From mboxrd@z Thu Jan 1 00:00:00 1970 From: dromede@gmail.com (=?UTF-8?B?TWFya28gS2F0acSH?=) Date: Tue, 4 Dec 2012 23:28:24 +0100 Subject: [PATCH] backlight: corgi_lcd: Use gpio_set_value_cansleep() for akita machines to avoid WARN_ON trigger In-Reply-To: <004301cdcec2$472385a0$d56a90e0$%han@samsung.com> References: <1354237577-4653-1-git-send-email-dromede@gmail.com> <004301cdcec2$472385a0$d56a90e0$%han@samsung.com> Message-ID: To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Fri, Nov 30, 2012 at 7:16 AM, Jingoo Han wrote: > On Friday, November 30, 2012 10:06 AM, Marko Katic wrote >> >> From: Marko Katic >> >> Changing backlight intensity on an Akita (Sharp Zaurus C-1000) >> will always trigger a WARN_ON: >> >> WARNING: at drivers/gpio/gpiolib.c:1672 __gpio_set_value+0x38/0xa4() >> Modules linked in: >> Backtrace: >> [] (dump_backtrace+0x0/0x110) from [] (dump_stack+0x18/0x1c) >> r6:c0158fc8 r5:00000009 r4:00000000 r3:c03d4f70 >> [] (dump_stack+0x0/0x1c) from [] (warn_slowpath_common+0x54/0x6c) >> [] (warn_slowpath_common+0x0/0x6c) from [] (warn_slowpath_null+0x24/0x2c) >> r8:c38d5c00 r7:c03f82c0 r6:00000000 r5:000000d0 r4:c384e4fc >> r3:00000009 >> [] (warn_slowpath_null+0x0/0x2c) from [] (__gpio_set_value+0x38/0xa4) >> [] (__gpio_set_value+0x0/0xa4) from [] (corgi_bl_set_intensity+0x44/0x74) >> r7:c3933418 r6:c3933400 r5:c392cdf0 r4:0000002f >> [] (corgi_bl_set_intensity+0x0/0x74) from [] (corgi_bl_update_status+0x5c/0x64) >> r5:c03d31f0 r4:c3933400 >> [] (corgi_bl_update_status+0x0/0x64) from [] (corgi_lcd_probe+0x1a8/0x258) >> r4:c392cdf0 r3:c0169bc0 >> [] (corgi_lcd_probe+0x0/0x258) from [] (spi_drv_probe+0x20/0x24) >> r8:00000052 r7:c0192d9c r6:c03da6e8 r5:c38d5c34 r4:c38d5c00 >> [] (spi_drv_probe+0x0/0x24) from [] (driver_probe_device+0xb0/0x208) >> [] (driver_probe_device+0x0/0x208) from [] (__driver_attach+0x70/0x94) >> r6:c03da6e8 r5:c38d5c34 r4:c38d5c00 r3:00000000 >> [] (__driver_attach+0x0/0x94) from [] (bus_for_each_dev+0x54/0x90) >> r6:c03da6e8 r5:c3827e80 r4:00000000 r3:00000000 >> [] (bus_for_each_dev+0x0/0x90) from [] (driver_attach+0x20/0x28) >> r7:00000000 r6:c03e29ec r5:c3932980 r4:c03da6e8 >> [] (driver_attach+0x0/0x28) from [] (bus_add_driver+0xd4/0x22c) >> [] (bus_add_driver+0x0/0x22c) from [] (driver_register+0xa4/0x134) >> r8:00000052 r7:c03ea900 r6:c03c32ac r5:c03bdfc8 r4:c03da6e8 >> [] (driver_register+0x0/0x134) from [] (spi_register_driver+0x4c/0x60) >> [] (spi_register_driver+0x0/0x60) from [] (corgi_lcd_driver_init+0x14/0x1c) >> [] (corgi_lcd_driver_init+0x0/0x1c) from [] (do_one_initcall+0x9c/0x174) >> [] (do_one_initcall+0x0/0x174) from [] (kernel_init+0xf4/0x2a8) >> [] (kernel_init+0x0/0x2a8) from [] (ret_from_fork+0x14/0x24) >> ---[ end trace a863a63f242ee38c ]--- >> >> Akita machines have backlight controls hooked to a gpio expander chip, max7310. >> The driver that controls the max7310 sets the cansleep flag and the corgi_lcd >> driver only uses plain gpio_set_value calls for changing backlight controls. >> This triggers the WARN_ON on akita machines. >> >> Akita is the only exception in this case since other users of corgi_bl access >> backlight gpio controls through a different gpio expander which does not set the cansleep flag. > > How about modifying the commit message as below? > > Akita machines have backlight controls hooked to a gpio expander chip, > max7310. In this case, pca953x_gpio_set_value() can be called to control > gpio, and pca953x_setup_gpio() sets can_sleep flag. Therefore, > gpio_set_value_cansleep() should be used in order to avoid WARN_ON on > akita machines. > Ok, i'll be more specific in my v2 commit message. > Akita is the only exception in this case since other users of corgi_lcd > access backlight gpio controls through a different gpio expander which > does not set the can_sleep flag. > >> >> Fix this by conditionally using gpio_set_value_cansleep() when changing backlight intensity on >> akita machines. > > I don't think this sentence is necessary. Fine by me. > >> >> Signed-off-by: Marko Katic >> --- >> drivers/video/backlight/corgi_lcd.c | 18 +++++++++++++----- >> 1 file changed, 13 insertions(+), 5 deletions(-) >> >> diff --git a/drivers/video/backlight/corgi_lcd.c b/drivers/video/backlight/corgi_lcd.c >> index c781768..f33e26f 100644 >> --- a/drivers/video/backlight/corgi_lcd.c >> +++ b/drivers/video/backlight/corgi_lcd.c >> @@ -26,7 +26,7 @@ >> #include >> #include >> #include >> - >> +#include >> #define POWER_IS_ON(pwr) ((pwr) <= FB_BLANK_NORMAL) >> >> /* Register Addresses */ >> @@ -408,11 +408,19 @@ static int corgi_bl_set_intensity(struct corgi_lcd *lcd, int intensity) >> /* Bit 5 via GPIO_BACKLIGHT_CONT */ >> cont = !!(intensity & 0x20) ^ lcd->gpio_backlight_cont_inverted; >> >> - if (gpio_is_valid(lcd->gpio_backlight_cont)) >> - gpio_set_value(lcd->gpio_backlight_cont, cont); >> + if (gpio_is_valid(lcd->gpio_backlight_cont)) { >> + if (machine_is_akita()) >> + gpio_set_value_cansleep(lcd->gpio_backlight_cont, cont); >> + else >> + gpio_set_value(lcd->gpio_backlight_cont, cont); >> + } > > Please don't use machine-specific code such as machine_is_akita(). > Using machine_is_xxx() in drivers is not preferable. > gpio_cansleep() would be better. > > However, there seems to be no difference between gpio_set_value_cansleep() > and gpio_set_value(), except for WARN_ON(chip->can_sleep); > So, if there is no side effect at other machines, please use only > gpio_set_value_cansleep(). > I'm not in favor of this approach. Differences between these functions might be small now but that might change sometime in the future. I thinks akpm's solution is good. > > Best regards, > Jingoo Han > >> >> - if (gpio_is_valid(lcd->gpio_backlight_on)) >> - gpio_set_value(lcd->gpio_backlight_on, intensity); >> + if (gpio_is_valid(lcd->gpio_backlight_on)) { >> + if (machine_is_akita()) >> + gpio_set_value_cansleep(lcd->gpio_backlight_on, intensity); >> + else >> + gpio_set_value(lcd->gpio_backlight_on, intensity); >> + } >> >> if (lcd->kick_battery) >> lcd->kick_battery(); >> -- >> 1.7.10.4 >