mbox series

[GIT,PULL] KVM/arm64 fixes for 5.12, take #1

Message ID 20210305164944.3729910-1-maz@kernel.org
State New
Headers show
Series [GIT,PULL] KVM/arm64 fixes for 5.12, take #1 | expand

Pull-request

git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm.git tags/kvmarm-fixes-5.12-1

Message

Marc Zyngier March 5, 2021, 4:49 p.m. UTC
Hi Paolo,

Here's the first batch of fixes for 5.12. We have a handful of low
level world-switch regressions, a page table walker fix, more PMU
tidying up, and a workaround for systems with creative firmware.

Note that this is based on -rc1 despite the breakage, as I didn't feel
like holding these patches until -rc2.

Please pull,

	M.

The following changes since commit fe07bfda2fb9cdef8a4d4008a409bb02f35f1bd8:

  Linux 5.12-rc1 (2021-02-28 16:05:19 -0800)

are available in the Git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm.git tags/kvmarm-fixes-5.12-1

for you to fetch changes up to e85583b3f1fe62c9b371a3100c1c91af94005ca9:

  KVM: arm64: Fix range alignment when walking page tables (2021-03-04 09:54:12 +0000)

----------------------------------------------------------------
KVM/arm64 fixes for 5.12, take #1

- Fix SPE context save/restore on nVHE
- Fix some subtle host context corruption on vcpu exit
- Fix panic handling on nVHE
- Prevent the hypervisor from accessing PMU registers when there is none
- Workaround broken firmwares advertising bogus GICv2 compatibility
- Fix Stage-2 unaligned range unmapping

----------------------------------------------------------------
Andrew Scull (1):
      KVM: arm64: Fix nVHE hyp panic host context restore

Jia He (1):
      KVM: arm64: Fix range alignment when walking page tables

Marc Zyngier (4):
      KVM: arm64: Turn kvm_arm_support_pmu_v3() into a static key
      KVM: arm64: Don't access PMSELR_EL0/PMUSERENR_EL0 when no PMU is available
      KVM: arm64: Rename __vgic_v3_get_ich_vtr_el2() to __vgic_v3_get_gic_config()
      KVM: arm64: Workaround firmware wrongly advertising GICv2-on-v3 compatibility

Suzuki K Poulose (1):
      KVM: arm64: nvhe: Save the SPE context early

Will Deacon (1):
      KVM: arm64: Avoid corrupting vCPU context register in guest exit

 arch/arm64/include/asm/kvm_asm.h        |  4 ++--
 arch/arm64/include/asm/kvm_hyp.h        |  8 ++++++-
 arch/arm64/kernel/image-vars.h          |  3 +++
 arch/arm64/kvm/hyp/entry.S              |  2 +-
 arch/arm64/kvm/hyp/include/hyp/switch.h |  9 +++++---
 arch/arm64/kvm/hyp/nvhe/debug-sr.c      | 12 ++++++++--
 arch/arm64/kvm/hyp/nvhe/host.S          | 15 +++++++------
 arch/arm64/kvm/hyp/nvhe/hyp-main.c      |  6 ++---
 arch/arm64/kvm/hyp/nvhe/switch.c        | 14 +++++++++---
 arch/arm64/kvm/hyp/pgtable.c            |  1 +
 arch/arm64/kvm/hyp/vgic-v3-sr.c         | 40 +++++++++++++++++++++++++++++++--
 arch/arm64/kvm/perf.c                   | 10 +++++++++
 arch/arm64/kvm/pmu-emul.c               | 10 ---------
 arch/arm64/kvm/vgic/vgic-v3.c           | 12 +++++++---
 include/kvm/arm_pmu.h                   |  9 ++++++--
 15 files changed, 116 insertions(+), 39 deletions(-)

Comments

Paolo Bonzini March 5, 2021, 5:27 p.m. UTC | #1
On 05/03/21 17:49, Marc Zyngier wrote:
> Hi Paolo,
> 
> Here's the first batch of fixes for 5.12. We have a handful of low
> level world-switch regressions, a page table walker fix, more PMU
> tidying up, and a workaround for systems with creative firmware.
> 
> Note that this is based on -rc1 despite the breakage, as I didn't feel
> like holding these patches until -rc2.
> 
> Please pull,
> 
> 	M.
> 
> The following changes since commit fe07bfda2fb9cdef8a4d4008a409bb02f35f1bd8:
> 
>    Linux 5.12-rc1 (2021-02-28 16:05:19 -0800)
> 
> are available in the Git repository at:
> 
>    git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm.git tags/kvmarm-fixes-5.12-1
> 
> for you to fetch changes up to e85583b3f1fe62c9b371a3100c1c91af94005ca9:
> 
>    KVM: arm64: Fix range alignment when walking page tables (2021-03-04 09:54:12 +0000)

Hi Marc,

due to a severe data corruption bug in 5.12-rc1, Linus suggested not 
including 5.12-rc1 in trees to avoid it eating our filesystems 
unwittingly during future bisections.

Would it be a problem for you to rebase on top of your merge window pull 
request?  If there are conflicts, another possibility is for you to just 
send me the patch series.  I will handle all the topic branch juggling.

This will mean rewriting kvmarm.git's history, but it does seem to be 
the lesser (or the most future-proof) evil.

Thanks,

Paolo

> ----------------------------------------------------------------
> KVM/arm64 fixes for 5.12, take #1
> 
> - Fix SPE context save/restore on nVHE
> - Fix some subtle host context corruption on vcpu exit
> - Fix panic handling on nVHE
> - Prevent the hypervisor from accessing PMU registers when there is none
> - Workaround broken firmwares advertising bogus GICv2 compatibility
> - Fix Stage-2 unaligned range unmapping
> 
> ----------------------------------------------------------------
> Andrew Scull (1):
>        KVM: arm64: Fix nVHE hyp panic host context restore
> 
> Jia He (1):
>        KVM: arm64: Fix range alignment when walking page tables
> 
> Marc Zyngier (4):
>        KVM: arm64: Turn kvm_arm_support_pmu_v3() into a static key
>        KVM: arm64: Don't access PMSELR_EL0/PMUSERENR_EL0 when no PMU is available
>        KVM: arm64: Rename __vgic_v3_get_ich_vtr_el2() to __vgic_v3_get_gic_config()
>        KVM: arm64: Workaround firmware wrongly advertising GICv2-on-v3 compatibility
> 
> Suzuki K Poulose (1):
>        KVM: arm64: nvhe: Save the SPE context early
> 
> Will Deacon (1):
>        KVM: arm64: Avoid corrupting vCPU context register in guest exit
> 
>   arch/arm64/include/asm/kvm_asm.h        |  4 ++--
>   arch/arm64/include/asm/kvm_hyp.h        |  8 ++++++-
>   arch/arm64/kernel/image-vars.h          |  3 +++
>   arch/arm64/kvm/hyp/entry.S              |  2 +-
>   arch/arm64/kvm/hyp/include/hyp/switch.h |  9 +++++---
>   arch/arm64/kvm/hyp/nvhe/debug-sr.c      | 12 ++++++++--
>   arch/arm64/kvm/hyp/nvhe/host.S          | 15 +++++++------
>   arch/arm64/kvm/hyp/nvhe/hyp-main.c      |  6 ++---
>   arch/arm64/kvm/hyp/nvhe/switch.c        | 14 +++++++++---
>   arch/arm64/kvm/hyp/pgtable.c            |  1 +
>   arch/arm64/kvm/hyp/vgic-v3-sr.c         | 40 +++++++++++++++++++++++++++++++--
>   arch/arm64/kvm/perf.c                   | 10 +++++++++
>   arch/arm64/kvm/pmu-emul.c               | 10 ---------
>   arch/arm64/kvm/vgic/vgic-v3.c           | 12 +++++++---
>   include/kvm/arm_pmu.h                   |  9 ++++++--
>   15 files changed, 116 insertions(+), 39 deletions(-)
>
Marc Zyngier March 5, 2021, 6:47 p.m. UTC | #2
Hi Paolo,

On Fri, 05 Mar 2021 17:27:36 +0000,
Paolo Bonzini <pbonzini@redhat.com> wrote:
> 
> On 05/03/21 17:49, Marc Zyngier wrote:
> > Hi Paolo,
> > 
> > Here's the first batch of fixes for 5.12. We have a handful of low
> > level world-switch regressions, a page table walker fix, more PMU
> > tidying up, and a workaround for systems with creative firmware.
> > 
> > Note that this is based on -rc1 despite the breakage, as I didn't feel
> > like holding these patches until -rc2.
> > 
> > Please pull,
> > 
> > 	M.
> > 
> > The following changes since commit fe07bfda2fb9cdef8a4d4008a409bb02f35f1bd8:
> > 
> >    Linux 5.12-rc1 (2021-02-28 16:05:19 -0800)
> > 
> > are available in the Git repository at:
> > 
> >    git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm.git tags/kvmarm-fixes-5.12-1
> > 
> > for you to fetch changes up to e85583b3f1fe62c9b371a3100c1c91af94005ca9:
> > 
> >    KVM: arm64: Fix range alignment when walking page tables (2021-03-04 09:54:12 +0000)
> 
> Hi Marc,
> 
> due to a severe data corruption bug in 5.12-rc1, Linus suggested not
> including 5.12-rc1 in trees to avoid it eating our filesystems
> unwittingly during future bisections.
> 
> Would it be a problem for you to rebase on top of your merge window
> pull request?  If there are conflicts, another possibility is for you
> to just send me the patch series.  I will handle all the topic branch
> juggling.
> 
> This will mean rewriting kvmarm.git's history, but it does seem to be
> the lesser (or the most future-proof) evil.

The problem is that this is not only kvmarm, but also the Android
tree, which directly pulls from the kvmarm stable branches. I guess
we'll have to live with it.

I'll reply to this email with the patch series.

Thanks,

	M.