Re: [GIT PULL] KVM: Fixes for 7.4-rcN
From: Paolo Bonzini
Date: Fri Nov 20 2026 - 04:27:10 EST
On Thu, Nov 19, 2026 at 9:18 PM Sean Christopherson <seanjc@xxxxxxxxxx> wrote:
>
> Please pull a somewhat random smattering of fixes. The biggest (but probably
> not scariest?) is to rework single-vCPU kicks to ensure waiting requests
> actually wait. It's big only because PPC is weird and required refactoring
> a bunch of code just to avoid throwing in more hacks. Including this many
> patches obviously isn't ideal, but failing to wait breaks the flow to force
> vCPUs to invalidate their SNP VMSA. Prior to 7.4 the race condition was all
> but impossible to hit, but in-place conversion makes it a *lot* easier to hit
> the window, which is why I included the fix here.
>
> Speaking of PPC, the scariest fix for me is the patch to actually implement
> kvm_arch_flush_shadow_all(). Some PPC folks reviewed it, so it's probably
> good? But I'm still flabbergasted that such a blatant bug has gone unnoticed
> for so long, which makes we worried I'm missing something.
Pulled, thanks. I'll give it another round of reviews...
Paolo
> The following changes since commit 7475ea8b58616b33c04afb976cf0fbaad9dd63fb:
>
> Linux 7.4-rc2 (2026-11-08 16:22:20 -0800)
>
> are available in the Git repository at:
>
> https://github.com/kvm-x86/linux.git tags/kvm-x86-fixes-7.4-rc4
>
> for you to fetch changes up to bcc359264f15095d3d167535cf5de0b7e4800279:
>
> KVM: nVMX: Clear nested.smm.guest_mode on a failed RSM-to-L2 (2026-11-10 14:42:26 -0800)
>
> ----------------------------------------------------------------
> KVM fixes for 7.4-rcN
>
> - Mitigate a class of (x86) bugs where KVM can read/write guest (userspace)
> memory using the wrong mm, i.e. using the wrong user page tables.
>
> - Don't force a full nested VMX VM-Exit when destroying a vCPU, as processing
> the MSR load/store lists and writing back the shadow vmcs12 is wildly wrong
> when VM is dying.
>
> - Implement "flush shadow all" on KVM PPC to ensure KVM's mappings are torn
> down when the VM and/or owning process is dying.
>
> - Provide empty installation rules for architectures that aren't supported by
> KVM selftests to fix build issues in CI environments.
>
> - Fix KVM selftests' rules for generating arm64's sysreg headers so that the
> headers are properly (re)generated on incremental builds.
>
> - Clear only the VMCB save area on an intercepted SHUTDOWN on SVM, partly to
> avoid unnecessary churn, but mostly to prevent KVM from clobbering the PML
> index.
>
> - Update the tracked pCPU on task migration for both vmcb01 and vmcb02 to fix
> an edge case where KVM can use a stale ASID across pCPUs.
>
> - Rework the single-vCPU kick logic to fix a bug where KVM would fail to wait
> for a target vCPU to respond if another in-flight, non-waiting kick changed
> the vCPU's mode to EXITING_GUEST_MODE.
>
> - Clear nVMX's smm.guest_most when freeing nested state, and if a RSM to L2
> ultimately fails (and queues a TRIPLE_FAULT to L1).
> ----------------------------------------------------------------
> Amirmohammad Eftekhar (1):
> KVM: nVMX: Clear nested.smm.guest_mode on a failed RSM-to-L2
>
> Fuad Tabba (1):
> KVM: selftests: Fix arm64 sysreg header dependencies
>
> Josef Bacik (1):
> KVM: nVMX: Clear nested.smm.guest_mode when freeing nested state
>
> Sasha Levin (1):
> KVM: selftests: Provide empty installation rules on unsupported ARCHs
>
> Sean Christopherson (10):
> KVM: Reject user accesses to guest memory if current->mm != kvm->mm
> KVM: nVMX: Don't force a full nested VM-Exit when destroying a vCPU
> KVM: PPC: Flush/zap all memslots on kvm_arch_flush_shadow_all()
> KVM: s390: Drop unused and unreachable kvm_arch_vcpu_should_kick()
> KVM: Add a @force param to __kvm_vcpu_kick() for PPC backwards compatibility
> KVM: PPC: Explicitly force IPIs when directly kicking vCPUs
> KVM: MIPS: Rework kvm_arch_vcpu_should_kick() to use standard implementation
> KVM: Move kvm_arch_vcpu_should_kick() to common code as kvm_vcpu_should_kick()
> KVM: Plumb the full request into __kvm_vcpu_kick()
> KVM: Drop kvm_vcpu_should_kick() and use kvm_request_needs_ipi()
>
> Shivansh Dhiman (1):
> KVM: SVM: Clear VMCB save area instead of entire VMCB on shutdown intercept
>
> Yosry Ahmed (1):
> KVM: SVM: Update tracked pCPU on pCPU switch
>
> arch/arm64/kvm/arm.c | 5 ---
> arch/loongarch/kvm/vcpu.c | 5 ---
> arch/mips/kvm/mips.c | 5 ---
> arch/powerpc/include/asm/kvm_host.h | 1 -
> arch/powerpc/include/asm/kvm_ppc.h | 13 +++++++-
> arch/powerpc/kvm/book3s.c | 4 +--
> arch/powerpc/kvm/book3s_pr.c | 2 +-
> arch/powerpc/kvm/booke.c | 6 ++--
> arch/powerpc/kvm/e500_emulate.c | 2 +-
> arch/powerpc/kvm/powerpc.c | 18 +++++++----
> arch/riscv/kvm/vcpu.c | 5 ---
> arch/s390/kvm/s390/s390.c | 7 -----
> arch/x86/kvm/svm/svm.c | 43 ++++++++++++++++++--------
> arch/x86/kvm/vmx/nested.c | 10 +++++-
> arch/x86/kvm/vmx/sgx.c | 2 +-
> arch/x86/kvm/vmx/vmx.c | 15 ++++++---
> arch/x86/kvm/x86.c | 5 ---
> include/linux/kvm_host.h | 32 ++++++++++++++++----
> tools/testing/selftests/kvm/Makefile | 4 +--
> tools/testing/selftests/kvm/Makefile.kvm | 16 +++++++---
> virt/kvm/kvm_main.c | 52 +++++++++++++++++---------------
> 21 files changed, 149 insertions(+), 103 deletions(-)
>