[PATCH 3/5] drm/ttm: replace busy placement with flags v6

Thomas Hellström thomas.hellstrom at linux.intel.com
Thu Jan 18 16:38:16 UTC 2024


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 at 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 at 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 at amd.com>
>>> Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram at 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);


More information about the Intel-gfx mailing list