Hi,
This is the first time I send mail to linux-pwm, I didn't read through
the mails in this list, so if somebody already asked this question, I'm
sorry about that.
I wanna set some fops in "struct platform_pwm_backlight_data". But the
currrent probe function in pwm_bl.c says:
-------
if (!data) {
ret = pwm_backlight_parse_dt(&pdev->dev, &defdata);
if (ret < 0) {
dev_err(&pdev->dev, "failed to find platform data\n");
return ret;
}
data = &defdata;
}
-------
This looks like if we set the platform data for pwm backlight device,
"pwm_backlight_parse_dt" will never have a chance to be called, which
means the stuffs I defined in backlight DT node will be ignored.
If I don't set the platform data for pwm backlight device, according to
the pwm_backlight_probe, I will never have a chance to set some fops
which I need(like "notify", "check_fb"...).
So, what I suppose to do now? Maybe there is a way to set function
pointers in DT?
Mark
On Thu, Oct 17, 2013 at 02:49:57PM +0800, Mark Zhang wrote:
> Hi,
>
> This is the first time I send mail to linux-pwm, I didn't read through
> the mails in this list, so if somebody already asked this question, I'm
> sorry about that.
>
> I wanna set some fops in "struct platform_pwm_backlight_data". But the
> currrent probe function in pwm_bl.c says:
>
> -------
> if (!data) {
> ret = pwm_backlight_parse_dt(&pdev->dev, &defdata);
> if (ret < 0) {
> dev_err(&pdev->dev, "failed to find platform data\n");
> return ret;
> }
>
> data = &defdata;
> }
> -------
>
> This looks like if we set the platform data for pwm backlight device,
> "pwm_backlight_parse_dt" will never have a chance to be called, which
> means the stuffs I defined in backlight DT node will be ignored.
>
> If I don't set the platform data for pwm backlight device, according to
> the pwm_backlight_probe, I will never have a chance to set some fops
> which I need(like "notify", "check_fb"...).
>
> So, what I suppose to do now? Maybe there is a way to set function
> pointers in DT?
Perhaps you could describe in more detail what you need the functions
for.
Generally you're not supposed to mix DT and platform data. Without more
info that's about all I can say.
Thierry
On 10/17/2013 03:14 PM, Thierry Reding wrote:
> On Thu, Oct 17, 2013 at 02:49:57PM +0800, Mark Zhang wrote:
>> Hi,
>>
>> This is the first time I send mail to linux-pwm, I didn't read through
>> the mails in this list, so if somebody already asked this question, I'm
>> sorry about that.
>>
>> I wanna set some fops in "struct platform_pwm_backlight_data". But the
>> currrent probe function in pwm_bl.c says:
>>
>> -------
>> if (!data) {
>> ret = pwm_backlight_parse_dt(&pdev->dev, &defdata);
>> if (ret < 0) {
>> dev_err(&pdev->dev, "failed to find platform data\n");
>> return ret;
>> }
>>
>> data = &defdata;
>> }
>> -------
>>
>> This looks like if we set the platform data for pwm backlight device,
>> "pwm_backlight_parse_dt" will never have a chance to be called, which
>> means the stuffs I defined in backlight DT node will be ignored.
>>
>> If I don't set the platform data for pwm backlight device, according to
>> the pwm_backlight_probe, I will never have a chance to set some fops
>> which I need(like "notify", "check_fb"...).
>>
>> So, what I suppose to do now? Maybe there is a way to set function
>> pointers in DT?
>
> Perhaps you could describe in more detail what you need the functions
> for.
>
Okay, I just want to set the "notify" function pointer in "struct
platform_pwm_backlight_data", because I want to tune the brightness
value before the pwm-bl sets the brightness to hardware. I don't know
how to do that, unless we define the platform data explicitly.
Mark
> Generally you're not supposed to mix DT and platform data. Without more
> info that's about all I can say.
>
> Thierry
>
Ping Thierry...
Mark
On 10/18/2013 12:48 PM, Mark Zhang wrote:
> On 10/17/2013 03:14 PM, Thierry Reding wrote:
>> On Thu, Oct 17, 2013 at 02:49:57PM +0800, Mark Zhang wrote:
>>> Hi,
>>>
>>> This is the first time I send mail to linux-pwm, I didn't read through
>>> the mails in this list, so if somebody already asked this question, I'm
>>> sorry about that.
>>>
>>> I wanna set some fops in "struct platform_pwm_backlight_data". But the
>>> currrent probe function in pwm_bl.c says:
>>>
>>> -------
>>> if (!data) {
>>> ret = pwm_backlight_parse_dt(&pdev->dev, &defdata);
>>> if (ret < 0) {
>>> dev_err(&pdev->dev, "failed to find platform data\n");
>>> return ret;
>>> }
>>>
>>> data = &defdata;
>>> }
>>> -------
>>>
>>> This looks like if we set the platform data for pwm backlight device,
>>> "pwm_backlight_parse_dt" will never have a chance to be called, which
>>> means the stuffs I defined in backlight DT node will be ignored.
>>>
>>> If I don't set the platform data for pwm backlight device, according to
>>> the pwm_backlight_probe, I will never have a chance to set some fops
>>> which I need(like "notify", "check_fb"...).
>>>
>>> So, what I suppose to do now? Maybe there is a way to set function
>>> pointers in DT?
>>
>> Perhaps you could describe in more detail what you need the functions
>> for.
>>
>
> Okay, I just want to set the "notify" function pointer in "struct
> platform_pwm_backlight_data", because I want to tune the brightness
> value before the pwm-bl sets the brightness to hardware. I don't know
> how to do that, unless we define the platform data explicitly.
>
> Mark
>> Generally you're not supposed to mix DT and platform data. Without more
>> info that's about all I can say.
>>
>> Thierry
>>
On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
> On 10/17/2013 03:14 PM, Thierry Reding wrote:
> > On Thu, Oct 17, 2013 at 02:49:57PM +0800, Mark Zhang wrote:
> >> Hi,
> >>
> >> This is the first time I send mail to linux-pwm, I didn't read through
> >> the mails in this list, so if somebody already asked this question, I'm
> >> sorry about that.
> >>
> >> I wanna set some fops in "struct platform_pwm_backlight_data". But the
> >> currrent probe function in pwm_bl.c says:
> >>
> >> -------
> >> if (!data) {
> >> ret = pwm_backlight_parse_dt(&pdev->dev, &defdata);
> >> if (ret < 0) {
> >> dev_err(&pdev->dev, "failed to find platform data\n");
> >> return ret;
> >> }
> >>
> >> data = &defdata;
> >> }
> >> -------
> >>
> >> This looks like if we set the platform data for pwm backlight device,
> >> "pwm_backlight_parse_dt" will never have a chance to be called, which
> >> means the stuffs I defined in backlight DT node will be ignored.
> >>
> >> If I don't set the platform data for pwm backlight device, according to
> >> the pwm_backlight_probe, I will never have a chance to set some fops
> >> which I need(like "notify", "check_fb"...).
> >>
> >> So, what I suppose to do now? Maybe there is a way to set function
> >> pointers in DT?
> >
> > Perhaps you could describe in more detail what you need the functions
> > for.
> >
>
> Okay, I just want to set the "notify" function pointer in "struct
> platform_pwm_backlight_data", because I want to tune the brightness
> value before the pwm-bl sets the brightness to hardware. I don't know
> how to do that, unless we define the platform data explicitly.
Okay, my question should have been what you need the functions for and
why you think you need them.
Thierry
On 10/22/2013 03:24 PM, Thierry Reding wrote:
> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
[...]
>>>
>>
>> Okay, I just want to set the "notify" function pointer in "struct
>> platform_pwm_backlight_data", because I want to tune the brightness
>> value before the pwm-bl sets the brightness to hardware. I don't know
>> how to do that, unless we define the platform data explicitly.
>
> Okay, my question should have been what you need the functions for and
> why you think you need them.
>
If I understanding you correctly, I suppose I've said that: "because I
want to tune the brightness value before the pwm-bl sets the brightness
to hardware".
Mark
> Thierry
>
On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
> On 10/22/2013 03:24 PM, Thierry Reding wrote:
> > On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
> [...]
> >>>
> >>
> >> Okay, I just want to set the "notify" function pointer in "struct
> >> platform_pwm_backlight_data", because I want to tune the brightness
> >> value before the pwm-bl sets the brightness to hardware. I don't know
> >> how to do that, unless we define the platform data explicitly.
> >
> > Okay, my question should have been what you need the functions for and
> > why you think you need them.
> >
>
> If I understanding you correctly, I suppose I've said that: "because I
> want to tune the brightness value before the pwm-bl sets the brightness
> to hardware".
Why do you want to tune the brightness value? What are you trying to
achieve?
Thierry
On 10/22/2013 08:49 PM, Thierry Reding wrote:
> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
>> [...]
>>>>>
>>>>
>>>> Okay, I just want to set the "notify" function pointer in "struct
>>>> platform_pwm_backlight_data", because I want to tune the brightness
>>>> value before the pwm-bl sets the brightness to hardware. I don't know
>>>> how to do that, unless we define the platform data explicitly.
>>>
>>> Okay, my question should have been what you need the functions for and
>>> why you think you need them.
>>>
>>
>> If I understanding you correctly, I suppose I've said that: "because I
>> want to tune the brightness value before the pwm-bl sets the brightness
>> to hardware".
>
> Why do you want to tune the brightness value? What are you trying to
> achieve?
>
Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
color value to make the display looks bright so that we can reduce the
backlight brightness to save power. So everytime PRISM is triggered, we
call "backlight_update_status", then in the "notify" callback, we change
the brightness value which pwm-bl provides by considering the PRISM
compensations.
Mark
> Thierry
>
On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
> On 10/22/2013 08:49 PM, Thierry Reding wrote:
> > On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
> >> On 10/22/2013 03:24 PM, Thierry Reding wrote:
> >>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
> >> [...]
> >>>>>
> >>>>
> >>>> Okay, I just want to set the "notify" function pointer in "struct
> >>>> platform_pwm_backlight_data", because I want to tune the brightness
> >>>> value before the pwm-bl sets the brightness to hardware. I don't know
> >>>> how to do that, unless we define the platform data explicitly.
> >>>
> >>> Okay, my question should have been what you need the functions for and
> >>> why you think you need them.
> >>>
> >>
> >> If I understanding you correctly, I suppose I've said that: "because I
> >> want to tune the brightness value before the pwm-bl sets the brightness
> >> to hardware".
> >
> > Why do you want to tune the brightness value? What are you trying to
> > achieve?
> >
>
> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
> color value to make the display looks bright so that we can reduce the
> backlight brightness to save power. So everytime PRISM is triggered, we
> call "backlight_update_status", then in the "notify" callback, we change
> the brightness value which pwm-bl provides by considering the PRISM
> compensations.
If you automatically call backlight_update_status() everytime PRISM
gives you new data, can't you just pass the correct value in in the
first place so that you don't have to tweak it in the .notify()
callback?
Thierry
On 10/23/2013 04:00 PM, Thierry Reding wrote:
> On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
>> On 10/22/2013 08:49 PM, Thierry Reding wrote:
>>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
>>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
>>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
>>>> [...]
>>>>>>>
>>>>>>
>>>>>> Okay, I just want to set the "notify" function pointer in "struct
>>>>>> platform_pwm_backlight_data", because I want to tune the brightness
>>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
>>>>>> how to do that, unless we define the platform data explicitly.
>>>>>
>>>>> Okay, my question should have been what you need the functions for and
>>>>> why you think you need them.
>>>>>
>>>>
>>>> If I understanding you correctly, I suppose I've said that: "because I
>>>> want to tune the brightness value before the pwm-bl sets the brightness
>>>> to hardware".
>>>
>>> Why do you want to tune the brightness value? What are you trying to
>>> achieve?
>>>
>>
>> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
>> color value to make the display looks bright so that we can reduce the
>> backlight brightness to save power. So everytime PRISM is triggered, we
>> call "backlight_update_status", then in the "notify" callback, we change
>> the brightness value which pwm-bl provides by considering the PRISM
>> compensations.
>
> If you automatically call backlight_update_status() everytime PRISM
> gives you new data, can't you just pass the correct value in in the
> first place so that you don't have to tweak it in the .notify()
> callback?
OK, how to do that? -- "pass the correct value in in the first place"?
Mark
>
> Thierry
>
On Wed, Oct 23, 2013 at 04:49:29PM +0800, Mark Zhang wrote:
> On 10/23/2013 04:00 PM, Thierry Reding wrote:
> > On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
> >> On 10/22/2013 08:49 PM, Thierry Reding wrote:
> >>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
> >>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
> >>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
> >>>> [...]
> >>>>>>>
> >>>>>>
> >>>>>> Okay, I just want to set the "notify" function pointer in "struct
> >>>>>> platform_pwm_backlight_data", because I want to tune the brightness
> >>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
> >>>>>> how to do that, unless we define the platform data explicitly.
> >>>>>
> >>>>> Okay, my question should have been what you need the functions for and
> >>>>> why you think you need them.
> >>>>>
> >>>>
> >>>> If I understanding you correctly, I suppose I've said that: "because I
> >>>> want to tune the brightness value before the pwm-bl sets the brightness
> >>>> to hardware".
> >>>
> >>> Why do you want to tune the brightness value? What are you trying to
> >>> achieve?
> >>>
> >>
> >> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
> >> color value to make the display looks bright so that we can reduce the
> >> backlight brightness to save power. So everytime PRISM is triggered, we
> >> call "backlight_update_status", then in the "notify" callback, we change
> >> the brightness value which pwm-bl provides by considering the PRISM
> >> compensations.
> >
> > If you automatically call backlight_update_status() everytime PRISM
> > gives you new data, can't you just pass the correct value in in the
> > first place so that you don't have to tweak it in the .notify()
> > callback?
>
> OK, how to do that? -- "pass the correct value in in the first place"?
Well, if you call backlight_update_status(), then you can pass in a
brightness value, right? You usually do that by setting the backlight's
props.brightness field.
So when PRISM gives you new data, you could just read out the current
brightness, compute the new one based on the current one and the PRISM
data, set the props.brightness field to that value and then call
backlight_update_status().
Thierry
On 10/23/2013 05:09 PM, Thierry Reding wrote:
> On Wed, Oct 23, 2013 at 04:49:29PM +0800, Mark Zhang wrote:
>> On 10/23/2013 04:00 PM, Thierry Reding wrote:
>>> On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
>>>> On 10/22/2013 08:49 PM, Thierry Reding wrote:
>>>>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
>>>>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
>>>>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
>>>>>> [...]
>>>>>>>>>
>>>>>>>>
>>>>>>>> Okay, I just want to set the "notify" function pointer in "struct
>>>>>>>> platform_pwm_backlight_data", because I want to tune the brightness
>>>>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
>>>>>>>> how to do that, unless we define the platform data explicitly.
>>>>>>>
>>>>>>> Okay, my question should have been what you need the functions for and
>>>>>>> why you think you need them.
>>>>>>>
>>>>>>
>>>>>> If I understanding you correctly, I suppose I've said that: "because I
>>>>>> want to tune the brightness value before the pwm-bl sets the brightness
>>>>>> to hardware".
>>>>>
>>>>> Why do you want to tune the brightness value? What are you trying to
>>>>> achieve?
>>>>>
>>>>
>>>> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
>>>> color value to make the display looks bright so that we can reduce the
>>>> backlight brightness to save power. So everytime PRISM is triggered, we
>>>> call "backlight_update_status", then in the "notify" callback, we change
>>>> the brightness value which pwm-bl provides by considering the PRISM
>>>> compensations.
>>>
>>> If you automatically call backlight_update_status() everytime PRISM
>>> gives you new data, can't you just pass the correct value in in the
>>> first place so that you don't have to tweak it in the .notify()
>>> callback?
>>
>> OK, how to do that? -- "pass the correct value in in the first place"?
>
> Well, if you call backlight_update_status(), then you can pass in a
> brightness value, right? You usually do that by setting the backlight's
> props.brightness field.
>
> So when PRISM gives you new data, you could just read out the current
> brightness, compute the new one based on the current one and the PRISM
> data, set the props.brightness field to that value and then call
> backlight_update_status().
>
The param of "backlight_update_status" is "struct backlight_device *".
So you mean after I get a pointer of correct backlight device, just set
the brightness value I want to it's "props.brightness" directly?
I think the "struct backlight_device" should be opaque(although it's
definition is in include/linux/backlight.h, I know that), so it's better
not to touch it's member directly, that's why I wanna use that "notify"
callback.
Mark
> Thierry
>
On 10/23/2013 05:09 PM, Thierry Reding wrote:
> On Wed, Oct 23, 2013 at 04:49:29PM +0800, Mark Zhang wrote:
>> On 10/23/2013 04:00 PM, Thierry Reding wrote:
>>> On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
>>>> On 10/22/2013 08:49 PM, Thierry Reding wrote:
>>>>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
>>>>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
>>>>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
>>>>>> [...]
>>>>>>>>>
>>>>>>>>
>>>>>>>> Okay, I just want to set the "notify" function pointer in "struct
>>>>>>>> platform_pwm_backlight_data", because I want to tune the brightness
>>>>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
>>>>>>>> how to do that, unless we define the platform data explicitly.
>>>>>>>
>>>>>>> Okay, my question should have been what you need the functions for and
>>>>>>> why you think you need them.
>>>>>>>
>>>>>>
>>>>>> If I understanding you correctly, I suppose I've said that: "because I
>>>>>> want to tune the brightness value before the pwm-bl sets the brightness
>>>>>> to hardware".
>>>>>
>>>>> Why do you want to tune the brightness value? What are you trying to
>>>>> achieve?
>>>>>
>>>>
>>>> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
>>>> color value to make the display looks bright so that we can reduce the
>>>> backlight brightness to save power. So everytime PRISM is triggered, we
>>>> call "backlight_update_status", then in the "notify" callback, we change
>>>> the brightness value which pwm-bl provides by considering the PRISM
>>>> compensations.
>>>
>>> If you automatically call backlight_update_status() everytime PRISM
>>> gives you new data, can't you just pass the correct value in in the
>>> first place so that you don't have to tweak it in the .notify()
>>> callback?
>>
>> OK, how to do that? -- "pass the correct value in in the first place"?
>
> Well, if you call backlight_update_status(), then you can pass in a
> brightness value, right? You usually do that by setting the backlight's
> props.brightness field.
>
> So when PRISM gives you new data, you could just read out the current
> brightness, compute the new one based on the current one and the PRISM
> data, set the props.brightness field to that value and then call
> backlight_update_status().
>
Okay, anyway, I got the idea. Actually it's simpler. :)
I'm just curious that if we can do that in this way, why we need
"notify" anymore?
Mark
> Thierry
>
On 10/23/2013 06:36 PM, Mark Zhang wrote:
> On 10/23/2013 05:09 PM, Thierry Reding wrote:
>> On Wed, Oct 23, 2013 at 04:49:29PM +0800, Mark Zhang wrote:
>>> On 10/23/2013 04:00 PM, Thierry Reding wrote:
>>>> On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
>>>>> On 10/22/2013 08:49 PM, Thierry Reding wrote:
>>>>>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
>>>>>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
>>>>>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
>>>>>>> [...]
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Okay, I just want to set the "notify" function pointer in "struct
>>>>>>>>> platform_pwm_backlight_data", because I want to tune the brightness
>>>>>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
>>>>>>>>> how to do that, unless we define the platform data explicitly.
>>>>>>>>
>>>>>>>> Okay, my question should have been what you need the functions for and
>>>>>>>> why you think you need them.
>>>>>>>>
>>>>>>>
>>>>>>> If I understanding you correctly, I suppose I've said that: "because I
>>>>>>> want to tune the brightness value before the pwm-bl sets the brightness
>>>>>>> to hardware".
>>>>>>
>>>>>> Why do you want to tune the brightness value? What are you trying to
>>>>>> achieve?
>>>>>>
>>>>>
>>>>> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
>>>>> color value to make the display looks bright so that we can reduce the
>>>>> backlight brightness to save power. So everytime PRISM is triggered, we
>>>>> call "backlight_update_status", then in the "notify" callback, we change
>>>>> the brightness value which pwm-bl provides by considering the PRISM
>>>>> compensations.
>>>>
>>>> If you automatically call backlight_update_status() everytime PRISM
>>>> gives you new data, can't you just pass the correct value in in the
>>>> first place so that you don't have to tweak it in the .notify()
>>>> callback?
>>>
>>> OK, how to do that? -- "pass the correct value in in the first place"?
>>
>> Well, if you call backlight_update_status(), then you can pass in a
>> brightness value, right? You usually do that by setting the backlight's
>> props.brightness field.
>>
>> So when PRISM gives you new data, you could just read out the current
>> brightness, compute the new one based on the current one and the PRISM
>> data, set the props.brightness field to that value and then call
>> backlight_update_status().
>>
>
> Okay, anyway, I got the idea. Actually it's simpler. :)
> I'm just curious that if we can do that in this way, why we need
> "notify" anymore?
Oh, by the way, how about "check_fb" fops? Is there some kind of
substitution as well? I mean, if I wanna set "check_fb" and also want to
define the backlight device via DT, is there a way to do that?
Mark
>
> Mark
>> Thierry
>>
On Wed, Oct 23, 2013 at 06:36:01PM +0800, Mark Zhang wrote:
> On 10/23/2013 05:09 PM, Thierry Reding wrote:
> > On Wed, Oct 23, 2013 at 04:49:29PM +0800, Mark Zhang wrote:
> >> On 10/23/2013 04:00 PM, Thierry Reding wrote:
> >>> On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
> >>>> On 10/22/2013 08:49 PM, Thierry Reding wrote:
> >>>>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
> >>>>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
> >>>>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
> >>>>>> [...]
> >>>>>>>>>
> >>>>>>>>
> >>>>>>>> Okay, I just want to set the "notify" function pointer in "struct
> >>>>>>>> platform_pwm_backlight_data", because I want to tune the brightness
> >>>>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
> >>>>>>>> how to do that, unless we define the platform data explicitly.
> >>>>>>>
> >>>>>>> Okay, my question should have been what you need the functions for and
> >>>>>>> why you think you need them.
> >>>>>>>
> >>>>>>
> >>>>>> If I understanding you correctly, I suppose I've said that: "because I
> >>>>>> want to tune the brightness value before the pwm-bl sets the brightness
> >>>>>> to hardware".
> >>>>>
> >>>>> Why do you want to tune the brightness value? What are you trying to
> >>>>> achieve?
> >>>>>
> >>>>
> >>>> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
> >>>> color value to make the display looks bright so that we can reduce the
> >>>> backlight brightness to save power. So everytime PRISM is triggered, we
> >>>> call "backlight_update_status", then in the "notify" callback, we change
> >>>> the brightness value which pwm-bl provides by considering the PRISM
> >>>> compensations.
> >>>
> >>> If you automatically call backlight_update_status() everytime PRISM
> >>> gives you new data, can't you just pass the correct value in in the
> >>> first place so that you don't have to tweak it in the .notify()
> >>> callback?
> >>
> >> OK, how to do that? -- "pass the correct value in in the first place"?
> >
> > Well, if you call backlight_update_status(), then you can pass in a
> > brightness value, right? You usually do that by setting the backlight's
> > props.brightness field.
> >
> > So when PRISM gives you new data, you could just read out the current
> > brightness, compute the new one based on the current one and the PRISM
> > data, set the props.brightness field to that value and then call
> > backlight_update_status().
> >
>
> Okay, anyway, I got the idea. Actually it's simpler. :)
> I'm just curious that if we can do that in this way, why we need
> "notify" anymore?
We don't need it any longer for DT, you've just proven that. =)
However we still need to support board files, and the .notify() callback
is the only way for board files to use board specific means to enable or
disable the backlight depending on the brightness value.
Thierry
On Wed, Oct 23, 2013 at 06:31:30PM +0800, Mark Zhang wrote:
> On 10/23/2013 05:09 PM, Thierry Reding wrote:
> > On Wed, Oct 23, 2013 at 04:49:29PM +0800, Mark Zhang wrote:
> >> On 10/23/2013 04:00 PM, Thierry Reding wrote:
> >>> On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
> >>>> On 10/22/2013 08:49 PM, Thierry Reding wrote:
> >>>>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
> >>>>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
> >>>>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
> >>>>>> [...]
> >>>>>>>>>
> >>>>>>>>
> >>>>>>>> Okay, I just want to set the "notify" function pointer in "struct
> >>>>>>>> platform_pwm_backlight_data", because I want to tune the brightness
> >>>>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
> >>>>>>>> how to do that, unless we define the platform data explicitly.
> >>>>>>>
> >>>>>>> Okay, my question should have been what you need the functions for and
> >>>>>>> why you think you need them.
> >>>>>>>
> >>>>>>
> >>>>>> If I understanding you correctly, I suppose I've said that: "because I
> >>>>>> want to tune the brightness value before the pwm-bl sets the brightness
> >>>>>> to hardware".
> >>>>>
> >>>>> Why do you want to tune the brightness value? What are you trying to
> >>>>> achieve?
> >>>>>
> >>>>
> >>>> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
> >>>> color value to make the display looks bright so that we can reduce the
> >>>> backlight brightness to save power. So everytime PRISM is triggered, we
> >>>> call "backlight_update_status", then in the "notify" callback, we change
> >>>> the brightness value which pwm-bl provides by considering the PRISM
> >>>> compensations.
> >>>
> >>> If you automatically call backlight_update_status() everytime PRISM
> >>> gives you new data, can't you just pass the correct value in in the
> >>> first place so that you don't have to tweak it in the .notify()
> >>> callback?
> >>
> >> OK, how to do that? -- "pass the correct value in in the first place"?
> >
> > Well, if you call backlight_update_status(), then you can pass in a
> > brightness value, right? You usually do that by setting the backlight's
> > props.brightness field.
> >
> > So when PRISM gives you new data, you could just read out the current
> > brightness, compute the new one based on the current one and the PRISM
> > data, set the props.brightness field to that value and then call
> > backlight_update_status().
> >
>
> The param of "backlight_update_status" is "struct backlight_device *".
> So you mean after I get a pointer of correct backlight device, just set
> the brightness value I want to it's "props.brightness" directly?
Yes.
> I think the "struct backlight_device" should be opaque(although it's
> definition is in include/linux/backlight.h, I know that), so it's better
> not to touch it's member directly, that's why I wanna use that "notify"
> callback.
Well, backlight_update_status() is the only way to change the brightness
of a backlight. If nobody ever calls backlight_update_status() then the
.notify() callback will never be called either.
By the way, what method do you use to control the backlight brightness?
Do you use the sysfs interface from userspace?
Thierry
On Wed, Oct 23, 2013 at 06:46:17PM +0800, Mark Zhang wrote:
> On 10/23/2013 06:36 PM, Mark Zhang wrote:
> > On 10/23/2013 05:09 PM, Thierry Reding wrote:
> >> On Wed, Oct 23, 2013 at 04:49:29PM +0800, Mark Zhang wrote:
> >>> On 10/23/2013 04:00 PM, Thierry Reding wrote:
> >>>> On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
> >>>>> On 10/22/2013 08:49 PM, Thierry Reding wrote:
> >>>>>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
> >>>>>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
> >>>>>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
> >>>>>>> [...]
> >>>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Okay, I just want to set the "notify" function pointer in "struct
> >>>>>>>>> platform_pwm_backlight_data", because I want to tune the brightness
> >>>>>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
> >>>>>>>>> how to do that, unless we define the platform data explicitly.
> >>>>>>>>
> >>>>>>>> Okay, my question should have been what you need the functions for and
> >>>>>>>> why you think you need them.
> >>>>>>>>
> >>>>>>>
> >>>>>>> If I understanding you correctly, I suppose I've said that: "because I
> >>>>>>> want to tune the brightness value before the pwm-bl sets the brightness
> >>>>>>> to hardware".
> >>>>>>
> >>>>>> Why do you want to tune the brightness value? What are you trying to
> >>>>>> achieve?
> >>>>>>
> >>>>>
> >>>>> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
> >>>>> color value to make the display looks bright so that we can reduce the
> >>>>> backlight brightness to save power. So everytime PRISM is triggered, we
> >>>>> call "backlight_update_status", then in the "notify" callback, we change
> >>>>> the brightness value which pwm-bl provides by considering the PRISM
> >>>>> compensations.
> >>>>
> >>>> If you automatically call backlight_update_status() everytime PRISM
> >>>> gives you new data, can't you just pass the correct value in in the
> >>>> first place so that you don't have to tweak it in the .notify()
> >>>> callback?
> >>>
> >>> OK, how to do that? -- "pass the correct value in in the first place"?
> >>
> >> Well, if you call backlight_update_status(), then you can pass in a
> >> brightness value, right? You usually do that by setting the backlight's
> >> props.brightness field.
> >>
> >> So when PRISM gives you new data, you could just read out the current
> >> brightness, compute the new one based on the current one and the PRISM
> >> data, set the props.brightness field to that value and then call
> >> backlight_update_status().
> >>
> >
> > Okay, anyway, I got the idea. Actually it's simpler. :)
> > I'm just curious that if we can do that in this way, why we need
> > "notify" anymore?
>
> Oh, by the way, how about "check_fb" fops? Is there some kind of
> substitution as well? I mean, if I wanna set "check_fb" and also want to
> define the backlight device via DT, is there a way to do that?
No, there's no substitution for that. .check_fb() is used to verify that
a backlight device is associated with a framebuffer device. There are
better ways to check for that with DT, although nothing has been merged
into mainline yet.
Thierry
On 10/23/2013 06:54 PM, Thierry Reding wrote:
> On Wed, Oct 23, 2013 at 06:31:30PM +0800, Mark Zhang wrote:
>> On 10/23/2013 05:09 PM, Thierry Reding wrote:
>>> On Wed, Oct 23, 2013 at 04:49:29PM +0800, Mark Zhang wrote:
>>>> On 10/23/2013 04:00 PM, Thierry Reding wrote:
>>>>> On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
>>>>>> On 10/22/2013 08:49 PM, Thierry Reding wrote:
>>>>>>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
>>>>>>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
>>>>>>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
>>>>>>>> [...]
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Okay, I just want to set the "notify" function pointer in "struct
>>>>>>>>>> platform_pwm_backlight_data", because I want to tune the brightness
>>>>>>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
>>>>>>>>>> how to do that, unless we define the platform data explicitly.
>>>>>>>>>
>>>>>>>>> Okay, my question should have been what you need the functions for and
>>>>>>>>> why you think you need them.
>>>>>>>>>
>>>>>>>>
>>>>>>>> If I understanding you correctly, I suppose I've said that: "because I
>>>>>>>> want to tune the brightness value before the pwm-bl sets the brightness
>>>>>>>> to hardware".
>>>>>>>
>>>>>>> Why do you want to tune the brightness value? What are you trying to
>>>>>>> achieve?
>>>>>>>
>>>>>>
>>>>>> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
>>>>>> color value to make the display looks bright so that we can reduce the
>>>>>> backlight brightness to save power. So everytime PRISM is triggered, we
>>>>>> call "backlight_update_status", then in the "notify" callback, we change
>>>>>> the brightness value which pwm-bl provides by considering the PRISM
>>>>>> compensations.
>>>>>
>>>>> If you automatically call backlight_update_status() everytime PRISM
>>>>> gives you new data, can't you just pass the correct value in in the
>>>>> first place so that you don't have to tweak it in the .notify()
>>>>> callback?
>>>>
>>>> OK, how to do that? -- "pass the correct value in in the first place"?
>>>
>>> Well, if you call backlight_update_status(), then you can pass in a
>>> brightness value, right? You usually do that by setting the backlight's
>>> props.brightness field.
>>>
>>> So when PRISM gives you new data, you could just read out the current
>>> brightness, compute the new one based on the current one and the PRISM
>>> data, set the props.brightness field to that value and then call
>>> backlight_update_status().
>>>
>>
>> The param of "backlight_update_status" is "struct backlight_device *".
>> So you mean after I get a pointer of correct backlight device, just set
>> the brightness value I want to it's "props.brightness" directly?
>
> Yes.
>
>> I think the "struct backlight_device" should be opaque(although it's
>> definition is in include/linux/backlight.h, I know that), so it's better
>> not to touch it's member directly, that's why I wanna use that "notify"
>> callback.
>
> Well, backlight_update_status() is the only way to change the brightness
> of a backlight. If nobody ever calls backlight_update_status() then the
> .notify() callback will never be called either.
>
Yes, so this is exactly my understanding before I posted this thread. I
think "struct backlight_device" is a handle, we get it and call
"backlight_update_status", then if we wanna change it's default behavior
or brightness value, we overwrite the fops it provides.
> By the way, what method do you use to control the backlight brightness?
> Do you use the sysfs interface from userspace?
>
Yes.
Mark
> Thierry
>
On 10/23/2013 06:58 PM, Thierry Reding wrote:
> On Wed, Oct 23, 2013 at 06:46:17PM +0800, Mark Zhang wrote:
>> On 10/23/2013 06:36 PM, Mark Zhang wrote:
>>> On 10/23/2013 05:09 PM, Thierry Reding wrote:
>>>> On Wed, Oct 23, 2013 at 04:49:29PM +0800, Mark Zhang wrote:
>>>>> On 10/23/2013 04:00 PM, Thierry Reding wrote:
>>>>>> On Wed, Oct 23, 2013 at 10:16:24AM +0800, Mark Zhang wrote:
>>>>>>> On 10/22/2013 08:49 PM, Thierry Reding wrote:
>>>>>>>> On Tue, Oct 22, 2013 at 04:55:09PM +0800, Mark Zhang wrote:
>>>>>>>>> On 10/22/2013 03:24 PM, Thierry Reding wrote:
>>>>>>>>>> On Fri, Oct 18, 2013 at 12:48:12PM +0800, Mark Zhang wrote:
>>>>>>>>> [...]
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Okay, I just want to set the "notify" function pointer in "struct
>>>>>>>>>>> platform_pwm_backlight_data", because I want to tune the brightness
>>>>>>>>>>> value before the pwm-bl sets the brightness to hardware. I don't know
>>>>>>>>>>> how to do that, unless we define the platform data explicitly.
>>>>>>>>>>
>>>>>>>>>> Okay, my question should have been what you need the functions for and
>>>>>>>>>> why you think you need them.
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> If I understanding you correctly, I suppose I've said that: "because I
>>>>>>>>> want to tune the brightness value before the pwm-bl sets the brightness
>>>>>>>>> to hardware".
>>>>>>>>
>>>>>>>> Why do you want to tune the brightness value? What are you trying to
>>>>>>>> achieve?
>>>>>>>>
>>>>>>>
>>>>>>> Oh, Tegra has a feature named PRISM(aka SmartDimmer). It changes the
>>>>>>> color value to make the display looks bright so that we can reduce the
>>>>>>> backlight brightness to save power. So everytime PRISM is triggered, we
>>>>>>> call "backlight_update_status", then in the "notify" callback, we change
>>>>>>> the brightness value which pwm-bl provides by considering the PRISM
>>>>>>> compensations.
>>>>>>
>>>>>> If you automatically call backlight_update_status() everytime PRISM
>>>>>> gives you new data, can't you just pass the correct value in in the
>>>>>> first place so that you don't have to tweak it in the .notify()
>>>>>> callback?
>>>>>
>>>>> OK, how to do that? -- "pass the correct value in in the first place"?
>>>>
>>>> Well, if you call backlight_update_status(), then you can pass in a
>>>> brightness value, right? You usually do that by setting the backlight's
>>>> props.brightness field.
>>>>
>>>> So when PRISM gives you new data, you could just read out the current
>>>> brightness, compute the new one based on the current one and the PRISM
>>>> data, set the props.brightness field to that value and then call
>>>> backlight_update_status().
>>>>
>>>
>>> Okay, anyway, I got the idea. Actually it's simpler. :)
>>> I'm just curious that if we can do that in this way, why we need
>>> "notify" anymore?
>>
>> Oh, by the way, how about "check_fb" fops? Is there some kind of
>> substitution as well? I mean, if I wanna set "check_fb" and also want to
>> define the backlight device via DT, is there a way to do that?
>
> No, there's no substitution for that. .check_fb() is used to verify that
> a backlight device is associated with a framebuffer device. There are
> better ways to check for that with DT, although nothing has been merged
> into mainline yet.
>
Okay, thanks for the explanation.
Mark
> Thierry
>