Re: [PATCH 5/6] mfd/ab8500: support of hierachical interrupt

[Samuel Ortiz <sameo@linux.intel.com>]

Hi Linus,

On Tue, Apr 17, 2012 at 09:30:49AM +0200, Linus Walleij wrote:
> +static irqreturn_t ab8500_hierarchical_irq(int irq, void *dev)
> +{
> +	struct ab8500 *ab8500 = dev;
> +	int i;
> +
> +	dev_vdbg(ab8500->dev, "interrupt\n");
> +
> +	/*  Hierarchical interrupt version */
> +	for (i = 0; i < AB8500_IT_LATCHHIER_NUM; i++) {
> +		int status;
> +		u8 hier_val;
> +
> +		status = get_register_interruptible(ab8500, AB8500_INTERRUPT,
> +			AB8500_IT_LATCHHIER1_REG + i, &hier_val);
> +		if (status < 0 || hier_val == 0)
> +			continue;
> +
> +		do {
> +			int latch_bit = __ffs(hier_val);
> +			u8 offset = i * 8 + latch_bit;
> +			u8 latch_val;
> +
> +			/* Fix inconsistent ITFromLatch25 bit mapping... */
> +			if (unlikely(offset == 17))
> +				offset = 24;
> +
> +			status = get_register_interruptible(ab8500,
> +					AB8500_INTERRUPT,
> +					AB8500_IT_LATCH1_REG + offset,
> +					&latch_val);
> +			if (status < 0 || latch_val == 0)
> +				goto discard;
> +
> +			do {
> +				int int_bit = __ffs(latch_val);
> +				int line;
> +				int j;
> +
> +				for (j = 0; j < ab8500->mask_size; j++)
> +					if (ab8500->irq_reg_offset[j] == offset)
> +						break;
> +
> +				if (j >= ab8500->mask_size) {
> +					dev_err(ab8500->dev,
> +							"Register offset 0x%2x not declared\n",
> +							offset);
> +					return IRQ_HANDLED;
> +				}
> +
> +				line = j * 8 + int_bit;
> +
> +				handle_nested_irq(ab8500->irq_base + line);
> +				latch_val &= ~(1 << int_bit);
> +			} while (latch_val);
> +discard:
> +			hier_val &= ~(1 << latch_bit);
> +		} while (hier_val);
> +	}
> +	return IRQ_HANDLED;
> +}
With 4 nested loops, this one is getting difficult to read. Have you
considered splitting the loops into separate routines ?

Cheers,
Samuel.

-- 
Intel Open Source Technology Centre
http://oss.intel.com/