pwm_bl: Add support for backlight enable GPIO
Subsystems (3)
PATCH 1/1 pwm_bl: Add support for backlight enable GPIO
The backlight enable GPIO is specified in the device tree node for backlight.
--- .../bindings/video/backlight/pwm-backlight.txt | 2 ++ drivers/video/backlight/pwm_bl.c | 32 +++++++++++++++++--- include/linux/pwm_backlight.h | 2 ++ 3 files changed, 32 insertions(+), 4 deletions(-)
Documentation/devicetree/bindings/video/backlight/pwm-backlight.txt +2 -0
@@ -14,6 +14,8 @@ Required properties: Optional properties: - pwm-names: a list of names for the PWM devices specified in the "pwms" property (see PWM binding[0]) + - enable-gpio: a GPIO that needs to be used to enable the backlight + - enable-gpio-active-high: polarity of GPIO is active high (default is low) [0]: Documentation/devicetree/bindings/pwm/pwm.txt
drivers/video/backlight/pwm_bl.c +28 -4
@@ -20,10 +20,13 @@#include <linux/pwm.h> #include <linux/pwm_backlight.h> #include <linux/slab.h> +#include <linux/of_gpio.h> struct pwm_bl_data {struct pwm_device *pwm;struct device *dev; +int enable_gpio; +unsigned int enable_gpio_flags;unsigned int period;unsigned int lth_brightness;unsigned int *levels;
@@ -146,10 +149,15 @@ static int pwm_backlight_parse_dt(struct device *dev, }/* -* TODO: Most users of this driver use a number of GPIOs to control -* backlight power. Support for specifying these needs to be -* added. +* If "enable-gpio" is present, use that GPIO to enable the backlight. +* The presence (or not) of "enable-gpio-active-high" will determine +* the value of the GPIO. */ + data->enable_gpio =of_get_named_gpio (node,"enable-gpio" ,0 ); +if (of_property_read_bool (node,"enable-gpio-active-high" )) + data->enable_gpio_flags =GPIOF_OUT_INIT_HIGH ; +else + data->enable_gpio_flags =GPIOF_OUT_INIT_LOW ;return 0 ; }
@@ -207,12 +215,23 @@ static int pwm_backlight_probe(struct platform_device *pdev) }else max = data->max_brightness; + pb->enable_gpio = data->enable_gpio; + pb->enable_gpio_flags = data->enable_gpio_flags; pb->notify = data->notify; pb->notify_after = data->notify_after; pb->check_fb = data->check_fb; pb->exit = data->exit; pb->dev = &pdev->dev; +if (gpio_is_valid (pb->enable_gpio)) { + ret =gpio_request_one (pb->enable_gpio, +GPIOF_DIR_OUT | pb->enable_gpio_flags,"bl_en" ); +if (ret) { +dev_err (&pdev->dev,"failed to allocate bl_en gpio" ); + goto err_alloc; + } + } + pb->pwm =devm_pwm_get (&pdev->dev,NULL );if (IS_ERR (pb->pwm)) {dev_err (&pdev->dev,"unable to request PWM, trying legacy API\n" );
@@ -221,7 +240,7 @@ static int pwm_backlight_probe(struct platform_device *pdev)if (IS_ERR (pb->pwm)) {dev_err (&pdev->dev,"unable to request legacy PWM\n" ); ret =PTR_ERR (pb->pwm); - goto err_alloc; + goto err_gpio; } }
@@ -255,6 +274,9 @@ static int pwm_backlight_probe(struct platform_device *pdev)platform_set_drvdata (pdev, bl);return 0 ; +err_gpio: +if (gpio_is_valid (data->enable_gpio)) +gpio_free (data->enable_gpio); err_alloc:if (data->exit) data->exit (&pdev->dev);
@@ -269,6 +291,8 @@ static int pwm_backlight_remove(struct platform_device *pdev)backlight_device_unregister (bl);pwm_config (pb->pwm,0 , pb->period);pwm_disable (pb->pwm); +if (gpio_is_valid (pb->enable_gpio)) +gpio_free (pb->enable_gpio);if (pb->exit) pb->exit (&pdev->dev);return 0 ;
include/linux/pwm_backlight.h +2 -0
@@ -8,6 +8,8 @@struct platform_pwm_backlight_data {int pwm_id; +int enable_gpio; +unsigned int enable_gpio_flags;unsigned int max_brightness;unsigned int dft_brightness;unsigned int lth_brightness;
Signature · 1 line
1.7.9.5
Quoted 117 lines · Andrew Chew
The backlight enable GPIO is specified in the device tree node for backlight.
Signed-off-by Andrew Chew <[email protected]>--- .../bindings/video/backlight/pwm-backlight.txt | 2 ++ drivers/video/backlight/pwm_bl.c | 32 +++++++++++++++++--- include/linux/pwm_backlight.h | 2 ++ 3 files changed, 32 insertions(+), 4 deletions(-)Documentation/devicetree/bindings/video/backlight/pwm-backlight.txt +2
-0@@ -14,6 +14,8 @@ Required properties: Optional properties: - pwm-names: a list of names for the PWM devices specified in the "pwms" property (see PWM binding[0]) + - enable-gpio: a GPIO that needs to be used to enable the backlight + - enable-gpio-active-high: polarity of GPIO is active high (default is low) [0]: Documentation/devicetree/bindings/pwm/pwm.txtdrivers/video/backlight/pwm_bl.c +28
-4@@ -20,10 +20,13 @@#include <linux/pwm.h> #include <linux/pwm_backlight.h> #include <linux/slab.h> +#include <linux/of_gpio.h> struct pwm_bl_data {struct pwm_device *pwm;struct device *dev; +int enable_gpio; +unsigned int enable_gpio_flags;unsigned int period;unsigned int lth_brightness;unsigned int *levels;@@ -146,10 +149,15 @@ static int pwm_backlight_parse_dt(struct device *dev, }/* -* TODO: Most users of this driver use a number of GPIOs to control -* backlight power. Support for specifying these needs to be -* added. +* If "enable-gpio" is present, use that GPIO to enable the backlight. +* The presence (or not) of "enable-gpio-active-high" will determine +* the value of the GPIO. */ + data->enable_gpio =of_get_named_gpio (node,"enable-gpio" ,0 ); +if (of_property_read_bool (node,"enable-gpio-active-high" )) + data->enable_gpio_flags =GPIOF_OUT_INIT_HIGH ; +else + data->enable_gpio_flags =GPIOF_OUT_INIT_LOW ;return 0 ; }@@ -207,12 +215,23 @@ static int pwm_backlight_probe(struct platform_device *pdev) }else max = data->max_brightness; + pb->enable_gpio = data->enable_gpio; + pb->enable_gpio_flags = data->enable_gpio_flags; pb->notify = data->notify; pb->notify_after = data->notify_after; pb->check_fb = data->check_fb; pb->exit = data->exit; pb->dev = &pdev->dev; +if (gpio_is_valid (pb->enable_gpio)) { + ret =gpio_request_one (pb->enable_gpio, +GPIOF_DIR_OUT | pb->enable_gpio_flags,"bl_en" ); +if (ret) { +dev_err (&pdev->dev,"failed to allocate bl_en gpio" ); + goto err_alloc; + } + } + pb->pwm =devm_pwm_get (&pdev->dev,NULL );if (IS_ERR (pb->pwm)) {dev_err (&pdev->dev,"unable to request PWM, trying legacy API\n" );@@ -221,7 +240,7 @@ static int pwm_backlight_probe(struct platform_device *pdev)if (IS_ERR (pb->pwm)) {dev_err (&pdev->dev,"unable to request legacy PWM\n" ); ret =PTR_ERR (pb->pwm); - goto err_alloc; + goto err_gpio; } }@@ -255,6 +274,9 @@ static int pwm_backlight_probe(struct platform_device *pdev)platform_set_drvdata (pdev, bl);return 0 ; +err_gpio: +if (gpio_is_valid (data->enable_gpio)) +gpio_free (data->enable_gpio); err_alloc:if (data->exit) data->exit (&pdev->dev);@@ -269,6 +291,8 @@ static int pwm_backlight_remove(struct platform_device *pdev)backlight_device_unregister (bl);pwm_config (pb->pwm,0 , pb->period);pwm_disable (pb->pwm); +if (gpio_is_valid (pb->enable_gpio)) +gpio_free (pb->enable_gpio);if (pb->exit) pb->exit (&pdev->dev);return 0 ;include/linux/pwm_backlight.h +0
-0
include/linux/pwm_backlight.h line 15
Quoted 8 lines
include/linux/pwm_backlight.h +2
-0@@ -8,6 +8,8 @@struct platform_pwm_backlight_data {int pwm_id; +int enable_gpio; +unsigned int enable_gpio_flags;unsigned int max_brightness;unsigned int dft_brightness;unsigned int lth_brightness;
Hi Andrew,
I'm Cc'ing Alexandre Courbot, who has been working on supporting this in a more generic way using power sequences. Generally this kind of support really belongs in the common display framework, but I guess we could add this one GPIO since it really is related only to the backlight. Usually more than just an enable for the backlight is required so I'm not sure how useful this really is.
Alex, any thought?
Thierry
Quoted 131 lines · Thierry Reding
* PGP Signed by an unknown key
Quoted 117 lines · Andrew Chew
The backlight enable GPIO is specified in the device tree node for backlight.
Signed-off-by Andrew Chew <[email protected]>--- .../bindings/video/backlight/pwm-backlight.txt | 2 ++ drivers/video/backlight/pwm_bl.c | 32 +++++++++++++++++--- include/linux/pwm_backlight.h | 2 ++ 3 files changed, 32 insertions(+), 4 deletions(-)Documentation/devicetree/bindings/video/backlight/pwm-backlight.txt +2
-0@@ -14,6 +14,8 @@ Required properties: Optional properties: - pwm-names: a list of names for the PWM devices specified in the "pwms" property (see PWM binding[0]) + - enable-gpio: a GPIO that needs to be used to enable the backlight + - enable-gpio-active-high: polarity of GPIO is active high (default is low) [0]: Documentation/devicetree/bindings/pwm/pwm.txtdrivers/video/backlight/pwm_bl.c +28
-4@@ -20,10 +20,13 @@#include <linux/pwm.h> #include <linux/pwm_backlight.h> #include <linux/slab.h> +#include <linux/of_gpio.h> struct pwm_bl_data {struct pwm_device *pwm;struct device *dev; +int enable_gpio; +unsigned int enable_gpio_flags;unsigned int period;unsigned int lth_brightness;unsigned int *levels;@@ -146,10 +149,15 @@ static int pwm_backlight_parse_dt(struct device *dev, }/* -* TODO: Most users of this driver use a number of GPIOs to control -* backlight power. Support for specifying these needs to be -* added. +* If "enable-gpio" is present, use that GPIO to enable the backlight. +* The presence (or not) of "enable-gpio-active-high" will determine +* the value of the GPIO. */ + data->enable_gpio =of_get_named_gpio (node,"enable-gpio" ,0 ); +if (of_property_read_bool (node,"enable-gpio-active-high" )) + data->enable_gpio_flags =GPIOF_OUT_INIT_HIGH ; +else + data->enable_gpio_flags =GPIOF_OUT_INIT_LOW ;return 0 ; }@@ -207,12 +215,23 @@ static int pwm_backlight_probe(struct platform_device *pdev) }else max = data->max_brightness; + pb->enable_gpio = data->enable_gpio; + pb->enable_gpio_flags = data->enable_gpio_flags; pb->notify = data->notify; pb->notify_after = data->notify_after; pb->check_fb = data->check_fb; pb->exit = data->exit; pb->dev = &pdev->dev; +if (gpio_is_valid (pb->enable_gpio)) { + ret =gpio_request_one (pb->enable_gpio, +GPIOF_DIR_OUT | pb->enable_gpio_flags,"bl_en" ); +if (ret) { +dev_err (&pdev->dev,"failed to allocate bl_en gpio" ); + goto err_alloc; + } + } + pb->pwm =devm_pwm_get (&pdev->dev,NULL );if (IS_ERR (pb->pwm)) {dev_err (&pdev->dev,"unable to request PWM, trying legacy API\n" );@@ -221,7 +240,7 @@ static int pwm_backlight_probe(struct platform_device *pdev)if (IS_ERR (pb->pwm)) {dev_err (&pdev->dev,"unable to request legacy PWM\n" ); ret =PTR_ERR (pb->pwm); - goto err_alloc; + goto err_gpio; } }@@ -255,6 +274,9 @@ static int pwm_backlight_probe(struct platform_device *pdev)platform_set_drvdata (pdev, bl);return 0 ; +err_gpio: +if (gpio_is_valid (data->enable_gpio)) +gpio_free (data->enable_gpio); err_alloc:if (data->exit) data->exit (&pdev->dev);@@ -269,6 +291,8 @@ static int pwm_backlight_remove(struct platform_device *pdev)backlight_device_unregister (bl);pwm_config (pb->pwm,0 , pb->period);pwm_disable (pb->pwm); +if (gpio_is_valid (pb->enable_gpio)) +gpio_free (pb->enable_gpio);if (pb->exit) pb->exit (&pdev->dev);return 0 ;include/linux/pwm_backlight.h +0
-0include/linux/pwm_backlight.h line 15
Quoted 8 lines
include/linux/pwm_backlight.h +2
-0@@ -8,6 +8,8 @@struct platform_pwm_backlight_data {int pwm_id; +int enable_gpio; +unsigned int enable_gpio_flags;unsigned int max_brightness;unsigned int dft_brightness;unsigned int lth_brightness;Hi Andrew,
Quoted 8 lines
I'm Cc'ing Alexandre Courbot, who has been working on supporting this in a more generic way using power sequences. Generally this kind of support really belongs in the common display framework, but I guess we could add this one GPIO since it really is related only to the backlight. Usually more than just an enable for the backlight is required so I'm not sure how useful this really is.
Alex, any thought?
It is very common for a GPIO to be involved in powering the backlight on, indeed. However it seems that in this patch the GPIO is set once and for all during probe and never touched afterwards. This means the backlight is still enabled (and consuming power) even when its value is zero - I'd at least like to see the GPIO disabled when this is the case to save power. Otherwise you can achieve the same result with a gpio-regulator defined to be always on in the DT, without touching the pwm-backlight driver.
Another issue is that if the GPIO is not explicitly set to -1 in the platform data, probe will try to acquire GPIO 0 and will fail. This would break compatibility with all existing users of pwm-backlight that rely on platform data.
And there is also the fact that the powering of backlights is often slightly more complicated than just an enabling GPIO - for Ventana we have at least one more regulator/GPIO involved. Maybe this regulator could be turned on forever in the DT - then pwm-backlight could use the enable GPIO to save power, but I suspect we would save even more power if we could turn the regulator off as well.
But overall I'm not against having enable GPIO support in this driver if this helps getting the job done while I finish proper power-sequence support. Andrew, does this single patch allow you to enable the backlight on some boards?
Alex.
Quoted 184 lines
From: Alex Courbot
Sent: Monday, March 04, 2013 7:00 PM
To: Thierry Reding
Cc: Andrew Chew; [email protected]
Subject: Re: [PATCH 1/1]pwm_bl: Add support for backlight enable GPIOQuoted 10 lines · Thierry Reding
* PGP Signed by an unknown key
Quoted 7 lines · Andrew Chew
The backlight enable GPIO is specified in the device tree node for backlight.
Signed-off-by Andrew Chew <[email protected]>--- .../bindings/video/backlight/pwm-backlight.txt | 2 ++ drivers/video/backlight/pwm_bl.c | 32 +++++++++++++++++---
Quoted 5 lines
Quoted 5 lines
include/linux/pwm_backlight.h | 2 ++ 3 files changed, 32 insertions(+), 4 deletions(-)diff --git a/Documentation/devicetree/bindings/video/backlight/pwm-
backlight.txt
Quoted 1 line
Quoted 1 line
b/Documentation/devicetree/bindings/video/backlight/pwm-
backlight.txt
Quoted 3 lines
Quoted 3 lines
index 1e4fc72..1ed4f0f 100644
---
a/Documentation/devicetree/bindings/video/backlight/pwm-backlight.txt
Quoted 1 line
Quoted 1 line
+++ b/Documentation/devicetree/bindings/video/backlight/pwm-
backlight
Quoted 23 lines
Quoted 23 lines
+++ .txt
Documentation/devicetree/bindings/video/backlight/pwm-backlight.txt +4
-0@@ -14,6 +14,8 @@ Required properties: Optional properties: - pwm-names: a list of names for the PWM devices specified in the "pwms" property (see PWM binding[0]) + - enable-gpio: a GPIO that needs to be used to enable the + backlight + - enable-gpio-active-high: polarity of GPIO is active high + (default is low) [0]: Documentation/devicetree/bindings/pwm/pwm.txtdrivers/video/backlight/pwm_bl.c +1
-0@@ -20,10 +20,13 @@#include <linux/pwm.h> #include <linux/pwm_backlight.h> #include <linux/slab.h> +#include <linux/of_gpio.h> Quoted 9 lines
drivers/video/backlight/pwm_bl.c static int pwm_backlight_parse_dt(struct device
Quoted 9 lines
drivers/video/backlight/pwm_bl.c +2
-0struct pwm_bl_data {struct pwm_device *pwm;struct device *dev; +int enable_gpio; +unsigned int enable_gpio_flags;unsigned int period;unsigned int lth_brightness;unsigned int *levels;@@ -146,10 +149,15 @@ static int pwm_backlight_parse_dt(struct device*dev,
Quoted 16 lines
Quoted 16 lines
drivers/video/backlight/pwm_bl.c +8
-3}/* -* TODO: Most users of this driver use a number of GPIOs to control -* backlight power. Support for specifying these needs to be -* added. +* If "enable-gpio" is present, use that GPIO to enable the backlight. +* The presence (or not) of "enable-gpio-active-high" will determine +* the value of the GPIO. */ + data->enable_gpio =of_get_named_gpio (node,"enable-gpio" ,0 ); +if (of_property_read_bool (node,"enable-gpio-active-high" )) + data->enable_gpio_flags =GPIOF_OUT_INIT_HIGH ; +else + data->enable_gpio_flags =GPIOF_OUT_INIT_LOW ;Quoted 3 lines
drivers/video/backlight/pwm_bl.c static int pwm_backlight_probe(struct
Quoted 3 lines
drivers/video/backlight/pwm_bl.c +0
-0return 0 ; }@@ -207,12 +215,23 @@ static int pwm_backlight_probe(struct
platform_device*pdev)Quoted 11 lines
Quoted 11 lines
drivers/video/backlight/pwm_bl.c +2
-0}else max = data->max_brightness; + pb->enable_gpio = data->enable_gpio; + pb->enable_gpio_flags = data->enable_gpio_flags; pb->notify = data->notify; pb->notify_after = data->notify_after; pb->check_fb = data->check_fb; pb->exit = data->exit; pb->dev = &pdev->dev;Quoted 13 lines
Quoted 12 lines
drivers/video/backlight/pwm_bl.c +9
-0+if (gpio_is_valid (pb->enable_gpio)) { + ret =gpio_request_one (pb->enable_gpio, +GPIOF_DIR_OUT | pb->enable_gpio_flags,"bl_en" ); +if (ret) { +dev_err (&pdev->dev,"failed to allocate bl_en gpio" ); + goto err_alloc; + } + } + pb->pwm =devm_pwm_get (&pdev->dev,NULL );if (IS_ERR (pb->pwm)) { dev_err(&pdev->dev, "unable to request PWM, trying legacyQuoted 1 line
API\n"); @@ -221,7 +240,7 @@ static int pwm_backlight_probe(struct
platform_device*pdev)Quoted 2 lines
Quoted 2 lines
if (IS_ERR(pb->pwm)) { dev_err(&pdev->dev, "unable to request legacyPWM\n");
Quoted 7 lines
drivers/video/backlight/pwm_bl.c static int pwm_backlight_probe(struct
Quoted 7 lines
ret = PTR_ERR(pb->pwm);drivers/video/backlight/pwm_bl.c +1
-1- goto err_alloc; + goto err_gpio; } }@@ -255,6 +274,9 @@ static int pwm_backlight_probe(struct
platform_device*pdev)Quoted 10 lines
drivers/video/backlight/pwm_bl.c static int pwm_backlight_remove(struct
Quoted 10 lines
drivers/video/backlight/pwm_bl.c +3
-0platform_set_drvdata (pdev, bl);return 0 ; +err_gpio: +if (gpio_is_valid (data->enable_gpio)) +gpio_free (data->enable_gpio); err_alloc:if (data->exit) data->exit (&pdev->dev);@@ -269,6 +291,8 @@ static int pwm_backlight_remove(struct
platform_device*pdev)Quoted 24 lines
Quoted 13 lines
drivers/video/backlight/pwm_bl.c +2
-0backlight_device_unregister (bl);pwm_config (pb->pwm,0 , pb->period);pwm_disable (pb->pwm); +if (gpio_is_valid (pb->enable_gpio)) +gpio_free (pb->enable_gpio);if (pb->exit) pb->exit (&pdev->dev);return 0 ;(quoted diff) +0
-0
b/include/linux/pwm_backlight.hindex 56f4a86..2706805 100644include/linux/pwm_backlight.h line 15
Quoted 8 lines
include/linux/pwm_backlight.h +2
-0@@ -8,6 +8,8 @@struct platform_pwm_backlight_data {int pwm_id; +int enable_gpio; +unsigned int enable_gpio_flags;unsigned int max_brightness;unsigned int dft_brightness;unsigned int lth_brightness;Hi Andrew,
Quoted 8 lines
I'm Cc'ing Alexandre Courbot, who has been working on supporting this in a more generic way using power sequences. Generally this kind of support really belongs in the common display framework, but I guess we could add this one GPIO since it really is related only to the backlight. Usually more than just an enable for the backlight is required so I'm not sure how useful this really is.
Alex, any thought?
It is very common for a GPIO to be involved in powering the backlight on, indeed. However it seems that in this patch the GPIO is set once and for all during probe and never touched afterwards. This means the backlight is still enabled (and consuming power) even when its value is zero - I'd at least like to see the GPIO disabled when this is the case to save power. Otherwise you can achieve the same result with a gpio-regulator defined to be always on in the DT, without touching the pwm-backlight driver.
Another issue is that if the GPIO is not explicitly set to -1 in the platform data, probe will try to acquire GPIO 0 and will fail. This would break compatibility with all existing users of pwm-backlight that rely on platform data.
And there is also the fact that the powering of backlights is often slightly more complicated than just an enabling GPIO - for Ventana we have at least one more regulator/GPIO involved. Maybe this regulator could be turned on forever in the DT - then pwm-backlight could use the enable GPIO to save power, but I suspect we would save even more power if we could turn the regulator off as well.
Quoted 4 lines
But overall I'm not against having enable GPIO support in this driver if this helps getting the job done while I finish proper power-sequence support. Andrew, does this single patch allow you to enable the backlight on some boards?
I did come to the same conclusion regarding the platform data breakage. I'm expecting that the use of platform data will go away, at least on ARM, since we are all aggressively moving what used to be in platform data into the device tree. Do other platforms use this driver?
I remember hearing that there is some work in progress to encapsulate gpios into a struct, rather than passing it around as a bare integer, so when that happens, we can use NULL for no-gpio, which should take care of the platform data issue as well. It's kind of difficult to work around this problem otherwise.
I agree that we should be turning on and off the backlight enable gpio as needed to save power. I just haven't gotten there yet. I can try to modify this patch if that's your preference, or I can follow up with a patch to add this in the very near future.
To answer your last question, yes, this single patch does allow me to enable the backlight on some boards (in particular, the one I'm working on).
Quoted 4 lines · Andrew Chew
I did come to the same conclusion regarding the platform data breakage. I'm expecting that the use of platform data will go away, at least on ARM, since we are all aggressively moving what used to be in platform data into the device tree. Do other platforms use this driver?
I can see at least 29 users of platform_pwm_backlight_data, all ARM with the exception of one unicore32. I guess at least for the foreseeable future platform data will remain.
Quoted 5 lines
I remember hearing that there is some work in progress to encapsulate gpios into a struct, rather than passing it around as a bare integer, so when that happens, we can use NULL for no-gpio, which should take care of the platform data issue as well. It's kind of difficult to work around this problem otherwise.
Yes, actually I am doing the GPIO rework. If you are not too much in a hurry you might want for it to happen (should not be too long now that the core has been reworked). At the same time, GPIO descriptors will also enable the power sequences, so if you wait even longer (or help me with it), this patch might not even be needed at all. Of course if you want to support this *now*, this is still the shortest path.
Quoted 4 lines
I agree that we should be turning on and off the backlight enable gpio as needed to save power. I just haven't gotten there yet. I can try to modify this patch if that's your preference, or I can follow up with a patch to add this in the very near future.
That's ultimately for Thierry to say, but submitting a new revision makes more sense IMHO - it is not a big change and there are other issues to address (uninitialized GPIO in platform data) anyway.
Quoted 2 lines
To answer your last question, yes, this single patch does allow me to enable the backlight on some boards (in particular, the one I'm working on).
Cool - may I ask which one? All the NV boards I tried to far required more complex sequences for their panels.
Alex.
I sent out a new patch that enables/disables the backlight enable gpio.
Quoted 9 lines · Andrew Chew
Quoted 4 lines
I did come to the same conclusion regarding the platform data breakage. I'm expecting that the use of platform data will go away, at least on ARM, since we are all aggressively moving what used to be in platform data into the device tree. Do other platforms use this driver?
I can see at least 29 users of
platform_pwm_backlight_data, all ARM with the exception of one unicore32. I guess at least for the foreseeable future platform data will remain.
I'm not sure how to solve this, then. Any suggestions?
Quoted 6 lines
Quoted 5 lines
I remember hearing that there is some work in progress to encapsulate gpios into a struct, rather than passing it around as a bare integer, so when that happens, we can use NULL for no-gpio, which should take care of the platform data issue as well. It's kind of difficult to work around this problem otherwise.
Quoted 6 lines
Yes, actually I am doing the GPIO rework. If you are not too much in a hurry you might want for it to happen (should not be too long now that the core has been reworked). At the same time, GPIO descriptors will also enable the power sequences, so if you wait even longer (or help me with it), this patch might not even be needed at all. Of course if you want to support this *now*, this is still the shortest path.
Sadly, I do need this now, and I'd rather do it as cleanly as possible rather than maintaining a hack. The project I am working on is very pedantic.
Quoted 8 lines
Quoted 4 lines
I agree that we should be turning on and off the backlight enable gpio as needed to save power. I just haven't gotten there yet. I can try to modify this patch if that's your preference, or I can follow up with a patch to add this in the very near future.
That's ultimately for Thierry to say, but submitting a new revision makes more sense IMHO - it is not a big change and there are other issues to address (uninitialized GPIO in platform data) anyway.
Done.
Quoted 6 lines
Quoted 2 lines
To answer your last question, yes, this single patch does allow me to enable the backlight on some boards (in particular, the one I'm working
on).
Cool - may I ask which one? All the NV boards I tried to far required more complex sequences for their panels.
This is for t114-dalmore. There may be other gpios that are needed that I'm not aware of off the top of my head. For the backlight itself, this seems to be the only one.
Quoted 8 lines · Andrew Chew
I sent out a new patch that enables/disables the backlight enable gpio.
Quoted 6 lines · Andrew Chew
Quoted 4 lines
I did come to the same conclusion regarding the platform data breakage. I'm expecting that the use of platform data will go away, at least on ARM, since we are all aggressively moving what used to be in platform data into the device tree. Do other platforms use this driver?
Quoted 5 lines
Quoted 3 lines
I can see at least 29 users of
platform_pwm_backlight_data, all ARM with the exception of one unicore32. I guess at least for the foreseeable future platform data will remain.I'm not sure how to solve this, then. Any suggestions?
In one of my (many) attempts to add power sequencing to pwm-backlight, I just added a boolean to the platform data that must be explicitly set in order to enable control by GPIO. I.e.
bool use_enable_gpio int enable_gpio; unsigned int enable_gpio_flags;
enable_gpio and enable_gpio_flags would then only be considered if use_enable_gpio is true. Granted, it's not the best solution here but that's the only way to handle this correctly with integer GPIOS, and it does not pollute the DT anyway (use_enable_gpio will only be set by pwm_backlight_parse_dt() if of_get_named_gpio() returned a valid GPIO. Btw, you also want to check if the enable-gpio property exists first because otherwise probe() will fails if no GPIO is specified).
Quoted 9 lines
Quoted 6 lines
Yes, actually I am doing the GPIO rework. If you are not too much in a hurry you might want for it to happen (should not be too long now that the core has been reworked). At the same time, GPIO descriptors will also enable the power sequences, so if you wait even longer (or help me with it), this patch might not even be needed at all. Of course if you want to support this *now*, this is still the shortest path.
Sadly, I do need this now, and I'd rather do it as cleanly as possible rather than maintaining a hack. The project I am working on is very pedantic.
Well, if you can get this right and make the GPIO optional, I think this is a reasonable feature to have in pwm-backlight, until a more generic powerseq-backlight driver takes over. ;)
Quoted 10 lines
Quoted 6 lines
Quoted 2 lines
To answer your last question, yes, this single patch does allow me to enable the backlight on some boards (in particular, the one I'm working
on).
Cool - may I ask which one? All the NV boards I tried to far required more complex sequences for their panels.
This is for t114-dalmore. There may be other gpios that are needed that I'm not aware of off the top of my head. For the backlight itself, this seems to be the only one.
I don't know the details of Dalmore but would think there must also be at least one regulator involved. If it is set to be always on in the DT, then your solution of using an enable GPIO should work then, even if not necessarily optimal wrt power usage.
Alex.
Quoted 2 lines · Alex Courbot
Btw, you also want to check if the enable-gpio property exists first because otherwise
probe()will fails if no GPIO is specified).
That's actually not true - I overlooked the fact that probe() checks for the GPIO validity before requesting it. My bad.
Alex.
Quoted 23 lines · Alex Courbot
Quoted 8 lines · Andrew Chew
I sent out a new patch that enables/disables the backlight enable gpio.
Quoted 6 lines · Andrew Chew
Quoted 4 lines
I did come to the same conclusion regarding the platform data breakage. I'm expecting that the use of platform data will go away, at least on ARM, since we are all aggressively moving what used to be in platform data into the device tree. Do other platforms use this driver?
Quoted 5 lines
Quoted 3 lines
I can see at least 29 users of
platform_pwm_backlight_data, all ARM with the exception of one unicore32. I guess at least for the foreseeable future platform data will remain.I'm not sure how to solve this, then. Any suggestions?
In one of my (many) attempts to add power sequencing to pwm-backlight, I just added a boolean to the platform data that must be explicitly set in order to enable control by GPIO. I.e.
bool use_enable_gpio int enable_gpio; unsigned int enable_gpio_flags;
Quoted 8 lines
enable_gpioandenable_gpio_flagswould then only be considered ifuse_enable_gpiois true. Granted, it's not the best solution here but that's the only way to handle this correctly with integer GPIOS, and it does not pollute the DT anyway (use_enable_gpiowill only be set bypwm_backlight_parse_dt()ifof_get_named_gpio()returned a valid GPIO. Btw, you also want to check if the enable-gpio property exists first because otherwiseprobe()will fails if no GPIO is specified).
There are two more options that I can see. The first involves updating all users to initialize the GPIO to -1 in the platform data, which will remove the requirement for an extra flag field. Another option would be to make this feature DT only, so that the GPIO can always be assumed to be -1 for non-DT and DT without an enable-gpio property.
There's a third alternative, namely using a regulator for this, which has better lookup support and therefore should be easier to make optional. And as Alex mentioned it already has support for always-on functionality and such.
Thierry