Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1688226 > unrolled thread

[PATCH 3.16 121/178] kvm: arm/arm64: Fix locking for kvm_free_stage2_pgd

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH 3.16 121/178] kvm: arm/arm64: Fix locking for  kvm_free_stage2_pgd Ben Hutchings <ben@decadent.org.uk> - 2017-07-16 16:10 +0200
    Re: [PATCH 3.16 121/178] kvm: arm/arm64: Fix locking for  kvm_free_stage2_pgd Suzuki K Poulose <Suzuki.Poulose@arm.com> - 2017-07-17 17:20 +0200
      Re: [PATCH 3.16 121/178] kvm: arm/arm64: Fix locking for  kvm_free_stage2_pgd Ben Hutchings <ben@decadent.org.uk> - 2017-07-18 18:20 +0200

#1688226 — [PATCH 3.16 121/178] kvm: arm/arm64: Fix locking for kvm_free_stage2_pgd

3.16.46-rc1 review patch.  If anyone has any objections, please let me know.

------------------

From: Suzuki K Poulose <suzuki.poulose@arm.com>

commit 8b3405e345b5a098101b0c31b264c812bba045d9 upstream.

In kvm_free_stage2_pgd() we don't hold the kvm->mmu_lock while calling
unmap_stage2_range() on the entire memory range for the guest. This could
cause problems with other callers (e.g, munmap on a memslot) trying to
unmap a range. And since we have to unmap the entire Guest memory range
holding a spinlock, make sure we yield the lock if necessary, after we
unmap each PUD range.

Fixes: commit d5d8184d35c9 ("KVM: ARM: Memory virtualization setup")
Cc: Paolo Bonzini <pbonzin@redhat.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Christoffer Dall <christoffer.dall@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
[ Avoid vCPU starvation and lockup detector warnings ]
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Christoffer Dall <cdall@linaro.org>
[bwh: Backported to 3.16:
 - unmap_stage2_range() is a wrapper around unmap_range(), which is also used for
   HYP page table setup.  So unmap_range() should do the cond_resched_lock(), but
   only if kvm != NULL.
 - Adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
---
 arch/arm/kvm/mmu.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

--- a/arch/arm/kvm/mmu.c
+++ b/arch/arm/kvm/mmu.c
@@ -199,6 +199,12 @@ static void unmap_range(struct kvm *kvm,
 		next = kvm_pgd_addr_end(addr, end);
 		if (!pgd_none(*pgd))
 			unmap_puds(kvm, pgd, addr, next);
+		/*
+		 * If the range is too large, release the kvm->mmu_lock
+		 * to prevent starvation and lockup detector warnings.
+		 */
+		if (kvm && next != end)
+			cond_resched_lock(&kvm->mmu_lock);
 	} while (pgd++, addr = next, addr != end);
 }
 
@@ -553,6 +559,7 @@ int kvm_alloc_stage2_pgd(struct kvm *kvm
  */
 static void unmap_stage2_range(struct kvm *kvm, phys_addr_t start, u64 size)
 {
+	assert_spin_locked(&kvm->mmu_lock);
 	unmap_range(kvm, kvm->arch.pgd, start, size);
 }
 
@@ -637,7 +644,10 @@ void kvm_free_stage2_pgd(struct kvm *kvm
 	if (kvm->arch.pgd == NULL)
 		return;
 
+	spin_lock(&kvm->mmu_lock);
 	unmap_stage2_range(kvm, 0, KVM_PHYS_SIZE);
+	spin_unlock(&kvm->mmu_lock);
+
 	free_pages((unsigned long)kvm->arch.pgd, S2_PGD_ORDER);
 	kvm->arch.pgd = NULL;
 }

[toc] | [next] | [standalone]

#1689167

On 16/07/17 14:56, Ben Hutchings wrote:
> 3.16.46-rc1 review patch.  If anyone has any objections, please let me know.
>
> ------------------
>
> From: Suzuki K Poulose <suzuki.poulose@arm.com>
>
> commit 8b3405e345b5a098101b0c31b264c812bba045d9 upstream.
>
> In kvm_free_stage2_pgd() we don't hold the kvm->mmu_lock while calling
> unmap_stage2_range() on the entire memory range for the guest. This could
> cause problems with other callers (e.g, munmap on a memslot) trying to
> unmap a range. And since we have to unmap the entire Guest memory range
> holding a spinlock, make sure we yield the lock if necessary, after we
> unmap each PUD range.
>
> Fixes: commit d5d8184d35c9 ("KVM: ARM: Memory virtualization setup")
> Cc: Paolo Bonzini <pbonzin@redhat.com>
> Cc: Marc Zyngier <marc.zyngier@arm.com>
> Cc: Christoffer Dall <christoffer.dall@linaro.org>
> Cc: Mark Rutland <mark.rutland@arm.com>
> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
> [ Avoid vCPU starvation and lockup detector warnings ]
> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
> Signed-off-by: Christoffer Dall <cdall@linaro.org>
> [bwh: Backported to 3.16:
>  - unmap_stage2_range() is a wrapper around unmap_range(), which is also used for
>    HYP page table setup.  So unmap_range() should do the cond_resched_lock(), but
>    only if kvm != NULL.
>  - Adjust context]
> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
> ---
>  arch/arm/kvm/mmu.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
>
> --- a/arch/arm/kvm/mmu.c
> +++ b/arch/arm/kvm/mmu.c
> @@ -199,6 +199,12 @@ static void unmap_range(struct kvm *kvm,
>  		next = kvm_pgd_addr_end(addr, end);
>  		if (!pgd_none(*pgd))
>  			unmap_puds(kvm, pgd, addr, next);
> +		/*
> +		 * If the range is too large, release the kvm->mmu_lock
> +		 * to prevent starvation and lockup detector warnings.
> +		 */
> +		if (kvm && next != end)
> +			cond_resched_lock(&kvm->mmu_lock);
>  	} while (pgd++, addr = next, addr != end);
>  }
>
> @@ -553,6 +559,7 @@ int kvm_alloc_stage2_pgd(struct kvm *kvm
>   */
>  static void unmap_stage2_range(struct kvm *kvm, phys_addr_t start, u64 size)
>  {
> +	assert_spin_locked(&kvm->mmu_lock);
>  	unmap_range(kvm, kvm->arch.pgd, start, size);
>  }
>
> @@ -637,7 +644,10 @@ void kvm_free_stage2_pgd(struct kvm *kvm
>  	if (kvm->arch.pgd == NULL)
>  		return;
>
> +	spin_lock(&kvm->mmu_lock);
>  	unmap_stage2_range(kvm, 0, KVM_PHYS_SIZE);
> +	spin_unlock(&kvm->mmu_lock);
> +
>  	free_pages((unsigned long)kvm->arch.pgd, S2_PGD_ORDER);
>  	kvm->arch.pgd = NULL;
>  }

Hi Ben,

There is a follow up patch for this one to make sure we check/free the PGD under the
mmu_lock.

commit 6c0d706b563af73 ("kvm: arm/arm64: Fix race in resetting stage2 PGD")

So unless you want to fold that in, this looks alright. Looks like we missed a Cc:
stable for that. I will send it to stable soon.

Cheers
Suzuki



>

[toc] | [prev] | [next] | [standalone]

#1690412

[Multipart message — attachments visible in raw view] — view raw

On Mon, 2017-07-17 at 16:12 +0100, Suzuki K Poulose wrote:
> On 16/07/17 14:56, Ben Hutchings wrote:
> > 3.16.46-rc1 review patch.  If anyone has any objections, please let me know.
> >
> > ------------------
> >
> > From: Suzuki K Poulose <suzuki.poulose@arm.com>
> >
[...]
> There is a follow up patch for this one to make sure we check/free the PGD under the
> mmu_lock.
> 
> commit 6c0d706b563af73 ("kvm: arm/arm64: Fix race in resetting stage2 PGD")
> 
> So unless you want to fold that in, this looks alright. Looks like we missed a Cc:
> stable for that. I will send it to stable soon.

That does look like a fix suitable for stable.  So far as I can see,
though, "kvm: arm/arm64: Fix locking for kvm_free_stage2_pgd" does not
introduce a regression - right?  So I don't have to apply the two at
the same time.

Ben.

-- 
Ben Hutchings
Quantity is no substitute for quality, but it's the only one we've got.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web