All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH 1/3] drm/amdgpu: move check from cpu address to bo
@ 2017-07-04  6:10 Huang Rui
       [not found] ` <1499148650-22409-1-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
  0 siblings, 1 reply; 8+ messages in thread
From: Huang Rui @ 2017-07-04  6:10 UTC (permalink / raw)
  To: Alex Deucher, Christian König,
	amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW
  Cc: Ken Wang, Huang Rui

Signed-off-by: Huang Rui <ray.huang@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 4 ++--
 drivers/gpu/drm/amd/amdgpu/psp_v3_1.c   | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
index 79db294..1469e6d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
@@ -435,10 +435,10 @@ static int psp_hw_fini(void *handle)
 
 	psp_ring_destroy(psp, PSP_RING_TYPE__KM);
 
-	if (psp->tmr_buf)
+	if (psp->tmr_bo)
 		amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf);
 
-	if (psp->fw_pri_buf)
+	if (psp->fw_pri_bo)
 		amdgpu_bo_free_kernel(&psp->fw_pri_bo,
 				      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
index 988ebd9..3d057d8 100644
--- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
@@ -341,7 +341,7 @@ int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type)
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
 			   0x80000000, 0x80000000, false);
 
-	if (ring->ring_mem)
+	if (adev->firmware.rbuf)
 		amdgpu_bo_free_kernel(&adev->firmware.rbuf,
 				      &ring->ring_mem_mc_addr,
 				      (void **)&ring->ring_mem);
-- 
2.7.4

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

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 2/3] drm/amdgpu: fix missed asd bo free when hw_fini
       [not found] ` <1499148650-22409-1-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
@ 2017-07-04  6:10   ` Huang Rui
       [not found]     ` <1499148650-22409-2-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
  2017-07-04  6:10   ` [PATCH 3/3] drm/amdgpu: make psp cmd buffer as a reserve memory Huang Rui
  2017-07-04  7:19   ` [PATCH 1/3] drm/amdgpu: move check from cpu address to bo Christian König
  2 siblings, 1 reply; 8+ messages in thread
From: Huang Rui @ 2017-07-04  6:10 UTC (permalink / raw)
  To: Alex Deucher, Christian König,
	amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW
  Cc: Ken Wang, Huang Rui

Signed-off-by: Huang Rui <ray.huang@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
index 1469e6d..7b43c60 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
@@ -446,6 +446,11 @@ static int psp_hw_fini(void *handle)
 		amdgpu_bo_free_kernel(&psp->fence_buf_bo,
 				      &psp->fence_buf_mc_addr, &psp->fence_buf);
 
+	if (psp->asd_shared_bo)
+		amdgpu_bo_free_kernel(&psp->asd_shared_bo,
+				      &psp->asd_shared_mc_addr,
+				      &psp->asd_shared_buf);
+
 	kfree(psp->cmd);
 	psp->cmd = NULL;
 
-- 
2.7.4

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

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH 3/3] drm/amdgpu: make psp cmd buffer as a reserve memory
       [not found] ` <1499148650-22409-1-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
  2017-07-04  6:10   ` [PATCH 2/3] drm/amdgpu: fix missed asd bo free when hw_fini Huang Rui
@ 2017-07-04  6:10   ` Huang Rui
       [not found]     ` <1499148650-22409-3-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
  2017-07-04  7:19   ` [PATCH 1/3] drm/amdgpu: move check from cpu address to bo Christian König
  2 siblings, 1 reply; 8+ messages in thread
From: Huang Rui @ 2017-07-04  6:10 UTC (permalink / raw)
  To: Alex Deucher, Christian König,
	amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW
  Cc: Ken Wang, Huang Rui

We would like to use a reserve vram to store all non-psp firmware data when it
is submmited. And needn't alloc/free when each firmware submits again and again,
we can reuse that just one page size buffer.

Signed-off-by: Huang Rui <ray.huang@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 39 +++++++++++++++++----------------
 drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h |  5 +++++
 2 files changed, 25 insertions(+), 19 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
index 7b43c60..59700a2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
@@ -118,33 +118,18 @@ psp_cmd_submit_buf(struct psp_context *psp,
 		   int index)
 {
 	int ret;
-	struct amdgpu_bo *cmd_buf_bo;
-	uint64_t cmd_buf_mc_addr;
-	struct psp_gfx_cmd_resp *cmd_buf_mem;
-	struct amdgpu_device *adev = psp->adev;
 
-	ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_VRAM,
-				      &cmd_buf_bo, &cmd_buf_mc_addr,
-				      (void **)&cmd_buf_mem);
-	if (ret)
-		return ret;
-
-	memset(cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE);
+	memset(psp->cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE);
 
-	memcpy(cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp));
+	memcpy(psp->cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp));
 
-	ret = psp_cmd_submit(psp, ucode, cmd_buf_mc_addr,
+	ret = psp_cmd_submit(psp, ucode, psp->cmd_buf_mc_addr,
 			     fence_mc_addr, index);
 
 	while (*((unsigned int *)psp->fence_buf) != index) {
 		msleep(1);
 	}
 
-	amdgpu_bo_free_kernel(&cmd_buf_bo,
-			      &cmd_buf_mc_addr,
-			      (void **)&cmd_buf_mem);
-
 	return ret;
 }
 
@@ -352,6 +337,13 @@ static int psp_load_fw(struct amdgpu_device *adev)
 				      &psp->fence_buf_mc_addr,
 				      &psp->fence_buf);
 	if (ret)
+		goto failed_mem2;
+
+	ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE,
+				      AMDGPU_GEM_DOMAIN_VRAM,
+				      &psp->cmd_buf_bo, &psp->cmd_buf_mc_addr,
+				      (void **)&psp->cmd_buf_mem);
+	if (ret)
 		goto failed_mem1;
 
 	memset(psp->fence_buf, 0, PSP_FENCE_BUFFER_SIZE);
@@ -379,9 +371,13 @@ static int psp_load_fw(struct amdgpu_device *adev)
 	return 0;
 
 failed_mem:
+	amdgpu_bo_free_kernel(&psp->cmd_buf_bo,
+			      &psp->cmd_buf_mc_addr,
+			      (void **)&psp->cmd_buf_mem);
+failed_mem1:
 	amdgpu_bo_free_kernel(&psp->fence_buf_bo,
 			      &psp->fence_buf_mc_addr, &psp->fence_buf);
-failed_mem1:
+failed_mem2:
 	amdgpu_bo_free_kernel(&psp->fw_pri_bo,
 			      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
 failed:
@@ -451,6 +447,11 @@ static int psp_hw_fini(void *handle)
 				      &psp->asd_shared_mc_addr,
 				      &psp->asd_shared_buf);
 
+	if (psp->cmd_buf_bo)
+		amdgpu_bo_free_kernel(&psp->cmd_buf_bo,
+				      &psp->cmd_buf_mc_addr,
+				      (void **)&psp->cmd_buf_mem);
+
 	kfree(psp->cmd);
 	psp->cmd = NULL;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
index 1a1c8b4..538fa9d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
@@ -108,6 +108,11 @@ struct psp_context
 	struct amdgpu_bo 		*fence_buf_bo;
 	uint64_t 			fence_buf_mc_addr;
 	void				*fence_buf;
+
+	/* cmd buffer */
+	struct amdgpu_bo		*cmd_buf_bo;
+	uint64_t			cmd_buf_mc_addr;
+	struct psp_gfx_cmd_resp		*cmd_buf_mem;
 };
 
 struct amdgpu_psp_funcs {
-- 
2.7.4

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

^ permalink raw reply related	[flat|nested] 8+ messages in thread

* Re: [PATCH 1/3] drm/amdgpu: move check from cpu address to bo
       [not found] ` <1499148650-22409-1-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
  2017-07-04  6:10   ` [PATCH 2/3] drm/amdgpu: fix missed asd bo free when hw_fini Huang Rui
  2017-07-04  6:10   ` [PATCH 3/3] drm/amdgpu: make psp cmd buffer as a reserve memory Huang Rui
@ 2017-07-04  7:19   ` Christian König
       [not found]     ` <f4402921-1085-5e4a-c6aa-582534f842eb-5C7GfCeVMHo@public.gmane.org>
  2 siblings, 1 reply; 8+ messages in thread
From: Christian König @ 2017-07-04  7:19 UTC (permalink / raw)
  To: Huang Rui, Alex Deucher, amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW
  Cc: Ken Wang

Actually amdgpu_bo_free_kernel is NULL save, so you should rather remove 
the "if"s here altogether.

Christian.

Am 04.07.2017 um 08:10 schrieb Huang Rui:
> Signed-off-by: Huang Rui <ray.huang@amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 4 ++--
>   drivers/gpu/drm/amd/amdgpu/psp_v3_1.c   | 2 +-
>   2 files changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> index 79db294..1469e6d 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> @@ -435,10 +435,10 @@ static int psp_hw_fini(void *handle)
>   
>   	psp_ring_destroy(psp, PSP_RING_TYPE__KM);
>   
> -	if (psp->tmr_buf)
> +	if (psp->tmr_bo)
>   		amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf);
>   
> -	if (psp->fw_pri_buf)
> +	if (psp->fw_pri_bo)
>   		amdgpu_bo_free_kernel(&psp->fw_pri_bo,
>   				      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
>   
> diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
> index 988ebd9..3d057d8 100644
> --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
> +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
> @@ -341,7 +341,7 @@ int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type)
>   	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
>   			   0x80000000, 0x80000000, false);
>   
> -	if (ring->ring_mem)
> +	if (adev->firmware.rbuf)
>   		amdgpu_bo_free_kernel(&adev->firmware.rbuf,
>   				      &ring->ring_mem_mc_addr,
>   				      (void **)&ring->ring_mem);


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

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 2/3] drm/amdgpu: fix missed asd bo free when hw_fini
       [not found]     ` <1499148650-22409-2-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
@ 2017-07-04  7:20       ` Christian König
  0 siblings, 0 replies; 8+ messages in thread
From: Christian König @ 2017-07-04  7:20 UTC (permalink / raw)
  To: Huang Rui, Alex Deucher, amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW
  Cc: Ken Wang

Am 04.07.2017 um 08:10 schrieb Huang Rui:
> Signed-off-by: Huang Rui <ray.huang@amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 5 +++++
>   1 file changed, 5 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> index 1469e6d..7b43c60 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> @@ -446,6 +446,11 @@ static int psp_hw_fini(void *handle)
>   		amdgpu_bo_free_kernel(&psp->fence_buf_bo,
>   				      &psp->fence_buf_mc_addr, &psp->fence_buf);
>   
> +	if (psp->asd_shared_bo)
> +		amdgpu_bo_free_kernel(&psp->asd_shared_bo,
> +				      &psp->asd_shared_mc_addr,
> +				      &psp->asd_shared_buf);
> +

Good catch, but again please remove the "if".

With that fixed the patch is Reviewed-by: Christian König 
<christian.koenig@amd.com>.

Regards,
Christian.

>   	kfree(psp->cmd);
>   	psp->cmd = NULL;
>   


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

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 1/3] drm/amdgpu: move check from cpu address to bo
       [not found]     ` <f4402921-1085-5e4a-c6aa-582534f842eb-5C7GfCeVMHo@public.gmane.org>
@ 2017-07-04  7:53       ` Huang Rui
  0 siblings, 0 replies; 8+ messages in thread
From: Huang Rui @ 2017-07-04  7:53 UTC (permalink / raw)
  To: Koenig, Christian
  Cc: Deucher, Alexander, Wang, Ken, amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW

On Tue, Jul 04, 2017 at 03:19:17PM +0800, Koenig, Christian wrote:
> Actually amdgpu_bo_free_kernel is NULL save, so you should rather remove
> the "if"s here altogether.
> 

I see, will remove all "if" in front of bo free.

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

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 3/3] drm/amdgpu: make psp cmd buffer as a reserve memory
       [not found]     ` <1499148650-22409-3-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
@ 2017-07-04  8:25       ` Zhang, Jerry (Junwei)
       [not found]         ` <595B50F5.6050509-5C7GfCeVMHo@public.gmane.org>
  0 siblings, 1 reply; 8+ messages in thread
From: Zhang, Jerry (Junwei) @ 2017-07-04  8:25 UTC (permalink / raw)
  To: Huang Rui, Alex Deucher, Christian König,
	amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW
  Cc: Ken Wang

Yeah, when I had a glance at this func, flashed a similar idea.
A little comment inline, please confirm it.

Regards,
Jerry

On 07/04/2017 02:10 PM, Huang Rui wrote:
> We would like to use a reserve vram to store all non-psp firmware data when it
> is submmited. And needn't alloc/free when each firmware submits again and again,
> we can reuse that just one page size buffer.
>
> Signed-off-by: Huang Rui <ray.huang@amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 39 +++++++++++++++++----------------
>   drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h |  5 +++++
>   2 files changed, 25 insertions(+), 19 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> index 7b43c60..59700a2 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> @@ -118,33 +118,18 @@ psp_cmd_submit_buf(struct psp_context *psp,
>   		   int index)
>   {
>   	int ret;
> -	struct amdgpu_bo *cmd_buf_bo;
> -	uint64_t cmd_buf_mc_addr;
> -	struct psp_gfx_cmd_resp *cmd_buf_mem;
> -	struct amdgpu_device *adev = psp->adev;
>
> -	ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE,
> -				      AMDGPU_GEM_DOMAIN_VRAM,
> -				      &cmd_buf_bo, &cmd_buf_mc_addr,
> -				      (void **)&cmd_buf_mem);
> -	if (ret)
> -		return ret;
> -
> -	memset(cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE);
> +	memset(psp->cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE);
>
> -	memcpy(cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp));
> +	memcpy(psp->cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp));
>
> -	ret = psp_cmd_submit(psp, ucode, cmd_buf_mc_addr,
> +	ret = psp_cmd_submit(psp, ucode, psp->cmd_buf_mc_addr,
>   			     fence_mc_addr, index);
>
>   	while (*((unsigned int *)psp->fence_buf) != index) {
>   		msleep(1);
>   	}
>
> -	amdgpu_bo_free_kernel(&cmd_buf_bo,
> -			      &cmd_buf_mc_addr,
> -			      (void **)&cmd_buf_mem);
> -
>   	return ret;
>   }
>
> @@ -352,6 +337,13 @@ static int psp_load_fw(struct amdgpu_device *adev)
>   				      &psp->fence_buf_mc_addr,
>   				      &psp->fence_buf);
>   	if (ret)
> +		goto failed_mem2;
> +
> +	ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE,
> +				      AMDGPU_GEM_DOMAIN_VRAM,
> +				      &psp->cmd_buf_bo, &psp->cmd_buf_mc_addr,
> +				      (void **)&psp->cmd_buf_mem);
> +	if (ret)
>   		goto failed_mem1;
>
>   	memset(psp->fence_buf, 0, PSP_FENCE_BUFFER_SIZE);
> @@ -379,9 +371,13 @@ static int psp_load_fw(struct amdgpu_device *adev)
>   	return 0;

We may also need to update the error handling for psp_ring_init()
as "goto failed_mem;"

>
>   failed_mem:
> +	amdgpu_bo_free_kernel(&psp->cmd_buf_bo,
> +			      &psp->cmd_buf_mc_addr,
> +			      (void **)&psp->cmd_buf_mem);
> +failed_mem1:
>   	amdgpu_bo_free_kernel(&psp->fence_buf_bo,
>   			      &psp->fence_buf_mc_addr, &psp->fence_buf);
> -failed_mem1:
> +failed_mem2:
>   	amdgpu_bo_free_kernel(&psp->fw_pri_bo,
>   			      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
>   failed:
> @@ -451,6 +447,11 @@ static int psp_hw_fini(void *handle)
>   				      &psp->asd_shared_mc_addr,
>   				      &psp->asd_shared_buf);
>
> +	if (psp->cmd_buf_bo)
> +		amdgpu_bo_free_kernel(&psp->cmd_buf_bo,
> +				      &psp->cmd_buf_mc_addr,
> +				      (void **)&psp->cmd_buf_mem);
> +
>   	kfree(psp->cmd);
>   	psp->cmd = NULL;
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
> index 1a1c8b4..538fa9d 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
> @@ -108,6 +108,11 @@ struct psp_context
>   	struct amdgpu_bo 		*fence_buf_bo;
>   	uint64_t 			fence_buf_mc_addr;
>   	void				*fence_buf;
> +
> +	/* cmd buffer */
> +	struct amdgpu_bo		*cmd_buf_bo;
> +	uint64_t			cmd_buf_mc_addr;
> +	struct psp_gfx_cmd_resp		*cmd_buf_mem;
>   };
>
>   struct amdgpu_psp_funcs {
>
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH 3/3] drm/amdgpu: make psp cmd buffer as a reserve memory
       [not found]         ` <595B50F5.6050509-5C7GfCeVMHo@public.gmane.org>
@ 2017-07-04 10:11           ` Huang Rui
  0 siblings, 0 replies; 8+ messages in thread
From: Huang Rui @ 2017-07-04 10:11 UTC (permalink / raw)
  To: Zhang, Jerry (Junwei)
  Cc: Alex Deucher, Ken Wang, Christian König,
	amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW

On Tue, Jul 04, 2017 at 04:25:25PM +0800, Zhang, Jerry (Junwei) wrote:
> Yeah, when I had a glance at this func, flashed a similar idea.
> A little comment inline, please confirm it.
> 
> Regards,
> Jerry
> 
> On 07/04/2017 02:10 PM, Huang Rui wrote:
> >We would like to use a reserve vram to store all non-psp firmware data when it
> >is submmited. And needn't alloc/free when each firmware submits again and again,
> >we can reuse that just one page size buffer.
> >
> >Signed-off-by: Huang Rui <ray.huang@amd.com>
> >---
> >  drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 39 +++++++++++++++++----------------
> >  drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h |  5 +++++
> >  2 files changed, 25 insertions(+), 19 deletions(-)
> >
> >diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> >index 7b43c60..59700a2 100644
> >--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> >+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
> >@@ -118,33 +118,18 @@ psp_cmd_submit_buf(struct psp_context *psp,
> >  		   int index)
> >  {
> >  	int ret;
> >-	struct amdgpu_bo *cmd_buf_bo;
> >-	uint64_t cmd_buf_mc_addr;
> >-	struct psp_gfx_cmd_resp *cmd_buf_mem;
> >-	struct amdgpu_device *adev = psp->adev;
> >
> >-	ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE,
> >-				      AMDGPU_GEM_DOMAIN_VRAM,
> >-				      &cmd_buf_bo, &cmd_buf_mc_addr,
> >-				      (void **)&cmd_buf_mem);
> >-	if (ret)
> >-		return ret;
> >-
> >-	memset(cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE);
> >+	memset(psp->cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE);
> >
> >-	memcpy(cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp));
> >+	memcpy(psp->cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp));
> >
> >-	ret = psp_cmd_submit(psp, ucode, cmd_buf_mc_addr,
> >+	ret = psp_cmd_submit(psp, ucode, psp->cmd_buf_mc_addr,
> >  			     fence_mc_addr, index);
> >
> >  	while (*((unsigned int *)psp->fence_buf) != index) {
> >  		msleep(1);
> >  	}
> >
> >-	amdgpu_bo_free_kernel(&cmd_buf_bo,
> >-			      &cmd_buf_mc_addr,
> >-			      (void **)&cmd_buf_mem);
> >-
> >  	return ret;
> >  }
> >
> >@@ -352,6 +337,13 @@ static int psp_load_fw(struct amdgpu_device *adev)
> >  				      &psp->fence_buf_mc_addr,
> >  				      &psp->fence_buf);
> >  	if (ret)
> >+		goto failed_mem2;
> >+
> >+	ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE,
> >+				      AMDGPU_GEM_DOMAIN_VRAM,
> >+				      &psp->cmd_buf_bo, &psp->cmd_buf_mc_addr,
> >+				      (void **)&psp->cmd_buf_mem);
> >+	if (ret)
> >  		goto failed_mem1;
> >
> >  	memset(psp->fence_buf, 0, PSP_FENCE_BUFFER_SIZE);
> >@@ -379,9 +371,13 @@ static int psp_load_fw(struct amdgpu_device *adev)
> >  	return 0;
> 
> We may also need to update the error handling for psp_ring_init()
> as "goto failed_mem;"
> 

Right, nice catch. Will update it.

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

^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2017-07-04 10:11 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-07-04  6:10 [PATCH 1/3] drm/amdgpu: move check from cpu address to bo Huang Rui
     [not found] ` <1499148650-22409-1-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
2017-07-04  6:10   ` [PATCH 2/3] drm/amdgpu: fix missed asd bo free when hw_fini Huang Rui
     [not found]     ` <1499148650-22409-2-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
2017-07-04  7:20       ` Christian König
2017-07-04  6:10   ` [PATCH 3/3] drm/amdgpu: make psp cmd buffer as a reserve memory Huang Rui
     [not found]     ` <1499148650-22409-3-git-send-email-ray.huang-5C7GfCeVMHo@public.gmane.org>
2017-07-04  8:25       ` Zhang, Jerry (Junwei)
     [not found]         ` <595B50F5.6050509-5C7GfCeVMHo@public.gmane.org>
2017-07-04 10:11           ` Huang Rui
2017-07-04  7:19   ` [PATCH 1/3] drm/amdgpu: move check from cpu address to bo Christian König
     [not found]     ` <f4402921-1085-5e4a-c6aa-582534f842eb-5C7GfCeVMHo@public.gmane.org>
2017-07-04  7:53       ` Huang Rui

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.