Re: Why do we still have 32 bit counters? Interrupt counters overflow within 50 days

[Thomas Gleixner <tglx@linutronix.de>]

On Sun, 5 Oct 2014, Christoph Lameter wrote:
> -DEFINE_PER_CPU(unsigned int, irq_count) __visible = -1;
> +DEFINE_PER_CPU(u64, irq_count) __visible = -1;

What's the point of this? irq_count is solely used to figure out
whether interrupts are nested, i.e. whether we need to switch the
stack on x86_64 or not. See the usage sites in arch/x86/kernel/entry_64.S
 
> @@ -51,7 +51,7 @@ struct irq_desc {
>  	unsigned int		core_internal_state__do_not_mess_with_it;
>  	unsigned int		depth;		/* nested irq disables */
>  	unsigned int		wake_depth;	/* nested wake enables */
> -	unsigned int		irq_count;	/* For detecting broken IRQs */
> +	u64			irq_count;	/* For detecting broken IRQs */

This is pointless as this count is solely used for the spurious
detector and reset to 0 when irq_count reaches 100000. See
note_interrupt().

> -extern unsigned int kstat_irqs_cpu(unsigned int irq, int cpu);
> +extern u64 kstat_irqs_cpu(unsigned int irq, int cpu);

Care to fixup the other call sites of this as well? git grep is your friend.

Thanks,

	tglx