Here is the patch solving the problem. I have a "feeling" that maybe
there is some kind of performance issue, but not sure. I'll test that
later.
--- gpio.c 2008-03-14 11:31:33.000000000 +0100
+++ gpio.c-new 2008-03-14 11:38:35.000000000 +0100
@@ -442,19 +442,26 @@
struct irq_desc *desc = &irq_desc[irq];
unsigned pin;
struct irq_desc *gpio;
+ struct at91_gpio_bank *bank;
void __iomem *pio;
u32 isr;
- pio = get_irq_chip_data(irq);
+ bank = get_irq_chip_data(irq);
+ pio = bank->regbase;
/* temporarily mask (level sensitive) parent IRQ */
desc->chip->ack(irq);
for (;;) {
isr = __raw_readl(pio + PIO_ISR) & __raw_readl(pio + PIO_IMR);
- if (!isr)
- break;
+ if (!isr) {
+ if (!bank->next)
+ break;
+ bank = bank->next;
+ pio = bank->regbase;
+ continue;
+ }
- pin = (unsigned) get_irq_data(irq);
+ pin = bank->chipbase;
gpio = &irq_desc[pin];
while (isr) {
On Fri, Mar 14, 2008 at 10:41 AM, Bosko Radivojevic
<[EMAIL PROTECTED]> wrote:
> On Thu, Mar 13, 2008 at 6:51 PM, Gilles Chanteperdrix
> <[EMAIL PROTECTED]> wrote:
>
> > Ok. The code which does this is __ipipe_mach_demux_irq in
> > arch/arm/mach-at91/gpio.c. Could you check if at least it triggers ?
>
> I put the following code into __ipipe_mach_demux_irq() in mach-at91/gpio.c:
>
> [..]
> void __iomem *pio;
> u32 isr;
>
> printk (KERN_INFO "__ipipe_mach_demux_irq: irq=%d\n", irq);
> pio = get_irq_chip_data(irq);
> printk (KERN_INFO "__ipipe_mach_demux_irq: pio=%x\n", pio);
> [..]
> for (;;) {
> isr = __raw_readl(pio + PIO_ISR) & __raw_readl(pio + PIO_IMR);
> printk (KERN_INFO "__ipipe_mach_demux_irq: isr=%x\n", isr);
> if (!isr) {
> printk (KERN_INFO "__ipipe_mach_demux_irq: !isr\n");
> break;
> }
> pin = (unsigned) get_irq_data(irq);
> printk (KERN_INFO "__ipipe_mach_demux_irq: pin: %d\n", pin);
> [...]
>
> And all I get is just once (first time I send a packet):
> __ipipe_mach_demux_irq: irq=3
> __ipipe_mach_demux_irq: pio=c02203d8
> __ipipe_mach_demux_irq: isr=0
> __ipipe_mach_demux_irq: !isr
>
> Maybe pio is wrongly calculated in get_irq_chip_data()?
>
_______________________________________________
Adeos-main mailing list
[email protected]
https://mail.gna.org/listinfo/adeos-main
