[PATCH] drm/amdgpu: conditionally compile amdgpu's amdkfd files

Felix Kuehling felix.kuehling at amd.com
Wed May 16 21:09:59 UTC 2018


Hi Oded,

Thanks for working on this! The Makefile changes look good.

Instead of checking and calling function pointers in amdgpu_amdkfd_...
functions at runtime, couldn't you just define empty stub functions in
amdgpu_amdkfd.h if KFD is not enabled? I think that would make the code
shorter and remove the runtime overhead.

Regards,
  Felix


On 2018-05-16 06:09 AM, Oded Gabbay wrote:
> In case CONFIG_HSA_AMD is not chosen, there is no need to compile amdkfd
> files that reside inside amdgpu dirver. In addition, because amdkfd
> depends on x86_64 architecture and amdgpu is not, compiling amdkfd files
> under i386 architecture can cause compiler errors and warnings.
>
> This patch modifies amdgpu's makefile to build amdkfd files only if
> CONFIG_HSA_AMD is chosen. The only file to be compiled unconditionally
> is amdgpu_amdkfd.c
>
> Direct calls from amdgpu driver proper to functions in other
> amdgpu_amdkfd_*.c files were changed to calls to functions inside
> amdgpu_amdkfd.c. These functions call the original functions using a
> function pointer to allow compilation without the original functions.
>
> Signed-off-by: Oded Gabbay <oded.gabbay at gmail.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/Makefile              | 13 +++--
>  drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c       | 66 ++++++++++++++++++++++--
>  drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h       | 48 ++++++++++++-----
>  drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c |  8 ++-
>  drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c         |  2 +-
>  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c          |  2 +-
>  6 files changed, 112 insertions(+), 27 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile
> index f3002020df6c..1464dff1b151 100644
> --- a/drivers/gpu/drm/amd/amdgpu/Makefile
> +++ b/drivers/gpu/drm/amd/amdgpu/Makefile
> @@ -56,8 +56,7 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
>  
>  # add asic specific block
>  amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \
> -	ci_smc.o ci_dpm.o dce_v8_0.o gfx_v7_0.o cik_sdma.o uvd_v4_2.o vce_v2_0.o \
> -	amdgpu_amdkfd_gfx_v7.o
> +	ci_smc.o ci_dpm.o dce_v8_0.o gfx_v7_0.o cik_sdma.o uvd_v4_2.o vce_v2_0.o
>  
>  amdgpu-$(CONFIG_DRM_AMDGPU_SI)+= si.o gmc_v6_0.o gfx_v6_0.o si_ih.o si_dma.o dce_v6_0.o si_dpm.o si_smc.o
>  
> @@ -126,13 +125,21 @@ amdgpu-y += \
>  	vcn_v1_0.o
>  
>  # add amdkfd interfaces
> +amdgpu-y += amdgpu_amdkfd.o
> +
> +ifneq ($(CONFIG_HSA_AMD),)
>  amdgpu-y += \
> -	 amdgpu_amdkfd.o \
>  	 amdgpu_amdkfd_fence.o \
>  	 amdgpu_amdkfd_gpuvm.o \
>  	 amdgpu_amdkfd_gfx_v8.o \
>  	 amdgpu_amdkfd_gfx_v9.o
>  
> +ifneq ($(CONFIG_DRM_AMDGPU_CIK),)
> +amdgpu-y += amdgpu_amdkfd_gfx_v7.o
> +endif
> +
> +endif
> +
>  # add cgs
>  amdgpu-y += amdgpu_cgs.o
>  
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> index cd0e8f192e6a..930d27dd6e27 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> @@ -27,7 +27,21 @@
>  #include "amdgpu_gfx.h"
>  #include <linux/module.h>
>  
> +#if defined(CONFIG_HSA_AMD_MODULE) || defined(CONFIG_HSA_AMD)
> +static const struct amdgpu_amdkfd_if amdkfd_if = {
> +	.fence_check_mm = amdkfd_fence_check_mm,
> +	.unreserve_system_memory_limit = amdkfd_unreserve_system_memory_limit,
> +	.gpuvm_destroy_cb = amdkfd_gpuvm_destroy_cb,
> +	.to_kfd_fence = to_amdgpu_amdkfd_fence,
> +	.evict_userptr = amdkfd_evict_userptr,
> +	.gfx_7_get_functions = amdgpu_amdkfd_gfx_7_get_functions,
> +	.gfx_8_0_get_functions = amdgpu_amdkfd_gfx_8_0_get_functions,
> +	.gfx_9_0_get_functions = amdgpu_amdkfd_gfx_9_0_get_functions
> +};
> +#endif
> +
>  const struct kgd2kfd_calls *kgd2kfd;
> +const struct amdgpu_amdkfd_if *amdgpu_amdkfd_if;
>  bool (*kgd2kfd_init_p)(unsigned int, const struct kgd2kfd_calls**);
>  
>  static const unsigned int compute_vmid_bitmap = 0xFF00;
> @@ -50,15 +64,22 @@ int amdgpu_amdkfd_init(void)
>  		kgd2kfd = NULL;
>  	}
>  
> +
>  #elif defined(CONFIG_HSA_AMD)
> +
>  	ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd);
>  	if (ret)
>  		kgd2kfd = NULL;
>  
>  #else
> +	amdgpu_amdkfd_if = NULL;
>  	ret = -ENOENT;
>  #endif
> +
> +#if defined(CONFIG_HSA_AMD_MODULE) || defined(CONFIG_HSA_AMD)
> +	amdgpu_amdkfd_if = &amdkfd_if;
>  	amdgpu_amdkfd_gpuvm_init_mem_limits();
> +#endif
>  
>  	return ret;
>  }
> @@ -75,14 +96,14 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev)
>  {
>  	const struct kfd2kgd_calls *kfd2kgd;
>  
> -	if (!kgd2kfd)
> +	if ((!kgd2kfd) || (!amdgpu_amdkfd_if))
>  		return;
>  
>  	switch (adev->asic_type) {
>  #ifdef CONFIG_DRM_AMDGPU_CIK
>  	case CHIP_KAVERI:
>  	case CHIP_HAWAII:
> -		kfd2kgd = amdgpu_amdkfd_gfx_7_get_functions();
> +		kfd2kgd = amdgpu_amdkfd_if->gfx_7_get_functions();
>  		break;
>  #endif
>  	case CHIP_CARRIZO:
> @@ -90,11 +111,11 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev)
>  	case CHIP_FIJI:
>  	case CHIP_POLARIS10:
>  	case CHIP_POLARIS11:
> -		kfd2kgd = amdgpu_amdkfd_gfx_8_0_get_functions();
> +		kfd2kgd = amdgpu_amdkfd_if->gfx_8_0_get_functions();
>  		break;
>  	case CHIP_VEGA10:
>  	case CHIP_RAVEN:
> -		kfd2kgd = amdgpu_amdkfd_gfx_9_0_get_functions();
> +		kfd2kgd = amdgpu_amdkfd_if->gfx_9_0_get_functions();
>  		break;
>  	default:
>  		dev_dbg(adev->dev, "kfd not supported on this ASIC\n");
> @@ -458,3 +479,40 @@ bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid)
>  
>  	return false;
>  }
> +
> +bool amdgpu_amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm)
> +{
> +	if (!amdgpu_amdkfd_if)
> +		return false;
> +
> +	return amdgpu_amdkfd_if->fence_check_mm(f, mm);
> +}
> +
> +void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo)
> +{
> +	if (amdgpu_amdkfd_if)
> +		amdgpu_amdkfd_if->unreserve_system_memory_limit(bo);
> +}
> +
> +void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
> +					struct amdgpu_vm *vm)
> +{
> +	if (amdgpu_amdkfd_if)
> +		amdgpu_amdkfd_if->gpuvm_destroy_cb(adev, vm);
> +}
> +
> +struct amdgpu_amdkfd_fence *amdgpu_amdkfd_to_kfd_fence(struct dma_fence *f)
> +{
> +	if (!amdgpu_amdkfd_if)
> +		return NULL;
> +
> +	return amdgpu_amdkfd_if->to_kfd_fence(f);
> +}
> +
> +int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm)
> +{
> +	if (!amdgpu_amdkfd_if)
> +		return 0;
> +
> +	return amdgpu_amdkfd_if->evict_userptr(mem, mm);
> +}
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> index 12367a9951e8..d40480887d49 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> @@ -79,8 +79,8 @@ struct amdgpu_amdkfd_fence {
>  
>  struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context,
>  						       struct mm_struct *mm);
> -bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm);
> -struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f);
> +bool amdgpu_amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm);
> +struct amdgpu_amdkfd_fence *amdgpu_amdkfd_to_kfd_fence(struct dma_fence *f);
>  
>  struct amdkfd_process_info {
>  	/* List head of all VMs that belong to a KFD process */
> @@ -120,10 +120,6 @@ int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine,
>  				uint32_t vmid, uint64_t gpu_addr,
>  				uint32_t *ib_cmd, uint32_t ib_len);
>  
> -struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void);
> -struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void);
> -struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void);
> -
>  bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid);
>  
>  /* Shared API */
> @@ -156,14 +152,14 @@ uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd);
>  
>  /* GPUVM API */
>  int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, void **vm,
> -					  void **process_info,
> -					  struct dma_fence **ef);
> +					void **process_info,
> +					struct dma_fence **ef);
>  int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd,
> -					   struct file *filp,
> -					   void **vm, void **process_info,
> -					   struct dma_fence **ef);
> +					struct file *filp,
> +					void **vm, void **process_info,
> +					struct dma_fence **ef);
>  void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
> -				    struct amdgpu_vm *vm);
> +				struct amdgpu_vm *vm);
>  void amdgpu_amdkfd_gpuvm_destroy_process_vm(struct kgd_dev *kgd, void *vm);
>  uint32_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm);
>  int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
> @@ -181,9 +177,35 @@ int amdgpu_amdkfd_gpuvm_sync_memory(
>  int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd,
>  		struct kgd_mem *mem, void **kptr, uint64_t *size);
>  int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info,
> -					    struct dma_fence **ef);
> +					struct dma_fence **ef);
>  
>  void amdgpu_amdkfd_gpuvm_init_mem_limits(void);
>  void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo);
>  
> +/* Function pointers interface between files inside amdgpu, to allow exclusion
> + * of amdkfd files from compilation of amdgpu when amdkfd driver is not
> + * enabled
> + */
> +
> +bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm);
> +void amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo);
> +void amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, struct amdgpu_vm *vm);
> +struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f);
> +int amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm);
> +struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void);
> +struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void);
> +struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void);
> +
> +struct amdgpu_amdkfd_if {
> +	bool (*fence_check_mm)(struct dma_fence *f, struct mm_struct *mm);
> +	void (*unreserve_system_memory_limit)(struct amdgpu_bo *bo);
> +	void (*gpuvm_destroy_cb)(struct amdgpu_device *adev,
> +					struct amdgpu_vm *vm);
> +	struct amdgpu_amdkfd_fence* (*to_kfd_fence)(struct dma_fence *f);
> +	int (*evict_userptr)(struct kgd_mem *mem, struct mm_struct *mm);
> +	struct kfd2kgd_calls* (*gfx_7_get_functions)(void);
> +	struct kfd2kgd_calls* (*gfx_8_0_get_functions)(void);
> +	struct kfd2kgd_calls* (*gfx_9_0_get_functions)(void);
> +};
> +
>  #endif /* AMDGPU_AMDKFD_H_INCLUDED */
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> index 4a6515ad94f8..c1b9865a240a 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> @@ -166,7 +166,7 @@ static void unreserve_system_mem_limit(struct amdgpu_device *adev,
>  	spin_unlock(&kfd_mem_limit.mem_limit_lock);
>  }
>  
> -void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo)
> +void amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo)
>  {
>  	spin_lock(&kfd_mem_limit.mem_limit_lock);
>  
> @@ -1079,8 +1079,7 @@ int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd,
>  	return 0;
>  }
>  
> -void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
> -				    struct amdgpu_vm *vm)
> +void amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, struct amdgpu_vm *vm)
>  {
>  	struct amdkfd_process_info *process_info = vm->process_info;
>  	struct amdgpu_bo *pd = vm->root.base.bo;
> @@ -1625,8 +1624,7 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd,
>   * restore, where we get updated page addresses. This function only
>   * ensures that GPU access to the BO is stopped.
>   */
> -int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem,
> -				struct mm_struct *mm)
> +int amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm)
>  {
>  	struct amdkfd_process_info *process_info = mem->process_info;
>  	int invalid, evicted_bos;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
> index 2d6f5ec77a68..82472f080a32 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
> @@ -96,7 +96,7 @@ static void *amdgpu_sync_get_owner(struct dma_fence *f)
>  	if (s_fence)
>  		return s_fence->owner;
>  
> -	kfd_fence = to_amdgpu_amdkfd_fence(f);
> +	kfd_fence = amdgpu_amdkfd_to_kfd_fence(f);
>  	if (kfd_fence)
>  		return AMDGPU_FENCE_OWNER_KFD;
>  
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index c713d30cba86..256497940a6b 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -1219,7 +1219,7 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
>  		for (i = 0; i < flist->shared_count; ++i) {
>  			f = rcu_dereference_protected(flist->shared[i],
>  				reservation_object_held(bo->resv));
> -			if (amdkfd_fence_check_mm(f, current->mm))
> +			if (amdgpu_amdkfd_fence_check_mm(f, current->mm))
>  				return false;
>  		}
>  	}



More information about the amd-gfx mailing list