[Qemu-devel] Re: [PATCH 0/5] piix_pci cleanup

[Juan Quintela]

Blue Swirl <address@hidden> wrote:
> On Fri, Aug 21, 2009 at 8:36 PM, Juan Quintela<address@hidden> wrote:
>> Hi
>>
>> This series:
>> - split piix4 from piix_pci.  The only shared code where piix_save/load, i.e.
>>  almost nothing.  Once there, compile piix4.o only for mips (it was not used
>>  anywhere else).
>> - Move global variables to PCII440FXState.  Nice and clean until....
>> - pci_irq_levels: This is great:
>>  * it is saved/loaded from i440fx
>>  * it is cleaned during reset in piix3
>>  * it is used in piix3_set_irq, that don't receive neither i440fx nor
>>    piix state, it needs to be a global variable. (created a global link
>>    until a better solution appear).
>>
>> I looked where to "hide" pci_irq_levels and piix3_dev (both needed in
>> piix3_set_irq), and didn't found where to pass them, out of:
>> - hack i8259 to hide it into PicState2 (that one got passed through the pic)
>>  ugly for words, but will work
>> - try to add <something> at setup_irq time, but at that point we have:
>>  * opaque -> pci_dev of device that we are working with
>>  * from there we can get to the bus that the device is attached to,
>>    but no way to go from there to the Host bridge (that is what we wanted
>>    in the 1st place)
>>
>> Notice that this is needed for both pc and mips_malta.c
>>
>> Ideas on where to hide a pci_dev inside a bus?
>
> pci_set_irq_fn should take an opaque state pointer (given at
> pci_register_bus time):
>
> typedef void (*pci_set_irq_fn)(void *opaque, qemu_irq *pic, int
> irq_num, int level);
>
> PCIBus *pci_register_bus(DeviceState *parent, const char *name,
>                          pci_set_irq_fn set_irq, pci_map_irq_fn map_irq,
>                          qemu_irq *pic, void *irq_opaque, int
> devfn_min, int nirq);

I fixed it changing the type of qemu_irq *pic to void *irq_state.
See the v2 of the series.

Thanks very much for the suggestion.

later, Juan.