Re: [PATCH 7/7] drm/amdgpu: flush TLB functions removal from kfd2kgd interface

[Christian König <ckoenig.leichtzumerken@xxxxxxxxx>]

Am 11.01.20 um 19:39 schrieb Alex Sierra:
> [Why]
> kfd2kgd interface will be deprecated. This removal only covers TLB
> invalidation for now. They have been replaced in amdgpu_amdkfd API.
>
> [How]
> TLB invalidate functions removed from the different amdkfd_gfx_v*
> versions.
>
> Change-Id: Ic2c7d4a0d19fe1e884dee1ff10a520d31252afee
> Signed-off-by: Alex Sierra <alex.sierra@xxxxxxx>
> Reviewed-by: Felix Kuehling <Felix.Kuehling@xxxxxxx>

Nice work for the patch set. Acked-by: Christian König 
<christian.koenig@xxxxxxx>

> ---
>   .../drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c   |  2 -
>   .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c    | 67 -------------
>   .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 41 --------
>   .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 41 --------
>   .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 96 -------------------
>   .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h |  2 -
>   .../gpu/drm/amd/include/kgd_kfd_interface.h   |  2 -
>   7 files changed, 251 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
> index f9011a07cb90..562e7a7f51a8 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
> @@ -317,7 +317,5 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = {
>   			kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
>   	.get_tile_config = kgd_gfx_v9_get_tile_config,
>   	.set_vm_context_page_table_base = kgd_set_vm_context_page_table_base,
> -	.invalidate_tlbs = kgd_gfx_v9_invalidate_tlbs,
> -	.invalidate_tlbs_vmid = kgd_gfx_v9_invalidate_tlbs_vmid,
>   	.get_hive_id = amdgpu_amdkfd_get_hive_id,
>   };
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
> index 61cd707158e4..6132b4874498 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
> @@ -686,71 +686,6 @@ static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
>   	return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK);
>   }
>   
> -static int invalidate_tlbs_with_kiq(struct amdgpu_device *adev, uint16_t pasid)
> -{
> -	signed long r;
> -	uint32_t seq;
> -	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
> -
> -	spin_lock(&adev->gfx.kiq.ring_lock);
> -	amdgpu_ring_alloc(ring, 12); /* fence + invalidate_tlbs package*/
> -	amdgpu_ring_write(ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0));
> -	amdgpu_ring_write(ring,
> -			PACKET3_INVALIDATE_TLBS_DST_SEL(1) |
> -			PACKET3_INVALIDATE_TLBS_PASID(pasid));
> -	amdgpu_fence_emit_polling(ring, &seq);
> -	amdgpu_ring_commit(ring);
> -	spin_unlock(&adev->gfx.kiq.ring_lock);
> -
> -	r = amdgpu_fence_wait_polling(ring, seq, adev->usec_timeout);
> -	if (r < 1) {
> -		DRM_ERROR("wait for kiq fence error: %ld.\n", r);
> -		return -ETIME;
> -	}
> -
> -	return 0;
> -}
> -
> -static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)
> -{
> -	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
> -	int vmid;
> -	uint16_t queried_pasid;
> -	bool ret;
> -	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
> -
> -	if (amdgpu_emu_mode == 0 && ring->sched.ready)
> -		return invalidate_tlbs_with_kiq(adev, pasid);
> -
> -	for (vmid = 0; vmid < 16; vmid++) {
> -		if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid))
> -			continue;
> -
> -		ret = get_atc_vmid_pasid_mapping_info(kgd, vmid,
> -				&queried_pasid);
> -		if (ret	&& queried_pasid == pasid) {
> -			amdgpu_gmc_flush_gpu_tlb(adev, vmid,
> -					AMDGPU_GFXHUB_0, 0);
> -			break;
> -		}
> -	}
> -
> -	return 0;
> -}
> -
> -static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid)
> -{
> -	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
> -
> -	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
> -		pr_err("non kfd vmid %d\n", vmid);
> -		return 0;
> -	}
> -
> -	amdgpu_gmc_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB_0, 0);
> -	return 0;
> -}
> -
>   static int kgd_address_watch_disable(struct kgd_dev *kgd)
>   {
>   	return 0;
> @@ -832,7 +767,5 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
>   			get_atc_vmid_pasid_mapping_info,
>   	.get_tile_config = amdgpu_amdkfd_get_tile_config,
>   	.set_vm_context_page_table_base = set_vm_context_page_table_base,
> -	.invalidate_tlbs = invalidate_tlbs,
> -	.invalidate_tlbs_vmid = invalidate_tlbs_vmid,
>   	.get_hive_id = amdgpu_amdkfd_get_hive_id,
>   };
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
> index 6e6f0a99ec06..8f052e98a3c6 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
> @@ -696,45 +696,6 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
>   		lower_32_bits(page_table_base));
>   }
>   
> -static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)
> -{
> -	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
> -	int vmid;
> -	unsigned int tmp;
> -
> -	if (adev->in_gpu_reset)
> -		return -EIO;
> -
> -	for (vmid = 0; vmid < 16; vmid++) {
> -		if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid))
> -			continue;
> -
> -		tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
> -		if ((tmp & ATC_VMID0_PASID_MAPPING__VALID_MASK) &&
> -			(tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid) {
> -			WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
> -			RREG32(mmVM_INVALIDATE_RESPONSE);
> -			break;
> -		}
> -	}
> -
> -	return 0;
> -}
> -
> -static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid)
> -{
> -	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
> -
> -	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
> -		pr_err("non kfd vmid\n");
> -		return 0;
> -	}
> -
> -	WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
> -	RREG32(mmVM_INVALIDATE_RESPONSE);
> -	return 0;
> -}
> -
>    /**
>     * read_vmid_from_vmfault_reg - read vmid from register
>     *
> @@ -771,7 +732,5 @@ const struct kfd2kgd_calls gfx_v7_kfd2kgd = {
>   	.set_scratch_backing_va = set_scratch_backing_va,
>   	.get_tile_config = get_tile_config,
>   	.set_vm_context_page_table_base = set_vm_context_page_table_base,
> -	.invalidate_tlbs = invalidate_tlbs,
> -	.invalidate_tlbs_vmid = invalidate_tlbs_vmid,
>   	.read_vmid_from_vmfault_reg = read_vmid_from_vmfault_reg,
>   };
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
> index bfbddedb2380..19a10db93d68 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
> @@ -657,45 +657,6 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
>   			lower_32_bits(page_table_base));
>   }
>   
> -static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)
> -{
> -	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
> -	int vmid;
> -	unsigned int tmp;
> -
> -	if (adev->in_gpu_reset)
> -		return -EIO;
> -
> -	for (vmid = 0; vmid < 16; vmid++) {
> -		if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid))
> -			continue;
> -
> -		tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
> -		if ((tmp & ATC_VMID0_PASID_MAPPING__VALID_MASK) &&
> -			(tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid) {
> -			WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
> -			RREG32(mmVM_INVALIDATE_RESPONSE);
> -			break;
> -		}
> -	}
> -
> -	return 0;
> -}
> -
> -static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid)
> -{
> -	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
> -
> -	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
> -		pr_err("non kfd vmid %d\n", vmid);
> -		return -EINVAL;
> -	}
> -
> -	WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
> -	RREG32(mmVM_INVALIDATE_RESPONSE);
> -	return 0;
> -}
> -
>   const struct kfd2kgd_calls gfx_v8_kfd2kgd = {
>   	.program_sh_mem_settings = kgd_program_sh_mem_settings,
>   	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
> @@ -717,6 +678,4 @@ const struct kfd2kgd_calls gfx_v8_kfd2kgd = {
>   	.set_scratch_backing_va = set_scratch_backing_va,
>   	.get_tile_config = get_tile_config,
>   	.set_vm_context_page_table_base = set_vm_context_page_table_base,
> -	.invalidate_tlbs = invalidate_tlbs,
> -	.invalidate_tlbs_vmid = invalidate_tlbs_vmid,
>   };
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
> index e7861f0ef415..932ae85d97e2 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
> @@ -617,100 +617,6 @@ bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
>   	return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK);
>   }
>   
> -static int invalidate_tlbs_with_kiq(struct amdgpu_device *adev, uint16_t pasid,
> -			uint32_t flush_type)
> -{
> -	signed long r;
> -	uint32_t seq;
> -	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
> -
> -	spin_lock(&adev->gfx.kiq.ring_lock);
> -	amdgpu_ring_alloc(ring, 12); /* fence + invalidate_tlbs package*/
> -	amdgpu_ring_write(ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0));
> -	amdgpu_ring_write(ring,
> -			PACKET3_INVALIDATE_TLBS_DST_SEL(1) |
> -			PACKET3_INVALIDATE_TLBS_ALL_HUB(1) |
> -			PACKET3_INVALIDATE_TLBS_PASID(pasid) |
> -			PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type));
> -	amdgpu_fence_emit_polling(ring, &seq);
> -	amdgpu_ring_commit(ring);
> -	spin_unlock(&adev->gfx.kiq.ring_lock);
> -
> -	r = amdgpu_fence_wait_polling(ring, seq, adev->usec_timeout);
> -	if (r < 1) {
> -		DRM_ERROR("wait for kiq fence error: %ld.\n", r);
> -		return -ETIME;
> -	}
> -
> -	return 0;
> -}
> -
> -int kgd_gfx_v9_invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)
> -{
> -	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
> -	int vmid, i;
> -	uint16_t queried_pasid;
> -	bool ret;
> -	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
> -	uint32_t flush_type = 0;
> -
> -	if (adev->in_gpu_reset)
> -		return -EIO;
> -	if (adev->gmc.xgmi.num_physical_nodes &&
> -		adev->asic_type == CHIP_VEGA20)
> -		flush_type = 2;
> -
> -	if (ring->sched.ready)
> -		return invalidate_tlbs_with_kiq(adev, pasid, flush_type);
> -
> -	for (vmid = 0; vmid < 16; vmid++) {
> -		if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid))
> -			continue;
> -
> -		ret = kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(kgd, vmid,
> -				&queried_pasid);
> -		if (ret && queried_pasid == pasid) {
> -			for (i = 0; i < adev->num_vmhubs; i++)
> -				amdgpu_gmc_flush_gpu_tlb(adev, vmid,
> -							i, flush_type);
> -			break;
> -		}
> -	}
> -
> -	return 0;
> -}
> -
> -int kgd_gfx_v9_invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid)
> -{
> -	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
> -	int i;
> -
> -	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
> -		pr_err("non kfd vmid %d\n", vmid);
> -		return 0;
> -	}
> -
> -	/* Use legacy mode tlb invalidation.
> -	 *
> -	 * Currently on Raven the code below is broken for anything but
> -	 * legacy mode due to a MMHUB power gating problem. A workaround
> -	 * is for MMHUB to wait until the condition PER_VMID_INVALIDATE_REQ
> -	 * == PER_VMID_INVALIDATE_ACK instead of simply waiting for the ack
> -	 * bit.
> -	 *
> -	 * TODO 1: agree on the right set of invalidation registers for
> -	 * KFD use. Use the last one for now. Invalidate both GC and
> -	 * MMHUB.
> -	 *
> -	 * TODO 2: support range-based invalidation, requires kfg2kgd
> -	 * interface change
> -	 */
> -	for (i = 0; i < adev->num_vmhubs; i++)
> -		amdgpu_gmc_flush_gpu_tlb(adev, vmid, i, 0);
> -
> -	return 0;
> -}
> -
>   int kgd_gfx_v9_address_watch_disable(struct kgd_dev *kgd)
>   {
>   	return 0;
> @@ -793,7 +699,5 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
>   			kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
>   	.get_tile_config = kgd_gfx_v9_get_tile_config,
>   	.set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base,
> -	.invalidate_tlbs = kgd_gfx_v9_invalidate_tlbs,
> -	.invalidate_tlbs_vmid = kgd_gfx_v9_invalidate_tlbs_vmid,
>   	.get_hive_id = amdgpu_amdkfd_get_hive_id,
>   };
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
> index 02b1426d17d1..dfafa28b7559 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
> @@ -57,7 +57,5 @@ uint32_t kgd_gfx_v9_address_watch_get_offset(struct kgd_dev *kgd,
>   
>   bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
>   					uint8_t vmid, uint16_t *p_pasid);
> -int kgd_gfx_v9_invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid);
> -int kgd_gfx_v9_invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid);
>   int kgd_gfx_v9_get_tile_config(struct kgd_dev *kgd,
>   		struct tile_config *config);
> diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
> index 2cd217e60125..a01ef836ad58 100644
> --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
> +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
> @@ -307,8 +307,6 @@ struct kfd2kgd_calls {
>   
>   	void (*set_vm_context_page_table_base)(struct kgd_dev *kgd,
>   			uint32_t vmid, uint64_t page_table_base);
> -	int (*invalidate_tlbs)(struct kgd_dev *kgd, uint16_t pasid);
> -	int (*invalidate_tlbs_vmid)(struct kgd_dev *kgd, uint16_t vmid);
>   	uint32_t (*read_vmid_from_vmfault_reg)(struct kgd_dev *kgd);
>   	uint64_t (*get_hive_id)(struct kgd_dev *kgd);
>   

_______________________________________________
amd-gfx mailing list
amd-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/amd-gfx