Re: Switching off LED triggers udev

[Jacek Anaszewski <jacek.anaszewski@gmail.com>]

Hi Juergen.

On 04/10/2017 04:29 PM, Juergen Schindele wrote:
> Hello List,
> i hope i found the right addres for this small problem.
> I have a userspace daemon which controls leds via sysfs API
> "/sys/class/leds/Soft1/brightness". Each time i turn the LED off
> by writing 0 to brightness the /proc/sys/kernel/hotplug deamon 
> is called with "ACTION:change SUBSYSTEM:leds DEVNAME:(null) 
> DEVPATH:/devices/...."
> 
> from led-class.c ----------------------------
>         if (state == LED_OFF)
>                 led_trigger_remove(led_cdev);
>         __led_set_brightness(led_cdev, state);
> 
> But in fact there is no change of trigger because "none" was setup between 
> enabling LED and disabling it. So from my point of view there is no change
> from "none" to "remove" which justifies a "kobject_uevent_env(&led_cdev->dev-
>> kobj, KOBJ_CHANGE, envp)".
> 
> To solve this problem of unusefull waste of CPU-time i suggest the following 
> patch. Please consider the implementation which saved a lot of valuable CPU-
> power in my case:
> 
> Index: drivers/leds/led-triggers.c
> ===================================================================
> --- drivers/leds/led-triggers.c (Revision 14968)
> +++ drivers/leds/led-triggers.c (Arbeitskopie)
> @@ -106,10 +106,11 @@
>         const char *name;
>  
>         name = trig ? trig->name : "none";
> -       event = kasprintf(GFP_KERNEL, "TRIGGER=%s", name);
>  
>         /* Remove any existing trigger */
>         if (led_cdev->trigger) {
> +               event = kasprintf(GFP_KERNEL, "TRIGGER=%s", name);
> +
>                 write_lock_irqsave(&led_cdev->trigger->leddev_list_lock, 
> flags);
>                 list_del(&led_cdev->trig_list);
>                 write_unlock_irqrestore(&led_cdev->trigger->leddev_list_lock,
> @@ -122,6 +123,8 @@
>                 led_set_brightness(led_cdev, LED_OFF);
>         }
>         if (trig) {
> +               event = kasprintf(GFP_KERNEL, "TRIGGER=%s", name);
> +
>                 write_lock_irqsave(&trig->leddev_list_lock, flags);
>                 list_add_tail(&led_cdev->trig_list, &trig->led_cdevs);
>                 write_unlock_irqrestore(&trig->leddev_list_lock, flags);
> 
> Thank you for your attention

In 4.9 we have the following patch that should fix your problem,
please let us know if it works for you:

commit fbfa197afddd13f9bdca1c822f5d5730b50639eb
Author: Jacek Anaszewski <jacek.anaszewski@gmail.com>
Date:   Sun Sep 18 20:24:29 2016 +0200

    leds: triggers: Return from led_trigger_set() if there is nothing to do

    If led_trigger_set() is called with "trig" argument set to NULL, and
there
    is no trigger to remove then the function should return immediately so
    as to avoid doing unnecessary allocation and sending uevent.

    Signed-off-by: Jacek Anaszewski <j.anaszewski@samsung.com>
    Reported-by: Daniel Romell <daro@hms.se>
    Acked-by Daniel Romell <daro@hms.se>

diff --git a/drivers/leds/led-triggers.c b/drivers/leds/led-triggers.c
index 59d6a3e..c7a38d4 100644
--- a/drivers/leds/led-triggers.c
+++ b/drivers/leds/led-triggers.c
@@ -110,6 +110,9 @@ void led_trigger_set(struct led_classdev *led_cdev,
struct led_trigger *trig)
        char *envp[2];
        const char *name;

+       if (!led_cdev->trigger && !trig)
+               return;
+
        name = trig ? trig->name : "none";
        event = kasprintf(GFP_KERNEL, "TRIGGER=%s", name);


-- 
Best regards,
Jacek Anaszewski