Re: [PATCH v2 10/11] iommu/amd: Add IO page fault notifier handler

[Baolu Lu <baolu.lu@linux.intel.com>]

On 9/13/23 2:46 AM, Jason Gunthorpe wrote:
>> +static void iommu_call_iopf_notifier(struct amd_iommu *iommu, u64 *raw)
>> +{
>> +	struct iopf_fault event;
>> +	struct pci_dev *pdev;
>> +	int ret = -EINVAL;
>> +	u16 devid = PPR_DEVID(raw[0]);
>> +
>> +	if (PPR_REQ_TYPE(raw[0]) != PPR_REQ_FAULT) {
>> +		pr_err_ratelimited("Unknown PPR request received\n");
>> +		return;
>> +	}
>> +
>> +	if (!ppr_is_valid(iommu, raw))
>> +		goto out;
>> +
>> +	pdev = pci_get_domain_bus_and_slot(iommu->pci_seg->id, PCI_BUS_NUM(devid),
>> +					   devid & 0xff);
>> +	if (!pdev)
>> +		goto out;
> Lu, here is another case where the core PRI code could make use of a
> core helper for a getting from the RID to the iommu world.
> 

Yeah! It's common for PCI/PRI. ARM SMMUv3 probably will also implement
PCI/PRI in the future.

AMD IOMMU driver also calls pci_get_domain_bus_and_slot() in the path of 
handling a DMA fault (unrecoverable fault). I don't think this is a
performance critical path, so we only need the common code to to serve
the PRI case. What's your opinion?

Best regards,
baolu