KVM: VMX: Update shared MSRs to be saved/restored on MSR_EFER.LMA changes
authorLiran Alon <liran.alon@oracle.com>
Tue, 20 Nov 2018 16:03:25 +0000 (18:03 +0200)
committerPaolo Bonzini <pbonzini@redhat.com>
Tue, 27 Nov 2018 11:50:08 +0000 (12:50 +0100)
When guest transitions from/to long-mode by modifying MSR_EFER.LMA,
the list of shared MSRs to be saved/restored on guest<->host
transitions is updated (See vmx_set_efer() call to setup_msrs()).

On every entry to guest, vcpu_enter_guest() calls
vmx_prepare_switch_to_guest(). This function should also take care
of setting the shared MSRs to be saved/restored. However, the
function does nothing in case we are already running with loaded
guest state (vmx->loaded_cpu_state != NULL).

This means that even when guest modifies MSR_EFER.LMA which results
in updating the list of shared MSRs, it isn't being taken into account
by vmx_prepare_switch_to_guest() because it happens while we are
running with loaded guest state.

To fix above mentioned issue, add a flag to mark that the list of
shared MSRs has been updated and modify vmx_prepare_switch_to_guest()
to set shared MSRs when running with host state *OR* list of shared
MSRs has been updated.

Note that this issue was mistakenly introduced by commit
678e315e78a7 ("KVM: vmx: add dedicated utility to access guest's
kernel_gs_base") because previously vmx_set_efer() always called
vmx_load_host_state() which resulted in vmx_prepare_switch_to_guest() to
set shared MSRs.

Fixes: 678e315e78a7 ("KVM: vmx: add dedicated utility to access guest's kernel_gs_base")
Reported-by: Eyal Moscovici <eyal.moscovici@oracle.com>
Reviewed-by: Mihai Carabas <mihai.carabas@oracle.com>
Reviewed-by: Liam Merwick <liam.merwick@oracle.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/vmx.c

index 065f1df..d09d673 100644 (file)
@@ -985,6 +985,7 @@ struct vcpu_vmx {
        struct shared_msr_entry *guest_msrs;
        int                   nmsrs;
        int                   save_nmsrs;
+       bool                  guest_msrs_dirty;
        unsigned long         host_idt_base;
 #ifdef CONFIG_X86_64
        u64                   msr_host_kernel_gs_base;
@@ -2898,6 +2899,20 @@ static void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu)
 
        vmx->req_immediate_exit = false;
 
+       /*
+        * Note that guest MSRs to be saved/restored can also be changed
+        * when guest state is loaded. This happens when guest transitions
+        * to/from long-mode by setting MSR_EFER.LMA.
+        */
+       if (!vmx->loaded_cpu_state || vmx->guest_msrs_dirty) {
+               vmx->guest_msrs_dirty = false;
+               for (i = 0; i < vmx->save_nmsrs; ++i)
+                       kvm_set_shared_msr(vmx->guest_msrs[i].index,
+                                          vmx->guest_msrs[i].data,
+                                          vmx->guest_msrs[i].mask);
+
+       }
+
        if (vmx->loaded_cpu_state)
                return;
 
@@ -2958,11 +2973,6 @@ static void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu)
                vmcs_writel(HOST_GS_BASE, gs_base);
                host_state->gs_base = gs_base;
        }
-
-       for (i = 0; i < vmx->save_nmsrs; ++i)
-               kvm_set_shared_msr(vmx->guest_msrs[i].index,
-                                  vmx->guest_msrs[i].data,
-                                  vmx->guest_msrs[i].mask);
 }
 
 static void vmx_prepare_switch_to_host(struct vcpu_vmx *vmx)
@@ -3437,6 +3447,7 @@ static void setup_msrs(struct vcpu_vmx *vmx)
                move_msr_up(vmx, index, save_nmsrs++);
 
        vmx->save_nmsrs = save_nmsrs;
+       vmx->guest_msrs_dirty = true;
 
        if (cpu_has_vmx_msr_bitmap())
                vmx_update_msr_bitmap(&vmx->vcpu);