On 1 July 2013 19:46, Alexander Spyridakis
<a.spyridakis@xxxxxxxxxxxxxxxxxxxxxx> wrote:
> On 29 June 2013 21:20, Peter Maydell <peter.maydell@xxxxxxxxxx> wrote:
>> So this is enabling support for emulating 32 bit VMs with a
>> QEMU running as a 64 bit process, but the commit message
>> doesn't mention this. Did you test that combination?
Actually I wasn't quite right here -- in fact
\( "$target_name" = "arm" -a "$cpu" = "aarch64" \)
is the case where you're building a QEMU running as a 64 bit binary
but only supporting 32 bit CPUs. In particular this would try to build
a 64 bit binary but with QEMU's copy of the 32 bit kernel headers:
I'm not sure that's expected (by the kernel folk) to work at all. If not,
we should just not put the line into configure in the first place.
Note that that still leaves the case of "built aarch64-softmmu
with KVM support and user used -cpu cortex-a15" which you can only
catch at runtime. My guess is that will either (a) work or (b) be
caught already by the KVM_ARM_VCPU_INIT ioctl failing because we asked
for KVM_ARM_TARGET_CORTEX_A15 and the kernel didn't support it.
Either way, it seems like a more sensible way to support the 32-on-64
case when we eventually choose to do so.
> Nope, this case is not handled. Currently it won't crash but of course
> won't boot either. Definitely worth looking into, at the very minimum
> it should gracefully deny execution.
Yes, in general I'm happy for us not to support 32-bit-VM-on-64-host
initially, as long as it fails gracefully.
thanks
-- PMM
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
