Re: [Qemu-arm] [Qemu-devel] [PATCH 14/27] iommu: Add IOMMU index concept to IOMMU API

[Peter Maydell]

On 22 May 2018 at 13:58, Auger Eric <address@hidden> wrote:
> Hi Peter,
> On 05/21/2018 04:03 PM, Peter Maydell wrote:
>> If an IOMMU supports mappings that care about the memory
>> transaction attributes, then it no longer has a unique
>> address -> output mapping, but more than one. We can
>> represent these using an IOMMU index, analogous to TCG's
>> mmu indexes.
>>
>> Signed-off-by: Peter Maydell <address@hidden>
>> ---
>>  include/exec/memory.h | 52 +++++++++++++++++++++++++++++++++++++++++++
>>  memory.c              | 23 +++++++++++++++++++
>>  2 files changed, 75 insertions(+)
>>
>> diff --git a/include/exec/memory.h b/include/exec/memory.h
>> index 309fdfb89b..f6226fb263 100644
>> --- a/include/exec/memory.h
>> +++ b/include/exec/memory.h
>> @@ -206,6 +206,20 @@ enum IOMMUMemoryRegionAttr {
>>   * to report whenever mappings are changed, by calling
>>   * memory_region_notify_iommu() (or, if necessary, by calling
>>   * memory_region_notify_one() for each registered notifier).
>> + *
>> + * Conceptually an IOMMU provides a mapping from input address
>> + * to an output TLB entry.
> actually it takes a source identifier too (ARM streamid + substreamID,
> this latter is not yet supported).
> At the moment we have 1 IOMMUMRRegion per sid on ARM. For each
> IOMMUMRRegion we would now have 2 indexes (1 for secure and 1 for
> unsecure). How do you see the implementation of PASIDs (substreamids).
> Would that be based on indexes or not?

Good question. I guess we would set that up so that the
substream ID is in the memory transaction attributes as the
requester_id and then use that as part of the IOMMU index ?
Or you could use the indirection between tx attrs and indexes
to allow you to map down a potentially large space of substream
ID values to a smaller set of actually different configurations.

How many substream IDs do we expect to see in practice?

>> +/**
>> + * memory_region_iommu_attrs_to_index: return the IOMMU index to
>> + * use for translations with the given memory transaction attributes.
>> + *
>> + * @iommu_mr: the memory region
>> + * @attrs: the memory transaction attributes
>> + */
>> +int memory_region_iommu_attrs_to_index(IOMMUMemoryRegion *iommu_mr,
>> +                                       MemTxAttrs attrs);
>> +
>> +/**
>> + * memory_region_iommu_num_indexes: return the total number of IOMMU
>> + * indexes that this IOMMU supports.
> is it IOMMU wide characteristics or per IOMMUMRRegion (SID)?

I generally in this API document have been treating "IOMMU" and
"IOMMURegion" as synonymous, since from the perspective of the API
we don't care whether there's a bigger underlying thing in common
between IOMMURegions.

thanks
-- PMM