Add dedicated helpers to emulate RESET (or at least, the parts of RESET
that KVM actually emulates) instead of having the relevant code scattered
throughout vcpu_create() and vcpu_reset().
Vitaly's innocuous suggestion to WARN on non-zero CR0 led to a wonderful
bit of git archaeology after, to my complete surprise, it fired on RESET
due to fx_init() setting CR0.ET (and XCR0.FP).
v2:
- Add patches to drop defunct fx_init() code.
- Add patch to zero CR3 at RESET/INIT.
- Add patch to mark all regs avail/dirty at creation, not at RESET/INIT.
- Add patch to WARN if CRs are non-zero at RESET. [Vitaly]
v1: https://lkml.kernel.org/r/20210914230840.3030620-1-seanjc@xxxxxxxxxx
Sean Christopherson (10):
KVM: x86: Mark all registers as avail/dirty at vCPU creation
KVM: x86: Clear KVM's cached guest CR3 at RESET/INIT
KVM: x86: Do not mark all registers as avail/dirty during RESET/INIT
KVM: x86: Remove defunct setting of CR0.ET for guests during vCPU
create
KVM: x86: Remove defunct setting of XCR0 for guest during vCPU create
KVM: x86: Fold fx_init() into kvm_arch_vcpu_create()
KVM: VMX: Drop explicit zeroing of MSR guest values at vCPU creation
KVM: VMX: Move RESET emulation to vmx_vcpu_reset()
KVM: SVM: Move RESET emulation to svm_vcpu_reset()
KVM: x86: WARN on non-zero CRs at RESET to detect improper
initalization
arch/x86/kvm/svm/sev.c | 6 ++--
arch/x86/kvm/svm/svm.c | 29 ++++++++++--------
arch/x86/kvm/svm/svm.h | 2 +-
arch/x86/kvm/vmx/vmx.c | 68 ++++++++++++++++++++----------------------
arch/x86/kvm/x86.c | 44 +++++++++++++--------------
5 files changed, 76 insertions(+), 73 deletions(-)