[PATCH 01/13] drm/amdkfd: add amdgpu_device entry to kfd_dev

[PATCH 01/13] drm/amdkfd: add amdgpu_device entry to kfd_dev

[Graham Sider]

Patch series to remove kgd_dev struct and replace all instances with
amdgpu_device objects.

amdgpu_device needs to be declared in kgd_kfd_interface.h to be visible
to kfd2kgd_calls.

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 drivers/gpu/drm/amd/amdkfd/kfd_device.c         | 1 +
 drivers/gpu/drm/amd/amdkfd/kfd_priv.h           | 1 +
 drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 1 +
 3 files changed, 3 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 0fffaf859c59..81ca00d7b3da 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -825,6 +825,7 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf)
 		return NULL;
 
 	kfd->kgd = kgd;
+	kfd->adev = adev;
 	kfd->device_info = device_info;
 	kfd->pdev = pdev;
 	kfd->init_complete = false;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 6d8f9bb2d905..c8bd062fb954 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -229,6 +229,7 @@ struct kfd_vmid_info {
 
 struct kfd_dev {
 	struct kgd_dev *kgd;
+	struct amdgpu_device *adev;
 
 	const struct kfd_device_info *device_info;
 	struct pci_dev *pdev;
diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
index c84bd7b2cf59..ba444cbf9206 100644
--- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
+++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
@@ -33,6 +33,7 @@
 #include <linux/dma-fence.h>
 
 struct pci_dev;
+struct amdgpu_device;
 
 #define KGD_MAX_QUEUES 128
 
-- 
2.25.1

[PATCH 02/13] drm/amdkfd: replace kgd_dev in static gfx v7 funcs

[Graham Sider]

Static funcs in amdgpu_amdkfd_gfx_v7.c now using amdgpu_device.

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 51 +++++++++----------
 1 file changed, 23 insertions(+), 28 deletions(-)

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 b91d27e39bad..d00ba8d65a6d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
@@ -87,38 +87,33 @@ static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
 	return (struct amdgpu_device *)kgd;
 }
 
-static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
+static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue);
 
 	mutex_lock(&adev->srbm_mutex);
 	WREG32(mmSRBM_GFX_CNTL, value);
 }
 
-static void unlock_srbm(struct kgd_dev *kgd)
+static void unlock_srbm(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	WREG32(mmSRBM_GFX_CNTL, 0);
 	mutex_unlock(&adev->srbm_mutex);
 }
 
-static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
+static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, queue_id, 0);
+	lock_srbm(adev, mec, pipe, queue_id, 0);
 }
 
-static void release_queue(struct kgd_dev *kgd)
+static void release_queue(struct amdgpu_device *adev)
 {
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
@@ -129,14 +124,14 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32(mmSH_MEM_CONFIG, sh_mem_config);
 	WREG32(mmSH_MEM_APE1_BASE, sh_mem_ape1_base);
 	WREG32(mmSH_MEM_APE1_LIMIT, sh_mem_ape1_limit);
 	WREG32(mmSH_MEM_BASES, sh_mem_bases);
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
@@ -174,12 +169,12 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, 0, 0);
+	lock_srbm(adev, mec, pipe, 0, 0);
 
 	WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
 			CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 
 	return 0;
 }
@@ -220,7 +215,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 
 	m = get_mqd(mqd);
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	/* HQD registers extend from CP_MQD_BASE_ADDR to CP_MQD_CONTROL. */
 	mqd_hqd = &m->cp_mqd_base_addr_lo;
@@ -239,16 +234,16 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	 * release srbm_mutex to avoid circular dependency between
 	 * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex.
 	 */
-	release_queue(kgd);
+	release_queue(adev);
 	valid_wptr = read_user_wptr(mm, wptr, wptr_val);
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 	if (valid_wptr)
 		WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask);
 
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
 	WREG32(mmCP_HQD_ACTIVE, data);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	return 0;
 }
@@ -271,7 +266,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
 	if (*dump == NULL)
 		return -ENOMEM;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0);
 	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1);
@@ -281,7 +276,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
 	for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_MQD_CONTROL; reg++)
 		DUMP_REG(reg);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	WARN_ON_ONCE(i != HQD_N_REGS);
 	*n_regs = i;
@@ -380,7 +375,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 	bool retval = false;
 	uint32_t low, high;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 	act = RREG32(mmCP_HQD_ACTIVE);
 	if (act) {
 		low = lower_32_bits(queue_address >> 8);
@@ -390,7 +385,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 				high == RREG32(mmCP_HQD_PQ_BASE_HI))
 			retval = true;
 	}
-	release_queue(kgd);
+	release_queue(adev);
 	return retval;
 }
 
@@ -426,7 +421,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 	if (amdgpu_in_reset(adev))
 		return -EIO;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 	WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0);
 
 	switch (reset_type) {
@@ -504,13 +499,13 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 			break;
 		if (time_after(jiffies, end_jiffies)) {
 			pr_err("cp queue preemption time out\n");
-			release_queue(kgd);
+			release_queue(adev);
 			return -ETIME;
 		}
 		usleep_range(500, 1000);
 	}
 
-	release_queue(kgd);
+	release_queue(adev);
 	return 0;
 }
 
@@ -651,9 +646,9 @@ static void set_scratch_backing_va(struct kgd_dev *kgd,
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 	WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va);
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-- 
2.25.1

Re: [PATCH 02/13] drm/amdkfd: replace kgd_dev in static gfx v7 funcs

[Felix Kuehling]

Am 2021-10-19 um 5:13 p.m. schrieb Graham Sider:
> Static funcs in amdgpu_amdkfd_gfx_v7.c now using amdgpu_device.

Doesn't this cause pointer type mismatch errors when assigning the
function pointers in gfx_v7_kfd2kgd? Those only get updated in patch 7.

Regards,
  Felix


>
> Signed-off-by: Graham Sider <Graham.Sider@amd.com>
> ---
>  .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 51 +++++++++----------
>  1 file changed, 23 insertions(+), 28 deletions(-)
>
> 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 b91d27e39bad..d00ba8d65a6d 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
> @@ -87,38 +87,33 @@ static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
>  	return (struct amdgpu_device *)kgd;
>  }
>  
> -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
> +static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
>  			uint32_t queue, uint32_t vmid)
>  {
> -	struct amdgpu_device *adev = get_amdgpu_device(kgd);
>  	uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue);
>  
>  	mutex_lock(&adev->srbm_mutex);
>  	WREG32(mmSRBM_GFX_CNTL, value);
>  }
>  
> -static void unlock_srbm(struct kgd_dev *kgd)
> +static void unlock_srbm(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = get_amdgpu_device(kgd);
> -
>  	WREG32(mmSRBM_GFX_CNTL, 0);
>  	mutex_unlock(&adev->srbm_mutex);
>  }
>  
> -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
> +static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
>  				uint32_t queue_id)
>  {
> -	struct amdgpu_device *adev = get_amdgpu_device(kgd);
> -
>  	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
>  	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
>  
> -	lock_srbm(kgd, mec, pipe, queue_id, 0);
> +	lock_srbm(adev, mec, pipe, queue_id, 0);
>  }
>  
> -static void release_queue(struct kgd_dev *kgd)
> +static void release_queue(struct amdgpu_device *adev)
>  {
> -	unlock_srbm(kgd);
> +	unlock_srbm(adev);
>  }
>  
>  static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
> @@ -129,14 +124,14 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
>  {
>  	struct amdgpu_device *adev = get_amdgpu_device(kgd);
>  
> -	lock_srbm(kgd, 0, 0, 0, vmid);
> +	lock_srbm(adev, 0, 0, 0, vmid);
>  
>  	WREG32(mmSH_MEM_CONFIG, sh_mem_config);
>  	WREG32(mmSH_MEM_APE1_BASE, sh_mem_ape1_base);
>  	WREG32(mmSH_MEM_APE1_LIMIT, sh_mem_ape1_limit);
>  	WREG32(mmSH_MEM_BASES, sh_mem_bases);
>  
> -	unlock_srbm(kgd);
> +	unlock_srbm(adev);
>  }
>  
>  static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
> @@ -174,12 +169,12 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
>  	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
>  	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
>  
> -	lock_srbm(kgd, mec, pipe, 0, 0);
> +	lock_srbm(adev, mec, pipe, 0, 0);
>  
>  	WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
>  			CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
>  
> -	unlock_srbm(kgd);
> +	unlock_srbm(adev);
>  
>  	return 0;
>  }
> @@ -220,7 +215,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
>  
>  	m = get_mqd(mqd);
>  
> -	acquire_queue(kgd, pipe_id, queue_id);
> +	acquire_queue(adev, pipe_id, queue_id);
>  
>  	/* HQD registers extend from CP_MQD_BASE_ADDR to CP_MQD_CONTROL. */
>  	mqd_hqd = &m->cp_mqd_base_addr_lo;
> @@ -239,16 +234,16 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
>  	 * release srbm_mutex to avoid circular dependency between
>  	 * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex.
>  	 */
> -	release_queue(kgd);
> +	release_queue(adev);
>  	valid_wptr = read_user_wptr(mm, wptr, wptr_val);
> -	acquire_queue(kgd, pipe_id, queue_id);
> +	acquire_queue(adev, pipe_id, queue_id);
>  	if (valid_wptr)
>  		WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask);
>  
>  	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
>  	WREG32(mmCP_HQD_ACTIVE, data);
>  
> -	release_queue(kgd);
> +	release_queue(adev);
>  
>  	return 0;
>  }
> @@ -271,7 +266,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
>  	if (*dump == NULL)
>  		return -ENOMEM;
>  
> -	acquire_queue(kgd, pipe_id, queue_id);
> +	acquire_queue(adev, pipe_id, queue_id);
>  
>  	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0);
>  	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1);
> @@ -281,7 +276,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
>  	for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_MQD_CONTROL; reg++)
>  		DUMP_REG(reg);
>  
> -	release_queue(kgd);
> +	release_queue(adev);
>  
>  	WARN_ON_ONCE(i != HQD_N_REGS);
>  	*n_regs = i;
> @@ -380,7 +375,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
>  	bool retval = false;
>  	uint32_t low, high;
>  
> -	acquire_queue(kgd, pipe_id, queue_id);
> +	acquire_queue(adev, pipe_id, queue_id);
>  	act = RREG32(mmCP_HQD_ACTIVE);
>  	if (act) {
>  		low = lower_32_bits(queue_address >> 8);
> @@ -390,7 +385,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
>  				high == RREG32(mmCP_HQD_PQ_BASE_HI))
>  			retval = true;
>  	}
> -	release_queue(kgd);
> +	release_queue(adev);
>  	return retval;
>  }
>  
> @@ -426,7 +421,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
>  	if (amdgpu_in_reset(adev))
>  		return -EIO;
>  
> -	acquire_queue(kgd, pipe_id, queue_id);
> +	acquire_queue(adev, pipe_id, queue_id);
>  	WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0);
>  
>  	switch (reset_type) {
> @@ -504,13 +499,13 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
>  			break;
>  		if (time_after(jiffies, end_jiffies)) {
>  			pr_err("cp queue preemption time out\n");
> -			release_queue(kgd);
> +			release_queue(adev);
>  			return -ETIME;
>  		}
>  		usleep_range(500, 1000);
>  	}
>  
> -	release_queue(kgd);
> +	release_queue(adev);
>  	return 0;
>  }
>  
> @@ -651,9 +646,9 @@ static void set_scratch_backing_va(struct kgd_dev *kgd,
>  {
>  	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
>  
> -	lock_srbm(kgd, 0, 0, 0, vmid);
> +	lock_srbm(adev, 0, 0, 0, vmid);
>  	WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va);
> -	unlock_srbm(kgd);
> +	unlock_srbm(adev);
>  }
>  
>  static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,

RE: [PATCH 02/13] drm/amdkfd: replace kgd_dev in static gfx v7 funcs

[Sider, Graham]

[AMD Official Use Only]

> -----Original Message-----
> From: Kuehling, Felix <Felix.Kuehling@amd.com>
> Sent: Tuesday, October 26, 2021 4:07 PM
> To: Sider, Graham <Graham.Sider@amd.com>; amd-
> gfx@lists.freedesktop.org
> Cc: Joshi, Mukul <Mukul.Joshi@amd.com>
> Subject: Re: [PATCH 02/13] drm/amdkfd: replace kgd_dev in static gfx v7
> funcs
> 
> Am 2021-10-19 um 5:13 p.m. schrieb Graham Sider:
> > Static funcs in amdgpu_amdkfd_gfx_v7.c now using amdgpu_device.
> 
> Doesn't this cause pointer type mismatch errors when assigning the function
> pointers in gfx_v7_kfd2kgd? Those only get updated in patch 7.
> 
> Regards,
>   Felix
> 

The function definitions changed in patches 2 through 6 are only used internally
within these files and aren't assigned to any of the gfx_v*_kfd2kgd entries. Patches
7 through 11 deal with the kfd2kgd functions that would cause a mismatch if done
file-by-file.

Best,
Graham

> 
> >
> > Signed-off-by: Graham Sider <Graham.Sider@amd.com>
> > ---
> >  .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 51
> > +++++++++----------
> >  1 file changed, 23 insertions(+), 28 deletions(-)
> >
> > 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 b91d27e39bad..d00ba8d65a6d 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
> > @@ -87,38 +87,33 @@ static inline struct amdgpu_device
> *get_amdgpu_device(struct kgd_dev *kgd)
> >  	return (struct amdgpu_device *)kgd;
> >  }
> >
> > -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t
> > pipe,
> > +static void lock_srbm(struct amdgpu_device *adev, uint32_t mec,
> > +uint32_t pipe,
> >  			uint32_t queue, uint32_t vmid)
> >  {
> > -	struct amdgpu_device *adev = get_amdgpu_device(kgd);
> >  	uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) |
> > QUEUEID(queue);
> >
> >  	mutex_lock(&adev->srbm_mutex);
> >  	WREG32(mmSRBM_GFX_CNTL, value);
> >  }
> >
> > -static void unlock_srbm(struct kgd_dev *kgd)
> > +static void unlock_srbm(struct amdgpu_device *adev)
> >  {
> > -	struct amdgpu_device *adev = get_amdgpu_device(kgd);
> > -
> >  	WREG32(mmSRBM_GFX_CNTL, 0);
> >  	mutex_unlock(&adev->srbm_mutex);
> >  }
> >
> > -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
> > +static void acquire_queue(struct amdgpu_device *adev, uint32_t
> > +pipe_id,
> >  				uint32_t queue_id)
> >  {
> > -	struct amdgpu_device *adev = get_amdgpu_device(kgd);
> > -
> >  	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
> >  	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
> >
> > -	lock_srbm(kgd, mec, pipe, queue_id, 0);
> > +	lock_srbm(adev, mec, pipe, queue_id, 0);
> >  }
> >
> > -static void release_queue(struct kgd_dev *kgd)
> > +static void release_queue(struct amdgpu_device *adev)
> >  {
> > -	unlock_srbm(kgd);
> > +	unlock_srbm(adev);
> >  }
> >
> >  static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t
> > vmid, @@ -129,14 +124,14 @@ static void
> > kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,  {
> >  	struct amdgpu_device *adev = get_amdgpu_device(kgd);
> >
> > -	lock_srbm(kgd, 0, 0, 0, vmid);
> > +	lock_srbm(adev, 0, 0, 0, vmid);
> >
> >  	WREG32(mmSH_MEM_CONFIG, sh_mem_config);
> >  	WREG32(mmSH_MEM_APE1_BASE, sh_mem_ape1_base);
> >  	WREG32(mmSH_MEM_APE1_LIMIT, sh_mem_ape1_limit);
> >  	WREG32(mmSH_MEM_BASES, sh_mem_bases);
> >
> > -	unlock_srbm(kgd);
> > +	unlock_srbm(adev);
> >  }
> >
> >  static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
> > @@ -174,12 +169,12 @@ static int kgd_init_interrupts(struct kgd_dev
> *kgd, uint32_t pipe_id)
> >  	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
> >  	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
> >
> > -	lock_srbm(kgd, mec, pipe, 0, 0);
> > +	lock_srbm(adev, mec, pipe, 0, 0);
> >
> >  	WREG32(mmCPC_INT_CNTL,
> CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
> >
> 	CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
> >
> > -	unlock_srbm(kgd);
> > +	unlock_srbm(adev);
> >
> >  	return 0;
> >  }
> > @@ -220,7 +215,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void
> > *mqd, uint32_t pipe_id,
> >
> >  	m = get_mqd(mqd);
> >
> > -	acquire_queue(kgd, pipe_id, queue_id);
> > +	acquire_queue(adev, pipe_id, queue_id);
> >
> >  	/* HQD registers extend from CP_MQD_BASE_ADDR to
> CP_MQD_CONTROL. */
> >  	mqd_hqd = &m->cp_mqd_base_addr_lo;
> > @@ -239,16 +234,16 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void
> *mqd, uint32_t pipe_id,
> >  	 * release srbm_mutex to avoid circular dependency between
> >  	 * srbm_mutex->mm_sem->reservation_ww_class_mutex-
> >srbm_mutex.
> >  	 */
> > -	release_queue(kgd);
> > +	release_queue(adev);
> >  	valid_wptr = read_user_wptr(mm, wptr, wptr_val);
> > -	acquire_queue(kgd, pipe_id, queue_id);
> > +	acquire_queue(adev, pipe_id, queue_id);
> >  	if (valid_wptr)
> >  		WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) &
> wptr_mask);
> >
> >  	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE,
> 1);
> >  	WREG32(mmCP_HQD_ACTIVE, data);
> >
> > -	release_queue(kgd);
> > +	release_queue(adev);
> >
> >  	return 0;
> >  }
> > @@ -271,7 +266,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
> >  	if (*dump == NULL)
> >  		return -ENOMEM;
> >
> > -	acquire_queue(kgd, pipe_id, queue_id);
> > +	acquire_queue(adev, pipe_id, queue_id);
> >
> >  	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0);
> >  	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1);
> > @@ -281,7 +276,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
> >  	for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_MQD_CONTROL;
> reg++)
> >  		DUMP_REG(reg);
> >
> > -	release_queue(kgd);
> > +	release_queue(adev);
> >
> >  	WARN_ON_ONCE(i != HQD_N_REGS);
> >  	*n_regs = i;
> > @@ -380,7 +375,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev
> *kgd, uint64_t queue_address,
> >  	bool retval = false;
> >  	uint32_t low, high;
> >
> > -	acquire_queue(kgd, pipe_id, queue_id);
> > +	acquire_queue(adev, pipe_id, queue_id);
> >  	act = RREG32(mmCP_HQD_ACTIVE);
> >  	if (act) {
> >  		low = lower_32_bits(queue_address >> 8); @@ -390,7 +385,7
> @@ static
> > bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
> >  				high == RREG32(mmCP_HQD_PQ_BASE_HI))
> >  			retval = true;
> >  	}
> > -	release_queue(kgd);
> > +	release_queue(adev);
> >  	return retval;
> >  }
> >
> > @@ -426,7 +421,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd,
> void *mqd,
> >  	if (amdgpu_in_reset(adev))
> >  		return -EIO;
> >
> > -	acquire_queue(kgd, pipe_id, queue_id);
> > +	acquire_queue(adev, pipe_id, queue_id);
> >  	WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0);
> >
> >  	switch (reset_type) {
> > @@ -504,13 +499,13 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd,
> void *mqd,
> >  			break;
> >  		if (time_after(jiffies, end_jiffies)) {
> >  			pr_err("cp queue preemption time out\n");
> > -			release_queue(kgd);
> > +			release_queue(adev);
> >  			return -ETIME;
> >  		}
> >  		usleep_range(500, 1000);
> >  	}
> >
> > -	release_queue(kgd);
> > +	release_queue(adev);
> >  	return 0;
> >  }
> >
> > @@ -651,9 +646,9 @@ static void set_scratch_backing_va(struct kgd_dev
> > *kgd,  {
> >  	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
> >
> > -	lock_srbm(kgd, 0, 0, 0, vmid);
> > +	lock_srbm(adev, 0, 0, 0, vmid);
> >  	WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va);
> > -	unlock_srbm(kgd);
> > +	unlock_srbm(adev);
> >  }
> >
> >  static void set_vm_context_page_table_base(struct kgd_dev *kgd,
> > uint32_t vmid,

Re: [PATCH 02/13] drm/amdkfd: replace kgd_dev in static gfx v7 funcs

[Felix Kuehling]

Am 2021-10-26 um 4:31 p.m. schrieb Sider, Graham:
> [AMD Official Use Only]
>
>> -----Original Message-----
>> From: Kuehling, Felix <Felix.Kuehling@amd.com>
>> Sent: Tuesday, October 26, 2021 4:07 PM
>> To: Sider, Graham <Graham.Sider@amd.com>; amd-
>> gfx@lists.freedesktop.org
>> Cc: Joshi, Mukul <Mukul.Joshi@amd.com>
>> Subject: Re: [PATCH 02/13] drm/amdkfd: replace kgd_dev in static gfx v7
>> funcs
>>
>> Am 2021-10-19 um 5:13 p.m. schrieb Graham Sider:
>>> Static funcs in amdgpu_amdkfd_gfx_v7.c now using amdgpu_device.
>> Doesn't this cause pointer type mismatch errors when assigning the function
>> pointers in gfx_v7_kfd2kgd? Those only get updated in patch 7.
>>
>> Regards,
>>   Felix
>>
> The function definitions changed in patches 2 through 6 are only used internally
> within these files and aren't assigned to any of the gfx_v*_kfd2kgd entries. Patches
> 7 through 11 deal with the kfd2kgd functions that would cause a mismatch if done
> file-by-file.

I see. I saw you're changing e.g. kgd_hqd_load, but you're not changing
its signature here, so no problem.


>
> Best,
> Graham
>
>>> Signed-off-by: Graham Sider <Graham.Sider@amd.com>
>>> ---
>>>  .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 51
>>> +++++++++----------
>>>  1 file changed, 23 insertions(+), 28 deletions(-)
>>>
>>> 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 b91d27e39bad..d00ba8d65a6d 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
>>> @@ -87,38 +87,33 @@ static inline struct amdgpu_device
>> *get_amdgpu_device(struct kgd_dev *kgd)
>>>  	return (struct amdgpu_device *)kgd;
>>>  }
>>>
>>> -static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t
>>> pipe,
>>> +static void lock_srbm(struct amdgpu_device *adev, uint32_t mec,
>>> +uint32_t pipe,
>>>  			uint32_t queue, uint32_t vmid)
>>>  {
>>> -	struct amdgpu_device *adev = get_amdgpu_device(kgd);
>>>  	uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) |
>>> QUEUEID(queue);
>>>
>>>  	mutex_lock(&adev->srbm_mutex);
>>>  	WREG32(mmSRBM_GFX_CNTL, value);
>>>  }
>>>
>>> -static void unlock_srbm(struct kgd_dev *kgd)
>>> +static void unlock_srbm(struct amdgpu_device *adev)
>>>  {
>>> -	struct amdgpu_device *adev = get_amdgpu_device(kgd);
>>> -
>>>  	WREG32(mmSRBM_GFX_CNTL, 0);
>>>  	mutex_unlock(&adev->srbm_mutex);
>>>  }
>>>
>>> -static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
>>> +static void acquire_queue(struct amdgpu_device *adev, uint32_t
>>> +pipe_id,
>>>  				uint32_t queue_id)
>>>  {
>>> -	struct amdgpu_device *adev = get_amdgpu_device(kgd);
>>> -
>>>  	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
>>>  	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
>>>
>>> -	lock_srbm(kgd, mec, pipe, queue_id, 0);
>>> +	lock_srbm(adev, mec, pipe, queue_id, 0);
>>>  }
>>>
>>> -static void release_queue(struct kgd_dev *kgd)
>>> +static void release_queue(struct amdgpu_device *adev)
>>>  {
>>> -	unlock_srbm(kgd);
>>> +	unlock_srbm(adev);
>>>  }
>>>
>>>  static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t
>>> vmid, @@ -129,14 +124,14 @@ static void
>>> kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,  {
>>>  	struct amdgpu_device *adev = get_amdgpu_device(kgd);
>>>
>>> -	lock_srbm(kgd, 0, 0, 0, vmid);
>>> +	lock_srbm(adev, 0, 0, 0, vmid);
>>>
>>>  	WREG32(mmSH_MEM_CONFIG, sh_mem_config);
>>>  	WREG32(mmSH_MEM_APE1_BASE, sh_mem_ape1_base);
>>>  	WREG32(mmSH_MEM_APE1_LIMIT, sh_mem_ape1_limit);
>>>  	WREG32(mmSH_MEM_BASES, sh_mem_bases);
>>>
>>> -	unlock_srbm(kgd);
>>> +	unlock_srbm(adev);
>>>  }
>>>
>>>  static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
>>> @@ -174,12 +169,12 @@ static int kgd_init_interrupts(struct kgd_dev
>> *kgd, uint32_t pipe_id)
>>>  	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
>>>  	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
>>>
>>> -	lock_srbm(kgd, mec, pipe, 0, 0);
>>> +	lock_srbm(adev, mec, pipe, 0, 0);
>>>
>>>  	WREG32(mmCPC_INT_CNTL,
>> CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
>> 	CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
>>> -	unlock_srbm(kgd);
>>> +	unlock_srbm(adev);
>>>
>>>  	return 0;
>>>  }
>>> @@ -220,7 +215,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void
>>> *mqd, uint32_t pipe_id,
>>>
>>>  	m = get_mqd(mqd);
>>>
>>> -	acquire_queue(kgd, pipe_id, queue_id);
>>> +	acquire_queue(adev, pipe_id, queue_id);
>>>
>>>  	/* HQD registers extend from CP_MQD_BASE_ADDR to
>> CP_MQD_CONTROL. */
>>>  	mqd_hqd = &m->cp_mqd_base_addr_lo;
>>> @@ -239,16 +234,16 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void
>> *mqd, uint32_t pipe_id,
>>>  	 * release srbm_mutex to avoid circular dependency between
>>>  	 * srbm_mutex->mm_sem->reservation_ww_class_mutex-
>>> srbm_mutex.
>>>  	 */
>>> -	release_queue(kgd);
>>> +	release_queue(adev);
>>>  	valid_wptr = read_user_wptr(mm, wptr, wptr_val);
>>> -	acquire_queue(kgd, pipe_id, queue_id);
>>> +	acquire_queue(adev, pipe_id, queue_id);
>>>  	if (valid_wptr)
>>>  		WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) &
>> wptr_mask);
>>>  	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE,
>> 1);
>>>  	WREG32(mmCP_HQD_ACTIVE, data);
>>>
>>> -	release_queue(kgd);
>>> +	release_queue(adev);
>>>
>>>  	return 0;
>>>  }
>>> @@ -271,7 +266,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
>>>  	if (*dump == NULL)
>>>  		return -ENOMEM;
>>>
>>> -	acquire_queue(kgd, pipe_id, queue_id);
>>> +	acquire_queue(adev, pipe_id, queue_id);
>>>
>>>  	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0);
>>>  	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1);
>>> @@ -281,7 +276,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
>>>  	for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_MQD_CONTROL;
>> reg++)
>>>  		DUMP_REG(reg);
>>>
>>> -	release_queue(kgd);
>>> +	release_queue(adev);
>>>
>>>  	WARN_ON_ONCE(i != HQD_N_REGS);
>>>  	*n_regs = i;
>>> @@ -380,7 +375,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev
>> *kgd, uint64_t queue_address,
>>>  	bool retval = false;
>>>  	uint32_t low, high;
>>>
>>> -	acquire_queue(kgd, pipe_id, queue_id);
>>> +	acquire_queue(adev, pipe_id, queue_id);
>>>  	act = RREG32(mmCP_HQD_ACTIVE);
>>>  	if (act) {
>>>  		low = lower_32_bits(queue_address >> 8); @@ -390,7 +385,7
>> @@ static
>>> bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
>>>  				high == RREG32(mmCP_HQD_PQ_BASE_HI))
>>>  			retval = true;
>>>  	}
>>> -	release_queue(kgd);
>>> +	release_queue(adev);
>>>  	return retval;
>>>  }
>>>
>>> @@ -426,7 +421,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd,
>> void *mqd,
>>>  	if (amdgpu_in_reset(adev))
>>>  		return -EIO;
>>>
>>> -	acquire_queue(kgd, pipe_id, queue_id);
>>> +	acquire_queue(adev, pipe_id, queue_id);
>>>  	WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0);
>>>
>>>  	switch (reset_type) {
>>> @@ -504,13 +499,13 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd,
>> void *mqd,
>>>  			break;
>>>  		if (time_after(jiffies, end_jiffies)) {
>>>  			pr_err("cp queue preemption time out\n");
>>> -			release_queue(kgd);
>>> +			release_queue(adev);
>>>  			return -ETIME;
>>>  		}
>>>  		usleep_range(500, 1000);
>>>  	}
>>>
>>> -	release_queue(kgd);
>>> +	release_queue(adev);
>>>  	return 0;
>>>  }
>>>
>>> @@ -651,9 +646,9 @@ static void set_scratch_backing_va(struct kgd_dev
>>> *kgd,  {
>>>  	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
>>>
>>> -	lock_srbm(kgd, 0, 0, 0, vmid);
>>> +	lock_srbm(adev, 0, 0, 0, vmid);
>>>  	WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va);
>>> -	unlock_srbm(kgd);
>>> +	unlock_srbm(adev);
>>>  }
>>>
>>>  static void set_vm_context_page_table_base(struct kgd_dev *kgd,
>>> uint32_t vmid,

[PATCH 03/13] drm/amdkfd: replace kgd_dev in static gfx v8 funcs

[Graham Sider]

Static funcs in amdgpu_amdkfd_gfx_v8.c now using amdgpu_device.

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 51 +++++++++----------
 1 file changed, 23 insertions(+), 28 deletions(-)

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 5ce0ce704a21..06be6061e4c0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
@@ -44,38 +44,33 @@ static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
 	return (struct amdgpu_device *)kgd;
 }
 
-static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
+static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t value = PIPEID(pipe) | MEID(mec) | VMID(vmid) | QUEUEID(queue);
 
 	mutex_lock(&adev->srbm_mutex);
 	WREG32(mmSRBM_GFX_CNTL, value);
 }
 
-static void unlock_srbm(struct kgd_dev *kgd)
+static void unlock_srbm(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	WREG32(mmSRBM_GFX_CNTL, 0);
 	mutex_unlock(&adev->srbm_mutex);
 }
 
-static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
+static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, queue_id, 0);
+	lock_srbm(adev, mec, pipe, queue_id, 0);
 }
 
-static void release_queue(struct kgd_dev *kgd)
+static void release_queue(struct amdgpu_device *adev)
 {
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
@@ -86,14 +81,14 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32(mmSH_MEM_CONFIG, sh_mem_config);
 	WREG32(mmSH_MEM_APE1_BASE, sh_mem_ape1_base);
 	WREG32(mmSH_MEM_APE1_LIMIT, sh_mem_ape1_limit);
 	WREG32(mmSH_MEM_BASES, sh_mem_bases);
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
@@ -132,12 +127,12 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, 0, 0);
+	lock_srbm(adev, mec, pipe, 0, 0);
 
 	WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
 			CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 
 	return 0;
 }
@@ -178,7 +173,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 
 	m = get_mqd(mqd);
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	/* HIQ is set during driver init period with vmid set to 0*/
 	if (m->cp_hqd_vmid == 0) {
@@ -226,16 +221,16 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	 * release srbm_mutex to avoid circular dependency between
 	 * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex.
 	 */
-	release_queue(kgd);
+	release_queue(adev);
 	valid_wptr = read_user_wptr(mm, wptr, wptr_val);
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 	if (valid_wptr)
 		WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask);
 
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
 	WREG32(mmCP_HQD_ACTIVE, data);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	return 0;
 }
@@ -258,7 +253,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
 	if (*dump == NULL)
 		return -ENOMEM;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0);
 	DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1);
@@ -268,7 +263,7 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
 	for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_HQD_EOP_DONES; reg++)
 		DUMP_REG(reg);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	WARN_ON_ONCE(i != HQD_N_REGS);
 	*n_regs = i;
@@ -375,7 +370,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 	bool retval = false;
 	uint32_t low, high;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 	act = RREG32(mmCP_HQD_ACTIVE);
 	if (act) {
 		low = lower_32_bits(queue_address >> 8);
@@ -385,7 +380,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 				high == RREG32(mmCP_HQD_PQ_BASE_HI))
 			retval = true;
 	}
-	release_queue(kgd);
+	release_queue(adev);
 	return retval;
 }
 
@@ -422,7 +417,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 	if (amdgpu_in_reset(adev))
 		return -EIO;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	if (m->cp_hqd_vmid == 0)
 		WREG32_FIELD(RLC_CP_SCHEDULERS, scheduler1, 0);
@@ -502,13 +497,13 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 			break;
 		if (time_after(jiffies, end_jiffies)) {
 			pr_err("cp queue preemption time out.\n");
-			release_queue(kgd);
+			release_queue(adev);
 			return -ETIME;
 		}
 		usleep_range(500, 1000);
 	}
 
-	release_queue(kgd);
+	release_queue(adev);
 	return 0;
 }
 
@@ -612,9 +607,9 @@ static void set_scratch_backing_va(struct kgd_dev *kgd,
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 	WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va);
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-- 
2.25.1

[PATCH 04/13] drm/amdkfd: replace kgd_dev in static gfx v9 funcs

[Graham Sider]

Static funcs in amdgpu_amdkfd_gfx_v9.c now using amdgpu_device.

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 52 ++++++++-----------
 1 file changed, 23 insertions(+), 29 deletions(-)

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 bcc1cbeb8799..a79f4d110669 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -51,32 +51,26 @@ static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
 	return (struct amdgpu_device *)kgd;
 }
 
-static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
+static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	mutex_lock(&adev->srbm_mutex);
 	soc15_grbm_select(adev, mec, pipe, queue, vmid);
 }
 
-static void unlock_srbm(struct kgd_dev *kgd)
+static void unlock_srbm(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	soc15_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 }
 
-static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
+static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, queue_id, 0);
+	lock_srbm(adev, mec, pipe, queue_id, 0);
 }
 
 static uint64_t get_queue_mask(struct amdgpu_device *adev,
@@ -88,9 +82,9 @@ static uint64_t get_queue_mask(struct amdgpu_device *adev,
 	return 1ull << bit;
 }
 
-static void release_queue(struct kgd_dev *kgd)
+static void release_queue(struct amdgpu_device *adev)
 {
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 void kgd_gfx_v9_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
@@ -101,13 +95,13 @@ void kgd_gfx_v9_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), sh_mem_config);
 	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_BASES), sh_mem_bases);
 	/* APE1 no longer exists on GFX9 */
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 int kgd_gfx_v9_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
@@ -180,13 +174,13 @@ int kgd_gfx_v9_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, 0, 0);
+	lock_srbm(adev, mec, pipe, 0, 0);
 
 	WREG32(SOC15_REG_OFFSET(GC, 0, mmCPC_INT_CNTL),
 		CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
 		CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 
 	return 0;
 }
@@ -245,7 +239,7 @@ int kgd_gfx_v9_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 
 	m = get_mqd(mqd);
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	/* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */
 	mqd_hqd = &m->cp_mqd_base_addr_lo;
@@ -308,7 +302,7 @@ int kgd_gfx_v9_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
 	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE), data);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	return 0;
 }
@@ -325,7 +319,7 @@ int kgd_gfx_v9_hiq_mqd_load(struct kgd_dev *kgd, void *mqd,
 
 	m = get_mqd(mqd);
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
@@ -361,7 +355,7 @@ int kgd_gfx_v9_hiq_mqd_load(struct kgd_dev *kgd, void *mqd,
 
 out_unlock:
 	spin_unlock(&adev->gfx.kiq.ring_lock);
-	release_queue(kgd);
+	release_queue(adev);
 
 	return r;
 }
@@ -384,13 +378,13 @@ int kgd_gfx_v9_hqd_dump(struct kgd_dev *kgd,
 	if (*dump == NULL)
 		return -ENOMEM;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR);
 	     reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
 		DUMP_REG(reg);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	WARN_ON_ONCE(i != HQD_N_REGS);
 	*n_regs = i;
@@ -508,7 +502,7 @@ bool kgd_gfx_v9_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 	bool retval = false;
 	uint32_t low, high;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 	act = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE));
 	if (act) {
 		low = lower_32_bits(queue_address >> 8);
@@ -518,7 +512,7 @@ bool kgd_gfx_v9_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 		   high == RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_BASE_HI)))
 			retval = true;
 	}
-	release_queue(kgd);
+	release_queue(adev);
 	return retval;
 }
 
@@ -555,7 +549,7 @@ int kgd_gfx_v9_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 	if (amdgpu_in_reset(adev))
 		return -EIO;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	if (m->cp_hqd_vmid == 0)
 		WREG32_FIELD15_RLC(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0);
@@ -584,13 +578,13 @@ int kgd_gfx_v9_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 			break;
 		if (time_after(jiffies, end_jiffies)) {
 			pr_err("cp queue preemption time out.\n");
-			release_queue(kgd);
+			release_queue(adev);
 			return -ETIME;
 		}
 		usleep_range(500, 1000);
 	}
 
-	release_queue(kgd);
+	release_queue(adev);
 	return 0;
 }
 
@@ -887,7 +881,7 @@ void kgd_gfx_v9_program_trap_handler_settings(struct kgd_dev *kgd,
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 
 	/*
 	 * Program TBA registers
@@ -905,7 +899,7 @@ void kgd_gfx_v9_program_trap_handler_settings(struct kgd_dev *kgd,
 	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TMA_HI),
 			upper_32_bits(tma_addr >> 8));
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
-- 
2.25.1

[PATCH 05/13] drm/amdkfd: replace kgd_dev in static gfx v10 funcs

[Graham Sider]

Static funcs in amdgpu_amdkfd_gfx_v10.c now using amdgpu_device.

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c    | 52 ++++++++-----------
 1 file changed, 23 insertions(+), 29 deletions(-)

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 960acf68150a..69aee5b52f64 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -44,32 +44,26 @@ static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
 	return (struct amdgpu_device *)kgd;
 }
 
-static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
+static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	mutex_lock(&adev->srbm_mutex);
 	nv_grbm_select(adev, mec, pipe, queue, vmid);
 }
 
-static void unlock_srbm(struct kgd_dev *kgd)
+static void unlock_srbm(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	nv_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 }
 
-static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
+static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, queue_id, 0);
+	lock_srbm(adev, mec, pipe, queue_id, 0);
 }
 
 static uint64_t get_queue_mask(struct amdgpu_device *adev,
@@ -81,9 +75,9 @@ static uint64_t get_queue_mask(struct amdgpu_device *adev,
 	return 1ull << bit;
 }
 
-static void release_queue(struct kgd_dev *kgd)
+static void release_queue(struct amdgpu_device *adev)
 {
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
@@ -94,13 +88,13 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, sh_mem_config);
 	WREG32_SOC15(GC, 0, mmSH_MEM_BASES, sh_mem_bases);
 	/* APE1 no longer exists on GFX9 */
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
@@ -159,13 +153,13 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, 0, 0);
+	lock_srbm(adev, mec, pipe, 0, 0);
 
 	WREG32_SOC15(GC, 0, mmCPC_INT_CNTL,
 		CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
 		CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 
 	return 0;
 }
@@ -231,7 +225,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	m = get_mqd(mqd);
 
 	pr_debug("Load hqd of pipe %d queue %d\n", pipe_id, queue_id);
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	/* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */
 	mqd_hqd = &m->cp_mqd_base_addr_lo;
@@ -296,7 +290,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
 	WREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE, data);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	return 0;
 }
@@ -313,7 +307,7 @@ static int kgd_hiq_mqd_load(struct kgd_dev *kgd, void *mqd,
 
 	m = get_mqd(mqd);
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
@@ -349,7 +343,7 @@ static int kgd_hiq_mqd_load(struct kgd_dev *kgd, void *mqd,
 
 out_unlock:
 	spin_unlock(&adev->gfx.kiq.ring_lock);
-	release_queue(kgd);
+	release_queue(adev);
 
 	return r;
 }
@@ -372,13 +366,13 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
 	if (*dump == NULL)
 		return -ENOMEM;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR);
 	     reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
 		DUMP_REG(reg);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	WARN_ON_ONCE(i != HQD_N_REGS);
 	*n_regs = i;
@@ -496,7 +490,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 	bool retval = false;
 	uint32_t low, high;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 	act = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE);
 	if (act) {
 		low = lower_32_bits(queue_address >> 8);
@@ -506,7 +500,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 		   high == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI))
 			retval = true;
 	}
-	release_queue(kgd);
+	release_queue(adev);
 	return retval;
 }
 
@@ -548,7 +542,7 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 	int retry;
 #endif
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	if (m->cp_hqd_vmid == 0)
 		WREG32_FIELD15(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0);
@@ -633,13 +627,13 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 			break;
 		if (time_after(jiffies, end_jiffies)) {
 			pr_err("cp queue preemption time out.\n");
-			release_queue(kgd);
+			release_queue(adev);
 			return -ETIME;
 		}
 		usleep_range(500, 1000);
 	}
 
-	release_queue(kgd);
+	release_queue(adev);
 	return 0;
 }
 
@@ -762,7 +756,7 @@ static void program_trap_handler_settings(struct kgd_dev *kgd,
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 
 	/*
 	 * Program TBA registers
@@ -781,7 +775,7 @@ static void program_trap_handler_settings(struct kgd_dev *kgd,
 	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TMA_HI),
 			upper_32_bits(tma_addr >> 8));
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
-- 
2.25.1

[PATCH 06/13] drm/amdkfd: replace kgd_dev in static gfx v10_3 funcs

[Graham Sider]

Static funcs in amdgpu_amdkfd_gfx_v10_3.c now using amdgpu_device.

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c  | 52 ++++++++-----------
 1 file changed, 23 insertions(+), 29 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index dac0d751d5af..b33a9fe715cd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -43,32 +43,26 @@ static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
 	return (struct amdgpu_device *)kgd;
 }
 
-static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe,
+static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	mutex_lock(&adev->srbm_mutex);
 	nv_grbm_select(adev, mec, pipe, queue, vmid);
 }
 
-static void unlock_srbm(struct kgd_dev *kgd)
+static void unlock_srbm(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	nv_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 }
 
-static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id,
+static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, queue_id, 0);
+	lock_srbm(adev, mec, pipe, queue_id, 0);
 }
 
 static uint64_t get_queue_mask(struct amdgpu_device *adev,
@@ -80,9 +74,9 @@ static uint64_t get_queue_mask(struct amdgpu_device *adev,
 	return 1ull << bit;
 }
 
-static void release_queue(struct kgd_dev *kgd)
+static void release_queue(struct amdgpu_device *adev)
 {
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 static void program_sh_mem_settings_v10_3(struct kgd_dev *kgd, uint32_t vmid,
@@ -93,13 +87,13 @@ static void program_sh_mem_settings_v10_3(struct kgd_dev *kgd, uint32_t vmid,
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, sh_mem_config);
 	WREG32_SOC15(GC, 0, mmSH_MEM_BASES, sh_mem_bases);
 	/* APE1 no longer exists on GFX9 */
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 /* ATC is defeatured on Sienna_Cichlid */
@@ -127,13 +121,13 @@ static int init_interrupts_v10_3(struct kgd_dev *kgd, uint32_t pipe_id)
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(kgd, mec, pipe, 0, 0);
+	lock_srbm(adev, mec, pipe, 0, 0);
 
 	WREG32_SOC15(GC, 0, mmCPC_INT_CNTL,
 		CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
 		CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 
 	return 0;
 }
@@ -201,7 +195,7 @@ static int hqd_load_v10_3(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	m = get_mqd(mqd);
 
 	pr_debug("Load hqd of pipe %d queue %d\n", pipe_id, queue_id);
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	/* HIQ is set during driver init period with vmid set to 0*/
 	if (m->cp_hqd_vmid == 0) {
@@ -281,7 +275,7 @@ static int hqd_load_v10_3(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
 	WREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE, data);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	return 0;
 }
@@ -298,7 +292,7 @@ static int hiq_mqd_load_v10_3(struct kgd_dev *kgd, void *mqd,
 
 	m = get_mqd(mqd);
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
@@ -334,7 +328,7 @@ static int hiq_mqd_load_v10_3(struct kgd_dev *kgd, void *mqd,
 
 out_unlock:
 	spin_unlock(&adev->gfx.kiq.ring_lock);
-	release_queue(kgd);
+	release_queue(adev);
 
 	return r;
 }
@@ -357,13 +351,13 @@ static int hqd_dump_v10_3(struct kgd_dev *kgd,
 	if (*dump == NULL)
 		return -ENOMEM;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR);
 	     reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
 		DUMP_REG(reg);
 
-	release_queue(kgd);
+	release_queue(adev);
 
 	WARN_ON_ONCE(i != HQD_N_REGS);
 	*n_regs = i;
@@ -481,7 +475,7 @@ static bool hqd_is_occupied_v10_3(struct kgd_dev *kgd, uint64_t queue_address,
 	bool retval = false;
 	uint32_t low, high;
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 	act = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE);
 	if (act) {
 		low = lower_32_bits(queue_address >> 8);
@@ -491,7 +485,7 @@ static bool hqd_is_occupied_v10_3(struct kgd_dev *kgd, uint64_t queue_address,
 		   high == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI))
 			retval = true;
 	}
-	release_queue(kgd);
+	release_queue(adev);
 	return retval;
 }
 
@@ -525,7 +519,7 @@ static int hqd_destroy_v10_3(struct kgd_dev *kgd, void *mqd,
 	uint32_t temp;
 	struct v10_compute_mqd *m = get_mqd(mqd);
 
-	acquire_queue(kgd, pipe_id, queue_id);
+	acquire_queue(adev, pipe_id, queue_id);
 
 	if (m->cp_hqd_vmid == 0)
 		WREG32_FIELD15(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0);
@@ -555,13 +549,13 @@ static int hqd_destroy_v10_3(struct kgd_dev *kgd, void *mqd,
 		if (time_after(jiffies, end_jiffies)) {
 			pr_err("cp queue pipe %d queue %d preemption failed\n",
 					pipe_id, queue_id);
-			release_queue(kgd);
+			release_queue(adev);
 			return -ETIME;
 		}
 		usleep_range(500, 1000);
 	}
 
-	release_queue(kgd);
+	release_queue(adev);
 	return 0;
 }
 
@@ -666,7 +660,7 @@ static void program_trap_handler_settings_v10_3(struct kgd_dev *kgd,
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
-	lock_srbm(kgd, 0, 0, 0, vmid);
+	lock_srbm(adev, 0, 0, 0, vmid);
 
 	/*
 	 * Program TBA registers
@@ -685,7 +679,7 @@ static void program_trap_handler_settings_v10_3(struct kgd_dev *kgd,
 	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_SHADER_TMA_HI),
 			 upper_32_bits(tma_addr >> 8));
 
-	unlock_srbm(kgd);
+	unlock_srbm(adev);
 }
 
 #if 0
-- 
2.25.1

[PATCH 07/13] drm/amdkfd: replace kgd_dev in hqd/mqd kfd2kgd funcs

[Graham Sider]

Modified definitions:

- hqd_load
- hiq_mqd_load
- hqd_sdma_load
- hqd_dump
- hqd_sdma_dump
- hqd_is_occupied
- hqd_destroy
- hqd_sdma_is_occupied
- hqd_sdma_destroy
---
 .../drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c   | 13 +++----
 .../drm/amd/amdgpu/amdgpu_amdkfd_arcturus.h   |  9 +++--
 .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c    | 36 +++++++-----------
 .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c  | 37 ++++++++-----------
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 33 +++++++----------
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 35 +++++++-----------
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 36 +++++++-----------
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h | 13 ++++---
 .../drm/amd/amdkfd/kfd_device_queue_manager.c |  6 +--
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c  | 12 +++---
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c  | 14 +++----
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c   | 14 +++----
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c   | 12 +++---
 .../gpu/drm/amd/include/kgd_kfd_interface.h   | 25 +++++++------
 14 files changed, 129 insertions(+), 166 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index 5a7f680bcb3f..c2b8d970195b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -123,10 +123,9 @@ static uint32_t get_sdma_rlc_reg_offset(struct amdgpu_device *adev,
 	return sdma_rlc_reg_offset;
 }
 
-int kgd_arcturus_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
+int kgd_arcturus_hqd_sdma_load(struct amdgpu_device *adev, void *mqd,
 			     uint32_t __user *wptr, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v9_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	unsigned long end_jiffies;
@@ -193,11 +192,10 @@ int kgd_arcturus_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-int kgd_arcturus_hqd_sdma_dump(struct kgd_dev *kgd,
+int kgd_arcturus_hqd_sdma_dump(struct amdgpu_device *adev,
 			     uint32_t engine_id, uint32_t queue_id,
 			     uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev,
 			engine_id, queue_id);
 	uint32_t i = 0, reg;
@@ -225,9 +223,9 @@ int kgd_arcturus_hqd_sdma_dump(struct kgd_dev *kgd,
 	return 0;
 }
 
-bool kgd_arcturus_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
+bool kgd_arcturus_hqd_sdma_is_occupied(struct amdgpu_device *adev,
+				void *mqd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v9_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t sdma_rlc_rb_cntl;
@@ -244,10 +242,9 @@ bool kgd_arcturus_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
 	return false;
 }
 
-int kgd_arcturus_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
+int kgd_arcturus_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 				unsigned int utimeout)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v9_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t temp;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.h
index ce08131b7b5f..756c1a5679c0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.h
@@ -20,11 +20,12 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 
-int kgd_arcturus_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
+int kgd_arcturus_hqd_sdma_load(struct amdgpu_device *adev, void *mqd,
 			     uint32_t __user *wptr, struct mm_struct *mm);
-int kgd_arcturus_hqd_sdma_dump(struct kgd_dev *kgd,
+int kgd_arcturus_hqd_sdma_dump(struct amdgpu_device *adev,
 			     uint32_t engine_id, uint32_t queue_id,
 			     uint32_t (**dump)[2], uint32_t *n_regs);
-bool kgd_arcturus_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
-int kgd_arcturus_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
+bool kgd_arcturus_hqd_sdma_is_occupied(struct amdgpu_device *adev,
+				void *mqd);
+int kgd_arcturus_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 				unsigned int utimeout);
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 69aee5b52f64..5927a8fcbc23 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -212,12 +212,11 @@ static inline struct v10_sdma_mqd *get_sdma_mqd(void *mqd)
 	return (struct v10_sdma_mqd *)mqd;
 }
 
-static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
-			uint32_t queue_id, uint32_t __user *wptr,
-			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm)
+static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			uint32_t __user *wptr, uint32_t wptr_shift,
+			uint32_t wptr_mask, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v10_compute_mqd *m;
 	uint32_t *mqd_hqd;
 	uint32_t reg, hqd_base, data;
@@ -295,11 +294,10 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	return 0;
 }
 
-static int kgd_hiq_mqd_load(struct kgd_dev *kgd, void *mqd,
+static int kgd_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
 			    uint32_t doorbell_off)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
 	struct v10_compute_mqd *m;
 	uint32_t mec, pipe;
@@ -348,11 +346,10 @@ static int kgd_hiq_mqd_load(struct kgd_dev *kgd, void *mqd,
 	return r;
 }
 
-static int kgd_hqd_dump(struct kgd_dev *kgd,
+static int kgd_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
 #define DUMP_REG(addr) do {				\
@@ -380,10 +377,9 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_sdma_load(struct amdgpu_device *adev, void *mqd,
 			     uint32_t __user *wptr, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v10_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	unsigned long end_jiffies;
@@ -450,11 +446,10 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 			     uint32_t engine_id, uint32_t queue_id,
 			     uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev,
 			engine_id, queue_id);
 	uint32_t i = 0, reg;
@@ -482,10 +477,10 @@ static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
 	return 0;
 }
 
-static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id)
+static bool kgd_hqd_is_occupied(struct amdgpu_device *adev,
+				uint64_t queue_address, uint32_t pipe_id,
+				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t act;
 	bool retval = false;
 	uint32_t low, high;
@@ -504,9 +499,8 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 	return retval;
 }
 
-static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
+static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v10_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t sdma_rlc_rb_cntl;
@@ -523,12 +517,11 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
 	return false;
 }
 
-static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
 	uint32_t temp;
@@ -637,10 +630,9 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 				unsigned int utimeout)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v10_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t temp;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index b33a9fe715cd..2fc7dfc4fc92 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -182,12 +182,11 @@ static inline struct v10_sdma_mqd *get_sdma_mqd(void *mqd)
 	return (struct v10_sdma_mqd *)mqd;
 }
 
-static int hqd_load_v10_3(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
-			uint32_t queue_id, uint32_t __user *wptr,
-			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm)
+static int hqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			uint32_t __user *wptr, uint32_t wptr_shift,
+			uint32_t wptr_mask, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v10_compute_mqd *m;
 	uint32_t *mqd_hqd;
 	uint32_t reg, hqd_base, data;
@@ -280,11 +279,10 @@ static int hqd_load_v10_3(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	return 0;
 }
 
-static int hiq_mqd_load_v10_3(struct kgd_dev *kgd, void *mqd,
+static int hiq_mqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
 			    uint32_t doorbell_off)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
 	struct v10_compute_mqd *m;
 	uint32_t mec, pipe;
@@ -333,11 +331,10 @@ static int hiq_mqd_load_v10_3(struct kgd_dev *kgd, void *mqd,
 	return r;
 }
 
-static int hqd_dump_v10_3(struct kgd_dev *kgd,
+static int hqd_dump_v10_3(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
 #define DUMP_REG(addr) do {				\
@@ -365,10 +362,9 @@ static int hqd_dump_v10_3(struct kgd_dev *kgd,
 	return 0;
 }
 
-static int hqd_sdma_load_v10_3(struct kgd_dev *kgd, void *mqd,
+static int hqd_sdma_load_v10_3(struct amdgpu_device *adev, void *mqd,
 			     uint32_t __user *wptr, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v10_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	unsigned long end_jiffies;
@@ -435,11 +431,10 @@ static int hqd_sdma_load_v10_3(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int hqd_sdma_dump_v10_3(struct kgd_dev *kgd,
+static int hqd_sdma_dump_v10_3(struct amdgpu_device *adev,
 			     uint32_t engine_id, uint32_t queue_id,
 			     uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev,
 			engine_id, queue_id);
 	uint32_t i = 0, reg;
@@ -467,10 +462,10 @@ static int hqd_sdma_dump_v10_3(struct kgd_dev *kgd,
 	return 0;
 }
 
-static bool hqd_is_occupied_v10_3(struct kgd_dev *kgd, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id)
+static bool hqd_is_occupied_v10_3(struct amdgpu_device *adev,
+				uint64_t queue_address, uint32_t pipe_id,
+				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t act;
 	bool retval = false;
 	uint32_t low, high;
@@ -489,9 +484,9 @@ static bool hqd_is_occupied_v10_3(struct kgd_dev *kgd, uint64_t queue_address,
 	return retval;
 }
 
-static bool hqd_sdma_is_occupied_v10_3(struct kgd_dev *kgd, void *mqd)
+static bool hqd_sdma_is_occupied_v10_3(struct amdgpu_device *adev,
+				void *mqd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v10_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t sdma_rlc_rb_cntl;
@@ -508,12 +503,11 @@ static bool hqd_sdma_is_occupied_v10_3(struct kgd_dev *kgd, void *mqd)
 	return false;
 }
 
-static int hqd_destroy_v10_3(struct kgd_dev *kgd, void *mqd,
+static int hqd_destroy_v10_3(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
 	uint32_t temp;
@@ -559,10 +553,9 @@ static int hqd_destroy_v10_3(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int hqd_sdma_destroy_v10_3(struct kgd_dev *kgd, void *mqd,
+static int hqd_sdma_destroy_v10_3(struct amdgpu_device *adev, void *mqd,
 				unsigned int utimeout)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v10_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t temp;
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 d00ba8d65a6d..1190f58c4a99 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
@@ -202,12 +202,11 @@ static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd)
 	return (struct cik_sdma_rlc_registers *)mqd;
 }
 
-static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
-			uint32_t queue_id, uint32_t __user *wptr,
-			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm)
+static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			uint32_t __user *wptr, uint32_t wptr_shift,
+			uint32_t wptr_mask, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct cik_mqd *m;
 	uint32_t *mqd_hqd;
 	uint32_t reg, wptr_val, data;
@@ -248,11 +247,10 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	return 0;
 }
 
-static int kgd_hqd_dump(struct kgd_dev *kgd,
+static int kgd_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t i = 0, reg;
 #define HQD_N_REGS (35+4)
 #define DUMP_REG(addr) do {				\
@@ -284,10 +282,9 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_sdma_load(struct amdgpu_device *adev, void *mqd,
 			     uint32_t __user *wptr, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct cik_sdma_rlc_registers *m;
 	unsigned long end_jiffies;
 	uint32_t sdma_rlc_reg_offset;
@@ -340,11 +337,10 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 			     uint32_t engine_id, uint32_t queue_id,
 			     uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t sdma_offset = engine_id * SDMA1_REGISTER_OFFSET +
 		queue_id * KFD_CIK_SDMA_QUEUE_OFFSET;
 	uint32_t i = 0, reg;
@@ -367,10 +363,10 @@ static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
 	return 0;
 }
 
-static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id)
+static bool kgd_hqd_is_occupied(struct amdgpu_device *adev,
+				uint64_t queue_address, uint32_t pipe_id,
+				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t act;
 	bool retval = false;
 	uint32_t low, high;
@@ -389,9 +385,8 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 	return retval;
 }
 
-static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
+static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct cik_sdma_rlc_registers *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t sdma_rlc_rb_cntl;
@@ -407,12 +402,11 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
 	return false;
 }
 
-static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t temp;
 	enum hqd_dequeue_request_type type;
 	unsigned long flags, end_jiffies;
@@ -509,10 +503,9 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 				unsigned int utimeout)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct cik_sdma_rlc_registers *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t temp;
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 06be6061e4c0..5eb4e9bbe463 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
@@ -160,12 +160,11 @@ static inline struct vi_sdma_mqd *get_sdma_mqd(void *mqd)
 	return (struct vi_sdma_mqd *)mqd;
 }
 
-static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
-			uint32_t queue_id, uint32_t __user *wptr,
-			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm)
+static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			uint32_t __user *wptr, uint32_t wptr_shift,
+			uint32_t wptr_mask, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct vi_mqd *m;
 	uint32_t *mqd_hqd;
 	uint32_t reg, wptr_val, data;
@@ -201,7 +200,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	 * on ASICs that do not support context-save.
 	 * EOP writes/reads can start anywhere in the ring.
 	 */
-	if (get_amdgpu_device(kgd)->asic_type != CHIP_TONGA) {
+	if (adev->asic_type != CHIP_TONGA) {
 		WREG32(mmCP_HQD_EOP_RPTR, m->cp_hqd_eop_rptr);
 		WREG32(mmCP_HQD_EOP_WPTR, m->cp_hqd_eop_wptr);
 		WREG32(mmCP_HQD_EOP_WPTR_MEM, m->cp_hqd_eop_wptr_mem);
@@ -235,11 +234,10 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	return 0;
 }
 
-static int kgd_hqd_dump(struct kgd_dev *kgd,
+static int kgd_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t i = 0, reg;
 #define HQD_N_REGS (54+4)
 #define DUMP_REG(addr) do {				\
@@ -271,10 +269,9 @@ static int kgd_hqd_dump(struct kgd_dev *kgd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_sdma_load(struct amdgpu_device *adev, void *mqd,
 			     uint32_t __user *wptr, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct vi_sdma_mqd *m;
 	unsigned long end_jiffies;
 	uint32_t sdma_rlc_reg_offset;
@@ -326,11 +323,10 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 			     uint32_t engine_id, uint32_t queue_id,
 			     uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t sdma_offset = engine_id * SDMA1_REGISTER_OFFSET +
 		queue_id * KFD_VI_SDMA_QUEUE_OFFSET;
 	uint32_t i = 0, reg;
@@ -362,10 +358,10 @@ static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
 	return 0;
 }
 
-static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id)
+static bool kgd_hqd_is_occupied(struct amdgpu_device *adev,
+				uint64_t queue_address, uint32_t pipe_id,
+				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t act;
 	bool retval = false;
 	uint32_t low, high;
@@ -384,9 +380,8 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 	return retval;
 }
 
-static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
+static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct vi_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t sdma_rlc_rb_cntl;
@@ -402,12 +397,11 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
 	return false;
 }
 
-static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t temp;
 	enum hqd_dequeue_request_type type;
 	unsigned long flags, end_jiffies;
@@ -507,10 +501,9 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 				unsigned int utimeout)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct vi_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t temp;
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 a79f4d110669..480014ada833 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -227,12 +227,11 @@ static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd)
 	return (struct v9_sdma_mqd *)mqd;
 }
 
-int kgd_gfx_v9_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
-			uint32_t queue_id, uint32_t __user *wptr,
-			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm)
+int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			uint32_t __user *wptr, uint32_t wptr_shift,
+			uint32_t wptr_mask, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v9_mqd *m;
 	uint32_t *mqd_hqd;
 	uint32_t reg, hqd_base, data;
@@ -307,11 +306,10 @@ int kgd_gfx_v9_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	return 0;
 }
 
-int kgd_gfx_v9_hiq_mqd_load(struct kgd_dev *kgd, void *mqd,
+int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
 			    uint32_t doorbell_off)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
 	struct v9_mqd *m;
 	uint32_t mec, pipe;
@@ -360,11 +358,10 @@ int kgd_gfx_v9_hiq_mqd_load(struct kgd_dev *kgd, void *mqd,
 	return r;
 }
 
-int kgd_gfx_v9_hqd_dump(struct kgd_dev *kgd,
+int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
 #define DUMP_REG(addr) do {				\
@@ -392,10 +389,9 @@ int kgd_gfx_v9_hqd_dump(struct kgd_dev *kgd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_sdma_load(struct amdgpu_device *adev, void *mqd,
 			     uint32_t __user *wptr, struct mm_struct *mm)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v9_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	unsigned long end_jiffies;
@@ -462,11 +458,10 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 			     uint32_t engine_id, uint32_t queue_id,
 			     uint32_t (**dump)[2], uint32_t *n_regs)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev,
 			engine_id, queue_id);
 	uint32_t i = 0, reg;
@@ -494,10 +489,10 @@ static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
 	return 0;
 }
 
-bool kgd_gfx_v9_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id)
+bool kgd_gfx_v9_hqd_is_occupied(struct amdgpu_device *adev,
+				uint64_t queue_address, uint32_t pipe_id,
+				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t act;
 	bool retval = false;
 	uint32_t low, high;
@@ -516,9 +511,8 @@ bool kgd_gfx_v9_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
 	return retval;
 }
 
-static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
+static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v9_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t sdma_rlc_rb_cntl;
@@ -535,12 +529,11 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
 	return false;
 }
 
-int kgd_gfx_v9_hqd_destroy(struct kgd_dev *kgd, void *mqd,
+int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
 	uint32_t temp;
@@ -588,10 +581,9 @@ int kgd_gfx_v9_hqd_destroy(struct kgd_dev *kgd, void *mqd,
 	return 0;
 }
 
-static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd,
+static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 				unsigned int utimeout)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct v9_sdma_mqd *m;
 	uint32_t sdma_rlc_reg_offset;
 	uint32_t temp;
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 c63591106879..3fd99968f463 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -29,19 +29,20 @@ void kgd_gfx_v9_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 int kgd_gfx_v9_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
 		unsigned int vmid);
 int kgd_gfx_v9_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
-int kgd_gfx_v9_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
+int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
 			uint32_t wptr_shift, uint32_t wptr_mask,
 			struct mm_struct *mm);
-int kgd_gfx_v9_hiq_mqd_load(struct kgd_dev *kgd, void *mqd,
+int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
 			    uint32_t doorbell_off);
-int kgd_gfx_v9_hqd_dump(struct kgd_dev *kgd,
+int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t (**dump)[2], uint32_t *n_regs);
-bool kgd_gfx_v9_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
-		uint32_t pipe_id, uint32_t queue_id);
-int kgd_gfx_v9_hqd_destroy(struct kgd_dev *kgd, void *mqd,
+bool kgd_gfx_v9_hqd_is_occupied(struct amdgpu_device *adev,
+			uint64_t queue_address, uint32_t pipe_id,
+			uint32_t queue_id);
+int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index f8fce9d05f50..1919efe06f30 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -2064,7 +2064,7 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data)
 		return 0;
 	}
 
-	r = dqm->dev->kfd2kgd->hqd_dump(dqm->dev->kgd,
+	r = dqm->dev->kfd2kgd->hqd_dump(dqm->dev->adev,
 					KFD_CIK_HIQ_PIPE, KFD_CIK_HIQ_QUEUE,
 					&dump, &n_regs);
 	if (!r) {
@@ -2086,7 +2086,7 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data)
 				continue;
 
 			r = dqm->dev->kfd2kgd->hqd_dump(
-				dqm->dev->kgd, pipe, queue, &dump, &n_regs);
+				dqm->dev->adev, pipe, queue, &dump, &n_regs);
 			if (r)
 				break;
 
@@ -2103,7 +2103,7 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data)
 		     queue < dqm->dev->device_info->num_sdma_queues_per_engine;
 		     queue++) {
 			r = dqm->dev->kfd2kgd->hqd_sdma_dump(
-				dqm->dev->kgd, pipe, queue, &dump, &n_regs);
+				dqm->dev->adev, pipe, queue, &dump, &n_regs);
 			if (r)
 				break;
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
index 064914e1e8d6..58a71cb0e537 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
@@ -170,7 +170,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, uint32_t pipe_id,
 	uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0);
 	uint32_t wptr_mask = (uint32_t)((p->queue_size / 4) - 1);
 
-	return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id,
+	return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
 					  wptr_shift, wptr_mask, mms);
 }
@@ -179,7 +179,7 @@ static int load_mqd_sdma(struct mqd_manager *mm, void *mqd,
 			 uint32_t pipe_id, uint32_t queue_id,
 			 struct queue_properties *p, struct mm_struct *mms)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd,
+	return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->adev, mqd,
 					       (uint32_t __user *)p->write_ptr,
 					       mms);
 }
@@ -271,7 +271,7 @@ static int destroy_mqd(struct mqd_manager *mm, void *mqd,
 			unsigned int timeout, uint32_t pipe_id,
 			uint32_t queue_id)
 {
-	return mm->dev->kfd2kgd->hqd_destroy(mm->dev->kgd, mqd, type, timeout,
+	return mm->dev->kfd2kgd->hqd_destroy(mm->dev->adev, mqd, type, timeout,
 					pipe_id, queue_id);
 }
 
@@ -284,7 +284,7 @@ static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd,
 				unsigned int timeout, uint32_t pipe_id,
 				uint32_t queue_id)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout);
+	return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->adev, mqd, timeout);
 }
 
 static bool is_occupied(struct mqd_manager *mm, void *mqd,
@@ -292,7 +292,7 @@ static bool is_occupied(struct mqd_manager *mm, void *mqd,
 			uint32_t queue_id)
 {
 
-	return mm->dev->kfd2kgd->hqd_is_occupied(mm->dev->kgd, queue_address,
+	return mm->dev->kfd2kgd->hqd_is_occupied(mm->dev->adev, queue_address,
 					pipe_id, queue_id);
 
 }
@@ -301,7 +301,7 @@ static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd,
 			uint64_t queue_address,	uint32_t pipe_id,
 			uint32_t queue_id)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd);
+	return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->adev, mqd);
 }
 
 /*
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
index c7fb59ca597f..fe03f2a77fb6 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v10.c
@@ -147,7 +147,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd,
 	/* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */
 	uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0);
 
-	r = mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id,
+	r = mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
 					  wptr_shift, 0, mms);
 	return r;
@@ -157,7 +157,7 @@ static int hiq_load_mqd_kiq(struct mqd_manager *mm, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
 			    struct queue_properties *p, struct mm_struct *mms)
 {
-	return mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->kgd, mqd, pipe_id,
+	return mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->adev, mqd, pipe_id,
 					      queue_id, p->doorbell_off);
 }
 
@@ -237,7 +237,7 @@ static int destroy_mqd(struct mqd_manager *mm, void *mqd,
 		       uint32_t queue_id)
 {
 	return mm->dev->kfd2kgd->hqd_destroy
-		(mm->dev->kgd, mqd, type, timeout,
+		(mm->dev->adev, mqd, type, timeout,
 		 pipe_id, queue_id);
 }
 
@@ -252,7 +252,7 @@ static bool is_occupied(struct mqd_manager *mm, void *mqd,
 			uint32_t queue_id)
 {
 	return mm->dev->kfd2kgd->hqd_is_occupied(
-		mm->dev->kgd, queue_address,
+		mm->dev->adev, queue_address,
 		pipe_id, queue_id);
 }
 
@@ -318,7 +318,7 @@ static int load_mqd_sdma(struct mqd_manager *mm, void *mqd,
 		uint32_t pipe_id, uint32_t queue_id,
 		struct queue_properties *p, struct mm_struct *mms)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd,
+	return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->adev, mqd,
 					       (uint32_t __user *)p->write_ptr,
 					       mms);
 }
@@ -360,14 +360,14 @@ static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd,
 		unsigned int timeout, uint32_t pipe_id,
 		uint32_t queue_id)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout);
+	return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->adev, mqd, timeout);
 }
 
 static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd,
 		uint64_t queue_address, uint32_t pipe_id,
 		uint32_t queue_id)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd);
+	return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->adev, mqd);
 }
 
 #if defined(CONFIG_DEBUG_FS)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
index 7f4e102ff4bd..643d3eb18ebc 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
@@ -198,7 +198,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd,
 	/* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */
 	uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0);
 
-	return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id,
+	return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
 					  wptr_shift, 0, mms);
 }
@@ -207,7 +207,7 @@ static int hiq_load_mqd_kiq(struct mqd_manager *mm, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
 			    struct queue_properties *p, struct mm_struct *mms)
 {
-	return mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->kgd, mqd, pipe_id,
+	return mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->adev, mqd, pipe_id,
 					      queue_id, p->doorbell_off);
 }
 
@@ -289,7 +289,7 @@ static int destroy_mqd(struct mqd_manager *mm, void *mqd,
 			uint32_t queue_id)
 {
 	return mm->dev->kfd2kgd->hqd_destroy
-		(mm->dev->kgd, mqd, type, timeout,
+		(mm->dev->adev, mqd, type, timeout,
 		pipe_id, queue_id);
 }
 
@@ -311,7 +311,7 @@ static bool is_occupied(struct mqd_manager *mm, void *mqd,
 			uint32_t queue_id)
 {
 	return mm->dev->kfd2kgd->hqd_is_occupied(
-		mm->dev->kgd, queue_address,
+		mm->dev->adev, queue_address,
 		pipe_id, queue_id);
 }
 
@@ -373,7 +373,7 @@ static int load_mqd_sdma(struct mqd_manager *mm, void *mqd,
 		uint32_t pipe_id, uint32_t queue_id,
 		struct queue_properties *p, struct mm_struct *mms)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd,
+	return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->adev, mqd,
 					       (uint32_t __user *)p->write_ptr,
 					       mms);
 }
@@ -415,14 +415,14 @@ static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd,
 		unsigned int timeout, uint32_t pipe_id,
 		uint32_t queue_id)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout);
+	return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->adev, mqd, timeout);
 }
 
 static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd,
 		uint64_t queue_address, uint32_t pipe_id,
 		uint32_t queue_id)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd);
+	return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->adev, mqd);
 }
 
 #if defined(CONFIG_DEBUG_FS)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
index 33dbd22d290f..8b8151597ec6 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
@@ -161,7 +161,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd,
 	uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0);
 	uint32_t wptr_mask = (uint32_t)((p->queue_size / 4) - 1);
 
-	return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id,
+	return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
 					  wptr_shift, wptr_mask, mms);
 }
@@ -262,7 +262,7 @@ static int destroy_mqd(struct mqd_manager *mm, void *mqd,
 			uint32_t queue_id)
 {
 	return mm->dev->kfd2kgd->hqd_destroy
-		(mm->dev->kgd, mqd, type, timeout,
+		(mm->dev->adev, mqd, type, timeout,
 		pipe_id, queue_id);
 }
 
@@ -277,7 +277,7 @@ static bool is_occupied(struct mqd_manager *mm, void *mqd,
 			uint32_t queue_id)
 {
 	return mm->dev->kfd2kgd->hqd_is_occupied(
-		mm->dev->kgd, queue_address,
+		mm->dev->adev, queue_address,
 		pipe_id, queue_id);
 }
 
@@ -343,7 +343,7 @@ static int load_mqd_sdma(struct mqd_manager *mm, void *mqd,
 		uint32_t pipe_id, uint32_t queue_id,
 		struct queue_properties *p, struct mm_struct *mms)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd,
+	return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->adev, mqd,
 					       (uint32_t __user *)p->write_ptr,
 					       mms);
 }
@@ -384,14 +384,14 @@ static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd,
 		unsigned int timeout, uint32_t pipe_id,
 		uint32_t queue_id)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout);
+	return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->adev, mqd, timeout);
 }
 
 static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd,
 		uint64_t queue_address, uint32_t pipe_id,
 		uint32_t queue_id)
 {
-	return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd);
+	return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->adev, mqd);
 }
 
 #if defined(CONFIG_DEBUG_FS)
diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
index ba444cbf9206..75e40acdd699 100644
--- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
+++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
@@ -238,36 +238,37 @@ struct kfd2kgd_calls {
 
 	int (*init_interrupts)(struct kgd_dev *kgd, uint32_t pipe_id);
 
-	int (*hqd_load)(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
+	int (*hqd_load)(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
 			uint32_t wptr_shift, uint32_t wptr_mask,
 			struct mm_struct *mm);
 
-	int (*hiq_mqd_load)(struct kgd_dev *kgd, void *mqd,
+	int (*hiq_mqd_load)(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
 			    uint32_t doorbell_off);
 
-	int (*hqd_sdma_load)(struct kgd_dev *kgd, void *mqd,
+	int (*hqd_sdma_load)(struct amdgpu_device *adev, void *mqd,
 			     uint32_t __user *wptr, struct mm_struct *mm);
 
-	int (*hqd_dump)(struct kgd_dev *kgd,
+	int (*hqd_dump)(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t (**dump)[2], uint32_t *n_regs);
 
-	int (*hqd_sdma_dump)(struct kgd_dev *kgd,
+	int (*hqd_sdma_dump)(struct amdgpu_device *adev,
 			     uint32_t engine_id, uint32_t queue_id,
 			     uint32_t (**dump)[2], uint32_t *n_regs);
 
-	bool (*hqd_is_occupied)(struct kgd_dev *kgd, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id);
-
-	int (*hqd_destroy)(struct kgd_dev *kgd, void *mqd, uint32_t reset_type,
-				unsigned int timeout, uint32_t pipe_id,
+	bool (*hqd_is_occupied)(struct amdgpu_device *adev,
+				uint64_t queue_address, uint32_t pipe_id,
 				uint32_t queue_id);
 
-	bool (*hqd_sdma_is_occupied)(struct kgd_dev *kgd, void *mqd);
+	int (*hqd_destroy)(struct amdgpu_device *adev, void *mqd,
+				uint32_t reset_type, unsigned int timeout,
+				uint32_t pipe_id, uint32_t queue_id);
+
+	bool (*hqd_sdma_is_occupied)(struct amdgpu_device *adev, void *mqd);
 
-	int (*hqd_sdma_destroy)(struct kgd_dev *kgd, void *mqd,
+	int (*hqd_sdma_destroy)(struct amdgpu_device *adev, void *mqd,
 				unsigned int timeout);
 
 	int (*address_watch_disable)(struct kgd_dev *kgd);
-- 
2.25.1

[PATCH 08/13] drm/amdkfd: replace kgd_dev in various kfd2kgd funcs

[Graham Sider]

Modified definitions:

- program_sh_mem_settings
- set_pasid_vmid_mapping
- init_interrupts
- address_watch_disable
- address_watch_execute
- wave_control_execute
- address_watch_get_offset
- get_atc_vmid_pasid_mapping_info
- set_scratch_backing_va
- set_vm_context_page_table_base
- read_vmid_from_vmfault_reg
- get_cu_occupancy
- program_trap_handler_settings

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c    | 33 +++++--------
 .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c  | 49 ++++++-------------
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 39 +++++----------
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 33 +++++--------
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 35 +++++--------
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h | 22 ++++-----
 .../gpu/drm/amd/amdkfd/cik_event_interrupt.c  |  4 +-
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c      |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c       | 18 +++----
 .../drm/amd/amdkfd/kfd_device_queue_manager.c | 14 +++---
 drivers/gpu/drm/amd/amdkfd/kfd_process.c      |  2 +-
 .../gpu/drm/amd/include/kgd_kfd_interface.h   | 29 ++++++-----
 12 files changed, 106 insertions(+), 174 deletions(-)

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 5927a8fcbc23..5f274b7c4121 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -80,14 +80,12 @@ static void release_queue(struct amdgpu_device *adev)
 	unlock_srbm(adev);
 }
 
-static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
+static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
 					uint32_t sh_mem_bases)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, sh_mem_config);
@@ -97,11 +95,9 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 	unlock_srbm(adev);
 }
 
-static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
+static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
 					unsigned int vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	/*
 	 * We have to assume that there is no outstanding mapping.
 	 * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because
@@ -144,9 +140,8 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
  * but still works
  */
 
-static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
+static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t mec;
 	uint32_t pipe;
 
@@ -669,11 +664,10 @@ static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 	return 0;
 }
 
-static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
+static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 					uint8_t vmid, uint16_t *p_pasid)
 {
 	uint32_t value;
-	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
 
 	value = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID0_PASID_MAPPING)
 		     + vmid);
@@ -682,12 +676,12 @@ static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
 	return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK);
 }
 
-static int kgd_address_watch_disable(struct kgd_dev *kgd)
+static int kgd_address_watch_disable(struct amdgpu_device *adev)
 {
 	return 0;
 }
 
-static int kgd_address_watch_execute(struct kgd_dev *kgd,
+static int kgd_address_watch_execute(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					uint32_t cntl_val,
 					uint32_t addr_hi,
@@ -696,11 +690,10 @@ static int kgd_address_watch_execute(struct kgd_dev *kgd,
 	return 0;
 }
 
-static int kgd_wave_control_execute(struct kgd_dev *kgd,
+static int kgd_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
 					uint32_t sq_cmd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t data = 0;
 
 	mutex_lock(&adev->grbm_idx_mutex);
@@ -721,18 +714,16 @@ static int kgd_wave_control_execute(struct kgd_dev *kgd,
 	return 0;
 }
 
-static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
+static uint32_t kgd_address_watch_get_offset(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					unsigned int reg_offset)
 {
 	return 0;
 }
 
-static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint64_t page_table_base)
+static void set_vm_context_page_table_base(struct amdgpu_device *adev,
+		uint32_t vmid, uint64_t page_table_base)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
 		pr_err("trying to set page table base for wrong VMID %u\n",
 		       vmid);
@@ -743,11 +734,9 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
 	adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base);
 }
 
-static void program_trap_handler_settings(struct kgd_dev *kgd,
+static void program_trap_handler_settings(struct amdgpu_device *adev,
 		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 
 	/*
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index 2fc7dfc4fc92..980430974aca 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -79,14 +79,12 @@ static void release_queue(struct amdgpu_device *adev)
 	unlock_srbm(adev);
 }
 
-static void program_sh_mem_settings_v10_3(struct kgd_dev *kgd, uint32_t vmid,
+static void program_sh_mem_settings_v10_3(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
 					uint32_t sh_mem_bases)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, sh_mem_config);
@@ -97,11 +95,9 @@ static void program_sh_mem_settings_v10_3(struct kgd_dev *kgd, uint32_t vmid,
 }
 
 /* ATC is defeatured on Sienna_Cichlid */
-static int set_pasid_vmid_mapping_v10_3(struct kgd_dev *kgd, unsigned int pasid,
+static int set_pasid_vmid_mapping_v10_3(struct amdgpu_device *adev, unsigned int pasid,
 					unsigned int vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	uint32_t value = pasid << IH_VMID_0_LUT__PASID__SHIFT;
 
 	/* Mapping vmid to pasid also for IH block */
@@ -112,9 +108,8 @@ static int set_pasid_vmid_mapping_v10_3(struct kgd_dev *kgd, unsigned int pasid,
 	return 0;
 }
 
-static int init_interrupts_v10_3(struct kgd_dev *kgd, uint32_t pipe_id)
+static int init_interrupts_v10_3(struct amdgpu_device *adev, uint32_t pipe_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t mec;
 	uint32_t pipe;
 
@@ -593,12 +588,12 @@ static int hqd_sdma_destroy_v10_3(struct amdgpu_device *adev, void *mqd,
 }
 
 
-static int address_watch_disable_v10_3(struct kgd_dev *kgd)
+static int address_watch_disable_v10_3(struct amdgpu_device *adev)
 {
 	return 0;
 }
 
-static int address_watch_execute_v10_3(struct kgd_dev *kgd,
+static int address_watch_execute_v10_3(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					uint32_t cntl_val,
 					uint32_t addr_hi,
@@ -607,11 +602,10 @@ static int address_watch_execute_v10_3(struct kgd_dev *kgd,
 	return 0;
 }
 
-static int wave_control_execute_v10_3(struct kgd_dev *kgd,
+static int wave_control_execute_v10_3(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
 					uint32_t sq_cmd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t data = 0;
 
 	mutex_lock(&adev->grbm_idx_mutex);
@@ -632,27 +626,23 @@ static int wave_control_execute_v10_3(struct kgd_dev *kgd,
 	return 0;
 }
 
-static uint32_t address_watch_get_offset_v10_3(struct kgd_dev *kgd,
+static uint32_t address_watch_get_offset_v10_3(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					unsigned int reg_offset)
 {
 	return 0;
 }
 
-static void set_vm_context_page_table_base_v10_3(struct kgd_dev *kgd, uint32_t vmid,
-		uint64_t page_table_base)
+static void set_vm_context_page_table_base_v10_3(struct amdgpu_device *adev,
+		uint32_t vmid, uint64_t page_table_base)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	/* SDMA is on gfxhub as well for Navi1* series */
 	adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base);
 }
 
-static void program_trap_handler_settings_v10_3(struct kgd_dev *kgd,
+static void program_trap_handler_settings_v10_3(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 
 	/*
@@ -676,11 +666,10 @@ static void program_trap_handler_settings_v10_3(struct kgd_dev *kgd,
 }
 
 #if 0
-uint32_t enable_debug_trap_v10_3(struct kgd_dev *kgd,
+uint32_t enable_debug_trap_v10_3(struct amdgpu_device *adev,
 				uint32_t trap_debug_wave_launch_mode,
 				uint32_t vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t data = 0;
 	uint32_t orig_wave_cntl_value;
 	uint32_t orig_stall_vmid;
@@ -707,10 +696,8 @@ uint32_t enable_debug_trap_v10_3(struct kgd_dev *kgd,
 	return 0;
 }
 
-uint32_t disable_debug_trap_v10_3(struct kgd_dev *kgd)
+uint32_t disable_debug_trap_v10_3(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	mutex_lock(&adev->grbm_idx_mutex);
 
 	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
@@ -720,11 +707,10 @@ uint32_t disable_debug_trap_v10_3(struct kgd_dev *kgd)
 	return 0;
 }
 
-uint32_t set_wave_launch_trap_override_v10_3(struct kgd_dev *kgd,
+uint32_t set_wave_launch_trap_override_v10_3(struct amdgpu_device *adev,
 						uint32_t trap_override,
 						uint32_t trap_mask)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t data = 0;
 
 	mutex_lock(&adev->grbm_idx_mutex);
@@ -749,11 +735,10 @@ uint32_t set_wave_launch_trap_override_v10_3(struct kgd_dev *kgd,
 	return 0;
 }
 
-uint32_t set_wave_launch_mode_v10_3(struct kgd_dev *kgd,
+uint32_t set_wave_launch_mode_v10_3(struct amdgpu_device *adev,
 					uint8_t wave_launch_mode,
 					uint32_t vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t data = 0;
 	bool is_stall_mode;
 	bool is_mode_set;
@@ -792,16 +777,14 @@ uint32_t set_wave_launch_mode_v10_3(struct kgd_dev *kgd,
  *	sem_rearm_wait_time      -- Wait Count for Semaphore re-arm.
  *	deq_retry_wait_time      -- Wait Count for Global Wave Syncs.
  */
-void get_iq_wait_times_v10_3(struct kgd_dev *kgd,
+void get_iq_wait_times_v10_3(struct amdgpu_device *adev,
 					uint32_t *wait_times)
 
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	*wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2));
 }
 
-void build_grace_period_packet_info_v10_3(struct kgd_dev *kgd,
+void build_grace_period_packet_info_v10_3(struct amdgpu_device *adev,
 						uint32_t wait_times,
 						uint32_t grace_period,
 						uint32_t *reg_offset,
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 1190f58c4a99..e31b03495db4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
@@ -116,14 +116,12 @@ static void release_queue(struct amdgpu_device *adev)
 	unlock_srbm(adev);
 }
 
-static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
+static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
 					uint32_t sh_mem_bases)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32(mmSH_MEM_CONFIG, sh_mem_config);
@@ -134,11 +132,9 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 	unlock_srbm(adev);
 }
 
-static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
+static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
 					unsigned int vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	/*
 	 * We have to assume that there is no outstanding mapping.
 	 * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because
@@ -160,9 +156,8 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
 	return 0;
 }
 
-static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
+static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t mec;
 	uint32_t pipe;
 
@@ -539,9 +534,8 @@ static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 	return 0;
 }
 
-static int kgd_address_watch_disable(struct kgd_dev *kgd)
+static int kgd_address_watch_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	union TCP_WATCH_CNTL_BITS cntl;
 	unsigned int i;
 
@@ -559,13 +553,12 @@ static int kgd_address_watch_disable(struct kgd_dev *kgd)
 	return 0;
 }
 
-static int kgd_address_watch_execute(struct kgd_dev *kgd,
+static int kgd_address_watch_execute(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					uint32_t cntl_val,
 					uint32_t addr_hi,
 					uint32_t addr_lo)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	union TCP_WATCH_CNTL_BITS cntl;
 
 	cntl.u32All = cntl_val;
@@ -590,11 +583,10 @@ static int kgd_address_watch_execute(struct kgd_dev *kgd,
 	return 0;
 }
 
-static int kgd_wave_control_execute(struct kgd_dev *kgd,
+static int kgd_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
 					uint32_t sq_cmd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t data;
 
 	mutex_lock(&adev->grbm_idx_mutex);
@@ -615,18 +607,17 @@ static int kgd_wave_control_execute(struct kgd_dev *kgd,
 	return 0;
 }
 
-static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
+static uint32_t kgd_address_watch_get_offset(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					unsigned int reg_offset)
 {
 	return watchRegs[watch_point_id * ADDRESS_WATCH_REG_MAX + reg_offset];
 }
 
-static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
+static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 					uint8_t vmid, uint16_t *p_pasid)
 {
 	uint32_t value;
-	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
 
 	value = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 	*p_pasid = value & ATC_VMID0_PASID_MAPPING__PASID_MASK;
@@ -634,21 +625,17 @@ static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
 	return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK);
 }
 
-static void set_scratch_backing_va(struct kgd_dev *kgd,
+static void set_scratch_backing_va(struct amdgpu_device *adev,
 					uint64_t va, uint32_t vmid)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 	WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va);
 	unlock_srbm(adev);
 }
 
-static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-			uint64_t page_table_base)
+static void set_vm_context_page_table_base(struct amdgpu_device *adev,
+			uint32_t vmid, uint64_t page_table_base)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
 		pr_err("trying to set page table base for wrong VMID\n");
 		return;
@@ -664,10 +651,8 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
   * @vmid: vmid pointer
   * read vmid from register (CIK).
   */
-static uint32_t read_vmid_from_vmfault_reg(struct kgd_dev *kgd)
+static uint32_t read_vmid_from_vmfault_reg(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	uint32_t status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS);
 
 	return REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
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 5eb4e9bbe463..9a30a4f4f098 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
@@ -73,14 +73,12 @@ static void release_queue(struct amdgpu_device *adev)
 	unlock_srbm(adev);
 }
 
-static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
+static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
 					uint32_t sh_mem_bases)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32(mmSH_MEM_CONFIG, sh_mem_config);
@@ -91,11 +89,9 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 	unlock_srbm(adev);
 }
 
-static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
+static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
 					unsigned int vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	/*
 	 * We have to assume that there is no outstanding mapping.
 	 * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because
@@ -118,9 +114,8 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
 	return 0;
 }
 
-static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
+static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t mec;
 	uint32_t pipe;
 
@@ -537,11 +532,10 @@ static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 	return 0;
 }
 
-static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
+static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 					uint8_t vmid, uint16_t *p_pasid)
 {
 	uint32_t value;
-	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
 
 	value = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 	*p_pasid = value & ATC_VMID0_PASID_MAPPING__PASID_MASK;
@@ -549,12 +543,12 @@ static bool get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
 	return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK);
 }
 
-static int kgd_address_watch_disable(struct kgd_dev *kgd)
+static int kgd_address_watch_disable(struct amdgpu_device *adev)
 {
 	return 0;
 }
 
-static int kgd_address_watch_execute(struct kgd_dev *kgd,
+static int kgd_address_watch_execute(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					uint32_t cntl_val,
 					uint32_t addr_hi,
@@ -563,11 +557,10 @@ static int kgd_address_watch_execute(struct kgd_dev *kgd,
 	return 0;
 }
 
-static int kgd_wave_control_execute(struct kgd_dev *kgd,
+static int kgd_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
 					uint32_t sq_cmd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t data = 0;
 
 	mutex_lock(&adev->grbm_idx_mutex);
@@ -588,28 +581,24 @@ static int kgd_wave_control_execute(struct kgd_dev *kgd,
 	return 0;
 }
 
-static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd,
+static uint32_t kgd_address_watch_get_offset(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					unsigned int reg_offset)
 {
 	return 0;
 }
 
-static void set_scratch_backing_va(struct kgd_dev *kgd,
+static void set_scratch_backing_va(struct amdgpu_device *adev,
 					uint64_t va, uint32_t vmid)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 	WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va);
 	unlock_srbm(adev);
 }
 
-static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint64_t page_table_base)
+static void set_vm_context_page_table_base(struct amdgpu_device *adev,
+		uint32_t vmid, uint64_t page_table_base)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
 		pr_err("trying to set page table base for wrong VMID\n");
 		return;
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 480014ada833..d7b31adcfd80 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -87,14 +87,12 @@ static void release_queue(struct amdgpu_device *adev)
 	unlock_srbm(adev);
 }
 
-void kgd_gfx_v9_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
+void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
 					uint32_t sh_mem_bases)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 
 	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), sh_mem_config);
@@ -104,11 +102,9 @@ void kgd_gfx_v9_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 	unlock_srbm(adev);
 }
 
-int kgd_gfx_v9_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
+int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
 					unsigned int vmid)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	/*
 	 * We have to assume that there is no outstanding mapping.
 	 * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because
@@ -165,9 +161,8 @@ int kgd_gfx_v9_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
  * but still works
  */
 
-int kgd_gfx_v9_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
+int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t mec;
 	uint32_t pipe;
 
@@ -620,11 +615,10 @@ static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 	return 0;
 }
 
-bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
+bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 					uint8_t vmid, uint16_t *p_pasid)
 {
 	uint32_t value;
-	struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
 
 	value = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID0_PASID_MAPPING)
 		     + vmid);
@@ -633,12 +627,12 @@ bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
 	return !!(value & ATC_VMID0_PASID_MAPPING__VALID_MASK);
 }
 
-int kgd_gfx_v9_address_watch_disable(struct kgd_dev *kgd)
+int kgd_gfx_v9_address_watch_disable(struct amdgpu_device *adev)
 {
 	return 0;
 }
 
-int kgd_gfx_v9_address_watch_execute(struct kgd_dev *kgd,
+int kgd_gfx_v9_address_watch_execute(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					uint32_t cntl_val,
 					uint32_t addr_hi,
@@ -647,11 +641,10 @@ int kgd_gfx_v9_address_watch_execute(struct kgd_dev *kgd,
 	return 0;
 }
 
-int kgd_gfx_v9_wave_control_execute(struct kgd_dev *kgd,
+int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
 					uint32_t sq_cmd)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	uint32_t data = 0;
 
 	mutex_lock(&adev->grbm_idx_mutex);
@@ -672,18 +665,16 @@ int kgd_gfx_v9_wave_control_execute(struct kgd_dev *kgd,
 	return 0;
 }
 
-uint32_t kgd_gfx_v9_address_watch_get_offset(struct kgd_dev *kgd,
+uint32_t kgd_gfx_v9_address_watch_get_offset(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					unsigned int reg_offset)
 {
 	return 0;
 }
 
-void kgd_gfx_v9_set_vm_context_page_table_base(struct kgd_dev *kgd,
+void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t page_table_base)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
 		pr_err("trying to set page table base for wrong VMID %u\n",
 		       vmid);
@@ -790,7 +781,7 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx,
  *
  *  Reading registers referenced above involves programming GRBM appropriately
  */
-void kgd_gfx_v9_get_cu_occupancy(struct kgd_dev *kgd, int pasid,
+void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid,
 		int *pasid_wave_cnt, int *max_waves_per_cu)
 {
 	int qidx;
@@ -804,10 +795,8 @@ void kgd_gfx_v9_get_cu_occupancy(struct kgd_dev *kgd, int pasid,
 	int pasid_tmp;
 	int max_queue_cnt;
 	int vmid_wave_cnt = 0;
-	struct amdgpu_device *adev;
 	DECLARE_BITMAP(cp_queue_bitmap, KGD_MAX_QUEUES);
 
-	adev = get_amdgpu_device(kgd);
 	lock_spi_csq_mutexes(adev);
 	soc15_grbm_select(adev, 1, 0, 0, 0);
 
@@ -868,11 +857,9 @@ void kgd_gfx_v9_get_cu_occupancy(struct kgd_dev *kgd, int pasid,
 				adev->gfx.cu_info.max_waves_per_simd;
 }
 
-void kgd_gfx_v9_program_trap_handler_settings(struct kgd_dev *kgd,
+void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev,
                         uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-
 	lock_srbm(adev, 0, 0, 0, vmid);
 
 	/*
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 3fd99968f463..24be49df26fd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -22,13 +22,13 @@
 
 
 
-void kgd_gfx_v9_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
+void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
 		uint32_t sh_mem_config,
 		uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit,
 		uint32_t sh_mem_bases);
-int kgd_gfx_v9_set_pasid_vmid_mapping(struct kgd_dev *kgd, u32 pasid,
+int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
 		unsigned int vmid);
-int kgd_gfx_v9_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
+int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id);
 int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
 			uint32_t wptr_shift, uint32_t wptr_mask,
@@ -46,25 +46,25 @@ int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
 				uint32_t queue_id);
-int kgd_gfx_v9_address_watch_disable(struct kgd_dev *kgd);
-int kgd_gfx_v9_address_watch_execute(struct kgd_dev *kgd,
+int kgd_gfx_v9_address_watch_disable(struct amdgpu_device *adev);
+int kgd_gfx_v9_address_watch_execute(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					uint32_t cntl_val,
 					uint32_t addr_hi,
 					uint32_t addr_lo);
-int kgd_gfx_v9_wave_control_execute(struct kgd_dev *kgd,
+int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
 					uint32_t sq_cmd);
-uint32_t kgd_gfx_v9_address_watch_get_offset(struct kgd_dev *kgd,
+uint32_t kgd_gfx_v9_address_watch_get_offset(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					unsigned int reg_offset);
 
-bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct kgd_dev *kgd,
+bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 					uint8_t vmid, uint16_t *p_pasid);
 
-void kgd_gfx_v9_set_vm_context_page_table_base(struct kgd_dev *kgd,
+void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t page_table_base);
-void kgd_gfx_v9_get_cu_occupancy(struct kgd_dev *kgd, int pasid,
+void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid,
 		int *pasid_wave_cnt, int *max_waves_per_cu);
-void kgd_gfx_v9_program_trap_handler_settings(struct kgd_dev *kgd,
+void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev,
 		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr);
diff --git a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c
index f6233019f042..3073361bb714 100644
--- a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c
+++ b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c
@@ -50,8 +50,8 @@ static bool cik_event_interrupt_isr(struct kfd_dev *dev,
 		*patched_flag = true;
 		*tmp_ihre = *ihre;
 
-		vmid = f2g->read_vmid_from_vmfault_reg(dev->kgd);
-		ret = f2g->get_atc_vmid_pasid_mapping_info(dev->kgd, vmid, &pasid);
+		vmid = f2g->read_vmid_from_vmfault_reg(dev->adev);
+		ret = f2g->get_atc_vmid_pasid_mapping_info(dev->adev, vmid, &pasid);
 
 		tmp_ihre->ring_id &= 0x000000ff;
 		tmp_ihre->ring_id |= vmid << 8;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index f1e7edeb4e6b..7c4f14410a74 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1130,7 +1130,7 @@ static int kfd_ioctl_set_scratch_backing_va(struct file *filep,
 	if (dev->dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS &&
 	    pdd->qpd.vmid != 0 && dev->kfd2kgd->set_scratch_backing_va)
 		dev->kfd2kgd->set_scratch_backing_va(
-			dev->kgd, args->va_addr, pdd->qpd.vmid);
+			dev->adev, args->va_addr, pdd->qpd.vmid);
 
 	return 0;
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c b/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c
index 159add0f5aaa..1e30717b5253 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c
@@ -41,7 +41,7 @@
 
 static void dbgdev_address_watch_disable_nodiq(struct kfd_dev *dev)
 {
-	dev->kfd2kgd->address_watch_disable(dev->kgd);
+	dev->kfd2kgd->address_watch_disable(dev->adev);
 }
 
 static int dbgdev_diq_submit_ib(struct kfd_dbgdev *dbgdev,
@@ -322,7 +322,7 @@ static int dbgdev_address_watch_nodiq(struct kfd_dbgdev *dbgdev,
 		pr_debug("\t\t%30s\n", "* * * * * * * * * * * * * * * * * *");
 
 		pdd->dev->kfd2kgd->address_watch_execute(
-						dbgdev->dev->kgd,
+						dbgdev->dev->adev,
 						i,
 						cntl.u32All,
 						addrHi.u32All,
@@ -420,7 +420,7 @@ static int dbgdev_address_watch_diq(struct kfd_dbgdev *dbgdev,
 
 		aw_reg_add_dword =
 				dbgdev->dev->kfd2kgd->address_watch_get_offset(
-					dbgdev->dev->kgd,
+					dbgdev->dev->adev,
 					i,
 					ADDRESS_WATCH_REG_CNTL);
 
@@ -431,7 +431,7 @@ static int dbgdev_address_watch_diq(struct kfd_dbgdev *dbgdev,
 
 		aw_reg_add_dword =
 				dbgdev->dev->kfd2kgd->address_watch_get_offset(
-					dbgdev->dev->kgd,
+					dbgdev->dev->adev,
 					i,
 					ADDRESS_WATCH_REG_ADDR_HI);
 
@@ -441,7 +441,7 @@ static int dbgdev_address_watch_diq(struct kfd_dbgdev *dbgdev,
 
 		aw_reg_add_dword =
 				dbgdev->dev->kfd2kgd->address_watch_get_offset(
-					dbgdev->dev->kgd,
+					dbgdev->dev->adev,
 					i,
 					ADDRESS_WATCH_REG_ADDR_LO);
 
@@ -457,7 +457,7 @@ static int dbgdev_address_watch_diq(struct kfd_dbgdev *dbgdev,
 
 		aw_reg_add_dword =
 				dbgdev->dev->kfd2kgd->address_watch_get_offset(
-					dbgdev->dev->kgd,
+					dbgdev->dev->adev,
 					i,
 					ADDRESS_WATCH_REG_CNTL);
 
@@ -752,7 +752,7 @@ static int dbgdev_wave_control_nodiq(struct kfd_dbgdev *dbgdev,
 
 	pr_debug("\t\t %30s\n", "* * * * * * * * * * * * * * * * * *");
 
-	return dbgdev->dev->kfd2kgd->wave_control_execute(dbgdev->dev->kgd,
+	return dbgdev->dev->kfd2kgd->wave_control_execute(dbgdev->dev->adev,
 							reg_gfx_index.u32All,
 							reg_sq_cmd.u32All);
 }
@@ -784,7 +784,7 @@ int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process *p)
 
 	for (vmid = first_vmid_to_scan; vmid <= last_vmid_to_scan; vmid++) {
 		status = dev->kfd2kgd->get_atc_vmid_pasid_mapping_info
-				(dev->kgd, vmid, &queried_pasid);
+				(dev->adev, vmid, &queried_pasid);
 
 		if (status && queried_pasid == p->pasid) {
 			pr_debug("Killing wave fronts of vmid %d and pasid 0x%x\n",
@@ -811,7 +811,7 @@ int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process *p)
 	/* for non DIQ we need to patch the VMID: */
 	reg_sq_cmd.bits.vm_id = vmid;
 
-	dev->kfd2kgd->wave_control_execute(dev->kgd,
+	dev->kfd2kgd->wave_control_execute(dev->adev,
 					reg_gfx_index.u32All,
 					reg_sq_cmd.u32All);
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 1919efe06f30..9cdf3fb8c58e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -130,7 +130,7 @@ void program_sh_mem_settings(struct device_queue_manager *dqm,
 					struct qcm_process_device *qpd)
 {
 	return dqm->dev->kfd2kgd->program_sh_mem_settings(
-						dqm->dev->kgd, qpd->vmid,
+						dqm->dev->adev, qpd->vmid,
 						qpd->sh_mem_config,
 						qpd->sh_mem_ape1_base,
 						qpd->sh_mem_ape1_limit,
@@ -216,7 +216,7 @@ static void program_trap_handler_settings(struct device_queue_manager *dqm,
 {
 	if (dqm->dev->kfd2kgd->program_trap_handler_settings)
 		dqm->dev->kfd2kgd->program_trap_handler_settings(
-						dqm->dev->kgd, qpd->vmid,
+						dqm->dev->adev, qpd->vmid,
 						qpd->tba_addr, qpd->tma_addr);
 }
 
@@ -257,14 +257,14 @@ static int allocate_vmid(struct device_queue_manager *dqm,
 	/* qpd->page_table_base is set earlier when register_process()
 	 * is called, i.e. when the first queue is created.
 	 */
-	dqm->dev->kfd2kgd->set_vm_context_page_table_base(dqm->dev->kgd,
+	dqm->dev->kfd2kgd->set_vm_context_page_table_base(dqm->dev->adev,
 			qpd->vmid,
 			qpd->page_table_base);
 	/* invalidate the VM context after pasid and vmid mapping is set up */
 	kfd_flush_tlb(qpd_to_pdd(qpd), TLB_FLUSH_LEGACY);
 
 	if (dqm->dev->kfd2kgd->set_scratch_backing_va)
-		dqm->dev->kfd2kgd->set_scratch_backing_va(dqm->dev->kgd,
+		dqm->dev->kfd2kgd->set_scratch_backing_va(dqm->dev->adev,
 				qpd->sh_hidden_private_base, qpd->vmid);
 
 	return 0;
@@ -775,7 +775,7 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm,
 
 	if (!list_empty(&qpd->queues_list)) {
 		dqm->dev->kfd2kgd->set_vm_context_page_table_base(
-				dqm->dev->kgd,
+				dqm->dev->adev,
 				qpd->vmid,
 				qpd->page_table_base);
 		kfd_flush_tlb(pdd, TLB_FLUSH_LEGACY);
@@ -953,7 +953,7 @@ set_pasid_vmid_mapping(struct device_queue_manager *dqm, u32 pasid,
 			unsigned int vmid)
 {
 	return dqm->dev->kfd2kgd->set_pasid_vmid_mapping(
-						dqm->dev->kgd, pasid, vmid);
+						dqm->dev->adev, pasid, vmid);
 }
 
 static void init_interrupts(struct device_queue_manager *dqm)
@@ -962,7 +962,7 @@ static void init_interrupts(struct device_queue_manager *dqm)
 
 	for (i = 0 ; i < get_pipes_per_mec(dqm) ; i++)
 		if (is_pipe_enabled(dqm, 0, i))
-			dqm->dev->kfd2kgd->init_interrupts(dqm->dev->kgd, i);
+			dqm->dev->kfd2kgd->init_interrupts(dqm->dev->adev, i);
 }
 
 static int initialize_nocpsch(struct device_queue_manager *dqm)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 21ec8a18cad2..8b3dc6462405 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -286,7 +286,7 @@ static int kfd_get_cu_occupancy(struct attribute *attr, char *buffer)
 	/* Collect wave count from device if it supports */
 	wave_cnt = 0;
 	max_waves_per_cu = 0;
-	dev->kfd2kgd->get_cu_occupancy(dev->kgd, proc->pasid, &wave_cnt,
+	dev->kfd2kgd->get_cu_occupancy(dev->adev, proc->pasid, &wave_cnt,
 			&max_waves_per_cu);
 
 	/* Translate wave count to number of compute units */
diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
index 75e40acdd699..8f4f3a1700e8 100644
--- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
+++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
@@ -229,14 +229,14 @@ struct tile_config {
  */
 struct kfd2kgd_calls {
 	/* Register access functions */
-	void (*program_sh_mem_settings)(struct kgd_dev *kgd, uint32_t vmid,
+	void (*program_sh_mem_settings)(struct amdgpu_device *adev, uint32_t vmid,
 			uint32_t sh_mem_config,	uint32_t sh_mem_ape1_base,
 			uint32_t sh_mem_ape1_limit, uint32_t sh_mem_bases);
 
-	int (*set_pasid_vmid_mapping)(struct kgd_dev *kgd, u32 pasid,
+	int (*set_pasid_vmid_mapping)(struct amdgpu_device *adev, u32 pasid,
 					unsigned int vmid);
 
-	int (*init_interrupts)(struct kgd_dev *kgd, uint32_t pipe_id);
+	int (*init_interrupts)(struct amdgpu_device *adev, uint32_t pipe_id);
 
 	int (*hqd_load)(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
@@ -271,20 +271,19 @@ struct kfd2kgd_calls {
 	int (*hqd_sdma_destroy)(struct amdgpu_device *adev, void *mqd,
 				unsigned int timeout);
 
-	int (*address_watch_disable)(struct kgd_dev *kgd);
-	int (*address_watch_execute)(struct kgd_dev *kgd,
+	int (*address_watch_disable)(struct amdgpu_device *adev);
+	int (*address_watch_execute)(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					uint32_t cntl_val,
 					uint32_t addr_hi,
 					uint32_t addr_lo);
-	int (*wave_control_execute)(struct kgd_dev *kgd,
+	int (*wave_control_execute)(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
 					uint32_t sq_cmd);
-	uint32_t (*address_watch_get_offset)(struct kgd_dev *kgd,
+	uint32_t (*address_watch_get_offset)(struct amdgpu_device *adev,
 					unsigned int watch_point_id,
 					unsigned int reg_offset);
-	bool (*get_atc_vmid_pasid_mapping_info)(
-					struct kgd_dev *kgd,
+	bool (*get_atc_vmid_pasid_mapping_info)(struct amdgpu_device *adev,
 					uint8_t vmid,
 					uint16_t *p_pasid);
 
@@ -292,16 +291,16 @@ struct kfd2kgd_calls {
 	 * passed to the shader by the CP. It's the user mode driver's
 	 * responsibility.
 	 */
-	void (*set_scratch_backing_va)(struct kgd_dev *kgd,
+	void (*set_scratch_backing_va)(struct amdgpu_device *adev,
 				uint64_t va, uint32_t vmid);
 
-	void (*set_vm_context_page_table_base)(struct kgd_dev *kgd,
+	void (*set_vm_context_page_table_base)(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t page_table_base);
-	uint32_t (*read_vmid_from_vmfault_reg)(struct kgd_dev *kgd);
+	uint32_t (*read_vmid_from_vmfault_reg)(struct amdgpu_device *adev);
 
-	void (*get_cu_occupancy)(struct kgd_dev *kgd, int pasid, int *wave_cnt,
-			int *max_waves_per_cu);
-	void (*program_trap_handler_settings)(struct kgd_dev *kgd,
+	void (*get_cu_occupancy)(struct amdgpu_device *adev, int pasid,
+			int *wave_cnt, int *max_waves_per_cu);
+	void (*program_trap_handler_settings)(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr);
 };
 
-- 
2.25.1

[PATCH 09/13] drm/amdkfd: replace kgd_dev in various amgpu_amdkfd funcs

[Graham Sider]

Modified definitions:

- amdgpu_amdkfd_submit_ib
- amdgpu_amdkfd_set_compute_idle
- amdgpu_amdkfd_have_atomics_support
- amdgpu_amdkfd_flush_gpu_tlb_pasid
- amdgpu_amdkfd_flush_gpu_tlb_pasid
- amdgpu_amdkfd_gpu_reset
- amdgpu_amdkfd_alloc_gtt_mem
- amdgpu_amdkfd_free_gtt_mem
- amdgpu_amdkfd_alloc_gws
- amdgpu_amdkfd_free_gws
- amdgpu_amdkfd_ras_poison_consumption_handler

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c    | 41 +++++++------------
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h    | 27 ++++++------
 drivers/gpu/drm/amd/amdkfd/kfd_device.c       | 18 ++++----
 .../drm/amd/amdkfd/kfd_device_queue_manager.c |  8 ++--
 .../gpu/drm/amd/amdkfd/kfd_int_process_v9.c   |  4 +-
 .../gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c   |  4 +-
 drivers/gpu/drm/amd/amdkfd/kfd_process.c      |  4 +-
 drivers/gpu/drm/amd/amdkfd/kfd_svm.c          | 10 ++---
 8 files changed, 53 insertions(+), 63 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 7077f21f0021..69fc8f0d9c45 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -233,19 +233,16 @@ int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev)
 	return r;
 }
 
-void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd)
+void amdgpu_amdkfd_gpu_reset(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	if (amdgpu_device_should_recover_gpu(adev))
 		amdgpu_device_gpu_recover(adev, NULL);
 }
 
-int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
+int amdgpu_amdkfd_alloc_gtt_mem(struct amdgpu_device *adev, size_t size,
 				void **mem_obj, uint64_t *gpu_addr,
 				void **cpu_ptr, bool cp_mqd_gfx9)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct amdgpu_bo *bo = NULL;
 	struct amdgpu_bo_param bp;
 	int r;
@@ -314,7 +311,7 @@ int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
 	return r;
 }
 
-void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
+void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void *mem_obj)
 {
 	struct amdgpu_bo *bo = (struct amdgpu_bo *) mem_obj;
 
@@ -325,10 +322,9 @@ void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
 	amdgpu_bo_unref(&(bo));
 }
 
-int amdgpu_amdkfd_alloc_gws(struct kgd_dev *kgd, size_t size,
+int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
 				void **mem_obj)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct amdgpu_bo *bo = NULL;
 	struct amdgpu_bo_user *ubo;
 	struct amdgpu_bo_param bp;
@@ -355,7 +351,7 @@ int amdgpu_amdkfd_alloc_gws(struct kgd_dev *kgd, size_t size,
 	return 0;
 }
 
-void amdgpu_amdkfd_free_gws(struct kgd_dev *kgd, void *mem_obj)
+void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj)
 {
 	struct amdgpu_bo *bo = (struct amdgpu_bo *)mem_obj;
 
@@ -675,11 +671,11 @@ int amdgpu_amdkfd_get_noretry(struct kgd_dev *kgd)
 	return adev->gmc.noretry;
 }
 
-int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine,
+int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
+				enum kgd_engine_type engine,
 				uint32_t vmid, uint64_t gpu_addr,
 				uint32_t *ib_cmd, uint32_t ib_len)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct amdgpu_ring *ring;
@@ -730,10 +726,8 @@ int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine,
 	return ret;
 }
 
-void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle)
+void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	amdgpu_dpm_switch_power_profile(adev,
 					PP_SMC_POWER_PROFILE_COMPUTE,
 					!idle);
@@ -747,10 +741,9 @@ bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid)
 	return false;
 }
 
-int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct kgd_dev *kgd, uint16_t vmid)
+int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
+				     uint16_t vmid)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	if (adev->family == AMDGPU_FAMILY_AI) {
 		int i;
 
@@ -763,10 +756,9 @@ int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct kgd_dev *kgd, uint16_t vmid)
 	return 0;
 }
 
-int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid,
-				      enum TLB_FLUSH_TYPE flush_type)
+int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+				      uint16_t pasid, enum TLB_FLUSH_TYPE flush_type)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	bool all_hub = false;
 
 	if (adev->family == AMDGPU_FAMILY_AI)
@@ -775,21 +767,18 @@ int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid,
 	return amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, flush_type, all_hub);
 }
 
-bool amdgpu_amdkfd_have_atomics_support(struct kgd_dev *kgd)
+bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	return adev->have_atomics_support;
 }
 
-void amdgpu_amdkfd_ras_poison_consumption_handler(struct kgd_dev *kgd)
+void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct ras_err_data err_data = {0, 0, 0, NULL};
 
 	/* CPU MCA will handle page retirement if connected_to_cpu is 1 */
 	if (!adev->gmc.xgmi.connected_to_cpu)
 		amdgpu_umc_process_ras_data_cb(adev, &err_data, NULL);
 	else
-		amdgpu_amdkfd_gpu_reset(kgd);
+		amdgpu_amdkfd_gpu_reset(adev);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 69de31754907..8d5c18953723 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -144,14 +144,16 @@ void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
 void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev);
 void amdgpu_amdkfd_device_init(struct amdgpu_device *adev);
 void amdgpu_amdkfd_device_fini_sw(struct amdgpu_device *adev);
-int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine,
+int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
+				enum kgd_engine_type engine,
 				uint32_t vmid, uint64_t gpu_addr,
 				uint32_t *ib_cmd, uint32_t ib_len);
-void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle);
-bool amdgpu_amdkfd_have_atomics_support(struct kgd_dev *kgd);
-int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct kgd_dev *kgd, uint16_t vmid);
-int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct kgd_dev *kgd, uint16_t pasid,
-				      enum TLB_FLUSH_TYPE flush_type);
+void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle);
+bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev);
+int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
+				uint16_t vmid);
+int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+				uint16_t pasid, enum TLB_FLUSH_TYPE flush_type);
 
 bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid);
 
@@ -159,7 +161,7 @@ int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev);
 
 int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev);
 
-void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd);
+void amdgpu_amdkfd_gpu_reset(struct amdgpu_device *adev);
 
 int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev,
 					int queue_bit);
@@ -198,12 +200,13 @@ int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm)
 }
 #endif
 /* Shared API */
-int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
+int amdgpu_amdkfd_alloc_gtt_mem(struct amdgpu_device *adev, size_t size,
 				void **mem_obj, uint64_t *gpu_addr,
 				void **cpu_ptr, bool mqd_gfx9);
-void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj);
-int amdgpu_amdkfd_alloc_gws(struct kgd_dev *kgd, size_t size, void **mem_obj);
-void amdgpu_amdkfd_free_gws(struct kgd_dev *kgd, void *mem_obj);
+void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void *mem_obj);
+int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
+				void **mem_obj);
+void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj);
 int amdgpu_amdkfd_add_gws_to_process(void *info, void *gws, struct kgd_mem **mem);
 int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem);
 uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
@@ -290,7 +293,7 @@ int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd,
 				      uint64_t *mmap_offset);
 int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd,
 				struct tile_config *config);
-void amdgpu_amdkfd_ras_poison_consumption_handler(struct kgd_dev *kgd);
+void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev);
 #if IS_ENABLED(CONFIG_HSA_AMD)
 void amdgpu_amdkfd_gpuvm_init_mem_limits(void);
 void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 81ca00d7b3da..a90ec8de213b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -893,7 +893,7 @@ static int kfd_gws_init(struct kfd_dev *kfd)
 			&& kfd->mec2_fw_version >= 0x30)
 		|| (kfd->device_info->asic_family == CHIP_ALDEBARAN
 			&& kfd->mec2_fw_version >= 0x28))
-		ret = amdgpu_amdkfd_alloc_gws(kfd->kgd,
+		ret = amdgpu_amdkfd_alloc_gws(kfd->adev,
 				amdgpu_amdkfd_get_num_gws(kfd->kgd), &kfd->gws);
 
 	return ret;
@@ -928,7 +928,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 	 * 32 and 64-bit requests are possible and must be
 	 * supported.
 	 */
-	kfd->pci_atomic_requested = amdgpu_amdkfd_have_atomics_support(kfd->kgd);
+	kfd->pci_atomic_requested = amdgpu_amdkfd_have_atomics_support(kfd->adev);
 	if (!kfd->pci_atomic_requested &&
 	    kfd->device_info->needs_pci_atomics &&
 	    (!kfd->device_info->no_atomic_fw_version ||
@@ -975,7 +975,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 	size += 512 * 1024;
 
 	if (amdgpu_amdkfd_alloc_gtt_mem(
-			kfd->kgd, size, &kfd->gtt_mem,
+			kfd->adev, size, &kfd->gtt_mem,
 			&kfd->gtt_start_gpu_addr, &kfd->gtt_start_cpu_ptr,
 			false)) {
 		dev_err(kfd_device, "Could not allocate %d bytes\n", size);
@@ -1069,10 +1069,10 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 kfd_doorbell_error:
 	kfd_gtt_sa_fini(kfd);
 kfd_gtt_sa_init_error:
-	amdgpu_amdkfd_free_gtt_mem(kfd->kgd, kfd->gtt_mem);
+	amdgpu_amdkfd_free_gtt_mem(kfd->adev, kfd->gtt_mem);
 alloc_gtt_mem_failure:
 	if (kfd->gws)
-		amdgpu_amdkfd_free_gws(kfd->kgd, kfd->gws);
+		amdgpu_amdkfd_free_gws(kfd->adev, kfd->gws);
 	dev_err(kfd_device,
 		"device %x:%x NOT added due to errors\n",
 		kfd->pdev->vendor, kfd->pdev->device);
@@ -1089,9 +1089,9 @@ void kgd2kfd_device_exit(struct kfd_dev *kfd)
 		kfd_doorbell_fini(kfd);
 		ida_destroy(&kfd->doorbell_ida);
 		kfd_gtt_sa_fini(kfd);
-		amdgpu_amdkfd_free_gtt_mem(kfd->kgd, kfd->gtt_mem);
+		amdgpu_amdkfd_free_gtt_mem(kfd->adev, kfd->gtt_mem);
 		if (kfd->gws)
-			amdgpu_amdkfd_free_gws(kfd->kgd, kfd->gws);
+			amdgpu_amdkfd_free_gws(kfd->adev, kfd->gws);
 	}
 
 	kfree(kfd);
@@ -1527,7 +1527,7 @@ void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd)
 void kfd_inc_compute_active(struct kfd_dev *kfd)
 {
 	if (atomic_inc_return(&kfd->compute_profile) == 1)
-		amdgpu_amdkfd_set_compute_idle(kfd->kgd, false);
+		amdgpu_amdkfd_set_compute_idle(kfd->adev, false);
 }
 
 void kfd_dec_compute_active(struct kfd_dev *kfd)
@@ -1535,7 +1535,7 @@ void kfd_dec_compute_active(struct kfd_dev *kfd)
 	int count = atomic_dec_return(&kfd->compute_profile);
 
 	if (count == 0)
-		amdgpu_amdkfd_set_compute_idle(kfd->kgd, true);
+		amdgpu_amdkfd_set_compute_idle(kfd->adev, true);
 	WARN_ONCE(count < 0, "Compute profile ref. count error");
 }
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 9cdf3fb8c58e..330fbbb15d83 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -283,7 +283,7 @@ static int flush_texture_cache_nocpsch(struct kfd_dev *kdev,
 	if (ret)
 		return ret;
 
-	return amdgpu_amdkfd_submit_ib(kdev->kgd, KGD_ENGINE_MEC1, qpd->vmid,
+	return amdgpu_amdkfd_submit_ib(kdev->adev, KGD_ENGINE_MEC1, qpd->vmid,
 				qpd->ib_base, (uint32_t *)qpd->ib_kaddr,
 				pmf->release_mem_size / sizeof(uint32_t));
 }
@@ -1844,7 +1844,7 @@ static int allocate_hiq_sdma_mqd(struct device_queue_manager *dqm)
 		dev->device_info->num_sdma_queues_per_engine +
 		dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ]->mqd_size;
 
-	retval = amdgpu_amdkfd_alloc_gtt_mem(dev->kgd, size,
+	retval = amdgpu_amdkfd_alloc_gtt_mem(dev->adev, size,
 		&(mem_obj->gtt_mem), &(mem_obj->gpu_addr),
 		(void *)&(mem_obj->cpu_ptr), false);
 
@@ -1994,7 +1994,7 @@ static void deallocate_hiq_sdma_mqd(struct kfd_dev *dev,
 {
 	WARN(!mqd, "No hiq sdma mqd trunk to free");
 
-	amdgpu_amdkfd_free_gtt_mem(dev->kgd, mqd->gtt_mem);
+	amdgpu_amdkfd_free_gtt_mem(dev->adev, mqd->gtt_mem);
 }
 
 void device_queue_manager_uninit(struct device_queue_manager *dqm)
@@ -2025,7 +2025,7 @@ static void kfd_process_hw_exception(struct work_struct *work)
 {
 	struct device_queue_manager *dqm = container_of(work,
 			struct device_queue_manager, hw_exception_work);
-	amdgpu_amdkfd_gpu_reset(dqm->dev->kgd);
+	amdgpu_amdkfd_gpu_reset(dqm->dev->adev);
 }
 
 #if defined(CONFIG_DEBUG_FS)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
index 543e7ea75593..20512a4e9a91 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c
@@ -231,7 +231,7 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev,
 				if (sq_intr_err != SQ_INTERRUPT_ERROR_TYPE_ILLEGAL_INST &&
 					sq_intr_err != SQ_INTERRUPT_ERROR_TYPE_MEMVIOL) {
 					kfd_signal_poison_consumed_event(dev, pasid);
-					amdgpu_amdkfd_ras_poison_consumption_handler(dev->kgd);
+					amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev);
 					return;
 				}
 				break;
@@ -253,7 +253,7 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev,
 			kfd_signal_event_interrupt(pasid, context_id0 & 0xfffffff, 28);
 		} else if (source_id == SOC15_INTSRC_SDMA_ECC) {
 			kfd_signal_poison_consumed_event(dev, pasid);
-			amdgpu_amdkfd_ras_poison_consumption_handler(dev->kgd);
+			amdgpu_amdkfd_ras_poison_consumption_handler(dev->adev);
 			return;
 		}
 	} else if (client_id == SOC15_IH_CLIENTID_VMC ||
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
index 643d3eb18ebc..3a5e484a4192 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
@@ -107,7 +107,7 @@ static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
 		mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL);
 		if (!mqd_mem_obj)
 			return NULL;
-		retval = amdgpu_amdkfd_alloc_gtt_mem(kfd->kgd,
+		retval = amdgpu_amdkfd_alloc_gtt_mem(kfd->adev,
 			ALIGN(q->ctl_stack_size, PAGE_SIZE) +
 				ALIGN(sizeof(struct v9_mqd), PAGE_SIZE),
 			&(mqd_mem_obj->gtt_mem),
@@ -299,7 +299,7 @@ static void free_mqd(struct mqd_manager *mm, void *mqd,
 	struct kfd_dev *kfd = mm->dev;
 
 	if (mqd_mem_obj->gtt_mem) {
-		amdgpu_amdkfd_free_gtt_mem(kfd->kgd, mqd_mem_obj->gtt_mem);
+		amdgpu_amdkfd_free_gtt_mem(kfd->adev, mqd_mem_obj->gtt_mem);
 		kfree(mqd_mem_obj);
 	} else {
 		kfd_gtt_sa_free(mm->dev, mqd_mem_obj);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 8b3dc6462405..704f07be2b3f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1896,10 +1896,10 @@ void kfd_flush_tlb(struct kfd_process_device *pdd, enum TLB_FLUSH_TYPE type)
 		 * only happens when the first queue is created.
 		 */
 		if (pdd->qpd.vmid)
-			amdgpu_amdkfd_flush_gpu_tlb_vmid(dev->kgd,
+			amdgpu_amdkfd_flush_gpu_tlb_vmid(dev->adev,
 							pdd->qpd.vmid);
 	} else {
-		amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->kgd,
+		amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->adev,
 					pdd->process->pasid, type);
 	}
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index b691c8495d66..5b67b14fd8be 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -1129,7 +1129,6 @@ svm_range_unmap_from_gpus(struct svm_range *prange, unsigned long start,
 	DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE);
 	struct kfd_process_device *pdd;
 	struct dma_fence *fence = NULL;
-	struct amdgpu_device *adev;
 	struct kfd_process *p;
 	uint32_t gpuidx;
 	int r = 0;
@@ -1145,9 +1144,9 @@ svm_range_unmap_from_gpus(struct svm_range *prange, unsigned long start,
 			pr_debug("failed to find device idx %d\n", gpuidx);
 			return -EINVAL;
 		}
-		adev = (struct amdgpu_device *)pdd->dev->kgd;
 
-		r = svm_range_unmap_from_gpu(adev, drm_priv_to_vm(pdd->drm_priv),
+		r = svm_range_unmap_from_gpu(pdd->dev->adev,
+					     drm_priv_to_vm(pdd->drm_priv),
 					     start, last, &fence);
 		if (r)
 			break;
@@ -1159,7 +1158,7 @@ svm_range_unmap_from_gpus(struct svm_range *prange, unsigned long start,
 			if (r)
 				break;
 		}
-		amdgpu_amdkfd_flush_gpu_tlb_pasid((struct kgd_dev *)adev,
+		amdgpu_amdkfd_flush_gpu_tlb_pasid(pdd->dev->adev,
 					p->pasid, TLB_FLUSH_HEAVYWEIGHT);
 	}
 
@@ -1243,8 +1242,7 @@ svm_range_map_to_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		struct kfd_process *p;
 
 		p = container_of(prange->svms, struct kfd_process, svms);
-		amdgpu_amdkfd_flush_gpu_tlb_pasid((struct kgd_dev *)adev,
-						p->pasid, TLB_FLUSH_LEGACY);
+		amdgpu_amdkfd_flush_gpu_tlb_pasid(adev, p->pasid, TLB_FLUSH_LEGACY);
 	}
 out:
 	return r;
-- 
2.25.1

[PATCH 10/13] drm/amdkfd: replace kgd_dev in get amdgpu_amdkfd funcs

[Graham Sider]

Modified definitions:

- amdgpu_amdkfd_get_fw_version
- amdgpu_amdkfd_get_local_mem_info
- amdgpu_amdkfd_get_gpu_clock_counter
- amdgpu_amdkfd_get_max_engine_clock_in_mhz
- amdgpu_amdkfd_get_cu_info
- amdgpu_amdkfd_get_dmabuf_info
- amdgpu_amdkfd_get_vram_usage
- amdgpu_amdkfd_get_hive_id
- amdgpu_amdkfd_get_unique_id
- amdgpu_amdkfd_get_mmio_remap_phys_addr
- amdgpu_amdkfd_get_num_gws
- amdgpu_amdkfd_get_asic_rev_id
- amdgpu_amdkfd_get_noretry
- amdgpu_amdkfd_get_xgmi_hops_count
- amdgpu_amdkfd_get_xgmi_bandwidth_mbytes
- amdgpu_amdkfd_get_pcie_bandwidth_mbytes

Also replaces kfd_device_by_kgd with kfd_device_by_adev, now
searching via adev rather than kgd.

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c    | 73 +++++++------------
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h    | 38 +++++-----
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c      | 16 ++--
 drivers/gpu/drm/amd/amdkfd/kfd_crat.c         | 16 ++--
 drivers/gpu/drm/amd/amdkfd/kfd_device.c       | 14 ++--
 drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c  |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_priv.h         |  2 +-
 .../amd/amdkfd/kfd_process_queue_manager.c    |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_topology.c     | 18 ++---
 9 files changed, 82 insertions(+), 99 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 69fc8f0d9c45..79a2e37baa59 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -358,11 +358,9 @@ void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj)
 	amdgpu_bo_unref(&bo);
 }
 
-uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
+uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
 				      enum kgd_engine_type type)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	switch (type) {
 	case KGD_ENGINE_PFP:
 		return adev->gfx.pfp_fw_version;
@@ -395,11 +393,9 @@ uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
 	return 0;
 }
 
-void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
+void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
 				      struct kfd_local_mem_info *mem_info)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	memset(mem_info, 0, sizeof(*mem_info));
 
 	mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
@@ -424,19 +420,15 @@ void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
 		mem_info->mem_clk_max = 100;
 }
 
-uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd)
+uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	if (adev->gfx.funcs->get_gpu_clock_counter)
 		return adev->gfx.funcs->get_gpu_clock_counter(adev);
 	return 0;
 }
 
-uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
+uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	/* the sclk is in quantas of 10kHz */
 	if (amdgpu_sriov_vf(adev))
 		return adev->clock.default_sclk / 100;
@@ -446,9 +438,8 @@ uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
 		return 100;
 }
 
-void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info)
+void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev, struct kfd_cu_info *cu_info)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct amdgpu_cu_info acu_info = adev->gfx.cu_info;
 
 	memset(cu_info, 0, sizeof(*cu_info));
@@ -469,13 +460,12 @@ void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info)
 	cu_info->lds_size = acu_info.lds_size;
 }
 
-int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
-				  struct kgd_dev **dma_buf_kgd,
+int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
+				  struct amdgpu_device **dmabuf_adev,
 				  uint64_t *bo_size, void *metadata_buffer,
 				  size_t buffer_size, uint32_t *metadata_size,
 				  uint32_t *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct dma_buf *dma_buf;
 	struct drm_gem_object *obj;
 	struct amdgpu_bo *bo;
@@ -503,8 +493,8 @@ int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
 		goto out_put;
 
 	r = 0;
-	if (dma_buf_kgd)
-		*dma_buf_kgd = (struct kgd_dev *)adev;
+	if (dmabuf_adev)
+		*dmabuf_adev = adev;
 	if (bo_size)
 		*bo_size = amdgpu_bo_size(bo);
 	if (metadata_buffer)
@@ -524,32 +514,28 @@ int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
 	return r;
 }
 
-uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd)
+uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct ttm_resource_manager *vram_man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
 
 	return amdgpu_vram_mgr_usage(vram_man);
 }
 
-uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd)
+uint64_t amdgpu_amdkfd_get_hive_id(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	return adev->gmc.xgmi.hive_id;
 }
 
-uint64_t amdgpu_amdkfd_get_unique_id(struct kgd_dev *kgd)
+uint64_t amdgpu_amdkfd_get_unique_id(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	return adev->unique_id;
 }
 
-uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *src)
+uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
+					  struct amdgpu_device *src)
 {
-	struct amdgpu_device *peer_adev = (struct amdgpu_device *)src;
-	struct amdgpu_device *adev = (struct amdgpu_device *)dst;
+	struct amdgpu_device *peer_adev = src;
+	struct amdgpu_device *adev = dst;
 	int ret = amdgpu_xgmi_get_hops_count(adev, peer_adev);
 
 	if (ret < 0) {
@@ -561,16 +547,18 @@ uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *s
 	return  (uint8_t)ret;
 }
 
-int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct kgd_dev *src, bool is_min)
+int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
+					    struct amdgpu_device *src,
+					    bool is_min)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)dst, *peer_adev;
+	struct amdgpu_device *adev = dst, *peer_adev;
 	int num_links;
 
 	if (adev->asic_type != CHIP_ALDEBARAN)
 		return 0;
 
 	if (src)
-		peer_adev = (struct amdgpu_device *)src;
+		peer_adev = src;
 
 	/* num links returns 0 for indirect peers since indirect route is unknown. */
 	num_links = is_min ? 1 : amdgpu_xgmi_get_num_links(adev, peer_adev);
@@ -585,9 +573,8 @@ int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct kgd_dev
 	return (num_links * 16 * 25000)/BITS_PER_BYTE;
 }
 
-int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min)
+int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)dev;
 	int num_lanes_shift = (is_min ? ffs(adev->pm.pcie_mlw_mask) :
 							fls(adev->pm.pcie_mlw_mask)) - 1;
 	int gen_speed_shift = (is_min ? ffs(adev->pm.pcie_gen_mask &
@@ -643,31 +630,23 @@ int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min)
 	return (num_lanes_factor * gen_speed_mbits_factor)/BITS_PER_BYTE;
 }
 
-uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev *kgd)
+uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	return adev->rmmio_remap.bus_addr;
 }
 
-uint32_t amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd)
+uint32_t amdgpu_amdkfd_get_num_gws(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	return adev->gds.gws_size;
 }
 
-uint32_t amdgpu_amdkfd_get_asic_rev_id(struct kgd_dev *kgd)
+uint32_t amdgpu_amdkfd_get_asic_rev_id(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	return adev->rev_id;
 }
 
-int amdgpu_amdkfd_get_noretry(struct kgd_dev *kgd)
+int amdgpu_amdkfd_get_noretry(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	return adev->gmc.noretry;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 8d5c18953723..7e3697a7a5cd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -209,29 +209,33 @@ int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
 void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj);
 int amdgpu_amdkfd_add_gws_to_process(void *info, void *gws, struct kgd_mem **mem);
 int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem);
-uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
+uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
 				      enum kgd_engine_type type);
-void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
+void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
 				      struct kfd_local_mem_info *mem_info);
-uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd);
+uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev);
 
-uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd);
-void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info);
-int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
-				  struct kgd_dev **dmabuf_kgd,
+uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev);
+void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev,
+			       struct kfd_cu_info *cu_info);
+int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
+				  struct amdgpu_device **dmabuf_adev,
 				  uint64_t *bo_size, void *metadata_buffer,
 				  size_t buffer_size, uint32_t *metadata_size,
 				  uint32_t *flags);
-uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd);
-uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd);
-uint64_t amdgpu_amdkfd_get_unique_id(struct kgd_dev *kgd);
-uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev *kgd);
-uint32_t amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd);
-uint32_t amdgpu_amdkfd_get_asic_rev_id(struct kgd_dev *kgd);
-int amdgpu_amdkfd_get_noretry(struct kgd_dev *kgd);
-uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *src);
-int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct kgd_dev *src, bool is_min);
-int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min);
+uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device *adev);
+uint64_t amdgpu_amdkfd_get_hive_id(struct amdgpu_device *adev);
+uint64_t amdgpu_amdkfd_get_unique_id(struct amdgpu_device *adev);
+uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct amdgpu_device *adev);
+uint32_t amdgpu_amdkfd_get_num_gws(struct amdgpu_device *adev);
+uint32_t amdgpu_amdkfd_get_asic_rev_id(struct amdgpu_device *adev);
+int amdgpu_amdkfd_get_noretry(struct amdgpu_device *adev);
+uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
+					  struct amdgpu_device *src);
+int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
+					    struct amdgpu_device *src,
+					    bool is_min);
+int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min);
 
 /* Read user wptr from a specified user address space with page fault
  * disabled. The memory must be pinned and mapped to the hardware when
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 7c4f14410a74..47acfef1aebd 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -850,7 +850,7 @@ static int kfd_ioctl_get_clock_counters(struct file *filep,
 	dev = kfd_device_by_id(args->gpu_id);
 	if (dev)
 		/* Reading GPU clock counter from KGD */
-		args->gpu_clock_counter = amdgpu_amdkfd_get_gpu_clock_counter(dev->kgd);
+		args->gpu_clock_counter = amdgpu_amdkfd_get_gpu_clock_counter(dev->adev);
 	else
 		/* Node without GPU resource */
 		args->gpu_clock_counter = 0;
@@ -1237,7 +1237,7 @@ bool kfd_dev_is_large_bar(struct kfd_dev *dev)
 	if (dev->use_iommu_v2)
 		return false;
 
-	amdgpu_amdkfd_get_local_mem_info(dev->kgd, &mem_info);
+	amdgpu_amdkfd_get_local_mem_info(dev->adev, &mem_info);
 	if (mem_info.local_mem_size_private == 0 &&
 			mem_info.local_mem_size_public > 0)
 		return true;
@@ -1306,7 +1306,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 			err = -EINVAL;
 			goto err_unlock;
 		}
-		offset = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd);
+		offset = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->adev);
 		if (!offset) {
 			err = -ENOMEM;
 			goto err_unlock;
@@ -1664,7 +1664,7 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
 {
 	struct kfd_ioctl_get_dmabuf_info_args *args = data;
 	struct kfd_dev *dev = NULL;
-	struct kgd_dev *dma_buf_kgd;
+	struct amdgpu_device *dmabuf_adev;
 	void *metadata_buffer = NULL;
 	uint32_t flags;
 	unsigned int i;
@@ -1684,15 +1684,15 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
 	}
 
 	/* Get dmabuf info from KGD */
-	r = amdgpu_amdkfd_get_dmabuf_info(dev->kgd, args->dmabuf_fd,
-					  &dma_buf_kgd, &args->size,
+	r = amdgpu_amdkfd_get_dmabuf_info(dev->adev, args->dmabuf_fd,
+					  &dmabuf_adev, &args->size,
 					  metadata_buffer, args->metadata_size,
 					  &args->metadata_size, &flags);
 	if (r)
 		goto exit;
 
 	/* Reverse-lookup gpu_id from kgd pointer */
-	dev = kfd_device_by_kgd(dma_buf_kgd);
+	dev = kfd_device_by_adev(dmabuf_adev);
 	if (!dev) {
 		r = -EINVAL;
 		goto exit;
@@ -2050,7 +2050,7 @@ static int kfd_mmio_mmap(struct kfd_dev *dev, struct kfd_process *process,
 	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
 		return -EINVAL;
 
-	address = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd);
+	address = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->adev);
 
 	vma->vm_flags |= VM_IO | VM_DONTCOPY | VM_DONTEXPAND | VM_NORESERVE |
 				VM_DONTDUMP | VM_PFNMAP;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index cfedfb1e8596..7143550becb0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
@@ -1993,16 +1993,16 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
 		if (adev->asic_type == CHIP_ALDEBARAN) {
 			sub_type_hdr->minimum_bandwidth_mbs =
 					amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(
-							kdev->kgd, NULL, true);
+							kdev->adev, NULL, true);
 			sub_type_hdr->maximum_bandwidth_mbs =
 					sub_type_hdr->minimum_bandwidth_mbs;
 		}
 	} else {
 		sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_PCIEXPRESS;
 		sub_type_hdr->minimum_bandwidth_mbs =
-				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, true);
+				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->adev, true);
 		sub_type_hdr->maximum_bandwidth_mbs =
-				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, false);
+				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->adev, false);
 	}
 
 	sub_type_hdr->proximity_domain_from = proximity_domain;
@@ -2044,11 +2044,11 @@ static int kfd_fill_gpu_xgmi_link_to_gpu(int *avail_size,
 	sub_type_hdr->proximity_domain_from = proximity_domain_from;
 	sub_type_hdr->proximity_domain_to = proximity_domain_to;
 	sub_type_hdr->num_hops_xgmi =
-		amdgpu_amdkfd_get_xgmi_hops_count(kdev->kgd, peer_kdev->kgd);
+		amdgpu_amdkfd_get_xgmi_hops_count(kdev->adev, peer_kdev->adev);
 	sub_type_hdr->maximum_bandwidth_mbs =
-		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd, peer_kdev->kgd, false);
+		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->adev, peer_kdev->adev, false);
 	sub_type_hdr->minimum_bandwidth_mbs = sub_type_hdr->maximum_bandwidth_mbs ?
-		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd, NULL, true) : 0;
+		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->adev, NULL, true) : 0;
 
 	return 0;
 }
@@ -2114,7 +2114,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
 	cu->flags |= CRAT_CU_FLAGS_GPU_PRESENT;
 	cu->proximity_domain = proximity_domain;
 
-	amdgpu_amdkfd_get_cu_info(kdev->kgd, &cu_info);
+	amdgpu_amdkfd_get_cu_info(kdev->adev, &cu_info);
 	cu->num_simd_per_cu = cu_info.simd_per_cu;
 	cu->num_simd_cores = cu_info.simd_per_cu * cu_info.cu_active_number;
 	cu->max_waves_simd = cu_info.max_waves_per_simd;
@@ -2145,7 +2145,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
 	 * report the total FB size (public+private) as a single
 	 * private heap.
 	 */
-	amdgpu_amdkfd_get_local_mem_info(kdev->kgd, &local_mem_info);
+	amdgpu_amdkfd_get_local_mem_info(kdev->adev, &local_mem_info);
 	sub_type_hdr = (typeof(sub_type_hdr))((char *)sub_type_hdr +
 			sub_type_hdr->length);
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index a90ec8de213b..00c726123207 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -894,7 +894,7 @@ static int kfd_gws_init(struct kfd_dev *kfd)
 		|| (kfd->device_info->asic_family == CHIP_ALDEBARAN
 			&& kfd->mec2_fw_version >= 0x28))
 		ret = amdgpu_amdkfd_alloc_gws(kfd->adev,
-				amdgpu_amdkfd_get_num_gws(kfd->kgd), &kfd->gws);
+				amdgpu_amdkfd_get_num_gws(kfd->adev), &kfd->gws);
 
 	return ret;
 }
@@ -911,11 +911,11 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 	unsigned int size, map_process_packet_size;
 
 	kfd->ddev = ddev;
-	kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd,
+	kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev,
 			KGD_ENGINE_MEC1);
-	kfd->mec2_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd,
+	kfd->mec2_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev,
 			KGD_ENGINE_MEC2);
-	kfd->sdma_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd,
+	kfd->sdma_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev,
 			KGD_ENGINE_SDMA1);
 	kfd->shared_resources = *gpu_resources;
 
@@ -996,9 +996,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 		goto kfd_doorbell_error;
 	}
 
-	kfd->hive_id = amdgpu_amdkfd_get_hive_id(kfd->kgd);
+	kfd->hive_id = amdgpu_amdkfd_get_hive_id(kfd->adev);
 
-	kfd->noretry = amdgpu_amdkfd_get_noretry(kfd->kgd);
+	kfd->noretry = amdgpu_amdkfd_get_noretry(kfd->adev);
 
 	if (kfd_interrupt_init(kfd)) {
 		dev_err(kfd_device, "Error initializing interrupts\n");
@@ -1016,7 +1016,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 	 */
 	if (kfd_gws_init(kfd)) {
 		dev_err(kfd_device, "Could not allocate %d gws\n",
-			amdgpu_amdkfd_get_num_gws(kfd->kgd));
+			amdgpu_amdkfd_get_num_gws(kfd->adev));
 		goto gws_error;
 	}
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
index c021519af810..7b4118915bf6 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
@@ -100,7 +100,7 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 	struct kfd_cu_info cu_info;
 	uint32_t cu_per_sh[KFD_MAX_NUM_SE][KFD_MAX_NUM_SH_PER_SE] = {0};
 	int i, se, sh, cu;
-	amdgpu_amdkfd_get_cu_info(mm->dev->kgd, &cu_info);
+	amdgpu_amdkfd_get_cu_info(mm->dev->adev, &cu_info);
 
 	if (cu_mask_count > cu_info.cu_active_number)
 		cu_mask_count = cu_info.cu_active_number;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index c8bd062fb954..499db2099775 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -969,7 +969,7 @@ struct kfd_topology_device *kfd_topology_device_by_proximity_domain(
 struct kfd_topology_device *kfd_topology_device_by_id(uint32_t gpu_id);
 struct kfd_dev *kfd_device_by_id(uint32_t gpu_id);
 struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev);
-struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd);
+struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device *adev);
 int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_dev **kdev);
 int kfd_numa_node_to_apic_id(int numa_node_id);
 void kfd_double_confirm_iommu_support(struct kfd_dev *gpu);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
index 243dd1efcdbf..d8462bd3b4a6 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
@@ -118,7 +118,7 @@ int pqm_set_gws(struct process_queue_manager *pqm, unsigned int qid,
 		return ret;
 
 	pqn->q->gws = mem;
-	pdd->qpd.num_gws = gws ? amdgpu_amdkfd_get_num_gws(dev->kgd) : 0;
+	pdd->qpd.num_gws = gws ? amdgpu_amdkfd_get_num_gws(dev->adev) : 0;
 
 	return pqn->q->device->dqm->ops.update_queue(pqn->q->device->dqm,
 							pqn->q);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index dd593ad0614a..31610b4a3e2e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -113,7 +113,7 @@ struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev)
 	return device;
 }
 
-struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd)
+struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device *adev)
 {
 	struct kfd_topology_device *top_dev;
 	struct kfd_dev *device = NULL;
@@ -121,7 +121,7 @@ struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd)
 	down_read(&topology_lock);
 
 	list_for_each_entry(top_dev, &topology_device_list, list)
-		if (top_dev->gpu && top_dev->gpu->kgd == kgd) {
+		if (top_dev->gpu && top_dev->gpu->adev == adev) {
 			device = top_dev->gpu;
 			break;
 		}
@@ -531,7 +531,7 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr,
 		sysfs_show_32bit_prop(buffer, offs, "sdma_fw_version",
 				      dev->gpu->sdma_fw_version);
 		sysfs_show_64bit_prop(buffer, offs, "unique_id",
-				      amdgpu_amdkfd_get_unique_id(dev->gpu->kgd));
+				      amdgpu_amdkfd_get_unique_id(dev->gpu->adev));
 
 	}
 
@@ -1106,7 +1106,7 @@ static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu)
 	if (!gpu)
 		return 0;
 
-	amdgpu_amdkfd_get_local_mem_info(gpu->kgd, &local_mem_info);
+	amdgpu_amdkfd_get_local_mem_info(gpu->adev, &local_mem_info);
 
 	local_mem_size = local_mem_info.local_mem_size_private +
 			local_mem_info.local_mem_size_public;
@@ -1189,7 +1189,7 @@ static void kfd_fill_mem_clk_max_info(struct kfd_topology_device *dev)
 	 * for APUs - If CRAT from ACPI reports more than one bank, then
 	 *	all the banks will report the same mem_clk_max information
 	 */
-	amdgpu_amdkfd_get_local_mem_info(dev->gpu->kgd, &local_mem_info);
+	amdgpu_amdkfd_get_local_mem_info(dev->gpu->adev, &local_mem_info);
 
 	list_for_each_entry(mem, &dev->mem_props, list)
 		mem->mem_clk_max = local_mem_info.mem_clk_max;
@@ -1372,7 +1372,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	 * needed for the topology
 	 */
 
-	amdgpu_amdkfd_get_cu_info(dev->gpu->kgd, &cu_info);
+	amdgpu_amdkfd_get_cu_info(dev->gpu->adev, &cu_info);
 
 	strncpy(dev->node_props.name, gpu->device_info->asic_name,
 			KFD_TOPOLOGY_PUBLIC_NAME_SIZE);
@@ -1384,13 +1384,13 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	dev->node_props.vendor_id = gpu->pdev->vendor;
 	dev->node_props.device_id = gpu->pdev->device;
 	dev->node_props.capability |=
-		((amdgpu_amdkfd_get_asic_rev_id(dev->gpu->kgd) <<
+		((amdgpu_amdkfd_get_asic_rev_id(dev->gpu->adev) <<
 			HSA_CAP_ASIC_REVISION_SHIFT) &
 			HSA_CAP_ASIC_REVISION_MASK);
 	dev->node_props.location_id = pci_dev_id(gpu->pdev);
 	dev->node_props.domain = pci_domain_nr(gpu->pdev->bus);
 	dev->node_props.max_engine_clk_fcompute =
-		amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev->gpu->kgd);
+		amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev->gpu->adev);
 	dev->node_props.max_engine_clk_ccompute =
 		cpufreq_quick_get_max(0) / 1000;
 	dev->node_props.drm_render_minor =
@@ -1404,7 +1404,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 				gpu->device_info->num_sdma_queues_per_engine;
 	dev->node_props.num_gws = (dev->gpu->gws &&
 		dev->gpu->dqm->sched_policy != KFD_SCHED_POLICY_NO_HWS) ?
-		amdgpu_amdkfd_get_num_gws(dev->gpu->kgd) : 0;
+		amdgpu_amdkfd_get_num_gws(dev->gpu->adev) : 0;
 	dev->node_props.num_cp_queues = get_cp_queues_num(dev->gpu->dqm);
 
 	kfd_fill_mem_clk_max_info(dev);
-- 
2.25.1

Re: [PATCH 10/13] drm/amdkfd: replace kgd_dev in get amdgpu_amdkfd funcs

[Felix Kuehling]

Am 2021-10-19 um 5:13 p.m. schrieb Graham Sider:
> Modified definitions:
>
> - amdgpu_amdkfd_get_fw_version
> - amdgpu_amdkfd_get_local_mem_info
> - amdgpu_amdkfd_get_gpu_clock_counter
> - amdgpu_amdkfd_get_max_engine_clock_in_mhz
> - amdgpu_amdkfd_get_cu_info
> - amdgpu_amdkfd_get_dmabuf_info
> - amdgpu_amdkfd_get_vram_usage
> - amdgpu_amdkfd_get_hive_id
> - amdgpu_amdkfd_get_unique_id
> - amdgpu_amdkfd_get_mmio_remap_phys_addr
> - amdgpu_amdkfd_get_num_gws
> - amdgpu_amdkfd_get_asic_rev_id
> - amdgpu_amdkfd_get_noretry
> - amdgpu_amdkfd_get_xgmi_hops_count
> - amdgpu_amdkfd_get_xgmi_bandwidth_mbytes
> - amdgpu_amdkfd_get_pcie_bandwidth_mbytes
>
> Also replaces kfd_device_by_kgd with kfd_device_by_adev, now
> searching via adev rather than kgd.

Some of these functions are so trivial, they could probably be replaced
with direct accesses to the respective fields in adev from KFD:

  * amdgpu_amdkfd_get_hive_id
  * amdgpu_amdkfd_get_unique_id
  * amdgpu_amdkfd_get_mmio_remap_phys_addr
  * amdgpu_amdkfd_get_num_gws
  * amdgpu_amdkfd_get_asic_rev_id
  * amdgpu_amdkfd_get_noretry

I'd do that with a separate follow-up patch, though.

Regards,
  Felix


>
> Signed-off-by: Graham Sider <Graham.Sider@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c    | 73 +++++++------------
>  drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h    | 38 +++++-----
>  drivers/gpu/drm/amd/amdkfd/kfd_chardev.c      | 16 ++--
>  drivers/gpu/drm/amd/amdkfd/kfd_crat.c         | 16 ++--
>  drivers/gpu/drm/amd/amdkfd/kfd_device.c       | 14 ++--
>  drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c  |  2 +-
>  drivers/gpu/drm/amd/amdkfd/kfd_priv.h         |  2 +-
>  .../amd/amdkfd/kfd_process_queue_manager.c    |  2 +-
>  drivers/gpu/drm/amd/amdkfd/kfd_topology.c     | 18 ++---
>  9 files changed, 82 insertions(+), 99 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> index 69fc8f0d9c45..79a2e37baa59 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> @@ -358,11 +358,9 @@ void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj)
>  	amdgpu_bo_unref(&bo);
>  }
>  
> -uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
> +uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
>  				      enum kgd_engine_type type)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	switch (type) {
>  	case KGD_ENGINE_PFP:
>  		return adev->gfx.pfp_fw_version;
> @@ -395,11 +393,9 @@ uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
>  	return 0;
>  }
>  
> -void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
> +void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
>  				      struct kfd_local_mem_info *mem_info)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	memset(mem_info, 0, sizeof(*mem_info));
>  
>  	mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
> @@ -424,19 +420,15 @@ void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
>  		mem_info->mem_clk_max = 100;
>  }
>  
> -uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd)
> +uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	if (adev->gfx.funcs->get_gpu_clock_counter)
>  		return adev->gfx.funcs->get_gpu_clock_counter(adev);
>  	return 0;
>  }
>  
> -uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
> +uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	/* the sclk is in quantas of 10kHz */
>  	if (amdgpu_sriov_vf(adev))
>  		return adev->clock.default_sclk / 100;
> @@ -446,9 +438,8 @@ uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
>  		return 100;
>  }
>  
> -void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info)
> +void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev, struct kfd_cu_info *cu_info)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
>  	struct amdgpu_cu_info acu_info = adev->gfx.cu_info;
>  
>  	memset(cu_info, 0, sizeof(*cu_info));
> @@ -469,13 +460,12 @@ void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info)
>  	cu_info->lds_size = acu_info.lds_size;
>  }
>  
> -int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
> -				  struct kgd_dev **dma_buf_kgd,
> +int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
> +				  struct amdgpu_device **dmabuf_adev,
>  				  uint64_t *bo_size, void *metadata_buffer,
>  				  size_t buffer_size, uint32_t *metadata_size,
>  				  uint32_t *flags)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
>  	struct dma_buf *dma_buf;
>  	struct drm_gem_object *obj;
>  	struct amdgpu_bo *bo;
> @@ -503,8 +493,8 @@ int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
>  		goto out_put;
>  
>  	r = 0;
> -	if (dma_buf_kgd)
> -		*dma_buf_kgd = (struct kgd_dev *)adev;
> +	if (dmabuf_adev)
> +		*dmabuf_adev = adev;
>  	if (bo_size)
>  		*bo_size = amdgpu_bo_size(bo);
>  	if (metadata_buffer)
> @@ -524,32 +514,28 @@ int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
>  	return r;
>  }
>  
> -uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd)
> +uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
>  	struct ttm_resource_manager *vram_man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
>  
>  	return amdgpu_vram_mgr_usage(vram_man);
>  }
>  
> -uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd)
> +uint64_t amdgpu_amdkfd_get_hive_id(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	return adev->gmc.xgmi.hive_id;
>  }
>  
> -uint64_t amdgpu_amdkfd_get_unique_id(struct kgd_dev *kgd)
> +uint64_t amdgpu_amdkfd_get_unique_id(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	return adev->unique_id;
>  }
>  
> -uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *src)
> +uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
> +					  struct amdgpu_device *src)
>  {
> -	struct amdgpu_device *peer_adev = (struct amdgpu_device *)src;
> -	struct amdgpu_device *adev = (struct amdgpu_device *)dst;
> +	struct amdgpu_device *peer_adev = src;
> +	struct amdgpu_device *adev = dst;
>  	int ret = amdgpu_xgmi_get_hops_count(adev, peer_adev);
>  
>  	if (ret < 0) {
> @@ -561,16 +547,18 @@ uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *s
>  	return  (uint8_t)ret;
>  }
>  
> -int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct kgd_dev *src, bool is_min)
> +int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
> +					    struct amdgpu_device *src,
> +					    bool is_min)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)dst, *peer_adev;
> +	struct amdgpu_device *adev = dst, *peer_adev;
>  	int num_links;
>  
>  	if (adev->asic_type != CHIP_ALDEBARAN)
>  		return 0;
>  
>  	if (src)
> -		peer_adev = (struct amdgpu_device *)src;
> +		peer_adev = src;
>  
>  	/* num links returns 0 for indirect peers since indirect route is unknown. */
>  	num_links = is_min ? 1 : amdgpu_xgmi_get_num_links(adev, peer_adev);
> @@ -585,9 +573,8 @@ int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct kgd_dev
>  	return (num_links * 16 * 25000)/BITS_PER_BYTE;
>  }
>  
> -int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min)
> +int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)dev;
>  	int num_lanes_shift = (is_min ? ffs(adev->pm.pcie_mlw_mask) :
>  							fls(adev->pm.pcie_mlw_mask)) - 1;
>  	int gen_speed_shift = (is_min ? ffs(adev->pm.pcie_gen_mask &
> @@ -643,31 +630,23 @@ int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min)
>  	return (num_lanes_factor * gen_speed_mbits_factor)/BITS_PER_BYTE;
>  }
>  
> -uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev *kgd)
> +uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	return adev->rmmio_remap.bus_addr;
>  }
>  
> -uint32_t amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd)
> +uint32_t amdgpu_amdkfd_get_num_gws(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	return adev->gds.gws_size;
>  }
>  
> -uint32_t amdgpu_amdkfd_get_asic_rev_id(struct kgd_dev *kgd)
> +uint32_t amdgpu_amdkfd_get_asic_rev_id(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	return adev->rev_id;
>  }
>  
> -int amdgpu_amdkfd_get_noretry(struct kgd_dev *kgd)
> +int amdgpu_amdkfd_get_noretry(struct amdgpu_device *adev)
>  {
> -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> -
>  	return adev->gmc.noretry;
>  }
>  
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> index 8d5c18953723..7e3697a7a5cd 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> @@ -209,29 +209,33 @@ int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
>  void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj);
>  int amdgpu_amdkfd_add_gws_to_process(void *info, void *gws, struct kgd_mem **mem);
>  int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem);
> -uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
> +uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
>  				      enum kgd_engine_type type);
> -void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
> +void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
>  				      struct kfd_local_mem_info *mem_info);
> -uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd);
> +uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev);
>  
> -uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd);
> -void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info);
> -int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
> -				  struct kgd_dev **dmabuf_kgd,
> +uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev);
> +void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev,
> +			       struct kfd_cu_info *cu_info);
> +int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
> +				  struct amdgpu_device **dmabuf_adev,
>  				  uint64_t *bo_size, void *metadata_buffer,
>  				  size_t buffer_size, uint32_t *metadata_size,
>  				  uint32_t *flags);
> -uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd);
> -uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd);
> -uint64_t amdgpu_amdkfd_get_unique_id(struct kgd_dev *kgd);
> -uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev *kgd);
> -uint32_t amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd);
> -uint32_t amdgpu_amdkfd_get_asic_rev_id(struct kgd_dev *kgd);
> -int amdgpu_amdkfd_get_noretry(struct kgd_dev *kgd);
> -uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *src);
> -int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct kgd_dev *src, bool is_min);
> -int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool is_min);
> +uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device *adev);
> +uint64_t amdgpu_amdkfd_get_hive_id(struct amdgpu_device *adev);
> +uint64_t amdgpu_amdkfd_get_unique_id(struct amdgpu_device *adev);
> +uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct amdgpu_device *adev);
> +uint32_t amdgpu_amdkfd_get_num_gws(struct amdgpu_device *adev);
> +uint32_t amdgpu_amdkfd_get_asic_rev_id(struct amdgpu_device *adev);
> +int amdgpu_amdkfd_get_noretry(struct amdgpu_device *adev);
> +uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
> +					  struct amdgpu_device *src);
> +int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
> +					    struct amdgpu_device *src,
> +					    bool is_min);
> +int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min);
>  
>  /* Read user wptr from a specified user address space with page fault
>   * disabled. The memory must be pinned and mapped to the hardware when
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> index 7c4f14410a74..47acfef1aebd 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> @@ -850,7 +850,7 @@ static int kfd_ioctl_get_clock_counters(struct file *filep,
>  	dev = kfd_device_by_id(args->gpu_id);
>  	if (dev)
>  		/* Reading GPU clock counter from KGD */
> -		args->gpu_clock_counter = amdgpu_amdkfd_get_gpu_clock_counter(dev->kgd);
> +		args->gpu_clock_counter = amdgpu_amdkfd_get_gpu_clock_counter(dev->adev);
>  	else
>  		/* Node without GPU resource */
>  		args->gpu_clock_counter = 0;
> @@ -1237,7 +1237,7 @@ bool kfd_dev_is_large_bar(struct kfd_dev *dev)
>  	if (dev->use_iommu_v2)
>  		return false;
>  
> -	amdgpu_amdkfd_get_local_mem_info(dev->kgd, &mem_info);
> +	amdgpu_amdkfd_get_local_mem_info(dev->adev, &mem_info);
>  	if (mem_info.local_mem_size_private == 0 &&
>  			mem_info.local_mem_size_public > 0)
>  		return true;
> @@ -1306,7 +1306,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
>  			err = -EINVAL;
>  			goto err_unlock;
>  		}
> -		offset = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd);
> +		offset = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->adev);
>  		if (!offset) {
>  			err = -ENOMEM;
>  			goto err_unlock;
> @@ -1664,7 +1664,7 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
>  {
>  	struct kfd_ioctl_get_dmabuf_info_args *args = data;
>  	struct kfd_dev *dev = NULL;
> -	struct kgd_dev *dma_buf_kgd;
> +	struct amdgpu_device *dmabuf_adev;
>  	void *metadata_buffer = NULL;
>  	uint32_t flags;
>  	unsigned int i;
> @@ -1684,15 +1684,15 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
>  	}
>  
>  	/* Get dmabuf info from KGD */
> -	r = amdgpu_amdkfd_get_dmabuf_info(dev->kgd, args->dmabuf_fd,
> -					  &dma_buf_kgd, &args->size,
> +	r = amdgpu_amdkfd_get_dmabuf_info(dev->adev, args->dmabuf_fd,
> +					  &dmabuf_adev, &args->size,
>  					  metadata_buffer, args->metadata_size,
>  					  &args->metadata_size, &flags);
>  	if (r)
>  		goto exit;
>  
>  	/* Reverse-lookup gpu_id from kgd pointer */
> -	dev = kfd_device_by_kgd(dma_buf_kgd);
> +	dev = kfd_device_by_adev(dmabuf_adev);
>  	if (!dev) {
>  		r = -EINVAL;
>  		goto exit;
> @@ -2050,7 +2050,7 @@ static int kfd_mmio_mmap(struct kfd_dev *dev, struct kfd_process *process,
>  	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
>  		return -EINVAL;
>  
> -	address = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd);
> +	address = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->adev);
>  
>  	vma->vm_flags |= VM_IO | VM_DONTCOPY | VM_DONTEXPAND | VM_NORESERVE |
>  				VM_DONTDUMP | VM_PFNMAP;
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
> index cfedfb1e8596..7143550becb0 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
> @@ -1993,16 +1993,16 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
>  		if (adev->asic_type == CHIP_ALDEBARAN) {
>  			sub_type_hdr->minimum_bandwidth_mbs =
>  					amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(
> -							kdev->kgd, NULL, true);
> +							kdev->adev, NULL, true);
>  			sub_type_hdr->maximum_bandwidth_mbs =
>  					sub_type_hdr->minimum_bandwidth_mbs;
>  		}
>  	} else {
>  		sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_PCIEXPRESS;
>  		sub_type_hdr->minimum_bandwidth_mbs =
> -				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, true);
> +				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->adev, true);
>  		sub_type_hdr->maximum_bandwidth_mbs =
> -				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, false);
> +				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->adev, false);
>  	}
>  
>  	sub_type_hdr->proximity_domain_from = proximity_domain;
> @@ -2044,11 +2044,11 @@ static int kfd_fill_gpu_xgmi_link_to_gpu(int *avail_size,
>  	sub_type_hdr->proximity_domain_from = proximity_domain_from;
>  	sub_type_hdr->proximity_domain_to = proximity_domain_to;
>  	sub_type_hdr->num_hops_xgmi =
> -		amdgpu_amdkfd_get_xgmi_hops_count(kdev->kgd, peer_kdev->kgd);
> +		amdgpu_amdkfd_get_xgmi_hops_count(kdev->adev, peer_kdev->adev);
>  	sub_type_hdr->maximum_bandwidth_mbs =
> -		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd, peer_kdev->kgd, false);
> +		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->adev, peer_kdev->adev, false);
>  	sub_type_hdr->minimum_bandwidth_mbs = sub_type_hdr->maximum_bandwidth_mbs ?
> -		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd, NULL, true) : 0;
> +		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->adev, NULL, true) : 0;
>  
>  	return 0;
>  }
> @@ -2114,7 +2114,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
>  	cu->flags |= CRAT_CU_FLAGS_GPU_PRESENT;
>  	cu->proximity_domain = proximity_domain;
>  
> -	amdgpu_amdkfd_get_cu_info(kdev->kgd, &cu_info);
> +	amdgpu_amdkfd_get_cu_info(kdev->adev, &cu_info);
>  	cu->num_simd_per_cu = cu_info.simd_per_cu;
>  	cu->num_simd_cores = cu_info.simd_per_cu * cu_info.cu_active_number;
>  	cu->max_waves_simd = cu_info.max_waves_per_simd;
> @@ -2145,7 +2145,7 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
>  	 * report the total FB size (public+private) as a single
>  	 * private heap.
>  	 */
> -	amdgpu_amdkfd_get_local_mem_info(kdev->kgd, &local_mem_info);
> +	amdgpu_amdkfd_get_local_mem_info(kdev->adev, &local_mem_info);
>  	sub_type_hdr = (typeof(sub_type_hdr))((char *)sub_type_hdr +
>  			sub_type_hdr->length);
>  
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> index a90ec8de213b..00c726123207 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> @@ -894,7 +894,7 @@ static int kfd_gws_init(struct kfd_dev *kfd)
>  		|| (kfd->device_info->asic_family == CHIP_ALDEBARAN
>  			&& kfd->mec2_fw_version >= 0x28))
>  		ret = amdgpu_amdkfd_alloc_gws(kfd->adev,
> -				amdgpu_amdkfd_get_num_gws(kfd->kgd), &kfd->gws);
> +				amdgpu_amdkfd_get_num_gws(kfd->adev), &kfd->gws);
>  
>  	return ret;
>  }
> @@ -911,11 +911,11 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
>  	unsigned int size, map_process_packet_size;
>  
>  	kfd->ddev = ddev;
> -	kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd,
> +	kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev,
>  			KGD_ENGINE_MEC1);
> -	kfd->mec2_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd,
> +	kfd->mec2_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev,
>  			KGD_ENGINE_MEC2);
> -	kfd->sdma_fw_version = amdgpu_amdkfd_get_fw_version(kfd->kgd,
> +	kfd->sdma_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev,
>  			KGD_ENGINE_SDMA1);
>  	kfd->shared_resources = *gpu_resources;
>  
> @@ -996,9 +996,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
>  		goto kfd_doorbell_error;
>  	}
>  
> -	kfd->hive_id = amdgpu_amdkfd_get_hive_id(kfd->kgd);
> +	kfd->hive_id = amdgpu_amdkfd_get_hive_id(kfd->adev);
>  
> -	kfd->noretry = amdgpu_amdkfd_get_noretry(kfd->kgd);
> +	kfd->noretry = amdgpu_amdkfd_get_noretry(kfd->adev);
>  
>  	if (kfd_interrupt_init(kfd)) {
>  		dev_err(kfd_device, "Error initializing interrupts\n");
> @@ -1016,7 +1016,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
>  	 */
>  	if (kfd_gws_init(kfd)) {
>  		dev_err(kfd_device, "Could not allocate %d gws\n",
> -			amdgpu_amdkfd_get_num_gws(kfd->kgd));
> +			amdgpu_amdkfd_get_num_gws(kfd->adev));
>  		goto gws_error;
>  	}
>  
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
> index c021519af810..7b4118915bf6 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
> @@ -100,7 +100,7 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
>  	struct kfd_cu_info cu_info;
>  	uint32_t cu_per_sh[KFD_MAX_NUM_SE][KFD_MAX_NUM_SH_PER_SE] = {0};
>  	int i, se, sh, cu;
> -	amdgpu_amdkfd_get_cu_info(mm->dev->kgd, &cu_info);
> +	amdgpu_amdkfd_get_cu_info(mm->dev->adev, &cu_info);
>  
>  	if (cu_mask_count > cu_info.cu_active_number)
>  		cu_mask_count = cu_info.cu_active_number;
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> index c8bd062fb954..499db2099775 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> @@ -969,7 +969,7 @@ struct kfd_topology_device *kfd_topology_device_by_proximity_domain(
>  struct kfd_topology_device *kfd_topology_device_by_id(uint32_t gpu_id);
>  struct kfd_dev *kfd_device_by_id(uint32_t gpu_id);
>  struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev);
> -struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd);
> +struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device *adev);
>  int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_dev **kdev);
>  int kfd_numa_node_to_apic_id(int numa_node_id);
>  void kfd_double_confirm_iommu_support(struct kfd_dev *gpu);
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
> index 243dd1efcdbf..d8462bd3b4a6 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
> @@ -118,7 +118,7 @@ int pqm_set_gws(struct process_queue_manager *pqm, unsigned int qid,
>  		return ret;
>  
>  	pqn->q->gws = mem;
> -	pdd->qpd.num_gws = gws ? amdgpu_amdkfd_get_num_gws(dev->kgd) : 0;
> +	pdd->qpd.num_gws = gws ? amdgpu_amdkfd_get_num_gws(dev->adev) : 0;
>  
>  	return pqn->q->device->dqm->ops.update_queue(pqn->q->device->dqm,
>  							pqn->q);
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
> index dd593ad0614a..31610b4a3e2e 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
> @@ -113,7 +113,7 @@ struct kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev)
>  	return device;
>  }
>  
> -struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd)
> +struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device *adev)
>  {
>  	struct kfd_topology_device *top_dev;
>  	struct kfd_dev *device = NULL;
> @@ -121,7 +121,7 @@ struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd)
>  	down_read(&topology_lock);
>  
>  	list_for_each_entry(top_dev, &topology_device_list, list)
> -		if (top_dev->gpu && top_dev->gpu->kgd == kgd) {
> +		if (top_dev->gpu && top_dev->gpu->adev == adev) {
>  			device = top_dev->gpu;
>  			break;
>  		}
> @@ -531,7 +531,7 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr,
>  		sysfs_show_32bit_prop(buffer, offs, "sdma_fw_version",
>  				      dev->gpu->sdma_fw_version);
>  		sysfs_show_64bit_prop(buffer, offs, "unique_id",
> -				      amdgpu_amdkfd_get_unique_id(dev->gpu->kgd));
> +				      amdgpu_amdkfd_get_unique_id(dev->gpu->adev));
>  
>  	}
>  
> @@ -1106,7 +1106,7 @@ static uint32_t kfd_generate_gpu_id(struct kfd_dev *gpu)
>  	if (!gpu)
>  		return 0;
>  
> -	amdgpu_amdkfd_get_local_mem_info(gpu->kgd, &local_mem_info);
> +	amdgpu_amdkfd_get_local_mem_info(gpu->adev, &local_mem_info);
>  
>  	local_mem_size = local_mem_info.local_mem_size_private +
>  			local_mem_info.local_mem_size_public;
> @@ -1189,7 +1189,7 @@ static void kfd_fill_mem_clk_max_info(struct kfd_topology_device *dev)
>  	 * for APUs - If CRAT from ACPI reports more than one bank, then
>  	 *	all the banks will report the same mem_clk_max information
>  	 */
> -	amdgpu_amdkfd_get_local_mem_info(dev->gpu->kgd, &local_mem_info);
> +	amdgpu_amdkfd_get_local_mem_info(dev->gpu->adev, &local_mem_info);
>  
>  	list_for_each_entry(mem, &dev->mem_props, list)
>  		mem->mem_clk_max = local_mem_info.mem_clk_max;
> @@ -1372,7 +1372,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
>  	 * needed for the topology
>  	 */
>  
> -	amdgpu_amdkfd_get_cu_info(dev->gpu->kgd, &cu_info);
> +	amdgpu_amdkfd_get_cu_info(dev->gpu->adev, &cu_info);
>  
>  	strncpy(dev->node_props.name, gpu->device_info->asic_name,
>  			KFD_TOPOLOGY_PUBLIC_NAME_SIZE);
> @@ -1384,13 +1384,13 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
>  	dev->node_props.vendor_id = gpu->pdev->vendor;
>  	dev->node_props.device_id = gpu->pdev->device;
>  	dev->node_props.capability |=
> -		((amdgpu_amdkfd_get_asic_rev_id(dev->gpu->kgd) <<
> +		((amdgpu_amdkfd_get_asic_rev_id(dev->gpu->adev) <<
>  			HSA_CAP_ASIC_REVISION_SHIFT) &
>  			HSA_CAP_ASIC_REVISION_MASK);
>  	dev->node_props.location_id = pci_dev_id(gpu->pdev);
>  	dev->node_props.domain = pci_domain_nr(gpu->pdev->bus);
>  	dev->node_props.max_engine_clk_fcompute =
> -		amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev->gpu->kgd);
> +		amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev->gpu->adev);
>  	dev->node_props.max_engine_clk_ccompute =
>  		cpufreq_quick_get_max(0) / 1000;
>  	dev->node_props.drm_render_minor =
> @@ -1404,7 +1404,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
>  				gpu->device_info->num_sdma_queues_per_engine;
>  	dev->node_props.num_gws = (dev->gpu->gws &&
>  		dev->gpu->dqm->sched_policy != KFD_SCHED_POLICY_NO_HWS) ?
> -		amdgpu_amdkfd_get_num_gws(dev->gpu->kgd) : 0;
> +		amdgpu_amdkfd_get_num_gws(dev->gpu->adev) : 0;
>  	dev->node_props.num_cp_queues = get_cp_queues_num(dev->gpu->dqm);
>  
>  	kfd_fill_mem_clk_max_info(dev);

RE: [PATCH 10/13] drm/amdkfd: replace kgd_dev in get amdgpu_amdkfd funcs

[Sider, Graham]

[AMD Official Use Only]

> -----Original Message-----
> From: Kuehling, Felix <Felix.Kuehling@amd.com>
> Sent: Tuesday, October 26, 2021 5:24 PM
> To: Sider, Graham <Graham.Sider@amd.com>; amd-
> gfx@lists.freedesktop.org
> Cc: Joshi, Mukul <Mukul.Joshi@amd.com>
> Subject: Re: [PATCH 10/13] drm/amdkfd: replace kgd_dev in get
> amdgpu_amdkfd funcs
> 
> Am 2021-10-19 um 5:13 p.m. schrieb Graham Sider:
> > Modified definitions:
> >
> > - amdgpu_amdkfd_get_fw_version
> > - amdgpu_amdkfd_get_local_mem_info
> > - amdgpu_amdkfd_get_gpu_clock_counter
> > - amdgpu_amdkfd_get_max_engine_clock_in_mhz
> > - amdgpu_amdkfd_get_cu_info
> > - amdgpu_amdkfd_get_dmabuf_info
> > - amdgpu_amdkfd_get_vram_usage
> > - amdgpu_amdkfd_get_hive_id
> > - amdgpu_amdkfd_get_unique_id
> > - amdgpu_amdkfd_get_mmio_remap_phys_addr
> > - amdgpu_amdkfd_get_num_gws
> > - amdgpu_amdkfd_get_asic_rev_id
> > - amdgpu_amdkfd_get_noretry
> > - amdgpu_amdkfd_get_xgmi_hops_count
> > - amdgpu_amdkfd_get_xgmi_bandwidth_mbytes
> > - amdgpu_amdkfd_get_pcie_bandwidth_mbytes
> >
> > Also replaces kfd_device_by_kgd with kfd_device_by_adev, now
> searching
> > via adev rather than kgd.
> 
> Some of these functions are so trivial, they could probably be replaced with
> direct accesses to the respective fields in adev from KFD:
> 
>   * amdgpu_amdkfd_get_hive_id
>   * amdgpu_amdkfd_get_unique_id
>   * amdgpu_amdkfd_get_mmio_remap_phys_addr
>   * amdgpu_amdkfd_get_num_gws
>   * amdgpu_amdkfd_get_asic_rev_id
>   * amdgpu_amdkfd_get_noretry
> 
> I'd do that with a separate follow-up patch, though.
> 
> Regards,
>   Felix
> 

Sounds good, I'll go ahead and replace these ones mentioned in a follow-up. Thanks!

Best,
Graham

> 
> >
> > Signed-off-by: Graham Sider <Graham.Sider@amd.com>
> > ---
> >  drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c    | 73 +++++++---------
> ---
> >  drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h    | 38 +++++-----
> >  drivers/gpu/drm/amd/amdkfd/kfd_chardev.c      | 16 ++--
> >  drivers/gpu/drm/amd/amdkfd/kfd_crat.c         | 16 ++--
> >  drivers/gpu/drm/amd/amdkfd/kfd_device.c       | 14 ++--
> >  drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c  |  2 +-
> >  drivers/gpu/drm/amd/amdkfd/kfd_priv.h         |  2 +-
> >  .../amd/amdkfd/kfd_process_queue_manager.c    |  2 +-
> >  drivers/gpu/drm/amd/amdkfd/kfd_topology.c     | 18 ++---
> >  9 files changed, 82 insertions(+), 99 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> > b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> > index 69fc8f0d9c45..79a2e37baa59 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> > @@ -358,11 +358,9 @@ void amdgpu_amdkfd_free_gws(struct
> amdgpu_device *adev, void *mem_obj)
> >  	amdgpu_bo_unref(&bo);
> >  }
> >
> > -uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
> > +uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
> >  				      enum kgd_engine_type type)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	switch (type) {
> >  	case KGD_ENGINE_PFP:
> >  		return adev->gfx.pfp_fw_version;
> > @@ -395,11 +393,9 @@ uint32_t amdgpu_amdkfd_get_fw_version(struct
> kgd_dev *kgd,
> >  	return 0;
> >  }
> >
> > -void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
> > +void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device
> *adev,
> >  				      struct kfd_local_mem_info *mem_info)  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	memset(mem_info, 0, sizeof(*mem_info));
> >
> >  	mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
> @@
> > -424,19 +420,15 @@ void amdgpu_amdkfd_get_local_mem_info(struct
> kgd_dev *kgd,
> >  		mem_info->mem_clk_max = 100;
> >  }
> >
> > -uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd)
> > +uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct
> amdgpu_device
> > +*adev)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	if (adev->gfx.funcs->get_gpu_clock_counter)
> >  		return adev->gfx.funcs->get_gpu_clock_counter(adev);
> >  	return 0;
> >  }
> >
> > -uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct
> kgd_dev
> > *kgd)
> > +uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct
> > +amdgpu_device *adev)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	/* the sclk is in quantas of 10kHz */
> >  	if (amdgpu_sriov_vf(adev))
> >  		return adev->clock.default_sclk / 100; @@ -446,9 +438,8 @@
> uint32_t
> > amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
> >  		return 100;
> >  }
> >
> > -void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct
> > kfd_cu_info *cu_info)
> > +void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev, struct
> > +kfd_cu_info *cu_info)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> >  	struct amdgpu_cu_info acu_info = adev->gfx.cu_info;
> >
> >  	memset(cu_info, 0, sizeof(*cu_info)); @@ -469,13 +460,12 @@ void
> > amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info
> *cu_info)
> >  	cu_info->lds_size = acu_info.lds_size;  }
> >
> > -int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int
> dma_buf_fd,
> > -				  struct kgd_dev **dma_buf_kgd,
> > +int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int
> dma_buf_fd,
> > +				  struct amdgpu_device **dmabuf_adev,
> >  				  uint64_t *bo_size, void *metadata_buffer,
> >  				  size_t buffer_size, uint32_t *metadata_size,
> >  				  uint32_t *flags)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> >  	struct dma_buf *dma_buf;
> >  	struct drm_gem_object *obj;
> >  	struct amdgpu_bo *bo;
> > @@ -503,8 +493,8 @@ int amdgpu_amdkfd_get_dmabuf_info(struct
> kgd_dev *kgd, int dma_buf_fd,
> >  		goto out_put;
> >
> >  	r = 0;
> > -	if (dma_buf_kgd)
> > -		*dma_buf_kgd = (struct kgd_dev *)adev;
> > +	if (dmabuf_adev)
> > +		*dmabuf_adev = adev;
> >  	if (bo_size)
> >  		*bo_size = amdgpu_bo_size(bo);
> >  	if (metadata_buffer)
> > @@ -524,32 +514,28 @@ int amdgpu_amdkfd_get_dmabuf_info(struct
> kgd_dev *kgd, int dma_buf_fd,
> >  	return r;
> >  }
> >
> > -uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd)
> > +uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device
> *adev)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> >  	struct ttm_resource_manager *vram_man =
> > ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
> >
> >  	return amdgpu_vram_mgr_usage(vram_man);  }
> >
> > -uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd)
> > +uint64_t amdgpu_amdkfd_get_hive_id(struct amdgpu_device *adev)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	return adev->gmc.xgmi.hive_id;
> >  }
> >
> > -uint64_t amdgpu_amdkfd_get_unique_id(struct kgd_dev *kgd)
> > +uint64_t amdgpu_amdkfd_get_unique_id(struct amdgpu_device *adev)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	return adev->unique_id;
> >  }
> >
> > -uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst,
> struct
> > kgd_dev *src)
> > +uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device
> *dst,
> > +					  struct amdgpu_device *src)
> >  {
> > -	struct amdgpu_device *peer_adev = (struct amdgpu_device *)src;
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)dst;
> > +	struct amdgpu_device *peer_adev = src;
> > +	struct amdgpu_device *adev = dst;
> >  	int ret = amdgpu_xgmi_get_hops_count(adev, peer_adev);
> >
> >  	if (ret < 0) {
> > @@ -561,16 +547,18 @@ uint8_t
> amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct
> kgd_dev *s
> >  	return  (uint8_t)ret;
> >  }
> >
> > -int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst,
> > struct kgd_dev *src, bool is_min)
> > +int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct
> amdgpu_device *dst,
> > +					    struct amdgpu_device *src,
> > +					    bool is_min)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)dst,
> *peer_adev;
> > +	struct amdgpu_device *adev = dst, *peer_adev;
> >  	int num_links;
> >
> >  	if (adev->asic_type != CHIP_ALDEBARAN)
> >  		return 0;
> >
> >  	if (src)
> > -		peer_adev = (struct amdgpu_device *)src;
> > +		peer_adev = src;
> >
> >  	/* num links returns 0 for indirect peers since indirect route is
> unknown. */
> >  	num_links = is_min ? 1 : amdgpu_xgmi_get_num_links(adev,
> peer_adev);
> > @@ -585,9 +573,8 @@ int
> amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev *dst, struct
> kgd_dev
> >  	return (num_links * 16 * 25000)/BITS_PER_BYTE;  }
> >
> > -int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev,
> bool
> > is_min)
> > +int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct
> amdgpu_device
> > +*adev, bool is_min)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)dev;
> >  	int num_lanes_shift = (is_min ? ffs(adev->pm.pcie_mlw_mask) :
> >  							fls(adev-
> >pm.pcie_mlw_mask)) - 1;
> >  	int gen_speed_shift = (is_min ? ffs(adev->pm.pcie_gen_mask & @@
> > -643,31 +630,23 @@ int
> amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool
> is_min)
> >  	return (num_lanes_factor *
> gen_speed_mbits_factor)/BITS_PER_BYTE;
> >  }
> >
> > -uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev
> *kgd)
> > +uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct
> amdgpu_device
> > +*adev)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	return adev->rmmio_remap.bus_addr;
> >  }
> >
> > -uint32_t amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd)
> > +uint32_t amdgpu_amdkfd_get_num_gws(struct amdgpu_device *adev)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	return adev->gds.gws_size;
> >  }
> >
> > -uint32_t amdgpu_amdkfd_get_asic_rev_id(struct kgd_dev *kgd)
> > +uint32_t amdgpu_amdkfd_get_asic_rev_id(struct amdgpu_device *adev)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	return adev->rev_id;
> >  }
> >
> > -int amdgpu_amdkfd_get_noretry(struct kgd_dev *kgd)
> > +int amdgpu_amdkfd_get_noretry(struct amdgpu_device *adev)
> >  {
> > -	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
> > -
> >  	return adev->gmc.noretry;
> >  }
> >
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> > b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> > index 8d5c18953723..7e3697a7a5cd 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
> > @@ -209,29 +209,33 @@ int amdgpu_amdkfd_alloc_gws(struct
> amdgpu_device
> > *adev, size_t size,  void amdgpu_amdkfd_free_gws(struct amdgpu_device
> > *adev, void *mem_obj);  int amdgpu_amdkfd_add_gws_to_process(void
> > *info, void *gws, struct kgd_mem **mem);  int
> > amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem);
> > -uint32_t amdgpu_amdkfd_get_fw_version(struct kgd_dev *kgd,
> > +uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
> >  				      enum kgd_engine_type type); -void
> > amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
> > +void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device
> *adev,
> >  				      struct kfd_local_mem_info *mem_info); -
> uint64_t
> > amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd);
> > +uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct
> amdgpu_device
> > +*adev);
> >
> > -uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct
> kgd_dev
> > *kgd); -void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct
> > kfd_cu_info *cu_info); -int amdgpu_amdkfd_get_dmabuf_info(struct
> kgd_dev *kgd, int dma_buf_fd,
> > -				  struct kgd_dev **dmabuf_kgd,
> > +uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct
> > +amdgpu_device *adev); void amdgpu_amdkfd_get_cu_info(struct
> amdgpu_device *adev,
> > +			       struct kfd_cu_info *cu_info); int
> > +amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int
> dma_buf_fd,
> > +				  struct amdgpu_device **dmabuf_adev,
> >  				  uint64_t *bo_size, void *metadata_buffer,
> >  				  size_t buffer_size, uint32_t *metadata_size,
> >  				  uint32_t *flags);
> > -uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd); -
> uint64_t
> > amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd); -uint64_t
> > amdgpu_amdkfd_get_unique_id(struct kgd_dev *kgd); -uint64_t
> > amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev *kgd); -
> uint32_t
> > amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd); -uint32_t
> > amdgpu_amdkfd_get_asic_rev_id(struct kgd_dev *kgd); -int
> > amdgpu_amdkfd_get_noretry(struct kgd_dev *kgd); -uint8_t
> > amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct
> kgd_dev
> > *src); -int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct kgd_dev
> > *dst, struct kgd_dev *src, bool is_min); -int
> > amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct kgd_dev *dev, bool
> > is_min);
> > +uint64_t amdgpu_amdkfd_get_vram_usage(struct amdgpu_device
> *adev);
> > +uint64_t amdgpu_amdkfd_get_hive_id(struct amdgpu_device *adev);
> > +uint64_t amdgpu_amdkfd_get_unique_id(struct amdgpu_device *adev);
> > +uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct
> amdgpu_device
> > +*adev); uint32_t amdgpu_amdkfd_get_num_gws(struct amdgpu_device
> > +*adev); uint32_t amdgpu_amdkfd_get_asic_rev_id(struct
> amdgpu_device
> > +*adev); int amdgpu_amdkfd_get_noretry(struct amdgpu_device *adev);
> > +uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device
> *dst,
> > +					  struct amdgpu_device *src);
> > +int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct
> amdgpu_device *dst,
> > +					    struct amdgpu_device *src,
> > +					    bool is_min);
> > +int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct
> amdgpu_device
> > +*adev, bool is_min);
> >
> >  /* Read user wptr from a specified user address space with page fault
> >   * disabled. The memory must be pinned and mapped to the hardware
> > when diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> > b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> > index 7c4f14410a74..47acfef1aebd 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> > @@ -850,7 +850,7 @@ static int kfd_ioctl_get_clock_counters(struct file
> *filep,
> >  	dev = kfd_device_by_id(args->gpu_id);
> >  	if (dev)
> >  		/* Reading GPU clock counter from KGD */
> > -		args->gpu_clock_counter =
> amdgpu_amdkfd_get_gpu_clock_counter(dev->kgd);
> > +		args->gpu_clock_counter =
> > +amdgpu_amdkfd_get_gpu_clock_counter(dev->adev);
> >  	else
> >  		/* Node without GPU resource */
> >  		args->gpu_clock_counter = 0;
> > @@ -1237,7 +1237,7 @@ bool kfd_dev_is_large_bar(struct kfd_dev *dev)
> >  	if (dev->use_iommu_v2)
> >  		return false;
> >
> > -	amdgpu_amdkfd_get_local_mem_info(dev->kgd, &mem_info);
> > +	amdgpu_amdkfd_get_local_mem_info(dev->adev, &mem_info);
> >  	if (mem_info.local_mem_size_private == 0 &&
> >  			mem_info.local_mem_size_public > 0)
> >  		return true;
> > @@ -1306,7 +1306,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct
> file *filep,
> >  			err = -EINVAL;
> >  			goto err_unlock;
> >  		}
> > -		offset =
> amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd);
> > +		offset =
> amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->adev);
> >  		if (!offset) {
> >  			err = -ENOMEM;
> >  			goto err_unlock;
> > @@ -1664,7 +1664,7 @@ static int kfd_ioctl_get_dmabuf_info(struct file
> > *filep,  {
> >  	struct kfd_ioctl_get_dmabuf_info_args *args = data;
> >  	struct kfd_dev *dev = NULL;
> > -	struct kgd_dev *dma_buf_kgd;
> > +	struct amdgpu_device *dmabuf_adev;
> >  	void *metadata_buffer = NULL;
> >  	uint32_t flags;
> >  	unsigned int i;
> > @@ -1684,15 +1684,15 @@ static int kfd_ioctl_get_dmabuf_info(struct file
> *filep,
> >  	}
> >
> >  	/* Get dmabuf info from KGD */
> > -	r = amdgpu_amdkfd_get_dmabuf_info(dev->kgd, args->dmabuf_fd,
> > -					  &dma_buf_kgd, &args->size,
> > +	r = amdgpu_amdkfd_get_dmabuf_info(dev->adev, args-
> >dmabuf_fd,
> > +					  &dmabuf_adev, &args->size,
> >  					  metadata_buffer, args-
> >metadata_size,
> >  					  &args->metadata_size, &flags);
> >  	if (r)
> >  		goto exit;
> >
> >  	/* Reverse-lookup gpu_id from kgd pointer */
> > -	dev = kfd_device_by_kgd(dma_buf_kgd);
> > +	dev = kfd_device_by_adev(dmabuf_adev);
> >  	if (!dev) {
> >  		r = -EINVAL;
> >  		goto exit;
> > @@ -2050,7 +2050,7 @@ static int kfd_mmio_mmap(struct kfd_dev *dev,
> struct kfd_process *process,
> >  	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
> >  		return -EINVAL;
> >
> > -	address = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev-
> >kgd);
> > +	address = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev-
> >adev);
> >
> >  	vma->vm_flags |= VM_IO | VM_DONTCOPY | VM_DONTEXPAND |
> VM_NORESERVE |
> >  				VM_DONTDUMP | VM_PFNMAP;
> > diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
> > b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
> > index cfedfb1e8596..7143550becb0 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
> > @@ -1993,16 +1993,16 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int
> *avail_size,
> >  		if (adev->asic_type == CHIP_ALDEBARAN) {
> >  			sub_type_hdr->minimum_bandwidth_mbs =
> >
> 	amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(
> > -							kdev->kgd, NULL,
> true);
> > +							kdev->adev, NULL,
> true);
> >  			sub_type_hdr->maximum_bandwidth_mbs =
> >  					sub_type_hdr-
> >minimum_bandwidth_mbs;
> >  		}
> >  	} else {
> >  		sub_type_hdr->io_interface_type =
> CRAT_IOLINK_TYPE_PCIEXPRESS;
> >  		sub_type_hdr->minimum_bandwidth_mbs =
> > -
> 	amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, true);
> > +
> 	amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->adev, true);
> >  		sub_type_hdr->maximum_bandwidth_mbs =
> > -
> 	amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, false);
> > +
> 	amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->adev, false);
> >  	}
> >
> >  	sub_type_hdr->proximity_domain_from = proximity_domain; @@ -
> 2044,11
> > +2044,11 @@ static int kfd_fill_gpu_xgmi_link_to_gpu(int *avail_size,
> >  	sub_type_hdr->proximity_domain_from = proximity_domain_from;
> >  	sub_type_hdr->proximity_domain_to = proximity_domain_to;
> >  	sub_type_hdr->num_hops_xgmi =
> > -		amdgpu_amdkfd_get_xgmi_hops_count(kdev->kgd,
> peer_kdev->kgd);
> > +		amdgpu_amdkfd_get_xgmi_hops_count(kdev->adev,
> peer_kdev->adev);
> >  	sub_type_hdr->maximum_bandwidth_mbs =
> > -		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd,
> peer_kdev->kgd, false);
> > +		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev-
> >adev,
> > +peer_kdev->adev, false);
> >  	sub_type_hdr->minimum_bandwidth_mbs = sub_type_hdr-
> >maximum_bandwidth_mbs ?
> > -		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd,
> NULL, true) : 0;
> > +		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev-
> >adev, NULL, true) :
> > +0;
> >
> >  	return 0;
> >  }
> > @@ -2114,7 +2114,7 @@ static int kfd_create_vcrat_image_gpu(void
> *pcrat_image,
> >  	cu->flags |= CRAT_CU_FLAGS_GPU_PRESENT;
> >  	cu->proximity_domain = proximity_domain;
> >
> > -	amdgpu_amdkfd_get_cu_info(kdev->kgd, &cu_info);
> > +	amdgpu_amdkfd_get_cu_info(kdev->adev, &cu_info);
> >  	cu->num_simd_per_cu = cu_info.simd_per_cu;
> >  	cu->num_simd_cores = cu_info.simd_per_cu *
> cu_info.cu_active_number;
> >  	cu->max_waves_simd = cu_info.max_waves_per_simd; @@ -2145,7
> +2145,7
> > @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
> >  	 * report the total FB size (public+private) as a single
> >  	 * private heap.
> >  	 */
> > -	amdgpu_amdkfd_get_local_mem_info(kdev->kgd,
> &local_mem_info);
> > +	amdgpu_amdkfd_get_local_mem_info(kdev->adev,
> &local_mem_info);
> >  	sub_type_hdr = (typeof(sub_type_hdr))((char *)sub_type_hdr +
> >  			sub_type_hdr->length);
> >
> > diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> > b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> > index a90ec8de213b..00c726123207 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> > @@ -894,7 +894,7 @@ static int kfd_gws_init(struct kfd_dev *kfd)
> >  		|| (kfd->device_info->asic_family == CHIP_ALDEBARAN
> >  			&& kfd->mec2_fw_version >= 0x28))
> >  		ret = amdgpu_amdkfd_alloc_gws(kfd->adev,
> > -				amdgpu_amdkfd_get_num_gws(kfd->kgd),
> &kfd->gws);
> > +				amdgpu_amdkfd_get_num_gws(kfd->adev),
> &kfd->gws);
> >
> >  	return ret;
> >  }
> > @@ -911,11 +911,11 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
> >  	unsigned int size, map_process_packet_size;
> >
> >  	kfd->ddev = ddev;
> > -	kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd-
> >kgd,
> > +	kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd-
> >adev,
> >  			KGD_ENGINE_MEC1);
> > -	kfd->mec2_fw_version = amdgpu_amdkfd_get_fw_version(kfd-
> >kgd,
> > +	kfd->mec2_fw_version = amdgpu_amdkfd_get_fw_version(kfd-
> >adev,
> >  			KGD_ENGINE_MEC2);
> > -	kfd->sdma_fw_version = amdgpu_amdkfd_get_fw_version(kfd-
> >kgd,
> > +	kfd->sdma_fw_version = amdgpu_amdkfd_get_fw_version(kfd-
> >adev,
> >  			KGD_ENGINE_SDMA1);
> >  	kfd->shared_resources = *gpu_resources;
> >
> > @@ -996,9 +996,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
> >  		goto kfd_doorbell_error;
> >  	}
> >
> > -	kfd->hive_id = amdgpu_amdkfd_get_hive_id(kfd->kgd);
> > +	kfd->hive_id = amdgpu_amdkfd_get_hive_id(kfd->adev);
> >
> > -	kfd->noretry = amdgpu_amdkfd_get_noretry(kfd->kgd);
> > +	kfd->noretry = amdgpu_amdkfd_get_noretry(kfd->adev);
> >
> >  	if (kfd_interrupt_init(kfd)) {
> >  		dev_err(kfd_device, "Error initializing interrupts\n"); @@ -
> 1016,7
> > +1016,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
> >  	 */
> >  	if (kfd_gws_init(kfd)) {
> >  		dev_err(kfd_device, "Could not allocate %d gws\n",
> > -			amdgpu_amdkfd_get_num_gws(kfd->kgd));
> > +			amdgpu_amdkfd_get_num_gws(kfd->adev));
> >  		goto gws_error;
> >  	}
> >
> > diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
> > b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
> > index c021519af810..7b4118915bf6 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
> > @@ -100,7 +100,7 @@ void mqd_symmetrically_map_cu_mask(struct
> mqd_manager *mm,
> >  	struct kfd_cu_info cu_info;
> >  	uint32_t
> cu_per_sh[KFD_MAX_NUM_SE][KFD_MAX_NUM_SH_PER_SE] = {0};
> >  	int i, se, sh, cu;
> > -	amdgpu_amdkfd_get_cu_info(mm->dev->kgd, &cu_info);
> > +	amdgpu_amdkfd_get_cu_info(mm->dev->adev, &cu_info);
> >
> >  	if (cu_mask_count > cu_info.cu_active_number)
> >  		cu_mask_count = cu_info.cu_active_number; diff --git
> > a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> > b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> > index c8bd062fb954..499db2099775 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> > @@ -969,7 +969,7 @@ struct kfd_topology_device
> > *kfd_topology_device_by_proximity_domain(
> >  struct kfd_topology_device *kfd_topology_device_by_id(uint32_t
> > gpu_id);  struct kfd_dev *kfd_device_by_id(uint32_t gpu_id);  struct
> > kfd_dev *kfd_device_by_pci_dev(const struct pci_dev *pdev); -struct
> > kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd);
> > +struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device
> *adev);
> >  int kfd_topology_enum_kfd_devices(uint8_t idx, struct kfd_dev
> > **kdev);  int kfd_numa_node_to_apic_id(int numa_node_id);  void
> > kfd_double_confirm_iommu_support(struct kfd_dev *gpu); diff --git
> > a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
> > b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
> > index 243dd1efcdbf..d8462bd3b4a6 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
> > @@ -118,7 +118,7 @@ int pqm_set_gws(struct process_queue_manager
> *pqm, unsigned int qid,
> >  		return ret;
> >
> >  	pqn->q->gws = mem;
> > -	pdd->qpd.num_gws = gws ? amdgpu_amdkfd_get_num_gws(dev-
> >kgd) : 0;
> > +	pdd->qpd.num_gws = gws ? amdgpu_amdkfd_get_num_gws(dev-
> >adev) : 0;
> >
> >  	return pqn->q->device->dqm->ops.update_queue(pqn->q->device-
> >dqm,
> >  							pqn->q);
> > diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
> > b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
> > index dd593ad0614a..31610b4a3e2e 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
> > @@ -113,7 +113,7 @@ struct kfd_dev *kfd_device_by_pci_dev(const
> struct pci_dev *pdev)
> >  	return device;
> >  }
> >
> > -struct kfd_dev *kfd_device_by_kgd(const struct kgd_dev *kgd)
> > +struct kfd_dev *kfd_device_by_adev(const struct amdgpu_device
> *adev)
> >  {
> >  	struct kfd_topology_device *top_dev;
> >  	struct kfd_dev *device = NULL;
> > @@ -121,7 +121,7 @@ struct kfd_dev *kfd_device_by_kgd(const struct
> kgd_dev *kgd)
> >  	down_read(&topology_lock);
> >
> >  	list_for_each_entry(top_dev, &topology_device_list, list)
> > -		if (top_dev->gpu && top_dev->gpu->kgd == kgd) {
> > +		if (top_dev->gpu && top_dev->gpu->adev == adev) {
> >  			device = top_dev->gpu;
> >  			break;
> >  		}
> > @@ -531,7 +531,7 @@ static ssize_t node_show(struct kobject *kobj,
> struct attribute *attr,
> >  		sysfs_show_32bit_prop(buffer, offs, "sdma_fw_version",
> >  				      dev->gpu->sdma_fw_version);
> >  		sysfs_show_64bit_prop(buffer, offs, "unique_id",
> > -				      amdgpu_amdkfd_get_unique_id(dev-
> >gpu->kgd));
> > +				      amdgpu_amdkfd_get_unique_id(dev-
> >gpu->adev));
> >
> >  	}
> >
> > @@ -1106,7 +1106,7 @@ static uint32_t kfd_generate_gpu_id(struct
> kfd_dev *gpu)
> >  	if (!gpu)
> >  		return 0;
> >
> > -	amdgpu_amdkfd_get_local_mem_info(gpu->kgd,
> &local_mem_info);
> > +	amdgpu_amdkfd_get_local_mem_info(gpu->adev,
> &local_mem_info);
> >
> >  	local_mem_size = local_mem_info.local_mem_size_private +
> >  			local_mem_info.local_mem_size_public;
> > @@ -1189,7 +1189,7 @@ static void kfd_fill_mem_clk_max_info(struct
> kfd_topology_device *dev)
> >  	 * for APUs - If CRAT from ACPI reports more than one bank, then
> >  	 *	all the banks will report the same mem_clk_max information
> >  	 */
> > -	amdgpu_amdkfd_get_local_mem_info(dev->gpu->kgd,
> &local_mem_info);
> > +	amdgpu_amdkfd_get_local_mem_info(dev->gpu->adev,
> &local_mem_info);
> >
> >  	list_for_each_entry(mem, &dev->mem_props, list)
> >  		mem->mem_clk_max = local_mem_info.mem_clk_max;
> @@ -1372,7 +1372,7
> > @@ int kfd_topology_add_device(struct kfd_dev *gpu)
> >  	 * needed for the topology
> >  	 */
> >
> > -	amdgpu_amdkfd_get_cu_info(dev->gpu->kgd, &cu_info);
> > +	amdgpu_amdkfd_get_cu_info(dev->gpu->adev, &cu_info);
> >
> >  	strncpy(dev->node_props.name, gpu->device_info->asic_name,
> >  			KFD_TOPOLOGY_PUBLIC_NAME_SIZE);
> > @@ -1384,13 +1384,13 @@ int kfd_topology_add_device(struct kfd_dev
> *gpu)
> >  	dev->node_props.vendor_id = gpu->pdev->vendor;
> >  	dev->node_props.device_id = gpu->pdev->device;
> >  	dev->node_props.capability |=
> > -		((amdgpu_amdkfd_get_asic_rev_id(dev->gpu->kgd) <<
> > +		((amdgpu_amdkfd_get_asic_rev_id(dev->gpu->adev) <<
> >  			HSA_CAP_ASIC_REVISION_SHIFT) &
> >  			HSA_CAP_ASIC_REVISION_MASK);
> >  	dev->node_props.location_id = pci_dev_id(gpu->pdev);
> >  	dev->node_props.domain = pci_domain_nr(gpu->pdev->bus);
> >  	dev->node_props.max_engine_clk_fcompute =
> > -		amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev-
> >gpu->kgd);
> > +		amdgpu_amdkfd_get_max_engine_clock_in_mhz(dev-
> >gpu->adev);
> >  	dev->node_props.max_engine_clk_ccompute =
> >  		cpufreq_quick_get_max(0) / 1000;
> >  	dev->node_props.drm_render_minor =
> > @@ -1404,7 +1404,7 @@ int kfd_topology_add_device(struct kfd_dev
> *gpu)
> >  				gpu->device_info-
> >num_sdma_queues_per_engine;
> >  	dev->node_props.num_gws = (dev->gpu->gws &&
> >  		dev->gpu->dqm->sched_policy !=
> KFD_SCHED_POLICY_NO_HWS) ?
> > -		amdgpu_amdkfd_get_num_gws(dev->gpu->kgd) : 0;
> > +		amdgpu_amdkfd_get_num_gws(dev->gpu->adev) : 0;
> >  	dev->node_props.num_cp_queues = get_cp_queues_num(dev-
> >gpu->dqm);
> >
> >  	kfd_fill_mem_clk_max_info(dev);

[PATCH 11/13] drm/amdkfd: replace kgd_dev in gpuvm amdgpu_amdkfd funcs

[Graham Sider]

Modified definitions:

- amdgpu_amdkfd_gpuvm_acquire_process_vm
- amdgpu_amdkfd_gpuvm_release_process_vm
- amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu
- amdgpu_amdkfd_gpuvm_free_memory_of_gpu
- amdgpu_amdkfd_gpuvm_map_memory_to_gpu
- amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu
- amdgpu_amdkfd_gpuvm_sync_memory
- amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel
- amdgpu_amdkfd_gpuvm_get_vm_fault_info
- amdgpu_amdkfd_gpuvm_import_dmabuf
- amdgpu_amdkfd_get_tile_config

Remove:

- get_amdgpu_device

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h    | 24 ++++++-----
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c  | 41 ++++++-------------
 .../gpu/drm/amd/amdkfd/cik_event_interrupt.c  |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c      | 22 +++++-----
 drivers/gpu/drm/amd/amdkfd/kfd_process.c      | 22 +++++-----
 5 files changed, 49 insertions(+), 62 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 7e3697a7a5cd..35f703dda034 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -265,37 +265,39 @@ int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_
 	(&((struct amdgpu_fpriv *)					\
 		((struct drm_file *)(drm_priv))->driver_priv)->vm)
 
-int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd,
+int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
 					struct file *filp, u32 pasid,
 					void **process_info,
 					struct dma_fence **ef);
-void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *drm_priv);
+void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev,
+					void *drm_priv);
 uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv);
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
-		struct kgd_dev *kgd, uint64_t va, uint64_t size,
+		struct amdgpu_device *adev, uint64_t va, uint64_t size,
 		void *drm_priv, struct kgd_mem **mem,
 		uint64_t *offset, uint32_t flags);
 int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
-		struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv,
+		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv,
 		uint64_t *size);
 int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
-		struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv, bool *table_freed);
+		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv,
+		bool *table_freed);
 int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
-		struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv);
+		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv);
 int amdgpu_amdkfd_gpuvm_sync_memory(
-		struct kgd_dev *kgd, struct kgd_mem *mem, bool intr);
-int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd,
+		struct amdgpu_device *adev, struct kgd_mem *mem, bool intr);
+int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct amdgpu_device *adev,
 		struct kgd_mem *mem, void **kptr, uint64_t *size);
 int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info,
 					    struct dma_fence **ef);
-int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd,
+int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev,
 					      struct kfd_vm_fault_info *info);
-int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd,
+int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
 				      struct dma_buf *dmabuf,
 				      uint64_t va, void *drm_priv,
 				      struct kgd_mem **mem, uint64_t *size,
 				      uint64_t *mmap_offset);
-int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd,
+int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
 				struct tile_config *config);
 void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev);
 #if IS_ENABLED(CONFIG_HSA_AMD)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index cdf46bd0d8d5..d632484b209e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -60,12 +60,6 @@ static const char * const domain_bit_to_string[] = {
 
 static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work);
 
-
-static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
-{
-	return (struct amdgpu_device *)kgd;
-}
-
 static bool kfd_mem_is_attached(struct amdgpu_vm *avm,
 		struct kgd_mem *mem)
 {
@@ -1269,12 +1263,11 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 	return ret;
 }
 
-int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd,
+int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
 					   struct file *filp, u32 pasid,
 					   void **process_info,
 					   struct dma_fence **ef)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct amdgpu_fpriv *drv_priv;
 	struct amdgpu_vm *avm;
 	int ret;
@@ -1350,12 +1343,12 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
 	}
 }
 
-void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *drm_priv)
+void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev,
+					    void *drm_priv)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct amdgpu_vm *avm;
 
-	if (WARN_ON(!kgd || !drm_priv))
+	if (WARN_ON(!adev || !drm_priv))
 		return;
 
 	avm = drm_priv_to_vm(drm_priv);
@@ -1383,11 +1376,10 @@ uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv)
 }
 
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
-		struct kgd_dev *kgd, uint64_t va, uint64_t size,
+		struct amdgpu_device *adev, uint64_t va, uint64_t size,
 		void *drm_priv, struct kgd_mem **mem,
 		uint64_t *offset, uint32_t flags)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
 	enum ttm_bo_type bo_type = ttm_bo_type_device;
 	struct sg_table *sg = NULL;
@@ -1520,7 +1512,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 }
 
 int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
-		struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv,
+		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv,
 		uint64_t *size)
 {
 	struct amdkfd_process_info *process_info = mem->process_info;
@@ -1608,10 +1600,9 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 }
 
 int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
-		struct kgd_dev *kgd, struct kgd_mem *mem,
+		struct amdgpu_device *adev, struct kgd_mem *mem,
 		void *drm_priv, bool *table_freed)
 {
-	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
 	int ret;
 	struct amdgpu_bo *bo;
@@ -1738,7 +1729,7 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
 }
 
 int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
-		struct kgd_dev *kgd, struct kgd_mem *mem, void *drm_priv)
+		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv)
 {
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
 	struct amdkfd_process_info *process_info = avm->process_info;
@@ -1799,7 +1790,7 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 }
 
 int amdgpu_amdkfd_gpuvm_sync_memory(
-		struct kgd_dev *kgd, struct kgd_mem *mem, bool intr)
+		struct amdgpu_device *adev, struct kgd_mem *mem, bool intr)
 {
 	struct amdgpu_sync sync;
 	int ret;
@@ -1815,7 +1806,7 @@ int amdgpu_amdkfd_gpuvm_sync_memory(
 	return ret;
 }
 
-int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd,
+int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct amdgpu_device *adev,
 		struct kgd_mem *mem, void **kptr, uint64_t *size)
 {
 	int ret;
@@ -1871,12 +1862,9 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd,
 	return ret;
 }
 
-int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd,
+int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev,
 					      struct kfd_vm_fault_info *mem)
 {
-	struct amdgpu_device *adev;
-
-	adev = (struct amdgpu_device *)kgd;
 	if (atomic_read(&adev->gmc.vm_fault_info_updated) == 1) {
 		*mem = *adev->gmc.vm_fault_info;
 		mb();
@@ -1885,13 +1873,12 @@ int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd,
 	return 0;
 }
 
-int amdgpu_amdkfd_gpuvm_import_dmabuf(struct kgd_dev *kgd,
+int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
 				      struct dma_buf *dma_buf,
 				      uint64_t va, void *drm_priv,
 				      struct kgd_mem **mem, uint64_t *size,
 				      uint64_t *mmap_offset)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
 	struct drm_gem_object *obj;
 	struct amdgpu_bo *bo;
@@ -2511,11 +2498,9 @@ int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem)
 }
 
 /* Returns GPU-specific tiling mode information */
-int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd,
+int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
 				struct tile_config *config)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
-
 	config->gb_addr_config = adev->gfx.config.gb_addr_config;
 	config->tile_config_ptr = adev->gfx.config.tile_mode_array;
 	config->num_tile_configs =
diff --git a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c
index 3073361bb714..e31ea107e998 100644
--- a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c
+++ b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c
@@ -113,7 +113,7 @@ static void cik_event_interrupt_wq(struct kfd_dev *dev,
 		kfd_process_vm_fault(dev->dqm, pasid);
 
 		memset(&info, 0, sizeof(info));
-		amdgpu_amdkfd_gpuvm_get_vm_fault_info(dev->kgd, &info);
+		amdgpu_amdkfd_gpuvm_get_vm_fault_info(dev->adev, &info);
 		if (!info.page_addr && !info.status)
 			return;
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 47acfef1aebd..11d917e18b38 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1039,7 +1039,7 @@ static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p,
 		}
 		mutex_unlock(&p->mutex);
 
-		err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kfd->kgd,
+		err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kfd->adev,
 						mem, &kern_addr, &size);
 		if (err) {
 			pr_err("Failed to map event page to kernel\n");
@@ -1151,7 +1151,7 @@ static int kfd_ioctl_get_tile_config(struct file *filep,
 	if (!dev)
 		return -EINVAL;
 
-	amdgpu_amdkfd_get_tile_config(dev->kgd, &config);
+	amdgpu_amdkfd_get_tile_config(dev->adev, &config);
 
 	args->gb_addr_config = config.gb_addr_config;
 	args->num_banks = config.num_banks;
@@ -1314,7 +1314,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 	}
 
 	err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
-		dev->kgd, args->va_addr, args->size,
+		dev->adev, args->va_addr, args->size,
 		pdd->drm_priv, (struct kgd_mem **) &mem, &offset,
 		flags);
 
@@ -1346,7 +1346,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 	return 0;
 
 err_free:
-	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem,
+	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, (struct kgd_mem *)mem,
 					       pdd->drm_priv, NULL);
 err_unlock:
 	mutex_unlock(&p->mutex);
@@ -1383,7 +1383,7 @@ static int kfd_ioctl_free_memory_of_gpu(struct file *filep,
 		goto err_unlock;
 	}
 
-	ret = amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd,
+	ret = amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev,
 				(struct kgd_mem *)mem, pdd->drm_priv, &size);
 
 	/* If freeing the buffer failed, leave the handle in place for
@@ -1468,7 +1468,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
 			goto get_mem_obj_from_handle_failed;
 		}
 		err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
-			peer->kgd, (struct kgd_mem *)mem,
+			peer->adev, (struct kgd_mem *)mem,
 			peer_pdd->drm_priv, &table_freed);
 		if (err) {
 			pr_err("Failed to map to gpu %d/%d\n",
@@ -1480,7 +1480,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
 
 	mutex_unlock(&p->mutex);
 
-	err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd, (struct kgd_mem *) mem, true);
+	err = amdgpu_amdkfd_gpuvm_sync_memory(dev->adev, (struct kgd_mem *) mem, true);
 	if (err) {
 		pr_debug("Sync memory failed, wait interrupted by user signal\n");
 		goto sync_memory_failed;
@@ -1577,7 +1577,7 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep,
 			goto get_mem_obj_from_handle_failed;
 		}
 		err = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
-			peer->kgd, (struct kgd_mem *)mem, peer_pdd->drm_priv);
+			peer->adev, (struct kgd_mem *)mem, peer_pdd->drm_priv);
 		if (err) {
 			pr_err("Failed to unmap from gpu %d/%d\n",
 			       i, args->n_devices);
@@ -1588,7 +1588,7 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep,
 	mutex_unlock(&p->mutex);
 
 	if (dev->device_info->asic_family == CHIP_ALDEBARAN) {
-		err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd,
+		err = amdgpu_amdkfd_gpuvm_sync_memory(dev->adev,
 				(struct kgd_mem *) mem, true);
 		if (err) {
 			pr_debug("Sync memory failed, wait interrupted by user signal\n");
@@ -1742,7 +1742,7 @@ static int kfd_ioctl_import_dmabuf(struct file *filep,
 		goto err_unlock;
 	}
 
-	r = amdgpu_amdkfd_gpuvm_import_dmabuf(dev->kgd, dmabuf,
+	r = amdgpu_amdkfd_gpuvm_import_dmabuf(dev->adev, dmabuf,
 					      args->va_addr, pdd->drm_priv,
 					      (struct kgd_mem **)&mem, &size,
 					      NULL);
@@ -1763,7 +1763,7 @@ static int kfd_ioctl_import_dmabuf(struct file *filep,
 	return 0;
 
 err_free:
-	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem,
+	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, (struct kgd_mem *)mem,
 					       pdd->drm_priv, NULL);
 err_unlock:
 	mutex_unlock(&p->mutex);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 704f07be2b3f..3db9433f7aa5 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -689,8 +689,8 @@ static void kfd_process_free_gpuvm(struct kgd_mem *mem,
 {
 	struct kfd_dev *dev = pdd->dev;
 
-	amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(dev->kgd, mem, pdd->drm_priv);
-	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, mem, pdd->drm_priv,
+	amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(dev->adev, mem, pdd->drm_priv);
+	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, mem, pdd->drm_priv,
 					       NULL);
 }
 
@@ -709,17 +709,17 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd,
 	int handle;
 	int err;
 
-	err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(kdev->kgd, gpu_va, size,
+	err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(kdev->adev, gpu_va, size,
 						 pdd->drm_priv, &mem, NULL, flags);
 	if (err)
 		goto err_alloc_mem;
 
-	err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->kgd, mem,
+	err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->adev, mem,
 			pdd->drm_priv, NULL);
 	if (err)
 		goto err_map_mem;
 
-	err = amdgpu_amdkfd_gpuvm_sync_memory(kdev->kgd, mem, true);
+	err = amdgpu_amdkfd_gpuvm_sync_memory(kdev->adev, mem, true);
 	if (err) {
 		pr_debug("Sync memory failed, wait interrupted by user signal\n");
 		goto sync_memory_failed;
@@ -738,7 +738,7 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd,
 	}
 
 	if (kptr) {
-		err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kdev->kgd,
+		err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kdev->adev,
 				(struct kgd_mem *)mem, kptr, NULL);
 		if (err) {
 			pr_debug("Map GTT BO to kernel failed\n");
@@ -756,7 +756,7 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd,
 	return err;
 
 err_map_mem:
-	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(kdev->kgd, mem, pdd->drm_priv,
+	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(kdev->adev, mem, pdd->drm_priv,
 					       NULL);
 err_alloc_mem:
 	*kptr = NULL;
@@ -938,10 +938,10 @@ static void kfd_process_device_free_bos(struct kfd_process_device *pdd)
 			if (!peer_pdd->drm_priv)
 				continue;
 			amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
-				peer_pdd->dev->kgd, mem, peer_pdd->drm_priv);
+				peer_pdd->dev->adev, mem, peer_pdd->drm_priv);
 		}
 
-		amdgpu_amdkfd_gpuvm_free_memory_of_gpu(pdd->dev->kgd, mem,
+		amdgpu_amdkfd_gpuvm_free_memory_of_gpu(pdd->dev->adev, mem,
 						       pdd->drm_priv, NULL);
 		kfd_process_device_remove_obj_handle(pdd, id);
 	}
@@ -967,7 +967,7 @@ static void kfd_process_destroy_pdds(struct kfd_process *p)
 
 		if (pdd->drm_file) {
 			amdgpu_amdkfd_gpuvm_release_process_vm(
-					pdd->dev->kgd, pdd->drm_priv);
+					pdd->dev->adev, pdd->drm_priv);
 			fput(pdd->drm_file);
 		}
 
@@ -1496,7 +1496,7 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd,
 	dev = pdd->dev;
 
 	ret = amdgpu_amdkfd_gpuvm_acquire_process_vm(
-		dev->kgd, drm_file, p->pasid,
+		dev->adev, drm_file, p->pasid,
 		&p->kgd_process_info, &p->ef);
 	if (ret) {
 		pr_err("Failed to create process VM object\n");
-- 
2.25.1

[PATCH 12/13] drm/amdkfd: replace/remove remaining kgd_dev references

[Graham Sider]

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 .../drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c   |  5 ---
 .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c    |  5 ---
 .../drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c  |  5 ---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c |  5 ---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c |  5 ---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c |  5 ---
 drivers/gpu/drm/amd/amdkfd/kfd_crat.c         |  6 +--
 drivers/gpu/drm/amd/amdkfd/kfd_device.c       |  2 +-
 .../drm/amd/amdkfd/kfd_device_queue_manager.c |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_priv.h         |  2 +-
 drivers/gpu/drm/amd/amdkfd/kfd_process.c      |  5 +--
 drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c   |  6 +--
 drivers/gpu/drm/amd/amdkfd/kfd_svm.c          | 43 +++++++------------
 drivers/gpu/drm/amd/amdkfd/kfd_topology.c     | 12 +++---
 14 files changed, 31 insertions(+), 77 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index c2b8d970195b..abe93b3ff765 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -57,11 +57,6 @@
 		(*dump)[i++][1] = RREG32(addr);		\
 	} while (0)
 
-static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
-{
-	return (struct amdgpu_device *)kgd;
-}
-
 static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd)
 {
 	return (struct v9_sdma_mqd *)mqd;
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 5f274b7c4121..7b7f4b2764c1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -39,11 +39,6 @@ enum hqd_dequeue_request_type {
 	SAVE_WAVES
 };
 
-static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
-{
-	return (struct amdgpu_device *)kgd;
-}
-
 static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index 980430974aca..1f37d3574001 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -38,11 +38,6 @@ enum hqd_dequeue_request_type {
 	SAVE_WAVES
 };
 
-static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
-{
-	return (struct amdgpu_device *)kgd;
-}
-
 static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
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 e31b03495db4..36528dad7684 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
@@ -82,11 +82,6 @@ union TCP_WATCH_CNTL_BITS {
 	float f32All;
 };
 
-static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
-{
-	return (struct amdgpu_device *)kgd;
-}
-
 static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
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 9a30a4f4f098..52832cd69a93 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
@@ -39,11 +39,6 @@ enum hqd_dequeue_request_type {
 	RESET_WAVES
 };
 
-static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
-{
-	return (struct amdgpu_device *)kgd;
-}
-
 static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t 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 d7b31adcfd80..ddfe7aff919d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -46,11 +46,6 @@ enum hqd_dequeue_request_type {
 	SAVE_WAVES
 };
 
-static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd)
-{
-	return (struct amdgpu_device *)kgd;
-}
-
 static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index 7143550becb0..1dc6cb7446e0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
@@ -1963,8 +1963,6 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
 			struct crat_subtype_iolink *sub_type_hdr,
 			uint32_t proximity_domain)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)kdev->kgd;
-
 	*avail_size -= sizeof(struct crat_subtype_iolink);
 	if (*avail_size < 0)
 		return -ENOMEM;
@@ -1981,7 +1979,7 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
 	/* Fill in IOLINK subtype.
 	 * TODO: Fill-in other fields of iolink subtype
 	 */
-	if (adev->gmc.xgmi.connected_to_cpu) {
+	if (kdev->adev->gmc.xgmi.connected_to_cpu) {
 		/*
 		 * with host gpu xgmi link, host can access gpu memory whether
 		 * or not pcie bar type is large, so always create bidirectional
@@ -1990,7 +1988,7 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
 		sub_type_hdr->flags |= CRAT_IOLINK_FLAGS_BI_DIRECTIONAL;
 		sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_XGMI;
 		sub_type_hdr->num_hops_xgmi = 1;
-		if (adev->asic_type == CHIP_ALDEBARAN) {
+		if (kdev->adev->asic_type == CHIP_ALDEBARAN) {
 			sub_type_hdr->minimum_bandwidth_mbs =
 					amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(
 							kdev->adev, NULL, true);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 00c726123207..402891a02a01 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -1031,7 +1031,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 
 	kfd_cwsr_init(kfd);
 
-	svm_migrate_init((struct amdgpu_device *)kfd->kgd);
+	svm_migrate_init(kfd->adev);
 
 	if(kgd2kfd_resume_iommu(kfd))
 		goto device_iommu_error;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 330fbbb15d83..f7d09e0a706f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -1131,7 +1131,7 @@ static int set_sched_resources(struct device_queue_manager *dqm)
 
 		res.queue_mask |= 1ull
 			<< amdgpu_queue_mask_bit_to_set_resource_bit(
-				(struct amdgpu_device *)dqm->dev->kgd, i);
+				dqm->dev->adev, i);
 	}
 	res.gws_mask = ~0ull;
 	res.oac_mask = res.gds_heap_base = res.gds_heap_size = 0;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 499db2099775..1fbc28c34c4c 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -876,7 +876,7 @@ struct kfd_process *kfd_lookup_process_by_pasid(u32 pasid);
 struct kfd_process *kfd_lookup_process_by_mm(const struct mm_struct *mm);
 
 int kfd_process_gpuidx_from_gpuid(struct kfd_process *p, uint32_t gpu_id);
-int kfd_process_gpuid_from_kgd(struct kfd_process *p,
+int kfd_process_gpuid_from_adev(struct kfd_process *p,
 			       struct amdgpu_device *adev, uint32_t *gpuid,
 			       uint32_t *gpuidx);
 static inline int kfd_process_gpuid_from_gpuidx(struct kfd_process *p,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 3db9433f7aa5..6e2f52dba16c 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1724,14 +1724,13 @@ int kfd_process_gpuidx_from_gpuid(struct kfd_process *p, uint32_t gpu_id)
 }
 
 int
-kfd_process_gpuid_from_kgd(struct kfd_process *p, struct amdgpu_device *adev,
+kfd_process_gpuid_from_adev(struct kfd_process *p, struct amdgpu_device *adev,
 			   uint32_t *gpuid, uint32_t *gpuidx)
 {
-	struct kgd_dev *kgd = (struct kgd_dev *)adev;
 	int i;
 
 	for (i = 0; i < p->n_pdds; i++)
-		if (p->pdds[i] && p->pdds[i]->dev->kgd == kgd) {
+		if (p->pdds[i] && p->pdds[i]->dev->adev == adev) {
 			*gpuid = p->pdds[i]->dev->id;
 			*gpuidx = i;
 			return 0;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
index ed4bc5f844ce..deae12dc777d 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
@@ -207,7 +207,6 @@ void kfd_smi_event_update_gpu_reset(struct kfd_dev *dev, bool post_reset)
 void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev,
 					     uint64_t throttle_bitmask)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)dev->kgd;
 	/*
 	 * ThermalThrottle msg = throttle_bitmask(8):
 	 * 			 thermal_interrupt_count(16):
@@ -223,14 +222,13 @@ void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev,
 
 	len = snprintf(fifo_in, sizeof(fifo_in), "%x %llx:%llx\n",
 		       KFD_SMI_EVENT_THERMAL_THROTTLE, throttle_bitmask,
-		       atomic64_read(&adev->smu.throttle_int_counter));
+		       atomic64_read(&dev->adev->smu.throttle_int_counter));
 
 	add_event_to_kfifo(dev, KFD_SMI_EVENT_THERMAL_THROTTLE,	fifo_in, len);
 }
 
 void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)dev->kgd;
 	struct amdgpu_task_info task_info;
 	/* VmFault msg = (hex)uint32_pid(8) + :(1) + task name(16) = 25 */
 	/* 1 byte event + 1 byte space + 25 bytes msg + 1 byte \n +
@@ -243,7 +241,7 @@ void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid)
 		return;
 
 	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, pasid, &task_info);
+	amdgpu_vm_get_task_info(dev->adev, pasid, &task_info);
 	/* Report VM faults from user applications, not retry from kernel */
 	if (!task_info.pid)
 		return;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index 5b67b14fd8be..9a46ba190ea7 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -193,7 +193,6 @@ svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap,
 
 	for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) {
 		struct kfd_process_device *pdd;
-		struct amdgpu_device *adev;
 
 		pr_debug("mapping to gpu idx 0x%x\n", gpuidx);
 		pdd = kfd_process_device_from_gpuidx(p, gpuidx);
@@ -201,9 +200,8 @@ svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap,
 			pr_debug("failed to find device idx %d\n", gpuidx);
 			return -EINVAL;
 		}
-		adev = (struct amdgpu_device *)pdd->dev->kgd;
 
-		r = svm_range_dma_map_dev(adev, prange, offset, npages,
+		r = svm_range_dma_map_dev(pdd->dev->adev, prange, offset, npages,
 					  hmm_pfns, gpuidx);
 		if (r)
 			break;
@@ -581,7 +579,7 @@ svm_range_get_adev_by_id(struct svm_range *prange, uint32_t gpu_id)
 		return NULL;
 	}
 
-	return (struct amdgpu_device *)pdd->dev->kgd;
+	return pdd->dev->adev;
 }
 
 struct kfd_process_device *
@@ -593,7 +591,7 @@ svm_range_get_pdd_by_adev(struct svm_range *prange, struct amdgpu_device *adev)
 
 	p = container_of(prange->svms, struct kfd_process, svms);
 
-	r = kfd_process_gpuid_from_kgd(p, adev, &gpuid, &gpu_idx);
+	r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpu_idx);
 	if (r) {
 		pr_debug("failed to get device id by adev %p\n", adev);
 		return NULL;
@@ -1255,7 +1253,6 @@ svm_range_map_to_gpus(struct svm_range *prange, unsigned long offset,
 {
 	struct kfd_process_device *pdd;
 	struct amdgpu_device *bo_adev;
-	struct amdgpu_device *adev;
 	struct kfd_process *p;
 	struct dma_fence *fence = NULL;
 	uint32_t gpuidx;
@@ -1274,19 +1271,18 @@ svm_range_map_to_gpus(struct svm_range *prange, unsigned long offset,
 			pr_debug("failed to find device idx %d\n", gpuidx);
 			return -EINVAL;
 		}
-		adev = (struct amdgpu_device *)pdd->dev->kgd;
 
 		pdd = kfd_bind_process_to_device(pdd->dev, p);
 		if (IS_ERR(pdd))
 			return -EINVAL;
 
-		if (bo_adev && adev != bo_adev &&
-		    !amdgpu_xgmi_same_hive(adev, bo_adev)) {
+		if (bo_adev && pdd->dev->adev != bo_adev &&
+		    !amdgpu_xgmi_same_hive(pdd->dev->adev, bo_adev)) {
 			pr_debug("cannot map to device idx %d\n", gpuidx);
 			continue;
 		}
 
-		r = svm_range_map_to_gpu(adev, drm_priv_to_vm(pdd->drm_priv),
+		r = svm_range_map_to_gpu(pdd->dev->adev, drm_priv_to_vm(pdd->drm_priv),
 					 prange, offset, npages, readonly,
 					 prange->dma_addr[gpuidx],
 					 bo_adev, wait ? &fence : NULL);
@@ -1320,7 +1316,6 @@ struct svm_validate_context {
 static int svm_range_reserve_bos(struct svm_validate_context *ctx)
 {
 	struct kfd_process_device *pdd;
-	struct amdgpu_device *adev;
 	struct amdgpu_vm *vm;
 	uint32_t gpuidx;
 	int r;
@@ -1332,7 +1327,6 @@ static int svm_range_reserve_bos(struct svm_validate_context *ctx)
 			pr_debug("failed to find device idx %d\n", gpuidx);
 			return -EINVAL;
 		}
-		adev = (struct amdgpu_device *)pdd->dev->kgd;
 		vm = drm_priv_to_vm(pdd->drm_priv);
 
 		ctx->tv[gpuidx].bo = &vm->root.bo->tbo;
@@ -1354,9 +1348,9 @@ static int svm_range_reserve_bos(struct svm_validate_context *ctx)
 			r = -EINVAL;
 			goto unreserve_out;
 		}
-		adev = (struct amdgpu_device *)pdd->dev->kgd;
 
-		r = amdgpu_vm_validate_pt_bos(adev, drm_priv_to_vm(pdd->drm_priv),
+		r = amdgpu_vm_validate_pt_bos(pdd->dev->adev,
+					      drm_priv_to_vm(pdd->drm_priv),
 					      svm_range_bo_validate, NULL);
 		if (r) {
 			pr_debug("failed %d validate pt bos\n", r);
@@ -1379,12 +1373,10 @@ static void svm_range_unreserve_bos(struct svm_validate_context *ctx)
 static void *kfd_svm_page_owner(struct kfd_process *p, int32_t gpuidx)
 {
 	struct kfd_process_device *pdd;
-	struct amdgpu_device *adev;
 
 	pdd = kfd_process_device_from_gpuidx(p, gpuidx);
-	adev = (struct amdgpu_device *)pdd->dev->kgd;
 
-	return SVM_ADEV_PGMAP_OWNER(adev);
+	return SVM_ADEV_PGMAP_OWNER(pdd->dev->adev);
 }
 
 /*
@@ -1962,7 +1954,6 @@ svm_range_handle_list_op(struct svm_range_list *svms, struct svm_range *prange)
 static void svm_range_drain_retry_fault(struct svm_range_list *svms)
 {
 	struct kfd_process_device *pdd;
-	struct amdgpu_device *adev;
 	struct kfd_process *p;
 	uint32_t i;
 
@@ -1974,9 +1965,9 @@ static void svm_range_drain_retry_fault(struct svm_range_list *svms)
 			continue;
 
 		pr_debug("drain retry fault gpu %d svms %p\n", i, svms);
-		adev = (struct amdgpu_device *)pdd->dev->kgd;
 
-		amdgpu_ih_wait_on_checkpoint_process(adev, &adev->irq.ih1);
+		amdgpu_ih_wait_on_checkpoint_process(pdd->dev->adev,
+						     &pdd->dev->adev->irq.ih1);
 		pr_debug("drain retry fault gpu %d svms 0x%p done\n", i, svms);
 	}
 }
@@ -2283,7 +2274,7 @@ svm_range_best_restore_location(struct svm_range *prange,
 
 	p = container_of(prange->svms, struct kfd_process, svms);
 
-	r = kfd_process_gpuid_from_kgd(p, adev, &gpuid, gpuidx);
+	r = kfd_process_gpuid_from_adev(p, adev, &gpuid, gpuidx);
 	if (r < 0) {
 		pr_debug("failed to get gpuid from kgd\n");
 		return -1;
@@ -2443,7 +2434,7 @@ svm_range *svm_range_create_unregistered_range(struct amdgpu_device *adev,
 		pr_debug("Failed to create prange in address [0x%llx]\n", addr);
 		return NULL;
 	}
-	if (kfd_process_gpuid_from_kgd(p, adev, &gpuid, &gpuidx)) {
+	if (kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx)) {
 		pr_debug("failed to get gpuid from kgd\n");
 		svm_range_free(prange);
 		return NULL;
@@ -2507,7 +2498,7 @@ svm_range_count_fault(struct amdgpu_device *adev, struct kfd_process *p,
 		uint32_t gpuid;
 		int r;
 
-		r = kfd_process_gpuid_from_kgd(p, adev, &gpuid, &gpuidx);
+		r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx);
 		if (r < 0)
 			return;
 	}
@@ -2915,7 +2906,6 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 	uint32_t best_loc = prange->prefetch_loc;
 	struct kfd_process_device *pdd;
 	struct amdgpu_device *bo_adev;
-	struct amdgpu_device *adev;
 	struct kfd_process *p;
 	uint32_t gpuidx;
 
@@ -2943,12 +2933,11 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 			pr_debug("failed to get device by idx 0x%x\n", gpuidx);
 			continue;
 		}
-		adev = (struct amdgpu_device *)pdd->dev->kgd;
 
-		if (adev == bo_adev)
+		if (pdd->dev->adev == bo_adev)
 			continue;
 
-		if (!amdgpu_xgmi_same_hive(adev, bo_adev)) {
+		if (!amdgpu_xgmi_same_hive(pdd->dev->adev, bo_adev)) {
 			best_loc = 0;
 			break;
 		}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index 31610b4a3e2e..8364d6614836 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -1286,7 +1286,6 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	void *crat_image = NULL;
 	size_t image_size = 0;
 	int proximity_domain;
-	struct amdgpu_device *adev;
 
 	INIT_LIST_HEAD(&temp_topology_device_list);
 
@@ -1477,16 +1476,17 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 
 	/* kfd only concerns sram ecc on GFX and HBM ecc on UMC */
 	dev->node_props.capability |=
-		((adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__GFX)) != 0) ?
+		((dev->gpu->adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__GFX)) != 0) ?
 		HSA_CAP_SRAM_EDCSUPPORTED : 0;
-	dev->node_props.capability |= ((adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__UMC)) != 0) ?
+	dev->node_props.capability |=
+		((dev->gpu->adev->ras_enabled & BIT(AMDGPU_RAS_BLOCK__UMC)) != 0) ?
 		HSA_CAP_MEM_EDCSUPPORTED : 0;
 
-	if (adev->asic_type != CHIP_VEGA10)
-		dev->node_props.capability |= (adev->ras_enabled != 0) ?
+	if (dev->gpu->adev->asic_type != CHIP_VEGA10)
+		dev->node_props.capability |= (dev->gpu->adev->ras_enabled != 0) ?
 			HSA_CAP_RASEVENTNOTIFY : 0;
 
-	if (KFD_IS_SVM_API_SUPPORTED(adev->kfd.dev))
+	if (KFD_IS_SVM_API_SUPPORTED(dev->gpu->adev->kfd.dev))
 		dev->node_props.capability |= HSA_CAP_SVMAPI_SUPPORTED;
 
 	kfd_debug_print_topology();
-- 
2.25.1

[PATCH 13/13] drm/amdkfd: remove kgd_dev declaration and initialization

[Graham Sider]

Signed-off-by: Graham Sider <Graham.Sider@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c      | 2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h      | 4 ++--
 drivers/gpu/drm/amd/amdkfd/kfd_device.c         | 4 +---
 drivers/gpu/drm/amd/amdkfd/kfd_priv.h           | 1 -
 drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 2 --
 5 files changed, 4 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 79a2e37baa59..83f863dca7af 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -72,7 +72,7 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev)
 	if (!kfd_initialized)
 		return;
 
-	adev->kfd.dev = kgd2kfd_probe((struct kgd_dev *)adev, vf);
+	adev->kfd.dev = kgd2kfd_probe(adev, vf);
 
 	if (adev->kfd.dev)
 		amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 35f703dda034..8c1ba8f258c8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -331,7 +331,7 @@ int kgd2kfd_schedule_evict_and_restore_process(struct mm_struct *mm,
 #if IS_ENABLED(CONFIG_HSA_AMD)
 int kgd2kfd_init(void);
 void kgd2kfd_exit(void);
-struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf);
+struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf);
 bool kgd2kfd_device_init(struct kfd_dev *kfd,
 			 struct drm_device *ddev,
 			 const struct kgd2kfd_shared_resources *gpu_resources);
@@ -355,7 +355,7 @@ static inline void kgd2kfd_exit(void)
 }
 
 static inline
-struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf)
+struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 {
 	return NULL;
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 402891a02a01..7677ced16a27 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -627,12 +627,11 @@ static void kfd_gtt_sa_fini(struct kfd_dev *kfd);
 
 static int kfd_resume(struct kfd_dev *kfd);
 
-struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf)
+struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 {
 	struct kfd_dev *kfd;
 	const struct kfd_device_info *device_info;
 	const struct kfd2kgd_calls *f2g;
-	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
 	struct pci_dev *pdev = adev->pdev;
 
 	switch (adev->asic_type) {
@@ -824,7 +823,6 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf)
 	if (!kfd)
 		return NULL;
 
-	kfd->kgd = kgd;
 	kfd->adev = adev;
 	kfd->device_info = device_info;
 	kfd->pdev = pdev;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 1fbc28c34c4c..32307b9f1ec2 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -228,7 +228,6 @@ struct kfd_vmid_info {
 };
 
 struct kfd_dev {
-	struct kgd_dev *kgd;
 	struct amdgpu_device *adev;
 
 	const struct kfd_device_info *device_info;
diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
index 8f4f3a1700e8..ac941f62cbed 100644
--- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
+++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
@@ -38,8 +38,6 @@ struct amdgpu_device;
 #define KGD_MAX_QUEUES 128
 
 struct kfd_dev;
-struct kgd_dev;
-
 struct kgd_mem;
 
 enum kfd_preempt_type {
-- 
2.25.1

Re: [PATCH 01/13] drm/amdkfd: add amdgpu_device entry to kfd_dev

[Felix Kuehling]

Am 2021-10-19 um 5:13 p.m. schrieb Graham Sider:
> Patch series to remove kgd_dev struct and replace all instances with
> amdgpu_device objects.
>
> amdgpu_device needs to be declared in kgd_kfd_interface.h to be visible
> to kfd2kgd_calls.
>
> Signed-off-by: Graham Sider <Graham.Sider@amd.com>

Nice cleanup. I responded with a suggestion for a possible follow-up to
patch 10. The series is

Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>


> ---
>  drivers/gpu/drm/amd/amdkfd/kfd_device.c         | 1 +
>  drivers/gpu/drm/amd/amdkfd/kfd_priv.h           | 1 +
>  drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 1 +
>  3 files changed, 3 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> index 0fffaf859c59..81ca00d7b3da 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> @@ -825,6 +825,7 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, bool vf)
>  		return NULL;
>  
>  	kfd->kgd = kgd;
> +	kfd->adev = adev;
>  	kfd->device_info = device_info;
>  	kfd->pdev = pdev;
>  	kfd->init_complete = false;
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> index 6d8f9bb2d905..c8bd062fb954 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
> @@ -229,6 +229,7 @@ struct kfd_vmid_info {
>  
>  struct kfd_dev {
>  	struct kgd_dev *kgd;
> +	struct amdgpu_device *adev;
>  
>  	const struct kfd_device_info *device_info;
>  	struct pci_dev *pdev;
> diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
> index c84bd7b2cf59..ba444cbf9206 100644
> --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
> +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
> @@ -33,6 +33,7 @@
>  #include <linux/dma-fence.h>
>  
>  struct pci_dev;
> +struct amdgpu_device;
>  
>  #define KGD_MAX_QUEUES 128
>