Thanks for taking a look.

> On Jul 4, 2026, at 3:30 PM, Mark Cave-Ayland <[email protected]> 
> wrote:
> 
> This doesn't look right: in general we should avoid accessing a device state 
> struct directly unless it is absolutely unavoidable.

I'll handle this as part of switching to a GPIO (below).

> Can you move this into a separate via1_swim_update(v1s) function as done for 
> the rtc and adb devices as below? If possible I'd like to minimise the 
> per-device logic in mos6522_q800_via1_write().

Yup, will do.

> Given that SWIM2 is also a TYPE_DEVICE, you should be able to expose the sel 
> signal as a GPIO on the SWIM2 device in a similar way as to how "nubus-irq" 
> is defined in mos6522_q800_via2_init() but with just 1 signal. This allows 
> you to connect the signal with qdev_connect_gpio_out() (see hw/m68k/q800.c) 
> and drive it using the qemu IRQ functions, which should avoid having to 
> reference the SWIM2 device itself directly at all.

Good idea -- I'll switch over to using a GPIO to drive this for v4.

Matt

Reply via email to