Re: [PATCH 10/10] hw/hppa: Add 715 machine type including NCR710 SCSI

[Richard Henderson <richard.henderson@linaro.org>]

On 10/17/25 15:06, deller@kernel.org wrote:
> @@ -363,8 +365,20 @@ static void machine_HP_common_init_tail(MachineState *machine, PCIBus *pci_bus,
>   
>       /* SCSI disk setup. */
>       if (drive_get_max_bus(IF_SCSI) >= 0) {
> -        dev = DEVICE(pci_create_simple(pci_bus, -1, "lsi53c895a"));
> -        lsi53c8xx_handle_legacy_cmdline(dev);
> +        if (pci_bus) {
> +            dev = DEVICE(pci_create_simple(pci_bus, -1, "lsi53c895a"));
> +            lsi53c8xx_handle_legacy_cmdline(dev);
> +        } else {
> +            dev = lasi_ncr710_init(addr_space,
> +                       translate(NULL, LASI_HPA_715 + 0x6000),
> +                       qdev_get_gpio_in(lasi_dev, LASI_IRQ_SCSI_HPA));

Indentation past (.

> +            if (dev) {
> +                lasi_ncr710_handle_legacy_cmdline(dev);
> +            } else {
> +                qemu_log("HPPA Machine: Warning - "
> +                         "Failed to create LASI NCR710 controller\n");

When can the init fail?
Should you be passing down an Error pointer?
Would an assert be sufficient?

> +    /* Create ISA bus, needed for PS/2 kbd/mouse port emulation */
> +    isa_bus = hppa_isa_bus(translate(NULL, IDE_HPA));
> +    assert(isa_bus);

... like here.

As an aside, we could usefully mark such functions returns_nonnull.
Or even __attribute__((malloc, returns_nonnull)) to indicate a new object.
That's obviously separate from adding the 715 machine, but it's something that we should 
think about rolling out tree-wide.


r~