[v2,0/4] KVM: extract lock_all_vcpus/unlock_all_vcpus

Message

Maxim Levitsky April 9, 2025, 1:41 a.m. UTC

Implement Paolo's suggestion of reusing
sev_lock/unlock_vcpus_for_migration in arm and riscv code
for the purpose of taking vcpu->mutex of all vcpus of a VM.

Because sev_lock/unlock_vcpus_for_migration already have a workaround
for lockdep max lock depth, this fixes the lockdep warnings on arm
which were the inspiration for this refactoring.

This patch series was only compile tested on all 3 architectures.

V2: added trylock option to kvm_lock_all_vcpus to be better compatible
with the orginal code.

Best regards,
	Maxim Levitsky

Maxim Levitsky (4):
  locking/mutex: implement mutex_trylock_nested
  KVM: x86: move sev_lock/unlock_vcpus_for_migration to kvm_main.c
  KVM: arm64: switch to using kvm_lock/unlock_all_vcpus
  RISC-V: KVM: switch to kvm_lock/unlock_all_vcpus

 arch/arm64/include/asm/kvm_host.h     |  3 --
 arch/arm64/kvm/arch_timer.c           |  4 +-
 arch/arm64/kvm/arm.c                  | 43 ----------------
 arch/arm64/kvm/vgic/vgic-init.c       |  4 +-
 arch/arm64/kvm/vgic/vgic-its.c        |  8 +--
 arch/arm64/kvm/vgic/vgic-kvm-device.c | 12 ++---
 arch/riscv/kvm/aia_device.c           | 34 +------------
 arch/x86/kvm/svm/sev.c                | 65 ++----------------------
 include/linux/kvm_host.h              |  6 +++
 include/linux/mutex.h                 |  8 +++
 kernel/locking/mutex.c                | 14 ++++--
 virt/kvm/kvm_main.c                   | 71 +++++++++++++++++++++++++++
 12 files changed, 116 insertions(+), 156 deletions(-)

Comments

Sean Christopherson April 9, 2025, 7:53 p.m. UTC | #1

+Adrian

On Tue, Apr 08, 2025, Maxim Levitsky wrote:
> Implement Paolo's suggestion of reusing

Ha!  I *knew* this felt familiar when I suggested extracting (un)lock_all_vcpus()
to common code in the context of the TDX series.

https://lore.kernel.org/all/[e-mail]

> sev_lock/unlock_vcpus_for_migration in arm and riscv code
> for the purpose of taking vcpu->mutex of all vcpus of a VM.
> 
> Because sev_lock/unlock_vcpus_for_migration already have a workaround
> for lockdep max lock depth, this fixes the lockdep warnings on arm
> which were the inspiration for this refactoring.
> 
> This patch series was only compile tested on all 3 architectures.
> 
> V2: added trylock option to kvm_lock_all_vcpus to be better compatible
> with the orginal code.
> 
> Best regards,
> 	Maxim Levitsky
> 
> Maxim Levitsky (4):
>   locking/mutex: implement mutex_trylock_nested
>   KVM: x86: move sev_lock/unlock_vcpus_for_migration to kvm_main.c
>   KVM: arm64: switch to using kvm_lock/unlock_all_vcpus
>   RISC-V: KVM: switch to kvm_lock/unlock_all_vcpus
> 
>  arch/arm64/include/asm/kvm_host.h     |  3 --
>  arch/arm64/kvm/arch_timer.c           |  4 +-
>  arch/arm64/kvm/arm.c                  | 43 ----------------
>  arch/arm64/kvm/vgic/vgic-init.c       |  4 +-
>  arch/arm64/kvm/vgic/vgic-its.c        |  8 +--
>  arch/arm64/kvm/vgic/vgic-kvm-device.c | 12 ++---
>  arch/riscv/kvm/aia_device.c           | 34 +------------
>  arch/x86/kvm/svm/sev.c                | 65 ++----------------------
>  include/linux/kvm_host.h              |  6 +++
>  include/linux/mutex.h                 |  8 +++
>  kernel/locking/mutex.c                | 14 ++++--
>  virt/kvm/kvm_main.c                   | 71 +++++++++++++++++++++++++++
>  12 files changed, 116 insertions(+), 156 deletions(-)
> 
> -- 
> 2.26.3
> 
>