linux-kernel - [PATCH v2 07/10] KVM: selftests: Skip ICR.BUSY test in xapic_state

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20240719235107.3023592-8-seanjc@google.com>
Date: Fri, 19 Jul 2024 16:51:04 -0700
From: Sean Christopherson <seanjc@...gle.com>
To: Sean Christopherson <seanjc@...gle.com>, Paolo Bonzini <pbonzini@...hat.com>
Cc: kvm@...r.kernel.org, linux-kernel@...r.kernel.org, 
	Maxim Levitsky <mlevitsk@...hat.com>, Suravee Suthikulpanit <suravee.suthikulpanit@....com>
Subject: [PATCH v2 07/10] KVM: selftests: Skip ICR.BUSY test in
 xapic_state_test if x2APIC is enabled

Don't test the ICR BUSY bit when x2APIC is enabled as AMD and Intel have
different behavior (AMD #GPs, Intel ignores), and the fact that the CPU
performs the reserved bit checks when IPI virtualization is enabled makes
it impossible for KVM to precisely emulate one or the other.

Signed-off-by: Sean Christopherson <seanjc@...gle.com>
---
 .../selftests/kvm/x86_64/xapic_state_test.c    | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)

diff --git a/tools/testing/selftests/kvm/x86_64/xapic_state_test.c b/tools/testing/selftests/kvm/x86_64/xapic_state_test.c
index 69849acd95b0..928d65948c48 100644
--- a/tools/testing/selftests/kvm/x86_64/xapic_state_test.c
+++ b/tools/testing/selftests/kvm/x86_64/xapic_state_test.c
@@ -70,12 +70,10 @@ static void ____test_icr(struct xapic_vcpu *x, uint64_t val)
 	vcpu_ioctl(vcpu, KVM_GET_LAPIC, &xapic);
 	icr = (u64)(*((u32 *)&xapic.regs[APIC_ICR])) |
 	      (u64)(*((u32 *)&xapic.regs[APIC_ICR2])) << 32;
-	if (!x->is_x2apic) {
+	if (!x->is_x2apic)
 		val &= (-1u | (0xffull << (32 + 24)));
-		TEST_ASSERT_EQ(icr, val & ~APIC_ICR_BUSY);
-	} else {
-		TEST_ASSERT_EQ(icr & ~APIC_ICR_BUSY, val & ~APIC_ICR_BUSY);
-	}
+
+	TEST_ASSERT_EQ(icr, val & ~APIC_ICR_BUSY);
 }
 
 #define X2APIC_RSVED_BITS_MASK  (GENMASK_ULL(31,20) | \
@@ -91,7 +89,15 @@ static void __test_icr(struct xapic_vcpu *x, uint64_t val)
 		 */
 		val &= ~X2APIC_RSVED_BITS_MASK;
 	}
-	____test_icr(x, val | APIC_ICR_BUSY);
+
+	/*
+	 * The BUSY bit is reserved on both AMD and Intel, but only AMD treats
+	 * it is as _must_ be zero.  Intel simply ignores the bit.  Don't test
+	 * the BUSY bit for x2APIC, as there is no single correct behavior.
+	 */
+	if (!x->is_x2apic)
+		____test_icr(x, val | APIC_ICR_BUSY);
+
 	____test_icr(x, val & ~(u64)APIC_ICR_BUSY);
 }
 
-- 
2.45.2.1089.g2a221341d9-goog