* [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
@ 2013-08-07 10:15 ` Bharat Bhushan
0 siblings, 0 replies; 20+ messages in thread
From: Bharat Bhushan @ 2013-08-07 10:15 UTC (permalink / raw)
To: paulus, scottwood, agraf, kvm-ppc, kvm; +Cc: Bharat Bhushan, Bharat Bhushan
When the MM code is invalidating a range of pages, it calls the KVM
kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
However, the Linux PTEs for the range being flushed are still valid at
that point. We are not supposed to establish any new references to pages
in the range until the ...range_end() notifier gets called.
The PPC-specific KVM code doesn't get any explicit notification of that;
instead, we are supposed to use mmu_notifier_retry() to test whether we
are or have been inside a range flush notifier pair while we have been
referencing a page.
This patch calls the mmu_notifier_retry() while mapping the guest
page to ensure we are not referencing a page when in range invalidation.
This call is inside a region locked with kvm->mmu_lock, which is the
same lock that is called by the KVM MMU notifier functions, thus
ensuring that no new notification can proceed while we are in the
locked region.
Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
---
arch/powerpc/kvm/e500_mmu_host.c | 19 +++++++++++++++++--
1 files changed, 17 insertions(+), 2 deletions(-)
diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
index ff6dd66..ae4eaf6 100644
--- a/arch/powerpc/kvm/e500_mmu_host.c
+++ b/arch/powerpc/kvm/e500_mmu_host.c
@@ -329,8 +329,14 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
int tsize = BOOK3E_PAGESZ_4K;
unsigned long tsize_pages = 0;
pte_t *ptep;
- int wimg = 0;
+ int wimg = 0, ret = 0;
pgd_t *pgdir;
+ unsigned long mmu_seq;
+ struct kvm *kvm = vcpu_e500->vcpu.kvm;
+
+ /* used to check for invalidations in progress */
+ mmu_seq = kvm->mmu_notifier_seq;
+ smp_rmb();
/*
* Translate guest physical to true physical, acquiring
@@ -458,6 +464,13 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
(long)gfn, pfn);
return -EINVAL;
}
+
+ spin_lock(&kvm->mmu_lock);
+ if (mmu_notifier_retry(kvm, mmu_seq)) {
+ ret = -EAGAIN;
+ goto out;
+ }
+
kvmppc_e500_ref_setup(ref, gtlbe, pfn, wimg);
kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
@@ -466,10 +479,12 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
/* Clear i-cache for new pages */
kvmppc_mmu_flush_icache(pfn);
+out:
+ spin_unlock(&kvm->mmu_lock);
/* Drop refcount on page, so that mmu notifiers can clear it */
kvm_release_pfn_clean(pfn);
- return 0;
+ return ret;
}
/* XXX only map the one-one case, for now use TLB0 */
--
1.7.0.4
^ permalink raw reply related [flat|nested] 20+ messages in thread* Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
2013-08-07 10:15 ` Bharat Bhushan
@ 2013-08-10 1:15 ` Scott Wood
-1 siblings, 0 replies; 20+ messages in thread
From: Scott Wood @ 2013-08-10 1:15 UTC (permalink / raw)
To: Bharat Bhushan; +Cc: paulus, agraf, kvm-ppc, kvm, Bharat Bhushan
On Wed, 2013-08-07 at 15:33 +0530, Bharat Bhushan wrote:
> When the MM code is invalidating a range of pages, it calls the KVM
> kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
> kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
> However, the Linux PTEs for the range being flushed are still valid at
> that point. We are not supposed to establish any new references to pages
> in the range until the ...range_end() notifier gets called.
> The PPC-specific KVM code doesn't get any explicit notification of that;
> instead, we are supposed to use mmu_notifier_retry() to test whether we
> are or have been inside a range flush notifier pair while we have been
> referencing a page.
>
> This patch calls the mmu_notifier_retry() while mapping the guest
> page to ensure we are not referencing a page when in range invalidation.
>
> This call is inside a region locked with kvm->mmu_lock, which is the
> same lock that is called by the KVM MMU notifier functions, thus
> ensuring that no new notification can proceed while we are in the
> locked region.
>
> Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
> ---
> arch/powerpc/kvm/e500_mmu_host.c | 19 +++++++++++++++++--
> 1 files changed, 17 insertions(+), 2 deletions(-)
>
> diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
> index ff6dd66..ae4eaf6 100644
> --- a/arch/powerpc/kvm/e500_mmu_host.c
> +++ b/arch/powerpc/kvm/e500_mmu_host.c
> @@ -329,8 +329,14 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> int tsize = BOOK3E_PAGESZ_4K;
> unsigned long tsize_pages = 0;
> pte_t *ptep;
> - int wimg = 0;
> + int wimg = 0, ret = 0;
> pgd_t *pgdir;
> + unsigned long mmu_seq;
> + struct kvm *kvm = vcpu_e500->vcpu.kvm;
> +
> + /* used to check for invalidations in progress */
> + mmu_seq = kvm->mmu_notifier_seq;
> + smp_rmb();
>
> /*
> * Translate guest physical to true physical, acquiring
> @@ -458,6 +464,13 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> (long)gfn, pfn);
> return -EINVAL;
> }
> +
> + spin_lock(&kvm->mmu_lock);
> + if (mmu_notifier_retry(kvm, mmu_seq)) {
> + ret = -EAGAIN;
> + goto out;
> + }
> +
> kvmppc_e500_ref_setup(ref, gtlbe, pfn, wimg);
>
> kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
> @@ -466,10 +479,12 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> /* Clear i-cache for new pages */
> kvmppc_mmu_flush_icache(pfn);
>
> +out:
> + spin_unlock(&kvm->mmu_lock);
> /* Drop refcount on page, so that mmu notifiers can clear it */
> kvm_release_pfn_clean(pfn);
>
> - return 0;
> + return ret;
> }
Acked-by: Scott Wood <scottwood@freescale.com> since it's currently the
standard KVM approach, though I'm not happy about the busy-waiting
aspect of it. What if we preempted the thread responsible for
decrementing mmu_notifier_count? What if we did so being a SCHED_FIFO
task of higher priority than the decrementing thread?
-Scott
^ permalink raw reply [flat|nested] 20+ messages in thread* Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
@ 2013-08-10 1:15 ` Scott Wood
0 siblings, 0 replies; 20+ messages in thread
From: Scott Wood @ 2013-08-10 1:15 UTC (permalink / raw)
To: Bharat Bhushan; +Cc: paulus, agraf, kvm-ppc, kvm, Bharat Bhushan
On Wed, 2013-08-07 at 15:33 +0530, Bharat Bhushan wrote:
> When the MM code is invalidating a range of pages, it calls the KVM
> kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
> kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
> However, the Linux PTEs for the range being flushed are still valid at
> that point. We are not supposed to establish any new references to pages
> in the range until the ...range_end() notifier gets called.
> The PPC-specific KVM code doesn't get any explicit notification of that;
> instead, we are supposed to use mmu_notifier_retry() to test whether we
> are or have been inside a range flush notifier pair while we have been
> referencing a page.
>
> This patch calls the mmu_notifier_retry() while mapping the guest
> page to ensure we are not referencing a page when in range invalidation.
>
> This call is inside a region locked with kvm->mmu_lock, which is the
> same lock that is called by the KVM MMU notifier functions, thus
> ensuring that no new notification can proceed while we are in the
> locked region.
>
> Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
> ---
> arch/powerpc/kvm/e500_mmu_host.c | 19 +++++++++++++++++--
> 1 files changed, 17 insertions(+), 2 deletions(-)
>
> diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
> index ff6dd66..ae4eaf6 100644
> --- a/arch/powerpc/kvm/e500_mmu_host.c
> +++ b/arch/powerpc/kvm/e500_mmu_host.c
> @@ -329,8 +329,14 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> int tsize = BOOK3E_PAGESZ_4K;
> unsigned long tsize_pages = 0;
> pte_t *ptep;
> - int wimg = 0;
> + int wimg = 0, ret = 0;
> pgd_t *pgdir;
> + unsigned long mmu_seq;
> + struct kvm *kvm = vcpu_e500->vcpu.kvm;
> +
> + /* used to check for invalidations in progress */
> + mmu_seq = kvm->mmu_notifier_seq;
> + smp_rmb();
>
> /*
> * Translate guest physical to true physical, acquiring
> @@ -458,6 +464,13 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> (long)gfn, pfn);
> return -EINVAL;
> }
> +
> + spin_lock(&kvm->mmu_lock);
> + if (mmu_notifier_retry(kvm, mmu_seq)) {
> + ret = -EAGAIN;
> + goto out;
> + }
> +
> kvmppc_e500_ref_setup(ref, gtlbe, pfn, wimg);
>
> kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
> @@ -466,10 +479,12 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> /* Clear i-cache for new pages */
> kvmppc_mmu_flush_icache(pfn);
>
> +out:
> + spin_unlock(&kvm->mmu_lock);
> /* Drop refcount on page, so that mmu notifiers can clear it */
> kvm_release_pfn_clean(pfn);
>
> - return 0;
> + return ret;
> }
Acked-by: Scott Wood <scottwood@freescale.com> since it's currently the
standard KVM approach, though I'm not happy about the busy-waiting
aspect of it. What if we preempted the thread responsible for
decrementing mmu_notifier_count? What if we did so being a SCHED_FIFO
task of higher priority than the decrementing thread?
-Scott
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
2013-08-07 10:15 ` Bharat Bhushan
@ 2013-10-04 13:38 ` Alexander Graf
-1 siblings, 0 replies; 20+ messages in thread
From: Alexander Graf @ 2013-10-04 13:38 UTC (permalink / raw)
To: Bharat Bhushan
Cc: Paul Mackerras, Scott Wood, kvm-ppc,
kvm@vger.kernel.org mailing list, Bharat Bhushan, Gleb Natapov,
Paolo Bonzini
On 07.08.2013, at 12:03, Bharat Bhushan wrote:
> When the MM code is invalidating a range of pages, it calls the KVM
> kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
> kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
> However, the Linux PTEs for the range being flushed are still valid at
> that point. We are not supposed to establish any new references to pages
> in the range until the ...range_end() notifier gets called.
> The PPC-specific KVM code doesn't get any explicit notification of that;
> instead, we are supposed to use mmu_notifier_retry() to test whether we
> are or have been inside a range flush notifier pair while we have been
> referencing a page.
>
> This patch calls the mmu_notifier_retry() while mapping the guest
> page to ensure we are not referencing a page when in range invalidation.
>
> This call is inside a region locked with kvm->mmu_lock, which is the
> same lock that is called by the KVM MMU notifier functions, thus
> ensuring that no new notification can proceed while we are in the
> locked region.
>
> Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
Acked-by: Alexander Graf <agraf@suse.de>
Gleb, Paolo, please queue for 3.12 directly.
Alex
> ---
> arch/powerpc/kvm/e500_mmu_host.c | 19 +++++++++++++++++--
> 1 files changed, 17 insertions(+), 2 deletions(-)
>
> diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
> index ff6dd66..ae4eaf6 100644
> --- a/arch/powerpc/kvm/e500_mmu_host.c
> +++ b/arch/powerpc/kvm/e500_mmu_host.c
> @@ -329,8 +329,14 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> int tsize = BOOK3E_PAGESZ_4K;
> unsigned long tsize_pages = 0;
> pte_t *ptep;
> - int wimg = 0;
> + int wimg = 0, ret = 0;
> pgd_t *pgdir;
> + unsigned long mmu_seq;
> + struct kvm *kvm = vcpu_e500->vcpu.kvm;
> +
> + /* used to check for invalidations in progress */
> + mmu_seq = kvm->mmu_notifier_seq;
> + smp_rmb();
>
> /*
> * Translate guest physical to true physical, acquiring
> @@ -458,6 +464,13 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> (long)gfn, pfn);
> return -EINVAL;
> }
> +
> + spin_lock(&kvm->mmu_lock);
> + if (mmu_notifier_retry(kvm, mmu_seq)) {
> + ret = -EAGAIN;
> + goto out;
> + }
> +
> kvmppc_e500_ref_setup(ref, gtlbe, pfn, wimg);
>
> kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
> @@ -466,10 +479,12 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> /* Clear i-cache for new pages */
> kvmppc_mmu_flush_icache(pfn);
>
> +out:
> + spin_unlock(&kvm->mmu_lock);
> /* Drop refcount on page, so that mmu notifiers can clear it */
> kvm_release_pfn_clean(pfn);
>
> - return 0;
> + return ret;
> }
>
> /* XXX only map the one-one case, for now use TLB0 */
> --
> 1.7.0.4
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe kvm-ppc" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 20+ messages in thread* Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
@ 2013-10-04 13:38 ` Alexander Graf
0 siblings, 0 replies; 20+ messages in thread
From: Alexander Graf @ 2013-10-04 13:38 UTC (permalink / raw)
To: Bharat Bhushan
Cc: Paul Mackerras, Scott Wood, kvm-ppc,
kvm@vger.kernel.org mailing list, Bharat Bhushan, Gleb Natapov,
Paolo Bonzini
On 07.08.2013, at 12:03, Bharat Bhushan wrote:
> When the MM code is invalidating a range of pages, it calls the KVM
> kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
> kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
> However, the Linux PTEs for the range being flushed are still valid at
> that point. We are not supposed to establish any new references to pages
> in the range until the ...range_end() notifier gets called.
> The PPC-specific KVM code doesn't get any explicit notification of that;
> instead, we are supposed to use mmu_notifier_retry() to test whether we
> are or have been inside a range flush notifier pair while we have been
> referencing a page.
>
> This patch calls the mmu_notifier_retry() while mapping the guest
> page to ensure we are not referencing a page when in range invalidation.
>
> This call is inside a region locked with kvm->mmu_lock, which is the
> same lock that is called by the KVM MMU notifier functions, thus
> ensuring that no new notification can proceed while we are in the
> locked region.
>
> Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
Acked-by: Alexander Graf <agraf@suse.de>
Gleb, Paolo, please queue for 3.12 directly.
Alex
> ---
> arch/powerpc/kvm/e500_mmu_host.c | 19 +++++++++++++++++--
> 1 files changed, 17 insertions(+), 2 deletions(-)
>
> diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
> index ff6dd66..ae4eaf6 100644
> --- a/arch/powerpc/kvm/e500_mmu_host.c
> +++ b/arch/powerpc/kvm/e500_mmu_host.c
> @@ -329,8 +329,14 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> int tsize = BOOK3E_PAGESZ_4K;
> unsigned long tsize_pages = 0;
> pte_t *ptep;
> - int wimg = 0;
> + int wimg = 0, ret = 0;
> pgd_t *pgdir;
> + unsigned long mmu_seq;
> + struct kvm *kvm = vcpu_e500->vcpu.kvm;
> +
> + /* used to check for invalidations in progress */
> + mmu_seq = kvm->mmu_notifier_seq;
> + smp_rmb();
>
> /*
> * Translate guest physical to true physical, acquiring
> @@ -458,6 +464,13 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> (long)gfn, pfn);
> return -EINVAL;
> }
> +
> + spin_lock(&kvm->mmu_lock);
> + if (mmu_notifier_retry(kvm, mmu_seq)) {
> + ret = -EAGAIN;
> + goto out;
> + }
> +
> kvmppc_e500_ref_setup(ref, gtlbe, pfn, wimg);
>
> kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
> @@ -466,10 +479,12 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
> /* Clear i-cache for new pages */
> kvmppc_mmu_flush_icache(pfn);
>
> +out:
> + spin_unlock(&kvm->mmu_lock);
> /* Drop refcount on page, so that mmu notifiers can clear it */
> kvm_release_pfn_clean(pfn);
>
> - return 0;
> + return ret;
> }
>
> /* XXX only map the one-one case, for now use TLB0 */
> --
> 1.7.0.4
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe kvm-ppc" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 20+ messages in thread* Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
2013-10-04 13:38 ` Alexander Graf
@ 2013-10-07 12:04 ` Paolo Bonzini
-1 siblings, 0 replies; 20+ messages in thread
From: Paolo Bonzini @ 2013-10-07 12:04 UTC (permalink / raw)
To: Alexander Graf
Cc: Bharat Bhushan, Paul Mackerras, Scott Wood, kvm-ppc,
kvm@vger.kernel.org mailing list, Bharat Bhushan, Gleb Natapov
Il 04/10/2013 15:38, Alexander Graf ha scritto:
>
> On 07.08.2013, at 12:03, Bharat Bhushan wrote:
>
>> When the MM code is invalidating a range of pages, it calls the KVM
>> kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
>> kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
>> However, the Linux PTEs for the range being flushed are still valid at
>> that point. We are not supposed to establish any new references to pages
>> in the range until the ...range_end() notifier gets called.
>> The PPC-specific KVM code doesn't get any explicit notification of that;
>> instead, we are supposed to use mmu_notifier_retry() to test whether we
>> are or have been inside a range flush notifier pair while we have been
>> referencing a page.
>>
>> This patch calls the mmu_notifier_retry() while mapping the guest
>> page to ensure we are not referencing a page when in range invalidation.
>>
>> This call is inside a region locked with kvm->mmu_lock, which is the
>> same lock that is called by the KVM MMU notifier functions, thus
>> ensuring that no new notification can proceed while we are in the
>> locked region.
>>
>> Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
>
> Acked-by: Alexander Graf <agraf@suse.de>
>
> Gleb, Paolo, please queue for 3.12 directly.
Here is the backport. The second hunk has a nontrivial conflict, so
someone please give their {Tested,Reviewed,Compiled}-by.
Paolo
diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
index 1c6a9d7..c65593a 100644
--- a/arch/powerpc/kvm/e500_mmu_host.c
+++ b/arch/powerpc/kvm/e500_mmu_host.c
@@ -332,6 +332,13 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
unsigned long hva;
int pfnmap = 0;
int tsize = BOOK3E_PAGESZ_4K;
+ int ret = 0;
+ unsigned long mmu_seq;
+ struct kvm *kvm = vcpu_e500->vcpu.kvm;
+
+ /* used to check for invalidations in progress */
+ mmu_seq = kvm->mmu_notifier_seq;
+ smp_rmb();
/*
* Translate guest physical to true physical, acquiring
@@ -449,6 +456,12 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
gvaddr &= ~((tsize_pages << PAGE_SHIFT) - 1);
}
+ spin_lock(&kvm->mmu_lock);
+ if (mmu_notifier_retry(kvm, mmu_seq)) {
+ ret = -EAGAIN;
+ goto out;
+ }
+
kvmppc_e500_ref_setup(ref, gtlbe, pfn);
kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
@@ -457,10 +470,13 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
/* Clear i-cache for new pages */
kvmppc_mmu_flush_icache(pfn);
+out:
+ spin_unlock(&kvm->mmu_lock);
+
/* Drop refcount on page, so that mmu notifiers can clear it */
kvm_release_pfn_clean(pfn);
- return 0;
+ return ret;
}
/* XXX only map the one-one case, for now use TLB0 */
^ permalink raw reply related [flat|nested] 20+ messages in thread* Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
@ 2013-10-07 12:04 ` Paolo Bonzini
0 siblings, 0 replies; 20+ messages in thread
From: Paolo Bonzini @ 2013-10-07 12:04 UTC (permalink / raw)
To: Alexander Graf
Cc: Bharat Bhushan, Paul Mackerras, Scott Wood, kvm-ppc,
kvm@vger.kernel.org mailing list, Bharat Bhushan, Gleb Natapov
Il 04/10/2013 15:38, Alexander Graf ha scritto:
>
> On 07.08.2013, at 12:03, Bharat Bhushan wrote:
>
>> When the MM code is invalidating a range of pages, it calls the KVM
>> kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
>> kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
>> However, the Linux PTEs for the range being flushed are still valid at
>> that point. We are not supposed to establish any new references to pages
>> in the range until the ...range_end() notifier gets called.
>> The PPC-specific KVM code doesn't get any explicit notification of that;
>> instead, we are supposed to use mmu_notifier_retry() to test whether we
>> are or have been inside a range flush notifier pair while we have been
>> referencing a page.
>>
>> This patch calls the mmu_notifier_retry() while mapping the guest
>> page to ensure we are not referencing a page when in range invalidation.
>>
>> This call is inside a region locked with kvm->mmu_lock, which is the
>> same lock that is called by the KVM MMU notifier functions, thus
>> ensuring that no new notification can proceed while we are in the
>> locked region.
>>
>> Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
>
> Acked-by: Alexander Graf <agraf@suse.de>
>
> Gleb, Paolo, please queue for 3.12 directly.
Here is the backport. The second hunk has a nontrivial conflict, so
someone please give their {Tested,Reviewed,Compiled}-by.
Paolo
diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
index 1c6a9d7..c65593a 100644
--- a/arch/powerpc/kvm/e500_mmu_host.c
+++ b/arch/powerpc/kvm/e500_mmu_host.c
@@ -332,6 +332,13 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
unsigned long hva;
int pfnmap = 0;
int tsize = BOOK3E_PAGESZ_4K;
+ int ret = 0;
+ unsigned long mmu_seq;
+ struct kvm *kvm = vcpu_e500->vcpu.kvm;
+
+ /* used to check for invalidations in progress */
+ mmu_seq = kvm->mmu_notifier_seq;
+ smp_rmb();
/*
* Translate guest physical to true physical, acquiring
@@ -449,6 +456,12 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
gvaddr &= ~((tsize_pages << PAGE_SHIFT) - 1);
}
+ spin_lock(&kvm->mmu_lock);
+ if (mmu_notifier_retry(kvm, mmu_seq)) {
+ ret = -EAGAIN;
+ goto out;
+ }
+
kvmppc_e500_ref_setup(ref, gtlbe, pfn);
kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
@@ -457,10 +470,13 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500,
/* Clear i-cache for new pages */
kvmppc_mmu_flush_icache(pfn);
+out:
+ spin_unlock(&kvm->mmu_lock);
+
/* Drop refcount on page, so that mmu notifiers can clear it */
kvm_release_pfn_clean(pfn);
- return 0;
+ return ret;
}
/* XXX only map the one-one case, for now use TLB0 */
^ permalink raw reply related [flat|nested] 20+ messages in thread* RE: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
2013-10-07 12:04 ` Paolo Bonzini
(?)
@ 2013-10-10 8:32 ` Bhushan Bharat-R65777
2013-10-10 9:01 ` Paolo Bonzini
-1 siblings, 1 reply; 20+ messages in thread
From: Bhushan Bharat-R65777 @ 2013-10-10 8:32 UTC (permalink / raw)
To: Paolo Bonzini, Alexander Graf
Cc: Paul Mackerras, Wood Scott-B07421, kvm-ppc@vger.kernel.org,
kvm@vger.kernel.org mailing list, Gleb Natapov
> -----Original Message-----
> From: Paolo Bonzini [mailto:paolo.bonzini@gmail.com] On Behalf Of Paolo Bonzini
> Sent: Monday, October 07, 2013 5:35 PM
> To: Alexander Graf
> Cc: Bhushan Bharat-R65777; Paul Mackerras; Wood Scott-B07421; kvm-
> ppc@vger.kernel.org; kvm@vger.kernel.org mailing list; Bhushan Bharat-R65777;
> Gleb Natapov
> Subject: Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress
> on page setup
>
> Il 04/10/2013 15:38, Alexander Graf ha scritto:
> >
> > On 07.08.2013, at 12:03, Bharat Bhushan wrote:
> >
> >> When the MM code is invalidating a range of pages, it calls the KVM
> >> kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
> >> kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
> >> However, the Linux PTEs for the range being flushed are still valid at
> >> that point. We are not supposed to establish any new references to pages
> >> in the range until the ...range_end() notifier gets called.
> >> The PPC-specific KVM code doesn't get any explicit notification of that;
> >> instead, we are supposed to use mmu_notifier_retry() to test whether we
> >> are or have been inside a range flush notifier pair while we have been
> >> referencing a page.
> >>
> >> This patch calls the mmu_notifier_retry() while mapping the guest
> >> page to ensure we are not referencing a page when in range invalidation.
> >>
> >> This call is inside a region locked with kvm->mmu_lock, which is the
> >> same lock that is called by the KVM MMU notifier functions, thus
> >> ensuring that no new notification can proceed while we are in the
> >> locked region.
> >>
> >> Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
> >
> > Acked-by: Alexander Graf <agraf@suse.de>
> >
> > Gleb, Paolo, please queue for 3.12 directly.
>
> Here is the backport. The second hunk has a nontrivial conflict, so
> someone please give their {Tested,Reviewed,Compiled}-by.
{Compiled,Reviewed}-by: Bharat Bhushan <bharat.bhushan@freescale.com>
Thanks
-Bharat
>
> Paolo
>
> diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
> index 1c6a9d7..c65593a 100644
> --- a/arch/powerpc/kvm/e500_mmu_host.c
> +++ b/arch/powerpc/kvm/e500_mmu_host.c
> @@ -332,6 +332,13 @@ static inline int kvmppc_e500_shadow_map(struct
> kvmppc_vcpu_e500 *vcpu_e500,
> unsigned long hva;
> int pfnmap = 0;
> int tsize = BOOK3E_PAGESZ_4K;
> + int ret = 0;
> + unsigned long mmu_seq;
> + struct kvm *kvm = vcpu_e500->vcpu.kvm;
> +
> + /* used to check for invalidations in progress */
> + mmu_seq = kvm->mmu_notifier_seq;
> + smp_rmb();
>
> /*
> * Translate guest physical to true physical, acquiring
> @@ -449,6 +456,12 @@ static inline int kvmppc_e500_shadow_map(struct
> kvmppc_vcpu_e500 *vcpu_e500,
> gvaddr &= ~((tsize_pages << PAGE_SHIFT) - 1);
> }
>
> + spin_lock(&kvm->mmu_lock);
> + if (mmu_notifier_retry(kvm, mmu_seq)) {
> + ret = -EAGAIN;
> + goto out;
> + }
> +
> kvmppc_e500_ref_setup(ref, gtlbe, pfn);
>
> kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
> @@ -457,10 +470,13 @@ static inline int kvmppc_e500_shadow_map(struct
> kvmppc_vcpu_e500 *vcpu_e500,
> /* Clear i-cache for new pages */
> kvmppc_mmu_flush_icache(pfn);
>
> +out:
> + spin_unlock(&kvm->mmu_lock);
> +
> /* Drop refcount on page, so that mmu notifiers can clear it */
> kvm_release_pfn_clean(pfn);
>
> - return 0;
> + return ret;
> }
>
> /* XXX only map the one-one case, for now use TLB0 */
>
>
^ permalink raw reply [flat|nested] 20+ messages in thread* Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
2013-10-10 8:32 ` Bhushan Bharat-R65777
@ 2013-10-10 9:01 ` Paolo Bonzini
0 siblings, 0 replies; 20+ messages in thread
From: Paolo Bonzini @ 2013-10-10 9:01 UTC (permalink / raw)
To: Bhushan Bharat-R65777
Cc: Alexander Graf, Paul Mackerras, Wood Scott-B07421,
kvm-ppc@vger.kernel.org, kvm@vger.kernel.org mailing list,
Gleb Natapov
Il 10/10/2013 10:32, Bhushan Bharat-R65777 ha scritto:
>
>
>> -----Original Message-----
>> From: Paolo Bonzini [mailto:paolo.bonzini@gmail.com] On Behalf Of Paolo Bonzini
>> Sent: Monday, October 07, 2013 5:35 PM
>> To: Alexander Graf
>> Cc: Bhushan Bharat-R65777; Paul Mackerras; Wood Scott-B07421; kvm-
>> ppc@vger.kernel.org; kvm@vger.kernel.org mailing list; Bhushan Bharat-R65777;
>> Gleb Natapov
>> Subject: Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress
>> on page setup
>>
>> Il 04/10/2013 15:38, Alexander Graf ha scritto:
>>>
>>> On 07.08.2013, at 12:03, Bharat Bhushan wrote:
>>>
>>>> When the MM code is invalidating a range of pages, it calls the KVM
>>>> kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
>>>> kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
>>>> However, the Linux PTEs for the range being flushed are still valid at
>>>> that point. We are not supposed to establish any new references to pages
>>>> in the range until the ...range_end() notifier gets called.
>>>> The PPC-specific KVM code doesn't get any explicit notification of that;
>>>> instead, we are supposed to use mmu_notifier_retry() to test whether we
>>>> are or have been inside a range flush notifier pair while we have been
>>>> referencing a page.
>>>>
>>>> This patch calls the mmu_notifier_retry() while mapping the guest
>>>> page to ensure we are not referencing a page when in range invalidation.
>>>>
>>>> This call is inside a region locked with kvm->mmu_lock, which is the
>>>> same lock that is called by the KVM MMU notifier functions, thus
>>>> ensuring that no new notification can proceed while we are in the
>>>> locked region.
>>>>
>>>> Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
>>>
>>> Acked-by: Alexander Graf <agraf@suse.de>
>>>
>>> Gleb, Paolo, please queue for 3.12 directly.
>>
>> Here is the backport. The second hunk has a nontrivial conflict, so
>> someone please give their {Tested,Reviewed,Compiled}-by.
>
> {Compiled,Reviewed}-by: Bharat Bhushan <bharat.bhushan@freescale.com>
Thanks, patch on its way to Linus.
Paolo
> Thanks
> -Bharat
>
>>
>> Paolo
>>
>> diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
>> index 1c6a9d7..c65593a 100644
>> --- a/arch/powerpc/kvm/e500_mmu_host.c
>> +++ b/arch/powerpc/kvm/e500_mmu_host.c
>> @@ -332,6 +332,13 @@ static inline int kvmppc_e500_shadow_map(struct
>> kvmppc_vcpu_e500 *vcpu_e500,
>> unsigned long hva;
>> int pfnmap = 0;
>> int tsize = BOOK3E_PAGESZ_4K;
>> + int ret = 0;
>> + unsigned long mmu_seq;
>> + struct kvm *kvm = vcpu_e500->vcpu.kvm;
>> +
>> + /* used to check for invalidations in progress */
>> + mmu_seq = kvm->mmu_notifier_seq;
>> + smp_rmb();
>>
>> /*
>> * Translate guest physical to true physical, acquiring
>> @@ -449,6 +456,12 @@ static inline int kvmppc_e500_shadow_map(struct
>> kvmppc_vcpu_e500 *vcpu_e500,
>> gvaddr &= ~((tsize_pages << PAGE_SHIFT) - 1);
>> }
>>
>> + spin_lock(&kvm->mmu_lock);
>> + if (mmu_notifier_retry(kvm, mmu_seq)) {
>> + ret = -EAGAIN;
>> + goto out;
>> + }
>> +
>> kvmppc_e500_ref_setup(ref, gtlbe, pfn);
>>
>> kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
>> @@ -457,10 +470,13 @@ static inline int kvmppc_e500_shadow_map(struct
>> kvmppc_vcpu_e500 *vcpu_e500,
>> /* Clear i-cache for new pages */
>> kvmppc_mmu_flush_icache(pfn);
>>
>> +out:
>> + spin_unlock(&kvm->mmu_lock);
>> +
>> /* Drop refcount on page, so that mmu notifiers can clear it */
>> kvm_release_pfn_clean(pfn);
>>
>> - return 0;
>> + return ret;
>> }
>>
>> /* XXX only map the one-one case, for now use TLB0 */
>>
>>
>
>
^ permalink raw reply [flat|nested] 20+ messages in thread* Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress on page setup
@ 2013-10-10 9:01 ` Paolo Bonzini
0 siblings, 0 replies; 20+ messages in thread
From: Paolo Bonzini @ 2013-10-10 9:01 UTC (permalink / raw)
To: Bhushan Bharat-R65777
Cc: Alexander Graf, Paul Mackerras, Wood Scott-B07421,
kvm-ppc@vger.kernel.org, kvm@vger.kernel.org mailing list,
Gleb Natapov
Il 10/10/2013 10:32, Bhushan Bharat-R65777 ha scritto:
>
>
>> -----Original Message-----
>> From: Paolo Bonzini [mailto:paolo.bonzini@gmail.com] On Behalf Of Paolo Bonzini
>> Sent: Monday, October 07, 2013 5:35 PM
>> To: Alexander Graf
>> Cc: Bhushan Bharat-R65777; Paul Mackerras; Wood Scott-B07421; kvm-
>> ppc@vger.kernel.org; kvm@vger.kernel.org mailing list; Bhushan Bharat-R65777;
>> Gleb Natapov
>> Subject: Re: [PATCH 2/2] kvm: ppc: booke: check range page invalidation progress
>> on page setup
>>
>> Il 04/10/2013 15:38, Alexander Graf ha scritto:
>>>
>>> On 07.08.2013, at 12:03, Bharat Bhushan wrote:
>>>
>>>> When the MM code is invalidating a range of pages, it calls the KVM
>>>> kvm_mmu_notifier_invalidate_range_start() notifier function, which calls
>>>> kvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.
>>>> However, the Linux PTEs for the range being flushed are still valid at
>>>> that point. We are not supposed to establish any new references to pages
>>>> in the range until the ...range_end() notifier gets called.
>>>> The PPC-specific KVM code doesn't get any explicit notification of that;
>>>> instead, we are supposed to use mmu_notifier_retry() to test whether we
>>>> are or have been inside a range flush notifier pair while we have been
>>>> referencing a page.
>>>>
>>>> This patch calls the mmu_notifier_retry() while mapping the guest
>>>> page to ensure we are not referencing a page when in range invalidation.
>>>>
>>>> This call is inside a region locked with kvm->mmu_lock, which is the
>>>> same lock that is called by the KVM MMU notifier functions, thus
>>>> ensuring that no new notification can proceed while we are in the
>>>> locked region.
>>>>
>>>> Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
>>>
>>> Acked-by: Alexander Graf <agraf@suse.de>
>>>
>>> Gleb, Paolo, please queue for 3.12 directly.
>>
>> Here is the backport. The second hunk has a nontrivial conflict, so
>> someone please give their {Tested,Reviewed,Compiled}-by.
>
> {Compiled,Reviewed}-by: Bharat Bhushan <bharat.bhushan@freescale.com>
Thanks, patch on its way to Linus.
Paolo
> Thanks
> -Bharat
>
>>
>> Paolo
>>
>> diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c
>> index 1c6a9d7..c65593a 100644
>> --- a/arch/powerpc/kvm/e500_mmu_host.c
>> +++ b/arch/powerpc/kvm/e500_mmu_host.c
>> @@ -332,6 +332,13 @@ static inline int kvmppc_e500_shadow_map(struct
>> kvmppc_vcpu_e500 *vcpu_e500,
>> unsigned long hva;
>> int pfnmap = 0;
>> int tsize = BOOK3E_PAGESZ_4K;
>> + int ret = 0;
>> + unsigned long mmu_seq;
>> + struct kvm *kvm = vcpu_e500->vcpu.kvm;
>> +
>> + /* used to check for invalidations in progress */
>> + mmu_seq = kvm->mmu_notifier_seq;
>> + smp_rmb();
>>
>> /*
>> * Translate guest physical to true physical, acquiring
>> @@ -449,6 +456,12 @@ static inline int kvmppc_e500_shadow_map(struct
>> kvmppc_vcpu_e500 *vcpu_e500,
>> gvaddr &= ~((tsize_pages << PAGE_SHIFT) - 1);
>> }
>>
>> + spin_lock(&kvm->mmu_lock);
>> + if (mmu_notifier_retry(kvm, mmu_seq)) {
>> + ret = -EAGAIN;
>> + goto out;
>> + }
>> +
>> kvmppc_e500_ref_setup(ref, gtlbe, pfn);
>>
>> kvmppc_e500_setup_stlbe(&vcpu_e500->vcpu, gtlbe, tsize,
>> @@ -457,10 +470,13 @@ static inline int kvmppc_e500_shadow_map(struct
>> kvmppc_vcpu_e500 *vcpu_e500,
>> /* Clear i-cache for new pages */
>> kvmppc_mmu_flush_icache(pfn);
>>
>> +out:
>> + spin_unlock(&kvm->mmu_lock);
>> +
>> /* Drop refcount on page, so that mmu notifiers can clear it */
>> kvm_release_pfn_clean(pfn);
>>
>> - return 0;
>> + return ret;
>> }
>>
>> /* XXX only map the one-one case, for now use TLB0 */
>>
>>
>
>
^ permalink raw reply [flat|nested] 20+ messages in thread