Skip to content

KVM: x86/hyperv: Fix racy usage of vcpu->arch.hyperv

Sean Christopherson kvm lkml 4 versions – Applied

Status from this message: Applied to kvm-x86 misc, thanks!

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v4 Latest11 patches 7 repliesApplied
  2. v310 patches 12 repliesSuperseded
  3. v28 patches 6 repliesSuperseded
Earlier versions · 1
VersionPostedPatchesRepliesStatus
v15 patches2 repliesSuperseded

Changes in v4 from the cover letter

- Route non-timer hypercalls at the very beginning of kvm_xen_hcall_vcpu_op() so that KVM doesn't unintentionally resume the guest on bad input (which might not even be bad since KVM would misinterpet the input). [Sashiko]
- Suck less at unwinding on failure, i.e. invalidate the vcpu_idx on any failure during vCPU creation. [Sashiko, syzbot]
- Reference U32_MAX, not -1u. [David]
- Add a compile-time assert to ensure XEN_VCPU_ID_INVALID can't collide with KVM's range of legal values.

Patches in v4 Applied

#Subject and filesSizeRepliesApplied
1/11KVM: x86/hyperv: Get target FIFO in hv_tlb_flush_enqueue(), not callerarch/x86/kvm/hyperv.c+9 -150 replies
2/11KVM: x86/hyperv: Check for NULL vCPU Hyper-V object in kvm_hv_get_tlb_flush_fifo()arch/x86/kvm/hyperv.c arch/x86/kvm/hyperv.h+11 -71 reply
3/11KVM: x86/hyperv: Ensure vCPU's Hyper-V object is initialized on cross-vCPU accessesarch/x86/kvm/hyperv.c arch/x86/kvm/hyperv.h+33 -80 replies
4/11KVM: x86/xen: Always route non-singleshot-timer vCPU hypercalls to userspacearch/x86/kvm/xen.c+8 -91 reply
5/11KVM: x86/xen: Consolidate checks on Xen vCPU ID for singleshot timer hypercallsarch/x86/kvm/xen.c+11 -91 reply
6/11KVM: x86/xen: Punt singleshot timer hcalls to userspace if Xen vCPU ID isn't setarch/x86/kvm/xen.c+7 -10 replies
7/11KVM: Initialize a vCPU's index to '-1' while it's being createdvirt/kvm/kvm_main.c+8 -01 reply
8/11KVM: Move nVMX's lockdep logic for vcpu->mutex to a common helperarch/x86/kvm/vmx/nested.h include/linux/kvm_host.h+8 -41 reply
9/11KVM: x86: Treat a vCPU as unreachable if its index is invalidinclude/linux/kvm_host.h+1 -01 reply
10/11KVM: x86/hyperv: Assert vCPU's mutex is held in to_hv_vcpu()arch/x86/kvm/hyperv.c arch/x86/kvm/hyperv.h+4 -40 replies
11/11KVM: x86/hyperv: Use {READ,WRITE}_ONCE for cross-task synic->active accessesarch/x86/kvm/hyperv.c+2 -20 replies

Reviews trailers per patch

7 of 11 patches have a Reviewed-by or Acked-by.

Patch[email protected]David WoodhousePhilippe Mathieu-Daudé
1/11
2/11R-b
3/11T-b
4/11R-b
5/11R-b R-b
6/11R-b
7/11R-b
8/11R-b
9/11R-b
10/11
11/11

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes. Scroll the table sideways for all 3 reviewers.

Read the v4 thread

Cover letter [PATCH v4 00/11] KVM: x86/hyperv: Fix racy usage of vcpu->arch.hyperv

Fix a bug found by syzkaller (originally on a Google-internal kernel, but now on upstream as well) where KVM consumes a vCPU's HyperV structure before it's fully initialized, by concurrently triggering PV TLB flushes (queues flushes into a vCPU's FIFO without holding the vCPU's mutex) on a vCPU that is in the process of activating HyperV.

Harden against similar bugs by asserting the vcpu->mutex is held when using the "normal" to_hv_vcpu(), same as we did for get_vmcs12() and get_shadow_vmcs12() (also in response to cross-task races). To avoid false positives when creating a vCPU, initialize vcpu_idx to -1, and treat the vCPU as unreachable (other than the caller, obviously) if its index is -1.

v4:
- Route non-timer hypercalls at the very beginning of kvm_xen_hcall_vcpu_op() so that KVM doesn't unintentionally resume the guest on bad input (which might not even be bad since KVM would misinterpet the input). [Sashiko]
- Suck less at unwinding on failure, i.e. invalidate the vcpu_idx on any failure during vCPU creation. [Sashiko, syzbot]
- Reference U32_MAX, not -1u. [David]
- Add a compile-time assert to ensure XEN_VCPU_ID_INVALID can't collide with KVM's range of legal values.

v3:
- https://lore.kernel.org/all/[email protected]
- Reset vcpu_idx back to -1 if adding the vCPU to the xarray fails. [syzbot]
- Use the safe accessor in kvm_hv_has_stimer_pending(). [sashiko]
- Explicitly initialize vcpu->arch.xen.vcpu_id to XEN_VCPU_ID_INVALID, and punt singleshot timer hypercalls to userspace if the vCPU ID hasn't been set. [sashiko, David]

v2:
- https://lore.kernel.org/all/[email protected]
- Init vcpu->vcpu_idx to -1, use that as a canary to detect the vCPU is unreachable, and allow accessing Hyper-V state if the vCPU is otherwise unreachable. [syzbot]

v1: https://lore.kernel.org/all/[email protected]

Sean Christopherson (11):
KVM: x86/hyperv: Get target FIFO in hv_tlb_flush_enqueue(), not caller
KVM: x86/hyperv: Check for NULL vCPU Hyper-V object in

    kvm_hv_get_tlb_flush_fifo()

KVM: x86/hyperv: Ensure vCPU's Hyper-V object is initialized on
cross-vCPU accesses
KVM: x86/xen: Always route non-singleshot-timer vCPU hypercalls to
userspace
KVM: x86/xen: Consolidate checks on Xen vCPU ID for singleshot timer
hypercalls
KVM: x86/xen: Punt singleshot timer hcalls to userspace if Xen vCPU ID
isn't set
KVM: Initialize a vCPU's index to '-1' while it's being created
KVM: Move nVMX's lockdep logic for vcpu->mutex to a common helper
KVM: x86: Treat a vCPU as unreachable if its index is invalid
KVM: x86/hyperv: Assert vCPU's mutex is held in to_hv_vcpu()
KVM: x86/hyperv: Use {READ,WRITE}_ONCE for cross-task synic->active
accesses

 arch/x86/kvm/hyperv.c     | 64 +++++++++++++++++++++------------------
 arch/x86/kvm/hyperv.h     | 27 ++++++++++++++---
 arch/x86/kvm/vmx/nested.h |  6 ++--
 arch/x86/kvm/xen.c        | 43 +++++++++++++++-----------
 include/linux/kvm_host.h  |  7 +++++
 virt/kvm/kvm_main.c       |  8 +++++
 6 files changed, 99 insertions(+), 56 deletions(-)

base-commit: a204badd8432f93b7e862e7dac6db0fe3d65f370

Signature · 1 line
2.55.0.rc0.799.gd6f94ed593-goog