| Message ID | 1394695879-22845-1-git-send-email-shc_work@mail.ru (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 13/03/14 09:31, Alexander Shiyan wrote: > Provide dummy functions for LCD register()/unregister() if > CONFIG_LCD_CLASS_DEVICE is not set. Hmm, why do you want to do that? If a driver needs the LCD class, it should depend on or select it. Tomi
???????, 13 ????? 2014, 10:23 +02:00 ?? Tomi Valkeinen <tomi.valkeinen@ti.com>: > On 13/03/14 09:31, Alexander Shiyan wrote: > > Provide dummy functions for LCD register()/unregister() if > > CONFIG_LCD_CLASS_DEVICE is not set. > > Hmm, why do you want to do that? If a driver needs the LCD class, it > should depend on or select it. I inspect my recent changes for the imxfb driver. I use the LCD class for power management and contrast, nevertheless, if it lack in the kernel this leads to an error. Yes, we can choose the LCD_CLASS_DEVICE symbol for the imxfb driver, but at the same time we must choose BACKLIGHT_LCD_SUPPORT. I do not think it's a good way. In any case, I would like to revise the patch to use NULL as a result if there is no support LCD_CLASS_DEVICE in the kernel. Additionally, I have a plan to convert "menuconfig" entry for BACKLIGHT_LCD_SUPPORT to the "menu". OK? ---
On 13/03/14 10:48, Alexander Shiyan wrote: > ???????, 13 ????? 2014, 10:23 +02:00 ?? Tomi Valkeinen <tomi.valkeinen@ti.com>: >> On 13/03/14 09:31, Alexander Shiyan wrote: >>> Provide dummy functions for LCD register()/unregister() if >>> CONFIG_LCD_CLASS_DEVICE is not set. >> >> Hmm, why do you want to do that? If a driver needs the LCD class, it >> should depend on or select it. > > I inspect my recent changes for the imxfb driver. > I use the LCD class for power management and contrast, nevertheless, > if it lack in the kernel this leads to an error. So is the CONFIG_LCD_CLASS_DEVICE optional for imxfb? It works fine with or without the LCD class support? Is there some reason to run it without LCD class support? > Yes, we can choose the LCD_CLASS_DEVICE symbol for the imxfb driver, > but at the same time we must choose BACKLIGHT_LCD_SUPPORT. > I do not think it's a good way. Why not? > In any case, I would like to revise the patch to use NULL as a result > if there is no support LCD_CLASS_DEVICE in the kernel. Why do you want to run imxfb without LCD_CLASS_DEVICE? Isn't it simpler to just depend on it? > Additionally, I have a plan to convert "menuconfig" entry for > BACKLIGHT_LCD_SUPPORT to the "menu". Hmm. That does make sense, as I don't see BACKLIGHT_LCD_SUPPORT affecting anything, except enabling the BL & LCD Kconfig menu. However, many of the items in BL & LCD menu have, for some reason, "default y" or "default m". So if you make BACKLIGHT_LCD_SUPPORT a menu, it means all those subitems are always enabled. So there definitely will be a side effect to that change. I guess there's legacy reasons why many of the items are enabled by default. It would make sense to me to have BACKLIGHT_LCD_SUPPORT as a menu, as you suggest, and having all the subitems disabled by default. But again, that would change the current behavior, which may or may not cause issues. Tomi
diff --git a/include/linux/lcd.h b/include/linux/lcd.h index 504f624..f1c94fd 100644 --- a/include/linux/lcd.h +++ b/include/linux/lcd.h @@ -110,14 +110,37 @@ static inline void lcd_set_power(struct lcd_device *ld, int power) mutex_unlock(&ld->update_lock); } -extern struct lcd_device *lcd_device_register(const char *name, - struct device *parent, void *devdata, struct lcd_ops *ops); -extern struct lcd_device *devm_lcd_device_register(struct device *dev, - const char *name, struct device *parent, +#if defined(CONFIG_LCD_CLASS_DEVICE) || defined(CONFIG_LCD_CLASS_DEVICE_MODULE) +struct lcd_device *lcd_device_register(const char *name, struct device *parent, void *devdata, struct lcd_ops *ops); -extern void lcd_device_unregister(struct lcd_device *ld); -extern void devm_lcd_device_unregister(struct device *dev, - struct lcd_device *ld); +struct lcd_device *devm_lcd_device_register(struct device *dev, + const char *name, struct device *parent, void *devdata, + struct lcd_ops *ops); +void lcd_device_unregister(struct lcd_device *ld); +void devm_lcd_device_unregister(struct device *dev, struct lcd_device *ld); +#else +static inline struct lcd_device *lcd_device_register(const char *name, + struct device *parent, void *devdata, struct lcd_ops *ops) +{ + return ERR_PTR(-ENOSYS); +} + +static inline struct lcd_device *devm_lcd_device_register(struct device *dev, + const char *name, struct device *parent, void *devdata, + struct lcd_ops *ops) +{ + return ERR_PTR(-ENOSYS); +} + +static inline void lcd_device_unregister(struct lcd_device *ld) +{ +} + +static inline void devm_lcd_device_unregister(struct device *dev, + struct lcd_device *ld) +{ +} +#endif #define to_lcd_device(obj) container_of(obj, struct lcd_device, dev)
Provide dummy functions for LCD register()/unregister() if CONFIG_LCD_CLASS_DEVICE is not set. Signed-off-by: Alexander Shiyan <shc_work@mail.ru> --- include/linux/lcd.h | 37 ++++++++++++++++++++++++++++++------- 1 file changed, 30 insertions(+), 7 deletions(-)