Here is an update to the SIGP handling I brought up a few weeks ago [1]. It doesn't share anything with v1, as one of the patches has already been pulled to master [2] and the other is now dropped in favor of this direction.
Here, we take advantage of a new KVM ioctl [3] to indicate when userspace has completed processing a particular SIGP order, based on David's suggestion [4]. This code resets the "busy" situation for every order, regardless of whether the Principles of Operation dictates whether the order would mark the vcpu busy. It's not as graceful as I'd like, but it does what I expect it to. Maybe some more refactoring on this side would help matters. Thoughts? [1] https://lore.kernel.org/r/20211008203811.1980478-1-far...@linux.ibm.com/ [2] https://lore.kernel.org/r/20211015091622.1302433-1-th...@redhat.com/ [3] https://lore.kernel.org/r/20211102194652.2685098-1-far...@linux.ibm.com/ [4] https://lore.kernel.org/r/3e3b38d1-b338-0211-04ab-91f913c1f...@redhat.com/ Eric Farman (2): Temporary linux-headers update s390x: Implement the USER_SIGP_BUSY capability linux-headers/linux/kvm.h | 4 ++++ target/s390x/cpu-sysemu.c | 7 +++++++ target/s390x/cpu.h | 4 ++++ target/s390x/kvm/kvm.c | 9 ++++++++ target/s390x/kvm/kvm_s390x.h | 1 + target/s390x/sigp.c | 40 ++++++++++++++++++++++++++++++++++++ 6 files changed, 65 insertions(+) -- 2.25.1
