On Sat, Aug 15, 2020 at 10:19:12AM +0800, Peng Liang wrote: > On 8/13/2020 8:21 PM, Andrew Jones wrote: > > On Thu, Aug 13, 2020 at 06:26:50PM +0800, Peng Liang wrote: > >> The implementation of CPUClass::parse_features only supports CPU > >> features in "feature=value" format. However, libvirt maybe send us a > >> CPU feature string in "+feature/-feature" format. > > > > Why would libvirt do that? If we have a boolean feature, then I'd think > > libvirt should be capable of sending feature=on/off just as easily as a > > +/-feature. Indeed libvirt seems to do the right thing with all the > > CPU features we already have: pmu, sve*, ... > > > > Thanks, > > drew > > > > . > > > > Libvirt before d47db7b16dd5422c7e487c8c8ee5b181a2f9cd66 ("qemu: command: > Support new cpu feature argument syntax") will send +/-feature. Does QEMU > need to support it? If not, I'll remote it.
No, we don't need to support pre-2016 libvirt, at least not for ARM CPU features, because libvirt didn't start querying QEMU for ARM CPU features until just last year. Thanks, drew