Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Reply-To: Sean Christopherson <seanjc@google.com>
Date:   Fri,  4 Feb 2022 21:42:00 +0000
In-Reply-To: <20220204214205.3306634-1-seanjc@google.com>
Message-Id: <20220204214205.3306634-7-seanjc@google.com>
Mime-Version: 1.0
References: <20220204214205.3306634-1-seanjc@google.com>
Subject: [PATCH 06/11] KVM: x86: WARN if KVM emulates an IPI without clearing
 the BUSY flag
From:   Sean Christopherson <seanjc@google.com>
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     Sean Christopherson <seanjc@google.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, Zeng Guang <guang.zeng@intel.com>,
        Chao Gao <chao.gao@intel.com>,
        Maxim Levitsky <mlevitsk@redhat.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

WARN if KVM emulates an IPI without clearing the BUSY flag, failure to do
so could hang the guest if it waits for the IPI be sent.

Opportunistically use APIC_ICR_BUSY macro instead of open coding the
magic number, and add a comment to clarify why kvm_recalculate_apic_map()
is unconditionally invoked (it's really, really confusing for IPIs due to
the existence of fast paths that don't trigger a potential recalc).

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 arch/x86/kvm/lapic.c | 10 +++++++++-
 arch/x86/kvm/x86.c   |  9 ++++-----
 2 files changed, 13 insertions(+), 6 deletions(-)

diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index 6e1f9e83eb68..4f57b6f5ebd4 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -1282,6 +1282,9 @@ void kvm_apic_send_ipi(struct kvm_lapic *apic, u32 icr_low, u32 icr_high)
 {
 	struct kvm_lapic_irq irq;
 
+	/* KVM has no delay and should always clear the BUSY/PENDING flag. */
+	WARN_ON_ONCE(icr_low & APIC_ICR_BUSY);
+
 	irq.vector = icr_low & APIC_VECTOR_MASK;
 	irq.delivery_mode = icr_low & APIC_MODE_MASK;
 	irq.dest_mode = icr_low & APIC_DEST_MASK;
@@ -2060,7 +2063,7 @@ int kvm_lapic_reg_write(struct kvm_lapic *apic, u32 reg, u32 val)
 	}
 	case APIC_ICR:
 		/* No delay here, so we always clear the pending bit */
-		val &= ~(1 << 12);
+		val &= ~APIC_ICR_BUSY;
 		kvm_apic_send_ipi(apic, val, kvm_lapic_get_reg(apic, APIC_ICR2));
 		kvm_lapic_set_reg(apic, APIC_ICR, val);
 		break;
@@ -2139,6 +2142,11 @@ int kvm_lapic_reg_write(struct kvm_lapic *apic, u32 reg, u32 val)
 		break;
 	}
 
+	/*
+	 * Recalculate APIC maps if necessary, e.g. if the software enable bit
+	 * was toggled, the APIC ID changed, etc...   The maps are marked dirty
+	 * on relevant changes, i.e. this is a nop for most writes.
+	 */
 	kvm_recalculate_apic_map(apic->vcpu->kvm);
 
 	return ret;
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index c25a6ef0ff06..9024e33c2add 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -2012,11 +2012,10 @@ static int handle_fastpath_set_x2apic_icr_irqoff(struct kvm_vcpu *vcpu, u64 data
 		return 1;
 
 	if (((data & APIC_SHORT_MASK) == APIC_DEST_NOSHORT) &&
-		((data & APIC_DEST_MASK) == APIC_DEST_PHYSICAL) &&
-		((data & APIC_MODE_MASK) == APIC_DM_FIXED) &&
-		((u32)(data >> 32) != X2APIC_BROADCAST)) {
-
-		data &= ~(1 << 12);
+	    ((data & APIC_DEST_MASK) == APIC_DEST_PHYSICAL) &&
+	    ((data & APIC_MODE_MASK) == APIC_DM_FIXED) &&
+	    ((u32)(data >> 32) != X2APIC_BROADCAST)) {
+		data &= ~APIC_ICR_BUSY;
 		kvm_apic_send_ipi(vcpu->arch.apic, (u32)data, (u32)(data >> 32));
 		kvm_lapic_set_reg(vcpu->arch.apic, APIC_ICR2, (u32)(data >> 32));
 		kvm_lapic_set_reg(vcpu->arch.apic, APIC_ICR, (u32)data);
-- 
2.35.0.263.gb82422642f-goog