thebeanogamer / rpms / qemu-kvm

Forked from rpms/qemu-kvm 5 months ago
Clone

Blame SOURCES/kvm-i386-do-kvm_put_msr_feature_control-first-thing-when.patch

bf143f
From f96220d64a31a4a52b2d132a503048579946f982 Mon Sep 17 00:00:00 2001
bf143f
From: Vitaly Kuznetsov <vkuznets@redhat.com>
bf143f
Date: Thu, 18 Aug 2022 17:01:13 +0200
bf143f
Subject: [PATCH 3/3] i386: do kvm_put_msr_feature_control() first thing when
bf143f
 vCPU is reset
bf143f
bf143f
RH-Author: Miroslav Rezanina <mrezanin@redhat.com>
bf143f
RH-MergeRequest: 219: Synchronize qemu-6.2.0-20.el8.1 build from RHEL 8.7 to RHEL 8.8
bf143f
RH-Bugzilla: 2125271
bf143f
RH-Acked-by: Vitaly Kuznetsov <vkuznets@redhat.com>
bf143f
RH-Acked-by: Jon Maloy <jmaloy@redhat.com>
bf143f
RH-Commit: [2/2] 08e1e67db96801e4a35aa6b60a93b2c2f1641220
bf143f
bf143f
kvm_put_sregs2() fails to reset 'locked' CR4/CR0 bits upon vCPU reset when
bf143f
it is in VMX root operation. Do kvm_put_msr_feature_control() before
bf143f
kvm_put_sregs2() to (possibly) kick vCPU out of VMX root operation. It also
bf143f
seems logical to do kvm_put_msr_feature_control() before
bf143f
kvm_put_nested_state() and not after it, especially when 'real' nested
bf143f
state is set.
bf143f
bf143f
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
bf143f
Message-Id: <20220818150113.479917-3-vkuznets@redhat.com>
bf143f
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
bf143f
(cherry picked from commit 45ed68a1a3a19754ade954d75a3c9d13ff560e5c)
bf143f
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
bf143f
---
bf143f
 target/i386/kvm/kvm.c | 17 ++++++++++++-----
bf143f
 1 file changed, 12 insertions(+), 5 deletions(-)
bf143f
bf143f
diff --git a/target/i386/kvm/kvm.c b/target/i386/kvm/kvm.c
bf143f
index 81d729dc40..a06221d3e5 100644
bf143f
--- a/target/i386/kvm/kvm.c
bf143f
+++ b/target/i386/kvm/kvm.c
bf143f
@@ -4255,6 +4255,18 @@ int kvm_arch_put_registers(CPUState *cpu, int level)
bf143f
 
bf143f
     assert(cpu_is_stopped(cpu) || qemu_cpu_is_self(cpu));
bf143f
 
bf143f
+    /*
bf143f
+     * Put MSR_IA32_FEATURE_CONTROL first, this ensures the VM gets out of VMX
bf143f
+     * root operation upon vCPU reset. kvm_put_msr_feature_control() should also
bf143f
+     * preceed kvm_put_nested_state() when 'real' nested state is set.
bf143f
+     */
bf143f
+    if (level >= KVM_PUT_RESET_STATE) {
bf143f
+        ret = kvm_put_msr_feature_control(x86_cpu);
bf143f
+        if (ret < 0) {
bf143f
+            return ret;
bf143f
+        }
bf143f
+    }
bf143f
+
bf143f
     /* must be before kvm_put_nested_state so that EFER.SVME is set */
bf143f
     ret = kvm_put_sregs(x86_cpu);
bf143f
     if (ret < 0) {
bf143f
@@ -4266,11 +4278,6 @@ int kvm_arch_put_registers(CPUState *cpu, int level)
bf143f
         if (ret < 0) {
bf143f
             return ret;
bf143f
         }
bf143f
-
bf143f
-        ret = kvm_put_msr_feature_control(x86_cpu);
bf143f
-        if (ret < 0) {
bf143f
-            return ret;
bf143f
-        }
bf143f
     }
bf143f
 
bf143f
     if (level == KVM_PUT_FULL_STATE) {
bf143f
-- 
bf143f
2.35.3
bf143f