]> rtime.felk.cvut.cz Git - can-eth-gw-linux.git/commitdiff
regmap: set MASK_ON_SUSPEND/SKIP_SET_WAKE if no wake_base
authorStephen Warren <swarren@nvidia.com>
Wed, 1 Aug 2012 17:40:49 +0000 (11:40 -0600)
committerMark Brown <broonie@opensource.wolfsonmicro.com>
Sat, 4 Aug 2012 11:13:53 +0000 (12:13 +0100)
If a regmap-irq chip has no wake base:

* There's no point calling .irq_set_wake, hence IRQCHIP_SKIP_SET_WAKE.

* If some IRQs in the chip are enabled for wake and some aren't, we
  should mask those interrupts that are not wake enabled, so that if
  they occur during suspend, the system is not awoken. Hence,
  IRQCHIP_MASK_ON_SUSPEND.

Note that IRQCHIP_MASK_ON_SUSPEND is handled by check_wakeup_irqs(),
which always iterates over every single interrupt in the system,
irrespective of whether an interrupt is a child of a controller whose
output interrupt has no wake-enabled inputs and hence is presumably
masked itself. Hence this change might cause interrupt unnecessary
masking operations and associated register I/O.

Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
drivers/base/regmap/regmap-irq.c

index bf1203cbd122b2036cd18af12ccf293e3e3671aa..9a6649c82daa3359654ad217078e469578d104a8 100644 (file)
@@ -287,6 +287,11 @@ int regmap_add_irq_chip(struct regmap *map, int irq, int irq_flags,
 
        d->irq_chip = regmap_irq_chip;
        d->irq_chip.name = chip->name;
+       if (!chip->wake_base) {
+               d->irq_chip.irq_set_wake = NULL;
+               d->irq_chip.flags |= IRQCHIP_MASK_ON_SUSPEND |
+                                    IRQCHIP_SKIP_SET_WAKE;
+       }
        d->irq = irq;
        d->map = map;
        d->chip = chip;