Re: linux-4.4: qemu starting VM emits vmwrite error

[Paolo Bonzini <pbonzini@redhat.com>]

On 12/01/2016 09:25, Han, Huaitong wrote:
> I don't have a old machine to test it, please patch it and give me
> feedback. Thanks a lot.
> 
> From f06a85c76db210ce7145be6c704c44a9cef70ee3 Mon Sep 17 00:00:00 2001
> From: Huaitong Han <huaitong.han@intel.com>
> Date: Tue, 12 Jan 2016 16:04:20 +0800
> Subject: [PATCH] kvm: x86: Fix vmwrite to SECONDARY_VM_EXEC_CONTROL
> 
> SECONDARY_VM_EXEC_CONTROL vmcs will be updated with vmx_cpuid_update,
> but it
> will vmwrite error on the older CPU because the code is written without
> CPU_BASED_ACTIVATE_SECONDARY_CONTROLS check.
> 
> This will get rid of the following trace:
> 
> vmwrite error: reg 401e value 10 (err 12)
> Call Trace:
> [<ffffffff8116e2b9>] ? dump_stack+0x40/0x57
> [<ffffffffa020b88d>] ?vmx_cpuid_update+0x5d/0x150 [kvm_intel]
> [<ffffffffa01d8fdc>] ?kvm_vcpu_ioctl_set_cpuid2+0x4c/0x70 [kvm]
> [<ffffffffa01b8363>] ?kvm_arch_vcpu_ioctl+0x903/0xfa0 [kvm]
> 
> Signen-off-by: Huaitong Han <huaitong.han@intel.com>
> ---
>  arch/x86/kvm/vmx.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 44976a5..80f21d3 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -8904,7 +8904,7 @@ static void vmx_cpuid_update(struct kvm_vcpu
> *vcpu)
>  {
>  	struct kvm_cpuid_entry2 *best;
>  	struct vcpu_vmx *vmx = to_vmx(vcpu);
> -	u32 secondary_exec_ctl = vmx_secondary_exec_control(vmx);
> +	u32 secondary_exec_ctl = -1;
>  
>  	if (vmx_rdtscp_supported()) {
>  		bool rdtscp_enabled = guest_cpuid_has_rdtscp(vcpu);
> @@ -8932,7 +8932,10 @@ static void vmx_cpuid_update(struct kvm_vcpu
> *vcpu)
>  			best->ebx &= ~bit(X86_FEATURE_INVPCID);
>  	}
>  
> -	vmcs_set_secondary_exec_control(secondary_exec_ctl);
> +	if (cpu_has_secondary_exec_ctrls()) {
> +		secondary_exec_ctl &= vmx_secondary_exec_control(vmx);
> +		vmcs_set_secondary_exec_control(secondary_exec_ctl);
> +	}
>  
>  	if (static_cpu_has(X86_FEATURE_PCOMMIT) && nested) {
>  		if (guest_cpuid_has_pcommit(vcpu))
> 


Looks good, thanks!

Paolo