[GIT PULL] KVM: x86 (and a PPC) fixes for 7.0-rc3+

[GIT PULL] KVM: x86 (and a PPC) fixes for 7.0-rc3+

[Sean Christopherson]

Please pull a larger-than-normal set of fixes for 7.0.  The PMU fix from
Namhyung is the only one that absolutely _needs_ to get into 7.0 (only because
no one cares about PPC e500).  The rest are fairly urgent stable@ fixes (and
most of 'em _just_ missed the merge window).

The following changes since commit 11439c4635edd669ae435eec308f4ab8a0804808:

  Linux 7.0-rc2 (2026-03-01 15:39:31 -0800)

are available in the Git repository at:

  https://github.com/kvm-x86/linux.git tags/kvm-x86-fixes-7.0-rc3

for you to fetch changes up to 3271085a7f1030ba9897cab60452acd370423ff1:

  KVM: PPC: e500: Rip out "struct tlbe_ref" (2026-03-03 12:29:07 -0800)

----------------------------------------------------------------
KVM x86 (and PPC e500) fixes for 7.0-rc3+

 - Synthesize features into KVM's set of supported CPU caps if and only if the
  feature is actually "enabled" by the kernel.

 - Validate _all_ GVAs when processing a range of GVAs for a PV Hyper-V TLB
   flush, not just the first GVA.

 - Fix a brown paper bug add_atomic_switch_msr().

 - Use hlist_for_each_entry_srcu() when traversing mask_notifier_list to fix a
   lockdep warning (KVM doesn't hold RCU, just irq_srcu).

 - Ensure AVIC VMCB fields are initialized if the VM has an in-kernel local
   APIC (and AVIC is enabled at the module level).

 - Update CR8 write interception when AVIC is (de)activated to fix a bug where
   KVM can run with the CR8 intercept in perpetuity.

 - Add a quirk to skip the consistency check on FREEZE_IN_SMM, i.e. to allow L1
   hypervisors to set FREEZE_IN_SMM to provide some amount of backwards
   compatibility with setups where L1 was freezing PMCs on VM-Entry to L2.

 - Increase the maximum number of NUMA nodes in the guest_memfd selftest to
   64 (from 8).

 - Fix a PPC e500 build error due to a long-standing wart that was exposed by
   the recent conversion to kmalloc_obj(), and rip out the underlying ugliness
   that led to the wart.

----------------------------------------------------------------
Carlos López (1):
      KVM: x86: synthesize CPUID bits only if CPU capability is set

Jim Mattson (1):
      KVM: x86: Introduce KVM_X86_QUIRK_VMCS12_FREEZE_IN_SMM_CC

Kai Huang (1):
      KVM: selftests: Increase 'maxnode' for guest_memfd tests

Li RongQing (1):
      KVM: x86: Fix SRCU list traversal in kvm_fire_mask_notifiers()

Manuel Andreas (1):
      KVM: x86: hyper-v: Validate all GVAs during PV TLB flush

Namhyung Kim (1):
      KVM: VMX: Fix a wrong MSR update in add_atomic_switch_msr()

Sean Christopherson (4):
      KVM: SVM: Initialize AVIC VMCB fields if AVIC is enabled with in-kernel APIC
      KVM: SVM: Set/clear CR8 write interception when AVIC is (de)activated
      KVM: PPC: e500: Fix build error due to using kmalloc_obj() with wrong type
      KVM: PPC: e500: Rip out "struct tlbe_ref"

 Documentation/virt/kvm/api.rst                 |  8 ++++++++
 arch/powerpc/kvm/e500.h                        |  6 +-----
 arch/powerpc/kvm/e500_mmu.c                    |  4 ++--
 arch/powerpc/kvm/e500_mmu_host.c               | 91 ++++++++++++++++++++++++++++++++++++++++++++-----------------------------------------------
 arch/x86/include/asm/kvm_host.h                |  3 ++-
 arch/x86/include/uapi/asm/kvm.h                |  1 +
 arch/x86/kvm/cpuid.c                           |  5 ++++-
 arch/x86/kvm/hyperv.c                          |  9 +++++----
 arch/x86/kvm/ioapic.c                          |  3 ++-
 arch/x86/kvm/svm/avic.c                        |  8 +++++---
 arch/x86/kvm/svm/svm.c                         | 11 ++++++-----
 arch/x86/kvm/vmx/nested.c                      | 22 ++++++++++++++++++----
 arch/x86/kvm/vmx/vmx.c                         |  2 +-
 tools/testing/selftests/kvm/guest_memfd_test.c |  2 +-
 14 files changed, 100 insertions(+), 75 deletions(-)

Re: [GIT PULL] KVM: x86 (and a PPC) fixes for 7.0-rc3+

[Sean Christopherson]

On Thu, Mar 05, 2026, Sean Christopherson wrote:
>  - Update CR8 write interception when AVIC is (de)activated to fix a bug where
>    KVM can run with the CR8 intercept in perpetuity.

...

> Sean Christopherson (4):
>       KVM: SVM: Initialize AVIC VMCB fields if AVIC is enabled with in-kernel APIC
>       KVM: SVM: Set/clear CR8 write interception when AVIC is (de)activated

The CR8 write patch broke SEV-ES guests.  If you haven't already prepared the
PULL request for Linus, can you apply the fix directly?

Thanks!

https://lore.kernel.org/all/20260310211841.2552361-2-seanjc@google.com

Re: [GIT PULL] KVM: x86 (and a PPC) fixes for 7.0-rc3+

[Paolo Bonzini]

On Tue, Mar 10, 2026 at 10:20 PM Sean Christopherson <seanjc@google.com> wrote:
>
> On Thu, Mar 05, 2026, Sean Christopherson wrote:
> >  - Update CR8 write interception when AVIC is (de)activated to fix a bug where
> >    KVM can run with the CR8 intercept in perpetuity.
>
> ...
>
> > Sean Christopherson (4):
> >       KVM: SVM: Initialize AVIC VMCB fields if AVIC is enabled with in-kernel APIC
> >       KVM: SVM: Set/clear CR8 write interception when AVIC is (de)activated
>
> The CR8 write patch broke SEV-ES guests.  If you haven't already prepared the
> PULL request for Linus, can you apply the fix directly?

Done. I also took the occasion to rename
KVM_X86_QUIRK_VMCS12_ALLOW_FREEZE_IN_SMM because not having a verb was
quite confusing.

More boringly, I changed one of the entries in the tag description like this:

- Fix cases where CPUID mitigation features were incorrectly marked as
available whenever the kernel used scattered feature words for them.

Paolo

>
> Thanks!
>
> https://lore.kernel.org/all/20260310211841.2552361-2-seanjc@google.com
>

Re: [GIT PULL] KVM: x86 (and a PPC) fixes for 7.0-rc3+

[Sean Christopherson]

On Wed, Mar 11, 2026, Paolo Bonzini wrote:
> On Tue, Mar 10, 2026 at 10:20 PM Sean Christopherson <seanjc@google.com> wrote:
> >
> > On Thu, Mar 05, 2026, Sean Christopherson wrote:
> > >  - Update CR8 write interception when AVIC is (de)activated to fix a bug where
> > >    KVM can run with the CR8 intercept in perpetuity.
> >
> > ...
> >
> > > Sean Christopherson (4):
> > >       KVM: SVM: Initialize AVIC VMCB fields if AVIC is enabled with in-kernel APIC
> > >       KVM: SVM: Set/clear CR8 write interception when AVIC is (de)activated
> >
> > The CR8 write patch broke SEV-ES guests.  If you haven't already prepared the
> > PULL request for Linus, can you apply the fix directly?
> 
> Done. 

No?  I see the pull request in kvm.git, but not the below fix-to-the-fix.

> > https://lore.kernel.org/all/20260310211841.2552361-2-seanjc@google.com

Re: [GIT PULL] KVM: x86 (and a PPC) fixes for 7.0-rc3+

[Paolo Bonzini]

On Thu, Mar 12, 2026 at 3:18 PM Sean Christopherson <seanjc@google.com> wrote:
>
> On Wed, Mar 11, 2026, Paolo Bonzini wrote:
> > On Tue, Mar 10, 2026 at 10:20 PM Sean Christopherson <seanjc@google.com> wrote:
> > >
> > > On Thu, Mar 05, 2026, Sean Christopherson wrote:
> > > >  - Update CR8 write interception when AVIC is (de)activated to fix a bug where
> > > >    KVM can run with the CR8 intercept in perpetuity.
> > >
> > > ...
> > >
> > > > Sean Christopherson (4):
> > > >       KVM: SVM: Initialize AVIC VMCB fields if AVIC is enabled with in-kernel APIC
> > > >       KVM: SVM: Set/clear CR8 write interception when AVIC is (de)activated
> > >
> > > The CR8 write patch broke SEV-ES guests.  If you haven't already prepared the
> > > PULL request for Linus, can you apply the fix directly?
> >
> > Done.
>
> No?  I see the pull request in kvm.git, but not the below fix-to-the-fix.

Ah, I had squashed it in directly from the original thread.  I have
now added on top the testcase.

Paolo

Re: [GIT PULL] KVM: x86 (and a PPC) fixes for 7.0-rc3+

[Sean Christopherson]

On Thu, Mar 12, 2026, Paolo Bonzini wrote:
> On Thu, Mar 12, 2026 at 3:18 PM Sean Christopherson <seanjc@google.com> wrote:
> >
> > On Wed, Mar 11, 2026, Paolo Bonzini wrote:
> > > On Tue, Mar 10, 2026 at 10:20 PM Sean Christopherson <seanjc@google.com> wrote:
> > > >
> > > > On Thu, Mar 05, 2026, Sean Christopherson wrote:
> > > > >  - Update CR8 write interception when AVIC is (de)activated to fix a bug where
> > > > >    KVM can run with the CR8 intercept in perpetuity.
> > > >
> > > > ...
> > > >
> > > > > Sean Christopherson (4):
> > > > >       KVM: SVM: Initialize AVIC VMCB fields if AVIC is enabled with in-kernel APIC
> > > > >       KVM: SVM: Set/clear CR8 write interception when AVIC is (de)activated
> > > >
> > > > The CR8 write patch broke SEV-ES guests.  If you haven't already prepared the
> > > > PULL request for Linus, can you apply the fix directly?
> > >
> > > Done.
> >
> > No?  I see the pull request in kvm.git, but not the below fix-to-the-fix.
> 
> Ah, I had squashed it in directly from the original thread.

OMG, -ENOCOFFEE x2.  I even looked at commit 87d0f901a9bd from your tree, and
the lack of a merge commit for fixes should have been a big clue.

> I have now added on top the testcase.

Thanks much!