Clemens Ladisch wrote: >wes schreiner wrote: > > >>The support for using lspci/setpci is in all 2.4.x kernels I think, but >>it doesn't always do what you want. I have a Zoran 36057 video capture >>device sitting on IRQ 10, same as my sound card. IRQ 9 is unused (no >>ACPI), so I just tried "setpci -v -s 03:09.0 INTERRUPT_LINE=09" to >>change its IRQ. lspci -v still says it is at IRQ 10, but lspci -b -v >>says it is at IRQ 9. Which is it? OK, I modprobe zr36067 and the module >>loads, and now both lspci -v and lspci -v -b agree that the Zoran chip >>is at IRQ 10, so nothing changed. >> >> > >The interrupt_line register doesn't affect the card in any way, it's >there for information purposes only. The BIOS sets this register so >that drivers know which interrupt to use. > >To change the interrupt routing, you have to reprogram some registers >of the PCI controller (starting at 60h in Intel's ICHx southbridges >IIRC, but read Intel's datasheet first). And I don't know if the >kernel is able to realize that the interrupt has changed. > > >HTH >Clemens > > Well there we go, so it's back to juggling cards in PCI slots and switching motherboards. I can report that irqtune does indeed change the interrupt priority though, using kernel 2.4.21. That might help someone. wes
