *placement)
{
struct ttm_buffer_object *bo = res->bo;
struct ttm_device *bdev = bo->bdev;
@@ -330,44 +308,25 @@ static bool ttm_resource_places_compat(struct ttm_resource *res,
if (res->placement & TTM_PL_FLAG_TEMPORARY)
return false;
- for (i = 0; i < num_placement; i++) {
- const struct ttm_place *heap = &places[i];
+ for (i = 0; i < placement->num_placement; i++) {
+ const struct ttm_place *place = &placement->placement[i];
+ struct ttm_resource_manager *man;
- if (!ttm_resource_compatible(bdev, res, heap, bo->base.size))
+ if (res->mem_type != place->mem_type)
+ continue;
+
+ man = ttm_manager_type(bdev, res->mem_type);
+ if (man->func->compatible &&
+ !man->func->compatible(man, res, place, bo->base.size))
continue;
- if ((res->mem_type == heap->mem_type) &&
- (!(heap->flags & TTM_PL_FLAG_CONTIGUOUS) ||
+ if ((!(place->flags & TTM_PL_FLAG_CONTIGUOUS) ||
(res->placement & TTM_PL_FLAG_CONTIGUOUS)))
return true;
}
return false;
}
-/**
- * ttm_resource_compat - check if resource is compatible with placement
- *
- * @res: the resource to check
- * @placement: the placement to check against
- *
- * Returns true if the placement is compatible.
- */
-bool ttm_resource_compat(struct ttm_resource *res,
- struct ttm_placement *placement)
-{
- if (ttm_resource_places_compat(res, placement->placement,
- placement->num_placement))
- return true;
-
- if ((placement->busy_placement != placement->placement ||
- placement->num_busy_placement > placement->num_placement) &&
- ttm_resource_places_compat(res, placement->busy_placement,
- placement->num_busy_placement))
- return true;
-
- return false;
-}
-
void ttm_resource_set_bo(struct ttm_resource *res,
struct ttm_buffer_object *bo)
{
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
index 2bfac3aad7b7..bfd41ce3c8f4 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
@@ -742,9 +742,21 @@ void vmw_bo_move_notify(struct ttm_buffer_object *bo,
vmw_resource_unbind_list(vbo);
}
+static u32 placement_flags(u32 domain, u32 desired, u32 fallback)
+{
+ if (desired & fallback & domain)
+ return 0;
+
+ if (desired & domain)
+ return TTM_PL_FLAG_DESIRED;
+
+ return TTM_PL_FLAG_FALLBACK;
+}
+
static u32
-set_placement_list(struct ttm_place *pl, u32 domain)
+set_placement_list(struct ttm_place *pl, u32 desired, u32 fallback)
{
+ u32 domain = desired | fallback;
u32 n = 0;
/*
@@ -752,35 +764,40 @@ set_placement_list(struct ttm_place *pl, u32 domain)
*/
if (domain & VMW_BO_DOMAIN_MOB) {
pl[n].mem_type = VMW_PL_MOB;
- pl[n].flags = 0;
+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_MOB, desired,
+ fallback);
pl[n].fpfn = 0;
pl[n].lpfn = 0;
n++;
}
if (domain & VMW_BO_DOMAIN_GMR) {
pl[n].mem_type = VMW_PL_GMR;
- pl[n].flags = 0;
+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_GMR, desired,
+ fallback);
pl[n].fpfn = 0;
pl[n].lpfn = 0;
n++;
}
if (domain & VMW_BO_DOMAIN_VRAM) {
pl[n].mem_type = TTM_PL_VRAM;
- pl[n].flags = 0;
+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_VRAM, desired,
+ fallback);
pl[n].fpfn = 0;
pl[n].lpfn = 0;
n++;
}
if (domain & VMW_BO_DOMAIN_WAITABLE_SYS) {
pl[n].mem_type = VMW_PL_SYSTEM;
- pl[n].flags = 0;
+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_WAITABLE_SYS,
+ desired, fallback);
pl[n].fpfn = 0;
pl[n].lpfn = 0;
n++;
}
if (domain & VMW_BO_DOMAIN_SYS) {
pl[n].mem_type = TTM_PL_SYSTEM;
- pl[n].flags = 0;
+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_SYS, desired,
+ fallback);
pl[n].fpfn = 0;
pl[n].lpfn = 0;
n++;
@@ -806,7 +823,7 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32 domain, u32 busy_domain)
u32 i;
pl->placement = bo->places;
- pl->num_placement = set_placement_list(bo->places, domain);
+ pl->num_placement = set_placement_list(bo->places, domain, busy_domain);
if (drm_debug_enabled(DRM_UT_DRIVER) && bo->tbo.resource) {
for (i = 0; i < pl->num_placement; ++i) {
@@ -821,8 +838,6 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32 domain, u32 busy_domain)
__func__, bo->tbo.resource->mem_type, domain);
}
- pl->busy_placement = bo->busy_places;
- pl->num_busy_placement = set_placement_list(bo->busy_places, busy_domain);
}
void vmw_bo_placement_set_default_accelerated(struct vmw_bo *bo)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index a84fffcef8e1..4d23d0a70bcb 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -46,15 +46,11 @@ static const struct ttm_place sys_placement_flags = {
struct ttm_placement vmw_vram_placement = {
.num_placement = 1,
.placement = &vram_placement_flags,
- .num_busy_placement = 1,
- .busy_placement = &vram_placement_flags
};
struct ttm_placement vmw_sys_placement = {
.num_placement = 1,
.placement = &sys_placement_flags,
- .num_busy_placement = 1,
- .busy_placement = &sys_placement_flags
};
const size_t vmw_tt_size = sizeof(struct vmw_ttm_tt);
diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
index 8e4a3b1f6b93..1e5edc7f0b85 100644
--- a/drivers/gpu/drm/xe/xe_bo.c
+++ b/drivers/gpu/drm/xe/xe_bo.c
@@ -38,22 +38,26 @@ static const struct ttm_place sys_placement_flags = {
static struct ttm_placement sys_placement = {
.num_placement = 1,
.placement = &sys_placement_flags,
- .num_busy_placement = 1,
- .busy_placement = &sys_placement_flags,
};
-static const struct ttm_place tt_placement_flags = {
- .fpfn = 0,
- .lpfn = 0,
- .mem_type = XE_PL_TT,
- .flags = 0,
+static const struct ttm_place tt_placement_flags[] = {
+ {
+ .fpfn = 0,
+ .lpfn = 0,
+ .mem_type = XE_PL_TT,
+ .flags = TTM_PL_FLAG_DESIRED,
+ },
+ {
+ .fpfn = 0,
+ .lpfn = 0,
+ .mem_type = XE_PL_SYSTEM,
+ .flags = TTM_PL_FLAG_FALLBACK,
+ }
};
static struct ttm_placement tt_placement = {
- .num_placement = 1,
- .placement = &tt_placement_flags,
- .num_busy_placement = 1,
- .busy_placement = &sys_placement_flags,
+ .num_placement = 2,
+ .placement = tt_placement_flags,
};
bool mem_type_is_vram(u32 mem_type)
@@ -230,8 +234,6 @@ static int __xe_bo_placement_for_flags(struct xe_device *xe, struct xe_bo *bo,
bo->placement = (struct ttm_placement) {
.num_placement = c,
.placement = bo->placements,
- .num_busy_placement = c,
- .busy_placement = bo->placements,
};
return 0;
@@ -251,7 +253,6 @@ static void xe_evict_flags(struct ttm_buffer_object *tbo,
/* Don't handle scatter gather BOs */
if (tbo->type == ttm_bo_type_sg) {
placement->num_placement = 0;
- placement->num_busy_placement = 0;
return;
}
@@ -1353,8 +1354,6 @@ static int __xe_bo_fixed_placement(struct xe_device *xe,
bo->placement = (struct ttm_placement) {
.num_placement = 1,
.placement = place,
- .num_busy_placement = 1,
- .busy_placement = place,
};
return 0;
@@ -2112,9 +2111,7 @@ int xe_bo_migrate(struct xe_bo *bo, u32 mem_type)
xe_place_from_ttm_type(mem_type, &requested);
placement.num_placement = 1;
- placement.num_busy_placement = 1;
placement.placement = &requested;
- placement.busy_placement = &requested;
/*
* Stolen needs to be handled like below VRAM handling if we ever need
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h
index 8074d0f6cae5..b510a4812609 100644
--- a/include/drm/ttm/ttm_placement.h
+++ b/include/drm/ttm/ttm_placement.h
@@ -64,6 +64,12 @@
/* For multihop handling */
#define TTM_PL_FLAG_TEMPORARY (1 << 2)
+/* Placement is never used during eviction */
+#define TTM_PL_FLAG_DESIRED (1 << 3)
+
+/* Placement is only used during eviction */
+#define TTM_PL_FLAG_FALLBACK (1 << 4)
+
/**
* struct ttm_place
*
@@ -86,16 +92,12 @@ struct ttm_place {
*
* @num_placement: number of preferred placements
* @placement: preferred placements
- * @num_busy_placement: number of preferred placements when need to evict buffer
- * @busy_placement: preferred placements when need to evict buffer
*
* Structure indicating the placement you request for an object.
*/
struct ttm_placement {
unsigned num_placement;
const struct ttm_place *placement;
- unsigned num_busy_placement;
- const struct ttm_place *busy_placement;
};
#endif
diff --git a/include/drm/ttm/ttm_resource.h b/include/drm/ttm/ttm_resource.h
index 78a226eba953..1afa13f0c22b 100644
--- a/include/drm/ttm/ttm_resource.h
+++ b/include/drm/ttm/ttm_resource.h
@@ -365,12 +365,8 @@ bool ttm_resource_intersects(struct ttm_device *bdev,
struct ttm_resource *res,
const struct ttm_place *place,
size_t size);
-bool ttm_resource_compatible(struct ttm_device *bdev,
- struct ttm_resource *res,
- const struct ttm_place *place,
- size_t size);
-bool ttm_resource_compat(struct ttm_resource *res,
- struct ttm_placement *placement);
+bool ttm_resource_compatible(struct ttm_resource *res,
+ struct ttm_placement *placement);
void ttm_resource_set_bo(struct ttm_resource *res,
struct ttm_buffer_object *bo);
--
2.34.1
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH 4/5] drm/ttm: improve idle/busy handling v3
2024-01-12 12:51 Rework TTMs busy handling Christian König
` (2 preceding siblings ...)
2024-01-12 12:51 ` [PATCH 3/5] drm/ttm: replace busy placement with flags v6 Christian König
@ 2024-01-12 12:51 ` Christian König
2024-01-18 14:24 ` Thomas Hellström
2024-01-12 12:51 ` [PATCH 5/5] drm/amdgpu: use GTT only as fallback for VRAM|GTT Christian König
2024-01-16 9:57 ` Rework TTMs busy handling Christian König
5 siblings, 1 reply; 19+ messages in thread
From: Christian König @ 2024-01-12 12:51 UTC (permalink / raw)
To: dri-devel, intel-gfx, thomas.hellstrom, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
Previously we would never try to move a BO into the preferred placements
when it ever landed in a busy placement since those were considered
compatible.
Rework the whole handling and finally unify the idle and busy handling.
ttm_bo_validate() is now responsible to try idle placement first and then
use the busy placement if that didn't worked.
Drawback is that we now always try the idle placement first for each
validation which might cause some additional CPU overhead on overcommit.
v2: fix kerneldoc warning and coding style
v3: take care of XE as well
Signed-off-by: Christian König <christian.koenig@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 2 +-
drivers/gpu/drm/ttm/ttm_bo.c | 131 ++++++++-------------
drivers/gpu/drm/ttm/ttm_resource.c | 16 ++-
drivers/gpu/drm/xe/xe_bo.c | 4 +-
include/drm/ttm/ttm_bo.h | 3 +-
include/drm/ttm/ttm_resource.h | 3 +-
7 files changed, 68 insertions(+), 93 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index b671b0665492..06fb3fc47eaa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -404,7 +404,7 @@ int amdgpu_bo_create_kernel_at(struct amdgpu_device *adev,
(*bo_ptr)->placements[i].lpfn = (offset + size) >> PAGE_SHIFT;
}
r = ttm_bo_mem_space(&(*bo_ptr)->tbo, &(*bo_ptr)->placement,
- &(*bo_ptr)->tbo.resource, &ctx);
+ &(*bo_ptr)->tbo.resource, &ctx, false);
if (r)
goto error;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 8722beba494e..f23cdc7c5b08 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -966,7 +966,7 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
placements.mem_type = TTM_PL_TT;
placements.flags = bo->resource->placement;
- r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx);
+ r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx, true);
if (unlikely(r))
return r;
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index a5e11a92e0b9..3783be24d832 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -414,7 +414,7 @@ static int ttm_bo_bounce_temp_buffer(struct ttm_buffer_object *bo,
hop_placement.placement = hop;
/* find space in the bounce domain */
- ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
+ ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx, true);
if (ret)
return ret;
/* move to the bounce domain */
@@ -454,7 +454,7 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo,
return ttm_bo_pipeline_gutting(bo);
}
- ret = ttm_bo_mem_space(bo, &placement, &evict_mem, ctx);
+ ret = ttm_bo_mem_space(bo, &placement, &evict_mem, ctx, true);
if (ret) {
if (ret != -ERESTARTSYS) {
pr_err("Failed to find memory space for buffer 0x%p eviction\n",
@@ -724,37 +724,6 @@ static int ttm_bo_add_move_fence(struct ttm_buffer_object *bo,
return ret;
}
-/*
- * Repeatedly evict memory from the LRU for @mem_type until we create enough
- * space, or we've evicted everything and there isn't enough space.
- */
-static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo,
- const struct ttm_place *place,
- struct ttm_resource **mem,
- struct ttm_operation_ctx *ctx)
-{
- struct ttm_device *bdev = bo->bdev;
- struct ttm_resource_manager *man;
- struct ww_acquire_ctx *ticket;
- int ret;
-
- man = ttm_manager_type(bdev, place->mem_type);
- ticket = dma_resv_locking_ctx(bo->base.resv);
- do {
- ret = ttm_resource_alloc(bo, place, mem);
- if (likely(!ret))
- break;
- if (unlikely(ret != -ENOSPC))
- return ret;
- ret = ttm_mem_evict_first(bdev, man, place, ctx,
- ticket);
- if (unlikely(ret != 0))
- return ret;
- } while (1);
-
- return ttm_bo_add_move_fence(bo, man, *mem, ctx->no_wait_gpu);
-}
-
/**
* ttm_bo_mem_space
*
@@ -763,6 +732,7 @@ static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo,
* @placement: Proposed new placement for the buffer object.
* @mem: A struct ttm_resource.
* @ctx: if and how to sleep, lock buffers and alloc memory
+ * @force_space: If we should evict buffers to force space
*
* Allocate memory space for the buffer object pointed to by @bo, using
* the placement flags in @placement, potentially evicting other idle buffer objects.
@@ -776,12 +746,14 @@ static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo,
int ttm_bo_mem_space(struct ttm_buffer_object *bo,
struct ttm_placement *placement,
struct ttm_resource **mem,
- struct ttm_operation_ctx *ctx)
+ struct ttm_operation_ctx *ctx,
+ bool force_space)
{
struct ttm_device *bdev = bo->bdev;
- bool type_found = false;
+ struct ww_acquire_ctx *ticket;
int i, ret;
+ ticket = dma_resv_locking_ctx(bo->base.resv);
ret = dma_resv_reserve_fences(bo->base.resv, 1);
if (unlikely(ret))
return ret;
@@ -790,19 +762,30 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
const struct ttm_place *place = &placement->placement[i];
struct ttm_resource_manager *man;
- if (place->flags & TTM_PL_FLAG_FALLBACK)
- continue;
-
man = ttm_manager_type(bdev, place->mem_type);
if (!man || !ttm_resource_manager_used(man))
continue;
- type_found = true;
- ret = ttm_resource_alloc(bo, place, mem);
- if (ret == -ENOSPC)
+ if (place->flags & (force_space ? TTM_PL_FLAG_DESIRED :
+ TTM_PL_FLAG_FALLBACK))
+ continue;
+
+ do {
+ ret = ttm_resource_alloc(bo, place, mem);
+ if (unlikely(ret != -ENOSPC))
+ return ret;
+ if (likely(!ret) || !force_space)
+ break;
+
+ ret = ttm_mem_evict_first(bdev, man, place, ctx,
+ ticket);
+ if (unlikely(ret == -EBUSY))
+ break;
+ if (unlikely(ret))
+ return ret;
+ } while (1);
+ if (ret)
continue;
- if (unlikely(ret))
- goto error;
ret = ttm_bo_add_move_fence(bo, man, *mem, ctx->no_wait_gpu);
if (unlikely(ret)) {
@@ -810,45 +793,19 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
if (ret == -EBUSY)
continue;
- goto error;
+ return ret;
}
return 0;
}
- for (i = 0; i < placement->num_placement; ++i) {
- const struct ttm_place *place = &placement->placement[i];
- struct ttm_resource_manager *man;
-
- if (place->flags & TTM_PL_FLAG_DESIRED)
- continue;
-
- man = ttm_manager_type(bdev, place->mem_type);
- if (!man || !ttm_resource_manager_used(man))
- continue;
-
- type_found = true;
- ret = ttm_bo_mem_force_space(bo, place, mem, ctx);
- if (likely(!ret))
- return 0;
-
- if (ret && ret != -EBUSY)
- goto error;
- }
-
- ret = -ENOSPC;
- if (!type_found) {
- pr_err(TTM_PFX "No compatible memory type found\n");
- ret = -EINVAL;
- }
-
-error:
- return ret;
+ return -ENOSPC;
}
EXPORT_SYMBOL(ttm_bo_mem_space);
static int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
struct ttm_placement *placement,
- struct ttm_operation_ctx *ctx)
+ struct ttm_operation_ctx *ctx,
+ bool force_space)
{
struct ttm_resource *mem;
struct ttm_place hop;
@@ -865,7 +822,7 @@ static int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
* stop and the driver will be called to make
* the second hop.
*/
- ret = ttm_bo_mem_space(bo, placement, &mem, ctx);
+ ret = ttm_bo_mem_space(bo, placement, &mem, ctx, force_space);
if (ret)
return ret;
bounce:
@@ -902,6 +859,7 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
struct ttm_placement *placement,
struct ttm_operation_ctx *ctx)
{
+ bool force_space;
int ret;
dma_resv_assert_held(bo->base.resv);
@@ -912,20 +870,27 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
if (!placement->num_placement)
return ttm_bo_pipeline_gutting(bo);
- /* Check whether we need to move buffer. */
- if (bo->resource && ttm_resource_compatible(bo->resource, placement))
- return 0;
+ force_space = false;
+ do {
+ /* Check whether we need to move buffer. */
+ if (bo->resource &&
+ ttm_resource_compatible(bo->resource, placement,
+ force_space))
+ return 0;
+
+ /* Moving of pinned BOs is forbidden */
+ if (bo->pin_count)
+ return -EINVAL;
- /* Moving of pinned BOs is forbidden */
- if (bo->pin_count)
- return -EINVAL;
+ ret = ttm_bo_move_buffer(bo, placement, ctx, force_space);
+ if (ret && ret != -ENOSPC)
+ return ret;
- ret = ttm_bo_move_buffer(bo, placement, ctx);
+ force_space = !force_space;
+ } while (force_space);
/* For backward compatibility with userspace */
if (ret == -ENOSPC)
return -ENOMEM;
- if (ret)
- return ret;
/*
* We might need to add a TTM.
diff --git a/drivers/gpu/drm/ttm/ttm_resource.c b/drivers/gpu/drm/ttm/ttm_resource.c
index fb14f7716cf8..65155f2013ca 100644
--- a/drivers/gpu/drm/ttm/ttm_resource.c
+++ b/drivers/gpu/drm/ttm/ttm_resource.c
@@ -295,11 +295,13 @@ bool ttm_resource_intersects(struct ttm_device *bdev,
*
* @res: the resource to check
* @placement: the placement to check against
+ * @evicting: true if the caller is doing evictions
*
* Returns true if the placement is compatible.
*/
bool ttm_resource_compatible(struct ttm_resource *res,
- struct ttm_placement *placement)
+ struct ttm_placement *placement,
+ bool evicting)
{
struct ttm_buffer_object *bo = res->bo;
struct ttm_device *bdev = bo->bdev;
@@ -315,14 +317,20 @@ bool ttm_resource_compatible(struct ttm_resource *res,
if (res->mem_type != place->mem_type)
continue;
+ if (place->flags & (evicting ? TTM_PL_FLAG_DESIRED :
+ TTM_PL_FLAG_FALLBACK))
+ continue;
+
+ if (place->flags & TTM_PL_FLAG_CONTIGUOUS &&
+ !(res->placement & TTM_PL_FLAG_CONTIGUOUS))
+ continue;
+
man = ttm_manager_type(bdev, res->mem_type);
if (man->func->compatible &&
!man->func->compatible(man, res, place, bo->base.size))
continue;
- if ((!(place->flags & TTM_PL_FLAG_CONTIGUOUS) ||
- (res->placement & TTM_PL_FLAG_CONTIGUOUS)))
- return true;
+ return true;
}
return false;
}
diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
index 1e5edc7f0b85..d4b8bd774ada 100644
--- a/drivers/gpu/drm/xe/xe_bo.c
+++ b/drivers/gpu/drm/xe/xe_bo.c
@@ -828,7 +828,7 @@ int xe_bo_evict_pinned(struct xe_bo *bo)
if (WARN_ON(!xe_bo_is_vram(bo)))
return -EINVAL;
- ret = ttm_bo_mem_space(&bo->ttm, &placement, &new_mem, &ctx);
+ ret = ttm_bo_mem_space(&bo->ttm, &placement, &new_mem, &ctx, false);
if (ret)
return ret;
@@ -893,7 +893,7 @@ int xe_bo_restore_pinned(struct xe_bo *bo)
if (WARN_ON(xe_bo_is_vram(bo) || !bo->ttm.ttm))
return -EINVAL;
- ret = ttm_bo_mem_space(&bo->ttm, &bo->placement, &new_mem, &ctx);
+ ret = ttm_bo_mem_space(&bo->ttm, &bo->placement, &new_mem, &ctx, false);
if (ret)
return ret;
diff --git a/include/drm/ttm/ttm_bo.h b/include/drm/ttm/ttm_bo.h
index 0223a41a64b2..64d738f152d1 100644
--- a/include/drm/ttm/ttm_bo.h
+++ b/include/drm/ttm/ttm_bo.h
@@ -397,7 +397,8 @@ vm_fault_t ttm_bo_vm_dummy_page(struct vm_fault *vmf, pgprot_t prot);
int ttm_bo_mem_space(struct ttm_buffer_object *bo,
struct ttm_placement *placement,
struct ttm_resource **mem,
- struct ttm_operation_ctx *ctx);
+ struct ttm_operation_ctx *ctx,
+ bool force_space);
void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo);
/*
diff --git a/include/drm/ttm/ttm_resource.h b/include/drm/ttm/ttm_resource.h
index 1afa13f0c22b..7561023db43d 100644
--- a/include/drm/ttm/ttm_resource.h
+++ b/include/drm/ttm/ttm_resource.h
@@ -366,7 +366,8 @@ bool ttm_resource_intersects(struct ttm_device *bdev,
const struct ttm_place *place,
size_t size);
bool ttm_resource_compatible(struct ttm_resource *res,
- struct ttm_placement *placement);
+ struct ttm_placement *placement,
+ bool evicting);
void ttm_resource_set_bo(struct ttm_resource *res,
struct ttm_buffer_object *bo);
--
2.34.1
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH 5/5] drm/amdgpu: use GTT only as fallback for VRAM|GTT
2024-01-12 12:51 Rework TTMs busy handling Christian König
` (3 preceding siblings ...)
2024-01-12 12:51 ` [PATCH 4/5] drm/ttm: improve idle/busy handling v3 Christian König
@ 2024-01-12 12:51 ` Christian König
2024-01-16 9:57 ` Rework TTMs busy handling Christian König
5 siblings, 0 replies; 19+ messages in thread
From: Christian König @ 2024-01-12 12:51 UTC (permalink / raw)
To: dri-devel, intel-gfx, thomas.hellstrom, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
Try to fill up VRAM as well by setting the busy flag on GTT allocations.
This fixes the issue that when VRAM was evacuated for suspend it's never
filled up again unless the application is restarted.
Signed-off-by: Christian König <christian.koenig@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index 06fb3fc47eaa..2752f2a67a44 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -173,6 +173,12 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
abo->flags & AMDGPU_GEM_CREATE_PREEMPTIBLE ?
AMDGPU_PL_PREEMPT : TTM_PL_TT;
places[c].flags = 0;
+ /*
+ * When GTT is just an alternative to VRAM make sure that we
+ * only use it as fallback and still try to fill up VRAM first.
+ */
+ if (domain & abo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM)
+ places[c].flags |= TTM_PL_FLAG_FALLBACK;
c++;
}
--
2.34.1
^ permalink raw reply related [flat|nested] 19+ messages in thread
* Re: Rework TTMs busy handling
2024-01-12 12:51 Rework TTMs busy handling Christian König
` (4 preceding siblings ...)
2024-01-12 12:51 ` [PATCH 5/5] drm/amdgpu: use GTT only as fallback for VRAM|GTT Christian König
@ 2024-01-16 9:57 ` Christian König
2024-01-17 3:13 ` Zack Rusin
5 siblings, 1 reply; 19+ messages in thread
From: Christian König @ 2024-01-16 9:57 UTC (permalink / raw)
To: dri-devel, intel-gfx, thomas.hellstrom, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
Am 12.01.24 um 13:51 schrieb Christian König:
> Hi guys,
just a gentle ping on this.
Zack any more comments for the VMWGFX parts?
Thanks,
Christian.
>
> same as the last time. Things I've changed:
>
> Implemented the requirements from Zack to correctly fill in the busy
> placements for VMWGFX.
>
> Renamed the placement flags to desired and fallback as suggested by
> Michel.
>
> Rebased on drm-tip instead of drm-misc-next and fixed XE as well.
>
> Please review and comment,
> Christian.
>
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: Rework TTMs busy handling
2024-01-16 9:57 ` Rework TTMs busy handling Christian König
@ 2024-01-17 3:13 ` Zack Rusin
0 siblings, 0 replies; 19+ messages in thread
From: Zack Rusin @ 2024-01-17 3:13 UTC (permalink / raw)
To: Christian König
Cc: thomas.hellstrom, kherbst, michel.daenzer, nouveau, intel-gfx,
dri-devel, zackr
On Tue, Jan 16, 2024 at 4:57 AM Christian König
<ckoenig.leichtzumerken@gmail.com> wrote:
>
> Am 12.01.24 um 13:51 schrieb Christian König:
> > Hi guys,
>
> just a gentle ping on this.
>
> Zack any more comments for the VMWGFX parts?
The new vmwgfx code looks great, thanks a lot for implementing it! In
fact the entire series looks good to me. For the series:
Reviewed-by: Zack Rusin <zack.rusin@broadcom.com>
z
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 3/5] drm/ttm: replace busy placement with flags v6
2024-01-12 12:51 ` [PATCH 3/5] drm/ttm: replace busy placement with flags v6 Christian König
@ 2024-01-17 8:01 ` Thomas Zimmermann
2024-01-17 10:47 ` Thomas Hellström
1 sibling, 0 replies; 19+ messages in thread
From: Thomas Zimmermann @ 2024-01-17 8:01 UTC (permalink / raw)
To: Christian König, dri-devel, intel-gfx, thomas.hellstrom,
nouveau, jani.nikula, kherbst, lyude, zackr, michel.daenzer
[-- Attachment #1.1: Type: text/plain, Size: 33883 bytes --]
Am 12.01.24 um 13:51 schrieb Christian König:
> From: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
>
> Instead of a list of separate busy placement add flags which indicate
> that a placement should only be used when there is room or if we need to
> evict.
>
> v2: add missing TTM_PL_FLAG_IDLE for i915
> v3: fix auto build test ERROR on drm-tip/drm-tip
> v4: fix some typos pointed out by checkpatch
> v5: cleanup some rebase problems with VMWGFX
> v6: implement some missing VMWGFX functionality pointed out by Zack,
> rename the flags as suggested by Michel, rebase on drm-tip and
> adjust XE as well
>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
For the VRAM helpers:
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 6 +-
> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 +---
> drivers/gpu/drm/drm_gem_vram_helper.c | 2 -
> drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 37 +++++------
> drivers/gpu/drm/loongson/lsdc_ttm.c | 2 -
> drivers/gpu/drm/nouveau/nouveau_bo.c | 59 +++++++----------
> drivers/gpu/drm/nouveau/nouveau_bo.h | 1 -
> drivers/gpu/drm/qxl/qxl_object.c | 2 -
> drivers/gpu/drm/qxl/qxl_ttm.c | 2 -
> drivers/gpu/drm/radeon/radeon_object.c | 2 -
> drivers/gpu/drm/radeon/radeon_ttm.c | 8 +--
> drivers/gpu/drm/radeon/radeon_uvd.c | 1 -
> drivers/gpu/drm/ttm/ttm_bo.c | 21 ++++---
> drivers/gpu/drm/ttm/ttm_resource.c | 73 +++++-----------------
> drivers/gpu/drm/vmwgfx/vmwgfx_bo.c | 33 +++++++---
> drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 4 --
> drivers/gpu/drm/xe/xe_bo.c | 33 +++++-----
> include/drm/ttm/ttm_placement.h | 10 +--
> include/drm/ttm/ttm_resource.h | 8 +--
> 19 files changed, 118 insertions(+), 197 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> index 425cebcc5cbf..b671b0665492 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> @@ -220,9 +220,6 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
>
> placement->num_placement = c;
> placement->placement = places;
> -
> - placement->num_busy_placement = c;
> - placement->busy_placement = places;
> }
>
> /**
> @@ -1397,8 +1394,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
> AMDGPU_GEM_DOMAIN_GTT);
>
> /* Avoid costly evictions; only set GTT as a busy placement */
> - abo->placement.num_busy_placement = 1;
> - abo->placement.busy_placement = &abo->placements[1];
> + abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>
> r = ttm_bo_validate(bo, &abo->placement, &ctx);
> if (unlikely(r == -EBUSY || r == -ERESTARTSYS))
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 75c9fd2c6c2a..8722beba494e 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -102,23 +102,19 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
> /* Don't handle scatter gather BOs */
> if (bo->type == ttm_bo_type_sg) {
> placement->num_placement = 0;
> - placement->num_busy_placement = 0;
> return;
> }
>
> /* Object isn't an AMDGPU object so ignore */
> if (!amdgpu_bo_is_amdgpu_bo(bo)) {
> placement->placement = &placements;
> - placement->busy_placement = &placements;
> placement->num_placement = 1;
> - placement->num_busy_placement = 1;
> return;
> }
>
> abo = ttm_to_amdgpu_bo(bo);
> if (abo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) {
> placement->num_placement = 0;
> - placement->num_busy_placement = 0;
> return;
> }
>
> @@ -128,13 +124,13 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
> case AMDGPU_PL_OA:
> case AMDGPU_PL_DOORBELL:
> placement->num_placement = 0;
> - placement->num_busy_placement = 0;
> return;
>
> case TTM_PL_VRAM:
> if (!adev->mman.buffer_funcs_enabled) {
> /* Move to system memory */
> amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU);
> +
> } else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
> !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) &&
> amdgpu_bo_in_cpu_visible_vram(abo)) {
> @@ -149,8 +145,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
> AMDGPU_GEM_DOMAIN_CPU);
> abo->placements[0].fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
> abo->placements[0].lpfn = 0;
> - abo->placement.busy_placement = &abo->placements[1];
> - abo->placement.num_busy_placement = 1;
> + abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
> } else {
> /* Move to GTT memory */
> amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT |
> @@ -966,8 +961,6 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> /* allocate GART space */
> placement.num_placement = 1;
> placement.placement = &placements;
> - placement.num_busy_placement = 1;
> - placement.busy_placement = &placements;
> placements.fpfn = 0;
> placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
> placements.mem_type = TTM_PL_TT;
> diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
> index b67eafa55715..75f2eaf0d5b6 100644
> --- a/drivers/gpu/drm/drm_gem_vram_helper.c
> +++ b/drivers/gpu/drm/drm_gem_vram_helper.c
> @@ -147,7 +147,6 @@ static void drm_gem_vram_placement(struct drm_gem_vram_object *gbo,
> invariant_flags = TTM_PL_FLAG_TOPDOWN;
>
> gbo->placement.placement = gbo->placements;
> - gbo->placement.busy_placement = gbo->placements;
>
> if (pl_flag & DRM_GEM_VRAM_PL_FLAG_VRAM) {
> gbo->placements[c].mem_type = TTM_PL_VRAM;
> @@ -160,7 +159,6 @@ static void drm_gem_vram_placement(struct drm_gem_vram_object *gbo,
> }
>
> gbo->placement.num_placement = c;
> - gbo->placement.num_busy_placement = c;
>
> for (i = 0; i < c; ++i) {
> gbo->placements[i].fpfn = 0;
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
> index 6b69ef0cdbb4..80c6cafc8887 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
> @@ -65,8 +65,6 @@ static const struct ttm_place sys_placement_flags = {
> static struct ttm_placement i915_sys_placement = {
> .num_placement = 1,
> .placement = &sys_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &sys_placement_flags,
> };
>
> /**
> @@ -157,32 +155,28 @@ i915_ttm_place_from_region(const struct intel_memory_region *mr,
>
> static void
> i915_ttm_placement_from_obj(const struct drm_i915_gem_object *obj,
> - struct ttm_place *requested,
> - struct ttm_place *busy,
> + struct ttm_place *places,
> struct ttm_placement *placement)
> {
> unsigned int num_allowed = obj->mm.n_placements;
> unsigned int flags = obj->flags;
> unsigned int i;
>
> - placement->num_placement = 1;
> + places[0].flags |= TTM_PL_FLAG_DESIRED;
> i915_ttm_place_from_region(num_allowed ? obj->mm.placements[0] :
> - obj->mm.region, requested, obj->bo_offset,
> + obj->mm.region, &places[0], obj->bo_offset,
> obj->base.size, flags);
>
> /* Cache this on object? */
> - placement->num_busy_placement = num_allowed;
> - for (i = 0; i < placement->num_busy_placement; ++i)
> - i915_ttm_place_from_region(obj->mm.placements[i], busy + i,
> - obj->bo_offset, obj->base.size, flags);
> -
> - if (num_allowed == 0) {
> - *busy = *requested;
> - placement->num_busy_placement = 1;
> + for (i = 0; i < num_allowed; ++i) {
> + i915_ttm_place_from_region(obj->mm.placements[i],
> + &places[i + 1], obj->bo_offset,
> + obj->base.size, flags);
> + places[i + 1].flags |= TTM_PL_FLAG_FALLBACK;
> }
>
> - placement->placement = requested;
> - placement->busy_placement = busy;
> + placement->num_placement = num_allowed + 1;
> + placement->placement = places;
> }
>
> static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev,
> @@ -789,7 +783,8 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj,
> int ret;
>
> /* First try only the requested placement. No eviction. */
> - real_num_busy = fetch_and_zero(&placement->num_busy_placement);
> + real_num_busy = placement->num_placement;
> + placement->num_placement = 1;
> ret = ttm_bo_validate(bo, placement, &ctx);
> if (ret) {
> ret = i915_ttm_err_to_gem(ret);
> @@ -805,7 +800,7 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj,
> * If the initial attempt fails, allow all accepted placements,
> * evicting if necessary.
> */
> - placement->num_busy_placement = real_num_busy;
> + placement->num_placement = real_num_busy;
> ret = ttm_bo_validate(bo, placement, &ctx);
> if (ret)
> return i915_ttm_err_to_gem(ret);
> @@ -839,7 +834,7 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj,
>
> static int i915_ttm_get_pages(struct drm_i915_gem_object *obj)
> {
> - struct ttm_place requested, busy[I915_TTM_MAX_PLACEMENTS];
> + struct ttm_place places[I915_TTM_MAX_PLACEMENTS + 1];
> struct ttm_placement placement;
>
> /* restricted by sg_alloc_table */
> @@ -849,7 +844,7 @@ static int i915_ttm_get_pages(struct drm_i915_gem_object *obj)
> GEM_BUG_ON(obj->mm.n_placements > I915_TTM_MAX_PLACEMENTS);
>
> /* Move to the requested placement. */
> - i915_ttm_placement_from_obj(obj, &requested, busy, &placement);
> + i915_ttm_placement_from_obj(obj, places, &placement);
>
> return __i915_ttm_get_pages(obj, &placement);
> }
> @@ -879,9 +874,7 @@ static int __i915_ttm_migrate(struct drm_i915_gem_object *obj,
> i915_ttm_place_from_region(mr, &requested, obj->bo_offset,
> obj->base.size, flags);
> placement.num_placement = 1;
> - placement.num_busy_placement = 1;
> placement.placement = &requested;
> - placement.busy_placement = &requested;
>
> ret = __i915_ttm_get_pages(obj, &placement);
> if (ret)
> diff --git a/drivers/gpu/drm/loongson/lsdc_ttm.c b/drivers/gpu/drm/loongson/lsdc_ttm.c
> index bf79dc55afa4..465f622ac05d 100644
> --- a/drivers/gpu/drm/loongson/lsdc_ttm.c
> +++ b/drivers/gpu/drm/loongson/lsdc_ttm.c
> @@ -54,7 +54,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo *lbo, u32 domain)
> pflags |= TTM_PL_FLAG_TOPDOWN;
>
> lbo->placement.placement = lbo->placements;
> - lbo->placement.busy_placement = lbo->placements;
>
> if (domain & LSDC_GEM_DOMAIN_VRAM) {
> lbo->placements[c].mem_type = TTM_PL_VRAM;
> @@ -77,7 +76,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo *lbo, u32 domain)
> }
>
> lbo->placement.num_placement = c;
> - lbo->placement.num_busy_placement = c;
>
> for (i = 0; i < c; ++i) {
> lbo->placements[i].fpfn = 0;
> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
> index 00cc7d1abaa3..56dcd25db1ce 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_bo.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
> @@ -404,27 +404,6 @@ nouveau_bo_new(struct nouveau_cli *cli, u64 size, int align,
> return 0;
> }
>
> -static void
> -set_placement_list(struct ttm_place *pl, unsigned *n, uint32_t domain)
> -{
> - *n = 0;
> -
> - if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
> - pl[*n].mem_type = TTM_PL_VRAM;
> - pl[*n].flags = 0;
> - (*n)++;
> - }
> - if (domain & NOUVEAU_GEM_DOMAIN_GART) {
> - pl[*n].mem_type = TTM_PL_TT;
> - pl[*n].flags = 0;
> - (*n)++;
> - }
> - if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
> - pl[*n].mem_type = TTM_PL_SYSTEM;
> - pl[(*n)++].flags = 0;
> - }
> -}
> -
> static void
> set_placement_range(struct nouveau_bo *nvbo, uint32_t domain)
> {
> @@ -452,10 +431,6 @@ set_placement_range(struct nouveau_bo *nvbo, uint32_t domain)
> nvbo->placements[i].fpfn = fpfn;
> nvbo->placements[i].lpfn = lpfn;
> }
> - for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
> - nvbo->busy_placements[i].fpfn = fpfn;
> - nvbo->busy_placements[i].lpfn = lpfn;
> - }
> }
> }
>
> @@ -463,15 +438,32 @@ void
> nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t domain,
> uint32_t busy)
> {
> - struct ttm_placement *pl = &nvbo->placement;
> + unsigned int *n = &nvbo->placement.num_placement;
> + struct ttm_place *pl = nvbo->placements;
>
> - pl->placement = nvbo->placements;
> - set_placement_list(nvbo->placements, &pl->num_placement, domain);
> + domain |= busy;
>
> - pl->busy_placement = nvbo->busy_placements;
> - set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
> - domain | busy);
> + *n = 0;
> + if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
> + pl[*n].mem_type = TTM_PL_VRAM;
> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_VRAM ?
> + TTM_PL_FLAG_FALLBACK : 0;
> + (*n)++;
> + }
> + if (domain & NOUVEAU_GEM_DOMAIN_GART) {
> + pl[*n].mem_type = TTM_PL_TT;
> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_GART ?
> + TTM_PL_FLAG_FALLBACK : 0;
> + (*n)++;
> + }
> + if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
> + pl[*n].mem_type = TTM_PL_SYSTEM;
> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_CPU ?
> + TTM_PL_FLAG_FALLBACK : 0;
> + (*n)++;
> + }
>
> + nvbo->placement.placement = nvbo->placements;
> set_placement_range(nvbo, domain);
> }
>
> @@ -1314,11 +1306,6 @@ vm_fault_t nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
> nvbo->placements[i].lpfn = mappable;
> }
>
> - for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
> - nvbo->busy_placements[i].fpfn = 0;
> - nvbo->busy_placements[i].lpfn = mappable;
> - }
> -
> nouveau_bo_placement_set(nvbo, NOUVEAU_GEM_DOMAIN_VRAM, 0);
> }
>
> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.h b/drivers/gpu/drm/nouveau/nouveau_bo.h
> index 70c551921a9e..e9dfab6a8156 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_bo.h
> +++ b/drivers/gpu/drm/nouveau/nouveau_bo.h
> @@ -15,7 +15,6 @@ struct nouveau_bo {
> struct ttm_placement placement;
> u32 valid_domains;
> struct ttm_place placements[3];
> - struct ttm_place busy_placements[3];
> bool force_coherent;
> struct ttm_bo_kmap_obj kmap;
> struct list_head head;
> diff --git a/drivers/gpu/drm/qxl/qxl_object.c b/drivers/gpu/drm/qxl/qxl_object.c
> index 06a58dad5f5c..1e46b0a6e478 100644
> --- a/drivers/gpu/drm/qxl/qxl_object.c
> +++ b/drivers/gpu/drm/qxl/qxl_object.c
> @@ -66,7 +66,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo *qbo, u32 domain)
> pflag |= TTM_PL_FLAG_TOPDOWN;
>
> qbo->placement.placement = qbo->placements;
> - qbo->placement.busy_placement = qbo->placements;
> if (domain == QXL_GEM_DOMAIN_VRAM) {
> qbo->placements[c].mem_type = TTM_PL_VRAM;
> qbo->placements[c++].flags = pflag;
> @@ -86,7 +85,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo *qbo, u32 domain)
> qbo->placements[c++].flags = 0;
> }
> qbo->placement.num_placement = c;
> - qbo->placement.num_busy_placement = c;
> for (i = 0; i < c; ++i) {
> qbo->placements[i].fpfn = 0;
> qbo->placements[i].lpfn = 0;
> diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
> index 1a82629bce3f..765a144cea14 100644
> --- a/drivers/gpu/drm/qxl/qxl_ttm.c
> +++ b/drivers/gpu/drm/qxl/qxl_ttm.c
> @@ -60,9 +60,7 @@ static void qxl_evict_flags(struct ttm_buffer_object *bo,
>
> if (!qxl_ttm_bo_is_qxl_bo(bo)) {
> placement->placement = &placements;
> - placement->busy_placement = &placements;
> placement->num_placement = 1;
> - placement->num_busy_placement = 1;
> return;
> }
> qbo = to_qxl_bo(bo);
> diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
> index 10c0fbd9d2b4..a955f8a2f7fe 100644
> --- a/drivers/gpu/drm/radeon/radeon_object.c
> +++ b/drivers/gpu/drm/radeon/radeon_object.c
> @@ -78,7 +78,6 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
> u32 c = 0, i;
>
> rbo->placement.placement = rbo->placements;
> - rbo->placement.busy_placement = rbo->placements;
> if (domain & RADEON_GEM_DOMAIN_VRAM) {
> /* Try placing BOs which don't need CPU access outside of the
> * CPU accessible part of VRAM
> @@ -114,7 +113,6 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
> }
>
> rbo->placement.num_placement = c;
> - rbo->placement.num_busy_placement = c;
>
> for (i = 0; i < c; ++i) {
> if ((rbo->flags & RADEON_GEM_CPU_ACCESS) &&
> diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
> index de4e6d78f1e1..2078b0000e22 100644
> --- a/drivers/gpu/drm/radeon/radeon_ttm.c
> +++ b/drivers/gpu/drm/radeon/radeon_ttm.c
> @@ -92,9 +92,7 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo,
>
> if (!radeon_ttm_bo_is_radeon_bo(bo)) {
> placement->placement = &placements;
> - placement->busy_placement = &placements;
> placement->num_placement = 1;
> - placement->num_busy_placement = 1;
> return;
> }
> rbo = container_of(bo, struct radeon_bo, tbo);
> @@ -114,15 +112,11 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo,
> */
> radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM |
> RADEON_GEM_DOMAIN_GTT);
> - rbo->placement.num_busy_placement = 0;
> for (i = 0; i < rbo->placement.num_placement; i++) {
> if (rbo->placements[i].mem_type == TTM_PL_VRAM) {
> if (rbo->placements[i].fpfn < fpfn)
> rbo->placements[i].fpfn = fpfn;
> - } else {
> - rbo->placement.busy_placement =
> - &rbo->placements[i];
> - rbo->placement.num_busy_placement = 1;
> + rbo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
> }
> }
> } else
> diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c b/drivers/gpu/drm/radeon/radeon_uvd.c
> index a2cda184b2b2..058a1c8451b2 100644
> --- a/drivers/gpu/drm/radeon/radeon_uvd.c
> +++ b/drivers/gpu/drm/radeon/radeon_uvd.c
> @@ -324,7 +324,6 @@ void radeon_uvd_force_into_uvd_segment(struct radeon_bo *rbo,
> rbo->placements[1].fpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
> rbo->placements[1].lpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
> rbo->placement.num_placement++;
> - rbo->placement.num_busy_placement++;
> }
>
> void radeon_uvd_free_handles(struct radeon_device *rdev, struct drm_file *filp)
> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
> index 8c1eaa74fa21..a5e11a92e0b9 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
> @@ -410,8 +410,8 @@ static int ttm_bo_bounce_temp_buffer(struct ttm_buffer_object *bo,
> struct ttm_resource *hop_mem;
> int ret;
>
> - hop_placement.num_placement = hop_placement.num_busy_placement = 1;
> - hop_placement.placement = hop_placement.busy_placement = hop;
> + hop_placement.num_placement = 1;
> + hop_placement.placement = hop;
>
> /* find space in the bounce domain */
> ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
> @@ -440,10 +440,9 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo,
> dma_resv_assert_held(bo->base.resv);
>
> placement.num_placement = 0;
> - placement.num_busy_placement = 0;
> bdev->funcs->evict_flags(bo, &placement);
>
> - if (!placement.num_placement && !placement.num_busy_placement) {
> + if (!placement.num_placement) {
> ret = ttm_bo_wait_ctx(bo, ctx);
> if (ret)
> return ret;
> @@ -791,6 +790,9 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
> const struct ttm_place *place = &placement->placement[i];
> struct ttm_resource_manager *man;
>
> + if (place->flags & TTM_PL_FLAG_FALLBACK)
> + continue;
> +
> man = ttm_manager_type(bdev, place->mem_type);
> if (!man || !ttm_resource_manager_used(man))
> continue;
> @@ -813,10 +815,13 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
> return 0;
> }
>
> - for (i = 0; i < placement->num_busy_placement; ++i) {
> - const struct ttm_place *place = &placement->busy_placement[i];
> + for (i = 0; i < placement->num_placement; ++i) {
> + const struct ttm_place *place = &placement->placement[i];
> struct ttm_resource_manager *man;
>
> + if (place->flags & TTM_PL_FLAG_DESIRED)
> + continue;
> +
> man = ttm_manager_type(bdev, place->mem_type);
> if (!man || !ttm_resource_manager_used(man))
> continue;
> @@ -904,11 +909,11 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
> /*
> * Remove the backing store if no placement is given.
> */
> - if (!placement->num_placement && !placement->num_busy_placement)
> + if (!placement->num_placement)
> return ttm_bo_pipeline_gutting(bo);
>
> /* Check whether we need to move buffer. */
> - if (bo->resource && ttm_resource_compat(bo->resource, placement))
> + if (bo->resource && ttm_resource_compatible(bo->resource, placement))
> return 0;
>
> /* Moving of pinned BOs is forbidden */
> diff --git a/drivers/gpu/drm/ttm/ttm_resource.c b/drivers/gpu/drm/ttm/ttm_resource.c
> index 02b96d23fdb9..fb14f7716cf8 100644
> --- a/drivers/gpu/drm/ttm/ttm_resource.c
> +++ b/drivers/gpu/drm/ttm/ttm_resource.c
> @@ -291,37 +291,15 @@ bool ttm_resource_intersects(struct ttm_device *bdev,
> }
>
> /**
> - * ttm_resource_compatible - test for compatibility
> + * ttm_resource_compatible - check if resource is compatible with placement
> *
> - * @bdev: TTM device structure
> - * @res: The resource to test
> - * @place: The placement to test
> - * @size: How many bytes the new allocation needs.
> - *
> - * Test if @res compatible with @place and @size.
> + * @res: the resource to check
> + * @placement: the placement to check against
> *
> - * Returns true if the res placement compatible with @place and @size.
> + * Returns true if the placement is compatible.
> */
> -bool ttm_resource_compatible(struct ttm_device *bdev,
> - struct ttm_resource *res,
> - const struct ttm_place *place,
> - size_t size)
> -{
> - struct ttm_resource_manager *man;
> -
> - if (!res || !place)
> - return false;
> -
> - man = ttm_manager_type(bdev, res->mem_type);
> - if (!man->func->compatible)
> - return true;
> -
> - return man->func->compatible(man, res, place, size);
> -}
> -
> -static bool ttm_resource_places_compat(struct ttm_resource *res,
> - const struct ttm_place *places,
> - unsigned num_placement)
> +bool ttm_resource_compatible(struct ttm_resource *res,
> + struct ttm_placement *placement)
> {
> struct ttm_buffer_object *bo = res->bo;
> struct ttm_device *bdev = bo->bdev;
> @@ -330,44 +308,25 @@ static bool ttm_resource_places_compat(struct ttm_resource *res,
> if (res->placement & TTM_PL_FLAG_TEMPORARY)
> return false;
>
> - for (i = 0; i < num_placement; i++) {
> - const struct ttm_place *heap = &places[i];
> + for (i = 0; i < placement->num_placement; i++) {
> + const struct ttm_place *place = &placement->placement[i];
> + struct ttm_resource_manager *man;
>
> - if (!ttm_resource_compatible(bdev, res, heap, bo->base.size))
> + if (res->mem_type != place->mem_type)
> + continue;
> +
> + man = ttm_manager_type(bdev, res->mem_type);
> + if (man->func->compatible &&
> + !man->func->compatible(man, res, place, bo->base.size))
> continue;
>
> - if ((res->mem_type == heap->mem_type) &&
> - (!(heap->flags & TTM_PL_FLAG_CONTIGUOUS) ||
> + if ((!(place->flags & TTM_PL_FLAG_CONTIGUOUS) ||
> (res->placement & TTM_PL_FLAG_CONTIGUOUS)))
> return true;
> }
> return false;
> }
>
> -/**
> - * ttm_resource_compat - check if resource is compatible with placement
> - *
> - * @res: the resource to check
> - * @placement: the placement to check against
> - *
> - * Returns true if the placement is compatible.
> - */
> -bool ttm_resource_compat(struct ttm_resource *res,
> - struct ttm_placement *placement)
> -{
> - if (ttm_resource_places_compat(res, placement->placement,
> - placement->num_placement))
> - return true;
> -
> - if ((placement->busy_placement != placement->placement ||
> - placement->num_busy_placement > placement->num_placement) &&
> - ttm_resource_places_compat(res, placement->busy_placement,
> - placement->num_busy_placement))
> - return true;
> -
> - return false;
> -}
> -
> void ttm_resource_set_bo(struct ttm_resource *res,
> struct ttm_buffer_object *bo)
> {
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
> index 2bfac3aad7b7..bfd41ce3c8f4 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
> @@ -742,9 +742,21 @@ void vmw_bo_move_notify(struct ttm_buffer_object *bo,
> vmw_resource_unbind_list(vbo);
> }
>
> +static u32 placement_flags(u32 domain, u32 desired, u32 fallback)
> +{
> + if (desired & fallback & domain)
> + return 0;
> +
> + if (desired & domain)
> + return TTM_PL_FLAG_DESIRED;
> +
> + return TTM_PL_FLAG_FALLBACK;
> +}
> +
> static u32
> -set_placement_list(struct ttm_place *pl, u32 domain)
> +set_placement_list(struct ttm_place *pl, u32 desired, u32 fallback)
> {
> + u32 domain = desired | fallback;
> u32 n = 0;
>
> /*
> @@ -752,35 +764,40 @@ set_placement_list(struct ttm_place *pl, u32 domain)
> */
> if (domain & VMW_BO_DOMAIN_MOB) {
> pl[n].mem_type = VMW_PL_MOB;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_MOB, desired,
> + fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> }
> if (domain & VMW_BO_DOMAIN_GMR) {
> pl[n].mem_type = VMW_PL_GMR;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_GMR, desired,
> + fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> }
> if (domain & VMW_BO_DOMAIN_VRAM) {
> pl[n].mem_type = TTM_PL_VRAM;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_VRAM, desired,
> + fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> }
> if (domain & VMW_BO_DOMAIN_WAITABLE_SYS) {
> pl[n].mem_type = VMW_PL_SYSTEM;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_WAITABLE_SYS,
> + desired, fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> }
> if (domain & VMW_BO_DOMAIN_SYS) {
> pl[n].mem_type = TTM_PL_SYSTEM;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_SYS, desired,
> + fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> @@ -806,7 +823,7 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32 domain, u32 busy_domain)
> u32 i;
>
> pl->placement = bo->places;
> - pl->num_placement = set_placement_list(bo->places, domain);
> + pl->num_placement = set_placement_list(bo->places, domain, busy_domain);
>
> if (drm_debug_enabled(DRM_UT_DRIVER) && bo->tbo.resource) {
> for (i = 0; i < pl->num_placement; ++i) {
> @@ -821,8 +838,6 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32 domain, u32 busy_domain)
> __func__, bo->tbo.resource->mem_type, domain);
> }
>
> - pl->busy_placement = bo->busy_places;
> - pl->num_busy_placement = set_placement_list(bo->busy_places, busy_domain);
> }
>
> void vmw_bo_placement_set_default_accelerated(struct vmw_bo *bo)
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
> index a84fffcef8e1..4d23d0a70bcb 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
> @@ -46,15 +46,11 @@ static const struct ttm_place sys_placement_flags = {
> struct ttm_placement vmw_vram_placement = {
> .num_placement = 1,
> .placement = &vram_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &vram_placement_flags
> };
>
> struct ttm_placement vmw_sys_placement = {
> .num_placement = 1,
> .placement = &sys_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &sys_placement_flags
> };
>
> const size_t vmw_tt_size = sizeof(struct vmw_ttm_tt);
> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> index 8e4a3b1f6b93..1e5edc7f0b85 100644
> --- a/drivers/gpu/drm/xe/xe_bo.c
> +++ b/drivers/gpu/drm/xe/xe_bo.c
> @@ -38,22 +38,26 @@ static const struct ttm_place sys_placement_flags = {
> static struct ttm_placement sys_placement = {
> .num_placement = 1,
> .placement = &sys_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &sys_placement_flags,
> };
>
> -static const struct ttm_place tt_placement_flags = {
> - .fpfn = 0,
> - .lpfn = 0,
> - .mem_type = XE_PL_TT,
> - .flags = 0,
> +static const struct ttm_place tt_placement_flags[] = {
> + {
> + .fpfn = 0,
> + .lpfn = 0,
> + .mem_type = XE_PL_TT,
> + .flags = TTM_PL_FLAG_DESIRED,
> + },
> + {
> + .fpfn = 0,
> + .lpfn = 0,
> + .mem_type = XE_PL_SYSTEM,
> + .flags = TTM_PL_FLAG_FALLBACK,
> + }
> };
>
> static struct ttm_placement tt_placement = {
> - .num_placement = 1,
> - .placement = &tt_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &sys_placement_flags,
> + .num_placement = 2,
> + .placement = tt_placement_flags,
> };
>
> bool mem_type_is_vram(u32 mem_type)
> @@ -230,8 +234,6 @@ static int __xe_bo_placement_for_flags(struct xe_device *xe, struct xe_bo *bo,
> bo->placement = (struct ttm_placement) {
> .num_placement = c,
> .placement = bo->placements,
> - .num_busy_placement = c,
> - .busy_placement = bo->placements,
> };
>
> return 0;
> @@ -251,7 +253,6 @@ static void xe_evict_flags(struct ttm_buffer_object *tbo,
> /* Don't handle scatter gather BOs */
> if (tbo->type == ttm_bo_type_sg) {
> placement->num_placement = 0;
> - placement->num_busy_placement = 0;
> return;
> }
>
> @@ -1353,8 +1354,6 @@ static int __xe_bo_fixed_placement(struct xe_device *xe,
> bo->placement = (struct ttm_placement) {
> .num_placement = 1,
> .placement = place,
> - .num_busy_placement = 1,
> - .busy_placement = place,
> };
>
> return 0;
> @@ -2112,9 +2111,7 @@ int xe_bo_migrate(struct xe_bo *bo, u32 mem_type)
>
> xe_place_from_ttm_type(mem_type, &requested);
> placement.num_placement = 1;
> - placement.num_busy_placement = 1;
> placement.placement = &requested;
> - placement.busy_placement = &requested;
>
> /*
> * Stolen needs to be handled like below VRAM handling if we ever need
> diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h
> index 8074d0f6cae5..b510a4812609 100644
> --- a/include/drm/ttm/ttm_placement.h
> +++ b/include/drm/ttm/ttm_placement.h
> @@ -64,6 +64,12 @@
> /* For multihop handling */
> #define TTM_PL_FLAG_TEMPORARY (1 << 2)
>
> +/* Placement is never used during eviction */
> +#define TTM_PL_FLAG_DESIRED (1 << 3)
> +
> +/* Placement is only used during eviction */
> +#define TTM_PL_FLAG_FALLBACK (1 << 4)
> +
> /**
> * struct ttm_place
> *
> @@ -86,16 +92,12 @@ struct ttm_place {
> *
> * @num_placement: number of preferred placements
> * @placement: preferred placements
> - * @num_busy_placement: number of preferred placements when need to evict buffer
> - * @busy_placement: preferred placements when need to evict buffer
> *
> * Structure indicating the placement you request for an object.
> */
> struct ttm_placement {
> unsigned num_placement;
> const struct ttm_place *placement;
> - unsigned num_busy_placement;
> - const struct ttm_place *busy_placement;
> };
>
> #endif
> diff --git a/include/drm/ttm/ttm_resource.h b/include/drm/ttm/ttm_resource.h
> index 78a226eba953..1afa13f0c22b 100644
> --- a/include/drm/ttm/ttm_resource.h
> +++ b/include/drm/ttm/ttm_resource.h
> @@ -365,12 +365,8 @@ bool ttm_resource_intersects(struct ttm_device *bdev,
> struct ttm_resource *res,
> const struct ttm_place *place,
> size_t size);
> -bool ttm_resource_compatible(struct ttm_device *bdev,
> - struct ttm_resource *res,
> - const struct ttm_place *place,
> - size_t size);
> -bool ttm_resource_compat(struct ttm_resource *res,
> - struct ttm_placement *placement);
> +bool ttm_resource_compatible(struct ttm_resource *res,
> + struct ttm_placement *placement);
> void ttm_resource_set_bo(struct ttm_resource *res,
> struct ttm_buffer_object *bo);
>
--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstrasse 146, 90461 Nuernberg, Germany
GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman
HRB 36809 (AG Nuernberg)
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 840 bytes --]
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 2/5] drm/ttm: return ENOSPC from ttm_bo_mem_space
2024-01-12 12:51 ` [PATCH 2/5] drm/ttm: return ENOSPC from ttm_bo_mem_space Christian König
@ 2024-01-17 10:18 ` Thomas Hellström
0 siblings, 0 replies; 19+ messages in thread
From: Thomas Hellström @ 2024-01-17 10:18 UTC (permalink / raw)
To: Christian König, dri-devel, intel-gfx, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
Hi,
On 1/12/24 13:51, Christian König wrote:
> Only convert it to ENOMEM in ttm_bo_validate.
>
> This allows ttm_bo_validate to distinct between an out of memory
NIT: s/distinct/distinguish/
> situation and just out of space in a placement domain.
In fact it would be nice if this could be propagated back to drivers as
well at some point, but then perhaps guarded with a flag in the
operation context.
In any case
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> ---
> drivers/gpu/drm/ttm/ttm_bo.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
> index edf10618fe2b..8c1eaa74fa21 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
> @@ -830,7 +830,7 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
> goto error;
> }
>
> - ret = -ENOMEM;
> + ret = -ENOSPC;
> if (!type_found) {
> pr_err(TTM_PFX "No compatible memory type found\n");
> ret = -EINVAL;
> @@ -916,6 +916,9 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
> return -EINVAL;
>
> ret = ttm_bo_move_buffer(bo, placement, ctx);
> + /* For backward compatibility with userspace */
> + if (ret == -ENOSPC)
> + return -ENOMEM;
> if (ret)
> return ret;
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 3/5] drm/ttm: replace busy placement with flags v6
2024-01-12 12:51 ` [PATCH 3/5] drm/ttm: replace busy placement with flags v6 Christian König
2024-01-17 8:01 ` Thomas Zimmermann
@ 2024-01-17 10:47 ` Thomas Hellström
2024-01-17 12:27 ` Thomas Hellström
1 sibling, 1 reply; 19+ messages in thread
From: Thomas Hellström @ 2024-01-17 10:47 UTC (permalink / raw)
To: Christian König, dri-devel, intel-gfx, thomas.hellstrom,
nouveau, jani.nikula, kherbst, lyude, zackr, michel.daenzer
Hi, Christian
Xe changes look good. Will send the series to xe ci to check for
regressions.
/Thomas
On 1/12/24 13:51, Christian König wrote:
> From: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
>
> Instead of a list of separate busy placement add flags which indicate
> that a placement should only be used when there is room or if we need to
> evict.
>
> v2: add missing TTM_PL_FLAG_IDLE for i915
> v3: fix auto build test ERROR on drm-tip/drm-tip
> v4: fix some typos pointed out by checkpatch
> v5: cleanup some rebase problems with VMWGFX
> v6: implement some missing VMWGFX functionality pointed out by Zack,
> rename the flags as suggested by Michel, rebase on drm-tip and
> adjust XE as well
>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 6 +-
> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 +---
> drivers/gpu/drm/drm_gem_vram_helper.c | 2 -
> drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 37 +++++------
> drivers/gpu/drm/loongson/lsdc_ttm.c | 2 -
> drivers/gpu/drm/nouveau/nouveau_bo.c | 59 +++++++----------
> drivers/gpu/drm/nouveau/nouveau_bo.h | 1 -
> drivers/gpu/drm/qxl/qxl_object.c | 2 -
> drivers/gpu/drm/qxl/qxl_ttm.c | 2 -
> drivers/gpu/drm/radeon/radeon_object.c | 2 -
> drivers/gpu/drm/radeon/radeon_ttm.c | 8 +--
> drivers/gpu/drm/radeon/radeon_uvd.c | 1 -
> drivers/gpu/drm/ttm/ttm_bo.c | 21 ++++---
> drivers/gpu/drm/ttm/ttm_resource.c | 73 +++++-----------------
> drivers/gpu/drm/vmwgfx/vmwgfx_bo.c | 33 +++++++---
> drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 4 --
> drivers/gpu/drm/xe/xe_bo.c | 33 +++++-----
> include/drm/ttm/ttm_placement.h | 10 +--
> include/drm/ttm/ttm_resource.h | 8 +--
> 19 files changed, 118 insertions(+), 197 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> index 425cebcc5cbf..b671b0665492 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> @@ -220,9 +220,6 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
>
> placement->num_placement = c;
> placement->placement = places;
> -
> - placement->num_busy_placement = c;
> - placement->busy_placement = places;
> }
>
> /**
> @@ -1397,8 +1394,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
> AMDGPU_GEM_DOMAIN_GTT);
>
> /* Avoid costly evictions; only set GTT as a busy placement */
> - abo->placement.num_busy_placement = 1;
> - abo->placement.busy_placement = &abo->placements[1];
> + abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>
> r = ttm_bo_validate(bo, &abo->placement, &ctx);
> if (unlikely(r == -EBUSY || r == -ERESTARTSYS))
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 75c9fd2c6c2a..8722beba494e 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -102,23 +102,19 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
> /* Don't handle scatter gather BOs */
> if (bo->type == ttm_bo_type_sg) {
> placement->num_placement = 0;
> - placement->num_busy_placement = 0;
> return;
> }
>
> /* Object isn't an AMDGPU object so ignore */
> if (!amdgpu_bo_is_amdgpu_bo(bo)) {
> placement->placement = &placements;
> - placement->busy_placement = &placements;
> placement->num_placement = 1;
> - placement->num_busy_placement = 1;
> return;
> }
>
> abo = ttm_to_amdgpu_bo(bo);
> if (abo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) {
> placement->num_placement = 0;
> - placement->num_busy_placement = 0;
> return;
> }
>
> @@ -128,13 +124,13 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
> case AMDGPU_PL_OA:
> case AMDGPU_PL_DOORBELL:
> placement->num_placement = 0;
> - placement->num_busy_placement = 0;
> return;
>
> case TTM_PL_VRAM:
> if (!adev->mman.buffer_funcs_enabled) {
> /* Move to system memory */
> amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU);
> +
> } else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
> !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) &&
> amdgpu_bo_in_cpu_visible_vram(abo)) {
> @@ -149,8 +145,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
> AMDGPU_GEM_DOMAIN_CPU);
> abo->placements[0].fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
> abo->placements[0].lpfn = 0;
> - abo->placement.busy_placement = &abo->placements[1];
> - abo->placement.num_busy_placement = 1;
> + abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
> } else {
> /* Move to GTT memory */
> amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT |
> @@ -966,8 +961,6 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> /* allocate GART space */
> placement.num_placement = 1;
> placement.placement = &placements;
> - placement.num_busy_placement = 1;
> - placement.busy_placement = &placements;
> placements.fpfn = 0;
> placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
> placements.mem_type = TTM_PL_TT;
> diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
> index b67eafa55715..75f2eaf0d5b6 100644
> --- a/drivers/gpu/drm/drm_gem_vram_helper.c
> +++ b/drivers/gpu/drm/drm_gem_vram_helper.c
> @@ -147,7 +147,6 @@ static void drm_gem_vram_placement(struct drm_gem_vram_object *gbo,
> invariant_flags = TTM_PL_FLAG_TOPDOWN;
>
> gbo->placement.placement = gbo->placements;
> - gbo->placement.busy_placement = gbo->placements;
>
> if (pl_flag & DRM_GEM_VRAM_PL_FLAG_VRAM) {
> gbo->placements[c].mem_type = TTM_PL_VRAM;
> @@ -160,7 +159,6 @@ static void drm_gem_vram_placement(struct drm_gem_vram_object *gbo,
> }
>
> gbo->placement.num_placement = c;
> - gbo->placement.num_busy_placement = c;
>
> for (i = 0; i < c; ++i) {
> gbo->placements[i].fpfn = 0;
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
> index 6b69ef0cdbb4..80c6cafc8887 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
> @@ -65,8 +65,6 @@ static const struct ttm_place sys_placement_flags = {
> static struct ttm_placement i915_sys_placement = {
> .num_placement = 1,
> .placement = &sys_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &sys_placement_flags,
> };
>
> /**
> @@ -157,32 +155,28 @@ i915_ttm_place_from_region(const struct intel_memory_region *mr,
>
> static void
> i915_ttm_placement_from_obj(const struct drm_i915_gem_object *obj,
> - struct ttm_place *requested,
> - struct ttm_place *busy,
> + struct ttm_place *places,
> struct ttm_placement *placement)
> {
> unsigned int num_allowed = obj->mm.n_placements;
> unsigned int flags = obj->flags;
> unsigned int i;
>
> - placement->num_placement = 1;
> + places[0].flags |= TTM_PL_FLAG_DESIRED;
> i915_ttm_place_from_region(num_allowed ? obj->mm.placements[0] :
> - obj->mm.region, requested, obj->bo_offset,
> + obj->mm.region, &places[0], obj->bo_offset,
> obj->base.size, flags);
>
> /* Cache this on object? */
> - placement->num_busy_placement = num_allowed;
> - for (i = 0; i < placement->num_busy_placement; ++i)
> - i915_ttm_place_from_region(obj->mm.placements[i], busy + i,
> - obj->bo_offset, obj->base.size, flags);
> -
> - if (num_allowed == 0) {
> - *busy = *requested;
> - placement->num_busy_placement = 1;
> + for (i = 0; i < num_allowed; ++i) {
> + i915_ttm_place_from_region(obj->mm.placements[i],
> + &places[i + 1], obj->bo_offset,
> + obj->base.size, flags);
> + places[i + 1].flags |= TTM_PL_FLAG_FALLBACK;
> }
>
> - placement->placement = requested;
> - placement->busy_placement = busy;
> + placement->num_placement = num_allowed + 1;
> + placement->placement = places;
> }
>
> static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev,
> @@ -789,7 +783,8 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj,
> int ret;
>
> /* First try only the requested placement. No eviction. */
> - real_num_busy = fetch_and_zero(&placement->num_busy_placement);
> + real_num_busy = placement->num_placement;
> + placement->num_placement = 1;
> ret = ttm_bo_validate(bo, placement, &ctx);
> if (ret) {
> ret = i915_ttm_err_to_gem(ret);
> @@ -805,7 +800,7 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj,
> * If the initial attempt fails, allow all accepted placements,
> * evicting if necessary.
> */
> - placement->num_busy_placement = real_num_busy;
> + placement->num_placement = real_num_busy;
> ret = ttm_bo_validate(bo, placement, &ctx);
> if (ret)
> return i915_ttm_err_to_gem(ret);
> @@ -839,7 +834,7 @@ static int __i915_ttm_get_pages(struct drm_i915_gem_object *obj,
>
> static int i915_ttm_get_pages(struct drm_i915_gem_object *obj)
> {
> - struct ttm_place requested, busy[I915_TTM_MAX_PLACEMENTS];
> + struct ttm_place places[I915_TTM_MAX_PLACEMENTS + 1];
> struct ttm_placement placement;
>
> /* restricted by sg_alloc_table */
> @@ -849,7 +844,7 @@ static int i915_ttm_get_pages(struct drm_i915_gem_object *obj)
> GEM_BUG_ON(obj->mm.n_placements > I915_TTM_MAX_PLACEMENTS);
>
> /* Move to the requested placement. */
> - i915_ttm_placement_from_obj(obj, &requested, busy, &placement);
> + i915_ttm_placement_from_obj(obj, places, &placement);
>
> return __i915_ttm_get_pages(obj, &placement);
> }
> @@ -879,9 +874,7 @@ static int __i915_ttm_migrate(struct drm_i915_gem_object *obj,
> i915_ttm_place_from_region(mr, &requested, obj->bo_offset,
> obj->base.size, flags);
> placement.num_placement = 1;
> - placement.num_busy_placement = 1;
> placement.placement = &requested;
> - placement.busy_placement = &requested;
>
> ret = __i915_ttm_get_pages(obj, &placement);
> if (ret)
> diff --git a/drivers/gpu/drm/loongson/lsdc_ttm.c b/drivers/gpu/drm/loongson/lsdc_ttm.c
> index bf79dc55afa4..465f622ac05d 100644
> --- a/drivers/gpu/drm/loongson/lsdc_ttm.c
> +++ b/drivers/gpu/drm/loongson/lsdc_ttm.c
> @@ -54,7 +54,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo *lbo, u32 domain)
> pflags |= TTM_PL_FLAG_TOPDOWN;
>
> lbo->placement.placement = lbo->placements;
> - lbo->placement.busy_placement = lbo->placements;
>
> if (domain & LSDC_GEM_DOMAIN_VRAM) {
> lbo->placements[c].mem_type = TTM_PL_VRAM;
> @@ -77,7 +76,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo *lbo, u32 domain)
> }
>
> lbo->placement.num_placement = c;
> - lbo->placement.num_busy_placement = c;
>
> for (i = 0; i < c; ++i) {
> lbo->placements[i].fpfn = 0;
> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
> index 00cc7d1abaa3..56dcd25db1ce 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_bo.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
> @@ -404,27 +404,6 @@ nouveau_bo_new(struct nouveau_cli *cli, u64 size, int align,
> return 0;
> }
>
> -static void
> -set_placement_list(struct ttm_place *pl, unsigned *n, uint32_t domain)
> -{
> - *n = 0;
> -
> - if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
> - pl[*n].mem_type = TTM_PL_VRAM;
> - pl[*n].flags = 0;
> - (*n)++;
> - }
> - if (domain & NOUVEAU_GEM_DOMAIN_GART) {
> - pl[*n].mem_type = TTM_PL_TT;
> - pl[*n].flags = 0;
> - (*n)++;
> - }
> - if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
> - pl[*n].mem_type = TTM_PL_SYSTEM;
> - pl[(*n)++].flags = 0;
> - }
> -}
> -
> static void
> set_placement_range(struct nouveau_bo *nvbo, uint32_t domain)
> {
> @@ -452,10 +431,6 @@ set_placement_range(struct nouveau_bo *nvbo, uint32_t domain)
> nvbo->placements[i].fpfn = fpfn;
> nvbo->placements[i].lpfn = lpfn;
> }
> - for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
> - nvbo->busy_placements[i].fpfn = fpfn;
> - nvbo->busy_placements[i].lpfn = lpfn;
> - }
> }
> }
>
> @@ -463,15 +438,32 @@ void
> nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t domain,
> uint32_t busy)
> {
> - struct ttm_placement *pl = &nvbo->placement;
> + unsigned int *n = &nvbo->placement.num_placement;
> + struct ttm_place *pl = nvbo->placements;
>
> - pl->placement = nvbo->placements;
> - set_placement_list(nvbo->placements, &pl->num_placement, domain);
> + domain |= busy;
>
> - pl->busy_placement = nvbo->busy_placements;
> - set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
> - domain | busy);
> + *n = 0;
> + if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
> + pl[*n].mem_type = TTM_PL_VRAM;
> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_VRAM ?
> + TTM_PL_FLAG_FALLBACK : 0;
> + (*n)++;
> + }
> + if (domain & NOUVEAU_GEM_DOMAIN_GART) {
> + pl[*n].mem_type = TTM_PL_TT;
> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_GART ?
> + TTM_PL_FLAG_FALLBACK : 0;
> + (*n)++;
> + }
> + if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
> + pl[*n].mem_type = TTM_PL_SYSTEM;
> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_CPU ?
> + TTM_PL_FLAG_FALLBACK : 0;
> + (*n)++;
> + }
>
> + nvbo->placement.placement = nvbo->placements;
> set_placement_range(nvbo, domain);
> }
>
> @@ -1314,11 +1306,6 @@ vm_fault_t nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
> nvbo->placements[i].lpfn = mappable;
> }
>
> - for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
> - nvbo->busy_placements[i].fpfn = 0;
> - nvbo->busy_placements[i].lpfn = mappable;
> - }
> -
> nouveau_bo_placement_set(nvbo, NOUVEAU_GEM_DOMAIN_VRAM, 0);
> }
>
> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.h b/drivers/gpu/drm/nouveau/nouveau_bo.h
> index 70c551921a9e..e9dfab6a8156 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_bo.h
> +++ b/drivers/gpu/drm/nouveau/nouveau_bo.h
> @@ -15,7 +15,6 @@ struct nouveau_bo {
> struct ttm_placement placement;
> u32 valid_domains;
> struct ttm_place placements[3];
> - struct ttm_place busy_placements[3];
> bool force_coherent;
> struct ttm_bo_kmap_obj kmap;
> struct list_head head;
> diff --git a/drivers/gpu/drm/qxl/qxl_object.c b/drivers/gpu/drm/qxl/qxl_object.c
> index 06a58dad5f5c..1e46b0a6e478 100644
> --- a/drivers/gpu/drm/qxl/qxl_object.c
> +++ b/drivers/gpu/drm/qxl/qxl_object.c
> @@ -66,7 +66,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo *qbo, u32 domain)
> pflag |= TTM_PL_FLAG_TOPDOWN;
>
> qbo->placement.placement = qbo->placements;
> - qbo->placement.busy_placement = qbo->placements;
> if (domain == QXL_GEM_DOMAIN_VRAM) {
> qbo->placements[c].mem_type = TTM_PL_VRAM;
> qbo->placements[c++].flags = pflag;
> @@ -86,7 +85,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo *qbo, u32 domain)
> qbo->placements[c++].flags = 0;
> }
> qbo->placement.num_placement = c;
> - qbo->placement.num_busy_placement = c;
> for (i = 0; i < c; ++i) {
> qbo->placements[i].fpfn = 0;
> qbo->placements[i].lpfn = 0;
> diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
> index 1a82629bce3f..765a144cea14 100644
> --- a/drivers/gpu/drm/qxl/qxl_ttm.c
> +++ b/drivers/gpu/drm/qxl/qxl_ttm.c
> @@ -60,9 +60,7 @@ static void qxl_evict_flags(struct ttm_buffer_object *bo,
>
> if (!qxl_ttm_bo_is_qxl_bo(bo)) {
> placement->placement = &placements;
> - placement->busy_placement = &placements;
> placement->num_placement = 1;
> - placement->num_busy_placement = 1;
> return;
> }
> qbo = to_qxl_bo(bo);
> diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
> index 10c0fbd9d2b4..a955f8a2f7fe 100644
> --- a/drivers/gpu/drm/radeon/radeon_object.c
> +++ b/drivers/gpu/drm/radeon/radeon_object.c
> @@ -78,7 +78,6 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
> u32 c = 0, i;
>
> rbo->placement.placement = rbo->placements;
> - rbo->placement.busy_placement = rbo->placements;
> if (domain & RADEON_GEM_DOMAIN_VRAM) {
> /* Try placing BOs which don't need CPU access outside of the
> * CPU accessible part of VRAM
> @@ -114,7 +113,6 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
> }
>
> rbo->placement.num_placement = c;
> - rbo->placement.num_busy_placement = c;
>
> for (i = 0; i < c; ++i) {
> if ((rbo->flags & RADEON_GEM_CPU_ACCESS) &&
> diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
> index de4e6d78f1e1..2078b0000e22 100644
> --- a/drivers/gpu/drm/radeon/radeon_ttm.c
> +++ b/drivers/gpu/drm/radeon/radeon_ttm.c
> @@ -92,9 +92,7 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo,
>
> if (!radeon_ttm_bo_is_radeon_bo(bo)) {
> placement->placement = &placements;
> - placement->busy_placement = &placements;
> placement->num_placement = 1;
> - placement->num_busy_placement = 1;
> return;
> }
> rbo = container_of(bo, struct radeon_bo, tbo);
> @@ -114,15 +112,11 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo,
> */
> radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_VRAM |
> RADEON_GEM_DOMAIN_GTT);
> - rbo->placement.num_busy_placement = 0;
> for (i = 0; i < rbo->placement.num_placement; i++) {
> if (rbo->placements[i].mem_type == TTM_PL_VRAM) {
> if (rbo->placements[i].fpfn < fpfn)
> rbo->placements[i].fpfn = fpfn;
> - } else {
> - rbo->placement.busy_placement =
> - &rbo->placements[i];
> - rbo->placement.num_busy_placement = 1;
> + rbo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
> }
> }
> } else
> diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c b/drivers/gpu/drm/radeon/radeon_uvd.c
> index a2cda184b2b2..058a1c8451b2 100644
> --- a/drivers/gpu/drm/radeon/radeon_uvd.c
> +++ b/drivers/gpu/drm/radeon/radeon_uvd.c
> @@ -324,7 +324,6 @@ void radeon_uvd_force_into_uvd_segment(struct radeon_bo *rbo,
> rbo->placements[1].fpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
> rbo->placements[1].lpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
> rbo->placement.num_placement++;
> - rbo->placement.num_busy_placement++;
> }
>
> void radeon_uvd_free_handles(struct radeon_device *rdev, struct drm_file *filp)
> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
> index 8c1eaa74fa21..a5e11a92e0b9 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
> @@ -410,8 +410,8 @@ static int ttm_bo_bounce_temp_buffer(struct ttm_buffer_object *bo,
> struct ttm_resource *hop_mem;
> int ret;
>
> - hop_placement.num_placement = hop_placement.num_busy_placement = 1;
> - hop_placement.placement = hop_placement.busy_placement = hop;
> + hop_placement.num_placement = 1;
> + hop_placement.placement = hop;
>
> /* find space in the bounce domain */
> ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
> @@ -440,10 +440,9 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo,
> dma_resv_assert_held(bo->base.resv);
>
> placement.num_placement = 0;
> - placement.num_busy_placement = 0;
> bdev->funcs->evict_flags(bo, &placement);
>
> - if (!placement.num_placement && !placement.num_busy_placement) {
> + if (!placement.num_placement) {
> ret = ttm_bo_wait_ctx(bo, ctx);
> if (ret)
> return ret;
> @@ -791,6 +790,9 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
> const struct ttm_place *place = &placement->placement[i];
> struct ttm_resource_manager *man;
>
> + if (place->flags & TTM_PL_FLAG_FALLBACK)
> + continue;
> +
> man = ttm_manager_type(bdev, place->mem_type);
> if (!man || !ttm_resource_manager_used(man))
> continue;
> @@ -813,10 +815,13 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
> return 0;
> }
>
> - for (i = 0; i < placement->num_busy_placement; ++i) {
> - const struct ttm_place *place = &placement->busy_placement[i];
> + for (i = 0; i < placement->num_placement; ++i) {
> + const struct ttm_place *place = &placement->placement[i];
> struct ttm_resource_manager *man;
>
> + if (place->flags & TTM_PL_FLAG_DESIRED)
> + continue;
> +
> man = ttm_manager_type(bdev, place->mem_type);
> if (!man || !ttm_resource_manager_used(man))
> continue;
> @@ -904,11 +909,11 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
> /*
> * Remove the backing store if no placement is given.
> */
> - if (!placement->num_placement && !placement->num_busy_placement)
> + if (!placement->num_placement)
> return ttm_bo_pipeline_gutting(bo);
>
> /* Check whether we need to move buffer. */
> - if (bo->resource && ttm_resource_compat(bo->resource, placement))
> + if (bo->resource && ttm_resource_compatible(bo->resource, placement))
> return 0;
>
> /* Moving of pinned BOs is forbidden */
> diff --git a/drivers/gpu/drm/ttm/ttm_resource.c b/drivers/gpu/drm/ttm/ttm_resource.c
> index 02b96d23fdb9..fb14f7716cf8 100644
> --- a/drivers/gpu/drm/ttm/ttm_resource.c
> +++ b/drivers/gpu/drm/ttm/ttm_resource.c
> @@ -291,37 +291,15 @@ bool ttm_resource_intersects(struct ttm_device *bdev,
> }
>
> /**
> - * ttm_resource_compatible - test for compatibility
> + * ttm_resource_compatible - check if resource is compatible with placement
> *
> - * @bdev: TTM device structure
> - * @res: The resource to test
> - * @place: The placement to test
> - * @size: How many bytes the new allocation needs.
> - *
> - * Test if @res compatible with @place and @size.
> + * @res: the resource to check
> + * @placement: the placement to check against
> *
> - * Returns true if the res placement compatible with @place and @size.
> + * Returns true if the placement is compatible.
> */
> -bool ttm_resource_compatible(struct ttm_device *bdev,
> - struct ttm_resource *res,
> - const struct ttm_place *place,
> - size_t size)
> -{
> - struct ttm_resource_manager *man;
> -
> - if (!res || !place)
> - return false;
> -
> - man = ttm_manager_type(bdev, res->mem_type);
> - if (!man->func->compatible)
> - return true;
> -
> - return man->func->compatible(man, res, place, size);
> -}
> -
> -static bool ttm_resource_places_compat(struct ttm_resource *res,
> - const struct ttm_place *places,
> - unsigned num_placement)
> +bool ttm_resource_compatible(struct ttm_resource *res,
> + struct ttm_placement *placement)
> {
> struct ttm_buffer_object *bo = res->bo;
> struct ttm_device *bdev = bo->bdev;
> @@ -330,44 +308,25 @@ static bool ttm_resource_places_compat(struct ttm_resource *res,
> if (res->placement & TTM_PL_FLAG_TEMPORARY)
> return false;
>
> - for (i = 0; i < num_placement; i++) {
> - const struct ttm_place *heap = &places[i];
> + for (i = 0; i < placement->num_placement; i++) {
> + const struct ttm_place *place = &placement->placement[i];
> + struct ttm_resource_manager *man;
>
> - if (!ttm_resource_compatible(bdev, res, heap, bo->base.size))
> + if (res->mem_type != place->mem_type)
> + continue;
> +
> + man = ttm_manager_type(bdev, res->mem_type);
> + if (man->func->compatible &&
> + !man->func->compatible(man, res, place, bo->base.size))
> continue;
>
> - if ((res->mem_type == heap->mem_type) &&
> - (!(heap->flags & TTM_PL_FLAG_CONTIGUOUS) ||
> + if ((!(place->flags & TTM_PL_FLAG_CONTIGUOUS) ||
> (res->placement & TTM_PL_FLAG_CONTIGUOUS)))
> return true;
> }
> return false;
> }
>
> -/**
> - * ttm_resource_compat - check if resource is compatible with placement
> - *
> - * @res: the resource to check
> - * @placement: the placement to check against
> - *
> - * Returns true if the placement is compatible.
> - */
> -bool ttm_resource_compat(struct ttm_resource *res,
> - struct ttm_placement *placement)
> -{
> - if (ttm_resource_places_compat(res, placement->placement,
> - placement->num_placement))
> - return true;
> -
> - if ((placement->busy_placement != placement->placement ||
> - placement->num_busy_placement > placement->num_placement) &&
> - ttm_resource_places_compat(res, placement->busy_placement,
> - placement->num_busy_placement))
> - return true;
> -
> - return false;
> -}
> -
> void ttm_resource_set_bo(struct ttm_resource *res,
> struct ttm_buffer_object *bo)
> {
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
> index 2bfac3aad7b7..bfd41ce3c8f4 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
> @@ -742,9 +742,21 @@ void vmw_bo_move_notify(struct ttm_buffer_object *bo,
> vmw_resource_unbind_list(vbo);
> }
>
> +static u32 placement_flags(u32 domain, u32 desired, u32 fallback)
> +{
> + if (desired & fallback & domain)
> + return 0;
> +
> + if (desired & domain)
> + return TTM_PL_FLAG_DESIRED;
> +
> + return TTM_PL_FLAG_FALLBACK;
> +}
> +
> static u32
> -set_placement_list(struct ttm_place *pl, u32 domain)
> +set_placement_list(struct ttm_place *pl, u32 desired, u32 fallback)
> {
> + u32 domain = desired | fallback;
> u32 n = 0;
>
> /*
> @@ -752,35 +764,40 @@ set_placement_list(struct ttm_place *pl, u32 domain)
> */
> if (domain & VMW_BO_DOMAIN_MOB) {
> pl[n].mem_type = VMW_PL_MOB;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_MOB, desired,
> + fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> }
> if (domain & VMW_BO_DOMAIN_GMR) {
> pl[n].mem_type = VMW_PL_GMR;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_GMR, desired,
> + fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> }
> if (domain & VMW_BO_DOMAIN_VRAM) {
> pl[n].mem_type = TTM_PL_VRAM;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_VRAM, desired,
> + fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> }
> if (domain & VMW_BO_DOMAIN_WAITABLE_SYS) {
> pl[n].mem_type = VMW_PL_SYSTEM;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_WAITABLE_SYS,
> + desired, fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> }
> if (domain & VMW_BO_DOMAIN_SYS) {
> pl[n].mem_type = TTM_PL_SYSTEM;
> - pl[n].flags = 0;
> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_SYS, desired,
> + fallback);
> pl[n].fpfn = 0;
> pl[n].lpfn = 0;
> n++;
> @@ -806,7 +823,7 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32 domain, u32 busy_domain)
> u32 i;
>
> pl->placement = bo->places;
> - pl->num_placement = set_placement_list(bo->places, domain);
> + pl->num_placement = set_placement_list(bo->places, domain, busy_domain);
>
> if (drm_debug_enabled(DRM_UT_DRIVER) && bo->tbo.resource) {
> for (i = 0; i < pl->num_placement; ++i) {
> @@ -821,8 +838,6 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32 domain, u32 busy_domain)
> __func__, bo->tbo.resource->mem_type, domain);
> }
>
> - pl->busy_placement = bo->busy_places;
> - pl->num_busy_placement = set_placement_list(bo->busy_places, busy_domain);
> }
>
> void vmw_bo_placement_set_default_accelerated(struct vmw_bo *bo)
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
> index a84fffcef8e1..4d23d0a70bcb 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
> @@ -46,15 +46,11 @@ static const struct ttm_place sys_placement_flags = {
> struct ttm_placement vmw_vram_placement = {
> .num_placement = 1,
> .placement = &vram_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &vram_placement_flags
> };
>
> struct ttm_placement vmw_sys_placement = {
> .num_placement = 1,
> .placement = &sys_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &sys_placement_flags
> };
>
> const size_t vmw_tt_size = sizeof(struct vmw_ttm_tt);
> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> index 8e4a3b1f6b93..1e5edc7f0b85 100644
> --- a/drivers/gpu/drm/xe/xe_bo.c
> +++ b/drivers/gpu/drm/xe/xe_bo.c
> @@ -38,22 +38,26 @@ static const struct ttm_place sys_placement_flags = {
> static struct ttm_placement sys_placement = {
> .num_placement = 1,
> .placement = &sys_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &sys_placement_flags,
> };
>
> -static const struct ttm_place tt_placement_flags = {
> - .fpfn = 0,
> - .lpfn = 0,
> - .mem_type = XE_PL_TT,
> - .flags = 0,
> +static const struct ttm_place tt_placement_flags[] = {
> + {
> + .fpfn = 0,
> + .lpfn = 0,
> + .mem_type = XE_PL_TT,
> + .flags = TTM_PL_FLAG_DESIRED,
> + },
> + {
> + .fpfn = 0,
> + .lpfn = 0,
> + .mem_type = XE_PL_SYSTEM,
> + .flags = TTM_PL_FLAG_FALLBACK,
> + }
> };
>
> static struct ttm_placement tt_placement = {
> - .num_placement = 1,
> - .placement = &tt_placement_flags,
> - .num_busy_placement = 1,
> - .busy_placement = &sys_placement_flags,
> + .num_placement = 2,
> + .placement = tt_placement_flags,
> };
>
> bool mem_type_is_vram(u32 mem_type)
> @@ -230,8 +234,6 @@ static int __xe_bo_placement_for_flags(struct xe_device *xe, struct xe_bo *bo,
> bo->placement = (struct ttm_placement) {
> .num_placement = c,
> .placement = bo->placements,
> - .num_busy_placement = c,
> - .busy_placement = bo->placements,
> };
>
> return 0;
> @@ -251,7 +253,6 @@ static void xe_evict_flags(struct ttm_buffer_object *tbo,
> /* Don't handle scatter gather BOs */
> if (tbo->type == ttm_bo_type_sg) {
> placement->num_placement = 0;
> - placement->num_busy_placement = 0;
> return;
> }
>
> @@ -1353,8 +1354,6 @@ static int __xe_bo_fixed_placement(struct xe_device *xe,
> bo->placement = (struct ttm_placement) {
> .num_placement = 1,
> .placement = place,
> - .num_busy_placement = 1,
> - .busy_placement = place,
> };
>
> return 0;
> @@ -2112,9 +2111,7 @@ int xe_bo_migrate(struct xe_bo *bo, u32 mem_type)
>
> xe_place_from_ttm_type(mem_type, &requested);
> placement.num_placement = 1;
> - placement.num_busy_placement = 1;
> placement.placement = &requested;
> - placement.busy_placement = &requested;
>
> /*
> * Stolen needs to be handled like below VRAM handling if we ever need
> diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h
> index 8074d0f6cae5..b510a4812609 100644
> --- a/include/drm/ttm/ttm_placement.h
> +++ b/include/drm/ttm/ttm_placement.h
> @@ -64,6 +64,12 @@
> /* For multihop handling */
> #define TTM_PL_FLAG_TEMPORARY (1 << 2)
>
> +/* Placement is never used during eviction */
> +#define TTM_PL_FLAG_DESIRED (1 << 3)
> +
> +/* Placement is only used during eviction */
> +#define TTM_PL_FLAG_FALLBACK (1 << 4)
> +
> /**
> * struct ttm_place
> *
> @@ -86,16 +92,12 @@ struct ttm_place {
> *
> * @num_placement: number of preferred placements
> * @placement: preferred placements
> - * @num_busy_placement: number of preferred placements when need to evict buffer
> - * @busy_placement: preferred placements when need to evict buffer
> *
> * Structure indicating the placement you request for an object.
> */
> struct ttm_placement {
> unsigned num_placement;
> const struct ttm_place *placement;
> - unsigned num_busy_placement;
> - const struct ttm_place *busy_placement;
> };
>
> #endif
> diff --git a/include/drm/ttm/ttm_resource.h b/include/drm/ttm/ttm_resource.h
> index 78a226eba953..1afa13f0c22b 100644
> --- a/include/drm/ttm/ttm_resource.h
> +++ b/include/drm/ttm/ttm_resource.h
> @@ -365,12 +365,8 @@ bool ttm_resource_intersects(struct ttm_device *bdev,
> struct ttm_resource *res,
> const struct ttm_place *place,
> size_t size);
> -bool ttm_resource_compatible(struct ttm_device *bdev,
> - struct ttm_resource *res,
> - const struct ttm_place *place,
> - size_t size);
> -bool ttm_resource_compat(struct ttm_resource *res,
> - struct ttm_placement *placement);
> +bool ttm_resource_compatible(struct ttm_resource *res,
> + struct ttm_placement *placement);
> void ttm_resource_set_bo(struct ttm_resource *res,
> struct ttm_buffer_object *bo);
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 3/5] drm/ttm: replace busy placement with flags v6
2024-01-17 10:47 ` Thomas Hellström
@ 2024-01-17 12:27 ` Thomas Hellström
2024-01-18 16:38 ` Thomas Hellström
0 siblings, 1 reply; 19+ messages in thread
From: Thomas Hellström @ 2024-01-17 12:27 UTC (permalink / raw)
To: Christian König, dri-devel, intel-gfx, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
On 1/17/24 11:47, Thomas Hellström wrote:
> Hi, Christian
>
> Xe changes look good. Will send the series to xe ci to check for
> regressions.
Hmm, there are some checkpatch warnings about author / SOB email mismatch,
But worserthere are some regressions in the dma-buf ktest (it tests
evicting of a dynamic dma-buf),
https://patchwork.freedesktop.org/series/128873/
I'll take a look later today or tomorrow.
/Thomas
>
> /Thomas
>
>
> On 1/12/24 13:51, Christian König wrote:
>> From: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
>>
>> Instead of a list of separate busy placement add flags which indicate
>> that a placement should only be used when there is room or if we need to
>> evict.
>>
>> v2: add missing TTM_PL_FLAG_IDLE for i915
>> v3: fix auto build test ERROR on drm-tip/drm-tip
>> v4: fix some typos pointed out by checkpatch
>> v5: cleanup some rebase problems with VMWGFX
>> v6: implement some missing VMWGFX functionality pointed out by Zack,
>> rename the flags as suggested by Michel, rebase on drm-tip and
>> adjust XE as well
>>
>> Signed-off-by: Christian König <christian.koenig@amd.com>
>> Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
>> ---
>> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 6 +-
>> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 +---
>> drivers/gpu/drm/drm_gem_vram_helper.c | 2 -
>> drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 37 +++++------
>> drivers/gpu/drm/loongson/lsdc_ttm.c | 2 -
>> drivers/gpu/drm/nouveau/nouveau_bo.c | 59 +++++++----------
>> drivers/gpu/drm/nouveau/nouveau_bo.h | 1 -
>> drivers/gpu/drm/qxl/qxl_object.c | 2 -
>> drivers/gpu/drm/qxl/qxl_ttm.c | 2 -
>> drivers/gpu/drm/radeon/radeon_object.c | 2 -
>> drivers/gpu/drm/radeon/radeon_ttm.c | 8 +--
>> drivers/gpu/drm/radeon/radeon_uvd.c | 1 -
>> drivers/gpu/drm/ttm/ttm_bo.c | 21 ++++---
>> drivers/gpu/drm/ttm/ttm_resource.c | 73 +++++-----------------
>> drivers/gpu/drm/vmwgfx/vmwgfx_bo.c | 33 +++++++---
>> drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 4 --
>> drivers/gpu/drm/xe/xe_bo.c | 33 +++++-----
>> include/drm/ttm/ttm_placement.h | 10 +--
>> include/drm/ttm/ttm_resource.h | 8 +--
>> 19 files changed, 118 insertions(+), 197 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> index 425cebcc5cbf..b671b0665492 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> @@ -220,9 +220,6 @@ void amdgpu_bo_placement_from_domain(struct
>> amdgpu_bo *abo, u32 domain)
>> placement->num_placement = c;
>> placement->placement = places;
>> -
>> - placement->num_busy_placement = c;
>> - placement->busy_placement = places;
>> }
>> /**
>> @@ -1397,8 +1394,7 @@ vm_fault_t
>> amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
>> AMDGPU_GEM_DOMAIN_GTT);
>> /* Avoid costly evictions; only set GTT as a busy placement */
>> - abo->placement.num_busy_placement = 1;
>> - abo->placement.busy_placement = &abo->placements[1];
>> + abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>> r = ttm_bo_validate(bo, &abo->placement, &ctx);
>> if (unlikely(r == -EBUSY || r == -ERESTARTSYS))
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> index 75c9fd2c6c2a..8722beba494e 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> @@ -102,23 +102,19 @@ static void amdgpu_evict_flags(struct
>> ttm_buffer_object *bo,
>> /* Don't handle scatter gather BOs */
>> if (bo->type == ttm_bo_type_sg) {
>> placement->num_placement = 0;
>> - placement->num_busy_placement = 0;
>> return;
>> }
>> /* Object isn't an AMDGPU object so ignore */
>> if (!amdgpu_bo_is_amdgpu_bo(bo)) {
>> placement->placement = &placements;
>> - placement->busy_placement = &placements;
>> placement->num_placement = 1;
>> - placement->num_busy_placement = 1;
>> return;
>> }
>> abo = ttm_to_amdgpu_bo(bo);
>> if (abo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) {
>> placement->num_placement = 0;
>> - placement->num_busy_placement = 0;
>> return;
>> }
>> @@ -128,13 +124,13 @@ static void amdgpu_evict_flags(struct
>> ttm_buffer_object *bo,
>> case AMDGPU_PL_OA:
>> case AMDGPU_PL_DOORBELL:
>> placement->num_placement = 0;
>> - placement->num_busy_placement = 0;
>> return;
>> case TTM_PL_VRAM:
>> if (!adev->mman.buffer_funcs_enabled) {
>> /* Move to system memory */
>> amdgpu_bo_placement_from_domain(abo,
>> AMDGPU_GEM_DOMAIN_CPU);
>> +
>> } else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
>> !(abo->flags &
>> AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) &&
>> amdgpu_bo_in_cpu_visible_vram(abo)) {
>> @@ -149,8 +145,7 @@ static void amdgpu_evict_flags(struct
>> ttm_buffer_object *bo,
>> AMDGPU_GEM_DOMAIN_CPU);
>> abo->placements[0].fpfn = adev->gmc.visible_vram_size
>> >> PAGE_SHIFT;
>> abo->placements[0].lpfn = 0;
>> - abo->placement.busy_placement = &abo->placements[1];
>> - abo->placement.num_busy_placement = 1;
>> + abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>> } else {
>> /* Move to GTT memory */
>> amdgpu_bo_placement_from_domain(abo,
>> AMDGPU_GEM_DOMAIN_GTT |
>> @@ -966,8 +961,6 @@ int amdgpu_ttm_alloc_gart(struct
>> ttm_buffer_object *bo)
>> /* allocate GART space */
>> placement.num_placement = 1;
>> placement.placement = &placements;
>> - placement.num_busy_placement = 1;
>> - placement.busy_placement = &placements;
>> placements.fpfn = 0;
>> placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
>> placements.mem_type = TTM_PL_TT;
>> diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c
>> b/drivers/gpu/drm/drm_gem_vram_helper.c
>> index b67eafa55715..75f2eaf0d5b6 100644
>> --- a/drivers/gpu/drm/drm_gem_vram_helper.c
>> +++ b/drivers/gpu/drm/drm_gem_vram_helper.c
>> @@ -147,7 +147,6 @@ static void drm_gem_vram_placement(struct
>> drm_gem_vram_object *gbo,
>> invariant_flags = TTM_PL_FLAG_TOPDOWN;
>> gbo->placement.placement = gbo->placements;
>> - gbo->placement.busy_placement = gbo->placements;
>> if (pl_flag & DRM_GEM_VRAM_PL_FLAG_VRAM) {
>> gbo->placements[c].mem_type = TTM_PL_VRAM;
>> @@ -160,7 +159,6 @@ static void drm_gem_vram_placement(struct
>> drm_gem_vram_object *gbo,
>> }
>> gbo->placement.num_placement = c;
>> - gbo->placement.num_busy_placement = c;
>> for (i = 0; i < c; ++i) {
>> gbo->placements[i].fpfn = 0;
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>> b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>> index 6b69ef0cdbb4..80c6cafc8887 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>> @@ -65,8 +65,6 @@ static const struct ttm_place sys_placement_flags = {
>> static struct ttm_placement i915_sys_placement = {
>> .num_placement = 1,
>> .placement = &sys_placement_flags,
>> - .num_busy_placement = 1,
>> - .busy_placement = &sys_placement_flags,
>> };
>> /**
>> @@ -157,32 +155,28 @@ i915_ttm_place_from_region(const struct
>> intel_memory_region *mr,
>> static void
>> i915_ttm_placement_from_obj(const struct drm_i915_gem_object *obj,
>> - struct ttm_place *requested,
>> - struct ttm_place *busy,
>> + struct ttm_place *places,
>> struct ttm_placement *placement)
>> {
>> unsigned int num_allowed = obj->mm.n_placements;
>> unsigned int flags = obj->flags;
>> unsigned int i;
>> - placement->num_placement = 1;
>> + places[0].flags |= TTM_PL_FLAG_DESIRED;
>> i915_ttm_place_from_region(num_allowed ? obj->mm.placements[0] :
>> - obj->mm.region, requested, obj->bo_offset,
>> + obj->mm.region, &places[0], obj->bo_offset,
>> obj->base.size, flags);
>> /* Cache this on object? */
>> - placement->num_busy_placement = num_allowed;
>> - for (i = 0; i < placement->num_busy_placement; ++i)
>> - i915_ttm_place_from_region(obj->mm.placements[i], busy + i,
>> - obj->bo_offset, obj->base.size, flags);
>> -
>> - if (num_allowed == 0) {
>> - *busy = *requested;
>> - placement->num_busy_placement = 1;
>> + for (i = 0; i < num_allowed; ++i) {
>> + i915_ttm_place_from_region(obj->mm.placements[i],
>> + &places[i + 1], obj->bo_offset,
>> + obj->base.size, flags);
>> + places[i + 1].flags |= TTM_PL_FLAG_FALLBACK;
>> }
>> - placement->placement = requested;
>> - placement->busy_placement = busy;
>> + placement->num_placement = num_allowed + 1;
>> + placement->placement = places;
>> }
>> static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev,
>> @@ -789,7 +783,8 @@ static int __i915_ttm_get_pages(struct
>> drm_i915_gem_object *obj,
>> int ret;
>> /* First try only the requested placement. No eviction. */
>> - real_num_busy = fetch_and_zero(&placement->num_busy_placement);
>> + real_num_busy = placement->num_placement;
>> + placement->num_placement = 1;
>> ret = ttm_bo_validate(bo, placement, &ctx);
>> if (ret) {
>> ret = i915_ttm_err_to_gem(ret);
>> @@ -805,7 +800,7 @@ static int __i915_ttm_get_pages(struct
>> drm_i915_gem_object *obj,
>> * If the initial attempt fails, allow all accepted
>> placements,
>> * evicting if necessary.
>> */
>> - placement->num_busy_placement = real_num_busy;
>> + placement->num_placement = real_num_busy;
>> ret = ttm_bo_validate(bo, placement, &ctx);
>> if (ret)
>> return i915_ttm_err_to_gem(ret);
>> @@ -839,7 +834,7 @@ static int __i915_ttm_get_pages(struct
>> drm_i915_gem_object *obj,
>> static int i915_ttm_get_pages(struct drm_i915_gem_object *obj)
>> {
>> - struct ttm_place requested, busy[I915_TTM_MAX_PLACEMENTS];
>> + struct ttm_place places[I915_TTM_MAX_PLACEMENTS + 1];
>> struct ttm_placement placement;
>> /* restricted by sg_alloc_table */
>> @@ -849,7 +844,7 @@ static int i915_ttm_get_pages(struct
>> drm_i915_gem_object *obj)
>> GEM_BUG_ON(obj->mm.n_placements > I915_TTM_MAX_PLACEMENTS);
>> /* Move to the requested placement. */
>> - i915_ttm_placement_from_obj(obj, &requested, busy, &placement);
>> + i915_ttm_placement_from_obj(obj, places, &placement);
>> return __i915_ttm_get_pages(obj, &placement);
>> }
>> @@ -879,9 +874,7 @@ static int __i915_ttm_migrate(struct
>> drm_i915_gem_object *obj,
>> i915_ttm_place_from_region(mr, &requested, obj->bo_offset,
>> obj->base.size, flags);
>> placement.num_placement = 1;
>> - placement.num_busy_placement = 1;
>> placement.placement = &requested;
>> - placement.busy_placement = &requested;
>> ret = __i915_ttm_get_pages(obj, &placement);
>> if (ret)
>> diff --git a/drivers/gpu/drm/loongson/lsdc_ttm.c
>> b/drivers/gpu/drm/loongson/lsdc_ttm.c
>> index bf79dc55afa4..465f622ac05d 100644
>> --- a/drivers/gpu/drm/loongson/lsdc_ttm.c
>> +++ b/drivers/gpu/drm/loongson/lsdc_ttm.c
>> @@ -54,7 +54,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo
>> *lbo, u32 domain)
>> pflags |= TTM_PL_FLAG_TOPDOWN;
>> lbo->placement.placement = lbo->placements;
>> - lbo->placement.busy_placement = lbo->placements;
>> if (domain & LSDC_GEM_DOMAIN_VRAM) {
>> lbo->placements[c].mem_type = TTM_PL_VRAM;
>> @@ -77,7 +76,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo
>> *lbo, u32 domain)
>> }
>> lbo->placement.num_placement = c;
>> - lbo->placement.num_busy_placement = c;
>> for (i = 0; i < c; ++i) {
>> lbo->placements[i].fpfn = 0;
>> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c
>> b/drivers/gpu/drm/nouveau/nouveau_bo.c
>> index 00cc7d1abaa3..56dcd25db1ce 100644
>> --- a/drivers/gpu/drm/nouveau/nouveau_bo.c
>> +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
>> @@ -404,27 +404,6 @@ nouveau_bo_new(struct nouveau_cli *cli, u64
>> size, int align,
>> return 0;
>> }
>> -static void
>> -set_placement_list(struct ttm_place *pl, unsigned *n, uint32_t domain)
>> -{
>> - *n = 0;
>> -
>> - if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
>> - pl[*n].mem_type = TTM_PL_VRAM;
>> - pl[*n].flags = 0;
>> - (*n)++;
>> - }
>> - if (domain & NOUVEAU_GEM_DOMAIN_GART) {
>> - pl[*n].mem_type = TTM_PL_TT;
>> - pl[*n].flags = 0;
>> - (*n)++;
>> - }
>> - if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
>> - pl[*n].mem_type = TTM_PL_SYSTEM;
>> - pl[(*n)++].flags = 0;
>> - }
>> -}
>> -
>> static void
>> set_placement_range(struct nouveau_bo *nvbo, uint32_t domain)
>> {
>> @@ -452,10 +431,6 @@ set_placement_range(struct nouveau_bo *nvbo,
>> uint32_t domain)
>> nvbo->placements[i].fpfn = fpfn;
>> nvbo->placements[i].lpfn = lpfn;
>> }
>> - for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
>> - nvbo->busy_placements[i].fpfn = fpfn;
>> - nvbo->busy_placements[i].lpfn = lpfn;
>> - }
>> }
>> }
>> @@ -463,15 +438,32 @@ void
>> nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t domain,
>> uint32_t busy)
>> {
>> - struct ttm_placement *pl = &nvbo->placement;
>> + unsigned int *n = &nvbo->placement.num_placement;
>> + struct ttm_place *pl = nvbo->placements;
>> - pl->placement = nvbo->placements;
>> - set_placement_list(nvbo->placements, &pl->num_placement, domain);
>> + domain |= busy;
>> - pl->busy_placement = nvbo->busy_placements;
>> - set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
>> - domain | busy);
>> + *n = 0;
>> + if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
>> + pl[*n].mem_type = TTM_PL_VRAM;
>> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_VRAM ?
>> + TTM_PL_FLAG_FALLBACK : 0;
>> + (*n)++;
>> + }
>> + if (domain & NOUVEAU_GEM_DOMAIN_GART) {
>> + pl[*n].mem_type = TTM_PL_TT;
>> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_GART ?
>> + TTM_PL_FLAG_FALLBACK : 0;
>> + (*n)++;
>> + }
>> + if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
>> + pl[*n].mem_type = TTM_PL_SYSTEM;
>> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_CPU ?
>> + TTM_PL_FLAG_FALLBACK : 0;
>> + (*n)++;
>> + }
>> + nvbo->placement.placement = nvbo->placements;
>> set_placement_range(nvbo, domain);
>> }
>> @@ -1314,11 +1306,6 @@ vm_fault_t
>> nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
>> nvbo->placements[i].lpfn = mappable;
>> }
>> - for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
>> - nvbo->busy_placements[i].fpfn = 0;
>> - nvbo->busy_placements[i].lpfn = mappable;
>> - }
>> -
>> nouveau_bo_placement_set(nvbo, NOUVEAU_GEM_DOMAIN_VRAM, 0);
>> }
>> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.h
>> b/drivers/gpu/drm/nouveau/nouveau_bo.h
>> index 70c551921a9e..e9dfab6a8156 100644
>> --- a/drivers/gpu/drm/nouveau/nouveau_bo.h
>> +++ b/drivers/gpu/drm/nouveau/nouveau_bo.h
>> @@ -15,7 +15,6 @@ struct nouveau_bo {
>> struct ttm_placement placement;
>> u32 valid_domains;
>> struct ttm_place placements[3];
>> - struct ttm_place busy_placements[3];
>> bool force_coherent;
>> struct ttm_bo_kmap_obj kmap;
>> struct list_head head;
>> diff --git a/drivers/gpu/drm/qxl/qxl_object.c
>> b/drivers/gpu/drm/qxl/qxl_object.c
>> index 06a58dad5f5c..1e46b0a6e478 100644
>> --- a/drivers/gpu/drm/qxl/qxl_object.c
>> +++ b/drivers/gpu/drm/qxl/qxl_object.c
>> @@ -66,7 +66,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo
>> *qbo, u32 domain)
>> pflag |= TTM_PL_FLAG_TOPDOWN;
>> qbo->placement.placement = qbo->placements;
>> - qbo->placement.busy_placement = qbo->placements;
>> if (domain == QXL_GEM_DOMAIN_VRAM) {
>> qbo->placements[c].mem_type = TTM_PL_VRAM;
>> qbo->placements[c++].flags = pflag;
>> @@ -86,7 +85,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo
>> *qbo, u32 domain)
>> qbo->placements[c++].flags = 0;
>> }
>> qbo->placement.num_placement = c;
>> - qbo->placement.num_busy_placement = c;
>> for (i = 0; i < c; ++i) {
>> qbo->placements[i].fpfn = 0;
>> qbo->placements[i].lpfn = 0;
>> diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c
>> b/drivers/gpu/drm/qxl/qxl_ttm.c
>> index 1a82629bce3f..765a144cea14 100644
>> --- a/drivers/gpu/drm/qxl/qxl_ttm.c
>> +++ b/drivers/gpu/drm/qxl/qxl_ttm.c
>> @@ -60,9 +60,7 @@ static void qxl_evict_flags(struct
>> ttm_buffer_object *bo,
>> if (!qxl_ttm_bo_is_qxl_bo(bo)) {
>> placement->placement = &placements;
>> - placement->busy_placement = &placements;
>> placement->num_placement = 1;
>> - placement->num_busy_placement = 1;
>> return;
>> }
>> qbo = to_qxl_bo(bo);
>> diff --git a/drivers/gpu/drm/radeon/radeon_object.c
>> b/drivers/gpu/drm/radeon/radeon_object.c
>> index 10c0fbd9d2b4..a955f8a2f7fe 100644
>> --- a/drivers/gpu/drm/radeon/radeon_object.c
>> +++ b/drivers/gpu/drm/radeon/radeon_object.c
>> @@ -78,7 +78,6 @@ void radeon_ttm_placement_from_domain(struct
>> radeon_bo *rbo, u32 domain)
>> u32 c = 0, i;
>> rbo->placement.placement = rbo->placements;
>> - rbo->placement.busy_placement = rbo->placements;
>> if (domain & RADEON_GEM_DOMAIN_VRAM) {
>> /* Try placing BOs which don't need CPU access outside of the
>> * CPU accessible part of VRAM
>> @@ -114,7 +113,6 @@ void radeon_ttm_placement_from_domain(struct
>> radeon_bo *rbo, u32 domain)
>> }
>> rbo->placement.num_placement = c;
>> - rbo->placement.num_busy_placement = c;
>> for (i = 0; i < c; ++i) {
>> if ((rbo->flags & RADEON_GEM_CPU_ACCESS) &&
>> diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c
>> b/drivers/gpu/drm/radeon/radeon_ttm.c
>> index de4e6d78f1e1..2078b0000e22 100644
>> --- a/drivers/gpu/drm/radeon/radeon_ttm.c
>> +++ b/drivers/gpu/drm/radeon/radeon_ttm.c
>> @@ -92,9 +92,7 @@ static void radeon_evict_flags(struct
>> ttm_buffer_object *bo,
>> if (!radeon_ttm_bo_is_radeon_bo(bo)) {
>> placement->placement = &placements;
>> - placement->busy_placement = &placements;
>> placement->num_placement = 1;
>> - placement->num_busy_placement = 1;
>> return;
>> }
>> rbo = container_of(bo, struct radeon_bo, tbo);
>> @@ -114,15 +112,11 @@ static void radeon_evict_flags(struct
>> ttm_buffer_object *bo,
>> */
>> radeon_ttm_placement_from_domain(rbo,
>> RADEON_GEM_DOMAIN_VRAM |
>> RADEON_GEM_DOMAIN_GTT);
>> - rbo->placement.num_busy_placement = 0;
>> for (i = 0; i < rbo->placement.num_placement; i++) {
>> if (rbo->placements[i].mem_type == TTM_PL_VRAM) {
>> if (rbo->placements[i].fpfn < fpfn)
>> rbo->placements[i].fpfn = fpfn;
>> - } else {
>> - rbo->placement.busy_placement =
>> - &rbo->placements[i];
>> - rbo->placement.num_busy_placement = 1;
>> + rbo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>> }
>> }
>> } else
>> diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c
>> b/drivers/gpu/drm/radeon/radeon_uvd.c
>> index a2cda184b2b2..058a1c8451b2 100644
>> --- a/drivers/gpu/drm/radeon/radeon_uvd.c
>> +++ b/drivers/gpu/drm/radeon/radeon_uvd.c
>> @@ -324,7 +324,6 @@ void radeon_uvd_force_into_uvd_segment(struct
>> radeon_bo *rbo,
>> rbo->placements[1].fpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
>> rbo->placements[1].lpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
>> rbo->placement.num_placement++;
>> - rbo->placement.num_busy_placement++;
>> }
>> void radeon_uvd_free_handles(struct radeon_device *rdev, struct
>> drm_file *filp)
>> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
>> index 8c1eaa74fa21..a5e11a92e0b9 100644
>> --- a/drivers/gpu/drm/ttm/ttm_bo.c
>> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
>> @@ -410,8 +410,8 @@ static int ttm_bo_bounce_temp_buffer(struct
>> ttm_buffer_object *bo,
>> struct ttm_resource *hop_mem;
>> int ret;
>> - hop_placement.num_placement = hop_placement.num_busy_placement
>> = 1;
>> - hop_placement.placement = hop_placement.busy_placement = hop;
>> + hop_placement.num_placement = 1;
>> + hop_placement.placement = hop;
>> /* find space in the bounce domain */
>> ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
>> @@ -440,10 +440,9 @@ static int ttm_bo_evict(struct ttm_buffer_object
>> *bo,
>> dma_resv_assert_held(bo->base.resv);
>> placement.num_placement = 0;
>> - placement.num_busy_placement = 0;
>> bdev->funcs->evict_flags(bo, &placement);
>> - if (!placement.num_placement && !placement.num_busy_placement) {
>> + if (!placement.num_placement) {
>> ret = ttm_bo_wait_ctx(bo, ctx);
>> if (ret)
>> return ret;
>> @@ -791,6 +790,9 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
>> const struct ttm_place *place = &placement->placement[i];
>> struct ttm_resource_manager *man;
>> + if (place->flags & TTM_PL_FLAG_FALLBACK)
>> + continue;
>> +
>> man = ttm_manager_type(bdev, place->mem_type);
>> if (!man || !ttm_resource_manager_used(man))
>> continue;
>> @@ -813,10 +815,13 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
>> return 0;
>> }
>> - for (i = 0; i < placement->num_busy_placement; ++i) {
>> - const struct ttm_place *place = &placement->busy_placement[i];
>> + for (i = 0; i < placement->num_placement; ++i) {
>> + const struct ttm_place *place = &placement->placement[i];
>> struct ttm_resource_manager *man;
>> + if (place->flags & TTM_PL_FLAG_DESIRED)
>> + continue;
>> +
>> man = ttm_manager_type(bdev, place->mem_type);
>> if (!man || !ttm_resource_manager_used(man))
>> continue;
>> @@ -904,11 +909,11 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
>> /*
>> * Remove the backing store if no placement is given.
>> */
>> - if (!placement->num_placement && !placement->num_busy_placement)
>> + if (!placement->num_placement)
>> return ttm_bo_pipeline_gutting(bo);
>> /* Check whether we need to move buffer. */
>> - if (bo->resource && ttm_resource_compat(bo->resource, placement))
>> + if (bo->resource && ttm_resource_compatible(bo->resource,
>> placement))
>> return 0;
>> /* Moving of pinned BOs is forbidden */
>> diff --git a/drivers/gpu/drm/ttm/ttm_resource.c
>> b/drivers/gpu/drm/ttm/ttm_resource.c
>> index 02b96d23fdb9..fb14f7716cf8 100644
>> --- a/drivers/gpu/drm/ttm/ttm_resource.c
>> +++ b/drivers/gpu/drm/ttm/ttm_resource.c
>> @@ -291,37 +291,15 @@ bool ttm_resource_intersects(struct ttm_device
>> *bdev,
>> }
>> /**
>> - * ttm_resource_compatible - test for compatibility
>> + * ttm_resource_compatible - check if resource is compatible with
>> placement
>> *
>> - * @bdev: TTM device structure
>> - * @res: The resource to test
>> - * @place: The placement to test
>> - * @size: How many bytes the new allocation needs.
>> - *
>> - * Test if @res compatible with @place and @size.
>> + * @res: the resource to check
>> + * @placement: the placement to check against
>> *
>> - * Returns true if the res placement compatible with @place and @size.
>> + * Returns true if the placement is compatible.
>> */
>> -bool ttm_resource_compatible(struct ttm_device *bdev,
>> - struct ttm_resource *res,
>> - const struct ttm_place *place,
>> - size_t size)
>> -{
>> - struct ttm_resource_manager *man;
>> -
>> - if (!res || !place)
>> - return false;
>> -
>> - man = ttm_manager_type(bdev, res->mem_type);
>> - if (!man->func->compatible)
>> - return true;
>> -
>> - return man->func->compatible(man, res, place, size);
>> -}
>> -
>> -static bool ttm_resource_places_compat(struct ttm_resource *res,
>> - const struct ttm_place *places,
>> - unsigned num_placement)
>> +bool ttm_resource_compatible(struct ttm_resource *res,
>> + struct ttm_placement *placement)
>> {
>> struct ttm_buffer_object *bo = res->bo;
>> struct ttm_device *bdev = bo->bdev;
>> @@ -330,44 +308,25 @@ static bool ttm_resource_places_compat(struct
>> ttm_resource *res,
>> if (res->placement & TTM_PL_FLAG_TEMPORARY)
>> return false;
>> - for (i = 0; i < num_placement; i++) {
>> - const struct ttm_place *heap = &places[i];
>> + for (i = 0; i < placement->num_placement; i++) {
>> + const struct ttm_place *place = &placement->placement[i];
>> + struct ttm_resource_manager *man;
>> - if (!ttm_resource_compatible(bdev, res, heap, bo->base.size))
>> + if (res->mem_type != place->mem_type)
>> + continue;
>> +
>> + man = ttm_manager_type(bdev, res->mem_type);
>> + if (man->func->compatible &&
>> + !man->func->compatible(man, res, place, bo->base.size))
>> continue;
>> - if ((res->mem_type == heap->mem_type) &&
>> - (!(heap->flags & TTM_PL_FLAG_CONTIGUOUS) ||
>> + if ((!(place->flags & TTM_PL_FLAG_CONTIGUOUS) ||
>> (res->placement & TTM_PL_FLAG_CONTIGUOUS)))
>> return true;
>> }
>> return false;
>> }
>> -/**
>> - * ttm_resource_compat - check if resource is compatible with placement
>> - *
>> - * @res: the resource to check
>> - * @placement: the placement to check against
>> - *
>> - * Returns true if the placement is compatible.
>> - */
>> -bool ttm_resource_compat(struct ttm_resource *res,
>> - struct ttm_placement *placement)
>> -{
>> - if (ttm_resource_places_compat(res, placement->placement,
>> - placement->num_placement))
>> - return true;
>> -
>> - if ((placement->busy_placement != placement->placement ||
>> - placement->num_busy_placement > placement->num_placement) &&
>> - ttm_resource_places_compat(res, placement->busy_placement,
>> - placement->num_busy_placement))
>> - return true;
>> -
>> - return false;
>> -}
>> -
>> void ttm_resource_set_bo(struct ttm_resource *res,
>> struct ttm_buffer_object *bo)
>> {
>> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>> b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>> index 2bfac3aad7b7..bfd41ce3c8f4 100644
>> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>> @@ -742,9 +742,21 @@ void vmw_bo_move_notify(struct ttm_buffer_object
>> *bo,
>> vmw_resource_unbind_list(vbo);
>> }
>> +static u32 placement_flags(u32 domain, u32 desired, u32 fallback)
>> +{
>> + if (desired & fallback & domain)
>> + return 0;
>> +
>> + if (desired & domain)
>> + return TTM_PL_FLAG_DESIRED;
>> +
>> + return TTM_PL_FLAG_FALLBACK;
>> +}
>> +
>> static u32
>> -set_placement_list(struct ttm_place *pl, u32 domain)
>> +set_placement_list(struct ttm_place *pl, u32 desired, u32 fallback)
>> {
>> + u32 domain = desired | fallback;
>> u32 n = 0;
>> /*
>> @@ -752,35 +764,40 @@ set_placement_list(struct ttm_place *pl, u32
>> domain)
>> */
>> if (domain & VMW_BO_DOMAIN_MOB) {
>> pl[n].mem_type = VMW_PL_MOB;
>> - pl[n].flags = 0;
>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_MOB, desired,
>> + fallback);
>> pl[n].fpfn = 0;
>> pl[n].lpfn = 0;
>> n++;
>> }
>> if (domain & VMW_BO_DOMAIN_GMR) {
>> pl[n].mem_type = VMW_PL_GMR;
>> - pl[n].flags = 0;
>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_GMR, desired,
>> + fallback);
>> pl[n].fpfn = 0;
>> pl[n].lpfn = 0;
>> n++;
>> }
>> if (domain & VMW_BO_DOMAIN_VRAM) {
>> pl[n].mem_type = TTM_PL_VRAM;
>> - pl[n].flags = 0;
>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_VRAM, desired,
>> + fallback);
>> pl[n].fpfn = 0;
>> pl[n].lpfn = 0;
>> n++;
>> }
>> if (domain & VMW_BO_DOMAIN_WAITABLE_SYS) {
>> pl[n].mem_type = VMW_PL_SYSTEM;
>> - pl[n].flags = 0;
>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_WAITABLE_SYS,
>> + desired, fallback);
>> pl[n].fpfn = 0;
>> pl[n].lpfn = 0;
>> n++;
>> }
>> if (domain & VMW_BO_DOMAIN_SYS) {
>> pl[n].mem_type = TTM_PL_SYSTEM;
>> - pl[n].flags = 0;
>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_SYS, desired,
>> + fallback);
>> pl[n].fpfn = 0;
>> pl[n].lpfn = 0;
>> n++;
>> @@ -806,7 +823,7 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32
>> domain, u32 busy_domain)
>> u32 i;
>> pl->placement = bo->places;
>> - pl->num_placement = set_placement_list(bo->places, domain);
>> + pl->num_placement = set_placement_list(bo->places, domain,
>> busy_domain);
>> if (drm_debug_enabled(DRM_UT_DRIVER) && bo->tbo.resource) {
>> for (i = 0; i < pl->num_placement; ++i) {
>> @@ -821,8 +838,6 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32
>> domain, u32 busy_domain)
>> __func__, bo->tbo.resource->mem_type, domain);
>> }
>> - pl->busy_placement = bo->busy_places;
>> - pl->num_busy_placement = set_placement_list(bo->busy_places,
>> busy_domain);
>> }
>> void vmw_bo_placement_set_default_accelerated(struct vmw_bo *bo)
>> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>> b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>> index a84fffcef8e1..4d23d0a70bcb 100644
>> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>> @@ -46,15 +46,11 @@ static const struct ttm_place sys_placement_flags
>> = {
>> struct ttm_placement vmw_vram_placement = {
>> .num_placement = 1,
>> .placement = &vram_placement_flags,
>> - .num_busy_placement = 1,
>> - .busy_placement = &vram_placement_flags
>> };
>> struct ttm_placement vmw_sys_placement = {
>> .num_placement = 1,
>> .placement = &sys_placement_flags,
>> - .num_busy_placement = 1,
>> - .busy_placement = &sys_placement_flags
>> };
>> const size_t vmw_tt_size = sizeof(struct vmw_ttm_tt);
>> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
>> index 8e4a3b1f6b93..1e5edc7f0b85 100644
>> --- a/drivers/gpu/drm/xe/xe_bo.c
>> +++ b/drivers/gpu/drm/xe/xe_bo.c
>> @@ -38,22 +38,26 @@ static const struct ttm_place sys_placement_flags
>> = {
>> static struct ttm_placement sys_placement = {
>> .num_placement = 1,
>> .placement = &sys_placement_flags,
>> - .num_busy_placement = 1,
>> - .busy_placement = &sys_placement_flags,
>> };
>> -static const struct ttm_place tt_placement_flags = {
>> - .fpfn = 0,
>> - .lpfn = 0,
>> - .mem_type = XE_PL_TT,
>> - .flags = 0,
>> +static const struct ttm_place tt_placement_flags[] = {
>> + {
>> + .fpfn = 0,
>> + .lpfn = 0,
>> + .mem_type = XE_PL_TT,
>> + .flags = TTM_PL_FLAG_DESIRED,
>> + },
>> + {
>> + .fpfn = 0,
>> + .lpfn = 0,
>> + .mem_type = XE_PL_SYSTEM,
>> + .flags = TTM_PL_FLAG_FALLBACK,
>> + }
>> };
>> static struct ttm_placement tt_placement = {
>> - .num_placement = 1,
>> - .placement = &tt_placement_flags,
>> - .num_busy_placement = 1,
>> - .busy_placement = &sys_placement_flags,
>> + .num_placement = 2,
>> + .placement = tt_placement_flags,
>> };
>> bool mem_type_is_vram(u32 mem_type)
>> @@ -230,8 +234,6 @@ static int __xe_bo_placement_for_flags(struct
>> xe_device *xe, struct xe_bo *bo,
>> bo->placement = (struct ttm_placement) {
>> .num_placement = c,
>> .placement = bo->placements,
>> - .num_busy_placement = c,
>> - .busy_placement = bo->placements,
>> };
>> return 0;
>> @@ -251,7 +253,6 @@ static void xe_evict_flags(struct
>> ttm_buffer_object *tbo,
>> /* Don't handle scatter gather BOs */
>> if (tbo->type == ttm_bo_type_sg) {
>> placement->num_placement = 0;
>> - placement->num_busy_placement = 0;
>> return;
>> }
>> @@ -1353,8 +1354,6 @@ static int __xe_bo_fixed_placement(struct
>> xe_device *xe,
>> bo->placement = (struct ttm_placement) {
>> .num_placement = 1,
>> .placement = place,
>> - .num_busy_placement = 1,
>> - .busy_placement = place,
>> };
>> return 0;
>> @@ -2112,9 +2111,7 @@ int xe_bo_migrate(struct xe_bo *bo, u32 mem_type)
>> xe_place_from_ttm_type(mem_type, &requested);
>> placement.num_placement = 1;
>> - placement.num_busy_placement = 1;
>> placement.placement = &requested;
>> - placement.busy_placement = &requested;
>> /*
>> * Stolen needs to be handled like below VRAM handling if we
>> ever need
>> diff --git a/include/drm/ttm/ttm_placement.h
>> b/include/drm/ttm/ttm_placement.h
>> index 8074d0f6cae5..b510a4812609 100644
>> --- a/include/drm/ttm/ttm_placement.h
>> +++ b/include/drm/ttm/ttm_placement.h
>> @@ -64,6 +64,12 @@
>> /* For multihop handling */
>> #define TTM_PL_FLAG_TEMPORARY (1 << 2)
>> +/* Placement is never used during eviction */
>> +#define TTM_PL_FLAG_DESIRED (1 << 3)
>> +
>> +/* Placement is only used during eviction */
>> +#define TTM_PL_FLAG_FALLBACK (1 << 4)
>> +
>> /**
>> * struct ttm_place
>> *
>> @@ -86,16 +92,12 @@ struct ttm_place {
>> *
>> * @num_placement: number of preferred placements
>> * @placement: preferred placements
>> - * @num_busy_placement: number of preferred placements when need
>> to evict buffer
>> - * @busy_placement: preferred placements when need to evict buffer
>> *
>> * Structure indicating the placement you request for an object.
>> */
>> struct ttm_placement {
>> unsigned num_placement;
>> const struct ttm_place *placement;
>> - unsigned num_busy_placement;
>> - const struct ttm_place *busy_placement;
>> };
>> #endif
>> diff --git a/include/drm/ttm/ttm_resource.h
>> b/include/drm/ttm/ttm_resource.h
>> index 78a226eba953..1afa13f0c22b 100644
>> --- a/include/drm/ttm/ttm_resource.h
>> +++ b/include/drm/ttm/ttm_resource.h
>> @@ -365,12 +365,8 @@ bool ttm_resource_intersects(struct ttm_device
>> *bdev,
>> struct ttm_resource *res,
>> const struct ttm_place *place,
>> size_t size);
>> -bool ttm_resource_compatible(struct ttm_device *bdev,
>> - struct ttm_resource *res,
>> - const struct ttm_place *place,
>> - size_t size);
>> -bool ttm_resource_compat(struct ttm_resource *res,
>> - struct ttm_placement *placement);
>> +bool ttm_resource_compatible(struct ttm_resource *res,
>> + struct ttm_placement *placement);
>> void ttm_resource_set_bo(struct ttm_resource *res,
>> struct ttm_buffer_object *bo);
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 4/5] drm/ttm: improve idle/busy handling v3
2024-01-12 12:51 ` [PATCH 4/5] drm/ttm: improve idle/busy handling v3 Christian König
@ 2024-01-18 14:24 ` Thomas Hellström
2024-01-18 16:48 ` Thomas Hellström
0 siblings, 1 reply; 19+ messages in thread
From: Thomas Hellström @ 2024-01-18 14:24 UTC (permalink / raw)
To: Christian König, dri-devel, intel-gfx, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
On Fri, 2024-01-12 at 13:51 +0100, Christian König wrote:
> Previously we would never try to move a BO into the preferred
> placements
> when it ever landed in a busy placement since those were considered
> compatible.
>
> Rework the whole handling and finally unify the idle and busy
> handling.
> ttm_bo_validate() is now responsible to try idle placement first and
> then
> use the busy placement if that didn't worked.
>
> Drawback is that we now always try the idle placement first for each
> validation which might cause some additional CPU overhead on
> overcommit.
>
> v2: fix kerneldoc warning and coding style
> v3: take care of XE as well
>
> Signed-off-by: Christian König <christian.koenig@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 2 +-
> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 2 +-
> drivers/gpu/drm/ttm/ttm_bo.c | 131 ++++++++-----------
> --
> drivers/gpu/drm/ttm/ttm_resource.c | 16 ++-
> drivers/gpu/drm/xe/xe_bo.c | 4 +-
> include/drm/ttm/ttm_bo.h | 3 +-
> include/drm/ttm/ttm_resource.h | 3 +-
> 7 files changed, 68 insertions(+), 93 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> index b671b0665492..06fb3fc47eaa 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> @@ -404,7 +404,7 @@ int amdgpu_bo_create_kernel_at(struct
> amdgpu_device *adev,
> (*bo_ptr)->placements[i].lpfn = (offset + size) >>
> PAGE_SHIFT;
> }
> r = ttm_bo_mem_space(&(*bo_ptr)->tbo, &(*bo_ptr)->placement,
> - &(*bo_ptr)->tbo.resource, &ctx);
> + &(*bo_ptr)->tbo.resource, &ctx, false);
> if (r)
> goto error;
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 8722beba494e..f23cdc7c5b08 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -966,7 +966,7 @@ int amdgpu_ttm_alloc_gart(struct
> ttm_buffer_object *bo)
> placements.mem_type = TTM_PL_TT;
> placements.flags = bo->resource->placement;
>
> - r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx);
> + r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx, true);
> if (unlikely(r))
> return r;
>
> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c
> b/drivers/gpu/drm/ttm/ttm_bo.c
> index a5e11a92e0b9..3783be24d832 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
> @@ -414,7 +414,7 @@ static int ttm_bo_bounce_temp_buffer(struct
> ttm_buffer_object *bo,
> hop_placement.placement = hop;
>
> /* find space in the bounce domain */
> - ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
> + ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx,
> true);
> if (ret)
> return ret;
> /* move to the bounce domain */
> @@ -454,7 +454,7 @@ static int ttm_bo_evict(struct ttm_buffer_object
> *bo,
> return ttm_bo_pipeline_gutting(bo);
> }
>
> - ret = ttm_bo_mem_space(bo, &placement, &evict_mem, ctx);
> + ret = ttm_bo_mem_space(bo, &placement, &evict_mem, ctx,
> true);
This is what breaks xe's selftest since the evict-flags preferred
placement is never tried so it changes the behavior.
The xe evict flags were set up to "Try to evict to TT first, but if
that causes recursive eviction, try to evict to system". This here
ignores the preferred eviction placement.
So "Preferred" in effect changes meaning from "Don't try this if it
will cause an eviction", to "Don't try this in the evict path", which
is hard for the driver to have any knowledge about.
Then also it sounds from the commit message with this patch "Preferred"
also gets overloaded with "Always retry preferred on validate", but
shouldn't that really be a dynamic driver decision and not something
TTM should try to enforce in a static way? Drivers could have short-
circuited the ttm_bo_validate() call if it succeeded once, and have a
deeper thought about when to migrate from, say TT to VRAM and vice
versa.
For the specific behaviour sought here, there is (or at least used to
be) a construct in the vmwgfx driver that first called
ttm_bo_validate() with VRAM as preferred placement and no fallback. If
that failed due to VRAM being full, it called ttm_bo_validate() again,
this time with fallback and VRAM allowing eviction.
/Thomas
> if (ret) {
> if (ret != -ERESTARTSYS) {
> pr_err("Failed to find memory space for
> buffer 0x%p eviction\n",
> @@ -724,37 +724,6 @@ static int ttm_bo_add_move_fence(struct
> ttm_buffer_object *bo,
> return ret;
> }
>
> -/*
> - * Repeatedly evict memory from the LRU for @mem_type until we
> create enough
> - * space, or we've evicted everything and there isn't enough space.
> - */
> -static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo,
> - const struct ttm_place *place,
> - struct ttm_resource **mem,
> - struct ttm_operation_ctx *ctx)
> -{
> - struct ttm_device *bdev = bo->bdev;
> - struct ttm_resource_manager *man;
> - struct ww_acquire_ctx *ticket;
> - int ret;
> -
> - man = ttm_manager_type(bdev, place->mem_type);
> - ticket = dma_resv_locking_ctx(bo->base.resv);
> - do {
> - ret = ttm_resource_alloc(bo, place, mem);
> - if (likely(!ret))
> - break;
> - if (unlikely(ret != -ENOSPC))
> - return ret;
> - ret = ttm_mem_evict_first(bdev, man, place, ctx,
> - ticket);
> - if (unlikely(ret != 0))
> - return ret;
> - } while (1);
> -
> - return ttm_bo_add_move_fence(bo, man, *mem, ctx-
> >no_wait_gpu);
> -}
> -
> /**
> * ttm_bo_mem_space
> *
> @@ -763,6 +732,7 @@ static int ttm_bo_mem_force_space(struct
> ttm_buffer_object *bo,
> * @placement: Proposed new placement for the buffer object.
> * @mem: A struct ttm_resource.
> * @ctx: if and how to sleep, lock buffers and alloc memory
> + * @force_space: If we should evict buffers to force space
> *
> * Allocate memory space for the buffer object pointed to by @bo,
> using
> * the placement flags in @placement, potentially evicting other
> idle buffer objects.
> @@ -776,12 +746,14 @@ static int ttm_bo_mem_force_space(struct
> ttm_buffer_object *bo,
> int ttm_bo_mem_space(struct ttm_buffer_object *bo,
> struct ttm_placement *placement,
> struct ttm_resource **mem,
> - struct ttm_operation_ctx *ctx)
> + struct ttm_operation_ctx *ctx,
> + bool force_space)
> {
> struct ttm_device *bdev = bo->bdev;
> - bool type_found = false;
> + struct ww_acquire_ctx *ticket;
> int i, ret;
>
> + ticket = dma_resv_locking_ctx(bo->base.resv);
> ret = dma_resv_reserve_fences(bo->base.resv, 1);
> if (unlikely(ret))
> return ret;
> @@ -790,19 +762,30 @@ int ttm_bo_mem_space(struct ttm_buffer_object
> *bo,
> const struct ttm_place *place = &placement-
> >placement[i];
> struct ttm_resource_manager *man;
>
> - if (place->flags & TTM_PL_FLAG_FALLBACK)
> - continue;
> -
> man = ttm_manager_type(bdev, place->mem_type);
> if (!man || !ttm_resource_manager_used(man))
> continue;
>
> - type_found = true;
> - ret = ttm_resource_alloc(bo, place, mem);
> - if (ret == -ENOSPC)
> + if (place->flags & (force_space ?
> TTM_PL_FLAG_DESIRED :
> + TTM_PL_FLAG_FALLBACK))
> + continue;
> +
> + do {
> + ret = ttm_resource_alloc(bo, place, mem);
> + if (unlikely(ret != -ENOSPC))
> + return ret;
> + if (likely(!ret) || !force_space)
> + break;
> +
> + ret = ttm_mem_evict_first(bdev, man, place,
> ctx,
> + ticket);
> + if (unlikely(ret == -EBUSY))
> + break;
> + if (unlikely(ret))
> + return ret;
> + } while (1);
> + if (ret)
> continue;
> - if (unlikely(ret))
> - goto error;
>
> ret = ttm_bo_add_move_fence(bo, man, *mem, ctx-
> >no_wait_gpu);
> if (unlikely(ret)) {
> @@ -810,45 +793,19 @@ int ttm_bo_mem_space(struct ttm_buffer_object
> *bo,
> if (ret == -EBUSY)
> continue;
>
> - goto error;
> + return ret;
> }
> return 0;
> }
>
> - for (i = 0; i < placement->num_placement; ++i) {
> - const struct ttm_place *place = &placement-
> >placement[i];
> - struct ttm_resource_manager *man;
> -
> - if (place->flags & TTM_PL_FLAG_DESIRED)
> - continue;
> -
> - man = ttm_manager_type(bdev, place->mem_type);
> - if (!man || !ttm_resource_manager_used(man))
> - continue;
> -
> - type_found = true;
> - ret = ttm_bo_mem_force_space(bo, place, mem, ctx);
> - if (likely(!ret))
> - return 0;
> -
> - if (ret && ret != -EBUSY)
> - goto error;
> - }
> -
> - ret = -ENOSPC;
> - if (!type_found) {
> - pr_err(TTM_PFX "No compatible memory type found\n");
> - ret = -EINVAL;
> - }
> -
> -error:
> - return ret;
> + return -ENOSPC;
> }
> EXPORT_SYMBOL(ttm_bo_mem_space);
>
> static int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
> struct ttm_placement *placement,
> - struct ttm_operation_ctx *ctx)
> + struct ttm_operation_ctx *ctx,
> + bool force_space)
> {
> struct ttm_resource *mem;
> struct ttm_place hop;
> @@ -865,7 +822,7 @@ static int ttm_bo_move_buffer(struct
> ttm_buffer_object *bo,
> * stop and the driver will be called to make
> * the second hop.
> */
> - ret = ttm_bo_mem_space(bo, placement, &mem, ctx);
> + ret = ttm_bo_mem_space(bo, placement, &mem, ctx,
> force_space);
> if (ret)
> return ret;
> bounce:
> @@ -902,6 +859,7 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
> struct ttm_placement *placement,
> struct ttm_operation_ctx *ctx)
> {
> + bool force_space;
> int ret;
>
> dma_resv_assert_held(bo->base.resv);
> @@ -912,20 +870,27 @@ int ttm_bo_validate(struct ttm_buffer_object
> *bo,
> if (!placement->num_placement)
> return ttm_bo_pipeline_gutting(bo);
>
> - /* Check whether we need to move buffer. */
> - if (bo->resource && ttm_resource_compatible(bo->resource,
> placement))
> - return 0;
> + force_space = false;
> + do {
> + /* Check whether we need to move buffer. */
> + if (bo->resource &&
> + ttm_resource_compatible(bo->resource, placement,
> + force_space))
> + return 0;
> +
> + /* Moving of pinned BOs is forbidden */
> + if (bo->pin_count)
> + return -EINVAL;
>
> - /* Moving of pinned BOs is forbidden */
> - if (bo->pin_count)
> - return -EINVAL;
> + ret = ttm_bo_move_buffer(bo, placement, ctx,
> force_space);
> + if (ret && ret != -ENOSPC)
> + return ret;
>
> - ret = ttm_bo_move_buffer(bo, placement, ctx);
> + force_space = !force_space;
> + } while (force_space);
> /* For backward compatibility with userspace */
> if (ret == -ENOSPC)
> return -ENOMEM;
> - if (ret)
> - return ret;
>
> /*
> * We might need to add a TTM.
> diff --git a/drivers/gpu/drm/ttm/ttm_resource.c
> b/drivers/gpu/drm/ttm/ttm_resource.c
> index fb14f7716cf8..65155f2013ca 100644
> --- a/drivers/gpu/drm/ttm/ttm_resource.c
> +++ b/drivers/gpu/drm/ttm/ttm_resource.c
> @@ -295,11 +295,13 @@ bool ttm_resource_intersects(struct ttm_device
> *bdev,
> *
> * @res: the resource to check
> * @placement: the placement to check against
> + * @evicting: true if the caller is doing evictions
> *
> * Returns true if the placement is compatible.
> */
> bool ttm_resource_compatible(struct ttm_resource *res,
> - struct ttm_placement *placement)
> + struct ttm_placement *placement,
> + bool evicting)
> {
> struct ttm_buffer_object *bo = res->bo;
> struct ttm_device *bdev = bo->bdev;
> @@ -315,14 +317,20 @@ bool ttm_resource_compatible(struct
> ttm_resource *res,
> if (res->mem_type != place->mem_type)
> continue;
>
> + if (place->flags & (evicting ? TTM_PL_FLAG_DESIRED :
> + TTM_PL_FLAG_FALLBACK))
> + continue;
> +
> + if (place->flags & TTM_PL_FLAG_CONTIGUOUS &&
> + !(res->placement & TTM_PL_FLAG_CONTIGUOUS))
> + continue;
> +
> man = ttm_manager_type(bdev, res->mem_type);
> if (man->func->compatible &&
> !man->func->compatible(man, res, place, bo-
> >base.size))
> continue;
>
> - if ((!(place->flags & TTM_PL_FLAG_CONTIGUOUS) ||
> - (res->placement & TTM_PL_FLAG_CONTIGUOUS)))
> - return true;
> + return true;
> }
> return false;
> }
> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> index 1e5edc7f0b85..d4b8bd774ada 100644
> --- a/drivers/gpu/drm/xe/xe_bo.c
> +++ b/drivers/gpu/drm/xe/xe_bo.c
> @@ -828,7 +828,7 @@ int xe_bo_evict_pinned(struct xe_bo *bo)
> if (WARN_ON(!xe_bo_is_vram(bo)))
> return -EINVAL;
>
> - ret = ttm_bo_mem_space(&bo->ttm, &placement, &new_mem,
> &ctx);
> + ret = ttm_bo_mem_space(&bo->ttm, &placement, &new_mem, &ctx,
> false);
> if (ret)
> return ret;
>
> @@ -893,7 +893,7 @@ int xe_bo_restore_pinned(struct xe_bo *bo)
> if (WARN_ON(xe_bo_is_vram(bo) || !bo->ttm.ttm))
> return -EINVAL;
>
> - ret = ttm_bo_mem_space(&bo->ttm, &bo->placement, &new_mem,
> &ctx);
> + ret = ttm_bo_mem_space(&bo->ttm, &bo->placement, &new_mem,
> &ctx, false);
> if (ret)
> return ret;
>
> diff --git a/include/drm/ttm/ttm_bo.h b/include/drm/ttm/ttm_bo.h
> index 0223a41a64b2..64d738f152d1 100644
> --- a/include/drm/ttm/ttm_bo.h
> +++ b/include/drm/ttm/ttm_bo.h
> @@ -397,7 +397,8 @@ vm_fault_t ttm_bo_vm_dummy_page(struct vm_fault
> *vmf, pgprot_t prot);
> int ttm_bo_mem_space(struct ttm_buffer_object *bo,
> struct ttm_placement *placement,
> struct ttm_resource **mem,
> - struct ttm_operation_ctx *ctx);
> + struct ttm_operation_ctx *ctx,
> + bool force_space);
>
> void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo);
> /*
> diff --git a/include/drm/ttm/ttm_resource.h
> b/include/drm/ttm/ttm_resource.h
> index 1afa13f0c22b..7561023db43d 100644
> --- a/include/drm/ttm/ttm_resource.h
> +++ b/include/drm/ttm/ttm_resource.h
> @@ -366,7 +366,8 @@ bool ttm_resource_intersects(struct ttm_device
> *bdev,
> const struct ttm_place *place,
> size_t size);
> bool ttm_resource_compatible(struct ttm_resource *res,
> - struct ttm_placement *placement);
> + struct ttm_placement *placement,
> + bool evicting);
> void ttm_resource_set_bo(struct ttm_resource *res,
> struct ttm_buffer_object *bo);
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 3/5] drm/ttm: replace busy placement with flags v6
2024-01-17 12:27 ` Thomas Hellström
@ 2024-01-18 16:38 ` Thomas Hellström
2024-01-26 22:16 ` Lucas De Marchi
0 siblings, 1 reply; 19+ messages in thread
From: Thomas Hellström @ 2024-01-18 16:38 UTC (permalink / raw)
To: Christian König, dri-devel, intel-gfx, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
On 1/17/24 13:27, Thomas Hellström wrote:
>
> On 1/17/24 11:47, Thomas Hellström wrote:
>> Hi, Christian
>>
>> Xe changes look good. Will send the series to xe ci to check for
>> regressions.
>
> Hmm, there are some checkpatch warnings about author / SOB email
> mismatch,
With those fixed, this patch is
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
>
> But worserthere are some regressions in the dma-buf ktest (it tests
> evicting of a dynamic dma-buf),
>
> https://patchwork.freedesktop.org/series/128873/
>
> I'll take a look later today or tomorrow.
These are from the next patch. Will continue the discussion there.
/Thomas
>
> /Thomas
>
>
>
>>
>> /Thomas
>>
>>
>> On 1/12/24 13:51, Christian König wrote:
>>> From: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
>>>
>>> Instead of a list of separate busy placement add flags which indicate
>>> that a placement should only be used when there is room or if we
>>> need to
>>> evict.
>>>
>>> v2: add missing TTM_PL_FLAG_IDLE for i915
>>> v3: fix auto build test ERROR on drm-tip/drm-tip
>>> v4: fix some typos pointed out by checkpatch
>>> v5: cleanup some rebase problems with VMWGFX
>>> v6: implement some missing VMWGFX functionality pointed out by Zack,
>>> rename the flags as suggested by Michel, rebase on drm-tip and
>>> adjust XE as well
>>>
>>> Signed-off-by: Christian König <christian.koenig@amd.com>
>>> Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
>>> ---
>>> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 6 +-
>>> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 +---
>>> drivers/gpu/drm/drm_gem_vram_helper.c | 2 -
>>> drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 37 +++++------
>>> drivers/gpu/drm/loongson/lsdc_ttm.c | 2 -
>>> drivers/gpu/drm/nouveau/nouveau_bo.c | 59 +++++++----------
>>> drivers/gpu/drm/nouveau/nouveau_bo.h | 1 -
>>> drivers/gpu/drm/qxl/qxl_object.c | 2 -
>>> drivers/gpu/drm/qxl/qxl_ttm.c | 2 -
>>> drivers/gpu/drm/radeon/radeon_object.c | 2 -
>>> drivers/gpu/drm/radeon/radeon_ttm.c | 8 +--
>>> drivers/gpu/drm/radeon/radeon_uvd.c | 1 -
>>> drivers/gpu/drm/ttm/ttm_bo.c | 21 ++++---
>>> drivers/gpu/drm/ttm/ttm_resource.c | 73
>>> +++++-----------------
>>> drivers/gpu/drm/vmwgfx/vmwgfx_bo.c | 33 +++++++---
>>> drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 4 --
>>> drivers/gpu/drm/xe/xe_bo.c | 33 +++++-----
>>> include/drm/ttm/ttm_placement.h | 10 +--
>>> include/drm/ttm/ttm_resource.h | 8 +--
>>> 19 files changed, 118 insertions(+), 197 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>> index 425cebcc5cbf..b671b0665492 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>> @@ -220,9 +220,6 @@ void amdgpu_bo_placement_from_domain(struct
>>> amdgpu_bo *abo, u32 domain)
>>> placement->num_placement = c;
>>> placement->placement = places;
>>> -
>>> - placement->num_busy_placement = c;
>>> - placement->busy_placement = places;
>>> }
>>> /**
>>> @@ -1397,8 +1394,7 @@ vm_fault_t
>>> amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
>>> AMDGPU_GEM_DOMAIN_GTT);
>>> /* Avoid costly evictions; only set GTT as a busy placement */
>>> - abo->placement.num_busy_placement = 1;
>>> - abo->placement.busy_placement = &abo->placements[1];
>>> + abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>>> r = ttm_bo_validate(bo, &abo->placement, &ctx);
>>> if (unlikely(r == -EBUSY || r == -ERESTARTSYS))
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> index 75c9fd2c6c2a..8722beba494e 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> @@ -102,23 +102,19 @@ static void amdgpu_evict_flags(struct
>>> ttm_buffer_object *bo,
>>> /* Don't handle scatter gather BOs */
>>> if (bo->type == ttm_bo_type_sg) {
>>> placement->num_placement = 0;
>>> - placement->num_busy_placement = 0;
>>> return;
>>> }
>>> /* Object isn't an AMDGPU object so ignore */
>>> if (!amdgpu_bo_is_amdgpu_bo(bo)) {
>>> placement->placement = &placements;
>>> - placement->busy_placement = &placements;
>>> placement->num_placement = 1;
>>> - placement->num_busy_placement = 1;
>>> return;
>>> }
>>> abo = ttm_to_amdgpu_bo(bo);
>>> if (abo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) {
>>> placement->num_placement = 0;
>>> - placement->num_busy_placement = 0;
>>> return;
>>> }
>>> @@ -128,13 +124,13 @@ static void amdgpu_evict_flags(struct
>>> ttm_buffer_object *bo,
>>> case AMDGPU_PL_OA:
>>> case AMDGPU_PL_DOORBELL:
>>> placement->num_placement = 0;
>>> - placement->num_busy_placement = 0;
>>> return;
>>> case TTM_PL_VRAM:
>>> if (!adev->mman.buffer_funcs_enabled) {
>>> /* Move to system memory */
>>> amdgpu_bo_placement_from_domain(abo,
>>> AMDGPU_GEM_DOMAIN_CPU);
>>> +
>>> } else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
>>> !(abo->flags &
>>> AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) &&
>>> amdgpu_bo_in_cpu_visible_vram(abo)) {
>>> @@ -149,8 +145,7 @@ static void amdgpu_evict_flags(struct
>>> ttm_buffer_object *bo,
>>> AMDGPU_GEM_DOMAIN_CPU);
>>> abo->placements[0].fpfn = adev->gmc.visible_vram_size
>>> >> PAGE_SHIFT;
>>> abo->placements[0].lpfn = 0;
>>> - abo->placement.busy_placement = &abo->placements[1];
>>> - abo->placement.num_busy_placement = 1;
>>> + abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>>> } else {
>>> /* Move to GTT memory */
>>> amdgpu_bo_placement_from_domain(abo,
>>> AMDGPU_GEM_DOMAIN_GTT |
>>> @@ -966,8 +961,6 @@ int amdgpu_ttm_alloc_gart(struct
>>> ttm_buffer_object *bo)
>>> /* allocate GART space */
>>> placement.num_placement = 1;
>>> placement.placement = &placements;
>>> - placement.num_busy_placement = 1;
>>> - placement.busy_placement = &placements;
>>> placements.fpfn = 0;
>>> placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
>>> placements.mem_type = TTM_PL_TT;
>>> diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c
>>> b/drivers/gpu/drm/drm_gem_vram_helper.c
>>> index b67eafa55715..75f2eaf0d5b6 100644
>>> --- a/drivers/gpu/drm/drm_gem_vram_helper.c
>>> +++ b/drivers/gpu/drm/drm_gem_vram_helper.c
>>> @@ -147,7 +147,6 @@ static void drm_gem_vram_placement(struct
>>> drm_gem_vram_object *gbo,
>>> invariant_flags = TTM_PL_FLAG_TOPDOWN;
>>> gbo->placement.placement = gbo->placements;
>>> - gbo->placement.busy_placement = gbo->placements;
>>> if (pl_flag & DRM_GEM_VRAM_PL_FLAG_VRAM) {
>>> gbo->placements[c].mem_type = TTM_PL_VRAM;
>>> @@ -160,7 +159,6 @@ static void drm_gem_vram_placement(struct
>>> drm_gem_vram_object *gbo,
>>> }
>>> gbo->placement.num_placement = c;
>>> - gbo->placement.num_busy_placement = c;
>>> for (i = 0; i < c; ++i) {
>>> gbo->placements[i].fpfn = 0;
>>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>>> b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>>> index 6b69ef0cdbb4..80c6cafc8887 100644
>>> --- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>>> @@ -65,8 +65,6 @@ static const struct ttm_place sys_placement_flags = {
>>> static struct ttm_placement i915_sys_placement = {
>>> .num_placement = 1,
>>> .placement = &sys_placement_flags,
>>> - .num_busy_placement = 1,
>>> - .busy_placement = &sys_placement_flags,
>>> };
>>> /**
>>> @@ -157,32 +155,28 @@ i915_ttm_place_from_region(const struct
>>> intel_memory_region *mr,
>>> static void
>>> i915_ttm_placement_from_obj(const struct drm_i915_gem_object *obj,
>>> - struct ttm_place *requested,
>>> - struct ttm_place *busy,
>>> + struct ttm_place *places,
>>> struct ttm_placement *placement)
>>> {
>>> unsigned int num_allowed = obj->mm.n_placements;
>>> unsigned int flags = obj->flags;
>>> unsigned int i;
>>> - placement->num_placement = 1;
>>> + places[0].flags |= TTM_PL_FLAG_DESIRED;
>>> i915_ttm_place_from_region(num_allowed ? obj->mm.placements[0] :
>>> - obj->mm.region, requested, obj->bo_offset,
>>> + obj->mm.region, &places[0], obj->bo_offset,
>>> obj->base.size, flags);
>>> /* Cache this on object? */
>>> - placement->num_busy_placement = num_allowed;
>>> - for (i = 0; i < placement->num_busy_placement; ++i)
>>> - i915_ttm_place_from_region(obj->mm.placements[i], busy + i,
>>> - obj->bo_offset, obj->base.size, flags);
>>> -
>>> - if (num_allowed == 0) {
>>> - *busy = *requested;
>>> - placement->num_busy_placement = 1;
>>> + for (i = 0; i < num_allowed; ++i) {
>>> + i915_ttm_place_from_region(obj->mm.placements[i],
>>> + &places[i + 1], obj->bo_offset,
>>> + obj->base.size, flags);
>>> + places[i + 1].flags |= TTM_PL_FLAG_FALLBACK;
>>> }
>>> - placement->placement = requested;
>>> - placement->busy_placement = busy;
>>> + placement->num_placement = num_allowed + 1;
>>> + placement->placement = places;
>>> }
>>> static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev,
>>> @@ -789,7 +783,8 @@ static int __i915_ttm_get_pages(struct
>>> drm_i915_gem_object *obj,
>>> int ret;
>>> /* First try only the requested placement. No eviction. */
>>> - real_num_busy = fetch_and_zero(&placement->num_busy_placement);
>>> + real_num_busy = placement->num_placement;
>>> + placement->num_placement = 1;
>>> ret = ttm_bo_validate(bo, placement, &ctx);
>>> if (ret) {
>>> ret = i915_ttm_err_to_gem(ret);
>>> @@ -805,7 +800,7 @@ static int __i915_ttm_get_pages(struct
>>> drm_i915_gem_object *obj,
>>> * If the initial attempt fails, allow all accepted
>>> placements,
>>> * evicting if necessary.
>>> */
>>> - placement->num_busy_placement = real_num_busy;
>>> + placement->num_placement = real_num_busy;
>>> ret = ttm_bo_validate(bo, placement, &ctx);
>>> if (ret)
>>> return i915_ttm_err_to_gem(ret);
>>> @@ -839,7 +834,7 @@ static int __i915_ttm_get_pages(struct
>>> drm_i915_gem_object *obj,
>>> static int i915_ttm_get_pages(struct drm_i915_gem_object *obj)
>>> {
>>> - struct ttm_place requested, busy[I915_TTM_MAX_PLACEMENTS];
>>> + struct ttm_place places[I915_TTM_MAX_PLACEMENTS + 1];
>>> struct ttm_placement placement;
>>> /* restricted by sg_alloc_table */
>>> @@ -849,7 +844,7 @@ static int i915_ttm_get_pages(struct
>>> drm_i915_gem_object *obj)
>>> GEM_BUG_ON(obj->mm.n_placements > I915_TTM_MAX_PLACEMENTS);
>>> /* Move to the requested placement. */
>>> - i915_ttm_placement_from_obj(obj, &requested, busy, &placement);
>>> + i915_ttm_placement_from_obj(obj, places, &placement);
>>> return __i915_ttm_get_pages(obj, &placement);
>>> }
>>> @@ -879,9 +874,7 @@ static int __i915_ttm_migrate(struct
>>> drm_i915_gem_object *obj,
>>> i915_ttm_place_from_region(mr, &requested, obj->bo_offset,
>>> obj->base.size, flags);
>>> placement.num_placement = 1;
>>> - placement.num_busy_placement = 1;
>>> placement.placement = &requested;
>>> - placement.busy_placement = &requested;
>>> ret = __i915_ttm_get_pages(obj, &placement);
>>> if (ret)
>>> diff --git a/drivers/gpu/drm/loongson/lsdc_ttm.c
>>> b/drivers/gpu/drm/loongson/lsdc_ttm.c
>>> index bf79dc55afa4..465f622ac05d 100644
>>> --- a/drivers/gpu/drm/loongson/lsdc_ttm.c
>>> +++ b/drivers/gpu/drm/loongson/lsdc_ttm.c
>>> @@ -54,7 +54,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo
>>> *lbo, u32 domain)
>>> pflags |= TTM_PL_FLAG_TOPDOWN;
>>> lbo->placement.placement = lbo->placements;
>>> - lbo->placement.busy_placement = lbo->placements;
>>> if (domain & LSDC_GEM_DOMAIN_VRAM) {
>>> lbo->placements[c].mem_type = TTM_PL_VRAM;
>>> @@ -77,7 +76,6 @@ static void lsdc_bo_set_placement(struct lsdc_bo
>>> *lbo, u32 domain)
>>> }
>>> lbo->placement.num_placement = c;
>>> - lbo->placement.num_busy_placement = c;
>>> for (i = 0; i < c; ++i) {
>>> lbo->placements[i].fpfn = 0;
>>> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c
>>> b/drivers/gpu/drm/nouveau/nouveau_bo.c
>>> index 00cc7d1abaa3..56dcd25db1ce 100644
>>> --- a/drivers/gpu/drm/nouveau/nouveau_bo.c
>>> +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
>>> @@ -404,27 +404,6 @@ nouveau_bo_new(struct nouveau_cli *cli, u64
>>> size, int align,
>>> return 0;
>>> }
>>> -static void
>>> -set_placement_list(struct ttm_place *pl, unsigned *n, uint32_t domain)
>>> -{
>>> - *n = 0;
>>> -
>>> - if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
>>> - pl[*n].mem_type = TTM_PL_VRAM;
>>> - pl[*n].flags = 0;
>>> - (*n)++;
>>> - }
>>> - if (domain & NOUVEAU_GEM_DOMAIN_GART) {
>>> - pl[*n].mem_type = TTM_PL_TT;
>>> - pl[*n].flags = 0;
>>> - (*n)++;
>>> - }
>>> - if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
>>> - pl[*n].mem_type = TTM_PL_SYSTEM;
>>> - pl[(*n)++].flags = 0;
>>> - }
>>> -}
>>> -
>>> static void
>>> set_placement_range(struct nouveau_bo *nvbo, uint32_t domain)
>>> {
>>> @@ -452,10 +431,6 @@ set_placement_range(struct nouveau_bo *nvbo,
>>> uint32_t domain)
>>> nvbo->placements[i].fpfn = fpfn;
>>> nvbo->placements[i].lpfn = lpfn;
>>> }
>>> - for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
>>> - nvbo->busy_placements[i].fpfn = fpfn;
>>> - nvbo->busy_placements[i].lpfn = lpfn;
>>> - }
>>> }
>>> }
>>> @@ -463,15 +438,32 @@ void
>>> nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t domain,
>>> uint32_t busy)
>>> {
>>> - struct ttm_placement *pl = &nvbo->placement;
>>> + unsigned int *n = &nvbo->placement.num_placement;
>>> + struct ttm_place *pl = nvbo->placements;
>>> - pl->placement = nvbo->placements;
>>> - set_placement_list(nvbo->placements, &pl->num_placement, domain);
>>> + domain |= busy;
>>> - pl->busy_placement = nvbo->busy_placements;
>>> - set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
>>> - domain | busy);
>>> + *n = 0;
>>> + if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
>>> + pl[*n].mem_type = TTM_PL_VRAM;
>>> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_VRAM ?
>>> + TTM_PL_FLAG_FALLBACK : 0;
>>> + (*n)++;
>>> + }
>>> + if (domain & NOUVEAU_GEM_DOMAIN_GART) {
>>> + pl[*n].mem_type = TTM_PL_TT;
>>> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_GART ?
>>> + TTM_PL_FLAG_FALLBACK : 0;
>>> + (*n)++;
>>> + }
>>> + if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
>>> + pl[*n].mem_type = TTM_PL_SYSTEM;
>>> + pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_CPU ?
>>> + TTM_PL_FLAG_FALLBACK : 0;
>>> + (*n)++;
>>> + }
>>> + nvbo->placement.placement = nvbo->placements;
>>> set_placement_range(nvbo, domain);
>>> }
>>> @@ -1314,11 +1306,6 @@ vm_fault_t
>>> nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
>>> nvbo->placements[i].lpfn = mappable;
>>> }
>>> - for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
>>> - nvbo->busy_placements[i].fpfn = 0;
>>> - nvbo->busy_placements[i].lpfn = mappable;
>>> - }
>>> -
>>> nouveau_bo_placement_set(nvbo, NOUVEAU_GEM_DOMAIN_VRAM, 0);
>>> }
>>> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.h
>>> b/drivers/gpu/drm/nouveau/nouveau_bo.h
>>> index 70c551921a9e..e9dfab6a8156 100644
>>> --- a/drivers/gpu/drm/nouveau/nouveau_bo.h
>>> +++ b/drivers/gpu/drm/nouveau/nouveau_bo.h
>>> @@ -15,7 +15,6 @@ struct nouveau_bo {
>>> struct ttm_placement placement;
>>> u32 valid_domains;
>>> struct ttm_place placements[3];
>>> - struct ttm_place busy_placements[3];
>>> bool force_coherent;
>>> struct ttm_bo_kmap_obj kmap;
>>> struct list_head head;
>>> diff --git a/drivers/gpu/drm/qxl/qxl_object.c
>>> b/drivers/gpu/drm/qxl/qxl_object.c
>>> index 06a58dad5f5c..1e46b0a6e478 100644
>>> --- a/drivers/gpu/drm/qxl/qxl_object.c
>>> +++ b/drivers/gpu/drm/qxl/qxl_object.c
>>> @@ -66,7 +66,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo
>>> *qbo, u32 domain)
>>> pflag |= TTM_PL_FLAG_TOPDOWN;
>>> qbo->placement.placement = qbo->placements;
>>> - qbo->placement.busy_placement = qbo->placements;
>>> if (domain == QXL_GEM_DOMAIN_VRAM) {
>>> qbo->placements[c].mem_type = TTM_PL_VRAM;
>>> qbo->placements[c++].flags = pflag;
>>> @@ -86,7 +85,6 @@ void qxl_ttm_placement_from_domain(struct qxl_bo
>>> *qbo, u32 domain)
>>> qbo->placements[c++].flags = 0;
>>> }
>>> qbo->placement.num_placement = c;
>>> - qbo->placement.num_busy_placement = c;
>>> for (i = 0; i < c; ++i) {
>>> qbo->placements[i].fpfn = 0;
>>> qbo->placements[i].lpfn = 0;
>>> diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c
>>> b/drivers/gpu/drm/qxl/qxl_ttm.c
>>> index 1a82629bce3f..765a144cea14 100644
>>> --- a/drivers/gpu/drm/qxl/qxl_ttm.c
>>> +++ b/drivers/gpu/drm/qxl/qxl_ttm.c
>>> @@ -60,9 +60,7 @@ static void qxl_evict_flags(struct
>>> ttm_buffer_object *bo,
>>> if (!qxl_ttm_bo_is_qxl_bo(bo)) {
>>> placement->placement = &placements;
>>> - placement->busy_placement = &placements;
>>> placement->num_placement = 1;
>>> - placement->num_busy_placement = 1;
>>> return;
>>> }
>>> qbo = to_qxl_bo(bo);
>>> diff --git a/drivers/gpu/drm/radeon/radeon_object.c
>>> b/drivers/gpu/drm/radeon/radeon_object.c
>>> index 10c0fbd9d2b4..a955f8a2f7fe 100644
>>> --- a/drivers/gpu/drm/radeon/radeon_object.c
>>> +++ b/drivers/gpu/drm/radeon/radeon_object.c
>>> @@ -78,7 +78,6 @@ void radeon_ttm_placement_from_domain(struct
>>> radeon_bo *rbo, u32 domain)
>>> u32 c = 0, i;
>>> rbo->placement.placement = rbo->placements;
>>> - rbo->placement.busy_placement = rbo->placements;
>>> if (domain & RADEON_GEM_DOMAIN_VRAM) {
>>> /* Try placing BOs which don't need CPU access outside of the
>>> * CPU accessible part of VRAM
>>> @@ -114,7 +113,6 @@ void radeon_ttm_placement_from_domain(struct
>>> radeon_bo *rbo, u32 domain)
>>> }
>>> rbo->placement.num_placement = c;
>>> - rbo->placement.num_busy_placement = c;
>>> for (i = 0; i < c; ++i) {
>>> if ((rbo->flags & RADEON_GEM_CPU_ACCESS) &&
>>> diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c
>>> b/drivers/gpu/drm/radeon/radeon_ttm.c
>>> index de4e6d78f1e1..2078b0000e22 100644
>>> --- a/drivers/gpu/drm/radeon/radeon_ttm.c
>>> +++ b/drivers/gpu/drm/radeon/radeon_ttm.c
>>> @@ -92,9 +92,7 @@ static void radeon_evict_flags(struct
>>> ttm_buffer_object *bo,
>>> if (!radeon_ttm_bo_is_radeon_bo(bo)) {
>>> placement->placement = &placements;
>>> - placement->busy_placement = &placements;
>>> placement->num_placement = 1;
>>> - placement->num_busy_placement = 1;
>>> return;
>>> }
>>> rbo = container_of(bo, struct radeon_bo, tbo);
>>> @@ -114,15 +112,11 @@ static void radeon_evict_flags(struct
>>> ttm_buffer_object *bo,
>>> */
>>> radeon_ttm_placement_from_domain(rbo,
>>> RADEON_GEM_DOMAIN_VRAM |
>>> RADEON_GEM_DOMAIN_GTT);
>>> - rbo->placement.num_busy_placement = 0;
>>> for (i = 0; i < rbo->placement.num_placement; i++) {
>>> if (rbo->placements[i].mem_type == TTM_PL_VRAM) {
>>> if (rbo->placements[i].fpfn < fpfn)
>>> rbo->placements[i].fpfn = fpfn;
>>> - } else {
>>> - rbo->placement.busy_placement =
>>> - &rbo->placements[i];
>>> - rbo->placement.num_busy_placement = 1;
>>> + rbo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>>> }
>>> }
>>> } else
>>> diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c
>>> b/drivers/gpu/drm/radeon/radeon_uvd.c
>>> index a2cda184b2b2..058a1c8451b2 100644
>>> --- a/drivers/gpu/drm/radeon/radeon_uvd.c
>>> +++ b/drivers/gpu/drm/radeon/radeon_uvd.c
>>> @@ -324,7 +324,6 @@ void radeon_uvd_force_into_uvd_segment(struct
>>> radeon_bo *rbo,
>>> rbo->placements[1].fpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
>>> rbo->placements[1].lpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
>>> rbo->placement.num_placement++;
>>> - rbo->placement.num_busy_placement++;
>>> }
>>> void radeon_uvd_free_handles(struct radeon_device *rdev, struct
>>> drm_file *filp)
>>> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c
>>> b/drivers/gpu/drm/ttm/ttm_bo.c
>>> index 8c1eaa74fa21..a5e11a92e0b9 100644
>>> --- a/drivers/gpu/drm/ttm/ttm_bo.c
>>> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
>>> @@ -410,8 +410,8 @@ static int ttm_bo_bounce_temp_buffer(struct
>>> ttm_buffer_object *bo,
>>> struct ttm_resource *hop_mem;
>>> int ret;
>>> - hop_placement.num_placement =
>>> hop_placement.num_busy_placement = 1;
>>> - hop_placement.placement = hop_placement.busy_placement = hop;
>>> + hop_placement.num_placement = 1;
>>> + hop_placement.placement = hop;
>>> /* find space in the bounce domain */
>>> ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
>>> @@ -440,10 +440,9 @@ static int ttm_bo_evict(struct
>>> ttm_buffer_object *bo,
>>> dma_resv_assert_held(bo->base.resv);
>>> placement.num_placement = 0;
>>> - placement.num_busy_placement = 0;
>>> bdev->funcs->evict_flags(bo, &placement);
>>> - if (!placement.num_placement && !placement.num_busy_placement) {
>>> + if (!placement.num_placement) {
>>> ret = ttm_bo_wait_ctx(bo, ctx);
>>> if (ret)
>>> return ret;
>>> @@ -791,6 +790,9 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
>>> const struct ttm_place *place = &placement->placement[i];
>>> struct ttm_resource_manager *man;
>>> + if (place->flags & TTM_PL_FLAG_FALLBACK)
>>> + continue;
>>> +
>>> man = ttm_manager_type(bdev, place->mem_type);
>>> if (!man || !ttm_resource_manager_used(man))
>>> continue;
>>> @@ -813,10 +815,13 @@ int ttm_bo_mem_space(struct ttm_buffer_object
>>> *bo,
>>> return 0;
>>> }
>>> - for (i = 0; i < placement->num_busy_placement; ++i) {
>>> - const struct ttm_place *place = &placement->busy_placement[i];
>>> + for (i = 0; i < placement->num_placement; ++i) {
>>> + const struct ttm_place *place = &placement->placement[i];
>>> struct ttm_resource_manager *man;
>>> + if (place->flags & TTM_PL_FLAG_DESIRED)
>>> + continue;
>>> +
>>> man = ttm_manager_type(bdev, place->mem_type);
>>> if (!man || !ttm_resource_manager_used(man))
>>> continue;
>>> @@ -904,11 +909,11 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
>>> /*
>>> * Remove the backing store if no placement is given.
>>> */
>>> - if (!placement->num_placement && !placement->num_busy_placement)
>>> + if (!placement->num_placement)
>>> return ttm_bo_pipeline_gutting(bo);
>>> /* Check whether we need to move buffer. */
>>> - if (bo->resource && ttm_resource_compat(bo->resource, placement))
>>> + if (bo->resource && ttm_resource_compatible(bo->resource,
>>> placement))
>>> return 0;
>>> /* Moving of pinned BOs is forbidden */
>>> diff --git a/drivers/gpu/drm/ttm/ttm_resource.c
>>> b/drivers/gpu/drm/ttm/ttm_resource.c
>>> index 02b96d23fdb9..fb14f7716cf8 100644
>>> --- a/drivers/gpu/drm/ttm/ttm_resource.c
>>> +++ b/drivers/gpu/drm/ttm/ttm_resource.c
>>> @@ -291,37 +291,15 @@ bool ttm_resource_intersects(struct ttm_device
>>> *bdev,
>>> }
>>> /**
>>> - * ttm_resource_compatible - test for compatibility
>>> + * ttm_resource_compatible - check if resource is compatible with
>>> placement
>>> *
>>> - * @bdev: TTM device structure
>>> - * @res: The resource to test
>>> - * @place: The placement to test
>>> - * @size: How many bytes the new allocation needs.
>>> - *
>>> - * Test if @res compatible with @place and @size.
>>> + * @res: the resource to check
>>> + * @placement: the placement to check against
>>> *
>>> - * Returns true if the res placement compatible with @place and @size.
>>> + * Returns true if the placement is compatible.
>>> */
>>> -bool ttm_resource_compatible(struct ttm_device *bdev,
>>> - struct ttm_resource *res,
>>> - const struct ttm_place *place,
>>> - size_t size)
>>> -{
>>> - struct ttm_resource_manager *man;
>>> -
>>> - if (!res || !place)
>>> - return false;
>>> -
>>> - man = ttm_manager_type(bdev, res->mem_type);
>>> - if (!man->func->compatible)
>>> - return true;
>>> -
>>> - return man->func->compatible(man, res, place, size);
>>> -}
>>> -
>>> -static bool ttm_resource_places_compat(struct ttm_resource *res,
>>> - const struct ttm_place *places,
>>> - unsigned num_placement)
>>> +bool ttm_resource_compatible(struct ttm_resource *res,
>>> + struct ttm_placement *placement)
>>> {
>>> struct ttm_buffer_object *bo = res->bo;
>>> struct ttm_device *bdev = bo->bdev;
>>> @@ -330,44 +308,25 @@ static bool ttm_resource_places_compat(struct
>>> ttm_resource *res,
>>> if (res->placement & TTM_PL_FLAG_TEMPORARY)
>>> return false;
>>> - for (i = 0; i < num_placement; i++) {
>>> - const struct ttm_place *heap = &places[i];
>>> + for (i = 0; i < placement->num_placement; i++) {
>>> + const struct ttm_place *place = &placement->placement[i];
>>> + struct ttm_resource_manager *man;
>>> - if (!ttm_resource_compatible(bdev, res, heap,
>>> bo->base.size))
>>> + if (res->mem_type != place->mem_type)
>>> + continue;
>>> +
>>> + man = ttm_manager_type(bdev, res->mem_type);
>>> + if (man->func->compatible &&
>>> + !man->func->compatible(man, res, place, bo->base.size))
>>> continue;
>>> - if ((res->mem_type == heap->mem_type) &&
>>> - (!(heap->flags & TTM_PL_FLAG_CONTIGUOUS) ||
>>> + if ((!(place->flags & TTM_PL_FLAG_CONTIGUOUS) ||
>>> (res->placement & TTM_PL_FLAG_CONTIGUOUS)))
>>> return true;
>>> }
>>> return false;
>>> }
>>> -/**
>>> - * ttm_resource_compat - check if resource is compatible with
>>> placement
>>> - *
>>> - * @res: the resource to check
>>> - * @placement: the placement to check against
>>> - *
>>> - * Returns true if the placement is compatible.
>>> - */
>>> -bool ttm_resource_compat(struct ttm_resource *res,
>>> - struct ttm_placement *placement)
>>> -{
>>> - if (ttm_resource_places_compat(res, placement->placement,
>>> - placement->num_placement))
>>> - return true;
>>> -
>>> - if ((placement->busy_placement != placement->placement ||
>>> - placement->num_busy_placement > placement->num_placement) &&
>>> - ttm_resource_places_compat(res, placement->busy_placement,
>>> - placement->num_busy_placement))
>>> - return true;
>>> -
>>> - return false;
>>> -}
>>> -
>>> void ttm_resource_set_bo(struct ttm_resource *res,
>>> struct ttm_buffer_object *bo)
>>> {
>>> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>> b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>> index 2bfac3aad7b7..bfd41ce3c8f4 100644
>>> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>> @@ -742,9 +742,21 @@ void vmw_bo_move_notify(struct
>>> ttm_buffer_object *bo,
>>> vmw_resource_unbind_list(vbo);
>>> }
>>> +static u32 placement_flags(u32 domain, u32 desired, u32 fallback)
>>> +{
>>> + if (desired & fallback & domain)
>>> + return 0;
>>> +
>>> + if (desired & domain)
>>> + return TTM_PL_FLAG_DESIRED;
>>> +
>>> + return TTM_PL_FLAG_FALLBACK;
>>> +}
>>> +
>>> static u32
>>> -set_placement_list(struct ttm_place *pl, u32 domain)
>>> +set_placement_list(struct ttm_place *pl, u32 desired, u32 fallback)
>>> {
>>> + u32 domain = desired | fallback;
>>> u32 n = 0;
>>> /*
>>> @@ -752,35 +764,40 @@ set_placement_list(struct ttm_place *pl, u32
>>> domain)
>>> */
>>> if (domain & VMW_BO_DOMAIN_MOB) {
>>> pl[n].mem_type = VMW_PL_MOB;
>>> - pl[n].flags = 0;
>>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_MOB, desired,
>>> + fallback);
>>> pl[n].fpfn = 0;
>>> pl[n].lpfn = 0;
>>> n++;
>>> }
>>> if (domain & VMW_BO_DOMAIN_GMR) {
>>> pl[n].mem_type = VMW_PL_GMR;
>>> - pl[n].flags = 0;
>>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_GMR, desired,
>>> + fallback);
>>> pl[n].fpfn = 0;
>>> pl[n].lpfn = 0;
>>> n++;
>>> }
>>> if (domain & VMW_BO_DOMAIN_VRAM) {
>>> pl[n].mem_type = TTM_PL_VRAM;
>>> - pl[n].flags = 0;
>>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_VRAM, desired,
>>> + fallback);
>>> pl[n].fpfn = 0;
>>> pl[n].lpfn = 0;
>>> n++;
>>> }
>>> if (domain & VMW_BO_DOMAIN_WAITABLE_SYS) {
>>> pl[n].mem_type = VMW_PL_SYSTEM;
>>> - pl[n].flags = 0;
>>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_WAITABLE_SYS,
>>> + desired, fallback);
>>> pl[n].fpfn = 0;
>>> pl[n].lpfn = 0;
>>> n++;
>>> }
>>> if (domain & VMW_BO_DOMAIN_SYS) {
>>> pl[n].mem_type = TTM_PL_SYSTEM;
>>> - pl[n].flags = 0;
>>> + pl[n].flags = placement_flags(VMW_BO_DOMAIN_SYS, desired,
>>> + fallback);
>>> pl[n].fpfn = 0;
>>> pl[n].lpfn = 0;
>>> n++;
>>> @@ -806,7 +823,7 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32
>>> domain, u32 busy_domain)
>>> u32 i;
>>> pl->placement = bo->places;
>>> - pl->num_placement = set_placement_list(bo->places, domain);
>>> + pl->num_placement = set_placement_list(bo->places, domain,
>>> busy_domain);
>>> if (drm_debug_enabled(DRM_UT_DRIVER) && bo->tbo.resource) {
>>> for (i = 0; i < pl->num_placement; ++i) {
>>> @@ -821,8 +838,6 @@ void vmw_bo_placement_set(struct vmw_bo *bo, u32
>>> domain, u32 busy_domain)
>>> __func__, bo->tbo.resource->mem_type, domain);
>>> }
>>> - pl->busy_placement = bo->busy_places;
>>> - pl->num_busy_placement = set_placement_list(bo->busy_places,
>>> busy_domain);
>>> }
>>> void vmw_bo_placement_set_default_accelerated(struct vmw_bo *bo)
>>> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>> b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>> index a84fffcef8e1..4d23d0a70bcb 100644
>>> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>> @@ -46,15 +46,11 @@ static const struct ttm_place
>>> sys_placement_flags = {
>>> struct ttm_placement vmw_vram_placement = {
>>> .num_placement = 1,
>>> .placement = &vram_placement_flags,
>>> - .num_busy_placement = 1,
>>> - .busy_placement = &vram_placement_flags
>>> };
>>> struct ttm_placement vmw_sys_placement = {
>>> .num_placement = 1,
>>> .placement = &sys_placement_flags,
>>> - .num_busy_placement = 1,
>>> - .busy_placement = &sys_placement_flags
>>> };
>>> const size_t vmw_tt_size = sizeof(struct vmw_ttm_tt);
>>> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
>>> index 8e4a3b1f6b93..1e5edc7f0b85 100644
>>> --- a/drivers/gpu/drm/xe/xe_bo.c
>>> +++ b/drivers/gpu/drm/xe/xe_bo.c
>>> @@ -38,22 +38,26 @@ static const struct ttm_place
>>> sys_placement_flags = {
>>> static struct ttm_placement sys_placement = {
>>> .num_placement = 1,
>>> .placement = &sys_placement_flags,
>>> - .num_busy_placement = 1,
>>> - .busy_placement = &sys_placement_flags,
>>> };
>>> -static const struct ttm_place tt_placement_flags = {
>>> - .fpfn = 0,
>>> - .lpfn = 0,
>>> - .mem_type = XE_PL_TT,
>>> - .flags = 0,
>>> +static const struct ttm_place tt_placement_flags[] = {
>>> + {
>>> + .fpfn = 0,
>>> + .lpfn = 0,
>>> + .mem_type = XE_PL_TT,
>>> + .flags = TTM_PL_FLAG_DESIRED,
>>> + },
>>> + {
>>> + .fpfn = 0,
>>> + .lpfn = 0,
>>> + .mem_type = XE_PL_SYSTEM,
>>> + .flags = TTM_PL_FLAG_FALLBACK,
>>> + }
>>> };
>>> static struct ttm_placement tt_placement = {
>>> - .num_placement = 1,
>>> - .placement = &tt_placement_flags,
>>> - .num_busy_placement = 1,
>>> - .busy_placement = &sys_placement_flags,
>>> + .num_placement = 2,
>>> + .placement = tt_placement_flags,
>>> };
>>> bool mem_type_is_vram(u32 mem_type)
>>> @@ -230,8 +234,6 @@ static int __xe_bo_placement_for_flags(struct
>>> xe_device *xe, struct xe_bo *bo,
>>> bo->placement = (struct ttm_placement) {
>>> .num_placement = c,
>>> .placement = bo->placements,
>>> - .num_busy_placement = c,
>>> - .busy_placement = bo->placements,
>>> };
>>> return 0;
>>> @@ -251,7 +253,6 @@ static void xe_evict_flags(struct
>>> ttm_buffer_object *tbo,
>>> /* Don't handle scatter gather BOs */
>>> if (tbo->type == ttm_bo_type_sg) {
>>> placement->num_placement = 0;
>>> - placement->num_busy_placement = 0;
>>> return;
>>> }
>>> @@ -1353,8 +1354,6 @@ static int __xe_bo_fixed_placement(struct
>>> xe_device *xe,
>>> bo->placement = (struct ttm_placement) {
>>> .num_placement = 1,
>>> .placement = place,
>>> - .num_busy_placement = 1,
>>> - .busy_placement = place,
>>> };
>>> return 0;
>>> @@ -2112,9 +2111,7 @@ int xe_bo_migrate(struct xe_bo *bo, u32 mem_type)
>>> xe_place_from_ttm_type(mem_type, &requested);
>>> placement.num_placement = 1;
>>> - placement.num_busy_placement = 1;
>>> placement.placement = &requested;
>>> - placement.busy_placement = &requested;
>>> /*
>>> * Stolen needs to be handled like below VRAM handling if we
>>> ever need
>>> diff --git a/include/drm/ttm/ttm_placement.h
>>> b/include/drm/ttm/ttm_placement.h
>>> index 8074d0f6cae5..b510a4812609 100644
>>> --- a/include/drm/ttm/ttm_placement.h
>>> +++ b/include/drm/ttm/ttm_placement.h
>>> @@ -64,6 +64,12 @@
>>> /* For multihop handling */
>>> #define TTM_PL_FLAG_TEMPORARY (1 << 2)
>>> +/* Placement is never used during eviction */
>>> +#define TTM_PL_FLAG_DESIRED (1 << 3)
>>> +
>>> +/* Placement is only used during eviction */
>>> +#define TTM_PL_FLAG_FALLBACK (1 << 4)
>>> +
>>> /**
>>> * struct ttm_place
>>> *
>>> @@ -86,16 +92,12 @@ struct ttm_place {
>>> *
>>> * @num_placement: number of preferred placements
>>> * @placement: preferred placements
>>> - * @num_busy_placement: number of preferred placements when need
>>> to evict buffer
>>> - * @busy_placement: preferred placements when need to evict buffer
>>> *
>>> * Structure indicating the placement you request for an object.
>>> */
>>> struct ttm_placement {
>>> unsigned num_placement;
>>> const struct ttm_place *placement;
>>> - unsigned num_busy_placement;
>>> - const struct ttm_place *busy_placement;
>>> };
>>> #endif
>>> diff --git a/include/drm/ttm/ttm_resource.h
>>> b/include/drm/ttm/ttm_resource.h
>>> index 78a226eba953..1afa13f0c22b 100644
>>> --- a/include/drm/ttm/ttm_resource.h
>>> +++ b/include/drm/ttm/ttm_resource.h
>>> @@ -365,12 +365,8 @@ bool ttm_resource_intersects(struct ttm_device
>>> *bdev,
>>> struct ttm_resource *res,
>>> const struct ttm_place *place,
>>> size_t size);
>>> -bool ttm_resource_compatible(struct ttm_device *bdev,
>>> - struct ttm_resource *res,
>>> - const struct ttm_place *place,
>>> - size_t size);
>>> -bool ttm_resource_compat(struct ttm_resource *res,
>>> - struct ttm_placement *placement);
>>> +bool ttm_resource_compatible(struct ttm_resource *res,
>>> + struct ttm_placement *placement);
>>> void ttm_resource_set_bo(struct ttm_resource *res,
>>> struct ttm_buffer_object *bo);
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 4/5] drm/ttm: improve idle/busy handling v3
2024-01-18 14:24 ` Thomas Hellström
@ 2024-01-18 16:48 ` Thomas Hellström
0 siblings, 0 replies; 19+ messages in thread
From: Thomas Hellström @ 2024-01-18 16:48 UTC (permalink / raw)
To: Christian König, dri-devel, intel-gfx, nouveau, jani.nikula,
kherbst, lyude, zackr, michel.daenzer
On 1/18/24 15:24, Thomas Hellström wrote:
> On Fri, 2024-01-12 at 13:51 +0100, Christian König wrote:
>> Previously we would never try to move a BO into the preferred
>> placements
>> when it ever landed in a busy placement since those were considered
>> compatible.
>>
>> Rework the whole handling and finally unify the idle and busy
>> handling.
>> ttm_bo_validate() is now responsible to try idle placement first and
>> then
>> use the busy placement if that didn't worked.
>>
>> Drawback is that we now always try the idle placement first for each
>> validation which might cause some additional CPU overhead on
>> overcommit.
>
>
>
>
>> v2: fix kerneldoc warning and coding style
>> v3: take care of XE as well
>>
>> Signed-off-by: Christian König <christian.koenig@amd.com>
>> ---
>> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 2 +-
>> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 2 +-
>> drivers/gpu/drm/ttm/ttm_bo.c | 131 ++++++++-----------
>> --
>> drivers/gpu/drm/ttm/ttm_resource.c | 16 ++-
>> drivers/gpu/drm/xe/xe_bo.c | 4 +-
>> include/drm/ttm/ttm_bo.h | 3 +-
>> include/drm/ttm/ttm_resource.h | 3 +-
>> 7 files changed, 68 insertions(+), 93 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> index b671b0665492..06fb3fc47eaa 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> @@ -404,7 +404,7 @@ int amdgpu_bo_create_kernel_at(struct
>> amdgpu_device *adev,
>> (*bo_ptr)->placements[i].lpfn = (offset + size) >>
>> PAGE_SHIFT;
>> }
>> r = ttm_bo_mem_space(&(*bo_ptr)->tbo, &(*bo_ptr)->placement,
>> - &(*bo_ptr)->tbo.resource, &ctx);
>> + &(*bo_ptr)->tbo.resource, &ctx, false);
>> if (r)
>> goto error;
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> index 8722beba494e..f23cdc7c5b08 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>> @@ -966,7 +966,7 @@ int amdgpu_ttm_alloc_gart(struct
>> ttm_buffer_object *bo)
>> placements.mem_type = TTM_PL_TT;
>> placements.flags = bo->resource->placement;
>>
>> - r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx);
>> + r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx, true);
>> if (unlikely(r))
>> return r;
>>
>> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c
>> b/drivers/gpu/drm/ttm/ttm_bo.c
>> index a5e11a92e0b9..3783be24d832 100644
>> --- a/drivers/gpu/drm/ttm/ttm_bo.c
>> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
>> @@ -414,7 +414,7 @@ static int ttm_bo_bounce_temp_buffer(struct
>> ttm_buffer_object *bo,
>> hop_placement.placement = hop;
>>
>> /* find space in the bounce domain */
>> - ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
>> + ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx,
>> true);
>> if (ret)
>> return ret;
>> /* move to the bounce domain */
>> @@ -454,7 +454,7 @@ static int ttm_bo_evict(struct ttm_buffer_object
>> *bo,
>> return ttm_bo_pipeline_gutting(bo);
>> }
>>
>> - ret = ttm_bo_mem_space(bo, &placement, &evict_mem, ctx);
>> + ret = ttm_bo_mem_space(bo, &placement, &evict_mem, ctx,
>> true);
> This is what breaks xe's selftest since the evict-flags preferred
> placement is never tried so it changes the behavior.
>
> The xe evict flags were set up to "Try to evict to TT first, but if
> that causes recursive eviction, try to evict to system". This here
> ignores the preferred eviction placement.
>
> So "Preferred" in effect changes meaning from "Don't try this if it
> will cause an eviction", to "Don't try this in the evict path", which
> is hard for the driver to have any knowledge about.
>
> Then also it sounds from the commit message with this patch "Preferred"
> also gets overloaded with "Always retry preferred on validate", but
> shouldn't that really be a dynamic driver decision and not something
> TTM should try to enforce in a static way? Drivers could have short-
> circuited the ttm_bo_validate() call if it succeeded once, and have a
> deeper thought about when to migrate from, say TT to VRAM and vice
> versa.
>
> For the specific behaviour sought here, there is (or at least used to
> be) a construct in the vmwgfx driver that first called
> ttm_bo_validate() with VRAM as preferred placement and no fallback. If
> that failed due to VRAM being full, it called ttm_bo_validate() again,
> this time with fallback and VRAM allowing eviction.
>
> /Thomas
>
To conclude here, a suggestion would be,
1) Could we separate the semantics of the "Preferred" flag. Perhaps
creating yet another one.
first flag: "Retry this placement even if bo is in a valid placement"
second flag: "Don't use this placement if it would cause an eviction"
(Retain old semantics)
Although I think the first flag behaviour can be easily implemented with
a driver specific xxx_bo_validate(), I'm not against having that in TTM
as well.
2) Please keep the old behaviour of the eviction flags? Meaning even for
eviction we have preferred and fallback placements. It comes in handy
when pipelining (always evicting to SYSTEM requires syncing), and if
opportunistically trying to evict to other VRAM memory types on
multi-tile. I see no easy way to otherwise hook into TTM to implement
that if the behaviour is changed as suggested.
Thanks,
Thomas
>
>
>> if (ret) {
>> if (ret != -ERESTARTSYS) {
>> pr_err("Failed to find memory space for
>> buffer 0x%p eviction\n",
>> @@ -724,37 +724,6 @@ static int ttm_bo_add_move_fence(struct
>> ttm_buffer_object *bo,
>> return ret;
>> }
>>
>> -/*
>> - * Repeatedly evict memory from the LRU for @mem_type until we
>> create enough
>> - * space, or we've evicted everything and there isn't enough space.
>> - */
>> -static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo,
>> - const struct ttm_place *place,
>> - struct ttm_resource **mem,
>> - struct ttm_operation_ctx *ctx)
>> -{
>> - struct ttm_device *bdev = bo->bdev;
>> - struct ttm_resource_manager *man;
>> - struct ww_acquire_ctx *ticket;
>> - int ret;
>> -
>> - man = ttm_manager_type(bdev, place->mem_type);
>> - ticket = dma_resv_locking_ctx(bo->base.resv);
>> - do {
>> - ret = ttm_resource_alloc(bo, place, mem);
>> - if (likely(!ret))
>> - break;
>> - if (unlikely(ret != -ENOSPC))
>> - return ret;
>> - ret = ttm_mem_evict_first(bdev, man, place, ctx,
>> - ticket);
>> - if (unlikely(ret != 0))
>> - return ret;
>> - } while (1);
>> -
>> - return ttm_bo_add_move_fence(bo, man, *mem, ctx-
>>> no_wait_gpu);
>> -}
>> -
>> /**
>> * ttm_bo_mem_space
>> *
>> @@ -763,6 +732,7 @@ static int ttm_bo_mem_force_space(struct
>> ttm_buffer_object *bo,
>> * @placement: Proposed new placement for the buffer object.
>> * @mem: A struct ttm_resource.
>> * @ctx: if and how to sleep, lock buffers and alloc memory
>> + * @force_space: If we should evict buffers to force space
>> *
>> * Allocate memory space for the buffer object pointed to by @bo,
>> using
>> * the placement flags in @placement, potentially evicting other
>> idle buffer objects.
>> @@ -776,12 +746,14 @@ static int ttm_bo_mem_force_space(struct
>> ttm_buffer_object *bo,
>> int ttm_bo_mem_space(struct ttm_buffer_object *bo,
>> struct ttm_placement *placement,
>> struct ttm_resource **mem,
>> - struct ttm_operation_ctx *ctx)
>> + struct ttm_operation_ctx *ctx,
>> + bool force_space)
>> {
>> struct ttm_device *bdev = bo->bdev;
>> - bool type_found = false;
>> + struct ww_acquire_ctx *ticket;
>> int i, ret;
>>
>> + ticket = dma_resv_locking_ctx(bo->base.resv);
>> ret = dma_resv_reserve_fences(bo->base.resv, 1);
>> if (unlikely(ret))
>> return ret;
>> @@ -790,19 +762,30 @@ int ttm_bo_mem_space(struct ttm_buffer_object
>> *bo,
>> const struct ttm_place *place = &placement-
>>> placement[i];
>> struct ttm_resource_manager *man;
>>
>> - if (place->flags & TTM_PL_FLAG_FALLBACK)
>> - continue;
>> -
>> man = ttm_manager_type(bdev, place->mem_type);
>> if (!man || !ttm_resource_manager_used(man))
>> continue;
>>
>> - type_found = true;
>> - ret = ttm_resource_alloc(bo, place, mem);
>> - if (ret == -ENOSPC)
>> + if (place->flags & (force_space ?
>> TTM_PL_FLAG_DESIRED :
>> + TTM_PL_FLAG_FALLBACK))
>> + continue;
>> +
>> + do {
>> + ret = ttm_resource_alloc(bo, place, mem);
>> + if (unlikely(ret != -ENOSPC))
>> + return ret;
>> + if (likely(!ret) || !force_space)
>> + break;
>> +
>> + ret = ttm_mem_evict_first(bdev, man, place,
>> ctx,
>> + ticket);
>> + if (unlikely(ret == -EBUSY))
>> + break;
>> + if (unlikely(ret))
>> + return ret;
>> + } while (1);
>> + if (ret)
>> continue;
>> - if (unlikely(ret))
>> - goto error;
>>
>> ret = ttm_bo_add_move_fence(bo, man, *mem, ctx-
>>> no_wait_gpu);
>> if (unlikely(ret)) {
>> @@ -810,45 +793,19 @@ int ttm_bo_mem_space(struct ttm_buffer_object
>> *bo,
>> if (ret == -EBUSY)
>> continue;
>>
>> - goto error;
>> + return ret;
>> }
>> return 0;
>> }
>>
>> - for (i = 0; i < placement->num_placement; ++i) {
>> - const struct ttm_place *place = &placement-
>>> placement[i];
>> - struct ttm_resource_manager *man;
>> -
>> - if (place->flags & TTM_PL_FLAG_DESIRED)
>> - continue;
>> -
>> - man = ttm_manager_type(bdev, place->mem_type);
>> - if (!man || !ttm_resource_manager_used(man))
>> - continue;
>> -
>> - type_found = true;
>> - ret = ttm_bo_mem_force_space(bo, place, mem, ctx);
>> - if (likely(!ret))
>> - return 0;
>> -
>> - if (ret && ret != -EBUSY)
>> - goto error;
>> - }
>> -
>> - ret = -ENOSPC;
>> - if (!type_found) {
>> - pr_err(TTM_PFX "No compatible memory type found\n");
>> - ret = -EINVAL;
>> - }
>> -
>> -error:
>> - return ret;
>> + return -ENOSPC;
>> }
>> EXPORT_SYMBOL(ttm_bo_mem_space);
>>
>> static int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
>> struct ttm_placement *placement,
>> - struct ttm_operation_ctx *ctx)
>> + struct ttm_operation_ctx *ctx,
>> + bool force_space)
>> {
>> struct ttm_resource *mem;
>> struct ttm_place hop;
>> @@ -865,7 +822,7 @@ static int ttm_bo_move_buffer(struct
>> ttm_buffer_object *bo,
>> * stop and the driver will be called to make
>> * the second hop.
>> */
>> - ret = ttm_bo_mem_space(bo, placement, &mem, ctx);
>> + ret = ttm_bo_mem_space(bo, placement, &mem, ctx,
>> force_space);
>> if (ret)
>> return ret;
>> bounce:
>> @@ -902,6 +859,7 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
>> struct ttm_placement *placement,
>> struct ttm_operation_ctx *ctx)
>> {
>> + bool force_space;
>> int ret;
>>
>> dma_resv_assert_held(bo->base.resv);
>> @@ -912,20 +870,27 @@ int ttm_bo_validate(struct ttm_buffer_object
>> *bo,
>> if (!placement->num_placement)
>> return ttm_bo_pipeline_gutting(bo);
>>
>> - /* Check whether we need to move buffer. */
>> - if (bo->resource && ttm_resource_compatible(bo->resource,
>> placement))
>> - return 0;
>> + force_space = false;
>> + do {
>> + /* Check whether we need to move buffer. */
>> + if (bo->resource &&
>> + ttm_resource_compatible(bo->resource, placement,
>> + force_space))
>> + return 0;
>> +
>> + /* Moving of pinned BOs is forbidden */
>> + if (bo->pin_count)
>> + return -EINVAL;
>>
>> - /* Moving of pinned BOs is forbidden */
>> - if (bo->pin_count)
>> - return -EINVAL;
>> + ret = ttm_bo_move_buffer(bo, placement, ctx,
>> force_space);
>> + if (ret && ret != -ENOSPC)
>> + return ret;
>>
>> - ret = ttm_bo_move_buffer(bo, placement, ctx);
>> + force_space = !force_space;
>> + } while (force_space);
>> /* For backward compatibility with userspace */
>> if (ret == -ENOSPC)
>> return -ENOMEM;
>> - if (ret)
>> - return ret;
>>
>> /*
>> * We might need to add a TTM.
>> diff --git a/drivers/gpu/drm/ttm/ttm_resource.c
>> b/drivers/gpu/drm/ttm/ttm_resource.c
>> index fb14f7716cf8..65155f2013ca 100644
>> --- a/drivers/gpu/drm/ttm/ttm_resource.c
>> +++ b/drivers/gpu/drm/ttm/ttm_resource.c
>> @@ -295,11 +295,13 @@ bool ttm_resource_intersects(struct ttm_device
>> *bdev,
>> *
>> * @res: the resource to check
>> * @placement: the placement to check against
>> + * @evicting: true if the caller is doing evictions
>> *
>> * Returns true if the placement is compatible.
>> */
>> bool ttm_resource_compatible(struct ttm_resource *res,
>> - struct ttm_placement *placement)
>> + struct ttm_placement *placement,
>> + bool evicting)
>> {
>> struct ttm_buffer_object *bo = res->bo;
>> struct ttm_device *bdev = bo->bdev;
>> @@ -315,14 +317,20 @@ bool ttm_resource_compatible(struct
>> ttm_resource *res,
>> if (res->mem_type != place->mem_type)
>> continue;
>>
>> + if (place->flags & (evicting ? TTM_PL_FLAG_DESIRED :
>> + TTM_PL_FLAG_FALLBACK))
>> + continue;
>> +
>> + if (place->flags & TTM_PL_FLAG_CONTIGUOUS &&
>> + !(res->placement & TTM_PL_FLAG_CONTIGUOUS))
>> + continue;
>> +
>> man = ttm_manager_type(bdev, res->mem_type);
>> if (man->func->compatible &&
>> !man->func->compatible(man, res, place, bo-
>>> base.size))
>> continue;
>>
>> - if ((!(place->flags & TTM_PL_FLAG_CONTIGUOUS) ||
>> - (res->placement & TTM_PL_FLAG_CONTIGUOUS)))
>> - return true;
>> + return true;
>> }
>> return false;
>> }
>> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
>> index 1e5edc7f0b85..d4b8bd774ada 100644
>> --- a/drivers/gpu/drm/xe/xe_bo.c
>> +++ b/drivers/gpu/drm/xe/xe_bo.c
>> @@ -828,7 +828,7 @@ int xe_bo_evict_pinned(struct xe_bo *bo)
>> if (WARN_ON(!xe_bo_is_vram(bo)))
>> return -EINVAL;
>>
>> - ret = ttm_bo_mem_space(&bo->ttm, &placement, &new_mem,
>> &ctx);
>> + ret = ttm_bo_mem_space(&bo->ttm, &placement, &new_mem, &ctx,
>> false);
>> if (ret)
>> return ret;
>>
>> @@ -893,7 +893,7 @@ int xe_bo_restore_pinned(struct xe_bo *bo)
>> if (WARN_ON(xe_bo_is_vram(bo) || !bo->ttm.ttm))
>> return -EINVAL;
>>
>> - ret = ttm_bo_mem_space(&bo->ttm, &bo->placement, &new_mem,
>> &ctx);
>> + ret = ttm_bo_mem_space(&bo->ttm, &bo->placement, &new_mem,
>> &ctx, false);
>> if (ret)
>> return ret;
>>
>> diff --git a/include/drm/ttm/ttm_bo.h b/include/drm/ttm/ttm_bo.h
>> index 0223a41a64b2..64d738f152d1 100644
>> --- a/include/drm/ttm/ttm_bo.h
>> +++ b/include/drm/ttm/ttm_bo.h
>> @@ -397,7 +397,8 @@ vm_fault_t ttm_bo_vm_dummy_page(struct vm_fault
>> *vmf, pgprot_t prot);
>> int ttm_bo_mem_space(struct ttm_buffer_object *bo,
>> struct ttm_placement *placement,
>> struct ttm_resource **mem,
>> - struct ttm_operation_ctx *ctx);
>> + struct ttm_operation_ctx *ctx,
>> + bool force_space);
>>
>> void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo);
>> /*
>> diff --git a/include/drm/ttm/ttm_resource.h
>> b/include/drm/ttm/ttm_resource.h
>> index 1afa13f0c22b..7561023db43d 100644
>> --- a/include/drm/ttm/ttm_resource.h
>> +++ b/include/drm/ttm/ttm_resource.h
>> @@ -366,7 +366,8 @@ bool ttm_resource_intersects(struct ttm_device
>> *bdev,
>> const struct ttm_place *place,
>> size_t size);
>> bool ttm_resource_compatible(struct ttm_resource *res,
>> - struct ttm_placement *placement);
>> + struct ttm_placement *placement,
>> + bool evicting);
>> void ttm_resource_set_bo(struct ttm_resource *res,
>> struct ttm_buffer_object *bo);
>>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: Re: [PATCH 3/5] drm/ttm: replace busy placement with flags v6
2024-01-18 16:38 ` Thomas Hellström
@ 2024-01-26 22:16 ` Lucas De Marchi
2024-01-26 22:22 ` Lucas De Marchi
0 siblings, 1 reply; 19+ messages in thread
From: Lucas De Marchi @ 2024-01-26 22:16 UTC (permalink / raw)
To: Thomas Hellström
Cc: kherbst, michel.daenzer, Christian König, intel-gfx,
dri-devel, nouveau, zackr
On Thu, Jan 18, 2024 at 05:38:16PM +0100, Thomas Hellström wrote:
>
>On 1/17/24 13:27, Thomas Hellström wrote:
>>
>>On 1/17/24 11:47, Thomas Hellström wrote:
>>>Hi, Christian
>>>
>>>Xe changes look good. Will send the series to xe ci to check for
>>>regressions.
>>
>>Hmm, there are some checkpatch warnings about author / SOB email
>>mismatch,
>
>With those fixed, this patch is
>
>Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
it actually broke drm-tip now that this is merged:
../drivers/gpu/drm/xe/xe_bo.c:41:10: error: ‘struct ttm_placement’ has no member named ‘num_busy_placement’; did you mean ‘num_placement’
41 | .num_busy_placement = 1,
| ^~~~~~~~~~~~~~~~~~
| num_placement
../drivers/gpu/drm/xe/xe_bo.c:41:31: error: excess elements in struct initializer [-Werror]
41 | .num_busy_placement = 1,
| ^
Apparently a conflict with another patch that got applied a few days
ago: a201c6ee37d6 ("drm/xe/bo: Evict VRAM to TT rather than to system")
Lucas De Marchi
>
>
>>
>>But worserthere are some regressions in the dma-buf ktest (it tests
>>evicting of a dynamic dma-buf),
>>
>>https://patchwork.freedesktop.org/series/128873/
>>
>>I'll take a look later today or tomorrow.
>
>These are from the next patch. Will continue the discussion there.
>
>/Thomas
>
>
>>
>>/Thomas
>>
>>
>>
>>>
>>>/Thomas
>>>
>>>
>>>On 1/12/24 13:51, Christian König wrote:
>>>>From: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
>>>>
>>>>Instead of a list of separate busy placement add flags which indicate
>>>>that a placement should only be used when there is room or if we
>>>>need to
>>>>evict.
>>>>
>>>>v2: add missing TTM_PL_FLAG_IDLE for i915
>>>>v3: fix auto build test ERROR on drm-tip/drm-tip
>>>>v4: fix some typos pointed out by checkpatch
>>>>v5: cleanup some rebase problems with VMWGFX
>>>>v6: implement some missing VMWGFX functionality pointed out by Zack,
>>>> rename the flags as suggested by Michel, rebase on drm-tip and
>>>> adjust XE as well
>>>>
>>>>Signed-off-by: Christian König <christian.koenig@amd.com>
>>>>Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
>>>>---
>>>> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 6 +-
>>>> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 +---
>>>> drivers/gpu/drm/drm_gem_vram_helper.c | 2 -
>>>> drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 37 +++++------
>>>> drivers/gpu/drm/loongson/lsdc_ttm.c | 2 -
>>>> drivers/gpu/drm/nouveau/nouveau_bo.c | 59 +++++++----------
>>>> drivers/gpu/drm/nouveau/nouveau_bo.h | 1 -
>>>> drivers/gpu/drm/qxl/qxl_object.c | 2 -
>>>> drivers/gpu/drm/qxl/qxl_ttm.c | 2 -
>>>> drivers/gpu/drm/radeon/radeon_object.c | 2 -
>>>> drivers/gpu/drm/radeon/radeon_ttm.c | 8 +--
>>>> drivers/gpu/drm/radeon/radeon_uvd.c | 1 -
>>>> drivers/gpu/drm/ttm/ttm_bo.c | 21 ++++---
>>>> drivers/gpu/drm/ttm/ttm_resource.c | 73
>>>>+++++-----------------
>>>> drivers/gpu/drm/vmwgfx/vmwgfx_bo.c | 33 +++++++---
>>>> drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 4 --
>>>> drivers/gpu/drm/xe/xe_bo.c | 33 +++++-----
>>>> include/drm/ttm/ttm_placement.h | 10 +--
>>>> include/drm/ttm/ttm_resource.h | 8 +--
>>>> 19 files changed, 118 insertions(+), 197 deletions(-)
>>>>
>>>>diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>>>b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>>>index 425cebcc5cbf..b671b0665492 100644
>>>>--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>>>+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>>>@@ -220,9 +220,6 @@ void amdgpu_bo_placement_from_domain(struct
>>>>amdgpu_bo *abo, u32 domain)
>>>> placement->num_placement = c;
>>>> placement->placement = places;
>>>>-
>>>>- placement->num_busy_placement = c;
>>>>- placement->busy_placement = places;
>>>> }
>>>> /**
>>>>@@ -1397,8 +1394,7 @@ vm_fault_t
>>>>amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
>>>> AMDGPU_GEM_DOMAIN_GTT);
>>>> /* Avoid costly evictions; only set GTT as a busy placement */
>>>>- abo->placement.num_busy_placement = 1;
>>>>- abo->placement.busy_placement = &abo->placements[1];
>>>>+ abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>>>> r = ttm_bo_validate(bo, &abo->placement, &ctx);
>>>> if (unlikely(r == -EBUSY || r == -ERESTARTSYS))
>>>>diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>>>b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>>>index 75c9fd2c6c2a..8722beba494e 100644
>>>>--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>>>+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>>>@@ -102,23 +102,19 @@ static void amdgpu_evict_flags(struct
>>>>ttm_buffer_object *bo,
>>>> /* Don't handle scatter gather BOs */
>>>> if (bo->type == ttm_bo_type_sg) {
>>>> placement->num_placement = 0;
>>>>- placement->num_busy_placement = 0;
>>>> return;
>>>> }
>>>> /* Object isn't an AMDGPU object so ignore */
>>>> if (!amdgpu_bo_is_amdgpu_bo(bo)) {
>>>> placement->placement = &placements;
>>>>- placement->busy_placement = &placements;
>>>> placement->num_placement = 1;
>>>>- placement->num_busy_placement = 1;
>>>> return;
>>>> }
>>>> abo = ttm_to_amdgpu_bo(bo);
>>>> if (abo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) {
>>>> placement->num_placement = 0;
>>>>- placement->num_busy_placement = 0;
>>>> return;
>>>> }
>>>> @@ -128,13 +124,13 @@ static void amdgpu_evict_flags(struct
>>>>ttm_buffer_object *bo,
>>>> case AMDGPU_PL_OA:
>>>> case AMDGPU_PL_DOORBELL:
>>>> placement->num_placement = 0;
>>>>- placement->num_busy_placement = 0;
>>>> return;
>>>> case TTM_PL_VRAM:
>>>> if (!adev->mman.buffer_funcs_enabled) {
>>>> /* Move to system memory */
>>>> amdgpu_bo_placement_from_domain(abo,
>>>>AMDGPU_GEM_DOMAIN_CPU);
>>>>+
>>>> } else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
>>>> !(abo->flags &
>>>>AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) &&
>>>> amdgpu_bo_in_cpu_visible_vram(abo)) {
>>>>@@ -149,8 +145,7 @@ static void amdgpu_evict_flags(struct
>>>>ttm_buffer_object *bo,
>>>> AMDGPU_GEM_DOMAIN_CPU);
>>>> abo->placements[0].fpfn =
>>>>adev->gmc.visible_vram_size >> PAGE_SHIFT;
>>>> abo->placements[0].lpfn = 0;
>>>>- abo->placement.busy_placement = &abo->placements[1];
>>>>- abo->placement.num_busy_placement = 1;
>>>>+ abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>>>> } else {
>>>> /* Move to GTT memory */
>>>> amdgpu_bo_placement_from_domain(abo,
>>>>AMDGPU_GEM_DOMAIN_GTT |
>>>>@@ -966,8 +961,6 @@ int amdgpu_ttm_alloc_gart(struct
>>>>ttm_buffer_object *bo)
>>>> /* allocate GART space */
>>>> placement.num_placement = 1;
>>>> placement.placement = &placements;
>>>>- placement.num_busy_placement = 1;
>>>>- placement.busy_placement = &placements;
>>>> placements.fpfn = 0;
>>>> placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
>>>> placements.mem_type = TTM_PL_TT;
>>>>diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c
>>>>b/drivers/gpu/drm/drm_gem_vram_helper.c
>>>>index b67eafa55715..75f2eaf0d5b6 100644
>>>>--- a/drivers/gpu/drm/drm_gem_vram_helper.c
>>>>+++ b/drivers/gpu/drm/drm_gem_vram_helper.c
>>>>@@ -147,7 +147,6 @@ static void drm_gem_vram_placement(struct
>>>>drm_gem_vram_object *gbo,
>>>> invariant_flags = TTM_PL_FLAG_TOPDOWN;
>>>> gbo->placement.placement = gbo->placements;
>>>>- gbo->placement.busy_placement = gbo->placements;
>>>> if (pl_flag & DRM_GEM_VRAM_PL_FLAG_VRAM) {
>>>> gbo->placements[c].mem_type = TTM_PL_VRAM;
>>>>@@ -160,7 +159,6 @@ static void drm_gem_vram_placement(struct
>>>>drm_gem_vram_object *gbo,
>>>> }
>>>> gbo->placement.num_placement = c;
>>>>- gbo->placement.num_busy_placement = c;
>>>> for (i = 0; i < c; ++i) {
>>>> gbo->placements[i].fpfn = 0;
>>>>diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>>>>b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>>>>index 6b69ef0cdbb4..80c6cafc8887 100644
>>>>--- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>>>>+++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
>>>>@@ -65,8 +65,6 @@ static const struct ttm_place sys_placement_flags = {
>>>> static struct ttm_placement i915_sys_placement = {
>>>> .num_placement = 1,
>>>> .placement = &sys_placement_flags,
>>>>- .num_busy_placement = 1,
>>>>- .busy_placement = &sys_placement_flags,
>>>> };
>>>> /**
>>>>@@ -157,32 +155,28 @@ i915_ttm_place_from_region(const struct
>>>>intel_memory_region *mr,
>>>> static void
>>>> i915_ttm_placement_from_obj(const struct drm_i915_gem_object *obj,
>>>>- struct ttm_place *requested,
>>>>- struct ttm_place *busy,
>>>>+ struct ttm_place *places,
>>>> struct ttm_placement *placement)
>>>> {
>>>> unsigned int num_allowed = obj->mm.n_placements;
>>>> unsigned int flags = obj->flags;
>>>> unsigned int i;
>>>> - placement->num_placement = 1;
>>>>+ places[0].flags |= TTM_PL_FLAG_DESIRED;
>>>> i915_ttm_place_from_region(num_allowed ? obj->mm.placements[0] :
>>>>- obj->mm.region, requested, obj->bo_offset,
>>>>+ obj->mm.region, &places[0], obj->bo_offset,
>>>> obj->base.size, flags);
>>>> /* Cache this on object? */
>>>>- placement->num_busy_placement = num_allowed;
>>>>- for (i = 0; i < placement->num_busy_placement; ++i)
>>>>- i915_ttm_place_from_region(obj->mm.placements[i], busy + i,
>>>>- obj->bo_offset, obj->base.size, flags);
>>>>-
>>>>- if (num_allowed == 0) {
>>>>- *busy = *requested;
>>>>- placement->num_busy_placement = 1;
>>>>+ for (i = 0; i < num_allowed; ++i) {
>>>>+ i915_ttm_place_from_region(obj->mm.placements[i],
>>>>+ &places[i + 1], obj->bo_offset,
>>>>+ obj->base.size, flags);
>>>>+ places[i + 1].flags |= TTM_PL_FLAG_FALLBACK;
>>>> }
>>>> - placement->placement = requested;
>>>>- placement->busy_placement = busy;
>>>>+ placement->num_placement = num_allowed + 1;
>>>>+ placement->placement = places;
>>>> }
>>>> static int i915_ttm_tt_shmem_populate(struct ttm_device *bdev,
>>>>@@ -789,7 +783,8 @@ static int __i915_ttm_get_pages(struct
>>>>drm_i915_gem_object *obj,
>>>> int ret;
>>>> /* First try only the requested placement. No eviction. */
>>>>- real_num_busy = fetch_and_zero(&placement->num_busy_placement);
>>>>+ real_num_busy = placement->num_placement;
>>>>+ placement->num_placement = 1;
>>>> ret = ttm_bo_validate(bo, placement, &ctx);
>>>> if (ret) {
>>>> ret = i915_ttm_err_to_gem(ret);
>>>>@@ -805,7 +800,7 @@ static int __i915_ttm_get_pages(struct
>>>>drm_i915_gem_object *obj,
>>>> * If the initial attempt fails, allow all accepted
>>>>placements,
>>>> * evicting if necessary.
>>>> */
>>>>- placement->num_busy_placement = real_num_busy;
>>>>+ placement->num_placement = real_num_busy;
>>>> ret = ttm_bo_validate(bo, placement, &ctx);
>>>> if (ret)
>>>> return i915_ttm_err_to_gem(ret);
>>>>@@ -839,7 +834,7 @@ static int __i915_ttm_get_pages(struct
>>>>drm_i915_gem_object *obj,
>>>> static int i915_ttm_get_pages(struct drm_i915_gem_object *obj)
>>>> {
>>>>- struct ttm_place requested, busy[I915_TTM_MAX_PLACEMENTS];
>>>>+ struct ttm_place places[I915_TTM_MAX_PLACEMENTS + 1];
>>>> struct ttm_placement placement;
>>>> /* restricted by sg_alloc_table */
>>>>@@ -849,7 +844,7 @@ static int i915_ttm_get_pages(struct
>>>>drm_i915_gem_object *obj)
>>>> GEM_BUG_ON(obj->mm.n_placements > I915_TTM_MAX_PLACEMENTS);
>>>> /* Move to the requested placement. */
>>>>- i915_ttm_placement_from_obj(obj, &requested, busy, &placement);
>>>>+ i915_ttm_placement_from_obj(obj, places, &placement);
>>>> return __i915_ttm_get_pages(obj, &placement);
>>>> }
>>>>@@ -879,9 +874,7 @@ static int __i915_ttm_migrate(struct
>>>>drm_i915_gem_object *obj,
>>>> i915_ttm_place_from_region(mr, &requested, obj->bo_offset,
>>>> obj->base.size, flags);
>>>> placement.num_placement = 1;
>>>>- placement.num_busy_placement = 1;
>>>> placement.placement = &requested;
>>>>- placement.busy_placement = &requested;
>>>> ret = __i915_ttm_get_pages(obj, &placement);
>>>> if (ret)
>>>>diff --git a/drivers/gpu/drm/loongson/lsdc_ttm.c
>>>>b/drivers/gpu/drm/loongson/lsdc_ttm.c
>>>>index bf79dc55afa4..465f622ac05d 100644
>>>>--- a/drivers/gpu/drm/loongson/lsdc_ttm.c
>>>>+++ b/drivers/gpu/drm/loongson/lsdc_ttm.c
>>>>@@ -54,7 +54,6 @@ static void lsdc_bo_set_placement(struct
>>>>lsdc_bo *lbo, u32 domain)
>>>> pflags |= TTM_PL_FLAG_TOPDOWN;
>>>> lbo->placement.placement = lbo->placements;
>>>>- lbo->placement.busy_placement = lbo->placements;
>>>> if (domain & LSDC_GEM_DOMAIN_VRAM) {
>>>> lbo->placements[c].mem_type = TTM_PL_VRAM;
>>>>@@ -77,7 +76,6 @@ static void lsdc_bo_set_placement(struct
>>>>lsdc_bo *lbo, u32 domain)
>>>> }
>>>> lbo->placement.num_placement = c;
>>>>- lbo->placement.num_busy_placement = c;
>>>> for (i = 0; i < c; ++i) {
>>>> lbo->placements[i].fpfn = 0;
>>>>diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c
>>>>b/drivers/gpu/drm/nouveau/nouveau_bo.c
>>>>index 00cc7d1abaa3..56dcd25db1ce 100644
>>>>--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
>>>>+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
>>>>@@ -404,27 +404,6 @@ nouveau_bo_new(struct nouveau_cli *cli, u64
>>>>size, int align,
>>>> return 0;
>>>> }
>>>> -static void
>>>>-set_placement_list(struct ttm_place *pl, unsigned *n, uint32_t domain)
>>>>-{
>>>>- *n = 0;
>>>>-
>>>>- if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
>>>>- pl[*n].mem_type = TTM_PL_VRAM;
>>>>- pl[*n].flags = 0;
>>>>- (*n)++;
>>>>- }
>>>>- if (domain & NOUVEAU_GEM_DOMAIN_GART) {
>>>>- pl[*n].mem_type = TTM_PL_TT;
>>>>- pl[*n].flags = 0;
>>>>- (*n)++;
>>>>- }
>>>>- if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
>>>>- pl[*n].mem_type = TTM_PL_SYSTEM;
>>>>- pl[(*n)++].flags = 0;
>>>>- }
>>>>-}
>>>>-
>>>> static void
>>>> set_placement_range(struct nouveau_bo *nvbo, uint32_t domain)
>>>> {
>>>>@@ -452,10 +431,6 @@ set_placement_range(struct nouveau_bo
>>>>*nvbo, uint32_t domain)
>>>> nvbo->placements[i].fpfn = fpfn;
>>>> nvbo->placements[i].lpfn = lpfn;
>>>> }
>>>>- for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
>>>>- nvbo->busy_placements[i].fpfn = fpfn;
>>>>- nvbo->busy_placements[i].lpfn = lpfn;
>>>>- }
>>>> }
>>>> }
>>>> @@ -463,15 +438,32 @@ void
>>>> nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t domain,
>>>> uint32_t busy)
>>>> {
>>>>- struct ttm_placement *pl = &nvbo->placement;
>>>>+ unsigned int *n = &nvbo->placement.num_placement;
>>>>+ struct ttm_place *pl = nvbo->placements;
>>>> - pl->placement = nvbo->placements;
>>>>- set_placement_list(nvbo->placements, &pl->num_placement, domain);
>>>>+ domain |= busy;
>>>> - pl->busy_placement = nvbo->busy_placements;
>>>>- set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
>>>>- domain | busy);
>>>>+ *n = 0;
>>>>+ if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
>>>>+ pl[*n].mem_type = TTM_PL_VRAM;
>>>>+ pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_VRAM ?
>>>>+ TTM_PL_FLAG_FALLBACK : 0;
>>>>+ (*n)++;
>>>>+ }
>>>>+ if (domain & NOUVEAU_GEM_DOMAIN_GART) {
>>>>+ pl[*n].mem_type = TTM_PL_TT;
>>>>+ pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_GART ?
>>>>+ TTM_PL_FLAG_FALLBACK : 0;
>>>>+ (*n)++;
>>>>+ }
>>>>+ if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
>>>>+ pl[*n].mem_type = TTM_PL_SYSTEM;
>>>>+ pl[*n].flags = busy & NOUVEAU_GEM_DOMAIN_CPU ?
>>>>+ TTM_PL_FLAG_FALLBACK : 0;
>>>>+ (*n)++;
>>>>+ }
>>>> + nvbo->placement.placement = nvbo->placements;
>>>> set_placement_range(nvbo, domain);
>>>> }
>>>> @@ -1314,11 +1306,6 @@ vm_fault_t
>>>>nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
>>>> nvbo->placements[i].lpfn = mappable;
>>>> }
>>>> - for (i = 0; i < nvbo->placement.num_busy_placement; ++i) {
>>>>- nvbo->busy_placements[i].fpfn = 0;
>>>>- nvbo->busy_placements[i].lpfn = mappable;
>>>>- }
>>>>-
>>>> nouveau_bo_placement_set(nvbo, NOUVEAU_GEM_DOMAIN_VRAM, 0);
>>>> }
>>>> diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.h
>>>>b/drivers/gpu/drm/nouveau/nouveau_bo.h
>>>>index 70c551921a9e..e9dfab6a8156 100644
>>>>--- a/drivers/gpu/drm/nouveau/nouveau_bo.h
>>>>+++ b/drivers/gpu/drm/nouveau/nouveau_bo.h
>>>>@@ -15,7 +15,6 @@ struct nouveau_bo {
>>>> struct ttm_placement placement;
>>>> u32 valid_domains;
>>>> struct ttm_place placements[3];
>>>>- struct ttm_place busy_placements[3];
>>>> bool force_coherent;
>>>> struct ttm_bo_kmap_obj kmap;
>>>> struct list_head head;
>>>>diff --git a/drivers/gpu/drm/qxl/qxl_object.c
>>>>b/drivers/gpu/drm/qxl/qxl_object.c
>>>>index 06a58dad5f5c..1e46b0a6e478 100644
>>>>--- a/drivers/gpu/drm/qxl/qxl_object.c
>>>>+++ b/drivers/gpu/drm/qxl/qxl_object.c
>>>>@@ -66,7 +66,6 @@ void qxl_ttm_placement_from_domain(struct
>>>>qxl_bo *qbo, u32 domain)
>>>> pflag |= TTM_PL_FLAG_TOPDOWN;
>>>> qbo->placement.placement = qbo->placements;
>>>>- qbo->placement.busy_placement = qbo->placements;
>>>> if (domain == QXL_GEM_DOMAIN_VRAM) {
>>>> qbo->placements[c].mem_type = TTM_PL_VRAM;
>>>> qbo->placements[c++].flags = pflag;
>>>>@@ -86,7 +85,6 @@ void qxl_ttm_placement_from_domain(struct
>>>>qxl_bo *qbo, u32 domain)
>>>> qbo->placements[c++].flags = 0;
>>>> }
>>>> qbo->placement.num_placement = c;
>>>>- qbo->placement.num_busy_placement = c;
>>>> for (i = 0; i < c; ++i) {
>>>> qbo->placements[i].fpfn = 0;
>>>> qbo->placements[i].lpfn = 0;
>>>>diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c
>>>>b/drivers/gpu/drm/qxl/qxl_ttm.c
>>>>index 1a82629bce3f..765a144cea14 100644
>>>>--- a/drivers/gpu/drm/qxl/qxl_ttm.c
>>>>+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
>>>>@@ -60,9 +60,7 @@ static void qxl_evict_flags(struct
>>>>ttm_buffer_object *bo,
>>>> if (!qxl_ttm_bo_is_qxl_bo(bo)) {
>>>> placement->placement = &placements;
>>>>- placement->busy_placement = &placements;
>>>> placement->num_placement = 1;
>>>>- placement->num_busy_placement = 1;
>>>> return;
>>>> }
>>>> qbo = to_qxl_bo(bo);
>>>>diff --git a/drivers/gpu/drm/radeon/radeon_object.c
>>>>b/drivers/gpu/drm/radeon/radeon_object.c
>>>>index 10c0fbd9d2b4..a955f8a2f7fe 100644
>>>>--- a/drivers/gpu/drm/radeon/radeon_object.c
>>>>+++ b/drivers/gpu/drm/radeon/radeon_object.c
>>>>@@ -78,7 +78,6 @@ void radeon_ttm_placement_from_domain(struct
>>>>radeon_bo *rbo, u32 domain)
>>>> u32 c = 0, i;
>>>> rbo->placement.placement = rbo->placements;
>>>>- rbo->placement.busy_placement = rbo->placements;
>>>> if (domain & RADEON_GEM_DOMAIN_VRAM) {
>>>> /* Try placing BOs which don't need CPU access outside of the
>>>> * CPU accessible part of VRAM
>>>>@@ -114,7 +113,6 @@ void radeon_ttm_placement_from_domain(struct
>>>>radeon_bo *rbo, u32 domain)
>>>> }
>>>> rbo->placement.num_placement = c;
>>>>- rbo->placement.num_busy_placement = c;
>>>> for (i = 0; i < c; ++i) {
>>>> if ((rbo->flags & RADEON_GEM_CPU_ACCESS) &&
>>>>diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c
>>>>b/drivers/gpu/drm/radeon/radeon_ttm.c
>>>>index de4e6d78f1e1..2078b0000e22 100644
>>>>--- a/drivers/gpu/drm/radeon/radeon_ttm.c
>>>>+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
>>>>@@ -92,9 +92,7 @@ static void radeon_evict_flags(struct
>>>>ttm_buffer_object *bo,
>>>> if (!radeon_ttm_bo_is_radeon_bo(bo)) {
>>>> placement->placement = &placements;
>>>>- placement->busy_placement = &placements;
>>>> placement->num_placement = 1;
>>>>- placement->num_busy_placement = 1;
>>>> return;
>>>> }
>>>> rbo = container_of(bo, struct radeon_bo, tbo);
>>>>@@ -114,15 +112,11 @@ static void radeon_evict_flags(struct
>>>>ttm_buffer_object *bo,
>>>> */
>>>> radeon_ttm_placement_from_domain(rbo,
>>>>RADEON_GEM_DOMAIN_VRAM |
>>>> RADEON_GEM_DOMAIN_GTT);
>>>>- rbo->placement.num_busy_placement = 0;
>>>> for (i = 0; i < rbo->placement.num_placement; i++) {
>>>> if (rbo->placements[i].mem_type == TTM_PL_VRAM) {
>>>> if (rbo->placements[i].fpfn < fpfn)
>>>> rbo->placements[i].fpfn = fpfn;
>>>>- } else {
>>>>- rbo->placement.busy_placement =
>>>>- &rbo->placements[i];
>>>>- rbo->placement.num_busy_placement = 1;
>>>>+ rbo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
>>>> }
>>>> }
>>>> } else
>>>>diff --git a/drivers/gpu/drm/radeon/radeon_uvd.c
>>>>b/drivers/gpu/drm/radeon/radeon_uvd.c
>>>>index a2cda184b2b2..058a1c8451b2 100644
>>>>--- a/drivers/gpu/drm/radeon/radeon_uvd.c
>>>>+++ b/drivers/gpu/drm/radeon/radeon_uvd.c
>>>>@@ -324,7 +324,6 @@ void
>>>>radeon_uvd_force_into_uvd_segment(struct radeon_bo *rbo,
>>>> rbo->placements[1].fpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
>>>> rbo->placements[1].lpfn += (256 * 1024 * 1024) >> PAGE_SHIFT;
>>>> rbo->placement.num_placement++;
>>>>- rbo->placement.num_busy_placement++;
>>>> }
>>>> void radeon_uvd_free_handles(struct radeon_device *rdev,
>>>>struct drm_file *filp)
>>>>diff --git a/drivers/gpu/drm/ttm/ttm_bo.c
>>>>b/drivers/gpu/drm/ttm/ttm_bo.c
>>>>index 8c1eaa74fa21..a5e11a92e0b9 100644
>>>>--- a/drivers/gpu/drm/ttm/ttm_bo.c
>>>>+++ b/drivers/gpu/drm/ttm/ttm_bo.c
>>>>@@ -410,8 +410,8 @@ static int ttm_bo_bounce_temp_buffer(struct
>>>>ttm_buffer_object *bo,
>>>> struct ttm_resource *hop_mem;
>>>> int ret;
>>>> - hop_placement.num_placement =
>>>>hop_placement.num_busy_placement = 1;
>>>>- hop_placement.placement = hop_placement.busy_placement = hop;
>>>>+ hop_placement.num_placement = 1;
>>>>+ hop_placement.placement = hop;
>>>> /* find space in the bounce domain */
>>>> ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
>>>>@@ -440,10 +440,9 @@ static int ttm_bo_evict(struct
>>>>ttm_buffer_object *bo,
>>>> dma_resv_assert_held(bo->base.resv);
>>>> placement.num_placement = 0;
>>>>- placement.num_busy_placement = 0;
>>>> bdev->funcs->evict_flags(bo, &placement);
>>>> - if (!placement.num_placement && !placement.num_busy_placement) {
>>>>+ if (!placement.num_placement) {
>>>> ret = ttm_bo_wait_ctx(bo, ctx);
>>>> if (ret)
>>>> return ret;
>>>>@@ -791,6 +790,9 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo,
>>>> const struct ttm_place *place = &placement->placement[i];
>>>> struct ttm_resource_manager *man;
>>>> + if (place->flags & TTM_PL_FLAG_FALLBACK)
>>>>+ continue;
>>>>+
>>>> man = ttm_manager_type(bdev, place->mem_type);
>>>> if (!man || !ttm_resource_manager_used(man))
>>>> continue;
>>>>@@ -813,10 +815,13 @@ int ttm_bo_mem_space(struct
>>>>ttm_buffer_object *bo,
>>>> return 0;
>>>> }
>>>> - for (i = 0; i < placement->num_busy_placement; ++i) {
>>>>- const struct ttm_place *place = &placement->busy_placement[i];
>>>>+ for (i = 0; i < placement->num_placement; ++i) {
>>>>+ const struct ttm_place *place = &placement->placement[i];
>>>> struct ttm_resource_manager *man;
>>>> + if (place->flags & TTM_PL_FLAG_DESIRED)
>>>>+ continue;
>>>>+
>>>> man = ttm_manager_type(bdev, place->mem_type);
>>>> if (!man || !ttm_resource_manager_used(man))
>>>> continue;
>>>>@@ -904,11 +909,11 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
>>>> /*
>>>> * Remove the backing store if no placement is given.
>>>> */
>>>>- if (!placement->num_placement && !placement->num_busy_placement)
>>>>+ if (!placement->num_placement)
>>>> return ttm_bo_pipeline_gutting(bo);
>>>> /* Check whether we need to move buffer. */
>>>>- if (bo->resource && ttm_resource_compat(bo->resource, placement))
>>>>+ if (bo->resource && ttm_resource_compatible(bo->resource,
>>>>placement))
>>>> return 0;
>>>> /* Moving of pinned BOs is forbidden */
>>>>diff --git a/drivers/gpu/drm/ttm/ttm_resource.c
>>>>b/drivers/gpu/drm/ttm/ttm_resource.c
>>>>index 02b96d23fdb9..fb14f7716cf8 100644
>>>>--- a/drivers/gpu/drm/ttm/ttm_resource.c
>>>>+++ b/drivers/gpu/drm/ttm/ttm_resource.c
>>>>@@ -291,37 +291,15 @@ bool ttm_resource_intersects(struct
>>>>ttm_device *bdev,
>>>> }
>>>> /**
>>>>- * ttm_resource_compatible - test for compatibility
>>>>+ * ttm_resource_compatible - check if resource is compatible
>>>>with placement
>>>> *
>>>>- * @bdev: TTM device structure
>>>>- * @res: The resource to test
>>>>- * @place: The placement to test
>>>>- * @size: How many bytes the new allocation needs.
>>>>- *
>>>>- * Test if @res compatible with @place and @size.
>>>>+ * @res: the resource to check
>>>>+ * @placement: the placement to check against
>>>> *
>>>>- * Returns true if the res placement compatible with @place and @size.
>>>>+ * Returns true if the placement is compatible.
>>>> */
>>>>-bool ttm_resource_compatible(struct ttm_device *bdev,
>>>>- struct ttm_resource *res,
>>>>- const struct ttm_place *place,
>>>>- size_t size)
>>>>-{
>>>>- struct ttm_resource_manager *man;
>>>>-
>>>>- if (!res || !place)
>>>>- return false;
>>>>-
>>>>- man = ttm_manager_type(bdev, res->mem_type);
>>>>- if (!man->func->compatible)
>>>>- return true;
>>>>-
>>>>- return man->func->compatible(man, res, place, size);
>>>>-}
>>>>-
>>>>-static bool ttm_resource_places_compat(struct ttm_resource *res,
>>>>- const struct ttm_place *places,
>>>>- unsigned num_placement)
>>>>+bool ttm_resource_compatible(struct ttm_resource *res,
>>>>+ struct ttm_placement *placement)
>>>> {
>>>> struct ttm_buffer_object *bo = res->bo;
>>>> struct ttm_device *bdev = bo->bdev;
>>>>@@ -330,44 +308,25 @@ static bool
>>>>ttm_resource_places_compat(struct ttm_resource *res,
>>>> if (res->placement & TTM_PL_FLAG_TEMPORARY)
>>>> return false;
>>>> - for (i = 0; i < num_placement; i++) {
>>>>- const struct ttm_place *heap = &places[i];
>>>>+ for (i = 0; i < placement->num_placement; i++) {
>>>>+ const struct ttm_place *place = &placement->placement[i];
>>>>+ struct ttm_resource_manager *man;
>>>> - if (!ttm_resource_compatible(bdev, res, heap,
>>>>bo->base.size))
>>>>+ if (res->mem_type != place->mem_type)
>>>>+ continue;
>>>>+
>>>>+ man = ttm_manager_type(bdev, res->mem_type);
>>>>+ if (man->func->compatible &&
>>>>+ !man->func->compatible(man, res, place, bo->base.size))
>>>> continue;
>>>> - if ((res->mem_type == heap->mem_type) &&
>>>>- (!(heap->flags & TTM_PL_FLAG_CONTIGUOUS) ||
>>>>+ if ((!(place->flags & TTM_PL_FLAG_CONTIGUOUS) ||
>>>> (res->placement & TTM_PL_FLAG_CONTIGUOUS)))
>>>> return true;
>>>> }
>>>> return false;
>>>> }
>>>> -/**
>>>>- * ttm_resource_compat - check if resource is compatible with
>>>>placement
>>>>- *
>>>>- * @res: the resource to check
>>>>- * @placement: the placement to check against
>>>>- *
>>>>- * Returns true if the placement is compatible.
>>>>- */
>>>>-bool ttm_resource_compat(struct ttm_resource *res,
>>>>- struct ttm_placement *placement)
>>>>-{
>>>>- if (ttm_resource_places_compat(res, placement->placement,
>>>>- placement->num_placement))
>>>>- return true;
>>>>-
>>>>- if ((placement->busy_placement != placement->placement ||
>>>>- placement->num_busy_placement > placement->num_placement) &&
>>>>- ttm_resource_places_compat(res, placement->busy_placement,
>>>>- placement->num_busy_placement))
>>>>- return true;
>>>>-
>>>>- return false;
>>>>-}
>>>>-
>>>> void ttm_resource_set_bo(struct ttm_resource *res,
>>>> struct ttm_buffer_object *bo)
>>>> {
>>>>diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>>>b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>>>index 2bfac3aad7b7..bfd41ce3c8f4 100644
>>>>--- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>>>+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
>>>>@@ -742,9 +742,21 @@ void vmw_bo_move_notify(struct
>>>>ttm_buffer_object *bo,
>>>> vmw_resource_unbind_list(vbo);
>>>> }
>>>> +static u32 placement_flags(u32 domain, u32 desired, u32 fallback)
>>>>+{
>>>>+ if (desired & fallback & domain)
>>>>+ return 0;
>>>>+
>>>>+ if (desired & domain)
>>>>+ return TTM_PL_FLAG_DESIRED;
>>>>+
>>>>+ return TTM_PL_FLAG_FALLBACK;
>>>>+}
>>>>+
>>>> static u32
>>>>-set_placement_list(struct ttm_place *pl, u32 domain)
>>>>+set_placement_list(struct ttm_place *pl, u32 desired, u32 fallback)
>>>> {
>>>>+ u32 domain = desired | fallback;
>>>> u32 n = 0;
>>>> /*
>>>>@@ -752,35 +764,40 @@ set_placement_list(struct ttm_place *pl,
>>>>u32 domain)
>>>> */
>>>> if (domain & VMW_BO_DOMAIN_MOB) {
>>>> pl[n].mem_type = VMW_PL_MOB;
>>>>- pl[n].flags = 0;
>>>>+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_MOB, desired,
>>>>+ fallback);
>>>> pl[n].fpfn = 0;
>>>> pl[n].lpfn = 0;
>>>> n++;
>>>> }
>>>> if (domain & VMW_BO_DOMAIN_GMR) {
>>>> pl[n].mem_type = VMW_PL_GMR;
>>>>- pl[n].flags = 0;
>>>>+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_GMR, desired,
>>>>+ fallback);
>>>> pl[n].fpfn = 0;
>>>> pl[n].lpfn = 0;
>>>> n++;
>>>> }
>>>> if (domain & VMW_BO_DOMAIN_VRAM) {
>>>> pl[n].mem_type = TTM_PL_VRAM;
>>>>- pl[n].flags = 0;
>>>>+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_VRAM, desired,
>>>>+ fallback);
>>>> pl[n].fpfn = 0;
>>>> pl[n].lpfn = 0;
>>>> n++;
>>>> }
>>>> if (domain & VMW_BO_DOMAIN_WAITABLE_SYS) {
>>>> pl[n].mem_type = VMW_PL_SYSTEM;
>>>>- pl[n].flags = 0;
>>>>+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_WAITABLE_SYS,
>>>>+ desired, fallback);
>>>> pl[n].fpfn = 0;
>>>> pl[n].lpfn = 0;
>>>> n++;
>>>> }
>>>> if (domain & VMW_BO_DOMAIN_SYS) {
>>>> pl[n].mem_type = TTM_PL_SYSTEM;
>>>>- pl[n].flags = 0;
>>>>+ pl[n].flags = placement_flags(VMW_BO_DOMAIN_SYS, desired,
>>>>+ fallback);
>>>> pl[n].fpfn = 0;
>>>> pl[n].lpfn = 0;
>>>> n++;
>>>>@@ -806,7 +823,7 @@ void vmw_bo_placement_set(struct vmw_bo *bo,
>>>>u32 domain, u32 busy_domain)
>>>> u32 i;
>>>> pl->placement = bo->places;
>>>>- pl->num_placement = set_placement_list(bo->places, domain);
>>>>+ pl->num_placement = set_placement_list(bo->places, domain,
>>>>busy_domain);
>>>> if (drm_debug_enabled(DRM_UT_DRIVER) && bo->tbo.resource) {
>>>> for (i = 0; i < pl->num_placement; ++i) {
>>>>@@ -821,8 +838,6 @@ void vmw_bo_placement_set(struct vmw_bo *bo,
>>>>u32 domain, u32 busy_domain)
>>>> __func__, bo->tbo.resource->mem_type, domain);
>>>> }
>>>> - pl->busy_placement = bo->busy_places;
>>>>- pl->num_busy_placement =
>>>>set_placement_list(bo->busy_places, busy_domain);
>>>> }
>>>> void vmw_bo_placement_set_default_accelerated(struct vmw_bo *bo)
>>>>diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>>>b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>>>index a84fffcef8e1..4d23d0a70bcb 100644
>>>>--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>>>+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
>>>>@@ -46,15 +46,11 @@ static const struct ttm_place
>>>>sys_placement_flags = {
>>>> struct ttm_placement vmw_vram_placement = {
>>>> .num_placement = 1,
>>>> .placement = &vram_placement_flags,
>>>>- .num_busy_placement = 1,
>>>>- .busy_placement = &vram_placement_flags
>>>> };
>>>> struct ttm_placement vmw_sys_placement = {
>>>> .num_placement = 1,
>>>> .placement = &sys_placement_flags,
>>>>- .num_busy_placement = 1,
>>>>- .busy_placement = &sys_placement_flags
>>>> };
>>>> const size_t vmw_tt_size = sizeof(struct vmw_ttm_tt);
>>>>diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
>>>>index 8e4a3b1f6b93..1e5edc7f0b85 100644
>>>>--- a/drivers/gpu/drm/xe/xe_bo.c
>>>>+++ b/drivers/gpu/drm/xe/xe_bo.c
>>>>@@ -38,22 +38,26 @@ static const struct ttm_place
>>>>sys_placement_flags = {
>>>> static struct ttm_placement sys_placement = {
>>>> .num_placement = 1,
>>>> .placement = &sys_placement_flags,
>>>>- .num_busy_placement = 1,
>>>>- .busy_placement = &sys_placement_flags,
>>>> };
>>>> -static const struct ttm_place tt_placement_flags = {
>>>>- .fpfn = 0,
>>>>- .lpfn = 0,
>>>>- .mem_type = XE_PL_TT,
>>>>- .flags = 0,
>>>>+static const struct ttm_place tt_placement_flags[] = {
>>>>+ {
>>>>+ .fpfn = 0,
>>>>+ .lpfn = 0,
>>>>+ .mem_type = XE_PL_TT,
>>>>+ .flags = TTM_PL_FLAG_DESIRED,
>>>>+ },
>>>>+ {
>>>>+ .fpfn = 0,
>>>>+ .lpfn = 0,
>>>>+ .mem_type = XE_PL_SYSTEM,
>>>>+ .flags = TTM_PL_FLAG_FALLBACK,
>>>>+ }
>>>> };
>>>> static struct ttm_placement tt_placement = {
>>>>- .num_placement = 1,
>>>>- .placement = &tt_placement_flags,
>>>>- .num_busy_placement = 1,
>>>>- .busy_placement = &sys_placement_flags,
>>>>+ .num_placement = 2,
>>>>+ .placement = tt_placement_flags,
>>>> };
>>>> bool mem_type_is_vram(u32 mem_type)
>>>>@@ -230,8 +234,6 @@ static int
>>>>__xe_bo_placement_for_flags(struct xe_device *xe, struct xe_bo
>>>>*bo,
>>>> bo->placement = (struct ttm_placement) {
>>>> .num_placement = c,
>>>> .placement = bo->placements,
>>>>- .num_busy_placement = c,
>>>>- .busy_placement = bo->placements,
>>>> };
>>>> return 0;
>>>>@@ -251,7 +253,6 @@ static void xe_evict_flags(struct
>>>>ttm_buffer_object *tbo,
>>>> /* Don't handle scatter gather BOs */
>>>> if (tbo->type == ttm_bo_type_sg) {
>>>> placement->num_placement = 0;
>>>>- placement->num_busy_placement = 0;
>>>> return;
>>>> }
>>>> @@ -1353,8 +1354,6 @@ static int
>>>>__xe_bo_fixed_placement(struct xe_device *xe,
>>>> bo->placement = (struct ttm_placement) {
>>>> .num_placement = 1,
>>>> .placement = place,
>>>>- .num_busy_placement = 1,
>>>>- .busy_placement = place,
>>>> };
>>>> return 0;
>>>>@@ -2112,9 +2111,7 @@ int xe_bo_migrate(struct xe_bo *bo, u32 mem_type)
>>>> xe_place_from_ttm_type(mem_type, &requested);
>>>> placement.num_placement = 1;
>>>>- placement.num_busy_placement = 1;
>>>> placement.placement = &requested;
>>>>- placement.busy_placement = &requested;
>>>> /*
>>>> * Stolen needs to be handled like below VRAM handling if
>>>>we ever need
>>>>diff --git a/include/drm/ttm/ttm_placement.h
>>>>b/include/drm/ttm/ttm_placement.h
>>>>index 8074d0f6cae5..b510a4812609 100644
>>>>--- a/include/drm/ttm/ttm_placement.h
>>>>+++ b/include/drm/ttm/ttm_placement.h
>>>>@@ -64,6 +64,12 @@
>>>> /* For multihop handling */
>>>> #define TTM_PL_FLAG_TEMPORARY (1 << 2)
>>>> +/* Placement is never used during eviction */
>>>>+#define TTM_PL_FLAG_DESIRED (1 << 3)
>>>>+
>>>>+/* Placement is only used during eviction */
>>>>+#define TTM_PL_FLAG_FALLBACK (1 << 4)
>>>>+
>>>> /**
>>>> * struct ttm_place
>>>> *
>>>>@@ -86,16 +92,12 @@ struct ttm_place {
>>>> *
>>>> * @num_placement: number of preferred placements
>>>> * @placement: preferred placements
>>>>- * @num_busy_placement: number of preferred placements when
>>>>need to evict buffer
>>>>- * @busy_placement: preferred placements when need to evict buffer
>>>> *
>>>> * Structure indicating the placement you request for an object.
>>>> */
>>>> struct ttm_placement {
>>>> unsigned num_placement;
>>>> const struct ttm_place *placement;
>>>>- unsigned num_busy_placement;
>>>>- const struct ttm_place *busy_placement;
>>>> };
>>>> #endif
>>>>diff --git a/include/drm/ttm/ttm_resource.h
>>>>b/include/drm/ttm/ttm_resource.h
>>>>index 78a226eba953..1afa13f0c22b 100644
>>>>--- a/include/drm/ttm/ttm_resource.h
>>>>+++ b/include/drm/ttm/ttm_resource.h
>>>>@@ -365,12 +365,8 @@ bool ttm_resource_intersects(struct
>>>>ttm_device *bdev,
>>>> struct ttm_resource *res,
>>>> const struct ttm_place *place,
>>>> size_t size);
>>>>-bool ttm_resource_compatible(struct ttm_device *bdev,
>>>>- struct ttm_resource *res,
>>>>- const struct ttm_place *place,
>>>>- size_t size);
>>>>-bool ttm_resource_compat(struct ttm_resource *res,
>>>>- struct ttm_placement *placement);
>>>>+bool ttm_resource_compatible(struct ttm_resource *res,
>>>>+ struct ttm_placement *placement);
>>>> void ttm_resource_set_bo(struct ttm_resource *res,
>>>> struct ttm_buffer_object *bo);
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: Re: Re: [PATCH 3/5] drm/ttm: replace busy placement with flags v6
2024-01-26 22:16 ` Lucas De Marchi
@ 2024-01-26 22:22 ` Lucas De Marchi
2024-01-27 15:43 ` Zeng, Oak
2024-01-29 11:00 ` Thomas Hellström
0 siblings, 2 replies; 19+ messages in thread
From: Lucas De Marchi @ 2024-01-26 22:22 UTC (permalink / raw)
To: Thomas Hellström
Cc: kherbst, michel.daenzer, nouveau, intel-gfx, dri-devel,
Christian König, zackr
On Fri, Jan 26, 2024 at 04:16:58PM -0600, Lucas De Marchi wrote:
>On Thu, Jan 18, 2024 at 05:38:16PM +0100, Thomas Hellström wrote:
>>
>>On 1/17/24 13:27, Thomas Hellström wrote:
>>>
>>>On 1/17/24 11:47, Thomas Hellström wrote:
>>>>Hi, Christian
>>>>
>>>>Xe changes look good. Will send the series to xe ci to check for
>>>>regressions.
>>>
>>>Hmm, there are some checkpatch warnings about author / SOB email
>>>mismatch,
>>
>>With those fixed, this patch is
>>
>>Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
>
>
>it actually broke drm-tip now that this is merged:
>
>../drivers/gpu/drm/xe/xe_bo.c:41:10: error: ‘struct ttm_placement’ has no member named ‘num_busy_placement’; did you mean ‘num_placement’
> 41 | .num_busy_placement = 1,
> | ^~~~~~~~~~~~~~~~~~
> | num_placement
>../drivers/gpu/drm/xe/xe_bo.c:41:31: error: excess elements in struct initializer [-Werror]
> 41 | .num_busy_placement = 1,
> | ^
>
>
>Apparently a conflict with another patch that got applied a few days
>ago: a201c6ee37d6 ("drm/xe/bo: Evict VRAM to TT rather than to system")
oh, no... apparently that commit is from a long time ago. The problem
was that drm-misc-next was not yet in sync with drm-next. Thomas, do you
have a fixup for this to put in rerere?
Lucas De Marchi
^ permalink raw reply [flat|nested] 19+ messages in thread
* RE: Re: Re: [PATCH 3/5] drm/ttm: replace busy placement with flags v6
2024-01-26 22:22 ` Lucas De Marchi
@ 2024-01-27 15:43 ` Zeng, Oak
2024-01-29 11:00 ` Thomas Hellström
1 sibling, 0 replies; 19+ messages in thread
From: Zeng, Oak @ 2024-01-27 15:43 UTC (permalink / raw)
To: De Marchi, Lucas, Thomas Hellström
Cc: kherbst, michel.daenzer, Christian König, intel-gfx,
dri-devel, nouveau, zackr
Hi Lucas,
I encountered this build issue as well. I submitted a fix for drm-tip.
Oak
> -----Original Message-----
> From: dri-devel <dri-devel-bounces@lists.freedesktop.org> On Behalf Of Lucas
> De Marchi
> Sent: Friday, January 26, 2024 5:23 PM
> To: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> Cc: kherbst@redhat.com; michel.daenzer@mailbox.org;
> nouveau@lists.freedesktop.org; intel-gfx@lists.freedesktop.org; dri-
> devel@lists.freedesktop.org; Christian König
> <ckoenig.leichtzumerken@gmail.com>; zackr@vmware.com
> Subject: Re: Re: Re: [PATCH 3/5] drm/ttm: replace busy placement with flags v6
>
> On Fri, Jan 26, 2024 at 04:16:58PM -0600, Lucas De Marchi wrote:
> >On Thu, Jan 18, 2024 at 05:38:16PM +0100, Thomas Hellström wrote:
> >>
> >>On 1/17/24 13:27, Thomas Hellström wrote:
> >>>
> >>>On 1/17/24 11:47, Thomas Hellström wrote:
> >>>>Hi, Christian
> >>>>
> >>>>Xe changes look good. Will send the series to xe ci to check for
> >>>>regressions.
> >>>
> >>>Hmm, there are some checkpatch warnings about author / SOB email
> >>>mismatch,
> >>
> >>With those fixed, this patch is
> >>
> >>Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> >
> >
> >it actually broke drm-tip now that this is merged:
> >
> >../drivers/gpu/drm/xe/xe_bo.c:41:10: error: ‘struct ttm_placement’ has no
> member named ‘num_busy_placement’; did you mean ‘num_placement’
> > 41 | .num_busy_placement = 1,
> > | ^~~~~~~~~~~~~~~~~~
> > | num_placement
> >../drivers/gpu/drm/xe/xe_bo.c:41:31: error: excess elements in struct initializer
> [-Werror]
> > 41 | .num_busy_placement = 1,
> > | ^
> >
> >
> >Apparently a conflict with another patch that got applied a few days
> >ago: a201c6ee37d6 ("drm/xe/bo: Evict VRAM to TT rather than to system")
>
> oh, no... apparently that commit is from a long time ago. The problem
> was that drm-misc-next was not yet in sync with drm-next. Thomas, do you
> have a fixup for this to put in rerere?
>
> Lucas De Marchi
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: Re: Re: [PATCH 3/5] drm/ttm: replace busy placement with flags v6
2024-01-26 22:22 ` Lucas De Marchi
2024-01-27 15:43 ` Zeng, Oak
@ 2024-01-29 11:00 ` Thomas Hellström
1 sibling, 0 replies; 19+ messages in thread
From: Thomas Hellström @ 2024-01-29 11:00 UTC (permalink / raw)
To: Lucas De Marchi
Cc: kherbst, michel.daenzer, Christian König, intel-gfx,
dri-devel, nouveau, zackr
On Fri, 2024-01-26 at 16:22 -0600, Lucas De Marchi wrote:
> On Fri, Jan 26, 2024 at 04:16:58PM -0600, Lucas De Marchi wrote:
> > On Thu, Jan 18, 2024 at 05:38:16PM +0100, Thomas Hellström wrote:
> > >
> > > On 1/17/24 13:27, Thomas Hellström wrote:
> > > >
> > > > On 1/17/24 11:47, Thomas Hellström wrote:
> > > > > Hi, Christian
> > > > >
> > > > > Xe changes look good. Will send the series to xe ci to check
> > > > > for
> > > > > regressions.
> > > >
> > > > Hmm, there are some checkpatch warnings about author / SOB
> > > > email
> > > > mismatch,
> > >
> > > With those fixed, this patch is
> > >
> > > Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> >
> >
> > it actually broke drm-tip now that this is merged:
> >
> > ../drivers/gpu/drm/xe/xe_bo.c:41:10: error: ‘struct ttm_placement’
> > has no member named ‘num_busy_placement’; did you mean
> > ‘num_placement’
> > 41 | .num_busy_placement = 1,
> > | ^~~~~~~~~~~~~~~~~~
> > | num_placement
> > ../drivers/gpu/drm/xe/xe_bo.c:41:31: error: excess elements in
> > struct initializer [-Werror]
> > 41 | .num_busy_placement = 1,
> > | ^
> >
> >
> > Apparently a conflict with another patch that got applied a few
> > days
> > ago: a201c6ee37d6 ("drm/xe/bo: Evict VRAM to TT rather than to
> > system")
>
> oh, no... apparently that commit is from a long time ago. The
> problem
> was that drm-misc-next was not yet in sync with drm-next. Thomas, do
> you
> have a fixup for this to put in rerere?
>
> Lucas De Marchi
I added this as a manual fixup and ran some quick igt tests.
Seems to work.
^ permalink raw reply [flat|nested] 19+ messages in thread
end of thread, other threads:[~2024-01-29 11:00 UTC | newest]
Thread overview: 19+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-01-12 12:51 Rework TTMs busy handling Christian König
2024-01-12 12:51 ` [PATCH 1/5] drm/vmwgfx: remove vmw_vram_gmr_placement Christian König
2024-01-12 12:51 ` [PATCH 2/5] drm/ttm: return ENOSPC from ttm_bo_mem_space Christian König
2024-01-17 10:18 ` Thomas Hellström
2024-01-12 12:51 ` [PATCH 3/5] drm/ttm: replace busy placement with flags v6 Christian König
2024-01-17 8:01 ` Thomas Zimmermann
2024-01-17 10:47 ` Thomas Hellström
2024-01-17 12:27 ` Thomas Hellström
2024-01-18 16:38 ` Thomas Hellström
2024-01-26 22:16 ` Lucas De Marchi
2024-01-26 22:22 ` Lucas De Marchi
2024-01-27 15:43 ` Zeng, Oak
2024-01-29 11:00 ` Thomas Hellström
2024-01-12 12:51 ` [PATCH 4/5] drm/ttm: improve idle/busy handling v3 Christian König
2024-01-18 14:24 ` Thomas Hellström
2024-01-18 16:48 ` Thomas Hellström
2024-01-12 12:51 ` [PATCH 5/5] drm/amdgpu: use GTT only as fallback for VRAM|GTT Christian König
2024-01-16 9:57 ` Rework TTMs busy handling Christian König
2024-01-17 3:13 ` Zack Rusin
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).