> > Sounds like a good idea. Also, doing inline ... __init is rather silly.
> > How's the idea sound Dan?
>
> The idea is OK, but I would prefer a different implementation. You
> are trying to do basically the same thing as all of the boards that
> use versions of OpenPIC, EPIC, etc. Rather than have a fixed size
> array, and fill most of it with zero, a generic solution should be
> able to work with variable sized arrays......just pass the size
> somehow. The sizeof() works pretty well for this :-). The boards
> initialize the array to whatever they need, and don't have to worry
> about others.
Isn't that exactly what I did (in cpci405.c)?
int __init
ppc405_map_irq(struct pci_dev *dev, unsigned char idsel, unsigned char pin)
{
static char pci_irq_table[][4] =
/*
* PCI IDSEL/INTPIN->INTLINE
* A B C D
*/
{
{28, 28, 28, 28}, /* IDSEL 15 - cPCI slot 8 */
{29, 29, 29, 29}, /* IDSEL 16 - cPCI slot 7 */
{30, 30, 30, 30}, /* IDSEL 17 - cPCI slot 6 */
{27, 27, 27, 27}, /* IDSEL 18 - cPCI slot 5 */
{28, 28, 28, 28}, /* IDSEL 19 - cPCI slot 4 */
{29, 29, 29, 29}, /* IDSEL 20 - cPCI slot 3 */
{30, 30, 30, 30}, /* IDSEL 21 - cPCI slot 2 */
};
const long min_idsel = 15, max_idsel = 21, irqs_per_slot = 4;
return PCI_IRQ_TABLE_LOOKUP;
};
It seems to me, that nearly all other boards do it this way also.
Perhaps you were referring to the walnut/ep405 implementations, which still
had some zero's in the array (I corrected this by now).
Best regards,
Stefan.
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
