On 25.01.2014, at 16:36, Victor Kamensky <victor.kamensky@xxxxxxxxxx> wrote: > On 25 January 2014 01:20, Alexander Graf <agraf@xxxxxxx> wrote: > >> It even >> does it on BE PPC if you access devices through swizzling >> buses, but we don't care as hypervisor. All we know in kvm is: >> >> - instruction used for access >> -> originating register (value) >> -> access size (len) >> -> virtual address (which we translate to phys) >> - guest endianness mode > > Do you agree that in above if you just change guest endianness > then device will see different (byteswapped) value so as far as > device concerned it is completely different transactions, different > effect. Why do we consider them? So if you want to see the > same transaction and have the same effect on device, then when > guest switches endianness it will need to add/drop byteswap, > in order to have the same effect. Yes. We consider it because KVM has to do the counter-swap manually based on the guest CPU's endianness setting because the only thing KVM knows are the 4 bits I mentioned above. Alex -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
