[PATCH] x86: correctly set nonlazy_xstate_used when loading full state

[PATCH] x86: correctly set nonlazy_xstate_used when loading full state

[Jan Beulich]

In this case, just like xcr0_accum, nonlazy_xstate_used should always be
set to the intended new value, rather than possibly leaving the flag set
from a prior state load.

Signed-off-by: Jan Beulich <jbeulich@suse.com>

--- a/xen/arch/x86/domctl.c
+++ b/xen/arch/x86/domctl.c
@@ -1187,8 +1187,7 @@ long arch_do_domctl(
                 vcpu_pause(v);
                 v->arch.xcr0 = _xcr0;
                 v->arch.xcr0_accum = _xcr0_accum;
-                if ( _xcr0_accum & XSTATE_NONLAZY )
-                    v->arch.nonlazy_xstate_used = 1;
+                v->arch.nonlazy_xstate_used = _xcr0_accum & XSTATE_NONLAZY;
                 compress_xsave_states(v, _xsave_area,
                                       evc->size - PV_XSAVE_HDR_SIZE);
                 vcpu_unpause(v);
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -1324,8 +1324,7 @@ static int hvm_load_cpu_xsave_states(str
 
     v->arch.xcr0 = ctxt->xcr0;
     v->arch.xcr0_accum = ctxt->xcr0_accum;
-    if ( ctxt->xcr0_accum & XSTATE_NONLAZY )
-        v->arch.nonlazy_xstate_used = 1;
+    v->arch.nonlazy_xstate_used = ctxt->xcr0_accum & XSTATE_NONLAZY;
     compress_xsave_states(v, &ctxt->save_area,
                           size - offsetof(struct hvm_hw_cpu_xsave, save_area));
 




_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [PATCH] x86: correctly set nonlazy_xstate_used when loading full state

[Wei Liu]

On Tue, Jun 26, 2018 at 01:47:12AM -0600, Jan Beulich wrote:
> In this case, just like xcr0_accum, nonlazy_xstate_used should always be
> set to the intended new value, rather than possibly leaving the flag set
> from a prior state load.
> 
> Signed-off-by: Jan Beulich <jbeulich@suse.com>

Why? Looking at the code it is in fact used as a boolean.

Also you seem to have missed hvm/svm/svm.c and xstate.c?

Wei.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [PATCH] x86: correctly set nonlazy_xstate_used when loading full state

[Jan Beulich]

>>> On 26.06.18 at 09:57, <wei.liu2@citrix.com> wrote:
> On Tue, Jun 26, 2018 at 01:47:12AM -0600, Jan Beulich wrote:
>> In this case, just like xcr0_accum, nonlazy_xstate_used should always be
>> set to the intended new value, rather than possibly leaving the flag set
>> from a prior state load.
>> 
>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> 
> Why? Looking at the code it is in fact used as a boolean.

Its boolean nature is irrelevant. The question is whether fully new state
is loaded (in which case the flag should be set according to this state,
without regard to the prior value of the flag) vs ...

> Also you seem to have missed hvm/svm/svm.c and xstate.c?

... existing state being updated.

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [PATCH] x86: correctly set nonlazy_xstate_used when loading full state

[Wei Liu]

On Tue, Jun 26, 2018 at 02:11:32AM -0600, Jan Beulich wrote:
> >>> On 26.06.18 at 09:57, <wei.liu2@citrix.com> wrote:
> > On Tue, Jun 26, 2018 at 01:47:12AM -0600, Jan Beulich wrote:
> >> In this case, just like xcr0_accum, nonlazy_xstate_used should always be
> >> set to the intended new value, rather than possibly leaving the flag set
> >> from a prior state load.
> >> 
> >> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> > 
> > Why? Looking at the code it is in fact used as a boolean.
> 
> Its boolean nature is irrelevant. The question is whether fully new state
> is loaded (in which case the flag should be set according to this state,
> without regard to the prior value of the flag) vs ...

I see. I misunderstood your intent.

Reviewed-by: Wei Liu <wei.liu2@citrix.com>

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Ping: [PATCH] x86: correctly set nonlazy_xstate_used when loading full state

[Jan Beulich]

Ping?

>>> On 26.06.18 at 09:47, <JBeulich@suse.com> wrote:
> In this case, just like xcr0_accum, nonlazy_xstate_used should always be
> set to the intended new value, rather than possibly leaving the flag set
> from a prior state load.
> 
> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> 
> --- a/xen/arch/x86/domctl.c
> +++ b/xen/arch/x86/domctl.c
> @@ -1187,8 +1187,7 @@ long arch_do_domctl(
>                  vcpu_pause(v);
>                  v->arch.xcr0 = _xcr0;
>                  v->arch.xcr0_accum = _xcr0_accum;
> -                if ( _xcr0_accum & XSTATE_NONLAZY )
> -                    v->arch.nonlazy_xstate_used = 1;
> +                v->arch.nonlazy_xstate_used = _xcr0_accum & XSTATE_NONLAZY;
>                  compress_xsave_states(v, _xsave_area,
>                                        evc->size - PV_XSAVE_HDR_SIZE);
>                  vcpu_unpause(v);
> --- a/xen/arch/x86/hvm/hvm.c
> +++ b/xen/arch/x86/hvm/hvm.c
> @@ -1324,8 +1324,7 @@ static int hvm_load_cpu_xsave_states(str
>  
>      v->arch.xcr0 = ctxt->xcr0;
>      v->arch.xcr0_accum = ctxt->xcr0_accum;
> -    if ( ctxt->xcr0_accum & XSTATE_NONLAZY )
> -        v->arch.nonlazy_xstate_used = 1;
> +    v->arch.nonlazy_xstate_used = ctxt->xcr0_accum & XSTATE_NONLAZY;
>      compress_xsave_states(v, &ctxt->save_area,
>                            size - offsetof(struct hvm_hw_cpu_xsave, save_area));
>  




_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [PATCH] x86: correctly set nonlazy_xstate_used when loading full state

[Andrew Cooper]

On 26/06/18 08:47, Jan Beulich wrote:
> In this case, just like xcr0_accum, nonlazy_xstate_used should always be
> set to the intended new value, rather than possibly leaving the flag set
> from a prior state load.
>
> Signed-off-by: Jan Beulich <jbeulich@suse.com>

Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel