On Fri, 2020-11-27 at 06:21 -0500, Paolo Bonzini wrote: > This is my take on the split irqchip bug that David reported. It's a > much more complicated patch, but I think it really gets to the bottom > of the issue and the code is clearer. Looks good to me; thanks. With the exception of the stray apostrophe already noted, both: Reviewed-by: David Woodhouse Tested-by: David Woodhouse There is a slight caveat that I think we're accepting ExtINT from userspace PIC sooner than we should, and queuing it up for delivery in the kernel. Whereas real hardware might not issue the INTA cycle to ask the PIC what vector it wants until later. By which time the PIC might give a different answer. But I don't think anyone cares, even if it can make an observable difference in practice. And I'm not sure that it's new with your patches either.