[PATCH v4 22/23] iommu/tegra241-cmdqv: Add user-space use support

Tian, Kevin kevin.tian at intel.com
Thu May 15 21:00:58 PDT 2025


> From: Nicolin Chen <nicolinc at nvidia.com>
> Sent: Friday, May 16, 2025 1:14 AM
> 
> On Thu, May 15, 2025 at 08:27:17AM +0000, Tian, Kevin wrote:
> > > From: Nicolin Chen <nicolinc at nvidia.com>
> > > Sent: Friday, May 9, 2025 11:03 AM
> > >
> > >  /**
> > >   * struct iommu_hw_info_arm_smmuv3 - ARM SMMUv3 hardware
> > > information
> > >   *                                   (IOMMU_HW_INFO_TYPE_ARM_SMMUV3)
> > >   *
> > > - * @flags: Must be set to 0
> > > - * @impl: Must be 0
> > > + * @flags: Combination of enum iommu_hw_info_arm_smmuv3_flags
> > > + * @impl: Implementation-defined bits when the following flags are set:
> > > + *        - IOMMU_HW_INFO_ARM_SMMUV3_HAS_TEGRA241_CMDQV
> > > + *          Bits[15:12] - Log2 of the total number of SID replacements
> > > + *          Bits[11:08] - Log2 of the total number of VINTFs per vIOMMU
> > > + *          Bits[07:04] - Log2 of the total number of VCMDQs per vIOMMU
> > > + *          Bits[03:00] - Version number for the CMDQ-V HW
> >
> > hmm throughout this series I drew an equation between VINTF
> > and vIOMMU. Not sure how multiple VINTFs can be represented
> > w/o introducing more objects. Do we want to keep such info here?
> 
> You are right that VINTF=vIOMMU. This is a per SMMU instance ioctl.
> So, each VM should only have one VTINF/vIOMMU per SMMU instance.
> 
> For multi-VINTF (multi-vIOMMU) case, there needs to be more SMMUs
> backing passthrough devices being assigned to the VM.
> 
> What exactly the concern of keeping this info here?

First, you agreed that VINTF=vIOMMU, then "total number of VINTFs
per vIOMMU" doesn't make sense as it's fixed to 1 in concept.

Then, each VM can only get one VINTF/vIOMMU per SMMU instance,
and this ioctl is per SMMU instance. This also implies that only one
VINTF can be reported in the ioctl.

In multi-VINTF case, the VM should get 1VINTF per ioctl from each
SMMU backing passthrough devices.

Then what is the point of " Bits[11:08] - Log2 of the total number
of VINTFs per vIOMMU "?

> 
> > > +	 * - suggest to back the queue memory with contiguous physical
> > > pages or
> > > +	 *   a single huge page with alignment of the queue size, limit
> > > vSMMU's
> > > +	 *   IDR1.CMDQS to the huge page size divided by 16 bytes
> > > +	 */
> > > +	IOMMU_HW_QUEUE_TYPE_TEGRA241_CMDQV = 1,
> >
> > Not sure about the last sentence. 'limit' refers to a certain action
> > which the user should perform?
> 
> Yes, set vSMMU's IDR1.CMDQS field up to the huge page size divided by
> 16 bytes, e.g. if using one 2MB huge page backing the queue memory,
> VMM should set IDR1.CMDQS no higher than 17:
> 	2MB = (1 << 17) * 16B
> 
> Certainly, it can set to lower than 17. So it's an upper "limit".
> 
> Or any better word in your mind that can be less confusing?

No. I misread 'vSMMU' as 'SMMU'.



More information about the linux-arm-kernel mailing list