On 07/16/2013 06:32 PM, Alexander Graf wrote: > > > Am 16.07.2013 um 08:20 schrieb Paolo Bonzini <pbonz...@redhat.com>: > >> Il 16/07/2013 07:19, Alexey Kardashevskiy ha scritto: >>> In the past, IO space could not be mapped into the memory address space >>> so we introduced a workaround for that. Nowadays it does not look >>> necessary so we can remove the workaround and make sPAPR PCI >>> configuration simplier. >>> >>> This also removes all byte swappings as it is not PHB's to take care >>> of endiannes - devices should do convertion if they want to. And almost >>> every PCI device which uses IO ports does that by registering IO ports >>> region with memory_region_init_io() (Intel e1000, RTL8139, virtio). >>> >>> However VGA uses MemoryRegionPortio which does not support endiannes >>> but it still expects the convertion to be done. For this case only, >>> this patch adds LITTLE_ENDIAN flag to portio_ops. Tests on PPC64 show >>> that other devices are not affected by this change. x86 systems should >>> not suffer either as having LITTLE_ENDIAN there has no effect. >>> >>> Signed-off-by: Alexey Kardashevskiy <a...@ozlabs.ru> >>> --- >>> >>> This removes bugs at least from SPAPR so any further fixes should be equal >>> for all platforms and hopefully will break all platform altogether but not >>> just PPC64-pSeries :) >>> >>> Did I miss anything here? >> >> No, I don't think so. The patch looks good. > > ... and will break all Mac targets again, no? Not to speak of non-ppc devices.
Like what? I do not mind/argue/discuss, may be it breaks, I am just looking for good examples to test. -- Alexey