On 03/28/2012 03:31 PM, Grant Likely wrote:
On Mon, 26 Mar 2012 12:31:19 -0700, David Daney<ddaney.cavm@xxxxxxxxx> wrote:
From: David Daney<david.daney@xxxxxxxxxx>
Create two domains. One for the GPIO lines, and the other for on-chip
sources.
Signed-off-by: David Daney<david.daney@xxxxxxxxxx>
---
[...]
+struct octeon_irq_gpio_domain_data {
+ unsigned int base_hwirq;
+};
Hmmm...
+static int octeon_irq_gpio_xlat(struct irq_domain *d,
+ struct device_node *node,
+ const u32 *intspec,
+ unsigned int intsize,
+ unsigned long *out_hwirq,
+ unsigned int *out_type)
+{
[...]
+ *out_hwirq = gpiod->base_hwirq + pin;
...base_hwirq is only used here...
[...]
+ gpiod = kzalloc(sizeof (*gpiod), GFP_KERNEL);
+ if (gpiod) {
+ /* gpio domain host_data is the base hwirq number. */
+ gpiod->base_hwirq = 16;
+ irq_domain_add_linear(gpio_node, 16,&octeon_irq_domain_gpio_ops, gpiod);
... and it is unconditionally set to 16. It looks to me like
base_hwirq and the associated kzalloc() is unnecessary.
There is a little information asymmetry here. You don't know that I
have a patch queued up to add another user of the GPIO irq_domain that
has a different base_hwirq.
I could re-do this to hard code it, and then add it back. But it would
really just be busy work.
David Daney
