s390x/kvm: call cpu_synchronize_state() on every kvm_arch_handle_exit()
commiteac53ac598cc43e325658a6889d9951fd7244e11
authorDavid Hildenbrand <david@redhat.com>
Fri, 6 Apr 2018 09:35:52 +0000 (6 11:35 +0200)
committerCornelia Huck <cohuck@redhat.com>
Mon, 9 Apr 2018 11:50:31 +0000 (9 13:50 +0200)
treefdaf3dfce8ab3dc5ccce72497e7c0f9087c7d520
parent2a6bcfdebe4115993a032395d459f5e0cf27a01e
s390x/kvm: call cpu_synchronize_state() on every kvm_arch_handle_exit()

Manually having to use cpu_synchronize_state() is error prone. And as
Christian Borntraeger discovered, e.g. handle_diag() is currently
missing a cpu_synchronize_state(), as decode_basedisp_s() uses a
general purpose register value internally.

So let's do an overall cpu_synchronize_state(), which fixes at least the
one mentioned BUG. We will clean up the superfluous cpu_synchronize_state()
calls later.

We now also call it (although maybe not neded) for
- KVM_EXIT_S390_RESET -> s390_reipl_request()
- KVM_EXIT_DEBUG -> kvm_arch_handle_debug_exit()
- unmanagable/unimplemented intercepts
- ICPT_CPU_STOP -> do_stop_interrupt() -> cpu gets halted
- Scenarios where we inject an operation exception
- handle_stsi()

I don't think any of these are performance critical. Especially as we
have all information directly contained in kvm_run, there are no
additional IOCTLs to issue on modern kernels.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180406093552.13016-1-david@redhat.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
target/s390x/kvm.c