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;
From:   Maxim Levitsky <mlevitsk@redhat.com>
To:     kvm@vger.kernel.org
Cc:     linux-kernel@vger.kernel.org, x86@kernel.org,
        Borislav Petkov <bp@alien8.de>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Sean Christopherson <seanjc@google.com>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        "H. Peter Anvin" <hpa@zytor.com>, Ingo Molnar <mingo@redhat.com>,
        Maxim Levitsky <mlevitsk@redhat.com>
Subject: [PATCH v2 1/3] KVM: x86: VMX: __kvm_apic_update_irr must update the IRR atomically
Date:   Wed, 26 Jul 2023 16:59:43 +0300
Message-Id: <20230726135945.260841-2-mlevitsk@redhat.com>
In-Reply-To: <20230726135945.260841-1-mlevitsk@redhat.com>
References: <20230726135945.260841-1-mlevitsk@redhat.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

If APICv is inhibited, then IPIs from peer vCPUs are done by
atomically setting bits in IRR.

This means, that when __kvm_apic_update_irr copies PIR to IRR,
it has to modify IRR atomically as well.

Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
---
 arch/x86/kvm/lapic.c | 20 +++++++++++++-------
 1 file changed, 13 insertions(+), 7 deletions(-)

diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index 113ca9661ab21d..b17b37e4d4fcd1 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -637,16 +637,22 @@ bool __kvm_apic_update_irr(u32 *pir, void *regs, int *max_irr)
 	*max_irr = -1;
 
 	for (i = vec = 0; i <= 7; i++, vec += 32) {
+		u32 *p_irr = (u32 *)(regs + APIC_IRR + i * 0x10);
+
+		irr_val = *p_irr;
 		pir_val = READ_ONCE(pir[i]);
-		irr_val = *((u32 *)(regs + APIC_IRR + i * 0x10));
+
 		if (pir_val) {
+			pir_val = xchg(&pir[i], 0);
+
 			prev_irr_val = irr_val;
-			irr_val |= xchg(&pir[i], 0);
-			*((u32 *)(regs + APIC_IRR + i * 0x10)) = irr_val;
-			if (prev_irr_val != irr_val) {
-				max_updated_irr =
-					__fls(irr_val ^ prev_irr_val) + vec;
-			}
+			do {
+				irr_val = prev_irr_val | pir_val;
+			} while (prev_irr_val != irr_val &&
+				 !try_cmpxchg(p_irr, &prev_irr_val, irr_val));
+
+			if (prev_irr_val != irr_val)
+				max_updated_irr = __fls(irr_val ^ prev_irr_val) + vec;
 		}
 		if (irr_val)
 			*max_irr = __fls(irr_val) + vec;
-- 
2.26.3