Re: [PATCH v2 15/16] iommu/amd: Initialize iommu_device->max_pasids

[Jason Gunthorpe <jgg@ziepe.ca>]

On Mon, Jul 31, 2023 at 12:33:18PM +0530, Vasant Hegde wrote:
> On 7/28/2023 8:16 PM, Jason Gunthorpe wrote:
> > On Fri, Jul 28, 2023 at 05:36:08AM +0000, Vasant Hegde wrote:
> >> Commit 1adf3cc20d69 ("iommu: Add max_pasids field in struct iommu_device")
> >> introduced a variable struct iommu_device.max_pasids to track max
> >> PASIDS supported by each IOMMU.
> >>
> >> Let us initialize this field for AMD IOMMU. IOMMU core will use this value
> >> to set max PASIDs per device (see __iommu_probe_device()).
> >>
> >> Also remove unused global 'amd_iommu_max_pasid' variable.
> >>
> >> Finally current code restricts max PASIDs to 16 bits and calls BUG_ON if
> >> max PASID is more than 16bit. This patch replaces BUG_ON with WARN_ON
> >> as system can continue to work with 16-bit PASID.
> > 
> >>  		pasmax = iommu->features & FEATURE_PASID_MASK;
> >>  		pasmax >>= FEATURE_PASID_SHIFT;
> >> -		max_pasid  = (1 << (pasmax + 1)) - 1;
> > 
> > This can be up to (1<<32)-1
> > 
> >> -
> >> -		amd_iommu_max_pasid = min(amd_iommu_max_pasid, max_pasid);
> >> +		iommu->iommu.max_pasids = (1 << (pasmax + 1)) - 1;
> > 
> > So why not
> > 
> >    	iommu->iommu.max_pasids = min((1 << (pasmax + 1)) - 1, PASID_MAX)
> 
> Because some older chips had support upto 16bit PASID only.

Sure, but I mean why not just directly cap it to PASID_MAX and not
bother with the WARN? If you WARN then a new chip reporting > 16 bits
of PASID support will cause pointless WARN_ONs at boot time even
though the driver will run fine?

Jason