Hello,

> >Anyway, here's another patch ( corrects some minor oversight, should now
> >catch all(?) lost interrupts, not only the PCI ones. Well, at least it
> >tries )
> 
> I'm getting
...
> pci1: i/o space, memory space enabled
> ofb0 at pci1 dev 11 function 0
> 
> and then it hung.
Damn. So Riccardo isn't the only one affected.

> I only found one line different from your last patch. Is this correct (comments excepted)?
The difference is minor, there should be one '& INT_LEVEL_MASK' commented out.

Could you define IRQ_DEBUG ( should be commented out somewhere ) and see what happens?

I think there's a race - establish_intr() installs a fake handler while calculating the masks, but the loop in do_pending_int() will cycle until an IRQ is cleared... I guess interrupts should be disabled there since the fake handler won't clear anything.

have fun
Michael