From: Joerg Roedel <redacted>
Subject: Re: [PATCH 2/2] ARM: IOMMU: Tegra30: Add iommu_ops for SMMU driver
Date: Tue, 24 Jan 2012 12:04:44 +0100
Message-ID: [ref]

quoted

quoted

Hmm, this looks like there is a 1-1 mapping between hardware SMMU
devices and domains. This is not consistent with IOMMU-API semantics
where a domain can contain devices behind different SMMUs. Please fix
that.

I'm a bit confused with the concept of "domain". I thought that
"domain" is equivalent to a "virtual address space". Usually a IOMMU
device provides a virtual address space for multiple client
devices. IOW, a IOMMU device provides a virtual address space, which
can be shared with multiple client devices.

Actually Tegra SMMU case, a single IOMMU device has 4 different
virtual address speace("smmu_as"). Each "smmu_as" has its own virtual
address space. "smmu_as[i]" has mutiple "smmu_client" devices.

  smmu_as[i] == domain[i]

I don't understand why "a domain can contain devices behind different
SMMUs" because those client devices belong to different virtual
address spaces, and they should belong to different "domains".

Could you please explain a bit more about "domain"?

A domain is, as you said, a virtual address space for IO devices. But
the important point is, an arbitrary number of devices can be part of a
domain. This also means that the devices can be behind different
hardware SMMUs. In this case your driver needs to program the page-table
pointer into more than one SMMU to give devices behind different SMMUs
the same address space.

Thank you for explaining.

Does the above mean that a buffer can be shared with different devices
which belong to different IOMMU devices(virtual address spaces)?

For example, assuming the following:

- We have "struct iommu_domain *domain1".
- "domain1" has iommu device "iommu_dev1" and "iommu_dev2".
- "iommu_dev1" has "client_dev1" and "client_dev2".
- "iommu_dev2" has "client_dev3" and "client_dev4".

"iommu_map(domain1, iova, pa, ...)" will create the following mapping
___at once___:

- (iova)-(pa) mapping in iommu_dev1(iommmu_dev1's virtual address space)
- (iova)-(pa) mapping in iommu_dev2(iommmu_dev2's virtual address space)

Is the above correct?

It seems that the same (iova) is used for different virtual address
spaces. What kind of case is this beneficial most in?