[PATCH RFC v3 0/9] KVM: allow dynamic kvm->vcpus array

[PATCH RFC v3 0/9] KVM: allow dynamic kvm->vcpus array

[Radim Krčmář]

The only common part with v2 is [v3 5/9].

The crucial part of this series is adding a separate mechanism for
kvm_for_each_vcpu() [v3 8/9] and with that change, I think that the
dynamic array [v3 9/9] would be nicer if protected by RCU, like in v2:
The protection can be nicely hidden in kvm_get_vcpu().  I just had the
split done before implementing [v3 8/9] and presented it for
consideration.

Smoke tested on x86 only.


Radim Kr?m?? (9):
  KVM: s390: optimize detection of started vcpus
  KVM: arm/arm64: fix vcpu self-detection in vgic_v3_dispatch_sgi()
  KVM: remember position in kvm->vcpus array
  KVM: arm/arm64: use locking helpers in kvm_vgic_create()
  KVM: remove unused __KVM_HAVE_ARCH_VM_ALLOC
  KVM: rework kvm_vcpu_on_spin loop
  KVM: add kvm_free_vcpus and kvm_arch_free_vcpus
  KVM: implement kvm_for_each_vcpu with a list
  KVM: split kvm->vcpus into chunks

 arch/mips/kvm/mips.c                |  19 ++----
 arch/powerpc/kvm/book3s_32_mmu.c    |   3 +-
 arch/powerpc/kvm/book3s_64_mmu.c    |   3 +-
 arch/powerpc/kvm/book3s_hv.c        |   7 +-
 arch/powerpc/kvm/book3s_pr.c        |   5 +-
 arch/powerpc/kvm/book3s_xics.c      |   2 +-
 arch/powerpc/kvm/book3s_xics.h      |   3 +-
 arch/powerpc/kvm/book3s_xive.c      |  18 +++---
 arch/powerpc/kvm/book3s_xive.h      |   3 +-
 arch/powerpc/kvm/e500_emulate.c     |   3 +-
 arch/powerpc/kvm/powerpc.c          |  16 ++---
 arch/s390/include/asm/kvm_host.h    |   1 +
 arch/s390/kvm/interrupt.c           |   3 +-
 arch/s390/kvm/kvm-s390.c            |  77 ++++++++--------------
 arch/s390/kvm/kvm-s390.h            |   6 +-
 arch/s390/kvm/sigp.c                |   3 +-
 arch/x86/kvm/hyperv.c               |   3 +-
 arch/x86/kvm/i8254.c                |   3 +-
 arch/x86/kvm/i8259.c                |   7 +-
 arch/x86/kvm/ioapic.c               |   3 +-
 arch/x86/kvm/irq_comm.c             |  10 +--
 arch/x86/kvm/lapic.c                |   5 +-
 arch/x86/kvm/svm.c                  |   3 +-
 arch/x86/kvm/vmx.c                  |   5 +-
 arch/x86/kvm/x86.c                  |  34 ++++------
 include/linux/kvm_host.h            |  81 ++++++++++++-----------
 virt/kvm/arm/arch_timer.c           |  10 ++-
 virt/kvm/arm/arm.c                  |  25 ++++----
 virt/kvm/arm/pmu.c                  |   3 +-
 virt/kvm/arm/psci.c                 |   7 +-
 virt/kvm/arm/vgic/vgic-init.c       |  31 ++++-----
 virt/kvm/arm/vgic/vgic-kvm-device.c |  30 +++++----
 virt/kvm/arm/vgic/vgic-mmio-v2.c    |   5 +-
 virt/kvm/arm/vgic/vgic-mmio-v3.c    |  22 ++++---
 virt/kvm/arm/vgic/vgic.c            |   3 +-
 virt/kvm/kvm_main.c                 | 124 +++++++++++++++++++++++-------------
 36 files changed, 278 insertions(+), 308 deletions(-)

-- 
2.13.3

[PATCH RFC v3 1/9] KVM: s390: optimize detection of started vcpus

[Radim Krčmář]

We can add a variable instead of scanning all online VCPUs to know how
many are started.  We can't trivially tell which VCPU is the last one,
though.

Suggested-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
---
 arch/s390/include/asm/kvm_host.h |  1 +
 arch/s390/kvm/kvm-s390.c         | 39 +++++++++++++++------------------------
 2 files changed, 16 insertions(+), 24 deletions(-)

diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h
index a409d5991934..be0d0bdf585b 100644
--- a/arch/s390/include/asm/kvm_host.h
+++ b/arch/s390/include/asm/kvm_host.h
@@ -735,6 +735,7 @@ struct kvm_arch{
 	struct mutex ipte_mutex;
 	struct ratelimit_state sthyi_limit;
 	spinlock_t start_stop_lock;
+	unsigned started_vcpus;
 	struct sie_page2 *sie_page2;
 	struct kvm_s390_cpu_model model;
 	struct kvm_s390_crypto crypto;
diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
index 9f23a9e81a91..1534778a3c66 100644
--- a/arch/s390/kvm/kvm-s390.c
+++ b/arch/s390/kvm/kvm-s390.c
@@ -3414,25 +3414,17 @@ static void __enable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
 
 void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
 {
-	int i, online_vcpus, started_vcpus = 0;
-
 	if (!is_vcpu_stopped(vcpu))
 		return;
 
 	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 1);
 	/* Only one cpu at a time may enter/leave the STOPPED state. */
 	spin_lock(&vcpu->kvm->arch.start_stop_lock);
-	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
 
-	for (i = 0; i < online_vcpus; i++) {
-		if (!is_vcpu_stopped(vcpu->kvm->vcpus[i]))
-			started_vcpus++;
-	}
-
-	if (started_vcpus == 0) {
+	if (vcpu->kvm->arch.started_vcpus == 0) {
 		/* we're the only active VCPU -> speed it up */
 		__enable_ibs_on_vcpu(vcpu);
-	} else if (started_vcpus == 1) {
+	} else if (vcpu->kvm->arch.started_vcpus == 1) {
 		/*
 		 * As we are starting a second VCPU, we have to disable
 		 * the IBS facility on all VCPUs to remove potentially
@@ -3441,6 +3433,8 @@ void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
 		__disable_ibs_on_all_vcpus(vcpu->kvm);
 	}
 
+	vcpu->kvm->arch.started_vcpus++;
+
 	atomic_andnot(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags);
 	/*
 	 * Another VCPU might have used IBS while we were offline.
@@ -3453,16 +3447,12 @@ void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
 
 void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
 {
-	int i, online_vcpus, started_vcpus = 0;
-	struct kvm_vcpu *started_vcpu = NULL;
-
 	if (is_vcpu_stopped(vcpu))
 		return;
 
 	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 0);
 	/* Only one cpu at a time may enter/leave the STOPPED state. */
 	spin_lock(&vcpu->kvm->arch.start_stop_lock);
-	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
 
 	/* SIGP STOP and SIGP STOP AND STORE STATUS has been fully processed */
 	kvm_s390_clear_stop_irq(vcpu);
@@ -3470,19 +3460,20 @@ void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
 	atomic_or(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags);
 	__disable_ibs_on_vcpu(vcpu);
 
-	for (i = 0; i < online_vcpus; i++) {
-		if (!is_vcpu_stopped(vcpu->kvm->vcpus[i])) {
-			started_vcpus++;
-			started_vcpu = vcpu->kvm->vcpus[i];
-		}
-	}
+	vcpu->kvm->arch.started_vcpus--;
+
+	if (vcpu->kvm->arch.started_vcpus == 1) {
+		struct kvm_vcpu *started_vcpu;
 
-	if (started_vcpus == 1) {
 		/*
-		 * As we only have one VCPU left, we want to enable the
-		 * IBS facility for that VCPU to speed it up.
+		 * As we only have one VCPU left, we want to enable the IBS
+		 * facility for that VCPU to speed it up.
 		 */
-		__enable_ibs_on_vcpu(started_vcpu);
+		kvm_for_each_vcpu(i, started_vcpu, vcpu->kvm)
+			if (!is_vcpu_stopped(started_vcpu)) {
+				__enable_ibs_on_vcpu(started_vcpu);
+				break;
+			}
 	}
 
 	spin_unlock(&vcpu->kvm->arch.start_stop_lock);
-- 
2.13.3

[PATCH RFC v3 1/9] KVM: s390: optimize detection of started vcpus

[Christian Borntraeger]

On 08/21/2017 10:35 PM, Radim Kr?m?? wrote:
> We can add a variable instead of scanning all online VCPUs to know how
> many are started.  We can't trivially tell which VCPU is the last one,
> though.
> 
> Suggested-by: Cornelia Huck <cohuck@redhat.com>
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>



> @@ -3453,16 +3447,12 @@ void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
> 
>  void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
>  {
> -	int i, online_vcpus, started_vcpus = 0;

here you remove i.

> -	struct kvm_vcpu *started_vcpu = NULL;
> -
>  	if (is_vcpu_stopped(vcpu))
>  		return;
> 
>  	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 0);
>  	/* Only one cpu at a time may enter/leave the STOPPED state. */
>  	spin_lock(&vcpu->kvm->arch.start_stop_lock);
> -	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
> 
>  	/* SIGP STOP and SIGP STOP AND STORE STATUS has been fully processed */
>  	kvm_s390_clear_stop_irq(vcpu);
> @@ -3470,19 +3460,20 @@ void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
>  	atomic_or(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags);
>  	__disable_ibs_on_vcpu(vcpu);
> 
> -	for (i = 0; i < online_vcpus; i++) {
> -		if (!is_vcpu_stopped(vcpu->kvm->vcpus[i])) {
> -			started_vcpus++;
> -			started_vcpu = vcpu->kvm->vcpus[i];
> -		}
> -	}
> +	vcpu->kvm->arch.started_vcpus--;
> +
> +	if (vcpu->kvm->arch.started_vcpus == 1) {
> +		struct kvm_vcpu *started_vcpu;
> 
> -	if (started_vcpus == 1) {
>  		/*
> -		 * As we only have one VCPU left, we want to enable the
> -		 * IBS facility for that VCPU to speed it up.
> +		 * As we only have one VCPU left, we want to enable the IBS
> +		 * facility for that VCPU to speed it up.
>  		 */
> -		__enable_ibs_on_vcpu(started_vcpu);
> +		kvm_for_each_vcpu(i, started_vcpu, vcpu->kvm)

here you need i.
> +			if (!is_vcpu_stopped(started_vcpu)) {
> +				__enable_ibs_on_vcpu(started_vcpu);
> +				break;
> +			}
>  	}
> 
>  	spin_unlock(&vcpu->kvm->arch.start_stop_lock);
> 

[PATCH RFC v3 1/9] KVM: s390: optimize detection of started vcpus

[David Hildenbrand]

On 21.08.2017 22:35, Radim Kr?m?? wrote:
> We can add a variable instead of scanning all online VCPUs to know how
> many are started.  We can't trivially tell which VCPU is the last one,
> though.

You could keep the started vcpus in a list. Then you might drop unsigned
started_vcpus;

No started vcpus: Start pointer NULL
Single started vcpu: Only one element in the list (easy to check)
> 1 started vcpus: More than one element int he list (easy to check)


> 
> Suggested-by: Cornelia Huck <cohuck@redhat.com>
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> ---
>  arch/s390/include/asm/kvm_host.h |  1 +
>  arch/s390/kvm/kvm-s390.c         | 39 +++++++++++++++------------------------
>  2 files changed, 16 insertions(+), 24 deletions(-)
> 
> diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h
> index a409d5991934..be0d0bdf585b 100644
> --- a/arch/s390/include/asm/kvm_host.h
> +++ b/arch/s390/include/asm/kvm_host.h
> @@ -735,6 +735,7 @@ struct kvm_arch{
>  	struct mutex ipte_mutex;
>  	struct ratelimit_state sthyi_limit;
>  	spinlock_t start_stop_lock;
> +	unsigned started_vcpus;
>  	struct sie_page2 *sie_page2;
>  	struct kvm_s390_cpu_model model;
>  	struct kvm_s390_crypto crypto;
> diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
> index 9f23a9e81a91..1534778a3c66 100644
> --- a/arch/s390/kvm/kvm-s390.c
> +++ b/arch/s390/kvm/kvm-s390.c
> @@ -3414,25 +3414,17 @@ static void __enable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
>  
>  void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
>  {
> -	int i, online_vcpus, started_vcpus = 0;
> -
>  	if (!is_vcpu_stopped(vcpu))
>  		return;
>  
>  	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 1);
>  	/* Only one cpu at a time may enter/leave the STOPPED state. */
>  	spin_lock(&vcpu->kvm->arch.start_stop_lock);
> -	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
>  
> -	for (i = 0; i < online_vcpus; i++) {
> -		if (!is_vcpu_stopped(vcpu->kvm->vcpus[i]))
> -			started_vcpus++;
> -	}
> -
> -	if (started_vcpus == 0) {
> +	if (vcpu->kvm->arch.started_vcpus == 0) {
>  		/* we're the only active VCPU -> speed it up */
>  		__enable_ibs_on_vcpu(vcpu);
> -	} else if (started_vcpus == 1) {
> +	} else if (vcpu->kvm->arch.started_vcpus == 1) {
>  		/*
>  		 * As we are starting a second VCPU, we have to disable
>  		 * the IBS facility on all VCPUs to remove potentially
> @@ -3441,6 +3433,8 @@ void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
>  		__disable_ibs_on_all_vcpus(vcpu->kvm);
>  	}
>  
> +	vcpu->kvm->arch.started_vcpus++;
> +
>  	atomic_andnot(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags);
>  	/*
>  	 * Another VCPU might have used IBS while we were offline.
> @@ -3453,16 +3447,12 @@ void kvm_s390_vcpu_start(struct kvm_vcpu *vcpu)
>  
>  void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
>  {
> -	int i, online_vcpus, started_vcpus = 0;
> -	struct kvm_vcpu *started_vcpu = NULL;
> -
>  	if (is_vcpu_stopped(vcpu))
>  		return;
>  
>  	trace_kvm_s390_vcpu_start_stop(vcpu->vcpu_id, 0);
>  	/* Only one cpu at a time may enter/leave the STOPPED state. */
>  	spin_lock(&vcpu->kvm->arch.start_stop_lock);
> -	online_vcpus = atomic_read(&vcpu->kvm->online_vcpus);
>  
>  	/* SIGP STOP and SIGP STOP AND STORE STATUS has been fully processed */
>  	kvm_s390_clear_stop_irq(vcpu);
> @@ -3470,19 +3460,20 @@ void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
>  	atomic_or(CPUSTAT_STOPPED, &vcpu->arch.sie_block->cpuflags);
>  	__disable_ibs_on_vcpu(vcpu);
>  
> -	for (i = 0; i < online_vcpus; i++) {
> -		if (!is_vcpu_stopped(vcpu->kvm->vcpus[i])) {
> -			started_vcpus++;
> -			started_vcpu = vcpu->kvm->vcpus[i];
> -		}
> -	}
> +	vcpu->kvm->arch.started_vcpus--;
> +
> +	if (vcpu->kvm->arch.started_vcpus == 1) {
> +		struct kvm_vcpu *started_vcpu;
>  
> -	if (started_vcpus == 1) {
>  		/*
> -		 * As we only have one VCPU left, we want to enable the
> -		 * IBS facility for that VCPU to speed it up.
> +		 * As we only have one VCPU left, we want to enable the IBS
> +		 * facility for that VCPU to speed it up.
>  		 */
> -		__enable_ibs_on_vcpu(started_vcpu);
> +		kvm_for_each_vcpu(i, started_vcpu, vcpu->kvm)
> +			if (!is_vcpu_stopped(started_vcpu)) {
> +				__enable_ibs_on_vcpu(started_vcpu);
> +				break;
> +			}
>  	}
>  
>  	spin_unlock(&vcpu->kvm->arch.start_stop_lock);
> 


-- 

Thanks,

David

[PATCH RFC v3 1/9] KVM: s390: optimize detection of started vcpus

[Cornelia Huck]

On Tue, 22 Aug 2017 13:31:27 +0200
David Hildenbrand <david@redhat.com> wrote:

> On 21.08.2017 22:35, Radim Kr?m?? wrote:
> > We can add a variable instead of scanning all online VCPUs to know how
> > many are started.  We can't trivially tell which VCPU is the last one,
> > though.  
> 
> You could keep the started vcpus in a list. Then you might drop unsigned
> started_vcpus;
> 
> No started vcpus: Start pointer NULL
> Single started vcpu: Only one element in the list (easy to check)
> > 1 started vcpus: More than one element int he list (easy to check)  

I'm not sure the added complication of keeping a list buys us much
here: We only have the "look for the last vcpu not stopped" operation
for the 2->1 transition.

[PATCH RFC v3 1/9] KVM: s390: optimize detection of started vcpus

[David Hildenbrand]

On 29.08.2017 13:23, Cornelia Huck wrote:
> On Tue, 22 Aug 2017 13:31:27 +0200
> David Hildenbrand <david@redhat.com> wrote:
> 
>> On 21.08.2017 22:35, Radim Kr?m?? wrote:
>>> We can add a variable instead of scanning all online VCPUs to know how
>>> many are started.  We can't trivially tell which VCPU is the last one,
>>> though.  
>>
>> You could keep the started vcpus in a list. Then you might drop unsigned
>> started_vcpus;
>>
>> No started vcpus: Start pointer NULL
>> Single started vcpu: Only one element in the list (easy to check)
>>> 1 started vcpus: More than one element int he list (easy to check)  
> 
> I'm not sure the added complication of keeping a list buys us much
> here: We only have the "look for the last vcpu not stopped" operation
> for the 2->1 transition.
> 

That is wrong, we also have to know the last remaining (started) VCPU.
For that, right now we have to iterate over all VCPUs.

There shouldn't be much complexity. We already perform changes under a
lock, so it is as simple as adding/removing from the list.

Detecting the transitions boils down to looking at two pointers.

-- 

Thanks,

David

[PATCH RFC v3 1/9] KVM: s390: optimize detection of started vcpus

[Cornelia Huck]

On Tue, 29 Aug 2017 14:05:40 +0200
David Hildenbrand <david@redhat.com> wrote:

> On 29.08.2017 13:23, Cornelia Huck wrote:
> > On Tue, 22 Aug 2017 13:31:27 +0200
> > David Hildenbrand <david@redhat.com> wrote:
> >   
> >> On 21.08.2017 22:35, Radim Kr?m?? wrote:  
> >>> We can add a variable instead of scanning all online VCPUs to know how
> >>> many are started.  We can't trivially tell which VCPU is the last one,
> >>> though.    
> >>
> >> You could keep the started vcpus in a list. Then you might drop unsigned
> >> started_vcpus;
> >>
> >> No started vcpus: Start pointer NULL
> >> Single started vcpu: Only one element in the list (easy to check)  
> >>> 1 started vcpus: More than one element int he list (easy to check)    
> > 
> > I'm not sure the added complication of keeping a list buys us much
> > here: We only have the "look for the last vcpu not stopped" operation
> > for the 2->1 transition.
> >   
> 
> That is wrong, we also have to know the last remaining (started) VCPU.
> For that, right now we have to iterate over all VCPUs.

Yes, but that's only for the 2->1 transition... and all in all that is
still much better that before.

> 
> There shouldn't be much complexity. We already perform changes under a
> lock, so it is as simple as adding/removing from the list.
> 
> Detecting the transitions boils down to looking at two pointers.

Having a private vcpu list feels a bit wrong... and I don't really see
much benefit.

[PATCH RFC v3 1/9] KVM: s390: optimize detection of started vcpus

[Cornelia Huck]

On Mon, 21 Aug 2017 22:35:22 +0200
Radim Kr?m?? <rkrcmar@redhat.com> wrote:

> We can add a variable instead of scanning all online VCPUs to know how
> many are started.  We can't trivially tell which VCPU is the last one,
> though.
> 
> Suggested-by: Cornelia Huck <cohuck@redhat.com>
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> ---
>  arch/s390/include/asm/kvm_host.h |  1 +
>  arch/s390/kvm/kvm-s390.c         | 39 +++++++++++++++------------------------
>  2 files changed, 16 insertions(+), 24 deletions(-)

Modulo the bug noticed by Christian, I think this looks good.

[PATCH RFC v3 2/9] KVM: arm/arm64: fix vcpu self-detection in vgic_v3_dispatch_sgi()

[Radim Krčmář]

The index in kvm->vcpus array and vcpu->vcpu_id are very different
things.  Comparing struct kvm_vcpu pointers is a sure way to know.

Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
---
 virt/kvm/arm/vgic/vgic-mmio-v3.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/virt/kvm/arm/vgic/vgic-mmio-v3.c b/virt/kvm/arm/vgic/vgic-mmio-v3.c
index 408ef06638fc..9d4b69b766ec 100644
--- a/virt/kvm/arm/vgic/vgic-mmio-v3.c
+++ b/virt/kvm/arm/vgic/vgic-mmio-v3.c
@@ -797,7 +797,6 @@ void vgic_v3_dispatch_sgi(struct kvm_vcpu *vcpu, u64 reg)
 	u16 target_cpus;
 	u64 mpidr;
 	int sgi, c;
-	int vcpu_id = vcpu->vcpu_id;
 	bool broadcast;
 
 	sgi = (reg & ICC_SGI1R_SGI_ID_MASK) >> ICC_SGI1R_SGI_ID_SHIFT;
@@ -821,7 +820,7 @@ void vgic_v3_dispatch_sgi(struct kvm_vcpu *vcpu, u64 reg)
 			break;
 
 		/* Don't signal the calling VCPU */
-		if (broadcast && c == vcpu_id)
+		if (broadcast && c_vcpu == vcpu)
 			continue;
 
 		if (!broadcast) {
-- 
2.13.3

[PATCH RFC v3 2/9] KVM: arm/arm64: fix vcpu self-detection in vgic_v3_dispatch_sgi()

[David Hildenbrand]

On 21.08.2017 22:35, Radim Kr?m?? wrote:
> The index in kvm->vcpus array and vcpu->vcpu_id are very different
> things.  Comparing struct kvm_vcpu pointers is a sure way to know.
> 
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> ---
>  virt/kvm/arm/vgic/vgic-mmio-v3.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/virt/kvm/arm/vgic/vgic-mmio-v3.c b/virt/kvm/arm/vgic/vgic-mmio-v3.c
> index 408ef06638fc..9d4b69b766ec 100644
> --- a/virt/kvm/arm/vgic/vgic-mmio-v3.c
> +++ b/virt/kvm/arm/vgic/vgic-mmio-v3.c
> @@ -797,7 +797,6 @@ void vgic_v3_dispatch_sgi(struct kvm_vcpu *vcpu, u64 reg)
>  	u16 target_cpus;
>  	u64 mpidr;
>  	int sgi, c;
> -	int vcpu_id = vcpu->vcpu_id;
>  	bool broadcast;
>  
>  	sgi = (reg & ICC_SGI1R_SGI_ID_MASK) >> ICC_SGI1R_SGI_ID_SHIFT;
> @@ -821,7 +820,7 @@ void vgic_v3_dispatch_sgi(struct kvm_vcpu *vcpu, u64 reg)
>  			break;
>  
>  		/* Don't signal the calling VCPU */
> -		if (broadcast && c == vcpu_id)
> +		if (broadcast && c_vcpu == vcpu)
>  			continue;
>  
>  		if (!broadcast) {
> 

Yes, this looks correct to me.

Reviewed-by: David Hildenbrand <david@redhat.com>

-- 

Thanks,

David

[PATCH RFC v3 2/9] KVM: arm/arm64: fix vcpu self-detection in vgic_v3_dispatch_sgi()

[Christoffer Dall]

On Mon, Aug 21, 2017 at 10:35:23PM +0200, Radim Kr?m?? wrote:
> The index in kvm->vcpus array and vcpu->vcpu_id are very different
> things.  Comparing struct kvm_vcpu pointers is a sure way to know.
> 
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>

Acked-by: Christoffer Dall <cdall@linaro.org>

> ---
>  virt/kvm/arm/vgic/vgic-mmio-v3.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/virt/kvm/arm/vgic/vgic-mmio-v3.c b/virt/kvm/arm/vgic/vgic-mmio-v3.c
> index 408ef06638fc..9d4b69b766ec 100644
> --- a/virt/kvm/arm/vgic/vgic-mmio-v3.c
> +++ b/virt/kvm/arm/vgic/vgic-mmio-v3.c
> @@ -797,7 +797,6 @@ void vgic_v3_dispatch_sgi(struct kvm_vcpu *vcpu, u64 reg)
>  	u16 target_cpus;
>  	u64 mpidr;
>  	int sgi, c;
> -	int vcpu_id = vcpu->vcpu_id;
>  	bool broadcast;
>  
>  	sgi = (reg & ICC_SGI1R_SGI_ID_MASK) >> ICC_SGI1R_SGI_ID_SHIFT;
> @@ -821,7 +820,7 @@ void vgic_v3_dispatch_sgi(struct kvm_vcpu *vcpu, u64 reg)
>  			break;
>  
>  		/* Don't signal the calling VCPU */
> -		if (broadcast && c == vcpu_id)
> +		if (broadcast && c_vcpu == vcpu)
>  			continue;
>  
>  		if (!broadcast) {
> -- 
> 2.13.3
> 

[PATCH RFC v3 3/9] KVM: remember position in kvm->vcpus array

[Radim Krčmář]

Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
---
 include/linux/kvm_host.h | 11 +++--------
 virt/kvm/kvm_main.c      |  5 ++++-
 2 files changed, 7 insertions(+), 9 deletions(-)

diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 6882538eda32..a8ff956616d2 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -220,7 +220,8 @@ struct kvm_vcpu {
 	struct preempt_notifier preempt_notifier;
 #endif
 	int cpu;
-	int vcpu_id;
+	int vcpu_id; /* id given by userspace at creation */
+	int vcpus_idx; /* index in kvm->vcpus array */
 	int srcu_idx;
 	int mode;
 	unsigned long requests;
@@ -516,13 +517,7 @@ static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
 
 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
 {
-	struct kvm_vcpu *tmp;
-	int idx;
-
-	kvm_for_each_vcpu(idx, tmp, vcpu->kvm)
-		if (tmp == vcpu)
-			return idx;
-	BUG();
+	return vcpu->vcpus_idx;
 }
 
 #define kvm_for_each_memslot(memslot, slots)	\
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index e17c40d986f3..caf8323f7df7 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -2498,7 +2498,10 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
 		goto unlock_vcpu_destroy;
 	}
 
-	BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
+	vcpu->vcpus_idx = atomic_read(&kvm->online_vcpus);
+
+	BUG_ON(kvm->vcpus[vcpu->vcpus_idx]);
+
 
 	/* Now it's all set up, let userspace reach it */
 	kvm_get_kvm(kvm);
-- 
2.13.3

[PATCH RFC v3 3/9] KVM: remember position in kvm->vcpus array

[David Hildenbrand]

On 21.08.2017 22:35, Radim Kr?m?? wrote:
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> ---
>  include/linux/kvm_host.h | 11 +++--------
>  virt/kvm/kvm_main.c      |  5 ++++-
>  2 files changed, 7 insertions(+), 9 deletions(-)
> 
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index 6882538eda32..a8ff956616d2 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -220,7 +220,8 @@ struct kvm_vcpu {
>  	struct preempt_notifier preempt_notifier;
>  #endif
>  	int cpu;
> -	int vcpu_id;
> +	int vcpu_id; /* id given by userspace at creation */
> +	int vcpus_idx; /* index in kvm->vcpus array */
>  	int srcu_idx;
>  	int mode;
>  	unsigned long requests;
> @@ -516,13 +517,7 @@ static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
>  
>  static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
>  {
> -	struct kvm_vcpu *tmp;
> -	int idx;
> -
> -	kvm_for_each_vcpu(idx, tmp, vcpu->kvm)
> -		if (tmp == vcpu)
> -			return idx;
> -	BUG();
> +	return vcpu->vcpus_idx;
>  }
>  
>  #define kvm_for_each_memslot(memslot, slots)	\
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index e17c40d986f3..caf8323f7df7 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -2498,7 +2498,10 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
>  		goto unlock_vcpu_destroy;
>  	}
>  
> -	BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
> +	vcpu->vcpus_idx = atomic_read(&kvm->online_vcpus);
> +
> +	BUG_ON(kvm->vcpus[vcpu->vcpus_idx]);
> +
>  
>  	/* Now it's all set up, let userspace reach it */
>  	kvm_get_kvm(kvm);
> 

Reviewed-by: David Hildenbrand <david@redhat.com>

-- 

Thanks,

David

[PATCH RFC v3 3/9] KVM: remember position in kvm->vcpus array

[Cornelia Huck]

On Mon, 21 Aug 2017 22:35:24 +0200
Radim Kr?m?? <rkrcmar@redhat.com> wrote:

> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> ---
>  include/linux/kvm_host.h | 11 +++--------
>  virt/kvm/kvm_main.c      |  5 ++++-
>  2 files changed, 7 insertions(+), 9 deletions(-)
> 
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index 6882538eda32..a8ff956616d2 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -220,7 +220,8 @@ struct kvm_vcpu {
>  	struct preempt_notifier preempt_notifier;
>  #endif
>  	int cpu;
> -	int vcpu_id;
> +	int vcpu_id; /* id given by userspace at creation */
> +	int vcpus_idx; /* index in kvm->vcpus array */

I predict confusion in the future over which is which... maybe
s/vcpus_idx/vcpus_array_pos/ or something like that?

>  	int srcu_idx;
>  	int mode;
>  	unsigned long requests;
> @@ -516,13 +517,7 @@ static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
>  
>  static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
>  {
> -	struct kvm_vcpu *tmp;
> -	int idx;
> -
> -	kvm_for_each_vcpu(idx, tmp, vcpu->kvm)
> -		if (tmp == vcpu)
> -			return idx;
> -	BUG();
> +	return vcpu->vcpus_idx;
>  }
>  
>  #define kvm_for_each_memslot(memslot, slots)	\
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index e17c40d986f3..caf8323f7df7 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -2498,7 +2498,10 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
>  		goto unlock_vcpu_destroy;
>  	}
>  
> -	BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
> +	vcpu->vcpus_idx = atomic_read(&kvm->online_vcpus);
> +
> +	BUG_ON(kvm->vcpus[vcpu->vcpus_idx]);
> +
>  
>  	/* Now it's all set up, let userspace reach it */
>  	kvm_get_kvm(kvm);

I like the concept.

[PATCH RFC v3 4/9] KVM: arm/arm64: use locking helpers in kvm_vgic_create()

[Radim Krčmář]

No new VCPUs can be created because we are holding the kvm->lock.
This means that if we successfuly lock all VCPUs, we'll be unlocking the
same set and there is no need to do extra bookkeeping.

Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
---
 virt/kvm/arm/vgic/vgic-init.c       | 24 +++++++++---------------
 virt/kvm/arm/vgic/vgic-kvm-device.c |  6 +++++-
 2 files changed, 14 insertions(+), 16 deletions(-)

diff --git a/virt/kvm/arm/vgic/vgic-init.c b/virt/kvm/arm/vgic/vgic-init.c
index 5801261f3add..feb766f74c34 100644
--- a/virt/kvm/arm/vgic/vgic-init.c
+++ b/virt/kvm/arm/vgic/vgic-init.c
@@ -119,7 +119,7 @@ void kvm_vgic_vcpu_early_init(struct kvm_vcpu *vcpu)
  */
 int kvm_vgic_create(struct kvm *kvm, u32 type)
 {
-	int i, vcpu_lock_idx = -1, ret;
+	int i, ret;
 	struct kvm_vcpu *vcpu;
 
 	if (irqchip_in_kernel(kvm))
@@ -140,18 +140,14 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)
 	 * vcpu->mutex.  By grabbing the vcpu->mutex of all VCPUs we ensure
 	 * that no other VCPUs are run while we create the vgic.
 	 */
-	ret = -EBUSY;
-	kvm_for_each_vcpu(i, vcpu, kvm) {
-		if (!mutex_trylock(&vcpu->mutex))
-			goto out_unlock;
-		vcpu_lock_idx = i;
-	}
+	if (!lock_all_vcpus(kvm))
+		return -EBUSY;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
-		if (vcpu->arch.has_run_once)
+	kvm_for_each_vcpu(i, vcpu, kvm)
+		if (vcpu->arch.has_run_once) {
+			ret = -EBUSY;
 			goto out_unlock;
-	}
-	ret = 0;
+		}
 
 	if (type == KVM_DEV_TYPE_ARM_VGIC_V2)
 		kvm->arch.max_vcpus = VGIC_V2_MAX_CPUS;
@@ -176,11 +172,9 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)
 	kvm->arch.vgic.vgic_cpu_base = VGIC_ADDR_UNDEF;
 	kvm->arch.vgic.vgic_redist_base = VGIC_ADDR_UNDEF;
 
+	ret = 0;
 out_unlock:
-	for (; vcpu_lock_idx >= 0; vcpu_lock_idx--) {
-		vcpu = kvm_get_vcpu(kvm, vcpu_lock_idx);
-		mutex_unlock(&vcpu->mutex);
-	}
+	unlock_all_vcpus(kvm);
 	return ret;
 }
 
diff --git a/virt/kvm/arm/vgic/vgic-kvm-device.c b/virt/kvm/arm/vgic/vgic-kvm-device.c
index 10ae6f394b71..c5124737c7fc 100644
--- a/virt/kvm/arm/vgic/vgic-kvm-device.c
+++ b/virt/kvm/arm/vgic/vgic-kvm-device.c
@@ -270,7 +270,11 @@ static void unlock_vcpus(struct kvm *kvm, int vcpu_lock_idx)
 
 void unlock_all_vcpus(struct kvm *kvm)
 {
-	unlock_vcpus(kvm, atomic_read(&kvm->online_vcpus) - 1);
+	int i;
+	struct kvm_vcpu *tmp_vcpu;
+
+	kvm_for_each_vcpu(i, tmp_vcpu, kvm)
+		mutex_unlock(&tmp_vcpu->mutex);
 }
 
 /* Returns true if all vcpus were locked, false otherwise */
-- 
2.13.3

[PATCH RFC v3 4/9] KVM: arm/arm64: use locking helpers in kvm_vgic_create()

[David Hildenbrand]

On 21.08.2017 22:35, Radim Kr?m?? wrote:
> No new VCPUs can be created because we are holding the kvm->lock.
> This means that if we successfuly lock all VCPUs, we'll be unlocking the
> same set and there is no need to do extra bookkeeping.
> 
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> ---
>  virt/kvm/arm/vgic/vgic-init.c       | 24 +++++++++---------------
>  virt/kvm/arm/vgic/vgic-kvm-device.c |  6 +++++-
>  2 files changed, 14 insertions(+), 16 deletions(-)
> 
> diff --git a/virt/kvm/arm/vgic/vgic-init.c b/virt/kvm/arm/vgic/vgic-init.c
> index 5801261f3add..feb766f74c34 100644
> --- a/virt/kvm/arm/vgic/vgic-init.c
> +++ b/virt/kvm/arm/vgic/vgic-init.c
> @@ -119,7 +119,7 @@ void kvm_vgic_vcpu_early_init(struct kvm_vcpu *vcpu)
>   */
>  int kvm_vgic_create(struct kvm *kvm, u32 type)
>  {
> -	int i, vcpu_lock_idx = -1, ret;
> +	int i, ret;
>  	struct kvm_vcpu *vcpu;
>  
>  	if (irqchip_in_kernel(kvm))
> @@ -140,18 +140,14 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)
>  	 * vcpu->mutex.  By grabbing the vcpu->mutex of all VCPUs we ensure
>  	 * that no other VCPUs are run while we create the vgic.
>  	 */
> -	ret = -EBUSY;
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> -		if (!mutex_trylock(&vcpu->mutex))
> -			goto out_unlock;
> -		vcpu_lock_idx = i;
> -	}
> +	if (!lock_all_vcpus(kvm))
> +		return -EBUSY;

Yes, this makes sense.

>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> -		if (vcpu->arch.has_run_once)
> +	kvm_for_each_vcpu(i, vcpu, kvm)
> +		if (vcpu->arch.has_run_once) {
> +			ret = -EBUSY;
>  			goto out_unlock;
> -	}
> -	ret = 0;
> +		}

somehow I prefer keeping the {}

>  
>  	if (type == KVM_DEV_TYPE_ARM_VGIC_V2)
>  		kvm->arch.max_vcpus = VGIC_V2_MAX_CPUS;
> @@ -176,11 +172,9 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)
>  	kvm->arch.vgic.vgic_cpu_base = VGIC_ADDR_UNDEF;
>  	kvm->arch.vgic.vgic_redist_base = VGIC_ADDR_UNDEF;
>  
> +	ret = 0;
>  out_unlock:
> -	for (; vcpu_lock_idx >= 0; vcpu_lock_idx--) {
> -		vcpu = kvm_get_vcpu(kvm, vcpu_lock_idx);
> -		mutex_unlock(&vcpu->mutex);
> -	}
> +	unlock_all_vcpus(kvm);
>  	return ret;
>  }
>  
> diff --git a/virt/kvm/arm/vgic/vgic-kvm-device.c b/virt/kvm/arm/vgic/vgic-kvm-device.c
> index 10ae6f394b71..c5124737c7fc 100644
> --- a/virt/kvm/arm/vgic/vgic-kvm-device.c
> +++ b/virt/kvm/arm/vgic/vgic-kvm-device.c
> @@ -270,7 +270,11 @@ static void unlock_vcpus(struct kvm *kvm, int vcpu_lock_idx)
>  
>  void unlock_all_vcpus(struct kvm *kvm)
>  {
> -	unlock_vcpus(kvm, atomic_read(&kvm->online_vcpus) - 1);
> +	int i;
> +	struct kvm_vcpu *tmp_vcpu;
> +
> +	kvm_for_each_vcpu(i, tmp_vcpu, kvm)
> +		mutex_unlock(&tmp_vcpu->mutex);
>  }
>  
>  /* Returns true if all vcpus were locked, false otherwise */
> 

Looks sane to me.

-- 

Thanks,

David

[PATCH RFC v3 4/9] KVM: arm/arm64: use locking helpers in kvm_vgic_create()

[Christoffer Dall]

On Mon, Aug 21, 2017 at 10:35:25PM +0200, Radim Kr?m?? wrote:
> No new VCPUs can be created because we are holding the kvm->lock.
> This means that if we successfuly lock all VCPUs, we'll be unlocking the
> same set and there is no need to do extra bookkeeping.
> 
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> ---
>  virt/kvm/arm/vgic/vgic-init.c       | 24 +++++++++---------------
>  virt/kvm/arm/vgic/vgic-kvm-device.c |  6 +++++-
>  2 files changed, 14 insertions(+), 16 deletions(-)
> 
> diff --git a/virt/kvm/arm/vgic/vgic-init.c b/virt/kvm/arm/vgic/vgic-init.c
> index 5801261f3add..feb766f74c34 100644
> --- a/virt/kvm/arm/vgic/vgic-init.c
> +++ b/virt/kvm/arm/vgic/vgic-init.c
> @@ -119,7 +119,7 @@ void kvm_vgic_vcpu_early_init(struct kvm_vcpu *vcpu)
>   */
>  int kvm_vgic_create(struct kvm *kvm, u32 type)
>  {
> -	int i, vcpu_lock_idx = -1, ret;
> +	int i, ret;
>  	struct kvm_vcpu *vcpu;
>  
>  	if (irqchip_in_kernel(kvm))
> @@ -140,18 +140,14 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)
>  	 * vcpu->mutex.  By grabbing the vcpu->mutex of all VCPUs we ensure
>  	 * that no other VCPUs are run while we create the vgic.
>  	 */
> -	ret = -EBUSY;
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> -		if (!mutex_trylock(&vcpu->mutex))
> -			goto out_unlock;
> -		vcpu_lock_idx = i;
> -	}
> +	if (!lock_all_vcpus(kvm))
> +		return -EBUSY;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> -		if (vcpu->arch.has_run_once)
> +	kvm_for_each_vcpu(i, vcpu, kvm)
> +		if (vcpu->arch.has_run_once) {
> +			ret = -EBUSY;
>  			goto out_unlock;
> -	}
> -	ret = 0;
> +		}

I also prefer the additional brace here.

>  
>  	if (type == KVM_DEV_TYPE_ARM_VGIC_V2)
>  		kvm->arch.max_vcpus = VGIC_V2_MAX_CPUS;
> @@ -176,11 +172,9 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)
>  	kvm->arch.vgic.vgic_cpu_base = VGIC_ADDR_UNDEF;
>  	kvm->arch.vgic.vgic_redist_base = VGIC_ADDR_UNDEF;
>  
> +	ret = 0;
>  out_unlock:
> -	for (; vcpu_lock_idx >= 0; vcpu_lock_idx--) {
> -		vcpu = kvm_get_vcpu(kvm, vcpu_lock_idx);
> -		mutex_unlock(&vcpu->mutex);
> -	}
> +	unlock_all_vcpus(kvm);
>  	return ret;
>  }
>  
> diff --git a/virt/kvm/arm/vgic/vgic-kvm-device.c b/virt/kvm/arm/vgic/vgic-kvm-device.c
> index 10ae6f394b71..c5124737c7fc 100644
> --- a/virt/kvm/arm/vgic/vgic-kvm-device.c
> +++ b/virt/kvm/arm/vgic/vgic-kvm-device.c
> @@ -270,7 +270,11 @@ static void unlock_vcpus(struct kvm *kvm, int vcpu_lock_idx)
>  
>  void unlock_all_vcpus(struct kvm *kvm)
>  {
> -	unlock_vcpus(kvm, atomic_read(&kvm->online_vcpus) - 1);
> +	int i;
> +	struct kvm_vcpu *tmp_vcpu;
> +
> +	kvm_for_each_vcpu(i, tmp_vcpu, kvm)
> +		mutex_unlock(&tmp_vcpu->mutex);
>  }
>  
>  /* Returns true if all vcpus were locked, false otherwise */
> -- 
> 2.13.3
> 

As noted on the other patch, it looks a bit strange to modify
unlock_all_vcpus() here without also doing something about the error
path in lock_all_vcpus().

Otherwise this patch looks fine to me.

Thanks,
-Christoffer

[PATCH RFC v3 5/9] KVM: remove unused __KVM_HAVE_ARCH_VM_ALLOC

[Radim Krčmář]

Moving it to generic code will allow us to extend it with ease.
Christian noted that it was only used in the removed ia64.

Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
---
 include/linux/kvm_host.h | 12 ------------
 virt/kvm/kvm_main.c      | 16 +++++++++++++---
 2 files changed, 13 insertions(+), 15 deletions(-)

diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index a8ff956616d2..abd5cb1feb9e 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -798,18 +798,6 @@ int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
 
-#ifndef __KVM_HAVE_ARCH_VM_ALLOC
-static inline struct kvm *kvm_arch_alloc_vm(void)
-{
-	return kzalloc(sizeof(struct kvm), GFP_KERNEL);
-}
-
-static inline void kvm_arch_free_vm(struct kvm *kvm)
-{
-	kfree(kvm);
-}
-#endif
-
 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index caf8323f7df7..d89261d0d8c6 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -640,10 +640,20 @@ static int kvm_create_vm_debugfs(struct kvm *kvm, int fd)
 	return 0;
 }
 
+static inline struct kvm *kvm_alloc_vm(void)
+{
+	return kzalloc(sizeof(struct kvm), GFP_KERNEL);
+}
+
+static inline void kvm_free_vm(struct kvm *kvm)
+{
+	kfree(kvm);
+}
+
 static struct kvm *kvm_create_vm(unsigned long type)
 {
 	int r, i;
-	struct kvm *kvm = kvm_arch_alloc_vm();
+	struct kvm *kvm = kvm_alloc_vm();
 
 	if (!kvm)
 		return ERR_PTR(-ENOMEM);
@@ -720,7 +730,7 @@ static struct kvm *kvm_create_vm(unsigned long type)
 		kfree(kvm_get_bus(kvm, i));
 	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
 		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
-	kvm_arch_free_vm(kvm);
+	kvm_free_vm(kvm);
 	mmdrop(current->mm);
 	return ERR_PTR(r);
 }
@@ -771,7 +781,7 @@ static void kvm_destroy_vm(struct kvm *kvm)
 		kvm_free_memslots(kvm, __kvm_memslots(kvm, i));
 	cleanup_srcu_struct(&kvm->irq_srcu);
 	cleanup_srcu_struct(&kvm->srcu);
-	kvm_arch_free_vm(kvm);
+	kvm_free_vm(kvm);
 	preempt_notifier_dec();
 	hardware_disable_all();
 	mmdrop(mm);
-- 
2.13.3

[PATCH RFC v3 6/9] KVM: rework kvm_vcpu_on_spin loop

[Radim Krčmář]

The original code managed to obfuscate a straightforward idea:
start iterating from the selected index and reset the index to 0 when
reaching the end of online vcpus, then iterate until reaching the index
that we started at.

The resulting code is a bit better, IMO.  (Still horrible, though.)

Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
---
 include/linux/kvm_host.h | 13 +++++++++++++
 virt/kvm/kvm_main.c      | 47 ++++++++++++++++++-----------------------------
 2 files changed, 31 insertions(+), 29 deletions(-)

diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index abd5cb1feb9e..cfb3c0efdd51 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -498,6 +498,19 @@ static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 	     (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
 	     idx++)
 
+#define kvm_for_each_vcpu_from(idx, vcpup, from, kvm) \
+	for (idx = from, vcpup = kvm_get_vcpu(kvm, idx); \
+	     vcpup; \
+	     ({ \
+		idx++; \
+		if (idx >= atomic_read(&kvm->online_vcpus)) \
+			idx = 0; \
+		if (idx == from) \
+			vcpup = NULL; \
+		else \
+			vcpup = kvm_get_vcpu(kvm, idx); \
+	      }))
+
 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
 {
 	struct kvm_vcpu *vcpu = NULL;
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index d89261d0d8c6..33a15e176927 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -2333,8 +2333,7 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
 	struct kvm_vcpu *vcpu;
 	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
 	int yielded = 0;
-	int try = 3;
-	int pass;
+	int try = 2;
 	int i;
 
 	kvm_vcpu_set_in_spin_loop(me, true);
@@ -2345,34 +2344,24 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
 	 * VCPU is holding the lock that we need and will release it.
 	 * We approximate round-robin by starting at the last boosted VCPU.
 	 */
-	for (pass = 0; pass < 2 && !yielded && try; pass++) {
-		kvm_for_each_vcpu(i, vcpu, kvm) {
-			if (!pass && i <= last_boosted_vcpu) {
-				i = last_boosted_vcpu;
-				continue;
-			} else if (pass && i > last_boosted_vcpu)
-				break;
-			if (!ACCESS_ONCE(vcpu->preempted))
-				continue;
-			if (vcpu == me)
-				continue;
-			if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
-				continue;
-			if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
-				continue;
-			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
-				continue;
+	kvm_for_each_vcpu_from(i, vcpu, last_boosted_vcpu, kvm) {
+		if (!ACCESS_ONCE(vcpu->preempted))
+			continue;
+		if (vcpu == me)
+			continue;
+		if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
+			continue;
+		if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
+			continue;
+		if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
+			continue;
 
-			yielded = kvm_vcpu_yield_to(vcpu);
-			if (yielded > 0) {
-				kvm->last_boosted_vcpu = i;
-				break;
-			} else if (yielded < 0) {
-				try--;
-				if (!try)
-					break;
-			}
-		}
+		yielded = kvm_vcpu_yield_to(vcpu);
+		if (yielded > 0) {
+			kvm->last_boosted_vcpu = i;
+			break;
+		} else if (yielded < 0 && !try--)
+			break;
 	}
 	kvm_vcpu_set_in_spin_loop(me, false);
 
-- 
2.13.3

[PATCH RFC v3 6/9] KVM: rework kvm_vcpu_on_spin loop

[David Hildenbrand]

On 21.08.2017 22:35, Radim Kr?m?? wrote:
> The original code managed to obfuscate a straightforward idea:
> start iterating from the selected index and reset the index to 0 when
> reaching the end of online vcpus, then iterate until reaching the index
> that we started at.
> 
> The resulting code is a bit better, IMO.  (Still horrible, though.)

I think I prefer dropping this patch and maybe _after_ we have the list
implementation in place, simply start walking the list from
last_boosted_vcpu? (store a pointer instead of an index then, of course)

If I understand correctly, this would then be simply, one walk from
last_boosted_vcpu until we hit last_boosted_vcpu again.


> 
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> ---
>  include/linux/kvm_host.h | 13 +++++++++++++
>  virt/kvm/kvm_main.c      | 47 ++++++++++++++++++-----------------------------
>  2 files changed, 31 insertions(+), 29 deletions(-)
> 
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index abd5cb1feb9e..cfb3c0efdd51 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -498,6 +498,19 @@ static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
>  	     (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
>  	     idx++)
>  
> +#define kvm_for_each_vcpu_from(idx, vcpup, from, kvm) \
> +	for (idx = from, vcpup = kvm_get_vcpu(kvm, idx); \
> +	     vcpup; \
> +	     ({ \
> +		idx++; \
> +		if (idx >= atomic_read(&kvm->online_vcpus)) \
> +			idx = 0; \
> +		if (idx == from) \
> +			vcpup = NULL; \
> +		else \
> +			vcpup = kvm_get_vcpu(kvm, idx); \
> +	      }))
> +
>  static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
>  {
>  	struct kvm_vcpu *vcpu = NULL;
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index d89261d0d8c6..33a15e176927 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -2333,8 +2333,7 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
>  	struct kvm_vcpu *vcpu;
>  	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
>  	int yielded = 0;
> -	int try = 3;
> -	int pass;
> +	int try = 2;
>  	int i;
>  
>  	kvm_vcpu_set_in_spin_loop(me, true);
> @@ -2345,34 +2344,24 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
>  	 * VCPU is holding the lock that we need and will release it.
>  	 * We approximate round-robin by starting at the last boosted VCPU.
>  	 */
> -	for (pass = 0; pass < 2 && !yielded && try; pass++) {
> -		kvm_for_each_vcpu(i, vcpu, kvm) {
> -			if (!pass && i <= last_boosted_vcpu) {
> -				i = last_boosted_vcpu;
> -				continue;
> -			} else if (pass && i > last_boosted_vcpu)
> -				break;
> -			if (!ACCESS_ONCE(vcpu->preempted))
> -				continue;
> -			if (vcpu == me)
> -				continue;
> -			if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
> -				continue;
> -			if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
> -				continue;
> -			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
> -				continue;
> +	kvm_for_each_vcpu_from(i, vcpu, last_boosted_vcpu, kvm) {
> +		if (!ACCESS_ONCE(vcpu->preempted))
> +			continue;
> +		if (vcpu == me)
> +			continue;
> +		if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
> +			continue;
> +		if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
> +			continue;
> +		if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
> +			continue;
>  
> -			yielded = kvm_vcpu_yield_to(vcpu);
> -			if (yielded > 0) {
> -				kvm->last_boosted_vcpu = i;
> -				break;
> -			} else if (yielded < 0) {
> -				try--;
> -				if (!try)
> -					break;
> -			}
> -		}
> +		yielded = kvm_vcpu_yield_to(vcpu);
> +		if (yielded > 0) {
> +			kvm->last_boosted_vcpu = i;
> +			break;
> +		} else if (yielded < 0 && !try--)
> +			break;
>  	}
>  	kvm_vcpu_set_in_spin_loop(me, false);
>  
> 


-- 

Thanks,

David

[PATCH RFC v3 6/9] KVM: rework kvm_vcpu_on_spin loop

[Cornelia Huck]

On Tue, 22 Aug 2017 16:06:57 +0200
David Hildenbrand <david@redhat.com> wrote:

> On 21.08.2017 22:35, Radim Kr?m?? wrote:
> > The original code managed to obfuscate a straightforward idea:
> > start iterating from the selected index and reset the index to 0 when
> > reaching the end of online vcpus, then iterate until reaching the index
> > that we started at.
> > 
> > The resulting code is a bit better, IMO.  (Still horrible, though.)  
> 
> I think I prefer dropping this patch and maybe _after_ we have the list
> implementation in place, simply start walking the list from
> last_boosted_vcpu? (store a pointer instead of an index then, of course)
> 
> If I understand correctly, this would then be simply, one walk from
> last_boosted_vcpu until we hit last_boosted_vcpu again.

Yes, doing this change at this point in the series trades an ugly piece
of code for a slightly less ugly one.

> > 
> > Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> > ---
> >  include/linux/kvm_host.h | 13 +++++++++++++
> >  virt/kvm/kvm_main.c      | 47 ++++++++++++++++++-----------------------------
> >  2 files changed, 31 insertions(+), 29 deletions(-)
> > 
> > diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> > index abd5cb1feb9e..cfb3c0efdd51 100644
> > --- a/include/linux/kvm_host.h
> > +++ b/include/linux/kvm_host.h
> > @@ -498,6 +498,19 @@ static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
> >  	     (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
> >  	     idx++)
> >  
> > +#define kvm_for_each_vcpu_from(idx, vcpup, from, kvm) \
> > +	for (idx = from, vcpup = kvm_get_vcpu(kvm, idx); \
> > +	     vcpup; \
> > +	     ({ \
> > +		idx++; \
> > +		if (idx >= atomic_read(&kvm->online_vcpus)) \
> > +			idx = 0; \
> > +		if (idx == from) \
> > +			vcpup = NULL; \
> > +		else \
> > +			vcpup = kvm_get_vcpu(kvm, idx); \
> > +	      }))

The loop below is better after the change, but this macro... it gets at
least a bit better if you push this behind patch 8.

> > +
> >  static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
> >  {
> >  	struct kvm_vcpu *vcpu = NULL;
> > diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> > index d89261d0d8c6..33a15e176927 100644
> > --- a/virt/kvm/kvm_main.c
> > +++ b/virt/kvm/kvm_main.c
> > @@ -2333,8 +2333,7 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
> >  	struct kvm_vcpu *vcpu;
> >  	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
> >  	int yielded = 0;
> > -	int try = 3;
> > -	int pass;
> > +	int try = 2;
> >  	int i;
> >  
> >  	kvm_vcpu_set_in_spin_loop(me, true);
> > @@ -2345,34 +2344,24 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
> >  	 * VCPU is holding the lock that we need and will release it.
> >  	 * We approximate round-robin by starting at the last boosted VCPU.
> >  	 */
> > -	for (pass = 0; pass < 2 && !yielded && try; pass++) {
> > -		kvm_for_each_vcpu(i, vcpu, kvm) {
> > -			if (!pass && i <= last_boosted_vcpu) {
> > -				i = last_boosted_vcpu;
> > -				continue;
> > -			} else if (pass && i > last_boosted_vcpu)
> > -				break;
> > -			if (!ACCESS_ONCE(vcpu->preempted))
> > -				continue;
> > -			if (vcpu == me)
> > -				continue;
> > -			if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
> > -				continue;
> > -			if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
> > -				continue;
> > -			if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
> > -				continue;
> > +	kvm_for_each_vcpu_from(i, vcpu, last_boosted_vcpu, kvm) {
> > +		if (!ACCESS_ONCE(vcpu->preempted))
> > +			continue;
> > +		if (vcpu == me)
> > +			continue;
> > +		if (swait_active(&vcpu->wq) && !kvm_arch_vcpu_runnable(vcpu))
> > +			continue;
> > +		if (yield_to_kernel_mode && !kvm_arch_vcpu_in_kernel(vcpu))
> > +			continue;
> > +		if (!kvm_vcpu_eligible_for_directed_yield(vcpu))
> > +			continue;
> >  
> > -			yielded = kvm_vcpu_yield_to(vcpu);
> > -			if (yielded > 0) {
> > -				kvm->last_boosted_vcpu = i;
> > -				break;
> > -			} else if (yielded < 0) {
> > -				try--;
> > -				if (!try)
> > -					break;
> > -			}
> > -		}
> > +		yielded = kvm_vcpu_yield_to(vcpu);
> > +		if (yielded > 0) {
> > +			kvm->last_boosted_vcpu = i;
> > +			break;
> > +		} else if (yielded < 0 && !try--)
> > +			break;
> >  	}
> >  	kvm_vcpu_set_in_spin_loop(me, false);
> >  
> >   
> 
> 

[PATCH RFC v3 7/9] KVM: add kvm_free_vcpus and kvm_arch_free_vcpus

[Radim Krčmář]

Generalize clearing of kvm->vcpus.  This should not be needed at all as
all accesses to VCPUs in the destruction path are bugs, but maybe helps
to catch them.  The call path crosses arch/common code way too much, so
extra untangling patch is welcome.  Doing the clearing later seems be
ok.

I don't see a reason for the locking the mutex, so there definitely is
a room for improvements.

Suggested-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
---
 arch/mips/kvm/mips.c       | 13 ++-----------
 arch/powerpc/kvm/powerpc.c | 17 +++++++++--------
 arch/s390/kvm/kvm-s390.c   |  9 +--------
 arch/x86/kvm/x86.c         |  9 +--------
 include/linux/kvm_host.h   |  2 ++
 virt/kvm/arm/arm.c         | 17 +++++++++--------
 virt/kvm/kvm_main.c        | 17 +++++++++++++++++
 7 files changed, 41 insertions(+), 43 deletions(-)

diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index bce2a6431430..770c40b9df37 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -160,7 +160,7 @@ int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
 	return 0;
 }
 
-void kvm_mips_free_vcpus(struct kvm *kvm)
+void kvm_arch_free_vcpus(struct kvm *kvm)
 {
 	unsigned int i;
 	struct kvm_vcpu *vcpu;
@@ -168,15 +168,6 @@ void kvm_mips_free_vcpus(struct kvm *kvm)
 	kvm_for_each_vcpu(i, vcpu, kvm) {
 		kvm_arch_vcpu_free(vcpu);
 	}
-
-	mutex_lock(&kvm->lock);
-
-	for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
-		kvm->vcpus[i] = NULL;
-
-	atomic_set(&kvm->online_vcpus, 0);
-
-	mutex_unlock(&kvm->lock);
 }
 
 static void kvm_mips_free_gpa_pt(struct kvm *kvm)
@@ -188,7 +179,7 @@ static void kvm_mips_free_gpa_pt(struct kvm *kvm)
 
 void kvm_arch_destroy_vm(struct kvm *kvm)
 {
-	kvm_mips_free_vcpus(kvm);
+	kvm_free_vcpus(kvm);
 	kvm_mips_free_gpa_pt(kvm);
 }
 
diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
index 3480faaf1ef8..1c563545473c 100644
--- a/arch/powerpc/kvm/powerpc.c
+++ b/arch/powerpc/kvm/powerpc.c
@@ -456,11 +456,17 @@ int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
 	return 0;
 }
 
-void kvm_arch_destroy_vm(struct kvm *kvm)
+void kvm_arch_free_vcpus(struct kvm *kvm)
 {
-	unsigned int i;
+	int i;
 	struct kvm_vcpu *vcpu;
 
+	kvm_for_each_vcpu(i, vcpu, kvm)
+		kvm_arch_vcpu_free(vcpu);
+}
+
+void kvm_arch_destroy_vm(struct kvm *kvm)
+{
 #ifdef CONFIG_KVM_XICS
 	/*
 	 * We call kick_all_cpus_sync() to ensure that all
@@ -471,14 +477,9 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
 		kick_all_cpus_sync();
 #endif
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
-		kvm_arch_vcpu_free(vcpu);
+	kvm_free_vcpus(kvm);
 
 	mutex_lock(&kvm->lock);
-	for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
-		kvm->vcpus[i] = NULL;
-
-	atomic_set(&kvm->online_vcpus, 0);
 
 	kvmppc_core_destroy_vm(kvm);
 
diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
index 1534778a3c66..bb6278d45a25 100644
--- a/arch/s390/kvm/kvm-s390.c
+++ b/arch/s390/kvm/kvm-s390.c
@@ -1941,20 +1941,13 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
 	kmem_cache_free(kvm_vcpu_cache, vcpu);
 }
 
-static void kvm_free_vcpus(struct kvm *kvm)
+void kvm_arch_free_vcpus(struct kvm *kvm)
 {
 	unsigned int i;
 	struct kvm_vcpu *vcpu;
 
 	kvm_for_each_vcpu(i, vcpu, kvm)
 		kvm_arch_vcpu_destroy(vcpu);
-
-	mutex_lock(&kvm->lock);
-	for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
-		kvm->vcpus[i] = NULL;
-
-	atomic_set(&kvm->online_vcpus, 0);
-	mutex_unlock(&kvm->lock);
 }
 
 void kvm_arch_destroy_vm(struct kvm *kvm)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index e10eda86bc7b..d021746f1fdf 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -8077,7 +8077,7 @@ static void kvm_unload_vcpu_mmu(struct kvm_vcpu *vcpu)
 	vcpu_put(vcpu);
 }
 
-static void kvm_free_vcpus(struct kvm *kvm)
+void kvm_arch_free_vcpus(struct kvm *kvm)
 {
 	unsigned int i;
 	struct kvm_vcpu *vcpu;
@@ -8091,13 +8091,6 @@ static void kvm_free_vcpus(struct kvm *kvm)
 	}
 	kvm_for_each_vcpu(i, vcpu, kvm)
 		kvm_arch_vcpu_free(vcpu);
-
-	mutex_lock(&kvm->lock);
-	for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
-		kvm->vcpus[i] = NULL;
-
-	atomic_set(&kvm->online_vcpus, 0);
-	mutex_unlock(&kvm->lock);
 }
 
 void kvm_arch_sync_events(struct kvm *kvm)
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index cfb3c0efdd51..a8b9aa563834 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -874,6 +874,8 @@ static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 void kvm_arch_destroy_vm(struct kvm *kvm);
 void kvm_arch_sync_events(struct kvm *kvm);
+void kvm_free_vcpus(struct kvm *kvm);
+void kvm_arch_free_vcpus(struct kvm *kvm);
 
 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
diff --git a/virt/kvm/arm/arm.c b/virt/kvm/arm/arm.c
index b9f68e4add71..d63aa1107fdb 100644
--- a/virt/kvm/arm/arm.c
+++ b/virt/kvm/arm/arm.c
@@ -166,6 +166,14 @@ int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
 	return VM_FAULT_SIGBUS;
 }
 
+void kvm_arch_free_vcpus(struct kvm *kvm)
+{
+	int i;
+	struct kvm_vcpu *vcpu;
+
+	for_each_online_vcpu(i, vcpu, kvm)
+		kvm_arch_vcpu_free(vcpu);
+}
 
 /**
  * kvm_arch_destroy_vm - destroy the VM data structure
@@ -173,17 +181,10 @@ int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
  */
 void kvm_arch_destroy_vm(struct kvm *kvm)
 {
-	int i;
-
 	free_percpu(kvm->arch.last_vcpu_ran);
 	kvm->arch.last_vcpu_ran = NULL;
 
-	for (i = 0; i < KVM_MAX_VCPUS; ++i) {
-		if (kvm->vcpus[i]) {
-			kvm_arch_vcpu_free(kvm->vcpus[i]);
-			kvm->vcpus[i] = NULL;
-		}
-	}
+	kvm_free_vcpus(kvm);
 
 	kvm_vgic_destroy(kvm);
 }
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 33a15e176927..0d2d8b0c785c 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -750,6 +750,23 @@ static void kvm_destroy_devices(struct kvm *kvm)
 	}
 }
 
+void kvm_free_vcpus(struct kvm *kvm)
+{
+	int i;
+
+	kvm_arch_free_vcpus(kvm);
+
+	mutex_lock(&kvm->lock);
+
+	i = atomic_read(&kvm->online_vcpus);
+	atomic_set(&kvm->online_vcpus, 0);
+
+	while (i--)
+		kvm->vcpus[i] = NULL;
+
+	mutex_unlock(&kvm->lock);
+}
+
 static void kvm_destroy_vm(struct kvm *kvm)
 {
 	int i;
-- 
2.13.3

[PATCH RFC v3 7/9] KVM: add kvm_free_vcpus and kvm_arch_free_vcpus

[David Hildenbrand]

> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index 33a15e176927..0d2d8b0c785c 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -750,6 +750,23 @@ static void kvm_destroy_devices(struct kvm *kvm)
>  	}
>  }
>  
> +void kvm_free_vcpus(struct kvm *kvm)
> +{
> +	int i;
> +
> +	kvm_arch_free_vcpus(kvm);

I wonder if it would be possible to get rid of kvm_arch_free_vcpus(kvm)
completely and simply call

kvm_for_each_vcpu(i, vcpu, kvm)
	kvm_arch_vcpu_free(vcpu);

at that point.

Would certainly require some refactoring, and I am not sure if we could
modify the special mmu handling for x86 ("Unpin any mmu pages first.").
But if in doubt, that part could be moved to kvm_arch_destroy_vm(), just
before calling kvm_free_vcpus().

> +
> +	mutex_lock(&kvm->lock);
> +
> +	i = atomic_read(&kvm->online_vcpus);
> +	atomic_set(&kvm->online_vcpus, 0);

i = atomic_xchg(&kvm->online_vcpus, 0);

> +
> +	while (i--)
> +		kvm->vcpus[i] = NULL;
> +
> +	mutex_unlock(&kvm->lock);
> +}
> +
>  static void kvm_destroy_vm(struct kvm *kvm)
>  {
>  	int i;
> 


-- 

Thanks,

David

[PATCH RFC v3 7/9] KVM: add kvm_free_vcpus and kvm_arch_free_vcpus

[Cornelia Huck]

On Tue, 22 Aug 2017 16:18:59 +0200
David Hildenbrand <david@redhat.com> wrote:

> > diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> > index 33a15e176927..0d2d8b0c785c 100644
> > --- a/virt/kvm/kvm_main.c
> > +++ b/virt/kvm/kvm_main.c
> > @@ -750,6 +750,23 @@ static void kvm_destroy_devices(struct kvm *kvm)
> >  	}
> >  }
> >  
> > +void kvm_free_vcpus(struct kvm *kvm)
> > +{
> > +	int i;
> > +
> > +	kvm_arch_free_vcpus(kvm);  
> 
> I wonder if it would be possible to get rid of kvm_arch_free_vcpus(kvm)
> completely and simply call
> 
> kvm_for_each_vcpu(i, vcpu, kvm)
> 	kvm_arch_vcpu_free(vcpu);
> 
> at that point.
> 
> Would certainly require some refactoring, and I am not sure if we could
> modify the special mmu handling for x86 ("Unpin any mmu pages first.").
> But if in doubt, that part could be moved to kvm_arch_destroy_vm(), just
> before calling kvm_free_vcpus().

Nod, that crossed my mind as well. Would avoid adding lots of
kvm_arch_free_vcpus() that do the same thing for many archs.

[PATCH RFC v3 8/9] KVM: implement kvm_for_each_vcpu with a list

[Radim Krčmář]

Going through all VCPUs is more natural with a list and the RCU list can
work as lockless with our constraints.

This makes kvm->vcpus lose most users, so it will be easier to make
something out of it.

A nice side-effect is that the first argument to the macro is gone.
ARM code was changed a bit to cope with the loss when working with a
range 0-n and most other places switched to vcpu->vcpus_idx.

Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
---
 arch/mips/kvm/mips.c                |  4 +---
 arch/powerpc/kvm/book3s_32_mmu.c    |  3 +--
 arch/powerpc/kvm/book3s_64_mmu.c    |  3 +--
 arch/powerpc/kvm/book3s_hv.c        |  7 +++----
 arch/powerpc/kvm/book3s_pr.c        |  5 ++---
 arch/powerpc/kvm/book3s_xics.c      |  2 +-
 arch/powerpc/kvm/book3s_xics.h      |  3 +--
 arch/powerpc/kvm/book3s_xive.c      | 18 ++++++++----------
 arch/powerpc/kvm/book3s_xive.h      |  3 +--
 arch/powerpc/kvm/e500_emulate.c     |  3 +--
 arch/powerpc/kvm/powerpc.c          |  3 +--
 arch/s390/kvm/interrupt.c           |  3 +--
 arch/s390/kvm/kvm-s390.c            | 31 ++++++++++---------------------
 arch/s390/kvm/kvm-s390.h            |  6 ++----
 arch/s390/kvm/sigp.c                |  3 +--
 arch/x86/kvm/hyperv.c               |  3 +--
 arch/x86/kvm/i8254.c                |  3 +--
 arch/x86/kvm/i8259.c                |  7 +++----
 arch/x86/kvm/ioapic.c               |  3 +--
 arch/x86/kvm/irq_comm.c             | 10 +++++-----
 arch/x86/kvm/lapic.c                |  5 ++---
 arch/x86/kvm/svm.c                  |  3 +--
 arch/x86/kvm/vmx.c                  |  3 +--
 arch/x86/kvm/x86.c                  | 25 ++++++++++---------------
 include/linux/kvm_host.h            | 30 +++++++++++++-----------------
 virt/kvm/arm/arch_timer.c           | 10 ++++------
 virt/kvm/arm/arm.c                  | 12 ++++--------
 virt/kvm/arm/pmu.c                  |  3 +--
 virt/kvm/arm/psci.c                 |  7 +++----
 virt/kvm/arm/vgic/vgic-init.c       | 11 +++++------
 virt/kvm/arm/vgic/vgic-kvm-device.c | 28 ++++++++++++++++------------
 virt/kvm/arm/vgic/vgic-mmio-v2.c    |  5 ++---
 virt/kvm/arm/vgic/vgic-mmio-v3.c    | 19 +++++++++++--------
 virt/kvm/arm/vgic/vgic.c            |  3 +--
 virt/kvm/kvm_main.c                 | 22 +++++++++++-----------
 35 files changed, 131 insertions(+), 178 deletions(-)

diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index 770c40b9df37..c841cb434486 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -162,12 +162,10 @@ int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
 
 void kvm_arch_free_vcpus(struct kvm *kvm)
 {
-	unsigned int i;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_arch_vcpu_free(vcpu);
-	}
 }
 
 static void kvm_mips_free_gpa_pt(struct kvm *kvm)
diff --git a/arch/powerpc/kvm/book3s_32_mmu.c b/arch/powerpc/kvm/book3s_32_mmu.c
index 1992676c7a94..1ac7cace49fc 100644
--- a/arch/powerpc/kvm/book3s_32_mmu.c
+++ b/arch/powerpc/kvm/book3s_32_mmu.c
@@ -353,11 +353,10 @@ static void kvmppc_mmu_book3s_32_mtsrin(struct kvm_vcpu *vcpu, u32 srnum,
 
 static void kvmppc_mmu_book3s_32_tlbie(struct kvm_vcpu *vcpu, ulong ea, bool large)
 {
-	int i;
 	struct kvm_vcpu *v;
 
 	/* flush this VA on all cpus */
-	kvm_for_each_vcpu(i, v, vcpu->kvm)
+	kvm_for_each_vcpu(v, vcpu->kvm)
 		kvmppc_mmu_pte_flush(v, ea, 0x0FFFF000);
 }
 
diff --git a/arch/powerpc/kvm/book3s_64_mmu.c b/arch/powerpc/kvm/book3s_64_mmu.c
index 29ebe2fd5867..7b043fcc8c88 100644
--- a/arch/powerpc/kvm/book3s_64_mmu.c
+++ b/arch/powerpc/kvm/book3s_64_mmu.c
@@ -534,7 +534,6 @@ static void kvmppc_mmu_book3s_64_tlbie(struct kvm_vcpu *vcpu, ulong va,
 				       bool large)
 {
 	u64 mask = 0xFFFFFFFFFULL;
-	long i;
 	struct kvm_vcpu *v;
 
 	dprintk("KVM MMU: tlbie(0x%lx)\n", va);
@@ -559,7 +558,7 @@ static void kvmppc_mmu_book3s_64_tlbie(struct kvm_vcpu *vcpu, ulong va,
 			mask = 0xFFFFFF000ULL;
 	}
 	/* flush this VA on all vcpus */
-	kvm_for_each_vcpu(i, v, vcpu->kvm)
+	kvm_for_each_vcpu(v, vcpu->kvm)
 		kvmppc_mmu_pte_vflush(v, va >> 12, mask);
 }
 
diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c
index 359c79cdf0cc..4d40537fad20 100644
--- a/arch/powerpc/kvm/book3s_hv.c
+++ b/arch/powerpc/kvm/book3s_hv.c
@@ -1253,9 +1253,8 @@ static void kvmppc_set_lpcr(struct kvm_vcpu *vcpu, u64 new_lpcr,
 	 */
 	if ((new_lpcr & LPCR_ILE) != (vc->lpcr & LPCR_ILE)) {
 		struct kvm_vcpu *vcpu;
-		int i;
 
-		kvm_for_each_vcpu(i, vcpu, kvm) {
+		kvm_for_each_vcpu(vcpu, kvm) {
 			if (vcpu->arch.vcore != vc)
 				continue;
 			if (new_lpcr & LPCR_ILE)
@@ -3347,7 +3346,7 @@ static int kvm_vm_ioctl_get_dirty_log_hv(struct kvm *kvm,
 {
 	struct kvm_memslots *slots;
 	struct kvm_memory_slot *memslot;
-	int i, r;
+	int r;
 	unsigned long n;
 	unsigned long *buf;
 	struct kvm_vcpu *vcpu;
@@ -3381,7 +3380,7 @@ static int kvm_vm_ioctl_get_dirty_log_hv(struct kvm *kvm,
 
 	/* Harvest dirty bits from VPA and DTL updates */
 	/* Note: we never modify the SLB shadow buffer areas */
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		spin_lock(&vcpu->arch.vpa_update_lock);
 		kvmppc_harvest_vpa_dirty(&vcpu->arch.vpa, memslot, buf);
 		kvmppc_harvest_vpa_dirty(&vcpu->arch.dtl, memslot, buf);
diff --git a/arch/powerpc/kvm/book3s_pr.c b/arch/powerpc/kvm/book3s_pr.c
index 69a09444d46e..9ef1b9b7e48a 100644
--- a/arch/powerpc/kvm/book3s_pr.c
+++ b/arch/powerpc/kvm/book3s_pr.c
@@ -251,7 +251,6 @@ static int kvmppc_core_check_requests_pr(struct kvm_vcpu *vcpu)
 static void do_kvm_unmap_hva(struct kvm *kvm, unsigned long start,
 			     unsigned long end)
 {
-	long i;
 	struct kvm_vcpu *vcpu;
 	struct kvm_memslots *slots;
 	struct kvm_memory_slot *memslot;
@@ -272,7 +271,7 @@ static void do_kvm_unmap_hva(struct kvm *kvm, unsigned long start,
 		 */
 		gfn = hva_to_gfn_memslot(hva_start, memslot);
 		gfn_end = hva_to_gfn_memslot(hva_end + PAGE_SIZE - 1, memslot);
-		kvm_for_each_vcpu(i, vcpu, kvm)
+		kvm_for_each_vcpu(vcpu, kvm)
 			kvmppc_mmu_pte_pflush(vcpu, gfn << PAGE_SHIFT,
 					      gfn_end << PAGE_SHIFT);
 	}
@@ -1593,7 +1592,7 @@ static int kvm_vm_ioctl_get_dirty_log_pr(struct kvm *kvm,
 		ga = memslot->base_gfn << PAGE_SHIFT;
 		ga_end = ga + (memslot->npages << PAGE_SHIFT);
 
-		kvm_for_each_vcpu(n, vcpu, kvm)
+		kvm_for_each_vcpu(vcpu, kvm)
 			kvmppc_mmu_pte_pflush(vcpu, ga, ga_end);
 
 		n = kvm_dirty_bitmap_bytes(memslot);
diff --git a/arch/powerpc/kvm/book3s_xics.c b/arch/powerpc/kvm/book3s_xics.c
index d329b2add7e2..9871b76368bd 100644
--- a/arch/powerpc/kvm/book3s_xics.c
+++ b/arch/powerpc/kvm/book3s_xics.c
@@ -966,7 +966,7 @@ static int xics_debug_show(struct seq_file *m, void *private)
 
 	seq_printf(m, "=========\nICP state\n=========\n");
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		struct kvmppc_icp *icp = vcpu->arch.icp;
 		union kvmppc_icp_state state;
 
diff --git a/arch/powerpc/kvm/book3s_xics.h b/arch/powerpc/kvm/book3s_xics.h
index 453c9e518c19..71e33af1e119 100644
--- a/arch/powerpc/kvm/book3s_xics.h
+++ b/arch/powerpc/kvm/book3s_xics.h
@@ -119,9 +119,8 @@ static inline struct kvmppc_icp *kvmppc_xics_find_server(struct kvm *kvm,
 							 u32 nr)
 {
 	struct kvm_vcpu *vcpu = NULL;
-	int i;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		if (vcpu->arch.icp && nr == vcpu->arch.icp->server_num)
 			return vcpu->arch.icp;
 	}
diff --git a/arch/powerpc/kvm/book3s_xive.c b/arch/powerpc/kvm/book3s_xive.c
index 08b200a0bbce..22222a540439 100644
--- a/arch/powerpc/kvm/book3s_xive.c
+++ b/arch/powerpc/kvm/book3s_xive.c
@@ -182,7 +182,7 @@ static int xive_check_provisioning(struct kvm *kvm, u8 prio)
 {
 	struct kvmppc_xive *xive = kvm->arch.xive;
 	struct kvm_vcpu *vcpu;
-	int i, rc;
+	int rc;
 
 	lockdep_assert_held(&kvm->lock);
 
@@ -193,7 +193,7 @@ static int xive_check_provisioning(struct kvm *kvm, u8 prio)
 	pr_devel("Provisioning prio... %d\n", prio);
 
 	/* Provision each VCPU and enable escalations */
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		if (!vcpu->arch.xive_vcpu)
 			continue;
 		rc = xive_provision_queue(vcpu, prio);
@@ -252,7 +252,7 @@ static int xive_try_pick_queue(struct kvm_vcpu *vcpu, u8 prio)
 static int xive_select_target(struct kvm *kvm, u32 *server, u8 prio)
 {
 	struct kvm_vcpu *vcpu;
-	int i, rc;
+	int rc;
 
 	/* Locate target server */
 	vcpu = kvmppc_xive_find_server(kvm, *server);
@@ -271,7 +271,7 @@ static int xive_select_target(struct kvm *kvm, u32 *server, u8 prio)
 	pr_devel(" .. failed, looking up candidate...\n");
 
 	/* Failed, pick another VCPU */
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		if (!vcpu->arch.xive_vcpu)
 			continue;
 		rc = xive_try_pick_queue(vcpu, prio);
@@ -1237,7 +1237,7 @@ static void xive_pre_save_queue(struct kvmppc_xive *xive, struct xive_q *q)
 static void xive_pre_save_scan(struct kvmppc_xive *xive)
 {
 	struct kvm_vcpu *vcpu = NULL;
-	int i, j;
+	int j;
 
 	/*
 	 * See comment in xive_get_source() about how this
@@ -1252,7 +1252,7 @@ static void xive_pre_save_scan(struct kvmppc_xive *xive)
 	}
 
 	/* Then scan the queues and update the "in_queue" flag */
-	kvm_for_each_vcpu(i, vcpu, xive->kvm) {
+	kvm_for_each_vcpu(vcpu, xive->kvm) {
 		struct kvmppc_xive_vcpu *xc = vcpu->arch.xive_vcpu;
 		if (!xc)
 			continue;
@@ -1418,9 +1418,8 @@ static bool xive_check_delayed_irq(struct kvmppc_xive *xive, u32 irq)
 {
 	struct kvm *kvm = xive->kvm;
 	struct kvm_vcpu *vcpu = NULL;
-	int i;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		struct kvmppc_xive_vcpu *xc = vcpu->arch.xive_vcpu;
 
 		if (!xc)
@@ -1787,14 +1786,13 @@ static int xive_debug_show(struct seq_file *m, void *private)
 	u64 t_vm_h_cppr = 0;
 	u64 t_vm_h_eoi = 0;
 	u64 t_vm_h_ipi = 0;
-	unsigned int i;
 
 	if (!kvm)
 		return 0;
 
 	seq_printf(m, "=========\nVCPU state\n=========\n");
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		struct kvmppc_xive_vcpu *xc = vcpu->arch.xive_vcpu;
 
 		if (!xc)
diff --git a/arch/powerpc/kvm/book3s_xive.h b/arch/powerpc/kvm/book3s_xive.h
index 5938f7644dc1..5bc33db0924c 100644
--- a/arch/powerpc/kvm/book3s_xive.h
+++ b/arch/powerpc/kvm/book3s_xive.h
@@ -175,9 +175,8 @@ struct kvmppc_xive_vcpu {
 static inline struct kvm_vcpu *kvmppc_xive_find_server(struct kvm *kvm, u32 nr)
 {
 	struct kvm_vcpu *vcpu = NULL;
-	int i;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		if (vcpu->arch.xive_vcpu && nr == vcpu->arch.xive_vcpu->server_num)
 			return vcpu;
 	}
diff --git a/arch/powerpc/kvm/e500_emulate.c b/arch/powerpc/kvm/e500_emulate.c
index 990db69a1d0b..56151681afcc 100644
--- a/arch/powerpc/kvm/e500_emulate.c
+++ b/arch/powerpc/kvm/e500_emulate.c
@@ -68,13 +68,12 @@ static int kvmppc_e500_emul_msgsnd(struct kvm_vcpu *vcpu, int rb)
 	ulong param = vcpu->arch.gpr[rb];
 	int prio = dbell2prio(rb);
 	int pir = param & PPC_DBELL_PIR_MASK;
-	int i;
 	struct kvm_vcpu *cvcpu;
 
 	if (prio < 0)
 		return EMULATE_FAIL;
 
-	kvm_for_each_vcpu(i, cvcpu, vcpu->kvm) {
+	kvm_for_each_vcpu(cvcpu, vcpu->kvm) {
 		int cpir = cvcpu->arch.shared->pir;
 		if ((param & PPC_DBELL_MSG_BRDCAST) || (cpir == pir)) {
 			set_bit(prio, &cvcpu->arch.pending_exceptions);
diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
index 1c563545473c..633d3bb501c1 100644
--- a/arch/powerpc/kvm/powerpc.c
+++ b/arch/powerpc/kvm/powerpc.c
@@ -458,10 +458,9 @@ int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
 
 void kvm_arch_free_vcpus(struct kvm *kvm)
 {
-	int i;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_arch_vcpu_free(vcpu);
 }
 
diff --git a/arch/s390/kvm/interrupt.c b/arch/s390/kvm/interrupt.c
index a619ddae610d..d503b25638af 100644
--- a/arch/s390/kvm/interrupt.c
+++ b/arch/s390/kvm/interrupt.c
@@ -2287,7 +2287,6 @@ static int flic_ais_mode_set_all(struct kvm *kvm, struct kvm_device_attr *attr)
 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
 {
 	int r = 0;
-	unsigned int i;
 	struct kvm_vcpu *vcpu;
 
 	switch (attr->group) {
@@ -2308,7 +2307,7 @@ static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
 		 * about late coming workers.
 		 */
 		synchronize_srcu(&dev->kvm->srcu);
-		kvm_for_each_vcpu(i, vcpu, dev->kvm)
+		kvm_for_each_vcpu(vcpu, dev->kvm)
 			kvm_clear_async_pf_completion_queue(vcpu);
 		break;
 	case KVM_DEV_FLIC_ADAPTER_REGISTER:
diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
index bb6278d45a25..3e64e3eb2a63 100644
--- a/arch/s390/kvm/kvm-s390.c
+++ b/arch/s390/kvm/kvm-s390.c
@@ -174,12 +174,11 @@ static int kvm_clock_sync(struct notifier_block *notifier, unsigned long val,
 {
 	struct kvm *kvm;
 	struct kvm_vcpu *vcpu;
-	int i;
 	unsigned long long *delta = v;
 
 	list_for_each_entry(kvm, &vm_list, vm_list) {
 		kvm->arch.epoch -= *delta;
-		kvm_for_each_vcpu(i, vcpu, kvm) {
+		kvm_for_each_vcpu(vcpu, kvm) {
 			vcpu->arch.sie_block->epoch -= *delta;
 			if (vcpu->arch.cputm_enabled)
 				vcpu->arch.cputm_start += *delta;
@@ -491,12 +490,10 @@ int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 
 static void icpt_operexc_on_all_vcpus(struct kvm *kvm)
 {
-	unsigned int i;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_s390_sync_request(KVM_REQ_ICPT_OPEREXC, vcpu);
-	}
 }
 
 static int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
@@ -705,7 +702,6 @@ static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu);
 static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
 {
 	struct kvm_vcpu *vcpu;
-	int i;
 
 	if (!test_kvm_facility(kvm, 76))
 		return -EINVAL;
@@ -743,7 +739,7 @@ static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
 		return -ENXIO;
 	}
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		kvm_s390_vcpu_crypto_setup(vcpu);
 		exit_sie(vcpu);
 	}
@@ -753,10 +749,9 @@ static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
 
 static void kvm_s390_sync_request_broadcast(struct kvm *kvm, int req)
 {
-	int cx;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(cx, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_s390_sync_request(req, vcpu);
 }
 
@@ -1943,10 +1938,9 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
 
 void kvm_arch_free_vcpus(struct kvm *kvm)
 {
-	unsigned int i;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_arch_vcpu_destroy(vcpu);
 }
 
@@ -2050,7 +2044,6 @@ static int sca_switch_to_extended(struct kvm *kvm)
 	struct bsca_block *old_sca = kvm->arch.sca;
 	struct esca_block *new_sca;
 	struct kvm_vcpu *vcpu;
-	unsigned int vcpu_idx;
 	u32 scaol, scaoh;
 
 	new_sca = alloc_pages_exact(sizeof(*new_sca), GFP_KERNEL|__GFP_ZERO);
@@ -2065,7 +2058,7 @@ static int sca_switch_to_extended(struct kvm *kvm)
 
 	sca_copy_b_to_e(new_sca, old_sca);
 
-	kvm_for_each_vcpu(vcpu_idx, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		vcpu->arch.sie_block->scaoh = scaoh;
 		vcpu->arch.sie_block->scaol = scaol;
 		vcpu->arch.sie_block->ecb2 |= ECB2_ESCA;
@@ -2491,14 +2484,13 @@ static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
 	struct kvm *kvm = gmap->private;
 	struct kvm_vcpu *vcpu;
 	unsigned long prefix;
-	int i;
 
 	if (gmap_is_shadow(gmap))
 		return;
 	if (start >= 1UL << 31)
 		/* We are only interested in prefix pages */
 		return;
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		/* match against both prefix pages */
 		prefix = kvm_s390_get_prefix(vcpu);
 		if (prefix <= end && start <= prefix + 2*PAGE_SIZE - 1) {
@@ -2856,13 +2848,12 @@ static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu)
 void kvm_s390_set_tod_clock(struct kvm *kvm, u64 tod)
 {
 	struct kvm_vcpu *vcpu;
-	int i;
 
 	mutex_lock(&kvm->lock);
 	preempt_disable();
 	kvm->arch.epoch = tod - get_tod_clock();
 	kvm_s390_vcpu_block_all(kvm);
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		vcpu->arch.sie_block->epoch = kvm->arch.epoch;
 	kvm_s390_vcpu_unblock_all(kvm);
 	preempt_enable();
@@ -3389,12 +3380,10 @@ static void __disable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
 
 static void __disable_ibs_on_all_vcpus(struct kvm *kvm)
 {
-	unsigned int i;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm)
 		__disable_ibs_on_vcpu(vcpu);
-	}
 }
 
 static void __enable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
@@ -3462,7 +3451,7 @@ void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
 		 * As we only have one VCPU left, we want to enable the IBS
 		 * facility for that VCPU to speed it up.
 		 */
-		kvm_for_each_vcpu(i, started_vcpu, vcpu->kvm)
+		kvm_for_each_vcpu(started_vcpu, vcpu->kvm)
 			if (!is_vcpu_stopped(started_vcpu)) {
 				__enable_ibs_on_vcpu(started_vcpu);
 				break;
diff --git a/arch/s390/kvm/kvm-s390.h b/arch/s390/kvm/kvm-s390.h
index 6fedc8bc7a37..6077a724630c 100644
--- a/arch/s390/kvm/kvm-s390.h
+++ b/arch/s390/kvm/kvm-s390.h
@@ -294,20 +294,18 @@ int kvm_s390_handle_diag(struct kvm_vcpu *vcpu);
 
 static inline void kvm_s390_vcpu_block_all(struct kvm *kvm)
 {
-	int i;
 	struct kvm_vcpu *vcpu;
 
 	WARN_ON(!mutex_is_locked(&kvm->lock));
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_s390_vcpu_block(vcpu);
 }
 
 static inline void kvm_s390_vcpu_unblock_all(struct kvm *kvm)
 {
-	int i;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_s390_vcpu_unblock(vcpu);
 }
 
diff --git a/arch/s390/kvm/sigp.c b/arch/s390/kvm/sigp.c
index 1a252f537081..a31e1c2a994d 100644
--- a/arch/s390/kvm/sigp.c
+++ b/arch/s390/kvm/sigp.c
@@ -158,7 +158,6 @@ static int __sigp_stop_and_store_status(struct kvm_vcpu *vcpu,
 static int __sigp_set_arch(struct kvm_vcpu *vcpu, u32 parameter)
 {
 	int rc;
-	unsigned int i;
 	struct kvm_vcpu *v;
 
 	switch (parameter & 0xff) {
@@ -167,7 +166,7 @@ static int __sigp_set_arch(struct kvm_vcpu *vcpu, u32 parameter)
 		break;
 	case 1:
 	case 2:
-		kvm_for_each_vcpu(i, v, vcpu->kvm) {
+		kvm_for_each_vcpu(v, vcpu->kvm) {
 			v->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
 			kvm_clear_async_pf_completion_queue(v);
 		}
diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c
index dc97f2544b6f..b43cb27bf783 100644
--- a/arch/x86/kvm/hyperv.c
+++ b/arch/x86/kvm/hyperv.c
@@ -109,13 +109,12 @@ static int synic_set_sint(struct kvm_vcpu_hv_synic *synic, int sint,
 static struct kvm_vcpu *get_vcpu_by_vpidx(struct kvm *kvm, u32 vpidx)
 {
 	struct kvm_vcpu *vcpu = NULL;
-	int i;
 
 	if (vpidx < KVM_MAX_VCPUS)
 		vcpu = kvm_get_vcpu(kvm, vpidx);
 	if (vcpu && vcpu_to_hv_vcpu(vcpu)->vp_index == vpidx)
 		return vcpu;
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		if (vcpu_to_hv_vcpu(vcpu)->vp_index == vpidx)
 			return vcpu;
 	return NULL;
diff --git a/arch/x86/kvm/i8254.c b/arch/x86/kvm/i8254.c
index af192895b1fc..fdd3bff598f3 100644
--- a/arch/x86/kvm/i8254.c
+++ b/arch/x86/kvm/i8254.c
@@ -241,7 +241,6 @@ static void pit_do_work(struct kthread_work *work)
 	struct kvm_pit *pit = container_of(work, struct kvm_pit, expired);
 	struct kvm *kvm = pit->kvm;
 	struct kvm_vcpu *vcpu;
-	int i;
 	struct kvm_kpit_state *ps = &pit->pit_state;
 
 	if (atomic_read(&ps->reinject) && !atomic_xchg(&ps->irq_ack, 0))
@@ -260,7 +259,7 @@ static void pit_do_work(struct kthread_work *work)
 	 * also be simultaneously delivered through PIC and IOAPIC.
 	 */
 	if (atomic_read(&kvm->arch.vapics_in_nmi_mode) > 0)
-		kvm_for_each_vcpu(i, vcpu, kvm)
+		kvm_for_each_vcpu(vcpu, kvm)
 			kvm_apic_nmi_wd_deliver(vcpu);
 }
 
diff --git a/arch/x86/kvm/i8259.c b/arch/x86/kvm/i8259.c
index bdcd4139eca9..c4f57fbd7316 100644
--- a/arch/x86/kvm/i8259.c
+++ b/arch/x86/kvm/i8259.c
@@ -50,14 +50,13 @@ static void pic_unlock(struct kvm_pic *s)
 {
 	bool wakeup = s->wakeup_needed;
 	struct kvm_vcpu *vcpu;
-	int i;
 
 	s->wakeup_needed = false;
 
 	spin_unlock(&s->lock);
 
 	if (wakeup) {
-		kvm_for_each_vcpu(i, vcpu, s->kvm) {
+		kvm_for_each_vcpu(vcpu, s->kvm) {
 			if (kvm_apic_accept_pic_intr(vcpu)) {
 				kvm_make_request(KVM_REQ_EVENT, vcpu);
 				kvm_vcpu_kick(vcpu);
@@ -270,7 +269,7 @@ int kvm_pic_read_irq(struct kvm *kvm)
 
 static void kvm_pic_reset(struct kvm_kpic_state *s)
 {
-	int irq, i;
+	int irq;
 	struct kvm_vcpu *vcpu;
 	u8 edge_irr = s->irr & ~s->elcr;
 	bool found = false;
@@ -287,7 +286,7 @@ static void kvm_pic_reset(struct kvm_kpic_state *s)
 	}
 	s->init_state = 1;
 
-	kvm_for_each_vcpu(i, vcpu, s->pics_state->kvm)
+	kvm_for_each_vcpu(vcpu, s->pics_state->kvm)
 		if (kvm_apic_accept_pic_intr(vcpu)) {
 			found = true;
 			break;
diff --git a/arch/x86/kvm/ioapic.c b/arch/x86/kvm/ioapic.c
index 1a29da6c0558..2216e6d34fb7 100644
--- a/arch/x86/kvm/ioapic.c
+++ b/arch/x86/kvm/ioapic.c
@@ -146,13 +146,12 @@ void kvm_rtc_eoi_tracking_restore_one(struct kvm_vcpu *vcpu)
 static void kvm_rtc_eoi_tracking_restore_all(struct kvm_ioapic *ioapic)
 {
 	struct kvm_vcpu *vcpu;
-	int i;
 
 	if (RTC_GSI >= IOAPIC_NUM_PINS)
 		return;
 
 	rtc_irq_eoi_tracking_reset(ioapic);
-	kvm_for_each_vcpu(i, vcpu, ioapic->kvm)
+	kvm_for_each_vcpu(vcpu, ioapic->kvm)
 	    __rtc_irq_eoi_tracking_restore_one(vcpu);
 }
 
diff --git a/arch/x86/kvm/irq_comm.c b/arch/x86/kvm/irq_comm.c
index 3cc3b2d130a0..a8f777f2b0e6 100644
--- a/arch/x86/kvm/irq_comm.c
+++ b/arch/x86/kvm/irq_comm.c
@@ -58,7 +58,7 @@ static int kvm_set_ioapic_irq(struct kvm_kernel_irq_routing_entry *e,
 int kvm_irq_delivery_to_apic(struct kvm *kvm, struct kvm_lapic *src,
 		struct kvm_lapic_irq *irq, struct dest_map *dest_map)
 {
-	int i, r = -1;
+	int r = -1;
 	struct kvm_vcpu *vcpu, *lowest = NULL;
 	unsigned long dest_vcpu_bitmap[BITS_TO_LONGS(KVM_MAX_VCPUS)];
 	unsigned int dest_vcpus = 0;
@@ -74,7 +74,7 @@ int kvm_irq_delivery_to_apic(struct kvm *kvm, struct kvm_lapic *src,
 
 	memset(dest_vcpu_bitmap, 0, sizeof(dest_vcpu_bitmap));
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		if (!kvm_apic_present(vcpu))
 			continue;
 
@@ -93,7 +93,7 @@ int kvm_irq_delivery_to_apic(struct kvm *kvm, struct kvm_lapic *src,
 				else if (kvm_apic_compare_prio(vcpu, lowest) < 0)
 					lowest = vcpu;
 			} else {
-				__set_bit(i, dest_vcpu_bitmap);
+				__set_bit(vcpu->vcpus_idx, dest_vcpu_bitmap);
 				dest_vcpus++;
 			}
 		}
@@ -335,13 +335,13 @@ int kvm_set_routing_entry(struct kvm *kvm,
 bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
 			     struct kvm_vcpu **dest_vcpu)
 {
-	int i, r = 0;
+	int r = 0;
 	struct kvm_vcpu *vcpu;
 
 	if (kvm_intr_is_single_vcpu_fast(kvm, irq, dest_vcpu))
 		return true;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		if (!kvm_apic_present(vcpu))
 			continue;
 
diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index 4f38818db929..a7cdd6baa38b 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -166,12 +166,11 @@ static void recalculate_apic_map(struct kvm *kvm)
 {
 	struct kvm_apic_map *new, *old = NULL;
 	struct kvm_vcpu *vcpu;
-	int i;
 	u32 max_id = 255; /* enough space for any xAPIC ID */
 
 	mutex_lock(&kvm->arch.apic_map_lock);
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		if (kvm_apic_present(vcpu))
 			max_id = max(max_id, kvm_x2apic_id(vcpu->arch.apic));
 
@@ -183,7 +182,7 @@ static void recalculate_apic_map(struct kvm *kvm)
 
 	new->max_apic_id = max_id;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		struct kvm_lapic *apic = vcpu->arch.apic;
 		struct kvm_lapic **cluster;
 		u16 mask;
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 1fa9ee5660f4..605c18003f55 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -3807,7 +3807,6 @@ static int avic_incomplete_ipi_interception(struct vcpu_svm *svm)
 		kvm_lapic_reg_write(apic, APIC_ICR, icrl);
 		break;
 	case AVIC_IPI_FAILURE_TARGET_NOT_RUNNING: {
-		int i;
 		struct kvm_vcpu *vcpu;
 		struct kvm *kvm = svm->vcpu.kvm;
 		struct kvm_lapic *apic = svm->vcpu.arch.apic;
@@ -3817,7 +3816,7 @@ static int avic_incomplete_ipi_interception(struct vcpu_svm *svm)
 		 * set the appropriate IRR bits on the valid target
 		 * vcpus. So, we just need to kick the appropriate vcpu.
 		 */
-		kvm_for_each_vcpu(i, vcpu, kvm) {
+		kvm_for_each_vcpu(vcpu, kvm) {
 			bool m = kvm_apic_match_dest(vcpu, apic,
 						     icrl & KVM_APIC_SHORT_MASK,
 						     GET_APIC_DEST_FIELD(icrh),
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index df8d2f127508..ae0f04e26fec 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -8472,7 +8472,6 @@ static void vmx_flush_pml_buffer(struct kvm_vcpu *vcpu)
  */
 static void kvm_flush_pml_buffers(struct kvm *kvm)
 {
-	int i;
 	struct kvm_vcpu *vcpu;
 	/*
 	 * We only need to kick vcpu out of guest mode here, as PML buffer
@@ -8480,7 +8479,7 @@ static void kvm_flush_pml_buffers(struct kvm *kvm)
 	 * vcpus running in guest are possible to have unflushed GPAs in PML
 	 * buffer.
 	 */
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_vcpu_kick(vcpu);
 }
 
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index d021746f1fdf..caea24d3ddb0 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -1734,7 +1734,6 @@ void kvm_make_mclock_inprogress_request(struct kvm *kvm)
 static void kvm_gen_update_masterclock(struct kvm *kvm)
 {
 #ifdef CONFIG_X86_64
-	int i;
 	struct kvm_vcpu *vcpu;
 	struct kvm_arch *ka = &kvm->arch;
 
@@ -1743,11 +1742,11 @@ static void kvm_gen_update_masterclock(struct kvm *kvm)
 	/* no guest entries from this point */
 	pvclock_update_vm_gtod_copy(kvm);
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
 
 	/* guest entries allowed */
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_clear_request(KVM_REQ_MCLOCK_INPROGRESS, vcpu);
 
 	spin_unlock(&ka->pvclock_gtod_sync_lock);
@@ -1945,14 +1944,13 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)
 
 static void kvmclock_update_fn(struct work_struct *work)
 {
-	int i;
 	struct delayed_work *dwork = to_delayed_work(work);
 	struct kvm_arch *ka = container_of(dwork, struct kvm_arch,
 					   kvmclock_update_work);
 	struct kvm *kvm = container_of(ka, struct kvm, arch);
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
 		kvm_vcpu_kick(vcpu);
 	}
@@ -5844,7 +5842,7 @@ static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long va
 	struct cpufreq_freqs *freq = data;
 	struct kvm *kvm;
 	struct kvm_vcpu *vcpu;
-	int i, send_ipi = 0;
+	int send_ipi = 0;
 
 	/*
 	 * We allow guests to temporarily run on slowing clocks,
@@ -5894,7 +5892,7 @@ static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long va
 
 	spin_lock(&kvm_lock);
 	list_for_each_entry(kvm, &vm_list, vm_list) {
-		kvm_for_each_vcpu(i, vcpu, kvm) {
+		kvm_for_each_vcpu(vcpu, kvm) {
 			if (vcpu->cpu != freq->cpu)
 				continue;
 			kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
@@ -6035,11 +6033,10 @@ static void pvclock_gtod_update_fn(struct work_struct *work)
 	struct kvm *kvm;
 
 	struct kvm_vcpu *vcpu;
-	int i;
 
 	spin_lock(&kvm_lock);
 	list_for_each_entry(kvm, &vm_list, vm_list)
-		kvm_for_each_vcpu(i, vcpu, kvm)
+		kvm_for_each_vcpu(vcpu, kvm)
 			kvm_make_request(KVM_REQ_MASTERCLOCK_UPDATE, vcpu);
 	atomic_set(&kvm_guest_has_master_clock, 0);
 	spin_unlock(&kvm_lock);
@@ -7783,7 +7780,6 @@ int kvm_arch_hardware_enable(void)
 {
 	struct kvm *kvm;
 	struct kvm_vcpu *vcpu;
-	int i;
 	int ret;
 	u64 local_tsc;
 	u64 max_tsc = 0;
@@ -7797,7 +7793,7 @@ int kvm_arch_hardware_enable(void)
 	local_tsc = rdtsc();
 	stable = !check_tsc_unstable();
 	list_for_each_entry(kvm, &vm_list, vm_list) {
-		kvm_for_each_vcpu(i, vcpu, kvm) {
+		kvm_for_each_vcpu(vcpu, kvm) {
 			if (!stable && vcpu->cpu == smp_processor_id())
 				kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
 			if (stable && vcpu->arch.last_host_tsc > local_tsc) {
@@ -7850,7 +7846,7 @@ int kvm_arch_hardware_enable(void)
 		u64 delta_cyc = max_tsc - local_tsc;
 		list_for_each_entry(kvm, &vm_list, vm_list) {
 			kvm->arch.backwards_tsc_observed = true;
-			kvm_for_each_vcpu(i, vcpu, kvm) {
+			kvm_for_each_vcpu(vcpu, kvm) {
 				vcpu->arch.tsc_offset_adjustment += delta_cyc;
 				vcpu->arch.last_host_tsc = local_tsc;
 				kvm_make_request(KVM_REQ_MASTERCLOCK_UPDATE, vcpu);
@@ -8079,17 +8075,16 @@ static void kvm_unload_vcpu_mmu(struct kvm_vcpu *vcpu)
 
 void kvm_arch_free_vcpus(struct kvm *kvm)
 {
-	unsigned int i;
 	struct kvm_vcpu *vcpu;
 
 	/*
 	 * Unpin any mmu pages first.
 	 */
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		kvm_clear_async_pf_completion_queue(vcpu);
 		kvm_unload_vcpu_mmu(vcpu);
 	}
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_arch_vcpu_free(vcpu);
 }
 
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index a8b9aa563834..5417dac55272 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -216,6 +216,7 @@ struct kvm_mmio_fragment {
 
 struct kvm_vcpu {
 	struct kvm *kvm;
+	struct list_head vcpu_list;
 #ifdef CONFIG_PREEMPT_NOTIFIERS
 	struct preempt_notifier preempt_notifier;
 #endif
@@ -393,6 +394,7 @@ struct kvm {
 	struct mm_struct *mm; /* userspace tied to this vm */
 	struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
 	struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
+	struct list_head vcpu_list;
 
 	/*
 	 * created_vcpus is protected by kvm->lock, and is incremented
@@ -402,7 +404,7 @@ struct kvm {
 	 */
 	atomic_t online_vcpus;
 	int created_vcpus;
-	int last_boosted_vcpu;
+	struct kvm_vcpu *last_boosted_vcpu;
 	struct list_head vm_list;
 	struct mutex lock;
 	struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
@@ -492,29 +494,23 @@ static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
 	return kvm->vcpus[i];
 }
 
-#define kvm_for_each_vcpu(idx, vcpup, kvm) \
-	for (idx = 0; \
-	     idx < atomic_read(&kvm->online_vcpus) && \
-	     (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
-	     idx++)
+#define kvm_for_each_vcpu(vcpup, kvm) \
+	list_for_each_entry_rcu(vcpup, &kvm->vcpu_list, vcpu_list)
 
-#define kvm_for_each_vcpu_from(idx, vcpup, from, kvm) \
-	for (idx = from, vcpup = kvm_get_vcpu(kvm, idx); \
+#define kvm_for_each_vcpu_from(vcpup, from, kvm) \
+	for (vcpup = from; \
 	     vcpup; \
 	     ({ \
-		idx++; \
-		if (idx >= atomic_read(&kvm->online_vcpus)) \
-			idx = 0; \
-		if (idx == from) \
+		vcpup = list_entry_rcu(vcpup->vcpu_list.next, typeof(*vcpup), vcpu_list); \
+		if (&vcpup->vcpu_list == &kvm->vcpu_list) \
+			vcpup = list_entry_rcu(kvm->vcpu_list.next, typeof(*vcpup), vcpu_list); \
+		if (vcpup == from) \
 			vcpup = NULL; \
-		else \
-			vcpup = kvm_get_vcpu(kvm, idx); \
-	      }))
+	     }))
 
 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
 {
 	struct kvm_vcpu *vcpu = NULL;
-	int i;
 
 	if (id < 0)
 		return NULL;
@@ -522,7 +518,7 @@ static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
 		vcpu = kvm_get_vcpu(kvm, id);
 	if (vcpu && vcpu->vcpu_id == id)
 		return vcpu;
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		if (vcpu->vcpu_id == id)
 			return vcpu;
 	return NULL;
diff --git a/virt/kvm/arm/arch_timer.c b/virt/kvm/arm/arch_timer.c
index 8e89d63005c7..3e311fa5c7a1 100644
--- a/virt/kvm/arm/arch_timer.c
+++ b/virt/kvm/arm/arch_timer.c
@@ -478,12 +478,11 @@ int kvm_timer_vcpu_reset(struct kvm_vcpu *vcpu)
 /* Make the updates of cntvoff for all vtimer contexts atomic */
 static void update_vtimer_cntvoff(struct kvm_vcpu *vcpu, u64 cntvoff)
 {
-	int i;
 	struct kvm *kvm = vcpu->kvm;
 	struct kvm_vcpu *tmp;
 
 	mutex_lock(&kvm->lock);
-	kvm_for_each_vcpu(i, tmp, kvm)
+	kvm_for_each_vcpu(tmp, kvm)
 		vcpu_vtimer(tmp)->cntvoff = cntvoff;
 
 	/*
@@ -622,7 +621,7 @@ void kvm_timer_vcpu_terminate(struct kvm_vcpu *vcpu)
 static bool timer_irqs_are_valid(struct kvm_vcpu *vcpu)
 {
 	int vtimer_irq, ptimer_irq;
-	int i, ret;
+	int ret;
 
 	vtimer_irq = vcpu_vtimer(vcpu)->irq.irq;
 	ret = kvm_vgic_set_owner(vcpu, vtimer_irq, vcpu_vtimer(vcpu));
@@ -634,7 +633,7 @@ static bool timer_irqs_are_valid(struct kvm_vcpu *vcpu)
 	if (ret)
 		return false;
 
-	kvm_for_each_vcpu(i, vcpu, vcpu->kvm) {
+	kvm_for_each_vcpu(vcpu, vcpu->kvm) {
 		if (vcpu_vtimer(vcpu)->irq.irq != vtimer_irq ||
 		    vcpu_ptimer(vcpu)->irq.irq != ptimer_irq)
 			return false;
@@ -720,9 +719,8 @@ void kvm_timer_init_vhe(void)
 static void set_timer_irqs(struct kvm *kvm, int vtimer_irq, int ptimer_irq)
 {
 	struct kvm_vcpu *vcpu;
-	int i;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		vcpu_vtimer(vcpu)->irq.irq = vtimer_irq;
 		vcpu_ptimer(vcpu)->irq.irq = ptimer_irq;
 	}
diff --git a/virt/kvm/arm/arm.c b/virt/kvm/arm/arm.c
index d63aa1107fdb..e105f6b307cc 100644
--- a/virt/kvm/arm/arm.c
+++ b/virt/kvm/arm/arm.c
@@ -168,10 +168,9 @@ int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
 
 void kvm_arch_free_vcpus(struct kvm *kvm)
 {
-	int i;
 	struct kvm_vcpu *vcpu;
 
-	for_each_online_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_arch_vcpu_free(vcpu);
 }
 
@@ -548,20 +547,18 @@ bool kvm_arch_intc_initialized(struct kvm *kvm)
 
 void kvm_arm_halt_guest(struct kvm *kvm)
 {
-	int i;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		vcpu->arch.pause = true;
 	kvm_make_all_cpus_request(kvm, KVM_REQ_SLEEP);
 }
 
 void kvm_arm_resume_guest(struct kvm *kvm)
 {
-	int i;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		vcpu->arch.pause = false;
 		swake_up(kvm_arch_vcpu_wq(vcpu));
 	}
@@ -1440,10 +1437,9 @@ static void check_kvm_target_cpu(void *ret)
 struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr)
 {
 	struct kvm_vcpu *vcpu;
-	int i;
 
 	mpidr &= MPIDR_HWID_BITMASK;
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		if (mpidr == kvm_vcpu_get_mpidr_aff(vcpu))
 			return vcpu;
 	}
diff --git a/virt/kvm/arm/pmu.c b/virt/kvm/arm/pmu.c
index 8a9c42366db7..84f526cfdb35 100644
--- a/virt/kvm/arm/pmu.c
+++ b/virt/kvm/arm/pmu.c
@@ -512,10 +512,9 @@ static int kvm_arm_pmu_v3_init(struct kvm_vcpu *vcpu)
  */
 static bool pmu_irq_is_valid(struct kvm *kvm, int irq)
 {
-	int i;
 	struct kvm_vcpu *vcpu;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		if (!kvm_arm_pmu_irq_initialized(vcpu))
 			continue;
 
diff --git a/virt/kvm/arm/psci.c b/virt/kvm/arm/psci.c
index f1e363bab5e8..876c965e0886 100644
--- a/virt/kvm/arm/psci.c
+++ b/virt/kvm/arm/psci.c
@@ -129,7 +129,7 @@ static unsigned long kvm_psci_vcpu_on(struct kvm_vcpu *source_vcpu)
 
 static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
 {
-	int i, matching_cpus = 0;
+	int matching_cpus = 0;
 	unsigned long mpidr;
 	unsigned long target_affinity;
 	unsigned long target_affinity_mask;
@@ -152,7 +152,7 @@ static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
 	 * If one or more VCPU matching target affinity are running
 	 * then ON else OFF
 	 */
-	kvm_for_each_vcpu(i, tmp, kvm) {
+	kvm_for_each_vcpu(tmp, kvm) {
 		mpidr = kvm_vcpu_get_mpidr_aff(tmp);
 		if ((mpidr & target_affinity_mask) == target_affinity) {
 			matching_cpus++;
@@ -169,7 +169,6 @@ static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
 
 static void kvm_prepare_system_event(struct kvm_vcpu *vcpu, u32 type)
 {
-	int i;
 	struct kvm_vcpu *tmp;
 
 	/*
@@ -181,7 +180,7 @@ static void kvm_prepare_system_event(struct kvm_vcpu *vcpu, u32 type)
 	 * after this call is handled and before the VCPUs have been
 	 * re-initialized.
 	 */
-	kvm_for_each_vcpu(i, tmp, vcpu->kvm)
+	kvm_for_each_vcpu(tmp, vcpu->kvm)
 		tmp->arch.power_off = true;
 	kvm_make_all_cpus_request(vcpu->kvm, KVM_REQ_SLEEP);
 
diff --git a/virt/kvm/arm/vgic/vgic-init.c b/virt/kvm/arm/vgic/vgic-init.c
index feb766f74c34..c926d7b30149 100644
--- a/virt/kvm/arm/vgic/vgic-init.c
+++ b/virt/kvm/arm/vgic/vgic-init.c
@@ -119,7 +119,7 @@ void kvm_vgic_vcpu_early_init(struct kvm_vcpu *vcpu)
  */
 int kvm_vgic_create(struct kvm *kvm, u32 type)
 {
-	int i, ret;
+	int ret;
 	struct kvm_vcpu *vcpu;
 
 	if (irqchip_in_kernel(kvm))
@@ -143,7 +143,7 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)
 	if (!lock_all_vcpus(kvm))
 		return -EBUSY;
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		if (vcpu->arch.has_run_once) {
 			ret = -EBUSY;
 			goto out_unlock;
@@ -266,7 +266,7 @@ int vgic_init(struct kvm *kvm)
 {
 	struct vgic_dist *dist = &kvm->arch.vgic;
 	struct kvm_vcpu *vcpu;
-	int ret = 0, i;
+	int ret = 0;
 
 	if (vgic_initialized(kvm))
 		return 0;
@@ -279,7 +279,7 @@ int vgic_init(struct kvm *kvm)
 	if (ret)
 		goto out;
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_vgic_vcpu_enable(vcpu);
 
 	ret = kvm_vgic_setup_default_irq_routing(kvm);
@@ -327,13 +327,12 @@ void kvm_vgic_vcpu_destroy(struct kvm_vcpu *vcpu)
 static void __kvm_vgic_destroy(struct kvm *kvm)
 {
 	struct kvm_vcpu *vcpu;
-	int i;
 
 	vgic_debug_destroy(kvm);
 
 	kvm_vgic_dist_destroy(kvm);
 
-	kvm_for_each_vcpu(i, vcpu, kvm)
+	kvm_for_each_vcpu(vcpu, kvm)
 		kvm_vgic_vcpu_destroy(vcpu);
 }
 
diff --git a/virt/kvm/arm/vgic/vgic-kvm-device.c b/virt/kvm/arm/vgic/vgic-kvm-device.c
index c5124737c7fc..3c3e865dc98b 100644
--- a/virt/kvm/arm/vgic/vgic-kvm-device.c
+++ b/virt/kvm/arm/vgic/vgic-kvm-device.c
@@ -257,31 +257,33 @@ int vgic_v2_parse_attr(struct kvm_device *dev, struct kvm_device_attr *attr,
 	return 0;
 }
 
-/* unlocks vcpus from @vcpu_lock_idx and smaller */
-static void unlock_vcpus(struct kvm *kvm, int vcpu_lock_idx)
+/* unlocks vcpus from up to @last_locked */
+static void unlock_vcpus(struct kvm *kvm, struct kvm_vcpu *last_locked)
 {
-	struct kvm_vcpu *tmp_vcpu;
+	if (last_locked) {
+		struct kvm_vcpu *tmp_vcpu;
 
-	for (; vcpu_lock_idx >= 0; vcpu_lock_idx--) {
-		tmp_vcpu = kvm_get_vcpu(kvm, vcpu_lock_idx);
-		mutex_unlock(&tmp_vcpu->mutex);
+		kvm_for_each_vcpu(tmp_vcpu, kvm) {
+			mutex_unlock(&tmp_vcpu->mutex);
+
+			if (tmp_vcpu == last_locked)
+				return;
+		}
 	}
 }
 
 void unlock_all_vcpus(struct kvm *kvm)
 {
-	int i;
 	struct kvm_vcpu *tmp_vcpu;
 
-	kvm_for_each_vcpu(i, tmp_vcpu, kvm)
+	kvm_for_each_vcpu(tmp_vcpu, kvm)
 		mutex_unlock(&tmp_vcpu->mutex);
 }
 
 /* Returns true if all vcpus were locked, false otherwise */
 bool lock_all_vcpus(struct kvm *kvm)
 {
-	struct kvm_vcpu *tmp_vcpu;
-	int c;
+	struct kvm_vcpu *tmp_vcpu, *last_locked = NULL;
 
 	/*
 	 * Any time a vcpu is run, vcpu_load is called which tries to grab the
@@ -289,11 +291,13 @@ bool lock_all_vcpus(struct kvm *kvm)
 	 * that no other VCPUs are run and fiddle with the vgic state while we
 	 * access it.
 	 */
-	kvm_for_each_vcpu(c, tmp_vcpu, kvm) {
+	kvm_for_each_vcpu(tmp_vcpu, kvm) {
 		if (!mutex_trylock(&tmp_vcpu->mutex)) {
-			unlock_vcpus(kvm, c - 1);
+			unlock_vcpus(kvm, last_locked);
 			return false;
 		}
+
+		last_locked = tmp_vcpu;
 	}
 
 	return true;
diff --git a/virt/kvm/arm/vgic/vgic-mmio-v2.c b/virt/kvm/arm/vgic/vgic-mmio-v2.c
index 37522e65eb53..bc5f5124be74 100644
--- a/virt/kvm/arm/vgic/vgic-mmio-v2.c
+++ b/virt/kvm/arm/vgic/vgic-mmio-v2.c
@@ -72,7 +72,6 @@ static void vgic_mmio_write_sgir(struct kvm_vcpu *source_vcpu,
 	int intid = val & 0xf;
 	int targets = (val >> 16) & 0xff;
 	int mode = (val >> 24) & 0x03;
-	int c;
 	struct kvm_vcpu *vcpu;
 
 	switch (mode) {
@@ -89,10 +88,10 @@ static void vgic_mmio_write_sgir(struct kvm_vcpu *source_vcpu,
 		return;
 	}
 
-	kvm_for_each_vcpu(c, vcpu, source_vcpu->kvm) {
+	kvm_for_each_vcpu(vcpu, source_vcpu->kvm) {
 		struct vgic_irq *irq;
 
-		if (!(targets & (1U << c)))
+		if (!(targets & (1U << vcpu->vcpus_idx)))
 			continue;
 
 		irq = vgic_get_irq(source_vcpu->kvm, vcpu, intid);
diff --git a/virt/kvm/arm/vgic/vgic-mmio-v3.c b/virt/kvm/arm/vgic/vgic-mmio-v3.c
index 9d4b69b766ec..62fd24431619 100644
--- a/virt/kvm/arm/vgic/vgic-mmio-v3.c
+++ b/virt/kvm/arm/vgic/vgic-mmio-v3.c
@@ -641,21 +641,24 @@ static void vgic_unregister_redist_iodev(struct kvm_vcpu *vcpu)
 
 static int vgic_register_all_redist_iodevs(struct kvm *kvm)
 {
-	struct kvm_vcpu *vcpu;
-	int c, ret = 0;
+	struct kvm_vcpu *vcpu, *last_registered = NULL;
+	int ret = 0;
 
-	kvm_for_each_vcpu(c, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		ret = vgic_register_redist_iodev(vcpu);
 		if (ret)
 			break;
+
+		last_registered = vcpu;
 	}
 
-	if (ret) {
+	if (ret && last_registered) {
 		/* The current c failed, so we start with the previous one. */
 		mutex_lock(&kvm->slots_lock);
-		for (c--; c >= 0; c--) {
-			vcpu = kvm_get_vcpu(kvm, c);
+		kvm_for_each_vcpu(vcpu, kvm) {
 			vgic_unregister_redist_iodev(vcpu);
+			if (vcpu == last_registered)
+				break;
 		}
 		mutex_unlock(&kvm->slots_lock);
 	}
@@ -796,7 +799,7 @@ void vgic_v3_dispatch_sgi(struct kvm_vcpu *vcpu, u64 reg)
 	struct kvm_vcpu *c_vcpu;
 	u16 target_cpus;
 	u64 mpidr;
-	int sgi, c;
+	int sgi;
 	bool broadcast;
 
 	sgi = (reg & ICC_SGI1R_SGI_ID_MASK) >> ICC_SGI1R_SGI_ID_SHIFT;
@@ -812,7 +815,7 @@ void vgic_v3_dispatch_sgi(struct kvm_vcpu *vcpu, u64 reg)
 	 * if we are already finished. This avoids iterating through all
 	 * VCPUs when most of the times we just signal a single VCPU.
 	 */
-	kvm_for_each_vcpu(c, c_vcpu, kvm) {
+	kvm_for_each_vcpu(c_vcpu, kvm) {
 		struct vgic_irq *irq;
 
 		/* Exit early if we have dealt with all requested CPUs */
diff --git a/virt/kvm/arm/vgic/vgic.c b/virt/kvm/arm/vgic/vgic.c
index fed717e07938..08662cd66e8d 100644
--- a/virt/kvm/arm/vgic/vgic.c
+++ b/virt/kvm/arm/vgic/vgic.c
@@ -758,13 +758,12 @@ int kvm_vgic_vcpu_pending_irq(struct kvm_vcpu *vcpu)
 void vgic_kick_vcpus(struct kvm *kvm)
 {
 	struct kvm_vcpu *vcpu;
-	int c;
 
 	/*
 	 * We've injected an interrupt, time to find out who deserves
 	 * a good kick...
 	 */
-	kvm_for_each_vcpu(c, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		if (kvm_vgic_vcpu_pending_irq(vcpu)) {
 			kvm_make_request(KVM_REQ_IRQ_PENDING, vcpu);
 			kvm_vcpu_kick(vcpu);
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 0d2d8b0c785c..6cec58cad6c7 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -207,7 +207,7 @@ static inline bool kvm_kick_many_cpus(const struct cpumask *cpus, bool wait)
 
 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req)
 {
-	int i, cpu, me;
+	int cpu, me;
 	cpumask_var_t cpus;
 	bool called;
 	struct kvm_vcpu *vcpu;
@@ -215,7 +215,7 @@ bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req)
 	zalloc_cpumask_var(&cpus, GFP_ATOMIC);
 
 	me = get_cpu();
-	kvm_for_each_vcpu(i, vcpu, kvm) {
+	kvm_for_each_vcpu(vcpu, kvm) {
 		kvm_make_request(req, vcpu);
 		cpu = vcpu->cpu;
 
@@ -667,6 +667,7 @@ static struct kvm *kvm_create_vm(unsigned long type)
 	mutex_init(&kvm->slots_lock);
 	refcount_set(&kvm->users_count, 1);
 	INIT_LIST_HEAD(&kvm->devices);
+	INIT_LIST_HEAD(&kvm->vcpu_list);
 
 	r = kvm_arch_init_vm(kvm, type);
 	if (r)
@@ -2348,10 +2349,9 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
 {
 	struct kvm *kvm = me->kvm;
 	struct kvm_vcpu *vcpu;
-	int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
-	int yielded = 0;
+	struct kvm_vcpu *last_boosted_vcpu = READ_ONCE(kvm->last_boosted_vcpu);
+	int yielded;
 	int try = 2;
-	int i;
 
 	kvm_vcpu_set_in_spin_loop(me, true);
 	/*
@@ -2361,7 +2361,7 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
 	 * VCPU is holding the lock that we need and will release it.
 	 * We approximate round-robin by starting at the last boosted VCPU.
 	 */
-	kvm_for_each_vcpu_from(i, vcpu, last_boosted_vcpu, kvm) {
+	kvm_for_each_vcpu_from(vcpu, last_boosted_vcpu, kvm) {
 		if (!ACCESS_ONCE(vcpu->preempted))
 			continue;
 		if (vcpu == me)
@@ -2375,11 +2375,12 @@ void kvm_vcpu_on_spin(struct kvm_vcpu *me, bool yield_to_kernel_mode)
 
 		yielded = kvm_vcpu_yield_to(vcpu);
 		if (yielded > 0) {
-			kvm->last_boosted_vcpu = i;
+			WRITE_ONCE(kvm->last_boosted_vcpu, vcpu);
 			break;
 		} else if (yielded < 0 && !try--)
 			break;
 	}
+
 	kvm_vcpu_set_in_spin_loop(me, false);
 
 	/* Ensure vcpu is not eligible during next spinloop */
@@ -2528,6 +2529,7 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
 	}
 
 	kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
+	list_add_tail_rcu(&vcpu->vcpu_list, &kvm->vcpu_list);
 
 	/*
 	 * Pairs with smp_rmb() in kvm_get_vcpu.  Write kvm->vcpus
@@ -3766,13 +3768,12 @@ static const struct file_operations vm_stat_get_per_vm_fops = {
 
 static int vcpu_stat_get_per_vm(void *data, u64 *val)
 {
-	int i;
 	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;
 	struct kvm_vcpu *vcpu;
 
 	*val = 0;
 
-	kvm_for_each_vcpu(i, vcpu, stat_data->kvm)
+	kvm_for_each_vcpu(vcpu, stat_data->kvm)
 		*val += *(u64 *)((void *)vcpu + stat_data->offset);
 
 	return 0;
@@ -3780,14 +3781,13 @@ static int vcpu_stat_get_per_vm(void *data, u64 *val)
 
 static int vcpu_stat_clear_per_vm(void *data, u64 val)
 {
-	int i;
 	struct kvm_stat_data *stat_data = (struct kvm_stat_data *)data;
 	struct kvm_vcpu *vcpu;
 
 	if (val)
 		return -EINVAL;
 
-	kvm_for_each_vcpu(i, vcpu, stat_data->kvm)
+	kvm_for_each_vcpu(vcpu, stat_data->kvm)
 		*(u64 *)((void *)vcpu + stat_data->offset) = 0;
 
 	return 0;
-- 
2.13.3

[PATCH RFC v3 8/9] KVM: implement kvm_for_each_vcpu with a list

[Christoffer Dall]

On Mon, Aug 21, 2017 at 10:35:29PM +0200, Radim Kr?m?? wrote:
> Going through all VCPUs is more natural with a list and the RCU list can
> work as lockless with our constraints.
> 
> This makes kvm->vcpus lose most users, so it will be easier to make
> something out of it.
> 
> A nice side-effect is that the first argument to the macro is gone.
> ARM code was changed a bit to cope with the loss when working with a
> range 0-n and most other places switched to vcpu->vcpus_idx.
> 
> Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
> ---
>  arch/mips/kvm/mips.c                |  4 +---
>  arch/powerpc/kvm/book3s_32_mmu.c    |  3 +--
>  arch/powerpc/kvm/book3s_64_mmu.c    |  3 +--
>  arch/powerpc/kvm/book3s_hv.c        |  7 +++----
>  arch/powerpc/kvm/book3s_pr.c        |  5 ++---
>  arch/powerpc/kvm/book3s_xics.c      |  2 +-
>  arch/powerpc/kvm/book3s_xics.h      |  3 +--
>  arch/powerpc/kvm/book3s_xive.c      | 18 ++++++++----------
>  arch/powerpc/kvm/book3s_xive.h      |  3 +--
>  arch/powerpc/kvm/e500_emulate.c     |  3 +--
>  arch/powerpc/kvm/powerpc.c          |  3 +--
>  arch/s390/kvm/interrupt.c           |  3 +--
>  arch/s390/kvm/kvm-s390.c            | 31 ++++++++++---------------------
>  arch/s390/kvm/kvm-s390.h            |  6 ++----
>  arch/s390/kvm/sigp.c                |  3 +--
>  arch/x86/kvm/hyperv.c               |  3 +--
>  arch/x86/kvm/i8254.c                |  3 +--
>  arch/x86/kvm/i8259.c                |  7 +++----
>  arch/x86/kvm/ioapic.c               |  3 +--
>  arch/x86/kvm/irq_comm.c             | 10 +++++-----
>  arch/x86/kvm/lapic.c                |  5 ++---
>  arch/x86/kvm/svm.c                  |  3 +--
>  arch/x86/kvm/vmx.c                  |  3 +--
>  arch/x86/kvm/x86.c                  | 25 ++++++++++---------------
>  include/linux/kvm_host.h            | 30 +++++++++++++-----------------
>  virt/kvm/arm/arch_timer.c           | 10 ++++------
>  virt/kvm/arm/arm.c                  | 12 ++++--------
>  virt/kvm/arm/pmu.c                  |  3 +--
>  virt/kvm/arm/psci.c                 |  7 +++----
>  virt/kvm/arm/vgic/vgic-init.c       | 11 +++++------
>  virt/kvm/arm/vgic/vgic-kvm-device.c | 28 ++++++++++++++++------------
>  virt/kvm/arm/vgic/vgic-mmio-v2.c    |  5 ++---
>  virt/kvm/arm/vgic/vgic-mmio-v3.c    | 19 +++++++++++--------
>  virt/kvm/arm/vgic/vgic.c            |  3 +--
>  virt/kvm/kvm_main.c                 | 22 +++++++++++-----------
>  35 files changed, 131 insertions(+), 178 deletions(-)
> 
> diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
> index 770c40b9df37..c841cb434486 100644
> --- a/arch/mips/kvm/mips.c
> +++ b/arch/mips/kvm/mips.c
> @@ -162,12 +162,10 @@ int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
>  
>  void kvm_arch_free_vcpus(struct kvm *kvm)
>  {
> -	unsigned int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_arch_vcpu_free(vcpu);
> -	}
>  }
>  
>  static void kvm_mips_free_gpa_pt(struct kvm *kvm)
> diff --git a/arch/powerpc/kvm/book3s_32_mmu.c b/arch/powerpc/kvm/book3s_32_mmu.c
> index 1992676c7a94..1ac7cace49fc 100644
> --- a/arch/powerpc/kvm/book3s_32_mmu.c
> +++ b/arch/powerpc/kvm/book3s_32_mmu.c
> @@ -353,11 +353,10 @@ static void kvmppc_mmu_book3s_32_mtsrin(struct kvm_vcpu *vcpu, u32 srnum,
>  
>  static void kvmppc_mmu_book3s_32_tlbie(struct kvm_vcpu *vcpu, ulong ea, bool large)
>  {
> -	int i;
>  	struct kvm_vcpu *v;
>  
>  	/* flush this VA on all cpus */
> -	kvm_for_each_vcpu(i, v, vcpu->kvm)
> +	kvm_for_each_vcpu(v, vcpu->kvm)
>  		kvmppc_mmu_pte_flush(v, ea, 0x0FFFF000);
>  }
>  
> diff --git a/arch/powerpc/kvm/book3s_64_mmu.c b/arch/powerpc/kvm/book3s_64_mmu.c
> index 29ebe2fd5867..7b043fcc8c88 100644
> --- a/arch/powerpc/kvm/book3s_64_mmu.c
> +++ b/arch/powerpc/kvm/book3s_64_mmu.c
> @@ -534,7 +534,6 @@ static void kvmppc_mmu_book3s_64_tlbie(struct kvm_vcpu *vcpu, ulong va,
>  				       bool large)
>  {
>  	u64 mask = 0xFFFFFFFFFULL;
> -	long i;
>  	struct kvm_vcpu *v;
>  
>  	dprintk("KVM MMU: tlbie(0x%lx)\n", va);
> @@ -559,7 +558,7 @@ static void kvmppc_mmu_book3s_64_tlbie(struct kvm_vcpu *vcpu, ulong va,
>  			mask = 0xFFFFFF000ULL;
>  	}
>  	/* flush this VA on all vcpus */
> -	kvm_for_each_vcpu(i, v, vcpu->kvm)
> +	kvm_for_each_vcpu(v, vcpu->kvm)
>  		kvmppc_mmu_pte_vflush(v, va >> 12, mask);
>  }
>  
> diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c
> index 359c79cdf0cc..4d40537fad20 100644
> --- a/arch/powerpc/kvm/book3s_hv.c
> +++ b/arch/powerpc/kvm/book3s_hv.c
> @@ -1253,9 +1253,8 @@ static void kvmppc_set_lpcr(struct kvm_vcpu *vcpu, u64 new_lpcr,
>  	 */
>  	if ((new_lpcr & LPCR_ILE) != (vc->lpcr & LPCR_ILE)) {
>  		struct kvm_vcpu *vcpu;
> -		int i;
>  
> -		kvm_for_each_vcpu(i, vcpu, kvm) {
> +		kvm_for_each_vcpu(vcpu, kvm) {
>  			if (vcpu->arch.vcore != vc)
>  				continue;
>  			if (new_lpcr & LPCR_ILE)
> @@ -3347,7 +3346,7 @@ static int kvm_vm_ioctl_get_dirty_log_hv(struct kvm *kvm,
>  {
>  	struct kvm_memslots *slots;
>  	struct kvm_memory_slot *memslot;
> -	int i, r;
> +	int r;
>  	unsigned long n;
>  	unsigned long *buf;
>  	struct kvm_vcpu *vcpu;
> @@ -3381,7 +3380,7 @@ static int kvm_vm_ioctl_get_dirty_log_hv(struct kvm *kvm,
>  
>  	/* Harvest dirty bits from VPA and DTL updates */
>  	/* Note: we never modify the SLB shadow buffer areas */
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		spin_lock(&vcpu->arch.vpa_update_lock);
>  		kvmppc_harvest_vpa_dirty(&vcpu->arch.vpa, memslot, buf);
>  		kvmppc_harvest_vpa_dirty(&vcpu->arch.dtl, memslot, buf);
> diff --git a/arch/powerpc/kvm/book3s_pr.c b/arch/powerpc/kvm/book3s_pr.c
> index 69a09444d46e..9ef1b9b7e48a 100644
> --- a/arch/powerpc/kvm/book3s_pr.c
> +++ b/arch/powerpc/kvm/book3s_pr.c
> @@ -251,7 +251,6 @@ static int kvmppc_core_check_requests_pr(struct kvm_vcpu *vcpu)
>  static void do_kvm_unmap_hva(struct kvm *kvm, unsigned long start,
>  			     unsigned long end)
>  {
> -	long i;
>  	struct kvm_vcpu *vcpu;
>  	struct kvm_memslots *slots;
>  	struct kvm_memory_slot *memslot;
> @@ -272,7 +271,7 @@ static void do_kvm_unmap_hva(struct kvm *kvm, unsigned long start,
>  		 */
>  		gfn = hva_to_gfn_memslot(hva_start, memslot);
>  		gfn_end = hva_to_gfn_memslot(hva_end + PAGE_SIZE - 1, memslot);
> -		kvm_for_each_vcpu(i, vcpu, kvm)
> +		kvm_for_each_vcpu(vcpu, kvm)
>  			kvmppc_mmu_pte_pflush(vcpu, gfn << PAGE_SHIFT,
>  					      gfn_end << PAGE_SHIFT);
>  	}
> @@ -1593,7 +1592,7 @@ static int kvm_vm_ioctl_get_dirty_log_pr(struct kvm *kvm,
>  		ga = memslot->base_gfn << PAGE_SHIFT;
>  		ga_end = ga + (memslot->npages << PAGE_SHIFT);
>  
> -		kvm_for_each_vcpu(n, vcpu, kvm)
> +		kvm_for_each_vcpu(vcpu, kvm)
>  			kvmppc_mmu_pte_pflush(vcpu, ga, ga_end);
>  
>  		n = kvm_dirty_bitmap_bytes(memslot);
> diff --git a/arch/powerpc/kvm/book3s_xics.c b/arch/powerpc/kvm/book3s_xics.c
> index d329b2add7e2..9871b76368bd 100644
> --- a/arch/powerpc/kvm/book3s_xics.c
> +++ b/arch/powerpc/kvm/book3s_xics.c
> @@ -966,7 +966,7 @@ static int xics_debug_show(struct seq_file *m, void *private)
>  
>  	seq_printf(m, "=========\nICP state\n=========\n");
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		struct kvmppc_icp *icp = vcpu->arch.icp;
>  		union kvmppc_icp_state state;
>  
> diff --git a/arch/powerpc/kvm/book3s_xics.h b/arch/powerpc/kvm/book3s_xics.h
> index 453c9e518c19..71e33af1e119 100644
> --- a/arch/powerpc/kvm/book3s_xics.h
> +++ b/arch/powerpc/kvm/book3s_xics.h
> @@ -119,9 +119,8 @@ static inline struct kvmppc_icp *kvmppc_xics_find_server(struct kvm *kvm,
>  							 u32 nr)
>  {
>  	struct kvm_vcpu *vcpu = NULL;
> -	int i;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		if (vcpu->arch.icp && nr == vcpu->arch.icp->server_num)
>  			return vcpu->arch.icp;
>  	}
> diff --git a/arch/powerpc/kvm/book3s_xive.c b/arch/powerpc/kvm/book3s_xive.c
> index 08b200a0bbce..22222a540439 100644
> --- a/arch/powerpc/kvm/book3s_xive.c
> +++ b/arch/powerpc/kvm/book3s_xive.c
> @@ -182,7 +182,7 @@ static int xive_check_provisioning(struct kvm *kvm, u8 prio)
>  {
>  	struct kvmppc_xive *xive = kvm->arch.xive;
>  	struct kvm_vcpu *vcpu;
> -	int i, rc;
> +	int rc;
>  
>  	lockdep_assert_held(&kvm->lock);
>  
> @@ -193,7 +193,7 @@ static int xive_check_provisioning(struct kvm *kvm, u8 prio)
>  	pr_devel("Provisioning prio... %d\n", prio);
>  
>  	/* Provision each VCPU and enable escalations */
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		if (!vcpu->arch.xive_vcpu)
>  			continue;
>  		rc = xive_provision_queue(vcpu, prio);
> @@ -252,7 +252,7 @@ static int xive_try_pick_queue(struct kvm_vcpu *vcpu, u8 prio)
>  static int xive_select_target(struct kvm *kvm, u32 *server, u8 prio)
>  {
>  	struct kvm_vcpu *vcpu;
> -	int i, rc;
> +	int rc;
>  
>  	/* Locate target server */
>  	vcpu = kvmppc_xive_find_server(kvm, *server);
> @@ -271,7 +271,7 @@ static int xive_select_target(struct kvm *kvm, u32 *server, u8 prio)
>  	pr_devel(" .. failed, looking up candidate...\n");
>  
>  	/* Failed, pick another VCPU */
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		if (!vcpu->arch.xive_vcpu)
>  			continue;
>  		rc = xive_try_pick_queue(vcpu, prio);
> @@ -1237,7 +1237,7 @@ static void xive_pre_save_queue(struct kvmppc_xive *xive, struct xive_q *q)
>  static void xive_pre_save_scan(struct kvmppc_xive *xive)
>  {
>  	struct kvm_vcpu *vcpu = NULL;
> -	int i, j;
> +	int j;
>  
>  	/*
>  	 * See comment in xive_get_source() about how this
> @@ -1252,7 +1252,7 @@ static void xive_pre_save_scan(struct kvmppc_xive *xive)
>  	}
>  
>  	/* Then scan the queues and update the "in_queue" flag */
> -	kvm_for_each_vcpu(i, vcpu, xive->kvm) {
> +	kvm_for_each_vcpu(vcpu, xive->kvm) {
>  		struct kvmppc_xive_vcpu *xc = vcpu->arch.xive_vcpu;
>  		if (!xc)
>  			continue;
> @@ -1418,9 +1418,8 @@ static bool xive_check_delayed_irq(struct kvmppc_xive *xive, u32 irq)
>  {
>  	struct kvm *kvm = xive->kvm;
>  	struct kvm_vcpu *vcpu = NULL;
> -	int i;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		struct kvmppc_xive_vcpu *xc = vcpu->arch.xive_vcpu;
>  
>  		if (!xc)
> @@ -1787,14 +1786,13 @@ static int xive_debug_show(struct seq_file *m, void *private)
>  	u64 t_vm_h_cppr = 0;
>  	u64 t_vm_h_eoi = 0;
>  	u64 t_vm_h_ipi = 0;
> -	unsigned int i;
>  
>  	if (!kvm)
>  		return 0;
>  
>  	seq_printf(m, "=========\nVCPU state\n=========\n");
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		struct kvmppc_xive_vcpu *xc = vcpu->arch.xive_vcpu;
>  
>  		if (!xc)
> diff --git a/arch/powerpc/kvm/book3s_xive.h b/arch/powerpc/kvm/book3s_xive.h
> index 5938f7644dc1..5bc33db0924c 100644
> --- a/arch/powerpc/kvm/book3s_xive.h
> +++ b/arch/powerpc/kvm/book3s_xive.h
> @@ -175,9 +175,8 @@ struct kvmppc_xive_vcpu {
>  static inline struct kvm_vcpu *kvmppc_xive_find_server(struct kvm *kvm, u32 nr)
>  {
>  	struct kvm_vcpu *vcpu = NULL;
> -	int i;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		if (vcpu->arch.xive_vcpu && nr == vcpu->arch.xive_vcpu->server_num)
>  			return vcpu;
>  	}
> diff --git a/arch/powerpc/kvm/e500_emulate.c b/arch/powerpc/kvm/e500_emulate.c
> index 990db69a1d0b..56151681afcc 100644
> --- a/arch/powerpc/kvm/e500_emulate.c
> +++ b/arch/powerpc/kvm/e500_emulate.c
> @@ -68,13 +68,12 @@ static int kvmppc_e500_emul_msgsnd(struct kvm_vcpu *vcpu, int rb)
>  	ulong param = vcpu->arch.gpr[rb];
>  	int prio = dbell2prio(rb);
>  	int pir = param & PPC_DBELL_PIR_MASK;
> -	int i;
>  	struct kvm_vcpu *cvcpu;
>  
>  	if (prio < 0)
>  		return EMULATE_FAIL;
>  
> -	kvm_for_each_vcpu(i, cvcpu, vcpu->kvm) {
> +	kvm_for_each_vcpu(cvcpu, vcpu->kvm) {
>  		int cpir = cvcpu->arch.shared->pir;
>  		if ((param & PPC_DBELL_MSG_BRDCAST) || (cpir == pir)) {
>  			set_bit(prio, &cvcpu->arch.pending_exceptions);
> diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
> index 1c563545473c..633d3bb501c1 100644
> --- a/arch/powerpc/kvm/powerpc.c
> +++ b/arch/powerpc/kvm/powerpc.c
> @@ -458,10 +458,9 @@ int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
>  
>  void kvm_arch_free_vcpus(struct kvm *kvm)
>  {
> -	int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_arch_vcpu_free(vcpu);
>  }
>  
> diff --git a/arch/s390/kvm/interrupt.c b/arch/s390/kvm/interrupt.c
> index a619ddae610d..d503b25638af 100644
> --- a/arch/s390/kvm/interrupt.c
> +++ b/arch/s390/kvm/interrupt.c
> @@ -2287,7 +2287,6 @@ static int flic_ais_mode_set_all(struct kvm *kvm, struct kvm_device_attr *attr)
>  static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
>  {
>  	int r = 0;
> -	unsigned int i;
>  	struct kvm_vcpu *vcpu;
>  
>  	switch (attr->group) {
> @@ -2308,7 +2307,7 @@ static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
>  		 * about late coming workers.
>  		 */
>  		synchronize_srcu(&dev->kvm->srcu);
> -		kvm_for_each_vcpu(i, vcpu, dev->kvm)
> +		kvm_for_each_vcpu(vcpu, dev->kvm)
>  			kvm_clear_async_pf_completion_queue(vcpu);
>  		break;
>  	case KVM_DEV_FLIC_ADAPTER_REGISTER:
> diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
> index bb6278d45a25..3e64e3eb2a63 100644
> --- a/arch/s390/kvm/kvm-s390.c
> +++ b/arch/s390/kvm/kvm-s390.c
> @@ -174,12 +174,11 @@ static int kvm_clock_sync(struct notifier_block *notifier, unsigned long val,
>  {
>  	struct kvm *kvm;
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  	unsigned long long *delta = v;
>  
>  	list_for_each_entry(kvm, &vm_list, vm_list) {
>  		kvm->arch.epoch -= *delta;
> -		kvm_for_each_vcpu(i, vcpu, kvm) {
> +		kvm_for_each_vcpu(vcpu, kvm) {
>  			vcpu->arch.sie_block->epoch -= *delta;
>  			if (vcpu->arch.cputm_enabled)
>  				vcpu->arch.cputm_start += *delta;
> @@ -491,12 +490,10 @@ int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
>  
>  static void icpt_operexc_on_all_vcpus(struct kvm *kvm)
>  {
> -	unsigned int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_s390_sync_request(KVM_REQ_ICPT_OPEREXC, vcpu);
> -	}
>  }
>  
>  static int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
> @@ -705,7 +702,6 @@ static void kvm_s390_vcpu_crypto_setup(struct kvm_vcpu *vcpu);
>  static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
>  {
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  
>  	if (!test_kvm_facility(kvm, 76))
>  		return -EINVAL;
> @@ -743,7 +739,7 @@ static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
>  		return -ENXIO;
>  	}
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		kvm_s390_vcpu_crypto_setup(vcpu);
>  		exit_sie(vcpu);
>  	}
> @@ -753,10 +749,9 @@ static int kvm_s390_vm_set_crypto(struct kvm *kvm, struct kvm_device_attr *attr)
>  
>  static void kvm_s390_sync_request_broadcast(struct kvm *kvm, int req)
>  {
> -	int cx;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(cx, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_s390_sync_request(req, vcpu);
>  }
>  
> @@ -1943,10 +1938,9 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
>  
>  void kvm_arch_free_vcpus(struct kvm *kvm)
>  {
> -	unsigned int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_arch_vcpu_destroy(vcpu);
>  }
>  
> @@ -2050,7 +2044,6 @@ static int sca_switch_to_extended(struct kvm *kvm)
>  	struct bsca_block *old_sca = kvm->arch.sca;
>  	struct esca_block *new_sca;
>  	struct kvm_vcpu *vcpu;
> -	unsigned int vcpu_idx;
>  	u32 scaol, scaoh;
>  
>  	new_sca = alloc_pages_exact(sizeof(*new_sca), GFP_KERNEL|__GFP_ZERO);
> @@ -2065,7 +2058,7 @@ static int sca_switch_to_extended(struct kvm *kvm)
>  
>  	sca_copy_b_to_e(new_sca, old_sca);
>  
> -	kvm_for_each_vcpu(vcpu_idx, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		vcpu->arch.sie_block->scaoh = scaoh;
>  		vcpu->arch.sie_block->scaol = scaol;
>  		vcpu->arch.sie_block->ecb2 |= ECB2_ESCA;
> @@ -2491,14 +2484,13 @@ static void kvm_gmap_notifier(struct gmap *gmap, unsigned long start,
>  	struct kvm *kvm = gmap->private;
>  	struct kvm_vcpu *vcpu;
>  	unsigned long prefix;
> -	int i;
>  
>  	if (gmap_is_shadow(gmap))
>  		return;
>  	if (start >= 1UL << 31)
>  		/* We are only interested in prefix pages */
>  		return;
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		/* match against both prefix pages */
>  		prefix = kvm_s390_get_prefix(vcpu);
>  		if (prefix <= end && start <= prefix + 2*PAGE_SIZE - 1) {
> @@ -2856,13 +2848,12 @@ static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu)
>  void kvm_s390_set_tod_clock(struct kvm *kvm, u64 tod)
>  {
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  
>  	mutex_lock(&kvm->lock);
>  	preempt_disable();
>  	kvm->arch.epoch = tod - get_tod_clock();
>  	kvm_s390_vcpu_block_all(kvm);
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		vcpu->arch.sie_block->epoch = kvm->arch.epoch;
>  	kvm_s390_vcpu_unblock_all(kvm);
>  	preempt_enable();
> @@ -3389,12 +3380,10 @@ static void __disable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
>  
>  static void __disable_ibs_on_all_vcpus(struct kvm *kvm)
>  {
> -	unsigned int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		__disable_ibs_on_vcpu(vcpu);
> -	}
>  }
>  
>  static void __enable_ibs_on_vcpu(struct kvm_vcpu *vcpu)
> @@ -3462,7 +3451,7 @@ void kvm_s390_vcpu_stop(struct kvm_vcpu *vcpu)
>  		 * As we only have one VCPU left, we want to enable the IBS
>  		 * facility for that VCPU to speed it up.
>  		 */
> -		kvm_for_each_vcpu(i, started_vcpu, vcpu->kvm)
> +		kvm_for_each_vcpu(started_vcpu, vcpu->kvm)
>  			if (!is_vcpu_stopped(started_vcpu)) {
>  				__enable_ibs_on_vcpu(started_vcpu);
>  				break;
> diff --git a/arch/s390/kvm/kvm-s390.h b/arch/s390/kvm/kvm-s390.h
> index 6fedc8bc7a37..6077a724630c 100644
> --- a/arch/s390/kvm/kvm-s390.h
> +++ b/arch/s390/kvm/kvm-s390.h
> @@ -294,20 +294,18 @@ int kvm_s390_handle_diag(struct kvm_vcpu *vcpu);
>  
>  static inline void kvm_s390_vcpu_block_all(struct kvm *kvm)
>  {
> -	int i;
>  	struct kvm_vcpu *vcpu;
>  
>  	WARN_ON(!mutex_is_locked(&kvm->lock));
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_s390_vcpu_block(vcpu);
>  }
>  
>  static inline void kvm_s390_vcpu_unblock_all(struct kvm *kvm)
>  {
> -	int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_s390_vcpu_unblock(vcpu);
>  }
>  
> diff --git a/arch/s390/kvm/sigp.c b/arch/s390/kvm/sigp.c
> index 1a252f537081..a31e1c2a994d 100644
> --- a/arch/s390/kvm/sigp.c
> +++ b/arch/s390/kvm/sigp.c
> @@ -158,7 +158,6 @@ static int __sigp_stop_and_store_status(struct kvm_vcpu *vcpu,
>  static int __sigp_set_arch(struct kvm_vcpu *vcpu, u32 parameter)
>  {
>  	int rc;
> -	unsigned int i;
>  	struct kvm_vcpu *v;
>  
>  	switch (parameter & 0xff) {
> @@ -167,7 +166,7 @@ static int __sigp_set_arch(struct kvm_vcpu *vcpu, u32 parameter)
>  		break;
>  	case 1:
>  	case 2:
> -		kvm_for_each_vcpu(i, v, vcpu->kvm) {
> +		kvm_for_each_vcpu(v, vcpu->kvm) {
>  			v->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
>  			kvm_clear_async_pf_completion_queue(v);
>  		}
> diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c
> index dc97f2544b6f..b43cb27bf783 100644
> --- a/arch/x86/kvm/hyperv.c
> +++ b/arch/x86/kvm/hyperv.c
> @@ -109,13 +109,12 @@ static int synic_set_sint(struct kvm_vcpu_hv_synic *synic, int sint,
>  static struct kvm_vcpu *get_vcpu_by_vpidx(struct kvm *kvm, u32 vpidx)
>  {
>  	struct kvm_vcpu *vcpu = NULL;
> -	int i;
>  
>  	if (vpidx < KVM_MAX_VCPUS)
>  		vcpu = kvm_get_vcpu(kvm, vpidx);
>  	if (vcpu && vcpu_to_hv_vcpu(vcpu)->vp_index == vpidx)
>  		return vcpu;
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		if (vcpu_to_hv_vcpu(vcpu)->vp_index == vpidx)
>  			return vcpu;
>  	return NULL;
> diff --git a/arch/x86/kvm/i8254.c b/arch/x86/kvm/i8254.c
> index af192895b1fc..fdd3bff598f3 100644
> --- a/arch/x86/kvm/i8254.c
> +++ b/arch/x86/kvm/i8254.c
> @@ -241,7 +241,6 @@ static void pit_do_work(struct kthread_work *work)
>  	struct kvm_pit *pit = container_of(work, struct kvm_pit, expired);
>  	struct kvm *kvm = pit->kvm;
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  	struct kvm_kpit_state *ps = &pit->pit_state;
>  
>  	if (atomic_read(&ps->reinject) && !atomic_xchg(&ps->irq_ack, 0))
> @@ -260,7 +259,7 @@ static void pit_do_work(struct kthread_work *work)
>  	 * also be simultaneously delivered through PIC and IOAPIC.
>  	 */
>  	if (atomic_read(&kvm->arch.vapics_in_nmi_mode) > 0)
> -		kvm_for_each_vcpu(i, vcpu, kvm)
> +		kvm_for_each_vcpu(vcpu, kvm)
>  			kvm_apic_nmi_wd_deliver(vcpu);
>  }
>  
> diff --git a/arch/x86/kvm/i8259.c b/arch/x86/kvm/i8259.c
> index bdcd4139eca9..c4f57fbd7316 100644
> --- a/arch/x86/kvm/i8259.c
> +++ b/arch/x86/kvm/i8259.c
> @@ -50,14 +50,13 @@ static void pic_unlock(struct kvm_pic *s)
>  {
>  	bool wakeup = s->wakeup_needed;
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  
>  	s->wakeup_needed = false;
>  
>  	spin_unlock(&s->lock);
>  
>  	if (wakeup) {
> -		kvm_for_each_vcpu(i, vcpu, s->kvm) {
> +		kvm_for_each_vcpu(vcpu, s->kvm) {
>  			if (kvm_apic_accept_pic_intr(vcpu)) {
>  				kvm_make_request(KVM_REQ_EVENT, vcpu);
>  				kvm_vcpu_kick(vcpu);
> @@ -270,7 +269,7 @@ int kvm_pic_read_irq(struct kvm *kvm)
>  
>  static void kvm_pic_reset(struct kvm_kpic_state *s)
>  {
> -	int irq, i;
> +	int irq;
>  	struct kvm_vcpu *vcpu;
>  	u8 edge_irr = s->irr & ~s->elcr;
>  	bool found = false;
> @@ -287,7 +286,7 @@ static void kvm_pic_reset(struct kvm_kpic_state *s)
>  	}
>  	s->init_state = 1;
>  
> -	kvm_for_each_vcpu(i, vcpu, s->pics_state->kvm)
> +	kvm_for_each_vcpu(vcpu, s->pics_state->kvm)
>  		if (kvm_apic_accept_pic_intr(vcpu)) {
>  			found = true;
>  			break;
> diff --git a/arch/x86/kvm/ioapic.c b/arch/x86/kvm/ioapic.c
> index 1a29da6c0558..2216e6d34fb7 100644
> --- a/arch/x86/kvm/ioapic.c
> +++ b/arch/x86/kvm/ioapic.c
> @@ -146,13 +146,12 @@ void kvm_rtc_eoi_tracking_restore_one(struct kvm_vcpu *vcpu)
>  static void kvm_rtc_eoi_tracking_restore_all(struct kvm_ioapic *ioapic)
>  {
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  
>  	if (RTC_GSI >= IOAPIC_NUM_PINS)
>  		return;
>  
>  	rtc_irq_eoi_tracking_reset(ioapic);
> -	kvm_for_each_vcpu(i, vcpu, ioapic->kvm)
> +	kvm_for_each_vcpu(vcpu, ioapic->kvm)
>  	    __rtc_irq_eoi_tracking_restore_one(vcpu);
>  }
>  
> diff --git a/arch/x86/kvm/irq_comm.c b/arch/x86/kvm/irq_comm.c
> index 3cc3b2d130a0..a8f777f2b0e6 100644
> --- a/arch/x86/kvm/irq_comm.c
> +++ b/arch/x86/kvm/irq_comm.c
> @@ -58,7 +58,7 @@ static int kvm_set_ioapic_irq(struct kvm_kernel_irq_routing_entry *e,
>  int kvm_irq_delivery_to_apic(struct kvm *kvm, struct kvm_lapic *src,
>  		struct kvm_lapic_irq *irq, struct dest_map *dest_map)
>  {
> -	int i, r = -1;
> +	int r = -1;
>  	struct kvm_vcpu *vcpu, *lowest = NULL;
>  	unsigned long dest_vcpu_bitmap[BITS_TO_LONGS(KVM_MAX_VCPUS)];
>  	unsigned int dest_vcpus = 0;
> @@ -74,7 +74,7 @@ int kvm_irq_delivery_to_apic(struct kvm *kvm, struct kvm_lapic *src,
>  
>  	memset(dest_vcpu_bitmap, 0, sizeof(dest_vcpu_bitmap));
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		if (!kvm_apic_present(vcpu))
>  			continue;
>  
> @@ -93,7 +93,7 @@ int kvm_irq_delivery_to_apic(struct kvm *kvm, struct kvm_lapic *src,
>  				else if (kvm_apic_compare_prio(vcpu, lowest) < 0)
>  					lowest = vcpu;
>  			} else {
> -				__set_bit(i, dest_vcpu_bitmap);
> +				__set_bit(vcpu->vcpus_idx, dest_vcpu_bitmap);
>  				dest_vcpus++;
>  			}
>  		}
> @@ -335,13 +335,13 @@ int kvm_set_routing_entry(struct kvm *kvm,
>  bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
>  			     struct kvm_vcpu **dest_vcpu)
>  {
> -	int i, r = 0;
> +	int r = 0;
>  	struct kvm_vcpu *vcpu;
>  
>  	if (kvm_intr_is_single_vcpu_fast(kvm, irq, dest_vcpu))
>  		return true;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		if (!kvm_apic_present(vcpu))
>  			continue;
>  
> diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> index 4f38818db929..a7cdd6baa38b 100644
> --- a/arch/x86/kvm/lapic.c
> +++ b/arch/x86/kvm/lapic.c
> @@ -166,12 +166,11 @@ static void recalculate_apic_map(struct kvm *kvm)
>  {
>  	struct kvm_apic_map *new, *old = NULL;
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  	u32 max_id = 255; /* enough space for any xAPIC ID */
>  
>  	mutex_lock(&kvm->arch.apic_map_lock);
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		if (kvm_apic_present(vcpu))
>  			max_id = max(max_id, kvm_x2apic_id(vcpu->arch.apic));
>  
> @@ -183,7 +182,7 @@ static void recalculate_apic_map(struct kvm *kvm)
>  
>  	new->max_apic_id = max_id;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		struct kvm_lapic *apic = vcpu->arch.apic;
>  		struct kvm_lapic **cluster;
>  		u16 mask;
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index 1fa9ee5660f4..605c18003f55 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -3807,7 +3807,6 @@ static int avic_incomplete_ipi_interception(struct vcpu_svm *svm)
>  		kvm_lapic_reg_write(apic, APIC_ICR, icrl);
>  		break;
>  	case AVIC_IPI_FAILURE_TARGET_NOT_RUNNING: {
> -		int i;
>  		struct kvm_vcpu *vcpu;
>  		struct kvm *kvm = svm->vcpu.kvm;
>  		struct kvm_lapic *apic = svm->vcpu.arch.apic;
> @@ -3817,7 +3816,7 @@ static int avic_incomplete_ipi_interception(struct vcpu_svm *svm)
>  		 * set the appropriate IRR bits on the valid target
>  		 * vcpus. So, we just need to kick the appropriate vcpu.
>  		 */
> -		kvm_for_each_vcpu(i, vcpu, kvm) {
> +		kvm_for_each_vcpu(vcpu, kvm) {
>  			bool m = kvm_apic_match_dest(vcpu, apic,
>  						     icrl & KVM_APIC_SHORT_MASK,
>  						     GET_APIC_DEST_FIELD(icrh),
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index df8d2f127508..ae0f04e26fec 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -8472,7 +8472,6 @@ static void vmx_flush_pml_buffer(struct kvm_vcpu *vcpu)
>   */
>  static void kvm_flush_pml_buffers(struct kvm *kvm)
>  {
> -	int i;
>  	struct kvm_vcpu *vcpu;
>  	/*
>  	 * We only need to kick vcpu out of guest mode here, as PML buffer
> @@ -8480,7 +8479,7 @@ static void kvm_flush_pml_buffers(struct kvm *kvm)
>  	 * vcpus running in guest are possible to have unflushed GPAs in PML
>  	 * buffer.
>  	 */
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_vcpu_kick(vcpu);
>  }
>  
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index d021746f1fdf..caea24d3ddb0 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -1734,7 +1734,6 @@ void kvm_make_mclock_inprogress_request(struct kvm *kvm)
>  static void kvm_gen_update_masterclock(struct kvm *kvm)
>  {
>  #ifdef CONFIG_X86_64
> -	int i;
>  	struct kvm_vcpu *vcpu;
>  	struct kvm_arch *ka = &kvm->arch;
>  
> @@ -1743,11 +1742,11 @@ static void kvm_gen_update_masterclock(struct kvm *kvm)
>  	/* no guest entries from this point */
>  	pvclock_update_vm_gtod_copy(kvm);
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
>  
>  	/* guest entries allowed */
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_clear_request(KVM_REQ_MCLOCK_INPROGRESS, vcpu);
>  
>  	spin_unlock(&ka->pvclock_gtod_sync_lock);
> @@ -1945,14 +1944,13 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)
>  
>  static void kvmclock_update_fn(struct work_struct *work)
>  {
> -	int i;
>  	struct delayed_work *dwork = to_delayed_work(work);
>  	struct kvm_arch *ka = container_of(dwork, struct kvm_arch,
>  					   kvmclock_update_work);
>  	struct kvm *kvm = container_of(ka, struct kvm, arch);
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
>  		kvm_vcpu_kick(vcpu);
>  	}
> @@ -5844,7 +5842,7 @@ static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long va
>  	struct cpufreq_freqs *freq = data;
>  	struct kvm *kvm;
>  	struct kvm_vcpu *vcpu;
> -	int i, send_ipi = 0;
> +	int send_ipi = 0;
>  
>  	/*
>  	 * We allow guests to temporarily run on slowing clocks,
> @@ -5894,7 +5892,7 @@ static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long va
>  
>  	spin_lock(&kvm_lock);
>  	list_for_each_entry(kvm, &vm_list, vm_list) {
> -		kvm_for_each_vcpu(i, vcpu, kvm) {
> +		kvm_for_each_vcpu(vcpu, kvm) {
>  			if (vcpu->cpu != freq->cpu)
>  				continue;
>  			kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
> @@ -6035,11 +6033,10 @@ static void pvclock_gtod_update_fn(struct work_struct *work)
>  	struct kvm *kvm;
>  
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  
>  	spin_lock(&kvm_lock);
>  	list_for_each_entry(kvm, &vm_list, vm_list)
> -		kvm_for_each_vcpu(i, vcpu, kvm)
> +		kvm_for_each_vcpu(vcpu, kvm)
>  			kvm_make_request(KVM_REQ_MASTERCLOCK_UPDATE, vcpu);
>  	atomic_set(&kvm_guest_has_master_clock, 0);
>  	spin_unlock(&kvm_lock);
> @@ -7783,7 +7780,6 @@ int kvm_arch_hardware_enable(void)
>  {
>  	struct kvm *kvm;
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  	int ret;
>  	u64 local_tsc;
>  	u64 max_tsc = 0;
> @@ -7797,7 +7793,7 @@ int kvm_arch_hardware_enable(void)
>  	local_tsc = rdtsc();
>  	stable = !check_tsc_unstable();
>  	list_for_each_entry(kvm, &vm_list, vm_list) {
> -		kvm_for_each_vcpu(i, vcpu, kvm) {
> +		kvm_for_each_vcpu(vcpu, kvm) {
>  			if (!stable && vcpu->cpu == smp_processor_id())
>  				kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
>  			if (stable && vcpu->arch.last_host_tsc > local_tsc) {
> @@ -7850,7 +7846,7 @@ int kvm_arch_hardware_enable(void)
>  		u64 delta_cyc = max_tsc - local_tsc;
>  		list_for_each_entry(kvm, &vm_list, vm_list) {
>  			kvm->arch.backwards_tsc_observed = true;
> -			kvm_for_each_vcpu(i, vcpu, kvm) {
> +			kvm_for_each_vcpu(vcpu, kvm) {
>  				vcpu->arch.tsc_offset_adjustment += delta_cyc;
>  				vcpu->arch.last_host_tsc = local_tsc;
>  				kvm_make_request(KVM_REQ_MASTERCLOCK_UPDATE, vcpu);
> @@ -8079,17 +8075,16 @@ static void kvm_unload_vcpu_mmu(struct kvm_vcpu *vcpu)
>  
>  void kvm_arch_free_vcpus(struct kvm *kvm)
>  {
> -	unsigned int i;
>  	struct kvm_vcpu *vcpu;
>  
>  	/*
>  	 * Unpin any mmu pages first.
>  	 */
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		kvm_clear_async_pf_completion_queue(vcpu);
>  		kvm_unload_vcpu_mmu(vcpu);
>  	}
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_arch_vcpu_free(vcpu);
>  }
>  
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index a8b9aa563834..5417dac55272 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -216,6 +216,7 @@ struct kvm_mmio_fragment {
>  
>  struct kvm_vcpu {
>  	struct kvm *kvm;
> +	struct list_head vcpu_list;
>  #ifdef CONFIG_PREEMPT_NOTIFIERS
>  	struct preempt_notifier preempt_notifier;
>  #endif
> @@ -393,6 +394,7 @@ struct kvm {
>  	struct mm_struct *mm; /* userspace tied to this vm */
>  	struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
>  	struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
> +	struct list_head vcpu_list;
>  
>  	/*
>  	 * created_vcpus is protected by kvm->lock, and is incremented
> @@ -402,7 +404,7 @@ struct kvm {
>  	 */
>  	atomic_t online_vcpus;
>  	int created_vcpus;
> -	int last_boosted_vcpu;
> +	struct kvm_vcpu *last_boosted_vcpu;
>  	struct list_head vm_list;
>  	struct mutex lock;
>  	struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
> @@ -492,29 +494,23 @@ static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
>  	return kvm->vcpus[i];
>  }
>  
> -#define kvm_for_each_vcpu(idx, vcpup, kvm) \
> -	for (idx = 0; \
> -	     idx < atomic_read(&kvm->online_vcpus) && \
> -	     (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
> -	     idx++)
> +#define kvm_for_each_vcpu(vcpup, kvm) \
> +	list_for_each_entry_rcu(vcpup, &kvm->vcpu_list, vcpu_list)
>  
> -#define kvm_for_each_vcpu_from(idx, vcpup, from, kvm) \
> -	for (idx = from, vcpup = kvm_get_vcpu(kvm, idx); \
> +#define kvm_for_each_vcpu_from(vcpup, from, kvm) \
> +	for (vcpup = from; \
>  	     vcpup; \
>  	     ({ \
> -		idx++; \
> -		if (idx >= atomic_read(&kvm->online_vcpus)) \
> -			idx = 0; \
> -		if (idx == from) \
> +		vcpup = list_entry_rcu(vcpup->vcpu_list.next, typeof(*vcpup), vcpu_list); \
> +		if (&vcpup->vcpu_list == &kvm->vcpu_list) \
> +			vcpup = list_entry_rcu(kvm->vcpu_list.next, typeof(*vcpup), vcpu_list); \
> +		if (vcpup == from) \
>  			vcpup = NULL; \
> -		else \
> -			vcpup = kvm_get_vcpu(kvm, idx); \
> -	      }))
> +	     }))
>  
>  static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
>  {
>  	struct kvm_vcpu *vcpu = NULL;
> -	int i;
>  
>  	if (id < 0)
>  		return NULL;
> @@ -522,7 +518,7 @@ static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
>  		vcpu = kvm_get_vcpu(kvm, id);
>  	if (vcpu && vcpu->vcpu_id == id)
>  		return vcpu;
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		if (vcpu->vcpu_id == id)
>  			return vcpu;
>  	return NULL;
> diff --git a/virt/kvm/arm/arch_timer.c b/virt/kvm/arm/arch_timer.c
> index 8e89d63005c7..3e311fa5c7a1 100644
> --- a/virt/kvm/arm/arch_timer.c
> +++ b/virt/kvm/arm/arch_timer.c
> @@ -478,12 +478,11 @@ int kvm_timer_vcpu_reset(struct kvm_vcpu *vcpu)
>  /* Make the updates of cntvoff for all vtimer contexts atomic */
>  static void update_vtimer_cntvoff(struct kvm_vcpu *vcpu, u64 cntvoff)
>  {
> -	int i;
>  	struct kvm *kvm = vcpu->kvm;
>  	struct kvm_vcpu *tmp;
>  
>  	mutex_lock(&kvm->lock);
> -	kvm_for_each_vcpu(i, tmp, kvm)
> +	kvm_for_each_vcpu(tmp, kvm)
>  		vcpu_vtimer(tmp)->cntvoff = cntvoff;
>  
>  	/*
> @@ -622,7 +621,7 @@ void kvm_timer_vcpu_terminate(struct kvm_vcpu *vcpu)
>  static bool timer_irqs_are_valid(struct kvm_vcpu *vcpu)
>  {
>  	int vtimer_irq, ptimer_irq;
> -	int i, ret;
> +	int ret;
>  
>  	vtimer_irq = vcpu_vtimer(vcpu)->irq.irq;
>  	ret = kvm_vgic_set_owner(vcpu, vtimer_irq, vcpu_vtimer(vcpu));
> @@ -634,7 +633,7 @@ static bool timer_irqs_are_valid(struct kvm_vcpu *vcpu)
>  	if (ret)
>  		return false;
>  
> -	kvm_for_each_vcpu(i, vcpu, vcpu->kvm) {
> +	kvm_for_each_vcpu(vcpu, vcpu->kvm) {
>  		if (vcpu_vtimer(vcpu)->irq.irq != vtimer_irq ||
>  		    vcpu_ptimer(vcpu)->irq.irq != ptimer_irq)
>  			return false;
> @@ -720,9 +719,8 @@ void kvm_timer_init_vhe(void)
>  static void set_timer_irqs(struct kvm *kvm, int vtimer_irq, int ptimer_irq)
>  {
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		vcpu_vtimer(vcpu)->irq.irq = vtimer_irq;
>  		vcpu_ptimer(vcpu)->irq.irq = ptimer_irq;
>  	}
> diff --git a/virt/kvm/arm/arm.c b/virt/kvm/arm/arm.c
> index d63aa1107fdb..e105f6b307cc 100644
> --- a/virt/kvm/arm/arm.c
> +++ b/virt/kvm/arm/arm.c
> @@ -168,10 +168,9 @@ int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
>  
>  void kvm_arch_free_vcpus(struct kvm *kvm)
>  {
> -	int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	for_each_online_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_arch_vcpu_free(vcpu);
>  }
>  
> @@ -548,20 +547,18 @@ bool kvm_arch_intc_initialized(struct kvm *kvm)
>  
>  void kvm_arm_halt_guest(struct kvm *kvm)
>  {
> -	int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		vcpu->arch.pause = true;
>  	kvm_make_all_cpus_request(kvm, KVM_REQ_SLEEP);
>  }
>  
>  void kvm_arm_resume_guest(struct kvm *kvm)
>  {
> -	int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		vcpu->arch.pause = false;
>  		swake_up(kvm_arch_vcpu_wq(vcpu));
>  	}
> @@ -1440,10 +1437,9 @@ static void check_kvm_target_cpu(void *ret)
>  struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr)
>  {
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  
>  	mpidr &= MPIDR_HWID_BITMASK;
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		if (mpidr == kvm_vcpu_get_mpidr_aff(vcpu))
>  			return vcpu;
>  	}
> diff --git a/virt/kvm/arm/pmu.c b/virt/kvm/arm/pmu.c
> index 8a9c42366db7..84f526cfdb35 100644
> --- a/virt/kvm/arm/pmu.c
> +++ b/virt/kvm/arm/pmu.c
> @@ -512,10 +512,9 @@ static int kvm_arm_pmu_v3_init(struct kvm_vcpu *vcpu)
>   */
>  static bool pmu_irq_is_valid(struct kvm *kvm, int irq)
>  {
> -	int i;
>  	struct kvm_vcpu *vcpu;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm) {
> +	kvm_for_each_vcpu(vcpu, kvm) {
>  		if (!kvm_arm_pmu_irq_initialized(vcpu))
>  			continue;
>  
> diff --git a/virt/kvm/arm/psci.c b/virt/kvm/arm/psci.c
> index f1e363bab5e8..876c965e0886 100644
> --- a/virt/kvm/arm/psci.c
> +++ b/virt/kvm/arm/psci.c
> @@ -129,7 +129,7 @@ static unsigned long kvm_psci_vcpu_on(struct kvm_vcpu *source_vcpu)
>  
>  static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
>  {
> -	int i, matching_cpus = 0;
> +	int matching_cpus = 0;
>  	unsigned long mpidr;
>  	unsigned long target_affinity;
>  	unsigned long target_affinity_mask;
> @@ -152,7 +152,7 @@ static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
>  	 * If one or more VCPU matching target affinity are running
>  	 * then ON else OFF
>  	 */
> -	kvm_for_each_vcpu(i, tmp, kvm) {
> +	kvm_for_each_vcpu(tmp, kvm) {
>  		mpidr = kvm_vcpu_get_mpidr_aff(tmp);
>  		if ((mpidr & target_affinity_mask) == target_affinity) {
>  			matching_cpus++;
> @@ -169,7 +169,6 @@ static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
>  
>  static void kvm_prepare_system_event(struct kvm_vcpu *vcpu, u32 type)
>  {
> -	int i;
>  	struct kvm_vcpu *tmp;
>  
>  	/*
> @@ -181,7 +180,7 @@ static void kvm_prepare_system_event(struct kvm_vcpu *vcpu, u32 type)
>  	 * after this call is handled and before the VCPUs have been
>  	 * re-initialized.
>  	 */
> -	kvm_for_each_vcpu(i, tmp, vcpu->kvm)
> +	kvm_for_each_vcpu(tmp, vcpu->kvm)
>  		tmp->arch.power_off = true;
>  	kvm_make_all_cpus_request(vcpu->kvm, KVM_REQ_SLEEP);
>  
> diff --git a/virt/kvm/arm/vgic/vgic-init.c b/virt/kvm/arm/vgic/vgic-init.c
> index feb766f74c34..c926d7b30149 100644
> --- a/virt/kvm/arm/vgic/vgic-init.c
> +++ b/virt/kvm/arm/vgic/vgic-init.c
> @@ -119,7 +119,7 @@ void kvm_vgic_vcpu_early_init(struct kvm_vcpu *vcpu)
>   */
>  int kvm_vgic_create(struct kvm *kvm, u32 type)
>  {
> -	int i, ret;
> +	int ret;
>  	struct kvm_vcpu *vcpu;
>  
>  	if (irqchip_in_kernel(kvm))
> @@ -143,7 +143,7 @@ int kvm_vgic_create(struct kvm *kvm, u32 type)
>  	if (!lock_all_vcpus(kvm))
>  		return -EBUSY;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		if (vcpu->arch.has_run_once) {
>  			ret = -EBUSY;
>  			goto out_unlock;
> @@ -266,7 +266,7 @@ int vgic_init(struct kvm *kvm)
>  {
>  	struct vgic_dist *dist = &kvm->arch.vgic;
>  	struct kvm_vcpu *vcpu;
> -	int ret = 0, i;
> +	int ret = 0;
>  
>  	if (vgic_initialized(kvm))
>  		return 0;
> @@ -279,7 +279,7 @@ int vgic_init(struct kvm *kvm)
>  	if (ret)
>  		goto out;
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_vgic_vcpu_enable(vcpu);
>  
>  	ret = kvm_vgic_setup_default_irq_routing(kvm);
> @@ -327,13 +327,12 @@ void kvm_vgic_vcpu_destroy(struct kvm_vcpu *vcpu)
>  static void __kvm_vgic_destroy(struct kvm *kvm)
>  {
>  	struct kvm_vcpu *vcpu;
> -	int i;
>  
>  	vgic_debug_destroy(kvm);
>  
>  	kvm_vgic_dist_destroy(kvm);
>  
> -	kvm_for_each_vcpu(i, vcpu, kvm)
> +	kvm_for_each_vcpu(vcpu, kvm)
>  		kvm_vgic_vcpu_destroy(vcpu);
>  }
>  
> diff --git a/virt/kvm/arm/vgic/vgic-kvm-device.c b/virt/kvm/arm/vgic/vgic-kvm-device.c
> index c5124737c7fc..3c3e865dc98b 100644
> --- a/virt/kvm/arm/vgic/vgic-kvm-device.c
> +++ b/virt/kvm/arm/vgic/vgic-kvm-device.c
> @@ -257,31 +257,33 @@ int vgic_v2_parse_attr(struct kvm_device *dev, struct kvm_device_attr *attr,
>  	return 0;
>  }
>  
> -/* unlocks vcpus from @vcpu_lock_idx and smaller */
> -static void unlock_vcpus(struct kvm *kvm, int vcpu_lock_idx)
> +/* unlocks vcpus from up to @last_locked */

from up to ?

I think you mean up to and including.

> +static void unlock_vcpus(struct kvm *kvm, struct kvm_vcpu *last_locked)
>  {
> -	struct kvm_vcpu *tmp_vcpu;
> +	if (last_locked) {
> +		struct kvm_vcpu *tmp_vcpu;

That would be more nicely written as 
	if (!last_locked)
		return

	kvm_for_each_vcpu(...)

But why not accept a NULL poiner here and call unlock_vcpus from
unlock_all_vcpus ?

>  
> -	for (; vcpu_lock_idx >= 0; vcpu_lock_idx--) {
> -		tmp_vcpu = kvm_get_vcpu(kvm, vcpu_lock_idx);
> -		mutex_unlock(&tmp_vcpu->mutex);
> +		kvm_for_each_vcpu(tmp_vcpu, kvm) {
> +			mutex_unlock(&tmp_vcpu->mutex);
> +
> +			if (tmp_vcpu == last_locked)
> +				return;
> +		}
>  	}
>  }
>  
>  void unlock_all_vcpus(struct kvm *kvm)
>  {
> -	int i;
>  	struct kvm_vcpu *tmp_vcpu;
>  
> -	kvm_for_each_vcpu(i, tmp_vcpu, kvm)
> +	kvm_for_each_vcpu(tmp_vcpu, kvm)
>  		mutex_unlock(&tmp_vcpu->mutex);
>  }
>  
>  /* Returns true if all vcpus were locked, false otherwise */
>  bool lock_all_vcpus(struct kvm *kvm)
>  {
> -	struct kvm_vcpu *tmp_vcpu;
> -	int c;
> +	struct kvm_vcpu *tmp_vcpu, *last_locked = NULL;
>  
>  	/*
>  	 * Any time a vcpu is run, vcpu_load is called which tries to grab the
> @@ -289,11 +291,13 @@ bool lock_all_vcpus(struct kvm *kvm)
>  	 * that no other VCPUs are run and fiddle with the vgic state while we
>  	 * access it.
>  	 */
> -	kvm_for_each_vcpu(c, tmp_vcpu, kvm) {
> +	kvm_for_each_vcpu(tmp_vcpu, kvm) {
>  		if (!mutex_trylock(&tmp_vcpu->mutex)) {
> -			unlock_vcpus(kvm, c - 1);
> +			unlock_vcpus(kvm, last_locked);

I don't fully understand the split of the patches (which I assume aim
for bisectability) and why unlock_all_vcpus had to be changed in the
previous patch but this call to unlock_vcpus() was allowed to remain
until here?

Thanks,
-Christoffer

[PATCH RFC v3 9/9] KVM: split kvm->vcpus into chunks

[Radim Krčmář]

This allows us to have high KVM_VCPU_MAX without wasting too much space
with small guests.  RCU is a viable alternative now that we do not have
to protect the kvm_for_each_vcpu() loop.

Suggested-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Kr?m?? <rkrcmar@redhat.com>
---
 arch/mips/kvm/mips.c     |  2 +-
 arch/x86/kvm/vmx.c       |  2 +-
 include/linux/kvm_host.h | 27 ++++++++++++++++++++-------
 virt/kvm/kvm_main.c      | 27 +++++++++++++++++++++++----
 4 files changed, 45 insertions(+), 13 deletions(-)

diff --git a/arch/mips/kvm/mips.c b/arch/mips/kvm/mips.c
index c841cb434486..7d452163dcef 100644
--- a/arch/mips/kvm/mips.c
+++ b/arch/mips/kvm/mips.c
@@ -488,7 +488,7 @@ int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu,
 	if (irq->cpu == -1)
 		dvcpu = vcpu;
 	else
-		dvcpu = vcpu->kvm->vcpus[irq->cpu];
+		dvcpu = kvm_get_vcpu(vcpu->kvm, irq->cpu);
 
 	if (intr == 2 || intr == 3 || intr == 4) {
 		kvm_mips_callbacks->queue_io_int(dvcpu, irq);
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index ae0f04e26fec..2b92c2de2b3a 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -11741,7 +11741,7 @@ static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
 
 	if (!kvm_arch_has_assigned_device(kvm) ||
 		!irq_remapping_cap(IRQ_POSTING_CAP) ||
-		!kvm_vcpu_apicv_active(kvm->vcpus[0]))
+		!kvm_vcpu_apicv_active(kvm_get_vcpu(kvm, 0)))
 		return 0;
 
 	idx = srcu_read_lock(&kvm->irq_srcu);
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 5417dac55272..5cc3ca8b92b3 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -388,12 +388,16 @@ struct kvm_memslots {
 	int used_slots;
 };
 
+#define KVM_VCPUS_CHUNK_SIZE 128
+#define KVM_VCPUS_CHUNKS_NUM \
+	(round_up(KVM_MAX_VCPUS, KVM_VCPUS_CHUNK_SIZE) / KVM_VCPUS_CHUNK_SIZE)
+
 struct kvm {
 	spinlock_t mmu_lock;
 	struct mutex slots_lock;
 	struct mm_struct *mm; /* userspace tied to this vm */
 	struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
-	struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
+	struct kvm_vcpu **vcpus[KVM_VCPUS_CHUNKS_NUM];
 	struct list_head vcpu_list;
 
 	/*
@@ -484,14 +488,23 @@ static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
 				      !refcount_read(&kvm->users_count));
 }
 
-static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
+static inline struct kvm_vcpu *__kvm_get_vcpu(struct kvm *kvm, int id)
 {
-	/* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
-	 * the caller has read kvm->online_vcpus before (as is the case
-	 * for kvm_for_each_vcpu, for example).
+	return kvm->vcpus[id / KVM_VCPUS_CHUNK_SIZE][id % KVM_VCPUS_CHUNK_SIZE];
+}
+
+static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int id)
+{
+	if (id >= atomic_read(&kvm->online_vcpus))
+		return NULL;
+
+	/*
+	 * Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu.  Ensures that the
+	 * pointers leading to an online vcpu are valid.
 	 */
 	smp_rmb();
-	return kvm->vcpus[i];
+
+	return __kvm_get_vcpu(kvm, id);
 }
 
 #define kvm_for_each_vcpu(vcpup, kvm) \
@@ -514,7 +527,7 @@ static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
 
 	if (id < 0)
 		return NULL;
-	if (id < KVM_MAX_VCPUS)
+	if (id < atomic_read(&kvm->online_vcpus))
 		vcpu = kvm_get_vcpu(kvm, id);
 	if (vcpu && vcpu->vcpu_id == id)
 		return vcpu;
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 6cec58cad6c7..f9d68ec332c6 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -759,11 +759,14 @@ void kvm_free_vcpus(struct kvm *kvm)
 
 	mutex_lock(&kvm->lock);
 
-	i = atomic_read(&kvm->online_vcpus);
+	i = round_up(atomic_read(&kvm->online_vcpus), KVM_VCPUS_CHUNK_SIZE) /
+		KVM_VCPUS_CHUNK_SIZE;
 	atomic_set(&kvm->online_vcpus, 0);
 
-	while (i--)
+	while (i--) {
+		kfree(kvm->vcpus[i]);
 		kvm->vcpus[i] = NULL;
+	}
 
 	mutex_unlock(&kvm->lock);
 }
@@ -2480,6 +2483,8 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
 {
 	int r;
 	struct kvm_vcpu *vcpu;
+	struct kvm_vcpu **vcpusp;
+	unsigned chunk, offset;
 
 	if (id >= KVM_MAX_VCPU_ID)
 		return -EINVAL;
@@ -2517,8 +2522,22 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
 
 	vcpu->vcpus_idx = atomic_read(&kvm->online_vcpus);
 
-	BUG_ON(kvm->vcpus[vcpu->vcpus_idx]);
+	chunk  = vcpu->vcpus_idx / KVM_VCPUS_CHUNK_SIZE;
+	offset = vcpu->vcpus_idx % KVM_VCPUS_CHUNK_SIZE;
 
+	if (!kvm->vcpus[chunk]) {
+		kvm->vcpus[chunk] = kzalloc(KVM_VCPUS_CHUNK_SIZE * sizeof(**kvm->vcpus),
+		                            GFP_KERNEL);
+		if (!kvm->vcpus[chunk]) {
+			r = -ENOMEM;
+			goto unlock_vcpu_destroy;
+		}
+
+		BUG_ON(offset != 0);
+	}
+
+	vcpusp = &kvm->vcpus[chunk][offset];
+	BUG_ON(*vcpusp);
 
 	/* Now it's all set up, let userspace reach it */
 	kvm_get_kvm(kvm);
@@ -2528,7 +2547,7 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
 		goto unlock_vcpu_destroy;
 	}
 
-	kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
+	*vcpusp = vcpu;
 	list_add_tail_rcu(&vcpu->vcpu_list, &kvm->vcpu_list);
 
 	/*
-- 
2.13.3