From: Grygorii Strashko <[email protected]>
Setting a dev_pm_ops suspend/resume pair of callbacks but not a set of
hibernation callbacks means those pm functions will not be
called upon hibernation - that leads to system crash on ARM during
freezing if gpio-led is used in combination with CPU led trigger.
It may happen after freeze_noirq stage (GPIO is suspended)
and before syscore_suspend stage (CPU led trigger is suspended)
- usually when disable_nonboot_cpus() is called.
Log:
PM: noirq freeze of devices complete after 1.425 msecs
Disabling non-boot CPUs ...
^ system may crash or stuck here with message (TI AM572x)
WARNING: CPU: 0 PID: 3100 at drivers/bus/omap_l3_noc.c:148 l3_interrupt_handler+0x22c/0x370()
44000000.ocp:L3 Custom Error: MASTER MPU TARGET L4_PER1_P3 (Idle): Data Access in Supervisor mode during Functional access
CPU1: shutdown
^ or here
Fix this by using SIMPLE_DEV_PM_OPS, which appropriately
assigns the suspend and hibernation callbacks and move
led_suspend/led_resume under CONFIG_PM_SLEEP to avoid
build warnings.
Signed-off-by: Grygorii Strashko <[email protected]>
---
This patch actually fixes commit:
73e1ab4 "leds: Convert led class driver from legacy pm ops to dev_pm_ops"
Changes in v2:
- fixed compile time warnings when !CONFIG_PM_SLEEP
- updated commit message
v1:
- https://lkml.org/lkml/2015/2/2/476
drivers/leds/led-class.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/drivers/leds/led-class.c b/drivers/leds/led-class.c
index dbeebac..5033e0d 100644
--- a/drivers/leds/led-class.c
+++ b/drivers/leds/led-class.c
@@ -183,6 +183,7 @@ void led_classdev_resume(struct led_classdev *led_cdev)
}
EXPORT_SYMBOL_GPL(led_classdev_resume);
+#ifdef CONFIG_PM_SLEEP
static int led_suspend(struct device *dev)
{
struct led_classdev *led_cdev = dev_get_drvdata(dev);
@@ -202,11 +203,9 @@ static int led_resume(struct device *dev)
return 0;
}
+#endif
-static const struct dev_pm_ops leds_class_dev_pm_ops = {
- .suspend = led_suspend,
- .resume = led_resume,
-};
+static SIMPLE_DEV_PM_OPS(leds_class_dev_pm_ops, led_suspend, led_resume);
/**
* led_classdev_register - register a new object of led_classdev class.
--
1.9.1
Hi Bryan, all,
On 02/05/2015 04:37 PM, [email protected] wrote:
> From: Grygorii Strashko <[email protected]>
>
> Setting a dev_pm_ops suspend/resume pair of callbacks but not a set of
> hibernation callbacks means those pm functions will not be
> called upon hibernation - that leads to system crash on ARM during
> freezing if gpio-led is used in combination with CPU led trigger.
> It may happen after freeze_noirq stage (GPIO is suspended)
> and before syscore_suspend stage (CPU led trigger is suspended)
> - usually when disable_nonboot_cpus() is called.
>
> Log:
> PM: noirq freeze of devices complete after 1.425 msecs
> Disabling non-boot CPUs ...
> ^ system may crash or stuck here with message (TI AM572x)
>
> WARNING: CPU: 0 PID: 3100 at drivers/bus/omap_l3_noc.c:148 l3_interrupt_handler+0x22c/0x370()
> 44000000.ocp:L3 Custom Error: MASTER MPU TARGET L4_PER1_P3 (Idle): Data Access in Supervisor mode during Functional access
>
> CPU1: shutdown
> ^ or here
>
> Fix this by using SIMPLE_DEV_PM_OPS, which appropriately
> assigns the suspend and hibernation callbacks and move
> led_suspend/led_resume under CONFIG_PM_SLEEP to avoid
> build warnings.
>
> Signed-off-by: Grygorii Strashko <[email protected]>
> ---
> This patch actually fixes commit:
> 73e1ab4 "leds: Convert led class driver from legacy pm ops to dev_pm_ops"
>
> Changes in v2:
> - fixed compile time warnings when !CONFIG_PM_SLEEP
> - updated commit message
>
> v1:
> - https://lkml.org/lkml/2015/2/2/476
>
> drivers/leds/led-class.c | 7 +++----
> 1 file changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/leds/led-class.c b/drivers/leds/led-class.c
> index dbeebac..5033e0d 100644
> --- a/drivers/leds/led-class.c
> +++ b/drivers/leds/led-class.c
> @@ -183,6 +183,7 @@ void led_classdev_resume(struct led_classdev *led_cdev)
> }
> EXPORT_SYMBOL_GPL(led_classdev_resume);
>
> +#ifdef CONFIG_PM_SLEEP
> static int led_suspend(struct device *dev)
> {
> struct led_classdev *led_cdev = dev_get_drvdata(dev);
> @@ -202,11 +203,9 @@ static int led_resume(struct device *dev)
>
> return 0;
> }
> +#endif
>
> -static const struct dev_pm_ops leds_class_dev_pm_ops = {
> - .suspend = led_suspend,
> - .resume = led_resume,
> -};
> +static SIMPLE_DEV_PM_OPS(leds_class_dev_pm_ops, led_suspend, led_resume);
>
> /**
> * led_classdev_register - register a new object of led_classdev class.
>
Could you pay attention on this patch, pls? It fixes a real issue for me.
Actually, I'd like to note that it fixes regression introduced by commit
73e1ab41a80d36207e8ea7cc2c9897afdeeef387
"leds: Convert led class driver from legacy pm ops to dev_pm_ops" from
Shuah Khan <[email protected]> (Jun 20 2013).
--
regards,
-grygorii