In vcpu_load_eoi_exitmap(), currently the eoi_exit_bitmap[4] array is
initialized only when Hyper-V context is available, in other path it is
just passed to kvm_x86_ops.load_eoi_exitmap() directly from on the stack,
which would cause unexpected interrupt delivery/handling issues, e.g. an
*old* linux kernel that relies on PIT to do clock calibration on KVM might
randomly fail to boot.
Fix it by passing ioapic_handled_vectors to load_eoi_exitmap() when Hyper-V
context is not available.
Signed-off-by: Huang Le <[email protected]>
---
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index dc7eb5fddfd3..0699832504c9 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -9547,11 +9547,14 @@ static void vcpu_load_eoi_exitmap(struct kvm_vcpu *vcpu)
if (!kvm_apic_hw_enabled(vcpu->arch.apic))
return;
- if (to_hv_vcpu(vcpu))
- bitmap_or((ulong *)eoi_exit_bitmap,
- vcpu->arch.ioapic_handled_vectors,
- to_hv_synic(vcpu)->vec_bitmap, 256);
+ if (!to_hv_vcpu(vcpu)) {
+ static_call(kvm_x86_load_eoi_exitmap)(
+ vcpu, (u64 *)vcpu->arch.ioapic_handled_vectors);
+ return;
+ }
+ bitmap_or((ulong *)eoi_exit_bitmap, vcpu->arch.ioapic_handled_vectors,
+ to_hv_synic(vcpu)->vec_bitmap, 256);
static_call(kvm_x86_load_eoi_exitmap)(vcpu, eoi_exit_bitmap);
}
黄乐 <[email protected]> writes:
> In vcpu_load_eoi_exitmap(), currently the eoi_exit_bitmap[4] array is
> initialized only when Hyper-V context is available, in other path it is
> just passed to kvm_x86_ops.load_eoi_exitmap() directly from on the stack,
> which would cause unexpected interrupt delivery/handling issues, e.g. an
> *old* linux kernel that relies on PIT to do clock calibration on KVM might
> randomly fail to boot.
>
> Fix it by passing ioapic_handled_vectors to load_eoi_exitmap() when Hyper-V
> context is not available.
>
> Signed-off-by: Huang Le <[email protected]>
Fixes: f2bc14b69c38 ("KVM: x86: hyper-v: Prepare to meet unallocated Hyper-V context")
Cc: [email protected]
> ---
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index dc7eb5fddfd3..0699832504c9 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -9547,11 +9547,14 @@ static void vcpu_load_eoi_exitmap(struct kvm_vcpu *vcpu)
> if (!kvm_apic_hw_enabled(vcpu->arch.apic))
> return;
>
> - if (to_hv_vcpu(vcpu))
> - bitmap_or((ulong *)eoi_exit_bitmap,
> - vcpu->arch.ioapic_handled_vectors,
> - to_hv_synic(vcpu)->vec_bitmap, 256);
> + if (!to_hv_vcpu(vcpu)) {
> + static_call(kvm_x86_load_eoi_exitmap)(
> + vcpu, (u64 *)vcpu->arch.ioapic_handled_vectors);
> + return;
> + }
>
> + bitmap_or((ulong *)eoi_exit_bitmap, vcpu->arch.ioapic_handled_vectors,
> + to_hv_synic(vcpu)->vec_bitmap, 256);
> static_call(kvm_x86_load_eoi_exitmap)(vcpu, eoi_exit_bitmap);
> }
>
Reviewed-by: Vitaly Kuznetsov <[email protected]>
My personal preference, however, would be to keep 'if
(to_hv_vcpu(vcpu))' check and not invert it, i.e.:
if (to_hv_vcpu(vcpu)) {
bitmap_or((ulong *)eoi_exit_bitmap,
vcpu->arch.ioapic_handled_vectors,
to_hv_synic(vcpu)->vec_bitmap, 256);
static_call(...)(vcpu, eoi_exit_bitmap)
return;
}
static_call(...)(vcpu, (u64 *)vcpu->arch.ioapic_handled_vectors);
to slightly reduce the code churn but it doesn't matter much.
Thanks!
--
Vitaly
> ???? <[email protected]> writes:
>
> > In vcpu_load_eoi_exitmap(), currently the eoi_exit_bitmap[4] array is
> > initialized only when Hyper-V context is available, in other path it is
> > just passed to kvm_x86_ops.load_eoi_exitmap() directly from on the stack,
> > which would cause unexpected interrupt delivery/handling issues, e.g. an
> > *old* linux kernel that relies on PIT to do clock calibration on KVM might
> > randomly fail to boot.
> >
> > Fix it by passing ioapic_handled_vectors to load_eoi_exitmap() when Hyper-V
> > context is not available.
> >
> > Signed-off-by: Huang Le <[email protected]>
>
> Fixes: f2bc14b69c38 ("KVM: x86: hyper-v: Prepare to meet unallocated Hyper-V context")
> Cc: [email protected]
Commit f2bc14b69c38 is not in stable tree I guess, it was merged in from 5.12,
do we still need Cc this patch to stable maintainers?
>
> > ---
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index dc7eb5fddfd3..0699832504c9 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -9547,11 +9547,14 @@ static void vcpu_load_eoi_exitmap(struct kvm_vcpu *vcpu)
> > if (!kvm_apic_hw_enabled(vcpu->arch.apic))
> > return;
> >
> > - if (to_hv_vcpu(vcpu))
> > - bitmap_or((ulong *)eoi_exit_bitmap,
> > - vcpu->arch.ioapic_handled_vectors,
> > - to_hv_synic(vcpu)->vec_bitmap, 256);
> > + if (!to_hv_vcpu(vcpu)) {
> > + static_call(kvm_x86_load_eoi_exitmap)(
> > + vcpu, (u64 *)vcpu->arch.ioapic_handled_vectors);
> > + return;
> > + }
> >
> > + bitmap_or((ulong *)eoi_exit_bitmap, vcpu->arch.ioapic_handled_vectors,
> > + to_hv_synic(vcpu)->vec_bitmap, 256);
> > static_call(kvm_x86_load_eoi_exitmap)(vcpu, eoi_exit_bitmap);
> > }
> >
>
> Reviewed-by: Vitaly Kuznetsov <[email protected]>
>
> My personal preference, however, would be to keep 'if
> (to_hv_vcpu(vcpu))' check and not invert it, i.e.:
>
> if (to_hv_vcpu(vcpu)) {
> bitmap_or((ulong *)eoi_exit_bitmap,
> vcpu->arch.ioapic_handled_vectors,
> to_hv_synic(vcpu)->vec_bitmap, 256);
> static_call(...)(vcpu, eoi_exit_bitmap)
> return;
> }
>
> static_call(...)(vcpu, (u64 *)vcpu->arch.ioapic_handled_vectors);
>
> to slightly reduce the code churn but it doesn't matter much.
Got it. Will send an updated one later. Thanks for suggestion!
>
> Thanks!
>
> --
> Vitaly
黄乐 <[email protected]> writes:
>> 黄乐 <[email protected]> writes:
>>
>> > In vcpu_load_eoi_exitmap(), currently the eoi_exit_bitmap[4] array is
>> > initialized only when Hyper-V context is available, in other path it is
>> > just passed to kvm_x86_ops.load_eoi_exitmap() directly from on the stack,
>> > which would cause unexpected interrupt delivery/handling issues, e.g. an
>> > *old* linux kernel that relies on PIT to do clock calibration on KVM might
>> > randomly fail to boot.
>> >
>> > Fix it by passing ioapic_handled_vectors to load_eoi_exitmap() when Hyper-V
>> > context is not available.
>> >
>> > Signed-off-by: Huang Le <[email protected]>
>>
>> Fixes: f2bc14b69c38 ("KVM: x86: hyper-v: Prepare to meet unallocated Hyper-V context")
>> Cc: [email protected]
>
> Commit f2bc14b69c38 is not in stable tree I guess, it was merged in from 5.12,
> do we still need Cc this patch to stable maintainers?
>
There are multiple stable trees, one for each major release. Not all of
them are still supported but you don't need to care about it, 'Cc:
[email protected]' is just an indication for everyone who has
f2bc14b69c38 in his tree (5.12+) that there's a fix available.
--
Vitaly