Re: [PATCH 38/49] drm/amdgpu/ttm: use bo manager subclassing for vram/gtt mgrs

[Christian König <christian.koenig@xxxxxxx>]

Am 31.07.20 um 06:05 schrieb Dave Airlie:
> From: Dave Airlie <airlied@xxxxxxxxxx>
>
> Signed-off-by: Dave Airlie <airlied@xxxxxxxxxx>

Reviewed-by: Christian König <christian.koenig@xxxxxxx>

> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c  | 35 +++++++++++--------
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c | 36 +++++++++++++-------
>   2 files changed, 44 insertions(+), 27 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c
> index 0b0d09d19b4f..83d88ee73468 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c
> @@ -25,11 +25,17 @@
>   #include "amdgpu.h"
>   
>   struct amdgpu_gtt_mgr {
> +	struct ttm_mem_type_manager manager;
>   	struct drm_mm mm;
>   	spinlock_t lock;
>   	atomic64_t available;
>   };
>   
> +static inline struct amdgpu_gtt_mgr *to_gtt_mgr(struct ttm_mem_type_manager *man)
> +{
> +	return container_of(man, struct amdgpu_gtt_mgr, manager);
> +}
> +
>   struct amdgpu_gtt_node {
>   	struct drm_mm_node node;
>   	struct ttm_buffer_object *tbo;
> @@ -87,11 +93,16 @@ static const struct ttm_mem_type_manager_func amdgpu_gtt_mgr_func;
>    */
>   int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size)
>   {
> -	struct ttm_mem_type_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT);
> +	struct ttm_mem_type_manager *man;
>   	struct amdgpu_gtt_mgr *mgr;
>   	uint64_t start, size;
>   	int ret;
>   
> +	mgr = kzalloc(sizeof(*mgr), GFP_KERNEL);
> +	if (!mgr)
> +		return -ENOMEM;
> +
> +	man = &mgr->manager;
>   	man->use_tt = true;
>   	man->func = &amdgpu_gtt_mgr_func;
>   	man->available_caching = TTM_PL_MASK_CACHING;
> @@ -99,16 +110,11 @@ int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size)
>   
>   	ttm_bo_init_mm_base(&adev->mman.bdev, man, gtt_size >> PAGE_SHIFT);
>   
> -	mgr = kzalloc(sizeof(*mgr), GFP_KERNEL);
> -	if (!mgr)
> -		return -ENOMEM;
> -
>   	start = AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS;
>   	size = (adev->gmc.gart_size >> PAGE_SHIFT) - start;
>   	drm_mm_init(&mgr->mm, start, size);
>   	spin_lock_init(&mgr->lock);
>   	atomic64_set(&mgr->available, gtt_size >> PAGE_SHIFT);
> -	man->priv = mgr;
>   
>   	ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_total);
>   	if (ret) {
> @@ -121,6 +127,7 @@ int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size)
>   		return ret;
>   	}
>   
> +	ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, &mgr->manager);
>   	ttm_bo_use_mm(man);
>   	return 0;
>   }
> @@ -136,7 +143,7 @@ int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size)
>   void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev)
>   {
>   	struct ttm_mem_type_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT);
> -	struct amdgpu_gtt_mgr *mgr = man->priv;
> +	struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man);
>   	int ret;
>   
>   	ttm_bo_disable_mm(man);
> @@ -148,13 +155,13 @@ void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev)
>   	spin_lock(&mgr->lock);
>   	drm_mm_takedown(&mgr->mm);
>   	spin_unlock(&mgr->lock);
> -	kfree(mgr);
> -	man->priv = NULL;
>   
>   	device_remove_file(adev->dev, &dev_attr_mem_info_gtt_total);
>   	device_remove_file(adev->dev, &dev_attr_mem_info_gtt_used);
>   
>   	ttm_bo_man_cleanup(man);
> +	ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_TT, NULL);
> +	kfree(mgr);
>   }
>   
>   /**
> @@ -184,7 +191,7 @@ static int amdgpu_gtt_mgr_new(struct ttm_mem_type_manager *man,
>   			      const struct ttm_place *place,
>   			      struct ttm_mem_reg *mem)
>   {
> -	struct amdgpu_gtt_mgr *mgr = man->priv;
> +	struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man);
>   	struct amdgpu_gtt_node *node;
>   	int r;
>   
> @@ -245,7 +252,7 @@ static int amdgpu_gtt_mgr_new(struct ttm_mem_type_manager *man,
>   static void amdgpu_gtt_mgr_del(struct ttm_mem_type_manager *man,
>   			       struct ttm_mem_reg *mem)
>   {
> -	struct amdgpu_gtt_mgr *mgr = man->priv;
> +	struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man);
>   	struct amdgpu_gtt_node *node = mem->mm_node;
>   
>   	if (node) {
> @@ -267,7 +274,7 @@ static void amdgpu_gtt_mgr_del(struct ttm_mem_type_manager *man,
>    */
>   uint64_t amdgpu_gtt_mgr_usage(struct ttm_mem_type_manager *man)
>   {
> -	struct amdgpu_gtt_mgr *mgr = man->priv;
> +	struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man);
>   	s64 result = man->size - atomic64_read(&mgr->available);
>   
>   	return (result > 0 ? result : 0) * PAGE_SIZE;
> @@ -275,7 +282,7 @@ uint64_t amdgpu_gtt_mgr_usage(struct ttm_mem_type_manager *man)
>   
>   int amdgpu_gtt_mgr_recover(struct ttm_mem_type_manager *man)
>   {
> -	struct amdgpu_gtt_mgr *mgr = man->priv;
> +	struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man);
>   	struct amdgpu_gtt_node *node;
>   	struct drm_mm_node *mm_node;
>   	int r = 0;
> @@ -303,7 +310,7 @@ int amdgpu_gtt_mgr_recover(struct ttm_mem_type_manager *man)
>   static void amdgpu_gtt_mgr_debug(struct ttm_mem_type_manager *man,
>   				 struct drm_printer *printer)
>   {
> -	struct amdgpu_gtt_mgr *mgr = man->priv;
> +	struct amdgpu_gtt_mgr *mgr = to_gtt_mgr(man);
>   
>   	spin_lock(&mgr->lock);
>   	drm_mm_print(&mgr->mm, printer);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
> index d33a750e07a8..9d4a13926b8c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
> @@ -29,12 +29,18 @@
>   #include "atom.h"
>   
>   struct amdgpu_vram_mgr {
> +	struct ttm_mem_type_manager manager;
>   	struct drm_mm mm;
>   	spinlock_t lock;
>   	atomic64_t usage;
>   	atomic64_t vis_usage;
>   };
>   
> +static inline struct amdgpu_vram_mgr *to_vram_mgr(struct ttm_mem_type_manager *man)
> +{
> +	return container_of(man, struct amdgpu_vram_mgr, manager);
> +}
> +
>   /**
>    * DOC: mem_info_vram_total
>    *
> @@ -170,29 +176,32 @@ static const struct ttm_mem_type_manager_func amdgpu_vram_mgr_func;
>    */
>   int amdgpu_vram_mgr_init(struct amdgpu_device *adev)
>   {
> -	struct ttm_mem_type_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
> +	struct ttm_mem_type_manager *man;
>   	struct amdgpu_vram_mgr *mgr;
>   	int ret;
>   
> +	mgr = kzalloc(sizeof(*mgr), GFP_KERNEL);
> +	if (!mgr)
> +		return -ENOMEM;
> +
> +	man = &mgr->manager;
> +
>   	man->available_caching = TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_WC;
>   	man->default_caching = TTM_PL_FLAG_WC;
>   
>   	ttm_bo_init_mm_base(&adev->mman.bdev, man, adev->gmc.real_vram_size >> PAGE_SHIFT);
>   
>   	man->func = &amdgpu_vram_mgr_func;
> -	mgr = kzalloc(sizeof(*mgr), GFP_KERNEL);
> -	if (!mgr)
> -		return -ENOMEM;
>   
>   	drm_mm_init(&mgr->mm, 0, man->size);
>   	spin_lock_init(&mgr->lock);
> -	man->priv = mgr;
>   
>   	/* Add the two VRAM-related sysfs files */
>   	ret = sysfs_create_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes);
>   	if (ret)
>   		DRM_ERROR("Failed to register sysfs\n");
>   
> +	ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, &mgr->manager);
>   	ttm_bo_use_mm(man);
>   	return 0;
>   }
> @@ -208,7 +217,7 @@ int amdgpu_vram_mgr_init(struct amdgpu_device *adev)
>   void amdgpu_vram_mgr_fini(struct amdgpu_device *adev)
>   {
>   	struct ttm_mem_type_manager *man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
> -	struct amdgpu_vram_mgr *mgr = man->priv;
> +	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
>   	int ret;
>   
>   	ttm_bo_disable_mm(man);
> @@ -220,11 +229,12 @@ void amdgpu_vram_mgr_fini(struct amdgpu_device *adev)
>   	spin_lock(&mgr->lock);
>   	drm_mm_takedown(&mgr->mm);
>   	spin_unlock(&mgr->lock);
> -	kfree(mgr);
> -	man->priv = NULL;
> +
>   	sysfs_remove_files(&adev->dev->kobj, amdgpu_vram_mgr_attributes);
>   
>   	ttm_bo_man_cleanup(man);
> +	ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, NULL);
> +	kfree(mgr);
>   }
>   
>   /**
> @@ -314,7 +324,7 @@ static int amdgpu_vram_mgr_new(struct ttm_mem_type_manager *man,
>   			       struct ttm_mem_reg *mem)
>   {
>   	struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev);
> -	struct amdgpu_vram_mgr *mgr = man->priv;
> +	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
>   	struct drm_mm *mm = &mgr->mm;
>   	struct drm_mm_node *nodes;
>   	enum drm_mm_insert_mode mode;
> @@ -430,7 +440,7 @@ static void amdgpu_vram_mgr_del(struct ttm_mem_type_manager *man,
>   				struct ttm_mem_reg *mem)
>   {
>   	struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev);
> -	struct amdgpu_vram_mgr *mgr = man->priv;
> +	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
>   	struct drm_mm_node *nodes = mem->mm_node;
>   	uint64_t usage = 0, vis_usage = 0;
>   	unsigned pages = mem->num_pages;
> @@ -562,7 +572,7 @@ void amdgpu_vram_mgr_free_sgt(struct amdgpu_device *adev,
>    */
>   uint64_t amdgpu_vram_mgr_usage(struct ttm_mem_type_manager *man)
>   {
> -	struct amdgpu_vram_mgr *mgr = man->priv;
> +	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
>   
>   	return atomic64_read(&mgr->usage);
>   }
> @@ -576,7 +586,7 @@ uint64_t amdgpu_vram_mgr_usage(struct ttm_mem_type_manager *man)
>    */
>   uint64_t amdgpu_vram_mgr_vis_usage(struct ttm_mem_type_manager *man)
>   {
> -	struct amdgpu_vram_mgr *mgr = man->priv;
> +	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
>   
>   	return atomic64_read(&mgr->vis_usage);
>   }
> @@ -592,7 +602,7 @@ uint64_t amdgpu_vram_mgr_vis_usage(struct ttm_mem_type_manager *man)
>   static void amdgpu_vram_mgr_debug(struct ttm_mem_type_manager *man,
>   				  struct drm_printer *printer)
>   {
> -	struct amdgpu_vram_mgr *mgr = man->priv;
> +	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
>   
>   	spin_lock(&mgr->lock);
>   	drm_mm_print(&mgr->mm, printer);

_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel