[0/6] KVM: nSVM: few random fixes

Message

Maxim Levitsky Jan. 6, 2021, 10:49 a.m. UTC

This is a series of fixes to nested SVM, that finally makes my kvm on kvm
stress test pass, and fix various other issues/regressions.

Patches 1-2 are a fix for disappearing interrupts in L2 on migration which
usually make the L2 hang.
Same issue happens on VMX and WIP, patches for this will be sent in a separate
series.
Paulo helped me to find the root cause of this issue.

Note that this patch likely breaks a nested guest that uses software interrupt
injections (SVM_EXITINTINFO_TYPE_SOFT) because currently kvm ignores these
on SVM.

Patch 3 is a fix for recent regression related to code that delayed the nested
msr bitmap processing to the next vm entry, and started to crash the L1 after
my on demand nested state allocation patches.

The problem was that the code assumed that we will still be in the nested
guest mode on next vmentry after setting the nested state, but a pending event
can cause a nested vmexit prior to that.

Patches 4,5,6 are few things I found while reviewing the nested migration code.
I don't have a reproducer for them.

Best regards,
	Maxim Levitsky

Maxim Levitsky (6):
  KVM: SVM: create svm_process_injected_event
  KVM: nSVM: fix for disappearing L1->L2 event injection on L1 migration
  KVM: nSVM: cancel KVM_REQ_GET_NESTED_STATE_PAGES on nested vmexit
  KVM: nSVM: correctly restore nested_run_pending on migration
  KVM: nSVM: always leave the nested state first on KVM_SET_NESTED_STATE
  KVM: nSVM: mark vmcb as dirty when forcingly leaving the guest mode

 arch/x86/kvm/svm/nested.c | 21 ++++++++++++--
 arch/x86/kvm/svm/svm.c    | 58 ++++++++++++++++++++++-----------------
 arch/x86/kvm/svm/svm.h    |  4 +++
 3 files changed, 55 insertions(+), 28 deletions(-)