Currently, use hypercall instruction in guest cpl3 would just skip
the instruction, however, that behaviour could trigger a exception
in Linux host. It is reasonable for hypervisor to inject a exception,
especially in nested guest, L1 guest could behaviour like host.
As for hypercall instruction emulation, hypervisor would replace
the wrong instruction with the right instruction instead of the real
instruction emulation. It's guest's responsibility to use the right
instruction, hypervisor could emulate it but shouldn't modify it
without guest's request. At present, Linux guest could use alternative
to choose right instruction, and hyperv guest could use hypercall to
modify instruction. So just do the real instruction emualtion job
for em_hypercall().
Hou Wenlong (3):
kvm: x86: Introduce hypercall x86 ops for handling hypercall not in
cpl0
kvm: x86: Refactor kvm_emulate_hypercall() to no skip instruction
kvm: x86: Emulate hypercall instead of fixing hypercall instruction
arch/x86/include/asm/kvm-x86-ops.h | 1 +
arch/x86/include/asm/kvm_host.h | 1 +
arch/x86/kvm/emulate.c | 20 +++++------
arch/x86/kvm/kvm_emulate.h | 2 +-
arch/x86/kvm/svm/svm.c | 5 +++
arch/x86/kvm/vmx/vmx.c | 9 +++++
arch/x86/kvm/x86.c | 55 +++++++++++++++++-------------
7 files changed, 58 insertions(+), 35 deletions(-)
--
2.31.1
