On 28 June 2013 21:41, Anthony Liguori <anth...@codemonkey.ws> wrote: > One thing I've been thinking about reviewing this code, what should we > be doing in virtio.c? > > We have barriers but we're relying on st[u][wlb]_phys having atomic > semantics. I think it's okay in practice but if we're taking a more > diligent approach here should we introduce atomic variants that work on > guest phys addresses?
Those accesses are weird anyway, because they implicitly depend on going through a path which is cache-coherent with the CPU that was used to run the guest VCPU. So they're not strictly speaking writing "guest physical memory" necessarily... -- PMM
