Re: [Qemu-devel] [Qemu-ppc] [PATCH 3/4] libqos: fix spapr qpci_map()

[Laurent Vivier]

On 13/12/2016 09:46, Thomas Huth wrote:
> A short patch description would be nice here.
> 
> On 13.12.2016 00:28, Laurent Vivier wrote:
>> Signed-off-by: Laurent Vivier <address@hidden>
>> ---
>>  tests/libqos/pci-spapr.c | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/tests/libqos/pci-spapr.c b/tests/libqos/pci-spapr.c
>> index 1e5d015..2043f1e 100644
>> --- a/tests/libqos/pci-spapr.c
>> +++ b/tests/libqos/pci-spapr.c
>> @@ -193,8 +193,8 @@ QPCIBus *qpci_init_spapr(QGuestAllocator *alloc)
>>      ret->pio.size = SPAPR_PCI_IO_WIN_SIZE;
>>  
>>      /* 32-bit portion of the MMIO window is at PCI address 2..4 GiB */
>> -    ret->mmio32_cpu_base = SPAPR_PCI_BASE + SPAPR_PCI_MMIO32_WIN_SIZE;
>> -    ret->mmio32.pci_base = 0x80000000; /* 2 GiB */
>> +    ret->mmio32_cpu_base = SPAPR_PCI_BASE;
>> +    ret->mmio32.pci_base = SPAPR_PCI_MMIO32_WIN_SIZE;
>>      ret->mmio32.size = SPAPR_PCI_MMIO32_WIN_SIZE;
> 
> Not sure whether it is really right to use SPAPR_PCI_MMIO32_WIN_SIZE for
> the pci_base? I mean, the WIN_SIZE could also be 1G for example, but the
> base could still be at 2G ? IMHO it's better to keep the old value for
> pci_base here.

in hw/ppc/spapr.c, we have:

    *mmio32 = SPAPR_PCI_BASE + (index + 1) * SPAPR_PCI_MEM32_WIN_SIZE;

So I think it's logic to have the PCI base at  SPAPR_PCI_MMIO32_WIN_SIZE.

Thanks for your review,
Laurent