On 12/2/22 16:49, Sean Christopherson wrote:
On Fri, Dec 02, 2022, Yuan ZhaoXiong wrote:
This patch fixes that VM rebooting itself will cause APICv
disabled when VM is started with APICv/x2AVIC enabled.
When a VM reboot itself, The Qemu whill reset LAPIC by invoking
ioctl(KVM_SET_LAPIC, ...) to disable x2APIC mode and set APIC_ID
to its vcpuid in xAPIC mode.
That will be handled in KVM as follows:
kvm_vcpu_ioctl_set_lapic
kvm_apic_set_state
kvm_lapic_set_base => disable X2APIC mode
kvm_apic_state_fixup
kvm_lapic_xapic_id_updated
kvm_xapic_id(apic) != apic->vcpu->vcpu_id
kvm_set_apicv_inhibit(APICV_INHIBIT_REASON_APIC_ID_MODIFIED)
memcpy(vcpu->arch.apic->regs, s->regs, sizeof(*s)) => update APIC_ID
kvm_apic_set_state invokes kvm_lapic_set_base to disable x2APIC mode
firstly, but don't change APIC_ID, APIC_ID is 32 bits in x2APIC mode
and 8 bist(bit 24 ~ bit 31) in xAPIC mode. So kvm_lapic_xapic_id_updated
will set APICV_INHIBIT_REASON_APIC_ID_MODIFIED bit inhibit and disable
APICv/x2AVIC.
kvm_lapic_xapic_id_updated must be called after APIC_ID is changed.
Fixes: 3743c2f02517 ("KVM: x86: inhibit APICv/AVIC on changes to APIC ID or APIC base")
Signed-off-by: Yuan ZhaoXiong <yuanzhaoxiong@xxxxxxxxx>
---
arch/x86/kvm/lapic.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index d7639d1..bf5ce86 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -2722,8 +2722,6 @@ static int kvm_apic_state_fixup(struct kvm_vcpu *vcpu,
icr = __kvm_lapic_get_reg64(s->regs, APIC_ICR);
__kvm_lapic_set_reg(s->regs, APIC_ICR2, icr >> 32);
}
- } else {
- kvm_lapic_xapic_id_updated(vcpu->arch.apic);
}
return 0;
@@ -2759,6 +2757,9 @@ int kvm_apic_set_state(struct kvm_vcpu *vcpu, struct kvm_lapic_state *s)
}
memcpy(vcpu->arch.apic->regs, s->regs, sizeof(*s));
+ if (!apic_x2apic_mode(apic))
+ kvm_lapic_xapic_id_updated(apic);
+
Already posted[*], along with a pile of other APIC fixes. Hopefully it will land
in 6.2.
I'll let this patch overtake the others since it is stable@ material.
I'll also volunteer someone to write a testcase (we can add an
apicv_enabled boolean statistic too for the sake of the test).
Paolo
[*] https://lore.kernel.org/all/20221001005915.2041642-7-seanjc@xxxxxxxxxx
