Hi Thomas.

Am Freitag 26 Oktober 2012, 20:45:05 schrieb Thomas Petazzoni:

Ah, this is interesting. It is not entirely surprising, since the gpio
driver is new. Even though it re-uses most of the previous gpio driver,
it is by far not impossible that there will be a few regressions.

Could you add a few debug prints to see if you're looping *inside* the
function (which I find pretty unlikely), or if the function gets called
over and over again? 

ah sorry, my bad. of course the handler is called forever:

mvebu_gpio_to_irq pin=8
mvebu_gpio_irq_set_type pin=8
mvebu_gpio_edge_irq_unmask
mvebu_gpio_irq_handler irq=103
mvebu_gpio_level_irq_mask irq=103
mvebu_gpio_irq_handler irq=100
mvebu_gpio_level_irq_mask irq=100
mvebu_gpio_irq_handler irq=101
mvebu_gpio_level_irq_mask irq=101
mvebu_gpio_irq_handler irq=102
mvebu_gpio_level_irq_mask irq=102
mvebu_gpio_irq_handler irq=103
mvebu_gpio_level_irq_mask irq=103
mvebu_gpio_irq_handler irq=112
mvebu_gpio_level_irq_mask irq=112
mvebu_gpio_irq_handler irq=103
mvebu_gpio_level_irq_mask irq=103
mvebu_gpio_irq_handler irq=103
mvebu_gpio_level_irq_mask irq=103
mvebu_gpio_irq_handler irq=103
mvebu_gpio_level_irq_mask irq=103
mvebu_gpio_irq_handler irq=103
mvebu_gpio_level_irq_mask irq=103
mvebu_gpio_irq_handler irq=103
mvebu_gpio_level_irq_mask irq=103
mvebu_gpio_irq_handler irq=103
mvebu_gpio_level_irq_mask irq=103
mvebu_gpio_irq_handler irq=103
mvebu_gpio_level_irq_mask irq=103
mvebu_gpio_irq_handler irq=103
[..]

Could it be that the hang occurs during the
initialization of the gpio-leds or gpio-keys drivers?

yeah, if i disable gpio-leds, gpio-keys and gpio-fan the hang will disappear. 
But once i enable one of it the board hangs again.

Also, even though I'm pretty sure it isn't going to fix your problem,
note the following mvebu-gpio fix:

  http://article.gmane.org/gmane.linux.ports.arm.kernel/195018

thats already in andrew's tree.

-- 
michael