[PATCH 00/20] iommu: Refactoring domain allocation interface

[Yi Liu]

On 2024/5/31 14:00, Baolu Lu wrote:
> On 5/31/24 11:16 AM, Yi Liu wrote:
>> On 2024/5/29 20:02, Baolu Lu wrote:
>>> On 2024/5/29 17:03, Yi Liu wrote:
>>>> On 2024/5/29 13:32, Lu Baolu wrote:
>>>>> The IOMMU subsystem has undergone some changes, including the removal
>>>>> of iommu_ops from the bus structure. Consequently, the existing domain
>>>>> allocation interface, which relies on a bus type argument, is no longer
>>>>> relevant:
>>>>>
>>>>>      struct iommu_domain *iommu_domain_alloc(struct bus_type *bus)
>>>>>
>>>>> This series is designed to refactor the use of this interface. It
>>>>> proposes two new interfaces to replace iommu_domain_alloc():
>>>>>
>>>>> - iommu_user_domain_alloc(): This interface is intended for allocating
>>>>>    iommu domains managed by userspace for device passthrough scenarios,
>>>>>    such as those used by iommufd, vfio, and vdpa. It clearly indicates
>>>>>    that the domain is for user-managed device DMA.
>>>>
>>>> user paging domain? It looks to me user domain includes the nested domains
>>>> as well.
>>>
>>> Yes, nested domain is a user domain. The iommu driver should implement
>>> iommu_ops->domain_alloc_user for nested domain allocation.
>>
>> will it be more clear to name iommu_user_domain_alloc() be
>> iommu_user_paging_domain_alloc() as it is mainly for paging domain
>> allocation?
> 
> That might be better; let's wait and see if there's another option.
> 
>>
>>>>
>>>>>    If an IOMMU driver does not implement iommu_ops->domain_alloc_user,
>>>>>    this interface will rollback to the generic paging domain allocation.
>>>>>
>>>>> - iommu_paging_domain_alloc(): This interface is for allocating iommu
>>>>>    domains managed by kernel drivers for kernel DMA purposes. It takes a
>>>>>    device pointer as a parameter, which better reflects the current
>>>>>    design of the IOMMU subsystem.
>>>>>
>>>>> The majority of device drivers currently using iommu_domain_alloc() do
>>>>> so to allocate a domain for a specific device and then attach that
>>>>> domain to the device. These cases can be straightforwardly migrated to
>>>>> the new interfaces.
>>>>>
>>>>> However, there are some drivers with more complex use cases that do
>>>>> not fit neatly into this new scheme. For example:
>>>>>
>>>>> $ git grep "= iommu_domain_alloc"
>>>>> arch/arm/mm/dma-mapping.c:      mapping->domain = 
>>>>> iommu_domain_alloc(bus);
>>>>> drivers/gpu/drm/rockchip/rockchip_drm_drv.c:    private->domain = 
>>>>> iommu_domain_alloc(private->iommu_dev->bus);
>>>>> drivers/gpu/drm/tegra/drm.c:            tegra->domain = 
>>>>> iommu_domain_alloc(&platform_bus_type);
>>>>> drivers/infiniband/hw/usnic/usnic_uiom.c:       pd->domain = domain = 
>>>>> iommu_domain_alloc(dev->bus);
>>>>>
>>>>> This series leave those cases unchanged and keep iommu_domain_alloc()
>>>>> for their usage. But new drivers should not use it anymore.
>>>>
>>>> does it mean there is still domains allocated via iommu_domain_alloc()
>>>> on VT-d platform?
>>>
>>> I think the drivers mentioned above do not run on x86 platforms, or do
>>> they?
>>
>> cool. BTW. I know out-of-tree drivers are not counted in upstream review.
>> Just out of curious, is there a formal way to let such drivers know it is
>> no longer allowed to use iommu_domain_alloc() on VT-d?
> 
> As Robin suggested, we should try to remove iommu_domain_alloc() from
> the tree in this series.

If it's completely dropped, that's fine. OOT drivers should fail in the
time of compiling.

-- 
Regards,
Yi Liu