[PATCH 6.12.y 0/3] Backport nested EPT violation GVA bit fix

[Bjoern Doebel <doebel@amazon.de>]

Backport of commit 96b067b59ad9 ("KVM: VMX: Synthesize nested EPT
violation GVA_IS_VALID/GVA_TRANSLATED bits") to 6.12.y, with its two
dependencies.

Patch 3 is the fix: KVM pulls the GVA_IS_VALID and GVA_TRANSLATED bits
from vmcs02.EXIT_QUALIFICATION even when the fault injected into L1 was
synthesized while emulating an L2 GVA access, i.e. when that field
describes an unrelated VM-Exit.

Patches 1 and 2 are prerequisites, tagged Stable-dep-of; neither changes
behavior on its own.  Patch 1 lets the EPT walker tell a paging-structure
access from a final access, patch 2 plumbs in whether a fault came from
hardware and renames kvm_inject_emulated_page_fault()'s exported symbol
to __kvm_inject_emulated_page_fault().

Context adjustments are noted per patch.  Based on v6.12.113, build tested
on x86_64 with CONFIG_KVM_INTEL=m and CONFIG_KVM_AMD=m.  Not runtime
tested under nested virtualization.

Kevin Cheng (1):
  KVM: VMX: Synthesize nested EPT violation GVA_IS_VALID/GVA_TRANSLATED
    bits

Paolo Bonzini (1):
  KVM: x86/mmu: pass PFERR_GUEST_PAGE/FINAL_MASK to kvm_translate_gpa

Sean Christopherson (1):
  KVM: x86: Tell ->inject_page_fault() whether or a fault came from
    hardware

 arch/x86/include/asm/kvm_host.h | 18 ++++++++++++++----
 arch/x86/kvm/hyperv.c           |  3 ++-
 arch/x86/kvm/mmu/mmu.c          |  3 ++-
 arch/x86/kvm/mmu/paging_tmpl.h  | 26 ++++++++++++++++++++++----
 arch/x86/kvm/regs.c             |  3 ++-
 arch/x86/kvm/svm/nested.c       |  3 ++-
 arch/x86/kvm/vmx/nested.c       | 26 +++++++++++++++++++++-----
 arch/x86/kvm/x86.c              | 16 +++++++++-------
 8 files changed, 74 insertions(+), 24 deletions(-)

-- 
2.50.1