hostap hangs in request_irq

[Pavel Roskin]

Hello, Bruno!

> i don't know if this is relevant to anyone except me, but to answer my own
> question, this is the solution i came up with:
>
> the attached patch adds deferred initialization of interrupts, so that all
> devices have been reset properly before the irqs are requested:
>
> this is a workaround for multiple cards that have to share the same hardware
> interrupt line. when the system reboots, the firmware obviously still
> generates interrupts even after a pci bus reset. these interrupts are never
> acknowledged from the newly booted kernel, so the interrupt stays active.

I believe this should not happen.  Unused PCI devices should be disabled
and should not generate any interrupts.  Maybe you should ask in LKML.

However, network devices in Linux 2.6.1 request interrupt only when the
interface goes up and free it when the interface goes down.  See e.g.
drivers/net/3c59x.c.  Maybe that's what you should try.

-- 
Regards,
Pavel Roskin