linux-kernel - [PATCH] x86/kernel/cpu/amd.c: don't apply the lahf

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 for Android: free password hash cracker in your pocket

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

Message-ID: <alpine.LRH.2.02.1707120318530.4739@file01.intranet.prod.int.rdu2.redhat.com>
Date:   Wed, 12 Jul 2017 03:19:56 -0400 (EDT)
From:   Mikulas Patocka <mpatocka@...hat.com>
To:     Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>,
        "H. Peter Anvin" <hpa@...or.com>
cc:     x86@...nel.org, linux-kernel@...r.kernel.org
Subject: [PATCH] x86/kernel/cpu/amd.c: don't apply the lahf_lm workaround in
 a hypervisor

When I start a virtual machine, kvm complains about accesses to an unknown
MSR ("vcpu0 unhandled rdmsr: 0xc001100d"). This is caused by the code in
init_amd_k8 that tries to work around a BIOS bug that incorrectly sets
lahf_lm.

This patch disables this workaround when running in a hypervisor.

Note that the processors for which this erratum applies do not support
virtualization at all, so if we are running in a hypervisor, we can be
sure that the workaronud doesn't apply.

Signed-off-by: Mikulas Patocka <mpatocka@...hat.com>

---
 arch/x86/kernel/cpu/amd.c |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Index: linux-2.6/arch/x86/kernel/cpu/amd.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/cpu/amd.c
+++ linux-2.6/arch/x86/kernel/cpu/amd.c
@@ -627,8 +627,12 @@ static void init_amd_k8(struct cpuinfo_x
 	 * Some BIOSes incorrectly force this feature, but only K8 revision D
 	 * (model = 0x14) and later actually support it.
 	 * (AMD Erratum #110, docId: 25759).
+	 *
+	 * Don't apply this workaround in a virtual machine because kvm
+	 * complains about reads from an unknown MSR.
 	 */
-	if (c->x86_model < 0x14 && cpu_has(c, X86_FEATURE_LAHF_LM)) {
+	if (c->x86_model < 0x14 && cpu_has(c, X86_FEATURE_LAHF_LM) &&
+	    !boot_cpu_has(X86_FEATURE_HYPERVISOR)) {
 		clear_cpu_cap(c, X86_FEATURE_LAHF_LM);
 		if (!rdmsrl_amd_safe(0xc001100d, &value)) {
 			value &= ~BIT_64(32);