Re: [PATCH 3/4] hw/pci-host/prep: Fix PCI swizzling in map_irq()

[Mark Cave-Ayland]

On 12/10/2020 08:19, Philippe Mathieu-Daudé wrote:

> In commit a01d8cadadf we changed the number of IRQs to 4 but
> forgot to update the map_irq() function. Do it now.
> 
> Fixes: a01d8cadadf ("Fix memory corruption ... in PreP emulation")
> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> ---
> Cc: Jocelyn Mayer <l_indien@magic.fr>
> Cc: Julian Seward <julian@valgrind.org>
> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> ---
>  hw/pci-host/prep.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/hw/pci-host/prep.c b/hw/pci-host/prep.c
> index 064593d1e52..2224135fedb 100644
> --- a/hw/pci-host/prep.c
> +++ b/hw/pci-host/prep.c
> @@ -195,7 +195,7 @@ static const MemoryRegionOps raven_io_ops = {
>  
>  static int raven_map_irq(PCIDevice *pci_dev, int irq_num)
>  {
> -    return (irq_num + (pci_dev->devfn >> 3)) & 1;
> +    return (irq_num + (pci_dev->devfn >> 3)) & 3;
>  }
>  
>  static void raven_set_irq(void *opaque, int irq_num, int level)

It feels like this should also have a corresponding change in OpenBIOS for
consistency, even though technically because of the OR on IRQ 15 it doesn't 
really
matter. The relevant part in OpenBIOS can be found here:
https://git.qemu.org/?p=openbios.git;a=blob;f=drivers/pci.c;h=34ae69a907b6312a3a7ab218afe8ba9efded1df7;hb=7f28286f5cb1ca682e3ba0a8706d8884f12bc49e#l2001
and in particular this section:

    /* Use the same "physical" routing as QEMU's raven_map_irq() although
       ultimately all 4 PCI interrupts are ORd to IRQ 15 as indicated
       by the PReP specification */
    props[(*ncells)++] = arch->irqs[((intno - 1) + (addr >> 11)) & 1];


ATB,

Mark.