[PATCH 1/2] drm/rockchip: create the fake mmap offset during gem creation

[PATCH 1/2] drm/rockchip: create the fake mmap offset during gem creation

[Heiko Stuebner]

Similarly to the equivalent exynos-fix in commit 48cf53f4343a
("drm/exynos: create a fake mmap offset with gem creation") the rockchip
drm driver also calls the drm_gem_create_mmap_offset fake-offset creation
on every invocation of rockchip_gem_dumb_map_offset.

Move that to the creation of the gem to only create that offset once.

Signed-off-by: Heiko Stuebner <heiko@sntech.de>
---
 drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
index 4ab2f20..e50d113 100644
--- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
+++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
@@ -126,8 +126,14 @@ struct rockchip_gem_object *
 	if (ret)
 		goto err_free_rk_obj;
 
+	ret = drm_gem_create_mmap_offset(obj);
+	if (ret)
+		goto err_free_buf;
+
 	return rk_obj;
 
+err_free_buf:
+	rockchip_gem_free_buf(rk_obj);
 err_free_rk_obj:
 	kfree(rk_obj);
 	return ERR_PTR(ret);
@@ -196,7 +202,6 @@ int rockchip_gem_dumb_map_offset(struct drm_file *file_priv,
 				 uint64_t *offset)
 {
 	struct drm_gem_object *obj;
-	int ret;
 
 	obj = drm_gem_object_lookup(dev, file_priv, handle);
 	if (!obj) {
@@ -204,14 +209,9 @@ int rockchip_gem_dumb_map_offset(struct drm_file *file_priv,
 		return -EINVAL;
 	}
 
-	ret = drm_gem_create_mmap_offset(obj);
-	if (ret)
-		goto out;
-
 	*offset = drm_vma_node_offset_addr(&obj->vma_node);
 	DRM_DEBUG_KMS("offset = 0x%llx\n", *offset);
 
-out:
 	drm_gem_object_unreference_unlocked(obj);
 
 	return 0;
-- 
2.6.2

[PATCH 2/2] drm/rockchip: unset pgoff when mmap'ing gems

[Heiko Stuebner]

Commit 371f0f085f629 ("ARM: 8426/1: dma-mapping: add missing range check
 in dma_mmap()") introduced offset-checking for mappings, which collides
with the fake-offset the drm sets for gems.

Other drm-drivers set this offset to 0 before doing the mapping, so
this looks like the correct way to go for rockchip as well.

Fixes: 371f0f085f629 ("ARM: 8426/1: dma-mapping: add missing range check in dma_mmap()")
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
---
This patch is definitly necessary to keep the rockchip-drm working
with kernel 4.4.

 drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
index e50d113..e9bc188 100644
--- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
+++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
@@ -68,6 +68,7 @@ static int rockchip_drm_gem_object_mmap(struct drm_gem_object *obj,
 	 * VM_PFNMAP flag that was set by drm_gem_mmap_obj()/drm_gem_mmap().
 	 */
 	vma->vm_flags &= ~VM_PFNMAP;
+	vma->vm_pgoff = 0;
 
 	ret = dma_mmap_attrs(drm->dev, vma, rk_obj->kvaddr, rk_obj->dma_addr,
 			     obj->size, &rk_obj->dma_attrs);
-- 
2.6.2

[PATCH 2/2] drm/rockchip: unset pgoff when mmap'ing gems

[Mark yao]

On 2015?11?10? 03:52, Heiko Stuebner wrote:
> Commit 371f0f085f629 ("ARM: 8426/1: dma-mapping: add missing range check
>   in dma_mmap()") introduced offset-checking for mappings, which collides
> with the fake-offset the drm sets for gems.
>
> Other drm-drivers set this offset to 0 before doing the mapping, so
> this looks like the correct way to go for rockchip as well.
>
> Fixes: 371f0f085f629 ("ARM: 8426/1: dma-mapping: add missing range check in dma_mmap()")
> Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> ---
> This patch is definitly necessary to keep the rockchip-drm working
> with kernel 4.4.
>
>   drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 1 +
>   1 file changed, 1 insertion(+)
>
> diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> index e50d113..e9bc188 100644
> --- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> +++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> @@ -68,6 +68,7 @@ static int rockchip_drm_gem_object_mmap(struct drm_gem_object *obj,
>   	 * VM_PFNMAP flag that was set by drm_gem_mmap_obj()/drm_gem_mmap().
>   	 */
>   	vma->vm_flags &= ~VM_PFNMAP;
> +	vma->vm_pgoff = 0;
>   
>   	ret = dma_mmap_attrs(drm->dev, vma, rk_obj->kvaddr, rk_obj->dma_addr,
>   			     obj->size, &rk_obj->dma_attrs);
Hi Heiko
      Sorry for miss this patch, thanks for the fix.
      Acked-by: Mark Yao <mark.yao@rock-chips.com>

-- 
?ark Yao

[PATCH 1/2] drm/rockchip: create the fake mmap offset during gem creation

[Mark yao]

Hi Heiko
      I don't think this patch is needed for rockchip drm, since 
rockchip drm only use kms.
I saw the discussion about ("drm/exynos: create a fake mmap offset with 
gem creation"),
Inki said that: "This patch makes drm_gem_create_mmap_offset to be 
called even in case of
not using dumb* interfaces. I.e., exynos_drm_gem_create_ioctl -> 
exynos_drm_gem_mmap"

So I think this patch doesn't make sense for rockchip drm.


On 2015?11?10? 03:50, Heiko Stuebner wrote:
> Similarly to the equivalent exynos-fix in commit 48cf53f4343a
> ("drm/exynos: create a fake mmap offset with gem creation") the rockchip
> drm driver also calls the drm_gem_create_mmap_offset fake-offset creation
> on every invocation of rockchip_gem_dumb_map_offset.
>
> Move that to the creation of the gem to only create that offset once.
>
> Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> ---
>   drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 12 ++++++------
>   1 file changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> index 4ab2f20..e50d113 100644
> --- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> +++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> @@ -126,8 +126,14 @@ struct rockchip_gem_object *
>   	if (ret)
>   		goto err_free_rk_obj;
>   
> +	ret = drm_gem_create_mmap_offset(obj);
> +	if (ret)
> +		goto err_free_buf;
> +
>   	return rk_obj;
>   
> +err_free_buf:
> +	rockchip_gem_free_buf(rk_obj);
>   err_free_rk_obj:
>   	kfree(rk_obj);
>   	return ERR_PTR(ret);
> @@ -196,7 +202,6 @@ int rockchip_gem_dumb_map_offset(struct drm_file *file_priv,
>   				 uint64_t *offset)
>   {
>   	struct drm_gem_object *obj;
> -	int ret;
>   
>   	obj = drm_gem_object_lookup(dev, file_priv, handle);
>   	if (!obj) {
> @@ -204,14 +209,9 @@ int rockchip_gem_dumb_map_offset(struct drm_file *file_priv,
>   		return -EINVAL;
>   	}
>   
> -	ret = drm_gem_create_mmap_offset(obj);
> -	if (ret)
> -		goto out;
> -
>   	*offset = drm_vma_node_offset_addr(&obj->vma_node);
>   	DRM_DEBUG_KMS("offset = 0x%llx\n", *offset);
>   
> -out:
>   	drm_gem_object_unreference_unlocked(obj);
>   
>   	return 0;


-- 
?ark Yao

[PATCH 1/2] drm/rockchip: create the fake mmap offset during gem creation

[Heiko Stuebner]

Hi Mark,

Am Dienstag, 10. November 2015, 14:17:43 schrieb Mark yao:
> Hi Heiko
>       I don't think this patch is needed for rockchip drm, since 
> rockchip drm only use kms.
> I saw the discussion about ("drm/exynos: create a fake mmap offset with 
> gem creation"),
> Inki said that: "This patch makes drm_gem_create_mmap_offset to be 
> called even in case of
> not using dumb* interfaces. I.e., exynos_drm_gem_create_ioctl -> 
> exynos_drm_gem_mmap"
> 
> So I think this patch doesn't make sense for rockchip drm.

ok :-) . I'm not that knowledgable when it comes to drm and just found that 
similar change in exynos while searching for the dma issue.

So we can drop this patch and you could just merge patch2 as well as

	[RESEND PATCH] drm/rockchip: vop: Correct enabled clocks during setup

from Sjoerd and send these two to Dave as fixes for 4.4?


Thanks
Heiko

> 
> 
> On 2015?11?10? 03:50, Heiko Stuebner wrote:
> > Similarly to the equivalent exynos-fix in commit 48cf53f4343a
> > ("drm/exynos: create a fake mmap offset with gem creation") the rockchip
> > drm driver also calls the drm_gem_create_mmap_offset fake-offset creation
> > on every invocation of rockchip_gem_dumb_map_offset.
> >
> > Move that to the creation of the gem to only create that offset once.
> >
> > Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> > ---
> >   drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 12 ++++++------
> >   1 file changed, 6 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c 
b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> > index 4ab2f20..e50d113 100644
> > --- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> > +++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> > @@ -126,8 +126,14 @@ struct rockchip_gem_object *
> >   	if (ret)
> >   		goto err_free_rk_obj;
> >   
> > +	ret = drm_gem_create_mmap_offset(obj);
> > +	if (ret)
> > +		goto err_free_buf;
> > +
> >   	return rk_obj;
> >   
> > +err_free_buf:
> > +	rockchip_gem_free_buf(rk_obj);
> >   err_free_rk_obj:
> >   	kfree(rk_obj);
> >   	return ERR_PTR(ret);
> > @@ -196,7 +202,6 @@ int rockchip_gem_dumb_map_offset(struct drm_file 
*file_priv,
> >   				 uint64_t *offset)
> >   {
> >   	struct drm_gem_object *obj;
> > -	int ret;
> >   
> >   	obj = drm_gem_object_lookup(dev, file_priv, handle);
> >   	if (!obj) {
> > @@ -204,14 +209,9 @@ int rockchip_gem_dumb_map_offset(struct drm_file 
*file_priv,
> >   		return -EINVAL;
> >   	}
> >   
> > -	ret = drm_gem_create_mmap_offset(obj);
> > -	if (ret)
> > -		goto out;
> > -
> >   	*offset = drm_vma_node_offset_addr(&obj->vma_node);
> >   	DRM_DEBUG_KMS("offset = 0x%llx\n", *offset);
> >   
> > -out:
> >   	drm_gem_object_unreference_unlocked(obj);
> >   
> >   	return 0;
> 
> 
> 

[PATCH 1/2] drm/rockchip: create the fake mmap offset during gem creation

[Mark yao]

On 2015?11?10? 14:46, Heiko Stuebner wrote:
> Hi Mark,
>
> Am Dienstag, 10. November 2015, 14:17:43 schrieb Mark yao:
>> Hi Heiko
>>        I don't think this patch is needed for rockchip drm, since
>> rockchip drm only use kms.
>> I saw the discussion about ("drm/exynos: create a fake mmap offset with
>> gem creation"),
>> Inki said that: "This patch makes drm_gem_create_mmap_offset to be
>> called even in case of
>> not using dumb* interfaces. I.e., exynos_drm_gem_create_ioctl ->
>> exynos_drm_gem_mmap"
>>
>> So I think this patch doesn't make sense for rockchip drm.
> ok :-) . I'm not that knowledgable when it comes to drm and just found that
> similar change in exynos while searching for the dma issue.
>
> So we can drop this patch and you could just merge patch2 as well as
>
> 	[RESEND PATCH] drm/rockchip: vop: Correct enabled clocks during setup
>
> from Sjoerd and send these two to Dave as fixes for 4.4?
>
Right, that patch is needed, and also I want to land two patch:
       [RESEND PATCH] drm/rockchip: import dma_buf to gem
       [PATCH] Revert "drm/rockchip: Convert the probe function to the 
generic drm_of_component_probe()"

> Thanks
> Heiko
>
>>
>> On 2015?11?10? 03:50, Heiko Stuebner wrote:
>>> Similarly to the equivalent exynos-fix in commit 48cf53f4343a
>>> ("drm/exynos: create a fake mmap offset with gem creation") the rockchip
>>> drm driver also calls the drm_gem_create_mmap_offset fake-offset creation
>>> on every invocation of rockchip_gem_dumb_map_offset.
>>>
>>> Move that to the creation of the gem to only create that offset once.
>>>
>>> Signed-off-by: Heiko Stuebner <heiko@sntech.de>
>>> ---
>>>    drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 12 ++++++------
>>>    1 file changed, 6 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
>>> index 4ab2f20..e50d113 100644
>>> --- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
>>> +++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
>>> @@ -126,8 +126,14 @@ struct rockchip_gem_object *
>>>    	if (ret)
>>>    		goto err_free_rk_obj;
>>>    
>>> +	ret = drm_gem_create_mmap_offset(obj);
>>> +	if (ret)
>>> +		goto err_free_buf;
>>> +
>>>    	return rk_obj;
>>>    
>>> +err_free_buf:
>>> +	rockchip_gem_free_buf(rk_obj);
>>>    err_free_rk_obj:
>>>    	kfree(rk_obj);
>>>    	return ERR_PTR(ret);
>>> @@ -196,7 +202,6 @@ int rockchip_gem_dumb_map_offset(struct drm_file
> *file_priv,
>>>    				 uint64_t *offset)
>>>    {
>>>    	struct drm_gem_object *obj;
>>> -	int ret;
>>>    
>>>    	obj = drm_gem_object_lookup(dev, file_priv, handle);
>>>    	if (!obj) {
>>> @@ -204,14 +209,9 @@ int rockchip_gem_dumb_map_offset(struct drm_file
> *file_priv,
>>>    		return -EINVAL;
>>>    	}
>>>    
>>> -	ret = drm_gem_create_mmap_offset(obj);
>>> -	if (ret)
>>> -		goto out;
>>> -
>>>    	*offset = drm_vma_node_offset_addr(&obj->vma_node);
>>>    	DRM_DEBUG_KMS("offset = 0x%llx\n", *offset);
>>>    
>>> -out:
>>>    	drm_gem_object_unreference_unlocked(obj);
>>>    
>>>    	return 0;
>>
>>
>
>
>


-- 
?ark Yao

[PATCH 1/2] drm/rockchip: create the fake mmap offset during gem creation

[Heiko Stuebner]

Am Dienstag, 10. November 2015, 17:26:10 schrieb Mark yao:
> On 2015?11?10? 14:46, Heiko Stuebner wrote:
> > Hi Mark,
> >
> > Am Dienstag, 10. November 2015, 14:17:43 schrieb Mark yao:
> >> Hi Heiko
> >>        I don't think this patch is needed for rockchip drm, since
> >> rockchip drm only use kms.
> >> I saw the discussion about ("drm/exynos: create a fake mmap offset with
> >> gem creation"),
> >> Inki said that: "This patch makes drm_gem_create_mmap_offset to be
> >> called even in case of
> >> not using dumb* interfaces. I.e., exynos_drm_gem_create_ioctl ->
> >> exynos_drm_gem_mmap"
> >>
> >> So I think this patch doesn't make sense for rockchip drm.
> > ok :-) . I'm not that knowledgable when it comes to drm and just found that
> > similar change in exynos while searching for the dma issue.
> >
> > So we can drop this patch and you could just merge patch2 as well as
> >
> > 	[RESEND PATCH] drm/rockchip: vop: Correct enabled clocks during setup
> >
> > from Sjoerd and send these two to Dave as fixes for 4.4?
> >
> Right, that patch is needed, and also I want to land two patch:
>        [RESEND PATCH] drm/rockchip: import dma_buf to gem

the import should go into 4.5 though (as it is a new feature), while
the other three are fixes that should go into the current 4.4

>        [PATCH] Revert "drm/rockchip: Convert the probe function to the 
> generic drm_of_component_probe()"
> 
> > Thanks
> > Heiko
> >
> >>
> >> On 2015?11?10? 03:50, Heiko Stuebner wrote:
> >>> Similarly to the equivalent exynos-fix in commit 48cf53f4343a
> >>> ("drm/exynos: create a fake mmap offset with gem creation") the rockchip
> >>> drm driver also calls the drm_gem_create_mmap_offset fake-offset creation
> >>> on every invocation of rockchip_gem_dumb_map_offset.
> >>>
> >>> Move that to the creation of the gem to only create that offset once.
> >>>
> >>> Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> >>> ---
> >>>    drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 12 ++++++------
> >>>    1 file changed, 6 insertions(+), 6 deletions(-)
> >>>
> >>> diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> > b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> >>> index 4ab2f20..e50d113 100644
> >>> --- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> >>> +++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> >>> @@ -126,8 +126,14 @@ struct rockchip_gem_object *
> >>>    	if (ret)
> >>>    		goto err_free_rk_obj;
> >>>    
> >>> +	ret = drm_gem_create_mmap_offset(obj);
> >>> +	if (ret)
> >>> +		goto err_free_buf;
> >>> +
> >>>    	return rk_obj;
> >>>    
> >>> +err_free_buf:
> >>> +	rockchip_gem_free_buf(rk_obj);
> >>>    err_free_rk_obj:
> >>>    	kfree(rk_obj);
> >>>    	return ERR_PTR(ret);
> >>> @@ -196,7 +202,6 @@ int rockchip_gem_dumb_map_offset(struct drm_file
> > *file_priv,
> >>>    				 uint64_t *offset)
> >>>    {
> >>>    	struct drm_gem_object *obj;
> >>> -	int ret;
> >>>    
> >>>    	obj = drm_gem_object_lookup(dev, file_priv, handle);
> >>>    	if (!obj) {
> >>> @@ -204,14 +209,9 @@ int rockchip_gem_dumb_map_offset(struct drm_file
> > *file_priv,
> >>>    		return -EINVAL;
> >>>    	}
> >>>    
> >>> -	ret = drm_gem_create_mmap_offset(obj);
> >>> -	if (ret)
> >>> -		goto out;
> >>> -
> >>>    	*offset = drm_vma_node_offset_addr(&obj->vma_node);
> >>>    	DRM_DEBUG_KMS("offset = 0x%llx\n", *offset);
> >>>    
> >>> -out:
> >>>    	drm_gem_object_unreference_unlocked(obj);
> >>>    
> >>>    	return 0;
> >>
> >>
> >
> >
> >
> 
> 
> 

[PATCH 1/2] drm/rockchip: create the fake mmap offset during gem creation

[Mark yao]

On 2015?11?10? 17:28, Heiko Stuebner wrote:
> Am Dienstag, 10. November 2015, 17:26:10 schrieb Mark yao:
>> On 2015?11?10? 14:46, Heiko Stuebner wrote:
>>> Hi Mark,
>>>
>>> Am Dienstag, 10. November 2015, 14:17:43 schrieb Mark yao:
>>>> Hi Heiko
>>>>         I don't think this patch is needed for rockchip drm, since
>>>> rockchip drm only use kms.
>>>> I saw the discussion about ("drm/exynos: create a fake mmap offset with
>>>> gem creation"),
>>>> Inki said that: "This patch makes drm_gem_create_mmap_offset to be
>>>> called even in case of
>>>> not using dumb* interfaces. I.e., exynos_drm_gem_create_ioctl ->
>>>> exynos_drm_gem_mmap"
>>>>
>>>> So I think this patch doesn't make sense for rockchip drm.
>>> ok :-) . I'm not that knowledgable when it comes to drm and just found that
>>> similar change in exynos while searching for the dma issue.
>>>
>>> So we can drop this patch and you could just merge patch2 as well as
>>>
>>> 	[RESEND PATCH] drm/rockchip: vop: Correct enabled clocks during setup
>>>
>>> from Sjoerd and send these two to Dave as fixes for 4.4?
>>>
>> Right, that patch is needed, and also I want to land two patch:
>>         [RESEND PATCH] drm/rockchip: import dma_buf to gem
> the import should go into 4.5 though (as it is a new feature), while
> the other three are fixes that should go into the current 4.4
>
OK, Got it
>>         [PATCH] Revert "drm/rockchip: Convert the probe function to the
>> generic drm_of_component_probe()"
>>
>>> Thanks
>>> Heiko
>>>
>>>> On 2015?11?10? 03:50, Heiko Stuebner wrote:
>>>>> Similarly to the equivalent exynos-fix in commit 48cf53f4343a
>>>>> ("drm/exynos: create a fake mmap offset with gem creation") the rockchip
>>>>> drm driver also calls the drm_gem_create_mmap_offset fake-offset creation
>>>>> on every invocation of rockchip_gem_dumb_map_offset.
>>>>>
>>>>> Move that to the creation of the gem to only create that offset once.
>>>>>
>>>>> Signed-off-by: Heiko Stuebner <heiko@sntech.de>
>>>>> ---
>>>>>     drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 12 ++++++------
>>>>>     1 file changed, 6 insertions(+), 6 deletions(-)
>>>>>
>>>>> diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
>>> b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
>>>>> index 4ab2f20..e50d113 100644
>>>>> --- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
>>>>> +++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
>>>>> @@ -126,8 +126,14 @@ struct rockchip_gem_object *
>>>>>     	if (ret)
>>>>>     		goto err_free_rk_obj;
>>>>>     
>>>>> +	ret = drm_gem_create_mmap_offset(obj);
>>>>> +	if (ret)
>>>>> +		goto err_free_buf;
>>>>> +
>>>>>     	return rk_obj;
>>>>>     
>>>>> +err_free_buf:
>>>>> +	rockchip_gem_free_buf(rk_obj);
>>>>>     err_free_rk_obj:
>>>>>     	kfree(rk_obj);
>>>>>     	return ERR_PTR(ret);
>>>>> @@ -196,7 +202,6 @@ int rockchip_gem_dumb_map_offset(struct drm_file
>>> *file_priv,
>>>>>     				 uint64_t *offset)
>>>>>     {
>>>>>     	struct drm_gem_object *obj;
>>>>> -	int ret;
>>>>>     
>>>>>     	obj = drm_gem_object_lookup(dev, file_priv, handle);
>>>>>     	if (!obj) {
>>>>> @@ -204,14 +209,9 @@ int rockchip_gem_dumb_map_offset(struct drm_file
>>> *file_priv,
>>>>>     		return -EINVAL;
>>>>>     	}
>>>>>     
>>>>> -	ret = drm_gem_create_mmap_offset(obj);
>>>>> -	if (ret)
>>>>> -		goto out;
>>>>> -
>>>>>     	*offset = drm_vma_node_offset_addr(&obj->vma_node);
>>>>>     	DRM_DEBUG_KMS("offset = 0x%llx\n", *offset);
>>>>>     
>>>>> -out:
>>>>>     	drm_gem_object_unreference_unlocked(obj);
>>>>>     
>>>>>     	return 0;
>>>>
>>>
>>>
>>
>>
>
>
>


-- 
?ark Yao

[PATCH 1/2] drm/rockchip: create the fake mmap offset during gem creation

[Daniel Vetter]

On Tue, Nov 10, 2015 at 02:17:43PM +0800, Mark yao wrote:
> Hi Heiko
>      I don't think this patch is needed for rockchip drm, since rockchip drm
> only use kms.
> I saw the discussion about ("drm/exynos: create a fake mmap offset with gem
> creation"),
> Inki said that: "This patch makes drm_gem_create_mmap_offset to be called
> even in case of
> not using dumb* interfaces. I.e., exynos_drm_gem_create_ioctl ->
> exynos_drm_gem_mmap"
> 
> So I think this patch doesn't make sense for rockchip drm.
> 
> 
> On 2015?11?10? 03:50, Heiko Stuebner wrote:
> >Similarly to the equivalent exynos-fix in commit 48cf53f4343a

This exynos patch and your rockchip patch look really strange - the point
_is_ to only create fake mmap offset when someone actually tries to mmap a
bo. drm_gem_create_mmap_offset internally (and in a race-free manner)
checks whether it's been called already, so it's perfectly safe to call
mutliple times.

Imo drop this (and probably revert 48cf53f4343a as unjustified).
-Daniel

> >("drm/exynos: create a fake mmap offset with gem creation") the rockchip
> >drm driver also calls the drm_gem_create_mmap_offset fake-offset creation
> >on every invocation of rockchip_gem_dumb_map_offset.
> >
> >Move that to the creation of the gem to only create that offset once.
> >
> >Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> >---
> >  drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 12 ++++++------
> >  1 file changed, 6 insertions(+), 6 deletions(-)
> >
> >diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> >index 4ab2f20..e50d113 100644
> >--- a/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> >+++ b/drivers/gpu/drm/rockchip/rockchip_drm_gem.c
> >@@ -126,8 +126,14 @@ struct rockchip_gem_object *
> >  	if (ret)
> >  		goto err_free_rk_obj;
> >+	ret = drm_gem_create_mmap_offset(obj);
> >+	if (ret)
> >+		goto err_free_buf;
> >+
> >  	return rk_obj;
> >+err_free_buf:
> >+	rockchip_gem_free_buf(rk_obj);
> >  err_free_rk_obj:
> >  	kfree(rk_obj);
> >  	return ERR_PTR(ret);
> >@@ -196,7 +202,6 @@ int rockchip_gem_dumb_map_offset(struct drm_file *file_priv,
> >  				 uint64_t *offset)
> >  {
> >  	struct drm_gem_object *obj;
> >-	int ret;
> >  	obj = drm_gem_object_lookup(dev, file_priv, handle);
> >  	if (!obj) {
> >@@ -204,14 +209,9 @@ int rockchip_gem_dumb_map_offset(struct drm_file *file_priv,
> >  		return -EINVAL;
> >  	}
> >-	ret = drm_gem_create_mmap_offset(obj);
> >-	if (ret)
> >-		goto out;
> >-
> >  	*offset = drm_vma_node_offset_addr(&obj->vma_node);
> >  	DRM_DEBUG_KMS("offset = 0x%llx\n", *offset);
> >-out:
> >  	drm_gem_object_unreference_unlocked(obj);
> >  	return 0;
> 
> 
> -- 
> ?ark Yao
> 
> 
> _______________________________________________
> dri-devel mailing list
> dri-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/dri-devel

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch