On Fri, 2005-03-11 at 10:44 +0000, Philip Blundell wrote:
> On Thu, 2005-03-10 at 17:06 -0800, Erik Hovland wrote:
> > Think you could expound on what stupid drivers don't like irq 0.
>
> It seems that serial.c is the main offender.
There have been several attempts at defining a global NO_IRQ constant,
though so far, it's been painful enough that none of them was complete.
At least, my latest rant about it caused a few more archs to define
it :) Ideally, serial.c could then be fixed to use it. The problem is
that some of the probe_irq crap has a strange API which is sort-of
designed around positive numbers beeing irqs, 0 beeing none and negative
numbers beeing errors. So x86 at least would need massaging.
Also, serial.c gets passed tables that are #defined in various fancy
ways by pretty much all architectures, if any of those happens to
contain a 0 for the IRQ, that must be fixed and turned into a NO_IRQ.
So it's not a simple task, certainly not for 2.4, but people are welcome
to send patches for 2.6.
Ben.
Received on Thu Mar 17 2005 - 02:49:36 EST
This archive was generated by hypermail 2.2.0 : Mon Jul 25 2005 - 17:20:11 EDT