Re: [Xen-devel] [PATCH v9 3/6] sysctl / libxl: report whether IOMMU/HAP page table sharing is supported

[Jan Beulich <jbeulich@xxxxxxxx>]

On 12.09.2019 15:18, Paul Durrant wrote:
>> -----Original Message-----
>> From: Jan Beulich <jbeulich@xxxxxxxx>
>> Sent: 12 September 2019 14:04
>> To: Paul Durrant <Paul.Durrant@xxxxxxxxxx>
>> Cc: xen-devel@xxxxxxxxxxxxxxxxxxxx; Julien Grall <julien.grall@xxxxxxx>; 
>> Andrew Cooper
>> <Andrew.Cooper3@xxxxxxxxxx>; Anthony Perard <anthony.perard@xxxxxxxxxx>; 
>> Christian Lindig
>> <christian.lindig@xxxxxxxxxx>; George Dunlap <George.Dunlap@xxxxxxxxxx>; Ian 
>> Jackson
>> <Ian.Jackson@xxxxxxxxxx>; Stefano Stabellini <sstabellini@xxxxxxxxxx>; 
>> Konrad Rzeszutek Wilk
>> <konrad.wilk@xxxxxxxxxx>; DavidScott <dave@xxxxxxxxxx>; Tim (Xen.org) 
>> <tim@xxxxxxx>; Wei Liu
>> <wl@xxxxxxx>
>> Subject: Re: [PATCH v9 3/6] sysctl / libxl: report whether IOMMU/HAP page 
>> table sharing is supported
>>
>> On 12.09.2019 13:17, Paul Durrant wrote:
>>> --- a/xen/arch/arm/sysctl.c
>>> +++ b/xen/arch/arm/sysctl.c
>>> @@ -15,6 +15,9 @@
>>>  void arch_do_physinfo(struct xen_sysctl_physinfo *pi)
>>>  {
>>>      pi->capabilities |= XEN_SYSCTL_PHYSCAP_hvm | XEN_SYSCTL_PHYSCAP_hap;
>>> +
>>> +    if ( iommu_enabled && iommu_hap_pt_share )
>>> +        pi->capabilities |= XEN_SYSCTL_PHYSCAP_iommu_hap_pt_share;
>>>  }
>>
>> I think this should be done in common code.
>>
>>> --- a/xen/arch/x86/sysctl.c
>>> +++ b/xen/arch/x86/sysctl.c
>>> @@ -164,7 +164,12 @@ void arch_do_physinfo(struct xen_sysctl_physinfo *pi)
>>>      if ( IS_ENABLED(CONFIG_PV) )
>>>          pi->capabilities |= XEN_SYSCTL_PHYSCAP_pv;
>>>      if ( hvm_hap_supported() )
>>> +    {
>>>          pi->capabilities |= XEN_SYSCTL_PHYSCAP_hap;
>>> +
>>> +        if ( iommu_enabled && iommu_hap_pt_share )
>>> +            pi->capabilities |= XEN_SYSCTL_PHYSCAP_iommu_hap_pt_share;
>>> +    }
>>>  }
>>
>> And if it's important to not have the bit set when !hvm_hap_supported(),
> 
> ...and that's what it’s not in common code... there is no hvm_hap_supported() 
> for Arm
> 
>> then iommu_hap_pt_share should be cleared in __init code in this case.
> 
> That would have been an alternative approach if you'd not wanted
> it #defined in patch #5. (Yes it's a later patch in the series,
> but this one is later chronologically and I didn't want to
> invalidate the other patch.)
> I could perhaps implement hvm_hap_supported() for Arm?

Well, implementing it for Arm is an option (at which point
XEN_SYSCTL_PHYSCAP_hap could also move to common code). But
personally I'd advise against providing more stubs than
necessary on Arm. I may not understand your remark regarding
patch 5: There's no problem with iommu_hap_pt_share now being
a #define on Arm - whether the system is HVM capable is an
x86 specific decision anyway, and hence the clearing of the
flag could occur e.g. in hvm_enable() when it bails early.
And btw., with !CONFIG_HVM it could be a #define on x86 as
well, just to "false" then.

Jan

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