Skip to contentSkip to conversation

KVM/arm64 fixes for 6.3, part #2

GIT PULL Oliver Upton 3 participants 6 messages Mar 17, 2023 – Mar 27, 2023

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hi Paolo,

Another week, another set of fixes for KVM/arm64.

Description can be found in the tag, but the teardown race when walking host page tables is particularly nasty and currently causing problems for folks. The fix is quite simple by disabling interrupts when walking host page tables, as the thread must be IPI'ed before the table memory can actually be freed.

Please pull,

Oliver

The following changes since commit 47053904e18282af4525a02e3e0f519f014fc7f9:

KVM: arm64: timers: Convert per-vcpu virtual offset to a global value (2023-03-11 02:00:40 -0800)

are available in the Git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm.git tags/kvmarm-fixes-6.3-2

for you to fetch changes up to 8c2e8ac8ad4be68409e806ce1cc78fc7a04539f3:

KVM: arm64: Check for kvm_vma_mte_allowed in the critical section (2023-03-16 23:42:56 +0000)

----------------------------------------------------------------
KVM/arm64 fixes for 6.3, part #2

Fixes for a rather interesting set of bugs relating to the MMU:

- Read the MMU notifier seq before dropping the mmap lock to guard against reading a potentially stale VMA

- Disable interrupts when walking user page tables to protect against the page table being freed

- Read the MTE permissions for the VMA within the mmap lock critical section, avoiding the use of a potentally stale VMA pointer

Additionally, some fixes targeting the vPMU:

- Return the sum of the current perf event value and PMC snapshot for reads from userspace

- Don't save the value of guest writes to PMCR_EL0.{C,P}, which could otherwise lead to userspace erroneously resetting the vPMU during VM save/restore

----------------------------------------------------------------
David Matlack (1):
KVM: arm64: Retry fault if vma_lookup() results become invalid

Marc Zyngier (2):
KVM: arm64: Disable interrupts while walking userspace PTs
KVM: arm64: Check for kvm_vma_mte_allowed in the critical section

Reiji Watanabe (2):
KVM: arm64: PMU: Fix GET_ONE_REG for vPMC regs to return the current value
KVM: arm64: PMU: Don't save PMCR_EL0.{C,P} for the vCPU

 arch/arm64/kvm/mmu.c      | 99 ++++++++++++++++++++++++++++++-----------------
 arch/arm64/kvm/pmu-emul.c |  3 +-
 arch/arm64/kvm/sys_regs.c | 21 +++++++++-
 3 files changed, 85 insertions(+), 38 deletions(-)
#1 lore raw

Paolo,

Pinging this PR in case you missed it, the issues around host page table walks are particularly urgent as the race on host page table teardown has been reproduced on some setups.

--
Thanks,
Oliver

Quoted 66 lines · Oliver Upton

Hi Paolo,

Another week, another set of fixes for KVM/arm64.

Description can be found in the tag, but the teardown race when walking host page tables is particularly nasty and currently causing problems for folks. The fix is quite simple by disabling interrupts when walking host page tables, as the thread must be IPI'ed before the table memory can actually be freed.

Please pull,

Oliver

The following changes since commit 47053904e18282af4525a02e3e0f519f014fc7f9:

KVM: arm64: timers: Convert per-vcpu virtual offset to a global value (2023-03-11 02:00:40 -0800)

are available in the Git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm.git tags/kvmarm-fixes-6.3-2

for you to fetch changes up to 8c2e8ac8ad4be68409e806ce1cc78fc7a04539f3:

KVM: arm64: Check for kvm_vma_mte_allowed in the critical section (2023-03-16 23:42:56 +0000)

----------------------------------------------------------------
KVM/arm64 fixes for 6.3, part #2

Fixes for a rather interesting set of bugs relating to the MMU:

- Read the MMU notifier seq before dropping the mmap lock to guard against reading a potentially stale VMA

- Disable interrupts when walking user page tables to protect against the page table being freed

- Read the MTE permissions for the VMA within the mmap lock critical section, avoiding the use of a potentally stale VMA pointer

Additionally, some fixes targeting the vPMU:

- Return the sum of the current perf event value and PMC snapshot for reads from userspace

- Don't save the value of guest writes to PMCR_EL0.{C,P}, which could otherwise lead to userspace erroneously resetting the vPMU during VM save/restore

----------------------------------------------------------------
David Matlack (1):
KVM: arm64: Retry fault if vma_lookup() results become invalid

Marc Zyngier (2):
KVM: arm64: Disable interrupts while walking userspace PTs
KVM: arm64: Check for kvm_vma_mte_allowed in the critical section

Reiji Watanabe (2):
KVM: arm64: PMU: Fix GET_ONE_REG for vPMC regs to return the current value
KVM: arm64: PMU: Don't save PMCR_EL0.{C,P} for the vCPU

 arch/arm64/kvm/mmu.c      | 99 ++++++++++++++++++++++++++++++-----------------
 arch/arm64/kvm/pmu-emul.c |  3 +-
 arch/arm64/kvm/sys_regs.c | 21 +++++++++-
 3 files changed, 85 insertions(+), 38 deletions(-)
Signature · 2 lines
Thanks,
Oliver
maintainer
#2 lore raw
Quoted 6 lines · Oliver Upton

Paolo,

Pinging this PR in case you missed it, the issues around host page table walks are particularly urgent as the race on host page table teardown has been reproduced on some setups.

If we don't hear from Paolo shortly, I suggest we route this via the arm64 tree.

Thanks,

	M.
Quoted 77 lines

--
Thanks,
Oliver

Quoted 66 lines · Oliver Upton

Hi Paolo,

Another week, another set of fixes for KVM/arm64.

Description can be found in the tag, but the teardown race when walking host page tables is particularly nasty and currently causing problems for folks. The fix is quite simple by disabling interrupts when walking host page tables, as the thread must be IPI'ed before the table memory can actually be freed.

Please pull,

Oliver

The following changes since commit 47053904e18282af4525a02e3e0f519f014fc7f9:

KVM: arm64: timers: Convert per-vcpu virtual offset to a global value (2023-03-11 02:00:40 -0800)

are available in the Git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm.git tags/kvmarm-fixes-6.3-2

for you to fetch changes up to 8c2e8ac8ad4be68409e806ce1cc78fc7a04539f3:

KVM: arm64: Check for kvm_vma_mte_allowed in the critical section (2023-03-16 23:42:56 +0000)

----------------------------------------------------------------
KVM/arm64 fixes for 6.3, part #2

Fixes for a rather interesting set of bugs relating to the MMU:

- Read the MMU notifier seq before dropping the mmap lock to guard against reading a potentially stale VMA

- Disable interrupts when walking user page tables to protect against the page table being freed

- Read the MTE permissions for the VMA within the mmap lock critical section, avoiding the use of a potentally stale VMA pointer

Additionally, some fixes targeting the vPMU:

- Return the sum of the current perf event value and PMC snapshot for reads from userspace

- Don't save the value of guest writes to PMCR_EL0.{C,P}, which could otherwise lead to userspace erroneously resetting the vPMU during VM save/restore

----------------------------------------------------------------
David Matlack (1):
KVM: arm64: Retry fault if vma_lookup() results become invalid

Marc Zyngier (2):
KVM: arm64: Disable interrupts while walking userspace PTs
KVM: arm64: Check for kvm_vma_mte_allowed in the critical section

Reiji Watanabe (2):
KVM: arm64: PMU: Fix GET_ONE_REG for vPMC regs to return the current value
KVM: arm64: PMU: Don't save PMCR_EL0.{C,P} for the vCPU

 arch/arm64/kvm/mmu.c      | 99 ++++++++++++++++++++++++++++++-----------------
 arch/arm64/kvm/pmu-emul.c |  3 +-
 arch/arm64/kvm/sys_regs.c | 21 +++++++++-
 3 files changed, 85 insertions(+), 38 deletions(-)
Signature · 2 lines
Thanks,
Oliver
Signature · 1 line
Without deviation from the norm, progress is not possible.
maintainer
#3 lore raw
Quoted 7 lines · Marc Zyngier
Quoted 5 lines

Paolo,

Pinging this PR in case you missed it, the issues around host page table walks are particularly urgent as the race on host page table teardown has been reproduced on some setups.

If we don't hear from Paolo shortly, I suggest we route this via the arm64 tree.

It missed the pull request I sent on March 17th by a few hours. I have queued it now and will send it to Linus later today.

Paolo

maintainer
#4 lore raw
Quoted 4 lines · Paolo Bonzini
Quoted 2 lines · Marc Zyngier
Quoted 2 lines

Paolo,

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

Pinging this PR in case you missed it, the issues around host page table walks are particularly urgent as the race on host page table teardown has been reproduced on some setups.

If we don't hear from Paolo shortly, I suggest we route this via the arm64 tree.

It missed the pull request I sent on March 17th by a few hours. I have queued it now and will send it to Linus later today.

Maybe you could help us here and state what is your schedule when it comes to sending these pull requests? It would certainly help coordinate and avoid wasting 10+ days to get things merged.

I appreciate that you don't need nor want to wait for us to send something to Linus, but if we know when the train is departing, we can make sure we're standing on the platform early enough.

Thanks,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
maintainer
#5 lore raw
Quoted 10 lines · Marc Zyngier
Quoted 2 lines

It missed the pull request I sent on March 17th by a few hours. I have queued it now and will send it to Linus later today.

Maybe you could help us here and state what is your schedule when it comes to sending these pull requests? It would certainly help coordinate and avoid wasting 10+ days to get things merged.

I appreciate that you don't need nor want to wait for us to send something to Linus, but if we know when the train is departing, we can make sure we're standing on the platform early enough.

In general, sending the pull request to me on or before Thursday will work best, though I have no problem sending stuff to Linus on Sunday morning (so that architectures don't need to time their request too carefully).

These weeks my Friday afternoon was more free than usual due to all meetings with American people moving one hour earlier, and that translated into different family plans for Friday itself and over the weekend.

Paolo