> > Wouldn't it be more correct for volatile PCI registers to be
> > read at driver initialisation time rather than PCI subsystem
> > initialisation? Perhaps in pci_find_device().
> I'll probably introduce an ioctl for setting of these kernel
> structures on a live system.
How about this:
Providing the PCI registers are only modified by the kernel or
drivers, why not get pci_write_*() to update pci_dev if it modifies
anything it has "cached"? (perhaps only irq and base registers).
> By the way, what interrupt problems do you exactly have? Unless
> there is a hardware bug or something rotten in the drivers, interrupt
> sharing should work fine on PCI.
You already responded to it. It was the problem with enabling USB
on a Toshiba 7020CT causing a permanent spurious interrupt to occur.
By redirecting a particular pci pin to an unused irq, the problem
disappears. Unfortunately that pin is used by the IR driver and doesn't
work after the setpci since the driver picks up the irq as it was
initially instead of what it is when toshoboe is loaded.
Cheers,
Paul
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/