[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 7 Jul 2022 16:37:18 +0000
From: Sean Christopherson <seanjc@...gle.com>
To: Wang Guangju <wangguangju@...du.com>
Cc: pbonzini@...hat.com, vkuznets@...hat.com, jmattson@...gle.com,
wanpengli@...cent.com, bp@...en8.de, joro@...tes.org,
suravee.suthikulpanit@....com, hpa@...or.com, tglx@...utronix.de,
mingo@...hat.com, kvm@...r.kernel.org,
linux-kernel@...r.kernel.org, chukaiping <chukaiping@...du.com>
Subject: Re: [PATCH] KVM: x86: Add EOI_INDUCED exit handlers for Hyper-V
SynIC vectors
This isn't adding a handler so much as it's signaling EOI for SynIC vectors. Maybe?
KVM: x86: Send EOI to SynIC vectors on accelerated EOI-induced VM-Exits
On Thu, Jul 07, 2022, Wang Guangju wrote:
> From: chukaiping <chukaiping@...du.com>
>
> When EOI virtualization is performed on VMX,
> kvm_apic_set_eoi_accelerated() is called upon
> EXIT_REASON_EOI_INDUCED but unlike its non-accelerated
> apic_set_eoi() sibling, Hyper-V SINT vectors are
> left unhandled.
Wrap changelogs closer to ~75 chars.
> This patch fix it, and add a new helper function to
> handle both IOAPIC and Hyper-V SINT vectors.
Avoid "this patch" and simply state what change is being made. E.g.
Send EOI to Hyper-V SINT vectors when handling acclerated EOI-induced
VM-Exits. KVM Hyper-V needs to handle the SINT EOI irrespective of
whether the EOI is acclerated or not.
Fixes: 5c919412fe61 ("kvm/x86: Hyper-V synthetic interrupt controller")
and probably Cc: stable@...r.kernel.org?
> Suggested-by: Vitaly Kuznetsov <vkuznets@...hat.com>
> Signed-off-by: wangguangju <wangguangju@...du.com>
> ---
> arch/x86/kvm/lapic.c | 19 ++++++++++++-------
> 1 file changed, 12 insertions(+), 7 deletions(-)
>
> diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> index f03facc..e046afe 100644
> --- a/arch/x86/kvm/lapic.c
> +++ b/arch/x86/kvm/lapic.c
> @@ -1269,6 +1269,16 @@ static void kvm_ioapic_send_eoi(struct kvm_lapic *apic, int vector)
> kvm_ioapic_update_eoi(apic->vcpu, vector, trigger_mode);
> }
>
> +static inline void apic_set_eoi_vector(struct kvm_lapic *apic, int vector)
> +{
> + if (to_hv_vcpu(apic->vcpu) &&
> + test_bit(vector, to_hv_synic(apic->vcpu)->vec_bitmap))
> + kvm_hv_synic_send_eoi(apic->vcpu, vector);
> +
> + kvm_ioapic_send_eoi(apic, vector);
> + kvm_make_request(KVM_REQ_EVENT, apic->vcpu);
> +}
Rather than add a third helper, what about renaming kvm_apic_set_eoi_accelerated()
and having the non-accelerated helper call the "acclerated" version? That will
document the delta between the non-accelerated patch and the accelerated path.
The only hiccup is tracing, but that's easy to resolve (or we could just not trace
if there's no valid vector to EOI), e.g.
/*
* Send EOI for a valid vector. The caller, or hardware when this is invoked
* after an accelerated EOI VM-Exit, is responsible for updating the vISR and
* vPPR.
*/
void kvm_apic_set_eoi(struct kvm_lapic *apic, int vector)
{
trace_kvm_eoi(apic, vector);
if (to_hv_vcpu(apic->vcpu) &&
test_bit(vector, to_hv_synic(apic->vcpu)->vec_bitmap))
kvm_hv_synic_send_eoi(apic->vcpu, vector);
kvm_ioapic_send_eoi(apic, vector);
kvm_make_request(KVM_REQ_EVENT, apic->vcpu);
}
EXPORT_SYMBOL_GPL(kvm_apic_set_eoi);
static int apic_set_eoi(struct kvm_lapic *apic)
{
int vector = apic_find_highest_isr(apic);
/*
* Not every write EOI will has corresponding ISR,
* one example is when Kernel check timer on setup_IO_APIC
*/
if (vector == -1) {
trace_kvm_eoi(apic, vector); <---- maybe just drop this?
return vector;
}
apic_clear_isr(vector, apic);
apic_update_ppr(apic);
kvm_apic_set_eoi(apic, vector);
return vector;
}
Powered by blists - more mailing lists