Re: [PATCH] KVM: nvmx: retry writing guest memory after page fault injected

Re: [PATCH] KVM: nvmx: retry writing guest memory after page fault injected

[linmiaohe]

Hi,

Liran Alon <liran.alon@oracle.com> wrote:
>> On 25 Dec 2019, at 4:21, linmiaohe <linmiaohe@huawei.com> wrote:
>> 
>> From: Miaohe Lin <linmiaohe@huawei.com>
>> 
>> We should retry writing guest memory when 
>> kvm_write_guest_virt_system() failed and page fault is injected in handle_vmread().
>> 
>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>
>Patch fix seems correct to me:
>Reviewed-by: Liran Alon <liran.alon@oracle.com>

Thanks for your review.

>However, I suggest to rephrase commit title & message as follows:
>
>"""
>KVM: nVMX: vmread should not set rflags to specify success in case of #PF
>
>In case writing to vmread destination operand result in a #PF, vmread should not call nested_vmx_succeed() to set rflags to specify success. Similar to as done in for VMPTRST (See handle_vmptrst()).
>"""

Thanks for your sueestion, I would rephrase commit title & message accordingly.

>
>In addition, it will be appreciated if you would also submit kvm-unit-test that verifies this condition.

I'd like to submit kvm-unit-test that verifies this condition, but I am not familiar with the kvm-unit-test code yet and
also not in my recent todo list. So such a patch may come late. It would be appreciated too if you could submit this
kvm-unit-test patch. :) 
Thanks again.

Re: [PATCH] KVM: nvmx: retry writing guest memory after page fault injected

[Wanpeng Li]

On Thu, 26 Dec 2019 at 10:32, linmiaohe <linmiaohe@huawei.com> wrote:
>
> Hi,
>
> Liran Alon <liran.alon@oracle.com> wrote:
> >> On 25 Dec 2019, at 4:21, linmiaohe <linmiaohe@huawei.com> wrote:
> >>
> >> From: Miaohe Lin <linmiaohe@huawei.com>
> >>
> >> We should retry writing guest memory when
> >> kvm_write_guest_virt_system() failed and page fault is injected in handle_vmread().
> >>
> >> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
> >
> >Patch fix seems correct to me:
> >Reviewed-by: Liran Alon <liran.alon@oracle.com>
>
> Thanks for your review.
>
> >However, I suggest to rephrase commit title & message as follows:
> >
> >"""
> >KVM: nVMX: vmread should not set rflags to specify success in case of #PF
> >
> >In case writing to vmread destination operand result in a #PF, vmread should not call nested_vmx_succeed() to set rflags to specify success. Similar to as done in for VMPTRST (See handle_vmptrst()).
> >"""
>
> Thanks for your sueestion, I would rephrase commit title & message accordingly.
>
> >
> >In addition, it will be appreciated if you would also submit kvm-unit-test that verifies this condition.
>
> I'd like to submit kvm-unit-test that verifies this condition, but I am not familiar with the kvm-unit-test code yet and
> also not in my recent todo list. So such a patch may come late. It would be appreciated too if you could submit this
> kvm-unit-test patch. :)

Hmm, did you verify your own patch? Please give the testcase.

Re: [PATCH] KVM: nvmx: retry writing guest memory after page fault injected

[linmiaohe]

Hi,

Wanpeng Li <kernellwp@gmail.com> wrote:
> On Thu, 26 Dec 2019 at 10:32, linmiaohe <linmiaohe@huawei.com> wrote:
> >
> > >
> > >In addition, it will be appreciated if you would also submit kvm-unit-test that verifies this condition.
> >
> > I'd like to submit kvm-unit-test that verifies this condition, but I 
> > am not familiar with the kvm-unit-test code yet and also not in my 
> > recent todo list. So such a patch may come late. It would be 
> > appreciated too if you could submit this kvm-unit-test patch. :)
>
> Hmm, did you verify your own patch? Please give the testcase.
>

I'am sorry but this is a code inspection path. And I think this fix is correct from the code logic and
this behavior is similar as handle_vmwrite, handle_vmptrst, handle_invept and so on. Thanks!

Re: [PATCH] KVM: nvmx: retry writing guest memory after page fault injected

[Liran Alon]

> On 25 Dec 2019, at 4:21, linmiaohe <linmiaohe@huawei.com> wrote:
> 
> From: Miaohe Lin <linmiaohe@huawei.com>
> 
> We should retry writing guest memory when kvm_write_guest_virt_system()
> failed and page fault is injected in handle_vmread().
> 
> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>

Patch fix seems correct to me:
Reviewed-by: Liran Alon <liran.alon@oracle.com>

However, I suggest to rephrase commit title & message as follows:

"""
KVM: nVMX: vmread should not set rflags to specify success in case of #PF

In case writing to vmread destination operand result in a #PF, vmread should
not call nested_vmx_succeed() to set rflags to specify success. Similar to as
done in for VMPTRST (See handle_vmptrst()).
"""

In addition, it will be appreciated if you would also submit kvm-unit-test that verifies this condition.

-Liran

> ---
> arch/x86/kvm/vmx/nested.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index 8edefdc9c0cb..c1ec9f25a417 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -4799,8 +4799,10 @@ static int handle_vmread(struct kvm_vcpu *vcpu)
> 					instr_info, true, len, &gva))
> 			return 1;
> 		/* _system ok, nested_vmx_check_permission has verified cpl=0 */
> -		if (kvm_write_guest_virt_system(vcpu, gva, &value, len, &e))
> +		if (kvm_write_guest_virt_system(vcpu, gva, &value, len, &e)) {
> 			kvm_inject_page_fault(vcpu, &e);
> +			return 1;
> +		}
> 	}
> 
> 	return nested_vmx_succeed(vcpu);
> -- 
> 2.19.1
> 

[PATCH] KVM: nvmx: retry writing guest memory after page fault injected

[linmiaohe]

From: Miaohe Lin <linmiaohe@huawei.com>

We should retry writing guest memory when kvm_write_guest_virt_system()
failed and page fault is injected in handle_vmread().

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
---
 arch/x86/kvm/vmx/nested.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index 8edefdc9c0cb..c1ec9f25a417 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -4799,8 +4799,10 @@ static int handle_vmread(struct kvm_vcpu *vcpu)
 					instr_info, true, len, &gva))
 			return 1;
 		/* _system ok, nested_vmx_check_permission has verified cpl=0 */
-		if (kvm_write_guest_virt_system(vcpu, gva, &value, len, &e))
+		if (kvm_write_guest_virt_system(vcpu, gva, &value, len, &e)) {
 			kvm_inject_page_fault(vcpu, &e);
+			return 1;
+		}
 	}
 
 	return nested_vmx_succeed(vcpu);
-- 
2.19.1