On 2021-01-25 05:49, Alex Shi wrote:
Hi Sean,
Thanks a lot for the detailed explanations.
Yes, they are all meaningful variables on x86. But for more archs,
guess the lock
issue on different arch's guest are similar. Maybe a abstraction on
the point would
be very helpful. Any comments?
It depends on what you have in mind. The whole "yield to another
vcpu" mechanism is already abstracted behind kvm_vcpu_on_spin(),
which at arm64, s390 and x86 are using. You could investigate
whether this is a valid approach for Power and MIPS.
Thanks,
M.
--
Jazz is not dead. It just smells funny...
