Re: nobody cared about IRQ 19 (firewire, on a HP 2510p notebook)

[Stefan Richter <stefanr@s5r6.in-berlin.de>]

Martin Michlmayr wrote:
> * Thomas Gleixner <tglx@linutronix.de> [2008-03-31 12:19]:
>> > >      Board: PM965/GM965/GL960 based
>> > >      The R5C832 is known to work with ohci1394 according to
>> > >      http://hardware4linux.info/component/14348/ and other reports.
>> > 
>> > Its also known to work with the juju firewire stack -- that's the controller 
>> > in my own laptop, as well as a few other folks here in the office, all 
>> > running the new stack.
>> 
>> Hmm, can you please check whether you can reproduce the problem with
>> the original stack ?
> 
> I get virtually the same message with the old and new stack.

I can confirm that what Martin previously posted in this thread shows
that firewire-ohci + firewire-core and ohci1394 + ieee1394 fail in the
same way:
  - All MMIO reads and writes leading up to chip initialization works,
  - first "self ID complete" interrupt and corresponding MMIO reads and
    DMAs work and lead up to recognition of the local node by the
    firewire/1394 mid layer,
  - a while later the IRQ is disabled with "nobody cared".
IRQ handlers are drivers/firewire/fw-ohci.c::irq_handler() and
drivers/ieee1394/ohci1394.c::ohci_irq_handler().  Both return IRQ_NONE
if readl() on the register which contains the interrupt event type
returns 0 or ~0, which both are impossible values _if_ the chip
generated the interrupt _and_ MMIO reads work.
-- 
Stefan Richter
-=====-==--- --== =====
http://arcgraph.de/sr/