* [PATCH] kvm: x86: remove vmx_vm_has_apicv() outside of hwapic_isr_update()
@ 2014-12-01 9:28 Tiejun Chen
2014-12-01 11:43 ` Paolo Bonzini
0 siblings, 1 reply; 6+ messages in thread
From: Tiejun Chen @ 2014-12-01 9:28 UTC (permalink / raw)
To: pbonzini; +Cc: kvm
In most cases calling hwapic_isr_update(), actually we always
check if kvm_apic_vid_enabled() == 1, and also actually,
kvm_apic_vid_enabled()
-> kvm_x86_ops->vm_has_apicv()
-> vmx_vm_has_apicv() or '0' in svm case
So its unnecessary to recall this inside hwapic_isr_update(), here
just remove vmx_vm_has_apicv() out and follow others.
Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
---
arch/x86/kvm/lapic.c | 3 ++-
arch/x86/kvm/vmx.c | 3 ---
2 files changed, 2 insertions(+), 4 deletions(-)
diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index e0e5642..2ddc426 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -1739,7 +1739,8 @@ void kvm_apic_post_state_restore(struct kvm_vcpu *vcpu,
if (kvm_x86_ops->hwapic_irr_update)
kvm_x86_ops->hwapic_irr_update(vcpu,
apic_find_highest_irr(apic));
- kvm_x86_ops->hwapic_isr_update(vcpu->kvm, apic_find_highest_isr(apic));
+ if (kvm_apic_vid_enabled(vcpu->kvm))
+ kvm_x86_ops->hwapic_isr_update(vcpu->kvm, apic_find_highest_isr(apic));
kvm_make_request(KVM_REQ_EVENT, vcpu);
kvm_rtc_eoi_tracking_restore_one(vcpu);
}
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 6a951d8..f0c16a9 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -7406,9 +7406,6 @@ static void vmx_hwapic_isr_update(struct kvm *kvm, int isr)
u16 status;
u8 old;
- if (!vmx_vm_has_apicv(kvm))
- return;
-
if (isr == -1)
isr = 0;
--
1.9.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH] kvm: x86: remove vmx_vm_has_apicv() outside of hwapic_isr_update()
2014-12-01 9:28 [PATCH] kvm: x86: remove vmx_vm_has_apicv() outside of hwapic_isr_update() Tiejun Chen
@ 2014-12-01 11:43 ` Paolo Bonzini
2014-12-19 2:32 ` Chen, Tiejun
0 siblings, 1 reply; 6+ messages in thread
From: Paolo Bonzini @ 2014-12-01 11:43 UTC (permalink / raw)
To: Tiejun Chen; +Cc: kvm
On 01/12/2014 10:28, Tiejun Chen wrote:
> In most cases calling hwapic_isr_update(), actually we always
> check if kvm_apic_vid_enabled() == 1, and also actually,
> kvm_apic_vid_enabled()
> -> kvm_x86_ops->vm_has_apicv()
> -> vmx_vm_has_apicv() or '0' in svm case
>
> So its unnecessary to recall this inside hwapic_isr_update(), here
> just remove vmx_vm_has_apicv() out and follow others.
If you want to do this, please NULL out the function pointer instead, as
KVM already does for hwapic_irr_update.
Paolo
> Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
> ---
> arch/x86/kvm/lapic.c | 3 ++-
> arch/x86/kvm/vmx.c | 3 ---
> 2 files changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> index e0e5642..2ddc426 100644
> --- a/arch/x86/kvm/lapic.c
> +++ b/arch/x86/kvm/lapic.c
> @@ -1739,7 +1739,8 @@ void kvm_apic_post_state_restore(struct kvm_vcpu *vcpu,
> if (kvm_x86_ops->hwapic_irr_update)
> kvm_x86_ops->hwapic_irr_update(vcpu,
> apic_find_highest_irr(apic));
> - kvm_x86_ops->hwapic_isr_update(vcpu->kvm, apic_find_highest_isr(apic));
> + if (kvm_apic_vid_enabled(vcpu->kvm))
> + kvm_x86_ops->hwapic_isr_update(vcpu->kvm, apic_find_highest_isr(apic));
> kvm_make_request(KVM_REQ_EVENT, vcpu);
> kvm_rtc_eoi_tracking_restore_one(vcpu);
> }
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 6a951d8..f0c16a9 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -7406,9 +7406,6 @@ static void vmx_hwapic_isr_update(struct kvm *kvm, int isr)
> u16 status;
> u8 old;
>
> - if (!vmx_vm_has_apicv(kvm))
> - return;
> -
> if (isr == -1)
> isr = 0;
>
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] kvm: x86: remove vmx_vm_has_apicv() outside of hwapic_isr_update()
2014-12-01 11:43 ` Paolo Bonzini
@ 2014-12-19 2:32 ` Chen, Tiejun
2014-12-19 11:12 ` Paolo Bonzini
0 siblings, 1 reply; 6+ messages in thread
From: Chen, Tiejun @ 2014-12-19 2:32 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: kvm
On 2014/12/1 19:43, Paolo Bonzini wrote:
>
>
> On 01/12/2014 10:28, Tiejun Chen wrote:
>> In most cases calling hwapic_isr_update(), actually we always
>> check if kvm_apic_vid_enabled() == 1, and also actually,
>> kvm_apic_vid_enabled()
>> -> kvm_x86_ops->vm_has_apicv()
>> -> vmx_vm_has_apicv() or '0' in svm case
>>
>> So its unnecessary to recall this inside hwapic_isr_update(), here
>> just remove vmx_vm_has_apicv() out and follow others.
>
> If you want to do this, please NULL out the function pointer instead, as
> KVM already does for hwapic_irr_update.
Are you saying something below?
if (enable_apicv)
...
else {
kvm_x86_ops->hwapic_irr_update = NULL;
But there's a little bit difference to NULL out hwapic_isr_update(),
static int vmx_vm_has_apicv(struct kvm *kvm)
{
return enable_apicv && irqchip_in_kernel(kvm);
}
Yes, I can do something like this,
static __init int hadware_setup(void)
{
...
if (enable_apicv) {
...
if (!irqchip_in_kernel(kvm))
kvm_x86_ops->hwapic_isr_update = NULL;
} else {
...
kvm_x86_ops->hwapic_isr_update = NULL;
But this means we have to revise hadware_setup() to get 'kvm' inside,
then rebase other callers to hwapic_isr_update(), is it really good?
Here what I will intend to do is trying to reduce some cost (reduplicate
check) with a little code, so its may not be worth changing much more.
Tiejun
>
> Paolo
>
>> Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
>> ---
>> arch/x86/kvm/lapic.c | 3 ++-
>> arch/x86/kvm/vmx.c | 3 ---
>> 2 files changed, 2 insertions(+), 4 deletions(-)
>>
>> diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
>> index e0e5642..2ddc426 100644
>> --- a/arch/x86/kvm/lapic.c
>> +++ b/arch/x86/kvm/lapic.c
>> @@ -1739,7 +1739,8 @@ void kvm_apic_post_state_restore(struct kvm_vcpu *vcpu,
>> if (kvm_x86_ops->hwapic_irr_update)
>> kvm_x86_ops->hwapic_irr_update(vcpu,
>> apic_find_highest_irr(apic));
>> - kvm_x86_ops->hwapic_isr_update(vcpu->kvm, apic_find_highest_isr(apic));
>> + if (kvm_apic_vid_enabled(vcpu->kvm))
>> + kvm_x86_ops->hwapic_isr_update(vcpu->kvm, apic_find_highest_isr(apic));
>> kvm_make_request(KVM_REQ_EVENT, vcpu);
>> kvm_rtc_eoi_tracking_restore_one(vcpu);
>> }
>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>> index 6a951d8..f0c16a9 100644
>> --- a/arch/x86/kvm/vmx.c
>> +++ b/arch/x86/kvm/vmx.c
>> @@ -7406,9 +7406,6 @@ static void vmx_hwapic_isr_update(struct kvm *kvm, int isr)
>> u16 status;
>> u8 old;
>>
>> - if (!vmx_vm_has_apicv(kvm))
>> - return;
>> -
>> if (isr == -1)
>> isr = 0;
>>
>>
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] kvm: x86: remove vmx_vm_has_apicv() outside of hwapic_isr_update()
2014-12-19 2:32 ` Chen, Tiejun
@ 2014-12-19 11:12 ` Paolo Bonzini
2014-12-22 9:01 ` Chen, Tiejun
0 siblings, 1 reply; 6+ messages in thread
From: Paolo Bonzini @ 2014-12-19 11:12 UTC (permalink / raw)
To: Chen, Tiejun; +Cc: kvm
On 19/12/2014 03:32, Chen, Tiejun wrote:
>
> Are you saying something below?
>
> if (enable_apicv)
> ...
> else {
> kvm_x86_ops->hwapic_irr_update = NULL;
Yes.
> But this means we have to revise hadware_setup() to get 'kvm' inside,
This would not even be possible, since hardware_setup() is only called once.
However, for the only caller of hwapic_isr_update (but presumably all of
them, as is the case for hwapic_irr_update), you already know that
irqchip_in_kernel(kvm) is true. You are in kvm_apic_post_state_restore,
which takes a kvm_lapic_state, and no lapic exists if
!irqchip_in_kernel(kvm).
(Yes, irqchip_in_kernel) is a bit weird and tests pic_irqchip(kvm)
instead, but it's the same. It tests pic_irqchip(kvm) only because the
LAPIC is per-cpu and irqchip_in_kernel takes a struct kvm).
So it's possible to NULL out hwapic_isr_update in hardware_setup. It
simply shouldn't happen that you call hwapic_isr_update without the
in-kernel irqchip. The kernel knows nothing about ISR/IRR without the
in-kernel irqchip.
Paolo
> then rebase other callers to hwapic_isr_update(), is it really good?
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] kvm: x86: remove vmx_vm_has_apicv() outside of hwapic_isr_update()
2014-12-19 11:12 ` Paolo Bonzini
@ 2014-12-22 9:01 ` Chen, Tiejun
2014-12-22 9:33 ` Paolo Bonzini
0 siblings, 1 reply; 6+ messages in thread
From: Chen, Tiejun @ 2014-12-22 9:01 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: kvm
On 2014/12/19 19:12, Paolo Bonzini wrote:
>
>
> On 19/12/2014 03:32, Chen, Tiejun wrote:
>>
>> Are you saying something below?
>>
>> if (enable_apicv)
>> ...
>> else {
>> kvm_x86_ops->hwapic_irr_update = NULL;
>
> Yes.
>
>> But this means we have to revise hadware_setup() to get 'kvm' inside,
>
> This would not even be possible, since hardware_setup() is only called once.
Yeah.
>
> However, for the only caller of hwapic_isr_update (but presumably all of
> them, as is the case for hwapic_irr_update), you already know that
There are two other callers to hwapic_isr_update(), apic_set_isr() and
apic_clear_isr(), but I think they still work here.
> irqchip_in_kernel(kvm) is true. You are in kvm_apic_post_state_restore,
> which takes a kvm_lapic_state, and no lapic exists if
> !irqchip_in_kernel(kvm).
>
> (Yes, irqchip_in_kernel) is a bit weird and tests pic_irqchip(kvm)
> instead, but it's the same. It tests pic_irqchip(kvm) only because the
> LAPIC is per-cpu and irqchip_in_kernel takes a struct kvm).
>
> So it's possible to NULL out hwapic_isr_update in hardware_setup. It
> simply shouldn't happen that you call hwapic_isr_update without the
> in-kernel irqchip. The kernel knows nothing about ISR/IRR without the
> in-kernel irqchip.
Thanks for your kind elaboration which always benefits me.
What about this revision as follows?
kvm: x86: vmx: NULL out hwapic_isr_update() in case of !enable_apicv
In most cases calling hwapic_isr_update(), we always check if
kvm_apic_vid_enabled() == 1, but actually,
kvm_apic_vid_enabled()
-> kvm_x86_ops->vm_has_apicv()
-> vmx_vm_has_apicv() or '0' in svm case
-> return enable_apicv && irqchip_in_kernel(kvm)
So its a little cost to recall vmx_vm_has_apicv() inside
hwapic_isr_update(), here just NULL out hwapic_isr_update() in
case of !enable_apicv inside hardware_setup() then make all
related stuffs follow this. Note we don't check this under that
condition of irqchip_in_kernel() since we should make sure
definitely any caller don't work without in-kernel irqchip.
Signed-off-by: Tiejun Chen <tiejun.chen@intel.com>
---
arch/x86/kvm/lapic.c | 7 ++++---
arch/x86/kvm/vmx.c | 4 +---
2 files changed, 5 insertions(+), 6 deletions(-)
diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index 4f0c0b9..eb4cd5e 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -402,7 +402,7 @@ static inline void apic_set_isr(int vec, struct
kvm_lapic *apic)
* because the processor can modify ISR under the hood. Instead
* just set SVI.
*/
- if (unlikely(kvm_apic_vid_enabled(vcpu->kvm)))
+ if (unlikely(kvm_x86_ops->hwapic_isr_update))
kvm_x86_ops->hwapic_isr_update(vcpu->kvm, vec);
else {
++apic->isr_count;
@@ -450,7 +450,7 @@ static inline void apic_clear_isr(int vec, struct
kvm_lapic *apic)
* on the other hand isr_count and highest_isr_cache are unused
* and must be left alone.
*/
- if (unlikely(kvm_apic_vid_enabled(vcpu->kvm)))
+ if (unlikely(kvm_x86_ops->hwapic_isr_update))
kvm_x86_ops->hwapic_isr_update(vcpu->kvm,
apic_find_highest_isr(apic));
else {
@@ -1742,7 +1742,8 @@ void kvm_apic_post_state_restore(struct kvm_vcpu
*vcpu,
if (kvm_x86_ops->hwapic_irr_update)
kvm_x86_ops->hwapic_irr_update(vcpu,
apic_find_highest_irr(apic));
- kvm_x86_ops->hwapic_isr_update(vcpu->kvm,
apic_find_highest_isr(apic));
+ if (kvm_x86_ops->hwapic_isr_update)
+ kvm_x86_ops->hwapic_isr_update(vcpu->kvm,
apic_find_highest_isr(apic));
kvm_make_request(KVM_REQ_EVENT, vcpu);
kvm_rtc_eoi_tracking_restore_one(vcpu);
}
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 96c84a8..e378dff 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -5895,6 +5895,7 @@ static __init int hardware_setup(void)
kvm_x86_ops->update_cr8_intercept = NULL;
else {
kvm_x86_ops->hwapic_irr_update = NULL;
+ kvm_x86_ops->hwapic_isr_update = NULL;
kvm_x86_ops->deliver_posted_interrupt = NULL;
kvm_x86_ops->sync_pir_to_irr = vmx_sync_pir_to_irr_dummy;
}
@@ -7471,9 +7472,6 @@ static void vmx_hwapic_isr_update(struct kvm *kvm,
int isr)
u16 status;
u8 old;
- if (!vmx_vm_has_apicv(kvm))
- return;
-
if (isr == -1)
isr = 0;
--
1.9.1
I can send out as a patch if we have on any objections.
Thanks
Tiejun
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH] kvm: x86: remove vmx_vm_has_apicv() outside of hwapic_isr_update()
2014-12-22 9:01 ` Chen, Tiejun
@ 2014-12-22 9:33 ` Paolo Bonzini
0 siblings, 0 replies; 6+ messages in thread
From: Paolo Bonzini @ 2014-12-22 9:33 UTC (permalink / raw)
To: Chen, Tiejun; +Cc: kvm
On 22/12/2014 10:01, Chen, Tiejun wrote:
> I can send out as a patch if we have on any objections.
No problem, I will apply it to kvm/queue.
Paolo
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2014-12-22 9:33 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-12-01 9:28 [PATCH] kvm: x86: remove vmx_vm_has_apicv() outside of hwapic_isr_update() Tiejun Chen
2014-12-01 11:43 ` Paolo Bonzini
2014-12-19 2:32 ` Chen, Tiejun
2014-12-19 11:12 ` Paolo Bonzini
2014-12-22 9:01 ` Chen, Tiejun
2014-12-22 9:33 ` Paolo Bonzini
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.