[PATCH v4 00/26] Nested SVM fixes, cleanups, and hardening

Yosry Ahmed posted 26 patches 3 weeks, 2 days ago
There is a newer version of this series
arch/x86/include/asm/svm.h                    |  96 ++--
arch/x86/kvm/svm/nested.c                     | 540 +++++++++++-------
arch/x86/kvm/svm/sev.c                        |   4 +-
arch/x86/kvm/svm/svm.c                        |  68 +--
arch/x86/kvm/svm/svm.h                        |  51 +-
arch/x86/kvm/x86.c                            |   3 +
tools/testing/selftests/kvm/Makefile.kvm      |   1 +
.../selftests/kvm/include/x86/processor.h     |   5 +
tools/testing/selftests/kvm/include/x86/svm.h |  14 +-
tools/testing/selftests/kvm/lib/x86/svm.c     |   2 +-
.../selftests/kvm/x86/svm_lbr_nested_state.c  | 155 +++++
11 files changed, 636 insertions(+), 303 deletions(-)
create mode 100644 tools/testing/selftests/kvm/x86/svm_lbr_nested_state.c
[PATCH v4 00/26] Nested SVM fixes, cleanups, and hardening
Posted by Yosry Ahmed 3 weeks, 2 days ago
A group of semi-related fixes, cleanups, and hardening patches for nSVM.
This series doubled in size between v2 and v3, but it's Sean's fault for
finding more bugs that needed fixing.

The series is essentially a group of related mini-series stitched
together for syntactic and semantic dependencies. The first 19 patches
(except patch 3) are all optimistically CC'd to stable as they are fixes
or refactoring leading up to bug fixes. Although I am not sure how much
of that will actually apply to stable trees.

Patches 1-3 here are v2 of the last 3 patches in the LBRV fixes series
[1]. The first 3 patches of [1] are already upstream.

Patches 4-14 are fixes for failure handling in the nested VMRUN and
#VMEXIT code paths, ending with a nice unified code path for handling
VMRUN failures as suggested by Sean. Within this block, patches 7-12 are
refactoring needed for patches 13-14.

Patches 15-19 are fixes for missing or made-up consistency checks.

Patches 20-22 are renames and cleanups.

Patches 23-26 add hardening to reading the VMCB12, caching all used
fields in the save area to prevent theoritical TOC-TOU bugs, sanitizing
used fields in the control area, and restricting accesses to the VMCB12
through guest memory.

v3 -> v4:
- Proerply leave guest mode on failed VMRUN [2].
- Fixed baseline and rebased on top of kvm-x86/next.

v3: https://lore.kernel.org/kvm/20251215192722.3654335-1-yosry.ahmed@linux.dev/

[1]https://lore.kernel.org/kvm/20251108004524.1600006-1-yosry.ahmed@linux.dev/
[2]https://lore.kernel.org/kvm/timpvklyyl5juo5ajjzuxwazc5w2t6ffcx7llnv6f2a5qzot3b@hnj3wqwtla6c/

Yosry Ahmed (26):
  KVM: SVM: Switch svm_copy_lbrs() to a macro
  KVM: SVM: Add missing save/restore handling of LBR MSRs
  KVM: selftests: Add a test for LBR save/restore (ft. nested)
  KVM: nSVM: Always inject a #GP if mapping VMCB12 fails on nested VMRUN
  KVM: nSVM: Triple fault if mapping VMCB12 fails on nested #VMEXIT
  KVM: nSVM: Triple fault if restore host CR3 fails on nested #VMEXIT
  KVM: nSVM: Drop nested_vmcb_check_{save/control}() wrappers
  KVM: nSVM: Call enter_guest_mode() before switching to VMCB02
  KVM: nSVM: Make nested_svm_merge_msrpm() return an errno
  KVM: nSVM: Call nested_svm_merge_msrpm() from enter_svm_guest_mode()
  KVM: nSVM: Call nested_svm_init_mmu_context() before switching to
    VMCB02
  KVM: nSVM: Refactor minimal #VMEXIT handling out of
    nested_svm_vmexit()
  KVM: nSVM: Unify handling of VMRUN failures with proper cleanup
  KVM: nSVM: Clear EVENTINJ field in VMCB12 on nested #VMEXIT
  KVM: nSVM: Drop the non-architectural consistency check for NP_ENABLE
  KVM: nSVM: Add missing consistency check for nCR3 validity
  KVM: nSVM: Add missing consistency check for hCR0.PG and NP_ENABLE
  KVM: nSVM: Add missing consistency check for EFER, CR0, CR4, and CS
  KVM: nSVM: Add missing consistency check for event_inj
  KVM: SVM: Rename vmcb->nested_ctl to vmcb->misc_ctl
  KVM: SVM: Rename vmcb->virt_ext to vmcb->misc_ctl2
  KVM: SVM: Use BIT() and GENMASK() for definitions in svm.h
  KVM: nSVM: Cache all used fields from VMCB12
  KVM: nSVM: Restrict mapping VMCB12 on nested VMRUN
  KVM: nSVM: Sanitize control fields copied from VMCB12
  KVM: nSVM: Only copy NP_ENABLE from VMCB01's misc_ctl

 arch/x86/include/asm/svm.h                    |  96 ++--
 arch/x86/kvm/svm/nested.c                     | 540 +++++++++++-------
 arch/x86/kvm/svm/sev.c                        |   4 +-
 arch/x86/kvm/svm/svm.c                        |  68 +--
 arch/x86/kvm/svm/svm.h                        |  51 +-
 arch/x86/kvm/x86.c                            |   3 +
 tools/testing/selftests/kvm/Makefile.kvm      |   1 +
 .../selftests/kvm/include/x86/processor.h     |   5 +
 tools/testing/selftests/kvm/include/x86/svm.h |  14 +-
 tools/testing/selftests/kvm/lib/x86/svm.c     |   2 +-
 .../selftests/kvm/x86/svm_lbr_nested_state.c  | 155 +++++
 11 files changed, 636 insertions(+), 303 deletions(-)
 create mode 100644 tools/testing/selftests/kvm/x86/svm_lbr_nested_state.c


base-commit: f62b64b970570c92fe22503b0cdc65be7ce7fc7c
-- 
2.52.0.457.g6b5491de43-goog
Re: [PATCH v4 00/26] Nested SVM fixes, cleanups, and hardening
Posted by Sean Christopherson 1 day, 13 hours ago
On Thu, Jan 15, 2026, Yosry Ahmed wrote:
> Yosry Ahmed (26):
>   KVM: SVM: Switch svm_copy_lbrs() to a macro
>   KVM: SVM: Add missing save/restore handling of LBR MSRs
>   KVM: selftests: Add a test for LBR save/restore (ft. nested)
>   KVM: nSVM: Always inject a #GP if mapping VMCB12 fails on nested VMRUN
>   KVM: nSVM: Triple fault if mapping VMCB12 fails on nested #VMEXIT
>   KVM: nSVM: Triple fault if restore host CR3 fails on nested #VMEXIT
>   KVM: nSVM: Drop nested_vmcb_check_{save/control}() wrappers
>   KVM: nSVM: Call enter_guest_mode() before switching to VMCB02
>   KVM: nSVM: Make nested_svm_merge_msrpm() return an errno
>   KVM: nSVM: Call nested_svm_merge_msrpm() from enter_svm_guest_mode()
>   KVM: nSVM: Call nested_svm_init_mmu_context() before switching to
>     VMCB02
>   KVM: nSVM: Refactor minimal #VMEXIT handling out of
>     nested_svm_vmexit()
>   KVM: nSVM: Unify handling of VMRUN failures with proper cleanup
>   KVM: nSVM: Clear EVENTINJ field in VMCB12 on nested #VMEXIT
>   KVM: nSVM: Drop the non-architectural consistency check for NP_ENABLE
>   KVM: nSVM: Add missing consistency check for nCR3 validity
>   KVM: nSVM: Add missing consistency check for hCR0.PG and NP_ENABLE
>   KVM: nSVM: Add missing consistency check for EFER, CR0, CR4, and CS
>   KVM: nSVM: Add missing consistency check for event_inj
>   KVM: SVM: Rename vmcb->nested_ctl to vmcb->misc_ctl
>   KVM: SVM: Rename vmcb->virt_ext to vmcb->misc_ctl2
>   KVM: SVM: Use BIT() and GENMASK() for definitions in svm.h
>   KVM: nSVM: Cache all used fields from VMCB12
>   KVM: nSVM: Restrict mapping VMCB12 on nested VMRUN
>   KVM: nSVM: Sanitize control fields copied from VMCB12
>   KVM: nSVM: Only copy NP_ENABLE from VMCB01's misc_ctl

All in all, looks good.  A few comments, but I don't anticipate a big jump in
the patch count :-)

Note, make sure to rebase on the latest kvm-x86 next, there are a handful of
minor conflicts.