On 11/7/2023 10:55 PM, isaku.yamahata@xxxxxxxxx wrote:
From: Isaku Yamahata <isaku.yamahata@xxxxxxxxx>
vmx_hardware_disable() accesses loaded_vmcss_on_cpu via
hardware_disable_all(). To allow hardware_enable/disable_all() before
kvm_init(), initialize it in vmx_hardware_setup() so that tdx module
initialization, hardware_setup method, can reference the variable.
In patch 004/116, vmx_init() and kvm_x86_vendor_init() are re-ordered,
tdx_hardware_setup() will be called after vmx_init(), the change of
this patch seems meaningless after patch 004/116.
Signed-off-by: Isaku Yamahata <isaku.yamahata@xxxxxxxxx>
---
arch/x86/kvm/vmx/vmx.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index 0e081c964e7a..0f3769cc3741 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -8263,8 +8263,12 @@ __init int vmx_hardware_setup(void)
{
unsigned long host_bndcfgs;
struct desc_ptr dt;
+ int cpu;
int r;
+ /* vmx_hardware_disable() accesses loaded_vmcss_on_cpu. */
+ for_each_possible_cpu(cpu)
+ INIT_LIST_HEAD(&per_cpu(loaded_vmcss_on_cpu, cpu));
store_idt(&dt);
host_idt_base = dt.address;
@@ -8507,11 +8511,8 @@ static int __init vmx_init(void)
if (r)
goto err_l1d_flush;
- for_each_possible_cpu(cpu) {
- INIT_LIST_HEAD(&per_cpu(loaded_vmcss_on_cpu, cpu));
-
+ for_each_possible_cpu(cpu)
pi_init_cpu(cpu);
- }
cpu_emergency_register_virt_callback(vmx_emergency_disable);
