On Wed, Feb 24, 2010 at 12:26:53PM +0000, Paul Brook wrote: > > > If we do have a software > > > fallback then the feature bit is just for backwards compatibility, and > > > should be enabled unconditionally (on current machine types). > > > > Software fallback might turn out to be slower than disabling the feature > > in the guest. For example, and extra pass over packet might cause extra CPU > > cache thrashing. In that case, it's not obvious whether enabling it > > unconditionally will turn out to be a good idea. But we'll have to have > > that code to be able to tell. > > IMO once you accept that these things can change, consistency is more > important than trying to guess what the "best" option may be.
Yes, SW fallback might be nice to have. What's important is likely to depend on specific user. > Starting qemu on machine A and migrating to machine B should give the same > guest environment as starting on machine B. > > Paul So currently, the way we try to ensure this is by checking feature bits against the list supported by backend, and failing migration if there's a discrepancy. -- MST
