Re: [PATCH 07/42] hw/intc/i8259: Introduce i8259 proxy "isa-pic"

[Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>]

On 01/09/2022 17:25, Bernhard Beschow wrote:

> Having an i8259 proxy allows for ISA PICs to be created and wired up in
> southbridges. This is especially interesting for PIIX3 for two reasons:
> First, the southbridge doesn't need to care about the virtualization
> technology used (KVM, TCG, Xen) due to in-IRQs (where devices get
> attached) and out-IRQs (which will trigger the IRQs of the respective
> virtzalization technology) are separated. Second, since the in-IRQs are
> populated with fully initialized qemu_irq's, they can already be wired
> up inside PIIX3.
> 
> Signed-off-by: Bernhard Beschow <shentey@gmail.com>
> ---
>   hw/intc/i8259.c         | 27 +++++++++++++++++++++++++++
>   include/hw/intc/i8259.h | 14 ++++++++++++++
>   2 files changed, 41 insertions(+)
> 
> diff --git a/hw/intc/i8259.c b/hw/intc/i8259.c
> index cc4e21ffec..531f6cca53 100644
> --- a/hw/intc/i8259.c
> +++ b/hw/intc/i8259.c
> @@ -458,9 +458,36 @@ static const TypeInfo i8259_info = {
>       .class_size = sizeof(PICClass),
>   };
>   
> +static void isapic_set_irq(void *opaque, int irq, int level)
> +{
> +    ISAPICState *s = opaque;
> +
> +    qemu_set_irq(s->out_irqs[irq], level);
> +}
> +
> +static void isapic_init(Object *obj)
> +{
> +    ISAPICState *s = ISA_PIC(obj);
> +
> +    qdev_init_gpio_in(DEVICE(s), isapic_set_irq, ISA_NUM_IRQS);
> +    qdev_init_gpio_out(DEVICE(s), s->out_irqs, ISA_NUM_IRQS);
> +
> +    for (int i = 0; i < ISA_NUM_IRQS; ++i) {
> +        s->in_irqs[i] = qdev_get_gpio_in(DEVICE(s), i);
> +    }
> +}
> +
> +static const TypeInfo isapic_info = {
> +    .name          = TYPE_ISA_PIC,
> +    .parent        = TYPE_ISA_DEVICE,
> +    .instance_size = sizeof(ISAPICState),
> +    .instance_init = isapic_init,
> +};
> +
>   static void pic_register_types(void)
>   {
>       type_register_static(&i8259_info);
> +    type_register_static(&isapic_info);
>   }
>   
>   type_init(pic_register_types)
> diff --git a/include/hw/intc/i8259.h b/include/hw/intc/i8259.h
> index e2b1e8c59a..0246ab6ac6 100644
> --- a/include/hw/intc/i8259.h
> +++ b/include/hw/intc/i8259.h
> @@ -1,6 +1,20 @@
>   #ifndef HW_I8259_H
>   #define HW_I8259_H
>   
> +#include "qom/object.h"
> +#include "hw/isa/isa.h"
> +#include "qemu/typedefs.h"
> +
> +#define TYPE_ISA_PIC "isa-pic"
> +OBJECT_DECLARE_SIMPLE_TYPE(ISAPICState, ISA_PIC)
> +
> +struct ISAPICState {
> +    ISADevice parent_obj;
> +
> +    qemu_irq in_irqs[ISA_NUM_IRQS];
> +    qemu_irq out_irqs[ISA_NUM_IRQS];
> +};
> +
>   /* i8259.c */
>   
>   extern DeviceState *isa_pic;

So effectively this proxy device is working around the issue that we don't yet have a 
qdev-ified PIC that can be used to wire up ISA bus IRQs via qdev_connect_gpio_out(). 
Whilst ultimately this is the goal, I think having the proxy is a nice intermediate 
step as long as we can ensure that the device is not user-visible (i.e. it won't be 
exposed via the command line) so that this can be improved later.


ATB,

Mark.