linux/arch/arm/mach-footbridge/cats-pci.c
Ben Dooks e38d523fd0 [ARM] CATS: Do not try and map bad PCI IRQ numbers
The cats_map_irq() function in the Simtec CATS support
code is mapping IRQ 255 (invalid IRQ) into what is a
supposedly valid interrupt numner which can cause problems
with other devices then seeing an interrupt they cannot
claim.

If the IRQ number if >= 255, then return -1 as this is not
something we can map.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
2008-08-09 13:51:53 +01:00

59 lines
1.3 KiB
C

/*
* linux/arch/arm/mach-footbridge/cats-pci.c
*
* PCI bios-type initialisation for PCI machines
*
* Bits taken from various places.
*/
#include <linux/kernel.h>
#include <linux/pci.h>
#include <linux/init.h>
#include <asm/irq.h>
#include <asm/mach/pci.h>
#include <asm/mach-types.h>
/* cats host-specific stuff */
static int irqmap_cats[] __initdata = { IRQ_PCI, IRQ_IN0, IRQ_IN1, IRQ_IN3 };
static int __init cats_map_irq(struct pci_dev *dev, u8 slot, u8 pin)
{
if (dev->irq >= 255)
return -1; /* not a valid interrupt. */
if (dev->irq >= 128)
return dev->irq & 0x1f;
if (dev->irq >= 1 && dev->irq <= 4)
return irqmap_cats[dev->irq - 1];
if (dev->irq != 0)
printk("PCI: device %02x:%02x has unknown irq line %x\n",
dev->bus->number, dev->devfn, dev->irq);
return -1;
}
/*
* why not the standard PCI swizzle? does this prevent 4-port tulip
* cards being used (ie, pci-pci bridge based cards)?
*/
static struct hw_pci cats_pci __initdata = {
.swizzle = NULL,
.map_irq = cats_map_irq,
.nr_controllers = 1,
.setup = dc21285_setup,
.scan = dc21285_scan_bus,
.preinit = dc21285_preinit,
.postinit = dc21285_postinit,
};
static int __init cats_pci_init(void)
{
if (machine_is_cats())
pci_common_init(&cats_pci);
return 0;
}
subsys_initcall(cats_pci_init);