On 12/06/2011 02:58 PM, Jan Kiszka wrote: > In this revision, I'm now trying the approach of backend/frontend > split-ups for the affected IRQ chips. That means we keep a single qdev > device description but fork off specific logic early during device init. > The backends support this by providing hooks that user space and KVM > models can implement differently. > > The result is slightly larger and comes with the not really beautiful > ioapic.kvm_gsi_base property but should otherwise meet expectations. > > Comments?
Looks good to me, much nicer than the previous approaches. I'll wait a bit for more reviews though. > PS: Series is still against old uq/master, therefore containing patches > that took/will take different routes. I just pushed a rebased uq/master. In the future, either ping me or just base on upstream (which uq/master supposedly tracks). -- error compiling committee.c: too many arguments to function
