Re: [PATCH -mm 2/2] PM: Disable usermode helper before hibernation and suspend

[Uli Luckas <u.luckas@road.de>]

On Monday, 18. June 2007, you wrote:
> On Monday, 18 June 2007 12:51, Uli Luckas wrote:
> > On Sunday, 17. June 2007, you wrote:
> > > On Friday, 15 June 2007 23:36, Rafael J. Wysocki wrote:
> > > > On Friday, 15 June 2007 15:08, Uli Luckas wrote:
> > > > > On Monday, 4. June 2007, Rafael J. Wysocki wrote:
> > > > > > From: Rafael J. Wysocki <rjw@sisk.pl>
> > > > > >
> > > > > > Use a hibernation and suspend notifier to disable the user mode
> > > > > > helper before a hibernation/suspend and enable it after the
> > > > > > operation.
> > > > >
> > > > > Hi Rafael,
> > > > > I have a couple of questions, regarding this patch ...
> > > > >
> > > > > 2) how does your patch prevent wait_for_completion(&done) to hang
> > > > > during freezing if usermodehelper_pm_callback is called _after_ the
> > > > > above check?
> > > >
> > > > It doesn't.  Once the helper is running, we can't distinguish it from
> > > > any other user land process.
> > > >
> > > > Still, it narrows the window quite a bit.
> > >
> > > Okay, I think we can help it a bit.  Please tell me what you think of
> > > the following patch (on top of 2.6.22-rc4-mm2).
> >
> > Hi Rafael,
> > Thanks for your work. I haven't found the time to actually test your
> > patch but in general it looks like a valid aproach.
> > I think there is one problem though. You need to have your wait queue
> > woken up when you update (atomic_dec) running_helpers. Otherwise you end
> > up always waiting the full RUNNING_HELPERS_TIMEOUT.
>
> Yeah, right.  I tend to forget about obvious things. :-(
>
Great. As far as I can see, this patch should now fix the situation except 
under heavy load where it would abort the attempted suspend. I can live with 
that and it is a whole lot better than it was before - Thanks.

> Besides, I think that the _pm_callback need not be compiled if CONFIG_PM is
> unset.  
>
Should work but is probably beneath what you intended to do. I'll comment 
further down...

> If you can, please have a look at the next version of the patch 
> (appended).
>
One more topic to keep in mind for the future. None of the callers of 
call_usermode_helper* expect a return value of EBUSY nor do they handle it 
correctly. If auto loading of a module for example fails, a permanent error 
is usually assumed and no retry attempts are made. 

> Greetings,
> Rafael
>
>
> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
> ---
>  kernel/kmod.c |   76
> ++++++++++++++++++++++++++++++++++++++++++++++++++-------- 1 file changed,
> 66 insertions(+), 10 deletions(-)
>
> Index: linux-2.6.22-rc4-mm2/kernel/kmod.c
> ===================================================================
> --- linux-2.6.22-rc4-mm2.orig/kernel/kmod.c
> +++ linux-2.6.22-rc4-mm2/kernel/kmod.c
> @@ -41,14 +41,6 @@ extern int max_threads;
>
>  static struct workqueue_struct *khelper_wq;
>
> -/*
> - * If set, both call_usermodehelper_keys() and call_usermodehelper_pipe()
> exit - * immediately returning -EBUSY.  Used for preventing user land
> processes from - * being created after the user land has been frozen during
> a system-wide - * hibernation or suspend operation.
> - */
> -static int usermodehelper_disabled;
> -
>  #ifdef CONFIG_KMOD
>
>  /*
> @@ -275,6 +267,30 @@ static void __call_usermodehelper(struct
>  	}
>  }
>
> +#ifdef CONFIG_PM
> +/*
> + * If set and if call_usermodehelper_exec() is supposed to wait, it will
> exit + * immediately returning -EBUSY (used for preventing user land
> processes from + * being created after the user land has been frozen during
> a system-wide + * hibernation or suspend operation).
> + */
> +static int usermodehelper_disabled;
> +
> +/* Number of helpers running */
> +static atomic_t running_helpers = ATOMIC_INIT(0);
> +
> +/*
> + * Wait queue head used by usermodehelper_pm_callback() to wait for all
> running + * helpers to finish.
> + */
> +static DECLARE_WAIT_QUEUE_HEAD(running_helpers_waitq);
> +
> +/*
> + * Time to wait for running_helpers to become zero before the setting of
> + * usermodehelper_disabled in usermodehelper_pm_callback() fails
> + */
> +#define RUNNING_HELPERS_TIMEOUT	(5 * HZ)
> +
>  static int usermodehelper_pm_callback(struct notifier_block *nfb,
>  					unsigned long action,
>  					void *ignored)
> @@ -283,7 +299,15 @@ static int usermodehelper_pm_callback(st
>  	case PM_HIBERNATION_PREPARE:
>  	case PM_SUSPEND_PREPARE:
>  		usermodehelper_disabled = 1;
> -		return NOTIFY_OK;
> +		wait_event_timeout(running_helpers_waitq,
> +					atomic_read(&running_helpers) == 0,
> +					RUNNING_HELPERS_TIMEOUT);
> +		if (atomic_read(&running_helpers) == 0) {
> +			return NOTIFY_OK;
> +		} else {
> +			usermodehelper_disabled = 0;
> +			return NOTIFY_BAD;
> +		}
>  	case PM_POST_HIBERNATION:
>  	case PM_POST_SUSPEND:
>  		usermodehelper_disabled = 0;
> @@ -293,6 +317,36 @@ static int usermodehelper_pm_callback(st
>  	return NOTIFY_DONE;
>  }
>
> +static void new_helper(void)
> +{
> +	atomic_inc(&running_helpers);
> +}
> +
> +static void helper_finished(void)
> +{
> +	if (atomic_dec_and_test(&running_helpers))
> +		wake_up(&running_helpers_waitq);
> +}
> +
> +#else /* CONFIG_PM */
> +#define usermodehelper_disabled	0
> +
> +static inline int usermodehelper_pm_callback(struct notifier_block *nfb,
          ^^^^^^
This kind of implies, the compiler might optimize this function away. Further 
down you call "pm_notifier(usermodehelper_pm_callback, 0);", thereby passing 
a pointer to usermodehelper_pm_callback which can't be inlined nor optimized 
away. If instead you had an inline function 
register_usermodehelper_pm_callback which called pm_notifier if CONFIG_PM was 
defined and did nothing otherwise...

> +						unsigned long action,
> +						void *ignored)
> +{
> +	return 0;
> +}
> +
> +static inline void new_helper(void)
> +{
> +}
> +
> +static inline void helper_finished(void)
> +{
> +}
> +#endif /* CONFIG_PM */
> +
>  /**
>   * call_usermodehelper_setup - prepare to call a usermode helper
>   * @path - path to usermode executable
> @@ -397,12 +451,13 @@ int call_usermodehelper_exec(struct subp
>  	DECLARE_COMPLETION_ONSTACK(done);
>  	int retval;
>
> +	new_helper();
>  	if (sub_info->path[0] == '\0') {
>  		retval = 0;
>  		goto out;
>  	}
>
> -	if (!khelper_wq || usermodehelper_disabled) {
> +	if (!khelper_wq || (wait != UMH_NO_WAIT && usermodehelper_disabled)) {
>  		retval = -EBUSY;
>  		goto out;
>  	}
> @@ -418,6 +473,7 @@ int call_usermodehelper_exec(struct subp
>
>    out:
>  	call_usermodehelper_freeinfo(sub_info);
> +	helper_finished();
>  	return retval;
>  }
>  EXPORT_SYMBOL(call_usermodehelper_exec);


-- 

------- ROAD ...the handyPC Company - - -  ) ) )

Uli Luckas
Software Development

ROAD GmbH
Bennigsenstr. 14 | 12159 Berlin | Germany
fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
url: www.road.de

Amtsgericht Charlottenburg: HRB 96688 B
Managing directors: Hans-Peter Constien, Hubertus von Streit