linux-kernel - Re: [PATCH] leds: make led_blink

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAK5ve-Kc8mjdbN0Nn-pGVjp2pj6ZV_FCBpeOtO5N92CkJZX3=g@mail.gmail.com>
Date:	Fri, 22 Aug 2014 17:21:30 -0700
From:	Bryan Wu <cooloney@...il.com>
To:	Hugh Dickins <hughd@...gle.com>, Tejun Heo <tj@...nel.org>
Cc:	Vincent Donnefort <vdonnefort@...il.com>, Valdis.Kletnieks@...edu,
	Linux LED Subsystem <linux-leds@...r.kernel.org>,
	lkml <linux-kernel@...r.kernel.org>,
	linux-wireless@...r.kernel.org
Subject: Re: [PATCH] leds: make led_blink_set IRQ safe

On Tue, Aug 19, 2014 at 6:51 PM, Hugh Dickins <hughd@...gle.com> wrote:
> On Tue, 19 Aug 2014, Vincent Donnefort wrote:
>
>> This patch introduces a work which take care of reseting the blink workqueue and
>> avoid calling the cancel_delayed_work_sync function which may sleep, from an IRQ
>> context.
>>

Vincent, I'm just wandering can we use cancel_delayed_work() instead
of sync version here.
cancel_delayed_work() can be called from IRQ context.

>> Signed-off-by: Vincent Donnefort <vdonnefort@...il.com>
>
> Thanks.  It does work for me.  Though the problem was more general than
> stated above: not just a problem in IRQ context, but in any atomic context.
>
> I don't suppose it has any effect on Valdis's lockdep issue, which I
> didn't get to see myself; but we should let Valdis report back on that.
>
> May I (most ungratefully!) say that your patch doesn't fill me with
> confidence that it's the right solution: adding yet another work_struct
> to get around the issue seemed dubious to me, I wonder if it might expose
> new races.
>

I agree with Hugh about this new cancel work_struct. But if we revert
it back, I saw led_blink_set() will call del_timer_sync() which might
also sleep and can't be used in IRQ context. Looks like we can't call
led_blink_set() in any IRQ/atomic context.

> But rest assured that I know nothing about this, and I'm not at all
> qualified to review your patch: I hope Bryan and others will do so.
>

Let me invite Tejun to give some advice on how to solve this problem.

Tejun, Vincent's commit 8b37e1bef5a6b60e949e28a4db3006e4b00bd758
convert a timer into work_struct, but Hugh found it will cause
sleeping BUGs [1]. Could you give some opinion about that?

Thanks,
-Bryan

[1]: https://lkml.org/lkml/2014/8/16/128


>
>>
>> diff --git a/drivers/leds/led-class.c b/drivers/leds/led-class.c
>> index 129729d..0971554 100644
>> --- a/drivers/leds/led-class.c
>> +++ b/drivers/leds/led-class.c
>> @@ -148,6 +148,24 @@ static void led_work_function(struct work_struct *ws)
>>                          msecs_to_jiffies(delay));
>>  }
>>
>> +static void reset_blink_work_delayed(struct work_struct *ws)
>> +{
>> +     struct led_classdev *led_cdev =
>> +             container_of(ws, struct led_classdev, reset_blink_work);
>> +
>> +     cancel_delayed_work_sync(&led_cdev->blink_work);
>> +
>> +     if (!led_cdev->blink_delay_on) {
>> +             __led_set_brightness(led_cdev, LED_OFF);
>> +             return;
>> +     } else if (!led_cdev->blink_delay_off) {
>> +             __led_set_brightness(led_cdev, led_cdev->blink_brightness);
>> +             return;
>> +     }
>> +
>> +     queue_delayed_work(system_wq, &led_cdev->blink_work, 1);
>> +}
>> +
>>  static void set_brightness_delayed(struct work_struct *ws)
>>  {
>>       struct led_classdev *led_cdev =
>> @@ -234,6 +252,7 @@ int led_classdev_register(struct device *parent, struct led_classdev *led_cdev)
>>       INIT_WORK(&led_cdev->set_brightness_work, set_brightness_delayed);
>>
>>       INIT_DELAYED_WORK(&led_cdev->blink_work, led_work_function);
>> +     INIT_WORK(&led_cdev->reset_blink_work, reset_blink_work_delayed);
>>
>>  #ifdef CONFIG_LEDS_TRIGGERS
>>       led_trigger_set_default(led_cdev);
>> diff --git a/drivers/leds/led-core.c b/drivers/leds/led-core.c
>> index 4bb1168..959510a 100644
>> --- a/drivers/leds/led-core.c
>> +++ b/drivers/leds/led-core.c
>> @@ -40,19 +40,7 @@ static void led_set_software_blink(struct led_classdev *led_cdev,
>>       led_cdev->blink_delay_on = delay_on;
>>       led_cdev->blink_delay_off = delay_off;
>>
>> -     /* never on - just set to off */
>> -     if (!delay_on) {
>> -             __led_set_brightness(led_cdev, LED_OFF);
>> -             return;
>> -     }
>> -
>> -     /* never off - just set to brightness */
>> -     if (!delay_off) {
>> -             __led_set_brightness(led_cdev, led_cdev->blink_brightness);
>> -             return;
>> -     }
>> -
>> -     queue_delayed_work(system_wq, &led_cdev->blink_work, 1);
>> +     schedule_work(&led_cdev->reset_blink_work);
>>  }
>>
>>
>> @@ -76,8 +64,6 @@ void led_blink_set(struct led_classdev *led_cdev,
>>                  unsigned long *delay_on,
>>                  unsigned long *delay_off)
>>  {
>> -     cancel_delayed_work_sync(&led_cdev->blink_work);
>> -
>>       led_cdev->flags &= ~LED_BLINK_ONESHOT;
>>       led_cdev->flags &= ~LED_BLINK_ONESHOT_STOP;
>>
>> diff --git a/include/linux/leds.h b/include/linux/leds.h
>> index 6a599dc..6e5523d 100644
>> --- a/include/linux/leds.h
>> +++ b/include/linux/leds.h
>> @@ -69,6 +69,7 @@ struct led_classdev {
>>
>>       unsigned long            blink_delay_on, blink_delay_off;
>>       struct delayed_work      blink_work;
>> +     struct work_struct       reset_blink_work;
>>       int                      blink_brightness;
>>
>>       struct work_struct      set_brightness_work;
>> --
>> 1.9.1
>>
>>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/