Re: [PATCH v2] hw/sparc64/sun4u: Mark unusable PCI busses as full to ease device plugging

On 6/7/26 23:19, Philippe Mathieu-Daudé wrote:
> On 6/7/26 23:16, Mark Cave-Ayland wrote:
>> On 06/07/2026 21:29, Philippe Mathieu-Daudé wrote:
>>

>>>> -    pci_bus = PCI_HOST_BRIDGE(sabre)->bus;
>>>>       pci_busA = pci_bridge_get_sec_bus(sabre->bridgeA);
>>>>       pci_busB = pci_bridge_get_sec_bus(sabre->bridgeB);
>>>>
>>>> -    /* Only in-built Simba APBs can exist on the root bus, slot 0 
>>>> on busA is
>>>> -       reserved (leaving no slots free after on-board devices) 
>>>> however slots
>>>> -       0-3 are free on busB */
>>>> -    pci_bus_set_slot_reserved_mask(pci_bus, 0xfffffffc);
>>>> +    /*
>>>> +     * Slot 0 on busA is reserved (leaving no slots free after on- 
>>>> board
>>>> +     * devices) however slots 0-3 are free on busB
>>>> +     */
>>>>       pci_bus_set_slot_reserved_mask(pci_busA, 0xfffffff1);
>>>>       pci_bus_set_slot_reserved_mask(pci_busB, 0xfffffff0);
>>>>
>>>> ---
>>>>
>>>> I'm still looking at the 2nd part, the reserved slots.
>>>
>>> (Because with this patch, the test passes as we can plug devices from
>>> the command line, but they end plugged on pciA instead of pciB -- thus
>>> the 2nd part missing).
>>
>> Thanks for taking a look, Phil. IIRC the reason I added 
>> slot_reserved_mask was so that it was possible to populate pciA 
>> programatically in sun4uv_init(), and then enable the mask so that 
>> specifying an additional PCI device using -device on the command line 
>> without an explict PCI bus would cause the automatic allocation to 
>> place the device on pciB.
> 
> The slot_reserved_mask calls make sense, but AFAICT they don't work as
> expected, since CLI devices are added on pciA. If you want to have a
> look on top of this snippet, be my guest! Otherwise I'll continue
> debugging that tomorrow, long day and late here :\

Generic PCI fix posted here:
https://lore.kernel.org/qemu-devel/[email protected]/