RE: [PATCH v6 18/22] iommufd: Introduce a helper function to extract vendor capabilities

[Duan, Zhenzhong]

>-----Original Message-----
>From: Nicolin Chen <nicol...@nvidia.com>
>Subject: Re: [PATCH v6 18/22] iommufd: Introduce a helper function to
>extract vendor capabilities
>
>On Wed, Sep 24, 2025 at 08:05:36AM +0000, Duan, Zhenzhong wrote:
>> >uint64_t host_iommu_extract_quirks(enum iommu_hw_info_type,
>> >VendorCaps *caps)
>> >{
>> >    uint64_t quirks = 0;
>> >
>> >#if defined(CONFIG_VTD)
>>
>> I have applied all suggested change except CONFIG_VTD here as it's a
>device config and iommufd.c is device agnostic, it doesn't recognize
>CONFIG_VTD.
>>
>> ../backends/iommufd.c:419:13: error: attempt to use poisoned
>"CONFIG_VTD"
>>
>> I thought this is trivial and OK for not having CONFIG_VTD?
>
>Hmm.. I didn't expect that. It seems that QEMU does encourage
>moving all vendor specific code to vendor specific file :-/

This make me think CONFIG_VTD should not be used here, it controls
intel_iommu emulation, we may have a custom virtual machine without
virtual intel_iommu but still need to check host IOMMU.

If we do want a control, maybe kind of CONFIG_HOST_VTD rather than CONFIG_VTD.

>
>Anyway, I think it's fine to drop the ifdef. The VTD type and
>cap structure are defined in the shared uAPI header.

Agree.

Thanks
Zhenzhong