Re: [PATCH v3 13/13] drivers/passthrough: arm: Add support for SMMU drivers

[Julien Grall <julien.grall@linaro.org>]

Hi Ian,

On 19/03/14 10:35, Ian Campbell wrote:
> On Tue, 2014-03-18 at 20:25 +0000, Julien Grall wrote:
>
>>>> +static __init void arm_smmu_device_reset(struct arm_smmu_device *smmu)
>>>> +{
>>>> +[...]
>>>> +    /* Don't upgrade barriers */
>>>> +    reg &= ~(SMMU_sCR0_BSU_MASK << SMMU_sCR0_BSU_SHIFT);
>>>
>>> No? Is that safe when a vcpu migrates around pCPUs?
>>
>>  From the SMMU doc 9.6.3, this field is only used when client devices are
>> not mapped to a translation context banks.
>>
>> By default, the policy in Xen is to deny every transaction that doesn't
>> have a valid mapping. So we are safe.
>>
>> I can update the comment if you want, or even better removing this code.
>
> I think adding the explanation you just gave would do the job. Even if
> you were to remove the code an explanation of why it doesn't mess with
> the BSU mask would still be valuable I think (but I think setting it to
> a known value even if it is unused would be best).

Ok. I will update the comment.

Regards,

-- 
Julien Grall