Re: [PATCH v2 05/15] iommu: Add QDA compute context bank bus to iommu_buses

On 17-08-2026 19:14, Jason Gunthorpe wrote:
> On Mon, Aug 17, 2026 at 10:17:40AM +0530, Ekansh Gupta wrote:
>> Register the QDA compute context bank bus (qda-compute-cb) with the
>> IOMMU subsystem by adding it to the iommu_buses[] array.
>>
>> The QDA driver creates synthetic devices on this bus to represent
>> IOMMU context banks (CBs). Each CB device needs its own IOMMU domain
>> so that the DSP memory manager can enforce per-session address space
>> isolation. Without this registration, the IOMMU subsystem does not
>> probe CB devices for IOMMU groups and of_dma_configure() in the bus
>> dma_configure callback has no IOMMU domain to attach to.
> 
> I didn't notice a clear explanation of the proposed DT schema for this
The context banks are described in DT as child nodes of the fastrpc
RPMsg endpoint with compatible "qcom,fastrpc-compute-cb" and a `reg`
property for the stream ID, this is the existing binding in
Documentation/devicetree/bindings/misc/qcom,fastrpc.yaml. QDA reuses
this schema unchanged.
> 
> But it looks awfully similar to the other driver that was creating a
> kernel synthetic bus for actual real HW, that wasn't well liked.
> 
> What is this for? Why can't you have a normal DT binding for the
> iommu with normal devices?
The CB child nodes are not separate addressable hardware — they are SMMU
stream-ID assignments described as sub-resources of the parent DSP
interface. Using platform devices for them was explicitly rejected by
Greg KH [1] and subsequently using a generic shared bus in drivers/base/
was also rejected [2]. Greg's conclusion was that each driver needing
this pattern should have its own custom bus type.

That is exactly what QDA does: the qda-compute-cb bus is a per-driver
bus type with a dma_configure callback, following the same pattern as
the existing host1x_context_device_bus_type that has been in the
iommu_buses[] array since 2021. The bus creates devices synchronously in
probe, calls of_dma_configure() via its bus callback so the IOMMU
subsystem probes them normally, and tears them down on remove.

[1] https://lore.kernel.org/all/2025062434-reviving-grumble-1e53@gregkh/
[2] https://lore.kernel.org/all/2026042442-luxurious-antonym-f20c@gregkh/

//Ekansh>
> Jason