Re: [patch 13/13] irqchip: spear_shirq: Simplify register access code

[Viresh Kumar <viresh.kumar@linaro.org>]

On 20 June 2014 13:35, Thomas Gleixner <tglx@linutronix.de> wrote:
> On Fri, 20 Jun 2014, Viresh Kumar wrote:
>> On Fri, Jun 20, 2014 at 3:04 AM, Thomas Gleixner <tglx@linutronix.de> wrote:
>>
>> > Index: linux/drivers/irqchip/spear-shirq.c
>> > -struct shirq_regs {
>> > -       u32 enb_reg;
>> > -       u32 reset_to_enb;
>>
>> I don't see something similar to 'reset_to_enb' is added again.
>> AFAICT, this field is being used by two blocks:
>>
>> spear300_shirq_ras1: writes 0 to this. i.e. we need to write 0 to
>> mask the interrupt
>>
>> and
>>
>> spear320_shirq_ras3: writes 1 to this. i.e. we need to write 1 to
>> mask the interrupt.
>
> AGAIN: spear320_shirq_ras3 is never instantiated as a chained irq. So
> the chip for these interrupts is never set.

Didn't knew it, yes spear320_shirq_ras3 is never registered.

> And if it ever is, you need a separate irq chip for it and not some
> conditional hackery.

Yeah.

>> And so the new code you have added breaks it for SPEAr300 ?
>
> +static void shirq_irq_mask(struct irq_data *d)
> +{
> +       struct spear_shirq *shirq = irq_data_get_irq_chip_data(d);
> +       u32 val, shift = d->irq - shirq->virq_base + shirq->offset;
> +       u32 __iomem *reg = shirq->base + shirq->regs.enb_reg;
> +
> +       raw_spin_lock(&shirq_lock);
> +       val = readl(reg) & ~(0x1 << shift);
> +       writel(val, reg);
> +       raw_spin_unlock(&shirq_lock);
> +}
>
> That's the mask function for 300 and it clears the bit, right?

Yeah, that looks fine. Misread it.