Re: swapper: page allocation failure. order:3, mode:0x20

[Andrew Morton <akpm@osdl.org>]

Anton Blanchard <anton@samba.org> wrote:
>
> 
> > So sure, ratelimit it, make it KERN_INFO and maybe add a dump_stack()?
> 
> Sounds good, I always end up adding a dump_stack there when debugging
> these problems anyway.
> 
> > (printk_ratelimit() may be a suitable name)
> 
> How does this look?

Good.  I guess we need to make net_ratelimit() use this sometime.

> +/* 
> + * printk rate limiting, lifted from the networking subsystem.
> + *
> + * This enforces a rate limit: not more than one kernel message
> + * every printk_ratelimit_jiffies to make a denial-of-service 
> + * attack impossible.
> + */ 
> +int printk_ratelimit(void)
> +{
> +	static spinlock_t ratelimit_lock = SPIN_LOCK_UNLOCKED;
> +	static unsigned long toks = 10*5*HZ;
> +	static unsigned long last_msg; 
> +	static int missed;
> +	unsigned long flags;
> +	unsigned long now = jiffies;
> +
> +	spin_lock_irqsave(&ratelimit_lock, flags);
> +	toks += now - last_msg;
> +	last_msg = now;
> +	if (toks > (printk_ratelimit_burst * printk_ratelimit_jiffies))
> +		toks = printk_ratelimit_burst * printk_ratelimit_jiffies;
> +	if (toks >= printk_ratelimit_jiffies) {
> +		int lost = missed;
> +		missed = 0;
> +		toks -= printk_ratelimit_jiffies;
> +		spin_unlock_irqrestore(&ratelimit_lock, flags);
> +		if (lost)
> +			printk(KERN_WARNING "printk: %d messages suppressed.\n", lost);
> +		return 1;
> +	}
> +	missed++;
> +	spin_unlock_irqrestore(&ratelimit_lock, flags);
> +	return 0;
> +}

This seems a bit odd.  It means that the further apart the message bursts
are, the longer they are allowed to be.  Or something.

Wouldn't it be better to say "after each greater-than-five second window,
allow up to ten printk's as long as they happen in the next five
milliseconds"?