Sheng Yang wrote: > Some guest device driver may leverage the "Non-Snoop" I/O, and explicitly > WBINVD or CLFLUSH to a RAM space. Since migration may occur before WBINVD or > CLFLUSH, we need to maintain data consistency either by: > 1: flushing cache (wbinvd) when the guest is scheduled out if there is no > wbinvd exit, or > 2: execute wbinvd on all dirty physical CPUs when guest wbinvd exits. > > For wbinvd VMExit capable processors, we issue IPIs to all physical CPUs to > do wbinvd, for we can't easily tell which physical CPUs are "dirty". wbinvd is a heavy weapon in the hands of a guest. Even if it is limited to pass-through scenarios, do we really need to bother all physical host CPUs with potential multi-millisecond stalls? Think of VMs only running on a subset of CPUs (e.g. to isolate latency sources). I would suggest to track the physical CPU usage of VCPUs between two wbinvd requests and only send the wbinvd IPI to that set. Also, I think the code is still too much vmx-focused. Only the trapping should be vendor specific, the rest generic. Jan -- Siemens AG, Corporate Technology, CT T DE IT 1 Corporate Competence Center Embedded Linux -- 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
