Re: [PATCH v4 1/9] hw/pci-host/i440fx: Inline sysbus_add_io()

["Philippe Mathieu-Daudé" <philmd@linaro.org>]

On 6/3/23 07:57, Bernhard Beschow wrote:
> 
> 
> Am 22. Februar 2023 18:05:51 UTC schrieb Bernhard Beschow <shentey@gmail.com>:
>>
>>
>> Am 22. Februar 2023 10:58:08 UTC schrieb "Philippe Mathieu-Daudé" <philmd@linaro.org>:
>>> On 13/2/23 17:19, Bernhard Beschow wrote:
>>>> sysbus_add_io() just wraps memory_region_add_subregion() while also
>>>> obscuring where the memory is attached. So use
>>>> memory_region_add_subregion() directly and attach it to the existing
>>>> memory region s->bus->address_space_io which is set as an alias to
>>>> get_system_io() by the pc machine.
>>>>
>>>> Signed-off-by: Bernhard Beschow <shentey@gmail.com>
>>>> Reviewed-by: Thomas Huth <thuth@redhat.com>
>>>> ---
>>>>    hw/pci-host/i440fx.c | 5 +++--
>>>>    1 file changed, 3 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/hw/pci-host/i440fx.c b/hw/pci-host/i440fx.c
>>>> index 262f82c303..9c6882d3fc 100644
>>>> --- a/hw/pci-host/i440fx.c
>>>> +++ b/hw/pci-host/i440fx.c
>>>> @@ -27,6 +27,7 @@
>>>>    #include "qemu/range.h"
>>>>    #include "hw/i386/pc.h"
>>>>    #include "hw/pci/pci.h"
>>>> +#include "hw/pci/pci_bus.h"
>>>>    #include "hw/pci/pci_host.h"
>>>>    #include "hw/pci-host/i440fx.h"
>>>>    #include "hw/qdev-properties.h"
>>>> @@ -217,10 +218,10 @@ static void i440fx_pcihost_realize(DeviceState *dev, Error **errp)
>>>>        PCIHostState *s = PCI_HOST_BRIDGE(dev);
>>>>        SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
>>>>    -    sysbus_add_io(sbd, 0xcf8, &s->conf_mem);
>>>> +    memory_region_add_subregion(s->bus->address_space_io, 0xcf8, &s->conf_mem);
>>>
>>> To avoid accessing internal fields we should stick to the PCI API:
>>>
>>>     memory_region_add_subregion(pci_address_space_io(PCI_DEVICE(dev)),
>>>                                 0xcf8, &s->conf_mem);
>>
>> dev is of type PCIHostState which derives from SysBusDevice, not PCIDevice. AFAICS there is no getter implemented on PCIBus.

You are right, there is no getter for PCIBus::address_space_io,
it is accessed directly:

   MemoryRegion *pci_address_space_io(PCIDevice *dev)
   {
       return pci_get_bus(dev)->address_space_io;
   }

If it is considered a property, probably no need for getter.

So:
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>

> 
>>
>>>
>>>>        sysbus_init_ioports(sbd, 0xcf8, 4);
>>>>    -    sysbus_add_io(sbd, 0xcfc, &s->data_mem);
>>>> +    memory_region_add_subregion(s->bus->address_space_io, 0xcfc, &s->data_mem);
>>>>        sysbus_init_ioports(sbd, 0xcfc, 4);
>>>
>>> Now all classes implementing PCI_HOST_BRIDGE register conf/data in I/O
>>> space, so this could be a pattern justifying reworking a bit the
>>> PCIHostBridgeClass or adding an helper in "hw/pci/pci_host.h" to do
>>> that generically.

(this comment is besides the scope of this patch)

>> What do you mean exactly? There are PCI hosts spawning two PCI buses and therefore have two such spaces.

I haven't checked but IIRC each PCI host exposing I/O registers
to access buses via ISA I/O open-code it. Even if it exposes
multiple buses, the same pattern is used (1 time per bus).

Anyway, no need to worry about that now...