[patches 6/8] pcmcia: remove racy try_irq()

Dominik Brodowski linux at dominikbrodowski.de
Tue Jan 11 17:41:20 EST 2005

On Tue, Jan 11, 2005 at 10:32:15PM +0000, Russell King wrote:
> On Tue, Jan 11, 2005 at 11:27:53PM +0100, Dominik Brodowski wrote:
> > > the kernel resource database can't be depended on for hot plug
> > > resource allocation.
> > 
> > A similar additon _may_ yet be needed for the IRQ resource database, for
> > ioport / iomem I tend to disagree.
> David is correct here as well - just look at the ports which VGA
> cards scatter around in reality, and the contents of /proc/ioports.
> The kernel's idea is no where complete.

That's why we still need a ioport/iomem resource database for non-statically mapped
sockets. That's true. What's incorrect is that the kernel resource database
doesn't distinguish between "occupied" and "used" resources.

Also, the ability to specify a IRQ to be used by a specific socket isn't
limited by my patches -- /sys/class/pcmcia_socket/pcmcia_socket%n/irq_mask
allows you to do so.

> Then there's these mysterious ACPI ports which lurk about and are
> (afaik) completely undetectable,

They are detectable parsing the ACPI tables and

a) checking all _CRS (current resource settings) methods _and_
b) checking for "unproduced" resources (ACPI always "produces" resources at
   a higher level and "consumes" them at a lower level).


More information about the linux-pcmcia mailing list