On Thu, 2021-10-28 at 16:11 +0200, Paolo Bonzini wrote: > > > There are potentially other ways of doing this, by shuffling the tail > > of kvm_arch_vcpu_ioctl_run() around a little and holding the lock once > > there instead of taking it within vcpu_run(). But the call to > > post_kvm_run_save() occurs even on the error paths, and it gets complex > > to untangle. This is the simple approach. > > > > This doesn't show up when I enable event channel delivery in > > xen_shinfo_test because we have pic_in_kernel() there. I'll fix the > > tests not to hardcode that, as I expand the event channel support and > > add more testing of it. > > Queued for 5.15, thanks. Thanks. Sean, since you actually went ahead and implemented the alternative approach along the lines I describe above, I'll let you submit that as a subsequent cleanup while keeping the simple version for stable as discussed?
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
